Immunoglobulin A Antibody Composition Is Sculpted to Bind the Self Gut Microbiome

Abstract

Despite being the most abundantly secreted immunoglobulin isotype, the pattern of reactivity of IgA antibodies towards each individual’s own gut commensal bacteria still remains elusive. By colonizing germ-free mice with defined commensal bacteria, we found the binding specificity of bulk fecal and serum IgA towards resident gut bacteria resolves well at the species level and has modest strain level specificity. IgA hybridomas generated from lamina propria B cells of gnotobiotic mice showed that most IgA clones recognized a single bacterial species, while a small portion displayed cross-reactivity. Orally administered hybridoma-produced IgAs still retained bacterial antigen binding capability, implying the potential for a new class of therapeutic antibodies. Species-specific IgAs had a range of strain specificities. Given the distinctive bacterial species and strain composition found in each individual’s gut, our findings suggest the IgA antibody repertoire is shaped uniquely to bind our “self” gut bacteria.

One Sentence Summary:

The intestinal IgA antibody repertoire is personalized to bind resident gut bacteria.

INTRODUCTION

Immunoglobulin A (IgA) is the most abundant immunoglobulin isotype and plays an essential role in maintaining the homeostasis of mucosal linings and other physiological processes (1–5). Usually, IgA has two forms: monomeric IgA mainly exists in systemic circulation, while secretory IgA (SIgA) dominates in secretions, such as stools, saliva and colostrum. Different from monomeric IgA, SIgA is composed of two pieces of monomeric IgA linked by a joining peptide and secretory component, which is the remnant of polymeric immunoglobulin receptor (pIgR) after transcytosis. SIgA also functions as a first line of barrier against pathogens and other environmental insults (5–7). SIgA can aggregate pathogenic bacteria to limit microbial motility leading to reduced invasion into host epithelial cells in the gut (8, 9), regulate gut bacteria colonization or clearance through binding bacterial surface epitopes (10, 11), and selectively coat disease-associated bacteria in the stool of patients (12–14). The production of SIgA is dependent on gut microbiota colonization with much less IgA present in the feces of germ-free (GF) mice than specific pathogen free mice (15, 16). Furthermore, monocolonization of GF mice with different bacterial strains leads to increased production of SIgA (17–20), although it remains unclear whether the same strains also function as dominant inducers in the context of a complex microbiota.

Considering that SIgA influences gut immune homeostasis in addition to shaping the topography and functions of gut bacteria (21–23), it is of paramount importance to better understand the reactivity of SIgA towards gut commensal bacteria, including those that are either colonized in the host (“self” bacteria) and those that are not currently colonized in the host (“non-self” bacteria), such as bacteria from other species or strains colonized in other individuals (24, 25). A few studies have used non-overlapping bacteria species to investigate the specificity of gut IgA toward commensals. By employing a reversible germ-free colonization system, Hapfelmeier and colleagues (26) observed a long-lived and highly specific SIgA antibody upon colonization of GF mice with Escherichia coli (E. coli). This SIgA persisted for weeks, even after E. coli was decolonized and the mouse returned to the germ-free state. Intriguingly, when new bacteria were then introduced to these pre-colonized mice, the E. coli-binding SIgA decreased significantly. Similarly, the gut IgA sequence repertoire of mice colonized with a given microbiota is largely stable but undergoes rapid transformation upon manipulation of the gut microbiota by fecal microbiota transplantation (FMT) (27). This finding implies that the gut IgA repertoire remains malleable and in fact can undergo profound changes in response to modifications of the gut microbiota.

Studies looking at the reactivity of gut-derived monoclonal antibodies in mouse or human indicate that these antibodies can recognize specific bacterial strains or species (28–30). Some of these gut-derived monoclonal antibodies also show polyreactivity, which involves recognition of a broad spectrum of bacteria combined with the binding to structurally different but very common microbial or autologous molecules, such as lipopolysaccharide (LPS), double-stranded DNA (dsDNA) and flagellin (31). These earlier studies have largely focused on simple monocolonized mice or mice colonized by complex microbiotas with well-established bacteria-host interactions. However, the reactivity of native intestinal antibodies or gut-derived recombinant monoclonal antibodies from mice colonized by defined communities of bacteria is still poorly investigated.

To elucidate the fine specificity of IgA towards the gut microbiota, we dissected the reactivity of both native gut IgA and gut-derived IgA monoclonal antibodies in germ-free mice colonized with diverse, but well-defined bacterial communities. We found that bulk fecal and serum IgA resolved well at the species level with modest strain specificity. By screening the specificity of hybridoma-derived gut IgA clones, we determined that the majority (>75%) of these clones recognized specific bacteria. In addition, these IgA clones retained the capability of binding bacterial surface antigen in the gut upon oral administration, supporting their potential use as new therapeutics (28, 29, 32). The binding-specificities of these monoclonal IgA antibodies (IgAs) largely confirmed our observations for total native IgA with the majority of antibodies resolving at least at the species level and several antibodies being specific towards only the strain of bacteria originally colonized in the gnotobiotic mouse and binding few or no other strains of the same species.

RESULTS

Bacteria-induced bulk IgAs have clear species-level and partial strain-level resolution

To explore the specificity of fecal and serum IgA towards the gut microbiota in the context of limited antigenic diversity, we monocolonized C57BL/6 germ-free mice with one of eight different gut bacterial species with representatives from the most prominent phyla of the human gut including Firmicutes, Bacteroidetes, Actinobacteria and Proteobacteria (8M community; table S1) (33). We have previously established that monocolonization with each of these strains increases fecal IgA significantly above the low baseline observed in germ-free mice (19). After three weeks of colonization to establish steady state IgA levels (34), we collected fecal IgA and serum IgA to measure the capability of IgA induced by one bacterial species to bind itself and all seven of the other tested species. The binding capability of the bulk fecal and serum IgA towards bacterial surface antigens was detected by individually growing each bacterial species in vitro, binding each cultured species onto a detection plate, and quantifying antibody binding by ELISA. Interestingly, IgAs from both stool and serum displayed better binding towards species used to colonize mice, relative to those that did not colonize the mice (Fig. 1A and fig. S1, A and B), as shown by the prominent signal along the diagonal.

Fig. 1. Fecal IgAs bind better towards bacteria that colonized them.

(A and B) Cross-reactivity of stool and serum IgA towards each bacterial species (A) or B. ovatus strains (B). (C and D) Stool and serum IgA binding capability towards each bacterial species of a cocktail of 8 bacterial species (8M Mix) or another cocktail of 8 bacterial species, which belong to the same species as 8M Mix but from different strains (8M* Mix). Stool and serum samples were harvested from gnotobiotic mice that were colonized with indicated bacterial strains or consortia for three weeks. The average value from 5 to 7 mice was used in heat map plotting. Data in A to D were normalized by row. Detailed information of bacterial strains is listed in table S1. p values were calculated by unpaired, nonparametric Mann-Whitney test: *p < 0.05, **p < 0.01).

We observed minor cross-species reactivity of fecal bulk IgA between organisms from the same genus. For example, fecal IgA from Bacteroides caccae colonized mice binds slightly to B. ovatus. However, the binding strength compared to that of fecal IgA from B. ovatus colonized mice is much weaker (fig. S1A). Overall, these results suggest that the majority of the bacteria-induced polyclonal fecal and serum IgA is generated with sufficient specificity to resolve organisms at the species level. Notable exceptions were C. bolteae and C. aerofaciens, which induced the lowest amount of total IgA amongst the eight tested organisms (19) and did not bind well to IgA induced from any of the organisms (fig. S1A), and R. gnavus, which was highly bound by IgA, especially fecal IgA, induced by all eight strains (fig. S1A). This promiscuous fecal IgA binding capability of R. gnavus was recently identified to be due, in part at least, to the unique capability of this species to bind murine IgA antibodies that express VH5, VH6, or VH7 variable regions (35). However, such promiscuous binding to R. gnavus by serum IgA was not as clear as fecal IgA, which may be explained by the structural differences of fecal IgA (dimer) and serum IgA (monomer) or the different developmental process and location of these IgA-secreting B cells (7).

Although unrelated individuals often have substantial overlap in the species composition of their gut microbiomes, there is typically no overlap at the strain level where unique strains of the same species are defined as those differing by at least 4% of their genomic content (36, 37). To test if bulk IgA from monocolonized germ-free mice was of sufficient specificity to differentially bind different strains from the same species, we monocolonized germ-free mice with one of four different strains of B. ovatus and measured the capability of IgA induced from one strain to bind to each of the other strains of B. ovatus. Once again, we found the dominant binding along the diagonal, although with perhaps more off diagonal cross-reactivity than was observed on the eight more phylogenetically diverse bacteria tested above (Fig. 1B and fig. S1C). Overall, these results suggest the specificity of the IgA induced by colonization with a bacterial strain is in part unique to antigens of that particular strain and in part to antigens that are shared across strains from the same species. Furthermore, the abundance of these shared antigens could also vary between strains.

To explore the strain level antigen specificity in the context of more complex antigenic diversity, we colonized one set of gnotobiotic mice with all eight bacterial species (8M) from our monocolonization experiments and another set of gnotobiotic mice with a different set of strains (8M*) from the same set of bacterial species (table S1). Fecal and serum IgA from germ-free controls did not bind well to any of the 16 strains. Although, in general, there was some binding of IgAs from 8M colonized mice to the 8M* strains and vice versa, we observed 11 cases that had significantly different IgA binding between the two strains from the same species. In 10 out of 11 cases, the “self” bulk IgAs better recognized the “self” strain. That is the 8M strain of a given species were bound better by the IgAs from 8M colonized mice, while the 8M* strain of a given species was bound better by IgAs from the 8M* colonized mice (Fig. 1, C and D, and fig. S1, D and E). Altogether, these results suggest that the remarkable diversity of the gut microbiota, where every individual has a largely unique set of bacterial strains, drives a similarly unique immune response that is shaped partially by the species composition and partially by the strain composition of each individual. In short, the immune system recognizes “self” microbes better than “non-self”.

We next examined whether mice colonized from birth and for longer duration also produce fecal IgAs that bind significantly better towards the “self” than “non-self” bacteria. We monocolonized germ-free mice (~8 weeks old) with B. ovatus for three weeks, bred them and analyzed fecal IgA specificity of the pups at 6, 8 and 12 weeks old (fig. S2A). Consistent with our previous observations in ex-germ-free mice, IgAs secreted in the stool of these pups colonized at birth bind B. ovatus better than other tested species, and do so consistently across extended time scales (fig. S2B). To examine the contribution of the CD4⁺ T cells to IgA specificity, we depleted CD4⁺ T cells in our germ-free mice with anti-CD4 antibody and colonized them with B. ovatus for three weeks (fig. S3A). Fecal IgA binding of B. ovatus and R. gnavus significantly decreased in CD4⁺ T cell depleted mice, while no significant change was observed in the amount of IgA antibodies that bind towards other tested species (fig. S3B). The above results indicate that the extended length of bacteria colonization has little influence on shaping gut IgA specificity (38) and CD4⁺ T cells play a role in the production of B. ovatus-specific instead of cross-reactive IgA antibodies (19).

Screening the distribution of lamina propria IgA specificity towards resident bacteria

The above specificities of serum and fecal IgA toward each bacterial strain represent the collective binding of many unique antibodies of different abundance. For more discrete understanding of the species-specific and strain-specific binding capability of IgA secreted by single IgA⁺ B cells induced by gut bacteria, we made hybridoma clones using the small intestinal and colonic lamina propria (LP) B cells of mice colonized for 3 weeks with the 8M consortia of bacteria and about 30% clones were IgA isotype (fig. S4A). In the 29 IgA hybridoma clones (2 out of 31 were excluded because of slow growth of cells), the majority of them (93%) expressed the preferred kappa light chain (fig. S4B). Like in the bulk IgA reactivity screening, an ELISA was performed to determine the specificity of each antibody towards each of the 8 different strains (at stationary phase) that were co-colonized in the gnotobiotic mice. Among the 29 hybridoma clones, we found that four bacterial species are specifically targeted by the majority of these IgA clones and 8 clones (27.6%) did not bind bacterial surface antigens from any of the 8M community members. Among the 21 bacteria-binding clones, 9 clones (42.9%) bound to only a single species, 10 clones (47.6%) bound to two species, and 2 clones (9.5%) bound to all the tested bacterial species (Fig. 2A and fig. S4C). We then quantified the relative abundance of each bacterial species in our gnotobiotic mice (fig. S4D) and wondered whether species abundance in the gut determined the frequency of isolating IgA clones towards a particular species. However, we did not find a significant correlation between the relative abundance of species and the percentage of IgA clones binding a particular species (R² = 0.21; p = 0.25) (fig. S4E).

Fig. 2. Majority of monoclonal IgA antibodies produced by gut lamina propria IgA⁺ B cells bind towards bacterial antigens.

(A) Binding capability of 29 monoclonal IgA antibodies towards bacteria that colonized the gut of gnotobiotic mice. Germ-free mice were colonized with all 8 bacterial species for three weeks, then immune cells isolated from intestinal lamina propria were used for hybridoma fusion. A OD₄₅₀ > 0.25 at antibody concentration of 10 μg/ml is regarded as positive signal (dark gray square). (B) Representative results of hybridoma-produced IgA antibodies binding towards B. ovatus (I), B. caccae (II), R. gnavus (III), both B. ovatus and R. gnavus (IV) and all 8 bacterial species (V). The relative binding capability of IgA clones was analyzed by ELISA and the average value of two technical replicates is displayed. (C) Representative flow cytometry plots of hybridoma-produced IgA antibodies (clones B11, C2 and F2) binding towards all 8 bacterial species. Detailed information of bacterial strains is listed in table S1. OD₄₅₀: optical density at 450 nm.

Similar to findings published previously showing the promiscuous binding capability of R. gnavus by murine IgA antibodies with VH5/6/7 variable regions (35), we also observed the clones recognizing R. gnavus and one other species from the 8M community expressed VH5 or VH6 variable regions (Table 1). Therefore, a more accurate interpretation of our results is that 19 out of 21 antibodies (90%) resolved at the species level, which is in line with the species specificity observed in experiments involving total serum or fecal IgA. In addition, the majority of these sequenced IgA clones in the IGHV region have 0.1–3% difference compared to the nearest mouse germline IGHV region sequence, which suggests that these IgA⁺ B cells underwent, at least minimal, somatic hypermutation after class-switch recombination (Table 1). Though a few clones are binding to the same bacterial strain, the complementarity determining region 3 sequence and the percent difference in the IGHV region of these clones are different, suggesting the uniqueness of each IgA clone (Table 1).

Table 1.

VDJ sequences of IgA hybridoma clones derived from lamina propria B cells of gnotobiotic mice.

Clone ID	VH	DH	JH	CDR3 sequence	Targeted Bacteria†	% Difference in IGHV Region§
A2	V5-16*01	D1-1*01 or D1-2*01 or D2-12*01	J4*01	gcaagagctcccttacggggggctatggactac	B. ovatus	0.20
B6	V1-82*01	D2-1*01 or D2-10*01 or D2-11*02	J1*03	gcaagacgctatggtaactactggtacttcgatgtc	All 8 bacterial strains	1.85
B11	V1-11*01	N/A	J2*01	ggaaggggggaagtttttgactac	B. ovatus	0.40
C1	V1-11*01	N/A	J2*01	ggaaggggggagatctttgactac	B. ovatus	2.15
C2	V14-3*01	D2-4*01 or D2-9*02	J3*01	gctattgattacgacgggtttgcttac	B. caccae	1.55
C4	V5-17*01	D2-1*01 or D2-10*01 or D2-10*02	J1*03	gcaaggtatggtaactactggtacttcgatgtc	B. ovatus & R. gnavus	1.25
C8	V1-62-2*01 or V1-71*01	D3-3*01 or D4-1*01 or D4-1*02	J3*01	gcaagacacgaagaagggggacgttttgcttac	N/A‡	0.40
D4	V5-16*01	D1-1*01 or D1-2*01 or D2-12*01	J4*01	gcaagagctcccttacggggggctatggactac	B. ovatus & R. gnavus	0.75
F1	V14-3*01	D2-4*01 or D2-9*02	J3*01	gctattgattacgacgggtttgcttac	B. caccae	1.60
F2	V5-4*01	D1-1*01	J2*01	gcaagagccccttactacggttactactttgactac	R. gnavus	0.00
F3	V1-66*01	N/A	J2*01	gcaagatgcttggactactttgactac	N/A‡	0.80
F11	V1-82*01	D2-1*01 or D2-10*01 or D2-10*02	J1*03	gcaagacgctaatggtaactactggtacttcgaatgtctg	All 8 bacterial strains	0.40
G7	V6-6*01	D1-1*01	J3*01	accaggaactacggtagtaccccgtttgcttac	B. ovatus & R. gnavus	0.00

^†All bacteria strains were from public repositories (see table S1) and were identical as used in Fig. 2.

^‡Do not bind any of the 8 tested bacterial strains and common antigens.

^§% difference in IGHV region was calculated by comparing the difference (%) between the IGHV region sequence of the IgA hybridoma clones and the closest germline sequence.

The difference in OD₄₅₀ signal of different IgA clones that bind to identical bacteria suggests that these clones recognize distinct epitopes or with different avidity/affinity (Fig. 2B and Table 1). Compared to IgA clones recognizing a single bacterial strain from community 8M, the two hybridomas that bound all 8 strains from this bacterial community had a lower overall binding to each microbe as reflected by low OD₄₅₀ value, which may reflect the surface antigen epitopes that are prevalent across diverse taxa (Fig. 2B). We also tested the binding capability of select IgA clones towards the 8M community using flow cytometry under non-blocking conditions to confirm the specific binding capability of these antibodies (39). Consistent with our ELISA results, clones B11, C2 and F2 bound significantly better towards their targeting species than non-targeting species, which implies that these IgA antibodies are recognizing specific antigens expressed on the surface of each bacterial species (Fig. 2C). We then examined whether these IgA antibodies were just binding towards common antigens, including lipopolysaccharide (fig. S5A), double stranded DNA (fig. S5B), insulin (fig. S5C), flagellin (fig. S5D) and albumin (fig. S5E). Intriguingly, we did not see specific binding to these antigens with mild promiscuous binding usually observed in clones B6 and F11 that also target all eight strains (fig. S5).

Screening the specificity of lamina propria IgA towards non-resident bacterial strains

We hypothesized that a fraction of our IgA clones could be strain-specific, or at least unable to bind all strains from a specific species. This could explain our findings of strain-level resolution with polyclonal mixtures from the in vivo bulk IgA samples (Fig. 1D). We used five to six additional strains within the species of B. ovatus, B. caccae, and R. gnavus (table S1) from different human donors to screen the specificity of antibody binding capability from the hybridomas generated from mice colonized with the 8M community. One of the four B. ovatus-binding IgA clones, B11, recognized all six newly tested B. ovatus strains (Fig. 3A and fig. S6A). IgA clone C1 recognized four out of six strains whereas clone C3 bound two out of six strains (Fig. 3A and fig. S6A). IgA clone A2 did not bind to any of the new strains (Fig. 3A and fig. S6A). Both B. caccae-binding IgA clones only recognize one out of six new strains (Fig. 3B and fig. S6B). Finally, the two R. gnavus specific monoclonal antibodies bound all five new strains, suggesting they were species-specific (Fig. 3C and fig. S6C). We also tested the binding capability of these bacteria-binding IgA clones towards a larger bacterial library, which included 18 different bacterial species belonging to different phyla. Consistent with our observations, none of the antibodies showed binding to these new species (fig. S6, D to F) confirming their species resolution. The bacterial strains within each species may be different in their genome sequence and each clone is binding to these strains at various capability, we wondered whether the genomic similarity between strains is correlated positively with the relative binding capability of each IgA clone. However, we did not find a significant positive correlation for any tested clone (fig. S7, A to C). To increase the power of calculation, we pooled all clones binding to the same species together and found a slight, but significant, and positive correlation in the B. ovatus-binding clones (Fig. 3D). Such correlation, however, was not seen in other bacteria-binding clones, together (Fig. 3, E and F).

Fig. 3. Monoclonal IgAs isolated from lamina propria bind bacteria at species and strain levels.

(A to C) Reactivity of monoclonal IgA antibodies towards different B. ovatus strains (A), B. caccae strains (B) and R. gnavus strains (C) that were not used to colonize mice. Strains denoted by the white/open bar belong to the 8M community that colonized gnotobiotic mice. (D to F) Correlation of genomic similarity (X-axis) and binding similarity (Y-axis) of IgA clones that bind to B. ovatus (D), B. caccae (E) and R. gnavus (F). The genomic similarity was calculated as the genome k-mer similarity between each strain and the particular strain that was used to colonize gnotobiotic mice. The binding similarity was calculated as the area under curve of each strain over the area under curve of the particular strain that was used to colonize gnotobiotic mice. Results are the average value of two technical replicates (A to C). Dotted lines denote A.U. in negative control. Detailed information of bacterial strains is listed in table S1. A.U.: arbitrary units.

Overall these in vitro results with monoclonal antibodies demonstrate the majority of monoclonal IgA antibodies resolve at least at the species level. Amongst these species-specific antibodies, the specificity appears to vary from being capable of binding all members of the species to only a subset of strains in the species – presumably driven by the variation of the IgA-binding antigen/epitopes in the species pan-genome. These observations for monoclonal antibodies parallel our ex vivo results with bulk serum and fecal IgA antibodies, where the resolution of polyclonal IgA was predominantly species-specific and, in some cases, strain-specific (Fig. 1).

Hybridoma-produced IgA antibodies still retain binding capability towards bacterial antigen in stool

As the most abundantly secreted immunoglobulin isotype, IgAs are secreted into saliva, colostrum and mucosal linings. We wondered whether hybridoma-produced IgA antibodies still retained binding capability towards bacterial surface antigens after passing through the gastrointestinal tract – thus representing a potential oral therapeutic with defined specificity towards components of the microbiome. Rag1^−/− mice, which lack IgAs in their stool, were orally administered with hybridoma-produced IgA antibodies. We then quantified IgA level in stool by ELISA (Fig. 4A and fig. S8A). Three hours later, IgA antibodies were successfully detected and decreased slightly 6 hours after the initial gavage, which was in line with the transit time of murine gastrointestinal tract (Fig. 4B and fig. S8B). Interestingly, each of these tested IgA antibody clones still retained binding capability towards their targeted bacterial species after passing through the gastrointestinal tract (Fig. 4C and fig. S8C). By adding IgAs in the drinking water of Rag1^−/− mice, we can successfully dose the level of IgA antibodies in the stool of these mice; fecal IgA level was higher in the samples collected in the morning than that collected in the evening reflecting their increased nocturnal activity (Fig. 4, D and E and fig. S8, D to G).

Fig. 4. Hybridoma-produced monoclonal IgA antibodies retain binding capability towards bacterial antigen after passing through intestinal tract.

(A) Schematic representation of antibody gavage in Rag1^−/− mice (100 μg IgA per gavage). (B) IgA quantification in the stool of Rag1^−/− mice at different time points after IgA clone F2 antibody gavage. (C) The binding capability of IgA clone F2 in the stool of Rag1^−/− mice at hour 9 after initial gavage. sup.: supernatant. (D) Correlation of IgA clone F2 concentration in drinking water (X-axis) and in stool (Y-axis). Mice were given water containing different concentration of IgA clone F2 at 7 pm. The next day at 7 am, stool samples were collected and IgA level was quantified by ELISA. (E) IgA clone F2 concentration in the stool of mice at different time points. Mice were given water containing 0.2 mg/ml IgA clone F2 at 7 pm. Stool samples were collected every 12 hours in the following 36 hours. Data are shown as means ± SEM.

Because the main form of secretory IgA antibodies is dimeric in structure, we wondered whether our hybridoma-produced IgAs had a monomeric or a dimeric structure. On an immunoblot analysis for clones B11, C2, F2 and F11, we detected IGJ, which covalently links two monomeric IgA fragments to form a dimeric IgA (fig. S9). The above results demonstrate our hybridoma-produced IgA antibodies, at least for the tested clones, are dimeric in structure, resistant to digestive and bacterial enzyme degradation and retaining binding capability after passing through gastrointestinal tract, which may be used potentially for therapeutic purposes (5, 7, 40).

DISCUSSION

Gut IgAs facilitate host-microbe interactions, constrain commensals within the intestinal lumen, modulate bacterial behaviors and shape the composition and topography of bacteria (11, 39, 41–45). However, the fine specificity of gut IgAs towards commensal bacteria still remains elusive. Here we demonstrated that bulk IgAs from both stool and serum recognize “self” bacteria that colonized the host more effectively than “non-self” bacteria of the same species but from a different host. We also observed that about 75% of hybridoma-derived IgA clones generated from single gut LP B cells of gnotobiotic mice resolve bacteria at least at the species level, and only a few IgA clones display cross-reactivity. Interestingly, some IgAs bind specifically to the host colonizing strain, while others bind to a variable number of strains that belong to the same species but are extraneous to the host.

Several studies have investigated the reactivity of gut IgA towards bacterial surface antigens. Upon monocolonization, bacteria-reactive IgAs can be detected in vitro by ELISA upon culturing fragments of Peyer’s patches and by ELISpot after culturing purified LP immune cells (46). These observations were later confirmed by cloning IgA-secreting B cells from the LP of monocolonized mice (34). In agreement with our results, most IgAs released by single B cell clones from the human gut LP have been shown to react against a single bacterial species (47). Intriguingly, after FMT in recurrent Clostridioides difficile patients, IgA-targeted bacterial taxa in recipients were dramatically altered and resembled more the IgA-targeting patterns of the donors, suggesting the transformation of mucosal IgA specificity towards new resident gut bacteria after FMT (48).

However, it has also been reported that most IgA clones from the LP of specific pathogen-free mice display polyreactivity for common antigens such as DNA, LPS, and flagellin in addition to recognizing a broad range of gut microbes (31). Many factors may account for the discrepancy. For example, there may be strain-level differences among bacteria, which implies that specificity screening should be run with carefully selected bacterial strains to distinguish between resident “self” and non-resident “non-self” strains. Also, the degree of complexity and the composition of a bacterial consortia used to colonize gnotobiotic mice likely has an impact on the production and specificity of IgA antibodies. We previously reported that C57BL/6 WT mice purchased from different vendors (i.e., Jackson Laboratory, Charles River Laboratories and Taconic Biosciences) had significantly different levels of total fecal IgA (19), probably due to gut microbiota differences, where the Taconic mice are well known for being colonized by segmented filamentous bacteria, a potent IgA inducer (17). In addition, the gut bacterial metabolite, acetate, also influences the reactivity of IgA to bacterial antigen through regulating the epithelial-immune cell interactions (49). Therefore, further experiments may be required to address the discrepancy.

Of note, our results also demonstrate the hybridoma-derived IgAs can transit through the harsh environment of the gastrointestinal tract and retain binding capability against bacterial antigens. This observation clearly implies the potential usage of in vitro engineered IgA antibodies as therapeutics (40, 50–52). Compared with naturally secreted IgA antibodies, those hybridoma-derived IgAs lack the bound secretory component, which may ultimately compromise their stability to bacterial proteases in the gut lumen (53, 54). Thus, it would be important to dissect the capability of gut bacteria, including both commensal and pathogens, to degrade secretory component-free IgAs prior to advancing hybridoma-derived IgAs as therapeutics.

In summary, our data show that gut bacteria-induced IgAs have higher reactivity towards “self” strains in mice colonized by a single microbe or a consortium of bacteria. We also demonstrate that the host generates species-specific and strain-specific rather than polyreactive gut IgA antibodies and species-specific IgA antibodies may or may not cross-react with a few to many different strains from the same species. By unraveling the complex specificity of bacteria-induced gut IgA antibodies in gnotobiotic models, our findings highlight the potential role of gut IgAs in the symbiotic host-microbe interactions, because the coating of bacteria by IgA antibodies influences not only the homeostasis of the mucosal immune system but also the physiology of bacteria themselves (21, 24, 25, 39, 55).

MATERIALS AND METHODS

Mice

Germ-free C57BL/6 mice were bred and maintained in flexible plastic gnotobiotic isolators (Class Biologically Clean, Ltd.) in Icahn School of Medicine at Mount Sinai. All mice were housed in groups under controlled temperature, a 12-hour light/dark cycle, and allowed ad libitum access to diet and water. Usage of animals for experiments was approved by the Institutional Committee on Use and Care of Animals (IACUC) in Icahn School of Medicine at Mount Sinai.

Inoculation of germ-free mice with cultured bacteria

In experiments involving gnotobiotic mice usage and colonization, mice were handled as previously described (37). Briefly, germ-free mice (~8 weeks old) were transferred outside from isolator into a laminar flow biosafety hood that was pre-cleaned with CLIDOX-S^®. Frozen bacteria cultures in anaerobic glass vials were sterilized with CLIDOX-S^®. After being thawed, ~200 μl aliquot of bacteria suspension was introduced into recipient mice via oral gavage. Colonized mice were then housed in clean filter top cages, which contained sterilized mouse diet, water and bedding, for the duration of the experiment.

Growth of bacterial strains

B. ovatus (ATCC 8483), B. caccae (ATCC 43185), B. thetaiotaomicron (VPI 5482), B. vulgatus (ATCC 8482), R. gnavus (ATCC 29149), C. bolteae (ATCC BAA-613), C. aerofaciens (ATCC 25986) and E. coli (K-12 MG1655), P. johnsonii (DSMZ 18315) and B. intestinalis (DSMZ 17393) were obtained from global repositories: the American Type Culture Collection (ATCC) and Deutsche Sammlung von Mikroorganismen und Zellkulturen (DSMZ). Apart from E. coli, the bacterial strains were grown under anaerobic condition at 37°C in Brain Heart Infusion (BHI) medium supplemented with 0.5% yeast extract (Difco Laboratories), 0.4% monosaccharide mixture, 0.3% disaccharide mixture, L-cysteine (0.5 mg/ml; Sigma-Aldrich), malic acid (1 mg/ml; Sigma-Aldrich) and 5 μg/ml hemin. LB Broth Miller (EMD Chemicals, Inc.) was used for E. coli culture under aerobic conditions at 37°C. After bacterial cultures reached an OD₆₀₀ of 1–2, glycerol was added to a final concentration of 15% (v/v). The mixture was aliquoted into glass vials with a crimped top and stored at −80°C freezer for further use.

Isolation of single bacterial strains from human stool samples

Except for strains obtained from public culture repositories (ATCC and DSMZ), individual bacterial strain in tables S1 was isolated from previously banked, deidentified stool samples (19, 56). These bacteria were cultured under anaerobic conditions, and the identity of each individual bacterium was verified by whole genome sequencing and mass spectrometry (Bruker Corp.).

Lymphocyte isolation from mice intestines

Laminar propria lymphocytes were isolated as described previously (56). Briefly, both small intestine and colon were excised from gnotobiotic mice, followed by cleaning visceral fat and intestinal content. Tissues were opened longitudinally, washed twice in Hank’s Balanced Salt Solution (HBSS) without Ca²⁺ and Mg²⁺ (GIBCO) and incubated in dissociation buffer (HBSS including 10% fetal bovine serum (FBS), 5 mM EDTA and 15 mM HEPES) for 30 min at 37°C with agitation to remove epithelium and intraepithelial lymphocytes. The remaining tissues were then washed three times in ice cold HBSS, cut into ~2 cm pieces and digested with collagenase (Sigma-Aldrich), DNase I (Sigma-Aldrich) and dispase I (Corning) for 40 min at 37°C with mild agitation. Cell suspensions were filtered through 70 μm cell strainers, washed three times and resuspended in IMDM/2%FBS.

Generation of hybridoma clones

Sp2/0-Ag14 myeloma cell line (ATCC CRL-158) was cultured in Dulbecco’s Modified Eagle’s Medium (DMEM) supplemented with 10% FBS. Before fusion, 80–90% confluence was achieved. After counting cell number, Sp2/0-Ag14 myeloma cell line and LP lymphocytes were mixed thoroughly and carefully at a ratio of 1:5. Mixed cells were washed in Hybridoma-SFM. Cell pellet was loosened gently. Then polyethylene glycol (PEG) was slowly added along the side of the tube. Tube was incubated at 37°C for 1 minute and then a 10 times volume of Hybridoma-SFM was slowly added to PEG treated cells. After spinning down, the fused cells were incubated overnight in complete media. The next day, the cells were centrifuged and resuspended in semi-solid hybridoma selection medium containing hypoxanthine, aminopterin and thymidine (HAT), mixed well and gently transferred to petri dishes. After incubation at 37°C for 2 weeks, single cell colonies were picked into 96 well plates by colony picker (Hamilton Company) and cultured for another 2 days for antibody isotyping. An aliquot of cell suspension was taken out and frozen for further examination.

Production of monoclonal IgA antibody

Selected clones of IgA hybridoma cells were thawed and expanded in DMEM buffer supplemented with 10% FBS (HyClone). When hybridoma cells grew to 90% confluence, they were washed three times in PBS and then transferred to the lower level chamber of CELLine Disposable Bioreactor (Corning). Hybridoma-SFM (Gibco by Life Technologies) was added to the upper level of the Bioreactor. In the first two weeks, Hybridoma-SFM was changed within 24 hours when medium changed color from red to orange. When a very high density of cells was observed, about 50% of cells and medium in the lower chamber were harvested. After spinning down, supernatant was kept for antibody purification. Cells in the bioreactor were cultured for four weeks and medium in the second level of the chamber were frequently harvested. IgA antibody in medium was precipitated in ammonium sulfate solution and then dialyzed in a cassette in PBS overnight at 4°C. Concentration of dialyzed IgA antibody was quantified by ELISA. Production of larger IgA quantifies was achieved from peritoneal cavity fluid of BALB/c mice after hybridoma cell injection. Ascites were harvested, precipitated and dialyzed.

Detection of bacteria-binding IgA antibodies from stool and serum by ELISA

Bacteria were cultured to OD₆₀₀ = 1~2 in appropriate condition and then fixed in 0.5% formaldehyde in PBS for 20 min at room temperature. After washing 3 times, bacteria density was adjusted to OD₆₀₀ = 1 in PBS. ELISA plates were coated with 30 μl of the adjusted bacterial suspension and incubated at 4°C overnight. After washing and blocking with 1% BSA, diluted fecal or seral samples containing polyclonal IgA were added and incubated overnight at 4°C. Captured IgA was detected by horseradish peroxidase (HRP)-conjugated goat anti-mouse IgA antibody (Sigma-Aldrich; A4789). ELISA plates were developed by 3,3’,5,5’-Tetramethylbenzidine (TMB) microwell peroxidase substrate (KPL, Inc.; 50–76-03) and quenched by 1 M H₂SO₄. The colorimetric reaction was measured at OD = 450 nm by a Synergy HTX Multi-Mode Microplate Reader (BioTek Instruments, Inc.).

Flow cytometry of bacteria bound with hybridoma-produced IgA antibodies

Bacteria were cultured to OD₆₀₀ = ~1 in rich broth and subsequently washed in PBS. Hybridoma produced IgA antibody at 10 μg/ml was added to samples and incubated for 30 min at room temperature. After washing 3 times in PBS/2%FBS/2mM EDTA, samples were further stained with monoclonal rat anti-mouse IgA antibody (Thermo Fisher Scientific, clone mA-6E1) for 30 min at room temperature. The samples were then washed 3 times. DAPI (4’,6-Diamidino-2-Phenylindole, Dihydrochloride) (Thermo Fisher Scientific, Cat# 62248) was used for staining bacterial nuclei. Samples were run through a BD LSR Fortessa™ cell analyzer and further analyzed in FlowJo software (Tree Star, Inc.). DAPI positive events were regarded as whole bacteria and gated for further analysis (fig. S10).

Screening the specificity of hybridoma-produced IgA by ELISA

For the screen of monoclonal IgA antibodies against bacterial antigens, reciprocal dilutions of hybridoma-produced IgA antibodies were added to ELISA plates, which were pre-coated with whole bacteria of the 8M community, respectively, and blocked with 1% BSA. The captured IgA was detected by horseradish peroxidase (HRP)-conjugated goat anti-mouse IgA antibody (Sigma-Aldrich; A4789). ELISA plates were then developed by TMB microwell peroxidase substrate (KPL, Inc.; 50–76-03), quenched by 1 M H₂SO₄, and read using a BioTek Synergy HTX plate reader. The polyreactivity of hybridoma-produced IgA was tested by adding antibody to ELISA plates that were pre-coated with individual common antigen including lipopolysaccharide (Sigma-Aldrich; L4391), double stranded DNA (Sigma-Aldrich; D4522), insulin (Sigma-Aldrich; 91077C), flagellin (Sigma-Aldrich; SRP8029) and albumin (Sigma-Aldrich; A1653). ELISA plates were then washed, developed and read as described above.

VDJ sequencing of IgA hybridoma clones

Hybridoma cell lines were cultured in DMEM buffer supplemented with 10% FBS. Cells were harvested at 90% confluence after washing three times with PBS. Total mRNA was extracted with the RNeasy Mini Kit (QIAGEN, 74104) according to manufacturer’s protocol. cDNA was synthesized with High-Capacity cDNA Reverse Transcription Kits (Applied Biosystems, 4368813) and approximately 200 ng cDNA was used for PCR amplification with Phusion High-Fidelity PCR Master Mix with HF Buffer (Thermo Fisher Scientific, F531L). The primers for amplicons of murine IgA are VH promiscuous 5’-GAGGTGCAGCTGCAGGAGTCTGG-3’ in combination with Cα (outer) 5’-GAGCTCGTGGGAGTGTCAGTG-3’ as described previously (27, 57). PCR conditions for IgA amplicons were set as follows: 30 s at 98°C, and then 28 cycles of (98°C for 10 s; 62°C for 30 s and 72°C for 45 s); and then 72°C for 5 min. Amplicons were sent for Sanger sequencing to Psomagen, Inc. The VDJ sequences are provided in data file S1. Sequenced reads were analyzed in IGBLAST (http://www.ncbi.nlm.nih.gov/igblast) with IMGT Mouse V, D and J Gene Databases (58) using reads from both primers. The VDJ assignments and CDR3 sequences are listed in Table 1. The % difference in IGHV region was calculated by comparing the difference (%) between the IGHV region sequence of the IgA hybridoma clones and the nearest germline sequence. The average percentage difference from both primers is displayed in Table 1.

Statistical analysis

Statistical significance between two groups was assessed by unpaired, nonparametric, Mann-Whitney tests unless stated otherwise. Comparisons among three or more groups were performed using one-way ANOVA. Pearson correlation coefficient was employed for correlation tests. Data plotting and statistical analysis were performed using GraphPad Prism 7.0 (GraphPad Software, La Jolla, CA). A p-value less than 0.05 was considered statistically significant.

Supplementary Material

reproducibility checklist

MDAR Reproducibility Checklist.

data file s2

Data file S2. Raw data file (Excel spreadsheet)

data file s1

Data file S1. VDJ sequences of IgA hybridoma clones

main supplementary

Fig. S1. Screening the binding capability of fecal and serum bulk IgAs towards different bacteria.

Fig. S2. Stool bulk IgAs from the F1 generation of B. ovatus monocolonized mice bind B. ovatus more than other tested bacterial species.

Fig. S3. CD4⁺ T cells play an essential role in the production of B. ovatus binding IgAs in B. ovatus monocolonized mice.

Fig. S4. Specificity screening of hybridoma-produced IgA towards bacterial surface antigens.

Fig. S5. Bacteria-specific monoclonal IgA antibodies do not bind to common antigens.

Fig. S6. Specificity of hybridoma-produced IgA towards different strains and species of bacteria that did not colonize host mice.

Fig. S7. The genomic similarity is not correlated positively with relative binding capability of IgA clones.

Fig. S8. Hybridoma-produced monoclonal IgA antibodies retain bacterial antigen binding capability after passing through the intestinal tract.

Fig. S9. Analysis of immunoglobulin J chain in hybridoma-produced IgA antibodies by immunoblot.

Fig. S10. Representative flow cytometry plots for gating IgA-bound bacteria.

Table S1: Detailed information of bacterial strains used in this study.

Data and materials availability:

All data needed to evaluate the conclusions in this paper are present in the paper or the Supplementary Materials. Bacterial genome sequencing data related to this study have been deposited in NCBI’s Assembly database under the individual accession numbers listed in Table S1. Unique/stable reagents generated in this study are available upon request from the corresponding author after completion of a Material Transfer Agreement.