Design of customized coronavirus receptors

Modular design of CVR

To understand the contribution of different domains of human ACE2 (hACE2) to its receptor function, we first assessed whether we could replace the hACE2 head domain with a computationally designed SARS-CoV-2-targeting miniprotein while maintaining its receptor function. Four hACE2 chimeric proteins were created, with the head domain or the head and neck domains replaced with one of the previously reported miniproteins designated LCB1 and LCB3²⁴. Using authentic SARS-CoV-2 infection assays, we demonstrated that these chimeric proteins effectively supported viral infection (Extended Data Fig. 1a,b).

We found even if we gradually reduce the remaining parts of ACE2 (Fig. 1a), these modified receptors maintained comparable binding efficiency to SARS-CoV-2 RBD (Fig. 1b,c). However, their ability to promote pseudovirus entry decreased concurrently with the shortening of the ACE2 anchor. The shortest 132aa protein maintained detectable receptor function but at a much lower level than that of full-length ACE2 (approximately 0.5%) (Fig. 1d,e and Supplementary Table 1). We subsequently designed modified receptors comprising parts from other viral or immune receptors (R1-R4), along with the addition of a receptor endocytosis-prevention motif (EPM) to enhance surface distribution (R5) (Fig. 1f)²⁵. All chimeric receptors expressed well and bound to RBD efficiently (Fig. 1g–h). Particularly, the chimeric protein with the ACE2 neck domain substituted with triple (3×) 23aa tandem repeats (TR23) from CLEC4M or human IgG Fc (hFc) supported efficient entry (Fig. 1i–j). Further substituting the remaining sequence with IL2Rα corresponding sequences maintained similar entry efficiency, suggesting that ACE2 is entirely replaceable (Fig. 1i,j). Among 31 tested transmembrane (TM) segments and several cytosolic domains from various receptors, the transmembrane and cytosolic domain (TMC) from the matrix remodeling-associated protein 8 (Mxra8) exhibited the best performance (Extended Data Fig. 1c–h and Supplementary Table 2). Constructs with EPM showed improved cell surface localization and improved entry-supporting efficiency (Fig. 1i–j, Extended Data Fig. 1i,j).

Moreover, constructs with a type-II transmembrane topology efficiently supported SARS-CoV-2 and MERS-CoV entry, indicating the feasibility of generating receptors with different transmembrane topologies (Extended Data Fig. 2a,b). We then explored the impact of spacer length and oligomerization on receptor function by testing spacers with various copies of TR23 tandem repeats or immunoglobulin-like domains from human IgG or mCEACAM1a (Fig. 1k–l, and Extended Data Fig. 2c–i). Results indicated the triple TR23 or two immunoglobulin (Ig) or Ig-like domains represent the optimal spacers while abolishing dimerization by Fc mutants has no significant impact on receptor function (Fig. 1m and Extended Data Fig. 2c–i)²⁶.

Different types of SARS-CoV-2 RBD-targeting viral binding domains (VBDs) were tested for receptor grafting, including designed helical frameworks, designed ankyrin repeat proteins (DARPins), nanobodies, scFv, and Fab (Extended Data Fig. 2j,m). We found that all these VBD types were functional, with nanobodies being preferred due to their relatively small size, single-chain nature, and compatibility with bio-panning (Extended Data Fig. 2k–l,n–o). We also demonstrated the functionality of a bi-specific receptor carrying two VBDs of different specificity, and tandem trimeric VBDs (Extended Data Fig. 3a–f). Additionally, we show entry facilitated by soluble receptor adapters connecting viral RBD and ACE2 or FcγRIIa, respectively (Extended Data Fig. 3g–l). In contrast, soluble antibodies without membrane anchoring ability are unable to efficiently promote viral entry, even when viruses are pre-attached to the cells (Extended Data Fig. 3m–o).

The functionality of CVRs compared with ACE2 was further demonstrated through a series of experiments showing membrane fusion, authentic SARS-CoV-2 infection, and virus specificity in different cell types (Extended Data Fig. 4a–e). The entry-promoting efficiencies of CVRs are significantly superior to several documented SARS-CoV-2 alternative receptors, coreceptors, entry factors, or binding proteins (Extended Data Fig.4f)¹⁷.

Together, we proposed a modular design strategy for generating CVRs comprising the customized VBD and modularized artificial receptor scaffolds (ARS) to support efficient coronavirus entry comparable to their native receptors (Fig. 1n).

Mapping of functional CVR epitopes

In our initial exploration of the relationship between CVR receptor function and binding affinity or neutralizing activity, we evaluated 25 neutralizing nanobodies targeting the SARS-CoV-2 RBD²⁷. However, we did not find a clear link between entry-supporting ability and binding affinity or neutralizing activity. These results suggest that there are other important factors at play, especially binding to epitopes that are not clearly defined for these 25 nanobodies (Extended Data Fig.5 and Supplementary Table 2).

Therefore, we engineered CVRs carrying scFvs derived from 22 well-characterized SARS-CoV-2 neutralizing antibodies covering a wide range of neutralizing epitopes (Fig. 2a)^28–35. These antibodies were converted to scFv-based VBDs with N-terminal heavy chain (HL) or N-terminal light chain (LH), resulting in 44 CVRs (Fig. 2b). All CVRs were well-expressed and efficiently bound to the SARS-CoV-2 spike trimer, except for 76E1 which recognizes a conformationally masked epitope exposed after receptor binding (Extended Data Fig. 6a–b)^33,36. The scFv-CVRs recognizing epitopes close to the canonical RBM (sites i, ii, and iii) supported efficient entry³⁷, as was the case for many other RBD core domain-targeting CVRs (Fig. 2c). However, a subset of RBD epitopes were not suitable for CVR design, such as those recognized by S309 and two antibodies binding to a quaternary epitope spanning two neighboring RBDs and locking the spike in a closed conformation (BG10-19, S2M11)^28–30. Unexpectedly, S2L20-CVR, which recognizes an NTD epitope (site iv), promoted robust entry, challenging the previous hypothesis that NTD-recognizing antibodies are insufficient to induce SARS-CoV-2 membrane fusion and entry without ACE2 (Fig. 2c)^22,32. We further demonstrated the expression, antigen binding, pseudovirus entry, and membrane fusion supported by ten selected CVRs (Fig. 2d).

To elucidate the molecular basis of functional CVR design, we hypothesized that CVR function depends on their ability to induce proper conformational changes that lead to down-stream events required for membrane fusion, as previously proposed for the S230 SARS-CoV-1 mAb³⁸, particularly RBD opening and exposure and cleavage of the S₂’ cleavage site for fusion activation (Fig. 2e)^20,33. Consistently, only scFv-mouse IgG Fc (scFv-mFc) corresponding to functional CVRs can induce the exposure of the 76E1 epitope in a dose-dependent manner (Fig. 2e–g, and Extended Data Fig. 6c), concurring with ACE2-induced exposure of the S₂’ cleavage site and fusion peptide³⁶. Furthermore, increased protease accessibility to and processing of the S₂’ cleavage site induced by functional scFv-mFc concurs with the data from 76E1 epitope exposure assays (Fig. 2h and Extended Data Fig.6d).

NTD-mediated entry by S2L20-CVR

We next sought to characterize the NTD-mediated entry promoted by S2L20-CVR. We confirmed that S2L20-CVR acts as a functional receptor for SARS-CoV-2, supporting membrane fusion, pseudovirus entry, and authentic virus infection (Fig. 3a). Additionally, S2L20-CVR effectively facilitated pseudovirus entry of the five SARS-CoV-2 variants of concern (VOCs) tested and of three other sarbecoviruses (BANAL-20-52, RaTG13, and GX_P2V) (Fig. 3b–d and Extended Data Fig.7a). SARS-CoV-1 and ZC45 cannot use S2L20-CVR for entry due to lack of binding³² (Fig. 3c,d).

Despite showing similar NTD-binding efficiency, S2L20 showed relatively lower efficiency in supporting RaTG13 and BANAL-20-52 entry than SARS-CoV-2 (Fig. 3c,d). Since the lack of an N370 glycan has been reported as a distinct feature of SARS-CoV-2, we generated several mutants to investigate the impact of CTD sequences and of the N370 glycan on S2L20-CVR promoted entry³⁹. A SARS-CoV-2 spike harboring either the RaTG13 RBD or an A372T (glycan knock-in) substitution bound more weakly to the soluble forms of hACE2 (sACE2) or S2L20-mFc than SARS-CoV-2 RBD-harboring spikes. We note that RaTG13 has a lower affinity for hACE2 than for an ACE2 orthologue from its host species (Rhinolophus. affinis) (Fig. 3e). We hypothesized that the N370 glycan might sterically interfere with S2L20 binding due to its proximity to the NTD-bound S2L20 Fab fragment. Consistently, the T372A mutation in the RaTG13 or BANAL-20-52 spike, abolishing the N370 glycosylation, significantly enhanced S2L20-CVR mediated entry (Fig. 3f–g and Extended Data Fig. 7b–d)⁴⁰.

We next investigated whether SARS-CoV-2 CTD-targeting neutralizing antibodies could interfere with NTD-mediated entry in cells expressing S2L20-CVR compared to hACE2-expressing cells. As expected, S2L20 exhibited higher neutralizing activity in S2L20-CVR-expressing cells. Importantly, although several antibodies (LY-CoV555, S309, and S2X259) showed reduced neutralizing efficiency in S2L20-CVR expressing cells, some CTD-binding antibodies exhibited similar neutralizing activity in both models (Extended Data Fig. 7e). These data suggest an association between RBD and S2L20-CVR-mediated entry.

A prior study described the CryoEM structure of S2L20 in complex with SARS-CoV-2 BA.5 S in which the three RBDs are in the “up” conformation, whereas the apo BA.5 S structure exhibits the three RBDs in the “down” conformation^41,42. In contrast, other NTD-targeting antibodies, including 4A8, CV3-13, and DH1052, do not appear to induce the RBD “up” conformation (Extended Data Fig. 7f) ^43–45. RBD opening can also be elicited by binding of ACE2, RBM-targeting antibodies and minibinders, or of any antibodies targeting an RBD epitope that is occluded in the closed state^{24,38,46–49}. Therefore, we hypothesized that S2L20-CVR induces RBD opening thereby promoting viral entry, similar to ACE2,. To gain further insights into how this might occur, we determined a CryoEM structure of the S2L20 Fab bound to SARS-CoV-2 Wuhan-Hu-1 S ectodomain with stabilizing HexaPro mutations (Fig. 3h, Extended Data Fig. 7g and Extended Data Table 1). Our structure features one S2L20 Fab bound to each of the three spike NTDs and three open RBDs instead of the previously described mixture of up and down RBD states for the apo spike (Fig. 3h)⁵⁰. However, we previously found that S2L20 can bind to the spike ectodomain when all the RBDs are locked closed due to the presence of the S2M11 Fab that recognizes a quaternary epitope spanning two neighboring RBDs (Fig.3i). Relative to closed S2L20-bound spike structures^51,52, the open S2L20-bound spike structure exhibits a repositioning of the NTDs and bound S2L20 along with stabilization of a conformation of the N234 glycan that would be sterically incompatible with subsequent RBD closing (Fig.3i). Overall, these data support a molecular ratchet mechanism for S2L20-mediated RBD-opening, whereby (i) S2L20 can bind NTDs adjacent to closed RBDs, and when an RBD opens, the NTD and S2L20 shift to a position that blocks subsequent RBD closing or (ii) S2L20 could bind to NTDs adjacent to already open RBDs, thereby preventing them from adopting the closed state. Consistent with this hypothesis, incubation of purified spike ectodomain (without stabilizing proline substitutions) with S2L20 triggered spike fusogenic conformational changes yielding postfusion trimers (Fig.3j, and Extended Data Fig. 7h).

Although the induction of three-RBD “up” conformation upon S2L20 binding appears to be crucial for S2L20-CVR receptor functionality, CEACAM1a is an entry receptor for MHV despite recognizing an NTD surface proximal to that targeted by the 4A8 SARS-CoV-2 antibody, the latter binder does not alter RBD opening nor support viral entry (Extended Data Fig. 7i)^5,43,53. This suggests that different coronaviruses might adopt distinct receptor recognition mechanisms to achieve NTD-mediated entry or that yet undiscovered triggering mechanisms lead to MHV RBD opening and viral entry, as recently discovered for HKU1^12,54,55.

CVR generalizability among coronaviruses

We extended our approach to generating CVRs for 12 coronaviruses from six subgenera, most of them without known receptors (Fig. 4a and Supplementary Table 1)¹. To develop effective VBDs, we used magnetic beads and immunotube-assisted phage display biopanning to screen for coronavirus spike-specific nanobodies from naïve libraries (Extended Data Fig. 8a). Lead CVR nanobodies were identified through validated binding to the RBD or S₁ subunit (Extended Data Fig. 8b and Supplementary Table 3). We also included a broadly-neutralizing nanobody Nb27, which cross-reacts with RsHuB2019A and several other sarbecoviruses (Extended Data Fig. 8c)⁵⁶. We determined the binding kinetics of the selected nanobodies to the 12 coronavirus antigens through biolayer interferometry (BLI) assays (Extended Data Fig. 8d,e). Efficient S-mediated pseudovirus entry and cell-cell membrane fusion were demonstrated in cells stably expressing the indicated CVRs, achieving approximately 10² to 10⁴-fold increase in pseudovirus entry compared with mock controls (Fig. 4b,c). The CVRs carrying EPM exhibited better cell surface localization and promoted even more efficient entry (Extended Data Fig. 8f,g). The HKU1-specific 2D1-CVR exhibited a comparably effective receptor function to TMPRSS2 (Extended Data Fig.8h–k). Moreover, several CVRs designed for 229E and MHV-A59 supported efficient membrane fusion and authentic virus propagation, including the 1B3-CVR targeting the MHV-CTD (Fig. 4d–g).

We next evaluated the CVR-based infection models for neutralizing antibody assessment. We compared the neutralizing activity of sera, collected from COVID-19 convalescents or vaccinated individuals, in HEK293T cells expressing ACE2, LCB1-CVR, and Nb24-CVR²⁷, respectively. We observed comparable overall serum inhibitory profiles testing the three receptors side-by-side, although slight differences were noted with Nb24-CVR, recognizing an RBD epitope distant from the RBM (Extended Data Fig.9a), possibly due to modulation of apparent neutralization potencies resulting from distinct receptor usage or expression levels. This indicates the utility of the CVR-based system for evaluating the effectiveness of humoral immunity, ideally with CVRs recognizing the RBM region. We evaluated the cross-reactivity of several broadly neutralizing β-CoV antibodies against a panel of coronaviruses and found that 76E1 exhibits the greatest neutralization breadth, consistent with the epitope conservation (Extended Data Fig.9b,c). Furthermore, we investigated the potential of the CVR-based infection system for evaluating antivirals targeting different entry steps, including proteolytic cleavage, endosome acidification, and membrane fusion^10,57–59. Comparable inhibitory efficacy was observed when comparing spike-mediated entry via ACE2 and LCB1-CVR (Extended Data Fig.9d). We further tested these inhibitors against SARS-CoV-2, HKU1, HKU3, and HKU5 entry into HEK293T cells mediated by the corresponding CVRs. The overall efficiencies of the entry inhibitory tested were similar among the four viruses except for HKU5 (HKU5-1, LMH03f isolate) which exhibited a higher sensitivity to the TMPRSS2 inhibitor camostat rather than the cathepsin inhibitor E64d (Extended Data Fig.9e). The fusion inhibitor EK1C4⁵⁷, the optimized form of OC43-HR2 lipopeptide, exhibited a broad-spectrum inhibitory effect (Extended Data Fig.9e).

To evaluate CVR-supported multiple-round virus propagation, we created propagation-competent (pc) VSV recombinant viruses by replacing the VSV-G gene with HKU3 or HKU5 S genes, and additionally incorporating a GFP-expressing cassette for visualization (Extended Data Fig.10a). The pcVSV-HKU3 and pcVSV-HKU5 were successfully rescued with the aid of VSV-G provided in trans (Extended Data Fig.10b). Efficient propagation of pcVSV-HKU3 and pcVSV-HKU5 was observed in CVR-expressing cells in a S-dependent manner, as evidenced by syncytia formation with green fluorescence and viral RNA accumulation, which was further enhanced by exogenous trypsin (Extended Data Fig.10c–g). Subsequently, we investigated whether CVRs can promote the propagation of authentic coronaviruses that are difficult to culture. RsHuB2019A is an ACE2-independent bat sarbecoviruses recently isolated from field samples³. Isolation and propagation of this virus were used to be carried out in Huh-7 under a serum-free culture condition with exogenous trypsin, with viral infection being difficult to detect while maintaining normal cell morphology (Fig. 4h). Our results demonstrate that Caco2-Nb27 efficiently supported RsHuB2019A propagation, even at very low multiplicity of infection (MOI) and in trypsin-free culture medium supplemented with 2% FBS, enabling observation of cytopathic effect (CPE) (Fig. 4i,j and Extended Data Fig. 10h,i).

Furthermore, we explored the feasibility of rescuing or isolating authentic HKU5 through CVR-expressing cells. Utilizing a coronavirus reverse genetics approach, we generated a full-length infectious clone of wild-type HKU5 (HKU5-1, LMH03f isolate), along with a fluorescence protein with ORF5 substituted by a ZsGreen-HiBit reporter (ZGH) (Extended Data Fig. 10j). Using the Caco2-1B4 cells, we successfully rescued both versions of HKU5-1 authentic viruses, and isolated another HKU5 strain (PaGD2014/15) from a bat anal swab sample (Fig. 4l, m and Extended Data Fig. 10k–n). Electron microscopy revealed typical morphology of “crown-shaped” HKU5-1 virions (Fig. 4k). Efficient amplification was observed in cells inoculated with HKU5-1 at different MOI, as indicated by nucleocapsid (N) protein immunostaining and the accumulation of viral RNA in the supernatant (Fig. 4l,m). Consistent with previous reports, HKU5 can weakly amplify in Vero E6 cells in the presence of exogenous trypsin (Extended Data Fig.10o)^60,61. Two mutations (T76R and K519T) in HKU5-1 S were detected after ten passages in Caco2-1B4 cells without significant altered biological features of trypsin dependence in infecting Caco2 cells (Extended Data Fig. 10p,q). Lastly, we assessed the activities of several antiviral reagents against HKU5-1 infection in Caco2-1B4 cells. Consistent with the pseudovirus entry assay data (Extended data Fig. 9e), most inhibitors tested blocked HKU5-1 infection, especially for EK1⁵⁹ and EK1C4 peptides (Extended data Fig. 10r–t). However, HKU5-1 infection was inhibited by Camostat but not E64d, in line with its sequence features at the typical protease cleavage sites⁶⁰ (Extended Data Fig. 10r,t,u).

Cell lines

HEK293T (CRL-3216), Vero E6 (CRL-1586), A549 (CCL-185), BHK-21 (CCL-10), Caco2 (HTB-37), Neuro2a (CCL-131) and Tb 1 Lu (CCL-88) were purchased from the American Type Culture Collection (ATCC). Huh-7 (SCSP-526) was obtained from the Cell Bank of Type Culture Collection, Chinese Academy of Sciences. All cells were maintained in Dulbecco’s Modified Eagle Medium (DMEM, Monad), supplemented with 10% fetal bovine serum (FBS), and with or without 1× Penicillin G (100U/ml) /Streptomycin (100 μg/mL) (Pen/Strep). An I1-hybridoma cell line (CRL-2700), producing a neutralizing mouse monoclonal antibody against VSV-G, was cultured in minimum essential medium with Earle’s balanced salts solution, 2 mM L-glutamine (Gibco), and 5% FBS. All cell lines were incubated at 37°C in 5% CO₂ with regular passage every 2-3 days.

Virus and host gene sequences

All viral genome or gene sequences were sourced from GenBank or GISAID databases with the following accession numbers. Viruses: SARS-CoV-1 (NC_004718), SARS-CoV-2 (NC_045512), MERS-CoV (NC_019843), HKU3 (DQ022305), Rp3 (DQ071615), HKU5-1 LMH03f (NC_009020), HKU31 (MK907286), HKU9 (NC_009021), Zhejiang2013 (NC_025217), Rs4081 (KY417143), MHV-A59 (NC_048217), NL63 (JX504050), 229E (OQ920101), HKU1 (NC_006577), OC43 (AY391777), RmYN02 (EPI_ISL_412977), ZC45 (MG772933), RsHuB2019A (OQ503498), GX_P2V (MW532698.1), BANAL-20-52 (MZ937000.1). The spike protein for Rs4075 (KC880993), B.1.1.7/α (EPI-ISL-601443), B.1.351/β (EPI_ISL_678597), P.1/γ (EPI_ISL_906075), B.1.617.2/δ (EPI_ISL_2378732), BA.2 (EPI_ISL_7580387). Receptors: ACE2 (NM_001371415), R.affinis ACE2 (MT394208), DPP4 (NM_001935), APN (NM_001150), mCEACAM1a (NM_001039186), AXL (NM_001699), NRP1 (NM_001024628), SCARB1 (BC143319), KREMEN1 (NM_032045), ASGR1 (NM_001671), CD147 (AB085790), CLEC4M (KJ902090), LRRC15 (NM_001135057)⁶⁴, TMEM106B (NM_018374)¹⁸, TMPRSS2 (NM_001135099). All receptor and viral gene sequences used in this study were commercially synthesized by Genewiz or GenScript. Supplementary Table 1 summarizes the detailed information on these genes.

Plasmids

All plasmids expressing type-I transmembrane CVRs were constructed by inserting the human codon-optimized CVR sequences into a lentiviral transfer vector (pLVX-EF1a-Puro, Genewiz) with an N-terminal CD5 leading sequence (MPMGSLQPLATLYLLGMLVASVL) and C-terminal 3×FLAG tag (DYKDHD-G-DYKDHD-IDYKDDDDK). Type-II transmembrane CVRs were constructed by replacing the C-terminal ectodomains with corresponding CVR modules, along with a N-terminal 3×FLAG tag. Chimeric protein-coding sequences were generated using overlapping PCR, direct sequence synthesis, homologous recombination (Hieff Clone, YEASEN, 10922ES20), or restriction endonuclease digestion and ligation. All modules were connected with two or four amino acids (commonly G, S, T) that were encoded by one or two restriction endonuclease sites. The sequences of CVRs for illustrating the modular design strategy are summarized in Supplementary Table 2.

Plasmids expressing the S glycoproteins from various coronaviruses for VSV pseudotyping were constructed by inserting human codon-optimized S coding sequences into either the pCAGGS or pcDNA3.1(-) vectors with C-terminal 13-18 residues substituted with an HA tag (YPYDVPDYA) to enhance VSV pseudotyping efficiency and facilitate detection⁶⁵. Several spike genes were also introduced into the pLVX-IRES-ZsGreen vectors for flow cytometry-related assays, including the scFv-mFc binding and the 76E1 epitope exposure assays.

Plasmids expressing secreted fusion proteins, including the coronavirus antigens, scFvs, and nanobodies fused with human or mouse Fc, were constructed by inserting the coding sequences into pCAGGS. These constructs featured an N-terminal CD5 secretion leading sequence (MPMGSLQPLATLYLLGMLVASVL) and a C-terminal Twin-Strep Tag II following 3×FLAG tandem sequences (WSHPQFEKGGGSGGGSGGSAWSHPQFEKGGGRSDYKDHDGDYKDHDIDYKDDDDK) for purification or detection. Plasmids encoding codon-optimized anti-ACE2 antibodies h11B11⁶⁶, or SARS-CoV-2 S-targeting neutralizing antibodies B6, S2P6, 76E1, S2H97, S2L20, and REGN10933, LY-CoV555, S309, S2X259, S2M11, BG10-19 were constructed by integrating the heavy-chain and light-chain coding sequences into pCAGGS with an N-terminal CD5 leader sequences, respectively. For DSP-based cell-cell fusion assays, the split protein genes were inserted into pLVX-EF1a-Puro. The coding sequences for the dual reporter split proteins, namely RLuc (1-155)-sfGFP (1-157) and sfGFP (158-231)-RLuc (156-311), are previously described¹³.

Stable cell lines

Cells stably expressing distinct CVRs and other receptors were established using lentivirus transduction and subsequent antibiotic selection. Lentiviruses carrying the target genes were generated by co-transfecting lentiviral transfer plasmid (pLVX-EF1a-Puro) with packaging plasmids pMD2G (Addgene, 12259) and psPAX2 (Addgene, 12260) into HEK293T cells through GeneTwin (Biomed, TG101). The lentivirus-containing supernatant was pooled at 24 and 48 hours post-transfection (hpt). Cell transduction was performed in the presence of 8 μg/mL polybrene. Stable cell lines were selected and maintained in a growth medium supplemented with puromycin (1 μg/mL). Typically, cells that remained stable for at least ten days were used in subsequent experiments.

SARS-CoV-2 reactive polyclonal sera

SARS-CoV-2 polyclonal sera were obtained from vaccinated individuals (SARS-CoV-2 CoronaVac, Sinovac), approximately 21 days post-vaccination and Wuhan COVID-19 convalescents around one-year post-infection, respectively. Ethical approval for the vaccinated individuals was granted by the Ethics Committee (seal) of Beijing Youan Hospital, Capital Medical University, with approval number LL-2021-042-K. The collection of sera from Wuhan COVID-19 convalescents was conducted in collaboration with the Hubei Provincial Center for Disease Control and Prevention and Hubei Provincial Academy of Preventive Medicine (HBCDC), following written consent and under the approval of the Institutional Review Boards with the identification number 2021-012-01. Sera were heat-inactivated at 56°C for 30 minutes.

Recombinant protein expression and purification

Proteins for binding, neutralizing, or biopanning-related assays were produced in HEK293T by transient transfection with plasmids using GeneTwin reagent (Biomed, TG101-01), following the manufacturer’s guidelines. Protein-containing supernatants were harvested every 2-3 days post-transfection, pooled, clarified, and proceeded to purification. Proteins fused with Fc were captured using Pierce Protein A/G Plus Agarose (Thermo Scientific, 20424), eluted with pH 3.0 glycine (100 mM in H₂O), and immediately pH-balanced by 1/10 volume of UltraPure 1 M Tris-HCI, pH 8.0 (Thermo Fisher Scientific, 15568025). Proteins with Twin-Strep Tag II (strep) were enriched using Strep-Tactin XT 4Flow high-capacity resin (IBA, 2-5030-002), washed, and eluted with buffer BXT (100 mM Tris/HCl, pH 8.0, 150 mM NaCl, 1 mM EDTA, 50 mM biotin). The eluted proteins were concentrated and buffer-exchanged to PBS through ultrafiltration, aliquoted, and stored at −80°C. Protein concentrations were determined using the Omni-Easy Instant BCA Protein Assay Kit (Epizyme, ZJ102).

Western blot

For detecting the cellular expression of CVRs or other receptors, cells were washed with PBS and lysed using RIPA buffer (50 mM Tris-pH 7.4, 150 mM NaCl, 1%TritonX-100, 0.5% sodium deoxycholate, 0.1 % SDS, 25 mM β-glycerophosphate, 1 mM EDTA, and 1 mM PMSF) on ice for 15 minutes. The lysate was clarified by centrifugation at 12,000g at 4°C for 15 minutes. The supernatant was combined with a 1:5 (v/v) ratio of 5×SDS-loading buffer and incubated at 95°C for 10 minutes. For detecting the spike packaging efficiency, the PSV-containing supernatant was concentrated with a 30% sucrose cushion (30% sucrose, 15 mM Tris-HCl, 100 mM NaCl, 0.5 mM EDTA) at 20,000×g for 1.5 hours at 4°C. The concentrated virus pellet was resuspended in 1×SDS loading buffer and incubated at 95°C for 30 minutes. For detecting the S2’ cleavage of PSV, the concentrated viruses were resuspended in DMEM in the presence of indicated concentrations of scFv-mFc or soluble ACE2 for 2 hours at 4 °C, and then treated with 10 μg/mL TPCK-trypsin (Sigma-Aldrich, T8820) for 30 minutes at 37°C, followed by mixing with a 1:5 (v/v) ratio 5×SDS-loading buffer and incubated at 95°C for 10 minutes.

After SDS-PAGE electrophoresis and PVDF membrane transfer, blots were blocked with 5% milk in PBS containing 0.1% Tween-20 (PBST) at room temperature for 1 h. Primary antibodies targeting FLAG tag (Sigma-Aldrich, F1804, 1:10,000), HA (MBL, M180-3, 1:5,000), VSV-M [23H12] (Kerafast, EB0011, 1:10,000), β-tubulin (Immunoway, YM3030, 1:10,000) and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (AntGene, ANT325, 1:10,000) were diluted in PBST with 1% milk and incubated with the blot overnight at 4 °C. The stem-helix targeting monoclonal antibody S2P6 for coronavirus spike detection was used at 1 μg/ml. After three washes with PBST, the blots were incubated with horseradish peroxidase (HRP)-conjugated secondary goat antibodies against human (Sigma-Aldrich, A0170), rabbit (Sino Biological, SSA003), or mouse (Jackson Lab, 15-035-003) IgG (1:5,000). After extensive wash, blots were visualized using the LI-COR Odyssey CLx or the Omni-ECL Femto Light Chemiluminescence Kit (EpiZyme, SQ201) and a ChemiDoc MP Imaging System (Bio-Rad). Image Lab (version 5.2.1) was utilized to analyze Western blot data. Uncropped and unprocessed full scans of gel source data are provided in Supplementary Fig. 1 and 2.

Live-cell binding assays

For detecting coronavirus antigens binding to cell surface expressed CVRs, NTD/CTD/S₁-Fc fusion proteins were diluted in DMEM and incubated with cells at 1 to 4 μg/mL for one hour at 37°C. Cells were washed twice with Hanks’ Balanced Salt solution (HBSS) and incubated with 2 μg/mL Alexa Fluor 594 or 488-conjugated goat anti-mouse IgG (Thermo Fisher Scientific; A32742/A32723) for visualization. For detecting the Twin-Strep Tag II labeled S-trimer or soluble ACE2 binding, the incubated cells were treated with 1 μg/mL anti-Twin-Strep Tag II monoclonal antibody (Abbkine; ABT2230) for 30 minutes at 4°C, washed twice with HBSS, and then subjected to fluorescence-labeled secondary antibody incubation. Finally, cells were incubated with Hoechst 33342 (1:5,000 in HBSS) for nuclear staining before imaging using a fluorescence microscope (MI52-N).

Immunofluorescence for detecting receptor expression

Immunofluorescence assays were performed to assess the expression of the CVRs or other receptors carrying the C-terminal 3×FLAG tags. In general, cells expressing the proteins were fixed with 100% methanol at room temperature for 10 minutes, washed once with PBS, and incubated with a mouse monoclonal antibody M2 specific to the FLAG-tag (Sigma-Aldrich, F1804, 1:1000) in 1% BSA/PBS at 37°C for 1 hour. After another wash with PBS, cells were incubated with 2 μg/mL Alexa Fluor 594-conjugated goat anti-mouse IgG (Thermo Fisher Scientific, A32742) diluted in 1% BSA/PBS for 1 hour at 37°C. Nuclei were stained with Hoechst 33342 (1:5,000 dilution in PBS). Images were captured and merged using a fluorescence microscope (Mshot, MI52-N).

Biolayer interferometry

Protein binding kinetics were evaluated through Bio-Layer Interferometry (BLI) assays conducted on the Octet RED96 instrument (Molecular Devices) at 25°C, and shaking at 1000 rpm. Briefly, 20 μg/mL of S₁/NTD/CTD-hFc recombinant proteins were immobilized on protein A (ProA) biosensors (ForteBio, 18-5010). Subsequently, the biosensors were washed and incubated with 2-fold serial-diluted monomeric nanobodies (Twin-Strep Tag II only) in the kinetic buffer (PBST) to record the association kinetics, followed by recording the dissociation kinetics in the same Kinetic buffer. The background was established using a kinetic buffer without the binding proteins. The kinetic parameters and binding affinities were determined using the Octet Data Analysis software (v.12.2.0.20) with global curve fitting using a 1:1 binding model.

VSV pseudovirus entry and inhibition assay

Single-round VSV-based pseudoviruses carrying the coronavirus spikes were produced following a modified version of a well-established protocol⁶⁷. The VSV-ΔG carrying GFP and firefly luciferase (VSV-ΔG-fLuc-GFP) was rescued using a reverse genetics system in the laboratory, along with helper plasmids (pBS-N, pBS-P, pBS-G, and pBS-L) from Karafast. For packaging coronavirus PSV, HEK293T cells were transfected with plasmids overexpressing the spike proteins. At 24-36 hpt, cells were inoculated with 1×10⁶ TCID₅₀/mL VSV-dG-fLuc-GFP for 4 hours at 37°C with 8 μg/mL polybrene. Following two DMEM washes, the culture medium was replenished with DMEM containing 1 μg/mL anti-VSV-G neutralizing antibody (from the I1-mouse hybridoma) to minimize background signals from parental viruses. The TCID₅₀ of the PSV was calculated using the Reed-Muench method^68,69.

For pseudovirus entry or entry inhibition assays, susceptible cells were cultured in 96-well plates at a density of 5×10⁴ cells per well and then incubated with around 1×10⁶ TCID₅₀/mL of pseudovirus (PSV), with 100 μL per well. The incubation allowed for attachment and viral entry with or without the indicated concentrations of antibodies or other inhibitors. In some cases, TPCK-treated trypsin (Sigma-Aldrich, T8802) of indicated concentrations was added to the medium to enhance entry efficiency. At 16-20 hours post-infection (hpi), 40 μL of One-Glo-EX substrate (Promega) was added to the cells and incubated for at least 5 minutes on a plate shaker in the dark. Relative light units (RLU) were determined using the GloMax 20/20 Luminometer (Promega). GFP intensity was analyzed using a fluorescence microscope (Mshot, MI52-N).

Recombinant VSV-CoV propagation assay

For VSV-CoV propagation assays, the propagation competent (pc) recombinant VSVs with VSV-G gene genomically replaced with HKU1, HKU3, or HKU5 spike genes (pcVSV-HKU1, pcVSV-HKU3, and pcVSV-HKU5) were generated by the VSV based reverse genetics system. The vector for the VSV genomes was modified based on pVSV-ΔG-fLuc-GFP, with fLuc replaced by the CoV S genes. In brief, the BHK-21 cells were infected with a recombinant vaccinia virus expressing T7 RNA polymerase (vvT7) for 45 minutes at 37°C (MOI=5). After removing vvT7, the cells were transfected with plasmids containing the pVSV-ΔG-GFP-HKU1/HKU3/HKU5-S vector and helper plasmids from Karafast. The virus-containing supernatant (P0) was collected 48 hpt and amplified in Vero E6 cells with in-trans provided VSV-G to yield P1 viruses. The P1 viruses were further amplified in Caco2-CVRs cells in a VSV-G independent manner and in the presence of anti-VSVG (I1-Hybridoma supernatant), generating P2 viruses that were dependent on the genomically encoded CoV S glycoproteins for amplification, which can be enhanced with the presence of 20 μg/mL TPCK-trypsin in DMEM+2% FBS.

Production of recombinant spike glycoprotein

The prefusion SARS-CoV-2 S ectodomain contains the Wuhan-Hu-1 sequence plus Hexapro stabilizing mutations (F817P, A892P, A899P, A942P, K986P, V987P), S₁/S₂ cleavage site mutations (R682G, R683S, R685S) as well as C-terminal foldon trimerization domain, 3C cleavable HIS- and StrepII- tags⁵⁰. For triggering assays, we used a construct encoding SARS-CoV-2 S G614 with a mu-phosphatase signal peptide placed before S residue Q14, with a mutated S₁/S₂ cleavage site (S₆₈₂GAR₆₈₅), and ending at residue K1211 followed by a TEV cleavage site, a foldon trimerization motif, and an 8× His tag cloned in a pCMV vector^70,71. S trimers were produced in 200 mL cultures of Expi293F Cells (ThermoFisher Scientific) grown in suspension using Expi293 Expression Medium (ThermoFisher Scientific) at 37°C in a humidified 8% CO2 incubator rotating at 130 rpm. Cells grown to a density of 3 million cells per mL were transfected using the ExpiFectamine 293 Transfection Kit (ThermoFisher Scientific) and cultivated for 5 days at which point the supernatant was harvested. S ectodomains were purified from clarified supernatant using a Cobalt affinity column (Cytiva, HiTrap TALON crude), washing with 20 column volumes of 50 mM Tris-HCl pH 8.0, 150 mM NaCl, and 1 mM imidazole, and eluted with 50 mM Tris-HCl pH 8.0, 150 mM NaCl, and 600 mM imidazole. The S ectodomain was then concentrated using a 100 kDa centrifugal filter (Amicon Ultra 0.5 mL centrifugal filters, MilliporeSigma), residual imidazole was washed away by consecutive dilutions in the centrifugal filter unit with 50 mM Tris-HCl pH 8.0 and 150 mM NaCl, and finally concentrated to 1 mg/mL for use immediately after purification.

CryoEM sample preparation and data collection

100 μL of 1 mg/mL prefusion-stabilized SARS-CoV-2 S ectodomain was incubated with 2.2 μL of 67 mg/mL S2L20 Fab for 15 min at 37°C. Most unbound Fab was then washed away with six consecutive dilutions in 400 μL of 50 mM Tris-HCl pH 8.0 and 150 mM NaCl and concentration using a 100 kDa centrifugal filter (Amicon Ultra 0.5 mL centrifugal filters, MilliporeSigma). The complex was concentrated to 3.5 mg/mL and 3 μL was immediately applied onto a freshly glow discharged 2.0/2.0 UltraFoil grid⁷² (200 mesh), plunge frozen using a vitrobot MarkIV (ThermoFisher Scientific) using a blot force of −1 and 6.0 s blot time at 100% humidity and 23°C. Data were acquired using the Leginon software⁷³ to control an FEI Titan Krios transmission electron microscope equipped with a Gatan K3 direct detector and operated at 300 kV with a Gatan Quantum GIF energy filter. The dose rate was adjusted to 3.75 counts/super-resolution pixel/s, and each movie was acquired in 75 frames of 40 ms with a pixel size of 0.843 Å and a defocus range comprised between −0.2 and −2.0 μm.

CryoEM data processing

Movie frame alignment (with a downsampled pixel size of 1.686 Å) was carried out using Warp⁷⁴. Estimation of the microscope contrast-transfer function parameters followed by manual micrograph curation, Blob particle picking, particle extraction (with a box size of 256 pixels²), reference-free 2D classification to generate templates for template-based particle picking (templates and micrographs low-pass filtered to 20 Å resolution during template-based particle picking), particle re-extraction, and then reference-free 2D classification were performed in cryoSPARC⁷⁵. Well-defined particle images ab initio reconstruction followed by non-uniform 3D refinement in CryoSPARC⁷⁶ before subjecting particle images to Bayesian polishing using Relion⁷⁷ during which particles were re-extracted with a box size of 512 Å at a pixel size of 0.843 Å. Reference-free 2D classification and non-uniform refinement in CryoSPARC followed by Bayesian polishing in Relion was repeated. At this stage, 3D classification with 50 iterations each (angular sampling 7.5 for 25 iterations and 1.8 with local search for 25 iterations) was carried out using Relion without imposing symmetry⁷⁸. Next, 42 optics groups were defined based on the beam tilt angle used for data collection, and another round of non-uniform refinement in CryoSPARC was then performed concurrently with defocus refinement and Global CTF refinement, fitting beam-tilt and trefoil⁷⁹. For focused classification of the NTD, particles were symmetry-expanded and 3D classified in Relion without alignment using a mask that encompasses the NTD and the S2L20 VH/VL region. Particles in well-defined 3D classes were then used for local refinement in CryoSPARC. Reported resolutions are based on the gold-standard Fourier shell correlation of 0.143 criterion and Fourier shell correlation curves were corrected for the effects of soft masking by high-resolution noise substitution^80,81.

CryoEM model building and analysis

UCSF Chimera⁸² and Coot⁸³ were used to fit atomic models of S2L20 and SARS-CoV-2 S (PDB 7SOB and 7LXY) into the CryoEM maps. The model was then refined and rebuilt into the map using Coot⁸³, Rosetta^84,85, and ISOLDE⁸⁶. Model validation and analysis used MolProbity⁸⁷, EMRinger⁸⁸, Privateer⁸⁹, and Phenix version 1.21⁹⁰. Figures were generated using UCSF ChimeraX⁹¹.

Triggering assay

For antibody-triggered S refolding, 20 ʼl of 0.21 mg/ml SARS-CoV-2 S G614 ectodomain (without 2P or Hexapro stabilizing mutations) was incubated at 37 ºC for 1 hour with and without 0.21 mg/ml S2H14, S2X28, or S2L20, immediately prior to negative staining. Samples were diluted to 0.01 mg/mL immediately prior to adsorption to glow-discharged carbon-coated copper grids for 30 sec prior to staining with 2% uranyl formate. Micrographs were recorded using the Leginon software⁷³ on a 120 kV FEI Tecnai G2 Spirit with a Gatan Ultrascan 4000 4k x 4k CCD camera at 67,000 nominal magnification. The defocus ranged from -1.0 to -2.0 ʼm and the pixel size was 1.6 Å.

Cell-cell fusion assays

A cell-cell fusion assay based on dual split proteins (DSPs) was performed on HEK293T or Caco2 cells stably expressing the CVRs or the native receptors¹³. Group A cells were transfected (Lipofectamine 2000, Biosharp) with plasmids expressing spike proteins and RLucN (1-155)-sfGFP(1-157), while the group B cells were transfection with plasmids expressing spike proteins (same as in group A) and sfGFP(158-231)-RLuc(156-311). Cells from both groups were trypsinized and co-cultured in a 96-well plate at a density of approximately 1×10⁵ cells per 96-well 12 hpt. After 24 hours, cells were washed once by DMEM to remove FBS. Following incubation of DMEM with or without 100 μg/mL of TPCK-treated trypsin at 37°C for 15 minutes, and subsequent culture in DMEM+10% FBS for 24-36 hours, cell nuclei were stained with Hoechst 33342 (1:5,000 dilution in HBSS) for 30 minutes at 37 °C, and then fluorescence imaging using a fluorescence microscope (MI52-N; Mshot). For the assessment of live-cell luciferase activity after reconstitution of split RlucN, 20 μM of EnduRen live-cell substrate (Promega, E6481) was added to the cells in DMEM and incubated for at least 1 hour before detection using the Varioskan LUX Multi-well Luminometer (Thermo Fisher Scientific).

Flow cytometry analysis

For live cell binding assays analyzing CoVs S₁/RBD-mFc recombinant proteins binding, proteins were diluted in DMEM at the indicated concentrations and then incubated with HEK293T cells expressing the CVRs for 1 hour at 37°C, followed by incubation with Alexa Fluor 594-conjugated goat anti-mouse IgG (Thermo Fisher Scientific; A32742; 1:5000) detecting antigen binding and Alexa Fluor 488-conjugated goat anti-human IgG (Thermo Fisher Scientific; A11013; 1:5000) detecting CVRs (with cell surface hFc) expression. If hFc is absent, the intracellular FLAG tags were utilized to assess the receptor expression. Specifically, cells were washed once with HBSS and fixed with 4% PFA, permeabilized with 0.1% Triton X-100, blocked with 1% BSA/PBS at 4°C for 30 minutes, and subsequently stained with Rabbit anti-Flag tag mAb (CST, 14793S; 1:1000) diluted in 1% BSA/PBS for 1 hour at 4°C. Following extensive washing, the cells were incubated with Alexa Fluor 647-conjugated goat anti-rabbit IgG (Thermo Fisher Scientific; A32733; 1:5000) and Alexa Fluor 488-conjugated goat anti-mouse IgG (Thermo Fisher Scientific; A32723; 1:5000) for 1 hour at 4°C for detecting both FLAG tags and SARS-CoV-2 RBD-mFc binding.

For scFv-mFc binding assays, the SARS2-CoV-2-S IRES-ZsGreen expressing cells were incubated with indicated concentrations of scFv-mFc for 1 hour at 37°C in DMEM/1%BSA, washed once with PBS, and then incubated with Alexa Fluor 594-conjugated goat anti-mouse IgG (Thermo Fisher Scientific; A32742; 1:5000). For 76E1 epitope exposure assays, the SARS2-CoV-2-S IRES-ZsGreen expressing cells were first incubated with scFv-mFc or soluble ACE2 for 1 hour in DMEM/1%BSA, washed once and then incubated with 76E1 (1 μg/mL) at 37°C for another hour in DMEM/1%BSA. Subsequently, the washed cells were incubated with DyLight 594-conjugated goat anti-human IgG (Thermo Fisher Scientific; SA5-10136; 1:5000) to detect 76E1 binding.

After all staining procedures, cells were washed twice with PBS and then analyzed using the CytoFLEX Flow Cytometer (Beckman). In each case, 5,000 cells expressing either receptors or spikes, gated based on FLAG/hFc/ZsGreen-fluorescence and SSC/FSC, were analyzed with the CytoFLEX Flow Cytometer (Beckman). FlowJo (version 10) was employed to analyze the flow cytometry data. The representative gating strategies for each experiment are demonstrated in Supplementary Fig.3.

HKU5 reverse genetics

The full-length cDNA clone of HKU5 (strain: HKU5-1 LMH03f, GenBank: NC_009020) was designed and synthesized as seven (from A to G) contiguous cDNAs flanked by unique class IIS restriction endonuclease site (BsaI or BsmBI) and cloned in pUC57 vector. Class II restriction endonuclease sites AvrII and AscI were introduced to the 5’ terminal of HKU5 A and the 3’ terminal of HKU5 G fragments, respectively. Several silent mutations were included to disrupt naturally occurring restriction cleavage sites. A poly-A (25 repeats) sequence was introduced to the 3’ terminal of the HKU5 G fragment. To assemble the full-length cDNA clone, HKU5 A-G fragments were digested by endonucleases, resolved on 1% agarose gels, purified with a gel extraction kit, extracted with chloroform, and precipitated with isopropyl alcohol. Digested HKU5 A-G inserts, and modified pBaloBAC11 vector were mixed, ligated overnight at 4°C, and transformed into DH10B competent cells. The correct full-length HKU5 cDNA clone was identified and verified by sequencing. The construction of HKU5-ZGH utilized the transformation-associated recombination (TAR) cloning technique. Specifically, a ZsGreen-HiBit (ZGH) DNA fragment was commercially synthesized (Tsingke) to replace the HKU5-ORF5. The PCC1 vector was used to clone the HKU5 genomic DNA carrying the ZGH substitution based on three segments amplified using the HKU5-WT infectious clone as a temperate. Subsequently, all the products were transformed into yeast using the high-efficiency lithium acetate/SS carrier DNA/PEG method. The yeast plasmid was extracted and transformed into EPI300 electrocompetent cells. The plasmid used for cell transfection was obtained from a 300 mL E. coli bacterial culture suspension. For transfection, 4 μg of both HKU5 WT and HKU5-ZGH plasmids were separately transfected into Caco2-1B4 cells (1×10⁶ cells) using GeneTwin (Biomed, TG101). Progeny viruses collected from the supernatant at 72 hpt (P0) were utilized to generate stocks for subsequent analyses.

HKU5 sampling and isolation

Bats were trapped in their natural habitat as previously described^92,93. An anal swab (Sample ID: NL140575) with confirmed HKU5 genomic sequence (strain HKU5_PaGD2014/15) was collected from a Pipistrellus abramus bat in Guangzhou City, Guangdong province on June 28, 2014. The sample was preserved in viral transport medium (VTM) comprising Hank’s balanced salt solution, BSA (1%), amphotericin (15 μg/mL), penicillin G (100 U/mL), and streptomycin (50 μg/mL)⁹⁴, transported to the laboratory and stored at −80°C until use. All bats trapped for this study were released back to their habitat after anal swab collection. All sampling processes were carried out by veterinarians with approval from the Animal Ethics Committee of the Wuhan Institute of Virology (WIVH05210201).

To isolate the HKU5, the sample preserved in VTM was centrifuged at 10,000g at 4°C for 10 minutes. Then, 100μl of the 0.45 μm-filtered sample was mixed with 100 μL DMEM + 2% FBS + 2× Pen/Strep and incubated with Caco2-1B4 cells at 37°C for 2 hours. This was followed by supplementation of 400 μL fresh DMEM + 2% FBS + 2× Pen/Strep. The cytopathic effect was monitored daily, and the supernatant (passage 1, P1) was collected on day 5 post-infection. Subsequently, 200 μL of the P1 supernatant was incubated with freshly plated Caco2-1B4 cells at 37°C for 1 hour. The inocula were then removed and replenished with 400 μL fresh DMEM with 2% FBS. The HKU5-containing supernatant (passage 2, P2) was collected on day 3 and day 6 post-infection and stored at −80°C. The Caco2 or Caco2-1B4 cells were challenged with P2 supernatant (collected on day 6) and fixed with methanol for 40 minutes at room temperature on day 3 post-infection, and HKU5 N proteins were detected by rabbit anti-HKU5 N protein serum (diluted at 1:4000), followed by Cy3-conjugated goat anti-rabbit IgG (Abcam, 1:250) staining. All HKU5 isolation-related experiments were conducted in the certified negative-pressure Biosafety Level 2 laboratory at the Wuhan Institute of Virology.

Transmission electron microscopy

Viral culture supernatant was fixed with formaldehyde (working concentration 0.1%) at 4°C overnight. Subsequently, it was concentrated by ultracentrifugation through OptiPrepTM Density Gradient Medium (D1556) at 154,000 g at 4°C for 2.5 hours using a SW41Ti rotor (Beckman). The pelleted viral particles were suspended in 100 μL of PBS, stained with 2% phosphotungstic acid (pH 7.0), and examined using a transmission electron microscope (Thermo Fisher, Talos L120C) at 120 kV.

Authentic coronavirus infection assays

Human coronavirus 229E (VR-740) is obtained from ATCC and amplified in Huh-7 cells. MHV-A59 is a gift from Professor Yu Chen’s lab (Wuhan University) and is amplified in Neuro2a cells. The SARS-CoV-2-(ΔN-GFP) with N protein substituted with GFP is rescued using an established protocol, and cultured in Caco2 cells overexpressing the SARS-CoV-2 N protein⁹⁵. All experiments involving RsHuB2019A, and HKU5-related authentic viruses infection were conducted in the certified negative-pressure Biosafety Level 2 laboratory at Wuhan Institute of Virology. RsHuB2019A is amplified in either Huh-7 or in Caco2-Nb27 cells. HKU5-related strains are amplified in Caco2-1B4 cells.

For authentic virus infection assays, target cells were seeded in 24-well plates, washed once with DMEM, and then inoculated with the indicated MOI of authentic viruses for 1 hour. Subsequently, the infected cells were washed with DMEM once and replenished with DMEM with or without 2% FBS. RsHuB2019A infection was conducted in Huh-7 cells in DMEM (must be FBS-free) with 100 μg/mL trypsin (Sigma-Aldrich, T4549). RsHuB2019A or HKU5-1 infection in Caco2-CVR cells was conducted with or without the presence of 100-200 μg/mL trypsin (in DMEM+2% FBS) or 5-10 μg/mL trypsin (in DMEM without FBS) during and post-inoculation. For qRT-PCR analysis of viral RNA accumulation, cell-free supernatants (50 μL per well each time) were collected at indicated time points post-infection and stored at −80°C. Viral RNA was extracted using Virus DNA/RNA Extraction Kit (Vazyme: RM501) and subjected to qRT-PCR as previously described⁹⁶. qRT-PCR was conducted using the C1000 Touch Thermal Cycler/CFX96 Real-time system (Bio-Red) and analyzed with Bio-Rad CFX Maestro (4.1.2433.1219) software. Primers for RsHuB2019A RdRp: 5’-TTGTTCTTGCTCGCAAACATA-3’ (forward) and 5’-CACACATGACCATCTCACTTAA-3’ (reverse). Primer for HKU5 nsp2: 5’-CTGCGCTTAATGCCCCATTC-3’ (forward) and 5’-GACGTGTAGACGTAGAGCCG-3’ (reverse). Primers for VSV L protein, forward primer: 5’-TCTTGAGTTGTGGAGACGGC-3’ (forward) and 5’- ACCGTCTTGAACATGGGACC-3’ (reverse). Primers for MHV-A59 N protein:5’-TATAAGAGTGATTGGCGTCC-3’(forward) and 5’-GAGTAATGGGGAACCACACT-3’ (reverse). All samples were analyzed in duplicate on two independent runs.

For immunofluorescence assays, cells were fixed with 4% PFA for at least 20 minutes at 25°C for 24-48 hpi at specified time points. N proteins for RsHuB2019A and HKU5 were detected using rabbit anti-Rp3 N protein serum (1:2000) and rabbit anti-HKU5 N protein serum (1:4000), respectively, followed by Cy3-conjugated goat anti-rabbit IgG (Abcam, 1:250) staining. Images were captured using an EVOS M5000 fluorescence microscope (Thermo Fisher Scientific). For MHV-A59 and 229E-VR740, spike proteins were detected using their spike-targeting nanobodies (1A1-mFc for 229E-VR740; 1F7-mFc for MHV-A59), followed by DL594-conjugated goat anti-mouse IgG antibodies (Thermo Fisher, 1:1000). Images were captured and merged using a fluorescence microscope (Mshot, MI52-N). The titer of the authentic viruses was determined by immunofluorescence at 4 days post-infection, and the TCID₅₀ was calculated using the Reed-Muench method^68,69.

Nanobody bio-panning

Specific viral antigens (30-100 μg) were immobilized on streptavidin-conjugated magnetic beads for one-hour incubation at 37°C and extensively washed to remove unbound antigens. Subsequently, the beads were incubated with the nanobody library (1×10¹⁰ PFU) (Naïve VHH libraries from Camelus bactrianus, Alpaca, and Llama from NBbiolab, China) for 1 hour. The bound phages were eluted using an Elution Buffer (50 mM Tris-pH 7.4, 150 mM NaCl, 50 mM biotin) after extensive washing with PBST to eliminate nonspecific binders. The eluted phage encoding the specific nanobodies was proliferated in E. coli (TG1). After one round of magnetic beads-based selection, 1-3 additional rounds of phage biopanning were conducted using magnetic beads or immunotubes. The positive clones were identified through the enzyme-linked immunosorbent assays (ELISA), sequenced, and verified by cell-based binding assays. Nanobody information was summarized in Supplementary Table 3.

Biosafety statement

Experiments related to authentic human coronavirus 229E, SARS-CoV-2-ΔN, and murine MHV-A59 were authorized by the Biosafety Committee of the State Key Laboratory of Virology, Wuhan University and conducted in accordance with standard operating procedures (SOPs) in a BSL-2 laboratory. SARS-CoV-2 authentic viruses-related experiments were conducted in the ABSL-3 facility at Wuhan University with the approval of the Biosafety Committee of the ABSL-3 laboratory. The SARS-CoV-2 WT strain (IVCAS 6.7512) was provided by the National Virus Resource, Wuhan Institute of Virology, Chinese Academy of Sciences and amplified in Vero E6 cells in the ABSL-3 facility at Wuhan University. Experiments related to authentic viruses RsHuB2019A, HKU5-WT, and HKU5-ZGH were approved by the Wuhan Institute of Virology (WIV) IBCs and performed in the BSL-2 laboratory according to SOPs at WIV facilities. In vitro experiments related to RsHuB2019A infection, HKU5-1 (LMH03f) rescue, and HKU5 (PaGD2014/15 strains) isolation were performed in the BSL-2 laboratory following SOPs with necessary personal protection and approved by the Wuhan Institute of Virology (WIV) IBCs according to a biorisk assessment procedure. In vitro experiments involving these bat coronaviruses were assessed as BSL-2 level primarily based on their narrow cell tropism and limited efficiency in using human receptors. The complete assessment includes the viral genome sequence and its phylogenetic relationship to known human viruses, human receptor utilization, cell tropism, prevalence in domestic animals and humans, and pathogenicity and transmission in animal models. All the facilities at both Wuhan University and WIV for this work adhere strictly to the safety requirements recommended by the China National Accreditation Service for Conformity Assessment.

We recommend that future research activities involving the use of CVR-expressing cells to isolate, rescue, or propagate uncharacterised authentic viruses should be authorized first and strictly adhere to biosafety requirements based on the biorisk assessment conducted by the institutional Biosafety Committee on a case-by-case basis, in accordance with the biosafety regulations of different countries. We recommend experiments based on pseudoviruses or replication-deficient viruses be conducted first whenever possible to gather essential data for biorisk assessment before isolating or rescuing authentic viruses. Considering the potential mutations during the propagation of viruses in human cells, biological feature changes during passages should be closely monitored, and viral genome sequencing should be conducted to check whether key adaptive mutations occurred. Furthermore, non-human cells expressing CVR can also be used for virus propagation to minimize the risk of viruses adapting to human cells. Large-scale viral propagation of the newly isolated/rescued uncharacterized viruses and initial virus characterization in animal models are suggested to be performed in a BSL-3 containment. The biorisk classification and containment levels of previously uncharacterized viruses should be re-assessed based on newly acquired data related to the biological and epidemiological characterizations to comply with the biosafety regulations of the country.

Bioinformatic and computational analyses

Multi-sequence alignment was analyzed by Mafft (v7.313) and Geneious Prime (2023.0.1) with default parameters. Phylogenetic trees were constructed by IQ-TREE (v2.0.6) with the WAG substitution model (1,000 Bootstraps) and rendered with iTOL (v6) (http://itol.embl.de).

Statistical analysis and reproducibility

Most experiments were repeated independently 2-4 times with similar results, each with approximately 2-4 biologically independent replicates (most are n=3). Results are presented as mean ± standard deviation (s.d.) as specified in the figure legends. Statistical analyses were primarily performed using GraphPad Prism (Version 8.0.2) through unpaired two-tailed Student’s t-tests for two independent groups or One-way Analysis of variance (ANOVA) analysis followed by Dunnett’s test for multiple comparisons. No adjustments were made for all multiple comparisons. P<0.05 was considered statistically significant; *P<0.05; **P<0.01; ***P<0.001; ****P< 0.0001; NS, non-significant, P>0.05.