Abstract
The vast majority of IgA production occurs in mucosal tissue following T cell–dependent and T cell–independent Ag responses. To study the nature of each of these responses, we analyzed the gene-expression and Ig-reactivity profiles of T cell–dependent CD27+IgA+ and T cell–independent CD27−IgA+ circulating memory B cells. Gene-expression profiles of IgA+ subsets were highly similar to each other and to IgG+ memory B cell subsets, with typical upregulation of activation markers and downregulation of inhibitory receptors. However, we identified the mucosa-associated CCR9 and RUNX2 genes to be specifically upregulated in CD27−IgA+ B cells. We also found that CD27−IgA+ B cells expressed Abs with distinct Ig repertoire and reactivity compared with those from CD27+IgA+ B cells. Indeed, Abs from CD27−IgA+ B cells were weakly mutated, often used Igλ chain, and were enriched in polyreactive clones recognizing various bacterial species. Hence, T cell–independent IgA responses are likely involved in the maintenance of gut homeostasis through the production of polyreactive mutated IgA Abs with cross-reactive anti-commensal reactivity.
Introduction
The microbiome of the human gastrointestinal tract contains large numbers of bacteria (up to 30,000 species) (1). The majority of these bacteria are coated with Igs (2) that are generated in dynamic responses (3, 4). Indeed, the mucosal surfaces of the intestinal tract, the oral cavity, and lungs are major sites of Ab production, mainly the secretory form of IgA (5).
Each B cell carries surface Ig generated through V(D)J recombination of IgH and Igκ and Igλ L chain genes during stepwise differentiation in the bone marrow (6, 7). Upon Ag recognition, these newly generated B cells undergo responses involving affinity maturation by induction of somatic hypermutations (SHMs) in the Ig variable domains and class-switch recombination (CSR) from the IgM, for example, to the IgA isotype (8). SHM and CSR are mediated by activation-induced cytidine deaminase (AID) (9), which is upregulated through CD40 signaling following interaction with CD40L on activated CD4+ T cells. Such T cell–dependent (TD) responses take place in germinal center reactions in lymphoid tissues. Alternatively, AID expression can be induced in T cell–independent (TI) B cell responses, which are associated with limited proliferation and affinity maturation to lipid or carbohydrate structures (8, 10–13). TI class-switching toward IgA is well supported by the microenvironment of the gut, especially by dendritic cells (DCs) in the GALT. These DCs secrete retinoic acid (RA) that activates circulating B cells to induce expression of adhesion molecule α4β7 and chemokine receptor CCR9, which mediate gut homing (14). Upon activation via TLRs, DCs and monocytes secrete BAFF and APRIL, which bind TACI on B cells and can induce CD40-independent class-switching toward IgA (15–18). In addition, DC-derived TGF-β and RA act in concert with IL-5, IL-6, and IL-10 to induce differentiation of B cells into Ab-secreting plasma cells (14, 18–20).
Although ∼25% of intestinal IgA-producing plasmablasts are polyreactive, they show molecular signs of Ag-mediated selection (21), fitting with Ag-induced production rather than secretion of natural Abs independent of Ag stimulation. It is tempting to speculate that TI IgA is directed against cell wall components of commensal bacteria to support the formation of a biofilm and to disable their translocation through the epithelial layer (22, 23). This would prevent priming of systemic high-affinity TD responses to beneficial gut microbiota. Indeed, MyD88/TRIF double-knockout mice deficient in TI IgA production spontaneously developed systemic responses against gut microbiota (24).
We recently distinguished two circulating human IgA+ memory B cell subsets: conventional CD27+IgA+ cells were dependent on T cell help, whereas unconventional CD27−IgA+ cells were present in CD40L-deficient individuals (25). Moreover, the limited replication history of CD27−IgA+ memory B cells, their low frequency of SHM, and increased IgA2 usage were features reminiscent of IgA+ B cells from the intestinal lamina propria (25, 26). We show in this study that both CD27+IgA+ and CD27−IgA+ B cell subsets are typical memory B cells, as is evident from their gene-expression profiles and detailed immunophenotypes. From single-cell–sorted CD27+IgA+ and CD27−IgA+ memory B cells we produced in vitro rAbs to assess their reactivity to various Ags and bacterial strains. We found that a large fraction of CD27−IgA+ memory B cells express polyreactive Abs with a unique repertoire and reactivity toward commensal bacteria, suggesting that these B cells play an important role in maintaining mucosal immunity.
Materials and Methods
Cell sorting and gene-expression profiling
Three naive and six human memory B cell subsets were purified from post-Ficoll mononuclear cells on a FACSAria I cell sorter (BD Biosciences) (25, 27). Naive B cells were separated into CD38+CD27−IgD+IgM+ transitional B cells, CD38dimCD27−IgD+IgM+CD5+ pre–naive B cells, and CD38dimCD27−IgD+IgM+CD5− mature naive B cells. Memory B cells were separated into CD38dimCD27+IgD+IgM+ natural effector B cells, CD38dimCD27+IgD−IgM+ IgM-only B cells, and CD38dimCD27+IgA+, CD38dimCD27+IgG+, CD38dimCD27−IgA+, and CD38dimCD27−IgG+ subsets. RNA was isolated from each sorted subset with the RNeasy Mini Kit (QIAGEN). Gene expression was quantified using Affymetrix HG-U133 Plus 2.0 GeneChip arrays (containing 54,675 probe sets), as previously described (7, 27, 28), and all data were deposited in the ArrayExpress database (http://www.ebi.ac.uk/arrayexpress) under accession numbers E-MEXP-3767 and E-MTAB-3637. Expression profiles of the three naive and six memory B cell subsets from three healthy donors were compared based on the perfect-match probe-intensity levels. RMA background removal and quantile normalization were performed, followed by a per-probe set two-way ANOVA (with factors probe and cell type). This resulted in average expression levels for each probe set in each cell type, as well as p values for the significance of the difference between cell types. The p values were adjusted for multiple testing using Šidák stepdown adjustment (29), and all differences with adjusted p values < 0.05 were considered significant.
The maximum absolute difference in expression between any two cell types was calculated to select probe sets that showed a signal. Only probe sets that showed larger differences than a log2 threshold value of 0.7 were selected for clustering. Correlation ρ between samples was calculated based on only these selected probe sets, and the data were hierarchically clustered (complete linkage) using 1 − ρ as a distance measure (28).
Single-cell sorting
Post-Ficoll mononuclear cells from healthy donors were enriched for B cells by magnetic separation with CD19 or CD20 MicroBeads (Miltenyi Biotec) and stained with CD20–PE–Cy7, CD27-allophycocyanin (both from BioLegend), CD38-FITC (BD Pharmingen), CD27-allophycocyanin, and IgA-PE (Southern Biotech) prior to purification. Single CD20+CD38dimCD27+IgA+ and CD20+CD38dimCD27−IgA+ memory B cells were sorted on a FACSAria flow cytometer (BD Biosciences) into 96-well PCR plates and immediately frozen on dry ice.
cDNA synthesis, Ig gene amplification, Ab production, and purification
RNA from single cells was reverse-transcribed in the original 96-well plate in 12.5-μl reactions containing 100 U Superscript II RT (Life Technologies) for 45 min at 42°C. RT-PCR reactions and primer sequences were as described previously (30–32), supplemented with an IGHA-specific primer (5′-CTTTCGCTCCAGGTCACACTGAG-3′) for the first PCR reaction. Cloning strategy, expression vectors, Ab expression, and purification were performed as described previously (30, 31); Ig sequences were analyzed by IgBLAST comparison with GenBank. IgH-CDR3 was defined as the interval between the conserved arginine/lysine at position 94 in IGHV-FR3 and the conserved tryptophan at position 103 in the IGHJ gene. SHM selection strengths were determined using the BASELINe program (http://selection.med.yale.edu/baseline/) (33).
ELISAs and immunofluorescence assays
Ab-reactivity analysis was performed as described previously with the highly polyreactive ED45 Ab as positive control for HEp-2 reactivity and polyreactivity (30, 31). Abs were considered polyreactive when they recognized the three distinct Ags: dsDNA, insulin, and LPS. ELISA plates for bacteria-reactivity testing were coated with purified flagellin from Bacillus subtilis (InvivoGen) or sonicated lysates from cultured Enterobacter cloacae (ATCC 13047), Enterococcus faecalis (ATCC 29212), Escherichia coli, or Streptococcus aureus or were obtained by multiple cycles of freezing and thawing lysates from Bacteroides fragilis (ATCC 2528) and Clostridium difficile (ATCC 9689) at a concentration of 1 ng/μl. For indirect immunofluorescence assays, HEp-2 cell–coated slides (Bion Enterprises) were incubated in a moist chamber at room temperature with purified rAbs at 50–100 μg/ml, according to the manufacturer’s instructions. FITC-conjugated goat anti-human IgG was used as detection reagent.
Statistical analysis
Statistical analyses were performed with the two-tailed Student t test, Mann–Whitney U test, or χ2 test, as indicated in the figure legends. The p values < 0.05 were considered statistically significant.
Results
Gene-expression profiling of naive and memory B cell subsets
Circulating CD27+IgA+ and CD27−IgA+ B cells both display a memory B cell phenotype with distinct features, suggesting that they may originate from different types of immune responses (25). To study whether the TI origin of CD27−IgA+ B cells results in a typical memory B cell transcription program, we compared their gene-expression profile with three naive and five other memory B cell subsets.
Unsupervised clustering analysis based on 399 probe sets that showed the greatest variation in expression between any two samples yielded three main clusters (Fig. 1A). Cluster 1 contains the three naive B cells (transitional, prenaive, and mature naive), cluster 2 contains the natural effector and IgM-only memory B cells, and cluster 3 contains the Ig class-switched CD27+IgA+, CD27+IgG+, CD27−IgA+, and CD27−IgG+ memory B cells (Fig. 1A). Among class-switched memory B cells, CD27+IgA+ cells mostly resembled CD27+IgG+ cells, and CD27−IgA+ cells were most similar to CD27−IgG+ cells (Fig. 1A). The expression patterns of these 399 probe sets in CD27−IgA+ B cells correlated well with all Ig class–switched cells, showed medium correlation with the IgM+ memory B cell subsets, and differed the most from naive B cells (Fig. 1B, Supplemental Table I).
Gene-expression profiling of naive and memory B cell subsets. (A) Hierarchical clustering (complete linkage) using 1 − correlation as a distance measure based on the 399 probe sets that showed the most variation between any two samples (threshold at log2 value 0.7). Clustering analyses was performed without bias for known genes or subsets. (B) Correlation of the gene-expression profiles of these 399 genes for CD27−IgA+ memory B cells compared with the naive and memory B cell subsets. (C) Heat map with normalized expression levels of selected probe sets in naive and memory B cell subsets. Z-scores were maximized to −2 and 2. *Statistically significant difference between CD27+IgA+ and mature naive B cells, #statistically significant difference between CD27−IgA+ and mature naive B cells, +statistically significant difference between CD27+IgA+ and CD27−IgA+ B cells.
Gene-expression profiling of naive and memory B cell subsets. (A) Hierarchical clustering (complete linkage) using 1 − correlation as a distance measure based on the 399 probe sets that showed the most variation between any two samples (threshold at log2 value 0.7). Clustering analyses was performed without bias for known genes or subsets. (B) Correlation of the gene-expression profiles of these 399 genes for CD27−IgA+ memory B cells compared with the naive and memory B cell subsets. (C) Heat map with normalized expression levels of selected probe sets in naive and memory B cell subsets. Z-scores were maximized to −2 and 2. *Statistically significant difference between CD27+IgA+ and mature naive B cells, #statistically significant difference between CD27−IgA+ and mature naive B cells, +statistically significant difference between CD27+IgA+ and CD27−IgA+ B cells.
All nine B cell subsets showed high gene-expression levels of pan-B markers CD19, CD20, BAFF-R, CD79A, and CD79B (Fig. 1C, Supplemental Table I), and the gene-expression levels for markers that were used to define the subsets (CD5, CD38, IgM, IgD, IgA, IgG, CD27) correlated well with their protein-expression levels. In addition, all memory B cell subsets showed the expected increase in activation markers and costimulatory molecules, such as CD80, CD86, TACI, FAS, and CD58, and downregulation of genes encoding inhibitory receptors CD22 and CD72 compared with naive B cell subsets (Fig. 1C) (25, 34). These expression patterns were confirmed at the protein level by flow cytometry (Fig. 2A) (25). TLR gene-expression levels were similar for IgA+ and other memory B cells (Supplemental Table I), although CD27−IgA+ cells contained fewer TLR1 and more TLR10 transcripts than did mature naive B cells. However, membrane TLR-1 and TLR-10 expression levels were similarly low in naive and memory B cell subsets (Fig. 2A). In addition, there were no signs of TLR signaling pathway deregulation in IgA+ cells. Furthermore, CD27−IgA+ memory B cells showed similar expression of signaling molecules involved in BCR and CD40 signaling pathways to all other memory B cell subsets. Comparison with mature naive B cells revealed upregulation of GAB2 and GRB2 (adaptor molecules in the BCR signaling pathway) and downregulation of LYN in CD27−IgA+ B cells, as well as downregulation of TANK in the CD40 signaling pathway. However, the majority of transcripts from both signaling pathways were not expressed significantly differently between the subsets. Both IgA+ B cell subsets showed low IL4R and high IL6R and IL10RA expression, supporting the role of IL-6 and IL-10 in IgA+ memory B cell differentiation (Supplemental Table I).
Expression levels of selected markers on memory B cell subsets. (A) Expression levels of selected costimulatory molecules (CD58, CD86), BCR signaling inhibitors (CD22, CD72), TLRs (TLR-6, TLR-10), and chemokine receptors (CCR7, CCR9) were analyzed by flow cytometry on IgM+, IgG+, and IgA+ memory B cell subsets. Each plot contains a filled graph representing the isotype control and a dark gray line graph representing mature naive B cells. (B) Expression levels of selected genes were analyzed by real time quantitative PCR. Each bar represents mean fold expression relative to control gene ABL; error bars are SEM. The number of analyzed samples is indicated below the name of the subset. Data were analyzed with Mann–Whitney U test. *p < 0.05, **p < 0.01.
Expression levels of selected markers on memory B cell subsets. (A) Expression levels of selected costimulatory molecules (CD58, CD86), BCR signaling inhibitors (CD22, CD72), TLRs (TLR-6, TLR-10), and chemokine receptors (CCR7, CCR9) were analyzed by flow cytometry on IgM+, IgG+, and IgA+ memory B cell subsets. Each plot contains a filled graph representing the isotype control and a dark gray line graph representing mature naive B cells. (B) Expression levels of selected genes were analyzed by real time quantitative PCR. Each bar represents mean fold expression relative to control gene ABL; error bars are SEM. The number of analyzed samples is indicated below the name of the subset. Data were analyzed with Mann–Whitney U test. *p < 0.05, **p < 0.01.
The Runx2 and Runx3 transcription factors act downstream of the TGF-β and RA signaling pathways to induce TI class-switching toward IgA in the gut (35). Although RUNX3, TGFBR, and RARA transcript levels were similarly high in all analyzed B cell subsets, RUNX2 was exclusively expressed by IgA+ B cells (Figs. 1, 2B, Supplemental Table I), especially in CD27−IgA+ B cells, supporting their TI origin. Thus, CD27−IgA+ B cells appear to be true memory B cells; they display the highest expression of RUNX2, which may play an important role in their development.
Differential expression of mucosa-homing–related genes by B cell subsets
Expression of chemokine receptors on lymphocytes determines their ability to migrate in response to stimuli. All analyzed B cell subsets expressed lymph node–homing receptors CXCR4 and CCR7, but their levels were significantly higher on naive than on class-switched memory B cells (Supplemental Table I). CCR7 protein was present on nearly all mature naive B cells but only on a fraction of cells within each memory B cell subset (Fig. 2A).
Stimulation of chemokine receptors induces surface expression of diverse adhesion molecules. All B cell subsets showed similarly high expression of genes encoding CD62L (SELL), α4β7 (ITGA4/ITGB7), and LFA-1 (ITGAL/ITGB2) involved in migration to lymph nodes (36, 37). In addition, ITGB1, which encodes the β1 subunit of the α4β1 integrin, was upregulated in all memory B cells (Supplemental Table I). Although none of the B cell subsets expressed the mucosal-homing marker CCR10 (38), CD27−IgA+ B cells specifically expressed the small intestine–homing receptor CCR9 (Supplemental Table I) (39). Membrane CCR9 protein expression was only detectable on a fraction of cells within this B cell subset (7.9% of CD27−IgA+ memory B cells; Fig. 2A). We conclude that CD27−IgA+ B cells contain clones with the capacity to home to the intestinal tract.
Distinct Ig gene repertoires in CD27+IgA+ and CD27−IgA+ B cells
The distinct maturation pathways of CD27+IgA+ and CD27−IgA+ memory B cells were barely reflected in their transcription program. However, previous observations indicated that these two IgA+ memory B cell subsets harbored distinct Ig repertoires (25). To study the Ig repertoire and reactivity of CD27+IgA+ and CD27−IgA+ memory B cells, we single-cell purified these from the blood of five healthy donors.
The IGHV subgroup and IGHJ gene distributions in CD27+IgA+ and CD27−IgA+ memory B cells were similar to each other (Fig. 3A, Supplemental Table II) (40), as well as to previously reported naive B cells (30). More than half of the cells used a member of the large IGHV3 subgroup, followed by IGHV4 (∼16%) and IGHV1 (∼10% in CD27+IgA+ B cells, ∼20% in CD27−IgA+ B cells). In both subsets, IGHV3-23, IGHV3-30, and IGHV3-33 were the most frequent IGHV3 genes, and IGHV4-59 was the most frequent IGHV4 gene. IGHJ4 was the most often used IGHJ gene, followed by IGHJ5 and IGHJ6. The IgH-CDR3 regions in rearrangements from both IgA+ subsets showed similar length distributions, with the majority of regions 10–14 aa in size, and they did not differ with regard to the content of positively charged amino acids (arginine, lysine, and histidine) (Fig. 3B). CD27−IgA+ B cells carried an average of 10 mutations, and ∼15% of sequences were unmutated. In contrast, nearly all IGHV genes of CD27+IgA+ memory B cells were mutated and contained an average of 19 mutations (Fig. 3C) (25). Notably, IgA1 transcripts showed a large difference in SHM levels between the two subsets (data not shown) (40). Despite difference in SHM levels, mutations were normally targeted to hypermutable motifs and appeared to be properly selected during immune responses, as reflected by a high ratio of replacement to silent mutations ≥ 3.2 in their CDRs (Supplemental Fig. 1A, 1B, Supplemental Table III).
IgH gene repertoire and characteristics in CD27+IgA+ and CD27−IgA+ memory B cells. (A) IGHV subgroup (upper panels) and IGHJ gene (lower panels) usage in IgA+ B cell subsets. The numbers of analyzed sequences are indicated in the center circles. (B) IgH-CDR3 length (upper panel) and charge (lower panels) distributions. (C) The number of SHMs in rearranged IGHV genes. Each gray dot represents an individual sequence; black horizontal lines represent median values. Data were analyzed with the χ2 test (A and B) or the Mann–Whitney U test (C). ****p < 0.0001.
IgH gene repertoire and characteristics in CD27+IgA+ and CD27−IgA+ memory B cells. (A) IGHV subgroup (upper panels) and IGHJ gene (lower panels) usage in IgA+ B cell subsets. The numbers of analyzed sequences are indicated in the center circles. (B) IgH-CDR3 length (upper panel) and charge (lower panels) distributions. (C) The number of SHMs in rearranged IGHV genes. Each gray dot represents an individual sequence; black horizontal lines represent median values. Data were analyzed with the χ2 test (A and B) or the Mann–Whitney U test (C). ****p < 0.0001.
The Igκ repertoire did not differ between the two IgA+ memory B cell subsets and showed predominant IGKV1 and IGKV3 subgroup and IGKJ1 and IGKJ4 gene usage. Still, CD27−IgA+ memory B cells showed increased Igλ usage (Supplemental Table II) (25, 41) and differed in the Igλ repertoire from CD27+IgA+ B cells (Fig. 4A) (30). Although almost 90% of IGL rearrangements in CD27+IgA+ cells involved the IGLV1 or IGLV2 subgroups, the rearrangements in CD27−IgA+ B cells reflected the pattern of mature naive B cells, with 29% IGLV3 and 29% IGLV2 subgroup usage (Fig. 4B) (42, 43). The increased IGLV3 usage in CD27−IgA+ B cells was primarily caused by abundant IGLV3-1 gene (20% versus 0% in CD27+IgA+ cells) (Fig. 4C). Further analysis of ∼550 IGLV3 rearrangements bulk amplified from seven additional donors revealed that IGLV3-1 usage within IGLV3 genes was similar between CD27+IgA+ and CD27−IgA+ B cells (Fig. 5). Thus, the observed increase in single-cell–sorted CD27−IgA+ B cells may parallel the overall increase in IGLV3 usage by these B cells (Fig. 4B). In addition, CD27−IgA+ B cells showed less usage of IGLJ1 than did CD27+IgA+ B cells (17% versus 38%) (Fig. 4B).
Igκ and Igλ L chain gene repertoire and characteristics in CD27+IgA+ and CD27−IgA+ memory B cells. (A) IGKV subgroup (upper panels) and IGKJ gene (lower panels) use in IgA+ memory B cell subsets. The numbers of analyzed sequences are indicated in the center circles. (B) IGLV subgroup (upper panels) and IGLJ gene (lower panels) use. (C) IGLV gene use. (D) The number of SHMs in rearranged IGKV (top panel) and IGLV (middle panel) in memory B cell subsets, as well as in IGLV3-1 and non–IGLV3-1 genes of CD27−IgA+ memory B cells (bottom panel). Each gray dot represents an individual sequence, and horizontal black lines represent median values. Data were analyzed with the χ2 test (A and B) or the Mann–Whitney U test (D). ***p < 0.001, ****p < 0.0001.
Igκ and Igλ L chain gene repertoire and characteristics in CD27+IgA+ and CD27−IgA+ memory B cells. (A) IGKV subgroup (upper panels) and IGKJ gene (lower panels) use in IgA+ memory B cell subsets. The numbers of analyzed sequences are indicated in the center circles. (B) IGLV subgroup (upper panels) and IGLJ gene (lower panels) use. (C) IGLV gene use. (D) The number of SHMs in rearranged IGKV (top panel) and IGLV (middle panel) in memory B cell subsets, as well as in IGLV3-1 and non–IGLV3-1 genes of CD27−IgA+ memory B cells (bottom panel). Each gray dot represents an individual sequence, and horizontal black lines represent median values. Data were analyzed with the χ2 test (A and B) or the Mann–Whitney U test (D). ***p < 0.001, ****p < 0.0001.
Usage and characteristics of IGLV3-1 gene. (A) Frequency of IGLV3-1 gene in IGLV3 subgroup in mature naive, CD27+IgA+, and CD27−IgA+ B cells from seven independent donors. IGL rearrangements were amplified with forward IGLV3-specific and two reverse IGLJ1- and IGLJ2/3-specific primers. In total, 557 unique productive rearrangements were analyzed. (B) Frequency of SHM in rearranged IGLV3-1 and non–IGLV3-1 genes in mature naive, CD27+IgA+, and CD27-IgA+ memory B cells from seven donors. Each black dot represents an individual sequence, and black horizontal lines represent median values. (C) Selection of SHMs in IGLV3-1 and other IGLV3 genes in CD27+IgA+ B cells (upper panel) and CD27−IgA+ B cells (lower panel), as measured with the BASELINe program (http://selection.med.yale.edu/baseline). *p < 0.05, ****p < 0.0001, Mann–Whitney U test.
Usage and characteristics of IGLV3-1 gene. (A) Frequency of IGLV3-1 gene in IGLV3 subgroup in mature naive, CD27+IgA+, and CD27−IgA+ B cells from seven independent donors. IGL rearrangements were amplified with forward IGLV3-specific and two reverse IGLJ1- and IGLJ2/3-specific primers. In total, 557 unique productive rearrangements were analyzed. (B) Frequency of SHM in rearranged IGLV3-1 and non–IGLV3-1 genes in mature naive, CD27+IgA+, and CD27-IgA+ memory B cells from seven donors. Each black dot represents an individual sequence, and black horizontal lines represent median values. (C) Selection of SHMs in IGLV3-1 and other IGLV3 genes in CD27+IgA+ B cells (upper panel) and CD27−IgA+ B cells (lower panel), as measured with the BASELINe program (http://selection.med.yale.edu/baseline). *p < 0.05, ****p < 0.0001, Mann–Whitney U test.
Similar to IGHV, CD27−IgA+ B cells carried significantly fewer SHMs in the expressed IGKV and IGLV genes than did CD27+IgA+ B cells (Fig. 4D), with slightly more mutations in IGLV than in IGKV in CD27+IgA+ memory B cells. Analogous to IGHV, CD27+IgA+ and CD27−IgA+ B cells showed similar selection of SHMs in their IGKV and IGLV genes, with high CDR replacement/silent ratios ranging from 2.8 to 3.4 (data not shown) and positive selection for replacement mutations (Supplemental Fig. 1B, 1C). Interestingly, the IGLV3-1 genes contained significantly fewer mutations than did other IGLV3 genes (Figs. 4D, 5B). These mutations in IGLV3-1 showed particularly strong selection against replacement mutations in framework regions (FRs), whereas selection in CDR was comparable between IGLV3-1 and other IGLV3 genes (Fig. 5C). Hence, the decreased mutation loads in IGH, IGK, and IGL, combined with the increased Igλ usage, in CD27−IgA+ B cells, suggest that these B cells were generated toward specific Ags through selection mechanisms that appear distinct from those shaping CD27+IgA+ B cells.
CD27−IgA+ B cells express Abs with distinct self-reactive features
To determine whether the Ig repertoire differences between CD27+IgA+ and CD27−IgA+ memory B cells reflect distinct Ab reactivity, we cloned and expressed the amplified Ig H and L chain genes as rAbs. Although not all Ig genes could be expressed in vitro, the repertoire of successfully produced rAbs was representative of the total set with regard to their Ig gene usage, SHM frequency and selection, and IgH-CDR3 characteristics. We tested Ab self-reactivity by immunofluorescence and ELISA against the human larynx carcinoma cell line HEp-2. We found that CD27+IgA+ B cells expressed significantly more HEp-2–reactive Abs than did their CD27−IgA+ counterparts, representing 35 and 26% of the clones, respectively (p < 0.01; Fig. 6A, 6B). In CD27−IgA+ B cells, the autoreactive Abs used IGHV1 more frequently, but otherwise these did not differ from nonautoreactive Abs with regard to SHM numbers or IgH-CDR3 length and amino acid composition (data not shown). Autoreactive CD27+IgA+, but not CD27−IgA+, B cells showed limited selection for replacement mutations in CDRs (Supplemental Fig. 1C, 1D). Immunofluorescence analyses revealed that 11.4% (10/88) of CD27+IgA+ memory B cells and only 5.3% (4/75) of CD27−IgA+ memory B cells reacted with cytoplasmic and/or nuclear Ags (Fig. 6C, 6D). In addition, CD27−IgA+ B cells were virtually devoid of anti-nuclear clones, whereas these represent 4% of CD27+IgA+ B cells (Fig. 6C, 6D). Thus, CD27−IgA+ memory B cells express Abs with decreased reactivity against cellular Ags compared with those produced by CD27+IgA+ B cells.
Autoreactivity in CD27+IgA+ and CD27−IgA+ memory B cell subsets. (A) Reactivity to HEp-2 cell lysates of 88 Abs from CD27+IgA+ cells and 75 Abs from CD27−IgA+ cells derived from five healthy donors. The dashed line represents the highly reactive ED45 positive control (31), and the red line represents the cut-off value of 0.5, above which Abs were considered HEp-2 reactive. (B) Frequencies of HEp-2–reactive CD27+IgA+ and CD27−IgA+ cells in five donors. Data points representing values for the same donor are connected with a black line. **p < 0.01, two-tailed Student t test for paired samples. (C) Autoreactivity patterns of CD27+IgA+ and CD27−IgA+ cells, as measured in an immunofluorescence assay with HEp-2 cell–coated slides. The numbers of analyzed Abs are indicated in the center. Data were analyzed with the χ2 test. (D) Representative pictures from immunofluorescence assay staining (original magnification ×40).
Autoreactivity in CD27+IgA+ and CD27−IgA+ memory B cell subsets. (A) Reactivity to HEp-2 cell lysates of 88 Abs from CD27+IgA+ cells and 75 Abs from CD27−IgA+ cells derived from five healthy donors. The dashed line represents the highly reactive ED45 positive control (31), and the red line represents the cut-off value of 0.5, above which Abs were considered HEp-2 reactive. (B) Frequencies of HEp-2–reactive CD27+IgA+ and CD27−IgA+ cells in five donors. Data points representing values for the same donor are connected with a black line. **p < 0.01, two-tailed Student t test for paired samples. (C) Autoreactivity patterns of CD27+IgA+ and CD27−IgA+ cells, as measured in an immunofluorescence assay with HEp-2 cell–coated slides. The numbers of analyzed Abs are indicated in the center. Data were analyzed with the χ2 test. (D) Representative pictures from immunofluorescence assay staining (original magnification ×40).
To determine the multispecificity of rAbs, we assessed the frequencies of polyreactive clones that recognized three structurally distinct Ags: dsDNA, insulin, and LPS (30). We found that CD27−IgA+ B cells contained significantly more polyreactive clones, which averaged 26% compared with 16% in CD27+IgA+ memory B cells (Fig. 7). Polyreactive Abs often used members of the IGHV1 subgroup, mainly at the expense of IGHV4 (Fig. 7C). The presence of positively charged amino acids in IgH-CDR3 was not different between polyreactive and nonpolyreactive clones, whereas polyreactive Abs were significantly enriched in very long IgH-CDR3s (≥20 aa) (9% in polyreactive versus 1% in nonpolyreactive cells; Fig. 7D). Interestingly, polyreactive Abs from CD27−IgA+ B cells harbored more IGHV SHMs than did nonpolyreactive CD27−IgA+ clones, at a level similar to that of CD27+IgA+ memory B cells (Fig. 7E). However, selection of SHM was similar between polyreactive and nonpolyreactive clones (Supplemental Fig. 1C, 1D, Supplemental Table III).
Polyreactivity in CD27+IgA+ and CD27−IgA+ memory B cell subsets. (A) Abs showing triple reactivity against dsDNA (top panels), insulin (middle panels), and LPS (bottom panels) were defined as polyreactive. In total, 88 Abs from CD27+IgA+ memory B cells and 75 Abs from CD27−IgA+ memory B cells were analyzed. Dashed lines represent the highly reactive ED45 positive control (31). (B) Frequencies of polyreactive CD27+IgA+ and CD27−IgA+ Abs in five donors. Data points representing values for one donor are connected with a black line. (C) IGHV subgroup (upper panels) and IGHJ gene (lower panels) use in polyreactive and nonpolyreactive IgA+ memory B cells. The numbers of analyzed sequences are indicated in the center circles. (D) IgH-CDR3 length (upper panel) and charge (lower panel) distributions. (E) The numbers of SHMs in rearranged IGHV genes. Each gray dot represents an individual sequence, and the horizontal black lines represent median values. Data were analyzed with the two-tailed Student t test for paired samples (B), the χ2 test (C), or the Mann–Whitney U test (E). *p < 0.05.
Polyreactivity in CD27+IgA+ and CD27−IgA+ memory B cell subsets. (A) Abs showing triple reactivity against dsDNA (top panels), insulin (middle panels), and LPS (bottom panels) were defined as polyreactive. In total, 88 Abs from CD27+IgA+ memory B cells and 75 Abs from CD27−IgA+ memory B cells were analyzed. Dashed lines represent the highly reactive ED45 positive control (31). (B) Frequencies of polyreactive CD27+IgA+ and CD27−IgA+ Abs in five donors. Data points representing values for one donor are connected with a black line. (C) IGHV subgroup (upper panels) and IGHJ gene (lower panels) use in polyreactive and nonpolyreactive IgA+ memory B cells. The numbers of analyzed sequences are indicated in the center circles. (D) IgH-CDR3 length (upper panel) and charge (lower panel) distributions. (E) The numbers of SHMs in rearranged IGHV genes. Each gray dot represents an individual sequence, and the horizontal black lines represent median values. Data were analyzed with the two-tailed Student t test for paired samples (B), the χ2 test (C), or the Mann–Whitney U test (E). *p < 0.05.
We conclude that CD27+IgA+ and CD27−IgA+ B cells express Abs with different self-reactive features; although CD27+IgA+ B cells are enriched in autoreactive clones recognizing cellular components, CD27−IgA+ B cells often produce polyreactive Abs that are rarely cross-reactive with self and displaying a distinct Ig repertoire associated with increased SHMs compared with their nonpolyreactive counterparts.
Polyreactive Abs from IgA+ memory B cells display anti-bacteria reactivity
Because IgA+ B cells are assumed to produce Abs protecting mucosa, we tested the reactivity of rAbs expressed by CD27+IgA+ and CD27−IgA+ memory B cells with specific micro-organisms, including commensal bacteria B. fragilis, E. cloacae, and E. faecalis, potentially pathogenic C. difficile, E. coli, and S. aureus, as well as bacterial flagellin. Irrespective of the tested Ags, the frequencies of reactive Abs were higher in CD27−IgA+ memory B cells than in CD27+IgA+ memory B cells (Fig. 8A). CD27−IgA+ cells showed high binding frequencies for commensal B. fragilis and E. faecalis strains, as well as potentially pathogenic E. coli and C. difficile, and a slightly lower frequency of Abs binding to S. aureus and E. cloacae. The frequencies of bacteria-reactive Abs were consistently higher in CD27−IgA+ B cells than in CD27+IgA+ cells for all six analyzed stains (Fig. 8B). Furthermore, nearly all of the strongly bacteria-reactive Abs were polyreactive (Fig. 8C). Thus, polyreactive Abs enriched in CD27−IgA+ B cells recognize mucosa-associated bacteria.
Reactivity of CD27+IgA+ and CD27−IgA+ memory B cells with commensal and potentially pathogenic bacteria. (A) Pie charts summarizing the frequencies of Abs from CD27+IgA+ and CD27−IgA+ memory B cells with reactivities against flagellin, commensal bacteria (B. fragilis, E. cloacae, E. faecalis), and potentially pathogenic bacteria (E. coli, S. aureus, C. difficile). The numbers of analyzed Abs are indicated in the center circles. (B) Overall bacteria reactivities of CD27+IgA+ and CD27−IgA+ memory B cells. (C) Reactivity levels against bacteria, as measured by ELISA for Abs from CD27+IgA+ and CD27−IgA+ memory B cells. Each gray dot represents a nonpolyreactive Ab, and each white dot represents a polyreactive Ab. Black horizontal lines represent the median values of all Abs together, and the dashed line represents a threshold value above which Abs were considered bacteria reactive. Data were analyzed with the χ2 test (A) or the two-tailed Student t test for paired samples (B). **p < 0.01.
Reactivity of CD27+IgA+ and CD27−IgA+ memory B cells with commensal and potentially pathogenic bacteria. (A) Pie charts summarizing the frequencies of Abs from CD27+IgA+ and CD27−IgA+ memory B cells with reactivities against flagellin, commensal bacteria (B. fragilis, E. cloacae, E. faecalis), and potentially pathogenic bacteria (E. coli, S. aureus, C. difficile). The numbers of analyzed Abs are indicated in the center circles. (B) Overall bacteria reactivities of CD27+IgA+ and CD27−IgA+ memory B cells. (C) Reactivity levels against bacteria, as measured by ELISA for Abs from CD27+IgA+ and CD27−IgA+ memory B cells. Each gray dot represents a nonpolyreactive Ab, and each white dot represents a polyreactive Ab. Black horizontal lines represent the median values of all Abs together, and the dashed line represents a threshold value above which Abs were considered bacteria reactive. Data were analyzed with the χ2 test (A) or the two-tailed Student t test for paired samples (B). **p < 0.01.
Discussion
We reported in this article the molecular characterization of unconventional CD27−IgA+ B cells, which share a common gene-expression profile with conventional CD27+IgA+ and IgG+ memory B cells, although they express a unique Ig repertoire favoring anti-commensal reactivity.
Despite their postulated origin from TI immune responses, CD27−IgA+ memory B cells displayed a gene-expression profile typical of other memory B cells. All memory subsets differed from naive B cells with regard to upregulation of costimulatory molecules and downregulation of BCR signaling inhibitors and naive B cell–specific transcription factors (34). These changes are thought to underlie the increased B cell responsiveness of memory B cells, which is crucial for adaptive immunity (34). In addition, the increased expression of IL6R and IL10R shared among memory B cell subsets promotes the synergistic effects of IL-6 and IL-10 to induce terminal B cell differentiation into plasma cells (44–46). RUNX2 was expressed exclusively by IgA+ B cells, and its expression level was significantly higher (2.4-fold) in CD27−IgA+ B cells than in CD27+IgA+ B cells. RUNX2 acts downstream of the TGF-β and RA signaling pathways to induce TI class-switching toward IgA in the gut and promote the terminal differentiation of IgA+ B cells into plasma cells (19, 20, 35). Hence, the specific expression of RUNX2 in IgA+ memory B cells further supports this scenario regarding their generation. CCR9 mediates homing to the gut, and CCR9 transcripts were specifically expressed in CD27−IgA+ B cells, and CCR9 protein was detected on the cell surface of some of these cells (14). Given the function of CCR9 to direct B cell migration toward mucous membranes (38, 39, 47), it is possible that protein expression is downregulated in circulating CD27−IgA+ B cells to allow their trafficking to other mucosal locations (48, 49). Altogether, our data indicate that human memory B cell subsets share a common genetic program and that RUNX2 and CCR9 more clearly define CD27−IgA+ memory B cells and likely play an important role in mucosal immunity.
We further analyzed the Ig H and L chain gene repertoires and confirmed that CD27−IgA+ B cells more frequently expressed Igλ than did CD27+IgA+ B cells and carried significantly fewer SHMs in their Ig (25). This could be a direct consequence of a TI origin associated with limited expression of AID (50). However, it is possible that not all IgA functions in the gut require extensive affinity maturation (51). It was postulated that long coevolution between host and microbiome may favor the selection of Ig variable genes that, in germline configuration, could recognize conserved bacterial Ags (52). Multiple studies showed that unmutated IgA isolated from mice colonized with a single bacterial strain can bind Ags with good affinity (3, 53). The distinct Igλ repertoire expressed by CD27−IgA+ B cells associated with their increased anti-commensal reactivity may also support this hypothesis. Indeed, IGLV3-1 genes from CD27−IgA+ memory B cells carried very few SHMs and showed a very strong selection against replacement mutations in FRs. The abundance of IGLV3-1 was described in diverse tissue-related conditions, such as amyloidosis (54). Thus, differences in the IGL repertoire of CD27−IgA+ B cells may reflect their involvement in local mucosal responses.
The increased usage of Igλ L chains in the intestinal tract was proposed to result from ongoing receptor-revision processes (41, 55). Our sequence analyses suggest otherwise for the following reasons: first, CD27−IgA+ B cells are enriched in IGLV3-1 genes, which are located downstream of the Igλ locus (41) and, in our study, contained nontemplated nucleotides, reflecting activity of the TdT enzyme. Because TdT is primarily expressed in B cell progenitors, we conclude that Igλ gene rearrangements in CD27−IgA+ memory B cells were generated during early B cell development in the bone marrow. Hence, the increased frequencies of Igλ L chains in CD27−IgA+ B cells are most likely the result of selection processes rather than receptor revision.
Differences in Ig repertoire between CD27−IgA+ and CD27+IgA+ B cell subsets were associated with differences in Ab reactivity. Because all of the studied Abs were expressed as IgG, we only took into account their Ag specificity and affinity, as mediated by the V region, and not avidity and effector functions associated with the constant regions. Previous studies reported that human CD27+IgG+ B cells displayed an increased frequency of autoreactive clones (47%) compared with mature naive B cells (20%) (56, 57). Similarly, we found that CD27+IgA+ B cells also expressed a higher frequency of HEp-2–reactive clones (35%). In contrast, CD27−IgA+ B cells contained fewer autoreactive clones (26%) and, therefore, were more similar to mature naive B cells (30). The lower frequency of autoreactive Abs in the CD27−IgA+ subset compared with CD27+IgA+ and CD27+IgG+ memory B cells may be explained by their low SHM levels. Autoreactivity in CD27+IgG+ memory B cells was shown to result from the introduction of SHMs (56). Although autoreactive and nonautoreactive CD27−IgA+ B cells had similarly mutated Ig genes, reduced AID activity induced by TI responses in these B cells might provide a basis for the generation of fewer autoreactive Abs than in conventional memory B cells.
The frequency of polyreactive Abs expressed by CD27−IgA+ B cells was significantly higher than in CD27+IgA+ B cells and comparable with those previously observed for intestinal IgA+ and IgG+ plasmablasts (21). This high Ab polyreactivity may be beneficial for responses toward gastrointestinal bacteria because polyreactive Abs generated from a specific immune response could serve as natural Abs that recognize other bacterial strains, even upon a first encounter. Interestingly, polyreactive anti-bacteria clones in CD27−IgA+ B cells display higher SHM frequencies than do their nonpolyreactive CD27−IgA+ counterparts, further supporting the importance of SHM in the generation of specific immunity. The role of SHM in the introduction of polyreactivity also was observed in several studies in mice and humans that demonstrated that reversion of mutated Ig genes to their germline counterparts results in a significant decrease in polyreactivity (42, 56). Moreover, mice carrying a mutated form of AID that allows CSR, but not SHM, display severe intestinal lymphoid hyperplasia accompanied by overwhelming microbial expansion in the mucosa (4). Thus, the production of mutated and Ag-selected, although polyreactive, Abs by CD27−IgA+ B cells may play an important role in controlling intestinal microbiota in humans.
In summary, we showed that CD27+IgA+ and CD27−IgA+ memory B cell subsets contain highly similar transcription profiles, despite their postulated origin from TD and TI immune responses, respectively. However, the distinct Ab repertoire and reactivity of these IgA+ memory B cell subsets reflect their unique physiological functions, with CD27−IgA+ B cells likely being involved in the maintenance of gut homeostasis through the production of polyreactive mutated IgA Abs with cross-reactive antibacterial reactivity.
Acknowledgements
We thank Dr. L. Devine, C. Wang, and S.J.W. Bartol for cell sorting and Dr. J.P. Hays for support with bacterial cultures.
Footnotes
This work was supported by Grants AI071087, AI082713, AI095848, and AI061093 from the National Institutes of Health-National Institute of Allergy and Infectious Diseases (to E.M.), as well as by a fellowship from the Ter Meulen Fund–Royal Netherlands Academy of Sciences (to M.A.B.) and fellowships from the Erasmus University Rotterdam and Erasmus MC (to M.C.v.Z.).
The microarray data presented in this article have been submitted to ArrayExpress under accession numbers E-MEXP-3767 and E-MTAB-3637.
The online version of this article contains supplemental material.
References
Disclosures
The authors have no financial conflicts of interest.