The retinoblastoma tumour suppressor in development and cancer

Key Points

• The retinoblastoma gene (RB) was the first tumour suppressor to be cloned, but the mechanism behind its role in tumour suppression remains unclear.

• The retinoblastoma protein (RB) has been implicated in many cellular processes, such as regulation of the cell cycle, DNA-damage responses, DNA repair, DNA replication, protection against apoptosis, and differentiation, all of which could contribute to its function as a tumour suppressor.

• RB is closely related to two genes in mice and humans (p107 and 130, which are not commonly mutated in tumours), which have been found to have partially redundant as well as opposing functions in genetic experiments in mice. Future studies are required to identify the similarities and differences between these proteins, which might explain the seemingly unique role of RB as a tumour suppressor.

• As retinoblastomas do not arise as a consequence of Rb loss in mice, this model system is not ideal for studying this disease. However, this model system has provided information regarding the role of RB, p107 and p130 in development and tumorigenesis. Further experiments with conditional knockouts, as well as inter-crossing experiments, will provide a better insight into their roles in biological processes.

• Biochemical experiments have provided information regarding the nature of various protein interactions with RB. Considering that there are more than 100 reported RB-binding proteins, much detailed structure/function mapping is required to clarify the biological relevance of all these interactions.

• RB has also been associated with tumour growth, based on studies in which a constitutively active Rb allele gives rise to dysplasia in transgenic mice.

Abstract

Since its discovery, the retinoblastoma (RB) tumour-suppressor protein has been a focal point of cancer research. Accumulating evidence indicates a complex role for RB in cell proliferation, differentiation and survival. To further complicate matters, proteins that are related to RB have redundant as well as antagonistic functions. Recent studies of knockout mice and cells that lack one or more of these proteins have begun to clarify their various context-specific functions and the unique activity of this tumour suppressor.

Main

It was first suggested more than a century ago that inherited abnormalities predispose some individuals to cancer (for review, see Ref. 1), but it is only in the past two decades, with the development of molecular cloning methodologies, that cancer-associated genes have been identified. These were found to include both oncogenes (dominant gain-of-function proteins) and tumour-suppressor genes (recessive loss-of-function proteins), and both classes of genes were discovered principally by virtue of their alteration in human cancers. Much of our current understanding of these genes has come from biochemical and functional studies in cultured cells; however, as described below, more recent clues to the mechanism by which the dysfunction of these genes contributes to human cancer has come from animal models.

The first tumour-suppressor gene to be cloned was the retinoblastoma ( RB ) gene^2,3,4,5. In familial cases of retinoblastoma, a germline mutation in the RB gene is inherited, and affected individuals develop retinal tumours in which the loss of the second RB allele is the rate-limiting step (reviewed in Ref. 6). The identification and subsequent cloning of the RB gene^3,4 launched a new era in cancer genetics, and RB was subsequently found to be mutated in other human cancers, such as small-cell lung carcinoma and osteosarcoma. The RB gene product, RB, was identified as an important target of oncoproteins that are expressed by DNA tumour viruses, such as adenovirus E1A, SV40 T antigen and the E7 proteins of human papillomaviruses. These findings indicated that RB is a regulator of cell proliferation^7,8,9,10 (Fig. 1). The discovery that RB overexpression caused cells to undergo arrest in the G1 phase of the cell cycle^11,12,13, whereas cells that are deficient in RB show an accelerated G1 transition^14,15,16, also provided support for the idea of RB as a cell-proliferation inhibitor.

Figure 1: The RB pathway and the many potential functions of RB.

The E2F transcription-factor family has been shown to regulate the expression of genes, products of which are involved in G1–S phase transition. The cell cycle is depicted at the bottom of the figure with its various phases (G0–G1–S–G/M). E2F-mediated transcription is negatively regulated by the RB family of proteins and their activity is, in turn, regulated by CDKs (CDK2, CDK4, and CDK6) and their inhibitors (p15, p16, p18, p19, p21 and p27). Molecules in this regulatory network (p16, cyclin D, CDK4/CDK6 and RB) (blue) have been found altered in a wide variety of tumours, with the exception of E2F. In addition to its ability to block S-phase progression, RB is believed to regulate a wide variety of processes, such as apoptosis, DNA replication, DNA repair, checkpoint control and differentiation (red). It is unclear whether these functions are all mediated through regulation of E2F or are dependent on one of the many (more than 100) factors with which RB can interact. Little is known about the impact the CDK regulatory network has on these alternative functions of RB.

The identification of viral oncoproteins that interact with and presumably interrupt the function of RB also raised the possibility that analogous cellular proteins could exist. Subsequent experimentation led to the identification of the E2F1 transcription factor as the first cellular target of RB^17,18,19 (reviewed in Refs 20,21). Verification that E2F1 regulates cell proliferation came from studies in which it was shown that E2F1 overexpression can promote transition from the G1 phase to the S phase of the cell cycle²¹. Additional studies revealed that E2F1 is a member of a family of closely related proteins, several of which have been found to interact with RB²² (Fig. 2). Most current models propose that the role of RB in controlling cell-cycle progression is exerted through its repressive effects on gene expression mediated by the E2Fs^{21,23,24,25,26,27,28,29,31}.

Figure 2: RB and E2F function.

a | RB is believed to bind the transcription factor E2F and its associated subunit DP. RB represses E2F-mediated transcription by recruiting chromatin remodelling complexes to the promoter in resting cells. At the G1–S-phase transition, RB is thought to be phosphorylated by CDK2, CDK4 and CDK6. Hyperphosphorylated RB releases E2F, allowing it to activate transcription of its target genes. b | The mammalian genome encodes at least six different E2F proteins, listed as E2F1–6. The E2F transcription factors can be subdivided into three categories based on their transcriptional properties as well as their interaction with different pocket proteins. The first group consists of E2F1, E2F2 and E2F3, which are potent transcriptional activators that can also interact with RB. This presumably represses the activating potential of these three E2Fs. E2F4 and E2F5, which represent the second group, are fairly weak transcriptional activators that seem to more readily form complexes with p107 and p130 (although E2F4 will interact with all three pocket proteins). The third category, E2F6, does not associate with pocket proteins and seems to act as a transcriptional repressor²². The RB family of proteins (in most cases, it has only been tested for RB) has been shown to interact with chromatin remodelling complexes, such as the SWI/SNF complex, the histone deacetylases (HDACs), polycomb group (PcG) proteins as well as methylases. Recruitment of these complexes to E2F-binding sites is thought to mediate the repressive function of RB-family members^32,33.

It was recently shown that the repressive action of RB in regulating gene expression occurs, in part, via the recruitment of chromatin-remodelling complexes to promoter regions. These complexes mediate chromatin condensation and subsequent inhibition of transcription^32,33,34. The ability of RB to repress E2F-mediated transcription is regulated by the phosphorylation of RB by cyclin-dependent kinases (CDKs). This phosphorylation disrupts the association between E2Fs and RB^35,36. RB is therefore linked directly to the intricate regulatory network that controls cell-cycle progression³⁷ (Figs 1,2). Alterations in cell-cycle regulatory genes that encode proteins that participate in the regulation of RB function are also commonly observed in a broad spectrum of tumour types. Taken together, these findings indicate that deregulation of the normal pathway in which RB functions is a common and important event in tumorigenesis (Fig. 1).

Efforts to clearly establish the mechanism by which RB acts as a tumour suppressor have been complicated by various factors. For example, in addition to its role in cell-cycle control, RB has been implicated in regulating a wide variety of cellular processes, including DNA replication^38,39,40, differentiation⁴¹ and apoptosis^41,42 (Fig. 1). It is easy to imagine that a decrease in differentiation potential, in addition to an increase in proliferative rates, both of which are seen in RB-deficient cells, could contribute to tumorigenesis. However, in light of accumulating evidence indicating that inhibition of apoptosis contributes to tumorigenesis, it is more difficult to reconcile RB's role as a tumour suppressor, with reports that loss of RB leads to increased apoptosis in some circumstances. Although there is accumulating evidence that de-repression of E2F is important in both apoptosis and cell-cycle regulation^{43,44,45,46,47,48,49}, it is unclear which aspects of RB's biological functions are achieved through its effects on E2F-mediated transcription.

Indeed, RB potentially interacts with more than 100 different cellular proteins⁵⁰, and the functional relevance of most of these interactions has yet to be established. A further complication is that RB is one member of a family of related proteins that are collectively called the 'pocket proteins' (Fig. 3) and that seem to have both redundant and opposing functions, depending on the context (for review, see Refs 51,52). Finally, the fact that most tumour-associated RB mutations lead to the production of a truncated or unstable protein has made it difficult to map precisely the domains of RB that are most important for its tumour-suppressive function. In short, there are several significant obstacles that have been encountered in the pursuit of a clear understanding of the cellular function of RB.

Figure 3: The 'pocket protein' family.

a | RB, p107 and p130 make up the 'pocket' family of proteins. The largest region of homology between these proteins lies in a 'pocket' domain (regions A and B, separated by a spacer), which is required for their interaction with E2Fs and many other factors. Following the publication of the crystal structure of an RB molecule bound to a peptide from the human papillomavirus (HPV) E7 protein, a more defined mapping of specific interactions have been initiated (again mainly in the RB molecule)^{124,125,126,127,128,129,130,131} to determine the binding sites of histone deacetylase (HDAC), as well as the viral oncoproteins adenovirus E1A (E1A), SV40 T antigen (T Ag) and HPV E7 protein (E7). Regions within the pocket, as well as a domain in the carboxyl terminus of RB, have been shown to be important for E2F binding, but the specific contacts made between E2F and RB have not yet been mapped in detail. Furthermore, it is still unclear whether the requirements are the same for all the E2F-family members that bind RB (p107 and p130). Arrows represent the many regulatory phosphorylation sites that have been mapped in RB, and two acetylation sites (stars) have been mapped in the carboxyl terminus of the protein¹³². The p107 and p130 pocket domains are also shown (regions A and B, which are separated by a longer spacer than in Rb). The cyclin-binding site maps to the spacer that is unique to p107 and p130, as does a conserved domain that binds to viral oncoproteins (E1A, E7 and T Ag). Other protein interactions are not yet well defined. b | The pocket proteins are differentially expressed throughout the cell cycle. Whereas RB is expressed steadily throughout the cell cycle (blue), p130 (orange) expression is high in arrested cells, whereas p107 (red) expression peaks during the S phase of the cell cycle.

The pocket family of proteins

The RB protein is a member of a family of three closely related mammalian proteins that includes p107 and p130 (Fig. 3a). Together, these proteins are referred to as the 'pocket proteins' because their main sequence similarity resides in a domain (the pocket domain) that mediates interactions with the viral oncoproteins. Overexpression experiments have indicated functional similarities between these proteins in the regulation of the cell cycle. For example, when overexpressed, they can all cause arrest of some cell types in the G1 phase of the cell cycle, and they can all interact with and repress E2F-mediated gene transcription. In addition, they are all phosphorylated by CDKs (for review, see Refs 51,52).

Despite these similarities, there are significant differences between these proteins. For example, RB interacts predominantly with E2Fs 1, 2, 3 and 4, whereas p107 and p130 primarily interact with E2F4 and E2F5 (see Fig. 2b and, for review, see Refs 22,52). These associations also occur at distinct times during the cell cycle, in part due to expression patterns (Fig. 3b). Whereas RB interacts with E2Fs in both quiescent and cycling cells, p130 interacts with E2Fs primarily in quiescent cells²². However, p107 is predominantly associated with E2Fs in the S phase of the cell cycle, a phenomenon that is not well understood, particularly since overexpression of p107 results in G1 arrest in some cells^53,54. In addition, p107 and p130 can recruit CDK2-containing complexes to promoters, and some experiments have indicated that p107 and p130 can act as inhibitors of CDK function^55,56. Although p107 and p130 contain unique binding sites for CDK2-binding cyclins (Fig. 3), it is unclear which of the many proteins that bind to RB also interact with p107 and p130. It is also unclear whether there are other proteins that associate with p107 and p130 that do not bind to RB⁵².

Although numerous studies of the RB family of proteins have led to an extensive understanding of the biochemical properties of these proteins, the ability to disrupt specific genes in mice has allowed researchers to study the role of this family of proteins in embryonic development as well as during tumorigenesis. This approach has also made feasible the establishment of loss-of-function tissue-culture models. Such reagents are potentially important as it is possible that subtle determinants of specificity are lost when these proteins are overexpressed.

Phenotypes of Rb -null mice

Although initial analysis indicated that Rb is ubiquitously expressed in adult mouse tissue, levels of Rb mRNA are temporally and spatially controlled during embryogenesis. Rb is highly expressed in a few tissues, including the nervous system, blood cells, skeletal muscle and lens⁵⁷. Rb^−/− mice die between days 13 and 15 of gestation, with pronounced defects in erythroid, neuronal and lens development, as well as skeletal muscle defects^{58,59,60,61,62} (Table 1). These abnormalities are associated with a partial failure of differentiation, increased apoptosis, as well as ECTOPIC CELL CYCLES. In addition to the cell-cycle changes and insensitivity to several G1-arrest signals that have been seen in Rb-deficient cells^{14,15,16,63,64,65}, cells established from Rb-deficient mouse embryos are defective in myogenesis as well as adipogenesis^{66,67,68,69,70,71}. Together, these findings highlight the multitude of functions of RB in vivo.

Table 1 Tumour phenotypes in mice

Although it seems that changes in E2f-regulated gene expression are responsible for the shortening of the G1 phase of the cell cycle in Rb-null cells, the challenge now is to understand the mechanisms that underlie the various phenotypes seen in Rb-deficient animals and cells. Overall, these phenotypes could potentially reflect cell changes that are mediated by the deregulation of E2f-target genes, but might also be the result of changes in additional transcriptional programmes that are involved in cellular processes, such as differentiation or apoptosis. In fact, it has been suggested that Rb can indirectly or directly potentiate the activity of transcription factors that are involved in differentiation^50,70,72,73.

Several lines of evidence indicate that regulation of E2F by RB is involved in apoptosis⁴². Ectopic expression of E2f1, E2f2 and E2f3 has been shown to induce apotosis in cultured cells as well as in transgenic mice^{42,43,44,45,48,49,74,75}. Furthermore, mutant mouse embryos that lack Rb, and either E2f1 or E2f3, show a significant reduction in the levels of apoptosis, as well as the number of ECTOPIC S-PHASE CELLS relative to those seen in mice lacking only Rb^76,77. Furthermore, the disruption of apoptotic regulatory genes that encode proteins such as Apaf1 and caspase-3, which have been shown to be activated by E2f overexpression, can rescue some of the developmental phenotypes seen in Rb-deficient embryos^78,79. However, it is important to note that the loss of these individual factors does not allow Rb-deficient embryos to survive until birth. Interestingly, targeted disruption of a gene encoding another Rb-binding protein, Id2 (Ref. 80), in combination with disruption of Rb, allows mice to survive until birth⁸¹. These mice, however, show a severe reduction in muscle tissue and die shortly after birth, indicating that loss of Id2 cannot rescue the muscle differentiation defect in Rb^−/− mice.

It is also interesting that CHIMERIC Rb animals in which Rb-deficient cells contribute substantially to all tissues are viable and show very mild phenotypes relative to those seen in Rb-deficient mice^{82,83,84,85,86}. This result indicates that Rb might have an additional cell non-autonomous role in development. Experiments with conditional knockout mice, however, have shown that depletion of Rb in the TELENCEPHALON results in neurons that are able to survive and differentiate but exhibit enhanced neurogenesis, due to an increase in cell division⁸⁷. Taken together, the analysis of Rb-knockout mice has highlighted the complex nature of Rb's normal function during embryonic development.

Phenotypes of p170 - or p130 -null mice

In contrast to the mice that are deficient in Rb, mice deficient in either p107 or 103 develop normally and exhibit no obvious adult phenotypes^88,89. However, differences in the genetic background of mice have recently been shown to be important determinants of the developmental consequences of the genetic loss of p107 and p130. Mice with disruptions in p107 and p130 in the C57/BL6-129 genetic background show defective chondrocyte development and die, shortly after birth, of unknown causes. This reveals an apparent functional redundancy between these proteins in some developmental processes⁸⁸. Mice with disruptions in p107 or p130 in a BALB/c background have more severe phenotypes^90,91. Cells isolated from p107- or p130-null mouse embryos have a shortened G1 phase of the cell cycle, similar to that of Rb-deficient cells^15,16. However, the genes that are deregulated as a result of p107 or p130 loss are different from those that are deregulated in Rb-deficient cells^15,92,93,94. These results indicate that although Rb-family members have distinct roles in the regulation of E2f-mediated gene expression, loss of their function seems to have a similar consequence in the context of cell-cycle progression.

Moreover, there are additional phenotypic similarities between Rb and p107- or p130-deficient cells, such as the inability of these cells to undergo arrest in response to p16 overexpression⁹⁵. The ability of cells that lack the various pocket proteins to respond to DNA damage is more controversial. Several studies showed that Rb-deficient cells failed to undergo cell-cycle arrest following DNA damage^64,65. However, in some studies, cells lacking both p107 and p130 showed a similar failure⁶⁵, whereas at least one study reported that such cells responded normally to DNA damage⁶⁴. It is therefore possible that subtle differences in the handling of these cells can substantially impact on their properties in culture.

As described above, cells that lack the various Rb-family proteins have several similar properties. It has become clear, however, that the presence or absence of particular members of the pocket-protein family can markedly affect these properties. For example, in an assay of adipogenic differentiation, RB^−/− 3T3 fibroblasts failed to respond to a cocktail of adipogenesis-inducing agents, whereas p107^−/−/p130^−/− cells differentiated at a much higher rate than wild-type cells⁶⁶. This difference potentially reflects the differential ability of these proteins to interact with the various E2fs. In fact, consistent with the known specificity of E2f interactions with the various pocket proteins, it was recently reported that E2f1-deficient cells showed a reduced potential for adipogenic differentiation, whereas cells lacking E2f4 showed increased adipogenesis⁹⁶. However, the possibility that Rb and p107/p130 additionally regulate other transcription factors that are involved in the differentiation programme cannot be ruled out.

The phenotypes of Rb^−/−/p107^−/− and Rb^−/−/p130^−/− double-mutant embryos further support the notion that Rb-family members show partial redundancy during development⁴¹. So, embryos of both of these genotypes have phenotypes that are similar to Rb^−/− embryos; however, these embryos die 2 days earlier. Similarly, chimeric animals that are generated from Rb/p107-deficient embryonal stem (ES) cells exhibit a very low contribution of mutant cells to the developing embryo, whereas ES cells that lack only Rb can contribute approximately 50% to the development of viable embryos^83,84,97. Whereas embryos that lack all three pocket proteins have not been characterized, embryo-derived fibroblasts lacking all three proteins have recently been reported^98,99. Such cells were found to be unresponsive to many G1-arrest signals (such as growth-factor withdrawal and DNA damage), are impaired in their ability to arrest at confluence and do not respond to senescence-inducing signals. As these defects are substantially more severe than those seen in cells harbouring any of the double-mutant combinations, it seems that, at least in some cellular contexts, all three of these proteins can function redundantly.

Based on the results obtained from the characterization of developmental defects and the analysis of embryo-derived cells from these various knockout mice, it is clear that Rb performs various in vivo functions related to cell proliferation, apoptosis and differentiation. Moreover, it has also become clear from this analysis that the additional Rb-family members, p107 and p130, perform some overlapping functions as well as some seemingly opposing functions. It might be the differences among the biological properties of the pocket proteins that account for the fact that Rb, but not p107 and p130, is selectively lost in tumours.

Tumour phenotypes in mice

Although the predominant tumour type that is seen in human patients harbouring a germline mutation in the RB gene is retinoblastoma, targeted disruption of one allele of Rb in mice does not lead to this cancer. In addition, studies of conditional knockout mice have shown that loss of Rb in mouse photoreceptor cells does not result in retinoblastoma or any other phenotypic changes in these cells¹⁰⁰. Instead, mice that are heterozygous for Rb show an increased predisposition to pituitary and thyroid tumours, which are associated with a loss of heterozygosity at the Rb locus^62,101. Significantly, when Trp53 (which encodes p53 in mice) is deleted from Rb-heterozygous mice, the animals develop retinal dysplasia^61,102. This indicates that loss of p53 could be a rate-limiting event in the development of this tumour in mice, but not in humans. As mentioned previously, loss of Rb in mouse development is associated with increased apoptosis in some tissues, supporting the notion that for loss of Rb to contribute to tumorigenesis, a block to apoptosis must arise. This is consistent with earlier observations that mice that express a T-antigen transgene in the retina develop retinoblastoma^103,104. T antigen inactivates all three pocket proteins, as well as p53. Similarly, combined transgenic expression of the papillomavirus E6 and E7 proteins, which inactivate p53 and the pocket proteins, respectively, also promotes retinoblastoma formation in mice¹⁰⁵.

Establishing the role of the pocket proteins in the development of retinoblastoma in transgenic and knockout mice has been further complicated by studies in which combinations of pocket proteins have been disrupted. Although mice with disruptions in all three pocket proteins have not been analysed in this context, chimaeras that contain cells with disruptions in both copies of the Rb and p107 genes develop retinoblastoma⁹⁷. Notably, these tumours seem to develop independently of their p53 status. Loss of p130, however, does not seem to impact on the development of retinoblastoma in an Rb-heterozygous background. Moreover, the loss of p107 and/or p130 does not lead to the development of retinal dysplasia or to the development of any tumour type⁵¹.

Collectively, these studies have shown that the development of retinoblastoma in mice might require distinct genetic changes in addition to loss of the Rb gene, and furthermore, that overcoming the tendency to undergo apoptosis is essential to the development of retinoblastoma in mice. Moreover, it is clear from observations in humans, as well as those from animal studies, that loss of RB contributes to tumorigenesis in a relatively small subset of tissue types, potentially reflecting partially redundant functions among the various pocket proteins in other tissues. Alternatively, it is possible that loss of RB is actually detrimental to the development of many tumour types, and that RB expression contributes to tumorigenesis. For example, removal of E2f1 from the mouse germline results in tumour formation in ageing mice, indicating that E2f1-induced apoptosis (repressed by Rb) protects against tumour development^106,107. Furthermore, studies have revealed that RB levels are substantially elevated in some cancers, such as colorectal tumours¹⁰⁸. Significantly, reduction of RB expression in those cells leads to apoptosis and reduced cell proliferation. High levels of RB might contribute to tumour development by increasing E2f-mediated transcriptional repression, as overexpression of E2f1 has also been shown to result in apoptosis in these same colon cancer cell lines¹⁰⁹. Furthermore, studies in transgenic mice have indicated that E2f1 overexpression, which could be analogous to loss of Rb, can, in fact, suppress the transforming ability of a constitutively active Ras allele in some tissues but not in others^110,111. Along these same lines, mice that express a 'hyperactive' Rb molecule (a mutated form of Rb that cannot be inactivated by phosphorylation) from the mouse mammary tumour virus (MMTV) promoter show reduced proliferation at the end buds, as well as precocious cellular differentiation and extended survival of the mammary epithelium¹¹². Some of these mice developed hyperplastic nodules as well as mammary adenocarcinoma, indicating that the extended survival/anti-apoptotic effect of hyperactive Rb could promote tumour progression in this setting.

Conclusion and perspectives

The importance of understanding the functional details of the RB regulatory pathway is underscored by the fact that many human tumours have deregulated some of its components. As described above, accumulating evidence from cell-culture experiments, as well as animal model studies, implicates RB and the related p107 and p130 in several cellular processes, thereby complicating the establishment of a precise role for RB as a tumour suppressor. Moreover, it is clear that these proteins perform context-specific overlapping and even opposing functions. This could account for RB's apparently unique role among the pocket proteins as a tumour suppressor, and for its selective loss in a relatively small subset of human tissues. Future studies will certainly reveal new insights into the interplay between the various RB-family members, and about the nature of downstream targets and signalling pathways that impinge on RB function. Importantly, the fruitfly Drosophila melanogaster and nematode worm Caenorhabditis elegans contain conserved pathways with fewer redundancies at each level. Specific components of these pathways have been shown to regulate cell growth, cell death, cell-fate determination and cell-cycle control^{113,114,115,116,117,118,119,120,121,122,123,124}. Genetic studies in these model systems are likely to reveal in vivo partners, upstream regulators and downstream genes of RB in each of these processes.

Finally, with regard to targeting the RB pathway for the treatment of cancer, we might have to revise and re-examine current models in light of recent 'counterintuitive' results, which indicated that loss of RB might, in fact, be detrimental to the development of tumours in some cell types. So, it is possible that targeting RB itself might be an effective antitumour strategy in some contexts.