Results

Structure of the N-Terminal Domain of the MERS-CoV N Protein
We determined the crystal structure of MERS-CoV N-NTD by molecular replacement (MR) using the structure of HCoV-OC43 N-NTD (PDB ID: 4J3K) as the search model.24 The final structure was refined to R-factor and R-free values of 0.26 and 0.29, respectively, at a resolution of 2.6 Å (Table S1). Each asymmetric unit contained four N-NTD molecules assembled into two identical dimers with an overall RMSD of 0.28 Å between the dimers (Figure S1A,B). The monomers shared a similar structural core preceded by a flexible region (Figure S1D). The core consisted of a five-stranded antiparallel β-sheet sandwiched between loops arranged in a right-handed, fist-shaped structure conserved among the CoVs.25 In our structure, however, the loop connecting strands β2 and β3 protruding out of the core into other CoV N proteins was absent. Unlike the reported structures that have a monomeric conformation, our structure was atypically dimeric.
Figure 1B shows the details of the interactions in the MERS-CoV N protein dimer. We named the units monomer 1 and monomer 2 (Figure 1A). According to the amino acid composition of the binding site on monomer 2, we divided the dimeric interface into two areas: one located on the N-terminus flexible region and the other on the loop between β4 and β5 of the N protein. In the first area, W43, N66, N68, Y102, and F135 of monomer 1 generated a conserved hydrophobic pocket permitting the side chain of M38 of monomer 2 to enter this hole by a hydrophobic contact (Figure 1C,D). H37 and N39 of monomer 2 were packed against W43 and F135 of monomer 1, respectively, and contributed to the hydrophobic interaction. The side chains of N39 of monomer 2 formed one hydrogen bond with the N68 backbone in monomer 1 at a distance of 2.6 Å. The second area was relatively more hydrophilic. The main chain oxygens of G104, F135, and T137 of monomer 2 formed three hydrogen bonds with the side chains of Q73 and T134 of monomer 1 at distances of 3.8, 3.2, and 3.7 Å, respectively. The side chain of N139 on monomer 2 formed a hydrogen bond with the main chain oxygen of T137 on monomer 1 at a distance of 3.6 Å. The interactions of the first and second areas comprise buried surface areas (BSA) of 289 and 103 Å2, respectively. The small surface area buried at the interface accounts for ∼5 kcal mol–1 binding energy,26 which translates to a dissociation constant of ∼200 mM. Thus, the dimer described here is unique in that it is non-native and relies on vector-fusion residues (H37 and M38) to maintain its dimeric status. This property may also explain why the present structure has an oligomeric status different from previously reported structures for the CoV N protein.24,27−30 We used cross-linking experiments to analyze the oligomeric capacity of MERS N-NTD containing the vector-fusion residues in solution. MERS N-NTD had a dimeric conformation in solution. Our structure indicated that W43 played an essential role in forming the hydrophobic pocket accommodating the vector-fusion residues and, therefore, mediated the N-NTD dimer formation. The W43A mutation significantly reduced the oligomeric tendency of N-NTD (Figure S1C). This further supports that the “exogenous residues” encoded by the vector backbone mediated the formation of the non-native dimer. We also superimposed the previously published structure of MERS-CoV N-NTD (PDB ID: 4ud1)27 containing a native N-terminal flexible region with our dimer structure. The side chain of N38 in the native structure could not interact with the hydrophobic pocket as the former is hydrophilic and short (Figure S1E). Thus, it may be possible to utilize small compounds to replace the vector-fusion residues and stabilize the PPI through hydrophobic interactions.
Figure 1  Structure and sequence of MERS-CoV N-NTD. (A) Overall structure of the MERS-CoV N-NTD dimer containing monomers 1 and 2 is depicted as a cartoon and colored yellow and green, respectively. Residues involved in dimerization are shown as sticks and highlighted in (B). (B) Interactions among MERS-CoV N-NTDs. Dimerization is mediated mainly by vector-fusion residues interacting with the conserved hydrophobic regions on the core structure (first area) along with the residues surrounding it (second area). Interacting residues on monomers 1 and 2 are labeled in black and blue, respectively. Vector-fusion and conserved hydrophobic regions are colored cyan and red, respectively. The color of all other interacting residues is the same as that for each monomer in (A). Polar contacts are indicated with red dashed lines. (C) Upper panel: close-up of the interacting region of vector-fusion residues. The surface was colored according to the hydrophobicity level at the protein surface. Vector-fusion residues (black) are shown as sticks to emphasize the hydrophobic pocket. Lower panel: 2D diagrams of the interaction between the hydrophobic pocket and the vector-fusion residues. The latter are labeled in black. The hydrophobic contacts are indicated with black dashed lines. (D) Sequence alignment of various CoV N proteins in the N-terminal region. Red letters indicate strictly conserved residues. Cyan indicates conservative substitution sites. Hydrophobic regions involved in unusual dimerization are indicated by black triangles.

Direct Targeting of the Non-native Dimer Interface for Antiviral Screening
We performed a structure-based virtual screening by targeting W43 in the hydrophobic pocket of the N-NTD dimeric interface. H37 and M38 were removed from the template to identify compounds that could replace the vector-fusion residues and, therefore, contribute to the stabilizing effect (Figures 2A and S2A,B). We chose the highest-scoring hits, listed in Table S2, based on shape complementarity, the presence of aromatic moieties, and the ability to stack onto W43 of N-NTD. Because the formation of the non-native dimers was primarily mediated by hydrophobic interactions in our structure (Figure 1C,D), we next considered the hydrophobic complementarity between the acquired ligands and N-NTD in the form of the lipophilic match surface (SL/L).31 We also took into account the ability of the drug to permeate cells by aiming for lower topological polar surface areas (TPSA). Based on the above criteria, three compounds were finally chosen for further study. Benzyl-2-(hydroxymethyl)-1-indolinecarboxylate (P1) and 5-benzyloxygramine (P3) had higher SL/L and docking scores and lower TPSA. The clinical drug etodolac (P2) had a comparable SL/L but a lower docking score. It too was selected as a candidate. However, only P3 induced a comparatively larger blue shift in the intrinsic N-NTD fluorescence spectrum, indicating that the microenvironment surrounding the tryptophans of the protein increased in rigidity and hydrophobicity in the presence of P3.32 The result also indicated that P3 bound more tightly to the N protein than P1 or P2 by interacting with the W43 pocket (Figure 2B). Fluorescent thermal stability assays disclosed that the N-NTD denaturation melting temperature had increased from 42 to 45 °C when P3 was added. The sigmoidal melting curve for MERS-CoV N-NTD changed in the presence of P3. The delay in protein denaturation suggests that P3 stabilized the MERS-CoV N-NTD dimer structure (Figure 2C). We then measured the cytotoxic concentration (CC50) and effective concentration (EC50) for each compound using Vero E6 cells infected with MERS-CoV. Table 1 shows that P3 had a favorable therapeutic index among the lead compounds tested in this study. Therefore, P3 is an excellent candidate inhibitor against MERS-CoV.
Figure 2  Compound P3 was a potent stabilizer of the MERS-CoV N protein. (A) Schematic depicting the rationale used in designing the allosteric stabilizer of this study. An orthosteric stabilizer may be used to bind to the non-native interaction interface of the N protein and stabilize the abnormal interaction between proteins. (B) Conformation and (C) stability analyses were performed based on the FL spectra of NTD (1 μM) incubated with P1–P3 (10 μM) for 1 h with a buffer consisting of 50 mM Tris-HCl (pH 8.3) and 150 mM NaCl.
Table 1  CC50 and EC50 and Therapeutic Indexes of Lead Compounds
quantal dose–response relationship (μM)
compound  CC50a  EC50b  TIc (CC50/EC50)
P1  459.69  >100  NAd
P2  569.77  >100  NAd
P3  805.32  32.1  25.1
a  CC50: Half maximal toxicity concentration.
b  EC50: Half maximum effective concentration.
c  TI: Therapeutic index.
d  NA: Nonavailable.

Structural Model of P3-Induced MERS-CoV N Protein Aggregation
We used SAXS to assess the effects of P3 on the full-length MERS-CoV N protein structure. The fitted distance distribution function of the protein with and without P3 are shown in Figure 3A. P3 increased the maximum dimension (Dmax) and radius of gyration (Rg) of the protein from 207 to 230 Å and from 58 to 65 Å, respectively. Thus, the size of the MERS-CoV N protein in solution was altered upon binding to P3.
Figure 3  P3-induced abnormal aggregation on the full-length MERS-CoV N protein. (A–E) SAXS analysis of the full-length MERS-CoV N protein. (A) Normalized results from GNOM showing pairwise distance distribution P(r) and maximum distance. The radius of gyration fitted to 207 and 230 Å for the N protein and the N-P3 complex, respectively. “r” represents pairwise distances. (B, C) Scattering profiles of the N protein (B) and the N-P3 complex (C) and normalization fitting with GNOM (dashed lines). (D, E) Representative models of the N protein (D) and the N-P3 complex (E) generated by CRYSOL simulations of the SAXS data. Only α carbons are shown. NTD (yellow), CTD (green), and disorder region (cyan). (F, G) Conformation (F) and stability (G) analyses based on FL spectra of the MERS-CoV N protein (1 μM) incubated with P3 (10 μM) for 1 h in a buffer consisting of 50 mM Tris-HCl, 150 mM NaCl (pH 8.3). (H) Schematic of the P3 inhibition mechanism. Left panel: in the absence of RNA, N proteins organize as a dimeric building block contributed by N-CTD dimerization. Middle panel: P3 promoted the dimerization of N-NTDs from different building blocks, by which the distance between CTD cuboids was shortened and N protein aggregation occurred. Right panel: octameric conformation of building blocks buried in the RNA-binding surface of N-CTDs. It hindered the formation of filamentous ribonucleocapsids. The presence of multiple intrinsically disordered regions in the N protein precluded the determination of its structure by X-ray crystallography. Instead, we used rigid body modeling of the SAXS data with the N-terminal domain (NTD; solved in this study) and the C-terminal domains (CTD, PDB ID: 6G13).23 In this way, we obtained structural models for the free N protein and its complex with P3 (Figure 3B,C). Excellent fits were obtained. Representative structural models for the full-length protein without and with P3 are shown in Figure 3D,E, respectively. The free N protein formed a tetramer through CTD with the NTD freely hanging in solution (Figure 3D). The conformation of the solution was consistent with structures previously reported for other CoV N proteins.33 The N-P3 complex formed a compact hexadecamer with a sunburst configuration (Figure 3E). The CTDs formed a central ring and non-native NTD dimers formed “spikes” protruding from the ring. Consistent with ligand-induced aggregation, we observed a “blue shift” in the fluorescence spectrum of the full-length MERS-CoV N protein in the presence of P3 (Figure 3F). The addition of P3 also delayed N protein thermal denaturation and changed the shape of the denaturation curve, further suggesting that large protein aggregates formed in the presence of P3 (Figure 3G). The structure explains how N-NTD dimerization decreased MERS-CoV viability. The N protein packages the viral genome into an RNP complex. Several models for N-CTD dimer assembly have been proposed for the formation of filamentous RNPs.33 All of the proposed interfaces between N-CTD dimers occurred on the side-faces of the CTD cuboid perpendicular to the proposed RNA-binding surface (Figure 3H). Combinatorial use of any region on the side-faces of the CTD dimer cuboid may facilitate manipulation of the RNP length and curvature without obstructing the RNA-binding surface.28,34 However, the SAXS results indicated that N-CTD aggregation occurred on the β-sheet floor of the CTD cuboid. For this reason, the RNA-binding surface of the CTD is occluded by the neighboring CTD on the ring and by the non-native NTD dimer making direct contact with the CTD (Figures 3H and S3). In addition, the CTD cuboids in the aggregation naturally form a topologically closed octamer, leaving no open ends for further addition of CTD cuboids to form a long filamentous RNP. Both the loss of the RNA-binding surface and the inability to incorporate further N protein molecules beyond an octamer may inhibit the formation of the RNP. Therefore, P3 may inhibit MERS-CoV RNP formation by inducing N protein aggregation.

P3 Inhibits MERS-CoV by Inducing N Protein Aggregation
We demonstrated that P3 had the best characteristics as a therapeutic candidate. To determine the anti-MERS-CoV activity of P3 in the cell, the effects of P3 incubation on extracellular viral titers and intracellular viral RNA levels were assessed by plaque assays on Vero E6 cells (Figure 4A) and by RT-qPCR (Figure 4B), respectively. At 50 μM, P3 marginally affected the viral titer after 48 h but suppressed viral RNA replication by 40%. At 100 μM, P3 halted both viral production and replication after 48 h. This result proved the capacity of P3 as an antiviral compound. We then examined MERS-CoV N protein distribution and expression in the infected cells with or without P3 treatment to confirm our SAXS findings. Immunofluorescence microscopy (Figure 4C) showed condensation of the intracellular N protein fluorescence signal in infected Vero E6 cells treated with 50 μM P3. Thus, P3 may induce intracellular N protein aggregation. At 100 μM, P3 suppressed N protein expression in most cells. However, a few presented with intense N protein signals. P3 may have restrained the MERS-CoV N proteins inside the infected cells that promoted the formation of new virions that could not be released. In this way, the adjacent cells could not be infected with MERS-CoV. The data, therefore, suggest that P3 may inhibit MERS-CoV by inducing abnormal aggregation of the N protein inside the cell. This finding is consistent with the results of our structure-based assays.
Figure 4  Compound P3 was a potential inhibitor against MERS-CoV. (A, B) Viral titers (A) and RNA (B) of MERS-CoV measured by plaque assay and RT-qPCR, respectively, decreased after P3 treatment for 48 h. Relative RNA levels were determined by comparing MERS alone at each time point. GAPDH RNA was the internal control. All values are presented as mean ± SE (standard error of mean). One-way Anova was used for statistics (*p < 0.05, **p < 0.01, ***p < 0.001). (C) MERS-CoV nucleocapsid protein decreased after 48 h P3 treatment. Nucleocapsid protein expressions (red) were examined under a confocal microscope at ×680. Nuclei were stained blue with DAPI.

Crystal Structure of MERS-CoV N-NTD Complexed with Potent Compounds
We attempted to obtain crystals of MERS-CoV N-NTD in complex with compounds P1, P2, and P3 by cocrystallization or ligand-soaking. With the exception of P2, the complex structures of N-NTD with P1 and P3 were solved at resolutions of 3.09 and 2.77 Å, respectively (Table S1). The overall structures of the complexes resembled that of apo-MERS-CoV. Both complexes revealed well-defined unbiased densities in the dimer interface and permitted detailed analysis of the interactions between the compounds and MERS-CoV N-NTD (Figure 5). The interactions between the N protein and each compound were calculated with the Discovery Studio Client (v19.1.0.18287). Most interactions were hydrophobic contacts, which were consistent with our selection rationale. In the P1 complex, N68, F135, and D143 on monomer 1 and V41, G106, P107, and T137 on monomer 2 packed against P1 to create a dimer (Figure 5A). In addition, two nonbonding interactions were detected between P1 and the monomers. There was a π-anion interaction between the benzene ring of the P1 indoline moiety and D143 of monomer 1. There was also a π-donor hydrogen bond between the other P1 benzene ring and the T137 side chain of monomer 2 (Figure 5B). Relative to P1, P3 bound more deeply into the dimer interface and interacted with a larger number of residues on both N-NTD monomers. The amino acid composition of this binding region was W43, N66, N68, S69, T70, N73, and F135 on monomer 1 and V41, G104, T105, G106, A109, and T137 on monomer 2. These residues along with P3 generated a massive hydrophobic driving force allowing the proteins and ligands to pack against each other and stabilize the dimeric conformation of the N protein (Figure 5C). Several nonbonding interactions were also observed at the P3-binding site. These included the interaction between the P3 benzene ring and N68 of monomer 1 and A109 of monomer 2 via π-lone pair and π-alkyl interactions. The dimethylaminomethyl moiety of P3 was a major source of nonbonding interactions. Three π-cation interactions formed between this moiety and the aromatic groups of W43 and F135 in monomer 1. This moiety also formed a π-lone pair interaction with N66 and a π-sigma interaction with W43 of monomer 1 (Figure 5D). The structural analyses explain the comparatively stronger binding of P3 to N-NTD (Figure 2B) and corroborated the thermal stabilization effects (Figure 2C) and antiviral activities (Table 1) of the compounds.
Figure 5  Structures of MERS-CoV N-NTD complexed with potent compounds. The structures were solved using HCoV-OC43 N-NTD (PDB:4J3K) as the search model.24 Left panel: (Upper) structural superimposition of the MERS-CoV N-NTD:P1 complex (monomers 1 and 2 are in purple and pink, respectively) and the MERS-CoV N-NTD:P3 complex (monomers 1 and 2 are in brown and green, respectively) with compounds depicted as stick structures. (Lower) Interactions involving vector-fusion residues in the non-native dimer of the apoprotein shown for comparison with (A) and (B). Color is the same as in Figure 1A. Right panel: detailed interactions among MERS-CoV N-NTD and P1 (A, B) and P3 (C, D). Different Fo–Fc maps were contoured at ∼2.5 σ. (A) Detailed stereoview of interactions at the P1-binding site. The color of each monomer is the same as in the left panel. Residues constructing the P1-binding pocket are labeled and showed as sticks. (B) Schematic of P1 bound to MERS-CoV N-NTD. Hydrophobic contacts between P1 and each monomer are displayed as dashed lines. Nonbonding interactions are indicated by cyan arrows. (C) Detailed stereoview of interactions at the P3-binding site. The color of each monomer is the same as in the left panel. Residues belonging to the P3-binding pocket are labeled and shown as sticks. (D) Schematic of P3 bound to MERS-CoV N-NTD. Hydrophobic contacts between P3 and each monomer are displayed as dashed lines. Nonbonding interactions are indicated by red arrows.