Cross-Inhibition of Chikungunya Virus Fusion and Infection by Alphavirus E1 Domain III Proteins

BHK-21 cells were cultured at 37°C in complete medium (Dulbecco's modified Eagle's medium containing 5% fetal calf serum, 10% tryptose phosphate broth, 100 U penicillin/ml, and 100 μg streptomycin/ml). Cholesterol-containing C6/36 mosquito cells and cholesterol-depleted C6/36 cells were cultured at 28°C as previously described (13). Wild-type (WT) SFV was a well-characterized plaque-purified stock (34). The chikungunya virus strain used was the CHIK vaccine strain 181/clone 25 (181/25), kindly provided by Robert B. Tesh (University of Texas Medical Branch, Galveston, TX). Sindbis virus (SINV) was derived from the Toto1101 infectious clone (35).

Virus fusion with the plasma membrane was quantitated using a previously described assay (24). In brief, dilutions of CHIKV, SFV, or SINV stocks were bound to BHK-21 cells on ice for 90 min and treated at 37°C for 1 min with medium buffered to the indicated pH. Cells were then cultured overnight at 28°C in medium containing 20 mM NH₄Cl to prevent secondary infection. Cells infected due to virus fusion with cell membrane were quantitated by immunofluorescence (24). SFV- and CHIKV-infected cells were detected with a polyclonal antibody to the SFV E1 and E2 proteins (24), and SINV-infected cells were detected with a mixture of the monoclonal antibodies (MAbs) R6 and R2 against SINV E1 and E2 proteins, respectively, obtained from hybridomas kindly provided by William Klimstra (36). Fusion and infection were similarly assayed on control and cholesterol-depleted C6/36 cells, using 1 min of treatment at the indicated pH at 28°C.

DNA sequences of CHIKV E1 DIII with or without the stem region or of CHIKV E1 DI/II were amplified by PCR from the plasmid p353T-EIC, kindly donated by Felix Rey and Karine Moncoq (Pasteur Institute). CHIKV DIII sequences were subcloned into the bacterial expression vector pET-14b (Novogen) to express DIII proteins with an added N-terminal methionine and with or without an N-terminal 6×His tag, consisting of 36 extra residues. The plasmids expressing SFV E1 DIII proteins and dengue virus DV2 DIIIH1 protein were previously described (24). The residue numbers in the CHIKV and SFV DIII proteins were 291 to 383 (DIII) or 291 to 412 (DIIIS). SFV E1 DI/II was amplified from the previously reported plasmid pMTA-SFV E1DI/II (33). The CHIKV and SFV E1 DI/II sequences were cloned into the bacterial expression vector pET 21a(+) (Novagen). All sequences were confirmed by DNA sequencing.

Plasmids were transformed in BL21(DE3) Rosetta cells (Stratagene). DIII proteins were expressed, refolded, and purified as previously described (24, 33). CHIKV and SFV E1 DI/II proteins were expressed and refolded using a modified version of a published protocol (37). In brief, bacterial cultures were induced with 1 mM IPTG (isopropyl-β-d-thiogalactopyranoside) for 3 h, bacterial pellets were treated to obtain inclusion bodies (24, 33), and proteins were refolded in vitro by a fast dilution method in a 1-liter oxidative refolding buffer (100 mM, Tris-HCl, 0.4 M nondetergent sulfobetaine 201 [NDSB-201], 5 mM reduced glutathione, 0.5 mM oxidized glutathione [pH 8.0]) (37). The refolded proteins were concentrated and purified by fast protein liquid chromatography on a Superdex S200 column (GE Healthcare). The protein concentration was determined by Bradford assay or by absorption at 205 nM (38).

C-terminally truncated forms of SFV or CHIKV E1 proteins were obtained by inducible expression in Drosophila S2 cells as previously described (33). The SFV E1 DI/II protein includes domains I and II and 13 residues of the DI-DIII linker region and ends at E1 P294, followed by two Strep-tags separated by a flexible linker. Our previous studies showed that the DI-DIII linker was essential to form the stable core trimer (39). The analogous CHIKV E1 DI/II protein was constructed by DNA amplification of the plasmid p353T-EIC mentioned above. Inducible S2 cells expressing these constructs were prepared by cotransfection with the pT371 puromycin selection vector. Proteins were expressed by growth of the cells in Drosophila serum-free medium (HyClone, Logan, UT) in the presence of 750 μM copper sulfate at 27°C for 8 days. Strep-tagged proteins were purified from the culture medium by affinity chromatography using a Strep-Tactin column (IBA, Lifesciences, Goettingen, Germany).

Protein thermostability was analyzed using an assay that detects the increase in fluorescence of SYPRO orange upon binding to exposed hydrophobic residues (Invitrogen, Carlsbad, CA). Proteins were diluted to 50 μM in 150 mM NaCl–100 mM HEPES (pH 7.5), and 40-μl samples were mixed with 1 μl of a 1:200 dilution of SYPRO orange dye stock. Replicate 20-μl aliquots were added to a standard 96-well PCR plate and covered with optical sealing film. Fluorescence was monitored versus temperature using a real-time PCR device (Bio-Rad iCycler iQ5 PCR thermal cycler; Bio-Rad Hercules, CA). The companion software was used to analyze the data and calculate the midpoint of the protein unfolding transition (T_m).

Liposomes were prepared in TAN buffer (20 mM triethanolamine [TEA] [pH 8.0], 130 mM NaCl) by freeze-thawing and extrusion through 2 stacked 400-nm polycarbonate filters as previously described, stored at 4°C under N₂, and used within 2 weeks of preparation (33). Complete liposomes were composed of a 1:1:1:3 molar ratio of 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine, 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine, bovine brain sphingomyelin (Avanti Polar Lipids, Alabaster, AL), and cholesterol (Steraloids, Inc., Wilton, NH), plus trace amounts of [³H]cholesterol (Amersham, Arlington Heights, IL). Cholesterol-minus liposomes maintained the 1:1:1 molar ratio of phospholipids without cholesterol.

SFV or CHIKV E1 DI/II proteins were mixed with liposomes at a final concentration of 50 μg protein/ml and 1 mM lipid in MES (morpholineethanesulfonic acid) buffer (20 mM MES [pH 8.0], 130 mM NaCl) or TAN buffer. DIII proteins were added as indicated, using a molar ratio of 1:1 with E1 DI/II. Samples were preincubated at 25°C for 5 min, adjusted to pH 5.75 by addition of precalibrated amounts of 0.3 M MES, incubated at 25°C for 30 min, and adjusted to pH 8.0 by addition of a precalibrated volume of 0.3 M TEA. Samples were then adjusted to a final concentration of 20% sucrose, loaded on top of a 300-μl 60% sucrose cushion in a TLS55 ultracentrifuge tube, and overlaid with 1.2 ml 15% sucrose and 200 μl 5% sucrose, all wt/wt in TAN buffer (pH 8.0). Gradients were centrifuged for 3 h at 54,000 rpm at 4°C in a TLS55 rotor and fractionated into seven 300-μl fractions. Fractions were pooled into the 3 top and 4 bottom fractions or into the 2 top, 2 middle, and 3 bottom fractions. The [³H]cholesterol marker was quantitated by scintillation counting; >70% of the marker was routinely recovered in the top 3 fractions. Aliquots of the pooled fractions were precipitated with 10% trichloroacetic acid and analyzed by SDS-PAGE on 11% acrylamide gels. Proteins were detected by Western blotting using domain-specific MAb E1-3 against SFV DI/II and MAb E1-2 against SFV DIII (33), a polyclonal antibody against SFV E1 and E2 proteins that also detects CHIKV E1 DI/II protein (33), and a MAb against the His tag (Lifetein LLC, South Plainfield, NJ) to detect His-tagged CHIKV DIII proteins, followed by Alexa Fluor 680-conjugated secondary antibodies (Molecular Probes, Grand Island, NY). Western blots were quantitated using an Odyssey infrared imaging system and Odyssey InCell Western software (LI-COR Biosciences, Lincoln, NE). The Prism program (Graphpad Software) was used to calculate P values by two-tailed Student's t test with Bonferroni's correction for multiple comparisons.

To produce the core trimer in bulk, 350 μg/ml of bacterially expressed SFV E1 DI/II protein was mixed with 1.5 mM cholesterol-containing liposomes, treated at pH 5.75 for 30 min, and adjusted to pH 8.0. An aliquot was taken to determine the efficiency of trimer formation using flotation in sucrose gradients or trypsin resistance (33). The mixture was then adjusted to a final concentration of 1.5% octyl-glucopyranoside (OG) (Sigma-Aldrich, St. Louis, MO) and 1 M NaCl, and solubilized for 1 h at room temperature. A mock reaction without DI/II protein was prepared in the same way, and used for trimer dilutions in fluorescence polarization (FP) studies as described below. SFV DIIIS and DIII proteins were labeled with an amino-reactive form of fluorescein (N-hydroxysuccinimide-fluorescein), purified by a desalting column, and eluted in TAN buffer, and the degree of labeling was determined, all as described by the manufacturer (Thermo Scientific, Waltham, MA). The labeling ratio for each DIII preparation was 2 to 3 mol of dye per mol of protein.

The labeled DIII was then used to establish an FP assay for binding of DIII to the core trimer. Serial 1:2 dilutions of labeled DIII proteins were analyzed for fluorescence and FP with core trimer targets; a concentration of 100 nM DIII proved optimal for FP measurements. The assay was thus performed using 100 nM DIII and serial 1:2 dilutions of trimer starting at 5 μM. Duplicate aliquots of the samples were loaded into 384-well plates and incubated at room temperature for 1 h. FP was then measured in a Tecan Infinite 200 PRO M1000 plate reader. The K_d (dissociation constant) value was estimated as the concentration of core trimer protein producing a half-maximal increase in fluorescence polarization and was calculated by nonlinear regression analysis of binding curves using the Prism program.

The CHIKV vaccine strain (181/25) was derived from the 15561 Southeast Asian human isolate by passage in MRC-5 cells (40). To develop this virus as a system to study inhibition by recombinant DIII, we first evaluated its membrane fusion requirements. The pH dependence of virus-membrane fusion was determined for CHIKV, SFV, and SINV by following the fusion of virus with the plasma membrane of BHK-21 cells (24) (Fig. 1). Virus was prebound to cells on ice and then pulsed with media of various pHs for 1 min at 37°C to trigger the fusion of virus with the cell membrane. As previously described, under these conditions, SFV showed a pH threshold of ∼pH 6.0 and maximal fusion at ∼pH 5.75 to 5.5, while SINV Toto1101 had a pH threshold of ∼pH 5.5 and maximal fusion at ∼pH 5.0 (34). Similar to SFV, CHIKV had a pH threshold of ∼pH 6.0, but maximal fusion occurred after treatment at pH 5.25, and unlike SFV, CHIKV did not show significant inactivation of fusion activity at lower pH. Fusion of this strain of CHIKV was also strongly promoted by target membrane cholesterol, similar to SFV and SINV (Table 1) (13, 14).

We then wished to test if CHIK virus fusion could be inhibited by exogenous DIII and to compare autologous and heterologous inhibition by CHIKV versus SFV DIII proteins. Recombinant CHIKV DIII proteins were expressed in Escherichia coli, refolded in vitro, and purified by gel filtration as previously described for SFV DIII proteins (24). We tested CHIKV DIII constructs with or without a His tag (H) at the amino terminus and with or without the C-terminal stem region (S). Unlike the comparable SFV DIII proteins, CHIKV DIII proteins without the His tag were neither efficiently expressed nor accumulated in inclusion bodies (data not shown). In contrast, His-tagged CHIKV DIII proteins with or without the stem were expressed and purified at levels comparable to those of SFV. SDS-PAGE analysis of these CHIKV DIII proteins demonstrated the presence of disulfide bonds (data not shown). As a test for correct folding and protein stability, we performed thermal denaturation assays using SYPRO orange to detect the exposure of hydrophobic regions. As a control, we tested His-tagged and untagged SFV proteins with or without the stem, which were extensively characterized in our previous work and shown to contain the correct disulfide bonding pattern and structure as DIII in SFV E1 (24) (Protein Data Bank [PDB] accession no. 2V33). The CHIKV HDIII and HDIIIS proteins showed low SYPRO orange binding below 70°C, a sharp rise in fluorescence of the probe at higher temperatures, and a single T_m of ∼81 to 82°C (Fig. 2). All of the SFV DIII proteins behaved similarly but showed lower melting temperatures ranging from ∼64 to 70°C (Fig. 2). Together, our results indicated that recombinant CHIKV HDIII and HDIIIS were correctly folded and highly stable, in keeping with the known immunoglobulin-like fold and 3 disulfide bonds of alphavirus E1 DIII.

We then compared the inhibitory activities of CHIKV HDIII and HDIIIS proteins with those of the analogous SFV proteins. The proteins were tested by inclusion during a 1-min low-pH treatment of CHIKV, SFV, or SINV prebound to BHK cells. In agreement with our previous studies (24), SFV HDIII efficiently inhibited membrane fusion and infection by SFV and SINV, and we observed that it also showed strong cross-inhibition of CHIKV fusion (Fig. 3A). In contrast, CHIKV HDIII protein showed no inhibition of CHIKV, SFV, or SINV fusion (Fig. 3A), and this lack of activity was observed even at high concentrations (up to 50 μM) of the CHIKV protein (data not shown). As previously described (24), the SFV stem region produced an increase in the efficiency of inhibition, with the SFV HDIIIS protein completely inhibiting SFV, SINV, and CHIKV fusion at a concentration of ∼0.5 μM (Fig. 3B). The CHIKV HDIIIS protein very efficiently inhibited CHIKV fusion, with complete inhibition at 0.5 μM (Fig. 3B). CHIKV HDIIIS also inhibited SFV and SINV fusion, although complete inhibition of these viruses required concentrations of ∼2 μM (Fig. 3B). Thus, CHIKV HDIIIS showed both homologous and cross-inhibition of alphavirus fusion. Unlike SFV, however, the CHIKV stem region was required for inhibition.

We hypothesized that the difference in inhibition by SFV HDIII versus CHIK HDIII was due to a difference in the binding affinity of these proteins for their target, the E1 core trimer intermediate formed during virus fusion (24, 25). The core trimer can be produced in vitro by low-pH treatment of purified DI/II proteins in the presence of cholesterol-containing liposomes (33). The membrane-inserted core trimer binds exogenous DIII, and the complex can be floated with the liposomes on sucrose gradients (33). We used this system to assay DIII binding to SFV or CHIKV DI/II core trimers. As predicted by our fusion results, the SFV HDIII and HDIIIS proteins interacted with the CHIKV core trimer (Fig. 4A) or the SFV core trimer (33; data not shown). In contrast, CHIKV HDIII did not cofloat with either the CHIKV core trimer (Fig. 4B, first two lanes) or the SFV core trimer (data not shown), in keeping with its lack of inhibition of fusion.

The E1 stem region is ∼28 residues long, extending from the end of DIII at P384 to K412 just prior to the membrane anchor (Fig. 5A). To define the extension of the stem required to block fusion, we produced CHIKV HDIIIS proteins truncated after position 388, 392, or 397, as well as HDIII and HDIIIS, which end at P384 and K412, respectively (Fig. 5A). We tested 10 μM concentrations of these proteins for inhibition of CHIKV or SFV fusion with BHK cells, using the assay described in Fig. 3. CHIKV HDIII proteins that extend to at least residue 392 inhibited membrane fusion by both viruses, while a version truncated at residue 388 showed little inhibition (Fig. 5B). Binding studies of these DIII proteins with the CHIKV core trimer demonstrated that DIII protein truncated at residue 392 cofloated with the core trimer, while DIII truncated at residue 388 did not (Fig. 4B). Thus, the stem residues ³⁸⁹NYPA³⁹² (Fig. 5A) and/or this stem length were important in promoting the binding of CHIKV HDIII protein to the core trimer and the inhibition of virus fusion.

To quantitatively analyze the affinity of the DIII-core trimer interaction, we developed an assay based on the change in fluorescence anisotropy upon binding of fluorescently labeled DIII protein to the DI/II core trimer. SFV or CHIKV DIII proteins were labeled with an amine-reactive form of fluorescein (details in Methods). While SFV DIII and DIIIS proteins were successfully labeled, the CHIKV DIII proteins precipitated under a variety of dye conjugation conditions. We also noted that the CHIKV DI/II core trimer was not stable in octylglucoside. We therefore used the SFV proteins for further studies. The labeled SFV DIIIS and DIII proteins showed comparable coflotation with the SFV DI/II core trimer as unlabeled DIII proteins (data not shown). We also established a bacterial expression and in vitro refolding system for the SFV E1 DI/II protein (details in Materials and Methods). Similar to the protein produced in Drosophila S2 cells, the SFV DI/II protein produced in bacteria formed stable core trimers in the presence of liposomes at low pH, but not at neutral pH or in the absence of target membranes. These core trimers bound DIII proteins at neutral pH, again similar to the S2 core trimers (data not shown).

Fluorescence polarization (FP) measurements were then performed to characterize the binding of labeled SFV DIII proteins to SFV E1 DI/II trimers. Dilutions of DI/II trimers or the DI/II monomer were mixed with DIII proteins and incubated for 1 h at neutral pH, and the FP was quantitated (Fig. 6A). The FP of SFV DIIIS showed a reproducible increase in the presence of DI/II trimers, while the DI/II monomer did not change the DIIIS FP. Comparable changes in FP were obtained using the S2 cell-expressed core trimer as the target for DIIIS binding (data not shown). The K_d for binding of SFV DIIIS with the DI/II core trimer was calculated as 85 ± 5 nM (Fig. 6A). Binding of labeled DIII without the stem did not reach saturation even at the maximal concentration of trimer used (Fig. 6A), and analysis using the Prism program estimated a K_d ∼100 times higher than that of DIIIS. These values are similar to those previously determined for DIII and DIIIS inhibition of virus fusion (24). The binding of DIIIS to the DI/II core trimer was efficiently competed by excess unlabeled DIIIS, confirming the specificity of the FP assay (Fig. 6B).

Our initial characterization experiments confirmed that CHIKV fusion was both low pH and cholesterol dependent. However, even though E1's sequence and structure are strongly conserved among the SFV, SINV, and CHIKV E1 proteins (20–22), we observed several differences in the properties of DIII proteins from CHIKV versus SFV. The CHIKV DIII proteins required the N-terminal His tag for successful expression. We have not tested alternative sequence tags for their effects. Once refolded in vitro, thermal denaturation tests showed that CHIKV HDIII and HDIIIS have a higher melting temperature (+ ∼10 to 18°C) than the analogous SFV proteins, suggesting higher stability. Unlike SFV DIII proteins, labeling of the CHIKV DIII proteins with amine-reactive fluorescein caused their aggregation under all conditions we tested, making them poor candidates for fluorescence anisotropy assays of DIII-core trimer interaction. In contrast, E1 DI/II proteins from both CHIKV and SFV were efficiently expressed and purified from S2 cells and showed similar activities in formation of core trimers that bound DIII. We also successfully produced SFV E1 DI/II by bacterial expression and refolding. Although in this system DI/II is not glycosylated, it was comparable to the Drosophila-expressed protein in formation of stable core trimers that efficiently bound DIII. This system avoids the higher cost of eukaryotic cell expression, and we anticipate that the successful bacterial production of active alphavirus DI/II and DIII will aid in large-scale production of these proteins as needed for structural studies and small molecule screens.

The structure of the SFV E1 postfusion trimer shows that DIII is positioned within the groove formed by the DI/II regions of two E1 molecules (23). The inhibition of the rapid virus fusion reaction by exogenous DIII proteins (24, 25) and the formation of a stable complex between exogenous DIII protein and the core trimer (24, 33) suggest that the affinity of this interaction is high. DIII proteins inhibit across alphavirus species (24; this work), and the CHIKV and SINV DIII-stem proteins are 49% and 45% identical to SFV DIII-stem, respectively. Here we observed a prominent difference between the SFV and CHIKV DIII proteins in their requirements for the stem region. Unlike SFV, the CHIKV DIII required the first 9 residues of the stem in order for the protein to inhibit fusion and stably bind the core trimer. Similar CHIKV DIII melting temperatures were observed with or without the stem, so this requirement was not due to effects on overall domain folding or stability.

The length of the alphavirus E1 stem region is conserved. A minimal length of the SFV E1 stem is required in order to allow DIII foldback, hairpin formation, and fusion (25, 41). The presence of the stem causes increased stability of the SFV DIII-DI/II core trimer complex (33). However, SFV mutagenesis studies showed that a specific stem sequence is not essential for fusion (41), and tests of several isolated SFV stem peptides showed that they did not inhibit fusion (41). These results suggested that DIII-core trimer interactions are important in promoting the hairpin. However, clearly the previous data and the results of this study also suggest a role for the stem region, probably occurring after the initial foldback of the DI-DIII linker and DIII (25, 39).

Using the SFV E1 homotrimer structure, we analyzed the residues in DIII-stem for interactions with the core trimer (Table 2 and Fig. 7). (Note that in the structure the stem ends at residue 391.) Initial analysis used the PISA (Protein Interfaces, Surface and Assemblies Service) program at the European Bioinformatics Institute (42) and a modified SFV E1 trimer PDB file containing one DI/II and the DIII from the same E1 protein or the adjacent E1 protein. All DIII residues containing surface area buried by contacts with DI/II were further analyzed with the programs CSU (Contact of Structural Units) (43) and PyMOL (44). Residues within 4 Å of each other were compared with the corresponding residues in CHIKV E1. Predicted interactions between SFV DIII-stem residues and the core trimer are listed in Table 2 and detailed in Fig. 7, and can occur between DIII and DI/II of the same E1 chain or the adjacent E1 chain.

We then compared these interactions with the residues at the same positions in CHIKV E1. The interacting residues are highly conserved between the sequences of SFV and CHIKV E1, in keeping with the cross-inhibition observed with DIII proteins from these two viruses. While a number of the predicted interactions occur via the main chain, both DIII and stem show several side-chain interactions with the core trimer. Most of these residues are conserved between SFV and CHIKV, with the exception of H333 on SFV DIII. In the case of CHIKV, the interaction of DIII H333 with DI D145 is replaced by an M333-T145 pair. Although prior studies with SFV showed that alanine substitution of H333 did not detectably affect virus growth or fusion (45), it is possible that these sequence differences contribute to the difference in CHIKV DIII interaction. The stem truncation experiments in Fig. 5 implicate stem residues ³⁸⁹NYPA³⁹² in stabilizing CHIKV DIII binding. The SFV homotrimer structure predicts interaction of Y390 with P237 in the ij loop of DII. This Y390-P237 pair is conserved between SFV and CHIKV and is relatively conserved among alphaviruses. Although Y390 is not essential for SFV fusion (41), it may be important in stabilizing the binding of CHIKV DIII protein, perhaps by compensating for the loss of the H333-D145 interaction. It is also possible that in the context of CHIKV E1, other residues of the stem promote binding and inhibition by exogenous DIII.

Subtle differences in fusion protein sequence or structure could result in a significant difference in the requirements for the fusion process. This is an important issue to consider in designing strategies for inhibitors of medically important alphaviruses, such as CHIKV. Since DIII inhibition of virus fusion and infection has been mainly developed and studied using the SFV model, we wanted to validate this approach for CHIKV. We here confirmed cross-inhibition of SFV and CHIKV fusion by SFV and CHIKV DIII proteins. CHIKV HDIIIS inhibition of membrane fusion was optimal for CHIKV, while somewhat higher concentrations were required to inhibit SFV and SINV. In contrast, SFV HDIIIS protein inhibited the fusion of all three viruses with comparable high efficiencies. In flotation experiments, the SFV and CHIKV HDIIIS proteins bound to the core trimer of SFV or CHIKV to a similar extent. The measured high affinity of SFV DIIIS for the DI/II core trimer (85 nM) from our fluorescence assay is in good agreement with the previously observed 50% inhibitory concentration (IC₅₀) of ∼100 nM for this protein in virus fusion assays (24). Our results suggest that the sites of contact for SFV HDIII or HDIIIS with the core trimer are present during its interaction with CHIKV core trimer and result in cross-inhibition. Thus, the assay we describe for the in vitro interaction of SFV DIIIS with the core trimer has potential for use as a screen for broad-specificity inhibitors of alphavirus membrane fusion.