ACE2-binding exposes the SARS-CoV-2 fusion peptide to broadly neutralizing coronavirus antibodies

1The Orthocoronavirinae subfamily of coronaviruses comprises four genera: alphacoronavirus, betacoronavirus, gammacoronavirus and deltacoronavirus. Alphacoronaviruses NL63 and 229E and betacoronaviruses OC43, HKU1, SARS-CoV-2, SARS-CoV and MERS-CoV account for almost all human coronavirus (hCoV) infections, with sporadic infections attributed to CCoV-HuPn-2018 and porcine deltacoronavirus (PDCoV). The spike (S) glycoprotein, which binds to various host receptors for viral entry (table S1), is composed of the S1 and S2 subunits, is highly divergent, with only ~30% sequence identity between alpha- and betacoronaviruses and is the main target of neutralizing antibodies. Previous studies have described neutralizing monoclonal antibodies (mAbs) that cross-react amongst sarbecoviruses by targeting the receptor binding domain (RBD) or more broadly across betacoronaviruses by targeting the stem helix. However, mAbs neutralizing both alpha- and betacoronaviruses have not been reported. Broadly neutralizing mAbs, as exemplified by those targeting influenza viruses or HIV-1, can potentially be used for prophylaxis or therapy and to guide the design of vaccines eliciting broadly protective immunity.

2To search for antibodies that cross-react with human-infecting alpha- and betacoronaviruses, we stimulated under limiting conditions, total peripheral blood mononuclear cells (PBMCs) from SARS-CoV-2 immune donors, in the presence of the TLR7/8 agonist R848 and IL-2, which selectively induce the proliferation and differentiation of memory B cells. On day 12, the specificities of IgGs secreted in the culture supernatants were tested by enzyme-linked immunosorbent assay (ELISA) against a panel of recombinant S proteins from alpha and beta hCoV (Fig. 1, A to C). The number of SARS-CoV-2 IgG-positive cultures was generally higher in COVID-19 convalescent patients and in SARS-CoV-2 vaccinees with prior infection (pre-immune) as compared to vaccinees without prior infection (naïve). Most SARS-CoV-2 S IgG-positive cultures were either monospecific or cross-reactive with the closely related SARS-CoV S, while a small fraction was also reactive with OC43 and HKU1 S (Fig. 1, A to D), consistent with previous serological analyses. Six cultures (out of >4,000) from 5 individuals (out of 43) cross-reacted with all alpha and beta hCoV S proteins tested (Fig. 1, A to D), suggesting that memory B cells producing broadly reactive coronavirus antibodies might exist at very low frequency.

3To isolate broadly reactive coronavirus mAbs, we combined the screening of polyclonally-activated memory B cells with the sorting and cloning of antibody secreting cells to retrieve paired heavy and light chain sequences (fig. S1). Using this approach, we isolated 16 SARS-CoV-2-S-specific mAbs that cross-reacted with various hCoV S proteins (Fig. 1E and fig. S2A). Six mAbs (Group 1) cross-reacted with the betacoronaviruses SARS-CoV, SARS-CoV-2, OC43 and HKU1 S proteins with EC50 values ranging from 45 ng/ml to 3,000 ng/ml. Three mAbs (Group 2) cross-reacted with high affinity to all beta hCoV S proteins, with EC50 values ranging from 18 ng/ml to 40 ng/ml (Fig. 1E). These mAbs were found to target the stem helix region and were described in a separate study. The remaining seven mAbs (Group 3) exhibited the broadest cross-reactivity to both alpha and beta hCoV S proteins with EC50 values ranging from 29 ng/ml to 800 ng/ml (Fig. 1E). These broadly reactive mAbs, which are the focus of the present study, were isolated from convalescent or vaccinated individuals, use different V genes (except for C13B8 and C13A7 that are clonally related) and displayed a high load of somatic mutations (7-14% in VH, 2-8% in VL at the nucleotide level; table S2). These results illustrate the utility of a simple high-throughput method based on multiple parallel screening steps of memory B cells to isolate broadly reactive coronavirus mAbs.

4Using SARS-CoV-2 S1, RBD and S2 proteins, as well as 15-mer linear peptides covering the entire S sequence, the specificities of all seven Group 3 mAbs were mapped to the K811PSKRSFIEDLLFNK825 sequence in the SARS-CoV-2 S2 subunit (Fig. 2A). This sequence spans the S2’ cleavage site (R815) and the fusion peptide N-terminal region, which is essential for membrane fusion and is highly conserved among all genera of the Orthocoronavirinae subfamily, including deltacoronavirus PDCoV and gammacoronavirus IBV, as well as all SARS-CoV-2 variants sequenced to date (fig. S3, A to C). Most of the seven Group 3 mAbs bound more tightly to the prefusion SARS-CoV-2 S trimer without stabilizing proline substitution in the fusion peptide (PentaPro: F817) relative to HexaPro (F817P) and all of them bound with greater avidity to prefusion SARS-CoV-2 S2 compared to postfusion S2 (Fig. 2A and fig. S2D).

5To explore the ontogeny of these fusion peptide-specific mAbs, we compared the binding properties of mature mAbs to their unmutated common ancestors (UCAs) (Fig. 2B and fig. S2B). The UCAs of C13B8, C13A7 (the two clonally related mAbs) and VN01H1 exhibited broad reactivity with hCoV S proteins but had lower affinity than their mature counterparts. In contrast, the C13C9 UCA bound only to the beta hCoV OC43 and MERS-CoV S, whereas the VP12E7 UCA bound only to the alpha hCoV NL63 S, with low affinity in both cases. Finally, the UCAs of C28F8 and C77G12 did not bind to any hCoV S proteins tested. Given the lack of a common V gene usage, these findings suggest that broadly reactive fusion peptide-specific mAbs can mature through multiple pathways and acquire high affinity and cross-reactivity through somatic mutations, possibly because of priming by endemic coronavirus infection followed by SARS-CoV-2 boosting upon infection or vaccination.

6Considering the high conservation of the fusion peptide region, we next assessed if broad reactivity is a property shared by most fusion peptide-specific mAbs. IgGs secreted upon polyclonal activation of memory B cells from 71 convalescent individuals were screened for binding to alpha and beta hCoV fusion peptide pool as well as to SARS-CoV-2 S protein. Cultures producing fusion peptide-specific mAbs were detected at low frequency and only in 19 individuals (fig. S4A). Although nearly all fusion peptide-reactive mAbs bound to SARS-CoV-2 S, only 9 out of 30 were broadly reactive and the remaining mAbs showed different degrees of cross-reactivity (fig. S4B). Thus, although the coronavirus S fusion peptide is conserved, broad reactivity is the property of a minority of fusion peptide-specific mAbs, consistent with the observation that only three out of the seven Group 3 mAbs cross-reacted with the PDCoV and IBV fusion peptides (fig. S3C).

7We next tested the neutralizing activity of all seven Group 3 mAbs against pseudotyped viruses carrying S proteins of alpha- and betacoronaviruses. Despite binding to the same motif, these mAbs exhibited distinct neutralizing potencies. Most notably, VN01H1 and VP12E7 neutralized all hCoV S pseudotyped viruses tested (SARS-CoV-2, SARS-CoV, MERS-CoV, NL63 and 229E), as well as bat sarbecovirus WIV-1 S pseudotyped viruses (Fig. 2C and fig. S5, A to C). In addition, VN01H1 also inhibited bat merbecovirus PDF-2180 S-mediated entry into cells (fig. S5B). C77G12 neutralized SARS-CoV-2 with the highest relative potency and all betacoronavirus S pseudotype viruses tested (Fig. 2C and fig. S5, A to C). SARS-CoV-2 neutralizing activity of VN01H1 and C77G12 was reduced by the addition of soluble fusion peptide (fig. S5D). Furthermore, VN01H1, C77G12 and C28F8 also inhibited SARS-CoV-2 S-mediated cell-cell fusion (Fig. 2D), suggesting that fusion peptide-specific mAbs could prevent S proteolytic activation or fusogenic rearrangements, thereby inhibiting membrane fusion and viral entry.

8We next assessed the protective efficacy of the VN01H1 and C77G12 mAbs in vivo. In the Syrian hamster model of SARS-CoV-2 P.1 (Gamma) infection, prophylactic administration of either mAb at high doses reduced viral RNA copies and lung titers, and ameliorated lung pathology at statistically significant levels (Fig. 2E). These findings demonstrate that fusion peptide-specific mAbs can reduce viral burden in vivo, albeit with moderate potency.

9To gain insight into the epitope recognized by fusion peptide-specific broadly reactive coronavirus mAbs, we performed substitution scan analysis on the six clonally unrelated mAbs and structural analysis. All mAbs bound to a core motif comprising I818EDLLFNK825 (fig. S6A), with C28F8, C77G12, VN01H1, and VP12E7 having a 3-amino acid expanded footprint spanning the N-terminal R815SF817 residues which includes the S2’ cleavage site R815.

10We determined the crystal structures of five Fabs (C13B8, C13C9, C77G12, VN01H1 and VP12E7) in complex with the K811PSKRSFIEDLLFNK825 fusion peptide at 2.1 Å, 2.1 Å, 1.7 Å, 1.86 Å and 2.5 Å, resolution, respectively (Fig. 3; fig. S6, B to D; and table S3). All five Fabs bind to overlapping epitopes in the fusion peptide through interactions involving the heavy and light chains. The strict conservation and the conservative substitution of key residues involved in mAb recognition (R815, S816, I818, E819, D820, L821, L822, F823, N824 and K825) across the Orthocoronavirinae subfamily explains the broad cross-reactivity of these fusion peptide-specific mAbs (figs. S3 and S6 and table S4). The overall architecture of the fusion peptide in the C13C9-bound complex structure is most similar to that observed in the prefusion S trimer (PDB 6VXX) (Fig. 3C and fig. S6B). In the two structures determined in complex with VN01H1 or VP12E7, the SARS-CoV-2 S fusion peptide adopts a similar conformation, which is distinct from the conformation observed in prefusion SARS-CoV-2 S trimeric structures (Fig. 3, B and C, and fig. S6C). Specifically, residues 813SKR815 refold from an extended conformation in prefusion S to an ɑ-helical conformation in the two Fab-bound peptide structures, thereby extending the ɑ-helix found at the N-terminal region of the fusion peptide. In the C77G12-bound complex structure, the fusion peptide residues P812-R815 adopt an extended conformation, distinct from prefusion S, (Fig. 3, A and C), whereas these residues are disordered in the C13B8-bound structure (Fig. 3C and fig. S6D). The conserved residue R815, which is the S2’ site of proteolytic processing upon receptor binding for membrane fusion activation, is engaged in electrostatic interactions with the C13C9, VN01H1, VP12E7 and C77G12 Fabs and therefore buried at the interface with their paratopes (table S4). Since pre-incubation of a soluble native-like SARS-CoV-2 S ectodomain trimer with fusion peptide-specific mAbs did not prevent S2X58-induced triggering of fusogenic conformational changes (fig. S7), these mAbs likely inhibit TMPRSS2 cleavage of the S2’ site (via steric hindrance) and in turn activation of membrane fusion. Although residue F817 is not part of the epitope (table S4), the F817P substitution present in the HexaPro construct likely prevents the adoption of the extended ɑ-helical conformation observed in the structures bound to VN01H1 or VP12E7 (due to restricted backbone torsion angles), resulting in dampened binding in ELISA (Fig. 2A and fig. S2D).

11Superimposition of the Fab-fusion peptide complexes with available prefusion S structure (PDB 6VXX) revealed that the targeted epitope is buried toward the core of the S trimer and is therefore inaccessible (Fig. 3, D and E, and fig. S6, B to D), likely explaining the lack of detectable complexes of Fabs with prefusion S trimers during single particle electron microscopy analysis. Taken together, these findings suggest that the epitope recognized by fusion peptide-specific mAbs is cryptic and may become accessible only transiently.

12To investigate the neutralization mechanism of fusion peptide-specific mAbs, we transfected HEK293T cells to express coronavirus S proteins on their cell surface. All fusion peptide-specific mAbs showed only marginal binding to SARS-CoV-2 S-expressing HEK293T cells, as compared to control mAbs targeting the RBD (C94) (fig. S2C) or the stem helix (C21E3) (Fig. 4A and fig. S8A). However, addition of soluble ACE2 enhanced binding of all fusion peptide mAbs to SARS-CoV-2 S to levels comparable to that of the C94 and C21E3 control mAbs, suggesting that receptor engagement induces a conformational change that exposes the cryptic fusion peptide epitope (Fig. 4, A and B, and fig. S8A). This ACE2-dependent enhancement of binding was also observed with SARS-CoV-2 S harboring a deletion in the polybasic S1/S2 cleavage site (ΔPRRA) (fig. S8A), suggesting that these mAbs can bind prior to S1 shedding or S fusogenic refolding to postfusion state, consistent with the preferential binding to prefusion S2 relative to postfusion S2 (fig. S2A). However, the 2P (K986P, V987P) prefusion-stabilizing mutations that lay outside of the epitope (Fig. 4A and fig. S8A), inhibited ACE2-mediated enhancement of mAb binding, implying an impediment of receptor-induced allosteric conformational changes, in line with recent findings. Enhanced binding of fusion peptide-specific mAbs was also observed with SARS-CoV S- and MERS-CoV S-expressing HEK293T cells in the presence of ACE2 and DPP4. In contrast, these mAbs bound efficiently to HEK293T cells displaying the alphacoronaviruses NL63 and 229E S, independently of receptor engagement by ACE2 or APN, respectively (Fig. 4A and fig. S8A). We observed that VN01H1 and C77G12 neutralized authentic SARS-CoV-2 Omicron BA.1 and BA.2 more potently than Wuhan-Hu-1, likely due to increased accessibility of the fusion peptide in Omicron S. Furthermore, the neutralization potencies of VN01H1 and C77G12 could be further improved by engineering mAbs with smaller formats (such as scFv), suggesting possible limited accessibility of the fusion peptide region (Fig. 4C).

13It was previously shown that certain mAbs can mimic ACE2 binding and trigger fusogenic activity. Indeed, the addition of S2E12, an ACE2-mimicking mAb, but not S2M11, a mAb which locks SARS-CoV-2 S trimer in the closed state enhanced binding of the fusion peptide mAbs (Fig. 4B). Consistent with this finding, we observed a synergy in pseudovirus neutralization when C77G12 was used in combination with S2E12 (Fig. 4D), suggesting that fusion-peptide mAbs may be more effective in the context of a polyclonal antibody response against the RBD. Collectively, these results indicate that the fusion peptide epitope only becomes accessible upon receptor-induced S conformational changes in sarbecovirus SARS-CoV-2, SARS-CoV and merbecovirus MERS-CoV, whereas it is more readily accessible in the alphacoronaviruses 229E and NL63, possibly due to molecular breathing to different extents in these S trimers.

14Using a high-throughput screening of memory B cells from immune donors with a panel of S proteins, we identified a new class of mAbs that target the fusion peptide region. These mAbs have different binding modalities, some of which exhibit unprecedented cross-reactivity to the fusion peptide of all four coronavirus genera. The finding that the UCAs of some of these mAbs bind preferentially to common cold coronaviruses and that the mAbs acquire affinity and breadth through somatic mutations suggest that their elicitation may require multiple rounds of selection, possibly by heterologous coronavirus infections. A complex developmental pathway has been reported for HIV-1 neutralizing mAbs and may be a general requirement for mAbs that recognize highly conserved epitopes.

15Previous studies have identified serum antibodies binding to the fusion peptide of SARS-CoV-2 and showed, through depletion or peptide inhibition experiments, that such antibodies can contribute to the serum neutralizing activity in a polyclonal setting. We show here that some fusion peptide-specific mAbs have direct neutralizing activity in vitro against alpha- and beta-coronaviruses and can ameliorate pathology and viral burden in vivo at high doses. Although the neutralizing activity of these mAbs is low when used alone, it is possible that in the context of a polyclonal response they may synergize with other antibodies that favor the exposure of the fusion peptide region, as shown here with S2E12.

16The biochemical events associated with S2’ proteolytic processing for SARS-CoV, SARS-CoV-2 and MERS-CoV and that are critical for unmasking the fusion peptide region have been described previously. The mAbs isolated in this study provide tools to characterize the S allosteric conformational changes triggered by receptor binding or by receptor-mimicking mAbs, a process that has thus far eluded structural definition. These events lead to exposure of the fusion peptide region for proteolytic cleavage by TMPRSS2 or endosomal cathepsins preceding membrane fusion. The finding that this region is readily accessible in alphacoronaviruses is consistent with the higher neutralizing activity of some fusion peptide-specific mAbs against NL63 and 229E and proposes that such conformational change might occur spontaneously as a consequence of molecular breathing.

17Broadly neutralizing mAbs recognizing the fusion peptide region of influenza virus hemagglutinin and HIV-1 gp120 have guided the design of universal vaccine candidates against these highly variable pathogens. Similarly, the conservation of the fusion peptide region across the Orthocoronavirinae subfamily provides a rationale to develop a vaccine capable of inducing fusion peptide-specific responses that may be less prone to viral escape due to the potential fitness loss of escape mutants. The finding that prefusion stabilizing 2P mutations abolishes receptor-induced exposure of the fusion peptide suggests that current mRNA vaccines may be suboptimal in inducing antibody responses against this region. Broadly neutralizing coronavirus mAbs described in this study may be used as probes to design immunogens that can better unmask the fusion peptide region and elicit broadly protective antibody responses. Interestingly, the fusion peptide region was also found to stimulate broadly reactive CD4 T cells, providing a cue for intramolecular help in the generation of such antibodies. Collectively, our study identifies neutralizing mAbs with unprecedented breadth and sheds light on a receptor-triggered conformational change that limits the immunogenicity of this conserved region, with potential impact on vaccine design.