Negative selection by IgM superantigen defines a B cell central tolerance compartment and reveals mutations allowing escape

Abstract

To analyze B lymphocyte central tolerance in a polyclonal immune system, mice were engineered to express a superantigen reactive to IgM of allotype b (IgM^b). IgM^b/b mice carrying superantigen were severely B cell lymphopenic, butsmall numbers of Bcells matured. Their sera contained low levels of IgG andoccasionally high levels of IgA. In bone marrow, immature B cells were normal in number, but internalized IgM and had a unique gene expression profile, compared to thoseexpressing high levels of surface IgM, including elevated recombinase activator gene expression. A comparable B cell population was defined in wild-type bone marrows, with an abundance suggesting that at steady state ∼20% of normal developing B cells are constantly encountering autoantigens in situ. In superantigen-expressing mice, as well as in mice carrying the 3H9 anti-DNA Ig heavy chain transgene, or 3H9 H along with mutation in the murine kappa deleting element RS, IgM internalization was correlated with CD19 downmodulation. CD19^low bone marrow cells from 3H9;RS^−/− mice were enriched in light chains that promote DNA binding. Our results suggest that central tolerance and attendant light chain receptor editing affect a large fraction of normal developing B cells.IgH^a/b mice carrying the superantigen had a ∼50% loss in follicular B cell numbers, suggesting that escape from central tolerance by receptor editing from one IgH allele to another was not a major mechanism. IgM^b superantigen hosts reconstituted withexperimental bone marrow were demonstrated to be useful in revealing pathways involved in central tolerance.

Introduction

Immunoglobulin gene assemblyin developing B lymphocytes often initially generates self-reactive receptors. Autoreactive B cells can be regulated in several ways, including receptor editing, clonal deletion, and the induction of anergy,with attendant reduced B cell lifespan(1, 2). Editing is a major mechanism of central tolerance in developing bone marrow (BM)⁶ cells that in mice mainly involves secondary rearrangements onIGKloci that can eliminate one functional light (L)-chainrearrangement and exchange it for another, thus altering specificity(3–9)_ENREF_3. However, secondary L-chain rearrangements also occur when sIg levels are insufficient to suppress recombination, Ig H/L pairing fails to occur, or signaling through the innocuous BCR is impaired (10–14). In addition, IGHV replacement might also contribute to escape from central tolerance(15, 16). Experiments in autoantibody transgenic(Tg) mice and studies involving antibody cloning from single human B cells show that autoreactivity is progressively diminished during normal B cell development, and is sometimes flawed in autoimmune-prone individuals(4, 8, 17–24)_ENREF_3_ENREF_3_ENREF_3. However, the frequency in the BMof B cells that are initially autoreactive, and the extent to which central tolerance and editing contribute to their control are not known.

B cells that are unable to edit efficiently might have mechanisms for altering specificity besides V(D)J recombination. In many species, hypermutation or gene conversion can occur in developing B cells(25–27). Although these pathways are minor in the mouse(28, 29), low levels of AID expression, class switching and somatic mutation occur in normal immature B cells. AID activity can be upregulated even in preB cells(30, 31).B cells of µMT/µMT mice, which lackIgM membrane exonsand exhibit a severe block in B cell development at the preB cell stage, can occasionally undergo class switch to downstream isotypes(32–35). However, what roles, if any, AID may play as a tolerance mechanism have not been investigated.

To control and visualize B cells undergoing central tolerance in a polyclonal immune system, we previously developed κ-macroself Tg mice, which express a superantigen reactive to Cκ. In these mice, there was efficientκ-to-λL-chain editing in the BM leading to significant escape of B cells carrying λ-chains(3). Here, we generated mice expressing an IgM^b superantigenderived from mAbAF6–78(36).We predicted that on anIgM^b/b backgroundL-chain editing would be ineffectual in eliminating superantigen reactivity, and that tolerance should either promote deletion and anergy, or reveal in the surviving cells a different type of receptor selection. IgM^b-macroself mice offer a model system to determine the phenotype of developmentally blocked “autoreactive” B cells that are otherwise normal in their Ig gene expression and editing. The modelallowed us to identify a similar population in normal mice that provides an estimate of the normal extent of central tolerance and receptor editing.

Materials and methods

IgM^b-macroself construct

IgM^b specific hybridoma AF6–78 was purchased from ATCC (Manassas, VA). The transgene encoding the IgM^b-macroself antigen was generated using methods essentially as described (37). Briefly, total RNA from AF6–78 was isolated using Trizol (Invitrogen, Carlsbad, CA) according to manufacture’s instruction. VL and VH cDNA were obtained by 5’-RACE (Ambion, Austin, TX) using Cκand Cγ antisense primers and subcloned into Zero Blunt TOPO vector (Invitrogen) and the sequence was determined. Leader (pUbF and iLAF6R), VL(AF6_VLF and AF6_VLR) and VH(AF6_VHF and AF6_VHR) encoding fragments were amplified. Purified fragments were fused using overlap PCR (pUbF and AF6_VHR) and cloned into SpeI and XmaI digested pUIiκ plasmid (3). Primer sequences used were as follows: Cκ(5’-ctgctcactggatggtgggaagatgg-3’); IgGs (5’-gctggacagggatccagagttcca-3’); pUbF(5’-ttttctccgtcgcaggacgcagggttcggg-3’); ILAF6R(5’-gtgagaacaatttgtgcacaggcacctgtaataattaataggc-3’);AF6_VLF (5’-tacaggtgcctgtgcacaaattgttctcacccagtctccagcaatcatgtctgcatctcc-3’); AF6_VLR (5’- cctcccgagccaccgcctccgctgcctccgcctccccgtttcagctccagcttggtcccag-3’); AF6_VHF (5’- cagcggaggcggtggctcgggaggcggaggctcgcaggtccagctgcagcagtctggggc-3’); AF6_VHR (5’-cccgggtttctgggggctgttgtttcagctgaggagacggtgactgaggttccttgac-3’).

Mice

Eight to 12 week-old mice were used in most experiments. C57BL/6 were purchased from The Scripps Research Institute (TSRI) breeding colony. Pronuclear injection was performed at TSRI Mouse Genetics Core facility using (C57BL/6 x BALB/cByJ)F2 mice. Three different lines (numbers 17, 68 and 72) were backcrossed to C57BL/6.Ly5a for further study. Based on the transgene expression, line 72 (low expresser) and line 17 (intermediate expresser) were used for most of the experiments. All IgM^b-macroself Tg mice had been bred>10 generations and maintained on C57BL/6.Ly5a background. κ-macroself Tg mice (pUIiκ line 2) were described (3). IgH^a/a mice (B6.Cg-Igh^a Thy1^a Gpi1^a/J) were purchased from Jackson Laboratories. Site-directed 3H9 H chain Tg mice on a C57BL/6 background were kindly provided by Dr. Martin Weigert (University of Chicago).Bcl2 Tg mice (line Eµ2–22) were kindly provided by Andreas Strasser and Alain Harris (Walter and Eliza Hall Institute). hCκ targeted mice (38) were kindly provided by Michel Nussenzweig (Rockefeller University). QM mice were kindly provided by Marilia Cascalho (University of Michigan). All mice were maintained at our facility.

Radiation chimeras

In BM transfer experiments, each mouse host was lethally irradiated with 1000 Rad one day prior to receiving i.v. 10⁷ BM cells, which were isolated from both upper and lower leg bones of donors using standard protocols. Donor mice were C57BL/6.IgH^a/b congenic mice.

Serum Ig ELISAs

Nunc Maxisorp plates (Nunc, Roskilde, Denmark) were coated with 2t;µg/ml antibody to mouse IgM (M41) or rat IgG-adsorbed donkey antibody to mouse IgG (H+L) (Jackson ImmunoResearch Laboratories) overnight at room temperature, blocked with 1% BSA in Tris buffered saline supplemented with 0.1% Tween-20 (ELISA buffer) for 1 h, and incubated for 1 h with mouse sera diluted in ELISA buffer. Horseradish peroxidase (HRP)-conjugateddonkey antibody to mouse IgM, or goat antibody to mouse IgG, Fcγ fragment specific (Jackson immune Research) antibodies were used to detect bound mouse Ig. For IgA ELISA, plates were coated with rat antibody to mouse IgA (11-44-2, eBiosciences (eBio), San Diego, CA) and biotinylated antibodies to κ and λ was used as secondary reagents, followed by incubation with streptavidin-HRP (BD Biosciences, San Diego, CA). Bound HRP was developed with Ultrasensitive TMB substrate (Millipore, Billerica, CA) per the manufacturer’s instructions. Absorbance signals were recorded at 450 nm using VersaMax microplate reader (Molecular Devices, Sunnyvale, CA).

Flow cytometry analysis

Flow cytometric analyses for surface markers were performed using standard protocols with appropriately diluted antibodies. Intracellular proteins were stained after permeabilization using Cytofix/ Cytoperm Kit (BD Biosciences) according to manufacturer’s recommendations. The following mAbs were used for the experiments: B220 (RA3–6B2; Pacific Blue (PB), FITC or APC; eBio); CD19 (1D3; PE or PE/Cy7; eBio); CD21 (7E9; FITC, Biolegend, San Diego, CA); CD93 (AA4.1; PE or APC; eBio); IgD (11–26; PE; eBio); IgM^b(AF6–78; FITC; Biolegend); TCR-β (H57; PerCP-Cy5.5; eBio); CD11c (N418; PerCP- eFluor® 710; eBio); PDCA1 (PDCA-1; PerCP- eFluor® 710; eBio); mAbs against mouse IgM (M41; PB, Alexa488 or Alexa647;331.12; Alexa 647), and κ (187.1; Alexa 647) were labeled in-house. All samples were run on an LSR-II instrument (BD) and analyzed using the FlowJo program (Tree Star, Ashland, OR).

Cell sorting

Bone marrow cells were harvested from three IgM^b-macroself, 3H9;RS^−/− or WT mice. B220⁺ cells were enriched with CD45R (B220) MicroBeads (Miltenyi Biotec, Auburn, CA) and stained with B220-FITC, CD19-PE, IgM-Alexa647 and dump channel mAb cocktail (H57-PerCP-Cy5.5 and PDCA1-PerCP- eFluor® 710). Subsequently B220^intCD19^lowIgM⁻, B220^intCD19^intIgM⁻ and B220^intCD19^intIgM⁺ fractions were isolated with a FACSAria (BD) sorter.

Quantitative PCR (qPCR)

Total RNA was purified from half to 1 million sorted cells using a RNEasy Plus kit (Qiagen, Valencia, CA). Reverse transcription was performed with a QuanteTect Reverse Transcription Kit (Qiagen) per the manufacturer’s protocol. Each gene was quantitated using SsoFast EvaGreen (BioRad, Hercules, CA) with 7900HT (Applied Biosystems, Carlsbad, CA) and normalized withβ-actin. Oligonucleotide primers forRag1 and Rag2 were used as previously reported (37). Other primer sequences were obtained from Origene.

5’-RACE and κ chain analysis

Total RNA was obtained from sorted cells and κ-chain variable sequence was determined as previously described (39). Briefly, adaptor was ligated to total RNA using 5’-RACE kit (FirstChoice RLM-RACE, Invitrogen) according to the manufacturer’s protocol, followed by reverse transcription using oligo dT primer with a Transcriptor High Fidelity cDNA synthesis kit (Roche, Indianapolis, IN), and PCR amplification with oligonucleotides corresponding to adapter and Cκ sequences. Amplified κ-chain variable product was plasmid cloned and the insert sequences of individual clones were determined. Obtained sequences were analyzed at the ImMunoGeneTics (IMGT) Web site (www.imgt.org)(40).

Results

Anti-IgM^b mAb AF6–78 fails to see the preBCR

To assess the ability of AF6–78 to bind to the preBCR, in which the µ^b-chain is associated with surrogate L-chain components, we took advantage of a preB cell line (70Z/3) that expresses a preBCR, but which can be induced to produce κ-chain and sIgM upon LPS activation(41). Flow cytometry analysis of permeabilized cells using mAb M41, an antibody reactive with µ-chain regardless of its associated L-chains(42), confirmed that unstimulated 70Z/3 cells express µ-chain but no κ-chain, (Fig.1A, left panel). By contrast, AF6–78 failed to react to unstimulated cells (Fig. 1B, solid lines), but bound avidly to LPS-stimulated 70Z/3 cells (dotted lines), indicating that AF6–78 only detects µ^b associated with conventional L-chain. To further assess the inability of AF6–78 to bind to the preBCR, we captured preBCR complexes from detergent lysates of unstimulated 70Z/3 cells on ELISA plates coated with antibodies to IgM and measured binding by AF6–78, M41, and antibodies to κ or the preBCR (Fig.1C). Again, AF6–78 failed to bind, whereas there was clear binding by M41 or antibodies to λ5 or the preBCR-specific antibody SL156. We conclude that AF6–78 has negligible affinity for µ^b-containing preBCRs, but binds well to IgM^b.

Figure 1.

Negligible reactivity of IgM-reactive mAbAF6-78to the preBCR and the generation of IgM^b-macroselfTg mice. (A) 70Z/3 murine preB lymphoma cells(41) were either left unstimulated or treated with 10 µg/mL LPS for 24 hours to induce expression of IgκL-chain. Cells were subsequently permeabilized and stained for total expression of IgM and Igκ. (B) Affinity of AF6-78 mAb for the IgM^b heavy chain before and after LPS induction was assessed by flow cytometric analyses of surface and total IgM expression. (C) ELISA was performed to assess relative affinities of antibodies to immunoglobulins for the captured pre-BCR. Plates were coated with goat F(ab’)2 antibodies to mouse IgM. 70Z/3 lysates containing the expressed pre-BCR were subsequently serially diluted and applied to each well. Detection was performed using equal concentration of each indicated antibody conjugated to biotin, and developed as described in Methods. The above experiments were repeated at least twice. (D) Design and features of IgM^b-macroself antigen construct including intron/exon structure, ubiquitin C promoter, leader exon (L), variable light (V_L) and heavy (V_H) codons, a linker peptide of the following sequence (GGGGSGGGGSGGGGS), rat IgG1 hinge, C_H2 and C_H3 sequences, and transmembrane and cytoplasmic sequences from H-2K^b. (E) Schematic representation of the predicted protein structure of membrane bound anti–mouse IgM^b-macroself Ag. Single-chain Fv generated from the IgM^b-reactivehybridoma AF6-78 is linked to the hinge and membrane proximal domains of rat IgG1 followed by transmembrane and cytoplasmic tail regions (Tm/Cy) of H-2K^b. (F) Flow cytometry analysis of superantigen expression in the tissues from transgenic line 17. Comparable results were obtained for at least two additional transgenic lines. Superantigen was detected with antibody to rat IgG1Fc.

Generation of an IgM^b-reactive superantigen and expression in Tg mice

Antibody genes derived from AF6–78 were then cloned and used to generate a single-chain antibody-coding construct (Fig.1D), based on theκ-macroself construct(3), which is predicted to form a dimeric membrane protein carrying C-terminal H-2K^b transmembrane and cytoplasmic domains (Fig.1E). Rat IgG1 H-chain hinge, CH2 and CH3 domains provide flexibility and facilitate superantigen detection. Several founder linesof Tg mice expressing this constructwere generated(e.g., Fig.1F)and bred over 10 generations onto the B6.CD45.1 background. All lines expressing superantigen had a similar phenotype. Studies in this paper mostly involved IgM^b-macroselflines72and 17, which gave nearly identical results.

Developmental block and antigen receptor downregulation in BM B cells

Flow cytometry was used to analyze BM, spleen (SP), lymph nodes(LN), and peritoneal cavity (PC) of IgM^b-macroself mice for B cell numbers and phenotype. Surface IgM-positive(sIgM⁺) cells were absent and B cell deletion was apparent from the reduction of B220⁺ cells in SP and LN (Fig.2A, center column). Staining with the pan B cell marker CD19 along with B220 (CD45R) revealed that B cell numbers were reduced to <1% of wild type in secondary lymphoid organs of IgM^b-macroself mice (Fig.2B–E). By contrast, κ-macroself mice had significant numbers of B cells in the periphery owing to Igκ-to-λ editing(3)_ENREF_25 (Fig.2A,B, right panels).Consistent with the massive reduction of peripheral B cells, recirculating CD19⁺B220^hi cells were absent from IgM^b-macroselfBM (Fig.2A,B top middle panels). Moreover, IgM^b-macroself mice had greatly reduced levels of serum IgM, IgG, and IgA. However, IgG and IgA levels were clearly above background, with 2of 12mice expressing IgA in the normal range and the mean level of IgG at ∼10µg/ml (Fig.2F,G). Similar findings of B cell deficiency and low serum Ig levels were obtained with IgM^b-macroselfline 17 (data not shown).Because the mice were always bred using IgM^b-macroself males and wild-type females, some serum Ig may have been maternal. However, we could generate from LPS-stimulated splenocytes pooled from threeIgM^b-macroselfmiceseveral IgG-secreting hybridomas, suggesting that some IgG came from cells that had avoided deletion by class switching (not shown). The small number of CD19⁺ cells in the spleens of IgM^b-macroselfTg mice appeared to be composed of a B220^hi population that had markers of follicular B cells butlittle detectable surface Ig L-chain(Fig.2H,I, left panels)or expressed lower levels of B220 and carriedL-chains (right panels).This latter population had a B1-like phenotype as it expressed CD43 and CD5 and low levels of IgD (Fig.2J). We conclude that B cells are profoundly reduced in number in IgM^b-macroself Tg mice, but that some cells survive and differentiate to secretion of class-switched antibodies.

Figure 2.

B cell deletion and hypogammaglobulinemia in IgM^b-macroselfTg mice. Lymphoid tissues from WT, IgM^b-macroself, or κ-macroself Tg mice were analyzed by flow cytometry using the indicated antibody combinations. Plots shown were gated on TCRβ lymphocytes (H57). (A–C) Flow cytometric analysis of B cells in bone marrow (BM), spleen (SP), lymph nodes (LN), and peritoneal cavity (PC) using antibodies to B220, IgM and CD19. (A) Top panels indicate with left, right bottom and right top boxes, B220^intsIgMpreB/immature B cells, B220^intsIgM⁺ immature B cells and B220^hiIgM⁺ recirculating B cells, respectively. (B) Top panels indicate with left, middle and right boxes B220^lowCD19^low B cells, B220^intCD19^int B cells and B220^hiCD19^hi recirculating B cells, respectively. (C) Analysis of PC B cells for CD19 and IgM expression (left panels) and CD19 and B220 (right panels). Analysis gates in lower panels define B2 vs B1 cells in upper vs lower gates, respectively. (D) Statistical analysis of total cell numbers in spleen and lymph nodes. WT, n=7; IgM^b-macroself, n=8. (E) Abundance of B1 and B2 cells in peritoneal cavity expressed as percent of viable cells. Each data point indicates the value in one mouse. (F) Serum IgM and IgG concentrations in WT and IgM^b-macroself Tg mice. WT, n=7; IgM^b-macroself, n=8. (G) Serum IgA concentrations. WT, n=7; IgM^b-macroself, n=12.Each data point indicates the value in one mouse. Shown are means and SEM. *, P < 0.05; **, P < 0.005; ***, P < 0.0005; ****, P<0.0001 (two-tailed Student’s T test).(H,I) Analysis of splenic B cells of IgM^b-macroselfmice. Cells gated as indicated in the central panel of part B were analyzed for (H) L-chain isotype or (I) CD21 and CD23 expression. (J) Cell surface phenotype of IgM^b-macroself B220^hi vs B220^low splenic B cell subsets compared to B220^hi wild type cells.

BM B cells of IgM^b-macroself mice were characterized in more detail. In addition to the lack of CD19⁺B220^hi recirculating cells, CD19^intB220⁺ cells were somewhat reducedin number compared to wild type, and there was aconcomitant increase in the numbers of CD19^lowB220⁺ cells(Fig.2B, top middle panel; Fig.3A).The CD19^lowB220⁺ cells appeared to be immature B cells that had downregulated sIgM because they expressed high levels of intracellular µ-chain and κ-chain (Fig.3B,C right panels). Moreover, intracellular IgM (iIgM) in CD19^lowB220⁺ cellswas recognized bymAb 331.12, which requires L-chain pairing (Fig.3D). By contrast, CD19^intB220⁺ cells in IgM^b-macroself Tg mice werepreB cells as indicated by their lack of cytoplasmic staining with 331.12 and antibody to κ, and by the expression of λ5 by a subset of cells (Fig.3C,E). In all respects, comparable results were obtained in analyses of line 17 (data not shown). Immature IgM^b-macroselfB cells with iIgM had CD19 levels only ∼1/10 ofthat of sIgM⁺ wild type cells (Hardy fraction E), whereas wild type cells that scored as iIgM⁺sIgM⁻ included a significant subset that had low CD19 levels (Fig. 3F).Thus,in two independent transgenic lines IgM^b-macroself antigen blocked B cell development at the immature B cell stage, leading to BCR and CD19 downregulation.

Figure 3.

Analysis of CD19⁺B220⁺ BM subpopulations gated as indicated in Fig.2B, top panels. (A) Statistical analysis of absolute numbers of CD19^low, CD19^int and CD19^hi B cells in bone marrow from five WT and five IgM^b-macroself mice.Shown are means and SEM. *, P < 0.05; ****, P<0.0001 (two-tailed Student’s T test). (B) B220⁺CD19 subpopulations from wild type or IgM^b-macroself mice were assessed for total (intracellular and surface) IgM, using an antibody (M41) that recognizes µ-chain in isolation or associated with conventional or surrogate L-chains. Solid gray histograms show CD19 levels on H57⁺ (TCRβ⁺) cells as a negative staining control. (C) The indicated B220⁺CD19 subpopulations from wild type or IgM^b-macroself mice were analyzed for intracellular staining with antibodies to κ. (D) Costaining of control,IgM^b-macroself, and κ-macroselfBM lymphocytes for expression of CD19 and IgM, using an IgM-reactive antibody (331.12) that requires µ-chain association with conventional L-chain for binding. (E)Analysis of CD19^intB220⁺ gated, permeabilized cells for expression of λ5 and IgM (331.12). (F) CD19 levels are shown for BM immature B cells from wild type or IgM^b-macroself mice. CD19 levels on B220^intermediate B cells expressing intracellular IgM detectable with 331.12 (iIgM) or both sIgM and iIgM were compared.

Identification of a natural BM population under negative selection

The foregoing experiments suggested that the B cells in wild typeBM undergoing central tolerance should have properties similar to B cells in IgM^b-macroself mice, namely intracellular IgM (iIgM) associated with conventional L-chain, but little surface IgM expression. By contrast, “innocuous” cells failing to see self-antigen in BMshould permit surface expression. To assess this, we compared wild type mice with IgM^b-macroself mice, in which we assume 100% of B cells are initially autoreactive,and κ-macroselfmice carrying either one or two reactive IGK alleles. On an otherwise wild type background, κ-macroself mice start out with >90% autoreactive cells (i.e., mouse κ⁺), while in κ-macroself mice carrying one mouse Cκ (mCκ) allele replaced byhuman Cκ(38)(hCκ) only half as many cells carry mouse κ. Using a flow cytometry strategy involving surface staining with anti-µ;MAb M41, followed by permeabilization and staining with 331.12 carrying a distinct fluorophore, we quantitated the frequency of iIgM⁺sIgM⁻ B cells (Fig.4A).In IgM^b-macroselfmice, as expected,nearly 100% of B220⁺iIgM⁺ cells lacked sIgM (Fig.4B, second panel). As a control for “innocuous” cells, we chose QM mice, which are IGK-deficient and express a transgenic H-chain that is considered to be non-autoreactive when paired with endogenous λ-chain(43). In these mice, only 20% of iIgM⁺ cells were sIgM⁻(Fig.4A, right panel;Fig.4B). Assuming that this proportion represents phenotypic lag between intracellular H/L assembly and surface expression, and that 100% IgM internalization represents 100% negative selection, we assessed the percentages of iIgM⁺ cells that were sIgM⁻ in WT mice and in κ-macroself mice that had either two mCκ alleles or one mCκ allele and one hCκ allele(38)(Fig.4A,B). In κ-macroself mice, ∼70%of iIgM⁺ cells were sIgM⁻, whereas this value was about 50% in κ-macroself;hCκ/mCκ. In fully wild type mice and in hCκ/mCκ mice lacking superantigen,the measured value was ∼40%of iIgM⁺ cells that were sIgM⁻.After subtracting 20% for phenotypic lag, these results suggest that ∼20% of normal developing B cells are undergoing central tolerance in the BM at steady state. CD19 downregulation was partly correlated with IgM internalization, being high in IgM^b-macroself mice and progressively lower in mice with fewer autoreactive cells(Fig.4C). However,the frequency of CD19^low cells was consistently lower than that of iIgM⁺sIgM⁻ cells. To see if B cells undergoing negative selection had a distinct RNA expression profile, we chose to sort CD19^low cells from IgM^b-macroself mice, as this strategy does not require fixation of cells and provides a good separation from CD19^int cells, which lacked an intact IgM BCR. Negativelyselecting CD19^low cells of IgM^b-macroself mice had a unique transcription profile including an intermediate level of expression of Rag1,Rag2, Foxo1, Il2ra, and Il7r, reduced expression ofCd36 andCcnd2, and high expression of Faim3(Fig.4D,E). Surface markers correlated with these changes in the cases of Il2ra (CD25), Il7r (CD127) (Fig. 4F) and Cd36 (CD36). These patterns distinguished them from CD19^intsIgM⁺ and CD19^intsIgM⁻bulk wild type populations.Rag1 and Rag2 mRNA levels in negatively selecting cells were intermediate between CD19^intsIgM⁻ and CD19^intsIgM⁺ cells, which had high and low levels, respectively. IgM internalization thus identifies a BM populationundergoing negative selection,and these cells have a distinctive gene expression profile.

Figure 4.

Analysis of BM B cells that internalize IgM in wild type and Tg mice. (A) BM cells of the indicated strains were analyzed for expression of surface and intracellular IgM, using mAb 331.12 to detect conventional intracellular IgM (iIgM) expression. Cells were gated to exclude recirculating and nonB lineage cells. Note that since QM B cells express only λ-chains, the plot demonstrates that 331.12 binds well to IgM composed of µ and λ-chains. (B) Analysis of the percentages of iIgM⁺sIgMBM cells as a function of total IgM⁺ immatureB cells as detected using mAb 331.12 (=100 x [sIgM⁻iIgM⁺]/[sIgM⁻iIgM⁺ + sIgM⁺]). (C) The CD19^low frequency among iIgM⁺ immature B cells was calculated(=100 x [CD19^low]/[sIgM⁻iIgM⁺ + sIgM⁺]). In B and C each dot represents the value measured in an individual mouse, with the mean represented by a horizontal bar.(D,E) Shown is quantitative RT-PCR expression analysis of selected mRNAs in the indicated BM B cell subsets. Expression levels found in wild type sIgM⁻CD19^int were set to 1 and relative levels given for other samples. (F) Staining profiles of cells from the indicated wild type (WT) or IgM^b-macroself BM B cell subpopulations using antibodies to the indicated cell surface markers. Data were from at least three mice per group.

To further test the validity of this approach, we analyzed additional mouse models lacking superantigen, but with altered extents of self reactivity and receptor editing, including mutants in the murine kappa deleting element RS(44) and mice carrying the3H9 site-directedH-chain transgene(4). RS^−/− B cells have defective editing of a subset of autoreactive B cells, particularly those that use Jκ5 (44). 3H9 H-chain B cells generate a BCR reactive to DNA/chromatin when paired with many L-chain partners, albeit with diverse affinities, leading to more extensive editing than normal (4, 8, 45, 46). 3H9 and 3H9;RS^−/− mice had an elevated frequency of CD19^lowBM cells(Fig.5A,C, E) and the percentages of iIgM⁺CD19^lowsIgM⁻and iIgM⁺sIgM⁻cells was as high or higher than in wild type (Fig.5B,D,E). B220⁺CD19^low cells were sorted from 3H9;RS^−/−BMand their expressed κ genes sequenced after 5’ rapid amplification of cDNA ends (5’-RACE) PCR cloning. Jκ usage analysis showed that the CD19^int sIgM⁺ population had skewed usage, favoring Jκ5 compared to CD19^low cells (Fig. 5F).Jκ5 usages of both samples were also significantly skewed compared to wild type peripheral B cells (39, 47). As predicted, IGKV genes that confer strong autoreactivity when paired with 3H9 H-chain were highly enriched in the CD19^low population, whereas IGKV genes predicted to edit autoreactivity were enriched among CD19^intsIgM⁺ cells(p <0.0001 by χ² by analysis of Vκ genes seen more than once in the combined samples). In particular IGKV1–135, IGKV10–96, andIGKV1–110were highly abundant in CD19^low cells, whereas editors such as IGKV17–127, IGKV6–17, and IGKV13–85 were enriched in CD19^intsIgM⁺ cells (Fig.5G,”a” vs “e”, respectively).IGKV genes predicted to confer weak reactivity, but not to induce editing, were also enriched among CD19^intsIgM⁺ cells (marked “w”).;These data demonstrate that cells internalizing IgM and CD19 are mainly composed of autoreactive cells undergoing central tolerance,not only in macroself Tg mice, but also in models that see physiological autoantigens.

Figure 5.

Increased numbers and distinct κL-chain usage in CD19^lowB220⁺BM cellsof 3H9;RS^−/−mice.(A) Flow cytometry analysis of CD19 and B220 expression of BMlymphocytesfrom the indicated mouse strains gated to excludeTCRβ⁺and plasmacytoid dendritic cells using H57 and PDCA-1 mAbs, respectively. (B) Quantitation ofiIgM⁺and CD19^low cells among the B220⁺sIgM⁻ population. (C) Plot showing the CD19^lowB220⁺ frequency within the immature B cell populationof the indicated strains with each point representing the value from one mouse. (D,E) iIgM⁺and CD19^lowiIgM⁺ frequencies, respectively, plottedas a percentage of(sIgM⁺ + sIgM⁻iIgM⁺) cells. (WT, n=11; RS^−/−, n=6; 3H9, n=8; 3H9;RS^−/−, n=8). Each point represents the value obtained from an individual mouse. Horizontal lines represent means of the individual values. ns; non significant, **; p<0.005, ***; p<0.0005.(F,G) CD19^lowB220⁺sIgM⁻ and CD19^intB220⁺sIgM⁺ were sorted from three 3H9;RS^−/− mice. RNA was extracted and κ-chain transcripts amplified by 5’-RACE. PCR fragments were cloned and Vκ and Jκ usage usages analyzed. Total sequences analyzed were CD19^lowB220⁺IgM⁻, n=78; CD19^intB220⁺IgM⁺, n=82.

Deletion and homeostasis of B cells in IgH^a/b;IgM^b-macroself mice

Analyses of the effects of superantigen on the B cell compartment in IgH^a/b mice were carried out both by introducing the line 17 transgene by breeding and by the use ofIgH^a/b→IgM^b-macroselfradiation BM chimeras. These experiments recapitulated the finding that mature IgM^b cells were deleted in the presence of superantigen (data not shown), but revealed subtle features of homeostasis among the IgM^a+ populations. In IgH^a/b;IgM^b-macroselfmice splenic marginal zone (MZ) and B1 cells that were IgM^a+developed in normal numbers, whereas B2 cell numbers were roughly half of normal (Fig.6A–E). Interestingly, however, in IgH^a/b→IgM^b-macroself chimeras, similar results were obtained with B2 and B1 cells, but MZ B cell numbers were also only half of normal (Fig.6F–K). We interpret these results to mean that B2 cell number is regulated mainly by BM output, whereas non-BM (presumably fetal)-derived MZ and B1 cells are regulated by homeostatic factors independently of adult BM-derived MZ and B1 cells.

Figure 6.

Effects of IgM^b superantigen on B cell homeostasis in a IgH^a/b background. (A–E) Analysis of peripheral B cell subset numbers in IgM^b-macroself; IgH^a/b mice generated by breeding. IgM^b-macroself line #17 mice analyzed were 9 months old. (A) Flow cytometry analysis of SPs revealing reduced B-2 compartment and disproportionate percentage of MZ B cells in IgM^b-macroself mice compared to IgH^a/b littermate controls. Left panels show gating scheme, with MZ B cells defined as CD21^hiCD23^lo cells in lower panels. In those plots B-2 cells are CD23⁺ and have a lower density of CD21. (B,C) Absolute numbers of total or IgM^a B-2 and MZ B cells in the indicated mice. (D) Analysis of B-1 vs B-2 cell populations in the peritoneum using CD19 and B220 costaining, showing gating scheme for B-1 cells (B220^loCD19^hi) and B-2 cells (B220^hiCD19^lo). (E) Shown are numbers of total or IgM^a B-1 and B-2 cells. (F–K) Analysis of B cell subsets in peripheral lymphoid tissues of IgH^a/b→IgM^b-macroself radiation chimeras. (F) Schematic of experimental strategy. Tg or control recipient mice were reconstituted with IgH^a/bBM as described in Materials and Methods. (G–K) B cell recoveries in the indicated chimeras for (G) MZ subset, (H) total SP B cells, (I) total LN B cells, and (J,K) B2 and B1 peritoneal subsets, respectively. The indicated B cell subsets identified by flow cytometry and scored for IgM allotype were enumerated in the indicated tissues. Open boxes and open circles show, respectively, total and IgM^a+ B cells in wild type recipients. Filled circles show total (IgM^a+) B cells in IgM-macroself recipients. Horizontal and vertical hash marks indicate means and SEM, respectively; n=5 mice/group.

Rescue by enforced Bcl2 expression in B cells

When autoreactive B cells are blocked at the BM stage and unable to edit, their loss is believed to take place through an apoptotic process inhibitable by Bcl2(48, 49). To assess this pathway, we bred IgM^b-macroself mice to mice expressing Bcl2 in the B cell compartment (Bcl2 Tg)(50). Young adult Bcl2Tgmice have a modest increase in B cell numbers compared to wild type mice (Fig.7A, left two columns), whereas IgM^b-macroself mice showed the predicted loss of peripheral B cells (third column). However, enforced Bcl2 expression in IgM^b-macroself;Bcl2 Tg micesubstantially rescued B cell numbers in the SP, LN and PC (Fig. 7A,B, right columns). Most of these B cells were still sIgM⁻.B cell maturation was evident also in the appearance in BM of cells with a recirculating phenotype (top right panel). Interestingly, the CD19^low population was dramatically expanded in IgM^b-macroself;Bcl2 Tg BM. Elevation of IgG but not IgM levels was clearly apparent in IgM^b-macroself;Bcl2 Tgsera(Fig.7C). The IgG generated was skewed to an excess usage of λ, suggesting a selective escape of B cells that previously attempted to edit their receptors.

Figure 7.

Effect of transgene-enforced Bcl2 expression on peripheral B cell elimination by IgM^b-superantigen. (A,B) Flow cytometry analysis of B cell numbers in the indicated lymphoid compartments of control, Bcl2 Tg or IgM^b-macroself single Tg, or double Tg (DTg) IgM^b-macroself;Bcl2 Tg mice. Data are representative of at least three independent experiments.(A) staining with antibodies to B220 and CD19; (B) staining with antibodies to IgM and CD19. Gating was as in Figure2. (C) SerumIg levels measured for IgM and IgG associated with κ or λ L-chains in the indicated mouse strains. Mice were 8–9 months old.

Discussion

Not just a poor man’s B cell knockout

We find in this study that ubiquitous expression of IgM superantigen leads to significant B cell deletionwith little escape. IgH^b/b;IgM^b-macroself mice are severely deficient in B cells, starting from the transitional stages of splenic B cell development, but early B cell developmental stages are unaltered because the superantigen fails to bind to the preBCR. Depletion of mature B cellsin these mice renders them hypogammaglobulinemic. The mice are susceptible to Pneumocystis carinii infection (not shown). Conventional receptor editing, VH replacement or V-region hypermutation cannot rescue IgH^b/bcells from reactivity to such a superantigen, potentially revealing alternative mechanisms of receptor alteration or escape. In this case, class switchingappears to allow a small number of B cells to develop.In many species, developing B cells diversify their receptors through expression of AID, which is also required for initiating H-chain class switch(51, 52). In mice, low, but detectable, levels of AID expression, class switching, and somatic mutation occur in immature B cells and in retrovirally infected preB cells(28, 31, 53). Under some circumstances, such as on theFas^lpr/lprbackground, B cells develop in µMT/µMT mice owing to class switch recombination to downstream isotypes in preB cells(32–35). However, Fasmutation did not have any obvious effects on B cell development in IgM^b-macroself mice (not shown). Overall, apart from the exception just mentioned, the phenotype of IgM-macroself mice is similar to that ofµ MT/µMT mice. As such, IgM^b-macroself mice provide an alternative way to render mice substantially, but not completely, B cell-deficient. In this respect, IgM^b-macroself mice provide advantages in that the effect is genetically dominant, and depletion can be effected in the setting of radiation BM chimeras using IgM^b-macroself hosts.Most importantly, however,IgM^b-macroselfchimerasshould beparticularly useful in identifying genes and mechanisms involved in regulating central tolerance.

Downregulation of IgM and, to a lesser extent, CD19 mark editing cells in the BM

By contrast to the depleting effects of IgM superantigen on mature peripheral B cells, immature B cells are abundant in the BMs of IgM^b-macroself mice. These cells downregulate CD19 and carry a significant level of intracellular IgM, detectable by staining permeabilized cells.Interestingly, a comparable, but less abundant, iIgM⁺sIgM^lowB220⁺, and partly CD19^low,population is apparent in normal BM.We propose that this population includes a high proportion of naturally autoreactive, editing cells.It has been suggested that such cells in humans, which have a high frequency of autoreactivity and polyreactivity,are “early immature” B cells, implying that they represent developmentally early B cells prior to antigen selection,though in that study they did not show the predicted skewing to reduced λ or upstream Jκ usage(18). Our results suggest rather that many or most such cells represent immature B cells that have internalized receptor upon BCR ligation. Jκ5 usage predominates in this population, which is clearly inconsistent with an early immature phenotype. Cells with downregulated BCR and CD19 are abundant not only in superantigen mice, but also in autoantibody Tg micewhere many cells are autoreactive(11, and this study, 54).CD19 downregulation may be of some functional importance because CD19 phosphorylation and attendant PI3K activation is known to contribute to the downregulation of editing (11, 55–59). By contrast, Ig transgenic mice generating “innocuous” receptors often have a reduced sIgM⁻ BM compartment owing to accelerated B cell development(54, 60, 61).

3H9H–chain Tg mice are predicted to generate autoreactive receptors at a higher than normal frequency due tothe antibody’s predisposition to bind to chromatin, DNA, and other self antigens, which leads to increased L-chain gene editing and bias to downstream Jκ usage(8, 46).However, IgM internalization in immature 3H9B cells was comparable to wild type, while the CD19^lowfraction was significantly increased. Similarly, RS^−/−B cells appear indistinguishable from normal in IgM and CD19 downregulation, but in combination with the 3H9 gene have an increased proportion of CD19^low cells. In 3H9;RS^−/−BM, autoreactive cells are highly enriched in the CD19^lowsIgM⁻iIgM⁺ fraction. Our working hypothesis for the pattern of IgM and CD19 modulation is that CD19 downregulation requires stronger or more prolongedBCR stimulation than is required for IgM downregulation alone. Consistent with this, Pesando et al. compared levels of sIg and CD19 in B cell lines treated with BCR ligands, and found that downregulation of CD19 was slower, less complete, and required higher ligand concentrations (62). If CD19 downregulation requires prolonged or continuous signaling relative to IgM downregulation, the CD19^low phenotype wouldpreferentially mark cells that have the most difficulty in rapidly editing away autoreactivity.Compared to wild type BM B cells, RS^−/− cells cannot efficiently silence receptors using Jκ5(44), while 3H9 B cells may be less able than wild type cells to extinguish autoreactivity by editing.

On the basis of the relative abundances in wild type versusIgM^b-macroself mice of theB220⁺iIgM⁺sIgM⁻BM population (Fig.4B), we estimate that approximately 20% of wild type BMimmature B cells are autoreactive and under negative selection, a number consistent with,or somewhat lower, than other estimates of the extent of editingor negative selection in the repertoire(18, 38, 61, 63, 64). It is important to note that in the approach taken here we exclude from analysis cells that fail to undergo positive selection because of an inability to pair H- and L-chains, since such cells would not be scored with the 331.12 antibody. Luning Prak and Weigert(65) and Casellas and colleagues(38) have shown with κ-chain targeted transgenic mice that functional VκJκ exons are often displaced by editing, at frequencies ranging from 18–71%, but the fraction of those arising from an inability to pair with H-chain was unclear.It is also possible that we might underestimate the frequency of central tolerance in the normal compartment because of the ability of cells to rapidly correct autoreactivity by editing after relatively little sIgM or CD19 downregulation. Our assumption that developing QM cells are completely non autoreactive also likely leads to an underestimate of as much as 15%, as some QM B cells express a non Tg H-chain and certain λ chains may confer autoreactivity when paired with the QM H-chain (43, 66).

Limited escape of IgH^bB cells

Although developing B cells encountering IgM^b superantigen rarely escape deletion, we often detect small numbers of residual B2-like cells in peripheral tissues of IgM^b-macroself mice. In the spleen, such cells are rare and haveeither a CD93⁺sIg⁻iIgM⁺ phenotype, which may represent the outcome of residual hematopoiesis in situ, or a B1-like phenotype.Classical anergic B cells have not been observed, probably because of the nature of the superantigen and its expression pattern in this particular model. In the peritoneum, B1 cells are occasionally present in significant numbers in older mice. Switching contributes to B cell escape because we detect low levels of IgG and IgA in the sera of most IgH^b/b mice carrying the IgM^b-macroself antigen. We suspect that the B1 cells still present in the peritoneum of IgM^b-macroself mice represent either rare class switch variants or are somatic cell mutants that can expand owing to the ability of B1 cells to self renew. In any case, these escape mechanisms seem inefficient and are unlikely to contribute significantly to normal development in mice. Our results are reminiscent of those obtained many years ago in experiments treatingimmature animals with µ-chain antibodies(67), which significantly blocked not only the production of IgM, but also that of IgG and IgA. In those studies,as in ours,low expression of IgG and IgA was often observed in a small subset of treated animals.

B cell homeostasis of IgH^a/b cells in the presence of IgM^b superantigen

In IgH^a/bmice, superantigen eliminatesIgH^b cells, resulting in a concomitant reduction in total BM output and total peripheral B-2 cell numbers to ∼50%. We interpret these data to indicate that BM output regulates steady state numbers of the B2 (i.e., follicular) subset. By contrast, MZ and B1 compartments were normal in size and made up of IgH^a cells, indicating that these compartments are regulated independently of BM output. These results are consistent with studies in which BMB cell output was suppressed in adult mice by lack of IL-7 or induced elimination of RAG expression(68, 69). However, in IgH^a/b→IgM^b-macroself chimeras,the MZ compartmentis also diminished by roughly half, suggesting that adult BMprecursors give rise to MZ cells that are more limited in the capacity for self-renewalthan those in intact mice. More subtly, these data appear to place limits on the potential contribution of H-chain editing to the opposite allele to facilitate B cell escape from deletion, as they fail to appreciably normalizeBM B cell output.However, if VH replacement is limited only to the initially “autoreactive” IgH^b allele, we would expect no escape from tolerance in theIgM^b-macroselfmodel, and hence our data only place limits on the extent of replacements leading to changes in allele usage.

Bcl2 overexpression in B cells in IgM^b-macroself;Bcl2 Tg micepromotes a significant escape from central tolerance and peripheralization of B cells. At face value, that might be evidence that central tolerance is solely an apoptotic process. However, we have previously found that the same Bcl2 transgenic line failed to promote escape in two other models, but instead facilitated receptor editing(3, 49). To explain these results we suggest that there must be an important component of competition between innocuous versus autoreactive B cells, with autoreactive B cells at a considerable disadvantage. When editing is blocked or ineffectual, as in IgM^b-macroself mice, and competing non-autoreactive B cells are rare, Bcl2 overexpression allows development to proceed. This is also consistent with our finding that Bcl2 Tg;RS^−/− B cells escape tolerance when they develop in κ-macroself mice, whereas Bcl2 Tg B cells do not(44). In that model, λ B cells would provide the competition, whereas in IgH^b/b;IgM^b-macroself mice all cells are equally reactive to superantigen.RS mutation hinders editing, which in turn reduces the abundance of competing λ cells.One interpretation of these results is that BM B cells compete for a limiting resource to enforce tolerance, a possibility with significant clinical implications. Although the nature of such a factor is unknown, we believe we can exclude BAFF as a candidate based on the inability of BAFF overexpression to rescue in this model (data not shown).