Loop Extrusion Mediates Physiological IgH Locus Contraction For RAG Scanning

Abstract

RAG endonuclease initiates IgH V(D)J recombination in pro-B cells by binding a J_H-recombination signal sequence (RSS) within a recombination center (RC) and then linearly scanning upstream chromatin, presented by cohesin-mediated loop extrusion, for convergent D-RSSs^1,2. Utilization of convergently-oriented RSSs and cryptic RSSs is intrinsic to long-range RAG scanning3. RAG scanning from the DJ_H-RC-RSS to upstream convergent V_H-RSSs is impeded by D-proximal CTCF-binding elements (CBEs)^2–5. Primary pro-B cells undergo a mechanistically-undefined V_H locus contraction proposed to provide distal V_Hs access to the DJ_H-RC^6–9. Here, we report that a 2.4 mega-base V_H locus inversion in primary pro-B cells abrogates rearrangement of both V_H-RSSs and normally convergent cryptic RSSs, even though locus contraction still occurs. In addition, this inversion activated both utilization of cryptic V_H-locus RSSs normally in opposite orientation and RAG scanning beyond the V_H locus through multiple convergent-CBE domains to the telomere. Together, these findings imply that broad deregulation of CBE impediments in primary pro-B cells promotes loop extrusion-mediated RAG V_H locus-scanning. We further found that expression of Wapl10, a cohesin-unloading factor, is low in primary pro-B cells versus v-Abl-transformed pro-B lines that lack contraction and RAG-scanning of the V_H locus. Correspondingly, Wapl depletion in v-Abl-tranformed lines activated both processes, further implicating loop extrusion in the locus contraction mechanism.

Introduction of ectopic RCs into non-antigen receptor loci across v-Abl-transformed pro-B cell lines (hereafter “v-Abl cells”) genomes allows RAG to scan convergent CBE-anchored chromatin domains until terminated at convergent CBE loop anchors³. During such scanning, orientation of initiating bona fide RSSs in ectopic RCs programs RC-bound RAG to scan, over mega-base (Mb) distances, flanking chromatin presented by cohesin-mediated loop extrusion in one direction or the other³. In this context, RAG utilizes 100s of cryptic RSSs, as short as the conserved first 3 bp CAC motif of a bona fide RSS, when in convergent orientation with the initiating bona fide RSS of the ectopic RC³. Cryptic RSSs in the same orientation as the RC-RSS (e.g. GTGs) are not used unless they lie very close (within 20kb) to the RC, allowing access by diffusion^1,3. Additional v-Abl cell studies demonstrated loop extrusion-mediated RAG scanning in promoting orientation-specific D to J_H joining in the downstream IgH loop domain¹. Thus, chromosomal orientation of J_H-RSSs in the IgH RC programs RC-bound RAG to scan upstream D-containing chromatin; and, even though Ds have downstream and upstream bona fide RSSs in opposite orientations, only downstream D-RSSs convergently-oriented with initiating J_H-RSSs are robustly used for D-to-J_H joining¹.

While all V_H-RSSs are in convergent orientation with the upstream DJ_H-RC-RSS³, their RC scanning access at the D-to-J_H rearrangement stage is impeded by CTCF-bound CBEs in intergenic control region 1 (IGCR1) upstream of the most distal D segment^3–5. Inactivation of IGCR1 CBEs allows scanning of the most D-proximal V_Hs, during which their RSS-associated CBEs promote robust rearrangement and impede further upstream scanning⁵. 100s of CBEs, mostly not closely-associated with V_Hs, lie throughout the upstream V_H locus and may also impede scanning if CTCF-bound². Distal V_Hs are thought to access the DJ_H-RC-RSS in primary pro-B cells by diffusion upon gaining proximity via V_H locus contraction^6–9. To test potential contributions of loop extrusion-mediated RAG-scanning to locus contraction and long-range V_H utilization in primary pro-B cells, we inverted the entire 2.4 Mb V_H locus upstream of D-proximal V_H81X in the genome of C57BL/6 mice; V_H81X was left in normal position and orientation as a control (Fig. 1a; Extended Data Fig. 1). The V_H locus contains 109 functional V_Hs within four V_H domains from D-proximal to distal^7,11,12 (Fig. 1a, top; Supplementary Table 1). The V_H locus inversion placed the distal J558/3609 V_Hs in the proximal V_H location and the proximal V_H7183/Q52 locus, other than V_H81X, in the distal location, with all 108 functional V_Hs and their RSSs across the locus inverted in orientation relative to the DJ_H-RC RSS (Fig. 1a middle, bottom). Based on our sensitive HTGTS-V(D)J-seq assay^13,14, this inversion essentially abrogated rearrangement of all V_H-RSSs within the inverted region in primary bone marrow pro-B cells, with utilization of normally-oriented and positioned V_H81X doubling, perhaps due to lack of competition from upstream V_Hs⁵ (Fig. 1b, top versus bottom panel; Supplementary Table 1).

Figure. 1 |. A 2.4 Mb inversion of the V_H locus nearly abrogates rearrangement of V_Hs within it in primary pro-B cells.

a, Schematic of the strategy of targeted IgH V_H locus inversion. Diagram of the murine IgH locus showing the first two proximal V_Hs (V_H81X and V_H2–2), the last two distal V_Hs (V_H1–86P and V_H1–85), Ds, J_Hs, C_Hs, and regulatory elements as indicated (not to scale), with RC that comprises J_H-proximal DQ52 segment, four J_H segments, and the intronic enhancer (iEμ) highlighted. All V_H segments are indicated and divided into four V_Hs domains (7183/Q52, Middle, J558 and J558/3609) from most proximal to distal. Yellow and dark orange triangles represent position and orientation of bona fide 23RSS and 12RSS, respectively. Purple and pink trapezoids represent position and orientation of CBEs. Green arrows denote the J_H1–4 coding end bait primers used for generating HTGTS-V(D)J-seq libraries. Cut1 and Cut2 show the location of 2 sgRNAs. Tel denotes telomere. Cen denotes centromere. b, Average utilization frequencies ± s.d. of all V_Hs in WT (top) and IgH V_H inversion (bottom) primary pro-B cells. Average percentage ± s.d. of V_HDJ_H and DJ_H rearrangements are shown. For comparison, the V_H usage data in the inverted locus is shown in inverted orientation (b, bottom), and several highly utilized V_Hs in each V_H domain are indicated. n, number of independent experiments. See Supplementary Table 1 and Methods for more details.

While abrogation of all bona fide V_H-RSS-mediated rearrangements within the inverted V_H locus strongly implicates RAG chromatin scanning in their utilization, such an inversion could conceivably prevent locus contraction, or in some other way, impede V_H-RSS access to the DJ_H-RC-RSS via diffusion. Therefore, we analyzed RAG utilization of cryptic RSSs across normal and inverted V_H loci in primary pro-B cells (Fig. 2a, top). These analyses demonstrated that cryptic RSSs across the length of the normal 2.4 Mb V_H locus in convergent orientation with the DJ_H-RC-RSS were utilized for RAG-mediated joining to the DJ_H complex, but those in the same orientation as the DJ_H-RC-RSS were not, further supporting linear scanning of the V_H locus by RC-bound RAG in primary pro-B cells (Fig. 2b, top and 2c; Extended Data Fig.2a, top, and 2b). Moreover, the V_H-inverted locus provided an even more stringent test of potential RAG scanning contributions. Thus, long-range RAG V_H locus scanning predicts that cryptic RSSs across the normal V_H locus in the same orientation as the DJ_H-RC-RSS that are not utilized by RAG scanning would be activated as scanning targets when placed in convergent orientation with the DJ_H-RC-RSS by the V_H locus inversion (Fig. 2a, bottom). Strikingly, this was the case (Fig. 2b, bottom and 2c; Extended Data Fig. 2a, bottom, 2c). These findings strongly support a major role for cohesin-mediated loop extrusion-driven RAG chromatin scanning of the upstream V_H locus in generating V_H to DJ_H rearrangements in primary pro-B cells.

Figure. 2 |. V(D)J recombination of cryptic RSSs in normal and inverted IgH loci in primary pro-B cells.

a, Illustration of possible joining outcomes between bona fide IgH locus RSSs from J_H1–4 coding-end baits to cryptic RSSs mostly represented by CAC motifs in the upstream D, V_Hs and domains upstream of V_Hs to the telomere in WT (top) and IgH V_H inversion (bottom) primary pro-B cells. Red and blue arches with arrows show possible deletional and inversional junctions, respectively, except for the IgH V_H inverted region (red for inversional, blue for deletional). Green arrows indicate the position and orientation of HTGTS primers. Details as shown in Fig. 1a. b, RAG off-target junction profiles at IgH locus and upstream 4 Mb region in WT and IgH V_H inversion primary pro-B cells. Junctions are displayed in a linear scale as stacked tracks. I-VI peaks indicated by pink lines are the RAG off-target examples plotted at single-base-pair resolution in Extended Data Fig. 2a. For comparison, RAG off-target junction profiles in the inverted region are shown in normal orientation (b, bottom). The IgH V_H inverted region is highlighted in transparent green (a, b, WT) and transparent gold (a, b, IgH V_H inversion), respectively. Del (+) indicates deletional junction. Inv (−) indicates inversional junction. c, Average frequencies ± s.d. of plus strand (red, +) and minus strand (blue, -) joining events within indicated regions in wide-type (WT, n=3) and IgH V_H inversion (INV, n=3) primary pro-B cells. n, number of independent experiments. Indicated P values were calculated using unpaired two-tailed t-test. See Methods and related Extended Data Fig. 2 for more details.

To examine effects of the 2.4 Mb inversion on interactions of V_H locus sequences with the RC, we performed high-resolution 3C-HTGTS⁵ with a RC bait in RAG1-deficient cultured primary pro-B cells with or without the inversion (Fig. 3a). RAG-deficient cells are used to prevent confounding V(D)J recombination events⁵. These studies revealed a series of robust RC interaction peaks across the V_H locus (Fig. 3b, labeled 1–15), many of which, including those involving Pax5-activated intergenic repeat (PAIR) elements¹⁵ (Fig. 3b, purple lines with gray background), are hallmarks of V_H locus contraction⁷. Notably, all major RC-V_H locus interaction peaks were maintained within the inverted V_H locus, despite the several Mb exchange in distance from the RC between proximal and distal portions of the locus (Fig. 3b, also labeled 1–15; Extended Data Fig. 3). Indeed, overall RC interactions within the different regions of the inverted locus were essentially mirror-image patterns of the normally-oriented locus (Fig. 3b; Extended Data Fig. 3). Cohesin and CTCF binding patterns determined by ChIP-seq and germline V_H transcription patterns determined by GRO-seq across the different V_H domains were maintained in detail in the inverted V_H-locus, also largely as mirror images of those of the normal V_H-locus (Extended Data Fig. 3,4). Notably, the majority of the major RC interaction peaks corresponded to peaks of sense/antisense V_H locus transcription maintained in the inverted V_H locus with orientation relative to the RC reversed (Fig. 3b; Extended Data Fig. 3). Thus, despite causing major alterations in V_H-RSS and cryptic RSS utilization, the 2.4 Mb inversion did not appear to disrupt V_H locus contraction, major looping factor binding patterns, and transcription per se, supporting the role of long-range RAG scanning in V_H utilization.

Figure. 3 |. 3C-HTGTS interactions in normal and inverted IgH loci in cultured RAG1-deficient primary pro-B cells.

a, Diagram of the entire murine IgH locus with details as shown in Fig. 1a. The primer regions of RC/iEμ bait and 3’CBE bait used for 3C-HTGTS are shown. b, Average RC/iEμ baited 3C-HTGTS signal counts ± s.e.m. across the four V_H domains in cultured WT (top) and IgH V_H inversion (bottom) RAG1-deficient primary pro-B cells. 15 representative major interaction peaks/clusters are indicated with color shades and numbers. PAIR elements are indicated as purple lines with gray background. n indicates the number of biological repeats. c, Average 3’CBE baited 3C-HTGTS signal counts ± s.e.m. across the four V_H domains in cultured WT (top) and IgH V_H inversion (bottom) RAG1-deficient primary pro-B cells. n indicates the number of biological repeats. The WT and inverted V_H locus/domains are diagrammed at the top of each panel (b, c). For comparison, the 3C-HTGTS results in the inverted region are shown in inverted orientation (b, bottom; c, bottom), and the highly utilized V_Hs in Fig. 1b are labeled at each panel (b, c). See Supplementary Data 1 for individual repeats of RC or 3’CBE interactions with entire IgH locus.

In contrast to our findings that, in v-Abl cells^1,2,5 and primary pro-B cells (see above), the RC serves as a major downstream loop extrusion anchor for RAG scanning of upstream chromatin, CBE loop anchors are proposed to play this role in primary pro-B cells, including 10 consecutive CBEs downstream of the IgH locus (3’IgH CBEs) in convergent orientation with V_H locus CBEs^8,16. Based on 3C-HTGTS analyses, we confirmed that 3’IgH CBEs interact with convergent CBEs across the V_H locus in primary pro-B cells (Fig. 3c, upper). Our V_H locus inversion places most CBEs across the V_H locus in the same orientation as 3’IgH CBEs. However, while 3’IgH CBE interaction levels with proximal V_H domain CBEs were diminished when inverted in the distal position; 3’IgH CBE interaction levels with CBEs of the CBE-rich distal V_H domain were mostly maintained or even increased (Fig. 3c, bottom). Based on this unanticipated finding, we directly addressed proposed critical functional roles of convergent V_H locus CBE interactions with the 3’CBEs by testing effects of complete deletion of all ten 3’IgH CBEs and found no measurable influence of their deletion on utilization of V_Hs across the V_H locus (Extended Data Fig. 5). While not ruling out unknown, redundant roles of convergent CBEs within IgH or further downstream upon 3’IgH CBE deletion¹⁷, these findings, coupled with those described in the preceding paragraph, support a major role of the RC as a dynamic downstream loop extrusion anchor for RAG V_H-locus scanning in primary pro-B cells.

In normal primary pro-B cells, cryptic RSS utilization via RAG scanning from the DJ_H-RC-RSS did not extend beyond the IgH locus (Fig. 2b, top and 2c; Extended Data Fig. 2a, top, and 2b; Extended Data Fig. 5 legend and discussion). Strikingly, the V_H locus inversion resulted in RC-based RAG scanning beyond the IgH locus, over distances of 4 Mb upstream through multiple convergent CBE-anchored loop domains to the telomere (Fig. 2b, bottom and 2c; Extended Data Fig. 2a, bottom, 2c). Consistent with linear scanning through these far upstream convergent CBE domains, only cryptic RSSs in convergent orientation with the DJ_H-RC-RSS were utilized (Fig. 2b, bottom and 2c; Extended Data Fig. 2a, bottom, 2c). Correspondingly, overall RC interactions within these upstream IgH locus domains were significantly increased upon the V_H locus inversion (Extended Data Fig. 2d; Supplementary Data 1a, b). Activation of RAG scanning beyond the V_H locus through multiple upstream convergent CBE-based domains in primary pro-B cells with inverted V_H loci is a striking phenomenon that indicates primary pro-B cells neutralize V_H locus-associated CBE-based impediments to loop-extrusion mediated scanning by mechanism that functions broadly within the genome and not just within V_H locus.

Very low-level IgH D-to-J_H V(D)J recombination occurs in cycling v-Abl cells presumably as they pass through the G1 cell cycle stage in which RAG is activated¹⁸. However, when v-Abl cells are viably arrested in G1 by treatment with a v-Abl tyrosine kinase inhibitor (STI-571)¹⁹, they activate robust D-to-J_H rearrangement; but essentially skip IgH locus contraction and V_H-to-DJ_H rearrangement pro-B cell processes and move-on to the precursor (pre) B cell-specific process of robustly rearranging Vκs across the 3Mb Igκ light chain locus to Jκs²⁰. Thus, to elucidate candidate factors for involvement in IgH locus contraction and distal V_H rearrangement, we employed GRO-seq to compare transcriptional activity of genes encoding proteins potentially involved in long-range RAG scanning in primary pro-B cells versus G1-arrested v-Abl cells (Extended Data Fig. 6a). Notably, the Wings apart-like (Wapl) factor gene¹⁰ stood out as having very low transcription, relative to other cohesin-complex factor genes in primary pro-B cells versus v-Abl cells (Extended Data Fig. 6a, b; see legend to Extended Data Fig. 6). Western blotting revealed Wapl protein levels in cycling v-Abl cells were approximately 9-fold higher than those of cultured primary pro-B cells (Fig. 4a, b). Wapl unloads cohesin from chromosomes by transiently opening the cohesin ring^21–23 and collaborates with transcription and CTCF-binding to position cohesin in genomes^24,25. Relevant to V_H locus contraction and RAG-scanning, Wapl-depletion extends CTCF-anchored loop sizes in association with increasing cohesin density²⁶, by increasing probability of extending loop extrusion past CTCF-bound CBE loop anchors^27,28. Thus, our findings potentially implicate Wapl down-regulation in mediating contraction and long-range RAG scanning of the IgH locus in primary pro-B cells. A study published while this paper was under review reached related conclusions regarding reduced pro-B cell Wapl expression due to Pax5 transcriptional repression¹⁶.

Figure. 4 |. Wapl depletion activates IgH V_H locus contraction and long-range V_H utilization in G1-arrested v-Abl cells.

a, Western blotting to detect Wapl protein levels in the primary #5 cycling v-Abl cells and the derived Wapl-degron #5–9 v-Abl cells before (day0) and after STI&IAA&Dox treatment (day4) and cultured primary pro-B cells. One of the two experiments is shown. b, Western blotting results to determine relative expression of Wapl protein levels in cycling and G1 arrested primary #5 v-Abl cells and cultured primary pro-B cells. Average values ± s.d. of Wapl protein levels are shown. For comparison, the intensity of the Wapl band in G1-arrested primary #5 v-Abl cells is set as 1.0. Average value is indicated at each bar. n, number of independent repeats. Indicated P values were calculated using unpaired two-tailed t-test. c, Average 3C-HTGTS signal counts ± s.e.m. across the four V_H domains in cultured RAG1-deficient primary pro-B cells (top), G1-arrested RAG1-deficient v-Abl cells without (middle) or with Wapl depletion (bottom). n, number of independent experiments. For comparison, 15 representative major interaction peaks/clusters are shown as in Fig. 3b. d, Average utilization frequencies ± s.d. of all V_H segments in RAG1-complemented, G1-arrested Wapl-degron v-Abl cells without (Untreated) or with (IAA&Dox) Wapl depletion are indicated. Average percentage ± s.d. of V_HDJ_H and DJ_H rearrangements are shown. n, number of independent experiments. See Supplementary Table 2 and Methods for more details.

To directly test ability of Wapl to impede IgH locus contraction and V_H-to-DJ_H rearrangement, we generated RAG1-deficient C57BL/6 v-Abl cells in which we could degrade Wapl via a mini auxin-inducible degron (mAID) approach^2,29 (see methods). In these “Wapl-degron v-Abl cells”, addition of doxycycline (Dox) and auxin (Indole-3-acetic acid, IAA) triggers proteasome-dependent degradation of endogenous mAID-tagged Wapl. To deplete mAID-tagged Wapl in the G1 stage, we treated Wapl-degron v-Abl cells with IAA and Dox from day1 to day4 during the STI-571-induced four-day G1 arrest period (day0–4) (Extended Data Fig. 6f). Western blotting revealed that mAID-tagged Wapl levels were further reduced about 2~3-fold in G1 arrested versus cycling Wapl-degron v-Abl cells and substantially depleted upon IAA&Dox-treatment of the G1 arrested lines (Extended Data Fig. 6g, i). Depletion of chromatin-bound mAID-tagged Wapl across the IgH locus and the genome in G1-arrested Wapl-depleted v-Abl cells was confirmed by ChIP-seq (Extended Data Fig. 7a, b). While Wapl depletion is lethal to dividing cells^30,31, Wapl-depleted v-Abl cells underwent normal STI-571-induced G1 arrest with modest effects on viability (Extended Data Fig. 6k, l).

Binding patterns of CTCF across V_H domains, RC transcription, and genome-wide transcription patterns were mostly maintained in IAA&Dox-treated Wapl-degron v-Abl cells (Extended Data Fig. 7c; 8a, b). As expected^24–26, Wapl depletion increased Rad21 redistribution to distal V_H locus CBEs and 3’ IgH CBEs to yielding a pattern significantly similar to that of cultured primary pro-B cells (Extended Data Fig. 7d). 3C-HTGTS-based interaction experiments of untreated versus IAA&Dox-treated G1-arrested RAG1-deficient Wapl-degron v-Abl cells revealed low-level peaks of RC interactions in all but the proximal 7183/Q52 V_H domain in untreated cells that were substantially increased upon Wapl depletion (Fig. 4c, middle and bottom). Indeed, the major interactions peaks across the J558/3609, J558, and Middle V_H domains that were substantially increased in IAA&Dox-treated Wapl-degron v-Abl cells closely corresponded to dominant peaks observed in locus-contracted RAG1-deficient primary pro-B cells (Fig. 4c, upper, labeled 1–15). Together, these findings support a role for reduced Wapl expression in enhancing V_H-locus contraction in primary pro-B cells, likely by increasing ability of cohesin-mediated loop extrusion to circumvent CTCF-bound CBE impediments.

To test effects of Wapl-depletion on IgH V(D)J recombination, we introduced RAG into Wapl-degron v-Abl cells prior to G1 arrest (Extended Data Fig. 8c). Like standard v-Abl cells^1,5, untreated RAG1-complemented Wapl degron v-Abl cells had robust D-to-J_H rearrangement (Extended Data Fig. 8d; Supplementary Table 2), but little V_H-to-DJ_H rearrangement (Fig. 4d, upper; Extended Data Fig. 9a, b). Wapl depletion markedly reduced distal DFL16.1 rearrangements, but promoted increased proximal DQ52 rearrangements (Extended Data Fig. 8d; Supplementary Table 2), a pattern observed in the context of IGCR1 CBE inactivation^5,32 (Extended Data Fig. 8e, bottom). Strikingly, Wapl depletion promoted an approximately 4-fold overall increase V_H-to-DJ_H recombination that, while about 12-fold lower than that of BM pro-B cells (Fig. 1b upper, 4d bottom; Extended Data Fig. 9c), was associated with significantly increased utilization of V_Hs across all four V_H domains of the 2.4 Mb IgH locus. An exception was decreased utilization of the most proximal V_Hs relative to that of more distal V_Hs (Fig. 4d, bottom; Extended Data Fig. 9c; Supplementary Table 2, 5), consistent with diminished impediment activity of the proximal V_H-CBEs (Extended Data Fig. 7d, middle; Extended Data Fig. 8e, bottom). Like primary pro-B cells, Wapl-depleted v-Abl cells utilized cryptic RSSs across the length of 2.4 Mb V_H locus in convergent, but not in the same orientation, as the DJ_H-RC-RSS (Extended Data Fig. 9e). We also made the same 2.4 Mb inversion of V_H locus in Wapl-degron v-Abl cells as in primary pro-B cells (see Methods for details). Strikingly, the v-Abl cell V_H locus inversion had the same effects as found in primary pro-B cells on bona fide V_H-RSS and cryptic RSS rearrangement, including extension of cryptic RSS rearrangements upstream of the V_H locus to the telomere (Extended Data Fig. 9d–f; compare with Fig. 2b–c). Together, these results strongly support a role for reduced Wapl expression in facilitating locus contraction and orientation-specific RAG scanning of the V_H locus.

The same G1-arrested v-Abl lines that require Wapl-depletion for robust IgH locus contraction and distal V_H utilization, robustly utilize Vκs across the 3Mb Vκ locus (a hallmark of Vκ locus contraction²⁰), with relative levels and patterns significantly similar to those of primary pre-B cells (Extended Data Fig. 8f). IAA&Dox-mediated Wapl depletion in these v-Abl cells reduced Vκ to Jκ rearrangement levels by about 7-fold compared to that of untreated v-Abl cells (Extended Data Fig. 8g), but did not substantially impact rearrangement patterns across the Vκ locus, with the most notable changes reflecting decreased proximal and increased very distal Vκ utilization, consistent with increased loop-extension mediated by further Wapl-depletion (Extended Data Fig. 8h). Together, the reduced Vκ and V_H rearrangement levels in IAA&Dox-treated v-Abl cells suggest that complete Wapl depletion may negatively impact V(D)J recombination activity per se, perhaps by impacting RC or RAG activity. The finding of primary pre-B cell patterns of Vκ rearrangement across the 3 Mb Vκ locus in untreated v-Abl cells is striking, suggesting that V_H and Vκ utilization across their respective loci, which are very differently organized⁸, may be regulated in different ways. As one possibility, G1 arrest of v-Abl cells reduces Wapl levels by 2–3 fold (Fig. 4b; Extended Data Fig. 6i), a reduction perhaps sufficient to promote contraction/scanning of the Vκ locus versus the V_H locus due to the latter having a much greater CBE density⁸. Also, potential roles for other cohesin-complex factors are conceivable given that down-regulation or modification of CTCF can neutralize CBE impediment activity and allow robust IgH locus contraction and scanning in experimental v-Abl cell systems².

We implicate Wapl down-regulation in primary pro-B cells as a major mechanism for partially neutralizing effects of CTCF-bound CBE impediments within the 2.4 Mb V_H locus to allow RC-bound RAG scanning of V_H locus chromatin propelled past it by cohesin-mediated loop extrusion. Correspondingly, V_H utilization in primary pro-B cells is significantly correlated with cohesin-binding in distal J558/3609, Middle and proximal 7183/Q52 domains and approaches significance in the J558 domain (Extended Data Fig. 7e). However, in IAA&Dox-treated v-Abl cells correlations are not significant in distal V_H domains, likely due to low overall V(D)J recombination levels and/or ectopic versus physiological RAG expression (Extended Data Fig. 7e and legend discussion). Consistent with findings of contemporaneous CTCF depletion studies in v-Abl lines², analyses of various parameters for each individual V_H in BM pro-B cells and Wapl-depleted v-Abl lines indicated that residual CBE activity and/or transcriptional mechanisms likely enhance utilization of many V_Hs along the scanning path provided by Wapl depletion; but rearrangement of other V_Hs, particularly in the middle domain, do not appear to employ either mechanism (Extended Data Fig. 3 and legend discussion; Supplementary Data 2, 3).

Nascent RCs formed in the absence of RAG serve as loop extrusion impediments that provide a “dynamic” downstream anchor for extrusion of upstream chromatin past the RC^2,5. However, RAG binding to the nascent RC has been implicated in enhancing activity of extrusion impediments within upstream chromatin being scanned¹. Thus, extrusion of upstream chromatin in pro-B cells that harbor nascent RCs, while impeded at major dynamic scanning impediments would likely proceed, on average, further upstream into the V_H locus than in pro-B cells in which the RC is RAG-bound¹. Further, in pro-B cells with nascent RCs, loop extrusion could generate a population of individual pro-B cells in which upstream chromatin extrusion progresses to various distances past the nascent RC in individual cells, allowing subsequent RAG-binding to be initiated at various points across the V_H locus to yield more normalized access to a RAG-bound RC via scanning (see Extended Data Fig.10 for model and legend for other implications). Finally, while we implicate linear RAG scanning as a dominant RC access mechanism for V_Hs across the locus, given that all V_Hs RSSs are in convergent orientation to the DJ_H-RC-RSS, diffusion of those brought into very close proximity at scanning impediments could contribute^1,3.

METHODS

Experimental procedures

No statistical methods were used to predetermine sample size. Experiments were not randomized, and investigators were not blinded to allocation during experiments and outcome assessment.

Mice

Wild-type (WT) C57BL/6 and 129SV mice were purchased from Charles River Laboratories International. RAG1-deficient (Rag1^−/−) mice (stock No.002216: B6.129S7-Rag1^tm1Mom/J) in C57BL/6 background were purchased from the Jackson Laboratory and maintained in the lab. All animal experiments were performed under protocols approved by the Institutional Animal Care and Use Committee of Boston Children’s Hospital. Animals were maintained on a 14-h light,10-h dark schedule in a temperature-controlled environment, with food and water provided ad libitum.

Generation and characterization of IgH V_H locus inversion mouse model

The CRISPR/Cas9-mediated entire IgH V_H locus inversion upstream V_H81X modifications were carried out on the C57BL/6 IgH allele in EF1 ES cell line, which was derived from a F1 hybrid mouse (129/SV:C57BL/6). The ES targeting was performed as previously described³³. sgRNAs targeting the upstream of V_H1–86P (Cut1) and upstream of V_H81-X (downstream of V_H2–2) (Cut2) were cloned by annealing pairs of oligos into pX459 with a puromycin selection marker (Addgene, #62988) following the standard protocol. For transfection, Cas9/sgRNAs (Cut1 and Cut2, each 2.5 µg) were nucleofected via Lonza 4D Nucleofector into 2.5×10⁶ EF1 cells. The electroporated ES cells were grown on a monolayer of MEF cells with puromycin resistance (Stemcell, #00325) for 1 day. Then the ES cells were selected with 2 µg/ml puromycin (Gibco, #A1113802) for another day. After selection, the ES cells were cultured with normal medium for 5–6 days, colonies were picked into 24 well plates that had been coated with MEFs. After 3 days, half of the colonies were frozen at −80°C and the other half were expanded for DNA isolation. Genomic DNA was analyzed with screening-PCR to identify clones in which the desired genetic modifications had taken place. After the primary screen, the potential positive clones with 2.4 Mb V_H locus inversion were amplified by PCR and sequenced. The IgH V_H inversion modifications were confirmed on the C57BL/6 IgH allele by single nucleotide polymorphisms (SNPs), each positive clone was further validated with a normal karyotype by metaphase analysis. All the cell lines were confirmed to be mycoplasma free. The ES clone with the IgH V_H inversion was injected into RAG2-deficient blastocysts to generate chimeras³⁴. The chimeric mice were bred with C57BL/6 mice for germline transmission of the targeted mutations. Heterozygous IgH V_H inversion mice were bred to homozygosity. For further analysis, the IgH V_H inversion mouse model was bred to RAG1-deficient homozygosity in C57BL/6 background. Sequences of all sgRNAs and oligos used are listed in Supplementary Table 6.

Generation of 3’IgH CBEs knockout mouse model

The ten 3’IgH CBEs deletion targeting was carried out in the TC1 ES cell line, which was derived from a 129/SV mouse. The 3’CBE locus was targeted with LoxP sites and deleted by Adeno-Cre (Extended Data Fig. 5b). The positive 3’CBEs-deleted ES cell clones were confirmed by PCR amplification (Extended Data Fig. 5c). All of the detailed methods of this type of ES targeting were performed as previously described³³. The ES clones with the homozygous 3’CBEs deletion were injected into RAG2-deficient blastocysts to generate chimeras³⁴. PCR primers are listed in Supplementary Table 6.

Bone marrow pro-B cells purification

For RAG on-target and off-target analysis, single cell suspensions were derived from bone marrows of 4–6-week-old C57BL/6 WT and IgH V_H inversion mice and incubated in Red Blood Cell Lysing Buffer (Sigma-Aldrich, #R7757) to deplete the erythrocytes. B220⁺CD43^hi+IgM^- pro-B cells were isolated by staining with anti-B220-APC (eBioscience, #17–0452-83), anti-CD43-PE (BD PharMingen, #553271), and anti-IgM-FITC (eBioscience, #11–5790-81) antibodies and purified by fluorescence-activated cell sorting (FACS), the purified primary pro-B cells were subjected to HTGTS-V(D)J-seq as described^13,14.

For 3C-HTGTS, ChIP-seq and GRO-seq, B220-positive WT and IgH V_H inversion primary pro-B cells were separately purified via anti-B220 MicroBeads (Miltenyi, #130–049-501) from individual 4–6-week-old RAG1-deficient mice (WT and IgH V_H locus inversion) and were cultured in opti-MEM medium containing 10% (v/v) FBS plus IL-7 and SCF for 4–5 days as previously described³⁵. Purified primary pro-B cells from different mice were separately cultured in different Petri dishes. Each sample was double-checked and confirmed by PCR prior to various assays as described below. PCR primers are listed in Supplementary Table 6.

Generation of doxycycline inducible Wapl-degron v-Abl cell lines

The targeting plasmid (pTet-On_OsTIR1_V5) for introducing the doxycycline inducible OsTIR1-V5 expression cassette into endogenous Rosa26 locus was constructed by modifying a published pEN113 plasmid (Plasmid #86233, Addgene). The donor plasmid (pW290) used to target the endogenous mouse Wapl locus was constructed by modifying a published pMK290 plasmid (Plasmid #72828, Addgene). Briefly, the doxycycline inducible Tet-On-3G_hPGK promoter sequence was PCR amplified from the plasmid pMK243 (Addgene, #72835) with 5’ SalI and 3’ MluI digestion sites. Then the pTet-On_OsTIR1_V5 targeting vector was generated by cutting out the enhancer and promoter sequence of OsTIR1_V5 expression cassette from the pEN113 plasmid via SalI and MluI digestion, and replacing the amplified Tet-On-3G_hPGK promoter sequence into pEN113 vector via the same restriction sites (Extended Data Fig.6c). To construct donor pW290 vector, 5’ and 3’ homology arms (247 bp each) flanking the BamHI sequence were synthesized via gBlocks by IDT (Integrated DNA Technologies) and cloned into pGEM-T easy vector (Promega, #A1360) to generate the pGEM-Wapl plasmid. The mAID-mClover cassette containing a hygromycin selection marker was excised from pMK290 by BamHI and cloned at the BamHI site of the pGEM-Wapl plasmid, between the homology arms, to generate the pW290 vector (Extended Data Fig. 6d). To modify the mouse Rosa26 and Wapl gene, sgRNAs targeting the endogenous Rosa26 locus or 3’ end sequence of the Wapl gene were cloned by annealing pairs of oligos into pX330 (Addgene, #42230) to construct the pX330_Rosa26 and pX330_Wapl-mAID, respectively.

The RAG1-deficient; Eμ-Bcl2⁺ v-Abl cells were co-transfected with pTet-On_OsTIR1_V5 and pX330_Rosa26 via Lonza 4D Nucleofector as shown before^3,5, selected with 0.4 mg/ml of G418 for 5–6 days, and subcloned by dilution. Candidate clones with desired gene modifications were screened by PCR. The expression of OsTIR1 in primary v-Abl clone #5 was confirmed by western blotting (Extended Data Fig. 6g). Similarly, the resultant clones were further co-transfected with donor pW290 with corresponding Cas9/sgRNA pX330_Wapl-mAID via nucleofection, selected with 800 μg/ml hygromycin B (Millipore Sigma, #10843555001) for 3 days, and subcloned by dilution. Candidate clones with desired gene modifications were screened by PCR and confirmed by southern blot in Extended Data Fig. 6d–e, and the expression of mAID-tagged Wapl protein in Wapl-degron v-Abl clone #5–3 and #5–9 was confirmed by western blotting and FACS as outlined in Extended Data Fig. 6g, j. The resulting lines were referred to as the Wapl-degron v-Abl lines and used in this study. Sequences of all sgRNAs and oligos used are listed in Supplementary Table 6.

Generation of 2.4 Mb IgH V_H inversion Wapl-degron v-Abl cell lines

To generate the 2.4 Mb IgH V_H inversion Wapl-degron v-Abl cell lines without corresponding WT IgH V_H region, two sgRNAs were firstly designed to target the upstream of V_H1–86P (Cut1) and downstream of 3’CBE (Cut3) respectively to delete the entire IgH locus and generate the Wapl-degron v-Abl cells harboring only one IgH allele. The resultant single IgH clones were further targeted by another two sgRNA (Cut2 and Cut4) to generated the Wapl-degron v-Abl cells with entire IgH V_H locus inversion upstream V_H81X. Candidate clones with desired gene modifications were screened by PCR and confirmed by Sanger sequencing. Sequences of all sgRNAs and oligos used are listed in Supplementary Table 6.

Treatment of Wapl-degron v-Abl cells with IAA&Dox

Wapl-degron v-Abl cells were first treated by 3 μM STI-571 (STI) for 1 day to induce G1 arrest, then apart from STI, cells were also treated with 150 μM Indole-3-acetic acid (auxin analog, IAA, Sigma-Aldrich, #I3750–25G-A) and 2 mg/mL doxycycline (Dox, Sigma-Aldrich, #D3447) from day1 to day4 to specifically deplete mAID-tagged Wapl protein in G1 arrested v-Abl cells (Extended Data Fig. 6f). All the chemicals were dissolved in DMSO to prepare the stock. DMSO solvent was applied as the mock to untreated cells. Both untreated and IAA&Dox-treated cells were then collected and examined by FACS for protein depletion confirmation (Extended Data Fig. 6j) prior to various assays as described below.

Western Blotting

Western blotting samples were prepared with short-term cultured RAG1-deficient primary pro-B cells, cycling RAG1-deficient v-Abl cells, and G1-arrested RAG1-deficient v-Abl cells with or without Wapl depletion. Western blotting experiments were performed according to published protocol with a few modifications³⁶. 10 million cells were harvested and lysed. Indicated antibodies (anti-Wapl, Thermo Fisher Scientific, #PA5–38024, 1:500; anti-V5-tag, #R960–25, 1:1000; anti-β-actin, Cell Signaling Technology, #3700, 1:3000) were used for the related western blotting experiments. For quantification and comparison of Wapl protein levels, the integrated density of Wapl and β-actin western blotting bands were calculated by Image J (version 1.42q) (Fig. 4b; Extended Data Fig. 6i). See Supplementary Figure 1 for uncropped scan blots with individual repeats.

Cell viability assays

For time-course cell viability assays, Wapl-degron v-Abl cells were treated with IAA&Dox from day1 to day4 during the STI-571 induced four-day G1 arrest experimental period. The viability of Wapl-degron v-Abl cells with or without Wapl depletion at different time points was determined by the percentage of viable lymphocyte in the population gated by FACS side (SSC) and forward (FSC) scatters out of the total cells.

GFP quantification and cell cycle stage analysis

For GFP (Clover) signaling quantification of Wapl depletion, both the untreated and IAA&Dox-treated Wapl-degron v-Abl cells at different time points were directly collected and examined by FACS. Primary v-Abl line #5 was processed as a GFP negative control.

For cell cycle analysis, Wapl-degron v-Abl cells without or with Wapl depletion at different time points were washed with ice-cold PBS and fixed in ethanol. The fixed cells were incubated with propidium iodide (PI) and RNase, and further analyzed by FACS.

RAG complementation

The Rag1-expressing vector pMSCV-RAG1-IRES-Bsr and Rag2-expressing vector pMSCV-FLAG-RAG2-GFP were kindly provided by Yu Zhang and Xuefei Zhang. RAG was reconstituted in RAG1-deficient v-Abl cells via retroviral infection of cells with the pMSCV-RAG1-IRES-Bsr and pMSCV-FLAG-RAG2-GFP vectors followed by 3–4 days of Blasticidin (ThermoFisher, #R21001) selection to enrich for cells with virus integration.

HTGTS-V(D)J-seq

For IgH V(D)J recombination analyses, primary pro-B cells from WT and IgH V_H locus inversion mice and RAG1-complemented, RAG1-deficient Wapl-degron v-Abl cells via the approach described above. For Igκ rearrangement analysis (Extended Data Fig. 8f–h), primary pre-B cells are from 129SV background mice, there is no difference in the Igκ locus between C57BL/6 and 129SV backgrounds. HTGTS-V(D)J-seq libraries were prepared as previously described^1,5,13,14. Briefly, 1 μg of gDNA from sorted primary pro-B cells or 8 μg of gDNA from G1-arrested RAG1-complemented Wapl-degron v-Abl cells was sonicated and subjected to LAM-PCR using biotinylated J_H1–4 bait primers. Single-stranded LAM-PCR products were purified using Dynabeads MyONE C1 streptavidin beads (ThermoFisher, #65002) and ligated to bridge adaptors. Adaptor-ligated products were amplified by nested PCR with indexed J_H1–4 primers and the primer annealed to the adaptor. HTGTS-V(D)J-seq libraries for mature splenic B cells (J_H4 primer) or primary pre-B cells (Jκ5 primer) were processed with the same procedure. The PCR products were further tagged with Illumina sequencing adaptor sequences, size-selected via gel extraction and loaded onto an Illumina NextSeq550 using paired-end 150-bp sequencing kit. Primer sequences are listed in Supplementary Table 6.

HTGTS-V(D)J-seq data processing and analyses

HTGTS-V(D)J-seq libraries were processed via the pipeline described previously¹⁴. The data were aligned to mm9 genome and analyzed with all duplicate junctions included in the analyses as previously described¹⁴. In the IgH repertoire libraries generated from C57BL/6 WT primary pro-B cells with J_H1–4 coding end baits^13,37, we detected in V_HDJ_H exons 109 functional V_Hs, as well as 23 pseudo V_Hs (Supplementary Table 1) (average rearrangement >1). For comparisons, utilization data of V_H and D segments in each library of WT or IgH V_H inverted primary pro-B cells was normalized to 118,475 total recovered junctions (Fig. 1b; Supplementary Table 1). Utilization data of V_H and D segments in each library of G1-arrested Wapl-degron v-Abl lines was normalized to 100,000 total recovered junctions (Fig. 4d; Extended Data Fig. 8d, 9a–d; Supplementary Table 2, 5). To show the absolute Vκ usage and relative Igκ recombination level, each HTGTS library was normalized to 20,000 total reads (Igκ junctions + germline reads) (Extended Data Fig. 8f–g); to show the percentage of individual Vκ usage pattern across the Igκ locus, each HTGTS library was normalized to 20,000 total Junctions (Vκ usage) (Extended Data Fig. 8h). Utilization data of V_H segments in each library of WT primary pro-B cells (129SV background) were normalized to 10,000 total recovered junctions; non-productive V_H segments in each library of WT and 3’CBE-knockout splenic B cells (129SV background) were normalized to 10,000 non-productive recovered junctions (Extended Data Fig. 5d–f; Supplementary Table 3). For the RAG off-targets in WT or IgH V_H inverted primary pro-B cells, each library was isolated from independent J_H1–4-baited library and was normalized to 402,241 isolated junctions (Fig. 2b; Extended Data Fig. 2). For comparison of RAG off-targets in primary pro-B cells, each of the biological independent library was normalized to 2,979 off-target junctions in the indicated region (chr12:114,666,712–121,257,530​) for statistical analysis (Fig. 2c). For the RAG off-targets in Wapl-degron v-Abl cells and derived IgH V_H inverted lines, each library was pooled from 3 independent J_H1–4-baited libraries and was normalized to 432,676 isolated junctions without repetitive sequences (Extended Data Fig. 9e). For comparison of RAG off-targets in Wapl-degron v-Abl cells, each library was normalized to 2,332 off-target junctions in the indicated region (chr12:114,666,712–121,257,530​) for statistical analysis (Fig. 9f). Junctions are denoted as in ‘‘+’’ orientation if prey sequence reads in centromere-to-telomere direction and in ‘‘-’’ orientation if prey sequence reads in telomere-to-centromere direction (Fig. 2; Extended Data Fig. 2, 9e–f).

ChIP-Seq library preparation and analyses

ChIP-seq libraries were prepared with short-term cultured RAG1-deficient primary pro-B cells or G1-arrested RAG1-deficient v-Abl cells and performed based on previously described procedures with a few modifications². RAD21 antibody (Abcam, #ab992, 1:300), CTCF antibody (Millipore, #07–729, 1:150) and Wapl antibody (Thermo Fisher Scientific, #PA5–38024, 1:300) were used for the related ChIP-seq experiments. For comparison of all the ChIP-seq results in this study, we used bowtie2 to align ChIP-seq reads to mm9 genome and run MACS2 callpeak with parameters ‘ -t IP.bam -c input.bam –nomodel –keep-dup all -extsize 51 -nolambda -B -SPMR -g mm -broad’ to generate bigwig graphs normalized to 10 million reads.

GRO-seq and data analyses

GRO-seq libraries were prepared as described previously^1,17 on short-term cultured RAG1-deficient primary pro-B cells and G1-arrested RAG1-deficient Wapl-degron v-Abl cells. To visualize the genome-wide RNA expression level, we run htseq-count to count read number on each gene and made scatter plot after down-sampling each sample to 10 million reads mappable to any genes. For comparison of GRO-seq results between WT and IgH inversion cultured primary pro-B cells, GRO-seq profiles were inverted in the IgH inversion primary pro-B cells (Extended Data Fig. 4c, bottom).

3C-HTGTS and data analyses

3C-HTGTS was performed as previously described⁵ on short-term cultured RAG1-deficient primary pro-B cells and G1-arrested Wapl-degron v-Abl cells. For comparison, sequencing reads of all the 3C-HTGTS libraries were aligned to mm9 genome and processed as previously described⁵. Data were plotted for comparison after normalizing junction from each experimental 3C-HTGTS library by random selection to the total number of genome-wide or entire IgH locus (± 10 kb, chr12:114,444,081–117,258,165) junctions recovered from the smallest library in the set of libraries being compared. Chromosomal interaction patterns were very comparable before and after normalization. For short-term cultured RAG1-deficient primary pro-B cells, both 3C-HTGTS libraries using iEμ (Fig. 3b) and 3’CBE (Fig. 3c) as baits were normalized to 885,651 total junctions of genome-wide, respectively. For comparison of iEμ-baited interactions between WT and IgH inversion primary pro-B cells in the inverted orientation, 3C-HTGTS results were inverted in the IgH inversion primary pro-B cells (Fig. 3b and 3c, bottom). For comparison of 3C-HTGTS libraries between short-term cultured RAG1-deficient primary pro-B cells and G1-arrested Wapl-degron v-Abl cells, we down-sampled each library to 43,658 junctions and removed the reads in the regions closely near to the iEμ bait (chr12:114,664,385–114,668,233), which did not contribute to the locus interaction, and presented them in the plots (Fig. 4c and Extended Data Fig. 8e, Supplementary Data 1g). The sequences of primers used for generating 3C-HTGTS libraries are listed in Supplementary Table 6.

Quantification and statistical analysis

Graphs were generated using GraphPad Prism 7.0 and R 3.6.3. Genomic heatmaps of ChIP-seq signals were generated using deeptools 3.3.2 (Extended Data Fig. 7b). After normalization in each sample, GRO-seq, 3C-HTGTS and ChIP–seq signals of multiple repeats were merged as average ± s.e.m. of the maximum value in each repeat in each bin, after dividing the plotting region into 1000 bins (Fig. 3b–c, 4c; Extended Data Fig. 3, 4, 6b, 7a, 7c–d, 8a and 8e). Unpaired two-tailed t-test was used to determine the statistical significance of difference in signal count (Fig. 2c; Extended Data Fig. 2d, 8d and 9f) or western band intensity (Fig. 4b; Extended Data Fig. 6i) between two experimental groups. Spearman’s rank correlation was calculated to determine the similarity in transcriptome-wide GRO-seq counts (Extended Data Fig. 6a and 8b), or the similarity in signal count across Vs (Extended Data Fig. 7d–e and 8f), after calculating the average among repeats. When comparing V_H usage, Rad21 and CTCF ChIP-seq signals, we counted the ChIP-seq reads within 5 kb of the RSS of each V_H (Extended Data Fig. 7d–e). Spearman’s correlation coefficient (rho) and P values that were determined by the two-sided Spearman’s correlation test between indicated samples are presented in the relevant figure legends. P values are calculated and shown in the relevant figures and Supplementary Tables as follows: *P ≤ 0.05, **P ≤ 0.01 and ***P ≤ 0.001.

Extended Data

Extended Data Figure. 1 |. Generation and characterization of IgH V_H locus inversion mouse model.

a, Schematic diagram showing CRISPR-Cas9-mediated entire IgH V_H locus inversion upstream V_H81X in embryonic stem cells (ES cells) on the IgH allele in C57BL/6 genetic background. Cut1 and Cut2 showing the location of 2 sgRNAs. Details as shown in Fig. 1a. b, Confirmation of the upstream and downstream inversion junctions by Sanger sequencing. The sgRNA-targeting sequence is underlined, and the PAM sequence is labelled in red. sgRNAs and oligos used are listed in Supplementary Table 6. c, Schematic showing the generation of IgH V_H locus inversion mouse model and further assays for phenotype and mechanism analyses. d, e, Representative flow cytometry analysis of IgM^- bone marrow (BM) B cell populations in 4~6-week-old WT (d) and IgH V_H locus inversion (e) mice. B220⁺IgM^- B cells were gated and shown in the left plot (d, e). The black arrow lines indicate the gating strategies. B220⁺CD43⁺ pro-B and B220⁺CD43^- pre-B cell populations are indicated in the right plot (d, e).

Extended Data Figure. 2 |. RAG utilization of cryptic RSSs across normal and inverted IgH loci in primary pro-B cells.

a, 6 examples (labeled in Fig. 2b, I-VI) of RAG off-target peaks in WT (repeat #1, middle) and IgH V_H locus inversion (repeat #1, bottom) primary pro-B cells plotted at single-base-pair resolution. Top panels are WT 2.4 Mb IgH locus and upstream 4 Mb region track. For comparison, the sequence of RAG off-target peaks (III-VI) in the inverted region are shown in inverted orientation. b, c, RAG off-target junction profiles at IgH locus and upstream 4 Mb region in WT (repeat #2, #3, middle) and IgH V_H inversion (repeat #2, #3, bottom) primary pro-B cells. For comparison, RAG off-target junction profiles in the inverted region are shown in normal orientation (c). The same regions of RAG off-target peaks (shown in a) are highlighted with pink lines (b, c). The IgH V_H inverted region is highlighted in transparent green (WT, a, b) and transparent gold (IgH V_H inversion, a, c), respectively. Del (+) indicates deletional junction. Inv (−) indicates inversional junction. d, Normalized RC-3C-HTGTS reads ± s.d. of WT (n=3) and INV (n=3) in entire upstream IgH locus and J558/3609 V_H region are shown. See Supplementary Data 1 for individual repeats of RC interactions with whole IgH locus and entire upstream IgH locus. Indicated P values were calculated using unpaired two-tailed t-test. n, number of independent experiments.

Extended Data Figure. 3 |. The major RC interactions, transcription patterns, and binding of key looping factors are maintained in the inverted IgH loci in cultured RAG1-deficient primary pro-B cells (Related to Fig. 3b).

For comparison, all the data in the inverted region are shown in normal orientation. Related to Fig. 3b. Zoom-in profiles of 3C-HTGTS, GRO-seq, Rad21/CTCF ChIP-seq signals for ± 10 kb regions of 15 representative peaks/clusters in Fig. 3b from WT (blue) and IgH inversion (INV, Red) cultured RAG1-deficient primary pro-B cells along with relevant bona fide CBE motif sites are shown, data are presented as average signal counts ± s.e.m., PAIR elements (green bars) that located in each peak were also shown above. 3C-HTGTS: WT, n=3; INV, n=3. GRO-seq: WT, n=4; INV, n=4. Rad21 ChIP-seq: WT, n=3; INV, n=3. CTCF ChIP-seq: WT, n=3; INV, n=3. n, number of independent experiments. A similarly detailed view of these parameters across the entire V_H locus is shown in Supplementary Data 2. Discussion: Overall these patterns indicate that proximal 7183/Q52 highly rearranged V_Hs are not transcribed but frequently have CTCF-bound closely associated CBEs; whereas distal J558/3609 highly rearranged V_Hs are often transcribed and/or have CBEs within 5kb that are CTCF-bound, which in cases where both occur may cooperate to enhance rearrangement. In the J558 and Middle V_H domains, some highly rearranged V_Hs are transcribed and/or have associated CTCF/cohesin-bound CBEs; but some highly rearranged V_Hs in these regions appear to have neither, suggesting other unknown mechanisms for their RAG-targeting during scanning. In some latter examples we cannot rule out very low level sense and/or anti-sense transcription as having a contributory role; since the threshold for transcriptional targeting of RAG scanning activity has not been determined. Overall, these patterns are somewhat similar to those discussed in recent report on study involving CTCF-depletion in v-Abl cells².

Extended Data Figure. 4 |. Cohesin/ CTCF binding patterns and germline V_H transcription patterns are shown in normal and inverted IgH loci in cultured RAG1-deficient primary pro-B cells (Related to Fig. 3b and Supplementary Data 2).

a, b, For comparison, the ChIP-seq data in the inverted locus are shown in normal orientation. Average signal counts ± s.e.m. of Rad21 (a) / CTCF (b) ChIP-seq across the 4 V_H domains as indicated in WT (upper, blue, n=3) and IgH V_H locus inversion (bottom, red, n=3) cultured RAG1-deficient primary pro-B cells. n, number of independent experiments. c, Average signal counts ± s.e.m. of GRO-seq profiles across the 4 V_H domains in WT (upper, n=4) and IgH V_H locus inversion (bottom, n=4) cultured RAG1-deficient primary pro-B cells. n, number of independent experiments. For comparison, the GRO-seq results in the inverted locus is shown in inverted orientation (c, bottom). The WT and inverted V_H locus/domains with PAIR elements are diagrammed at the top of each panel. Both the sense and antisense transcription are relative to the entire IgH V_H locus upstream V_H81X with or without inversion and indicated, respectively. For comparison, 15 representative major interaction peaks/clusters as shown in Fig. 3b were indicated with colour shades and numbers. The IgH V_H inverted region is highlighted in transparent green (WT) and transparent gold (IgH V_H inversion), respectively. See Supplementary Data 2 for more details.

Extended Data Figure. 5 |. The 3’CBEs are not required for V_H utilization.

a, Diagram of the entire murine IgH locus with or without ten 3’IgH CBEs, details as shown in Fig. 1a. b, Schematic of the strategy to generate the mouse ES cells with entire 3’CBE locus deletion. Ten 3’IgH CBEs are shown in gray peaks on the top panel. LoxP sites flanking the pGK-Neo selection marker are shown as black triangles. Red filled boxes ‘A’ and ‘B’ indicates the left and right homology arms, respectively, flanking all ten 3’IgH CBEs. c, Gel images show the confirmation results of two independent 3’CBE-deleted ES cell clones, 3’CBEs∆^HOMO #1 and #2 (n=2 biologically independent repeats with similar results). P1-P3 are the genotyping primers and listed in Supplementary Table 6. See Methods for more details. d, Average utilization frequencies ± s.d. of all V_H segments in WT primary pro-B cells (n=6). V_H usage data were extracted from Ba et al.² (GEO: GSE142781). n, number of independent experiments. e, f, Average utilization frequencies ± s.d. of all non-productive (NP) V_H segments in splenic B cells from WT (e, n=3) and 3’CBE deletion (f, n=3) mice, respectively. n, number of independent experiments. All the data are from 129SV background mice (d-f). Prior studies show that non-productive allele rearrangement patterns of splenic B cells are highly correlated with those productive and non-productive patterns in primary pro-B cells⁵. See Supplementary Table 3 and Methods for all further details of this analysis. Discussion: Termination of loop extrusion-mediated V_H locus RAG-scanning within the V_H locus in normal primary pro-B cells may be largely impeded by bona fide convergent V_HRSS-mediated rearrangements. In this regard, inversion of the V_H locus would eliminate any potential contribution of V_HRSS-mediated contributions. Beyond this, such termination of RAG-scanning also may be contributed by cooperative activities of highly frequent distal V_H locus scanning impediments including active transcription sites and CBEs. Due to the number of such impediments, and the possibility that their activity may be collective, assessing their functions in this context may be difficult via targeted mutational analyses. However, our findings of differential effects of the inversion on proximal versus distal V_H locus CBEs interactions with convergent 3’CBEs in the V_H inverted locus provides indirect support of such a potential role for distal CBEs (Fig. 3c, bottom).

Extended Data Figure. 6 |. Generation and characterization of RAG1-deficient Wapl-degron v-Abl cell lines.

a, Scatter plots of average transcriptome-wide GRO-seq counts in G1-arrested v-Abl cells (x axis, n=3) and primary pro-B cells (y axis, n=4). n, number of independent experiments. Representative known requisite genes implicated in the cohesin-complex function for V(D)J recombination and chromatin interactions are highlighted by red circles and blue arrows. Representative known genes implicated in the DNA repair and B cell development were also analyzed to determine if there were any potential transcriptional defects in these essential genes for V(D)J recombination and none were found and highlighted. Analyses of scatter plots indicate that Wapl is expressed at significantly higher levels in G1-arrested v-Abl cells than in primary pro-B cells (Spearman’s correlation coefficient (rho) and P values determined by two-sided Spearman’s correlation test are presented). These transcription finding were confirmed by western blotting studies (Fig. 4a). b, Comparison of Wapl transcription levels by GRO-seq in primary pro-B cells and G1-arrested v-Abl cells from 4 and 3 independent repeats, respectively. Data are presented as average signal counts ± s.e.m. of GRO-seq. n, number of independent experiments. Although other genes upstream and downstream of the Wapl gene show altered transcription in primary pro-B cells compared to the v-Abl cells, their products thus far have not been implicated in loop extrusion or chromatin scanning. c, Schematic of the targeting strategy for introducing Tet-On OsTIR1 expression cassette at the mouse Rosa26 locus. d, Strategy to generate Wapl-degron v-Abl cell lines. Positions of homology arms (gray box), Cas9/sgRNAs and southern blot probe are indicated. e, Southern blot confirmation of two correctly targeted clones (#5–3 and #5–9) with Wapl-mAID on both alleles (n=2 biologically independent repeats with similar results). f, Diagram of the experimental strategy to specifically deplete mAID-tagged Wapl protein in G1-arrested v-Abl cells. g, Western blotting to detect Wapl and Wapl-mAID protein. The indicated clones (#5–3 and #5–9) were grown without or with Wapl depletion at indicated time points before immunoblotting (n=2 biologically independent repeats with similar results). The specific western blotting bands of WT Wapl and Wapl-mAID were labeled. OsTIR1 was detected by anti-V5 antibody. Primary #5 clone was used for the WT Wapl control and β-actin was a loading control. h, Western blotting to detect Wapl protein levels in G1-arrested RAG1-deficient primary #5 v-Abl cells and cultured RAG1-deficient primary pro-B cells. Each sample was loaded with 3 indicated titrations. One of the two experiments is shown. i, Western blotting results to determine relative Wapl protein levels in cycling and G1 arrested Wapl-degron v-Abl cell clones (#5–3 and #5–9) with indicated conditions. For comparison, the intensity of the Wapl band in G1-arrested primary #5 v-Abl cells is set as 1.0. Average value is indicated at each bar. n, number of independent repeats. Indicated P values were calculated using unpaired two-tailed t-test. j, Representative flow-cytometry plots showing the percentage of Clover-positive Wapl-degron v-Abl cells that are without (Untreated) or with (IAA&Dox) Wapl depletion at indicated time points (n=2 biologically independent repeats with similar results). Primary #5 was processed as a Clover-negative control. k, Cell viability assay for G1-arrested v-Abl cells with indicated conditions. Data are presented as average percentage ± s.d. of viable cells for each condition. n, number of independent experiments. l, Representative flow-cytometry plots of propidium iodide (PI) stained G1-arrested v-Abl cells with indicated conditions (n=2 biologically independent repeats with similar results). Percentages in the top-right corner represent the percentage of cells at G1, S and G2/M stage. One of the two experiments is shown. Indicated P values were calculated using unpaired two-tailed t-test (i, k). Plasmids, sgRNAs and oligos used are listed in Supplementary Table 6.

Extended Data Figure. 7 |. Characterization of Wapl/ CTCF/ Rad21-binding in G1-arrested RAG1-deficient Wapl-degron v-Abl cells.

a, Average signal counts ± s.e.m. of Wapl ChIP-seq across the entire IgH locus are plotted as indicated for G1-arrested RAG1-deficient v-Abl cells without (Untreated, blue) or with (IAA&Dox, red) Wapl depletion and cultured RAG1-deficient primary pro-B cells (green). n, number of independent experiments. Wapl ChIP-seq showed that IAA&Dox treatment leads to a depletion of chromatin-bound Wapl at IgH locus, which largely resembles that of primary pro-B cells at IgH locus. b, Three independent repeats of Wapl ChIP-seq signal within ±1.0 kb region across all peaks genome-wide called in G1-arrested RAG1-deficient v-Abl cells without (Untreated) or with (IAA&Dox) Wapl depletion. Top: Average enrichment. c, d, Average signal counts ± s.e.m. of CTCF (c) and Rad21 (d) ChIP-seq across the entire IgH locus are plotted as indicated for G1-arrested RAG1-deficient v-Abl cells without (Untreated, blue) or with (IAA&Dox, red) Wapl depletion, and cultured RAG1-deficient primary pro-B cells (green). n, number of independent experiments. Rad21 ChIP-seq showed that Wapl depletion in G1-arrested v-Abl cells influenced Rad21 (cohesin) redistribution across the IgH locus to give a pattern significantly similar to primary pro-B cells (Spearman’s correlation r=0.84, P<2e-16) (d). e, Spearman correlation analyses of V_H usage with Rad21 and CTCF binding in WT primary pro-B cells and Wapl depleted v-Abl cells. Data for Spearman correlation analyses are from Fig. 1b, 4d, and Extended Data Fig. 4a–b, and 7c–d. Indicated P values were calculated by a two-sided Spearman’s correlation test and shown as: *: 0.01≤P<0.05, **: 0.001≤P<0.01, and ***: P<0.001. Discussion: As noted in the text, we found significant correlations between V_H usage and Rad21 or CTCF binding in 3 of the 4 V_H families including the distal V_H J558/3609 family in primary pro-B cells. However, in the IAA&Dox-treated v-Abl lines, we did not observe a significant correlation between V_H usage and Rad21 or CTCF binding in the distal V_H J558/3609 family. To explore the reason for this correlation discrepancy, we did an analysis of the correlation of primary pro-B cell V_H usage patterns with cohesin and CTCF binding patterns in IAA&Dox-treated v-Abl cells, which indeed revealed significant correlations of these patterns in distal J558/3609 V_H usage for both cohesin and CTCF (Extended Data Fig. 7e). This finding indicates that the lack of correlation of distal J558/3609 V_H usage with cohesin and CTCF binding patterns in Wapl-depleted v-Abl cells stems from their V_H usage patterns. In this regard, primary pro-B cells generate their V_H repertoires from a steady state population in which these cells presumably have RAG scanning across the locus fully engaged and coordinated with V_H locus contraction. However, in Wapl-depleted v-Abl cells, RAG scanning and V_H locus contraction may not be physiologically matched due to ectopic RAG expression and ectopically-induced Wapl depletion. The further impact of the nearly complete Wapl-depletion in IAA&Dox-treated v-Abl cells versus Wapl expression levels in primary pro-B cells (Fig. 4a) is reflected by greatly reduced levels of proximal versus distal V_H rearrangements in v-Abl lines versus those primary pro-B cells (Fig. 4d, bottom; Extended Data Fig. 9c; Supplementary Table 2, 5). Moreover, complete Wapl depletion decreases overall V(D)J recombination activity per se in v-Abl lines as noted in the text, which might also impact these correlations.

Extended Data Figure. 8 |. Characterization IgH gene transcription, D usage, and Igκ rearrangements in G1-arrested Wapl-degron v-Abl cells.

a, Average signal counts ± s.e.m. of GRO-seq across the 4 V_H domains (left) and RC region (right) are plotted as indicated in G1-arrested RAG1-deficient v-Abl cells without (Untreated, upper) or with (IAA&Dox, middle) Wapl depletion and cultured RAG1-deficient primary pro-B cells (bottom). n, number of independent experiments. For comparison, 15 representative major interaction peaks/clusters as Fig. 3b are indicated. PAIR elements are indicated as purple lines with gray background. b, Scatter plots of average transcriptome-wide GRO-seq counts in G1-arrested RAG1-deficient v-Abl cells without (Untreated, x axis) and with (IAA&Dox, y axis) Wapl depletion. n, number of independent experiments. Representative known requisite genes for V(D)J recombination and chromatin interaction are highlighted by red circles and blue arrows in each of the three scatter plots. Spearman’s correlation coefficient (rho) and P value determined by two-sided Spearman’s correlation test are presented. c, Diagram of the experimental strategy including retrovirus-mediated RAG complementation in Wapl-degron v-Abl cells for HTGTS-V(D)J-seq assay. d, Average utilization frequencies ± s.d. of all D segments from DJ_H plus V_HDJ_H joins in RAG1-complemented, G1-arrested Untreated primary, Untreated Wapl-degron and IAA&Dox-treated Wapl-degron v-Abl lines. n, number of independent experiments. Indicated P values were calculated using unpaired two-tailed t-test. e, Average 3C-HTGTS signal counts ± s.e.m. baiting from RC for interactions with the RC domains that includes 3’CBE downstream and the most proximal V_Hs upstream in RAG1-deficient cultured primary pro-B cells (top), G1-arrested RAG1-deficient v-Abl cells without (middle) or with Wapl depletion (bottom). n, number of independent experiments. See Supplementary Data 1 for whole IgH locus interactions. Related very low-level peaks of RC interactions with IGCR1 in IAA&Dox treated cells suggest IGCR1 impediments were neutralized by Wapl depletion in v-Abl cells. f, Absolute individual Vκ usage in total reads. Each library was normalized to 20,000 total reads including Igκ junctions and germline reads. Data are presented as average absolute Vκ usage ± s.d. across the Igκ locus in WT primary pre-B cells and untreated Wapl-degron v-Abl cells (G1_Untreated). For the IAA&Dox-treated Wapl-degron v-Abl cells (G1_IAA&Dox), data are presented as average absolute Vκ usage. Spearman correlation of Vκ usage patterns between WT primary pre-B cells and untreated Wapl-degron v-Abl cells is statistically significant (r=0.96, P<0.001). P values are calculated by two-sided Spearman’s correlation test. n, number of independent experiments. g, Relative Vκ to Jκ rearrangement levels in RAG1-complemented cycling v-Abl cells and G1-arrested Wapl-degron v-Abl cells without (G1_Untreated) or with (G1_IAA&Dox) Wapl depletion. Relative Vκ to Jκ rearrangement levels = Igκ junctions / (Igκ junctions + germline reads) %.Data are presented as mean percentage values ± s.d. for cycling v-Abl cells and untreated Wapl-degron v-Abl cells (G1_Untreated). For IAA&Dox treated Wapl-degron v-Abl cells (G1_IAA&Dox), data are presented as mean percentage values. Average Igκ recombination levels are indicated at each bar. n, number of independent experiments. Vκ usage data of RAG1-complemented G1-arrested Wapl-degron v-Abl cells without or with Wapl depletion from Panel f (middle and bottom) are used to generate the bar graphs (G1_Untreated and G1_IAA&Dox). n, number of independent experiments. h, Percentage of individual Vκ usage in total Vκ to Jκ rearrangements. The same library as panel f was normalized to 20,000 total Igκ junctions. Data are presented as average percentage ± s.d. in WT primary pre-B cells and untreated Wapl-degron v-Abl cells (G1_Untreated). For the IAA&Dox treated Wapl-degron v-Abl cells (G1_IAA&Dox), data are presented as average percentage. n, number of independent experiments. See Supplementary Table 2, 4 and Methods for all further details of the analyses.

Extended Data Figure. 9 |. The large V_H inversion has similar impacts on normal and cryptic RSS utilization within and beyond the V_H locus in G1-arrested Wapl-degron v-Abl cells and primary pro-B cells.

a-d, Average utilization frequencies ± s.d. of all V_H segments in RAG1-complemented, G1-arrested v-Abl cells with indicated conditions. Average percentage ± s.d. of V_HDJ_H and DJ_H rearrangements are shown. Untreated primary, n=3 (a); Untreated Wapl-degron (single IgH allele), n=3 (b); IAA&Dox Wapl-degron (single IgH allele), n=3 (c); IAA&Dox Wapl-degron IgH V_H inversion (single IgH allele), n=3 (d) were used for independent experiments for statistical analyses with error bars. n, number of independent experiments. All V_H segments divided into four domains from most proximal to distal. See Supplementary Table 2, 5 and Methods for more details. e, Pooled HTGTS junction profiles at IgH locus and upstream 4 Mb region for deletional and inversional joining in RAG1-complemented, G1-arrested v-Abl cells without (Untreated, upper, pooled n=3) or with Wapl depletion (IAA&Dox, middle, pooled n=3), or with 2.4 Mb inversion with Wapl depletion (IAA&Dox, IgH V_H inversion, bottom, pooled n=3). For comparison, the V_H usage (d) and RAG off-target (e, bottom) data in the V_H-loci inverted v-Abl cells are shown in normal IgH orientation. A smaller scale is used to present recombination to cryptic recombination signal sequences in Wapl-degron v-Abl lines (e) compared with that in primary pro-B cells (Fig. 2b); the need for this likely reflects lower levels of overall V(D)J recombination following Wapl depletion. f, Average frequencies ± s.d. of plus strand (red, +) and minus strand (blue, -) joining events within indicated regions in RAG1-complemented, G1-arrested v-Abl cells: Untreated Wapl-degron, n=3; IAA&Dox Wapl-degron, n=3; IAA&Dox Wapl-degron IgH V_H inversion, n=3. n, number of independent experiments. Indicated P values were calculated using unpaired two-tailed t-test. See Methods for more details.

Extended Data Figure. 10 |. Working model for loop extrusion-mediated IgH locus contraction for RAG chromatin scanning.

a, In C57BL/6 mice, all 109 V_H segments are located within one or another four V_Hs domains as indicated in figure from proximal to distal: 7183/Q52 (blue), Middle (pink), J558 (green) and J558/3609 (red). Beyond V_H domains relevant elements and proteins including the DJ_HRC, 12 and 23 RSSs, cohesin, and RAG1 and RAG2 subunits and complexes are indicated in the box. b-j, Model for loop-extrusion-mediated physiological locus contraction for distal V_H utilization based on RAG1-deficient background analyses and RAG on- and off-target analyses in primary pro-B cells with reduced Wapl expression. See text for overall description. In brief: Loop extrusion past the nascent (non-RAG-bound) DJ_H-RC may cover much greater distances of upstream chromatin (that would vary from cell) than would occur when RAG is bound to the RC, as directly suggested by results of a recent publication from our lab¹ (b-f). If RAG bound to the DJ_H-RC subsequently in such cells, it could form active DJ_HRCs that could initiate loop-extrusion-mediated scanning at different extrusion points across the V_H locus (g-j), which could avoid downstream rearrangements and potential impediments in those cells to provide more equal scanning access to all V_Hs across the locus⁶ for V(D)J recombination and contribution to diverse antibody repertoires. This model and findings that form its basis¹ may also be relevant to why V_H utilization patterns, which must be done in RAG sufficient cells, in some cases, do not correlate as well as might be anticipated with various peaks found for CTCF/cohesin-binding and 3C-HTGTS interaction studies done in RAG-deficient cells^2,this study. In this regard, future experiments with catalytically dead RAG mutants in which RAG still binds to the RC, as opposed to RAG-deficient cells, may be informative. Finally, please see related discussion in Extended Data Fig. 7.