2
Last, Almeida et al. present a tantalizing speculation. Maybe the no-signalling criter- ion is to blame for the existence of useless Bell inequalities in a multi-party situation; that criterion is an inherently two-party statement. If one found a proper extension to many play- ers, these useless inequalities might become trivial and go away. This would indeed provide a new principle, bringing physicists closer to understanding what makes quantum corre- lations so special. But it is also possible that, even for two players, there are ‘useless’ facet Bell inequalities and that even their existence could lead to a positive insight. Andreas Winter is in the Department of Mathematics, University of Bristol, Bristol BS8 1TW, UK, and at the Centre for Quantum Technologies, National University of Singapore, Block S15, 3 Science Drive 2, Singapore 117543. e-mail: [email protected] 1. Almeida, M. L. et al. Phys. Rev. Lett. 104, 230404 (2010). 2. Bell, J. S. Physics 1, 195–200 (1964). 3. Einstein, A., Podolsky, B. & Rosen, N. Phys. Rev. 47, 777–780 (1935). 4. Clauser, J. F., Horne, M. A., Shimony, A. & Holt, R. A. Phys. Rev. Lett. 23, 880–884 (1969). 5. Aspect, A., Grangier, P. & Roger, G. Phys. Rev. Lett. 49, 91–94 (1982). 6. Tsirelson, B. Hadronic J. Suppl. 8, 329–345 (1993). 7. Popescu, S. & Rohrlich, D. Found. Phys. 24, 379–385 (1994). 8. Gill, R. Bell inequalities holding for all quantum states: problem 26B; available at www.imaph.tu-bs.de/qi/ problems/26.html (28 February 2006). 9. Linden, N., Popescu, S., Short, A. J. & Winter, A. Phys. Rev. Lett. 99, 180502 (2007). 10. Acín, A. et al. Phys. Rev. Lett. 104, 140404 (2010). 11. Cirel’son, B. S. Lett. Math. Phys. 4, 93–100 (1980). CANCER Viruses’ backup plan Kevin M. Ryan Tumour viruses can cause cancer by altering gene expression and protein activity in the host cell. Tumour adenoviruses, however, seem to go to great lengths to ensure that one particular host cell protein, p53, is suppressed. Analyses of tumour-causing viruses have been instrumental in understanding cancer biology. The first vertebrate oncogene, for example, was discovered on the basis of its sequence simi- larity to the cancer-causing gene of a chicken retrovirus. Numerous tumour-suppressor genes have also been identified because they are cellular targets of tumour-causing DNA viruses 1 . On page 1076 of this issue, Soria et al. 2 describe a previously unknown mechanism that tumour-inducing adenoviruses use to suppress the activity of the tumour-suppressor protein p53. Their findings call for a revision of the dogma on how these viruses make cells cancerous. Many tumour-inducing DNA viruses com- monly cause abnormal activity of the cellular transcription factor E2F1, through binding to proteins of the retinoblastoma family such as Rb, which would normally be bound to E2F1 (refs 3, 4). In cells that are infected with cancer-causing strains of adenovirus, the viral E1A protein binds to proteins of the host cell’s retinoblastoma family, freeing E2F1 to promote not only progression of the cell cycle — a cen- tral requirement of the viral life cycle — but also transcription of many viral genes 3 (Fig. 1a). Stimulation of cell-cycle progression by E1A would be too simple a route to inducing cancer, and host cells often have safeguard mechanisms to overcome a single precancerous change. In this case, the host cell can counter the possible change because deregulated E2F1 also acti- vates the cellular ARF protein, which inhibits MDM2 — the main inhibitor of p53 (ref. 5). In the absence of MDM2 activity, p53 levels increase, either inhibiting cell-cycle progression or destroying the infected cell. no-signalling, that prevent nature from being even more non-local? A flurry of recent work, as discussed by Almeida et al. 1 , has addressed this question by attempting to show that a world with ‘super-quantum’ correlations would indeed be a stranger place — one that would differ markedly from the quantum universe as we know it. From the point of view of no-signalling correlations, quantum correlations are mere special cases, and even more special are local realistic correlations, but it is this that makes them more interesting. The no-signalling prop- erty is entirely captured by the positivity of the values P(ab|xy…) — after all, they are probabilities — and by certain identities between them. All of these constraints are expressed as Bell inequalities, properly called trivial Bell inequalities. But, in the context of no-signalling correlations, what is interesting are the non-trivial ones, those that express restrictions on correlations that are not a priori implied by the no-signalling principle. It is thus natural to ask if every non-trivial Bell inequality is violated by quantum mechanics 8 . Almeida et al. 1 now answer this question in the negative, by describing and analysing a non-local game, ‘guess your neighbour’s input’. For any number of players greater than two, their game results in a Bell inequality that is true even for quantum correlations, yet it is non-trivial because its violation is consistent with no signalling. In fact, it is a ‘facet inequal- ity’, meaning that it is not implied by any other Bell inequality — that is, the authors identify a demonstrably useless yet necessary Bell inequality. The corresponding experiment would not be able to discriminate between clas- sical and quantum physics. Instead, it would distinguish both of these from the non-phys- ical realm of no-signalling correlations. (For two players, earlier investigations on a class of games called ‘non-local computation’ 9 showed inviolable Bell inequalities — but it was found later that these are not facet inequalities; M. L. Almeida, N. Brunner and P. Skrzypczyk, personal communication.) So why will experts get excited about Almeida and colleagues’ result 1 ? First, it has an immediate application in showing that a multi-party extension of Gleason’s theorem 10 , a notable contribution to quantum foundations, fails for three or more parties. Second, and more importantly, it ties in with other recent attempts to find common ground between classical and quantum physics that would distinguish them from weird, no-signalling, super-quantum theories. Correlations x y a b Player 1 Player 2 Figure 1 | The CHSH game. In this illustration of the game, two cooperating players receive an input, x, y (where x and y can be 0 or 1), from a referee, and they have to reply with a, b (where a and b can also be 0 or 1), respectively. The game’s winning condition is that the parity of a + b has to equal the product of x and y — that is, a and b have to be different from each other if x = y =1 and equal to each other for the other three possible combinations of x and y. If the players adopt a classical strategy (that is, if they are correlated according to the laws of classical physics), only any three of the four pairs of inputs — (0,0), (0,1), (1,0) and (1,1) — can be satisfied, and the maximum probability of winning, P win , is 0.75 (ref. 4). However, if they adopt a strategy based on quantum correlations, this value can be increased to 0.851 (refs 4–7, 11). The maximum value of P win , P win = 1, is consistent with ‘no-signalling’ correlations 6,7 . Almeida et al. 1 describe a game much like this, for three or more players, in which quantum mechanics does not provide an advantage over classical physics, although no-signalling correlations do. 1054 NATURE|Vol 466|26 August 2010 NEWS & VIEWS © 20 Macmillan Publishers Limited. All rights reserved 10

Cancer: Viruses' backup plan

  • Upload
    kevin-m

  • View
    219

  • Download
    4

Embed Size (px)

Citation preview

Page 1: Cancer: Viruses' backup plan

Last, Almeida et al. present a tantalizing speculation. Maybe the no-signalling criter-ion is to blame for the existence of useless Bell in equalities in a multi-party situation; that criterion is an inherently two-party statement. If one found a proper extension to many play-ers, these useless inequalities might become trivial and go away. This would indeed provide a new principle, bringing physicists closer to understanding what makes quantum corre-lations so special. But it is also possible that, even for two players, there are ‘useless’ facet Bell inequalities and that even their existence could lead to a positive insight. ■

Andreas Winter is in the Department of Mathematics, University of Bristol, Bristol BS8 1TW, UK, and at the Centre for Quantum Technologies, National University of Singapore, Block S15, 3 Science Drive 2, Singapore 117543.e-mail: [email protected]

1. Almeida, M. L. et al. Phys. Rev. Lett. 104, 230404 (2010).2. Bell, J. S. Physics 1, 195–200 (1964).3. Einstein, A., Podolsky, B. & Rosen, N. Phys. Rev. 47, 777–780

(1935).4. Clauser, J. F., Horne, M. A., Shimony, A. & Holt, R. A. Phys.

Rev. Lett. 23, 880–884 (1969).5. Aspect, A., Grangier, P. & Roger, G. Phys. Rev. Lett. 49,

91–94 (1982). 6. Tsirelson, B. Hadronic J. Suppl. 8, 329–345 (1993).7. Popescu, S. & Rohrlich, D. Found. Phys. 24, 379–385 (1994).8. Gill, R. Bell inequalities holding for all quantum states:

problem 26B; available at www.imaph.tu-bs.de/qi/problems/26.html (28 February 2006).

9. Linden, N., Popescu, S., Short, A. J. & Winter, A. Phys. Rev. Lett. 99, 180502 (2007).

10. Acín, A. et al. Phys. Rev. Lett. 104, 140404 (2010).11. Cirel’son, B. S. Lett. Math. Phys. 4, 93–100 (1980).

CANCER

Viruses’ backup planKevin M. Ryan

Tumour viruses can cause cancer by altering gene expression and protein activity in the host cell. Tumour adenoviruses, however, seem to go to great lengths to ensure that one particular host cell protein, p53, is suppressed.

Analyses of tumour-causing viruses have been instrumental in understanding cancer biology. The first vertebrate oncogene, for example, was discovered on the basis of its sequence simi-larity to the cancer-causing gene of a chicken retrovirus. Numerous tumour-suppressor genes have also been identified because they are cellular targets of tumour-causing DNA viruses1. On page 1076 of this issue, Soria et al.2 describe a previously unknown mechanism that tumour-inducing adenoviruses use to suppress the activity of the tumour-suppressor protein p53. Their findings call for a revision of the dogma on how these viruses make cells cancerous.

Many tumour-inducing DNA viruses com-monly cause abnormal activity of the cellular transcription factor E2F1, through binding to proteins of the retinoblastoma family such

as Rb, which would normally be bound to E2F1 (refs 3, 4). In cells that are infected with cancer-causing strains of adeno virus, the viral E1A protein binds to proteins of the host cell’s retino blastoma family, freeing E2F1 to promote not only progression of the cell cycle — a cen-tral requirement of the viral life cycle — but also transcription of many viral genes3 (Fig. 1a).

Stimulation of cell-cycle progression by E1A would be too simple a route to inducing cancer, and host cells often have safeguard mechanisms to overcome a single precancerous change. In this case, the host cell can counter the possible change because deregulated E2F1 also acti-vates the cellular ARF protein, which inhibits MDM2 — the main inhibitor of p53 (ref. 5). In the absence of MDM2 activity, p53 levels increase, either inhibiting cell-cycle progression or destroying the infected cell.

no-signalling, that prevent nature from being even more non-local? A flurry of recent work, as discussed by Almeida et al.1, has addressed this question by attempting to show that a world with ‘super-quantum’ correlations would indeed be a stranger place — one that would differ markedly from the quantum universe as we know it.

From the point of view of no-signalling correlations, quantum correlations are mere special cases, and even more special are local realistic correlations, but it is this that makes them more interesting. The no-signalling prop-erty is entirely captured by the positivity of the values P(ab…|xy…) — after all, they are probabilities — and by certain identities between them. All of these constraints are expressed as Bell inequalities, properly called trivial Bell inequalities. But, in the context of no-signalling correlations, what is interesting are the non-trivial ones, those that express restrictions on correlations that are not a priori implied by the no-signalling principle. It is thus natural to ask if every non-trivial Bell inequality

is violated by quantum mechanics8.Almeida et al.1 now answer this question

in the negative, by describing and analysing a non-local game, ‘guess your neighbour’s input’. For any number of players greater than two, their game results in a Bell inequality that is true even for quantum correlations, yet it is non-trivial because its violation is consistent with no signalling. In fact, it is a ‘facet inequal-ity’, meaning that it is not implied by any other Bell inequality — that is, the authors identify a demonstrably useless yet necessary Bell in equality. The corresponding experiment would not be able to discriminate between clas-sical and quantum physics. Instead, it would distinguish both of these from the non-phys-ical realm of no-signalling correlations. (For two players, earlier investigations on a class of games called ‘non-local computation’9 showed inviolable Bell inequalities — but it was found later that these are not facet inequalities; M. L. Almeida, N. Brunner and P. Skrzypczyk, personal communication.)

So why will experts get excited about Almeida and colleagues’ result1? First, it has an immediate application in showing that a multi-party extension of Gleason’s theorem10, a notable contribution to quantum foundations, fails for three or more parties. Second, and more importantly, it ties in with other recent attempts to find common ground between classical and quantum physics that would distinguish them from weird, no-signalling, super-quantum theories.

Correlations

x y

a b

Player 1 Player 2

Figure 1 | The CHSH game. In this illustration of the game, two cooperating players receive an input, x, y (where x and y can be 0 or 1), from a referee, and they have to reply with a, b (where a and b can also be 0 or 1), respectively. The game’s winning condition is that the parity of a + b has to equal the product of x and y — that is, a and b have to be different from each other if x = y =1 and equal to each other for the other three possible combinations of x and y. If the players adopt a classical strategy (that is, if they are correlated according to the laws of classical physics), only any three of the four pairs of inputs — (0,0), (0,1), (1,0) and (1,1) — can be satisfied, and the maximum probability of winning, Pwin, is 0.75 (ref. 4). However, if they adopt a strategy based on quantum correlations, this value can be increased to 0.851 (refs 4–7, 11). The maximum value of Pwin, Pwin = 1, is consistent with ‘no-signalling’ correlations6,7. Almeida et al.1 describe a game much like this, for three or more players, in which quantum mechanics does not provide an advantage over classical physics, although no-signalling correlations do.

1054

NATURE|Vol 466|26 August 2010NEWS & VIEWS

© 20 Macmillan Publishers Limited. All rights reserved10

Page 2: Cancer: Viruses' backup plan

In the ‘cat-and-mouse’ games of evolution, tumour-causing adenoviruses have devised ways to deal with this p53 activity that arises as a result of the host cell’s response to E1A. Adeno viruses also produce a protein called E1B-55k, which binds to p53 and both inhibits its transcription-promoting activity and directs its degradation by the proteasome6 (Fig. 1b). For many years, this has stood as the central mechanism by which adeno viruses — and by analogy other tumour viruses — are thought to inactivate Rb and p53.

Soria et al.2 were intrigued by their observa-tion that infection with an adenoviral mutant lacking E1B-55k results in high levels of p53 — as expected — but not in the activation of p53’s target genes. They proposed that the virus must use a mechanism to inactivate p53 in addition to the effects it exerts through E1B-55k.

The authors’ analysis of adenoviruses with several mutations — deletion of the gene encoding E1B-55k, as well as other genes — led to the identification of E4-ORF3 as a sec-ond repressor of p53 (Fig. 1c). This protein selectively silences the activation of p53 target genes by mediating the methylation of histone proteins at these genes’ promoters. E4-ORF3 does this through its effect on the cellular methylase enzymes SUV39H1 and SUV39H2. Intriguingly, E1A also seems to be required for effective transcription of the gene encoding E4-ORF3.

At first, these results might seem paradoxi-cal — why would an adenovirus, which does not have a genome big enough to afford func-tional redundancy, have multiple mechanisms to inactivate p53? This question might point to the importance and potency of p53, such that it may be too risky for the virus to rely on just one mechanism for inactivating the protein. Previous work7 also revealed that only a robust decrease in p53 production would inactivate its tumour-suppressive activity. The effects of E1B-55k might therefore be insuf-ficient to inactivate p53 completely, leading to the evolution of other strategies to work in concert.

Alternatively, complete repression of p53 through E1B-55k might not be ideal for the virus. In this case, E4-ORF3 could selectively inhibit p53 target genes that impede viral replication, sparing other targets that might benefit the virus. One target that might fit into the latter category is the gene encoding TIGAR, which redirects intermediates from one metabolic process, glycolysis, to another, the pentose-phosphate pathway8. This altera-tion in metabolism could be beneficial for the high biosynthetic demands of a virally infected cell.

A pertinent question is whether Soria and colleagues’ findings are relevant for human cancers that do not involve a viral component. The authors correctly point out that many of the cellular proteins that bind to E4-ORF3 are mutated in human cancer. Consequently, these mutations might affect the expression of p53

target genes. Nonetheless, in response to stress, many tumour-cell lines carrying wild-type p53 can activate p53 target genes, indicating that it is not a common event in cancer for these genes to be silenced by methylation of the histones at their promoters. Does this therefore mean that the effects that Soria et al. report are specific to tumour viruses or occur only in certain tumour settings? Alternatively, perhaps the methyla-tion of histones at p53-binding sites occurs frequently in human cancer but is lost during the process of establishing tumour-cell lines. Further investigations should yield answers to these intriguing questions.

Undoubtedly, the greatest significance of this study2 will be its contribution to devising strategies to treat cancer. Adenoviruses must inactivate p53 so that they can replicate and subsequently induce the breakdown of the infected cell. Because many tumours lack p53 function5, researchers have engineered viruses that lack E1B-55k with the idea that these viruses would replicate selectively in tumour cells lacking p53 but not in normal cells, even-tually leading to the death of the tumour cells9. Although these engineered viruses have proven to be therapeutically beneficial, their replication

does not seem to depend on the p53 ‘status’ of the cell10. Soria and co-workers’ finding that E4-ORF3 also, at least partly, inactivates p53 function provides an explanation for why this would be the case. Following on from these insights is the exciting prospect that adeno-viruses lacking both E1B-55k and E4-ORF3 could be selective and even more potent anticancer agents than viruses lacking just E1B-55k. ■

Kevin M. Ryan is at the Beatson Institute for Cancer Research, Glasgow G61 1BD, UK.e-mail: [email protected]

1. Howley, P. M. & Livingston, D. M. Virology 384, 256–259 (2009).

2. Soria, C., Estermann, F. E., Espantman, K. C. & O’Shea, C. C. Nature 466, 1076–1081 (2010).

3. Frisch, S. M. & Mymryk, J. S. Nature Rev. Mol. Cell Biol. 3, 441–452 (2002).

4. Helgason, G. V., O’Prey, J. & Ryan, K. M. Cancer Res. 70, 4074–4080 (2010).

5. Sherr, C. J. & McCormick, F. Cancer Cell 2, 103–112 (2002).

6. Berk, A. J. Oncogene 24, 7673–7685 (2005).7. Hemann, M. T. et al. Nature Genet. 33, 396–400

(2003).8. Bensaad, K. et al. Cell 126, 107–120 (2006).9. O’Shea, C. C. Oncogene 24, 7640–7655 (2005).10. O’Shea, C. C. et al. Cancer Cell 6, 611–623 (2004).

Ub

UbUb

p53

p53 p53Ub

MeMeMeMe

E1A

AdenovirusRb

E2F1

ARF

MDM2

p53

p53 p53

p53

E1B-55k

E4-ORF3

Promoters of p53 target genes

SUV39H1

SUV39H2

Inhibition of p53-mediatedgene transcription

Cell death

b

c

a

p53 p53

p53p53

55k

55k55k

55k

p53

XX

Histone

Figure 1 | Inactivating p53. a, The adenoviral protein E1A binds to cellular proteins of the retinoblastoma family (such as Rb). This results in the release of E2F1 from binding to retinoblastoma proteins, which — among other effects — leads to the ARF-mediated inhibition of MDM2. This causes p53 to accumulate because it is not targeted (black cross) for degradation by MDM2-mediated ubiquitylation (Ub is a ubiquitin protein). This increase in p53, in turn, arrests the cell’s proliferation and directs its death. b, To counter this effect on the cell’s viability, the virus produces another protein, E1B-55k, which (together with the E4-ORF6 protein, not shown) not only binds to p53’s transactivation domain, impairing the transcription-factor activity of p53, but also directs p53 to be degraded. c, Soria et al.2 find that, as an additional measure, the virus expresses the gene encoding E4-ORF3 (a process for which E1A is required). E4-ORF3 either selectively or completely silences the expression of p53 target genes by mediating the methylation (Me is a methyl group) of histones at their promoters through its effect on the cellular histone methyltransferase enzymes SUV39H1 and SUV39H2.

1055

NATURE|Vol 466|26 August 2010 NEWS & VIEWS

© 20 Macmillan Publishers Limited. All rights reserved10