General Approach for Tetramer-Based Identification of Autoantigen-Reactive B Cells: Characterization of La- and snRNP-Reactive B Cells in Autoimmune BXD2 Mice

Autoantibody production by autoreactive B cells is characteristic of many autoimmune diseases, including systemic lupus erythematosus (SLE) and rheumatoid arthritis (1, 2). Studies using mouse models indicate that certain autoantibodies can drive the development of these diseases (3–5). In humans, the close association of some autoantibodies with disease activity and progression, together with the therapeutic effects of B cell depletion, suggests their role in clinical disease (6, 7). Although disrupted regulation of autoreactive B cells is considered central to the development of autoimmunity, the relative contributions of different subsets of B cells (8, 9) remains unclear. Progress in this area is hindered by the low frequency of the autoreactive B cells and their diversity, which encompasses the broad spectrum of autoantigens recognized, the isotype of the Abs produced, and the subtle phenotypic distinctions that differentiate B cell subsets. The most commonly used approach for the analysis of autoantigen-specific B cell subsets in autoimmunity has been the creation of transgenic mice in which the cells can be expanded clonally through experimental manipulation (10).

Labeled monomeric and tetrameric Ag conjugates can be used to brightly label cells on the basis of their ligand specificity (11, 12). This approach has been applied successfully to the identification and isolation of specific types of cells that occur at low frequency (13, 14). However, it is technically difficult to construct a labeled autoantigen tetramer using most full-length Ags, because the process requires ligation of the Ag-coding material into an expression vector with a biotinylated site and, subsequently, stringent purification of the Ag. One approach to overcome this issue is the use of small, linear-peptide autoepitopes. In 2003, Newman et al. (15) described a system in which a DNA mimetope peptide could be conjugated to PE-labeled streptavidin (SA) and used to detect B cells reactive to this DNA mimetope in immunized BALB/c mice and later in humans with SLE (16). This tetramer strategy has since been adapted for the isolation of B cells specific for various epitopes on citrullinated fibrinogen (17), HLA (18), HIV gp41 (19, 20), and tetanus toxoid C fragment (11). Recently, Taylor et al. (21) used a novel detection and tetramer enrichment strategy to assess polyclonal self-Ag–specific B cells by rigorously analyzing regulation of OVA and glucose-6-phosphate isomerase Ag-specific B cells in OVA-expressing and wild-type mice.

Potentially, the epitope tetramer approach represents a powerful tool for analysis of B cell reactivity in autoimmunity, especially if it could be applied to the analysis of B cells that are reactive with the predominant autoantigens that have been identified in SLE and rheumatoid arthritis. However, full realization of this potential depends on a feasible strategy for the identification and characterization of linear autoepitopes and the demonstration that these linear autoepitopes represent authentic, clinically relevant Ags. Since the first description of the development of large-scale autoantigen arrays for autoantibody determination in patient sera (22), advances in peptide microarray chip technologies have led to the development of comprehensive Ag arrays (23, 24). We reasoned that these arrays, which contain thousands of known B cell epitopes and their modified variants, can be used to identify authentic linear autoepitopes that are potential candidates for the generation of autoantigen-tetramer panels.

To test this strategy, we used a step-wise approach to enable identification of lupus La and small nuclear ribonucleoprotein (snRNP) autoepitopes reactive with B cells from the BXD2 mouse model of systemic autoimmunity. The emergence of the autoantibody repertoire and, importantly, its age-associated transition to the expression of pathogenic autoantibodies, is well characterized in these mice (4, 25, 26). The use of these tetramers for FACS of B cells from the spleens of BXD2 and C57BL/6 (B6) mice confirmed an expanded frequency and number of La_13–27 and snRNP_357–373 epitope-reactive, activated (CD69⁺ and CD86⁺) marginal zone (MZ) precursor (MZ-P) B cells in BXD2 mice compared with age-matched B6 mice and further revealed an expanded frequency and higher number of La_13–27 and snRNP_357–373 reactive transitional T3 and memory B cells. Thus, the present work validates the use of linear autoepitopes for analysis of the autoantibody repertoire and establishes a systematic, experimental approach to autoepitope identification and Ag tetramer–based B cell isolation. This approach is readily adaptable to analysis of other autoimmune models and, potentially, analysis of patient-derived samples.

B6 and BXD2/TyJ recombinant inbred mice were obtained from The Jackson Laboratory. All mice were housed in the University of Alabama at Birmingham Mouse Facility under specific pathogen–free conditions in a room equipped with an air-filtering system. The cages, bedding, water, and food were sterilized. All mouse procedures were approved by the University of Alabama at Birmingham Institutional Animal Care and Use Committee. Sera were obtained from B6 and BXD2 mice at the specified ages by retro-orbital eye bleeding. After separation from blood, sera were stored at −80°C.

Autoantibody profiles for linear peptides were determined using PEPperCHIP technology (PEPperPRINT, Heidelberg, Germany). For the PEPperCHIP Autoimmune Epitope Microarray, 2733 autoimmune disease–associated linear B cell epitopes were selected without bias using all linear epitopes that have been curated for humans or mouse models of autoimmune diseases from the Immune Epitope Database (27) (IEDB; http://www.iedb.org/), including 192 citrullinated peptides. Longer peptides with >15 aa were translated into overlapping 15-aa peptides with a peptide–peptide overlap of 14 aa, resulting in 3830 total peptides on the chip. Peptides shorter than 3 aa and peptides including noncitrulline side-chain modifications were not included on the chip. Peptides were all printed in duplicate spots and framed by a fusion tag (Flag) peptide (DYKDDDDKGG, 186 spots) and influenza virus hemagglutinin (HA) epitope tag peptide (YPYDVPDYAG, 186 spots) as controls. Controls were detected by monoclonal anti-FLAG(M2)-LL-DyLight 800 and monoclonal anti-HA (12CA5)-LL-DyLight 680 (1:1000) (Rockland Immunochemicals, Gilbertsville, PA).

The microarray was initially incubated with the secondary goat anti-mouse IgG (H+L) DyLight 680 Ab at a dilution of 1:5000 for 60 min at room temperature to analyze background interactions with the autoimmune epitopes, ensuring that there were no background interactions due to nonspecific binding of the secondary Ab to the peptides. Serum from six 8–10-mo-old BXD2 mice was pooled. Before serum incubation with chip, peptide arrays were incubated for 60 min in Blocking Buffer for Near Infra Red Fluorescent Western blotting (Rockland Immunochemicals). The microarray was washed twice in PBS (pH 7.4) with 0.05% Tween 20 (PBS-T) and incubated for an additional 30 min in washing buffer. The array was incubated overnight at 4°C with mouse sera diluted 1:1000 for anti-mouse IgG (H+L) analysis or 1:200 for γ-chain–specific analysis and 1:1000 for μ-chain–specific analysis in PBS-T (secondary anti-mouse Ig Abs from Rockland Immunochemicals). After multiple washes in washing buffer, the microarrays were incubated for 30 min with the secondary Ab in PBS-T at room temperature. After two additional washes in washing buffer, the microarrays were rinsed with ultrapure water and dried in a stream of air.

Green/red fluorescence intensities were acquired on an Odyssey Imager (LI-COR, Lincoln, NE) at scanning intensities of 7/7 in both channels (700/800 nm), 0.8–1.0 mm offset, at a spatial resolution of 21 μm. Staining of the Flag and HA control peptides that frame the arrays gave rise to high and homogeneous spot intensities with a coefficient of variation < 2%. PepSlide Analyzer software was used to analyze the data. This program breaks down the fluorescence intensities of each spot into raw, foreground, and background signals and calculates the SD of the foreground median intensities. Epitopes resulting in a binding intensity >5 SD from mean chip intensity were considered positive.

A custom epitope microarray with 80 epitopes selected based upon the first array analysis was prepared (PEPperCHIP). Analysis of serum from individual B6 and BXD2 mice of different ages was carried out in the same fashion as the first microarray (serum dilution 1:200 for IgG analysis, 1:1000 for IgM analysis). IgG and IgM reactivity was detected by goat anti-mouse IgG–conjugated DyLight 800 and goat anti-mouse IgM (μ chain) conjugated to DyLight 680, at 1:1000 and 1:5000 dilutions, respectively (both from Rockland Immunochemicals). For these analyses, epitopes resulting in a binding intensity ≥5-fold compared with B6 control mice were considered positive. Lastly, pooled sera from B6 and BXD2 mice were used to probe the 2.0v Autoimmune Epitope Microarray, containing 4389 peptides from the IEDB (PEPperCHIP).

Lupus La_13–27 (LEAKICHQIEYYFGD) and snRNP_357–373 (SHRSERERRRDRDRDRD) were produced and biotinylated at the N terminus by Sigma-Aldrich and supplied as a lyophilized powder. Each peptide was suspended in DMSO to a stock concentration of 10 mM and diluted to 1 mM in pure dH₂O. Suspended peptides were aliquoted and stored at −80°C. Tetramers were generated by adding biotinylated peptide step-wise in 1/10 volumes to 6.7 μM SA–Rhodophyta-PE (ProZyme) at a molar ratio of 30:1 and allowed to incubate for 60 min at room temperature or overnight at 4°C. Tetramers were purified on a Sephacryl S-300 FPLC size-exclusion column. The tetramer fraction was concentrated using a 100-kDa molecular mass–cutoff Amicon Ultra filter (Millipore). The concentration of tetramer was calculated by comparison with a standard curve of PE absorbance at 540 nM, which was measured using an EMax Precision Microplate Reader (Molecular Devices, Sunnyvale, CA).

The nonspecific tetramer control was prepared by conjugating the core fluorochrome SA-PE to Alexa Fluor 647 (AF647; Molecular Probes), according to the manufacturer’s protocol, for 60 min at room temperature. The free AF647 was removed by centrifugation in a 100-kDa molecular mass–cutoff Amicon Ultra filter (Millipore). The SA-PE*AF647 complex concentration was calculated by measuring the absorbance of PE at 540 nm. The SA-PE*AF647 complex was incubated with a 10-fold molar excess of free biotin for 30 min at room temperature.

Spleens were harvested from individual mice, and single-cell suspensions were prepared in RPMI 1640 supplemented with 5% FBS. Cell suspensions were prepared by gentle teasing apart of tissue with the plunger of a 3-ml syringe. Cells were passed through a cell strainer to eliminate clumps and debris and washed with RPMI 1640. RBCs were lysed from the resuspended pellet with 3–5 ml ACK lysing buffer and washed twice in RPMI 1640. Lymphocytes were resuspended to 200 μl in anti-CD16/32 Fc block (2.4G2; BioLegend) in 2% rat serum. Next, PE*AF647-conjugated nonspecific tetramer was added at a concentration of 20–30 nM and incubated at 4°C for 5–10 min. PE-conjugated peptide tetramer was added at a concentration of 10–20 nM and incubated on ice for 30 min, followed by one wash in 15 ml cold fresh sorter buffer (PBS, 2 mM EDTA, 2% FBS). Tetramer-stained cells were resuspended to a volume of 80 μl sorter buffer/10⁷ cells, mixed with 5–10 μl anti-PE–conjugated magnetic MicroBeads (Miltenyi Biotec) per 10⁷ cells, and incubated on ice for 30 min, followed by one wash with 10 ml sorter buffer. In experiments in which single-cell suspensions were divided into 1/2 and 1/4 mouse equivalents before tetramer labeling/enrichment, the volumes were scaled down appropriately. The cells were then resuspended in 3 ml cold sorter buffer and passed over a magnetized LS column (Miltenyi Biotec). Bound cells were eluted according to the manufacturer’s instructions. Free peptide–blocking experiments were performed on unenriched samples. Free peptide (300 μM) was allowed to incubate with cells 30 min prior to tetramer staining.

Cell pellets from the enriched and column flow-through fractions were resuspended in FACS buffer (PBS + 5% FBS) and incubated with surface Abs for 30 min on ice. Surface Abs included Pacific Blue–anti-CD19 (6D5) or Brilliant Violet 650–anti-CD19 (6D5), Brilliant Violet 605–anti-CD86 (GL-1), Brilliant Violet 510–anti-CD69 (H1.2F3), FITC–anti-CD21/35 (7E9), PE-Cy7–anti-F4/80 (BM8), PE-Cy7–anti-Thy1.2 (30-H12), and FITC–anti-CD93 (AA4.1; all from BioLegend). Dead cells were excluded from analysis with allophycocyanin–eFluor 780 Organic Viability Dye (eBioscience). After cell surface staining, cells were washed twice with 3 ml FACS buffer and fixed in 1% paraformaldehyde/FACS solution for cell surface marker analysis. Cells (300,000–1 × 10⁶/sample) were analyzed by flow cytometry. FACS data were acquired with an LSR II FACS analyzer (BD Biosciences) and analyzed with FlowJo software (TreeStar, Ashland, OR). All flow cytometry analysis was carried out using a combination of forward light scatter and side scatter height, area, and width parameters to exclude aggregated cells. B cell population gating strategy was carried out based on the method described by Allman and Pillai (28).

The levels of autoantibodies specific for selected linear peptides in the sera of B6 and BXD2 mice were determined by ELISA using a NeutrAvidin Coated High Capacity 96-well plate (Thermo Scientific Pierce). Briefly, biotinylated peptides were conjugated to the plate overnight at 4°C at a concentration of 30 μM (all peptides were purchased from Sigma-Aldrich). ELISAs were developed with either an HRP-labeled goat anti-mouse IgG Ab or a goat anti-mouse IgM Ab (Southern Biotechnology Associates) and tetramethylbenzidine substrate (Sigma-Aldrich). OD_450–650 was measured on an EMax Microplate Reader.

NeutrAvidin Coated High Capacity plates (96 well) (Thermo Scientific Pierce) were coated overnight with 50 μM peptide at 4°C, washed, and blocked with complete medium. Single-cell suspensions of spleens isolated from B6 or BXD2 mice were prepared, as described above, followed by erythrocyte removal by ACK lysis. Cells were washed twice and adjusted to a final volume of 200 μl containing 1 × 10⁵ cells/well in the presence or absence of PMA (50 ng/ml) and ionomycin (750 ng/ml; both from Sigma-Aldrich), which stimulate the calcineurin pathway and B cell–affinity maturation (29). After incubation for 24 h, plates were washed six times with PBS-T and incubated for 4 h with 1 μg/ml HRP-conjugated goat anti-mouse IgM mAb or HRP-conjugated goat anti-mouse IgG (both from Southern Biotechnology Associates) in PBS/5% BSA. Plates were washed six times with PBS/0.05% Tween 20 before spots were developed using 3-amino-9-ethylcarbazole. Plates were read using an automatic ELISPOT reader and analyzed using Immunospot 3.1 software (CTL). All experiments were repeated in duplicate.

For some experiments, different populations of tetramer⁺, nonspecific and tetramer⁻ B cells were sorted on a FACSAria and cultured in the presence of LPS (Sigma-Aldrich; 20 μg/ml) and IL-4 (100 U/ml) for 4 d. The supernatant was collected and used immediately or was stored at −80°C until analysis of reactivity for the secreted IgG autoantibodies.

Full-length human recombinant La was purchased from ProSpec-Tany TechnoGene. La protein was uniformly uploaded, electrophoresed on 12% SDS polyacrylamide gels, and transferred onto polyvinylidene difluoride membranes. The membranes were incubated with supernatant from cultured cells overnight at 4°C. As a positive control, anti-La Ab (Cell Signaling Technology) was used at a dilution of 1:1000 to confirm the identity of the La protein on the blot. Anti-mouse or anti-rabbit HRP-conjugated Abs (Life Technologies) were used at a 1:250 dilution. HRP Abs were detected using a chemiluminescent substrate (Pierce).

All results are shown as mean ± SEM. A two-tailed t test was used when two groups were compared for statistical differences, and p values < 0.05 were considered significant. For microarray Ag-distribution analyses, χ² analysis was performed, and a p value < 0.05 was considered significant.

Microarray data were deposited in the National Center for Biotechnology Information’s Gene Expression Omnibus (GEO) under master accession number GSE65290 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE65290). GEO accession numbers for data shown in Figs. 1 and 2 are GSE65276 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE65276) and GSE65234 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE65234), respectively. GEO accession numbers for data shown in Supplemental Fig. 1 are GSE65277 and GSE65278 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE65277 and http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE65278).

To test whether an autoimmunity epitope microarray could be used to identify linear autoepitopes in the sera of autoimmune BXD2 mice, we used pooled sera from six BXD2 mice that developed spontaneous lupus-like features (8–10 mo old). The microarray consists of 2733 experimentally characterized autoimmune epitopes and 192 citrullinated variants. All of these epitopes have been associated with autoimmune disease and are cataloged in the IEDB (30). For our analyses, chip epitopes derived from nonmammalian species were not included. The BXD2 serum peptide-reactivity profile was detected by goat anti-mouse IgG, revealing that the majority (79%) of BXD2-reactive autoepitopes derive from three main groups: nuclear proteins, endoplasmic reticulum or mitochondrial enzymatic proteins, and matrix proteins (Fig. 1A). There was a significantly increased prevalence of reactivities to peptides deriving from nuclear Ags and endoplasmic reticulum/mitochondrial enzymatic proteins (Fig. 1A) and, within the nuclear Ags, an increase in peptides deriving from ribonucleoprotein (RNP) Ags (Fig. 1B). These results are consistent with our previous MALDI-TOF–mass spectrometry analysis of BXD2 autoantibody-precipitated autoantigens resolved by two-dimensional electrophoresis, in which the major autoantigens targeted by serum from 8–10-mo-old BXD2 mice were identified as nuclear proteins (snRNP, Ro, histone), heat-shock proteins (GRP78 or “BiP”), enzymes (enolase, aldolase), and structural proteins (keratin, actin) (4, 26).

The identified BXD2 autoepitope profile advances knowledge concerning the autoantibody repertoire in BXD2 mice. Common lupus-associated autoantigens, such as snRNP, smD1, CENP-A, and Ro52, were represented by 53, 37, 39, and 46 reported B cell linear epitopes on the microarray, respectively (Fig. 1C). This extensive coverage allowed identification of multiple BXD2⁺ epitopes on a given autoantigen. Ags displaying extensive reactivity with BXD2 serum (six or more positive epitopes) were 70-kDa U1RNP, smD1, CENP-A, and Ro52 (Fig. 1C). Similarly, the inclusion of citrulline-modified variants in the array content revealed that pooled BXD2 serum was reactive with a citrullinated and a noncitrullinated version of fibromodulin, fibrinogen-β, and BiP peptides (Fig. 1D). This finding is consistent with reports that human autoantibodies recognize both an noncitrullinated and citrullinated variant of peptides deriving from common autoantigens (31, 32).

Given that the number of potential autoepitopes in BXD2 mice is extremely large, further analysis required development of an efficient, high-throughput strategy that would generate the data required for informed selection of the linear autoepitopes of interest. A secondary microarray was printed that included linear peptides representing the predominant epitopes identified in the original screening of pooled BXD2 sera, encompassing peptides from the three major categories of proteins targeted by BXD2 autoantibodies and present in human disease (33): nuclear, matrix, and enzyme/chaperone Ags (Table I).

Because class switching is associated with the production of pathogenic autoantibodies (34), the secondary microarrays were probed with serum from individual BXD2 mice and control normal B6 mice using an isotype-specific analysis (Fig. 2). In addition, because the emergence of class-switched, pathogenic autoantibodies follows a well-defined age-associated progression in BXD2 mice (4), sera from BXD2 mice of different ages and age-matched B6 controls were included in this assay. The results obtained from analysis of BXD2 sera indicated an increase in IgG autoantibodies to peptides deriving from nuclear autoantigens, with the highest levels of anti-La, RNP, and Ro60 in older BXD2 mice (Fig. 2A, 2B). The predominant nuclear peptides that were reactive with BXD2 sera were associated with a cluster of key Ags, including La, the U1 snRNP complex, Ro60, CENP-A/E, DNA topoisomerase, and RNA polymerase II (Fig. 2A). Epitopes derived from matrix proteins fibrinogen-β and fibromodulin (Fig. 2C) and enzyme proteins BiP and clusterin (Fig. 2D) were positive in most BXD2 mice.

In contrast to the IgG autoantibodies, which were detected largely in BXD2 mice only, IgM autoantibodies were broadly present in both B6 and BXD2 mice (Fig. 2, Supplemental Fig. 1). These results are consistent with the previously described age-associated emergence of class-switched Ro60, followed by heat-shock proteins, including BiP, and, finally, histone and DNA autoantibodies in BXD2 mice (4).

To verify that the peptides discovered by screening the epitope microarray are authentic epitopes, selected peptides were synthesized with an N-terminal biotin (Table II). These biotinylated peptides were analyzed for reactivity with IgM and IgG autoantibodies in the sera of B6 and BXD2 mice by standard ELISA and ELISPOT analyses. For these assays, we used NeutrAvidin Coated High Capacity 96-well plates for anti-peptide Ab detection. Using this modified ELISA, we found significantly higher levels of IgG autoantibodies directed against all tested linear epitopes from BiP, histone, CENP-A, La, and snRNP, as well as structural Ag epitopes fibrinogen-β-cit and fibromodulin-cit, in the sera of 7–9-mo-old BXD2 mice compared with younger BXD2 mice and normal B6 mice (Fig. 3A, 3B). Also, consistent with the microarray data, we found IgM autoantibody reactivity with all peptides tested and, for most peptides, a lack of elevation in the levels of these autoantibodies between young or old B6 mice compared with young or old BXD2 mice (Fig. 3C).

Using the modified ELISPOT, we also were able to enumerate anti-peptide Ab–producing B cells from the spleens of 5–6-mo-old B6 and BXD2 mice and observed a significantly increased number of IgG-producing B cells from BXD2 mice that were reactive with La_13–27, histone H1b_205–219, and snRNP_357–373; this was increased by stimulation with PMA + ionomycin (Fig. 3D). However, there were few to no detectable IgG-producing autoantibody-reactive B cells from B6 mice (Fig. 3D). For IgM isotype ELISPOT analysis, under unstimulated conditions, there was a significantly lower number of IgM spots for H1b_205–219 and snRNP_357–373 in BXD2 mice compared with B6 mice (Fig. 3E).

From the above assays, the La_13–27 epitope was selected for construction of an autoepitope tetramer using a protocol modified from Taylor et al. (21). Briefly, a 10–20 fold molar excess of biotinylated peptide was conjugated to Rhodophyta-PE–labeled SA. To exclude cells binding irrelevant epitopes on PE or SA, a nonspecific tetramer also was produced by conjugating AF647 to SA-PE loaded with biotin to block free binding sites. For all tetramer experiments, single-cell suspensions were costained with the nonspecific biotin-PE*AF647 tetramer and the peptide PE tetramer to discriminate between peptide binding and nonspecific binding (21). Lymphocytes were gated by forward and side scatter, followed by exclusion of doublets, nonviable cells, and non-B cells. The resulting two-dimensional plot with biotin-PE*AF647 nonspecific tetramer on the y-axis and the peptide PE tetramer on the x-axis yields a small population of La_13–27-binding B cells (Fig. 4A). To test the specificity of tetramer-labeled B cells, cells were preincubated with free peptide prior to tetramer staining. Preincubation of cells with La_13–27, but not irrelevant OVA_323–339 peptide, resulted in a significant decrease in La_13–27 tetramer–stained cells (Fig. 4B, 4C).

Because epitope-specific B cells are rare within the B6 and BXD2 polyclonal repertoire (Fig. 4D, left panels), tetramer-stained cells were enriched with magnetic anti-PE beads and a magnetized MACS column using the protocol described by Taylor et al. (21). Although most cells obtained in the final enriched fraction bind nonspecifically to the column or biotin-PE*AF647 tetramer, a smaller number of cells bind only the peptide PE tetramer. This tetramer-enrichment strategy enables isolation and analysis of rare peptide-specific B cells with significantly increased sensitivity compared with a similar analysis of pre-enriched B cells and column flow-through fractions (Fig. 4D, 4E). A similar method was used to generate, enrich, and determine the B cell–binding specificity of a second snRNP_357–373 tetramer (Supplemental Fig. 2).

Identifying a BXD2 autoepitope repertoire for tetramer design enables study of epitope-reactive B cells in nontransgenic autoimmune mice. Therefore, we tested this approach in a proof-of-principle study designed to determine whether autoantigen-reactive B cells are skewed toward separate B cell subsets in B6 and BXD2 mice. Based on results from the aforementioned analyses, epitopes derived from La and snRNP were selected. These two RNA-binding protein (RBP) autoantigens are targets of human SLE autoantigens (6) and are present in apoptotic debris (35, 36). Tetramer-reactive cell populations were gated as described in Fig. 4. There were comparable percentages of La_13–27- or snRNP_357–373-reactive B cells in the spleens of 6–8-mo-old B6 and BXD2 mice (Fig. 5A), although the total numbers of La_13–27⁺ or snRNP_357–373⁺ B cells were significantly higher in the spleens of BXD2 mice (Fig. 5B) as a result of splenomegaly (37).

We showed previously that IgM^hiCD21^hiCD23^hi MZ-P B cells can capture and present Ags derived from membrane OVA apoptotic debris to directly stimulate OT-II TCR- specific CD4 T cells (38). We also showed previously that splenic MZ-P B cells are increased in BXD2 mice and can capture and transport TNP Ag directly into the germinal center (GC) of BXD2 mice (39). Because both MZ and MZ-P B cells are increasingly implicated in the pathogenesis of SLE (9, 40, 41), and both populations are critical in BXD2-autoreactive GC development (42), we analyzed the expression of IgM, CD21, and CD23 markers (28) to delineate the subsets of B cells that were reactive to La_13–27 and snRNP_357–373 peptides. From these analyses, both the IgM^hiCD21^hi population, containing the MZ and MZ-P B subsets, as well as IgM^lo/−CD21^lo/− population, containing the follicular (FO) B subset, were represented in the tetramer⁺ population (Fig. 5C). There was an increased percentage of La_13–27- and snRNP_357–373-reactive IgM^int and IgM^lo FO B cells in the BXD2 mouse compared with the B6 mouse. Furthermore, within the IgM^hiCD21^hi population that contains both IgM^hiCD21^hiCD23^lo/− MZ and IgM^hiCD21^hiCD23⁺ MZ-P B cells, there was a significant increase in the percentage of La_13–27 and snRNP_357–373 tetramer-reactive CD23^hi MZ-P B cells in BXD2 mice (Fig. 5C). There were also significantly increased numbers of La_13–27⁺ and snRNP_357–373⁺ MZ-P B cells in the BXD2 mouse spleen (Fig. 5D). These results are consistent with our previous finding that BXD2 mice have a higher frequency and greater number of FO and MZ-P B cells than do B6 mice (42) and further demonstrate that a fraction of these cells exhibit reactivity to La_13–27 and snRNP_357–373 epitopes.

In vivo studies by other investigators showed that counterselection against autoreactivity may take place at the transitional stage of B cell development (43, 44). We further used the CD93 (AA4) marker in combination with CD23 and surface IgM staining to determine whether there is abnormal deletion or selection of a transitional T1, T2, or T3 population of B cells within tetramer⁺ B cells in BXD2 mice. The results showed that, although T1 is the dominant La_13–27⁺ or snRNP_357–373⁺ population of CD93⁺ transitional B cells from B6 mouse spleens, there was abnormal skewing of La_13–27⁺ and snRNP_357–373⁺ B cells to the T3 population in BXD2 mice (Fig. 5E, 5F). Such abnormal expansion of T3 transitional B cells is also observed in the total B cell population in BXD2 mice (Supplemental Fig. 3).

The unusual expansion of MZ-P B cells and otherwise anergic T3 B cells (45) in BXD2 mice suggests that specific activation signals must exist to lead to expansion of these Ag-specific B cells. We showed previously that type I IFN–mediated upregulation of CD69 and downregulation of S1P1 results in their inward migration from the MZ into the follicle (40), where higher CD86 expression enables CD4 Th cell stimulation (46). Therefore, we analyzed whether the La_13–27-reactive and snRNP_357–373-reactive MZ-P B cells exhibited an activation phenotype characterized by CD69 and CD86 expression. Compared with B6 mice, there was a significant upregulation of CD69 and CD86 in the tetramer⁺ MZ-P B cell population in BXD2 mice (Fig. 6A, 6B). Similar analysis of these markers in the tetramer⁺ IgM^lo/−CD21^lo/− compartment, containing mainly FO and potentially GC B cells, revealed a slight increase in CD86, but not CD69, expression on both La_13–27- and snRNP_357–373-reactive cells (Fig. 6A, 6B). These results demonstrate that autoantigen La_13–27- and snRNP_357–373-specific MZ-P B cells in BXD2 mice exhibit upregulation of CD69 and CD86.

Because ELISPOT analysis demonstrated significantly elevated numbers of IgG anti-La_13–27– and anti-snRNP_357–373–producing B cells, one important question is whether tetramer⁺ B cells from BXD2 mice can mature into IgG-secreting cells reactive with the peptide epitope and the whole self-Ag under appropriate stimulation. This was tested by sorting of La_13–27⁺, nonspecific PE*AF647⁺La_13–27^+/−, and double-negative PE*AF647⁻La_13–27⁻ B cell subsets, followed by stimulation of these cells in vitro with LPS + IL-4. We first tested cell culture supernatant reactivity against the La_13–27⁺ peptide and an irrelevant histone H1b peptide (His_205–219). ELISA results verified that La_13–27⁺ B cells produced significantly higher levels of IgG Abs that reacted specifically with the La_13–27⁺ peptide and not an irrelevant His_205–219 peptide. In contrast, the nonspecific PE*AF647⁺La_13–27^+/− B cells produced low levels of Abs for both His_205–219 and La_13–27 (Fig. 6C). The selective production of full-length 47-kDa La-reactive IgG Abs by LPS + IL-4–stimulated La_13–27⁺ B cells, but not by other subsets of B cells, was verified using a Western immunoblotting analysis with whole recombinant La (Fig. 6D).

Finally, we applied the current tetramer approach to determine whether, in vivo, BXD2 mice exhibited elevated numbers of mature and mutated B cells. This was carried out by analyzing the presence of CD80⁺PD-L2⁺ memory B cells within the La_13–27- and snRNP_357–373-reactive populations. B cells upregulating these two molecules were shown to be enriched in isotype-switched cells and in highly mutated cells, even when IgM bearing (47). Using this classification, La_13–27 and snRNP_357–373 tetramer-reactive B cells in BXD2 mice displayed a significantly higher frequency of CD80⁺PD-L2⁺ B cells compared with the counterpart B cells from B6 mice (Fig. 6E, 6F). There was also an increased frequency of CD80⁺PD-L2⁺ B cells in tetramer⁺ cells compared with nontetramer-gated total B cells from BXD2 mice (Fig. 6E, 6F). Together, these results indicate that this tetramer approach can be used to identify abnormal development of autoreactive B cells at various developmental stages.

In this study, we demonstrate a step-wise approach to enable the enrichment, isolation, and characterization of autoantigen-reactive B cells in a mouse model that spontaneously develops autoantibody-mediated systemic autoimmune disease (4, 48). We demonstrate that an IEDB-based peptide array of 2733 autoimmune disease–associated B cell epitopes can reveal linear epitopes on known autoantigens in BXD2 mice. Further, the age-related emergence of class-switched autoantibodies identified using the peptide arrays paralleled previous observations using standard whole-Ag ELISAs (4). Moreover, the serum reactivity of a panel of these identified autoepitopes was confirmed using ELISA and ELISPOT analysis of synthetic peptides. Peptide-reactive tetramer B cells also can be stimulated to produce IgG Abs that react with the full-length protein Ag. In the future, tetramer⁺ B cells and Ig genes from these cells would need to be analyzed at the single cell level to enable cloning and expression of the autoantibody to better determine the specificity of the tetramer⁺ B cells.

The peptides identified as BXD2 autoepitopes by the epitope array assay indicate which of these may be considered immunodominant. Potentially, this apparent immunodominance may simply reflect the array assay conditions. However, the reactivity pattern of BXD2 autoantibodies showed a preference toward nuclear and RNP autoantigens. Since initial observations of autoantigen clustering in apoptotic blebs (49), there has been growing awareness that chemical and structural modifications to autoantigens during cell death and neutrophil NETosis may provide B cells with access to normally concealed epitopes that may drive an aberrant adaptive-immune response (50–52). Multiple studies demonstrated that the majority of autoantigens targeted in systemic autoimmune diseases are substrates for granzymes, particularly granzyme B (53–55), and La itself is cleaved during apoptosis (56, 57). Similarly, the 70-kDa fragment of U1 snRNP, another hallmark autoantigen in patients with SLE, is specifically cleaved by caspase-3 during apoptosis. This cleavage converts the molecule into a truncated 40-kDa fragment and a smaller 96-residue C-terminal fragment (36, 58) containing the snRNP_357–373 epitope identified in this study as a major BXD2 autoantigen. Notably, this particular epitope is located in a structurally disordered region of the protein, a characteristic that may influence Ag capture, processing, presentation, and immune dominance during cellular processes, including apoptosis (59). The peptide array identification of these autoepitopes is consistent with our previous observations of defective apoptotic body clearance and the ability of MZ-P B cells to directly capture Ag derived from uncleared apoptotic debris in BXD2 mice (38).

The present peptide array results also provide a clear picture of the transition from autoreactive IgM to IgG in BXD2 mice and confirm that this transition is absent for most linear autoepitopes in B6 mice, as we described previously for full-length autoantigens (4). The autoreactive IgGs in BXD2 mice are highly pathogenic and form immune complexes that deposit in the kidney and joints (4). However, it should be noted that the current studies were focused on the identification of Ag-specific subsets of autoreactive B cells. It has not been confirmed that the peptide-specific IgG autoantibodies that we identified are capable of forming pathogenic immune complexes or eliciting tissue damage.

The IgM autoantibody pattern, which is often overlooked, also revealed some interesting features. Although both B6 and BXD2 mice produce broadly autoreactive IgM, the IgM repertoire in both mice exhibits specificity in that only some autoantigens, but not all, are recognized. Thus, in both strains, IgM appears to recognize a similar set of key autoantigens. These results are consistent with the observation that autoreactive and polyreactive autoantibodies are present in both healthy and autoimmune humans (44, 60, 61), an observation that is not limited to IgM but occurs commonly, even within the IgG memory B cell pool (62, 63). These results suggest that BCR ability to recognize and engage self-Ag does not necessarily cause disease. However, within a permissive environment, as in the BXD2 mouse, even small perturbations in B cell development, selection, or phenotype may drive normally benign autoreactivity toward an autotoxic response. They further support the concept that, although normal individuals may manifest some degree of autoimmunity to a set of self-Ags, this is benign as long as the regulatory capacity is intact (64).

Consistent with the detection of IgM autoantibodies in both B6 and BXD2 mice, most of the autoantigen tetramer⁺ B cells that are obtained directly from the spleen of a mouse are IgM⁺. The most likely reason for this is that there is a larger population of splenic B cells that expresses both surface and secretory IgM compared with a relatively small number of IgG B cells that express both surface and secretory IgG (65). However, in the spleens of BXD2 mice, there are decreased percentages of IgM^hi B cells compared to B6 mice. This is the case both within the pan-B cell population (39), as well as within the autoantigen-reactive tetramer⁺ B cell population (Fig. 5C).

Although there was an increase in the number of La and snRNP tetramer⁺ B cells in the spleen of BXD2 mice compared with B6 mice, the percentages were not significantly different. However, the phenotype of these self-reactive tetramer⁺ B cells in normal and autoimmune mice is different, which could be directed by autoantigen stimulation, the effects of type I IFN, and developmental differences. We propose that these phenotypic changes reflect common factors and events that are generally present in autoimmune disease.

The most prevalent population of tetramer⁺ B cells in both B6 and BXD2 mice is IgM^int FO B cells. These results are consistent with the finding that autoreactive B cells persist in the mature repertoire, even in normal mice (66). Interestingly, compared with B6 mice, there is a higher percentage of IgM^int tetramer⁺ B cells and a distinct population of IgM^lo tetramer⁺ B cells in BXD2 mouse spleens. In B6 mice, IgM^lo FO B cells were recently shown to be polyreactive B cells enriched for nuclear-reactive specificities (67). The presence of IgM^lotetramer⁺ B cells in BXD2 mice is consistent with our previous findings that pathogenic autoantibodies produced from BXD2 mice exhibit polyreactivity (4) and further suggests that these B cells may have derived from IgM^hi B cells that encountered chronic autoantigen stimulation.

Although La_13–27- and snRNP_357–373-reactive IgM^hi B cells from B6 mice are primarily the CD21^hiCD23^lo MZ B cells, the majority of these B cells in BXD2 mice display the CD21^hiCD23^hi MZ-P B cell phenotype. Furthermore, these Ag⁺ MZ-P B cells are skewed toward a hyperreactive CD86⁺CD69⁺ phenotype in BXD2 mice. In mice, B cells with regulatory function have been observed within the transitional, B1, and MZ compartments (68). We showed previously that, although the expression of Il10 and Tgfb was significantly higher, the expression of Il6 was significantly lower in MZ-P B cells of B6 mice compared with BXD2 mice (38). Consistent with this, CD23⁺CD21^hiCD1d^hi B cells were reported as the key pathogenic B cells in other mouse models of autoimmunity (69, 70). In contrast, Evans et al. (71) demonstrated that MZ-P B cells from spleens of healthy naive DBA/1 mice adoptively transferred into immunized DBA/1 mice significantly prevented and ameliorated disease. This and other reports (72) of MZ-P–like B cell regulatory functions suggest that, in the nonautoimmune state, these cells may indeed serve a regulatory role. Regulatory roles of human B cells are less understood, but transitional B cells with an IL-10–mediated regulatory function were similarly reported to be defective in patients with SLE (40). Increased CD80⁺ and CD86⁺ B cells also coincide with the observation that naive populations of B cells from SLE patients appear to be activated (73).

Analysis of the CD93⁺ transitional population of La_13–27⁺ or snRNP_357–373⁺ B cells in BXD2 mice further revealed that there is an abnormal expansion of T3 transitional B cells, a phenotype that has not been identified in other autoimmune mouse models (45). T3 B cells generally are not considered strictly transitional but, rather, anergic B cells maintaining self-tolerance through rapid turnover in vivo (74, 75). Interestingly, self-Ag stimulation was shown to promote the regression of mature B cells into the T3 compartment (75, 76). Thus, expansion of these cells coupled with abnormal T cell help (25, 77) may present a risk for B cell tolerance loss to La_13–27 or snRNP_357–373 in BXD2 mice. Consistent with the possible B cell tolerance defects that can occur before or at the MZ-P B cell stage in BXD2 mice, there is an approximate 3-fold increase in CD80⁺CD273⁺ memory/activated phenotype B cells in BXD2 mice compared with B6 mice. In contrast, the percentage of B cells with the CD80⁺CD273⁺ phenotype in BXD2 mice is enriched ∼6-fold in the La_13-27⁺ population and 20-fold in the snRNP_357–373⁺ population compared with the same populations in B6 mice. These results suggest that specific Ag stimulation can lead to expansion of an Ag-specific memory B cell population in BXD2 mice. CD80⁺PD-L2⁺ memory B cells were shown to differentiate rapidly into Ab-producing B cells upon rechallenge, without the need to go through another GC response (78). Thus, the present results may help to identify mechanisms related to the high titers of IgG autoantibody production in BXD2 mice.

The present proof-of-principle study establishes that the approach described in this work should provide a platform for integrating autoantibody profiles with underlying B cell defects. The ability to collect and compare comprehensive global autoantibody profiles, coupled with the usefulness of the tetramer approach, provides a feasible strategy to address other clinically relevant questions. For example, tetramer⁺ B cells and Ig genes from these cells should be analyzed at the single cell level to identify the differences between benign and pathogenic autoreactive Ig sequences. Such observations may be highly applicable for characterization of autoreactive B cells in human disease, because tetramer⁺ B cells that exhibit abnormal phenotypes can be detected in well-defined B cell subsets in human PBMCs (79).

We thank Dr. Fiona Hunter for critical review of this manuscript.

The authors have no financial conflicts of interest.