PROGRESS Pathways of mammalian replication fork restart Eva Petermann and Thomas Helleday
Abstract | Single-molecule analyses of DNA replication have greatly advanced our understanding of mammalian replication restart. Several proteins that are not part of the core replication machinery promote the efficient restart of replication forks that have been stalled by replication inhibitors, suggesting that bona fide fork restart pathways exist in mammalian cells. Different models of replication fork restart can be envisaged, based on the involvement of DNA helicases, nucleases, homologous recombination factors and the importance of DNA double-strand break formation. The faithful and complete replication of chromosomes is essential to maintain genetic stability and prevent the accumulation of cancer-promoting mutations. Replication forks are vulnerable to stalling or collapse as they encounter obstacles on the DNA template, which can be unrepaired DNA damage, DNA-bound proteins or secondary structures. Similarly, chemical agents like hydroxyurea and aphidicolin inhibit replication elongation, leading to fork stalling or collapse. A stalled replication fork is arrested but is capable of resuming replication (replication fork restart) once the inhibition is removed, whereas a collapsed fork has become inactivated through dissociation of the replication machinery or the generation of DNA double-strand breaks (DSBs). In eukaryotic cells, fork-associated DSBs are generated by the structurespecific endonuclease complex of methyl methanesulphonate (MMS) and ultravioletsensitive 81 (MUS81) and essential meiotic endonuclease 1 (EME1)1. After the removal of replication inhibitors such as aphidicolin and hydroxyurea, replication resumes by mechanisms that have been poorly understood in mammalian cells. More is known about replication restart in Escherichia coli, in which stalled or collapsed forks are reactivated by recombination-dependent or recombinationindependent pathways, depending on their structure2 (Supplementary information S1 (box)). The proteins involved in these pathways are not conserved in eukaryotes. E. coli
replication is dependent on fork reactivation as bacterial chromosomes and plasmids contain only one replication origin and forks cannot move beyond a specific termination site. By contrast, eukaryotic chromosomes contain more replication origins than are activated during each round of replication and these can serve as backups for completion of replication if progression of nearby forks is impaired3,4. It has therefore been unclear whether eukaryotes use specialized mechanisms for replication restart after the removal of replication inhibitors. Although it has been established that replication inhibitors activate homologous recombination5,6, analysing replication restart has been difficult owing to the lack of methods for studying individual replication forks. However, in recent years, single-molecule analyses of replication using chromosome combing or the DNA fibre technique7 have led to a better understanding of mammalian replication restart8–13. Various proteins that are not part of the core replication machinery promote efficient replication fork restart after the removal of replication inhibitors, suggesting that bona fide fork restart pathways exist in mammalian cells. This Progress article focuses on possible restart mechanisms for mammalian replication forks that have been stalled by hydroxyurea and aphidicolin, replication inhibitors that do not seem to modify the DNA template. Broadly, three fork stabilization and restart pathways can be postulated (FIG. 1). In the first mechanism, a stalled fork structure
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
might undergo remodelling before restart, such as the re-annealing of excess singlestrand DNA (ssDNA) or fork regression and pairing of the newly synthesized strands to form a Holliday junction in a structure termed a ‘chicken foot’ (FIG. 1a). This fork remodelling would involve fork collapse by the dissociation of the replication machinery but could theoretically be reversed. In a second mechanism, the Holliday junction is used to facilitate fork restart by recombining into the template by strand invasion to form a displacement loop (D-loop; FIG. 1b). In a D‑loop, one strand of a double-stranded DNA molecule is displaced by a third strand of homologous sequence; in this case, a newly synthesized daughter strand. In E. coli, this allows origin-independent loading of the replication machinery 14. In the third mechanism, the Holliday junction is processed into a one-ended DSB. Fork restart is achieved through homologous recombination-mediated repair of the DSB in a mechanism analogous to break-induced replication (BIR)15 (FIG. 1c). Here, we discuss recent evidence, obtained using single-molecule analyses in mammalian cells, for and against each of these fork restart mechanisms. As the ability to analyse replication forks in mammalian cells is a recent advance in the field, we do not cover evidence from non-mammalian model systems. Furthermore, owing to different underlying concepts, we do not cover fork progression on DNA templates modified by DNA-damaging agents. It is important to keep in mind that single-molecule approaches directly measure fork progression and cannot distinguish defects in fork stabilization, which could underlie a restart defect (see Supplementary information S2 (box)). The pathways described below may therefore be active during the replication block to stabilize forks and after the removal of the inhibitor to restart forks. Restart by fork remodelling DNA helicases may promote restart by remodelling and stabilizing stalled fork structures (FIG. 1a). Fork regression into Holliday junctions has not been shown in mammalian cells, but there is no direct evidence against this model. Such a process VOLUME 11 | O CTOBER 2010 | 683
© 2010 Macmillan Publishers Limited. All rights reserved
PROGRESS a
5′ 3′ 5′ 3′
Polymerase–helicase uncoupling 5′ 3′
Restart
ssDNA annealing
Reversal or degradation
5′ 3′
Chicken foot Fork regression
b
D-loop
Double Holliday junction
D-loop formation Restart
c Holliday junction cleavage
One-ended DSB
D-loop
D-loop formation
Figure 1 | Models of replication fork restart. a | Restart by fork remodelling. A stalled replication fork might be stabilized by the re-annealing of single-strand DNA (ssDNA) generated by excessive unwinding of the template, or might undergo regression and pairing of the newly synthesized strands to form a Holliday junction in a structure termed a ‘chicken foot’. Restart after Holliday junction formation may be difficult if it requires the removal and subsequent re-loading of the replication machinery. b | Holliday junction‑mediated fork restart. The double-stranded DNA (dsDNA) end of the Holliday junction is recombined into the template through strand invasion, forming a displacement loop (D‑loop). In E. coli, the D‑loop allows re-loading of the replication machinery. The invading strand re-anneals with
may be beneficial (by stabilizing stalled forks) or detrimental to restart, and both possibilities have to be considered. Three non-replicative helicases are known to promote fork restart after short to mediumlength periods (14 hours) of replication inhibition. BLM facilitates fork restart. The DNA helicase Bloom syndrome helicase (BLM) promotes efficient replication fork restart after 2–6 hours of aphidicolin or hydroxy urea treatment through its helicase activity 8. Reduced fork restart in BLM-deficient cells is accompanied by increased origin firing, suggesting a novel role of BLM in controlling replication initiation8. BLM might stabilize forks either by facilitating fork regression into a Holliday junction or by reversing fork regression through its Holliday junction migration activity 16,17. Additionally, BLM could promote fork restart by repressing the aberrant formation of recombination intermediates at stalled forks18. BLM-deficient cells treated with
Holliday junction dissolution
Restart
Holliday junction resolution
the template, forming a double Holliday junction that can be removed by NaturetoReviews | Molecular exchanges, Cell Biology Holliday junction resolution, which leads sister chromatid or Holliday junction dissolution, which avoids the generation of recombination products42. c | Double-strand break (DSB)-mediated restart. The Holliday junction is processed into a one-ended DSB and fork restart is achieved through homologous recombination repair of the DSB in a mechanism analogous to break-induced replication. The resulting single Holliday junction would be resolved by Holliday junction resolution. The leading strand and leading-strand template are shown in light and dark blue, respectively, and the lagging strand and lagging-strand template are shown in light and dark red, respectively.
replication inhibitors display nuclear foci of the central homologous recombination factor RAD51, which is a sign of increased homologous recombination initiation19. This could be interpreted as elevated, unscheduled recombination, but also as an increased need for the repair of forks collapsed into DSBs or a failure to complete homologous recombination. WRN enables fork progression during restart. Like BLM, Werner syndrome ATPdependent helicase (WRN) can catalyse both fork regression and Holliday junction migration in vitro20,21. WRN depletion in fibroblasts does not affect the percentage of forks restarting after release from short (4–7 hour) hydroxyurea treatment; instead, the speed of the restarting forks is reduced10. This suggests that WRN is not needed to stabilize forks but is required for efficient fork progression during restart. In WRNdepleted HeLa cells, however, fewer forks restart after 6–14 hours of hydroxyurea treatment, indicating fork inactivation9.
684 | O CTOBER 2010 | VOLUME 11
This discrepancy may not reflect a specific feature of WRN, but could be due to the different time points used or different kinetics of fork inactivation in these cells. In fibroblasts, WRN also promotes fork progression after MMS treatment 10, suggesting that WRN acts on structures such as single-strand gaps left behind the fork. This prompted the speculation that WRN remodels forks to coordinate fork progression with repair of gaps or lesions10,22. ssDNA annealing promotes fork restart. SWI/SNF-related, matrix-associated, actin-dependent regulator of chromatin, subfamily A‑like 1 (SMARCAL1) is a ssDNA annealing helicase that is mutated in Schimke immunoosseous dysplasia23 and was recently implicated in the response to replication inhibitors. SMARCAL1 is recruited to stalled replication forks or sites of DNA damage containing ssDNA12,24. Cells depleted of SMARCAL1 accumulate ssDNA and DNA damage and display increased sensitivity to aphidicolin and hydroxyurea12,24. www.nature.com/reviews/molcellbio
© 2010 Macmillan Publishers Limited. All rights reserved
PROGRESS DNA fibre analyses show that replication fork restart after 2 hours of aphidicolin treatment is reduced in SMARCAL1‑depleted cells12. These data suggest that SMARCAL1 promotes fork stability and restart by reannealing long stretches of ssDNA generated at stalled replication forks.
Glossary Aphidicolin
Holliday junction
A small-molecule inhibitor that directly blocks the activity of the replicative DNA polymerases. Aphidicolin treatment leads to replication fork stalling and eventually DNA DSB formation.
A four-way junction between two dsDNA molecules of homologous sequence. Holliday junctions are mobile and can be translocated by DNA helicases (branch migration).
Hydroxyurea Break-induced replication
Considerations against restart by remodelling. The main conceptual weakness of fork restart by remodelling is that the replication machinery presumably needs to dissociate from the replication fork to allow remodelling and it is not clear how it would be re-loaded. All fork remodelling might therefore channel forks into the Holliday junction‑mediated or DSB-mediated restart pathways.
A mechanism for origin-independent replication restart, whereby a resected DNA end invades a homologous DNA molecule, thus establishing a replication fork.
Holliday junction-mediated restart The idea that fork restart can be achieved by a Holliday junction intermediate without DSB formation is mainly supported by the observation that fork restart requires the recombination factor RAD51 (which catalyses Holliday junction formation) but not necessarily MUS81 (which cleaves Holliday junctions into one-ended DSBs)9,13. RAD51‑mediated restart is seen after short (2 hour) but not long (24 hour) hydroxyurea treatments13. RAD51 is loaded onto ssDNA to form protein–DNA filaments, which then catalyse homology search and strand exchange25, activities that would be required for strand invasion and D‑loop formation during Holliday junction‑mediated restart. As regression into a Holliday junction prob ably requires dissociation of the replication machinery, this restart pathway may act on forks that have collapsed by losing most of their replication factors.
Core replication machinery
Generation of the Holliday junction. Both BLM and WRN could facilitate fork regression into a Holliday junction through their helicase activities17,21. Another way of generating a Holliday junction could be through RAD51. The bacterial RAD51 homologue, RecA, can convert specific types of stalled forks into a Holliday junction by pairing the template strands26,27, and RAD51 can facilitate a similar mechanism28. Requirement of DNA end processing. RAD51 loading onto DNA requires a lagging-strand gap or a 3′-overhang and, therefore, probably involves end processing. These structures may be generated by the exonuclease activity of meiotic recombination 11 (MRE11) as MRE11‑dependent DNA resection has been shown to promote
A radical scavenger that inhibits ribonucleotide reductase, which results in cells producing less of the desoxyribo nucleotides that are used for DNA synthesis. Hydroxyurea treatment leads to replication fork stalling and eventually DNA DSB formation.
Chromosome combing A method for single-molecule analysis of DNA replication forks. Newly replicated DNA is labelled in vivo using halogenated thymidine analogues and genomic DNA is isolated and spread out on microscope coverslips. Immunofluorescence staining of the thymidine analogues is used to visualize the labelled tracks that are left on the DNA by moving replication forks.
The complex of proteins that is essential for all DNA replication and includes the replicative DNA helicase, primase, clamp loader, sliding clamp and leading- and lagging-strand DNA polymerases.
Lagging strand The nascent strand of DNA that is synthesized discontinuously in short pieces (Okazaki fragments) at the replication fork.
Origin firing The start of replication fork progression at a replication origin. Mammalian origin firing is restricted to S phase and is controlled by cell cycle signalling.
Replication fork A structure formed when the template strands have been separated by helicases and a newly formed copy of the DNA is synthesized. The fork moves in the direction of leading-strand synthesis.
DNA fibre technique A technique that is similar to chromosome combing, but in which cells on microscope slides are treated with detergent and the DNA is spread directly out of the lysed nuclei. Because of the lysis method, this technique is used in vertebrate cells but not in yeast.
Replication fork restart The resumption of fork progression after removal or bypass of a replication block.
Replication fork stabilization The maintenance of viable replication fork structures and the replication machinery during a replication block.
DNA helicase An enzyme that translocates on DNA and unwinds the double helix into ssDNA in an ATP-driven reaction. Annealing helicases use ATP to catalyse the reverse reaction.
homologous recombination29,30. Supporting this, DNA fibre analyses show that MRE11 facilitates replication fork restart after release from a short hydroxyurea-mediated block11 acting in the same pathway as the DNA damage signalling protein poly(ADPribose) polymerase 1 (PARP1). PARP1 seems to recruit MRE11 to facilitate fork restart by the resection of DNA ends11. BLM and WRN might function in the same pathway, as both physically or functionally interact with MRE11 (REFS 31,32). BLM promotes MRE11 foci formation in response to hydroxyurea31. Interestingly, WRN interacts with MRE11 during short (4 hour), but not longer (8 hour) hydroxyurea treatments, suggesting that the role of this interaction is specific to short replication blocks32. Observations from budding yeast suggest that MRE11 might alternatively or additionally promote fork restart by facilitating sister chromatid cohesion33, which also regulates normal replication fork progression in human cells34.
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
Replication origin The chromosomal location at which new replication factories are assembled and replication is initiated. Replication fork movement from origins is bidirectional.
RAD51 loading and strand invasion. The RAD51 paralogue X‑ray repair crosscomplementing protein 3 (XRCC3), which mediates RAD51 filament formation35, also promotes fork restart after a short hydroxyurea-mediated block 13. This implicates RAD51‑mediated filament formation in the restart process. Nevertheless, RAD51‑mediated fork restart does not involve RAD51 foci formation or detectable, long patch recombination in a reporter construct 13. This suggests that the extent of RAD51 loading and the recombination mechanism differ from homologous recombination‑mediated DSB repair. Visible foci would require numerous RAD51 molecules to be present at the forks in, for example, long filaments. Detectable recombination at a distant site might be prevented in the absence of a DSB at restarting forks. Fork restart that does not generate recombination products would be supported by BLM, which, in a complex with DNA topoisomerase IIIα (TOPIIIα), resolves double VOLUME 11 | O CTOBER 2010 | 685
© 2010 Macmillan Publishers Limited. All rights reserved
PROGRESS Hydroxyurea-mediated block (S phase)
Length of hydroxyurea treatment
0 hours
Stalled fork 5′ 3′ 5′ 3′ MUS81
Release from block (S phase)
SMARCAL1 BLM WRN PARP1 MRE11 XRCC3 RAD51
Fork restart
Completed replication
5′ 3′
5′ 3′
5′ 3′
5′ 3′
5′ 3′
5′ 3′ Collapsed fork 24 hours
Late S–G2 phase
New origin firing
PARP1 MRE11 XRCC3 RAD51 BLM WRN
5′ 3′ 5′ 3′ Homologous recombinationmediated repair
Figure 2 | Early replication fork restart and late replication fork repair. Most replication forks Nature Reviews | Molecular Cell Biology resume progression after release from short (2–4 hour) hydroxyurea-mediated blocks. Restart is promoted by the DNA helicases Bloom syndrome helicase (BLM) and Werner syndrome ATP-dependent helicase (WRN), the annealing helicase SWI/SNF-related matrix-associated actin-dependent regu lator of chromatin subfamily A-like protein 1 (SMARCAL1), the DNA damage signalling protein poly (ADP-ribose) polymerase 1 (PARP1; which recruits the nuclease and double-strand break (DSB) repair protein meiotic recombination 11 (MRE11)), and the homologous recombination proteins X-ray repair cross-complementing protein 3 (XRCC3) and RAD51. Stalled forks become inactivated during long (24 hour) hydroxyurea treatments. Within one hour after release from these long blocks, replication restarts by the firing of new origins. Fork inactivation coincides with the accumulation of DSBs that are dependent on the structure-specific endonuclease MMS and UV-sensitive protein 81 (MUS81). DSB formation does not seem to promote restart but does promote the RAD51‑dependent, postreplicative recombination repair of forks, shown here as synthesis-dependent strand annealing. PARP1, MRE11, BLM and WRN promote DSB repair. Hydroxyurea ‑induced DSBs can persist for more than 48 hours. Thus, after short hydroxyurea treatments, DNA replication resumes mainly by fork restart, whereas after long hydroxyurea treatments, replication restarts by the firing of new origins, followed by DSB repair of collapsed forks.
Holliday junctions in a process that avoids crossing over 36–38. DSB-mediated restart If mammalian replication forks are stalled for many hours, fork-associated DSBs are generated by MUS81–EME1 (REFS 1,6) (FIG. 1c), suggesting that DSB formation may play a part in replication fork restart, especially after long periods of replication inhibition.
Evidence for DSB-mediated fork restart. MUS81‑deficient mouse embryonic stem cells display a fork restart defect, especially after 24 hours of hydroxyurea treatment 1. This suggests that DSB generation does indeed promote fork restart. Interestingly, during treatments with low doses of aphidicolin that do not completely block replication, BLM and MUS81 facilitate a transient surge of DSBs during the first hour of treatment, and promote residual replication fork progression during that time39. This suggests that DSB-mediated fork restart can promote replication fork progression in the presence of low levels of a replication inhibitor.
Evidence against DSB-mediated fork restart. Other recent evidence suggests that DSB generation does not play a part in fork restart in human cancer cells. Although most stalled replication forks in human U2OS cells resume progression after short hydroxyurea treatments, they do not restart after long hydroxyurea treatments13 (FIG. 2). Furthermore, small interfering RNA depletion of MUS81 has no effect on fork restart in HeLa and U2OS cells (REF. 9; E.P and T.H, unpublished observations). These observations suggest that DSB formation at stalled forks is not necessarily part of a restart process. In Saccharomyces cerevisiae, initiation of BIR, a mechanism equivalent to DSB-mediated restart, takes several hours40. A similar delay could account for the observation that in U2OS cells, replication restarts by a new origin firing within one hour of being released from a long hydroxyureamediated block13. The conflicting reports described here could be due to different cell types and organisms used. In some rodent cells, increased fork inactivation could channel forks into DSB-mediated restart. Mouse embryonic stem cells and Chinese hamster ovary (CHO) cells display earlier
686 | O CTOBER 2010 | VOLUME 11
hydroxyurea‑induced RAD51 foci formation than human cells, and in CHO cells this is accompanied by increased fork inactivation1,13. WRN-deficient cells also display early fork inactivation and require MUS81 for efficient S phase progression after release from a hydroxyurea-mediated block, suggesting that MUS81 contributes to replication restart in these cells9. Fork inactivation and repair. Although inactivation of stalled replication forks coincides with the accumulation of DSBs, DSB formation does not seem to be the cause of fork inactivation as MUS81 depletion does not improve fork restart after long hydroxyurea treatments13. WRN-deficient cells display increased MUS81 activity, but MUS81 co-depletion does not rescue their fork restart defect 9. Instead, MUS81 promotes homologous recombination-mediated repair of the inactivated forks9. This suggests that fork inactivation might be caused by the dissociation of replication proteins and possibly fork remodelling. Fork cleavage by MUS81 is a result of fork inactivation and is a prerequisite for DSB-mediated fork repair, a process that also requires RAD51 (REF. 13). Other proteins that promote early fork restart (BLM, WRN, PARP1 and MRE11) also have roles in homologous recombination11,16,20,29,30 and may be involved in fork repair, depending on the length of replication inhibition. With most stalled forks inactivated after long hydroxyurea treatments, global replication restart is achieved by new origin firing 13. If homologous recombination (and possibly non-homologous end joining) can repair the vast number of DSBs, then genomic instability may be prevented. Conclusions The findings described here have helped to increase our understanding of mammalian replication fork restart. The proposed pathways are required for efficient restart of a subset of replication forks, suggesting that most forks are able to restart directly. A notable exception is the SET domain and mariner transposase fusion gene-containing protein (METNASE; also known as SETMAR), a primate-specific methyltransferase and nuclease that is required for the restart of most forks after hydroxyurea treatment 41,42. However, replication initiation is also suppressed in METNASE-depleted cells, suggesting that there is a general block to replication rather than a specific fork restart defect 42. Replication forks requiring special pathways for reactivation may have undergone additional structural changes. www.nature.com/reviews/molcellbio
© 2010 Macmillan Publishers Limited. All rights reserved
PROGRESS In such cases, remodelling of stalled fork structures by DNA helicases and homologous recombination factors promotes the efficient resumption of fork progression. DSB formation is not necessarily a first step in fork restart but can be a result of fork inactivation. Homologous recombination is activated for the post-replicative repair of collapsed forks. However, many aspects remain to be investigated. The molecular details of replication fork stabilization and Holliday junction‑mediated restart are not yet clear. Analysing the requirement for homologous recombination proteins that promote different steps during Holliday junction‑mediated restart will help elucidate the molecular mechanism of this process. Although most proteins mentioned are known to promote homologous recombination, it is currently unknown whether they act in the same restart pathway and subset of replication forks. It is also unclear how replication complexes are re-loaded at remodelled replication forks. Other pathways, such as checkpoint signalling 43–45 and sister chromatid cohesion46,47, stabilize stalled replication forks and may also promote fork restart (Supplementary information S3 (box)). Global inhibitors of replication, such as hydroxyurea and aphidicolin, differ from more physiological, stochastic inhibitors of individual forks in that they are more likely to activate cell cycle checkpoint signalling, which will affect the processing of stalled forks. It will therefore be interesting to develop methods to study rare fork‑stalling events. Inaccurate replication fork restart may be an important source of genomic instability. In the future, single-molecule analyses of DNA replication will become increasingly important for addressing this issue in mammalian cells. Eva Petermann is at the School of Cancer Sciences, University of Birmingham, Birmingham, B15 2TT, UK. Thomas Helleday is at the Gray Institute for Radiation Oncology and Biology, University of Oxford, Oxford, OX3 7DQ, UK, and at the Department of Genetics Microbiology and Toxicology, Stockholm University, S‑106 91 Stockholm, Sweden.
4.
5.
6. 7. 8.
9.
10.
11. 12. 13.
14.
15. 16.
17.
18.
19.
20.
21.
22. 23. 24.
e-mails:
[email protected];
[email protected] doi:10.1038/nrm2974 Published online 15 September 2010 1.
2. 3.
Hanada, K. et al. The structure-specific endonuclease Mus81 contributes to replication restart by generating double-strand DNA breaks. Nature Struct. Mol. Biol. 14, 1096–1104 (2007). Heller, R. C. & Marians, K. J. Replisome assembly and the direct restart of stalled replication forks. Nature Rev. Mol. Cell Biol. 7, 932–943 (2006). Ibarra, A., Schwob, E. & Mendez, J. Excess MCM proteins protect human cells from replicative stress by
25.
26.
27.
licensing backup origins of replication. Proc. Natl Acad. Sci. USA 105, 8956–8961 (2008). Ge, X. Q., Jackson, D. A. & Blow, J. J. Dormant origins licensed by excess MCM2–7 are required for human cells to survive replicative stress. Genes Dev. 21, 3331–3341 (2007). Lundin, C. et al. Different roles for nonhomologous end joining and homologous recombination following replication arrest in mammalian cells. Mol. Cell. Biol. 22, 5869–5878 (2002). Saintigny, Y. et al. Characterization of homologous recombination induced by replication inhibition in mammalian cells. EMBO J. 20, 3861–3870 (2001). Tuduri, S., Tourriere, H. & Pasero, P. Defining replication origin efficiency using DNA fiber assays. Chromosome Res. 18, 91–102 (2010). Davies, S. L., North, P. S. & Hickson, I. D. Role for BLM in replication-fork restart and suppression of origin firing after replicative stress. Nature Struct. Mol. Biol. 14, 677–679 (2007). Franchitto, A. et al. Replication fork stalling in WRN-deficient cells is overcome by prompt activation of a MUS81‑dependent pathway. J. Cell Biol. 183, 241–252 (2008). Sidorova, J. M., Li, N., Folch, A. & Monnat, R. J. Jr. The RecQ helicase WRN is required for normal replication fork progression after DNA damage or replication fork arrest. Cell Cycle 7, 796–807 (2008). Bryant, H. E. et al. PARP is activated at stalled forks to mediate Mre11‑dependent replication restart and recombination. EMBO J. 28, 2601–2615 (2009). Ciccia, A. et al. The SIOD disorder protein SMARCAL1 is an RPA-interacting protein involved in replication fork restart. Genes Dev. 23, 2415–2425 (2009). Petermann, E., Orta, M. L., Issaeva, N., Schultz, N. & Helleday, T. Hydroxyurea-stalled replication forks become progressively inactivated and require two different RAD51‑mediated pathways for restart and repair. Mol. Cell 37, 492–502 (2010). Liu, J., Xu, L., Sandler, S. J. & Marians, K. J. Replication fork assembly at recombination intermediates is required for bacterial growth. Proc. Natl Acad. Sci. USA 96, 3552–3555 (1999). Llorente, B., Smith, C. E. & Symington, L. S. Breakinduced replication: what is it and what is it for? Cell Cycle 7, 859–864 (2008). Karow, J. K., Constantinou, A., Li, J. L., West, S. C. & Hickson, I. D. The Bloom’s syndrome gene product promotes branch migration of holliday junctions. Proc. Natl Acad. Sci. USA 97, 6504–6508 (2000). Ralf, C., Hickson, I. D. & Wu, L. The Bloom’s syndrome helicase can promote the regression of a model replication fork. J. Biol. Chem. 281, 22839–22846 (2006). Bugreev, D. V., Yu, X., Egelman, E. H. & Mazin, A. V. Novel pro- and anti-recombination activities of the Bloom’s syndrome helicase. Genes Dev. 21, 3085–3094 (2007). Rassool, F. V., North, P. S., Mufti, G. J. & Hickson, I. D. Constitutive DNA damage is linked to DNA replication abnormalities in Bloom’s syndrome cells. Oncogene 22, 8749–8757 (2003). Constantinou, A. et al. Werner’s syndrome protein (WRN) migrates Holliday junctions and co-localizes with RPA upon replication arrest. EMBO Rep. 1, 80–84 (2000). Machwe, A., Xiao, L., Groden, J. & Orren, D. K. The Werner and Bloom syndrome proteins catalyze regression of a model replication fork. Biochemistry 45, 13939–13946 (2006). Sidorova, J. M. Roles of the Werner syndrome RecQ helicase in DNA replication. DNA Repair (Amst.) 7, 1776–1786 (2008). Yusufzai, T. & Kadonaga, J. T. HARP is an ATP-driven annealing helicase. Science 322, 748–750 (2008). Bansbach, C. E., Betous, R., Lovejoy, C. A., Glick, G. G. & Cortez, D. The annealing helicase SMARCAL1 maintains genome integrity at stalled replication forks. Genes Dev. 23, 2405–2414 (2009). Baumann, P., Benson, F. E. & West, S. C. Human Rad51 protein promotes ATP-dependent homologous pairing and strand transfer reactions in vitro. Cell 87, 757–766 (1996). Seigneur, M., Ehrlich, S. D. & Michel, B. RuvABCdependent double-strand breaks in DnaBts mutants require RecA. Mol. Microbiol. 38, 565–574 (2000). Robu, M. E., Inman, R. B. & Cox, M. M. RecA protein promotes the regression of stalled replication forks in vitro. Proc. Natl Acad. Sci. USA 98, 8211–8218 (2001).
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
28. Yoon, D., Wang, Y., Stapleford, K., Wiesmuller, L. & Chen, J. p53 inhibits strand exchange and replication fork regression promoted by human Rad51. J. Mol. Biol. 336, 639–654 (2004). 29. Buis, J. et al. Mre11 nuclease activity has essential roles in DNA repair and genomic stability distinct from ATM activation. Cell 135, 85–96 (2008). 30. Sartori, A. A. et al. Human CtIP promotes DNA end resection. Nature 450, 509–514 (2007). 31. Franchitto, A. & Pichierri, P. Bloom’s syndrome protein is required for correct relocalization of RAD50/ MRE11/NBS1 complex after replication fork arrest. J. Cell Biol. 157, 19–30 (2002). 32. Franchitto, A. & Pichierri, P. Werner syndrome protein and the MRE11 complex are involved in a common pathway of replication fork recovery. Cell Cycle 3, 1331–1339 (2004). 33. Tittel-Elmer, M., Alabert, C., Pasero, P. & Cobb, J. A. The MRX complex stabilizes the replisome independently of the S phase checkpoint during replication stress. EMBO J. 28, 1142–1156 (2009). 34. Terret, M. E., Sherwood, R., Rahman, S., Qin, J. & Jallepalli, P. V. Cohesin acetylation speeds the replication fork. Nature 462, 231–234 (2009). 35. Bishop, D. K. et al. Xrcc3 is required for assembly of Rad51 complexes in vivo. J. Biol. Chem. 273, 21482–21488 (1998). 36. Raynard, S., Bussen, W. & Sung, P. A double Holliday junction dissolvasome comprising BLM, topoisomerase IIIα, and BLAP75. J. Biol. Chem. 281, 13861–13864 (2006). 37. Singh, T. R. et al. BLAP18/RMI2, a novel OB‑fold‑containing protein, is an essential component of the Bloom helicase-double Holliday junction dissolvasome. Genes Dev. 22, 2856–2868 (2008). 38. Wu, L. & Hickson, I. D. The Bloom’s syndrome helicase suppresses crossing over during homologous recombination. Nature 426, 870–874 (2003). 39. Shimura, T. et al. Bloom’s syndrome helicase and Mus81 are required to induce transient double-strand DNA breaks in response to DNA replication stress. J. Mol. Biol. 375, 1152–1164 (2008). 40. Malkova, A., Naylor, M. L., Yamaguchi, M., Ira, G. & Haber, J. E. RAD51‑dependent break-induced replication differs in kinetics and checkpoint responses from RAD51‑mediated gene conversion. Mol. Cell Biol. 25, 933–944 (2005). 41. Lee, S. H. et al. The SET domain protein Metnase mediates foreign DNA integration and links integration to nonhomologous end-joining repair. Proc. Natl Acad. Sci. USA 102, 18075–18080 (2005). 42. De Haro, L. P. et al. Metnase promotes restart and repair of stalled and collapsed replication forks. Nucleic Acids Res. 10 May 2010 (doi:10.1093/nar/gkq339). 43. Feijoo, C. et al. Activation of mammalian Chk1 during DNA replication arrest: a role for Chk1 in the intra‑S phase checkpoint monitoring replication origin firing. J. Cell Biol. 154, 913–923 (2001). 44. Zachos, G., Rainey, M. D. & Gillespie, D. A. Chk1‑dependent S‑M checkpoint delay in vertebrate cells is linked to maintenance of viable replication structures. Mol. Cell Biol. 25, 563–574 (2005). 45. Scorah, J. & McGowan, C. H. Claspin and Chk1 regulate replication fork stability by different mechanisms. Cell Cycle 8, 1036–1043 (2009). 46. Leman, A. R., Noguchi, C., Lee, C. Y. & Noguchi, E. Human Timeless and Tipin stabilize replication forks and facilitate sister-chromatid cohesion. J. Cell Sci. 123, 660–670 (2010). 47. Tanaka, H. et al. Replisome progression complex links DNA replication to sister chromatid cohesion in Xenopus egg extracts. Genes Cells 14, 949–963 (2009).
Acknowledgements
We thank the Medical Research Council, the Swedish Research Council, the Swedish Children’s Cancer Foundation, the Swedish Pain Relief Foundation and the Swedish Cancer Society for financial support.
Competing interests statement
The authors declare no competing financial interests.
FURTHER INFORMATION Thomas Helleday’s homepage: http://helleday.gmt.su.se
SUPPLEMENTARY INFORMATION See online article: S1 (box) | S2 (box) | S3 (box) ALL LINKS ARE ACTIVE IN THE ONLINE PDF
VOLUME 11 | O CTOBER 2010 | 687 © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS 1 0 - Y E A R A N N I V E R S A RY S E R I E S
Revitalizing membrane rafts: new tools and insights Kai Simons and Mathias J. Gerl
Abstract | Ten years ago, we wrote a Review on lipid rafts and signalling in the launch issue of Nature Reviews Molecular Cell Biology. At the time, this field was suffering from ambiguous methodology and imprecise nomenclature. Now, new techniques are deepening our insight into the dynamics of membrane organization. Here, we discuss how the field has matured and present an evolving model in which membranes are occupied by fluctuating nanoscale assemblies of sphingolipids, cholesterol and proteins that can be stabilized into platforms that are important in signalling, viral infection and membrane trafficking. Caveola A 50–80‑nm, flask-shaped pit that forms in the plasma membrane and is enriched in caveolins, cavins, sphingolipids and cholesterol.
Max Planck Institute of Cell Biology and Genetics, Pfotenhauerstraβe 108, 01307 Dresden, Germany. e-mails:
[email protected];
[email protected] doi:10.1038/nrm2977
Cell membranes contain hundreds of lipids in two asymmetric leaflets1 and a plethora of proteins. For several decades, membrane research was dominated by the idea that proteins were the key factors for membrane functionality, whereas lipids were regarded as a passive, fluid solvent 2. Introducing the lipid raft concept in 1997, we postulated that sphingolipid–cholesterol–protein assemblies could function in membrane trafficking and signalling 3. These assemblies, or rafts, were thought to be characterized by their tight lipid packing, similar to the sterol-dependent, liquid-ordered phase in model membranes. The novelty of the raft concept was that it brought lipids back into the picture by giving them a function and by introducing chemical specificity into the lateral heterogeneity of membranes. When we wrote our first Review in this journal4, the emerging raft field had become increasingly confused by ambiguous methodology and imprecise nomenclature. Caveolae, for example, became synonymous with rafts but clearly represented only a subset of membrane assemblies defined by the action of the protein caveolin5. Complicating matters further was the size of the sphingolipid–cholesterol–protein assemblies being studied, which were below the resolution of light microscopy. Only after cross-linking did raft proteins and lipid constituents cluster together to form micrometre-size, quilt-like patches. Our focus in the first Review was to emphasize that rafts are small and dynamic and can be stabilized to form larger microdomains that function in membrane trafficking and signalling. We proposed that three types of assembly should be recognized in cell membranes — rafts, clustered rafts and caveolae (a subset of clustered rafts) — and that the residue remaining insoluble after detergent extraction should be called
detergent-resistant membrane (DRM) fractions. We also summarized the tools that were available for defining rafts and discussed their strengths and shortcomings. Obviously, what was known about lipid rafts and membrane organization at the time was dependent on the available methodology. The rationale of the present Review is to summarize where we stand today and to highlight the important role that new technology has had in moving the field forwards (BOX 1). We describe how membrane rafts are now defined as dynamic, nanoscale, sterol– sphingolipid-enriched, ordered assemblies of proteins and lipids, in which the metastable raft resting state can be stimulated to coalesce into larger, more stable raft domains by specific lipid–lipid, protein–lipid and protein–protein oligomerizing interactions (FIG. 1). The lipids in these assemblies are thought to be enriched in saturated hydrocarbon chains. We describe advances in our understanding of how lipid rafts function as a membrane organizing principle in cellular processes such as T cell signalling, viral infection and membrane trafficking, and also try to identify issues that need to be resolved.
Controversies then and now A key issue ten years ago was the methodology used to define a raft component. An increasing number of papers used detergents as the main criterion for raft association; raft constituents were defined simply as the insoluble residue or DRM remaining after non-ionic detergent solubilization at 4ºC. This criterion was usually combined with the use of methyl‑β-cyclodextrin to extract cholesterol from cell membranes. If the protein became detergent-soluble after cyclodextrin treatment, this strengthened the conclusion that it’s
688 | O CTOBER 2010 | VOLUME 11
www.nature.com/reviews/molcellbio © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS Box 1 | Changes in the raft concept Our concept of rafts has shifted with the realization that the association of components is dynamic and sizes range from small, short-lived, nanoscale assemblies to more stable membrane domains, the size and lifetime of which also vary. The apical membranes of epithelial cells, for example, can behave as large percolating raft domains116 or as a super-raft91,117. Similarly, the myelin sheets that oligodendrocytes produce to be wrapped around neuronal axons are another specific raft membrane type118. The situation is like that of logs in a river: one or several logs can function as a raft for one or more loggers and these can pile up into a raft jam. Discussing raft size as a criterion is irrelevant and the dynamics of raft lipids and proteins must be considered. Also, different methods and conditions used to study raft behaviour will give differing results. For example, fluorescence recovery after photobleaching (FRAP) may measure no difference in diffusion rates between raft and non-raft constituents, whereas stimulated emission depletion (STED) microscopy will30 (FIG. 2). The definition of a raft marker has evolved. Protein association with rafts was previously defined by sterol dependency and detergent resistance. Among putative raft lipids, the ganglioside GM1 was a commonly used marker because a fluorescently labelled probe was available. Patching of putative raft constituents was also done by antibody cross-linking in a way that would induce large raft clusters, potentially recruiting every membrane constituent with an affinity for the patched raft membrane environment. Now when nanoscale assemblies are analysed, one does not expect every marker to be enriched, but instead only a limited subset of proteins and lipids16. When these assemblies are clustered into raft platforms, there is no obligatory reason why GPI-anchored proteins, GM1 or other raft constituents should be enriched. Only when larger patches are produced, such as in phase-segregated plasma membrane preparations, would such markers be considered useful.
Förster resonance energy transfer A fluorescence-based method for detecting interactions between fluorophores that are <10 nm apart. It is dependent on the spectral overlap between donor and acceptor chromophores and uses non-radiative energy transfer from an excited donor molecule to excite an acceptor molecule.
Fluorescence polarization anisotropy A technique to measure rotational diffusion using changes in fluorescence polarization that are due to fluorophore rotation.
GPI-anchored protein (Glycosyl phosphatidylinositolanchored protein). One of a class of proteins that become post-translationally linked to GPI in the lumen of the ER.
Total internal reflection fluorescence (TIRF) microscopy An optical technique based on evanescent wave illumination (~150 nm into the sample) that is created by a totally internally reflected beam at the glass–water interface.
DRM association was indicative of it being a raft component. Finally, if a biological process in living cells was disrupted by cyclodextrin treatment, the process was considered to be raft-based. The uncritical use of these methods to study complex cellular processes caused an inflation of claims that were difficult to interpret and reconcile. Rightfully, this raised questions as to whether rafts were a real physio logical phenomenon6,7. The main criticisms were of course directed towards the use of detergent resistance as a defining factor for raft components8. Whereas physiologically induced changes in DRM composition can reflect lateral biases in the membrane, detergent solubilization is an inherently artificial method giving different results depending on the concentration and type of detergent, duration of extraction and temp erature9. Similarly, the other methods used to define raft-dependent processes were prone to artefacts and misinterpretations; for example, cyclodextrin treatment led to serious side effects such as lateral protein immobilization10. As plasma membranes can contain up to 40 mol% cholesterol11, it is perhaps not surprising that many cellular functions can be perturbed by cholesterol depletion12. For example, the plasma membrane can depolarize and Ca2+ stores can be induced to empty 13, leading to global cellular effects. The difficulty of visualizing rafts in cell membranes was an important concern, discussed critically even ten years ago4. Although microscopically observable patches were present after cross-linking with exogenous ligands, the variability of the colocalization and sizes seen led to doubts about their relevance. Moreover, approaches used to study membrane protein diffusion, such as fluorescence recovery after photobleaching (FRAP)14,
and molecular complexes in membranes, such as Förster resonance energy transfer (FRET)15, led to mixed results and thus caused scepticism about the raft concept. These studies also brought to the fore the issue of what constitutes a raft marker 16 (BOX 1). There were claims that the lipid raft field was at a technical impasse because the physical tools to study biological membranes were lacking 17.
Advances with new technology Despite the difficulties, the situation gradually clarified. The controversies inspired renewed efforts to find methodology that could detect and follow the behaviour of these small and elusive rafts. If they were present as postulated, one should be able to get glimpses of them before they clustered into more stable platforms. The problem required the development of high temporal and spatial resolution techniques to compare the location of different molecular constituents in the membrane. Visualizing rafts with new microscopy techniques. Single-molecule spectroscopy and microscopy techniques were, in principle, well-suited for this challenge because they should give access to the dynamic state of the membrane18–20. Indeed, an influx of novel methods, such as hetero- and homo-FRET and fluorescence polarization anisotropy, revealed that GPI-anchored proteins and other lipid-modified proteins form cholesteroldependent, nanoscale clusters21,22. Single-particle tracking also revealed the existence of dynamic nanoscale domains, and, on this basis, resting rafts were proposed to consist of a few raft proteins23. Another tracking method employed dual-colour total internal reflection fluorescence (TIRF) microscopy and single quantum dot tracking to study the cholesterol-dependent diffusion behaviour of a GPI-anchored protein and showed that this protein dynamically partitioned into and out of cell surface clusters of the ganglioside GM1 (REF. 24). Furthermore, fluorescence correlation spectroscopy (FCS) analysis of how membrane proteins behave in live cells revealed cholesterol- and sphingolipid-based nanoscale domains, into which proteins and lipids dynamically partition or assemble with a timescale of tens to hundreds of milliseconds25. Moreover, super-resolution optical microscopy methods have been introduced that provide resolution well beyond the diffraction limit 26; these include stim ulated emission depletion (STED) microscopy 27, photoactivated localization microscopy 28 and stochastic optical reconstruction microscopy (PALM and STORM)29. STED microscopy showed that sphingolipids and GPIanchored proteins are transiently trapped in cholesteroldependent molecular complexes in live cells30 (FIG. 2a). Super-resolution near-field scanning optical microscopy (NSOM)31 was also used in combination with quantum dots to show that T cell stimulation triggers nanoscale organization of T cell receptors (TCRs) in live cells32. Together, these methods revealed different aspects of how the behaviour of lipids and proteins is correlated in the plasma membrane.
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
VOLUME 11 | O CTOBER 2010 | 689 © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS a
Transmembrane non-raft protein GPI-anchored protein Transmembrane raft protein GPL
Cholesterol GSL
Actin Nanoscale assembly
Nanoscale assembly Acylated protein
b
c
Raft platform
Raft phase
Figure 1 | Raft-based heterogeneity in cell membranes. a | Nanoscale assemblies of Nature Reviews Cell Biology sterols such as cholesterol, sphingolipids such as sphingomyelin and| Molecular glycosphingolipids (GSLs), and proteins in the plasma membrane fluctuate in composition. GPI-anchored proteins, transmembrane raft proteins and acylated cytosolic proteins are postulated constituents of these assemblies, which can be modulated by actin filaments. Not much is known about the state of nanoscale assemblies in the cytosolic leaflet of the membrane. Transmembrane non-raft proteins are excluded from these assemblies. b | In response to external signals or the initiation of membrane trafficking events, raft platforms are formed from fluctuating assemblies through lipid–lipid, lipid–protein and protein–protein oligomerizing interactions. These platforms are important for membrane signalling and membrane trafficking. c | Micrometre-sized raft ‘phases’ can be induced at equilibrium. This state can be seen in model systems such as giant unilamellar vesicles (GUVs) and also in giant plasma membrane vesicles (GMPVs) or plasma membrane spheres released from the cell. Figure modified, with permission, from REF. 55 © (2010) the American Association for the Advancement of Science.
Quantum dot A nanoscale semiconductor crystal used as a label in fluorescence microscopy owing to its high emission and photostability.
Despite the recent wealth of information gained from peering below the previously inviolable ‘diffraction limit’, all of the techniques mentioned above have their inherent disadvantages. FCS is diffraction-limited and therefore must be extrapolated25,33,34 or combined with a super-resolution technique like STED30 to reach the nanoscopic world. NSOM requires nanometre proximity to a tip or surface, which may influence the system under study 35. A further issue is the use of labels that are bulky or chemically foreign to the cell, usually fluorescent tags,
beads or quantum dots. Although there are attempts to tackle the bulkiness of these tags36, it remains to be seen whether these tagged (often overexpressed) constructs faithfully mimic the behaviour of the unmodified native membrane components. Despite these limitations, data from these methods conclusively support the existence of nanoscale, cholesterol-assisted, dynamic and selective assemblies. Lipid diversity revealed by lipidomics. Lipid analysis has dramatically improved in the past decade37. The field has been energized by the improved technical capabilities of tandem mass spectrometers that now allow fast quantitative profiling of lipidomes from minute amounts of sample. The analytical precision of modern instruments has made it possible to routinely identify lipids as molecular species and to analyse the compositional diversity of lipids in an unprecedented manner. Previously, most lipid analyses of membranes quantified composition in terms of lipid classes and not the molecular diversity in each lipid class (FIG. 2c). To understand how this diversity is used by the cell, sensitive and quantitative lipidomic techniques will have to be used to define the specificity that governs lipid assembly 38. The first analyses of raft clusters in activated TCR domains39, raft transport carriers40 and raft viruses41 reveal selective enrichment of lipid species, as predicted by the raft hypothesis. Insights from biophysical analysis. In parallel with biochemical and analytical studies of cell membrane organization, biophysicists have been characterizing model systems using monolayers and bilayers to explore liquid– liquid immiscibility (the inability to mix)42,43. There are two relevant phases of membranes: liquid disordered and liquid ordered. The liquid-ordered phase is typically enriched in raft lipids, such as sphingolipids, cholesterol and saturated phospholipids42. In three-component lipid mixtures reconstituted into giant unilamellar vesicles (GUVs) or supported bilayers44,45, varying sizes of the liquid-disordered and liquid-ordered phases can be seen with light and atomic force microscopes, depending on the composition46. The sizes of these domains can be altered from large to microscopically undetectable by adjusting the composition of the lipid mixture. Similarly, raft-like domains are visible in GUVs from isolated rat kidney membrane lipids47. Oligomerizing raft lipids (for example, by cross-linking a minor solute such as the ganglioside GM1, a typical raft lipid) was shown to have a dramatic effect in promoting phase separation48 (FIG. 2b). How do these findings in simple model systems relate to cell membranes? Given that cell membrane bilayers are asymmetric in their composition, the phase behaviour in each leaflet could be unique and coupled by interactions at the bilayer midplane49. Cell membranes are certainly not at equilibrium; they are continually perturbed by compositional fluctuations caused by lipid metabolism, the action of flippases, lipid transporters and membrane traffic, and the return to equilibrium can be slow. It has been argued that the chemical composition of the outer monolayer of cell plasma membranes endows them with
690 | O CTOBER 2010 | VOLUME 11
www.nature.com/reviews/molcellbio © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS a
b
STED
Confocal
CTX cross-linking Synthetic lipids
GUV Phase-separated GUV CTX cross-linking
Cell swelling
PMS Phase-separated PMS
250nm
Cooling Cell blebbing
<50nm
GPMV Phase-separated GPMV
c Membrane purifications
Mass spectrometry analysis
Lipidome analysis 35
Lipid classes
30
Total cell extract FusMid vesicles TGN-E
Detection of molecular lipid species
Mol (%)
25 20 15 10
2
M
ra
st
Ce
go Er
)C
MIPC
(IP
id
e
ol
IPC
m
er
ro
ns
dG
dI
Pt
Pt
dC Pt
Pt
Pt
dE
tn
er
PA
dS
DAG
ho
5 0
Lipid species 815
m/z
10
820 Mol (%)
810
5
:1
ho
18
:0-
18
:1
:1
18 :118 dC Pt
Pt
dC
ho
:0-
18 :1Pt
dC
ho
16
16
18
:1
:1 16 :0-
Pt
dC
ho
16 :1-
16
16 Pt
dC
ho
ho
16 :0-
14
dC Pt
ho
:1
:0
:1
:1
16 :0dC
Pt
Pt
dC
ho
14
14
:1-
16
:0
Pt
dC
ho
0-
16
:1 16
12:
ho dC
Pt
Pt
dC
ho
12:
0-
16 :0-
10 ho
dC Pt
Pt
dC
ho
10
:0-
16
:1
:0
0
Figure 2 | Novel methodology for the study of rafts. a | High-resolution stimulated emission depletion (STED) microscopy (right) is one of several novel nanoscopic techniques that can go beyond the diffraction limit of standard confocal laser scanning microscopes (left). The large detection area of confocal microscopes (~250 nm) cannot discern | Molecular Cell to Biology details such as molecules moving freely or being transiently trapped on small spatialNature scales.Reviews However, STED is able discriminate between molecules that diffuse freely (red) and those that are hindered (blue) during their passage through the subdiffraction spot (<50 nm). Red and blue lines indicate diffusion traces. b | Homogenous giant unilamellar vesicles (GUVs) can be prepared from defined lipids, and phase separation can be induced by cross-linking the ganglioside GM1 with cholera toxin (CTX)48. Plasma membrane spheres (PMS), produced from a cell swelling procedure, can also recapitulate the same effect in membranes containing the lipid and protein complexity of a plasma membrane53. Similarly, giant plasma membrane vesicles (GPMVs) produced by blebbing can be induced to phase separate by cooling51 or CTX cross-linking122. c | A simplified workflow of a lipidome analysis. Biochemically purified membranes from trans-Golgi network (TGN)-derived (FusMid) vesicles (red), purified membranes from the donor organelle (TGN‑E; blue), and total cell membranes (green) are analysed by mass spectrometry to reveal the amount of each molecular species in the sample. The sum of the species can be further analysed for specific features such as fatty acid length, saturation and abundance37,123. The mole percentage of each lipid class in yeast post-Golgi vesicles is shown together with the species distribution of the phosphatidylcholine (PtdCho) lipid class. Lipid species are denoted as, for example, PtdCho 10:0-16:1, in which a fatty acid with 10 carbon atoms and no double bonds and a fatty acid with 16 carbon atoms and 1 double bond are present. DAG, diacylglycerol; IPC, inositolphosphoceramide; MIPC, mannosyl-IPC; M(IP)2C, mannosyldiinositolphosphoceramide; PA, phosphatidic acid; PtdEtn, phosphatidylethanolamine; PtdGro, phosphatidylglycerol; PtdIns, phosphatidylinositol; PtdSer, phosphatidylserine. Image in part a is modified, with permission, from Nature REF. 30 © (2009) Macmillan Publishers Ltd. All rights reserved. Data in part c is reproduced, with permission, from REF. 40 © (2009) The Rockefeller University Press. NATURE REVIEWS | MOLECULAR CELL BIOLOGY
VOLUME 11 | O CTOBER 2010 | 691 © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS Box 2 | Critical behaviour in membranes
Fluorescence correlation spectroscopy A technique that measures diffusion by correlating the fluorescence signal of a diffusing fluorophore with time.
Stimulated emission depletion A nanoscopic technique that uses a red-shifted beam to deplete the emission of the periphery of the excitation spot and create a smaller excitation region, thus overcoming the diffraction limit.
PALM and STORM (Photoactivated localization microscopy and stochastic optical reconstruction microscopy). Super-resolution microscopy techniques that use stochastically photoactivated fluorescent probes to reconstitute the full image from individual point spread functions.
Near-field scanning optical microscopy A super-resolution technique that exploits the evanescent wave near the surface of the sample by placing the detector close to the sample.
Glycosphingolipid A lipid that contains at least one sugar residue and a ceramide (N-acylated sphingoid).
Palmitoylation The reversible covalent attachment of fatty acids to Cys residues of membrane proteins, which promotes their membrane association.
Major histocompatibility complex A complex of genetic loci in higher vertebrates that encodes a family of cellular antigens that allow the immune system to recognize self from non-self.
Domain-forming lipid mixtures exhibit critical fluctuations near miscibility transition points before passing the boundary to stable microscopic liquid ordered–liquid disordered phase separation119. These fluctuations arise because the energy required to maintain regions of different composition becomes vanishingly small, and thermal motions lead to composition fluctuations over a wide range of time and length scales120. A remarkable similarity between plasma and model membranes is that the giant plasma membrane vesicles (GPMVs) produced by blebbing from plasma membranes display critical behaviour. In the two-phase region, the energetic cost (line tension) of the interface between two coexisting domains approach zero as the temperature approaches the miscibility transition boundary. Micrometre-scale fluctuations occur and suggest that the plasma membrane composition is tuned to a critical point. This behaviour can be extrapolated to physiological temperature, suggesting that, with the composition of the plasma membrane, the heterogeneity would correspond to less than the 50 nm-sized compositional fluctuations seen in some cell types121.
the propensity for phase separation (if at equilibrium)50. Even if no microscopically observable phase separation can be detected, the propensity for sub-microscopic domain formation may still exist. Clearly, the phase segregation seen in liquid-phase model systems is a manifestation of liquid–liquid immiscibility. So if one were to leave a plasma membrane in a resting state (no lipid synthesis or metabolism, no exocytosis or endocytosis, no inter actions with peripheral proteins and no actin cytoskeleton), would membrane components phase‑separate into large domains with time? This question has yet to be answered, but lipid-based phase separation into liquid-ordered-like and liquiddisordered-like phases has been seen in isolated giant plasma membrane vesicles (GPMVs) formed by membrane blebbing 51, and the temperature- and cholesteroldependence of this phase separation resembles that of simple model systems52 (FIG. 2b). Furthermore, plasma membrane spheres (PMSs), produced using a swelling procedure and separated from the influence of cyto skeletal and trafficking events, showed sterol-dependent coalescence into a micrometre-scale phase on clustering by cholera toxin at 37ºC (FIG. 2b). The selective lateral reorganization of proteins and lipids correlated with their predicted affinity for raft domains53. This phase-separating behaviour is similar to the cholera toxin-induced phase separation seen in model membranes, suggesting that the protein and lipid composition of the plasma membranes is positioned close to a phase boundary and can be induced to phase separate by the clustering of raft components such as the ganglioside GM1. Together, these studies have provided an exciting and unexpected new dimension in membrane research, showing that cell membranes of complex lipid and protein compositions can phase separate similarly to
simple model membranes and that the composition might be tuned close to a critical point (BOX 2). It should be pointed out that the two phases induced by cholera toxin in PMS behave differently from liquid-ordered and liquid-disordered phases in model membranes in terms of the amount of order and packing 54. These differences in organization are perhaps not surprising, considering that the plasma membrane is crowded with proteins and specific lipid–protein interactions. Hence, we are left with a description of plasma membranes that could correspond to three states, of which two are seen in living cells55 (FIG. 1). The first state is represented by dynamic nanoscale assemblies that associate and dissociate on a subsecond timescale56,57. These can be clustered into the second state to generate more stable, selective and functional platforms. The third state is the complete micrometre-scale phase separation that is seen only in isolated membranes at equilibrium. How the critical fluctuations observed in GPMVs relate to associating and dissociating nanoscale assemblies in living cells is still unclear. Obviously, the dynamic behaviour will depend on the composition of the membranes, and therefore on the cell type that the plasma membrane is derived from (BOX 1).
Biological roles of raft heterogeneity The recent advances of high-resolution microscopic methodology together with observations of large-scale phase separation in plasma membranes have confirmed the potential for lipid phase separation. To illuminate the possible roles of rafts in cellular functions, we focus on four areas of cell biology —TCR signalling, HIV assembly, endoplasmic reticulum (ER)-to-Golgi and postGolgi trafficking to the cell surface and glycosphingolipidmediated endocytosis — and review progress made in the past decade. T cell signalling. An early clue that rafts might affect T cell signalling was the observation that antibody-mediated cross-linking of GPI-anchored proteins (which do not span the membrane) could stimulate signalling 58. Later, DRM analysis showed that factors important for T cell signalling were detergent-insoluble, whereas engineered palmitoylation -deficient proteins became detergentsoluble and impaired T cell activation. Cholesterol depletion inhibited T cell activation, whereas co-patching experiments using cholera toxin induced part of the T cell-activation programme and lead to microscopically observable domains containing essential T cell-activation proteins. Taken together, these data suggested that lipid rafts are involved in T cell signalling. T cells are activated by conjugation with cognate antigen-presenting cells (APCs). TCRs interact with the antigenic peptide bound to the major histocompatibility complex (pMHC) on APCs59, and this leads to the phosphorylation of immunoreceptor Tyr-based activation motifs in the TCR multisubunit complex by lymphocyte cell-specific protein Tyr kinase (LCK; a Src family kinase) and subsequent recruitment and activation of the Tyr kinase 70 kDa ζ-associated protein (ZAP70) (FIG. 3). In turn, activated ZAP70 phosphorylates linker for T cell
692 | O CTOBER 2010 | VOLUME 11
www.nature.com/reviews/molcellbio © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS APC Plasma membrane
pMHC
TCR APC
cSMAC
T cell
LCK Plasma membrane
P P P
Actin
T cell
ZAP70
P P P
P P P
LAT
cSMAC
Raft platform
P P P
LAT
CD3 Lipid modification
Downstream signals
Figure 3 | T cell receptor activation. T cell receptors (TCRs) in the resting state might associate with raft lipids to form Nature Reviews Molecular Cell Biology nanoscale assemblies that stabilize into a raft platform (TCR microclusters) on activation of the T cell|by the antigenic peptide-bound major histocompatibility complex (pMHC) group on an antigen-presenting cell (APC). Here, the TCR gets phosphorylated by the Src family kinase lymphocyte cell-specific protein Tyr kinase (LCK), and recruits and activates the Tyr kinase 70 kDa ζ-associated protein (ZAP70). The main target of ZAP70 is linker for T cell activation (LAT), which initiates further downstream signalling. The TCR clusters are anchored to actin filaments and are transported towards the central supramolecular activation cluster (cSMAC), the central part of the immunological synapse. The TCR microcluster signalling is envisaged to take place in a raft platform39. For clarity, only a subset of the involved proteins is shown.
activation (LAT), a transmembrane adaptor protein that recruits several signalling molecules to initiate downstream events such as actin polymerization, Ca2+ influx, Ras activation and transcriptional changes. This cascade of events results in the formation of an ‘immunological synapse’ between the contacting cells, with a bull’s eye structure that consists of a central supramolecular activation cluster (cSMAC) and a surrounding ring of adhesion molecules. The model of T cell synapse formation by stimulated raft condensation generated considerable criticism. The effects of cholesterol depletion, for example, could be explained by the previously mentioned pleiotropic effects of this treatment on, for instance, Ca2+ influx 13. Moreover, the immunological synapse itself, first considered to be a stabilized raft platform, was found to be a dynamic structure59. TCRs are first engaged and activated in TCR microclusters, which are subsequently transported radially to the centres of the contact by actin filaments, generating a cSMAC, in which many TCRs are already dephosphorylated and inactive60,61. Using dual-colour, single-molecule microscopy, microclusters have been observed with protein–protein interactions involving TCR, LAT, LCK and CD2 (but not the phosphatase CD45 (REF. 62)). The data suggested that diffusional trapping through protein–protein interactions created microdomains that include specific cell surface proteins to facilitate T cell signalling. A non-raft mutant of LAT lacking palmitoylation sites was also trapped by the microclusters, whereas a raft-associated construct, the lipid-modified amino terminus of LCK, was not concentrated in the microclusters63 (BOX 1). Thus, there was no direct evidence for lipid raft involvement, but it should be noted that lipids were not included in the analysis. An important issue in T cell signalling is how TCRs connect to 10–100 cognate pMHC molecules among a total cell surface pool of 104–105 MHC molecules on an APC64. The kinetics of binding between TCRs and the
cognate pMHC have been evaluated using engineered proteins in solution, and only weak affinities and dissociation rates were seen. However, when binding was measured between a TCR in a T cell plasma membrane and the cognate pMHC integrated into another membrane, accelerated kinetics and a more than 100‑fold higher affinity were observed65,66. These data explain the rapid and efficient recognition of an APC and also show that TCRs can serially engage a few cognate pMHCs in a large, self-MHC background. Cholesterol depletion was found to reduce the effective two-dimensional affinities between TCRs and pMHCs66. One issue in the field is the oligomeric state of TCRs before interaction with pMHCs. There are mixed reports as to whether it is monovalent 67,68 or multivalent 69, but high-resolution microscopy methods show complexes with up to 7–20 TCRs, which are manifested as protein islands that are 70–140 nm in diameter 70. How can these findings be reconciled with monovalent TCRs? What is the molecular size of a resting state TCR? One suggested possibility is that the contact of a cell with a glass surface perturbs membrane organization such that weak signalling of TCRs occurs without ligation71. This is reminiscent of live-cell analysis by photonic force microscopy, which concluded that GPI-anchored protein domains have a cholesterol-dependent size of 50 nm72, much larger than the <10 nm measured by homo- and hetero-FRET21. In these experiments, beads coated with antibodies against the GPI-anchored protein were immobilized by an optical trap to measure the unhindered diffusion on the cell surface, and effects of the trap on the lifetime and size of the nanoscale, GPI-anchored protein assemblies cannot be excluded. These studies therefore highlight how experimental conditions can easily lead to divergent results. If the resting states of raft proteins were indeed associated with nanoscale assemblies of raft lipids, such as sphingolipids and cholesterol, then such variances in size could arise from the coalescent capacity of these
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
VOLUME 11 | O CTOBER 2010 | 693 © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS Env
GPI-anchored protein
Budding Multimerization Membrane binding Plasma membrane
Raft
Polyunsaturated fatty acid
PtdIns(4,5)P2 Matrix domain Myristate
P P
P P
P P
P P
P P
P P
Gag Raft platform
Figure 4 | HIV assembly and release. Gag binding induces the formation of raft platforms Nature Reviewsdrives | Molecular Cell Biology at the plasma membrane of the host cell and Gag multimerization the assembly of the virus particle. The viral Env protein is incorporated during the budding process and the virus detaches from the plasma membrane. To emphasize that not all host proteins are excluded from the virus particle, a GPI-anchored protein is shown included124,125. One possibility is that during raft platform formation, Gag binding to phosphatidylinositol‑ 4,5‑bisphosphate (PtdIns(4,5)P2) through its matrix domain releases the bound myristate group, which inserts into the plasma membrane83. The polyunsaturated fatty acid of PtdIns(4,5)P2 has been proposed to flip into the hydrophobic cleft of the matrix domain, and multimerization of the Gag protein stabilizes the raft platform. Figure modified, with permission, from REF. 76 © (2009) Elsevier.
Lipid shell A model proposing that specific membrane proteins bind to complexes of cholesterol and sphingolipids or laterally organize specific contacting lipids.
STALL (Stimulation-induced temporary arrest of lateral diffusion). The cholesterolassisted, sub-second trapping of GPI-anchored signalling proteins with downstream signalling proteins.
fluctuating structures. Stabilization of fluctuating nanoscale structures could mirror a physiological process in which increased access to LCK mediates the stimulation of TCR signalling 73. Lipidomics has been used to measure immunoisolated TCR activation domains purified by coating magnetic beads with TCR-activating antibodies that induce TCR cross-linking. When the lipidome of the immuno-isolated TCR domains was compared to similarly cross-linked transferrin receptor plasma membrane domains from T cells, TCR domains were enriched in sphingolipids, saturated phosphatidylcholines (PtdChos) and plasmenyl phosphatidylserine (PtdSer)39. Together, these data are difficult to explain without assuming that lipid heterogeneity is functionalized to activate T cells. What is missing are data on whether and how the lipid context around the receptors is changing from the inactive to the activated state and back again. This is a general issue in the raft field: how are the ‘lipid shells’, the lipids contacting the raft proteins, composed and organized74? Do lipid–protein interactions
allosterically change receptor conformations? The potential relevance of dynamic, nanoscale raft assemblies for other signalling processes in which rafts have been implicated is obvious but requires further analysis. An example is the STALL concept, in which clusters of GPI-anchored CD59 recruit the G protein subunit Gαi2 and the Tyr kinase LYN through protein–protein and protein–raft interactions, resulting in temporary immobilization of the cluster by binding to filamentous actin (F‑actin) and signal activation by phospholipase Cγ2 (PLCγ2)75. Virus budding. Many viruses acquire a membrane envelope when budding off from the host cell plasma membrane. Some viruses, including HIV76 and influenza77, seem to do this by organizing a lipid raft domain around their nucleocapsid that includes viral glycoproteins and excludes most host cell surface proteins from the budding viral envelope. The Gag protein of HIV, the matrix domain of which assembles with the Env glycoprotein in the plasma membrane, becomes detergent-resistant while driving the budding process (FIG. 4); furthermore, budding is cholesterol- and sphingolipid-dependent. If labelled cholera toxin is applied to HIV-expressing cells, Gag, GM1 and the virus proteins co-patch in distinct clusters that segregate away from clusters of non-raft transferrin receptors76. These data suggested that the assembly of the virus envelope at the host cell plasma membrane involves the clustering of rafts. In support of this hypothesis, the lipidome of purified HIV particles showed that sphingolipids, cholesterol, plasmenyl phosphatidylethanolamine (PtdEtn), PtdSer and saturated PtdChos were enriched in the HIV membrane relative to total host cell membranes41. Comparison of the HIV lipidome with that of the host cell plasma membrane found that only cholesterol, the ganglioside GM3 and ceramide78 were highly enriched in the virus envelope79. Consistent with the enrichment of raft lipids, the viral membrane was shown to have an ordered lipid packing by spectroscopy using the fluorescent probe laurdan80, which reports relative membrane order, with lipid composition being its main determinant, in an experimental setup that is not affected by geometry 54. The phosphoinositides phosphatidylinositol phosphate (PtdInsP) and PtdIns‑4,5‑bisphosphate (PtdIns(4,5)P2) are also enriched in the HIV envelope79 and PtdIns(4,5)P2 is bound by Gag. Structural models are now emerging for how the association of PtdIns(4,5)P2 with Gag may drive membrane reorganization and viral budding. PtdIns(4,5)P2 is negatively charged owing to two phosphates in the inositol headgroup and it contains both a saturated fatty acid and a polyunsaturated arachidonic acid. Why would a phospholipid with a polyunsaturated fatty acid be included in a raft domain? Proteins such as myristoylated Ala-rich C‑kinase substrate (MARCKS) and growth-associated protein 43 (GAP43) that bind to PtdIns(4,5)P2 have been proposed to polymerize and tightly cluster PtdIns(4,5)P2 in the cytosolic leaflet of the membrane and partition PtdIns(4,5)P2 into raft domains 81,82. But the matrix domain of the Gag protein may do this differently 83.
694 | O CTOBER 2010 | VOLUME 11
www.nature.com/reviews/molcellbio © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS The matrix domain of Gag, like in MARCKS, has a cluster of basic amino acids that binds to PtdIns(4,5)P2 and it also has a saturated myristate fatty acid at its N terminus buried in a hydrophobic cleft. The hypothesis has been put forward that Gag multimerizes during budding and, on binding to PtdIns(4,5)P2, the N‑terminal myristate becomes exposed on the matrix domain surface to insert into the cytosolic leaflet of the plasma membrane, to be replaced by the polyunsaturated arachidonic acid flipping from the membrane into the myristate pocket on the matrix domain. This exchange would result in a multimeric cluster of saturated fatty acid tails that penetrate into the bilayer to promote raft clustering around the HIV Gag nucleocapsid83. Polyunsaturated fatty acids such as arachidonic acid have an extremely flexible and highly disordered structure, and poor packing with cholesterol could promote the flipping out of the polyunsaturated chain into the pocket on the matrix domain84. Viruses ingeniously make use of the cellular machinery to replicate in host cells. If this postulated mechanism of PtdIns(4,5)P2 interaction with the matrix domain is correct, it is likely that native cellular proteins could make use of similar tricks for association with, or condensation of, raft domains.
Rafts in membrane trafficking ER-to-Golgi traffic in yeast. Secretory proteins are inserted into the ER and transported by coat protein II (COPII) vesicles to the Golgi and subsequently to their final destinations. In the ER of yeast, secretory proteins are sorted into at least three types of ER exit site85. Two of these sites concentrate mainly soluble or transmembrane cargo, and early COPII machinery was found to be necessary for these functions. The third exit site also produces COPII vesicles, but COPII machinery is not required to concentrate its predominantly GPI-anchored cargo85. This concentration depends on the remodelling of GPI anchors with a saturated, long-chain fatty acid or a ceramide that confers detergent resistance86,87. Defects in GPI-anchor synthesis dramatically reduce the total sphingolipid levels88, and ER exit of GPI-anchored proteins depends on ceramide synthases89,90. Therefore, the sorting mechanism for this type of exit site might be a concentration of ceramides that attracts GPI-anchored proteins but tends to exclude most transmembrane proteins85.
Myristate A group that is attached to a protein through an amide bond by the irreversible, co-translational process of myristoylation. This is an important modification for membrane targeting.
Shiga toxin One of a family of protein toxins produced by bacteria that can cause dysentery.
Post-Golgi traffic to the cell surface. One of the main tenets of the lipid raft hypothesis was the prediction that the transport machinery in the trans-Golgi network (TGN) of epithelial cells sorts lipids and proteins into common carrier vesicles for targeted delivery to the apical or basolateral surface91. The postulated apicalsurface raft carriers have remained elusive and have so far not been isolated. In yeast, there are also two main pathways to the cell surface92: one that transports plasma membrane proteins (including GPI-anchored proteins) directly, and one that uses endosomal compartments as an intermediate station to transport soluble secreted proteins such as invertase to the cell surface (akin to what has been suggested for the basolateral route
in epithelial Madin–Darby canine kidney (MDCK) cells). A genome-wide visual screen identified several enzymes involved in sterol and sphingolipid synthesis that are essential for the delivery of a raft marker protein to the cell surface93. Thus, the yeast raft lipids, which are the cholesterol homologue ergosterol and three classes of sphingolipids (inositolphosphoceramide, mannosyl-inositolphosphoceramide and mannosyldiinositolphosphoceramide)94, were shown to regulate the delivery of detergent-resistant cargo to the plasma membrane. Consistent with the notion of raft-enriched carriers, immuno-isolation and subsequent quantitative analysis by mass spectrometry of TGN-derived vesicles isolated using a raft marker protein as bait showed that ergosterol and the most complex yeast sphingolipid, mannosyl-diinositolphosphoceramide, were selectively enriched and were the most abundant lipids in the transport vesicles40 (FIG. 2). These were the first data to give direct experimental support to the hypothesis that raft cargo proteins are delivered from the Golgi complex to the cell surface in a raft carrier. Interestingly, laurdan analysis of the immuno-isolated yeast carrier vesicles showed that their membranes were more condensed than the membrane of the donor compartment, supporting a change in membrane packing during sorting. The protein and lipid-sorting process therefore probably involves raft clustering to drive segregation in the membrane of the TGN95. Biophysical studies of yeast raft lipids have shown that they can phase separate into liquid-ordered and liquid-disordered domains in GUVs96. Therefore, the formation of the raft-selective 100‑nm carrier vesicles at the yeast Golgi membrane is envisioned as an interplay between raft clustering (induced by unidentified proteins), line tension and curvature, and additional machinery, including tensional forces generated by actin-based motors (FIG. 5b). Curvature is probably facilitated by bending proteins, such as the yeast BAR (Bin–Amphiphysin–Rvs) protein Rvs161 (reduced viability on starvation protein 161), which was identified in the plasma membrane delivery screen. Glycosphingolipid-mediated endocytosis. Toxins and viruses have proven to be important tools for the study of endocytosis as they can be taken up by various internalization mechanisms97. Here, we consider Shiga toxin and the polyoma virus Simian virus 40 (SV40), both of which bind glycosphingolipids as surface receptors to become internalized by a glycosphingolipid-mediated clustering process. Shiga toxin uses both clathrin-mediated and nonclathrin-dependent endocytosis 98 . Non-clathrindependent endocytosis has been studied in detail. The homopentameric Shiga toxin can bind up to 15 copies of its receptor globotriaosylceramide (Gb3)99. Binding on the plasma membrane causes the formation of narrow tubular invaginations that use the dynamin scission machinery to be released from the plasma membrane into the cytoplasm100. Tubule formation requires no energy but is regulated by membrane tension. Studies using phase-separated GUVs showed that the binding
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
VOLUME 11 | O CTOBER 2010 | 695 © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS a
b Coat-mediated trafficking Domain-mediated trafficking StxB
Raft platform
Protein cluster
Gb3 Plasma membrane 4
3
Coat
Adaptor protein
2
Cytosol Sorting signal
Acylated protein
Raft protein
1 Lumen
Cargo
GSL GPI-anchored protein
Lectin
Raft transport container
Plasma membrane Coated vesicle Endosomal compartment
TGN
Figure 5 | Rafts in membrane trafficking. a | Glycosphingolipid-mediated endocytosis. The pentavalent Shiga toxin B (StxB) binds to its receptor, the glycosphingolipid Gb3, inducing a clustering process that leads to invagination of the plasma membrane into the cytosol. b | Post-Golgi traffic to the cell surface. Cargo destined for the endosomal compartments in yeast or the basolateral plasma membrane in epithelial cells is sorted by adaptor proteins and coat proteins such as clathrin (shown as coat-mediated trafficking). Proteins containing the appropriate cytoplasmic sorting signals (1) are recognized by adaptor proteins on which the coat proteins assemble (2). For clathrin-mediated sorting, membrane bending (3) and subsequent budding (4) is driven by the formation of the coat, aided by specific bending proteins (not shown). Proteins sorted into a raft transport container (shown as domain-mediated trafficking) are clustered into a raft platform as described in FIG. 1. The platform includes raft constituents and excludes non-raft proteins and lipids. Membrane bending is facilitated by the line tension between the domains. Sorting is brought about by clustering agents, which might be lumenal lectins or cytosolic peripheral proteins. GSL, glycosphingolipid. Figure in part a is modified, with permission, from REF. 102.
Line energy The energy arising from the unfavourable interaction or ‘tension’ of two phasesegregating membrane domains. It is the product of line tension and interaction length.
of Shiga toxin to Gb3 takes place in the liquid-ordered phase, which induces a more condensed phase and segregation of Gb3 into high-density clusters. Moreover, Shiga toxins may locally create an asymmetric stress in the external leaflet of the bilayer that leads to bending towards the protein; this effect seems to depend on wedge-shaped Gb3 species100 (FIG. 5a). Additionally, the toxin molecules in the Gb3 cluster can experience attractive interactions, arising solely as a result of membrane bending 101. These effects could together drive the formation of tubules102.
Furthermore, tubule fission has also been proposed to occur by a line energy-driven mechanism103. When dynamin is inhibited in cells, long tubules are created in response to Shiga toxin, which can be released in a cholesterol-dependent manner by conditions that favour domain formation such as cooling or actin polymerization103. The scission is therefore proposed to derive from physical pinching as the newly created domains minimize the unfavourable interaction with the bulk membrane, and not from the ‘pinching’ activity of GTPases.
696 | O CTOBER 2010 | VOLUME 11
www.nature.com/reviews/molcellbio © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS Similarly to Shiga toxin, the SV40 virus binds a glyco sphingolipid, the ganglioside GM1 (REF. 97). In this case, the binding is mediated by the pentameric capsid protein VP1. After binding, the GM1 patch invaginates towards the cytoplasm, usually enclosing one virus into a vesicle destined for the ER. Binding of SV40 can also occur in caveolae, but these mostly remain at the cell surface97. In phase-separated GUVs containing GM1, the viral structures bind to the liquid-ordered phase104. Unlike native GM1 molecules containing long acyl chains that support line tension-driven tubulation, synthetic GM1 molecular species with short hydrocarbon chains partition to the liquid-disordered phase and impair internalization and infection when introduced into host cells lacking the native version of this glycosphingolipid. The saturation level of the hydrocarbon chains affects GM1 partitioning in a similar manner. Finally, the spacing of multivalent binding by toxins and the virus seems to be important as antibody clustering fails to induce internalization104. Studies of GPI-anchored proteins using clathrinindependent internalization pathways have shown that proteins with artificial, unsaturated lipid anchors105 can be endocytosed together with GPI-anchored proteins that have normal, saturated anchors, although it is unclear how. The linkers could form hydrogen bonds with more ‘raftophilic’ molecules, such as native GPI-anchored proteins, and the protein ectodomains could help drive internalization by curvature-mediated interactions101. An alternative explanation is that GPI-anchored proteins use a simple, nonspecific, default (or bulk) process for the uptake of lipid-bound proteins present in the exoplasmic (lumenal) leaflet of the plasma membrane105. The emerging principles from these studies are that raft clustering can lead to formation of membrane domains that can bud off to form raft carriers that distribute specific sets of lipids and proteins to different post-Golgi destinations in the secretory and endocytic pathways. These mechanisms differ from those used by the well-known coat-mediated transport carriers, but the machinery governing their generation is poorly understood.
Moving forwards with rafts Although many cell functions take place in membranes where proteins and lipids are intimately mixing, research in this field has long avoided the study of how lipids and proteins function together. This neglect is now being remedied by the realization that membranes are made functional by lipid–lipid, lipid–protein and protein– protein interactions. For example, it has been recently shown that proteins adjust their transmembrane length and composition to the specific physical properties of the different subcellular membranes in which they reside106, which themselves have evolved highly specific lipid compositions1. Cell membranes are lipid bilayers, crowded with proteins occupying around 20% of the bilayer area107. This means that lipids and proteins in membranes should really be studied as collectives55. Membrane proteins alter their lipid environment not only by binding specific lipids but also by influencing their surrounding lipid environment 108. Therefore, one area of increasing
importance will be the study of lipid–protein interactions that regulate the nanoscale raft protein assemblies and how they can coalesce to form functional platforms. This work will require sophisticated and difficult image technology, as well as biochemistry that must overcome the technical hurdles that plague work with hydrophobic proteins and lipids. The well-known asymmetry of the bilayer and the proteins spanning the membrane add additional barriers to reconstituting membrane organization and function in vitro. In addition, most saturated lipids that are thought to underlie raft formation reside in the exoplasmic leaflet of the membrane, and the principles of raft organization in the cytosolic leaflet remain unknown109. One issue we have not discussed is the role of cortical actin in regulating membrane organization. Clearly, actin can shape the lateral distribution of membrane components. For example, transient binding of a clustered GPI-anchored protein requires actin and depends on CSK-binding protein (CBP), ERM-binding protein 50 (EBP50), Src-family kinase phosphorylation and cholesterol110. Actin also seems to nucleate raftbased heterogeneity, both at the nanoscale level111,112 and as a scaffold. Here, actin is functional, for example during high-affinity immunoglobin-ε receptor (FcERI) immobilization113 or TCR signalling 114. We suggest that the cell, not aware of our present distinction between biophysics, biochemistry and structural mechanics, couples the specificity of peripheral decorating agents, such as actin, to those of cholesterol and sphingolipids, with the goal of making heterogeneity functional in the lateral dimension115. Realizing the power of this synergy will likewise require the coming together of these fields in membrane research. The emergence of lipidomics methodology is an important advance. Until now, few studies have existed in which the diversity of membrane lipid composition has been analysed. We need to go beyond lipid classes and also analyse molecular diversity. With the new mass spectrometers and with an increasing availability of lipid standards, quantitative lipidomics is becoming possible that could characterize for the first time the molecular lipid composition of different organelle membranes. This possibility demands improved methodology for fractionating and purifying subcellular membrane compartments. However, the superior sensitivity of the mass spectrometric methods will allow analysis of minute samples. The analysis of membrane protein–lipid complexes will also become routine. Elucidation of the fine-tuned composition of membranes by lipidomics and proteomics is bound to open up new perspectives for basic biology and biomedical research. During the past ten years, this field has been energized by the introduction of new and sophisticated method ology that is providing unprecedented resolution in space and time. The combination of methods and ideas will continue to challenge our views on how membranes are organized. A decade from now, we will have more accurate insights into how cells use their vast lipid and protein variability to construct functional membrane collectives.
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
VOLUME 11 | O CTOBER 2010 | 697 © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS 1.
2. 3. 4. 5. 6. 7. 8.
9. 10.
11.
12.
13. 14. 15.
16.
17. 18. 19.
20.
21.
22. 23.
24.
25.
26. 27.
van Meer, G., Voelker, D. R. & Feigenson, G. W. Membrane lipids: where they are and how they behave. Nature Rev. Mol. Cell Biol. 9, 112–124 (2008). Singer, S. J. & Nicolson, G. L. The fluid mosaic model of the structure of cell membranes. Science 175, 720–731 (1972). Simons, K. & Ikonen, E. Functional rafts in cell membranes. Nature 387, 569–572 (1997). Simons, K. & Toomre, D. Lipid rafts and signal transduction. Nature Rev. Mol. Cell Biol. 1, 31–39 (2000). Parton, R. G. & Simons, K. The multiple faces of caveolae. Nature Rev. Mol. Cell Biol. 8, 185–194 (2007). Munro, S. Lipid rafts: elusive or illusive? Cell 115, 377–388 (2003). Shaw, A. S. Lipid rafts: now you see them, now you don’t. Nature Immunol. 7, 1139–1142 (2006). Lichtenberg, D., Goñi, F. M. & Heerklotz, H. Detergent-resistant membranes should not be identified with membrane rafts. Trends Biochem. Sci. 30, 430–436 (2005). Lingwood, D. & Simons, K. Detergent resistance as a tool in membrane research. Nature Protoc. 2, 2159–2165 (2007). Kenworthy, A. K. Have we become overly reliant on lipid rafts? Talking point on the involvement of lipid rafts in T‑cell activation. EMBO Rep. 9, 531–535 (2008). Kalvodova, L. et al. The lipidomes of vesicular stomatitis virus, semliki forest virus, and the host plasma membrane analyzed by quantitative shotgun mass spectrometry. J. Virol. 83, 7996–8003 (2009). Zidovetzki, R. & Levitan, I. Use of cyclodextrins to manipulate plasma membrane cholesterol content: evidence, misconceptions and control strategies. Biochim. Biophys. Acta 1768, 1311–1324 (2007). Pizzo, P. et al. Lipid rafts and T cell receptor signaling: a critical re-evaluation. Eur. J. Immunol. 32, 3082–3091 (2002). Kenworthy, A. K. et al. Dynamics of putative raftassociated proteins at the cell surface. J. Cell Biol. 165, 735–746 (2004). Glebov, O. O. & Nichols, B. J. Lipid raft proteins have a random distribution during localized activation of the T‑cell receptor. Nature Cell Biol. 6, 238–243 (2004). He, H. & Marguet, D. T‑cell antigen receptor triggering and lipid rafts: a matter of space and time scales. Talking point on the involvement of lipid rafts in T‑cell activation. EMBO Rep. 9, 525–530 (2008). Jacobson, K., Mouritsen, O. G. & Anderson, R. G. W. Lipid rafts: at a crossroad between cell biology and physics. Nature Cell Biol. 9, 7–14 (2007). Day, C. A. & Kenworthy, A. K. Tracking microdomain dynamics in cell membranes. Biochim. Biophys. Acta 1788, 245–253 (2009). Lagerholm, B. C., Weinreb, G. E., Jacobson, K. & Thompson, N. L. Detecting microdomains in intact cell membranes. Annu. Rev. Phys. Chem. 56, 309–336 (2005). Meyer, B. H. et al. FRET imaging reveals that functional neurokinin‑1 receptors are monomeric and reside in membrane microdomains of live cells. Proc. Natl Acad. Sci. USA 103, 2138–2143 (2006). Sharma, P. et al. Nanoscale organization of multiple GPI-anchored proteins in living cell membranes. Cell 116, 577–589 (2004). The first demonstration of cholesterol-assisted nanoscale clusters in living cells. Vyas, N. et al. Nanoscale organization of hedgehog is essential for long-range signaling. Cell 133, 1214–1227 (2008). Kusumi, A., Koyama-Honda, I. & Suzuki, K. Molecular dynamics and interactions for creation of stimulationinduced stabilized rafts from small unstable steadystate rafts. Traffic 5, 213–230 (2004). Pinaud, F. et al. Dynamic partitioning of a glycosyl‑phosphatidylinositol‑anchored protein in glycosphingolipid-rich microdomains imaged by singlequantum dot tracking. Traffic 10, 691–712 (2009). Lenne, P. et al. Dynamic molecular confinement in the plasma membrane by microdomains and the cytoskeleton meshwork. EMBO J. 25, 3245–3256 (2006). Schermelleh, L., Heintzmann, R. & Leonhardt, H. A guide to super-resolution fluorescence microscopy. J. Cell Biol. 190, 165–175. Hell, S. W. Far-field optical nanoscopy. Science 316, 1153–1158 (2007).
28. Shroff, H., Galbraith, C. G., Galbraith, J. A. & Betzig, E. Live-cell photoactivated localization microscopy of nanoscale adhesion dynamics. Nature Meth. 5, 417–423 (2008). 29. Rust, M. J., Bates, M. & Zhuang, X. Sub‑diffraction‑limit imaging by stochastic optical reconstruction microscopy (STORM). Nature Meth. 3, 793–795 (2006). 30. Eggeling, C. et al. Direct observation of the nanoscale dynamics of membrane lipids in a living cell. Nature 457, 1159–1162 (2009). A super-resolution microscopy study showing that sphingolipids and GPI-anchored proteins are transiently trapped in cholesterol-dependent molecular complexes in live cells. 31. van Zanten, T. S., Cambi, A. & Garcia-Parajo, M. F. A nanometer scale optical view on the compartmentalization of cell membranes. Biochim. Biophys. Acta 1798, 777–787 (2010). 32. Zhong, L. et al. NSOM/QD-based direct visualization of CD3‑induced and CD28‑enhanced nanospatial coclustering of TCR and coreceptor in nanodomains in T cell activation. PLoS ONE 4, e5945 (2009). 33. Lasserre, R. et al. Raft nanodomains contribute to Akt/PKB plasma membrane recruitment and activation. Nature Chem. Biol. 4, 538–547 (2008). 34. Wawrezinieck, L., Rigneault, H., Marguet, D. & Lenne, P. F. Fluorescence correlation spectroscopy diffusion laws to probe the submicron cell membrane organization. Biophys. J. 89, 4029–4042 (2005). 35. Wenger, J. et al. Diffusion analysis within single nanometric apertures reveals the ultrafine cell membrane organization. Biophys. J. 92, 913–919 (2007). 36. Sahl, S. J., Leutenegger, M., Hilbert, M., Hell, S. W. & Eggeling, C. Fast molecular tracking maps nanoscale dynamics of plasma membrane lipids. Proc. Natl Acad. Sci. USA 107, 6829–6834 (2010). 37. Shevchenko, A. & Simons, K. Lipidomics: coming to grips with lipid diversity. Nature Rev. Mol. Cell Biol. 11, 593–598 (2010). 38. Dennis, E. A. Lipidomics joins the omics evolution. Proc. Natl Acad. Sci. USA 106, 2089–2090 (2009). 39. Zech, T. et al. Accumulation of raft lipids in T‑cell plasma membrane domains engaged in TCR signalling. EMBO J. 28, 466–476 (2009). The first lipidomic analysis of raft clusters in activated TCR domains. 40. Klemm, R. W. et al. Segregation of sphingolipids and sterols during formation of secretory vesicles at the trans-Golgi network. J. Cell Biol. 185, 601–612 (2009). The first direct experimental support for a post-Golgi raft pathway that transports proteins towards the plasma membrane. 41. Brugger, B. et al. The HIV lipidome: a raft with an unusual composition. Proc. Natl Acad. Sci. USA 103, 2641–2646 (2006). 42. Feigenson, G. W. Phase behavior of lipid mixtures. Nature Chem. Biol. 2, 560–563 (2006). 43. Jorgensen, K. & Mouritsen, O. G. Phase separation dynamics and lateral organization of two-component lipid membranes. Biophys. J. 69, 942–954 (1995). 44. Kahya, N. & Schwille, P. Fluorescence correlation studies of lipid domains in model membranes. Mol. Membr. Biol. 23, 29–39 (2006). 45. Johnston, L. J. Nanoscale imaging of domains in supported lipid membranes. Langmuir 23, 5886–5895 (2007). 46. Elson, E. L., Fried, E., Dolbow, J. E. & Genin, G. M. Phase separation in biological membranes: integration of theory and experiment. Annu. Rev. Biophys. 39, 207–226 (2010). 47. Dietrich, C. et al. Lipid rafts reconstituted in model membranes. Biophys. J. 80, 1417–1428 (2001). 48. Hammond, A. T. et al. Crosslinking a lipid raft component triggers liquid ordered-liquid disordered phase separation in model plasma membranes. Proc. Natl Acad. Sci. USA 102, 6320–6325 (2005). 49. Collins, M. D. & Keller, S. L. Tuning lipid mixtures to induce or suppress domain formation across leaflets of unsupported asymmetric bilayers. Proc. Natl Acad. Sci. USA 105, 124–128 (2008). 50. Simons, K. & Vaz, W. L. Model systems, lipid rafts, and cell membranes. Annu. Rev. Biophys. Biomol. Struct. 33, 269–295 (2004). 51. Baumgart, T. et al. Large-scale fluid/fluid phase separation of proteins and lipids in giant plasma membrane vesicles. Proc. Natl Acad. Sci. USA 104, 3165–3170 (2007).
698 | O CTOBER 2010 | VOLUME 11
52. 53.
54. 55. 56. 57. 58.
59. 60.
61.
62.
63.
64. 65. 66. 67. 68.
69.
70.
71. 72.
73. 74.
Shows the presence of liquid-ordered-like and liquid-disordered-like phases in plasma membrane vesicles containing native lipids and proteins. Levental, I. et al. Cholesterol-dependent phase separation in cell-derived giant plasma-membrane vesicles. Biochem. J. 424, 163–167 (2009). Lingwood, D., Ries, J., Schwille, P. & Simons, K. Plasma membranes are poised for activation of raft phase coalescence at physiological temperature. Proc. Natl Acad. Sci. USA 105, 10005–10010 (2008). Shows that inflated plasma membranes can separate into large-scale domains at 37ºC following cholera toxin cross-linking. Kaiser, H. J. et al. Order of lipid phases in model and plasma membranes. Proc. Natl Acad. Sci. USA 106, 16645–16650 (2009). Lingwood, D. & Simons, K. Lipid rafts as a membrane-organizing principle. Science 327, 46–50 (2010). Kholodenko, B. N., Hancock, J. F. & Kolch, W. Signalling ballet in space and time. Nature Rev. Mol. Cell Biol. 11, 414–426 (2010). Hancock, J. F. Lipid rafts: contentious only from simplistic standpoints. Nature Rev. Mol. Cell Biol. 7, 456–462 (2006). Stefanová, I., Horejsí, V., Ansotegui, I. J., Knapp, W. & Stockinger, H. GPI-anchored cell-surface molecules complexed to protein tyrosine kinases. Science 254, 1016–1019 (1991). Seminario, M. C. & Bunnell, S. C. Signal initiation in T‑cell receptor microclusters. Immunol. Rev. 221, 90–106 (2008). Campi, G., Varma, R. & Dustin, M. L. Actin and agonist MHC-peptide complex-dependent T cell receptor microclusters as scaffolds for signaling. J. Exp. Med. 202, 1031–1036 (2005). Kaizuka, Y., Douglass, A. D., Varma, R., Dustin, M. L. & Vale, R. D. Mechanisms for segregating T cell receptor and adhesion molecules during immunological synapse formation in Jurkat T cells. Proc. Natl Acad. Sci. USA 104, 20296–20301 (2007). Kaizuka, Y., Douglass, A. D., Vardhana, S., Dustin, M. L. & Vale, R. D. The coreceptor CD2 uses plasma membrane microdomains to transduce signals in T cells. J. Cell Biol. 185, 521–534 (2009). Douglass, A. D. & Vale, R. D. Single-molecule microscopy reveals plasma membrane microdomains created by protein-protein networks that exclude or trap signaling molecules in T cells. Cell 121, 937–950 (2005). Davis, M. et al. T cells as a self-referential, sensory organ. Annu. Rev. Immunol. 25, 681–695 (2007). Huang, J. et al. The kinetics of two-dimensional TCR and pMHC interactions determine T‑cell responsiveness. Nature 464, 932–936 (2010). Huppa, J. et al. TCR‑peptide‑MHC interactions in situ show accelerated kinetics and increased affinity. Nature 463, 963–967 (2010). Dunne, P. D. et al. DySCo: quantitating associations of membrane proteins using two-color single-molecule tracking. Biophys. J. 97, L5–L7 (2009). James, J. R. et al. Single-molecule level analysis of the subunit composition of the T cell receptor on live T cells. Proc. Natl Acad. Sci. USA 104, 17662–17667 (2007). Lillemeier, B., Pfeiffer, J. R., Surviladze, Z., Wilson, B. S. & Davis, M. Plasma membrane-associated proteins are clustered into islands attached to the cytoskeleton. Proc. Natl Acad. Sci. USA 103, 18992–18997 (2006). Lillemeier, B. et al. TCR and Lat are expressed on separate protein islands on T cell membranes and concatenate during activation. Nature Immunol. 11, 90–96 (2010). van der Merwe, P. A., Dunne, P. D., Klenerman, D. & Davis, S. J. Taking T cells beyond the diffraction limit. Nature Immunol. 11, 51–52 (2010). Pralle, A., Keller, P., Florin, E. L., Simons, K. & Horber, J. K. Sphingolipid-cholesterol rafts diffuse as small entities in the plasma membrane of mammalian cells. J. Cell Biol. 148, 997–1008 (2000). Nika, K. et al. Constitutively active Lck kinase in T cells drives antigen receptor signal transduction. Immunity 32, 766–777 (2010). Anderson, R. G. & Jacobson, K. A role for lipid shells in targeting proteins to caveolae, rafts, and other lipid domains. Science 296, 1821–1825 (2002).
www.nature.com/reviews/molcellbio © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS 75. Suzuki, K. G. et al. GPI-anchored receptor clusters transiently recruit Lyn and G α for temporary cluster immobilization and Lyn activation: single-molecule tracking study 1. J. Cell Biol. 177, 717–730 (2007). This paper introduced the STALL concept of GPI-anchored protein signalling, which shows that clusters undergoing STALL generate short-lived, digital-like signalling bursts. 76. Waheed, A. A. & Freed, E. O. Lipids and membrane microdomains in HIV‑1 replication. Virus Res. 143, 162–176 (2009). 77. Scheiffele, P., Rietveld, A., Wilk, T. & Simons, K. Influenza viruses select ordered lipid domains during budding from the plasma membrane. J. Biol. Chem. 274, 2038–2044 (1999). 78. Grassme, H., Riethmuller, J. & Gulbins, E. Biological aspects of ceramide-enriched membrane domains. Prog. Lipid Res. 46, 161–170 (2007). 79. Chan, R. et al. Retroviruses human immunodeficiency virus and murine leukemia virus are enriched in phosphoinositides. J. Virol. 82, 11228–11238 (2008). 80. Lorizate, M. et al. Probing HIV‑1 membrane liquid order by Laurdan staining reveals producer celldependent differences. J. Biol. Chem. 284, 22238–22247 (2009). 81. McLaughlin, S., Wang, J., Gambhir, A. & Murray, D. PIP2 and proteins: interactions, organization, and information flow. Annu. Rev. Biophys. Biomol. Struct. 31, 151–175 (2002). 82. Tong, J. et al. Role of GAP‑43 in sequestering phosphatidylinositol 4,5‑bisphosphate to raft bilayers. Biophys. J. 94, 125–133 (2008). 83. Saad, J. S. et al. Structural basis for targeting HIV‑1 Gag proteins to the plasma membrane for virus assembly. Proc. Natl Acad. Sci. USA 103, 11364–11369 (2006). 84. Wassall, S. R. & Stillwell, W. Polyunsaturated fatty acid-cholesterol interactions: domain formation in membranes. Biochim. Biophys. Acta 1788, 24–32 (2009). 85. Castillon, G. A., Watanabe, R., Taylor, M., Schwabe, T. M. & Riezman, H. Concentration of GPIanchored proteins upon ER exit in yeast. Traffic 10, 186–200 (2009). 86. Fujita, M., Umemura, M., Yoko, O. T. & Jigami, Y. PER1 Is required for GPI-phospholipase A2 activity and involved in lipid remodeling of GPI-anchored proteins. Mol. Biol. Cell 17, 5253–5264 (2006). 87. Fujita, M., Yoko, O. T. & Jigami, Y. Inositol deacylation by Bst1p is required for the quality control of glycosylphosphatidylinositol-anchored proteins. Mol. Biol. Cell 17, 834–850 (2006). 88. Kajiwara, K. et al. Yeast ARV1 is required for efficient delivery of an early GPI intermediate to the first mannosyltransferase during GPI assembly and controls lipid flow from the endoplasmic reticulum. Mol. Biol. Cell 19, 2069–2082 (2008). 89. Barz, W. P. & Walter, P. Two endoplasmic reticulum (ER) membrane proteins that facilitate ER‑to‑Golgi transport of glycosylphosphatidylinositol-anchored proteins. Mol. Biol. Cell 10, 1043–1059 (1999). 90. Watanabe, R., Funato, K., Venkataraman, K., Futerman, A. H. & Riezman, H. Sphingolipids are required for the stable membrane association of glycosylphosphatidylinositol-anchored proteins in yeast. J. Biol. Chem. 277, 49538–49544 (2002). 91. Simons, K. & van Meer, G. Lipid sorting in epithelial cells. Biochemistry 27, 6197–6202 (1988). 92. Harsay, E. & Bretscher, A. Parallel secretory pathways to the cell surface in yeast. J. Cell Biol. 131, 297–310 (1995).
93. Proszynski, T. J. et al. A genome-wide visual screen reveals a role for sphingolipids and ergosterol in cell surface delivery in yeast. Proc. Natl Acad. Sci. USA 102, 17981–17986 (2005). 94. Ejsing, C. et al. Global analysis of the yeast lipidome by quantitative shotgun mass spectrometry. Proc. Natl Acad. Sci. USA 106, 2136–2141 (2009). 95. Schuck, S. & Simons, K. Polarized sorting in epithelial cells: raft clustering and the biogenesis of the apical membrane. J. Cell Sci. 117, 5955–5964 (2004). 96. Klose, C. et al. Yeast lipids can phase separate into micrometer-scale membrane domains. J. Biol. Chem. 20 Jul 2010 (doi:10.1074/jbc.M110.123554) . 97. Mercer, J., Schelhaas, M. & Helenius, A. Virus entry by endocytosis. Ann. Rev. Biochem. 79, 803–833 (2010). 98. Sandvig, K. et al. Pathways followed by protein toxins into cells. Int. J. Med. Microbiol. 293, 483–490 (2004). 99. Ling, H. et al. Structure of the shiga-like toxin I B‑pentamer complexed with an analogue of its receptor Gb3. Biochemistry 37, 1777–1788 (1998). 100. Romer, W. et al. Shiga toxin induces tubular membrane invaginations for its uptake into cells. Nature 450, 670–675 (2007). Shows that Shiga toxin can induce a lipid reorganization in plasma membranes that forms energy-independent endocytic tubules in cells. 101. Reynwar, B. J. et al. Aggregation and vesiculation of membrane proteins by curvature-mediated interactions. Nature 447, 461–464 (2007). 102. Windschiegl, B. et al. Lipid reorganization induced by Shiga toxin clustering on planar membranes. PLoS ONE 4, e6238 (2009). 103. Romer, W. et al. Actin dynamics drive membrane reorganization and scission in clathrin-independent endocytosis. Cell 140, 540–553 (2010). 104. Ewers, H. et al. GM1 structure determines SV40‑induced membrane invagination and infection. Nature Cell Biol. 12, 11–18 (2010). 105. Bhagatji, P., Leventis, R., Comeau, J., Refaei, M. & Silvius, J. R. Steric and not structure-specific factors dictate the endocytic mechanism of glycosylphosphatidylinositol-anchored proteins. J. Cell Biol. 186, 615–628 (2009). 106. Sharpe, H. J., Stevens, T. J. & Munro, S. A comprehensive comparison of transmembrane domains reveals organelle-specific properties. Cell 142, 158–169. 107. Dupuy, A. D. & Engelman, D. M. Protein area occupancy at the center of the red blood cell membrane. Proc. Natl Acad. Sci. USA 105, 2848–2852 (2008). 108. Niemelä, P. S. et al. Membrane proteins diffuse as dynamic complexes with lipids. J. Am. Chem. Soc. 132, 7574–7575 (2010). 109. Devaux, P. F. & Morris, R. Transmembrane asymmetry and lateral domains in biological membranes. Traffic 5, 241–246 (2004). 110. Chen, Y., Veracini, L., Benistant, C. & Jacobson, K. The transmembrane protein CBP plays a role in transiently anchoring small clusters of Thy‑1, a GPIanchored protein, to the cytoskeleton. J. Cell Sci. 122, 3966–3972 (2009). 111. Suzuki, K. G., Fujiwara, T. K., Edidin, M. & Kusumi, A. Dynamic recruitment of phospholipase C γ at transiently immobilized GPI-anchored receptor clusters induces IP3‑Ca2+ signaling: single-molecule tracking study 2. J. Cell Biol. 177, 731–742 (2007). 112. Goswami, D. et al. Nanoclusters of GPI-anchored proteins are formed by cortical actin-driven activity. Cell 135, 1085–1097 (2008).
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
113. Andrews, N. L. et al. Actin restricts FcεRI diffusion and facilitates antigen-induced receptor immobilization. Nature Cell Biol. 10, 955–963 (2008). 114. Viola, A. & Gupta, N. Tether and trap: regulation of membrane-raft dynamics by actin-binding proteins. Nature Rev. Immunol. 7, 889–896 (2007). 115. Gupta, N. et al. Quantitative proteomic analysis of B cell lipid rafts reveals that ezrin regulates antigen receptor-mediated lipid raft dynamics. Nature Immunol. 7, 625–633 (2006). 116. Meder, D., Moreno, M. J., Verkade, P., Vaz, W. L. & Simons, K. Phase coexistence and connectivity in the apical membrane of polarized epithelial cells. Proc. Natl Acad. Sci. USA 103, 329–334 (2006). 117. Danielsen, E. M. & Hansen, G. H. Lipid rafts in epithelial brush borders: atypical membrane microdomains with specialized functions. Biochim. Biophys. Acta 1617, 1–9 (2003). 118. Simons, M. & Trotter, J. Wrapping it up: the cell biology of myelination. Curr. Opin. Neurobiol. 17, 533–540 (2007). 119. Veatch, S. L., Soubias, O., Keller, S. L. & Gawrisch, K. Critical fluctuations in domain-forming lipid mixtures. Proc. Natl Acad. Sci. USA 104, 17650–17655 (2007). 120. Honerkamp-Smith, A. R., Veatch, S. L. & Keller, S. L. An introduction to critical points for biophysicists; observations of compositional heterogeneity in lipid membranes. Biochim. Biophys. Acta 1788, 53–63 (2009). 121. Veatch, S. L. et al. Critical fluctuations in plasma membrane vesicles. ACS Chem. Biol. 3, 287–293 (2008). The first observation of critical behaviour in plasma membranes, suggesting that nanoscale domains form under physiological conditions. 122. Johnson, S. A. et al. Temperature-dependent phase behavior and protein partitioning in giant plasma membrane vesicles. Biochim. Biophys. Acta 1798, 1427–1435 (2010). 123. Ejsing, C. S. et al. Automated identification and quantification of glycerophospholipid molecular species by multiple precursor ion scanning. Anal. Chem. 78, 6202–6214 (2006). 124. Ivanchenko, S. et al. Dynamics of HIV‑1 assembly and release. PLoS Pathog. 5, e1000652 (2009). 125. Nguyen, D. H. & Hildreth, J. E. Evidence for budding of human immunodeficiency virus type 1 selectively from glycolipid-enriched membrane lipid rafts. J. Virol. 74, 3264–3272 (2000).
Acknowledgements
We thank H. He for critically reading the manuscript and the members of the K.S. laboratory for input, especially D. Lingwood and I. Levental. Work in the K.S. laboratory was supported by the EUFP6 PRISM grant LSHB‑CT2007–037,740, DFG Schwerpunktprogramm1175, the BMBF BioChance Plus grant 0313,827 and the BMBF ForMaT grant 03FO1212.
Competing interests statement
The authors declare competing financial interests: see web version for details.
FURTHER INFORMATION Kai Simons’s homepage: http://www.mpi-cbg.de/research/ research-groups/kai-simons.html ALL LINKS ARE ACTIVE IN THE ONLINE PDF
VOLUME 11 | O CTOBER 2010 | 699 © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS
Molecular mechanisms of necroptosis: an ordered cellular explosion Peter Vandenabeele*, Lorenzo Galluzzi‡§, Tom Vanden Berghe* and Guido Kroemer‡||
Abstract | For a long time, apoptosis was considered the sole form of programmed cell death during development, homeostasis and disease, whereas necrosis was regarded as an unregulated and uncontrollable process. Evidence now reveals that necrosis can also occur in a regulated manner. The initiation of programmed necrosis, ‘necroptosis’, by death receptors (such as tumour necrosis factor receptor 1) requires the kinase activity of receptorinteracting protein 1 (RIP1; also known as RIPK1) and RIP3 (also known as RIPK3), and its execution involves the active disintegration of mitochondrial, lysosomal and plasma membranes. Necroptosis participates in the pathogenesis of diseases, including ischaemic injury, neurodegeneration and viral infection, thereby representing an attractive target for the avoidance of unwarranted cell death.
*Molecular Signalling and Cell Death Unit, Department for Molecular Biomedical Research, VIB, and Department of Biomedical Molecular Biology, Ghent University, B‑9052 Ghent, Belgium. ‡ INSERM, U848, F‑94805 Villejuif, France. § Institut Gustave Roussy, and Université Paris‑Sud XI, F‑94805 Villejuif, France. || Metabolomics Platform, Institut Gustave Roussy, F‑94805 Villejuif, France; Centre de Recherche des Cordoliers, F‑75,005 Paris, France; Pôle de Biologie, Hôpital Européen Georges Pompidou, AP‑HP, F‑75908 Paris, France; and Université Paris Descartes V, F‑75270 Paris, France. Correspondence to G.K. and P.V. e-mails:
[email protected]; peter.vandenabeele@dmbr. vib-ugent.be doi:10.1038/nrm2970 Published online 8 September 2010
Biased by their focus on life, biologists have neglected cell death for a long time. Although the first morphological descriptions of cellular demise date back to the mid-nineteenth century, the notion of ‘programmed cell death’ was formulated by Lockshin as late as 1964 (REF. 1) . In the early 1970s, Kerr, Wyllie and Currie discovered a peculiar type of mammalian cell death that they dubbed ‘apoptosis’ (REF. 2). The stereotyped features of apoptosis (BOX 1) suggested that it would constitute a regulated cell death process, a notion that was elegantly shown in Caenorhabiditis elegans by the Horvitz laboratory in 1980–1990 (reviewed in REF. 3). Textbooks soon thought of apoptosis and necrosis as opposed mechanisms (BOX 1), necrosis being considered as a purely accidental and passive cell death subroutine. The first morphological classification of cell death was proposed by Schweichel and Merker 4 who described, in rat embryos exposed to toxicants, type I cell death associated with heterophagy, type II cell death associated with autophagy and type III cell death without digestion. Today, these cell death modes are referred to as apoptosis, autophagic cell death and necrosis, respectively 5. The purely unregulated nature of necrosis was questioned in 1988, when it was discovered that distinct cell types died in response to the same trigger, tumour necrosis factor (TNF), while manifesting either the ‘classical’ features of apoptosis or a ‘balloon-like’ morphology without nuclear disintegration6. Since then, accumulating evidence has paved the way to the concept of ‘programmed necrosis’ (TIMELINE) ,
culminating in the introduction, in 2005, of the neo logism necroptosis to describe one instance of regulated (as opposed to accidental) necrotic cell death7. Over the past two decades, a plethora of molecules and pro cesses have been characterized as initiators, modulators or effectors of necroptosis. These include (but are not limited to): receptor-interacting protein 1 (RIP1; also known as RIPK1), RIP3 (also known as RIPK3)8–12, caspase inhibitors13, ubiquitin E3 ligases, deubiquitylating enzymes 11,14, reactive oxygen species (ROS) generated by mitochondria or NADPH oxidase 1 (NOX1) 15–17, bioenergetic reactions such as glyco genolysis12 and glutaminolysis12,18, pro-apoptotic B cell lymphoma 2 (BCL‑2) family members14, poly(ADP– ribose) polymerase (PARP)19, the mitochondrial perme‑ ability transition pore complex (PTPC)20–22, lysosomal membrane permeabilization (LMP)23,24 and lysosomal, mitochondrial and cytosolic hydrolases23–25. Altogether, it seems that multiple signal transducers and metabolic processes can ignite or mediate cellular demolition by necrosis (Supplementary information S1 (table)), thereby constituting targets for the therapeutic suppression of necroptosis, a possibility that has raised huge expectations26. As the underlying molecular mechanisms have only recently begun to emerge, a comprehensive review on necroptosis is timely and may shed new light on research areas that, until now, have been dominated by apoptosis. Here, we provide a detailed description of the molecular mechanisms of necroptosis and briefly discuss its immunological outcomes and pathophysiological implications.
700 | O CTOBER 2010 | VOLUME 11
www.nature.com/reviews/molcellbio © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS
Heterophagy A term of Greek origin indicating the cellular digestion of an exogenous substance, cell or subcellular particle that has been taken up from the extracellular microenvironment.
Autophagy A pathway for the recycling of cellular contents, in which materials inside the cell are packaged into vesicles and are then targeted to the vacuole or lysosome for bulk turnover. Autophagy is thought to be prominently cytoprotective.
Caspase A Cys protease that cleaves its substrate after an Asp residue. Caspases play a crucial part in both the initiation (caspase 2, caspase 8, caspase 9 and caspase 10) and execution (caspase 3, caspase 6 and caspase 7) of apoptosis, and they are also required for many processes that are unrelated to cell death, such as the differentiation of several cell types161.
Glutaminolysis The bioenergetic pathway by which Glu or Gln is converted to α‑ketoglutarate, an intermediate of the Krebs cycle. Thus, glutaminolysis can provide substrates for ATP generation by oxidative phosphorylation or stimulate the generation of pyruvate through malate decarboxylation.
Mitochondrial permeability transition Long-lasting openings of the PTPC lead to an abrupt increase in the inner mitochondrial membrane’s permeability to ions and low-molecular-mass solutes, thus provoking osmotic swelling of the mitochondrial matrix and rupture of the mitochondrial outer membrane.
Apoptotic body A membrane-surrounded vesicle that is shed from dying cells during the late stages of apoptosis and that may include portions of the nucleus and/or apparently normal organelles.
Initiation of necroptosis: the receptors A sizeable fraction of cells dying in vivo in response to ischaemia–reperfusion, physical or chemical trauma, viral or bacterial infection, or neurodegenerative processes exhibit morphological features of necrosis27 (BOX 1). Although necrosis was initially believed to be triggered by excessively harsh microenvironmental conditions, killing cells in an uncontrollable manner, it turned out that the molecular mechanisms of pathological cell loss (in particular ischaemia–reperfusion-induced necrosis) partially overlap with the biochemical cascades that mediate necroptosis (reviewed in REF. 28). Death receptors in the initiation of necroptosis. Necroptosis can be induced by the ligation of death receptors, including CD95 (also known as FAS; which binds the ligand CD95L (also known as FASL))29, TNF receptor 1 (TNFR1), TNFR2 (REFS 6,13,30) , TNF-related apoptosis-inducing ligand receptor 1 (TRAILR1) and TRAILR2 (REF. 9.) These receptors usually activate the apoptotic machinery, and their cytotoxicity often requires the presence of transcriptional or translational inhibitors, suggesting the existence of short-lived cytoprotective proteins that are continuously being synthesized (reviewed in REFS 31,32). Nevertheless, in some cell lines and primary cells, the
presence of caspase inhibitors (which block apoptosis) unveils a caspase-independent cell death pathway that emanates from death receptors and culminates in a necrotic morphology 33. PRRs in the initiation of necroptosis. Although the underlying molecular mechanisms remain elusive, it seems that necroptosis can also be initiated by members of the pathogen recognition receptor (PRR) family, which include plasma membrane or endosome membrane-associated Toll-like receptors, cytosolic NOD-like receptors and retinoic acid-inducible gene Ilike receptors. All of these are expressed by cells of the innate immune system to sense pathogen-associated molecular patterns (PAMPs), such as viral or bacterial nucleotides, lipoproteins, lipopolysaccharide or peptido glycans, and respond by triggering inflammation or cell death34. Various PAMPs have been shown to induce necroptosis by activating PRRs in different cell types. For example, viral dsRNA induces necroptotic cell death in human Jurkat T lymphocytes and murine fibrosarcoma L929 cells 35, and lipopolysaccharide does so in macrophages when caspase 8 activity is inhibited36. Similarly, the Gram-negative bacterium Shigella flexneri triggers necroptosis in neutrophils37 and in
Box 1 | Morphological aspects of necrosis versus apoptosis In 1972, Kerr and colleagues introduced the term ‘apoptosis’ (a Greek word describing falling leaves) to indicate a type of cell death that is morphologically distinct from necrosis2. For more than three decades, apoptosis was considered the sole mechanism of developmental and homeostatic cell death, as well as the only outcome of the activation of a specific class of proteases, caspases161. Now, multiple types of cell death have been classified according to morphological, biochemical or functional aspects, generating a rather diversified nomenclature5. Although biochemical definitions are expected to gradually replace the current vocabulary, the terms apoptosis and necrosis are firmly established in scientific literature5. Apoptosis exhibits peculiar morphological traits, including pseudopod retraction, the rounding up of cells, decreased cellular volume (pyknosis), chromatin condensation and nuclear fragmentation (karyorrhexis), blebbing of the intact plasma membrane, shedding of vacuoles containing cytoplasmic portions and apparently unchanged organelles (known as apoptotic bodies), and the in vivo uptake of apoptotic corpses by neighbouring cells or professional phagocytes (see the figure, part a). When phagocytosis is inefficient, apoptotic bodies progressively lose integrity and their content spills into the extracellular milieu (secondary necrosis). Dying cells were initially catalogued as necrotic in a negative manner; that is, when they failed to display the morphology of apoptotic or autophagic cell death5. However, necrotic cells exhibit some common morphological features, including an increasingly translucent cytoplasm, swelling of organelles, minor ultrastructural modifications of the nucleus (specifically, dilatation of the nuclear membrane and condensation of chromatin into small, irregular, circumscribed patches) and increased cell volume (oncosis), culminating in the disruption of the plasma membrane (see the figure, part b). Necrotic cells do not fragment into discrete corpses as their apoptotic counterparts do. Moreover, their nuclei remain intact and can aggregate and accumulate in necrotic tissues. a b Importantly, although the signalling pathways and/or biochemical events leading to necroptosis, accidental necrosis and secondary necrosis are clearly distinct, these cell death modes are accompanied by similar end-stage degradation and disintegration processes, implying that it is impossible to discriminate among them based on single end-point morphological 5 µm 5 µm assessments24,162,163. Nature Reviews | Molecular Cell Biology
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
VOLUME 11 | O CTOBER 2010 | 701 © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS Timeline | Evolution of the concept of programmed necrosis
Kerr et al. introduce the term apoptosis2.
1972
ROS shown to be involved in TNF-induced cytotoxicity 85.
1988
1992
Discovery that TNF can induce both apoptosis and necrosis6.
Discovery that caspase inhibition favours necrosis13.
1996
1998
Gln metabolism implicated in necrosis18. Discovery that TNFR1 recruits RIP1 (REF. 8).
Discovery that RIP1 mediates TNFR1‑induced caspase-independent cell death9.
1999
Molecular characterization of RIP3 (REF. 171).
2000
Description of a regulated form of necrosis activated by DNA damage19.
2003
2004
Chan et al. introduce the term ‘programmed necrosis’ (REF. 30).
Degterev et al. identify necrostatin 1 and introduce the term ‘necroptosis’7. Characterization of CYPD-deficient mice20,21.
2005
ANT implicated in necroptosis22.
Discovery of the role of CYLD in TNFR1 complex I167,168.
ANT, adenine nucleotide translocase; CYLD, cylindromatosis; RIP, receptor-interacting protein (also known as RIPK); ROS, reactive oxygen species; TNF, tumour necrosis factor; TNFR1, TNF receptor 1.
Inflammasome A supramolecular complex comprising a pattern recognition receptor (such as NLRP3) and an adaptor protein (such as ASC) that is required for the autocatalytic activation of pro-caspase 1. Active caspase 1 catalyses the proteolytic maturation of interleukin‑1β, a potent pro-inflammatory cytokine.
Activation-induced cell death After an adaptive immune response, superfluous lymphocytes are eliminated on T cell receptor re-stimulation by a mechanism that may involve the CD95–CD95L system.
Polyubiquitylation The attachment of chains of the small protein ubiquitin to Lys residues of proteins, often as a tag for rapid cellular degradation.
human monocyte-derived macrophages38. The lethal response of macrophages to S. flexneri depends on the inflammasome component NACHT, LRR and PYD domains-containing protein 3 (NLRP3; also known as NALP3 and cryopyrin) and shares features with the excessive necrosis seen in monocytes of patients affected by autoinflammatory disorders caused by NLRP3 gain-of-function mutations39. Viral infections have repeatedly been reported to promote cell death with necrotic features 30, although this often results from supraphysiologically high viral loads that directly perturb the plasma membrane40. Infection by vaccinia virus, which encodes the caspase inhibitor B13R (also known as Spi2), has been shown to shift to necroptosis the otherwise apoptotic demise of T cells succumbing to activation-induced cell death and of mouse embryonic fibroblasts (MEFs) killed by TNF10. Similarly, whereas the infection of pig kidney cells by strains of cowpox virus expressing cytokine response modifier protein A (CrmA; a potent and specific inhibitor of caspase 8) resulted in cytopathic effects consistent with necrotic death, CrmA-deficient viruses generated an apoptotic cell death phenotype41. These examples underscore the notion that viral infection and PAMP-activated PRRs can facilitate necroptosis. However, our Review will focus on TNFR1‑initiated necroptosis, as this is the most extensively studied model of programmed necrosis to date.
Initiation of necroptosis: TNFR1 decides The most extensively characterized pathway leading to necroptosis is initiated by ligation of TNFR1 (TABLE 1). Depending on the cell type, cell activation state and microenvironment factors, TNF administration can result in cell survival, apoptosis or necroptosis, reflecting an intricate network of signals that operate downstream of TNFR1 and that can ‘switch’ between different patterns of response32 (FIG. 1). In particular, the ubiquitin-editing system and initiator caspases such as caspase 8 modulate the molecular switches that dictate the biological response to TNFR1 activation.
2006
2008
Identification of RIP3 as a crucial modulator of necroptosis10–12.
2009
First systems biology study on necroptosis14 Identification of RIP1 as a specific molecular target of necrostatins63.
TNFR1 complex I promotes cell survival. In the absence of TNF, TNFR1 subunits spontaneously assemble at the plasma membrane to generate trimeric receptors owing to the so-called pre-ligand assembly domain (PLAD), which is localized in the extracellular Cys-rich domain 1 (CRD1) of the protein42. On ligand binding, TNFR1 trimers undergo a conformational change that allows the cytosolic portion of the receptor to recruit multiple proteins, including TNFR-associated death domain (TRADD), RIP1, cellular inhibitor of apoptosis 1 (cIAP1), cIAP2, TNFR-associated factor 2 (TRAF2) and TRAF5. This membrane-proximal supramolecular structure has been named complex I43 (FIG. 1a). cIAPs — E3 ubiquitin ligases that were previously known as apoptosis inhibitors owing to their ability to interfere with caspase activation44 — are recruited (by an amino‑terminal domain that contains baculovirus IAP repeats) to complex I by TRAF2, which stabilizes them by preventing their polyubiquitylation45,46. cIAPs catalyse the addition of Lys63‑linked polyubiquitin moieties to Lys377 of RIP1 (REF. 47). Lys63‑polyubiquitylated RIP1 provides a docking site for transforming growth factor‑β-activated kinase 1 (TAK1), TAK1‑binding protein 2 (TAB2) and TAB3, which together (the TAK1–TAB2–TAB3 complex) constitute the apical stimulator of the canonical nuclear factor-κB (NF‑κB) activation pathway (reviewed in REF. 31; BOX 2). NF-κB transactivates cytoprotective genes and facilitates cell survival. Recent results48 challenge the common notion that RIP1 constitutes an absolute requirement for NF-κB activation 49. In some experimental scenarios, complex I constitutes the molecular platform that recruits the ROS-generating NADPH oxidase NOX1 to the plasma membrane, an event that might be involved in the execution of necroptosis16,17 (see below). Depending on cell type and lethal trigger, complex I might therefore exert either cytoprotective or cytotoxic functions (through NF-κB activation or NOX1 recruitment, respectively), suggesting that complex I regulates an intricate network of pro-survival and pro-death signalling pathways.
702 | O CTOBER 2010 | VOLUME 11
www.nature.com/reviews/molcellbio © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS Table 1 | The functional interactome of TNFR1 in necroptosis Factor*
Localization
Roles in necroptosis
Outcome
Refs
A20 (TNFAIP3)
Cytoplasm and plasma membrane‡
RIP1-deubiquitylating enzyme
Inhibits the NF-κB system to favour necroptosis
Caspase 8
Cytoplasm
TNFR1-interacting protein in complex II
Cleaves and inactivates RIP1 and RIP3
Ceramidase
Mitochondria and plasma membrane
Converts ceramide into sphingosine on TNFR1 ligation
Sphingosine induces lysosomotropic LMP
Cezanne (OTUD7B)
Cytoplasm and plasma membrane‡
RIP1-deubiquitylating enzyme
Inhibits the NF-κB system to favour necroptosis
cIAPs
Cytoplasm and plasma membrane‡
RIP1-ubiquitylating enzymes
Facilitates NF-κB activation and inhibits necroptosis
cPLA2
Cytoplasm
Produces arachidonic acid in response to TNFR1 ligation
Induces lysosomotropic LMP
CYLD
Cytoplasm and plasma membrane‡
RIP1-deubiquitylating enzyme
Inhibits the NF-κB system to favour necroptosis
FADD
Cytoplasm
TNFR1-interacting protein in complex II
Adaptor for TNF-induced necroptosis in some cells
JNK1
Cytoplasm and mitochondria§
Degrades ferritin on TNFR1 ligation
Favours ROS overgeneration downstream of RIP1
97
LOX
Cytoplasm
Converts cPLA2-generated arachidonic acid into lipid hydroperoxides
Induces lysosomotropic LMP
23
NOX1
Plasma membrane‡
NAPDH oxidase that generates O2– in a TRADDand RIP1-dependent manner on TNFR1 ligation
Induces pro-necrotic ROS generation
16
RFK
Cytoplasm and plasma membrane‡
TNFR1-interacting protein in complex I
Couples TNFR1 to NOX1
17
RIP1 (RIPK1)
Cytoplasm, plasma membrane and possibly mitochondria
Crucial component of the necrosome
Triggers necroptosis (which requires RIP1 kinase activity)
9,14,30, 62
RIP3 (RIPK3)
Cytoplasm, mitochondria Crucial component of the necrosome and plasma membrane
Triggers necroptosis (which requires RIP3 kinase activity)
10–12,55
SMases
Lysosomes and plasma membrane
Transform sphingomyelin into ceramide in response to TNF
Trigger ROS generation and lipid peroxidation to induce lysosomotropic LMP
96,112, 115
TNF
Extracellular milieu and plasma membrane‡
Pleiotropic pro-inflammatory cytokine I
Activates necroptosis in the absence of caspase activity
6,13,30
TNFR2
Plasma membrane
Death receptor that potentiates RIP1 recruitment at TNFR1 complex
Triggers necroptosis by facilitating RIP1 activation
30
TRADD
Cytoplasm and plasma membrane‡
TNFR1-interacting protein in complex I and II
Adaptor for TNF-induced necroptosis in some cells
43,58
TRAF2 and TRAF5
Cytoplasm and plasma membrane‡
TNFR1-interacting proteins in complex I
Promotes NF-κB activation, which inhibits necroptosis
30,57
USP21
Cytoplasm and plasma membrane‡
RIP1-deubiquitylating enzyme
Inhibits the NF-κB system to favour necroptosis
52 11,55 112,113 53 11,47 111 14 9,43
54
cIAPs, cellular inhibitor of apoptosis proteins; cPLA2, cytosolic phospholipase A2; CYLD, cylindromatosis; FADD, FAS-associated protein with a death domain; JNK1, JUN N‑terminal kinase 1; LMP, lysosomal membrane permeabilization; LOX, lipoxygenase; NF-κB, nuclear factor κB; NOX1, NADPH oxidase 1; O2–, superoxide anion; RFK, riboflavin kinase; RIP, receptor-interacting protein; ROS, reactive oxygen species; SMases, sphingomyelinases; TNF, tumor necrosis factor; TNFR, TNF receptor; TRADD, TNFR-associated death domain; TRAF, TNFR-associated factor; USP21, ubiquitin-specific peptidase 21. *Alternative names are provided in brackets. ‡Associated with TNFR. §Associated with the mitochondrial outer membrane.
TNFR1 complex II promotes apoptosis or necroptosis. Ligand-bound TNFR1 is internalized, leading to a shift in the molecular composition of the TNFR1 interactome (TABLE 1) and to the formation of a cytosolic death-inducing signalling complex (DISC), better known as complex II (REFS 43,50) (FIG. 1). RIP1 poly ubiquitylation not only affects NF-κB activation but also influences the transition from complex I to complex II (REFS 10,11,51). On deubiquitylation of RIP1 by the Lys63‑deubiquitylating enzyme cylindromatosis
(CYLD)14, RIP1 (together with its cognate kinase RIP3) is recruited to a supramolecular complex that includes TRADD, FAS-associated protein with a death domain (FADD) and caspase 8 (REF. 43). In line with this model, RNA interference (RNAi)-mediated knockdown of CYLD inhibits TNF-induced necroptosis14. It remains unclear whether other deubiquitylating enzymes, including A20 (also known as TNFAIP3)52, cezanne (also known as OTUD7B) 53 and ubiquitin-specific peptidase 21 (USP21)54, all of which inhibit NF-κB
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
VOLUME 11 | O CTOBER 2010 | 703 © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS ROS
TNF
NOX1
TNFR1
Plasma membrane
FAD p22phox NADPH
a
FMN Complex I
Cytoplasm
RF
RFK
TRADD TRAF2 and TRAF5 RIP1 cIAPs
Lys63-linked polyubiquitin USP21, A20 or cezanne
?
TAK1–TAB2–TAB3 complex
NF-κB activation
CYLD RIP1
Cytosolic formation of complex II
b
c TRADD RIP1
FADD RIP3
Caspase 8
Caspase 8 activity Caspase-dependent executioner mechanisms
TRADD RIP1
Active kinase
FADD RIP3
Caspase 8
P X?
Caspase 8 inhibitor Caspase-independent executioner mechanisms
and TNFR1 ligation results (at least in some cell types) in programmed necrosis9,13. Whether FADD or TRADD are strictly required to assemble the necroptosissignalling complex, or ‘necrosome’, is less clear. The absence of FADD sensitizes some cells, including Jurkat lymphocytes, to necrotic cell death9,35. In contrast, MEFs isolated from FADD-deficient mice are resistant to TNF-induced necroptosis57. Both TNF-induced apoptosis and necroptosis (obtained in the presence of the chemical pan-caspase inhibitor Z‑VAD.fmk) are blocked in TRADD-deficient cells58, suggesting that, at least in some experimental settings, TRADD (which is also part of TNFR complex I; see above) constitutes an indispensable cell deathinducing adaptor protein43. In contrast to these observations, TRADD cannot be detected in complex II formed on TNFR ligation in the presence of second mitochondria-derived activator of caspase (SMAC; also known as Diablo) mimetics (chemical agents that block cIAPs by mimicking the activity of SMAC, a mitochondrial cIAP inhibitor). Moreover, RNAi-mediated knockdown of TRADD stimulates (rather than inhibits) the formation of complex II in some cell types, suggesting that TRADD is not required for the assembly and function of complex II (REFS 56,59).
The necrosome signalling complex. The term necrosome refers to a multiprotein complex containing RIP1 and RIP3 that stimulates necroptosis59. The formation Figure 1 | TNFR1‑elicited signalling pathways. a | On tumour necrosis factor (TNF) Nature Reviews | Molecular Cell Biology binding, TNF receptor 1 (TNFR1) undergoes a conformational change, allowing for the of the necrosome is highly regulated by ubiquitylation intracellular assembly of the so-called TNFR complex I, which includes TNF receptor(see above) and mutual RIP1 and RIP3 phosphorylaassociated death domain (TRADD), receptor-interacting protein 1 (RIP1; also known as tion (see below). Whereas many cell lines are protected RIPK1), cellular inhibitor of apoptosis proteins (cIAPs), TNF receptor-associated factor 2 against TNF-induced apoptosis by Z‑VAD.fmk, others (TRAF2) and TRAF5. On cIAP-mediated Lys63‑ubiquitylation, RIP1 can serve as a scaffold respond to TNF plus Z‑VAD.fmk by activating necropfor the recruitment of transforming growth factor‑β activated kinase 1 (TAK1), tosis60, a phenomenon that has recently been correlated TAK1‑binding protein 2 (TAB2) and TAB3, which initiate the canonical nuclear factor-κB with the expression of RIP3 (REF. 11). RIP3 contains an (NF-κB) activation pathway (BOX 2). Riboflavin kinase (RFK) physically bridges the TNFR1 N‑terminal kinase domain and a C‑terminal RIP homodeath domain to p22phox (also known as CYBA), the common subunit of multiple typic interaction motif (RHIM), which mediates its NADPH oxidases, including NADPH oxidase 1 (NOX1), which also contributes to interaction with RIP1 (REF. 61). Necroptosis induced by TNFα-induced necroptosis by generating reactive oxygen species (ROS). Conversely, on deubiquitylation by cylindromatosis (CYLD; and perhaps also by A20 (also known as CD95L, TRAIL or TNF in combination with Z‑VAD.fmk TNFAIP3), cezanne (also known as OTUD7B) or ubiquitin-specific peptidase 21 (USP21)), is abrogated in RIP1‑deficient T cells9, and enforced RIP1 exerts lethal functions, which can be executed by two distinct types of cell death. dimerization of RIP1 can induce necroptosis in FADDb | The internalization of TNFR1 is accompanied by a change in its binding partners that deficient Jurkat lymphocytes7. Consistent with a role for leads to the cytosolic assembly of TNFR complex II, which often (but not invariably) RIP1 in NF-κB-mediated pro-survival signalling, mice contains TRADD, FAS-associated protein with a death domain (FADD), caspase 8, RIP1 lacking RIP1 display extensive apoptosis in lymphoid and RIP3 (also known as RIPK3). Normally, caspase 8 triggers apoptosis by activating and adipose tissues and die 1–3 days after birth62. In conthe classical caspase cascade. It also cleaves, and hence inactivates, RIP1 and RIP3. trast to RIP1, RIP3 is not involved in NF-κB activation10. c | If caspase 8 is blocked by pharmacological or genetic interventions, RIP1 and RIP3 Recent experiments with cells that have been stably or become phosphorylated (perhaps by an unidentified kinase) and engage the effector mechanisms of necroptosis. FAD, flavin adenine nucleotide; FMN; flavin mononucleotide. temporarily depleted of RIP3 showed that this kinase is required for necroptosis and revealed the existence of a RIP1- and RIP3‑containing complex that is assembled activation, also stimulate the lethal functions of RIP1. In in response to TNF and is stabilized in the presence of complex II, caspase 8 inactivates RIP1 and RIP3 by pro- SMAC mimetics or caspase inhibitors10–12. teolytic cleavage and initiates the pro-apoptotic caspase In 2005, Yuan and colleagues identified necrostatin 1 activation cascade11,55. Moreover, genetic or pharmaco- and necrostatin 3, small molecules that allosterically logical inhibition of cIAPs prevents RIP1 ubiquitylation block the kinase activity of RIP1, thereby inhibiting Necrostatin 1 A Trp-based molecule and favours the formation of complex II, thus sensitiz- necroptosis but leaving RIP1‑mediated activation of (5‑(1H‑indol-3‑ylmethyl)ing cells to RIP1‑dependent activation of caspase 8 and NF-κB, mitogen-activated protein kinase p38 and JUN 3‑methyl‑2‑thioxo-4‑ apoptosis47,56. By contrast, when caspase 8 is deleted, N‑terminal kinase 1 (JNK1) unaffected7,63. Several imidazolidinone) that was first depleted or inhibited by CrmA or pharmacological other necrostatins have been identified by virtue of identified as a specific and potent inhibitor of necroptosis7. agents, complex II cannot enter the ‘apoptotic mode’ their capacity to suppress necrosis induced by TNF Apoptosis
Necroptosis
704 | O CTOBER 2010 | VOLUME 11
www.nature.com/reviews/molcellbio © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS Box 2 | The NF-κB system Nuclear factor-κB (NF-κB) refers to a heterogeneous group of dimeric transcription factors belonging to the REL protein family, which can be activated by tumour necrosis factor receptor 1 (TNFR1), pathogens, toxins, drugs and oxidants. In mammals, five NF-κB subunits share a highly conserved REL homology domain (RHD), which mediates dimerization, DNA binding and the interaction with inhibitor of NF-κB (IκB) proteins. These subunits are NFKB1 (p50 and its precursor p105), NFKB2 (p52 and its precursor p100), cREL, RELA (p65) and RELB164. NF-κB homodimers or heterodimers are normally sequestered in the cytoplasm by IκB proteins. In the canonical pathway (see the figure, part a), the IκB kinase (IKK) complex (composed of one regulatory subunit, IKKγ (also known as NEMO), and two catalytic subunits, IKKα and IKKβ) responds to specific signals (including TNFR1 ligation) or nonspecific stress by phosphorylating IκB to target it for destruction by E1–E2–E3‑mediated ubiquitylation and proteasomal degradation31. IκB degradation unmasks the RHD and a nuclear localization signal (NLS; which is common to all REL proteins) on associated NF-κB dimers, allowing them to access the nucleus and bind DNA31. The IKK complex is stabilized by the linear ubiquitin chain assembly complex (LUBAC), which linearly adds ubiquitin (Ub) moieties to IKKγ165. In the non-canonical pathway (see the figure, part b), which responds to a specific set of differentiating or developmental stimuli, the IKK complex comprises IKKα dimers and is activated by NF-κB-inducing kinase (NIK)-mediated phosphorylation. In turn, active IKKα phosphorylates p100 to promote its proteolytic processing to p52, which can dimerize with other NF-κB subunits and enter the nucleus. Once bound to nuclear DNA, NF-κB dimers regulate the expression of genes implicated in a plethora of patho physiological processes, including innate and adaptive immune responses, inflammation, cell proliferation, cell death and cell survival. Alterations of the IKK–NF-κB signalling module (most often resulting in constitutive NF-κB activation) contribute to oncogenesis and tumour development in many solid or haematopoietic malignancies166. One example is provided by the negative NF-κB regulator cylindromatosis (CYLD), a deubiquitylating enzyme, the loss-of-function mutation of which leads to familial cylindromatosis167,168. Moreover, NF-κB activation reduces the apoptotic potential of anticancer chemotherapeutics, thereby favouring resistance169. Pharmacological inhibitors of the NF-κB system might therefore directly target oncogene addiction170 or sensitize tumour cells to chemotherapy. Multiple NF-κB inhibitors are being evaluated in clinical trials, alone or in combination with radiotherapy or chemotherapy169.
b Non-canonical pathway
a Canonical pathway IκB p65 IκB p50
Ub Ub
IKKγ
Ub
P
IKKα IKKβ
P P
Linear ubiquitylation
P
Kγ
Ub
IK
LUBAC Ub Ubiquitylation
Ub Ub Ub Ub Ub
IKKα IKKα
P
IκB p65 IκB p50
NIK
IKKα IKKα
P
P
P100 RELB
Processing
E1 E2 E3
0 p5 0 p5
P
P100 RELB
RELB NF-κB p52 dimers
p6 p6 5 5
Ub
IKK complex
IKK complex
p65 p50
REL p52 B
cREL cREL
IκB IκB
cREL cREL
p65 p50 Degradation
p52 RELB
Oncogene addiction An expression coined by Weinberg in 2002 (REF. 170) to describe the observation that tumour maintenance often depends on the continued activity of some oncogenes.
Immunity
Inflammation
Cell death
Cell survival
Proliferation Nature Reviews | Molecular Cell Biology VOLUME 11 | O CTOBER 2010 | 705
NATURE REVIEWS | MOLECULAR CELL BIOLOGY © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS plus Z‑VAD.fmk, but they inhibit RIP1 indirectly by interfering with upstream signals that are yet to be elucidated63–65. Necrostatin 1 abolishes the assembly of the RIP1–RIP3 complex, suggesting that the kinase activity of RIP1 is required for necrosome formation10,11. Necroptosis depends on a tightly regulated mutual relationship between RIP1 and RIP3 kinase activities, involving the autophosphorylation of RIP1 on Ser161 and direct or indirect RIP3‑mediated phosphorylation of RIP1 (REFS 11,63). Recently, murine cytomegalovirus infection was shown to induce RIP3‑dependent but RIP1‑independent necroptosis66. Moreover, overexpression of catalytically active RIP3 can trigger necroptosis irrespective of the presence of RIP1 (REF. 12). Thus, at least in some cases, RIP1 may not constitute an absolute requirement for necroptosis induction. Altogether, these results point to the existence of a highly complex, tightly regulated signal transduction pathway that connects death receptors to pro-inflammatory, apoptotic or necrotic signal transducers.
Execution of necroptosis Several distinct molecular mechanisms contribute to the execution of TNFR1‑initiated necroptosis. Some of these effectors can also be activated by other necroptotic triggers, including PAMPs and DNA damage (see above).
Mitochondrial transmembrane potential (Δψm)
The electrochemical gradient built across the inner mitochondrial membrane by the proton pumps associated with the respiratory chain. The Δψm creates a proton-moving force that is required for mitochondrial ATP generation by the F1–FO ATP synthase, and its permanent dissipation is considered an early sign of apoptosis.
Bioenergetic aspects of the execution of necroptosis. During apoptosis, ATP-consuming processes including PARP1 activity 67, translation68 and proteasome-mediated degradation69 are rapidly shut off by caspases. By contrast, during TNF-induced necroptosis, these processes persist and hence may contribute to the lethal decline in intracellular ATP70. PARP1 is a nuclear enzyme involved in DNA repair and transcriptional regulation71. The overactivation of PARP1, perhaps due to ROS-mediated DNA damage, is critically involved in the necroptotic response of L929 fibrosarcoma cells to TNF72 and eventually results in the depletion of ATP and NAD (FIG. 2). In response to DNA alkylation, PARP1 activation and the consequent NAD depletion and/or PAR accumulation stimulates the release of apoptosis-inducing factor (AIF) from the mitochondrial intermembrane space, a process that reportedly depends on calpains — Ca2+activated non-caspase Cys proteases73–75. Cytosolic AIF rapidly relocalizes to the nuclear compartment, where it mediates caspase-independent, large-scale DNA fragmentation25, which in turn can further stimulate PARP activation, thereby initiating a vicious cycle. Harlequin mice, which bear a hypomorphic mutation of Aifm1 and therefore express reduced amounts of AIF, are protected against several necrotic stimuli, including ischaemia– reperfusion injury of the brain76–78. Similarly, pharmaco logical and genetic inhibition of PARP1 has consistent cytoprotective effects79. Surprisingly, both RIP1‑deficient and TRAF2-deficient MEFs are resistant to PARP1‑induced cell death in response to DNA alkylating agents80, indicating that RIP1 activation can also occur downstream of PARP1, at least in specific experimental settings. In line with this notion,
PARP1 hyperactivation not only causes mitochondrial dysfunction but also activates JNKs, two processes that have been shown to enhance necrotic cell death in some experimental setups80,81. A direct link between RIP1 and decreasing ATP concentrations (which occur during necroptosis) was postulated when the existence of a RIP1‑dependent signal that results in the inhibition of adenine nucleotide translocase (ANT) was uncovered22. In physiological circumstances, ANT, an integral protein of the inner mitochondrial membrane, exchanges mitochondrially neosynthesized ATP with cytosolic ADP25. Inhibition of ANT by RIP1 can be expected to reduce intramitochondrial ADP levels, leading first to the inhibition of F1–FO ATP synthase (as ADP is its substrate) and then to the reversal of F1–FO ATP synthase activity, which causes the ATP hydrolysis-driven extrusion of protons from the mitochondrial matrix, resulting in a net increase in the mitochondrial transmembrane potential (Δψm). This model is apparently corroborated by the fact that mitochondria show transiently increased Δψm during the early phases of necroptosis15,24. ANT has also been suggested to interact with the voltage-dependent anion channel (VDAC; present on the outer mitochondrial membrane) and cyclophilin D (CYPD; present in the mitochondrial matrix) to generate the PTPC (reviewed in REF. 25). In response to some lethal triggers, including oxidative stress and Ca2+ overload, the PTPC adopts a high conductance conformation, permitting the unregulated entry of solutes and water into the mitochondrial matrix, a phenomenon that has been dubbed the mitochondrial permeability transition25. The PTPC is a highly dynamic entity that interacts with multiple proteins, including pro- and anti-apoptotic members of the BCL‑2 protein family 82. However, it is not known whether BCL‑2‑modifying factor (BMF), a BH3‑only protein required for TNF-induced necroptosis14, functionally or physically interacts with the PTPC. Both pharmacological and genetic interventions aimed at inhibiting backbone components of the PTPC, including VDAC, ANT and CYPD, mediate cytoprotective effects against numerous insults in vitro and in vivo (reviewed in REFS 25,78). As it stands, CYPD seems to be the leading player of the PTPC, as genetic ablation of peptidylprolyl isomerase F (Ppif; the CYPD-encoding gene), but not of the genes coding for all known VDAC and ANT isoforms83,84, consistently protects mice against ischaemic injury of the brain and heart in vivo20,21. These results underscore the importance of mitochondrial events in pathological necroptosis. ROS and RNS contribute to the execution of necroptosis. Mitochondrial energy metabolism was first linked to the execution of necrosis in the early 1990s, when the Fiers group showed that ROS production by mitochondrial respiratory complex I is crucial for the necrotic response of L929 cells to TNF85. Mitochondrial ROS also mediate cell death-associated ultrastructural changes of the mitochondria and endoplasmic reticulum (ER)85,86. Although ROS production is not essential for all instances of TNFinduced necrosis11,86, the kinase activity of RIP3 may link TNFR1 signalling, mitochondrial bioenergetics
706 | O CTOBER 2010 | VOLUME 11
www.nature.com/reviews/molcellbio © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS UV or ROS
Nucleus
DNA damage
PARP activation
8
Glycogenolysis Glycogen PYGL G1P PGM
PARP activation
G6P
Glycolysis
Pyruvate
[ATP]↓
P
GLUL
Ser161 RIP1
Glutaminase
Pyruvate
P
GLUD1 α-Ketoglutarate ara ratte te
TCA 2 cycle Fumarate CYPD Succinate ADP ATP T H+ NADH NAD+ SDH ATP AN T ½O2 H2O AC AN V VD AC II I VD IV III + PTPC ∆ψm↑ ROS H Lipid 7 H+ + H Mitochondrion peroxidation
Respiratory complex component Coenzyme Q10 Cytochrome c
P Ser199 RIP3 Active kinase
X? 3
ANT
AIF
Gln
Glu + NH4+
ROS Calpains
RIP1–RIP3 necrosome
Glutaminolysis
Methylglyoxal↑ 6 AGEs↑
?
1 Glu + NH4+
JNK
4 SMases
Ferritin↓
Ceramide Calpains
Labile iron pool↑
Sphingosine
ROS [ATP]↓
5 Lipid peroxidation Phospholipids
Ca2+
cPLA2 Arachidonic acid Lipoxygenase Lipid hydroperoxides
Lipid peroxidation
Lysosomes LMP
Necroptosis
Figure 2 | Execution of necroptosis. When caspase activation is prevented, receptor-interacting protein 1 (RIP1; also known as RIPK1) and RIP3 (also known as RIPK3) are phosphorylated and elicit necroptosis. The RIP1–RIP3 necrosome Nature Reviews | Molecular Cell Biology stimulates glycogenolysis and glutaminolysis by enhancing glycogen phosphorylase (PYGL), glutamate–ammonia ligase (GLUL) and glutamate dehydrogenase 1 (GLUD1) activity (1), inhibits the mitochondrial adenine nucleotide translocase (ANT) to deplete cytosolic ATP (2), activates JUN N‑terminal kinase (JNK)-mediated degradation of ferritin, thus increasing the labile iron pool (3), and favours sphingomyelinase (SMase)-mediated generation of ceramide, which is converted into sphingosine by ceramidase and promotes a cytosolic Ca2+ wave that activates calpains and cytosolic phospholipase A2 (cPLA2; 4). cPLA2 triggers lipid peroxidation by mobilizing the lipoxygenase substrate arachidonic acid and may be required for SMase-mediated ceramide generation (5). Sphingosine, calpains and lipid hydroperoxides induce lysosome membrane permeabilization (LMP), resulting in the leakage of cytotoxic hydrolases into the cytosol. Oxidative metabolism favours the generation of reactive oxygen species (ROS) by the mitochondrial respiratory chain and the formation of redox-active advanced glycation end products (AGEs; 6). ROS (which also derive from NADPH oxidase 1 (NOX1; see FIG. 1), ceramide metabolism and labile iron pool elevation), initiate vicious cycles of damage by exacerbating mitochondrial uncoupling and lipid peroxidation and favour the opening of the permeability-transition pore complex (PTPC; 7). This results in the permeabilization of mitochondrial membranes and the translocation of cytotoxic proteins, including apoptosis-inducing factor (AIF), from the mitochondrial intermembrane space to the cytosol. Alternatively, AIF release can follow a poly(ADP-ribose) polymerase 1 (PARP1)–calpain cascade triggered by DNA damage (8). As cytosolic AIF enters the nucleus to exert endonucleolytic functions, and PARP1 overactivation rapidly depletes cytosolic ATP, DNA damage can initiate a feed-forward signalling loop towards necroptosis. Notably, RIP1 can also operate downstream of PARP1 to execute necroptosis. Δψm, mitochondrial transmembrane potential; CYPD, cyclophilin D; G1P, glucose‑1‑phosphate; G6P, glucose‑6‑phosphate; PGM, phosphoglucomutase; SDH, succinate dehydrogenase; TCA, tricarboxylic acid; VDAC, voltage-dependent anion channel.
and ROS overproduction (FIG. 2). RIP3 physically interacts with and allosterically activates several metabolic enzymes, including glycogen phosphorylase (PYGL), glutamate–ammonia ligase (GLUL) and glutamate dehydrogenase 1 (GLUD1). RNAi-mediated knockdown of any of these enzymes attenuates TNF plus Z-VAD.fmk-mediated ROS production and necroptosis12. PYGL catalyses the breakdown of glycogen
into glucose‑1‑phosphate (glycogenolysis), which can be converted into the glycolytic substrate glucose‑6‑ phosphate87, thereby stimulating glycolysis (which eventually contributes to ROS generation). The RIP3‑mediated necrotic boost on glycogenolysis can also favour the production of methylglyoxal, a cytotoxic compound for which the synthetic rate is proportional to glycolytic flux 88. Methylglyoxal covalently binds to proteins and
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
VOLUME 11 | O CTOBER 2010 | 707 © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS
Advanced glycation end product (AGE) The product of a chain of chemical reactions that most often is initiated by non-enzymatic protein glycosylation. Increased extracellular glucose favours the accumulation of AGEs, which interact with specific receptors on the plasma membrane to stimulate the generation of intracellular ROS.
Haber–Weiss reaction The generation of hydroxyl radicals from hydrogen peroxide and superoxide (H2O2 + O•2– → OH•+ HO– + O2). The reaction is very slow, but is catalysed by ferric ions (Fe3+).
Fenton reaction The ferrous ion (Fe2+)-dependent decomposition of dihydrogen peroxide, generating the highly reactive hydroxyl radical (Fe2+ + H2O2 → Fe3+ + OH• + OH–).
Lipid peroxidation The biochemical reaction whereby free radicals ‘steal’ electrons from lipids in cell membranes, resulting in ultrastructural damage to organelles.
Labile iron pool A cytosolic fraction of iron ions loosely bound to macromolecules (for example, ferritin) — also known as a chelatable iron pool — that harbours the metabolically active (and hence potentially toxic) forms of ferrous (Fe2+) and ferric (Fe3+) ions.
Oxidative phosphorylation The process whereby respiratory chain complexes embedded in the inner mitochondrial membrane catalyse a series of redox reactions that provide the free energy to generate the Δψm.
forms advanced glycation end products (AGEs), which alter protein function and constitute new centres of sustained ROS generation88. Mitochondrial proteins seem to be particularly prone to methylglyoxal‑mediated post-translational modifications89. Thus, inhibition of glycolysis reportedly attenuates cell death by apoptosis and necroptosis (as both these processes are stimulated by ROS), whereas the blockage of methylglyoxal‑ detoxifying pathways accelerates it 88. Both GLUL, a cytosolic enzyme that condensates glutamate and free ammonia into Glu, and GLUD1, a mitochondrial enzyme that converts glutamate to α‑ketoglutarate, are required for glutaminolysis (reviewed in REF. 90). Glutaminolysis results in the generation of α‑ketoglutarate, which feeds into the Krebs cycle to generate reduced equivalents and pyruvate (by malate decarboxylase), in turn favouring lactate accumulation90. Moreover, mitochondrial Glu catabolism increases the local concentration of ammonia, thus facilitating ROS generation by the respiratory chain91. Altogether, there seem to be several mechanisms by which enhanced glycogenolysis, glycolysis and glutaminolysis can contribute to the respiratory burst that characterizes necrotic cell death. Non-mitochondrial ROS production by the plasma membrane NADPH oxidase NOX1 (which is recruited by RIP1) also contributes to TNF-induced necrotic cell death16. NOX1 activation is dependent on ribof lavin kinase, which physically bridges the TNFR1 death domain and p22phox (also known as CYBA), the common subunit of multiple NADPH oxidases17. NOX1 is activated within minutes of the administration of TNF, and it is possible that NOX1‑generated ROS trigger or sustain the subsequent production of ROS by the mitochondrial respiratory chain92. ROS generation is auto-amplified through several reactions. For instance, interaction of hydrogen peroxide with the superoxide anion in the Haber–Weiss reaction or with ferrous (Fe2+) ions in the Fenton reaction generates the highly reactive hydroxyl radical, further promoting lipid peroxidation28. Interestingly, whereas low levels of ROS can favour a mild mitochondrial uncoupling that is detrimental to ATP synthesis but exerts cytoprotective effects93, ROS overgeneration engages the respiratory chain in a potentially lethal vicious cycle that also entails the generation of reactive nitrogen species (RNS)94 (see below). Similar to mitochondrial ROS, NOX1‑derived ROS are not a requisite for necroptosis, as shown by the fact that small interfering RNA-mediated downregulation of NOX1 almost abrogates TNF-induced ROS generation but only marginally rescues L929 fibrosarcoma cells from necroptosis16. Accordingly, ROS scavengers such as tert-butyl4‑hydroxyanisole (BHA) exert anti-necroptotic effects in some (but not all) experimental settings95,96. Thus, the relative contribution of ROS from distinct sources to necroptosis may be dictated by the cell type. TNF also stimulates ROS formation by favouring JNK1‑dependent degradation of the ubiquitous ironbinding protein ferritin, resulting in an increase in the labile iron pool97. Redox-active iron is a threat to cells and needs to be cautiously transported and stored in an inactive form98. Ferritin-deficient cells (in which the
iron storage capacity is reduced) are more resistant to TNF-induced labile iron pool elevation, ROS generation and necroptosis than their wild-type counterparts99. Intriguingly, RIP1‑deficient MEFs also failed to elevate the labile iron pool on TNF administration99, suggesting that RIP1 might modulate the induction of ROS through an effect on ferritin. However, the exact molecular mechanisms underlying this phenomenon and the possible implication of RIP1 remain to be elucidated. At low intracellular concentrations, nitric oxide functions as a second messenger in a myriad of signalling pathways. If overproduced, nitric oxide is highly toxic and leads to the generation of RNS with distinctive chemical and biological properties100. Similar to ROS, RNS are potent oxidants and can initiate or propagate lipid101 and protein oxidation and peroxidation102. Recently, nitration has been shown to elicit RIP1- and RIP3‑mediated necroptosis, with respiratory complex I subunit NDUFB8 being involved103. This is apparently in contrast with the well-known cytoprotective effects of nitrite, which attenuates oxidative stress, mitochondrial damage and dysfunction, hypothermia, tissue infarction and organismal death in a murine model of TNF-induced shock104. Nitrites also confer protection against ischaemia–reperfusion injuries in vivo, presumably owing to nitrite-dependent inhibition of mitochondrial ROS generation105 or to an effect on the soluble guanylate cyclase α1 subunit, one of the main intracellular receptors for nitric oxide and signal transducers in the cardiovascular system104. Involvement of LMP in the execution of necroptosis. ROS can react with polyunsaturated fatty acids in cellular membranes to generate reactive aldehydes (such as 4‑hydroxynonenal), which in turn can attack protein and lipid moieties in membranes, thereby compromising their integrity 106. In mitochondria, the products of lipid peroxidation inhibit oxidative phosphorylation, compromise the permeability of the inner membrane, dissipate the Δψm and reduce the Ca2+ buffering capacity, thus contributing to necrosis107. Lipid peroxidationmediated destabilization of cellular membranes (including the plasma, lysosomal and ER membranes) results in a leakage of proteases or an elevation of cytosolic Ca2+ concentrations, two phenomena that participate in necrotic cell death. Lysosomes are the only intracellular compartment in which redox-active iron temporarily resides before it is incorporated into the catalytic centre of specific enzymes or stored in ferritin108. Typically, the Fenton reaction is favoured in the lumen of lysosomes, not only because lysosomes are enriched in reduced iron (Fe2+) and reducing equivalents (provided by Cys, ascorbic acid and reduced glutathione), but also because they are permeable to hydrogen peroxide and lack hydrogen peroxidedetoxifying enzymes, such as catalases and glutathione peroxidases108. Accordingly, oxidative stress-induced lipid peroxidation, LMP and cell death can be prevented by the iron chelator desferrioxamine 24,109. Cytosolic phospholipase A2 (cPLA2) and ceramide also act upstream of lipid peroxidation to stimulate LMP. PLA2 is an esterase that produces arachidonic acid from
708 | O CTOBER 2010 | VOLUME 11
www.nature.com/reviews/molcellbio © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS arachidonate-containing phospholipids110. Treatment of L929 cells with TNF leads to PLA2 activation, and overexpression of cPLA2 sensitizes TNF-resistant L929 variants to necroptosis111. Arachidonic acid is converted by lipoxygenase into membrane-damaging lipid hydroperoxides23 (FIG. 2). In response to TNF, both the lysosomal enzyme acid sphingomyelinase (aSMase) and its neutral counterpart at the plasma membrane (nSMase) transform sphingomyelin into ceramide, which in turn can be converted to sphingosine by ceramidase112. Sphingosine has been characterized as a lysosomotropic LMP inducer (see below)113. Depending on the specific experimental setting (the cell type or lethal trigger), ceramide can induce either apoptosis or necroptosis114. Isoformspecific pharmacological inhibition of nSMase protects breast cancer MCF7 cells against TNF-induced apoptosis115. Notably, both RIPK1–/– human Jurkat lymphocytes and cPLA2-deficient murine L929 cells fail to accumulate ceramide on TNF administration and are protected against TNF-induced necroptosis, suggesting that RIP1, as well as cPLA2, might be required for SMase-mediated generation of ceramide and consequent cell death96. A connection between Ca2+ homeostasis and LMP was first suggested by the observation that TNF induces a moderate increase in intracellular Ca2+ concentrations, resulting in the generation of enlarged lysosomes that are particularly prone to LMP116. In some experimental settings, for example in vivo during the neuronal response to ischaemia–reperfusion, lysosomal membranes can be destabilized by calpains78,117. Calpain-mediated LMP results in the cytosolic spillage of lysosomal hydrolases such as proteases of the cathepsin family, which play an important part during necrotic cell death118. Accordingly, pharmacological inhibitors of cathepsins confer consistent neuroprotection in vivo119. Importantly, calpain has also been shown to proteolytically inactivate the plasma membrane Na+–Ca2+ exchanger, thereby engaging a positive feedback loop of self-activation mediated by the irreversible accumulation of cytosolic Ca2+ (REF. 120). Additional evidence showing the important role of LMP in necrotic cell death has been provided by the genetic manipulation of 70 kDa heat shock protein (HSP70), a guardian of lysosomal membrane integrity 121. HSP70 specifically interacts with the endo lysosomal anionic phospholipid bis(monoacylglycero) phosphate121 and may also constitute the lysosomal target of calpain-mediated proteolysis on its ROS-mediated carbonyl ation 118. HSP70 delays LMP and necrosis induced by TNF, heat shock or oxidative stress122–124. In response to an ischaemic insult, mitochondria from HSP70‑overexpressing cells exhibit reduced levels of ROS production and lipid peroxidation125. Whether this represents a primary effect of HSP70 on LMP, mitochondrial membranes or iron homeostasis124 is not yet clear.
Macropinosome A large intracellular vesicle filled with extracellular fluids and macromolecules that is formed by macropinocytosis.
Disposal of necrotic cells When confronted with cell death, the immune system clears corpses, stimulates the replacement of lost cells, alerts host defences if infectious agents are detected and possibly eliminates cells approaching oncogenic transformation. The type and nature of dying cells, the history
of prior attempts to cope with stress and the biochemical routes leading to death influence the cell surface characteristics while affecting the release of ‘find-me’ signals (for the attraction of phagocytes), the exposure of ‘eat-me’ signals (for corpse engulfment) and the disclosure of ‘danger’ signals (which are often part of otherwise ‘hidden’ molecules). The combination of these cell death-associated molecules (CDAMs) determines which engulfing cells are recruited, how they are activated and how they interpret the dead cell’s antigens. A particular set of CDAMs can be decoded by the microenvironment of dying cells to alternatively trigger silent corpse removal, tissue repair responses, recruitment of additional inflammatory effectors, or full-blown immune responses126,127. So, what impact does programmed necrosis have on the inflammatory or immune system? Apoptotic cells emit a series of well-defined ‘find-me’ (such as soluble lysophosphatidylcholine (LPC)128 and ATP129) and ‘eat-me’ (such as surface-exposed and oxidized phosphatidylserine130) signals, allowing them to engage in synapse-like interactions with macrophages and to be recruited into tight-fitting phagosomes through a zipper-like mechanism131. Often, apoptotic corpses are taken up by phagocytic cells in the absence of inflamma tory or immunogenic reactions. In some cases, cells that are en route to necrosis also externalize phosphatidylserine before plasma membrane permeabilization132, thus facilitating their recognition and internalization by phagocytes133,134. However, fully necrotic cells are internalized by macrophages through the formation of spacious macropinosomes135, a process that is accompanied by macrophage ruffling and involves the sorting of fluid-phase macromolecules, as judged by the colocalization of fluid-phase tracers131. Thus, the handling of apoptotic and necrotic cells by the immune system is radically distinct. In spite of this fundamental difference, both apoptotic and necrotic cells are efficiently cleared by professional and non-professional phagocytes and hence are rarely found in tissues. Defective clearance of dying cells, however, may contribute to the persistence of inflammation, excessive tissue injury and the pathogenesis of chronic obstructive pulmonary disease 136, diabetes137, atherosclerosis138 or autoimmune diseases such as systemic lupus erythematosus139. It has been a common paradigm that apoptosis is antiphlogistic (anti-inflammatory) and tolerogenic (producing immunological tolerance) but necrosis triggers inflammation and an immune response. This paradigm must be refined because in some cases, in particular on ER stress, apoptosis can be interpreted by the immune system as immunogenic 140, and the immunogenicity of apoptosis is lost when the same cells undergo necrotic lysis on freeze–thaw cycles141. Moreover, depending on the cell type, necrotic cells can even inhibit inflammatory reactions. Necrotic (ATP depleted or subjected to freeze–thaw cycles) and apoptotic (but not heat-killed) Jurkat lymphocytes have been shown to inhibit Escherichia coli-induced TNF secretion by human macrophages to a similar extent 134. Moreover, macrophages can engulf necrotic L929 cells (which have been killed by TNF) without producing
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
VOLUME 11 | O CTOBER 2010 | 709 © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS inflammatory cytokines133. These findings reveal an unexpected complexity in the interaction between necrotic cells and the phagocytic system. In spite of these caveats, it must be noted that necrotic cells can release multiple pro-inflammatory factors, including the alarmin SAP130, heat-shock proteins (such as HSP70, HSP90 and GP96), histones, high mobility group protein B1 (HMGB1) and several nonproteinaceous factors (such as RNA, DNA and monosodium urate microcrystals), all of which act on different PRRs on immune effector cells to activate inflammatory reactions (reviewed in REF. 126). Histones released from necrotic cells have a major pathogenic role in sepsis, and their neutralization by antibodies or activated protein C can prevent organismal lethality 142. Recently, mitochondrial damage-associated molecular patterns (DAMPs), including N‑formylated peptides and mitochondrial DNA, were found to be released by necrotic cells into the circulation and to contribute to neutrophil-mediated organ injury similarly to bacterial PAMPs143, underscoring an evolutionarily conserved link between distinct routes to innate immunity. Recent genetic manipulations suggest an important role for necrosis in the outcome of viral infections and immunosurveillance. Ripk3–/– mice fail to control vaccinia virus infection, because virus-elicited necrosis can limit viral replication and/or stimulate the antiviral immune response10. Intriguingly, some viral genomes encode inhibitors of necrotic cell death that may facilitate their propagation and their subversion of the immune response. For example, murine cytomegalovirus (MCMV) expresses the proteins M36 and M45, which inhibit caspase 8‑mediated apoptosis144 and RIP1- and RIP3‑mediated necroptosis145, respectively. We suspect that multiple virus-encoded necrosis-inhibitory factors will be discovered as the comprehension of signalling events in necroptosis advances. It has not been investigated in detail whether necrotic cell death might exert a tumour-suppressive function like apoptotic cell death does. However, mice lacking CYLD (which is required for TNF-induced necrosis)14 are highly susceptible to developing a wide array of tumours, including skin and colon cancers146,147. It remains to be determined whether CYLD merely acts as a cell-autonomous tumour suppressor or whether it is required for stimulating immunosurveillance, and whether these effects can be attributed to its role in necroptosis or NF-κB signalling.
Pathophysiological facets of necroptosis Necrosis can occur in a programmed manner during development (for example, the death of chondrocytes controlling the longitudinal growth of bones)148 and in adult tissue homeostasis (for example, in intestinal epithelial cells)149. Moreover, cells in which apoptosisassociated caspase activation has been blocked often succumb to necrosis in response to the same stimuli that would usually induce apoptosis. Thus, interdigital cells or thymocytes from apoptotic peptidase-activating factor 1 (Apaf1–/–) embryos or adult mice, respectively, undergo necrotic cell death to the same extent and with
the same timing as cells from wild-type mice would undergo apoptosis150. Nonetheless, necrosis is mostly associated with pathological conditions, including neurodegeneration, ischaemia–reperfusion and infection. Excitotoxicity, oxidative stress and mitochondrial dysfunction, all of which contribute to the execution of necroptosis (see above), are indeed implicated in stroke as well as in Alzheimer’s, Huntington’s and Parkinson’s diseases (reviewed in REF. 151). The ageing brain accumulates iron, copper and zinc, resulting in increased oxidative stress by the Fenton reaction, which contributes to necrotic cell death152. Accordingly, iron chelation and ROS scavenging may delay the manifestation of neurodegenerative diseases109. Intriguingly, necroptosis has recently emerged as a prominent antiviral mechanism, as shown by the fact that Ripk3–/– mice are more susceptible to viral infection than their wild-type counterparts10, and by the existence of viral factors that contain the RHIM domain and interfere with the RIP1–RIP3 interaction66. Ripk1–/– mice are not viable62, and tissue-specific knockout and kinase-dead knock-in models will be required to elucidate the contribution of RIP1 to pathological cell loss. Pharmacological RIP inhibitors, including the RIP1‑specific agent necrostatin 1 and geldanamycin (which downregulates RIP1, RIP3 and several other HSP90 client proteins)10,153, exert cyto protective effects in vitro in several distinct experimental settings (Supplementary information S2 (table)). Intriguingly, in some cell types, geldanamycin induces a switch from TNF-induced necroptosis to apoptosis154. The inhibition of RIP1 kinase activity also attenuates neurodegenerative diseases155, brain ischaemia7, myocardial infarction 156 and head trauma 157 in vivo (Supplementary information S3 (table)), underscoring the contribution of RIP1 to pathological cell death. Similarly, pharmacological or genetic inhibition of PARP1, CYPD, cPLA2 or RIP3 limits cell loss in vivo in several rodent models of injury (Supplementary information S3 (table) and Supplementary information S4 (table)). Parp1–/– mice are protected from haemorrhagic shock158, acute pancreatitis and consequent lung injury 159. Mice lacking CYPD are more resistant to ischaemia–reperfusion damage of the brain21 and the heart 20 than their wild-type counterparts. Similarly, cPLA2-deficient mice exhibit reduced injury after brain ischaemia160. Finally, cerulein-induced pancreatic acinar cell loss and pancreatitis are greatly reduced in Ripk3–/– mice11,12. Altogether, these results support the idea that specific inhibitors of RIP1, RIP3, PARP1, CYPD and cPLA2 can attenuate pathological cell loss in vivo in rodent models of human disease.
Problems and perspectives Necroptosis can be conceived as a partially programmed event of cellular explosion. Physiological signals or cellular damage are perceived by specific receptors or sensors that ignite a detonator, which in turn activates the blasting agent. It is important to accurately distinguish and molecularly identify the upstream signals — the detonator and the explosives — for several important reasons.
710 | O CTOBER 2010 | VOLUME 11
www.nature.com/reviews/molcellbio © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS First, the elucidation of the precise molecular hierarchy involved in different cell death scenarios may clarify whether one ‘core programme’ or several independent pathways of necrosis exist. Undoubtedly, there are different ways to induce necrosis, be it through the activation of specific receptors or by inflicting distinct types of cellular damage. However, it is still debatable whether the core features of necrosis (such as RIP1 activation, bioenergetic and redox crisis, and lysosomal and mitochondrial perturbation) are built in a single interdependent circuit or several independent, mutually stimulatory (self-amplifying) circuits. Understanding this is important for the development of necroptosis-inhibitory cytoprotective drugs. The existence of several distinct pathways that ignite necrosis would imply that they all need to be interrupted simultaneously for cytoprotection, suggesting the need for combination therapies. Although this has not been addressed systematically, it may be beneficial to combine cytoprotective agents that target different lethal subroutines (for example, apoptosis and necroptosis) and processes (for example, LMP and the mitochondrial permeability transition) for optimal therapeutic results. Second, upstream signals (as opposed to downstream effectors) may constitute better targets for the specific pharmacological suppression of unwarranted cell death. The interception of a pro-necrotic signal transduction pathway would be more efficient if it occurred at an early step, for instance at the level of TNF–TNFR1 interaction, rather than downstream. Moreover, the interruption of specific (stimulus-dependent) pro-necrotic signals should decrease negative side effects. As has previously been shown for several components of the apoptotic machinery (reviewed in REF. 161), necrosis-relevant molecules and processes have ‘day jobs’ and hence exert physiological functions, in particular in the response to cell stress and infection and in immune or inflammatory responses, that should not be perturbed. Third, downstream signals, which are usually (but not always) activated late in the pathway (when the initial signalling cascade has already been engaged), are also attractive therapeutic targets, as (at least in some settings) they could be blocked after the primary lesion (such as stroke, trauma, infarction and sepsis).
1.
2.
3. 4. 5.
Lockshin, R. A. & Williams, C. M. Programmed cell death — II. Endocrine potentiation of the breakdown of the intersegmental muscles of silkmoths. J. Insect Physiol. 10, 643–649 (1964). Kerr, J. F., Wyllie, A. H. & Currie, A. R. Apoptosis: a basic biological phenomenon with wide-ranging implications in tissue kinetics. Br. J. Cancer 26, 239–257 (1972). Lettre, G. & Hengartner, M. O. Developmental apoptosis in, C. elegans: a complex CEDnario. Nature Rev. Mol. Cell Biol. 7, 97–108 (2006). Schweichel, J. U. & Merker, H. J. The morphology of various types of cell death in prenatal tissues. Teratology 7, 253–266 (1973). Kroemer, G. et al. Classification of cell death: recommendations of the Nomenclature Committee on Cell Death 2009. Cell Death Differ. 16, 3–11 (2009). This article provides up-to-date guidelines for the use of cell death-related terminology in scientific publications, as provided by the Nomenclature Committee on Cell Death, an organization composed of reputed researchers in the field of cell death worldwide.
6.
7.
8.
9.
Although intercepting upstream signals is the most desirable therapeutic choice, early interventions are rarely (if ever) achievable in the treatment of, for example, patients with stroke and trauma. Moreover, targeting downstream events is desirable when the upstream signals are not uniform or when they are transduced by multiple, interconnected pathways. These considerations underscore the importance of appropriately dissecting the chronological and functional aspects of necrotic demolition and defining the exact ‘point of no return’ beyond which cytoprotection can no longer be achieved. Fourth, cytotoxic T lymphocytes and natural killer cells, the cell death-inducing activity of which can contribute to the pathophysiology of human diseases, including AIDS and autoimmune disorders, reportedly ‘overkill’ their targets by transferring multiple proteases and membranepermeabilizing proteins into them, thereby triggering both apoptotic and necrotic programmes. Thus, rescuing targets from this type of cytotoxic attack may require a multipronged strategy that is yet to be optimized. Fifth, multiple cancer cell lines display an altered propensity to undergo necroptosis, which, at least partially, correlates with the expression levels of RIP3 (REF. 11). Further work is required to elucidate the importance of this finding in vivo and, in particular, whether it would be possible to stimulate the necrotic demise of RIP3‑proficient tumour cells to circumvent apoptosis resistance. It is also unknown whether, and which, necrotic pathways might elicit immunogenic tumour cell death and hence ignite a highly desirable anticancer immune response that would eliminate residual tumour (stem) cells. We anticipate that resolving these questions will help in the design of cytoprotective and cytotoxic therapies, with important implications for neuroprotection, cardioprotection, organ preservation and cancer therapy. Although the molecular exploration of programmed necrosis is still in its infancy, it is clear that interrupting pro-necrotic signals may prevent pathological cell loss in many human diseases. In this respect, the development of necrostatins63 may have paved the way for the development of a new class of potentially powerful therapeutic agents for clinical applications.
Laster, S. M., Wood, J. G. & Gooding, L. R. Tumor necrosis factor can induce both apoptic and necrotic forms of cell lysis. J. Immunol. 141, 2629–2634 (1988). Degterev, A. et al. Chemical inhibitor of nonapoptotic cell death with therapeutic potential for ischemic brain injury. Nature Chem. Biol. 1, 112–119 (2005). Hsu, H., Huang, J., Shu, H. B., Baichwal, V. & Goeddel, D. V. TNF-dependent recruitment of the protein kinase RIP to the TNF receptor‑1 signaling complex. Immunity 4, 387–396 (1996). Holler, N. et al. Fas triggers an alternative, caspase‑8‑independent cell death pathway using the kinase RIP as effector molecule. Nature Immunol. 1, 489–495 (2000). The authors discovered that, in some cell types, FAS can trigger non-apoptotic cell death that is independent of caspases but dependent on the adaptor protein FADD and the presence and enzymatic activity of the protein kinase RIP1. This milestone paper is the first report of RIP1‑dependent necroptosis.
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
10. Cho, Y. S. et al. Phosphorylation-driven assembly of the RIP1–RIP3 complex regulates programmed necrosis and virus-induced inflammation. Cell 137, 1112–1123 (2009). 11. He, S. et al. Receptor interacting protein kinase‑3 determines cellular necrotic response to TNF-α. Cell 137, 1100–1111 (2009). 12. Zhang, D. W. et al. RIP3, an energy metabolism regulator that switches TNF-induced cell death from apoptosis to necrosis. Science 325, 332–336 (2009). References 10–12 independently uncovered the obligate role of RIP3 in necroptosis. By multiple experimental approaches, RIP3 was shown to functionally and physically interact with RIP1, leading to a mitochondrial metabolic burst that underlies necroptosis execution. The pathophysiological role of RIP3 in vivo was substantiated by animal models of viral infection and acute pancreatitis. 13. Vercammen, D. et al. Inhibition of caspases increases the sensitivity of L929 cells to necrosis mediated by tumor necrosis factor. J. Exp. Med. 187, 1477–1485 (1998).
VOLUME 11 | O CTOBER 2010 | 711 © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS
14.
15.
16.
17. 18.
19.
20.
21.
22.
23. 24.
25.
26. 27.
28.
29. 30.
The sensitivity of murine L929 fibrosarcoma cells to TNF-induced necrosis was dramatically increased by expression of a serpin-like viral protein (CrmA) and by pharmacological caspase inhibitors, showing for the first time a major role for caspases in the apoptotis–necrosis switch. Hitomi, J. et al. Identification of a molecular signaling network that regulates a cellular necrotic cell death pathway. Cell 135, 1311–1323 (2008). The first time systems biology was applied to compare necroptosis to apoptosis in murine cells. Genome wide RNAi-based screens coupled to in silico and in vitro analyses allowed the delineation of a signalling network that regulates the molecular bifurcation between necroptosis and apoptosis. Goossens, V., Stange, G., Moens, K., Pipeleers, D. & Grooten, J. Regulation of tumor necrosis factorinduced, mitochondria- and reactive oxygen speciesdependent cell death by the electron flux through the electron transport chain complex I. Antioxid. Redox Signal. 1, 285–295 (1999). Kim, Y. S., Morgan, M. J., Choksi, S. & Liu, Z. G. TNFinduced activation of the Nox1 NADPH oxidase and its role in the induction of necrotic cell death. Mol. Cell 26, 675–687 (2007). Yazdanpanah, B. et al. Riboflavin kinase couples TNF receptor 1 to NADPH oxidase. Nature 460, 1159–1163 (2009). Goossens, V., Grooten, J. & Fiers, W. The oxidative metabolism of glutamine. A modulator of reactive oxygen intermediate-mediated cytotoxicity of tumor necrosis factor in L929 fibrosarcoma cells. J. Biol. Chem. 271, 192–196 (1996). Zong, W. X., Ditsworth, D., Bauer, D. E., Wang, Z. Q. & Thompson, C. B. Alkylating DNA damage stimulates a regulated form of necrotic cell death. Genes Dev. 18, 1272–1282 (2004). Nakagawa, T. et al. Cyclophilin D‑dependent mitochondrial permeability transition regulates some necrotic but not apoptotic cell death. Nature 434, 652–658 (2005). Schinzel, A. C. et al. Cyclophilin D is a component of mitochondrial permeability transition and mediates neuronal cell death after focal cerebral ischemia. Proc. Natl Acad. Sci. USA 102, 12005–12010 (2005). References 20 and 21 show that CYPD-deficient mice are protected against ischaemic insults, as compared to their wild-type littermates. Moreover, multiple types of primary cells isolated from Ppif –/– mice exhibit enhanced resistance to necrotic (but not apoptotic) stimuli. Together, these studies provide unequivocal evidence that CYPD regulates necroptosis in vivo. Temkin, V., Huang, Q., Liu, H., Osada, H. & Pope, R. M. Inhibition of ADP/ATP exchange in receptorinteracting protein-mediated necrosis. Mol. Cell. Biol. 26, 2215–2225 (2006). Boya, P. & Kroemer, G. Lysosomal membrane permeabilization in cell death. Oncogene 27, 6434–6451 (2008). Vanden Berghe, T. et al. Necroptosis, necrosis and secondary necrosis converge on similar cellular disintegration features. Cell Death Differ. 17, 922–930 (2010). Kroemer, G., Galluzzi, L. & Brenner, C. Mitochondrial membrane permeabilization in cell death. Physiol. Rev. 87, 99–163 (2007). This review provides a comprehensive analysis of the mitochondrial pathway of cell death and of its multifaceted implications for human physiology and pathology. Green, D. R. & Kroemer, G. Pharmacological manipulation of cell death: clinical applications in sight? J. Clin. Invest. 115, 2610–2617 (2005). Rosenbaum, D. M. et al. Necroptosis, a novel form of caspase-independent cell death, contributes to neuronal damage in a retinal ischemia-reperfusion injury model. J. Neurosci. Res. 88, 1569–1576 (2010). Vanlangenakker, N., Vanden Berghe, T., Krysko, D. V., Festjens, N. & Vandenabeele, P. Molecular mechanisms and pathophysiology of necrotic cell death. Curr. Mol. Med. 8, 207–220 (2008). Vercammen, D. et al. Dual signaling of the Fas receptor: initiation of both apoptotic and necrotic cell death pathways. J. Exp. Med. 188, 919–930 (1998). Chan, F. K. et al. A role for tumor necrosis factor receptor‑2 and receptor-interacting protein in programmed necrosis and antiviral responses. J. Biol. Chem. 278, 51613–51621 (2003).
31. Hacker, H. & Karin, M. Regulation and function of IKK and IKK-related kinases. Sci. STKE 2006, re13 (2006). 32. Wilson, N. S., Dixit, V. & Ashkenazi, A. Death receptor signal transducers: nodes of coordination in immune signaling networks. Nature Immunol. 10, 348–355 (2009). 33. Fiers, W. et al. TNF-induced intracellular signaling leading to gene induction or to cytotoxicity by necrosis or by apoptosis. J. Inflamm. 47, 67–75 (1995). 34. Martinon, F., Gaide, O., Petrilli, V., Mayor, A. & Tschopp, J. NALP inflammasomes: a central role in innate immunity. Semin. Immunopathol. 29, 213–229 (2007). 35. Kalai, M. et al. Tipping the balance between necrosis and apoptosis in human and murine cells treated with interferon and dsRNA. Cell Death Differ. 9, 981–994 (2002). 36. Ma, Y., Temkin, V., Liu, H. & Pope, R. M. NF‑κB protects macrophages from lipopolysaccharideinduced cell death: the role of caspase 8 and receptorinteracting protein. J. Biol. Chem. 280, 41827–41834 (2005). 37. Francois, M., Le Cabec, V., Dupont, M. A., Sansonetti, P. J. & Maridonneau-Parini, I. Induction of necrosis in human neutrophils by Shigella flexneri requires type III secretion, IpaB and IpaC invasins, and actin polymerization. Infect. Immun. 68, 1289–1296 (2000). 38. Koterski, J. F., Nahvi, M., Venkatesan, M. M. & Haimovich, B. Virulent Shigella flexneri causes damage to mitochondria and triggers necrosis in infected human monocyte-derived macrophages. Infect. Immun. 73, 504–513 (2005). 39. Willingham, S. B. et al. Microbial pathogen-induced necrotic cell death mediated by the inflammasome components CIAS1/cryopyrin/NLRP3 and ASC. Cell Host Microbe 2, 147–159 (2007). 40. Chu, J. J. & Ng, M. L. The mechanism of cell death during West Nile virus infection is dependent on initial infectious dose. J. Gen. Virol. 84, 3305–3314 (2003). 41. Ray, C. A. & Pickup, D. J. The mode of death of pig kidney cells infected with cowpox virus is governed by the expression of the CrmA gene. Virology 217, 384–391 (1996). 42. Chan, F. K. et al. A domain in TNF receptors that mediates ligand-independent receptor assembly and signaling. Science 288, 2351–2354 (2000). 43. Micheau, O. & Tschopp, J. Induction of TNF receptor Imediated apoptosis via two sequential signaling complexes. Cell 114, 181–190 (2003). This paper elucidated the early biochemical events that are engaged on TNFR1 ligation and, in particular, showed the existence of two distinct signalling complexes, one at the plasma membrane that includes TNFR1, TRAF2 and RIP1, and one at an endocytic compartment that involves internalized TNFR1, TRADD, FADD and caspase 8. 44. Deveraux, Q. L. et al. IAPs block apoptotic events induced by caspase‑8 and cytochrome c by direct inhibition of distinct caspases. EMBO J. 17, 2215–2223 (1998). 45. Csomos, R. A., Brady, G. F. & Duckett, C. S. Enhanced cytoprotective effects of the inhibitor of apoptosis protein cellular IAP1 through stabilization with TRAF2. J. Biol. Chem. 284, 20531–20539 (2009). 46. Rothe, M., Pan, M. G., Henzel, W. J., Ayres, T. M. & Goeddel, D. V. The TNFR2‑TRAF signaling complex contains two novel proteins related to baculoviral inhibitor of apoptosis proteins. Cell 83, 1243–1252 (1995). 47. Bertrand, M. J. et al. cIAP1 and cIAP2 facilitate cancer cell survival by functioning as E3 ligases that promote RIP1 ubiquitination. Mol. Cell 30, 689–700 (2008). 48. Wong, W. W. et al. RIPK1 is not essential for TNFR1‑induced activation of NF‑κB. Cell Death Differ. 17, 482–487 (2010). 49. Ea, C. K., Deng, L., Xia, Z. P., Pineda, G. & Chen, Z. J. Activation of IKK by TNFα requires site-specific ubiquitination of RIP1 and polyubiquitin binding by NEMO. Mol. Cell 22, 245–257 (2006). 50. Christofferson, D. E. & Yuan, J. Necroptosis as an alternative form of programmed cell death. Curr. Opin. Cell Biol. 22, 263–268 (2010). 51. Ting, A. T., Pimentel-Muinos, F. X. & Seed, B. RIP mediates tumor necrosis factor receptor 1 activation of NF‑κB but not Fas/APO‑1‑initiated apoptosis. EMBO J. 15, 6189–6196 (1996). 52. Shembade, N., Ma, A. & Harhaj, E. W. Inhibition of NF‑κB signaling by A20 through disruption of ubiquitin enzyme complexes. Science 327, 1135–1139 (2010).
712 | O CTOBER 2010 | VOLUME 11
53. Enesa, K. et al. NF‑κB suppression by the deubiquitinating enzyme Cezanne: a novel negative feedback loop in pro-inflammatory signaling. J. Biol. Chem. 283, 7036–7045 (2008). 54. Xu, G. et al. Ubiquitin-specific peptidase 21 inhibits tumor necrosis factor α-induced nuclear factor κB activation via binding to and deubiquitinating receptor-interacting protein 1. J. Biol. Chem. 285, 969–978 (2010). 55. Feng, S. et al. Cleavage of RIP3 inactivates its caspaseindependent apoptosis pathway by removal of kinase domain. Cell Signal 19, 2056–2067 (2007). 56. Wang, L., Du, F. & Wang, X. TNF-α induces two distinct caspase‑8 activation pathways. Cell 133, 693–703 (2008). 57. Lin, Y. et al. Tumor necrosis factor-induced nonapoptotic cell death requires receptor-interacting protein-mediated cellular reactive oxygen species accumulation. J. Biol. Chem. 279, 10822–10828 (2004). 58. Ermolaeva, M. A. et al. Function of TRADD in tumor necrosis factor receptor 1 signaling and in TRIFdependent inflammatory responses. Nature Immunol. 9, 1037–1046 (2008). 59. Declercq, W., Vanden Berghe, T. & Vandenabeele, P. RIP kinases at the crossroads of cell death and survival. Cell 138, 229–232 (2009). 60. Vandenabeele, P., Vanden Berghe, T. & Festjens, N. Caspase inhibitors promote alternative cell death pathways. Sci. STKE 2006, pe44 (2006). 61. Sun, X., Yin, J., Starovasnik, M. A., Fairbrother, W. J. & Dixit, V. M. Identification of a novel homotypic interaction motif required for the phosphorylation of receptor-interacting protein (RIP) by RIP3. J. Biol. Chem. 277, 9505–9511 (2002). 62. Kelliher, M. A. et al. The death domain kinase RIP mediates the TNF-induced NF‑κB signal. Immunity 8, 297–303 (1998). 63. Degterev, A. et al. Identification of RIP1 kinase as a specific cellular target of necrostatins. Nature Chem. Biol. 4, 313–321 (2008). 64. Teng, X. et al. Structure–activity relationship study of [1,2,3]thiadiazole necroptosis inhibitors. Bioorg. Med. Chem. Lett. 17, 6836–6840 (2007). 65. Vandenabeele, P., Declercq, W. & Berghe, T. V. Necrotic cell death and ‘necrostatins’: now we can control cellular explosion. Trends Biochem. Sci. 33, 352–355 (2008). 66. Upton, J. W., Kaiser, W. J. & Mocarski, E. S. Virus inhibition of RIP3‑dependent necrosis. Cell Host Microbe 7, 302–313 (2010). 67. Soldani, C. & Scovassi, A. I. Poly(ADP-ribose) polymerase‑1 cleavage during apoptosis: an update. Apoptosis 7, 321–328 (2002). 68. Saelens, X. et al. Protein synthesis persists during necrotic cell death. J. Cell Biol. 168, 545–551 (2005). 69. Sun, X. M. et al. Caspase activation inhibits proteasome function during apoptosis. Mol. Cell 14, 81–93 (2004). 70. Leist, M., Single, B., Castoldi, A. F., Kuhnle, S. & Nicotera, P. Intracellular adenosine triphosphate (ATP) concentration: a switch in the decision between apoptosis and necrosis. J. Exp. Med. 185, 1481–1486 (1997). 71. Kraus, W. L. Transcriptional control by PARP‑1: chromatin modulation, enhancer-binding, coregulation, and insulation. Curr. Opin. Cell Biol. 20, 294–302 (2008). 72. Los, M. et al. Activation and caspase-mediated inhibition of PARP: a molecular switch between fibroblast necrosis and apoptosis in death receptor signaling. Mol. Biol. Cell 13, 978–988 (2002). 73. Yu, S. W. et al. Mediation of poly(ADP-ribose) polymerase‑1‑dependent cell death by apoptosisinducing factor. Science 297, 259–263 (2002). 74. Cao, G. et al. Critical role of calpain I in mitochondrial release of apoptosis-inducing factor in ischemic neuronal injury. J. Neurosci. 27, 9278–9293 (2007). 75. Moubarak, R. S. et al. Sequential activation of poly(ADP-ribose) polymerase 1, calpains, and Bax is essential in apoptosis-inducing factor-mediated programmed necrosis. Mol. Cell. Biol. 27, 4844–4862 (2007). 76. Slemmer, J. E. et al. Causal role of apoptosis-inducing factor for neuronal cell death following traumatic brain injury. Am. J. Pathol. 173, 1795–1805 (2008). 77. Culmsee, C. et al. Apoptosis-inducing factor triggered by poly(ADP-ribose) polymerase and Bid mediates neuronal cell death after oxygen-glucose deprivation and focal cerebral ischemia. J. Neurosci. 25, 10262–10272 (2005).
www.nature.com/reviews/molcellbio © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS 78. Galluzzi, L., Blomgren, K. & Kroemer, G. Mitochondrial membrane permeabilization in neuronal injury. Nature Rev. Neurosci. 10, 481–494 (2009). 79. Boujrad, H., Gubkina, O., Robert, N., Krantic, S. & Susin, S. A. AIF-mediated programmed necrosis: a highly regulated way to die. Cell Cycle 6, 2612–2619 (2007). 80. Xu, Y., Huang, S., Liu, Z. G. & Han, J. Poly(ADP-ribose) polymerase‑1 signaling to mitochondria in necrotic cell death requires RIP1/TRAF2‑mediated JNK1 activation. J. Biol. Chem. 281, 8788–8795 (2006). 81. Alano, C. C., Ying, W. & Swanson, R. A. Poly(ADPribose) polymerase‑1‑mediated cell death in astrocytes requires NAD+ depletion and mitochondrial permeability transition. J. Biol. Chem. 279, 18895–18902 (2004). 82. Brenner, C. et al. Bcl‑2 and Bax regulate the channel activity of the mitochondrial adenine nucleotide translocator. Oncogene 19, 329–336 (2000). 83. Baines, C. P., Kaiser, R. A., Sheiko, T., Craigen, W. J. & Molkentin, J. D. Voltage-dependent anion channels are dispensable for mitochondrial-dependent cell death. Nature Cell Biol. 9, 550–555 (2007). 84. Kokoszka, J. E. et al. The ADP/ATP translocator is not essential for the mitochondrial permeability transition pore. Nature 427, 461–465 (2004). 85. Schulze-Osthoff, K. et al. Cytotoxic activity of tumor necrosis factor is mediated by early damage of mitochondrial functions. Evidence for the involvement of mitochondrial radical generation. J. Biol. Chem. 267, 5317–5323 (1992). This paper provided the first evidence for a role of complex I‑mediated ROS production in TNF-induced necrotic cell death. 86. Festjens, N. et al. Butylated hydroxyanisole is more than a reactive oxygen species scavenger. Cell Death Differ. 13, 166–169 (2006). 87. Bouche, C., Serdy, S., Kahn, C. R. & Goldfine, A. B. The cellular fate of glucose and its relevance in type 2 diabetes. Endocr. Rev. 25, 807–830 (2004). 88. Van Herreweghe, F. et al. Tumor necrosis factorinduced modulation of glyoxalase I activities through phosphorylation by PKA results in cell death and is accompanied by the formation of a specific methylglyoxal-derived AGE. Proc. Natl Acad. Sci. USA 99, 949–954 (2002). 89. Rabbani, N. & Thornalley, P. J. Dicarbonyls linked to damage in the powerhouse: glycation of mitochondrial proteins and oxidative stress. Biochem. Soc. Trans. 36, 1045–1050 (2008). 90. Mates, J. M. et al. Glutamine homeostasis and mitochondrial dynamics. Int. J. Biochem. Cell Biol. 41, 2051–2061 (2009). 91. Albrecht, J. & Norenberg, M. D. Glutamine: a Trojan horse in ammonia neurotoxicity. Hepatology 44, 788–794 (2006). 92. Morgan, M. J., Kim, Y. S. & Liu, Z. G. TNFα and reactive oxygen species in necrotic cell death. Cell Res. 18, 343–349 (2008). 93. Brookes, P. S. Mitochondrial H+ leak and ROS generation: an odd couple. Free Radic. Biol. Med. 38, 12–23 (2005). 94. Poyton, R. O., Ball, K. A. & Castello, P. R. Mitochondrial generation of free radicals and hypoxic signaling. Trends Endocrinol. Metab. 20, 332–340 (2009). 95. Chen, T. Y., Chi, K. H., Wang, J. S., Chien, C. L. & Lin, W. W. Reactive oxygen species are involved in FasLinduced caspase-independent cell death and inflammatory responses. Free Radic. Biol. Med. 46, 643–655 (2009). 96. Thon, L. et al. Ceramide mediates caspaseindependent programmed cell death. FASEB J. 19, 1945–1956 (2005). 97. Antosiewicz, J., Ziolkowski, W., Kaczor, J. J. & Herman-Antosiewicz, A. Tumor necrosis factor‑αinduced reactive oxygen species formation is mediated by JNK1‑dependent ferritin degradation and elevation of labile iron pool. Free Radic. Biol. Med. 43, 265–270 (2007). 98. Kurz, T., Terman, A. & Brunk, U. T. Autophagy, ageing and apoptosis: the role of oxidative stress and lysosomal iron. Arch. Biochem. Biophys. 462, 220–230 (2007). 99. Xie, C. et al. Distinct roles of basal steady-state and induced H‑ferritin in tumor necrosis factor-induced death in L929 cells. Mol. Cell. Biol. 25, 6673–6681 (2005). 100. Brune, B. The intimate relation between nitric oxide and superoxide in apoptosis and cell survival. Antioxid. Redox Signal. 7, 497–507 (2005).
101. Goss, S. P., Singh, R. J., Hogg, N. & Kalyanaraman, B. Reactions of *NO, *NO2 and peroxynitrite in membranes: physiological implications. Free Radic. Res. 31, 597–606 (1999). 102. Tien, M., Berlett, B. S., Levine, R. L., Chock, P. B. & Stadtman, E. R. Peroxynitrite-mediated modification of proteins at physiological carbon dioxide concentration: pH dependence of carbonyl formation, tyrosine nitration, and methionine oxidation. Proc. Natl Acad. Sci. USA 96, 7809–7814 (1999). 103. Davis, C. W. et al. Nitration of the mitochondrial complex I subunit NDUFB8 elicits RIP1- and RIP3‑mediated necrosis. Free Radic. Biol. Med. 48, 306–317 (2010). 104. Cauwels, A. et al. Nitrite protects against morbidity and mortality associated with TNF- or LPS-induced shock in a soluble guanylate cyclase-dependent manner. J. Exp. Med. 206, 2915–2924 (2009). 105. Shiva, S. et al. Nitrite augments tolerance to ischemia/ reperfusion injury via the modulation of mitochondrial electron transfer. J. Exp. Med. 204, 2089–2102 (2007). 106. Benedetti, A., Comporti, M. & Esterbauer, H. Identification of 4‑hydroxynonenal as a cytotoxic product originating from the peroxidation of liver microsomal lipids. Biochim. Biophys. Acta 620, 281–296 (1980). 107. Orrenius, S., Gogvadze, V. & Zhivotovsky, B. Mitochondrial oxidative stress: implications for cell death. Annu. Rev. Pharmacol. Toxicol. 47, 143–183 (2007). 108. Kurz, T., Terman, A., Gustafsson, B. & Brunk, U. T. Lysosomes in iron metabolism, ageing and apoptosis. Histochem. Cell Biol. 129, 389–406 (2008). 109. Kurz, T., Gustafsson, B. & Brunk, U. T. Intralysosomal iron chelation protects against oxidative stress-induced cellular damage. FEBS J. 273, 3106–3117 (2006). 110. Burke, J. E. & Dennis, E. A. Phospholipase A2 structure/function, mechanism, and signaling. J. Lipid Res. 50, S237–S242 (2009). 111. Suffys, P. et al. Tumour‑necrosis‑factor-mediated cytotoxicity is correlated with phospholipase‑A2 activity, but not with arachidonic acid release per se. Eur. J. Biochem. 195, 465–475 (1991). 112. Won, J. S. & Singh, I. Sphingolipid signaling and redox regulation. Free Radic. Biol. Med. 40, 1875–1888 (2006). 113. Kagedal, K., Zhao, M., Svensson, I. & Brunk, U. T. Sphingosine-induced apoptosis is dependent on lysosomal proteases. Biochem. J. 359, 335–343 (2001). 114. Kim, W. H., Choi, C. H., Kang, S. K., Kwon, C. H. & Kim, Y. K. Ceramide induces non-apoptotic cell death in human glioma cells. Neurochem. Res. 30, 969–979 (2005). 115. Luberto, C. et al. Inhibition of tumor necrosis factorinduced cell death in MCF7 by a novel inhibitor of neutral sphingomyelinase. J. Biol. Chem. 277, 41128–41139 (2002). 116. Ono, K., Kim, S. O. & Han, J. Susceptibility of lysosomes to rupture is a determinant for plasma membrane disruption in tumor necrosis factor αinduced cell death. Mol. Cell. Biol. 23, 665–676 (2003). 117. Yamashima, T. et al. Sustained calpain activation associated with lysosomal rupture executes necrosis of the postischemic CA1 neurons in primates. Hippocampus 13, 791–800 (2003). 118. Yamashima, T. & Oikawa, S. The role of lysosomal rupture in neuronal death. Prog. Neurobiol. 89, 343–358 (2009). 119. Yamashima, T. et al. Inhibition of ischaemic hippocampal neuronal death in primates with cathepsin B inhibitor CA‑074: a novel strategy for neuroprotection based on ‘calpain-cathepsin hypothesis’. Eur. J. Neurosci. 10, 1723–1733 (1998). Monkeys undergoing temporary whole brain ischaemia were shown to be remarkably protected against hippocampal necrosis by the intravenous administration of a cathepsin B inhibitor, strongly supporting the pathophysiological relevance of the ‘calpain–cathepsin’ hypothesis in vivo and pointing to cathepsin inhibitors as potential neuroprotective agents. 120. Bano, D. et al. Cleavage of the plasma membrane Na+/Ca2+ exchanger in excitotoxicity. Cell 120, 275–285 (2005). 121. Kirkegaard, T. et al. Hsp70 stabilizes lysosomes and reverts Niemann-Pick disease-associated lysosomal pathology. Nature 463, 549–553 (2010).
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
122. Nylandsted, J. et al. Heat shock protein 70 promotes cell survival by inhibiting lysosomal membrane permeabilization. J. Exp. Med. 200, 425–435 (2004). 123. Tang, D. et al. Nuclear heat shock protein 72 as a negative regulator of oxidative stress (hydrogen peroxide)-induced HMGB1 cytoplasmic translocation and release. J. Immunol. 178, 7376–7384 (2007). 124. Doulias, P. T. et al. Involvement of heat shock protein‑70 in the mechanism of hydrogen peroxideinduced DNA damage: the role of lysosomes and iron. Free Radic. Biol. Med. 42, 567–577 (2007). 125. Williamson, C. L., Dabkowski, E. R., Dillmann, W. H. & Hollander, J. M. Mitochondria protection from hypoxia/reoxygenation injury with mitochondria heat shock protein 70 overexpression. Am. J. Physiol. Heart Circ. Physiol. 294, H249–H256 (2008). 126. Zitvogel, L., Kepp, O. & Kroemer, G. Decoding cell death signals in inflammation and immunity. Cell 140, 798–804 (2010). 127. Poon, I. K., Hulett, M. D. & Parish, C. R. Molecular mechanisms of late apoptotic/necrotic cell clearance. Cell Death Differ. 17, 381–397 (2010). 128. Lauber, K. et al. Apoptotic cells induce migration of phagocytes via caspase‑3‑mediated release of a lipid attraction signal. Cell 113, 717–730 (2003). 129. Elliott, M. R. et al. Nucleotides released by apoptotic cells act as a find-me signal to promote phagocytic clearance. Nature 461, 282–286 (2009). 130. Martin, S. J. et al. Early redistribution of plasma membrane phosphatidylserine is a general feature of apoptosis regardless of the initiating stimulus: inhibition by overexpression of Bcl‑2 and Abl. J. Exp. Med. 182, 1545–1556 (1995). 131. Krysko, D. V. et al. Macrophages use different internalization mechanisms to clear apoptotic and necrotic cells. Cell Death Differ. 13, 2011–2022 (2006). 132. Krysko, O., De Ridder, L. & Cornelissen, M. Phosphatidylserine exposure during early primary necrosis (oncosis) in JB6 cells as evidenced by immunogold labeling technique. Apoptosis 9, 495–500 (2004). 133. Brouckaert, G. et al. Phagocytosis of necrotic cells by macrophages is phosphatidylserine dependent and does not induce inflammatory cytokine production. Mol. Biol. Cell 15, 1089–1100 (2004). 134. Hirt, U. A. & Leist, M. Rapid, noninflammatory and PS‑dependent phagocytic clearance of necrotic cells. Cell Death Differ. 10, 1156–1164 (2003). 135. Krysko, D. V., Brouckaert, G., Kalai, M., Vandenabeele, P. & D’Herde, K. Mechanisms of internalization of apoptotic and necrotic L929 cells by a macrophage cell line studied by electron microscopy. J. Morphol. 258, 336–345 (2003). 136. Hodge, S., Hodge, G., Scicchitano, R., Reynolds, P. N. & Holmes, M. Alveolar macrophages from subjects with chronic obstructive pulmonary disease are deficient in their ability to phagocytose apoptotic airway epithelial cells. Immunol. Cell Biol. 81, 289–296 (2003). 137. O’Brien, B. A. et al. A deficiency in the in vivo clearance of apoptotic cells is a feature of the NOD mouse. J. Autoimmun. 26, 104–115 (2006). 138. Schrijvers, D. M., De Meyer, G. R., Herman, A. G. & Martinet, W. Phagocytosis in atherosclerosis: Molecular mechanisms and implications for plaque progression and stability. Cardiovasc. Res. 73, 470–480 (2007). 139. Munoz, L. E. et al. SLE — a disease of clearance deficiency? Rheumatology (Oxford) 44, 1101–1107 (2005). 140. Zitvogel, L. et al. Immune response against dying tumor cells. Adv. Immunol. 84, 131–179 (2004). 141. Casares, N. et al. Caspase-dependent immunogenicity of doxorubicin-induced tumor cell death. J. Exp. Med. 202, 1691–1701 (2005). 142. Xu, J. et al. Extracellular histones are major mediators of death in sepsis. Nature Med. 15, 1318–1321 (2009). 143. Zhang, Q. et al. Circulating mitochondrial DAMPs cause inflammatory responses to injury. Nature 464, 104–107 (2010). 144. Menard, C. et al. Role of murine cytomegalovirus US22 gene family members in replication in macrophages. J. Virol. 77, 5557–5570 (2003). 145. Mack, C., Sickmann, A., Lembo, D. & Brune, W. Inhibition of proinflammatory and innate immune signaling pathways by a cytomegalovirus RIP1‑interacting protein. Proc. Natl Acad. Sci. USA 105, 3094–3099 (2008).
VOLUME 11 | O CTOBER 2010 | 713 © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS 146. Massoumi, R., Chmielarska, K., Hennecke, K., Pfeifer, A. & Fassler, R. CYLD inhibits tumor cell proliferation by blocking Bcl‑3‑dependent NF‑κB signaling. Cell 125, 665–677 (2006). 147. Zhang, J. et al. Impaired regulation of NF‑κB and increased susceptibility to colitis-associated tumorigenesis in CYLD-deficient mice. J. Clin. Invest. 116, 3042–3049 (2006). 148. Roach, H. I. & Clarke, N. M. Physiological cell death of chondrocytes in vivo is not confined to apoptosis. New observations on the mammalian growth plate. J. Bone Joint Surg. Br. 82, 601–613 (2000). 149. Barkla, D. H. & Gibson, P. R. The fate of epithelial cells in the human large intestine. Pathology 31, 230–238 (1999). 150. Chautan, M., Chazal, G., Cecconi, F., Gruss, P. & Golstein, P. Interdigital cell death can occur through a necrotic and caspase-independent pathway. Curr. Biol. 9, 967–970 (1999). 151. Lin, M. T. & Beal, M. F. Mitochondrial dysfunction and oxidative stress in neurodegenerative diseases. Nature 443, 787–795 (2006). 152. Doraiswamy, P. M. & Finefrock, A. E. Metals in our minds: therapeutic implications for neurodegenerative disorders. Lancet Neurol. 3, 431–434 (2004). 153. Lewis, J. et al. Disruption of Hsp90 function results in degradation of the death domain kinase, receptorinteracting protein (RIP), and blockage of tumor necrosis factor-induced nuclear factor-κB activation. J. Biol. Chem. 275, 10519–10526 (2000). 154. Vanden Berghe, T., Kalai, M., van Loo, G., Declercq, W. & Vandenabeele, P. Disruption of HSP90 function reverts tumor necrosis factor-induced necrosis to apoptosis. J. Biol. Chem. 278, 5622–5629 (2003). 155. Yuan, J., Lipinski, M. & Degterev, A. Diversity in the mechanisms of neuronal cell death. Neuron 40, 401–413 (2003). 156. Lim, S. Y., Davidson, S. M., Mocanu, M. M., Yellon, D. M. & Smith, C. C. The cardioprotective effect of necrostatin requires the cyclophilin‑D component of the mitochondrial permeability transition pore. Cardiovasc. Drugs Ther. 21, 467–469 (2007). 157. You, Z. et al. Necrostatin‑1 reduces histopathology and improves functional outcome after controlled
cortical impact in mice. J. Cereb. Blood Flow Metab. 28, 1564–1573 (2008). 158. Liaudet, L. et al. Protection against hemorrhagic shock in mice genetically deficient in poly(ADP-ribose) polymerase. Proc. Natl Acad. Sci. USA 97, 10203–10208 (2000). 159. Mota, R. A. et al. Inhibition of poly(ADP-ribose) polymerase attenuates the severity of acute pancreatitis and associated lung injury. Lab. Invest. 85, 1250–1262 (2005). 160. Bonventre, J. V. et al. Reduced fertility and postischaemic brain injury in mice deficient in cytosolic phospholipase A2. Nature 390, 622–625 (1997). 161. Galluzzi, L. et al. No death without life: vital functions of apoptotic effectors. Cell Death Differ. 15, 1113–1123 (2008). 162. Galluzzi, L. et al. Guidelines for the use and interpretation of assays for monitoring cell death in higher eukaryotes. Cell Death Differ. 16, 1093–1107 (2009). 163. Krysko, D. V., Vanden Berghe, T., D’Herde, K. & Vandenabeele, P. Apoptosis and necrosis: detection, discrimination and phagocytosis. Methods 44, 205–221 (2008). 164. Hayden, M. S. & Ghosh, S. Signaling to NF‑κB. Genes Dev. 18, 2195–2224 (2004). 165. Tokunaga, F. et al. Involvement of linear polyubiquitylation of NEMO in NF‑κB activation. Nature Cell Biol. 11, 123–132 (2009). 166. Baud, V. & Karin, M. Is NF‑κB a good target for cancer therapy? Hopes and pitfalls. Nature Rev. Drug Discov. 8, 33–40 (2009). 167. Brummelkamp, T. R., Nijman, S. M., Dirac, A. M. & Bernards, R. Loss of the cylindromatosis tumour suppressor inhibits apoptosis by activating NF‑κB. Nature 424, 797–801 (2003). 168. Trompouki, E. et al. CYLD is a deubiquitinating enzyme that negatively regulates NF‑κB activation by TNFR family members. Nature 424, 793–796 (2003). References 167 and 168 elucidate the molecular determinants linking CYLD mutations and familial cylindromatosis. CYLD turned out to be a deubiquitylating enzyme that exerts oncosuppressive functions by negatively regulating NF-κB activation.
714 | O CTOBER 2010 | VOLUME 11
169. Nakanishi, C. & Toi, M. Nuclear factor-κB inhibitors as sensitizers to anticancer drugs. Nature Rev. Cancer 5, 297–309 (2005). 170. Weinstein, I. B. Addiction to oncogenes — the Achilles heal of cancer. Science 297, 63–64 (2002). 171. Sun, X. et al. RIP3, a novel apoptosis-inducing kinase. J. Biol. Chem. 274, 16871–16875 (1999).
Acknowledgements
We apologize to our colleagues for not citing all primary research papers owing to space restrictions, and we thank W. Declercq for fruitful discussions. Electron microscopy pictures in Box 1 were kindly provided by D. Krysko, Ghent University, VIB, Belgium. P.V. holds a Methusalem grant from the Flemish Government (BOF09/01M00709) and is supported by the Flanders Institute for Biotechnology (VIB), the Interuniversity Poles of Attraction-Belgian Science Policy (IAP6/18), Fonds voor Wetenschappelijk Onderzoek – Vlaanderen (FWO, G.0133.05 and 3G.0218.06), The Special Research Fund of Ghent University (Geconcerteerde Onderzoekstacties 12.0505.02) and the European Commission (EU Marie Curie Training and Mobility Program, ApopTrain, MRTN-CT‑035,624; EU FP7 Integrated Project, APO-SYS, HEALTH‑F4‑2007‑200,767; EU FP6 Integrated Project, Epistem, LSHB-CT‑2005‑019,067; Marie Curie Training and Mobility Program). L.G. and T.V.B. are financed by APO-SYS and FWO, respectively. G.K. is supported by Ligue Nationale contre le Cancer (Equipe labellisée), Agence Nationale pour la Recherche (ANR), the European Commission (APO-SYS, ChemoRes, ApopTrain, Active p53), Fondation pour la Recherche Médicale (FRM), Institut National du Cancer (INCa) and Cancéropôle Ile-de-France.
Competing interests statement
The authors declare no competing financial interests.
FURTHER INFORMATION Peter Vandenabeele’s homepage: http://www.dmbr.ugent.be/
SUPPLEMENTARY INFORMATION See online article: S1 (table) | S2 (table) | S3 (table) | S4 (table) ALL LINKS ARE ACTIVE IN THE ONLINE PDF
www.nature.com/reviews/molcellbio © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS
The engine driving the ship: metabolic steering of cell proliferation and death Marisa R. Buchakjian and Sally Kornbluth
Abstract | Metabolic activity is a crucial determinant of a cell’s decision to proliferate or die. Although it is not fully understood how metabolic pathways such as glycolysis and the pentose phosphate pathway communicate to cell cycle and apoptotic effectors, it is clear that a complex network of signalling molecules is required to integrate metabolic inputs. D‑type cyclins, cyclin-dependent kinases, the anaphase-promoting complex, p53, caspase 2 and B cell lymphoma 2 proteins, among others, have been shown to be regulated by metabolic crosstalk. Elucidating these pathways is of great importance, as metabolic aberrations and their downstream effects are known to contribute to the aetiology of cancer and degenerative disorders. Pentose phosphate pathway A metabolic pathway that generates NADPH and pentose sugars from glucose‑6phosphate. NADPH is important for the biosynthesis of many cell components and serves as a major cellular antioxidant.
Department of Pharmacology and Cancer Biology, Duke University Medical Center, Durham, North Carolina 27710, USA. Correspondence to S.K. e‑mail:
[email protected] doi:10.1038/nrm2972
A wide range of metabolites participate in cellular energy commerce; glucose, lipids, amino acids and nucleic acids are extensively broken down by the cell and remodelled to form countless metabolic intermediates and byproducts. The number of potential metabolic signalling molecules is staggering, and research is only beginning to uncover how these metabolites communicate with the cell cycle and the cell death machinery to affect cell fate. Of importance to this Review are several central meta‑ bolic pathways, including glycolysis, the tricarboxylic acid (TCA) cycle, the pentose phosphate pathway (PPP) and a range of biosynthetic pathways, most notably those that culminate in fatty acid biosynthesis (FIG. 1). Glucose enters the cell through glucose transporters (GLUTs) and, once intracellular, is phosphorylated to glucose‑6‑phosphate (G6P) by hexokinases. G6P can then proceed through glycolysis to produce ATP, the coenzyme NADH and pyruvate, or through the PPP, resulting in the produc‑ tion of ribose‑5‑phosphate and NADPH. In addition to hexokinases, glycolytic enzymes that are crucial for link‑ ing metabolism and the cellular events discussed in this Review include phosphofructokinase 1, which catalyses the conversion of fructose‑6‑phosphate to fructose‑1,6‑ bisphosphate, and pyruvate kinase, which yields pyruvate and ATP in the final step of glycolysis. Pyruvate produced by glycolysis is converted to acetylCoA, which enters the TCA cycle and undergoes a series of oxidative reactions, ultimately resulting in the produc‑ tion of two ATP molecules and six NADH molecules per glucose. NADH is then used in mitochondrial oxidative phosphorylation, which produces abundant ATP from nutrients in the cell. In addition to glucose, amino acids
can also funnel into the TCA cycle, as their catabolism results in the production of TCA cycle intermediates (α-ketoglutarate, succinyl-CoA, fumarate, oxaloacetate and acetyl-CoA). For example, glutamine is converted to glutamate by glutaminase, and glutamate in turn may be converted to α‑ketoglutarate to feed the TCA cycle. In addition to participating in the TCA cycle, acetyl-CoA is as a key precursor for fatty acid biosynthesis. AcetylCoA cannot cross the inner mitochondrial membrane, but intramitochondrial acetyl-CoA and oxaloacetate combine to form citrate, which is transported out of the mitochondria and broken back down into its constituents by ATP citrate lyase (ACL). Acetyl-CoA is converted to malonyl-CoA by acetyl-CoA carboxylase (ACC), and acetyl-CoA and malonyl-CoA are then both used by the multi-subunit enzyme fatty acid synthase (FAS) for the synthesis and elongation of fatty acid chains. Cytosolic and nuclear acetyl-CoA is also a precursor for the posttranslational modification of proteins (for example, histones) by acetylation. In the PPP, G6P is converted to ribose‑5‑phosphate while producing two molecules of NADPH. The enzyme that governs entry of G6P into this pathway is glucose-6‑ phosphate dehydrogenase (G6PD), which is regulated by the availability of its substrate (for example, the expres‑ sion of the gene encoding G6PD increases when animals transition from a fasting to a fed state) and the NADPH to NADP+ ratio1. NADPH is both a major cellular anti‑ oxidant, maintaining glutathione in a reduced state to prevent oxidative damage, and a required cofactor in the reductive biosynthesis of fatty acids, nucleotides and amino acids.
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
VOLUME 11 | O CTOBER 2010 | 715 © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS Glucose
Cytosol
GLUT Glucose 1
Hexokinase
G6PD
6-phosphogluconolactone
G6P PGI Fructose-2,6bisphosphate
Plasma membrane
Histone acetylation
PFKFB
NADPH
FAS Malonyl-CoA
ACL
3
F6P
NADPH
PFK1
Fructose-1,6-bisphosphate
ACC Acetyl-CoA 6
7 Fatty acid elongation
NADPH
Citrate
Ribose-5-phosphate
Glyceraldehyde 3-phosphate
Glutamine 5
GAPDH 1,3-bisphosphoglycerate
Glutaminase Glutamate
PK
PDH
Pyruvate
4 Acetyl-CoA
NADH
Citrate α-ketoglutarate
Oxaloacetate NADH NADH Malate
Succinyl-CoA
Electron transport chain
2 ATP
NADH
Mitochondrion
PEP
ATPase
H++ H+ H+ H H++ H+ H+ H H+
ATP
Figure 1 | Overview of metabolism. Metabolic pathways important to this Review include glycolysis, the tricarboxylic acid (TCA) cycle, the pentose phosphate pathway (PPP) and fatty acid biosynthesis. Glucose enters the cell and is Nature Reviews | Molecular Cell Biology phosphorylated to glucose‑6‑phosphate (G6P) by hexokinase (1). G6P either proceeds through glycolysis to produce ATP, NADH and pyruvate (2), or through the PPP, producing ribose‑5‑phosphate and NADPH (3). G6P dehydrogenase (G6PD) dictates entry of G6P into the PPP, and G6P oxidation produces NADPH. NADPH is an important cellular antioxidant and is a cofactor in the reductive biosynthesis of fatty acids, nucleotides and amino acids. Pyruvate produced by glycolysis is converted to acetyl-CoA, which enters the TCA cycle and produces two ATP molecules and six NADH molecules per glucose (4). NADH is then used in mitochondrial oxidative phosphorylation for ATP production. Glutamine and other amino acids also feed into the TCA cycle; glutamine, for example, is converted to glutamate by glutaminase, and glutamate can be converted to α‑ketoglutarate (5). Acetyl-CoA is also an important precursor for fatty acid biosynthesis. Intramitochondrial acetyl-CoA and oxaloacetate combine to form citrate, which is transported out of the mitochondria and broken back down by ATP citrate lyase (ACL) (6). Acetyl-CoA can be used as a precursor for post-translational acetylation of proteins and can also be converted to malonyl-CoA by acetyl-CoA carboxylase (ACC), resulting in the synthesis and elongation of fatty acid chains (7). GAPDH, glyceraldehyde‑3‑phosphate dehydrogenase; GLUT, glucose transporter; F6P, fructose‑6‑phosphate; FAS, fatty acid synthase; PDH, pyruvate dehydrogenase; PEP, phosphoenolpyruvate; PFK1, phosphofructokinase 1; PFKFB, 6‑phosphofructo‑2‑kinase/fructose‑2,6‑bisphosphatase; PGI, phosphoglucose isomerase; PK, pyruvate kinase.
Anaphase-promoting complex A large multisubunit E3 ubiquitin ligase with a RING-containing subunit (APC11) that ubiquitylates, among other proteins, several proteins that are crucial for the transition from M phase to G1.
This Review provides a glimpse into the ways in which the metabolic pathways introduced briefly above influence key cell cycle and apoptotic effectors to pro‑ mote cell survival or death. Molecular responses to nutrient flux are complex, and many reviews on the topic have focused on upstream, multimodal nutrientsensing pathways such as those initiated by phospho inositide 3‑kinase (PI3K)–AKT, 5′-AMP-activated protein kinase (AMPK) and mammalian target of rapamycin2–5. In this Review we dissect the modu lation of cell cycle and apoptotic effectors by meta bolism, taking a bottom-up approach to examine how nutrient-initiated signalling pathways impinge on key
cellular processes. The effects of metabolism on cell cycle targets are discussed first, with specific emphasis on cyclins and cyclin-dependent kinases (CDKs), CDK inhibitors, the anaphase-promoting complex (APC) and chromatin remodelling. Crosstalk between metabolism and apoptosis, and the roles of p53, caspase 2, B cell lymphoma 2 (BCL‑2) family proteins, cytochrome c and the apoptosome at the interface of these cellular processes, is also examined. The identification and understanding of metabolic effectors in the cell cycle and cell death machinery provides a platform for future inquiry and for rational drug design to specifically target metabolism-responsive effectors.
716 | O CTOBER 2010 | VOLUME 11
www.nature.com/reviews/molcellbio © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS PFKFB3 p27 Cdc25C Cyclin B CDK1
Glucose uptake
APC–CDH1 M
G2
D-type cyclins CDK4 and CDK6 G1
Cell cycle
APC–CDH1 Cyclin A
Hexokinase II PK FAS ACC
Insulin secretion
S
CDK2 E2F1 ATP/AMP imbalance
Cyclin E CDK2
Figure 2 | Crosstalk between cell cycle transitions and metabolism. Transitions between phases of the cell cycle (G1, S, G2 and M) are orchestrated by cell cycle Nature Reviews | Molecular Cell Biology activators and inhibitors, including cyclin–cyclin-dependent kinase (CDK) pairs, Cdc25C, the anaphase-promoting complex (APC) and p27. Signalling by metabolic enzymes (including 6‑phosphofructo‑2‑kinase/fructose‑2,6‑bisphosphatase isoenzyme 3 (PFKFB3), hexokinase II, pyruvate kinase (PK), fatty acid synthase (FAS), acetyl-CoA carboxylase (ACC)) and small molecule metabolites (for example, glucose and fruc‑ tose‑2,6‑bisphosphate), influences cell cycle transitions to determine cell fate. For example, PFKFB3, which promotes fructose‑2,6‑bisphosphate production and helps to drive glycolysis, dictates cell cycle progression by upregulating cyclin D3 and Cdc25C and downregulating the cell cycle inhibitor p27, all of which contribute to an increase in CDK1 activity. In a reciprocal manner, APC– Cdc20 homologue 1 (CDH1) ubiquitylates PFKFB3, targeting it for degradation, thereby moving metabolic intermediates from glycolysis to the pentose phosphate pathway. Further crosstalk from the cell cycle back to metabolism is exemplified by the observation that cyclin D1 overexpression results in decreased activity of the key glycolytic enzyme hexokinase II, and cyclin D1 knockdown promotes increases in PK, FAS and ACC, which influence glycolysis and fatty acid metabolism. Furthermore, in Drosophila melanogaster, cyclin E is degraded in response to a decrease in ATP following mitochondrial dysfunction, which results in cell cycle arrest. In pancreatic β‑cells, glucose uptake ultimately signals for cyclin D2 expression, and increased CDK4 activity in the G1 phase promotes E2F1‑mediated expression of cell components involved in glucose-mediated insulin release.
Powering on: metabolism and division To proliferate, a cell must cycle through interphase, which consists of G1, S and G2 phases, and then mitosis, the stage at which a cell divides into two daughter pro geny. Given the energy required to replicate the entire contents of the cell, including DNA as well as organelle and membrane components, cells must be able to assess whether there are adequate carbohydrates, nucleotides, amino acids and fatty acids to initiate and complete these processes. From the perspective of a single cell, if nutrients were unavailable, it would be unwise to expend energy on division instead of the housekeeping
functions required for continued survival. Thus, inform ation on metabolic status must be taken into account by the cell cycle machinery when deciding when to divide and when to conserve energy and abandon the path towards mitosis (FIG. 2; TABLE 1). Cyclins–CDKs and metabolism to fuel interphase. Dynamic regulation of the cell cycle machinery ensures that progression is unidirectional and DNA is replicated only once per cycle. This process is tightly controlled by the master regulatory proteins cyclins and CDKs. The cyclin–CDK interaction confers activity to the pair, and specific cyclin–CDK combinations predominate during different phases of the cell cycle to orchestrate accurate cell cycle progression. CDKs are constitutively expressed and regulate cell cycle substrates by phosphorylation, whereas the activity of cyclins is primarily determined by oscillatory changes in their protein expression in response to molecular cues6. Phosphorylation of down‑ stream targets by cyclins–CDKs alters their activity to promote controlled entry into the next phase of the cell cycle6,7. Following mitosis, cells must decide whether to commit to G1 phase and prepare for subsequent division or whether to exit the cycle into G0 phase. G0 is a metabolic ally conservative phase in which cells display decreased gene expression and lowered activity of proteins involved in nucleotide, carbohydrate and lipid biosynthesis, the products of which are required in subsequent cell cycle phases for division8,9. Given the importance of this com‑ mitment to division in G1, it is perhaps not surprising that recent literature emphasizes new links between cell metabolism and G1 phase regulators. Of note, D‑type cyclins, which act in G1, have emerged as crucial targets in metabolism–cell cycle crosstalk10–14. Overexpression of nuclear 6‑phosphofructo‑2‑kinase/fructose‑2,6‑ bisphosphatase isoenzyme 3 (PFKFB3) was recently shown to promote proliferation through its effects on central cell cycle regulators11. PFKFBs are bifunctional enzymes that interconvert fructose‑6‑phosphate and fructose‑2,6‑bisphosphate (Fru‑2,6‑BP), the latter of which is an essential co-activator of the rate-limiting glycolytic enzyme phosphofructokinase 1. Accordingly, cytoplasmic PFKFB3, which is an isoform with a high kinase to phosphatase activity ratio (740/1), seems to have a mainly glycolytic role11,15,16. However, over expression of PFKFB3 results in increased expression of several central cell cycle regulators, including cyclin D3 (REF. 11). In addition to promoting G1 phase by upregu‑ lating cyclin D3, nuclear Fru‑2,6‑BP was also shown to increase the expression of the M phase-promoting phosphatase Cdc25C and decrease the expression of the CDK1 inhibitor p27 (REF. 11). Surprisingly, Fru‑2,6‑BP had direct effects on CDK1-mediated p27 phosphoryla‑ tion, as addition of exogenous metabolite to cell lysates promoted Thr187 phosphorylation, which is a cue for p27 degradation. Levels of Fru‑2,6‑BP may reflect the robustness of glucose uptake and glycolysis, resulting in an accumulation of active cyclin–CDK1 complexes and an efficient drive towards proliferation when glucose is readily available and Fru‑2,6‑BP levels are high11.
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
VOLUME 11 | O CTOBER 2010 | 717 © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS Table 1 | Crosstalk between cell cycle and metabolism Metabolic pathway Proximal metabolic Cell cycle or or cell cycle factor metabolic target
Effector or metabolic outcome
Cell cycle outcome
Glycolysis
Cyclin D3
↑ Cyclin D3 expression
Promotes G1
11
CDK1
↑ CDK1 expression
Promotes mitotic entry
11
Cdc25C
↑ Cdc25C expression
Promotes mitotic entry
PFKFB3
Refs
11
APC–CDH1
21,22
↓ PFKFB3 levels (↑ APC–CDH1mediated degradation)
↓ Glycolysis and ↓ G1 progression
Fructose‑2,6‑ bisphosphate
p27
↓ p27 levels (↑ Thr187 phosphorylation and protein degradation) and↑ CDK1 activity
Promotes G1
11
Cyclin D1
Hexokinase II
↑ Cyclin D1 causes ↓ hexokinase II levels and ↓ glycolysis
NA
12
Pyruvate kinase
Cyclin D1 knockdown causes ↑ pyruvate kinase levels
NA
12
Histones
PFK1 and hexokinase II
↑ PFK1 levels; hexokinase II transcription causes ↑ glycolysis
NA
25
Glucose uptake and glycolysis
ACL
Histones
↑ Acetyl-CoA; promotes global histone acetylation
Promotes S phase
25
Glucose uptake
Intracellular glucose
Cyclin D2
↑ Cyclin D2 transcription (↓ cyclin D2 repressor BCL‑6 levels)
Promotes G1–S transition
14
Histones
GLUT4
↑ GLUT4 transcription causes ↑ insulin-regulated glucose uptake
NA
25
Anaerobic glycolysis
Histones
LDHA
↑ LDHA transcription causes ↑ anaerobic glycolysis
NA
25
Insulin secretion
CDK4
KIR6.2*
↑ CDK4 activity at G1 promotes ↑ insulin release
NA
27
Fatty acid synthesis
Cyclin D1
FAS and ACC
Cyclin D1 knockdown causes ↑ ACC and ↑ FAS
NA
12
Respiration
ATP/AMP imbalance
Cyclin E
SCF-mediated cyclin E degradation
G1–S arrest
18,19
ACC, acetyl-CoA carboxylase; ACL, ATP citrate lyase; APC, anaphase-promoting complex; BCL-6, B cell lymphoma 6; CDK, cyclin-dependent kinase; CDH1, Cdc20 homologue 1; FAS, fatty acid synthase; GLUT4, glucose transporter 4; LDHA, l-lactate dehydrogenase A chain; NA, not applicable; PFK1, phosphofructokinase 1; PFKFB3, 6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase 3; SCF, SKP1–cullin 1–F-box protein. * A component of KATP mitochondrial channels.
However, it is not yet clear on a molecular level how nuclear Fru‑2,6‑BP promotes changes in the abundance and activity of cell cycle regulators. Not only do metabolic intermediates control G1 and S phase progression, cyclins that promote the G1 to S transition can also provide feedback to the meta‑ bolic machinery. For example, cyclin D1 overexpres‑ sion decreases the abundance and activity of the key glycolytic enzyme hexokinase II in breast cancer cells12. Similarly, cyclin D1 knockdown results in increased levels of pyruvate kinase, thereby promoting glyco‑ lysis, and increased levels of FAS and ACC, which are crucial for fatty acid biosynthesis12. Further analysis of gene expression from cyclin D1‑null mice revealed that cyclin D1 inhibits numerous target genes involved in glycolysis, lipogenesis and mitochondrial activity, potentially leading to downregulation of metabolic activity 12. The effects of cyclin D1 outside of the cell cycle have also been studied by profiling cyclin D1 interacting proteins using epitope-tagged cyclin D1 knock-in mice to isolate binding partners13. Although this work primarily focused on a transcriptional role for cyclin D1 in promoting the expression of the trans‑ membrane receptor Notch1, examination of extensive
published data uncovered interactions between cyclin D1 and numerous metabolic proteins, such as FAS and ACC, as well as the mitochondrial electron transport chain components cytochrome c oxidase and ATP syn‑ thase13. The consequences of these interactions have not yet been delineated, but these binding partners suggest that cyclin D1 may play a new part in meta‑ bolic regulation in addition to its duties in driving the cell cycle. Indeed, cyclin D1 expression is high during G1 phase; however, once the G1 to S transition has occurred, cyclin D1 levels drop as DNA replication is initiated17. If metabolic activity increases cyclin D1 expression and cyclin D1 suppresses metabolism, it is attractive to speculate that an increase in cyclin D1 lev‑ els during G1 phase steadily tempers metabolic activ‑ ity, thus supporting a negative feedback loop to help prevent re-initiation of G1 until the G1 to S transition is complete. Furthermore, a drop in cyclin D1 levels during S phase might increase the availability of meta‑ bolic intermediates, providing energy for the crucial tasks of DNA replication and repair. Work in model systems has recently provided addi‑ tional insight into links between metabolism and cyclins– CDKs. For example, in Drosophila melanogaster it
718 | O CTOBER 2010 | VOLUME 11
www.nature.com/reviews/molcellbio © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS has been shown that cyclin E is a crucial effector that ensures G1 to S arrest in the setting of mitochondrial dysfunction and imbalances in the ATP/AMP ratio18,19. Mutation of the cytochrome c oxidase subunit Va, which leads to decreased ATP levels, did not seem to compro‑ mise cell survival, but induced a loss of cyclin E and resulted in cell cycle arrest 19. Subsequent work showed that a decrease in mitochondrial ATP production pro‑ motes downstream transcription of the F‑box protein archipelago through an AMPK- and p53‑dependent pathway 18. Archipelago then recruits cyclin E to the SKP1–cullin 1–F-box protein (SCF) ubiquitin ligase complex, resulting in cyclin E degradation and a stall at the G1 to S transition until ATP levels are restored and cyclin E reaccumulates18. APC substrates in the metabolic cascade. APC is a central regulator of the cell cycle, controlling the ubiq‑ uitylation of key substrates to affect both exit from M phase and transition through G1. Degradation of specific substrates by the APC depends on one of two activators, Cdc20 or Cdc20 homologue 1 (CDH1)20. APC–CDH1 is active from late mitosis and throughout G1 phase and ubiquitylates mitotic substrates for pro‑ teasomal degradation, thus preventing premature entry into S phase and inhibiting proliferation20. Interestingly, it was recently reported that loss of APC–CDH1 activity affects cell proliferation and also enhances glycolysis, suggesting that the APC might control components of metabolic pathways as well as cell cycle regulators21. Indeed, an examination of PFKFB3 dynamics in rat cortical neurons and astrocytes revealed that the dif‑ ferential abundance of PFKFB3 in these cells (detect‑ able in astrocytes, undetectable in cortical neurons) was due to a difference in the activity of APC–CDH1 (low in astrocytes, high in cortical neurons). In fact, this complex was found to be directly responsible for ubiquitylating PFKFB3 (REFS 21,22). Examination of the effects of APC–CDH1 in other cell lines suggested that CDH1 overexpression abrogates glycolytic flux, whereas depletion of CDH1 by small hairpin RNA promotes gly‑ colysis and entry into S phase21. Interestingly, this study confirmed the nuclear localization of PFKFB3 and sug‑ gested that inhibition of APC–CDH1‑mediated nuclear PFKFB3 destruction increases the levels of nuclear and cytoplasmic PFKFB3 and enhances glycolysis and cell proliferation. In addition to linking glycolysis and proliferation in cycling cells, APC–CDH1 seems to provide a crucial link between metabolism and oxidant-induced apop‑ tosis in terminally differentiated neurons by regulat‑ ing PFKFB3 expression 22. As mentioned above, the activity of APC–CDH1 is high in terminally differen‑ tiated neurons to help prevent cell cycle progression. Accordingly, neurons contain undetectable levels of PFKFB3, which is continuously targeted for ubiquityl ation22. This dampening of glycolysis is beneficial to neurons because high levels of glycolysis can move the shared intermediate, G6P, away from the antioxidantproducing PPP and towards glycolysis, thus promoting oxidative stress-induced apoptosis 22. In neurons,
APC–CDH1 helps to maintain a reducing intracellular environment, protecting cells from apoptosis22. These examples highlight the fact that interactions between the cell cycle machinery and metabolic intermediates vary among different cell types. In cycling cells, PFKFB3 promotes proliferation, as evidenced by increased cyclin D3, CDK1 and Cdc25C levels; however, in ter‑ minally differentiated neurons, APC–CDH1 serves primarily as a locus of control to ensure that PFKFB3 is completely degraded so that the cell can funnel glucose into the PPP to promote cell survival11,21,22. Linking chromatin structure to metabolism. DNA must be made accessible to the replication machinery to be replicated in S phase. This can be achieved by global core histone acetylation, which remodels chromatin structure, thereby freeing DNA for replication and transcriptional upregulation and enabling cell cycle progression23,24. ACL has recently emerged as an integral link between glucose metabolism and global histone acetylation25 (BOX 1). Increased glycolysis and pyruvate production results in the generation of the downstream metabolite citrate, which can be used for fatty acid syn‑ thesis when converted to acetyl-CoA in the cytoplasm by ACL26. ACL is crucial for citrate-derived acetyl-CoA production, which, in addition to its use in lipid syn‑ thesis, has recently been shown to act in the nucleus to promote core histone acetylation, global transcriptional upregulation and S phase progression25. These findings reveal an important role for acetyl-CoA availability as determined by changes in glucose uptake, in histone acetyl transferase activity and histone acetylation. Furthermore, ACL-dependent histone acetylation pro‑ motes a positive feedback metabolic circuit by driving the transcription of genes encoding glycolytic proteins, such as phosphofructokinase 1, hexokinase II, l-lactate dehydrogenase A chain (LDHA) and GLUT4 (REF. 25). Therefore, ACL-mediated acetyl-CoA production serves as an inducer of chromatin remodelling, providing an additional means for cells to coordinate proliferation and metabolic flux. Cyclins–CDKs and metabolic crosstalk in β‑cells. Crosstalk between metabolism and cyclins–CDKs is readily evident in cells that are exquisitely sensitive to metabolite levels, such as pancreatic β‑cells. These cells continuously sample extracellular glucose to determine appropriate insulin release and maintain organismal glucose homeostasis. Thus, their proliferation and metabolic machinery must be tightly coupled. A study of pancreatic β‑cells revealed that glucose uptake pro‑ motes cell proliferation mediated by the downstream effector cyclin D2 (REF. 14). Specifically, it was deter‑ mined that glucose activates the upstream targets PI3K and AKT14, which in turn inactivate forkhead box O (FOXO) transcription factors, thus decreasing the expression of the cyclin D2 repressor BCL‑6 (REF. 14). An increase in cyclin D2 transcription supports β‑cell proliferation and provides a link between glucose metabolism and pancreatic islet expansion14. Similarly, crosstalk has also been discovered in β‑cells between
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
VOLUME 11 | O CTOBER 2010 | 719 © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS Box 1 | Effects of ATP citrate lyase on cell cycle and metabolism Glucose and glycolysis Inactive transcription Pro Pyruvate Mitochondrion
Acetyl-CoA Citrate
Lipid synthesis
Histone acetylation
HDACs
HATs Acetyl-CoA
Active transcription
Ac
Ac
Ac
Act Pro Ac
Ac
Ac
Ac
Ac
Ac
Citrate
ACL
Ac
Ac
Ac
Transcripts
PFK1 Hexokinase II GLUT4 LDHA
Global core histone acetylation remodels chromatin structure to increase DNA Reviews Molecularbeen Cell linked Biology replication and transcription. Interestingly, histoneNature acetylation has| recently to metabolism through the function of the enzyme ATP citrate lyase (ACL) (see the figure), an important contributor to fatty acid biosynthesis that acts in the cytoplasm to convert the tricarboxylic acid (TCA) cycle intermediate citrate into the fatty acid synthetic precursor acetyl-CoA. In addition to its role in fatty acid synthesis, acetyl-CoA is also an important source of acetyl for histone acetylation, which results in global transcriptional upregulation and S phase progression. It was recently discovered that ACL activity also links upstream glucose metabolism with downstream replication and transcription, and this signalling pathway also feeds back to metabolism by regulating the transcription of metabolic enzymes. Increased citrate levels from glycolytic flux drive ACL-mediated acetyl-CoA production and histone acetylation, thereby promoting DNA replication and cell cycle progression. Among the downstream transcriptional targets affected by ACL are glucose transporter 4 (GLUT4), hexokinase II, phosphofructokinase 1 (PFK1) and l-lactate dehydrogenase A chain (LDHA), which are further involved in determining the metabolic status of the cell by regulating glucose uptake, glycolysis and anaerobic glycolysis. Therefore, ACL-mediated acetyl-CoA production influences chromatin remodelling and the expression of metabolic genes, helping to coordinate proliferation and metabolism. Act, activator; HAT, histone acetylase; HDAC, histone deacetylase; Pro, promoter.
Initiator caspase A caspase lying at the apex of apoptotic signalling cascades (for example, caspase 2, caspase 8 and caspase 9). These cleave and activate executioner caspases.
Executioner caspase A caspase (caspase 3, caspase 6 and caspase 7) that cleaves a range of cellular substrate proteins, resulting in apoptotic cell death. Also termed effector caspases.
Senescence A cellular state of prolonged G1 cell cycle arrest with characteristic metabolic, morphological and protein expression alterations.
the CDK4–retinoblastoma (RB)–E2F pathway at the G1 to S transition and glucose-mediated insulin secre‑ tion27. In this model, increased CDK4 activity in G1 releases the transcription factor E2F1 from RB repres‑ sion, allowing it to bind target promoters6,27. In β‑cells, E2F1 activates the transcription of Kir6.2 (also known as KCNJ11), the product of which is a subunit of the ATP-sensitive potassium (KATP) mitochondrial channels involved in glucose-mediated insulin release27,28. This study describes the mechanisms linking upstream nutrient-sensing pathways, intermediate cell cycle CDK4–E2F1 activity and downstream endocrine signal‑ ling through insulin release and β‑cell responsiveness. Given the interplay between cell cycle mediators and metabolism in β‑cells, it will be interesting to determine whether similar mechanisms exist in other insulinresponsive cell types, such as skeletal muscle cells and adipocytes.
Powering off: metabolism and apoptosis Apoptosis is a form of programmed cell death that uses the caspase family of Cys proteases to orchestrate sig‑ nal-mediated cell destruction29,30. Following receipt of pro-apoptotic signals, such as DNA damage and oxida‑ tive stress, the cell undergoes rapid and orderly demoli‑ tion that is marked by chromatin condensation, DNA fragmentation and membrane blebbing 29. Membraneenclosed cell components are ultimately engulfed by phagocytes, and the intracellular milieu of the dying cell is never exposed to the extracellular environment 31. Under normal physiological conditions, apoptosis of damaged or unneeded cells is balanced by cell regener ation, thus maintaining proper homeostasis. A defect in the apoptotic process can manifest pathologically as cancer or autoimmunity, whereas excessive apoptosis is seen in neurodegenerative and immunodeficiency disorders29. Canonical apoptosis proceeds through either an extrinsic or intrinsic signalling pathway. In the extrinsic pathway, plasma membrane death receptors are engaged by pro-apoptotic ligands to activate downstream caspasedependent signalling. By contrast, the intrinsic death pathway is initiated by intracellular pro-apoptotic stim‑ uli and involves the convergence of initiator caspases and BCL‑2 family proteins (see below) on mitochondria to promote cytochrome c release from the intermembrane space. Cytochrome c molecules then form a complex with the adaptor protein apoptotic protease-activating factor 1 (APAF1) to recruit and activate pro-caspase 9 in the apoptosome complex 32–34. Active caspase 9 promotes apoptosis by activating executioner caspases, which cause the demise of the cell by cleaving and inactivating sub‑ strates that are important for normal cell function or by cleaving substrates to produce fragments with new functions35,29,30. When glucose is available and metabolic activity is robust, cells can often withstand apoptotic stimuli. Conversely, when nutrients wane, cells may be unable to carry out crucial tasks and become more vulnerable to apoptosis. There are many points when metabolism can impinge on apoptotic signalling pathways; we highlight here some of the important effectors that communicate signals between metabolism and apoptosis. Fuel depletion: connecting p53 to cell death. The tumour suppressor p53 integrates intracellular signals and induces the transcription of target genes to help deter‑ mine cell fate and promote cell cycle arrest, senescence or apoptosis. p53 is itself modified in response to metabo‑ lism; it is phosphorylated by AMPK when intracellu‑ lar glucose levels are low, and this phosphorylation is required for the p53‑mediated G1–S cell cycle arrest that allows cells to recover following a period of low ambi‑ ent glucose36. Sustained activation of AMPK-mediated p53 phosphorylation during conditions of prolonged glucose deprivation can, however, lead to senesence36. p53 has an established role as a pro-apoptotic effec‑ tor in response to genotoxic stress as it upregulates the transcription of pro-apoptotic molecules such as the BCL‑2 homology 3 (BH3)-only proteins PUMA
720 | O CTOBER 2010 | VOLUME 11
www.nature.com/reviews/molcellbio © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS Glucose
GLUT
↓ Glucose Cytosol
p53 Nucleus
↓ PGM
↓ Glucose AMPK
P
p53 Promoter
↑ TIGAR
↑ Glutaminase 2
↑ SCO2
↑ GAMT
↓ Fructose2,6-bisphosphate
↑ Glutamate
↑ Cytochrome c oxidase
↑ Fatty acid oxidation
↑ α-ketoglutarate ↓ Glycolysis ↑ PPP shunt
↓ Glycolysis ↑ PPP shunt
↓ Glycolysis
↓ ROS
↓ ROS
↑ ATP
Plasma membrane
Cell cycle arrest and recovery
p53
↑ PUMA, NOXA, BAX and PIDD
↑ Glutathione ↑ Cytochrome c release susceptibility
↑ ETC function ↓ ROS
↑ ATP
↑ ATP
Apoptosis
Cell cycle
Figure 3 | p53 activation and its effects on metabolism and cellular fate. Active p53 induces the transcription of many Nature Reviews | Molecular Cell Biology target substrates, several of which are closely involved in metabolism. For example, p53 influences metabolism, and consequently cell fate, by upregulating guanidinoacetate N-methyltransferase (GAMT), synthesis of cytochrome c oxidase 2 (SCO2), glutaminase 2 and TP53‑induced glycolysis and apoptosis regulator (TIGAR), and downregulating phosphoglycerate mutase (PGM). Glutaminase 2 converts glutamine to glutamate, which feeds either into the tricarboxylic acid cycle when converted to α‑ketoglutarate, or into glutathione synthesis to help dictate the antioxidant status of the cell. Increased glutaminase 2 expression by p53 therefore results in increased ATP production and decreased levels of reactive oxygen species (ROS), promoting cell cycle progression and protecting the cell from apoptosis. The p53‑mediated regulation of PGM, SCO2 and GAMT also promotes cell cycle progression by decreasing ROS, or increasing ATP, as indicated. Note that GAMT can also cause an increase in ATP through increased fatty acid oxidation to provide sufficient energy for apoptosis. p53 also directly regulates apoptosis by upregulating the expression of the pro-apoptotic proteins PUMA, NOXA, BAX and PIDD. Interestingly, p53‑mediated PUMA expression is suppressed by high glucose, suggesting an additional layer of complexity in the crosstalk between p53, metabolism and apoptosis. AMPK, 5′-AMP-activated protein kinase; ETC, electron transport chain; GLUT, glucose transporter; PPP, pentose phosphate pathway.
(also known as BBC3) and NOXA (also known as PMAIP1), the BCL‑2‑family member BAX and the caspase 2‑activating adaptor protein PIDD (also known as LRDD)37–41. It has also been suggested that p53 acts as a metabolic sensor to control PUMA expression. Consistent with this, p53‑mediated PUMA induction is suppressed in the presence of abundant glucose, and glucose can suppress the induction of PUMA that would normally occur after growth factor withdrawal42. In a transcriptionally independent manner, p53 can also promote mitochondrial outer membrane permeabili‑ zation (MOMP) and apoptosis by forming a complex with the anti-apoptotic proteins BCL-XL and PUMA40,43. The metabolic crosstalk between p53 and metabolism is shown in FIG. 3. In addition to activating cell death proteins, p53 influences metabolic pathways by upregulating the expression of guanidinoacetate N-methyltransferase (GAMT), synthesis of cytochrome c oxidase 2 (SCO2),
glutaminase 2 and TP53‑induced glycolysis and apop‑ tosis regulator (TIGAR), and decreasing the expression of phosphoglycerate mutase (PGM)44–49. p53 modulates creatine biosynthesis during both genotoxic and nutri‑ ent stress by upregulating GAMT44, which converts the glycine metabolite guanidoacetate to creatine for ADP/ ATP energy metabolism and promotes genotoxic- and glucose starvation-induced apoptosis44. Interestingly, upregulation of GAMT by p53 in the setting of glu‑ cose deprivation not only affects creatine biosynthesis but actually provides an alternative source of ATP by increasing fatty acid oxidation. This consequence of GAMT upregulation may be important for providing sufficient ATP to execute apoptosis, particularly under conditions in which energy generation by glycolysis is compromised.44 The effect of p53 on the cytochrome c oxidase complex assembly protein SCO2 also promotes altera‑ tions in cell respiration and oxygen consumption46.
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
VOLUME 11 | O CTOBER 2010 | 721 © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS Inactive caspase 2 Glucose Hexokinase II
Active caspase 2
Glucose-6phosphate
↓ NADPH ↓ CDK1 or cyclin B ↓ Hexokinase II binding ↑ C16:0 PAF
NADPH Ribose-5phosphate
CaMKII
14-3-3 P PP1 Ser135 Caspase 2 Ser308 P
14-3-3
Stress PIDD
p53
P PP1 PIDD RAIDD Caspase 2
↑ NADPH ↑ CDK1 or cyclin B ↑ Hexokinase II binding ↓ C16:0 PAF
Cyclin B
BID Cytochrome c release
CDK1 Mitosis
Caspase 3 and caspase 7 Apoptosis
Figure 4 | Metabolic regulation of caspase 2. The initiator caspase, caspase 2, is intricately involved in the crosstalk between cell metabolism and apoptosis. Caspase 2 Nature Reviews | Molecular Cell Biology is inhibited in response to pentose phosphate pathway glucose flux and abundant NADPH, thereby preventing cell death. Caspase 2 inhibition by NADPH is mediated by Ca2+/calmodulin-dependent protein kinase II (CaMKII)-directed phosphorylation at Ser135 (in Xenopus laevis), which promotes the association of the phosphorylated Serbinding protein 14‑3‑3. As NADPH levels wane, 14‑3‑3 is released from caspase 2, leaving the inhibitory phosphorylation vulnerable to removal by the constitutively bound and active protein phosphatase 1 (PP1). Once dephosphorylated, caspase 2 is poised for activation by induced proximity oligomerization through its adaptor proteins RIP-associated protein with a death domain (RAIDD) and PIDD. In addition to NADPH, caspase 2 is also sensitive to the lipid metabolite 1‑O-hexadecyl‑2‑acetyl-sn-glycero3‑phosphocholine (C16:0 PAF) and the glycolytic enzyme hexokinase II; caspase 2 is required for C16:0 PAF-dependent neurotoxicity, and caspase 2‑mediated cell death can be enhanced by detachment of hexokinase II from mitochondria. Caspase 2 signalling is suppressed during mitosis by cyclin-dependent kinase 1 (CDK1) through inhibitory phosphorylation at Ser308 (in X. laevis). Thus, caspase 2 has emerged as a central integrator of cell metabolism and apoptosis that is also responsive to cell cycle status. BID, BH3-interacting domain death agonist.
p53 increases SCO2 expression and upregulates aerobic mitochondrial respiration; conversely, loss of p53 com‑ promises cytochrome c oxidase function and supports a switch to glycolytic metabolism, which contributes to the metabolic phenotype observed in cancer cells, in which glycolytic pathways are used for ATP generation and aerobic respiration is downregulated46. p53 also influences glucose flux by regulating the expression of PGM and TIGAR. p53 activity decreases the expression of PGM, which produces the glycolytic substrate 2‑phosphogylcerate; this results in a drop in glycolysis and entry into senescence45. Thus, loss of functional p53, as is common in cancer cells, might drive increased PGM expression, thereby propelling aerobic glycolysis45. This change promotes glycolysis and the PPP and renders cells resistant to oxidative stress in the absence of p53, thereby forestalling senescence. In contrast to the effects of PGM on glycolysis, TIGAR
expression decreases intracellular Fru‑2,6‑BP levels, concomitantly decreasing glycolysis47. As described previously, a decrease in glycolysis allows the diversion of G6P into the PPP, in which it generates reducing equivalents and limits oxidative stress-induced cell death22. Therefore, TIGAR might help the cell to assess the appropriate response to p53‑dependent stres‑ sors by metabolically fine-tuning the cell’s reaction to apoptotic stimuli. Recent publications have delineated interesting con‑ nections between p53, cell metabolism and antioxidant responses48,49. As mentioned above, p53 promotes an increase in glutaminase 2 expression, which drives the conversion of glutamine to glutamate. Glutamate can support mitochondrial respiration and ATP produc‑ tion by its interconversion into the TCA cycle substrate α‑ketoglutarate, and can also participate in glutathione synthesis to help regulate the antioxidant status of the cell. By upregulating glutaminase 2, p53 influences both mitochondrial respiration and intracellular levels of reactive oxygen species. Therefore, glutaminase 2 upreg‑ ulation and glutamate production may be mechanisms by which p53 protects against apoptosis by promoting an intracellular environment conducive to recovery from subthreshold stressors48,49. Caspase 2: metabolism and cell death at mitochondria. In the past several years caspase 2 has emerged as a crucial locus for communication between metabolic and cell death pathways upstream of mitochondria. In response to specific stimuli, such as DNA damage and heat shock, caspase 2 is activated by induced proximity oligomerization mediated by adaptor proteins 50,51. Following oligomerization, caspase 2 is autocatalytically processed to amplify its enzymatic activity and can then cleave the pro-apoptotic protein BH3‑interacting domain death agonist (BID) to promote mitochondrial cytochrome c release and cell death50,51. Caspase 2 was known to be important for pro‑ grammed oocyte death during mouse development, and this observation has been extended to Xenopus laevis oocytes, in which caspase 2 has been extensively charac‑ terized as a metabolism-regulated apoptotic effector 52–54 (FIG. 4). X. laevis egg cell-free extracts spontaneously undergo apoptosis, marked by successive initiator and executioner caspase activation and cytochrome c release, when they are incubated at room temperature over time55. Apoptosis in this case is initiated by the depletion of a key PPP byproduct, NADPH53, which is primarily produced by glucose flux through the PPP. NADPH is an important reducing agent in the cell and is also involved in the reductive biosynthesis of fatty acids and nucleotides. An endogenous drop in NADPH levels over time promotes caspase 2 activation and downstream apoptosis, and caspase 2 activation is inhib‑ ited by supplementing the extract with either G6P, to ectopically increase NADPH levels through the PPP, or malate, which produces NADPH through a non-PPP mechanism (through the malic enzyme (also known as malate dehydrogenase))53. High PPP flux and abun‑ dant NADPH production is communicated to effector
722 | O CTOBER 2010 | VOLUME 11
www.nature.com/reviews/molcellbio © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS
Bimolecular fluorescence complementation A method for detecting protein–protein interactions using non-fluorescent protein halves fused to the proteins of interest. When the proteins of interest interact, the non-fluorescent halves associate to form a fluorescent complex.
caspase 2 through Ca2+/calmodulin-dependent protein kinase II (CaMKII), which phosphorylates caspase 2 at Ser135 to prevent binding of the caspase 2 adaptor RIP-associated protein with a death domain (RAIDD; also known as CRADD) and thus caspase 2 activation53. Therefore, NADPH and glucose flux through the PPP dictate the viability of X. laevis oocytes by maintaining caspase 2 in an inhibitory state when nutrient levels are high53. Interestingly, caspase 2 activation in the setting of low NADPH levels is also regulated by metabolism in X. laevis egg extracts54. As PPP metabolism slows down, NADPH begins to drop, and apoptosis occurs following caspase 2 dephosphorylation53,54. It was shown that caspase 2 activation depends on the metabolismregulated release of the phosphorylated Ser-binding pro‑ tein 14‑3‑3 (REF. 54). 14‑3‑3 release from caspase 2 allows the inhibitory phosphate on Ser135 to be removed by the constitutively active protein phosphatase 1 (PP1), thus freeing caspase 2 for induced proximity activation54. In the presence of excess G6P to prevent a decrease in NADPH levels, 14‑3‑3 remains bound to phosphorylated caspase 2, resulting in inhibition of caspase 2 activation and apoptosis54. The mechanism of 14‑3‑3 removal by changes in metabolic inter mediates is being investigated, and elucidation of these details will help to complete the story of how caspase 2 is regulated by NADPH. An exciting extension of the work on NADPH and caspase 2 recently showed that caspase 2 activity, as visualized by real-time oligomerization, might also be regulated by metabolism56. This study used bimolecular fluorescence complementation to examine induced proximity oligomerization of recombinant caspase 2 constructs fused to non-fluorescent halves of Venus fluorescent protein (VFP)56. Treatment of cells with dehydroepiandrosterone (DHEA), which inhibits the rate-limiting enzyme of the PPP and leads to a decrease in NADPH, induced caspase 2 oligomerization as observed by VFP fluorescence visualization, further revealing a role for the PPP in caspase 2 activation56. Caspase 2 is also involved in the crosstalk between cell cycle and apoptotic machinery in X. laevis egg extracts57. Caspase 2 is directly suppressed during mitosis by the M phase-promoting kinase CDK1 (REFS 6,57). CDK1–cyclin B1 phosphorylates caspase 2 at Ser308 in X. laevis, thus inhibiting its activation and ensuring that caspase 2‑mediated apoptosis will not be aberrantly initiated during normal mitosis57. In investigating PPPand mitosis-mediated caspase 2 suppression, it emerged that abundant NADPH was sufficient to prevent caspase 2 activation in mitotic X. laevis egg extracts57. Furthermore, high CDK1–cyclin B1 activity was also sufficient to prevent metabolism-mediated caspase 2 activation57. This detailed crosstalk between meta bolism, mitosis and apoptosis in X. laevis egg extracts suggests that caspase 2 is a central metabolic effector in the decision of whether to proliferate or die. In addition to its role as an integrator of the PPP and mitotic signals, caspase 2 potentiates apoptosis in the setting of glucose and lipid signalling 58,59. Caspase 2
modulates mitochondrial activity and the apoptotic threshold by cooperating with hexokinase II at mito‑ chondria58. Hexokinase II catalyses the conversion of glucose to G6P, which feeds into both glycolysis and the PPP, and the interaction between hexokinase II and the mitochondrial voltage-dependent anion channel (VDAC) is important for mitochondrial function and cell survival60. Detachment of hexokinase II from mito‑ chondria enhances caspase 2‑dependent apoptosis that is induced by the platinum-based chemotherapeutic cisplatin, suggesting an interplay between caspase 2, BCL‑2 family members and metabolic machinery at the mitochondria 58. In another interesting para‑ digm, alterations in lipid signalling, and specifically in the choline-containing lipid alkylacylglycerophospho‑ choline, was shown to promote caspase 2‑mediated apoptosis in a β‑amyloid oligomer model of Alzheimer’s disease59,61. Using an unbiased lipidomics approach, it was determined that metabolic alterations in the alkylacylglycerophosphocholine second messenger 1‑ O-hexadecyl‑2‑acetyl-sn-glycero‑3‑phosphocholine (C16:0 PAF) signal for cell death in response to toxic β‑amyloid oligomers59. C16:0 PAF mediated neurotoxic‑ ity was also shown to depend on caspase 2, as caspase 2 inhibition protects neurons from β‑amyloid- and C16:0 PAF-induced apoptosis59,62. This new link between lipid metabolism and caspase 2 mediated β‑amyloid neuro toxicity opens the possibility of using metabolismtargeted treatments to modulate the course of Alzheimer’s disease as well as that of other caspase 2‑dependent neurodegenerative disorders. Given the apparent role of caspase 2 in integrating metabolic signals and progression of apoptosis, it is inter‑ esting that the only obvious phenotype of caspase 2 defi‑ cient mice is an excess of oocytes (indicative of a failure of apoptosis in these cells)52. This may be because cas‑ pase 2 is more important for cell death in response to cell stress than in response to developmental cues or because of compensatory upregulation of other caspases in the caspase 2‑null setting. However, given the poten‑ tially important role of caspase 2 in linking metabolism and cell death, an alternative explanation is that oocytes depend more on stockpiled internal energy stores than other cells in the body, and it is advantageous to imme‑ diately eliminate any oocytes that are nutrient deficient as they would not have sufficient energy stores to suc‑ cessfully undergo fertilization and initial cell divisions. Thus, the brake on apoptosis in oocytes may primarily depend on the metabolic suppression of caspase 2 in a way that does not occur in cells that use glucose as their primary source of energy. BCL‑2 family proteins: balancing metabolism and death. Mitochondria use downstream metabolites from cytosolic pathways, such as glycolysis and the PPP, in addition to housing the TCA cycle, glutamine metabo‑ lism, fatty acid β‑oxidation and oxidative phosphoryla‑ tion. Importantly, mitochondria are also responsible for integrating cell death signals to initiate cytochrome c release from their intermembrane space, triggering apoptosis. Mitochondria are therefore prime candidates
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
VOLUME 11 | O CTOBER 2010 | 723 © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS Table 2 | Crosstalk between BCL‑2 family members and metabolism Metabolic pathway
Proximal metabolic factor or BCL‑2 protein
BCL‑2 protein or metabolic target
Effector or metabolic outcome
Apototic outcome
Refs
Growth factor stimulation and glucose uptake
GSK3
MCL1
AKT inhibits GSK3-mediated MCL1 ubiquitylation and degradation
Anti-apoptotic
76,77
AKT and mitochondrionassociated PKA
BAD
BAD phosphorylation and inhibition
Anti-apoptotic
65–68
Ceramide
BCL‑2
↑ Ceramide causes PP2A-mediated BCL‑2 dephosphorylation
Pro-apoptotic
98
Ceramide
BAK
Ceramides cooperate with BAK to promote MOMP
Pro-apoptotic
99
BAK
Long-chain ceramides
↑ Long-chain ceramide production
Pro-apoptotic
100
Glucose-responsive p53
PUMA
↓ PUMA expression causes ↓ BAX activation
Anti-apoptotic
42
AKT
BAX
↓ BAX activity (BAX held in inactive conformation)
Anti-apoptotic
72,73
Glucose deprivation
MCL1–NOXA interaction
NOXA
↓ MCL1 causes ↑ NOXA-induced cell death
Pro-apoptotic
71
Glycolysis
Hexokinase II
BAX
Hexokinase II binding to VDAC limits BAX’s ability to promote MOMP
Anti-apoptotic
60
Glycolysis and mitochondrial respiration
BAD
Glucokinase
↓ BAD causes ↓ glucokinase activity and ATP production
Anti-apoptotic
70
Sphingolipid metabolism
Glucose uptake and glycolysis
BAD, BCL-2 antagonist of cell death; BAK, BCL-2 antagonist/killer; BCL-2, B cell lymphoma 2; GSK3, glycogen synthase kinase 3; MCL1, myeloid leukaemia cell differentiation 1; MOMP, mitochondrial outer membrane permeabilization; PKA, protein kinase A; PP2A, protein phosphatase 2A; VDAC, voltage-dependent anion channel.
for regulating metabolic and apoptotic crosstalk. In particular, BCL‑2 family proteins dynamically interact with mitochondria, and the balance between pro- and anti-apoptotic BCL‑2 proteins is a central determinant of downstream caspase activation. Intriguingly, BCL‑2 family members are extensively regulated by nutrients to create a metabolic threshold for MOMP, a process that is required for the activation of downstream caspases in the intrinsic apoptotic pathway 63 (TABLE 2). Furthermore, BCL‑2 family proteins have also been shown to have a role in modulating metabolism and in particular in mitochondrial respiration. The BCL‑2 family members are generally classi‑ fied as either the anti-apoptotic BCL‑2, BCL-XL and myeloid leukaemia cell differentiation 1 (MCL1) pro‑ teins, the pro-apoptotic BAX and BCL‑2 antagonist/ killer (BAK) proteins, or the pro-apoptotic BH3‑only proteins, which include BID, BCL‑2 antagonist of cell death (BAD), BCL‑2‑interacting mediator of cell death (BIM), PUMA and NOXA64. Anti-apoptotic BCL‑2, BCL-XL and MCL1 interact with and inhibit the activity of BAX and BAK, thereby preventing cytochrome c release. When active, the pro-apoptotic BH3‑only pro‑ teins sequester BCL‑2, BCL-XL and MCL1 from BAX and BAK, leaving them free to interact with mitochon‑ drial membrane proteins and induce MOMP. In addi‑ tion, some BH3‑only proteins, such as BID and BIM, also serve as direct activators of BAK and BAX (for a general BCL‑2 family review, see REF. 64). We touch briefly here on the crosstalk between metabolism and BCL‑2 proteins. Pro-apoptotic BH3‑only proteins such as BAD have emerged as central metabolic effectors that dictate downstream BCL‑2 family protein activation. Growth
factor stimulation (for example, by interleukin‑3 (IL‑3)) results in increased glucose uptake and metabolism as well as AKT-mediated inhibitory BAD phosphorylation, which protects cells exposed to enough nutrients from undergoing apoptosis65–67. Interestingly, BAD is also phosphorylated and inhibited in response to IL‑3 by mitochondrion-associated protein kinase A (PKA)68. Dissociation of PKA from mitochondria disrupts the PKA–BAD interaction, promotes BAD hypophosphoryl ation and is thought to sensitize the cell to apoptosis68. Indeed, BAD phosphorylation is crucial for its antiapoptotic functions, as replacement of endogenous BAD with a non-phosphorylatable variant reduces the threshold for MOMP69. BAD itself can influence mito‑ chondrial metabolism by interacting with a membraneassociated holoenzyme containing glucokinase, which catalyses the first committed step in glycolysis70. BAD is required for holoenzyme assembly and efficient gluco kinase activity, as Bad–/– hepatocytes show decreased glucokinase function and compromised mitochondrial respiration70. Parodoxically, growth factor stimulation of T cells, which promotes glucose uptake and metabolism, has also been shown to upregulate the pro-apoptotic BH3‑only protein NOXA71. When glucose is readily available, the anti-apoptotic BCL‑2 protein MCL1 binds and inhibits NOXA, whereas the expression of NOXA renders these cells poised for death in the event of glucose insufficiency 71. The authors of this study termed this NOXA–MCL1 pair a glucose-sensitive apoptotic ‘rheostat’: when glucose levels are low, MCL1 degradation is accelerated, thereby liberating the preexisting pool of NOXA to trigger mitochondrial cyto‑ chrome c release and apoptosis, whereas when there is
724 | O CTOBER 2010 | VOLUME 11
www.nature.com/reviews/molcellbio © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS sufficient glucose, NOXA is inhibited by MCL1 (REF. 71). Loss of glucose uptake following growth factor with‑ drawal has also been shown to induce the expression of the p53‑regulated, pro-apoptotic BH3‑only protein PUMA42. Interestingly, maintenance of glucose uptake, even in the absence of growth factor stimulation, is suf‑ ficient to downregulate PUMA levels, thus abrogating BAX activation and downstream caspase induction42. These studies suggest that PUMA, which is induced by glucose-responsive p53, is a metabolism-regulated apoptotic effector 42 (FIG. 3). Pro-apoptotic BAX has also been shown to be directly regulated by glucose levels, as BAX is maintained in an inactive conformation by AKT signalling in the presence of glucose, thereby preventing apoptosis72,73. Apoptosis is further linked to metabolism by the interaction between hexokinase II and VDAC, which is thought to prevent cytochrome c release both in the presence and absence of BAX or BAK signalling 60,74,75. The anti-apoptotic proteins BCL‑2, BCL-XL and MCL1 have also been shown to receive metabolic signals and influence mitochondrial physiology; for example, MCL1 is also regulated by the AKT signalling axis to promote cell survival76,77. When active, the AKT substrate glycogen synthase kinase 3 (GSK3) phos‑ phorylates MCL1, targeting it for ubiquitylation and proteasomal degradation, thus predisposing the cell to cytochrome c release76,77. In response to growth factor stimulation, AKT inhibits GSK3, thereby preserving MCL1 levels and contributing to cell survival76,77. The interplay detailed above between metabolic and apop‑ totic proteins localized to the mitochondria raises the question of whether additional crosstalk at the level of mitochondria might remain to be discovered. As an interesting side note, rho-zero (ρ0) cells, which lack mitochondrial DNA, die by apoptosis despite the absence of respiratory chain function. Moreover, this apoptosis can be inhibited by BCL‑2 expression, high‑ lighting the fact that the survival function of BCL‑2 does not depend on mitochondrial respiration 78. Analysis of mouse embryos deficient for transcrip‑ tion factor A mitochondrial (TFAM) or mouse hearts with a tissue-specific TFAM deficiency, revealed that cells lacking mitochondrial respiration in vivo are even more susceptible to apoptotic stimuli than their normal counterparts79.
Autophagy A process in which intracellular contents are destroyed by bulk enclosure of cytoplasmic material in membrane-enclosed vesicles that are then targeted for lysosomal degradation.
Cytochrome c: at the metabolism–apoptosis crossroads. Cytochrome c is at the crossroads of energy generation and the apoptotic cascade80–82. Thus, it is not difficult to imagine a scenario in which upstream metabolic messages affect electron transport chain fitness and modulate cell death signalling through cyto‑ chrome c-dependent pathways. In fact, cells that rely heavily on glucose utilization, such as neurons and can‑ cer cells, have been shown to inhibit the pro-apoptotic functions of cytochrome c in a glucose-dependent manner 83–85. The pro-apoptotic activity of cytochrome c is influenced by its redox state, with reduced cyto‑ chrome c lacking the ability to promote cell death85. Glucose flux through the PPP to produce abundant
NADPH is crucial in determining the redox state of cytochrome c, and inhibition of the PPP by DHEA or incubation of cells in glucose-free media sensitizes the cells to cytochrome c-induced apoptosis85. This mecha‑ nism couples glucose flux through the PPP with the potency of pro-apoptotic cytochrome c, providing an additional layer of crosstalk between metabolism and cell death. Outside of the canonical intrinsic pathway of apop‑ tosis, glycolysis and glyceraldehyde‑3‑phosphate de hydrogenase (GAPDH) activity can protect cells from caspase-independent cell death, a type of cell death that can occur following MOMP even when downstream caspases are inhibited86–88. In CICD, caspase inhibition is not sufficient to prevent MOMP-induced cell death, most likely owing to compromised mitochondrial membrane potential and the release of pro-apoptotic mediators from the intermembrane space88,89. MOMP is typically accompanied by a loss of GAPDH levels and decreased glycolysis and membrane potential86,90. However, overexpression of GAPDH was found to rescue cell death following MOMP and cytochrome c release, not only by inducing ATP production, but also by triggering autophagy, presumably resulting in the removal of damaged mitochondria86. Thus, upregu‑ lation of GAPDH following mitochondrial damage may provide a cell with the ability to recover following MOMP86,91. Additional studies have demonstrated the importance of mitochondrial respiratory chain dysfunc‑ tion, rather than release of pro-apoptotic mitochondrial constituents, in promoting MOMP-induced caspaseindependent cell death92. Specifically, MOMP results in progressive loss of complex I and complex IV activities by 8 hours after mitochondrial permeabilization, with a concomitant decrease in oxidative phosphorylation. These events lead to cell death despite the absence of caspase activity 92. Evidence of crosstalk between metabolism and the apoptosome outside of cytochrome c regulation is limited, although it is certainly plausible that inhibiting executioner caspase activity in conditions of nutrient abundance would contribute to cell survival. Despite the paucity of evidence for regulation of the apoptosome by metabolism at the post-translational level, some interesting work has emerged that suggests an inter‑ play between lipid metabolism and caspase 9 regulation at the post-transcriptional level93. The caspase 9 splice variant caspase 9b has an inhibitory role in the apop‑ totic cascade in contrast to the canonical pro-apoptotic caspase 9 isoform93,94. Ceramides are lipid molecules involved in bilayer cell membrane structure, as well as various cell signalling pathways, and are known to induce apoptosis when present in excess. Excess cera‑ mide promotes a decrease in the caspase 9b to caspase 9 ratio by affecting the proteins responsible for alterna‑ tive splicing, thus sensitizing cells to apoptosis by the intrinsic pathway 93. This crosstalk between lipid mol‑ ecules and caspase 9 mRNA represents a mechanism by which ceramide-producing death stimuli sensitize the cell to apoptosis and communicate pro-death signals by a metabolic messenger 93.
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
VOLUME 11 | O CTOBER 2010 | 725 © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS Conclusion and perspectives The metabolic signalling networks communicating with cell cycle and apoptotic machinery are complex and have crucial roles in determining the fitness of the cell. It is likely that most metabolic pathways impinge at some point on the machinery that controls cell pro‑ liferation and death. We have not in this Review been able to touch on all of the many metabolism-regulated factors that modulate effectors of cell proliferation and cell death. For example, there is an emerging literature on sirtuins (a class of proteins that have either histone deacetylase or mono-ribosyltransferase activity) as potential bridges between metabolism and the apop‑ totic machinery, and continued characterization of new microRNAs is also likely to reveal additional means of linking metabolism with cell death and division tar‑ gets (REFS 95,96). Moreover, broad-based genetic, RNA
1. 2.
3. 4. 5. 6. 7. 8. 9. 10. 11.
12.
13. 14.
15.
16. 17.
Salati, L. M. & Amir-Ahmady, B. Dietary regulation of expression of glucose‑6‑phosphate dehydrogenase. Annu. Rev. Nutr. 21, 121–140 (2001). Engelman, J. A., Luo, J. & Cantley, L. C. The evolution of phosphatidylinositol 3‑kinases as regulators of growth and metabolism. Nature Rev. Genet. 7, 606–619 (2006). Manning, B. D. & Cantley, L. C. AKT/PKB signaling: navigating downstream. Cell 129, 1261–1274 (2007). Hardie, D. G. AMP-activated/SNF1 protein kinases: conserved guardians of cellular energy. Nature Rev. Mol. Cell Biol. 8, 774–785 (2007). Wullschleger, S., Loewith, R. & Hall, M. N. TOR signaling in growth and metabolism. Cell 124, 471–484 (2006). Nurse, P. A long twentieth century of the cell cycle and beyond. Cell 100, 71–78 (2000). Sherr, C. J. G1 phase progression: cycling on cue. Cell 79, 551–555 (1994). Coller, H. A., Sang, L. & Roberts, J. M. A new description of cellular quiescence. PLoS Biol. 4, e83 (2006). Zwerschke, W. et al. Metabolic analysis of senescent human fibroblasts reveals a role for AMP in cellular senescence. Biochem. J. 376, 403–411 (2003). Sherr, C. J. D‑type cyclins. Trends Biochem. Sci. 20, 187–190 (1995). Yalcin, A. et al. Nuclear targeting of 6‑phosphofructo‑2‑kinase (PFKFB3) increases proliferation via cyclin-dependent kinases. J. Biol. Chem. 284, 24223–24232 (2009). This work shows that unlike other PFKFB isoforms, PFKFB3 is localized to nuclei and that this nuclear localization is required for its ability to control cell cycle progression by altering the expression of multiple cell cycle regulators. Sakamaki, T. et al. Cyclin D1 determines mitochondrial function in vivo. Mol. Cell. Biol. 26, 5449–5469 (2006). Shows that cyclin D1 also suppresses mitochondrial function and aerobic glycolysis by decreasing the expression of many metabolic enzymes. Bienvenu, F. et al. Transcriptional role of cyclin D1 in development revealed by a genetic-proteomic screen. Nature 463, 374–378. Glauser, D. A. & Schlegel, W. The FoxO/Bcl‑6/cyclin D2 pathway mediates metabolic and growth factor stimulation of proliferation in Min6 pancreatic β-cells. J. Recept Signal Transduct Res. 29, 293–298 (2009). Chesney, J. et al. An inducible gene product for 6‑phosphofructo‑2‑kinase with an AU‑rich instability element: role in tumor cell glycolysis and the Warburg effect. Proc. Natl Acad. Sci. USA 96, 3047–3052 (1999). Okar, D. A. & Lange, A. J. Fructose‑2, 6‑bisphosphate and control of carbohydrate metabolism in eukaryotes. Biofactors 10, 1–14 (1999). Yang, K., Hitomi, M. & Stacey, D. W. Variations in cyclin D1 levels through the cell cycle determine the proliferative fate of a cell. Cell Div. 1, 32 (2006).
interference and proteomic screens are likely to reveal new interactions between the cell cycle, the apoptotic and the metabolic machinery. For example, a recent genome-wide screen for genes important in DNA damage-induced apoptosis in D. melanogaster S2 cells identified many proteins involved in metabolic regula‑ tion, although it remains to be determined where and how they impinge on cell death pathways97. Agents that can alter pathways linking metabolism and cell pro‑ liferation and/or cell death, identified both through high-throughput small molecule screening and rational drug design, may become viable therapeutic agents for altering the course of metabolism-driven diseases. The potential for usurping metabolic crosstalk to alter disease progression is exciting, and a complete understanding of metabolic effectors will help to reach this goal.
18. Mandal, S., Freije, W. A., Guptan, P. & Banerjee, U. Metabolic control of G1‑S transition: cyclin E degradation by p53‑induced activation of the ubiquitin-proteasome system. J. Cell Biol. 188, 473–479 (2010). 19. Mandal, S., Guptan, P., Owusu-Ansah, E. & Banerjee, U. Mitochondrial regulation of cell cycle progression during development as revealed by the tenured mutation in Drosophila. Dev. Cell 9, 843–854 (2005). 20. Li, M. & Zhang, P. The function of APC/CCdh1 in cell cycle and beyond. Cell Div. 4, 2 (2009). 21. Almeida, A., Bolanos, J. P. & Moncada, S. E3 ubiquitin ligase APC/C-Cdh1 accounts for the Warburg effect by linking glycolysis to cell proliferation. Proc. Natl Acad. Sci. USA 107, 738–741. Reports that PFKFB3 is a substrate of the APC and that glycolysis and cell proliferation are therefore dependent on a drop in APC activity. 22. Herrero-Mendez, A. et al. The bioenergetic and antioxidant status of neurons is controlled by continuous degradation of a key glycolytic enzyme by APC/C-Cdh1. Nature Cell Biol. 11, 747–752 (2009). 23. Li, B., Carey, M. & Workman, J. L. The role of chromatin during transcription. Cell 128, 707–719 (2007). 24. Knosp, O., Talasz, H. & Puschendorf, B. Histone acetylation and histone synthesis in mouse fibroblasts during quiescence and restimulation into S‑phase. Mol. Cell Biochem. 101, 51–58 (1991). 25. Wellen, K. E. et al. ATP-citrate lyase links cellular metabolism to histone acetylation. Science 324, 1076–1080 (2009). Shows that acetyl-CoA generated by ACL is required for histone acetylation; control of ACL activity by glucose uptake therefore allows metabolic input into the histone acetylation required for DNA replication and transcription following growth factor stimulation. 26. Kaplan, R. S., Mayor, J. A. & Wood, D. O. The mitochondrial tricarboxylate transport protein. cDNA cloning, primary structure, and comparison with other mitochondrial transport proteins. J. Biol. Chem. 268, 13682–13690 (1993). 27. Annicotte, J. S. et al. The CDK4‑pRB‑E2F1 pathway controls insulin secretion. Nature Cell Biol. 11, 1017–1023 (2009). 28. Fajas, L. et al. Impaired pancreatic growth, β cell mass, and beta cell function in E2F1–/– mice. J. Clin. Invest. 113, 1288–95 (2004). 29. Danial, N. N. & Korsmeyer, S. J. Cell death: critical control points. Cell 116, 205–219 (2004). 30. Nicholson, D. W. Caspase structure, proteolytic substrates, and function during apoptotic cell death. Cell Death Differ. 6, 1028–1042 (1999). 31. Taylor, R. C., Cullen, S. P. & Martin, S. J. Apoptosis: controlled demolition at the cellular level. Nature Rev. Mol. Cell Biol. 9, 231–241 (2008). 32. Acehan, D. et al. Three-dimensional structure of the apoptosome: implications for assembly, procaspase‑9 binding, and activation. Mol. Cell 9, 423–432 (2002).
726 | O CTOBER 2010 | VOLUME 11
33. Li, P. et al. Cytochrome c and dATP-dependent formation of Apaf‑1/caspase‑9 complex initiates an apoptotic protease cascade. Cell 91, 479–489 (1997). 34. Schafer, Z. T. & Kornbluth, S. The apoptosome: physiological, developmental, and pathological modes of regulation. Dev. Cell 10, 549–561 (2006). 35. Zou, H., Henzel, W. J., Liu, X., Lutschg, A. & Wang, X. Apaf‑1, a human protein homologous to C. elegans CED‑4, participates in cytochrome c-dependent activation of caspase‑3. Cell 90, 405–413 (1997). 36. Jones, R. G. et al. AMP-activated protein kinase induces a p53‑dependent metabolic checkpoint. Mol. Cell 18, 283–293 (2005). 37. Oda, E. et al. Noxa, a BH3‑only member of the Bcl‑2 family and candidate mediator of p53‑induced apoptosis. Science 288, 1053–1058 (2000). 38. Miyashita, T. & Reed, J. C. Tumor suppressor p53 is a direct transcriptional activator of the human bax gene. Cell 80, 293–299 (1995). 39. Nakano, K. & Vousden, K. H. PUMA, a novel proapoptotic gene, is induced by p53. Mol. Cell 7, 683–694 (2001). 40. Chipuk, J. E., Bouchier-Hayes, L., Kuwana, T., Newmeyer, D. D. & Green, D. R. PUMA couples the nuclear and cytoplasmic proapoptotic function of p53. Science 309, 1732–1735 (2005). 41. Tinel, A. & Tschopp, J. The PIDDosome, a protein complex implicated in activation of caspase‑2 in response to genotoxic stress. Science 304, 843–846 (2004). 42. Zhao, Y. et al. Glucose metabolism attenuates p53 and Puma-dependent cell death upon growth factor deprivation. J. Biol. Chem. 283, 36344–36353 (2008). Shows that cell death induced by PUMA is linked to nutrient levels, as high glucose metabolism can suppress p53‑mediated PUMA induction even under conditions of growth factor withdrawal. 43. Mihara, M. et al. p53 has a direct apoptogenic role at the mitochondria. Mol. Cell 11, 577–590 (2003). 44. Ide, T. et al. GAMT, a p53‑inducible modulator of apoptosis, is critical for the adaptive response to nutrient stress. Mol. Cell 36, 379–392 (2009). 45. Kondoh, H. et al. Glycolytic enzymes can modulate cellular life span. Cancer Res. 65, 177–185 (2005). 46. Matoba, S. et al. p53 regulates mitochondrial respiration. Science 312, 1650–1653 (2006). 47. Bensaad, K. et al. TIGAR, a p53‑inducible regulator of glycolysis and apoptosis. Cell 126, 107–120 (2006). Identifies a new p53‑inducible gene, TIGAR, that can decrease intracellular levels of Fru‑2,6‑BP, thereby dampening glycolysis and protecting cells from reactive oxygen species-induced apoptosis through PPP-mediated generation of NADPH. 48. Hu, W. et al. Glutaminase 2, a novel p53 target gene regulating energy metabolism and antioxidant function. Proc. Natl Acad. Sci. USA 107, 7455–7460 (2010). 49. Suzuki, S. et al. Phosphate-activated glutaminase (GLS2), a p53‑inducible regulator of glutamine metabolism and reactive oxygen species. Proc. Natl Acad. Sci. USA 107, 7461–7466 (2010).
www.nature.com/reviews/molcellbio © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS 50. Baliga, B. C., Read, S. H. & Kumar, S. The biochemical mechanism of caspase‑2 activation. Cell Death Differ. 11, 1234–1241 (2004). 51. Krumschnabel, G., Manzl, C. & Villunger, A. Caspase‑2: killer, savior and safeguard — emerging versatile roles for an ill-defined caspase. Oncogene 28, 3093–3096 (2009). 52. Bergeron, L. et al. Defects in regulation of apoptosis in caspase‑2‑deficient mice. Genes Dev. 12, 1304–1314 (1998). 53. Nutt, L. K. et al. Metabolic regulation of oocyte cell death through the CaMKII-mediated phosphorylation of caspase‑2. Cell 123, 89–103 (2005). Shows that nutrients sufficient to promote the generation of NADPH through the PPP can promote NADPH-dependent suppressive phosphorylation of caspase 2 by CaMKII. 54. Nutt, L. K. et al. Metabolic control of oocyte apoptosis mediated by 14‑3‑3ζ-regulated dephosphorylation of caspase‑2. Dev. Cell 16, 856–866 (2009). 55. Newmeyer, D. D., Farschon, D. M. & Reed, J. C. Cell-free apoptosis in Xenopus egg extracts: inhibition by Bcl‑2 and requirement for an organelle fraction enriched in mitochondria. Cell 79, 353–364 (1994). 56. Bouchier-Hayes, L. et al. Characterization of cytoplasmic caspase‑2 activation by induced proximity. Mol. Cell 35, 830–840 (2009). This work employs bimolecular fluorescence complementation to demonstrate cytoplasmic activation of caspase 2 in response to several stimuli, including DHEA, an inhibitor of the PPP. 57. Andersen, J. L. et al. Restraint of apoptosis during mitosis through interdomain phosphorylation of caspase‑2. EMBO J. 28, 3216–3227 (2009). 58. Shulga, N., Wilson-Smith, R. & Pastorino, J. G. Hexokinase II detachment from the mitochondria potentiates cisplatin induced cytotoxicity through a caspase‑2 dependent mechanism. Cell Cycle 8, 3355–3364 (2009). 59. Ryan, S. D. et al. Amyloid-β42 signals tau hyperphosphorylation and compromises neuronal viability by disrupting alkylacylglycerophosphocholine metabolism. Proc. Natl Acad. Sci. USA 106, 20936–20941 (2009). 60. Majewski, N. et al. Hexokinase-mitochondria interaction mediated by Akt is required to inhibit apoptosis in the presence or absence of Bax and Bak. Mol. Cell 16, 819–830 (2004). 61. Klein, J. Membrane breakdown in acute and chronic neurodegeneration: focus on choline-containing phospholipids. J. Neural Transm. 107, 1027–1063 (2000). 62. Troy, C. M. et al. Caspase‑2 mediates neuronal cell death induced by β-amyloid. J. Neurosci. 20, 1386–1392 (2000). 63. Tait, S. W. & Green, D. R. Mitochondria and cell death: outer membrane permeabilization and beyond. Nature Rev. Mol. Cell Biol. 11, 621–632 (2010). 64. Danial, N. N. BCL‑2 family proteins: critical checkpoints of apoptotic cell death. Clin. Cancer Res. 13, 7254–7263 (2007). 65. del Peso, L., Gonzalez-Garcia, M., Page, C., Herrera, R. & Nunez, G. Interleukin‑3‑induced phosphorylation of BAD through the protein kinase Akt. Science 278, 687–689 (1997). 66. Datta, S. R. et al. Akt phosphorylation of BAD couples survival signals to the cell-intrinsic death machinery. Cell 91, 231–241 (1997). 67. Rathmell, J. C., Vander Heiden, M. G., Harris, M. H., Frauwirth, K. A. & Thompson, C. B. In the absence of extrinsic signals, nutrient utilization by lymphocytes
68. 69. 70.
71. 72.
73.
74.
75.
76.
77.
78. 79.
80. 81. 82.
83. 84. 85.
is insufficient to maintain either cell size or viability. Mol. Cell 6, 683–692 (2000). Harada, H. et al. Phosphorylation and inactivation of BAD by mitochondria-anchored protein kinase A. Mol. Cell 3, 413–422 (1999). Datta, S. R. et al. Survival factor-mediated BAD phosphorylation raises the mitochondrial threshold for apoptosis. Dev. Cell 3, 631–643 (2002). Danial, N. N. et al. BAD and glucokinase reside in a mitochondrial complex that integrates glycolysis and apoptosis. Nature 424, 952–956 (2003). Shows that BAD is required for assembly of a glucokinase-containing protein complex necessary both for mitochondrial glucokinase activity and optimal glucose-dependent mitochondrial respiratory function. Alves, N. L. et al. The Noxa/Mcl‑1 axis regulates susceptibility to apoptosis under glucose limitation in dividing T cells. Immunity 24, 703–716 (2006). Rathmell, J. C. et al. Akt-directed glucose metabolism can prevent Bax conformation change and promote growth factor-independent survival. Mol. Cell Biol. 23, 7315–7328 (2003). Yamaguchi, H. & Wang, H. G. The protein kinase PKB/Akt regulates cell survival and apoptosis by inhibiting Bax conformational change. Oncogene 20, 7779–7786 (2001). Pastorino, J. G., Shulga, N. & Hoek, J. B. Mitochondrial binding of hexokinase II inhibits Bax-induced cytochrome c release and apoptosis. J. Biol. Chem. 277, 7610–7618 (2002). Majewski, N., Nogueira, V., Robey, R. B. & Hay, N. Akt inhibits apoptosis downstream of BID cleavage via a glucose-dependent mechanism involving mitochondrial hexokinases. Mol. Cell. Biol. 24, 730–740 (2004). Zhao, Y. et al. Glycogen synthase kinase 3α and 3β mediate a glucose-sensitive antiapoptotic signaling pathway to stabilize Mcl‑1. Mol. Cell. Biol. 27, 4328–4339 (2007). Maurer, U., Charvet, C., Wagman, A. S., Dejardin, E. & Green, D. R. Glycogen synthase kinase‑3 regulates mitochondrial outer membrane permeabilization and apoptosis by destabilization of MCL‑1. Mol. Cell 21, 749–760 (2006). Jacobson, M. D. et al. Bcl‑2 blocks apoptosis in cells lacking mitochondrial DNA. Nature 361, 365–369 (1993). Wang, J., Silva, J. P., Gustafsson, C. M., Rustin, P. & Larsson, N. G. Increased in vivo apoptosis in cells lacking mitochondrial DNA gene expression. Proc. Natl Acad. Sci. USA 98, 4038–4043 (2001). Ow, Y. P., Green, D. R., Hao, Z. & Mak, T. W. Cytochrome c: functions beyond respiration. Nature Rev. Mol. Cell Biol. 9, 532–542 (2008). Wikstrom, M. K. Proton pump coupled to cytochrome c oxidase in mitochondria. Nature 266, 271–273 (1977). Liu, X., Kim, C. N., Yang, J., Jemmerson, R. & Wang, X. Induction of apoptotic program in cell-free extracts: requirement for dATP and cytochrome c. Cell 86, 147–157 (1996). Schubert, D. Glucose metabolism and Alzheimer’s disease. Ageing Res. Rev. 4, 240–257 (2005). Warburg, O. On the origin of cancer cells. Science 123, 309–314 (1956). Vaughn, A. E. & Deshmukh, M. Glucose metabolism inhibits apoptosis in neurons and cancer cells by redox inactivation of cytochrome c. Nature Cell Biol. 10, 1477–1483 (2008).
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
86. Colell, A. et al. GAPDH and autophagy preserve survival after apoptotic cytochrome c release in the absence of caspase activation. Cell 129, 983–997 (2007). Shows that GAPDH supports survival in cells with permeabilized mitochondria by increasing glycolysis and, in a new nuclear role, increasing autophagy. 87. Rathmell, J. C. & Kornbluth, S. Filling a GAP(DH) in caspase-independent cell death. Cell 129, 861–863 (2007). 88. Green, D. R. & Kroemer, G. The pathophysiology of mitochondrial cell death. Science 305, 626–629 (2004). 89. Chipuk, J. E. & Green, D. R. Do inducers of apoptosis trigger caspase-independent cell death? Nature Rev. Mol. Cell Biol. 6, 268–275 (2005). 90. Majeski, A. E. & Dice, J. F. Mechanisms of chaperonemediated autophagy. Int. J. Biochem. Cell Biol. 36, 2435–2444 (2004). 91. Colell, A., Green, D. R. & Ricci, J. E. Novel roles for GAPDH in cell death and carcinogenesis. Cell Death Differ. 16, 1573–1581 (2009). 92. Lartigue, L. et al. Caspase-independent mitochondrial cell death results from loss of respiration, not cytotoxic protein release. Mol. Biol. Cell 20, 4871–4884 (2009). 93. Chalfant, C. E. et al. De novo ceramide regulates the alternative splicing of caspase 9 and Bcl‑x in A549 lung adenocarcinoma cells. Dependence on protein phosphatase‑1. J. Biol. Chem. 277, 12587–12595 (2002). 94. Seol, D. W. & Billiar, T. R. A caspase‑9 variant missing the catalytic site is an endogenous inhibitor of apoptosis. J. Biol. Chem. 274, 2072–2076 (1999). 95. Kim, H. S. et al. SIRT3 is a mitochondria-localized tumor suppressor required for maintenance of mitochondrial integrity and metabolism during stress. Cancer Cell 17, 41–52 (2010). 96. Xu, P., Vernooy, S. Y., Guo, M. & Hay, B. A. The Drosophila microRNA Mir‑14 suppresses cell death and is required for normal fat metabolism. Curr. Biol. 13, 790–795 (2003). 97. Yi, C. H. et al. A genome-wide RNAi screen reveals multiple regulators of caspase activation. J. Cell Biol. 179, 619–626 (2007). 98. Ruvolo, P. P., Deng, X., Ito, T., Carr, B. K. & May, W. S. Ceramide induces Bcl2 dephosphorylation via a mechanism involving mitochondrial PP2A. J. Biol. Chem. 274, 20296–300 (1999). 99. Ganesan, V. et al. Ceramide and activated Bax act synergistically to permeabilize the mitochondrial outer membrane. Apoptosis 15, 553–562 (2010). 100. Siskind, L. J. et al. The BCL‑2 protein BAK is required for long-chain ceramide generation during apoptosis. J. Biol. Chem. 285, 11818–11826 (2010).
Acknowledgements
We are grateful to J. Rathmell and M. Kurokawa for critical reading and feedback on the manuscript
Competing interests statement
The authors declare no competing financial interests.
FURTHER INFORMATION Sally Kornbluth’s homepage: http://cmb.duke.edu/faculty/kornbluth.html ALL LINKS ARE ACTIVE IN THE ONLINE PDF
VOLUME 11 | O CTOBER 2010 | 727 © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS
Eukaryotic DNA replication origins: many choices for appropriate answers Marcel Méchali
Abstract | At each cell division in humans, 30,000–50,000 DNA replication origins are activated, and it remains unclear how they are selected and recognized by replication factors. DNA replication in multicellular organisms must accommodate variations in growth conditions and DNA damage. It must also adapt to changes in chromatin organization associated with cell differentiation and development. The selection of replication origins in metazoans seems to involve multiple choices, with the appropriate answers depending on the identity of the cell or the conditions of growth. This suggests that during evolution, the use of replication origins became more controlled by epigenetic mechanisms affecting chromosome dynamics and expression than by DNA synthesis per se. Replication fork When replication starts, the opened DNA forms two branched structures on both sides of the replication origin that resemble forks. Fork progression is mediated by the action of DNA helicases that unwind the DNA and facilitate the movement of the DNA synthesis machinery.
Checkpoint One of many points in the cell cycle at which the cell checks whether the cycle can progress normally or should be delayed or stopped to allow time for the current phase to be completed properly.
Institute of Human Genetics, CNRS, Replication and Genome Dynamics, 141 Rue de la Cardonille, 34396 Montpellier cedex 5, France. e-mail:
[email protected] doi:10.1038/nrm2976
DNA replication is required for faithful inheritance of the genome at each cell division. To what extent this process uses similar mechanisms in bacteria and complex organisms is still debated. Multicellular organisms require additional layers of complexity to adapt to larger genomes and various cell fates, and the mechanism of DNA replication initiation is an example of such adaptation. To be duplicated, a DNA double helix must open to allow the DNA synthesis machinery to copy each DNA strand. These opening sites, called replication origins, are recognized by specific proteins, and DNA synthesis progresses from these sites in a bidirectional manner (FIG. 1a). In Escherichia coli, DNA replication starts from a single, sequence-specific element, and the speed of the two replication forks (60 kb min–1) keeps pace with a rapid cell cycle (less than 30 min). The human genome is 700‑fold larger than the E. coli genome, but the replication fork speed is 20‑fold slower (2–3 kb min–1). Thus, it would take at least 20 days to achieve a single division if there was one origin per chromosome. Pioneering work by Huberman and Riggs1 showed that in mammals, 30,000–50,000 origins are active at each cell cycle. Not all origins are activated at the same time; their activation follows the specific timing of DNA replication during the cell cycle (FIG. 1b). No consensus sequence has been identified to predict the localization of DNA replication origins in metazoans, and the degree of similarity between origin recognition in metazoans and the replicon model in E. coli 2 (whereby sequence-specific proteins recognize a sequence-specific DNA element) is also questioned. Moreover, it is now recognized that potential replication origins are in excess and only a subset
are activated at each cell cycle. This leads to the notion of flexible origins, which are used stochastically at each cell cycle or with an increased use in stress conditions or in specific cell cycles. Here, I describe replication origin features and the specificity of origin selection. I discuss the use of origins, the notion of flexible or dormant origins and the relationship of origins with cell cycle checkpoint controls. The contribution of chromosomes and chromatin structure to origin selection and firing and the influence of transcription will also be discussed, as well as how origin selection is linked with differentiation and development. Other aspects of DNA replication, particularly the function of the proteins involved, have been reviewed elsewhere3.
Assembly of the replication complex In metazoans, replication origins are set by a three‑step process: recognition of origins, assembly of a pre-replication complex (preRC) during G1 phase and activation of the preRC (BOX 1). This process should be tightly controlled as any origin that has been activated once should not be activated a second time in the same cell cycle. Proteins involved in origin recognition are relatively well conserved but the main origin binding factor, origin-recognition complex (ORC), varies in its sequence specificity. Saccharomyces cerevisiae ORCs specifically recognize a 12 bp consensus sequence4, but Schizo saccharomyces pombe and metazoan ORCs do not exhibit sequence specificity 5,6. In metazoans, ORC1 seems to be more involved than the other ORC subunits in the selection of the active DNA replication origin (reviewed in REF. 7). S. pombe Orc4 has an AT‑hook domain that
728 | O CTOBER 2010 | VOLUME 11
www.nature.com/reviews/molcellbio © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS a
5′ 3′
DNA
Origin
3′ 5′
RNA primer DNA leading strand 3′ 5′
5′ 3′ DNA lagging strand
Replication forks
b
G1
3
1
2
4
Origin
Origin
Origin
Origin
preRC
Bidirectional replication from activated origins 1
2
3
1
2
3
1
S 2
4
Figure 1 | Replication origins. a | At each replication origin, DNA synthesis starts with short RNA primers that are synthesized by DNA polymerase‑α. As DNA synthesis always occurs in the 5′–3′ direction, one strand of the DNA (the leading strand) will be synthesized | Molecular Cell Biology continuously, whereas the other strand (the laggingNature strand)Reviews will be synthesized discontinuously by short RNA-primed DNA fragments. Two other DNA polymerases (δ and ε) are recruited for the elongation of lagging and leading strands, respectively. b | Activation of replication origins during S phase. Pre-replication complexes (preRCs) are assembled at replication origins during G1 phase. Activation of replication origins occurs throughout S phase, some during early (1 and 2), and some in mid (3) or late (4) S phase.
CpG island A genomic region of at least 200 bp with a high frequency of CpG sites. CpG islands are often found in the transcription promoter regions of mammalian genomes and are unmethylated when the gene is expressed.
Chromatin immunoprecipitation A method used to localize the DNA-binding site of a protein using an antibody that specifically recognizes the protein on chromatin, which is previously broken into small pieces.
recognizes the AT‑rich elements of the region8, and a putative AT‑hook is also present in S. cerevisiae Orc2 (REF. 9.) Drosophila melanogaster ORC does not exhibit sequence specificity but has higher affinity for negatively supercoiled DNA5 and its binding sequences are more AT‑rich than the average genome10. Although AT‑richness alone is not sufficient to define an ORCbinding site10, it may influence replication origin activity 11,12. Similarly, the well-characterized human lamin B2 origin has an AT‑rich region and a CpG island that participate in its activity 13. Analysis in S. cerevisiae and D. melanogaster showed interesting similarities of ORC proteins with E. coli DnaA14. DnaA and ORC are both AAA proteins that use ATP to bind DNA. DnaA proteins bind to repeated elements at the E. Coli origin (DnaA boxes) and open the adjacent AT‑rich element. ATP binding induces the DnaA transition from a monomeric state to a right-handed
helical oligomer (which remodels replication origins for preRC assembly 15) that is a similar conformation to the D. melanogaster ORC16. All the other known factors that participate in the assembly of the replication initiation complex do not show much sequence specificity and depend on ORC for their binding to replication origins. Therefore, although the architectural features necessary for the assembly of the replication initiation complex are similar, none of the known preRC factors, including ORC, explain how specific rep lication origins are selected in multicellular eukaryotes. This is in contrast to transcription control, in which several transcription activators are sequence‑specific.
DNA replication origin features Several questions regarding replication origins in metazoans remain unanswered. Does the sequence of origins contribute to their features? Is there a different replication origin code in different cell lineages? To what extent is their activation stochastic during the cell cycle? S. cerevisiae, in which the mechanism seems simpler, has been valuable for unravelling preRC protein functions. Origins in the yeast genome. In S. cerevisiae, autonomous replication sequence (ARS) elements have a common, specific 12 bp consensus sequence (the autonomous consensus sequence (ACS)) that behaves as a replicator when inserted into plasmids17 and to which ORCs bind4. However, the presence of an ACS is not sufficient to predict a functional DNA replication origin — of the 12,000 ACS sites in S. cerevisiae genomes, only 400 are functional18. Genome-wide analyses to study replication timing 19 and map replication origins by chromatin immunoprecipitation (ChIP) using anti-ORC and anti-minichromosome maintenance protein (MCM) antibodies 20 did not uncover the known ACS consensus sequence element, unless ORC- and Mcm2‑binding sites were considered together 21. However, recent ChIP studies with anti-ORC antibodies coupled to high-throughput sequencing have identified the ACS motif 22. Comparative genomics using different S. cerevisiae strains18 confirmed that the ACS element was essential but not sufficient for origin activity and that a region of helical instability close to the ACS was also important. They also showed that S. cerevisiae origins are mainly located in intergenic regions. Thus, sequence specificity is essential but is not the only determinant of origin selection, even in S. cerevisiae. Other features, not obligatorily shared by all origins, may influence their selection, including transcription and/or the chromatin status (see below). In S. pombe, ARS elements that allow autonomous replication of plasmids do not share a specific consensus sequence as in S. cerevisiae. Origins are characterized by AT‑rich islands23–25, and poly-dA–poly-dT tracks can replace important regions in S. pombe ARS elements26. S. pombe origins are also found in intergenic regions, and genome-wide mapping of Orc1- and Mcm6‑binding sites and the identification of early 5‑bromodeoxyuridine (BrdU)-labelled origins confirmed that preRCs are formed at long AT‑rich intergenic regions27. Their main
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
VOLUME 11 | O CTOBER 2010 | 729 © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS be because of the abundance of some of the preRC proteins (such as MCMs) or because they are also involved in On your marks... functions other than origin localization (such as ORCs). DNA Origin 5′ 3′ In addition, only a small fraction of assembled preRCs is 3′ 5′ ????? activated at each cell cycle, and, therefore, ORC-binding sites are expected to outnumber active initiation sites (see below). Recent data in D. melanogaster show that 5′ 3′ ORC 3′ 5′ only 30% of ORC-binding sites are associated with early CDC6 CDT1–Geminin replication origins29. Two-thirds of ORC-binding sites are MCM9 at promoter regions, and no clear sequence specificity but CDC6 preferential localization at open chromatin and sites of 5′ 3′ CDT1 ORC cohesin loading were found29. 3′ 5′ MCM9 In Xenopus laevis eggs, which contain replication proMCM2–7 MCM2–7 teins in excess, any injected DNA replicates well30,31, with Ready... CDC6 an efficiency that increases with DNA size31. Initiation 5 5 2 3 2 3 5′ 3′ CDT1 ORC occurs at any site, with some preference for AT‑rich 3′ 5′ 6 6 7 7 MCM9 4 4 sequences, but this does not increase the overall efficiency Pre-replication complex of plasmid replication32. The same applies to endogenous CDC45, TOPBP1, GINS, chromosomes in X. laevis and D. melanogaster early SLD2, treslin, GEMC1, MCM10, embryos33,34. During early development, transcription is CDKs, DBF4 and RPA off and replication origins are set randomly at short, regSteady... Preinitiation complex ular intervals. A transition from random to site-specific Pol-ε and Pol-α DNA replication occurs when transcription starts later in Go! Initiation complex development35, and, in somatic cells, origins are located Geminin at specific sites in chromosomes, although no consensus sequence is discernible. CDT1 S Recent genome-wide studies to map origins in mouse and human cells36,37 confirmed that there is a correlation In eukaryotes, the origin recognition complex (ORC) — a heterohexamer with Nature Reviews | Molecular Cell Biology with unmethylated CpG islands and promoter regions38 DNA-dependant ATPase activity — is the only initiation protein complex thought to (C. Carou, P. Coulombe and M.M., unpublished observadirectly recognize origins. After ORC binds to an origin, two factors, cell division cycle 6 (CDC6) and CDT1, are recruited and have the role of loading the minichromosome tions). CpG islands could be a mark of replication origins maintenance protein (MCM) complex MCM2–7 onto the replication origin (see the or of the associated transcription promoters, although figure). The MCM2–7 complex is a heterohexamer that has ATPase-dependent DNA some known replication origins do not have strict CpG helicase activity and forms a ring around the DNA at replication origins159. After islands, such as those at the β-globin and dihydrofolate this helicase has bound DNA, the origins are licensed, marking the end of pre-replication reductase (DHFR) loci. CpG island methylation might complex (preRC) assembly. CDT1 is a major regulator of this reaction, as it is negatively regulate the timing of replication as origins at non‑ regulated by Geminin to restrict licensing to only once per cell cycle. In Xenopus laevis, methylated CpG islands replicate earlier than those Geminin is already present as a subcomplex with CDT1 (REF. 160), and MCM9, a recently at methylated CpG islands39. However, origin activity discovered new protein from the MCM family, cooperates with CDT1 in the binding of 162,163 does not seem to be altered by CpG methylation as oriMCM2–7 (REF. 161). At least two MCM complexes assemble at replication origins . gins are similarly used on both methylated, inactive and The preRC is further activated by a growing list of several other factors, including CDC45 and the GINS complex, CDC7–DBF4 and cyclin E–cyclin-dependent kinase 2 unmethylated, active X chromosomes40. Sequence asym(CDK2). This reaction enables the association of the DNA polymerase (Pol) machinery metries were also proposed as identifiers of replication and MCM2–7 to travel ahead of the replication fork to open the double-stranded origins41, but this remains debated as only a few origins DNA and allow the synthesis of the complementary strand. As cells enter S phase, have this property 42. CDT1 is inactivated by both its release from origins by Geminin and by degradation. In conclusion, CpG islands and AT‑rich stretches seem RPA, replication protein A; TOPBP1, DNA topoisomerase 2‑binding protein 1. to characterize metazoan DNA replication origins, but no consensus sequence has yet been revealed. A reason could be that metazoan origins are modular and are identified characteristic is the presence of AT‑rich islands23 that can and regulated by defined combinations of sequence elebe targeted by the S. pombe Orc4 subunit 8, which con- ments, similarly to transcription promoters. In early tains an AT‑hook domain not found in ORC proteins D. melanogaster and X. laevis embryos, where all compoMinichromosome from other species. nents necessary for DNA replication are stored in excess, maintenance protein sequence specificity might be more relaxed, alleviating the (MCM). One of a group of Metazoan origins. In multicellular organisms, ARS activity need for cooperative effects of different modules. proteins that belong to the assays were unsuccessful and it was difficult to identify It is therefore possible that metazoans have several AAA+ ATPase family and common features of replication origins. FIGURE 2 summa- classes of replication origins that use different modhave a conserved MCM box motif. The main eukaryotic rizes the elements that have been found at origins, but ules recognized by different subsets of proteins. Higher DNA-dependent and none of these alone is predictive of a replication origin. eukaryote genomes are packaged in chromosomes that ATPase-dependent DNA The use of ChIP methods to localize replication ori- contain longer regions of non-coding sequences than helicase is MCM2–7, a gins in higher eukaryotes was not as efficient as it was for yeast genomes. Heterochromatic regions may also require complex of six different MCM subunits. transcription factors or chromatin proteins28. This might specific factors to be recognized and opened. Thus, a Box 1 | Assembly of the pre-replication complex at DNA replication origins
730 | O CTOBER 2010 | VOLUME 11
www.nature.com/reviews/molcellbio © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS AT
GC
AT-rich DUE
CpG islands
Sequence
DNA topology
Loop MAR
Structure
Nucleosome DNase Isensitive site Chromatin
Promoter, enhancer or insulator
Start site Level?
Transcription
Figure 2 | Features of DNA replication origins. Summary of the features seen in Nature Reviews | Molecular Cell Biology DNA replication origins in eukaryotes. Several characteristics have been described at metazoan replication origins, but they are not present at all origins. Rather, they represent different marks or modules that can contribute to the selection of a given origin. At the sequence level, AT-rich elements and CpG islands have been reported as well as DNA regions that easily unwind (DNA unwinding elements (DUEs)), but their importance or role remains elusive. At the DNA structure level, bent DNA (or cruciform DNA) and the formation of loops has been described. At the chomatin level, nucleosome-free regions, histone acetylation and DNase-sensitive sites have been seen, but their direct participation in origin recognition as opposed to being a consequence of chromatin organization for transcription is sometimes difficult to estimate. The possible links of transcription features with replication origin recognition have been described but evidence for direct interactions between replication origin factors and transcription factors remains scarce. MAR, matrix attachment region.
universal origin consensus sequence might not be compatible with genome structure and expression plasticity of multicellular organisms.
Replication origins: a multiple choice At each cell cycle, only a subset of replication origins are used to replicate eukaryotic genomes.
GINS complex A complex comprising four subunits — SLD5, PSF1, PSF2 and PSF3 — that are ubiquitous and evolutionarily conserved in eukaryotes. It interacts with the MCM2–7 complex and CDC45 to activate the MCM2–7 helicase activity.
DNA combing A method used to produce arrays of uniformly stretched DNA molecules on silanized glass. The in vivo incorporation of fluorescent deoxynucleotides during replication permits, after combing, the localization of DNA replication origins and the progression of the replication forks on the combed DNA molecules.
Different classes of origins. During G1 phase of the cell cycle, inactive MCM helicases are loaded on the preRCs. Activation of preRCs is regulated at two levels. First, origins are activated at different times throughout S phase and are classified as early-, mid- and late-activated origins. Second, only a fraction of all potential origins is used at each cell cycle, whereas others are ‘woken’ and used only in conditions that affect S phase, such as DNA damage or changes in growth conditions43. At first, DNA replication origin firing seems to be a relatively inefficient process in metazoans and yeast. In S. cerevisiae (where origins have a consensus sequence) and in S. pombe, the overall efficiency of origin firing is less than 50%25,44 but varies considerably across individual origins — some are used at almost every cell cycle and others are nearly inactive. In metazoans, the efficiency measured in specific replication domains is 5–20%45,46. Analysis of yeast ribosomal DNA (rDNA) replication origins in the rDNA cluster by DNA combing showed that not all origins in the 100–200 identical copies of rDNA are activated. Instead, origins are activated by clusters of two or three consecutive units separated by large regions in which origins are silenced47. A similar situation is seen in human cells45. According to their use, DNA replication origins can fall into three classes: flexible, dormant (or inactive) and constitutive (FIG. 3). Constitutive origins, which are a minority of the eukaryotic origins, are used all the time in any cell cycle or cell type. Flexible origins are potential origins that can be used stochastically in different cells and explain the low origin use seen in eukaryotic cells. They also elucidate the notion of an initiation zone, in which several origins are found at relatively close intervals
in a domain, such as the DHFR locus. Here, multiple origins are used indifferently over a 50 kb region48. In reality, if individual cells activate a single origin in a locus but at different places along the domain, the analysis of the whole cell population will score all the origins. The resulting pattern will reflect the sum of all individual situations and the stochastic nature of origin activation in this locus. If some origins are deleted, others nearby become more active or more efficient, reflecting a large choice of origins48,49. Flexible origins follow the ‘Jesuit Model’ proposed 17 years ago50 — “For many are called, but few are chosen” (Matthew 22:14, The Bible) — but flexibility could be lost in two ways. First, by increasing origin use in cases of poor growth conditions or DNA damage. Second, by decreasing the choice of origins in specific domains in cells engaged in specific differentiation programmes (FIG. 3). Inactive or dormant origins are potential origins that are never used in the cell cycle in normal conditions but that can be woken in specific cell programmes or in stress conditions. An important question is whether origin selection in a given cell is transmitted through cell divisions. In this case, flexibility would be linked to cell-to-cell variation rather than to a complete stochastic use of origins at each cell division. Chromatin fibre analysis showed that 70–90% of active origins were the same during two consecutive cell cycles51, suggesting that there is a memory for the selection of origins in a given cell. However, this memory effect seems to apply to replication clusters and replication foci52 rather than to individual origins inside the clusters53. Why are there so many potential replication origins? First, flexible origins could be abortive origins that on activation failed to elongate. Over-replication of short DNA regions (less than 200 bp) was seen during S phase in human cells, close to origin sites54, suggesting that all potential origins are activated but only a few succeed in elongating. Second, the efficiency of DNA replication origin firing could be regulated by a limiting factor, or factors, that is not present in sufficient amount to be recruited to all origins. Such factors may include the HSK1 and homologous cell division cycle 7 (Cdc7) kinases55, cyclin-dependent kinase (CDK) activities and CDC45 (REFS 56,57). These factors become available for late origins after being released from early replicated DNA. Third, several potential origins could provide a choice for the most suitable origin to be activated in a given chromatin context, which might vary in different cells or tissues (see below). This could explain why highly clustered origins are found in heterochromatic regions, such as subtelomeric areas27. The apparently stochastic nature of origin activation may therefore reflect the necessity to adapt DNA replication to the chromosomal context, developmental stage or growth conditions. A link with the cell cycle and its checkpoints. When in the cell cycle are origin sites chosen? A series of elegant experiments showed that during G1 there is a specific point — the origin decision point (ODP) — at which replication origins are selected58. Mitosis also seems to be a crucial stage for the reorganization of the nucleus that is needed for the selection of origins. Differentiated
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
VOLUME 11 | O CTOBER 2010 | 731 © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS Potential origin sites (Pre-licensed by CpG islands and other sequence features)
PreRC
Origin selection
Origin Cell 1 Cell 2 Cell 3 Cell 4 Flexible cluster Use Growth conditions
Stress Inactive or dormant Change in cell programme
Constitutive
Chromatin, differentiation and transcription
Figure 3 | Different types of DNA replication origins. Potential DNA replication origins are set during mitosis–G1 phase by the assembly of pre-replication complex (preRC) proteins. The selection of the origins that will be activated at the next S phase occurs at G1 phase and may vary according to the cell fate or environmental conditions. Four examples of DNA replication origin positions are shown in different cells in a growing cell population. A cluster of flexible origins contains origins that can be used differently in different cells. Their use could increase or decrease according to physiological or abnormal growth conditions. Inactive or dormant origins are rarely used or are not used at all. Constitutive origins are fixed origins that are always set at the same position by chromatin or transcriptional constraints. Replication stress can activate dormant origins or increase the use of flexible origins, resulting in an increased number of origins per replication cluster.
Topoisomerase II An enzyme that cuts and reseals both strands of DNA to remove DNA supercoiling.
Hypomorphic allele An allele with a mutation that decreases gene expression.
Insulator element A regulatory DNA sequence that serves as a genetic barrier to protect a gene against positional effects and the spreading of condensed chromatin or to block enhancer activity.
nuclei adopt short inter-origin spacing characteristics of early development only after exposure to a mitotic context59. This phenomenon is controlled by topoisomerase II and correlates with the size of chromatin loops. Topoisomerase II also interacts with the human replication origin located on the lamin B gene60. Similarly, mitosis is important to reset inter-origin spacing in Chinese hamster ovary cells61. Origin selection seems to be relatively independent from the timing of replication, which is defined earlier in the cell cycle than origin selection58,62. In addition, the timing of replication seems to be determined at the level of clusters of origins or even larger replication domains rather than at the level of individual origins63,64. The relationship between cell cycle checkpoints and increased use or activation of dormant origins has been the focus of recent investigations. Dormant origins might be spare origins that rescue DNA replication when fork progression is perturbed during replication stress65. Usually, inhibitors of DNA damage or DNA synthesis activate the S phase checkpoint, which prevents the firing of late origins, allowing DNA repair to take place before further S phase progression65–67. In conditions of reduced nucleotide precursor concentration, late origins are inhibited by the checkpoint activity, whereas dormant origins are activated68,69. Similarly, a double-strand DNA break enhances the use of origins proximal to the break70. Therefore, the checkpoint response to poor cell growth conditions prevents S phase progression but could also increase the use of flexible origins or activate dormant origins to rescue under-replicated domains. This suggests that the excess of potential origins in G1 represents
an efficient safeguard against possible S phase problems. Spare origins must be ‘put on stand-by’ before S phase because during S phase, new preRC formation (by the licensing reaction) is not possible in order to strictly avoid re-replication. All potential origins are therefore licensed at the end of G1 phase, and a few of them are used during the cell cycle. According to this model, the excess of MCM and ORC proteins in a cell allows the licensing of all potential origins. Indeed, it was found that lowering the MCM concentration by small interfering RNA suppresses the use of dormant origins69. In agreement, a hypomorphic allele of MCM4 causes chromosomal instability in mice68,71. These observations led to the question of whether checkpoint kinases are involved in origin activation in the absence of DNA damage. The ataxia telangiectasia mutated (ATM) and the ATM-related (ATR) genes72 encode kinases that are involved in the checkpoint response to DNA damage. ATR is activated by the stalling of replication forks73, and two other checkpoint kinases — CHK1 (also known as CHEK1) and CHK2 (also known as CHEK2) — are downstream targets that are involved in the inhibition of late-origin activation74. Replication stress activates ATR, leading to phosphorylation of CHK1, the inhibition of late-origin firing and the stabilization of stalled replication forks. The movement of replication forks can also be hindered by physiological constraints of the genome. First, some DNA sequence elements may be more difficult to duplicate than others. Second, chromatin-associated factors, insulator elements and specific nuclear structures may slow down or stall a replication fork in the absence of replication stress. Indeed, in X. laevis, treatment with caffeine, an inhibitor of the ATM and ATR checkpoint, slows down replication forks and increases the number of replication origins75,76. Similarly, in CHK1‑depleted cells, the speed of replication forks is decreased and more origins are activated77–79. So why are late origins inhibited and why does the use of early origins increase following replication stress? I suggest the existence of two pathways (FIG. 4). The first is a general ATR-dependant pathway that inhibits late-origin firing and limits the origin usage in clusters of early origins (lateral inhibition), and the activity of which is increased by DNA damage. The second pathway is specifically induced during replication stress and overrides ATR inhibition of dormant origins. So, when ATR is inhibited by caffeine, both late and dormant origins are activated. When ATR levels are increased by replication stress, late origins are suppressed, whereas early-origin use is increased by overriding the ATR inhibitory effect on early origins. It is also possible that in some regions, a passive mechanism is at play, whereby slowing down of a replication fork gives time for a potential origin to become activated before the fork passes through it (see also REFS 69,74). ATR and CHK1 might regulate the activation of replication origins by controlling CDC45 assembly at replication origins. Activation of the S phase checkpoints inhibits the association of CDC45 with chromatin80–82 and therefore inhibits the transition of the preRC to the preinitiation complex. Two newly described factors are involved in this
732 | O CTOBER 2010 | VOLUME 11
www.nature.com/reviews/molcellbio © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS Replication stress • Nucleotide depletion • DNA damage • Slowing down or stalling of the replication fork ATR
PreRC at origin
Early origins
Late origins
Nature Reviews | Molecular Cell Biologya Figure 4 | Checkpoint control of replication origins firing. During DNA replication, DNA lesion caused by stresses such as ultraviolet damage, oxidative damage, genotoxic drugs or growth medium deprivation may slown down or even arrest progression of the replication fork. Replication stress can induce or increase a signal transduction cascade, called the checkpoint response, which tries to maintain the integrity of the replication forks and facilitate DNA repair in coordination with the cell cycle. The checkpoint signalling through ataxia telangiectasia mutated-related (ATR) results in the inhibition of late origins. The frequency of origin usage in clusters of early origins might also be negatively regulated by the ATR pathway, through a lateral inhibition of activated origins on the potential neighbouring origins. However, replication stress might induce a second pathway that cancels this inhibitory effect on early origins, resulting in activation of dormant early origins.
transition: treslin (also known as TICRR)83,84 and Geminin coiled-coil domain-containing protein 1 (GEMC1)85. Both factors interact with DNA topoisomerase 2‑binding protein 1 (TOPBP1; homologous to Cut5 (also known as Rad4)), a general activator of ATR involved in the checkpoint response that regulates the association of CDC45 with replication origins86–89.
Epigenetic features and nuclear organization The selection of a replication origin amid the many potential ones is predominantly regulated by local chromatin structure and epigenetics (FIG. 5).
Y RNA A small non-coding RNA component of the Ro ribonucleoprotein particle that is frequently recognized by antibodies present in autoimmune sera. There are four Y RNAs in humans.
Macronucleus The larger of the two nuclei present in ciliate protozoans.
Chromatin structure. Generally, it is thought that to be efficient a replication origin should be located in a relatively open chromatin domain. In S. cerevisiae, positioning of a single nucleosome on the ACS consensus sequence is sufficient to block the use of the corresponding origin90. Similarly, nucleosome positioning by ORC facilitates DNA replication initiation91 by promoting preRC assembly. A recent study on the nucleosome patterns at replication origins in S. cerevisiae shows that ACS is sufficient to exclude nucleosomes, whereas ORC is necessary for ordered nucleosome positioning around the origin22. Therefore, chromatin arrangement at origins seems to be crucial for a two-step process of origin selection and function. Chromatin remodelling complexes also facilitate the formation of the origin complex in S. cerevisiae and higher organisms92. In yeast, a genome-wide scan showed that origins are more active at open chromatin structures93. Similarly, histone acetylation, which is associated with higher chromatin accessibility, correlates with origin activation94,95. In agreement, Sir2 histone deacetylase inhibits the activity of five replication origins in S. cerevisiae by favouring an unsuitable positioning of nucleosomes at replication origins96. Recently, it was shown that multiple acetylation of histones H3 and H4 enhances DNA synthesis in a replicating plasmid97. Histone acetyltransferase
binding to ORC1 (HBO1; also known as MYST2), a major H4 histone acetylase, is required for DNA replication licensing of origins98–100. However, acetylation is not a universal feature of replication origins36,101,102. It can be used to select some origins, but generally it seems to be linked to the timing of origin activation rather than to their selection64,103–105. It should also be stressed that histone modifications might only be transient at replication origins and that only highly synchronized systems will unravel them. An unexpected finding was the link between heterochromatin and replication origins. In S. cerevisiae, silencer elements contain replication origins106, which assemble the ORC107. The ORC subunit Orc1 is more related to Sir3, which is involved in heterochromatin silencing, than to any other ORC subunits of other species108. Orc1 interacts with Sir1 (REF. 109) to establish a silenced state independently from its role in DNA replication. In higher eukaryotes, Sir1 is not present and the ORC1 domain that interacts with Sir1 is lost. However, this interplay is replaced by the interaction of ORC with heterochromatin protein 1 (HP1)110. These conserved links between heterochromatin and ORC are intriguing in view of the general idea that an open chromatin structure is more suitable for DNA replication-origin activity. Three explanations could be envisaged. Because heterochromatin is less accessible to preRC proteins, a direct interaction between HP1 and the main origin-recognition protein, ORC, would help heterochromatin to assemble the preRC. This would constitute an important sequence-independent recruitment of ORC to origins at less accessible chromatin domains. In agreement with this hypothesis, ISWI, a chromatin remodelling factor, is required for replication of heterochromatin in mammalian cells92 but not for the less compacted chromatin of early X. laevis embryos111. Another possibility would be that HP1 is used not only at heterochromatin but also at subnuclear structures implicated in preRC assembly at euchromatin regions. Finally, in S. pombe, Swi6 (the HP1 homologue) activates origins by recruiting Dbf4‑dependent kinase (DDK) and Cdc7 to heterochromatin foci112 and by allowing Sld3 assembly at origins. Sld3 is needed to recruit DNA polymerase and to establish a replication fork at replication origins in yeast113. No Sld3 homologue has been identified in higher eukaryotes, but functional homologues have recently been detected83,84. Non-coding RNAs. At present, evidence of non-coding RNAs regulating DNA replication initiation is scarce. However, as non-coding RNAs play an essential part in the epigenetic control of transcription, a function in DNA replication would be predicted. A specific class of RNAs called Y RNAs has been proposed to be involved in replication initiation in vertebrates, although no interaction of such RNAs with replication proteins has been detected so far 114. In Tetrahymena thermophila, rDNA amplification, which occurs during macronucleus development, is regulated by the binding of ORC to a non-coding RNA that corresponds to the 3′ end of the 26S rDNA115. However, this reaction is specific for the amplification of rDNA and not for other replication origins116. The replication
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
VOLUME 11 | O CTOBER 2010 | 733 © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS Chromatin remodelling complexes HBO1 TF Nucleosome HATS
Figure 5 | Chromatin and replication origins. Binding of Nature Reviews | Molecular Cell Biology the origin recognition complex (ORC) to replication origins may be favoured by the chromatin remodelling complexes of nucleosomes that can also facilitate the access of histone acetylases (HATs) and thus the acetylation of nucleosomes. This, in turn, would facilitate pre-replication complex (preRC) assembly at origins. Proteins from the preRC, such as the minichromosome maintenance protein (MCM) complex, might also be modified by histone acetyltransferase binding to ORC1 (HBO1; also known as MYST2), a histone H4 acetylase and coactivator of CDT1. Associated factors, such as transcription factors (TFs) and heterochromatin protein 1 (HP1), can cooperate. The assembly of the preRC also results in an ordered positioning of the nucleosome around the region, leaving the origin free of nucleosomes.
of the Epstein-Barr virus (EBV) requires EBV nuclear antigen 1 (EBNA1), a virus-encoded protein that allows the recruitment of ORC to the viral replication origin by an RNA-dependent interaction117.
Cohesin A protein complex that is responsible for the association of the two sister chromatids during S phase.
Organization of DNA replication origins in the nucleus. Nakamura et al. were the first to show that DNA replication takes place at discrete nuclear structures118 that form up to 1,000 foci per nucleus52,119. Each focus can contain 10–100 replicons, constituting a replication domain52,120,121. These replication factories comprise various proteins involved in the elongation of DNA replication. It was proposed that replication foci are stable chromatin territor ies120,121 but recent data suggest that they are characterized by small replicon clusters that associate in large replication domains63. It is thought that each focus is formed by several synchronously activated replicons. These replication clusters are often thought to form chromatin loops that are anchored to a matrix structure122. Although this longstanding hypothesis remains controversial, the correlation between replicons, chromatin loops and replication foci deserves to be experimentally revisited using genomewide approaches for origin mapping and new imaging technologies. The excess of potential origins in the genome might be explained by additional roles of ORC. Several studies now point to a function of ORC in cohesin recruitment and chromosome structure during S phase. ORC-binding sites seem to correlate with cohesin-binding sites in D. mela nogaster 29,123,124, and ORC depletion inhibits cohesin loading in X. laevis 129. Sister chromatid cohesion is also impaired in Orc2‑depleted yeast cells, although cohesin seems to be normally associated with chromatin 130. Therefore, all ORC-bound origins, including those that are inactive, might play a part in sister chromatid cohesion during S phase.
It is important to know whether preRCs are already assembled in preformed nuclear foci before DNA synthesis is initiated. Studies in X. laevis suggest that this could be the case125,126 as replication protein A (RPA) foci can be clearly detected before S phase initiation. However, it is still debated whether these foci form only when initiation of DNA synthesis is engaged. By definition, replication foci are detected with BrdU pulses or by immunofluorescence using antibodies against elongation proteins, therefore only after initiation of DNA synthesis. As they are not clearly seen with antibodies against preRC proteins, it remains questionable whether replication origins already assemble in replication foci at the preRC stage. Is the mechanism for selecting clusters of replication origins independent from the mechanism for selecting individual origins? Firing at a given DNA replication origin may inhibit firing at nearby origins by an origin interference mechanism45,127 that leads to the 100–120 kb average size of a replicon. A specific origin among several flexible origins would be chosen by this mechanism and a local checkpoint response might inhibit this mechanism (FIG. 4). Interestingly, although individual replication origins are flexible, large replication domains128 and replication clusters are not, and the same clusters of replication origins are activated in subsequent cell cycles52,53. Control elements, either distant from the origin cluster or in the domain, could control the firing of the whole cluster. At the level of global chromosome organization, rep lication domains are not uniformly distributed along chromosomes. Instead, early and late replication domains are interspersed; in mammalian cells they have a size ranging from 200 kb up to 2 megabases63,129 (C. Carou, P. Coulombe and M.M., unpublished results), whereas they are much shorter in D. melanogaster 10,64 and S. pombe27. Our recent data show that early replication domains are origin-rich, whereas late replication domains are origin-poor. In yeast and mammalian cells, late-activated origins preferentially localize at the nuclear periphery, suggesting that the localization of the domains in the nucleus correlates with the timing of their replication130. Some sequence determinants control the timing of replication origin firing 130,131 but do not correlate with the localization of origins at the nuclear periphery.
Transcription and DNA replication origins Links between transcription and replication result in either negative or positive regulation. Transcription may affect the activity or the choice of replication origins or may influence chromatin structure at replication origins. DNA synthesis itself may provide a window of opportunity to assemble or erase transcription factors as chromatin is opened and reassembled during the passage of the replication fork. A clear connection exists in viral genomes, in which the replication initiation protein can be both a transcription factor and a replication factor 132. The interplay between metazoan replication origin selection and transcription correlates more with the timing of replication. First, in animal cells, there is a well-known relationship between replication origins that are activated early and actively transcribed genes, whereas origins activated late in S phase are associated with non-transcribed
734 | O CTOBER 2010 | VOLUME 11
www.nature.com/reviews/molcellbio © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS a Somatic cells
b Early embryos M
Unspecified G2 M
Unspecified ODP
G2 G1
Specific
between the DNA polymerase and RNA polymerase machineries, as seen in bacteria145. As well as being associated with promoters36,37, many origins are also found inside genes. The enrichment of origins in intergenic regions is more convincing in yeast than in metazoans. However, as most replication origins are obviously present in non-coding regions, other mechanisms of origin selection must be at work.
S
S
Figure 6 | Selection of specific origins of DNA replication during the cell cycle. a | In somatic cells, the events occurring from mid‑G1 phase (the origin decision point Natureto Reviews | Molecular Cell Biology (ODP)) in the cell cycle select the DNA replication origins be used from the many potential origins set between the end of mitosis (M) and early G1 phase (blue),58. After the ODP, replication origin sites are set (red). b | Early Drosophila melanogaster and Xenopus laevis embryos go through a series of cell divisions that consist of overlapping S and M phases without clear G1 and G2 phases. DNA replication origins, which have been deprogrammed in mitosis59, are not reprogrammed and remain unspecified because of the absence of G1 phase. Embryonic stem cells, which are characterized by a very short G1 period, may also have a larger choice of DNA replication origins.
regions as heterochromatin133–135. Transcriptionally active genes have more efficient origins, and this has been shown for individual genes72,136 and in genome-wide studies10,36,37. However, this correlation is not seen in S. cerevisiae19,20, and a recent genome-wide analysis showed that it is not always the case in D. melanogaster 64 or mouse cells63. In several cases, the assembly of a chromatin domain devoted to transcription rather than transcription per se affects origin activity 95,137. Second, active transcription in a gene silences replication origins inside that gene138–140 or reduces the size of the initiation zone137,141. This implies that origins activated in non-transcribed genes might be erased in transcribed alleles, further complicating genome-wide analysis. In addition, an origin is activated on a gene only during a specific time in the cell cycle; if transcription of the corresponding domain does not occur at the same time, the origin might not be affected. Third, although active transcription may prevent firing of a potential replication origin (negative regulation), the presence of a promoter may help the selection of an active origin (positive regulation)49,142. This might be due to increased chromatin accessibility at the promoter region and to crosstalk between replication initiation proteins and transcription factors. In D. melanogaster, ORC is often associated with RNA polymerase II-binding sites10 but direct interactions between preRC proteins and transcription factors have not been convincingly shown. Promising results were obtained in D. melanogaster, in which MYB, E2F1 and retinoblastoma (RB) were found to be associated with elements that control the amplification of the Chorion locus and to interact with ORC143,144. Such or similar associations have not yet been described for standard origins. The clustering of origins in promoter regions may also eliminate possible head-to-head collision
Development and DNA replication origins The often discussed relationship between transcription and DNA replication may reflect a link between DNA replication and cell identity 146. Indeed, during X. laevis or D. melanogaster early embryonic development, when transcription is off, replication origins are activated at very short intervals, every 10–20 kb33,35,147. When transcription resumes in the embryos, origins are set at specific sites and at larger intervals35,148, and the experimental assembly of a transcription-competent promoter is sufficient to drive site-specific replication origins95. The explanation might be that early X. laevis or D. melanogaster embryos use all potential origins to accelerate S phase as they have no transcription constraints. In somatic cells or in late embryos, transcription would restrain origin use and contribute to the establishment of a flexible origin usage, adapted to the cell fate. Another possibility is that a specific organization of the chromosomes during early development imposes a regular setting of origins at short intervals. A link between chromatin loops and replicon size, seen in different organisms149 and in reprogramming experiments59, may contribute to the organization of replication origins. Changes in the choice of origins were observed in other developmental or differentiation processes, such as during cell differentiation in Physarum polycephalum136, neural differentiation of mouse cells101, embryonic development of the fly Sciara coprophila137, B cell development72, and at the chicken β-globin locus during erythroid differentiation102. Links between replication and development are also established at the level of the timing of replication rather than at the level of origin localization and selection. A series of extensive analyses in mouse and human embryonic stem cells have emphasized the evidence for changes in replication timing during differentiation63,150, which were also seen during D. melanogaster cell differentiation64. A more recent analysis carried out using cells from different germ layers during mouse development confirms these results and the interplay between replication timing and transcription151. A link between origins and differentiation programmes is also suggested by the unexpected involvement of preRC proteins in differentiation. For instance, in D. melanogaster, Latheo, a protein that interacts with ORC, is involved in neural differentiation152. Furthermore, Geminin plays a crucial part in inactivating the preRC complex after DNA synthesis has started and also has a function in neural differentiation that is related to the transcriptional control of the Hox genes153–155 and chromatin remodelling 153. In Arabidopsis thaliana, GEM (which, like Geminin, interacts with CDT1), is involved in root epidermis patterning 156. More recently, the homeotic protein HOXC13 was found at three human replication origins157.
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
VOLUME 11 | O CTOBER 2010 | 735 © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS Perspectives Recent work emphasizes that we have many more replication origins than needed at each cell cycle and that their use is highly flexible. Flexibility is probably needed to adapt to environmental cues, obstacles encountered by the replication machinery and developmental features that require reorganization of the genome or nuclear structures. This might explain the difficulties in finding a common rule that specifies replication origins. Origins are probably multi-modular and their use might depend on the combination of modules and the concentration and affinity of the proteins that bind them. An interesting question is to what extent genome-wide studies can define these modules and predict their usage according to the combination. Flexibility, however, does not mean that initiation is totally random. Recent genome-wide studies indicate that this is not the case36,37. Origins are reproducibly found at the same sites in a given cell population and flexibility refers to the choice of the activated origins among these sites in a given cell or tissue. A remaining issue is how constitutive or master origins and secondary origins that are used only in specific cell contexts are established. Origins that are activated by the checkpoint response following a replication stress may fall into this second category. A more open chromatin structure, owing to the presence of a transcription promoter region, might be suitable for setting a replication origin, but other chromatin features, not obligatorily linked to transcriptional permissiveness, might be important as well. The flexible use of replication origins in somatic cells and the extreme use of origins in transcriptionally quies cent embryos strongly suggest that a strict positioning of origins in the metazoan genome is not required for a regulated, once-per-cell-cycle type of DNA synthesis. The role of transcription promoters and the associated chromatin features that are present in some specific cases Huberman, J. A. & Riggs, A. D. Autoradiography of chromosomal DNA fibers from Chinese hamster cells. Proc. Natl Acad. Sci. 55, 599–606 (1966). 2. Jacob, F., Brenner, J. & Cuzin, F. On the regulation of DNA replication in bacteria. Cold Spring Harbor Symp. Quant. Biol. 28, 329–348 (1963). 3. DePamphilis, M. L. DNA Replication and Human Disease (Cold Spring Harbor Laboratory Press, 2006). 4. Bell, S. P. & Stillman, B. ATP-dependent recognition of eukaryotic origins of DNA replication by a multiprotein complex. Nature 357, 128–134 (1992). The first article describing the ORC. 5. Remus, D., Beall, E. L. & Botchan, M. R. DNA topology, not DNA sequence, is a critical determinant for Drosophila ORC-DNA binding. EMBO J. 23, 897–907 (2004). 6. Vashee, S. et al. Sequence-independent DNA binding and replication initiation by the human origin recognition complex. Genes Dev. 17, 1894–1908 (2003). 7. DePamphilis, M. L. et al. Regulating the licensing of DNA replication origins in metazoa. Curr. Opin. Cell Biol. 18, 231–239 (2006). 8. Chuang, R. Y. & Kelly, T. J. The fission yeast homologue of Orc4p binds to replication origin DNA via multiple AT‑hooks. Proc. Natl Acad. Sci. USA 96, 2656–2661 (1999). 9. Duncker, B. P., Chesnokov, I. N. & McConkey, B. J. The origin recognition complex protein family. Genome Biol. 10, 214 (2009). 10. MacAlpine, D. M., Rodriguez, H. K. & Bell, S. P. Coordination of replication and transcription along a 1.
11.
12.
13. 14. 15.
16.
17. 18.
does not seem to be a general determinant for origins. Inversely, setting a replication origin in a given chromatin domain does not necessarily favour transcriptional activity. It might be that neither replication origins control transcription nor transcription controls replication, but that an underlying structure of a chromosome fixes the regulation of both transcription and the replication origins of a domain. It might be argued that S phase rules are in fact set during the formation of mitotic chromosomes at the previous cell cycle. Indeed, in X. laevis 59, mammalian cells61 and yeast57, origin use during a given cell cycle is regulated by events that occur during previous mitoses. In this case, the selection of replication origins among several potential origins might be achieved in two steps. Mitosis might be an important first step to reset the origin code of a given cell and might be used to deprogramme its corresponding genome. The replication origins to be activated might be further fixed in G1 phase (FIG. 6) and adapted to the gene expression programme. Interestingly, pluripotent stem cells have a short G1 period and replication origin choice might be more flexible. In agreement, nuclear transfer experiments showed that adult somatic chromosomes are efficiently reprogrammed in mitotic zygotes but not in interphase zygotes158. The reprogramming of replication origins in mitosis might therefore be crucial for somatic cell reprogramming, even in the induction of pluripotent stem cells. The possibility of multicellular organisms reprogramming their origins at each cell cycle might contribute to the segmentation of the genome into autonomous units of replication and transcription. This might help to organize chromosomes for different transcription programmes in different cell lineages during development and differentiation. Therefore, DNA replication origins might be much more than just entry sites for replication factories; they might be essential regulatory elements that organize the genome according to cell fate.
Drosophila chromosome. Genes Dev. 18, 3094–3105 (2004). Wang, L., Lin, C. M., Lopreiato, J. O. & Aladjem, M. I. Cooperative sequence modules determine replication initiation sites at the human β-globin locus. Hum. Mol. Genet. 15, 2613–2622 (2006). Kong, D., Coleman, T. R. & DePamphilis, M. L. Xenopus origin recognition complex (ORC) initiates DNA replication preferentially at sequences targeted by Schizosaccharomyces pombe ORC. EMBO J. 22, 3441–3450 (2003). Paixao, S. et al. Modular structure of the human lamin B2 replicator. Mol. Cell Biol. 24, 2958–2967 (2004). Mott, M. L. & Berger, J. M. DNA replication initiation: mechanisms and regulation in bacteria. Nature Rev. Microbiol. 5, 343–354 (2007). Erzberger, J. P., Mott, M. L. & Berger, J. M. Structural basis for ATP-dependent DnaA assembly and replication-origin remodeling. Nature Struct. Mol. Biol. 13, 676–683 (2006). Clarey, M. G. et al. Nucleotide-dependent conformational changes in the DnaA-like core of the origin recognition complex. Nature Struct. Mol. Biol. 13, 684–690 (2006). References 14–16 describe how DnaA recognizes the E. coli replication origin. Stinchcomb, D. T., Struhl, K. & Davis, R. W. Isolation and characterisation of a yeast chromosomal replicator. Nature 282, 39–43 (1979). Nieduszynski, C. A., Knox, Y. & Donaldson, A. D. Genome-wide identification of replication origins in yeast by comparative genomics. Genes Dev. 20, 1874–1879 (2006).
736 | O CTOBER 2010 | VOLUME 11
19. Raghuraman, M. K. et al. Replication dynamics of the yeast genome. Science 294, 115–121 (2001). 20. Wyrick, J. J. et al. Genome-wide distribution of ORC and MCM proteins in S. cerevisiae: high-resolution mapping of replication origins. Science 294, 2357–2360 (2001). 21. Xu, W., Aparicio, J. G., Aparicio, O. M. & Tavare, S. Genome-wide mapping of ORC and Mcm2p binding sites on tiling arrays and identification of essential ARS consensus sequences in S. cerevisiae. BMC Genomics 7, 276 (2006). 22. Eaton, M. L., Galani, K., Kang, S., Bell, S. P. & MacAlpine, D. M. Conserved nucleosome positioning defines replication origins. Genes Dev. 24, 748–753 (2010). 23. Segurado, M., de Luis, A. & Antequera, F. Genome-wide distribution of DNA replication origins at A+T-rich islands in Schizosaccharomyces pombe. EMBO Rep. 4, 1048–1053 (2003). 24. Dai, J., Chuang, R. Y. & Kelly, T. J. DNA replication origins in the Schizosaccharomyces pombe genome. Proc. Natl Acad. Sci. USA 102, 337–342 (2005). 25. Heichinger, C., Penkett, C. J., Bahler, J. & Nurse, P. Genome-wide characterization of fission yeast DNA replication origins. EMBO J. 25, 5171–5179 (2006). 26. Okuno, Y., Satoh, H., Sekiguchi, M. & Masukata, H. Clustered adenine/thymine stretches are essential for function of a fission yeast replication origin. Mol. Cell Biol. 19, 6699–6709 (1999). 27. Hayashi, M. et al. Genome-wide localization of pre-RC sites and identification of replication origins in fission yeast. EMBO J. 26, 1327–1339 (2007).
www.nature.com/reviews/molcellbio © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS 28. Schepers, A. & Papior, P. Why are we where we are? Understanding replication origins and initiation sites in eukaryotes using ChIP-approaches. Chromosome Res. 18, 63–77 (2010). 29. MacAlpine, H. K., Gordan, R., Powell, S. K., Hartemink, A. J. & MacAlpine, D. M. Drosophila ORC localizes to open chromatin and marks sites of cohesin complex loading. Genome Res. 20, 201–211 (2010). 30. Laskey, R. A. & Harland, R. M. Replication origins in the eukaryotic chromosome. Cell 24, 283–284 (1981). 31. Mechali, M. & Kearsey, S. Lack of specific sequence requirement for DNA replication in Xenopus eggs compared with high sequence specificity in yeast. Cell 38, 55–64 (1984). 32. Stanojcic, S., Lemaitre, J. M., Brodolin, K., Danis, E. & Mechali, M. In Xenopus egg extracts DNA replication initiates preferentially at or near asymmetric AT sequences. Mol. Cell Biol. 28, 5265–5274 (2008). 33. Hyrien, O. & Méchali, M. Chromosomal replication initiates and terminates at random sequences but at regular intervals in the ribosomal DNA of Xenopus early embryos. EMBO J. 12, 4511–4520 (1993). 34. Nishiyama, T. et al. Rac p21 is involved in insulininduced membrane ruffling and Rho p21 is involved in hepatocyte growth factor- and 12‑O‑tetradecanoylphorbol‑13‑acetate (TPA)-induced membrane ruffling in KB cells. Mol. Cell Biol. 14, 2447–2456 (1994). 35. Hyrien, O., Maric, C. & Mechali, M. Transition in specification of embryonic metazoan DNA replication origins. Science 270, 994–997 (1995). This study shows that a transition from non-specific to specific locations of DNA replication origins occurs during development in X. laevis. 36. Cadoret, J. C. et al. Genome-wide studies highlight indirect links between human replication origins and gene regulation. Proc. Natl Acad. Sci. USA 105, 15837–15842 (2008). The first genome-wide study of human DNA replication origins. 37. Sequeira-Mendes, J. et al. Transcription initiation activity sets replication origin efficiency in mammalian cells. PLoS Genet. 5, e1000446 (2009). 38. Delgado, S., Gomez, M., Bird, A. & Antequera, F. Initiation of DNA replication at CpG islands in mammalian chromosomes. EMBO J. 17, 2426–2435 (1998). This study shows a link between CpG islands and initiation of DNA replication. 39. Gomez, M. & Brockdorff, N. Heterochromatin on the inactive X chromosome delays replication timing without affecting origin usage. Proc. Natl Acad. Sci. USA 101, 6923–6928 (2004). 40. Cohen, S. M., Brylawski, B. P., Cordeiro-Stone, M. & Kaufman, D. G. Same origins of DNA replication function on the active and inactive human X chromosomes. J. Cell Biochem. 88, 923–931 (2003). 41. Touchon, M. et al. Replication-associated strand asymmetries in mammalian genomes: toward detection of replication origins. Proc. Natl Acad. Sci. USA 102, 9836–9841 (2005). 42. Necsulea, A., Guillet, C., Cadoret, J. C., Prioleau, M. N. & Duret, L. The relationship between DNA replication and human genome organization. Mol. Biol. Evol. 26, 729–741 (2009). 43. Gilbert, D. M. Replication origin plasticity, Taylor-made: inhibition vs recruitment of origins under conditions of replication stress. Chromosoma 116, 341–347 (2007). 44. Friedman, K. L., Brewer, B. J. & Fangman, W. L. Replication profile of Saccharomyces cerevisiae chromosome VI. Genes Cells 2, 667–678 (1997). 45. Lebofsky, R., Heilig, R., Sonnleitner, M., Weissenbach, J. & Bensimon, A. DNA replication origin interference increases the spacing between initiation events in human cells. Mol. Biol. Cell 17, 5337–5345 (2006). 46. Hamlin, J. L. et al. A revisionist replicon model for higher eukaryotic genomes. J. Cell Biochem. 105, 321–329 (2008). 47. Pasero, P., Bensimon, A. & Schwob, E. Single-molecule analysis reveals clustering and epigenetic regulation of replication origins at the yeast rDNA locus. Genes Dev. 16, 2479–2484 (2002). 48. Mesner, L. D., Li, X., Dijkwel, P. A. & Hamlin, J. L. The dihydrofolate reductase origin of replication does not contain any nonredundant genetic elements required for origin activity. Mol. Cell. Biol. 23, 804–814 (2003). 49. Kalejta, R. F. et al. Distal sequences, but not ori-β/OBR‑1, are essential for initiation of DNA replication in the Chinese hamster DHFR origin. Mol. Cell 2, 797–806 (1998).
50. DePamphilis, M. L. Eukaryotic DNA replication: anatomy of an origin. Annu. Rev. Biochem. 62, 29–63 (1993). 51. Li, F., Chen, J., Solessio, E. & Gilbert, D. M. Spatial distribution and specification of mammalian replication origins during G1 phase. J. Cell Biol. 161, 257–266 (2003). 52. Jackson, D. A. & Pombo, A. Replicon clusters are stable units of chromosome structure: evidence that nuclear organization contributes to the efficient activation and propagation of S phase in human cells. J. Cell Biol. 140, 1285–1295 (1998). 53. Takebayashi, S. I., Manders, E. M., Kimura, H., Taguchi, H. & Okumura, K. Mapping sites where replication initiates in mammalian cells using DNA fibers. Exp. Cell Res. 271, 263–268 (2001). 54. Gomez, M. & Antequera, F. Overreplication of short DNA regions during S phase in human cells. Genes Dev. 22, 375–385 (2008). 55. Patel, P. K. et al. The Hsk1(Cdc7) replication kinase regulates origin efficiency. Mol. Biol. Cell 19, 5550–5558 (2008). 56. Krasinska, L. et al. Cdk1 and Cdk2 activity levels determine the efficiency of replication origin firing in Xenopus. EMBO J. 27, 758–769 (2008). 57. Wu, P. Y. & Nurse, P. Establishing the program of origin firing during S phase in fission Yeast. Cell 136, 852–864 (2009). 58. Wu, J. R. & Gilbert, D. M. A distinct G1 step required to specify the Chinese hamster DHFR replication origin. Science 271, 1270–1272 (1996). This study unravels a major event occurring during G1 that enables the localization of replication origins to be specified. 59. Lemaitre, J. M., Danis, E., Pasero, P., Vassetzky, Y. & Mechali, M. Mitotic remodeling of the replicon and chromosome structure. Cell 123, 1–15 (2005). Shows that mitosis has a big influence on the organization of the genome for DNA replication, allowing the remodelling of chromosome structure and DNA replication origin spacing of differentiated adult nuclei, in a reaction that is topoisomerase II-dependent. 60. Abdurashidova, G. et al. Functional interactions of DNA topoisomerases with a human replication origin. EMBO J. 26, 998–1009 (2007). 61. Courbet, S. et al. Replication fork movement sets chromatin loop size and origin choice in mammalian cells. Nature 455, 557–560 (2008). Shows the effect of growth conditions on the loop size and localization of DNA replication origins. 62. Dimitrova, D. S. & Gilbert, D. M. The spatial position and replication timing of chromosomal domains are both established in early G1 phase. Mol. Cell 4, 983–993 (1999). 63. Hiratani, I. et al. Global reorganization of replication domains during embryonic stem cell differentiation. PLoS Biol. 6, e245 (2008). 64. Schwaiger, M. et al. Chromatin state marks cell‑type‑ and gender-specific replication of the Drosophila genome. Genes Dev. 23, 589–601 (2009). References 63 and 64 are detailed genome-wide analyses of replication timing domains in mouse and D. melanogaster cells, respectively. 65. Branzei, D. & Foiani, M. The DNA damage response during DNA replication. Curr. Opin. Cell Biol. 17, 568–575 (2005). 66. Santocanale, C. & Diffley, J. F. A Mec1- and Rad53‑dependent checkpoint controls late-firing origins of DNA replication. Nature 395, 615–618 (1998). 67. Shirahige, K. et al. Regulation of DNA-replication origins during cell-cycle progression. Nature 395, 618–621 (1998). 68. Ibarra, A., Schwob, E. & Mendez, J. Excess MCM proteins protect human cells from replicative stress by licensing backup origins of replication. Proc. Natl Acad. Sci. USA 105, 8956–8961 (2008). 69. Ge, X. Q., Jackson, D. A. & Blow, J. J. Dormant origins licensed by excess MCM2–7 are required for human cells to survive replicative stress. Genes Dev. 21, 3331–3341 (2007). 70. Doksani, Y., Bermejo, R., Fiorani, S., Haber, J. E. & Foiani, M. Replicon dynamics, dormant origin firing, and terminal fork integrity after double-strand break formation. Cell 137, 247–258 (2009). References 68–70 emphasize the control of replication origin firing by replication stresses. 71. Shima, N. et al. A viable allele of MCM4 causes chromosome instability and mammary adenocarcinomas in mice. Nature Genet. 39, 93–98 (2007).
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
72. Norio, P. et al. Progressive activation of DNA replication initiation in large domains of the immunoglobulin heavy chain locus during b cell development. Mol. Cell 20, 575–587 (2005). 73. Abraham, R. T. Cell cycle checkpoint signaling through the ATM and ATR kinases. Genes Dev. 15, 2177–2196 (2001). 74. Zegerman, P. & Diffley, J. F. DNA replication as a target of the DNA damage checkpoint. DNA Repair (Amst) 8, 1077–1088 (2009). 75. Shechter, D., Costanzo, V. & Gautier, J. ATR and ATM regulate the timing of DNA replication origin firing. Nature Cell Biol. 6, 648–655 (2004). 76. Marheineke, K. & Hyrien, O. Control of replication origin density and firing time in Xenopus egg extracts: role of a caffeine-sensitive, ATR-dependent checkpoint. J. Biol. Chem. 279, 28071–28081 (2004). References 75 and 76 show the role of ATM and ATR on the regulation of DNA replication. 77. Petermann, E. et al. Chk1 requirement for high global rates of replication fork progression during normal vertebrate S phase. Mol. Cell Biol. 26, 3319–3326 (2006). 78. Seiler, J. A., Conti, C., Syed, A., Aladjem, M. I. & Pommier, Y. The intra‑S‑phase checkpoint affects both DNA replication initiation and elongation: single-cell and -DNA fiber analyses. Mol. Cell Biol. 27, 5806–5818 (2007). 79. Maya-Mendoza, A., Petermann, E., Gillespie, D. A., Caldecott, K. W. & Jackson, D. A. Chk1 regulates the density of active replication origins during the vertebrate S phase. EMBO J. 26, 2719–2731 (2007). 80. Aparicio, O. M., Stout, A. M. & Bell, S. P. Differential assembly of Cdc45p and DNA polymerases at early and late origins of DNA replication. Proc. Natl Acad. Sci. USA 96, 9130–9135 (1999). 81. Costanzo, V. et al. Reconstitution of an ATM-dependent checkpoint that inhibits chromosomal DNA replication following DNA damage. Mol. Cell 6, 649–659 (2000). 82. Liu, P. et al. The Chk1‑mediated S‑phase checkpoint targets initiation factor Cdc45 via a Cdc25A/Cdk2‑ independent mechanism. J. Biol. Chem. 281, 30631–30644 (2006). 83. Kumagai, A., Shevchenko, A. & Dunphy, W. G. Treslin collaborates with TopBP1 in triggering the initiation of DNA replication. Cell 140, 349–359 (2010). 84. Sansam, C. L. et al. A vertebrate gene, Ticrr, is an essential checkpoint and replication regulator. Genes Dev. 24, 183–194 (2010). 85. Balestrini, A., Cosentino, C., Errico, A., Garner, E. & Costanzo, V. GEMC1 is a TopBP1‑interacting protein required for chromosomal DNA replication. Nature Cell Biol. 12, 484–491 (2010). 86. Mordes, D. A., Glick, G. G., Zhao, R. & Cortez, D. TopBP1 activates ATR through ATRIP and a PIKK regulatory domain. Genes Dev. 22, 1478–1489 (2008). 87. Kumagai, A., Lee, J., Yoo, H. Y. & Dunphy, W. G. TopBP1 activates the ATR-ATRIP complex. Cell 124, 943–955 (2006). 88. Hashimoto, Y. & Takisawa, H. Xenopus Cut5 is essential for a CDK-dependent process in the initiation of DNA replication. EMBO J. 22, 2526–2535 (2003). 89. Kubota, Y. et al. A novel ring-like complex of Xenopus proteins essential for the initiation of DNA replication. Genes Dev. 17, 1141–1152 (2003). 90. Simpson, R. T. Nucleosome positioning can affect the function of a cis-acting DNA element in vivo. Nature 343, 387–389 (1990). 91. Lipford, J. R. & Bell, S. P. Nucleosomes positioned by ORC facilitate the initiation of DNA replication. Mol. Cell 7, 21–30 (2001). 92. Collins, N. et al. An ACF1‑ISWI chromatin-remodeling complex is required for DNA replication through heterochromatin. Nature Genet. 32, 627–632 (2002). 93. Field, Y. et al. Distinct modes of regulation by chromatin encoded through nucleosome positioning signals. PLoS Comput. Biol. 4, e1000216 (2008). 94. Aggarwal, B. D. & Calvi, B. R. Chromatin regulates origin activity in Drosophila follicle cells. Nature 430, 372–376 (2004). 95. Danis, E. et al. Specification of a DNA replication origin by a transcription complex. Nature Cell Biol. 6, 721–730 (2004). 96. Crampton, A., Chang, F., Pappas, D. L., Jr, Frisch, R. L. & Weinreich, M. An ARS element inhibits DNA replication through a SIR2‑dependent mechanism. Mol. Cell 30, 156–166 (2008). 97. Unnikrishnan, A., Gafken, P. R. & Tsukiyama, T. Dynamic changes in histone acetylation regulate origins of DNA replication. Nature Struct. Mol. Biol. 17, 430–437 (2010).
VOLUME 11 | O CTOBER 2010 | 737 © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS 98. Doyon, Y. et al. ING tumor suppressor proteins are critical regulators of chromatin acetylation required for genome expression and perpetuation. Mol. Cell 21, 51–64 (2006). 99. Miotto, B. & Struhl, K. HBO1 histone acetylase activity is essential for DNA replication licensing and inhibited by Geminin. Mol. Cell 37, 57–66 (2010). 100. Iizuka, M., Matsui, T., Takisawa, H. & Smith, M. M. Regulation of replication licensing by acetyltransferase Hbo1. Mol. Cell. Biol. 26, 1098–1108 (2006). 101. Gregoire, D., Brodolin, K. & Mechali, M. HoxB domain induction silences DNA replication origins in the locus and specifies a single origin at its boundary. EMBO Rep. 7, 812–816 (2006). 102. Dazy, S., Gandrillon, O., Hyrien, O. & Prioleau, M. N. Broadening of DNA replication origin usage during metazoan cell differentiation. EMBO Rep. 7, 806–811 (2006). 103. Aparicio, J. G., Viggiani, C. J., Gibson, D. G. & Aparicio, O. M. The Rpd3‑Sin3 histone deacetylase regulates replication timing and enables intra‑S origin control in Saccharomyces cerevisiae. Mol. Cell. Biol. 24, 4769–4780 (2004). 104. Vogelauer, M., Rubbi, L., Lucas, I., Brewer, B. J. & Grunstein, M. Histone acetylation regulates the time of replication origin firing. Mol. Cell 10, 1223–1233 (2002). 105. Knott, S. R., Viggiani, C. J., Tavare, S. & Aparicio, O. M. Genome-wide replication profiles indicate an expansive role for Rpd3L in regulating replication initiation timing or efficiency, and reveal genomic loci of Rpd3 function in Saccharomyces cerevisiae. Genes Dev. 23, 1077–1090 (2009). 106. Fox, C. A., Loo, S., Rivier, D. H., Foss, M. A. & Rine, J. A transcriptional silencer as a specialized origin of replication that establishes functional domains of chromatin. Cold Spring Harb. Symp. Quant. Biol. 58, 443–455 (1993). 107. Foss, M., McNally, F. J., Laurenson, P. & Rine, J. Origin recognition complex (ORC) in transcriptional silencing and DNA replication in S. cerevisiae. Science 262, 1838–1844 (1993). 108. Gavin, K. A., Hidaka, M. & Stillman, B. Conserved initiator proteins in eukaryotes. Science 270, 1667–1671 (1995). 109. Triolo, T. & Sternglanz, R. Role of interactions between the origin recognition complex and SIR1 in transcriptional silencing. Nature 381, 251–253 (1996). 110. Pak, D. T. et al. Association of the origin recognition complex with heterochromatin and HP1 in higher eukaryotes. Cell 91, 311–323 (1997). 111. Demeret, C., Bocquet, S., Lemaitre, J. M., Francon, P. & Mechali, M. Expression of ISWI and its binding to chromatin during the cell cycle and early development. J. Struct. Biol. 140, 57–66 (2002). 112. Hayashi, M. T., Takahashi, T. S., Nakagawa, T., Nakayama, J. & Masukata, H. The heterochromatin protein Swi6/HP1 activates replication origins at the pericentromeric region and silent mating-type locus. Nature Cell Biol. 11, 357–362 (2009). 113. Remus, D. & Diffley, J. F. Eukaryotic DNA replication control: lock and load, then fire. Curr. Opin. Cell Biol. 21, 771–777 (2009). 114. Christov, C. P., Gardiner, T. J., Szuts, D. & Krude, T. Functional requirement of noncoding Y RNAs for human chromosomal DNA replication. Mol. Cell. Biol. 26, 6993–7004 (2006). 115. Mohammad, M. M., Donti, T. R., Sebastian Yakisich, J., Smith, A. G. & Kapler, G. M. Tetrahymena ORC contains a ribosomal RNA fragment that participates in rDNA origin recognition. EMBO J. 26, 5048–5060 (2007). 116. Donti, T. R., Datta, S., Sandoval, P. Y. & Kapler, G. M. Differential targeting of Tetrahymena ORC to ribosomal DNA and non-rDNA replication origins. EMBO J. 28, 223–233 (2009). 117. Norseen, J. et al. RNA-dependent recruitment of the origin recognition complex. EMBO J. 27, 3024–3035 (2008). 118. Nakamura, H., Morita, T. & Sato, C. Structural organizations of replicon domains during DNA synthetic phase in the mammalian nucleus. Exp. Cell Res. 165, 291–297 (1986). 119. Ma, H. et al. Spatial and temporal dynamics of DNA replication sites in mammalian cells. J. Cell Biol. 143, 1415–1425 (1998). 120. Berezney, R., Dubey, D. D. & Huberman, J. A. Heterogeneity of eukaryotic replicons, replicon clusters, and replication foci. Chromosoma 108, 471–484 (2000).
An excellent review on the elusive replication cluster and replication foci. 121. Shaw, A., Olivares-Chauvet, P., Maya-Mendoza, A. & Jackson, D. A. S‑phase progression in mammalian cells: modelling the influence of nuclear organization. Chromosome Res. 18, 163–178 (2010). 122. Gilbert, D. M. & Gasser, S. M. In DNA Replication and Human Disease (ed. DePamphilis, M. L.) (Cold Spring Harbor Laboratory Press, 2006). 123. Takahashi, T. S., Yiu, P., Chou, M. F., Gygi, S. & Walter, J. C. Recruitment of Xenopus Scc2 and cohesin to chromatin requires the pre-replication complex. Nature Cell Biol. 6, 991–996 (2004). 124. Shimada, K. & Gasser, S. M. The origin recognition complex functions in sister-chromatid cohesion in Saccharomyces cerevisiae. Cell 128, 85–99 (2007). Together with references 29 and 123, this study establishes a link between cohesin and replication origins. 125. Adachi, Y. & Laemmli, U. K. Study of the cell cycledependent assembly of the DNA pre-replication centres in Xenopus egg extracts. EMBO J. 13, 4153–4164 (1994). 126. Francon, P. et al. A hypophosphorylated form of RPA34 is a specific component of pre-replication centers. J. Cell Sci. 117, 4909–4920 (2004). 127. Brewer, B. J. & Fangman, W. L. Initiation at closely spaced replication origins in a yeast chromosome. Science 262, 1728–1731 (1993). 128. Labit, H., Perewoska, I., Germe, T., Hyrien, O. & Marheineke, K. DNA replication timing is deterministic at the level of chromosomal domains but stochastic at the level of replicons in Xenopus egg extracts. Nucleic Acids Res. 36, 5623–5634 (2008). 129. White, E. J. et al. DNA replication-timing analysis of human chromosome 22 at high resolution and different developmental states. Proc. Natl Acad. Sci. USA 101, 17771–17776 (2004). 130. Heun, P., Laroche, T., Raghuraman, M. K. & Gasser, S. M. The positioning and dynamics of origins of replication in the budding yeast nucleus. J. Cell Biol. 152, 385–400 (2001). 131. Friedman, K. L. et al. Multiple determinants controlling activation of yeast replication origins late in S phase. Genes Dev. 10, 1595–1607 (1996). 132. Kohzaki, H. & Murakami, Y. Transcription factors and DNA replication origin selection. Bioessays 27, 1107–1116 (2005). 133. Gilbert, D. M. Replication timing and transcriptional control: beyond cause and effect. Curr. Opin. Cell Biol. 14, 377–383 (2002). 134. Schwaiger, M. & Schubeler, D. A question of timing: emerging links between transcription and replication. Curr. Opin. Genet. Dev. 16, 177–183 (2006). 135. Donaldson, A. D. Shaping time: chromatin structure and the DNA replication programme. Trends Genet. 21, 444–449 (2005). 136. Maric, C., Benard, M. & Pierron, G. Developmentally regulated usage of Physarum DNA replication origins. EMBO Rep. 4, 474–478 (2003). 137. Lunyak, V. V., Ezrokhi, M., Smith, H. S. & Gerbi, S. A. Developmental changes in the sciara II/9a initiation zone for DNA replication. Mol. Cell. Biol. 22, 8426–37 (2002). 138. Haase, S. B., Heinzel, S. S. & Calos, M. P. Transcription inhibits the replication of autonomously replicating plasmids in human cells. Mol. Cell. Biol. 14, 2516–2524 (1994). 139. Mesner, L. D. & Hamlin, J. L. Specific signals at the 3′ end of the DHFR gene define one boundary of the downstream origin of replication. Genes Dev. 19, 1053–1066 (2005). 140. Sasaki, T. et al. The Chinese hamster dihydrofolate reductase replication origin decision point follows activation of transcription and suppresses initiation of replication within transcription units. Mol. Cell. Biol. 26, 1051–1062 (2006). 141. Saha, S., Shan, Y., Mesner, L. D. & Hamlin, J. L. The promoter of the Chinese hamster ovary dihydrofolate reductase gene regulates the activity of the local origin and helps define its boundaries. Genes Dev. 18, 397–410 (2004). 142. Ghosh, M., Liu, G., Randall, G., Bevington, J. & Leffak, M. Transcription factor binding and induced transcription alter chromosomal c‑Myc replicator activity. Mol. Cell. Biol. 24, 10193–10207 (2004). 143. Beall, E. L. et al. Role for a Drosophila Myb-containing protein complex in site-specific DNA replication. Nature 420, 833–837 (2002).
738 | O CTOBER 2010 | VOLUME 11
144. Bosco, G., Du, W. & Orr-Weaver, T. L. DNA replication control through interaction of E2F‑RB and the origin recognition complex. Nature Cell Biol. 3, 289–295 (2001). 145. Kunst, F. et al. The complete genome sequence of the Gram-positive bacterium Bacillus subtilis. Nature 390, 249–256 (1997). 146. Mechali, M. DNA replication origins: from sequence specificity to epigenetics. Nature Rev. Genet. 2, 640–645 (2001). 147. Blumenthal, A. B., Kriegstein, H. J. & Hogness, D. S. The units of DNA replication in Drosophila melanogaster chromosomes. Cold Spring Harb. Symp. Quant. Biol. 38, 205–223 (1974). 148. Sasaki, T., Sawado, T., Yamaguchi, M. & Shinomiya, T. Specification of regions of DNA replication initiation during embryogenesis in the 65‑kilobase DNApolαdE2F locus of Drosophila melanogaster. Mol. Cell. Biol. 19, 547–555 (1999). 149. Buongiorno-Nardelli, M., Micheli, G., Carri, M. T. & Marilley, M. A relationship between replicon size and supercoiled loop domains in the eukaryotic genome. Nature 298, 100–102 (1982). 150. Desprat, R. et al. Predictable dynamic program of timing of DNA replication in human cells. Genome Res. 19, 2288–2299 (2009). 151. Hiratani, I. et al. Genome-wide dynamics of replication timing revealed by in vitro models of mouse embryogenesis. Genome Res. 20, 155–169 (2010). 152. Pinto, S. et al. Latheo encodes a subunit of the origin recognition complex and disrupts neuronal proliferation and adult olfactory memory when mutant. Neuron 23, 45–54 (1999). 153. Seo, S. et al. Geminin regulates neuronal differentiation by antagonizing Brg1 activity. Genes Dev. 19, 1723–1734 (2005). 154. Luo, L., Yang, X., Takihara, Y., Knoetgen, H. & Kessel, M. The cell-cycle regulator geminin inhibits Hox function through direct and polycomb-mediated interactions. Nature 427, 749–753 (2004). 155. Del Bene, F., Tessmar-Raible, K. & Wittbrodt, J. Direct interaction of geminin and Six3 in eye development. Nature 427, 745–749 (2004). 156. Caro, E., Castellano, M. M. & Gutierrez, C. A chromatin link that couples cell division to root epidermis patterning in Arabidopsis. Nature 447, 213–217 (2007). 157. Comelli, L. et al. The homeotic protein HOXC13 is a member of human DNA replication complexes. Cell Cycle 8, 454–459 (2009). 158. Egli, D., Rosains, J., Birkhoff, G. & Eggan, K. Developmental reprogramming after chromosome transfer into mitotic mouse zygotes. Nature 447, 679–685 (2007). 159. Remus, D. et al. Concerted loading of MCM2–7 double hexamers around DNA during DNA replication origin licensing. Cell 139, 719–730 (2009). 160. Lutzmann, M., Maiorano, D. & Mechali, M. A Cdt1‑geminin complex licenses chromatin for DNA replication and prevents rereplication during S phase in Xenopus. EMBO J. 25, 5764–5774 (2006). 161. Lutzmann, M. & Mechali, M. MCM9 binds Cdt1 and is required for the assembly of prereplication complexes. Mol. Cell 31, 190–200 (2008). 162. Ilves, I., Petojevic, T., Pesavento, J. J. & Botchan, M. R. Activation of the MCM2–7 helicase by association with Cdc45 and GINS proteins. Mol. Cell 37, 247–258 (2010). 163. Pacek, M., Tutter, A. V., Kubota, Y., Takisawa, H. & Walter, J. C. Localization of MCM2–7, Cdc45, and GINS to the site of DNA unwinding during eukaryotic DNA replication. Mol. Cell 21, 581–587 (2006).
Acknowledgements
I would like to thank E. Andermarcher for critical reading of this manuscript. The M.M laboratory is supported by the European Research Council (ERC advanced grant FP7/2007-2013, Grant Agreement no. 233339), the ANR, the ‘Association pour la Recherche contre le Cancer’, and the ‘Ligue Nationale contre le Cancer’.
Competing interests statement
The authors declare no competing financial interests.
FURTHER INFORMATION Marcel Méchali’s homepage: http://www.igh.cnrs.fr/equip/mechali/ ALL LINKS ARE ACTIVE IN THE ONLINE PDF
www.nature.com/reviews/molcellbio © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS
Non-vesicular lipid transport by lipid-transfer proteins and beyond Sima Lev
Abstract | The movement of lipids within and between intracellular membranes is mediated by different lipid transport mechanisms and is crucial for maintaining the identities of different cellular organelles. Non-vesicular lipid transport has a crucial role in intracellular lipid trafficking and distribution, but its underlying mechanisms remain unclear. Lipid-transfer proteins (LTPs), which regulate diverse lipid-mediated cellular processes and accelerate vectorial transport of lipid monomers between membranes in vitro, could potentially mediate non-vesicular intracellular lipid trafficking. Understanding the mechanisms by which lipids are transported and distributed between cellular membranes, and elucidating the role of LTPs in intracellular lipid transport and homeostasis, are currently subjects of intensive study. Vesicular transport An active process in which materials move into or out of the cell enclosed in vesicles. This process is mediated by a sequence of events involving the budding of the vesicles from a donor membrane and their subsequent fusion with an acceptor membrane.
Peroxisome An organelle present in most eukaryotic cells that is involved in the oxidation of fatty acids and the production and destruction of hydrogen peroxide.
Molecular Cell Biology Department, Weizmann Institute of Science, Rehovot, Israel. e‑mail:
[email protected] doi:10.1038/nrm2971 Published online 8 September 2010
Eukaryotic cells are organized into separate membrane-bound compartments or organelles, each with a specialized function and unique protein and lipid composition 1. This compartmentalization ensures efficient segregation of diverse metabolic processes mediated by distinct sets of enzymes, regulatory and structural proteins. The distribution of proteins among cellular organelles is often mediated by specific protein-targeting motifs and can be regulated by chemical modifications and/or conformational changes. By contrast, lipids lack any intrinsic motifs that mediate their distinct intracellular distribution; nevertheless, different cellular membranes vary in their lipid composition2,3. For example, the plasma membrane, which exhibits a transverse lipid compositional asymmetry, is enriched in sphingolipids and sterols, whereas the endoplasmic reticulum (ER), which displays a symmetrical transbilayer lipid distribution, contains low levels of both lipids4,5. Eukaryotic cells contain more than 1,000 chemically distinct lipid species6, which can be divided into three major classes: glycerophospholipids, sphingolipids and sterols2. It is not fully understood how these lipids are delivered to their target destinations. Increasing lines of evidence, however, suggest that intracellular lipid trafficking is mediated by both vesicular transport and non-vesicular transport mechanisms7. Vesicular transport plays a major part in protein transport along the exocytic and endocytic pathways, and requires metabolic energy and an intact cytoskeleton. Given that lipids are the basic constituents of transport vesicles, large amounts
of lipids must be transported between organelles by vesicular transport. Nevertheless, lipid transport has been detected under conditions in which vesicular transport was blocked by either ATP depletion, reduction in temperature or treatment with specific pharmacological drugs (such as brefeldin A and colchicine)8,9. Lipid transport has also been seen between organelles that are not connected to the vesicular transport machinery (for example, mitochondria and peroxisomes)10,11. These observations suggest that non-vesicular transport mechanisms have an important role in intracellular lipid trafficking. Furthermore, current studies suggest that intracellular lipid trafficking is greatly facilitated at membrane contact sites (MCSs)12–14. These sites are defined as small cytosolic gaps (10–20 nm) between the membranes of the ER and virtually all cellular organelles and enable the transport of Ca2+, metabolites and lipids by a non-vesicular transport mechanism11,15. Non-vesicular lipid transport could occur, in principle, by spontaneous lipid desorption of a lipid monomer from a bilayer and its free diffusion through the cytosol, but this process is slow and not sufficient to support substantial transport of most lipids16,17. Lipid-transfer proteins (LTPs) can greatly facilitate lipid transport between membranes in vitro18,19. LTPs were initially discovered as soluble factors that accelerate the exchange or net transfer of different lipid species between membranes in vitro20. Since then, many LTPs have been isolated, cloned and crystallized. LTPs have been identified in all eukaryotes, in plants and in bacteria, and have been subdivided into different protein families according to
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
VOLUME 11 | O CTOBER 2010 | 739 © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS Spontaneous transfer
LTP-mediated transfer
Lateral diffusion
Lid closed Release
Exchange Lid open
Flip-flop
Figure 1 | Modes of non-vesicular lipid transport. There are three mechanisms Nature Reviews | Molecular Cell Biology involved in non-vesicular lipid transport: monomeric lipid exchange, lateral diffusion and transbilayer flip-flop. Monomeric lipid exchange can be spontaneous or mediated by lipid-transfer proteins (LTPs). LTPs can transfer a lipid to the acceptor membrane or exchange it with a lipid of the acceptor membrane. Lateral diffusion occurs in the lateral plane of the membrane (at 0.1–1 μm per second)10. Transbilayer flip-flop can be either spontaneous (not shown) or mediated by proteins such as flippases and translocases.
Lipid desorption The release of a lipid molecule from a lipid bilayer to the surrounding aqueous phase. This process involves both the disruption of lipid–lipid interactions in the bilayer and the formation of a cavity in the aqueous phase that accommodates the diffusing lipid molecule.
Vectorial lipid transport A directional transport of lipids that is driven by a concentration gradient.
Lipid droplet An organelle that stores neutral lipids and has a crucial role in lipid metabolism.
Flippase A membrane protein that catalyses the transport of lipids across the membrane bilayer in an ATP-dependent manner. Flippases commonly transport lipids towards the cytoplasm, whereas floppases transport lipids from the cytofacial surface to the opposite side of the membrane.
Membrane curvature The bending of the membrane, which can be influenced by the relative distribution of cone-like and inverted cone-like lipids (for example, diacylglcerol and phosphatidic acid, and lysophospholipids, respectively) between the inner or outer leaflets of the bilayer.
their sequence and structure similarity 21. It seems that most LTPs can bind lipid monomers in a hydrophobic pocket and transfer hydrophobic lipids through an aqueous phase. Although LTPs have been extensively studied over the past 30 years, their modes of action in intact cells have not been fully explored. Increasing lines of evidence suggest that LTPs do not mediate a simple vectorial lipid transport from one membrane to another. Instead, they facilitate lipid transport between membranes according to their membrane environment. LTPs can, therefore, locally modulate the lipid composition of membranes and consequently regulate various cellular processes, including vesicular trafficking, signal transduction and lipid metabolism22–24. This Review discusses the latest advances in our understanding of non-vesicular lipid trafficking, the mechanisms by which LTPs act in vitro and their diverse modes of action in intact cells.
Non-vesicular transport of lipids Lipid movement in and between biological membranes is mediated by three mechanisms: monomeric lipid exchange, lateral diffusion and transbilayer flip-flop4,6 (FIG. 1). Transport of lipids between membranes is mainly mediated by monomeric lipid exchange, in which a lipid molecule is transported through an aqueous phase from the outer leaflet of a donor membrane to the outer leaflet of an acceptor membrane. This process does not require metabolic energy and can be either spontaneous or mediated by LTPs. Lateral diffusion is the process by which lipids move in the lateral plane of the membrane bilayer. Although it mainly occurs within membranes, it can mediate lipid transport between membranes that are connected by membrane bridges. Membrane continuities have been seen between the plasma membrane and lipid droplets, the ER and the outer mitochondrial membrane, and lipid droplets and the outer mitochondrial membrane, among others6,25.
Transbilayer flip-flop is the process by which lipids are moved between the two leaflets of the membrane bilayer either spontaneously or with the assistance of proteins such as flippases and translocases2. In some cases, protein-assisted flip-flop requires metabolic energy. Although transbilayer flip-flop is not directly involved in lipid transport between different organelles, it may indirectly influence inter-organelle lipid transport mediated by monomeric lipid exchange or vesicular transport. The flip-flop of lipids from the inner to the outer leaflet would enable monomeric lipid exchange. Furthermore, flip-flop of lipids with either a small or a non-polar headgroup, such as phosphatidic acid or diacylglycerol, could markedly affect membrane curvature, and consequently vesicle budding, vesicle fission and vesicle fusion2,26. Thus, the interplay between vesicular and non-vesicular lipid transport is crucial for the establishment and maintenance of intracellular lipid distribution.
Monomeric lipid exchange Spontaneous movement of lipids between membranes. Although most cellular lipids are highly insoluble in water, numerous studies have shown that lipid monomers can spontaneously move between membranes27–31. For most classes of lipids, this process is extremely slow (FIG. 2a,b), with half-times on the order of days. The rate of spontaneous lipid exchange usually correlates with their aqueous-phase solubility 29,32. Lyso-phosphatidylcholine (Lyso-PtdCho), for example, which has only one acyl chain, exchanges more rapidly than PtdCho. Cholesterol, which is more soluble than PtdCho, exchanges much faster between synthetic vesicles with a similar lipid composition. It was estimated that six molecules of cholesterol and one molecule of PtdCho are transferred between such vesicles in 2 minutes28. The lipid exchange rate is also influenced by membrane curvature, and indeed cholesterol exchanges more rapidly from small donor vesicles with high membrane curvature than vesicles with low membrane curvature33. However, the interaction of cholesterol with phospholipids in the bilayer can form condensed complexes, which markedly attenuate its membrane desorption34. Based on kinetic measurements, three types of spontaneous lipid exchange mechanisms have been proposed35. The first mechanism is aqueous diffusion, which occurs at low membrane concentrations and is determined by the concentration of donor, but not acceptor membranes30. In this case, a lipid desorbs out of one bilayer, enters the aqueous phase, diffuses across it and inserts into a second bilayer (FIG. 2c). This is a first-order process, in which lipid desorption from the bilayer is the rate-limiting step27,29,36. The second possible mechanism occurs at high membrane concentrations as a result of lipid–membrane collision16 and is dependent on the concentration of both donor and acceptor membranes (FIG. 2d). This is a second-order process, in which the rates of lipid desorption from donor membrane and its collision with the acceptor membrane are the first- and second-order rate constants, respectively.
740 | O CTOBER 2010 | VOLUME 11
www.nature.com/reviews/molcellbio © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS Half-time transfer Lyso-PtdCho (16:0) 30 sec PtdCho (14:0, 14:0)
2.0 h
PtdCho (16:0, 18:1)
60 h
PtdCho (16:0, 16:0)
83 h
c Aqueous diffusion
b Hydrophobicity
Hydrophobicity
a
Half-time transfer 25OH cholesterol 2 min Cholesterol Cholesterol oleate
2h 107 h
d Collision
e Activated collision
Figure 2 | Spontaneous lipid exchange between membranes. a, b | Rates of Nature Reviews | Molecular Cell Biology exchange of various phosphatidylcholine (PtdCho; a) or sterol (b) species correlate 29,30,110 with their aqueous-phase solubility . a | Lyso-PtdCho, which has only one acyl chain, exchanges most rapidly. Because most PtdCho in cells has acyl chains with 16 or more carbons, spontaneous PtdCho exchange between membranes is probably not physiologically relevant. The ratios in brackets refer to the length of the fatty acid chain at sn1 and sn2 positions of the phospholipids (14:0, myristic acid; 16:0, palmitic acid; 18:0, stearic acid; 18:1, oleic acid). The structures of the lipids can be found in Supplementary information S3 (figure). b | 25‑hydroxylcholesterol (25OH) is more hydrophilic than cholesterol and therefore exchanges more rapidly, whereas cholesteryl oleate is more hydrophobic than cholesterol and exchanges more slowly. c–e | Mechanisms of spontaneous lipid exchange. c | During aqueous diffusion, a lipid desorbs out of one bilayer and diffuses across an aqueous phase to insert into a different bilayer. d | Lipids can also transfer during the collision of two membranes. e | Activated collision can occur when a lipid partially extends from the bilayer, increasing the probability of transfer.
Vesicle fission The pinching-off of a vesicle from a membrane bilayer.
Vesicle fusion The merging of a vesicle with a membrane bilayer.
Condensed complex A complex formed between cholesterol and saturated phospholipids with long fatty acid chains or with sphingomyelin.
First-order process A reaction with a rate that is proportional to the concentration of only one reactant. Other reactants can be present but have no influence on the reaction rate.
The third mechanism is activated collision (FIG. 2e). In this case, a lipid partially extends from the bilayer (and is referred to as activated), increasing the probability of its transfer to a second membrane during collision 37 by decreasing the energy required for transfer. Lipid activation can be stochastic, as thermal motion probably drives lipids partially in and out of membranes. The activated collision model is consistent with studies showing that hydration forces between liposomes prevent them from getting closer than about 20 Å. This suggests that lipids can only be exchanged during membrane collision if they protrude from the bilayer. Collisional lipid transfer is a concentration-dependent process (see Supplementary information S1 (table) for an example of the effect of concentration on transfer); it could, therefore, occur in intact cells owing to a high concentration (~40 mM) of cellular membranes16,38.
LTPs as accelerators of lipid transport in vitro. The slow rate of spontaneous lipid exchange between membranes can be accelerated by LTPs, possibly by increasing the rate of lipid desorption from membranes39. This could be accomplished by decreasing the energy barrier for the lipid monomer–membrane equilibrium reaction, thereby facilitating the dissociation rate of lipid monomers from membranes40. Thus, LTPs may act as both lipid carriers and catalysts for monomeric exchange41. LTPs carry a lipid monomer in their hydrophobic pocket and show specificity for one or more lipid types. Nonspecific LTP (NSLTP; also known as SCP2), for example, transfers all common diacyl phospholipids, glycolipids and cholesterol19,38, whereas ceramide-transfer protein (CERT) displays high specificity towards natural ceramide with C14–C20 but not longer acyl chains42, thus showing specificity to both the lipid headgroup and the backbone. According to their lipid-binding specificity and transfer capability, LTPs can be grouped into three main classes: phospholipid‑, sterol- and sphingolipid-transfer proteins (FIG. 3a). In mammals, PtdCho‑transfer protein (PCTP), phosphatidylinositol (PtdIns)-transfer protein (PITP) and NSLTP are the three main classes of phospholipidtransfer proteins. Steroidogenic acute regulatory protein (STAR) is an example of a cholesterol-transfer protein43, and oxysterol-binding protein (OSBP) and OSBP-related proteins (ORPs) are considered to be either sterol-sensing and/or sterol-transfer proteins. Examples of sphingolipidtransfer proteins include CERT and FAPP2 (also known as PLEKHA8)44. LTPs can contain a single structural lipid-transfer domain (LTD) or can have additional structural domains with varying functions (FIG. 3a). Based on the structure of the LTD, several LTP families have been defined, including SEC14, PITP, STAR-related lipid transfer (START), glycolipid-transfer protein (GLTP), SCP2, OSBP and ORP21. Mechanisms of LTP action. LTPs have been extensively studied using different in vitro lipid-transfer assays (see Supplementary information S2 (box)). Together with molecular, biophysical and crystallographic approaches, these studies have shed light on their mechanisms of action. The crystal structures of many LTPs have been resolved both in the presence and absence (known as the apo form) of their lipid ligands. In general, the structures are dominated by β‑sheet motifs, such as β‑barrels, β‑cups and β‑grooves. Several α‑helices (2–4) are closely packed with central β‑sheet motifs and together form a hydrophobic binding tunnel (FIG. 3b–d). The shape and size of the tunnel dictate the specificity for a cognate lipid ligand. A network of hydrophobic interactions and hydrogen bonds stabilize the lipid binding and affect the binding affinity. Typically, a ‘lid’ covers the hydrophobic tunnel and acts as a gate for lipid uptake and release45,46. The interaction of an LTP with membranes is thought to induce conformational changes, leading to displacement of the lid from the lipid-binding cavity and, consequently, tunnel opening.
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
VOLUME 11 | O CTOBER 2010 | 741 © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS This simple mechanism raises two important questions. How does an LTP associate with and dissociate from the donor and acceptor membranes? What are the mechanisms that dictate the direction, specificity and efficiency of lipid transport events? PITP FFAT DDHD LNS2 NIR2 1,244 aa The lipid composition and biophysical properties of the membranes (curvature and membrane fluidity) START directly affect the kinetic parameters and rate constants PCTP 214 aa of the transfer reactions19. High transfer activities are Thiolase SCP2 Trp203 Trp204 generally seen with more fluid or highly curved memLipid-exchange loop NSLTP 574 aa branes. Under such conditions, the interactions between c Osh4 d CERT (Cholesterol-bound) (C16-ceramide-bound) the lipid and the bilayer are much weaker and, there Sterol-transfer proteins fore, the absorption of the lipid to the LTP, the most FFAT ORD PH OSBP1 807 aa energy-consuming step, is more efficient49. Furthermore, membrane curvature can also influence the dissociation START STAR 285 aa rate of the LTP from the membrane. PCTP, for example, dissociates 100‑fold faster from highly curved vesicles Sphingolipid-transfer proteins than liposomes and so transfers PtdCho between such PH FFAT START vesicles more efficiently than between liposomes50. 624 aa CERT In contrast to the membrane-association steps, which mostly depend on physical parameters such GLTP PH Trp473 Lid as membrane charge, curvature and size38,41, the lipid 519 aa FAPP2 Trp562 absorption step mainly depends on the lipid type and Figure 3 | Domain organization and three-dimensional structure of LTPs. its surface concentration. Indeed, LTPs can mediate a a | The domain organization of representatives of theNature three main classes of human Reviews | Molecular Cell Biology net lipid transfer between donor and acceptor memlipid-transfer proteins (LTPs). The lipid-transfer domains are shown in green. b | A ribbon branes down the lipid concentration gradient 51. PITPs, diagram of the open (left panel; protein data bank (PDB) code 1KCM)111 and closed phosfor example, which exhibit a dual specificity for both phatidylcholine (PtdCho)‑bound (right panel; PDB code 1T27)112 conformation of phosphatidylinositol (PtdIns) transfer protein-α (PITPα). The PtdCho- and PtdIns-binding site is PtdIns and PtdCho, can transfer PtdIns to membranes formed by the concave surface of a central eight-stranded β‑sheet and two α-helices. with low PtdIns content in return for PtdCho, and can The lipid-exchange loop and the carboxy‑terminal tail act as a lid. Trp203 and Trp204 are thereby mediate a net PtdIns transfer according to the crucial for membrane association and lipid-transfer activity57,58. c | The oxysterol-related PtdIns concentration gradient 52. Furthermore, PCTP, domain (ORD) of oxysterol-binding protein (OSBP) homologue 4 (Osh4) consists of a which generally mediates a PtdCho‑exchange reaction, β‑barrel (17 strands) and three α‑helices (PDB code 1ZHX). A short amino-terminal region can also mediate a net transport of PtdCho to acceptor forms a lid over the tunnel. The 3‑hydroxyl of cholesterol is buried at the bottom of the membranes that lack PtdCho53. Thus, the lipid compositunnel46. d | A ribbon representation of the steroidogenic acute regulatory protein tion of the donor and acceptor membranes has a crucial (STAR)-related lipid transfer (START) domain of ceramide transfer protein (CERT) in role in dictating the direction of the lipid transport event complex with C16-ceramide (PDB code 2E3P). The overall structure contains two and the rate of lipid desorption from the bilayer 38,54,55. α‑helices at the N and C termini, separated by nine β‑strands and two shorter α‑helices. Trp473 and Trp562 are crucial for membrane association59. PCTP, PtdCho-transfer However, the association and dissociation of LTPs with protein; PH, pleckstrin homology; NSLTP, nonspecific LTP (also known as SCP2). and from the membranes also influence the efficiency of transport. GLTP, for example, which is positively charged at neutral pH, strongly interacts with negatively Thus, LTPs may exist in two distinct conformations: charged donor membranes. This strong electrostatic a ‘closed’ conformation, which reflects a transport- interaction slows its dissociation from a donor surface competent conformation in which one lipid molecule is and apparently diminishes GLTP-mediated transfer 56. Second-order process enclosed in the tunnel; and an ‘open’, membrane-bound Currently, the mechanisms responsible for the A reaction with a rate that conformation45,46 (FIG. 3b). In the closed conformation, the targeting of LTPs to specific donor or acceptor memis proportional to the square lid is often stabilized by hydrogen bonds and hydrophobic branes are not clear. Nevertheless, studies over the past concentration of a single reactant or to the concentration interactions with the rest of the protein. Polar interactions few years suggest that many LTPs interact with memof two reactants. between the lid and membrane phospholipids could facil- branes through specific protein motifs. Membraneitate the opening of the tunnel and stabilize membrane binding sites containing a Trp residue, which has a Thermal motion anchoring. Indeed, truncation of the lid in various LTPs high propensity for membrane interaction, have been The random motion of lipid molecules in the bilayer that is markedly affects their ability to interact with membranes identified in several LTPs, including PITPs57,58 (FIG. 3b), due to temperature. and abolishes their lipid-transfer activity 47,48. Hence, CERT59 (FIG. 3d) and GLTPs56. These Trp residues are the transport of lipids by LTPs could be mediated by crucial for membrane binding and therefore for lipid Hydration force sequential events involving the interaction with a donor transport activity in vitro 57– 59, but were proposed The repulsive force acting membrane and the opening of the lipid-binding tunnel, to mediate nonspecific binding with membranes 60. between apposing lipid bilayers in aqueous solution. extraction of a lipid from the bilayer, dissociation from Distinct binding sites may contribute to the specificity the donor membrane and diffusion through the aqueous of LTP–membrane interactions. These binding sites Liposome phase in a closed conformation. The transport is termin could be found either on the LTD or in other strucAn artificial microscopic vesicle ated by interaction with an acceptor membrane, tunnel tural domains of large LTPs. For example, the LTD of consisting of an aqueous core surrounded by a lipid bilayer. opening and lipid desorption. OSBP homologue 4 (Osh4; also known as Kes1), a yeast a
Phospholipid-transfer proteins PITP PITPα 271 aa
b PITPα
PITPα (PtdCho-bound)
742 | O CTOBER 2010 | VOLUME 11
www.nature.com/reviews/molcellbio © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS ORP, contains at least two membrane-binding surfaces that bind phosphoinositides, one near the mouth of the sterol-binding pocket and another at a distal site that binds PtdIns‑4,5‑bisphosphate (PtdIns(4,5)P2)61. It has recently been shown that the transport of sterol by Osh4 is enhanced when PtdIns(4,5)P2 is present on the acceptor membrane. PtdIns(4,5)P2 may, therefore, enhance the directional transport of sterol from PtdIns(4,5)P2poor to PtdIns(4,5)P2-rich membranes. Likewise, PtdIns‑4‑phosphate (PtdIns4P), which binds the pleckstrin homology domain (PH domain) of OSBP62, stimulates in vitro transport of sterol by OSBP when present in the donor membranes63. These examples show how the transport of specific lipid monomers by LTPs can be influenced by the relative distribution of other lipid species in the acceptor and donor membranes, which can bind to specific lipid-binding motifs on the LTPs. Thus, membrane-binding determinants on the LTPs, which may interact with specific membrane lipids and/or proteins, could affect the membrane specificity and influence the directionality of the lipid transport events. Transport efficiency can be greatly enhanced by close proximity between the donor and the acceptor membranes, as has been recently shown for Osh4‑mediated sterol transport in vitro61. This close apposition could reduce the diffusion distance between donor and acceptor membranes or even enable simultaneous interaction with the two membranes8. Such close proximity between different membrane compartments can be conferred by MCSs and has been identified in intact cells64. Membrane fluidity The viscosity of the lipid bilayer, which is determined by the length and saturation of the fatty-acid side chains of phospholipids and the content of cholesterol and sphingolipids..
Phosphoinositide The phosphorylated form of PtdIns. The inositol ring of PtdIns can be phosphorylated in three different positions (3, 4 and 5), yielding seven distinct phosphoinositides. Phosphoinositides play a key part in signal transduction and membrane trafficking.
PH domain A protein domain of ~100 amino acids that is present in numerous proteins and in many cases binds phosphoinositides with high affinity and specificity.
FFAT motif A short sequence motif, containing the EFFDAxE consensus sequence, that has been identified in 17 eukaryotic proteins, most of which are involved in lipid transfer, sensing or binding.
Lipid transport at MCSs MCSs are currently recognized as dynamic structures that facilitate non-vesicular lipid transport and are involved in the regulation of cellular lipid and Ca2+ homeostasis11,64,65. MCSs have been identified in all eukaryotes by morphological and biochemical studies, and some have been physically isolated. These include the mitochondrion-associated membrane (MAM) fraction, which contains unique regions of ER membranes attached to the outer mitochondrial membrane, and the plasma membrane-associated membrane (PAM) fraction, which contains many types of intracellular membrane (mainly from the ER and mitochondria) that co-isolate with the plasma membrane. In addition to MAM and PAM, MCSs have been identified between the ER membranes and those of the Golgi apparatus, vacuoles, peroxisomes, lipid droplets, late endosomes and lysosomes11,64. Despite their heterogeneity (TABLE 1) , different MCSs share several common features: they are typically enriched in proteins involved in lipid biosynthesis and trafficking, they are formed and/or stabilized by the tethering of apposing membranes through protein–protein or protein–lipid interactions and they are dynamic structures that can be formed and/or stabilized in response to different physiological conditions15,66. Although the molecular mechanisms underlying MCS formation and/or stabilization have not been fully characterized, current efforts have led to the identification of tethering complexes and bridging
proteins that are involved in MCS formation and/or stabilization66–75 (TABLE 1). Such tethers can affect the lipid composition of their tethered membranes and/or Ca2+ transport between the membranes. For example, the ER–mitochondrion encounter structure (ERMES) complex, which acts as an ER–mitochondrion tether in yeast, affects the lipid composition of the mitochondrial membrane67. Mitofusin 2, which has been implicated in ER–mitochondrion tethering in mammals, is required for Ca2+ uptake into mitochondria68 (FIG. 4a). Lipid transport at MCSs can occur by both the spontaneous and LTP-mediated routes. The production of phosphatidylethanolamine (PtdEtn) from phosphatidyl serine (PtdSer) at the MAM fraction is thought to be mediated by spontaneous lipid transport 15,66 (FIG. 4a,b). In mammalian cells, the formation of PtdSer occurs through the exchange of l‑Ser with either the choline moiety of PtdCho or the ethanolamine moiety of PtdEtn, and is catalysed by two different enzymes, PtdSer synthase 1 (PSS1) and PSS2, respectively 76 (FIG. 4c,d). These enzymes are highly enriched in the MAM fraction. The newly formed PtdSer at the MAM is transported into the mitochondrial intermembrane space, where it undergoes decarboxylation by PtdSer decarboxylase to produce PtdEtn. When the mitochondrial PtdSer decarboxylase 1 is inhibited, for example by hydroxylamine, accumulation of PtdSer in the MAM fraction is observed, indicating that transport of PtdSer from ER to mitochondria occurs through the MAM compartment 77. The decarboxylation of PtdSer in the mitochondria seems to account for most PtdEtn synthesis in mammalian cells. PtdEtn is then exported from mitochondria back to the ER. Hence, the transport of PtdSer from the ER to mitochondria and PtdEtn back from mitochondria to the ER represents bidirectional non-vesicular lipid transport and demonstrates the importance of MCSs in regulating cellular lipid biosynthesis and distribution. Several LTPs have been localized to MCSs and implicated in diverse functions (TABLE 1). OSBP-related protein 1L (ORP1L), for example, was implicated in the formation of ER–late endosome MCSs under low cholesterol conditions75. Osh1, which is recruited to nucleus– vacuole junctions (NVJs) by direct interaction with the outer nuclear membrane protein Nvj1, may regulate the lipid composition of NVJs in yeast 70. It thus seems that many LTPs function at MCSs. Perhaps the most convincing evidence of LTPmediated lipid transport at MCSs has emerged from studies of CERT-mediated ceramide transport at ER–Golgi MCSs 12,78. In mammalian cells, ceramide is synthesized in the ER and primarily transported by CERT to the Golgi complex, where it is converted into sphingomyelin by sphingomyelin synthase (SMS). It was proposed that CERT efficiently transports ceramide at ER–Golgi MCSs owing to its dual membrane targeting determinants: the PH domain, which binds PtdIns4P and mediates its interaction with the Golgi complex, and an FFAT motif (two Phe residues in an acidic tract)79, which interacts with the ER membrane proteins of the VAMP-associated protein (VAP) family (VAPA and VAPB)80, thereby mediating interaction with the ER14.
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
VOLUME 11 | O CTOBER 2010 | 743 © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS Table 1 | Tethering of MCSs Organelles
Lipid-modifying or binding proteins
Tethering complexes
Comments
Refs
ER and mitochondria (MAM)
PSS1, PSS2, FACL4 and acyl-CoA synthetase
ERMES: Mmm1* and Mdm12 (ER), Mdm10 and Mdm34 (mitochondrion)
ERMES was identified by a synthetic biological screen in yeast as a molecular tether between the ER and mitochondria
67
MFN: MFN2* (ER), MFN1* and MFN2* (mitochondrion)
MFN1 and MFN2 are dynamin-like GTPases in the OMM that mediate mitochondrial fusion; a fraction of MFN2 is also present in the ER membrane and tethers the ER membrane to mitochondria by interacting with MFN1 or MFN2 on the OMM
68
IP3R* (ER), VDAC* and GRP75 (mitochondrion)
VDAC in the OMM is physically linked to the ER Ca2+-release channel IP3R by interacting with GRP75, which enhances mitochondrial Ca2+ uptake
69
PACS2: calnexin‡ (ER) and unknown (mitochondrion)
PACS2 is a cytosolic multifunctional sorting protein that mediates ER–mitochondrion communication by an unknown mechanism
66
The autophagic process PMN, which degrades portions of the yeast nucleus in the hydrolytic vacuole lumen, occurs at NVJs; Tsc13 and the OSH family are required for efficient PMN
70
ONM and vacuole (NVJ)
Tsc13 and Osh1
Nvj1* (ONM) and Vac8 (vacuole)
ER and plasma membrane (PAM)
Erg1, Erg6, Erg9, PtdSer‑ and PtdIns‑ synthase (yeast); and RDGB (fly)
PTP1B (ER) and insulin receptor* The Tyr phosphatase PTP1B resides in the ER (plasma membrane) membrane and directly interacts with insulin receptors in the plasma membrane, thereby enhancing the formation of ER–plasma membrane junctions
71
STIM1* (ER) and ORAI1* (plasma membrane)
The interaction between the ER Ca2+ sensor STIM1 and the ORAI1 subunit of the CRAC channel is involved in the formation and stabilization of ER–plasma membrane MCSs and capacitative Ca2+ influx
72
IP3R* (ER), TRP* (plasma membrane) and homer
An interaction between TRP on the plasma membrane and IP3R in the ER is mediated by the cytoplasmic adaptor protein homer and regulates store-operated Ca2+ entry
73
ORP3
VAPA* (ER) and ORP3 (plasma membrane)
ORP3 interacts with the plasma membrane mainly through its PH domain and with the ER through its FFAT motif that binds VAPA
74
ER and late endosome
ORP1L
VAPA* (ER) and ORP1L (late endosome)
ORP1L on late endosomes interacts with VAPA on the ER through its FFAT motif and thereby tethers the ER to late endosomes under low cholesterol levels
75
ER and trans-Golgi
OSBP, CERT and NIR2
VAPA* (ER) and OSBP (trans-Golgi)
25OH enhances the Golgi targeting of OSBP, which also interacts with the ER through its FFAT motif; OSBP may interact with both the Golgi and the ER in the presence of 25OH, and induces their tethering
12,82,84
CRAC, Ca2+-release-activated Ca2+ ; ERMES, endoplasmic reticulum (ER)–mitochondrion encounter structure; GRP75, 75 kDa glucose-regulated protein; IP3R, inositol 1,4,5-triphosphate receptor; MAM, mitochondrion-associated membrane; MCS, membrane contact site; Mdm, mitochondrial distribution and morphology protein; MFN, mitofusin; Mmm, mitochondrial morphology protein; NVJ, nucleus–vacuole junction; 25OH, 25-hydroxycholesterol; OMM, outer mitochondrial membrane; ONM, outer nuclear membrane; ORP1L, OSBP-related protein 1L; ORP3, OSBP-related protein 3; Osh1, oxysterol-binding protein homologue 1; PACS2, phosphofurin acidic cluster sorting protein 2; PAM, plasma membrane-associated membrane; PH, pleckstrin homology; PMN, piecemeal microautophagy of the nucleus; PSS, phosphatidylserine (PtdSer) synthase; PtdIns, phosphatidylinositol; PTP1B, protein Tyr phosphatase 1B; RDGB, retinal degeneration B; STIM1, stromal interaction molecule 1; Tsc13, temperature-sensitive CSG2 suppressor protein 13; TRP, transient receptor potential; Vac8, vacuolar protein 8; VAPA, VAMP-associated protein A; VDAC, voltage-dependent anion channel. *Transmembrane protein. ‡Unclear involvement.
PtdIns-transfer domain A protein domain that is present in PITPs and mediates the exchange of PtdIns for PtdCho, and vice versa. PITPs have a ~16‑fold higher binding affinity for PtdIns than PtdCho.
Mutations in the CERT FFAT motif abrogate not only the VAP–CERT interaction, but also CERT-mediated ER‑to-Golgi ceramide transport in intact cells12,81. Similarly, mutations in the CERT PH domain abolish its targeting to the Golgi, and, consequently, sphingomyelin synthesis. These observations suggest that the Golgi and ER targeting of CERT spatially restrict ceramide transport to MCSs, thereby ensuring efficient ER‑to-Golgi ceramide transport. OSBP also contains a PH domain and an FFAT motif that mediate its interaction with the Golgi and the ER membranes, respectively 82. OSBP binds
25‑hydroxycholesterol (25OH) through its carboxy‑ terminal sterol-binding domain and facilitates CERTmediated ER‑to-Golgi ceramide transport in the presence of 25OH82. The PtdIns-transfer domain-containing protein NIR2 (also known as PITPNM1) also has an FFAT motif, which mediates its interaction with the ER through VAP binding 83. It seems that CERT, OSBP and NIR2, which are all localized to the Golgi, function coordinately at ER–Golgi MCSs to regulate ceramide transport. Treatment with 25OH enhances the Golgi targeting of these proteins and the recruitment of VAPs to the ER–Golgi MCSs, and could thereby facilitate
744 | O CTOBER 2010 | VOLUME 11
www.nature.com/reviews/molcellbio © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS a
b
ER ERMES
MFN2
Mdm12 Mmm1 Mdm34 Mdm10
GTP
GTP
100 nM Mitochondrion
c
O
PtdCho
O O
O
O
O– P
N+
O O O
O
O
P
O O
COO– O
NH3+
O
O
PtdEtn
O–
CH3
CH3 CH3
O
PtdSer
O
O– P
O
NH2
O
O
d
O
ER PtdSer
PtdCho
PSS1
PtdEtn
Ethanolamine
Choline L-Ser
PtdSer CO 2 PtdSer decarboxylase
PSS2
L-Ser
PtdEtn Mitochondrion
Figure 4 | Lipid transport at ER‑mitochondrion MCSs. a | Endoplasmic reticulum (ER)‑mitochondrion tethers. The ER-mitochondrion encounter structure (ERMES) Nature tether. ReviewsIt| is Molecular CellofBiology complex was identified in yeast as an ER–mitochondrion composed proteins residing in the ER (maintenance of mitochondrial morphology protein 1 (Mmm1) and mitochondrial distribution and morphology protein 12 (Mdm12)) and mitochondria (Mdm10 and Mdm34). ERMES-mutant strains suffer from mitochondrial phospholipid abnormalities owing to impaired phospholipid transport and biosynthesis67. ER–mitochondrion tethers can also affect Ca2+ transport. Mitofusin 2 (MFN2), a dynamin-like GTPase, is enriched in the mitochondrion-associated membrane (MAM) and was proposed to tether the ER to the mitochondria in mammalian cells by trans homotypic interaction with MFN2 or heterotypic interaction with MFN1 (not shown) on the mitochondrial surface68. In MFN2‑deficient cells, both the number of ER–mitochondrion membrane contact sites (MCSs) and the uptake of Ca2+ into mitochondria were markedly reduced, showing a crucial role for MCSs in Ca2+ transport and signalling68. b | An electron microscopy image of MAM. Two ER membranes are seen closely associated with the mitochondrion and are indicated by arrowheads25. c | Structures of phosphatidylcholine (PtdCho), phosphatidylserine (PtdSer) and phosphatidylethanolamine (PtdEtn). d | In mammalian cells, PtdSer can be formed through the exchange of l‑Ser for ethanolamine from PtdEtn or for choline from PtdCho. PtdSer is transported through the MAM and is converted to PtdEtn by decarboxylation in the mitochondrial intermembrane space. PtdEtn is subsequently exported back to the ER. PSS, PtdSer synthase. Image in part b reproduced, with permission, from REF. 25 © (1997) Elsevier.
non-vesicular lipid transport between the ER and Golgi. It was proposed that NIR2 is required for transport of PtdIns from the ER to the Golgi and that the sub sequent production of PtdIns4P in the Golgi membrane enhances the recruitment of OSBP and CERT to the Golgi and consequently ER‑to-Golgi ceramide transport 84. Hence, NIR2 and CERT can mediate PtdIns and ceramide transport, respectively. It is not clear, however, whether OSBP mediates sterol transfer at ER–Golgi MCSs or exerts other functions, such as lipid sensing or MCS stabilization. Together, these observations suggest that MCSs are ideal sites for non-vesicular lipid transport. Spontaneous lipid transport could be greatly facilitated by local membrane concentrations16, whereas LTPs might be involved in various functions, including the formation and stabilization of MCSs, as well as lipid transport and/or sensing.
Functions of LTPs in cells LTPs have been implicated in the regulation of several cellular processes. Their ability to transfer lipids between membranes in vitro suggests that they could also mediate lipid trafficking in intact cells. Although CERT is perhaps the most established example of an LTP that mediates intracellular lipid trafficking, several sterol transfer proteins have also been implicated in intracellular sterol trafficking85. ORP2 was proposed to function as an ER–plasma membrane sterol transporter, as its overexpression in mammalian cells enhanced efflux of newly synthesized cholesterol from the ER to extracellular cyclodextrin without perturbing the plasma membrane cholesterol content 86. STAR has a crucial role in steroidogenesis by facilitating the delivery of cholesterol from the outer to the inner mitochondrial membrane, where the first step of steroid biosynthesis is catalysed43. It can transfer cholesterol between membranes in vitro, and robustly transfers cholesterol in steroidogenic cells. It was estimated that STAR can deliver ~400 molecules of cholesterol into mitochondria per minute87, but its precise mechanism of action remains unclear. The OSBP homologues (Osh proteins) in budding yeast also mediate intracellular sterol transport, as a yeast strain deficient in all of the seven Osh proteins exhibits a severe reduction (~80%) in plasma membrane–ER sterol transfer 88. However, Osh proteins may have an impact on sterol transport indirectly, by affecting actin dynamics22 or modulating other cellular functions. Indeed, many LTPs have been implicated in the regulation of diverse cellular functions, such as signal transduction, vesicular transport and lipid transport and metabolism. These diverse functions might be mediated by distinct modes of action. First, LTPs could facilitate vector ial lipid transfer in cells as they do in vitro. Second, they could act as lipid sensors that alter their interactions with partner proteins in response to binding lipids or membranes. For example, an LTP might activate a partner protein in response to lipid exchange with a membrane. Furthermore, LTPs might control the access of other lipid-binding proteins to lipids in the membrane,
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
VOLUME 11 | O CTOBER 2010 | 745 © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS Vesicular transport Golgi
PKD
DAG
PI4P
SMS1 PtdCho + ceramide
SM + DAG GSL PtdIns4P PtdIns PtdCho CPT DAG
GlcCer
PI4K
GlcCer synthase
CDP-choline
FAPP2
CERT
OSBP 25OH
a
PITP
p-Choline CDP-choline
b
ER Ceramide
PITP
PtdIns
PtdCho
CCT
Figure 5 | Networking of coordinated functions of LTPs Reviews in the Golgi complex Nature | Molecular Cell Biology of mammalian cells. a | Ceramide is synthesized in the endoplasmic reticulum (ER) and transported by ceramide transfer protein (CERT) to the trans-Golgi complex at ER–Golgi membrane contact sites (MCSs)14,78. Ceramide is converted into sphingomyelin (SM) at the lumenal leaflet of the trans-Golgi by sphingomyelin synthase 1 (SMS1), which transfers phosphocholine from phosphatidylcholine (PtdCho) to ceramide, yielding diacylglycerol (DAG) as a by-product. Diacylglycerol affects membrane curvature and also stimulates several protein kinases, including protein kinase D (PKD), which in turn facilitates budding and fission of secretory vesicles90. The transfer of ceramide by CERT is facilitated by 25‑hydroxycholesterol (25OH)-bound oxysterol-binding protein (OSBP). b | Phosphatidylinositol (PtdIns) is synthesized in the ER and can be transported to the Golgi complex by PtdIns-transfer proteins (PITPs), possibly through ER–Golgi MCSs84. PITPs exchange PtdIns with PtdCho, and thereby may transport PtdCho back from the Golgi to the ER, where PtdCho can inhibit CCT (cholinephosphotransferase (CPT)-phosphocholine cytidylyltransferase) and consequently the cytidine diphosphate (CDP)-choline pathway103. PtdIns is phosphorylated by Golgi-localized PtdIns‑4‑phosphate (PtdIns4) kinase (PI4K) to produce PtdIns4P, which facilitates the recruitment of several membrane-trafficking regulatory proteins, as well as CERT, OSBP and FAPP2 (also known as PLEKHA8) through their pleckstrin homology (PH) domains14,44,62. Glycosphingolipids (GSLs) are produced at the lumenal leaflet of the trans-Golgi from glucosylceramide (GlcCer), which is synthesized from ceramide by GlcCer synthase. GlcCer is transported by FAPP2, which is essential for glycosphingolipid (GSL) production113,114. However, the mechanism by which FAPP2 transfers GlcCer and consequently leads to the production of glycosphingolipids, as well as the localization of GlcCer synthase at the cis- or trans-Golgi, are currently subjects of controversy113,114. For simplicity, only the trans-Golgi is shown here.
Steroidogenesis The biosynthesis of steroid hormones.
either by presenting a lipid to a second protein or by preventing a lipid-binding protein from accessing a lipid in a membrane. Finally, an LTP might help to establish transient changes in the distribution of lipids in a membrane by extracting or delivering a lipid to a particular region of the membrane, or by affecting the lipid phase in the portion of a membrane to which it is bound. These modes of action are not mutually exclusive, and it seems possible that an LTP could employ more than one of them.
LTPs can mediate intracellular lipid transport. The transport of ceramide from its synthesis site at the ER to the Golgi complex by CERT is an example of LTP-mediated vectorial lipid transport in mammalian cells. CERT was discovered as a protein that can restore sphingomyelin synthesis in LY‑A cells, a mutant Chinese hamster ovary (CHO) cell line81. The reduced level of sphingomyelin in LY‑A cells results from a defect in ceramide transport from the ER to the Golgi owing to a point mutation in CERT (Gly67Glu) that abrogates its PtdIns4P binding and consequently its Golgi targeting. This finding suggests that CERT does not mediate a simple vectorial transport of ceramide from the ER to the Golgi. Instead, it transfers ceramide on the basis of the PtdIns4P content of the Golgi membranes. The transport of ceramide from the ER to the Golgi by CERT, most likely at ER–Golgi MCSs, not only provides a major route for intracellular ceramide trafficking, but also initiates a signal transduction cascade that eventually regulates Golgi-mediated membrane trafficking (FIG. 5). Ceramide is converted into sphingomyelin at the lumenal leaflet of the trans-Golgi by SMS1, which transfers phosphocholine from PtdCho to ceramide, yielding diacylglycerol as a by-product89 (BOX 1). Diacylglycerol activates several Golgi-localized protein kinases, including protein kinase D (PKD), which in turn regulates vesicular fission and consequently protein transport from the Golgi to the plasma membrane90. In addition, PKD phosphorylates PtdIns 4‑kinase IIIβ (PI4KIIIβ) and stimulates its lipid-kinase activity, thereby enhancing the production of PtdIns4P in the Golgi complex 91. PtdIns4P can recruit numerous proteins to the Golgi complex; some, such as the adaptor protein 1 complex (AP1 complex)92, are directly involved in the regulation of vesicular transport. Remarkably, it has been shown that CERT regulates PKD activation and is essential for protein transport from the Golgi complex. Furthermore, activated PKD phosphorylates CERT, leading to its dissociation from the Golgi and thereby attenuating ER‑to-Golgi ceramide transport 93. This phosphorylation could provide a feedback mechanism for regulating ceramide transport to the Golgi and, consequently, for the production of diacylglycerol by SMS. This example shows how lipid transfer mediated by an LTP can induce a network of lipid modifications in a specific membrane compartment, consequently regulating other cellular processes, such as membrane transport (FIG. 5). Another example of an LTP that could mediate lipid transport in cells has emerged from studies of retinal degeneration B (RDGB) in Drosophila melanogaster. RDGB, a PtdIns‑transfer domain-containing protein related to NIR2, is required for photoreceptor cell viability and light response94,95. Phototransduction in D. melanogaster is a light-induced G protein-coupled receptor cascade mediated by phospholipase C (PLC) activation, which in turn catalyses the hydrolysis of PtdIns(4,5)P2 to generate inositol‑1,4,5‑trisphosphate (Ins(1,4,5)P3) and diacylglycerol. RDGB-mutant flies exhibit light-induced PtdIns(4,5)P2 depletion96, and rescue experiments suggest that the PtdIns‑transfer domain of RDGB is crucial for all of RDGB’s essential functions97.
746 | O CTOBER 2010 | VOLUME 11
www.nature.com/reviews/molcellbio © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS Box 1 | Homeostasis of diacylglycerol in the Golgi The steady-state level of Choline diacylglycerol (DAG) in Choline-K the Golgi complex is p-Choline determined by several CTP metabolic pathways, which CCT regulate its production Golgi CPT DAGK (dashed arrows) and PA DAG + CDP-choline PtdCho consumption (dotted PAP 26,89,105 PLD arrows) . SMS1 PtdCho + ceramide PtdCho + ceramide Diacylglycerol is mainly produced by sphingomyelin SM (SM) synthase 1 (SMS1), which catalyses the transfer of phosphocholine from phosphatidylcholine (PtdCho) to ceramide, thereby Nature Reviews | Molecular Cellproducing Biology diacylglycerol and sphingomyelin. Diacylglycerol is also produced by phosphatidic acid (PA) phosphatase (PAP), which dephosphorylates phosphatidic acid. Diacylglycerol consumption is predominantly mediated by the cytidine diphosphate (CDP)-choline pathway for PtdCho biosynthesis. This pathway begins in the endoplasmic reticulum, where choline is phosphorylated by choline kinase (choline‑K) to phosphocholine (p-choline), which is then converted to CDP-choline by CCT (cholinephosphotransferase (CPT)–phosphocholine cytidylyltransferase); the rate-limiting enzyme of this pathway. CPT catalyses the production of PtdCho from diacylglycerol and CDP-choline. Inactivation of choline kinase, CCT or CPT in yeast can ‘bypass’ the requirement for Sec14 (REF. 101). The consumption of diacylglycerol is also regulated by diacylglycerol kinase (DAGK), which phosphorylates diacylglycerol to generate phosphatidic acid, or by SMS1, which converts diacylglycerol and sphingomyelin back to PtdCho and ceramide. Phosphatidic acid is produced from PtdCho by phospholipase D (PLD). The level of diacylglycerol in the Golgi is crucial for maintaining the Golgi structure and could directly affect Golgi-mediated transport events.
These observations suggest that RDGB is required for the regeneration of PtdIns(4,5)P2 in response to light excitation. Accordingly, it was proposed that RDGB uses its PtdIns‑transfer domain to transfer PtdIns from its synthesis site at the subrhabdomeric cisternae (part of the ER) to the rhabdomeric microvilli (part of the plasma membrane) in D. melanogaster photoreceptors. PtdIns is then phosphorylated to PtdIns(4,5)P2, which promotes light-induced PtdIns(4,5)P2 hydrolysis by PLC98. Indeed, inactivation of PLC in flies suppresses rdgB degeneration94. These RDGB-mediated PtdIns‑transport events may also operate at MCSs (ER–plasma membrane), as RDGB is localized to the subrhabdomeric cisternae and is also associated with the base of the rhabdomeric microvilli98. Furthermore, RDGB contains an FFAT motif 79 and could possibly interact with ER‑localized VAP proteins, which are found in many MCSs12,75,80,82.
AP1 complex A heterotetrameric complex with a role in protein sorting at the trans-Golgi network and endosomes. AP1 mediates the recruitment of clathrin to membranes and the recognition of sorting signals in the cytosolic tails of transmembrane cargo proteins.
LTPs act as lipid sensors and/or lipid-presenting proteins. The function of LTPs as lipid sensors has been proposed by numerous studies. Perhaps the most established example is Sec14, the main PITP in Saccharomyces cerevisiae. Similarly to other PITPs, Sec14 facilitates the monomeric exchange of either PtdIns or PtdCho between membranes in vitro 49,99. Sec14 is required for protein transport from a late Golgi compartment and is essential for yeast cell viability 100. This requirement, however, can be bypassed by the inactivation of one of seven different genes101,102. Three of these genes encode structural enzymes of the cytidine
diphosphate (CDP)-choline pathway for PtdCho biosynthesis, which consumes diacylglycerol to produce PtdCho101 (BOX 1). In its PtdCho‑bound form, Sec14 negatively regulates CCT (cholinephosphotransferase (CTP)–phosphocholine cytidylyltransferase), the ratelimiting enzyme of the CDP-choline pathway 103. These observations, along with the low PtdIns and high PtdCho levels in the Golgi of sec14‑mutant strains, suggest that Sec14 does not modulate Golgi secretory function through its PtdIns–PtdCho exchange activity, but instead through its direct effect on the CDP-choline pathway and therefore diacylglycerol consumption103,104. Thus, Sec14 was proposed to sense the PtdCho levels in yeast Golgi and to respond to increased PtdCho levels by inhibiting the activity of CCT. Hence, Sec14 is required to maintain a critical pool of diacylglycerol in the Golgi by regulating its consumption by the CDP-choline pathway, a process that is essential for Golgi-mediated trafficking. Remarkably, mammalian NIR2 uses a similar mechanism to regulate the levels of diacylglycerol in the Golgi complex 105. In both cases, however, the PtdCho‑exchange activity must be coupled to changes in phosphoinositides. It was recently proposed that as PtdCho levels increase, Sec14 is activated for heterotypic PtdIns– PtdCho exchange, which in turn stimulates the production of PtdIns4P by PtdIns4‑kinase (Pik1) in the yeast Golgi complex 106. In heterotypic reactions that exchange PtdIns for PtdCho, an invading PtdCho may force PtdIns out of the Sec14 lipid-binding pocket, thereby generating a kinase-susceptible PtdIns intermediate; in this case, Sec14 acts as a lipid-presenting protein, as it presents PtdIns to the kinase106. Therefore, Sec14 may act both as a PtdCho sensor and as a PtdIns‑presenting protein, which transmits PtdCho metabolic information to PtdInsP synthesis106. This example shows that a single LTP may have more than one mode of action in cells. Members of the OSBP, ORP and OSH family may also function as sterol sensors22. ORP1L, for example, has been implicated in the sensing of cholesterol levels in late endosomes, and thereby in regulating their subcellular distribution. Specifically, by undergoing conformational changes in response to cholesterol content in late endosomes, ORP1L can mediate the assembly of a protein complex that may or may not be linked to the microtubule cytoskeleton. This causes late endosomes to cluster towards the microtubule minus end at high cholesterol levels and scatter at low cholesterol levels75. This example shows how an LTP can regulate protein–protein interactions according to its membrane environment. A related example is associated with the production of a protein complex between OSBP and two phosphatases, which dephosphorylate extracellular signal-regulated kinase (ERK)107. This complex disassembles at low cholesterol levels, losing its ability to dephosphorylate ERK, and consequently leads to an increase in phosphorylated ERK levels. It was, therefore, proposed that OSBP functions as a cholesterol sensor that regulates mitogenic signals in response to cellular levels of cholesterol through the assembly of a protein complex consisting of ERK phosphatases107.
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
VOLUME 11 | O CTOBER 2010 | 747 © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS OSH proteins also respond to their membrane environment, but in this case they modulate the sterol content of membranes according to changes in phosphoinositide levels. It was shown that inhibition of PtdInsP biosynthesis markedly attenuates plasma membrane–ER sterol transport 88 and that OSH proteins bind phospho inositides61,108,109. Consequently, OSH proteins may not mediate the simple vectorial transport of sterol from one membrane to another. Instead, they might dynamically modulate the sterol content of membranes that can recruit them; that is, membranes with specific lipid composition (for example, PtdInsP content) and/or biophysical properties (for example, curvature)17. This mode of action shows how an LTP uses its lipid-transfer activity to coordinate the intracellular distribution of different lipid species between different cellular membranes.
Concluding remarks The mechanisms by which lipids are transported and distributed between and in cellular membranes have not been fully explored. The transport of several cellular lipids, such as sterols and ceramide, is mainly mediated by a non-vesicular transport mechanism. This mode of transport could potentially occur by spontaneous lipid exchange or by the action of LTPs. Spontaneous lipid exchange is a slow process and is therefore insufficient to support large lipid fluxes. However, it could be greatly facilitated by high membrane concentrations, suggesting that MCSs could be optimal sites for spontaneous lipid exchange. MCSs could also be ideal for the lipid transfer activity of LTPs, and indeed, many LTPs have been found in MCSs12,70,74,75. Whether MCSs represent the predominant sites of lipid transport in intact cells is unclear and difficult to show experimentally. It is also not known whether LTPs function exclusively at MCSs. Elucidating the mechanisms by which LTPs interact with specific membranes, identifying targeting determinants that could bind different membrane compartments and characterizing the subcellular localization of LTPs using ultrastructural analysis and live-cell imaging, could deepen our understanding of LTP function.
1.
2. 3. 4. 5.
6. 7. 8. 9.
Holthuis, J. C., van Meer, G. & Huitema, K. Lipid microdomains, lipid translocation and the organization of intracellular membrane transport (Review). Mol. Membr. Biol. 20, 231–241 (2003). Sprong, H., van der Sluijs, P. & van Meer, G. How proteins move lipids and lipids move proteins. Nature Rev. Mol. Cell Biol. 2, 504–513 (2001). Voelker, D. R. Organelle biogenesis and intracellular lipid transport in eukaryotes. Microbiol. Rev. 55, 543–560 (1991). van Meer, G. Lipid traffic in animal cells. Annu. Rev. Cell Biol. 5, 247–275 (1989). Pomorski, T., Hrafnsdottir, S., Devaux, P. F. & van Meer, G. Lipid distribution and transport across cellular membranes. Semin. Cell Dev. Biol. 12, 139–148 (2001). Sleight, R. G. Intracellular lipid transport in eukaryotes. Annu. Rev. Physiol. 49, 193–208 (1987). Voelker, D. R. Lipid transport pathways in mammalian cells. Experientia 46, 569–579 (1990). Kaplan, M. R. & Simoni, R. D. Intracellular transport of phosphatidylcholine to the plasma membrane. J. Cell Biol. 101, 441–445 (1985). Vance, J. E., Aasman, E. J. & Szarka, R. Brefeldin A does not inhibit the movement of
10. 11. 12.
13. 14. 15.
Regardless of their sites of action, it seems that LTPs do not mediate a simple vectorial lipid transport from one membrane to another. Instead, LTPs use their lipidtransfer activity to modulate the lipid content of membranes according to their local membrane environment. LTPs can, therefore, control lipid homeostasis and the diverse cellular processes associated with it, such as signal transduction, membrane trafficking and lipid meta bolism, by capturing and responding to local membrane modifications. This ability enables a tight coordination of various cell responses that occur in the same or even different membranes. For example, the production of diacylglycerol in the Golgi is coordinately regulated by its consumption by the CDP-choline pathway. This pathway, however, is regulated by CCT, which resides on the ER (BOX 1). This coordination could be regulated by PITPs, such as NIR2, that may transfer PtdIns from the ER to the Golgi and PtdCho from the Golgi to the ER at the ER–Golgi MCSs (FIG. 5). The ability of LTPs to coordinate and integrate different lipid-mediated cellular responses could be even more pronounced in large LTPs consisting of multiple structural domains. Although most have not been studied extensively, many large LTPs contain structural domains with cellular functions or enzymatic activities that are well established. Domains that regulate the activity of the Rho, Rac and/or Ras small GTPases have been identified in the Sec14 and START families21. These small GTPases are involved in the regulation of cytoskeletal remodelling and cell proliferation, both of which are tightly regulated by certain lipids. Thus, the spatial proximity between the LTD and these domains could provide a mechanism for coupling their corresponding functions, thereby enhancing the efficiency of their integrated physiological role. The coupling of lipid modifications to cytoskeletal remodelling events, for example, operates in diverse cellular processes such as membrane trafficking, cell migration and signal transduction. Thus, characterizing the functions of large LTPs and elucidating their mode of action and regulation is expected to shed light on some of the unsolved mysteries associated with this class of protein.
phosphatidylethanolamine from its sites for synthesis to the cell surface. J. Biol. Chem. 266, 8241–8247 (1991). Holthuis, J. C. & Levine, T. P. Lipid traffic: floppy drives and a superhighway. Nature Rev. Mol. Cell Biol. 6, 209–220 (2005). Levine, T. Short-range intracellular trafficking of small molecules across endoplasmic reticulum junctions. Trends Cell Biol. 14, 483–490 (2004). Kawano, M., Kumagai, K., Nishijima, M. & Hanada, K. Efficient trafficking of ceramide from the endoplasmic reticulum to the Golgi apparatus requires a VAMPassociated protein-interacting FFAT motif of CERT. J. Biol. Chem. 281, 30279–30288 (2006). This paper provides molecular insight into CERT-mediated ceramide transport at ER–Golgi MCSs. Funato, K. & Riezman, H. Vesicular and nonvesicular transport of ceramide from ER to the Golgi apparatus in yeast. J. Cell Biol. 155, 949–959 (2001). Hanada, K. Discovery of the molecular machinery CERT for endoplasmic reticulum‑to‑Golgi trafficking of ceramide. Mol. Cell Biochem. 286, 23–31 (2006). Lebiedzinska, M., Szabadkai, G., Jones, A. W., Duszynski, J. & Wieckowski, M. R. Interactions between the endoplasmic reticulum, mitochondria,
748 | O CTOBER 2010 | VOLUME 11
plasma membrane and other subcellular organelles. Int. J. Biochem. Cell Biol. 41, 1805–1816 (2009). 16. Jones, J. D. & Thompson, T. E. Spontaneous phosphatidylcholine transfer by collision between vesicles at high lipid concentration. Biochemistry 28, 129–134 (1989). 17. Mesmin, B. & Maxfield, F. R. Intracellular sterol dynamics. Biochim. Biophys. Acta 1791, 636–645 (2009). 18. Zilversmit, D. B. Lipid transfer proteins: overview and applications. Methods Enzymol. 98, 565–573 (1983). 19. Helmkamp, G. M. Jr. Phospholipid transfer proteins: mechanism of action. J. Bioenerg. Biomembr. 18, 71–91 (1986). 20. Wirtz, K. W. & Zilversmit, D. B. Exchange of phospholipids between liver mitochondria and microsomes in vitro. J. Biol. Chem. 243, 3596–3602 (1968). 21. D’Angelo, G., Vicinanza, M. & De Matteis, M. A. Lipid-transfer proteins in biosynthetic pathways. Curr. Opin. Cell Biol. 20, 360–370 (2008). 22. Fairn, G. D. & McMaster, C. R. Emerging roles of the oxysterol-binding protein family in metabolism, transport, and signaling. Cell. Mol. Life Sci. 65, 228–236 (2008).
www.nature.com/reviews/molcellbio © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS 23. De Matteis, M. A., Di Campli, A. & D’Angelo, G. Lipidtransfer proteins in membrane trafficking at the Golgi complex. Biochim. Biophys. Acta 1771, 761–768 (2007). 24. Cockcroft, S. Mammalian phosphatidylinositol transfer proteins: emerging roles in signal transduction and vesicular traffic. Chem. Phys. Lipids 98, 23–33 (1999). 25. Perkins, G. et al. Electron tomography of neuronal mitochondria: three-dimensional structure and organization of cristae and membrane contacts. J. Struct. Biol. 119, 260–272 (1997). 26. Lev, S. Lipid homoeostasis and Golgi secretory function. Biochem. Soc. Trans. 34, 363–366 (2006). 27. Nichols, J. W. & Pagano, R. E. Kinetics of soluble lipid monomer diffusion between vesicles. Biochemistry 20, 2783–2789 (1981). 28. McLean, L. R. & Phillips, M. C. Mechanism of cholesterol and phosphatidylcholine exchange or transfer between unilamellar vesicles. Biochemistry 20, 2893–2900 (1981). 29. McLean, L. R. & Phillips, M. C. Kinetics of phosphatidylcholine and lysophosphatidylcholine exchange between unilamellar vesicles. Biochemistry 23, 4624–4630 (1984). 30. Phillips, M. C., Johnson, W. J. & Rothblat, G. H. Mechanisms and consequences of cellular cholesterol exchange and transfer. Biochim. Biophys. Acta 906, 223–276 (1987). 31. Massey, J. B. et al. Measurement and prediction of the rates of spontaneous transfer of phospholipids between plasma lipoproteins. Biochim. Biophys. Acta 794, 274–280 (1984). 32. Bai, J. & Pagano, R. E. Measurement of spontaneous transfer and transbilayer movement of BODIPYlabeled lipids in lipid vesicles. Biochemistry 36, 8840–8848 (1997). 33. McLean, L. R. & Phillips, M. C. Cholesterol transfer from small and large unilamellar vesicles. Biochim. Biophys. Acta 776, 21–26 (1984). 34. Radhakrishnan, A. & McConnell, H. M. Chemical activity of cholesterol in membranes. Biochemistry 39, 8119–8124 (2000). 35. Roseman, M. A. & Thompson, T. E. Mechanism of the spontaneous transfer of phospholipids between bilayers. Biochemistry 19, 439–444 (1980). 36. Martin, F. J. & MacDonald, R. C. Phospholipid exchange between bilayer membrane vesicles. Biochemistry 15, 321–327 (1976). 37. Steck, T. L., Kezdy, F. J. & Lange, Y. An activationcollision mechanism for cholesterol transfer between membranes. J. Biol. Chem. 263, 13023–13031 (1988). 38. Gadella, T. W. Jr & Wirtz, K. W. Phospholipid binding and transfer by the nonspecific lipid-transfer protein (sterol carrier protein 2). A kinetic model. Eur. J. Biochem. 220, 1019–1028 (1994). 39. Lalanne, F. & Ponsin, G. Mechanism of the phospholipid transfer protein-mediated transfer of phospholipids from model lipid vesicles to high density lipoproteins. Biochim. Biophys. Acta 1487, 82–91 (2000). 40. Gadella, T. W. Jr & Wirtz, K. W. The low-affinity lipid binding site of the non-specific lipid transfer protein. Implications for its mode of action. Biochim. Biophys. Acta 1070, 237–245 (1991). 41. Nichols, J. W. Kinetics of fluorescent-labeled phosphatidylcholine transfer between nonspecific lipid transfer protein and phospholipid vesicles. Biochemistry 27, 1889–1896 (1988). 42. Kumagai, K. et al. CERT mediates intermembrane transfer of various molecular species of ceramides. J. Biol. Chem. 280, 6488–6495 (2005). 43. Alpy, F. & Tomasetto, C. Give lipids a START: the StARrelated lipid transfer (START) domain in mammals. J. Cell Sci. 118, 2791–2801 (2005). 44. Yamaji, T., Kumagai, K., Tomishige, N. & Hanada, K. Two sphingolipid transfer proteins, CERT and FAPP2: their roles in sphingolipid metabolism. IUBMB Life 60, 511–518 (2008). 45. Wirtz, K. W., Schouten, A. & Gros, P. Phosphatidylinositol transfer proteins: from closed for transport to open for exchange. Adv. Enzyme Regul. 46, 301–311 (2006). 46. Im, Y. J., Raychaudhuri, S., Prinz, W. A. & Hurley, J. H. Structural mechanism for sterol sensing and transport by OSBP-related proteins. Nature 437, 154–158 (2005). The crystal structure of Osh4 was resolved in this study, and the mechanisms by which Osh4 mediates sterol transport and interacts with membranes were proposed.
47. Feng, L., Chan, W. W., Roderick, S. L. & Cohen, D. E. High-level expression and mutagenesis of recombinant human phosphatidylcholine transfer protein using a synthetic gene: evidence for a C‑terminal membrane binding domain. Biochemistry 39, 15399–15409 (2000). 48. Arakane, F. et al. Steroidogenic acute regulatory protein (StAR) retains activity in the absence of its mitochondrial import sequence: implications for the mechanism of StAR action. Proc. Natl Acad. Sci. USA 93, 13731–13736 (1996). 49. Wirtz, K. W. Phospholipid transfer proteins. Annu. Rev. Biochem. 60, 73–99 (1991). 50. Wirtz, K. W., Vriend, G. & Westerman, J. Kinetic analysis of the interaction of the phosphatidylcholine exchange protein with unilamellar vesicels and multilamellar liposomes. Eur. J. Biochem. 94, 215–221 (1979). 51. Rueckert, D. G. & Schmidt, K. Lipid transfer proteins. Chem. Phys. Lipids 56, 1–20 (1990). 52. Kasper, A. M. & Helmkamp, G. M. Jr. Intermembrane phospholipid fluxes catalyzed by bovine brain phospholipid exchange protein. Biochim. Biophys. Acta 664, 22–32 (1981). 53. Wirtz, K. W., Devaux, P. F. & Bienvenue, A. Phosphatidylcholine exchange protein catalyzes the net transfer of phosphatidylcholine to model membranes. Biochemistry 19, 3395–3399 (1980). 54. Zilversmit, D. B. Lipid transfer proteins. J. Lipid Res. 25, 1563–1569 (1984). 55. Li, H., Tremblay, J. M., Yarbrough, L. R. & Helmkamp, G. M. Jr. Both isoforms of mammalian phosphatidylinositol transfer protein are capable of binding and transporting sphingomyelin. Biochim. Biophys. Acta 1580, 67–76 (2002). 56. Mattjus, P. Glycolipid transfer proteins and membrane interaction. Biochim. Biophys. Acta 1788, 267–272 (2009). 57. Tilley, S. J. et al. Structure-function analysis of phosphatidylinositol transfer protein a bound to human phosphatidylinositol. Structure 12, 317–326 (2004). 58. Shadan, S. et al. Dynamics of lipid transfer by phosphatidylinositol transfer proteins in cells. Traffic 9, 1743–1756 (2008). 59. Kudo, N. et al. Structural basis for specific lipid recognition by CERT responsible for nonvesicular trafficking of ceramide. Proc. Natl Acad. Sci. USA 105, 488–493 (2008). 60. Phillips, S. E., Ile, K. E., Boukhelifa, M., Huijbregts, R. P. & Bankaitis, V. A. Specific and nonspecific membranebinding determinants cooperate in targeting phosphatidylinositol transfer protein b-isoform to the mammalian trans-Golgi network. Mol. Biol. Cell 17, 2498–2512 (2006). 61. Schulz, T. A. et al. Lipid-regulated sterol transfer between closely apposed membranes by oxysterolbinding protein homologues. J. Cell Biol. 187, 889–903 (2009). 62. Levine, T. P. & Munro, S. The pleckstrin homology domain of oxysterol-binding protein recognises a determinant specific to Golgi membranes. Curr. Biol. 8, 729–739 (1998). 63. Ngo, M. & Ridgway, N. D. Oxysterol binding proteinrelated protein 9 (ORP9) is a cholesterol transfer protein that regulates Golgi structure and function. Mol. Biol. Cell 20, 1388–1399 (2009). 64. Levine, T. & Loewen, C. Inter-organelle membrane contact sites: through a glass, darkly. Curr. Opin. Cell Biol. 18, 371–378 (2006). 65. Voelker, D. R. Bridging gaps in phospholipid transport. Trends Biochem. Sci. 30, 396–404 (2005). 66. Giorgi, C., De Stefani, D., Bononi, A., Rizzuto, R. & Pinton, P. Structural and functional link between the mitochondrial network and the endoplasmic reticulum. Int. J. Biochem. Cell Biol. 41, 1817–1827 (2009). 67. Kornmann, B. et al. An ER–mitochondria tethering complex revealed by a synthetic biology screen. Science 325, 477–481 (2009). This study used an elegant synthetic biology screen in yeast and identified an ER–mitochondrion tethering complex, the ERMES, which is essential for inter-organelle phospholipid transport. 68. de Brito, O. M. & Scorrano, L. Mitofusin 2 tethers endoplasmic reticulum to mitochondria. Nature 456, 605–610 (2008). 69. Szabadkai, G. et al. Chaperone-mediated coupling of endoplasmic reticulum and mitochondrial Ca2+ channels. J. Cell Biol. 175, 901–911 (2006).
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
70. Kvam, E. & Goldfarb, D. S. Nvj1p is the outer‑nuclear‑membrane receptor for oxysterolbinding protein homolog Osh1p in Saccharomyces cerevisiae. J. Cell Sci. 117, 4959–4968 (2004). 71. Anderie, I., Schulz, I. & Schmid, A. Direct interaction between ER membrane-bound PTP1B and its plasma membrane-anchored targets. Cell Signal. 19, 582–592 (2007). 72. Wu, M. M., Buchanan, J., Luik, R. M. & Lewis, R. S. Ca2+ store depletion causes STIM1 to accumulate in ER regions closely associated with the plasma membrane. J. Cell Biol. 174, 803–813 (2006). 73. Yuan, J. P. et al. Homer binds TRPC family channels and is required for gating of TRPC1 by IP3 receptors. Cell 114, 777–789 (2003). 74. Lehto, M. et al. Targeting of OSBP-related protein 3 (ORP3) to endoplasmic reticulum and plasma membrane is controlled by multiple determinants. Exp. Cell Res. 310, 445–462 (2005). 75. Rocha, N. et al. Cholesterol sensor ORP1L contacts the ER protein VAP to control Rab7‑RILP‑p150Glued and late endosome positioning. J. Cell Biol. 185, 1209–1225 (2009). 76. Stone, S. J. & Vance, J. E. Phosphatidylserine synthase‑1 and ‑2 are localized to mitochondriaassociated membranes. J. Biol. Chem. 275, 34534–34540 (2000). 77. Daum, G. & Vance, J. E. Import of lipids into mitochondria. Prog. Lipid Res. 36, 103–130 (1997). 78. Hanada, K., Kumagai, K., Tomishige, N. & Kawano, M. CERT and intracellular trafficking of ceramide. Biochim. Biophys. Acta 1771, 644–653 (2007). 79. Loewen, C. J., Roy, A. & Levine, T. P. A conserved ER targeting motif in three families of lipid binding proteins and in Opi1p binds VAP. EMBO J. 22, 2025–2035 (2003). This study identified a new motif, FFAT, in various LTPs that mediates the interaction with the ER through binding of the integral ER membrane proteins of the VAP family. 80. Lev, S., Ben Halevy, D., Peretti, D. & Dahan, N. The VAP protein family: from cellular functions to motor neuron disease. Trends Cell Biol. 18, 282–290 (2008). 81. Hanada, K. et al. Molecular machinery for non-vesicular trafficking of ceramide. Nature 426, 803–809 (2003). CERT, which was cloned in this study, was shown to have a major role in intracellular ceramide transport and, therefore, sphingomyelin production. 82. Perry, R. J. & Ridgway, N. D. Oxysterol-binding protein and vesicle-associated membrane proteinassociated protein are required for sterol-dependent activation of the ceramide transport protein. Mol. Biol. Cell 17, 2604–2616 (2006). 83. Amarilio, R., Ramachandran, S., Sabanay, H. & Lev, S. Differential regulation of endoplasmic reticulum structure through VAP–Nir protein interaction. J. Biol. Chem. 280, 5934–5944 (2005). 84. Peretti, D., Dahan, N., Shimoni, E., Hischberg, K. & Lev, S. Coordinated lipid transfer between the endoplasmic reticulum and the Golgi complex requires the VAP proteins and is essential for Golgi-mediated transport. Mol. Biol. Cell 19, 3871–3884 (2008). 85. Maxfield, F. R. & Mondal, M. Sterol and lipid trafficking in mammalian cells. Biochem. Soc. Trans. 34, 335–339 (2006). An insightful review on sterol transport in cells that provides interesting quantitative data and calculations. 86. Hynynen, R. et al. Overexpression of OSBP-related protein 2 (ORP2) induces changes in cellular cholesterol metabolism and enhances endocytosis. Biochem. J. 390, 273–283 (2005). 87. Ikonen, E. & Jansen, M. Cellular sterol trafficking and metabolism: spotlight on structure. Curr. Opin. Cell Biol. 20, 371–377 (2008). 88. Raychaudhuri, S., Im, Y. J., Hurley, J. H. & Prinz, W. A. Nonvesicular sterol movement from plasma membrane to ER requires oxysterol-binding proteinrelated proteins and phosphoinositides. J. Cell Biol. 173, 107–119 (2006). 89. Pagano, R. E. What is the fate of diacylglycerol produced at the Golgi apparatus? Trends Biochem. Sci. 13, 202–205 (1988). 90. Baron, C. L. & Malhotra, V. Role of diacylglycerol in PKD recruitment to the TGN and protein transport to the plasma membrane. Science 295, 325–328 (2002).
VOLUME 11 | O CTOBER 2010 | 749 © 2010 Macmillan Publishers Limited. All rights reserved
REVIEWS 91. Hausser, A. et al. Protein kinase D regulates vesicular transport by phosphorylating and activating phosphatidylinositol‑4 kinase IIIβ at the Golgi complex. Nature Cell Biol. 7, 880–886 (2005). 92. Wang, Y. J. et al. Phosphatidylinositol 4 phosphate regulates targeting of clathrin adaptor AP‑1 complexes to the Golgi. Cell 114, 299–310 (2003). 93. Fugmann, T. et al. Regulation of secretory transport by protein kinase D‑mediated phosphorylation of the ceramide transfer protein. J. Cell Biol. 178, 15–22 (2007). 94. Harris, W. A. & Stark, W. S. Hereditary retinal degeneration in Drosophila melanogaster. A mutant defect associated with the phototransduction process. J. Gen. Physiol. 69, 261–291 (1977). 95. Hotta, Y., Benzer, S. Abnormal electroretinogram in visual mutants in Drosophila. Nature 222, 354–356 (1969). 96. Hardie, R. C. et al. Calcium influx via TRP channels is required to maintain PIP2 levels in Drosophila photoreceptors. Neuron 30, 149–159 (2001). 97. Milligan, S. C., Alb, J. G. Jr, Elagina, R. B., Bankaitis, V. A. & Hyde, D. R. The phosphatidylinositol transfer protein domain of Drosophila retinal degeneration B protein is essential for photoreceptor cell survival and recovery from light stimulation. J. Cell Biol. 139, 351–363 (1997). 98. Vihtelic, T. S., Goebl, M., Milligan, S., O’Tousa, J. E. & Hyde, D. R. Localization of Drosophila retinal degeneration B, a membrane-associated phosphatidylinositol transfer protein. J. Cell Biol. 122, 1013–1022 (1993). 99. Bankaitis, V. A., Aitken, J. R., Cleves, A. E. & Dowhan, W. An essential role for a phospholipid transfer protein in yeast Golgi function. Nature 347, 561–562 (1990). 100. Bankaitis, V. A., Malehorn, D. E., Emr, S. D. & Greene, R. The Saccharomyces cerevisiae SEC14 gene encodes a cytosolic factor that is required for transport of secretory proteins from the yeast Golgi complex. J. Cell Biol. 108, 1271–1281 (1989). 101. Cleves, A. E. et al. Mutations in the CDP-choline pathway for phospholipid biosynthesis bypass the
requirement for an essential phospholipid transfer protein. Cell 64, 789–800 (1991). A functional link between Sec14 and the CDP-choline pathway was shown here by a genetic approach that led to the isolation of ‘bypass‑Sec14’ mutants. 102. Kearns, B. G. et al. Essential role for diacylglycerol in protein transport from the yeast Golgi complex. Nature 387, 101–105 (1997). 103. Skinner, H. B. et al. The Saccharomyces cerevisiae phosphatidylinositol-transfer protein effects a ligand-dependent inhibition of choline-phosphate cytidylyltransferase activity. Proc. Natl Acad. Sci. USA 92, 112–116 (1995). 104. McGee, T. P., Skinner, H. B., Whitters, E. A., Henry, S. A. & Bankaitis, V. A. A phosphatidylinositol transfer protein controls the phosphatidylcholine content of yeast Golgi membranes. J. Cell Biol. 124, 273–287 (1994). 105. Litvak, V., Dahan, N., Ramachandran, S., Sabanay, H. & Lev, S. Maintenance of the diacylglycerol level in the Golgi apparatus by the Nir2 protein is critical for Golgi secretory function. Nature Cell Biol. 7, 225–234 (2005). 106. Schaaf, G. et al. Functional anatomy of phospholipid binding and regulation of phosphoinositide homeostasis by proteins of the Sec14 superfamily. Mol. Cell 29, 191–206 (2008). This study provides mechanistic insights into PtdIns–PtdCho exchange by Sec14 and its dual role in sensing PtdCho and presenting PtdIns to control phosphoinositide homeostasis. 107. Wang, P. Y., Weng, J. & Anderson, R. G. OSBP is a cholesterol-regulated scaffolding protein in control of ERK 1/2 activation. Science 307, 1472–1476 (2005). 108. Fairn, G. D., Curwin, A. J., Stefan, C. J. & McMaster, C. R. The oxysterol binding protein Kes1p regulates Golgi apparatus phosphatidylinositol‑4‑ phosphate function. Proc. Natl Acad. Sci. USA 104, 15352–15357 (2007). 109. Li, X. et al. Analysis of oxysterol binding protein homologue Kes1p function in regulation of Sec14p‑dependent protein transport from the yeast Golgi complex. J. Cell Biol. 157, 63–77 (2002).
750 | O CTOBER 2010 | VOLUME 11
110. Ferrell, J. E. Jr, Lee, K. J. & Huestis, W. H. Lipid transfer between phosphatidylcholine vesicles and human erythrocytes: exponential decrease in rate with increasing acyl chain length. Biochemistry 24, 2857–2864 (1985). 111. Schouten, A. et al. Structure of apo-phosphatidylinositol transfer protein a provides insight into membrane association. EMBO J. 21, 2117–2121 (2002). 112. Yoder, M. D. et al. Structure of a multifunctional protein. Mammalian phosphatidylinositol transfer protein complexed with phosphatidylcholine. J. Biol. Chem. 276, 9246–9252 (2001). 113. D’Angelo, G. et al. Glycosphingolipid synthesis requires FAPP2 transfer of glucosylceramide. Nature 449, 62–67 (2007). 114. Halter, D. et al. Pre- and post-Golgi translocation of glucosylceramide in glycosphingolipid synthesis. J. Cell Biol. 179, 101–115 (2007).
Acknowledgements
Sima Lev is the incumbent of the Joyce and Ben B. Eisenberg Chair of Molecular Biology and Cancer Research. This work was supported by the Israel Science Foundation, Grant number 548/08. The author thanks R. Sertchook from the Weizmann Institute of Science for assistance in collecting the three-dimensional images, A. Menon and O. Laufman for productive discussion, and especially W. Prinz for the critical reading of this manuscript and his intellectual contribution.
Competing interests statement
The author declares no competing financial interests.
DATABASES Protein Data Bank: http://www.pdb.org 1KCM | 1T27 | 1ZHX | 2E3P
FURTHER INFORMATION Sima Lev’s homepage: http://www.weizmann.ac.il/mcb/Lev/
SUPPLEMENTARY INFORMATION See online article: S1 (table) | S2 (box) | S3 (figure) ALL LINKS ARE ACTIVE IN THE ONLINE PDF
www.nature.com/reviews/molcellbio © 2010 Macmillan Publishers Limited. All rights reserved
RESEARCH HIGHLIGHTS
1 0 - Y E A R A N N I V E R S A RY
Great expectations of small RNAs During the 1990s, two small RNAs in Caenorhabditis elegans, lin‑4 and let‑7, were found to regulate developmental timing through a unique mechanism, by annealing to a target mRNA and preventing its translation. This unusual behaviour later proved to be the first glimpse of a novel tier of gene expression control that is conserved from plants to mammals and is governed by small or ‘micro’ RNAs. The first indication that microRNAs (miRNAs) were not a peculiarity of worms came in 2000, when Pasquinelli et al. showed that let‑7 is highly conserved across a wide range of species — from flies to humans.
This idea, that miRNAs have many targets and act to destabilize mRNA, was heretic at the time but turned out to be exactly right. Eric Miska, University of Cambridge, UK
COMSTOCK IMAGES
Importantly, let‑7 sequence length was conserved at ~21 nucleotides and, in some cases, its target gene was also retained. On the basis of this, and its temporal regulation, the author’s proposed a conserved role of let‑7 in regulating gene expression during development, and they also predicted that there may be other such miR‑ NAs. Indeed, a year later, the Ambros, Bartel and Tuschl laboratories reported that there are numerous, diverse miRNAs in human cells, flies and worms. The next challenge was to identify the potential targets of these miRNAs. Computational approaches were initially used for this, but in 2004 Doench and Sharp went back to basics and experimentally investi‑ gated the pairing rules for a miRNA– mRNA interaction using a mutational analysis. Surprisingly, they found that the first eight nucleotides in the 5′ region of a miRNA contribute the most to target specificity and activity, forming a ‘seed’ that drives the interaction with a target mRNA. This principle formed the basis of modern miRNA target prediction. Computational predictions also suggested that miRNAs might have several targets. In 2005, Lim et al. made the experimental case for this. Using microarray analysis, they showed that a miRNA could drive the downregulation of several target mRNAs in human cells. In addition
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
to reducing protein levels of these targets, the transcripts themselves were also reduced. “This idea, that miRNAs have many targets and act to destabilize mRNA, was heretic at the time but turned out to be exactly right”, says Eric Miska (University of Cambridge, UK). The field has continued to gain momentum since these studies and numerous insights have been gained into the diverse functions of miRNAs both during development and in the adult, for example during mammalian tumorigenesis. The endogenous path‑ ways that drive their production and processing are also being unravelled. This, together with the therapeutic potential of miRNAs that is now emerging, suggests that we may be right to have such high hopes for these small RNAs.
Alison Schuldt
ORIGINAL RESEARCH PAPERS Pasquinelli, A. E. et al. Conservation of the sequence and temporal expression of let‑7 heterochronic regulatory RNA. Nature 408, 86–89 (2000) | Lee, R. C. & Ambros, V. An extensive class of small RNAs in Caenorhabditis elegans. Science 294, 862–864 (2001) | Lau, N. C., Lim, L. P., Weinstein, E. G. & Bartel, D. P. An abundant class of tiny RNAs with probable regulatory roles in Caenorhabditis elegans. Science 294, 858–862 (2001) | LagosQuintana, M., Rauhut, R., Lendeckel, W. & Tuschl, T. Identification of novel genes coding for small expressed RNAs. Science 294, 853–858 (2001) | Doench, J. G. & Sharp, P. A. Specificity of microRNA target selection in translational repression. Genes Dev. 18, 504–511 (2004) | Lim, L. P. et al. Microarray analysis shows that some microRNAs downregulate large numbers of target mRNAs. Nature 433, 769–773 (2005)
VOLUME 11 | O CTOBER 2010 © 2010 Macmillan Publishers Limited. All rights reserved
RESEARCH HIGHLIGHTS
1 0 - Y E A R A N N I V E R S A RY
A ciliary antenna
GETTY
The biological relevance of the primary cilium, a microtubulebased projection on the surface of most vertebrate cells, was a mystery 10 years ago compared with the established role of motile cilia in force and flow generation. However, a report in 2003 by Huangfu et al. revealed that primary cilia provide a
new way for cells to interact with their environment. Huangfu et al. were carrying out a phenotypic study to identify embryonic patterning mutations in mice and observed that some of the mutants showed abnormalities that are typical of loss of Sonic Hedgehog (SHH) signalling. The genes mutated in these embryos encode the intraflagellar transport (IFT) proteins IFT88, IFT172 and KIF3A (part of the kinesin‑2 motor), which are required for cilium maintenance and growth; this suggested a link between cilia and SHH signalling. Further analysis revealed that IFT proteins are important in SHH signalling and that their loss interferes with the SHH pathway, leading to phenotypic abnormalities. Specifically, although embryos mutated for Patched 1 (the receptor for SHH, which inhibits SHH in the absence of ligand) showed increased levels of SHH signalling, the concomitant mutation of IFT proteins blocked SHH activation and resulted in embryos that were morphologically similar to those mutated for IFT proteins alone. This indicated that
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
IFT proteins function downstream from Patched 1. Similarly, IFT proteins were found to act downstream of RAB23 (which acts downstream of Patched 1 and Smoothened and negatively regulates SHH signalling). By contrast, the authors showed that IFT proteins act upstream of the transcription factor GLI3 (which represses SHH target gene activation in the absence of ligand). Several studies have since focused on how cilia contribute to vertebrate SHH signalling. Together, these reports have revealed that primary cilia function as signalling ‘antennae’, probing the extracellular environment for signalling components, and also as dynamic platforms for signalling, where signalling components localize and interact.
Rachel David
ORIGINAL RESEARCH PAPER Huangfu, D. et al. Hedgehog signalling in the mouse requires intraflagellar transport proteins. Nature 426, 83–87 (2003) FURTHER READING Corbit, K. C. et al. Vertebrate Smoothened functions at the primary cilium. Nature 437, 1018–1021 (2005) | Caspary, T. et al. The graded response to Sonic Hedgehog depends on cilia architecture. Dev. Cell 12, 767–778 (2007)
VOLUME 11 | O CTOBER 2010 © 2010 Macmillan Publishers Limited. All rights reserved
RESEARCH HIGHLIGHTS
1 0 - Y E A R A N N I V E R S A RY
Achieving pluripotency Nuclear reprogramming — whereby the nucleus from a differentiated somatic cell is reprogrammed to a pluripotent embryonic-like state — was initially achieved by two methods: the fusion of somatic cells with embryonic stem (ES) cells, and nuclear transfer, in which the nucleus of a differentiated cell is injected into an enucleated, undifferentiated cell such as an oocyte. Both these methods require the availability of embryonic cells to generate pluripotent stem cells, which remains an ethical issue in addition to being technically challenging. In 2006, Takahashi and Yamanaka revolutionized the stem cell field by showing that it is possible to convert adult somatic cells directly into pluripotent embryonic-like cells by expressing a limited number of transcription factors and culturing the transformed cells under ES cell-like conditions.
They selected 24 genes as candidate factors to induce pluripotency, based on their known role in the maintenance of ES cell identity, and introduced them into mouse fibroblasts harbouring a selective marker that is expressed only in ES cells. Although none of these factors alone could induce the expression of the marker, the 24 genes together did, generating some cells with the growth characteristics and morphology of ES cells, which they called induced pluripotent stem (iPS) cells. Next, they narrowed the set of factors necessary and sufficient to obtain iPS cells down to four: OCT3 (also known as POU5F1 or OCT4), Sry box-containing factor 2 (SOX2), Krüppel-like factor 4 (KLF4) and MYC. Cells expressing these four factors were able to induce the formation of teratomas with all three germ layers in vivo and embryoid bodies in vitro, showing the successful reprogramming of differentiated cells into pluripotent cells.
The following year, two other studies from the Yamanaka and Thompson groups showed that transcription factor-mediated reprogramming to a pluripotent state using just four factors can also be achieved in human cells. iPS cells opened up an entirely new field of stem cell research and the potential for patient-specific therapies. Although the generation of iPS cells has low efficiency and the developmental potential of human iPS cells remains to be fully explored, they are already a key tool for the study of the molecular mechanisms underlying pluripotency. Kim Baumann ORIGINAL RESEARCH PAPERS Takahashi, K. & Yamanaka, S. Induction of pluripotent stem cells from mouse embryonic and adult fibroblast cultures by defined factors. Cell 126, 663–676 (2006) | Takahashi, K. et al. Induction of pluripotent stem cells from adult human fibroblasts by defined factors. Cell 131, 861–872 (2007) | Yu, J. et al. Induced pluripotent stem cell lines derived from human somatic cells. Science 381, 1917–1920 (2007) FURTHER READING Nishikawa, S., Goldstein, R. A. & Nierras, C. R. The promise of human induced pluripotent stem cells for research and therapy. Nature Rev. Mol. Cell Biol. 9, 725–729 (2008)
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
GETTY
VOLUME 11 | O CTOBER 2010 © 2010 Macmillan Publishers Limited. All rights reserved
DIGITAL VISION
RESEARCH HIGHLIGHTS
1 0 - Y E A R A N N I V E R S A RY
The dynamic nucleus Compared with the membranebound organelles of the cytoplasm, the way in which subcompartments of the nucleus are formed and maintained has remained enigmatic. New imaging techniques made it possible to probe this and what emerged was a more dynamic nucleus than anti cipated, capable of self-organization and rapid remodelling. In 2000, Phair and Misteli assessed how proteins move in the nucleus using photobleaching analysis. They measured the mobility of three proteins with diverse nuclear roles and subnuclear localization patterns: the nucleosome-binding protein HMG17, the pre-mRNA splicing factor SF2 (also known as ASF) and the ribosomal RNA processing protein fibrillarin. They fused each
of these proteins to green fluorescent protein (GFP) and used fluorescence recovery after photobleaching (FRAP) to determine their mobility in mammalian cells. Surprisingly, each protein moved throughout the nucleus rapidly and independently of energy. Photobleaching of single nuclear compartments revealed that all three proteins rapidly associate and dissociate from particular bodies, and, using kinetic modelling of data from fluorescence loss in photobleaching (FLIP) analysis, they confirmed that these proteins have short residence times and show high flux through nuclear compartments. Later the same year, Misteli et al. showed that this principle could be extended to a DNA-binding protein on chromatin, the linker histone H1.
Using similar techniques, they assessed the mobility of histone H1 and found that although a large proportion binds chromatin at any time, this pool is constantly exchanging and moving to alternative binding sites. Moreover, hyperacetylation of chromatin, a modification that was known to correlate with chromatin remodelling and transcriptional activation, also led to increased exchange of histone H1. Thus, Misteli et al. proposed that “the dynamic nature of binding is an essential feature of linker histones in their functions as regulators of chromatin remodelling and chromatin structure in vivo.” The repercussions of this work were broad, as they called for a re-assessment of how molecular complexes are organized throughout a cell. It soon became widely accepted that assemblies previously assumed to be static — from nuclear bodies to DNA-bound transcription complexes and multiprotein machines such as the polymerases — are in constant flux, and that this high mobility is important for rapid responses to external stimuli.
What emerged was a more dynamic nucleus than anticipated, capable of selforganization and rapid remodelling.
Alison Schuldt
ORIGINAL RESEARCH PAPERS Phair, R. D. & Misteli, T. High mobility of proteins in the mammalian cell nucleus. Nature 404, 604–609 (2000) | Misteli, T. et al. Dynamic binding of histone H1 to chromatin in living cells. Nature 408, 877–881 (2000)
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
VOLUME 11 | O CTOBER 2010 © 2010 Macmillan Publishers Limited. All rights reserved
RESEARCH HIGHLIGHTS
1 0 - Y E A R A N N I V E R S A RY
Keeping genes quiet RNA interference (RNAi), the use of double-stranded RNAs (dsRNAs) to inhibit the expression of specific genes, had been successfully used in various organisms, including plants and invertebrates, but it was not thought to work in mammals. This is because the introduction of dsRNAs longer than 30 nucleotides activates the interferon response, which mediates the general degradation of RNAs by activating RNase L. However, a study by Tuschl and colleagues in 2001 showed, for the first time, that genes can be effectively silenced by RNAi in mammalian cell lines. To assess whether RNAi can be used in mammals, the authors transfected different cells lines with luciferase
reporter genes together with 21- or 22‑nucleotide small interfering RNAs (siRNAs) that were specific for the reporter genes. The expression of the reporter genes was inhibited by the siRNAs in a sequence-specific manner, albeit at lower levels than was observed in transfected Drosophila melanogaster cells. The introduction of longer (50–500 bp) dsRNAs specific for the reporter genes also induced sequence-specific gene silencing, but this could only be detected when taking into account the non-specific effects of silencing mediated by the interferon response. Together, these results indicated that RNAi is effective in mammalian cells, but the silencing effect of longer dsRNAs is difficult to detect in vivo.
So can RNAi be used to inhibit endogenous RNAs? Although not effective for suppressing vimentin, possibly owing to using a non-optimalsequence siRNA, the introduction of cognate siRNAs for lamin A/C, lamin B and nuclear mitotic apparatus protein successfully decreased their expression. This confirmed that RNAi is not an oddity specific to plants and invertebrates, but can also be used in mammalian cells to rapidly inhibit gene expression. As a result, RNAi is today one of the most widely used tools to study gene function. Rachel David
RNAi ... can also be used in mammalian cells to rapidly inhibit gene expression.
ORIGINAL RESEARCH PAPER Elbashir, S. M. et al. Duplexes of 21‑nucleotide RNAs mediate RNA interference in cultured mammalian cells. Nature 411, 494–498 (2001) FURTHER READING Dykxhoorn, D. M., Novina, C. D. & Sharp, P. A. Killing the messenger: short RNAs that silence gene expression. Nature Rev. Mol. Cell Biol. 4, 457–467 (2003) | Rana, T. M. Illuminating the silence: understanding the structure and function of small RNAs. Nature Rev. Mol. Cell Biol. 8, 23–36 (2007)
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
PHOYOALTO
VOLUME 11 | O CTOBER 2010 © 2010 Macmillan Publishers Limited. All rights reserved
RESEARCH HIGHLIGHTS 1 0 - Y E A R A N N I V E R S A RY
The limits of light Despite the vital role that conventional and confocal light microscopy have had in driving advances in cell biology, their resolution is ultimately limited by the diffraction of light. Light microscopy cannot resolve beyond ~200 nm and, given that most molecular complexes are smaller than this, much of what we want to see has remained beyond our grasp. The development of super-resolution fluorescence microscopy techniques over the past decade allowed this diffraction barrier to be bypassed. In 2000, Klar et al. demonstrated that one way to increase resolution was by deactivating the fluorophores at the periphery of the excitation focal spot of a scanning microscope with a beam of light for stimulated emission depletion (STED) of their excited state. They further showed that this technique worked in bacteria and yeast cells, improving three-dimensional resolution to ~100 nm. They were able, for example, to improve the resolution of labelled vacuolar membranes in live Saccharomyces cerevisiae cells, resolving features not recognized by conventional or confocal microscopy. An alternative solution, photoactivated localization microscopy (PALM), was found in 2006 by Betzig et al. PALM and related techniques, such as stochastic optical reconstruction microscopy (STORM), developed by Rust et al., work on the principle that sparse subsets of labels can be individually photoactivated and then bleached, and their positional information collected en masse can then be assembled into a full image with a more precise position. This approach allowed Betzig et al. to image diverse structures in fixed cells, including focal adhesions and mitochondria, with increased resolution. These techniques and more recent variations have been embraced in diverse fields and are providing unprecedented insights into the organization and function of complex subcellular structures. The hope is that their improved use in live-cell analysis will further increase their potential. Alison Schuldt ORIGINAL RESEARCH PAPERS Klar, T. A. et al. Fluorescence microscopy with diffraction resolution limit broken by stimulated emission. Proc. Natl Acad. Sci. USA 97, 8206–8210 (2000) | Betzig, E. et al. Imaging intracellular fluorescent proteins at nanometer resolution. Science 313, 1642–1645 (2006) | Rust, M. J., Bates, M. & Zhuang, X. Sub-diffraction-limit imaging by stochastic optical reconstruction microscopy (STORM). Nature Methods 3, 793–796 (2006)
MACMILLAN SOUTH AFRICA
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
VOLUME 11 | O CTOBER 2010 © 2010 Macmillan Publishers Limited. All rights reserved
Nikki Walker
RESEARCH HIGHLIGHTS
1 0 - Y E A R A N N I V E R S A RY
Environment dictates behaviour Although the extracellular matrix (ECM) was originally thought to function to keep cells together, it is now widely recognized that it has a more influential and active role as it also controls cell behaviour. In the 1980s, Bissell and colleagues laid a strong foundation for studies on the role of the ECM by showing that the physical properties of the environment dictate epithelial cell differentiation and tumour induction by Rous sarcoma virus; but the functional importance and robustness of the physical properties of the ECM have only been revealed in the past decade. In 2004, McBeath et al. showed that human mesenchymal stem cells (MSCs) differentiate into adipocytes or osteoblasts depending on their shape, which is determined by the density at which they are grown and thus the degree of adhesion to their substrate. The authors also showed that the mechanical cues that drive MSC differentiation are mediated by the small GTPase RHOA, which signals to the cytoskeleton. Two years later, Engler et al. found that different degrees of ECM stiffness direct human MSC fate — MSCs differentiate into neuron-like cells when cultured on soft matrices, into muscle cells on stiffer matrices and into osteoblasts on rigid matrices. They also reported that this information is transmitted by focal adhesions and requires myosin II contractility. These studies showed that the mechanical properties of the environment are sensed by MSCs and can direct lineage specificity, similarly to growth factors. More recently, Levental et al. showed that ECM stiffness affects tumour cell invasion. They showed that breast cancer tumorigenesis is accompanied by collagen cross-linking, which stiffens the ECM. This, in turn, promotes the formation of adhesions, which function as mechanical sensors, and integrin signalling, which induces tumour invasion. The importance of the ECM in controlling cell behaviour has led to many studies aiming to elucidate the principles of mechanosensing, which are only now beginning to emerge. Understanding how cells respond to signals from the microenvironment should increase the efficiency of cancer and stem cell therapies and thus may have important clinical implications. Kim Baumann ORIGINAL RESEARCH PAPERS McBeath, R. et al. Cell shape, cytoskeletal tension, and RhoA regulate stem cell lineage commitment. Dev. Cell 6, 483–495 (2004) | Engler, A. J. et al. Matrix elasticity directs stem cell lineage specification. 126, 677–689 (2006) | Levental, K. R. et al. Matrix crosslinking forces tumor progression by enhancing integrin signaling. Cell 139, 891–906 (2009) FURTHER READING Nelson, C. M. & Bissell, M. J. Of extracellular matrix, scaffolds, and signaling: tissue architecture regulates development, homeostasis, and cancer. Ann. Rev. Cell Dev. Biol. 22, 287–309 (2006)
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
VOLUME 11 | O CTOBER 2010 © 2010 Macmillan Publishers Limited. All rights reserved
RESEARCH HIGHLIGHTS
1 0 - Y E A R A N N I V E R S A RY
All wrapped up in histones The post-translational modification of histones, along with the choice of histone variants, are key factors involved in flagging chromatin for transcriptional activation or repression. Although the late 1990s brought insights into the control of histone acetylation, the regulation of histone methylation was not well understood, and it was unclear whether it is a permanent modification or one that is removed by a potential demethylating enzyme. This changed in 2004 when Shi et al. found that a previously described protein, KIAA0601, was the demethylase that scientists had been seeking for almost 40 years. The enzyme, which they renamed Lys-specific histone demethylase 1 (LSD1), was found to have Lys demethylase activity in vitro. Moreover, they observed that LSD1 demethylates histone H3 on Lys4 in vivo, which coincides with repression of gene expression. Thus, this study “revealed the dynamic nature of histone methylation, and really opened up the idea of multiple dynamic modification of histones controlling gene activity,” says Bill Earnshaw (Edinburgh University, UK). LSD1 targets histone H3, which occurs in four different variants (H3.1, H3.2, H3.3 and centromere protein A (CENPA)) that are incorporated into nucleosomes and can potentially alter gene expression. Although major advances in our understanding of H3 variant deposition during DNA replication had been made, a study in Drosophila melanogaster by Ahmad and Henikoff in 2002 revealed that H3.3 can also be deposited in a DNA
[this study] revealed the dynamic nature of histone methylation, and really opened up the idea of multiple dynamic modification of histones controlling gene activity. Bill Earnshaw, Edinburgh University, UK
replication-independent manner to regulate gene expression. They observed that H3.3 deposition can occur throughout the cell cycle and marks sites of active transcription. A first insight into the mechanism by which H3.3 is deposited was provided in 2004 by Tagami et al., who used a biochemical strategy to purify histone variants and their targeting factors from human cells. They showed that in the H3.3 complex the histone chaperone HIRA, which is distinct from any of the histone chaperones found in the H3.1 complex, is necessary for the deposition of H3.3 in vitro during DNA replicationindependent nucleosome assembly. This study started the search for the exact role of many other histone chaperones in histone dynamics. Together with previous studies identifying the dynamic nature of other histone modifications such as acetylation and the role of histone tails in transcription activation, these findings provided important insights into the role of epigenetics in genome function.
Rachel David
ORIGINAL RESEARCH PAPERS Shi, Y. et al. Histone demethylation mediated by nuclear amine oxidase homolog LSD1. Cell 119, 941–953 (2004) | Ahmad, K. & Henikoff, S. The histone variant H3.3 marks active chromatin by replication-independent nucleosome assembly. Mol. Cell 9, 1191–1200 (2002) | Tagami, H. et al. Histone H3.1 and H3.3 complexes mediate nucleosome assembly pathways dependent or independent of DNA synthesis. Cell 116, 51–61 (2004) FURTHER READING Klose, R. J. & Zhang, Y. Regulation of histone methylation by demethylimination and demethylation. Nature Rev. Mol. Cell Biol. 8, 307–318 (2007) | Talbert, P. & Henikoff, S. Histone variants — ancient wrap artists of the epigenome. Nature Rev. Mol. Cell Biol. 11, 264–275 (2010)
PHOYOALTO
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
VOLUME 11 | O CTOBER 2010 © 2010 Macmillan Publishers Limited. All rights reserved
RESEARCH HIGHLIGHTS
1 0 - Y E A R A N N I V E R S A RY
Sensing and controlling protein dynamics Several studies in the past decade highlighted the value of biosensors as research tools and revealed new insights into the dynamics of small GTPases in living cells. The small GTPases RAC1, RHOA and CDC42, which regulate the cytoskeletal dynamics that drive cell motility, only interact with effectors in their active, GTP-bound form. In 2000, Kraynov et al. developed a method to quantify the spatiotemporal dynamics of RAC1 activity in live cells. A fragment of the RAC1 effector protein p21‑activated kinase 1 (PAK1) was labelled with Alexa‑546 to create a biosensor. In
The RAC1 biosensor in a motile fibroblast. RAC1 activity is elevated in protrusions at the leading edge (red and yellow). Image courtesy of K. M. Hahn, University of North Carolina, Chapel Hill, North Carolina, USA.
cells expressing this biosensor and green fluorescent protein (GFP)-tagged RAC1, the Alexa and GFP fluorophores undergo fluorescence resonance energy transfer (FRET) to emit a unique fluorescence signal when PAK1 binds RAC1–GTP. This technique revealed precise spatial control of RAC1 activation at the leading edge of motile cells. In 2009, the Hahn group developed the means to go beyond the visualization of RAC1 activation by controlling local, acute RAC1 activation in live cells. Wu et al. described a genetically encoded photoactivatable RAC1 (PA‑RAC1), consisting of the photoactivatable LOV domain from Avena sativa phototropin fused to constitutively active RAC1. In this construct, a helix linking LOV to RAC1, which blocks RAC1 binding to effectors in the dark, is unwound by light; this activates RAC1 to generate cell protrusions and ruffles. Importantly, this construct was also used by Yoo et al. and Wang et al. in 2010 to manipulate cell motility in live Danio rerio and Drosophila melanogaster, respectively. Also in 2009, Machacek et al. used an improved version of the RAC1 biosensor, together with biosensors for RHOA and CDC42, to assess the dynamics of, and relationship between, small GTPases during cell protrusion. RAC1, RHOA and CDC42 biosensor activity was measured separately at the leading edge of a cell. Computational multiplexing (a
mathematical method that correlates multiple time-dependent variables obtained during time-lapse imaging) revealed that RHOA is active at the leading edge synchronously with protrusion, and RAC1 and CDC42 are activated 40 seconds later, 2 μm behind the leading edge. The visualization of RHOA and CDC42 biosensors simultaneously, using four-channel imaging, also showed that CDC42 is activated after RHOA. Thus, coupling biosensors with image analysis tools further enhances their potential. Together, these studies allowed the activity of small GTPases to be visualized and controlled at the leading edge of live cells. They also paved the way for the development of other biosensors, such as the one recently used by Grashoff et al. to measure the mechanical force across proteins. Katharine H. Wrighton
The work of Klaus Hahn has transformed biosensors into major cell biological tools that have revealed new and important insights. Rick Horwitz, University of Virginia, USA
ORIGINAL RESEARCH PAPERS Kraynov, V. S. et al. Localized Rac activation dynamics visualized in living cells. Science 290, 333–337 (2000) | Wu, Y. et al. A genetically encoded photoactivatable Rac controls the motility of living cells. Nature 461, 104–108 (2009) | Machacek, M. et al. Coordination of Rho GTPase activities during cell protrusion. Nature 461, 99–103 (2009) | Yoo, S. K. et al. Differential regulation of protrusion and polarity by PI(3)K during neutrophil motility in live zebrafish. Dev. Cell 18, 226–236 (2010) | Wang, X. et al. Lightmediated activation reveals a key role for Rac in collective guidance of cell movement in vivo. Nature Cell Biol. 12, 591–597 (2010) | Grashoff, C. et al. Measuring mechanical tension across vinculin reveals regulation of focal adhesion dynamics. Nature 466, 263–266 (2010)
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
VOLUME 11 | O CTOBER 2010 © 2010 Macmillan Publishers Limited. All rights reserved
RESEARCH HIGHLIGHTS 1 0 - Y E A R A N N I V E R S A RY
The importance of ‘self-eating’ During autophagy, double-membrane structures called autophagosomes engulf cytosol or organelles and deliver them to lysosomes to be degraded and released as nutrients. Few scientists need this explanation now, but this wasn’t true a decade ago when the molecular control of mammalian autophagy was only beginning to emerge. The finding that autophagy is implicated in human pathophysiologies, including tumorigenesis and neurodegeneration, highlighted its crucial role as a dynamic and selective cellular process. In 2003, beclin 1 (BECN1; also known as autophagy-related gene 6 (ATG6)), the protein product of which is required for autophagosome formation, was established as a haploinsufficient tumour suppressor gene. Driven to understand why BECN1 was monoallelically deleted in up to 75% of human sporadic breast, ovarian and prostate cancers, Qu et al. generated Becn1+/– mice, which showed an increase in the frequency of spontaneous malignancies and cell proliferation, and reduced autophagy, in vivo. Yue et al. found that Becn1–/– mice died early in embryogenesis and that Becn1+/– mice had a high incidence of tumours. Furthermore, Becn1–/– mouse embryonic stem cells were deficient in their autophagic response. These studies established that BECN1 and consequently autophagy have a role in tumour suppression. Autophagy was linked to neurodegeneration in 2002 when Ravikumar et al. found that aggregate-prone proteins that are typically associated with Alzheimer’s disease accumulated in cells treated with autophagy inhibitors, whereas the stimulation of autophagy enhanced their clearance, suggesting that autophagy
protects against neurodegeneration. Komatsu et al. provided further evidence for this in 2006 when they found that mice lacking the essential autophagy gene Atg7 in the nervous system showed behavioural defects and neuronal loss. At the same time, Hara et al. showed that mice deficient for Atg5 in neural cells develop defects in motor function and accumulate cytoplasmic inclusion bodies in neurons. Thus, these studies showed that basal autophagy has a role in preventing neurodegeneration, highlighting the possibility that it might be protective in other diseases. Indeed, “all of us are relying on autophagy to protect us from various diseases, even if we are healthy”, says Daniel Klionsky at the University of Michigan, USA. Now that autophagy is firmly linked to human pathophysiology, the emphasis is on understanding its selectivity in particular conditions and the subcellular membrane trafficking events that contribute to this. The hope is that the next decade could see the development of disease therapies that target the autophagic pathway.
all of us are relying on autophagy to protect us from various diseases, even if we are healthy. Daniel Klionsky, University of Michigan, USA
Katharine H. Wrighton ORIGINAL RESEARCH PAPERS Qu, X. et al. Promotion of tumorigenesis by heterozygous disruption of the beclin 1 autophagy gene. J. Clin. Invest. 112, 1809–1820 (2003) | Yue, Z. et al. Beclin 1, an autophagy gene essential for early embryonic development, is a haploinsufficient tumor suppressor. Proc. Natl Acad. Sci. USA 100, 15077–15082 (2003) | Ravikumar, B. Duden, R. & Rubinsztein D. C. Aggregate-prone proteins with polyglutamine and polyalanine expansions are degraded by autophagy. Hum. Mol. Genet. 11, 1107–1117 (2002) | Komatsu, M. et al. Loss of autophagy in the central nervous system causes neurodegeneration in mice. Nature 441, 880–884 (2006) | Hara, T. et al. Suppression of basal autophagy in neural cells causes neurodegenerative disease in mice. Nature 441, 885–889 (2006) FURTHER READING Klionsky, D. J. Autophagy: from phenomenology to molecular understanding in less than a decade. Nature Rev. Mol. Cell Biol. 8, 931–937 (2007)
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
VOLUME 11 | O CTOBER 2010 © 2010 Macmillan Publishers Limited. All rights reserved