Burke, Hura, and Rubin

1

Burke, Hura, and Rubin

Supplemental Figure Legends

Supplemental Figure 1. S608E mimics the inhibitory effect of RbPL phosphorylation on E2FTD binding. (A) Table of calorimetry data for E2FTD (E2F1372-437) binding to Rb pocket domain (Rb380-787) wild type and mutant constructs. Previous results indicate that E2FTD binding is reduced 14-fold upon phosphorylation of the S608/S612 sites in RbPL(Burke et al. 2010). (B) E2FTD affinity (Kd = 1.2  0.2 M) for a phosphorylated construct containing a truncated RbPL (∆616-642) and an S780A mutation is also reduced. The S780A mutation is used to achieve homogeneous phosphorylation of constructs used in structural studies, but does not affect E2FTD binding(Burke et al. 2010). (C) E2FTD affinity for the truncated construct containing S608E/S612A/S780A mutations (Kd = 0.6  0.1 M) is similarly reduced, which indicates that the crystallization construct RbPL-Precapitulates the inhibitory mechanism. (D) Comparison of the RbPL helix with modeled phosS608 (left) and the S608E mutant observed in the RbPL-P structure (right). The phosS608 was built in by simple mutation in Coot with no further refinement, and the interatomic distances are appropriate for the indicted hydrogen bonding. In the RbPL-P structure with S608E (Fig. 2C), D604 and Y606 in RbPL make specific contacts with R467 and F482 in the pocket domain. R467 and F482 mutation disrupts phosphorylated RbPL binding to the pocket in trans(Burke et al. 2010), further validating 608E as a mutation that mimics S608 phosphorylation in properly structuring RbPL.

Supplemental Figure 2. Effects of T356/T373 phosphorylation on E2FTD binding affinities in Rb mutants. (A) Table of calorimetry data for E2FTD (E2F1372-437) binding to wild type and mutant Rb constructs containing RbN and the pocket domains. Previous results indicate that E2FTD binding is reduced 10-fold upon phosphorylation of the T356/T373 sites in RbIDL(Burke et al. 2010). (B-D) E2FTD affinity was quantified for a construct containing both RbN and the pocket but lacking the internal loops in each domain (Rb∆LoopsAffinities for unphosphorylated (Kd = 0.14  0.06 M) and phosphorylated (Kd = 6  1 M) Rb∆Loops are similar to values previously reported for an analogous construct containing the loops(Burke et al. 2010), indicating that deletion of internal loops neither affects E2FTD affinity nor inhibition induced by RbIDL phosphorylation. An S780A mutation, which facilitates homogeneous phosphorylation for structural studies, also does not affect inhibition. (E) E2FTD affinity for the phosphorylated crystallization construct Rb∆Loops,K289A,Y292A,S780A (RbN-P; Kd = 4  1 M) is similar to its affinity for the other phosphorylated constructs. The K289A and Y292A mutations in the crystal structure appear close to a crystal-packing interface and are not involved in forming the E2FTD binding site or RbN-pocket interfaces. (F-G) E2FTD affinity for a T373A mutant (phosRb∆Loops,T373A,S780A; Kd = 0.32  0.06 M) is similar to its affinity for unphosphorylated Rb, while its affinity for a T356A mutant (phosRb∆Loops,T373A,S780A; Kd = 2  1 M) is similar to phosphorylated Rb. Considering that T373 mutation but not T356 mutation abrogates the inhibitory effect of RbIDL phosphorylation, we conclude that T373 phosphorylation is necessary and sufficient for E2FTD inhibition by RbN docking. This observation is consistent with the structural effects of T373 phosphorylation (Fig.4B and Supplemental Fig.3).

Supplemental Figure 3. Small angle x-ray scattering (SAXS) curves for Rb mutant constructs. (A) The required mutations for crystallization (K289A and Y292A) do not affect the solution conformation of the phosphorylated protein. SAXS scattering curves for phosRb∆Loops,S780A and RbN-P (phosRb∆Loops,K289A,Y292A,S780A) are similar. (B) Phosphorylation of T373 and not T356 induces a global conformational change in Rb. The left panel shows that small angle scattering data for phosRb∆Loops,T373A,S780A (purple) are similar to data for unphosphorylated Rb∆Loops,T373A,S780A (yellow) and unphosphorylated Rb∆Loops,S780A (pink), indicating the phosphorylated T373A mutant protein does not have the same conformation as phosRb∆Loops,T373A,S780A (blue). The panel on the right shows that data for phosRb∆Loops,T356A,S780A (brown) are similar to phosRb∆Loops,S780A (blue) and distinct from the unphosphorylated proteins (pink and green), demonstrating that T356 phosphorylation is not necessary for the conformational change.

Supplemental Figure4. Modeling Rb∆Loops SAXS data utilizing atomic resolution coordinates. (A) Modeller model of Rb∆Loops(magenta)superimposed on the RbN-P crystal structure (purple). (B) Loops and missing termini were allowed to flex in subsequent BilboMD analysis (cyan), while RbIDL (yellow) was allowed to act as a flexible tether between the RbN and pocket domains. (C) BilboMD was used to generate potential solution conformations for fitting SAXS data. A sample trajectory from BilboMD, in which the rigid part of the pocket domain from each snapshot is displayed as magenta ribbon and moving loops and the RbN domain are shown in black wire. The RbN domain itself is rigid but moves in the trajectory because of the flexible RbIDL. A model from one of 36 molecular dynamic trajectories from BilboMD is visualized. (D) The experimental SAXS profile (black) for phosphorylated Rb∆Loops is compared with SAXS curves calculated from the RbN-P crystal structure (red), the Modeller model (green dashes), the best fitting model from BilboMD (magenta, also shown in Fig. 5A), and an ensemble of four models (cyan). (E) Dividing the experimental and calculated model curves highlights the quality of fits. (F) The pair distribution functions calculated from all curves in panel D. (G-I) The same analysis as in panels D-F is shown for the unphosphorylated Rb∆Loops. (J) Minimal ensembles fitting the SAXS data. The MES algorithm was used to identify 4 models, which together (optimally weighted), maximize the fit to the SAXS data for unphosphorylated and phosphorylated Rb∆Loops (cyan curves in panel D-I). The pocket domain is shown as ribbons in the same orientation for all models. The RbN domain is represented in spheres. The size of the model correlates with its contribution to the overall population. Because many ensembles differing in atomic resolution details may fit the SAXS data equally well, only lower resolution properties are described. For phosphorylated Rb∆Loops, 96% of the MES population was compact. The 2 improvement, relative to the best single model seems to be due to small changes throughout the scattering curve as shown in panels D and E. In contrast, the unphosphorylated state when modeled in the same way shows significant improvements in specific regions of the SAXS curve by including both extended and compact structures (G and H). In the single best-fit model to the unphosphorylated state the RbN and pocket domains are dissociated (Fig. 5A). Applying MES, an ensemble with 60% dissociated structures was complimented with 40% associated structures. The compact structures enable the SAXS curve determined from the ensemble to fit frequencies apparent in the experimental SAXS curve that the dissociated models cannot fit alone (G and H).

Supplemental Figure 5. Isothermal titration calorimetry data for binding of E7LxCxE and phosRbC to Rb. Representative ITC data used for binding measurements reported in Table 2.

Supplemental Figure 6. E2FTD binding data supporting the allosteric model for inhibition. (A) Table of calorimetry data for E2FTD binding to Rb wild type and mutant constructs. (B) The K652Q mutation to the pocket domain results in weaker E2FTD binding (Kd = 1.9  0.8 M), indicating that K652 interaction is required. K652 is misaligned for E2F binding in the RbN-P structure, supporting its role in the allosteric model for E2FTD inhibition upon phosphorylation-induced RbN docking. (C and D) The affinity of E2FTD for Rb∆Loops,Q736A,K840Ais similar when unphosphorylated (Kd = 0.15 0.05M) and phosphorylated (Kd = 0.3 0.1M). Considering the role of Q736 and K740 in forming the pocket subdomain B – RbN subdomain A interface (Fig. 4C), these data demonstrate the requirement of both pocket-RbN interfaces in inhibiting E2FTD binding.