B-1 B Cell–Derived Natural Antibodies against N-Acetyl-d-Glucosamine Suppress Autoimmune Diabetes Pathogenesis Free

This article is featured in Top Reads, p. 1253

Although the notion that environmental factors contribute to precipitation of type 1 diabetes (T1D) in genetically susceptible individuals is a widely cited paradigm, mechanistic understanding of this phenomenon is limited. One intriguing possibility, outlined by the “hygiene hypothesis,” suggests that exposure to environmental Ags and microorganisms during critical stages of immune system development suppresses the development of autoimmunity (1). Consistent with this possibility, housing cleanliness and the commensal microbiota exhibit a well-documented impact on T1D penetrance in the commonly studied NOD mouse model (2). Multiple prior studies additionally reported decreased T1D incidence in diabetes-prone rodents after treatment with Streptococcus pyogenes (group A Streptococcus [GAS]) (3–5), supporting specific involvement of particular infectious microorganisms on influencing T1D onset in these models. Although mechanistic understanding of these observations is limited, retrospective epidemiological studies have similarly reported a reduced risk for T1D in humans after childhood cases of scarlet fever, highlighting the potential relevance of host–GAS interactions on human T1D pathogenesis (6, 7).

The group-specific cell-wall carbohydrate expressed by S. pyogenes Lancefield group A carbohydrate (GAC) elicits clonally restricted N-acetyl-d-glucosamine (GlcNAc)-specific B lymphocyte responses in both mice and humans (8, 9). Although not previously examined in the context of T1D protection, GlcNAc-specific Abs induced by GAS exhibit reactivity for carbohydrate epitopes generated through posttranslational glycosylation of host proteins (9), including the dynamic O-GlcNAc modifications abundant in pancreatic β cells, where they are vital in nutrient sensing (10). Although posttranslational modification (PTM)-associated neoepitopes have been suggested to drive loss of self-tolerance in T1D and other autoimmune diseases (11), we asked whether GlcNAc-specific B cell responses are involved in delaying development of T1D in NOD mice after exposure to GAS.

C57BL/6J, B6.129S7-Rag1^tm1Mom/J (C57BL/6J Rag1 knockout [KO]), NOD/ShiLtJ, and NOD/ShiLtJ.BDC2.5 TCR transgenic mice were purchased from Jackson Laboratories, and B6-H2^g7 congenic mice were generously provided by Dr. H. Tse (University of Alabama at Birmingham [UAB], Birmingham, AL). All animal use was in accordance with Institutional Animal Care and Use Committee protocols; mice were given standard chow and acidified water ad libitum. Heat-killed, pepsin-treated vaccine stocks prepared as previously described (12) and 5 × 10⁷ bacteria or 50 μg OK-432, a lyophilizate of penicillin-treated Streptococcus pyogenes A3su (Picibanil; Chugai Pharmaceutical), were administered i.p. to NOD mice at 14 d of age. Litters derived from individual dams were randomized with respect to immunization to reduce litter-dependent effects during T1D incidence studies. Diabetes incidence was monitored in female NOD mice from 10 to 30 wk of age by weekly measurements of glycosuria (Bayer Diastix); mice were classified as diabetic and were sacrificed after two consecutive urine glucose measurements of ≥1/4% and a blood glucose measurement of ≥250 mg/dl measured with an Accu-Chek Comfort Curve glucose meter (Bayer). HGAC78, HGAC41, and HGAC39 (13, 14) were generously provided by Dr. M. Nahm (UAB), CTD110.6 (15) was provided by Drs. G. Hart (Johns Hopkins University) and M. A. Accavitti (Hybridoma Core Facility, UAB), RL-2 (16) was provided by Dr. L. Gerace (Johns Hopkins University), and Dr. J. Thomas (Vanderbilt University) provided the insulin-specific hybridoma (clone 7CD9). Hybridomas were grown in serum-free media (Hybridoma SF media; Life Technologies), and culture supernatants were passed over protein G–Sepharose or GlcNAc–Sepharose columns for purification of mAbs and concentrated using AmiconUltra centrifugal filters (Ultracel-10K; Millipore). Appropriate m.w., Ag binding, and endotoxin levels were confirmed for purified mAbs by SDS-PAGE, ELISA, and limulus assay (Limulus Amebocyte Lysate Pyrogent; Lonza), respectively, and concentrations of mAb preparations were determined by spectrophotometric analysis (ND-1000; NanoDrop Technologies) before storage at 4°C. In some cases, mAbs were directly conjugated to Alexa Fluor dyes (Life Technologies). C57BL/6J-derived MIN6 insulinoma cells were obtained from Dr. H. Tse and maintained in DMEM supplemented with 10% FCS. MIN6 cells are not represented in the database of commonly misidentified cell lines maintained by International Cell Authentifcation Committe; they were authenticated through analysis of insulin expression and were free of Mycoplasma contamination as determined through luminescent mycoplasma detection (MycoAlert; Lonza).

Half-area (ELISA) or full-area (ELISPOT) 96-well flat-bottom EIA/RIA plates (Costar) were coated overnight at 4°C with 2 µg/ml GAC (5S PG-PS; BD Lee Labs), GlcNAc₃₅BSA (Pyxis Laboratories), or Ig isotype-specific capture Abs (goat anti-mouse; Southern Biotechnology). Plates were blocked with PBS + 2% BSA and after adsorption of diluted sera, serum Ig was detected by alkaline phosphatase–conjugated mouse Ig isotype-specific secondary Abs (goat; Southern Biotechnology) and phosphatase substrate reactions. Purified mAbs were used as standards for quantitation of serum Ig, and plate absorbance was read at 405 nm (SPECTROstar Omega; BMG Labtech). ELISPOT assays to enumerate Ab-secreting cells (ASCs) were blocked with PBS + 2% gelatin before incubation of spleen and bone marrow (BM) cells overnight at 37°C in RPMI + 10% FCS. Spot detection was facilitated by alkaline phosphatase–conjugated mouse Ig isotype-specific secondary Abs (goat; Southern Biotechnology) and 5-bromo-4-chloro-3-indolyl phosphate substrate reactions.

For whole-pancreas histology, pancreata were dissected from the abdominal cavity after euthanasia and whole-body perfusion with PBS + 1× heparin, flash frozen in optimal cutting temperature medium (Tissue-Tek; Sakura Finetek), and stored at −80°C until sectioning. Six-micrometer sections were cut using a cryostat at −20°C (Leica CM1850, Laboratory of Dr. F. Lund, UAB) and air dried before storage at −80°C. Tissue sections were fixed with either PBS + 0.5% paraformaldehyde and 0.5% Triton X-100 or ice-cold acetone, and nonspecific binding was blocked using PBS + 5% normal horse serum or PBS + 2% BSA. Spontaneous insulitis was examined at 10- to 12-wk-old prediabetic time points, and histological scoring of islet lesions was completed by a third party in a blind manner on a scale of 0–3 (0: no infiltrate; 1: peri-insulitis; 2: progressed peri-insulitis with breakdown of peri-islet membrane; 3: advanced insulitis) after immunofluorescence labeling of laminin (rabbit polyclonal; Novus Biologicals), CD4 (clone RM4-5; BioLegend) and IgM (goat polyclonal; ThermoFisher). Distribution of GlcNAc epitopes was evaluated using various GlcNAc-specific mAbs, wherein islets were differentiated from exocrine tissue by insulin (mAb 7CD9) and laminin staining. Detection of passively administered mAbs during high-dose STZ treatments (single 200 mg/kg dose i.p.) was accomplished with the IgH^a allotype-specific mAb RS-3.1. Complement deposition was measured with C1q-, C4- (Hycult Biotechnology), and C3b-specific (Cedar Lanes Laboratories) Abs, and signals were later quantitated using ImageJ analysis software. Mouse islets were purified as described by Leiter (17), whereas islets from other species were kindly provided by Drs. A. Thompson and L. Cui (Islet Procurement Facility, UAB). Purified islets were cytospun before staining as described earlier. Coverslips were mounted using Fluoromount, which in some cases contained DAPI for nuclear staining (Southern Biotechnology), and prepared tissues were imaged using a Leica Leitz DMRB fluorescence microscope.

Approximately 10⁹ MIN6 insulinoma cells were grown in culture with DMEM (Life Technologies) + 10% FCS (Hyclone) and postnuclear supernatant (PNS) prepared as described by Hutton et al. (18); PNS was resuspended in isotonic media, loaded onto a 8–30% continuous Nycodenz (Sigma-Aldrich) gradient, and centrifuged at 107,000 × g (Optima XPN 80k-IVD and SW 41 Ti rotor [Beckman Coulter], UAB Microbiology Department). The resulting gradient was subsequently fractionated using an automatic fractionator (Frac-920; Amersham Biosciences), and fraction activity was assessed as described previously (18), in addition to binding assays with Ag-specific mAbs.

For analysis of low-frequency Ag-specific B cells, ∼10⁷ splenocytes were pelleted in 96-well round-bottom plates by centrifugation at 1200 rpm for 2 min and blocked with PBS + 2.5% FCS and 2 μg/ml anti-CD16/32 (Ab93) (19). Non–B cells were gated out using a dump mixture containing biotinylated mAbs against CD3, CD4, CD8, CD11c, F4/80, and Streptavidin PerCP. Dead cells were excluded from analysis by propidium iodide (Sigma-Aldrich). B cell phenotypes were subsequently identified using various fluorescence-conjugated Abs. Ag-specific B cells were detected using GAC (BD Lee Labs) for single-epitope multiple staining of GAC-binding B cells (20) and were further interrogated with GAC-associated IdI-3a idiotope-specific mAb IA.1 (13). During analysis of pancreas-infiltrating B cells, pancreas tissue was digested in HBSS + 10 mM HEPES (Life Technologies) supplemented with 1 mg/ml Collagenase Type IV and 10 U/ml DNase1 (Sigma-Aldrich) after whole-body perfusion and dissection of pancreatic lymph nodes (PanLNs) and omentum. Liberated lymphocytes were isolated with lymphocyte separation media (CellGro). For flow cytometric experiments involving MIN6 insulinoma cells, apoptotic MIN6 cells were generated by X-ray irradiation (30 Gy) or treatment with 25 μM STZ and assessed for apoptosis with Annexin-V and 7AAD or active caspase-3 (all from BD Biosciences). In some cases, apoptotic MIN6 cells were labeled with a lipophilic membrane tracker (10⁻⁶ M PKH26; Sigma), and their uptake by BM-derived dendritic cells (DCs; BMDCs) was evaluated by flow cytometry after a 30-min coculture at 37°C with apoptotic MIN6 cells at 1:1 cell ratios. Data were acquired on a LSRII or FACSAria (Becton Dickinson) and subsequently analyzed using FlowJo software (Tree Star).

Bulk CD19⁺ peritoneal cavity (PerC) B cells were FACS sorted from adult naive NOD mice and adult NOD mice that had been immunized with J17A4 at day 14 (d14) at after birth. A total of 2 × 10⁵ CD19⁺ B cells from each source were administered i.p. to nondiabetic naive adult NOD mice. The mice receiving transfers together with a group of nontransferred NOD mice were monitored for 30 wk for the emergence of diabetes.

Single GAC⁺ B cells were sorted directly into 384-well plates containing 3 μl of 10 mM Tris and 0.75 U/ml RNAsinPlus (Promega). Resulting lysates were used to generate cDNA using the High-Capacity cDNA synthesis kit (Applied Biosystems) following the manufacturer’s protocol. After the generation of cDNA, rearranged BCR genes were amplified by nested PCR using IGHV family– and C region–specific primers as previously described (21). The resulting amplicons were subsequently sequenced by Sangers sequencing at the Heflin Center Genomics Core (UAB). Sequence fasta files were submitted to the IGMT server for IgH gene identification and analyzed using scripts modified from ImmuneDiversity (22) and VDJ tools (23), and R. Clonal profiles were rendered using Circos software package (14), wherein pairing (edges) of IGHV (right vertex) and IGHD (left vertex) genes, as well as IGHJ gene identities (left vertex inlay) and mutation number (right vertex inlay), was depicted.

T lymphocytes from the TCR transgenic NOD.BDC2.5 mice (15) were used to evaluate presentation of islet cell–derived Ags by BMDCs (generation described by Inaba et al. [16]). In brief, peripheral T cells were collected from secondary lymphoid tissues, purified using a negative naive T cell isolation kit (Miltenyi Biotechnology), and labeled with CellTrace CFSE (Life Technologies). A total of 5 × 10⁴ purified T cells were subsequently cocultured with C57BL/6 H-2^g7 BMDCs at a 10:1 ratio and stimulated with purified islet cells, the insulin-derived peptide (InsB_9–23) as a negative control, and the BDC2.5 mimotope. T cell activation was evaluated after 5 d of coculture and after 6-h treatment with brefeldin A, via the dilution of CFSE and intracellular staining for TNF-α (BD Biosciences).

For each experiment, the mean and SEM were determined for the respective measured parameter, and statistical analyses were performed using Student t test (single-factorial comparisons) or one-way or two-way ANOVA when appropriate (multifactorial comparisons), with Tukey’s and Bonferroni’s post hoc tests, respectively. For mouse experiments, sample sizes were not predetermined through statistical methods but were instead chosen based on pilot studies and previously reported results such that appropriate statistical testing could yield significant results. In all cases, samples were tested for similar variance to ensure the assumptions of the statistical comparisons used were met; blinding was not done in allocation of animals to experimental groups but was done for histological scoring of insulitis, which was completed by a third party. No specific randomization or exclusion criteria were applied to mouse samples because of using inbred mouse strains. Analysis was completed using GraphPad Prism software (GraphPad Software).

Program scripts used during the analysis of Ig sequencing data are available through previously published analysis software, described earlier in the IgH sequencing subsection.

GAC⁺ B cell IgH nucleotide sequence data are available through GenBank. Sequences for naive and d14 GAS-immunized NOD mice may be found at the accession numbers KY210724–KY210801 and KY210802–KY210877, respectively. Mammalian glycan microarray data are publicly available through the Consortium for Functional Glycomics gateway (http://www.functionalglycomics.org/static/index.shtml). The data that support the findings of this study are available from the corresponding author upon reasonable request.

Previous studies reported that repeated administration of the GAS preparation OK-432 to NOD mice suppressed spontaneous T1D development (3, 4). Similar to these findings, we found that a single immunization with OK-432 at 14 d of age (d14) reduced T1D penetrance in female NOD mice relative to naive controls (Fig. 1A). We further found that immunization with a heat-killed, pepsin-treated vaccine GAS preparation (strain J17A4) also protected NOD mice from T1D. Conversely, similar immunization with group C Streptococci (strain C74), which expresses a cell-wall–bearing N-acetyl-galactosamine epitope, did not impact T1D penetrance (Fig. 1B). Unlike neonatal GAS exposure, immunizing NOD mice with GAS as adults did not protect mice from T1D (Fig. 1C), suggesting that the timing of exposure was central to GAS-related efficacy in delaying T1D.

Consistent with T1D pathology, histological examination of pancreata from naive 10- to 12 wk-old prediabetic NOD mice revealed significant B and T lymphocyte infiltration, as well as a breakdown of the peri-islet membrane in most islets examined (24, 25). In marked contrast, islets of NOD mice immunized with GAS as neonates were predominantly free of lymphocytic infiltration and perturbations to the peri-islet membrane (Fig. 1D). Flow-cytometric analysis of pancreatic-tissue digests similarly revealed that fewer J17A4-immunized NOD mice developed significant lymphocytic infiltrates, relative to age-matched naive controls (Fig. 1E).

Serum ELISAs revealed that NOD mice immunized with J17A4 and OK-432 as neonates produced IgM Abs that bound both GAC- and GlcNAcylated-BSA (GlcNAc₃₅BSA) evident by 4 wk of age and remained elevated relative to control mice for >16 wk (Fig. 1F), whereas no significant increases in GlcNAc-reactive IgG Abs were observed (data not shown). ELISPOT analysis further showed that GAC-reactive IgM ASC numbers were increased in the spleen and BM of d14 J17A4-immunized mice relative to unimmunized controls (Fig. 1G). Collectively, these findings indicate specific involvement of GAC-specific B cell responses during early life in delaying T1D development.

We therefore investigated a role for B cells in dampening diabetogenesis by neonatal immunization with GAS. As we showed previously in the C57BL/6 mice (26), GAC-binding (GAC⁺) B cells were enriched in the PerC of 10- to 12-wk-old naive NOD mice and exhibited predominantly B220^loCD5⁺CD43⁺CD23⁻ (B-1a), B220^loCD5⁻CD43⁺CD23⁻ (B-1b), and to a lesser extent B220⁺CD5⁻CD43⁻CD23⁺ (B-2) B cell phenotypes (Fig. 1H). The numbers of PerC-localized GAC⁺ B cells were increased 5-fold in mice immunized with GAS (J17A4) as neonates (Fig. 1I) and were skewed toward a B-1b phenotype that exhibited increased expression of CD11b (Fig. 1H–J). Likewise, innate-like (B-1 and MZ) GAC⁺ B cells, typified by increased CD36 and CD9 expression, were increased in spleens of d14 J17A4-immunized mice (Fig. 1K).

GAC⁺ B-1b B cell expansion in NOD mice immunized with GAS as neonates was accompanied by expansion of B cell clones expressing the IdI-1 idiotope (Fig. 1K), previously associated with a subset of GlcNAc-specific IGHV6 (VH-J606)-encoded Abs (13, 26, 27). Sequencing the IgH chain genes of single FACS-sorted GAC⁺ B cells, isolated from 10- to 12 wk-old naive and d14 J17A4-immunized NOD mice, revealed that similar to other mouse strains (28, 29), GAC⁺ B cells in NOD mice were pauciclonal, >80% expressed IGHV6-3 genes, and GAC⁺ B cell IGH gene sequences were predominantly germline with little somatic mutation, consistent with their early emergence as B-1 B cells in ontogeny (Fig. 1M) (26). Relative to GAC⁺ B cells of naive mice, which expressed predominantly IGHV6-3 D1-1 J2 and IGHV6-3 D3-3 J2 rearrangements, several notable V(D)J configurations, including IGHV6-3 D2-5 J3 and IGHV6-3 D2-4 J3 BCR, were enriched in and conserved between individual d14 J17A4-immunized NOD mice (Fig. 1L). Collectively, these data indicate that neonatal immunization with GAS leads to altered clonal selection in the GAC-reactive B cell repertoire and enhances clonal expansion of otherwise low-frequency GAC-reactive B-1b B cell clonotypes.

Considering involvement of self-reactive IgM Abs in regulating autoimmunity (30–32) and the potential for GAS-specific Abs to engage endogenous GlcNAc-containing PTMs (10), we initially explored the potential for a panel of hybridoma-derived GlcNAc-specific mAbs with distinct IG-gene composition and GAC reactivity profiles to bind GlcNAc PTMs in β cells (28, 33–35). Intriguingly, we found that the GAC-reactive mAb HGAC78 (μ, IGHV6-3) exhibited robust reactivity for islets purified from human (Fig. 2A) and mouse (data not shown) pancreata. We used a mammalian glycan microarray to further probe the endogenous glycan specificity of the GAC-reactive mAbs HGAC78 and HGAC41, relative to the O-GlcNAc–specific mAbs CTD110.6 and RL2 (γ1, IGHV1–54). Each mAb exhibited exquisite specificity for GlcNAc-terminating carbohydrate structures and were particularly reactive with monomeric GlcNAc, as well as GlcNAc-β-1,4– and GlcNAc-β-1,6–terminating glycans (Fig. 2B). Importantly, most structures bound in microarray analysis represented cryptic epitopes, generally contained on the interior of mature carbohydrates. Further, there were subtle differences in the binding profiles of individual mAbs, suggesting heterogenous fine epitope specificity among the B cell clonotypes that comprise the GlcNAc-specific B cell repertoire. GlcNAc-specific Abs appeared to bind a subset of insulin⁺ granule-like puncta (Fig. 2A) in a manner strongly inhibited by adding soluble GlcNAc monomers (data not shown). Further analysis of fractionated MIN6 PNSs confirmed HGAC78 reactivity correlated closely with insulin secretory granule (ISG) abundance, as measured by ELISA with anti-insulin (clone 7CD9) and anti–glutamic acid decarboxylase 65 mAbs (Fig. 2C). Further supporting this notion, ELISA and SPR analysis revealed the O-GlcNAc–specific mAb CTD110.6 did not react with GAC-associated GlcNAc epitopes, despite exhibiting a higher affinity for GlcNAc-haptenated BSA Ags than anti-GAC clones (Fig. 2D). Interestingly, lack of apparent ISG reactivity exhibited by CTD110.6 correlated with lack of binding to GAC.

In summary, GlcNAc-specific Abs elicited by GAS immunization bound epitopes associated with ISG that most likely represented cryptic epitopes generated during autophagosomal degradation of mature carbohydrates.

We next investigated whether GlcNAc-specific Abs elicited by GAS immunization may serve to promote clearance of apoptotic β cells during pancreatic remodeling, similar to previously reported natural Ab (NAb) activity (36). We first explored this possibility by evaluating GlcNAc-specific Ab binding to MIN6 insulinoma cells that were treated with apoptosis-inducing agents, relative to live cells. HGAC78 exhibited no surface reactivity with live MIN6 cells detectable by flow cytometry, despite exhibiting robust binding to intracellular Ags after cell permeabilization (Supplemental Fig. 1A). Conversely, we observed dramatic increases in GlcNAc epitope accessibility on apoptotic MIN6 cells after irradiation (Supplemental Fig. 1B). Binding of GAC-specific mAbs HGAC41 and HGAC78 to irradiated MIN6 cells coincided with Annexin-V and activated caspase-3 staining, was inhibited by addition of soluble GlcNAc, but not GalNAc, and was superior to that of the O-GlcNAc-specific mAb CTD110.6 (Supplemental Fig. 1D).

We next examined whether anti-GlcNAc Abs engaged apoptotic β cells in vivo using a protocol of high-dose STZ treatment to synchronously induce β cell apoptosis in C57BL/6 Rag1 KO mice after passive administration of allotype-marked HGAC78 or control IgM^a mAb (Fig. 2E). Similar to our in vitro analyses, we detected robust IgM deposition in pancreatic islets of STZ-treated mice administered HGAC78, but not in HGAC78 + vehicle control–treated mice, nor in mice administered a control IgM Ab (Fig. 2F, 2G). Moreover, IgM^a deposition in HGAC78 + STZ–treated mice coincided with staining for early complement components C1q and C3b, indicating HGAC78 mediated classical complement pathway activation on apoptotic β cells (Fig. 2F). Although islet-localized C1q was clearly increased in the presence of HGAC78, C3b staining intensity did not differ between STZ-treated mice, suggesting that innate pathways mediated complement activation on apoptotic β cells in the absence of serum Abs. In further support of this possibility, flow-cytometric analysis of irradiated MIN6 cells incubated in 10% C57BL/6 Rag1 KO mouse serum revealed robust deposition of complement component C3b in the absence of detectable C1q binding. Chelation of either Ca²⁺ or Mg²⁺ abrogated C3b deposition specifically implicating the lectin-dependent and alternative complement pathways, respectively, in Ab-independent opsonization of apoptotic β cells (Supplemental Fig. 1F). Surprisingly, preopsonization of irradiated MIN6 cells with HGAC78 did not affect C1q binding (Supplemental Fig. 1G) but diminished detectable C3b deposition compared with MIN6 cells pretreated with control IgM, in which complement activity presumably stemmed from innate pathways (Supplemental Fig. 1G, 1H). We reasoned that serum IgM may compete with GlcNAc-reactive serum lectin opsonins for apoptotic β cell–associated substrates, leading to differential production of C3-derived ligands. Indeed, preincubation of MIN6 cells with 10% normal mouse serum partially inhibited HGAC78 binding to apoptosis-associated β cell epitopes, an effect that was recapitulated with recombinant GlcNAc-reactive innate serum lectins human ficolin-3 and mouse mannose binding lectin-2 (Supplemental Fig. 1I).

We next used PKH-26 membrane-tracking dye to label apoptotic MIN6 cells and analyze the effect of HGAC78 opsonization on their uptake by BMDCs in the presence or absence of complement. These results clearly demonstrate that preopsonization by HGAC78 promotes significant complement-dependent increases in PKH-26⁺ MIN6 cell–bearing BMDCs, relative to pretreatment with control IgM (Supplemental Fig. 1J, 1K). Moreover, these data imply that classical complement pathway activation by HGAC78 mediates more efficient uptake of apoptotic β cells than does C3b deposited via innate pathways.

Lastly, we sought to investigate whether passive administration of GlcNAc-specific Abs to NOD mice was able to suppress spontaneous T1D in vivo. Antisera was collected from NOD mice after a neonatal prime-adult boost immunization strategy with either GAS J17A4 or OK-432 vaccine preparations, as well as serum from naive NOD donors (Fig. 2H). Serum collected from GAS-immunized mice contained GAC-specific IgM at concentrations increased >300-fold over that of naive serum (Fig. 2I). We pooled serum from each donor group and normalized these preparations by their total IgM content, before administering six independent doses of each IgM pool to NOD pups between 7 and 21 d of age (Fig. 2J). Based on quantified concentrations of GAC-specific IgM present in these preparations, we estimated that serum preparations from naive, GAS J17A4–, and GAS OK-432–immunized mice provided 0.001, 7.5, and 12 μg of GAC-reactive Abs per treatment, respectively. However, we were unable to reproducibly demonstrate that passive administration of polyclonal serum Ab enriched for GlcNAc-specific IgM reduced T1D penetrance (Fig. 2K). Similarly, administration of the GlcNAc-specific mAb HGAC78 to naive NOD pups also did not appear to affect penetrance of T1D relative to mice treated with control IgM (Fig. 2J, lower right). Thus, although GlcNAc-reactive B cell responses to GAS appear to significantly delay T1D protection potentially by facilitating efficacious clearance of apoptotic β cells and thereby limiting T cell activation, we could not demonstrate that passive serum anti-GlcNAc Abs were sufficient to dampen spontaneous T1D in NOD mice.

To determine whether Ab opsonization impacted subsequent presentation of β cell–derived Ags, we cocultured Ag-loaded BMDCs with CFSE-labeled diabetogenic NOD.BDC2.5 TCR transgenic T cells (15, 37). Islet cell–loaded BMDCs stimulated moderate BDC2.5 T cell activation; however, preopsonization of islet cells with HGAC78 and other anti-GlcNAc mAbs significantly decreased frequencies of CFSE^loBDC2.5 T cells (Fig. 3A, 3B). As expected, BMDCs loaded with mimotope potently stimulated proliferation and TNF-α production by BDC2.5 T cells relative to those loaded with an irrelevant control peptide. By contrast, HGAC78 did not affect activation of T cells stimulated by BMDCs loaded with the BDC2.5 TCR mimotope (Fig. 3C), suggesting that HGAC78 specifically modulated the uptake and/or processing of islet cell Ags by APCs.

Other GlcNAc-specific mAbs, including HGAC41, CTD110.6, and RL2, similarly suppressed BDC2.5 T cell activation and TNF-α production. In contrast HGAC39 (γ3, IGHV6-6), which did not interact with apoptosis-associated β cells, was unable to mediate suppression of T cells toward their islet antigenic cargo HGAC39 (γ3, IGHV6-6) (Fig. 3B), associating suppressive activity with specific Ag reactivity (Fig. 3B, 3D, 3E).

The abundance of Foxp3⁺CD25⁺ regulatory T (Treg) cells, as well as the frequency of Treg cells expressing Helios in pancreatic tissue and draining PanLNs, was not influenced by neonatal immunization with GAS, indicating that neonatal immunization with GAS did not suppress T1D pathogenesis through the induction of Treg cells (Fig. 3F–H). Further analysis of 12 immunomodulating cytokines revealed no significant differences in serum concentrations between mice immunized with GAS as neonates and GCS-immunized or naive control mice (Supplemental Fig. 2).

Thus, delayed development of T1D in NOD mice after neonatal immunization with GAS is associated with the enhanced production of GlcNAc-reactive B cells and little or no apparent direct T cell–mediated suppression.

When we directly examined pancreas-infiltrating B cells in 10- to 12-wk-old prediabetic NOD mice, we found that the numbers of GAC⁺ B cells were dramatically increased in pancreatic-tissue digests and PanLNs when NOD mice were immunized with GAS J17A4 as d14 neonates relative to the inguinal lymph nodes (ILNs) (Fig. 4A). By contrast, phosphorylcholine-binding B cells, a specificity largely restricted to B-1 B cells, were not enriched in the pancreas from naive NOD mice, nor were GAC⁺ B cells detected in nondiabetic C57BL/6 mouse pancreas (Fig. 4G). Collectively, these results demonstrate that specific recruitment of GlcNAc-reactive B-1 B cells to the pancreas was associated with diabetogenesis initiation.

Unlike the pancreas-infiltrating GAC⁺ B cells present in naive NOD mice that exhibited predominantly IgD⁺ phenotypes, those present in neonatally J17A4-immunized mice were almost exclusively IgD^−/loIgM⁺ (Fig. 4A) and highly enriched in a population of IgD⁻IgM⁺Fas⁺ GAC-binding B cells (Fig. 4B, left), indicating in situ activation. The frequency of these activated GAC⁺ B cells, uniquely localized to the pancreas, was increased >20-fold by prior neonatal d14 immunization (Fig. 4B, right). Although GAC⁺ B cells in d14 GAS-immunized mice exhibited signs of activation in pancreatic tissue, we did not detect significant differentiation of pancreas-localized GAC⁺ B cells into CD138-expressing ASCs.

We sought to understand the mechanisms by which GAC⁺ B cells modulated diabetogenesis in pancreatic tissue. Relative to mice immunized with GCS via the neonatal prime:adult boost regimen, GAS-immunized mice exhibited increased GAC⁺ B cell numbers in the spleen, moderate increases in the PanLNs, and a 20-fold increase in pancreatic-tissue digests. In addition, more GAC⁺ ASCs were found in both PanLN and pancreas relative to control mice. By contrast, secondary and tertiary GAS vaccinations in the absence of perinatal priming gave rise to fewer GAC⁺ B cells and GAC⁺ plasmablasts in the pancreas and PanLNs, whereas ASC abundance within the spleens was equivalent between each mouse group (Fig. 4C, 4D). These results suggest that the decrease in diabetogenesis of NOD mice immunized with GAS as neonates correlated with the striking propensity of GAC⁺ B cells generated by neonatal B cell responses to localize in NOD mouse pancreatic tissue.

To examine this possibility further, we adoptively transferred FACS-sorted PerC B cells from d14 GAS-immunized or from naive adult NOD mouse donors i.p. to neonatal NOD recipient mice (Fig. 4E). Mice that were adoptively transferred with B cells from naive donors exhibited T1D incidence that did not differ significantly from that of unmanipulated NOD mice. However, mice that received B cells from d14 GAS-immunized donors exhibited a significantly reduced incidence of T1D by 30 wk of age (Fig. 4E). Collectively, these results indicate that the delay in T1D development after neonatal immunization with GAS is linked to enhanced localization of GAC binding B cells in the pancreas of NOD mice. Because B-1 B cells have been shown to suppress the development of autoimmune disease through the production of immunosuppressive cytokines, such as IL-10, we sought to investigate whether neonatal immunization with GAS modulated the capacity of GAC⁺ B cells to produce IL-10. Interestingly, the overall GAC⁺ B cell compartment was enriched for IL-10–producing cells relative to the bulk CD19⁺ B cell compartment, with ∼50% of GAC⁺ B cells staining for intracellular IL-10 after ex vivo stimulation. GAC⁺ B cells represented in the spleen and PerC were similar in their ability to produce IL-10. However, when we compared naive mice with those immunized with GAS as neonates, the relative frequency of IL-10–producing GAC⁺ B cells was not significantly affected by immunization (Fig. 4H). Nevertheless, because GAC⁺ B cells were expanded after neonatal immunization, it stands to reason that the overall numbers of IL-10–producing GAC⁺ B cells are increased. By extension, one could reason that because of their increased propensity to localize to the pancreas during diabetogenesis, there is a greater likelihood that they will produce IL-10 in situ.

The observations presented in this article portray, to our knowledge, a novel scenario in which exposure to exogenous microbial Ags during perinatal life stimulates T1D-inhibitory B cell clonotype development. T1D protection after early immunization with GAS was associated with significant increases in GlcNAc-reactive serum Ab levels and B-1 B cell numbers, as well as more efficient localization of GlcNAc-reactive B cells to pancreatic tissue during T1D pathogenesis. B lymphocytes play a critical role in the pathogenesis of T1D through their Ag-presenting capabilities (38, 39). Although this function is likely key to the formation of pancreas-localized ectopic lymphoid follicles (25) and the priming of T cells that ultimately kill β cells (40), our data indicate that PTM-reactive innate-like B cells may subvert this process during the early phases of T1D initiation.

Formation of immunologically relevant PTM-associated epitopes driven by constitutively high levels of ER stress in β cells may contribute to failure of peripheral tolerance during T1D pathogenesis (41). O-GlcNAc protein modifications are not only represented at uniquely high levels in β cells relative to other tissues (42), but aberrant O-GlcNAcylation is associated with β cell stress and dysfunction during T2D (43, 44). Our findings reveal a potential interplay between innate pattern recognition receptors and B cell clonotypes during immune recognition of GlcNAc modifications represented on a subset of ISGs and apoptosis-associated β cell epitopes.

Indeed, GlcNAc epitopes are known to serve as damage-associated molecular patterns to facilitate recognition of apoptotic cells by macrophages during efferocytosis (45). Perhaps of significance in this context, NOD mouse macrophages exhibit defective efferocytosis, leading to reduced clearance of senescent cells and potentiating induction of adaptive immune responses toward β cell–associated Ags (46, 47). Although waves of β cell apoptosis occurring through pancreatic remodeling during early life have long been considered an initiation point for T1D pathogenesis (40), recent reports describing early type 1 IFN signatures in T1D patients, similar to those observed during apoptotic cell–specific immune responses of SLE patients, may support this concept (48). Furthermore, Batf3-dependent DCs that drive potent Ag cross-presentation and are essential to T1D pathogenesis express a diverse array of glycan-specific innate receptors (49–51). Effective clearance of glycosylated autoantigens may therefore represent a key step in subverting T1D pathogenesis and maintaining tolerance toward β cell Ags.

Our studies indicate that GlcNAc-specific Abs act analogously to natural IgM to promote maintenance of tolerance to β cells. Although present in naive mice in the absence of deliberate immunization, early immunization with GlcNAc-bearing GAS leads to long-lasting increases in the serum concentrations of GlcNAc-specific Abs. Similar to previously reported natural IgM Abs (36), certain GlcNAc-reactive IgM Ab clonotypes engage neoepitopes generated during β cell apoptosis and initiate classical complement pathways, leading to significant increases in the phagocytosis of apoptotic β cell fragments by DCs.

Apoptosis-associated carbohydrate epitopes additionally are substrates for innate complement activation on β cells, and GlcNAc-reactive serum Abs may exist in competition with the GlcNAc-reactive serum opsonins ficolin and mannose-binding lectin. Although the differences in classical versus innate complement activation on the immunological outcome of β cell engagement remain unclear, we find that, relative to the classical complement pathway, C3b deposition driven by innate pathways is inefficient at promoting β cell uptake by BMDCs. Differential processing of C3b can drastically affect trafficking of opsonized Ags, with generation of C3dg/C3d opsonins facilitating adaptive immune responses by targeting Ags to follicular DC networks (52).

The role of complement activation in T1D therefore requires further investigation. Although often ignored because of the absence of functional C5 genes in NOD mice (53), early components of the complement system remain relevant to T1D initiation, and expression of the complement receptor from the Ig superfamily, as well as the polysaccharide-specific C4 opsonin and other complement system genes by intraislet macrophage, was recently correlated with slowing of T1D development in NOD mice (54). Human T1D patients similarly exhibit dysregulated serum concentrations of early complement components, and postmortem tissue specimens from T1D patients demonstrate clear signs of islet-localized complement activation (55, 56). Although our studies indicate that glycan epitopes serve as important ligands for innate immune system engagement, a better understanding of the driving forces and outcome of β cell–targeted complement activity is needed.

That GlcNAc-specific IgM-mediated suppression of diabetogenic CD4⁺ T cell activation in the absence of complement activity indicates an additional mechanism of immune modulation likely occurs at the point of DC Ag engagement. NAbs can attenuate inflammatory signals driven by innate receptors in DCs (57), and glycan-bearing Ags can elicit potent cell-mediated immunity through engagement of various lectin receptors on DCs (58, 59). Thus, the NAb repertoire content during early life may significantly impact immune system responsiveness to PTMs presented on T1D-associated autoantigen.

B-1 cell numbers induced by neonatal immunization were increased in the pancreas and PanLNs and were sufficient to inhibit progression of T1D upon adoptive transfer to naive NOD mice. However, neither passively administered immune sera nor anti-GlcNAc mAb were found to inhibit the progression of T1D. How can these findings be reconciled with the clear demonstration of passively administered anti-GlcNAc mAb binding to β cells in vivo after streptozotocin treatment? Interpretation of these results may be impacted by the ability of streptozotocin to induce islet blood hyperperfusion and vascular permeability (60, 61), thereby permitting access of passively administered IgM Ab to islets that may not occur during spontaneous diabetogenesis of NOD mice.

Further, we have shown that passively administered mAb HGAC78 is rapidly depleted from circulation in mice, presumably because of target-mediated disposition resulting from accessible GlcNAc epitopes, which likely precluded reaching sufficient local concentrations in pancreatic tissue to dampen T1D. In contrast, GlcNAc-specific B cells bearing IgD^loFas^hi phenotypes that enrich in pancreatic tissue may mediate the immunosuppressive effect via in situ secretion of IgM. Although we were unable to demonstrate expression of the canonical plasma cell marker CD138 by these cells, recent demonstration of important protective roles for CD138⁻ nonproliferating B-1 cells in lung in mouse models of influenza virus infection expression via their secretion of IgM highlight the likelihood that as of yet poorly understood subsets of plasma cells contribute to immunological outcomes in vivo (62). It remains possible that B-1 B cells delay development of T1D in NOD mice through other mechanisms. For example, receptors for the metabolite γ-aminobutyric acid expressed by T cells are implicated in T1D progression in the T1D-prone BB rat (63), and more recent studies have reported that B cell– or plasma cell–derived γ-aminobutyric acid can dampen CD8⁺ and CD4⁺ T cells inflammatory responses in mouse models of cancer (64).

Thus, it is possible that soluble mediators beyond Abs are produced by GlcNAc⁺ B-1 bells and are implicated in dampening of T1D in NOD mice and warrant further investigation.

Collectively, our observations indicate a significant involvement of cryptic glycan epitopes in immunological recognition of senescent β cells, and that clonal composition of the GlcNAc-specific B cell repertoire during early life can significantly influence immune responsiveness to T1D autoantigens. Early exposure to GlcNAc-bearing microorganisms may represent an important environmental factor that influences T1D progression in humans. Although maturation of GlcNAc-specific B cell responses in humans is delayed relative to initiation of T1D pathogenesis, we postulate that NAb repertoire defects precede emergence of β cell–specific autoantibodies associated with and currently used to determine the risk of T1D progression.

The authors have no financial conflicts of interest.

We thank Lisa Jia, Dr. Jeffrey Sides for technical assistance, Enid Keyser (UAB Consolidated Flow Cytometry Core) for assistance and expertise in cell sorting, and Dr. Denise Kaminski for editorial assistance. Glycan microarray analysis was completed through the Consortium for Functional Glycomics (Protein-Glycan Interaction Core, Emory University).