Trends in Peptide and Protein Sciences 2017ˏVol. 1ˏ No. 4ˏ153-166 DOI: http://dx.doi.org/10.22037/tpps.v1i4.18019 Original Research Article

Evaluating the Difficulty Involved in Designing Small Molecule Drugs to Inhibit Protein-Protein Interactions

Delaram Ahmadi and David Barlow*

Department of Pharmacy, King’s College London, London SE1 9NH, UK

Article history: HIGHLIGHTS Received: 17 July 2017 • Mean interfacial surface curvatures have been determined for protein-protein interaction (PPI) partners in Accepted: 24 July 2017 their complexed and uncomplexed states. • Mean interfacial surface roughnesses have been determined for protein-protein interaction (PPI) partners in their complexed and uncomplexed states. • Amino acid compositions have been determined for PPI interface surfaces and these compared with that for the average protein surface. • Quantification of the PPI interfacial surface properties is used to assess the druggability of these targets.

ABSTRACT

The targeting of drugs to block protein-protein interactions (PPIs) has attracted great Keywords: interest over recent years. Such targets, however, have been held to be difficult to inhibit Drug Design using low molecular weight compounds, and as a consequence they are often branded Druggability as “undruggable”. This is partly because the interfaces involved are seen to be large, Molecular Modelling and the fact that they are generally regarded as being too smooth and too flat. In the Protein-protein work reported here, a series of quantitative systematic studies have been performed to interactions determine the molecular area, roughness, curvature, and amino acid composition of the interfacial surfaces of PPIs, to determine the feasibility of designing small molecule drugs to inhibit these interactions. The X-ray crystal structures are analysed for a set of 48 PPIs involving G-protein, membrane receptor extracellular domain, and enzyme-inhibitor complexes. The protein partners involved in these PPIs are shown to have much larger interfacial areas than those for protein-small molecule complexes (≥ 900 Å2 vs ~250 Å2 respectively), and they have interfaces that are fairly smooth (with fractal dimensions close to 2) and quite flat (with mean surface curvatures in the order of ± 0.1 Å-1). The mean interfacial surface curvatures of the PPI protein partners, however, are seen to change upon complexation, some very significantly so. Despite the fact that the amino acid compositions of the PPI interface surfaces are found to be significantly different from that of the average protein surface (with variations according to the type of PPI), it is concluded that the prospects for designing low molecular weight PPI inhibitors that act in an orthosteric manner remain rather limited.

* Corresponding Author: Email: [email protected] (D. Barlow) D. Ahmadi & D. Barlow / TPPS 2017 1(4) 153-166 Figure 1 Introduction

Protein-protein interactions (PPIs) play a fundamental role in mediating key biological processes, such as the mammalian immune response, cell-cell recognition, cell proliferation, growth, differentiation, signal transduction and apoptosis. Changes in the specificity and affinity of these interactions can lead to cellular malfunction like uncontrolled cell growth that results in cancer (Toogood, 2002; White et al., 2008; Raj et al., 2012; Nero et al., 2014). Inevitably, interest has arisen amongst scientists within the pharmaceutical industry to develop inhibitors of these interactions as therapeutic agents. In the past decade, therapeutic proteins such as monoclonal antibodies, which target PPIs that are accessible to the extracellular environment, have proven tremendously successful. However, these macromolecules Figure 1. Cartoon illustrating the relation between the mean surface can be expensive and are rarely orally bioavailable. curvature values (k) and the notional external sphere fitted to the Thus, the relative lower cost, ease of use and potential to surface, with radius = k-1. Negative values of k indicate convex target intracellular PPIs all favour the discovery of small surfaces, positive values indicate concave surfaces. synthetic molecule (<500 Da) PPI modulators (Nero et al., 2014). However, PPIs occur over a large surface area, with approximately 750 to 1,500 Å2 buried on each side of the flat, we aimed to evaluate the difficulty in devising low protein interface (Jones and Thornton, 1996; Conte et al., molecular weight drugs targeted against PPIs of the kind 1999). Moreover, the contact surfaces involved at these relevant to anticancer drug discovery. interfaces are qualitatively observed to be rather smooth, flat and featureless, and thus lack the deeply buried grooves Materials and Methods and pockets that provide for the high affinity binding sites for small molecules like those found in the substrate A database of 48 PPIs (with X-ray single crystal structures binding sites of enzymes (Wells and McClendon, 2007). determined to ≤ 2.5 Å resolution) was compiled from Despite the ‘undruggability’ of these PPIs, recent the information provided by Kastritis et al., (2011). The studies show that there have been some successes seen in atomic co-ordinates of the protein complexes of interest24 the development of small molecule inhibitors as cancer (together with those of the unbound protein partners) therapeutics. For example, with the recent case of the were obtained from the Protein Data Bank in Europe MDM2/p53 interaction inhibitors, there are interaction (PDBe) (Gutmanas et al., 2013) accessed at http://www. ‘hot spots’ on the PPI surfaces detected, which can provide ebi.ac.uk/pdbe/. The molecular surface areas and amino a focus for small molecule inhibitor interactions [Conte et acid compositions of the PPI interfaces for the bound and al., 1999; Vassilev et al., 2004; Morrow and Zhang, 2012; unbound proteins involved in G-protein (OG) complexes, Thangudu et al., 2012). At the same time, there has been receptor extracellular domain (OR) complexes, and much ongoing research in the challenges governing the enzyme inhibitor (EI) complexes were obtained using inhibition of PPIs with small molecule drugs (Arkin and PDBePISA (Krissinel et al., 2007). Mean interfacial Wells, 2004; Laraia et al., 2015). surface curvature values for the component proteins The complexes of interest within this study – involved in PPI complexes were computed (in Å-1, see specifically, those reflecting PPIs which might be targeted Fig. 1) using the SurfRace program (Tsodikov et al., in cancer therapy – were chosen from those identified by 2002), employing a probe radius of 1.2 Å. Kastritis et al., (2011) and are binary complexes involving The roughness of each PPI interface was computed either G-proteins (OG), receptor extracellular domains from the fractal dimension of the surface (D), using the (OR) or enzyme inhibitors (EI). gradient (2 – D) of a plot of log(A) vs. log(r), (Lewis and The research reported here focuses on providing a Rees, 1985) where A is the molecular surface area of systematic quantitative analysis of the characteristics of the interface atoms (calculated using SurfRace) given a these PPI surfaces and the structural changes involved spherical probe of radius, r (varied over the range 0.75 Å at the interfaces following complexation. By testing to 4.0 Å), using the atomic van der Waals radii given by the hypothesis that the PPIs are relatively extensive in Chothia (1976). D can take values from 2 (which indicates interface area, have smooth interfaces that are relatively a perfectly smooth surface) through to 3 (which indicates

154 Figure 2 Designing small molecule drugs to inhibit protein-protein interactions

1800 Pairwise statistical analyses were performed using a Student’s t-test, and three-way comparisons using 1-way 1600 ANOVA, with statistical significance taken as p < 0.05. 2

1400 Results and Discussion

Tables 1 and 2 present the details of the 48 protein-protein 1200 complexes considered for analysis: Table 1 shows details for the 12 G-protein (OG) and 7 receptor extracellular domain complexes (OR) complexes, and Table 2 shows Molecular surface area / Å / area surface Molecular 1000 the details for the 29 enzyme inhibitor (EI) complexes. The interface areas of the PPIs (calculated as half 1 2 3 45 the difference in the surface areas of the complexed and Probe radius, r / Å uncomplexed protein partners), ranged from 476 Å2 to 1712 Å2, and were significantly greater, therefore, than Figure 2. PPI interfacial molecular surface area (Å2) of the those for protein-small molecule interactions, which uncomplexed form of ADP-ribosylation factor 1 (PDB code, 1J2J) average around 500 Å2 (Wells & McClendon, 2007). There as a function of probe radius, r (Å). The slope of the plot allows were modest but statistically significant differences found determination of the surface fractal dimension, D, in this case 2.25. for the interfacial areas of the OG and OR complexes versus the EI complexes, with the mean areas determined as 952 (± 415) Å2 and 875 (± 229) Å2, respectively. Log-log plots of the variation in PPI interface surface a very rough surface). area as a function of probe radius were in all cases Mean interfacial surface curvature and roughness linear (with r 2 ≥ 0.95; cf., Fig. 2), and allowed reliable values were only calculated for PPI protein partners determination of the interfacial surface irregularity/ whose structures in the unbound state were determined surface roughness, described by means of the associated by X-ray crystallography. In cases where the structure of fractal dimension (D). All the PPI interface surfaces were the unbound form was determined by NMR, the pairwise shown to be smooth (Table 3; Fig. 3), with the mean fractal comparisons between its bound and unbound species dimensions for their unbound protein components of 2.25, were not performed. 2.18 and 2.23 for OG, OR and EI complexes, respectively. The generic mean frequencies of amino acids across One-way ANOVA analysis for the variation in the fractal protein surfaces were taken from Miller et al., (1987),25 dimensions of the unbound components of the OG, OR and the frequencies of the residue types found within PPI and EI complexes showed no significant difference interfaces were computed using the PDBGoodies utility between the mean interfacial surface roughnesses of the (Hussain et al., 2002). three types of complex.

Figure 3

(a) (b)

Figure 3. Space-filling models showing the ligand binding domain of the ephrin type B receptor 4 (a; PDB code, 2HLE) and the catalytic domain of P120 gap protein (b; PDB code 1WQ1), with the PPI interface surfaces highlighted yellow. The interface surface of the ephrin domain is fairly smooth, with a fractal dimension of D = 2.25, and the P120 protein interface surface is somewhat more rough, with D = 2.46.

155

26 D. Ahmadi & D. Barlow / TPPS 2017 1(4) 153-166

Table 1. Data set of G-protein (OG) and receptor extracellular domain (OR) complexes with their unbound components and corresponding structural and interfacial properties. Complex Interface Protein partner 1 Number Protein partner 2 Number Resolution Resolution Resolution Class (PDB Area b PDB b of PDB b of 2 (Å) Name (Å) Name (Å) entry) (Å ) entry a residues entry a residues GTP-binding Nuclear 1QG4 1OUN OG 1A2K 1595 2.50 nuclear protein 2.50 22 transport 2.30 27 (D) (A) Ran factor 2 Ras-related Neutrophil 1MH1 1HH8 OG 1E96 590 2.40 C3 botulinum 1.38 18 cytosol factor 1.80 20 (A) (B) toxin substrate 2 Guanine nucleotide- Regulator 1TND binding 1FQI OG 1FQJ 899 2.02 2.20 24 of G-protein 1.94 28 (A) protein G(t) (B) signalling 9 subunit alpha-1 GTP-binding Regulator of 1QG4 1A12 OG 1I2M 1389 1.76 nuclear protein 2.50 34 chromosome 1.70 47 (A) (B) Ran condensation Ras-related 1I49 1MH1 C3 botulinum OG 1I4D 703 2.50 Arfaptin-2 2.80 19 1.38 20 (A) (D) toxin substrate 1 GTP-binding 1QG4 1F59 Importin OG 1IBR 1695 2.30 nuclear protein 2.50 49 2.80 54 (C) (D) subunit beta-1 Ran ADP- ADP- 1O3Y 1OXZ ribosylation OG 1J2J 604 1.60 ribosylation 1.50 18 1.50 14 (A) (B) factor-binding factor 1 protein GGA1 Ral guanine 5P21 1LXD nucleotide OG 1LFD 619 2.10 GTPase HRas 1.35 18 2.40 15 (D) (C) dissociation stimulator Son of sevenless 1LF0 2II0 OG 1NVU 1712 2.20 GTPase HRas 1.70 43 homolog 2.02 57 (R) (S) 1 variant (Fragment) Ras GTPase- 6Q21 1WER OG 1WQ1 1420 2.50 GTPase HRas 1.95 37 activating 1.60 45 (R) (G) protein 1 Ras-related 2BME 1YZM OG 1Z0K 927 1.92 protein Rab- 1.57 27 Rabenosyn-5 1.50 24 (C) (D) 4A Ras-related 2ZKM Phospholipase 1MH1 C3 botulinum OG 2FJU 626 2.20 1.62 19 1.38 14 (B) C-beta-2 (A) toxin substrate 1 2DTQ Ig gamma-1 1FNL Fc-gamma OR 1E4K 477 2.30 2.00 15 1.80 12 (B) chain C region (C) RIIIb 1BUY 1ERN Erythropoietin OR 1EER 975 1.90 Erythropoietin NMR 25 2.40 28 (A) (B) receptor Transforming TGF-beta 1TGK 1M9Z OR 1KTZ 493 2.15 growth factor 3.30 10 receptor 1.05 15 (A) (B) beta-3 type-2

156 Designing small molecule drugs to inhibit protein-protein interactions

Table 1. Continued.

Complex Interface Protein partner 1 Number Protein partner 2 Number Resolution Resolution Resolution Class (PDB Area b PDB b of PDB b of 2 (Å) Name (Å) Name (Å) entry) (Å ) entry a residues entry a residues 1ACC Protective 1SHU Anthrax toxin OR 1T6B 967 2.50 2.10 32 1.50 26 (X) antigen (Y) receptor 2 TNFR 13B 1U5Y TNF domain 1XUT OR 1XU1 646 1.90 0.96 32 TACI CRD2 NMR 14 (B) of APRIL (S) domain

2BBA Ephrin type-B 1IKO OR 2HLE 1052 2.05 1.65 37 Ephrin-B2 1.92 25 (A) receptor 4 (B) a) PDB entry codes of the protein-protein complexes or their unbound protein partners, with the chain identifiers for partners 1 and 2 given in parentheses. b) Where the 3-D structure of an unbound protein partner has been determined by NMR, this is indicated; otherwise, the resolution of the X-ray structure determination is given.

A comparison of the mean interfacial surface 4A (1Z0K) – are quite substantial. The mean absolute roughness of the bound and uncomplexed protein partners difference in surface curvatures for all the complexed and involved in OR and EI complexes showed no statistically uncomplexed species is 0.028. Figure 4 significant difference, but for the OG complexes, the FigureStatistical 4 comparisons of the frequencies of the interfacing surfaces of the partner proteins were found to 20 (a)differentOR complexes amino acid types across protein surfaces become a little less rough on complexation, with the mean in (a)generalOR complexes and their frequencies within PPI interfaces interfacial surface fractal dimension decreasing from D = revealed significant differences between OG, OR and EI 2.25 to D = 2.21 (p = 0.015). complexes, and the findings for the former two classes All the PPI interfaces studied were likewise of complex are shown graphically in Fig.4. Arginine and determined to be relatively flat, with their mean surface leucine show a statistically significant enrichment at the curvatures lying in the range, -0.12 Å-1 – 0.11 Å-1 (Table PPI interfaces of OG and OR complexes, whereas EI 4). For most of the proteins involved in these PPIs complexes show a significant enrichment of histidine, (92%), however, there are differences between the mean leucine and cysteine (data not shown). The residue types interfacial surface curvatures of their complexed and that are significantly underrepresented at PPI interfaces uncomplexed forms, some of which – like phospholipase tend to be those with small and/or polar side chains, C‬‫‭3 (2FJU), staphopain B (1PXV), and Ras-related Rab-‫ including alanine and glycine in the PPI interfaces of OR

Figure 4

Figure(a) OR4 complexes

(a) OR complexes (b) OG complexes (b) OG complexes

Figure 4. Percentage frequencies of amino acid residue types at protein surfaces and across PPI interfaces. Residue types showing a statistically significant(b) OG complexes enrichment across PPI interfaces are highlighted red; those showing statistically significant depletion at PPI interfaces are highlighted blue. Data relate to OR complexes (a) and OG complexes (b). (b) OG complexes 27 27

157

27 27 D. Ahmadi & D. Barlow / TPPS 2017 1(4) 153-166 of 17 20 20 28 81 30 45 20 20 14 24 35 31 31 18 30 Number residues b 2.50 2.76 2.00 2.68 1.75 1.40 1.55 1.55 1.03 1.40 (Å) NMR NMR NMR NMR NMR NMR Resolution A TEM Name Elafin Ecotin Barstar Eglin C inhibitor inhibitor Hirustasin Proteinase inhibitor 3 Colicin-E9 Cystatin-A Major pepsin Staphostatin B Alpha-amylase Beta-lactamase Cationic trypsin Trypsin inhibitor Trypsin immunity protein Protease inhibitor inhibitor 1 Trypsin Protein partner 2 a (I) (I) PDB entry 1LU0 (I) 2REL 1BX8 (I) 2RN4 (I) 1F32 (B) 1EGL 1A19 (D) 1PMC (I) 1ZG4 (A) 1BA7 (B) 2PTN (B) 1ECZ (A) 1HOE (T) 1IMQ (A) 1DVC (L) 1NYC (C) of 30 32 23 42 80 22 57 30 29 24 37 39 27 13 35 36 Number residues b 2.0 1.68 1.63 1.50 2.20 1.80 2.30 1.80 1.90 1.68 2.05 2.00 1.98 2.10 1.55 2.50 (Å) Resolution Name protein Trypsin Alkaline Pepsin A Colicin-E9 Ribonuclease Staphopain B Cationic trypsin Pro-cathepsin H Anionic trypsin-2 metalloproteinase Chymotrypsin-like Alpha-1-antitrypsin Glandular kallikrein Chymotrypsinogen A Chymotrypsinogen A Chymotrypsinogen Pancreatic alpha-amylase elastase family member 1 Beta-lactamase inhibitory Protein partner 1 a (P) (A) (E) (A) (A) 1PIG (P) 1FSJ (B) 9EST 4PEP (A) 4PEP 1A2P 2PTN (E) 1AKL 1QLP 2PKA (B) 2PKA 1X9Y 4CHA (E) 4CHA 8PCH (D) 4CHA (A) 4CHA 1QQU (A) 1TRM (D) 3GMU (B) PDB entry b 2.1 2.00 1.90 2.00 2.50 1.70 2.40 2.45 1.90 2.40 1.74 1.73 2.40 2.30 2.00 1.80 (Å) Resolution ) 2 770.8 792.1 778.5 766.8 934.0 883.5 799.1 631.2 894.1 677.0 842.2 1106.6 1164.1 1567.1 1004.1 1297.7 Interface Area (Å 1F34 1JIW 1JTG 1PPE 1HIA 1FLE 1GL1 1NB5 1BRS 1EZU 1OPH 1PXV 1AVX 1ACB 1BVN 1EMV Complex (PDB entry) Data set of enzyme inhibitor (EI) complexes with their unbound components and corresponding structural interfacial properties. Class EI EI EI EI EI EI EI EI EI EI EI EI EI EI EI EI Table 2. Table

158 Designing small molecule drugs to inhibit protein-protein interactions of 15 32 19 41 20 19 28 15 15 18 12 14 12 Number residues b 1.16 2.20 1.70 1.70 1.22 2.30 2.00 1.40 1.50 (Å) NMR NMR NMR NMR Resolution Name Metallo- Metallo- inhibitor inhibitor inhibitor inhibitor Chagasin Subtilisin- inhibitor 1 glycosylase Ovomucoid Ovomucoid inhibitor-2A Uracil-DNA chymotrypsin carboxypeptidase carboxypeptidase nuclease inhibitor Tryptase inhibitor Tryptase Pancreatic trypsin Metalloproteinase Xylanase inhibitor Subtilisin inhibitor Sugar-non-specific Protein partner 2 a PDB entry 3SSI (I) 9PTI (I) 2CI2 (I) 1J57 (B) 1H20 (J) 1ZFI (C) 1T6E (A) 2GKR (I) 2UGI (D) 2OVO (I) 1D2B (E) 2NNR (B) 2UUX (B) of 31 31 26 43 29 31 33 31 33 30 31 27 24 Number residues b 1.90 1.43 1.39 1.40 1.90 1.90 2.90 1.55 1.60 2.40 1.55 1.18 2.00 (Å) Resolution A1 A1 glycosylase Name Nuclease Xylanase Falcipain 2 Streptogrisin-B Subtilisin BPN' Subtilisin BPN' Cationic trypsin Cationic trypsin Subtilisin Carlsberg Carboxypeptidase Carboxypeptidase Interstitial collagenase Uracil-DNA Protein partner 1 a (B) (E) 3I1U (A) 966C (B) 1SUP 8CPA (B) 8CPA 2PTN (E) 3BPF (A) 2QA9 (E) 1SCN (E) 2PTN (A) 3EUG (C) 1ZM8 (A) 1UBN (E) 2DCY PDB entry b 1.10 2.40 2.16 2.50 2.54 2.30 2.20 1.90 1.80 2.10 1.15 1.80 2.50 (Å) Resolution ) 2 705.3 712.6 774.8 815.7 966.9 714.3 806.9 813.9 637.2 633.2 558.7 1056.7 1259.9 Interface Area (Å 2J0T 2SIC 2SNI 2B42 1R0R 2PTC 4CPA 2O3B 3SGB 2ABZ 2OUL 1UUG 2UUY Complex (PDB entry) Continued. Where the 3-D structure of an unbound protein partner has been determined by NMR, this is indicated; otherwise, resolution X-ray determination given. Class EI EI EI EI EI EI EI EI EI EI EI EI EI Table 2. Table a) PDB entry codes of the protein-protein complexes or their unbound protein partners, with chain identifiers for partners 1 and 2 given in parentheses. b)

159 D. Ahmadi & D. Barlow / TPPS 2017 1(4) 153-166

complexes, alanine and aspartic acid in EI complexes weight PPI inhibitors of high affinity can be successfully (data not shown), and alanine, glycine, and serine in OG achieved provided they are targeted to bind to the complexes. interfaces’ key ‘hot spot’ residues (Bogan and Thorn, Previous systematic analyses of PPIs have sometimes 1998; Wells and McClendon, 2007; Thangudu et al., included consideration of complexes whose crystal 2012; Cukuroglu et al., 2014). It should be noted though structures were determined to resolutions > 2.5 Å (Jones that even if this is so, the resulting enthalpic contribution and Thornton, 1996; Conte et al., 1999), and the dataset to the free energy of complexation is very unlikely to selected for study here was deliberately restricted to compensate the fact that van der Waals interactions will those whose structures were refined at medium to high be far fewer than for similar-sized ligands binding within resolution, in order to give greater confidence in atomic buried pockets, and unlikely also to compensate for the co-ordinates and any quantities derived from these. Many minimal contribution made to binding affinity by ligand of the earlier analyses of PPIs also considered a wider desolvation. spectrum of interaction types than was considered here To further compound the issue, we find that virtually – including those between the sub-units of oligomeric all the proteins involved in the PPIs studied here undergo proteins, and those between antigens and antibodies changes in their mean interfacial surface curvature upon (Jones and Thornton, 1996; Conte et al., 1999). Given that complexation, and these changes – taken together with these PPIs interfaces are of limited relevance as regards the changes in interface residue conformation shown by the design of (intracellularly acting) anticancer drugs, Kastritis et al. (2011) – would pose problems not only they too were excluded from consideration in the studies for the identification of so-called ‘hot spot’ residues but reported here. problems too for the quantitative assessment of putative Notwithstanding these restrictions to the dataset, ligand binding affinities. however, and in full accord with the observations recorded by previous workers, we find that the interfaces Conclusions of the PPIs included in our dataset cover areas around 5 to 10 times larger than those seen in complexes involving We thus note that while various reviews have reported proteins and low molecular ligands. Moreover, these recent good progress in the design of PPI inhibitors interfaces are seen to have surfaces that are quite smooth (Corbi-Verge and Kim, 2016), with successes cited where and flat, with very little change in the roughness of the molecules have been taken into clinical trials (Nero et al., relevant surfaces of the protein partners in their transition 2014; Laraia et al., 2015), we note too that the molecular from their unbound to bound states. weights of these compounds are still quite high (typically Targeting of these interfaces using non-protein ligands around 1000 to 1500 Da), and all of the compounds that bind across the whole interface area would thus (assuming they are even successful in trials) are thus require compounds of relatively high molecular weight, likely to be formulated for parenteral (and not oral) use. which would then preclude their administration via the The large size of these compounds can also be expected to oral route (given that such drugs would ideally have result in them having a rather low membrane permeability, molecular weights < 500 Da (Lipinski et al., 2001)). In and their high numbers of hydrogen bond donor and addition, because the PPI interfaces are so smooth (and acceptor groups may mean that they will likely satisfy the also quite flat), these high molecular weight ligands pharmacophores of a great number of off-target proteins, would of necessity bind to the target protein surface leading to innumerable side-effects. Moreover, many of with around half of their structure remaining exposed these compounds are shown to bind as allosteric (rather to the surrounding solvent, and this would greatly limit than orthosteric) inhibitors, and so their inhibitory activity the entropic contribution to the free energy of their is explained not by their binding at the PPI interface but complexation provided through desolvation. While such by their binding elsewhere on the protein surface and a deficiency might be mitigated if the PPI interfaces thereby triggering a conformational change in the protein, involved significant numbers of charged and polar which then indirectly interferes with the protein-protein groups which would allow for extensive electrostatic and interaction. Taken in conjunction with the observations hydrogen bond interactions, our analysis of the amino that are recorded here, therefore, the prospects for acid composition of the PPI interface surfaces shows that successful design of orthosteric PPI inhibitors for use in this is not generally the case. Most of the PPI interface cancer therapy would seem very limited. surfaces are depleted in small polar residues (like serine, threonine), and they are also often deficient in the charged Acknowledgements residues, lysine and glutamic acid, although the interface surfaces in OG and OR complexes are enriched in The authors gratefully acknowledge helpful discussions arginine, and EI complexes are enriched in histidine. with Dr Oleg Tsodikov (University of Kentucky, USA), It has been proposed, however, that low molecular in respect of his SurfRace program.

160 Designing small molecule drugs to inhibit protein-protein interactions

Table 3. Interfacial surface roughness/fractal dimension (D) of PPI protein partners.

Fractal Dimension (D) Class PDB code (chain id) Component Unbound Bound

EI 1ACB (E) Chymotrypsinogen A 2.27 2.41

EI 1AVX (A) Trypsin 2.34 2.36

EI 1AVX (B) Trypsin inhibitor A 2.17 2.16

EI 1BRS (A) Ribonuclease 2.21 2.22

EI 1BVN (P) Pancreatic alpha-amylase 2.39 2.30

EI 1EMV (B) Colicin-E9 2.14 2.15

EI 1EZU (D) Anionic trypsin-2 2.31 2.27

EI 1F34 (A) Pepsin A 2.18 2.33

EI 1F34 (B) Major pepsin inhibitor 3 2.23 2.19

EI 1GL1 (A) Chymotrypsinogen A 2.23 2.31

EI 1HIA (B) Glandular Kallikrein 2.37 2.41

EI 1HIA (I) Hirustasin 2.18 2.17

Alkaline EI 1JIW (P) 2.33 2.37 metalloproteinase

EI 1JTG (A) Beta-lactamase TEM 2.15 2.19

Beta-lactamase inhibitory EI 1JTG (B) 2.19 2.27 protein

EI 1NB5 (D) Pro-cathepsin H 2.16 2.17

EI 1OPH (A) Alpha-1-antitrypsin 2.13 2.17

EI 1OPH (B) Cationic trypsin 2.24 2.35

EI 1PPE (E) Cationic trypsin 2.31 2.37

EI 1PPE (I) Trypsin inhibitor 1 2.19 2.14

EI 1PXV (A) Staphopain B 2.50 2.18

EI 1PXV (C) Staphostatin B 2.15 2.16

EI 1R0R (E) Subtilisin Carlsberg 2.22 2.22

EI 1R0R (I) Ovomucoid 2.13 2.13

EI 1UUG (C) Uracil-DNA glycosylase 2.17 2.23

Uracil-DNA glycosylate EI 1UUG (D) 2.20 2.25 inhibtor

EI 2ABZ (A) Carboxypeptidase A1 2.29 2.30

Metallocarboxypeptidase EI 2ABZ (C) 2.21 2.19 inhibitor

EI 2B42 (A) Xylanase inhibitor 2.22 2.30

EI 2B42 (B) Xylanase 2.23 2.45

EI 2J0T (B) Interstitial collagenase 2.24 2.25

EI 2O3B (A) Nuclease 2.19 2.16

EI 2OUL (B) Chagasin 2.19 2.18

161 D. Ahmadi & D. Barlow / TPPS 2017 1(4) 153-166

Table 3. Continued. Fractal Dimension (D) Class PDB code (chain id) Component Unbound Bound

Panctreatic trypsin 33 EI 2PTC (I) 2.11 2.12 inhibitor

EI 2SIC (E) Subtilisin BPN’ 2.27 2.32

EI 2SIC (I) Subtilisin inhibitor 2.17 2.09

EI 2SNI (E) Subtilisin BPN’ 2.37 2.30

Subtilisin-chymotrypsin EI 2SNI (I) 2.13 2.13 inhibitor 2A

EI 2UUY (A) Cationic trypsin 2.23 2.33

EI 2UUY (B) Tryptase inhibitor 2.08 2.10

EI 3SGB (E) Streptogrisin-B 2.24 2.24

EI 3SGB (I) Ovomucoid 2.12 2.09

EI 4CPA (B) Carboxypeptidase 2.36 2.38

OG 1A2K (A) Nuclear transport factor 2 2.34 2.32

GTP-binding nuclear OG 1A2K (D) 2.28 2.20 protein Ran

Ras-related C3 botulinum OG 1E96 (A) 2.14 2.17 toxin

Neutrophil cytosol factor OG 1E96 (B) 2.26 2.19 2

Guanine nucleotide- OG 1FQJ (A) binding protein G(t) 2.22 2.16 subunit alpha- 1

Regulator of G-protein OG 1FQJ (B) 2.17 2.24 signalling 9

GTP-binding nuclear OG 1I2M (A) 2.23 2.26 protein Ran

Regulator of chromosome OG 1I2M (B) 2.29 2.22 condensation

Ras-related C3 botulinum OG 1I4D (D) 2.29 2.14 toxin substrate 1

GTP-binding nuclear OG 1IBR (C) 2.28 2.13 protein Ran

OG 1J2J (A) ADP-ribosylation factor 1 2.21 2.26

ADP-ribosylation factor- OG 1J2J (B) 2.15 2.20 binding protein GGA1

Ral guanine nucleotide OG 1LFD (C) 2.15 2.12 dissociation stimulator

OG 1LFD (D) GTPase HRas 2.27 2.28

Son of sevenless homolog OG 1NVU (S) 2.30 2.34 1 variant (fragment)

OG 1NVU (R) GTPase HRas 2.30 2.19

162 Designing small molecule drugs to inhibit protein-protein interactions

Table 3. Continued. Fractal Dimension (D) Class PDB code (chain id) Component Unbound Bound

Ras GTPase- activating OG 1WQ1 (G) 2.46 2.32 protein 1

Ras-related protein Rab- OG 1Z0K (C) 2.24 2.15 4A

OG 1Z0K (D) Rabenosyn-5 2.19 2.10

Ras-related C3 botulinum OG 2FJU (A) 2.29 2.23 toxin substrate 1

OG 2FJU (B) Phospholipase C- beta- 2 2.15 2.16

Ig gamma-1 chain C OR 1E4K (B) 2.20 2.16 rgeion

OR 1E4K (C) Fc- gamma RIIIb 2.17 2.15

OR 1EER (B) Erythropoietin receptor 2.23 2.15

OR 1KTZ (B) TGF- beta receptor type-2 2.11 2.15

Interleukin-6 receptor OR 1PVH (A) 2.22 2.22 subunit beta

OR 1T6B (X) Protective antigen 2.18 2.23

OR 1T6B (Y) Anthrax toxin receptor 2 2.17 2.17

OR 1XU1 (B) TNF domain of April 2.21 2.17

OR 2HLE (A) Ephrin type-B receptor 4 2.20 2.25

OR 2HLE (B) Ephrin-B2 2.17 2.21

Table 4. Mean interfacial surface curvatures (Å-1) of PPI proteins. Curvature (Å-1) Curvature (Å-1) Class PDB entry (chain id) Complex Unbound Bound

EI 1AVX (A) Trypsin 0.04 0.07

EI 1AVX (B) Trypsin inhibitor A -0.02 -0.02

EI 1F34 (A) Pepsin A 0.03 0.03

EI 1F34 (B) Major pepsin inhibitor 3 -0.05 -0.03

EI 1HIA (B) Glandular Kallikrein 0.11 0.08

EI 1HIA (I) Hirustasin -0.05 -0.08

EI 1JTG (A) Beta-lactamase TEM 0.03 0.07

Beta-lactamase inhibitory EI 1JTG (B) 0.00 0.01 protein

EI 1OPH (A) Alpha-1-antitrypsin -0.09 -0.10

EI 1OPH (B) Cationic trypsin 0.06 0.05

EI 1PPE (E) Cationic trypsin 0.05 0.06

EI 1PPE (I) Trypsin inhibitor 1 -0.07 -0.06

163 D. Ahmadi & D. Barlow / TPPS 2017 1(4) 153-166

Table 4. Continued. Curvature (Å-1) Curvature (Å-1) Class PDB entry (chain id) Complex Unbound Bound

EI 1PXV (C) Staphostatin B -0.09 -0.05

EI 1R0R (E) Subtilisin Carlsberg 0.03 0.01

EI 1UUG (C) Uracil-DNA glycosylase -0.03 -0.06

Uracil-DNA glycosylate EI 1UUG (D) 0.02 0.00 inhibitor

EI 2ABZ (A) Carboxypeptidase A1 0.05 0.06

Metallocarboxypeptidase EI 2ABZ (C) -0.02 -0.04 inhibitor

EI 2B42 (A) Xylanase inhibitor 0.01 0.01

EI 2B42 (B) Xylanase 0.00 0.01

EI 2PTC (E) Cationic trypsin 0.04 0.07

Panctreatic trypsin EI 2PTC (I) -0.06 -0.05 inhibitor

EI 2SIC (E) Subtilisin BPN’ 0.04 0.02

EI 2SIC (I) Subtilisin inhibitor -0.07 -0.09

EI 2SNI (E) Subtilisin BPN’ 0.07 0.04

Subtilisin-chymotrypsin EI 2SNI (I) -0.07 -0.06 inhibitor 2A

EI 2UUY (A) Cationic trypsin 0.09 0.06

EI 2UUY (B) Tryptase inhibitor -0.12 -0.10

EI 3SGB (E) Streptogrisin-B 0.02 -0.02

EI 3SGB (I) Ovomucoid -0.11 -0.08

OG 1A2K (A) Nuclear transport factor 2 0.06 0.02

GTP-binding nuclear OG 1A2K (D) 0.01 0.02 protein Ran

Ras-related C3 botulinum OG 1E96 (A) -0.06 -0.04 toxin

Neutrophil cytosol factor OG 1E96 (B) 0.06 0.04 2

Guanine nucleotide- OG 1FQJ (A) binding protein G(t) -0.03 -0.02 subunit alpha- 1

Regulator of G-protein OG 1FQJ (B) -0.05 -0.06 signalling 9

GTP-binding nuclear OG 1I2M (A) 0.01 -0.01 protein Ran

Regulator of chromosome OG 1I2M (B) 0.04 -0.04 condensation

Ras-related C3 botulinum OG 1I4D (D) -0.09 -0.04 toxin substrate 1

OG 1J2J (A) ADP-ribosylation factor 1 -0.01 0.04

164 Designing small molecule drugs to inhibit protein-protein interactions

Table 4. Continued.

Curvature (Å-1) Curvature (Å-1) Class PDB entry (chain id) Complex Unbound Bound

Ral guanine nucleotide OG 1LFD (C) -0.08 -0.05 dissociation stimulator

OG 1LFD (D) GTPase HRas 0.04 0.00

Son of sevenless homolog OG 1NVU (S) 0.03 0.04 1 variant (fragment)

OG 1NVU (R) GTPase HRas 0.03 -0.01

OG 1WQ1 (R) GTPase HRas -0.03 -0.04

Ras GTPase- activating OG 1WQ1 (G) 0.02 0.05 protein 1

OG 1Z0K (D) Rabenosyn-5 -0.06 -0.07

Ras-related C3 botulinum OG 2FJU (A) -0.22 -0.06 toxin substrate 1

OG 2FJU (B) Phospholipase C- beta- 2 -0.05 0.08

Ig gamma-1 chain C OR 1E4K (B) -0.05 -0.06 rgeion

OR 1E4K (C) Fc- gamma RIIIb -0.03 -0.03

OR 1T6B (X) Protective antigen -0.03 0.00

OR 1T6B (Y) Anthrax toxin receptor 2 -0.02 -0.03

OR 2HLE (A) Ephrin type-B receptor 4 -0.01 0.04

OR 2HLE (B) Ephrin-B2 -0.05 -0.04

E., Conroy, M. J., Dana, J. M., Fernandez Montecelo, M. A., van Ginkel, Competing Interests G., Gore, S. P. and P. Haslam, (2013). “PDBe: Protein Data Bank in Europe.” Nucleic Acids Research. 42(1): 285-291. The authors declare that they have no competing financial, Hussain, A., Shanthi, V., Sheik, S., Jeyakanthan, J., Selvarani, P. and direct or indirect interests in connection with the work K. Sekar, (2002). “PDB Goodies – a web-based GUI to manipulate the reported here. Protein Data Bank file.” Acta Crystallographica Section D Biological Crystallography, 58(8): 1385-1386. Jones, S., and J. M. Thornton, (1996). “Principles of protein-protein References interactions.” Proceedings of the National Academy of Sciences of the United States of America, 93(1): 13–20. Arkin, M. and J. Wells, (2004). “Small-molecule inhibitors of Kastritis, P., Moal, I., Hwang, H., Weng, Z., Bates, P., Bonvin, A. protein–protein interactions: progressing towards the dream.” Nature and J. Janin, (2011). “A structure-based benchmark for protein-protein Reviews Drug Discovery, 3(4): 301-317. binding affinity.” Protein Science, 20(3): 482-491. Bogan, A. and K. Thorn, (1998). “Anatomy of hot spots in protein Krissinel, E. and K. Henrick, (2007). “Inference of Macromolecular interfaces.” Journal of Molecular Biology, 280(1): 1-9. Assemblies from Crystalline State.” Journal of Molecular Biology, Chothia, C. (1976). “The nature of the accessible and buried 372(3): 774-797. surfaces in proteins.” Journal of Molecular Biology, 105(1): 1-12. Laraia, L., McKenzie, G., Spring, D., Venkitaraman, A. and Conte, L., Chothia, C. and J. Janin, (1999). “The atomic structure D. Huggins, (2015). “Overcoming Chemical, Biological, and of protein-protein recognition sites.” Journal of Molecular Biology, Computational Challenges in the Development of Inhibitors Targeting 285(5): 2177-2198. Protein-Protein Interactions.” Chemistry & Biology, 22(6): 689-703. Corbi-Verge, C. and P. Kim, (2016). “Motif mediated protein- Lipinski, C., Lombardo, F., Dominy, B. and P. Feeney, (2001). protein interactions as drug targets.” Cell Communication and Signaling, “Experimental and computational approaches to estimate solubility and , 14(8): 1-12. permeability in drug discovery and development settings.” Advanced Gutmanas, A., Alhroub, Y., Battle, G. M., Berrisford, J. M., Bochet, Drug Delivery Reviews, 46(1-3): 3-26.

165 D. Ahmadi & D. Barlow / TPPS 2017 1(4) 153-166

Lewis, M. and D. Rees, (1985). “Fractal surfaces of proteins.” Chemistry, 45(8): 1543-1558. Science, 230: 1163 ‑ 1165. Tsodikov, O., Record, M. and Y. Sergeev, (2002). “Novel computer Miller, S., Janin, J., Lesk, A. and C. Chothia, (1987). “Interior and program for fast exact calculation of accessible and molecular surface surface of monomeric proteins” Journal of Molecular Biology, 196: 641 areas and average surface curvature.” Journal of Computational ‑ 656. Chemistry, 23(6): 600-609. Nero, T., Morton, C., Holien, J., Wielens, J. and M. Parker, (2014). Vassilev, L. (2004). “In Vivo Activation of the p53 Pathway by “Oncogenic protein interfaces: small molecules, big challenges.” Nature Small-Molecule Antagonists of MDM2.” Science, 303(5659): 844-848. Reviews Cancer, 14(4): 248-262. Wells, J., and C. McClendon, (2007). “Reaching for high-hanging Raj, M., Bullock, B. and P. Arora, (2013). “Plucking the high fruit in drug discovery at protein–protein interfaces.” Nature, 450(7172): hanging fruit: A systematic approach for targeting protein–protein 1001-1009. interactions.” Bioorganic & Medicinal Chemistry, 21(14): 4051-4057. White, A., Westwell, A. and G. Brahemi, (2008). “Protein–protein Thangudu, R., Bryant, S., Panchenko, A. and T. Madej, (2012). interactions as targets for small-molecule therapeutics in cancer.” “Modulating Protein–Protein Interactions with Small Molecules: The Expert Reviews in Molecular Medicine, 10: e8. DOI: 10.1017/ Importance of Binding Hotspots.” Journal of Molecular Biology, S1462399408000641. 415(2): 443-453. Toogood, P. (2002). “Inhibition of Protein−Protein Association by This open-access article distributed under the terms of the Creative Commons Small Molecules: Approaches and Progress.” Journal of Medicinal Attribution-NonCommercial 4.0 International License (CC BY-NC 4.0).

166