|Home | About | Journals | Submit | Contact Us | Français|
Effective central tolerance is required to control the large extent of autoreactivity normally present in the developing B cell repertoire. Insulin-reactive B cells are required for T1D in the NOD mouse, as engineered mice lacking this population are protected from disease. The VH125Tg model is used to define this population, which is found with increased frequency in the periphery of NOD vs. non-autoimmune C57BL/6 VH125Tg mice, but the ontogeny of this disparity is unknown. To better understand the origins of these pernicious B cells, anti-insulin B cells were tracked during development in the polyclonal repertoire of VH125Tg mice. An increased proportion of insulin-binding B cells is apparent in NOD mice at the earliest point of antigen commitment in the bone marrow. Two predominant light chains were identified in B cells that bind heterologous insulin. Interestingly, Vκ4-57-1 polymorphisms that confer a CDR3 Pro-Pro motif enhance self-reactivity in VH125Tg/NOD mice. Despite binding circulating autoantigen in vivo, anti-insulin B cells transition from the parenchyma to the sinusoids in the bone marrow of NOD mice and enter the periphery unimpeded. Anti-insulin B cells expand at the site of autoimmune attack in the pancreas and correlate with increased numbers of IFN-γ producing cells in the repertoire. These data identify failure to cull autoreactive B cells in the bone marrow as the primary source of anti-insulin B cells in NOD mice, and suggest that dysregulation of central tolerance permits escape into the periphery to promote disease.
Central tolerance is a critical barrier for the development of autoreactive B lymphocytes. Dangerous self specificities are seeded into the peripheral B lymphocyte repertoire to promote several autoimmune diseases, suggesting defective central tolerance (1–4). A better understanding of how and why central tolerance fails is necessary to enhance treatment modalities for complex autoimmune disorders. A key goal is to correct the underlying immune defects in these patients, rather than only managing symptoms after the destruction has run its course.
Developmental checkpoints exist in the bone marrow (BM) that revise BCR autoreactivity or limit BM egress and maturational progression. Antigen availability and BCR interaction strength control whether and how developing B cells are censored. Receptor editing is the primary mechanism through which autoreactivity is removed from the developing repertoire; if receptor editing fails, deletion may ensue (5). Developing B cells migrate from the BM parenchyma into the BM sinusoids as they mature and prepare for egress into the blood (6). Autoreactive B cells which undergo receptor editing show impaired transition from the BM parenchyma into the sinusoids (7). Furthermore, loss of cannabinoid receptor 2, required for sinusoidal retention, reduces the frequency of lambda+ B cells in the periphery, consistent with the hypothesis that limiting the window of time in the BM reduces receptor editing (6). These data highlight the transition between these anatomical niches as a key central tolerance checkpoint.
The NOD mouse model develops type 1 diabetes spontaneously, and mirrors many aspects of human disease. Insulin autoantibodies are predictive of T1D development in both mice and humans (8–10), suggesting that breaches in B lymphocyte tolerance occur. Insulin is a critical T1D autoantigen; limiting insulin recognition by either the T cell (11) or B cell (12) repertoires is protective against disease development in NOD mice. To better understand how autoreactive B cells are censored differently in the context of autoimmune disease, anti-insulin B cell development was investigated in NOD mice. Despite the clear presence of insulin autoantibodies in WT/NOD mice, anti-insulin B cells are difficult to reliably track in the mature WT/NOD repertoire (13). The VH125Tg model harbors an anti-insulin heavy chain which pairs with endogenous light chains. This model permits clear tracking of this important specificity within the context of a polyclonal repertoire (12). This approach thus preserves competition of autoreactive B cells with non-autoreactive B cells for survival factors, follicular entry, and other biological aspects known to impact how central tolerance is imposed (14,15). Previous studies demonstrate that whereas anti-insulin B cells are readily observed in the mature repertoire of VH125Tg/NOD mice, this specificity is reduced or absent in the mature repertoire of non-autoimmune VH125Tg/B6 mice, despite restricted usage of the same anti-insulin heavy chain (13,16). The ontogeny of this disparity is unknown, but could contain clues to how similar tolerance defects may promote autoimmune disease in humans.
Using the VH125Tg model, we show that anti-insulin B cells form with increased frequency in BM of the autoimmune-prone NOD strain, compared to the non-autoimmune B6 strain. This disparity is first evident among immature B cells present in the BM sinusoids, after they have transitioned from the parenchyma. BM culture also enhanced anti-insulin B cell formation in vitro, suggesting extrinsic factors that might be unique to the autoimmune environment in vivo do not entirely account for this difference. Germline Vκ polymorphisms that confer the potential for generation of a Pro-Pro motif in CDR3 enhance the autoreactivity of the NOD insulin-binding light chain, Vκ4-57-1. In addition, negative selection of anti-insulin BCRs is not evident in VH125Tg/NOD mice, despite insulin encounter at the earliest stages of anti-insulin B cell development. Once this central tolerance defect seeds anti-insulin B cells into the mature repertoire, anti-insulin B cells are further enriched at the site of autoimmune attack in the pancreas. The presence of anti-insulin B cells in the repertoire is associated with an increased frequency of cells that produce IFN-γ, presumably insulin-reactive T cells. These data demonstrate that “original sin” against the insulin autoantigen is traced to enhanced bone marrow production of anti-insulin B cells. The failure of central tolerance to cull this key autoimmune specificity from the developing B cell repertoire ultimately imparts critical APC to facilitate islet autoreactivity.
The anti-insulin VH125Tg [Cg-Tg(Igh-6/Igh-V125)2Jwt/JwtJ] and non-insulin-binding VH281Tg [Tg(Igh-6/Igh-V281)3Jwt/JwtJ] (The Jackson Laboratory) H chain Tg mice used in this study harbor a randomly integrated VH Tg on C57BL/6 (B6) or NOD backgrounds as described previously (12,17). Age ranges are indicated in figure legends. All data are derived from lines that have beenbackcrossed >20 generations to B6 or NOD, and which are hemizygous for all Tg indicated. All mice were housed under sterile housing conditions, and all studies were approved by the Institutional Animal Care and Use Committee of Vanderbilt University, fully accredited by the AAALAC.
BM was eluted from femurs, tibias, and humeri with HBSS (Invitrogen) + 10% FBS (HyClone). RBC were lysed using Tris-NH4Cl and cells were used for flow cytometry analysis, or resuspended at 2 × 106 cells/ml in complete culture medium: DMEM + 10% FBS + L-glutamine + HEPES + MEM sodium pyruvate + Non-essential aa (NEAA) + gentamycin + 2 × 10−5 M 2- ME + 15 ng/ml human rIL-7 (Peprotech) and cultured for 5 d in a 37°C CO2 incubator (all from Invitrogen unless otherwise specified). FBS contains fg/ml amounts of bovine insulin, which is below the threshold necessary to induce any B cell responsiveness in all assays tested. To remove IL-7, 5 d cultures were washed with HBSS + 10% FBS and were resuspended at 2 × 106 cells/ml in culture media without IL-7 and grown for an additional 2 d, at which point cells were harvested and stained for flow cytometry analysis. Spleens were harvested, macerated, and RBC were lysed. Freshly isolated pancreata were digested with 3 mL of 1 mg/mL collagenase P diluted in HBSS at 37°C for 30 min., then tissue was disrupted using an 18G needle. HBSS + 10% FBS was immediately added to inhibit collagenase activity. Cells were resuspended and used for flow cytometry analysis.
Flow cytometry analysis was performed using an LSRII (BD Biosciences). Ab reagents reactive with B220 (6B2), IgMa (DS-1), IgMb (AF6-78), CD4 (RM4-5), CD19 (1D3), CD21 (7G6), CD23 (B3B4), 7-aminoactinomycin D (7AAD), or DAPI were used for flow cytometry (BD Biosciences), or IgM (μ chain specific, Invitrogen). Human insulin (Sigma-Aldrich) was biotinylated at pH 8.0 in bicine buffer using biotin N-hydroxysuccinimide ester (Sigma-Aldrich) and detected with fluorochrome-labeled streptavidin (BD Biosciences). Insulin-specific B cells were confirmed among B220+ IgM+ live lymphocytes by competitive inhibition with 10-fold excess unlabeled insulin as well as a linear relationship of insulin-binding and IgMa expression. The percentage of insulin-specific B cells was calculated by subtracting the percentage of insulin-binding B cells in the presence of 10X inhibition with unlabeled insulin competitor (e.g. Fig 1B, right panels) from the percentage of insulin-binding B cells in the absence of unlabeled insulin competitor (e.g. Fig. 1B, left panels) to include only antigen-specific B cells in computational analyses. BCR occupancy with endogenous insulin was detected using a second anti-insulin antibody, mAb123 (10–20 μg/mL), which was biotinylated. Pre-incubation with Fc Block (2.4G2, BD Biosciences) did not impact the frequency of insulin-binding B cells (multiple experiments, not shown). mAb123 binds a distinct insulin epitope from mAb125 (from which VH125Tg is derived), and it has been successfully utilized to detect 125Tg BCR occupancy with endogenous insulin (17). These mAb do not recognize insulin bound to the hormone receptor (18). FlowJo (Tree Star, Inc.) software was used for analysis.
VH125Tg/NOD or VH125Tg/B6 mice were immunized with human or porcine insulin covalently conjugated to Brucella abortus ring test antigen (U.S. Department of Agriculture, Ames, IA) using insulin acylated with m-maleimidobenzoyl-N-hydroxysuccinimide ester (Pierce Chemical Co., Rockville, IL) and B. abortus thiolated with methyl-4-mercaptobutyrimidate (Pierce Chemical Co.) (19). This approach has been previously shown to interrogate the germline repertoire for anti-insulin B cells in Balb/c mice (19–22). Splenocytes were harvested after 3–5 d and used in a standard fusion protocol (19) with the mouse myeloma line NSO, or to facilitate capture of autoreactive B cells, NSO-BCL2 line (a gift from Dr. Betty Diamond, Feinstein Institute, Manhasset, NY) as fusion partners (23). Hybridomas were selected in hypoxanthine-aminopterin-TdR (Sigma-Aldrich) supplemented hybridoma medium: DMEM + 10% FBS + L-glutamine + HEPES + MEM sodium pyruvate + penicillin/streptomycin + 2 × 10−5 M 2-ME + NCTC-109 (Invitrogen) and production of anti-insulin Ab was verified through ELISA as described previously (19). RNA was purified from expanded clones and Vκ genes were amplified and sequenced as described below.
Rows of the same 96-well plate were coated with either 1 μg/mL human (Sigma) or rodent insulin (Novo, Bagsvaerd, Denmark) and then incubated with parallel hybridoma supernatant samples in duplicate. Anti-insulin IgMa antibody was measured as described previously (19), and wells were monitored for a human insulin O.D. of ~1.0 to normalize for differences in total antibody levels among hybridoma supernatants. To assess the level of self reactivity of hybridoma clones with rodent insulin, the “autoreactivity index” was calculated by dividing the average rodent insulin O.D. by the average human insulin O.D. read at the same time point on the same plate. A value of zero would indicate no self-reactivity, whereas a value of 1 would indicate comparable reactivity with both rodent and human insulin. ELISA duplicates were very precise (not shown).
BM cells were cultured 5 d + IL-7 (described above) to enrich for naïve developing B cells. Splenocytes were enriched for B cells by CD43+ cell depletion using anti-CD43 magnetic beads through magnetic sorting (Miltenyi). Developing (BM) or peripheral (spleen) B cells from 5 mice (8–12 wk of age) were incubated with anti-B220, anti-IgMa, DAPI, and biotinylated insulin/streptavidin, and insulin-binding B cells were purified using a FACSAria I or II cell sorter. RNA was isolated from flow cytometry-sorted insulin-binding B cells using an Ambion RNAqueous-Micro kit, or from insulin-binding hybridoma cell lines using an Ambion RNAqueous kit (Applied Biosystems). First strand cDNA was generated from total RNA using Superscript II RT (Invitrogen) and 0.67 μg oligo-dT primer (GE Healthcare) in a standard protocol. Vκ sequences were amplified from first strand cDNA using the following primers: murine Cκ primer - 5′GGA TAC AGT TGG TGC AGC ATC and murine VκA - 5′ATT GTK MTS ACM CAR TCT CCA, where K=G or T, M=A or C, S=C or G, R=A or G. Vκ sequences were amplified, cloned, and sequenced as described previously (16,24). Vκ gene segment sequence alignments (excluding sequence provided by the degenerate 5′ primer) were assigned using the ImMunoGeneTics (IMGT) database (www.imgt.cines.fr:8104/) and IgBLAST (http://www.ncbi.nlm.nih.gov/igblast/). Individual clones are each derived from independent pools of RNA, accession numbers are provided in the legends of Fig. 2, ,44–5.
Mice were injected i.v. with 1μg anti-CD19-PE (1D3, BD Biosciences) and sacrificed after 2 min. BM was immediately eluted from femurs, tibias, and humeri and cells were isolated as above. Cells were stained with antibodies reactive with cell surface markers, including anti-CD19-APC to aid in detection of sinusoidal (CD19-PE hi, CD19-APC low) versus parenchymal (CD19-PE low, CD19-APC hi) cells using flow cytometry, as described above. Specific labeling of B cells present in these anatomical niches using this method has been previously reported (6).
96-well multiscreen filter plates (Millipore) were pre-wet with 70% methanol, then washed with sterile 1X PBS (no Mg2+ or Ca2+) and coated with 10 μg/mL unlabeled anti-mouse IFN-γ (14-7313-85, eBiosciences) in 1X PBS, and incubated overnight at 4°C in the dark. Plates were washed with 1X PBS and blocked with complete RPMI culture medium: RPMI (Cellgro) containing 10% FBS, 1% L-glutamine, 1% HEPES, 0.2% Gentamycin, and 0.1% 2-ME (GIBCO) for 1h at RT. Splenocytes from VH281Tg NOD and VH125Tg NOD female mice were isolated and plated at 5×105 cells/well with or without 100 μg/mL human insulin (Sigma), or 5 μg/mL anti-mouse CD3 (hybridoma 2C11, American Type Culture Collection) as positive control and incubated for 72–96h at 37°C in a CO2 incubator. Plates were washed with 1X PBS and then with wash buffer (1X PBS/1% FBS/0.05% Tween-20), then coated with 2 μg/mL biotinylated anti-mouse IFN-γ (13-7312-85, eBiosciences) in 1X PBS/0.5% FBS and rocked for 3h at RT. Plates were washed with wash buffer and rocked with avidin peroxidase complex (PK-6100, Vectastain) prepared in wash buffer for 1h at RT, then washed with wash buffer followed by 1X PBS. 30% H2O2 was added immediately before coating the plate with one 3-amino-9-ethyl-carbazole (205-057-7, Sigma) tablet dissolved in 2.5 mL di-methyl formamide and mixed with 0.1M acetate buffer, which was incubated 10–15 min at RT. Cold tap water was used to stop the reaction, and dry plates were read on an Immunospot plate reader (Cellular Technology Limited). Data are expressed as the average of technical triplicates of the number of spots per well per 104 CD4+ T cells (calculated using total cell count and CD4+ live lymphocyte frequency identified using flow cytometry).
An increased frequency of insulin-binding B cells is found in the mature B cell repertoire of NOD mice despite identical H chain transgenes that provide similar potential to generate anti-insulin B cells in both B6 and NOD strains (13). The small population of anti-insulin B cells present in the mature VH125Tg/NOD B cell repertoire is sufficient to promote type 1 diabetes (T1D) (12,25). To identify the ontogeny of this disparity and uncover where tolerance defects arise, high throughput flow cytometry was used to investigate these populations in the BM. Insulin autoantibodies are harbingers of disease in WT/NOD mice (9), however anti-insulin B cells are difficult to detect in the BM (Fig. 1). This issue was circumvented by the use of the VH125Tg model, in which a small, but reproducible population of insulin-specific B cells (calculated as in Methods) was detected among B6 and NOD VH125Tg immature B cells (B220mid IgMa+ live lymphocytes) in the BM (Fig. 1A). This population is not observed in VH281Tg mice (12), confirming antigen-specificity in the VH125Tg mice. Of note, the percentage of insulin-specific immature B cells observed was significantly increased ~3–4 fold in the BM of VH125Tg/NOD (0.32 ± 0.17), compared with VH125Tg/B6 BM (0.08 ± 0.14), p < 0.001 (Fig. 1A). Contamination of CD23+ mature recirculating B cells in the B220mid IgM+ gate is minimal; similar results are observed when immature B cells are defined as B220+ IgM+ CD23− (not shown). These data show that the VH125Tg model can be used to track the otherwise rare population of anti-insulin B cells as they navigate tolerance hurdles during development. Furthermore, these data suggest that the autoimmune strain enhances formation of anti-insulin B cells, despite heavy chain repertoire restriction.
IL-7-driven BM culture generates antigen-naïve immature (IgM+ CD23−) B cells in vitro (26,27), and thus provides a useful tool to eliminate tolerance induced by circulating insulin, and other environmental influences present in vivo which may differentially shape autoreactive B cell development within the two strains. B6 or NOD VH125Tg BM was therefore cultured with IL-7 in vitro in the absence of insulin to investigate generation of anti-insulin immature B cells. Flow cytometry analysis of live B220+ IgMa+ VH125Tg/NOD BM cultured with IL-7 showed a nearly 10-fold increased percentage of insulin-binding B cells (0.66 ± 0.29) compared with VH125Tg/B6 BM (0.07 ± 0.06, p < 0.001) (Fig. 1B, summarized in Fig. 1C). VH125Tg/B6 mice showed a significantly increased percentage of insulin-binding B cells compared with negative control VH281Tg/B6 mice (p < 0.01). This demonstrates that enhanced formation of insulin-binding B cells in the developing VH125Tg/NOD BM repertoire occurs independently of the in vivo environment, in which circulating insulin exposure occurs.
Immature B cells migrate from the BM parenchyma into the blood-exposed sinusoids as they mature and prepare for emigration to the periphery and are subject to tolerance checkpoints at these transitions (6,28). To investigate the developmental stage at which an increased frequency of anti-insulin B cells is first apparent in the NOD strain, in vivo labeling with anti-CD19 as described in Methods was used to identify immature B cells (B220+ IgM+ CD23− lymphocytes) within the parenchyma and sinusoids in the BM. Insulin-binding immature B cells are present in both parenchyma (0.52 ± 0.11) and sinusoids (0.71 ± 0.13) of VH125Tg/B6 mice, as well as in the parenchyma (0.59 ± 0.10) and sinusoids (1.06 ± 0.21) of VH125Tg/NOD mice (Fig. 2A–B). These data suggest that anti-insulin B cells are not counter-selected during the transition of immature B cells from the parenchyma into the sinusoids in either VH125Tg/B6 or VH125Tg/NOD mice. Furthermore, the increased frequency of insulin-binding B cells within the NOD strain is first significantly apparent in immature B cells that have reached the sinusoids (p < 0.001).
B6 and NOD VH125Tg mice which express identical anti-insulin heavy chains should possess the same potential to generate insulin-binding B cells, but the observed frequency is disparate (Fig. 1, 2A–B). To identify whether Vκ differences are found in the insulin-binding repertoire of VH125Tg C57BL/6 and NOD mice, flow cytometry sorting was used to purify insulin-binding immature B cells from freshly isolated BM. Degenerate PCR primers were used to amplify Vκ genes from cDNA that were cloned, sequenced, and analyzed as in Methods. Fig. 2C shows that ~70% of the insulin-binding repertoire of VH125Tg C57BL/6 and NOD mice consists of two Vκ 4 genes, Vκ 4-57-1 and Vκ 4-74. CDR comparisons between these genes show germline polymorphisms present in the NOD strain (Fig. 2D), consistent with previously published data identifying germline polymorphisms in many NOD Vκ sequences (24).
Due to the polymorphic nature of NOD Vκ, definitive assignment of specific Vκ identity was previously not possible based on sequence homology alone, as NOD Vκ 4-74 and Vκ 4-57-1 each shared similar homology with the corresponding C57BL/6 germline sequences (24). The Wellcome Trust Sanger Institute (WTSI) is sequencing the genomes of several strains of mice, including NOD. The Mouse Genomes Project online tool LookSeq was used to compare polymorphisms determined by WTSI sequencing of the NOD genome that mapped to the Vκ4-74 region with those experimentally determined for NOD “Vκ4-74” identified in these and prior studies. The current level of coverage is sufficient to show a high degree of agreement between polymorphisms, suggesting the Vκ 4-74 designation is correct and that the nt differences are germline encoded. Sequence coverage in the Vκ4-57-1 region was less robust; however after ruling out Vκ4-74 as its homolog, it is sufficiently dissimilar to other potential Vκ4 that an alternative identity is unlikely. These data show that two polymorphic Vκ4 genes with altered CDR composition dominate the insulin-binding repertoire in the BM of VH125Tg/NOD mice.
Polymorphic changes in NOD Vκ CDR might alter autoantigen binding. The anti-insulin antibody, mAb123, recognizes a separate insulin epitope from VH125, and can be used to identify B cells whose BCR are occupied by endogenous insulin (13,17). To investigate the anatomical niche in which the insulin autoantigen is first encountered, and whether it is differentially recognized by VH125Tg C57BL/6 and NOD mice, in vivo labeling with anti-CD19 was used to identify parenchymal and sinusoidal immature B cells in VH125Tg NOD and B6 mice as in Fig. 2. This was combined with mAb123-biotin staining to enumerate the frequency of insulin-binding B cells whose BCR were endogenously occupied with insulin in these anatomical niches using flow cytometry. Fig. 3A demonstrates that mAb123 staining is specific; the insulin-binding population identified by mAb123 is not observed with isotype control staining (right panels) or in VH281Tg mice that harbor a similar heavy chain but lack insulin-binding B cells (bottom, (12)). As shown in Fig. 3B–C, BM B cells are clearly stained with mAb123-biotin in both parenchyma (0.74 ± 0.22) and sinusoids (1.17 ± 0.21) of VH125Tg/NOD mice, whereas smaller insulin-occupied populations are observed in the parenchyma (0.23 ± 0.03) and sinusoids (0.43 ± 0.10) of VH125Tg/B6 mice. As expected, insulin-occupied B cells are not detected in VH281Tg B6 or NOD mice, confirming mAb123 specificity. The frequencies of mAb123+ B cells detected in VH125Tg mice are significantly higher than the comparative VH281Tg populations, regardless of strain or anatomical niche (p < 0.001). The frequency of insulin-occupied immature B cells is significantly increased in the sinusoids of VH125Tg mice.
To investigate whether the amount of insulin occupancy varies with the anatomical niche or with background strain, the mAb123 MFI of mAb123+ B cells identified in Fig. 3B was compared among parenchymal and sinusoidal immature B cells of VH125Tg NOD and B6 mice. Interestingly, Fig. 3D shows that the mAb123 MFI is elevated in the sinusoids of VH125Tg/NOD mice (5630 ± 708) versus the parenchyma (3344 ± 687). In contrast, the mAb123 MFI is unchanged between the parenchyma (2397 ± 787) and sinusoids (2445 ± 161) of VH125Tg/B6 mice. Whereas there is no significant difference between the mAb123 MFI in the parenchyma of VH125Tg NOD and B6 mice, insulin-occupied BCR in the sinusoids of NOD mice show a significantly higher mAb123 MFI than B6 mice. These data show that despite the low concentration of autoantigen physiologically present (~1–5ng/mL, (29)), insulin occupies the BCR at the earliest detectable stage of antigen-binding commitment of anti-insulin B cells in the BM of VH125Tg mice. The blood-exposed sinusoids of VH125Tg/NOD autoimmune mice are pinpointed as the niche in which insulin-occupied immature B cells are present at higher frequency and with a higher degree of insulin occupancy, compared to the non-autoimmune B6 strain.
Increased levels of insulin autoantigen occupy a higher frequency of insulin-binding B cells in VH125Tg/NOD mice, compared to the non-autoimmune strain (Fig. 3). To functionally investigate whether CDR polymorphisms identified might provide an explanation for this discrepancy, hybridomas were generated from VH125Tg/NOD or VH125Tg/B6 mice 3 d following immunization with human insulin (the original mAb125 immunogen) or beef insulin that were conjugated to a T cell-independent carrier (Brucella abortus). The conjugate has been shown to capture a pre-immune (unmutated) repertoire for insulin (19). This approach combines potent T cell-independent immunization and enhanced fusion techniques to facilitate the capture of autoreactive B cells (19,23). ELISA of hybridoma supernatants verified the production of anti-insulin Ab (screened against human insulin), binding of which was inhibited by the addition of excess human insulin in solution (not shown).
The large majority of Vκ genes identified among the insulin Ab-secreting NOD hybridoma clones were Vκ4-57-1 or Vκ4-74 and were germline encoded (B6 Vκ4-57-1: 9/9, B6 Vκ4-74: 29/34, NOD Vκ4-57-1: 13/15, NOD Vκ4-74: 20/25), based on comparison to published sequences (24). The self-reactive potential of the antibodies was determined by measuring reactivity with both human (foreign) and rodent (self) insulin in ELISA and an autoreactivity index was calculated (Methods). The average autoreactivity index was calculated for each individual Vκ/Jκ species, defined by different CDR aa sequences that arise from diverse Vκ/Jκ rearrangement, thus error bar variability is due to Vκ/Jκ heterogeneity, rather than assay variability; in the event of multiple identical isolates, the average was used (Fig. 4A). The autoreactivity index is graphed for each individual isolate in Fig. 4B. The autoreactivity index is significantly higher (p < 0.01) in NOD Vκ4-57-1 hybridomas, compared to B6 (Fig. 4A–B, left), but is unchanged (p = 0.08), in NOD vs. B6 Vκ4-74 hybridomas (Fig. 4A–B, right). These data show that B6 mice possess at least one endogenous light chain (Vκ4-74) which can pair with VH125 to generate an insulin autoantigen-binding BCR.
Two Jκ(Jκ2 and Jκ4) are polymorphic in the NOD strain (Fig. 4E), clouding interpretation of the specific contribution of Vκ4-57-1 polymorphisms to autoantigen recognition; however Jκ1 and Jκ5 are the same between the two strains. Fig. 4C shows that in the presence of identical Jκ5 rearrangements, the autoreactivity index is increased 3-fold by the polymorphic NOD Vκ4-57-1 compared with the B6 Vκ4-57-1. Vκ125Tg (Genbank Accession Number: M34530 http://www.ncbi.nlm.nih.gov/genbank/) is a well characterized Vκ4-74 insulin-binding light chain that contains a Pro-Pro motif at the Vκ/Jκ join; this motif is found in other insulin-binding light chains (17,20,30). This Pro-Pro motif was found in B6 Vκ4-74 and NOD Vκ4-57-1 hybridomas that exhibited significantly higher insulin autoreactivity indices compared to NOD and B6 Vκ4-74 and Vκ4-57-1 antibodies that did not contain this motif (Fig. 4D, p < 0.001). Pro-Pro was not found in any of the B6 Vκ4-57-1 or NOD Vκ4-74 hybridomas captured, and germline sequence analysis confirms that these light chain sequences cannot encode the Pro-Pro motif, regardless of Jκjoin, as the second Pro can only be contributed by the Vκ(Fig. 4E). Of note, the more highly autoreactive NOD Vκ4-74 P-L clones possess the germline sequence, whereas those with lower autoreactivity have framework mutations, but identical CDR3. These data suggest that polymorphisms in Vκ4-57-1 confer the potential for a Pro-Pro CDR3 motif that increases insulin autoreactivity in VH125Tg/NOD mice. In contrast, an opposite trend is observed for NOD Vκ4-74, in which the Pro-Pro motif potential is absent.
Flow cytometry has revealed the presence of insulin-binding B cells that are occupied by insulin in both the developing repertoire of the BM (Fig. 3) as well as the mature repertoire of the spleen (13,17). Functional studies of antibodies captured through hybridoma screens confirm insulin autoantigen binding specificity (Fig. 4). To assess whether these polymorphic anti-insulin Vκ are counter-selected by central tolerance mechanisms aimed at developing B cells emigrating from the bone marrow into the spleen, the insulin-binding repertoire of VH125Tg/NOD bone marrow and spleen was compared. To ensure capture of anti-insulin B cells present in the developing repertoire in the absence of counter-selection due to autoantigen engagement, VH125Tg/NOD BM cells were cultured with IL-7 in the absence of insulin (as in Fig. 1B–C), and anti-insulin B cells were isolated using flow cytometry sorting. Spleen B cells were purified by CD43− MACS negative selection of the same mice. Insulin-binding B cells were sorted using FACS, and expressed Ig genes were amplified, cloned, and sequenced.
As shown in Fig. 5A, preferential usage of Vκ4-57-1 and Vκ4-74 is found in both developing (73%, 11/15 clones) and peripheral (90%, 27/30 clones) VH125Tg/NOD anti-insulin B cells, similar to what was observed in the anti-insulin repertoire of immature B cells derived from BM ex vivo (Fig. 2C). In many instances, the same CDR sequences that were captured in the insulin-binding hybridoma screen were also found among primary isolates: B6 Vκ4-57-1: 0/9, NOD Vκ4-57-1: 10/15, B6 Vκ4-74: 21/34, NOD Vκ4-74: 23/25 (not shown). There was no significant difference between the usage of particular insulin-binding light chains in the naïve BM culture and peripheral spleen B cells, as determined by a two-sample binomial test to compare the proportions (p = 0.16). These data indicate that light chains confirmed to bind autologous insulin when paired with VH125 (Fig. 4) are eluding central tolerance checkpoints to populate the periphery of T1D-prone NOD mice.
Discrete populations (diagonal binding) of anti-insulin B cells are apparent by flow cytometry in the BM and spleens of VH125Tg/NOD mice (Fig. 5B). These populations were sorted separately into low or high MFI populations based on the intensity of staining with biotinylated insulin. Expressed Vκ were isolated, cloned, and identified as above. Among the low MFI population present in BM and spleen, Vκ4-57-1 predominated (57%, 13/23 clones), and among the high MFI population, Vκ4-74 was present in the majority of isolates (74%, 20/27 clones, Fig. 5C). Taken together, these data suggest that the majority of anti-insulin B cells isolated from VH125Tg/NOD mice arise from the Vκ4 family, specifically from the Vκ4-57-1 and Vκ4-74 genes, and that BCR using these light chains are not eliminated from the repertoire through negative selection.
Previous attempts failed to detect insulin-binding B cells in the pancreas of VH125Tg/NOD mice, despite identification of Vκ4-57-1 among pancreatic B cell isolates (31). The finding that VH125Tg/NOD B cells expressing Vκ4-57-1 react with autoantigen (Fig. 4) suggests that insulin-reactive B cells are present in the pancreas, but have evaded detection by staining with biotinylated-insulin. To assess the frequency of anti-insulin B cells that are present among pancreatic infiltrates, mAb123 was employed to identify insulin-binding B cells whose BCR may be too fully occupied with endogenous insulin to detect through exogenous staining with biotinylated insulin. Using this method, insulin-binding B cells were detected among B220+ IgMa+ live lymphocytes in the pancreas (Fig. 6A). This frequency was compared among B cell subsets present at every developmental stage from origination in the BM to pancreas infiltration. The frequency of BCR occupied by insulin observed in the pancreas was significantly higher than in any other subset compared (2.66 ± 0.58%, p < 0.001 for pancreas vs. every other subset, Fig. 6B). These data show that anti-insulin B cells are increased in the pancreas at least ~3-4-fold over the frequency which initially forms in the developing repertoire, suggesting that anti-insulin B cells are positively selected into the organ targeted by autoimmune destruction.
Vκ repertoire analysis of VH125Tg/NOD pancreata identifies Vκ4-57-1 as 7% of the light chain repertoire based on clone frequency (31 and not shown), a proportion consistent with flow cytometry findings (Fig. 6). CDR aa alignment with NOD germline reference sequences (24) shows evidence of somatic hypermutation (SHM) in VH125Tg/NOD spleen and pancreatic Vκ4-57-1 isolates; the mutated CDR aa are shown in Fig. 7A. Of the Vκ4-57-1 clones isolated from spleen or pancreata, 28% or 47% of clones show evidence of SHM, respectively (Fig. 7B). These data indicate that a substantial proportion of anti-insulin B cells emerging from the BM ultimately undergo SHM.
The presence of SHM in anti-insulin Vκ(Fig. 7A) implies autoantigen-specific T-B cell interactions in the disease process and is consistent with our recent data that show anti-insulin B cells can process and present insulin epitopes to T cells (32). We hypothesize that this is due to anti-insulin B cell crosstalk with cognate T cells. To investigate whether anti-insulin B cells influence the frequency or function of this important subset, splenocytes were harvested from VH125Tg/NOD or VH281Tg/NOD mice and were used in ELISPOT assay to measure Th1 (IFN-γ) and Th2 (IL-4) cytokine production following insulin stimulation. Splenocytes from VH125Tg/NOD mice that harbor anti-insulin B cells show an increase in the number of IFN-γ spots in response to insulin compared to VH281Tg/NOD mice that lack anti-insulin B cells (p < 0.01, Fig. 7C). Interestingly, the numbers of spontaneous IFN-γ spots was also observed to be increased in VH125Tg/NOD splenocytes at baseline (p < 0.01, Fig. 7C). In contrast, IL-4 was undetectable (data not shown). Activated Th1 cells that produce IFN-γ are associated with T1D pathogenesis (33,34). These findings are consistent with the association of anti-insulin B cells in VH125Tg/NOD mice with acceleration of the disease process (12,25), and further suggest that flawed central tolerance for anti-insulin B cells generates a pre-T1D environment that promotes expansion of pathogenic anti-insulin T cells.
These studies highlight the BM sinusoids as the point of “original sin” for the genesis of anti-insulin B cells in type 1 diabetes-prone mice. Thomas Francis initially coined the term “original antigenic sin” to describe how early foreign antigenic encounter guides subsequent immune responses (35). Although the stages of B cell development are different, both processes govern the subsequent focus of B cell/antibody interaction with antigen. In NOD mice, we find a more pernicious source of repertoire bias, in which B lymphocytes display increased insulin autoreactivity from birth in the BM, rather than requiring subsequent affinity maturation to pose an islet threat. In addition to the higher frequency of insulin-binding B cells that form in the autoimmune strain, insulin autoantigen occupancy of cognate BCR is also higher in the BM. Two Vκ4 genes dominate the BM and spleen anti-insulin repertoire of NOD mice, and show germline potential for insulin autoreactivity. Polymorphisms in NOD Vκ4-57-1 confer a key CDR3 motif that enhances insulin autoreactivity. The consequence of BM proclivity for insulin autoimmunity is that anti-insulin B cells transit to the periphery and are positively selected into the pancreas, where they show evidence of somatic hypermutation. The presence of anti-insulin B cells in the repertoire is associated with increased production of the inflammatory cytokine, IFN-γ in ELISpot. These data suggest that the BM “original sin” seeds anti-insulin B cells into the periphery where they likely collaborate with cognate T cells to promote inflammatory attack of the pancreas.
Anti-insulin B cells develop with increased frequency in VH125Tg/NOD BM (Fig. 1A), despite the identical heavy chain present in both NOD and B6 VH125Tg mice. One potential explanation for this difference is that environmental factors in the autoimmune strain may enhance autoreactive B cell development. These could include increased survival factor availability or changes in chemokine or homing receptor expression that increase the kinetics of BM exit, limiting the window for central tolerance in the autoimmune strain. Studies on developing B cells that reach mature subsets in WT/NOD mice suggests that negative selection collapses at the transitional stage (36), and are complementary with data suggesting that similar B cell checkpoint defects underlie autoimmunity in patients (2,3). A receptor editing defect has also been proposed for T1D mice and human patients based on a lower observed RS rearrangement frequency (37). The current investigation clearly indicates that receptor editing is failing to eliminate insulin-binding B cells in VH125Tg/NOD mice. Future studies to directly address whether receptor editing efficiency is altered in autoimmune mice are necessary to outline the role that this key central tolerance mechanism plays in preventing insulin autoimmunity.
IL-7 driven BM culture in vitro limits the influence of environmental factors and eliminates autoantigen exposure to produce naïve immature B cells (26,27). Nonetheless, increased formation of anti-insulin B cells in the NOD strain is still observed in BM cultured in the absence of insulin in vitro (Fig. 1B–C). This suggests that while central tolerance may certainly be a component of the enhanced anti-insulin B cell frequency, it is not the only contributing factor. Vκ polymorphisms alter CDR in the NOD strain (24), and offer one potential explanation for enhanced formation of autoreactive B cells. Two polymorphic Vκ4 family members, Vκ4-57-1 and Vκ4-74, dominate the insulin-binding repertoire in VH125Tg/NOD mice (Fig. 2 and and5).5). The nt changes of the Vκ sequences are of germline origin, and do not arise from SHM, based on comparison with published NOD germline sequences (24). Binding studies confirm that even when the J contribution is the same, Vκ4-57-1 polymorphisms enhance insulin autoantigen recognition in VH125Tg/NOD mice (Fig. 4). A CDR3 Pro-Pro motif has been identified previously in several anti-insulin Vκ(20). These studies show that Vκ4-57-1 polymorphisms in CDR3 confer the potential for generation of this motif, which is shown to enhance insulin autoantigen recognition (Fig. 4).
The serine-rich CDR1 of both Vκ4-57-1 and Vκ4-74 is found among other anti-insulin mAb (20,38). This motif is reminiscent of that found in the insulin receptor ligand binding domain, and is thus a likely contributor to antigen binding. Similar Vκ4 genes are also found in autoreactive antibodies derived from insulin immunization in combination with different VH as well, suggesting this structure may predispose towards insulin recognition (20,38,39). As the polymorphisms present in NOD Vκ are shared among other SLE-prone autoimmune strains (24), and since other anti-insulin autoantibodies characterized from NOD mice have shown cross-reactivity with other self antigens such as DNA, thyroglobulin, IgG, and cardiolipin (39), we speculate that polymorphic Vκ may enhance recognition of other autoantigens as well. The α chain of the TCR is the structural correlate to the light chain of the BCR. Interestingly, a polymorphic TCRα chain also contributes to the formation of TCR on pathological anti-insulin T cells in NOD mice (40,41). Polymorphic Vκ are not the only explanation for the higher frequency of insulin-binding B cells generated in the NOD strain, as Vκ4-74 shows potential for autoreactivity in both strains. Differences in regulatory elements that control Vκ usage may also exist between strains. Central and peripheral tolerance may also be differentially applied, in addition to any repertoire bias in the autoimmune strain.
Some anti-insulin mAb that recognize human insulin do not bind rodent insulin (Fig. 4). This suggests that a fraction of the cells identified by biotinylated human insulin staining may be clonally ignorant. However, it is clear that a proportion of anti-insulin B cells have autoantigen-occupied BCR when antigen specificity is first acquired in the BM parenchyma (Fig. 3); this frequency is elevated in the sinusoids of the NOD strain compared to B6 VH125Tg mice. The increased frequency of insulin-binding B cells observed in the sinusoids, compared to the parenchyma, suggests that positive selection may be occurring (Fig. 2–3). An alternative explanation is that receptor editing prolongs the time anti-insulin B cells spend in the sinusoids, thus increasing the frequency; however, as the anti-insulin B cell percentage in the sinusoids is similar to that observed in transitional and subsequent mature B cell compartments, it is unlikely that receptor editing is culling a major fraction of anti-insulin B cells in VH125Tg/NOD mice.
Naïve immature anti-insulin B cells can sense insulin and show signs of anergy induction following autoantigen exposure, supporting the ability of insulin to interact with the BCR in a perceptible way (26). A higher level of BCR occupancy with insulin autoantigen is first observed in this blood-exposed BM niche in VH125Tg/NOD mice (Fig. 3). We speculate that this may be due to a higher average affinity of BCR for endogenous insulin in VH125Tg/NOD mice, compared to the non-autoimmune strain. Insulin-occupied BCR are also clearly observed in VH125Tg/NOD mature peripheral B cell subsets (13), thus clonal ignorance fails to explain the escape of anti-insulin B cells to the periphery of NOD mice. If anergy is being applied in this model, it certainly does not sufficiently limit the competitive survival of anti-insulin B cells to prevent their infiltration of the pancreas (Fig. 6).
Tracking Vκ usage by anti-insulin B cells in the BM and periphery documents a direct connection between the two compartments (Fig. 5). This identifies BM genesis, rather than peripheral selection of a rare specificity, as the key factor that contributes to autoreactive B cell presence in the mature repertoire. Notably, the Vκ4-57-1 light chain identified in the BM and spleen that was confirmed to be insulin-reactive (Fig. 4) is also found among B cells infiltrating the pancreas of VH125Tg/NOD mice (31). Anti-insulin B lymphocytes exit the BM and infiltrate the pancreas with 3–4 fold increased frequency (Fig. 6). Insulin-reactive B cells that escape from the BM can thus impact the invading repertoire of the pancreas. Vκ1-110, Vκ9-120, and Vκ9-124 were also observed in a minor fraction of insulin-binding B cells isolated from BM and spleen. These light chains have also been isolated from other anti-insulin B cell screens (16), as well as from the pancreas, the site of autoimmune destruction (31 and not shown), however functional studies are required to confirm their specificity for insulin.
Flow cytometry identifies two distinct insulin-binding populations in the BM and spleen (Fig. 5), however only one is apparent in the pancreas (Fig. 6). Interestingly, the Vκ4-74 light chain was not observed in the pancreas of VH125Tg/NOD mice (31), despite its ability to bind insulin autoantigen (Fig. 4). One explanation is that Vκ4-57-1 may have higher affinity for autoantigen and thus is more heavily selected into the site of autoimmune attack. In support of this, the autoreactivity index is higher on average for Vκ4-57-1 than for Vκ4-74, suggesting that Vκ4-57-1 more efficiently binds rodent (self) insulin (Fig. 4). Biotinylated insulin staining is impaired by higher BCR occupancy with endogenous insulin in the pancreas of VH125Tg/NOD mice, [(31) and Fig. 6]. Of the two insulin-binding populations observed in BM and spleens, Vκ4-57-1 is primarily associated with the population that shows lower biotinylated-insulin MFI (Fig. 5). We therefore propose that Vκ4-57-1 has higher endogenous antigen affinity (and thus higher occupancy) due to unique polymorphisms, which may lead to its selection into the pancreas.
It is clear that many Vκ4-57-1+ B cells have undergone SHM (Fig. 7A–B and (31)). Tertiary lymphoid structures with germinal centers are observed to form in the pancreas, and it has been proposed that SHM occurs at the site of autoimmune attack (31). Further work will be required to determine how SHM impacts the disease process. Regardless of when SHM initiates, productive B lymphocyte collaboration with T cells is also suggested by increased IFN-γ production in the spleens of VH125Tg/NOD, but not VH281Tg/NOD mice, in which insulin-binding B cells are present, or absent, respectively (Fig. 7C). This is further supported by the finding that anti-insulin B cells can present antigen to cognate T cells isolated from NOD mice (32). We therefore favor the hypothesis that enhanced IFN-γ production results from antigen-specific T cells activated by anti-insulin B cells in vivo. The production of IFN-γ, but not IL-4, is consistent with Th1 cell activation, which is known to mediate the islet inflammatory process in T1D (33,34). These data therefore serve as indirect evidence that anti-insulin B cells are presenting antigen to insulin-reactive Th1 CD4+ T cells to stimulate their production of IFN-γ within a polyclonal repertoire, and suggest inflammatory consequences of anti-insulin B cell escape through tolerance checkpoints.
Based on these findings, we propose the following model. Anti-insulin B cells form with increased frequency in the BM of autoimmune VH125Tg mice, which is first apparent in the BM sinusoids. Polymorphisms that alter CDR composition enhance self-reactivity for insulin in VH125Tg/NOD mice. Despite autoantigen encounter as early as antigen-commitment in the BM parenchyma, insulin autoreactive B cells escape into the periphery of the autoimmune strain in the absence of detectable negative selection. Anti-insulin B cells interact with anti-insulin T cells to drive islet-directed inflammation, and are enriched in the pancreas, the site of autoimmune attack. The BM is thus identified as the point of “original sin”, from which dangerous autoimmunity first arises in the B cell repertoire that leads to the peripheral consequence of islet attack.
We would like to acknowledge James B. Case, Chrys Hulbert, Allison M. Sullivan, Emily J. Woodward, and Guowu Yu for technical support (Vanderbilt University), as well as the VMC Flow Cytometry Shared Resource and the VDDRC, the Vanderbilt DNA Sequencing Facility, the Vanderbilt Antibody and Protein Resource, the Vanderbilt Hormone Assay & Analytical Services Core, and the Vanderbilt Diabetes Research and Training Center. We would also like to thank Dr. Betty Diamond (Feinstein Institute, Manhasset, NY) for kindly providing the NSO-BCL2 myeloma line.
1This work was supported by NIH grants 5T32-HL069765, 5T32-AR059039, and R01-AI051448, R01 DK084246, K08 DK070924 and Core Laboratories by P30 CA68485, DK058404, CA68485, DK20593, and HL65962.
2Abbreviations: B6: C57BL/6, BCR: B cell receptor, T1D: Type 1 diabetes, VH: Heavy chain Ig Tg