RASP: rapid antibody functional screening by pentavalent phage display

ABSTRACT

The heavy-chain antibody variable domain (VHH) is the smallest antigen-binding domain of such antibodies, which are derived from camelids. In the past three decades, VHHs, which are also called single-domain antibodies, have been extensively used to target pathogens and/or toxins. Conventional screening methods, such as phage display, rely only on antibody-antigen binding as the sole criterion for selection. Despite being robust and high-throughput, such methods often require additional downstream experiments to identify VHH that neutralize their target. Here, we describe an innovative, high-throughput functional screening method, Rapid Antibody functional Screening by Pentavalent phage display (RASP), that incorporates purified antibody-displaying phages for virus neutralization assays, and thus can be used to directly identify neutralizing VHHs. As a proof-of-concept, we first displayed previously identified neutralizing VHHs specific for the spike proteins of SARS-CoV-2 and respiratory syncytial virus on phages and demonstrated a dose-dependent blockade of viral infection. We further improved our method by utilizing the pentavalent display feature of hyperphages. We showed that hyperphage-derived VHH phages were superior to helper phage-derived VHH phages in assaying viral neutralization potential. Thereafter, we applied RASP to identify multiple candidates by screening a semi-synthetic VHH library against recombinant vesicular stomatitis viruses pseudotyped with spike glycoproteins from SARS-CoV-2, Junin virus, and Ebola virus, featuring as case studies in antiviral antibody discovery. Further, we benchmarked RASP against established phage ELISA and next-generation sequencing methods. Overall, we successfully used RASP in the context of the discovery of antiviral VHHs, highlighting its broader applicability as a platform that can be used either in isolation or in conjunction with other functional screening methods to accelerate the discovery of antiviral VHHs.

Introduction

Functional heavy-chain antibodies are only found in camelids and cartilaginous fish such as sharks.^1,2 They were first discovered in the 1990s¹ in camelids from camelid sera samples, and later, similar types of naturally occurring antibodies were also found in sharks.² In camelids, heavy-chain antibodies (HCAbs) are present only as the IgG isotype due to point mutations at the intron-exon junction of the first constant domain of the heavy chain (CH1), resulting in the removal of CH1.³ Several other conserved key mutations have also been reported in the HCAbs. These mutations play an essential role in their increased stability, but also make them unable to pair with light chains.⁴ HCAbs have only one domain for antigen binding – a variable domain (VHH) also termed a single-domain antibody (sdAb) or Nanobody (hereafter referred to as VHH). The VHH is the smallest known antibody-based domain that can bind its antigen independent of a light chain. Other features of VHHs, including their small size (~15 kDa vs. ~150 kDa for an IgG), high thermal stability,⁵ and longer complementary-determining region 3 (CDR3) sequences, which afford binding to epitopes that are typically inaccessible to conventional antibodies, have made them especially desirable as reagents, diagnostics, and therapeutics. To date, over a thousand VHHs have been reported,⁶ including those targeting a variety of viral pathogens; several of these latter have been advanced.^7,8

The single-chain nature of VHHs (which obviates the chain-pairing issues associated with conventional monoclonal antibodies) has made them especially amenable to discovery through phage display and a variety of other platforms.⁹ In typical VHH discovery campaigns, antigen-specific VHHs are first identified through binding-based selections and screens (‘panning’) and then cloned and purified. Purified VHHs are then evaluated in functional assays and down-selected. Although this approach has proven highly successful, it nevertheless results in a substantial fraction of the isolated VHHs being discarded due to a lack of functional activity.

Given the potential for using VHHs in the treatment of infectious diseases, it is essential to improve screening methods. In most screening methods, the first step is to select candidates based on their ability to bind to the target antigen. Later, after subcloning and purification, they are tested for their functionality as part of a laborious and time-consuming process. Here, we describe a new approach – Rapid Antibody functional Screening by Pentavalent phage display (RASP) – that inverts this traditional discovery paradigm. Featuring case studies in antiviral antibody discovery, we demonstrate that screening for viral infection-blocking (neutralizing) VHHs can be performed directly on hyperphage-derived VHH phages, isolating only those VHHs that exhibit functional activity. We establish that RASP improves upon conventional screening methods by accelerating the discovery of antiviral VHHs. RASP and other approaches that triage antibodies lacking functional activity at the outset should facilitate rapid antibody discovery in response to new viral outbreaks

Materials and methods

Design of a semi-synthetic VHH library

We retrieved the VHH sequences from the Protein Data Bank (PDB) and followed the Aho numbering to obtain information on the diversity of CDRs, as described earlier.^10,11 It covers all the VHH with known structures (either from camelids (Naïve and immunized), semi-synthetic or synthetic origin). Duplicate sequences were removed using BBEdit, and then 206 unique VHH sequences were further analyzed using EXCEL macros. We calculated the frequency of each amino acid at each position in the CDRs and removed those residues that occurred less than 5% of the time at a given position. Briefly, we compared the frequency of amino acids at a given position in the respective CDRs from the PDB database. There were 1–4 key residues that were present in 10–15% frequency at each position within a specific CDR. As expected, CDR3 was the most diverse region, with its hypervariable innermost core exhibiting a frequency of 1–5% per amino acid per position (residues 3–15) and variable lengths. Secondly, we removed all residues with a frequency lower than 5% in CDR1, CDR2, and the outermost residues (1–4 and 15–18) of CDR3, and redistributed the frequency of key residues at each position. Thirdly, we incorporated CDR3 length variation, ranging from 11 to 26 amino acids, with a maximum frequency observed for 16 to 19 residues. We avoided methionine, cysteine, 3x tyrosine, and rare codons for both humans and Escherichia coli (E. coli). With all the possible combinations, the theoretical diversity was 3.9 × 10^28, which could not be tested entirely due to practical limitations; hence, we used only 100 ng of VHH fragments (2.17–1.62 ×10^11 ds DNA copies for 450–650 bp synthesized VHH fragments, including adaptors) for library cloning. We used the sdAb^D10 scaffold.¹²

Cloning of semi-synthetic VHH library

pADL23c phagemid (Antibody Design Laboratories, #PD0111) was modified by cloning, i.e., we: 1) inserted toxic gene ccdB to get minimal self-ligation,¹³ and 2) added two new restriction enzyme sites; SbfI (NEB, #R3642S) at 5’ and PspXI (NEB, #R0656S) at 3’end of ccdB (named as pDL23c-ccdB phagemid) (Supplementary Figure S1A). VHH fragments were synthesized by Twist Bioscience (California, USA). To add overlapping regions specific to phagemid, 100 ng of VHH fragments were amplified by using Phusion High-Fidelity PCR master mix (Fisher Scientific, # F531L) (at a rate of 5 ng in 100 µL reaction) with MK_VHH_Phagemid_FP and MK_VHH_Phagemid_RP primers in low cycle PCR (Initial denaturation at 98°C for 30 sec, 10 cycles for denaturation at 98°C for 10 sec, annealing at 66°C for 30 sec, extension at 72°C for 10 sec, final extension at 72°C for 5 min) and PCR product ~500 bp were gel extracted by QIAquick Gel Extraction Kit (Qiagen, #28704). For ligation, 4.9 µg of VHH were mixed with 7.1 µg of the digested, purified phagemid vector, and ligated using the NEBuilder® HiFi DNA Assembly Master Mix (#E2621S). The ligation reaction mix was purified using a PCR purification kit, and the entire purified ligated product was used to transform 1.6 mL of TG1 electrocompetent cells (Biosearch Technologies, # 60,502–1). To calculate transformation efficiency, 100 µL of the transformation mix was taken after 1 h of recovery and serially diluted in 2x Yeast Extract Tryptone (2xYT) Medium (Fisher Scientific, #DF0440-17). The diluted mix was then spread on 2xYT Carb plates. The remaining transformed bacteria were superinfected with CM13-Interference-resistant helper phages (Antibody Design Laboratories, # PH020L) at a 20 multiplicity of Infection (MOI), incubated for 30 min at 37°C, then spun down at 4000 rpm to remove unbound phages. The resulting pellet was resuspended in 2 L of 2xYT media supplemented with carbenicillin (100 µg/mL) and kanamycin (50 µg/mL). After growing overnight, the phages were purified by the PEG/NaCl method.¹⁴ To calculate the total number of infective phages carrying the phagemid genome, 10 µL purified phages were serially diluted in 1 mL phosphate-buffered saline (PBS), and a further 10 µL was used to infect 100 µL TG1 (O.D. 0.4 at 600 nm). Purified phages were aliquoted in PCR tubes and stored at −80°C (Supplementary Figure S1B). The total number of transformants and phages carrying phagemid genome (size of phage library), and the total number of phages were calculated by UV absorbance-based method as described previously.^15,16 After electroporation of TG1 E. coli, we obtained 1.92 × 10^11 colony-forming units and 1.4 × 10^11 infective M13 virions with a phagemid genome. The entire phage library was divided into sub-libraries of 10^9 infective units (helper phage-derived VHH phages with phagemid genome) and amplified such that each clone was represented 5 × 10^3 times before panning.

Viruses

We used recombinant vesicular stomatitis viruses (rVSV) in which the G glycoprotein of VSV was replaced with the glycoproteins of EBOV, SARS-CoV-2, or JUNV (rVSV-EBOV-GFP,^17,18 rVSV-JUNV-GFP,¹⁹ rVSV-SARS-CoV-2-GFP,²⁰ and that contained enhanced green fluorescent protein (eGFP) to measure the infection in Vero cells. In brief, these rVSVs were rescued by co-transfecting the VSV N, P, and L plasmids, along with the plasmid expressing the glycoprotein gene of interest, into 293 FT cells. They were plaque-purified, sequence-verified, and propagated in Vero cells. Respiratory syncytial virus (RSV), kindly provided by Mark Peepels (The Ohio State University, Center for Microbe and Immunity Research), was subtype A with mKate2 as red fluorescent protein to score the infection. RSV was grown on Vero cells and purified by rapid freeze-thaw cycles, as described earlier.²¹

SARS-CoV-2 spike expression and purification

SARS-CoV-2 spike (Wuhan-Hu-1 (GenBank MN908947) construct was a generous gift from Jason McLellan (The University of Texas at Austin, Department of Molecular Biosciences).²² 25 mL of ExpiCHO-S^TM Cells (ThermoFisher Scientific, #A29127) at a density of 3 × 10^6 cells/mL were seeded 1 day before transfection. When the cells reached a density of 6 × 10^6 cells/mL, they were transfected with 30 μg of plasmid. The plasmid was first diluted in 920 µL of OptiPRO^TM SFM (ThermoFisher Scientific, #12309050) to achieve a final volume of 1 mL. In a separate tube, 80 µL of Expifectamine (ExpiFectamine^TM CHO Transfection Kit, ThermoFisher Scientific, #A29131) was diluted with 920 µL of OptiPRO^TM SFM. The contents of both tubes are mixed and added to the cells dropwise. Cells were incubated at 37°C with 8% CO₂ under humid conditions. Enhancer and feeder (ExpiFectamine^TM CHO Transfection Kit, ThermoFisher Scientific, #A29131) were added after 19 h post-transfection, and the cells were further incubated for 8 days post-transfection. As the spike protein contains a 6xHis tag as well as a double strep tag, it was purified by using the double strep tag II purification system as per the manufacturer’s instructions (IBA Lifesciences, Strep-Tactin® Sepharose® resin, Cat. No. 2–1201-002). The eluted protein was concentrated and buffer-exchanged with PBS. Its concentration was measured using a Nanodrop, and purity was checked by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE).

Phage purification and titer check

Supernatant of overnight grown cultures was incubated with 5x PEG/NaCl solution (20% PEG w/v and 2.5 M sodium chloride w/v) at 4:1 for 45 min–1 h on ice and then spun at 4000 rpm for 45 min-1 hour at 4°C. Supernatant was discarded, and a thin white pellet of phages was resuspended in PBS. Phage concentration was measured by the UV absorbance-based method as described by the manufacturer (Antibody Design Laboratories).

Virion per mL = (Absorbance at 269 nm – Absorbance at 320 nm) ×(6 × 10^16)/(number of bases/virion)

Panning for purified proteins and viruses

Helper phage-derived VHH phages from two vials of library stock were pooled and amplified by infecting TG1 (O.D. 0.4 at 600 nm) in 16 mL of media. After 30 min of infection, TG1 was superinfected with helper phages at a 20 MOI and incubated for an additional 30 min at 37°C. Unbound phages were removed by centrifugation, and the pellet was resuspended in 500 mL of fresh 2xYT media supplemented with carbenicillin (100 µg/mL) and kanamycin (50 µg/mL). After 19 h at 30°C, the phages were purified by the PEG/NaCl method. For panning, 100 µL of 10 µg/mL SARS-CoV-2 spike was coated in PBS overnight at 4°C in a full-area high-binding enzyme-linked immunosorbent assay (ELISA) plate (Corning, #9018). For viruses, we first selected the optimal amount of rVSV-EBOV-GFP and rVSV-JUNV-GFP for coating using ELISA, where different virus dilutions were coated overnight. The next day, after blocking in bovine serum albumin (BSA) and washings, 10 nM of positive-control antibody ADI 15878 mab for rVSV-EBOV-GFP (previously identified by our group and collaborators),²³ and CR1-07 for rVSV-JUNV-GFP²⁴ were added. The total protein concentration of the virus coated for panning was ~20 µg/mL for a 1:1000 dilution (1 OD = 1 mg/mL). We started the first round of panning at a 1:1000 dilution for both viruses and subsequently decreased the dilution in the next two rounds, completing three rounds in total. Panning was performed in 10 replicates, and ~10^13 phages were added per well.

Colony forming unit method:

After panning, eluted helper phage-derived VHH phages from test antigen-coated wells and in negative control wells (coated with BSA) were used to infect fresh TG1 E. coli at 0.4 O.D. After 30 min of infection, TG1 E. coli was spread on carbenicillin-2xYT plates and incubated overnight at 37°C. The next day, colonies of TG1 were counted, and their numbers were compared between the negative control (BSA) and the test antigen.

Phage ELISA

For monoclonal phage ELISA

Helper phage-derived VHH phages or hyperphage-derived VHH phages were prepared by the same method. Hyperphages were obtained as a lyophilized powder (PROGEN, #PRHYPE-XS) and resuspended in PBS to a concentration of 1 × 10^12 PFU/mL. Single bacterial colonies (obtained from eluted phages) were picked and inoculated into 96-well deep plates (one colony per well). After overnight incubation, 1% of these primary cultures were inoculated into fresh 200 µL 2xYT Carb⁺ media per well and incubated for 3 h at 37°C with 200rpm. After 3 h, the bacterial cultures reached approximately 0.4 O.D., and then helper phages or hyperphages were added to each well at an approximate 20 MOI. Incubations were performed at 37°C for 30 min, and then unbound residual helper phages or hyperphages were removed by centrifugation of the 96-well plates at 4000rpm for 10 min. The supernatant was removed, and 200 µL of fresh 2xYT Carb⁺ Kan⁺ media was added per well. Plates were incubated at 37°C for an additional 14–16 h with shaking at 200 rpm. The next day, the plates were centrifuged at 4000 rpm for 20 min, and ~200 µL of culture supernatant was transferred to new 96-well deep plates. VHH phages in culture supernatant were pre-blocked with 1x blocking buffer (3% skim milk in PBS with Tween 20 (PBS-T) at room temperature for 1 h at 100 rpm. 50 µL of blocked phages were added per well and incubated for 1 h at 37°C. After five washes with a plate washer, 50 µL of a 1:1000 diluted M13 Phage coat protein monoclonal antibody (A5B3) HRP (ThermoFisher Scientific, #MA5-36125) in blocking buffer was added per well. Plates were washed after 1 h, and 50 µL of 3,3‘,5,5’-tetramethylbenzidine (TMB) 1-step substrate (ThermoFisher Scientific, #34028) was added. The reaction was stopped with 50 µL of 0.5 M sulfuric acid.

For polyclonal phage ELISA

Eluted helper phage-derived VHH phages after each round were first amplified by infecting fresh TG1 E. coli and then purified the next day. The phage number was calculated by measuring the absorbance at 260 nm and used 10^12 VHH phages per well for ELISA.

ELISA

For the VHH-Fc concentration ELISA

The initial rough concentration of each VHH-Fc protein was estimated using a nanodrop. Starting from 10 µg/mL as the initial concentration, four-fold dilutions were prepared to coat ELISA plates (Corning, #9018) with 50 µL per well. The following day, plates were washed three times with PBS-T buffer (PBS pH 7.4 with 0.05% Tween 20) using a plate washer (Biotek). Then, they were blocked with 3% BSA in PBS-T for 1 h at 37°C. After 1 h of incubation at 37°C, 50 µL of a 1:10,000 diluted goat anti-human IgG (H+L) cross-adsorbed secondary antibody (Invitrogen, # A18811) in PBS-T was added to each well. Plates were washed after 1 h, and 50 µL of TMB 1-step substrate was added. The reaction was stopped with 50 µL of 0.5 M sulfuric acid.

For the VHH-Fc binding ELISA

100 µL of 10 µg/mL pure SARS-CoV-2 spike or ~10 µg/mL of rVSV-SARS-CoV-2-GFP/rVSV-EBOV-GFP/rVSV-JUNV-GFP in PBS was coated to full area high-binding ELISA plates (Corning, #9018) for overnight at 4°C (the amount of virus was measured in terms of 1 absorbance at 280 is equal to 1 mg/mL). The next day, plates were processed as described above, and four-fold dilutions of VHH-Fc protein were prepared in the TBS buffer, and 50 µL of diluted VHH-Fc was added per well. After 1 h of incubation at 37°C, 50 µL of a 1:10,000 diluted anti-human Fc in PBS-T was added per well. Plates were washed after 1 h, and 50 µL of TMB 1-step substrate was added. The reaction was stopped with 50 µL of 0.5 M sulfuric acid.

VHH-Fc expression and purification

VHH-Fc was cloned into the pTwist expression vector from Twist Bioscience. For expression, we used 24-well U-bottom plates and seeded 3 × 10^6 cells/mL ExpiCHO-S^TM cells 1 day before transfection. The next day, 6 × 10^6 cells/mL were transfected using a transfection kit (ExpiFectamine^TM CHO Transfection Kit, ThermoFisher Scientific, #A29131). Briefly, 3 µg of plasmid was diluted in 100 µL OptiPro^TM SFM and mixed with diluted expifectamine. This mixture was then added to cells after 1–5 min of incubation at room temperature. After 19–22 h post-transfection, 15 µL of enhancer and 600 µL of feeder were added per well. Cells were incubated for 8 days, and then protein was purified using protein A agarose beads according to the manufacturer’s protocol (GenScript, #L00210).

SDS-PAGE

30 μL of ELISA-normalized 1 μM for all VHH-Fcs were mixed with 10 µL of 4x sample Laemmli buffer with dithiothreitol and boiled for 5 min at 95°C, spun down, and loaded 25 µL of sample per well in 4–20% tricine gels. After electrophoresis, the gels were stained with Coomassie Blue stain and destained with a destaining solution (45% methanol, 10% acetic acid, and 45% water). The gels were stored in water until imaging. IBright was used for imaging.

Cell lines

Vero cells procured from ATCC, grown in DMEM with 2% fetal bovine serum, 1X Penicillin-streptomycin, 1x Glutamax at 37°C/5%CO2. These were cultured every 3 to 4 days. ExpiCHO-S^TM Cells were maintained in ExpiCHO^TM expression media (ThermoFisher, # A2910001) at 37°C/8% CO2/120 rpm with 6–8 ×10^6 cells per mL.

Neutralization assay with phages/VHH-Fcs

One day before the neutralization assay, Vero cells were seeded in a 96-well plate and incubated at 37°C with 5% CO₂. We used different combinations of virus MOI and incubation times to get 30–40% infection. For low-MOI neutralization assays, we seeded 10,000 Vero cells per well; for high-MOI neutralization assays, we seeded 15,000 Vero cells per well. The next day, purified phages or diluted VHH-Fc were preincubated with the virus for 1 h at room temperature and then added to the cells. Premixes were prepared by combining 210 μL of each concentration of VHH-Fc with 70 μL of diluted virus, resulting in a total volume of 280 μL. After a one-hour incubation, 80 μL of the mix was added to each well of Vero cells. Cells were fixed after 7–8 h for a high MOI of 0.4 to 0.5 or 20–24 h for a low MOI of 0.01. The endpoint of infection is typically 20–40% GFP-positive cells, depending on the MOI and time frame. Infection was scored using the Biotek Cytation 5 imaging reader.

Next-generation sequencing (NGS) sample preparation

The phagemid genome was isolated by the Phenol-Chloroform-Isoamyl alcohol (PCI) method from ~10^13 phages from each round of panning. For CDR3 amplification, 10 μg ssDNA phagemid was used as a template for gradient PCR (at a concentration of 100 ng in a 10 μL reaction) with Phage_NGS_FP_MK and Phage_NGS_FP_MK primers. For full-length VHH amplification, 600 ng ssDNA phagemid was used as a template for gradient PCR (at a concentration of 100 ng in a 100 µL reaction) with MK_NGS_VHH_FP and MK_NGS_VHH_RP primers. Initial denaturation at 98°C for 30 sec, 10 cycles for denaturation at 98°C for 25 sec, annealing at 60.5–70.5°C for 50 sec (12–15 cycles), extension at 72°C for 50 sec, final extension at 72°C for 5 min. PCR products, ~130 bp for CDR3 and ~500 bp for full-length VHH, were gel-extracted using the Qiagen Gel Extraction Kit. Samples were prepared for Illumina sequencing using the KAPA HyperPrep Kit with Library Amplification (Roche, # 07962347001). The final product was bead-purified and used at a concentration of 10 ng/µL for Illumina MiSeq 150 × 2 paired-end sequencing of the CDR3 region and Illumina MiSeq 250 × 2 paired-end sequencing for the full VHH. DNA concentrations were measured using the Qubit dsDNA BR reagent on a fluorometer.

NGS sample analysis

Paired-end read files were merged by NGmerge, and then unique molecular identifiers (UMIs) were located by using the primer information at the 5’ and 3’ ends of paired reads. Duplicated sequences were discarded based on the same UMIs. To calculate the overall diversity of all three CDRs, the sequence from the last six nucleotide residues of framework region 1 (FR1) to the starting six nucleotide residues of FR4 was used, and for the calculations of only CDR3 diversity, sequences were extracted between the last six nucleotide residues of FR3 and the starting six nucleotide residues of FR4 regions. To assess protein-level diversity, reads were translated, and sequences containing premature stop codons and/or duplicated reads were excluded. The raw reads of NGS from all samples have been submitted to the European Nucleotide Archive (ENA) with Accession no. PRJEB101623.

The following formulas were used to calculate:

Diversity at the amino protein level = (No. of unique reads in frame)/(Total no. of reads in frame)

The fold change of each sequence = (no. of reads in round “x”)/(No. of reads in round “x-1”)

Sequencing

0.5–1 µg of plasmid was sequenced by plasmidsaurus using Oxford Nanopore sequencing and Azenta by Sanger sequencing.

Statistical analysis

GraphPad Prism was used to plot Heat maps, area under the curve (AUC) plots, and non-linear regression curves. AUCs were plotted using the same concentration range of VHH-Fc for ELISA and virus neutralization assay. Statistical comparisons were performed using the two-tailed t-test, with *p < 0.05, **p < 0.01, and ***p < 0.001.

Results

Pentavalent VHH phages show enhanced viral neutralization compared to monovalent VHH phages

In the phage display platform, filamentous phages are used to display antibodies (Fab/VHH/scFv) or peptides on their surface.²⁵ We have used CM13-Interference-resistant helper phage, which is a derivative of M13K07. It contains a full-length pIII protein and displays VHH as a fusion protein on its surface.²⁶ It provides a monovalent display of VHH. We hypothesized that “helper phage-derived VHH phages” (hereafter termed monovalent VHH phages) could be used to neutralize viral infection in vitro (Figure 1(A)). We engineered monovalent VHH phages carrying previously published single-domain antibodies against viral glycoproteins, either VHH72 targeting the SARS-CoV-2 spike²⁷ or VHH4, which binds the prefusion F protein of respiratory syncytial virus (RSV),²⁸ and an anti-beta-lactamase (αBL) VHH targeting β-lactamase of E. coli (clone cAbTem13) as a control.²⁹ Upon incubation of PEG/NaCl-purified phages displaying monovalent VHH72/VHH4 with rVSV-SARS-CoV-2-GFP and RSV mKate2, respectively, we observed a dose-dependent neutralization of both viruses (Figure 1(B and C)). We reasoned that the relatively low neutralization efficiency observed with monovalent VHH phages could be due to their limited ability to display a single copy of VHH. It is known that the helper phages-derived monovalent display of VHH/scFv or peptides is always less than 10% and polyvalent display is further less than 1% of total pIII display³⁰; hence, neutralization with monovalent VHH phages may not be consistently reproducible.

Figure 1.

Pentavalent VHH phages neutralize the virus infection more effectively than monovalent VHH phages. A. Schematic of the VHH-Fc discovery platform (created in BioRender. Kaur, M. (2025) https://BioRender.com/fyzytws). B. Neutralization of rVSV-SARS-CoV-2-GFP at MOI 0.004 with monovalent VHH72 phages and control VHH phages (cells fixed after 18 hours of infection). C. Neutralization of respiratory syncytial virus (RSV) at MOI 0.4 with monovalent VHH4 and control VHH (cells fixed after 24 h of infection). Because of the VSV’s replication rate, a very low MOI was used to achieve 20–30% infection in 18 hours, compared to RSV. D. Structural and functional comparison between helper phage and hyperphage-derived VHH phages, along with neutralization of rVSV-SARS-CoV-2-GFP at MOI 0.032 (cells fixed after 22 hours of infection).

We speculated that neutralization based on monovalent VHH phages can be improved further by utilizing the pentavalent display feature of hyperphages. Hyperphages are like helper phages except that they do not have the pIII gene; therefore, they rely on a special strain of E. coli named E. coli DH5α/pIII, in which hyperphage progeny is produced only when this E. coli is transfected with M13KO7ΔpIII helper phage. All hyperphage-derived phages display five copies of VHH on their surface as a fusion protein with pIII, allowing for improved uniformity.³⁰ Therefore, we hypothesized that hyperphage-derived VHH phages (also referred to as pentavalent VHH phages), would be more effective at neutralization compared to monovalent VHH phages. To test this hypothesis, dose-dependent differences in neutralization by VHH phage-derived from hyperphages and helper phages were examined. We fixed the cells when there was > 90% infection in the control groups (No phage group). Interestingly, pentavalent VHH72 phages showed enhanced neutralization of rVSV-SARS-CoV-2-GFP compared to monovalent VHH72 phages (Figure 1D), even when the infection was almost saturated. Overall, this suggests that pentavalent VHH phages can neutralize viral infection even at high MOI, whereas monovalent VHH phages are unable to do so. These results prompted us to test pentavalent VHH phage-mediated neutralization as a screening method on our semi-synthetic VHH library (Supplementary Figure S1 and Table S1) to select potent antiviral VHHs.

Construction of a semi-synthetic phagemid library

We designed and constructed a semi-synthetic VHH library as described in the Materials and Methods section. The size of the library was 10^13 colony-forming units (CFU), of which only a subset (2 ×10^9 CFU) was used for downstream processing. The library was further amplified 5 × 10^3 times (final CFU: 10^13) before panning to increase the initial population of individual clones, mainly because obtaining positive hits from semi-synthetic libraries has been more challenging compared to libraries made from camelids immunized with the antigen of interest.³¹ We tested whether our RASP method could be used to shortlist virus-neutralizing VHHs from our semi-synthetic library. Before using the sub-library, we verified its quality through Illumina sequencing. The results were comparable to the diversity from PDB sequences and the designed diversity (Figure 2(A–C)). The length distribution of our library closely matched that of PDB sequences and the initial design (Figure 2(D–F)). We also noted a slight variation in CDR2 at position 6, and a very low to negligible presence of aspartic acid between positions 8 and 11. Overall, the diversity patterns of amino acid percentages across different CDRs and the CDR3 length distribution were consistent with our design.

Figure 2.

Construction of a semi-synthetic VHH library. A-C. Diversities of CDR1, CDR2, and CDR3 from the protein data Bank (PDB), designed versus observed from next-generation sequencing in the helper phage display library, respectively (B-C, showing CDR3 with 18 amino acids long). D-F. CDR3 length distribution from the protein data Bank, designed versus observed from next-generation sequencing in the helper phage display library, respectively.

RASP can filter anti-viral VHHs from the top hits identified by ELISA/NGS

Our next step was to assess whether we could implement RASP to downselect VHH hits identified by traditional screening methods. To isolate VHHs from our semi-synthetic library, we performed three to five rounds of panning using a recombinant SARS-CoV-2 spike protein. Enrichment for VHH binders was verified by polyclonal phage ELISA, in which the capacity of VHH phages eluted and amplified after each round of panning to bind to SARS-CoV-2 spike vs. BSA was tested (Supplementary Figure S2A). We also checked the enrichment by CFU (Supplementary Figure S2B), suggesting non-significant differences among different panning rounds. In the polyclonal phage ELISA, we found a significant increment in the binding signal for SARS-CoV-2 spike. Accordingly, we performed monoclonal phage ELISA from round 5. We identified 32 hits above a 5-fold cutoff representing only one unique VHH sequence (Figure 3(A)). The CDR3 of this exact clone was also captured by NGS with a frequency of 12% in round 3 (second most representative), 28% in round 4 (second most representative), and the topmost representative with 60% frequency in round 5, which suggests its over-representation in monoclonal phage ELISA (Figure 3(B)). As the enrichment was already observed in panning round 3, to test RASP as a filtering method, we picked the top 10 CDR3 sequences from round 3 of NGS, retrieved the whole VHH sequence from the round 3 phagemid DNA pool (Supplementary Figure S2C-D) by overlap PCR, and then cloned the sequences into modified phagemid as well as mammalian expression vectors. First, we demonstrated that upon display of these top 10 VHHs, either on helper phage or hyperphage-derived VHH phages, only VHH2 showed binding to SARS-CoV-2 spike and rVSV-SARS-CoV-2-GFP, comparable to the positive control (VHH72).

Figure 3.

RASP as a filtering method. A. Monoclonal phage ELISA (using monovalent VHH phages) after the 5th round of panning. B. Plot showing the distribution of the top 100 CDR3 hits from round 3 across different rounds of panning by next-generation sequencing. C. Phage ELISA for SARS-CoV-2 spike protein and rVSV-SARS-CoV-2-GFP (Fold change = binding signal for test antigen/BSA). D. Neutralization of rVSV-SARS-CoV-2-GFP at MOI 0.4 by monovalent and pentavalent VHH phages displaying the top 1–10 VHHs. E-F. ELISA and neutralization assays with purified top 1–10 VHH-Fc proteins for rVSV-SARS-CoV-2-GFP. Statistical comparisons were performed between test VHH phages and anti-BL-VHH phages (negative control) using the two-tailed t-test, with *p < 0.05, **p < 0.01, and ***p < 0.001 (n = 3). G. Heatmap showing IC₅₀ nM of the rVSV-SARS-CoV-2-GFP neutralization at different MOIs (0.003 and 0.3) by the top 1–10 VHH-Fc.

In contrast, the remaining VHHs had only 1.5- to 2-fold binding over the BSA-coated wells (negative control) (Figure 3(C)). To test their neutralization capability, we preincubated 2.5 × 10^11 monovalent or pentavalent VHH phages with the virus at a high MOI of 0.48 (20–30% infection in ~7 h) and then added them to Vero cells. None of the monovalent VHH phage neutralized rVSV-SARS-CoV-2-GFP infection. However, three of 10 VHHs from the pentavalent VHH phage group showed VSV neutralization, which suggests that not all the top 10 hits are neutralizers (Figure 3(D)). To further validate our findings, we normalized the concentration of purified VHH-Fcs by Fc ELISA (Supplementary Figure S2E-F) and then performed rVSV-SARS-CoV-2-GFP binding ELISA (Figure 3(E)) and neutralization assays (Figure 3(F)). We observed that all VHHs tested bind to rVSV-SARS-CoV-2-GFP at high concentrations (8^4 infection units per well, 1:50 virus dilution based on the signal for the positive control VHH72); however, only 6 of 10 VHH-Fcs neutralize VSVs with IC₅₀ values in the 5–300 nM range (Figure 3(G)). Overall, the RASP accurately predicted 8 of 10 test clones and 2 of 2 controls, indicating that only 3 of the top 10 VHHs were neutralizing. It suggests that not all top binders are neutralizers; hence, they can be identified and selected directly by this method. Second, as NGS top hits are also present in output colonies formed by eluted helper phages, NGS may be bypassed, allowing more colonies to be screened directly by RASP.

Panning using rVSV-EBOV-GFP and rVSV-JUNV-GFP

To further characterize our novel VHH screening method, we used it to pan for neutralizing VHHs against rVSV-EBOV-GFP and rVSV-JUNV-GFP. We aimed to apply this method to other viruses and select neutralizing VHHs within a short time frame. We performed panning for Ebola and Junin virus glycoproteins by using our recombinant VSVs displaying the respective viral glycoproteins (rVSV-EBOV-GFP and rVSV-JUNV-GFP) as antigens.^32,33 For both viruses, we observed an atypical enrichment pattern of virus binders. Specifically, in the rVSV-EBOV-GFP – specific polyclonal ELISA, the phage population after each round of panning showed decreased binding to rVSV-EBOV-GFP until round two and then increased in round three. However, we observed decreases in enrichment when using the CFU method (Supplementary Figure S3A-C). Panning with rVSV-JUNV-GFP showed that enrichment increased after panning round one, followed by a dip after round two and a sharp increase after round three (Supplementary Figure S3D-F). There was no enrichment in this panning according to the CFU method. Nonetheless, our findings indicate that low dilutions of VSV-based surrogate viruses can be used for panning.

Screening for rVSV-EBOV-GFP VHHs by RASP

We screened for neutralizing VHHs against rVSV-EBOV-GFP from phages eluted in panning round 3 using three different methods: phage ELISA (Method I), phage neutralization (Method II), and NGS (Method III) (Figure 4A). For methods I and II, we used TG1 E. coli colonies formed by eluted monovalent VHH phages. We inoculated a single colony per well of 96-deep-well plates and then infected them with hyperphages at 0.4 O.D. For method III, we used the phagemid genome of the amplified monovalent VHH phage population after each round of panning for NGS. For direct comparisons in methods I and II, we used only pentavalent VHH phages to ensure the uniform display of VHHs. From method I, five of 272 screened clones showed at least a two-fold signal enrichment for rVSV-EBOV-GFP as compared to BSA (Figure 4(B), top panel, and Supplementary Figure S4A). We selected a total of eight clones with a fold change greater than 1.9 for further analysis. For method II, we tested different combinations of incubation time and virus doses. We observed that higher MOI (~20–40% infection in 7 h) yielded better resolution than lower MOI (to ~ 20–40% infection in 20 h), as almost all the clones neutralized rVSV-EBOV at low MOI, but only a few clones did so at high MOI. We shortlisted 24 hits (combining from high and low MOI neutralization) from 272 clones based on a cutoff set by the neutralizing activity of a negative control VHH (anti-ß-lactamase VHH) (Figure 4(B), middle panel, and Supplementary Figure S4B). For method III, we first compared the enrichment of the top 100 VHH sequences from each panning round with the starting panning phage pool. Here, we used the top 100 VHH sequences from all rounds of rVSV-JUNV-GFP panning as a negative control. We observed that not all the top 100 sequences from the rVSV-EBOV-GFP round 1 were enriched in subsequent rounds, possibly due to the low antigen amounts in rounds 2 and 3 (Figure 4(B), right panel). Further, some of these sequences were also present across the different rounds of rVSV-JUNV-GFP panning, suggesting the enrichment of some nonspecific binders or “sticky” VHHs. Sequences from rVSV-EBOV-GFP panning round 3 were highly enriched and minimally shared with rVSV-JUNV-GFP panning; therefore, we selected only the top 20 enriched VHH sequences from panning round 3 (Supplementary Figure S4C). During subcloning of these shortlisted VHHs, we observed that only three clones from method I and 20 from method II were PCR-positive (potentially due to recombinational loss of these VHH sequences from the phagemids in TG1 E. coli), and their enrichment varied across different rounds of rVSV-EBOV-GFP panning (Supplementary Figure S4D).

Figure 4.

RASP as a screening method after rVSV-EBOV-GFP panning. A. Schematic of panning and screening. B. rVSV-EBOV-GFP binding ELISA (method I) (Fold change = binding signal for test antigen/BSA) and neutralization assay at high MOI 0.56 (cells fixed after 7 hours of infection) and at low MOI 0.01 (cells fixed after 20 hours of infection) with pentavalent VHH phages from different rounds of panning (method II) (left panel). Enrichment of the top 100 VHH hits across three rounds of panning of rVSV-EBOV-GFP versus nonspecific enrichment in rVSV-JUNV-GFP panning (method III) (right panel). B. ELISA and neutralization (MOI 0.2, with cells fixed after approximately 9 hours) using shortlisted VHH-Fc from all three methods.

Furthermore, some clones could not be mapped to the starting library using NGS, likely due to the incomplete sequencing coverage of this large semi-synthetic library. We proceeded with three clones from method I, 20 clones from method II, and 20 clones from method III for expression in a mammalian system. All VHHs were expressed as Fc fusions, and their concentrations were normalized by Fc ELISA. EBOV_R3_VHH 116 was poorly expressed and was not used further.

From method I, only one VHH clone (EBOV_R3_93) showed both remarkably high binding to the virus with moderate neutralization potency (Figure 4(C), Table S2, and Supplementary Figures S4E-F). Still, this clone was not specific to rVSV-EBOV panning as it was also enriched to some extent in rVSV-JUNV-GFP panning, suggesting that it is sticky or cross-reactive. Among the other two VHH clones, EBOV_R3_99 showed moderate binding, while EBOV_R3_64 showed low binding to the virus, but neither neutralized rVSV-EBOV. EBOV_R3_64 was also present in method II. From method II, all clones exhibited binding to the virus, but only 8 of 20 clones demonstrated viral neutralization (Figure 4(C), Table S2, and Supplementary Figures S4G-H). Clones EBOV_R3_72, EBOV_R3_195, and EBOV_R3_226 were among the most potent neutralizers. From method III, all clones showed moderate to low-level binding, but this was on a scale from poor to non-neutralizing (Figure 4(C), Table S2, and Supplementary Figures S4I-J).

Our data indicate that only three clones were identified from phage ELISA screens, and just one VHH clone was able to neutralize infection with moderate efficiency. Conversely, using pentavalent VHH phage-mediated neutralization (RASP), we successfully selected those VHH clones, a method I had previously failed to achieve, demonstrating the superior sensitivity of this approach/platform. As a result, we obtained several VHHs that neutralize rVSV-EBOV-GFP more effectively within 7 h at an MOI of 0.56. Lastly, from method III, we did not observe any neutralization from the top 20 VHH hits. All these hits exhibited strong binding, suggesting that further hits needed to be screened, which is a time-consuming and labor-intensive process. Compared to other methods, the hyperphage-mediated neutralization method proves to be highly efficient and reliable.

Screening for rVSV-JUNV-GFP VHHs by RASP

To further validate the pentavalent VHH phage-mediated neutralization approach (RASP), we replicated the workflow described above using rVSV-JUNV-GFP. We shortlisted 10 clones with method I (Figure 5(A), top panel, and Supplementary Figure S5A), 33 clones with method II (Figure 5(A), middle panel, and Supplementary Figure S5B), and the top 20 enriched VHHs from method III (Figure 5(A), right panel, and Supplementary Figure S5C). Only 36 clones were PCR-positive for VHH. Overall, we used six clones from method I and 30 from method II (including 26 from the High MOI screen and 4 from the Low MOI screen) in downstream assays. Nearly all shortlisted VHHs from methods I and II were captured in the NGS. VHH clones from 1 to 2 rounds of panning were not enriched in later rounds, except clone JUNV_R2_111. VHH clone JUNV_R3_18 from method II, clone JUNV_R3_45, and clone JUNV-R3_83 from method I were also the top three hits in the NGS analysis (Supplementary Figure S5D). Only JUNV_R3_62 from method I showed high binding with moderate neutralization (Figure 5(B), Table S3, and Supplementary Figures S5E-F). Two clones, JUNV_R3_164 and JUNV_R3_161, neutralized infection by approximately 60% at the maximum VHH-Fc dose used (Figure 5(B), Table S3, and Supplementary Figures S5G-H). Method III did not produce any clones with significant neutralizing potential (Figure 5(B), Table S3, and Supplementary Figures S5I-J). All VHHs were expressed as Fc fusions, and their concentrations were normalized by Fc ELISA (Supplementary Figures S6,7, 8).

Figure 5.

RASP as a screening method after rVSV-JUNV-GFP panning. A. rVSV-JUNV-GFP binding ELISA (method I) (Fold change = binding signal for test antigen/BSA) and neutralization assay at high MOI 0.2 (cells fixed after 7 hours of infection) and at low MOI 0.04 (cells fixed after 14 hours of infection) with pentavalent VHH phages from different rounds of panning (method II) (left panel), enrichment of the top 100 VHH hits across three rounds of panning of rVSV-JUNV-GFP versus nonspecific enrichment in rVSV-EBOV-GFP panning (method III) (right panel), B. ELISA and neutralization (MOI 0.03 and cells fixed after 14 hours of infection) with shortlisted VHH-Fc from all three methods.

Discussion

Phage display is a widely used platform for the discovery of single-domain antibodies against target antigens. One advantage of this approach is its capacity to accommodate large and highly diverse libraries. However, panning these large semi-synthetic or naïve libraries for binders can be more challenging than panning libraries derived from hyperimmune animals due to the much lower proportion of antigen binders in the former. A typical solution is to perform several rounds of panning to sufficiently enrich for clones that bind the target antigen, which may also lead to the unintended enrichment of high-replicating, nonspecific, or poorly binding clones. Traditional selection methods are time-consuming, with an initial ELISA-based selection. The shortlisted clones are subsequently expressed in bacteria or mammalian cells to assess their functional properties and down-select to highly active molecules. Several functional screening methods have been described in which antibodies are selected directly based on their functionality. For example, the discovery of protease inhibitors by expressing a scFv or VHH library inside bacteria³⁴ or selecting agglutinating antibodies based on the altered morphology of target bacteria³⁵ has been studied. Similarly, mammalian cell surface display has been used to identify anti-viral VHHs,³⁶ as well as activators or inhibitors of cell surface receptors, such as G protein-coupled receptors, utilizing cytoplasmic signals and fluorescent reporters.³⁷ Other indirect functional screens, which are based on enrichment, utilize mass spectrometry and NGS. Using proteomics, HCAbs or single-domain antibodies from the serum of immunized or naive animals can capture antigens, and positive hits can be evaluated by mass spectrometry.^38,39 NGS, another high-throughput technique, has been extensively used alone or in conjunction with various screening platforms, such as phage display, mass spectrometry, and yeast display, to identify lead candidates based on their binding properties during screenings.⁴⁰ All these methods require either one or more subcloning/processing or additional equipment for functional readouts. Here, we have demonstrated that the discovery timeline of neutralizing antibodies can be significantly decreased by using purified pentavalent VHH phages (hyperphage-derived VHH phages). Our RASP platform follows the same steps as phage ELISA, but the final step involves purifying the VHH phages, which can then be used in cell culture. In a prior study,⁴¹ Stefan et al. monitored the panning progress by rVSV-SARS-CoV-2-GFP neutralization with polyclonal VHH phages, and a CFU method was used to equalize the number of VHH phages. It is known that the CFU method, which reflects the infective titer of phages, is always lower than the physical phage titer because not all phages remain infectious after purification. To address this, we used a phage neutralization assay, modifying it by first using absorbance-based titer matching rather than infectivity units and second by using hyperphage-derived VHH phages (pentavalent VHH phages) to enhance VHH display. In the absorbance-based method, the phage concentration was measured at 269 nm, and then virions per mL were calculated (see Methods).

We have demonstrated that the neutralization potential of VHH phages can be improved by using hyperphages instead of helper phages. Helper phages display monovalent VHH on pIII, with up to 10% showing one VHH on their surface, and less than 1% displaying more than one VHH.³⁰ This variation in VHH display among helper phage-derived VHH phage populations may result in inconsistent results. We also observed frequent disruption of the genotype-to-phenotype linkage in helper phages compared to hyperphage-derived VHH phages. Although helper phage-infected TG1 bacteria can still form colonies on antibiotic-containing LB plates, PCR did not show VHH amplification. This is likely due to recombination events in TG1 bacteria, since they are recA compared to commonly used E. coli cloning strains.⁴² To confirm the disruption of genotype-to-phenotype linkage, we performed western blots on a few clones of purified pentavalent VHH phages, where VHH was detected as a fusion protein with pIII (data not shown). Additionally, sequencing the entire phagemids revealed multiple recombined variants, with only some carrying a full-length VHH gene (data not shown). Overall, this suggests that the phagemids have undergone recombination, resulting in the loss of genotype-phenotype linkage. Using hyperphages instead of helper phages allows maintenance of this linkage over multiple rounds of panning, as the phagemid is the sole source of pIII for pentavalent VHH phages assembly and release from the host cell. We also tested different combinations of virus MOI and endpoint times, resulting in 20–30% infection rates, and found that 7 h of infection was optimal for rVSV-SARS-CoV2 and rVSV-EBOV-GFP and 14 h for rVSV-JUNV-GFP for pentavalent VHH phages-mediated neutralization-based screening.

We mapped the VHH hits obtained through the pentavalent VHH phages-mediated neutralization method (RASP) using NGS and found that only a few of them were sequenced, highlighting its coverage limitations. Another reason could be the use of eluted phages for RASP, whereas for NGS, we first amplified the eluted phages to recover enough phagemid for sample preparation. It is also possible that those positive clones did not replicate as efficiently as other nonspecific or weak binders in the eluted pool.

We have also shown that VSV-based surrogate systems can be used for panning, although we did not get an enrichment pattern from round one to round three. One potential explanation for this finding is that we started with a very low concentration of virus, which was detectable by 10 nM mAb within 20–30 min by ELISA (Supplementary Figure S3A and S4A), and then further reduced the amount of virus in each round of panning. Thus, only strong glycoprotein binders were expected to be retained in the subsequent rounds of selection. Another reason could be that eluted helper phages failed to infect the TG1, or that the reduction of non-specific/weak binders in round one, which did not bind at lower virus doses, occurred. One limitation of RASP is that it must be used in viral assays where a reporter gene or another functional readout is available to measure infection in a high-throughput manner. Neutralization results are semi-quantitative, as pentavalent VHH phages are not counted for each well.

In the rapidly evolving field of antibody engineering, our method offers an efficient way to identify functional nanobodies that target viral glycoproteins from a semi-synthetic library. Since RASP is a modular and standalone technique, it can be used independently or integrated into standard phage ELISA or NGS-based antibody discovery pipelines to shortlist potent antivirals. It provides a clear advantage over traditional phage-based screening methods by delivering comparable results within the same timeframe, while also eliminating the need for additional subcloning steps.

Funding Statement

The work was supported by the National Institutes of Health [R01AI132256].

Abbreviations

BSA

Bovine serum albumin

CDR

Complementary-determining regions

CH1

First constant domain of heavy chain

CHO

Chinese hamster ovary

CFU

Colony-forming units

EBOV

Ebola virus

ELISA

Enzyme-linked immunosorbent assay

GFP

green fluorescent protein

HCAb

Heavy chain antibodies

JUNV

Junin virus

LB

Luria-Bertani

MOI

Multiplicity of infection

NGS

Next-generation sequencing

OD

optical density

PBS

Phosphate-buffered saline

PCR

Polymerase chain reaction

PDB

Protein Data Bank

PEG

Polyethylene glycol

RSV

Respiratory syncytial virus

SARS-CoV-2

Severe Acute Respiratory Syndrome Coronavirus 2

SdAb

Single domain antibody

ScFv

Single chain fragment

VHH

Variable domain of heavy chain antibodies

VSV

Vesicular stomatitis virus

Disclosure statement

K.C. holds equity in Integrum Scientific, LLC and Eitr Biologics, Inc.

Supplementary Information

Supplemental data for this article can be accessed online at https://doi.org/10.1080/19420862.2026.2627708