A single point mutation in S2 affords successful rescue of rVSV bearing the SHC014-CoV bat CoV spike

We sought to generate a panel of replication-competent rVSVs bearing the spikes of bat-origin CoVs as their only entry glycoproteins. We initially selected four sarbecoviruses from clades 1, 2, and 3 with potential for zoonotic transmission in humans—WIV-1-CoV, SHC014-CoV, RmYN02-CoV, and BtKY72-CoV [4,38,39]. Previous work has shown that truncations of ≥21 amino acids in the cytoplasmic tails of divergent betacoronavirus spikes enhance the infectivities of VSV and retrovirus surrogates bearing them [40,41]. Accordingly, we replaced the VSV glycoprotein G gene with each CoV S gene engineered to encode a spike protein lacking 21 residues at its C–terminus. We also inserted a sequence encoding the enhanced green fluorescent protein (eGFP) as an independent transcriptional unit at the first position of the VSV genome, as described previously [42,43].

We attempted to generate each rVSV using a well-established plasmid-based rescue system [42,44–46]. rVSV-WIV-1-CoV S rescued readily, consistent with previous work using coronavirus reverse genetics and describing WIV-1-CoV’s capacity to use ACE2 from a wide variety of hosts, including humans, civets, and Chinese horseshoe bats [4]. By contrast, numerous attempts failed to yield rVSVs for RmYN02-CoV and BtKY72-CoV (Fig 1A). rVSV-SHC014-CoV S was similarly challenging to generate; however, a single experiment yielded a replicating viral stock. Nanopore DNA sequencing of an RT-PCR product derived from this viral population revealed that a single point mutation that results in an A835D substitution in the spike protein (corresponding to position 852 in the SARS-CoV-2 spike), had reached fixation in the population (S1 File). Sanger sequencing confirmed the presence of A835D as the only mutation in all six plaques isolated from this population. Importantly, a VSV cDNA clone bearing SHC014-CoV S(A835D) was successfully recovered in each of five attempts, whereas its WT counterpart was not (Fig 1A and 1B), providing evidence that A835D drives the generation and propagation of rVSV-SHC014-CoV S.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. SHC014-CoV S A835D allows for rVSV rescue and is conserved amongst sarbecoviruses.

(a) Summary of recovery attempts of rVSVs bearing SHC014-CoV spike proteins. Asterisk refers to rescue that yielded the A835D substitution. (b) Supernatants from 293FT cells co-transfected with plasmids encoding for VSV genomes expressing eGFP and WT or A835D variants of SHC014-CoV spike as well as helper plasmids, were used to infect Vero cells. Representative images show eGFP expression in Vero cells at indicated time points (days post-infection [dpi]). Scale bar, 100 μm. (c) Alignment of amino acid sequences in the FPPR region (rounded rectangle) for selected coronavirus spike proteins. Subgenera of the genus Betacoronavirus are indicated in italics (Sarbecovirus, Hibecovirus, Nobecovirus, Merbecovirus, and Embecovirus). Sarbecoviruses are color-coded by clade (1a: SARS-CoV–like, red; 1b: SARS-CoV-2–like, green; 2: Southeast Asian bat-origin CoV, blue; 3: non-Asian bat-origin CoV, purple). Spikes investigated in the current study are in bold.

https://doi.org/10.1371/journal.ppat.1012704.g001

Cognate substitutions facilitate the rescue of rVSVs bearing spikes from divergent bat sarbecoviruses

Amino acid sequence alignment of SHC014-CoV S with those of divergent coronavirus spikes indicated that residue 835 is in the FPPR of S2 (Fig 1C). The FPPR forms part of an intersubunit interface that is largely buried in the pre-fusion trimer, is highly conserved, and is proposed to play a key role in regulating spike conformation [47]. Most of the betacoronavirus spike sequences we examined, including all sarbecoviruses and merbecoviruses, bore an Ala at the corresponding residue (Fig 1C). By contrast, hibecoviruses, nobecoviruses, and embecoviruses do not. Accordingly, we postulated that a substitution cognate to A835D in the FPPR of other bat-origin sarbecovirus spikes could also enhance entry in the rVSV backbone. Consistent with this hypothesis, we could recover rVSV-RmYN02-CoV S(A806D) and rVSV-BtKY72-CoV S(A837D), even as their WT counterparts failed to rescue (Fig 1A). Thus, the effect of this amino acid substitution in the spike FPPR appears generalizable to multiple bat CoVs, at least in the genetic and structural context of spikes belonging to the sarbecovirus subgenus.

A835D greatly enhances the infectivity of authentic coronavirus-like vectors

We next evaluated the possibility that A835D arose as a spike adaptation specific to the 21-amino acid cytoplasmic tail truncation and/or the heterologous VSV context. To investigate this, we used a recently developed SARS-CoV-2 replicon-based system to generate and evaluate single-cycle replicon-delivery particles (RDPs) trans-complemented with full-length WT or A835D SHC014-CoV spikes [48]. We also generated single-cycle VSV pseudotypes (scVSV) bearing WT or A835D SHC014-CoV spikes, as described previously [43,49]. We compared the infectivities of the VSV (Fig 2A) and RDP (Fig 2B) preparations in DBT-9 murine astrocytoma cells previously shown to be ACE2-null and engineered to ectopically express human and intermediate horseshoe bat (Rhinolophus affinis) orthologs of ACE2 (HsACE2 and RaACE2, respectively) [12,50]. Neither the VSVs nor the RDPs could transduce the parental ACE2-null DBT-9 cells, indicating that they require ACE2. This is consistent with previous findings that the SHC014-CoV spike recognizes ACE2 from humans, civets, and horseshoe bats and uses them as entry receptors [13,20]. However, ACE2 overexpression also had little effect on cell entry by both VSVs and RDPs bearing the SHC014-CoV(WT) spike. By contrast, the SHC014-CoV(A835D) spike mediated dose-dependent increases in viral entry, particularly in the cells expressing human and horseshoe bat ACE2 (Fig 2). Further, rVSV-WIV-1-CoV S(A835D) also exhibited enhanced ACE2-dependent entry relative to WT (S1 Fig). We conclude that the infection-enhancing activity of the A835D mutation is autonomous to the spike protein for multiple sarbecoviruses and can be generalized to authentic CoV virions bearing full-length spikes.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. A835D increases infectivity of cells in an ACE2-dependent manner.

(a) Parental DBT-9 cells or DBT-9 cells overexpressing HsACE2 or RaACE2 were infected with pre-titrated amounts of scVSV-SHC014-CoV WT or scVSV-SHC014-CoV A835D. Infection was scored by eGFP expression at 16–18 hours post-infection (average±SD, n = 5–8 from 3 independent experiments). A range of 7.2×10² to 4.5×10⁶ viral genome-equivalents (GEQ) was used. (b) Parental DBT-9 cells or DBT-9 cells overexpressing HsACE2 or RaACE2 were infected with CoV RDPs bearing WT or A835D SHC014-CoV spikes. Infection was scored by mNeonGreen expression at 16–18 hours post-infection (average±SD, n = 10–12 from 4–5 independent experiments). Ranges of 8.2×10³ to 1.8×10⁷ GEQ for WT and 6.1×10³ to 4.0×10⁷ for A835D were used. Infectivity (%) shown is the proportion of infected cells to the total number of cells for each viral dilution. Groups (WT vs. mutant for each cell line) were compared with Welch’s t-test with Holm-Šídák correction for multiple comparisons. ns p>0.05; ** p<0.01; *** p<0.001; **** p<0.0001. Only the statistically significant comparisons are shown.

https://doi.org/10.1371/journal.ppat.1012704.g002

A835D does not alter spike expression or VSV incorporation

We postulated that the A835D substitution acts in part by increasing the cellular expression and viral incorporation of the SHC014-CoV spike. To test this hypothesis, we transfected 293T cells with plasmid expression vectors encoding the WT and A835D spikes. We observed similar levels of WT and A835D full-length spikes, both in total and at the cell surface (S2A Fig), indicating that A835D does not enhance SHC014-CoV spike protein expression or steady-state plasma membrane localization. Despite this, scVSVs bearing the A835D spike were much more infectious than their WT counterparts in Vero cells, recapitulating the phenotypes observed in the rVSV and CoV RDP systems (S2C Fig).

We next used an ELISA to directly assess spike incorporation into viral particles. Specifically, scVSV-coated plates were probed with antibodies specific for the SARS-CoV-2 spike and the VSV internal matrix protein, M, and the Spike-to-M ratio was used as a measure of the relative incorporation of spike into virions. A835D had little or no effect in this assay (S2B Fig). Taken together, these findings suggest that A835D exerts its effects at a post-assembly step in the viral life cycle, likely during cell entry.

A835D enhances SHC014-CoV spike:ACE2 binding

The preceding findings in cell lines overexpressing ACE2 suggested that A835D acts at least in part by enhancing the capacity of the SHC014-CoV spike to use ACE2 for cell entry. Accordingly, we considered the possibility that, despite its location at a distance from the RBD, A835D may nonetheless augment spike:ACE2 recognition. To test this hypothesis, we first evaluated the capacity of HsACE2 to capture scVSV-SHC014-CoV S particles in an ELISA. Soluble HsACE2-coated plates were incubated with genome-normalized amounts of scVSVs, and particle capture was detected with a CoV spike-specific monoclonal antibody (mAb). Despite previous evidence that the isolated SHC014-CoV spike RBD recognizes HsACE2 with high avidity [20], WT SHC014-CoV spike could capture HsACE2 only poorly, whereas its A835D counterpart exhibited a dramatic, dose-dependent increase in HsACE2 capture (Fig 3A).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. A835D increases SHC014-CoV S binding to HsACE2.

(a) Genome-normalized amounts of scVSV particles bearing WT or A835D SHC014-CoV spike were diluted with serial 3-fold dilutions and loaded on an ELISA plate precoated with soluble HsACE2 and detected with a spike-specific mAb followed by an anti-human HRP-conjugated secondary antibody (average±SD, n = 6–8 from 3–4 independent experiments). A range of 1.6×10⁶ to 1.3×10⁸ viral GEQ was used. ELISA signal values were normalized to the highest absorbance for scVSV-SHC014-CoV WT for each replicate. Groups (WT vs. mutant) were compared with Welch’s t-test with Holm-Šídák correction for multiple comparisons. ns p>0.05; ** p<0.01; *** p<0.001; **** p<0.0001. (b) Sensorgrams of SHC014-CoV variant spikes and a pre-fusion SARS-CoV-2 spike stabilized with 6 proline substitutions (HexaPro) binding to HsACE2 by BLI. Similar response levels of spike were captured using an anti-T4 fibritin (foldon) antibody bound to BLI biosensors and subsequently dipped into wells containing 0.5 μM HsACE2. Traces of only one replicate of three are shown for clarity. (c) Graph of BLI binding response collected in triplicate for the indicated SHC014-CoV spike variants (average±SD). WT and A835D were compared by one-way ANOVA, with Dunnett’s correction for multiple comparisons. ns p>0.05; ** p<0.01; *** p<0.001; **** p<0.0001. Only the statistically significant comparisons are shown.

https://doi.org/10.1371/journal.ppat.1012704.g003

We next deployed an established biolayer interferometry (BLI) assay to measure binding of the WT and A835D spikes to HsACE2. Because the A835D substitution is not located in the receptor-binding motif, the binding response can be used as a proxy for RBD exposure or ‘availability’ to bind ACE2, as shown previously [51]. We began by producing trimeric ectodomains of both spike proteins with a C–terminal T4 fibritin (foldon) domain as described previously for other spikes [52]. Then, recombinant SHC014-CoV spike ectodomains were captured to similar levels onto BLI probes using an anti-foldon IgG and subsequently dipped into solutions containing HsACE2. A SARS-CoV-2 spike stabilized with six prolines (Hexapro) served as a positive control and efficiently captured HsACE2 from solution, as expected [53]. In contrast, WT SHC014-CoV spike bound weakly to HsACE2, and its activity was substantially boosted by the A835D substitution, corroborating our ELISA results (Fig 3B and 3C).

Finally, to further investigate the effect of A835D on the conformation of the SHC014-CoV spike, we assessed the binding of a panel of spike-specific monoclonal antibodies (mAbs) by scVSV-SHC014-CoV S ELISA, as above (S3 Fig). Adagio-2 (ADG-2), an RBD binder that can only recognize the spike in its open, “RBD-up” conformation [54], resembled HsACE2 in showing greatly enhanced binding to the A835D SHC014-CoV spike compared to WT. By contrast, no such difference was observed with S309, an RBD-specific mAb that can recognize both open and closed spikes [55,56], and with two S2-directed antibodies (S3 Fig). Together, these findings strongly suggest that the A835D substitution augments the SHC014-CoV spike’s capacity to bind HsACE2 by increasing the availability of its RBDs for receptor engagement, but without causing large-scale changes in the pre-fusion conformation of the spike.

High-resolution structure of a SHC014-CoV spike ectodomain

We performed cryo-electron microscopy (cryo-EM) studies to define the interactions made by residue 835 in the WT SHC014-CoV spike. Purified, soluble spike ectodomain was applied to a cryo-EM grid and vitrified. Sufficient particle density with little aggregation was observed, and a dataset containing 2,025 micrographs was collected (S1 Table). Extracted particles were subjected to two-dimensional (2D) classification, which resulted in class averages populated with fully closed trimers, wherein all three RBDs are in the ‘down’ position. This finding stands in contrast to previous cryo-EM studies of SARS-CoV-2 spike ectodomains, in which substantial proportions of partially open spikes with at least one RBD in the ‘up’ position were observed [57–59]. A 3.1-Å–resolution 3D reconstruction of the WT SHC014-CoV spike trimer was obtained and found to resemble that of the closed SARS-CoV-2 spike (Figs 4A, S4 and S5), which was expected given the ≈80% sequence identity between the proteins. The flexibility of the NTD led to poorly resolved regions of the map for this domain, preventing the modeling of some loops based on our reconstruction. Additionally, the FP was not visible in our cryo-EM map, as also observed in many SARS-CoV-2 spike structures. In all, we were able to build residues 33–1124, excluding 45–46, 71–78, 94–97, 121–123, 134–155, 167–182, 205–212, 237–251, 268–270, 617–621, 665–671, and 812–831. C–terminal FPPR residues were resolved (residues 832–840), affording an examination of residue A835 and its local environment.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Structural determination of SHC014-CoV spike.

(a) Cryo-EM map and model of the WT SHC014-CoV spike. The cryo-EM map is shown for two protomers and the third protomer is shown as a ribbon representation. (b) Ala835 is proximal to Asp555 of the neighboring protomer, which is colored by electrostatic surface potential, red (negative) to blue (positive). Scale bar, -10 to +10 kT/e. (c) Ala835 makes van der Waals interactions with Val556 of a neighboring protomer (blue) and Val946 of the same protomer (orange). Lys837 (orange) makes a salt bridge with Asp601 of a neighboring protomer (blue). (d) S2 of one protomer (orange) and subdomain 1 (SD1) and RBD of an adjacent protomer (both in blue) are shown as ribbons to depict the distance between the Ala835-containing hydrophobic pocket to the RBD of the adjacent protomer. Labelled residues are shown as spheres.

https://doi.org/10.1371/journal.ppat.1012704.g004

A835 is part of an α-helix proximal to the FP, and it lies at the interface between protomers (buried surface area upon oligomerization: 22.7%). The A835 sidechain is pointed inward toward V946 in the S2 heptad repeat 1 (HR1) of the same protomer and toward residues D555, V556 and S557 in the S1 subdomain I (SD1) of the neighboring protomer (Fig 4B). Residues 555–557 in SD1 connect A835 to the RBD in the neighboring protomer (Fig 4D). Structure-based in silico analyses consistently suggested that A835D substitution increases the binding free energy between protomers, substantially affecting the stability of the trimer (ΔΔG_trimer: > 3 Kcal/mol; ΔΔG_interface: +1.25 Kcal/mol by FoldX; +1.08 Kcal/mol by flexddG). We propose that the introduction of a negatively charged residue at position 835 decreases the stability of the spike trimer by disrupting a network of hydrophobic interactions within (A835–V963) and between protomers (A835–V556; Fig 4B) and triggering unfavorable electrostatic interactions between protomers by virtue of its close proximity to an electronegative surface region in the neighboring protomer (Fig 4C). In support of this hypothesis, scVSVs bearing a Glu at position 835 (A835E) showed a similar enhancement of infectivity over WT as A835D (S6 Fig). Further evidence for the reduced stability of the A835D spike includes a decrease in the expression of its ectodomain relative to WT (S7A Fig). Moreover, the A835D spike ectodomain was found to be unstable on cryo-EM grids with mostly dissociated particles and few trimeric spikes, thereby preventing high-resolution structure determination (S7B and S7C Fig). Together, these results strongly suggest that the A835D substitution has a destabilizing effect on the pre-fusion conformation of the SHC014-CoV spike, resulting in increased transition of RBDs from ‘down’ to ‘up’ positions and enhanced ACE2 binding.

A835D reduces spike thermostability and alters the temperature dependence of RBD availability

To test the hypothesis that A835D destabilizes the SHC014-CoV spike, we measured the temperature dependence of viral infectivity loss (inactivation) as a surrogate for the thermal stability of the viral membrane-embedded pre-fusion spike, as we described previously for VSVs bearing filovirus glycoproteins [60]. scVSV particles bearing WT and A835D SHC014-CoV spikes were preincubated at different temperatures, then shifted to 4°C, and titrated for infectivity on DBT-RaACE2 cells. WT spike-containing particles suffered loss of infectivity between 48.5–51°C, with no detectable infectivity at 53°C. We observed a left shift in temperature-dependent infectivity loss for A835D—the mutant particles underwent inactivation between 41–46°C, with no detectable infectivity at 48.5°C (Fig 5A). Thus, the A835D substitution renders the SHC014-CoV spike more susceptible to thermal inactivation in the context of intact viral particles.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. A835D reduces thermostability of the SHC014-CoV spike.

(a) scVSV-SHC014-CoV S particles were incubated at various temperatures for 1 h and used to infect DBT-9 cells overexpressing RaACE2. (average±SD, n = 4–6 from 2–3 independent experiments). Infectivity levels were normalized to the infectivity percentage at 41°C for each virus. Area under the curve (AUC) values were calculated for each curve, and groups were compared by one-way ANOVA with Dunnett’s correction for multiple comparisons. (b) Pre-titrated amounts of scVSV-SHC014-CoV S particles were incubated at various temperatures for 1h and loaded onto HsACE2-coated ELISA plates. A spike-specific mAb and anti-human HRP-conjugated secondary antibody were used to detect the spike protein. 9.2×10⁶ viral GEQs per well were used. (average±SD, n = 12 from 6 independent experiments). ELISA signals were normalized to the absorbance observed for scVSV-SHC014-CoV WT at 53.5°C. Groups were compared by two-way ANOVA with Tukey’s correction for multiple comparisons, ns p>0.05; ** p<0.01; *** p<0.001; **** p<0.0001.

https://doi.org/10.1371/journal.ppat.1012704.g005

In parallel experiments, we measured the activity of scVSVs bearing WT and A835D SHC014-CoV spikes in the HsACE2 capture ELISA as a function of temperature. We observed an increase in spike:ACE2 binding for WT particles between 43.5–48°C, a temperature range just below and overlapping that of viral inactivation for infection. Strikingly, A835D particles exhibited higher levels of HsACE2 binding at all temperatures tested, with an increase that titrated between 41–43.5°C, consistent with their temperature range for inactivation (Fig 5B). We conclude that the A835D substitution has a destabilizing effect on the pre-fusion conformation of the SHC014-CoV spike, sensitizing it to undergo conformational changes in response to heat that result in enhanced RBD exposure and culminate in viral inactivation. We surmise that this phenotype, together with data in Figs 4 and 5, reflects the increased propensity of the A835D spike to sample open conformers with RBDs in the ‘up’ position at physiological temperatures (see below).

NTD substitution F294L enhances viral entry and ACE2 recognition by WT and A835D spikes

In seeking to identify key intra-spike interactions altered by the A835D substitution while structural studies on the SHC014-CoV spike (Fig 4) were in progress, we modeled both WT and A835D SHC014-CoV spikes using AlphaFold2 [61,62]. These studies suggested that A835D may engage in a new polar interaction with residue T842 in the FPPR. Although we could not structurally corroborate this hypothesis given the relative instability of the A835D ectodomain, we nonetheless attempted to generate an rVSV bearing an SHC014-CoV(A835D/T842A) double-mutant spike. As with the WT spike, the double mutant was challenging to rescue, but one attempt yielded a replicating viral population. Nanopore sequencing of viral cDNA from this population indicated that both parental mutations were present but uncovered one additional point mutation in the spike NTD in most reads (S2 File), causing the F294L substitution (corresponding to position F306 in the SARS-CoV-2 spike). Sanger sequencing confirmed the acquisition of F294L and retention of both parental substitutions in all three plaques isolated from this population, strongly suggesting that F294L and/or T842A played functional roles in rVSV recovery.

To dissect the effects of these substitutions on viral entry, we generated scVSVs bearing WT and mutant SHC014-CoV spikes. Spikes containing F294L alone exhibited increases relative to WT in cell entry (Fig 6A) and HsACE2 binding (Fig 7A) resembling those observed for A835D. F294L+A835D conferred a further enhancement in infectivity relative to the single mutants in DBT-9 cells over-expressing HsACE2 and RaACE2 (Fig 6A). Because F294L was originally selected in the A835D/T842A background, we additionally tested the infectivity of scVSVs bearing SHC014-CoV spikes with A835D+T842A and F294L+A835D+T842A (S8 Fig). We saw no significant difference in infectivity when T842A was introduced into scVSV-SHC014-CoV S particles bearing A835D or F294L+A835D. These findings strongly suggested that F294L, but not T842A, afforded a growth advantage to viral clones in the rVSV rescue experiments.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. A835D and F294L substitutions enhance infectivity.

(a) Parental DBT-9 cells or DBT-9 cells overexpressing HsACE2 or RaACE2 were infected with pre-titrated amounts of scVSV-SHC014-CoV particles bearing WT, A835D, F294L, or F294L+A835D spike. Infection was scored by eGFP expression at 16–18 hours post-infection (average±SD, n = 5–8 from 3 independent experiments). A range of 1.5×10³ to 9.6×10⁶ viral GEQ was used. Infectivity (%) shown is the proportion of infected cells to the total number of cells for each viral dilution. Groups (WT vs. mutant for each cell line) were compared with Welch’s t-test with Holm-Šídák correction for multiple comparisons. ns p>0.05; ** p<0.01; *** p<0.001; **** p<0.0001. Only the statistically significant comparisons are shown. (b) Supernatants from 293FT cells co-transfected with plasmids encoding for VSV genomes expressing eGFP and WT, A835D, F294L, or F294L+A835D variants of SHC014-CoV spike as well as helper plasmids, were used to infect Vero cells. Representative images show eGFP expression in Vero cells at 24, 60, 108, 204, 252, 276, and 273 h post-infection [hpi]. Representative images are shown. Scale bar, 100 μm.

https://doi.org/10.1371/journal.ppat.1012704.g006

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. A835D and F294L enhance ACE2 binding but act through distinct mechanisms.

(a) Genome normalized amounts of scVSV particles bearing WT, A835D, F294L, or F294L+A835D SHC014-CoV spike were diluted with serial 3-fold dilutions and loaded on an ELISA plate precoated with soluble HsACE2 and detected with a spike-specific mAb followed by an anti-human HRP-conjugated secondary antibody (average±SD, n = 6–8 from 3–4 independent experiments). A range of 5.5×10³ to 1.2×10⁷ viral GEQ was used. ELISA signals were normalized to the absorbance observed for scVSV-SHC014-CoV WT at the highest number of viral genomes. Groups (WT vs. mutant) were compared with Welch’s t-test with Holm-Šídák correction for multiple comparisons. ns p>0.05; ** p<0.01; *** p<0.001; **** p<0.0001. (b) Sensorgrams of SHC014-CoV spike variants binding to HsACE2 by BLI. Spike variants were captured to similar response levels using an anti-T4 fibritin antibody bound to BLI biosensors and subsequently dipped into wells containing 0.5 μM HsACE2. One of three replicates is shown for clarity. (c) scVSV-SHC014-CoV S particles were incubated at various temperatures for 1h and used to infect DBT-9 cells overexpressing RaACE2. Infection levels were normalized to values observed at 41°C for each virus (average±95%CI, n = 4–8 from 2–4 independent experiments). AUC values were calculated for each curve, and groups were compared by one-way ANOVA with Dunnett’s correction for multiple comparisons. Statistical significance is shown for WT vs. A835D, WT vs. F294L+A835D, F294L vs. A835D, and F294L vs. F294L+A835D. Differences among all other two-way comparisons were not statistically significant (d) Pre-titrated amounts of scVSV-SHC014-CoV S particles were incubated at various temperatures for 1 h and loaded onto HsACE2-coated ELISA plates. A spike-specific mAb and anti-human HRP-conjugated secondary antibody were used to detect the spike protein. Absorbance values were normalized to that observed for scVSV-SHC014-CoV WT particles at 54.5°C. (average±95%CI, n = 12 from 3 independent experiments). 5.1×10⁶ viral GEQs were used per well. AUC values were calculated for each curve, and groups were compared by one-way ANOVA with Dunnett’s correction for multiple comparisons. (e) Pre-titrated amounts of scVSV-SHC014-CoV S bearing WT, A835D, F294L, or F294L+A835D spikes were incubated with serial 3-fold dilutions of mAb ADG-2 starting at 500 nM, at 37°C for 1 h. Virus:mAb mixtures were applied to monolayers of DBT-9 cells overexpressing RaACE2 cells. At 18 h post-infection, cells were fixed, nuclei were counterstained, and infected cells were scored by eGFP expression (average±SD, n = 9 from 3 independent experiments). Infectivity levels were normalized to the infectivity percentage with no mAb present for each virus. AUC values were calculated for each curve, and groups were compared by one-way ANOVA with Dunnett’s post hoc test, ns p>0.05; ** p<0.01; *** p<0.001; **** p<0.0001. Only the statistically significant comparisons are shown.

https://doi.org/10.1371/journal.ppat.1012704.g007

Accordingly, we attempted to rescue rVSVs with SHC014-CoV(F294L) and SHC014-CoV(F294L+A835D) spikes in parallel with rVSV-SHC014-CoV(A835D) and WT spikes. Rescues of rVSV-SHC014-CoV bearing spikes with F294L, A835D, and F294L+A835D were all successful, with F294L+A835D rescuing the fastest, followed by A835D, and then F294L (Fig 6B). As we observed previously (Fig 1B), WT did not rescue. Sanger sequencing of the spike gene in isolated plaques confirmed the presence of only these mutations in each respective virus prep.

The boosts in infectivity and rVSV rescue seen in the context of F294L+A835D were also associated with a commensurate, dramatic gain in HsACE2 binding to viral particles at 37°C (Fig 7A). BLI analyses of HsACE2 binding to SHC014-CoV spike ectodomains bearing F294L and F294L+A835D yielded largely concordant results, but with some differences (Fig 7B). Specifically, although F294L did significantly enhance spike:HsACE2 binding, its effect was smaller than that observed for A835D. Further, F294L+A835D did not boost HsACE2 binding over A835D alone in the BLI assay. These discordances between the virus- and spike ectodomain-based ACE2 binding assays may simply reflect the greater instability of the ectodomains bearing A835D relative to viral membrane-embedded spikes—juxtamembrane and transmembrane sequences in the latter are predicted to counteract the destabilizing effects of the ectodomain substitutions [63–65]. Alternatively, or in addition, they may reflect total saturation of the BLI biosensors, preventing further increases in RBD availability from being measured.

Finally, we considered the possibility that F294L+A835D constitutes an intramolecular epistasis, in which the combination of the two substitutions affords significantly different fitness than the sum of their effects when considered independently. Specifically, we used the scVSV infectivity (DBT-HsACE2; Fig 6A) and HsACE2 binding (Fig 7A) datasets as proxies for viral fitness to assess the likelihood of an epistatic interaction between F294L and A835D with the linear-modeling based approach of Miton and colleagues [66]. We found that our data fit best to a first-order model in which the enhanced biological activity of the double-mutant relative to its single-mutant parents could be largely explained as the additive effect of the independent contributions of each substitution (S2 Table). We conclude that F294L and A835D independently enhance ACE2 recognition and spike-dependent entry, with their combined effect providing the double mutant a selective growth advantage over the single mutants (Fig 6B).

F294L and A835D act through distinct mechanisms

The location of F294L in the NTD, at a substantial distance from the receptor-binding motif, indicated that its effect on spike:HsACE2 binding is indirect, and as with A835D, likely mediated by a change in the RBD availability of pre-fusion spikes. To begin to uncover F294L’s mechanism of action, we investigated its effects on spike thermal stability in both the WT and A835D contexts. As described above (Fig 5A), scVSV particles bearing each spike were preincubated at different temperatures and then titrated for infectivity on DBT-RaACE2 cells (Fig 7C). Unlike with A835D, we observed little or no left shift in the temperature dependence of thermal inactivation for F294L alone, relative to WT. Moreover, the F294L+A835D double mutant resembled its A835D parent in thermal stability, sharing a large left shift in thermal inactivation, as observed in Figs 5A and 7C. Therefore, F294L does not contribute to spike destabilization, which is concordant with our observation that the F294L spike ectodomain resembles its WT counterpart in expression level and pre-fusion stability. Expression levels were compared by overlaying absorbance traces from size-exclusion chromatography (SEC) of different variants. All cultures were prepared in parallel in a similar manner, allowing for direct comparison (S7A Fig).

Finally, we examined the capacity of F294L to alter RBD availability in a temperature-dependent manner by HsACE2 capture ELISA (Fig 7D). Consistent with its behavior in the thermal inactivation studies (Fig 7C), F294L afforded an increase in HsACE2 binding that titrated over a similar temperature range as WT. Strikingly, however, the F294L+A835D double mutant exhibited a large increase in HsACE2 binding at all temperatures tested, with no discernible temperature-dependent effect between 37–55°C. We conclude that F294L enhances RBD availability via a mechanism distinct from that of A835D. Moreover, the additive effects of F294L and A835D afford the double mutant a substantial increase in RBD availability at physiological temperature and an attendant enhancement in viral infectivity relative to WT and its single-mutant counterparts.

RBD-mobilizing substitutions enhance viral sensitivity to antibody neutralization

Previous work with SARS-CoV-2 variants of concern has shown that amino acid substitutions which modulate spike dynamics can alter the exposure of neutralizing antibody epitopes [67–69]. The large increases in RBD availability mediated by A835D and F294L suggested that these substitutions might also impact viral susceptibility to antibody neutralization. To investigate this possibility, we measured the activity of ADG-2, a broad neutralizer of clade 1 sarbecoviruses, which binds an RBD epitope accessible only in the ‘up’ RBD [54,70] against VSVs bearing WT and mutant SHC014-CoV spikes. We observed little neutralization of WT at mAb concentrations as high as 500 nM. In contrast, A835D, F294L, and A835D+F294L spikes were all neutralized by ADG-2 with sensitivities proportional to their capacity to recognize ACE2 (Fig 7A) and mediate entry (A835D+F294L > A835D > F294L) (Fig 6B). Remarkably, incorporation of both substitutions into the SHC014-CoV spike afforded an enhancement in neutralization IC₅₀ of >500-fold (Fig 7E). These findings provide additional evidence that A835D and F294L enhance viral entry by increasing RBD exposure. Moreover, they suggest that substitutions of this type would be subject to a substantial fitness tradeoff in the form of enhanced sensitivity to neutralizing antibodies, with likely implications for their evolutionary trajectories during adaptation to new hosts following zoonotic transmission.

Cell lines

293FT cells (Thermo Fisher) were cultured in high-glucose Dulbecco’s Modified Eagle Medium (DMEM, Thermo Fisher) supplemented with 10% fetal bovine serum (FBS, Gemini), 1% Glutamax (Thermo Fisher), and 1% penicillin-streptomycin (P/S, Thermo Fisher). African green monkey kidney Vero Cells (ATCC) were cultured in high-glucose DMEM supplemented with 2% FBS, 1% Glutamax, and 1% P/S. Embryonic kidney fibroblast 293T cells (ATCC) were cultured in DMEM supplemented with 10% FBS, 1% Glutamax, and 1% P/S.

DBT-9 cells (gift of Ralph Baric, source unknown) [103,104] were cultured in Minimum Essential Medium α (MEMα, Thermo Fisher) supplemented with 10% FBS, 1% Glutamax, 1% P/S, and 1% Amphotericin (Thermo Fisher). Cells overexpressing ACE2 were also cultured with 5ug/mL puromycin (Gibco).

These cell lines were subcultured every 2–3 days using 0.05% Trypsin/EDTA (Gibco). All cell lines were maintained in a humidified 37°C incubator supplied with 5% CO2.

Plasmids

Plasmids encoding human codon-optimized spike (S) of SHC014-CoV (GenBank accession number KC881005.1), WIV-1-CoV (GenBank accession number KC881007), BtKY72-CoV (GenBank accession number KY352407), and RmYN02-CoV (GenBank accession number MW201982) in the VSV genome with an eGFP gene, were generated as previously described [42,43,45,46,105].

Sarbecovirus sequence alignment

Sequence alignment of amino acid sequences of the fusion peptide proximal region (FPPR) of the S protein of selected sarbecoviruses were aligned using CLUSTAL Omega [106].

Generation of recombinant VSVs

A plasmid encoding the VSV genome was modified to replace its glycoprotein, G, with the wild-type or mutant spike glycoprotein gene of SHC014-CoV, WIV-1-CoV, BtKY72-CoV, or RmYN02-CoV with a 21-amino acid C-tail truncation. The VSV genome also encodes for an eGFP reporter gene as a separate transcriptional unit. Replication-competent, recombinant VSVs (rVSVs) bearing these S proteins were generated via a plasmid-based rescue system in 293FT cells as previously described [42,43,45,46,105]. Briefly, 293FT cells were transfected with the VSV plasmid and plasmids expressing T7 polymerase and VSV N, P, M, G, and L proteins using polyethylenimine. Forty-eight hours post transfection, supernatants from the transfected cells were transferred to Vero cells. No exogenous trypsin was added to cell cultures. Rescues of SHC014-CoV S(A835D) and WIV-1-CoV S(WT) and S(A835D) were conducted at 37°C. BtKY72-CoV S(A835D) and RmYN02-CoV S(A806D) could be rescued at 32°C but not 37°C. Viral growth was monitored by an eGFP reporter every day. Viral stocks were plaque purified on Vero cells. Spike sequences were amplified from viral genomic RNA by RT-PCR and analyzed by Sanger sequencing and/or minION sequencing.

Generation of pseudotyped VSVs

Single cycle VSV pseudotypes bearing SHC014-CoV S proteins and an eGFP reporter, were produced in 293T cells as described previously [107,108]. Briefly, 293T cells were transfected with a plasmid encoding an expression vector and the SHC014-CoV S. Two days later, cells were infected with a passage stock of VSVG/ΔG for 1h at 37°C. Cells were washed eight times with high glucose media to remove any residual VSV-G. Viral supernatant was harvested two days later and pelleted by ultracentrifugation.

Generation of replication-competent delivery particles (RDPs)

Single-cycle SARS-CoV-2 RDPs bearing WT or mutant SHC014-CoV spikes were generated as described previously [48]. Briefly, BHK-21 cells were transfected with plasmids encoding the full-length WT or mutant SHC014-CoV spikes. The following day, cells were electroporated with ΔSpike nsp1- SARS-CoV-2-mNeonGreen replicon RNA and N mRNA. At 24h postelectroporation, the RDP-containing supernatants were collected and filtered through a 0.4-μM filter to remove cell debris.

SHC014-CoV spike glycoprotein ectodomain expression and purification

Amino acids 1–1191 of SHC014-CoV spike were coupled to a C–terminal foldon motif, 3C protease site, 8X histidine tag, and Twin-Strep tag and cloned into a pαH vector. For variants, substitutions F294L and A835D were incorporated into this background. All variants were expressed by polyethylenimine-induced transient transfection of FreeStyle 293-F cells (Thermo Fisher). After 6 days, the cell supernatants were harvested by centrifugation and clarified via passage through a 0.22 μm filter. Variants were purified from filtered supernatant via gravity flow over StrepTactin resin (IBA) followed by gel-filtration chromatography on a Superose 6 10/300 column (GE Healthcare) into a buffer consisting of 2 mM Tris pH 8.0, 200 mM NaCl, 0.02% NaN₃.

HsACE2 expression and purification

Amino acids 1–615 of HsACE2 were coupled to a PreScission site, 8X histidine tag, and Strep Tag II and cloned into a pαH vector. HsACE2 was expressed by polyethylenimine-induced transient transfection of FreeStyle 293-F cells (Thermo Fisher). After 6 days, the cell supernatant was harvested by centrifugation and clarified via passage through a 0.22 μm filter. HsACE2 was then purified from filtered supernatant via gravity flow over StrepTactin XT resin (IBA) followed by gel-filtration chromatography on a Superdex 200 Increase 10/300 column (GE Healthcare) into a buffer consisting of 2 mM Tris pH 8.0, 200 mM NaCl, 0.02% NaN₃.

Nanopore sequencing

Viral RNA was isolated from rescue population supernatants (Zymogen Quick-RNA viral Kit). cDNA was then generated through reverse transcription (Invitrogen Superscript III). PCR reactions were performed to amplify the cDNA population (NEB Q5 Hotstart). Following this Oxford Nanopore, MinION long read DNA sequencing was performed to identify mutations present in the viral population. Sequenced bases were called by Guppy (https://community.nanoporetech.com/docs/prepare/library_prep_protocols/Guppy-protocol/v/gpb_2003_v1_revax_14dec2018/guppy-software-overview) and the resulting sequences were assembled using the de novo assembly tool Shasta (https://github.com/chanzuckerberg/shasta/blob/master/docs/ComputationalMethods.html). The resulting assembled contigs were aligned to the reference genome utilizing Smith-Waterman local alignments.

Detection of SHC014-CoV spike surface expression

293FT cells were seeded in six-well plates. 24 hours later, cells were transfected with 2ug of expression plasmids encoding either nothing, SHC014-CoV S WT, or A835D, and 1ug of a plasmid encoding for RFP. 24 hours post transfection, cells were blocked with 0.5% bovine serum albumin (BSA, Sigma Aldrich) in PBS for 30 minutes at 4°C. SHC014-CoV S was stained by a spike-specific mAb, S309 (20ug/mL) followed by anti-human Alexa Fluor 488 (4ug/mL) for 1 hour at 4°C each. After extensive washing, stained cells were filtered through a 41-μm nylon net filter (Millipore) and analyzed using a BD FACSCalibur Flow Cytometer and FlowJo software.

Detection of SHC014-CoV spike total expression

293FT cells were seeded in six-well plates. 24 hours later, cells were transfected with 2ug of expression plasmids encoding either nothing, SHC014-CoV S WT, or A835D, and 1ug of a plasmid encoding for RFP. 24 hours post transfection, cells were collected at fixed with 4% formaldehyde at room temperature (RT) for 5 minutes. Cells were washed and permeabilized with 1X permeabilization buffer (Tonbo Biosciences) for 15 minutes at RT. Cells were blocked with 10% FBS in permeabilization buffer for 30 minutes at RT. Following washing, cells were stained with S309 (20ug/mL) followed by anti-human Alexa Fluor 488 (4ug/mL). After extensive washing, stained cells were filtered through a 41-μm nylon net filter (Millipore) and analyzed using a BD FACSCalibur Flow Cytometer and FlowJo software.

ELISA for assessment of spike incorporation into scVSV particles

High-protein binding 96-well ELISA plates (Corning) were coated overnight at 4°C with scVSV-SHC014-CoV particles bearing WT or A835D mutant spikes. Plates were blocked with 3% PBSA for 1h at 37°C and particles were subsequently probed with 100nM of S309 or anti-VSV M mAb (1:1000 dilution) for 1h at 37°C. Following washing, anti-human IgG (for S309) or anti-mouse IgG (for anti-VSV M mAb) secondary antibody conjugated to horseradish peroxidase was added for 1h at 37°C. Plates were washed, 1-Step Ultra TMB-ELISA Substrate Solution (Thermo Fisher) was added and quenched with the addition of 2 M sulfuric acid. Absorbance was read at 450 nm. The ratio of spike-to-VSV M was calculated for evaluation of spike incorporation per scVSV particle.

ELISA for detection of scVSV-SHC014-CoV S binding to soluble HsACE2

High-protein binding 96-well ELISA plates (Corning) were coated with 2ug/mL of soluble HsACE2 overnight at 4°C. Plates were then washed with PBS and blocked with PBS containing 3% BSA for 1h at 37°C. Blocked plates were exposed to scVSV-SHC014-CoV, in duplicate, at pre-titrated amounts that were diluted in a 3-fold serial dilution, for 1 hour at 37°C. Plates were washed with PBS and incubated with 10nM of ADG-2 for 1h at 37°C. Following washing, anti-human IgG secondary antibody conjugated to horseradish peroxidase was added for 1h at 37°C. Plates were washed, 1-Step Ultra TMB-ELISA Substrate Solution (Thermo Fisher) was added and quenched with the addition of 2 M sulfuric acid. Absorbance was read at 450 nm.

ELISA for detection of scVSV-SHC014-CoV S binding to antibodies

High-protein binding 96-well ELISA plates (Corning) were coated with genome normalized amounts of scVSV-SHC014-CoV particles bearing WT or A835D mutant spikes. Plates were blocked with 3% PBSA for 1h at 37°C and particles were subsequently probed with 100nM of S309, ADG-2, S2-21, SARS-CoV-2 Spike S2 rabbit polyclonal antibody (Spike S2 PAb) (https://www.sinobiological.com/antibodies/cov-spike-40590-t62) (1:5000 dilution), or anti-VSV M mAb (1:1000 dilution) for 1h at 37°C. Following washing, anti-human IgG (for S309, ADG-2, and S2-21), anti-mouse IgG (for anti-VSV M mAb), or anti-rabbit IgG (for SARS-CoV-2 Spike S2 rabbit polyclonal antibody) secondary antibody conjugated to horseradish peroxidase was added for 1h at 37°C. Plates were washed, 1-Step Ultra TMB-ELISA Substrate Solution (Thermo Fisher) was added and quenched with the addition of 2 M sulfuric acid. Absorbance was read at 450 nm.

Biolayer interferometry (BLI)

An IgG targeting T4 fibritin (foldon) was immobilized to anti-human Fc (AHC) Octet biosensors (FortéBio) to a response level of 1.0 nm. Tips were then submerged into solutions of SHC014-CoV spike variants in 20 mM Tris pH 8.0, 150 mM NaCl, 1 mg/mL bovine serum albumin, and 0.01% Tween-20, capturing variants to similar response levels. Spike-bound biosensors were subsequently dipped into 500 nM HsACEII in 20 mM Tris pH 8.0, 150 mM NaCl, 1 mg/mL bovine serum albumin, and 0.01% Tween-20 [109] to observe receptor association. Dissociation of HsACEII was then observed by submersion into a buffer consisting of 20 mM Tris pH 8.0, 150 mM NaCl, 1 mg/mL bovine serum albumin, and 0.01% Tween-20. The relative proportion of accessible RBDs was quantified as previously described [109]. Data were collected in triplicate.

Cryo-EM data collection

The SHC014-CoV S ectodomain data set was collected on a Thermo Scientific Glacios operating at 200 kV utilizing a Falcon 4 direct electron detector. C-Flat 1.2/1.3 grids (Electron Microscopy Sciences) were glow discharged at 20 mA for 30 seconds. Samples were prepared by application of 4 uL of a 1.0 mg/mL solution of spike onto grids. Grids were then blotted for 4 seconds using a blot force of -3 and a wait time of 5 seconds on a Vitrobot Mark IV (Thermo Scientific) set to 100% humidity at room temperature. The blotted grid was subsequently plunge-frozen into liquid ethane. All data were collected at 150,000x magnification, which gave a pixel size of 0.94 Å. In total, 2,025 exposures were taken using a defocus range of -1 to -2 microns, an exposure time of 13.2 seconds, and a total electron exposure of ~50 e^-/Ų. Data were collected using SerialEM.

Cryo-EM data processing

The cryo-EM data were pre-processed using cryoSPARC Live v3 [110] for motion correction and patch CTF estimation. Exposures were subsequently transferred to cryoSPARC v3 for manual curation, blob particle picking, particle extraction, and 2D classification. 1,653 exposures were accepted after curation, which resulted in 649,312 particle picks. After 2D classification, 150,716 particles remained and were used in a 4-class ab initio reconstruction job. The four output volumes were used with all particles for heterogeneous refinement. The best class contained 79,540 particles, which was used for homogenous refinement with both C1 and C3 symmetry, leading to reconstructions at 3.7 Å and 3.2 Å resolution, respectively. A total of 76,442 particles were then subjected to a non-uniform refinement with C3 symmetry, leading to a 3.1 Å resolution reconstruction.

An initial model of the WT SHC014-CoV spike was generated using AlphaFold2 [61] and fitting into the 3.1 Å cryo-EM map in ChimeraX [111,112]. The model was subsequently refined iteratively in Isolde [113], Coot [114], and Phenix [115]. Structural figures were generated in ChimeraX.

In silico binding free energy calculation between spike protomers

Optimization of the cryo-EM structure: Missing loops were modeled via various tools including Swiss Model [116], AlphaFold2 [61], Prime Homology Modeling [117], and Modeller [118]. Resultant loops were grafted onto the resolved Cryo-EM structure via in-house Visual Molecular Dynamics scripts. Uniprot and comparison to human SARS-CoV-2 spike system structure were used to identify necessary disulfide bonds (15 per monomer). Ectodomain glycosites were predicted and confirmed as described [119,120]. Our modeled construct is fully N-/O-glycosylated following a similar human glycoprofile as described [109], which was consistent with Watanabe and co-workers [121]. Protonation states for all titratable residues were assigned using PROPKA [122] at pH 7.4, and histidine protonation states were assigned with Schrodinger’s Protein Preparation Wizard [123]. The resultant SHC014-CoV glycoprotein was parameterized for minimization and short equilibration according to CHARMM36 all-atom additive force fields for proteins and glycans [124,125]. The SHC014-CoV construct was fully solvated with explicit water molecules using the TIP3P model [126] and neutralized with a 150 mM concentration of sodium and chloride ions. The system was then minimized, heated to 310 K, and briefly equilibrated using NAMD3 [127] with standard molecular dynamics simulations protocols. For a complete set of grafting, system construction, glycoprofile details, minimization, heating, equilibration scripts, and complete methodological details, please see all shared scripts and shared model files which will be provided on the AmaroLab website (https://amarolab.ucsd.edu/covid19.php).

Binding free energy calculation (interface ΔΔG): We ran two separate methods (FoldX and flex ddG) to estimate the binding affinity change between interacting protomers in the SCH014 spike trimer [128,129]. Using FoldX, we performed five rounds of internal structure optimization (FoldX repairPDB), and iteratively modeled and assessed A835D (25 iterations; FoldX BuildModel and AnalyseComplex). Flex ddG: calculation parameters were set as follows, nrstruct = 35, max_minimization_iter = 5000, abs_score_convergence_thresh = 1m, number_backrub_trials = 35000. In order to make the calculation tractable with flex ddG, we assessed the effect of A835D using a subset of residues (chain A: residues 691 to 1124; chain C: residues 18 to 689). The reported ΔΔG for each method corresponds to the averaged ΔΔG.

Virus infection assays

24 hours prior to infection, adherent cells were seeded in high glucose media on 96 well plates (Corning). Vero cells were plated at 2 x 10⁴ cells per well. DBT-9 cells were plated at 1.8 x 10⁴ cells per well. Prior to infection, virus was diluted in corresponding media and infected cells were incubated at 37°C for 12–18 hours. Cells were then fixed with 4% paraformaldehyde, stained with Hoechst-33342 (Invitrogen) before eGFP+ cells were counted using a Cytation 5 reader (BioTek Instruments). Infection levels were calculated as a percentage of eGFP+ cells over the total number of cells.

Temperature-dependent infection assays

DBT-9 cells overexpressing RaACE2 were seeded 24 hours prior to infection in high glucose media on 96 well plates at 1.8 x 10⁴ cells per well. scVSV-SHC014-CoV particles were diluted in PBS and incubated at a temperature range from 37°C to 55°C (actual range per virus indicated in Results) for 1 hour and subsequently placed on ice. After cooling, virus was added to cells and incubated at 37°C for 12–18 hours. Cells were then fixed with 4% paraformaldehyde, stained with Hoechst 33342 (Invitrogen) before eGFP+ cells were counted using a Cytation 5 reader (BioTek Instruments). Infection levels were calculated as a percentage of eGFP+ cells over the total number of cells.

Temperature-dependent binding assays

scVSV-SHC014-CoV particles were diluted in PBS and incubated at a temperature range from 37°C to 55°C (actual range per virus indicated in Results) for 1 hour and subsequently placed on ice. After cooling, virus was added onto high-binding 96-well plates precoated with soluble HsACE2. Plates were blocked with 3% BSA in PBS. SHC014-CoV S was detected with ADG-2. Bound antibody was then detected with an anti-human antibody conjugated to HRP. Plates were washed, 1-Step Ultra TMB-ELISA Substrate Solution (Thermo Fisher) was added and quenched with the addition of 2 M sulfuric acid, per manufacturer recommendations. Absorbance was read at 450 nm. All binding steps were carried out at 37°C for 1h. Binding curves were generated using Prism (GraphPad Software, La Jolla, CA).

ADG-2 neutralization assay

mAb ADG-2 was serially diluted and mixed with pre-diluted scVSV-SHC014-CoV WT, A835D, F294L, or F294L A835D in infection media (DMEM, 2% FBS, 1% P/S, 1% Q). Mixtures were incubated for 1 hr at 37°C. Virus/antibody inoculum was added to DBT-RaACE2 cells, pre-seeded in 96 well plates and incubated for 12–18 hours. Cells were fixed with 4% paraformaldehyde (Sigma), washed with PBS, and stored in PBS containing Hoechst-33342 (Invitrogen) at a dilution of 1:4000. Viral infectivity was measured by automated enumeration of GFP-positive cells from captured images using a Cytation5 automated fluorescence microscope (BioTek) and analyzed using the Gen5 data analysis software (BioTek).

Linear model for non-additive mutant effects

Infectivity data on DBT-HsACE2 cells and HsACE2 binding ELISA data for scVSVs bearing WT, F294L, A835D, or F294L+A835D SHC014-CoV spikes were analyzed according to the simplex regression method of Miton and colleagues to extract additive and non-additive mutational effects, essentially as described [66]. Log-transformed replicate infectivity and binding ELISA data were calculated for each genotype, and mutant–WT values were processed with python script ‘simplex_regression.py’ available at https://github.com/danderson8/Epistasis to determine the best-fit linear model and extract the contribution of each substitution to the phenotype of the double-mutant. See S2 Table Legend for more details.

Statistical analysis

Statistical details for each experiment, including number of replicates (n) and types of statistical tests used, are reported in respective figure legends. Statistical analysis was carried out using GraphPad Prism (V.10).