Article pubs.acs.org/jmc

Structure-Based Development of a Protein−Protein Interaction Inhibitor Targeting Tumor Necrosis Factor Receptor-Associated Factor 6 Jun Moriya,† Koh Takeuchi,‡ Kenji Tai,† Kenzo Arai,† Naoki Kobayashi,† Naoki Yoneda,† Yoshifumi Fukunishi,‡ Atsushi Inoue,† Miho Kihara,† Takumi Murakami,§ Kenichi Chiba,† and Ichio Shimada*,∥ †

Eisai Product Creation Systems, Eisai Co., Ltd., Tokodai 5-1-3, Tsukuba-shi, Ibaraki 300-2635, Japan Biological Information Research Center and Molecular Profiling Research Center for Drug Discovery, National Institute of Advanced Industrial Science and Technology, Aomi 2-3-26, Koto-ku, Tokyo 135-0064, Japan § Pharmacological Evaluation Unit, Tsukuba Division, Sunplanet Co., Ltd., Tokodai 5-11-1, Tsukuba-shi, Ibaraki 300-2635, Japan ∥ Graduate School of Pharmaceutical Sciences, The University of Tokyo, Hongo 7-3-1, Bunkyo-ku, Tokyo 113-0033, Japan ‡

S Supporting Information *

ABSTRACT: The interactions between tumor necrosis factor (TNF) receptor-associated factor 6 (TRAF6) and TNF superfamily receptors (TNFRSFs) are promising targets for rheumatoid arthritis (RA) treatment. However, due to the challenging nature of protein− protein interactions (PPIs), a potent inhibitor that surpasses the affinity of the TRAF6− TNFRSF interactions has not been developed. We developed a small-molecule PPI inhibitor of TRAF6−TNFRSF interactions using NMR and in silico techniques. The most potent compound, TRI4, exhibited an affinity higher than those of TNFRSFs and competitively inhibited a TRAF6−TNFRSF interaction. Structural characterization of the TRAF6−TRI4 complex revealed that TRI4 supplants key interactions in the TRAF6− TNFRSF interfaces. In addition, some TRAF6−TRI4 interactions extend beyond the TRAF6−TNFRSF interfaces and increase the binding affinity. Our successful development of TRI4 provides a new opportunity for RA treatment and implications for structure-guided development of PPI inhibitors.



INTRODUCTION Protein−protein interactions (PPIs) are fundamental to cellular functions. Thus, PPIs represent an appealing target for development of small-molecule therapeutics.1−3 However, the generally flat PPI surfaces make design of small-molecule inhibitors quite challenging. In addition, unlike conventional binding sites for small-molecule inhibitors, such as catalytic pockets of enzymes, natural small-molecule ligands that can be used as starting points for inhibitor development rarely exist for PPIs. Tumor necrosis factor (TNF) receptor-associated factor 6 (TRAF6) is a cytoplasmic protein that mediates signaling from the TNF superfamily receptors (TNFRSFs).4,5 Binding between TRAF6 and the intracellular sequence of TNFRSFs occurs via the C-terminal domain (C-domain) of TRAF6. Molecules downstream of the interleukin-1 (IL-1) receptor, called IL-1 receptor-associated kinases (IRAKs), utilize the same conserved Pro-x-Glu-x-x-(aromatic/acidic) motif to bind to the C-domain of TRAF6.5 These interactions activate nuclear factor-κB (NF-κB) family and activator protein 1 (AP1) family transcription factors (TFs), triggering immune and inflammatory responses and modulating bone homeostasis.6 These characteristics make the C-domain of TRAF6 a © XXXX American Chemical Society

promising target for treatment of rheumatoid arthritis (RA).5,7 Therefore, inhibiting these upstream TRAF6 Cdomain interactions is a promising strategy for comprehensive suppression of the signaling pathways.8 In contrast, downstream TRAF6 signaling occurs through multiple pathways,9,10 and simultaneous inhibition by one specific molecule would be difficult. The crystal structures of TRAF6 in complex with two TNFRSFs, the receptor activator of NF-κB (RANK) and cluster of differentiation 40 (CD40), revealed that the TNFRSF-binding interface in TRAF6 is located in a shallow trench at the side of the molecule (Figure 1a). Although the binding affinities between TRAF6 and TNFRSFs are rather weak (dissociation constant (KD) ∼100 μM),11 several key interactions can be identified in the TRAF6−TNFRSF interfaces, and those key interactions are sterically configured both horizontally and perpendicularly in the interfaces. Extensive side chain interactions between TRAF6 and the conserved Pro-x-Glu-x-x-(aromatic/acidic) motif of TNFRSFs are horizontally extended to span the PPI surface, while the Received: May 22, 2015

A

DOI: 10.1021/acs.jmedchem.5b00778 J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

Figure 1. Structure and key interactions of the TRAF6−RANK complex. (a) Ribbon and molecular surface representations of the X-ray structure of TRAF6 (gray) in complex with the RANK peptide (cyan sticks) (PDB code 1LB5). The interacting residues of TRAF6 are shown as yellow sticks. Red dashed lines indicate main chain−main chain hydrogen bonds, while blue dashed lines indicate other hydrogen bonds and/or electrostatic interactions. (b) Schematic representation of interactions between TRAF6 and the RANK peptide. Red and blue dashed lines correspond to those in (a). van der Waals contacts are indicated by orange curves. Only the conserved motif in the RANK peptide (PTEDEY) is shown.

approach reportedly achieves a higher probability to pick up genuine hit compounds than the conventional docking approaches that use only one target protein.13 Starting from one million compounds in LigandBox,14,15 20 000 compounds were selected through the MSM-MTS strategy. Subsequently, the 20 000 compounds were evaluated by another in silico docking program, Glide (Schrödinger, Inc.), to cross-validate the result with a different program and to narrow down the number of hit compounds to 2000. The number of candidate compounds was further reduced to 682 on the basis of commercial availability and drug-likeness while retaining structural diversity. The 682 in silico hits and their analogue compounds were subjected to follow-up competition assays, using surface plasmon resonance (SPR) and osteoclast differentiation of mouse bone marrow cells to verify the inhibitory activity of the compounds in vitro and in a cellular condition, respectively. We identified four compounds with clear inhibitory activity in both assays (Figure S1, Supporting Information). Notably, these compounds did not show cell toxicity in the osteoclast differentiation assay. These four hit compounds induced chemical shift perturbations (CSPs) to TRAF6 NMR signals. One of them was classified into pan assay interference compounds (PAINS),16 and thus was excluded from further evaluations. Three other hit compounds did not show notable differences regarding calculated membrane permeability, solubility, hERG inhibition, sodium channel inhibition, unbound fraction in human plasma, and microsomal stability. However, among these three compounds, TRI1 was the only compound that did not cause aggregation of TRAF6 in the NMR titration experiments. Sufficient solubility of the TRAF6−compound complex is necessary to obtain a clear structure−activity relationship and to conduct compound optimization based on the complex structure. Thus, despite less inhibitory activity in the SPR and osteoclast differentiation assays (Figure S1), we selected TRI1 for further optimization. The residues with large CSPs upon the addition of TRI1 were located in the TNFRSF-binding site (Figure 2a−c), and TRI1 was competitively displaced from TRAF6 by the RANK peptide (Figure 2d). These data experimentally support a model in which TRI1 and TNFRSF share a binding site. It should be noted that the scaffold of TRI1 seems to have limited

main chain−main chain hydrogen bonds are oriented perpendicularly along with the exposed β-strand (β7) of TRAF6, which is located underneath the TRAF6−TNFRSF interfaces (Figure 1). Furthermore, biochemical experiments revealed that the TRAF6−TNFRSF interactions are severely impaired by point mutations at Arg-392, Phe-471, and Tyr-473 of TRAF6.11 Arg-392 electrostatically interacts with an acidic side chain (Asp) in RANK, and Phe-471 and Tyr-473 form a hydrophobic pocket to accommodate the conserved Pro in TNFRSFs.11 Despite the availability of these structures and biochemical data, a potent PPI inhibitor that surpasses the affinities of the TRAF6−TNFRSF interactions has not been developed. Given the flat shape and weak affinities for the endogenous peptides, the TNFRSF interfaces of TRAF6 seem to be challenging for drug development. Herein, we report the development of a potent smallmolecule PPI inhibitor of TRAF6−TNFRSF interactions, TRI4. TRI4 was developed through structure-based approaches and has a higher affinity for TRAF6 than TNFRSFs. Structural characterization of the TRAF6−TRI4 complex by a combination strategy using solution NMR and in silico techniques revealed that TRI4 supplants several key interactions in the TRAF6−TNFRSF interfaces and forms unique interactions that are extended beyond the TRAF6−TNFRSF interfaces. Those interactions result in its high affinity for TRAF6 and may contribute to the specificity of TRI4 for TRAF6 among other TRAF family proteins.



RESULTS AND DISCUSSION To obtain initial compounds with structures similar to that of an endogenous binding partner, the RANK peptide, we adopted machine-learning score modification multiple-target screening (MSM-MTS) as an initial screening method.12,13 In the MSM strategy, the optimal matrix, which improves the docking score of known active molecules (RANK peptide) against the target protein (TRAF6), is used to increase the chance to have compounds with interaction features similar to those of the known active molecule. The MTS approach examines the order of docking scores of tested compounds against the target protein and 181 standard proteins. Compounds that have higher scores with the target protein than standard proteins are selected as hit compounds. This B

DOI: 10.1021/acs.jmedchem.5b00778 J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

Figure 2. NMR interaction analysis between TRAF6 and TRI1. (a) 1H−15N HSQC spectra of 50 μM TRAF6 in the absence (black) and presence (red) of 1 mM TRI1. (b) Chemical shift perturbations (CSPs) induced in the backbone amide groups of TRAF6 by TRI1 binding. Twenty residues with resonances experiencing the largest CSPs upon addition of TRI1 are shown in red. (c) Distribution of the significantly affected residues upon the addition of TRI1 in the structure of TRAF6 (left). The 20 residues with the largest CSPs upon the addition of TRI1 are shown in red. The structure is in the same direction as the complex of TRAF6−RANK peptide (PDB code 1LB5) (right). The RANK peptide is shown as cyan sticks. (d) Competition between TRI1 and the RANK peptide on the surface of TRAF6. The signals in the 1H 1D spectrum of 20 μM TRI1 (black) were shifted by the addition of 20 μM uniformly 2H-labeled TRAF6 (red). The signals of TRI1 were reverted to the original position upon the addition of 50 μM RANK peptide (blue).

On the basis of the observed binding mode, other key interactions in the TRAF6−TNFRSF interfaces can be incorporated into the compound by extending the carboxyl group of TRI2 (Figure 3c). Two hundred derivatives of TRI2 were explored to find the optimal modification to the carboxyl group. As a result, the modification with a methanesulfonamide group (TRI3; Figure 4) or a ((3-chloro-4-methoxyphenyl)sulfonyl)benzamide group (TRI4; Figure 4) greatly improved the affinities of the compounds. Especially, the affinity of TRI4 (KD = 35 μM) was higher than those of any other evaluated compounds and the RANK peptide (KD = 78 μM). The inhibition activity in the osteoclast differentiation assay was also improved, and 40% inhibition was achieved by 150 μM TRI4. It should be mentioned that there are discrepancies between IC50 in the cell-based assay and the KD values of the TRI compounds. Off-target effects might potentiate the activity in the cell-based assay (IC50) of TRI1 compared to its KD value.

promiscuity as TRI1 and its derivative were active in only 1 out of 753 assays that were published in the PubChem database. A complex model structure with TRI1 docked in the TRAF6−TNFRSF interface showed that TRI1 forms a hydrogen bond to the main chain carbonyl group of Gly-470 of TRAF6 (Figure 3a,b). This interaction is also observed in the TRAF6−RANK interaction (Figure 1). In addition, the benzothiazolone ring (ring-1) occupies the binding site for the conserved Pro residue in the Pro-x-Glu-x-x-(aromatic/ acidic) motif (Figure 3). The Pro site is essential for the interaction between TRAF6 and the RANK peptide (Figure 1), as mutations to Phe-471 and Tyr-473 abolish the TRAF6− RANK interaction.11 The introduction of chlorine but not bulkier substituents to the fifth position of ring-1 improved the affinity of the compound (TRI2; Figure 4), indicating that ring1 with a small chlorine atom fits well in the Pro site. C

DOI: 10.1021/acs.jmedchem.5b00778 J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

Figure 4. Chemical structures and affinities of TRI compounds. Chemical structures of TRI1−4 and KD values derived from NMR titration experiments are shown.

concentrations used in our assays (Figure S3, Supporting Information), indicating that our data are not affected by undesirable effects originated from compound aggregations. Titration of TRI4 into TRAF6 induced substantial CSPs of the TRAF6 resonances, and some of the TRAF6 resonances disappeared upon TRI4 titration (Figure 5a,b). Mapping of the perturbed resonances indicated that TRI4 binds to the same site as TNFRSFs (Figure 5c). Furthermore, some of the perturbed residues converged in the β-strands behind the TRAF6−TNFRSF interfaces (Figure 5c). The significant perturbations did not indicate the presence of a second binding site because a mutation introduced into the region (E498A; Figure 5c) did not change the affinity between TRAF6 and TRI4. In addition, a linear correlation between the CSPs and the signal intensity reductions indicated the presence of single free-bound chemical exchange on a fast-to-intermediate time scale (Figure 5d); there was no indication of binding with slower conformational exchange, which would cause additional broadening of the TRAF6 resonances. The overall rotational correlation time of TRAF6 did not change upon TRI4 binding. Our initial attempts to crystallize the TRAF6−TRI4 complex and to soak TRI4 into free TRAF6 crystals were not successful, most likely due to a strong tendency of TRAF6 to crystallize without the compound and/or overlap between the TRAF6 crystal contact surface and the TRI4 binding site. Therefore, taking advantage of the ability to characterize protein−smallmolecule complexes in solution, we utilized NMR spectroscopy to obtain the TRAF6−TRI4 complex model. Since TRI4 and the RANK peptide compete for binding to TRAF6, we performed an INPHARMA experiment.17,18 In this experiment, indirect nuclear Overhauser effects (NOEs), INPHARMA NOEs, between two competitive ligands indicate overlapping binding moieties. A ternary mixture of TRI4, RANK peptide, and a substoichiometric amount of TRAF6 was used to obtain the relative orientations of TRI4 and the RANK peptide on the surface of TRAF6. The INPHARMA NOEs were observed between the ring-1 protons (r1-2,3) of TRI4 and

Figure 3. Structural model of the TRAF6−TRI1 complex. (a) The top-scored structure of the TRAF6−TRI1 complex from the docking calculation is shown. TRI1 is shown as magenta sticks, and the TRAF6 residues that interact with TRI1 are shown as yellow sticks. A red dashed line indicates a hydrogen bond with the carbonyl group of Gly470, and a blue dashed line indicates a hydrogen bond to the side chain of Arg-392. (b) Schematic representation of the interaction. Dashed lines correspond to those in (a). van der Waals contacts are indicated by orange curves. (c) TRI1 (magenta sticks) is overlaid with the RANK peptide (transparent cyan sticks) on the surface of TRAF6. The Pro site is highlighted by a red dashed circle.

The higher IC50 compared to the KD value for TRI4 could be explained by its low membrane permeability due to its relatively high molecular weight. To confirm that TRI4 can inhibit TRAF6−TNFRSF interactions, displacement of the RANK peptide from TRAF6 was analyzed by solution NMR. As expected, titration of TRI4 into the TRAF6−RANK complex recovered the intensity of signals from the RANK peptide (Figure S2, Supporting Information), confirming that TRI4 is a competitive inhibitor of TRAF6−TNFRSF interactions. It should be noted that TRI1−TRI4 showed little intermolecular interaction and no tendency of aggregation within the D

DOI: 10.1021/acs.jmedchem.5b00778 J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

Figure 5. NMR interaction analysis between TRAF6 and TRI4. (a) 1H−15N HSQC spectra of 50 μM TRAF6 in the absence (black) and presence (red) of 180 μM TRI4. (b) CSPs induced by TRI4 binding observed in the backbone amide groups of TRAF6. Six residues with resonances that disappear upon addition of TRI4 are indicated with orange bars. Fourteen residues with resonances experiencing the largest chemical shift changes upon the addition of TRI4 are shown in red. (c) Distribution of the significantly affected residues upon the addition of TRI4 in the structure of TRAF6. The structure is shown in the same orientation as in Figure 2c. The residues with resonances that disappeared or experienced large CSPs upon the addition of TRI4 are shown in orange or red, respectively. Glu-498 is shown in green. (d) Correlation between the CSPs and signal intensity reductions (Ibound/Ifree) of TRAF6 amide resonances induced by the addition of TRI4. 180 μM TRI4 was added to 50 μM uniformly 15Nlabeled TRAF6. CSPs are calculated as Δδ = (ΔδN2 + ΔδH2)1/2 in units of hertz.

the β and γ protons of the conserved Pro residue of the RANK peptide (Figure S4a, Supporting Information). These correlations were not observed in a control experiment using uniformly 2H-labeled TRAF6, indicating that the NOE signals were derived from the indirect magnetization transfer between TRI4 and the RANK peptide in the shared binding site. Thus, the experiment clearly indicates that ring-1 of TRI4 occupies the same pocket as the conserved Pro residue of TNFRSFs. For more accurate model building of the TRAF6−TRI4 complex, we carried out a DIRECTION experiment.19 Intermolecular cross relaxation rates, which reflect the surrounding proton density of each TRI4 site in the TRAF6−TRI4 interface, were measured in the DIRECTION experiment (Figure S4b, Supporting Information), and docking

calculations were performed to satisfy the result of the DIRECTION experiment (Figure S4c).20 The resulting structure model of the TRAF6−TRI4 complex is shown in Figure 6. In the complex structure model, ring-1 of TRI4 is located in the hydrophobic pocket for the conserved Pro residue of TNFRSFs (Figure 6), which is consistent with the INPHARMA experiment. Although there were structures in which TRI4 bound to TRAF6 in the opposite direction, the possibility was excluded by the result of the INPHARMA experiment. The chlorine on ring-1 makes van der Waals interactions with the surrounding Met-450, Phe-471, and Tyr473 residues of TRAF6 (Figure 6a,b). Ring-2 forms a stacking interaction with Phe-410, and ring-3 forms a CH−π interaction with the side chain of His-394 (Figure 6a,b). Ring-3 occupies a site that is not utilized in the TRAF6−TNFRSF interactions E

DOI: 10.1021/acs.jmedchem.5b00778 J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

(Figure 6c), and the H394A mutation in TRAF6 decreased the affinity of TRI4 4-fold (KD = 150 μM). Furthermore, the amide oxygen in the ring-1/ring-2 linker forms a hydrogen bond with the amide proton of Gly-472 (Figure 6a,b), which is also observed in the TRAF6−RANK interface (Figure 1). All of these interactions were preserved within 4 ns of a molecular dynamic (MD) calculation, indicating that TRI4 fits well in the binding interface of TRAF6. The sulfonyl oxygen in the ring-2/ring-3 linker makes a hydrogen bond with the guanidinium group of Arg-392 (Figure 6a,b). It should be noted that the acylsulfonamide group of TRI4 has a pKa around 5.4 as the solubility of the compound dropped significantly below the pH value. The hydrogen bond was observed in 25% of structures in the MD trajectory and contributes to the potency of TRI4, as the R392A mutation of TRAF6 reduced the affinity for TRI4 3-fold (KD = 100 μM) compared to the wild type. As the carboxylic acid of TRI1 has a similar interaction with Arg-392 (Figure 3a), retaining the acidic centers would have been crucial in the compound optimization process. The interaction between Arg-392 of TRAF6 and Asp of the RANK peptide is important for TRAF6−RANK binding (Figure 1); thus, it is reasonable that TRI compounds target the site for competition.11 The chlorine on ring-3 forms van der Waals interactions with Gln-396, Cys403, and Val-474 of TRAF6, consistent with data showing that elimination of the chlorine drastically decreases the potency (data not shown). In this study, we developed a competitive PPI inhibitor with an affinity for TRAF6 higher than that of the endogenous TNFRSFs by taking advantage of the known structures of TRAF6−TNFRSF complexes and targeting multiple key interactions in TRAF6−TNFRSF interfaces that are energetically required for TRAF6−TNFRSF complex formation. While TRI4 binds to the same surface as TNFRSFs and competitively inhibits RANK peptide binding to TRAF6, substantial CSPs upon TRI4 binding were also observed in the β-strands behind the TRAF6−RANK interface (Figure 5c). This reflects an allosteric structural change coupled to the TRI4 binding. The allosteric structural change is also induced by binding of the RANK peptide (Figure S5a−c, Supporting Information) but not reflected in the X-ray structure of the TRAF6−RANK complex (Figure S5d).11 Overall, the bound TRI4 extends horizontally along the surface of TRAF6, and a hydrogen bond between TRI4 and a main chain amide proton of TRAF6 is vertically arranged in the interface (Figure 6 and Figure S6a, Supporting Information). The methylene group in the ring-1/ring-2 linker of TRI4 provides an additional rotational degree of freedom that enables this vertical hydrogen bond, which is also observed in the TRAF6−RANK interface and is therefore essential to supplant the endogenous interactions (Figure 1 and Figure S6b). This structural feature of TRI4 is derived from the primary hit, TRI1, which was identified in the in silico MSM-MTS approach, where the docking score was modified to recapture the interactions observed in the structure of the RANK peptide. The high affinity of TRI4 is also ascribed to additional interactions outside of the TRAF6−TNFRSF interface (Figure 6c). Previous examples in the literature have also indicated that inhibitor potencies can be improved by targeting surfaces outside the binding sites of endogenous ligands.3 This feature might be advantageous in improving the specificity of TRI4 for TRAF6 relative to other TRAF family proteins (TRAF2, TRAF3, and TRAF5; TRAF2/3/5), although experimental

Figure 6. Structural model of the TRAF6−TRI4 complex. (a) Structure and key interactions of the TRAF6−TRI4 complex. Red and blue dashed lines indicate interactions involving the main chain and side chains of TRAF6, respectively. The black dashed line indicates an intramolecular hydrogen bond. The TRAF6 residues that interact with TRI4 are shown as yellow sticks. The residues that form van der Waals interactions with ring-1 and ring-3 are highlighted by transparent orange surface representations. (b) Schematic representation of the interactions between TRAF6 and TRI4. The representation schemes are the same as in Figure 1. (c) TRI4 (magenta sticks) is overlaid with the RANK peptide (transparent cyan sticks) on the surface of TRAF6. The Pro site and 6877002 site are highlighted by red and blue dashed circles, respectively. F

DOI: 10.1021/acs.jmedchem.5b00778 J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

verification would be necessary. The β3−β4 loop of TRAF6, which is the binding site for ring-3 of TRI4 (Figure 6a), has a one-residue insertion in TRAF6 and forms a structure distinct from that of TRAF2/3/5 (Figure S7, Supporting Information).21 In addition, His-394 in the β3-strand of TRAF6 is involved in the interaction with ring-3 of TRI4 (Figure 6a) and is mutated to Tyr in other TRAF family proteins. Predicted binding modes of TRI4 to TRAF2/3/5 are different from that to TRAF6, and the β3−β4 loop in TRAF2/3/5 no longer participated in the interaction with the ring-3 of TRI4. In addition, the docking score of TRI4 to TRAF6 was higher than those to any other TRAF family proteins. As TRAF6 and TRAF2/3/5 signaling reportedly have opposite effects on inflammation and metabolic dysfunction,8 specific binding to the TRAF6 structure may improve the specificity and clinical efficacy of TRI4. While the binding site of TRI4 is in the interface for the N-terminus of a TNFRSF peptide, another TRAF6 inhibitor, compound 6877002, was shown to target the binding pocket for the C-terminus of a TNFRSF peptide.8 The binding sites for these two inhibitors are adjacent but do not overlap (Figure 6c). Therefore, linking these two TRAF6− TNFRSF inhibitors could lead to a more potent inhibitor. The different binding sites might reflect differences in the hitidentification approaches. We further emphasize that NMR techniques are important for experimentally obtaining structural information for bound PPI inhibitors in solution without distortions originating from crystallization. When structural information for the targeted PPI is available, our integrated strategy utilizing NMR and in silico methods could be applied generally to drug discovery targeting PPIs.

of the crystal structure of the TRAF6−RANK complex (PDB code 1LB5) after removal of the RANK peptide. Hydrogen atom coordinates were generated by Maestro (Schrödinger, Inc.), and the whole structure was optimized by energy minimization before the MSM-MTS procedures were started. For the machine-learning procedure, the RANK peptide structure in the TRAF6−RANK complex (PTEDEY sequence; PDB code 1LB5) was used as an active molecule, following the procedure provided by the developer (http:// presto.protein.osaka-u.ac.jp/myPresto4/). Then 2000 compounds were chosen from the 20 000 MSM-MTS hits using the standard precision (SP) mode of the Glide software (Schrödinger, Inc.); the initial coordinates of TRAF6 were the same as those used in the MSMMTS screening. The force field used was OPLS2001. These compounds were then classified into 840 groups using twodimensional fingerprint analysis by the improved MACCS key in the MOE program (Chemical Computing Group Inc.). The classified compounds were prioritized on the basis of the docking score of the extra precision (XP) mode of the Glide software, commercial availability, and drug-likeness, while their structural diversity was retained. The protein coordinate and force field used in Glide XP calculation were the same as those in Glide SP. This analysis resulted in selection of 682 compounds and their analogues for further analysis. Compounds were purchased from Namiki Shoji Co., Ltd., and 100 mM solutions were prepared by dissolution in dimethyl sulfoxide (DMSO). Peptide and Protein Production. Unless otherwise noted, all chemicals were obtained from WAKO chemicals, and stable isotopes were obtained from Cambridge Isotope Laboratories. Preparation of the RANK Peptide. The TRAF6-binding sequence from human RANK receptor (343-QMPTEDEY-349) was synthesized and purified by reversed-phase high-performance liquid chromatography (HPLC) (Toray Research Center). For surface plasmon resonance (SPR) analysis, the N-terminus of the RANK peptide was biotinylated for immobilization to the sensor chip. Preparation of the C-Terminal Domain (C-Domain) of TRAF6. The C-domain of human TRAF6 (residues 346−504) was cloned into a pET-24d vector (Novagen, Madison, WI) with a C-terminal Histag.23 Unlabeled TRAF6 was obtained by growing Escherichia coli BL21(DE3) cells (Novagen) harboring the plasmid encoding TRAF6 in Luria−Bertani (LB) medium containing 50 μg/mL kanamycin. For the expression of uniformly 15N- or 13C,15N-labeled TRAF6, the E. coli cells were grown in CHL medium (Chlorella Industry) containing 50 μg/mL kanamycin and corresponding stable isotope labeling. For the expression of uniformly 2H-labeled TRAF6, M9 medium with 1 g/L NH4Cl, 2 g/L 2H7-glucose, and 1 g/L 2H-Celtone in D2O was used. For the expression of uniformly 2H,15N-labeled TRAF6, M9 medium with 1 g/L 15NH4Cl, 2 g/L 2H7-glucose, and 1 g/L 2H,15N-Celtone in D2O was used. In the expression procedure, 0.5 mM isopropyl β-Dthiogalactopyranoside (IPTG) was added when the optical density at 600 nm (OD600) reached 0.6, and protein expression was induced for 12 h at 20 °C. The cells were harvested and then disrupted using 50 mL BugBuster (Merck) containing 50 000 units of lysozyme (Epicenter), 4000 units of Benzonase (Merck), and one tablet of Complete Tablet (Roche) per 1 L of culture. After centrifugation, the supernatant was loaded onto a Ni affinity column (His Trap FF clude 5 mL; GE Healthcare) and equilibrated with buffer A (20 mM Tris− HCl (pH 8.5), 100 mM NaCl, and 1 mM dithiothreitol (DTT)) in an AKTA purification system (GE Healthcare). The TRAF6 protein was eluted by an imidazole gradient from 0 to 500 mM with approximately 110 mM imidazole. The fractions containing TRAF6 were concentrated to less than 10 mL and loaded onto a HiLoad 26/60 Superdex 75 pg column (GE Healthcare) equilibrated with buffer B (20 mM sodium phosphate (pH 7.2), 100 mM NaCl). The fractions containing TRAF6 were collected, and purity was confirmed by sodium dodecyl sulfate−polyacrylamide gel electrophoresis (SDS− PAGE). The final yield was 1−3 mg/L of culture. The plasmids encoding mutants of TRAF6 were constructed by Quikchange mutagenesis (Agilent Technologies) of the plasmid encoding wildtype TRAF6. The mutants of TRAF6 were overexpressed and purified using the same procedure described for wild-type TRAF6.



CONCLUSION In summary, we developed TRI4, a potent small-molecule inhibitor of PPIs between TRAF6 and TNFRSFs, using a structure-guided approach. TRI4 is the first inhibitor that binds to TRAF6 with a higher affinity than those of the endogenous binding partners and directly competes with the N-terminal conserved motif in TNFRSFs for the TRAF6 interface. In this study, we showed that an initial compound, which weakly but specifically inhibits multiple key interactions in the targeted PPI, could be a good starting compound for development of a relatively high-potency compound. The TRAF6−TRI4 complex structure obtained from docking calculations based on NMR experiments (INPHARMA and DIRECTION) suggested that an additional binding site outside of the endogenous PPI interface could allow high affinity and target specificity. Finally, throughout this study, we showed that the combination of NMR and in silico methods provides an effective approach for structural analysis for rational drug design and elucidation of the molecular mechanism of a PPI inhibitor. Our results indicate new possibilities for the treatment of RA using a smallmolecule inhibitor of PPIs and provide a foundation for further development of novel PPI inhibitors.



EXPERIMENTAL SECTION

in Silico Screening and Selection of Compounds by Docking Calculation. Twenty thousand compounds were selected from one million starting compounds in LigandBox14,15 using the machinelearning score modification multiple target screening (MSM-MTS) protocol implemented in myPresto (http://presto.protein.osaka-u.ac. jp/myPresto4/).12,13,22 The template three-dimensional (3D) structures of TRAF6 for the in silico screening were prepared on the basis G

DOI: 10.1021/acs.jmedchem.5b00778 J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

EtOAc (2:1), and the solid was filtered and dried under reduced pressure to give a white solid (77 mg, 54%). TFA (570 μL, 7.4 mmol) was added to an ice-cooled solution of tert-butyl-4-[2-(5-chloro-2-oxo2,3-dihydro-1,3-benzothiazol-3-yl)acetamido]benzoate (75 mg, 0.179 mmol) in CH2Cl2 (2 mL), and the mixture was stirred at room temperature for 2 h. The reaction mixture was concentrated in vacuo, and the residue was treated with diethyl ether. The resulting solid was filtered and dried under reduced pressure to give a white solid (60 mg, 92%). 1H NMR (400 MHz, DMSO-d6): 12.75 (br s, 1H), 10.77 (s, 1H), 7.91 (d, J = 9.2 Hz, 2H), 7.74 (d, J = 8.4 Hz, 1H), 7.69 (d, J = 8.8 Hz, 2H), 7.60 (d, J = 2.0 Hz, 1H), 7.22 (dd, J = 8.4, 2.0 Hz, 1H), 4.89 (s, 2H). LC−MS: 1.15 min. HRMS (m/z): [M + H]+ calcd for C16H11ClN2O4S, 363.0206; found, 363.0208. 4-(2-(5-Chloro-2-oxo-2,3-dihydro-1,3-benzothiazol-3-yl)acetamido)-N-(methylsulfonyl)benzamide (TRI3). N,N-Diisopropylethylamine (144 μL, 0.826 mmol) was added to a mixture of TRI2 (150 mg, 0.413 mmol), methanesulfonamide (196 mg, 2.07 mmol), and HATU (157 mg, 0.328 mmol) in CH2Cl2 (4 mL) at room temperature, and the mixture was stirred overnight. The reaction mixture was diluted with CH2Cl2 (30 mL), washed with H2O (10 mL) and saturated NaCl (10 mL), and dried over anhydrous Na2SO4. The precipitate was removed by filtration, and the resulting solution was evaporated under reduced pressure. The obtained residue was purified by HPLC (H2O/MeCN/CH3COOH), and the eluent was evaporated under reduced pressure to give a white solid (42.3 mg, 23%). 1H NMR (700 MHz, DMSO-d6): 12.02 (br, 1H), 10.63 (br, 1H), 7.90 (d, J = 8.7 Hz, 2H), 7.74 (d, J = 8.4 Hz, 1H), 7.61 (d, J = 2.0 Hz, 1H), 7.59 (d, J = 7.3 Hz, 2H), 7.29 (dd, J = 8.4, 2.0 Hz, 1H), 4.88 (s, 2H), 3.02 (br, 3H). LC−MS: 1.15 min. HRMS (m/z): [M + H]+ calcd for C17H14ClN3O5S2, 440.0141; found, 440.0140. 4-(2-(5-Chloro-2-oxo-2,3-dihydro-1,3-benzothiazol-3-yl)acetamido)-N-[(3-chloro-4-methoxyphenyl)sulfonyl]benzamide (TRI4). N,N-Diisopropylethylamine (29 μL, 0.165 mmol) was added to a mixture of TRI2 (30 mg, 0.083 mmol), 3-chloro-4-methoxybenzene1-sulfonamide (27.5 mg, 0.124 mmol), and HATU (31.4 mg, 0.083 mmol) in CH2Cl2 (3 mL) at room temperature, and the mixture was stirred overnight. The reaction mixture was diluted with CH2Cl2 (20 mL), washed with H2O (7 mL) and saturated NaCl (7 mL), and dried over anhydrous Na2SO4. The precipitate was removed by filtration, and then the resulting solution was evaporated under reduced pressure. The obtained residue was purified by HPLC (H2O/ MeCN/CH3COOH), and the eluent was evaporated under reduced pressure to give a white solid (12.9 mg, 28%). 1H NMR (700 MHz, DMSO-d6): 12.41 (br, 1H), 10.63 (br, 1H), 7.88 (br, 1H), 7.84 (d, J = 8.6 Hz, 2H), 7.82 (br, 1H), 7.74 (d, J = 8.4 Hz, 1H), 7.60 (d, J = 1.6 Hz, 1H), 7.57 (br, 2H), 7.28 (dd, J = 8.4, 1.8 Hz, 1H), 7.24 (br, 1H), 4.87 (s, 2H). 3.91 (s, 3H). LC−MS: 1.40 min. HRMS (m/z): [M + H]+ calcd for C23H17Cl2N3O6S2, 566.0014; found, 566.0014. NMR Spectroscopy. The purified TRAF6 solution was exchanged into 20 mM phosphate buffer (pH 7.2) with 100 mM NaCl and 5 mM deuterated DTT. For titration analysis, the RANK peptide (10 mM in D2O) and the compounds (10 or 100 mM in DMSO-d6) were added to protein samples to reach the desired concentrations. All NMR spectra were recorded on a Bruker Avance 600, 700, or 800 MHz spectrometer equipped with a cryo-cooled triple-resonance probe. Unless otherwise described, all experiments were performed at 25 °C. The carrier positions and sweep widths were set at 54 and 34 ppm, respectively, for the 13C dimension, 118 and 36 ppm, respectively, for the 15N dimension, and 4.7 and 16 ppm, respectively, for the 1H dimension. Data were processed using the TOPSPIN program and analyzed by using the SPARKY3 program.25 Main chain amide resonances of TRAF6 in the absence and presence of the RANK peptide were assigned on the basis of conventional 3D experiments. For the measurements, uniformly 13 15 C, N-labeled TRAF6 was concentrated to 160 μM in the free state and 500 μM in the presence of 1 mM RANK peptide. Totals of 72% and 74% of the main chain resonances were assigned for the free and RANK-bound states, respectively. Upon compound titration, most of the signals exhibited fast to intermediate CSPs, although some disappeared. Dissociation constants (KD) of the compounds were

SPR Analysis. All experiments were performed at 25 °C using a flow rate of 20 μL/min in HBS-EP buffer. The biotinylated RANK peptide was immobilized on a standard BIAcore SA sensor chip. TRAF6 was injected over RANK-immobilized and reference flow cells at a concentration of 2.5 μg/mL in the absence and presence of test compounds. The changes in maximal RU were obtained for various concentrations of compounds, and the IC50 was calculated on the basis of the concentration dependence of the inhibition. Osteoclast Differentiation Assay. Primary mouse bone marrowderived monocytes (BMMs) were prepared from femurs of 6−10week-old mice (BALB/c, Male). BMM cells were suspended in αMEM with 10% charcoal-treated fetal bovine serum (FBS) and cultured in a 96-well tissue culture plate at a density of 6 × 105 cells per well. They were treated with 30 ng/mL RANKL, 10 ng/mL MCSF, and varying concentrations of test compounds or DMSO as a control at the beginning of the culture. After 7 days, BMM cell differentiation and viability were assessed on the basis of tartrateresistant acid phosphatase (TRAP) activity using p-nitrophenyl phosphate (pNPP) as the substrate and AlamarBlue, respectively. For evaluation of cell differentiation, after the culture medium was removed, 100 μL of substrate solution containing 10 mM pNPP, 0.1 M sodium acetate (pH 5.8), 0.15 M KCl, 0.1% (v/v) Triton X-100, 10 mM sodium tartrate, 1 mM ascorbic acid, and 0.1 mM FeCl3 was added to each well. After 30 min of incubation at 37 °C, 100 μL of 0.3 M NaOH was added, and absorbance was measured at 405 nm.24 For the evaluation of cell viability, 10 μL of AlamarBlue medium (Life Technologies Japan) was added to 100 μL of cultured medium. After 7 h of incubation at 37 °C in a 5.0% CO2 incubator, fluorescence was measured with 560 nm excitation and 590 nm emission detection. Compound Synthesis. NMR spectra were recorded on a Bruker Avance 400 or 700 MHz spectrometer (Bruker Biospin). 1H chemical shifts were reported using tetramethylsilane (TMS) as the internal chemical shift standard in deuterated DMSO (DMSO-d6) or CDCl3. Data are reported as follows: chemical shift (δ), multiplicity (s = singlet, d = doublet, br = broad), coupling constant (Hz), integration. High-resolution mass spectra were recorded on a Waters Q-TOF LCT Premier XE system with electrospray ionization (Waters). All compounds had almost 100% purity as analyzed by liquid chromatography−mass spectrometry (LC−MS). Analytical HPLC was performed on a Waters Acquity system with photodiode array (PDA) and evaporative light scattering (ELS) detection. LC−MS parameters were as follows: Acquity UPLC BEH C8 column, 1.7 μm, 50 × 2.1 mm, 1.9 min gradient, 1% (0.1% formic acid/MeCN)/99% (0.1% formic acid/H2O) to 100% (0.1% formic acid/MeCN). Preparative purification was performed on a Shimadzu HPLC system (CAPCELL PAK-C18 ACR, 50 × 20 mm; mobile phase A = water containing 0.1% AcOH, mobile phase B = MeCN containing 0.1% AcOH; gradient: 0−7 min, linear 0−100% B, 7−9.2 min, isocratic 100% B, 9.2−10 min, isocratic 1% B). 4-[2-(2-Oxo-2,3-dihydro-1,3-benzothiazol-3-yl)acetamido]benzoic Acid (TRI1). This compound was purchased from Namiki Shoji Co., Ltd. 1H NMR (400 MHz, DMSO-d6): 12.75 (br s, 1H), 10.77 (s, 1H), 7.91 (d, J = 9.2 Hz, 2H), 7.74 (d, J = 8.4 Hz, 1H), 7.69 (d, J = 8.8 Hz, 2H), 7.60 (d, J = 2.0 Hz, 1H), 7.22 (dd, J = 8.4, 2.0 Hz, 1H), 4.89 (s, 2H). LC−MS: 1.05 min. HRMS (m/z): [M + H]+ calcd for C16H12N2O4S, 329.0596; found, 329.0594. 4-[2-(5-Chloro-2-oxo-2,3-dihydro-1,3-benzothiazol-3-yl)acetamido]benzoic Acid (TRI2). As a precursor, tert-butyl-4-[2-(5chloro-2-oxo-2,3-dihydro-1,3-benzothiazol-3-yl)acetamido]benzoate was first synthesized as described below. To a mixture of 2-(5-chloro2-oxo-2,3-dihydro-1,3-benzothiazol-3-yl)acetic acid (80 mg, 0.328 mmol), tert-butyl 4-aminobenzoate (63.4 mg, 0.328 mmol), and HATU (O-(7-azabenzotriazol-1-yl)-N,N,N′,N′-tetramethyluronium hexafluorophosphate) (125 mg, 0.328 mmol) in CH2Cl2 (6 mL) was added N,N-diisopropylethylamine (114 μL, 0.656 mmol) added at room temperature, and the mixture was stirred overnight. The reaction mixture was concentrated in vacuo, and the residue was diluted with EtOAc and washed with 10% citric acid, water, saturated NaHCO3, and saturated NaCl. The resulting solution was dried over anhydrous Na2SO4. After solvent removal, the residue was treated with hexane− H

DOI: 10.1021/acs.jmedchem.5b00778 J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

determined from concentration dependence of normalized chemical shift changes upon compound titration.26 A series of 1H−15N HSQC spectra at increasing compound concentrations were acquired, and the normalized chemical shift change was defined as Δδ = ([0.2ΔδN]2 + ΔδH2)1/2. The obtained plots were fitted by an equation described previously.27 The displacement of TRI1 by the RANK peptide was monitored by acquiring 1H 1D spectra of the sample with 20 μM TRI1 and 20 μM uniformly 2H-labeled TRAF6 in the absence and presence of 50 μM RANK peptide. The displacement of the RANK peptide by TRI4 was monitored by acquiring 1H 1D spectra of the sample with 100 μM RANK peptide and 20 μM uniformly 2H-labeled TRAF6 in the absence and presence of 200 μM TRI4. For INPHARMA experiments, a NOESY spectrum was acquired with the sample containing 200 μM RANK peptide, 200 μM TRI4, and 17 μM unlabeled TRAF6. For the control measurement, uniformly 2H-labeled TRAF6 was used instead of unlabeled TRAF6. The mixing time for NOESY experiments was set to 800 ms, and measurement was performed at 10 °C for efficient magnetization transfer. For DIRECTION measurements, the inversion recovery experiments with and without irradiation were performed using the pulse scheme as previously described.19 150 μM TRI4 was mixed with 16 μM TRAF6 for the experiment. The inversion recovery times were set to 0.1, 0.2, 0.4, and 0.6 s. The irradiation frequency was set at 0 and −30 ppm for the spectra with and without irradiation of TRAF6 protons, respectively. The irradiation time was set at 2 s, and the delay for relaxation was 6 s. The normalized signal intensities of each ligand proton were plotted against the time of inversion recovery, and the intermolecular cross relaxation rates were calculated as the differences between longitudinal relaxation rates with and without irradiation of TRAF6 protons. Docking and Molecular Dynamic (MD) Calculation with the DIRECTION Data. The compounds were first docked to the protein by Sievgene/myPresto28 and equilibrated by MD calculations using Cosgene/myPresto22 in solvent to generate candidate complex structures. On the basis of the structures, simulated DIRECTION data were calculated by SievgeneNMR/myPresto19,20 and fitted against the experimental data to obtain the rational model of the complex satisfying the DIRECTION data. To confirm the stability of each interaction observed in the structure of the TRAF6−TRI4 complex, a 4 ns MD calculation from the obtained structure was performed, and 100 structures were collected at 40 ps intervals from the MD trajectory. Details are given below. The force field and atomic charges of the protein atoms were obtained from AMBER parm99.29 The atomic charge of the ligand was determined by the restricted electrostatic point charge (RESP) procedure using HF/6-31G*-level quantum chemical calculations with Gaussian98.30,31 MD calculation was performed in a sphere of TIP3P water molecules (CAP water),32 including ion particles of 0.1% Na+ and Cl−, to neutralize the total charge of the systems. The center of the sphere was set at the mass center of the protein. The shortest distance between the protein atom and the CAP sphere wall was set to 10 Å. Before an MD calculation was performed for the entire system, an MD calculation only for the solvent parts (solvent water and counterions) was performed with fixed protein, ligand, and metal ion coordinates. Thereafter, MD calculations for the entire system were performed with increasing temperatures of 100, 200, and 300 K using 2.0 fs time steps while the SHAKE method was applied for hydrogen atoms.33 The fast multipole method was used to calculate Coulombic interactions.34 The cutoff distance for van der Waals interactions was 12 Å. Docking Calculation of TRI4 to TRAF2/3/5/6 for Comparison of Complex Structures. Glide XP was used for the docking calculations. The template 3D structures of TRAF2/3/5/6 were prepared on the basis of their respective crystal structures without ligand (TRAF2, PDB code 1CA4; TRAF3, PDB code 1FLK; TRAF5, PDB code 4GJH; TRAF6, PDB code 1LB5). Hydrogen atom coordinates were generated by Maestro (Schrödinger, Inc.), and the whole structure was then optimized by energy minimization. The 3D

structure of TRI4 was constructed by Maestro. The ionization state was predicted using EPIC, and the coordinate was optimized by a quantum mechanics calculation using Jaguar. The force field used was OPLS2001. The structure models with the highest scores were chosen for the comparison.



ASSOCIATED CONTENT

S Supporting Information *

Additional figures illustrating hit compounds in the SPR and osteoclast differentiation assays, NMR competition analysis between TRI4 and the RANK peptide on the surface of TRAF6, 1H 1D spectra of TRI1 and TRI4 at various concentrations, results of INPHARMA and DIRECTION experiments, NMR interaction analysis between TRAF6 and the RANK peptide, hydrogen bonds between the amide proton of Gly-472 of TRAF6 and TRI4 or the RANK peptide, and comparison of β3−β4 loop structures among TRAF family proteins. The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/ acs.jmedchem.5b00778.



AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Phone/fax: (+81)-33815-6540. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS This work was supported by a grant from the Japan New Energy and Industrial Technology Development Organization (NEDO) and the Ministry of Economy, Trade, and Industry (METI; to I.S.). We thank Dr. S. Fujimoto for the X-ray analysis, Dr. S. Suzuki for the compound preparation, Dr. K. Hagiwara for the HRMS measurements, Dr. M. Nakao for the PubChem search, Dr. M. Ikemori-Kawada, Dr. K. Sawada, Dr. T. Yoshinaga, and Dr. T. Dodo for the prediction of physicochemical and ADME parameters of compounds, and Dr. I. Kushida for the pKa analysis of TRI compounds.



ABBREVIATIONS USED AP1, activator protein 1; BMMs, bone marrow-derived monocytes; CD, cluster of differentiation; CSP, chemical shift perturbation; FBS, fetal bovine serum; HATU, O-(7-azabenzotriazol-1-yl)-N,N,N′,N′-tetramethyluronium hexafluorophosphate; IL, interleukin; IPTG, isopropyl β-D-thiogalactopyranoside; IRAK, interleukin-1 receptor-associated kinase; LB, Luria−Bertani; MSM-MTS, machine-learning score modification multiple target screening; PAINS, pan assay interference compounds; pNPP, p-nitrophenyl phosphate; PPI, protein− protein interaction; RA, rheumatoid arthritis; RANK, receptor activator of NF-kB; RANKL, receptor activator of the NF-kB ligand; TNF, tumor necrosis factor; TNFRSF, tumor necrosis factor superfamily receptor; TRAF, tumor necrosis factor receptor-associated factor; TRAP, tartrate-resistant acid phosphatase



REFERENCES

(1) Arkin, M. R.; Wells, J. A. Small-molecule inhibitors of proteinprotein interactions: progressing towards the dream. Nat. Rev. Drug Discovery 2004, 3, 301−317. (2) Pagliaro, L.; Felding, J.; Audouze, K.; Nielsen, S. J.; Terry, R. B.; Krog-Jensen, C.; Butcher, S. Emerging classes of protein-protein

I

DOI: 10.1021/acs.jmedchem.5b00778 J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

interaction inhibitors and new tools for their development. Curr. Opin. Chem. Biol. 2004, 8, 442−449. (3) Wells, J. A.; McClendon, C. L. Reaching for high-hanging fruit in drug discovery at protein-protein interfaces. Nature 2007, 450, 1001− 1009. (4) Aggarwal, B. B. Signalling pathways of the TNF superfamily: a double-edged sword. Nat. Rev. Immunol. 2003, 3, 745−756. (5) Wu, H.; Arron, J. R. TRAF6, a molecular bridge spanning adaptive immunity, innate immunity and osteoimmunology. BioEssays 2003, 25, 1096−1105. (6) Kobayashi, T.; Walsh, M. C.; Choi, Y. The role of TRAF6 in signal transduction and the immune response. Microbes Infect. 2004, 6, 1333−1338. (7) Joosten, L. A.; Helsen, M. M.; van de Loo, F. A.; van den Berg, W. B. Anticytokine treatment of established type II collagen-induced arthritis in DBA/1 mice. A comparative study using anti-TNF alpha, anti-IL-1 alpha/beta, and IL-1Ra. Arthritis Rheum. 1996, 39, 797−809. (8) Chatzigeorgiou, A.; Seijkens, T.; Zarzycka, B.; Engel, D.; Poggi, M.; van den Berg, S.; van den Berg, S.; Soehnlein, O.; Winkels, H.; Beckers, L.; Lievens, D.; Driessen, A.; Kusters, P.; Biessen, E.; GarciaMartin, R.; Klotzsche-von Ameln, A.; Gijbels, M.; Noelle, R.; Boon, L.; Hackeng, T.; Schulte, K. M.; Xu, A.; Vriend, G.; Nabuurs, S.; Chung, K. J.; Willems van Dijk, K.; Rensen, P. C.; Gerdes, N.; de Winther, M.; Block, N. L.; Schally, A. V.; Weber, C.; Bornstein, S. R.; Nicolaes, G.; Chavakis, T.; Lutgens, E. Blocking CD40-TRAF6 signaling is a therapeutic target in obesity-associated insulin resistance. Proc. Natl. Acad. Sci. U. S. A. 2014, 111, 2686−2691. (9) Kobayashi, N.; Kadono, Y.; Naito, A.; Matsumoto, K.; Yamamoto, T.; Tanaka, S.; Inoue, J. Segregation of TRAF6-mediated signaling pathways clarifies its role in osteoclastogenesis. EMBO J. 2001, 20, 1271−1280. (10) Hostager, B. S. Roles of TRAF6 in CD40 signaling. Immunol. Res. 2007, 39, 105−114. (11) Ye, H.; Arron, J. R.; Lamothe, B.; Cirilli, M.; Kobayashi, T.; Shevde, N. K.; Segal, D.; Dzivenu, O. K.; Vologodskaia, M.; Yim, M.; Du, K.; Singh, S.; Pike, J. W.; Darnay, B. G.; Choi, Y.; Wu, H. Distinct molecular mechanism for initiating TRAF6 signalling. Nature 2002, 418, 443−447. (12) Fukunishi, Y.; Kubota, S.; Nakamura, H. Noise reduction method for molecular interaction energy: application to in silico drug screening and in silico target protein screening. J. Chem. Inf. Model. 2006, 46, 2071−2084. (13) Fukunishi, Y.; Mikami, Y.; Kubota, S.; Nakamura, H. Multiple target screening method for robust and accurate in silico ligand screening. J. Mol. Graphics Modell. 2006, 25, 61−70. (14) Fukunishi, Y.; Sugihara, Y.; Mikami, Y.; Sakai, K.; Kusudo, H.; Nakamura, H. Advanced in-silico drug screening to achieve high hit ratio - development of 3D-compound database. Synthesiology 2009, 2, 60−68. (15) Kawabata, T.; Sugihara, Y.; Fukunishi, Y.; Nakamura, H. LigandBox: A database for 3D structures of chemical compounds. Biophysics (Biophys. Soc. Jpn.) 2013, 9, 113−121. (16) Baell, J. B.; Holloway, G. A. New substructure filters for removal of pan assay interference compounds (PAINS) from screening libraries and for their exclusion in bioassays. J. Med. Chem. 2010, 53, 2719− 2740. (17) Sanchez-Pedregal, V. M.; Reese, M.; Meiler, J.; Blommers, M. J.; Griesinger, C.; Carlomagno, T. The INPHARMA method: proteinmediated interligand NOEs for pharmacophore mapping. Angew. Chem., Int. Ed. 2005, 44, 4172−4175. (18) Ono, K.; Takeuchi, K.; Ueda, H.; Morita, Y.; Tanimura, R.; Shimada, I.; Takahashi, H. Structure-based approach to improve a small-molecule inhibitor by the use of a competitive peptide ligand. Angew. Chem., Int. Ed. 2014, 53, 2597−2601. (19) Mizukoshi, Y.; Abe, A.; Takizawa, T.; Hanzawa, H.; Fukunishi, Y.; Shimada, I.; Takahashi, H. An accurate pharmacophore mapping method by NMR spectroscopy. Angew. Chem., Int. Ed. 2012, 51, 1362−1365.

(20) Fukunishi, Y.; Mizukoshi, Y.; Takeuchi, K.; Shimada, I.; Takahashi, H.; Nakamura, H. Protein-ligand docking guided by ligand pharmacophore-mapping experiment by NMR. J. Mol. Graphics Modell. 2011, 31, 20−27. (21) Wu, H. Assembly of post-receptor signaling complexes for the tumor necrosis factor receptor superfamily. Adv. Protein Chem. 2004, 68, 225−279. (22) Fukunishi, Y.; Mikami, Y.; Nakamura, H. The Filling Potential Method: A Method for Estimating the Free Energy Surface for Protein−Ligand Docking. J. Phys. Chem. B 2003, 107, 13201−13210. (23) Ye, H.; Cirilli, M.; Wu, H. The use of construct variation and diffraction data analysis in the crystallization of the TRAF domain of human tumor necrosis factor receptor associated factor 6. Acta Crystallogr., Sect. D: Biol. Crystallogr. 2002, 58, 1886−1888. (24) Perez-Amodio, S.; Jansen, D. C.; Schoenmaker, T.; Vogels, I. M.; Reinheckel, T.; Hayman, A. R.; Cox, T. M.; Saftig, P.; Beertsen, W.; Everts, V. Calvarial osteoclasts express a higher level of tartrateresistant acid phosphatase than long bone osteoclasts and activation does not depend on cathepsin K or L activity. Calcif. Tissue Int. 2006, 79, 245−254. (25) Goddard, T. D.; Kneller, D. G. SPARKY3; University of California: San Francisco. (26) Clubb, R. T.; Omichinski, J. G.; Clore, G. M.; Gronenborn, A. M. Mapping the binding surface of interleukin-8 complexed with an Nterminal fragment of the type 1 human interleukin-8 receptor. FEBS Lett. 1994, 338, 93−97. (27) Williamson, M. P. Using chemical shift perturbation to characterise ligand binding. Prog. Nucl. Magn. Reson. Spectrosc. 2013, 73, 1−16. (28) Fukunishi, Y.; Mikami, Y.; Nakamura, H. Similarities among receptor pockets and among compounds: analysis and application to in silico ligand screening. J. Mol. Graphics Modell. 2005, 24, 34−45. (29) Case, D. A.; Darden, T.; Cheatham, T. E., III; Simmerling, C. L.; Wang, J.; Duke, R. E.; Luo, R.; Merz, K. M.; Wang, B.; Pearlman, D. A.; Crowley, M.; Brozell, S.; Tsui, V.; Gohlke, H.; Mongan, J.; Hornak, V.; Cui, G.; Beroza, P.; Schafmeister, C.; Caldwell, J. W.; Ross, W. S.; Kollman, P. A. AMBER 8; University of California: San Francisco, 2004. (30) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.; Robb, M. A.; Cheeseman, J. R.; Zakrzewski, V. G.; Montgomery, J. A., Jr.; Stratmann, R. E.; Burant, J. C.; Dapprich, S.; Millam, J. M.; Daniels, A. D.; Kudin, K. N.; Strain, M. C.; Farkas, O.; Tomasi, J.; Barone, V.; Cossi, M.; Cammi, R.; Mennucci, B.; Pomelli, C.; Adamo, C.; Clifford, S.; Ochterski, J.; Petersson, G. A.; Ayala, P. Y.; Cui, Q.; Morokuma, K.; Malick, D. K.; Rabuck, A. D.; Raghavachari, K.; Foresman, J. B.; Cioslowski, J.; Ortiz, J. V.; Stefanov, B. B.; Liu, G.; Liashenko, A.; Piskorz, P.; Komaromi, I.; Gomperts, R.; Martin, R. L.; Fox, D. J.; Keith, T.; Al-Laham, M. A.; Peng, C. Y.; Nanayakkara, A.; Gonzalez, C.; Challacombe, M.; Gill, P. M. W.; Johnson, B. G.; Chen, W.; Wong, M. W.; Andres, J. L.; Head-Gordon, M.; Replogle, E. S.; Pople, J. A. Gaussian 98, revision A.9; Gaussian, Inc.: Pittsburgh, PA, 1998. (31) Wang, J.; Cieplak, P.; Kollman, P. A. How well does a restrained electrostatic potential (RESP) model perform in calculating conformational energies of organic and biological molecules? J. Comput. Chem. 2000, 21, 1049−1074. (32) Jorgensen, W. L.; Chandrasekhar, J.; Madura, J. D.; Impey, R. W.; Klein, M. L. Comparison of simple potential functions for simulating liquid water. J. Chem. Phys. 1983, 79, 926−935. (33) Ryckaert, J. P.; Ciccotti, G.; Berendsen, H. J. C. Numerical integration of the cartesian equations of motion of a system with constraints: molecular dynamics of n-alkanes. J. Comput. Phys. 1977, 23, 327−341. (34) Greengard, L.; Rokhlin, V. A fast algorithm for particle simulations. J. Comput. Phys. 1987, 73, 325−348.

J

DOI: 10.1021/acs.jmedchem.5b00778 J. Med. Chem. XXXX, XXX, XXX−XXX

Structure-Based Development of a Protein-Protein Interaction Inhibitor Targeting Tumor Necrosis Factor Receptor-Associated Factor 6.

The interactions between tumor necrosis factor (TNF) receptor-associated factor 6 (TRAF6) and TNF superfamily receptors (TNFRSFs) are promising target...
5MB Sizes 0 Downloads 5 Views