Llama mRNA-LNP immunization enables rapid isolation of target specific nanobodies using cell-based panning

Abstract

Background

As a clinical vaccine platform, lipid nanoparticle (LNP)-formulated mRNA has demonstrated potent and broad antibody responses, prompting speculation about its potential for antibody discovery. Membrane proteins remain among the most challenging targets for antibody development, highlighting the urgent need for technologies that preserve their native conformation during immunization and screening. Nanobodies (V_HHs), the single-domain fragments of heavy-chain-only antibodies naturally found in camelids, offer unique advantages over conventional antibodies, including high solubility, stability, and the ability to access cryptic or membrane-proximal epitopes.

Methods

Here, we report the first demonstration of target specific V_HH discovery in a llama using an mRNA–LNP immunization platform. A llama was immunized with LNPs encapsulating mRNA encoding the severe acute respiratory syndrome coronavirus 2 spike protein, a well-characterized model antigen. The resulting immune V_HH library was screened using a cell-based panning strategy that maintains antigen conformation and enables recovery of binders recognizing native epitopes.

Results

This approach yielded multiple high-affinity spike-specific V_HHs, including two clones that effectively blocked the spike–angiotensin-converting enzyme 2 interaction in a surrogate cell-based assay.

Conclusions

These findings demonstrate the feasibility of mRNA–LNP immunization as a rapid and efficient method for isolating target specific, functionally promising V_HHs from camelids. When combined with conformationally relevant screening, this platform may provide a versatile and accelerated route for generating high-quality V_HHs against complex or recalcitrant targets, particularly membrane proteins, thereby expanding the potential of V_HH-based therapeutics and diagnostics.

Statement of Significance This work demonstrates the feasibility of an mRNA-LNP immunization and cell-based panning strategy for the rapid discovery of camelid nanobodies (V_HHs) recognizing conformational epitopes on native membrane proteins. The approach yielded high-affinity V_HHs capable of blocking spike-angiotensin-converting enzyme 2 interactions in a surrogate cell-based assay, demonstrating its utility for generating binders with functional potential against complex viral antigens.

Introduction

Nanobodies (V_HHs), the variable domains of heavy-chain-only antibodies (HCAbs) naturally found in Camelidae species, have emerged as valuable tools for research, diagnostic, and therapeutic applications. Compared to conventional antibodies (mAbs), V_HHs display several favourable properties [1–4]. In addition to their strong affinity and specificity, they have low immunogenicity, exceptional solubility and stability, and superior tissue penetration. Owing to their small size (~15 kDa) and unique structural features, V_HHs can also access cryptic or membrane-proximal epitopes that are often inaccessible to full-sized Abs. Furthermore, they are readily produced in large quantities in microbial and mammalian systems and can be modularly engineered into multivalent or multispecific formats, enabling a wide range of functional configurations [5, 6].

Despite their promise, current V_HH discovery methods face several limitations. Most protocols rely on the production of soluble recombinant proteins for animal immunization and screening, which can be time-consuming, technically challenging, and poorly suited for quickly emerging infectious pathogens or structurally complex membrane proteins where production and purification of sufficient quantities in native conformation is hard to achieve [3, 7–9]. V_HHs generated from synthetic peptides derived from the target protein usually fail to recognize the protein in its native conformation whereas conjugation of the peptides to larger, highly immunogenic carrier proteins results in unpredictable specificity and affinity necessitating considerable screening efforts to identify good therapeutic candidates. Additionally, traditional screening strategies can generate V_HHs that recognize linear epitopes, increasing the likelihood of off-target binding and toxicity, and most isolation campaigns yield binders to a single dominant epitope, limiting functional diversity [10]. To address these challenges, various approaches have been explored, including immunization with transfected cells overexpressing the protein of interest, or incorporation of the purified protein into liposomes or viral particles to better mimic the native membrane environment [7, 11, 12]. However, these methods suffer from the presence of numerous other membrane components potentially interfering with the immune response directed against the target protein of interest.

Genetic immunization, using DNA or RNA, offers a promising alternative. In this approach, host cells are transfected with vectors encoding the target antigen, enabling faithful expression and post-translational modification in a native cellular context [13]. This strategy allows for rapid and flexible immunogen design, avoids impurities commonly associated with protein immunization, and supports the presentation of conformational epitopes [14, 15]. While DNA immunization has been successfully utilized in small inbred laboratory animals (mostly rodents), this method has proven less effective in large outbred species such as camelids, potentially due to low plasmid uptake and limited antigen expression in peripheral tissues [16]. Some of these limitations have been partially addressed through improved delivery methods, including tattoo-based vaccination, which has provided safe and robust immune responses in rodents and non-human primates [17–23], although this approach has not yet been evaluated in members of the Camelidae family. Nevertheless, a few groups have successfully generated functional V_HHs in llamas using DNA immunization against challenging membrane targets including the chemokine receptors CXCR7 [24] and ChemR23 [25], the ribosyltransferase ART2.2 [26], the ecto-enzyme CD38 [27], the G-protein coupled receptor (GPCR) VPAC1 [28], the ion channel P2X7 [29] and the epidermal growth factor receptor (EGFR) [30]. However, the overall success rate in previous studies was variable and the isolated V_HHs were generally of modest affinity. In addition, all attempts to date included either repeated dosing of the genetic material over an extended period of time or a final boost with recombinant proteins or overexpressing cells. In contrast, mRNA immunization platforms have gained significant attention following their clinical success during the Coronavirus disease 2019 (COVID-19) pandemic. mRNA vaccines allow for rapid and efficient expression of the protein antigen as they bypass the nuclear transcription step involved with DNA immunogens. When combined with lipid nanoparticle (LNP) delivery, which protects mRNA from degradation and facilitates cellular uptake, this approach has proven markedly more effective than previous genetic immunization strategies [15]. Recently, one group successfully utilized mRNA-LNP immunogens to raise functional mAbs in humanized mice against membrane proteins such as the Omicron BA.1 spike, CD22, and GPRC5D [31]. To date, however, mRNA immunization has not been employed for V_HH discovery in camelids.

Here, we report the first successful generation of target specific, functionally promising V_HHs in llamas using mRNA-LNP immunization. We formulated LNPs encapsulating mRNA encoding the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) Wuhan spike protein as a model antigen and administered these LNPs to a llama. The resulting V_HH immune phage display library was screened using cell-based panning methods that preserve the native conformation of the membrane-bound spike protein. From this screen, we identified high-affinity V_HHs, including two lead candidates that effectively blocked spike-angiotensin-converting enzyme 2 (ACE2) receptor interaction in a surrogate cell-based assay. To our knowledge, this study represents the first published demonstration of target specific V_HH discovery using an mRNA-LNP platform in a camelid species. This approach, when combined with conformationally relevant screening techniques, may offer a powerful and accelerated strategy for generating high-quality V_HHs against complex or recalcitrant targets, particularly membrane proteins. It holds promise for expanding the scope and speed of V_HH-based therapeutics and diagnostics across a broad range of applications.

Materials and methods

Recombinant antigens, antibodies, and reagents

Purified recombinant spike proteins from SARS-CoV-2 Wuhan were described previously [32]. Full length spike proteins (aa 16–1208), C-terminally biotinylated and non-biotinylated (spike-bio and spike), or containing a C-terminal Flag tag (trimeric spike) [33] and recombinant domains (S1, S2, receptor binding domain (RBD) or N-terminal domain (NTD)) of the spike protein were produced in-house. The S1 domain contains a N-terminal Flag tag followed by 6xHis (FH6-S1, aa 16–685), the S2 domain is C-terminally tagged with resistin, flag, and 6xHis (S2-T1, aa 686–1208), the RBD (aa 319–591) is either fused to a human IgG1 hinge/Fc fragment (RBD-Fc) or C-terminally tagged with Flag and 6xHis (RBD-FH6) and finally, the NTD contains a C-terminal Flag tag followed by 6xHis (NTD-FH6, aa 16–305). Recombinant ACE2 with a N-terminal Flag tag followed by 6xHis (FH6-ACE2, aa 18–615) was produced in-house and has been previously described [32]. Recombinant 6 × His-tagged human EGFR ectodomain was from GenScript (control Ag, Cat#Z03194). Ovalbumin was obtained from Sigma (Cat#A5503) and dissolved in milliQ water at 1 mg/ml. Abs and other detection reagents used in this study included mouse mAb anti-M13 major coat protein conjugated to horseradish peroxidase or phycoerythrin (anti-M13 HRP and PE) (Santa Cruz Biotechnology, Cat#Sc-53004 and Sc-53004 PE), rabbit polyclonal anti-6-His Tag conjugated to HRP or Biotin (anti-His HRP and bio) (Bethyl Laboratories, Cat#A190-114P and Abcam, Cat#27025), streptavidin R-PE (SA-PE) (ThermoFisher, Cat#S866), goat anti-llama IgG conjugated to HRP or fluorescein isothiocyanate (anti-llama HRP and FITC) (Fisher Scientific, Cat#NC9656984 and Agrisera Cat# AS10 1419), donkey anti-Human IgG Fc conjugated to Alexa Flour 488 (anti-human Fc-488) (Jackson ImmunoResearch, Cat#709–546-098), mouse THE™ DYKDDDK tag Ab conjugated to Alexa Flour 488 (anti-Flag-488) (GenScript, Cat#A01809–100) and goat anti-human IgG Fc conjugated to HRP (anti-human Fc HRP) (Sigma, Cat#A0170). The positive control V_HH-1d targeting SARS-CoV-2 spike RBD [32] and the negative control A20.1 against Clostridioides difficile toxin A [34] were previously described. M13K07 helper phage (Cat#N0315S) was purchased from New England Biolabs.

Preparation of mRNA/LNPs

The procedure to generate SARS-CoV-2 spike mRNA/LNPs has been described previously [35]. Briefly, a linearized plasmid DNA construct was used as the template for in vitro transcription (IVT) using the Megascript T7 transcription kit (ThermoFisher, Cat#AM1334), CleanCap Reagent AG (TriLink Biotechnologies, Cat# N-7113) and m1ψ (TriLink Biotechnologies, Cat#N-1081). RNA was encapsulated within LNPs at an N:P ratio of 6 by use of the NanoAssemblr Ignite system (Precision NanoSystems) and Genvoy-ILM ionizable lipid mixture (Precision NanoSystems). Quality and quantity of RNA was resolved by formaldehyde (2.6% [v/v]) denaturing agarose (1% [w/v]) gel and RiboGreen assay, respectively.

Llama immunization and assessment of the immune response

Immunization of one llama was performed at Cedarlane Laboratories (Burlington, Canada) essentially as described [34, 36]. Briefly, the llama was injected with three doses (Days 0, 21, and 35) of 50 μg of LNPs containing SARS-CoV-2 Wuhan spike-encoding mRNA in a final volume of 500 μl of 1xPBS. At Day 42, 100 ml of blood from the llama was drawn and peripheral blood mononuclear cells (PBMCs) were purified by ficoll gradient at Cedarlane Laboratories.

The immune response against the spike was assessed by enzyme-linked immunosorbent assay (ELISA) and cell binding experiments using pre-immune (Day 0) and immune (Day 42) sera. The detailed procedures can be found under the respective sections (ELISA and Cell binding assays).

Phage-displayed V_HH library construction

Phage-displayed V_HH library construction was performed essentially as described previously [30, 34, 36]. Briefly, total RNA was extracted from PBMCs using TRIzol™ Plus RNA Purification Kit (ThermoFisher, Cat#12183555) following the manufacturer’s instructions and used to reverse transcribe cDNA with SuperScript™ IV VILO™ Master Mix supplemented with random hexamer (ThermoFisher, Cat#SO142) and oligo (dT) (ThermoFisher, Cat#AM5730G) primers. V_HH genes were amplified using semi-nested PCR and cloned into the phagemid vector pMED1, followed by transformation of Escherichia coli TG1 (Lucigen, Cat#60502–02) to construct a library representing the llama’s V_HH repertoire. Phage particles displaying the V_HHs were rescued from E. coli cell libraries using M13K07 helper phage (New England Biolabs, Whitby, Canada, Cat#N0315S) and used for panning.

Panning

• a) Cell-based panning. Chinese hamster ovary (CHO)^55E1™ cells expressing the full-length (including transmembrane and C-terminal domains) SARS-CoV-2 Wuhan spike protein (CHO-Spike) under the control of the cumate-inducible CR5 promoter were generated by methionine sulfoximine selection of plasmid-transfected cells, as described previously [37, 38]. Non-transfected CHO^55E1™ cells (Parental CHO) were used as negative antigen control. Cells were grown in BalanCD™ CHO Growth A medium (Irvine Scientific, Cat#91128) supplemented with 50 μM methionine sulfoximine at 120 rpm and 37 °C in a humidified 5% CO₂ atmosphere. Expression of the spike protein was induced by adding cumate at 2 μg/ml for 48 h at 37 °C. After 48 h, the expression of the spike was confirmed by Mirrorball® fluorescence cytometry using a control V_HH directed against the spike fused to the human IgG Fc domain (V_HH-1-Fc) [32] followed by detection using anti-human Fc-488 (data not shown). Parental CHO and two independent cumate-induced CHO lines expressing non-related membrane receptors (decoy antigens) were used for negative subtraction of the phage population prior to incubation with spike-expressing cells. The decoy antigens, LPAR1 and GALR2, were selected as unrelated, structurally distinct cell-surface proteins with no sequence similarity to spike. The decoy antigen-expressing cells were maintained under the same methionine sulfoximine selection pressure and cumate induction conditions as the target cells. This design enables subtraction against extraneous proteins that may be up- or down-regulated by the induction process itself. The cumate system is known for its low basal leakiness, however, even minimal background expression of the target antigen can risk loss of high-affinity binders if non-induced CHO-Spike cells are used for subtraction. Therefore, two independent inducible decoy lines were used instead, reducing the likelihood of off-target enrichment and preventing dominant non-specific clones from monopolizing downstream TG-1 infection.

  For panning, decoy antigen CHO cells (1 × 10⁷) were centrifuged (200 × g, 5 min) and rehydrated in CHO media with 1% BSA and 1.5 × 10¹⁰ phage library then incubated at 4 °C for 30 min on a gentle rocking platform to remove non-specific binding phage. The subtraction cells were pelleted by centrifugation (200 × g, 5 min) and the saved library supernatant was subjected to additional subtractions with CHO decoy antigen cells followed by parental CHO cells. The final subtracted phage library was incubated (1.5 h at 4 °C) with CHO-Spike cells to capture target selective binding phage. Following the phage incubation, cells were washed five times with 10 ml of phosphate buffer saline (PBS) and bound phage was eluted with 1000 μl of 100 mM triethylamine for 10 min at room temperature, passed through a 1 ml syringe 18-gauge needle to shear the DNA and reduce sample viscosity. The elution product was neutralized with 2 M Tris–HCl pH 8 and used to infect exponentially growing E. coli TG1 cells. The cultured E. coli were then superinfected with M13K07 helper phage at a multiplicity of infection of 20:1 at 37 °C for 30 min with no shaking, after which they were grown overnight at 37 °C and 250 rpm in fresh 2 × YT media in the presence of 100 μg/ml of ampicillin and 50 μg/ml of kanamycin. The next day, amplified phage was purified by polyethylene glycol precipitation from 10 ml overnight cultures and used in subsequent panning rounds. After four rounds of selection, individual V_HH-displaying phage clones were screened for antigen binding by monoclonal phage ELISA or cell binding as described below. The sequences of positive V_HHs were identified by Sanger DNA sequencing.

• b) Recombinant antigen panning. Antigens were captured or passively adsorbed in appropriate wells overnight at 4 °C with agitation. For round 1 and 3, 1.0 μg/well of spike in 100 μl of PBS was directly adsorbed in wells of NUNC® Immulon 4 HBX microtiter plates (ThermoFisher, Cat#3855) whereas for round 2 and 4, 1.0 μg/well of spike-bio was captured in streptavidin- or neutravidin-coated wells respectively already pre-blocked with Superblock™ Buffer (ThermoFisher, Cat#15501 and Cat#15127, respectively). For all rounds, negative selection was achieved on empty wells prior to incubation with the target antigen. The next morning, wells that were not pre-blocked were emptied and 300 μl/well of blocking buffer (PBS containing either 1% [w/v] of casein (PBSC) for round 1 or 1% [w/v] of BSA (PBSB) for round 3) was added in each well for 1 h at room temperature prior to incubation with library phages (1.5 × 10¹⁰ colony-forming units in blocking buffer containing 0.05% [v/v] Tween 20) for 2 h at room temperature. The blocking buffer (PBSC and PBSB) was switched in alternate rounds of panning, except for panning on streptavidin- and neutravidin-coated wells which were already pre-blocked. In the latter case, the capture protein (streptavidin and neutravidin) was switched in alternate rounds of panning. Following the phage incubation, cells were washed with PBS +0.05% [v/v] tween (PBST) and bound phage was eluted with 100 μl of 100 mM triethylamine for 10 min at room temperature, then neutralized with 2 M Tris–HCl, pH 8, and used to infect exponentially growing E. coli TG1 cells as described in a). After four rounds of selection, individual V_HH-displaying phage clones were screened for antigen binding by monoclonal phage ELISA or cell binding as described below. The sequences of positive V_HHs were identified by Sanger DNA sequencing.

Enzyme-linked immunosorbent assay

• a) Serum ELISA. Wells of NUNC® Immulon 4 HBX microtiter plates were coated with 200 ng/well of the spike or control Ag (EGFR) in 100 μl of PBS overnight at 4 °C with slow agitation. The next morning, wells were blocked with 300 μl of PBSC for 1 h at room temperature followed by addition of pre-immune or immune serum serially diluted in PBSC 0.05% [v/v] Tween 20 (PBSCT). Following incubation for 2 h at room temperature, wells were washed three times with PBST and once with PBS before adding 100 μl of PBSCT containing anti-llama HRP for 1 h at room temperature. Wells were then washed as above prior to incubation with 100 μl of TMB substrate (SeraCare, Cat#50–76-00) for 5–30 min. The reaction was stopped by adding 100 μl of 1 M H₂SO₄, and A_450nm was read with a microtiter plate reader.

• b) Monoclonal phage ELISA. Individual colonies obtained after four rounds of panning were grown in 2 × YT medium supplemented with 100 μg/ml of ampicillin (2 × YT-Amp) and 0.1% glucose until A_600nm of 0.6 and infected with M13K07 helper phage for 30 min at 37 °C with no agitation followed by an additional 30 min at 37 °C with agitation at 250 rpm. The cells were then pelleted by centrifugation at 6129 × g for 15 min at 4 °C and resuspended in fresh 2 × YT medium containing 50 μg/ml of kanamycin and 100 μg/ml of ampicillin for overnight growth at 37 °C and 250 rpm. The next day, cultures were centrifuged at 6129 × g for 10 min at 4 °C to pellet the cells and 100 μl of the supernatant from each sample was added to pre-coated microtiter plate wells for ELISA as described in a) and detection using anti-M13 HRP.

Cloning and expression of V_HHs

For expression of V_HHs, selected sequences were cloned into the SfiI sites of a modified pET expression vector containing biotin acceptor peptide and poly-histidine tags (pMRO-BAP-H6) [36] for their production in BL21(DE3) E. coli. Following electroporation of BL21(DE3) cells with the plasmids, individual colonies were cultured overnight in 10 ml of 2 × YT media supplemented with 50 μg/ml of kanamycin (2 × YT-Kan) at 37 °C and 250 rpm. The next day, cultures were added to 250 ml of 2 × YT-Kan and grown to an A_600nm of 0.6. Expression of V_HHs was induced with 10 μM of isopropyl β-D-1-thiogalactopyranoside overnight at 28 °C and 250 rpm. Bacterial pellets were then harvested by centrifugation at 6129 × g for 15 min at 4 °C and V_HHs were extracted by sonication followed by biotinylation by incubating extracts with 10 μM of ATP (Alfa Aesar, Cat#CAAAJ61125–09), 100 μM of D-biotin (VWR, Cat#97061–446) and bacterial cell extracts overexpressing E. coli biotin ligase (BirA) as previously described [36]. The V_HHs were then purified by immobilized metal affinity chromatography (IMAC) and analysed by mass spectrometry to confirm that they were intact and fully biotinylated.

Size-exclusion chromatograph

Purified V_HHs were subjected to size-exclusion chromatography (SEC) to determine their aggregation tendencies and to isolate monomeric fractions. Briefly, ~1 mg of each V_HH was injected into a Superdex™ 75 10/300 GL or a HiLoad 16/600 Superdex 75 column (Cytiva, Cat#29148721 and 28989333, respectively) connected to an ÄKTA FPLC protein purification system (Cytiva) as previously described [39]. PBS was used as the running buffer at a flow rate of 0.5 ml/min at room temperature. Fractions corresponding to the monomeric peak were collected, aliquoted, and stored at −80 °C until further use. Percent (%) monomers represent the fraction of the area under the monomeric peak over the total peak area and were calculated using the UNICORN™ 7.6 software (Cytiva).

Surface plasmon resonance

All SEC purified monomeric V_HHs were supplemented with 3 mM ethylenediaminetetraacetic acid (EDTA) and 0.05% [v/v] Tween 20 prior to surface plasmon resonance (SPR) analyses. All SPR analyses were performed on a Biacore™ T200 instrument (Cytiva) at 25 °C in PBST running buffer (135 mM NaCl, 2.7 mM KCl, 4.3 mM Na₂HPO₄, 1.4 mM KH₂PO₄, 0.05% [v/v] Tween-20 (Teknova, Cat#P1192) and 3 mM EDTA (Teknova, Cat#E0307), pH 7.4. Briefly, proteins (spike, ˜2000 resonance units (RUs); spike-bio, ˜2000 RUs; ovalbumin, ˜500 RUs) were immobilized via amine coupling on a CM5 series S sensor chip (Cytiva, Cat#29149603) using 10 mM acetate buffer, pH 4.0 or 4.5 buffer (Cytiva, Cat#BR100350) according to the manufacturer’s instructions. These immobilization densities resulted in surfaces with theoretical R_maxs for V_HHs ranging from 190–254 RUs. An ethanolamine-blocked flow cell served as a reference surface. V_HHs were injected over all four surfaces at various concentrations ranging from 5 nM to 4 μM, at a flow rate of 20 μl/min, with a contact time and a dissociation time of 180 s. All surfaces were regenerated using 10 mM glycine, pH 1.5, for 60 s at 100 μl/min. Single cycle kinetics data were fitted to a 1:1 interaction model and binding affinities (K_Ds) and kinetic rate constants (k_as, k_ds) were determined using BIAevaluation 3.2.1 software (Cytiva).

For SPR-based epitope binning, a dual injection approach was used by injecting one V_HH over the spike surface followed immediately by a second injection of a mixture of the first V_HH with a second V_HH. The concentrations of the V_HHs injected ranged from 5× K_D to 2000× K_D, at a flow rate of 20 μl/min. Between cycles, all surfaces were regenerated using 10 mM glycine, pH 1.5, for 60 s at 100 μl/min.

For SPR-based domain mapping, spike protein fragments were amine coupled on CM5 series S sensor chips in 10 mM acetate buffer, pH 4.5. The following surfaces were prepared: RBD (RBD-FH6; 466 RUs), NTD (NTD-FH6; 455 RUs), and S2 (S2-T1; 996 RUs). Each V_HH was injected at 500 nM for 180 s of contact time and 180 s of dissociation time, at a flow rate of 20 μL/min. Surfaces were regenerated as above.

ACE2 - V_HH competition experiments were performed by SPR on a S1 (FH6-S1, 1106 RUs) surface using a dual injection method, essentially as previously described [32]. In the first injection, a V_HH (5 μM) was flowed for 180 s at 20 μl/min and immediately followed by a second injection that contained ACE2 (FH6-ACE2; 1.5 μM) + V_HH (5 μM) for 180 s. After the second injection, 30 s of dissociation time precluded surface regeneration (conditions as above). The A20.1 V_HH [34] served as a non-binding control.

Melting temperature (T_m) measurements

T _ms were determined by circular dichroism (CD) as previously described [38]. Ellipticity of V_HHs at 100 μg/ml was measured at a wavelength of 202 nm in 100 mM sodium phosphate buffer, pH 7.4. Ellipticity measurements were normalized to a percentage scale and then T_ms were determined by plotting % V_HH folded vs temperature and fitting the Boltzmann sigmoidal (non-linear regression) model to the plot using GraphPad Prism version 10.1.2 for Windows (GraphPad Software, www.graphpad.com).

Cell binding assays

For cell cytometry experiments, CHO-Spike or parental CHO cells were harvested by centrifugation and resuspended at 1 × 10⁶ cells/ml in PBS with 1% [w/v] BSA (PBSB). Cells were then kept on ice until use.

• a) Serum cell binding. Parental CHO cells were incubated with stock pre-immune and immune sera for 1 h at 4 °C to preclear the samples of non-specific binding V_HH. Serial dilutions of pre-immune and immune sera were then incubated with 100 μl (1x10⁵ cells) of CHO-Spike or parental CHO cells in V-Bottom 96-well microtiter plates (Globe Scientific, Cat#120130). Plates were incubated for 1.5 h on ice, washed three times with PBS by centrifugation for 2 min at 200 × g and then incubated for an additional 1 h with 100 μl of anti-llama FITC at 2.5 μg/ml diluted in PBSB. Then, cells were washed as above and resuspended in 100 μl of 500 nM Draq 5™ (ThermoFisher, Cat#65–0880-92) and transferred to 96-well microplate (ThermoFisher, Cat#165–305) for data acquisition on a Mirrorball® instrument (SPT Labtech) and analysis with the Cellista software (SPT Labtech).

• b) Monoclonal phage cell binding. Individual culture supernatant samples of clones obtained after four rounds of panning were produced as described in the Monoclonal phage ELISA section above. To remove non-specific binding phage, each monoclonal phage sample was pre-cleared with parental CHO cells (1 h at 4 °C). The pre-cleared supernatant (200 μl) was incubated with 0.5 ×10⁶ CHO-Spike or parental CHO cells (1 h at 4 °C), washed four times with PBS by centrifugation for 1 min at 200 × g and incubated for an additional hour with 100 μl of anti-M13 PE at 0.4 μg/ml diluted in PBSB. Then, cells were washed three times with PBS and resuspended in 100 μl of 500 nM Draq 5™ (ThermoFisher, Cat#65–0880-92) and transferred to 96-well microplate (ThermoFisher, Cat#165–305) for data acquisition on a Mirrorball® instrument (SPT Labtech) and analysis with the Cellista software (SPT Labtech).

• c) V_HH cell binding. Starting at 1000 nM, five-fold dilutions of V_HHs as well as A20.1 V_HH negative control [34] were incubated with CHO-Spike and parental CHO cells (2 h at 4 °C) followed by detection with anti-His bio (1 μg/ml) and SA-PE (2 μg/ml) for 1 h at 4 °C. Half-maximal effective concentrations (EC₅₀s) for the binding of V_HHs to CHO-Spike cells were obtained by plotting the mean fluorescence intensity (MFI) vs [V_HH] and fitting the specific binding with Hill slope (non-linear regression) model to the plot using the GraphPad Prism version 10.1.2 for Windows.

Surrogate ACE2-blocking cell-based assay

The surrogate ACE2-blocking cell-based assay has been described previously [40]. Briefly, the ability of V_HHs to inhibit the interaction of trimeric spike or RBD-Fc with the ACE2 receptor expressed at the cell surface was determined by combining 100 ng of the trimeric spike or RBD-Fc proteins with 1 × 10⁵ human embryonic kidney cells recombinantly expressing ACE2 on its surface (HEK293T-ACE2, BEI Resources, Cat#NR-52511) in the presence of three-fold dilutions of each V_HH starting at a concentration of 1000 nM in a final volume of 150 μl of PBSB. Following incubation on ice for 1 h, cells were washed twice with PBSB by centrifugation for 2 min at 200 × g and then incubated for an additional hour on ice with 50 μl of anti-Flag-488 or anti-human Fc-488 diluted to 1.25 and 1.87 μg/ml, respectively in PBSB. Cells were subsequently washed as above and resuspended in 100 μl of fresh PBSB for data acquisition on a Mirrorball® instrument (SPT Labtech) and analysis with the Cellista software (SPT Labtech). Half-maximal inhibitory concentrations (IC₅₀s) for the blocking of spike or RBD/ACE2 binding by the V_HHs were obtained by plotting MFI vs [V_HH] and fitting the non-linear regression [inhibitor] vs response – variable slope (four parameters) model to the plot using the GraphPad Prism version 10.1.2 for Windows. Data is presented as percent inhibition of spike or RBD-Fc binding to HEK293T-ACE2 cells which was determined using the following formula;

where, Fn is the measured fluorescence at any given competitor V_HH dilution.

Fmin is the background fluorescence measured in the presence of cells and detection agent only.

Fmax is the maximum fluorescence measured in the absence of competitor V_HH.

Results

Production of mRNA-LNPs for llama immunization and induction of an immune response

To demonstrate the applicability of mRNA vaccination for generating V_HHs that recognize proteins in their native conformation in camelid species, mRNA encoding the SARS-CoV-2 spike protein was prepared using a plasmid template based on the design used by Pfizer-BioNTech, in which the spike coding sequence is flanked by the 5′ UTR of alpha-globin and the 3′ UTR of mtRNR1-AES followed by a poly(A) tail (Fig. 1a). The sequence of spike was based on the reference strain originally identified in Wuhan and as in the clinical vaccine [41], includes two stabilizing prolines in the S2 domain to improve stability of the trimer in the prefusion state. The 5’ Cap-1 structure was incorporated into the synthetic RNA co-transcriptionally. Additionally, similar to the clinically relevant RNA vaccines [42], N1-methyl-pseudouridine was used in place of the canonical uridine. Purified RNA was then incorporated into LNPs and quantified by RiboGreen Assay to determine the levels of encapsulated RNA. Notably, we have previously shown that mRNA-LNPs generated using this method have demonstrated robust immunogenicity within a mouse model [35].

Figure 1.

A novel llama mRNA-LNP immunization platform for generating specific immune responses against target antigens. (a) Schematic representation of the immunization schedule. mRNA encoding the SARS-CoV-2 Wuhan spike protein was encapsulated in LNPs and injected in one llama at day 0, 21 and 35. At Day 42, blood and PMBCs were collected from the animal to assess the immune response and generate an immune V_HH phage display library. (b, c) Screening of the serum collected from Day 0 and Day 42 by (b) ELISA and (c) cell cytometry assay.

One llama was immunized with three doses of spike mRNA-LNPs (50 μg per dose) administered on Days 0, 21, and 35 (Fig. 1a). Blood was collected one week after the final injection (Day 42), and the specific immune response against SARS-CoV-2 Wuhan spike was assessed by ELISA (Fig. 1b) and cell cytometry assays (Fig. 1c), comparing pre-immune (Day 0) and immune sera (Day 42). ELISA results demonstrated a strong polyclonal response in the immune serum against the spike protein, while no signal was detected with the pre-immune serum (Fig. 1b). Specificity controls using an irrelevant antigen (Control Ag) and casein showed no detectable binding, confirming that the response was directed specifically against the spike antigen. Cell cytometry analysis further revealed that the immune serum recognized CHO-Spike cells, indicating recognition of the antigen in its native conformation (Fig. 1c). No fluorescence signal was observed in parental CHO cells, confirming the absence of nonspecific binding.

Isolation of V_HHs targeting the SARS-CoV-2 spike protein using cell-based panning

To identify specific binders to the native SARS-CoV-2 spike protein, a phage display library representing the llama’s V_HH repertoire was constructed using PBMCs collected at day 42 after the first mRNA-LNP injection. The immune library underwent three rounds of cell-based panning using CHO-Spike cells (Fig. 2a). Each round included pre-incubation steps with parental CHO or decoy antigen cells to remove nonspecific phages before exposure to CHO-Spike cells. At the end of panning, ~200 clones from each round were sequenced. From this pool, 19 candidates that were present exclusively in the CHO-Spike population and absent from the parental CHO pool were selected for specificity assessment by monoclonal phage ELISA (Fig. 2b) and cell binding assays (Fig. 2c). Clones showing positive responses in either assay were advanced, resulting in a final panel of 12 candidate V_HHs (Fig. S1, Table S1). Among these, one clone, S-4 exhibited a slightly higher sequence frequency of 8.6% while others were below 5%, ranging from 0.3% to 2.7% (Fig. 2d, Table S1).

Figure 2.

Isolation of antigen-specific V_HHs by cell-based panning. (a) Schematic representation of the cell-based panning strategy. The phage-displayed V_HH library was applied to decoy antigen-expressing cells or parental CHO cells for three rounds of negative selection to reduce background prior to positive selection on CHO-spike cells. Cells were then washed thoroughly and bound phages were eluted and purified for subsequent rounds of panning. (b, c) screening of phage-displayed V_HHs by ELISA (b) and cell cytometry (c) following three rounds of panning according to the strategy described in (a). M13 phage particles were used as a negative control in both assays. (d) Frequency of isolated clones amongst all sequences analyzed (total of 694).

To directly evaluate the mRNA-LNP immunization platform without the added complexity of cell-based selection, a traditional panning method using recombinant spike antigen was performed in parallel. In this case, four rounds were carried out, alternating between biotinylated and non-biotinylated spike (Fig. S2A). From round four, 14 clones were selected for specificity analysis as described above. Monoclonal phage ELISA results showed that 12 V_HHs bound specifically to the spike protein, with no detectable interaction with the casein control (Fig. S2B, Table S1). Among these, five V_HHs also recognized CHO-Spike cells but not parental CHO cells (Fig. S2C, Table S1). One clone (S-41) exhibited specific binding to the spike protein by cell cytometry but not by ELISA. Of the 13 spike-binding clones identified, 11 overlapped with those recovered from cell-based panning, while two (S-21 and S-41) were unique to the recombinant antigen campaign. Since characterization of the recombinant spike-derived V_HHs was beyond the scope of this study, S-21 and S-41 were not pursued further. Unlike the cell-based panning results (Fig. 2d, Table S1), recombinant antigen selection yielded a dominant clone, S-43, representing 41% of the population (Fig. S2D, Table S1).

Characterization of the biophysical and binding properties of selected V_HHs

The 12 spike-specific V_HHs isolated by cell-based panning were subcloned for bacterial expression and purified using IMAC. Purified samples were analyzed by SEC to assess aggregation tendencies. Most V_HHs showed no or very low levels of aggregation, with monomer content ranging from 79.6% to 94.9% (Fig. 3a, Table 1). V_HH stability was assessed by determining their T_ms in CD assays. All V_HHs were highly thermostable, with T_m values ranging from 61.6 to 75.9 °C (Fig. 3b, Table 1).

Figure 3.

Biophysical properties of lead V_HH candidates. (a) SEC profiles assessing the aggregation tendency of V_HHs. mAU, milliabsorbance unit. (b) Thermal stability of V_HHs determined by CD experiments. V_H ellipticities measured as a function of temperature are normalized to percentage scales (% V_HH folded). Melting temperatures (T_ms) correspond to temperatures at which 50% of the protein is unfolded. Ellipticity values represent means of three technical replicates.

Table 1.

Physico-chemical and functional properties of V_HHs selected against SARS-CoV-2 spike antigens

VH H	% monomer	V e M (ml) a	T m (°C) b	K D (nM) b  ,  c		EC 50 (nM) b  ,  d	IC 50 (nM) b  ,  e
				Spike	Spike-bio		RBD-Fc	Spike
S-3	79.6	13.1	67.1 ± 0.32	79.2 ± 3.07	107.3 ± 3.09	7.75 ± 11.08	-	-
S-4	88.5	15.1	70.2 ± 0.17	1.92 ± 0.04	2.53 ± 0.07	0.39 ± 0.06	20.95 ± 4.77	4.03 ± 0.93
S-10	86.8	13.5	74.0 ± 0.01	677 ± 14.7	978 ± 26.3	23.82 ± 4.96	-	-
S-11	91.9	12.8	70.7 ± 0.05	-	-	N/A	N/A	N/A
S-17	89.7	80.3	75.1 ± 0.5	9.56 ± 0.2	13.2 ± 0.47	1.97 ± 2.52	-	-
S-36	92.8	13.1	63.5 ± 0.15	1463.3 ± 30.9	1933.3 ± 74.09	50.61 ± 17.18	-	-
S-42	94.2	80.7	69.6 ± 0.03	112 ± 2.45	142.3 ± 1.7	3.74 ± 0.41	-	-
S-43	87.5	12.4	69.5 + 0.05	33.27 ± 0.42	37.1 ± 0.5	750.7 ± 177.52	364.5 ± 7.76	45.59 ± 35.26
S-44	94.9	73.3	61.6 ± 0.06	19.9 ± 0.65	19.47 ± 1.07	9.07 ± 4.03	-	-
S-88	87.5	13.2	70.4 ± 0.19	142.3 ± 15.37	191.3 ± 50.8	1.29 ± 0.28	-	-
S-94	83	13.1	75.9 ± 0.05	1440 ± 43.2	1953.3 ± 69.4	40.14 ± 16.65	-	-
S-188	94.8	14.7	69.4 ± 1.35	-	-	N/A	N/A	N/A

^aElution volume of the monomeric peak (V_e M) obtained from size-exclusion chromatograms. V_HHs S-17, S-42 and S-44 were purified on a HiLoad 16/600 Superdex 75 column whereas all other samples were processed on a Superdex 75 10/300 column.

^bValues are mean ± SD of two to three technical replicates.

^c K _D values were obtained by SPR against spike antigens. k_a and k_d values are reported in Table S2. None of the V_HHs bound to ovalbumin.

^d EC ₅₀s were determined by cell cytometry assays of V_HHs against CHO-Spike cells.

^e IC50s were determined through surrogate blocking assays against RBD-Fc and trimeric spike. Dash indicates that no significant binding or inhibition was observed. N/A indicates that the V_HH was not analysed in the respective assay.

Binding kinetics of the purified V_HHs were evaluated by SPR against spike-bio, spike, and ovalbumin, the latter serving as a polyspecificity reagent to detect potential nonspecific interactions [43]. Two clones (S-11 and S-188) showed no measurable binding and were therefore excluded from subsequent analyses (data not shown). All remaining V_HHs bound specifically to the spike protein, exhibiting comparable affinities for both spike-bio and spike, and no detectable binding to ovalbumin (Fig. 4a, Tables 1 and S2). Affinity measurements revealed three groups of binders. The first group, comprising S-4, S-17, S-43, and S-44, displayed the highest affinities (1.92–37.1 nM), with S-4 showing the strongest binding of all clones, exhibiting K_D values of 1.92 nM against spike and 2.55 nM against spike-bio. The second group (S-3, S-42, and S-88) showed moderate affinities ranging from 79.2 to 191.3 nM, while the third group (S-10, S-36, and S-94) exhibited lower affinities between 677 and 1953.3 nM. To further characterize binding profiles and assess epitope diversity, the 10 spike-specific V_HHs were subjected to epitope binning experiments by SPR. Paired combinations of V_HHs were sequentially injected over a spike surface to determine competitive or non-competitive binding patterns. The analysis identified five distinct epitope bins, with S-4 and S-43 each recognizing a unique, non-overlapping epitope, indicating that the isolated V_HHs target multiple distinct regions of the spike protein (Fig. 4b).

Figure 4.

Characterization of the binding properties of spike-specific V_HHs by SPR. (a) Representative SPR sensorgrams and steady-state plots of V_HHs flown over immobilized spike, spike-bio or the ovalbumin control. Binders were classified as high (top; S-4, S-17, S-43 and S-44), medium (middle; S3-, S-42 and S-88) and low (bottom; S-10, S-36 and S-94) and the results of one representative V_HH from each group are shown. Binding affinity and kinetic constants, determined by single-cycle kinetic analysis, are reported in Table 1 and Table S2. Raw data points are shown with corresponding fits to a 1:1 binding model. SPR assays were performed in triplicates and the results from one replicate are shown. RU, resonance unit (b) epitope binning of binders identified in (a).

Finally, the ability of the 10 spike-specific V_HHs to recognize the spike protein in its native, membrane-associated context was evaluated by cell cytometry using CHO-Spike cells (Fig. 5). All V_HHs bound specifically to spike-expressing cells in a dose-dependent manner, with no detectable binding to parental CHO controls. Calculated EC₅₀ values ranged from 0.39 to 750.7 nM (Fig. 5, Table 1). Notably, clone S-4 had an EC₅₀ of 0.39 nM, surpassing the binding performance of the positive control V_HH-1d (EC₅₀ of 1.42 nM) previously isolated by recombinant antigen panning by our group [32].

Figure 5.

Recognition of the native spike receptor on overexpressing cells by V_HH candidates. Cell cytometry assessment of the binding of lead V_HHs to CHO-spike cells expressing spike protein on their surface or parental CHO cells. V_HH-1d [32] was used as a positive control and A20.1 [34] served as a negative control. Error bars indicate standard deviations of means of three technical replicates.

ACE2-blocking potency of spike-specific V_HHs

To assess whether binding to the spike protein translated into the ability of the V_HHs to block the interaction of the viral protein with the ACE2 receptor, a cell cytometry–based surrogate blocking assay was performed [32, 38, 40]. This assay measured the ability of each V_HH to block the interaction between the SARS-CoV-2 RBD-Fc or trimeric spike and the human ACE2 receptor expressed on the surface of HEK293T (HEK293T-ACE2) cells, yielding half-maximal inhibitory concentration (IC₅₀) values. As shown in Fig. 6a and b and Table 1, one V_HH, S-4, strongly inhibited binding of both viral proteins to ACE2, with IC₅₀ values of 20.95 nM against RBD-Fc and 4.03 nM against trimeric spike. The positive control V_HH-1d also demonstrated efficient inhibition (IC₅₀ values of 38.43 nM and 5.19 nM against RBD-Fc and spike, respectively), whereas the negative control A20.1 showed no measurable activity. The blocking activity of S-43 was substantially lower with an IC₅₀ value of 364.5 nM against RBD-Fc (Fig. 6a) and 45.59 nM against the trimeric spike (Fig. 6b). Notably, for all V_HHs that exhibited inhibitory activity (S-4, S-43 and positive control V_HH-1d), inhibition was more pronounced against the trimeric spike (Fig. 6b) than against the isolated RBD (Fig. 6a), suggesting that structural components of the spike trimer contribute to efficient blockade by these V_HHs. These results were further corroborated by an SPR-based ACE2 competition assay (Fig. S4), which confirmed complete inhibition of ACE2 binding to the S1 subunit of SARS-CoV-2 spike protein in the presence of S-4, while S-43 only partially blocked the interaction. In this assay, S-17 was included as a comparative control because it binds strongly to spike (Fig. 4a and Table 1) but does not inhibit ACE2 binding at the cell surface (Fig. 6a and b). Consistent with these findings, S-17 did not compete with ACE2 for binding to S1 (Fig. S4).

Figure 6.

ACE2-blocking potency of lead V_HHs against SARS-CoV-2. Surrogate ACE2-blocking cell-based assay assessing the potency of lead V_HHs in inhibiting the binding of (a) RBD-Fc or (b) full-length trimeric spike to HEK293T-ACE2 cells. V_HH-1d [32] was used as a positive control whereas A20.1 [34] served as a negative control. Data is expressed as % inhibition. Error bars indicate standard deviations of means of two to three technical replicates. (c) Domain-level SPR analysis showing the binding of S-4, S-43 and S-17 to immobilized RBD-FH6, NTD-FH6, and S2-T1. SPR assays were performed in triplicates and representative sensorgrams from one replicate are shown. RU, resonance unit.

To gain further insight into the mechanistic basis of these differences, additional SPR analyses were performed with S-4, S-43 and S-17. All three clones were assessed for binding to spike subunits (RBD, NTD and S2). Both S-4 and S-43 bound specifically to the RBD, whereas S-17 recognized the NTD (Fig. 6c). Based on the earlier epitope binning results (Fig. 4b), S-4 and S-43 were previously shown to recognize unique, non-overlapping epitopes, indicating that these distinct epitopes are both located within the RBD. Together, these findings corroborate the cell-based blocking data and provide mechanistic insight into the epitope localization and functional behavior of these V_HHs.

Discussion

This study demonstrates that llama immunization with mRNA-LNP effectively elicits a strong, specific Ab response against the SARS-CoV-2 spike protein and enables the generation of diverse, functionally blocking V_HHs. To our knowledge, this represents the first use of mRNA-LNP immunization for V_HH discovery in llamas, demonstrating the feasibility of this platform for rapid Ab generation in large outbred species.

The strength of this approach lies in the combination of mRNA-LNP immunization and cell-based panning. mRNA-LNP immunization enables in vivo expression of the antigen in its native conformation, eliciting a focused immune response without the off-target activation often associated with cell- or VLP-based immunizations that expose animals to thousands of unrelated surface proteins. This results in an immune repertoire highly enriched for antibodies specific to the target antigen. Cell-based panning then complements this by maintaining the antigen’s native membrane-associated structure during selection, allowing recovery of conformationally specific V_HHs. Together, these two components provide a powerful, integrated strategy for isolating high-affinity antibodies against complex membrane proteins.

The three-dose mRNA-LNP regimen used here contrasts with previous DNA immunization studies from our group [30], where no detectable Ab response was observed even after six injections and multiple additional boosts were required. Other DNA immunization campaigns in llamas have used four to eight DNA injections, often combined with cell or protein boosts [24–29, 44]. However, many of those studies did not include pre-immune controls or direct serum analyses, limiting assessment of their true efficacy. The present work highlights the advantage of mRNA-based immunization in eliciting rapid and robust humoral responses in large outbred mammals such as camelids. Future studies should explore the reproducibility of the approach across animals and incorporate deeper repertoire analyses to further characterize V_HH diversity and maturation dynamics.

Sequence analysis of V_HHs obtained from cell-based panning revealed modest sequence enrichment across rounds, with no single clone strongly dominating. The most frequent clone recovered was S-4 and represented 8.6% of total sequences. The relatively even sequence distribution likely reflects the fact that most clones were recovered during the first round of panning, before substantial selection pressure developed. In cell-based systems, phages displaying V_HHs that recognize abundant cell-surface components often expand preferentially during amplification, which can dilute or obscure specific antigen binders over successive rounds. Consistent with this, previous studies combining genetic immunization with cell- or virus-like particle (VLP)-based panning have typically limited screening to one or two rounds to reduce nonspecific amplification and preserve recovery of relevant binders [24–26, 28, 29]. Despite these challenges, the strategy employed here successfully yielded multiple spike-specific antibodies. Importantly, the robustness of this approach was further validated by applying the same cell-based panning protocol to a previously described immune llama V_HH library derived from a spike protein immunization [32], which resulted in the recovery of similar V_HH candidates (Fig. S3). This reproducibility underscores the reliability and broad applicability of the cell-based panning method.

In contrast, recombinant spike antigen panning yielded a repertoire skewed toward a single dominant clone, S-43, which accounted for 41% of all sequences. In this context, lower-frequency clones such as S-4, present at only 5.6%, can easily be overlooked despite their functional relevance. This outcome reflects a known limitation of antigen-based selections, where restricted epitope presentation on immobilized proteins can bias recovery toward antibodies recognizing exposed or linear epitopes [10, 45]. By comparison, cell-based panning maintained a more balanced sequence distribution, enabling detection and recovery of potent clones like S-4 that recognize conformational epitopes in the native membrane-associated form of the antigen.

SPR and cell cytometry analyses confirmed that the isolated V_HHs bind specifically and with high affinity to the spike protein in both recombinant and cell surface contexts. Epitope binning identified five distinct binding regions, indicating that the selected panel spans a broad epitope landscape. Among these, the two ACE2-blocking V_HHs, S-4, and S-43, each recognize unique epitopes within the RBD, in line with their distinct inhibitory profiles. S-4, which exhibited the strongest affinity and most potent inhibitory activity, likely binds a conformational epitope accessible in both recombinant and membrane-associated spike, reflecting its efficient cell-based binding. In contrast, S-43 targets a distinct epitope that appears to be more linear or less exposed in the native membrane context, explaining its comparatively weak inhibitory activity and minimal binding to spike-expressing cells despite high affinity by SPR. Because ACE2 blocking involves soluble spike engagement, S-43 can still exert partial inhibitory activity when its epitope becomes accessible in solution. Consistent with its NTD-binding profile, S-17 did not inhibit ACE2 interaction, as this domain is not involved in receptor engagement. This observation agrees with previous reports of NTD-specific V_HHs [32] and further supports that the behaviors observed here reflect known domain functions rather than unexpected cross-reactivity.

In general, the relative binding strengths observed by cell cytometry did not correlate directly with affinities determined by SPR. Several V_HHs that displayed high affinities in SPR assays showed comparatively weaker binding to cell-associated spike, and vice versa, suggesting that epitope accessibility and the structural context of the native membrane environment exert a greater influence on cell binding than intrinsic affinity alone. Collectively, these findings emphasize that epitope accessibility and presentation strongly influence apparent binding and functional outcomes, underscoring the advantage of cell-based selection for identifying Abs that effectively target native, membrane-associated antigens while minimizing dominance by sequence-amplification artifacts.

Overall, these findings demonstrate the feasibility of mRNA-LNP immunization as an efficient platform for generating spike specific ACE2-blocking V_HHs from llamas and demonstrate that cell-based panning can effectively isolate Abs recognizing native, conformational epitopes on complex membrane proteins such as the SARS-CoV-2 spike. The spike protein served as an ideal model antigen for proof-of-concept evaluation because it is well characterized and can be expressed in both soluble and membrane-bound forms. While spike is relatively accessible as a recombinant antigen, the full potential of this approach will be realized when applied to more structurally demanding multi-span membrane proteins such as GPCRs and ion channels, where purification of stable soluble formats remains challenging. Ab discovery against these targets has historically been limited by the difficulty of maintaining native conformation and presentation for Ab selection [3, 7, 8]. The strategy described here provides a versatile framework for overcoming these limitations and advancing the discovery of Abs against complex, clinically relevant membrane proteins. Ongoing work in our laboratory is focused on extending this approach to additional targets to further validate its general applicability.

Abbreviations

ACE2, human angiotensin-converting enzyme 2; Amp, ampicillin; BirA, E. coli biotin ligase; BSA, bovine serum albumin; CD, circular dichroism; CHO, Chinese hamster ovary; CHO-Spike, CHO cell recombinantly expressing spike protein on its surface; COVID-19, Coronavirus disease 2019; EC₅₀, half-maximal effective concentration; EDTA, ethylenediaminetetraacetic acid; EGFR, epidermal growth factor receptor; ELISA, enzyme-linked immunosorbent assay; FITC, fluorescein isothiocyanate; GPCR, G protein-coupled receptor; HEK293T-ACE2, human embryonic kidney cells recombinantly expressing ACE2; HCAb, heavy-chain only antibody; HRP, horseradish peroxidase; IC₅₀, half-maximal inhibitory concentration; IMAC, immobilized metal-ion affinity chromatography; IPTG, isopropyl β-D-1-thiogalactopyranoside; IVT, in vitro transcription; Kan, kanamycin; K_D, equilibrium dissociation constant; LNP, lipid nanoparticle; mAb, monoclonal antibody; MFI, mean fluorescence intensity; NTD, N-terminal domain; PBMC, peripheral blood mononuclear cell; PBS, phosphate-buffered saline; PBSB, PBS containing 1% (w/v) BSA; PBSC, PBS containing 1% (w/v) casein; PBSCT, PBSC containing 0.05% (v/v) Tween 20; PBST, PBS containing 0.05% (v/v) Tween 20; PCR, polymerase chain reaction; PE, phycoerythrin; RBD, receptor-binding domain of SARS-CoV-2 spike protein; RBD-Fc, human IgG1 hinge/Fc-fused RBD; RU, resonance unit; SA PE, streptavidin-PE; SARS-CoV-2, severe acute respiratory syndrome coronavirus 2; SEC, size-exclusion chromatography; Spike-bio, biotinylated spike; SPR, surface plasmon resonance; T_m, melting temperature; V_e, elution volume; V_HH, nanobody; VLP, virus-like particle.

Author contributions

Kasandra Belanger (Conceptualization [lead], Data curation [lead], Formal analysis [equal], Supervision [lead], Writing—original draft [lead], Writing—review & editing [lead]), Debbie Callaghan (Data curation [equal], Formal analysis [equal], Methodology [lead], Writing—review & editing [supporting]), Tyler Renner (Methodology [equal], Writing—review & editing [supporting]), Hiva Azizi (Methodology [supporting], Writing—review & editing [supporting]), Jasmine Hu (Methodology [supporting]), Shalini Raphael (Methodology [supporting], Writing—review & editing [supporting]), Marie-France Goneau (Methodology [supporting]), Greg Hussack (Supervision [supporting], Writing—review & editing [equal]), Michael McCluskie (Funding acquisition [equal]), and Bassel Akache (Methodology [supporting], Supervision [supporting], Writing—review & editing [supporting])

Supplementary Data

Supplementary data is available at ABT Online.

Conflict of interest

The authors report there are no competing interests to declare.

Funding

This work was supported by funding from the National Research Council Canada.

Data availability

The data underlying this article will be shared on reasonable request to the corresponding author.

Ethics and consent statement

Not applicable.

Animal research statement

All experiments involving animals were conducted according to protocols approved by the National Research Council Canada Animal Care Committee and in accordance with the guidelines set out in the OMAFRA Animals for Research Act, R.S.O. 1990, c. A.22.