Clefts, grooves, and (small) pockets: The structure of the retinoblastoma tumor suppressor in complex with its cellular target E2F unveiled

Tumor suppressors are functional antagonists of oncogenes and their loss of function importantly contributes to human carcinogenesis. The sequence of the first human tumor suppressor, the retinoblastoma susceptibility gene product pRb, was published in 1986 (1). In this issue of PNAS, Xiao et al. (2) report the crystal structure of a pRb fragment in complex with a peptide derived from the E2F transcription factor, a key cellular target of pRb. An article by Lee et al. (3) recently published elsewhere reports similar findings. Thus, in less than four National Institutes of Health grant funding cycles since its initial discovery, we have arrived at a detailed structural understanding of how pRb inhibits cell growth, how this activity is regulated in normal cells, and how pRb is put out of action in tumor cells.

Loss of pRb function is the rate-limiting step in the development of retinoblastoma, a rare childhood eye tumor. There is compelling evidence that pRb is at the center of a crucial cellular regulatory circuit that is perturbed in most human tumors (reviewed in ref. 4) (Fig. 1). Thus, the pRb pathway offers attractive targets for cancer therapy, and the delineation of the E2F/pRb structure is good news and highly significant.

Figure 1.

The retinoblastoma tumor suppressor pathway contains multiple oncogenes (green) and tumor suppressors (red) and is dysfunctional in almost every human tumor. Even though E2F is an important mediator of the growth suppressor function of pRb, no activating E2F mutants have been detected in human tumors.

One important function of pRb is to inhibit cellular DNA synthesis. In resting cells, pRb is hypophosphorylated and in complex with E2F, a group of heterodimeric transcription factors consisting of an E2F family member (E2F-1 to 6) and a dimerization partner (DP-1 or DP-2) (reviewed in ref. 5). E2F/pRb complexes are transcriptional repressors and prevent cells from undergoing DNA replication (6). In preparation for DNA synthesis, pRb is phosphorylated by cyclin-dependent kinases (cdks) and E2F/pRB complexes break apart. Free E2F acts as a transcriptional activator to drive DNA synthesis (reviewed in ref. 7) (Fig. 1). After cell division, pRb is dephosphorylated, binds up free E2F, and prevents cell cycle reentry. This delicately tuned regulatory circuit is almost universally subverted in tumors. Hyperactivity of cyclin D, cdk4, or cdk6, or inactivation of the cdk inhibitor p16^INK4A each cause constitutive pRb phosphorylation. In cervical cancers that are caused by infections with high-risk human papillomaviruses (HPVs) the E7 oncoprotein binds and degrades pRb and disrupts E2F/pRb complexes (reviewed in ref. 8) (Fig. 1).

The adenovirus E1A protein, simian virus 40 large tumor antigen (SV40 TAg) and HPV E7 oncoproteins can each interact with pRb (9–11) and thwart the pRb regulatory circuit to accomplish expression of host cellular enzymes necessary for viral genome replication. These viral oncoproteins each contain an LXCXE domain, a small block of highly conserved amino acid residues that include the sequence leucine-X-cysteine-X-glutamate (X denotes any amino acid residue), which represents the core pRb-binding site and is necessary for induction of DNA synthesis and cellular transformation (reviewed in ref. 12).

Whereas some cellular pRb interacting proteins also contain LXCXE domains, E2F proteins do not share significant sequence similarity to the viral oncoproteins and contain no LXCXE domains. Nevertheless, the viral oncoproteins and E2F transcription factors interact with similar pRb sequences, a central “pocket domain,” which consists of two densely packed subdomains that are connected by a flexible linker. Two additional cellular proteins p107 and p130 also contain pocket domains. Like pRb, they interact with E2F transcription factors and are targeted by viral oncoproteins. Despite these functional similarities, p107 and p130 are not frequently mutated in human tumors and may not be tumor suppressors. Proteins related to pRb and E2F have been discovered in many multicellular organisms including plants, suggesting that this regulatory principle is highly conserved (reviewed in refs. 13 and 14). Consistent with this notion, certain plant viruses encode replication proteins that interact with pRb (reviewed in ref. 15).

Crystal structures of the pRB pocket domain in complex with LXCXE peptides of HPV E7 and SV40 TAg have been solved (16, 17). The LXCXE motif binds a conserved shallow groove within pRb pocket domain B. The E2F binding site is located on the opposite face of the pRb pocket within the central cleft formed by the interface of domains A and B and is >30 Å apart from the E7 binding site (2, 3) (Fig. 2). Similar to the LXCXE binding site it corresponds to a domain that is highly conserved in p107 and p130 and pRb-related proteins from different organisms.

Figure 2.

Schematic representation of the pRb, E2F, and E7 domains discussed here. E2F interacts through a linear sequence within the activation domain (AD) shown in green with a cleft formed by three non- contiguous regions within domains A and B of the pRb pocket (green). The HPV E7 protein interacts with the LXCXE domain (yellow) with a groove formed by two noncontiguous sequences within domain B of the pocket. The two interaction domains are ≈30 Å apart and on different faces of the pocket. The marked box (M) in E2F and the amino (N)- and carboxyl (C)-terminal domains as well as the flexible spacer (S) between domains A and B of the pRb pocket are also indicated. Areas of pRb, E2F, and E7 shown in gray represent structural terra incognita. See text for details.

As insightful as the E2F/pRb structure is it does not directly explicate a number of interesting questions. How does pRb phosphorylation regulate E2F binding? Are there, as suggested by other studies, additional E2F sequences that contribute to pRb binding? How does binding of E7 disrupt E2F/pRb complexes? The authors have addressed each of these issues by using isothermal titration calorimetry, and these studies offer some additional interesting insights (2).

Most of the cdk phosphorylation sites are within the C-terminal domain and are not included on the crystallized pRb fragment. It has been proposed that initial phosphorylation of the pRb C terminus may disrupt intramolecular interactions between the C terminus and the pocket. This renders the pocket accessible for phosphorylation, which in turn alters the pocket structure and bound E2F is displaced (18). Consistent with this model, phosphorylation of the pRb fragment in vitro by cyclin/cdk complexes reduces E2F peptide binding to background levels (2). The E2F/pRB structure also showed that an important cdk phosphorylation target, serine 567, is localized at the interface of the A and B domain near the E2F binding site. Even though serine 567 does not directly participate in E2F binding, its phosphorylation may disorganize the E2F binding groove at the A/B pocket interface (2). Additional structural studies are warranted, however, to prove this attractive model.

Binding of the E2F peptide does not confer dramatic structural changes to the pRb pocket, and E2F peptides bind a preformed E7 LXCXE peptide/pRb pocket complex and the naked pRb pocket with similar affinity. Interestingly, however, a larger E2F peptide that also includes an additional domain referred to as the “marked box” binds the pRb pocket domain at a substantially higher affinity than the smaller E2F peptide used for crystallization, supporting the notion that the “marked box” also contributes to pRb binding (2). This is reminiscent of studies with E7 that showed that an LXCXE peptide that includes the E7 C terminus binds to pRb with substantially higher affinity than the core LXCXE peptide (19, 20), and that the E7 C terminus is necessary to disrupt the E2F/pRb complex (21, 22). Interestingly, E7 LXCXE peptides that include the C terminus inhibited binding of marked box containing E2F peptides to pRb. These results suggest that the E7 C-terminal domain hinders interaction of E2F marked box sequences with pRb potentially by competition for a binding site within the pocket.

Unlike the pRb C terminus, which contains the bulk of the cdk phosphorylation sites (23) and may play a vital role in regulating pocket function (18) the extensive N-terminal domain does not directly contribute to tumor suppression (24). Nevertheless, delineation of the structures of these domains, ideally in complex with the central pocket, will yield valuable insights into the intramolecular regulatory mechanisms of pRb activity.

Even though E2F scores as a modifier of pRb-mediated carcinogenesis in a mouse model (25), no oncogenic E2F mutants have been detected in human cancers. This suggests that additional cellular targets of pRb that control cellular processes other than proliferation may also contribute to pRb's tumor suppressor function. It was noted early on that retinoblastoma survivors frequently develop osteosarcomas (26, 27). In addition, pRb mutations are restricted to a specific set of human tumors (28). Hence, in addition to generic dysregulation of cell cycle control, pRb likely possesses cell type-specific tumor suppressor activities that are communicated through a distinct set of cellular targets. Reexpression of pRb in RB-deficient osteosarcoma cell lines causes cell cycle arrest, induction of a senescence-like phenotype (29, 30) and expression of bone differentiation markers (31). Senescence and differentiation induction appears E2F independent and is an activity not shared by p107 and p130 (31). Interestingly pRb binds to and acts as a transcriptional coactivator of CBFA1/Runx2, a transcription factor that is required for bone differentiation (32).

In addition to controlling proliferation and differentiation, pRb may regulate apoptosis, proliferative lifespan and genomic stability and some of these activities may be E2F independent. A dazzling number of pRb-interacting cellular proteins have been reported but the functional relevance of most of these interactions remains largely obscure (reviewed in ref. 33). With the structural information of the E2F/pRb complex in hand it is now possible to generate and analyze the biological activities of E2F-binding deficient pRb mutants. Structure-based studies with LXCXE binding deficient pRb mutants have already revealed that interaction with LXCXE containing cellular proteins is dispensable for pRb growth suppression (34). Moreover, mutation of some conserved basic pRb residues that turned out to be E2F contact sites retain some growth arrest activity and interact with E2F family members (35). It will be important to structurally define additional protein interaction surfaces within or outside of the pocket that mediate some of the pRb-specific tumor suppressor functions such as the CBFA1/Runx2 interaction. Such binding sites may not be conserved in p107 and p130 and hence represent particularly attractive drug targets.