Introduction

Three zoonotic spillover events of highly pathogenic coronaviruses have occurred in the past 20 years (SARS-CoV, MERS-CoV and SARS-CoV-2), highlighting the significant and continued threat that coronaviruses pose to human health. SARS-CoV-2 has now joined four other endemic human coronaviruses (HCoV-HKU1, -OC43, -NL63, -229E), that generally cause mild upper respiratory tract infections in a seasonally recurrent manner1,2. Coronaviruses that infect humans can be divided into an alpha genus (alphacoronaviruses: NL63, 229E) and beta genus (betacoronaviruses: HKU1, OC43, SARS-CoV, MERS-CoV, SARS-CoV-2). Infection with HCoVs and seroconversion is typically observed in early childhood3, with periodic re-infections seen throughout life, sometimes as often as 9 to 12 months following the last infection1. NL63 is of particular interest given it binds to the same cellular receptor (ACE2) as highly pathogenic SARS-CoV and SARS-CoV-2 pandemic viruses4, though NL63 generally causes mild respiratory illness. First discovered in 2004 in a 7-month-old baby in the Netherlands5, NL63 is thought to have been circulating in humans for many centuries, with a likely zoonotic bat origin6. Whether NL63 caused more severe infection when it first crossed into humans is unknown, though emergence of a subgenotype associated with severe lower respiratory tract infections in children was reported in China in 20187. The capability to cause widespread infection, coupled with evidence that low pathogenicity is not a given, suggests alphacoronaviruses like NL63 constitute a pandemic threat that warrants the development of effective countermeasures.

The coronavirus spike protein mediates entry into host cells and is the major target for vaccines and therapeutics. The receptor binding domain (RBD) represents the major neutralising target for SARS-CoV-28,9, though neutralising antibodies against the N terminal domain (NTD) and S2 domains have been reported10,11,12. Sites vulnerable to antibody-mediated neutralisation are poorly defined for seasonal coronaviruses such as NL63. Notably, there are significant structural differences between the spike protein of NL63 and other coronaviruses. NL63 RBD (domain S1B) is locked in a “down” conformation in the prefusion state of spike and has two N terminal domains (S10 and S1A), which is a shared feature of several alphacoronaviruses13. Interestingly, NL63 spike is unable to readily bind to ACE2 in the prefusion state as the residues directly interacting with ACE2 within the RBD are occluded through interactions with the NTD of the same protomer13. It is thought that NTD interactions with heparan sulfate proteoglycans on the cell surface are requisite to enable RBD to subsequently bind ACE2.

While our knowledge base for developing vaccines and antiviral therapeutics against SARS-CoV-2 and related pathogenic betacoronaviruses has rapidly expanded, there remains limited research on alphacoronaviruses. Here we characterised the humoral response to NL63 spike and isolated a panel of NL63 spike-specific human monoclonal antibodies (mAbs). We found neutralising mAbs directed against the receptor binding motif within S1B that neutralised NL63 by blocking receptor engagement, and a neutralising mAb against the membrane proximal heptad repeat 2 (HR2) region of S2. Despite not blocking receptor engagement, the S2 mAb neutralised NL63 at similar potencies to S1B mAbs and targeted a region of S2 unique to alphacoronaviruses. Our data provide insights into protective antibody immunity against NL63 and potential avenues for further vaccine and therapeutic mAb development against alphacoronaviruses.

Results

Humoral immune responses to NL63 spike in a healthy human cohort

We recruited a cohort of 14 healthy adults in early 2022 and assessed antibody and B cell responses to NL63 spike. Consistent with reports of high seropositivity in adults14,15,16, we detected plasma IgG binding full length spike and domain S1B (RBD) in all subjects, with endpoint titres ranging from 1:678 to 1:6887 for S and 1:2506 to 1:8299 for S1B (Fig. 1a). Neutralisation activity in plasma was assessed using a live virus assay (Amsterdam1 isolate), with IC50 titres ranging from 1:29 to 1:745 (Fig. 1a). Next, we used fluorescently labelled spike probes to flow cytometrically measure memory B cells (MBCs; CD19+IgD-IgG + ) specific for NL63, and in comparison, for SARS-CoV-2 spike since all subjects had received prior COVID vaccination (Fig. 1a, b; gating in Supplementary Fig. 1). MBCs against NL63 spike were clearly detectable but found at frequencies significantly lower ( ~ 21 fold) than those against SARS-CoV-2 spike, likely indicative of more recent exposure to SARS-CoV-2 infection and/or immunisation (Fig. 1c).

Fig. 1: Humoral responses against NL63 spike in healthy subjects.
figure 1

a The table depicts plasma IgG endpoint titres against NL63 full length spike and the S1B domain (RBD), neutralising titres against the Amsterdam1 live virus isolate in HAT24 cells, and frequencies of NL63 spike-specific memory B cells (MBC). Subjects are ranked in order of NL63 spike-specific MBC frequencies. The two subjects that mAbs were isolated from are highlighted in yellow. The ELISA and neutralisation data were from two technical replicates. b Representative flow cytometry plots of NL63 and SARS-CoV-2 spike-specific memory B cells from two subjects stained with recombinant spike proteins fluorescently labelled with PE or APC. c Frequencies of NL63 and SARS-CoV-2 spike-specific memory B cells as a proportion of CD19+ IgD- IgG+ B cells in PBMCs (N = 14). The line and error bars depict median and interquartile range respectively. Statistical analyses between matched pairs were performed with a Wilcoxon signed-rank test (***P < 0.001). Results in this figure are representative data from single experiments.

Full size image

Isolation of monoclonal antibodies against NL63 spike

The resolved structure of NL63 spike17 with subdomains highlighted is illustrated in Fig. 2a. The S1 subunit consists of S10 + S1A (NTD), S1B (RBD), S1C and S1D while the S2 subunit contains the viral fusion machinery and transmembrane domains. From two subjects (COR344 and COR004) with high frequencies of spike-specific memory B cells and potent plasma neutralisation titres, we sorted NL63 spike or NL63 S1B-specific single MBCs using fluorescent probes. A total of 97 heavy chain and 44 light chain sequences were recovered using multiplex RT-PCR, with selected heavy and light chain pairs chosen for expression as mAbs (denoted with prefix NLH) in mammalian cell culture. The V allele usage, germline identity and CDR3 sequences of NL63 mAbs are shown in Fig. 2b, with full sequences in Supplementary Table 1. Two mAbs (NLH18 and NH28) were derived from S1B-specific MBCs while the rest were derived from MBCs recognising full length spike. Reactivity for NL63 spike and subdomains was assessed by ELISA. Eight mAbs bound full length spike and five were specific for domain S1B (Fig. 2c, binding curves in Supplementary Fig. 2A, B), with only one mAb (NLH28) not binding to spike but binding to domain S1B weakly (EC50 of 8.9 µg/ml). While domain S1B is usually locked in a “down” conformation in the prefusion state of spike with the ACE2-interacting surface occluded13, negative stain electron microscopy revealed a heterogeneous mixture of trimeric and non-trimeric species of NL63 spike (Supplementary Fig. 3). As such, epitopes within domain S1B that are usually obscured could have been available for mAb and MBC binding. NLH03 displayed weak binding to full length spike and the NTD (S10 and S10 + S1A) (Fig. 2c, Supplementary Fig. 2C, D). NLH02, NLH05 and NLH36 bound to full length spike strongly (EC50 of 18.9–23 ng/ml), but did not display any reactivity for the subdomains S10, S1A or S1B, indicating they might bind to an epitope within the S2 domain. To examine binding breadth against other coronavirus spike proteins, we tested binding against a recent pathogenic variant of NL63 (ChinaGD027), alphacoronaviruses (HCoV-229E and swine acute diarrhoea syndrome (SADS-CoV)) and the betacoronavirus SARS-CoV-2 (Fig. 2c, Supplementary Fig. 2E–H). All spike-binding mAbs bound to the ChinaGD02 NL63 strain with equivalent potency. No mAbs apart from the NTD mAb NLH03 displayed cross-reactivity to the other spike proteins tested, though NLH03 bound weakly to 229E and SARS-CoV-2 spikes with EC50 values of 2 µg/ml and 1.6 µg/ml respectively (Supplementary Fig. 2G).

Fig. 2: Monoclonal antibodies against NL63 spike.
figure 2

a Organisation of the NL63 spike protein. Upper part: linear depiction of NL63 spike genome (not to scale). The S1 subunit consists of domains 0, A, B, C and D. Lower part: cryo-ET structure of the trimeric NL63 spike (PDB: 8FR7) with the subdomains of one protomer shown in coloured ribbons. The other two protomers are shown in grey. b Germline identity and CDR3 sequences of 9 NL63 spike-specific human mAbs. c Binding and neutralisation specificity of NL63 mAbs. Binding of mAbs was assessed via ELISA against full length spike and subdomains S1B, S10, S10 + S1A from NL63. Breadth was assessed against spike proteins from the ChinaGD02 NL63 strain, 229E, SADS-CoV (swine acute diarrhoea syndrome coronavirus) and SARS-CoV-2. Receptor binding inhibition was assessed via a S1B (RBD)-ACE2 inhibition ELISA while neutralisation was assessed using live NL63 virus isolates (Amsterdam1 and contemporary isolates) in HAT24 cells. ELISA binding and neutralisation titration curves are provided in Supplementary Fig. 2 and 4. Dashes indicate no binding or neutralising activity. The receptor binding inhibition ELISA and neutralisation assays were performed in duplicate and are representative data from single experiments.

Full size image

Neutralising activity against NL63

Spike-binding mAbs were screened for neutralising activity against live NL63 (Amsterdam1 isolate) (Fig. 2c, neutralisation curves in Supplementary Fig. 4A). Five mAbs showed neutralisation activity with IC50 values ranging from 4.9 ng/ml to 448.9 ng/ml, with four mAbs binding to S1B (NLH04, NLH06, NLH07, NLH18) and one potently neutralising mAb recognising a non-RBD and non-NTD epitope within spike (NLH02, IC50 of 8 ng/ml). Neutralisation activity was assessed against more contemporary NL63 isolates collected between 2018-2019, since the Amsterdam1 isolate was isolated in 20045. NL63 clinical isolates were first grown in primary human nasal epithelial cell cultures18 before further passage in LLC-AT cells (LLC-MK2 cells overexpressing ACE2 and TMPRSS2; Supplementary Fig. 5A). These primary NL63 isolates (18111908, 18091206, 19071116) grew to lower titres compared to Amsterdam1 but still caused complete cytopathic effect (CPE) in 293 T cells overexpressing ACE2 and TMPRSS2 (Supplementary Fig. 5B). All five neutralising mAbs maintained neutralising activity against the contemporary NL63 isolates, with IC50 values largely comparable to those against Amsterdam1 (Fig. 2c, Supplementary Fig. 4A). The four S1B-directed neutralising mAbs were able to block recombinant ACE2 from binding to S1B in vitro (Fig. 2c, Supplementary Fig. 4B), suggesting viral neutralisation is achieved via direct blocking of cellular receptor recognition. mAb NLH02 was able to neutralise all NL63 isolates at comparably high potencies (IC50 of 4.3–11.7 ng/ml) despite not blocking ACE2 engagement.

Mapping neutralising epitopes on NL63 spike

To better understand epitopes targeted by NL63 neutralising antibodies, we generated escape variants via serial passaging of NL63 (Amsterdam1 strain) in the presence of NLH02, which only bound to full length spike, and the three most potent S1B neutralising mAbs (NLH04, NLH07 and NLH18). Viral supernatants were recovered after each passage and sequenced, with a bona fide escape variant defined as a variant that (a) is absent in baseline passages, (b) imparts a protein coding change, and (c) increases in frequency under selection with increasing concentrations of mAb. For all mAbs, escape variants were generated within 3-9 passages (Fig. 3a). Mutations conferring full neutralisation escape were E1256A and Q1246L for NLH02 and K501N for NLH04. For NLH07, two mutations were fixed at passage 3 (S10: N115S; S1B: S535L) that conferred partial neutralisation escape, while the S496F mutation in S1B that appeared at passage 8 conferred full escape. For NLH18, a mutation in S1B outside of the receptor binding motif (E582V) became fixed at passage 6, though full neutralisation escape was only achieved when H586Y (in the receptor binding motif) became fixed in passage 9.

Fig. 3: Viral escape from NL63 neutralising mAbs.
figure 3

a Antibody escape viruses were generated by serial passaging of NL63 (Amsterdam1 strain) in HAT24 cells in the presence of neutralising mAbs (NLH02, NLH04, NLH07, NLH18). At each passage, the wells with detectable cytopathic effect (CPE) at the highest mAb concentration were harvested and re-passaged in a titration of mAbs starting at a higher concentration. Serial passaging was repeated until escape occurred (presence of CPE) at a mAb concentration of 50 µg/ml. Whole genome sequencing was performed on viral RNA extracted from cell supernatants. Amino acid mutations within spike that become fixed are indicated when they first appear. b mAb escape mutations mapped onto NL63 spike (PDB: 8FR7) with the top view of spike depicted on the left and side view depicted on the right. S1B (RBD) domains of each spike protomer are coloured. Escape mutations for NLH04, NLH07 and NLH18 are within domain S1B (RBD) while the escape mutations for NLH02 are within the S2 domain.

Full size image

Amino acid substitutions conferring antibody resistance were mapped onto the cryoET structure17 of NL63 spike trimer (Fig. 3b). Escape mutations for the S1B neutralising mAbs were concentrated on the three discontinuous receptor binding motif loops of NL63 RBD (Fig. 2c, Supplementary Fig. 4A). The mutations for NLH07 (S496F, S535L) and NLH18 (H586Y) are of residues that make direct contacts with ACE219. In contrast, the non-S1B NLH02 mAb drove the emergence of E1256A and Q1264L mutations within the membrane proximal HR2 region of S2, suggesting a neutralisation mechanism via interference of viral fusion machinery, as described previously for SARS-CoV20 and SARS-CoV-221.

We next examined amino acid conservation of NL63 spike in over 200 genomic sequences of viral isolates ranging from 1983 to 2023. We found limited divergence in NL63 spike as previously reported22, with most sequence variability concentrated within domains S10 and S1A (Fig. 4A; PDB: 8FR717). Domain S1B (RBD) is relatively conserved, with areas of variability concentrated on the exterior surface of S1B, distal from the receptor binding motif that interacts with cellular receptor ACE2 (Fig. 4B; PDB: 3KBH19). Viral mutations we identified that conferred escape from neutralising antibodies were not observed within available genomic sequences, with strong sequence conservation of these and most surrounding residues within putative binding sites (Fig. 4C). The S2 subunit of coronaviruses is more highly conserved, with several pan-coronavirus or pan-betacoronavirus neutralising S2 mAbs reported23,24,25,26. We contrasted conservation of S2 region of NL63 with other coronaviruses, finding that the HR1 and HR2 regions of alphacoronaviruses are generally extended by 14 amino acids compared to that of betacoronaviruses (Supplementary Fig. 6), with the exception of SADS-CoV and the related bat HKU2 virus. While there are a few residues conserved across the alphacoronavirus sequences, the putative epitope of NLH02 in HR2 is not very well conserved and is absent in betacoronaviruses, in line with our finding that NLH02 did not cross-react with 229E, SADS-CoV or SARS-CoV-2 spikes.

Fig. 4: Amino acid conservation within NL63 spike.
figure 4

A Amino acid conservation across NL63 strains (N = 207 sequences spanning 1983—2023 from Genbank) is illustrated on a monomer of NL63 spike (PDB: 8FR7). Shading indicates areas of low ( < 80%; deep red), partial ( > 80%; salmon), near complete ( > 0.98%; light pink) and complete conservation (100%; grey). B Amino acid conservation depicted on NL63 domain S1B (RBD) in complex with the ACE2 receptor (PDB: 3KBH). C Amino acid sequence conservation of the residues around the escape mutations (in black boxes) for the S2 neutralising mAb (NLH02) and three RBD neutralising mAbs (NLH04, NLH07 and NLH18). Residues are shaded according to how well conserved they are with orange/red indicating lower conservation.

Full size image

Discussion

In line with past reports15,16, we observed universal seropositivity against NL63 in a healthy adult population. Serological neutralisation activity and detectable populations of memory B cells against NL63 spike suggest a high degree of pre-existing immunity within adults. Interestingly, despite this prior immunity, NL63 does not undergo significant adaptive evolution in comparison to HCoVs 229E and OC4322,27, which have much higher rates of sequence changes concentrated in the S1 subunit (containing the RBD). SARS-CoV-2, which uses the same receptor as NL63, also undergoes significant genetic drift and escape from neutralising antibodies22,28. The high conservation of the NL63 spike suggests replication and transmission occurs in a setting of low immune pressure. However, it remains unclear whether this is due to low concentrations of spike-specific mucosal antibodies in the upper respiratory tract, or instead reflects virological constraints of NL63 itself, for example, the relatively locked down RBD13. Nevertheless, previous studies have demonstrated protective immunity against all HCoVs is relatively short-lived1,29, with frequent re-infections in adults, suggesting immunity elicited by prior infection does little to curb spread in human populations.

To better understand antibody recognition of NL63, we isolated a panel of human monoclonal antibodies targeting NL63 spike. Five of these were neutralising, with four recognising the receptor binding domain S1B and one recognising the membrane proximal HR2 region of S2. Similar to SARS-CoV-2 where RBD is the predominant neutralisation target8,9, we find that the majority of our neutralising mAbs targeted the RBD and blocked ACE2 engagement in vitro. While we were unable to isolate neutralising mAbs against the NTD of NL63 (S10 and S1A), neutralisation supersites within the NTD have been reported for SARS-CoV-210,12, OC4330, and 229E31. A previous study of alphacoronavirus 229E found NTD antibodies displayed a degree of cross-reactivity to NL63, although any cross-neutralising activity was not assessed31. There have been multiple reports of pan-coronavirus or pan-betacoronavirus mAbs directed to conserved elements within the S2 domain23,24,25,26. For example, mAb 76E1 recognises a highly conserved epitope between the S2’ site and fusion peptide and could neutralise six human coronaviruses with IC50 values ranging from 0.4–4.8 µg/ml26. However, while S2 mAbs for betacoronaviruses to date have tended to display less potent neutralisation activity than those directed at the RBD, we find the S2 mAb NLH02 was unusually potent (4.3–11.7 ng/ml), comparable to highly potent NL63 RBD mAbs. Since alphacoronaviruses have an extended HR1 and HR2 region compared to betacoronaviruses (by 14 amino acids32), this region might constitute a unique site of vulnerability for alphacoronavirus neutralisation. During viral entry, the HR1 and HR2 regions of class I fusion proteins form a six-helix bundle that drives fusion of viral and cellular membranes. Antibodies that disrupt this fusion process have been described for SARS-CoV20, SARS-CoV-221,33, HIV-134 and Ebola virus35, and it is likely NLH02 neutralises viral entry in an analogous manner. However, the putative HR2 epitope of NLH02 is not highly conserved amongst alphacoronaviruses and is unlikely to constitute a pan-alphacoronavirus neutralising epitope. Future studies should investigate whether the extended HR1 and HR2 regions are potent neutralisation targets for other alphacoronaviruses.

Upon serial passage in the presence of neutralising mAbs, we find that NL63 can readily generate escape variants that replicate to fixation. However, further experiments are required to confirm whether these escape mutations have any impact on NL63 replication. Analogous escape mutations were not observed in available genomic NL63 sequences spanning decades (1983–2023), with contemporary NL63 strains isolated from 2018-2019 remaining equally sensitive to human neutralising mAbs. Overall, this suggests that the NL63 spike is able to structurally tolerate variation to escape neutralisation by human antibodies, but is not under selective pressure to do so22,27.

A key limitation of our study is the lack of in vivo protective efficacy data. Despite significant effort at mouse adaptation in vitro or establishing mouse challenge models, we were unable to derive a replicative model of NL63 in either ACE2-transgenic or wildtype mice (data not shown). Development of such an informative animal model is a key limiting step to directly assessing the protective capacity of vaccines and therapeutics against NL63 and related alphacoronaviruses. Nevertheless, it is likely that our mAbs are protective in vivo, as seen for similarly potent RBD mAbs that block receptor engagement for SARS-CoV36,37, SARS-CoV-238,39 and MERS-CoV40,41. A HR2 targeting S2 neutralising mAb also displayed therapeutic efficacy in a hamster challenge model of SARS-CoV-233, despite having an IC50 of 12.2 µg/ml ( ~ 1400-fold higher than NLH02). Additionally, while we were able to generate escape mutants in vitro, it is unclear whether viral escape will occur in vivo in response to prolonged exposure to NL63 neutralising mAbs, despite viral escape not occurring in humans.

While no alphacoronaviruses have caused a pandemic in recent human history, the presence of two endemic viruses, one that shares a receptor with highly pathogenic betacoronaviruses, suggests that alphacoronaviruses similarly pose an emerging threat. While disease severity has generally been low to date, alphacoronaviruses are known to recombine in bats42, and can cause zoonotic infections in humans43. For example, a strain of canine alphacoronavirus (CCoV-HuPn-2018) was isolated from children with pneumonia in Malaysia and found to use the same receptor as 229E (aminopeptidase N)43. Polyclonal human antibodies elicited by 229E infection could cross-neutralise CCoV-HuPn-2018, highlighting that pre-existing alphacoronavirus immunity provides a level of protection against antigenically related zoonotic viruses. Our study here provides a deeper understanding of NL63 recognition by human antibodies, providing insights into sites of vulnerability with protective potential. Additional studies are needed to gain an appropriate understanding of how to design vaccines and therapeutics to bolster pandemic preparedness against future alphacoronavirus threats.

Materials and methods

Human subjects and ethics

A cohort of 14 healthy adults was recruited in early 2022 to donate blood via venepuncture into 9 ml blood collection tubes with sodium heparin anticoagulant. Plasma was collected and stored at -80°C, and PBMCs were isolated via Ficoll Paque separation, cryopreserved in fetal calf serum (FCS) with 10% DMSO and stored in liquid nitrogen. The study protocols were approved by the University of Melbourne Human Research Ethics Committee (#2056689) and all associated procedures were carried out in accordance with the approved guidelines. All participants provided written informed consent in accordance with the Declaration of Helsinki.

Cell lines

LLC-MK2 cells (macaque kidney epithelial cells) and HAT24 cells (a clone of transduced HEK-293T cells stably expressing ACE2 and TMPRSS244) were maintained in cDMEM (DMEM with 10% FCS and 1% penicillin-streptomycin-glutamine). LLC-AT cells were generated by transducing LLC-MK2 cells with lentiviral constructs expressing human ACE2 (pHAGE2 containing the angiotensin-converting enzyme 2 gene; NR-52512, BEI Resources, NIAID, NIH) and TMPRSS2 (pscALPSblasti-TMPRSS2 Blasti, Addgene plasmid #158088, a gift from Jeremy Luban). Following lentiviral transduction, cells were surface stained with goat anti-human ACE2 (R&D #AF933) at 5 µg/ml and anti-TMPRSS2 PE (BioLegend #378403) antibodies. After washing with PBS containing 1 mM EDTA, cells were stained with the Donkey anti-goat IgG AF647 secondary antibody (Invitrogen #A21447). Cells expressing high levels of ACE2 and TMPRSS2 were bulk sorted and maintained in cDMEM.

NL63 virus strains and propagation

HCoV-NL63 (Amsterdam1) isolate was grown in LLC-MK2 cells in DMEM with 2% FCS and 1 µg/ml TPCK trypsin at 35 °C. Contemporary NL63 isolates (NL63/AUS/18091206/2018, NL63/AUS/18111908/2018, NL63/AUS/19071116/2019) were propagated from nasopharyngeal swabs collected between 2018-2019 by the Victorian Infectious Disease Reference Laboratory. Human nasal epithelial cell cultures were established as previously described18, inoculated with clinical material for 2 h at 33 °C and cultured up to 7 days. Apical washes were collected daily and stored at −70 °C. RNA was extracted from apical washes using the QiaAmp 96 Virus QiaCube HT kit (Qiagen, Cat. 57731) and RT-PCR reactions set up using SensiFast Probe No-ROX One-Step Kit (Bioline, Cat. BIO-76005) with the following primers/probes: NL63 Forward: 5’-AGGTTGACTTGTATAATGGTGCT-3’, NL63 Reverse: 5’-GCCAACACAAAGAAAAATATCA-3’ and NL63 Probe: 5’- TGCCGAAGAGCCTGTTGTTGGT -3’. Apical washes that were positive for NL63 RNA were then used to inoculate LLC-AT cells (LLC-MK2 cells overexpressing ACE2 and TMPRSS2) in DMEM with 2% FCS and 1 µg/ml TPCK trypsin at 35 °C for 1-2 passages. For all isolates, cell culture supernatants containing infectious virus were harvested on Day 6 or 7 when ample CPE could be detected, clarified via centrifugation, aliquoted and stored at −80 °C.

NL63 neutralisation assay with viability dye readout

A live virus neutralisation assay was adapted for NL63 using a previously established protocol for SARS-CoV-245. Infectivity of NL63 virus stocks was determined by titration on HAT-24 cells (a clone of transduced HEK293T cells stably expressing human ACE2 and TMPRSS244). In a 96-well flat bottom plate, virus stocks were serially diluted three-fold (1:5-1:10,935) in DMEM with 5% FCS, added with 15,000 freshly trypsinised HAT-24 cells per well and incubated at 35°C. After 70 h, 10 µl of alamarBlue™ Cell Viability Reagent (ThermoFisher) was added into each well and incubated at 35 °C for 1 h. The reaction was then stopped with 1% SDS and read on a FLUOstar Omega plate reader (excitation wavelength 560 nm, emission wavelength 590 nm). The relative fluorescent units (RFU) measured were used to calculate %viability (‘sample’ ÷ ‘no virus control’ ×100), which was then plotted as a sigmoidal dose response curve on Graphpad Prism to obtain the virus dilution that induces 50% cell death (50% infectious dose; ID50). Each virus was titrated in quintuplicate to obtain mean ID50 values.

In 96-well flat bottom plates, 3-fold dilutions of heat-inactivated plasma samples (1:20-1:43,740) or NL63 mAbs (10,000 ng/ml – 4.6 ng/ml) were incubated in duplicate in DMEM-5%FCS with NL63 virus at a final concentration of 2× ID50 at 35 °C for 1 h. Next, 15,000 freshly trypsinised HAT-24 cells in serum-free DMEM were added and incubated at 35 °C. ‘Cells only’ and ‘Virus+Cells’ controls were included to represent 0% and 100% infectivity respectively. After 70 h, 10 µl of alamarBlue™ Cell Viability Reagent (ThermoFisher) was added into each well and incubated at 37 °C for 1 h. The reaction was then stopped with 1% SDS and read on a FLUOstar Omega plate reader (excitation wavelength 560 nm, emission wavelength 590 nm). The relative fluorescent units (RFU) measured were used to calculate %neutralisation with the following formula: (‘Sample’ – ‘Virus+Cells’) ÷ (‘Cells only’ – ‘Virus+Cells’) × 100. IC50 values were determined using four-parameter non-linear regression in GraphPad Prism with curve fits constrained to have a minimum of 0% and maximum of 100% neutralisation.

Expression of coronavirus antigens

NL63 and SARS-CoV-2 spike proteins were generated for serological and flow cytometric assays using previously described techniques46. Genes encoding the ectodomain of SARS-CoV-2 S (NC_045512; aa1-1209) with 6 proline stabilisation mutations and furin cleavage site removal (Hexapro47), or spike ectodomains with 2 proline stabilisation mutations (S-2P; aa1-1291) from isolates Amsterdam57 (DQ445911.1), Amsterdam1 (NC_005831.2), ChinaGD02 (MK334043) and SADS-CoV/CN/GDWT/2017 (MG557844.1; aa1-1065) were synthesised (GeneArt) and cloned into mammalian expression vectors with a C-terminal T4 fibritin foldon, Avitag and polyhistidine tag. Genes encoding NL63 spike domain S10 (aa1-205), domain S10 and S1A (aa1-454), or domain S1B (aa474-633; N-terminal prolactin signal sequence) were synthesised with a C-terminal polyhistidine tag and cloned. Proteins were expressed in Expi293 or ExpiCHO cells (Thermo Fisher, Massachusetts USA) using manufacturer’s instructions and purified using Ni-NTA and size exclusion chromatography. Protein integrity was confirmed using SDS-PAGE.

Negative-staining electron microscopy of the NL63 Spike

Purified recombinant NL63 Spike was diluted in 20 mM HEPES pH 7.5, 150 mM NaCl to a final concentration of 0.1 mg/ml. 4 µl of the sample was applied to a glow-discharged (GloQube Plus, Quorum Technologies) carbon coated 400 micron mesh copper grid and allowed to adsorb for 1 min. Prepared grids were then stained with 4 µl 1% Uranyl Acetate for 1 min before blotting away excess stain. Micrographs were collected at 57,000x magnification with a calibrated pixel size of 2.44 Å on a Talos 120 C microscope (Thermo Fisher Scientific USA) running at 120 keV equipped with a CETA 4x4k CMOS camera. Particle picking and reference-free 2D classification were performed in CryoSPARC 4.6.248.

ELISA against coronavirus spike proteins

Antibody binding to spike and spike subdomain proteins was assessed using ELISA. 96-well Maxisorp plates (Thermo Fisher) were coated overnight at 4 °C with 2 µg/mL recombinant spike or spike subdomain proteins. After blocking with PBS containing 1% FCS, duplicate wells of 4-fold serially diluted plasma (starting from 1:100) or mAbs (starting from 10 µg/ml) were added and incubated for two hours at room temperature. Bound antibody was detected using 1:20,000 dilution of HRP-conjugated anti-human IgG (Sigma) and plates developed using TMB substrate (Sigma), stopped using sulphuric acid and read at 450 nm. Plates were washed with PBS containing 0.05% Tween 20 3–6 times between incubations. Endpoint titres were calculated using Graphpad Prism as the reciprocal serum dilution giving signal 2× background using a fitted curve (4 parameter log regression).

S1B (RBD)-ACE2 inhibition ELISA

An ELISA to measure the ability of mAbs to block the interaction between ACE2 and NL63 S1B (RBD) was adapted from a previous protocol used for SARS-CoV-249. Recombinant full length human ACE2 ectodomain with mutations to enhance binding to SARS-CoV-2 RBD (T27Y, L79T, N330Y50) was fused to the constant regions of an IgG1 heavy chain (CH1-CH2-CH3) and to the light chain constant domain (CL). Protein expression and purification were as previously described51, except co-transfection of the CH and CL fusion constructs expressed a tetravalent ACE2 ‘IgG-like’ molecule. This tetravalent ACE2-Fc was then biotinylated with the EZ-Link™ Sulfo NHS-LC-LC-Biotin, No-Weigh™ Format kit (Thermo Fisher, #A35358). 96-well Maxisorp plates (Thermo Fisher) were coated overnight at 4 °C with 1.5 µg/mL of recombinant NL63 RBD protein (R&D systems, #10605-CV-100) or 1.5 µg/ml of BSA in PBS. After blocking with PBS containing 4% BSA and 0.1% Tween-20, duplicate wells of three-fold serially diluted mAbs (10,000 ng/ml–4.6 ng/ml) in PBS with 0.1% BSA were added and incubated for 1 h at room temperature. Plates were then incubated with 0.5 μg/ml of biotinylated tetravalent ACE2-Fc for 1 h at room temperature followed by incubation with HRP-conjugated streptavidin (1:10,000 dilution; Thermo Fisher, #21130) for 1 h at room temperature. Plates were developed with TMB substrate (Thermo Fisher), stopped with 0.16 M sulphuric acid and read at 450 nm. The OD450 values were used to calculate %inhibition using the following formula: (‘No antibody’ – ‘Sample’) ÷ (‘No antibody’ – ‘BSA control’) × 100. IC50 values were determined using four-parameter non-linear regression in GraphPad Prism with curve fits constrained to have a minimum of 0% and maximum of 100% inhibition.

Flow cytometric detection of spike-reactive B cells

Spike-specific memory B cells were measured as previously described14. Biotinylated recombinant SARS-CoV-2 or NL63 (Amsterdam57) spike trimers were conjugated to streptavidin-APC or -PE fluorophores. PBMCs were thawed and stained with Aqua viability dye (Thermo Fisher Scientific) and then surface stained with Spike probes, CD19 ECD (J3-119) (Beckman Coulter), IgD PE-Cy7 (IA6-2), IgG BV786 (G18-145), streptavidin BV510 (BD Biosciences), CD14 BV510 (M5E2), CD3 BV510 (OKT3), CD8a BV510 (RPA-T8), CD16 BV510 (3G8), and CD10 BV510 (HI10a) (BioLegend). Cells were washed twice with PBS containing 1% FCS and fixed with 1% formaldehyde (Polysciences), acquired on a BD LSR Fortessa using BD FACS Diva and analysed using FlowJo 10.

Recovery of human monoclonal antibodies from NL63 spike-specific B cells

NL63 spike-specific B cells were identified within cryopreserved human PBMC using fluorescent spike probes as described above. Single antigen-specific class-switched B cells (S or RBD + , CD19+ IgD- IgG + ) were sorted using a BD Aria II into 96-well plates, subject to cDNA generation and multiplex PCR and Sanger sequencing, as previously described in refs. 46,52. Productive, recombined heavy (V-D-J) and light (V-J) chain immunoglobulin sequences were synthesised (Geneart) and cloned into human IgG1 expression vectors for recombinant production in Expi293 or ExpiCHO mammalian cell culture using transient transfection. After 4 – 5 days, IgG1 was purified from culture supernatants using Protein-A affinity chromatography.

Viral escape assay and sequencing

NL63 escape mutants were generated via serial passaging of NL63 viruses with neutralising mAbs. Five-fold dilutions of mAbs were incubated with NL63 virus (Amsterdam1) in 24 well flat bottom plates in DMEM with 5% FCS for 1 h at 35 °C. 90,000 freshly trypsinised HAT-24 cells in serum-free DMEM were then added to the virus-mAb mixtures and incubated for 3-4 days in the presence of 1 µg/ml TPCK trypsin. Supernatants from wells with visible CPE at the highest concentration of mAb were collected and clarified via centrifugation. An aliquot was frozen at −80 °C while the rest of the supernatant was used to infect a new plate of HAT-24 cells with a new titration of mAbs. mAb concentrations were increased five-fold if there was evidence of neutralisation escape (if CPE was visible at higher mAb concentrations compared to the previous passage) until CPE was observed at 50 µg/ml of mAb. NL63 was also serially passaged in the absence of mAb to control for cell culture adaptation.

Viral RNA was extracted from supernatants collected at each passage using the QIAamp® viral RNA isolation kit (Qiagen) following manufacturer’s instructions. cDNA and next-generation sequencing libraries were prepared following the Twist Bioscience Total Nucleic Acids Library Preparation Kit (Twist Bioscience), with quality assessed using Qubit (ThermoFisher Scientific) and Agilent 2200/4200 TapeStation platforms (Agilent). Prepared libraries were enriched for NL63 by applying the Comprehensive Viral Research Panel (Twist Bioscience) and the manufacturer’s standard hybridisation protocol. Libraries were sequenced using an Illumina P1 cartridge on a NextSeq2000 platform (Illumina) or an i1 v2 cartridge on an iSeq100 (Illumina) with a paired end, 150 base pair sequencing format. Sequence reads were demultiplexed and quality and adapter trimmed before aligning to the NL63 reference genome (NC_005831) using minimap2 and the short-read preset (2.24, https://github.com/lh3/minimap2), retaining paired reads which aligned. NL63 specific reads were then down sampled to 400,000 reads per sample to make further sequence analysis tractable. These down sampled reads were realigned to the NL63 reference as above, with samples achieving a mean coverage of 75 reads or more classified as suitable for further analysis. Sequence variants were called if they reached a minimum coverage of ten reads and a variant frequency of 0.5 or above. Annotations for NL63 were used to determine coding changes associated with each variant, and each alignment manually inspected in Geneious Prime (2022.1.1).

An escape variant is defined here as a variant that (a) is absent in baseline passages, (b) imparts a protein coding change, and (c) increases in frequency under selection with increasing concentrations of mAb.

Sequence analysis of viral isolates

To characterise sequence conservation of the NL63 spike, full genomic sequences (n = 207) spanning depositions from 1983–2023 were exported from Genbank and spike protein sequences aligned using Clustal Omega 1.2.3 and Geneious Prime. Amino acid conservation at each position was visualised using Pymol 2.5.5.1 (Schrodinger).