Locus folding mechanisms determine modes of antigen receptor gene assembly

The dynamic organization of three-dimensional genome structure regulates important aspects of DNA transcription, replication, recombination, and repair, and is controlled by two fundamental mechanisms in interphase cells. Interactions among chromatin regions of similar transcriptional activity and covalent histone modifications, referred to as homotypic chromatin interactions, contribute to genome-wide architecture in all eukaryotes (Eagen, 2018; Hildebrand and Dekker, 2020; Luppino and Joyce, 2020; Penagos-Puig and Furlan-Magaril, 2020; Rowley and Corces, 2018). These homotypic interactions are mediated by multiple mechanisms, which include coalescence of heterochromatin binding proteins or transcription factors via phase separation, homotypic contacts between transcriptionally active or repressive histone marks, and promoter–promoter or promoter–enhancer contacts in microenvironments known as transcription factories (Boija et al., 2018; Chong et al., 2018; Csink and Henikoff, 1996; Falk et al., 2019; Ganai et al., 2014; Gotzmann and Foisner, 1999; Harrison et al., 2021; Hilbert et al., 2021; Li et al., 2012; Rowley et al., 2017; Wang et al., 2019). In higher-order eukaryotes, point-to-point contacts that fold chromosomes are mediated by active extrusion of DNA to form loops that are anchored by DNA-binding proteins (Dixon et al., 2012; Rao et al., 2014). Loop extrusion relies on the cohesin complex, which loads onto chromosomes mainly at enhancers and promoters (Kagey et al., 2010), then translocates back and forth along DNA until its movement in one direction is halted by an occupied CTCF binding element (CBE). The blocked cohesin spools DNA from the other direction until reaching another bound CBE, generating a transient loop (Davidson et al., 2019; Fudenberg et al., 2016; Kim et al., 2019; Sanborn et al., 2015). As CBEs have asymmetrical sequences, and the CTCF N-terminus binds cohesin (Li et al., 2020), the effectiveness of CBEs for impeding cohesin depends on their linear genomic orientations, with CBEs of convergent orientation providing more durable loop anchors than CBEs in the same orientation (de Wit et al., 2015; Rao et al., 2014). Additional impediments to cohesin can anchor loops, including active transcription and chromatin-bound proteins other than CTCF (Busslinger et al., 2017; Jeppsson et al., 2022; Wutz et al., 2017; Zhang et al., 2019). Homotypic chromatin interactions and chromosome looping each also orchestrate functional interactions between cis-acting DNA elements, most commonly transcriptional promoters and enhancers (Dowen et al., 2014; Flavahan et al., 2016; Hnisz et al., 2016; Ji et al., 2016; Lupiáñez et al., 2015; Narendra et al., 2015). Importantly, the precise molecular mechanisms and functional relationships between these two fundamental means of genome folding and how they regulate biological processes remain matters of broad interest.

A particularly relevant example of the interplay between genome architecture and function is the diversification of antigen receptor (AgR) genes through the process of V(D)J recombination, which is the molecular basis for adaptive immunity in jawed vertebrates. Precursor B and T cells employ lineage- and developmental stage–specific mechanisms that facilitate physical interactions between recombining gene segments separated by up to 3 Mb (Allyn et al., 2020; Carico and Krangel, 2015; Ribeiro de Almeida et al., 2015; Ebert et al., 2015; Jhunjhunwala et al., 2008; Kenter and Feeney, 2019; Proudhon et al., 2015; Zhang et al., 2022). The lymphocyte-specific, heterotetrametric RAG1₂/RAG2₂ (RAG) endonuclease mediates recombination between variable (V), diversity (D), and joining (J) gene segments within a given AgR locus. The RAG complex initially binds a recombination signal sequence (RSS) flanking one gene segment, captures a compatible RSS of a different segment (synapsis), and then cleaves DNA at each segment/RSS junction (Bassing et al., 2002; Schatz and Swanson, 2011). The four liberated DNA ends are repaired by non-homologous end-joining proteins, generating a precise RSS–RSS signal join and a processed V(D)J coding join (Bassing et al., 2002; Schatz and Swanson, 2011). The assembled V(D)J gene segments and downstream constant region exons comprise a complete AgR gene. Notably, RAG-mediated recombination can proceed through deletion or inversion of intervening sequences depending on whether the participating RSSs reside in convergent or identical linear genomic orientation, respectively.

V(D)J recombination is regulated at numerous levels with cell type, developmental stage, and allele specificity. At endogenous AgR loci, RAG targeting is facilitated by the sequential activation of transcriptional enhancers and gene segment-proximal promoters, which render RSSs accessible (Barajas-Mora et al., 2023; Bhat et al., 2023; Krangel, 2003; Majumder et al., 2015a; Sakamoto et al., 2012; Schlissel, 2004; Sleckman et al., 1996). In developing B and T cells, RAG is initially allowed access to DJ segments at Igh and Tcrb, respectively, forming recombination centers (RCs) that mediate short-range D-to-J recombination (Ji et al., 2010b). RAG-bound DJ complexes must then capture a V RSS to conduct V-to-DJ rearrangement. In the later stages of B and T cell development, RAG establishes RCs over Igk and Tcra J segments, respectively, to capture a V RSS for V-to-J recombination (Ji et al., 2010b). The importance of efficiently targeting and completing V(D)J recombination is evidenced by mutations that attenuate RAG endonuclease activity and AgR gene assembly, which cause immunodeficiencies characterized by limited numbers of mature lymphocytes and restricted AgR diversity (Delmonte et al., 2018).

The germline configurations of AgR loci pose a common challenge for their efficient assembly, since V and (D)J segments, separated by up to 3 Mb, must be brought into close spatial proximity for RSS capture. In the absence of locus folding, long-range V-to-(D)J synapsis through diffusion-based collision is highly unlikely and would lead to extremely biased usage of RC-proximal V segments. Indeed, the Igh, Igk, Tcrb, and Tcra/d loci each undergo lineage- and developmental stage–specific contraction to bring V segments into close spatial proximity with their (D)J segments (Fuxa et al., 2004; Jhunjhunwala et al., 2008; Kosak et al., 2002; Roldán et al., 2005; Shih and Krangel, 2010; Skok et al., 2007). Long-range contacts within AgR loci are likely mediated by CBEs that are interspersed among the V segments and oriented convergently with CBEs flanking their respective RCs (Loguercio et al., 2018). Indeed, cohesin and/or CTCF-dependent contacts between convergent V and RC CBEs enhance long-range V-to-(D)J recombination at these four loci (Hill et al., 2023; Seitan et al., 2011; Shih et al., 2012; Majumder et al., 2015b; Zhang et al., 2019). In addition, homotypic interactions between transcriptionally active V and (D)J chromatin have been proposed to fold AgR loci and stimulate V-to-(D)J recombination by facilitating diffusion-based synapsis (Verma-Gaur et al., 2012).

The mechanisms that generate AgR architectures also contribute to which RSSs are allowed to synapse for RAG-mediated recombination. Theoretically, locus compaction through cohesin/CBE-mediated chromosome looping or homotypic chromatin interactions could promote synapsis of RSSs via diffusion-based collisions independent of their genomic orientations, thereby permitting deletional or inversional recombination (Bassing et al., 2008; Jhunjhunwala et al., 2008; Ji et al., 2010b; Ranganath et al., 2008; Wu et al., 2003, 2007). However, next-generation sequencing (NGS) of RAG-mediated rearrangements in the mouse Igh and Tcra/d loci revealed nearly exclusive participation of bona fide or cryptic RSSs in convergent genomic orientation, resulting in only deletional V-to-(D)J rearrangements (Dai et al., 2021; Hill et al., 2020; Zhao et al., 2016). The observed restriction in RSS orientation at these loci led to a mechanistic model wherein cohesin-driven chromosome loop extrusion originates from a RAG-bound RSS and progressively spools DNA past this obstacle, enabling the open RAG active site to linearly and unidirectionally scan loci for capture of a convergent RSS (Ba et al., 2020; Dai et al., 2021; Hill et al., 2020; Jain et al., 2018; Zhang et al., 2019). In support of this RAG scanning model for synapsis, Igh folding in RAG-deficient pro-B cells proceeds predominantly by loop extrusion anchored at the D_HJ_H end of the locus that spools through upstream V_H segments (Hill et al., 2023). Although it remains possible that RAG scanning from RCs is a primary mechanism for long-range V-to-(D)J recombination at all AgR loci, it cannot conduct rearrangements that proceed through inversion between RSSs with identical linear genomic orientations in Igk and Tcrb loci (Ribeiro de Almeida et al., 2015; Lee and Bassing, 2020; Wu et al., 2020). These inversional Igk and Tcrb rearrangements must depend on topological changes that place V and (D)J RSSs in spatial proximity to facilitate synapsis through random diffusion-based collisions.

We now show that Tcrb loci adopt intralocus contacts formed via a combination of homotypic chromatin interactions and bidirectional chromosome loop extrusion between Vβ gene segments and the DβJβ RC in double-negative (CD4⁻CD8⁻, DN) thymocytes poised to undergo recombination. To determine the independent contributions of homotypic interactions and looping in these processes, we focused on the most RC-distal Vβ gene segment (Trbv1) and its single proximal CBE. Because Trbv1 is isolated from all other Tcrb gene segments and CBEs by 150 kb, we could cleanly dissect the impact of promoter-driven compartmentalization and CBE-mediated looping on locus folding and recombination, independent of neighboring transcriptional and architectural cis elements. Indeed, mutation of its proximal CBE had no significant impact on local chromatin but dramatically diminished loop extrusion initiating from the Trbv1 region, nearly eliminating Trbv1-RC loops, contacts, and recombination. Likewise, promoter deletion profoundly impaired loop extrusion emanating from Trbv1, as well as Trbv1-RC loops, contacts, and recombination, despite retaining CTCF occupancy over the Trbv1 CBE. Moreover, unlike at Igh, an RSS engineered to have a flipped linear genomic orientation mediates inversional Trbv1 rearrangements with DβJβ segments, which cannot occur by a RAG scanning mechanism. Therefore, we conclude that Vβ CBE-dependent loop extrusion and resultant chromosome loops, in addition to promoter-driven homotypic chromatin interactions, fold activated Tcrb loci for recombination in DN thymocytes. These cooperative architectural mechanisms augment physical interactions between Vβ and DβJβ segments, facilitating their recombination via diffusion-based synapsis. Our findings have broad implications for how the mechanisms of genome folding interface to regulate other biological processes.

As a prelude to mechanistic studies, we generated a map of DNA contacts for Tcrb loci that were poised to undergo V(D)J recombination. For this purpose, we performed HiC analyses on DN thymocytes from RAG-deficient mice, in which Tcrb gene segments are transcriptionally active but cannot rearrange due to the lack of active recombinase complexes. The most salient features of the Tcrb locus, which spans 670 kb on mouse chromosome 6, are shown in Fig. 1 A (Majumder et al., 2015a). In brief, the most 5′Vβ segment, Trbv1, resides 150 kb upstream of the main Vβ cluster (Trbv2 through Trbv30; 250 kb), with all these Vβ segments configured to rearrange by deletion. In turn, this Vβ cluster lies 250 kb upstream of two Dβ-Jβ-Cβ clusters (Dβ1-Jβ1-Cβ1, Dβ2-Jβ2-Cβ2) and, at the 3′ terminus of the locus, a lone Vβ segment (Trbv31) in an orientation that restricts it to inversional recombination (Fig. 1 A; Majumder et al., 2015a). Regions on either side of the main Vβ cluster contain multiple trypsinogen genes, none of which are expressed in T lineage cells (Majumder et al., 2015a). Transcription within Tcrb is controlled by a single known enhancer (Eβ), which is positioned between the second Dβ-Jβ-Cβ cluster and Trbv31, as well as promoters that lie upstream of each Vβ and Dβ segment (Gopalakrishnan et al., 2013; Mathieu et al., 2000; McMillan and Sikes, 2008; Sikes et al., 2002; Whitehurst et al., 1999). The Dβ promoters and Eβ are required for RAG to access DβJβ segments and for the formation of two independent RCs (Ji et al., 2010a). Tcrb architectural elements include 18 CBEs interspersed among its Vβ segments, all of which are oriented convergently relative to three CBEs flanking the DβJβ segments: 5′PC and CBE1 (27 and 3 kb upstream of Dβ1, respectively) and CBE3 (between Eβ and Trbv31; Fig. 1 A; Loguercio et al., 2018; Majumder et al., 2015b). For simplicity, we refer to Tcrb sequences from 5′PC through CBE3 as the RC region.

To elucidate topological mechanisms that fold Tcrb loci, we complemented HiC analyses for DNA contacts with RNA sequencing (RNA-Seq) and chromatin immunoprecipitation sequencing (ChIP-Seq) to characterize transcriptional and chromatin activities. Our use of RAG-deficient (Rag1^−/−) DN thymocytes eliminated complications that arise from the multitude of Tcrb configurations present in actively recombining thymocytes. A heatmap for DNA contacts was generated from six independent HiC experiments, each performed on thymocytes pooled from at least five mice (Fig. 1 B and Table S1). Because HiC is a population-based analysis, these data presumably reflect a range of Tcrb architectures present in DN thymocytes but reveal several prominent patterns of interactions (Fig. 1 B). First, the main Vβ cluster partitions into a distinct topological domain that interacts more with itself than sequences of flanking domains (Fig. 1 B). This feature may reflect homotypic chromatin interactions since all functional Vβ segments within the main cluster are transcribed and enriched for H3K27ac (active chromatin), whereas its flanking domains lack transcripts and are decorated by H3K9me2 (inactive chromatin; Fig. 1 C). Second, the Vβ cluster exhibits robust contacts with the RC region, which harbors highly transcribed DβJβ segments and chromatin enriched for H3K27ac, while sharply segregating from upstream inactive chromatin spanning silent trypsinogen genes (Fig. 1, B and C). The Trbv12.2 and Trbv13.2 Vβ segments have the most intense contacts with RC chromatin, which correlates with their very high levels of germline transcripts and H3K27ac (Fig. 1, B and C). Third, the isolated and transcriptionally active Trbv1 segment makes a trail of contacts through active chromatin over the main Vβ cluster and the RC, but interacts infrequently with intervening regions of silent chromatin, together forming a segmented stripe originating at Trbv1 (Fig. 1, B and C). These findings are consistent with a role for homotypic chromatin interactions mediated by alternating active and inactive chromatin domains, which folds Tcrb to position all Vβ, Dβ, and Jβ gene segments near each other, but spatially segregate them from silent regions.

To identify anchored loops in Tcrb, we analyzed the HiC data from DN thymocytes with the MUSTACHE algorithm. As shown in Fig. 1, B and C, nearly all of the loop anchors map near (i) CBEs occupied by CTCF and the cohesin subunit RAD21, (ii) transcriptionally active gene segments, or (iii) both (Fig. 1, B and C). Most loops form between Vβ regions with CBEs that have the same linear genomic orientation or between Vβ and RC CBEs that are in a convergent orientation (Fig. 1, B and C). This range of anchoring is exemplified by Trbv1, whose CBE forms loops with a CBE of the same orientation located between Trbv3 and Trbv4, with a region between Trbv10 and Trbv11, and with two of the convergent CBEs, 5′PC, and CBE3, which flank the RC (Fig. 1, B and C). Similarly, the CBE between Trbv3 and Trbv4 loops with CBEs of the same orientation lying near a number of Vβ gene segments (Trbv11, Trbv12.1, Trbv14, Trbv20, or Trbv28), with regions lacking CBEs near Trbv17 or Trbv22, and with all three convergent CBEs flanking the RC (Fig. 1, B and C).

Stripes of contacts emanating from the axis of HiC heatmaps from a cellular population provide information about the direction(s) of loop extrusion across a genetic locus. These architectural features are considered to represent asymmetrical loop extrusion, wherein sequences on the axis are the stationary anchor for loop extrusion, and the stripe results from cells captured at different points along the trajectory of the loop extrusion event (Barrington et al., 2019; Fudenberg et al., 2016; Mirny et al., 2019; Vian et al., 2018). In addition to the segmented stripe originating near Trbv1, striking features emerging from these HiC data include similar segmented stripes emanating from either Vβ or RC regions (Fig. 1 B). As with Trbv1, stripes anchored at Trvb4 or RC CBEs interact more with Vβ gene segments than with intervening silent trypsinogen genes (Fig. 1 B). The presence of these segmented stripes is consistent with cohesin-driven loop extrusion anchored at Vβ or RC CBEs and progressing downstream or upstream, respectively. We envision that this bidirectional loop extrusion is dispersed across cells whereby loop extrusion proceeds from a Vβ CBE to an RC CBE in some cells and from an RC CBE to a Vβ CBE in other cells. The HiC maps indicate that such loop extrusion is impeded more frequently when crossing chromatin stretches that are transcriptionally active and enriched for CTCF, leading to a higher frequency of captured contacts. In contrast, regions that are transcriptionally repressed and depleted for CTCF, particularly the spans of silent trypsinogen genes, likely pose little impediment to loop extrusion, leading to a paucity of captured contacts with the anchor. Together, HiC data support roles for both homotypic chromatin interactions and chromosome looping as cooperative mechanisms that fold Tcrb loci, positioning the collection of Vβ segments into spatial proximity with the two RCs to mediate efficient recombination across large genomic distances and normalized distribution of Vβ segments.

To independently assess functions of chromosome looping and homotypic chromatin interactions in long-range Vβ-to-DβJβ recombination, we focused on Trbv1 for several reasons, including (i) it is the most RC-distal Vβ segment, (ii) it has a downstream CBE and its own promoter, (iii) it is 150 kb away from the remaining cluster of Vβ cluster segments, allowing us to test regulatory elements without influence from neighboring counterparts, (iv) a distinct contact stripe, indicative of loop extrusion, emanates from this region, and (v) Trbv1 is flanked on both sides by transcriptionally inactive, CTCF-depleted chromatin in DN thymocytes. Accordingly, we took a genetic approach, using CRISPR/Cas9-mediated genomic editing to either replace the Trbv1 CBE with a scramble sequence (Jain et al., 2018) or to delete a 1,245 bp region that contains the Trbv1 promoter (Fig. 2 A; Gopalakrishnan et al., 2013). We reasoned that the CBE mutation would abolish CTCF binding without altering Trbv1 transcriptional activity, allowing us to determine whether it loops with convergent RC CBEs by anchoring and/or terminating loop extrusion. Conversely, mutation of the Trbv1 promoter would abrogate transcription and chromatin activation, potentially without reducing CTCF occupancy, enabling us to assess whether the promoter orchestrates homotypic chromatin interactions. Notably, we also expected that Trbv1 promoter inactivation would render the Trbv1 RSS inaccessible to RAG, thereby preventing Trbv1 recombination independent of promoter functions in folding Tcrb.

We first established RAG-sufficient mice with homozygous inactivation of the Trbv1 CBE (V1C^Scr/Scr mice) or promoter (V1P^KO/KO mice) to determine whether these cis elements shape TCRβ repertoires by stimulating long-range Trbv1-to-DβJβ rearrangements. As revealed by flow cytometry, the fraction of thymic αβ T cells with surface expression of TRBV1 was dramatically reduced in each mutant strain (Fig. 2 B). Specifically, TRBV1 was displayed on 5.3% of wild-type (WT) cells, 1.5% of V1C^Scr/Scr cells (a 3.5-fold decrease), and 0.13% of V1P^KO/KO cells (a 40.8-fold decrease; Fig. 2, Β and C). To directly ascertain the effects of each mutation on Trbv1 recombination in DNA from unselected DN thymocytes, we used Adaptive Immunosequencing, a commercial NGS platform. In WT cells, 4.4% of unique Vβ-to-DβJβ rearrangements (in-frame and out-of-frame combined) involved Trbv1, compared with 0.8% in V1C^Scr/Scr cells [a 6-fold decrease] and 0.1% in V1P^KO/KO cells (a 48-fold decrease; Fig. 2 D and Table S2). Notably, residual Trbv1 recombination in both mutants retained a normal distribution of coding joins to Dβ1Jβ1 versus Dβ2Jβ2 clusters (Fig. 2 E), precluding a bias for shorter-range events resulting from secondary Vβ rearrangements to the latter cluster. Collectively, our flow cytometry and sequencing data demonstrate that the Trbv1 CBE and promoter are important determinants of Trbv1-to-DβJβ recombination in DN thymocytes, cooperating to ensure robust representation of TRBV1 within the TCRβ repertoire.

To investigate underlying mechanisms for CBE-mediated recombination of Trbv1, we bred V1C^Scr/Scr mice onto a Rag1^−/− background and profiled Tcrb contacts and chromatin in DN thymocytes. For each Trbv1 mutant, we conducted HiC on two, independent DN samples and combined data (Fig. 3 A and Table S1). The HiC data from WT and mutant samples were down-sampled to reflect an identical number of reads. As compared with WT cells, V1C^Scr/Scr thymocytes had a striking reduction of Trbv1 contacts with other downstream Tcrb elements (Fig. 3, A and B). The prominent segmented stripe emanating from Trbv1 was significantly attenuated in the mutant thymocytes, including its progression through the main Vβ cluster and RC regions (Fig. 3, A and B). Consistent with this observation, MUSTACHE failed to identify loops between the inactivated Trbv1 CBE and other Tcrb elements in V1C^Scr/Scr cells (Fig. 3, A and C). Notably, while very low-level contacts between Trbv1 and sequences spanning other Vβ segments or the RC were detectable in V1C^Scr/Scr thymocytes, interactions between the mutant Trbv1 CBE and convergent 5′PC CBE, which still binds CTCF, were absent (Fig. 3 B). As anticipated, the mutant Trbv1 CBE lacked CTCF occupancy but retained transcriptional activity and H3K27ac levels at Trbv1 (Fig. 3 D). However, we detected a modest extension of H3K9me2 and reductions of H3K7ac and transcripts on V1C^Scr alleles relative to WT alleles (Fig. 3, C and D). These modest changes in chromatin could reflect, respectively, (i) loop extrusion events initiating from the RC that now extend past the inactivated CBE and pull Trbv1 away from the nuclear lamina and (ii) reduced Trbv1 contacts with the RC, which may diminish cooperative effects of homotypic interactions on reinforcing active chromatin modifications. It is possible that the slightly lower activity of Trbv1 chromatin in V1C^Scr/Scr thymocytes contributes to reduced Trbv1 recombination. Nevertheless, these results indicate that the Trbv1 CBE functions as a primary anchor for loop extrusion that originates from the Trbv1 region, generating contacts/loops with downstream Tcrb sequences (Trbv1 stripe), including the RC, and promoting Trbv1 recombination. In addition to this function, our data are consistent with the Trbv1 CBE acting as a terminal anchor for loop extrusion that originates from RC regions, forming loops between the RC and Trbv1 that drive its rearrangement. The low levels of Trbv1 contacts and rearrangements with the RC in V1C^Scr/Scr cells could still proceed through the Trbv1 promoter, which may serve as a less efficient initiating or terminal anchor for loop extrusion compared with alleles that have the functional Trbv1 CBE. We conclude that cohesin-mediated extrusion occurs in both directions across Tcrb, creating chromosome loops anchored at the Trbv1 CBE to juxtapose Trbv1 with DβJβ segments and stimulate their long-range recombination.

To elucidate how the promoter of Trbv1 stimulates its long-range recombination, we profiled Tcrb contacts and chromatin in DN thymocytes from V1P^KO/KORag1^−/− mice. As compared with WT cells, V1P^KO/KO thymocytes had a profound reduction in Trbv1 interactions with all other downstream Tcrb elements (Fig. 4, A and B). Indeed, we observed a nearly complete loss of the segmented stripe originating from Trbv1, as well as a complete loss of loops anchored near this gene segment (Fig. 4, A and C). As expected, Trbv1 promoter deletion abrogated its transcription and dramatically reduced H3K27ac levels across Trbv1 but had no significant impact on CTCF occupancy at the Trbv1 CBE (Fig. 4 D). Our interpretation of these findings is that the promoter activates Trbv1, driving its homotypic interaction with transcriptionally active chromatin spanning the main Vβ and RC regions. Importantly, the promoter mutants lack a detectable loop between convergent Trbv1 and 5′PC CBEs, despite CTCF occupancy at both sites (Fig. 4, C and D). This finding could be explained by several mechanisms that are not mutually exclusive, including (i) Trbv1 promoter-mediated homotypic chromatin interactions stabilize contacts/loops formed with downstream Tcrb elements during loop extrusion from the Trbv1 CBE, (ii) Trbv1 promoter-mediated homotypic chromatin interactions function to stabilize Trbv1-RC loops created by cohesin-driven extrusion from the opposite direction, initially anchored at RC CBEs, and (iii) the Trbv1 promoter serves as a loading site for cohesin complexes that mediate loop extrusion emanating from the Trbv1 CBE. Collectively, these data support multifaceted roles for the Trbv1 promoter in generating Tcrb architecture by contributing to homotypic chromatin interactions and cohesin-mediated loop extrusion emanating at both ends of the locus.

Our evidence supports a mechanistic model in which homotypic chromatin interactions and chromosome loop extrusion emanating from Trbv1 cooperate to fold Tcrb. Importantly, neither mechanism is compatible with synapsis of RSSs solely by unidirectional RAG scanning from the RC, which requires loop extrusion anchored at RCs to progress linearly to Trbv1. Indeed, loop extrusion anchored at the Trbv1 CBE would reel the RC region toward Vβ RSSs rather than move Vβ RSSs past the unoccupied active site of RAG-bound RCs—the foundation of RAG scanning (Fig. 6). Thus, Trbv1-anchored loop extrusion, as well as homotypic interactions, are expected to position Trbv1 and RC RSSs close enough to facilitate diffusion-based synapsis, which is orientation independent.

HiC heatmaps from RAG-deficient lymphocytes provide information regarding topological regulation of AgR loci and potential mechanisms for synapsis of RSSs across enormous genomic distances. However, the only current means of determining whether diffusion-based synapsis can mediate a V-to-(D)J recombination event at Trbv1 is to invert the genomic orientation of its RSS (Dai et al., 2021; Hill et al., 2020; Zhao et al., 2016). Any inversional rearrangements between the reoriented RSS and its normal partner(s) must occur by diffusion-based synapsis, not by active RAG scanning. To determine whether Tcrb can employ diffusion-based synapsis for Trbv1 recombination, we first replaced its RSS with a more efficient version (3′Dβ1 RSS) in either the normal (Wu et al., 2022) or inverted orientation (the V1^R or V1^Ri allele, respectively; Fig. 5 A). As sequences flanking an RSS influence recombination (Boubnov et al., 1995; Gerstein and Lieber, 1993; Hesse et al., 1989; Nadel et al., 1998; Yu and Lieber, 1999), we inserted the last 10 nucleotides of Trbv1 coding sequence fused to the inverted 3′Dβ1 RSS (Fig. 5 A). As expected, the V1^R/R allele rearranged efficiently to produce TRBV1⁺ thymocytes, whereas V1^Ri/Ri mice lack TRBV1⁺ αβ T cells because any inversional rearrangement would fail to create a functional Tcrb gene (Fig. 5, B and C). To quantify inversional Trbv1 rearrangements in the engineered mice, we used Taqman quantitative PCR (qPCR) on DNA isolated from DN3 thymocytes. No inversional rearrangements were found in V1^R/R thymocytes because the RSS orientation only permits deletional recombination (Fig. 5 D). In contrast, signal joins and coding joins created by inversions were detected in V1^Ri/Ri thymocytes (Fig. 5 D). Notably, these inversions occurred with Dβ1Jβ1 segments (Fig. 5 D), indicating that Tcrb can support long-range primary Vβ rearrangement through diffusion-based synapsis.

To determine the relative levels of inversional versus deletional Trbv1 rearrangements on V1^Ri alleles and V1^R alleles, respectively, we developed a Taqman assay to quantify RAG cleavage at each RSS. Using primers that anneal to sequences surrounding each Trbv1 RSS, we can detect RAG cleavage by loss of the amplification signal (Fig. 5 E). Applying this assay to sorted DN3 thymocytes, we detected cleavage at 16.6% of V1^Ri alleles, compared with 31.7% cleavage at V1^R alleles (Fig. 5 F). These data demonstrate that inversional events on V1^Ri alleles proceed at a frequency comparable to deletional events on V1^R alleles. Accordingly, Tcrb can employ diffusion-based synapsis to mediate long-range V-to-(D)J recombination, in addition to potential unidirectional RAG scanning from the RC as at Igh loci (Ba et al., 2020; Dai et al., 2021; Hill et al., 2020; Jain et al., 2018; Zhang et al., 2019). Thus, we conclude that the pathways involved in folding a given AgR locus determine the range of mechanisms deployed for long-range RSS synapsis and, likely, contribute to V gene segment usage in primary Ig and TCR repertoires.

The folding of chromosomes is a dynamic process that creates topological domains to control numerous biological processes by promoting or inhibiting functional interactions between distal DNA elements. We examined how these dynamic mechanisms impact long-range contacts within Tcrb loci, a critical aspect of their assembly into functional AgR genes during T cell development. We found that homotypic chromatin interactions and bidirectional chromosome loop extrusion both contribute to long-range contacts in Tcrb loci, enabling efficient recombination between distal Vβ and DβJβ gene segments. Integrated analysis of our new -omics data supported an important role for homotypic chromatin interactions in establishing contact domains, with robust interactions between transcriptionally active chromatin at the Trbv1 segment, the main Vβ cluster, and the RC region, and segregation of these domains from intervening stretches of silent chromatin. Our work also supported roles for bidirectional loop extrusion initiating from the RC or upstream Vβ segments in distinct cells for folding Tcrb, with resulting chromosome loops placing these Vβ segments near the RC region.

The independent contributions of chromosome loops and homotypic chromatin interactions in forming Tcrb structures and driving its recombination were tested by directed mutations of the most distal Vβ gene segment, Trbv1. Deletion of the Trbv1-proximal CBE revealed that it serves a dual role at Tcrb—acting as a primary anchor for loop extrusion originating at or near this distal gene segment and as a terminal anchor for extrusion emanating from the RC. Accordingly, loss of this single CBE severely diminished locus-wide Trbv1 contacts and its recombination, despite retaining substantial levels of transcription and active chromatin. In this regard, inactivation of the Trbv1 promoter confirmed that a transcriptionally active chromatin environment was critical for its long-range contacts with the main Vβ cluster and with the RC. Indeed, these contacts were retained at a residual level in the CBE but not in the promoter mutant, suggesting that homotypic interactions of active Trbv1 chromatin can drive some long-range contacts independent of loop anchoring, the Trbv1 promoter can act as an alternative anchor for cohesin-mediated loop extrusion, or both. HiC heatmaps from the promoter mutant are consistent with a multifaceted role for this cis element, serving as an anchor for Trbv1→RC loop extrusion, driving Trbv1 compartmentalization to favor its stable interaction with other CBEs, and enabling the nearby CBE to be a terminal anchor for RC→Trbv1 loop extrusion. Finally, we directly tested how these architectural mechanisms determine the range of substrates that can be utilized by RAG complexes to generate AgR genes. Inversion of the Trbv1-RSS continued to support its long-range recombination with DβJβ clusters at an appreciable level, precluding RAG scanning from the RC as the sole mechanism for Tcrb rearrangement. Indeed, these findings further substantiate the involvement of bidirectional loop extrusion, in combination with homotypic chromatin interactions, as vehicles for promoting diffusion-based contacts between distal elements.

Our study also highlights the range of mechanisms employed by developing lymphocytes to ensure efficient assembly of Ig and TCR genes, while also normalizing usage of V segments that are spread over large chromosomal distances from the RC. Prior studies have elegantly shown that Igh is folded in pro-B cells predominantly by unidirectional, cohesin-driven loop extrusion emanating from the D_H-J_H-C_H end of the locus, progressing upstream into the large span of V_H chromatin (Ba et al., 2020; Bhat et al., 2023; Dai et al., 2021; Hill et al., 2020, 2023; Jain et al., 2018; Zhang et al., 2019). This means of locus folding promotes RAG scanning of V_H segments, as the latter spool past the open active site of RC-bound RAG complexes, restricting recombination to only convergently oriented RSSs (Dai et al., 2021; Hill et al., 2020). In this mechanism, V_H CBEs act as a pause for loop extrusion rather than anchors that generate loops to enhance the frequency of diffusion-based synapsis (Dai et al., 2021; Hill et al., 2020). Similarly, our contact maps indicate that one facet of Tcrb folding is a cohesin-driven loop extrusion originating from the RC and progressing through Trbv1, which is consistent with its use of RAG scanning for long-range rearrangements. However, unlike Igh, loop extrusion also originates at high frequencies in the V chromatin regions of Tcrb, progressing in the opposite direction, toward the RC. Together with homotypic chromatin interactions, this bidirectional loop extrusion should allow for diffusion-based synapsis in an RSS orientation-independent manner. Indeed, in marked contrast to Igh, flipping the linear genomic orientation of the RSS attached to Trbv1 enables recombination through inversion, which is not possible by RAG scanning (Dai et al., 2021; Hill et al., 2020). Recent studies have shown that folding of Igk in pre-B cells is accomplished by loop extrusion from RC-proximal CBEs proceeding up through the large Vκ cluster, as well as loop extrusion that is confined within the Vκ cluster (Hill et al., 2023). This contrasts with Tcrb where frequent loop extrusion originating in the Vβ regions proceeds down through the RC. The dual origins of loop extrusion in Igk form multiple adjacent Vκ loops and a single Jκ-Cκ loop that positions Vκ segments near the RC. Creation of this architecture likely is important to facilitate diffusion-based synapsis at Igk since its assembly can occur by either inversional or deletional Vκ-to-Jκ recombination, depending on RSS orientations (Hill et al., 2023).

It remains an open question as to why Tcrb employs mechanisms that promote diffusion-based synapsis given that all upstream Vβ gene segments have RSSs oriented convergently with their partner RSSs in the RC, rearranging by deletion. Perhaps diffusion augments the frequency of long-range contacts that would arise during RAG scanning to enhance overall recombination efficiencies. Alternatively, a chromosome loop existing before RAG binding establishes an RC might increase the efficiency of RAG scanning by providing a structure that facilitates loop extrusion by cohesin anchored at RAG or other features of RC chromatin. Another intriguing possibility may derive from the unique structure of Tcrb, which has two Dβ-Jβ-Cβ clusters and a solitary Vβ gene segment, Trbv31, which lies downstream of the RC and rearranges by inversion. In DN thymocytes that harbor a non-functional Vβ31-Dβ1Jβ1 coding join, secondary recombination is permitted between upstream Vβs and the Dβ2Jβ cluster (Lee and Bassing, 2020). However, in this configuration, all remaining Vβ RSSs and their Dβ2 RSS target are in an identical linear orientation, and any secondary rearrangements must occur by inversion, necessitating diffusion-mediated synapsis.

Our data, coupled with previous observations, support a layered model for the folding of Tcrb into structures that favor long-range Vβ contacts and recombination in DN thymocytes. First, the Eβ super-enhancer engages germline Dβ promoters to generate transcriptionally activated chromatin over DβJβ clusters, enabling RAG binding, RC formation, and DβJβ recombination. Concurrently, promoter-dependent transcription of each Vβ gene segment generates active chromatin locally and renders each downstream RSS accessible to RAG. Simultaneous activation of Vβ and DβJβ chromatin drives homotypic chromatin interactions to stabilize spatial proximity when these regions encounter one another during loop extrusion. In that regard, Eβ, Dβ, and some Vβ promoters likely serve as loading sites for cohesin, which migrates randomly back and forth until it is impeded by CTCF at RC or Vβ CBEs, or by RAG at a DβJβ complex, which then prompts cohesin-driven extrusion in either direction. Regardless of direction, extrusion anchored at an RC or Vβ CBE can terminate at the opposite element, generating a loop that juxtaposes DβJβ complexes and Vβ segments to facilitate their recombination by diffusion-based synapsis (Fig. 6). It is also likely that collision of converging extrusion events forms side-by-side loops that similarly promote long-range Vβ-to-DβJβ recombination. These structure-forming mechanisms likely occur on top of RAG scanning, as loop extrusion anchored at the RC and progressing into Vβ chromatin allows RAG to unidirectionally scan for an accessible Vβ RSS (Fig. 6).

Our discovery, linking distinct mechanisms of folding with recombination at Tcrb, has important implications for other biological processes. As one example, current evidence suggests that during the cellular response to a DNA double-strand break (DSB), cohesin is anchored at each DNA end. Cohesin docking activates unidirectional loop extrusion that allows histone H2Ax phosphorylation (forming γ-H2Ax) by the ATM kinase as chromatin is reeled past kinases at the damage site (Arnould et al., 2021). This process creates γ-H2Ax domains around DSBs that promote end-joining by holding together DNA ends (Yin et al., 2009). The apparent importance of homotypic chromatin interactions in forming Tcrb structures could explain why DSBs within the locus produce γ-H2Ax patterns that reproduce intralocus contacts, presumably by interacting with DSB-proximal ATM, rather than simply spreading out linearly from the break. Specifically, γ-H2Ax accumulates at much higher levels across the interacting active chromatin regions than intervening silent heterochromatin regions (Collins et al., 2020). Thus, either DSBs induced within Tcrb do not activate loop extrusion emanating from each DNA end or homotypic chromatin interactions stabilize contacts during this process to locally increase γ-H2Ax. Beyond orchestrating DNA recombination and repair, regulation of anchoring sites and directions of loop extrusion contributes to proper expression of complex genetic loci at specific developmental stages, including those encoding globin, homeobox, protocadherin, and olfactory receptor genes. Approaches like those employed here can be used to study the contributions of directional loop extrusion and homotypic chromatin interactions toward contacts between the requisite cis-acting elements in health and disease.

Mice used for all experiments were 4–6 wk old, of mixed sex, and housed under specific pathogen–free conditions at the Children’s Hospital of Philadelphia (CHOP) or the Ohio State University (OSU) College of Medicine. All animal husbandry, breeding, and experiments were performed in accordance with national guidelines and regulations and approved by the CHOP Institutional Animal Care and Use Committee and the OSU Institutional Animal Care and Use Committee. Wild-type (C57BL/6J) and Rag1^−/−(B6.129S7-Rag1^tm1Mom/J) mice were obtained from Jackson Laboratories. The V1^R/R mice were previously described (Wu et al., 2022). We used the Easi-CRISPR method (Quadros et al., 2017) of CRISPR/Cas9-mediated genomic editing in C57BL/6 zygotes to create mice with a scrambled Trbv1 CBE (V1C^Scr allele), a deletion of the Trbv1 promoter (V1P^KO allele), or an inverted 3′Dβ1RSS replacement of the Trbv1 RSS (V1^Ri allele).

For the V1C^Scr/Scr mice, we used CRISPOR to identify a suitable guide sequence on the antisense strand to position the cut site centrally within the Trbv1 CBE. The crispr RNA (crRNA) was purchased from Integrated DNA Technologies (IDT) with the identified guide sequence (see Table S3 for a complete list of oligos). The Easi-CRISPR method uses single stranded DNA as donors (ssODN) for homology-directed repair, which carried a previously published scrambled CBE sequence that was confirmed not to bind CTCF (Jain et al., 2018). The CHOP Transgenic Core electroporated zygotes with a mixture of ctRNA (8 μM; crRNA + trans-activating crispr RNA [tracrRNA]), Cas9 protein (3.2 μM), ssODN (20 μM) in Duplex Buffer diluted 1:1 with Opti-MEM buffer. A similar approach was used to generate the V1P^KO/KO mice, with the exception that two different guide RNAs were used to generate a deletion between the two cut sites, with no homology-directed repair. The mixture used to electroporate zygotes contained two different ctRNAs (4 μM each; two different crRNAs independently annealed to tracrRNA) and no ssODN. For the V1^Ri/Ri mice, we utilized the same crRNA as in a previous study to replace the Trbv1 RSS with the 3′Dβ1 RSS in its normal orientation (V1^R; Wu et al., 2022). This crRNA targets the sense strand within the spacer of the Trbv1 RSS. The ssODN harbors the 3′Dβ1 RSS in the inverted orientation plus 10 bases of Trbv1 flanking sequence. V1C^Scr/Scr founders were screened by PCR of tail DNA with the V1Cdelta5′ and CBEScrRev or CBEScrFwd and V1Cdelta3′ primers to test for the presence of the scrambled sequence and to check for appropriate insertion of the 5′ and 3′ homology arms, respectively. V1P^KO/KO founders were screened with the V1PKO5′ and V1PKO3′ primers, which span the region we targeted for deletion. Finally, V1^Ri/Ri founders were screened with the V1RSSComFwd and 3′Dβ1Fwd or 3′Dβ1Rev and V1RSSComRev primers. For each mutant, two founder lines were selected and backcrossed to WT mice for two generations to ensure the propagation of a single validated mutant allele and limit potential off-target CRISPR effects. We crossed heterozygous F2 mice to generate homozygous mice, and genotypes were verified by PCR with the flanking primers, followed by Sanger Sequencing from either end of the PCR product. The sequence-validated founder lines were analyzed by flow cytometry using WT littermate controls to confirm the same phenotype between founder lines. One line was selected for Adaptive Immunosequencing analysis and to backcross onto the RAG-deficient background.

Four replicates of conventional HiC were performed on Rag1^−/− mouse thymocytes as previously described (Collins et al., 2020; Rao et al., 2014). Briefly, 5 × 10⁶ formaldehyde-crosslinked DN cells were lysed with 250 μl of ice-cold HiC lysis buffer (10 mM Tris-HCl, pH 8.0, 10 mM NaCl, 0.2% IGEPAL CA630) containing protease inhibitors (Roche) for 15 min on ice. Chromatin was digested at 37°C for 6 h with DpnII (100 U); ends were filled and marked with biotin using Klenow and ligated together with T4 DNA ligase. Following the reversal of crosslinking, DNA was fragmented on a Covaris E220 Evolution Sonicator and size-selected for 300–500 bp with AMPure XP Beads (Beckman Coulter). DNA ends were repaired with the NEBNext Ultra II DNA Library Prep Kit according to the manufacturer’s instructions using 1 μg of the HiC DNA. The adapter-ligated DNA was size-selected for 300–400 bp with AMPure XP beads and biotinylated DNA fragments were pulled down using MyOne Streptavidin T1 beads (Life Technologies). The final HiC library was generated with 5 PCR cycles using the NEBNext Ultra II DNA Library Prep Kit and NEBNext Dual Index primers (NEB) for Illumina sequencing. For comparison of our mutants, we performed two replicates each with RAG-deficient thymocytes of the WT, V1C^Scr/Scr, or V1P^KO/KO background using the Arima-HiC Kit (A510008) according to the manufacture’s protocol with the KAPA Hyper Prep Kit. We performed these HiC on cells pooled from at least five mice for each genotype. Paired-end sequencing for both HiC methods was performed on an Illumina NovaSeq 6000 (300 cycles) and specific read depths were provided (Table S1).

RNA-Seq was performed on sorted (B220⁻, CD19⁻, CD11b⁻, CD11c⁻, NK1.1⁻, TER119⁻, TCRβ⁻, and TCRγ/δ⁻) DN thymocytes from two Rag1^−/− mice per experiment (see Table S4 for the list of antibodies). RNA extraction, library preparations, sequencing reactions, and bioinformatics analysis were conducted at GENEWIZ, LLC. Total RNA was extracted from frozen cell pellet samples using Qiagen Rneasy Plus Universal mini kit following the manufacturer’s instructions (Qiagen). RNA samples were quantified using Qubit 2.0 Fluorometer (Life Technologies) and RNA integrity was checked with 4200 TapeStation (Agilent Technologies). RNA-Seq library preparation was prepared using Rib0-Zero rRNA Removal Kit and TruSeq Stranded Total RNA library Prep kit following the manufacturer’s protocol (RS-122-2101; Illumina). Briefly, rRNA was depleted with Ribp-Zero rRNA Removal Kit. rRNA-depleted RNAs were fragmented for 8 min at 94°C. First-strand and second-strand cDNA were subsequently synthesized. The second strand of cDNA was marked by incorporating dUTP during the synthesis. cDNA fragments were adenylated at 3′ends and the indexed adapter was ligated to cDNA fragments. Limited cycle PCR was used for library enrichment. The incorporated dUTP in second-strand cDNA quenched the amplification of the second strand, which helped to preserve the strand specificity. Sequencing libraries were validated using DNA Analysis Screen Tape on the Agilent 2200 TapeStation (Agilent Technologies) and quantified by using Qubit 2.0 Fluorometer (Invitrogen) as well as by qPCR (Applied Biosystems). The sequencing libraries were multiplexed and clustered on three lanes of a flowcell. After clustering, the flowcell was loaded on the Illumina HiSeq instrument according to the manufacturer’s instructions. The samples were sequenced using a 2 × 150 Pair-End High Output configuration. Image analysis and base calling were conducted by the HiSeq Control Software on the HiSeq instrument. Raw sequence data (.bcl files) generated from Illumina HiSeq were converted into fastq files and demultiplexed using Illumina bcl2fastq program version 2.17. One mismatch was allowed for index sequence identification.

ChIP-Seq was performed on DN thymocytes pooled from five to seven mice of mixed sex of the RAG-deficient background. 10 million cells were fixed with 2% formaldehyde in complete RPMI media (10% FBS, 1% Pen/Strep) for 5 min (CTCF) or 1% formaldehyde for 10 min (histone marks) at room temperature on an orbital shaker. The crosslinking reaction was quenched with glycine at a final concentration of 120 mM for 5 min on ice and then washed once with PBS. Fixed cells were then resuspended in nuclei isolation buffer (10 mM Tris-HCl, pH 7.5, 10 mM NaCl, 1% IGEPAL) with protease inhibitors (5871S; Cell Signaling Technology) and incubated with rotation at 4°C for 10 min. Next, nuclei were lysed with SDS lysis buffer (1% SDS; 10 mMM EDTA; 50 mM Tris-HCl, pH 8) with protease inhibitor for 20 min on ice and then sonicated with a Diagenode Bioruptor for 20 cycles (1 cycle = 30 s on, 30 s off). Sonicated lysate was spun down to remove cellular debris. At this stage, portions were set aside for input control and confirmation of successful sonication. The rest was diluted with ChIP dilution buffer (0.01% SDS; 1.1% Triton X-100; 1.2 mM EDTA; 16.7 mM Tris-HCl, pH 8; and 167 mM NaCl) plus protease inhibitors and then split between four preprepared antibody–protein A bead conjugates (including an IgG control). Protein A beads (ab214286; Abcam) were washed three times with PBS/T (PBS with 0.02% Tween) and then incubated with rotation for at least 4 h at 4°C with antibody in PBS/T (see Table S4 for a list of antibodies). Following incubation, bead–antibody conjugate was washed three times with PBS/T, added to diluted, sonicated chromatin, and incubated with rotation overnight at 4°C. Next day, the supernatant was aspirated off and the beads were washed with rotation at 4°C for three min each with low salt wash (0.1% SDS; 1% Triton X-100; 2 mM EDTA; 20 mM Tris-HCl, pH 8; 150 mM NaCl), high salt wash (0.1% SDS; 1% Triton X-100; 2 mM EDTA; 20 mM Tris-HCl, pH 8; 500 mM NaCl), LiCl wash (0.25 M LiCl; 1% IGEPAL; 1% sodium deoxycholate; 1 mM EDTA; 10 mM Tris-HCl, pH 8), and then twice with Tris EDTA buffer. Chromatin was then eluted from the beads in two rounds with 50 μl of freshly prepared elution buffer (1% SDS 0.1 M NaHCO₃) at 56°C with rotation. Elute was treated with RNAseA (R1253; Thermo Fisher Scientific) for 4 h and then proteinase K (BP1700100; Thermo Fisher Scientific) overnight at 56°C. Decrosslinked DNA was purified using AMPure XP beads and libraries were generated with the NEBNext Ultra II DNA Library Prep Kit for Illumina (E7645). Libraries were pooled for a target of 20 million (CTCF) or 40 million (histone marks) paired-end reads per sample and sequenced on an Illumina NovaSeq 6000 (200 or 300 cycles). For Rad21 ChIP, we used publicly available data (GSM2973690; Loguercio et al., 2018) and ran it through our computational analysis pipeline.

For HiC, paired-end raw reads were mapped with bwa version (Li and Durbin, 2009), treating forward and reverse reads separately, as described and implemented with the HiCExplorer command tool version 3.6 (Ramírez et al., 2018). For Fig. 1, we combined experimental data from conventional and Arima HiC because they had sufficient pairwise Spearman correlations. For Figs. 3 and 4, we downsampled data from the genotype with the most reads to be equivalent to the number of reads from the genotype with the least reads. Visualization was done via the UCSC genome browser (Kent et al., 2002), using the interact format and juicebox (Robinson et al., 2019). MUSTACHE version 1.2.0 (Roayaei Ardakany et al., 2020) was used to identify significant looping at resolutions 10, 5, and 3 K. Raw RNA-Seq reads were trimmed using bbduk as part of bbmap (version 38.92; https://sourceforge.net/projects/bbmap/). Reads were then aligned to mm10 using STAR (version 2.7.9a; Dobin et al., 2013). bamCoverage from the deeptools package (version 3.5.1; Ramírez et al., 2016) yielded the bigwig coverage files and was run using flags to separate the coverage tracks by strandedness and to normalize to counts per million. Normalized bigwig coverage files were uploaded to the UCSC genome browser (Keng et al., 2002).

For ChiP-Seq, paired-end reads were aligned to the mm10 genome using bowtie2 (version 2.4.5; Langmead and Salzberg, 2012). Peak calling was done using the macs2 algorithm (version 2.2.7.1; Zhang et al., 2008), with an exception for the H3K9me2 mark where the edd (version 1.1.9) software (Lund et al., 2014) was used to identify enriched domains. Bigwig files for the visualization in the UCSC genome browser were yielded by the bamCoverage tool inside the deeptools package (version 3.5.1; Ramírez et al., 2016) applying the reads per genomic content normalization. Finally, the bioconductor package chipqc (version 1.30.0; Carroll et al., 2014) that helped determine our samples had a Reads in Peaks score of at least 4.

Flow cytometry was performed on thymocytes from individual mice. Single-cell suspensions were prepared and treated with red blood cell lysis buffer (140 mM NH₄Cl; 17 mM Tris, pH 7.4). Fc receptors were blocked with anti-CD16/CD32 and antibodies were stained in PBS with 2% FBS and 2 mM EDTA. To determine any effect on gross αβ T cell development and expression of TCRβ chains utilizing the Trbv1 gene segment, we stained with the following panel: CD4, CD8, TCRβ, Vβ2 (TRBV1), and live/dead aqua (L34957; Life Technologies). Data were collected on an LSR Fortessa and analyzed with FlowJo software. Single cells were gated based on forward and side scatter. For statistical analyses, we performed one-way ANOVA followed by Tukey’s post-tests for multiple comparisons.

Adaptive Immunosequencing was performed on sorted DN3 thymocytes (CD4⁻, CD8⁻, B220⁻, CD19⁻, CD11b⁻, CD11c⁻, NK1.1⁻, TER119⁻, TCRβ⁻, TCRγ/δ⁻, CD44⁻, and CD25⁺) pooled from two mice per experiment. Genomic DNA was isolated using the DNeasy Blood and Tissue Kit (69506; Qiagen) and submitted to Adaptive Biotechnologies for their Mouse TCRβ assay at the survey resolution. Read depths for each sample are listed (Table S2). For statistical analyses, we performed multiple unpaired t-tests.

TaqMan qPCR was performed with PrimeTime Gene Expression Master Mix (1055772; IDT) on sorted DN3 thymocytes (CD4⁻, CD8⁻, B220⁻, CD19⁻, CD11b⁻, CD11c⁻, NK1.1⁻, TER119⁻, TCRβ⁻, TCRγ/δ⁻, CD44⁻, and CD25⁺) from two RAG-sufficient mice per experiment. Primer/probe placement to assay for rearrangements (Fig. 5 A) was designed to amplify both coding joins and signal joins resulting from inversional rearrangement to either DJβ1.1 or DJβ2.1 (Fig. 5 C). This PCR strategy will also amplify any inversional rearrangements directly to Jβ1.1 or Jβ2.1. Primer/probe placement to assay for cleavage at Trbv1 (Fig. 5 E) was designed to span the RSS in both V1^R/R and V1^Ri/Ri alleles such that a loss in amplification signal corresponds with RAG cleavage.

Table S1 lists sequencing details for all HiC samples, showing numbers of sequenced reads, numbers, and percentages of paired mappable reads, and numbers and percentages of HiC contact reads. Table S2 lists sequencing details for all Adaptive Immunosequencing samples, showing numbers of total unique complete Tcrb rearrangements, numbers of unique Trbv1 rearrangements, and percentages of Tcrb rearrangements involving Trbv1. Table S3 lists all oligo sequences for generation of genotyping of mouse lines, as well as qPCR. Table S4 lists all antibodies used for the indicated experiment.

The data in Figs. 1, 3, and 4 are openly available in NCBI’s Gene Expression Omnibus and are accessible through GEO Series accession number GSE249649.

We thank Adele Harman of the CHOP Transgenic Core for establishing all of the novel genetically modified mouse strains in this study.

Support provided by National Institutes of Health grants R01 AI130231 (E.M. Oltz and C.H. Bassing), R37 AI118852 (E.M. Oltz), and R21 AI174545 (C.H. Bassing).

Author contributions: E.M. Oltz and C.H. Bassing designed and supervised the study. K. Lee and B.M. Allyn created and analyzed the V1P^KO/KO and V1C^Scr/Scr mice by flow cytometry and Adaptive Immunosequencing. B.M. Allyn bred V1P^KO/KO and V1C^Scr/Scr mice onto the RAG-deficient background and conducted Arima HiC, RNA-Seq, or ChIP-Seq on thymocytes. C. Oyeniran and V. Nganga performed conventional HiC experiments and library preparation for Arima HiC and ChIP-Seq. K.E. Hayer conducted all computational analyses and helped B.M. Allyn prepare figures. A. Sacan mentored K.E. Hayer on computational data. B.M. Allyn wrote the manuscript and prepared figures under the guidance of E.M. Oltz and C.H. Bassing, who worked together to edit the assembled manuscript.