Selectivity by small-molecule inhibitors of protein interactions can be driven by protein surface fluctuations

Abstract

Small-molecules that inhibit interactions between specific pairs of proteins have long represented a promising avenue for therapeutic intervention in a variety of settings. Structural studies have shown that in many cases, the inhibitor-bound protein adopts a conformation that is distinct from its unbound and its protein-bound conformations. This plasticity of the protein surface presents a major challenge in predicting which members of a protein family will be inhibited by a given ligand. Here, we use biased simulations of Bcl-2-family proteins to generate ensembles of low-energy conformations that contain surface pockets suitable for small molecule binding. We find that the resulting conformational ensembles include surface pockets that mimic those observed in inhibitor-bound crystal structures. Next, we find that the ensembles generated using different members of this protein family are overlapping but distinct, and that the activity of a given compound against a particular family member (ligand selectivity) can be predicted from whether the corresponding ensemble samples a complementary surface pocket. Finally, we find that each ensemble includes certain surface pockets that are not shared by any other family member: while no inhibitors have yet been identified to take advantage of these pockets, we expect that chemical scaffolds complementing these "distinct" pockets will prove highly selective for their targets. The opportunity to achieve target selectivity within a protein family by exploiting differences in surface fluctuations represents a new paradigm that may facilitate design of family-selective small-molecule inhibitors of protein-protein interactions.

Fig 1. Building “exemplars” from surface pockets.

(A) Bcl-xL (grey surface) is shown in complex with an inhibitor (spheres). (B) The protein surface features a large pocket (small white spheres) that is complementary in shape to the inhibitor. (C) From this surface pocket, an “exemplar” is built: a map of the “perfect” ligand to complement this protein surface, without considerations of atom connectivity. The exemplar is comprised of hydrogen bond donors (yellow) and acceptors (magenta) that complement surface polar groups on the protein, and hydrophobic atoms that fill the remainder of the surface pocket (cyan). In the studies we report here, we will use the shape and chemical features of the exemplar as a proxy for the shape and chemical features of the protein surface pocket.

Fig 2. Bcl-2 family members recognize different inhibitors using distinct surface pockets.

In all cases color gradient indicates the similarity between complexes, expressed as Z-scores (green are most similar, red are most dissimilar). (A) Chemical similarity of the inhibitors. (B) Three-dimensional similarity of the inhibitors’ active conformation. (C) Similarity of protein surface pockets, measured using exemplar similarity. Numbering in all cases corresponds to complexes in S1 Table. A representative subset of the complexes are included in this figure; a corresponding figure containing all available complexes is available as S2 Fig.

Fig 3. Ensembles of low-energy pocket-containing conformations.

(A) To examine the effect of the particular target residues used in generating the ensemble of pocket-containing conformations, we generated separate ensembles from the top-scoring pair of target residues (Ala93 and Arg139, green) and the second-best pair (Ala93 and Ala142, red). We use exemplars to compare the similarity of surface pockets on each conformation, and we show each conformation on the two-dimensional projection that best reflects the pairwise distances between them. The overlap between the two ensembles highlights the robustness of the conformations to the particular target residues. (B) For each member of the Bcl-2 family, we generated an ensemble of conformations from unbiased simulations; the distribution of these energies is shown (black). The range of energies for pocket-containing conformations generated using a biasing potential target residues derived from the Bcl-xL protein interaction site (magenta) or the Mcl-1 protein interaction site (red) suggest that many of these conformations are energetically accessible to these proteins under physiological conditions. All energies shown here were evaluated in the absence of the biasing potential, for fair comparison. Each simulation is started from the structure of the unbound protein; a corresponding figure starting from the peptide-bound structures containing all available complexes is available as S5 Fig.

Fig 4. Maps of “pocket space” sampled by individual Bcl-2 family members.

The ensemble of pockets observed from simulation: individual conformations are represented as points on a two-dimensional projection that reflects the pairwise distances between their exemplars. The relative position of exemplars from experimentally-derived Bcl-xL unbound (“U”) and peptide-bound (“P”) structures are indicated, as are the positions of exemplars from Bcl-xL structures solved in complex with various inhibitors (numbers correspond to complexes listed in S1 Table). Exemplars marked “D” correspond to the same “distinct” conformations described in Fig. 5.

Fig 5. Comparison of “pocket space” sampled by each Bcl-2 family member.

(A) A projection built using ensembles collected from simulations of several Bcl-2 family members: Bcl-xL (magenta), Bcl-2 (green), Mcl-1 (red), Bcl-w (orange), and Bax (cyan). Bid and Ced-9 were used in generating the projection, but for clarity are not included on this map. Target residues derived from the Bcl-xL protein interaction site were used in generating exemplars shown on the Bcl-xL and Bcl-2 MDS plots, whereas target residues derived from the Mcl-1 protein interaction site were used in generating exemplars shown on the Mcl-1 MDS plots. For each of Bcl-xL, Bcl-2, and Mcl-1 we observe a distinct region of conformational space (“D”) that is not sampled by any other Bcl-2 family member. (B) Comparison of an exemplar from each “distinct” region to the corresponding unbound (“U”) or peptide-bound (“P”) protein structure from which the simulation was initiated. (C) Comparison of the conformation harboring the “distinct” pocket to the corresponding unbound/peptide-bound protein structure from which the simulation was initiated.

Fig 6. Ensembles of available pocket shapes contain distinct pocket shapes.

We define the “distinctness” of a pocket as the difference in exemplar distances of the closest conformation from a different family member, and the closest conformation from one’s own ensemble. Histograms are shown over conformations that comprise the ensembles used above. By this measure, all known inhibitors of all Bcl-2 family members bind to pockets that are not unique to their cognate target protein (i.e. low “distinctness”). Data are shown for three representative Bcl-2 family members, the complete set are included as S7 Fig.

Fig 7. Ensembles of available pocket shapes explain ligand selectivity across the Bcl-2 family.

(A) For each inhibitor-bound pocket, the exemplar distance to the most similar pocket is indicated by color gradient, with all distances expressed as Z-scores (green are most similar, red are most dissimilar; the range of colors for each row is normalized across that row). Markings inside the cells denote experimental reports [,,,–76] for a given protein-ligand pair (Kd or Ki where available, otherwise IC50): ✔✔✔ indicates < 0.1 μM, ✔✔ indicates 0.1–1 μM, ✔ indicates binding weaker than 1 μM, and ✗ indicates that binding was not detectable/quantifiable. Cells that do not include any markings correspond to protein-ligand pairs for which binding data has not been reported. Numbering corresponds to complexes as in S1 Table. A representative subset of the complexes are included in this figure; a corresponding figure containing all complexes is available as S8 Fig. The underlying raw data are included in S5 Table (exemplar distances) and S7 Table (citations to binding data). (B) A receiver operating characteristic (ROC) plot demonstrating the performance of exemplar distances for predicting whether a given compound is active against a particular member of the Bcl-2 protein family.

Fig 8. Pocket shapes explain (+)-JQ1 selectivity across bromodomains.

For each bromodomain, the exemplar distance of (+)-JQ1 to the most similar pocket is indicated by color gradient, with all distances expressed as Z-scores (green are most similar, red are most dissimilar; the range of colors for each row is normalized across that row). Markings inside the cells denote experimental binding measurements [10] for each protein-ligand pair: ✔✔✔ indicates ΔTm > 7°C, ✔✔ indicates ΔTm = 3–5°C, ✔ indicates ΔTm = 0–1°C, and ✗ indicates no detectable binding. The underlying raw data are included in S8 Table.