Cell lines

FreeStyle293F cells (Thermo Fisher Scientific) and HEK293-S cells (Thermo Fisher Scientific) were cultured in FreeStyle293 expression medium (Life Technologies), cultured at 37 °C with 8% CO₂ while shaking at 130 rpm. HEK293T cells (ATCC) and Vero E6 cells (ATCC) were cultured at 37 °C in the presence of 5% CO₂ in DMEM supplemented with 10% heat-inactivated FBS, 1% penicillin, 1% streptomycin, 2 mM L-glutamine, non-essential amino acids (Invitrogen) and 1 mM sodium pyruvate. ExpiCHO-S cells (GIBCO) were cultured at 37 °C with 8% CO₂ while shaking at 130 rpm in ExpiCHO expression media (GIBCO). VeroE6/TMPRSS2 cells that stably express human TMPRSS2 (NIBIOHN, JCRB1819)⁴⁷ were cultured in DMEM containing 10% FBS, Penicillin (100 unit/mL), Streptomycin (100 μg/mL), Geneticin (G418) (1 mg/mL). VeroE6/TMPRSS2 cells were seeded in medium without Geneticin for infectivity assays.

Expression and purification of recombinant SARS-CoV-2 spike

ExpiCHO cells were transfected with a plasmid encoding the SARS-CoV-2 spike ectodomain (residues 1-1208), with 2P (K986 and V987) and RRAR-GSAS (residues 682-685) mutations to stabilize the protein in the prefusion conformation⁴⁸, a kind gift from Dr. Jason McLellan. Twelve days after transfection, the supernatant was collected, and the protein was purified from the supernatant with a 5 mL HisTrap^TM High Performance column (Cytiva), which had been equilibrated with 20 mM NaH₂PO₄ pH 7.4, 0.5 M NaCl, and 20 mM imidazole. The protein was eluted with 20 mM NaH₂PO₄ pH 7.4, 0.5 M NaCl and 0.5 M imidazole. Peak fractions were desalted to PBS using a HiPrep^TM 26/10 desalting column (Cytiva) and sterile filtered with a low protein binding 0.22 µm filter.

VHH phage display and panning

A llama that was previously immunized with recombinant prefusion stabilized SARS-CoV-1 and MERS spike protein was additionally immunized 3 times with recombinant SARS-CoV-2 spike protein (S-2P, Wuhan sequence) stabilized in its prefusion conformation^49,50. Five days after the third immunization, peripheral blood lymphocytes were isolated from the llama and a VHH-displaying phagemid library was constructed. Phages that display SARS-CoV-2 spike specific VHHs were enriched from this phage library by 2 successive rounds of biopanning on 100 ng of HIS-tagged SARS-CoV-2 spike 2P protein⁴⁹, that was immobilized in a well of a microtiter plate (type II, F96 Maxisorp, Nunc) via precoated anti-HIS antibodies in the presence of 10 μg/mL RBD-SD1-mouse IgG. Two additional rounds of biopanning were performed using anti-HIS captured spike proteins (R3C and R4C series) or using directly coated spike proteins (R3DC and R4DC series). Also these two series of additional rounds of biopanning were performed in the presence of 10 μg/mL RBD-SD1-mouse IgG. For each panning round an uncoated well was used as a negative control. The wells were then washed 5 times with phosphate-buffered saline (PBS) + 0.05% Tween 20 and blocked with 4% milk powder in PBS (Regilait) in the first panning round and Pierce protein free blocking buffer (Thermo Fisher Scientific), SEA blocking buffer (Thermo Fisher Scientific) and 1% BSA in PBS in the subsequent panning rounds. Non-specifically bound phages were removed by extensive washing with PBS + 0.05% Tween 20. The retained phages were eluted with TEA-solution (14% trimethylamine (Sigma) pH 10) and subsequently neutralized with 1M Tris-HCl pH 8.0. The collected phages were amplified in exponentially growing E. coli TG1 cells, infected with VCS M13 helper phages and subsequently purified using PEG 8000/NaCl precipitation for the next round of selection. Enrichment after each panning round was evaluated by infecting TG1 cells with 10-fold serial dilutions of the collected phages after which the bacteria were plated on LB agar plates with 100 μg/mL− ampicillin and 1% glucose.

Preparation of periplasmic extracts

After 3 or 4 panning rounds individual colonies of phage infected bacteria were randomly selected for further analysis. The selected colonies were inoculated in 2 mL of terrific broth (TB) medium with 100 μg/mL ampicillin in 24-well deep well plates. After 5 h incubation at 37 °C, isopropyl β-D-1-thiogalactopyranoside (IPTG) (1 mM) was added to induce VHH expression during overnight incubation at 37 °C. To prepare periplasmic extract, the bacterial cells were pelleted and resuspended in 250 μL TES buffer (0.2 M Tris-HCl pH 8.0, 0.5 mM EDTA, 0.5 M sucrose) and incubated at 4 °C for 30 min. Subsequently 350 μL water was added to induce an osmotic shock. After 1 h incubation at 4 °C followed by centrifugation, the periplasmic extract was collected.

Production of VHHs in E. coli and purification

For the production of VHH in E. coli, a pMECS vector containing the VHH of interest was transformed into WK6 cells (the non-suppressor E. coli strain) and plated on an LB plate containing ampicillin. The next day clones were picked and grown overnight in 2 mL LB containing 100 μg/mL ampicillin and 1% glucose at 37 °C while shaking at 200 rpm. One mL of this preculture was used to inoculate 25 mL of TB supplemented with 100 µg/mL ampicillin, 2 mM MgCl₂ and 0.1% glucose and incubated at 37 °C with shaking (200–250 rpm) until an OD₆₀₀ of 0.6–0.9 was reached. VHH expression was induced by addition of IPTG to a final concentration of 1 mM followed by overnight incubation at 28 °C while shaking at 200 rpm. The VHH-containing fraction was extracted from the periplasm and purified as described in Wrapp et al.⁵⁰. In short, the VHHs were purified from the periplasmic extract using Ni Sepharose beads (GE Healthcare). After elution with 500 mM imidazole, the VHH-containing fractions were buffer-exchanged with PBS using a Vivaspin column (5 kDa cutoff, GE Healthcare). The purified VHHs were analyzed by SDS-PAGE and Coomassie staining and by intact mass spectrometry. For crystallography, R3DC23 was produced in a 1 L culture of E. coli WK6. The periplasmic extract was loaded on a 1 mL HisTrap^TM High Performance column (Cytiva), which had been equilibrated with 20 mM NaH₂PO₄ pH 7.5, 0.5 M NaCl, and 20 mM imidazole. Bound VHH was eluted with 20 mM NaH₂PO₄, 20 mM NaCl and 0.5 M imidazole. The peak fractions were pooled and further purified by size exclusion chromatography, using a HiLoad 16/600 Superdex® 75 prep grade column (Sigma-Aldrich), in PBS.

Enzyme-linked immunosorbent assay

Wells of microtiter plates (type II, F96 Maxisorp, Nunc) were coated overnight at 4 °C with 100 ng of recombinant SARS-Cov-2 S-2P, SARS-CoV-2 S-6P protein, recombinant SARS-CoV-1 spike protein, recombinant MERS-CoV spike protein, recombinant HKU1 spike protein (all kindly provided by Dr. Jason McLellan), SARS-CoV-2 Omicron BA.1 S protein (ACROBiosystems), mouse Fc-tagged SARS-CoV-2 RBD (Sino Biological, 40592-V05H), the SARS-CoV-2 spike S2 subunit (ACROBiosystems, S2NC52H5) or BSA. The coated plates were blocked with 5% milk powder in PBS. Dilution series of the VHHs, VHH-Fcs or antibodies were added to the wells and plates were further incubated for 90 min at room temperature. After washing, binding of VHHs was detected by HRP-conjugated rabbit anti-camelid VHH antibodies (Genscript, catalog number A01861-200, 1/5000) or a mouse anti-HA antibody (clone 16B12, BioLegend, catalog number 901501, 1/2000), followed by anti-mouse IgG-HRP (Cytiva, NA931V, 1/2000). Binding of VHH-Fcs or conventional human monoclonal antibodies was detected by HRP-conjugated rabbit anti-human IgG (Sigma, catalog number A8792, 1/2000). After washing, 50 μL of TMB substrate (tetramethylbenzidine, BD OptEIA) was added to the plates and the reaction was stopped by addition of 50 μL of 1 M H₂SO₄. The absorbance at 450 nm was measured with an iMark Microplate Absorbance Reader (Bio Rad). Curve fitting was performed using nonlinear regression using Graphpad 8.0.

In the PE-ELISA shown in Fig. 1a, the blue-white heat map shows for each PE sample (10-fold diluted), the ratio of the ELISA OD₄₅₀ signal for the indicated antigen over the ELISA OD₄₅₀ signal of the corresponding PE sample for the control antigen (BSA), normalized to the signal for a well with TES buffer (used to prepare the PEs) (0%) and the highest OD₄₅₀ ratio in the ELISA (100%). PEs prepared from E. coli cells that express an RBD binding VHH were used as control. Buffer used to prepare the PEs was used as negative control (TES).

Generation of replication-deficient VSV pseudotyped viruses

To generate replication-deficient VSV pseudotyped viruses, HEK293T cells, transfected with SARS-CoV-1 S or SARS-CoV-2 S expression vectors were inoculated with a replication-deficient VSV vector containing eGFP and firefly luciferase expression cassettes^48,51. After 1 h incubation at 37 °C, the inoculum was removed, cells were washed with PBS and incubated in media supplemented with an anti-VSV G mAb (ATCC, catalog number CRL-2700, RRID:CVCL_G654; 1/1000) for 16 h. Pseudotyped particles were then harvested and clarified by centrifugation⁵¹.

Pseudovirus neutralization assay with periplasmic extracts

Pseudoviruses expressing the SARS-CoV-2 spike (D614G) were incubated for 30 min at 37 °C with a 1/100 dilution of periplasmic extract in Fluorobrite DMEM medium (Invitrogen), supplemented with 5% heat-inactivated FBS, 1% penicillin, 1% streptomycin, 2 mM L-glutamine, non-essential amino acids (Invitrogen) and 1 mM sodium pyruvate. The incubated pseudoviruses were subsequently added to subconfluent monolayers of Vero E6 or Vero E6/TMPRSS2 cells from which the original growth medium was removed. Sixteen hours later, the cells were lysed using passive lysis buffer (Promega). The transduction efficiency was quantified by measuring the GFP fluorescence in the prepared cell lysates using a Tecan infinite 200 pro plate reader. GFP fluorescence was normalized using the GFP fluorescence of non-infected cells and infected cells treated with PBS.

Flow cytometric analysis of VHH-binding to HEK293T cells expressing SARS-CoV spike proteins or fragments thereof

Binding of VHHs to spike proteins on the surface of mammalian cells was determined by flow cytometry using pcG1-expression plasmids containing the coding sequence of the ancestral SARS-CoV-2 spike protein from which the C-terminal 18 amino acids were deleted (AA1-1255) and in which the D614G substitution was introduced by QuickChange site-directed mutagenesis (Agilent) according to the manufacturer’s instructions. For analysis of VHH-binding to cells expressing the SARS-CoV-2 HR2 coiled coil, a gene fragment containing the SARS-CoV-2 signal peptide (MFVFLVLLPVVSSQ), fused via a GSG-linker to an HA-tag and fused via a GSG-linker to the SARS-CoV-2 P1162-K1255 sequence was ordered at IDT and cloned into a pcDNA3.4 expression vector via TOPO-cloning (pcDNA^TM 3.4 TOPO^TM TA cloning kit, Thermo Fisher Scientific). Two days after transfecting HEK293T cells with a GFP expression plasmid in combination with either a spike expression plasmid or a control expression plasmid the cells were collected. All further procedures were performed on ice. The cells were washed once with PBS and blocked with 1% BSA. The cells were then stained with antibody or VHH dilution series for 90 min and subsequently washed 3 times with PBS containing 1% BSA. Binding of VHHs was detected using a mouse anti-HIS-tag antibody (Biorad, MCA1396, 1/2000) and an AF647 conjugated donkey anti-mouse IgG antibody (Invitrogen, catalog number A31571, 1/1000). Binding of VHH-Fcs or antibodies was detected using an AF633 donkey goat anti-human IgG antibody (Invitrogen, catalog number A21091, 1/1000) and dead cells were stained using Live/Dead stain (Invitrogen, 15560607, 1/1000). Following 3 washes with PBS containing 0.5% BSA, the cells were analyzed by flow cytometry using a BD LSRII flow cytometer (BD Biosciences). Binding was calculated as the ratio of the AF647 MFI of antibodies binding GFP⁺ cells over that of GFP^– cells. The binding curves were fitted using nonlinear regression (Graphpad 9.10).

For cells transfected with 6HB and 5HB expression vectors, the expression of the 6HB and 5HB was confirmed by mouse anti-HA staining of the HA-tag that was fused to the N-terminus of these protein constructs, and for the 5HB also by binding of monomeric HIS-SUMO-HR2, which was detected with a mouse anti-HIS-tag antibody. Binding of HIS-SUMO-HR2 or HR2-foldon to spike-expressing cells expressing were detected using a mouse anti-HIS-tag antibody and a mouse anti-strep-tag antibody (Qiagen, catalog number 34850, 1/1000), respectively, in combination with an AF647-conjugated anti-mouse IgG antibody (Invitrogen, A31571, 1/1000). Binding was calculated as the ratio of the AF647 MFI of antibodies binding GFP⁺ cells over that of GFP^– cells. An example of the gating strategy for data analysis is shown in Supplementary Fig. 13A.

Generation of spike protein expression vectors for the production of VSVdelG pseudovirus particles expressing spike proteins of SARS-CoV-2 variants

The pCG1 expression vector for the SARS-CoV-2 spike protein containing the D614G mutation was generated from the pCG1-SARS-2-Sdel18 vector by introducing the specific RBD mutation(s) via QuickChange mutagenesis using appropriate primers, according to the manufacturer’s instructions (Aligent). For the pCG1-SARS-2-Sdel18 expression vector for the omicron BA.1 variant, a codon-optimized spike protein nucleotide sequence containing the BA.1 mutations (A67V, Δ69-70, T95I, G142D, Δ143-145, N211I, Δ212, ins215EPE, G339D, S371L, S373P, S375F, K417N, N440K, G446S, S477N, T478K, E484A, Q493R, G496S, Q498R, N501Y, Y505H, T547K, D614G, H655Y, N679K, P681H, N764K, D796Y, N856K, Q954H, N969K, L981F) and flanking BamHI and SalI restriction sites was ordered at GeneArt (Thermo Fischer Scientific) and cloned in the pCG1 vector as a BamHI/SalI fragment. For the pCG1-SARS-2-BA.2 Sdel18 expression vector, a codon-optimized spike protein nucleotide sequence containing the BA.2 mutations (T19I, ΔL14-P26, A27S, G142D, V213G, G339D, S371F, S373P, S375F, T376A, D405N, R408S, K417N, N440K, S477N, T478K, E484A, Q493R, Q498R, N501Y, Y505H, D614G, H655Y, N679K, P681H, N764K, D796Y, Q954H, N969K) and flanking BamHI and SalI restriction sites was ordered at GeneArt (Thermo Fischer Scientific) and cloned in the pCG1 vector as a BamHI/SalI fragment. After sequencing, clones containing the correct spike coding sequence were prepared using the Qiagen plasmid kit. For the pCG1-SARS-2-BA.2.75 Sdel18 expression vector, a codon-optimized spike protein nucleotide sequence containing the BA.2.75 mutations (T19I, ΔL14-P26, A27S, G142D, S147E, W152R, F157L, I210V, V213G, G257S, G339D, S371F, S373P, S375F, T376A, D405N, R408S, K417N, N440K, G446S, S477N, N460K, T478K, E484A, Q498R, N501Y, Y505H, D614G, H655Y, N679K, P681H, N764K, D796Y, Q954H, N969K) and flanking BamHI and SalI restriction sites was ordered at GeneArt (Thermo Fischer Scientific) and cloned in the pCG1 vector as a BamHI/SalI fragment. To generate expression plasmids for the BA.5 spike protein, mutations ΔLH96-V70, L452R, F486V and R493Q were introduced in the BA.2 spike expression construct by QuickChange mutagenesis according to the manufacturer’s instructions (Aligent). Likewise, to generate the BA.4.6 and BQ.1.1 spike expression vectors, mutations R346T and N658S and mutations R346T and K444T were respectively introduced in the BA.5 spike expression vector by QuickChange mutagenesis. The BF.7 spike expression vector was generated by introducing the R346T substitution in the BA.5 spike expression vector by QuickChange mutagenesis. After sequencing, plasmids from clones containing the correct spike coding sequence were prepared using the Qiagen plasmid kit. To generate the pCG1-SARS-2-XBB, a gBlock corresponding to the SARS-CoV-2 amino acid sequence V70-V510 containing the XBB mutations (V83A, G142D, ΔY144, H146Q, Q183E, V213E, D339H, R346T, L368I, S371F, S373P, S375F, T376A, D405N, R408S, K417N, N440K, V445P, G446S, S477N, T478K, E484A, F486S, F490S, Q498R, N501Y, Y505H) was ordered (IDT) and cloned into the pCG1-SARS-2 BA.2 vector via Gibson assembly according to the manufacturer’s instructions (New England Biolabs). To this end, the pCG1-SARS-2-BA.2 vector was amplified by PCR using appropriate primers (5’ GTGCCATTGGTGCCGGACACG 3’ and 5’ CACCAGCCTTACAGAGTGG 3’) and the Phusion High-Fidelity DNA polymerase (New England Biolabs). The XBB.1.5(-G252V) spike expression vector was generated by introducing the F496P substitutions in the XBB spike expression vector by QuickChange mutagenesis. The KP.3 spike expression vector was generated by introducing F456L, Q493E, V1104L in the JN.1 spike expression vector by Gibson assembly (using primers 5’ CCCAGATCATCACCACCGACAACACCTTCGTGTCTGGC 3’ and 5’ TAATCGTAGTTGCCGgaGTGTTTGGAGTCCAGCTTGTTG 3’). The KP.2.3 spike expression vector was generated by introducing Δ31, H146Q, R346T and E493Q (reversion to WT) in the KP.3 spike expression vector by Gibson assembly (using primers 5’ CAGACCCACATATGGCGTGGGCCACCAGCCTTACAGAG 3’ and 5’ TCACACACTGGCTGGACACCAGAGGCAGCAGCACCAGAAAC 3’. The KP.3.1.1 spike expression vector was generated from the KP.3 spike expression vector by introducing Δ31 with QuickChange mutagenesis. The LB.1 spike expression vector was generated from the JN.1 spike expression vector by introducing Δ31, Q183H, R346T and F456L by Gibson assembly (using primers 5’ TCCACCGAGATCTATCAGGCCGGCAACAAGCCCTGTAAG 3’ and 5’ TTCACACACTGGCTGGACACCAGAGGCAGCAGCACCAG 3’). The KP.1.1 spike expression vector was generated from the KP.3 spike expression vector by introducing R346T, E493Q (reversion to WT) and K1086R by Gibson assembly (using primers 5’ AGAGAAGGCGTGTTCGTGTCCAACGGCACCCATTGGTTC 3’). The XEC spike expression vector was generated from the KP.3 spike expression vector by introducing T22N and F59S by Gibson assembly (using primers 5’ ATGGCACCAAGAGATTCGACAACCCCGTGCTGCCCTTC 3’ and 5’ GGCAGCAGCACCAGAAACACGAACATGGTGGCGGATCCG 3’). The nucleotide sequence of all purified spike expression plasmids was verified by Sanger sequencing. The Khosta-2 spike expression vector was a kind gift from Dr. Michael Letko.

For the validation of the amino acids that are important for binding of R3DC23 as determined by X-ray crystallography, every position in HR2 between N1194 and L1203 that was not involved in the viral escape analysis was substituted to at least one amino acid that occurred at that position in the GISAID database by Quickchange mutagenesis. Expression levels of each of the mutated spikes were confirmed by staining with S309, an RBD-binding monoclonal antibody.

SARS-CoV pseudovirus neutralization assay

For the VSV pseudotype neutralization experiments, the pseudoviruses were incubated for 30 min at 37 °C with different dilutions of purified VHH or VHH-Fc fusions or with GFP-binding protein (GBP: a VHH specific for GFP). The incubated pseudoviruses were subsequently added to subconfluent monolayers of Vero E6 or VeroE6/TMPRSS2 cells. Sixteen hours later, the cells were lysed using passive lysis buffer (Promega). The transduction efficiency was quantified by measuring the GFP fluorescence in the prepared cell lysates using a Tecan infinite 200 pro plate reader. GFP fluorescence was normalized using either the GFP fluorescence of non-infected cells and infected cells treated with PBS or the lowest and highest GFP fluorescence value of each dilution series. The IC₅₀ was calculated by non-linear regression curve fitting, log(inhibitor) vs. response (four parameters). Alternatively, dilution series of VHHs or antibodies were mixed with 100 PFU of GFP-expressing replication competent VSV virus particles pseudotyped with the SARS-CoV-2 spike protein derived from an early isolate. Of note, during propagation of this viral clone (S1-10a) on Vero E6 cells, the furin cleavage site was mutated and as such inactivated²¹. After 30 min incubation at 37 °C the virus-antibody mixtures were added to monolayers of Vero E6, VeroE6/TMPRSS2 or A549-hACE2 cells and allowed to infect and replicate for three days. In all neutralization assays using pseudotyped VSV viral particles, FluoroBrite DMEM medium (Invitrogen) supplemented with 5% heat-inactivated FBS, 100 unit/mL penicillin, 100 unit/mL streptomycin, 2 mM L-glutamine, non-essential amino acids (Invitrogen) and 1 mM sodium pyruvate was used to prepare the VHH- or antibody-virus mixtures. The mixtures were added to cells from which the original growth medium was removed. For all VSV pseudotype neutralization assays using huR3DC23-Fc_LS, huR3DC23-Fc_LS from stable cell pools was used.

Neutralization assays using authentic SARS-CoV-2 performed at VIB-UGent CMB

In vitro neutralization experiments with authentic SARS-CoV-2 viruses were performed in the Biosafety level 3 laboratory at VIB-UGent Center for Medical Biotechnology. The plaque reduction assays using authentic viruses were performed with SARS-CoV-2 D614G strain SARS-CoV-2/human/FRA/702/2020, obtained from the European Virus Archive (EVAG) and with an SARS-CoV-2 BA.1 virus that was obtained from KULeuven⁵² and grown at KULeuven on Vero E6 cells. Further propagation of the virus was performed on VeroE6/TMPRSS2 cells. Both viruses were titrated using a plaque assay in which monolayers of VeroE6/TMPRSS2 cells were infected with dilutions series prepared in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 2% fetal bovine serum (FBS) in duplicate for 2 vials of each virus. Two hours after infection, Avicel was added to a final concentration of 0.3% (w/v). Dose-dependent neutralization was assessed by mixing the constructs at different concentrations (5-fold serial dilutions) with 40 PFU of SARS-CoV-2, followed by incubation of the mixture at 37 °C for 1 h. The VHH-virus mixes were then added to VeroE6/TMPRSS2 cell monolayers in 24-well plates and incubated at 37 °C for 1 h. Subsequently, Avicel was added to a final concentration of 0.3% (w/v). After 2 days of incubation at 37 °C, the overlays were removed, and the cells were fixed with 3.7% paraformaldehyde (PFA) and stained with 0.5% crystal violet. Half-maximum neutralization titers (PRNT₅₀) were defined as the VHH or VHH-Fc concentration that resulted in a plaque reduction of 50% across two independent plates.

Analysis of binding kinetics and stoichiometry by biolayer interferometry

The SARS-CoV-2 spike 2 P (S2-P) binding kinetics of R3DC23 monomer were assessed by biolayer interferometry on an Octet RED96 system (FortéBio). Streptavidin (SA) biosensors (Sartorius) were soaked in 1x kinetics buffer (10 mM HEPES pH 7.4, 150 mM NaCl, 0.1 mg/mL bovine serum albumin, 0.02% Tween-20 and 0.02% sodium azide) for 20 min before use. Trimeric S2-P-foldon-His-Strep (in-house HEK293S produced) was thawed from −20 °C and incubated at 37 °C for 24 h followed by 30 °C for 30 min, before immobilizing on the SA biosensors at 36.5 µg/mL (80 nM) for 450 s, to a signal of 1.5–1.7 nm. Association (120 s) and dissociation (480 s) of a two-fold dilution series starting from 20 nM monomeric R3DC23 VHH in 1x kinetics buffer were measured at 30 °C. Between analyses, biosensors were regenerated by three times 5 s exposure to regeneration buffer (10 mM glycine pH 3). The response in nm for stoichiometry analysis was determined from the raw data curves. Data for kinetics analysis were double reference-subtracted and aligned to each other in Octet Data Analysis software v9.0 (FortéBio). Association and dissociation of the full dilution series were fit in a global 1:1 model. In Fig. 1f, the blue lines are the sensorgrams for the indicated concentrations of R3DC23 and the black lines are the fitted curves. The poor fitting of the association sensorgram using a 1:1 model suggests that most likely the spike trimer is bound by multiple R3DC23 VHHs. If a ligand is captured to a biosensor in such a way it is fully available for analyte binding, stoichiometry of the interaction between ligand and analyte can be described upon saturation of the captured ligand as:

Generation of R3DC23-Fc(YTE)

A humanized (Q1D, Q5V, A13P, D15G, T19R, M63V, S73N, T84L, K83R and Q98L; according to Kabat numbering) version of R3DC23 was fused via a (G₄S)₂ linker to a human IgG1 Fc (EPKSCdel_YTE_K447del) and ordered synthetically at IDT as gBlocks. Upon arrival, gBlocks were solubilized in ultraclean water at a concentration of 20 ng/µL. gBlocks were A-tailed using the NEBNext-dA-tailing module (NEB), purified using CleanPCR magnetic beads (CleanNA) and inserted in pcDNA3.4-TOPO vector (Thermo Fisher Scientific). The ORF of positive clones was fully sequenced, and pDNA of selected clones was prepared using the NucleoBond Xtra Midi kit (Machery-Nagel).

Production of YTE variants of VHH-Fc in mammalian cells

HEK293S cells were transfected with VHH-Fc encoding plasmids using poly-ethylene imine (PEI). Briefly, suspension-adapted and serum-free HEK293S cells were seeded at 3 × 10⁶ cells/mL in Freestyle-293 medium (Thermo Fisher Scientific). Next, 4.5 μg/mL of pcDNA3.4-VHH-Fc plasmid DNA was added to the cells and the cells were incubated on a shaking platform at 37 °C and 8% CO₂, for 5 min. Next, 9 μg of PEI/mL was added to the cultures, and cells were further incubated for 5 h, after which an equal culture volume of Ex-Cell-293 (Sigma) was added to the cells. Transfected cells were incubated for 4 days, after which the cells were pelleted (10 min, 300 × g) and supernatants were filtered before further use.

Additionally, huR3DC23-Fc_YTE was expressed in ExpiCHO-S^TM cells (Thermo Fisher Scientific), according to the manufacturer’s protocol. Briefly, a 50 mL culture of 6 ×106 cells per mL, grown at 37 °C and 8% CO₂, was transfected with 40 µg of pcDNA3.4-huR3DC23-Fc_YTE plasmid DNA using ExpiFectamine^TM CHO reagent. One day after transfection, 300 µL of ExpiCHO^TM enhancer and 8 mL ExpiCHO^TM feed was added to the cells, and cultures were further incubated at 32 °C and 5% CO₂. Cells were fed a second time on day 5 after transfection. Productions were collected as soon as cell viability dropped below 75%.

For purification of the VHH-Fc proteins, supernatants were loaded on a 5 mL MAbSelect SuRe column (Cytiva). Unbound proteins were removed by a wash step with McIlvaine buffer pH 7.2, and bound proteins were eluted using McIlvaine buffer pH 3.0. Immediately after elution, protein-containing fractions were neutralized with 30% (v/v) of a saturated Na₃PO₄ buffer. Next, these fractions were pooled, and loaded on a HiPrep Desalting column for buffer exchange to PBS pH 7.4.

Hydrophobic interaction chromatography (HIC) assay

Apparent hydrophobicity was assessed using a hydrophobic interaction chromatography (HIC) assay employing a Dionex ProPac HIC-10 column, 100 mm ×4.6 mm (Thermo Fisher Scientific, catalog nr. 063655), containing a stationary phase consisting of a mixed population of ethyl and amide functional groups bonded to silica. All separations were carried out on an Agilent 1100/1260 HPLC equipped with a UV/VIS detector. The column temperature was maintained at 25 °C throughout the run and the flow rate was 0.8 mL/min. The mobile phases used for HIC were 1.6 M ammonium sulfate and 50 mM phosphate pH 7.0 (buffer A), and 50 mM phosphate pH 7.0 (buffer B). Protein and calibrator samples were diluted 1:1 with buffer A and injected onto the column. Following a 5 min hold at 50% B, bound protein was eluted using a linear gradient from 50 to 100% B in 50 min followed by 5 min hold at 100% B. The column was washed with 100% B, followed by 50 mM ammonium acetate pH 5.0 and re-equilibration in 50% B for 10 min prior to the next sample. The separation was monitored by absorbance at 280 nm with a 16 nm bandwidth, without reference subtraction.

SARS-CoV-2 infection in K18-hACE2 Tg mice

K18-hACE2^Tg/+ mice were obtained from The Jackson Laboratory (Strain #034860). Male and female K18-hACE2^Tg/+ mice were housed under specific pathogen-free conditions before they were transferred to the Biosafety level 3 laboratory at VIB-UGent Center for Medical Biotechnology. In the BSL-3 laboratory, mice were housed with a 16 h light (6 A.M.–10 P.M.), 8 h dark (10 P.M.–6 A.M.) cycle, at room temperature (21 °C) and 45-65% humidity. Mice were used between 8 and 11 weeks of age. All experiments were approved by the animal ethics committee at Ghent University (EC2020-125) and were in accordance with Belgian animal protection law. Mice received 150 µg huR3DC23-Fc via an intraperitoneal injection in 200 µL. The irrelevant antibody palivizumab (anti-RSV monoclonal antibody) was used as a negative control. Twenty-four hours later, animals were anesthetized by isoflurane inhalation and 3*10^2 PFU of SARS-CoV-2 614 G (SARS-CoV-2/human/FRA/702/2020, obtained from the European Virus Archive) was administered by intratracheal instillation. Each treatment group contained 5 mice. Animals were monitored on a daily base by a blind observer who measured weight change and scored for humane endpoints (hunchback (1 point), piloerection (1 point), less movement upon opening cage (1 point), motionless upon touching (2 points), neurological symptoms (shaking, balance, 3 points), heavy breathing (3 points)). Mice that lost more than 25% of their initial bodyweight or reached a humane endpoint with a score of 5 points were euthanized.

In a separate experiment, mice were sacrificed with an overdose of pentobarbital on day five post infection to measure viral replication and inflammatory cell influx in the lung. The left lung lobe was collected in PBS and frozen for virological analysis. The right lung lobes were isolated, cut with scissors and then digested for 30 min in RPMI-1640 (Gibco, Thermo Fisher Scientific) containing 20 µg/mL liberase TM (Roche), 10 U/mL DNase I (Roche), and 5% of FCS (Bodinco) at 37 °C. Next, lungs were filtered through a 70 µm cell strainer. For quantification of the viral lung load, the left lung lobe was homogenized using a Precellys Tissue Homogenizer (Bertin) and 2 mL Precellys tubes. After centrifugation (1000 × g for 5 min), 5-fold dilution series of the supernatant were prepared in duplicate and added to a monolayer of VeroE6/TMPRSS2 cells in 24-well plates and incubated at 37 °C for 1 h. Subsequently, Avicel was added to a final concentration of 0.3% (w/v). After 2 days of incubation at 37 °C, the overlays were removed, the cells were fixed with 3.7% paraformaldehyde (PFA) and stained with 0.5% crystal violet. Per dilution series, the viral plaques were counted in at least 2 wells and used to calculate the number of PFU. Finally, for each mouse the PFU per g was calculated as the mean of 2 duplicates. For the quantification of the viral RNA, 150 µL of cleared lung homogenate supernatant was used to prepare RNA using the Macherey-Nagel NucleoSpin RNA Virus, Mini kit for viral RNA. cDNA was prepared using random hexamer primers (Roche Transcriptor First strand cDNA synthesis kit) and used to performing qPCR using the following Nucleocapsid specific primers and Roche probe: FW primer 5’-TTA CAA ACA TTG GCC GCA AA-3’, RV primer 5’-TTA CAA ACA TTG GCC GCA AA-3’ and probe 5’-FAM-ACA ATT TGC CCC CAG CGC TTC AG-BHQ1-3’.

The following statistical tests were used to test statistically significant differences between groups: Log-rank, Mantel-cox test for animal survival (shown in Fig. 2j, left panel), two-way ANOVA with Tukey’s multiple comparisons test for mean relative bodyweight (Fig. 2j, right panel), Kruskal-Wallis test with Dunn’s multiple comparisons test for lung viral titers (Fig. 2k), one-way ANOVA with Tukey’s multiple comparisons test when data are normally distributed (AMs, eosinophils, monocytes, T cells and B cells) or Kruskal-Wallis test with Dunn’s multiple comparisons test when data are not normally distributed (DCs, neutrophils) (Fig. 2l).

Flow cytometric analysis of the leukocyte influx in the lungs

Lung single cell suspensions were incubated with the eBioscience™ Fixable Viability Dye eFluor™ 506 (Thermo Fisher Scientific, 1/300) to identify dead cells. Fc Block 2.4.G2 (Bioceros, 1/800) was used to prevent aspecific antibody binding. Cell surface markers were stained for 30 min at 4 °C in the dark.

Lung single-cell suspensions were stained for flow cytometry using PerCP/Cy5.5-conjugated CD88 (clone 20/70; BioLegend, catalog number 135812, 1/200), eFluor450-conjugated Ly-6C (clone HK1.4; eBioscience, catalog number 48-5932-82, 1/500), BV605-conjugated CD11b (cline M1/70; BioLegend, catalog number 101237, 1/800), BV650-conjugated Ly-6G (clone 1A8; BD Biosciences, catalog number 740554, 1/400), BV711-conjugated CD64 (clone X54-5/7.1; BioLegend, catalog number 139311, 1/100), BV786-conjugated Siglec-F (clone E50-2440; BD Biosciences, catalog number 740956, 1/400), AF700-conjugated CD45 (clone 30-F11, BioLegend, catalog number 103127, 1/800), APC-eFluor780 conjugated anti-I-A/I-E (clone M5/114.15.2; Thermo Fisher Scientific, catalog number 47-5321-82, 1/1000), PE-Cy5-conjugated CD3e (clone 145-2c11; Thermo Fisher Scientific, catalog number 15-0031-81, 1/200), PE-Cy5-conjugated NK1.1 (clone PK136, BioLegend, catalog number 108713, 1/200), PE-Cy5-conjuaged CD19 (clone 1D3; Thermo Fisher Scientific, catalog number 15-0193-82, 1/400) and PE-Cy7-conjugated CD11c (clone N418; Thermo Fisher Scientific, catalog number 25-0114-82, 1/800).

After staining, lung cells were washed and fixation was performed using BD cytofix fixation buffer (BD Biosciences) for 20 min. Samples were washed and 123count eBeads™ Counting Beads (Thermo Fisher Scientific) were added to each sample to determine absolute cell numbers. Settings were calibrated using UltraComp eBead™ Compensation Beads (Thermo Fisher Scientific). Data were acquired and analyzed on a NovoCyte Quanteon flow cytometer (Agilent). All procedures were performed in BSL-3 conditions. The gating strategy for data analysis is shown in Supplementary Fig. 13B.

Production of huR3DC23-Fc_LS following transient transfection

The gene encoding huR3DC23-Fc_LS was codon optimized, synthesized, and cloned into the pXLG6 backbone vector at ATUM’s laboratories. Upon gene and codon optimization, the R3DC23 DNA sequence was inserted into pXLG6 expression vector and transfected in CHOExpress™ cells at a cell density of 4 ×106 cells/mL. TGE supernatant was harvested by centrifugation and clarified by filtration (0.2 µm) after 10 days when cell viability dropped below 90%. The protein was further purified by Protein A affinity chromatography.

Fed batch production of huR3DC23-Fc_LS from stable pool at 1 L scale

The gene encoding huR3DC23-Fc_LS was codon optimized, synthesized, and cloned into the pXLG6 backbone vector at ATUM’s laboratories. Upon expansion to a density of about 4 ×106 cells/mL, parental CHOExpress™ cells were co-transfected with the expression vector and the pXLG5 helper vector. The stable pool was generated under 50 mg/L puromycin selective pressure (applied daily) and further expanded. The stable pool research cell bank was banked at day 14 when cell viability reached 95%.

The RCB pool was then expanded for protein production at 1 L scale and cultured until day 12 (cell density 3.5 ×107 cells/mL, cell viability 96%). The supernatant was harvested by centrifugation and clarified by filtration (0.2 µm). The protein was further purified by Protein A using MabSelect SuRe LX resin. Consecutive washed were performed with 20 mM sodium phosphate and 110 mM NaCl at pH 7.2; 100 mM sodium acetate and 500 mM NaCl at pH 5.5; and 20 mM sodium phosphate at pH 7.2. The eluate in 100 mM sodium acetate pH 3.5 was neutralized to pH 7.0 by addition of 1 M Tris pH 11.0 (10%_v/v). After filter sterilization (0.22 µm), the protein was aliquoted at 2 mg/mL.

Protein preparation for biophysical analysis of Fc-fusions

Preceding biophysical characterization, MAbSelect SuRe-purified protein samples were further purified via size-exclusion chromatography (SEC) on a 12 °C-cooled Superdex 200 column (Cytiva) equilibrated with either phosphate-buffered saline (PBS, Sigma-Aldrich) supplemented with 0.02% sodium azide to prevent microbial growth, or a sample buffer comprising 50 mM L-histidine and 150 mM L-arginine (Sigma-Aldrich), 0.02% polysorbate-20 and 0.02% sodium azide, set to pH 7.0 at 25 °C. After filter sterilization (0.22 µm), 1 mg/mL aliquots were snap-frozen in polypropylene tubes in liquid nitrogen and stored at −80 °C.

Sample composition

Purified VHH-Fc samples were characterized by analytical SEC to determine the molecular composition of each sample. After rapid thawing in a water bath at 25 °C, 10 min centrifugation at 16,000 × g and transfer of supernatant to fresh tubes, 5 µg was injected on an AdvanceBio SEC column, 4.6 ×300 mm (Agilent) with 2.7 µm porous particle size and 300 Å pore size, calibrated with PBS. The separation was monitored by absorbance at 280 nm with a 16 nm bandwidth, without reference subtraction. For additional quality control, proteins were separated on reducing 15% SDS-PAGE with Coomassie staining.

Acquisition of sotrovimab (without LS), cilgavimab, bebtelovimab biosimilars and palivizumab

Bebtelovimab biosimilar (PX-TA1750), cilgavimab biosimilar (PX-TA1033) and sotrovimab biosimilar (without LS) (PX-TA1637) were purchased from Proteogenix. Clinical grade palivizumab was obtained from the Ghent University hospital.

Neutralization assays using authentic SARS-CoV-2 to test the neutralizing activity of huR3DC23-Fc_LS

SARS-CoV-2 viruses belonging to different lineages (D614G, Delta, Omicron BA.1, Omicron BA.2 and Omicron BA.5) were isolated from nasopharyngeal swabs taken from patients/travelers between January 2020 and July 2022. More specifically, the following clinical isolates were used: SARS-CoV-2 Isolate BavPat1/2020/Germany (09 Feb 2020); Delta variant SARS-Related Coronavirus 2, Isolate hCoV-19/USA/MD-HP05647/2021; Omicron BA.1 variant SARS-CoV-2 hCoV-19/Netherlands/NH-RIVM-72291/2021; Omicron BA.2 variant Clinical isolate hCoV-19/Netherlands/VCB-20220303-1/2022; and Omicron BA.5 variant Clinical isolate hCov19/NL/VCB-20220714-2/2022. Dose-dependent neutralization of the test item (huR3DC23-Fc_LS), the positive controls (bebtelovimab biosimilar, cilgavimab biosimilar, sotrovimab biosimilar (without LS) and a negative control (isotype control) were assessed in an authentic virus neutralization assay. For all assays in which authentic D614G, Delta, BA.1, and BA.2 SARS-CoV-2 were tested, huR3DC23-Fc_LS produced from transiently transfected cells was used. For all assays in which SARS-CoV-2 BA.5 was tested, huR3DC23_Fc_LS produced from stable cell pools was used. For each variant, three independent runs were performed. Different system controls were included in the assay: cell only (medium only), virus only, and an internal positive control (human serum). Briefly, 5-fold or 7-fold serial dilutions of the test items and controls were incubated with a fixed amount of plaque-forming units (PFUs) of the virus for 1 h at room temperature. Afterwards, the Vero E6 cell monolayer was inoculated with virus-antibody mixtures for 1 h at 37 °C. In a next step, the inoculum was removed and cells were incubated at 37 °C with infection medium (up to 18–24 h post-infection). Afterwards, the SARS-CoV2 infected cells were fixed and immunostained with a SARS-CoV Nucleocapsid Antibody (Sino Biological, catalog number: 40143-MM05, clone #5), followed by HRP-conjugated Goat anti-Mouse IgG (H + L) Secondary Antibody (Invitrogen, catalog number A16072). Spots (infected cells) were counted using an Immunospot Image Analyser. For each test item, the compound concentration showing 50 % reduction in infection (IC₅₀) was calculated based on the Zielinska method. The geometric mean values were calculated based on three independent runs. (ND) Not possible to determine IC50 value for sotrovimab (without LS) within the tested concentration range (0.128-10 000 ng/mL). (NT) Cilgavimab was not tested for Omicron BA.5. (*) The IC50 of cilgavimab for BA.1 is based on 2 biological replicates in which the highest concentration of cilgavimab (500 ng/mL) resulted in a more than 50 % reduction of viral replication but not in a third biological replicate. (Sotrovimab biosimilar) = sotrovimab biosimilar without LS mutation.

SARS-CoV-2 challenge model in Syrian golden hamsters

In brief, 9-to 10-weeks-old male Syrian golden hamsters (Mesocricetus auratus) weighing 89.8–132.3 g were obtained from Janvier (France). Housing conditions and experimental procedures were approved by the ethics committee in the Netherlands (study was registered under number 27700202114492-WP49). The hamsters were housed according to SOP VCX-P073 (Animal housing and welfare management) in elongated type 2 IVC group cages with one or two animals per cage under BSL-II conditions during acclimatisation and in elongated type 2 group cages under BSL-III conditions (isolators) during challenge. Six hamsters per group were infected intranasally with 100 TCID₅₀/dose SARS-CoV-2 (Wuhan strain) in a total dose volume of 100 µL, divided equally over both nostrils. The test item huR3DC23-Fc_LS (2 and 10 mg/kg), palivizumab (10 mg/kg) and bebtelovimab (10 mg/kg) were administered by intraperitoneal injection 4 h after the SARS-CoV-2 challenge. The irrelevant antibody palivizumab (anti-RSV monoclonal antibody) was used as a negative control, while bebtelovimab was used as a positive control. The huR3DC23-Fc_LS used in the hamster study was produced from stable cells pools. Hamsters were monitored daily for behavior, appearance and body weight.

On day 4 post-infection, animals were sacrificed. At the time of necropsy, gross pathology was performed and abnormalities were recorded. Samples from the right lung lobes were collected and frozen for virological analysis. To determine virus titers, quadruplicate 10-fold serial dilutions were used in confluent layers of Vero E6 cells. To this end, serial dilutions of the samples (lung tissue homogenates) were incubated on Vero E6 monolayers for 1 h at 37 °C. Vero E6 monolayers were then washed and incubated for 5 or 6 days at 37 °C, after which plates were stained and scored based on cytopathic effect (CPE) by using the vitality marker WST8 (colorimetric readout). Viral titers (Log₁₀ TCID₅₀/g) were calculated using the Spearman-Karber method. For the viral titration, the lower limit of detection (LLOD) ranged between 1.1 and 1.3 log₁₀ TCID₅₀/g. To detect viral RNA, lung tissue and homogenates were used. RNA was isolated and Taqman PCR was performed. The number of copies (Log₁₀ CP/g) in the different samples was calculated against a standard included in each run. For the viral RNA, the LLOD was 3.5 Log₁₀ CP/g. In the graphs in Fig. 2h, the dotted horizontal lines indicate the LLOD.

Blood samples were collected prior to the start of the study on day -2 ( ~ 200 μl of blood was collected for serum preparation under isoflurane anesthesia) and on day 4 post-infection (p.i.), at time of necropsy for pharmacokinetic analysis. Blood samples for serum preparation were immediately transferred to appropriate tubes containing a clot activator. Serum was collected and stored frozen. To inactivate potential infectious material present and to allow the testing of the sera samples in a BSL-2 environment, day 4 post-infection sera was heat-treated at 56 °C for 30 min.

One-way ANOVA with Dunn’s multiple comparisons test was used to test statistical differences in TCID50/gram lung tissue and viral RNA copies/gram lung tissue between treatment groups in hamsters (Fig. 2n).

To assess whether there was a difference in hamster body weight (BW) at any time during the follow up of 4 days, we used the per hamster normalized BW. This was calculated as the BW at each of the 4 days day minus the BW 2 days before the start of the experiment. This normalized BW was the outcome variable in a linear mixed model, run in R⁵³ using the nlme^54,55 and splines packages. As predictor variables, we used the time (in days) and the treatment group (positive control, negative control, VHH-Fc low and VHH-Fc high dose groups). The time was modeled with a natural spline with 1 internal knot (df = 2) at day 2.5. An interaction between time and treatment group allowed for different courses of BW in the different treatment groups. Further, we allowed for a random intercept and a random slope for time in each hamster to model the within-hamster correlation between the time points. We obtained confidence intervals for pairwise differences in BW between the negative control group and all other treatment groups using the contrast⁵⁶ and rms⁵⁷ packages. The intervals were calculated with robust variance-covariance estimators and corrected for multiple testing with the multcomp⁵⁸ package.

ELISA for detection of huR3DC23-Fc_LS in hamster sera

Pharmacokinetic analysis (PK) of the hamster serum samples was done using an ELISA based assay. Streptavidin-coated 96-wells microtiter plates were pre-blocked with superblock T20 (Thermo Fisher Scientific, catalog number 37516) at room temperature. Afterwards, the plates were washed 3 times with 200 µL of washing buffer (PBS with 0.05% Tween20). Biotinylated anti-VHH monoRab monoclonal antibody (Genscript, catalog number A01995-200) was captured at a concentration of 0.1 μg/mL (10 ng/well in 0.1 mL) onto the streptavidin plate for 2 h at room temperature with gentle shaking (400 rpm). Next, the supernatant was removed, and plates were washed three times with washing buffer. As a next step, 100 µL of 1.7-fold dilution series of standard calibrator (huR3DC23-Fc_LS), freshly prepared quality control samples (QCs) or hamster serum samples were added to the plates (according to a minimal required dilution of 10 or 50), which allowed huR3DC23-Fc_LS to bind to the captured anti-VHH antibody. After 1 h incubation with shaking at room temperature, plates were again washed three times with washing buffer. To detect bound huR3DC23-Fc_LS onto the plates, anti-VHH monoRab antibody conjugated to horse-radish peroxidase (Genscript, catalog number A01861-200) was added at a concentration of 0.2 µg/mL (0.1 mL per well) and incubated for 30 min at room temperature with shaking. Plates were washed three times with PBS buffer to remove the remaining Tween20. The final step was the addition of 100 µL of 1-Step Ultra TMB-ELISA Substrate solution (Thermo Fisher Scientific, catalog number 34029) into wells for 5 min to allow a colorimetric reaction. Following the color development, the reaction was stopped by adding 50 µL of Stop solution (Thermo Fisher Scientific, Catalog number N600). The colorimetric output was read on a Tecan Spark instrument with SparkControl software.

S1 shedding assay

Antibody or VHH was added at a final concentration of 10 µg/mL to 500 000 HEK293T cells that were previously transfected with a SARS-CoV-2 spike (D614G)-expressing plasmid, or an empty control vector. The antibody-cell mixture was incubated for 30 min at 37 °C and 5% CO₂. After incubation, cells were pelleted by centrifugation, supernatant was transferred to a fresh tube and the cell pellet was lysed with 250 µL of RIPA lysis buffer (50 mM Tris-HCl pH 8.0, 100 mM NaCl, 1 mM EDTA, 1 mM EGTA, 0.1% SDS, 1% NP-40). Twenty µL samples of supernatant and lysate were separated on 8% SDS-PAGE gels, and electroblotted onto nitrocellulose membranes. Membranes were blocked with 4% milk, stained with rabbit anti-SARS-S1 antibody (1/1000, Sino Biological, 40591-T62) followed by anti-rabbit IgG-HRP (1/2000, Cytiva, NA934V) and developed using Pierce™ ECL Western Blotting Substrate (Thermo Fisher Scientific).

Fusion inhibition assay using spike-expressing Vero E6 cells

Vero E6 cells were transfected with a GFP expression vector together with either a control expression vector (no spike) or a SARS-CoV-2 spike expression vector using Fugene (Promega). Two hours after transfection, PBS, monoclonal antibodies, or VHHs were added to a final concentration of 10 μg/mL. Twenty-two or forty hours later, the cells were imaged using an Olympus fluorescence microscope using a 10x lens. Alternatively, monoclonal antibodies or VHHs were added to transfected cells at a final concentration of 1 µg/mL and GFP expression was monitored hourly over time with an Incucyte Zoom live cell analysis device (Sartorius) and analyzed with the provided Incucyte software. To derive fusogenicity of spike variants, the mean slope of the curve derived from the area of green objects at time points 24 h after transfection (maximum GFP signals in negative control cells, indicating maximal GFP expression) and 38 h after transfection (maximal size of syncytia, before cell death) was calculated.

Fusion inhibition assay using replication-competent GFP report VSV virus pseudotyped with Wuhan SARS-CoV spikes(del-18)

VeroE6/TMPRSS2 cells were infected with 40 PFU of SARS-CoV-2 spike GFP-expressing pseudotyped replication-competent VSV-GFP virus, generously provided by Dr. Florian Schmidt²¹. Two hours later the indicated monoclonal antibodies or VHHs were added. Non-infected cells were used as negative controls. Infected cells were incubated overnight and imaged with an Olympus fluorescence microscope using a 4x lens. GFP fluorescence was measured with a fluorimeter. Of note, different from the clone of replication-competent pseudotyped VSV particles used in the neutralization assays, the furin cleavage site of the virus used in the fusion assays was intact as confirmed by Sanger sequencing.

Fusion inhibition assay using split GFP

HEK293T cells in a 96 well plate were transfected with a GFP11 expression vector (Addgene, plasmid #68716), and HEK293T cells in a 6 well plate were cotransfected with a GFP1-10 expression vector (Addgene, plasmid #68715) and a surface-expressed SARS-CoV-2 spike expression vector, or an empty vector. One day after transfection, the HEK293T cells expressing GFP1-10 and the spike protein were detached with trypsin, incubated for 15 min at room temperature with CB6 (which destabilizes the spike by arresting the RBD in an up conformation and as such induces S1 shedding and subsequent cell-cell fusion was induced by CB6) or with a control antibody, and subsequently incubated with VHH or antibody at a final concentration of 10 µg/mL, for 30 min at room temperature. After incubations, the cells were added to the HEK293T cells that express GFP11. Cell-cell fusion was monitored by measuring the GFP fluorescence with a Tecan infinite 200 pro plate reader 16 h after co-incubation of the cells.

Viral escape selection

Monolayers of VeroE6/TMPRSS2 cells seeded in 96 well plates were infected with 200 PFU of GFP expressing replication-competent VSV virus particles pseudotyped with the SARS-CoV-2 spike protein containing an intact furin cleavage site, generously provided by Dr. Florian Schmidt²¹. Two hours after infection, 10 μg/mL of VHH R3DC23 was added to 59 wells, and as control 1 well was treated with PBS. Rapid syncytium formation and spread of the infection to neighboring cells was observed in the PBS treated well, whereas these events were blocked or strongly delayed in wells with R3DC23. From the growth medium of wells that displayed syncytia formation or viral replication in the presence of VHH R3DC23, nine single clones were isolated by limiting dilution on fresh VeroE6/TMPRSS2 cells seeded in 96 well plates. Growth medium of wells with a single PFU was used to propagate possible escape viruses on monolayers of VeroE6/TMPRSS2 cells seeded in 6 well plates in the presence of VHH R3DC23. From these infected cells, RNA was prepared using a nucleospin RNA virus kit (Macherey Nagel Bioanalysis) and used to generate cDNA using random hexamer primers. This cDNA was used to amplify the spike coding sequences by PCR. The resulting PCR fragments were purified and sequenced using Sanger sequencing. The obtained nucleotide sequences were analyzed and aligned to spike proteins of WT SARS-CoV-2 and clade 1, 2, and 3 sarbecoviruses using CLC Main Workbench 20.0.4. Each of these viral clones had acquired a single amino acid substitution at one out of four positions: N1192D (1 clone), L1197P (2 clones), L1200P (1 clone), Q1201R (4 clones), and Q1201K (1 clone). Mutations were visualized on a model of full-length glycosylated spike protein obtained from Charmm-gui.org (PDB: 6VXX_1_1_1 model) or the SARS-CoV-2 HR2 coiled-coil as determined by NMR (PDB: 2FXP) using Pymol.

Growth kinetics of viral escape variants

Vero E6 cells seeded in a 96 well-plate were infected with 50 PFU of GFP-expressing replication-competent VSV virus particles pseudotyped with the SARS-CoV-2 spike protein, that were obtained during escape selection. GFP-expression was monitored hourly with an Incucyte Zoom device and analyzed with accompanying software.

HDX-MS epitope mapping

SARS-CoV-2 S-2P trimer at 3.33 μM was incubated overnight at 37 °C. The protein was then diluted to 1.66 μM trimer in the presence or absence of 6.25 μM R3DC23 in 1x PBS (pH 7.4, Sigma-Aldrich P4417). To initiate exchange, the protein was diluted tenfold into 25 °C temperature-equilibrated deuterated buffer made by lyophilizing 1x PBS and resuspending in D₂O (Sigma-Aldrich 151882). Samples were quenched at each time point (15 s, 3 min, 30 min, 3 h) by mixing 60 µL of the exchange reaction with 60 µL of ice-cold 2x quench buffer (3.6 M guanidinium chloride, 500 mM TCEP, 200 mM glycine pH 2.4). The quenched samples were incubated on ice for 1 min and then flash frozen in liquid nitrogen and stored at −80 °C until LC-MS. LC-MS and data analysis was conducted as previously described²⁷. The glycosylated peptide has taken up more deuterium than the number of backbone exchangeable sites (11 sites) because the glycan can uptake and retain deuterium at amide sites similarly to the backbone as noted by Guttman et al.⁵⁹. No bimodal spectra were noted outside of populations <5% in regions previously described in Costello et al.²⁷ which are associated with the prefusion and open-interface trimer conformational change. One biological replicate (same VHH preparation, other S-2P preparation) was performed for R3DC23 bound to S-2P in the prefusion conformation.

To allow access to the HDX data of this study, the HDX data summary table (Supplementary Table 1) and the HDX data tables (included in the Source Data file) are included as per consensus guidelines⁶⁰.

Generation of 5HB and 6HB constructs

For the 5HB construct, the following sequence was ordered synthetically at IDT as a gBlock: the transmembrane domain and part of the stalk of an influenza NA mutant that assembles in monomers instead of tetramers⁶¹, followed by HR1 (DNA encoding AA911-988), a SGGRGG linker, HR2 (DNA encoding AA1162-1206), a GGGGKGGSG linker, HR1, SGGRGG linker, HR2, GGGGKGGSG linker, HR1, GGGGS linker and finally an HA tag. This gBlock was cloned into the pcDNA^TM 3.4 TOPO^TM vector with the pcDNA^TM 3.4 TOPO^TM TA cloning kit (Thermo Fisher Scientific) according to the manufacturer’s instructions. A secreted version of such a 5HB was shown to bind to HR2 peptides and to neutralize SARS-CoV-2 with µM-range potency (31). Similarly, we observed binding of monomeric SUMO-HR2 fusion proteins but not SUMO-HR1 fusion proteins to 5HB expressing cells, demonstrating correct folding of the 5HB (Supplementary Fig. S15). To create the 6HB construct, an extra HR2 was cloned into this construct, after the third HR1 repeat, with Gibson assembly. In brief, a gBlock encoding the last residues of HR1 (KVE), the SGGRGG linker and HR2 sequence was ordered synthetically at IDT. Both the HR2 gBlock and existing 5HB were amplified with PCR using primers that created homologous regions at the 3’ and 5’ ends. Finally, the insert and vector were joined using the NEBuilder® HiFi DNA assembly kit (New England Biolabs). The ORFs were sequence-verified with Sanger sequencing.

HR2 expression and purification

For structural biology purposes the HR2 protein was expressed in a bacterial expression system. Therefore, the synthetic gene encoding residues H1159-K1211 of the SARS-CoV-2 spike was cloned into a pFloat-SUMO vector, generating a His-tagged SUMO-HR2 fusion protein. The construct also contained a 3C protease cleavage site to remove the His-SUMO-tag. The pFloat-SUMO-HR2 plasmid was transformed in BL21(DE3) cells and plated on kanamycin (100 µg/mL) containing LB agar plates. A small LB culture, supplemented with 100 µg/mL kanamycin, was inoculated with a single colony of BL21(DE3)(pFloat-SUMO-HR2) and grown overnight at 37 °C. One L LB cultures were subsequently inoculated with 20 mL of this preculture and grown at 37 °C until OD₆₀₀ reached 0.8. At this point protein expression was induced by adding 0.5 mM IPTG to the E. coli culture. Cells were incubated further overnight at 20 °C and subsequently harvested by centrifugation (Beckman rotor 8.1000, 5000 rpm (4752 g), 15 min, 4 °C). The pellet was resuspended in PBS, 500 mM NaCl, 10 mM imidazole, 5 mM ß-mercaptoethanol, 0.1 mg/mL 4-(2-aminoethyl) benzenesulfonyl fluoride hydrochloride (AEBSF), 1 µg/mL leupeptine, 50 µg/mL DNaseI and 20 mM MgCl₂. The cells were lysed using a French press (Constant Systems) at 20 kpsi and the cell debris was removed by centrifugation. The cell lysate was loaded on a Ni-sepharose FF HiLoad column (GE Healthcare), equilibrated in 20 mM Tris-HCl pH 7.5, 500 mM NaCl, 10 mM imidazole, 5 mM ß-mercaptoethanol. The bound proteins were eluted using a linear gradient to 500 mM imidazole. Fractions containing the His-SUMO-HR2 protein were pooled and dialyzed overnight to 20 mM Tris-HCl pH 7.5, 150 mM NaCl at 4 °C, followed by 2 h incubation with 3 C protease at room temperature. The cleaved sample was loaded again on a Ni-sepharose FF HiLoad column, equilibrated in the same buffer. The flow through, containing the HR2 protein, was concentrated and applied to a BioRad Enrich70 10/30 size exclusion column (SEC), equilibrated in 20 mM Tris-HCl pH 7.5, 150 mM NaCl. The HR2-containing SEC fractions were pooled.

Generation, production and purification of HR2-foldon constructs

For the HR2-foldon constructs, the following sequences were ordered synthetically at IDT as gBlocks: the SARS-CoV-2 signal peptide, followed by the HR2 sequence (in Wuhan numbering: for the SARS-CoV-2 HR2 (QELGKYEQ-GS-GYIPE): AA1159-1208, for the SARS-CoV-2 HR2 (QELGKY-IPE): AA1159-1206), for the SARS-CoV-2 HR2 (QEL-QEAQGYIPE): AA1159-1203), for the MERS-CoV HR2: AA1242-1291) linked by a GS linker to the foldon sequence, followed by a PreCission cleavage site, 8x His-tag and 2x StrepII tag. This gBlock was cloned into the pcDNA^TM 3.4 TOPO^TM vector with the pcDNA^TM 3.4 TOPO^TM TA cloning kit (Thermo Fisher Scientific) according to the manufacturer’s instructions. HEK293T cells were transfected with the sequence verified construct for production. Five days after transfection, the supernatant of cells was collected and cleared from cell debris by centrifuging twice (10 min, 400 × g). HR2-foldon constructs were purified with Ni Sepharose^TM High Performance beads (Cytiva), which were first equilibrated with 20 mM NaH₂PO₄ pH 7.5, 0.5 M NaCl, and 20 mM imidazole. Bound HR2-foldon was eluted with 20 mM NaH₂PO₄, 20 mM NaCl and 0.5 M imidazole. The eluted fraction was then buffer exchanged to PBS with 3 kDa cutoff centrifugal units (Amicon).

R3DC23-HR2 crystallization, X-ray data collection, processing, and structure determination

For crystallization, R3DC23 was added to HR2 in 1.2 times molar excess and concentrated to 19 mg/mL using Amicon Ultra 3 kDa cut off centrifugal filter devices. Crystallization screens were set up using the sitting drop vapor diffusion technique, mixing 0.1 µL of R3DC23-HR2 and 0.2 µL bottom solution. Crystals were grown from the Molecular Dimensions Proplex crystallization screen, in 0.1 M magnesium chloride hexahydrate, 0.1 M sodium citrate pH 5.0, and 15% PEG4000. For X-ray data collection, crystals were flash frozen in liquid nitrogen. X-ray data were collected on the i24 beamline at the Diamond Light Source synchrotron facility (Didcot, UK). X-ray data were processed using autoPROC+Staraniso^62,63. The structure of the R3DC23-HR2 complex was solved using the automatic molecular replacement workflow in the CCP4 cloud⁶⁴. The initial model was further build manually in Coot⁶⁵ and refined using phenix.refine⁶⁶ from the Phenix crystallographic software suite⁶⁷. Data collection parameters, as well as processing and refinement statistics are shown in Supplementary Table S2. The electron density map of R3DC23 bound to two HR2 helices is presented in Supplementary Fig. 14.

SARS-CoV-2 spike sequence variant analysis

SARS-CoV-2 spike protein sequences were downloaded from GISAID (N = 17,063,827 sequences available on November 10^th 2024). Only sequences originating from infected human hosts were retained. Very short spike sequences retrieved from incompletely sequences genomes were removed. In the next step, we removed spike protein sequences with a high frequency of undetermined amino acids (X), derived from poor sequencing results (Ns). X frequency threshold was defined as (IQR * 1.5) + Q3 = 5.0% X per protein, where IQR and Q3 are inter-quartile range and the third quartile, respectively. Further, proteins with premature stop codon(s) were excluded. The resulting 13,564,753 quality-controlled spike protein sequences were aligned using the Muscle algorithm in R version 4.1.1 package msa version 1.24.0 with 12 as gap opening and 3 as gap extension penalties and the BLOSUM80 substitution matrix.

To calculate sequence variation, amino acid positions present in the reference spike sequence annotated in NC_045512.2 (Wuhan-Hu-1 isolate, NCBI RefSeq) were extracted from multiple sequence alignment (MSA), disregarding insertions. Conservation matrixes were calculated using Biostrings version 2.60.0 for different subsets of the MSA: sequences derived from the Omicron variant (N = 7,305,876), sequences of Omicron sub-lineages (see figure for respective sequence counts) or sequences obtained before 31/10/2021 (prior to the onset of Omicron, N = 4,697,133). Variant frequencies were derived from conservation matrixes, while undetermined amino acid X frequencies were excluded. Minor variant frequencies were summed to obtain overall sequence variability and calculate conservation percentage as compared to the Wuhan reference.

Sequence variation was overlaid on the surface of spike protein structure model published by the Amaro lab (https://amarolab.ucsd.edu/files/covid19/PSF-PDB_spike_open_prot_glyc_memb_wat_ions_amarolab.tar.gz). R package Bios2cor 2.2.1 was used to calculate Shannon’s entropy score of each position in the MSA, except positions with more than 20 percent of gaps, which were excluded.

HDX-MS analysis of the transition kinetics of prefusion to open-interface spike trimer

3.33 µM SARS-CoV-2 S-2P trimer was incubated at 37 °C for 2 days. The protein was then diluted to 1.66 µM trimer in the presence or absence of 6.25 µM R3DC23 in 1× PBS (pH 7.4, Sigma-Aldrich P4417). The conversion from prefusion to open-interface trimer was initiated by incubating the protein at 4 °C. At 1 h, 3.5 h, 7.5 h, 25 h and 100 h, 6 µL of protein was removed from the incubation, diluted tenfold with deuterated 1× PBS (made as indicated in other HDX method sections), allowed to label for 60 seconds at 25 °C and then quenched with 60 µL of ice-cold 2× quench buffer (3.6 M guanidinium chloride, 500 mM TCEP, 200 mM glycine pH 2.4). The quenched samples were incubated on ice for 1 min and then flash frozen in liquid nitrogen and stored at −80 °C until LC-MS. At 100 h of incubation at 4 °C the final timepoint for measuring the prefusion to open-interface trimer conversion was taken. One experiment was performed. LC-MS and data analysis was conducted as previously described²⁷.

HDX-MS analysis of the transition kinetics of open-interface trimer to prefusion spike

After 100 h of incubation at 4 °C to convert spike from predominantly prefusion to predominantly open-interface trimer, the same aliquot of protein (1.66 µM trimer in the presence or absence of 6.25 µM R3DC23) was moved from 4 °C to 37 °C to initiate the conformational change from open-interface trimer to prefusion. At 10 s, 1 min and 24 h, 6 µL of protein was removed from the incubation, diluted tenfold with deuterated 1× PBS (made as indicated in other HDX method sections), allowed to label for 60 sec at 25 °C and then quenched with 60 µL of ice-cold 2× quench buffer (3.6 M guanidinium chloride, 500 mM TCEP, 200 mM glycine pH 2.4). The quenched samples were incubated on ice for 1 min and then flash frozen in liquid nitrogen and stored at −80 °C until LC-MS. One experiment was performed. LC-MS and data analysis was conducted as previously described²⁷.

Flow cytometric analysis of binding of HIS-SUMO-HR2 or HR2-foldon to cells displaying the extended spike intermediate after antibody-evoked S1 shedding

HEK293T cells were co-transfected with a GFP expression vector and a SARS-CoV-2 spike(del18) expression vector. Two days after transfection, the cells were detached using ice-cold trypsin and subsequently blocked with 1% BSA. Alternatively, when indicated the cells were detached with enzyme-free dissociation buffer and subsequently activated by trypsin. To refold the prefusion spike into the extended intermediates containing the HR1 coiled-coil, the cells were stimulated with 10 µg/mL VHH72_S56A-Fc for 10 min at 37 °C to trigger shedding of S1, or as control stimulated with BSA. This was performed after pre-incubating the cells with either 10 µg/mL huR3DC23-Fc or palivizumab control antibody on ice for 20 min. Refolding to the extended intermediate spike conformation containing the HR1 coiled-coil was monitored by staining with 10 µg/mL HIS-SUMO-HR2, or as control with HIS-SUMO-HR1. Binding of HIS-SUMO-HR2 and HIS-SUMO-HR1 was detected using a mouse anti-HIS antibody and an AF647-conjugated anti-mouse IgG antibody. Binding was calculated as the ratio AF647 MFI of GFP⁺ cells over that of GFP^– cells. Alternatively, when indicated, stable display of the extended intermediate was obtained by pre-treating cells with huR3DC23-Fc (10 µg/mL) for 20 min on ice before inducing S1 shedding by treatment with 10 µg/mL VHH72_S56A-Fc for 10 min at 37 °C. After washing twice, the cells were stained with 5 µg/mL HR2(SARS-CoV-2)-foldon or HR2(MERS)-foldon preincubated with a 2-fold molar excess of either monomeric VHH R3DC23 or a control VHH (GBP). Binding of HR2-foldon was detected using a mouse anti-strep-tag antibody and AF647-conjugated anti-mouse IgG antibody. Binding was calculated as the ratio of the AF647 MFI of GFP⁺ cells over that of GFP^– cells

The following tests were used to test statistical significance between groups: two-way ANOVA with Sidak’s multiple comparisons for binding of SUMO-HR2 to shedded spike when added before S1 shedding (Fig. 8d), one-way ANOVA with Dunn’s multiple comparisons test for binding of SUMO-HR2 (or SUMO-HR1) to the HR1 coiled coil when added after S1 shedding (Fig. 8e), two-way ANOVA with Sidak’s multiple comparisons test for binding of HR2-foldon to the HR1 coiled coil (Fig. 8f).

Prediction of protein structures with AlphaFold-Multimer

The trimer structures were predicted with AlphaFold-Multimer (v2.3.1)⁶⁸. Here, a protein complex was predicted from a FASTA file with the sequence of interest three times. Multiple sequence alignments were computed with the default databases and default HMMER strategy. Each of the five AlphaFold-Multimer models were used to predict five structures, and of the 25 resulting structures the optimal prediction was selected based on the highest predicted template modeling (pTM) score. Finally, Amber relaxation is applied to the final structure.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.