Discovering novel therapeutic VHHs for emerging perspectives from VEEV selection strategies

Authors: Autumn T. LaPointe, Fortunato Ferrara, Jennifer M. Zupancic, Alba L. Montoya, Jurgen Schmidt, Li-Wei Hung, Alison M. Kell, Nileena Velappan

Abstract

Introduction

Evolution or emergence of a new viral variant is a significant public health concern. Alphaviruses, such as Venezuelan equine encephalitis virus (VEEV), are mosquito-borne viruses which are becoming more prevalent due to expansion of vector habitats. Despite this, there are currently no antiviral therapies or FDA-approved vaccines available to treat or prevent VEEV infection. The increased prevalence of such viruses provides opportunities for novel variants to evolve. Key therapeutic molecules that could be developed against viral pathogens are recombinant antibodies or antibody fragments, such as the variable heavy domain of heavy chain antibodies (VHHs).

Methods

In vitro selections offer a promising pathway for identification of therapeutic antibodies, here we explored isolation of VHHs using phage and yeast display methodology with three antigen formats 1) recombinant E2, 2) linear peptides of E2, selected based on molecular dynamics analysis, and 3) UV inactivated virus.

Results

Here we report four novel “human” VHHs which bind to the VEEV E2 protein selected using different strategies that include both computational and biochemical design of suitable antigens and whole virus selections. These VHHs have distinct complementarity-determining regions (CDRs). Multiple VHHs bind to the VEEV viral particles in ELISAs, and we report the peptide epitope recognized by these VHHs.

Discussion

Though non-neutralizing, these VHHs bind to and sequester VEEV viral particles preventing infection, demonstrating the potential of these VHHs to perform viral “sponging” which represents a novel therapeutic approach. The selection strategies we report may have applications to further antibody developments against other viruses.

Introduction

Emerging pathogens with pandemic potential pose a serious threat to global public health. These viruses may emerge through natural evolution and zoonosis transmission or may be potentially released into susceptible populations either accidentally (e.g., laboratory leaks) or intentionally. Mosquito-borne viruses represent a particularly concerning risk in these scenarios due to socioeconomic conditions in tropical regions and the ongoing expansion of mosquito habitats driven by climate variations. Consequently, the development of novel therapeutics and innovative methods for generating these molecules is critical to safeguarding public health.

Alphaviruses are positive-sense, single-stranded RNA arboviruses capable of causing severe disease. Their natural enzootic transmission cycle involves a mosquito vector and mammalian or avian hosts; however, epizootic spillover events can lead to infection in humans and equids. Alphaviruses are broadly classified based on disease outcome as either arthritogenic or encephalitic. Arthritogenic alphaviruses, such as Chikungunya virus (CHIKV) and Ross River virus (RRV), cause illnesses ranging from mild febrile episodes to severe, persistent polyarthralgia that can last weeks to years’ post-infection (1–3). In contrast, encephalitic alphaviruses, such as Venezuelan equine encephalitis virus (VEEV), produce wide range of neurological symptoms, including potentially fatal encephalitis (3, 4). VEEV is of particular concern as biodefense threat due to its high infectivity and potential for aerosol transmission (5, 6). Despite the significant threat to public health, there are currently no FDA-approved vaccines or antiviral therapies available for prevention or treatment.

Antibody-based immunotherapy has proven to be a powerful tool against infectious diseases as demonstrated recently during the Coronavirus disease 2019 (COVID-19) pandemic (7–9). Antibodies are highly effective therapeutics due to their target specificity, structural adaptability, and diversity (9). The variable heavy domain of heavy chain antibodies (VHHs), also referred to as nanobodies, retain the binding ability of an antibody with a single protein domain and offer several advantages over conventional IgGs, including enhanced solubility, simplified engineering (10), and superior stability across a broad pH range (11–13).

Moreover, VHH domains are easily engineered in combination with other antibody formats. For instance, IgGs fused to additional VHH domain(s) have been actively pursued as therapeutics because they enable binding to multiple epitopes (14). Recently, Mei et al. (10) described VHHs developed against various viruses as well as currently available VHH fusion proteins. In their recent review on VHH antibody discovery and potential application, Alexander et al. (11) denotes several potential therapeutic applications for “human” or humanized VHH antibodies in cancer biology, neurodegenerative diseases, autoimmune diseases, and infectious diseases. These publications highlight the unique niche for VHHs as protein domains that can be tethered to other molecules to facilitate multi-specificity and the corresponding multifunctionality.

In vitro selections have been widely applied to the selection of anti-viral VHHs due to the control provided over both the target antigen and the composition of VHHs within the immune or synthetic library interrogated for binding (15). Synthetic VHH libraries that use human variable heavy chain as the scaffold for inserting Complementarity-Determining Regions (CDRs) allow for construction of VHH libraries (16). This design aims to address “humanness”, stability, and manufacturing issues (17). Libraries based on therapeutic VHH scaffolds that have been extensively studied and used in clinical trials provide frameworks that are well behaved with excellent developability properties (18). If the library and consequently the selected VHHs contain human CDRs, this further facilitates their potential use as therapeutics.

VHH-based therapeutics have not been reported for VEEV. More importantly methods to select therapeutic VHH against emerging viruses that have complex viral coat proteins remain challenging. In the past couple of decades, in vitro binder (antibodies, peptides, other engineered molecules) selections using phage and yeast display have resulted in high quality moieties that recognize their intended target with high specificity and affinity (9, 19–22). However, isolation of high quality binding proteins depends on the presentation of the antigen during these selections. Therefore, antigen formats that can be used in binder selection and/or evolution are key for developing novel therapeutics against emerging viruses.

Transmembrane coat protein ectodomains in viral particles are often solvent accessible or immunogenic and hence present the best antigen for in vitro or in silico binder selection and verification. However, recombinant expression and purification of these ectodomains for binder selection may be challenging for some viruses, especially if the coat protein is a heteromultimer. The ectodomain of a given protein may not be expressible or foldable in bacterial, mammalian, or cell free systems. Use of virus-like particle (VLP, pseudovirus) based protocols require extensive optimization and remain expensive. An alternative option is to use linear peptides from the viral coat protein as a selection antigen, followed by verification on infected cells or virus particles. In vitro antibody fragment selection on inactivated virus is also possible.

VEEV presents a unique opportunity to explore various antigen formats to select therapeutic VHH antibodies against a mosquito-borne pathogen with high pandemic potential. The major coat protein of VEEV is the heteromeric complex of three envelop glycoproteins E1/E2/E3 (23). The availability of various fragments of this virus allowed us to explore in vitro VHH antibody selection using multiple antigen 1) recombinant E2, 2) linear peptides of E2, selected based on molecular dynamics analysis, and 3) UV inactivated virus. Here we report on both the potentially therapeutic VHHs discovered using these targets as well as the strategies and challenges associated with each of these antigen formats.

Results

Antigen selection

The availability of both recombinantly produced viral proteins and a crystal structure of the VEEV (23) coat proteins enabled us to explore three antigen formats in our VHH selections (Figure 1a**):** (1) recombinant proteins, (2) viral peptides, and (3) inactivated viruses. The purified recombinant ectodomain of E2, purchased from Creative Diagnostics, was initially chosen as the primary antigen. This small protein domain represents the dominant epitope targeted by previously reported antibodies (24). The antigen was dissolved in 1× PBS containing 8 M urea, as per the manufacturer’s instructions, and subsequently dialyzed into PBS prior to biotinylation. The requirement for urea raised concerns about protein stability in PBS during phage and yeast sorting. Structural analysis revealed the intertwined nature of the E1/E2 complex, indicating challenges could arise if regions of E2 that are not solvent exposed in the E1/E2 complex become accessible to binding only when recombinant E2 protein is used in isolation. To present a more native coat protein conformation, we explored availability of the recombinant E1/E2 ectodomain complex. However, the yield and purity were insufficient for the phage selection pipeline (data not shown). These challenges prompted exploration of alternative antigen formats.

Previously, we demonstrated that antibodies recognizing peptides corresponding to the ectodomain of the influenza A M2 protein can bind the full-length transmembrane protein on cell and viral surfaces (25). This finding supported the feasibility of using linear peptides as antigens. Four immunogenic fragments of E2 (P1, P2, P3, and P4), were assessed for VHH-binding propensity. Peptide locations were mapped to the VEEV structure (PDB IDs: 7FFE/7FFO) as shown in Figure 1b. These peptides are all well-exposed surface fragments since their per-residue solvent accessible surface area (SASA) are all comparable, taking into account the side-chain variability. To semi-quantitatively evaluate surface clustering property (suitable for VHH anchoring) of these fragments, we implemented a dimensionless SASA / (radius of gyration, Rg^2^). Higher scores represent more congregated patches while lower scores may indicate more elongated or fragmented surfaces. In this test, P2 ranked highest (22.02), and visual inspection confirmed this analysis (Figure 1b). Therefore, P2 was prioritized for further VHH antibody selection. Peptide synthesis also offers additional controls, such as unlabeled and scrambled peptides, to enhance specificity during selection pipeline. The peptides were synthesized in good to excellent yields by microwave assisted solid phase syntheses on a CEM Liberty Prime™ using conditions outlined in the materials and methods section. The peptides were initially tested against control single chain variable fraction (scFv) antibody displayed on yeast to examine non-specific interactions against previously characterized antibodies. The norleucine (Nle) version of the P2 peptide was found to bind non-specifically, while the P2 peptide with original sequence and scramble peptide with Nle substitution showed no non-specific interaction during this initial interrogation on antigen quality (Supplementary Figure 1b).

While various forms of the VEEV virus present different opportunities for antibody selection, there are also potential challenges associated with each form of the target antigen. Recombinant proteins and peptides may differ in folding compared to E1/E2 on the viral surface. For peptides, the termini present in the peptide but not the full version of the protein may participate in the epitope of antibodies when used for selection. As a result, viral peptides may direct antibody binding to a highly specific region, or they have the potential to result in lower or no binding if this C- or N-terminal interaction is needed to mediate binding. On the other hand, the inactivated virus offers the most biologically relevant form of viral proteins, but in this antigen other proteins beyond E2 or the protein of interest are also present, and antibodies may be selected to sites other than the epitope of highest interest. As a result, it is often necessary to either perform alternating selection against different antigens or to check binding against different versions of the target to ensure binding to both the desired site and within the larger viral context. Consequently, we selected sucrose-gradient–purified, UV-inactivated virus as the final antigen format. In particular, the vaccine TC-83 strain of VEEV was used as opposed to the virulent Trinidad Donkey parental strain due to its reduced biosafety constraints and the conserved nature of the glycoproteins (26).

Selection of recombinant VHHs against VEEV from a semi-synthetic antibody library

To generate VHHs reactive to the E2 protein of VEEV, we employed a semi-synthetic VHH library (11, 17). Library panning was performed using three distinct antigen formats (Figure 1a). The phage library underwent two rounds of panning, and the phage populations obtained after the second round were evaluated by polyclonal phage ELISA, in which the phage outputs were tested against their corresponding targets (Figures 2a–c). A significant enrichment of specific polyclonal antibody phage occurred after the second round of selection with all three antigen formats. Comparatively, the population selected against the recombinant E2 protein delivered a lower signal in the polyclonal phage ELISA (Figure 2a).

The output selected solely using the full-length E2 protein exhibited a specific yet relatively weak signal against the intended target. In contrast, the peptide-based campaign yielded a robust and highly specific signal for the target peptide, with no detectable reactivity toward the control peptide composed of the same amino acids in a randomized sequence (scramble peptide). Furthermore, inactivated viral particles also demonstrated a strong and specific signal, confirming the recognition of the native antigenic context. The outputs from the second round were subcloned into a yeast display vector and transformed into yeast, following a well-established selection pipeline (8, 19). During the yeast-based enrichment for binders, we implemented a strategy involving one or two sorts on the same target used during phage selection (either protein, peptide or virus). For the selection on the viral particles, the phage output on yeast contained streptavidin binders (4% of the population) hence a negative sort (collection of population that does not bind to streptavidin) was collected (Figure 1a ss-3). Subsequently, the population was enriched for virus-binders with two rounds of sorting on biotinylated virus, followed by one more round of negative sort on streptavidin binding population. In addition, this population underwent two additional sorts on the peptide (P2) target at 250 nM and 1 µM (Figure 2f) prior to next generation sequencing (NGS) to identify most abundant clones. This approach was designed to ensure that the antibodies selected on the virus would also recognize the epitope of interest in the E2 protein. The final yeast populations (Figures 2d–f) were subjected to sequencing via NGS using Oxford Nanopore Technologies (ONT) to identify the enriched VHH clones. The nine most abundant clones identified using the protein as target, the three most abundant clones obtained using the peptide, and the seven most abundant clones obtained when the virus/peptide strategy were selected for further analysis. A total of 19 unique antibodies were gene synthetized, expressed and purified as VHH-Fc (human IgG1 Fc) fusion proteins.

All the expressed VHH-Fc fusions were assessed in ELISA for their ability to bind the peptide, the E2 protein (Figures 3a–c). ELISA analysis using soluble VHH-Fc fusions revealed that all the three clones selected using the VEEV P2 peptide bound to the same peptide, the recombinant E2 protein. In contrast, all the clones selected using solely the recombinant E2 protein were unable to bind the peptide, and only one clone showed a specific binding signal to the E2 protein in ELISA (Supplementary Figure 1a) but exhibited no binding signal to the viral particles. All the clones obtained from the selection against inactivated biotinylated virus showed binding in ELISA against the peptide, but only one showed binding when tested with the recombinant protein. The VHH sequences showed diversity in amino acid sequence, particularly in complementarity-determining region 3. The VHHs were selected from a library based on four clinical scaffolds (17). We focused particularly on further characterization of four VHHs (Figure 3d). Of these four, the VHHs selected using only peptide antigen (selection strategy 2), all had the framework region corresponding to clinical antibody sonelokimab. The VHH selected using both inactivated virus and peptide had the framework corresponding to clinical antibody caplacizumab (17). While these VHHs showed greatest diversity in their CDR3 sequences, they also differed from each other substantially in CDR1. However, all VHHs had similar CDR2 sequences.

The four identified VHH antibodies were further validated by serial-dilution ELISA using both the P2 peptide and the recombinant E2 protein as targets. Serial dilutions of each antibody, ranging from 300 nM to 0.1 nM, were used to estimate the EC50 values of the antibodies (Figures 3e, f). The apparent affinities ranged from 12 to 79 nM when the peptide was used as the antigen and from 11 to 85 nM when the recombinant protein served as the target (Table 1).

Virus binding validations (ELISA, neutralization, and viral sponge properties)

Once a suite of VHHs had been selected, we determined whether they bind to viral particles, indicating recognition of E2 in its native conformation. We coated an ELISA plate with 1x10^7^ PFU/mL VEEV TC-83 and then incubated with 0.25 µg VHH-Fc fusion, the commercially available VEEV-57 mAb, or an IgG isotype control (10). We identified four of the VHH-Fc fusions that bound substantially greater than the IgG one antibody obtained from the selection on inactivated viral particles recognizing the peptide in ELISA (VHH-ss3-6) and the three VHHs identified at the end of the selection on the peptide (VHH-ss2-1, VHH-ss2–2 and VHH-ss2-3) (Figure 3c**).**

Next, we performed plaque reduction neutralization assays to determine if any of the VHHs could neutralize viral particles in liquid. We incubated 20 µg of VHH-Fc fusion or VEEV-57 antibody with 2.5x10^2^ PFU/mL VEEV TC-83 for 1 h, then performed a plaque assay. Mock treated virus (inoculum) served as a negative control. As previously published, all concentrations of the mAb VEEV-57 antibody reduced plaque number and size (10), however, none of the VHH-Fc fusions were found to be neutralizing in solution (Supplementary Figure 2).

A noteworthy property of VHHs is that they may be readily tethered to other proteins or antibodies, allowing multiple antibodies to be grouped in proximity creating a “sponge” which can be used to bind virions and sequester them in a solution (27). Because four of the VHHs were found to bind to VEEV virions (Figure 3c), we wanted to determine if they could act as viral sponges and immunoprecipitate virions out of solution. To test this, we bound 20 µg of VHH-ss3-6, VHH-ss2-1, VHH-ss2–2 and VHH-ss2-3, VEEV-57, or IgG isotype control on protein G Dynabeads and then incubated them with 2.5x10^2^ PFU/mL VEEV TC-83 for 1 h. After removing the beads, the supernatant was titered by plaque assay. All the VHHs and the VEEV-57 antibody reduced infectious particles in the supernatant to varying degrees (Figure 4**).** VEEV-57 and VHH-ss2–3 demonstrated the greatest efficacy, showing 100% and 98% plaque reduction, respectively, compared to IgG. VHH-ss3–6 reduced plaque formation by 80%, while VHH-ss2–1 showed a 61% reduction, and VHH-ss2–2 showed a 46% reduction.

Discussion

Mosquito-borne diseases remain a significant global health threat, causing hundreds of millions of infections each year (28). The rising incidence of drug resistance, limited effectiveness of current interventions, and absence of vaccines for several key arboviruses highlight an urgent need for novel, broad-spectrum therapeutics that target both viral and vector-specific mechanisms. In particular, VEEV represents a significant threat to public health due to its propensity for epizootic spillover and potential use as a biothreat. This study reports the screening and isolation of anti-VEEV VHHs from a semisynthetic humanized naïve VHH library using phage and yeast display technologies. A key determinant of success in any selection campaign is the choice of target antigen. In this study, we evaluated multiple antigen formats and integrated phage and yeast display to identify VHHs targeting the E2 glycoprotein of VEEV. Our findings demonstrate that antigen presentation significantly shapes the outcome of VHH discovery campaigns and that peptide- and virion-based approaches outperformed recombinant E2 protein in yielding virus-reactive clones. Indeed, the highest performing VHH-ss2–1 has EC50 values in 10–20nM range for the P2 peptide and recombinant E2 protein, indicating its high potency. The purity and availability of well-designed control antigens have proven to be crucial quality requirements for the antigen formats used in phage and yeast display selection processes. Furthermore, our results demonstrate the potential use for non-neutralizing antibodies to be used as viral “sponges”, which could have therapeutic impacts.

Recombinant E2 ectodomain was initially selected as the antigen based on its immunological relevance; however, its requirement for solubilization in urea and subsequent refolding into PBS likely indicate compromised structural integrity. The limited enrichment of phage binders and the weak ELISA reactivity of the resulting VHH-Fc fusions support the conclusion that the recombinant antigen did not fully recapitulate the native conformation of E2. The intertwined architecture of the E1/E2 complex, as observed in published alphavirus structures, further suggests that isolated E2 may not adopt the conformations present on the virion surface. Although expression of the E1/E2 heterodimeric ectodomain was attempted, insufficient yield and purity restricted its use in selection campaigns (data not shown).

In contrast, structural mapping and surface accessibility analyses identified a short linear region within E2 (P2 peptide) that is solvent exposed and part of the known antigenic epitope for VEEV (24). Phage selections using this peptide produced strong enrichment, high specificity relative to scrambled control, and clones that recognized the peptide, recombinant E2, and inactivated virions. These results align with earlier observations (25) that antibodies raised against short extracellular peptides of viral membrane proteins can retain reactivity to the intact virus. The ability of P2-selected VHHs to bind multiple antigen formats indicates that this peptide effectively mimics a naturally accessible linear epitope on the virion surface.

Selections performed using inactivated viral particles yielded a second set of VHHs with strong virion reactivity compared to recombinant protein selections. Most virus-selected clones also recognized the P2 peptide, indicating convergence on the same immunodominant region identified structurally and through peptide-based selection. However, only one virus-derived clone bound the recombinant protein, again underscoring discrepancies between recombinant E2 and its native viral context.

Neutralizing antibodies typically block viral entry to host cells by physically blocking binding to host receptor, often facilitated by the binding angle and the Fc mediated steric hindrance (29). This blockage reduces the infection rate and lowers the viral titer. Neutralizing antibodies typically bind to highly variable epitopes such as the receptor binding domain of the SARS-CoV-2 spike protein (30). These epitopes undergo mutations as the virus tries to evade host immune response, reducing the efficacy of these antibodies as new variants arise. A strategy used in this scenario is developing broadly neutralizing antibodies that bind to conserved regions, however identification of such epitopes for emerging pathogens may take extensive time and research. Non-neutralizing (nNAbs) antibodies have previously been isolated from immunization with multiple viral species, and their effect in lowering viral load has been demonstrated (31). Known mechanisms of action for nNAbs include antibody dependent cytotoxicity (ADCC), complement-mediated cytotoxicity, steric inhibition of proteins important in viral replication, and antibody-dependent cell-mediated phagocytosis (ADCP), which are mediated by the Fc region of antibodies.

Although four VHHs demonstrated binding to immobilized virions, none neutralized the virus in plaque reduction assays. These findings suggest that the targeted epitope, while accessible and immunodominant, is not essential for entry-related functions. The robust neutralization observed with the positive-control mAb VEEV-57 confirms assay sensitivity and highlights the functional distinction between high-affinity binding and neutralizing capacity. It should be noted that the peptides used for development of these VHHs are not in proximity of mutation sites (Supplementary Figure S4), including the major virulence determinant at aa120, that differ between TC-83 and the virulent Trinidad donkey strain (32, 33). Validation of binding to virulent VEEV strains under BSL-3 conditions in future studies would further clarify the utility of the VHHs described herein.

Despite the absence of neutralization, all virus-binding VHHs effectively removed infectious virions from solution when bound to protein G Dynabeads. This “viral sponge” activity demonstrates that these VHHs can sequester virus with high efficiency, with one clone approaching the performance of mouse monoclonal VEEV-57 in reducing infectious titers. These data indicate that VHHs targeting non-neutralizing epitopes can still perform valuable functions in virion capture, concentration, and removal applications. Further work is needed to determine the therapeutic potential of VHH viral “sponges”.

An advantage of non-neutralizing antibodies and VHHs is that they can be functionalized and/or modified to increase their utility while taking advantage of the lack of viral evolution directed toward their epitopes. Popular mechanisms explored include modifications or switching the Fc domain to increase the Fc effector function or creation of multivalent antibodies to increase binding avidity (14). In addition to Fc domain, VHHs can also be functionalized using multimerizing units (e.g., multimeric proteins, self-assembling nanomaterials) to facilitate integration of favorable characteristics to further improve potential therapeutics against viral pathogens, especially emerging viruses with pandemic potential. A key advantage of VHHs is the ability to tether them with compatible materials to increase their functionality. The viral sponge property of these nanobodies could be improved by combining them as bispecific or multispecific binders. Self-assembling biopolymers is an example of suitable material for these purposes.

Anti-VEEV VHHs isolated from immunized of llamas followed by phage display screening have been characterized previously (5, 6). These single domain nanobodies when presented in bivalent format provide viral neutralization in vitro and in vivo and structural analysis of binding epitopes reveal that these VHHs bind in close proximity to the LDLRAD3 receptor binding site (34). In contrast, the VHHs described herein have framework region and CDRs derived from human antibodies. In addition, the P2 peptide used as the target antigen in our selections present an orthogonal viral epitope (Supplementary Figure 4) to the published animal derived moieties. Together, these differences suggest the potential for combining these VHHs with existing molecules to enhance therapeutic efficacy in future studies.

Overall, this work demonstrates that antigen format is a critical determinant of success in VHH discovery against structurally complex viral glycoproteins. Peptide-based and virion-based campaigns converged on a shared, surface-exposed epitope and yielded VHHs with robust virion-binding and viral capture properties. Future studies will benefit from exploring additional epitopes beyond P2, applying affinity maturation, and evaluating multivalent or Fc-extended constructs that may enhance neutralizing potential. This antigen-selection framework may be broadly applicable to other enveloped viruses with conformationally sensitive surface proteins. Furthermore, the human VHHs described in this study represent a promising medical countermeasure (MCM) platform with potential for rapid deployment and stockpiling. Specifically, the anti-VEEV VHHs developed here with the ability to act as viral sponges could serve as a valuable foundation for MCM preparedness strategies aimed at protecting deployed personnel and vulnerable populations against emerging mosquito-borne viral threats.

Materials and methods

Viruses and mammalian cells

VEEV TC-83 was obtained through BEI Resources, NIAID, NIH: Venezuelan Equine Encephalitis Virus, TC-83, NR-63. Vero cells (ATCC, CRL-81) were cultured in Dulbecco’s modified Eagle’s medium (DMEM, VWR 45000-304) supplemented with 1% (v/v) nonessential amino acids 100x solution (Gibco/Fisher Scientific, 11140050), 1% (v/v) Penicillin Streptomycin 100x Solution (Corning, 30-002-Cl), 2.5% HEPES (Gibco/Fisher Scientific, 15-630-106), and 10% heat-inactivated FBS. Cell lines were cultured at 37 °C and 5% CO2 in a humidified incubator. Regular passaging using standard subculturing techniques was used to maintain low-passage-number stocks.

Bacteria and yeast strains

OmniMAX™ 2 T1R E. coli (ThermoFisher, C854003): F´ [proAB+lacIqlacZΔM15 Tn10(TetR) Δ(ccdAB)] mcrA Δ(mrr-hsdRMS-mcrBC) φ80lacZΔM15 Δ(lacZYA-argF)U169 endA1 recA1 supE44 thi-1 gyrA96 (NalR) relA1 tonA panD was used for phage display.

S. cerevisiae EBY100 (ATCC, MYA-4941): (GAL1-AGA1::URA3 ura3–52 trp1 leu21 his3200 :HIS2 prb11.6R can1 GAL) was used for yeast display.

Antigen preparation

Protein antigen

Purified VEEV E2 glycoprotein (Creative Diagnostics, DAGA-268) was purchased and resuspended in 1xPBS with 8M Urea per manufacturer’s instruction and then dialyzed into 1xPBS prior to biotinylation using EZ-Link™ Sulfo NHS-LC-LC-Biotin (ThermoFisher, 21343) following manufacturer’s instructions.

Peptide antigens

Six different peptides were synthesized (Supplementary Figure 3). The biotin labeled peptides were used as antigens in phage and yeast selections and the unlabeled peptides were synthesized for use as competition during the selection process.

Computational modeling

SASA of selected E2 fragments were calculated using the BIO.PDB.SASA module in the BioPython package. Rg values were analyzed with the HullRad software (35) using the coordinates of the corresponding fragments extracted from the VEEV structure (PDBID:7FFO). The Ribbon diagram in Figure 1b was generated with the CCP4mg (36) package.

Automated solid phase peptide synthesis

A CEM Liberty Prime™ microwave peptide synthesizer was used for solid phase synthesis at high temperature (90 °C). All syntheses were performed at a 0.1 mmol scale using the recommended standard instrument chemistry, with modified cycles in which drain times were increased from 5 to 10 s to accommodate for any resin volume increases over successive synthesis cycles. Coupling of Fmoc-protected amino acids (5 eq, 0.5 M in DMF) was achieved by treatment with DIC (10 eq, 2 M in DMF) and Oxyma Pure (5 eq, 0.25 M in DMF). For the peptides described in this study, double-coupling instrument cycles were employed, resulting in average coupling yields of >98% per cycle. The synthesis deploys direct addition of 25% pyrrolidine in DMF to the heated resin for Fmoc deprotection at 105 °C for 60 s, a 120 s coupling at 90 °C, and three wash steps of 20 s each, resulting in a total single coupling cycle time of 4 min. To prevent hydrolysis of acid-labile side-chain protecting groups during the syntheses, 0.1 M DIPEA was added to the Oxyma solution.

C-terminal biotinylation

Peptides bearing a C-terminal biotin were synthesized on Fmoc-Lys(Biotin)-Wang resin using standard Fmoc SPPS. The preloaded biotinylated lysine served as the C-terminal residue. After resin swelling, the N-terminal Fmoc group was removed with 25% pyrrolidine in DMF, and the Fmoc-AEEAC-OH linker was installed using DIC/Oxyma activation. Then, the remaining peptide sequence was assembled by iterative cycles of Fmoc deprotection and amino acid coupling, also using DIC/Oxyma activation. Following completion of the sequence, peptides were cleaved from resin and globally deprotected under standard TFA cleavage conditions. The biotin moiety is stable in these conditions, yielding the peptides with a biotin label.

Peptide cleavage and global deprotection

Deprotection used 25 mL of modified “reagent K” mixture consisting of TIPS (1.25 mL), thioanisole (0.625 mL), DODT (1.25 mL) - a less odorous substitute for EDT (ethylendithiol), water (1.25 mL), and TFA (trifluoroacetic acid). The resin was pretreated with the quencher solution for 5 min, then the TFA was added (to a final volume of 25 mL). The deprotections were carried out in 50 mL conical tubes of high-density polypropylene, under a blanket of Argon to prevent side reactions from air for 3 h at room temperature. The solutions were filtered, and the filtrate was concentrated to 10 mL. The peptide was then precipitated into ice cold ether and collected by centrifugation.

Purification and analysis of synthetic peptides

Purifications (to > 98%) were performed on a Waters HPLC preparative work-station with a 2545 pump, at a flow rate of 20 mL/min. C18 reverse-phase columns (Waters BEH (Ethylene Bridged Hybrid) 130 Å, 10 μm, 19x150 mm) with a gradient from water to acetonitrile with 0.1% TFA added. Peaks were collected based on monitoring at 215 nm using a PDA 2998 detector. Combined product fractions were lyophilized, yielding white fluffy solids. Reaction yields were between 70 and 90%, indicating coupling yields of > 99% per amino residue incorporated. Peptides were then analyzed for purity by analytical HPLC on a C18 reverse phase column (Waters BEH 130 Å, 5 μm, 4.6x150 mm) with a linear gradient from 95% to 50% acetonitrile water with 0.1% TFA in a 10 min gradient, and by mass spectrometry on a Thermo LTQ mass spectrometer in ESI+ mode.

Inactivated whole virus as antigen

To propagate the VEEV TC-83 stocks, 5x10^6^ Vero E6 cells were infected with VEEV TC-83 at an MOI of 0.01 PFU/cell. Supernatants were collected at 3 days post-infection and clarified by centrifuging at 1000g for 10 minutes. Virus used as antigen for antibody production was inactivated by applying UV radiation at 2500 mJ/cm^2^ (FisherBrand UV crosslinker, 13-245-221). The inactivated viral stock was then pelleted though a 30% w/v sucrose cushion in TNE buffer (10 mM Tris-HCl pH 7.5, 100 mM NaCl, 1 mM EDTA) for 2 h at 30,000 rpm and resuspended in 100 µL PBS. The viral particles were biotinylated similar to the protein antigens as described above.

VHH selections

For phage selections, 10 µL of streptavidin-conjugated magnetic beads (Dynabeads M-280, ThermoFisher, 11205D) were coated with an excess of biotinylated antigen to ensure complete coverage. The selection process was performed using the automated KingFisher magnetic bead system (ThermoFisher Scientific), which included multiple washing steps to remove unbound phage. Bound phage were eluted from the beads by acid treatment and used to infect F′ pilus-carrying bacteria. After propagation of the eluted phage, the selection cycle was repeated. After two rounds of phage selection, the enriched VHHs were subcloned into the yeast display vector as previously described (19, 37, 38). The selected VHH genes were amplified using primers that introduced overlapping sequences to the yeast display vector pDNL6. The vector and amplified fragments were co-transformed into yeast cells to enable cloning by homologous recombination (8, 19). The resulting yeast mini-libraries were further enriched for binders using fluorescence-activated cell sorting following established protocols (19, 37, 38). After induction, 2 × 10^6^; yeast cells were stained with 100 nM biotinylated antigens or 10^6^ biotinylated viral particles. Cells were then labeled with streptavidin–AlexaFluor633 (ThermoFisher, S21375) to detect antigen binding and anti-SV5- phycoerythrin (PE) to assess VHH display levels. Yeast clones positive for both antigen binding (AlexaFluor633) and display (PE) were sorted. Sorted cells were grown at 30 °C for two days and induced for the next round of sorting at 20 °C for 16 h.

ONT sequencing and VHH-Fc expression

Yeast plasmid DNA was purified from the final saturated yeast cultures using a Qiagen miniprep kit (Qiagen, 27106) with the addition of acid-washed glass beads and vortexing to promote lysis. Purified DNA was eluted in 50 µL Buffer EB. To amplify the cloned VHH for NGS, 100 µL PCR reactions were prepared each with 25 µL of yeast miniprep as template, 0.2 µM forward and reverse primers, and 1 unit NEB Q5 Hot-Start Polymerase. The same forward and reverse primers were used as for Sanger sequencing. The reactions were cycled at 98 °C for 30 s; 20 cycles of 98 °C for 5 s, 65 °C for 5 s, and 72 °C for 25 s for VHH; 72 °C for 2 min. The amplified VHH (around 640 bp) were purified from a 1% agarose gel and submitted to Plasmidsaurus^®^ for sequencing using the Standard Premium PCR Service. The antibody sequences were submitted to Biointron^®^ for gene synthesis, cloning and protein expression as VHH humanIgG1 Fc-fusion, and purification.

VHH validation

ELISA

Nunc-Immuno™ MicroWell™ 96 well MaxiSorp™ (ThermoFisher, 442404) were coated overnight with neutravidin (ThermoFisher, 31050, 2.5 µg/mL). The following day, after removing the neutravidin solution, 2.5 µg/mL of biotinylated-E2, biotinylated-target peptide, biotinylated-scrambled peptide, and biotinylated- interferon gamma (IFNγ) (negative control), were added to the plate. After binding, wells were washed three times with 1x PBS and blocked with 5% bovine serum albumin (BSA) at room temperature for 1 h. The purified VHH-Fc (100 µL) were added to the wells at a concentration of 1 µg/mL in 1x PBS and incubated for 1 h. Wells were washed three times with PBS and 100 µL of anti-human IgG (1:2000 in 1x PBS 0.5% BSA, Jackson ImmunoResearch, 109–035–008) were added per well. After 1 h incubation at room temperature, wells were washed three times with 1x PBS and 50 µL of 3,3’, 5,5’ tetramethylbenzidine (TMB) substrate (Sigma, T8665-1 L) were added per well. To stop the reactions, 25 µL of 1 M sulfuric acid (H2SO4) were added per well. Absorbance was read at 450 nm using a Tecan Infinite 200 PRO plate reader.

Inactivated VEEV TC-83 was diluted to 1x10^7^ PFU/mL to coat 96-well ELISA plates (ThermoFisher Scientific, 3455) with 100 µL/well and incubated overnight at 4 °C. The plates were then blocked for 1 h at room temperature with 5% FBS in PBS. VHH-Fc fusions, VEEV-57 mAb (BioXCell, BE0435), or IgG isotype control (MilliporeSigma, M5284) were diluted to 2.5 µg/mL in 5% FBS, added at a volume of 50 µL per well, and incubated for 2 h at room temperature. Plates were washed three times with PBS and then secondary antibody, either goat anti-human (VHH; Jackson ImmunoResearch 109-035-088) or goat anti-mouse (VEEV-57 or IgG; Jackson ImmunoResearch, 115-035-003), was diluted 5000 in 5% FBS and incubated for 2 h at room temperature. The plates were then washed three times with PBS and 100 µL TMB solution was added per well and incubated 5 min, then the reaction was stopped with 1% HCl. Absorbance was then measured at OD 450nm with a Synergy HTX plate reader.

For the titration ELISA, the Nunc-Immuno™ MicroWell™ 96 well MaxiSorp™ were coated overnight with neutravidin (2.5 µg/mL) at 4 °C. The following day, after removing the neutravidin solution, 2.5 µg/mL of biotinylated-E2 and biotinylated-target peptide (P2), were added to the plate. After binding, wells were washed three times with 1x PBS and blocked with 5% bovine serum albumin (BSA) at room temperature for 1 h. Serial dilution of the purified VHH-Fc (300 nM to 0.1 nM) were added to the wells and incubated for 1 h. Wells were washed three times with PBS and 100 µL of anti-human IgG (1:2000 in 1x PBS 0.5% BSA) were added per well. After 1 h incubation at room temperature, wells were washed three times with 1x PBS and 50 µL of 3,3’, 5,5’ tetramethylbenzidine (TMB) substrate were added to each well. To stop the reactions, 25 µL of 1 M sulfuric acid (H2SO4) were added to each well. Absorbance was read at 450 nm using a spectrophotometer. Raw absorbance values were background-corrected and plotted against antibody concentration. EC50 values were calculated by fitting a four-parameter logistic (4-PL) nonlinear regression curve using GraphPad Prism software.

Plaque reduction neutralization titer assay

VHH (15 µg/mL) or VEEV-57 (ranging from 2.5 µg/mL–10 µg/mL) antibodies were incubated with 2.5 x 10^2^ PFU/mL VEEV TC-83 at 37 °C for 1 h. The antibody-virus mixture was then added to 80% confluent Vero cells and allowed to adsorb for 1 h at 37 °C. A 10-fold serial dilution of the inoculum without antibody was used to calculate inoculum titer and as a positive control. Afterwards, cells were overlaid with a solution of 1.2% cellulose (Sigma Aldrich, 435244-250G) in 1x complete DMEM medium for 4 days post infection (dpi). The monolayers were then fixed with formaldehyde solution (4% formaldehyde-1x PBS) for at least 1 h. The overlay was then removed, and the plaques were visualized via crystal violet staining.

VEEV immunoprecipitation assay

20 µg of VHH, VEEV-57, or IgG (Millipore Sigma), isotype control antibody were bound to 50 µL of Protein G Dynabeads (ThermoFisher, 10004D) in 500 µL PBS and rotated at room temperature for 2 h. 2.5x10^2^ PFU/mL VEEV TC-83 was then added to the antibody bound beads and the mixture was rotated at 37 °C for 1h. The beads were then removed using a magnet, and the supernatant was added to 80% confluent Vero cells and allowed to adsorb for 1 h at 37 °C. Afterwards, cells were overlaid with a solution of 1.2% cellulose in 1x complete DMEM medium for 4 days post-infection. The monolayers were then fixed with formaldehyde solution (4% formaldehyde-1x PBS) for at least 1 h. The overlay was then removed, and the plaques were visualized via crystal violet staining.