Abstract During B lymphopoiesis, B cell progenitors progress through alternating and mutually exclusive stages of clonal expansion and immunoglobulin (Ig) gene rearrangements. Great diversity is generated through the stochastic recombination of Ig gene segments encoding heavy and light chain variable domains. However, this commonly generates autoreactivity. Receptor editing is the predominant tolerance mechanism for self-reactive B cells in the bone marrow (BM). B cell receptor editing rescues autoreactive B cells from negative selection through renewed light chain recombination first at Igκ then Igλ loci. Receptor editing depends upon BM microenvironment cues and key transcription factors such as NF-κB, FOXO and E2A. The specific BM factor required for receptor editing is unknown. Furthermore, how transcription factors coordinate these developmental programs to promote usage of the λ-chain remain poorly defined. Therefore, we utilized two mouse models that recapitulate pathways by which Igλ light chain positive B cells develop. The first possess deleted J kappa (Jκ) genes and, as such, models Igλ expression resulting from failed Igκ recombination (Igκdel). The second models autoreactivity by ubiquitous expression of a single-chain chimeric anti-Igκ antibody (κ-mac). Here, we demonstrated that autoreactive B cells transit asymmetric forward and reverse developmental trajectories. This imparted a unique epigenetic landscape on small pre-B cells, which opened chromatin to transcription factors essential for Igλ recombination. The consequences of this asymmetric developmental path were both amplified and complemented by CXCR4 signaling. These findings reveal how intrinsic molecular programs integrate with extrinsic signals to drive receptor editing. One sentence summary: The molecular pathways of BCR editing consist of a unique epigenetic landscape activated and amplified by both BCR and CXCR4 signals. Introduction During B lymphopoiesis, B cell progenitors progress through alternating and mutually exclusive stages of clonal expansion and immunoglobulin (Ig) gene rearrangements ([48]1-[49]3). Great diversity is generated through the stochastic, nonhomologous recombination of Ig gene segments encoding heavy (IgH) and light (IgL) chain variable domains. However, this commonly generates autoreactivity. Indeed, up to 75% of immature B cells express autoreactive antigen receptors ([50]4-[51]6). Therefore, the initial B cell repertoire must be purged of autoreactivity in the bone marrow (BM), before transit to the periphery. B cells are primarily tolerized in the BM by either negative selection (apoptosis) or B cell receptor editing. Receptor editing is the predominant BM central tolerance mechanism ([52]7-[53]12). Receptor editing rescues autoreactive B cells by inducing renewed light chain recombination in attempts to edit antigen specificity away from “self” ([54]13-[55]15). Light chain editing occurs first at Igκ then proceeds to Igλ loci. In mice, once at Igλ, the choice of recombination partners is limited to four major products: λ1, λ2, λ3 and λx ([56]16). λ-chains, both in humans and mice, have distinct physiochemical properties in comparison to κ-chain that are postulated to aid in rescuing polyreactive BCRs during development ([57]17). In particular, λx has acidic Asp residues in the complementarity-determining regions 3 (CDR3) loop that quench IgH autoreactivity to DNA ([58]12, [59]13, [60]16, [61]18). In this manner, autoreactivity promotes Igλ-chain usage. Failure to successfully rearrange the Igκ alleles can also lead to Igλ recombination. Indeed, single cell studies indicate that most initial Vκ-Jκ rearrangements are out of frame ([62]3). Therefore, there are two very different scenarios leading to Igλ recombination: autoreactivity and failed Igκ recombination. In the instance of failed Igκ rearrangement, recombination progresses to Igλ and continues until recombination is successful or the cell is deleted. In contrast, mouse models indicate a more complex developmental path for editing B cells ([63]11, [64]13, [65]19-[66]21). At the immature B cell stage, self-antigen induces loss of BCR surface expression and ‘de-differentiation’ back towards the small pre-B cell stage ([67]20). This enables RAG1/2 expression ([68]22) and renewed light chain recombination at both Igκ and Igλ ([69]13, [70]21, [71]22). Receptor editing is also uniquely dependent on NF-κB signals, which promote survival; thereby, prolonging the developmental window for successive Igλ recombination ([72]23). E2A proteins are required for Igλ germline transcription ([73]24). It is unlikely that simple back-differentiation is sufficient to provide all the molecular conditions required for efficient receptor editing. Finally, the BM provides environmental cues critical for receptor editing ([74]25). The nature of these BM-derived extrinsic signals is unknown. One candidate is the C-X-C motif chemokine ligand 12 (CXCL12), which engages the C-X-C chemokine receptor type 4 (CXCR4), to mediate retention of immature B cells in the BM ([75]26). CXCR4 also activates signaling pathways necessary for efficient Igκ recombination ([76]27). However, if CXCR4 plays a similar role in Igλ recombination is unknown. Here, we demonstrated that editing B cells transit asymmetric retrograde and anterograde epigenetic trajectories that impart a unique landscape containing accessible regions from both early and late developmental stages. These open sites are enriched for E2A and NF-κB transcription factors binding motifs. In addition to this intrinsic program, CXCR4 provides critical extrinsic cues including opening the Igλ locus to recombination. The resulting cellular state enables Igλ recombination and ensures cell fate decisions are determined by BCR specificity in the appropriate environmental context. This study furthers our understanding of the molecular mechanisms of early B cell development that promote self-tolerance and antibody diversification. Results Two developmental paths to Igλ recombination Igλ expressing B cells represent a small (~10%) proportion of the wild-type (WT) mouse B cell repertoire. To circumvent this limitation, previous work has relied on transgenic expression of rearranged Igλ BCRs in mice or hybridomas ([77]15, [78]28). However, our understanding of B cell receptor editing in a polyclonal repertoire is limited. Therefore, we utilized two complementary polyclonal mouse models that recapitulate the different pathways by which Igλ bearing B cells develop and are selected. The first has a deletion in the J kappa (Jκ) genes and as such models Igλ expression resulting from failed Igκ recombination (Igκ^del) ([79]29). The second model utilizes a ‘κ-macroself’ transgene, which encodes for ubiquitous expression of a single-chain chimeric anti-Igκ antibody (κ-mac) ([80]10). In this model, newly formed Igκ expressing immature B cells are “autoreactive” and forced to edit away from Igκ-chain expression in favor of Igλ expression to continue differentiation. Flow cytometric analysis of WT (CD45.1), Igκ^del and κ-mac bone marrow revealed equivalent BM cellularity but different distribution of B cell progenitor subsets ([81]Fig. 1A and [82]1B). Pre-pro B cells (B220+CD19−) were expanded in the κ-mac mice compared to WT and Igκ^del mice ([83]Fig. 1C). However, fully committed pro-B (B220^+CD19^+CD43^+) and subsequent large pre-B (B220^+CD19^+CD43^−IgM^−FSC^hi) cell numbers were similar between the three genotypes ([84]Fig. 1D and [85]1E). Notably, Igκ^del mice had an expanded small pre-B cell (B220^+CD19^+CD43^−IgM^−FSC^lo) compartment compared to WT and κ-mac mice ([86]Fig. 1F). Because light chain recombination is sequential, it was possible that preventing Igκ recombination impaired subsequent Igλ recombination. However, we observed similar numbers of immature B (B220^loCD19^+CD43^−IgM^+) and recirculating mature B cells between WT and Igκ^del mice ([87]Fig. 1G and [88]1H). These populations were reduced in κ-mac mice likely due to increased negative selection associated with extensive receptor editing. These findings were similar to those found in other studies using Igκ^del and κ-mac mice ([89]10, [90]12). Nevertheless, BCR surface expression on immature B cells was similar across all three genotypes ([91]Fig. 1I). Together, these results demonstrate that neither loss of productive Igκ recombination nor the κ-macroself transgene severely impair B cell development. Figure 1. Two molecular paths to Igλ recombination. [92]Figure 1. [93]Open in a new tab A, Representative flow cytometric analysis of different B lymphopoiesis developmental stages in the BM of WT (CD45.1), Igκ^del and κ-mac mice. B cell progenitor subpopulations are defined as follows: Pre-pro-B cells (B220+CD19−), pro-B cells (B220+CD19+CD43+IgM−), large pre-B cells (B220loCD19+CD43−IgM−FSChi) and small pre-B cells (B220loCD19+CD43−IgM−FSClo) immature B cells (B220loCD19+CD43−IgM+) and mature recirculating B cells (B220hiCD19+CD43−IgM+). FSC; Forward Scatter. B, Absolute numbers of bone marrow cells from WT (CD45.1) (n=7), Igκ^del (n=7) and κ-mac mice (n=8). C, Absolute number of Pre-Pro B cells from WT (CD45.1), Igκ^del and κ-mac mice. D, Absolute number of Pro B cells from WT (CD45.1), Igκ^del and κ-mac mice E, Absolute number of large pre-B cells from WT (CD45.1), Igκ^del and κ-mac mice. F, Absolute number of small pre-B cells from WT (CD45.1), Igκ^del and κ-mac mice. G, Absolute number of immature B cells from WT (CD45.1), Igκ^del and κ-mac mice. H, Absolute number of mature recirculating B cells from WT (CD45.1), Igκ^del and κ-mac mice. I, Flow cytometric analysis of the corresponding cell-surface expression of IgM (top panel) and quantification of geometric mean fluorescence intensity (gMFI, bottom panel) on immature B cells from WT (CD45.1), Igκ^del and κ-mac mice. J, Flow cytometric analysis of Igκ surface expression on WT (CD45.1), Igκ^del and κ-mac B cells. B-I, Data are presented as means ± standard error of the mean (S.E.M). P values were determined by ANOVA with Tukey multiple testing (* p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001). To investigate gene-regulatory networks that govern receptor editing and Igλ recombination, we first used FACS to isolate BM cells from WT, Igκ^del and κ-mac small pre-B cells as well as WT Igκ+, Igκ^del and κ-mac Igκ− immature B cells ([94]Fig. 1SA). Consistent with published studies, Igκ^del and κ-mac mice did not produced Igκ+ B cells, while WT mice overwhelmingly produced Igκ+ B cells ([95]Fig. 1J) ([96]10, [97]29). This demonstrates κ-mac and Igκ^del mice are suitable models to investigate dynamics of Igλ recombination. Increased accessibility at E2A and NF-κB motifs in editing B cells For these and subsequent studies, we sorted Igκ− immature B cells from Igκ^del and κ-mac mice because commercially available anti-Igλ antibodies are specific for 3 of the 4 known Igλ products: λ1, λ2, λ3 but not λx. We first analyzed the ATAC-seq data using ChromVAR to assess transcription factor motif enrichment within open chromatin regions (OCRs) ([98]30). ChromVAR defines a ‘deviation score’, which is a measure of the change in accessibility at a particular motif within a sample compared to the average accessibility of that motif in all samples analyzed. Clustering using Pearson correlations of normalized deviation scores between sorted samples revealed that small pre-B and immature B cells possessed divergent chromatin accessibility landscapes ([99]Fig. 2SA). Within each developmental stage, there were differences between WT, Igκ^del and κ-mac genotypes. In particular, within small pre-B cells, κ-mac cells clustered away from WT and Igκ^del cells, which themselves were relatively similar. In contrast, within immature B cells, WT and κ-mac cells were more similar to each other than Igκ^del cells. These data suggest that receptor editing is associated with an epigenetic reprogramming in small pre-B cells. Next, we assessed the transcription factor motifs selectively enriched between cell types ([100]Fig. 2A). Numerous transcription factor motifs necessary for early B cell development such as EBF1 ([101]31), PU.1 ([102]32-[103]34) and MEF2C/MEF2D ([104]35, [105]36) showed large changes in accessibility across the small pre-B to immature B cell transitions. WT, Igκ^del small pre-B cells and Igκ^del immature B cells showed increased accessibility at transcription factor motifs involved in chromatin structure (CCCTC-binding factor, CTCF), and the CTCF paralog BORIS, compared to WT and κ-mac small pre-B and immature B cells. κ-mac small pre-B cells were selectively enriched for accessibility at E-box motifs (HEB and E2A) and Early B cell Factor 1 (EBF1) motifs. Figure 2. κ-mac cells enriched for E2A and NF-κB motifs and Igλ accessibility. [106]Figure 2. [107]Open in a new tab A, Heatmap of average top transcription factor motif deviation scores in indicated cell populations. Red indicates an increase in deviation score and blue indicates a decrease. Hierarchical clustering performed on Euclidean distances. B, Top transcription factor motifs ranked by motif variability score. C, T-stochastic neighbor embedding of transcription factor motifs for SPI-B, E2A and composite motifs for NF-κB-p65-REL in indicated cell populations. Scale bar represents z score of deviations calculated using the ChromVAR R package. D, Meta-analysis of accessibility (ATAC-seq) over the length of each Igλ gene body or regulatory element for WT (CD45.1) (Blue), Igκ^del (Red) and κ-mac (Gold) small pre-B cells. E, Accessibility calculated from – 10% to +10% relative to the transcription start site (TSS) for Vλ and Jλ small pre-B cells in Igκ^del (Red) and κ-mac (Gold) small pre-B cells and Igκ^del (Orange) and κ-mac (Purple) immature B cells. F, Accessibility calculated from – 10% to +10% relative to the transcription start site (TSS) for Igκ genes (Vκ, Jκ, Cκ) in indicated cell populations. G, Accessibility calculated from – 10% to +10% relative to the transcription start site (TSS) for Jκ in indicated cell populations. D-G, Error bars represent S.E.M. We next ranked transcription factors with the greatest change in accessibility by using ChromVAR’s ‘variability score’, which is the standard deviation of the deviation z-scores across samples ([108]Fig. 2B). The highest enrichment was for E protein motifs, which were highly represented in κ-mac small pre-B cells. IRF, ETS and NF-κB motifs were also among the highly ranked transcription factor motifs. T-distributed stochastic neighbor embedding (t-SNE) showed developmental stage dependent clustering with small pre-B and immature B cells forming separate clusters ([109]Fig. 2C and [110]2SB). Within these clusters, NF-κB and SPIB motifs were enriched in WT and κ-mac immature B cells compared to Igκ^del immature B cells, while E2A was enriched within κ-mac small pre-B cells. These findings are consistent with the requirement of both E2A and NF-κB for receptor editing ([111]23, [112]24, [113]37-[114]39). Furthermore, Foxhead motifs (FOXO3) and FOXO:Ebox composite motifs were specifically enriched in κ-mac small pre-B and immature B cells, while enrichment for IRF4 motifs were shared between WT and κ-mac immature B cells ([115]Fig. 2SB). Together, these results demonstrate that κ-mac small pre-B OCRs are enriched for E2A sites, while both κ-mac small pre-B and immature B cells preferentially open FOXO and FOXO:E2A binding sites. Selective opening of Igλ locus in editing B cells Next, we quantified chromatin accessibility at Igκ and Igλ loci across all encoding genes and regulatory segments. These results were then plotted as a function of relative position across all genetic elements (i.e., gene body and enhancers). To capture accessibility at recombination signal sequences (RSS) and gene promoters, flanking regions immediately before and after gene segments were included. Remarkably, the Igλ locus of κ-mac small pre-B cells (gold) showed increased accessibility compared to both Igκ^del (red) and WT (blue) small pre-B cells ([116]Fig. 2D). Accessibility at Vλ and Jλ gene segments was increased in κ-mac small pre-B, while being moderately increased in immature B cells ([117]Fig. 2E). In contrast, accessibility at Igκ remained largely unchanged in Igκ^del and κ-mac small pre-B and immature B cells ([118]Fig. 2F). As expected, Igκ^del mice had no accessibility signal at the Jκ locus in small pre-B and immature B cells due to genomic deletion ([119]Fig. 2G). Altogether, these data suggest receptor editing is associated with selective opening of the Igλ locus. Unique chromatin landscape in editing small pre-B cells We next performed an in-depth analysis of OCRs in WT, Igκ^del and κ-mac small pre-B cells. As expected, most OCRs (25,142) were shared between all three small pre-B genotypes ([120]Fig. 3A). In contrast, κ-mac small pre-B cells had ~11k unique OCRs compared to ~1.4k and ~1.7k OCRs in WT and Igκ^del cells, respectively. We then annotated the unique accessible regions to genomic locations such as promoters, exons, introns, and intergenic regions ([121]Fig. 3B). Unique κ-mac OCRs were found at a higher frequency at promoters and exons, compared to WT and Igκ^del OCRs that were somewhat frequent at promoters but not at exons. Igκ^del and WT small pre-B accessible regions were mostly found at intergenic regions. Annotation enrichment analysis of observed/expected genomic annotations confirmed enrichment at promoters and exons in κ-mac small pre-B cells compared to Igκ^del and WT small pre-B cells ([122]Fig. 3C). Figure 3. Unique chromatin landscape in editing small pre-B cells. [123]Figure 3. [124]Open in a new tab A, Total and overlapping open chromatin regions (ATAC-Seq) in flow-purified WT (CD45.1), Igκ^del and κ-mac small pre-B cell populations. B, Distribution of unique accessible regions identified in A across the genome in promoters (Prom), Exon, Intron, and intergenic regions (Inter.) Roman numerals indicate unique regions: i) WT (CD45.1) (n=2), ii) Igκ^del (n=2), iii) κ-mac ( n=3) small pre-B cells. C, Enrichment statistic for peak overlap with different sets of annotation identified in B. D, Gene ontology analysis of genes near unique accessible regions identified in A. ‘Wikipathway’ gene sets were used for this analysis. Scale bar represents log transformed P values obtained using HOMER (see methods). E&H, De Novo transcription factor binding motifs associated with unique and common accessible regions. P value and percent of Target (% of Target) determined using HOMER (see methods). F-H, ATAC-seq signal single-base-pair resolution at motif-centered peaks containing the de novo discovered motifs of CTCF, E2A and SPI-B in WT (CD45.1) (Magenta), Igκ^del (Orange), or κ-mac (Purple) small pre-B cells. A 300bp (top panel) or 3000bp (middle panel) window around centered motif is shown. Bottom panel shows quantification of accessibility with a 300bp window of centered motif. Boxes represent interquartile ranges (IQRs; Q1–Q3 percentile) and black horizontal lines represent median values. Maximum and minimum values (ends of whiskers) are defined as Q3 + 1.5× the IQR and Q1 − 1.5× the IQR, respectively. Outliers are indicated as black dots along the whiskers. Statistical significance determined using ANOVA followed by Kruskal-Wallis multiple comparisons (* p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001). We then performed gene ontology (GO) pathway analysis of genes in the vicinity of the unique accessible regions ([125]Fig. 3D). Genes near κ-mac OCRs were involved in NF-κB and P53 signaling, which are important for light chain recombination and DNA repair, respectively ([126]23, [127]37, [128]38, [129]40, [130]41). κ-mac OCRs were found near genes involved in cytoprotective response to reactive oxygen species (Keap1-Nrf2 pathway) and chemotaxis. These observations could reflect unique epigenetic programs in editing κ-mac small pre-B cells. Analysis of motifs within accessible regions shared by all three genotypes termed ‘Common - Motifs’ revealed enrichment for SPIB, CTCF, E2A, PRMD1, RUNX, and MEF2C ([131]Fig. 3E). However, single base-pair resolution mapping of accessibility around these shared transcription factor motifs suggested preferential binding of CTCF, SPI-B and E2A in κ-mac small pre-B cells ([132]Fig. 3F-[133]H). Next, we examined transcription factor binding motifs within OCRs in WT, Igκ^del and κ-mac small pre-B cells identified in [134]Figure 3A ([135]Fig. 3I). CTCF was a common motif in all accessible regions. Nuclear Factor of Activated T cells (NFAT) was enriched in WT accessible regions and TWIST1 was enriched in Igκ^del accessible regions. In contrast, unique κ-mac accessible regions were enriched for the NF-κB subunit RELB and PU.1 containing composite motifs with IRF and SPIB transcription factors ([136]Fig. 3I). These findings suggest editing small pre-B cells adopt a chromatin landscape poised for Igλ recombination with increased accessibility and transcription factor motif enrichment for known mediators of receptor editing and light chain recombination ([137]23, [138]24, [139]37, [140]42). Unique κ-mac small pre-B epigenetic state Autoreactive immature B cells back-differentiate towards the small pre-B compartment upon down-modulation of the autoreactive BCR ([141]20). We wondered whether the ~11k chromatin regions unique to κ-mac small pre-B cells were shared with WT immature B cells. Indeed, 89% of κ-mac unique peaks were represented in immature B cells ([142]Fig. 3SA). Moreover, these common OCRs were enriched for key transcription factor motifs such as those for bound by SPIB, E2A and NF-κB. Gene pathway enrichment analysis of genes in the vicinity of these shared OCRs were strongly enriched for the MAPK and P53 signaling pathways ([143]Fig. 3SB). Indeed, accessibility at genes important for these pathways such as Cxcr4, Keap1, Ptpn22 and Traf6 all showed increased accessibility in κ-mac small pre-B cells compared WT and Igκ^del small pre-B cells ([144]Fig. 3SC-[145]3SF). Thus, the epigenetic landscape of κ-mac small pre-B cells is a unique blend of small pre-B and immature B cell epigenetic states reflecting the complex developmental trajectory of receptor editing B cells. Small pre-B and immature κ-mac B cells have unique transcriptional programs To investigate the transcriptional differences between editing cells and non-editing cells, we performed RNA-sequencing on FACS isolated small pre-B and immature B cells from WT, Igκ^del and κ-mac mice. As expected, principal component analysis (PCA) indicated distinct differences between small pre-B and immature B cells, regardless of genotype ([146]Fig. 4A). Strikingly, small pre-B cells clustered together suggesting remarkable similarity in transcriptional states. In contrast, immature B cells showed genotype dependent clustering. Figure 4. Receptor editing initiates unique transcriptional programs. [147]Figure 4. [148]Open in a new tab A, Principal component analysis (PCA) of differentially expressed genes (DEGs) RNA-seq data from these cell populations: WT (CD45.1) immature B cells (Magenta), Igκ^del immature B (Orange), κ-mac immature B (Purple), WT (CD45.1) small pre-B (Light Blue), Igκ^del small pre-B (Red) and κ-mac small pre-B (Gold). Replicates are shown. B, Spearman correlation plot of DEGs from cell populations. Scale indicates spearman correlation coefficients. C, Heatmap of DEGs between κ-mac (Gold) and Igκ^del (Red) small pre-B cells. Scale bar indicated z score of log transformed counts per million: z score log2(CPM). Right panel indicates a subset of DEGs upregulated in κ-mac small pre-B cells. Representative gene pathways these genes belong to identified using Metascape program. Replicates are shown. D, Volcano plot of DEGs expressed genes between Igκ^del and κ-mac immature B cells. Red dots are genes with significantly increased (right side) or decreased (left side) expression in κ-mac immature B cells. False discovery rate (FDR) was performed on P values using Benjamini-Hochberg correction. E, Gene set enrichment analysis (GSEA) for indicated terms between κ-mac and Igκ^del immature B cells. NES: Normalized enrichment score”. F, Flow cytometric analysis of MitoSOX staining in immature B cells from indicated populations. Data presented as mean MitoSOX+ cells and ± 90% confidence interval. N=3 for all genotypes. G, MFI of MitoSOX staining in WT (CD45.1) (Magenta), Igκ^del (Orange) and κ-mac (purple) immature B cells. N=3 for all genotypes. H, Representative flow cytometric analysis of immature B cells. IgM^lo population is indicated by red box in upper-left panel and IgM^hi population indicated in bottom left panel. MitoSOX MFI was quantified in for the IgM^lo and IgM^hi population in the upper and lower right panels respectively. N=3 for all genotypes. F-H, Data are presented as means ± standard error of the mean (S.E.M). P values were determined by ANOVA followed by Tukey multiple comparisons (* p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001). Spearman correlation of the 10,408 (FDR < 0.05) differentially expressed genes between WT, Igκ^del and κ-mac small pre-B and immature B cells confirmed the PCA clustering ([149]Fig. 4B). Independent of genotype, small pre-B cells were highly correlated with Igκ^del and κ-mac small pre-B cells being virtually identical (spearman correlation: 0.98). In contrast, spearman correlation values indicated that WT, Igκ^del and κ-mac immature B cells were transcriptionally more distinct with values ranging from 0.67 to 0.83. Consistent with their transcriptional similarity, pairwise comparisons between Igκ^del and κ-mac small pre-B cell transcriptional profiles revealed only 285 differentially expressed genes ([150]Fig. 4C). 230 genes were upregulated in κ-mac small pre-B cells and 55 genes were upregulated in Igκ^del cells. Metascape gene annotation analysis revealed κ-mac upregulated genes involved in DNA repair, histone methylation and NF-κB signaling ([151]Fig. 4C; right panel)([152]43). Altogether, these data suggest that at the small pre-B cells stage, κ-mac cells begin to establish a distinct transcriptional program predicted to be permissive for receptor editing. Next, we performed differential expression analysis on Igκ^del and κ-mac immature B cells. 783 genes were significantly (FDR < 0.05) upregulated in Igκ^del immature B cells, and 844 genes were significantly (FDR < 0.05) upregulated in κ-mac immature B cells ([153]Fig. 4D). κ-mac immature B cells were enriched for DNA repair pathways ([154]Fig. 4SA), including Rag2 and Gadd45a. Gadd45a is a gene induced by DNA damage and has been shown to regulate Rag1 and Rag2 expression during receptor editing ([155]42, [156]44). Metabolic pathways such as oxidative phosphorylation (OxPhos), glycolysis and fatty acid oxidation as well as translation initiation, P53 signaling, DNA repair were also enriched in κ-mac immature B cells ([157]Fig. 4E and [158]4SA). In contrast, GSEA identified gene networks involved in chromatin organization and phosphoinositide 3-kinase (PI3K) signaling in Igκ^del immature B cells ([159]45, [160]46). Consistent with this, we observed increased expression of Brwd1, Brd4, Cbl in Igκ^del immature B cells ([161]Fig. 4D). PI3K signaling represses DNA recombination ([162]1, [163]40, [164]47, [165]48). These data are consistent with the persistence Igλ recombination in κ-mac immature B cells. To validate whether OxPhos was heightened in κ-mac immature B cells, we measured mitochondrial reactive oxygen species production (mitoROS) by flow cytometry. We used the MitoSOX™, which is rapidly oxidized by superoxide produced in mitochondria and becomes highly fluorescent. Consistent with our GSEA results, κ-mac immature B cells showed increased mitoROS production compared to WT and Igκ^del immature B cells ([166]Fig. 4F). Within the immature B cell compartment, cells with lower BCR expression (IgM^lo) exhibited higher mitoROS than cells with higher BCR expression (IgM^hi) ([167]Fig. 4G and [168]4H). Moreover, IgM^lo κ-mac cells exhibited the highest overall mitoROS levels indicating that cells initiating receptor editing increase OxPhos ([169]37). Increases in mitoROS can cause mitochondrial dysfunction and lead to cell death ([170]49). We thus searched the genes within the GSEA ROS pathway to identify antioxidant genes that may regulate the levels of mitoROS production. Indeed, superoxide dismutase 1 and 2 (Sod1, Sod2) were increased in κ-mac immature B cells compared to Igκ^del immature B cells ([171]Fig. 4SB and [172]4SC). Altogether, these data show cells undergoing receptor editing upregulate the metabolic processes OxPhos and ROS production. Antigen recognition by the BCR initiates both signaling cascades and antigen delivery along the endocytic pathway to MHC class II antigen-presenting compartment ([173]50). We sought to determine whether BCR downregulation associated with receptor editing was associated with targeting of endocytosed BCRs to the MHC class II antigen-presenting compartment. Thus, we sorted small pre-B and immature B cells from WT and κ-mac mice and stimulated them in vitro through the BCR for 30 minutes. Cells were then fixed and stained with endosomal marker Lamp-1 ([174]Fig. 4SD and [175]4SE). As expected, in small pre-B cells, the pre-BCR was excluded from the MHC class II antigen-presenting compartment ([176]51). In contrast in WT immature B cells, the endocytosed BCR targeted the MHC class II antigen-presenting compartment. However, in κ-mac immature B cells, these endocytosed receptors were largely excluded from the MHC class II antigen-presenting compartment ([177]Fig. 4SF-[178]4SG). These results suggest κ-mac immature B cells maintain some small pre-B cell metabolic and endocytic pathways. CXCR4 deficiency impairs Igλ+ B cell formation Given that Igκ^del and κ-mac mice differ at genomic regions encoding CD45.1 and CD45.2 respectively, we next determined whether this significantly influenced the differences we observed between Igκ^del and κ-mac mice ([179]52). We used FACS to isolate large pre-B, small pre-B and immature B cells from CD45.1 and CD45.2 mice and performed RNA-seq. PCA of mRNA expression revealed cell type dependent grouping of sorted populations that was not influenced by CD45 allotype ([180]Fig 5SA). Next, we performed pairwise comparisons between cells from CD45.1 and CD45.2 genotypes and observed 15 genes were significantly upregulated in CD45.1 cells compared to 134 genes upregulated in CD45.2 cells ([181]Fig 5SB). Furthermore, when we repeated the comparison focusing on small pre-B or immature B cells only 35 genes were differentially expressed between small pre-B cells and only 67 genes were differentially expressed between immature B cells ([182]Fig. 5SC-[183]5SD). Lastly, when these genes were compared to the DEGs found between Igκ^del and κ-mac mice, only 85 genes were in common ([184]Fig. 5SE). Taken together, the underlying genetic difference between the Igκ^del and κ-mac mice were unlikely to be strongly influencing the results of our experiments. ATAC-seq analysis demonstrated chemokine signaling pathways were enriched in κ-mac cells ([185]Fig. 3D). In particular, Cxcr4, a key regulator of B cell chemotaxis, was among the upregulated genes in κ-mac immature B cells ([186]Fig. 4D). This was confirmed in both immature B cells and small pre-B cells by flow cytometry ([187]Fig. 5SF-[188]5SG). Of note, the observed difference in CXCR4 expression was likely an underestimation as κ-mac are on the B6-CD45.1 background, which possesses genomic mutations that diminish expression of CXCR4 relative to B6-CD45.2 mice, which is the Igκ^del background ([189]53). RNA-seq analysis of the Cxcr4 locus confirmed the presence of mutations in κ-mac, but not Igκ^del, mice associated with decreased CXCR4 expression ([190]Fig. 5SH). Yet, despite this genetic predisposition, κ-mac small pre-B and immature B cells highly expressed CXCR4. CXCR4 not only mediates small pre-B chemotaxis but the CXCR4/CXCl12 signaling axis regulates small pre-B cell differentiation and Igκ recombination ([191]27). To determine if CXCR4 also has a role in receptor editing, we generated κ-mac Cxcr4^fl/flmb1-cre^+ (hereafter κ-mac Cxcr4KO) and κ-mac Cxcr4^fl/fl mice. Consistent with our previous results ([192]27), pro-B and large pre-B populations were largely unaffected in κ-mac Cxcr4KO mice ([193]Fig. 5A-[194]5B). However, small pre-B cells were 3-fold reduced, while immature and recirculating mature B cells were 8-fold and 36-fold reduced, respectively. Figure 5. CXCR4 deficiency impairs Igλ+ B cell development. [195]Figure 5. [196]Open in a new tab A, Representative flow cytometric analysis of different developmental stages of B lymphopoiesis in the BM of κ-mac Cxcr4^fl/fl mb1-Cre− (κ-mac Cxcr4^fl/fl ) and κ-mac Cxcr4^fl/fl mb1-cre+ (κ-mac Cxcr4KO) mice. B, Absolute cell numbers for populations shown in A. Data are presented as means ± standard error of the mean (S.E.M). P values were determined by Student’s T test (* p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001). C, Heatmap of DEGs between κ-mac Cxcr4^fl/fl (n=2) and κ-mac Cxcr4KO (n=3) small pre-B cells. Scale indicates z-score of Log[2] transformed normalized counts: z score log[2](CPM). Replicates are shown. D, Metascape gene pathway analysis of differentially expressed genes identified in C. Scale bar indicates Log[10] transformed P values determined by Metascape software. E, Heatmap of representative genes in indicated pathways. Scale indicates z-score of Log[2] transformed normalized counts: z score log[2](CPM). Replicates are shown. Bulk RNA-seq of FACS isolated small pre-B cells revealed 869 (FDR < 0.05) differentially expressed genes between κ-mac Cxcr4^fl/fl and κ-mac Cxcr4KO mice ([197]Fig. 5C). Loss of CXCR4 expression was associated with 230 upregulated and 639 repressed genes. Gene annotation revealed CXCR4 regulated genes involved in endocytosis, cell activation, cell chemotaxis, MAPK signaling and NF-κB signaling gene programs ([198]Fig. 5D). As expected, the cell chemotaxis pathway included mediators of cell mobility such as CXCR4, Sipra and cell adhesion molecules Itgam and Itgb2 ([199]Fig. 5E). Strikingly, CXCR4 was required for inducing genes that inhibit cellular activation including Tbc1d10c, Tnfrsf21 and Ptpn22 ([200]54, [201]55, [202]56,Stanford, 2014 #363) ([203]Fig. 5E). CXCR4 also induced positive regulators of PTEN, including Usp7, predicted to attenuate PI3K activation ([204]57). Attenuation of autoreactive BCR signaling is necessary for back differentiation, induction of the RAG proteins and receptor editing. This attenuation has been ascribed to down-modulation of the receptor from the cell surface ([205]20). Indeed, CXCR4 was required for inducing mediators of endocytosis. Our data suggest that CXCR4 coordinates a program that inhibits specific BCR signaling mechanisms. Additionally, CXCR4 was required for other signaling pathways, notably MAPK activation ([206]Figure 5E). Indeed, we have previously demonstrated that CXCR4 is a strong inducer of ERK in developing B cells ([207]27). Furthermore, ERK activation, and downstream induction of E2A, is required for opening both the Igκ and Igλ loci ([208]24, [209]27). Those findings, in addition to our results herein, suggest CXCR4 coordinates multiple signaling pathways required for receptor editing. Next, we assessed how many of the genes differentially regulated in Igκ^del and κ-mac cells were regulated by CXCR4. Thus, we combined genes that were differentially expressed from the pairwise comparisons of small pre-B Igκ^del and κ-mac cells with immature B Igκ^del and κ-mac cells. Genes that were specifically upregulated in κ-mac cells were those ascribed to receptor editing. Overlaps of these gene sets with CXCR4 dependent genes revealed a subset of editing and non-editing genes dependent on CXCR4 expression ([210]Fig. 6SA). CXCR4 dependent editing genes included Cλ1, Cλ2, Jλ2, Cox7a2 and Cox6c and Ptpn22. Pathway analysis revealed editing-associated gene programs including receptor-mediated endocytosis, ROS/OxPhos, and MAPK and chemokine signaling pathways ([211]Fig. 6SB). Conversely, non-editing CXCR4 dependent genes were enriched for processes associated with normal B cell development including chromatin remodeling/organization and protein degradation pathways. Thus, CXCR4 signaling appears to provide specific functions in the context of receptor editing. CXCR4 mediates Igλ recombination in editing cells Next, we investigated whether CXCR4 regulates Igλ recombination. As expected, cell surface staining for IgM, Igκ and Igλ on κ-mac B cell progenitors demonstrated a complete lack of Igκ+ B cells ([212]Fig. 6A). Moreover, Igλ+ B cells were greatly diminished in κ-mac Cxcr4KO mice. We next FACS isolated small pre-B cells from κ-mac Cxcr4^fl/fl and κ-mac Cxcr4KO BM, isolated DNA and performed semi-qPCR for Igλ recombination. We observed a marked decrease in Vλ2-Cλ2 recombination in the absence of CXCR4 suggesting CXCR4 is required for Igλ recombination ([213]Fig. 6B). Figure 6. Igκ recombination is normal in κ-mac CXCR4KO mice. [214]Figure 6. [215]Open in a new tab A, Representative flow cytometry of surface Igκ and Igλ expression from κ-mac CXCR4^fl/fl vs κ-mac CXCR4KO. Cells are gated on B220+CD19+ cells. Quantification of highlighted populations. a) Pink: Igκ+ Igλ− cells. b) Orange: Igκ− Igλ+ cells. Each dot represents a mouse. B, Semi-quantitative PCR analysis of Igλ gene rearrangements from κ-mac CXCR4KO and κ-mac CXCR4^fl/fl small pre-B cells. Five-fold Serial dilutions started from 25ng are shown. C-D, Quantified and integrated transcription across all Igκ (C) and Igλ (D) gene segments from FACS sorted κ-mac CXCR4^fl/fl (n=2) vs κ-mac CXCR4KO (n=3) small pre-B cells. E-F, Complementary DNA was synthesized from RNA obtained from sorted κ-mac CXCR4^fl/fl (n=4) vs κ-mac CXCR4KO (n=4) small pre-B cells and used to perform qPCR for Vλ1-Jλ1 and Vλ1-Jλ3. (E) recombination products and Rag1 and Rag2 (F). G-J, Top panels: Representative flow cytometry of surface IgM against Igκ (G) and Igλ (H) and intracellular IgM against intracellular Igκ (I) and Igλ (J) from κ-mac CXCR4^fl/fl vs κ-mac CXCR4KO. Cells are gated on B220+ cells. Bottom panel: Quantification of highlighted gate in red: I) Surface IgM^+ and Intracellular Igκ^+, II) Surface IgM^+ and Intracellular Igλ^+, III) Intracellular IgM^+ and intracellular Igκ^+, IV) Intracellular IgM^+ and Intracellular Igλ+. Data are presented as means ± standard error of the mean (S.E.M). P values were determined by Student’s T test (* p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001). However, our previous work demonstrated CXCR4 is necessary for efficient Igκ accessibility and recombination ([216]27). Thus, the decrease in Igλ+ B cells could be secondary to lack of preceding Igκ recombination and not a unique requirement of CXCR4 for receptor editing. To test whether Igκ recombination was significantly impaired in κ-mac Cxcr4KO mice, we first examined Igκ and Igλ transcription in κ-mac Cxcr4^fl/fl and κ-mac Cxcr4KO small pre-B cells using RNA-seq. We quantified transcription at Igκ and Igλ loci across all coding gene segments. Surprisingly, in the setting of κ-macroself expression, CXCR4 deficiency did not impair transcription across Igκ genes ([217]Fig. 6C). In fact, transcription across the gene body was slightly elevated in the absence of CXCR4. However, CXCR4 deficiency resulted in a substantial decrease in transcription across Igλ gene elements and at regions flanking the gene bodies ([218]Fig. 6D). This suggests that, in the context of self-reactivity, CXCR4 is necessary for efficient transcription of Igλ. To confirm that Igλ recombination was impaired in the absence of CXCR4, we measured Vλ-Jλ recombination for λ1 and λ3 using quantitative PCR (qPCR) on cDNA from κ-mac Cxcr4KO and κ-mac Cxcr4^fl/fl small pre-B cells. Recombination of both Vλ1-Jλ1 and Vλ1-Jλ3 was severely diminished in κ-mac Cxcr4KO sorted small pre-B cells ([219]Fig. 6E). Moreover, Rag1 and Rag2 expression was also diminished in the absence of CXCR4 ([220]Fig. 6F). Thus, CXCR4 is necessary for Igλ transcription and recombination in editing B cells. An inability to initiate Igκ recombination is a possible explanation for the lack of Igλ+ B cells in κ-mac Cxcr4KO mice. Thus, we measured intracellular Igκ and Igλ expression using flow cytometry. κ-mac Cxcr4KO (blue) and κ-mac Cxcr4^fl/fl (black) cells expressing surface IgM were negative for intracellular Igκ ([221]Fig. 6G). As might be expected, κ-mac Cxcr4KO cells expressed much less intracellular Igλ ([222]Fig. 6H). Interestingly, the presence or absence of CXCR4 did not impact intracellular IgM+Igκ+ cell frequency ([223]Fig. 6I). However, in the absence of CXCR4 intracellular IgM+Igλ+ cells were greatly diminished ([224]Fig. 6J). These data excluded the possibility that lack of Igλ+ B cells in κ-mac Cxcr4KO mice is due to a block in Igκ recombination. From this, we conclude that the proportion of small pre-B cells that complete primary Igκ recombination in the absence of CXCR4 is sufficient to generate a pool of cells for receptor editing. However, subsequent recombination of Igλ in response to autoreactivity is critically dependent upon CXCR4. CXCR4 is required for epigenetic landscape of receptor editing We next investigated whether CXCR4 regulated the chromatin landscape of editing small pre-B cells. In the absence of CXCR4, there were approximately 2,300 regions where chromatin accessibility was altered in κ-mac small pre-B cells ([225]Fig. 7A). Accessibility at ~1,100 of these regions were significantly decreased in κ-mac Cxcr4KO, while ~1,200 regions showed increased accessibility. Interestingly, of the 1,100 CXCR4 dependent peaks, roughly 70% were located at intergenic or intronic regions with the remaining 30% located at promoter or exonic regions ([226]Fig. 7B). In contrast, regions that gained accessibility in κ-mac Cxcr4KO were almost exclusively found at intergenic or intronic regions (90%) with the remaining regions distributed between promoter or exonic regions. Analysis of genes near CXCR4-dependent OCRs identified constituents of several pathways including metabolism, NF-κB and MAPK signaling ([227]Fig. 7C). Genes in the vicinity of chromatin sites repressed by CXCR4 were involved in P53 signaling and oxidative stress. These findings suggest CXCR4 preferentially increases accessibility at genes involved in receptor editing. Figure 7. CXCR4 dictates epigenetic landscape of editing B cells. [228]Figure 7. [229]Open in a new tab A, Heatmap depicting differentially accessible regions (DARs) as determined by ATAC-Seq of FACS sorted κ-mac CXCR4^fl/fl (n=3) vs κ-mac CXCR4KO (n=4) small pre-B cells. The row z-score of Log[2] transformed normalized counts were used to construct the heatmap. Biological replicates are indicated. B, Frequency of DARs depicted in A at annotated genomic regions: Promoter (blue), Exon (red), Intron (green) and Intergenic (purple). C, Gene ontology analysis of genes near DAR depicted in A. ‘Wikipathway’ gene sets were used for this analysis. Scale bar represents Log[10] transformed P values. D-E, DARs were analyzed for transcription factor motif enrichment using HOMER’s ‘findmotifsgenome.pl’. De Novo motifs are depicted with significance determined using hypergeometric distribution. F-G, Quantified and integrated accessibility determined by ATAC-seq immediately before and after the length of each Igκ (F) or Igλ (G) gene body from κ-mac CXCR4^fl/fl (n=3) vs κ-mac CXCR4KO (n=4) small pre-B cells. H, Histogram depicting ATAC-Seq defined accessibility at Vλ, Jλ and Cλ genes separately. Data are presented as means ± standard error of the mean (S.E.M). P values were determined by Student’s T test (* p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001). In sites both opened and closed by CXCR4 there was enrichment for transcription factor binding sites including those for SPIB, PU.1 and E-box proteins ([230]Fig. 7D-[231]E). However, CXCR4 specifically opened regions containing FOXO1 binding motifs, a transcription factor which induces Rag1/2 transcription ([232]42). Thus, CXCR4 modulates accessibility at binding motifs for transcription factors critical for receptor editing. Next, we assessed accessibility at the light chain loci. In agreement with transcriptional analysis at Igκ, accessibility was nearly identical in the presence or absence of CXCR4 ([233]Fig. 7F). Chromatin accessibility was highest at the flanking 5’ and 3’ regions of Igκ genes indicating the locus is open and capable of undergoing recombination. However, unlike the transcriptional analysis, Igλ accessibility was only marginally affected in the absence of CXCR4 ([234]Fig. 7G). Assessing accessibility separately at Vλ, Jλ and Cλ segments revealed that loss of accessibility was most significantly observed at Cλ genes in the absence of CXCR4 ([235]Fig. 7H). Thus, CXCR4 signaling promotes Igλ chromatin accessibility. CXCR4 signaling directly induces Igλ recombination It was possible that the observed functions of CXCR4 were a consequence of BM positioning rather than a direct effect of CXCR4 signaling on Igλ recombination. To discern between these possibilities, we used an established stromal cell-free system of early B cell differentiation ([236]27). MACS column purified B220^+IgM^− progenitors from κ-mac mice were grown in the presence of IL-7 (16ng/ml) for five days, followed by three days in culture with IL-7 (16ng/ml, +IL7) or low IL-7 (0.2ng/ml, referred hereafter as −IL-7) with or without CXCL12 (100ng/ml, +CXCL12). Cells were harvested and subjected to bulk RNA-Seq. Overall, in κ-mac small pre-B cells withdrawing IL-7 and adding CXCL12 resulted in 2,101 differentially expressed genes (q < .05) ([237]Fig. 8A). K-means clustering identified four clusters with distinct gene expression patterns. CXCL12 amplified gene transcription changes induced by IL-7 withdrawal (clusters 1 and 4), induced new transcriptional programs (cluster 2) and even repressed programs induced by IL-7 withdrawal (cluster 3). Genes found in cluster 1 were enriched for cell cycle, DNA replication and homologous recombination. Examples of genes in cluster 1 include Myc, Ccnd2, Ccnd3 and genes involved in homologous recombination include Aurka, Aurkb, Brca1 and Brca1 ([238]Fig. 8B, [239]D). Cluster 2 contained genes such as Prdx1 and Xrcc6 and was enriched for negative regulation of cell transduction, detoxification, endosomal trafficking ([240]Fig. 8E-[241]G). Genes in cluster 3 included Itga2 and Mmp12 and were enriched for immune response, cell adhesion and PI3K-Akt signaling ([242]Fig. 8H-[243]J). Lastly, in cluster 4 were genes such as Cybb, G6pdx, Gpx3 involved in lysosome function, cell migration, ROS metabolism and receptor mediated endocytosis ([244]Fig. 8K-[245]M). Other genes induced in cluster 4 included Rag1, and Traf6 ([246]Fig. 8N). Consistent with our previous experiments using WT small pre-B cells ([247]27), CXCL12 did not substantially modulate the expression of important developmental transcription factors including Tcf3 (E2A), Pax5, Foxo1, Irf4 or Irf8 ([248]Fig. 8O). These data indicated that CXCL12-CXCR4 signaling axis directly modulates multiple transcriptional programs in editing small pre-B cells. Figure 8. CXCR4 signaling directs Igλ recombination. [249]Figure 8. [250]Open in a new tab A, RNA-seq heatmap depicting differential gene expression (q < 0.01) of κ-mac pre-B cells (replicates shown) cultured for 72 hrs with 16 ng ml–1 of IL-7 (+IL-7), 0.2 ng ml–1 of IL-7 (−IL-7) and 0.2 ng ml–1 of IL-7 with 100 ng ml–1 of CXCL12 (−IL-7+CXCL12). CPM, counts per million. Clusters were determined using consensus k-means clustering with 1000 iterations. Numbers of genes in each cluster are indicated. Scale bar indicated z score of Log[2] transformed counts per million: z score log2(CPM). B,E,H,K, Boxplot indicating average z-score for clusters identified in A. Boxes represent interquartile ranges (IQRs) and black horizontal lines represent median values. Maximum and minimum values (error bars) are defined as Q3 + 1.5× the IQR and Q1 − 1.5× the IQR, respectively. C,F,I,L, Corresponding Metascape gene ontology analysis for the indicated clusters identified in A and B. D,G,J,M,N, Histograms of example genes belonging to the indicated clusters. O, Histograms of example genes from bulk RNA-seq from experiment shown in A. P, Quantified unique RNA-seq reads that mapped to Vλ-Jλ-Cλ gene bodies for Igλ light chains: λ1, λ2, λ3, λx. Data are presented as means ± standard error of the mean (S.E.M). P values were determined by One-way ANOVA with Tukey multiple comparisons (* p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001). We next assessed the impact of CXCR4 signaling on Igλ recombination in vitro. We utilized our RNA-seq data from the cultured κ-mac pre-B cells and quantified the number of unique single reads that mapped to specific Vλ, Jλ, and Cλ gene segments. As expected, the presence of IL-7 (+IL7) repressed Igλ recombination ([251]Fig. 8P). The withdrawal of IL-7 alone (−IL-7) resulted in a substantial increase in Igλ recombination but addition of CXCL12 (−IL-7 +CXCL12) resulted in the highest amount of Igλ recombination in κ-mac cells ([252]Fig. 8P). Overall, our data demonstrated a direct requirement for CXCR4 signaling in orchestrating receptor editing and Igλ recombination. Igλ enhancers regulate recombination within chromatin domains The Igλ locus spans 0.2 megabases and is divided into two recombination modules ([253]Fig. 7SA). One module contains Vλ1 in close proximity to Jλ1, Jλ3, Cλ1, Cλ3 and a nearby enhancer element Enhancer 1-3 (Eλ1-3), while in the other module Vλ2 and Vλ3 are in close proximity to Jλ2, Jλ4, Cλ2, Cλ4 and Enhancer 2-4 (Eλ2-4; [254]Fig. 7SA). Transcriptional induction is associated with E2A binding to both enhancers ([255]24). However, it is not known if these enhancers regulate recombination. It is also not known if they provide redundant roles or if each is required for specific recombination events. CTCF bound at convergent motifs most commonly define topological associating domain (TAD) boundaries ([256]58-[257]60). Indeed, ChIP-seq of CTCF and the Cohesin subunit RAD21 ([258]61, [259]62), in sorted WT pro-B cells showed binding across the Igλ locus ([260]Fig. 7SB). The placement and orientation of CTCF bound sites predicts that the Igλ locus can be divided into two TADs, each with its own enhancer ([261]Fig. 7SC). To confirm this prediction, we performed in situ Hi-C Seq on sorted WT small pre-B and immature B cells ([262]60, [263]63). In situ Hi-C contact maps of the Igλ locus showed increased chromosomal interactions in immature B cells compared to small pre-B ([264]Fig. 7SD). As a control, WT double positive T cells showed no apparent chromosomal interactions at the Igλ locus. We assessed the significance of the chromosomal contacts strengthened in WT immature B cells compared to small pre-B cell and observed many statistically significant (P<0.05) DNA contacts clustered around E2-4 in immature-B cells ([265]Fig. 7SE-[266]F). Interestingly, these loops each contained an enhancer (Eλ1-3 or Eλ2-4), Jλ and Cλ genes, and either one or two Vλ genes. Given the high sequence homology of the enhancers and their location in different TADs this raised the possibility that these enhancers orchestrate Vλ-Jλ recombination locally within their respective TADs. To test this hypothesis, we used CRISPR to delete Eλ2-4 from WT C57BL/6J mice using the indicated guide RNA pair ([267]64) ([268]Fig. 7SA). Flow cytometric analysis of BM populations revealed a slight reduction of immature B cells in Eλ2-4 CRISPRKO mice compared to control mice derived from embryos inject with Eλ2-4 gRNAs but which did not delete Eλ2-4. Staining for Igκ and Igλ revealed an increase in the frequency Igκ+ B cells in Eλ2-4 CRISPRKO and a slight decrease in Igλ+ (Igλ1, 2 and 3) B cells ([269]Fig. 7SG-[270]H). Interestingly, we observed a complete loss of the Igκ− Igλ− cells, which represents the Igλx product that arises from rearrangement of Vλx-Jλ2-Cλ2. These results suggested that the Igλ locus is modular with each enhancer controlling recombination within its respective TAD. Discussion Downmodulation of self-reactive BCRs is associated with ‘back differentiation’ and induction of Igλ recombination ([271]20). Here, we demonstrated that this back differentiation was selective, with κ-mac small pre-B cells developing an epigenetic landscape that is a blending of that observed in WT small pre-B and immature B cells. Indeed, most of the unique OCRs in κ-mac small pre-B cells were those characteristic of immature B cells. These immature B cell-mapped OCRs were enriched for binding sites for E2A, SPIB, NF-κB, CTCF and FOXO many of which are necessary for B cell receptor editing ([272]23, [273]24, [274]37, [275]42, [276]65). Likewise, Igλ preferentially opened in κ-mac small pre-B cells. Interestingly, some binding sites, such as those for E2A, were primarily enriched in small pre-B cells and this enrichment did not carry forward into the immature B cell compartment. For others, such as those for NF-kB, they became much more accessible in immature B cells. These findings indicate editing cells transit asymmetrical and non-equivalent forward and reverse developmental trajectories that dictate epigenetic landscape and subsequent function ([277]Fig. 8SA). Large changes in small pre-B cell accessibility were associated with rather modest concurrent differences in transcription ([278]Fig. 8SB). These findings indicate epigenetic regulation initiates the mechanisms of B cell receptor editing. Indeed, the epigenetic landscape of κ-mac small pre-B cells predicted subsequent large differences in the transcriptional state of immature B cells. Genes in the vicinity of κ-mac small pre-B cell OCRs were enriched for multiple pathways involved in receptor editing including chemokine receptors and NF-κB signaling ([279]23, [280]26). These pathways then became active in immature B cells. These data demonstrate how the complex developmental trajectory of autoreactive B cells, and the resulting imprinting of this trajectory on the epigenetic landscape, primes for a subsequent different developmental path than non-autoreactive B cells. In addition to those predicted by the small pre-B cell epigenome, other transcriptional programs were specifically upregulated in κ-mac immature B cells critical for receptor editing including effectors of gene recombination such as Rag1/2 and DNA repair. There were also striking changes in metabolic pathways including strong induction of oxidative phosphorylation and ROS. High levels of ROS, a consequence of oxidative phosphorylation, is associated with DNA damage and could therefore complicate receptor editing ([281]66, [282]67). However, there was also an attendant upregulation of antioxidant genes in κ-mac immature B cells which likely mitigate genomic risk. In the periphery, BCR stimulation induces the oxidative phosphorylation pathway ([283]68). Therefore, it is likely that this metabolic state is, in part, a direct consequence of BCR autoreactivity. These data suggest that both BCR signaling, and subsequent loss of BCR surface expression, contribute to the functional state of editing B cells. BCR editing was associated with modulation of endocytic effectors and a reduction in trafficking of endocytosed BCRs to the MIIC. The MIIC is a central hub for secondary signals including those activated by endosomal toll-like receptors ([284]69, [285]70). Such signals overlap with those downstream of the BCR and could therefore subvert BCR-driven cell fate decisions. Indeed, regulation of BCR endocytic trafficking prevents antigen complexes containing TLR ligands from spuriously activating anergic B cells ([286]71). In the case of receptor editing, endosomal signaling might prevent the back-differentiation associated with BCR downmodulation. We propose that control of BCR endocytic fate ensures fidelity of the receptor editing molecular program. Autoreactive B cells in the BM increase surface expression of CXCR4 ([287]20, [288]26, [289]72, [290]73). This was thought to facilitate tolerance by retaining autoreactive B cells in the BM. Herein, we demonstrated that CXCR4 signaling played a direct role in inducing Igλ recombination in editing B cells. Both in vivo and in vitro studies indicated that CXCR4 makes specific contributions to the molecular programs of receptor editing including induction of MAPK signaling, repressing proliferation, and inducing oxidative phosphorylation. In vitro studies revealed that CXCR4 both amplified pathways initiated by IL-7 escape, such as cell cycle repression and PI3K signaling, and induced unique programs, including those associated with the endosomal pathway. Finally, CXCR4 induced repressors of BCR signaling, including Ptpn22, predicted to facilitate back differentiation and anti-apoptotic mediators (Bcl2, Bcl2a1d), predicted to prolong the developmental window. These data suggest that the interplay between CXCR4 and BCR signaling, and BCR internalization, coordinate B cell antigen receptor editing. Also notable was the almost complete dependence of in vivo Igλ recombination on CXCR4. In contrast, Igκ recombination was surprisingly preserved. CXCR4 feeds forward to amplify pre-BCR initiated signals and drive efficient Igκ recombination ([291]27). However, CXCR4 was not absolutely required and, in the context of self-reactivity, Igκ recombination appeared sufficient for progression to Igλ. CXCR4 was not required to open the Igλ locus. Rather, it was required for Igλ gene segment transcription with this effect being most pronounced at flanking sequences containing RSSs. Transcription through the T cell receptor-α locus is required for Vα-to-Jα recombination ([292]74). Our data are consistent with this and suggest that locus transcription is more important than accessibility for Igλ recombination. Transcription of the Igλ locus is dependent upon two nearly homologous enhancers, located in separate TADs associated with divergent Vλ and Jλ segments ([293]61, [294]62). This modular genomic organization could allow specific signaling contexts to favor particular Igλ gene products. For example, it might have been expected that autoreactivity would preferentially induce Igλx, the light chain most associated with receptor editing. However, either pathway to Igλ recombination, failure at Igκ or autoreactivity, leads to a similar distribution of expressed Igλ chains. This likely reflects the high sequence homology between the Igλ enhancers. That autoreactivity does not skew Igλ repertoire suggests that either selection for repertoire diversity drives Igλ gene evolution or that other Igλ chains, besides Igλx, edit autoimmunity in different contexts. Previous studies on receptor editing have relied on fixed BCR specificity toward model antigens. While the κ-mac mice possess a polyclonal B cell repertoire, we recognize that a caveat of our study is the use of the κ-macroself transgene to induce editing. In a WT unmanipulated mice, the nature of the autoreactive antigen may further modulate transcription factor networks and chromatin accessibility. Ultimately, a complete understanding of receptor editing would require studying editing B cells in WT mice. A significant hurdle for completing these studies has been distinguishing cells undergoing receptor editing from those that are not. In the absence of a cell surface marker for editing cells, single cell analysis in combination with the molecular features of receptor editing outlined in this paper may enable identification of editing cells in WT BM. Our data provide a model in which asymmetric “forward” and “reverse” developmental trajectories, developmental plasticity, sets a unique epigenetic landscape permissive for BCR editing. This then becomes the substrate for both BCR and CXCR4 signals, which activate and amplify the molecular pathways of receptor editing. Therefore, plasticity is not just a feature of lymphocyte development ([295]75, [296]76). Rather, it is critical for enabling specific cell fate decisions. Materials and Methods Study design This study aimed to understand the molecular mechanisms of receptor editing and Igλ recombination. We utilized two mouse models. The first contained a deletion at the J kappa (Jκ) genes and, as such, models Igλ expression resulting from failed Igκ recombination (Igκdel). The second models autoreactivity by expressing a single-chain chimeric anti-Igκ antibody (κ-mac). We then performed RNA-seq and ATAC-seq on bone marrow B cells to investigate transcriptional networks and chromatin accessibility, respectively associated with these two pathways to Igλ recombination. We also derived κ-macxCxcr4^fl/flxMb1-Cre^+ mice to determine the role CXCR4 in receptor editing and Igλ recombination. Experimenters were not blinded. Mice were randomly assigned to experimental groups and analyzed without excluding outliers. The number of independent experiments is indicated in the figure legend. Mice κ-macroself (stock #006259) were cryo-recovered from Jackson laboratories. All κ-macroself mice used were hemizygous for the κ-macroself gene. Note: breeding should be performed with hemizygous κ-macroself female mice paired with WT CD45.1 male mice. Igκ^del mice were obtained from David Nemazee. Wildtype CD45.1 C57BL/6J (stock #002014) mice were purchased from Jackson laboratories and housed in the University of Chicago animal facilities. Cxcr4^fl/fl mb1-cre+ mice were obtained from Malay Mandal ([297]27). Female and male mice, randomly distributed between control and experimental groups, were used at 7–12 weeks of age, and studies carried out in accordance with the guidelines of the Institutional Animal Care and Use Committee at the University of Chicago (Protocol No. 71577). Five-week-old B6 CD45.1 (#002014) and B6 CD45.2 (#000664) mice were purchased from Jackson laboratories for the RNA-seq experiment comparing strain specific gene expression. Isolation and flow cytometry of BM B cell progenitors. BM was collected from WT, Igκ^del, κ-mac, κ-mac Cxcr4^fl/fl mb1-cre− or κ-mac Cxcr4^fl/fl mb1-cre+ mice and cells were resuspended in staining buffer: (3% (v/v) fetal bovine serum (FBS) in 1x phosphate buffered saline. Erythrocytes were lysed with ACK lysis buffer (Lonza cat # 10-548E) and cells were stained with rat anti-CXCR4 (2B11), rat anti-CD43 (S7), rat anti-IgM (R6–60.2), rat anti-IgD ([298]11-[299]36), rat anti-CD19 (1D3) and rat anti-B220 (RA3-6B2l), (all from BD Biosciences) as described previously ([300]77, [301]78) and viability dye eFluor 506 (eBioscience). Pre-pro-B cells (CD19−B220+IgM−), pro-B cells (CD19+B220+CD43+IgM−), large pre-B cells (B220+CD43−IgM−FSChi), small pre-B cells (B220+CD43− IgM−FSC^low) and immature B cells (B220+CD43−IgM+) were isolated by cell sorting with a FACSAria II (BD Biosciences). In Vitro Culture B cell progenitors (B220+IgM−) from WT, Igκ^del , κ-mac mice were isolated from BM with a MACS separation column (Miltenyi Biotec) and cultured in complete Opti-MEM containing 10% (v/v) FBS and IL-7 (16 ng ml−1) for 5 d. Further culture for 72 hours with 16 ng ml−1 IL-7 (+IL-7), 0.2 ng ml−1 IL-7 (−IL-7) or 0.2 ng ml−1 IL-7 with 100 ng CXCL12 (−IL-7+CXCL12) without any stromal cells was performed before analysis. These in vitro culture conditions were adapted from ([302]27). Hi-C preparation and Analysis Flow sorted WT (CD45.2) small pre-B and WT (CD45.2) immature B cells (5×10^6 for each stage) were crosslinked with 1% formaldehyde following wash with ice cold PBS. Membranes were lysed keeping the nuclei intact following by restriction digestion with 100U of MboI (NEB R0147). The DNA ends were then marked with biotin, ligated proximally and crosslinking reversed. DNA shearing and Size selection were then performed for fragments 300-500bp. DNA amount was quantified by Qubit dsDNA High Sensitivity Assay (Life technologies, [303]Q32854) and biotin pull down was done to prepare final in-situ Hi-C library that was quantified and sequenced using the Illumina HiSeq 4000. For further details on sequencing, see [304]supplementary materials and methods. qPCR QPCR reactions were analyzed in triplicate in a 25 μl reaction containing 10 μM of each primer. Cycling conditions for all qPCRs were: 50°C for 2 min, 95°C for 3 min, followed by 45 cycles of 95°C for 15 s and 61°C for 1 min. C[T] values were determined using the Applied Biosystems 7300 Real-Time PCR System and the provided application-specific software. Data were exported and analyzed with Microsoft Excel. Data were analyzed according to the ΔΔC[T] method. The C[T] mean for each sample was calculated and standard deviations (s) were calculated for each mean C[T] value. C[T] means were first defined as the difference between sample and corresponding input control and the amount of target gene was normalized to the negative control Hprt1. Differential expression (RNA-seq and ATAC–seq) Differential expression statistics (fold-change and P value) were computed using edgeR, on raw expression counts obtained from quantification (either genes or ATAC open chromatin regions) ([305]79, [306]80). Group comparisons were made using the generalized linear modeling capability in edgeR. In all cases, P values were adjusted for multiple testing using the FDR correction of Benjamini and Hochberg. Significant genes were determined based on an FDR threshold of 5% (0.05) in the group comparison and greater than a fold change. For differential expression between κ-mac Cxcr4^fl/fl mb1-cre− and κ-mac Cxcr4^fl/fl mb1-cre+ small pre-B cells, batch correction was performed using the Remove Unwanted Variation from RNA-Seq Data (RUV-Seq) package from bioconductor in R. Specifically the RUVs with RUV-Seq program was used. Principal component analysis (PCA) was performed using the R function prcomp with the parameter “scale=TRUE”. For further details on RNA analysis, see [307]supplemental methods. Pathway analysis and motif analysis (ATAC-seq) ATAC-seq analysis for WT (CD45.1), Igκ^del and κ-mac small pre-B cells was performed using HOMER (Hypergeometric Optimization of Motif EnRichment)([308]81). Motif enrichment analysis was performed using ‘findmotifsgenome.pl’ command with -size 300 and other default parameters. On occasion when the observed chromatin sites exceed background chromatin sites statistical significance was determined using a hypergeometric distribution by setting the -h flag. Open chromatin regions were assigned to their nearest using the ‘annotatepeaks.pl’ command. Gene pathway enrichment of genes near open chromatin regions was performed by setting the -go flag. The results from the ‘wikipathway’ were explored and data was replotted using ‘ComplexHeatmap’ in R. For [309]Figure 3 G-[310]I, histograms of reads around transcription factor binding motifs were generated by centering open chromatin regions containing the investigated motifs using the ‘annotatepeaks.pl’ command, followed by counting reads from individual experiments at single base pair resolution in a radius of 1500 bp (or 150 bp) around the peak centers using the ‘annotatepeaks.pl’ function with flags ‘-hist -fragLength 1.’ Overlaps of open chromatin regions from ATAC-seq data was performed using the HOMER “mergePeaks” command. For further details on ATAC-seq analysis, see [311]supplemental methods. CRISPR Mice Guide sites were identified using Integrated DNA Technologies’ (IDT) selection tool ([312]https://www.idtdna.com/site/order/designtool/index/CRISPR_CUSTOM) by inputting ~500bp surrounding the region of interest. For the 2-4 enhancer deletion, the protospacer sequences were GCTTGGAGCATTACCTGAAT and GGGTGAGTATCAATCTGTCC. Three hours prior to microinjection, tracrRNA (IDT #1072532) and crRNAs (IDT) for the protospacers were annealed and complexed with Cas9 protein (IDT #1081058). For further details on generating CRISPR mice, see [313]supplemental methods. Detection of mitochondrial reactive oxygen species (MitoSOX™) See [314]supplemental methods for further details on conditions used for MitoSOX™ experiments. Endocytic trafficking and imaging See [315]supplemental methods for further details on conditions used for endocytic trafficking and image microscopy. Primers and PCR conditions See [316]supplemental methods for list of primers and PCR conditions. Gene Pathway analysis (RNA-seq) Metascape web portal ([317]www.metascape.org) was used to pathway analysis of differential expressed genes found from analyzing RNA-seq data ([318]43). Visualization of ATAC-seq Visualization of ATAC-seq data was done using Integrated Genome Browser ([319]82). GSEA Gene Set Enrichment Analysis (GSEA) was performed on normalized Log[2] [counts per million] values from RNA-seq data. ‘Ratio of Classes’ metric was used to determine statistical significance. Gene sets found within the HALLMARK and C5 GO molecular signatures pathways were used. GSEA program can downloaded here: [320]http://software.broadinstitute. Results from GSEA analysis were replotted using the custom R function ReplotGSEA.R ([321]https://github.com/PeeperLab/Rtoolbox/blob/master/R/ReplotGSEA.R) . Clustering and heatmaps Clustering for heatmaps was performed using complete linkage hierarchical clustering of z-scored normalized values from RNA-seq or ATAC-seq and plotted as a heatmap using the ‘hclust’ and ‘heatmap.3’ functions in or the ‘ComplexHeatmap’ from Bioconductor in R ([322]83, [323]84). Computer software versions R_3.5.3, edgeR_3.28.1, circlize_0.4.10, ComplexHeatmap_2.2.0, chromVAR_1.4.1, RUVSeq_1.20.0, Homer_v.4.10.3, bedtools_v2.27.1-9-g5f83cacb, macs2_2.2.6, GSEA_4.0.3., IGB_9.1.4, ggplot2_3.3.2, ggrepel_0.8.2, dplyr_1.0.1, magrittr_1.5, Star 2.7.0. Statistical analysis Statistical analyses were performed with GraphPad Prism. For multiple comparisons, data were analyzed by analysis of variance in combination with Tukey’s multiple comparisons test. Bar graphs are displayed as the mean ± S.E.M. Significance as defined by P value or FDR are provided in the Figures, figure legends, corresponding text or in tables. Additional quantitative methods and statistical criteria are mentioned above based on their respective technology and analysis approaches. Supplementary Material Data file S1 [324]NIHMS1840951-supplement-Data_file_S1.xlsx^ (44.9MB, xlsx) Data file S2 [325]NIHMS1840951-supplement-Data_file_S2.docx^ (45.7KB, docx) main supplementary [326]NIHMS1840951-supplement-main_supplementary.docx^ (6.2MB, docx) Acknowledgements