CTCF-mediated cis-regulatory chromatin insulation enforces a central B-cell tolerance checkpoint

27 The generation of a diverse and self-tolerant B cell repertoire is essential for 28 adaptive immunity and is achieved through V(D)J recombination. In mice, Igκ is the 29 dominant light chain, whereas Igλ rearrangement typically occurs in response to 30 nonproductive or autoreactive Igκ recombination, a process termed receptor editing. 31 Recombination at the RS element deletes the Igκ constant exon, silencing the locus 32 and enabling Igλ expression. However, the epigenetic regulatory framework that 33 orchestrates and governs receptor editing remains poorly defined. Here, we identify 34 a CTCF-binding insulator element (CBE) within the 3′ Igκ super-enhancer (3′-SEκ) 35 that regulates receptor editing and directs the κ-to-λ switch required for Igλ⁺ B-cell 36 development. Mechanistically, loss of this CBE activates an insulated enhancer 37 within the 3′-SEκ, causing aberrant Vκ rearrangements and altered chromatin 38 interactions through disrupted loop extrusion dynamics. Notably, loss of this CBE in 39 mice leads to increased autoantibody production by ten weeks of age, demonstrating 40 that CBE-mediated chromatin architecture shapes B cell fate by constraining 41 autoreactive potential. Collectively, our findings define a novel CTCF-dependent cis-42 regulatory insulation checkpoint that connects chromatin loop extrusion to antigen-43 driven receptor editing, thereby enforcing B-cell tolerance. 44 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 3

45 The generation of a diverse and self-tolerant B cell repertoire is fundamental to 46 adaptive immunity and relies on the precise orchestration of immunoglobulin (Ig) 47 gene rearrangements1-4. Antigen receptor diversity is primarily achieved through 48 V(D)J recombination, wherein variable (V), diversity (D), and joining (J) gene 49 segments are stochastically rearranged to encode functional Ig molecules5,6. In 50 developing B lymphocytes, rearrangement of the immunoglobulin kappa (Igκ) locus 51 follows successful assembly of the heavy chain and is subject to stringent spatial 52 and temporal control1,7. The murine Igκ locus encompasses over 100 Vκ gene 53 segments alongside four functional Jκ segments, undergoing somatic recombination 54 at the pre-B cell stage7-12. This process is tightly regulated by a constellation of cis-55 regulatory elements that coordinate chromatin accessibility, spatial genome 56 architecture, and recombination potential13-22. 57 58 Higher-order genome architecture plays an integral role in orchestrating antigen 59 receptor gene assembly7,23. Topologically associating domains (TADs) are 60 evolutionarily conserved chromatin domains defined by frequent intra-domain 61 interactions and insulation from adjacent regions24,25. Within TADs, finer-scale 62 subdomains known as subTADs often exhibit cell–type–specific interactions that 63 coordinate local gene regulation24,25. Both TADs and subTADs are shaped through 64 cohesin-driven loop extrusion and are typically delimited by convergently oriented 65 CTCF Binding Elements (CBEs)26-34. The Igκ locus is governed by a multilayered 66 regulatory architecture comprising key enhancer elements, including the intronic 67 enhancer (iEκ)13,18,19,35,36, the 3′ enhancer (3′-Eκ)13-16,19, and the distal enhancer 68 (dEκ)14,15,19. Together with co-enhancers such as HS1014 and Dm17, these elements 69 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 4 form the 3′-Igκ super-enhancer (3′-SEκ)37, which regulates chromatin activation and 70 shapes the Vκ gene repertoire. The Igκ locus is organized into CTCF-anchored 71 chromatin loops that define spatial compartments, including subTADs within the Vκ 72 gene array19,38,39. Structural boundaries between the Vκ and Jκ domains are 73 reinforced by two prominent CBEs, the Contraction Element for Recombination 74 (CER) and Silencer in the Intervening Sequences (SIS), which insulate the Vκ region 75 from the Jκ recombination centre (RC)40,41, thereby limiting proximal Vκ usage and 76 promoting distal rearrangements through long-range chromatin interactions11,22,42-44. 77 The CER element enables locus contraction and diffusion-based accessibility 78 required for antigen receptor assembly in murine pre-B cells40,42. In the absence of 79 CTCF protein, the Igκ light chain locus similarly exhibits elevated proximal and 80 diminished distal Vκ usage20. As B cells progress from early progenitor to precursor 81 B cell stages, the Igκ locus undergoes extensive chromatin remodelling 82 characterized by dynamic shifts in epigenetic marks, transcriptional activation, and 83 the emergence of de novo CTCF-mediated chromatin loops that prime the locus for 84 recombination41,45-48. 85 86 Receptor editing is a vital mechanism of B cell tolerance, allowing autoreactive or 87 nonfunctional B cells to revise their antigen receptor specificity through secondary 88 Vκ–Jκ recombination events or by initiating Igλ light chain rearrangement49, which is 89 enabled by the RS element located downstream of the Cκ exon. This process 90 promotes secondary recombination by excising previous Igκ rearrangements49,50. 91 Although significant progress has been made in identifying the enhancers and 92 boundary elements that regulate Igκ recombination, the epigenetic mechanisms 93 connecting chromatin architecture to receptor editing and light chain isotype 94 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 5 selection remain unknown. In this study, we explore the newly identified function of a 95 previously uncharacterized CBE within the 3′-SEκ, which influences immunoglobulin 96 light chain recombination, isotype balance, receptor editing, and immune tolerance 97 maintenance. 98 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 6

99 CBE is essential for maintaining light chain isotype balance and enabling 100 bidirectional transcriptional insulation 101 To address whether CBEs located downstream to CER and SIS elements within the 102 3′-SEκ (Fig. 1a) modulate chromatin architecture or participate directly in Vκ–Jκ 103 recombination, we initially conducted CTCF ChIP–sequencing analysis in IL-7–104 expanded wild-type murine pre-B cells. We identified two prominent CBEs within the 105 3’-SEκ region. One element is positioned immediately downstream of the dEκ 106 enhancer (hereafter termed rsCBE), while the other resides within the eighth intron 107 of the ubiquitously expressed Rpia gene (hereafter termed rpiaCBE) (Fig. 1a). Both 108 elements display conserved CTCF occupancy, are embedded within a topologically 109 dynamic region of the Igκ locus, and feature binding motifs oriented toward the Vκ 110 domain. Importantly, these elements exhibit conserved CTCF-binding patterns in 111 humans, as evidenced by CTCF-ChIP analysis (Fig. 1b). To dissect their functional 112 roles in Igκ recombination, we employed CRISPR–Cas9-mediated genome editing to 113 generate mice with germline deletion of 571 bp at rsCBE (ΔrsCBE) and 528 bp at 114 rpiaCBE (ΔrpiaCBE). These mouse models allowed us to investigate how the 115 disruption of CTCF-mediated insulation within the 3′-SEκ region influences 116 epigenetic remodelling and recombination dynamics in B cells. First, we assessed 117 whether rsCBE and rpiaCBE modulate the composition of the Igκ repertoire. We 118 performed high-throughput sequencing of RNA from bone marrow-derived pre-B 119 cells isolated from ΔrsCBE and ΔrpiaCBE mice. Comparative repertoire analysis 120 revealed that Vκ gene usage across both mutant lines was essentially 121 indistinguishable from that of wild-type (WT) controls, suggesting that loss of these 122 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 7 CTCF-binding elements does not perturb Vκ selection during early B cell 123 development (Extended Data Fig. 1a,b). 124 125 To assess the impact of CBEs on Vκ–Jκ rearrangement, pre-B cells were isolated 126 from the bone marrow and subjected to quantitative PCR analysis using both 127 genomic DNA and RNA. Deletion of rsCBE led to a pronounced increase in Vκ 128 rearrangement, exceeding 1.8-fold relative to WT controls at the RNA levels (Fig. 129 1c). In contrast, loss of rpiaCBE resulted in a small yet significant reduction in 130 rearrangement efficiency (Fig. 1c). These findings were also corroborated by 131 increased sterile germline transcripts (GLT) (non-coding transcripts originating from 132 promoters located upstream of Jκ51,52) in rsCBE mice (Fig. 1c), and genomic DNA 133 rearrangement closely mirrored RNA expression patterns (Extended Data Fig. 1c). 134 These results collectively indicate that rsCBE, positioned within the 3′-SEκ region, 135 acts as an Igκ chromatin insulator, regulating recombination dynamics. 136 137 To assess whether loss of insulation impacts the transcriptional regulation of 138 neighbouring genes, we examined expression of the housekeeping gene Rpia14, 139 situated approximately 2 kb downstream of the HS10 element (Fig. 1a). Strikingly, 140 disruption of individual CBEs, rsCBE or rpiaCBE, led to a measurable loss of 141 insulation accompanied by an upregulation of Rpia transcripts by ~1.8-fold and ~2.8-142 fold, respectively, in pre-B cells (Fig. 1d). Rpia encodes ribose-5-phosphate 143 isomerase A, a pivotal enzyme in the pentose phosphate pathway, and its aberrant 144 expression has been implicated in autophagy53,54 and tumorigenesis55,56. These data 145 reveal a previously unrecognized role for CBEs within the 3′-SEκ region in 146 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 8 safeguarding local transcriptional fidelity, underscoring their importance in preserving 147 the regulatory integrity of the Igκ locus. 148 149 Given that approximately 15% of pro-B cells are estimated to initiate Vκ 150 rearrangement57, we hypothesized that deletion of rsCBE may induce premature 151 activation of Vκ–Jκ recombination during the pro-B cell stage. In agreement, pro-B 152 cells from ΔrsCBE mice showed a significant over twofold increase in Vκ–Jκ 153 rearrangement compared to WT (Extended Fig. 1d). Notably, rsCBE deletion also 154 resulted in over a twofold upregulation of Rpia expression in pro-B cells (Extended 155 Data Fig. 1d). Together, these findings underscore the role of CBEs within the 3′-SEκ 156 region of the Igκ locus as bidirectional transcriptional insulators, functioning to 157 regulate gene expression and recombination events not only in pre-B cells but also 158 at the pro-B cell stage. 159 160 Given the altered Igκ expression after CBE deletion, we hypothesized that this may 161 influence the κ⁺/λ⁺ B cell ratio. To investigate this, we first examined the immature B 162 cell compartment in the bone marrow of the ΔrsCBE mouse to ascertain our 163 hypothesis. Interestingly, flow cytometric analysis demonstrated a significant 164 elevation in the proportion of IgM⁺κ⁺ B cells, concomitant with a marked reduction 165 exceeding twofold in the frequency of IgM⁺λ⁺ B cells within the bone marrow of 166 ΔrsCBE mice compared to the control (Extended Data Fig. 1e-g). These findings 167 indicate that rsCBE regulates light chain isotype distribution during early B-cell 168 development, with its deletion skewing the balance toward κ⁺ B cells at the expense 169 of λ⁺ populations in the bone marrow. 170 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 9 We next analyzed splenic B220⁺ cells from the mutant mouse lines to evaluate the 171 role of CBEs in peripheral B-cell populations. Flow cytometry showed that deleting 172 rsCBE caused a significant expansion of the κ⁺ B cell subset, along with more than a 173 twofold reduction in the λ⁺ population (Fig. 1e, Extended Data Fig. 1h,i). As a result, 174 the ratio of splenic Igκ⁺ to Igλ⁺ B cells was significantly increased in rsCBE-deficient 175 mice (Fig. 1f). Importantly, the frequency of Igκ⁺Igλ⁺ double-positive cells remained 176 unchanged between wild-type and rsCBE-deficient groups, indicating that light chain 177 isotype exclusion stays intact. Unlike the rsCBE mutant mice, loss of rpiaCBE alone 178 did not affect light chain isotype distribution (Fig. 1e, Extended Data Fig. 1h,i). 179 Overall, these findings confirm rsCBE as a key regulator of light chain isotype choice 180 in mouse B cells in vivo. Thus, our subsequent experiments will focus on the ΔrsCBE 181 mouse, with the ΔrpiaCBE mice serving as a comparison where relevant. 182 183 CBE controls early B cell development 184 To define the role of CBEs in B cell development, we performed flow cytometric 185 analysis of bone marrow B cell subsets—pro-B, large pre-B, small pre-B, and 186 immature B cells in wild-type and CBE-deficient mice. Deletion of the rsCBE element 187 resulted in a marked reduction in the frequency of small pre-B cells (Fig. 2a, 188 Extended Data Fig. 2a). Conversely, the frequencies of pro-B and large pre-B cell 189 compartments were unaffected significantly, although there was an increasing trend. 190 To determine whether the reduced pre-B cell frequency reflected an absolute 191 numerical loss or a shift in progenitor distributions, we quantified the absolute 192 numbers of each B cell subset isolated from the femur, tibia, and fibula of each 193 mouse. As expected, ΔrsCBE mice exhibited a pronounced decrease in the number 194 of small pre-B cells, whereas pro-B and immature B-cell numbers remained 195 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 10 unchanged (Fig. 2b). In contrast, ΔrpiaCBE mice displayed no alterations in B-cell 196 subset distribution (Extended Fig. 2b), suggesting a specific requirement for rsCBE 197 in maintaining early B-cell development. 198 199 To dissect the developmental impairment in ΔrsCBE mice further, we performed 200 genome-wide transcriptomic profiling of Fluorescence-Activated Cell Sorting (FACS)-201 purified pre-B cells. RNA-sequencing identified 70 differentially expressed genes, the 202 majority (~66) of which were significantly downregulated as compared with WT pre-B 203 cells (Fig. 2c). Gene set enrichment analysis of downregulated genes revealed 204 strong enrichment for pathways linked to cell cycle regulation (Fig. 2d). Consistently, 205 cell cycle analysis of bone marrow B-cell subsets by Hoechst staining in WT and 206 ΔrsCBE mice showed a marked accumulation of pro-B, large pre-B, small pre-B, and 207 immature B cells in the G1–S phase, with a smaller fraction in S and G2–M phases, 208 indicating cell-cycle arrest. (Fig. 2e, Extended Data Fig 2c). Given this defect, we 209 next assessed apoptosis using Annexin V staining. In WT and ΔrsCBE mice, pro-B 210 and large pre-B cells exhibited a similar level of apoptosis, while small pre-B and 211 immature B cells showed significantly higher frequencies of Annexin V⁺ early 212 apoptotic cells, indicating elevated cell death (Fig. 2f, Extended Data Fig. 2d). 213 Impaired proliferation combined with increased apoptosis reduces the small pre-B 214 cell compartment in ΔrsCBE mice. This also raises the possibility that the reduction 215 in λ⁺ cells is an indirect consequence of increased cell death among κ-rearranging 216 cells, potentially reflecting enhanced deletion of autoreactive B cells, a possibility that 217 warrants further investigation. 218 219 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 11 Splenic B cell subsets were analysed by flow cytometry in ΔrsCBE and ΔrpiaCBE 220 mice, revealing no significant differences in Marginal Zone B (MZB), Follicular B 221 (FoB), or Transitional 1 B (T1B) cells (Extended Data Fig. 2e). Altogether, these 222

identify rsCBE as a pivotal cis-regulatory element that sustains proliferation 223 and developmental progression of early B cell precursors. Its loss disrupts cell-cycle 224 dynamics and diminishes pre-B cell output, a phenotype not observed upon rpiaCBE 225 deletion, highlighting a unique role for rsCBE in safeguarding pre-B cell expansion. 226 227 Co-ordinated molecular epigenetic mechanisms orchestrate Igκ expression 228 Previous studies have demonstrated a strong correlation between chromatin 229 accessibility and the efficiency of Vκ–Jκ rearrangement58-62. To explore changes in 230 the epigenetic landscape across the Igκ locus, we performed ATAC-seq on ex vivo–231 purified pre-B cells. Our analysis specifically examined whether deletion of rsCBE, 232 and the consequent loss of insulation, alters chromatin accessibility both upstream 233 and downstream of the deleted regions. Given that the pre-B cells were derived from 234 Rag-proficient mice lacking an intact germline configuration, our investigation 235 concentrated mainly on the 3′-SEκ region of the Igκ locus. Loss of rsCBE-mediated 236 insulation resulted in increased chromatin accessibility downstream of the deletion 237 site, unveiling a previously insulated regulatory element, termed rsEκ (Fig. 3a). 238 Notably, rsEκ displayed significantly greater accessibility in ΔrsCBE mice compared 239 to wild-type controls and was markedly more accessible than the neighbouring HS10 240 co-enhancer element (Fig. 3b). In contrast, deletion of rpiaCBE induced a modest, 241 non-significant decrease in accessibility across the three canonical enhancers (Fig. 242 3a,b). These results reveal a complex interplay between the rsCBE and other 243 regulatory elements in shaping the Igκ chromatin accessibility landscape. 244 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 12 245 CTCF-binding elements are established regulators of RAG substrate accessibility21. 246 Given that the rsCBE, a CTCF-binding site, lies ~160 bp downstream of the RS 247 element, we assessed chromatin accessibility at the RS RSS. Of note, deletion of 248 rsCBE caused a pronounced decrease in accessibility at the Rag1 binding site within 249 the RS element (Fig. 3c). These findings reveal that the structural integrity of rsCBE 250 is crucial for maintaining recombination signal sequences (RSS) accessibility at the 251 RS locus, and its disruption impairs Vκ–RS recombination efficiency (see below). 252 253 To assess whether alterations in chromatin accessibility correlate with changes in 254 active enhancer marks, we evaluated H3K27ac enrichment. Wild-type and CBE-255 deleted mutant mice were bred onto a Rag1-deficient background and crossed with 256 human IgM (hIgM) transgenic mice. This strategy ensured that the Igκ locus 257 remained in the germline configuration at the pre-B cell stage. Bone marrow-derived 258 CD19⁺ cells, composed of over 95% pre-B cell purity, were isolated from WT-259 Rag1⁻/⁻.hIgM and mutant-Rag1⁻/⁻.hIgM mice and subjected to H3K27 acetylation 260 ChIP sequencing. 261 262 Since increased acetylation has been correlated to greater recombination 263 efficiency63-65, we first evaluated its enrichment at the 3′-SEκ. Notably, within the 3’-264 SEκ, the rsEκ and HS10 elements showed hyperacetylation only in ΔrsCBE-265 Rag1⁻/⁻.hIgM mice (Fig. 3d,e). In ΔrsCBE-Rag1⁻/⁻.hIgM cells, elevated H3K27 266 acetylation extended to the CER, 5′-GLT and 3′-GLT promoters, RC, and the 267 canonical 3′-Eκ regions. By contrast, ΔrpiaCBE-Rag1⁻/⁻.hIgM pre-B cells showed no 268 significant changes in H3K27ac across these regulatory elements within the 3′-SEκ 269 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 13 (Fig. 3d,e). H3K27ac enrichment was also markedly increased across the Ig Vκ gene 270 segments in ΔrsCBE-Rag1⁻/⁻.hIgM pre-B cells as compared to WT (Extended Data 271 Fig. 3a). We also assessed the H3K27ac levels at RSS in the Vκ region. We found 272 increased acetylation at the 12-RSS across the Ig Vκ gene segments in ΔrsCBE-273 Rag1⁻/⁻.hIgM pre-B cells (Fig. 3f). We observed no significant changes in H3K27 274 acetylation at the Ig Vκ gene segments or their RSS in ΔrpiaCBE-Rag1⁻/⁻.hIgM pre-275 B cells compared with WT. These results identify rsEκ as a previously insulated 276 putative shadow enhancer that gains chromatin accessibility and active histone 277 modifications upon loss of rsCBE-mediated insulation. Together, our findings reveal 278 that disruption of a specific CTCF-mediated insulation within the 3’-SEκ reshapes 279 chromatin architecture, influencing Igκ locus regulation. 280 281 CBE controls the intricate Igκ super-enhancer architecture through chromatin 282 insulation 283 Within the Igκ locus, the 3′-SEκ orchestrates chromatin interactions necessary for 284 proper light chain recombination and B-cell development13-19. CBEs play a pivotal 285 role in constraining these interactions through mechanisms such as chromatin 286 insulation11,20,22,41,42,44, yet how they influence the internal configuration of super-287 enhancer domains remains incompletely understood. To investigate the chromosome 288 topology of the 3′-SEκ architecture and its relationship to Igκ recombination, we 289 performed Micro-C-TALE in WT-Rag1⁻/⁻.hIgM and ΔrsCBE-Rag1⁻/⁻.hIgM mice. 290 Micro-C-TALE is a BAC-based targeted chromatin conformation capture method, 291 which utilizes micrococcal nuclease digestion of cross-linked chromatin to enable 292 high-resolution profiling of chromatin folding at target regions. We applied this 293 approach to ex vivo purified CD19⁺ pre-B cells to map chromatin organization across 294 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 14 the Igκ locus with fine structural detail. Micro-C-TALE yielded approximately 7-12 295 million high-quality paired-end reads per sample, producing high-resolution contact 296 maps across biological replicates from WT-Rag1⁻/⁻.hIgM and ΔrsCBE-Rag1⁻/⁻.hIgM 297 pre-B cells. While chromatin contact frequencies across the Igκ variable region were 298 broadly similar in both genotypes (Extended Data Fig. 4a), local short-range 299 interaction differences emerged upon detailed inspection of the contact maps at the 300 3′ regulatory region (3′-RR) of the Igκ locus (Fig. 4a). Notably, we observed 301 genotype-specific contact frequency at key regulatory nodes involving the chromatin 302 boundary element SIS, and other 3′-RRs (Fig. 4a, upper and lower triangles). 303 304 In the Micro-C-TALE interaction matrix, contact strength is indicated in the regions 305 marked by black and blue arrows for WT-Rag1⁻/⁻.hIgM and ΔrsCBE-Rag1⁻/⁻.hIgM 306 pre-B cells, respectively (Fig. 4a). Specifically, we compared chromatin interactions 307 within the 3′-SEκ. In WT cells, contacts among the canonical enhancers were 308 spatially constrained, primarily confined to the region between the SIS and rsCBE 309 (Fig. 4a, upper black triangle), and, to a lesser extent, between the SIS and rpiaCBE 310 (Fig. 4a, upper extended black triangle). The SIS–rsCBE domain encompasses the 311 three canonical enhancers, thereby restricting their activity within this domain and 312 limiting stronger interactions to the nearest downstream co-enhancers, such as 313 HS10. By contrast, loss of rsCBE insulation in ΔrsCBE-Rag1⁻/⁻.hIgM pre-B cells led 314 to pronounced changes in 3′-SEκ chromatin architecture. Contacts between the SIS 315 and rsCBE were lost (Fig. 4a, lower black triangle), while spatial interactions within 316 the SIS–rpiaCBE domain were enhanced (Fig. 4a, lower extended black triangle), 317 permitting increased contacts in the region between the putative rsEκ shadow 318 enhancer and dEκ (Fig. 4a, lower white triangle). These findings indicate that rsCBE 319 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 15 functions as a key insulator, ensuring domain-specific chromatin interactions, thereby 320 shaping enhancer connectivity and chromatin folding. Consistent with this altered 321 chromatin topology, the rsEκ region exhibited elevated chromatin accessibility and 322 H3K27ac enrichment in ΔrsCBE-Rag1⁻/⁻.hIgM pre-B cells (Fig. 3a,d). 323 324 The Igκ locus is organized into discrete sub-topologically associating domains 325 (subTADs)38,39,66, demarcated by boundary elements enriched for architectural 326 proteins such as CTCF and the cohesin subunit Rad2130. To assess the insulating 327 capacity of the rsCBE region, we computed the Insulation Score (IS) in WT-328 Rag1⁻/⁻.hIgM and ΔrsCBE-Rag1⁻/⁻.hIgM pre-B cells. The IS of a locus measures the 329 average interaction frequency between its upstream loci and downstream loci, such 330 that dips in the IS reflect potential insulating regions67. Notably, deletion of rsCBE led 331 to reduced insulation activity (higher IS) at this site, highlighting its essential role in 332 preserving the 3′-SEκ chromatin architecture and subTAD boundary integrity (Fig. 333 4b). 334 335 To delineate the insulating function of rsCBE within the 3′-SEκ region further, we 336 generated virtual 4C profiles from normalized Micro-C-TALE data at 1-kb resolution. 337 Comparative analysis between WT-Rag1⁻/⁻.hIgM and ΔrsCBE-Rag1⁻/⁻.hIgM pre-B 338 cells revealed that enhancer interactions across the 3′-SEκ were markedly altered 339 upon rsCBE deletion. Using the SIS element as a virtual 4C viewpoint, we found that 340 in WT-Rag1⁻/⁻.hIgM cells, the SIS interacts with rsCBE and rpiaCBE at comparable 341 frequencies, consistent with published data (Fig. 4c)40,68. In ΔrsCBE-Rag1⁻/⁻.hIgM 342 pre-B cells, loss of rsCBE led to aberrant chromatin looping toward the downstream 343 rpiaCBE at higher frequency, incorporating the putative rsEκ shadow enhancer within 344 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 16 the expanded interaction domain (Fig. 4c). Conversely, virtual 4C analysis from the 345 rsCBE viewpoint revealed a loss of contact with the SIS element in ΔrsCBE-346 Rag1⁻/⁻.hIgM cells compared with WT-Rag1⁻/⁻.hIgM cells, indicating disruption of 347 regular boundary activity (Extended Data Fig. 4b). Notably, the rsEκ viewpoint 348 exhibited strengthened interactions with dEκ, 3′-Eκ, and the Rpia promoter in the 349 absence of rsCBE (Fig. 4d). RC viewpoint analysis revealed enhanced contacts with 350 canonical enhancers, including 3′-Eκ, dEκ, and the putative rsEκ shadow enhancer 351 in ΔrsCBE-Rag1⁻/⁻.hIgM pre-B cells (Fig. 4e). Similarly, chromatin interactions from 352 the rsCBE, 3′-Eκ, and dEκ viewpoints were evaluated (Extended Data Fig. 4b-d). 353 These findings highlight a complex enhancer network driving elevated Igκ and Rpia 354 expression in ΔrsCBE pre-B cells. 355 356 To reinforce our in vivo observations, we extended our analysis to the v-Abl–357 transformed pro-B cell line 445.3 to dissect the contribution of rsCBE to enhancer 358 network organization within the 3′-SEκ. v-Abl–transformed lines are developmentally 359 arrested at the pro-B to pre-B cell transition. Treatment of 445.3 WT cells with the Abl 360 kinase inhibitor STI571 alleviates this block, inducing robust Igκ rearrangement and 361 GLT69. The WT 445.3 line originates from a Rag1-deficient background, thereby 362 preserving an intact germline Igκ locus and permitting the generation of deletion 363 mutants without prior V(D)J recombination events. Using CRISPR–Cas9 genome 364 editing, we deleted the 571-bp rsCBE region to recapitulate the in vivo deletion, 365 hereafter referred to as ΔrsCBE-571bp 445.3, respectively. Single-cell clones were 366 isolated and screened by DNA sequencing to confirm homozygous deletions. In 367 particular, we investigated whether the putative rsEκ shadow enhancer functions as 368 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 17 an active node within the reconfigured chromatin architecture that emerges following 369 rsCBE deletion. 370 371 We conducted 4C-seq in the Rag1-deficient v-Abl pro–B cell line 445.3, enabling 372 high-resolution mapping of chromatin contacts from a defined genomic viewpoint. 373 Using rsEκ as a viewpoint in ΔrsCBE-571bp 445.3 cells, we observed increased 374 contacts with IgκC and the RC compared with WT 445.3 cells, indicating that loss of 375 rsCBE insulation promotes its integration into the canonical super-enhancer 376 interaction hub. (Fig. 4f). To further explore this reorganization, we used RC as the 377 reciprocal viewpoint in 4C-seq. In ΔrsCBE-571bp cells, the RC exhibited elevated 378 interactions with 3′-Eκ, dEκ, rsEκ, and HS10 region, suggesting a substantial 379 remodelling of the enhancer network upon loss of rsCBE insulation (Fig. 4g). When 380 dEκ was used as the viewpoint, we noted reduced interactions with IgκC but 381 maintained contact with the RC, alongside interaction with rsEκ (Extended Data Fig. 382 4e). These data indicate that rsCBE deletion leads to a reconfiguration of enhancer 383 connectivity, with rsEκ emerging as a key regulatory node within the remodelled 384 super-enhancer architecture. These findings reveal that rsCBE serves as a critical 385 architectural boundary that orchestrates the structural integrity of the Igκ super-386 enhancer through chromatin insulation. 387 388 rsCBE-mediated chromatin loop extrusion controls B-cell tolerance and 389 autoimmunity 390 It is well established that murine λ⁺ B cells predominantly arise through receptor 391 editing, which commences after secondary Igκ rearrangement49,70-73. The RS 392 element, located approximately 25 kb downstream of the Cκ exon, harbours a 393 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 18 canonical recombination signal sequence and undergoes V(D)J recombination with 394 upstream Vκ segments or cryptic recombination sites within the Jκ–Cκ intron49,74. RS 395 recombination results in the deletion of the Cκ exon and the three canonical Igκ 396 enhancers, thereby functionally silencing the Igκ locus49,71,72,75. As RS rearrangement 397 is predominant in murine λ⁺ B cells (77%)76 but occurs in only 12% of κ⁺ B cells73, we 398 analysed Vκ–RS recombination in purified splenic λ⁺ B cells. Remarkably, Vκ–RS 399 recombination was reduced approximately fivefold in ΔrsCBE mice compared to 400 controls. Moreover, λ⁺ B cells from these mutants displayed decreased usage of Vκ–401 Jκ1 and enhanced usage of Vκ–Jκ5 segments, reflecting sequential κ locus 402 rearrangements characteristic of receptor editing (Fig. 5a). This suggests that, in the 403 absence of rsCBE, κ⁺ cells fail to transition into λ⁺ B cells, reflecting defective 404 receptor editing77. In contrast, deletion of rpiaCBE alone had no detectable effect on 405 Vκ–RS recombination (Fig. 5a). Collectively, these findings identify rsCBE as a key 406 regulator of Vκ–RS recombination and an essential determinant of λ⁺ B cell 407 development. 408 409 To investigate whether CTCF-mediated chromatin loop extrusion drives Vκ–RS 410 recombination, we focused on the v-Abl–transformed pro-B cell line. We generated 411 an additional deletion targeting the 18 bp CTCF-binding motif within rsCBE, hereafter 412 referred to as ΔrsCBE-18bp 445.3, respectively. Single-cell clones were isolated and 413 screened by DNA sequencing to confirm homozygous deletions. Additionally, the 414 ΔrsCBE-18bp clones were validated by ChIP-qPCR to confirm loss of CTCF binding 415 (Extended Data Fig. 5a). Given that 445.3 cells are Rag1-deficient, both the 445.3 416 WT and rsCBE-deleted lines were transduced with a Rag1-expressing retrovirus to 417 restore recombination capability. Following transduction, cells were treated with 418 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 19 STI571 for 20 hours, and Vκ–RS recombination levels were quantified by qPCR. 419 Remarkably, the ΔrsCBE-18bp 445.3 clone lacking the CTCF-binding motif 420 recapitulated the reduction in Vκ–RS recombination observed with the larger 421 ΔrsCBE-571bp deletion, both in vitro and in vivo. (Fig. 5b). These results strongly 422 support the model that CTCF-mediated chromatin loop extrusion at the rsCBE locus 423 is a critical driver of Vκ–RS recombination, facilitating the development of λ⁺ B cells. 424 425 Primary rearrangements, whether by deletion or inversion, typically delete or invert 426 the CER and SIS elements40. In vitro studies with v-Abl pro-B cells suggest that RAG 427 scanning from the primary Jκ–RC is halted by the CTCF-dependent SIS element40. 428 We therefore propose that the SIS element acts as a barrier to rsCBE-mediated 429 CTCF interactions with the Vκ region. Consequently, deletion and inversion of CER 430 and SIS during primary rearrangement may enable rsCBE to interact with the Vκ 431 region, thereby promoting Vκ–RS secondary recombination. 432 433 To investigate long-range chromatin interactions between rsCBE and Vκ gene 434 segments, and to assess whether the SIS functions as a barrier to rsCBE-mediated 435 CTCF contacts with the Vκ region, we performed in vitro studies using the v-Abl–436 transformed pro-B cell line. Using 4C-seq with a viewpoint positioned upstream of 437 rsCBE, we found that rsCBE establishes contacts with dEκ, 3′-Eκ, IgκC, RC, SIS, 438 and the CER element in WT 445.3 cells (Extended Data Fig. 5b). By contrast, both 439 ΔrsCBE-571bp and ΔrsCBE-18bp 445.3 cells showed markedly reduced interactions 440 with these regulatory elements, with the most pronounced loss observed for contacts 441 with SIS (as expected) and CER, (Extended Data Fig. 5b). To quantitatively assess 442 alterations in long-range chromatin interaction, we performed cumulative frequency 443 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 20 distribution (CFD) analysis of 4C-seq read density upstream of the rsCBE viewpoint. 444 This approach allowed us to systematically measure the extent of rsCBE–Vκ region 445 interactions across the locus. Strikingly, both the frequency and intensity of rsCBE–446 Vκ contacts were markedly diminished in ΔrsCBE-571bp and ΔrsCBE-18bp 445.3 447 cells, with the ΔrsCBE-571bp deletion exhibiting an even more pronounced defect, 448 underscoring the critical role of rsCBE in maintaining productive long-range 449 interactions with distal Vκ gene segments. (Fig. 5c). 450 451 To explore chromatin interactions during secondary rearrangement, we also deleted 452 the ~3 kb SIS element, which harbours two CTCF-binding sites oriented toward 453 rsCBE, and is thought to constrain rsCBE–Vκ interactions in the v-Abl–transformed 454 line. Homozygous ΔSIS 445.3 single-cell clones were confirmed by DNA sequencing 455 and used for downstream assays. As shown above, 4C-seq analysis revealed that in 456 WT 445.3 cells, rsCBE engages in interactions with dEκ, 3′-Eκ, IgκC, RC, SIS, and 457 the CER element (Extended Data Fig. 5c). By contrast, in ΔSIS 445.3 cells, rsCBE 458 exhibited a pronounced loss of contact with IgκC and SIS, accompanied by 459 enhanced interactions with the Vκ region, consistent with SIS functioning as a 460 chromatin barrier that modulates rsCBE–Vκ contact dynamics (Extended Data Fig. 461 5c). In ΔSIS 445.3 cells, CFD analysis revealed a marked increase in upstream 462 interactions, toward the Vκ region, compared to WT 445.3 cells (Fig. 5c). Indeed, 463 deletion of the SIS element in vivo has been shown to increase Vκ–RS 464 recombination by ∼1.2-fold in splenic λ⁺ B cells22. These findings suggest that SIS 465 constrains rsCBE activity during primary rearrangement. At the same time, rsCBE 466 promotes chromatin looping toward the Vκ gene cluster to drive secondary 467 rearrangements, when the SIS element is absent. 468 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 21 469 We subsequently investigated how deletion of rsCBE influences allelic exclusion, 470 with particular emphasis on assessing the emergence of dual κ-expressing B cells in 471 ΔrsCBE mice. Given that allelic inclusion is a consequence of receptor editing, it may 472 facilitate the emergence and persistence of autoreactive B cells by circumventing 473 central tolerance mechanisms, thereby contributing to autoimmunity78-82. To evaluate 474 this, we crossed ΔrsCBE with human Cκ knock-in mice to generate Igκm/h 475 heterozygotes and analysed IgM⁺ immature B cells in the bone marrow (Extended 476 Data Fig. 5d). Flow cytometric analysis revealed that rsCBE deficiency resulted in a 477 modest yet significant increase in the proportion of cells expressing the mouse Cκ 478 (mCκ) allele, accompanied by a modest, non-significant decrease in human Cκ 479 (hCκ). (Fig. 5d, Extended Data Fig. 5e). Notably, ΔrsCBE heterozygotes also 480 exhibited a significant increase in the frequency of dual-expressing mCκ⁺hCκ⁺ cells, 481 indicative of allelic inclusion (Fig. 5d). These findings suggest that rsCBE promotes 482 preferential usage of the Igκ allele in cis and plays a role in enforcing allelic exclusion 483 during early B cell development. 484 485 We further assessed allelic usage in the spleen of Igκm/h heterozygous mice. Flow 486 cytometric analysis of splenic B220⁺ B cells confirmed that loss of rsCBE resulted in 487 a modest but significant increase in the proportion of cells expressing the mCκ allele 488 (Fig. 5e, Extended Data Fig. 5e). Our data demonstrate that deletion of rsCBE 489

in a small breakdown of allelic exclusion in the bone marrow, as evidenced by 490 the emergence of dual Igκ allele-expressing (mCκ⁺hCκ⁺) immature B cells. 491 492 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 22 Receptor editing is a key mechanism of B cell tolerance, whereby κ⁺ B cells first 493 attempt secondary κ rearrangements to replace an autoreactive κ chain with a non-494 autoreactive one. If autoreactivity persists, RS recombination is triggered to activate 495 the λ locus and express a λ light chain, thereby restoring self-tolerance49. Inefficient 496 RS rearrangement can compromise this process, allowing autoreactive B cells to 497 persist in the periphery, a feature observed in murine models of systemic lupus 498 erythematosus (SLE) and type 1 diabetes (T1D), as well as in patients with SLE83,84. 499 Given that RS rearrangement is diminished in such models, we hypothesized that 500 impaired RS recombination in splenic λ⁺ B cells from ΔrsCBE mice may underlie 501 defects in central B cell tolerance. Thus, we measured serum levels of 502 autoantibodies targeting canonical self-antigens85-88 in ΔrsCBE mice. Notably, 503 ΔrsCBE mice exhibited elevated titres of IgG-class anti-nuclear antigen (ANA) and 504 anti-LPS antibodies at 14 months of age, as assessed by ELISA—hallmarks of 505 systemic autoimmune pathology (Fig. 5f). Serum levels of anti-DNA and anti-insulin 506 also showed an elevated trend compared to wild-type controls (Fig. 5f). Analysis of 507 splenic B cell subsets in ΔrsCBE mice revealed no major alterations, although the 508 mice exhibited splenomegaly, indicating inflammation (Extended Data Fig. 5f,g). 509 These findings collectively suggest that loss of rsCBE compromises immune 510 tolerance and precipitates autoantibody production. 511 512 To examine whether elevated Igκ expression in ΔrsCBE mice contributes to early-513 onset of systemic-like autoimmunity, we assessed serum autoantibody levels at 10 514 weeks of age. Notably, ΔrsCBE mice displayed an increase in IgG autoantibodies 515 targeting DNA and insulin compared to wild-type controls (Fig. 5g). These findings 516 implicate dysregulated Igκ expression, driven by the loss of rsCBE insulation, in 517 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 23 possibly promoting early autoimmune responses. These findings underscore the 518 regulatory importance of enhancer insulation in preserving B-cell immune tolerance 519 from a young age. 520 521 522 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 24

523 The RS element is a central regulator of Igλ expression, light-chain isotype 524 exclusion, and B cell receptor editing49, yet the epigenetic basis of these processes 525 has remained elusive. Here, we show that rsCBE-mediated chromatin loop extrusion 526 establishes a CTCF-dependent axis that mechanistically underpins this regulation. 527 Loss of rsCBE insulation unmasks a previously sequestered putative shadow 528 enhancer, rsEκ, which aberrantly integrates into the 3′-SEκ regulatory hub to drive 529 coordinated misregulation of Igκ and the closely linked gene Rpia. This architectural 530 disruption not only enhances Igκ expression but also predisposes to autoantibody 531 production, directly linking chromatin topology to the maintenance of central B cell 532 tolerance. More broadly, our findings demonstrate that individual cis-regulatory CBEs 533 within the 3′-SEκ function as distinct, bidirectional transcriptional insulators that 534 safeguard the integrity of transcription. 535 536 Our findings establish insulation within the 3′-SEκ as a critical determinant of gene 537 regulation in B cells. Loss of rsCBE produced a ~1.8-fold increase in Igκ expression 538 in pre-B cells, whereas rpiaCBE deletion caused a modest but significant reduction. 539 Strikingly, both deletions derepressed the neighbouring Rpia gene, demonstrating 540 that individual CBEs within the 3′-SEκ function as non-redundant architectural 541 boundaries that differentially tune enhancer–promoter communication. This 542 mechanistic insight contrasts with earlier reports showing that deletion of SIS, CER, 543 or both elements within the 3′-SEκ had minimal impact on Igκ or Rpia expression, 544 leaving unresolved how Rpia was insulated from the potent regulatory activity of the 545 3′-SEκ11,22,42-44. Even loss of HS10, the co-enhancer closest to Rpia, failed to perturb 546 its expression14, underscoring that enhancer proximity alone does not dictate 547 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 25 insulation. Furthermore, CTCF gene deletion in pre-B cells was shown to reduce 548 both 5′- and 3′-GLTs at the Jκ region while increasing proximal Vκ GLT and elevating 549 GLTs at the Jλ1 and Jλ3 regions20. Similarly, mutations in Igκ enhancers within the 3′-550 SEκ decrease Igκ expression but enhance Igλ recombination13,18,74,89,90. The 551 simultaneous upregulation of Igκ and Rpia, or of either gene individually, has not 552 been observed in previous studies, emphasizing the unique role of rsCBE-mediated 553 insulation in preserving balanced gene regulation within B cells. Given that aberrant 554 Rpia activation has been implicated in autophagy and diverse cancers53-56, our 555 findings highlight CBE-dependent chromatin architecture as an essential mechanism 556 that safeguards Rpia insulation within the 3′-SEκ. Whether the increased expression 557 of the Rpia gene affects B cell function or survival warrants further investigation. 558 559 Our study identifies rsCBE as a critical architectural element in early B cell 560 development, functioning to insulate the putative rsEκ shadow enhancer and prevent 561 its premature activation. Loss of rsCBE insulation in adult mice most probably 562 unleashes rsEκ activity in pro-B cells, driving untimely Igκ recombination. This 563 premature recombination imposes a G1–S checkpoint arrest during the pro-B to 564 immature B cell transition, resulting in attrition of the pre-B cell pool through 565 apoptosis. In earlier studies, Igκ rearrangement was induced in fetal-derived B1 pro-566 B cells91. Thus, rsCBE safeguards developmental timing by constraining enhancer 567 activity until the appropriate stage, and its loss reveals that Igκ recombination can be 568 induced in adult pro-B cells. 569 570 Loss of rsCBE and rpiaCBE insulation triggers profound epigenetic remodeling within 571 the 3′-SEκ. Disruption of rsCBE enhances chromatin accessibility at rsEκ, HS10, and 572 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 26 rpiaCBE, while concomitantly reducing accessibility at the RS element in pre-B cells. 573 This suggests that rsCBE functions as a gatekeeper of Rag1 accessibility, with 574 CTCF-mediated loop extrusion directing Vκ–RS synapsis and recombination21. The 575 accompanying increase in H3K27ac across rsEκ, HS10, 3′-Eκ, and the RC, together 576 with enrichment at Vκ RSSs and Vκ genes, indicates that enhancer activation and 577 chromatin remodelling are tightly coupled to recombination potential63-65. Thus, 578 rsCBE insulation safeguards the balance between enhancer-driven transcriptional 579 activation and recombinase accessibility, and its loss disrupts this architecture to 580 prematurely couple Igκ expression with aberrant Vκ–RS recombination in pre-B cells. 581 582 The 3′-SEκ functions as a highly organized regulatory hub of enhancer interactions, 583 integrating the CER, SIS, rsCBE, and rpiaCBE into its chromatin architecture37. 584 Within the 3′-SEκ network, rsCBE and SIS CBE form a boundary domain that 585 encapsulates the three canonical enhancers, thereby constraining their activity to 586 maintain physiological levels of Igκ and Rpia expression. Disruption of rsCBE 587 insulation rewires this architecture by looping toward rpiaCBE and incorporating the 588 previously sequestered putative rsEκ shadow enhancer, leading to simultaneous 589 upregulation of both Igκ and Rpia. Beyond transcriptional control, rsCBE is 590 indispensable for Vκ–RS recombination in pre-B cells, akin to the RS element49. 591 Loss of rsCBE reduces RS element accessibility in vivo and abrogates Rag1-592 mediated recombination. This is in line with an established study showing that CBEs 593 mediate accessibility of Rag substrates during chromatin scanning21. It is also 594 consistent with models of RAG scanning, which propose that RAG complexes 595 search for RSSs located within convergent CTCF sites in the IgH locus92-99. 596 597 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 27 Functionally, increased Igκ expression skews light-chain isotype usage, producing a 598 twofold reduction in λ⁺ cells, due to impaired receptor editing and hyperactivation of 599 the Igκ locus. Interestingly, in vitro analyses demonstrate that rsCBE facilitates Vκ–600 RS secondary rearrangements through direct interactions with the Vκ variable 601 region, contacts that are normally constrained by the SIS element40. Importantly, 602 primary rearrangements involving deletion or inversion of CER and SIS remove this 603 barrier, allowing rsCBE to engage during secondary editing. Together, these findings 604 establish rsCBE as a dual-function regulator that coordinates enhancer insulation, 605 Vκ–RS recombination, and light-chain isotype balance. 606 607 While light chain isotype exclusion remained intact, allelic exclusion was 608 compromised. Hybrid reporter mice revealed an increased frequency of immature B 609 cells co-expressing both Igκ alleles (mCκ⁺hCκ⁺), demonstrating a requirement for 610 rsCBE in maintaining allelic exclusion. The presence of such dual-mCκ⁺hCκ⁺ cells in 611 the bone marrow suggests defective clonal deletion, possibly because of dilution of 612 autoreactive BCRs78-80. An additional layer of peripheral clonal deletion is still active 613 and effective, as dual-κ+ cells are absent in the periphery. However, this peripheral 614 deletion mechanism is not absolute, as evidenced by increased autoantibody titres in 615 the serum. These findings establish rsCBE as a cis-regulatory element that 616 safeguards B cell tolerance. By coordinating receptor editing and allelic exclusion, 617 rsCBE prevents the emergence of autoreactive clones and preserves immune 618 homeostasis. Its loss uncovers a critical checkpoint in central tolerance, where Igκ 619 hyperactivation and defective editing converge to accelerate the generation of 620 autoreactive B cells. 621 622 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 28 B cells are subject to stringent selection in the bone marrow to eliminate autoreactive 623 clones as part of the central tolerance mechanism. Such cells can undergo receptor 624 editing, clonal deletion, or anergy to generate a non-autoreactive repertoire. Deletion 625 of rsCBE in vivo impaired receptor editing, leading to a reduction in λ⁺ B cells and 626 spontaneous production of IgG autoantibodies as early as ten weeks of age. Notably, 627 rsCBE-deficient mice exhibited serum autoantibodies against DNA and insulin—628 hallmarks of systemic autoimmune disease such as lupus, well before they typically 629 arise in C57BL/6 mice, which normally do not develop anti-DNA antibodies until late 630 in life100. These findings indicate that rsCBE is required to preserve B-cell tolerance 631 at a young age. At 14 months of age, rsCBE-deficient mice displayed elevated IgG 632 autoantibody titres against nuclear antigens (ANA) and lipopolysaccharide (LPS), 633 further underscoring rsCBE long-term role in immune homeostasis. 634 635 Previous studies have shown that germline RS mutations reduce the frequency of λ⁺ 636 B cells by impairing receptor editing, while leaving Igκ rearrangement and allelic 637 exclusion intact49. In contrast, rsCBE deletion not only recapitulates editing defects 638 but also drives excessive Igκ rearrangement in pre-B cells. This shift skews light 639 chain usage, with splenic λ⁺ B cells showing increased detection of Vκ–Jκ5 640 segments, reflecting a failure to undergo Vκ–RS recombination in the absence of 641 rsCBE, a defining feature of secondary rearrangements. The marked reduction in λ⁺ 642 cells, coupled with increased κ⁺ cells in bone marrow and spleen, highlights the 643 essential role of rsCBE in sustaining λ⁺ B cell development. 644 645 Transgenic mouse models of defective receptor editing manifest enhanced clonal 646 deletion without spontaneous autoantibody production101-103. Even under conditions 647 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 29 that block apoptosis, such as Bcl-2 overexpression, inefficiently edited B cells 648 generally remain subject to tolerance mechanisms with minimal autoantibody 649 production104-106. Only when the RS mutation is combined with anti-apoptotic signals 650 has autoimmunity been observed, and then only in non-physiological settings49,105-651 107. By contrast, loss of rsCBE alone is sufficient to disrupt receptor editing and 652 trigger early breakage of B cell tolerance. Unlike transgenic models of central 653 tolerance, which express a high-avidity monoclonal BCR that primarily promotes 654 clonal deletion and peripheral anergy, rsCBE-deficient mice retain a normal 655 polyclonal BCR repertoire. This physiological diversity reveals a broader escape 656 from both central and peripheral tolerance checkpoints. 657 658 Our findings position rsCBE as a dual-function regulatory element that couples 659 chromatin architecture with B cell tolerance. By insulating the putative rsEκ shadow 660 enhancer and directing CTCF–cohesin loop extrusion, rsCBE safeguards the 661 developmental timing of Igκ recombination, maintains allelic exclusion, and 662 preserves the balance of light-chain isotypes. Its loss disrupts enhancer insulation, 663 deregulates Igκ and Rpia expression, and impairs Vκ–RS recombination, thereby 664 compromising receptor editing. The early production of autoantibodies in rsCBE-665 deficient mice at a young age directly links chromatin architecture to central 666 tolerance. Our study demonstrates that defects in central B cell tolerance predispose 667 to failures in peripheral tolerance, culminating in early autoantibody production. More 668 broadly, our study demonstrates that individual CBEs within the 3′-SEκ act as non-669 redundant architectural boundaries that orchestrate enhancer cross-talk and 670 promoter engagement to preserve transcriptional integrity. 671 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 30

672 Mice 673 Transgenic mice expressing a human immunoglobulin heavy chain (Igh-Tg⁺), 674 originally described, were generously provided by Dr. Cornelis Murre (University of 675 California, San Diego)108. Mice carrying a human Cκ knock-in gene were kindly 676 provided by Michel C. Nussenzweig (The Rockefeller University, New York, NY)78. 677 All animal procedures were approved by the Institutional Animal Care and Use 678 Committee of the Hebrew University of Jerusalem (MD-24-17515-4). Wild-type 679 C57BL/6 and genetically modified mice were housed under specific pathogen-free 680 (SPF) conditions. Animals were maintained on a 12-hour light/dark cycle at 22 ± 2 °C 681 and 55 ± 15% relative humidity. Unless otherwise indicated in the figure legends, 682 male mice aged 6–12 weeks were used for experiments. All lines were maintained 683 on a C57BL/6 genetic background. For the allelic exclusion experiments, reporter 684 mice were maintained on a B6/BALB/c hybrid background. 685 686 Generation of rsCBE and rpiaCBE deletions in mice using CRISPR–Cas9 687 To generate an 18-bp rsCBE deletion in 445.3 cells, a single-stranded 688 oligodeoxynucleotide (ssODN; IDT) containing 50-bp homology arms flanking the 18-689 bp CTCF-binding motif was used as a homology-directed repair (HDR) template. A 690 single-guide RNA (sgRNA) targeting the CTCF-binding motif within the 571-bp 691 rsCBE region (rsCBE-sgRNA3) was designed using the CRISPR design tool 692 (http://crispr.mit.edu) and cloned into pSpCas9(BB)-2A-GFP (pX458; Addgene 693 #48138). To generate the 571-bp rsCBE deletion, two sgRNAs (rsCBE-sgRNA1 and 694 rsCBE-sgRNA2) were individually cloned into pX458, and all constructs were 695 sequence-verified. The ~3 kb SIS deletion was produced using two sgRNAs (SIS-696 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 31 sgRNA1 and SIS-sgRNA2), which were likewise cloned and verified. pX458 carrying 697 rsCBE-sgRNA3 together with the ssODN were co-transfected into 445.3 cells using 698 the Neon electroporation system to induce the 18-bp deletion, whereas the 571-bp 699 deletion was achieved by co-transfecting the two rsCBE pX458 constructs. Similarly, 700 the SIS pX458 pair was co-transfected to generate the ~3 kb deletion. After 48 h, 701 GFP⁺ cells were single-cell sorted by FACS and cultured for two weeks to allow 702 colony formation. Individual clones were screened by PCR for biallelic deletions, and 703 successful editing was confirmed by Sanger sequencing. CRISPR/Cas9-mediated 704 genomic modifications were validated by PCR and DNA sequencing. 705 706 Cells 707 Bone marrow-derived cells were cultured on an ST2 stromal feeder layer (kindly 708 provided by A. Rolink)109 following density gradient separation using Histopaque. 709 Cells were maintained in RPMI 1640 medium supplemented with 10% fetal bovine 710 serum (FBS), 2 mM L-glutamine, 100 U/ml penicillin, 100 μg/ml streptomycin, and 50 711 μM 2-mercaptoethanol. Interleukin-7 (IL-7; PeproTech, Cat. 217-17) was added to 712 the culture at a final concentration of 5 ng/ml. The Abelson virus-transformed pro-B 713 cell line 445.3 was maintained in RPMI 1640 medium supplemented with 10% FBS, 714 2 mM L-glutamine, 100 U/ml penicillin, 100 μg/ml streptomycin, and 50 μM 2-715 mercaptoethanol. The 445.3 cells are Rag1−/− and were kindly provided by Dr. 716 Cornelis Murre (UCSD). Hep-2 cells were also cultured in RPMI 1640 medium 717 supplemented with 10% FBS, 2 mM L-glutamine, 100 U/ml penicillin, 100 μg/ml 718 streptomycin, and 50 μM 2-mercaptoethanol. HEK 293T cells were maintained in 719 DMEM supplemented with 10% FBS, 2 mM L-glutamine, and 100 U/ml penicillin–100 720 μg/ml streptomycin. 721 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 32 722 Bone marrow and splenic B cell extraction 723 The femur, tibia, and fibula (and humeri in some instances) were dissected, rinsed 724 with PBS, and mechanically homogenized in 2 ml of FACS buffer (0.5% BSA in PBS) 725 using a mortar and pestle until completely depigmented. The resulting cell 726 suspension was filtered through a 100 μm nylon mesh into a 50 ml conical tube and 727 centrifuged at 300 × g for 5 min at 4 °C. The cell pellet was subsequently 728 resuspended in an appropriate volume of FACS buffer for downstream use. 729 Lymphocytes were enriched by layering cell suspensions onto Histopaque-1077 730 (Sigma-Aldrich), followed by centrifugation at 700 × g for 30 min at 4 °C without 731 brake. The interface containing enriched lymphocytes was carefully collected, 732 washed with FACS buffer, and pelleted by centrifugation at 1,200 rpm for 5 min at 4 733 °C. The cell pellet was resuspended in 300 µl of chilled FACS buffer and labelled for 734 flow cytometric sorting. For flow cytometric analysis, bone marrow and splenic cells 735 were treated with 1× Ammonium–Chloride–Potassium (ACK) lysis buffer to remove 736 red blood cells, then washed and resuspended in an appropriate volume for analysis. 737 738 Flow Cytometry 739 Pre-B cells (B220+IgM-CD43-CD25+) were isolated from the bone marrow of pooled 740 cohorts of five to six female mice (6–12 weeks old) for Igκ RNA library preparation. 741 DNA and RNA were extracted from samples of individual mice for subsequent qPCR 742 analysis. Bone marrow cell suspensions were counted, washed in 10 ml of 743 phosphate-buffered saline (PBS), and resuspended at 1 × 10⁶ cells per 100 μl in 744 FACS buffer containing Fc receptor blocking reagent (anti-CD16/CD32, Invitrogen, 745 #14-0161-82; 1:100 dilution). Cells were stained for 25 min at 4 °C with the following 746 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 33 antibodies (1:200–1:400): anti-B220 (RA3-6B2, PerCP/Cy5.5, BioLegend, #103235); 747 anti-IgM (II/41, APC, eBioscience, #17-5790-82; II/41, eFluor™ 450, Invitrogen, #48-748 5790-82); anti-CD25 (PC61.5, Alexa Fluor 488, Invitrogen, #53-0251-82; PC61, 749 BV786, BD Horizon, #564023); anti-CD19 (1D3/CD19, PE-Cy7, BioLegend, 750 #152417; 1D3/CD19, APC, BioLegend, #152410); and anti-CD43 (eBioR2/60, PE, 751 Invitrogen, #12-0431-81). After staining, cells were washed and resuspended in 752 FACS buffer for analysis on a Beckman CytoFLEX cytometer. Data were analysed 753 using FlowJo v10.7 (BD Biosciences), and cells were sorted on a BD FACSAria. 754 755 For splenic B cell analysis, single-cell suspensions were first incubated with Fc 756 receptor–blocking antibody for 10 min at 4 °C. After washing, cells were stained for 757 25 min at 4 °C with the following antibodies: anti-B220; anti-IgM (II/41, eFluor™ 450, 758 Invitrogen, #48-5790-82); anti-CD93 (AA4.1, APC, eBioscience/Invitrogen, #17-759 5892-81); anti-CD21 (7G6, FITC, BD Pharmingen, #561769); and anti-CD23 (B3B4, 760 PE, BD Pharmingen, #561773). 761 762 For analysis of immunoglobulin light chain isotypes, splenic single-cell suspensions 763 were stained for 25 min at 4 °C with anti-mouse Cκ (H139-52.1, PE, 764 SouthernBiotech, #1180-09), anti-Igλ1–3 (R26-46, BD Pharmingen, #553434), and 765 anti-B220. For bone marrow immature B cell analysis, cells were stained with anti-766 CD19 (1D3/CD19, PE-Cy7, BioLegend, #152417), anti-B220, anti-IgM (II/41, APC, 767 eBioscience, #17-5790-82), as well as anti-mouse Cκ and anti-Igλ1–3. 768 769 For assessment of allelic exclusion, single-cell suspensions were stained with anti-770 IgM (II/41, eFluor™ 450, Thermo Fisher #42-5790-82) in combination with anti-771 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 34 mouse Cκ. After washing, cells were further incubated with anti-human Cκ (SB81a, 772 APC, SouthernBiotech, #9230-11), followed by a final wash and filtration through a 773 40 μm cell strainer. Samples were acquired on a flow cytometer and analysed using 774 FlowJo v10.7 (BD Biosciences). 775 776 CRISPR-Cas9 mediated mutagenesis 777 To generate an 18-bp rsCBE deletion in 445.3 cells, a single-stranded 778 oligodeoxynucleotide (ssODN; IDT) containing 50-bp homology arms flanking the 18-779 bp CTCF-binding motif was used as a homology-directed repair (HDR) template. A 780 single-guide RNA (sgRNA) targeting the CTCF-binding motif within the 571-bp 781 rsCBE region (rsCBE-sgRNA3) was designed using the CRISPR design tool 782 (http://crispr.mit.edu) and cloned into pSpCas9(BB)-2A-GFP (PX458; Addgene 783 #48138). To generate the 571-bp rsCBE deletion, two sgRNAs (rsCBE-sgRNA1 and 784 rsCBE-sgRNA2) were individually cloned into PX458, and all constructs were 785 sequence-verified. The ~3 kb SIS deletion was produced using two sgRNAs (SIS-786 sgRNA1 and SIS-sgRNA2), which were likewise cloned and verified. PX458 carrying 787 rsCBE-sgRNA3 together with the ssODN were co-transfected into 445.3 cells using 788 the Neon electroporation system to induce the 18-bp deletion, whereas the 571-bp 789 deletion was achieved by co-transfecting the two rsCBE PX458 constructs. Similarly, 790 the SIS PX458 pair was co-transfected to generate the ~3 kb deletion. After 48 h, 791 GFP⁺ cells were single-cell sorted by FACS and cultured for two weeks to allow 792 colony formation. Individual clones were screened by PCR for biallelic deletions, and 793 successful editing was confirmed by Sanger sequencing. CRISPR/Cas9-mediated 794 genomic modifications were validated by PCR and DNA sequencing. 795 796 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 35 RNA and genomic DNA analysis 797 Total RNA was extracted from 4–6 × 10⁵ FACS-purified cells using the High Pure 798 RNA Isolation Kit (Roche, #11828665001) or the Nucleospin RNA Kit (Macherey-799 Nagel, #740955.50) according to the manufacturers’ instructions. First-strand cDNA 800 was synthesized from 30–60 ng of total RNA using the qScript cDNA Synthesis Kit 801 (Quanta Bio, #95047-100). Genomic DNA was isolated from a similar number of cells 802 using the DNeasy Blood & Tissue Kit (Qiagen, #96504). Quantitative real-time PCR 803 (qPCR) was performed on a Bio-Rad CFX Connect system using SYBR Green 804 Master Mix. 805 806 4C Sequencing 807 4C template and library preparation was performed as described previously110. In 808 brief, 5–10 × 10⁶ Rag1⁻/⁻ 445.3-WT, ΔrsCBE-571 bp 445.3, ΔrsCBE-18 bp 445.3, 809 and ΔSIS 445.3 cells were crosslinked with 1% formaldehyde following stimulation 810 with kinase inhibitors. Primary digestion was performed using NlaIII (NEB, R0125S), 811 followed by dilution and ligation. Cross-links were then reversed by proteinase K 812 treatment, and templates were trimmed with Csp6I (NEB, R0639S) before re-ligation. 813 Inverted PCR was carried out from multiple viewpoints using the Expand Long 814 Template PCR System (Roche, #11681842001) with indexed Illumina sequencing 815 adapters, and products were purified using the Qiaquick PCR Purification Kit 816 (Qiagen, #28104). Libraries were sequenced on an Illumina NovaSeq 6000 with 817 122–160 bp single-end reads. 4C-Seq data were analysed using the pipe4C 818 pipeline110, and genomic interaction profiles were visualized by smoothing WIG files 819 with the npreg R package (spar = 0.75). 4C data were normalized to reads per 820 million (RPM). CFD differences upstream of the rsCBE viewpoint among WT 445.3, 821 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 36 ΔrsCBE-571 bp 445.3, ΔrsCBE-18 bp 445.3, and ΔSIS 445.3 were assessed for 822 significance using the Mann–Whitney test in Prism 8.0.1. 823 824 Retrovirus preparation and transduction 825 To generate a Rag1-expressing retrovirus, HEK 293T cells at 70–80% confluence 826 were co-transfected with pMSCV-IRES-Bsr-Rag1, encoding full-length wild-type 827 murine Rag1 and a blasticidin resistance cassette, and the pEco packaging plasmid, 828 which expresses the ecotropic envelope protein under the CMV immediate-early 829 promoter. Transfections were performed using Mirus reagent according to the 830 manufacturer’s instructions, and viral supernatants were collected 48 h post-831 transfection. For in vitro recombination assays, 445.3 pro-B cells were transduced 832 with the Rag1 retrovirus and cultured in blasticidin (20 µg/mL) for seven days to 833 select for stable expression. Cells were subsequently treated with STI571 to induce 834 recombination. 835 836 Igκ repertoire RNA sequencing and analysis 837 DNase-treated RNA was isolated from FACS-purified pre-B cells from bone marrow. 838 The library preparation protocol was adapted from Rena Levin-Klein et al. 2017111. 839 RNA was poly-A enriched using poly-dT beads (Life Technologies) in two selection 840 cycles and RT then performed using AffinityScript QPCR cDNA Synthesis Kit 841 (Agilent) with an RT primer specific for the Cκ region 5′-842 ATGCTGTAGGTGCTGTCTTT-3′. The residual RNA was degraded with 0.1 N NaOH, 843 neutralized with 0.1 M acetic acid and the single strand cDNA then purified using 844 Silane beads (Life Technologies). A 3TR3 adapter (5′-/Phos/ 845 AGATCGGAAGAGCACACGTCTG/3SpC3/-3′) was ligated to the 3′ end by overnight 846 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 37 incubation with T4 RNA ligase (NEB) at 22 °C, and the cDNA then purified from 847 excess adapter with Silane beads and PCR amplified for 12 cycles using the reverse 848 complement of the 3RT3 adapter as the forward primer and the upstream Cκ region 849 as the reverse primer with the partial Truseq Illumina adapter added to the beginning 850 (5′-TACACGACGCTCTTCCGATCT-ACTGGATGGTGGGAAGATGGAT-3′). The PCR 851 product was cleaned with 0.7 × ampure XT beads, amplified with indexed universal 852 Illumina adapter primers for an additional seven cycles to obtain ∼550 bp libraries. 853 These libraries were sequenced on a Miseq platform yielding 151 bp paired-end. 854 Sequencing reads were processed to retain only read pairs (R1 and R2) with an 855 average Phred quality score ≥30. R2 reads were reverse complemented, and both 856 R1 and R2 were aligned independently to the C57BL/6 mouse kappa light chain V 857 and J gene reference sequences from the IMGT9 database using IgBLAST112. The 858 proportion of Vĸ gene usage was calculated. Statistical comparisons between 859 genotypes were performed using the two-sided Mann–Whitney U test. 860 861 Genome-wide RNA sequencing and analysis 862 Total RNA was extracted from ex vivo FACS-purified pre-B cells using the 863 NucleoSpin RNA Kit according to the manufacturer’s instructions. RNA integrity was 864 assessed using an Agilent TapeStation, and concentrations were quantified with a 865 Qubit fluorometer (Thermo Fisher). A total of 1 µg of DNase-treated RNA was used. 866 Ribosomal RNA was depleted prior to library construction. RNA-seq libraries were 867 prepared using the SMARTer Stranded Total RNA Sample Prep Kit – HI Mammalian 868 (Takara) following the manufacturer’s recommendations. This kit generates libraries 869 compatible with Illumina platforms for mammalian samples. Libraries were pooled to 870 a final concentration of 10 nM and loaded on a NextSeq 500 using the NextSeq 871 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 38 500/550 High-Output v2.5 Kit (75 cycles). Sequencing generated 122-bp single-end 872 reads. Raw FASTQ files were quality-checked and trimmed using Trim Galore 873 (v0.6.10). Trimmed reads were aligned to the mm10 reference genome using STAR 874 (v2.7.11) with default parameters113. Resulting BAM files were used for gene-level 875 quantification with HTSeq (v2.0.3)114. Raw read counts were analyzed in R (v4.4.1), 876 and normalization and differential gene expression analysis were performed using 877 the DESeq2 package115. Genes with a fold change >1.85 and a Benjamini–878 Hochberg–adjusted p-value (Padj) <0.05 were considered significantly differentially 879 expressed. 880 881 ATAC sequencing 882 ATAC-seq libraries were prepared from 50–100 × 10³ ex vivo FACS-purified pre-B 883 cells using the Active Motif ATAC-Seq Kit (Active Motif, #53150) according to the 884 manufacturer’s protocol. Briefly, nuclei were isolated, tagmented with pre-assembled 885 transposomes, and DNA was purified. Libraries were amplified using the Nextera 886 Illumina kit following the manufacturer’s instructions. Final libraries were size-887 selected with a 1.2× SPRI bead cleanup and sequenced on an Illumina NovaSeq 888 6000 platform with 61-bp paired-end reads. 889 890 H3K27ac and CTCF ChIP sequencing 891 For H3K27ac ChIP-sequencing, approximately 1.8–5.3 × 10⁶ CD19⁺ pre-B cells were 892 enriched from WT-Rag1⁻/⁻.hIgM and mutant-Rag1⁻/⁻.hIgM mice using anti-CD19 893 magnetic beads (Miltenyi Biotec) and MACS column separation (Miltenyi Biotec). For 894 CTCF ChIP-sequencing, 3–6 × 10⁶ pre-B cells cultured in IL-7 from WT and ΔrsCBE 895 mice were used. Cells were cross-linked with 1% formaldehyde for 10 min at room 896 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 39 temperature and quenched with 0.1 M glycine. Cells were washed twice with cold 897 phosphate-buffered saline (PBS) and lysed with lysis buffer [0.5% SDS, 10 mM 898 EDTA, 50 mM Tris-HCl (pH 8), and protease inhibitor]. DNA was sonicated in an 899 ultrasonic bath (Diagenode, Bioruptor) to an average length of 300–500 bp. 900 Sonicated chromatin was centrifuged at 16,000 rpm for 15 min, and the supernatants 901 were immunoprecipitated overnight with anti-H3K27ac (Active Motif, #39133) or anti-902 CTCF (Cell Signaling Technology, #3418S) antibodies. Protein G beads (Cell 903 Signaling Technology, #9006) were added to immunoprecipitated samples and 904 incubated for 2.5 h at 4°C. Beads were sequentially washed for 5 min each in low-905 salt RIPA [20 mM Tris-HCl (pH 8), 150 mM NaCl, 2 mM EDTA, 1% Triton X-100, 906 0.1% SDS], high-salt RIPA [20 mM Tris-HCl (pH 8), 500 mM NaCl, 2 mM EDTA, 1% 907 Triton X-100, 0.1% SDS], LiCl buffer [10 mM Tris (pH 8.0), 1 mM EDTA, 250 mM 908 LiCl, 1% NP-40, 1% Na-deoxycholate], and Tris-EDTA buffer. Beads were eluted in 909 TE (pH 8.0) with 0.1% SDS and 150 mM NaCl for 1 h at 65 °C, followed by RNase 910 treatment for 30 min at 37 °C. Proteinase K was added and incubated at 50 °C for 2 911 h. Reverse cross-linking was performed by adding NaCl to a final concentration of 912 200 mM and incubating for at least 6 h at 65°C. DNA was purified by phenol–913 chloroform extraction, and libraries were prepared from the immunoprecipitated DNA 914 using the KAPA Hyper Prep Kit according to the manufacturer’s protocol. H3K27ac 915 ChIP–seq libraries were sequenced on an Illumina NovaSeq 6000, generating 122-916 bp single-end reads, while CTCF ChIP–seq libraries were sequenced on an Illumina 917 NextSeq, generating 122-bp single-end reads. 918 919 ChIP-qPCR 920 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 40 To assess CTCF occupancy at the targeted deletion site, 445.3-WT and ΔrsCBE-18 921 bp 445.3 v-Abl Rag1-deficient cells were subjected to anti-CTCF ChIP analysis. 922 Experiments were performed using 1 × 10⁷ cells per condition, following the 923 protocols described above, with 1% formaldehyde. Immunoprecipitated and input 924 DNA samples were analysed by quantitative real-time PCR, and enrichment was 925 calculated relative to input to account for variability in chromatin quantity. 926 927 ChIP sequencing and ATAC sequencing data analysis 928 Raw FASTQ files from ATAC-seq and ChIP-seq experiments (paired-end and single-929 end, respectively) were processed. Quality control and adapter trimming were 930 performed using Trim Galore (v0.6.10). Cleaned reads were aligned to the mm10 931

genome using Bowtie2 (v2.5.2) with default parameters116, and the 932 resulting alignments were saved as BAM files. Samtools (v1.19.2) merge was then 933 used with default parameters to combine data from the two biological replicates117. 934 The merged BAM files were converted to BigWig format for visualization using 935 bamCoverage, normalized to counts per million (CPM)118. Peak calling was 936 performed on the merged BAM files using MACS2 (v2.2.8) callpeak119, with peaks 937 identified using MACS2 for ATAC-seq and de novo peak detection for ChIP-seq. 938 Read counts from both experiments were normalized using the csaw package. The 939 normalized counts were converted into a DGEList object in edgeR to identify 940 differentially accessible or enriched regions120. Normalized BigWig signal intensities 941 were extracted for defined genomic regions and visualized as heatmaps using 942 plotHeatmap for ATAC-seq and H3K27ac ChIP–seq118. 943 944 Micro-C-TALE sequencing 945 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 41 Micro-C-TALE was carried out as described below (unpublished data). Briefly, 1.6–4 946 × 10⁶ fixed and permeabilized cells were treated with micrococcal nuclease (MNase) 947 (ThermoFisher, #88216) to obtain predominantly mononucleosomal DNA. The DNA 948 was then ligated using T4 DNA ligase (NEB, #M0202L), extracted by phenol–949 chloroform and ethanol precipitation, and ligation products were enriched by size 950 selection using Ampure XP beads (Beckman Coulter, #A63881). End repair and 951 adaptor ligation were performed using the NEBNext Ultra II DNA Library Prep Kit 952 (NEB, #E7645S), followed by PCR amplification with indexed primers. Four Micro-C 953 libraries were generated and pooled in equimolar proportions. 954 955 Capture probes were prepared from fifteen BACs spanning the target genomic 956 region. BAC DNA was mixed equimolarly, sheared to ~200-bp fragments by 957 sonication, end-repaired, and ligated to specialized adaptors using the NEBNext 958 Ultra II kit. The BAC library was then PCR-amplified using biotinylated primers. 959 Hybridization-based capture was performed by incubating the pooled Micro-C 960 libraries with the biotinylated BAC-derived probes in multiple replicates. Hybridized 961 products were pulled down using streptavidin-coated magnetic beads (Dynabeads 962 MyOne sterptavidin C1, Invitrogen, #65001) and amplified by PCR with Illumina-963 compatible primers. The capture cycle was repeated once, and the final libraries 964 were sequenced on a DNBSEQ-G400 using paired-end 150-bp reads. 965 966 Micro-C-TALE analysis 967 Reads were mapped, filtered, and balanced using nf-distiller 968 (https://github.com/open2c/distiller-nf) and cooler 969 (https://academic.oup.com/bioinformatics/article/36/1/311/5530598). 970 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 42 971 Apoptosis Assay 972 Apoptosis was assessed using the FITC Annexin V Apoptosis Detection Kit with 973 propidium iodide (PI) (BioLegend, #640914). Fc receptors were blocked with anti-974 CD16/CD32 antibody in 1 × 10⁶ Histopaque-purified bone marrow lymphocytes. 975 Following centrifugation, cells were stained with fluorochrome-conjugated antibodies 976 against surface markers for 25 min at 4 °C, washed twice with FACS buffer, and 977 resuspended in Annexin V binding buffer. Annexin V and PI were added (5 µl each 978 per 100 µl of cell suspension) and incubated for 15 min at room temperature in the 979 dark. Samples were then diluted with an additional 100 µl of binding buffer prior to 980 flow cytometric acquisition. 981 982 Cell cycle Analysis 983 Bone marrow cells were subjected to ACK lysis, followed by surface marker staining 984 of 2 × 10⁶ cells with fluorochrome-conjugated antibodies for cell cycle analysis. Cells 985 were then incubated with Hoechst 33342 (Sigma-Aldrich, #B2261) (1 µl of 5 mg ml⁻¹ 986 stock per 500 µl DPBS) for 30 min at room temperature with mixing, resuspended in 987 200 µl FACS buffer, and analysed on a CytoFLEX flow cytometer. 988 989 Serum isolation 990 Blood was collected from mice via tail vein or cardiac puncture into 1.5 ml 991 microcentrifuge tubes. Samples were allowed to coagulate at room temperature (20–992 25 °C) for 30–60 min, followed by centrifugation at 1,500 × g for 10 min. The resulting 993 serum was carefully aspirated and stored at −80 °C until further analysis. 994 995 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 43 ELISA 996 ELISA reactions were carried out using flat-bottom MaxiSorp™ 96-well Nunc-997 Immuno plates (Thermo Scientific, #442404). The assay was adapted from a 998 previously published protocol121. Antigens (Hep-2 lysate; lipopolysaccharides from 999 Escherichia coli O55:B5, Merck, #L4005-100MG; Salmon Sperm DNA Solution, 1000 Invitrogen, #15632011; Human Insulin Solution, Merck, #I9278-5ML) were coated in 1001 PBS at 100 µl per well and incubated overnight at 4 °C. For comparative ELISA 1002 assays with multiple targets, antigens were plated at 0.5 µg/well. Plates were 1003 washed five times with washing buffer (1× PBS, 0.05% Tween-20) and incubated 1004 with 100 µl blocking buffer (1× PBS with 1% BSA) for 1 h at room temperature. The 1005 blocking solution was then replaced with serum samples for 2 h at room 1006 temperature. Serum samples were assayed at a 1:50 dilution. Plates were washed 1007 five times with washing buffer and incubated with anti-mouse IgG secondary 1008 antibody conjugated to horseradish peroxidase (HRP) (Jackson ImmunoResearch, 1009 #115-035-062) in PBS at a 1:6,000 dilution. After an additional five washes, plates 1010 were developed using 3,3',5,5'-tetramethylbenzidine (TMB) (BioFX TMB One 1011 Component HRP Microwell Substrate, Surmodics), and absorbance was measured 1012 at 630 nm using an ELISA plate reader (Tecan Spark). 1013 1014 Quantification and Statistical analysis 1015 Statistical analyses were performed using two-tailed unpaired t-tests in GraphPad 1016 Prism (v8.0.1). Data are expressed as mean ± SD, with p-values < 0.05 considered 1017 statistically significant. Details of specific statistical tests are provided in the 1018 corresponding sections. 1019 1020 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 44

1021 We thank Ihab Abd-Elrahman and Kirill Makedonski from the National Genetically 1022 Engineered Mouse Models (GEMM) Unit, The Hebrew University of Jerusalem, for 1023 generating the transgenic mice. We are grateful to A. Nasereddin and I. Shiff from 1024 the Genomic Applications Laboratory, Core Research Facility, Faculty of Medicine – 1025 Ein Kerem, Hebrew University of Jerusalem, for their scientific advice and RNA 1026 sequencing services. We also thank Dr. Hadas Segev-Yekutiel and Dr. Eleonora 1027 Medvedev from the Core Research Facility at the Hebrew University School of 1028 Medicine for their assistance with flow cytometry analysis. This work was supported 1029 by research grants from the Israel Academy of Sciences (grant no. 1228/18 to Y .B.), 1030 the Israel Cancer Research Foundation (grant no. 211410 to Y.B.) , the Isr ael Cancer 1031 Association (grant no. 20241009 to Y.B.), the United States–Israel Binational Science 1032 Foundation (grant no. 2100289 to Y.B.), and the Emanuel Rubin Chair in Medical 1033 Sciences (Y.B.). 1034 1035 Author contributions 1036 D.G. conceived the study, designed and performed all experiments, and analysed 1037 and interpreted the data. E.G. performed bioinformatic analyses of all high-1038 throughput datasets. D.G. and A.G. conducted the Micro-C-TALE sequencing 1039 experiments under the supervision of N.K. We adapted the Micro-C-TALE method 1040 from a protocol originally developed by AKG. D.G. performed the CTCF ChIP–seq 1041 and H3K27ac ChIP–seq experiments under the supervision of M.H and Y.D. A.H. 1042 provided the 4C sequencing scripts and primer information. Bar Avidov analysed the 1043 Vκ RNA repertoire for the rsCBE genotype and provided the corresponding analysis 1044 scripts. F.C. and L.H. contributed to the FACS experiments and offered valuable 1045 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 45 technical guidance for the analysis. Batia Azria contributed to the initial analysis of 1046 the RNA-seq dataset for the mutant genotypes. D.G. wrote the entire manuscript 1047 with input from all authors. Y.B. oversaw, supervised and directed the entire study. 1048 1049 Competing interests 1050 The authors declare that they have no competing interests. 1051 1052 Data and materials availability: 1053 High-throughput sequencing data generated in this study have been deposited in the 1054 Gene Expression Omnibus (GEO) database. Igκ repertoire RNA-seq data are 1055 available under accession number GSE313431; genome-wide RNA-seq data under 1056 GSE313430; ATAC-seq data under GSE313428; CTCF ChIP-seq data under 1057 GSE313432; H3K27ac ChIP-seq data under GSE313433; 4C-seq data under 1058 GSE313429; and Micro-C-TALE-seq data under GSE313783. 1059 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 46

1060 1 Alt, F. W ., Zhang, Y ., Meng, F. L., Guo, C. & Schwer, B. Mechanisms of programmed DNA lesions 1061 and genomic instability in the immune system. Cell 152, 417 -429, 1062 doi:10.1016/j.cell.2013.01.007 (2013). 1063 2 Deguine, J. & Xavier, R. J. B cell tolerance and autoimm unity: Lessons from repertoires. J Exp 1064 Med 221, doi:10.1084/jem.20231314 (2024). 1065 3 Young, C., Lau, A. W. Y. & BurneZ, D. L. B cells in the balance: Offse]ng self-reac^vity avoidance 1066 with protec^on against foreign. Front Immunol 13, 951385, doi:10.3389/fimmu.2022.951385 1067 (2022). 1068 4 Watanabe, A. et al. Self-tolerance curtails the B cell repertoire to microbial epitopes. JCI Insight 1069 4, doi:10.1172/jci.insight.122551 (2019). 1070 5 Jung, D. & Alt, F. W . Unraveling V(D)J recombina^on; insights into gene regula^on. Cell 116, 1071 299-311, doi:10.1016/s0092-8674(04)00039-x (2004). 1072 6 Schatz, D. G. & Swanson, P . C. V(D)J recombina^on: mechanisms of ini^a^on. Annu Rev Genet 1073 45, 167-202, doi:10.1146/annurev-genet-110410-132552 (2011). 1074 7 Jhunjhunwala, S., van Zelm, M. C., Peak, M. M. & Murre, C. Chroma^n architecture and the 1075 genera^on of an^gen receptor diversity. Cell 138, 435 -448, doi:10.1016/j.cell.2009.07.016 1076 (2009). 1077 8 Harrow, J. et al. GENCODE: producing a reference annota^on for ENCODE. Genome Biol 7 Suppl 1078 1, S4 1-9, doi:10.1186/gb-2006-7-s1-s4 (2006). 1079 9 Lefranc, M. P . et al. IMGT, the interna^onal ImMunoGeneTics informa^on system. Nucleic 1080 Acids Res 37, D1006-1012, doi:10.1093/nar/gkn838 (2009). 1081 10 Brekke, K. M. & Garrard, W. T. Assembly and analysis of the mouse immunoglobulin kappa 1082 gene sequence. Immunogene@cs 56, 490-505, doi:10.1007/s00251-004-0659-0 (2004). 1083 11 Kleiman, E., Xu, J. & Feeney, A. J. Cu]ng Edge: Proper Orienta^on of CTCF Sites in Cer Is 1084 Required for Normal Jkappa-Distal and Jkappa-Proximal Vkappa Gene Usage. J Immunol 201, 1085 1633-1638, doi:10.4049/jimmunol.1800785 (2018). 1086 12 Aoki-Ota, M., Torkamani, A., Ota, T., Schork, N. & Nemazee, D. Skewed primary Igkappa 1087 repertoire and V-J joining in C57BL/6 mice: implica^ons for recombina^on a ccessibility and 1088 receptor edi^ng. J Immunol 188, 2305-2315, doi:10.4049/jimmunol.1103484 (2012). 1089 13 Inlay, M., Alt, F. W ., Bal^more, D. & Xu, Y . Essen^al roles of the kappa light chain intronic 1090 enhancer and 3' enhancer in kappa rearrangement and demethy la^on. Nat Immunol 3, 463-1091 468, doi:10.1038/ni790 (2002). 1092 14 Zhou, X., Xiang, Y ., Ding, X. & Garrard, W . T. A new hypersensi^ve site, HS10, and the enhancers, 1093 E3' and Ed, differen^ally regulate Igkappa gene expression. J Immunol 188, 2722 -2732, 1094 doi:10.4049/jimmunol.1102758 (2012). 1095 15 Zhou, X., Xiang, Y ., Ding, X. & Garrard, W . T. Loss of an Igkappa gene enhancer in mature B cells 1096

in rapid gene silencing and par^al reversible dedifferen^a^on. Mol Cell Biol 33, 2091-1097 2101, doi:10.1128/MCB.01569-12 (2013). 1098 16 Gorman, J. R. et al. The Ig(kappa) enhancer influences the ra^o of Ig(kappa) versus Ig(lambda) 1099 B lymphocytes. Immunity 5, 241-252, doi:10.1016/s1074-7613(00)80319-2 (1996). 1100 17 Levin-Klein, R., Kirillov, A., Rosenbluh, C., Cedar, H. & Bergman, Y. A n o ve l p a x 5-binding 1101 regulatory element in the igkappa locus. Front Immunol 5, 240, 1102 doi:10.3389/fimmu.2014.00240 (2014). 1103 18 Xu, Y., Davidson, L., Alt, F. W. & Bal^more, D. Dele^on of the Ig kappa light chain intronic 1104 enhancer/matrix aZachment region impairs but does not abolish V kappa J kappa 1105 rearrangement. Immunity 4, 377-385, doi:10.1016/s1074-7613(00)80251-4 (1996). 1106 19 Barajas-Mora, E. M. & Feeney, A. J. Enhancers as regulators of an^gen receptor loci three -1107 dimensional chroma^n structure. Transcrip@on 11, 37 -51, 1108 doi:10.1080/21541264.2019.1699383 (2020). 1109 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 47 20 Ribeiro de Almeida, C. et al. The DNA -binding protein CTCF limits proximal Vkappa 1110 recombina^on and restricts kappa enhancer interac^ons to the immunoglobulin kappa light 1111 chain locus. Immunity 35, 501-513, doi:10.1016/j.immuni.2011.07.014 (2011). 1112 21 Jain, S., Ba, Z., Zhang, Y ., Dai, H. Q. & Alt, F. W . CTCF-Binding Elements Mediate Accessibility of 1113 RAG Substrates During Chroma^n Scanning. Cell 174, 102 -116 e114, 1114 doi:10.1016/j.cell.2018.04.035 (2018). 1115 22 Xiang, Y ., Zhou, X., HewiZ, S. L., Skok, J. A. & Garrard, W . T. A mul^func^onal element in the 1116 mouse Igkappa locus that specifies repertoire and Ig loci subnuclear loca^on. J Immunol 186, 1117 5356-5366, doi:10.4049/jimmunol.1003794 (2011). 1118 23 Ebert, A., Hill, L. & Busslinger, M. Spa^al Regula^on of V -(D)J Recombina^on at An^gen 1119 Receptor Loci. Adv Immunol 128, 93-121, doi:10.1016/bs.ai.2015.07.006 (2015). 1120 24 Beagan, J. A. & Phillips -Cremins, J. E. On the existence and func^onality of topolo gically 1121 associa^ng domains. Nat Genet 52, 8-16, doi:10.1038/s41588-019-0561-1 (2020). 1122 25 Phillips-Cremins, J. E. et al. Architectural protein subclasses shape 3D organiza^on of genomes 1123 during lineage commitment. Cell 153, 1281-1295, doi:10.1016/j.cell.2013.04.053 (2013). 1124 26 Schwarzer, W. et al. Two independent modes of chroma^n organiza^on revealed by cohesin 1125 removal. Nature 551, 51-56, doi:10.1038/nature24281 (2017). 1126 27 Fudenberg, G. et al. Forma^on of Chromosomal Domains by Loop Extrusion. Cell Rep 15, 2038-1127 2049, doi:10.1016/j.celrep.2016.04.085 (2016). 1128 28 Rao, S. S. P . et al. Cohesin Loss Eliminates All Loop Domains. Cell 171, 305 -320 e324, 1129 doi:10.1016/j.cell.2017.09.026 (2017). 1130 29 Sanborn, A. L. et al. Chroma^n extrusion explains key features of loop and domain forma^on 1131 in wild -type and engineered genomes. Proc Natl Acad Sci U S A 112, E6456 -6465, 1132 doi:10.1073/pnas.1518552112 (2015). 1133 30 Rao, S. S. et al. A 3D map of the human genome at kilobase resolu^ on reveals principles of 1134 chroma^n looping. Cell 159, 1665-1680, doi:10.1016/j.cell.2014.11.021 (2014). 1135 31 Vos, E. S. M. et al. Interplay between CTCF boundaries and a super enhancer controls cohesin 1136 extrusion trajectories and gene expression. Mol Cell 81, 3082 -3095 e3086, 1137 doi:10.1016/j.molcel.2021.06.008 (2021). 1138 32 Liu, N. Q. et al. WAPL maintains a cohesin loading cycle to preserve cell -type-specific distal 1139 gene regula^on. Nat Genet 53, 100-109, doi:10.1038/s41588-020-00744-4 (2021). 1140 33 Zhu, Y ., Denholtz, M., Lu, H. & Murre, C. Calcium signaling instructs NIPBL recruitment at ac^ve 1141 enhancers and promoters via dis^nct mechanisms to reconstruct genome 1142 compartmentaliza^on. Genes Dev 35, 65-81, doi:10.1101/gad.343475.120 (2021). 1143 34 Denholtz, M. et al. Upon microbial challenge, human neutrophils undergo rapid changes in 1144 nuclear architecture and chroma^n folding to orchestrate an immediate inflammatory gene 1145 program. Genes Dev 34, 149-165, doi:10.1101/gad.333708.119 (2020). 1146 35 Lichtenstein, M., Keini, G., Cedar, H. & Bergman, Y . B cell-specific demethyla^on: a novel role 1147 for the intronic kappa chain enhancer sequence. Cell 76, 913 -923, doi:10.1016/0092 -1148 8674(94)90365-4 (1994). 1149 36 Levin-Klein, R. & Bergman, Y. Regula^on of IgL Chain Recombina^on. Encyclopedia of 1150 Immunobiology, Vol 1: Development and Phylogeny of the Immune System , 71 -77, 1151 doi:10.1016/B978-0-12-374279-7.01008-0 (2016). 1152 37 Qian, J. et al. B cell super-enhancers and regulatory clusters recruit AID tumorigenic ac^vity. 1153 Cell 159, 1524-1537, doi:10.1016/j.cell.2014.11.013 (2014). 1154 38 Barajas-Mora, E. M. et al. A B -Cell-Specific Enhancer Orchestrates Nuclear Architecture to 1155 Generate a Diverse An^gen Receptor Repertoire. Mol Cell 73, 48 -60 e45, 1156 doi:10.1016/j.molcel.2018.10.013 (2019). 1157 39 Barajas-Mora, E. M. et al. Enhancer-instructed epigene^c landscape and chroma^n 1158 compartmentaliza^on dictate a primary an^body repertoire protec^ve against specific 1159 bacterial pathogens. Nat Immunol 24, 320-336, doi:10.1038/s41590-022-01402-z (2023). 1160 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 48 40 Zhang, Y . et al. Molecular basis for differen^al Igk versus Igh V(D)J joining mechanisms. Nature 1161 630, 189-197, doi:10.1038/s41586-024-07477-y (2024). 1162 41 Ribeiro de Almeida, C., Hendriks, R. W . & Stadhouders, R. Dynamic Control of Long -Range 1163 Genomic Interac^ons at the Immunoglobulin kappa Light-Chain Locus. Adv Immunol 128, 183-1164 271, doi:10.1016/bs.ai.2015.07.004 (2015). 1165 42 Xiang, Y ., Park, S. K. & Garrard, W . T. Vkappa gene repertoire and locus contrac^on are specified 1166 by cri^cal DNase I hypersensi^ve sites within the Vkappa-Jkappa intervening region. J Immunol 1167 190, 1819-1826, doi:10.4049/jimmunol.1203127 (2013). 1168 43 Liu, Z. et al. A recombina^on silencer that specifies heterochroma^n posi^oning and ikaros 1169 associa^on in the immunoglobulin kappa locus. Immunity 24, 405 -415, 1170 doi:10.1016/j.immuni.2006.02.001 (2006). 1171 44 Xiang, Y ., Park, S. K. & Garrard, W . T. A major dele^on in the Vkappa-Jkappa intervening region 1172

in hyperelevated transcrip^on of proximal Vkappa genes and a severely restricted 1173 repertoire. J Immunol 193, 3746-3754, doi:10.4049/jimmunol.1401574 (2014). 1174 45 Matheson, L. S. et al. Local Chroma^n Features Including PU.1 and IKAROS Binding and H3K4 1175 Methyla^on Shape the Repertoire of Immunoglobulin Kappa Genes Chosen for V(D)J 1176 Recombina^on. Front Immunol 8, 1550, doi:10.3389/fimmu.2017.01550 (2017). 1177 46 Kleiman, E., Loguercio, S. & Feeney, A. J. Epigene^c Enhancer Marks and Transcrip^on Factor 1178 Binding Influence Vkappa Gene Rearrangement in Pre-B Cells and Pro-B Cells. Front Immunol 1179 9, 2074, doi:10.3389/fimmu.2018.02074 (2018). 1180 47 Stadhouders, R. et al. Pre-B cell receptor signaling induces immunoglobulin kappa locus 1181 accessibility by func^onal redistribu^on of enhancer -mediated chroma^n interac^ons. PLoS 1182 Biol 12, e1001791, doi:10.1371/journal.pbio.1001791 (2014). 1183 48 Lazorchak, A. S., Schlissel, M. S. & Zhuang, Y . E2A and IRF -4/Pip promote chroma^n 1184 modifica^on and transcrip^on of the immunoglobulin kappa locus in pre-B cells. Mol Cell Biol 1185 26, 810-821, doi:10.1128/MCB.26.3.810-821.2006 (2006). 1186 49 Vela, J. L., Ait -Azzouzene, D., Duong, B. H., Ota, T. & Nemazee, D. Rearrangement of mouse 1187 immunoglobulin kappa dele^ng element recombining sequence promotes immune tolerance 1188 and lambda B cell produc^on. Immunity 28, 161 -170, doi:10.1016/j.immuni.2 007.12.011 1189 (2008). 1190 50 Durdik, J., Moore, M. W. & Selsing, E. Novel kappa light-chain gene rearrangements in mouse 1191 lambda light chain -producing B lymphocytes. Nature 307, 749 -752, doi:10.1038/307749a0 1192 (1984). 1193 51 Mar^n, D. J. & van Ness, B. G. Ini^a^on an d processing of two kappa immunoglobulin germ 1194 line transcripts in mouse B cells. Mol Cell Biol 10, 1950-1958, doi:10.1128/mcb.10.5.1950-1195 1958.1990 (1990). 1196 52 Amin, R. H. et al. Biallelic, ubiquitous transcrip^on from the distal germline Igkappa locus 1197 promoter during B cell development. Proc Natl Acad Sci U S A 106, 522 -527, 1198 doi:10.1073/pnas.0808895106 (2009). 1199 53 Nieh, Y . C. et al. Suppression of Ribose -5-Phosphate Isomerase a Induces ROS to Ac^vate 1200 Autophagy, Apoptosis, and Cellular Senescence in Lung Canc er . Int J Mol Sci 23, 1201 doi:10.3390/ijms23147883 (2022). 1202 54 Heintze, J., Costa, J. R., Weber, M. & KeZeler, R. Ribose 5 -phosphate isomerase inhibits LC3 1203 processing and basal autophagy. Cell Signal 28, 1380-1388, doi:10.1016/j.cellsig.2016.06.015 1204 (2016). 1205 55 Ciou, S. C. et al. Ribose-5-phosphate isomerase A regulates hepatocarcinogenesis via PP2A and 1206 ERK signaling. Int J Cancer 137, 104-115, doi:10.1002/ijc.29361 (2015). 1207 56 Chou, Y . T. et al. Iden^fica^on of a noncanonical func^on for ribose-5-phosphate isomerase A 1208 promotes colorectal cancer forma^on by stabilizing and ac^va^ng beta-catenin via a novel C-1209 terminal domain. PLoS Biol 16, e2003714, doi:10.1371/journal.pbio.2003714 (2018). 1210 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 49 57 Novobrantseva, T. I. et al. Rearrangement and expression of immunoglobulin light chain genes 1211 can precede heavy chain expression during normal B cell development in mice. J Exp Med 189, 1212 75-88, doi:10.1084/jem.189.1.75 (1999). 1213 58 Espinoza, C. R. & Feeney, A. J. Chroma^n accessibility and epigene^c modifica^ons differ 1214 between frequently and infrequently rearranging VH genes. Mol Immunol 44, 2675 -2685, 1215 doi:10.1016/j.molimm.2006.12.002 (2007). 1216 59 Pulivarthy, S. R. et al. Regulated large -scale nucleosome density paZerns and precise 1217 nucleosome posi^oning correlate with V(D)J recombina^on. Proc Natl Acad Sci U S A 113, 1218 E6427-E6436, doi:10.1073/pnas.1605543113 (2016). 1219 60 Yancopoulos, G. D., Blackwell, T. K., Suh, H., Hood, L. & Alt, F. W. Introduced T cell receptor 1220 variable region gene segments recombine in pre -B cells: evidence that B and T cells use a 1221 common recombinase. Cell 44, 251-259, doi:10.1016/0092-8674(86)90759-2 (1986). 1222 61 Yancopoulos, G. D. & Alt, F. W. Developm entally controlled and ^ssue -specific expression of 1223 unrearranged VH gene segments. Cell. 1985. 40: 271-281. J Immunol 188, 10-20 (2012). 1224 62 Stanhope-Baker, P ., Hudson, K. M., Shaffer, A. L., Constan^nescu, A. & Schlissel, M. S. Cell type-1225 specific chroma^n structure determines the targe^ng of V(D)J recombinase ac^vity in vitro. 1226 Cell 85, 887-897, doi:10.1016/s0092-8674(00)81272-6 (1996). 1227 63 McBlane, F. & Boyes, J. S^mula^on of V(D)J recombina^on by histone acetyla^on. Curr Biol 10, 1228 483-486, doi:10.1016/s0960-9822(00)00449-8 (2000). 1229 64 Nigh^ngale, K. P . et al. Acetyla^on increases access of remodelling complexes to their 1230 nucleosome targets to enhance ini^a^on of V(D)J recombina^on. Nucleic Acids Res 35, 6311-1231 6321, doi:10.1093/nar/gkm650 (2007). 1232 65 Goldmit, M. et al. Epigene^c ontogeny of the Igk locus during B cell development. Nat Immunol 1233 6, 198-203, doi:10.1038/ni1154 (2005). 1234 66 Lin, Y . C. et al. Global changes in the nuclear posi^oning of genes and intra - and interdomain 1235 genomic interac^ons t hat orchestrate B cell fate. Nat Immunol 13, 1196 -1204, 1236 doi:10.1038/ni.2432 (2012). 1237 67 Crane, E. et al. Condensin-driven remodelling of X chromosome topology during dosage 1238 compensa^on. Nature 523, 240-244, doi:10.1038/nature14450 (2015). 1239 68 Hill, L. et al. Igh and Igk loci use different folding principles for V gene recombina^on due to 1240 dis^nct chromosomal architectures of pro -B and pre -B cells. Nat Commun 14, 2316, 1241 doi:10.1038/s41467-023-37994-9 (2023). 1242 69 Muljo, S. A. & Schlissel, M. S. A small molecule Abl kinase inhibitor induces differen^a^on of 1243 Abelson virus-transformed pre-B cell lines. Nat Immunol 4, 31-37, doi:10.1038/ni870 (2003). 1244 70 Tiegs, S. L., Russell, D. M. & Nemazee, D. Receptor edi^ng in self-reac^ve bone marrow B cells. 1245 J Exp Med 177, 1009-1020, doi:10.1084/jem.177.4.1009 (1993). 1246 71 ReZer, M. W. & Nemazee, D. Receptor edi^ng occurs frequently during normal B cell 1247 development. J Exp Med 188, 1231-1238, doi:10.1084/jem.188.7.1231 (1998). 1248 72 Moore, M. W., Durdik, J., Persiani, D. M. & Selsing, E. Dele^ons of kappa chain constant region 1249 genes in mouse lambda chain -producing B cells involve intrachromosomal DNA 1250 recombina^ons similar to V -J joining. Proc Natl Acad Sci U S A 82, 6211 -6215, 1251 doi:10.1073/pnas.82.18.6211 (1985). 1252 73 Dunda, O. & Corcos, D. Recombining sequence recombina^on in normal kappa -chain-1253 expressing B cells. J Immunol 159, 4362-4366 (1997). 1254 74 Zou, Y. R., Takeda, S. & Rajewsky, K. Gene targe^ng in the Ig kappa locus: efficient genera^on 1255 of lambda chain-expressing B cells, independent of gene rearrangements in Ig kappa. EMBO J 1256 12, 811-820, doi:10.1002/j.1460-2075.1993.tb05721.x (1993). 1257 75 Daitch, L. E., Moore, M. W., Persiani, D. M., Durdik, J. M. & Selsing, E. Transcrip^on and 1258 recombina^on of the murine RS element. J Immunol 149, 832-840 (1992). 1259 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 50 76 Nadel, B., Cazenave, P . A. & Sanchez, P . Murine lambda gene rearrangements: the stochas^c 1260 model prevails over the ordered model. EMBO J 9, 435 -440, doi:10.1002/j.1460 -1261 2075.1990.tb08128.x (1990). 1262 77 Dorner, T., Foster, S. J., Farner, N. L. & Lipsky, P. E. Immunoglobulin kappa chain receptor edi^ng 1263 in systemic lupus erythematosus. J Clin Invest 102, 688-694, doi:10.1172/JCI3113 (1998). 1264 78 Casellas, R. et al. Igkappa allelic inclusion is a consequence of receptor edi^ng. J Exp Med 204, 1265 153-160, doi:10.1084/jem.20061918 (2007). 1266 79 Liu, S. et al. Receptor edi^ng can lead to allelic inclusion and development of B cells that retain 1267 an^bodies reac^ng with high avidity autoan^gens. J Immunol 175, 5067 -5076, 1268 doi:10.4049/jimmunol.175.8.5067 (2005). 1269 80 Gerdes, T. & Wabl, M. Autoreac^vity and allelic inclusion in a B cell nuclear transfer mouse. 1270 Nat Immunol 5, 1282-1287, doi:10.1038/ni1133 (2004). 1271 81 Velez, M. G. et al. Ig allotypic inclusion does not prevent B cell development or res ponse. J 1272 Immunol 179, 1049-1057, doi:10.4049/jimmunol.179.2.1049 (2007). 1273 82 Fournier, E. M. et al. Dual-reac^ve B cells are autoreac^ve and highly enriched in the 1274 plasmablast and memory B cell subsets of autoimmune mice. J Exp Med 209, 1797 -1812, 1275 doi:10.1084/jem.20120332 (2012). 1276 83 Panigrahi, A. K. et al. RS rearrangement frequency as a marker of receptor edi^ng in lupus and 1277 type 1 diabetes. J Exp Med 205, 2985-2994, doi:10.1084/jem.20082053 (2008). 1278 84 Fraser, L. D. et al. Immunoglobulin light chain allelic inclusion in systemic lupus erythematosus. 1279 Eur J Immunol 45, 2409-2419, doi:10.1002/eji.201545599 (2015). 1280 85 Lipsky, P . E. Systemic lupus erythematosus: an autoimmune disease of B cell hyperac^vity. Nat 1281 Immunol 2, 764-766, doi:10.1038/ni0901-764 (2001). 1282 86 Grammer, A. C. & Lipsky, P . E. B cell abnormali^es in systemic lupus erythematosus. Arthri@s 1283 Res Ther 5 Suppl 4, S22-27, doi:10.1186/ar1009 (2003). 1284 87 Kitaura, Y . et al. Control of the B cell-intrinsic tolerance programs by ubiqui^n ligases Cbl and 1285 Cbl-b. Immunity 26, 567-578, doi:10.1016/j.immuni.2007.03.015 (2007). 1286 88 Fritzler, M. J. Clinical relevance of autoan^bodies in systemic rheuma^c diseases. Mol Biol Rep 1287 23, 133-145, doi:10.1007/BF00351161 (1996). 1288 89 Takeda, S. et al. Dele^on of the immunoglobulin kappa chain intron enhancer abolishes kappa 1289 chain gene rearrangement in cis but not lambda chain gene rearrangement in trans. EMBO J 1290 12, 2329-2336, doi:10.1002/j.1460-2075.1993.tb05887.x (1993). 1291 90 Chen, J. et al. B cell development in mi ce that lack one or both immunoglobulin kappa light 1292 chain genes. EMBO J 12, 821-830, doi:10.1002/j.1460-2075.1993.tb05722.x (1993). 1293 91 Wong, J. B. et al. B-1a cells acquire their unique characteris^cs by bypassing the pre -BCR 1294 selec^on stage. Nat Commun 10, 4768, doi:10.1038/s41467-019-12824-z (2019). 1295 92 Hu, J. et al. Chromosomal Loop Domains Direct the Recombina^on of An^gen Receptor Genes. 1296 Cell 163, 947-959, doi:10.1016/j.cell.2015.10.016 (2015). 1297 93 Ba, Z. et al. CTCF orchestrates long -range cohesin -driven V(D)J recombina^onal scanning. 1298 Nature 586, 305-310, doi:10.1038/s41586-020-2578-0 (2020). 1299 94 Lin, S. G., Ba, Z., Alt, F. W. & Zhang, Y . RAG Chroma^n Scanning During V(D)J Recombina^on 1300 and Chroma^n Loop Extrusion are Related Processes. Adv Immunol 139, 93 -135, 1301 doi:10.1016/bs.ai.2018.07.001 (2018). 1302 95 Zhao, L. et al. Orienta^on-specific RAG ac^vity in chromosomal loop domains contributes to 1303 Tcrd V(D)J recombina^on during T cell development. J Exp Med 213, 1921 -1936, 1304 doi:10.1084/jem.20160670 (2016). 1305 96 Zhang, Y . et al. The fundamental role of chroma^n loop extrusion in physiological V(D)J 1306 recombina^on. Nature 573, 600-604, doi:10.1038/s41586-019-1547-y (2019). 1307 97 Dai, H. Q. et al. Loop extrusion mediates physiological Igh locus contrac^on for RAG scanning. 1308 Nature 590, 338-343, doi:10.1038/s41586-020-03121-7 (2021). 1309 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 51 98 Zhang, Y ., Zhang, X., Dai, H. Q., Hu, H. & Alt, F. W . The role of chroma^n loop extrusion in 1310 an^body diversifica^on. Nat Rev Immunol 22, 550 -566, doi:10.1038/s41577 -022-00679-3 1311 (2022). 1312 99 Hill, L. et al. Wapl repression by Pax5 promotes V gene recombina^on by Igh loop extrusion. 1313 Nature 584, 142-147, doi:10.1038/s41586-020-2454-y (2020). 1314 100 Morel, L. et al. Func^onal dissec^on of systemic lupus erythematosus using congenic mouse 1315 strains. J Immunol 158, 6019-6028 (1997). 1316 101 Spanopoulou, E. et al. Func^onal immunoglobulin transgenes guide ordered B -cell 1317 differen^a^on in Rag -1-deficient mice. Genes Dev 8, 1030-1042, doi:10.1101/gad.8.9.1030 1318 (1994). 1319 102 Xu, H., Li, H., Suri -Payer, E., Hardy, R. R. & Weigert, M. Regula^on of an^ -DNA B cells in 1320 recombina^on-ac^va^ng gene -deficient mice. J Exp Med 188, 1247 -1254, 1321 doi:10.1084/jem.188.7.1247 (1998). 1322 103 Halverson, R., Torres, R. M. & Pelanda, R. Receptor edi^ng is the main mechanism of B cell 1323 tolerance toward membrane an^gens. Nat Immunol 5, 645-650, doi:10.1038/ni1076 (2004). 1324 104 Hartley, S. B. et al. Elimina^on of self-reac^ve B lymphocytes proceeds in two stages: arrested 1325 development and cell death. Cell 72, 325-335, doi:10.1016/0092-8674(93)90111-3 (1993). 1326 105 Lang, J. et al. Enforced Bcl -2 expression inhibits an^gen -mediated clonal elimina^on of 1327 peripheral B cells in an an^gen dose -dependent manner and promotes receptor edi^ng in 1328 autoreac^ve, immature B cells. J Exp Med 186, 1513 -1522, doi:10.1084/jem.186.9.1513 1329 (1997). 1330 106 Strasser, A. et al. Enforced BCL2 expression in B -lymphoid cells prolongs an^body responses 1331 and elicits autoimmune disease. Proc Natl Acad Sci U S A 88, 8661 -8665, 1332 doi:10.1073/pnas.88.19.8661 (1991). 1333 107 Ait-Azzouzene, D. et al. An immunoglobulin C kappa -reac^ve single chain an^body fusion 1334 protein induces tolerance through receptor edi^ng in a normal polyclonal immune system. J 1335 Exp Med 201, 817-828, doi:10.1084/jem.20041854 (2005). 1336 108 Nussenzweig, M. C. et al. Allelic exclusion in transgenic mice that express the membrane form 1337 of immunoglobulin mu. Science 236, 816-819, doi:10.1126/science.3107126 (1987). 1338 109 Rolink, A., Kudo, A., Karasuyama, H., Kikuchi, Y. & Melchers, F. Long-term prolifera^ng early pre 1339 B cell lines and clones with the poten^al to develop to surface Ig-posi^ve, mitogen reac^ve B 1340 cells in vitro and in vivo. EMBO J 10, 327-336, doi:10.1002/j.1460-2075.1991.tb07953.x (1991). 1341 110 Krijger, P. H. L., Geeven, G., Bianchi, V., Hilvering, C. R. E. & de Laat, W . 4C-seq from beginning 1342 to end: A detailed protocol for sample prepara^on and data analysis. Methods 170, 17-32, 1343 doi:10.1016/j.ymeth.2019.07.014 (2020). 1344 111 Levin-Klein, R. et al. Clonally stable Vkappa allelic choice instr ucts Igkappa repertoire. Nat 1345 Commun 8, 15575, doi:10.1038/ncomms15575 (2017). 1346 112 Ye, J., Ma, N., Madden, T. L. & Ostell, J. M. IgBLAST: an immunoglobulin variable domain 1347 sequence analysis tool. Nucleic Acids Res 41, W34-40, doi:10.1093/nar/gkt382 (2013). 1348 113 Dobin, A. et al. STAR: ultrafast universal RNA -seq aligner . Bioinforma@cs 29, 15 -21, 1349 doi:10.1093/bioinforma^cs/bts635 (2013). 1350 114 Anders, S., Pyl, P . T. & Huber, W . HTSeq--a Python framework to work with high -throughput 1351 sequencing data. Bioinforma@cs 31, 166-169, doi:10.1093/bioinforma^cs/btu638 (2015). 1352 115 Love, M. I., Huber, W. & Anders, S. Moderated es^ma^on of fold change and dispersion for 1353 RNA-seq data with DESeq2. Genome Biol 15, 550, doi:10.1186/s13059-014-0550-8 (2014). 1354 116 Langmead, B. & Salzberg, S. L. Fast gapped-read alignment with Bow^e 2. Nat Methods 9, 357-1355 359, doi:10.1038/nmeth.1923 (2012). 1356 117 Danecek, P . et al. Twelve years of SAMtools and BCFtools. Gigascience 10, 1357 doi:10.1093/gigascience/giab008 (2021). 1358 118 Ramirez, F. et al. deepTools2: a next genera^on web server for deep-sequencing data analysis. 1359 Nucleic Acids Res 44, W160-165, doi:10.1093/nar/gkw257 (2016). 1360 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 52 119 Zhang, Y . et al. Model-based analysis of ChIP -Seq (MACS). Genome Biol 9, R137, 1361 doi:10.1186/gb-2008-9-9-r137 (2008). 1362 120 Reske, J. J., Wilson, M. R. & Chandler, R. L. ATAC -seq normaliza^on method can significantly 1363 affect differen^al accessibility analysis and interpreta^on. Epigene@cs Chroma@n 13, 22, 1364 doi:10.1186/s13072-020-00342-y (2020). 1365 121 Mazor, R. D. et al. Tumor-reac^ve an^bodies evolve from non -binding and autoreac^ve 1366 precursors. Cell 185, 1208-1222 e1221, doi:10.1016/j.cell.2022.02.012 (2022). 1367 1368 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 53 Figures 1369 Fig. 1 : CBE is essential for maintaining light chain isotype balance and enabling 1370 bidirectional transcriptional insulation. a, Schematic of the 3′-SEκ (red) with CTCF ChIP–1371 seq in IL-7–cultured WT pre-B cells. Regulatory elements and the downstream Rpia gene are 1372 shown; rsCBE and rpiaCBE (black boxes) were targeted by CRISPR–Cas9 in vivo. b, Genome 1373 browser view of the Igκ locus showing the positions of the RPIA gene and the Jκ region, along 1374 with CTCF binding profiles (GSM749762) in EBV -transformed human B lymphocytes. CBEs 1375 are highlighted by black boxes. c, Quantitative PCR analysis of 3′ -GLT , 5′-GLT and total Igκ 1376 transcripts in bone marrow–derived small pre-B cells from WT, ΔrsCBE and ΔrpiaCBE mice, 1377 normalized to Ubc and Ppia (n = 3-4). d, Quantitative PCR analysis of Rpia transcripts in small 1378 pre-B cells (n = 3 –6). e, Igκ and Ig λ isotype exclusion in B220 ⁺ splenic B cells from WT, 1379 ΔrsCBE and ΔrpiaCBE mice, expressed as percentages (n = 4). f, Ratio of Igκ⁺ to Igλ⁺ B220⁺ 1380 splenic cells. Data are presented as mean ± s.d.; P values were calculated using an unpaired 1381 t-test. 1382 a b c d e 0.007156p= p=0.003208 p=0.010749 Relative normalized expresssion 1.0 0.0 0.5 1.5 2.0 2.5 3’-Germline transcript 5’-Germline transcript Vĸ-Cĸ ∆rsCBE ∆rpiaCBE WT Small pre-B 0 2 4 6 8 75 80 85 90 95 % of light chain postive cells p=0.017112 p=0.007652 B220+ Igĸ+ B220+ Igλ1-3+B220+ Igĸ+Igλ1-3+ ∆rsCBE ∆rpiaCBE WT Splenic B p=0.0023 p=0.0008 Rpia Relative normalized expresssion 2.0 0.0 1.0 3.0 4.0 ∆rsCBE ∆rpiaCBE WT Small pre-B 0 10 20 30 40 Ratio of Igĸ+/Igλ1-3+ B cells p=0.0108 B220+ ∆rsCBE ∆rpiaCBE WT Splenic B Promoter Rpia iEκ dEκ3’-Eκ rsCBE rpiaCBEHS10 IgκJ1 IgκJ2 IgκJ4 IgκJ5Igκc CERSIS IgκJ3 3’-SEκ 70,710 70,720 70,730 70,740 70,750 70,760 70,770 70,780 70,790 Kb Chr6 0 500 CTCF motif orientation WT IL-7+ pre-B CTCF ChIP-seq Mouse f Rpia 0 120 IGKC IGKJ5 IGKJ4 IGKJ3 IGKJ2 IGKJ1 CTCF motif orientation B-Lymphocyte CTCF ChIP-seq 89,000 89,050 89,100 89,150 89,200 Kb Chr2 IGKV4-1 IGKV5-2 IGKV7-3 IGKV2-4 Human (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 54 1383 Fig. 2: CBE controls early B cell development . a, Flow cytometric quantification of bone 1384 marrow B cell populations in WT and ΔrsCBE mice, shown as percentages of the parental 1385 population; b, corresponding cell counts (in millions) (n = 4 –5). c, Heatmap showing 1386 differentially expressed genes (fold change > 1.85, adj . p < 0.1) in small pre -B cells from 1387 ΔrsCBE and WT mice. Biological replicates include three WT males and three ΔrsCBE mice 1388 (one male and two females). d, Gene ontology analysis of downregulated genes in small pre-1389 B cells of ΔrsCBE mice. e, Hoechst-based cell-cycle analysis of bone marrow B cell subsets 1390 in WT and ΔrsCBE mice; percentages of cells in cycle (G1/S/G2–M DNA content) are shown 1391 (n = 4). f, Apoptosis in B cell subsets assessed by Annexin V and PI staining; early (Annexin 1392 V⁺) and late (Annexin V⁺ PI⁺) apoptotic cells are shown as percentages (n = 4). Data are mean 1393 ± s.d.; P values were calculated using an unpaired t-test.1394 H2bc3 Aurkb Ncapg Ccnf Kif20a Aspm Birc5 Dlgap5 Ccna2 Gas2l3 Cip2a H2bc7 Kif15 Tbc1d31 Hmmr Pclaf Nusap1 Ect2 Mis18bp1 Kif14 Tpx2 H2ac10 Nuf2 H3c2 Spag5 Cep55 Polq Cdc20 Kif4 Arhgap19 Plk1 Tubb4b Anp32b−ps1 Cenpf Mki67 Fignl1 Top2a Prc1 Cenpe Spc25 Cbx5 Espl1 H2ac20 Sgo2a Bub1b Knl1 Neil3 Ckap2l H2ax Kif20b H2bc11 Kif23 Racgap1 Cdca2 Tmpo Tacc3 Fbxo5 Kif11 Kif22 Ptges3−ps Gm42048 Ndc80 Ncapd2 Bub1 H4c6 Xist Iqcn Rpia Fos Klf4 Condition Condition −1.5 −1 −0.5 0 0.5 1 1.5 ∆rsCBE WT 0.3 0.4 0.5 0.6 GeneRatio q-value 1.646462e−46 2.553965e−24 5.107930e−24 7.661894e−24 1.021586e−23 Count 15 20 25 30 35 chromosome segregation nuclear division nuclear chromosome segregation sister chromatid segregation mitotic nuclear division mitotic sister chromatid segregation spindle organization microtubule cytoskeleton organization involed in mitosis mitotic spindle organization chromosome separation a b c d e 0 20 40 60 80 0.000563p= large pre-Bsmall pre-Blmmature B pro-B % of parental population ∆rsCBE WT p=0.012732 large pre-Bsmall pre-Blmmature B pro-B Cell count (x10⁶) 0 1 2 3 4 ∆rsCBE WT f G1 64.0 G2-M 6.45 S 17.6 5.0M 10M 15M FL17-A :: Hoechst 375__450-45-A 0 10 20 30 40 50 Count ∆rsCBE WT pro-B G1 S G2-M 0 20 40 60 80 100% of cells p=0.001741 p=0.000008 ∆rsCBE WT pro-B Early apoptosis Late apoptosis % of apoptotic cells0 5 10 15 ∆rsCBE WT pro-B G1 70.3 G2-M 5.92 S 14.0 5.0M 10M 15M FL17-A :: Hoechst 375__450-45-A 0 100 200 300 Count ∆rsCBE WT Small pre-B % of cells G1 S G2-M 0 20 40 60 80 100 p=0.002478 p=0.000077 ∆rsCBE WT Small pre-B p=0.033638 % of apoptotic cells 0 2 4 6 8 Early apoptosis Late apoptosis ∆rsCBE WT Small pre-B (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 55 1395 a b d e WT IL-7+ pre-B WT-Rag1−/−.hIgM ΔrsCBE-Rag1−/−.hIgM ΔrpiaCBE-Rag1−/−.hIgM [CTCF ChIP-seq] [H3K27Ac seq] [H3K27Ac seq] [H3K27Ac seq] CTCF motif orientation 9 0 9 0 9 0 500 0 Rpia iEκ dEκ3’-Eκ rsCBE rpiaCBEHS10 IgκJ1 IgκJ2 IgκJ4 IgκJ5 IgκcCER SIS IgκJ3 rsEκ * * * * * * * c WT IL-7+ pre-B WT pre-B ex vivo ΔrsCBE pre-B ex vivo ΔrpiaCBE pre-B ex vivo [CTCF ChIP-seq] [ATAC seq] [ATAC seq] [ATAC seq] 950 0 950 0 950 0 500 0 Rpia iEκ dEκ3’-Eκ rsCBE rpiaCBEHS10 IgκJ1 IgκJ2 IgκJ4IgκJ5 IgκcCER SIS IgκJ3 rsEκ CTCF motif orientation * * ** *** * c f [Rag-2 ChIP-seq] [ATAC seq] [ATAC seq] [ATAC seq] [Rag-1 ChIP-seq] [CTCF ChIP-seq] WT pre-B ex vivo ΔrsCBE pre-B ex vivo ΔrpiaCBE pre-B ex vivo WT IL-7+ pre-B Abelson pro-B line Abelson pro-B line 0 1000 100 0 1000 90 0 90 0 500 rsCBERS * * CTCF motif orientation 1 2 3 genes and elements Start End Start End Start End 1 2 3 4 5 6 7 8CER SIS 5'-GLT 3'-GLT IgKJ iEK IgKC 3'-EK dEK RS rsCBE rsEΚ HS10 rpiaCBE WT Pre-B ΔrsCBE Pre-B ΔrpiaCBE Pre-B Signal intensity -75bp +200bp 0.0 0.2 0.4 0.6 0.8 1.0 WT Rag1−/−.hIgM ΔrsCBE Rag1−/−.hIgM ΔrpiaCBE Rag1−/−.hIgM elements0.3 0.25 0.2 0.15 Vκ RSS TES -75bp +200bpTES -75bp +200bpTES Signal intensity1.50 1.75 2.00 2.25 Start End Start End Start End 0 1 2 3 4 5 6CER SIS 5'-GLT 3'-GLT Ig�J iE� Ig�C 3'-E� dE� RS rsCBE rsEκ HS10 rpiaCBE WT Rag1−/−. hIgM ΔrsCBE Rag1−/−. hIgM ΔrpiaCBE Rag1−/−. hIgM genes and elements Signal intensity (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 56 Fig. 3: Co-ordinated molecular epigenetic mechanisms orchestrate Ig κ expression. a, 1396 Genome browser tracks across the Igκ locus showing CTCF ChIP–seq from IL-7–cultured WT 1397 pre-B cells and ATAC –seq signal from WT -Rag1⁻/⁻.hIgM, ΔrsCBE-Rag1⁻/⁻.hIgM, and 1398 ΔrpiaCBE-Rag1⁻/⁻.hIgM pre-B cells. Peaks were called with MACS2 (FDR < 0.05); differential 1399 peaks were tested using edgeR (P 1.5); asterisks denote significance 1400 (n = 2). b, Heat map of ATAC–seq signal intensity across regulatory elements within the 3′-RR 1401 of the Ig κ locus. c, Browser view of the 23 -RSS within the RS region showing CTCF, Rag1 1402 (GSM4176366) and Rag2 (GSM4176369) ChIP –seq, together with ATAC–seq signal in WT-1403 Rag1⁻/⁻.hIgM, ΔrsCBE-Rag1⁻/⁻.hIgM, and ΔrpiaCBE-Rag1⁻/⁻.hIgM pre -B cells; significa nt 1404 differences are indicated by asterisks. d, Genome browser tracks showing CTCF ChIP –seq 1405 from WT pre-B cells and H3K27ac ChIP–seq from WT-Rag1⁻/⁻.hIgM, ΔrsCBE-Rag1⁻/⁻.hIgM, 1406 and ΔrpiaCBE-Rag1⁻/⁻.hIgM pre-B cells. Peaks were identified with csaw ; differential peaks 1407 were assessed by edgeR ( FDR 1.5; n = 2). e, Heat map of H3K27ac 1408 enrichment across regulatory elements within the 3′-RR. f, Heat map of H3K27ac enrichment 1409 at 12 -RSS across V κ gene segments in WT -Rag1⁻/⁻.hIgM, ΔrsCBE-Rag1⁻/⁻.hIgM, and 1410 ΔrpiaCBE-Rag1⁻/⁻.hIgM pre-B cells, quantified within a 275-bp window (75 bp upstream and 1411 200 bp downstream of transcription end site (TES) of Vκ genes). Data are based on the mean 1412 of two biological replicates from female mice. 1413 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 57 Fig. 4: CBE controls the intricate Ig κ super-enhancer architecture through chromatin 1414 insulation. a, High-resolution (1 kb) Micro-C-TALE contact maps of the 3′-RR, smoothed with 1415 a Gaussian filter (σ = 1 kb), overlaid with ATAC–seq, H3K27ac ChIP–seq, and CTCF ChIP–1416 seq tracks. Notable interactions include SIS–rsCBE (black triangle), SIS–rpiaCBE (extended 1417 black triangle), and dEκ–rsEκ (white triangle). b, Insulation score profile plotted as a difference 1418 between ΔrsCBE-Rag1⁻/⁻.hIgM and WT -Rag1⁻/⁻.hIgM pre–B cells calc ulated with a 10 -kb 1419 window. c, Virtual 4C profiles derived from Micro-C-TALE data (1-kb resolution; Gaussian σ = 1420 1 kb) with SIS d, rsEκ, and e, RC as anchor viewpoints. f, 4C-seq interaction profiles across 1421 the 3′-RR with rsE κ, and g, RC as viewpoints in W T 445.3 and ΔrsCBE-571bp 445.3 cells. 1422 Interaction frequencies are shown as reads per million (RPM) with a 100 -bp window and 21-1423 bp running window. Data represent the mean of two biological replicates. 1424 a chr6:70695000-70825000 WT-Rag1−/−. hIgM ΔrsCBE-Rag1−/−. hIgM SIS rsCBE dEκ rsE κ rpiaCBE 70695000 70727000 70759000 70791000 70823000 CTCF H3K27acATAC 3’-E κ iEκ f g 4C Signal rsEK Promoter Rpia iEκ dEκ3’-Eκ rsCBE rpiaCBE HS10 IgJκ Igκc CER SIS WT 445.3 ΔrsCBE-571bp 445.3 CTCF binding sites 4C Signal 0 200 400 600 800 1000 * WT 445.3 ΔrsCBE-571bp 445.3 Promoter Rpia iEκ dEκ3’-Eκ rsCBE rpiaCBE HS10 IgJκ Igκc CER SIS RC 0 200 400 600 8001000 rsEK b c d e Rpia PromoteriEκ dEκ3’-Eκ rsCBE rpiaCBErsEκSIS RC Genomic positions70700kb 70814kb Insulation Score-4 -2 4 2 6 0 10 8 Difference Insulation score w=10 Promoter Rpia iEκ dEκ3’-Eκ rsCBE rpiaCBE HS10 IgJκ Igκc CER SIS WT-Rag1−/−.hIgM ΔrsCBE-Rag1−/−.hIgM rsEκ 0 3 2 1 4 5 CTCF binding sites Cumulative interaction frequency * Promoter Rpia iEκ dEκ3’-Eκ rsCBE rpiaCBE HS10 IgJκ Igκc CER SIS rsEκ WT-Rag1−/−. hIgM ΔrsCBE-Rag1−/−. hIgM 0 3 2 1 4 5Cumulative interaction frequency70700000 7074000070720000 70760000 70760000 70800000 70820000 RC Promoter Rpia iEκ dEκ3’-Eκ rsCBE rpiaCBE HS10 IgJκ Igκc CER SIS rsEκ WT-Rag1−/−.hIgM ΔrsCBE-Rag1−/−.hIgM 0 3 2 1 4 5Cumulative interaction frequency rsEκ SIS (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 58 1425 Fig. 5 : rsCBE-mediated chromatin loop extrusion control B -cell tolerance and 1426 autoimmunity. a, Relative levels of rearranged J κ, Jλ1, and V κ–RS segments in FACS -1427 purified B220 ⁺Igλ⁺ splenic B cells, quantified by qPCR and normalized to the E μ genomic 1428 region (n = 3; mean ± s.d.). b, Relative Vκ–RS recombination in Rag1 -proficient WT 445.3, 1429 ΔrsCBE-571 bp 445.3, and ΔrsCBE-18 bp 445.3 clones, measured by qPCR and normalized 1430 to the Eμ genomic region (n = 3). P values were determined by unpaired t-test. c, Cumulative 1431 frequency distributions (CFDs) from rsCBE vie wpoints showing cumulative 4C signals in 1432 ΔrsCBE-571 bp, ΔrsCBE-18 bp, and ΔSIS 445.3 cells relative to WT 445.3 (p < 0.0001, Mann–1433 Whitney test). Data represent the mean of two independent biological replicates. d, Allelic 1434 exclusion in bone marrow IgM⁺ B cells from Igκm/h and IgκΔrsCBE-m/h mice, shown as percentages 1435 of mCκ⁺, hCκ⁺, and dual mCκ⁺hCκ⁺ cells (n = 4). e, Allelic exclusion in splenic IgM ⁺ B cells 1436 from Igκm/h and IgκΔrsCBE-m/h mice, shown as percentages (n = 4). Data are mean ± s.d.; P values 1437 were determined by unpaired t -test. f, Serum IgG autoantibodies against nuclear antigens 1438 (NA), DNA, LPS, and insulin quantified by ELISA in 14-month-old mice (n = 4–10) and g, 10-1439 week-old mice (n = 4–5), measured by ELISA. Data are mean ± s.d.; P values were determined 1440 by unpaired t-test. 1441 445.3-WT 445.3-ΔrsCBE-571bp 445.3-ΔrsCBE-18bp 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 Relative levels Vĸ-RS p=0.0207 p=0.0203 a b c d e f gAbsorbance (650nm) 0.0 0.5 1.0 anti-NA anti-DNA anti-LPS Anti-Insulin ∆rsCBE WTp=0.045253 p=0.042731 10 weeks old mice Absorbance (650nm) 0.0 0.5 1.0 anti-NA anti-DNA anti-LPS Anti-Insulin 1.5 p=0.003553 p=0.001640 ∆rsCBE WT 14 months old mice 0 1 2 3 4 5 40 45 50 % of IgM+ cells in bone marrow mC ĸ+ hC ĸ+ mC ĸ+ hC ĸ+ p=0.042120 IgκΔrsCBE-m/h Igκm/h p=0.010933 mC ĸ+ hC ĸ+ mC ĸ+ hC ĸ+ 25000 20000 15000 10000 5000 Reads/ million 0 Igkv9-120 Igkv20-101-2 Igkv1-108 Igkv14-126-1 Igkv1-115 Igkv1-133 Igkv13-85 Igkv4-63 Igkv4-70 Igkv4-92 Igkv4-80 Igkv10-96 Igkv12-49 Igkv13-55-1 Igkv4-58 Igkv13-74-1 Igkv8-24 Igkv8-31 Igkv12-42 Igkv3-4 Igkv3-11 Igkv6-17 Igkv18-36 WT 445.3 ΔrsCBE-571bp 445.3 ΔrsCBE-18bp 445.3 ΔSIS 445.3 Viewpoint rsCBE p<0.0001 p<0.0001 p<0.0001 mC ĸ+ hC ĸ+ mC ĸ+ hC ĸ+ IgκΔrsCBE-m/h Igκm/h mC ĸ+ hC ĸ+ mC ĸ+ hC ĸ+ 0 1 2 3 40 45 50 % of IgM+ Splenic B cells p=0.001778 0.0 0.5 1.0 1.5 2.0 Relative levels Vĸ-Jĸ1 Vĸ-Jĸ5 Vλ-Jλ1Vĸ-RS ∆rsCBE ∆rpiaCBE WT p=0.013420 p=0.000057 Splenic B (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 59 Extended Data Figures 1442 Lymphocytes 16.6 0 500K 1.0M 1.5M 2.0M 0 500K 1.0M 1.5M 2.0M CD19+ 82.3 0 104 105 106 107 0 500K 1.0M 1.5M 2.0M CD19+B220+IgM- 59.4 Immature B 18.6 0 104 105 106 107 0 -103 103 104 105 106 107 Q1 85.5 Q2 2.79 Q3 9.57 Q4 2.13 0 104 105 106 107 0 -103 103 104 105 106 107 Lymphocytes 14.2 0 500K 1.0M 1.5M 2.0M 0 500K 1.0M 1.5M 2.0M CD19+B220+IgM- 51.4 Immature B 17.0 0 104 105 106 107 0 -103 103 104 105 106 107 Q1 89.5 Q2 2.84 Q3 6.25 Q4 1.42 0 104 105 106 107 0 -103 103 104 105 106 107 CD19+ 76.5 0 104 105 106 107 0 500K 1.0M 1.5M 2.0M WTΔrsCBE SSC-A SSC-H B220 PerCP-Cy5.5 Igκ+ PE Bone marrow FSC-A CD19 PE-Cy-7 IgM-e405 Igλ1-3+ FITC Immature B WT ΔrpiaCBEΔrsCBE Spleen Q1 80.1 Q2 7.14 Q3 5.09 Q4 7.71 10 2 10 3 10 4 10 5 10 6 10 2 10 3 10 4 10 5 10 6 10 7 Q1 81.3 Q2 4.88 Q3 5.50 Q4 8.37 10 2 10 3 10 4 10 5 10 6 10 2 10 3 10 4 10 5 10 6 10 7 Q1 86.3 Q2 5.19 Q3 2.89 Q4 5.64 10 2 10 3 10 4 10 5 10 6 10 2 10 3 10 4 10 5 10 Igλ1-3+ FITC-488 Igκ+ PE a b c d e f g h i 0.05 0 0.1 0.05Vĸ gene proportion 0.1 WT ΔrsCBE n=5 n=5 ** Small pre-B 0.05 0 0.1 0.05Vĸ gene proportion WT ΔrpiaCBE n=5 n=3 Small pre-B p=0.006573 p=0.046835 p=0.000763 p=0.024168 p=0.049491 1.0 0.0 0.5 1.5 2.0Relative level Vĸ-Jĸ1 Vĸ-Jĸ2 Vĸ-Jĸ4 Vĸ-Jĸ5 ∆rsCBE ∆rpiaCBE WT Small pre-B ∆rsCBE WT p=0.000107 p=0.049232 Relative normalized expresssion 2.0 0.0 1.0 3.0 4.0 Vĸ-Cĸ Rpia pro-B 0.0 0.2 0.4 0.6 0.8 5 10 15 B220+ Igĸ+ B220+ Igλ1-3+B220+ Igĸ+Igλ1-3+ Cell count (x10⁶) ∆rsCBE ∆rpiaCBE WT Splenic B p=0.000301 p=0.000099 0.0 5 10 70 80 90 100 Igĸ+ Igλ1-3+ Igĸ+Igλ1-3+ Immature B % of light chain positive cells ∆rsCBE WT 0.01 0.1 1 p=0.011215 Igĸ+ Igλ1-3+ Igĸ+Igλ1-3+ Immature B Cell count (x10⁶) ∆rsCBE WT 10 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 60 Extended Data Fig. 1: CBE is essential for maintaining light chain isotype balance and 1443 enabling bidirectional transcriptional insulation. a, b, Vκ gene rearrangement 1444 frequencies in RNA from bone marrow–derived small pre-B cells of ΔrsCBE and ΔrpiaCBE 1445 mice. Error bars represent s.d. for each Vκ gene. Statistical significance was determined by 1446 t-test (p-adj < 0.001, ***; p-adj < 0.01, **; p-adj < 0.05, *). c, Relative levels of rearranged Jκ 1447 segments in small pre-B cells from WT, ΔrsCBE and ΔrpiaCBE mice, measured by qPCR 1448 and normalized to the Eμ region (n = 5–6). d, Rpia and total Igκ transcript levels in bone 1449 marrow–derived pro-B cells from WT and ΔrsCBE mice, measured by qPCR and normalized 1450 to Ubc and Ppia (n = 3). e, Igκ and Igλ isotype exclusion in IgM⁺ B cells from WT and 1451 ΔrsCBE mice, shown as percentages; f, corresponding cell counts (in millions) (n = 5–6 mice 1452 per genotype). Absolute numbers of each B cell subset were quantified from the femur, tibia, 1453 and fibula of each mouse. g, Gating strategy for Igκ⁺ and Igλ⁺ Immature B cells in bone 1454 marrow from WT and ΔrsCBE mice; values shown as percentages in plots. h, Igκ and Igλ 1455 isotype exclusion in B220⁺ splenic B cells from WT , ΔrsCBE and ΔrpiaCBE mice, shown as 1456 cell counts (in millions) (n = 4). Data are mean ± s.d.; P values were calculated by unpaired 1457 Student’s t-test. Absolute numbers of each B cell subset were quantified from the femur, 1458 tibia, and fibula of each mouse. i, Representative gating strategy of B220⁺ splenic B cells for 1459 Igκ⁺ and Igλ⁺ populations, with percentages indicated in each quadrant. 1460 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 61 1461 Extended Data Fig. 2 : CBE controls early B cell development . a, Representative gating 1462 strategy for identifying bone marrow B cell subsets in WT and ΔrsCBE mice. b, Flow cytometric 1463 quantification of bone marrow B cell populations in WT and ΔrpiaCBE mice (n = 4), presented 1464 as percentages of the parental population. c, Hoechst-based cell -cycle analysis of bone 1465 marrow B cell subsets in WT and ΔrsCBE mice; percentages of cells in cycle (G1/S/G2 –M 1466 DNA content) are shown (n = 4). d, Apoptosis in B cell subsets assessed by Annexin V and PI 1467 staining; early (Annexin V ⁺) and late (Annexin V ⁺ PI⁺) apoptotic cells are presented as 1468 percentages (n = 4). Data are mean ± s.d.; P values were determined using an unpaired t -1469 test. e, Flow cytometric quantification of splenic B cell subsets in WT, ΔrsCBE and ΔrpiaCBE 1470 mice (n = 8–10), presented as percentages of the parental population (mean ± s.d.); P values 1471 were determined using an unpaired t-test. 1472 CD19+ 16.9 0 104 105 106 0 500K 1.0M 1.5M 2.0M CD19+B220+IgM- 61.4 Immature B 9.17 102 103 104 105 106 0 -103 103 104 105 106 107 CD19+B220+IgM- 62.3 Immature B 10.7 102 103 104 105 106 0 -103 103 104 105 106 107 large pre-B 20.5 pro-B 18.3 small pre-B 56.4 103 104 105 0 500K 1.0M 1.5M 2.0M large pre-B 20.5 pro-B 10.8 small pre-B 64.4 3 104 105 0 500K 1.0M 1.5M 2.0M CD19+ 13.1 104 105 106 0 500K 1.0M 1.5M 2.0M WTΔrsCBE B220-PerCP-Cy5.5 FSC-A CD19-APC IgM-e405 CD43-PE SSC-A a b d c large pre-Bsmall pre-Blmmature B pro-B 0 20 40 60 80 % of parental population ∆rpiaCBE WT ∆rsCBE ∆rpiaCBE WT FoB MZB T1BB220+ % of parental population 0 5 60 80 10 40 e Early apoptosis Late apoptosis % of apoptotic cells 0 5 10 15 ∆rsCBE WT large pre-B Early apoptosis Late apoptosis % of apoptotic cells 0 5 10 15 p=0.016219 ∆rsCBE WT Immature B G1 S G2-M 0 20 40 60 80 100% of cells p=0.018543 p=0.018581 ∆rsCBE WT large pre-B G1 83.1 G2-M 1.47 S 4.57 5.0M 10M 15M FL17-A :: Hoechst 375__450-45-A 0 20 40 60 Count ∆rsCBE WT large pre-B G1 S G2-M 0 20 40 60 80 100 p=0.002509 p=0.000562 p=<0.000001 % of cells ∆rsCBE WT Immature B G1 71.5 G2-M 7.11 S 14.5 5.0M 10M 15M FL17-A :: Hoechst 375__450-45-A 0 50 100 150 Count ∆rsCBE WT Immature B (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 62 1473 Extended Data Fig. 3: Co-ordinated molecular epigenetic mechanisms orchestrate Igκ 1474 expression. a, Heat map of H3K27ac enrichment across V κ gene segments in WT -1475 Rag1⁻/⁻.hIgM, ΔrsCBE-Rag1⁻/⁻.hIgM, and ΔrpiaCBE-Rag1⁻/⁻.hIgM pre -B cells. Signal 1476 intensity was quantified within a 2-kb window spanning 1 kb upstream of the transcription start 1477 site (TSS) to 1 kb downstream of the TES. Data are averages of two biological replicates. 1478 a 0.1 0.2 0.3 0.4 genes -1.0 TSS TES 1Kb 0.0 0.2 0.4 0.6 0.8 1.0 1.2 WT Rag1−/−. hIgM ΔrsCBE Rag1−/−. hIgM ΔrpiaCBE Rag1−/−. hIgM Variable Igκ genes -1.0 TSS TES 1Kb -1.0 TSS TES 1Kb -1.0 TSS TES 1Kb -1.0 TSS TES 1Kb -1.0 TSS TES 1Kb Signal intensity (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 63 1479 Extended Data Fig. 4 : CBE controls the intricate Ig κ super-enhancer architecture 1480 through chromatin insulation. a, Micro-C-TALE contact maps across the Ig κ locus in WT-1481 Rag1⁻/⁻.hIgM and ΔrsCBE-Rag1⁻/⁻.hIgM pre–B cells at 5-kb resolution and smoothed with a 1482 Gaussian filter ( σ = 5 kb). b, Virtual 4C interaction profiles from Micro -C-TALE data (1 -kb 1483 resolution, Gaussian smoothing, σ = 1 kb) in WT -Rag1⁻/⁻.hIgM and ΔrsCBE-Rag1⁻/⁻.hIgM 1484 pre–B cells with rsCBE c, 3′ -Eκ, and d, dEκ as anchor viewpoints. e, 4C-seq interaction 1485 profiles across the 3 ′-RR with dE κ as the viewpoint in WT 445.3 and ΔrsCBE-571bp 445.3 1486 cells. Interaction frequencies are shown as reads per million (RPM) with a 100-bp window and 1487 21-bp running window. Data represent the mean of two biological replicates. 1488 Promoter Rpia iEκ dEκ3’-Eκ rsCBE rpiaCBE HS10 IgJκ Igκc CER rsEκSIS rsCBE 0 3 2 1 4 5Cumulative interaction frequency CTCF binding sites WT-Rag1−/−.hIgM ΔrsCBE-Rag1−/−.hIgM b c d Promoter Rpia iEκ dEκ3’-Eκ rsCBE rpiaCBE HS10CER rsEκSIS 3’-Eκ 0 3 2 1 4 5Cumulative interaction frequency WT-Rag1−/−.hIgM ΔrsCBE-Rag1−/−.hIgM IgJκ Igκc Promoter Rpia iEκ 3’-Eκ rsCBE rpiaCBE HS10 IgJκ Igκc CER rsEκSIS dEκ 70700000 7074000070720000 70760000 70760000 70800000 70820000 0 3 2 1 4 5Cumulative interaction frequency WT-Rag1−/−.hIgM ΔrsCBE-Rag1−/−.hIgM * e Promoter Rpia iEκ dEκ3’-Eκ rsCBE rpiaCBE HS10 IgJκ Igκc CER SIS dEκ CTCF binding sites 4C Signal 0 200 400 600 8001000 WT 445.3 ΔrsCBE-571bp 445.3 rsEK CTCF H3K27acATAC WT-Rag1−/−. hIgM ΔrsCBE-Rag1−/−. hIgM 67470000 68630000 69790000 Igκ 3’-RR a (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 64 Extended Data Fig. 5 : rsCBE-mediated chromatin loop extrusion governs B -cell 1489 tolerance and autoimmunity. a, CTCF occupancy at the rsCBE region in Rag1-deficient WT 1490 445.3 and ΔrsCBE–18 bp 445.3 single-cell clones, measured by ChIP–qPCR and expressed 1491 as fold enrichment. The murine Kifc5b locus served as a positive control for CTCF occupancy, 1492 while Hsf2bp was used as a negative control lacking CTCF binding. Data are presented as 1493 mean ± s.d. (n = 2 –3). Statistical significance was determined using a n unpaired t-test; p < 1494 0.05 was considered significant. b, 4C-seq interaction profiles across the Ig κ locus in Rag1-1495 deficient WT 445.3, ΔrsCBE-571 bp 445.3, ΔrsCBE-18 bp 445.3, and c, ΔSIS 445.3 cells. 1496 Interaction frequencies are shown as reads per million (RPM) using a 100 -bp window with a 1497 Igκ ΔrsCBE-m/h Igκ m/h Balbc C57B6 hCκ rpiaCBErsCBEmCκJκ Q1 42.5 Q2 3.11 Q3 46.8 Q4 7.56 100 102 104 106 100 101 102 103 104 105 106 107 Q1 44.9 Q2 4.09 Q3 45.7 Q4 5.29 100 102 104 106 100 101 102 103 104 105 106 107 Bone MarrowSpleen Q1 48.0 Q2 2.38 Q3 44.4 Q4 5.21 0 104 105 106 107 0 -103 103 104 105 106 107 Q1 49.0 Q2 2.23 Q3 43.3 Q4 5.45 0 104 105 106 107 0 -103 103 104 105 106 107 WT ΔrsCBE hCκ-APC mCκ-PE p=0.0201 Spleen 0.05 0.10 0.15 0.20 0.00 weight (gm) ∆rsCBE WT ∆rsCBEWT ∆rsCBE WT % of parental population 0 20 80 100 40 60 FoB MZB T1BB220+ b c d e f g a REIR Promoter Rpia iEκ dEκ3’-Eκ rpiaCBE HS10 IgJκ Igκc CER SIS Igκv3-4 Igκv3-3 Igκv3-2 Igκv3-1 WT 445.3 ΔrsCBE-571bp 445.3 ΔrsCBE-18bp 445.3 rsCBE viewpoint interaction 0 200 400 600 800 1000 CTCF binding sites rsCBE Igkv2−137 Igkv1−136 Igkv14−134−1 Igkv1−132 Igkv14−130 Igkv9−129Igkv17−127 Igkv14−126 Igkv9−124 Igkv1−122 Igkv9−120 Igkv14−118−1Igkv11−118 Igkv2−116 Igkv2−113 Igkv2−112 Igkv14−111 Igkv2−109 Igkv2−107Igkv2−105 Igkv16−104 Igkv15−102 Igkv14−100Igkv1−99 Igkv12−98 Igkv10−96 Igkv2−95−1 Igkv10−94 Igkv19−93 Igkv4−91 Igkv4−90 Igkv12−89 Igkv1−88 Igkv13−87 Igkv13−85 Igkv4−81 Igkv4−80 Igkv4−79 Igkv4−77 Igkv13−76 Igkv13−74−1 Igkv4−72 Igkv4−70 Igkv4−68 Igkv12−66 Igkv13−64 Igkv13−62−1 Igkv4−61Igkv4−59 Igkv4−60 Igkv4−58Igkv13−57−2 Igkv13−57−1 Igkv13−55−1 Igkv4−54 Igkv4−51 Igkv12−49 Igkv12−47 Igkv5−45 Igkv12−44 Igkv12−41 Igkv5−39 Igkv12−38 Igkv18−36 Igkv8−34 Igkv6−32 Igkv8−31 Igkv6−29 Igkv8−27Igkv8−26Igkv6−25 Igkv6−23 Igkv8−22 Igkv6−20 Igkv6−17 Igkv6−15 Igkv6−14Igkv6−13 Igkv3−12−1 Igkv3−11 Igkv3−9 Igkv3−6 Igkv3−4 Igkv3−2 rsCBE viewpoint interaction 0 50 100 150 200 WT 445.3 ΔrsCBE-571bp 445.3 ΔrsCBE-18bp 445.3 68000000 70000000 rsCBE CTCF binding sites Igkv2−137 Igkv1−136 Igkv14−134−1 Igkv1−132 Igkv14−130 Igkv9−129Igkv17−127 Igkv14−126 Igkv9−124 Igkv1−122 Igkv9−120 Igkv14−118−1Igkv11−118 Igkv2−116 Igkv2−113 Igkv2−112 Igkv14−111 Igkv2−109 Igkv2−107Igkv2−105 Igkv16−104 Igkv15−102 Igkv14−100Igkv1−99 Igkv12−98 Igkv10−96 Igkv2−95−1 Igkv10−94 Igkv19−93 Igkv4−91 Igkv4−90 Igkv12−89 Igkv1−88 Igkv13−87 Igkv13−85 Igkv4−81 Igkv4−80 Igkv4−79 Igkv4−77 Igkv13−76 Igkv13−74−1 Igkv4−72 Igkv4−70 Igkv4−68 Igkv12−66 Igkv13−64 Igkv13−62−1 Igkv4−61Igkv4−59 Igkv4−60 Igkv4−58Igkv13−57−2 Igkv13−57−1 Igkv13−55−1 Igkv4−54 Igkv4−51 Igkv12−49 Igkv12−47 Igkv5−45 Igkv12−44 Igkv12−41 Igkv5−39 Igkv12−38 Igkv18−36 Igkv8−34 Igkv6−32 Igkv8−31 Igkv6−29 Igkv8−27Igkv8−26Igkv6−25 Igkv6−23 Igkv8−22 Igkv6−20 Igkv6−17 Igkv6−15 Igkv6−14Igkv6−13 Igkv3−12−1 Igkv3−11 Igkv3−9 Igkv3−6 Igkv3−4 Igkv3−2 68000000 70000000 rsCBE CTCF binding sites rsCBE viewpoint interaction 0 50 100 150 200 WT 445.3 ΔSIS 445.3 rsCBE viewpoint interaction 0 200 400 600 800 1000 REIR WT 445.3 ΔSIS 445.3 Promoter Rpia iEκ dEκ3’-Eκ rpiaCBE HS10 IgJκ Igκc CER SIS rsCBE CTCF binding sites Igκv3-4 Igκv3-3 Igκv3-2 Igκv3-1 c 0 1 2 3 445.3-WT 445.3-ΔrsCBE-18bp Hsf2bp Kifc5b rsCBE-18bp p=0.002926 Fold enrichement 4 5 6 7 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint 65 21-bp running window. The anchor denotes the rsCBE viewpoint, and a magnified view of the 1498 3′-RR (dotted line) shows mean interaction profiles from two independent biological replicates. 1499 d, Schematic of mCκ⁺ and hCκ⁺ alleles in Igκm/h and IgκΔrsCBE-m/h hybrid mice. e, Representative 1500 flow cytometric gating of bone marrow and splenic IgM ⁺ B cells in hybrid mice, showing the 1501 proportions of mCκ⁺, hCκ⁺, and dual-mCκ⁺hCκ⁺ B cells within the gated populations. Values 1502 are shown as percentages in each quadrant. f, Flow cytometric quantification of splenic B cell 1503 populations in 14-month-old mice (n = 3), shown as percentages of the parental population. 1504 g, Splenic weight and images from 14-month-old mice (n = 3). P values were determined by 1505 unpaired t-test. 1506 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.12.717890doi: bioRxiv preprint