key: cord-0682745-meg9f20w
authors: Burke, Matthew J.; Scott, James N.F.; Minshull, Thomas; Stockley, Peter G.; Calabrese, Antonio N.; Boyes, Joan
title: An Ultralong Bovine CDRH3 that Targets a Conserved, Cryptic Epitope on SARS-CoV and SARS-CoV-2
date: 2022-04-06
journal: bioRxiv
DOI: 10.1101/2022.04.06.487306
sha: 5707b83e188de9ff5306efef80619a44c32607d0
doc_id: 682745
cord_uid: meg9f20w

The ability of broadly neutralising antibodies to target conserved epitopes gives them huge potential as antibody-based therapeutics, particularly in the face of constant viral antigen evolution. Certain bovine antibodies are highly adept at binding conserved, glycosylated epitopes, courtesy of their ultralong complementarity determining region (CDR)H3. Here, we used a SARS-naïve, bovine ultralong CDRH3 library and mammalian cell display, to isolate a bovine paratope that engages the SARS-CoV and SARS-CoV-2 receptor-binding domain (RBD). This neutralises viruses pseudo-typed with SARS-CoV Spike protein but not by competition with RBD binding to ACE2. Instead, using differential hydrogen-deuterium exchange mass spectrometry and site-directed mutagenesis, we demonstrate that this ultralong CDRH3 recognises a rarely identified, conserved, cryptic epitope that overlaps the target of pan-sarbecovirus antibodies (7D6/6D6). The epitope is glycan-shielded and becomes accessible only transiently via inter-domain movements. This represents the first bovine anti-sarbecovirus paratope and highlights the power of this approach in identifying novel tools to combat emerging pathogens.

defined role in generating and stabilising its structure [25, 26, 29] and which lends itself well to the specific isolation of ultralong CDRH3 sequences (Fig. 1a) .

Here, we sought to isolate ultralong bovine heavy chains with neutralising activity against SARS-CoV-2 and related coronaviruses. We capitalised on previous findings that these pair with a relatively invariable Vλ light chain [25, 29] , to generate a single chain variable fragment (scFv) scaffold into which ultralong heavy chain-only libraries can be cloned and expressed.

Using mammalian cell surface display and His-tagged S2P SARS-CoV-2 Spike glycoprotein, we isolated an ultralong scFv (B9-scFv) from a SARS-CoV-2-naïve heavy-chain library that binds to SARS-CoV-2 RBD, all current SARS-CoV-2 variants and, with a ≈50x stronger interaction, to SARS-CoV RBD. B9-scFv does not compete with ACE2 binding but neutralises SARS-CoV pseudo-typed lentiviruses with an IC50 of 468 nM, likely by destabilising the prefusion complex. Consistent with this, the epitope localises to a cryptic cleft on the inner face of the RBD that becomes available only transiently by inter-domain movements. This epitope has been identified only once previously as the target for two other pan-sarbecovirus antibodies [30] when multiple, repeated immunisations were required. Remarkably, here we isolate a bovine broadly reactive CDRH3 from a library of <1 x 10 4 sequences. This attests to the huge potential of the bovine system as a source of broadly reactive antibodies that can protect against emerging pathogens and their variants.

To establish a cell-surface display platform for the isolation of ultralong CDRH3s, we generated a library of ultralong bovine paratopes by amplifying variable exons from the leukocyte genomic DNA (gDNA) of two adult cows. Ultralong CDRH3s were enriched by nested PCR and gel excision, resulting in an initial heavy chain library with >96% ultralong CDRH3s ( Supplementary Fig. 1a) . The purified amplicons were inserted into the pBovShow cassette ( Fig. 1b) that resulted in them being joined to an invariant Vλ light chain (Vλ-LC) via a flexible linker. Ultralong scFvs were expressed following transient transfection into 293T cells, as determined by flow cytometry (Fig. 1c) . Notably, cell surface expression of over 70% of scFv clones was achieved ( Supplementary Fig. 1b) , suggesting the invariant Vλ light chain pairs with most ultralong heavy chains, even in an scFv format.

Next, we screened our ultralong scFv library for binding to the recombinant full-length SARS-CoV-2 S-2P glycoprotein [31] by transient transfection into 293Ts and subsequent flow cytometry. Initially, a small population of cells expressing scFvs that bind Spike protein were isolated (0.02%; Fig. 1c ); these were enriched by two further rounds of plasmid recovery, retransfection and flow cytometry (Fig. 1d) , resulting in a sharp increase in the proportion of Spike-binding scFvs (Fig. 1d) . Sequencing the scFv library at each stage of enrichment allowed the original library diversity to be estimated at <1 x 10 4 unique sequences ( Supplementary Fig. 1c) . Given this low level of initial diversity, it is remarkable that scFvs that bind SARS-CoV-2 Spike protein are present.

To isolate these Spike-binding scFv(s), we cloned our enriched library into lentivirus (LV) vectors and generated LV particles pseudo-typed with VSV-G. These were transduced into 293T cells at a low titre to achieve few scFv sequences per cell. After puromycin selection and expansion, only 0.42% of the transduced cells bound Spike protein ( Supplementary Fig. 2a,   b) . From this population, we isolated 15 single cell clones (SCCs) that interacted with 40 nM Spike. Three of these SCCs harboured a single scFv and remarkably, the sequence of all three scFvs was identical at the nucleotide level. This sequence, termed B9-scFv (Fig. 1e) , encodes a 53 amino acid ultralong CDRH3 that interacts with the SARS-CoV-2 Spike ( Supplementary Fig. 2b, lower) . B9-scFv accounted for 53% of all scFvs from the LVtransduced cells after a single selection for Spike binding, and this increased to 83% upon a further round of enrichment by flow cytometry. These data therefore suggest that the ultralong B9-scFv likely accounts for much the anti-Spike activity in our library.

To further characterise B9-scFv, we purified the SARS-CoV-2 Spike sub-domains (Fig. 2a) , S1 (aa 2-682), S2 (aa 686-1211), and RBD-SD1 (aa 319-591) via IMAC ( Supplementary Fig.   3a ). Cells transiently expressing B9-scFv display clear binding to the S1 domain and the RBD, but not to equivalent concentrations of the S2 domain (Fig. 2b) . These data therefore localise the B9-scFv epitope to RBD residues 319-591.

Although B9-scFv readily binds 40 nM Spike, markedly higher levels of RBD (≈2 μM) are necessary to achieve a similar effect ( Supplementary Fig. 3b ). Similar disparities in apparent affinity have been reported previously for the interactions of engineered ACE2-Fc constructs with RBD-monomers and full-length trimeric Spike [32] . These were attributed to differences in avidity due to the increased number of binding sites provided by the trimeric Spike compared to the RBD. While these data suggest B9-scFv has an epitope within SARS-CoV-2 RBD, they also suggest its binding may be relatively weak.

Bovine ultralong CDRH3s typically recognise conserved epitopes. If this is the case for B9-scFv, we would expect its binding will be unaffected by mutations in any of the SARS-CoV-2 variants of concern (VOC). To investigate this, the B9-scFv sequence was cloned upstream of a poly-His tag, expressed in 293Ts and purified by IMAC ( Supplementary Fig. 4a ). 293T cells were then transiently transfected with an expression vector for SARS-CoV-2 Spike (Wuhan-Hu-1 + D614G) and cell-surface Spike expression was confirmed by staining with a positive control scFv (CR3022-scFv), which binds to a conserved RBD epitope between SARS-CoV-2 and SARS-CoV [33] (Supplementary Fig. 4b , upper). Titration of purified B9-scFv gives concentration-dependent binding to Spike transfected cells, that is enriched compared to a negative control bovine ultralong scFv (137-scFv; Supplementary Fig. 4b ). We next examined B9-scFv binding to SARS-CoV-2 Spike protein variants; crucially, binding is maintained to all of the mutations in the commonly circulating SARS-CoV-2 variants, including D614G, N501Y, E484K, Y453F, L452R and K417N from the Alpha, Beta, Gamma, Delta and Omicron variants (Fig. 3a, b) . Not only does this imply B9-scFv recognises SARS-CoV-2 Spike glycoprotein in its native state, but also suggests B9-scFv may be broadly reactive.

To further confirm B9-scFv binding to the SARS-CoV-2 VOC, we purified SARS-CoV-2 RBDs (aa 319-591) harbouring various mutations and incubated these with 293T cells expressing B9-scFv on the cell surface. Consistent with the data in Fig. 3b , B9-scFv bound to RBDs carrying the mutations associated with the Alpha lineage as well as to the RBD of the hypermutated Omicron variant, with no significant loss of affinity compared to wild type Spike ( Fig. 3c) . Notably, these RBD mutations found in Omicron are also found in the Beta and Gamma variants. These data therefore add weight to the idea that B9-scFv binds a conserved epitope.

Given its potential broad reactivity, we next sought to determine if B9-scFv binds to other coronvirus RBDs. The corresponding sequences of SARS-CoV RBD (aa 319-591) and MERS-CoV RBD (aa 368-586) were cloned, and the proteins purified from 293T cells ( Supplementary   Fig. 3c) . Surprisingly, the interaction between cell surface expressed B9-scFv and SARS-CoV RBD was markedly stronger than that observed with equivalent amounts of SARS-CoV-2 RBD. Maximal detectable binding by FACS was observed at 2 μM SARS-CoV RBD and this was only partially reduced at 200 nM. In contrast, the interaction with SARS CoV-2 RBD was modest at 2 μM RBD and scarcely detectable at 200 nM (Fig. 3d ). Both B9-scFv and the SARS-specific human CR3022-scFv were relatively unreactive with the MERS-CoV RBD at all concentrations tested (Fig. 3d) , suggesting B9-scFv has specificity for SARS-CoVs.

The small difference in binding of B9-scFv to 2 μM and 200 nM SARS-CoV RBD indicates a nanomolar affinity. To further investigate this, binding of cell surface expressed B9-scFv to a range of RBD concentrations (10 nM -4 μM) was measured (Fig. 3e) , allowing an approximate KD of 139 nM to be calculated. Using this same method, we approximated the KD for CR3022-scFv interactions with the SARS-CoV (0.7 nM) and SARS-CoV-2 (33 nM) RBDs, values that are comparable to those reported in the literature [34] . The binding of B9-scFv to SARS-CoV-2 RBD increased almost linearly up to the maximum analysed concentration of 5 μM RBD and therefore the affinity of the interaction could not be estimated (Fig. 3e) .

We next tested if purified B9-scFv recognises the SARS-CoV RBD in the context of the Spike trimer. Consistent with our previous results, B9-scFv binds more strongly to cells expressing the SARS-CoV Spike than the SARS-CoV-2 Spike, as much lower concentrations of B9-scFv (<200 nM) are needed to label cells expressing this glycoprotein (Fig. 3f) . Collectively, these data suggest that B9-scFv cross-reacts with viruses in the Sarbecovirus subgenus and binds with much higher affinity to SARS-CoV compared to SARS-CoV-2

Given that B9-scFv appears to be broadly reactive, based on its resistance to all current SARS-CoV-2 RBM mutations and its cross-reactivity with the SARS-CoV RBD (Fig. 3b , c, d, e, f), we next sought to determine if it also neutralises virus infectivity. We therefore capitalised on the higher affinity of B9-scFv for SARS-CoV Spike to ask whether B9-scFv neutralises pseudo-typed lentivirus particles. As can be seen in Figure 4a , B9-scFv fully neutralises lentiviral particles pseudo-typed with the SARS-CoV Spike (Urbani variant) when tested at 70 μg/ml but has no consistent effect on an equivalent titre of SARS-CoV-2 (Wu-1-D614G) pseudo-typed virus, correlating with previously observed differences in estimated affinity (Fig.   3d ). Importantly, B9-scFv does not reduce the infectivity of control VSV-G pseudo-typed lentivirus at 70 μg/ml (Fig. 4a) whereas titration experiments demonstrate that it neutralises SARS-CoV pseudo-typed lentiviruses with an IC50 of 468 nM (Fig. 4b) .

In complementary experiments, we used B9-scFv in a competition-binding assay to test if it prevents purified SARS-CoV RBD from binding to hACE2-expressing cells. Although there is significant competition between soluble hACE2-Fc (300 nM) and cell surface ACE2 for RBD binding (Fig. 4c) , high concentrations of B9-scFv (5 μM) or CR3022-scFv (5 μM) did not prevent SARS-CoV RBD from binding to cell surface ACE2. The distinct lack of competition in this context suggests that the mechanism of B9-scFv neutralisation is unlikely to directly involve interference with ACE2 binding.

Given that the Spike proteins of SARS-CoV and SARS-CoV-2 are 76% identical [35], we took advantage of the higher affinity of B9-scFv for SARS-CoV RBD to better localise its epitope.

To this end, we performed differential hydrogen-deuterium exchange mass spectrometry (HDX-MS) of SARS-CoV RBD in the absence (SARS-CoV RBD only) or presence of B9-scFv (SARS-CoV RBD + B9-scFv). A 2-minute exposure to deuterium revealed two main regions of the SARS-CoV RBD with significantly reduced deuterium uptake in the presence of B9-scFv ( Fig. 5a and Supplementary Fig. 5a ). Protected region 1 includes three peptides involving RBD residues 449-467 and spans from β7 near the ACE2 interacting region, through the β7-β8 loop to an inner face of the RBD ( Supplementary Fig. 5b ). Region 1 can be further resolved to just residues 456-467 due to the identification of a shorter protected peptide ( Supplementary   Fig. 5a ). In contrast, region 2 comprises only a single peptide of residues 551-565 within subdomain 1 and appears to be markedly less protected than region 1 (Fig. 5a) . Given that residues 542-591 can be removed with only a small reduction in B9 binding to SARS-CoV RBD ( Supplementary Fig. 5c ), Region 2 does not appear to be the main epitope. We therefore focused on region 1.

Region 1 contains a candidate motif, 463PFERD467, that is fully conserved between SARS-CoV-2 and SARS-CoV and may explain the cross-reactivity of B9-scFv (Fig. 5b) . To further investigate this, residues surrounding this motif were mutated in full-length SARS-CoV-2 Spike to their equivalents in SARS-CoV ( Supplementary Fig. 5d) , with the hypothesis that only mutations proximal to the bona fide epitope will improve binding. The various SARS-CoV-2 mutants were transfected into 293T cells and subsequently incubated with a concentration of B9-scFv (1 μM) at which binding to SARS-CoV-2 Spike is minimal. Three mutations strengthen the interaction with B9-scFv, namely A348P, N354E and Mut455-462 (Fig. 5c, d) . Of these, Mut455-462 has the largest effect and is located adjacent to the 463PFERD467 motif in both the primary amino acid sequence and tertiary structure of the RBD (Fig. 5c, d, e) . Similarly, although the A348P and N354E mutations are distal from 463PFERD467 in the primary amino acid sequence, they are in very close proximity in the tertiary structure (Fig. 5e , Supplementary   Fig. 5b ) and are situated on the opposite side of 463PFERD467 relative to Mut455-462 (Fig. 5e ).

In contrast, mutations located away from 463PFERD467 on the globular RBD, such as P384A, Mut402-406, Mut438-447, I434L and L452K, had no effect on B9-scFv binding to the mutated SARS-CoV-2 Spike (Fig. 5c , d, e). When combined, the A348P, N354E and Mut455-462 mutations increased binding of B9-scFv to SARS-CoV-2 to levels comparable with the positive control (CR3022-scFv; Fig. 5c ). The region mapped by these experiments therefore corresponds to the smaller HDX protected peptide of residues 456-467 in region 1 ( Supplementary Fig. 5a ), supporting the idea that this is the epitope recognised by B9-scFv.

Interestingly, in the context of a 1-up, 2-down conformation SARS-CoV Spike, the epitope for B9-scFv is relatively inaccessible on all protomers (Fig. 6a) . This site has, however, previously been proposed to be transiently exposed by interdomain movements [30] . Notably, a glycan from the NTD of the adjacent protomer (N165) is thought to block access to this region by inserting itself in the volume left by the RBD when it is in the "up" conformation [36] . This glycan forms the remaining contact between NTDB and RBDA to stabilise RBD-up via a "lock and load" mechanism ( [36] ; Fig 6b) . Transient domain movements that allow B9-scFv binding would break this glycan contact, potentially destabilising the Spike complex. Likewise, two previous neutralising antibodies that target this site cause destabilisation of the prefusion Spike complex [30] and shedding of the SARS-CoV-2 S1 domain.

Notably, B9-scFv engages this region with knob domain only binding, as disruption of the ultralong knob domain by mutagenesis abolishes the interaction ( Supplementary Fig. 5e ).

Taken together, these data provide strong evidence that B9-scFv binds at a site largely made up of conserved residues in the β7-β8 loop and causes destabilisation of the Spike trimer. The increased strength of its interaction with SARS-CoV can largely be accounted for by a handful of residues that vary between SARS-CoV-2 (Wu-1-D614G) and SARS-CoV (Urbani).

The increasing frequency of zoonotic transfers, together with the possibility of rapid viral evolution, highlights the urgent need for broadly active therapeutics to fight emerging pathogens. Using our novel scFv display system, which specifically expresses bovine ultralong heavy chains in mammalian cells, we have isolated a bovine ultralong scFv (B9) that engages the receptor binding domain (RBD) of the SARS-CoV-2 and SARS-CoV Spikes at a conserved but cryptic site. Only two neutralising antibodies targeting this site have previously been isolated (7D6/6D6) and they were shown to induce destabilisation of the prefusion Spike complex [30] . Consistent with this, B9-scFv neutralises SARS-CoV pseudo-typed lentiviral particles (IC50 468 nM) but not by competing with RBD binding to ACE2. Instead, it likely causes destabilisation of the prefusion Spike complex in part by interfering with a critical glycan contact. It is notable that isolation of the 7D6/6D6 antibodies required five immunizations of mice with either SARS-CoV-2 S-2P or a combination of SARS-CoV-2 Spike, SARS-CoV Spike and MERS-RBD; by contrast, our present study isolated B9-scFv from a small library of bovine CDRH3s. In fact, the identification of an ultralong CDRH3 targeting a seemingly rare neutralising epitope is made even more remarkable by the limited size of our initial library (<1 x 10 4 unique sequences). Overall, this study clearly demonstrates how our novel approach can be used to rapidly isolate ultralong CDRH3s that target conserved sites of vulnerability on viral antigens.

Bovine ultralong antibodies possess the longest known CDRH3 regions. It is likely that the extended β-stranded stalk helps to punch through glycan coats on some viral antigens to reach occluded, functionally conserved epitopes, while it has been shown that the disulphide-bonded loops can engage a target using a compact surface area. Together, these structural features increase the resistance of ultralong neutralising antibodies to viral escape mutations [21] . In keeping with previously studied ultralong Abs, B9 engages its epitope with knob domain-only binding since disruption of the knob by mutation completely abrogates its interaction with SARS-CoV RBD ( Supplementary Fig. 5e ). Unusually, however, B9-scFv has a truncated ascending and descending β-stranded stalk [25, 26] , with fewer residues at the VD junction (101RD102) and no alternating tyrosine motif at the 3' end of D8-2. These features will likely impact the length, angle, and flexibility of the stalk and may influence how the disulphidebonded loops of the B9 knob domain engage the RBD.

Notably, when the epitope for B9-scFv was mapped onto a model of a trimeric SARS-CoV Spike in the 1-up, 2-down RBD conformation, it appears relatively inaccessible in every context. Our HDX and mutagenesis data show B9-scFv's epitope shares several key contacts with the previously identified cross-neutralising antibodies 7D6 and 6D6, including N/E354 on β1 and residues 457-467 on the β7-β8 loop of the RBD. This implies that our scFv is also likely to clash with the N-terminal domain (NTD) from the adjacent protomer in the context of the trimeric Spike (Fig. 6c) . The adjacent NTD also contains a crucial aminoglycan moiety at N165.

While this has been shown to be critical for gating of the RBD opening, it may also act to shield this conserved region from antibody recognition (Fig. 6b) [30, 36] . Indeed, previous models suggest this cryptic epitope is only made accessible by transient movements of the RBD and NTD, and subsequent antibody binding acts to destabilise the trimer. This ultimately leads to neutralisation by the destruction of the pre-fusion SARS-CoV and SARS-CoV-2 Spikes and shedding of the S1 domain in the latter [30] . Due to its significant overlap with the footprint of 7D6, B9-scFv should likely be considered a 7D6-like RBD-targeting scFv [30] . Overall, B9-scFv binding to this cryptic epitope correlates with its relative cross-reactivity, lack of competition with ACE2 and resistance to RBM mutations found in SARS-CoV-2 VOC.

Despite its modest initial affinity for SARS-CoV-2, it may be possible to improve the binding of an isolated bovine scFv, either by directing AID-mediated mutagenesis, using CRISPR-x [37] , or through other diversification methods, such as error-prone PCR or structure-guided affinity maturation [38] . Indeed, all mutagenesis efforts could be entirely focussed on the region encoding the knob domain. It is also probable that with larger initial libraries, binders against more diverse, and cross-reactive epitopes will be isolated. At least some of these would be expected to have a high initial affinity and require fewer mutations, if any, to achieve strong target binding. Previous efforts to select and engineer human antibodies for increased breadth and potency against SARS-like viruses have been remarkably successful [14] . It should be noted, however, that bovine ultralong CDRH3s are unlikely to be amenable to typical oligonucleotide-based CDR diversification methods, due to the length and unique structural requirements of the ultralong CDRH3 [25, 26] . To ensure that improvements in an scFvs potency are not to the detriment of its breadth, any mutagenesis and selection efforts would need to use multiple related antigens for screening.

Clearly, identifying a novel paratope is just the first step in developing a treatment against a new pathogen, as a non-immunogenic scaffold with which to deliver the neutralising agent is crucial. Fortunately, various possibilities exist. Firstly, bovine paratopes have been successfully transferred to a human antibody scaffold with minimal loss of activity [20] , although optimisation of a scaffold may be required to achieve stability and good manufacturability [23] . Other options include scFv-Fc fusions [39] or PEGylated scFvs [40] , both of which have significantly improved pharmacokinetic profiles over standard scFvs.

Nonetheless, a risk remains that the bovine ultralong CDRH3s will be immunogenic when administered intravenously. This has not proven to be a problem for llama paratopes but these CDRH3s are considerably shorter (median 16 aa versus >50 aa) than their bovine counterparts. If this does prove to be an issue for the ultralong bovine CDRH3s, it may be feasible to nebulize modified bovine scFvs, or smaller neutralising fragments such as knob peptides [41] , for delivery to the sites of virus entry [42] , as has been suggested for heavy chain only (VHH) nanobodies.

Pandemic preparedness and responsiveness hinges on the rapid identification of neutralising epitopes. The emergence of three beta-coronaviruses of pandemic potential in the last 20 years indicates the huge risk posed by these viruses and their high incidence of zoonotic transfers. Indeed, with over 50 known SARS-like viruses circulating in bats alone, this will not be the last emergence. Therefore, the generation of libraries of bNAbs that recognise this virus group will allow rapid screening against emerging related viruses. Indeed, even if the antibody has a relatively weak affinity to one virus sub-type, it may be useful against a related pathogen.

With a significantly expanded ultralong CDRH3 library, our pipeline has the potential to identify ultralong neutralising antibodies that can be rapidly deployed against new pathogens, without the need for animal immunizations.

293T cells were grown in Dulbecco's modified essential medium (DMEM) supplemented with 10 % foetal calf serum, 4 mM L-glutamine, 50 U/ml penicillin and 50 μg/ml streptomycin. Cells were grown in a humified incubator at 37 °C with 5% CO2.

To generate the pBovShow expression vector, DNA encoding an scFv expression cassette, with Ig kappa (IGK) leader sequence, (GGGS)x3 linker, Myc epitope tag and platelet derived growth factor receptor transmembrane domain (PDGFR-TM), was synthesized by IDT (Coralville, Iowa) and cloned into the EcoRV and XbaI sites of the pCS2-MT+ vector (Addgene plasmid #2296). Sequences encoding Vλ-light chains (LCs) were amplified from bovine whole blood genomic DNA using the primers described [43] . Individual clones were Sanger sequenced and screened for homology to Vλ-LCs that are known to productively pair with ultralong heavy chains [43] . A sequence with 99% homology to the BLV1H12 Vλ-LC was coli cells and used to inoculate an overnight midi culture for preparation of polyclonal plasmid DNA encoding the scFv library.

Full length trimeric Spike protein, residues 1-1208, was kindly provided by the Oxford protein 

The scFv library was subjected to amplicon sequencing following rounds 0, 2 and 3 of the 

A lentiviral plasmid for the stable expression of ultralong scFvs was generated by modifying 

For HDX-MS experiments, a robot for automated HDX (LEAP Technologies) was coupled to an Acquity M-Class liquid chromatography (LC) system and HDX manager (Waters). Samples Table 1 .

The sequence encoding B9-scFv was mutated in the BovShow cell surface expression vector; residues 123YNCRPAVWY131 of the B9-scFv knob domain (B9-WT) were replaced with the irrelevant amino acid sequence 123ETCYYGSGL131 by site-directed mutagenesis (B9-Mut) with Q5 polymerase (New England Bioloabs). 293T cells were transfected with either B9-WT or B9-Mut plasmid DNA and cell-surface expressed scFvs were tested for binding to purified SARS-CoV RBD (200 nM) using the standard staining protocol.

The PDB files used in this study were 4K3D, 6M0J, 5X5B and 6VXX. All structural figures used in this study were generated in UCSF Chimera.

Flow cytometry experiments include a positive and unstained negative control and were performed at least in triplicate and/or with sufficient replicates to ensure statistically significant data (except Fig. 5c that was performed in duplicate). Quantification of binding is determined using mean fluorescence intensity via CytExpert2. 4 and is plotted to show the standard error of the mean. The KD for interactions between cell surface scFvs and recombinant RBD proteins was estimated by non-linear analyses of the log(molarity)-response plots on

GraphPad.

Pseudotype neutralisation assays were performed at least in triplicate to calculate the standard error of % neutralisation (compared to the negative control) at each concentration of scFv. IC50 values are calculated from the nonlinear regression of log(molarity) of scFv versus % neutralisation.

For differential HDX-MS, peptide-level significance testing was implemented in Deuteros 2.0 to identify peptides with significant differences in deuterium uptake in the bound state. A hybrid significance test was used that first evaluates if the difference in deuterium uptake between two states is greater than a threshold value that corresponds to a significance level of p < 0.01.

This was followed by a Welch's t-test to confirm that the differences are significant.

The raw HDX-MS data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset: PXD032965 

We are grateful to the Oxford protein production group for the trimeric SARS-CoV-2 Spike protein and BEI for a series of reagents. Specifically, the following reagents were obtained through BEI Resources, NIAID, NIH: (a) Human Embryonic Kidney cells (HEK-293T) expressing the human angiotensin-converting enzyme 2, HEK-293T-hACE2 cell line, NR-52511 and (b) Modified pαH Vector Containing the Human Angiotensin-Converting Enzyme 2, NR-52565. The following reagents were also obtained through BEI resources and produced under HHSN272201400008C: (a) Plasmid Set for Anti-SARS Coronavirus Human Monoclonal Antibody CR3022, NR-53260 and (b) Vector pCAGGS Containing the SARS-Related Coronavirus 2, Wuhan-Hu-1 Spike Glycoprotein Gene (soluble, stabilized), NR-52394. We also gratefully acknowledge Professor Nicola Stonehouse and Dr

Efficacy and Safety of the mRNA-1273 SARS-CoV-2 Vaccine

Safety and efficacy of the ChAdOx1 nCoV-19 vaccine (AZD1222) against SARS-CoV-2: an interim analysis of four randomised controlled trials in Brazil, South Africa, and the UK. The Lancet

Safety and Efficacy of the BNT162b2 mRNA Covid-19 Vaccine

Monoclonal antibodies for the treatment of COVID-19 in a patient with high-risk acute leukaemia

Mapping Neutralizing and Immunodominant Sites on the SARS-CoV-2 Spike Receptor-Binding Domain by Structure-Guided High-Resolution Serology

Potently neutralizing and protective human antibodies against SARS-CoV-2

Receptor Recognition by the Novel Coronavirus from Wuhan: an Analysis Based on Decade-Long Structural Studies of SARS Coronavirus

Complete Mapping of Mutations to the SARS-CoV-2 Spike Receptor-Binding Domain that Escape Antibody Recognition

Current status of therapeutic monoclonal antibodies against SARS-CoV-2

Deep Mutational Scanning of SARS-CoV-2 Receptor Binding Domain Reveals Constraints on Folding and ACE2 Binding

Omicron-B.1.1.529 leads to widespread escape from neutralizing antibody responses. bioRxiv

Omicron escapes the majority of existing SARS-CoV-2 neutralizing antibodies

Cross-neutralization of SARS-CoV-2 by a human monoclonal SARS-CoV antibody

Broad and potent activity against SARS-like viruses by an engineered human monoclonal antibody

SARS-CoV-2 RBD antibodies that maximize breadth and resistance to escape

Broad neutralization of SARS-related viruses by human monoclonal antibodies

Broad sarbecovirus neutralization by a human monoclonal antibody

Immunizations with diverse sarbecovirus receptor-binding domains elicit SARS-CoV-2 neutralizing antibodies against a conserved site of vulnerability

Defining variant-resistant epitopes targeted by SARS-CoV-2 antibodies: A global consortium study

Rapid elicitation of broadly neutralizing antibodies to HIV by immunization in cows

Structural basis of broad HIV neutralization by a vaccine-induced cow antibody

Development of Foot-and-Mouth Disease Virus-Neutralizing Monoclonal Antibodies Derived From Plasmablasts of Infected Cattle and Their Germline Gene Usage

Mechanistic principles of an ultra-long bovine CDR reveal strategies for antibody design

Broadly Neutralizing Bovine Antibodies: Highly Effective New Tools against Evasive Pathogens? Viruses, 2020

Conservation and diversity in the ultralong third heavy-chain complementarity-determining region of bovine antibodies

Exceptionally long CDR3H region with multiple cysteine residues in functional bovine IgM antibodies

Extensive CDR3H length heterogeneity exists in bovine foetal VDJ rearrangements

Structural Diversity of Ultralong CDRH3s in Seven Bovine Antibody Heavy Chains. Front Immunol

Cross-neutralizing antibodies bind a SARS-CoV-2 cryptic site and resist circulating variants

Site-specific glycan analysis of the SARS-CoV-2 spike

Engineered ACE2 receptor traps potently neutralize SARS-CoV-2

A highly conserved cryptic epitope in the receptor binding domains of SARS-CoV-2 and SARS-CoV

A natural mutation between SARS-CoV-2 and SARS-CoV determines neutralization by a cross-reactive antibody

A glycan gate controls opening of the SARS-CoV-2 spike protein

Directed evolution using dCas9-targeted somatic hypermutation in mammalian cells

Structure-guided affinity maturation of a single-chain variable fragment antibody against the Fu-bc epitope of the dengue virus envelope protein

Pharmacokinetic properties of IgG and various Fc fusion proteins in mice

Tailoring structure-function and pharmacokinetic properties of singlechain Fv proteins by site-specific PEGylation

The allosteric modulation of complement C5 by knob domain peptides

Nanobodies(R) as inhaled biotherapeutics for lung diseases

Bovine IgM antibodies with exceptionally long complementaritydetermining region 3 of the heavy chain share unique structural properties conferring restricted VH + Vlambda pairings

MiXCR: software for comprehensive adaptive immunity profiling

Protocol and Reagents for Pseudotyping Lentiviral Particles with SARS-CoV-2 Spike Protein for Neutralization Assays

We gratefully acknowledge funding from the BBSRC (BB/V01384X/1 

The authors have no competing interests to declare.

Requests for materials and manuscript correspondence should be sent to Joan Boyes (j.m.boyes@leeds.ac.uk)