Patterns in the tapestry of chromatin-bound RB

Abstract

The Retinoblastoma protein (RB)-mediated regulation of E2F is a component of a highly conserved cell cycle machine. However, RB’s tumor suppressor activity, like RB’s requirement in animal development, is tissue-specific, context-specific, and sometimes appears uncoupled from cell proliferation. Detailed new information about RB’s genomic distribution provides a new perspective on the complexity of RB function, suggesting that some of its functional specificity results from context-specific RB association with chromatin. Here we summarize recent evidence showing that RB targets different types of chromatin regulatory elements at different cell cycle stages. RB controls traditional RB/E2F targets prior to S-phase but, when cells proliferate, RB re-distributes to cell type-specific chromatin loci. We discuss the broad implications of the new data for RB research.

The RB conundrum

The RB tumor suppressor has been studied intensively in the 37 years since it was first cloned [1]. While much is known about RB, there is a contradiction at the heart of the RB literature.

On one hand, RB seems to be an important cog in the cell cycle machine. RB suppresses E2F-dependent transcription, allowing the expression of a suite of genes needed for cell proliferation to be controlled by the periodic activation of Cyclin-Dependent Kinases (CDKs). RB is broadly expressed and its roles in E2F-regulation and cell cycle control have been conserved during evolution. CDK deregulation occurs frequently in human tumors and increased CDK activity prevents RB-mediated cell cycle arrest. The idea that RB is a tumor suppressor because it blocks E2F-driven transcription/proliferation makes a great deal of sense because ectopic expression of activator E2Fs is sufficient to drive quiescent cells into S-phase [2], and many E2F-regulated genes are essential for cell proliferation. It has been said that if one considers all of the different ways that RB can be inactivated then RB may be functionally compromised in most tumor cells [3,4].

On the other hand, the consequences of RB loss are context-specific. RB1 was initially cloned because it is mutated in retinoblastomas, a rare childhood cancer. Inheritance of a mutant RB1 allele predisposes individuals to specific types of cancer (notably retinoblastoma and osteosarcoma) but not to the most common types of human cancer. Mutation of the mouse Rb ortholog also causes cancer but gives a tumor spectrum that is completely different from that seen in humans [5]. Studies of mosaic mice show that Rb mutation does not cause ectopic proliferation in most differentiated or quiescent cells, and developmental defects are rescued in chimeric embryos [6,7]. Furthermore, studies of cells from the human eye indicate that retinoblastomas develop after RB1 loss in a specific cell population [8–10]. While RB1 is mutated in an assortment of human cancers, the frequency of these mutations is uneven. In retinoblastoma, osteosarcoma, and small cell lung cancer (SCLC), RB1 mutation occurs in most cases and is present in most tumor cells, consistent with it being an early event. In other types of cancer, RB1 mutations are not as frequent. In certain tumor types, such as castrate-resistant prostate cancers, RB1 mutation is typically a late event [11,12]. Here, RB loss is associated with tumor progression. RB loss is also associated with the conversion of the non-small cell lung cancer to SCLC [13]. Studies of Kras- and p53-mutant mouse lung adenocarcinomas show that Rb ablation increases the malignancy of the tumor but does not significantly increase cell proliferation [14], instead this progression is associated with reduced lineage fidelity.

If RB is simply a cog in a universal cell cycle machine, then why are the effects of RB loss so context-dependent, and why is RB’s tumor suppressor activity so variable? In this article, we focus on the recently identified classes of RB-binding sites within human chromatin. As described below, the recent discovery that RB targets different types of regulatory elements at different stages of the cell cycle introduces an appealing new model for RB’s mechanism of action: RB controls the RB/E2F targets before the G1/S-checkpoint and switches to different, cell type-specific programs of gene expression once cells start to replicate. Cell type-specific differences in the genomic distribution of RB likely have functional consequences. We hypothesize that context-specific binding sites allow RB to play context-specific roles. If so, it will be necessary to know precisely which loci are bound by RB in the relevant cell types, and to understand how RB acts at these categories of binding sites, to be able to explain the context-specific functions of RB.

Components that shape the context-specific requirement for RB.

Three general concepts have been used to explain why RB has context-specific functions. These ideas, and the underlying literature, have been described thoroughly [15,16], and are only briefly summarized here.

One idea is that context-specificity stems from the fact that RB has partly overlapping roles with other negative regulators of the cell cycle, such as the CDK inhibitors p21 or p27, or the RB-related proteins p107 and p130. Of these, the functional overlap between RB and p107 or p130, has been documented most extensively: the proteins have similar structures, interact with similar types of proteins, regulate transcription, and target to many of the same genes [17,18]. RB’s overlap with p107 and p130 underscores the point that CDKs have important substrates aside from RB [19,20]. Indeed, compound mutations result in more severe phenotypes [21–25], and enhance the resistance to CDK4/6 inhibition [26]. Why the functional overlap between RB and p107/p130 varies between cell types is not known. Nevertheless, tumors that acquire lesions that increase CDK activity may do so because, in those cells, the inactivation of RB alone is insufficient to drive proliferation. Thus, one could imagine that RB performs essentially the same general role in most cell types, that it is inactivated in most tumor cells, and that the context-specific consequences of RB mutation stem from mysterious (largely unknown) determinants of functional overlap/functional compensation between RB, its homologs, and other cell cycle regulators.

A second idea is that the consequences of RB loss vary because RB acts differently in different cells. To some degree this must be true, as RB associates with many cellular proteins in addition to E2F and several of these have cell type-specific expression patterns and/or cell type-specific functions [16]. Examples of these include (but are not limited to) the lymphoid-specific transcription factor Elf1 [27], CBFA1 and Runx2 in osteoblasts [28,29], Pu1 in erythroleukemia cells [30], and KDM5A during myoblast differentiation [31]. Hence, RB’s targets do vary. The key unknowns are: how great is this variation, and how significant are the differences? Context-specific differences in RB’s interactors have certainly been invoked to explain cell type-specific consequences of RB inactivation [15,16].

A third idea, is the complication that RB may possess multiple tumor-suppressive activities, with the importance of individual roles varying between contexts. Potentially, RB may prevent tumorigenesis in some situations by stopping cell cycle progression. However, in other instances, RB may promote differentiation , or it may trigger senescence, or promote lineage fidelity, or help to maintain genome integrity, and so forth. In the right circumstances, several of RB’s outputs may be a rate-limiting barrier to tumorigenesis.

These ideas are not mutually exclusive and each likely provides part of the solution. For example, the functional overlap between pRB and p107/p130 may vary depending on their portfolios of associated proteins and depending on the precise mechanism of tumor suppression. Answering these questions is challenging because several of most fundamental questions about RB have not yet been answered: exactly which proteins/genes does RB target, how much cell-type variability is there in the proteins that RB associates with and the genes that it regulates? How is the pool of RB divided among its potential targets?

A major obstacle to answering these questions is that, for many years, there has been limited information about the genomic loci bound by RB. Obviously, if it is not certain where RB acts, then explaining how RB functions is problematic. This situation arose because RB ChIP-seq (see Glossary) is technically difficult, with most experiments giving a low signal-to-noise ratio. Consequently, models for RB’s mechanism of action have been constructed from fragments of information. Over time, as RB was proposed to play multiple roles and was linked to increasingly large numbers of proteins, it has become increasingly difficult to describe an overarching framework that accommodates multiple ideas [15].

Recent advances in ChIP-seq methods are starting to give detailed information about the genomic sites bound by RB (Table 1). This data offers a more complete picture of RB in action and there have been some surprises. Below, we summarize this data and describe some of the implications. One of the mysteries in the RB literature is why RB has context-specific effects, and we highlight features in the ChIP-seq data that may explain how RB bridges the gap between being a core cell cycle regulator and performing cell type-specific functions.

Table 1.

RB ChIP-seq studies

Target	Cell type / Species	growth condition	Reference:
endogenous RB	IMR-90 primary human fetal lung fibroblasts	normal growth, quiescence and senescence	[38]*
endogenous RB	IMR-90 primary human fetal lung fibroblasts	oncogenic transformation by adenovirus small e1a	[39]
endogenous RB	mouse embryonic fibroblasts	non-stress	[52]
endogenous RB	mouse embryonic fibroblasts	G1-arrested and proliferating cells	[37]**
endogenous RB	human GM12878	non-stress	ENCODE project (ENCSR785OKZ) GEO: GSE106046
endogenous RB	human K562	non-stress	ENCODE project (ENCSR670JDQ) GEO: GSE105638
exogenous FLAG-RB wild-type FLAG-RBΔCDK	human h-TERT immortalized RPE1 cells human h-TERT immortalized BJ fibroblasts human ER-positive breast cancer T47D human mammary epithelial MCF10A	normal growth, cells synchronized in G1 or S-phase, G1-arrested cells	[40]
endogenous RB	human h-TERT immortalized epithelial RPE1	non-stress	[40]

^*low signal-to-background ratio in this study impeded identifying many RB-bound loci in non-promoter regions.

^**most of the RB “peaks’ in this data are PCR duplicates.

What ChIP-seq experiments tell us about the distribution of RB.

RB is primarily nuclear and a chromatin-associated protein. Here are a few key facts about the RB-binding sites in human chromatin.

First, RB associates with many thousands of loci; far more than the number of genes typically listed in RB-loss or E2F-activation transcriptional signatures. Such a difference is not uncommon in chromatin studies, but it is unknown what distinguishes binding sites where RB is rate-limiting for gene expression from those where it isn’t. One potential interpretation is that many RB-binding sites are not critical for gene regulation. Another possibility is that RB-binding sites may have additional roles that extend beyond transcription, such as the RB-dependent changes in chromatin organization [32]. An alternative scenario is that the large number of binding sites illustrates the potential reach of RB. In this interpretation, many RB-binding sites may be functionally important when examined in the right context, and the challenge is to identify the appropriate situation. Context-specific synergies may explain why the effects of RB on the transcription of its targets can vary significantly between cell types, growth conditions or when individual phosphorylation isoforms of RB are expressed [33–36].

Second, the strongest RB-binding sites tend to be located in promoter regions. Although RB is reported to interact with repeat sequences [37], most of the mapped RB-binding sites are in euchromatin.

Third, the distribution of RB-binding sites is not fixed but it varies. Variation is seen between cell lines but is also evident within a single cell type. ChIP-seq experiments in human lung fibroblast cells IMR-90 found that the genomic loci bound by RB differed, depending on whether cells were growing, quiescent, or senescent [38]. In senescent cells, this redistribution was accompanied by a reduced ability of p130 to compensate for the removal of RB, increasing the functional significance of RB [38]. However, whether alterations in p107 or p130 expression influence RB association with chromatin is unknown. Other studies have shown that RB inactivation by the Adenovirus E1A protein redistributes RB to p300-targeted genes involved in TGFβ-, TNF-, and interleukin-signaling [39]. Recently, Sanidas et al., showed that cell cycle progression is accompanied by the redistribution of RB toward enhancer regions enriched in the Activator Protein-1 complex (AP-1) [40].

Fourth, there are multiple categories of RB-binding sites. The most recent data shows that RB-binding sites are located in promoters, enhancers, and CTCF-bound loci, and sites in these locations have distinct features (Table 2). Not only do RB-binding sites reside in different types of chromatin, but they can be subdivided by the largely mutually exclusive presence of sequence-specific DNA-binding proteins: E2F1 marks most RB-binding sites in promoters, cJun marks many RB-bound enhancers, a third set of RB peaks are found at CTCF sites. Since RB lacks any sequence-specific DNA binding activity of its own, these enriched transcription factors (TFs) are likely important for the recruitment of RB. Notably, these RB-binding sites differ in their cell cycle dynamics. In arrested cells, RB accumulates on promoters. In cycling cells, or when arrested cells are stimulated to enter the cell cycle and progress into S-phase, RB redistributes away from promoters and towards enhancers. RB ChIP peaks in promoters are generally stronger than the binding sites in non-promoter regions, explaining perhaps why RB’s affinity for chromatin appears to drop as cells enter S-phase (BOX1 and Figure I). For years, RB has been portrayed as a protein that oscillates on and off chromatin in response to CDK phosphorylation. This new data suggests that a more accurate description is that RB is a chromatin-associated protein that is redistributed by cell cycle transitions.

Table 2.

Categories of RB-binding sites^*

Location	Active Promoters	Active Enhancers	CTCF sites:
Chromatin marks	H3K4me3 and H3K27ac	H3K4me1 and H3K27ac	CTCF peaks
Number of peaks	10,024	15,646	2445
Enriched motifs	E2F ETS SP1 NFY	AP-1 (Fos/Jun) Bach2 NF-E2	CTCF BORIS
Co-localizing TFs shown Predicted by PanChIP software [63]	E2F NELFA, GTF2B, NELFE, RPE, XRN2, E2F1, PHF8, E2F4, POLR2B.	c-Jun FOSL2, JUN, E2F7 FOS, FOSL1, JUND, TEAD1, TEAD4, CEBPB, DAAX.	CTCF SUMO1, ZNF654, KDM3B, JMJD6, SMC3, CTCF, BRD9, SMC1A, ZBTB2
Relative peak strength	Broad range from strong to weak	Broad range from moderate to weak	Broad range from strong to weak
Cell cycle oscillation (see also BOX1)	Strongest binding in G1 arrested cells	Strongest binding in S/G2	Little change
functions enriched in proximal genes [40]	Cell cycle DNA replication Metabolism RNA processing Ribosome biogenesis	Focal adhesion Adherens junctions Signal transduction Axon guidance	Neurogenesis Neuron differentiation
Conservation between RPE1 and BJ fibroblasts [40]	70%	10%	30%

^*RB ChIP-seq data from RPE1 cells [40].

BOX 1. Cell cycle progression changes RB’s affinity for chromatin.

The overall affinity of RB for chromatin drops as cells progress through G1/S. This change was first visualized in experiments that used RB-specific antibodies for immunofluorescence (IF) [64,65]. When cells with wild-type RB were arrested in G1, the RB IF signal was found to be resistant to extraction with low-salt buffers. But when cells progressed through G1/S, the low-salt conditions washed away the RB IF signal. This observation agreed nicely with the discoveries that G1 CDKs phosphorylate RB and reduce RB/E2F binding. As a result, RB/E2F regulation is often depicted with figures showing RB being released from DNA at G1/S.

These diagrams overlook the fact that the RB IF signal remains nuclear throughout the cell cycle. The RB IF pattern does change during mitosis, and some studies have detected a fraction of RB in the cytoplasm [66–70], but in most cells, most RB protein remains in the nucleus; so, where is it, and what is it doing? ChIP-seq experiments shed new light on this puzzle. By comparing the chromatin distribution of RB in cells arrested by constitutively unphosphorylated RB (RBΔCDK) with RB’s chromatin distribution in cycling cells, it is possible to quantify the change in RB signal at each site. The results show, as expected, that in arrested cells RB binds preferentially to sites in promoters that contain E2F motifs. However, other RB peaks display a reciprocal pattern and give stronger signals in cycling cells or in cells released from G1-arrest. Such peaks are most commonly located in non-promoter regions and are generally weaker than those in E2F-regulated promoters. If we accept the argument that ChIP-signals provide a crude measure of relative affinity, then this suggests that RB moves from high-affinity sites in G1 to lower-affinity sites in S/G2, a change consistent with the RB IF studies.

Figure I:

RB-binding data from RPE1 cells [40] are presented in contrast groups of RB-binding sites classified by the TF-binding motifs that they contain. 1xE2F and 2xE2F indicate RB peaks that contain one or two E2F-binding motifs, while 1xAP-1 and 2xAP-1 indicate peaks with one or two AP-1 motifs, respectively. RB peaks containing CTCF, ETS, SP1, or other TF motifs are also shown. Peaks with more complex combinations of motifs were excluded. For each group, the RBΔCDK ChIP-seq signal is displayed as a percent change relative to the wild-type RB ChIP-seq signal. The scale ranges from an increase of 40% (dark red) to a decrease of 40% (dark blue). The figure illustrates that the cell cycle-dependent change in RB binding varies depending on the TFs that likely recruit RB to DNA. Note that RB-binding sites with 2 E2F motifs have the strongest binding in arrested cells, while sites with 2 AP-1 motifs show the strongest binding in cycling cells. Motifs bound by SP1 or ETS proteins are enriched in RB peaks located in promoters, but peaks with ETS motifs behave differently from those with E2F motifs. Thus, cell cycle-dependent binding, a feature commonly regarded as a hallmark of RB-mediated regulation, actually varies in degree and timing from site to site.

Re-distribution likely changes the role of RB.

The binding of “active” (i.e. unphosphorylated or mono-phosphorylated) RB to promoters is associated with decreased histone acetylation at H3K27, a mark of transcriptionally active chromatin. As RB is known to recruit histone deacetylases to promoters, this link is consistent with RB-mediated repression of transcription. In contrast, in cycling cells with hyper-phosphorylated (“inactive”) RB, RB binding to enhancers is associated with increased H3K27ac, suggesting different models [40]. Currently, nothing is known about the reason for RB’s presence at CTCF-binding sites, or the consequences. RB has been linked to complexes that organize chromosomal regions; RB physically interacts with the Condensin II complex to regulate chromosome condensation [41–44], and the loss of RB has implications for chromatin cohesion [45]. Furthermore, the long-term expression of active RB promotes structural changes that are visible under the microscope [32]. However, as of now, it is unknown whether any of these involve the effects of RB at CTCF-dependent boundaries.

Importantly, RB-binding sites in promoters, enhancers, and CTCF-bound loci are associated with genes that function in different cellular processes. RB-bound promoters include the classic E2F targets; genes that function in cell cycle and DNA replication, and metabolic genes, genes involved in oxidative phosphorylation and mitochondrial function. RB-bound enhancers are linked with genes that function in adhesion and cell signaling. Genes proximal to RB-bound CTCF sites are enriched for functions in neurogenesis. Interestingly, all of these cellular processes are known to be altered in Rb-knockout mice [46–51].

RB binding to promoters and enhancers impacts gene expression. But, just as RB is rate-limiting for transcription at only a subset of RB-bound promoters, only a subset of genes with RB-bound enhancers showed changes in transcript levels when RB was deleted. While RB mostly represses transcription of E2F-bound promoters, the loss of RB has varied effects on the expression of genes with RB-bound enhancers and different consequences were noted in different cell lines. Genes with RB-bound enhancers that display altered expression when RB is deleted include genes like EGFR and TGFB2 that have critical roles in cell signaling and tumorigenesis [40]. RB binding to an enhancer element for Sox2 has been proposed to restrict pluripotency [52]. Thus, the effects of RB at enhancers are likely to be important, and when RB redistributes from promoters to enhancers it also moves from the regulatory elements of one cohort of genes, to regulatory elements of different sets of genes.

RB has conserved binding sites in promoters and cell type-specific binding sites in enhancers.

Perhaps the most striking difference between RB-binding sites in promoter and non-promoter regions is their degree of conservation between cell lines. Greater than 70% of RB-binding sites in promoters are conserved between RPE1 human retinal pigment epithelial cells and BJ human skin fibroblasts [40]. In contrast, only 11% of RB-binding sites in enhancers are similarly conserved. There are thousands of RB peaks at enhancers in both RPE1 and BJ cells, and these peaks show a similar enrichment for AP-1 motifs and cJun-binding sites, but the RB-binding sites are located in different sets of enhancers.

A wider analysis, aggregating RB-binding site data from 7 cell lines, underscores this point and highlights additional features ([40], BOX2 and Figure II). The number, and positions, of RB-binding sites in enhancers vary tremendously. Of the cell lines examined to date, the largest numbers of RB enhancer peaks are seen in non-transformed cells; very curiously, some tumor-derived lines have far fewer. While some RB-binding sites in promoters are private (peaks seen only in one cell line), many others are conserved across cell lines. Cluster analysis suggests that there may be as many as 50 different categories of RB-binding sites in the 7 lines examined [40]. Clusters of private peaks are often enriched for motifs bound by lineage-specific factors , supporting previous studies that RB has tissue-specific interactors [28,29,31]. This is true for RB peaks in promoters as well as peaks in non-promoter regions. Although the connection between RB and CTCF has not yet been studied, RB peaks that are strongly bound by CTCF are among the RB-binding sites that are well preserved between cell lines.

BOX 2. Conservation of RB-binding sites between ChIP-seq datasets.

It is challenging to compare the peaks between RB ChIP-seq profiles because each study uses a different methodology, each dataset has a different signal/noise ratio, and these parameters allow very different numbers of peaks to be called. To get a general picture we took the set of 28,115 RB peaks identified in RPE1 cells [40] and calculated a conservation score, which was the number of cell lines with an RB peak called at the same location. Peaks were called using MACS2. In accordance with ENCODE guidelines, during the alignment stage, we utilized PCR deduplication, mapping quality >= 30, and number of mismatches <=2. The total number of RB peaks with fold change enrichment score >2 relative to the background signal in the other cell lines were BJ: 20526, GM12878: 3201, IMR-90: 548, K562: 16714, MCF10A: 1359, T47D: 4786 [40].

Figure II:

The ternary plot on the left is reproduced from Sanidas et al., 2022, Figure 2B [40] and displays the RB binding sites colored to show RB-peaks in promoters (green), enhancers (orange) and CTCF sites (purple). The axes indicate the percentage of E2F1 signal, c-Jun signal and CTCF ChIP signal detected at each RB-bound locus. The ternary plot on the right displays the conservation scores and illustrates that RB-binding sites in promoters are best conserved between cell lines. As noted previously [40], most binding sites in enhancers and at CTCF sites are cell line-specific.

In this comparison, 3589 peaks were detected in two cell lines, 3971 in three lines, 2279 in four lines, 804 in five lines, 333 in six, and 104 in all seven lines. Gene Ontology analysis of the genes proximal to these binding sites shows that RB peaks that are unique to RPE1 cells or conserved in only one other cell line, are proximal to genes enriched for functions in adhesion, cell morphogenesis, and cell migration. RB peaks conserved among 3 or 4 cell lines are proximal to genes enriched for functions in translation, catabolism, mRNA metabolism, amide biosynthesis, and RNA processing. The genes proximal to RB peaks detected in 5 or more lines give the strongest enrichment ratios, with enrichment in genes that function in DNA replication, DNA repair, cellular response to DNA damage, DNA metabolic processes and mitotic cell cycle. This suggests that these cell cycle functions represent “core” targets of RB, but we note that it remains to be determined whether the peaks with the highest conservation scores are truly the most conserved targets of RB, or whether these are simply robust peaks that are most resilient to the varied experimental conditions used in independent ChIP-seq studies.

This pattern of conservation/variation adds another feature to the tapestry of RB-binding sites: in arrested cells, RB binds to sites in promoters that are generally well-conserved, but when cells proliferate, RB redistributes to non-promoter sites that are largely cell type-specific (BOX2 and Figure II). Some specific roles of RB are likely mediated via cell type-specific interactions at promoters [31], but the genome-wide data suggests that far greater variation stems from RB’s interactions with non-promoter regions.

Concluding remarks.

The fact that the chromatin-bound RB is redistributed by changes in cell cycle position and cell state has important implications for our reading of the RB literature. RB was conventionally viewed as an integral component of a cell cycle machine, with its biological function frequently associated with the regulation of the E2F transcription program that is universally applied. The genomic data shows a kaleidoscope of RB-binding sites rather than a single pattern. RB’s targets change, and it is necessary to study the full spectrum of RB-binding sites on chromatin to understand RB-dependent phenotypes. The RB peaks are not all the same: the regulation and consequences of RB binding vary, depending on the chromatin environment and how RB is brought to DNA. The ChIP-seq data also cautions us not to assume that G1 is the only phase of the cell cycle where RB acts on chromatin.

The stage is now set for a thorough examination of RB-binding patterns in a wide variety of experimental models. This will reveal how many different RB-binding patterns exist, and how extensively RB’s targets change during animal development, cell differentiation and tumorigenesis. Further research is needed to identify the rules that determine the distribution of RB. RB can be recruited to sites of DNA damage [53] and it will be fascinating to learn more about the processes that target RB to other types of loci. Since RB’s distribution varies between cell types, and depends on growth conditions, a variety of context-specific factors are likely involved. CDK-mediated phosphorylation of RB is typically viewed as a way to relieve E2F-mediated repression, but it may also provide a means to re-target RB. Mono-phosphorylation is known to control RB’s interactions with protein partners and to influence the partition of RB between promoters and enhancers [35,40]. Stress-induced phosphorylation of RB at specific residues has been correlated with distinct RB functions triggered by stress signals [54,55]. However, it remains unclear whether these phosphorylation events direct RB to specific chromatin regions to control distinct transcriptional programs. Additional post-translational modifications regulate RB activity, and these may also help to direct RB to specific loci [56–60].

The highly detailed map of RB-bound chromatin loci also underscores the significance of revisiting studies on the RB family members. A crucial objective of these studies should be to investigate the extent to which the various RB-binding sites are also bound by p107 and p130 and to compare the precise location of the respective elements. Additional investigations can answer whether the absence of these complexes influences the chromatin distribution of RB and vice versa, providing valuable mechanistic insights into the functional redundancy and distinctions between the pocket-family proteins.

The identification of transcription factors that are present, in a largely mutually-exclusive pattern, at RB-bound promoters, enhancers, and insulators suggests that RB is brought to chromatin in multiple ways. More information is needed about RB’s mode of action at these different types of binding sites. Studies are also needed to identify and characterize the subsets of RB-binding sites that are critical for RB-mediated control of transcription, epigenetic regulation, or for chromatin organization. RB-binding sites that are rate-limiting for the transcription of E2F-targets represent only a small fraction of chromatin-associated RB. Does RB act differently at these loci, or is it the significance of RB’s action that changes? In order to have a full understanding of RB’s mechanism of action, it is necessary to know where RB binds and how important this association is for transcription regulation and chromatin structure. To date, that has not been done because although the RB field has been proficient at expression profiling, detailed information about the distribution of RB-binding sites has been lacking. RB association with enhancers adds an extra challenge in integrating RB ChIP-seq data with the RB-mediated transcription since enhancers act at a distance and regulate gene expression in complex ways.

Currently, it is uncertain whether RB is functional at all loci. It is possible that some types of RB-binding sites simply provide a storage mechanism that preserves a pool of latent nuclear RB. Finally, we acknowledge the possibility that some RB-binding sites may have no function at all; some peaks may simply mark accessible sites in open chromatin and persist because they have no selective disadvantage. It is important to note that ChIP-seq studies merely provide a list of candidate binding sites. For each peak, further studies are needed to test whether mutation of the enriched motif eliminates RB-binding, and to determine whether the specific elimination of an RB-bound locus alters gene expression or chromatin structure. To date, only a handful of E2F/RB binding sites have been examined in this kind of detail [61]. Initial results confirm that some binding sites are important, but it is clear that the mere presence of a binding site is not a reliable predictor of functional significance.

We have learned so much about RB, yet so many unanswered questions remain. If the past 37 years of RB research are any guide, there will be more surprises ahead. Intriguingly, genetic studies in Drosophila indicate an essential role for RBF (Retinoblastoma-family protein), and DP (E2F Dimerization Partner), in muscle development [62]. In this context, E2F/DP and RBF are needed for tissue-specific activation of gene expression, rather than repression. Similarly, there are examples where RB is reported to increase transcription of certain E2F targets in mammalian cells [33]. So perhaps we should not be completely surprised if we learn of situations in human cells, where everything that we think we know about RB and E2F, is transformed again (see Outstanding Questions).

Outstanding questions:

1. How many different patterns of chromatin-associated RB occur during normal development, do they change during tumorigenesis, and how are changes in RB distribution controlled?

2. Are the molecular functions of RB the same or different at different types of chromatin loci?

3. What distinguishes RB-binding sites that are rate-limiting for gene expression from those that are not?

4. Which biological functions of RB depend on its effects at promoters, and which functions (if any) can be attributed to its effects at non-promoter regions?

5. Which RB-bound chromatin loci are important for RB’s cell type-specific functions?

Highlights:

• RB regulation of E2F is an important component of a highly-conserved cell cycle machine, but the consequences of RB loss are mostly context-specific.

• ChIP-seq studies indicate that RB does not act exclusively at promoters but is also associated with enhancers and chromatin insulators.

• Cell cycle transitions alter the distribution of RB; when cells proliferate, RB redistributes away from promoters and towards enhancers.

• RB-binding sites in promoters are conserved; in contrast, RB-binding sites in non-promoter regions are largely cell type-specific and are mostly independent of E2F.

• RB’s cell type-specific interactions with enhancers and insulators may help to explain some of RB’s context-specific activities and its non-canonical functions.

Glossary

active RB

The unphosphorylated and mono-phosphorylated forms of RB expressed in G1 that suppress E2F transcription

ChIP-seq

Chromatin Immunoprecipitation followed by DNA sequencing

promoter

DNA region upstream of a gene where Transcription Factors and RNA polymerase bind to initiate transcription

enhancer

non-promoter chromatin region that associates with Transcription Factors and regulates gene expression

non-promoter

chromatin region outside gene promoters

H3K27ac

Histone 3 Lysine 27 acetylation (a typical marker of transcriptionally active chromatin)

E2F/DP

E2F and DP Transcription Factors heterodimer

RBΔCDK

RB mutant allele in which all 14 in vivo CDK phosphorylation sites are mutated to alanine