Preformulation Characterization and Stability Assessments of Secretory IgA Monoclonal Antibodies as Potential Candidates for Passive Immunization by Oral Administration

Repository Citation Hu Y, Kumru OS, Xiong J, Antunez LR, Hickey J, Wang Y, Cavacini L, Klempner MS, Joshi SB, Volkin DB. (2019). Preformulation Characterization and Stability Assessments of Secretory IgA Monoclonal Antibodies as Potential Candidates for Passive Immunization by Oral Administration. Open Access Articles. https://doi.org/10.1016/j.xphs.2019.07.018. Retrieved from https://escholarship.umassmed.edu/oapubs/3922


INTRODUCTION
Diarrheal diseases are the second leading cause of death in developing countries, especially in Sub-Saharan Africa and South Asia, 1-3 with ~0.6 million children under 5 years of age dying each year due to complications caused by severe diarrhea. 4,5 A major cause of diarrhea is from drinking water contaminated by pathogenic bacteria, viruses, or parasites. 4 Enterotoxigenic Escherichia coli (ETEC) is the most common bacterial cause of diarrhea-associated mortality, which leads to approximately one quarter of all diarrheal episodes for infants and children less than 5years of age. [6][7][8][9] To further complicate these problems, enhanced antibiotic resistance has been found in many ETEC strains. [10][11][12] Thus, the development of an ETEC vaccine is considered the most effective and feasible strategy to prevent diarrheal diseases among children in developing countries, 13,14 and has become a high priority for the World Health Organization. 15 Currently, however, there are no ETEC vaccines commercially available and there are numerous scientific challenges (e.g., heterogeneity of potential target antigens, 4 poor mucosal immunogenicity responses, and potential safety issues of with antigens) as well as cost hurdles (e.g., develop, manufacture and commercialize for use in the developing world) that impede ETEC vaccine development. 7,16 Due to these challenges, there is growing interest in the use of passive immunization strategies to treat ETEC-induced diarrheal diseases in targeted populations by oral delivery of neutralizing immunoglobulins. For example, local delivery of antibodies that bind and neutralize ETEC in the GI tract could be utilized to prevent infection. Multiple virulence factors from ETEC have been recognized as potential M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT 5 antigens for passive immunity, 10,17 including secretion heat-labile enterotoxin (LT) which directly induces diarrhea by prompting solute retention and loss of water absorption in the intestinal lumen. LT is a heterohexameric A-B subunit toxin comprised of a catalytically active A-subunit and five B subunits. 17 Subunit A has ADP-ribosylation activity, which covalently modifies the subunit of the GTP-binding protein (Gs), leading to the constitutive activation of adenylate cyclase and production of 3',5'-cyclic AMP (cAMP). 18 Consequently, continuous release of chloride and water into intestinal lumen occurs causing watery diarrhea. The five B subunits mediate LT binding to glycolipid and glycoprotein receptors on host cells. 18 Thus, antibody-induced neutralization of LT enzymatic activity and inhibition of adhesion could potentially be effective in controlling ETEC infection.
Secretory IgA (sIgA) antibodies are of particular interest for passive immunization during oral administration due to their natural abundance in secretions and mucosal surfaces. 19 As the most prevalent immunoglobulin isotype in mucosal membranes, secretory IgAs (sIgAs) play crucial roles in protecting gut mucosal surfaces from pathogens and toxins. [20][21][22] Secretory IgAs function to promote clearance of pathogens, maintenance of intestinal homeostasis, direct neutralization of bacterial virulence factors (e.g., enterotoxins), and modulation of proinflammatory responses. [20][21][22][23] Therefore, sIgA mAbs are a potential therapeutic platform for passive immunization by oral administration 24 . Secretory IgA antibodies consist of dimeric IgG-like molecules, linked by a joining chain (J-chain), and complexed with a secretory component (SC) chain 25 .
The SC protein is acquired as the polymeric immunoglobulin receptor cleaves upon transport across epithelial cells into mucosal surfaces and secretions. Secretory IgA M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT 6 antibodies are inherently more resistant to proteolysis by digestive enzymes when compared to IgG in the gastrointestinal tract. 26,27 In this work, three anti-LT isotype variants (sIgA1, sIgA2 and IgG1) were expressed and purified from CHO cells in quantities of ~5-10 mg. A series of physiochemical methods were developed (to accommodate limited availability of material) and utilized for preformulation characterization of anti-LT sIgA1, sIgA2, and IgG1 mAbs including evaluating various structural attributes (i.e., primary structure, post-translational modifications, size heterogeneity/aggregation, conformational stability, relative solubility, and antibody binding), and identifying several key structural attributes of the sIgA mAbs to monitor during stability assessments. To this end, we examined the stability profile of the three anti-LT mAbs under conditions that mimic the gastric phase of oral delivery using simulated gastric fluids in a modified, scaled-down version of an in vitro gastric digestive model. These results are evaluated in terms of relative rankordering of the pharmaceutical stability of the three anti-LT mAbs from the point of view of future formulation development work to optimize both storage stability as well as stability during oral delivery.

Sample Preparation
The three anti-heat labile toxin (LT) immunoglobulins (sIgA1, sIgA2, and IgG1) were expressed in CHO cells and purified by MassBiologics (University of Massachusetts Medical School, Boston, MA). The antibodies were prepared in 10 mM containing 100 mM iodoacetamide (IAM, Life Technologies), and incubated at 100 o C for 5 min. Samples were cooled to room temperature (RT) and separated by SDS-PAGE using NuPAGE 10% Bis-tris gels (Life Technologies) and MOPS running buffer (Life Technologies) at 150 V for 75 min. Gels were stained with Coomassie Blue R-250 (Teknova, Hollister, CA) and destained with 40% methanol 10% acetic acid. Gels were digitized using an AlphaImager (Protein Simple, Santa Clara, CA) gel imaging system.

Size Exclusion Chromatography (SEC)
A Shimadzu Prominence ultra-fast liquid chromatography HPLC system equipped with a diode array detector (with absorbance detection at 214 nm) was utilized. The system was equilibrated at 0.5 mL/min flow rate in 0.2 M sodium phosphate buffer at pH 6.8 for at least 2 hours. Ten µL of each Ig (10 µg total protein) was injected and separated by a TOSOH TSKgel G4000SWXL column (8 µm particle M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT 8 size, 7.8 mm ID × 30 cm) for sIgA or a TOSOH TSK-Gel BioAssist G3SWxl column (5 µm size, 7.8 mm ID × 30 cm) for IgG1 with a corresponding guard column operated at ambient temperature (Tosoh Biosciences) using a 30-minute run time. Gel filtration molecular weight standards (Bio-Rad, Hercules, CA) were injected before and after the Ig sample sets to ensure integrity of the column and HPLC system. Potential presence of larger aggregates were determined by running Ig samples with and without the SEC column (i.e., protein percentage recovery). Greater than 95% protein recovery was obtained for each of the three mAbs by SE-HPLC, indicating minimal loss of protein (e.g., larger aggregates) by using optimized SE-HPLC conditions for sIgA vs IgG1. Data were analyzed using LC-Solution software (Shimadzu, Kyoto, Japan).

Sedimentation Velocity Analytical Ultracentrifugation (SV-AUC)
SV-AUC was performed using a Proteome Lab XL-I (Beckman Coulter) analytical ultracentrifuge equipped with a scanning ultraviolet-visible optical system. Samples were diluted to 0.2 mg/mL in PBS pH 7.2 and transferred into Beckman charcoal-epon two sector cells with a 12 mm centerpiece and sapphire windows. All experiments were performed at 20 o C after at least 1 hour of equilibration after the rotor reached 20 o C. SV-AUC was performed at a rotor speed of 40,000 RPM and with detection at 280 nm. The data were analyzed using Sedfit software (Dr. Peter Schuck, NIH). The partial specific volume was calculated using Sednterp software (Professor Thomas Laue, University of New Hampshire) based on the primary sequence. The buffer density and viscosity used in the analysis were also calculated using Sednterp based on the composition of the buffer. The density and viscosity of PBS buffer were calculated to be 1.0059 g/mL and 0.01021 Poise, respectively. A continuous c(s) distribution model was applied with a M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT 9 range from 0 to 15 svedbergs, with a resolution of 300 points per distribution and a confidence level of 0.95. Baseline, radial independent noise, and time independent noise were fit parameters, while the meniscus and bottom positions were set manually.

LC-MS Peptide Mapping
Ninety µl of 0.5 mg/mL Ig samples were reduced with 3 µl of 0. Lysozyme, BSA, ovalbumin, Apo-transferrin, fetuin, and α1-acid were used as glycoprotein standards to construct a standard curve.

N-Glycan Oligosaccharide Analysis
A GlycoWorks RapiFluor-MS N-Glycan Kit (Waters Corporation, Milford, MA) was used to identify and quantify N-linked glycans following the manufacturer instructions.
Briefly, Ig samples were centrifuged at 10,000 rpm for 5 mins, 7.5 µL of 2 mg/mL Igs were mixed with 15.3 µL ultrapure water and 6 µL Rapi-surfactant and then heated at overnight before performing fluorescence measurements as outlined above, but without the silicone oil overlay and at a fixed temperature (10 o C). Data analysis was performed as described above.

Relative Protein Solubility (Polyethylene Glycol Precipitation Assay)
Relative solubility of Igs was performed by adapting the method by Gibson et al. 32 and Toprani et al. 33 using smaller volumes. Briefly, 384-well polystyrene filter plates (Corning Life Sciences, Corning, NY) were used. Thirty percent w/v PEG 10,000 stock solutions were prepared in either PBS pH 7.2 or SGF containing 10 mM CP buffer pH 3.0. Various concentrations of PEG 10,000 solutions ranging from 0 to 25% w/v were prepared with Ig concentration of 0.2 mg/mL in both buffer conditions. Samples were incubated overnight at RT in dark. The next day, plates were centrifuged at 1,233 × g for 15 min and directly eluted into a clean 384-well plate. Relative protein concentration in each well was determined using a SpectraMax M5 plate reader (Molecular Devices) using detection at 214 nm. %PEG midpt values were then calculated as described previously. 32

In vitro Model of Gastric Digestion
To determine the stability of the proteins under simulated gastric conditions, each

Immobilized Pepsin Digestion
The Ig samples were diluted in SGF/CP buffer (see composition above) at a final concentration of 0.2 mg/mL. Immobilized pepsin-agarose (Thermo-Fisher) was washed three times in SGF by centrifugation at 12,000 x g prior to addition to the diluted Ig mAb mixture to a final pepsin concentration of 2000 U/mL. Samples were incubated at 37°C with end over end rotation to keep the beads in suspension. Beads were removed by centrifugation at 12,000 x g for 1 min upon completion of the desired incubation times, and the supernatant was removed. Samples were then neutralized by addition of 400mM NaOH prior to analysis by SE-HPLC. SE-HPLC was performed as described above, but with an injection volume of 25 µl.

Characterization of Purity, Primary Structure and Post-translational Modifications of sIgA vs. IgG mAbs
In this work, we utilized various analytical tools to perform preformulation characterization of three anti-LT mAbs (sIgA1, sIgA2 and IgG1) to identify key structural attributes of the mAbs to then subsequently monitor for various stability assessments and for future formulation development work. As shown schematically in Figure 1A samples, the dimeric sIgA and higher molecular weight (HMW) bands were shown to contain covalently cross-linked disulfide bonded species upon comparison to the reduced samples (consistent with previous reports). [39][40][41] Under reducing conditions, three major components were identified for the sIgA mAb samples: the SC (~70 kDa), heavy chain (~50 kDa), and light chain (~25 kDa). Although the J-chain (~16 kDa) was not observed by SDS-PAGE (consistent with literature results; see discussion), its presence was confirmed by LC-MS peptide mapping (see below). Specifically for sIgA2, we also observed two bands at relatively lower molecular weights (~17 kDa and ~40 kDa), which could represent a sIgA2 fragment (for the ~17 kDa band), rather than J chain, due to their disappearance after reduction. The heavy and light chains of the PNGase F-treated, reduced sIgAs migrated at slightly lower MW on the gel, indicating M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT 17 deglycosylation of these molecules. As expected, no migration differences were observed for the light chain bands independent of PNGase F treatment. In contrast, for the reduced IgG1 sample, both heavy (~50 kDa) and light (~25 kDa) chains were observed, and small amount of fragments (~17 kDa) were also seen. Since the IgG1 heavy chain is N-glycosylated (see below), it also displayed lower MW migration after PNGase F treatment.
To confirm the primary sequence and identify potential post-translational modifications (PTMs), LC-MS peptide mapping was performed with the three anti-LT mAbs. Due to the requirement for PNGase F for successful chromatographic resolution (data not shown), contributions of the N-glycans were not detected, and this PTM was examined separately (see below). Due to the sequence similarity (>97%) of sIgA1 and sIgA2, including both the variable and constant regions, many peptides were similar in terms of elution profile (Figure 2A). At the same time, some differences in peptide elution profiles were also observed thus demonstrating a unique profile for each sIgA.
For IgG1, the base peak chromatogram was significantly different when compared to the sIgAs, indicating a distinct digestion profile. Therefore, "fingerprint" chromatograms were obtained for each of the three mAbs. The sequence coverage obtained for each of the polypeptide chains is shown in Figure 2B. Overall, >85% coverage was obtained for each polypeptide chain for each mAbs. The light chain displayed the best coverage (97-100%), the heavy chain coverage was from 83% to 97%, and SC and J chain displayed 86-87% and 83-96% sequence coverage, respectively. In terms of PTMs, no notable chemical modifications on sIgA1 and sIgA2 mAbs were observed. For IgG1, N-terminal pyroglutamic acid formation and C-terminal lysine residual truncation were identified in the heavy chain, which are commonly observed PTMs with IgG1 mAbs. 42,43 Glycosylation of antibodies plays an important role in functional activity (effector function and potentially antigen binding) as well as physical properties such as solubility and stability. [44][45][46][47] A combination of total carbohydrate content as well the identification and quantification of the N-glycan oligosaccharide profile was determined for each of the three anti-LT mAbs. As shown in Figure 3A, a substantial difference in the total carbohydrate content between sIgAs mAbs (18.7% and 18.5% for sIgA1 and sIgA2, respectively), and IgG1 (1.2%) was observed. This result is consistent with the known structure and post-translational modifications of each mAb ( Figure 1A). Further analysis was performed to identify specific N-glycan type and relative quantification was performed by removal and derivatization of the N-glycans followed by chromatographic separation with detection by a combination of MS analysis and fluorescence measurements (see methods). Twenty-four and twenty-three different N-glycans for sIgA1 ( Figure 3B) and sIgA2 ( Figure 3C) were identified, respectively, with G2+NANA and G2F+NANA observed to be the most dominant glycan types. In contrast, as shown in Figure 3D, the IgG1 mAb displayed a much simpler N-glycan profile, with 5 major Nglycan oligosaccharides, in which glycan G0F was the most dominant type (>80%).
Each N-glycan type and corresponding percent composition found in each anti-LT mAb are summarized in Figure 4. It can be seen that the N-linked oligosaccharide composition and distribution greatly differs between the sIgA and IgG1 mAbs, as well as between the sIgA1 and sIgA2 mAbs.

Characterization of Size and Aggregation Profile of sIgA vs IgG mAbs
In addition to size analysis under denaturing conditions by SDS-PAGE (see Figure 1B), size distribution profiles under non-denaturing conditions were determined for each of the three anti-LT mAbs using two orthogonal methods, SV-AUC and SE-HPLC, as shown in Figure 5A and 5B, respectively. Two size categories (main peak and higher-order molecular weight (HMW) species) were used to classify the size distribution of each sample. To be consistent with literature nomenclature, the main peak of the sIgAs is referred to as dimeric sIgA while the main peak for IgG is simply referred to as IgG1. Multiple species were identified by both SV-AUC and SE-HPLC, with SV-AUC displaying superior peak resolution between the main peak and the HMW species. Overall, similar percent composition results were observed in comparing SV-AUC and SE-HPLC size distribution results after peak area integration ( Figure 5C). Both sIgA1 and sIgA2 samples displayed a combination of main peak (dimeric sIgA) as well as relatively higher amounts of HMW species in solution (~50% and ~80% of total protein peak area, respectively, for sIgA1 and sIgA2 samples). For IgG1, a more homogeneous peak distribution was observed (it should be noted that to optimize separation of different species and percent recovery, different SEC columns were employed for the sIgA vs. IgG1 samples, and thus the IgG1 eluted at an earlier retention time; see methods) with <10% of total protein peak area in the form of HMW species. and ~4M GdnHCl. Lower conformational stability was also noted at pH 3.0 for the IgG1 mAb. Finally, in terms of pH effects on size/aggregation profiles, a preliminary SV-AUC experiment (n=1) was performed to compare results at pH 7.2 (PBS buffer) to pH 3.0 (SGF without CP buffer), and no notable differences in the percent area of the major peaks (see Figure 5) were noted for either the sIgA1, sIgA2 or IgG mAbs (data not shown).
Interestingly, in terms of relative apparent solubility as measured by PEG-10,000 precipitation assay, a higher concentration of PEG-10,000 was required at pH 3.0 to precipitate each of the three anti-LT mAbs compared to pH 7.2 in the relative rank order of sIgA1 > sIgA2 > IgG1 ( Figure 6C). In fact, the sIgA1 remained soluble and failed to precipitate despite addition of the highest concentration PEG-10,000 (25%, w/v) when the solution pH was 3.0. Thus, higher relative apparent solubility was observed for each of three mAbs, albeit to various extents, by decreasing the solution pH from 7.2 to 3.0.

Examination of sIgA vs IgG Stability in an In Vitro Gastric Digestion Model to Mimic Oral Administration
To investigate and compare stability profiles of each of the mAbs under conditions that mimic oral delivery, we adapted an in vitro digestion model that focused on the gastric phase using simulated adult conditions for food digestion. 34 Figure 7 A, B). In contrast, IgG1 lost its LT binding ability to a much greater extent (shift in the midpoint values as well as decreased total signal), and much more rapidly, when compared to the sIgAs ( Figure 7C). The majority of the binding capacity of IgG1 sample was lost after 5-10 minutes of digestion. To better compare these results across the three anti-LT mAbs, the percent loss of binding signal was calculated and the relative loss rates were then compared ( Figure 7D). For IgG1, the loss of mAb binding to LT antigen was rapid compared to the slower rates observed for both of the sIgAs. The coaddition of sodium bicarbonate buffer, which neutralizes the acidic pH leading to M A N U S C R I P T

A C C E P T E D ACCEPTED MANUSCRIPT
23 irreversible inactivation of pepsin, 48,49 resulted in ~100% retention of LT binding even after overnight incubation in the in vitro gastric digestion model as shown in Figure 7D.
Second, non-reducing SDS-PAGE was performed on the same sIgA1, sIgA2 and IgG1 samples incubated in the in vitro gastric digestion model (Figure 8). The intact sIgA1 mAb (containing dimeric sIgA and HMW species as described above) was gradually digested, and a series of digestion byproducts were observed including a major species ~100 kDa, which presumably corresponds to the F(ab') 2 fragment ( Figure   7A). After 3 hours, sIgA1 degraded mostly to the ~100 kDa species. This result is consistent with known Fc susceptibility to pepsin digestion into smaller MW peptides while the more resistant F(ab') 2 fragment remains intact. 50,51 As expected, the intact sIgA1 species were essentially completely protected with co-addition of a sodium bicarbonate buffer ( Figure 8A). Overall similar observations were made with the sIgA2 mAb as shown in Figure 8B, however, one difference was noted: in addition to the major digestion species of F(ab') 2 , another protein species was detected at ~50 kDa which was likely the Fab fragment. Addition of the bicarbonate buffer played a similar role in protecting sIgA2 from digestion by increasing solution pH. In contrast, IgG1 displayed an accelerated digestion profile when compared to the sIgAs ( Figure 8C). After the first time point (5 min), almost all of the IgG1 was digested to F(ab') 2 fragments, and these remained after overnight incubation ( Figure 8C). The protective effect of bicarbonate buffer addition was also observed for IgG1. To more directly compare digestion profiles of the three mAbs by non-reduced SDS-PAGE, densitometry analysis of the native mAb band was performed ( Figure 8D). Although each of the intact mAbs were fully digested after overnight incubation (without addition of bicarbonate buffer), the rate of digestion M A N U S C R I P T

A C C E P T E D ACCEPTED MANUSCRIPT
24 of the sIgAs was much slower when compared to IgG1, indicating an increased resistance to acidic pH and pepsin digestion.
Finally, SE-HPLC was also used to determine the size degradation profile of the three anti-LT mAbs (under non-denaturing conditions) by quantifying the decrease of the intact protein species and increase of the corresponding degradation products ( Figure 9). In this experiment, immobilized pepsin was utilized to easily remove the pepsin from the solution, since pepsin co-eluted with the sIgA degradation products in the SEC chromatograms (data not shown). Three major peaks were identified in the SEC chromatograms of the sIgA samples: the intact species (containing dimeric sIgA and HMW species as described above), large fragments, and small fragments ( Figure   9A). Presumably, native protein is the intact species, while the F(ab') 2 /F(ab) are the large fragments and smaller peptide byproducts represent the small fragments. As shown in Figure 9A, digestion of intact sIgA1 was observed as a function of time where the main peak area decreased at each time point, while there was a concurrent increase in the large and small fragments. For sIgA2, similar trends were observed ( Figure 9B). For IgG1, three peaks were also observed, however, the digestion occurred more rapidly (when compared to the sIgAs) based on the reduction of the main peak area ( Figure 9C). To facilitate comparisons, the percent of intact mAb as a function of digestion time was determined. Both sIgA1 and sIgA2 demonstrated greater resistance to pepsin digestion when compared to IgG1, with no notable differences between sIgA1 and sIgA2 under these conditions ( Figure 9D).

25
To better summarize and compare the stability results described above as a function of solution pH and molecule type (sIgA1, sIgA2, IgG1), a "relative stability index" was determined ( Figure 10). Briefly, Figure 10 displays the results of the relative stability comparisons between the three anti-LT mAbs in terms of conformational stability at pH 7.2 and 3.0 (vs. temperature and vs. GdnHCl) as shown in Figure 10A, the relative apparent solubility at pH 7.2 and 3.0 as shown in Figure 10B, and finally, the stability profile during incubation in the in vitro digestion model (37°C, pH 3.5 with pepsin) as shown in Figure 10C. For each condition, three values (1, 2, and 3) were assigned to each of the three mAbs corresponding to their relative rank ordering in stability (highest, intermediate, and lowest). These values in Figure 10A and B were determined from the replotting of Figure 6 data as shown in Supplemental Figure S1.
As shown in Figure 10A (top panel), the physical properties of the three mAbs at pH 7.2 were ranked ordered as described above, and then the results were combined (bottom panel). It can be seen that sIgA1 scored as having the best physical properties (combination of results from thermal and denaturant unfolding as well as relative solubility), followed by sIgA2 with intermediate behavior, and IgG1 as the least desirable properties overall. The same evaluation was carried out at pH 3.0 as shown in Figure   10B, and the same relative rank ordering of desirable physical properties was calculated at pH 3.0 with sIgA1 > sIgA2 > IgG1. Results of relative stability index during incubation in the in vitro gastric digestion model are shown in Figure 10C based on rank ordering the results from the ELISA, SDS-PAGE and SEC analyses (see Figures 7,8,9). The including antigen binding, cell-based assays, and in some cases, effector function assays. [53][54][55] Once established, these structural attributes can be closely monitored during manufacturing, storage and transport of an IgG1 mAb to ensure product quality.

Analytical/Formulation Development Challenges for Orally Administered sIgA mAbs
In this work, we have evaluated some unique analytical and formulation development challenges with a different class of monoclonal antibodies (secretory IgAs, sIgAs) for administration by a different route (oral administration for local delivery) for a M A N U S C R I P T

A C C E P T E D ACCEPTED MANUSCRIPT
27 different application (passive immunization to protect against enteric diseases in the developing world). Specifically, sIgA1, sIgA2 and IgG1 mAbs targeting heat labile enterotoxin (LT), a major virulence factor of Enterotoxigenic E. coli (ETEC), were examined in this work. The potential therapeutic use of sIgAs for passive immunization are of particular interest since they are the predominant immunoglobulin isotype in tears, saliva, breast milk, colostrum, and mucosal surfaces such as the gastrointestinal as well as genitourinary tracts. 26 Regardless of ultimate success of using anti-LT sIgA mAbs for passive immunization against ETEC infections in vivo (preclinical animal studies ongoing), generation of milligram quantities of these anti-LT sIgA monoclonal antibodies provided the opportunity to evaluate sIgA mAbs in terms of pharmaceutical development challenges including preformulation characterization, stabilization and formulation for oral delivery.
One key challenge for performing the preformulation characterization and stability evaluations reported in this work was the limited amount of purified sIgA material available. Given the current preclinical stage of development, only ~5-10 mg of material was available for this work. To this end, we first evaluated a series of analytical tools to assess structural integrity, post-translational modifications, size/aggregation, conformational stability, relative solubility and antigen binding activity with minimal material. In addition, we aimed to perform many of these assessments under conditions of neutral pH as well as more acidic pH (using a scaled down version of an in vitro gastric digestion model; see below). The main objective was to not only better understand the key structural attributes of sIgA mAbs when formulated for oral administration, but also to compare the results to the much more widely studied IgG1 conditions that mimic oral delivery as discussed in more detail below. The monitoring of these structural attributes during process development and scale-up geared toward lowering the cost of producing sIgAs (while maintaining product quality) will be of great importance to the overall success of this passive immunization approach.

M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT
29 When considering the total amount of carbohydrate and N-linked oligosaccharide profiles, significant differences were observed between sIgA1 vs. sIgA2 (~18% total carbohydrate with 23-24 different N-glycan oligosaccharides) vs. IgG1 (~1% total carbohydrate with 5 different N-glycan oligosaccharides) expressed in CHO cells. It is expected the glycosylation pattern for sIgA mAbs will be a critical structural attribute to monitor since their heavily glycosylated nature facilitates antibody binding to various pathogens and receptors. 38 For example, the N-glycans on the J chain are usually required for dimer or oligomer formation of sIgAs, and can also bind to polymeric immunoglobulin receptors (plgR). 57 The secretory components (SC) of sIgAs are also heavily glycosylated, and the wide range of N-glycans on the SC creates diverse glycan epitopes, which can function as targets for lectins and bacterial adhesins. 38,58,59 As a result, glycosylated SC can inhibit bacteria adhesion and prevent the establishment of an infection. 59,60 In addition, the galactose-terminating N-glycans are potential ligands for the asialoglycoprotein receptor (ASGP-R) that could mediate the clearance and halflife of IgAs. 61,62 Although a relatively simpler N-glycan profile was obtained for IgG1, these glycans are required for maintaining protein stability, increasing solubility, maintaining Fc effector functions, and receptor binding (e.g., Fcγ). 38,63 In terms of future work, analysis of the total carbohydrate content by mass spectrometry methods prior to sample manipulation will be of interest to determine. In addition, identification of the Oglycosylation profile for sIgA mAbs will need to be evaluated, both for process consistency as well to better understand the role it may play in pathogen binding.
Finally, batch-to-batch variability of the glycan profiles as well as their effects on sIgA M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT 30 mAb stability (in terms of overall flexibility of the hinge region and protection of the hinge region from protease digestion 38 ) will be important topics to further study.
As for size heterogeneity of these three anti-LT mAbs, the IgG1 mAb was relatively more homogeneous containing 91-96% main peak with smaller amounts of higher molecular weight (HMW) species (4-9%) as measured by SE-HPLC and SV-AUC. In contrast, both sIgA mAbs contained lower amounts of the main peak (dimeric sIgA at 50-57% for sIgA1 and 18-22% for sIgA2), and had higher levels of HMW species (43-50% for sIgA1 and 78-82% for sIgA2). Since there are several cysteine residues in each J-chain that usually form both inter-and intra-chain disulfide bonds, it is likely that disulfide bond scrambling (leading to formation of inter-chain disulfide bonds between the tailpiece and cysteine residues in the heavy chains) can occur. 64,65 Thus, the J chain has the potential to be a hotspot for cross-linking oligomers for sIgAs. [64][65][66][67] In this work and consistent with published data, the J-chain was not detected by SDS-PAGE under reducing or non-reducing conditions 68 , although it was readily identified by LC-MS peptide mapping. One possible reported explanation is that the Jchain remains associated with light chain of sIgAs as a complex and thus co-migrated with the light chain 68,69 , although we did not observe such a complex by SDS-PAGE in terms of MW migration in this work ( Figure 1B).
It is expected that the presence and formation of HMW species observed in this work will be a key structural attribute to monitor in the future with various preparations of sIgAs. Aggregation is of concern with parenterally administered mAbs due to the loss of potency and the potential for anti-drug immune responses that limit efficacy and potentially affect safety. 70 However, it is not known to what extent this would be a M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT 31 concern during oral delivery of sIgAs. In fact, the polymeric nature of sIgA may not necessarily be a negative attribute in terms of efficacy during oral delivery for passive immunization (as long as the mAb-based drug itself is not lost due to irreversible precipitation and no unwanted immune responses are generated). The biological potency of polymeric sIgAs has been previously reported to be preserved along with some protease resistance. 71 Furthermore, polymeric sIgA may elicit intracellular signaling by binding to pIgRs, and potentially inhibit intracellular virus replication. [72][73][74] Interestingly, polymeric sIgA can display greater activity, when compared to dimeric sIgA, with regards to neutralizing toxins or whole bacterial cells, such as neutralizing proinflammatory antigens located in the apical recycling endosome. 64,75,76 In addition, it has been recently revealed that the functionality of an sIgA against influenza A viruses is notably enhanced in a specific polymeric form (tetrameric) due to significant improvement of target breadth. 77 Since it is likely that both covalent crosslinking as well as non-covalent interactions between sIgA molecules play a key role in formation of HMW species, future work will focus on better understanding the mechanism(s) of aggregate formation during production and during long-term storage. The batch-tobatch variability of the percent content and size distribution of the HMW species for various sIgAs mAbs will also be of interest to further evaluate (both at time zero and during storage) as well as determining the effect of sIgA oligomerization on biological activity.

Stability Profiles and Formulation Challenges of sIgA mAbs for Oral Delivery
The last key structural attribute identified in this work is the stability profile of sIgA1, sIgA2 and IgG1 mAbs under in vitro conditions that mimic oral delivery. We . 30 Due to limited material availability, we focused our experiments on a scaled down version of the gastric phase of digestion, since this is the first major stage that is encountered in vivo with pepsin readily degrading proteins. 78 As a preliminary formulation assessment, we also tested a bicarbonate formulation buffer, which has been successfully used as part of a rotavirus vaccination program during oral vaccination. 34 The stability profile of the three anti-LT mAbs was monitored by ELISA, SDS-PAGE and SEC. Although possessing a trend toward relatively lower binding affinity at time zero, sIgA mAbs showed greatly improved stability of antigen binding properties as measured by ELISA over 24 hr incubation in the in vitro gastric digestion conditions (37°C, pH 3.5 in the presence of pepsin). The major digestion product after pepsin digestion was the F(ab') 2 fragment for each of the three mAbs as determined by SDS-PAGE and SE-HPLC. Nevertheless, the antigen binding values varied over time between the three mAbs. These observations indicate that not all F(ab') 2 fragments retained antigen binding activity to the same extent in comparison to the full length, undigested mAbs. Potential conformational structure changes, allosteric effects, and/or glycosylation may influence these properties. 56,[79][80][81][82] In terms of future work with the scaled-down in vitro digestion model, sIgA stability profiles under conditions that mimic sequential digestion (e.g., oral, gastric, and intestinal phases) will be evaluated to better understand the stability profile of sIgA candidates under varying conditions that mimic the entire oral delivery pathway for local delivery to the GI tract. In addition, analytical techniques with better sensitivity and M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT 33 higher resolution, such as mass spectrometry, will be applied to assess the most sensitive sites for proteolytic digestion of the sIgA mAbs in the in vitro digestion model.
In addition, this scaled-down in vitro digestion model can be utilized in the future to screen for formulation excipients that may help improve stability and retain potency during oral delivery. The concept was established in this work by demonstrating the protective effect of co-addition of bicarbonate buffer in terms of stabilizing the three anti-LT mAbs during incubation in the in vitro gastric digestion model (see Figures 7, 8 and   9).
Smaller molecular weight protein therapeutic drugs (e.g., insulin) have been evaluated for systemic use by oral delivery [83][84][85] , and face several significant barriers including poor stability (due to acidic pH and digestive enzymes) and low bioavailability.
It has been reported that advanced drug delivery systems can be used to improve oral delivery of insulin, e.g. polymeric nanoparticles, micelles, liposomes, microspheres, or pH responsive complexation gels. 83,[85][86][87] In contrast, the goal of this work is local delivery of the sIgA mAbs to bind and neutralize Enterotoxigenic E. coli (ETEC) in the GI tract. Thus, passive immunization with sIgA mAbs may be a more successful approach than systematic delivery by the oral route of administration. However, for passive immunization applications in low-income and developing countries, low cost formulations of sIgA mAbs for treatment of diarrheal diseases is critical, making the use of complex formulations such as advanced delivery technologies described above less desirable.
To this end, future sIgA mAb formulation development efforts will focus on simple, low-cost liquid formulations that provide good long-term storage stability (ideally M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT 34 at room temperature), and concomitantly provide protection from acidic pH/proteases degradation during oral delivery, in a single final container. Ideally, such a low-cost liquid dosage form could be designed as an oral supplement during infant feeding.
Based on results of this work, the commonly used PBS buffer formulation of sIgAs is not sufficient to meet these goals, so improved formulations will certainly need to be identified including optimization of solution pH and addition of excipients to minimize aggregate formation during processing as well as during long-term storage. In the shorter term, to facilitate first-in-human clinical studies of orally delivered sIgA mAbs in adults, stable refrigerated/frozen preparations of sIgAs, combined with bed-side mixing with additives to protect during oral delivery (for adults), can be considered. Thus, the analytical tools utilized in this work were selected for their ability to monitor key structural attributes of sIgA mAbs from a formulation development perspective. These tools can be used to not only ensure therapeutic sIgA mAb drug candidates are reproducibly produced, but also can be formulated in a low-cost dosage form for oral administration, to pursue the long term goal of protecting infants in the developing world against certain enteric diseases by targeted passive immunization.     (C) Relative amount of main peak and HMW species calculated based on the total peak areas for both SV-AUC and SE-HPLC. Estimated molecular weight determinations were calculated as shown in Supplemental Table S1. Percent species were determined as average value; n = 2 for SV-AUC and n = 3 for SE-HPLC, with range and standard deviation values of 0.1 to 4.4% and from 0.1 to 0.6%, respectively.