Rapid discovery and characterization of synthetic neutralizing antibodies against anthrax edema toxin

Abstract

Anthrax, a lethal, weaponizable disease caused by Bacillus anthracis, acts through exotoxins which are primary mediators of systemic toxicity and also targets for neutralization by passive immunotherapy. The ease of engineering Bacillus anthracis strains resistant to established therapy and the historic use of the microbe in bioterrorism present a compelling test case for platforms that permit the rapid and modular development of neutralizing agents. In vitro antigen-binding fragment (Fab) selection offers the advantages of speed, sequence level molecular control, and engineering flexibility compared to traditional monoclonal antibody pipelines. By screening an unbiased, chemically-synthetic phage Fab library and characterizing hits in cell-based assays, we identified two high-affinity neutralizing Fabs, A4 and B7, against anthrax edema factor (EF), a key mediator of anthrax pathogenesis. Engineered homodimers of these Fabs exhibited potency comparable to the best reported neutralizing monoclonal antibody against EF at preventing EF-induced cyclic AMP production. Using internalization assays in COS cells, B7 was found to block steps prior to EF internalization. This work demonstrates the efficacy of synthetic alternatives to traditional antibody therapeutics against anthrax, while also demonstrating a broadly generalizable, rapid, and modular screening pipeline for neutralizing antibody generation.

Graphical Abstract

Introduction

Anthrax, a disease caused by the Gram-positive bacteria Bacillus anthracis, can present through cutaneous, gastrointestinal, or respiratory illness depending on the route of entry for anthrax spores. Inhalational anthrax, the deadliest form of the disease, is of particular concern (1, 2). Anthrax primarily acts through a secreted tripartite toxin composed of protective antigen (PA), edema factor (EF), and lethal factor (LF) to disrupt the host immune response and cause edema and multi-system organ failure (3). These toxins are sufficient to induce characteristic symptoms of anthrax when injected without bacteria and can persist even after elimination of B. anthracis (4–7). As a result, there is a narrow window for effective antibiotic therapy, making antitoxin development an important feature of preparedness against a potential anthrax bioterrorism attack (8).

PA, EF, and LF act cooperatively to disrupt cellular function in anthrax pathogenesis (Figure 1). PA is an 83 kDa protein which binds the mammalian surface receptors capillary morphogenesis gene 2 (CMG2) and tumor endothelial marker 8 (TEM8) and is cleaved by furin to an activated 63 kDa form (7, 9). Cleavage enables PA oligomerization and the binding of EF and LF, inducing receptor mediated endocytosis of the toxin complex (7, 10). The EF-PA complex is referred to as edema toxin (ET) and the LF-PA complex lethal toxin (LT). Endosomal maturation and acidification induces a conformational change in PA, which forms a translocation pore through which EF and LF enter the cytosol or intraluminal vesicles in the late endosome (11, 12). Storage in intraluminal vesicles extends toxin action across prolonged periods and greater distances by delaying release and allowing secretion in exosomes (11).

Figure 1. Mechanism of B. anthracis PA mediated effects of EF toxicity and proposed action of characterized Fabs.

B. anthracis secretes EF and PA, which binds to cell surface receptors, is cleaved by furin, and oligomerizes. PA oligomerization enables EF association, which results in endocytosis and entry of EF into the cytoplasm. Cytoplasmic EF acts as a calmodulin-dependent adenylyl cyclase, resulting in edema and organismal death. Identified Fabs act synergistically prior to endocytosis of the toxin complex to block EF toxicity. While not shown, LF secreted by B. anthracis has a similar route of entrance to EF. LF is a metalloprotease that cleaves the N-terminal end of MAPKK to prevents the activation of MAP kinases by MAPKK.

EF belongs to a family of nucleotidyl cyclase toxins, which also include CyaA from Bordetella pertussis, ExoY from Pseudomonas aeruginosa, and ExoY-like toxins from Vibrio sp.(13). These toxins selectively convert nucleotide triphosphates to 3’,5’-cyclic nucleotide monophosphates in response to the activation by cellular factors, namely calmodulin for EF and CyaA, and actin for ExoY and ExoY-like proteins (14–16) (Figure 1). The highly-active calmodulin-dependent adenylyl cyclase activity of EF raises intracellular levels of the second messenger cyclic AMP (cAMP) (14). Through elevation of intracellular cAMP, EF exerts pleiotropic effects which benefit bacterial expansion during anthrax infection (17–20). For example, ET suppresses innate and adaptive immune functions and promotes the migration of infected macrophages (18).

LF acts as a zinc-dependent metalloprotease which cleaves mitogen-activated protein kinase kinases (MAPKKs), preventing downstream activation of the ERK, JNK, and p38 pathways (7, 21). ET and LT work together to impair innate immunity (18, 20). The accumulation of anthrax toxins and other virulence factors cause tissue damage and ultimately host death (6, 7).

Currently, the only FDA-approved anti-toxin, raxibacumab, targets and inhibits the action of PA – a natural target due to its central role in the pathogenesis of both ET and LT (22). However, the ease of engineering antibody-resistant PA (23) and the long-lived intracellular effects of LF and EF exposure (11) reveal the need for additional therapeutics which target EF and LF. Most approaches to developing neutralizing antibodies rely on an in vivo immunization step as part of library generation (24–27). While effective, this in vivo step can be slow, rate-limiting, and lacking in sequence-level engineering capabilities – all of which may be important for the rapid development of therapeutics against new or engineered outbreaks. An alternative, completely ex vivo, and sequence-controllable approach to neutralizing antibody development is the use of phage-based synthetic antibody libraries to identify candidate high-affinity neutralizing antibodies. Similar Fab-based screening approaches have led to the successful identification of neutralizing agents against the influenza envelope protein hemagglutinin (28).

In this work, we apply a previously developed, chemically synthetic Fab library using a single antibody framework (29) against anthrax EF. This library approach has a number of advantages, including the ability to develop an automatable, high-throughput, and modular screening pipeline against proteins of diverse structures and sizes (30). We show that phage display is capable of identifying Fabs that bind at nanomolar affinity to EF, and that in cyto screening readily identifies a subset of these Fabs that are neutralizing and comparable to the best reported anti-EF monoclonal antibodies (24). Together, these results offer a potentially generalizable model for the rapid generation of neutralizing antibodies against novel bacterial or viral targets.

Materials and Methods:

Anthrax Toxins:

An AviTag was introduced by ligation to the C-terminal end of EF in a pPro-EX expression plasmid containing an N-terminal polyhistidine tag. pProEx-EF-6x His-Avi was confirmed by sequencing and transformed into E. coli BL21 cells and grown at 37°C in T7 medium containing 100 μg/ml ampicillin to A₅₉₅ = 0.8. Expression was induced at 25°C with 200 μM isopropyl-1-thiogalactopyranoside with 50 μM biotin. Cells were harvested by centrifugation and frozen 18-hours post-induction. The pellet was lysed in 0.1 mg/mL lysozyme and sonicated in T₂₀β₁N₁₀₀P_0.1 buffer (20 mM Tris-HCl pH 8, 1 mM β-mercaptoethanol, 0.1 mM phenylmethylsulfonyl fluoride, 100 mM NaCl) and centrifuged for one hour at 35,000 rpm. The supernatant was loaded to a Ni²⁺-NTA column equilibrated with T₂₀β₁N₁₀₀P_0.1. The column was washed with T₂₀β₅N₁₀₀P_0.1 followed by T₂₀β₅N₁₀₀P_0.1 plus 20 mM imidazole and eluted with T₂₀β₅N₁₀₀P_0.1 with 150 mM imidazole. Peak fractions were pooled and diluted tenfold with T₂₀P_0.1, loaded onto a Source Q anion exchange column, and eluted by a 0–1M NaCl gradient. Purified EF was concentrated to approximately 20 mg/mL and frozen at −80°C. Protein quantitation was performed using extinction coefficients calculated from the known primary sequence of each protein on the Expasy ProtParam webserver. PA was purchased from List Labs.

Streptavidin Pull-Down Assay:

Streptavidin MagneSphere Paramagnetic Particles (Promega) were transferred to two microtubes on a magnetic stand and washed three times with PBS. Beads in one tube were blocked with 25 μM biotin for 15 minutes at room temperature. Both tubes were then incubated with 10 μL of 0.08 mg/mL purified biotinylated EF under the same conditions. The supernatant and the beads were collected separately and run on a 9% acrylamide (v/v) SDS-PAGE gel alongside a marker and a biotinylated EF control.

Phage Display Library Screening:

Biotinylated EF (500 nM) was immobilized using Streptavidin MagneSphere Paramagnetic Particles (Promega) and incubated with phage-display libraries, as has been previously described (31). Additional streptavidin binding sites were blocked with biotin and beads were washed and incubated with approximately 10¹² cfu phage in PBST-BSA for 15 minutes. Non-binding phages were washed away and binding phages were amplified by infecting XL1-Blue E. coli cells with magnetic beads containing immobilized EF and EF-binding Fabs displayed on the phage surface. Amplified phage were then used in the next round of selection (29). Subsequent rounds used decreasing concentrations of EF, from 500 nM to 10 nM to increase the stringency of the selection for high-affinity binders.

Synthetic Antibody Purification:

Fab purification was conducted as previously described (31). Briefly, anti-EF Fabs were cloned into a pSFV4 E. coli expression vector and transformed into BL21(DE3) cells. Cells were then grown in 2xYT medium with 100 μg/ml ampicillin at 37°C, 250 pm. At A₅₉₅ = 0.7, expression was induced at 37°C with 1 mM isopropyl-1-thiogalactopyranoside for 5 hours. The cells were harvested by centrifugation, frozen at - 80°C, and lysed the next day by lysozyme (0.1 mg/mL) and sonication in PBS with 500 mM NaCl and 0.2 mg/mL DNase I. The lysate was centrifuged at 24,000 rpm for 30 min at 4°C and applied to a Protein-G-A1 column (32). Filtrate was collected and applied to the column a second time, after which the column was washed with 45 mL of a high salt (0.5 M NaCl) PBS wash buffer. Fabs were eluted into 1.2 mL of 1M Tris-HCL, pH 7.4 using 20 mL 0.01 mM Glycine pH 2.4 and dialyzed overnight in 3L PBS. Fabs were concentrated to ~10 mg/mL and stored at −80°C. Engineering, expression and purification of the F(ab)₂ homodimers conducted as previously described (33). EF13D IgG was cloned, expressed and purified as previously described (24).

Surface Plasmon Resonance:

The affinity of Fab-EF interactions was determined using a Ni-NTA biosensor chip (GE Healthcare Life Sciences) using a Bioacore3000 instrument at 25 °C. Briefly, a solution of 500 μM NiCl₂ in eluent buffer (EB) (10 mM HEPES, 150 mM NaCl, 50 μM EDTA, 0.005% Tween20, pH 7.4) was applied to the Ni-NTA biosensor chip at 5 μL/min. 100 nM 6xHis-EF was then applied to the biosensor flowpaths 2–4 at 5 μL/min with the protein immobilization reaching ~50–100 response units. The biosensor was washed with EB at a flowrate of 30 μL/min to establish the baseline. Ligand was applied to the sensor in a series of five, twofold dilutions, ranging from 50 to 3.2 nM for 300 seconds at 30 μL/min and subsequent dissociation monitored at 50 μL/min in EB. Every third cycle was performed with EB alone as the analyte to enable subsequent double referencing. Regeneration of the chip was achieved by addition of 150 μL of regeneration solution (10 mM HEPES, 150 mM NaCl, 350 mM EDTA, 0.005% Tween20, pH 8.3) and subsequently 8 μL NaOH. Data were processed and fit with a bimolecular model using double referencing in Scrubber2 (Biologic Software Pty. Ltd., Australia).

Biolayer Interferometry:

Fab A4 was analyzed using the Octet HTX system (Fremont, CA). Biotinylated EF was diluted to 25 nM in Kinetics Buffer (PBS, 0.1% BSA, 0.02% Tween20 and 0.05% sodium azide, pH 7.4) and applied to a Streptavidin biosensor to a final response of ~1 nm at 25 °C. The sensor was then subjected to Kinetics Buffer to establish baseline followed by a series of five, threefold dilutions of Fab A4, ranging from 500–6.2 nM. The Fab-EF complex was allowed to dissociate for 250 seconds in Kinetics Buffer. The experiment was repeated in triplicate. The data were analyzed using a 1:1 biomolecular model using the Octet software with subtraction of the baseline with an analyte of Kinetics Buffer.

Fab Epitope Competition ELISA:

The Fab of interest was immobilized onto an immunoplate at 1 μg/mL in 50 mM sodium carbonate buffer (pH 9.5) and blocked in PBS, 1% (w/v) BSA. A sandwich ELISA was performed using 25 nM biotinylated EF in solution alone or in combination with a 50-fold molar excess of soluble Fab competitor for 15 minutes. After washing, the plate was incubated in streptavidin-HRP (ThermoFisher, 1:4000 dilution) for 30 minutes. The plate was washed and incubated with peroxidase-TMB substrate. Reaction progression was monitored and quenched with 1 M HCl. The resultant color was quantitated at 410 nm.

Calmodulin Competition ELISA:

A neutravidin-coated ELISA plate was coated with 50 nM biotinylated EF, washed, and blocked with free biotin. The immobilized EF was then incubated with 500 nM Fabs or PBS alone, washed, and subsequently incubated with biotinylated human CaM in the presence of 10μM CaCl₂. After washing, the plate was incubated with Neutravidin-HRP, washed, and incubated with peroxidase-TMB substrate. Reaction progression was monitored, quenched with 5% (v/v) phosphoric acid, and quantitated at 450 nM.

Cell Culture:

Y1 mouse adrenocortical cells and Chinese hamster ovary (CHO) cells were purchased from ATCC (ATCC CCL-79 and CCL-6 respectively). Cos7 cells were a generous gift from Gopal Thinakaran. Cells were maintained at 37°C and 5% CO₂ in DMEM/F-12 supplemented with 2mM L-glutamine and 10 units/mL penicillin/streptomycin. For Y1 cells, fetal bovine serum and horse serum were added to 2.5% and 12.5% (v/v), respectively. For CHO cells calf serum was added to 10% (v/v). To facilitate attachment, Y1 cells were grown on plates coated with 1% gelatin (w/v). Cos7 cells were maintained at 37°C and 5% CO2 in DMEM/F-12 High Glucose + GlutaMax (Thermo-Fisher) supplemented with 10 units/mL penicillin/streptomycin and 10% (v/v) FBS.

Y1 Adrenocortical Cell Round-Up Assay:

Y1 mouse adrenocortical cells were plated in gelatin-coated 48-well plates in 0.5 mL media at 25% confluency. The assay was conducted when the cells reached 40–50% confluency. Cells were incubated for 2 hours in 200 μL media containing 25 ng/μL EF, 100 ng/μL PA in the presence or absence of Fabs at 1 μM. Images were taken two hours after toxin exposure.

Intracellular cAMP accumulation in Chinese Hamster Ovary (CHO) cells:

CHO cells were plated in 48-well plates at 50% confluency and used at 75% confluency. Cells were incubated for 2 hours in 200 μL media containing 25 ng/μL EF, 100 ng/μL PA, and Fabs between 10 nM to 1000 nM (calculated as concentration of monomeric Fab subunits). The Direct cAMP ELISA kit and protocol (Enzo Life Sciences) was used to measure cAMP concentrations. After 2 hours, cells were lysed with 200 μL 0.1M HCl for 10 minutes at room temperature. Lysate was then centrifuged at 600xg for 5 minutes and the supernatant removed. On a 96-well plate coated with goat anti-rabbit IgG antibody, 100 μL of this supernatant was added to 100 μL of neutralization buffer, 50 μL cAMP conjugated to an alkaline phosphatase, and 50 μL of a rabbit anti-cAMP antibody. This mixture was incubated at room temperature and shaking at 500 rpm for 2 hours, followed by washing and addition of 200 μL of the substrate solution. The reaction was incubated at room temperature for another hour, after which 50 μL of stop solution was added and the optical density was read at 405 nM.

COS7 EF-Strep Internalization Assay

Cos7 cells were plated onto poly-L-lysine treated coverslips at 50% confluency and grown overnight. Cells were transferred to serum free media 45 min prior to experiments. Biotinylated EF (5 μg/mL), Streptavidin-550 (3 μg/mL) (Thermo-Fisher), and respective antibodies were first pre-bound 1 hr at room temperature in serum free media, followed by addition of PA (20 μg/mL) and Transferrin-488 (Invitrogen). Treatment cocktails were added to each coverslip and incubated at 37°C for 30 min. Cells were washed in ice cold PBS with 1 mM CaCl₂ and 1 mM MgCl₂, fixed with 4% (w/v) paraformaldehyde, mounted, and images acquired using a confocal 40X objective. Image processing and particle quantification was done with ImageJ using the Analyze Particles function.

Statistical Analyses

For the CHO intracellular cAMP assay, significance in neutralization by Fab monomers, homodimers, and combinations relative to the ET-alone treatment was determined by 2-way ANOVA adjusted for multiple comparisons. A two-sided t-test found no significant difference between the dose response curves of B7₂/A4₂ and EF13D. The IC₅₀s were determined by using nonlinear fitting of the slope of log(inhibitor) and normalized response. For the COS7 EF-Strep internalization assay, significance was determined by Student’s t-test with Bonferroni’s correction for multiple comparisons, with counts averaged across 4 biological replicates over 2 experiments.

Results:

Identification of Synthetic Antibodies against EF:

Phage display screening for synthetic binders in this pipeline requires target immobilization through biotin-streptavidin interactions. We introduced an AviTag to the C-terminal end of the EF and biotinylated the protein through expression in an E. coli BL21 strain overexpressing biotin ligase BirA in the presence of exogenously added biotin. The purified EF was effectively pulled-down using streptavidin-coated magnetic beads, confirming efficient in vivo biotinylation of EF (Figure S1).

A diverse synthetic phage antibody library was employed to uncover high affinity binders against EF. The design of the library utilized a “reduced genetic code” that emphasizes the use of a small number of amino acid types, particularly Tyr/Ser. This greatly expands the number of sites that can be randomized within the size limit of a phage display library (~10¹⁰), while keeping the fraction of the functional synthetic antibodies high (26). The libraries are built on an invariant trastuzumab 4D5 Fab scaffold, with diversity introduced into all three complementarity determining regions (CDRs) of the heavy chain CDRs, as well as CDR L3 of the light chain (31). Immobilized EF was incubated with pools of Fab-displaying phages and binding phages were amplified in E. coli and used again in subsequent rounds of selection. Increasing the stringency of selection in subsequent rounds by reducing EF concentrations enabled identification of high affinity binders. Seven anti-EF Fabs were identified and confirmed to be genetically distinct by sequencing of the phagemid DNA (Figure 2A, S2). Fabs were cloned into an E. coli expression vector and purified as cytoplasmic proteins. The affinity of the interaction with EF was determined by surface plasmon resonance (Figure S3) and revealed strong interactions with K_D values between 0.3 – 8.6 nM (Figure 2B).

Figure 2. Fabs display diversity of CDR sequences, kinetics and target epitopes in EF binding.

(A) Summary of Fab CDR sequences, numbered using the Chothia Scheme. (B) Summary of Fab binding affinity and binding parameters based on biolayer interferometry (for A4) and SPR (for remaining Fabs). (C) B7 focused competitive ELISA assay, which monitors the ability of immobilized B7 to bind EF in the presence or absence of the indicated Fab. Binding of EF to A4, A11, A12 and B1 permits simultaneous EF binding to immobilized B7 whereas binding of B7 itself to EF does not. Plot depicts mean with SD.

The ability of the Fabs to simultaneously bind distinct epitopes on EF was assessed by ELISA. The binding of Fabs A4, A11, A12 or B1 to EF did not prevent the interaction between EF and immobilized Fab B7, while competitive addition of B7 itself did. This suggests that the binding epitopes of B7 and the remaining Fabs are non-overlapping (Figure 2C). A similar assessment utilizing immobilized A4 confirmed the ability of A4 and B7 to bind EF concurrently and also suggested epitope overlap between Fabs B1 and A11 with Fab A4 (Figure S4). Since calmodulin binding is key for EF function, we also characterized the ability of identified Fabs to inhibit the EF-calmodulin interaction, revealing competition between B7 and calmodulin (Figure S5).

Characterization of the effect of Fab hits on ET-mediated intoxication in cultured cells:

Fabs were screened for their ability to block ET-induced toxicity in Y1 mouse adrenocortical cells. Elevated intracellular cAMP in Y1 cells triggers paxillin phosphorylation and release from cell-matrix adhesion complexes, yielding a rounded shape readily distinguishable from the characteristic flat morphology of epithelial cells. This change is detectable within thirty minutes of the initial cAMP spike, allowing rapid testing for EF inhibition (34, 35). As expected, Y1 cells assumed a normal, flat morphology under standard growth conditions, whereas addition of ET resulted in Y1 cell rounding within an hour (Figure 3A). Cells exposed to either EF, PA or the synthetic antibodies alone remained flat, confirming the specificity of the morphological change to the action of edema toxin (data not shown). Anti-EF Fabs were added to toxin-exposed Y1 cells at a 1μM concentration, exceeding the K_D by over 100-fold. Whereas most Fabs showed minimal reduction of ET-induced cell rounding, Fab B7 completely rescued normal cell morphology and Fab A4 displayed an intermediate effect (Figure 3A). Subsequent analyses thus focused on A4 and B7, in particular B7, which shows the strongest neutralizing effect.

Figure 3. Monomeric and homodimeric Fabs block EF toxicity in vitro.

(A) Y1 cells incubated with ET showed characteristic cAMP-induced rounding relative to the media only control. Fabs added at 1μM showed complete neutralization (B7), intermediate neutralization (A4), or minimal benefit (A9, A11, A12, B1) in this assay. (B) CHO cells were incubated with three concentrations of Fabs, F(ab)₂, or monoclonal antibodies – 2, 10, and 50 nM. F(ab)₂ concentration calculated as concentration of monomeric Fab subunits. Subsequent addition of ET and measurement of cAMP by ELISA reveals concentration-dependent neutralization of ET-induced cAMP increases with select treatments (* indicates p<0.05, ** indicates p<0.01, *** indicates p<0.001, **** indicates p<0.0001. Plot depicts mean with SD. (C) B7₂/A4₂ cocktail approximates monoclonal antibody EF13D inhibition of ET-induced cAMP toxicity CHO cells (B7₂/A4₂ and EF13D not statistically different, p=0.8). Plot depicts mean with SD.

The ability of these anti-EF Fabs to block EF toxicity in mammalian cells was then investigated by measuring the cAMP response in Chinese hamster ovary (CHO) cells. We also compared our neutralizing antibodies with EF13D, a chimpanzee-based monoclonal antibody which binds EF with high affinity (0.05–0.12 nM K_D), and has the highest reported potency to neutralize EF to date (24). While EF13D is bivalent, our identified Fabs were only monovalent. To gain the avidity effects present in bivalent antibody-antigen interactions, we generated bivalent variants of Fabs A4 and B7 by engineering a C-terminal linker which allows for spontaneous covalent linkage. The Fab homodimers, or F(ab)₂, were then isolated by cation exchange chromatography (33).

Consistent with the adenylyl cyclase activity of ET, intracellular cAMP levels spiked in ET-exposed cells compared to negative controls by direct cAMP ELISA. EF13D neutralized ET with an IC₅₀ of 1.4 nM. Fab B7 neutralized ET with an IC₅₀ of 24 nM, comparable to the effect of the Fab B7/A4 monomer cocktail (Figure 3B). F(ab)₂ B7₂ neutralized ET with a much lower IC₅₀, 9 nM. Because oligoclonal therapies using several monoclonal antibodies have been shown to improve neutralization potency, the combined effect of a B7₂/A4₂ homodimer cocktail was also explored (36, 37). Indeed, the combined B7₂/A4₂ cocktail neutralized ET with an IC₅₀ of 3 nM, approaching the potency of the monoclonal antibody EF13D (Figure 3B, C).

Fab B7 and B7₂/A4₂ homodimer cocktail act by preventing EF internalization:

To elucidate the mechanism for B7 and B7₂/A4₂ neutralization of ET, cellular uptake of EF was visualized in COS7 C. aethiops kidney cells, which have been previously shown to be sensitive to anthrax toxin (38). Receptor mediated endocytosis of EF was measured by visualization of internalized streptavidin-550 bound to biotinylated EF, with visualization of internalized transferrin-488 serving as a positive control for undisrupted endocytosis. Quantification of internalized EF-streptavidin puncta and normalization to the ET (full uptake) and EF-only (no uptake) counts measured the efficacy of cellular uptake under each treatment condition (Figure 4A–E). All treated cells showed a significant decrease in number of internalized streptavidin particles (p < 0.001 for all treatment conditions, Figure 4F), with EF13D and B7₂/A4₂ treatment resulting in the lowest number of counted particles (Figure 4D–E). Because EF-streptavidin positive particles would be visible at all steps of the toxin entry pathway following endocytosis, these results suggest EF13D and B7 both act upstream of endocytosis, possibly blocking EF-PA binding, PA-receptor binding, or engagement in receptor mediated endocytosis.

Figure 4. EF internalization is blocked by antibody treatment.

Cos7 cells were allowed to uptake EF-biotin conjugated to Strep-550 (green) and transferrin-488 (red) under treatment conditions: EF-only (A), EF+PA (B), EF+PA+B7 (C), EF+PA+B7₂/A4₂ (D), EF+PA+13D (E). Particles were quantified and normalized to EF-only and EF+PA (F). Mean ± SEM: EF-only: 0 ± 0.05, EF+PA: 1 ± 0.08, EF+PA+B7: 0.33 ± 0.08, EF+PA+A4₂/B7₂: 0.23 ± 0.04, EF+PA+13D: 0.131 ± 0.08. *** indicates p<0.001

Discussion:

Oligoclonal treatment has emerged as a promising lead in IgG-based anti-toxin efforts, capturing both the specificity of monoclonal antibodies and the potency of polyclonal sera (36, 37, 39). Here, we present the entirely in vitro identification and characterization of synthetic Fabs with engineered bivalency that demonstrate individual and synergistic neutralization of EF, a critical component of anthrax toxin pathogenesis.

Anthrax presents a test case for rapid anti-toxin therapeutic development due to the potential for widespread dissemination of weaponized spores and the ease of engineering new disease variants. Using high-throughput screening of an unbiased, chemically synthetic, phage-display library, we identified several Fabs that strongly bind EF. Fabs B7 and A4 demonstrated neutralization capability in an initial phenotypic screen, and combination therapy with homodimerized variants of these two Fabs reduced toxin-induced cAMP production in CHO cells to levels comparable to those achieved by a well-characterized, potent anti-EF monoclonal antibody, EF13D (24).

Both Fab B7 and EF13D likely act prior to endocytosis of the EF-PA complex (Figure 1). The decreased number of EF particles that are internalized after treatment suggest it is unlikely that B7 acts downstream on EF translocation. Due to the complex mechanism of EF internalization, it remains to be determined whether B7 prevents EF-PA binding, oligomeric complex assembly, or endocytosis. Mapping the step at which B7 interferes with EF could allow targeted perturbation and better characterization of EF-mediated toxicity. Interestingly, both EF13D (40) and B7 appear to compete with calmodulin for binding to EF, suggesting possible targeting of a similar functional epitope on EF. Although we did not specifically target calmodulin binding in our functional screens for EF neutralization, the fact that both B7 and EF13D compete with calmodulin suggests this epitope of EF may be uniquely neutralizable. Further screening using our platform (e.g. in vitro affinity maturation coupled with competition screening with calmodulin) could potentially identify high-affinity calmodulin targeting variants of B7 with increased potency.

More broadly, generation of synthetic antibodies may be a particularly important strategy against bacterial nucleotidyl cyclase toxins. Identification of small molecules which target EF at either the calmodulin binding or active sites with selectivity and potency has been challenging due to the high concentration of endogenous substrate and the presence of mammalian adenylyl cyclases (35, 41–43). While the repurposing of antiviral drug adefovir, an ATP analog, represents a successful case in overcoming aforementioned challenges to reduce anthrax lethality in mice (42, 44), the development of such potent inhibitors remains time-consuming and challenging. Synthetic antibodies offer a highly selective alternative to small molecules, which can be designed modularly and with high potency. Developing oligoclonal antibody therapies against secreted nucleotidyl cyclase toxins also enables identification of key epitopes to target for immunization. Mapping of key epitopes on EF targeted by oligoclonal therapy allows targeted delivery of these epitopes as a subunit vaccine against anthrax, and is generalizable to the other nucleotidyl cyclase toxins (45, 46).

This strategy for generating anti-toxin agents approximates the efficacy of monoclonal antibodies for both therapeutic and mechanistic in vitro analyses while retaining the speed and flexibility of synthetic approaches, an observation consistent with previous success with single chain variable fragment (scFv) targeting of viral coat proteins (28). Importantly, while the Fab library used here has previously generated crystallization chaperones and cryo-EM fiducials (47–49), we extend its utilization to identification of neutralizing agents. This approach of screening chemically synthetic libraries has been successfully employed previously, producing therapeutic candidates currently in clinical trials (50).

Simultaneous action of the two Fab homodimers may contribute to the heightened efficacy of the B7₂/A4₂ cocktail relative to the individual homodimers. By screening for non-overlapping epitopes through competitive interaction assays, we confirmed that B7 and A4 can bind EF in concert, allowing for a combinatorial inhibitory effect on toxicity. Notably, dimerization is necessary for the combinatorial effect, as the A4 + B7 monomer cocktail does not show improvement from B7 treatment. Dimerization has been commonly used to heighten neutralization by adding avidity to the antibody-antigen interaction, mimicking natural immunoglobulins. Thus, the increased efficacy of the B7₂/A4₂ cocktail can likely be attributed to the avidity effects of dimerization and the combinatorial efficacy of oligoclonal therapy. Because B7 and A4 bind different epitopes of EF, we speculate that the avidity effect observed in the cocktail of homodimers is due to cross-linking of different EF monomers, potentially resulting in oligomeric aggregates or disrupting the ordered interaction between EF monomers and the PA oligomer (Figure 1). Together, the effects of multivalency and synergistic action appear to strengthen neutralization of EF by the B7₂/A4₂ cocktail, possibly by increasing functional affinity to EF (36). Additional modifications, such as affinity maturation or adding Fc domain, represent opportunities to further strengthen binding and heighten the therapeutic efficacy of Fab-based treatment. This approach thus demonstrates a generalizable platform for high-throughput therapeutic discovery that captures the advantages of antibody specificity, avidity, and synergy within the flexibility of an entirely in vitro system.