Global analysis of B cell selection using an immunoglobulin light chain–mediated model of autoreactivity

We established a mouse model in which fixed Igκ LC transgenes influence the preselection autoreactivity of the B cell repertoire when paired with a polyclonal repertoire of endogenously rearranged HCs. To do so, we used mice containing a predicted “innocuous” αHelκ LC and “self-reactive prone” Vκ4 LC knocked into the variable region of one κ allele. To monitor receptor editing, these LC transgenic mice were crossed with mice containing the human κ constant (hCκ) region knocked in downstream of unrearranged variable region genes (Igκ^h) of the endogenous κ locus (Fig. 1 A). In these Igκ^αhel/h or Igκ^Vκ4/h mice, upon receptor editing, the hCκ allele undergoes VJ rearrangement and pairs with the endogenous HCs (Casellas et al., 2001, 2007). Thus, we can track the frequency of receptor-edited B cells and their subsequent selection by flow cytometry for mouse versus human Igκ and/or by RT-PCR and sequence analysis of the VκJκ region of the expressed LC.

We first verified that the fixed transgenic LC in these mice is not significantly skewing or restricting the HC repertoire. Comparison of the HC VH gene usage in the splenic mature compartment showed similar and diverse usage of VH genes in Igκ^αhel/h or Igκ^Vκ4/h mice in both edited and unedited B cells (Fig. 1 B and not depicted). Likewise, spectratyping of VH genes J558.85 and J606.1 in sorted B cell subsets showed a varied, polyclonal HC repertoire at all stages of development, comparable with a wild-type B6 mouse for both HCs (Fig. 1, C–E). This was true in cells that expressed the transgenic LC or endogenous LC. J558.85 was specifically chosen because its germline sequence is similar to the 3H9 HC and has been found in other autoantibodies in its unmutated form (Shlomchik et al., 1990). Conversely, J606.1 is normally abundantly rearranged (Meng et al., 2011) and has not been described among autoantibodies to our knowledge. Collectively, these results suggest that alterations in LC rearrangement between Igκ^αhel/h or Igκ^Vκ4/h mice are not caused by severe limitations in HC/LC pairing or strongly altered HC selection at the pro-B to pre-B transition.

Next, we confirmed that antibodies encoded by randomly rearranged VH genes paired with the Vκ4 LC were indeed more self-reactive than those encoded by the αHelκ LC. For this, we directly compared the self-reactivity of the BCR composed of the Vκ4 versus the αHELκ LC transgene paired with 43 VH genes that were isolated from single-cell sorted mature (AA4⁻) B cells from the spleens of Igκ^αhel/h and Igκ^Vκ4/h mice. VH genes from both types of mice were sequenced and cloned into expression vectors and cotransfected with either transgenic LC into 293A cells to express recombinant mAbs. The mAbs were expressed as chimeric antibodies using identical human IgG1/Cκ constant regions to normalize detection in all assays. There was no clonal relation between any two individual B cells (not depicted). The 43 pairs of VH/Vκ4 and VH/αHELκ encoded mAbs were tested side by side for their ability to bind HEp-2 cells either by ELISA or immunofluorescence (IMF). When VH genes isolated from either mouse strain were paired with the Vκ4 LC, in all but two cases, the resulting antibody was more HEp-2 reactive by ELISA or IMF than when paired with the αHelκ LC (Fig. 2, A–C). As we are testing antibodies with mouse-derived variable regions on a human cell line, we further confirmed that we are measuring reactivity to self and not foreign antigen by testing the ability of the antibodies to bind 3T3 mouse fibroblast cells. As with HEp-2 cells, binding of VH/Vκ4 antibodies to 3T3 cells was higher than VH/αHELκ antibodies by IMF (Fig. 2 D). None of the antibodies bound better to HEp-2 cells compared with 3T3 cells, although several of the VH/Vκ4 antibodies had more intensive staining on 3T3 cells than HEp-2 cells. Whether this is caused by differences in the species from which the cell is derived or the cell fixation process is unclear. We went on further to measure by ELISA the reactivity of antibody pairs to the common self-antigen, dsDNA. Antibody binding to dsDNA mirrored reactivity to LPS and insulin (not depicted), and we therefore used this assay as a measure of polyreactivity. Similar to self-reactivity as determined by HEp-2 binding, the affinity of polyreactive antibodies was also reduced when the HC was paired with the αHelκ LC (Fig. 2 E). Thus, the antibodies encoded by the Vκ4 LC are indeed self-reactive prone and more polyreactive when paired with HCs isolated from either LC transgenic mouse. In contrast, the αHELκ LC tends to control or reduce HC-mediated self-reactivity and polyreactivity.

In the fixed LC knockin mice used herein, the endogenous Vκ/Jκ gene segments on the hCκ allele are rearranged almost solely for receptor editing, either with deletion of the original LC (editing by deletion: mCκ⁻hCκ⁺ cells) or by coexpression with the transgene (editing by inclusion: mCκ⁺hCκ⁺ cells). This dilutes the self-reactivity of the cell to levels where the cell can avoid deletion (Gay et al., 1993; Li et al., 2002; Wardemann et al., 2004; Liu et al., 2005; Huang et al., 2006; Casellas et al., 2007; Velez et al., 2007). In either case, the majority of HCs from hCκ⁺ B cells are predicted to be less self-reactive when paired to hCκ rather than with the Vκ4 or αHELκ LC transgenes. This provides an easily measurable biomarker identifying B cells with random specificities but reduced self-reactivity. To formally demonstrate this, we isolated B cells from the LC transgenic mice that express hCκ, indicating they had undergone receptor editing. Some HCs may not pair well with the knockin transgenes, so in mCκ⁻hCκ⁺ B cells, the use of the hCκ gene may sometimes occur because of structural reasons rather than selection to reduce self-reactivity. Therefore, we first looked at B cells that had undergone receptor editing by allelic inclusion (mCκ⁺hCκ⁺ cells) because in these cells we could verify that both the mCκ transgene (the edited LC) and the hCκ endogenous LC (the editor) could efficiently pair with the HC expressed by that particular B cell. Using single-cell RT-PCR, we isolated and sequenced the HC and the hCκ⁺ VκJκ rearrangements. The HC genes were then cotransfected with either the hCκ⁺ (editor) gene or the original Vκ4 or αHEL transgenes, and the resulting antibodies were tested for self-reactivity by HEp-2 ELISA and IMF and dsDNA ELISAs. As indicated in Fig. 3 A, most (24/32) of the mAbs cloned from allelically included (mCκ⁺hCκ⁺) B cells were significantly more self-reactive when the HC was paired with the Vκ4 transgene rather than with the hCκ. This observation was consistent when we grouped the antibodies based on the individual mouse they were cloned from (Fig. 3 B). Similarly, by all three assays, all but one of the antibodies from allelically excluded mCκ⁻hCκ⁺ B cells from an Igκ^Vκ4/h mouse bound better to self-antigen when expressed with Vκ4 compared with hCκ (Fig. 3 C). Thus, although structural incompatibility between the HC and LC may play some role in receptor editing, LC editing primarily occurs to reduce autoreactivity in Igκ^Vκ4/h mice. It is more difficult to detect a difference in self-reactivity between hCκ- and αHelκ-expressing antibodies as very few mature B cells in Igκ^αhel/h mice, regardless of the LC, bind self-antigen at detectable levels in the assays used here. However, the HCs from both mCκ⁺hCκ⁺ and mCκ⁻hCκ⁺ B cells cloned from an Igκ^αhel/h mouse were more often dsDNA reactive when expressed with the αHELκ transgene (Fig. 3, D and E). Furthermore, though not significantly increased on average, there were several clear examples where HEp-2 ELISA and or IMF positivity was increased when the HC was paired with the αHELκ transgene compared with the hCκ editor (Fig. 3, D and E). We can reliably conclude, then, that in this transgenic LC system, receptor editing and expression of the hCκ allele serve as a tolerance mechanism to reduce the autoreactivity of the mature B cell repertoire. Thus, by quantifying the loss of transgenic mCκ LC using B cells (and the concomitant accumulation of edited hCκ⁺ and λ⁺ B cells), we have a biomarker to analyze central and peripheral immune tolerance events in mice containing a diverse, polyclonal HC repertoire.

Given their relative self-reactive properties, we would expect substantially more counter-selection of B cells expressing the Vκ4 over the αHELκ transgene beginning at the earliest stage of B cell development when the HC/LC pair is first expressed on the cell surface. Indeed, when we measured by flow cytometry the percentage of immature B cells (B220⁺IgM⁺CD23⁻; Fig. S1 A) in the BM that were mCκ⁺, hCκ⁺, or λ⁺, a substantially higher percentage of immature cells in the Igκ^Vκ4/h mice had undergone receptor editing and expressed an hCκ or λ LC (Fig. 4, A and B), similar to data published previously (Casellas et al., 2001, 2002). On average, 85% of immature B cells from Igκ^αhel/h mice expressed the mCκ LC, whereas only ∼65% of the immature B cells did so in the Igκ^Vκ4/h mice (Fig. 4 B).

The different frequencies of edited cells in the two strains could be caused by either increased levels of receptor editing or greater deletion of unedited mCκ⁺ cells in the Igκ^Vκ4/h mice during development in the BM. In an attempt to differentiate these two possibilities, we compared the relative levels of receptor editing in Igκ^αhel/h mice versus Igκ^Vκ4/h mice by measuring levels of recombining sequence (RS) rearrangements formed during the primary stage of receptor editing in BM B220⁺CD43⁻IgM⁻ pre-B cells (Panigrahi et al., 2008), leading to the inactivation of the original κ LC allele. There was no significant difference in the relative frequency of RS rearrangements between Igκ^Vκ4/h and Igκ^αhel/h mice at the pre-B cell stage (Fig. 4 C), suggesting overall levels of receptor editing might be the same. However, editing can occur without terminal RS rearrangements (Nemazee, 2006), allowing the possibility of allelic inclusion. When we looked in more detail, we found that as expected, there were lower levels of RS rearrangements in sorted, largely unedited, mCκ⁺ immature B cells of either strain in comparison with total Igκ⁺ splenic mature B cells from wild-type B6 mice. Also, predictably, in the Igκ^αhel/h mice there were more RS rearrangements in the immature cells that had undergone receptor editing (the hCκ⁺ fraction) compared with the unedited mCκ⁺ fraction. Interestingly however, RS rearrangement levels were virtually unaltered between the edited and unedited immature B cell fractions sorted from Igκ^Vκ4/h mice (Fig. 4 C). This reduction in Vκ RS terminal rearrangements correlates with a twofold increase in the percentage of allelically included mCκ⁺hCκ⁺ surface double-positive immature B cells in Igκ^Vκ4/h mice (Fig. 4 D). More strikingly, when we single-cell sorted splenic transitional edited mCκ⁻hCκ⁺ cells and determined by RT-PCR the κ chain or chains expressed, we found that an estimated 57% of cells isolated from Igκ^Vκ4/h mice still expressed at the transcript level both the original fixed LC and the newly rearranged hCκ⁺ LC, whereas only ∼20% of the transitional hCκ⁺ cells from Igκ^αhel/h mice expressed two LCs (Fig. 4 E). This difference in the editing mechanism between the two transgenic mice makes it difficult to determine exactly the relative level of receptor editing. However, significantly higher levels of self-antigen–mediated deletion in the BM were not apparent in Igκ^Vκ4/h mice as total numbers of B cells at all developmental stages were not altered in the BM or spleen (Fig. 4 F). We conclude from these experiments that increased receptor editing and limited levels of deletion are induced to mitigate LC-mediated autoreactivity at the immature stage of development in the BM.

When we compared the level of edited cells between immature BM and mature splenic follicular mature (FM; AA4⁻CD21^intCD23⁺) B cells in Igκ^Vκ4/h mice, we found that the percentage of edited hCκ⁺ cells had almost doubled from ∼30% in the immature B cell subset to >60% in the splenic FM compartment (Fig. 5, A and B). The percentage of hCκ⁺ cells also increased in the periphery of Igκ^αhel/h from a mean of 13% in the BM immature subset to around 20% at the mature stage in the spleen (Fig. 5, A and B). There was little difference in LC usage between mature B cells in the spleen, LNs, and BM in both LC transgenic mice (not depicted). To further delineate at what point of peripheral B cell development the loss of mCκ⁺ cells is occurring, we measured by flow cytometry LC usage in AA4⁺IgM^hiCD24^hiCD21^loCD23⁻ early transitional (T1) and AA4⁺IgM^hiCD24^hiCD21^intCD23⁺ late transitional (T2) splenic B cell subsets (Fig. S1 B). No difference in LC usage was detected between immature B cells in the BM and T1 B cells in the spleen in either mouse strain, and only a small decrease in mCκ usage was observed between the T1 and T2 compartment in the Igκ^Vκ4/h mice (Fig. 5 B). Thus, the loss of transgenic LC-expressing cells is primarily occurring between the late transitional (T2) and the FM B cell developmental subsets.

Some receptor editing can occur on the mCκ allele of both the Vκ4 and αHelκ transgene as both still contain downstream endogenous Jκ gene segments. Therefore, to determine more accurately the percentage of cells still expressing the knockin LC in the peripheral B cell compartment, we single-cell sorted mCκ⁺hCκ⁻ cells, amplified by RT-PCR the expressed κ chain from each cell, and sequenced it. Based on the analysis of >80 sequences from each B cell population sorted from six mice, we determined that >90% of AA4⁺CD24^hiIgM^hi transitional (inclusive of both T1 and T2; Tr) B cells from both mouse strains still expressed the knockin LC. However, although 87% of the mCκ⁺hCκ⁻ FM B cells from Igκ^αhel/h mice expressed the αHelκ LC, only 62% of FM B cells in Igκ^Vκ4/h mice remained Vκ4⁺. If we combine this with the flow cytometry data, the proportion of cells between the Tr and FM compartment in Igκ^Vκ4/h mice that had a receptor-edited LC almost doubled (43 to 82%; Fig. 5 C). Even in Igκ^αhel/h mice expressing an innocuous LC, the level of edited cells increased by about a third (21% in Tr to 29% in FM) during peripheral B cell development, presumably because of largely HC-mediated autoreactivity that the αHelκ LC is unable to mask.

To measure the impact of peripheral tolerance on removing self-reactive and dsDNA-binding polyreactive B cells from the naive mature B repertoire, we single-cell sorted mCκ⁺hCκ⁻ Tr and FM B cells from Igκ^Vκ4/h mice (Fig. S2 A), verified by sequence analysis that they expressed the Vκ4 LC, and produced mAbs with the endogenously rearranged HCs and Vκ4 LC from 30 Tr and 25 FM B cells. When we compared self-antigen binding of antibodies expressed by Tr versus FM B cells, we saw a significant reduction by HEp-2 cell ELISA in the overall level of autoreactivity between Tr and the FM B cell compartment (Fig. 5, D and E). We detected little dsDNA-binding polyreactivity, however, in either peripheral B cell compartment (Fig. 5, D and E).

These experiments show that a substantial proportion of primary tolerance occurs in peripheral locations rather than just in the BM. This is particularly striking in the Igκ^Vκ4/h mice but is also evident in the Igκ^αhel/h mice. Finally, we show in mice with a polyclonal B cell repertoire that the transition from T2 to FM is an important stage in primary selection during peripheral B cell development.

Deletion of self-reactive B cells is thought to be the default negative selection mechanism operating in the periphery, not receptor editing as in the BM. Indeed, similar to past studies of RAG reporter mice (Yu et al., 1999; Gärtner et al., 2000), we were unable to detect RAG transcript expression by single-cell RT-PCR in transitional cells in the periphery of Igκ^Vκ4/h or Igκ^αhel/h mice (not depicted), nor was there a significant number of Ig-low cells, which can be indicative of ongoing receptor editing (Fig. S1 B). These results indicate that continuing receptor editing during the transitional phase of development does not account for the substantial change in proportion of edited cells within the mature compartment.

We also looked for evidence of Vκ4⁺ cells being rendered anergic by encounter with self-antigen, thus preventing development into FM B cells. The absolute number of T3 (An1) AA4⁺CD23⁺IgM^lo cells previously described as anergic B cells (Merrell et al., 2006) was not different between Igκ^αhel/h and Igκ^Vκ4/h mice, and surface IgM levels of other subsets were normal (Fig. 4 F and Fig. S1 B). In addition, Vκ4⁺ cells were not enriched in the T3 B cell compartment, which in fact contained fewer Vκ4⁺ cells than the IgM^hi Tr compartment (Fig. 6 A). Interestingly, however, >60% of the T3 population in Igκ^αhel/h mice no longer expressed the αHelκ LC compared with ∼30% in the FM compartment (Fig. 6 A). We hypothesized that these AA4⁺CD23⁺IgM^lo receptor-edited cells likely express an HC with intrinsic self-reactive properties, triggering receptor editing and eventually anergy when LC rearrangement proved unsuccessful at completely correcting self-reactivity. To test whether these hCκ⁺ T3 cells were cells in which receptor editing failed to reduce self-reactivity, we single-cell sorted T3 cells from Igκ^αhel/h mice (Fig. S2 B) and cloned the HC and hCκ LC and expressed antibodies containing the HC paired with either the unedited αHELκ or the rearranged hCκ LC. In contrast to FM B cells (Fig. 3 E), in 9/11 T3 antibody pairs tested, expression of a rearranged LC did not lead to reduction in self-reactivity compared with antibodies with the unedited αHELκ LC (Fig. 6 B). Indeed, when we directly compared the self-reactivity of antibody pairs from edited FM and T3 B cells, although there was no difference between the two populations in the ability of αHELκ-containing antibodies to bind HEp-2 cells, T3 hCκ⁺ antibodies were more self-reactive than their FM counterparts (Fig. 6 C). Though not statistically significant, a similar trend was observed in the ability to bind dsDNA (Fig. 6 C). Together, these data indicate that unedited self-reactive cells primarily undergo deletion in the periphery, whereas cells that have undergone editing in BM but retain some level of HC-mediated autoreactivity are maintained in an anergic state in the periphery.

To better understand the level of BCR self-reactivity that triggers different forms of tolerance, we compared the levels of autoreactivity of cells that had or had not undergone receptor editing in Ig^Vκ4/h mice. To estimate the preselection reactivity of cells induced to undergo receptor editing, we took HC genes from cells that no longer expressed the Vκ4 LC (Vκ4⁻hCκ⁺) or were allelically included (Vκ4⁺hCκ⁺) and paired them with the original Vκ4 LC. The autoreactivity of these preselection HC/LC pairs was then compared with antibodies containing the HC paired with an edited, rearranged hCκ LC (postselection hCκ⁺) and to antibodies cloned from cells that had reached maturity without undergoing receptor editing (Vκ4⁺hCκ⁻; Fig. 7 A). As expected, the two groups of antibodies representing the preselection condition had the highest levels of autoreactivity as measured by HEp-2 ELISA and IMF (Fig. 7 A). The affinity and number of dsDNA-binding polyreactive antibodies was also increased in preselected antibodies. Notably, as suggested by our previous publication (Casellas et al., 2007), many allelically included B cells escaped primary immune tolerance controls despite continued expression of HC/LC self-reactive pairs. It is also interesting that both Tr B cells, the earliest population to have exited central selection but have not completed peripheral selection, and edited and unedited post-selection antibodies isolated from mature B cells had very low levels of dsDNA-binding compared with preselection antibodies (Figs. 5 E and 7 A). This suggests that B cells expressing polyreactive antibodies capable of binding this autoantigen were strongly counter-selected during central B cell development before these cells reached the periphery. HEp-2 reactivity, however, declined between unedited cells at the Tr developmental stage and full maturation (Figs. 5 E and 7 A). The level of autoreactivity was even lower in mature cells that had undergone receptor editing. Overall, these experiments indicate first that receptor editing results in a significant reduction in the mean degree of autoreactivity to a level that allows them to reach the mature naive B cell stage. Second, whereas polyreactive cells capable of binding dsDNA were removed or edited primarily during the central stage of tolerance, those binding other autoantigens as detected by HEp-2 ELISA and IMF were predominantly removed during the peripheral stage of tolerance.

Despite the extensive peripheral deletion occurring in the Ig^Vκ4/h mice, the remaining Vκ4⁺ cells in the naive mature compartment were, as a whole, more HEp-2 reactive compared with edited hCκ⁺ cells in the same mice or mature αHelκ LC–expressing cells (Fig. 7, A–E). We therefore looked for evidence of further selection against Vκ4⁺ cells during B cell activation, or secondary tolerance. We saw similar LC usage when we compared splenic FM naive B cells and IgG⁺ B cells in Igκ^Vκ4/h mice but detected a small, but reproducible decrease in mCκ⁺ cells in isotype-switched IgG⁺ compartment in the Igκ^αhel/h mice (Fig. 8 A). This may be because the naive mature cells in the small, but more diverse hCκ⁺ mature pool are more likely to encounter foreign antigen and enter an immune response. The proportion of serum IgG antibodies expressing mCκ versus hCκ (edited) LC and IgG antibody–secreting cells also correlated with the number of naive mature B cells expressing these LCs (Fig. 8, B and C). Thus, despite the higher level of autoreactivity still present in the unedited fraction of naive FM B cells in Igκ^Vκ4/h compared with Igκ^αhel/h mice (Fig. 7, B–E), we detected no further selection against cells expressing the self-reactive prone Vκ4 LC during B cell activation and differentiation into antibody-secreting cells and isotype-switched cells on a B6 background.

To study the importance of tolerance mechanisms during B cell development, we crossed the Igκ^Vκ4/h and Igκ^αhel/h mice onto the autoimmune MRL/lpr background where defects in central and peripheral B and T cell tolerance lead to systemic lupus–like disease. On the MRL/lpr background, there was no significant difference in the relative level of RS elements in pre-B or immature B cells in the BM (Fig. 9 A) or in the levels of allelically included cells at the transcript or protein level (not depicted). This suggests that the levels of receptor editing in Ig^Vκ4/h mice were not altered between the two genetic backgrounds. However, as published previously (Panigrahi et al., 2008), we did detect lower levels of RS elements in MRL/lpr mice without a fixed LC compared with B6 mice (not depicted). Although there was only a modest increase in the levels of mCκ⁺ (self-reactive prone) cells at the BM immature stage in the Igκ^Vκ4/h MRL/lpr mice, there was a substantial increase in these cells at the FM stage of development compared with the B6 background (Fig. 9 B). Indeed, when we determined the percentage of splenic B cells by flow cytometry and single-cell sequence analysis of the κ locus that still expressed the transgenic LC, we saw little to no difference in LC usage at the Tr stage in Igκ^Vκ4/h mice between B6 and MRL/lpr mice (Fig. 9 C). At the FM stage, however, whereas 80% of the cells in the B6 mice were receptor edited, in the MRL/lpr mice, only 65% of the cells had an edited LC (Fig. 9 C). Likewise, fewer receptor-edited FM B cells were detected in MRL/lpr Igκ^αhel/h mice. Thus, in the MRL/lpr mouse model of lupus, a significant proportion of the defect in primary immune tolerance occurs during the peripheral stages of B cell selection.

We then wanted to determine whether this difference in peripheral tolerance impacted the progression of autoantibody production in MRL/lpr mice. Total serum IgG levels were very high in both Igκ^αhel/h and Igκ^Vκ4/h mice on the MRL/lpr background (Fig. 9 D). However, in agreement with the decreased levels of hCκ⁺ FM B cells, the ratio of hCκ to mCκ⁺ serum IgG levels decreased in the MRL/lpr Igκ^Vκ4/h mice (Fig. 9 E). Thus, altered peripheral selection of Vκ4 LC–expressing cells on the MRL/lpr background resulted in more Vκ4⁺ serum IgG antibodies. Most importantly, this alteration in peripheral selection corresponded with accelerated development of IgG dsDNA antibodies, which were detectable as early as 2 mo of age in the Igκ^Vκ4/h MRL/lpr mice, in contrast to normal MRL/lpr or Igκ^αhel/h MRL/lpr mice (Fig. 9 F, left). This increased production of dsDNA antibodies continued as the mice aged, although total serum IgG levels were comparable between the two transgenic mice at all ages tested (Fig. 9 F, right). IgG dsDNA-specific serum antibodies were almost undetectable in both LC knockin mice on the B6 background up to 11 mo of age (not depicted). Thus, expression of an autoimmune-permissive LC can accelerate production of self-reactive antibodies when peripheral tolerance mechanisms are impaired.

It has long been appreciated that a portion of primary tolerance must occur as newly formed B cells transition to maturity in the periphery. However, it had been difficult to establish the relative contributions of central versus peripheral tolerance mechanisms in honing the mature repertoire that, as a pool, recognizes a multitude of self and nonself antigens. We therefore established a mouse model to track tolerance mechanisms operating in a diverse polyclonal B cell repertoire in the context of two significantly different starting levels of LC-mediated autoreactivity. The HC CDR3 forms the epicenter of the Ig antigen–binding site and thus more easily trumps LC effects on specificity (Kindt and Capra, 1984) and has been demonstrated to have profound associations with B cell selection and tolerance (Zemlin et al., 2005, 2008; Ippolito et al., 2006; Schelonka et al., 2008). However, as illustrated by the high-affinity anti-DNA 3H9 antibody isolated from MRL/lpr mice, the LC can have a profound influence on binding affinity and to a lesser extent on antigen specificity (Radic et al., 1991; Li et al., 2001). When paired with the 3H9 HC, variant LCs with acidic CDRs are able to reduce anti-DNA binding, whereas LCs with more neutral or basic CDRs such as the Vκ4 LC used in the study fail to do so. Thus, the Vκ4 LC in 3H9/Vκ4 and 3H9R/Vκ4 double transgenic BALB/c mice is almost invariably replaced by endogenously rearranged LCs (Gay et al., 1993; Chen et al., 1997). This, together with studies demonstrating extensive receptor editing in Vκ4 LC knockin mice, suggests that this LC has strong inherent self-binding properties (Prak and Weigert, 1995; Casellas et al., 2001). Here, we formally demonstrate that the Vκ4 LC is self-reactive when paired with a wide selection of HCs isolated from mature B cells. This led to profound selection against Vκ4⁺ cells both centrally and in the periphery, making this a useful mouse model to decipher the key selection checkpoints. We coupled this transgenic mouse system with cloning and expression of BCRs to track qualitatively and quantitatively the self-reactivity of B cells that have undergone tolerance at different stages of development. Together, these methods have allowed us to gain unprecedented insight into stage-specific B cell tolerance in mice containing a diverse BCR repertoire.

We demonstrate that receptor editing of the original fixed LC by the endogenous VJκ genes as detected by hCκ⁺ usage reduced autoreactivity. Thus, changes in LC usage by receptor editing that we detected appear to be largely caused by negative selection aimed at reducing autoreactivity, as has been shown in many other mouse tolerance models (Retter and Nemazee, 1998; Wardemann et al., 2004; Ait-Azzouzene et al., 2005). Although the possibility that changes in B cell LC usage over the course of development are caused by positive selection cannot be formally ruled out, the demonstrated autoreactivity of Vκ4-encoded antibodies strongly suggests that changes in the repertoire were mediated by negative selection. Therefore, it was not surprising that in Igκ^Vκ4/h mice, a much larger percentage of immature B cells compared with Igκ^αhel/h mice no longer expressed the LC knockin as the result of central tolerance. Strikingly, however, although the percentage of edited cells changed little between splenic early (T1) and late transitional (T2) cells, there was an almost 50% reduction in Vκ4⁺-expressing cells between the T2 developmental stage and FM B cells. Although it is well established that deletion of self-reactive cells occurs in the periphery (Russell et al., 1991; Fulcher et al., 1996; Wardemann et al., 2003; Aït-Azzouzene et al., 2006; Merrell et al., 2006), the specific transitional stage at which this occurs in a polyclonal repertoire has not been well characterized. Factors regulating B cell survival clearly change between the AA4⁺CD23⁻CD21^lo T1 and AA4⁺CD23⁺CD21^int T2 transitional subsets, although in vitro they both undergo apoptosis upon BCR stimulation (Allman et al., 2001; Duong et al., 2010). In T2 cells, tonic BCR signaling becomes vital for survival, and up-regulation of BAFFR levels coincides with responsiveness to the survival signal provided by BAFF (Hsu et al., 2002; Chung et al., 2003). Mice lacking BCR signaling components and BAFF expression have a block in development between the T1 and T2 stage, and conversely, mice overexpressing BAFF or lacking the proapoptotic gene Bcl2l11 have selective survival of cells at the T2 stage and beyond (Batten et al., 2000; Schiemann et al., 2001; Craxton et al., 2005). Given these functional differences, we expected that negative selection might occur to a significant degree between the T1 and T2 stage. However, we show that the majority of self-reactive cells that reach the periphery are lost between the T2 and FM compartments. This would suggest that positive selection through appropriate BCR/BAFF signals can occur at the T2 stage of development, but attrition is primarily occurring later for self-reactive B cells. Although maintenance of self-reactive cells this late into maturity is potentially dangerous, it allows for selection of certain polyreactive low-affinity B cells that may be important in innate-like responses (i.e., marginal zone or B1 cells) as previously demonstrated in M167 and 81x transgenic mice (Martin and Kearney, 2001; Meyer-Bahlburg et al., 2008).

In contrast to our results, in mice expressing an Igκ superantigen only in the periphery, negative selection appears to primarily occur earlier, between the T1 and T2 stage (Aït-Azzouzene et al., 2006; Ota et al., 2011). However, in this model, B cells are first encountering a high-affinity autoantigen in the periphery, resulting in deletion at first exposure to self-antigen at the early T1 transitional stage. By using an LC-mediated model of autoreactivity, we are able to study tolerance against a broad range of specificities (evident from the various affinities to HEp-2 cell extracts or dsDNA and the variant staining patterns of HEp-2 cells that were detected from the Vκ4-encoded antibodies) encountered in their physiological settings. Under these conditions, it appears that high-affinity self-reactive cells are removed earlier in the BM. Although the peripheral selection stage is still significant in magnitude, it appears to primarily remove moderate-affinity autoantibodies after the T2 stage.

Importantly, we detected very limited Vκ4-encoded polyreactive dsDNA binding in peripheral B cells, although anti-dsDNA reactivity was commonly encoded by VH genes from edited (hCκ⁺) cells when expressed with the Vκ4 transgene to resemble the cell’s preselection specificity. It is clear that different factors including the affinity to self-reactivity influences the mechanism of tolerance (Shlomchik, 2008), but simple differences in binding affinity cannot fully explain what dictates central versus peripheral tolerance. Indeed, very low-affinity anti-MHC BCRs have been shown to be efficiently tolerized in the BM (Lang et al., 1996), suggesting that ignorance, such as seen in the AM14 rheumatoid factor transgenic mouse model (Hannum et al., 1996), not just affinity, can explain why some self-reactive B cells reach and persist in the periphery. In any case, we can clearly establish a stepwise progression in tolerance as B cells develop.

Anergic B cells and B cells that have undergone receptor editing by allelic inclusion remain self-reactive in the periphery (Goodnow et al., 1988; Hartley et al., 1991; Kenny et al., 2000; Li et al., 2002; Liu et al., 2005; Huang et al., 2006; Casellas et al., 2007; Velez et al., 2007; Duty et al., 2009; Isnardi et al., 2010), but the prevalence of these two partial forms of tolerance varied between the two LC transgenic models. Given the high level of peripheral attrition of Vκ4⁺ B cells, it is somewhat surprising that more Vκ4⁺ cells weren’t found in the T3 IgM^lo B cell population. Work with Hel, ArsA1, superantigen-induced tolerance, and other model systems indicates that a major mechanism of tolerance is the presence of anergic cells in the periphery with reduced survival (Goodnow et al., 1988; Erikson et al., 1991; Santulli-Marotto et al., 1998; Benschop et al., 2001; Ota et al., 2011). But it is unclear from the literature whether peripheral deletion is necessarily preceded by low surface Ig levels and induction of an anergic state and that all deleted cells would transit through the T3 stage before undergoing apoptosis. It may be that induction of anergy (whether caused by antigen encounter in the BM or periphery) and peripheral induction of apoptosis through strong BCR signals could be two separate tolerance mechanisms. In contrast to other model systems in which peripheral attrition of cells anergized by antigen encounter is observed, we are characterizing tolerance induced in response to a myriad of different self-antigens present in different anatomical locations and with different affinities and avidities. It is probable that a given antigen that would induce frank apoptosis in the BM may only be encountered in the periphery whereupon deletion occurs.

Although there were no significant increases in the numbers of anergic Vκ4⁺ cells in Ig^Vκ4/h mice, edited hCκ and Igλ⁺ cells were enriched in the T3 IgM^lo B cell subset in Igκ^αhel/h mice, and these edited cells retained a high level of self-reactivity compared with edited FM B cells. As autoreactivity is largely mediated by the HC in these mice, it is likely that these T3 B cells represent cells with self-reactive prone HCs in which receptor editing failed to mitigate the self-reactive nature of the BCR, leading to anergy. If the LC is the major contributor of autoreactivity, however, as is more often the case in Ig^Vκ4/h mice, receptor editing of the offending LC is more likely to result in a self-tolerant BCR. Cells that have both a highly self-reactive HC and LC are most likely deleted early during central tolerance. Collectively, it seems that receptor editing and deletion primarily deal with LC-mediated autoreactivity, whereas anergy plays a more significant role in HC-mediated autoreactivity.

Allelic inclusion or coexpression of two different LCs in the same cell has also been described in several different systems as a byproduct of receptor editing that allows self-reactive BCRs to escape detection and reach maturity (Gay et al., 1993; Kenny et al., 2000; Li et al., 2002; Liu et al., 2005; Huang et al., 2006; Casellas et al., 2007; Velez et al., 2007). In support of this model, when we compared the self-reactivity of BCRs with either the original Vκ4 LC or with the hCκ editor cloned from mature cells coexpressing similar levels of both LCs, hCκ⁺ BCRs were more self-tolerant. The proportion of allelically included cells detected by flow cytometry was doubled in Ig^Vκ4/h mice compared with Igκ^αhel/h mice at the same stage in the BM. Furthermore, as detected by single-cell PCR, about half of the Ig^Vκ4/h cells isolated by flow cytometry as hCκ⁺ but mCκ negative at the protein level expressed the Vκ4 transgene at the transcript level, whereas much fewer did so in the Ig^αhel/h mice. This suggests strong selection for cells in which the rearranged VJκ gene on the hCκ allele outcompeted the preexisting Vκ4 LC for pairing with the HC, which in effect resulted in receptor editing and self-tolerance. The reason for this increase in allelically included cells in Ig^Vκ4/h mice is not entirely clear. It is possible that this is an artifact of the transgenic LC as it’s been shown that the Vκ4 LC, but not the αHel LC, has impaired allelic exclusion in an OcaB^−/− knockout (Casellas et al., 2002). This phenotype is associated with a specific need for this transcriptional regulator for Vκ4 but not αHel expression. However, whether allelic exclusion of Vκ4 transgenic cells undergoing editing is reduced on a wild-type background, as in the case here, was not investigated. Alternatively, exacerbated allelic inclusion in the Igκ^Vκ4/h mice could be a physiological mechanism. It may be that the LC-mediated autoreactivity of Vκ4-expressing cells is more easily diluted by the presence of any competing LC. Thus, during receptor editing, any additional LC that functionally rearranges and can more favorably interact with the HC can remove the LC-mediated autoreactivity. Conversely, HC-induced autoreactivity that predominates in the Igκ^αhel/h mice must be directly controlled at the chemical level dependent on the properties of the editing Vκ gene. For example, the presence of LC aspartate residues adjacent to HC arginines in the CDRs is essential to reduce anti-DNA reactivity in the 3H9 system (Li et al., 2001). Our observation demonstrates that Vκ genes that are poor editors and the autoantibodies they encode are particularly prone to be maintained in mature B cells by allelic inclusion, which may have contributed to the exacerbated autoimmune disease when the Vκ4 transgene was expressed on the MRL/lpr background herein.

The last stage of immune tolerance is to control the activation of self-reactive B cells during immune responses, some of which likely arise as the result of somatic mutations that alter specificity. Despite increased receptor editing and peripheral attrition, the fully mature Vκ4⁺ B cells from Ig^Vκ4/h mice still bind better on average to HEp-2 antigens than their counterpart in Igκ^αhelκ/h mice. We hypothesized that the level of self-reactive cells might be further reduced during B cell activation and secondary tolerance. This did not appear to be case, however, as the levels of edited cells did not change between the naive and IgG compartments in Ig^Vκ4/h mice. On the MRL/lpr background, however, we saw accelerated development of IgG dsDNA antibodies in Ig^Vκ4/h mice likely caused by defects in both primary and secondary tolerance. In the classical 3-83 BCR transgenic tolerance model, tolerance against central or peripheral expression of the H-2^k self-antigen is intact on the MRL/lpr background (Rubio et al., 1996; Kench et al., 1998). Likewise, tolerance for soluble and membrane-bound Hel is unchanged in lpr/lpr mice (Rathmell and Goodnow, 1994). However, in anti-DNA models of autoimmunity, there are breaks in tolerance at multiple levels, leading to accelerated production of autoantibodies in MRL/lpr mice. For example, in 3H9 HC or 3H9/Vκ8 HC/LC transgenic mice on the MRL/lpr background, central tolerance appears to be largely intact, but peripheral tolerance is compromised. In these mice, DNA-reactive B cells normally limited to the peripheral immature or anergic compartment with a high turnover rate reach maturity and with T cell help become activated and differentiate into anti-DNA–secreting plasmablasts (Brard et al., 1999; Mandik-Nayak et al., 2000; Chen et al., 2006). Allelically included cells found in the 3H9/Vκ8 transgenic mice with greater frequency in the FM compartment further contributed to the loss of tolerance and DNA-reactive B cells in MRL/lpr mice (Brard et al., 1999). Similarly, though we saw little change in central tolerance in these LC transgenic mice, there was a significant increase in unedited Vκ4⁺ and αHelκ LC–expressing cells in the periphery. Thus, it appears that defects in peripheral B cell tolerance increasing the level of naive precursor self-reactive cells, in combination with the availability of self-reactive T cell help, are instrumental in leading to accelerated autoimmunity in MRL/lpr mice.

This LC-mediated model of self-reactivity highlights the importance and complexity of peripheral tolerance. Furthermore, we demonstrate that LC-mediated autoreactivity alone can lead to accelerated development of autoantibodies. LCs matter.

Igκ^αhel/h and Igκ^Vκ4/h mice were provided by R. Casellas at the National Institute of Allergy and Infectious Diseases (Bethesda, MD), and the MRL/MPJ and MRL/lpr mice were purchased from the Jackson Laboratory. Igκ^αhel/h and Igκ^Vκ4/h mice were bred as previously described (Casellas et al., 2001). The Igκ^αhel/αhel, Igκ^Vκ4/Vκ4, and Igκ^hCκ/hCκ were backcrossed to MRL/MPJ mice for six generations and then crossed with MRL/lpr mice for two generations. To get Igκ^αhel/h or Igκ^Vκ4/h on the MRL/lpr background, mice that were MRL/lpr/Igκ^αhel/αhel or MRL/lpr/Igκ^Vκ4/Vκ4 were crossed with MRL/lpr/Igκ^hCκ/hCκ mice. B6 and MRL/lpr without a LC transgene were bred separately in the same facility. Mice were kept and bred in the animal facility at the Oklahoma Medical Research Foundation and University of Chicago, and experiments were approved by the Institutional Animal Care and Use Committee. Unless stated otherwise, all mice were 7–10 wk of age and age matched for each experiment.

Single-cell suspensions were derived from mouse BM, LNs, and spleens. Cells were then stained with antibodies recognizing mouse κ (187.1; BD), human κ (goat anti–human κ F(ab′)2 mouse absorbed; SouthernBiotech), λ (JC-5; SouthernBiotech), B220 (RA3-6B2; BD), IgM (1B4B1; SouthernBiotech), CD93 (AA4.1; eBioscience), CD21 (7G6; BD), CD23 (B3B4; eBioscience), CD24 (M1/69; BioLegend), IgG1/2 (A85-1/R2-40; SouthernBiotech), and CD43 (S7; BD). Flow cytometry was performed on an LSR II (BD), and data were analyzed using FlowJo software (Tree Star).

B cells from mouse spleens were enriched using a RosetteSep murine B cell isolation kit (STEMCELL Technologies) or MACS CD43 depletion (Miltenyi Biotec) and stained with antibodies described above. Cells were presorted in bulk using a FACSAria cytometer (BD), and single cells were resorted into 96-well plates containing 10 mM Tris-HCl with 40 U/µl RNase inhibitor (Promega). Single-cell RT-PCR for RAG expression was performed as previously described (Yannoutsos et al., 2001). For antibody variable gene amplification, each cell was amplified in a one-step RT-PCR reaction (OneStep RT-PCR kit; QIAGEN) using a cocktail of sense primers specific for the leader region and antisense primers specific for either mCκ or hCκ, or Fcµ/δ for HC genes (primer sequences are available in Table S1). RT-PCR conditions were as follows: 50°C for 30 min, 95°C for 15 min, followed by 39 cycles of 95°C for 1 min, 55°C for 1 min, 72°C for 1 min, and, finally, 72°C for 10 min. 1 µl from each RT-PCR reaction was reamplified using nested PCR primers specific for FW1 regions of the various Vκ genes and antisense primers specific for either mCκ or hCκ, or Fcµ/δ for HC genes. Nested PCR reaction conditions were as follows: 95°C for 15 min, followed by 39 cycles of 95°C for 1 min, 55°C for 1 min, 72°C for 1 min, and finally 72°C for 10 min. To amplify LCs from both human and mouse alleles in allelically included cells, cDNA synthesis from single cells was first performed using random primers, followed by two-step amplification of Ig transcripts using specific primers as described above. Efficiency of LC amplification was typically ∼70%.

mAbs were produced as previously published (Wardemann et al., 2003; Wrammert et al., 2008; Duty et al., 2009; Smith et al., 2009). In brief, the Ig HC and LC variable genes from single-sorted B cells amplified by RT-PCR were sequenced, amplified, and cloned into expression vectors and expressed with human IgG constant regions in the 293A human cell line. The antibodies produced were then purified from the serum-free tissue culture media using Protein A beads.

To measure the autoreactivity of the mAbs, we tested each antibody for HEp-2 binding activity using a human HEp-2 ANA ELISA kit (Alpha Diagnostic) according to the manufacturer’s instructions. For the anti-dsDNA ELISAs, microtiter plates (Costar; Corning) were coated with 10 µg/ml calf thymus dsDNA (Invitrogen). Goat anti–human IgG (γ chain specific) peroxidase conjugate (Jackson ImmunoResearch Laboratories, Inc.) was used to detect binding of the recombinant antibodies followed by development with AquaBlue ELISA substrate (eBioscience). Paired antibodies sharing the same HC were tested on the same plate at the same time. Absorbencies were measured at OD405 when the control antibody 3H9 included on every plate reached an OD of 3.0. mAbs were diluted to a starting concentration of 25 µg/ml followed by three additional threefold serial dilutions for HEp-2 cell ELISAs and 1 µg/ml followed by three additional fourfold serial dilutions for dsDNA ELISAs. Data were plotted, and area under the curve (AUC) was determined with Prism software (GraphPad Software). To determine Ig concentrations in mouse sera, 96-well plates were coated with anti–mouse IgG and blocked with PBS/0.5% BSA. Sera were diluted to 1:10,000 (B6) or 1:100,000 (MRL/lpr) to fall within the concentration range of the IgG standard control. Detection was performed with HRP-conjugated anti–mouse IgG (SouthernBiotech). To determine the levels of anti-dsDNA antibodies in mouse sera, 1:100 dilutions of serum were added to microtiter plates coated with calf-thymus DNA and blocked with PBS + 5% BSA. Bound Igs were detected with HRP-conjugated anti–mouse IgG. To normalize results across plates, each plate had serum from the same MRL/lpr mouse, and the absorbance was read when this sample reached an OD405 of 3.0.

Multiscreen HTS filter plates (EMD Millipore) were coated overnight with mouse or human anti-κ antibody followed by incubation with 3 × 10⁶/well total splenocytes overnight and detection of secreted antibodies with biotin-conjugated anti–mouse IgG (SouthernBiotech) and streptavidin peroxidase.

mAbs at 50 µg/ml were screened on commercial ANA (HEp-2 cell) slides (Bion Enterprises, Ltd.) according to the manufacturer’s recommended protocol. 3T3 cells were grown on chambered slides and fixed with acetone before staining with mAbs according to the same protocol as with the HEp-2 slides. Paired antibodies sharing the same HC were tested on the same slide at the same time. HEp-2 and 3T3 slides were analyzed using a fluorescent microscope (Axioplan II; Carl Zeiss). All HEp-2 and 3T3 reactivity was compared with positive and negative control sera included in the ANA commercial kit. Binding to HEp-2 and 3T3 cells by IMF was independently scored in a blinded manner on a scale of 0–4 by several individuals.

Detection of RS rearrangements was performed as described previously (Panigrahi et al., 2008). In brief, genomic DNA was isolated from sorted cells using a Puregene kit (QIAGEN). Quantitative PCR was then performed using Taqman technology (Applied Biosystems) with primers and probes specific to RS elements and actin (Meng et al., 2011).

Genomic DNA was isolated from sorted cells, purified according to the manufacturer’s directions using Puregene, and amplified using the J606.1 or J558.85 VH primers and a labeled JH2 reverse primer as described previously (Meng et al., 2011). All primers were synthesized by Integrated DNA Technologies. 2 µl PCR product per reaction was resolved by capillary electrophoresis on an ABI 3100 analyzer (Applied Biosciences). Peak scanner version 1.0 was used to generate and analyze the spectratypes (Applied Biosciences) To measure the frequency of individual rearrangements accurately, the input DNA was diluted to very low concentrations (up to 2,000 cells) in each independent PCR reaction, so that each spectratype of a given size corresponds to only one rearrangement.

Statistics were performed using Prism software (analyzing column and grouped data by Student’s t test and by ANOVA) or Microsoft Excel (for χ² test statistics). Unless stated otherwise in the figure legend, statistical significance was determined using a two-tailed Student’s t test of independent sample means.

Fig. S1 shows B cell developmental subsets in LC transgenic mice. Fig. S2 shows LC usage in mature B cell populations and sort purity. Table S1 lists primers used in this study.

We thank Marin Weigert for helpful discussions.

This work was supported by National Institutes of Health grants to P.C. Wilson (including RO1AI76585-01 and U19AI082724-02) and grants from the Arthritis Foundation and Lupus Research Institute to E.T.L. Prak.

The authors declare no financial or commercial conflicts of interest.