Multiple Mechanisms for E2F Binding Inhibitionby Phosphorylation of the Retinoblastoma Protein C-terminal Domain

Abstract

The retinoblastoma proteinC-terminal domain (RbC) is necessary for the tumor suppressor protein's activities in growth suppression and E2F transcription factor inhibition. Cyclin-dependent kinase phosphorylation of RbC contributes to Rb inactivation and weakens the Rb-E2F inhibitory complex. Here we demonstratetwo mechanisms for how RbC phosphorylation inhibits E2F binding. We find that phosphorylation of S788 and S795 weakens the direct association between the N-terminal portion of RbC (RbC^N) and the marked-box domains of E2F and its heterodimerization partner DP. Phosphorylation of these sites and S807/S811 also induces an intramolecular association between RbC and the pocket domain,which overlaps with the site of E2F transactivation domain binding. Areduction in E2F binding affinity occurs with S788/S795 phosphorylation that is additive with the effects of phosphorylation at other sites, and we propose a structural mechanism that explains this additivity. We find that different Rb phosphorylation events have distinct effects on activating E2F family members, which suggests a novel mechanism for how Rb may differentially regulate E2F activities.

Introduction

The retinoblastoma protein (Rb) is a broad-functioning tumor suppressor that is frequently deregulated in human cancers.^1;2 The loss of functional Rb is associated with several hallmarks of cancer including chromosomal instability and aberrant cell proliferation. Rb acts as a negative regulator of cell division at the G₁-S transition of the cell cycle.^3;4;5;6 In G₀ and early G₁, Rb forms a growth-repressive complex with E2F transcription factors.^7;8 The Rb-E2F complex is stabilized through two cohesive interactions (Fig. 1a and 1b): the pocket domain of Rb binds and represses the E2F transactivation domain (E2F^TD),^9;10;11 and the C-terminal domain of Rb (RbC) associates with the E2F1-DP1 marked-box and coiled-coil domains (E2F1-DP1^CM).^12;13;14 These structured interactions are consistent with the finding that both the pocket domain and RbC are required for full growth suppression and E2F binding.^15;16;17

Fig. 1.

Schematic of Rb protein structure.(a)Domain organization and Cdk-consensus phosphorylation sites. Structured domains are colored, unstructured domains are uncolored, and domain boundaries used in this study are indicated. (b) Summary of interactions that contribute to the overall Rb-E2F complex. The Rb pocket domain bindsthe E2F transactivation domain (E2F^TD), while the Rb C-terminal domain (RbC) binds the E2F-DP coiled-coil and marked box (E2F-DP^CM) domains.(c) Summary of phosphorylation events and their structural effects that disrupt the Rb-E2F^TD complex. Phosphorylation of S608/S612 in the pocket loop (Rb^PL) induces binding of Rb^PL to the pocket at the E2F^TD site. Phosphorylation of T373 in the interdomain linker (Rb^IDL) induces N-terminal domain (RbN) docking tothe pocket,which allosterically disrupts the E2F^TD-binding cleft.

Cyclin-dependent kinases (CDKs) phosphorylate Rb at specific CDK-consensus sites in late G₁(Fig. 1).^{3;4;5;6;18;19;20;21} Hyperphosphorylated Rb dissociates from E2F, allowing for up-regulation of E2F-mediated transcription and entry into S-phase.^15;22;23 Protein crystal structures have revealed how several key phosphorylation events induce conformational changes to Rb that disrupt the Rb-E2F interfaces(Fig. 1c).²⁴ Phosphorylation of S608/S612 in the pocket loop (Rb^PL) promotes a binding interaction between Rb^PL and the pocket domain that is structurally analogous to the Rb-E2F^TD binding interaction.^25;26 Phosphorylation of T373 in the interdomain linker (Rb^IDL) stabilizes binding between the pocket domain and the N-terminal domain (RbN), inducing an allosteric change to the E2F^TD binding site in the pocket.^26;27 Phosphorylation of T821 and T826 in RbC also induces an intramolecular association between RbC and the pocket at the ‘LxCxE’ binding site;^14;28;29 data suggest that this interaction dissociates proteins involved in chromatin remodeling and gene silencing.²⁸ Quantitative binding studies have revealed that phosphorylation of sites in RbC also reduces binding between RbC and E2F1-DP1^CM, although this inhibitory mechanism has not been clarified.¹⁴

RbC phosphorylation is a critical component of E2F activation and is necessary for full transactivation activity at E2F-bound promoters.^30;31 We therefore sought here to determine whether RbC phosphorylation destabilizes binding between Rb and the E2F transactivation domain. In this study, we use isothermal titration calorimetry (ITC) and nuclear magnetic resonance (NMR) to observe phosphorylation-dependent changes in binding between Rb and E2F. We find that phosphorylation of S788 and S795 in the N-terminal region of RbC (RbC^N) inhibits the Rb-E2F^TDassociationbyinducing binding of RbC^Nto the pocket domain. In addition, phosphorylation of RbC^N abrogates the association between RbC^N and the E2F1-DP1^CM complex. We find that phosphorylation of RbC^N at S788/S795 is additive to the effects of Rb^IDL and Rb^PL phosphorylation in inhibiting Rb-E2F^TD binding, indicating structural compatibilities exist between these distinct mechanisms. Finally, we identify differences in how these phosphorylation-induced inhibitory mechanisms affect the binding of paralogous ‘activating’ E2F^TDs (E2F1-3) to Rb. Together, these binding studies contribute to a complete understanding of how specific post-translational phosphorylation events regulate the distinct functional interfaces of this critical cell cycle regulatory protein.

Results

RbC^N (S788/S795) Phosphorylation Inhibits E2F1^TD Binding to Rb Pocket by ITC

We used ITC to measure binding affinities of E2F1^TD(residues 372-437) for a series of Rb constructs, each engineered to contain specific RbC^N phosphorylation sites(Table1 and Supplementary Fig. S1). The proteins were phosphorylated quantitatively with recombinant Cdk as needed (Supplementary Fig. S2). We first tested binding ofE2F1^TDto an Rb construct(Rb^{380-816∆PL/S780A}) that contains the pocket domain and 4 phosphoacceptor sites (S788/S795/S807/S811)but lacks the Rb^PL sites (S608/S612) and S780. S780 phosphorylation does not influence E2F^TD binding,²⁵ and the S780A mutation facilitates homogeneous phosphorylation in the preparative in vitro kinase reaction. We found that E2F1^TD binds to phosphorylatedRb^{380-816∆P/S780A} (K_d = 0.47 ± 0.04 μM)with an affinity that is 7-fold weaker than its affinity for unphosphorylated protein(K_d = 0.07 ± 0.03μM). When both S807 and S811 are substituted for alanine in this construct, E2F1^TD retains the reduced binding affinity for phosphorylated Rb (K_d= 0.51 ± 0.07 μM). This result indicates that S807 and S811 do not contribute to the inhibition of the Rb-E2F1^TD interaction.

When S788 and S795 are both substituted to alanine, we find that phosphorylated Rb (K_d=0.13 ± 0.06 μM)binds E2F1^TD similar to unphosphorylated Rb(K_d=0.09 ± 0.03 μM), demonstrating that phosphorylation of S788 and S795 negatively affects Rb-E2F1^TD binding. Consistent with this result, we observe reduced E2F1^TD binding to a phosphorylated construct that is truncated to exclude both S807 and S811 and contains the S780 to alanine mutation. This construct (Rb^{380-800∆PL/S780A}) has only two phosphoacceptor sites intact (S788, S795) and binds E2F1^TD 10-fold more weakly when it is phosphorylated (K_d=0.51 ± 0.10 μM) compared to unphosphorylated(K_d=0.05 ± 0.01 μM). A similar construct, truncated to include only phosphorylation at S788 (Rb^{380-794∆PL/S780A}),has a relatively minor effect (K_d=0.27 ± 0.02 μM (phosphorylated), K_d=0.12 ± 0.01 μM (unphosphorylated)). In summary, these results reveal that phosphorylation of RbC^N at S788/S795 negatively regulates binding between E2F1^TD and the Rb pocket domain.

Phosphorylation of RbC^N Induces Binding to Rb Pocket

Rb^PLphosphorylation on S608/S612 induces an intramolecular association with the pocket domain that overlaps with the E2F^TD-binding cleft.^25;26 We hypothesized that phosphorylation of RbC^N similarly promotes intramolecular binding to the pocket domain. To test this idea, we generated ¹⁵N-labeled RbC^N peptide (RbC^787-816) to detect the associationin trans by NMR. This fragment is phosphorylated on S788, S795, S807, and S811. The ¹H-¹⁵N HSQC spectrum of the phosphorylated, ¹⁵N-labeled RbC^787-816 alone shows minimal peak dispersion in the proton dimension, typical of intrinsically disordered polypeptides. Titration of unlabeled Rb pocket into the sample reveals small chemical shift changes and considerable broadening for several peaks (Fig. 2a and 2b). The broadening isprotein concentration dependent (Fig. 2b) and anticipated for binding between the relatively small labeled peptide and the larger unlabeled pocket domain (molecular weight ~43kDa). Binding between phosphorylated RbC^787-816 and the pocket domain is too weak to be detectedin trans by ITC (data not shown); this weak binding(K_d> ~100 μM) is consistent with the high protein concentrations needed to observe the broadening effect in the NMR experiment. Peak broadening is not observed for the ¹⁵N-labeled unphosphorylated peptide in the presence of excess Rb pocket, demonstrating that the RbC^N-Rb pocket interaction is dependent on phosphorylation of the RbC^N peptide (Fig. 2c and Supplementary Fig. S3).

Fig. 2.

Phosphorylation of RbC^787-816 promotes intramolecular binding to the Rb pocket domain. (a)¹H-¹⁵N HSQC spectra of 50 μM ¹⁵N-labeled phosphorylated RbC^787-816 alone (black) and in the presence of 900 μM unlabeled Rb pocket^380-787∆PL (cyan). Broadening of select resonance peaks indicates an in trans association between the phosphorylated RbC^N peptide and the pocket domain. (b)The ratio of peak intensity in the presence (I) and absence (I(0)) of pocket domain is plotted for each residue in phosRbC^787-816. Ratios are plotted for addition of 300μM (dark blue) and 900 μM (cyan) unlabeled Rb pocket^380-787∆PLand for addition of 900 μM pocket in the presence of 900 μM unlabeled E2F2^TD(red). Intensity ratios were not quantified for overlapping peaks and prolines, marked with *.(c) HSQC spectra of 60 μM ¹⁵N-labeled unphosphorylated RbC^787-816 alone (black) and in the presence of 900μM unlabeled Rb pocket^380-787∆PL (cyan). The absence of peak broadening indicates a lack of binding between unphosphorylated RbC^787-816 and Rb pocket, which demonstrates that binding between RbC^787-816 and the pocket is dependent upon phosphorylation. The ratio of peak intensities in the presence and absence of pocket domain are plotted in Supplementary Fig. S3.(d) ¹H-¹⁵N HSQC spectra of 50 μM¹⁵N-labeled phosphorylated RbC^787-816 alone (black) and in the presence of 900 μM unlabeled Rb pocket^380-787∆PL and 900 μM unlabeled E2F2^TD(pink). The signal broadening, plotted in (b),is less extensive than in the absence of E2F2^TD, suggesting that binding between E2F2^TD and Rb pocket excludes phosphorylated RbC^787-816.

NMR peaks in the phosphorylated RbC^787-816 spectrum were assigned using standard methods. The peaksthat undergo the most pronounced broadeningcorrespond to clusters of residues surrounding phosphorylatedS788, S795 and S807 (Fig. 2b). The most straightforward interpretation of this result is that residues in these sequences directly contact the pocket domain. However, we cannot rule out the possibility that a subset of these spectral perturbations result from structural changes in RbC^Nthat occur upon association. Peaks corresponding to residues surrounding S807 are influenced in the pocket titration, even though these residues do not contribute to inhibition of E2F1^TD in the ITC assay. The chemical environments of these residues are perhaps influenced by pocket binding in a manner that is independent of the effect of phosphorylated RbC on E2F inhibition. We next tested for whether binding of phosphorylated RbC^N and E2F^TDto the pocket is mutually exclusive. We titrated a labeled, phosphorylated RbC^N sample with unlabeled Rb pocket that is first saturated with one molar equivalent of E2F2^TD. In this case, we observe reduced peak broadening effects, demonstrating that E2F^TDwhen present in excess prevents phosphorylated RbC^Nfrom bindingthe pocket (Fig. 2b and 2d).

We examined the structure of the pocket-E2F2^TD complex to identify potential binding sites between the pocket and phosphorylated RbC^N (Fig. 3a). We found two clusters of conserved sidechains that are capable of coordinating a phosphate. These sidechains make interactions with E2F^TD that would likely be disrupted by RbC^N binding and are in close proximity to the structured C-terminus of the pocket (I785 in the E2F2^TD-pocket structure).¹⁰ One site is near K652 and R656. K652 forms a salt bridge with E417 in E2F2. A second potential site is formed by K548, H555, R661, and H733. K548 and H555 hydrogen bond with D410 and D411 in E2F2^TD respectively. We used the NMR assay to examine whether mutation of these sites influences the binding of phosRbC^N to the pocket domain (Fig. 3b). We added unlabeled wild-type, K652A/R656A, and H555A/H733A pocket to ¹⁵N-labelled phosRbC^787-816 and compared the line-broadening in the spectra. The K652A/R656A pocket mutant shows reduced broadening consistent with loss of affinity, while the H555A/H733A mutant shows similar behavior as wild-type. These dataindicate that phosRbC^N binding requires an interaction with the conserved K652A/R656A pocket.

Fig. 3.

Phosphorylated RbC^N binds the pocket domain at a site that overlaps the E2F^TD binding site. (a) Putative interactions sites in the pocket domain for phosphorylated RbC^N sidechains. The structure of the E2F2^TD-pocket complex is shown (PDB code: 1N4M) along with two clusters of residues that could be reached by S795 (K548, H555, R661, and H733) and S788 or S795 (K652 or R656). Binding of RbC^N in this region of the pocket would disrupt interactions with the N-terminal portion of E2F^TD.(b) Mutation of K652/R656 but not H555/H733 inhibits binding of phosRbC^N to the pocket domain. The ratio of peak intensity in the presence (I) and absence (I(0)) of 300 μM wild-type (dark blue), K652A/R656A (green), and H555A/H733A (yellow) pocket domain is plotted for each residue in phosRbC^787-816. The K652A/R656A mutant induces relatively less line broadening, suggesting that the association depends on those residues.

Phosphorylation of RbC^N Dissociates RbC^N from E2F1-DP1^CM

It has been observed that that phosphorylation of S788/S795 in the context of the entire RbC (residues 771-928) weakens the RbC affinity for E2F1-DP1^CM.¹⁴ However, because RbC^N was not included in the ternary complex that crystallized, the mechanismhas not been characterized. We examined here whether RbC^Nbinds directly with E2F1-DP1^CM and whether this association is modulated by S788/S795 phosphorylation. We collected ¹H-¹⁵N HSQC spectra of unphosphorylated ¹⁵N-labeled RbC^787-816both alone and in the presence of unlabeledE2F1-DP1^CM (Fig. 4a). By comparing the spectra, we observe extensive signal broadening indicative of complex formation between unphosphorylated RbC^N and E2F1-DP1^CM. When we conduct this experiment using phosphorylated¹⁵N-labeled RbC^N, we see less signal-broadening (Fig. 4b), suggesting that phosphorylation of RbC^N destabilizes the direct binding interaction with E2F-DP1^CM. Taken together with the NMR data demonstrating an RbC^N-pocket interaction (Fig. 2), these results suggest that RbC^N phosphorylation has two distinct roles in disrupting the overall Rb-E2F complex (Fig. 4c). First, phosphorylation of RbC^N destabilizes the direct association between RbC^N and E2F1-DP1^CM. Second, phosRbC^Nbinds the pocket domain to destabilize the E2F1^TD interaction.

Fig. 4.

Phosphorylation of RbC^787-816 abrogates binding between RbC^N and E2F1-DP1^CM.(a) ¹H-¹⁵N HSQC spectra of 60 μM¹⁵N-labeled unphosphorylated RbC^787-816 alone (black), and in the presence of 300 μM unlabeled E2F1-DP1^CM(red). The peak broadening observed in this spectrum indicates binding. (b) HSQC spectra of 100 μM ¹⁵N-labeled phosphorylated RbC^787-816 alone (black) and in the presence of 300 μM unlabeled E2F1-DP1^CM (red). Less peak broadening is observed here than in (a), indicating that phosphorylation of RbC^787-816 inhibits the RbC-E2F1-DP1^CMcomplex. (c) Summary of interactions identified by NMR: phosphorylated RbC^787-816 dissociates from E2F1-DP1^CMand associates with Rb pocket in a manner that is inconsistent with E2F^TDbinding.

Phosphorylation of RbC^Nis Additivewith the Effects of Rb^PL and Rb^IDL Phosphorylation in Inhibiting E2F1^TD-Rb Pocket Binding

With these observed effects ofRbC^N phosphorylation, we have now identified three distinct structural mechanisms by which Rb phosphorylation inhibits E2F1^TD binding to the pocket domain. Phosphorylation of Rb^IDL (T356/T373) induces RbN-pocket docking and inhibits Rb-E2F1^TD binding approximately 45-fold.^25;26 Most of this effect is attributable to T373 phosphorylation alone, which mediates the conformational change,although T356 phosphorylation does enhance the inhibition of Rb-E2F1^TD binding.²⁶ Phosphorylation of Rb^PL at S608 induces association of Rb^PL and the pocket,which inhibits Rb-E2F1^TD binding approximately 10-fold. S612 phosphorylation alone in Rb^PL reduces E2F1^TD binding 3-fold, but has no additive effect when S608 is also phosphorylated.^25;26 Finally, we report here that phosphorylation of S788 and S795 in RbC^Ninduces an association with the pocket domain, which inhibits Rb-E2F1^TD binding approximately 7-fold.

We previously found that Rb^PL and Rb^IDL phosphorylation together have an additive effect in inhibiting Rb-E2F1^TD.²⁵ We next investigated whether RbC^N phosphorylation contributes additively to the inhibitory effects of Rb^PL and Rb^IDL phosphorylation. We first generated an Rb construct (Rb^380-816)that contains both the RbC^N(S788/S795) and Rb^PL(S608/S612) phosphorylation sites but lacks RbN and the Rb^IDL(T356/T373) phosphorylation sites. E2F1^TD binds unphosphorylated Rb^380-816/780Awith a K_d of 0.07 ± 0.03 μM, and phosphorylated Rb^380-816/780A with a K_d of 3.1 ± 0.5 μM (Fig. 5a and 5b). This 45-fold reduction in E2F1^TD binding is nearly the product of the 10-fold and 7-fold effects observed independently,indicating that the Rb^PL and RbC^N mechanisms work together to inhibit the Rb-E2F1^TD complex.

Fig. 5.

ITC data demonstrate that Rb-E2F binding inhibition occurs through the additive effects of multiple mechanisms. Dissociation constants were determined from binding isotherms of E2F1^TD with (a)unphosphorylated Rb^380-816/780A,(b) phosphorylated Rb^380-816/780A, (c) unphosphorylated Rb^{53-800∆NL/∆PL/S780A}, and (d) phosphorylated Rb^{53-800∆NL/∆PL/S780A}. As shown in the schematic diagrams, these constructs each contain elements needed for two inhibitory mechanisms. Their phosphorylation results in reduced affinity that is greater than constructs containing single inhibitory mechanisms.

To test whether RbC^N and Rb^IDL phosphorylation similarly add in inhibiting E2F^TD binding, we used a construct of Rb thatcontains RbN,Rb^IDL, the pocket, and RbC^N, but lacks the large loops in the pocket and RbN (Rb^{55-800∆NL/∆ PL/S780A}). Weobserve that E2F1^TD binds 230-fold more weakly when this Rbconstruct is phosphorylated (K_d = 30 ± 10 μM)compared to when it is unphosphorylated(K_d = 0.13 ± 0.05 μM) (Fig. 5c and 5d). RbC^N and Rb^IDL phosphorylation mechanisms are therefore also additive, as phosphorylation at these sites has a 7-fold and 45fold effect alone respectively. These dataindicate that RbC^N phosphorylation functions in a manner that is functionally compatible with the mechanisms induced by phosphorylation of Rb^IDL and Rb^PL, lending insight into the way that these three separate structural mechanisms contribute to the inhibition of Rb-E2F^TD binding.

Phosphorylation Mechanisms Regulate E2Fs Differently

The E2F family of transcription factors consists of members that both activate transcription (E2Fs 1-3) and repress transcription (E2Fs 4-8). During G₀ and early G₁, Rb negatively regulates transcription by binding and repressing activating E2Fs 1-3, whereas the Rb paralogs p107 and p130 associate with repressive E2F4 and E2F5.³² E2F1-5 display a large degree of sequence and structural similarity, and crystal structures of E2F1^TD and E2F2^TD bound to Rb pocket reveal mostly similar binding modes. There are some notable, albeit subtle, differences in binding. For example, E2F2^TD makes additional salt bridge and hydrogen bond contacts through D410 and D411, which are not observed in the structure of Rb-E2F1^TD.^10;11 These differences in binding contacts suggest that E2Fs may be differentially affected by the distinct phosphorylation-induced mechanisms for E2F release.

We compared binding of E2F1^TD, E2F2^TD and E2F3^TD to Rb constructs designed to promote the specific phosphorylation-induced structural mechanisms we have identified for inhibiting E2F^TD binding. These ITC data are summarized in Table 2and shown in Supplementary Fig. S4. Previously we found that phosphorylation of Rb^IDL (T356/T373) inhibits E2F1^TD binding to Rb pocket 45-fold.²⁵ When we perform this experiment using the other activating E2Fs, we find that E2F3^TD binding is inhibited 25-fold, while E2F2^TD binding is inhibited only 2-fold. Phosphorylation of Rb^PL (S608/S612) and RbC^N (S788/S795) raise the K_d for Rb-E2F3^TD binding 9-fold and 10-fold respectively. These effects are similar to what is observed for E2F1^TD;²⁵ however, we also observe that either of these mechanisms alone produces only a relatively smaller3 to 4fold inhibition of E2F2^TD binding. When we test the effect of phosphorylating Rb^IDL and RbC^N together, we observe a larger22-fold reduction in the binding affinity of E2F2^TD, suggesting that these mechanisms have a cooperative effect in dissociating E2F2^TD from Rb.

These results support several conclusions regarding how phosphorylation modulates differently the interaction between Rb and the activating E2Fs. First, the Rb-E2F2^TD interaction is distinct from E2F1^TD and E2F3^TD in that individual phosphorylation events produce little effect. Second, E2F1^TD and E2F2^TD have considerably higher affinity for unphosphorylated Rb than E2F3^TD. Therefore, like for E2F2^TD, multiple phosphorylation events on Rb are required for changing the E2F1^TD affinity to the micromolar range. With its weaker initial affinity, E2F3^TD only requires any one phosphorylation event for micromolar affinity. Finally, we note that in the context of full-length phosphorylated Rb, it was observed that deletion of RbC has no effect on the inhibition of E2F1^TD binding.²⁵ We suggest that the inhibitory effect of RbC^N phosphorylation observed here may not be additive with both Rb^IDL and Rb^PLphosphorylation together. Another possible explanation is that the association between RbC and the pocket induced by T821/T826 phosphorylation negatively influences the ability of phosphorylated RbC^N to bind the pocket and inhibit E2F^TD. Further exploration of the interdependence of phosphorylation events and their corresponding structural changes is needed to address these possibilities.

Discussion

Rb inactivation by multisite phosphorylation is required for cell cycle advancement into S phase and is commonly found in tumors. The studies presented here reveal a novel phosphorylation-induced mechanism that utilizes RbC and contributes to inhibition of the Rb-E2F growth-repressive complex. Our datademonstrate that S788/S795 phosphorylation plays a dual role in disrupting both the RbC-E2F1-DP1^CM and Rb pocket-E2F^TD interfaces. Our NMR data support a model in which phosphorylation of these sites stabilizes binding of RbC^N to the pocket domain. The interdomain association is likely mediated by other amino acids surrounding the phosphate moieties and takes place at a site that overlaps with the E2F^TD binding site.

Our analysis indicates that RbC^N phosphorylation enhances repression of E2F^TD binding induced by either Rb^PLorRb^IDL phosphorylation, indicating that these pairwise mechanisms are functionally compatible. The additive inhibitionof these phosphorylation events isconsistent with theirknown and proposed structural effects. The binding site for phosphorylated RbC^Nin the pocket domain near K652/R656suggest that phosphorylated RbC^N, like Rb^IDL phosphorylation,disrupts interactions between the pocket and the N-terminal section of E2F^TD(Fig. 3).²⁶ Alternatively, Rb^PLphosphorylation targets a separate Rb-E2F^TD interface near the C-terminus of E2F^TD.²⁶ We propose that the additive effect of RbC^N and Rb^PL phosphorylation in inhibiting Rb-E2F^TD binding occurs because these two mechanisms each targets a distinct, stabilizing subsection of the overall Rb-E2F^TD interface. We find that S807/S811 phosphorylation does not contribute to inhibition of the Rb-E2F^TD interaction either alone or in the presence of S788/S795. Although it remains a formal possibility that these residues contribute in some unique way to the regulation of Rb-E2F, other roles for these phosphoacceptor sites have been suggested,including modulating interactions with other proteins and serving as priming sites for subsequent phosphorylation events.^24;29;33

The quantitative binding studies presented here are consistent with cell-based studies, which indicate S788/S795 phosphorylation plays a role in the dissociation of Rb-E2F complexes and E2F transactivation that is additive with other sites. Transcriptional assays using alanine substitutions at S788 and S795 show that these mutations alone are not sufficient to suppress E2F activation in the presence of kinase.³⁰ However, these mutations do greatly enhance the effects of additional phosphoacceptor mutations in suppressing E2F activity in the presence of kinase. A similar study confirms that S795 phosphorylation alone is insufficient to induce E2F activity.³⁴ In cells, S795 phosphorylation is most likely the target of CDK4, and not CDK2.^21;35;36 More recent studies have shown that S795 phosphorylation and E2F activation also occur in response to p38 activation.^37;38

We tested for the first time here the specific effects of each phosphorylation mechanism on different activating E2F transactivation domains. We found that while discrete phosphorylation at T356/T373, S608/S612, or S788/S795 reduces the affinity of Rb for E2F1^TD or E2F3^TD nearly 10-fold in each case, the effect of phosphorylating one pair of sites on E2F2^TD binding is modest. Phosphorylation of both the Rb^IDL and RbC^N sites together dramatically reduces E2F2^TD. While we do not understand the details explaining this synergy and its specificity for E2F2^TD, we note that E2F2^TD makes specific additional contacts relative to E2F1^TD in the region of the pocket domain most affected by the structural changes induced by RbC^N and Rb^IDL phosphorylation.

The observation that E2F1^TDandE2F2^TD is less responsive to activation by specific phosphorylation events than E2F3^TD is noteworthyconsidering activating E2Fs have distinct cellular functions,regulate only partially overlapping sets of genes, and show differences in how they regulate expression in conjunction with other transcription factors and post-translational modifications.^32;39;40 We suggest that such differential effects of phosphorylation provide a possible mechanism for tuning different E2F family member activities separately. Further identification of specific Rb phosphorylation events in different cellular contexts remains an important goal for understanding when and how these mechanisms contributeto Rb regulation of E2F activity.

Materials and Methods

Protein Expression, Purification and Phosphorylation

Rb and E2F protein constructs were expressed in E. coli as fusion proteins with glutathione-S-transferase (GST). Cells were induced overnight at room temperature with the exception of RbC^787-816, which was induced for 2.5 hr at 37°C. Cells were re-suspended in a lysate buffer containing 100mM NaCl, 25mM Tris-HCl pH 8.0, 1mM PMSF and 1mM dithiotheritol, passed twice through a cell homogenizer, and centrifuged at 27,000 × g for 30 minutes. The supernatant fraction of the lysate was loaded to GS4B affinity resin and eluted with 10mM glutathione. Elute was further purified using anion exchange chromatography. The GST tag was cleaved overnight at 4°C with 1% (percentage mass of substrate in the reaction) GST-tagged tobacco etch virus (TEV), and the cleaved protein was passed back over GS4B resin to remove GST. E2F1-DP1^CMwas prepared and purified as described previously.¹⁴

Purified Rb protein was phosphorylated by concentrating to 4 mg/mL and incubating in a reaction containing 5% CDK2-CyclinA, 10mM ATP, 10mM MgCl₂, 100mM NaCl, 25mM TrisHCl pH 8.0, 1mM dithiotheritol at 4°C overnight. Quantitative phosphorylation of RbC^787-816 could be achieved but required using5% CDK2-CyclinA and 2% CDK6-CyclinK. Proteins were analyzed for phosphate incorporation using electrospray ionization mass spectrometry as previously described.²⁵A sample mass spectrometry experiment demonstrating quantitative phosphorylation is shown in SupplementaryFig.S2.

Nuclear Magnetic Resonance Spectroscopy

All samples were prepared in a buffer containing 50mM NaPO₄, 5mM dithiotheritol and 10% D₂O (pH 6.1). ¹H-¹⁵N HSQC Spectra were recorded at 25°C on a Varian INOVA 600-MHz spectrometer equipped with a HCN 5-mm cryoprobe. HSQC assignments for ¹⁵N-labeled phosphorylated RbC^787-816 were obtained using ¹H-¹⁵N sidechain correlations observed in a three-dimensional ¹N-¹⁵N TOCSY spectrum (120ms mixing time) and a three-dimensional ¹N-¹⁵N NOESY spectrum (360ms mixing time).^41;42 All NMR spectra were processed with NMRpipe and analyzed with NMRViewJ.^43;44

Isothermal Titration Calorimetry

ITC experiments were conducted with a MicroCal VP-ITC calorimeter. Proteins were dialyzed overnight in a buffer containing 100mM NaCl, 25mM Tris-HCl (pH 8.0), and 1mM dithiotheritol. The data were fit to one-site binding using the Origin software package. Reported errors values reflect the standard deviation of 2-4 separate binding experiments. In all cases the stoichiometry parameter was determined to be around n ~ 1.

Supplementary Material

01

Supplementary Fig.S1 ITC data corresponding to the measurements summarized in Table 1.

02

Supplementary Fig.S2 Electrospray ionization mass spectrometry data demonstrating the preparative in vitro kinase reaction yields quantitative phosphorylation on available Cdk phosphoacceptor sites. Rb^{380-816∆ PL/S780A}, which contains four acceptor sites (S788/S795/S807/S811), was phosphorylated as described in Materials and Methods. The spectra for the unphosphorylated (above, blue) and phosphorylated (below, red) proteins are shown. The difference in molecular weight (+320) corresponds to the addition of four phosphates.

03

Supplementary Fig.S3 Titration of Rb pocket domain into unphosphorylated ¹⁵N-labelled RbC^787-816 shows little perturbation to the HSQC spectrum, which is consistent with a lack of an association. The ratio of RbC peak intensities in the presence (I) and absence (I(0)) of pocket are plotted for the 19 unassigned, non-overlapping peaks in the spectrum.

04

Supplementary Fig.S4 ITC data corresponding to the measurements summarized in Table 2.

Highlights.

• Phosphorylation of the Rb C-terminal domain (RbC) inhibits binding of the E2F transcription factor transactivation domain.

• Phosphorylation in RbC induces a conformational change that blocks access to an E2F binding site.

• The RbC mechanism is additive with other phosphorylation induced inhibitory mechanisms.

• Rb complexes with different E2F family members are inhibited differently by each mechanism.