Effects of single α-to-β residue replacements on recognition of an extended segment in a viral fusion protein

Abstract

Partial replacement of α-amino acid residues with β-amino acid residues has been established as a strategy for preserving target-engagement by helix-forming polypeptides while altering other properties. The impact of β-residue incorporation within polypeptides that adopt less regular conformations, however, has received less attention. The HRC domains of fusion glycoproteins from pathogenic paramyxoviruses contain a segment that must adopt an extended conformation in order to co-assemble with the HRN domain in the post-fusion state and drive merger of the viral envelope with a target cell membrane. Here we examine the impact of single α-to-β substitutions within this extended N-terminal segment of an engineered HRC peptide designated VIQKI. Stabilities of hexameric co-assemblies formed with the native HPIV3 HRN have been evaluated, the structures of five co-assemblies have been determined, and antiviral efficacies have been measured. Many sites within the extended segment show functional tolerance of α-to-β substitution. These results offer a basis for future development of paramyxovirus infection inhibitors with novel biological activity profiles, possibly including resistance to proteolysis.

The disruption of interactions within or between proteins is an attractive basis for drug design. This strategy requires molecules that can engage large surfaces on target proteins, a goal that is generally difficult to achieve with small molecules (MW<1000).^1-3 Clinically useful inhibitors of pathogenic protein-protein interactions are typically engineered proteins or large peptides. Antibodies and other large proteins are expensive to produce.⁴ Smaller polypeptides offer the advantage of chemical synthesis but suffer from rapid degradation by endogenous proteases.^1-3 One strategy for retaining the benefits of polypeptides while minimizing proteolytic susceptibility is to incorporate non-natural subunits into the peptidic backbone.^5-12 The challenge inherent in this approach is to introduce a sufficient number and distribution of backbone modifications to hinder protease action without loss of target protein engagement. We have shown that the recognition properties of an α-helix can be retained after periodic replacement of α-amino acid residues with β residues.^13-22 The resulting “α/β-peptides” adopt an α-helix-like conformation but resist enzymatic degradation. Less is known about the impact of β residue incorporation in non-helical segments.

Paramyxoviruses, such as parainfluenza, measles, Nipah, and Hendra viruses, are negative-strand RNA viruses that negatively impact global human health. Here, we describe fundamental studies of α-to-β modifications in the context of an extended segment that plays a critical role in a paramyxovirus infection mechanism. We focus on an eight-residue portion of the fusion (F) glycoprotein of human parainfluenza virus 3 (HPIV3). The F protein orchestrates fusion of the HPIV3 envelope and the outer membrane of a target cell. This process involves a complex set of structural transitions within an F trimer that ultimately leads to the formation of a “six-helix bundle” (6HB) assembly between N-terminal heptad repeat (HRN) and C-terminal heptad repeat (HRC) domains of F.^23-25 In the post-fusion 6HB state, the N-terminal portion of the HRC domain (residues 449-456) adopts an extended conformation, while the majority of the HRC domain residues (457-484) form an α-helix (Figure 1A).²⁶

Figure 1.

(A)HPIV3 F ectodomain in postfusion conformation (PDB 1ZTM), (B) VIQKI co-assembled with HPIV3-HRN (PDB 6NRO), (C) sequence of VIQKI, (D) structure of a β³-residue.

The few previous evaluations of α-to-β modification in extended peptide ligands had variable outcomes. For example, three groups have explored systematic “β scans” of short peptide antigens (8 or 9 residues) that bind to class I major histocompatibility complexes (MHC I).^27-29 In each system, MHC binding was retained for β substitution at some positions, but most of the resulting peptide-MHC complexes were not effective in stimulating T cells that respond to the natural antigen. A subsequent study found greater functional tolerance of single α-to-β replacements in a 17-mer antigen that binds to a class II MHC.³⁰ A systematic α-to-β analogue series was recently reported for a 10-residue phosphopeptide that binds in an extended conformation to 14-3-3 proteins.³¹ β substitution near either end caused modest declines in affinity (~10-fold), but more severe losses in affinity were observed for substitutions closer to the center of the peptide.

Based on these precedents, it was not clear whether α-to-β substitutions would be structurally or functionally tolerated within the extended portion of an HPIV3 inhibitor. We were motivated to proceed, however, by two considerations. First, peptide binding to an MHC or a 14-3-3 protein involves multiple H-bonds between the backbone of the peptide ligand and the protein partner. In contrast, there are relatively few H-bonds between the extended segment of the HPIV3 HRC domain and the HRN trimer core in the post-fusion 6HB state of F. Second, coronaviruses (CoVs) feature even longer extended segments within the post-fusion HRC domains of their fusion glycoproteins (S) than do paramyxoviruses, but the backbones of these extended CoV segments form relatively few H-bonds with the HRN trimer core.^32-34 α-to-β replacement seems likely to disrupt backbone H-bonding in an extended conformation; therefore, we predicted that this type of modification might be better tolerated in an extended segment of a viral fusion protein HRC domain than in other linear peptide epitopes.

METHODS

Our study began with a previously reported peptide derived from the HRC domain of HPIV3, designated VIQKI (Figure 1C), which blocks infection by both HPIV3 and respiratory syncytial virus (RSV).^35-37 These two viruses are leading causes of fatal lower respiratory infections in young children.^38-40 A crystal structure of VIQKI co-assembled with HPIV3-HRN (Figure 1B) revealed a 6HB assembly similar to the HRC/HRN assembly in the post-fusion conformation of the HPIV3 F protein (Figure 1A).³⁵ As expected, two distinct conformations were observed within the VIQKI peptide, an α-helix spanning residues corresponding to 457–484 of the F protein, and an extended conformation for residues corresponding to 449–456. Our studies focused on backbone modifications in the extended portion of VIQKI (Figure 1, dashed boxes).

A set of VIQKI analogues was synthesized in which each of the eight N-terminal residues was replaced individually with the homologous β³-amino acid residue (Figure 1D). Thus, each replacement preserved the side chain found in VIQKI but introduced an extra CH₂ unit into the backbone. The effect of each single α-to-β³ substitution on assembly stability with HPIV3-HRN was examined using variable-temperature circular dichroism (VT-CD) spectroscopy. When combined with a 51-residue peptide corresponding to the HPIV3-HRN domain, each α/β-VIQKI variant formed a helical assembly, as detected by CD minima at 208 and 222 nm. Each assembly exhibited a cooperative thermal transition at 222 nm, which allowed us to determine apparent thermal denaturation (T_m,app) values, based on the mid-point of each transition (Table 1). For each α/β-VIQKI variant, a ΔT_m,app value was calculated based on comparison with T_m,app for VIQKI itself [ΔT_m,app = T_m,app (HPIV3-HRN+VIQKI β³-variant) – T_m,app (HPIV3-HRN+VIQKI)].

Table 1.

Apparent thermal denaturation temperatures (T_m,app) of co-assemblies formed between HPIV3-HRN+VIQKI β³-variant pairs

Peptide	Tm,app (°C)a	ΔTm,appb
VIQKI	88.6±0.4
VIQKI-V449(β3V)	89.0±0.8	+0.4
VIQKI-A450(β3A)	89.2±1.0	+0.6
VIQKI-L451 (β3L)	88.7±0.8	+0.1
VIQKI-D452(β3D)	89.1±0.4	+0.5
VIQKI-P453(β3P)	86.1±0.9	−2.5
VIQKI-I454(β3I)	86.7±1.0	−1.9
VIQKI-D455(β3D)	81.5±0.5	−7.1
VIQKI-I456(β3I)	86.2±1.1	−2.4

^a1:1 mixture of Inhibitor (50 μM) and HRN (50 μM) peptides. T_m,app values were determined as mean values of three replicates ± standard deviation.

^bΔT_m.app = T_m,app (HPIV3-HRN+VIQKI β³-variant) - T_m,app (HPIV3-HRN+VIQKI)

The VT-CD analysis revealed that most α-to-β replacements in the N-terminal region of VIQKI caused little or no change in ΔT_m,app. Since omission of the three N-terminal residues (N449-L451) has little effect HRC-HRN assembly stability,³⁵ it is perhaps not surprising that the β³ residue was well tolerated at each of these positions. The D452-I456 segment, on the other hand, was shown by truncation studies to be crucial for assembly stability,³⁵ and it was therefore not expected that only one α-to-β replacement among these five residues, D455(β³D), would be highly destabilizing.

To assess the functional impact of each single α-to-β substitution, we evaluated inhibitory activity toward HPIV3 infection using plaque reduction assays. Hep2 cell monolayers were infected with engineered recombinant HPIV3 that expresses green fluorescent protein (GFP) in the presence of various concentrations of each VIQKI variant, and the extent of infection was determined by quantification of fluorescence. The results are provided in two formats. Figure 2A shows extent of inhibition of infection as a function of VIQKI variant concentration. Figure 2C shows the ratio of extent of inhibition for each variant relative to inhibition for VIQKI itself at the experimental concentration closest to the IC₅₀ of VIQKI (16 nM). Consistent with the VT-CD results, the functional data show that seven of the eight β-substituted peptides display substantial antiviral activity. Activity was lowest for the substitution that was most destabilizing to HRC-HRN assembly, D455(β³D).

Figure 2.

Anti-viral efficacy of VIQKI β³-variants against HPIV3 or RSV. Peptide activity against HPIV3 (left) or RSV (right) were determined by plaque reduction assay in infected Hep-2 cell monolayers. Data are shown as (A and B) percent inhibition of infection as a function of inhibitor concentration and (C and D) a ratio of the percent inhibition of each β³ variant with the percent inhibition of VIQKI. Gray zones represent the standard deviation of the VIQKI control. Data are expressed as mean ± standard deviation (n=3 separate experiments).

The high sensitivity of D455 to β residue replacement, in terms of both assembly stability and antiviral activity, was intriguing because of a critical role suggested by previous structural studies for the D455 side chain. The carboxylate of D455 in the post-fusion 6HB form of the HPIV3 F protein engages in a H-bond with the backbone amide N-H of S457.²⁶ The same H-bond is seen in the co-crystal structure of VIQKI with the HPIV3 HRN domain.³⁵ This i,i+2 interaction appears to serve as an N-terminal cap⁴¹ for the α-helical secondary structure that starts at S457 and extends to the C-terminus.

To try to gain insight into the structural impact of the α-to-β substitutions, we co-crystallized five VIQKI variants with the HPIV3-HRN peptide, D452(β³D), P453(β³P), I454(β³I), D455(β³D) and I456(β³I). Each structure was solved by molecular replacement using a search model based on the HPIV3-HRN+VIQK1 structure (PDB 6NRO).³⁵ Chains A and B from 6NRO were truncated to include residues 153–173 (chain A) and 460–480 (chain B). All hydrogen atoms and side chains were removed prior to molecular replacement. Crystallographic models were refined to 1.70 Å [HPIV3-HRN+VIQKI-D452(β³D), (PDB 6PZ6)]; 1.87 Å [HPIV3-HRN+VIQKI-P453(β³P), (PDB 6PRL)]; 1.49 Å [HPIV3-HRN+VIQKI-I454(β³I), (PDB 6VAS)]; 1.79 Å [HPIV3-HRN+VIQKI-D455(β³D), (PDB 6PYQ)]; and 2.17 Å [HPIV3-HRN+VIQKI-I456(β³I), (PDB 6V3V)].

In each of the five new structures, the single β³ residue substitution within the extended N-terminal segment did not affect the α-helical conformation of residues 457–484. However, the impact of α-to-β³ substitution on the conformation of the N-terminal segment differed among the five variants (Figure 3).

Figure 3.

X-ray crystal structures of VIQKI and β³-variants co-assembled with HPIV3-HRN. (A) HPIV3-HRN+VIQKI, PDB:6NRO; (B) HPIV3-HRN+VIQKI-D452(β³D), (PDB:6PZ6); (C) HPIV3-HRN+VIQKI-P453(β³P), (PDB:6PRL); (D) HPIV3-HRN+VIQKI-I454(β³I), (PDB:6VAS); (E) HPIV3-HRN+VIQKI-D455(β³D), (PDB:6PYQ); (F) HPIV3-HRN+VIQKI-I456(β³I), (PDB:6V3V). HPIV3-HRN (orange), VIQKI and β³-variants (green) with β³-residues highlighted in violet.

In three of the five co-crystal structures, those containing P453(β³P), I454(β³I), or D455(β³D) substitutions, the backbone modification resulted in a significant remodeling of the extended conformation in the N-terminal region relative to VIQKI. For VIQKI-P453(β³P), contacts involving the I454 and I456 side chains and hydrophobic residues from the HPIV3-HRN trimer were retained, as was the helix capping interaction of D455. However, this substitution nucleated a hairpin turn that reoriented residues 449–452, with loss of contacts observed between these residues in VIQKI and HPIV3-HRN (Figures 3C, 4B). Since previous work showed that contacts involving residues 449–451 do not significantly contribute to six-helix bundle stability,³⁵ this distortion may reflect effects of packing with lattice neighbors. For VIQKI-I454(β³I), only residues 455–484 could be resolved. No significant electron density was observed for the β³I residue or residues 449–453.

Figure 4.

2mFo-DFc weighted electron density maps for β³-residues in (A) HPIV3-HRN+VIQKI-D452(β³D), (B) HPIV3-HRN+VIQKI-P453(β³P), (C) HPIV3-HRN+VIQKI-D455(β³D), (D) HPIV3-HRN+VIQKI-I456(β³I). β³-residues are shown in violet.

In the co-assembly of VIQKI-D455(β³D) and HPIV3-HRN, residues 452–484 of the VIQKI variant were resolved. However, the transition from extended to α-helical conformation, which occurred at S457 in the HPIV3-HRN+VIQKI structure, was not observed for the D455(β³D) variant. Instead, VIQKI-D455(β³D) adopted a helical conformation from residue 453 to 484 (Figures 3E, 4C). We hypothesize that the additional backbone carbon atom between the carboxylate side chain of β³D455 and the backbone amide N-H of S457 in this variant, relative to VIQKI itself, interferes with the helix-capping interaction involving the side chain of D455 in VIQKI. This conformational rearrangement in the N-terminal segment may explain the loss of 6HB stability (Table 1) and diminished antiviral activity (Figure 2) observed for VIQKI-D455(β³D).

In the crystal structures of VIQKI-D452(β³D) and VIQKI-I456(β³I) co-assembled with HPIV3-HRN, the extended conformation observed for VIQKI itself was retained (Figures 3B, 3F), although in both variants the two terminal residues, V449 and A450, could not be resolved. In both new structures, key interactions between the N-terminal segment of the inhibitor and the HPIV3-HRN core were preserved. These interactions include contacts of side chains from L451, I454, and I456 (or β³I456) with hydrophobic residues in the HRN trimer, as well as H-bonds between the backbone carbonyl of D452 with the side chain hydroxyl of HRN residue Y178. The helix capping interaction was present in both of these VIQKI variants.

Although assessing activity of the β-substituted analogues of VIQKI toward RSV was tangential to the main goals of this work, we evaluated these peptides for the ability to inhibit RSV infection of Hep-2 cells. Figures 2B,D show that the effect of α-to-β substitutions on efficacy against RSV was more pronounced than the effect on efficacy against HPIV3. Most VIQKI variants were significantly less active relative to VIQKI itself. However, variants L451(β³L) and D452(β³D) seemed to match VIQKI in efficacy against RSV infection. The N-terminal segment of VIQKI adopts different conformations when co-assembled with HPIV3-HRN vs. RSV-HRN.³⁰ Residues P453-I456 are part of the extended segment in the assembly with HPIV3-HRN, but these residues have been incorporated into the α-helix in the assembly with RSV-HRN. It is noteworthy that only two variants, L451(β³L) and D452(β³D), maintained inhibition of infection by both viruses.

We have used a peptide that inhibits HPIV3 infection via a well-characterized mode of target engagement to explore the effects of replacing α-amino acid residues by homologous β³ residues within a segment that is extended in the bioactive conformation. This modification preserves the side chain but alters the backbone via introduction of a CH₂ unit. Based on previous studies involving β replacements in extended peptide ligands, it could not have been predicted that most sites in the extended portion of the α-peptide inhibitor VIQKI would tolerate α-to-β³ substitution while retaining the ability to inhibit HPIV3 infection. We hypothesize that the extended segments of viral HRC domains in the 6HB assemblies of post-fusion forms may generally prove to be tolerant of backbone modifications because the backbone amide groups in these segments form few H-bonds with HRN domain trimers. This work offers a step toward development of protease-resistant dual HPIV3/RSV inhibitors that might address life-threatening child-hood infections for which there are no current drugs. (We have not examined the effects of proteases on the β-substituted peptides discussed here, because, with only one backbone modification, these 36-mers will undoubtedly be susceptible at most sites of potential cleavage.) In addition, given the presence of long extended segments in the post-fusion HRC/HRN assemblies formed by coronavirus fusion glycoproteins,^32-34 these results encourage application of the backbone modification strategy to inhibitor development directed toward these pathogens.