Engineered Single-Domain Antibodies with High Protease Resistance and Thermal Stability

1The gastrointestinal (GI) tract is the site of numerous microbial infections caused by a range of pathogens, including: Helicobacter pylori, Salmonella Typhi, Vibrio cholerae, Escherichia coli, Campylobacter jejuni, and C. difficile. The current approach for treating most of these infections involves administration of antibiotics, which places selection pressure on the organism, can lead to antibiotic resistance, and suppresses or eliminates beneficial commensal microbes. Disease-causing pathogens of the GI tract rely on a myriad of virulence factors for colonization, adherence, motility, cellular entry, and pathogenesis. These include, but are not limited to: surface-layer proteins, adhesins, invasins, flagella, high-molecular weight toxins, and quorum sensing molecules. Inhibition of bacterial virulence factors that are essential for disease pathogenesis therefore represents a novel, non-antibiotic based strategy to treat infectious diseases, while reducing the risk of microbial resistance and maintaining commensal gut populations,,.

2Several approaches are being explored for antivirulence microbial therapy. Inhibition of E. coli pilus assembly, Bacillus anthracis lethal factor,, Type III secretion systems,, Staphylococcus aureus quorum sensing pathways, cholera toxin and C. difficile toxins A and B,, with small molecules and peptides, are examples currently under development. One of the most pursued antivirulence strategies is targeting bacterial toxins with antibodies. Neutralizing antibodies against anthrax, shiga toxin, cholera toxin, botulinum toxin and C. difficile toxins,,,, have all been successfully isolated and a number of clinical trials involving antibodies to bacterial targets are underway. For human pathogens that secrete toxins into the GI lumen before cellular entry, such as C. difficile, it may be advantageous to neutralize the toxins within the GI tract. Several studies indicate that oral administration of immunoglobulins (i.e., bovine Ig, human IgA, chicken IgY) may be successful at controlling various GI pathogens, including C. difficile,, rotavirus, shigella, and enterotoxigenic E. coli in humans and neonatal pigs. However, there are major limitations facing orally administered immunotherapeutics, including the susceptibility of antibodies to proteolytic degradation, instability at low pH, high dosing requirements and cost.

3Recombinant antibody fragments, such as single-domain antibodies (sdAbs), isolated from conventional IgGs (i.e., VHs, VLs), from the heavy-chain IgG of Camelidae species (i.e., VHHs) and from cartilagous shark IgNARs (i.e., VNARs), are ideal agents to explore for oral immunotherapy because of their small size (12 kDa–15 kDa), high affinity, high protease and thermal stability, high expression, amenability to library selection under denaturing conditions for isolating superstable species and ease of genetic manipulation. Despite possessing relatively high intrinsic protease and pH stability, a limited number of studies have shown that, when administered orally, sdAbs are readily degraded in the low pH pepsin-rich environment of the stomach and by digestive enzymes in the duodenum,,. Several engineering and selection-based approaches have been undertaken to improve the thermal stability and protease resistance of sdAbs and other recombinant antibody fragments (i.e., scFvs and Fabs). Engineered disulfide bonds,,, and other stabilizing mutations have increased the thermal stability of various recombinant fragments. Library selection of antibodies in the presence of proteases, denaturants, extreme pH, and elevated temperatures has lead to the isolation of antibody fragments with favorable characteristics such as improved thermal and chemical stability, increased protease resistance, and resistance to aggregation,,,,,,,. Random mutagenesis approaches have been used to increase the proteolytic stability of VHHs. There has been no universal strategy to increase recombinant antibody thermal and protease stability simultaneously.

4In this work, we hypothesized the addition of a non-canonical disulfide bond into the hydrophobic core of llama VHHs between framework region 2 (FR2) and FR3 would not only increase thermal stability at neutral pH, as previously reported,,, but would also impart resistance to proteolytic degradation and increase antibody stability at low pH. To test this hypothesis, we introduced two cysteine residues into a panel of VHHs which neutralize C. difficile toxin A (TcdA). Then, the mutant VHHs were compared to the wild-type VHH counterparts with respect to expression yield, tendency for aggregation, antigen binding affinity, CD structural signatures, thermal stability at neutral and acidic pH, susceptibility to GI proteases, and toxin-neutralization capacity.

5All chemicals used in this study were of analytical grade supplied by various companies. Oligonucleotides were synthesized by Operon (Huntsville, AL). The vectors pSJF2H or pMED2 (a modified version of pSJF2H containing SfiI cloning sites) were used for all VHH expression in E. coli cells (strain TG1) supplied by Stratagene (La Jolla, CA).

6The nomenclature used throughout this work to distinguish between wild-type and mutant VHHs is exemplified as follows: “A4.2” denotes a wild-type VHH, “A4.2m” denotes a mutant VHH. To construct mutant VHHs with a second disulfide bond, splice-overlap extension-polymerase chain reaction (SOE-PCR) was performed using 4 primers for each VHH (Table S1) and two rounds of PCR essentially as described. Ala or Gly and Ile codons at positions 54 and 78 (IMGT numbering system; http://imgt.cines.fr/), respectively, were changed to Cys codons through primer-forced mutation. In the first PCR, two mutagenized overlapping sub-fragments were generated for each VHH. The primer pairs used for each VHH were as follows: A4.2m (BbsI-VHH and A4.2mR-Cys, A4.2mF-Cys and BamHI-VHH); A5.1m (BbsI-VHH and A5.1mRCys, A4.2mFCys and BamHI-VHH); A19.2m (BbsI-VHH and A19.2mR-Cys, A19.2mF-Cys and BamHI-VHH); A20.1m (A20.1mSfiI-F and A20.1mR-Cys, A20.1mF-Cys and A20.1mSfiI-R); A24.1m (A20.1mSfiI-F and A24.1mR-Cys, A24.1mF-Cys and A20.1mSfiI-R); A26.8m (BbsI-VHH and A26.8mR-Cys, A26.8mF-Cys and BamHI-VHH). Each sub-fragment was gel purified and spliced with its partner fragment in a second PCR. Briefly, 160 ng of each sub-fragment were added to a 50 µL PCR mixture containing Pfu DNA polymerase, dNTPs and reaction buffer. The reaction was placed in a thermal cycler and the two fragments were spliced together using a program consisting of a preheating step at 94°C for 5 min and 10 cycles of 94°C for 30 s, 55°C for 30 s, and 72°C for 1 min. To amplify the spliced products, the reaction was heated to 94°C for 3 min, 5 pmol (0.5 µL) of each primer pair was added (BbsI-VHH and BamHI-VHH for A4.2m, A5.1m, A19.2m, and A26.8m; A20.1mSfiI-F and A20.1mSfiI-R for A20.1m and A24.1m), and 35 PCR cycles were performed exactly as described above. The resulting fragments were gel purified, digested with BbsI and BamHI (A4.2m, A5.1m, A19.2m, and A26.8) or SfiI (A20.1m and A24.1m) restriction enzymes, ligated into similarly digested expression vectors (pSJF2H or pMED2), and transformed into TG1 E. coli for VHH expression. Positive colonies were identified by colony-PCR and DNA sequencing, using the M13RP and M13FP primers (Table S1).

7Mutant VHHs were expressed in the same vector as wild-type VHHs. Expression and purification of wild-type and mutant VHHs were performed as described, followed by dialysis into phosphate-buffered saline pH 7.3 (PBS), into distilled, deionized water (ddH2O) for mass spectrometry (MS) analysis, or into 10 mM sodium phosphate buffer pH 7.3 for CD experiments.

8Proteolytic peptide fragments of mutant VHHs were created by digestion with cyanogen bromide (CNBr) and trypsin. Briefly, 100 µL reactions containing 50 µg of mutant VHH (diluted in PBS), 10 µL of 1 M HCl and 40 µL of CNBr (10 mg/mL stock prepared in 1 M HCl) were digested for 14 h at ambient temperature in the dark. The next day, 100 µL of 1 M Tris-HCl, pH 8.6, and 60 µL of trypsin (100 µg/mL stock; sequencing grade, Roche, Mississauga, ON, Canada) were added directly to the CNBr reaction mixture and incubated for 2 h at 37°C. Samples were then analyzed by non-reducing SDS-PAGE to ensure digestion prior to MS analysis. Nano-flow reversed-phase HPLC MS (nanoRPLC-ESI-MS) with data dependent analysis (DDA) was performed to confirm disulfide bond formation in the mutant VHHs. An aliquot of the CNBr/trypsin digested VHHs was re-suspended in 0.1% formic acid (aq) and analyzed by nanoRPLC-ESI-MS using a nanoAcquity UPLC system coupled to a Q-TOF Ultima™ hybrid quadrupole/TOF mass spectrometer (Waters). The peptides were first loaded onto a 180 µm I.D. ×20 mm 5 µm Symmetry®C18 trap (Waters), then eluted to a 100 µm I.D. ×10 cm 1.7 µm BEH130C18 column (Waters) using a linear gradient from 0% to 36% solvent B (acetonitrile + 0.1% formic acid) in 36 min, 36%–90% solvent B for 2 min. Solvent A was 0.1% formic acid in water. The peptide MS2 spectra were searched against mutant VHH protein sequences using the Mascot™ database searching algorithm (Matrix Science, London, UK). The MS2 spectra of the disulfide-linked peptides were deconvoluted using the MaxEnt 3 program (Waters) for de novo sequencing to determine the exact disulfide-linked positions.

9Mutant VHHs were passed over a Superdex™ 75 (GE Healthcare, Baie-d'Urfé, QC, Canada) size exclusion chromatography column as described to determine their aggregation state. Briefly, VHHs were applied at concentrations ranging from 0.75–1 mg/mL (≅45–60 µM) with a flow rate of 0.5 mL/min in a mobile phase that consisted of HBS-EP running buffer (10 mM HEPES, pH 7.4, 150 mM NaCl, 3 mM EDTA, and 0.005% (v/v) P20 surfactant). The collected fractions from the Superdex™ 75 column were then used directly for surface plasmon resonance (SPR) analysis. All kinetic rate and equilibrium constants were determined as described using a Biacore 3000 instrument (GE Healthcare) and 10,287 resonance units (RUs) of immobilized TcdA. In addition, the dissociation rate constants (k offs) of mutant VHHs before and after digestion with pepsin were compared by SPR (see below).

10Wild-type and mutant VHHs were analyzed by CD spectroscopy using a Jasco J-815 spectropolarimeter (Jasco, Easton, MD) at pH 7.3 (10 mM sodium phosphate buffer) and at pH 2.0 (10 mM sodium phosphate buffer+50 mM HCl). For all CD experiments performed at pH 2.0, proteins were equilibrated in the above buffer for a minimum of 2 h before scanning. The 50 mM Cl− concentration did contribute to a minor amount of light scatter at wavelengths less than 200 nm. For far-UV CD secondary structure scans and thermal unfolding experiments a 5 mm cuvette containing 1.5 mL of VHH at 50 µg/mL (3.2 µM; A280≅0.1) was used. VHH concentrations of up to 10 µM were initially tested, but signal intensities, expressed in molar ellipticity, were identical to that of 3.2 µM VHH concentrations and this concentration also avoided generating compromising signals from protein aggregates formed at high temperatures in thermal unfolding experiments. In these experiments, 4 accumulations were collected for each sample between 190 nm–250 nm with a 1 mm bandwidth, 20 nm/min scan speed and 0.5 nm data pitch. Raw ellipticity data, given in millidegrees (mdeg), was smoothed using the Jasco software, exported, and converted to molar ellipticity, [θ]. To convert from mdeg to molar ellipticity ([θ]) in deg cm2/dmol, Equation 1 was used,where the mean residue weight, MRW = (molecular weight of the antibody in Da/number of backbone amino acids), pathlength = cell pathlength in mm, and [VHH] = concentration of VHH in mg/mL. Thermal unfolding was followed at 215 nm with CD measurements taken every 2°C from 30°C to 96°C with a temperature increase of 1°C/min. It should be noted that 0.5°C and 1°C temperature interval measurements, on a select test VHH, gave nearly identical T m values to 2°C intervals. Molar ellipticity ([θ]) was used to calculate the fraction of protein folded (FF), which is shown in Equation 2,where [θF] and [θU] is the molar ellipticity of the folded (30°C) and unfolded (96°C) states, respectively. The thermal unfolding midpoint temperature (T m) was obtained by plotting FF against temperature (T) and fitting with a sigmoidal Boltzmann function in GraphPad Prism (GraphPad Software, La Jolla, CA). We assumed a temperature of 30°C represented a fully folded VHH (FF = 1.0) and a temperature of 96°C represented a fully unfolded VHH (FF = 0). In the case of some VHHs with a limited number of lower baseline data points, our T m values are minimum estimates. We followed unfolding at 215 nm because of a large difference in ellipticity between folded and unfolded states at this wavelength and because of very low light scattering in samples measured at neutral and acidic pH. A single T m replicate for each VHH was collected because of the very small standard error in CD-determined T m values. For example, a number of previous VHH T m replicates in our lab, using identical conditions, produced a standard error ranging from ±0.03%–0.63% with an average error of ±0.33%.

11To compare the tertiary structures of wild-type and mutant VHHs at neutral and acidic pH, near-UV CD experiments were performed in the range of 250 nm–340 nm using the conditions described above with the exception of a 10 mm cuvette containing 2 mL of protein at 250 µg/mL. In all cases, the ellipticity of buffer blanks were subtracted from experimental values and the reported data is the average of two independent experiments with 4 data accumulations in each.

12The sensitivity of wild-type and mutant VHHs to the three major GI proteases pepsin, trypsin, and chymotrypsin was explored. All reactions were performed in 20 µL volumes with 4.8 µg of VHH diluted in PBS. For pepsin digestions, reactions contained 17 µL of VHH, 2 µL of porcine stomach pepsin (460 U/mg; Sigma, Mississauga, ON, Canada), and 1 µL of 1 M HCl (final pH: 2.0). Final pepsin concentrations in each reaction ranged from 0.1 µg/mL to 100 µg/mL. Digestions were incubated at 37°C for 1 h and neutralized with 1 µL of 1 M NaOH. For trypsin and chymotrypsin digestions, reactions contained 18 µL of VHH (diluted in PBS supplemented with 10 mM CaCl2) and 2 µL of either trypsin or chymotrypsin (sequencing grade, Roche). Final trypsin/chymotrypsin concentrations ranged from 0.1 µg/mL to 100 µg/mL. Digestions were incubated at 37°C for 1 h and neutralized with 1 µL of protease inhibitor cocktail (Sigma). All neutralized VHH-protease reactions and controls (VHHs with no protease) were separated by SDS-PAGE, stained with Coomassie and photographed using an AlphaImager3400 (Alpha Innotech Corporation, San Leandro, CA). To determine the percent of VHH retained after protease digestions, densitometry analysis was performed using the AlphaEaseFc software package (Version 7.0.1, Alpha Innotech Corporation) on control and digested VHHs. A total of three independent digestion reactions were performed on all of the VHHs at each protease concentration and replicate digestions were run on separate SDS-PAGE gels. Digestions at the highest protease concentration (100 µg/mL) that were not analyzed by SDS-PAGE were buffer exchanged into ddH2O using Millipore Biomax 5K MWCO spin columns (Millipore, Billerica, MA) and subjected to MS analysis to identify the cleavage products, or analyzed by SPR for TcdA binding activity.

13In vitro TcdA neutralization assays were performed essentially as described. Human lung fibroblast cell rounding was reported 24 h post addition of TcdA (100 ng/mL), TcdA+wild-type VHH (1000 nM) or TcdA+mutant VHH (1000 nM). Specifically, VHHs were added as pooled mixtures of A4.2, A5.1, A20.1, and A26.8 (250 nM each, 1000 nM total) or A4.2m, A5.1m, A20.1m, and A26.8m (250 nM each, 1000 nM total). The percentage of cell rounding was scored visually using light microscopy and the reported values are the average of two independent experiments in which each VHH mixture was tested in triplicate.

14The SWISS-MODEL online workspace (http://swissmodel.expasy.org/workspace/) was used to construct homology models of A4.2 (wild-type) and A4.2m (mutant) VHHs. The 1qd0A (PDB) VHH was used as a template, sharing 73.5% and 71.8% homology, respectively. Images of the modeled VHHs were generated using PyMOL (www.pymol.org).

15Previously, a unique dromedary “VHH” was isolated that possessed a naturally occurring disulfide bond between Cys54 and Cys78 residues. When incorporated into several “wild-type” VHHs which possessed only the conserved Cys23/Cys104 disulfide bond, the Cys54/Cys78 disulfide bond increased VHH thermal and chemical stabilities,. To examine the stabilizing effects of an engineered disulfide bond on llama-derived VHHs, we followed this strategy and chose to introduce two cysteine residues into the hydrophobic core of six C. difficile TcdA-specific VHHs by incorporating Ala/Gly54Cys and Ile78Cys point mutations (Fig. 1A, Fig. S1), creating VHHs with two disulfide bonds. Soluble VHHs were extracted from the periplasm of TG1 E. coli and purified by immobilized-metal affinity chromatography (IMAC) with purified yields ranging from 3–12 mg/L of bacterial culture. Non-reducing SDS-PAGE and Western blot analysis of the purified products revealed the mutant VHHs were of high purity and did not form interdomain disulfide bonds (Fig. 1B). On non-reducing SDS-PAGE gels, mutant VHHs consistently ran slower than their corresponding wild-type VHHs (Fig. 1C).

16The molecular weights of all mutant VHHs were determined, but were not accurate enough to confirm the formation of the engineered disulfide bond. To precisely confirm the presence of the introduced disulfide bond, mutant VHHs were digested with CNBr and trypsin (Fig. 2A, B) and their digests subjected to MS2 analysis. The identification coverage of the mutant VHHs from the analysis of their CNBr/trypsin digests using nanoRPLC-ESI-MS with DDA was more than 30%. The disulfide-linked peptide ions appeared prominent in the survey scan of the DDA experiment when the proteins were digested with a combination of CNBr and trypsin. Peptide fragments linked by the engineered Cys54–Cys78 disulfide bond (shown in blue text in Fig. S1) were positively identified for all mutant VHHs by manual de-novo sequencing (Table 1). For example, the protein sequence coverage of A5.1m was 43% and a prominent ion at m/z 526.25 (3+) was sequenced as a disulfide-linked peptide EFVCVITR (P1) and FTCSR (P2) as shown (Fig. 2C, Fig. S1, Table 1). An almost complete disulfide-linked y fragment ion series was observed from one peptide with the other peptide attached as a modification via a disulfide bond, which remains intact under collision induced dissociation (CID).

17Analysis of mutant VHHs on a Superdex™ 75 size exclusion chromatography column produced single, monomeric peaks nearly identical to the profile for wild-type VHHs (Fig. 1D, E), confirming the mutant VHHs are non-aggregating. SPR analysis revealed the specific and high-affinity binding of 4 of 6 mutant VHHs to TcdA (Fig. 3, Table 2). These four were also the strongest TcdA neutralizers. Two mutants (A19.2m and A24.1m) exhibited non-specific binding to reference cell proteins and as a result specific interaction data could not be generated, even at antibody concentrations as high as 3.2 µM. When compared to their wild-type counterparts, the K Ds of 3 TcdA-binding mutants were reduced approximately 2–6 fold (Table 2), while the affinity of one VHH was relatively unchanged (K Ds of 24 nM and 20 nM for A4.2 and A4.2m, respectively). The K D reductions were largely a result of faster k off values and to a much lesser extent influenced by slower k on values. In general, these data suggest the Cys54–Cys78 disulfide bond may slightly distort the VHH structure leading to decreases in target binding affinities and decreases in antibody specificity.

18CD experiments were used to examine VHH secondary structure, tertiary structure, and thermal stability at both neutral and acidic pH. We first examined VHH secondary structure by far-UV CD (Fig. 4A, Fig. S2). Although the overall shape of the far-UV CD spectra from wild-type and mutant VHH pairs was similar at a given pH, spectra intensity shifts were observed for all wild-type/mutant pairs. In general, peak minima were seen at 216 nm–218 nm and at 230 nm–235 nm wavelengths but, in almost all cases, the intensity of the peak at 216 nm–218 nm was lower (decreased negative ellipticity) for mutant VHHs. Another prominent feature in the far-UV CD spectra was that mutant VHHs exhibited a near-UV shift in the peak range of 230 nm–235 nm. Wild-type VHHs possessed peak minima around 230 nm–232 nm whereas mutants displayed peak minima in this region around 232 nm–235 nm. Interestingly, A4.2/A4.2m, which of all the wild-type/mutant pairs had the most similar CD spectra at neutral pH, also had the same binding affinity for TcdA.

19We next examined VHH tertiary structures with near-UV CD spectroscopy (Fig. 4B, Fig. S3). The CD spectra in this region (250 nm–320 nm) come primarily from aromatic residues within the VHH, with Phe contributing in the range of 250 nm–270 nm, Tyr contributing in the range of 270 nm–290 nm, and Trp contributing in the range of 280 nm–300 nm. Overall, the near-UV spectra profiles were similar between wild-type and mutant VHH pairs. Spectra from wild-type and mutant pairs shared nearly identical peak wavelengths; however, between 250 nm to 295 nm, the ellipticity of mutant VHHs was consistently more negative than wild-type VHHs. There were also subtle differences in peaks occurring around 297 nm, with mutant VHHs exhibiting a minor but consistent shift to the right. Three of the four wild-type/mutant pairs (A4.2/A4.2m, A5.1/A5.1m, and A20.1m/A20.1m) produced predominantly negative ellipticity, whereas the A26.8/A26.8m pair remained positive. The contributions of the second disulfide bond cannot be ruled out as a factor which may augment the contribution of aromatic residues to ellipticity (increasing negatively) of the mutants.

20Finally, temperature-induced unfolding experiments were conducted in order to determine VHH T ms and T onsets by following changes in VHH ellipticity at 215 nm (Fig. 5, Fig. S4, Table 3, Table S2). All VHHs exhibited sigmoidal melting curves, indicative of cooperative unfolding of a protein that exists in either a folded or unfolded state. The wild-type VHHs already have high T ms (as high as 84.7°C) – significantly higher than those reported for other VHHs. At neutral pH, all mutant VHHs had significantly higher thermal unfolding midpoint temperatures (p = 0.031, unpaired two-tailed t-test) than their wild-type VHH counterparts. The T m values of mutants ranged from 78.8°C to 93.6°C, with one mutant, A5.1m, having a T m 11.6°C higher than wild-type (A5.1). The increase in mutant VHH T ms relative to wild-type ranged from 3.7°C to 11.6°C. Overall, at neutral pH, the mean T m ± SEM was 76.2°C±1.8°C and 83.6°C±2.3°C for wild-type and mutant VHHs, respectively (Fig. 5B). These findings are in agreement with previous reports that showed significant increases in the T ms of disulfide bond engineered VHHs,,. In a second series of experiments, temperature-induced unfolding was conducted at pH 2.0 by once again following VHH ellipticity changes at 215 nm (Fig. 5, Fig. S4, Table 3). At acidic pH a considerable reduction in T m was observed for both wild-type (22.1°C to 32.4°C) and mutant VHHs (23.7°C to 31.2°C) when compared to the T m values recorded at pH 7.3. However, at acidic pH the T m of all six mutants was still significantly higher than the corresponding wild-type VHHs (p = 0.002, unpaired two-tailed t-test). In acid, the increase in mutant VHH T ms relative to wild-type ranged from 2.1°C to 11.6°C, which is a nearly identical spread in temperature increases to that seen at neutral pH. Overall, at pH 2.0, the mean T m ± SEM was 49.3°C±1.2°C and 56.6°C±1.2°C for wild-type and mutant VHHs, respectively (Fig. 5B). Interestingly, the highest T m gains at both pHs were seen for the four strongest neutralizers. The T m differences between wild-type/mutant pairs are more significant at acidic pH than neutral pH. Taken together, these results (Table 3; Fig. 5) suggest the Cys54–Cys78 disulfide bond may stabilize the VHHs from acid-induced denaturation. Using our thermal unfolding curves, we also identified VHH T onset temperatures, the temperature at which 5% of the VHH was unfolded (Fig. 5C; Table S2). The T onset of mutant VHHs was significantly higher than wild-type VHHs at both neutral and acidic pH (p = 0.027 and p = 0.006, respectively, unpaired two-tailed t-test). The T onset differences between wild-type/mutant pairs are more significant at acidic pH than neutral pH. At pH 7.3, the mean T onset ± SEM was 68.9°C±1.8°C and 74.9°C±1.5°C for wild-type and mutant VHHs, respectively. At pH 2.0, the mean T onset ± SEM was 41.2°C±1.3°C and 47.3°C±1.3°C for wild-type and mutant VHHs, respectively. Therefore, the lowest T onset for the mutants was 45.0°C, whereas two of the wild-type VHHs (A5.1, A20.1) already had T onsets of ∼37°C at pH 2.0 (physiological stomach conditions).

21Proteins traveling through the GI tract encounter low pH and digestive enzymes in the stomach. We therefore asked if the Cys54–Cys78 disulfide bond improved VHH resistance to proteolytic degradation. We compared the effects of the major GI proteases pepsin, trypsin, and chymotrypsin on wild-type and mutant VHHs through SDS-PAGE and MS analysis. Initially, protease concentrations of 0.1 µg/mL, 1 µg/mL, 10 µg/mL, and 100 µg/mL were explored. When the lowest concentrations of proteases (0.1 µg/mL and 1 µg/mL) were used in digestion reactions, wild-type and mutants appeared similar to undigested controls on SDS-PAGE (data not shown). Similarly, VHHs were only moderately susceptible to protease degradation at 10 µg/mL (data not shown). In order to see clear differences in the proteolytic susceptibility of wild-type and mutant VHHs, all remaining digestions were performed at protease concentrations of 100 µg/mL. SDS-PAGE analysis of pepsin-digested wild-type and mutant VHHs showed a reduction in VHH size from ∼16 kDa (control) to either ∼14 kDa, or complete digestion to smaller fragments (Fig. 6A). The band at ∼14 kDa routinely appeared in digestions with each of the proteases. Similar to VH protease digestion studies, MS mass analysis on the ∼14 kDa products revealed cleavage at various positions within the VHH C-terminal c-Myc epitope tag. Loss of the epitope tag corresponded to reductions of 1641.7 Da, 1754.8 Da, and 1641.7 Da for pepsin, trypsin, and chymotrypsin digested VHHs, respectively (data not shown).

22Overall, significant increases in pepsin resistance were found for all mutant VHHs compared to their wild-type counterparts (p = 0.026, Mann-Whitney U test) (Fig. 6B; Fig. 7; Table 4). The increase in mutant VHH pepsin resistance relative to corresponding wild-type ranged from almost 4.5% to 63% (Table 4). For example, A5.1 was completely degraded after incubation with pepsin, while nearly 50% of A5.1m remained intact (Fig. 6A, B). The biggest increase in pepsin resistance was found for A4.2m, where an almost 63% increase in intact VHH structure was found relative to A4.2. Interestingly, A4.2m also had the highest T m and T onset at pH 2.0 (Table 3; Table S2), the same pH at which the pepsin digestions were performed. Increases in mutant VHH resistance to chymotrypsin were not as universal (Fig. 7; Fig. S5, Table 4) but, nonetheless, 4 of 6 mutant VHHs showed increased resistance to chymotrypsin, with significant increases found in clones A5.1m, A24.1m, and A26.8m (p<0.05) compared to their wild-type counterparts. No statistical differences were found between trypsin digested wild-type and mutant VHHs (Fig. 7; Fig. S5, Table 4), except for A4.2m, where trypsin resistance was actually reduced from almost 36% in the wild-type VHH to almost 5% in the mutant. Both the wild-type and mutant versions of A19.2 and A26.8 were very susceptible to trypsin degradation.

23A correlation was observed between VHH pepsin resistance and T ms at pH 2.0 (r2 = 0.735, Fig. 8A). The wild-type VHHs with lower T ms occupied the low protease resistance region of the graph, the mutants with higher T ms occupied the high protease resistance region of the graph. There was also a moderate correlation between VHH pepsin resistance and T ms at pH 7.3 (r2 = 0.500, data not shown). No correlation was evident between VHH trypsin resistance and T ms at pH 7.3 or pH 2.0 (r2 = 0.138 and r2 = 0.138, respectively) or between VHH chymotrypsin resistance and T ms at pH 7.3 or pH 2.0 (r2 = 0.012 and r2 = 0.004, respectively). In addition, a strong correlation between wild-type VHH pepsin resistance and wild-type VHH T onset at pH 2.0 was noted (r2 = 0.975, Fig. 8B, Table S2). No correlation was evident between mutant VHH pepsin resistance and mutant VHH T onset at pH 2.0 (r2 = 0.191), presumably because mutant VHH T onset temperatures were much higher than the temperature at which pepsin digestions were performed (37°C). Interestingly, we also noted a correlation between VHH trypsin resistance and the theoretical number of trypsin cleavage sites located within the whole VHH (r2 = 0.822) or located within the VHH CDR (r2 = 0.681) regions (Table S3, Fig. S6). No correlation was found between VHH pepsin or chymotrypsin resistance and the theoretical number of pepsin or chymotrypsin cleavage sites, respectively (Fig. S6).

24The ability of pepsin-treated mutants (A4.2m, A5.1m, A20.1m, and A26.8m) to bind TcdA was evaluated by SPR. SPR analyses confirmed the mutants (“VHH−tag”; see Fig. 6A) retained TcdA binding as their k off values were essentially the same as those of untreated controls (Table 2; Fig. 6C). SPR analysis on pepsin-digested wild-type VHHs could not be performed since these VHHs were significantly degraded by pepsin. These experiments highlight the profound impact a second disulfide bond in the hydrophobic core has on VHH conformational stability at low pH and resistance to proteolytic degradation by pepsin.

25Mutant VHHs retained their ability to neutralize to cytotoxic effects of TcdA on monolayers of fibroblast cells. Comparison of the neutralization capacity of pooled mixtures (1000 nM total) of wild-type and mutant VHHs revealed mutants performed nearly as well as wild-types at reducing TcdA-mediated cell rounding (Fig. 9). Given that 3 of 4 mutants showed weaker affinity for TcdA the reduction in neutralizing capacity relative to wild-type VHHs was not unexpected.

26The rapid development of bacterial resistance to most major classes of antibiotics has created a demand for novel therapeutics in the fight against infectious diseases. One of the most pursued non-antibiotic strategies involves targeting bacterial virulence factors with small molecules and antibodies. For some pathogens, inhibition of toxins and colonization factors within the GI tract may be an effective means of disease control. Oral immunotherapy for treating infectious diseases has had limited success due to the instability of immunoglobulins in the extreme pH and protease-rich environment of the GI tract. Here, through protein engineering, we increased the protease, acid and thermal stability of llama-derived sdAbs (VHHs) which target and neutralize C. difficile toxin A without dramatically affecting biological function.

27Our stabilization strategy involved the substitutions of two amino acid residues at positions 54 and 78 for cysteine, allowing for the formation of a second, non-native disulfide bond between FR2 and FR3 in the VHH hydrophobic core. Incorporation of a disulfide bond at these positions has been previously reported in camelid VHHs,, and was found to increase VHH chemical and thermal stability. We hypothesized that the additional disulfide bond may also enhance VHH resistance to proteases, especially in denaturing acidic conditions.

28To test this hypothesis, we generated the disulfide bond mutants and compared them to their wild-type counterparts containing only the native disulfide bond between residues 23 and 104. Mutant VHHs were well expressed in E. coli when targeted to the periplasmic space, although with lower yields compared to wild-type VHH counterparts, and all were non-aggregating monomers as determined by size exclusion chromatography. To confirm disulfide bond formation, we used a combination of proteolytic and chemical digestion coupled with MS2 to precisely identify VHH peptide fragments harboring the introduced disulfide bond. This approach is preferred over the Ellman's assay approach for the determination of disulfide linkage formation, as it requires less quantities of protein and reveals the positional identity of Cys pairs in a given disulfide bond. The latter information is important, as there is also the possibility that the two engineered Cys residues, besides forming the desired disulfide bond may form undesired disulfide bonds with the two conserved Cys residues at positions 23 and 104. After confirming disulfide bond formation in our mutants, SPR binding experiments revealed most mutant VHHs possessed 1- to 5-fold weaker affinity constants relative to wild-type, which is consistent with observations by others of up to 3-fold reductions in the affinities of VHHs containing the same introduced disulfide bond,. However, for the two weak neutralizing VHHs, A19.2m and A24.1m, the non-canonical disulfide linkage compromised specificity.

29We used CD spectroscopy to compare wild-type and mutant VHH secondary structure, tertiary structure and thermal stability (T m and T onset). Comparisons of VHH secondary and tertiary structure with far-UV and near-UV CD spectroscopy strongly suggested structural differences between wild-type and mutants, at both neutral and acidic pH. For all mutants, peak intensity and selective peak minima shifts were observed, although the overall spectral profiles remained very similar in all wild-type/mutant pairs. More specifically, mutants consistently showed rightward peak shifts in the peak range of 230 nm–235 nm (far-UV CD spectra) and around 297 nm (near-UV CD spectra) compared to wild-type VHHs. Such patterns may be used as signatures that could be used to quickly identify VHHs containing a properly formed non-canonical disulfide bond, as could SDS-PAGE motility values since, compared to wild-type VHHs, mutants consistently moved slower in SDS-PAGE gels. Thus, the far- and near-UV CD spectral data suggests the introduced disulfide bond changes the structure of VHHs. This is consistent with the observed perturbations in affinities and specificities and increased GI protease resistance of the mutant VHHs compared to the wild-types (see below). We used CD spectroscopy thermal denaturation experiments to show a profound and significant increase in the T ms and T onsets of mutant VHHs at both neutral and acidic pH. These mutants are more thermostable than previously reported VHs, which were affinity selected from a VH phage display library under stability pressure. The beneficial effect of the non-canonical disulfide linkage on T ms varies widely, with T m increases ranging from ≈4°C to ≈12°C. This suggests that for the mutant VHHs with a higher thermostability gain, the non-canonical disulfide linkage may have been a better fit to the overall fold. A19.2m and A24.1m showed the lowest thermostability gains and, if it is true that this is because of an unfit disulfide linkage, it would explain why they were transformed into non-specific binders upon mutation. For A4.2m on the other hand, the non-canonical disulfide linkage seems to be a natural fit, as it increased its T m the most (by almost 12°C) and significantly improved GI protease resistance (with the highest increase in pepsin resistance; see below), all without adversely affecting the K D. We also observed a correlation between pepsin resistance and T m, and this has implications in terms of using heat as the selective pressure for selecting pepsin resistant antibody fragments by in vitro evolutionary approaches.

30Most likely, mutants (exhibiting higher T ms) also have higher thermodynamic stability since thermodynamic stability generally increases with T m. This has been shown to be the case for both VH and VHH domains as well,. In the instance of VHHs, it has been shown that the introduction of the Cys54/Cys78 disulfide linkage used in our study into VHHs led to increases in both T m and thermodynamic stability. Proteins with higher T m are also less likely to unfold. These may be the reasons why our mutants were more resistant to acid-induced unfolding at 37°C, supported by the higher T onsets and pepsin resistance of our mutant VHHs (see below). Consistent with this, in a previous study, human VHs which were more resistant to acid-induced aggregation, a phenomenon encouraged/initiated by protein unfolding, had higher T ms and thermodynamic stabilities. The improved reversibility of thermal unfolding of mutant VHHs compared to their wild-type counterparts under acidic conditions in our work (data not shown) indicates that the introduced disulfide linkage may also render VHHs with aggregation resistant unfolded states, in addition to higher thermodynamic stability. Hagihara et al showed that the introduction of the same Cys54/Cys78 disulfide linkage into a VHH, in addition to increasing its T m, led to decreases in its enthalpy and entropy changes of unfolding. The enthalpy and entropy measurements indicated that the stabilization effect of the extra disulfide linkage in VHHs may be related to factors such as loop entropy, internal interactions such as hydrogen bonding and van der Waals interactions and hydration of the native and unfolded states.

31We also examined the resistance profiles of the disulfide bond mutants to the major GI proteases. Mutant VHHs were universally more resistant to pepsin and many were more resistance to chymotrypsin when compared to their wild-type counterparts. Protease sensitivity is a function of many variables including the location of proteolytic sites (e.g., loops vs protein core in antibodies), the theoretical number of proteolytic sites, and protein compactness and thermodynamic stability,. Since each wild-type and mutant VHH pair possessed the identical number of theoretical protease cleavage sites, we speculate that the second disulfide bond presents a more compact and thermodynamically stable VHH structure, preventing pepsin and chymotrypsin from accessing proteolytic cleavage sites. This view is consistent with the increased T ms in mutants (an indicator of mutants' increased thermostability), the positive correlation between pepsin resistance and T m (Fig. 8), and the lack of correlation between pepsin/chymotrypsin resistance and the number of theoretical protease cleavage sites (Fig. S6). The pepsin resistance vs T m/T onset correlation curves also point to the fact that structural compactness and thermodynamic stability plays a more prominent role in pepsin resistance, which is understandable given that pepsin requires protein unfolding for efficient digestion. This benefit is not realized for mutants against trypsin, possibly because their cleavage sites are at hydrophilic residues (Lys or Arg) which must be in more exposed regions of the VHH, possibly located in the CDR regions. Further, these regions would not be protected by stabilizing the core of the structure. The positive correlation between VHH trypsin resistance and the number of theoretical trypsin cleavage sites is a testament to this (Fig. S6). Harmsen et al have suggested the CDR regions of VHHs to be the most sensitive sites to proteolysis due to their flexibility and exposed position relative to the VHH core. Indeed, there are more predicted trypsin-cleavage sites in the CDR regions (Table S3; Fig. S6) of trypsin-sensitive VHHs (A4.2, A19.2 and A26.8) compared to trypsin-resistant VHHs (A5.1, A20.1 and A24.1). This is not the case for pepsin and chymotrypsin sensitivities (Table S3; Fig. S6).

32Importantly, we also observed an increase in T onset temperatures for mutants at the physiological conditions representative of the stomach (pH ≅ 2.0 and 37°C) to values significantly above 37°C (T onsets from 45°C–53°C). This suggests that the mutants should remain fully folded at 37°C in the stomach, hence resisting pepsin degradation (and denaturation) to a higher extent than wild-type VHHs, a statement supported by our in vitro pepsin digestion experiments. In contrast to the mutants, 3 wild-type VHHs, for example, have low T onset values of 37.8°C (A5.1 and A20.1) and 40.3°C (A26.8) which suggests they would partially unfold in the stomach (pH ≅ 2.0, 37°C), increasing their proteolytic susceptibility. This indeed is the case in an in vitro setting as A5.1 and A20.1, VHHs with T onset temperatures overlapping the physiological temperature, are completely pepsin sensitive, and A26.8 with a T onset slightly above the physiological temperature, although somewhat better than the former two, is barely resistant to pepsin (pepsin resistance: ≈3%). In the corresponding pepsin resistant mutants, acquiring resistance parallels an increase in T onset. In line with these findings, we observe a strong positive correlation between pepsin resistance and T onset (Fig. 8), and depending on the melting curve profile, T onsets may be better predictors of protein pepsin resistance than T ms.

33Compared to other studies involving in vitro VHH proteolysis, our mutant VHHs performed remarkably well, withstanding near physiological concentrations of pepsin and chymotrypsin and retaining functionality thereafter. Additionally, half of the mutants were trypsin resistant and for those which were not, identification and removal of their cleavage site(s) should be straightforward, e.g., by MS analysis and site-directed mutagenesis. Balan et al note the human stomach contains pepsin concentrations ranging from 500 µg/mL to 1 mg/mL, while Schmidt et al found the average pepsin concentration in the stomach of piglets to be 155 U/mL. Our pepsin digestion assays were performed at 100 µg/mL concentrations, which correspond to 46 U/mL. The most stable VHH mutant produced by Harmsen et al using a DNA shuffling approach showed only 21% residual VHH remaining after digestion with 100 µg/mL of pepsin. In contrast, our most stable VHH (A4.2m) showed 74% residual VHH remaining after digestion, while 4 others had residual pepsin resistance values of 24% or higher. In addition, all 4 disulfide bond mutant VHHs retained binding to TcdA after pepsin treatment, confirming their resistance to the protease and retention of functionality.

34We also examined the toxin A neutralizing efficacy of our disulfide bond mutant VHHs. Compared to the wild-type VHHs, the mutants were 3–4 fold weaker with respect to toxin A neutralization in cell-based assays, presumably a reflection in the reduced affinities of 3 of 4 VHHs for the toxin. If a more thorough analysis was performed on individual VHHs, it is possible that clone A4.2m, which showed the same affinity as A4.2 for toxin A, might be a more potent neutralizer due to its higher stability. Under stringent conditions in vivo, the lower affinity mutants may actually be more efficacious than the higher affinity wild-type VHHs due to their greater stability, as shown elsewhere. Also, a number of methods are available to increase the affinity of the disulfide-stabilized domains, allowing for the creation of superpotent toxin A neutralizing antibodies capable of withstanding a wide range of harsh conditions.

35In conclusion, we have shown that the introduction of a second disulfide bond into the hydrophobic core of a panel of llama VHHs increased thermal stability and GI protease resistance; the approach is both effective and general. The approach does not come without some drawbacks, including, reduced affinity, specificity, and expression yield. However, the mutants outperformed the wild-type VHHs under more stringent physiological conditions, which outweighs the reductions in affinity, as noted above. Whether the mutant VHHs are more efficacious than the wild-type VHHs in vivo remains to be determined. Based on our results and those of others, we suggest incorporating the non-canonical disulfide bond between position 54 and 78 at the library construction phase and not after the selection/screening phase to avoid adverse side effects on affinity and specificity seen here and in other studies. Other approaches, such as affinity maturation, could be used to overcome losses in target affinity as a result of disulfide bond incorporation. Our mutant VHHs are ideal building blocks for oral therapeutic agents that must survive the harsh GI tract, and provide promising alternatives to antibiotics. The oral administration of therapeutic proteins is of interest to the pharmaceutical and biotechnological industries,,,. Protein-based oral therapeutics have several conceived advantages over systemic administration: convenience, patience compliance, lower cost, pain-free administration, drug purity, flexibility in production source (i.e., bacterial, plant, etc.), and fewer concerns over immunogenicity. Despite the many advantages of orally administering protein therapeutics, few successes have been realized due to the destabilizing environment of the GI tract. Of the major GI proteases, pepsin is considered the primary cause of antibody degradation,, and hence a major obstacle facing orally delivered antibody therapeutics. Regarding the mutant VHHs generated in this study, the therapeutic efficacy can be further enhanced by improving their affinity (through selection of affinity maturation display libraries) and by formulation. The affinity maturation libraries could yield VHHs which are hyper-stabilized (e.g., high GI protease resistance) in addition to being of ultra-high affinity, if selection pressures (acid, proteases, heat) are applied during the panning stage,. Indeed, the correlation between VHH pepsin resistance and T m suggests that selection under heat should produce pepsin-resistant VHHs. Given their stability profile, the mutants may be resistant to serum degradation, making them efficacious systemic therapeutics if they are coupled to a half-life extending molecule. Other applications for our stabilized domains include: (i) use as delivery agents for mucosal vaccines or (ii) use as robust affinity purification reagents resistant to acidic and heat elution steps. Furthermore, the recent incorporation of these engineered disulfide bonds into human VH sdAbs not only resulted in increased thermal stability, but also markedly reduced VH aggregation, suggesting that the introduced disulfide bond imparts a universal stabilizing effect in all immunoglobulin variable domains.