Skip to main content
NCBI home page
Protein Science : A Publication of the Protein Society logoLink to Protein Science : A Publication of the Protein Society
. 2023 Sep 1;32(9):e4744. doi: 10.1002/pro.4744

Modulation of a conformational ensemble by a small molecule that inhibits key protein–protein interactions involved in cell adhesion

Akinobu Senoo 1,2, Satoru Nagatoishi 3,4,✉, Daisuke Kuroda 1,4,5, Sho Ito 6, Go Ueno 7, Jose M M Caaveiro 2, Kouhei Tsumoto 1,3,4,✉
PMCID: PMC10443342  PMID: 37531208

Abstract

Small molecules that regulate protein–protein interactions can be valuable drugs; however, the development of such small molecules is challenging as the molecule must interfere with an interaction that often involves a large surface area. Herein, we propose that modulating the conformational ensemble of the proteins participating in a given interaction, rather than blocking the interaction by directly binding to the interface, is a relevant strategy for interfering with a protein–protein interaction. In this study, we applied this concept to P‐cadherin, a cell surface protein forming homodimers that are essential for cell–cell adhesion in various biological contexts. We first determined the crystal structure of P‐cadherin with a small molecule inhibitor whose inhibitory mechanism was unknown. Molecular dynamics simulations suggest that the inhibition of cell adhesion by this small molecule results from modulation of the conformational ensemble of P‐cadherin. Our study demonstrates the potential of small molecules altering the conformation ensemble of a protein as inhibitors of biological relevant protein–protein interactions.

Keywords: cadherin, conformational change, inhibition mechanism, protein–protein interaction, small molecule

1. INTRODUCTION

Rather than a single, very stable conformation, most proteins adopt conformational ensembles that are in equilibrium (Frauenfelder et al., 1991). This is the case not only in proteins that act as monomers, but also for protein complexes stabilized by protein–protein interactions (PPIs; Liauw et al., 2021; Wang et al., 2021; Wei et al., 2016). The function of a protein complex depends on some of the conformers within the conformational ensemble. For example, in the interaction between Keap1 and Nrf2, the conformational state of the complex regulates the ubiquitination cycle (Baird et al., 2013). Likewise, the bacterial enzyme dihydropteroate synthase functions as a homodimer, and its catalytic activity depends on the conformational state of the homodimer (Hammoudeh et al., 2014). In these cases, a limited number of substates in the conformational ensemble of a protein complex are responsible for its specific biological function.

The majority of small molecule inhibitors that block PPIs rely on an orthosteric inhibitory mechanism (Fischer et al., 2015), in which the small molecule directly binds to the interface on one protein to prevent interaction with the other molecule of protein. However, given that the interface surface area is inevitably much larger than the accessible surface area of the small molecule (Ran & Gestwicki, 2018; Smith & Gestwicki, 2012), this strategy does not work for many PPIs, explaining why the regulation of PPI by a small molecule is thought to be challenging (Arkin & Wells, 2004; Scott et al., 2016; Wells & Mcclendon, 2007). As an alternative to the orthosteric mechanism, ensemble modulation, where a small molecule binds to a protein and shifts the conformational ensemble to one non‐optimal for the function, is an attractive strategy (Fischer et al., 2015; Lo et al., 2021; Quiniou et al., 2014).

P‐cadherin is a classical type I cadherin that is important for adhesion among epithelial cells, but that has also been important in cancer cell proliferation and metastasis (Van Roy, 2014; Vieira & Paredes, 2015). Small molecules that inhibit its homodimerization also block cell adhesion and have potential to be anti‐metastatic agents (Han et al., 2011; Senoo et al., 2018; Zhang et al., 2010). Proteins belonging to the type I cadherin family form two types of sequential homodimers (Kudo et al., 2016), the X dimer and the strand‐swap dimer, both of which are stabilized by interactions between the first two extracellular domains of P‐cadherin termed EC1 and EC2 (EC12) (Figure 1a). The X dimer assembles first, and facilitates the formation of the more stable and biological relevant strand‐swap dimer (Kudo et al., 2016). A previous study suggested that the small molecule 2‐(5‐chloro‐2‐methyl‐1H‐indole‐3‐yl)ethan‐1‐amine (Hit1, Figure 1b) modulates X dimerization and inhibits cell adhesion (Senoo et al., 2021). Structural and functional analysis revealed that Hit1 binds to the cavity located between EC1 and EC2 in the X dimer. This observation led to synthesis of another ligand, 2‐(2‐methyl‐5‐phenyl‐1H‐indole‐3‐yl)ethan‐1‐amine (PhHit1, Figure 1c), which also inhibits cell adhesion effectively (Senoo et al., 2021). However, the precise mechanism of inhibition by PhHit1 has not been elucidated.

FIGURE 1.

FIGURE 1

Open in a new tab

Formation of P‐cadherin dimers and previously identified dimerization inhibitors. (a) Schematic illustration of cell adhesion mediated by P‐cadherin dimerization. The intermediate X dimer is converted to the strand‐swap dimer, the final state responsible for cell adhesion. (b) Chemical structure of Hit1. (c) Chemical structure of PhHit1.

Herein, we report the crystal structure of P‐cadherin X dimer with the optimized inhibitor PhHit1. We performed molecular dynamics (MD) simulations using the coordinates of this structure as our starting point. In the presence of PhHit1, we observed that the conformational ensemble of X dimer was clearly altered. From the viewpoint of conformational ensemble, we discuss a reasonable mechanism underlying the inhibition of cell adhesion by PhHit1.

2. RESULTS

2.1. PhHit1 binds to the X dimer as shown by crystallographic analysis

Based on the structural similarities between Hit1 and PhHit1, we assumed that the two ligands bind to the same cavity on the P‐cadherin X dimer. We previously showed that Hit1 binds to the cavity between the EC1 and EC2 domains of the X dimer (Senoo et al., 2021). To confirm the binding mode of PhHit1, we analyzed the complex of PhHit1 and a previously described construct that forms X dimer but not the strand‐swap dimer (Kudo et al., 2016). The crystal of X dimer was soaked in a solution containing PhHit1, and the structure of this holo form was determined to a resolution of 2.85 Å.

As expected from the location of Hit1 binding (Senoo et al., 2021), we found electron density that could be modeled as PhHit1 in the cavity between the EC1 and EC2 domains (Figure 2a, Supporting Information Figure S1a). The conformation of PhHit1 in the structure reflects the best match to the electron density, although from its shape, we cannot rule out other alternative conformation. Based on the 3D interaction map generated with BINANA (Durrant & McCammon, 2011), PhHit1 binding appears to be stabilized through π–π interactions, a salt bridge and hydrogen bonds with residues surrounding PhHit1 (Supporting information Figure S1b). Although the establishment of these interactions depends on the charge condition of PhHit1 molecule, because Hit1 did not form either a salt bridge or hydrogen bonds, but only π–π interaction with the X dimer (Supporting information Figure S1c), the contribution of the phenyl group of PhHit1 could be more relevant in the X dimer, resulting in greater affinity. This could explain the stronger activity of PhHit1 with respect to Hit1. PhHit1 caused a structural change upon binding; in the complex with the small molecule, the relative positions of chains A and B in the X dimer are visibly different from those in the apo X dimer, the form not bound to PhHit1 (Figure 2b). The difference seems to be caused by the disruption of one of a pair of hydrogen bonds, which are known hot spots of X dimerization (Kudo et al., 2016). The hydrogen bond between K14‐A138', but not the hydrogen bond between K14'‐A138 (where the lack of a prime symbol indicates a residue of chain A, and the prime symbol indicates a residue of chain B) was disrupted in the presence of PhHit1 (Figure 2c). Besides, in correlation with this hydrogen bond impairment, the conformation of the loop where K14 is located was fluctuated (Figure 2d). Indeed, PhHit1 molecules have some contact with the residues on the loop such as Q19 and L21. These contacts could have caused such loop fluctuation. Interestingly, the bound PhHit1 moiety did not directly inhibit hydrogen bond formation; rather, a structural change induced by PhHit1 binding seems to disrupt the interaction indirectly (Figure 2c). Although with the structural change, PhHit1 bound to the cavity of the X dimer; therefore, the inhibitory effect of PhHit1 should result at a later step in the process of cell adhesion.

FIGURE 2.

FIGURE 2

Open in a new tab

Crystal structure of PhHit1‐bound state of X dimer. (a) Structure of PhHit1‐bound state of X dimer. Chain A is colored in sky blue; chain B in pink. Black arrows indicate the two PhHit1‐binding sites. Green spheres show calcium ions. (b) Superposition of backbone traces of PhHit1‐bound state of X dimer with apo form of X dimer (PDB ID: 4zmq, black). Chain A was fixed. The dotted lines show the angle among Cα atoms from K183'(holo)‐D100‐K183'(apo). (c) Enlarged views show impacts of PhHit1 on hydrogen bonds between K14 on one chain and A138 on the other, which were observed in the crystal structure. The box to the left of the structure shows the K14'‐A138 hydrogen bond (indicated by dotted line) and the box to the right shows that the K14‐A138' hydrogen bond is disrupted (indicated by an arrow). (d) Fluctuation of the loop including K14 residue. Only the backbone trace is shown. The loop of interest is colored in black in the apo form or in blue in the holo form. PhHit1 established interactions with Q19 and L21, presumably causing the difference in the loop conformation.

2.2. Alternation of domain angle in the presence of PhHit1

Since the crystal structure of the PhHit1‐bound state of X dimer suggested that inhibition of cell adhesion by PhHit1 is explained by an indirect mechanism different from an orthosteric one, we sought to analyze the dynamics of the X dimer structure beyond the single snapshot obtained by X‐ray crystallography. We used MD simulations to analyze the conformational dynamics of the PhHit1‐bound state of X dimer. Three independent simulations each for 100 ns were performed using the coordinates of X dimer in complex with PhHit1 as the starting point. Since several binding orientations of PhHit1 could be assumed from the crystal structure, the charge of PhHit1 was set to zero so that the compound could easily bind in different orientations. The convergence of the MD simulations was confirmed by calculation of the root mean square deviation of Cα atoms of the protein (Supporting Information Figure S2a). The MD trajectories during 40–100 ns, when the convergence was confirmed, were used for further analysis. We also employed the previously reported MD trajectories of the apo form of the X dimer (Senoo et al., 2021) and compared them with those of PhHit1‐bound X dimer.

We first focused on the difference in the angle of the EC1–EC2 domain in holo form versus that in the apo form. The binding site of PhHit1 is in the linkage region of two domains (Figure 2a), so we expected that the binding of PhHit1 might impact this angle. The dihedral angles between the two planes composed of Cα atoms from R30‐Q101‐H104 and from Q101‐H104‐E182 were calculated from the MD trajectories (Figure 3a). On average, this angle was larger in the apo form than the holo form (Figure 3b), suggesting that binding of PhHit1 to the X dimer modifies the conformational ensemble of P‐cadherin, which could alter the equilibrium between the X dimer and the strand‐swap dimer.

FIGURE 3.

FIGURE 3

Open in a new tab

PhHit1 binding alters the angle between EC1 and EC2 domains. (a) Illustration of the dihedral angle calculated from the MD trajectories. Red spheres indicate Cα atoms used for the calculation. (b) Frequency histograms of dihedral angles from MD simulation Runs 1, 2, and 3 for apo form (upper) and holo form (lower). Counts from chain A are shown in black or red bars; those from chain B in blue or yellow. The arrows indicate the dihedral angles in crystal structures of the apo form (PDB ID; 4zmq) and the holo form.

Despite the clear effect on the domain angle, both PhHit1 molecules did not remain simultaneously bound to the cavity between EC1 and EC2 during all simulations, as shown by measuring the distance between PhHit1 (C6) and Y140 (CZ), a residue that interacts with PhHit1 in the crystal structure (Supporting Information Figures S1b and S2b). When this distance became greater than 20 Å, PhHit1 was found completely outside of the pocket of interest. According to this criterion, in run 1, PhHit1 dissociated from chain A at around 80 ns; in run 2, it dissociated from chain B at around 43 ns; and in run 3, it dissociated from chain A at around 8 ns (Supporting Information Figure S2c). Considering this binding behavior, the binding stoichiometry in solution may be a single PhHit1 molecule for each X dimer. After dissociation of one molecule of PhHit1, only the other PhHit1 molecule remained bound, but the dihedral angle still differed from that in the apo X dimer even in the chain from which PhHit1 dissociated. For example, the dihedral angles in run 2 were similar for chains A and B throughout the run, although PhHit1 dissociated from chain B for most of the simulation time (Supporting Information Figures S2c and S3). This suggests that the change in domain angle is induced by binding of a single PhHit1 molecule per X dimer. Considering the greater potency of PhHit1 compared to Hit1, we speculate that Hit1 requires two molecules to show effect on the X dimer and that the smaller number of compounds required to show effect to the X dimer in the case of PhHit1 explains the difference in potency.

2.3. Asymmetric nature of PhHit1‐bound X dimer

In the crystal structure of the PhHit1‐bound X dimer, one of the two hydrogen bonds between hot spot residues (i.e., K14‐A138') was disrupted. Although no other hydrogen bond appeared to be affected during the analysis of the crystal structure, the MD simulations reflect the pseudo‐solutions conditions beyond a snapshot of a stable state, and so we expected that MD simulations could identify other hydrogen bonds that were disrupted due to PhHit1 binding. We analyzed four hydrogen bonds at the interface of the two monomers in the X dimer in the MD: K14‐A138' and K14'‐A138, which are interactions at the outer region of the interface, and D100‐Q101' and D100'‐Q101, which are located in the inner region (Supporting Information Figure S4). These hydrogen bonds are representative of the state of the interface.

To determine whether the hydrogen bonds were maintained during the MD trajectories, the distances between the hydrogen bond donor (NZ for K14, NE for Q101) and the acceptor (O for A138 and D100) were computed as a function of time. If the distance between hydrogen bond donor and acceptor was less than 3.5 Å, we assumed that the hydrogen bond was formed. In the presence of PhHit1, the hydrogen bond between K14 and A138' was stable but that between K14' and A138 was not (Supporting Information Figure S5). Furthermore, in the presence of PhHit1, the distances between hydrogen bond donors and acceptors on D100 and Q101 were larger than 5 Å at most analyzed times. In contrast, these hydrogen bonds were generally detected in the apo form (Figure 4a, Supporting Information Figure S6). When the frequency histogram of the distances between hydrogen bond donors and acceptors was generated, the disruption of K14'‐A138, D100‐Q101', and D100'‐Q101 by PhHit1 binding was clear (Figure 4b,c). For K14‐A138' and K14'‐A138, the peak patterns from chains A and B of holo form were quite distinct, whereas those for the apo form were similar (Figure 4b). Thus, the asymmetric nature of the X dimer observed in the crystal structure was also apparent in the MD trajectories. For D100‐Q101' and D100'‐Q101, the highest peak in the apo form was approximately 3 Å. In contrast, the highest peak in the holo form was observed at around 6 Å (Figure 4c), indicating disruption of both hydrogen bonds even in the inner region of the X dimer interface upon PhHit1 binding.

FIGURE 4.

FIGURE 4

Open in a new tab

Disruption of hydrogen bonds and asymmetric nature of X dimer in complex with PhHit1. (a) Upper: Superposition of representative asymmetric X dimers from MD simulation Run 1 at 61.9 ns and apo form of X dimer (PDB ID; 4zmq, colored in black). Lower: The region indicated by the dotted line is enlarged to show the disruption of K14'‐A138, D100'‐Q101, and D100‐Q101'; relevant amino acid residues are shown in stick and ribbon format. The distances of between hydrogen bond donors and receptors are shown in red, and the distance between donor and acceptor in K14‐A138' is shown in black. Yellow and blue regions indicate inner and outer regions of the PPI interface of X dimer, respectively. (b,c) Frequency histograms of the distances between hydrogen bond donors and acceptors in b) K14‐A138' and K14'‐A138 and c) D100‐Q101' and D100'‐Q101 pairs. The dotted line shows the distance of 3.5 Å, the criterion for formation of a hydrogen bond. (d) The putative mechanism by which PhHit1 indirectly disrupts the hydrogen bond between K14' and A138. The mechanism is divided into four phases (Phases I ~ IV). The black arrows indicate directions of the movements of side chains pivotal in this mechanism.

This finding prompted us to decipher how the binding of PhHit1 resulted in hydrogen bond disruption. We reasoned that disruption of K14'‐A138 resulting from the binding of PhHit1 propagated to disorganize internal hydrogen bonds such as that between D100 and Q101, so we specifically focused on the effect of PhHit1 on severing of K14'‐A138. In the MD trajectories of the holo form, we found two points when the hydrogen bond of interest was broken, and the disrupted state was maintained for a relatively long period: around 30 ns in Run 1 and around 2 ns in Run 2 (Supporting Information Figure S5). At these time points, the distance between hydrogen bond donor and acceptor in the K14'‐A138 pair (NZ in K14' and O in A138) was changed drastically (2.8–8.2 Å in Run 1 and 4.2–7.3 Å in Run 2). Upon careful examination, we found common movements at these time points. First, the interaction between PhHit1 and Y140 altered the conformation of the loop that includes Y140, and the change in the conformation of the loop caused a side chain flip of Y143 (Figure 4d, Phase I to II). This flip made room for K14' to move more freely, resulting in the disruption of the hydrogen bond with A138 (Figure 4d, Phase II to III). Once broken, this hydrogen bond did not recover possibly because of the steric hindrance between K14' and Y143 (Figure 4d, Phase IV). This mechanism suggests the importance of the conformation of the Y140‐containing loop in maintaining critical hydrogen bonds that stabilize the interface between monomers of the X dimer. This mechanism suggests that the asymmetric nature of the X dimer was triggered only by one PhHit1 molecule. Moreover, once this symmetric X dimer was induced by the binding of a single PhHit1 to one cavity, the other binding cavity of interest becomes narrower, which may make the binding of the next PhHit1 molecule difficult. For example, in Run 2 at 58 ns, PhHit1 compound came closer to the pocket, but never bound again. This seems to be because the pocket was almost closed at this time (Supporting Information Figure S2d). Therefore, the first PhHit1 molecule that binds to one of the two pockets of the X dimer could even stop the second PhHit1 molecule to bind to the other pocket.

2.4. Alterations in the conformational ensemble visualized through principal component analysis

The analyses of the angle between domains and distances between pairs of residues that form hydrogen bonds in the apo X dimer indicated that binding of PhHit1 causes changes in conformation at the residue and the domain level. To validate this and to more directly visualize such alterations in the conformational ensemble, we performed principal component analysis (PCA; Amadei et al., 1993; Kitao, 1999) of the Cartesian coordinates of Cα atoms from MD trajectories of apo and holo forms. The projection of each trajectory onto the PC1 (the first eigenvector) and PC2 (the second eigenvector) plane is shown in Supporting Information Figure S7. The contributions of PC1 and PC2 were 28% and 24%, respectively.

When this projection was depicted as a histogram, the values on the PC1 axis were distributed in a roughly monomodal manner in the case of holo form but were distributed in a bimodal manner in the case of the apo form (Figure 5a). This result indicates that the conformational ensemble in holo form is distinct from that in apo form. The PC1 value at the peak apex is smaller in holo form than in apo form (Figure 5a), indicating that there is less fluctuation in ensemble conformations in the holo form than in the apo form. The motion represented by PC1 is mainly due to bending of EC2 domains (Figure 5b, Supporting Information Movie S1). The smaller fluctuation in holo form appear to reflect restricted angles between EC1 and EC2 domains (Figure 3).

FIGURE 5.

FIGURE 5

Open in a new tab

Principal component analysis reveals altered conformational ensemble of X dimer upon PhHit1 binding. (a) Frequency histogram showing the number of projections from MD trajectories along the PC1 axis. Red and blue histograms are of holo form and apo form of X dimer, respectively. Arrows show the peak tops of each distribution. (b) The first eigenvectors drawn as black arrows onto each Cα atom in the average structure. In each average structure, chain A is colored in sky blue; chain B in pink. Top view, side view, and bottom view are shown. See also Supporting information Movie S1. (c) Frequency histogram showing the number of projections from MD trajectories along the PC2 axis in each interval. (d) The second eigenvectors drawn as black arrows onto each Cα atom in the average structure. See also Supporting information Movie S2.

A clear difference in the conformational ensemble was also observed on the PC2 axis. The values on the PC2 axis had a bimodal distribution for the holo form and a monomodal distribution for the apo form (Figure 5c). The wider distribution of PC2 values in the holo form suggested that motion was more prominent in the holo form. The motion represented by PC2 occurs within the EC1 domain and is asymmetric (Figure 5d, Supporting Information Movie S2). Considering the asymmetric nature of the disruption of hydrogen bonds, the motion represented by PC2 likely reflects breaking of hydrogen bonds. Overall, PCA suggested that upon the binding of PhHit1, the ensemble of conformations of the X dimer is altered in a manner consistent with the alternation of domain angle and hydrogen bond disruption discussed in Figures 3 and 4, respectively. The ensemble of conformations adopted upon binding of PhHit1 was less likely to sample the conformation of the apo form, which may be necessary for the transition from the X dimer to the strand‐swap dimer that ultimately mediates cell adhesion.

2.5. Effect of PhHit1 on the strand‐swap dimer

Through structural analysis and MD simulation, we draw a hypothesis that PhHit1 indirectly contributes to inhibit the strand‐swap dimer. To quantify the extent of inhibition, interaction analysis using surface plasmon resonance (SPR) was performed. Here, based on the previous report (Kudo et al., 2016; Tsukasaki et al., 2014), the construct forming the strand‐swap dimer via the X dimer was employed. The construct was immobilized on the sensor chip and the same construct was injected into the sensor chip to observe the homophilic interaction forming the strand‐swap dimer. We observed a dose‐dependent binding response (Supporting Information Figure S8), and the dissociation constant (K D) of strand‐swap dimerization was determined to be 4.7 μM (Figure 6a). This value was almost identical to the previously reported K D value measured by analytical ultra‐centrifugation (Kudo et al., 2016), suggesting that the homophilic interaction measured in this assay system provides reliable information. When 100 μM PhHit1 was added to the analyte solution, the strand‐swap dimer became less stable, that is, the apparent K D value increased to 8.3 μM (Figure 6a). This result supports our hypothesis that the binding of PhHit1 to the cavity formed in X dimer inhibited the strand‐swap dimerization. Although the difference in the K D values of the strand‐swap dimer is small, on the cell surface such small difference might be amplified because numerous molecules of cadherin are necessary to form the characteristic clusters appearing in cadherin‐mediated cell‐adhesion (Katasamba et al., 2009). In our assay, we did not perform this assay with concentrations of PhHit1 higher than 100 μM to avoid the aggregation of the compound.

FIGURE 6.

FIGURE 6

Open in a new tab

Demonstration of inhibition of the strand‐swap dimer. (a) Comparison of K D values measured by SPR in forming the strand‐swap dimer in the presence or absence of PhHit1 (N = 3). **Significant difference (p < 0.05) by Student's t‐test. (b) Schematic illustration of mechanism underlying the inhibition of strand‐swap dimer by PhHit1.

3. DISCUSSION

In this study, we determined the crystal structure of the PhHit1‐bound state of the X dimer of P‐cadherin and, using coordinates from this crystal structure and those of the previously determined structure of the apo X dimer (Kudo et al., 2016), we performed MD simulations to evaluate and compare their conformational ensembles. PhHit1 has previously been reported to inhibit cell adhesion (Senoo et al., 2021), but the mechanism of this inhibition was unknown. By comparing MD simulations of apo and holo form of the X dimer and using PCA, we demonstrated that binding of PhHit1 alters the conformational ensemble of the X dimer. The binding of PhHit1 altered the angle between EC1 and EC2 domains and disrupted hydrogen bonds, resulting in an asymmetric X dimer state. We hypothesize that the alteration of the angle between domains and the disruption of hydrogen bonds in the X dimer indirectly prevented the formation of the strand‐swap dimer (Figure 6b). This change in the conformational ensemble observed upon PhHit1 binding suggests that PhHit1 induced the formation of a conformation that was present at low frequency in the ensemble of apo form and stabilized an otherwise unfavored conformation of the X dimer. In this sense, PhHit1 worked as an ensemble modulator of X dimer to inhibit the subsequent PPI, the strand‐swap dimer. Hit1 regulates the kinetics of X dimer, while PhHit1 inhibits the strand‐swap dimer, a biologically functional homodimer. Given the higher potency of PhHit1 compared to Hit1, the compounds that can affect the formation of strand‐swap dimer may provide a more effective way to inhibit cell adhesion.

In general, the buried surface area of PPIs varies from 1000 to 6000 Å2. When this surface area is less than 2000 Å2, the contact region is limited to a single patch, while the larger interacting surface is usually composed of several patches separated by solvent‐exposed regions (Scott et al., 2016). We reason that large interacting surfaces composed of several patches could be more easily targeted by ensemble modulators than small interacting surfaces. That is because, when one of the patches fails to form the interaction with the other protein, this should provide an unfavored conformation of PPI to be stabilized by an ensemble modulator. Since it is even more difficult to obtain an orthosteric binder that blocks PPI interface with larger area, the mechanism by ensemble modulator could compensate those of the normal orthosteric PPI inhibitors when the buried surface area is large.

Typical orthosteric inhibitors mainly affect the on‐rate of protein–protein binding as the inhibitor binds to the PPI interface of one of the protein partners and directly blocks the interaction. In contrast, ensemble modulators, which stabilize an incomplete PPI interface, likely affect both the on‐rate and off‐rate because the stabilized non‐optimal conformation can arise in either association or dissociation phases of the PPI. In the case of cell adhesion mediated by classical cadherin, an inhibitor that slows both the on‐rate and off‐rate of X dimer would be ideal for the following two reasons. First, slowing the on‐rate would reduce the accumulation of the X dimer, and second, slowing the off‐rate of X dimer would prevent the transition from the X dimer to the biological active strand‐swap dimer. From this viewpoint, inhibitor screening based on kinetic parameters of a PPI should identify small molecule inhibitors targeting cadherin molecules that inhibit cell adhesion. Furthermore, the non‐standard, asymmetric state of X dimer observed in our MD trajectory could serve as a template structure for in silico docking of potential inhibitors of P‐cadherin strand‐swap dimer formation.

4. MATERIALS AND METHODS

4.1. Expression and purification of P‐cadherin recombinant proteins

The MEC12 (amino acids 1–213), which forms an X dimer but not the strand‐swap dimer and the EC12 (amino acids 1–241), which forms the strand‐dimer via the X dimer, were expressed and purified as described previously (Senoo et al., 2021). Briefly, both proteins were expressed as a fusion protein with SUMO‐tag. The fusion protein was first purified over a Ni‐NTA agarose column (QIAGEN). The eluted sample was treated with Ulp1 to remove the SUMO tag. The desired protein was collected as the flowthrough of a second Ni‐NTA chromatography. The protein was further purified using size exclusion chromatography on a Hiload 26/60 Superdex 200 column (Cytiva).

4.2. Crystallization of MEC12

For X‐ray crystallography, purified MEC12 (12.5 mg/mL) was crystallized in 0.075 M HEPES sodium salt, pH 7.5, 1.125 M lithium sulfate monohydrate, 25% v/v glycerol. The crystals were soaked in cryoprotectant solution containing 10% DMSO and 10 mM PhHit1, synthesized and stored as described (Senoo et al., 2021), for 1 h.

4.3. Data collection and refinement

Diffraction data were collected on the RIKEN Structural Genomics Beamline II (BL26B2) at SPring‐8 (Ueno et al., 2006). The KAMO programs (Evans, 2011; Kabsch, 2010; Yamashita et al., 2018) were used to process the diffraction data, and the structure was solved by molecular replacement using the reported structure of P‐cadherin MEC12 (PDB ID 4zmq) as a search model with Phenix phaser. The resultant structures were iteratively refined using Phenix refine (Adams et al., 2010) and manually rebuilt in Coot (Adams et al., 2010; Emsley et al., 2010). Final refinement statistics are summarized in Table 1. UCSF Chimera (Pettersen et al., 2004) was used to prepare the figures.

TABLE 1.

Crystallographic data collection and refinement statistics.

Data collection MEC12‐PhHit1
PDB ID 8HYI
Space group P 21 21 21
Unit cell dimensions
a, b, c (Å) 80.00, 99.50, 107.57
α, β, γ (°) 90.0, 90.0, 90.0
Wavelength (Å) 1.0000
Resolution (Å) a 47.31–2.86 (2.98–2.85)
R merge 0.252 (2.168)
R meas 0.271 (2.311)
CC 1/2 0.994 (0.545)
<I/σ(I)> 9.70 (1.04)
Completeness (%) 99.87 (99.4)
Redundancy 7.33 (7.42)
Refinement statistics
Resolution (Å) 47.31–2.85
R work 0.22
R free 0.27
No. of non‐hydrogen atoms 3360
Macromolecules 3308
Ligands 94
Solvent 0
Unique reflections 20,515 (2504)
Average B‐factor (Å2) 65.48
R. M. S. deviations from ideal
Bonds (Å) 0.011
Angles (°) 1.43
Ramachandran plot (%)
Favored region 90.57
Allowed region 8.73
Outlier region 0.71
Open in a new tab
a

Values in parentheses are for the highest‐resolution shell.

4.4. MD simulation

GROMACS 2016.3 (Abraham et al., 2015) with the CHARMM36m force field (Huang et al., 2016) was used for MD simulations of PhHit1‐bound state of X dimer. For solvation of the protein, TIP3P (Jorgensen et al., 1983) water molecules were placed in a rectangular box. The minimum distance to the edge of the box was 15 Å under periodic boundary conditions. Through the CHARMM‐GUI, 150 mM of Na+ and Cl− ions were added to imitate a salt solution. Before the MD simulation, energy minimization for 5000 steps and equilibration with the NVT ensemble (303 K) for 1 ns were conducted. Simulations were performed with the NPT ensemble at 303 K. The time step was set to 2 fs, and snapshots were saved every 10 ps through the simulations. The cutoff distances for Coulomb and van der Waals interactions was 12 Å. The particle mesh Ewald method (Darden et al., 1993) was used for the evaluation of long‐range electrostatic interactions. The LINCS algorithm (Hess et al., 1997) was used to constrain covalent bonds involving hydrogen atoms. For the calculation of domain angle and the distance between hydrogen bond donor and acceptor, trajectories were analyzed using GROMACS with the converged trajectories. For the PCA, a Python package Prody (Bakan et al., 2011) was used. The PCs were visualized using the Normal Mode Wizard of VMD (Bakan et al., 2011; Humphrey et al., 1996).

4.5. Surface plasmon resonance

For K D determination, we performed SPR using a Biacore X100 instrument (Cytiva). The construct forming the strand‐swap dimer was immobilized on Sensor Chip CM5 via amine coupling at pH 4.0. The same protein samples in 10 mM HEPES, 150 mM NaCl, 3 mM CaCl2, 0.05% Tween20, 5% dimethyl sulfoxide (DMSO) was injected on the sensor chip surface in a concentration series (10, 5, 2.5, 1.25, 0.625 μM). The contact time was 60 s, and the dissociation time was 50 s. After every cycle, the sensor chip was regenerated with a flush of the buffer. To correct the bulk response in each flow cell, the solvent correction was performed using buffer solutions containing from 4% to 6% DMSO. The dissociation constant K D was calculated using the Scatchard method in a BIA evaluation software.

AUTHOR CONTRIBUTIONS

Akinobu Senoo, Satoru Nagatoishi, Daisuke Kuroda, and Kouhei Tsumoto designed experiments, performed MD simulations, and analyzed and discussed the results. Akinobu Senoo, Sho Ito, and Go Ueno, performed crystallization experiments and processed and determined the crystal structure. Akinobu Senoo, Daisuke Kuroda, and Satoru Nagatoishi wrote the manuscript with input from Jose M. M. Caaveiro and Kouhei Tsumoto. All the authors discussed the data, reviewed the manuscript, and approved the final manuscript.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

Supporting information

Data S1: Supporting Information.

Click here for additional data file. (1.9MB, docx)

Supporting Information Movie S1. The dynamic movement shown by the first eigenvectors. Engen vectors are drawn as black arrows onto each Cα atom in the average structure. Chain A is colored in sky blue; chain B in pink.

Click here for additional data file. (1.8MB, mp4)

Supporting Information Movie S2. The dynamic movement shown by the second eigenvectors. Engen vectors are drawn as black arrows onto each Cα atom in the average structure. Chain A is colored in sky blue; chain B in pink.

Click here for additional data file. (1.5MB, mp4)

ACKNOWLEDGMENTS

The super‐computing resource was provided by Human Genome Center, the Institute of Medical Science, and the University of Tokyo. We thank Dr Y. Saito and Dr S. Sando for technical support in synthesizing PhHit1. This work was supported by JSPS Grant‐in‐Aid for Scientific Research on Innovative Areas “Bio‐metal” (19H05766 to Kouhei Tsumoto), JSPS Grants‐in‐Aid for Scientific Research (18H05425 to Satoru Nagatoishi, 20H02531 and 16H02420 to Kouhei Tsumoto), Platform Project for Supporting Drug Discovery and Life Science Research (Basis for Supporting Innovative Drug Discovery and Life Science Research [BINDS] from Japan Agency for Medical Research and Development [AMED; JP21am0101094 to Kouhei Tsumoto]), and a Grant‐in‐Aid for JSPS fellows (19J14451 to Akinobu Senoo). The synchrotron radiation experiments were performed at BL26B2 of SPring‐8 and were supported by BINDS from AMED (JP21am0101070). The SPR experiment was supported by Research Support Project for Life Science and Drug Discovery (Basis for Supporting Innovative Drug Discovery and Life Science Research [BINDS]) from AMED under Grant Number JP22ama121031. This work was partly supported by the World‐leading Innovative Graduate Study Program for Life Science and Technology, The University of Tokyo, as part of the WISE Program (Doctoral Program for World‐leading Innovative & Smart Education), MEXT, Japan.

Senoo A, Nagatoishi S, Kuroda D, Ito S, Ueno G, Caaveiro JMM, et al. Modulation of a conformational ensemble by a small molecule that inhibits key protein–protein interactions involved in cell adhesion. Protein Science. 2023;32(9):e4744. 10.1002/pro.4744

Review Editor: Aitziber L. Cortajarena

Contributor Information

Satoru Nagatoishi, Email: s-nagatoishi@g.ecc.u-tokyo.ac.jp.

Kouhei Tsumoto, Email: tsumoto@bioeng.t.u-tokyo.ac.jp.

DATA AVAILABILITY STATEMENT

Coordinates and structure factors were deposited in the Protein Data Bank under accession code 8HYI. The authors declare that the data supporting the findings of this study are available within the paper and its Supporting Information files.

REFERENCES

  1. Abraham MJ, Murtola T, Schulz R, Páll S, Smith JC, Hess B, et al. Gromacs: high performance molecular simulations through multi‐level parallelism from laptops to supercomputers. SoftwareX. 2015;1–2:19–25. [Google Scholar]
  2. Adams PD, Afonine PV, Bunkóczi G, Chen VB, Davis IW, Echols N, et al. PHENIX: a comprehensive python‐based system for macromolecular structure solution. Acta Crystallogr D Biol Crystallogr. 2010;66:213–221. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Amadei A, Linssen ABM, Berendsen HJC. Essential dynamics of proteins. Proteins. 1993;425:412–425. [DOI] [PubMed] [Google Scholar]
  4. Arkin MMR, Wells JA. Small‐molecule inhibitors of protein–protein interactions: progressing towards the dream. Nat Rev Drug Discov. 2004;3:301–317. [DOI] [PubMed] [Google Scholar]
  5. Baird L, Llères D, Swift S, Dinkova‐Kostova AT. Regulatory flexibility in the Nrf2‐mediated stress response is conferred by conformational cycling of the Keap1‐Nrf2 protein complex. Proc Natl Acad Sci U S A. 2013;110:15259–15264. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Bakan A, Meireles LM, Bahar I. ProDy: protein dynamics inferred from theory and experiments. Bioinformatics. 2011;27:1575–1577. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Darden T, York D, Pedersen L. Particle mesh Ewald: an N·log(N) method for Ewald sums in large systems. J Chem Phys. 1993;98:10089–10092. [Google Scholar]
  8. Durrant JD, McCammon JA. BINANA: a novel algorithm for ligand‐binding characterization. J Mol Graph Model. 2011;29:888–893. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Emsley P, Lohkamp B, Scott WG, Cowtan K. Features and development of coot. Acta Crystallogr D Biol Crystallogr. 2010;66:486–501. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Evans PR. An introduction to data reduction: space‐group determination, scaling and intensity statistics. Acta CrystallogrD Biol Crystallogr. 2011;67:282–292. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Fischer G, Rossmann M, Hyvo M. Alternative modulation of protein–protein interactions by small molecules. Curr Opin Biotechnol. 2015;35:78–85. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Frauenfelder H, Sligar SG, Wolynes PG. The energy landscapes and motions of proteins. Science. 1991;254:1598–1603. [DOI] [PubMed] [Google Scholar]
  13. Hammoudeh DI, Daté M, Yun MK, Zhang W, Boyd VA, Viacava Follis A, et al. Identification and characterization of an allosteric inhibitory site on dihydropteroate synthase. ACS Chem Biol. 2014;9:1294–1302. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Park J, Park E, Han S‐W, Im S‐A, Kim T‐Y, Kim W‐H, et al. Down‐regulation of P‐cadherin with PF‐03732010 inhibits cell migration and tumor growth in gastric cancer. Invest New Drugs. 2011;30:1404–1412. [DOI] [PubMed] [Google Scholar]
  15. Hess B, Bekker H, Berendsen HJC, Fraaije JGEM. LINCS: a linear constraint solver for molecular simulations. J Comput Chem. 1997;18:1463–1472. [Google Scholar]
  16. Huang J, Rauscher S, Nawrocki G, Ran T, Feig M, de Groot BL, et al. CHARMM36m: an improved force field for folded and intrinsically disordered proteins. Nat Methods. 2016;14:71–73. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Humphrey W, Dalke A, Schulten K. Sartorius products. J Mol Graph. 1996;14:33–38. [DOI] [PubMed] [Google Scholar]
  18. Jorgensen WL, Chandrasekhar J, Madura JD, Impey RW, Klein ML. Comparison of simple potential functions for simulating liquid water. J Chem Phys. 1983;79:926–935. [Google Scholar]
  19. Kabsch W. XDS. Acta Crystallogr D Biol Crystallogr. 2010;66:125–132. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Katasamba P, Carroll K, Ahlsen G, Bahna F, Vendome J, Posy S, et al. Linking molecular affinity and cellular specificity in cadherin‐mediated adhesion. Proc Natl Acad Sci. 2009;106:11594–11599. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Kitao A. Investigating protein in collective coordinate space. Curr Opin Struct Biol. 1999;9:164–169. [DOI] [PubMed] [Google Scholar]
  22. Kudo S, Caaveiro JMM, Tsumoto K. Adhesive dimerization of human P‐cadherin catalyzed by a chaperone‐like mechanism. Structure. 2016;24:1523–1536. [DOI] [PubMed] [Google Scholar]
  23. Liauw BWH, Afsari HS, Vafabakhsh R. Conformational rearrangement during activation of a metabotropic glutamate receptor. Nat Chem Biol. 2021;17:291–297. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Lo CH, Schaaf TM, Thomas DD, Sachs JN. Fluoresence‐based TNFR1 biosensor for monitoring receptor structural and conformational dynamics and discovery of small molecule modulators. Methods Mol Biol. 2021;2248:121–137. 10.1007/978-1-0716-1130-2_9 [DOI] [PubMed] [Google Scholar]
  25. Pettersen EF, Goddard TD, Huang CC, Couch GS, Greenblatt DM, Meng EC, et al. UCSF Chimera—a visualization system for exploratory research and analysis. J Comput Chem. 2004;25:1605–1612. [DOI] [PubMed] [Google Scholar]
  26. Quiniou C, Domínguez‐Punaro M, Cloutier F, Erfani A, Ennaciri J, Sivanesan D, et al. Specific targeting of the IL‐23 receptor, using a novel small peptide noncompetitive antagonist, decreases the inflammatory response. Am J Physiol Regul Integr Comp Physiol. 2014;307:R1216–R1230. [DOI] [PubMed] [Google Scholar]
  27. Ran X, Gestwicki JE. Inhibitors of protein–protein interactions (PPIs): an analysis of scaffold choices and buried surface area. Curr Opin Chem Biol. 2018;44:75–86. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Scott DE, Bayly AR, Abell C, Skidmore J. Small molecules, big targets: drug discovery faces the protein‐protein interaction challenge. Nat Rev Drug Discov. 2016;15:533–550. [DOI] [PubMed] [Google Scholar]
  29. Senoo A, Ito S, Nagatoishi S, Saito Y, Ueno G, Kuroda D, et al. Regulation of cadherin dimerization by chemical fragments as a trigger to inhibit cell adhesion. Commun Biol. 2021;4:1041. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Senoo A, Nagatoishi S, Moberg A, Babol LN, Mitani T, Tashima T, et al. Inhibition of homophilic dimerization and disruption of cell adhesion by P‐cadherin‐specific small molecules from SPR‐based assays. Chem Commun. 2018;54:5350–5353. [DOI] [PubMed] [Google Scholar]
  31. Smith MC, Gestwicki JE. Features of protein–protein interactions that translate into potent inhibitors: topology, surface area and affinity. Expert Rev Mol Med. 2012;14:e16. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Tsukasaki Y, Miyazaki N, Matsumoto A, Nagae S, Yonemura S. Giant cadherins fat and dachsous self‐bend to organize properly spaced intercellular junctions. Proc Natl Acad Sci U S A. 2014;111:16011–16016. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Ueno G, Kanda H, Hirose R, Ida K, Kumasaka T, Yamamoto M. RIKEN structural genomics beamlines at the SPring‐8; high throughput protein crystallography with automated beamline operation. J Struct Funct Genomics. 2006;7:15–22. [DOI] [PubMed] [Google Scholar]
  34. Van Roy F. Beyond E‐cadherin: roles of other cadherin superfamily members in cancer. Nat Rev Cancer. 2014;14:121–134. [DOI] [PubMed] [Google Scholar]
  35. Vieira AF, Paredes J. P‐cadherin and the journey to cancer metastasis. Mol Cancer. 2015;14:178. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Wang Y, Xu C, Wang Y, Hong Q, Zhang C, Li Z, et al. Conformational dynamics of the Beta and Kappa SARS‐CoV‐2 spike proteins and their complexes with ACE2 receptor revealed by cryo‐EM. Nat Commun. 2021;12:7345. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Wei G, Xi W, Nussinov R, Ma B. Protein ensembles: how does nature harness thermodynamic fluctuations for life? The diverse functional roles of conformational ensembles in the cell. Chem Rev. 2016;116:6516–6551. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Wells JA, Mcclendon CL. Reaching for high‐hanging fruit in drug discovery at protein – protein interfaces. Nature. 2007;450:1001–1009. [DOI] [PubMed] [Google Scholar]
  39. Yamashita K, Hirata K, Yamamoto M. KAMO: towards automated data processing for microcrystals. Acta Crystallogr D Struct Biol. 2018;74:441–449. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Zhang CC, Yan Z, Zhang Q, Kuszpit K, Zasadny K, Qiu M, et al. PF‐03732010: a fully human monoclonal antibody against P‐cadherin with antitumor and antimetastatic activity. Clin Cancer Res. 2010;16:5177–5188. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Data S1: Supporting Information.

Click here for additional data file. (1.9MB, docx)

Supporting Information Movie S1. The dynamic movement shown by the first eigenvectors. Engen vectors are drawn as black arrows onto each Cα atom in the average structure. Chain A is colored in sky blue; chain B in pink.

Click here for additional data file. (1.8MB, mp4)

Supporting Information Movie S2. The dynamic movement shown by the second eigenvectors. Engen vectors are drawn as black arrows onto each Cα atom in the average structure. Chain A is colored in sky blue; chain B in pink.

Click here for additional data file. (1.5MB, mp4)

Data Availability Statement

Coordinates and structure factors were deposited in the Protein Data Bank under accession code 8HYI. The authors declare that the data supporting the findings of this study are available within the paper and its Supporting Information files.

Articles from Protein Science : A Publication of the Protein Society are provided here courtesy of The Protein Society

RESOURCES