J Mol Model (2014) 20:2192 DOI 10.1007/s00894-014-2192-x
ORIGINAL PAPER
Mutations in herpes simplex virus gD protein affect receptor binding by different molecular mechanisms Joachim D. Stump & Heinrich Sticht
Received: 18 November 2013 / Accepted: 24 February 2014 # Springer-Verlag Berlin Heidelberg 2014
Abstract Glycoprotein D (gD) is an essential protein of herpes simplex virus-1 (HSV-1) that targets the structurally unrelated receptors HVEM and nectin-1. Receptor binding of gD is accompanied by intramolecular structural rearrangements including the detachment of the C-terminus or formation of an N-terminal hairpin structure. We have investigated several gD mutations that were reported to affect receptor binding affinity or specificity in order to identify their molecular mode of action. Molecular dynamics simulations and subsequent energetic analyses of the gD-receptor complexes reveal that some mutations (M11A, N15A, L28A, T29A) play a more prominent role for HVEM binding than for nectin-1 binding, thereby conferring specificity to receptor recognition. However, our studies show that mutations can also affect the intramolecular structural rearrangement processes in gD. W294A and Q27A mutations facilitate the detachment of the C-terminus, and Q27A additionally hampers the formation of an intramolecular hairpin in gD that is exclusively established upon HVEM binding. The finding that a Q27A mutation affects multiple steps of the receptor binding process offers a molecular explanation for its enhanced nectin-1 affinity and the pronounced receptor specificity. This study also indicates that an inspection of the gD-receptor interfaces alone may be insufficient for predicting the effect of novel mutations that alter receptor specificity. Instead, such an analysis will additionally require to assess the effect of candidate mutation on the preceding steps of gD activation.
This paper belongs to a Topical Collection on the occasion of Prof. Tim Clark’s 65th birthday J. D. Stump : H. Sticht (*) Bioinformatik, Institut für Biochemie, Friedrich-Alexander-Universität Erlangen-Nürnberg, Fahrstraße 17, 91054 Erlangen, Germany e-mail: [email protected]
Keywords Glycoprotein D . HSV-1 . Molecular dynamics . Mutational analysis . Protein interaction
Introduction Herpes simplex virus-1 (HSV-1) belongs to the subfamily of alphaherpesviruses and has a high prevalence in most human populations. Clinical symptoms range from oral lesions with watery blisters to more severe diseases like meningitis and encephalitis. After initial infection the virus persists lifelong and can reactivate any time possibly due to stress factors [1, 2]. One essential protein for the infection process is HSV glycoprotein D (gD), which can use two structurally unrelated cell surface proteins as receptors. The first receptor, herpesvirus entry mediator (HVEM), belongs to the tumor necrosis factor receptor family [3] and is generally found on immune cells. The second receptor recognized by gD is nectin-1, which is a homophilic adhesion molecule present on the surface of neurons, keratinocytes, and epithelial cells [4–6]. It has also been possible to engineer mutant gD molecules that are able to recognize other receptors than those mentioned above resulting in an altered cell tropism of HSV-1 [7–11]. The ability of gD molecules to recognize structurally dissimilar receptors renders this protein a potential candidate to be included in a viral vector system for gene therapy that allows to target distinct cell types. Due to its interesting properties, the structures of unbound gD and of gD-receptor complexes were the subject of numerous studies. gD itself consists of 369 amino acids with a 316 residue ectodomain followed by a transmembrane helix (residues 317–339) and a short intracellular C-terminus. Residues 307–369 are not necessary for functionality, as experiments showed that C-terminally truncated forms of gD were able to bind their receptor [12, 13] and even restore the infectivity of a gD-deficient virus [14, 15]. The gD ectodomain can roughly be divided into the following regions: The N-terminus (residues 1–37) involved in receptor
2192, Page 2 of 11
binding, the globular core domain (residues 38–267), the flexible profusion domain (PFD; residues 268–285), and a short C-terminal stretch (residues 286–306) that precedes the transmembrane helix [14, 16]. Up to present, crystal structures are available for unliganded gD [14, 17], as well as for gD in complex with its receptors HVEM [17] and nectin-1 [18, 19]. These structures revealed that receptor binding requires major structural changes of the gD N- and/or C-terminus suggesting that receptor binding is a multi-step-process [14]. This process can be dissected into several major steps, which are shown in Fig. 1: Unliganded gD (Fig. 1, left) adopts a closed conformation, in which the Cterminus tightly interacts with the gD core and with residues 23–37 of the N-terminus, whereas residues 1–22 are flexible and not resolved in the crystal structure. Receptor binding requires the detachment of the C-terminus, thereby uncovering the gD core and N-terminal residues, which thereby become available for further interactions. In case of nectin-1, the respective conformation is directly recognized by the receptor (Fig. 1, top right). In the respective complex, the N-terminal residues 1–22 and the C-terminal residues 253–306 remain disordered. In contrary, interaction of gD with HVEM requires a conformational rearrangement of the N-terminus. In this step, residues 1–22 of gD become ordered and align in an antiparallel fashion with residues 23–36 to form an intramolecular hairpin structure where residues 1–17 lie within the same groove as the C-terminus in the closed gD conformation. Both stretches of the N-terminal hairpin together constitute the HVEM binding site (Fig. 1, bottom right). After the first crystal structure of the gD-HVEM complex became available in 2001 [17], numerous mutational studies have been performed to obtain a more detailed understanding of receptor binding [16, 20–26]. By experimental alanine scans, several mutations have been identified, which inhibit Fig. 1 Schematic presentation of the individual steps of HSV gD structural rearrangement upon receptor binding. Unliganded gD adopts a closed conformation (left), in which the C-terminus (red) is attached to the protein core (gray). The N-terminus (residues1-37; green) is disordered for residues 1–22 as indicated by a dashed line. Detachment of the C-terminus from the core leads to an open gD conformation. This form can interact with nectin-1 (top right) without rearrangement of the N-terminus. Alternatively, the N-terminus can form a hairpin structure (‘hairpin-gD’), which constitutes the HVEM binding site (bottom right)
J Mol Model (2014) 20:2192
HVEM binding (M11A, N15A, L25A, Q27A, L28A, T29A) [20]. Unexpectedly, the L25A, Q27A, L28A, and T29A mutations still bind nectin-1, although these residues are located in a sequence stretch that is both involved in nectin-1 and HVEM binding. Even more striking, the Q27A mutation results in an enhanced nectin-1 affinity. This specific effect on nectin-1 affinity is clearly distinct from the effect of the W294A mutation, which nonspecifically enhances both nectin-1 and HVEM binding ∼100-fold [14]. The results above demonstrate that some of the mutational effects can hardly be rationalized based on the static crystal structures alone. Therefore, we employed molecular dynamics (MD) simulations to investigate the effect of different mutations in more detail. There was a particular focus on the Q27A mutant because it exhibits the most prominent effect on receptor binding specificity. MD simulations were performed for unbound gD and the two gD-receptor complexes. Furthermore, in silico alanine scans were used to quantify the effect of the mutations. These studies were complemented by explicit simulations of selected mutants to monitor larger-scale structural changes. Our simulations demonstrate that mutations do not only affect receptor binding itself but also the preceding steps of gD’s structural rearrangement, thereby explaining the experimentally observed receptor binding selectivity. These findings should be helpful for a more detailed understanding of the gD structural changes and for the design of gD mutants that exhibit an altered receptor specificity.
Methods In the present work, a total of four different gD conformations (Fig. 1) were studied by MD simulations: Unliganded gD in the closed and hairpin conformation, as well as liganded gD in
J Mol Model (2014) 20:2192
complex with HVEM and nectin-1. Simulation of closed-gD was based on one monomer from PDB entry 2C36 [14]. Cysteine 307, which was inserted to stabilize a dimer for crystallization, was deleted from this structure. The gDnectin complex was taken from PDB entry 3SKU [18], and nectin-1 was truncated after its first Ig domain leaving residues 32–144 for simulation. The complex crystal structure 1JMA [17] was used for simulation of the gD-HVEM complex and also for the simulation of hairpin-gD after deletion of the HVEM moiety. Missing residues in the structures of closed-gD (residues 257–267) and gD-HVEM (residues 93–94 of HVEM) were modeled by ModLoop [27, 28]. N- and C-termini were capped with acetyl and N-methylamide, respectively. Full-atom MD simulations were performed for each system with AMBER 10 [29] using the ff99SB force field [30] and the particle mesh Ewald summation method [31, 32]. Charges were neutralized by adding sodium or chloride counter ions, respectively. A rectangular water box with TIP3P water model [33] was generated with at least 10 Å distance between any atom of the solute and the end of the periodic box. Furthermore, the SHAKE procedure [34] was applied to constrain all bonds containing hydrogen atoms. All structures were minimized in three subsequent steps with the SANDER module of AMBER followed by a 20 ps equilibration of the systems to 310 K using a previously established protocol [35–38]. Nonbonded interactions were calculated with a cutoff of 10 Å and updated every 15 steps. The following production stage was used for gathering data. Coordinates were saved every 2 ps in a trajectory over 50 ns for the complex structures (gD-HVEM, gD-nectin) and 100 ns for unbound gD (closed-gD and hairpin-gD). Clustering was performed using ptraj from AMBER Tools [39]. We applied the average-linkage algorithm with RMSD as distance metric to generate three clusters over simulation [40]. For the analysis of unbound hairpin-gD, the representative structure was taken from the largest cluster (Q27A >87 %, WT>94 %). For visualization and structural analysis VMD 1.91 [41], Sybyl [42], Xmgrace [43], and AMBER Tools 1.4 were used [39]. The molecular mechanics generalized Born surface area (MM/GBSA) method implemented in AMBER 11 [39] was applied for energetic analysis. GB model 2 was selected in combination with recommended bondi settings for atomic radii [44, 45]. In solution, the binding free energy ΔG of a protein and its ligand is defined as: ΔG ¼ Gcomplex − Gprotein þ Gligand : Each free energy term is the sum of the molecular mechanics (MM) interaction energy (EMM) and a solvation term Gsol: ΔGb ¼ ΔE MM þ ΔGsol −TΔS: ΔEMM is calculated with the sander module of AMBER. It represents the MM interaction energy between protein and its
Page 3 of 11, 2192
ligand and comprises bond, angle, and torsion-angle energies as well as electrostatic and van der Waals energies. The solvation term ΔGsol is composed of polar and nonpolar energy contributions: ΔGsol ¼ ΔGele þ ΔGnonpolar : Nonpolar contributions were calculated as a function of the solvent accessible surface area (SASA): ΔGnonpolar ¼ γSASA þ b where γ=0.0072 kcal mol−1 Å−2 and b=0.00 kcal mol−1. Dielectric constants for solute and solvent were set to 1 and 80, respectively. MM/GBSA energetic analyses were performed over a timeframe of 40 ns for the gD-receptor complexes (11th– 50th ns), and 90 ns for unliganded gD simulations (11th– 100th ns). Snapshots were taken at a time interval of 20 ps. Intramolecular interaction energies of the C-terminus in unliganded gD were calculated between residues 23–288 and 291–306. The standard deviation σ and the number of snapshots N were taken for calculation of standard errors (SE) of the ΔG values: σ SE ¼ pffiffiffiffi : N The relative change in free energy of binding (ΔΔGb) for the selected gD mutations was calculated by a singletrajectory in silico alanine scan. In the scan, the structures from the trajectory were mutated by truncating the side-chains of the residues one at a time. For the calculation, we followed the standard protocol described by Massova and Kollman [46], which assumes that the entropy of the mutant and the wildtype do not differ significantly, and therefore entropic contributions can be neglected in the calculation of ΔΔGb. Standard errors for ΔΔGb were calculated according to the error propagation law.
Results and discussion Effect of mutations on the interaction of gD with HVEM and nectin-1 Since most mutations are located in the surface patch of gD that interacts both with HVEM and nectin-1 (Fig. 2), we first analyzed the effect of the mutations on receptor interaction and binding affinity. For that purpose, gD was simulated in complex with its receptors to analyze similarities and differences in interaction with HVEM and nectin-1. The root mean square deviation (RMSD) is ∼2 Å over the simulation time (Fig. 3). The two RMSD peaks for gD-
2192, Page 4 of 11
J Mol Model (2014) 20:2192
Fig. 2 Structure of the gD-HVEM (a) and gD-nectin (b) complexes indicating the residues investigated by the in silico alanine scan (spacefilled presentation). HVEM and nectin-1 are depicted in orange and blue, respectively, and the residues considered for mutation are colored
magenta (M11, V24, L28) and yellow (N15, Q27, T29). M11 and N15 are not shown for the gD-nectin complex, because they are not involved in the interaction and therefore not resolved in the respective crystal structure
HVEM at 2.5 ns and for gD-nectin at 13 ns arise from reversible motions of flexible regions that are distant from the binding interface. For gD-HVEM this peak can be ascribed to some flexible regions in HVEM (residues 4–13, 52–60). For gD-nectin-1 the peak in the RMSD arises from motions of the flexible loops of gD (region 83–93, 118–126). Apart from these local fluctuations both complexes are conformationally stable and no significant rearrangements are observed. These simulations thus provide a basis to assess the role of individual residues for complex formation. For that purpose, an in silico alanine scan was performed for several residues that were known to abrogate HVEM binding from experiment [20]: M11, N15, Q27, L28, and T29. Consistent with the experimental finding, MM/GBSA analysis reveals that a replacement of these residues by alanine decreases binding affinity by 2–4 kcal mol−1 (Table 1). Furthermore, the V24A mutation, which does not disrupt the gD-HVEM interaction,
also has a much smaller effect of ∼1 kcal mol−1 according to the computational alanine scan. For the gD-nectin system, the in silico alanine scan shows that mutation of V24, L28, or T29 has only a minor energetic effect on the interaction (Table 1). This is again consistent with the experiment, showing that the respective mutations do not abolish nectin-1 binding. The predicted loss in binding affinity of more than 2 kcal mol−1 for the Q27A mutation, however, is in strong disagreement with the experimental observation of an enhanced nectin-1 binding for this mutant [20, 23]. Thus, an energetic analysis of the gD-receptor complexes offers a plausible explanation for the experimentally observed effects of all mutations, with the exception of Q27A suggesting that the latter mutation exerts its effect by a more complex mechanism. To clarify the role of Q27 for binding in more detail, we also investigated the gD conformational rearrangement steps, which occur upon the receptor binding (Fig. 1).
Fig. 3 Backbone RMSD of the gD-HVEM (orange) and gDnectin (blue) complexes. The following residues were included in the RMSD calculation: gD (residues 23–252), HVEM (residues 4–60, 72–79), nectin-1 (residues 32–144)
J Mol Model (2014) 20:2192
Page 5 of 11, 2192
Table 1 Results of an in silico alanine scan performed for the gD-HVEM and gD-nectin complexes. ΔG and ΔΔG values are in kcal mol-1; SE denotes the standard error. No values were calculated for M11 and N15 in the gD-nectin complex, because these residues are not involved in the interaction and therefore not resolved in the respective crystal structure gD-HVEM
WT M11A N15A V24A Q27A L28A T29A
Table 2 Results of an in silico alanine scan performed for the closed conformation of gD. Intramolecular interaction energies were calculated between residues 291–306 of the C-terminus and residues 23–288 of the gD core. ΔG and ΔΔG values are in kcal mol−1; SE denotes the standard error Closed gD
gD-nectin
ΔG
ΔΔG
SE (ΔΔG)
ΔG
ΔΔG
SE (ΔΔG)
−49.95 −45.79 −47.69 −48.81 −47.86 −47.42 −46.40
– +4.16 +2.26 +1.14 +2.09 +2.53 +3.55
(±0.179) (±0.177) (±0.172) (±0.179) (±0.179) (±0.181) (±0.175)
−74.14 – – −73.56 −71.87 −73.95 −74.21
– – – +0.58 +2.27 +0.19 −0.07
(±0.275) – – (±0.274) (±0.264) (±0.275) (±0.275)
Structural properties of the closed conformation of HSV gD The receptor binding of HSV gD can be dissected into several distinct steps starting with the intramolecular detachment of the C-terminus (residues 268–306) from the core domain (residues 38–259) of HSV gD. Opening of this autoinhibited conformation is a prerequisite for a conformational rearrangement of the N-terminus and/or receptor binding (Fig. 1) [14, 25, 26]. An analysis of the crystal structure of the closed conformation shows that several of the residues investigated above for receptor binding are also located close to the interface between gD core and C-terminus (V24, Q27, L28, T29). In contrary, M11 and N15 are flexible in the closed conformation and therefore not resolved in the respective crystal structure. To estimate the effect of individual residues on the interaction of the C-terminus with the remaining parts of gD, a MD simulation of the closed gD conformation followed by in silico alanine scans was performed. Furthermore, a W294A mutation was included in the analysis since this mutant was previously shown to exhibit receptor binding properties similar to a gD that lacks the entire C-terminus [14] suggesting that a W294A-gD can no longer efficiently stabilize the closed gD conformation. The results for this alanine scan (Table 2) show that the Q27A and W294A mutation have the strongest effect by weakening the interaction of gD with the C-terminus by 3.43 and 11.90 kcal mol−1, respectively. This is consistent with the properties of the crystal structure, in which these two residues are directly located in the interface between core and C-terminus. For the remaining mutations (V24A, L28A, T29A), much smaller effects are detected (Table 2), which is in agreement with the fact that these residues do not significantly interact with the C-terminus. The pronounced effects predicted for the Q27A and W294A mutations from the in silico alanine scan suggested that both mutations might cause large-scale structural rearrangements in
WT V24A Q27A L28A T29A W294A
ΔG
ΔΔG
SE (ΔΔG)
−61.03 −59.60 −57.60 −60.95 −61.07 −49.13
– +1.43 +3.43 +0.08 −0.04 +11.90
(±0.131) (±0.130) (±0.128) (±0.131) (±0.131) (±0.129)
the gD structure or even the dissociation of the C-terminus. To address these aspects in more detail, explicit MD simulations were performed for Q27A and W294A mutant gD. Comparison of the RMSD values (Fig. 4) shows that both mutations increase the flexibility of the C-terminus, which is also reflected in the structural overlay shown in Fig. 5. As already expected from the results of the in silico alanine scan, this effect is more pronounced for the W294A mutant. Although the C-terminal residues in the W294A mutant do not completely dissociate, the high RMSD values (Fig. 4) and the structural changes observed over the simulation (Fig. 5b) indicate a low conformational stability of the C-terminus. These results are in agreement with the experimental findings of Krummenacher [14] indicating the importance of W294 as anchor residue. Loss of this anchor leads to easier detachment of the C-terminus, thus facilitating the subsequent receptor binding steps. Experimentally, W294A-gD exhibits enhanced binding affinities to HVEM and nectin-1 that are similar to affinities of a truncated gD, which lacks the C-terminal residues 286–306 [14]. Similar to the W294A mutation, the Q27A mutation also increases the flexibility of the C-terminus, although the effect is less pronounced (Fig. 5c). From an energetic point of view, the interaction of the C-terminus with the core is weakened by 3.59 kcal mol−1 by the Q27A mutation. This value, which was obtained from the explicit simulation of the mutant, is in remarkably good agreement with the value of 3.43 kcal mol−1 obtained from the previous single-trajectory alanine scan (Table 2). Thus, one can conclude that the Q27A mutation facilitates the opening of the closed conformation and therefore also all subsequent steps of receptor binding. This is a nonspecific effect, which should facilitate both the interaction with HVEM and nectin-1. Putting this energetic analysis into context with the above analysis of the gD-receptor complexes (Tables 1 and 2) leads to the following picture: The Q27A lowers the energetic barrier for the detachment of the C-terminus by ∼3.5 kcal mol−1, while it decreases both HVEM and nectin-1 binding affinity by
2192, Page 6 of 11
J Mol Model (2014) 20:2192
Fig. 4 RMSD of the C-terminal 16 residues in the closed gD conformation for wildtype (black), W294A (red), and Q27A (green) gD. The values were calculated after fitting the gD core on the starting structure
∼2 kcal mol−1. Considering the energetics of both steps together, one would expect that the binding of both receptors is favored by ∼1.5 kcal mol−1 due to the Q27A mutation. This would explain the experimentally observed increase in binding affinity of Q27A-gD for nectin-1, but it cannot explain the observed loss of interaction with HVEM. Thus, the inability to interact with HVEM must result from another step of the gD receptor binding process. Therefore, we investigated the properties of the N-terminal hairpin in gD, which is formed upon HVEM binding. Conformational stability of the N-terminal hairpin in HSV gD Interaction of gD with HVEM requires a conformational rearrangement of the N-terminus (Fig. 1) [17]. In this step,
Fig. 5 Conformational stability of the C-terminus in the closed gD conformation. (a) Overlay of starting structure of wildtype gD in the closed conformation with the final structure obtained after 100 ns simulation. The core of the starting structure is shown in gray with the N- and C-terminus colored in green and red, respectively. The final structure after
residues 1–22 of gD become ordered and align in an antiparallel fashion with residues 23–36 to form an intramolecular hairpin structure. Both stretches of this hairpin together constitute the HVEM binding site. In contrary, no hairpin structure is formed upon gD interaction with nectin-1 and residues 1–22 remain disordered in the gD-nectin complex structure. Thus, hairpin formation represents a candidate step which might differentially affect HVEM and nectin-1 binding specificity. Therefore, we investigated the conformational stability of the N-terminus in unliganded wildtype and Q27A hairpin-gD. The RMSD analysis indicates that the N-terminal 22 residues deviate stronger from the starting structure in the Q27A mutant compared to the wildtype (Fig. 6a). This deviation also becomes apparent when monitoring the interactions of the
100 ns is shown in lighter/opaque colors. Residues Q27 and W294 are depicted as sticks in both structures. b, c Same type of overlay for the simulations of the W294A (b) and the Q27A (c) mutant. Residues at sequence positions 27 and 294 are shown as sticks
J Mol Model (2014) 20:2192 Fig. 6 Conformational stability of the N-terminus in the hairpin gD conformation. a RMSD of the N-terminal 22 residues for wildtype (black) and Q27A (green) in unliganded gD. The RMSD values for the same sequence stretch in the gDHVEM complex (orange) are shown for comparison. All values were calculated after fitting the gD core on the starting structure. b Distance between the backbone oxygen of K10 and the amide nitrogen of I224 located in the gD core. c Distance between the Cα- carbons of M11 and N227. Color coding as in (a)
Page 7 of 11, 2192
2192, Page 8 of 11
hairpin with the gD core; in the starting structure, K10 forms a backbone hydrogen bond with I224 (Fig. 6b). This hydrogen bond is stable in the simulations of the unbound and HVEM-bound wildtype, whereas it is only formed over a very short period of the simulation time in the Q27A mutant. The loss of interactions between the hairpin and core is also evidenced by the increase of the Cα-distances between M11 and N227 for Q27A-gD (Fig. 6c). In contrast, for the unbound wildtype the respective hairpin-core distances are highly similar to those observed in the simulation of the HVEM-gD complex (Fig. 6b and c). This indicates that a proper orientation of the hairpin with respect to the core can be maintained in the unbound wildtype but not in the Q27A mutant. The sampling of the hairpin also becomes evident from an inspection of several snapshots collected over the simulation time. Those snapshots reveal that this increased RMSD of the N-terminus does not result from an overall enhanced flexibility but rather from an altered sampling of the Q27A mutant (Fig. 7a and b). In particular the conformation of the hairpin tip deviates strongly from the starting structure. In contrary, sampling of the unliganded wildtype is rather similar to that of HVEM-bound gD, although the magnitude of the fluctuations is slightly larger (Fig. 7a and c). To identify the structural origin for the altered sampling of the mutant, we analyzed the interactions of the N-terminal hairpin in more detail. Interestingly, there are almost no structural changes caused by the Q27A replacement at the site of mutation itself. However, the mutation has a strong effect on residue M11, which forms sidechain packing interactions with Q27 in the wildtype (Figs. 7d and 8a). The shorter sidechain of A27 causes a conformational rearrangement in the vicinity of M11 to re-establish a tight packing within the hairpin. The most prominent motion is that of the R18 sidechain, which becomes oriented toward M11 and packs with M11 into a cavity that is occupied by the Q27 sidechain in the wildtype (Fig. 8). This rearrangement occurs within the first nanoseconds and persists over the entire simulation, therefore resulting in an altered hairpin conformation. The observation that the hairpin of Q27A-gD exhibits a rather low flexibility and samples conformations different from that observed in the HVEM-gD complex suggest that the mutant gets trapped in a conformation with reduced HVEM-binding capacity. This idea is further corroborated by the analysis of the clashes that would result upon HVEM-binding of the mutant hairpin structure (Fig. 9). For that purpose, gD-HVEM complexes were modeled using representative conformations of unbound wildtype and mutant gD (Fig. 9b and c) and compared to the complex crystal structure. In the complex crystal structure, Y23 protrudes into a crevice of the N-terminal hairpin and represents a hotspot residue of the gD-HVEM recognition [17, 47] (Fig. 9a).
J Mol Model (2014) 20:2192
Fig. 7 Sampling of the N-terminal hairpin in wildtype and mutant gD. a Initial structure of wildtype hairpin-gD (core: gray; N-terminus: green). For the N-terminal 32 residues, the snapshots collected every 10th nanosecond from the simulation were overlaid with the starting structure and are colored according to time from red to blue. b Sampling of the unliganded Q27A mutant. c Sampling of wildtype gD in complex with HVEM. Same type of presentation as in (a). HVEM is not shown for clarity. d Crystal structure showing the location of M11 and Q27 (spacefilled) in the N-terminal hairpin of gD
For the unbound Q27A-gD, major clashes would be formed with Y23 of HVEM upon receptor binding (Fig. 9c). Y23 forms intermolecular clashes of >1 Å with A12 and P14, and of >2 Å with R18. In contrary, the Y23 binding pocket is stable in the unbound wildtype (Fig. 9b) and already exhibits a similar shape as in the complex crystal structure (Fig. 9a). Consequently, the unbound wildtype exhibits much smaller clashes with HVEM of
ORIGINAL PAPER
Mutations in herpes simplex virus gD protein affect receptor binding by different molecular mechanisms Joachim D. Stump & Heinrich Sticht
Received: 18 November 2013 / Accepted: 24 February 2014 # Springer-Verlag Berlin Heidelberg 2014
Abstract Glycoprotein D (gD) is an essential protein of herpes simplex virus-1 (HSV-1) that targets the structurally unrelated receptors HVEM and nectin-1. Receptor binding of gD is accompanied by intramolecular structural rearrangements including the detachment of the C-terminus or formation of an N-terminal hairpin structure. We have investigated several gD mutations that were reported to affect receptor binding affinity or specificity in order to identify their molecular mode of action. Molecular dynamics simulations and subsequent energetic analyses of the gD-receptor complexes reveal that some mutations (M11A, N15A, L28A, T29A) play a more prominent role for HVEM binding than for nectin-1 binding, thereby conferring specificity to receptor recognition. However, our studies show that mutations can also affect the intramolecular structural rearrangement processes in gD. W294A and Q27A mutations facilitate the detachment of the C-terminus, and Q27A additionally hampers the formation of an intramolecular hairpin in gD that is exclusively established upon HVEM binding. The finding that a Q27A mutation affects multiple steps of the receptor binding process offers a molecular explanation for its enhanced nectin-1 affinity and the pronounced receptor specificity. This study also indicates that an inspection of the gD-receptor interfaces alone may be insufficient for predicting the effect of novel mutations that alter receptor specificity. Instead, such an analysis will additionally require to assess the effect of candidate mutation on the preceding steps of gD activation.
This paper belongs to a Topical Collection on the occasion of Prof. Tim Clark’s 65th birthday J. D. Stump : H. Sticht (*) Bioinformatik, Institut für Biochemie, Friedrich-Alexander-Universität Erlangen-Nürnberg, Fahrstraße 17, 91054 Erlangen, Germany e-mail: [email protected]
Keywords Glycoprotein D . HSV-1 . Molecular dynamics . Mutational analysis . Protein interaction
Introduction Herpes simplex virus-1 (HSV-1) belongs to the subfamily of alphaherpesviruses and has a high prevalence in most human populations. Clinical symptoms range from oral lesions with watery blisters to more severe diseases like meningitis and encephalitis. After initial infection the virus persists lifelong and can reactivate any time possibly due to stress factors [1, 2]. One essential protein for the infection process is HSV glycoprotein D (gD), which can use two structurally unrelated cell surface proteins as receptors. The first receptor, herpesvirus entry mediator (HVEM), belongs to the tumor necrosis factor receptor family [3] and is generally found on immune cells. The second receptor recognized by gD is nectin-1, which is a homophilic adhesion molecule present on the surface of neurons, keratinocytes, and epithelial cells [4–6]. It has also been possible to engineer mutant gD molecules that are able to recognize other receptors than those mentioned above resulting in an altered cell tropism of HSV-1 [7–11]. The ability of gD molecules to recognize structurally dissimilar receptors renders this protein a potential candidate to be included in a viral vector system for gene therapy that allows to target distinct cell types. Due to its interesting properties, the structures of unbound gD and of gD-receptor complexes were the subject of numerous studies. gD itself consists of 369 amino acids with a 316 residue ectodomain followed by a transmembrane helix (residues 317–339) and a short intracellular C-terminus. Residues 307–369 are not necessary for functionality, as experiments showed that C-terminally truncated forms of gD were able to bind their receptor [12, 13] and even restore the infectivity of a gD-deficient virus [14, 15]. The gD ectodomain can roughly be divided into the following regions: The N-terminus (residues 1–37) involved in receptor
2192, Page 2 of 11
binding, the globular core domain (residues 38–267), the flexible profusion domain (PFD; residues 268–285), and a short C-terminal stretch (residues 286–306) that precedes the transmembrane helix [14, 16]. Up to present, crystal structures are available for unliganded gD [14, 17], as well as for gD in complex with its receptors HVEM [17] and nectin-1 [18, 19]. These structures revealed that receptor binding requires major structural changes of the gD N- and/or C-terminus suggesting that receptor binding is a multi-step-process [14]. This process can be dissected into several major steps, which are shown in Fig. 1: Unliganded gD (Fig. 1, left) adopts a closed conformation, in which the Cterminus tightly interacts with the gD core and with residues 23–37 of the N-terminus, whereas residues 1–22 are flexible and not resolved in the crystal structure. Receptor binding requires the detachment of the C-terminus, thereby uncovering the gD core and N-terminal residues, which thereby become available for further interactions. In case of nectin-1, the respective conformation is directly recognized by the receptor (Fig. 1, top right). In the respective complex, the N-terminal residues 1–22 and the C-terminal residues 253–306 remain disordered. In contrary, interaction of gD with HVEM requires a conformational rearrangement of the N-terminus. In this step, residues 1–22 of gD become ordered and align in an antiparallel fashion with residues 23–36 to form an intramolecular hairpin structure where residues 1–17 lie within the same groove as the C-terminus in the closed gD conformation. Both stretches of the N-terminal hairpin together constitute the HVEM binding site (Fig. 1, bottom right). After the first crystal structure of the gD-HVEM complex became available in 2001 [17], numerous mutational studies have been performed to obtain a more detailed understanding of receptor binding [16, 20–26]. By experimental alanine scans, several mutations have been identified, which inhibit Fig. 1 Schematic presentation of the individual steps of HSV gD structural rearrangement upon receptor binding. Unliganded gD adopts a closed conformation (left), in which the C-terminus (red) is attached to the protein core (gray). The N-terminus (residues1-37; green) is disordered for residues 1–22 as indicated by a dashed line. Detachment of the C-terminus from the core leads to an open gD conformation. This form can interact with nectin-1 (top right) without rearrangement of the N-terminus. Alternatively, the N-terminus can form a hairpin structure (‘hairpin-gD’), which constitutes the HVEM binding site (bottom right)
J Mol Model (2014) 20:2192
HVEM binding (M11A, N15A, L25A, Q27A, L28A, T29A) [20]. Unexpectedly, the L25A, Q27A, L28A, and T29A mutations still bind nectin-1, although these residues are located in a sequence stretch that is both involved in nectin-1 and HVEM binding. Even more striking, the Q27A mutation results in an enhanced nectin-1 affinity. This specific effect on nectin-1 affinity is clearly distinct from the effect of the W294A mutation, which nonspecifically enhances both nectin-1 and HVEM binding ∼100-fold [14]. The results above demonstrate that some of the mutational effects can hardly be rationalized based on the static crystal structures alone. Therefore, we employed molecular dynamics (MD) simulations to investigate the effect of different mutations in more detail. There was a particular focus on the Q27A mutant because it exhibits the most prominent effect on receptor binding specificity. MD simulations were performed for unbound gD and the two gD-receptor complexes. Furthermore, in silico alanine scans were used to quantify the effect of the mutations. These studies were complemented by explicit simulations of selected mutants to monitor larger-scale structural changes. Our simulations demonstrate that mutations do not only affect receptor binding itself but also the preceding steps of gD’s structural rearrangement, thereby explaining the experimentally observed receptor binding selectivity. These findings should be helpful for a more detailed understanding of the gD structural changes and for the design of gD mutants that exhibit an altered receptor specificity.
Methods In the present work, a total of four different gD conformations (Fig. 1) were studied by MD simulations: Unliganded gD in the closed and hairpin conformation, as well as liganded gD in
J Mol Model (2014) 20:2192
complex with HVEM and nectin-1. Simulation of closed-gD was based on one monomer from PDB entry 2C36 [14]. Cysteine 307, which was inserted to stabilize a dimer for crystallization, was deleted from this structure. The gDnectin complex was taken from PDB entry 3SKU [18], and nectin-1 was truncated after its first Ig domain leaving residues 32–144 for simulation. The complex crystal structure 1JMA [17] was used for simulation of the gD-HVEM complex and also for the simulation of hairpin-gD after deletion of the HVEM moiety. Missing residues in the structures of closed-gD (residues 257–267) and gD-HVEM (residues 93–94 of HVEM) were modeled by ModLoop [27, 28]. N- and C-termini were capped with acetyl and N-methylamide, respectively. Full-atom MD simulations were performed for each system with AMBER 10 [29] using the ff99SB force field [30] and the particle mesh Ewald summation method [31, 32]. Charges were neutralized by adding sodium or chloride counter ions, respectively. A rectangular water box with TIP3P water model [33] was generated with at least 10 Å distance between any atom of the solute and the end of the periodic box. Furthermore, the SHAKE procedure [34] was applied to constrain all bonds containing hydrogen atoms. All structures were minimized in three subsequent steps with the SANDER module of AMBER followed by a 20 ps equilibration of the systems to 310 K using a previously established protocol [35–38]. Nonbonded interactions were calculated with a cutoff of 10 Å and updated every 15 steps. The following production stage was used for gathering data. Coordinates were saved every 2 ps in a trajectory over 50 ns for the complex structures (gD-HVEM, gD-nectin) and 100 ns for unbound gD (closed-gD and hairpin-gD). Clustering was performed using ptraj from AMBER Tools [39]. We applied the average-linkage algorithm with RMSD as distance metric to generate three clusters over simulation [40]. For the analysis of unbound hairpin-gD, the representative structure was taken from the largest cluster (Q27A >87 %, WT>94 %). For visualization and structural analysis VMD 1.91 [41], Sybyl [42], Xmgrace [43], and AMBER Tools 1.4 were used [39]. The molecular mechanics generalized Born surface area (MM/GBSA) method implemented in AMBER 11 [39] was applied for energetic analysis. GB model 2 was selected in combination with recommended bondi settings for atomic radii [44, 45]. In solution, the binding free energy ΔG of a protein and its ligand is defined as: ΔG ¼ Gcomplex − Gprotein þ Gligand : Each free energy term is the sum of the molecular mechanics (MM) interaction energy (EMM) and a solvation term Gsol: ΔGb ¼ ΔE MM þ ΔGsol −TΔS: ΔEMM is calculated with the sander module of AMBER. It represents the MM interaction energy between protein and its
Page 3 of 11, 2192
ligand and comprises bond, angle, and torsion-angle energies as well as electrostatic and van der Waals energies. The solvation term ΔGsol is composed of polar and nonpolar energy contributions: ΔGsol ¼ ΔGele þ ΔGnonpolar : Nonpolar contributions were calculated as a function of the solvent accessible surface area (SASA): ΔGnonpolar ¼ γSASA þ b where γ=0.0072 kcal mol−1 Å−2 and b=0.00 kcal mol−1. Dielectric constants for solute and solvent were set to 1 and 80, respectively. MM/GBSA energetic analyses were performed over a timeframe of 40 ns for the gD-receptor complexes (11th– 50th ns), and 90 ns for unliganded gD simulations (11th– 100th ns). Snapshots were taken at a time interval of 20 ps. Intramolecular interaction energies of the C-terminus in unliganded gD were calculated between residues 23–288 and 291–306. The standard deviation σ and the number of snapshots N were taken for calculation of standard errors (SE) of the ΔG values: σ SE ¼ pffiffiffiffi : N The relative change in free energy of binding (ΔΔGb) for the selected gD mutations was calculated by a singletrajectory in silico alanine scan. In the scan, the structures from the trajectory were mutated by truncating the side-chains of the residues one at a time. For the calculation, we followed the standard protocol described by Massova and Kollman [46], which assumes that the entropy of the mutant and the wildtype do not differ significantly, and therefore entropic contributions can be neglected in the calculation of ΔΔGb. Standard errors for ΔΔGb were calculated according to the error propagation law.
Results and discussion Effect of mutations on the interaction of gD with HVEM and nectin-1 Since most mutations are located in the surface patch of gD that interacts both with HVEM and nectin-1 (Fig. 2), we first analyzed the effect of the mutations on receptor interaction and binding affinity. For that purpose, gD was simulated in complex with its receptors to analyze similarities and differences in interaction with HVEM and nectin-1. The root mean square deviation (RMSD) is ∼2 Å over the simulation time (Fig. 3). The two RMSD peaks for gD-
2192, Page 4 of 11
J Mol Model (2014) 20:2192
Fig. 2 Structure of the gD-HVEM (a) and gD-nectin (b) complexes indicating the residues investigated by the in silico alanine scan (spacefilled presentation). HVEM and nectin-1 are depicted in orange and blue, respectively, and the residues considered for mutation are colored
magenta (M11, V24, L28) and yellow (N15, Q27, T29). M11 and N15 are not shown for the gD-nectin complex, because they are not involved in the interaction and therefore not resolved in the respective crystal structure
HVEM at 2.5 ns and for gD-nectin at 13 ns arise from reversible motions of flexible regions that are distant from the binding interface. For gD-HVEM this peak can be ascribed to some flexible regions in HVEM (residues 4–13, 52–60). For gD-nectin-1 the peak in the RMSD arises from motions of the flexible loops of gD (region 83–93, 118–126). Apart from these local fluctuations both complexes are conformationally stable and no significant rearrangements are observed. These simulations thus provide a basis to assess the role of individual residues for complex formation. For that purpose, an in silico alanine scan was performed for several residues that were known to abrogate HVEM binding from experiment [20]: M11, N15, Q27, L28, and T29. Consistent with the experimental finding, MM/GBSA analysis reveals that a replacement of these residues by alanine decreases binding affinity by 2–4 kcal mol−1 (Table 1). Furthermore, the V24A mutation, which does not disrupt the gD-HVEM interaction,
also has a much smaller effect of ∼1 kcal mol−1 according to the computational alanine scan. For the gD-nectin system, the in silico alanine scan shows that mutation of V24, L28, or T29 has only a minor energetic effect on the interaction (Table 1). This is again consistent with the experiment, showing that the respective mutations do not abolish nectin-1 binding. The predicted loss in binding affinity of more than 2 kcal mol−1 for the Q27A mutation, however, is in strong disagreement with the experimental observation of an enhanced nectin-1 binding for this mutant [20, 23]. Thus, an energetic analysis of the gD-receptor complexes offers a plausible explanation for the experimentally observed effects of all mutations, with the exception of Q27A suggesting that the latter mutation exerts its effect by a more complex mechanism. To clarify the role of Q27 for binding in more detail, we also investigated the gD conformational rearrangement steps, which occur upon the receptor binding (Fig. 1).
Fig. 3 Backbone RMSD of the gD-HVEM (orange) and gDnectin (blue) complexes. The following residues were included in the RMSD calculation: gD (residues 23–252), HVEM (residues 4–60, 72–79), nectin-1 (residues 32–144)
J Mol Model (2014) 20:2192
Page 5 of 11, 2192
Table 1 Results of an in silico alanine scan performed for the gD-HVEM and gD-nectin complexes. ΔG and ΔΔG values are in kcal mol-1; SE denotes the standard error. No values were calculated for M11 and N15 in the gD-nectin complex, because these residues are not involved in the interaction and therefore not resolved in the respective crystal structure gD-HVEM
WT M11A N15A V24A Q27A L28A T29A
Table 2 Results of an in silico alanine scan performed for the closed conformation of gD. Intramolecular interaction energies were calculated between residues 291–306 of the C-terminus and residues 23–288 of the gD core. ΔG and ΔΔG values are in kcal mol−1; SE denotes the standard error Closed gD
gD-nectin
ΔG
ΔΔG
SE (ΔΔG)
ΔG
ΔΔG
SE (ΔΔG)
−49.95 −45.79 −47.69 −48.81 −47.86 −47.42 −46.40
– +4.16 +2.26 +1.14 +2.09 +2.53 +3.55
(±0.179) (±0.177) (±0.172) (±0.179) (±0.179) (±0.181) (±0.175)
−74.14 – – −73.56 −71.87 −73.95 −74.21
– – – +0.58 +2.27 +0.19 −0.07
(±0.275) – – (±0.274) (±0.264) (±0.275) (±0.275)
Structural properties of the closed conformation of HSV gD The receptor binding of HSV gD can be dissected into several distinct steps starting with the intramolecular detachment of the C-terminus (residues 268–306) from the core domain (residues 38–259) of HSV gD. Opening of this autoinhibited conformation is a prerequisite for a conformational rearrangement of the N-terminus and/or receptor binding (Fig. 1) [14, 25, 26]. An analysis of the crystal structure of the closed conformation shows that several of the residues investigated above for receptor binding are also located close to the interface between gD core and C-terminus (V24, Q27, L28, T29). In contrary, M11 and N15 are flexible in the closed conformation and therefore not resolved in the respective crystal structure. To estimate the effect of individual residues on the interaction of the C-terminus with the remaining parts of gD, a MD simulation of the closed gD conformation followed by in silico alanine scans was performed. Furthermore, a W294A mutation was included in the analysis since this mutant was previously shown to exhibit receptor binding properties similar to a gD that lacks the entire C-terminus [14] suggesting that a W294A-gD can no longer efficiently stabilize the closed gD conformation. The results for this alanine scan (Table 2) show that the Q27A and W294A mutation have the strongest effect by weakening the interaction of gD with the C-terminus by 3.43 and 11.90 kcal mol−1, respectively. This is consistent with the properties of the crystal structure, in which these two residues are directly located in the interface between core and C-terminus. For the remaining mutations (V24A, L28A, T29A), much smaller effects are detected (Table 2), which is in agreement with the fact that these residues do not significantly interact with the C-terminus. The pronounced effects predicted for the Q27A and W294A mutations from the in silico alanine scan suggested that both mutations might cause large-scale structural rearrangements in
WT V24A Q27A L28A T29A W294A
ΔG
ΔΔG
SE (ΔΔG)
−61.03 −59.60 −57.60 −60.95 −61.07 −49.13
– +1.43 +3.43 +0.08 −0.04 +11.90
(±0.131) (±0.130) (±0.128) (±0.131) (±0.131) (±0.129)
the gD structure or even the dissociation of the C-terminus. To address these aspects in more detail, explicit MD simulations were performed for Q27A and W294A mutant gD. Comparison of the RMSD values (Fig. 4) shows that both mutations increase the flexibility of the C-terminus, which is also reflected in the structural overlay shown in Fig. 5. As already expected from the results of the in silico alanine scan, this effect is more pronounced for the W294A mutant. Although the C-terminal residues in the W294A mutant do not completely dissociate, the high RMSD values (Fig. 4) and the structural changes observed over the simulation (Fig. 5b) indicate a low conformational stability of the C-terminus. These results are in agreement with the experimental findings of Krummenacher [14] indicating the importance of W294 as anchor residue. Loss of this anchor leads to easier detachment of the C-terminus, thus facilitating the subsequent receptor binding steps. Experimentally, W294A-gD exhibits enhanced binding affinities to HVEM and nectin-1 that are similar to affinities of a truncated gD, which lacks the C-terminal residues 286–306 [14]. Similar to the W294A mutation, the Q27A mutation also increases the flexibility of the C-terminus, although the effect is less pronounced (Fig. 5c). From an energetic point of view, the interaction of the C-terminus with the core is weakened by 3.59 kcal mol−1 by the Q27A mutation. This value, which was obtained from the explicit simulation of the mutant, is in remarkably good agreement with the value of 3.43 kcal mol−1 obtained from the previous single-trajectory alanine scan (Table 2). Thus, one can conclude that the Q27A mutation facilitates the opening of the closed conformation and therefore also all subsequent steps of receptor binding. This is a nonspecific effect, which should facilitate both the interaction with HVEM and nectin-1. Putting this energetic analysis into context with the above analysis of the gD-receptor complexes (Tables 1 and 2) leads to the following picture: The Q27A lowers the energetic barrier for the detachment of the C-terminus by ∼3.5 kcal mol−1, while it decreases both HVEM and nectin-1 binding affinity by
2192, Page 6 of 11
J Mol Model (2014) 20:2192
Fig. 4 RMSD of the C-terminal 16 residues in the closed gD conformation for wildtype (black), W294A (red), and Q27A (green) gD. The values were calculated after fitting the gD core on the starting structure
∼2 kcal mol−1. Considering the energetics of both steps together, one would expect that the binding of both receptors is favored by ∼1.5 kcal mol−1 due to the Q27A mutation. This would explain the experimentally observed increase in binding affinity of Q27A-gD for nectin-1, but it cannot explain the observed loss of interaction with HVEM. Thus, the inability to interact with HVEM must result from another step of the gD receptor binding process. Therefore, we investigated the properties of the N-terminal hairpin in gD, which is formed upon HVEM binding. Conformational stability of the N-terminal hairpin in HSV gD Interaction of gD with HVEM requires a conformational rearrangement of the N-terminus (Fig. 1) [17]. In this step,
Fig. 5 Conformational stability of the C-terminus in the closed gD conformation. (a) Overlay of starting structure of wildtype gD in the closed conformation with the final structure obtained after 100 ns simulation. The core of the starting structure is shown in gray with the N- and C-terminus colored in green and red, respectively. The final structure after
residues 1–22 of gD become ordered and align in an antiparallel fashion with residues 23–36 to form an intramolecular hairpin structure. Both stretches of this hairpin together constitute the HVEM binding site. In contrary, no hairpin structure is formed upon gD interaction with nectin-1 and residues 1–22 remain disordered in the gD-nectin complex structure. Thus, hairpin formation represents a candidate step which might differentially affect HVEM and nectin-1 binding specificity. Therefore, we investigated the conformational stability of the N-terminus in unliganded wildtype and Q27A hairpin-gD. The RMSD analysis indicates that the N-terminal 22 residues deviate stronger from the starting structure in the Q27A mutant compared to the wildtype (Fig. 6a). This deviation also becomes apparent when monitoring the interactions of the
100 ns is shown in lighter/opaque colors. Residues Q27 and W294 are depicted as sticks in both structures. b, c Same type of overlay for the simulations of the W294A (b) and the Q27A (c) mutant. Residues at sequence positions 27 and 294 are shown as sticks
J Mol Model (2014) 20:2192 Fig. 6 Conformational stability of the N-terminus in the hairpin gD conformation. a RMSD of the N-terminal 22 residues for wildtype (black) and Q27A (green) in unliganded gD. The RMSD values for the same sequence stretch in the gDHVEM complex (orange) are shown for comparison. All values were calculated after fitting the gD core on the starting structure. b Distance between the backbone oxygen of K10 and the amide nitrogen of I224 located in the gD core. c Distance between the Cα- carbons of M11 and N227. Color coding as in (a)
Page 7 of 11, 2192
2192, Page 8 of 11
hairpin with the gD core; in the starting structure, K10 forms a backbone hydrogen bond with I224 (Fig. 6b). This hydrogen bond is stable in the simulations of the unbound and HVEM-bound wildtype, whereas it is only formed over a very short period of the simulation time in the Q27A mutant. The loss of interactions between the hairpin and core is also evidenced by the increase of the Cα-distances between M11 and N227 for Q27A-gD (Fig. 6c). In contrast, for the unbound wildtype the respective hairpin-core distances are highly similar to those observed in the simulation of the HVEM-gD complex (Fig. 6b and c). This indicates that a proper orientation of the hairpin with respect to the core can be maintained in the unbound wildtype but not in the Q27A mutant. The sampling of the hairpin also becomes evident from an inspection of several snapshots collected over the simulation time. Those snapshots reveal that this increased RMSD of the N-terminus does not result from an overall enhanced flexibility but rather from an altered sampling of the Q27A mutant (Fig. 7a and b). In particular the conformation of the hairpin tip deviates strongly from the starting structure. In contrary, sampling of the unliganded wildtype is rather similar to that of HVEM-bound gD, although the magnitude of the fluctuations is slightly larger (Fig. 7a and c). To identify the structural origin for the altered sampling of the mutant, we analyzed the interactions of the N-terminal hairpin in more detail. Interestingly, there are almost no structural changes caused by the Q27A replacement at the site of mutation itself. However, the mutation has a strong effect on residue M11, which forms sidechain packing interactions with Q27 in the wildtype (Figs. 7d and 8a). The shorter sidechain of A27 causes a conformational rearrangement in the vicinity of M11 to re-establish a tight packing within the hairpin. The most prominent motion is that of the R18 sidechain, which becomes oriented toward M11 and packs with M11 into a cavity that is occupied by the Q27 sidechain in the wildtype (Fig. 8). This rearrangement occurs within the first nanoseconds and persists over the entire simulation, therefore resulting in an altered hairpin conformation. The observation that the hairpin of Q27A-gD exhibits a rather low flexibility and samples conformations different from that observed in the HVEM-gD complex suggest that the mutant gets trapped in a conformation with reduced HVEM-binding capacity. This idea is further corroborated by the analysis of the clashes that would result upon HVEM-binding of the mutant hairpin structure (Fig. 9). For that purpose, gD-HVEM complexes were modeled using representative conformations of unbound wildtype and mutant gD (Fig. 9b and c) and compared to the complex crystal structure. In the complex crystal structure, Y23 protrudes into a crevice of the N-terminal hairpin and represents a hotspot residue of the gD-HVEM recognition [17, 47] (Fig. 9a).
J Mol Model (2014) 20:2192
Fig. 7 Sampling of the N-terminal hairpin in wildtype and mutant gD. a Initial structure of wildtype hairpin-gD (core: gray; N-terminus: green). For the N-terminal 32 residues, the snapshots collected every 10th nanosecond from the simulation were overlaid with the starting structure and are colored according to time from red to blue. b Sampling of the unliganded Q27A mutant. c Sampling of wildtype gD in complex with HVEM. Same type of presentation as in (a). HVEM is not shown for clarity. d Crystal structure showing the location of M11 and Q27 (spacefilled) in the N-terminal hairpin of gD
For the unbound Q27A-gD, major clashes would be formed with Y23 of HVEM upon receptor binding (Fig. 9c). Y23 forms intermolecular clashes of >1 Å with A12 and P14, and of >2 Å with R18. In contrary, the Y23 binding pocket is stable in the unbound wildtype (Fig. 9b) and already exhibits a similar shape as in the complex crystal structure (Fig. 9a). Consequently, the unbound wildtype exhibits much smaller clashes with HVEM of
