Sarbecoviruses interact with their receptor, angiotensin-converting enzyme 2 (ACE2), via the receptor-binding domain (RBD) of Spike, the immunodominant target for neutralising antibodies. Understanding the interplay and correlation between ACE2-determined host range and antigenicity is vitally important for understanding the zoonotic potential of related bat sarbecoviruses. Using binding assays, pseudotype-entry assays and a diverse panel of mammalian ACE2 proteins, we examined the host range and related antigenicity of multiple bat coronaviruses. Broad bat ACE2 usage (a generalist phenotype) was most common in clade 1 sarbecoviruses, including SARS-CoV-2 and the BANAL isolates from Laos. In contrast, clade 3 (e.g., RhGB07) and 5 (e.g., Rc-o319) sarbecoviruses exhibited more restricted ACE2 usage (a specialist phenotype). A novel structure for RhGB07 Spike further helped to identify RBD residues associated with this receptor specialism. Interestingly, the generalist phenotypes were largely maintained with more diverse mammalian receptor libraries, including human, non-human primate, livestock, rodent ACE2 and potential intermediate reservoir hosts (e.g., civet, racoon dog, pangolin), while specialists, like RhGB07, exhibited wider phenotypic diversity. The impact of SARS-CoV-2’s continued evolution in humans was also examined, identifying an expanding and/or shifting pattern of generalism for variants, especially Omicron and its sub-lineages. Furthermore, we compared and correlated these entry phenotypes with antigenicity using sera from SARS-CoV-2 convalescent individuals. Clade 1 viruses, phylogenetically related to SARS-CoV-2, were antigenically the most similar, with robust evidence for cross-neutralisation; however, there was still evidence for limited cross-neutralisation across the entire sub-genus. Finally, using monoclonal antibodies, derived from COVID-19 vaccinees with breakthrough infections, we pin-pointed the antibody epitope classes responsible for wider neutralisation. Our research indicates that generalist ACE2-using sarbecoviruses are phylogenetically and antigenically related to SARS-CoV-2.

Cells

HEK293T cells were used to generate lentiviral pseudoparticles, and BHK-21 cells for receptor screens. HEK293T cells stably expressing human angiotensin-converting enzyme 2 (HEK293T-hACE2) were used for pseudotype neutralisation assays. Cells were maintained in Dulbecco’s Modified Eagle Medium (DMEM, Sigma-Aldrich) supplemented with 10% FBS (Life Science Production), 1% 100mM sodium pyruvate (ThermoFisher Scientific), and 1% penicillin/streptomycin (10,000 Uml − 1; Life Technologies) at 37 °C in a humidified atmosphere of 5% CO₂. Stable cell lines were generated, as described previously [28], using a lentiviral transduction system under 1 μg/mL puromycin (ThermoFisher Scientific) selection.

Plasmids

Codon-optimised bat sarbecovirus Spike constructs were synthesised and subcloned into pcDNA3.1 (+) with a C-terminal FLAG tag (BioBasic, Canada). Human codon-optimised, Δ19-truncated Spike genes from RhGB07 and RfGB02 were subcloned into pcDNA3.1 (+) [10]. Mammalian ACE2 constructs were subcloned into pDISPLAY with an N-terminal signal peptide (the murine Ig k-chain leader sequence) and HA-tag (BioBasic, Canada; Twist Bioscience, USA). Accession numbers can be found in S1 and S2 Tables, respectively. All SARS-CoV-2 variant Spike expression plasmids were based on codon-optimised SARS-CoV-2 lineage B (Pango) reference sequence (GenBank ID NC_045512.2), with an additional Δ19 cytoplasmic tail truncation. Mutants of B.1 and BA.1, mutants of RhGB07, RBD chimeras of SARS-CoV-2 and RhGB07, and RBD chimeras of SARS-CoV-2 and Rs4231 were generated by site-directed mutagenesis, using the QuikChange II Lightning Multi Site-Directed Mutagenesis kit (Agilent).

Lentiviral pseudotype production

Viral pseudotypes bearing the Spike of different bat sarbecoviruses were generated as described previously [29]. Briefly, HEK293T cells were transfected with p8.91 (gag/pol), CSFLW (lentivirus backbone expressing Firefly luciferase) and the Spike construct of interest using PEI transfection reagent (Merck). Virus was harvested 48 and 72 h post-transfection and centrifuged and frozen at −80 °C to remove cellular debris. Viruses were titrated on HEK293T-hACE2 stable cell lines, or where viruses used hACE2 poorly, HEK293T cells were transfected with the cognate bat ACE2 receptor (i.e., Rh. cornutus for Rc-o319, Rh. ferrumequinum for RhGB07 and RfGB02). Luciferase signals were measured 3 days later using Bright-Glo (Promega).

ACE2 receptor screen

Receptor usage screens were conducted as described previously [15,21,29]. BHK-21 cells were seeded in 6-well plates at 7.5 × 10⁵/well in DMEM-10% 1 day prior to transfection with 500 ng of different species, ACE2-expressing constructs or empty vector (pDISPLAY) (S2 Table) in OptiMEM (ThermoFisher Scientific) and TransIT-X2 (Mirus Bio) transfection reagent, according to the manufacturer’s recommendations. The next day, cells were re-plated into white-bottomed 96-well plates at 2 × 10⁴/well and infected with pseudoparticles equivalent to 10⁵ to 10⁶ relative light units (RLU), pre-determined by titrating on BHK-21 cells overexpressing Rh. cornutus ACE2 (Rc-o319), Rh. alcyone ACE2 (RfGB02, RhGB07) or human ACE2 (all others) and the volume of pseudoparticles added to the assay was adjusted accordingly. For non-ACE2-using pseudoviruses, a fixed volume of 100 µL was used in subsequent assays. Transfected cells were incubated with pseudotyped virus for 48 h at 37 °C, 5% CO₂. To quantify Firefly luciferase, media was replaced with 50 μL Bright-Glo substrate (Promega) diluted 1:2 with PBS. The plate was incubated in the dark for 2 min and then read on a Glomax Multi+ Detection System (Promega) as above. Comma-separated values (CSV) files were exported for analysis. Biological replicates were performed 3 times.

Micro virus neutralisation assays

Neutralisation assays were conducted as described previously [30]. COVID-19 convalescent plasma/serum were diluted in serum-free media for a final dilution of 1:40 and added in triplicate to a white-bottomed 96-well plate and titrated 3-fold. Serial dilution of mAb were prepared. Convalescent plasma/serum or mAbs were incubated with a fixed titred volume of sarbecovirus pseudoparticles equivalent to 10⁵–10⁶ signal luciferase units and incubated for 1 h at 37 °C and 5% CO₂. Target HEK293T cells expressing human ACE2, or the relevant bat ACE2 receptor (Rh. alcyone for RhGB07 and RfGB02, Rh. cornutus for Rc-o319) were then added (at a density of 2 × 10⁵ cells/mL for convalescent serum/plasma, and at 4x10⁵cell/mL for mAbs) and incubated at 37 °C, 5% CO₂ for 72 h. Firefly luciferase activity was then measured after the addition of Bright-Glo luciferase reagent on a luminescence plate reader (GloMax-Multi+ Detection System (Promega) for convalescent serum/plasma or Victor X3 multilabel plate reader (PerkinElmer) for mAbs). Pseudotyped virus neutralisation titres were calculating by interpolating the point at which there was a 50% reduction in luciferase activity, relative to untreated controls (IC₅₀).

Cell–cell fusion assays

HEK293T rLuc-GFP 1–7 [28] effector cells were transfected in OptiMEM (ThermoFisher Scientific) using Transit-X2 transfection reagent (Mirus Bio), as per the manufacturer’s recommendations, with sarbecovirus Spikes (500ng) alongside mock-transfection with empty plasmid vector (pcDNA3.1+) (S1 Table). BHK-21 rLuc-GFP 8–11 target cells were transfected with 500 ng of different ACE2-expressing constructs (S2 Table). Effector and target cells were co-cultured at a ratio of 1:1 in white 96-well plates to a final density of 4 × 10⁴ cells/well 2 days post-transfection. Quantification of cell–cell fusion was measured based on Renilla luciferase activity, 24 h later by adding 1 μM of Coelenterazine-H (Promega) at 1:400 dilution in PBS. The plate was incubated in the dark for 2 min and then read on a Glomax Multi+ Detection System (Promega) as above and CSV files were exported for analysis. To quantify GFP expression, co-cultures were set up of effector and target cells as above into a clear flat-bottomed 96-well plate (Nunc) and imaged 24 h later using the IncuCyte S3 live cell imaging system (Essen BioScience). Four fields of view were taken per well at 10× magnification. Analysis of images generated was performed using the IncuCyte S3 software, setting fluorescence (“masks”) and cellular thresholds (“objects”) to remove background fluorescence. The total sum of objects with fluorescent masks was then calculated using the total integrated intensity metric, expressed as green count units (GCU) µm².

Recombinant protein production

Secreted sarbecovirus RBDs and mammalian ACE2 proteins were produced by cloning into the pOPIN-BAP-His and pOPIN-Fc expression vectors, respectively. These plasmids were then expressed in Expi293F cells using the PEI40K transfection reagent (Polysciences) with additives (300 mM valproic acid, 500 mM of sodium propionate, 2.5 M glucose). Supernatants were harvested 4 days post-transfection, centrifuged for 45 mins at 6,000g and filter through a 0.45 µM filter. Presence of secreted protein was determined by Coomassie and by blotting for anti-His or anti-Fc. Supernatants were purified using HisTrapHP or HiTrap Protein A columns (Cytivia), desalted using Zeba Spin Desalting Columns (ThermoFisher Scientific) and concentrated using Amicon Ultra 15 columns (Merck Millipore, #C7715). The proteins were then quantified using a Nanodrop and a Pierce BCA Protein assay (ThermoFisher Scientific) and run on an SDS-PAGE gel to assess protein purity by Coomassie.

Structural analysis

Surface plasmon resonance

Kinetics and epitope binning assays were performed on the Carterra LSA^XT using HC30M chips. Chips were pre-conditioned using sequential 60s injections of 50 mM NaOH, 1 M NaCl, and 10 mM pH2 glycine. For the kinetics assay, an anti-human Fc lawn was prepared by covalently coupling goat anti-human-Fc polyclonal antibody (BioRad 5211-8004) to the chip surface. Free carboxyl groups on the chip surface were activated for 5 min using 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride (EDC) and 115 mg N-hydroxysuccinimide (NHS) (final concentrations of 130 and 33 mM, respectively, in PBS). Goat anti-human-Fc was diluted to 50 μg/ml in 10 mM sodium acetate (NaOAc) to pre-concentrate the protein into the negatively charged mix and applied to the chip for 7 min in two independent locations. The coupling reaction was quenched for 5 min using 500 mM ethanolamine, and the chip washed with three 30-second pulses of pH 1.9 glycine (10 mM). The human anti-RBD monoclonal antibodies were diluted in PBST (PBS + 0.005% tween 20) to 1 µg/ml and 0.25 µg/ml and captured to the anti-human-Fc lawn for 7 min. Non-regenerative kinetics was performed using RBD diluted 2-fold in PBST (5 μg/ml to 2.4 ng/ml) allowing a seven-minute association and 14- minute dissociation. After the highest concentration of RBD the anti-human Fc lawn was regenerated using three 30-second injections of pH 1.9 glycine (10 mM) and fresh human anti-RBD antibodies were captured for each sarbecovirus RBD kinetics cycle. Kinetic analysis was performed in Kinetics v1.9.2 to calculate K_D by fitting to the 1:1 model following channel and double referencing. Average K_D was calculated using the values obtained from both coupling concentrations and the two independent couplings on the chip giving n = 4.

For epitope binning, human anti-RBD monoclonal antibodies were diluted to 1 μg/ml in NaOAc and directly coupled to the pre-conditioned HC30M chip as above allowing a 10-min coupling period. SARS-CoV-2 wild-type RBD was diluted to 0.625 μg/ml in PBST and injected for 5 min, followed by the competing antibody at 1 μg/ml for 5 min. The chip was regenerated using pH 1.9 glycine (10 mM) as previously before injecting the next RBD and competitor antibody. Epitope binning and community analysis was performed in Epitope v1.9.2.

Flow cytometry

BHK-21 cells were transfected using Transit-X2 transfection reagent (Mirus Bio), as per the manufacturer’s instructions with 500 ng of each ACE2-expressing construct (S2 Table) or mock-transfected with empty plasmid vector (pDISPLAY) for 48 h. Cells were resuspended in cold PBS and washed in cold stain buffer (PBS with 1% BSA (Sigma-Aldrich), 0.01% NaN₃ and protease inhibitors (ThermoFisher Scientific). Cells were stained with anti-HA Phycoerythrin (PE)-conjugated antibody (Miltenyi Biotech) at 1:500 dilution for 10⁶ cells for 30 min on ice. In parallel, 10 μg RBD recombinant proteins were stained with anti-His-Allophycocyanin (APC)-conjugated antibody (Miltenyi Biotech) at 1:500, and 50 µl of stained RBD were mixed with cells for 1 h to allow binding. Cells were then fixed with 2% paraformaldehyde (Fisher Scientific) on ice for 20 mins, then resuspended in PBS before being analysed using the MACSQuant Analyzer 10 (Miltenyi Biotech), and RBD-ACE2 binding was statistically calculated by SE Dymax % Positive. The mean SE Dymax % Positive was calculated from three biological replicates and analysed using FlowJo_v10.8.1. The gating strategy can be found in S4A Fig and the same gating strategy was applied in all experiments.

Western blotting

BHK-21 cells were transfected using Transit-X2 transfection reagent (Mirus Bio), as per the manufacturer’s instructions with 500 ng of different ACE2-expressing constructs (S2 Table) or mock-transfected with empty plasmid vector (pDISPLAY) and harvested 48 h post-transfection. Pseudotyped virus was purified and concentrated under a 20% sucrose cushion by ultracentrifugation (23,000 rpm for 2 h, 4 °C). All protein samples were generated using 2 × Laemmli buffer (Bio-Rad) and reduced at 95˚C for 5 min. Samples were resolved on 7.5% acrylamide gels by SDS-PAGE, using semidry transfer onto nitrocellulose membrane. Blots were probed with anti-HA primary antibody (Cell Signalling, #2367) for ACE2-transfected cells at 1:1,000 or with anti-GAPDH (Cell Signalling, #2118) internal control at 1:3000. Purified psuedovirus was resolved using an anti-FLAG primary antibody (Cell Signalling, #2368) at 1:1,000, anti-Spike antibody (Novus Biologicals, #NB100-56578) at 1:500 or anti-p24 (ThermoFisher Scientific, #MA1-71515) as an internal lentiviral control at 1:1,000 in PBS-Tween 20 (PBS-T, 0.1%) with 5% (w/v) milk powder overnight at 4˚C. Blots were washed in PBS-T and incubated with anti-mouse (for HA and p24) or anti-rabbit secondary (for GAPDH, Spike and FLAG), antibody conjugated with horseradish peroxidase (Cell Signalling, #7076 and #7074, respectively) at 1:3,000 in PBS-T for 1 h at room temperature. Membranes were exposed to Clarity Western ECL substrate (Bio-Rad) according to the manufacturer’s guidelines and exposed to autoradiographic film.

RBD ELISA

Recombinant RBD was coated onto 96-well ELISA plates at 1 µg/ml per well in carbonate/bicarbonate coating buffer (0.6 M pH9.6) at 4–8 °C overnight. Plates were blocked for 1 at RT with PBS-T containing 1% BSA, after which a 3-fold dilution series of monoclonal antibodies (starting concentration 5 µg/ml) or polyclonal COVID-19 plasma/serum (starting concentration 50 µg/ml), or a 4-fold dilution series of ACE2 recombinant proteins (starting concentration 8 µg/ml), were diluted in PBS and added for 1 h, at 37 °C. Plates were washed 3 times with PBS-T, followed by the addition of anti-human-Fc HRP conjugate diluted 1:10,000 added for 1 h at 37 °C. 1-step Ultra TMB (3,3′,5,5′- 40 Tetramethylbenzidine) (Thermo Fisher Scientific) was added to each well, incubated for 5–10 min at RT and the reactions stopped with an equivalent volume of 1 M Sulphuric acid solution, after which the optical density (OD) at 450 nm was measured using a GloMax Microplate Reader (Promega). Samples were analysed by plotting 50% effective concentration (EC50) using GraphPad Prism (8.2.1) software.

Different bat Sarbecovirus clades show varied bat ACE2 usage

Sarbecoviruses have been isolated from different bat species across Afro-Eurasia, phylogenetically clustering into established virus clades, when aligned using the RBD amino acid sequences. Representative viruses from each clade were selected for analysis based on sequence availability and experimental data in publications [5,6,12,17,36] (Fig 1A; highlighted in red). This included the clade Ia viruses SARS-CoV-1, WIV-1 and Rs4231 (yellow), clade Ib viruses SARS-CoV-2, RaTG13, BANAL-20-52, BANAL-20-236 and BANAL-20-103 (orange), clade II viruses RacCS203, RpYN06, Rf1/2004 and Rp3/2004 (blue), clade III viruses RhGB07, RfGB02 (green), and the clade V virus Rc-o319 (red). Clade IV viruses were not included in this study. The selected viruses have been isolated primarily in Asia (clade I, II, IV and V), and in parts of Europe and Africa (clade III), as indicated (Fig 1B). The bat species from which these viruses were isolated are globally widespread (individual geographic distributions of each species are shaded as shown; overlay in Fig 1B and individually in S1 Fig). Of note, Rhinolophus species, Pteropus giganteus, Lyroderma lyra, Rousettus leschenaultii and Taphozous melanaopogon bat species overlap in their geographic reach, particularly in Southeast Asia, whereas the Desmodus rotundus, Antrozous pallidus and Myotis lucifugus species intersect in their range only in the Americas (Fig 1B).

To understand how bat ACE2 tropism varies across the sarbecovirus sub-genus, a receptor usage screen was conducted in cells overexpressing human ACE2 (hACE2) or 16 different bat ACE2 receptors, which included a range of Rhinolophus species (Rh. cornutus, Rh. pusillus, Rh. macrotis, Rh. sinicus, Rh. pearsonii, Rh. affinis, Rh. shameli, Rh. ferrumequinum, Rh. Alcyone) and representative bat species from other families (L. lyra, P.giganteus, R. leschenaultii, T. melanopogon, D. rotundus, A. pallidus and M. lucifugus). When their sequence conservation was analysed in the context of the structure of SARS-CoV-2 RBD bound to Rh. affinis ACE2 (PDB 7XA7), the selected bat sarbecoviruses and bat ACE2s showed high levels of divergence at the binding interface, suggesting that the receptor usage of these viruses with bat ACE2s assessed in this study would likely vary (Fig 1C). Transfected cells were subsequently transduced with a fixed volume of bat sarbecovirus Spike-based pseudotypes equivalent to 10⁵–10⁶ RLU (those highlighted in red in Figs 1A and S2A; S1 and S2 Tables; S1 Data). Expression of ACE2 proteins in transfected cells, as well as SARS-CoV-2 Spike and p24 (internal lentivirus control) in purified pseudotypes, was confirmed in parallel (S2B and S2C Fig). To summarise, a diverse pattern of receptor usage was seen across the viruses tested (Fig 1D; S1 Data). Spike proteins from clade Ib sarbecoviruses, which includes SARS-CoV-2, had a broader (more generalist) receptor usage pattern, with RaTG13 being a notable exception. Indeed, only a few species of bat ACE2s were less permissive to SARS-CoV-2 or BANAL Spike-mediated entry. Clade Ia, which includes SARS-CoV-1, had a more intermediate receptor usage pattern, with clearer examples of restriction, even within the Rhinolophus bat family, e.g., Rh. ferrumequinum and Rh. pusillus. Nevertheless, WIV-1 and Rs4231, which were both isolated from Rh. sinicus, showed robust usage of this ACE2 receptor. In contrast, the clade III and V Spikes exhibited more restricted (specialist) ACE2 usage, primarily using the cognate bat ACE2 from which they were isolated, e.g., Rh. cornutus ACE2 usage by Rc-o319. As expected, the clade II viruses were unable to use any of the bat ACE2 receptors tested (Fig 1D, S1 data) [1,2,5,18,37,38]. Sequence alignment of the bat sarbecovirus RBDs, relative to SARS-CoV-2, revealed various deletions: at positions 446−449 in clade III viruses, at positions 480−487 for clade V viruses, and at positions 445−449 and 475−487 for clade II viruses (equivalent SARS-CoV-2 Wu-hu-1 numbering used throughout this study). The RBD-ACE2 interaction sites of Rh. affinis and human ACE2 were also compared based on known structures (human ACE2: PDB 6LZG, Rh. affinis ACE2: PDB 7XA7), noting differences in amino acid residues at positions 24, 27, 31, 34, 38 and 82. We also noted unique binding sites for human ACE2 that were not present for Rh. affinis ACE2, and vice versa (Fig 1E).

Coronavirus Spike proteins are also capable of inducing cell-cell fusion. To examine concordance between pseudotype entry and fusion, we repeated the assays in a quantitative cell-cell fusion assay, showing similar trends in quantifiable luciferase signals and GFP⁺ syncytia formation (S3A and S4 Figs; S1 Data). Our cell-cell fusion assay also showed high concordance with the pseudotyped-based entry assay by Pearson’s correlation (S3B Fig; S1 Data). The entry of bat sarbecoviruses with bat ACE2s also correlated to protein binding and affinity, i.e., those viruses that showed the highest levels of virus entry also bound most strongly in RBD-based flow cytometry assays (S5A and S5B Fig). Further, purified recombinant bat sarbecovirus RBDs were used to assess binding to recombinant human ACE2 and two bat ACE2s (Rh. cornutus, Rh. leschenaultti) in ELISA; however, these assays appeared to be less sensitive and do not allow determination of affinity, rather provide an approach to relatively rank binding across ACE2 orthologues (S5C Fig; S1 Data). Recombinant SARS-CoV-2 and RhGB07 RBDs were also used to assess binding to the three ACE2 species for which RhGB07 entry was observed (human, Rh. Alcyone, M. lucifugus). We identified good binding of SARS-CoV-2 RBD to these ACE2s, which was markedly reduced for RhGB07 RBD, correlating to the pseudotype entry phenotypes (Figs 1D and S5D; S1 Data).

RhGB07 structure reveals a modified putative RBD-ACE2 interaction interface

Our functional screens revealed interesting differences between viruses within the same clade, particularly clade III viruses. For example, although RhGB07 exhibited a specialist receptor usage profile, it was able to enter cells overexpressing human ACE2, whereas RfGB02 could not (Fig 1D). To understand the structural determinants of generalism and specialism, as well as zoonotic potential across sarbecoviruses we therefore focused on resolving the structure of the specialist Spike, RhGB07. Previously, we produced recombinant 2P RhGB07 Spike with a putative basic region mutated and tested this in BLI binding assays [10]. To reveal the molecular details of potential interactions between ACE2 and RhGB07 Spike, we attempted to determine a cryoEM structure of human ACE2 and RhGB07 Spike. We prepared grids for cryoEM by co-incubating either human or bat ACE2 (Rh. ferrumequinum) with RhGB07 Spike before freezing. Initial representative 2D classes did not reveal the presence of any bound ACE2 to Spike (S6A Fig; S3 Table). Further classification revealed the presence of only 3-RBD closed/down Spike particles (in the dataset containing hACE2 and RhGB07). We could clearly define a limited number of particles and when C3 symmetry was imposed we could refine a map to 2.97 Å (S6A Fig). The model was built by docking Alphafold2 predictions as individual domains using phenix.voyager followed by manual model-building, specifically within the NTD, RBD and the putative basic region of the RhGB07. The structure revealed a trimeric organisation of protomers, following a domain organisation typical of sarbecoviruses. As noted in the 2D classes, the structure revealed a completely closed structure that shows no variation between RBD up or down (Fig 2A).

Comparison with other Clade III RBD apo-structures, PRD-0038, Khosta-1, Khosta-2 and BtKY72 revealed high similarity between the C-alpha traces (S7A and S7B Fig). However, regions of differences were noted that coincide with putative ACE2-interacting residues based, particularly in the ins1-loop and Δ4-loop, which could impact ACE2 usage and receptor tropism. It was also noted that the closed-state PRD-0038 RBD structure resembled SARS-CoV-2 RBD with a fatty acid, linoleic acid, bound [14]. An examination of the RhGB07 RBD revealed no observable density (S7C Fig) [39]. Similarly, the NTD, that in some clade 3 binds biliverdin, also did not reveal any density (S7C Fig) [40]. Comparisons of the sequence conservation between RhGB07 and the other bat sarbecovirus Spikes examined in this study revealed the RBD and NTD as regions of highest variability (S7D Fig). Further comparisons between the structures of the generalist sarbecovirus Spikes from SARS-CoV-1 and SARS-CoV-2 with RhGB07 revealed structural similarity except at the sites of deletion and insertions (Fig 2B). Within the RBD, we determined the loops that mediate ACE2 interaction in SARS-CoV-1 and SARS-CoV-2 adopt strikingly different 3D conformations as compared to RhGB07 and also contain residues of different identities to RhGB07 despite the overall RBD structure appearing highly identical (Fig 2B). For example, the ‘ins1-loop’ where there is a single residue insertion in RhGB07 compared to SARS-CoV-2, positions serine 486 (RhGB07 numbering) in place of the Q498 residue in SARS-CoV-2. Changes at these hotspot loops may be determinants for RhGB07-ACE2 binding selectivity.

To examine the extent to which the Δ4-loops contribute to the restricted receptor tropism of RhGB07, mutants of RhGB07 and SARS-CoV-2 were generated for comparison. Firstly, full RBD chimeric swaps were generated, resulting in, as expected, a modest increase in human ACE2 usage for RhGB07_{SARS-CoV-2 RBD} pseudotypes, and a slight decrease in pseudotype entry for SARS-CoV_{RhGB07 RBD} (Fig 2C; S1 Data). Next, the direct importance of the Δ4-loop was investigated by inserting the SARS-CoV-2 GGNY residues between positions 437 and 438 in the RhGB07 Spike (RhGB07 437_438insGGNY) or deleting the corresponding region in SARS-CoV-2 (SARS-CoV-2 Δ446−449). A further decrease in human ACE2 usage was noted with the SARS-CoV Δ446−449 mutant, and a similar human ACE2 usage for RhGB07 437_438insGGNY, when compared to RhGB07_{SARS-CoV-2 RBD} (Fig 2C; S1 Data). This indicates that this region is important in mediating virus entry, likely due to altered ACE2 binding affinity, confirmed by ELISA (Fig 2D; S1 Data). To examine the wider significance of this result to sarbecovirus biology, further mutants of RhGB07 were generated to examine whether host range could also be altered by mutating the Δ4-loop region to reflect other generalist sarbecoviruses. RhGB07 was mutated to make it more SARS-CoV-2-like (RhGB07 437_438insVGGNY), SARS-CoV-1-like (RhGB07 437_438insSTGNY), Rs4231-like (RhGB07 437_438insTSGNY), RaTG13-like (RhGB07 437_438insEGGNF), or Rc-o319-like (RhGB07 437_438insTSGNF). Importantly, the receptor tropism of these mutants was examined with a small panel ACE2s that were previously shown to be differentially used by our bat sarbecovirus (Fig 1D). The use of human, Rh.alycone and M. lucifugus ACE2s increased with all mutants except the SARS-CoV-1-like mutant (RhGB07 437_438insSTGNY), which remained similar to wildtype RhGB07 (Fig 2E; S1 Data). Interestingly, the RhGB07 437_438insGGNY mutation alone was insufficient to broaden bat ACE2 tropism, but additional mutations to make the virus SARS-CoV-2-like (RhGB07 437_438insVGGNY) (generalist), increased usage of Rh. pusillus, Rh affinis, Rh. ferrumequinum and Rh. cornutus ACE2s (Fig 2E; S1 Data), highlighting the importance of the Q437V substitution. Also of note, an increased usage of Rh. affinis ACE2 was observed with the SARS-CoV-1-like mutant (RhGB07,437_438insSTGNY), but not the other bat ACE2s examined, mirroring the ACE2-usage profile of SARS-CoV-1 (Fig 1D). However, while the Rc-o319-like mutant (RhGB07 437_438insTSGNF) also exhibited an increase in Rh. affinis ACE2 usage, this was not observed for the RaTG13-like mutant (RhGB07 437_438insEGGNF) or Rs4231-like mutant (RhGB07 437_438insTSGNY), despite Fig 1D showing both RaTG13 and Rs4231 sarbecovirus pseudotypes are able to enter cells overexpressing Rh. affinis ACE2 (Fig 2E; S1 Data). Additionally, RhGB07 437_438QGGNF and RhGB07 437_438insEGGNY mutants alone (reflecting the Δ4-loop region sequences of RaTG13 and Rc-o319) were unable to increase receptor tropism, and in fact exhibited a decrease in human, Rh. alcyone and M. lucifugus ACE2 usage (Fig 2E; S1 Data). These data indicate a synergistic role for these residues in the Δ4-loop, but also the likely involvement of other residues in the RBD or Spike that contribute to receptor tropism.

Together, these functional and structural data suggest that sarbecoviruses have distinct receptor usage profiles, differences which can likely be attributed to variations within the RBD-ACE2 interaction interface, for example in the Δ4-loop. To understand whether patterns of generalism or specialism in the bat reservoir host are conserved in potential intermediate reservoirs we expanded our analysis to a wider range of mammalian ACE2 proteins.

Bat sarbecoviruses also have broad tropism for mammalian ACE2

Based on previously established ACE2 restrictions [15,17,18,21,41], evidence for a role in spillover of sarbecoviruses into humans [42–44], and/or the identification of SARS-CoV-2 reverse zoonotic spillover [25–27,44,45] we selected 19 additional mammalian ACE2s for analysis including human, livestock/pets (dog, cat, pig), rodents (mouse, rat, Syrian hamster), potential intermediate reservoirs (red fox, civet, ferret, pangolin, mink, raccoon dog, white-tailed deer, red deer), and monkeys (marmoset, squirrel monkey, lemur, macaque) (S2 Table). Clade II Spikes that did not use any of the screened bat ACE2s in the initial screen were excluded from further analysis. Similar to the bat ACE2s, clade Ib Spike pseudotyped viruses exhibited broad tropism with the majority of mammalian ACE2s tested, with notable restrictions for SARS-CoV-2 with mouse, rat, ferret, mink, marmoset and squirrel monkey ACE2, as published previously [15]. Interestingly, these restrictions were less obvious for the closely related BANAL viruses, including BANAL-20-52 whose RBD is almost identical to SARS-CoV-2, whereas RaTG13 was much more restricted. The mammalian ACE2 usage profile of the clade III specialists RfGB02 and RhGB07, as well as the clade V specialist Rc-o319, were more divergent than with the bat ACE2s. For example, whilst RhGB07 Spike was able to use the ACE2 from many mammals, including pangolin and raccoon dog to enter cells, Rc-o319 was entirely restricted with this panel. The lack of an absolute correlation between generalism and specialism in bats and other mammals was also seen for clade Ia pseudoviruses, which demonstrated almost universal usage of all ACE2s in our mammalian ACE2 receptor usage, aside from New World monkeys (Fig 3A; S1 Data). Comparison of mammalian ACE2 protein sequences (relative to human ACE2) revealed a much higher sequence conservation than seen for bat ACE2s (Fig 1C), particularly at the RBD-binding interface, with divergence observed at amino acid residues 24, 34, 79 and 82, and to a lesser extent at positions 30, 31 and 353 (Fig 3B and 3C).

Bat and mammalian ACE2 usage has changed in line with continued SARS-CoV-2 evolution in humans

The D614G-containing SARS-CoV-2 lineage, B.1 was responsible for the first wave of the COVID-19 pandemic, with subsequent waves caused by genetically divergent variants including Alpha, Beta, Delta, and Omicron. All these variants, in particular Omicron, were defined by the introduction of important mutations within the Spike protein (Figs 4A and S8A; S1 Data). To examine changes to bat and mammalian ACE2 receptor usage by these SARS-CoV-2 variants, we measured the entry efficiency of Spike-based pseudotypes in cells overexpressing ACE2 orthologues. For bat ACE2s, we observed enhanced (compared to B.1) bat ACE2 usage with the Alpha, Beta and Delta variants approaching or exceeding the levels recorded for human ACE2, however, the overall pattern of receptor usage did not change. In contrast, a divergent pattern was seen with Omicron BA.1, including shifted receptor usage of Rh. cornutus, Rh. pusillus, Rh. ferrumequinum and L. lyra ACE2 (Fig 4B, left panel; S1 Data).

In contrast, the species tropism for a wider selection of mammalian ACE2s was somewhat unchanged, maintaining the generalist properties of B.1, with the same notable restrictions, e.g., New World monkey (marmoset and squirrel monkey) and ferret. Interestingly, an increase in mouse and rat ACE2 usage was detected in all variants harbouring N501Y mutations in the Spike RBD, as previously shown (Fig 4B, right panel; S1 Data) [21].

Focussing on the significant differences in bat ACE2 receptor usage between B.1 and Omicron BA.1, we next investigated which mutations in the Spike were responsible. Compared to B.1, Omicron BA.1 has a significant number of amino acid substitutions in Spike, with 13 mutations in the RBD alone, particularly at residues that interact directly with ACE2, with the 417N and 484A mutations reducing potential interactions (Fig 4C). Based on our structural understanding of RBD residues that interact directly with ACE2, and/or previous evidence of a direct contribution to altered species tropism, we generated B.1→BA.1 substitution mutants (K417N, E484A, Q493R, N501Y) either individually or in combination. In addition, inverse mutations (BA.1→B.1) were built into a BA.1 backbone, with all mutants being functional when titrated on cells stably expressing human ACE2 (S8B Fig; S1 Data).

We then assessed the entry phenotype of these mutated Spikes with bat ACE2s that showed a switch towards increased (Rh. cornutus, Rh. pusillus, Rh. ferrumequinum) or decreased receptor usage (L. lyra) with BA.1 compared to B.1. We noted that whilst the K417N and E484A mutants on their own, or in combination, had little impact on receptor tropism, the Q493R and N501Y mutations, either in isolation or in combination with other mutations, resulted in increased receptor usage of Rh. cornutus and Rh. pusillus ACE2. For the Rh. ferrumequinum ACE2, the Q493R mutation alone was sufficient to switch receptor usage. For L. lyra, the decrease in receptor usage was largely attributable to the change at position 417 (Fig 4D, top panel; S1 Data). When the corresponding mutations were introduced into the BA.1 backbone, a similar phenotype was observed, with amino acid changes at positions 493 (R493Q) and 501 (Y501N) being detrimental to Rh. cornutus Rh. pusillus and Rh. ferrumequinum ACE2 receptor usage, especially when in combination. Likewise, the 417 (N→K) and 493 (R→Q) mutations in BA.1, as single mutations or together, were able to revert the observed restriction between BA.1 Spike and L. lyra ACE2 usage (Fig 4D, bottom panel; S1 Data).

Finally, we also examined the receptor usage of contemporaneous Omicron sub-lineages that have emerged and risen to dominance since 2022. Interestingly, aside from BA.4/5, and to a lesser extent BA.2.86, many of the Omicron variants tested (BA.2, XBB and XBB1.5), exhibited a reduced or complete loss of receptor usage for many bat ACE2 receptors (Fig 4E, left panel; S1 Data). This was not evident for the wider library of mammalian ACE2 species, which remained widely used by the Omicron sub-linages tested (Fig 4E, right panel; S1 Data). Indeed, there was evidence of previous restrictions being overcome by variants such as XBB and BA.2.86, e.g., New World monkey and ferret ACE2s. In summary, the continued evolution of SARS-CoV-2 in humans has shifted the receptor usage pattern of this sarbecovirus away from an ancestral bat phenotype, towards one that is more generalistic for other mammals, including humans.

Multiple bat sarbecoviruses share antigenic similarity to SARS-CoV-2

Several bat sarbecovirus Spikes in our study showed a broad host range for mammalian ACE2s, including human ACE2 (Fig 3A), suggesting the potential for spillover into human populations. To ascertain whether pre-existing immunity following infection with SARS-CoV-2 can provide cross-protection against related bat sarbecoviruses should they emerge in the human population, we examined their antigenic similarity by ELISA and VNT with a panel of COVID-19 convalescent plasma/serum obtained from the Medicines and Healthcare products Regulatory Agency (MHRA), UK. The antibody panel includes the World Health Organization (WHO) International Standards for anti-SARS-CoV-2 immunoglobulins, and the International Reference Panel material for antibodies to SARS-CoV-2 variants of concern B.1 (wildtype, WT), Alpha, Gamma, Delta, and Omicron BA.1. We also tested plasma from vaccinees with high and low neutralising antibody titres to SARS-CoV-2, as well as negative plasma collected from healthy donors pre-2019 (Fig 5A). By ELISA, we observed high levels of human IgG-based cross-specific binding to the RBDs of clade Ia and Ib viruses. Interestingly, we also noted binding to the RhGB07 and Rc-o319 RBD by most of the sera/plasma tested despite their phylogenetic divergence. (S9 Fig; S1 Data). The same sera/plasma also exhibited moderate-to-high neutralisation titres against equivalent pseudotyped viruses (BANAL-20-103, BANAL-20-236: ns; BANAL-20-52, * p < 0.05; Rs4231: ** p < 0.005; WIV-1: *** p < 0.0005, compared to SARS-CoV-2) (Fig 5B and 5C; S1 Data). Due to the inefficient usage of human ACE2 by Rc-o319 and RhGB07 pseudotypes, the target cells for these virus neutralisation assays were instead transiently transfected with Rh. cornutus ACE2 and Rh. alcyone ACE2, respectively. Compared to SARS-CoV-2 we noted lower neutralisation titres against RhGB07 (***, p < 0.0005), and lower still against Rc-o319 (***, p < 0.0005) (Fig 5B and 5C; S1 Data). Interestingly, the binding and neutralisation phenotype observed for these two viruses, which were the most antigenically distant in our screen, were similar to those observed for Omicron (BA.1). This is perhaps unsurprising given that most of the sera/plasma tested was taken before BA.1’s emergence; however, it does highlight the role of immune pressure in driving SARS-CoV-2 evolution (S8A–S5C Fig; S1 Data). Another interesting observation was that in some ELISAs we did not detect binding, e.g., S2-03 did not bind to the RhGB07 RBD, and, 21/296 and S2-04 did not bind to WIV-1 RBD, however, we were still able to detect neutralisation of respective pseudotypes, suggesting neutralisation sites outside the RBD (Figs 5B and S8; S1 Data). In summary, both polyclonal sera and plasma from SARS-CoV-2 convalescent individuals shows robust evidence for the presence of cross-neutralising pan-sarbecovirus antibodies.

Pan-sarbecovirus monoclonal antibodies are identifiable following breakthrough SARS-CoV-2 infection

Building on this cross-neutralisation data, we confirmed the RBD as the predominant target for this immune recognition. Using SARS-CoV-2 as an exemplar ACE2 generalist and Rs4231 as a specialist (Figs 1D, 3A, and 5), we generated RBD chimeras of these two Spikes and confirmed their functionality (S10A Fig; S1 Data). We opted against the more phylogenetically and phenotypically distant viruses, RhGB07 and Rc-o319, because of their inefficient human ACE2 usage, which would limit direct comparison with SARS-CoV-2. Using a panel of RBD-specific monoclonal antibodies (mAbs) isolated from SARS-CoV-2 breakthrough infections [46], we then conducted neutralisation assays with these chimeras. We detected a reduction in IC₅₀ when the Rs4231 RBD was introduced into the SARS-CoV-2 backbone (p < 0.0001), whereas the converse was observed when the SARS-CoV-2 RBD was inserted into the Rs4231 backbone (p < 0.001) (Fig 6A; S1 Data). We also confirmed that the differences in ACE2 receptor usage tropism between these two Spikes mapped to the RBD (S10B Fig; S1 Data), e.g., switching mouse and marmoset ACE2 usage (S10B Fig; S1 Data). These data confirmed the RBD as a major target for cross-specific antibodies, and that both receptor generalism/specialism and antigenicity functionally partition to the same Spike protein sub-domain.

The mAbs used in our screen have been characterised in more detail previously, based on binding and competition assays with mAbs to SARS-CoV-2 RBD (Group 1 [purple], Group 2 [orange], Group 3 [red], Group 3.5 [blue], Group 4 [green] and Group 8 [non-RBD binder]) [46] (Fig 6B). To confirm these mAb groups and to provide additional resolution, binding ELISAs and epitope binning competition assays were performed by assessing binding to SARS-CoV-2 RBD. In initial ELISAs, all RBD-specific mAbs bound to the SARS-CoV-2 RBD, whereas the NTD-specific mAb VA14_16 failed to recognise this protein, as expected (Fig 6C; S1 Data). Epitope binning was then performed by high-throughput surface plasmon resonance (SPR) on the Carterra platform. In general, the same antibody binding groups were preserved, with additional blocking interactions between groups observed (Figs 6D and S11). Blocking interactions between group 3, 3.5, and 4 antibodies varied, indicating the presence of subtle differences in antibody binding within these groups (Fig 6D). For example, antibodies VA14_26 (group 2) and V1WT_41 (group 3) were characterised as having similar blocking phenotypes, indicative of epitope overlap (Fig 6D). Interestingly, the V1WT_26 antibody (group 1) showed blockage of group 2, 3, and 4 antibodies, despite group 1 mAbs occupying binding a distinct region of the RBD [46], suggesting possible steric hindrance (Fig 6D).

Building on the partial neutralisation of Rs4231 with these mAbs, we performed a more comprehensive assessment of their binding (by ELISA) phenotype, against other bat sarbecoviruses in our library. As with the polyclonal sera tested above, SARS-CoV-2, BANAL-20-52 and BANAL-20-236 RBDs were bound well by all the mAbs tested. In contrast, the more divergent WIV-1, Rs4231, Rc-o319 and RhGB07 RBDs were bound less robustly (Fig 6E; S1 Data). However, significant binding was maintained by antibodies in group 2 (VA14_26, except RhGB07), group 3 (V1WT_41) and group 4 (V1WT_06) (Fig 6C; S1 Data). The lack of binding of VA14_26 with RhGB07 RBD is likely due to differences in amino acid residues present at the antibody-RBD interface that result in weaker, repulsive interactions (K370E and R400K in RhGB07 compared to SARS-CoV-2, WIV-1 and Rc-o319) (S12A Fig). Interestingly, these 3 mAbs were also able to bind the RBD of the non-ACE2-using clade II virus, RacCS203, albeit to a lesser degree (S12B Fig; S1 Data). As expected, none of the group 8 mAbs (NTD-specific) bound to any RBD (Fig 6E; S1 Data).

A representative subset of these sarbecovirus RBDs and mAbs were next analysed for complementary binding kinetics and affinity by ELISA and SPR, respectively. Generally, there was good correlation between the assays, wherein mAbs with low EC₅₀ values also exhibited lower K_D (dissociation constant) values, indicated higher affinity (Fig 7A; S1 Data). However, some RBD/mAb interactions characterised as non-binding by ELISA showed low binding by SPR and a greatly increased K_D (Figs 7A and S12C; S4 Table; S1 Data). For example, V3Om_03 showed minimal to no binding to the more divergent RBDs in ELISA, but in SPR a low level of binding was detected for Rs4231 and Rc-o319 (Fig 7A; S4 Table; S1 Data). Interestingly, the converse trend was observed for mAb V3WT_20 with RhGB07 RBD, with binding identified by ELISA but not by SPR (Fig 7A; S4 Table; S1 Data).

Finally, we examined the neutralisation of various pseudotyped spikes by these mAbs. These experiments showed a similar trend, with BANAL-20-52 exhibiting binding EC₅₀s and neutralisation IC₅₀ values of <0.08 µg/ml, close to SARS-CoV-2 (Fig 7B; S4 Table; S1 Data). Mirroring the ELISA data, Rs4231 (clade Ia), RhGB07 (clade III) and Rc-o319 (clade V) were less well neutralised; however, the same three monoclonal antibodies (VA14_26, V1WT_41 and V1WT_06) showed good efficacy, highlighting competition groups 3, 3.5 and 4 antibodies as being important mediators of cross-neutralisation (Fig 7B; S4 Table; S1 Data). Neutralisation IC₅₀ titres also correlated with K_D, with a larger K_D correlating with weakened or abrogated neutralisation. In contrast, the three antibodies which exhibited broad RBD binding and neutralisation (VA14_26, V1WT_41, and V1WT_06) displayed a much narrower fluctuation in K_D consistent with their ability to bind all the RBDs tested (Figs 7A, 7B, and S12C; S4 Table; S1 Data). Mapping the broad footprints of where these mAb competition groups might bind onto the Spike trimer of RhGB07 highlighted a significant number of amino acid residues that differ to SARS-CoV-2 RBD (Fig 7C). RhGB07 has a 3 RBD-down, closed conformation, which may contribute to mAb accessibility, particularly the Group, 3.5 and 4 mAbs, which target the most exposed epitopes on in the RhGB07 RBD down/closed state (Fig 7C). Our companion article describes in more detail the structural basis for the broad specificity of these antibodies to sarbecovirus RBDs [49].

S1 Fig. Geographic range of bat species investigated in this study.

The geographic location in which the bat sarbecoviruses analysed in this study have been isolated is coloured in blue on the world map, with separate maps depicting the range the bat species investigated in this study span. Rhinolophidae species have been grouped together and span across Europe, Asia and some parts of southern Africa. The T. melanopogon, P. giganeteus, L. Lyra and R. leschenautii species all span across southeast Asia, in particular overlapping in regions of India and parts of China. The M.lucifugus, D. rotundus and A.pallidus bat species can be found nesting in the Americas.

https://doi.org/10.1371/journal.pbio.3003944.s001

(TIF)

S2 Fig. Pseudotype virus titrations and protein expression of ACE2s and sarbecovirus Spikes.

(A) Lentiviral pseudotypes expressing different bat sarbecovirus Spikes were titrated on BHK-21 cells overexpressing human ACE2 (SARS-CoV-2, BANAL-20–52, BANAL-20–103, BANAL-20–236, RaTG13, SARS-CoV-1, WIV-1, Rs4231, Rf1, Rp3, RacCS203, RpYN06), Rh. cornutus ACE2 (Rc-o319) or Rh. alcyone ACE2 (RhGB07, RfGB02). Virus titres were used to normalise input virus for subsequent assays. (B) Expression of tagged ACE2 proteins was carried out in lysates from BHK-21 cells transfected with ACE2 constructs and probed for HA expression by Western blotting. (C) Sarbecovirus Spikes were used to generate lentivral-based pseudotypes, and the viral supernatant purified by ultracentrifugation under a 20% sucrose cushion. Purified viruses were probed with an anti-FLAG antibody (except RhGB07 and RfGB02, which are untagged), an anti-Spike antibody (S, full-length Spike; S1/S2, cleaved variant) and the lentivirus internal control, anti-p24 antibody. The data underlying this Figure can be found in S1 Data and S1 Raw Images.

https://doi.org/10.1371/journal.pbio.3003944.s002

(TIF)

S3 Fig. Entry of bat sarbecoviruses by cell-cell fusion.

(A) Heatmap showing virus entry by cell-cell fusion. HEK293T cells stably expressing half of a split GFP-Renilla (RLuc-GFP 1−7) luciferase reporter were transfected with bat sarbecovirus Spikes, and co-cultured with BHK-21 cells expressing the other half of the split reporter (RLuc-GFP 8−11), transfected with bat ACE2s. Productive cell-cell fusion results in reconstitution of the split reporter, from which luciferase signals can be measured. Data is shown as log (RLU) and is the average of three separate experiments. A pDISPLAY vector control was included to set the baseline value to account for background signals. The red boxes indicate the predicted cognate bat receptor for each bat sarbecovirus, where known. (B) XY scatter plots for each of the viruses assessed for their receptor usage profile, correlating the log (RLU) of pseudoparticle entry (X) against cell-cell fusion entry (Y), excluding clade II viruses that do not use ACE2 as their entry receptor (RfGB02: r = 0.8244, p < 0.0001; RhGB07: r = 0.9335, p < 0.0001; Rc-o319: r = 0.9467, p < 0.0001; SARS-CoV-2: r = 0.6684, p = 0.0024; BANAL-20–103: r = 0.5308, p = 0.0234; BANAL-20–236: r = 0.6096, p = 0.0072; BANAL-20–52: r = 0.7343, p = 0.0005; RaTG13: r = 0.7719, p = 0.0002; SARS-CoV-1: r = 0.9005, p < 0.0001; WIV-1: r = 0.7724, p = 0.0002; Rs4231: r = 0.8259, p < 0.0001). Each data point is labelled with the ACE2-corresponding ACE2 receptor and a Pearson’s correlation calculated with 95% confidence intervals. The Pearson r value and p value are both documented. The data underlying this Figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003944.s003

(TIF)

S4 Fig. GFP+ syncytia formation of bat sarbecovirus entry by cell–cell fusion.

(A) GFP and brightfield images of cell-cell fusion of representative bat sarbecovirus Spikes (transfected into HEK293T effector cells stably expressing rLuc-GFP 1–7) with bat ACE2s (transfected into BHK-21 cells expressing rLuc-GFP 8–11) 24 h after co-culture. Four fields of view were taken per well at 10x magnification. Scale bar is shown, 400 µm. (B) Analysis of images generated was performed using the IncuCyte S2 software, calculating the total sum of object with fluorescence above the background threshold, measured as total integrated intensity, and expressed as green count units (GCU) µm² and shown in a heatmap. The data underlying this Figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003944.s004

(TIF)

S5 Fig. Gating strategy and analysis for flow cytometry of ACE2-expressing constructs bound to sarbecovirus RBDs.

(A) BHK-21 cells were transfected with HA-tagged bat ACE2-expressing constructs (S2 Table) and baited with His tagged-purified RBD proteins for bat sarbecoviruses. Cells were stained with anti-HA PE-conjugated antibody (ACE2) and anti-His APC-conjugated antibody (RBD). Live and singlet BHK-21 cells were gated as PE + APC+ (ACE2 + RBD+), relative to mock-transfected cells (pDISPLAY). (B) The SE% Dymax metric was used to calculate the percentage of ACE2 + RBD+ cells, shown for all RBDs tested with our library of bat ACE2 and human ACE2. Representative datasets are shown for mock-transfected cells and raccoon dog ACE2 with SARS-CoV-2 RBD, with their respective SE Dymax % positive value noted. (C) Heatmap showing the binding of purified bat sarbecovirus proteins to ACE2s of human, M. lucifugus Rh. cornutus by ELISA (pink, EC₅₀ µg/ml) or flow cytometry (blue, % ACE2⁺RBD⁺ cells). (D) ELISA binding curves of human, M. lucifugus and Rh. alcyone ACE2s with SARS-CoV-2 and RhGB07 RBDs, depicted also as a heatmap with calculated EC₅₀ µg/ml. EC₅₀ values were also correlated with entry data from Fig 1D. Scale bars indicate direction of increased binding (pink) or entry (blue). Data is representative of a mean of 3 separate experiments. A cross in a box indicates this data was not obtained. The data underlying this Figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003944.s005

(TIF)

S6 Fig. Cryo-EM Processing Workflow.

(A) Processing workflow for the determination of the RhGB07 full-length trimeric spike. (B) Cryo-EM density coloured by local resolution of the full-length trimer and of the isolated RBD. Selected regions represented with their cryo-EM density are also shown. (C) FSC curve showing a determined resolution for the full-length spike of 2.97 Å (D) Viewing direction distribution of particles used in the reconstruction. (E) Relative Signal versus Viewing direction plot of particles. Cryo-EM data collection, refinement and validation statistics can be sound in S3 Table.

https://doi.org/10.1371/journal.pbio.3003944.s006

(TIF)

S7 Fig. Comparison of RhGB07 with other sarbecoviruses.

(A) The C- α trace of the RBD of RhGB07 rendered as balls-and-sticks superimposed with the RBD apostructures of PRD-0038 (PDB 8U29), Khosta-1 (PDB 9J6G), Khosta-2 (PDB: 9J5U) and BtKY72 (PDB 8ZY0). The insets highlight regions of interest of the RBD that are either involved in binding of ACE2 or have presented structural changes across clades. (B) The C-α trace of the RBD of RhGB07 rendered as balls-and-sticks superimposed with the RBD of PRD-0038 apo or PRD-0038 bound to R. Alcyone ACE2 (PDB: 8U0T). The insets show a zoomed-in view of two regions of the RBD that show the greatest changes in the position of the C-α of the RBD. The region containing the ins1-loop and Δ4-loop coincides with putative ACE2-interacting residues based on mapping the equivalent residues. (C) RBD (top) and NTD (bottom) of RhGB07 cartoon depicted with the cryo-EM density contoured in this region. Residues comprising a putative fatty acid binding pocket as seen in SARS-CoV-2 RBD [39] or a putative biliverdin binding pocket as seen in other coronavirus NTD [40] are highlighted. No density was present in either of these pockets corresponding to a fatty acid or biliverdin, respectively. (D) Sequence conservation of selected sarbecoviruses were mapped onto the RhGB07 spike. Areas least conserved are in red and most conserved are blue. The least conserved regions are at the top of the RBD, the NTD and close to a putative polybasic site.

https://doi.org/10.1371/journal.pbio.3003944.s007

(TIF)

S8 Fig. SARS-CoV-2 variants entry and binding with a library of bat ACE2.

(A) Schematic of the Spike proteins for B.1, Alpha, Delta, Omicron BA.1, Omicron BA.2 and its subvariants BA.4/5, XBB, XBB1.5 and BA.2.86. Key regions of the N-terminal domain (NTD), receptor-binding domain (RBD), S1 and S2 are highlighted and the mutations that arose in these variants have been annotated. The BA.4 and BA.5 Spike is identical, and the difference in XBB1.5 compared to XBB is denoted by an asterisk, which is a reversion of the mutation at position 494 (R493Q). (B) Infection of HEK293T cells stably expressing hACE2 with SARS-CoV-2, WT and BA.1 pseudoparticles, along with individual mutants generated of these variants at positions 417, 484, 493 and 501, either singly or in combination. Virus was titrated 5-fold on cells and Firefly luciferase signals measured; a negative control (non-enveloped, NE) was also included. The data underlying this Figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003944.s008

(TIF)

S9 Fig. Binding of convalescent sera from SARS-CoV-2 infections with sarbecovirus RBDs.

(A) Heatmap of calculated human IgG binding end-point titres and (B) absorbance binding curves for convalescent sera (details in Fig 5A) with SARS-CoV-2, BANAL-20–52, BANAL-20–103, BANAL-20–236, Rs4231, WIV-1, RhGB07 and Rc-o319 RBDs. Boxes with crosses indicate that sample was not tested. Dotted lines represent the limit of detection for the assay (mean + 3SD of no sera control well), with error bars shown for replicate values (mean + SD). The data underlying this Figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003944.s009

(TIF)

S10 Fig. RBD Chimeras of SARS-CoV-2 and Rs4231 show a switch in ACE2 entry phenotype.

(A) Infection of HEK293T cells stably expressing hACE2 with SARS-CoV-2, Rs4231 or chimeric pseudoparticles (Rs4231 + SARS-CoV-2 RBD, SARS-CoV-2 + Rs4231 RBD). Virus was titrated 5-fold on cells and Firefly luciferase signals measured; a negative control (non-enveloped, NE) was also included. (B) Pseudoparticle entry assay using SARS-CoV-2 and Rs4231 chimeras, on a subset of ACE2s that had striking entry profile in the initial screen (human, Rh. sinicus, Rh. shameli, Rh. macrotis, T. melanopogan, mouse, Syrian hamster, pig, raccoon dog and marmoset ACE2). The heatmap shows log(RLU) of psuedotype entry, with pDISPLAY included as a control to establish background luciferase signal. The data underlying this Figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003944.s010

(TIF)

S11 Fig. Binding dendrogram of SARS-CoV-2 monoclonal antibodies analysed by epitope binning.

(A) Dendrogram represents the relatedness of the epitopes of the antibodies. The cut-off for creating antibody communities is shown by the red line at 0.45. (B) Heatmap of antibody blocking and sandwiching interactions. Red boxes indicate blocking interactions, green indicates sandwiching interactions. Self-self blocking is indicated by red boxes with thicker black outlines. Antibodies immobilised as ligands are listed on the left and analytes injected are listed across the top. Where a monoclonal antibody failed to bind as a ligand, it has been excluded.

https://doi.org/10.1371/journal.pbio.3003944.s011

(TIF)

S12 Fig. Binding of sarbceovirus RBDs with monoclonal antibodies isolated from SARS-CoV-2 breakthrough infections.

The (A) WIV-1-RBD (blue)/ VA14_26 complex (green) (WIV-1 PDB 8WLZ [49]) is shown in cartoon representation. Only the VH and VL domains of the Fab of VA14_26 are shown for clarity. Residues of interest on the WIV-1 RBD that make contacts with VA14_26, R396, K366, A372 are coloured in orange. Contact residues on VA14_26 are coloured in purple. The inset shows a zoomed in highlight of this interface. A surface representation of VA14_26 coloured by electrostatic potential highlights that the WIV-1 RBD residues R396 and K366 that are positively charged bind and localise in a highly negative patch of the Fab demonstrating that VA14_26 interactions are mediated by charge attraction. (B) The WIV-1 RBD, along with Rc-o319 (PDB 9W78), SARS-CoV-2 (PDB 7DF3) and RhGB07 (PDB 9RDT), are rendered as cartoon representation or surfaces with electrostatic potential colouring. The residues of interest on the RBD are highlighted on the cartoon representation or of the surface. A critical interacting residue of the VA14_26, E106, is highlighted. This residue is attracted to the positively charged surface of the WIV-1 RBD. Homologous residues present in other RBD, decrease (Rco319) or reverse (RhGB07) this positive charge surface that decreases affinity of the VA14_26 antibody for these RBDs. (C) Each plot shows a single mAb binding, coloured by its designated antibody competition group (Group 1 = purple, Group 2 = orange, Group 3 = red, Group 3.5 = blue, Group 4 = green, Group 8 = non-RBD, NTD-specific) in an ELISA using purified RacCS203 RBD. The curves were calculated using a least-squares regression non-linear fit. Binding (EC₅₀, µg/ml) titres are shown for each mAb against RacCS203, with limits of detection indicated (lower limit: <0.08g/ml upper limit: 5 µg/ml). Red boxes indicate mAbs that exhibit high binding with all sarbecoviruses tested. (C) Representative SPR kinetics binding curves showing the association and dissociation of mAb VA14_26 with bat sarbecovirus RBDs. Each line represents a RBD concentration, starting at 5 µg/ml and titrated 2-fold (8 dilutions). Calculated K_D values are also shown. The data underlying this Figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003944.s012

(TIF)