The past decade has increased our understanding of how genome topology controls RAG endonuclease-mediated assembly of lymphocyte AgR genes. New technologies have illuminated how the large IgH, Igκ, TCRα/δ, and TCRβ loci fold into compact structures that place their numerous V gene segments in similar three-dimensional proximity to their distal recombination center composed of RAG-bound (D)J gene segments. Many studies have shown that CTCF and cohesin protein–mediated chromosome looping have fundamental roles in lymphocyte lineage- and developmental stage–specific locus compaction as well as broad usage of V segments. CTCF/cohesin–dependent loops have also been shown to direct and restrict RAG activity within chromosome domains. We summarize recent work in elucidating molecular mechanisms that govern three-dimensional chromosome organization and in investigating how these dynamic mechanisms control V(D)J recombination. We also introduce remaining questions for how CTCF/cohesin–dependent and –independent genome architectural mechanisms might regulate compaction and recombination of AgR loci.
The generation of diverse lymphocyte AgR genes is essential for adaptive immunity in jawed vertebrates (1). The lymphocyte-specific RAG endonuclease generates this diversity by assembling the second exons of Ig and TCR genes via recombination of germline V, D, and J gene segments. RAG binds a J or D recombination signal sequence (RSS), captures a D or V RSS, and upon this synapsis, induces DNA double-strand breaks (DSBs) at the border of each RSS and gene segment (2). RAG functions with DSB repair factors to process and repair these DSBs to create V(D)J coding joins and signal joins (2). V(D)J recombination and lymphocyte development are interdependently regulated. Lymphocyte lineage- and development stage–specific activation of specific AgR locus enhancers and promoters directs transcription-linked accessibility of RSSs to RAG proteins (3, 4). The assembly of TCRβ genes in double-negative (DN) thymocytes or IgH genes in pro-B cells, respectively, proceeds through d-to-J joining followed by V-to-DJ recombination on one allele at a time (5–7). In-frame VDJ rearrangements lead to expression of TCRβ or IgH proteins that, respectively, signal differentiation into double-positive (DP) thymocytes, which assemble TCRα genes or pre-B cells, which assemble Igκ and sometimes Igλ genes. The expression of αβ TCRs or IgH/Igκ or IgH/Igλ BCRs signals differentiation into mature αβ T cells or κ or λ B cells. Likewise, the assembly and expression of TCRγ and TCRδ genes in DN thymocytes through VJ or VDJ joining, respectively, signals development into γδ T cells. The importance for broad use of gene segments in AgR repertoires is highlighted by the requirement of particular V or J gene segments for development of specialized T cell lineages or immunity from specific pathogens (8–11).
IgH, Igκ, TCRβ, and TCRα/δ loci offer a challenge for diverse V segment usage in V(D)J rearrangements as their 20–120 V segments are dispersed across ∼0.6–3 Mb of linear genomic distance (Figs. 1, 2). If synapsis of accessible V and (D)J RSSs occurred solely through stochastic diffusion, then because of polymer physics of DNA, V segment usage would be biased for those located closer to distal (D)J segments (12), which generally does not happen. A seminal paper in 2002 implied that lymphocyte lineage- and developmental stage–specific modulation of AgR locus topology might control V(D)J recombination (13). Fluorescence in situ hybridization revealed that the far distal and proximal sequences of Igh are physically separated in lymphoid progenitors and thymocytes but together in pro-B cells independent of V(D)J recombination (13). Igk, Tcrb, and Tcra/d were found to similarly compact within the lineage and developmental stage in which their V segments recombine (14–18). The discoveries that loss of the B lineage–specific transcription factor (TF) Pax5 impairs Igh compaction and recombination of distal VHs without reducing VH accessibility indicated that compaction and accessibility can be independently regulated (14, 19). The application of chromosome conformation capture–based methods, which quantify interactions between DNA sequences within a cell population, indicated chromosome looping underlies compaction and increases contacts between V and (D)J segments (17, 20–22). About the same time, studies revealed that RAG binds accessible (D)J segments to create a focal recombination center (RC) at each locus (23, 24). These findings formulated the prevailing model that compaction directs efficient assembly of broad AgR gene repertoires by positioning all V segments of a locus in similar spatial proximity to the distal RC (23). In these compacted structures, spatial confinement of V and D-J segments would increase diffusion-based capture (synapsis) of accessible V RSSs by RAG-bound D/J RSSs (25). Computational-based models predict that the relative recombination level of each V segment is determined by its RSS strength, accessibility, and frequency of contact with the RC by means of chromosome looping (12, 26–29). However, these machine learning platforms have not included the influence of coding end sequence on initiation of V(D)J recombination (30–33). Many excellent reviews have thoroughly discussed the numerous studies that shape our knowledge of how genome topology controls V(D)J recombination (34–46). We build on these reviews by emphasizing recent work on elucidating roles of specific CTCF loops.
Principles, mechanisms, and functions of three-dimensional genome organization
The last decade has extensively increased our understanding of how genomes compact in interphase cells. This progress has been fueled by Hi-C, a chromosome conformation capture–based method that employs sequencing to quantify pairwise DNA fragment interactions throughout a genome (47, 48). The first Hi-C analysis of a mammalian cell population revealed that chromosomes broadly segregate into two different compartments: 1) mainly transcriptionally active chromatin or 2) predominantly silent chromatin (47). Higher-resolution Hi-C indicated that chromosomes partition into 0.2–1.0–Mb topologically associated domains (TADs), whose sequences interact with each other more often than with sequences of other TADs (49–52). Although TADs are largely conserved across cell types, even higher-resolution Hi-C revealed TADs often partition further into different classes of tissue-specific sub-TAD contact domains (53–56). Within a TAD, chromosome regions of similar chromatin activity that contact each other define a compartmental domain (CD), whereas sequences between a point-to-point contact of binding sites for the CTCF chromosome-looping protein comprise a CTCF loop. These structures are not mutually exclusive. A CD can reside in one CTCF loop or span many, whereas CTCF loops may contain one CD or more of alternating activities. TADs and sub-TAD CDs are all dynamic, dissolving upon mitosis, reforming as cells enter the G1 phase, and continually dissolving and reforming in G1 (57–60).
The field is seeking to elucidate how CDs and CTCF loops are formed and revised during lineage-commitment and cellular differentiation. The coincidence of CD formation and initiation of transcription in embryos implies that transcription, transcripts, TFs, and/or histone modifications could promote compartmentalization (61–63). The process of liquid–liquid phase separation (LLPS) might also help establish CDs. TFs and additional proteins with disordered domains spontaneously, stochastically, and cooperatively yield droplets via LLPS (64–67). The targeting of such proteins to chromatin drives LLPS, which expels nontargeted sequences and coalesces targeted regions (68). Many CTCF loops occur between convergent CTCF binding elements (CBEs) that are bound by CTCF and the ring-like cohesin protein complexes that link sister chromatids in S/G2/M phases (49, 53, 69). The loss of CTCF or cohesin abolishes CTCF loops (55, 70, 71). Together, these findings yielded an extrusion model for CTCF loop formation, wherein cohesin complexes load onto DNA and thread strands through their rings until halted by CTCF bound at convergent CBEs (72, 73). Consistent with this model, cohesin protein complexes can compact naked and nucleosome-bound DNA in vitro by extending DNA loops symmetrically in both directions (74, 75). Notably, loop extrusion does not encompass how CTCF loops arise between CBEs of the same orientation, as among V segments within the AgR loci. In addition, many CTCF-bound CBEs do not form loops (53, 70) but rather create boundary elements (BEs) that separate CDs by blocking the spread of active chromatin (76). Moreover, inactivation of cohesin but not CTCF alters CDs (55, 70, 71). Such observations suggest that overlapping and distinct mechanisms govern chromosome compaction.
There also is much ongoing effort to determine how chromosome architecture regulates genome function. Most enhancers and their cognate promoters reside within the same CTCF loop anchored by convergent CBEs (76–81). Deletion of a CBE loop anchor abolishes the loop and frees the other anchor to loop with another convergent CBE (76–81). This typically alters gene expression by impairing normal enhancer–promoter contacts and allowing enhancers and promoters previously insulated in adjacent loops to now interact (76–81). Inversion of a CBE loop anchor also abolishes the loop, forms a new loop with another convergent CBE, and can alter enhancer–promoter interactions and gene expression (82). Within CTCF loops, enhancer/promoter interactions (regulatory loops) often occur between the CTCF-related YY1 TF and cohesins with subunits distinct from those at CTCF loop anchors (83–85). Mutation of a YY1 binding site at an enhancer or promoter can reduce enhancer/promoter looping and gene expression without affecting the parental CTCF loop (84). It has been proposed that CDs might stimulate gene expression by concentrating TFs, promoters, and enhancers within transcription factories (86). However, elucidating CD functions in gene regulation will require genetic and biochemical means to alter these contact domains.
The roles of CTCF loops in regulating compaction and recombination of AgR loci
The knowledge that CTCF folds genomes and directs enhancer/promoter contacts fueled studies of CTCF loops in regulating AgR locus compaction and recombination. In compacted Igh, Igk, Tcra/d, and Tcrb loci, CTCF/cohesin binds to numerous CBEs interspersed among V segments and a smaller number of CBEs surrounding (D)J segments (Figs. 1, 2) (87). CTCF or cohesin inactivation in developing mouse lymphocytes stimulates germline transcription and recombination of RC-proximal V segments preferentially over RC-distal V segments, concomitant with increased interactions between proximal Vs and RC enhancers (20, 88, 89). Although these data support roles of CTCF loops in control of AgR locus transcription and recombination, an important caveat when inactivating trans factors is potential indirect effects from altered expression of other genes caused by global changes in genome topology. Accordingly, the field has moved to mutating CBEs. We outline recent studies that identify and elucidate roles of chromosome loops in compacted loci. We first discuss how loops restrict RAG-mediated recombination as this is relevant for all AgR loci.
Chromosome loops control RAG-mediated rearrangements.
The application of high-throughput genome-wide translocation sequencing (HTGTS) has demonstrated that chromosome loops direct and constrain V(D)J recombination events. HTGTS employs primers that anneal near RSSs to identify sequences to which RSSs or flanking coding sequences recombine (90, 91). This assay revealed that RSSs inserted into random genomic locations recombine to fortuitous RSSs only within the same CTCF loop anchored by convergent CBEs (92). RSSs have polarity and recombine through deletion when in opposite orientation or inversion if they are in the same orientation (2). The randomly inserted RSSs preferentially recombine by deletion to fortuitous RSSs of convergent orientation (92). HTGTS showed similar orientation bias and topological restriction in recombination between endogenous DH and JH RSSs or between one of these RSSs and a fortuitous RSS within Igh (92). These observations suggested that RSS-bound RAG protein unidirectionally tracks along the chromosome to capture a bona fide or fortuitous RSS of convergent orientation for synapsis (92). Such unidirectional RAG scanning from a 5′DH RSS was shown to extend further throughout a bigger loop when the 5′CBE anchor of the normal DH–JH–containing loop is deleted (92, 93). In contrast, binding of an inactive Cas9 nuclease or transcription of a repetitive DNA sequence hinders RAG scanning from a 3′DH RSS and increases DH recombination with fortuitous RSSs near the hindrance point (93). The halted RAG scanning is associated with new Igh topology, altered in a manner that implies impediment of cohesin-mediated loop extrusion at the locations where RAG scanning stops (93). These data are consistent with the original model that RAG bound at an RSS moves linearly down the chromosome within preformed loops (92). However, these data also support the current model, in which cohesin-directed threading of DNA past RAG bound to a DH or JH RSS allows RAG scanning-based synapsis and serves a fundamental role in promoting DH-to-JH recombination via deletion rather than inversion (93). Regardless of the mechanism of RAG scanning-based synapsis, the JH-proximal DH DQ52 (Fig. 1) appears to rearrange to JH segments mainly via diffusion-based synapsis, with DQ52 RSSs favoring deletional rearrangements (93), as previously attributed (94). Similarly, diffusion-based contacts of sequences of a CTCF loop may account for synapses that yield rare inversional rearrangements and some fraction of deletional rearrangements. Nevertheless, HTGTS offers firm evidence that looping restricts genomic distances over which RAG-mediated synapsis, cleavage, and joining occurs in cis along a chromosome.
The Igh locus.
Igh is the most studied AgR locus for identifying and elucidating how chromosome loops control recombination (15, 34, 36–40, 43, 44). The locus has over 100 VH gene segments spread across 2.7-Mb upstream of 13 DH segments, 4 JH segments, and CH exons (Fig. 1A). The compacted locus folds into three domains: 1) RC-proximal VHs through the 3′ end of Igh; 2) middle VHs; and 3) RC-distal VHs (Fig. 1A) (95, 96). Domains A and C each have many internal loops between VH CBEs of the same orientation, whereas domain B lacks such internal loops (22, 95–98). Domain A has loops between convergent CBEs at proximal VHs, the 5′DH intergenic control region 1 (IGCR1), and the 3′ end of Igh (3′CBEs) (22, 95–98). In addition to these CTCF/cohesin–dependent loops, the iEμ intronic enhancer exhibits YY1-dependent looping with IGCR1, the 3′Igh enhancer, or 3′CBEs. Compacted Igh loci have long-range interactions between contact domains that include the following: CTCF-dependent loops between domains A, B, or C; YY1-dependent loops between iEμ and the far end of domain A or the middle of domain C; and Pax5-dependent folding of domain C and interactions between domains A and C (95, 99). Unlike YY1 loops identified in other cells and loci, YY1-dependent Igh loops are anchored by ring-link condensin protein complexes (100), which compact chromosomes in M phase (101). Igh configurations in pro-B cells lacking CTCF, cohesin, YY1, or Pax5 suggest a stepwise mechanism for compaction. In this model, CTCF folds Igh into several VH subdomains and an RC subdomain, Pax5 coalesces distal VH subdomains, and YY1 juxtaposes the VH and RC regions (100). Yet, considering lower than normal Pax5 expression in YY1-inactivated cells and potential YY1/CTCF contacts (99, 102–104), elucidating how Igh folds to regulate recombination requires inactivation of specific binding sites for CTCF, YY1, Pax5, and other possibly relevant proteins. It also will be important to ascertain if the ubiquitously expressed YY1 protein and/or T lineage–specific TFs help fold TCR loci and regulate TCR recombination.
The mutation of CBEs has proven that CTCF loops regulate Igh topology, transcription, and recombination. Igh has ∼100 VH CBEs that are convergent with 9 3′CBEs (87, 105) (Fig. 1A) VH CBEs are convergent with the 5′CBE of IGCR1 (CBE1), and 3′CBEs are convergent with the 3′CBE of IGCR1 (CBE2) (Fig. 1A). Inactivation of CBE1 and CBE2 in mice abolishes IGCR1 looping with iEμ and 3′CBEs, reduces iEμ looping with distal VHs, and enables looping between proximal VH CBEs and iEμ or 3′CBEs (98, 102). These new loops require CTCF, YY1, and iEμ and allow iEμ to increase transcription and recombination of proximal VH segments (98, 102). CBE1 and CBE2 mutation also allows proximal VH segments to escape IgH-mediated feedback inhibition and to recombine in thymocytes, where iEμ is normally active and drives DH-to-JH rearrangements (98). In pro-B cells, the Igh RC forms on a 4-kb loop between iEμ and the 3′DH promoter (22). HTGTS of Abelson-transformed (v-Abl) pro-B cells with IGCR1+ or IGCR1− alleles showed that IGCR1 halts RAG scanning from DJH complexes (92). IGCR1 deletion allows RAG binding at proximal VH segments (102), likely from extended RAG scanning, increased diffusional contact of these VH segments with the RC, or both. Inactivation of CBE1 or CBE2 has similar albeit lesser effects as mutation of both (98, 102, 106); whereas, the deletion of seven 3′CBEs similarly reduces IGCR1/3′CBE looping and increases proximal VH rearrangements (107). These data support a model wherein a CBE2/3′CBE loop promotes DH-to-JH recombination by focusing activities of RAG and iEμ and by increasing DH/JH interactions (98, 102, 106, 108). A simultaneous CBE1/VH CBE loop would restrict VH recombination by lowering VH/RC contacts and insulating VH segments from iEμ-mediated accessibility and RAG scanning (98, 102, 106, 108). In B lineage cells, CTCF loop extrusion appears to progress mainly from 3′CBEs (109). This, coupled with the large number of 3′CBEs, might direct ordered Igh assembly by rapidly, frequently, and repeatedly establishing stable CBE2/3′CBE loops. Potentially slower and less frequent formation of loops between 3′CBEs and proximal VH CBEs after occasional extrusion past IGCR1 CBEs might lower the likelihood of biallelic initiation of VH rearrangements before feedback inhibition halts further VH recombination to enforce Igh allelic exclusion.
Despite their proposed role in restraining VH recombination, proximal VH CBEs are vital for efficient VH rearrangements. A CBE lies just 3′ of each proximal VH RSS (87, 108). Mutation of the VH81X CBE in mice nearly abolishes recombination of VH81X, but increases rearrangements of upstream proximal VH gene segments (110). In a pro-B cell line with a preassembled DJH complex, VH81X loops with 3′CBEs and the IGCR1/DJH/iEμ region and recombines to the DJH complex (110). Upon VH81X CBE deletion, these loops are abolished, and rearrangements decrease ∼50-fold (110). Mutation of the VH2-2 CBE in v-Abl cells nearly abolishes VH2-2 recombination and increases upstream proximal VH rearrangements without affecting recombination of the downstream VH18X (110). These data imply that proximal VH CBE looping with IGCR1 and/or 3′CBEs stimulates proximal VH rearrangements. The loops could promote RAG-mediated synapsis by placing VH RSSs in spatial proximity to DJH RSSs or by halting RAG scanning near VH RSSs, depending on whether a loop forms before or after RAG binds a DJH RSS. Notably, VH81X CBE inversion in v-Abl cells downregulates only 2-fold VH81X looping and rearrangement to DJH complexes (110). It is not clear how the inverted VH81X CBE loops with 3′CBEs of the same orientation and whether the loops or CTCF binding near the VH81X RSS, independent of looping, halt RAG scanning. Furthermore, whether VH81X CBE deletion or inversion alters transcription-linked accessibility of the VH81X RSS remains unreported. For Igh and other AgR loci, most V segments lack a CBE at their RSS and, instead, have a CBE upstream of their promoter or further away, even on the far side of several other V segments (87, 105). Therefore, the deletion and inversion of distinct types of V region CBEs linked with analyses of loop extrusion, RAG binding, synapsis, and accessibility is required to rigorously elucidate how CTCF loops control AgR locus compaction and recombination.
The Igκ locus.
As compared with Igh, less is known regarding how CTCF loops regulate Igk topology and recombination. The locus has 96 functional Vκ gene segments spread across 3.2 Mb upstream of 4 functional Jκ segments and Cκ exons (Fig. 1B). The compacted locus folds into five contact domains, four spanning Vκs (domains A–D) and one spanning 3′Vκs and the RC (domain E) (Fig. 1B). Convergent CBEs bound by CTCF/cohesin demark each domain (Fig. 1B) (41, 111–113). Loops form among Vκ segments as well as between Vκ segments and RC-flanking regions (Fig. 1B) (41, 111–113). These RC regions include the contracting element for recombination (Cer), silencer in the intervening sequence (Sis), iEκ enhancer, and 3′Eκ enhancer (Fig. 1B) (41, 111, 112, 114, 115). Most RC-proximal Vκ CBEs are in the same orientation as the Cer CBEs, one Sis CBE, and several 3′Igκ CBEs. In contrast, most RC-distal Vκ CBEs are convergent with one Sis CBE and other 3′Igκ CBEs (Fig. 1B) (87). CTCF loss in pre-B cells increases interactions of proximal Vκ segments with iEκ and 3′Eκ and stimulates transcription and recombination of these Vκ segments (41). This observation implies that CTCF looping within Igk is not required for long-range contacts and rearrangements but rather reorganizes contacts and insulates proximal Vκ segments from RC enhancers (41). The deletion of Sis in mice elevates transcription and recombination of proximal Vκ segments without effecting Igκ compaction as assayed by fluorescence in situ hybridization (115). Sis deletion decreases Igk localization with heterochromatin, likely from loss of binding sites for the Ikaros TF rather than deletion of CBEs (115). Cer deletion in mice and v-Abl cells impairs compaction and recombination of distal Vκ segments and increases transcription and recombination of proximal Vκ segments, even allowing proximal Vκ rearrangements in thymocytes (114, 116). The deletion of Cer and Sis has the combined phenotype of each deletion (117). Inversion of Cer in v-Abl cells increases rearrangement but not transcription of proximal Vκ gene segments and redirects Cer contacts from Vκs to sequences 3′ of Igk (116). These data suggest that both Sis and Cer anchor CTCF loops to insulate proximal Vκ segments from Igk enhancers, whereas only Cer establishes loops with Vκ CBEs to fold Igk and thereby promote Vκ rearrangements. However, specific inactivation of Cer and/or Sis CBEs without affecting other TF binding sites is required to unequivocally elucidate functions of Cer- and Sis-anchored CTCF loops. Moreover, analyses of Cer and Sis mutants on RAG-deficient backgrounds in which Vκ rearrangements cannot occur might provide further insights into how CTCF loops regulate Igk topology, transcription, and recombination.
The Tcrα/δ locus.
The control of Tcrα/δ topology, transcription, and recombination is unique among AgR loci, as the locus contains two different AgR genes that assemble within distinct developmental stages. The locus has ∼110 Vα/δ segments upstream of 2 Dδ segments, 2 Jδ segments, Cδ exons, and a Vδ (Trdv5) in opposite transcriptional orientation, which is followed by 61 Jα segments and Cα exons (Fig. 2A). In DN thymocytes, Tcra/d is fully compacted, and 3′Vα/δ segments rearrange to Dδ segments; whereas, in DP thymocytes, the locus adopts a 5′-decontracted/3′-contracted topology, and 3′Vα/δ segments recombine with 5′Jα segments (18). In most DP cells, many successive Vα-to-Jα rearrangements occur on each allele before positive selection of αβ TCRs (118). The extended Tcra/d conformation likely facilitates progressive 3′-to-5′ usage of V segments to ensure broad representation of Vα segments within the naive αβ TCR repertoire (18). In both DN and DP cells, CTCF/cohesin binds to ∼100 V CBEs, 2 intergenic CBEs (INT1, INT2) between Trdv2-2 and Trdv4, a CBE 5′ of the T early α (TEA) 5 Jα promoter, a CBE 3′ of the Eα enhancer, and a CBE further downstream of Eα (20, 87, 89, 119) (Fig. 2A). The similar patterns and extents of CTCF/cohesin binding between DN and DP thymocytes implies that additional factors help CTCF loops shape developmental stage–specific Tcra/d locus topologies (87, 89).
The deletion of Tcrα/δ CBEs has shown CTCF loops can control topology and recombination independent of transcription-linked accessibility. The INT1 and INT2 CBEs and a CBE (INT3) between Trdv2-1 and Trdv1 are convergent with the TEA CBE (Fig. 2A) (119). In DN thymocytes, INT2 and TEA form a loop that contains Trdv4, Dδ-Jδ segments, and Trdv5, whereas INT1 contacts proximal Vδ segments (119). Deletion of INT1 and INT2 abolishes the INT2/TEA loop and yields a new INT3/TEA loop that includes Trdv2-2, which is more accessible than Trdv4, Trdv5, and other Vδ segments (119). The INT3/TEA loop has no effect on Trdv2-2 accessibility, perhaps because of a limited linear distance across which Eδ functions and/or Eδ-independent activity of the Trdv2-2 promoter. However, this new loop elevates Trdv2-2 contact with the Dδ-Jδ segment, causing Trdv2-2 to outcompete Trdv4, Trdv5, and other Vδ segments for rearrangement (119). RAG normally binds Dδ RSSs and scans the INT2/TEA loop, generating deletions with convergent fortuitous RSSs (120). On INT1/INT2-deleted alleles, RAG scanning extends throughout the larger INT3/TEA loop (120). These data imply that an INT2/TEA loop insulates Trdv2-2 from RC contact and RAG scanning to prevent Trdv2-2 from outcompeting the invariably less accessible Trdv4 and Trdv5 segments for rearrangement to Dδ segments (119). Simultaneous heterogenous loops between INT1 and other various proximal Vδ CBEs in the DN thymocyte population likely facilitate broad Vδ usage during TCRδ gene assembly (119).
The insertion of a CBE into Tcrα/δ implies that CTCF loops can also control V(D)J recombination independent of accessibility and RAG scanning. A CBE inserted between Eδ and Trdv5 binds CTCF and loops with the INT2 CBE, despite these CBEs residing in the same orientation (121). RAG scanning from 3′Dδ RSSs is blocked at the new CBE, indicating that a chromosome loop between CBEs of the same orientation can restrict rearrangements (121). The new CBE also decreases interaction and recombination of Trdv5 with Dδ segments, without altering Trdv5 transcription or chromatin activity (121). As the Trdv5 RSS resides in the same orientation as its partner 5′Dδ RSS and thus rearranges by inversion, a halt in RAG scanning cannot explain the reduced Trdv5 recombination. On a normal allele, RAG scanning from 3′Dδ RSSs extends to the Trdv5 RSS (121), but the incompatibility of these RSSs blocks deletional Trdv5-to-Dδ rearrangements. Thus, the simplest most plausible explanation for the ectopic CBE reducing Trdv5 rearrangements is that Trdv5 exclusion from the new loop reduces normal diffusion-based contacts and synapses between Trdv5 and 5′Dδ RSSs (121).
CTCF loops have been implicated in regulating Tcrα/δ topology, transcription, and recombination in DP thymocytes. In this developmental stage, contacts form among 3′Vα segments, TEA, 5′Jα segments, and Eα, all of which have a CBE bound by CTCF/cohesin (20, 87, 89). CTCF loss in DP thymocytes decreases these contacts, TEA transcripts, and Vα-to-Jα recombination (89). At the same time, Eα gains interactions with Tcrδ segments, and Tcrd transcripts are elevated (89). Inactivation of cohesin in DP cells also reduces Eα contacts with TEA and 5′Jα segments and Vα-to-Jα recombination, although it increases Tcrδ transcripts (20). Upon DN-to-DP development, CTCF binding increases at 3′Vα segments and a 5′Jα concomitant with both activated transcription of these segments and their interactions among each other, TEA, and Eα (89). The deletion of Eα and its associated CBE blocks increased CTCF binding, interactions, and transcription of these elements and nearly abolishes germline Jα transcription and Vα-to-Jα recombination (89). Deletion of TEA and its CBE allows Eα to contact and drive transcription and recombination of Tcrδ segments in DP cells (89). However, inhibition of TEA transcription does not have these effects (89). Collectively, these data imply that Tcra/d CTCF loops are not required for long-range contacts and rearrangements in DP cells. Instead, these loops might reorganize contacts between gene segments and insulate gene segments from Eα to sculpt Vα and Jα usage. Finally, the transcriptional activation and looping of TEA and Eα in DP cells without changes in CTCF binding at either element implies topological roles for TFs, transcripts, and/or active chromatin at TEA and Eα (89). Independent mutation of CBEs or other TF binding sites within TEA, Eα, and Vα promoters is essential to elucidate how CTCF loops form and control Tcrα/δ topology, transcription, and recombination in DP thymocytes.
The Tcrβ locus.
Tcrb is the least studied AgR locus in regard to the role chromosome looping plays in regulating V(D)J recombination. It contains 21 Vβ segments upstream of 2 Dβ-Jβ-Cβ clusters, followed by a single Vβ (Trdv31) in the opposite transcriptional orientation (Fig. 2B). In DN thymocytes, CTCF/cohesin binds to 18 Vβ CBEs, 3 CBEs (5′PC, C1, and C2) upstream of Dβ1, and a CBE (C3) between the Tcrb enhancer (Eβ) and Trbv31 (Fig. 2B) (26, 87, 122). Compacted Tcrb loci show interactions between upstream Vβ segments as well as between upstream Vβ segments and the RC end of the locus (17, 26, 122). Although CTCF loops have not been identified within Tcrb, the inactivation of CTCF decreases recombination of Vβ segments located within 1 kb of a CBE and increases recombination of Vβ segments lacking a nearby CBE (123), implying CTCF loops might promote Vβ rearrangements. Deletion of a 3-kb region spanning C1 and C2 increases recombination of proximal Vβ segments (Trbv15–Trbv30) and reduces recombination of distal Vβ segments (Trbv1–Trbv14) (122). The deleted region normally serves as a BE that blocks the spread of active chromatin from the RC. On normal alleles, 5′PC contacts distal Vβ segments and tethers them near the RC, likely by forming CTCF loops with convergent Vβ CBEs. However, when the normal BE is lost, 5′PC instead functions as a BE that prevents the spread of active chromatin, which inactivates the 5′PC tethering function. CTCF binding to C1 or C2 and/or the insulating activity of a residual viral long-terminal repeat upstream of C1 could establish the normal BE (122, 124). Notably, the loss of 5′PC tethering also reduces interactions between distal and proximal Vβ segments without affecting contacts between distal Vβ segments, between proximal Vβ segments, or between proximal Vβ segments and the RC (122). This suggests that proximal and distal Vβ segments fold into distinct contact domains whose interaction with each other depends on tethering of distal Vβ segments to the 5′PC. Testing this possibility and elucidating how CTCF loops direct compaction and recombination of Tcrb requires mutation of specific CBEs within the locus.
General concepts and questions for chromosome looping control of AgR loci.
Current knowledge implies general concepts but also raises locus-specific questions for how chromosome looping controls Igh, Igk, Tcra/d, and Tcrb topology, transcription, chromatin, and rearrangement. In each of the activated loci, the RC and associated transcriptional enhancer(s) reside within a prevalent CTCF loop that focuses enhancer-mediated activation of transcription and chromatin (Figs. 1, 2). For the loci studied by specific inactivation (Igh) or insertion (Tcra/d) of a CBE, this loop also restrains RAG scanning and normalizes interaction and recombination of D/J segments with all V segments. Immediately upstream of the RC loop of each locus lies at least one CBE of convergent orientation with many V CBEs (Figs. 1, 2). Dynamic CTCF/cohesin–mediated looping between these 5′RC CBE(s) and V CBEs appears to drive efficient long-range recombination and normalize usage of individual V segments. The many V CBEs themselves seem to form loops that organize V segments into subdomains and promote interactions among the subdomains (Figs. 1, 2). One question is whether the concentration of these CBEs, additional factors, or both facilitate loops between V CBEs of the same orientation. Notable differences among loci include that all V CBEs of Igh and Tcrb are in convergent orientation with 5′RC CBE(s), although some V CBEs of Igk and Tcra/d reside in the same orientation as 5′RC CBE(s). Furthermore, of all the loci, only the TCRα RC lacks a 3′CBE in convergent orientation with V CBEs (Figs. 1, 2). Igk is unique in that it contains many distal V segments in opposite transcriptional orientation as Jκ segments and supports long-range V recombination through inversion of intervening sequences. As many as four successive Vκ-to-Jκ rearrangements can occur on each allele before assembly and positive selection of an Igκ gene or initiation of Igλ locus recombination (125). Another question is whether the Vκ CBEs originally in the same orientation as RC CBEs promote initial Vκ-to-Jκ rearrangement through inversion and then subsequent deletional recombination of Vκ segments once they are in the same orientation as Jκ segments. Additional questions include whether the presence of convergent CBEs among Vα/δ segments and/or the lack of a 3′RC CBE in convergent orientation with V CBEs is important for assembly and selection of Tcra genes. The ability to conduct up to 50 successive Vα-to-Jα rearrangements on each allele is necessary for the development of a replete and diverse αβ T cell population (118). It is possible that formation of loops between convergent Vα/δ CBEs helps establish the 5′-decontracted Tcra/d conformation and thereby lowers usage of the RC-distal Vα/δ segments in initial Vα-to-Jα rearrangements. In addition, the infrequent formation or limited duration of CTCF loops between Vα/δ CBEs and the TCRα 3′RC CBE of the same orientation may lower the efficiency of successive Vα-to-Jα recombination to facilitate expression of Tcra genes and subsequent selection of αβ TCRs. Clearly, additional studies that test these general concepts and questions are essential to fully elucidate how chromosome looping regulates AgR gene assembly.
Contributions of CTCF/cohesin–independent architectural mechanisms in recombination
Neither inactivation of CTCF nor cohesin abolishes interactions and rearrangements between V and D-J segments, implying that CTCF/cohesin–independent mechanisms help promote locus compaction and V rearrangements (20, 88, 89). Indeed, the YY1 and Pax5 TFs also fold Igh and stimulate distal VH rearrangements (14, 19, 99, 100, 103). Although YY1 likely functions with condensins to form loops between VH gene segments and the RC (100), it is unclear how Pax5 promotes interaction and recombination of VH segments with the RC. A possibility is that incorporation of distal VH regions, Pax5-activated intergenic repeat (PAIR) promoters, and the iEμ enhancer into the same transcription factory enables locus folding independent of transcription per se (99). A similar promoter/enhancer–dependent mechanism has been proposed to drive contacts and rearrangements between Vα and Jα segments (89). Compacted Igk loci contain many interaction hubs, with most exhibiting binding of the E2A TF and histone modifications indicative of active enhancers (111, 113). The major hub is a Vκ intergenic enhancer (E88) near the domain C and D border (113). E88 contacts CBEs and other enhancers throughout the Vκ and RC regions (113). Deletion of E88 lowers RC interaction and recombination of Vκ segments of domains C and D, whereas it increases RC contact and rearrangement of Vκ segments of domains A, B, and E (113). Inactivation of E88 reduces germline transcripts of only a few flanking Vκ segments, suggesting that E88 functions as an architectural element that folds Igk to shape Vκ usage rather than a widespread transcriptional enhancer (113). Although the effect of E88 deletion on CTCF or cohesin occupancy was not reported, the fact that CTCF inactivation does not discernably alter interactions between the RC and domains A–D suggests E88 structural activity is CTCF-independent (88, 113). The E88 binding sites for E2A and the EBF TF are both required for normal Igk folding (113). Such data suggest enhancers, promoters, and TFs may drive AgR locus compaction through mechanisms that underlie contact domain formation. Consistent with this idea, published HiC data imply that V segments and RC-proximal enhancers of compacted Igh or Igk loci contact each other within a CD (95, 113). Yet, normal levels of contacts between Vβ segments and the RC on Eβ-deleted alleles argues such transcription-associated mechanisms do not function at Tcrb (122) or possibly involve enhancers downstream of the RC. A major challenge is to elucidate how CTCF/cohesin–independent mechanisms might orchestrate compaction and recombination of AgR loci.
Basic mechanisms that regulate genome topology have been shown to also control V(D)J recombination. However, we still know little of how AgR loci topologies form, change throughout B and T lymphocyte development, and orchestrate tissue- and developmental stage–specific gene assembly. The small number of reported CBE manipulations have indicated that specific CTCF chromosome loops can do the following: restrict and direct RAG scanning-mediated synapsis, insulate proximal V segment promoters from RC enhancers, and/or enable interactions between V and D/J segments. The inactivation, inversion, and switching of additional CBEs throughout AgR loci is vital to ascertain the complete array of mechanisms by which CTCF loops regulate AgR locus architecture and assembly. The modest effects of CTCF or cohesin inactivation on the topology and recombination of Ig and TCR loci are consistent with the current consensus in the chromosome biology field that some of the dynamic mechanisms of three-dimensional genome architecture operate independently of CTCF and cohesin. However, it is important to determine whether these modest effects are due to incomplete loss of protein; potentially, the use of the auxin-inducible degron system could discern this (70). Regardless, the recognition that some CTCF/cohesin–independent genome-folding mechanisms involve promoters, enhancers, TFs, and the transcripts that they produce might fuel new avenues of study that investigate potential mechanistic relationships between AgR locus accessibility and topology. It will also be important to investigate whether chromosome architectural components function in potential lymphocyte lineage-specific (e.g., YY1) and AgR locus-specific (e.g., Pax5) capacities. Future studies also should consider how association with nuclear lamina and additional nuclear structures may influence locus topology to regulate V(D)J recombination (126). It also is vital to move beyond population-based analyses by adapting single cell technologies, including live cell imaging (25), to completely elucidate the diverse dynamic mechanisms that organize AgR locus topology, accessibility, RAG scanning, and V(D)J recombination. Finally, considering that many human diseases may arise from alterations in genome topology that impact gene expression (127), the knowledge that emerges from studies of genome topology regulation of AgR gene assembly may illuminate pathological mechanisms and possibly novel therapeutic avenues for clinical management of some types of immunodeficiency, autoimmunity, and lymphoid malignancies.
We apologize to our many colleagues whose primary papers we regrettably are not able to cite because of space limitations. We thank Rebecca Glynn for help editing the manuscript.
This works was supported by the National Institute of Allergy and Infectious Diseases, National Institutes of Health Grants R01 AI112621 and R01 AI130231 (to C.H.B.) supported this work.
Abbreviations used in this article:
CTCF binding element
5′CBE of IGCR1
3′CBE of IGCR1
3′ end of Igh
contracting element for recombination
high-throughput genome-wide translocation sequencing
intergenic control region 1
liquid–liquid phase separation
recombination signal sequence
silencer in the intervening sequence
topologically associated domain
T early α
The authors have no financial conflicts of interest.