Transcription-Dependent Mobilization of Nucleosomes at Accessible TCR Gene Segments In Vivo Free

Antigen receptor loci contain multiple variable (V), diversity (D), and joining (J) gene segments that are assembled into complete Ag receptor genes by V(D)J recombination (1–3). This sequence-directed recombination process is mediated by the RAG 1 and 2 proteins (hereafter referred to as RAG1/2). Ag receptor gene segments are flanked by recombination signal sequences (RSSs) composed of conserved heptamer and nonamer elements separated by 12 or 23 bp spacers (12 RSS and 23 RSS, respectively). RAG1/2 binds to these RSSs and catalyzes the formation of double-strand breaks at the heptamers of two different RSSs (one 12 and one 23 RSS) held in a synaptic complex. The four ends are subsequently repaired via the nonhomologous DNA end joining pathway to create an imprecise coding joint between the two gene segments and a precise signal joint between the two RSSs.

Because the introduction of double-strand breaks can lead to genomic instability, the process of V(D)J recombination is highly regulated (4–6). V(D)J recombination is restricted to developing T and B lymphocytes, because only these cells express the RAG1/2 proteins. In addition, each Ag receptor locus recombines in a particular cell lineage and developmental stage. For example, complete V(D)J recombination of the Igh and Igk loci occurs in developing pro- and pre-B cells, respectively. Similarly, V(D)J recombination of the Tcrb, Tcrd, and Tcrg loci occurs in CD4⁻ CD8⁻ double-negative (DN) thymocytes, whereas the Tcra locus recombines in CD4⁺ CD8⁺ double positive thymocytes.

It is generally accepted that locus-specific regulation of V(D)J recombination is enforced largely by alterations in the accessibility of chromatin-embedded RSSs to the RAG1/2 enzymatic complex (7, 8). Studies of Ag receptor loci in vivo have correlated RSS targeting by the recombinase to an active chromatin state, established by enhancer and promoter elements, that is characterized by germline transcription, increased nuclease sensitivity and active histone modifications (e.g., histone H3 acetylation [H3Ac], histone H4 acetylation, and histone H3 lysine 4 di- and trimethylation [H3K4me2 and H3K4me3]) (6, 9). Although an active chromatin state positively correlates with recombinase accessibility, an inactive chromatin state negatively correlates with accessibility. Moreover, establishment of inactive chromatin can suppress V(D)J recombination, arguing that an active chromatin state is necessary for V(D)J recombination (10).

In vitro studies have demonstrated that assembly of RSSs into nucleosomes reduces the efficiency of the recombination reaction (11–14), suggesting that nucleosomes can impede RAG1/2 binding or catalytic activity. This barrier may then be surmounted by ATP-dependent chromatin remodeling complexes, such as switch/sucrose nonfermenting (SWI/SNF), which disrupt nucleosome structure and organization. Indeed, Brg1, the catalytic subunit of SWI/SNF, is bound to Ag receptor gene segments when they are accessible to RAG1/2 in vivo (15, 16). Addition of SWI/SNF increased the ability of RAG1/2 to cleave RSSs assembled into nucleosomes in vitro (11, 12, 17). Moreover, in vivo experiments have shown that recombination is stimulated by recruitment of Brg1 and suppressed by depletion of Brg1 (16). Thus, it is clear that SWI/SNF-dependent chromatin remodeling is critical for RSS accessibility and V(D)J recombination in vivo.

Chromatin is now understood to have not only a suppressive role in V(D)J recombination but a critical positive role as well. RAG2 contains a plant homeodomain finger in its carboxy-terminus that specifically binds H3K4me3 (18–20). This histone modification is tightly linked with RNA polymerase II activity and transcription (21), and marks accessible regions of Ag receptor gene loci in vivo (18). In vitro and in vivo experiments have shown that RAG2 interactions with H3K4me3 are necessary for stable recruitment (18, 19) and optimal catalytic activity (22) of the recombinase complex. Together, these findings suggest a need for H3K4me3 modified nucleosomes at or adjacent to an RSS, and provide an explanation for the critical role of germline transcription at Ag receptor loci in vivo (23, 24).

Despite the importance of chromatin structure to developmental regulation of V(D)J recombination, very little is known about nucleosome organization at Ag receptor loci. In vitro experiments have suggested that the RSS nonamer functions as a nucleosome positioning sequence (25). However, it is unclear whether this in vitro observation pertains at Ag receptor loci in vivo. Maes et al. (26) analyzed nucleosome organization at J_H gene segments in vivo, and suggested the presence of distinct and nonrandom nucleosome arrays in pro-B and liver cells. Further high-resolution analysis of pro-B cells demonstrated susceptibility to micrococcal nuclease digestion at the J_H RSSs. However, a complete picture of nucleosome organization and occupancy in accessible versus inaccessible states was not provided.

Genome-wide analyses have indicated that nucleosome positioning depends in part on DNA sequence features that promote the appropriate bending of dsDNA around the core histone octamer, and various algorithms have been designed for the purpose of predicting nucleosome organization in vivo (27–29). Such studies have also shown that promoter sequences are either nucleosome-free or occupied by unstable nucleosomes, and are flanked by highly organized nucleosome arrays in vivo (30–32). However, the only genome-wide study of nucleosome organization in the lymphoid compartment examined peripheral human CD4⁺ T cells, which contain accessible but heterogeneously rearranged TCR loci and inaccessible Ig loci (33). Thus, current data do not allow evaluation of individual unrearranged Ag receptor loci in both the accessible and inaccessible states.

In this study, we characterize the organization of nucleosomes at accessible and inaccessible TCR loci in primary mouse thymocytes. We find that nucleosome organization is dictated at least in part by the local DNA sequence, but that RSSs do not play a dominant role in nucleosome positioning. Promoter- and enhancer-mediated accessibility is characterized by loss and, to a lesser extent, repositioning of histone octamers. At least some of these changes are a direct consequence of germline transcription across RSSs. We discuss the implications of these observations for regulated RAG1/2-mediated recombination at RSSs in vivo.

Rag2^−/− mice (34), E_β^−/− Rag1^−/− mice (35) (kind gift from P. Ferrier, Centre d'Immunologie Marseille-Luminy, Marseille, France), PD_β1^−/− Rag2^−/− mice (36) (kind gift from J. Chen, Massachusetts Institute of Technology, Cambridge, MA), Rag2^−/− Tcrb transgene (tg) mice (37), E_α^−/− Rag2^−/− Tcrb tg mice (38), T early α (TEA)^−/− Rag2^−/− Tcrb tg mice (39), mice homozygous for the T early α-transcription terminator (TEA-T) insertion on a Rag2^−/− Tcrb tg background (24) and Tcrb^−/− Tcrd^−/− mice (40) were used following protocols approved by the Duke University and Vanderbilt University Animal Care and Use committees.

Primary thymocytes were obtained from 2- to 3-wk-old mice. After cell lysis, thymocyte nuclei were incubated with micrococcal nuclease and mononucleosomes were isolated using a 10–40% sucrose gradient as previously described (41). Mononucleosome DNA was purified and quantified using SYBR Green real-time PCR (Roche LightCycler2.0) using a genomic DNA standard curve. All PCR amplifications used a linear touchdown protocol with annealing temperatures decreasing from 68°C to 56°C in 10 cycles, followed by 40 cycles with annealing at 56°C. Amplification signals from a positioned nucleosome in the promoter of the IL-12 gene (42) were used to normalize the experimental values. A complete list of primer sequences is provided in Supplemental Table I online.

A portion of the D_β1–J_β1 cluster (nucleotides 152520–154707; GenBank accession number MMAE000665; www.ncbi.nlm.nih.gov/nuccore/2358100) was PCR amplified using PFU Turbo (Stratagene, La Jolla, CA) and cloned using a TOPO Cloning kit (Invitrogen, Carlsbad, CA). Nucleosomes were assembled on 1 μg plasmid DNA using a Chromatin Assembly kit (Millipore, Bedford, MA, 17-410) according to the manufacturer’s instructions, with the exception that assembled chromatin was digested with enzyme mixture for 14 min at 27°C. After the addition of the Enzymatin Stop Solution, nucleosomes were fractionated on a 10–40% sucrose gradient. Fractions containing pure mononucleosomes were identified by PCR amplification of purified DNA using primers designed to amplify mononucleosome-sized and dinucleosome-sized fragments. Mononucleosomal DNA was then analyzed as outlined previously, with the exception that normalization to IL-12 was omitted.

To study nucleosome organization at Ag receptor loci in vivo, we used micrococcal nuclease to prepare mononucleosomes from thymocyte chromatin and we subsequently analyzed purified mononucleosomal DNA by real-time PCR (41, 43). Primer sets were selected to define overlapping 100–140 bp regions whose efficient amplification required protection of the target DNA within a positioned nucleosome. In contrast, target sequences that span internucleosomal DNA would be amplified inefficiently because they are cleaved by micrococcal nuclease. In all experiments, nucleosome occupancy at TCR loci was scored by normalization to signals obtained with a primer pair contained within a defined positioned nucleosome in the IL-12 promoter (42).

D_β to J_β recombination at the Tcrb locus depends on the chromatin remodeling functions of the Tcrb enhancer (E_β) (44, 45) and the individual promoters (PD_β1 and PD_β2) associated with D_β gene segments (46–48). E_β and PD_β1 collaborate to provide accessibility to D_β1 and J_β1.1–1.6 (49), whereas E_β and PD_β2 are thought to provide accessibility to D_β2 and J_β2.1–2.7. To study nucleosome organization at accessible and inaccessible versions of the D_β1–J_β1 cluster, we analyzed wild-type, E_β-deficient, and PD_β1-deficient alleles in DN thymocytes from Rag^−/− mice (Fig. 1).

We first focused on an inaccessible D_β1–J_β1 cluster lacking the active enhancer element (E_β^−/−). This analysis revealed a nucleosome free region spanning PD_β1, followed by a highly organized array of positioned nucleosomes (A–K). The nucleosome-free promoter is consistent with studies showing that PD_β1 is preloaded with transcription factors such as Ikaros and CREB in DN thymocytes, independent of E_β activity (45). Nucleosome A spans the D_β1 gene segment and 5′RSS, whereas nucleosomes B and C are situated downstream of the D_β1 3′RSS. The D_β1 3′RSS resides in a relatively long internucleosomal linker (50 bp) between nucleosomes A and B; in contrast, the B/C internucleosomal space is smaller (34 bp). Notably, the amplification pattern across J_β1.1–J_β1.3 revealed an array of closely spaced peaks indicative of overlapping nucleosomes that cannot all be accommodated on the same DNA molecule. For instance, DNA molecules that have assembled nucleosome E cannot also have assembled nucleosomes D and F. Similarly, DNA molecules that have assembled nucleosome G cannot have assembled nucleosomes F and H, and DNA molecules that have assembled nucleosome I cannot have assembled nucleosome J. As a result of this heterogeneous nucleosome positioning, the J_β1.2 RSS is nucleosomal on some DNA molecules and internucleosomal on others. Downstream of J_β1.3, nucleosome K was positioned more uniformly.

Inaccessible alleles that carry an active enhancer but lack the promoter element (PD_β1^−/−) displayed a nearly identical nucleosome organization, except for the reduced abundance of nucleosome B, and smaller reductions in abundance of several other nucleosomes. Reduced occupancy of histone octamers on PD_β1-deleted as compared with E_β-deleted alleles is consistent with previous studies demonstrating E_β-dependent histone modifications and partial accessibility on PD_β1-deleted alleles (47, 49). Nucleosome A was not examined because it falls within the region disrupted by the PD_β1 deletion.

We also determined the nucleosome organization on wild-type alleles in which E_β and PD_β1 can physically interact and create a chromatin environment that is permissive for RAG1/2 binding and RSS cleavage (49). Accessible alleles displayed nucleosome positioning that was identical to inaccessible alleles from D_β1 to J_β1.1. However, there were differences downstream of this region, with a repositioned nucleosome (G′) replacing G and H on many alleles, and repositioning of nucleosome J further 5′ (to I) on most alleles. Importantly, although nucleosome positioning did not change in the 5′ region, the average occupancy of histone octamers was substantially diminished on accessible alleles compared with inaccessible alleles. This almost certainly reflects loss of histone octamers rather than mobilization to new positions, because there was no evidence for newly positioned nucleosomes on a measurable fraction of alleles in this region. We conclude that PD_β1 and E_β are both necessary for maximal removal and repositioning of histone octamers across D_β1–J_β1.3.

Chromatin accessibility at the D_β2 gene segment depends on E_β, and presumably PD_β2, but is independent of PD_β1 activity (36). Notably, PD_β2 consists of two distinct promoter elements situated 5′ and 3′ of the D_β2 gene segment (48) (Fig. 2). We found that on inaccessible E_β-deficient alleles the entire D_β2 region is organized into an array of positioned nucleosomes (A, B, C, and D) with high occupancy of histone octamers. Moreover, D_β2 and its 5′ and 3′ RSSs are entirely contained within nucleosome C. It is striking that, unlike PD_β1, PD_β2 is fully assembled into nucleosomes in the absence of E_β. This difference may have important implications for how these promoters are activated during development, because germline transcription and rearrangements appear to occur earlier and more efficiently at D_β1–J_β1 as compared with D_β2–J_β2 in developing thymocytes (50–52). Nucleosome packaging may delay or reduced the efficiency of PD_β2 activation, thereby directing the majority of initial recombination events to D_β1–J_β1.

Wild-type and PD_β1-deficent alleles are both accessible at D_β2 (47). Accordingly, these alleles were found to display nucleosome organizations that are identical to each other but distinct from the inaccessible E_β-deficient alleles in three major ways: 1) promoter nucleosomes A and B were absent, 2) nucleosome C was slightly repositioned (to C′), and 3) nucleosomes C and D were diminished in abundance. Loss of histone octamers at 5′PD_β2 suggests that factor loading to this promoter is E_β-dependent rather than -independent, as for PD_β1. We conclude that loss of histone octamers may be the primary mechanism associated with accessibility of D_β2 RSSs.

We identified no consistent pattern of nucleosome phasing relative to the D_β and J_β RSSs (Fig. 3). This raised the question of how nucleosome positioning is established. A potential determinant of nucleosome organization is regional features of the DNA sequence (27, 29, 53). To test this, we analyzed the D_β1–J_β1 DNA sequence using algorithms that predict nucleosome positioning from DNA sequence information (http://genie.weizmann.ac.il/software/nucleo_prediction.html). An algorithm designed based on global analysis of nucleosome positioning across the chicken genome was found to correctly predict the in vivo positions of multiple nucleosomes in the D_β1–J_β1 cluster (nucleosome A = 2, B = 3, C = 4, D = 5, and F = 6) (Fig. 4A). However, this algorithm did not predict one of the two alternate nucleosome phasing patterns at J_β1.1, and failed to predict positioned nucleosomes between J_β1.2 and J_β1.3. Moreover, the algorithm predicted a positioned nucleosome at PD_β1. This suggests that PD_β1 DNA sequence features, per se, do not exclude nucleosomes, and that the low histone octamer occupancy at PD_β1 may be caused by protein binding to PD_β1.

To further address the impact of DNA sequence on nucleosome positioning, we assembled nucleosomes on D_β1–J_β1 DNA in the absence of most cellular factors in vitro. We found that the majority of nucleosome positions detected on E_β-deficient alleles were also detected on in vitro assembled chromatin (Fig. 4B). However, in vitro assembled chromatin also displayed histone octamers at new positions that were not apparent in vivo. Thus, DNA sequence features are likely to be one of several factors contributing to nucleosome positioning in vivo. Notably, PD_β1 was heterogeneously assembled into nucleosomes on in vitro assembled chromatin, supporting the notion that PD_β1 is unoccupied in vivo because of the binding of cellular factors rather than because the PD_β1 sequences tend to exclude nucleosomes.

We note that the nucleosome organization of in vitro-assembled templates is more heterogeneous than is observed in vivo or than is predicted by the computer algorithm. This might reflect underassembly of nucleosome arrays in vitro, which could diminish the organizational constraints imposed by well-positioned neighboring nucleosomes. Such influences are explicitly accounted for in the computer model.

Our nucleosome mapping data indicate that promoters and enhancers cooperate to generate accessibility via loss and repositioning of histone octamers. These changes in nucleosome organization could reflect enhancer-dependent recruitment of chromatin remodeling complexes to the promoters (16, 45) as well as the effects of enhancer-dependent transcription originating at the promoters (23, 24). We specifically tested the role of transcription by comparing nucleosome organization on wild-type versus mutant Tcra alleles in which transcriptional elongation was disrupted in vivo.

At the Tcra locus, V_α-to-J_α recombination depends on the chromatin remodeling effects of the Tcra enhancer (E_α) and the TEA promoter that together provide accessibility to the 5′ J_α gene segments, including J_α61. To distinguish the effects of transcription from other activities of E_α and the TEA promoter, we previously introduced a transcription terminator downstream of the TEA promoter and showed that this manipulation suppressed both recombination and covalent histone modifications at downstream J_α gene segments (24). Therefore, to better understand the relative roles of enhancer activity, promoter activity, and transcription in remodeling nucleosome arrays, we determined the nucleosome organization at J_α61, ∼1.8 kb downstream of TEA, on wild type, E_α-deficient, TEA promoter-deficient, and TEA-T alleles.

The J_α61 region of inaccessible E_α-deficient alleles was characterized by three positioned nucleosomes (A, B, and C), one of which (B) encompassed J_α61 and its RSS (Fig. 5). Given the broad and closely spaced peaks, we inferred that nucleosomes A and B must adopt alternative positions on a fraction of alleles (A′, B′). Accessible wild-type alleles were characterized by repositioning of nucleosomes A and A′ to a more 5′ location (A″) and partial loss of nucleosomes B and C. Although all these changes resulted from E_α activity on wild-type alleles, only some depended on an intact TEA promoter. This was true for nucleosome C, whose position and occupancy were similar on E_α-deficient and TEA-deficient alleles. This was also true for nucleosome A, which reverted to a more 3′ position (favoring A′) on TEA-deficient alleles. To be compatible with A′, nucleosome B must adopt primarily the B′ position on TEA-deficient alleles. However, overall occupancy at nucleosome B (B and B′) was lower on TEA-deficient as compared with E_α-deficient alleles, indicating that occupancy at this position is influenced by E_α in a manner that is independent of TEA.

Because the repositioning of nucleosome A to the A″ position and the abundance of nucleosome C depend on both TEA and E_α, we asked whether these changes are mediated by transcriptional read-through from TEA. Indeed, the abundance of nucleosome C was nearly identical on TEA-T as compared with E_α-deficient and TEA-deficient alleles. Moreover, the positioning of nucleosome A was identical to that on the E_α-deficient alleles, and very similar to that on TEA-deficient alleles. Importantly, the properties of both nucleosomes A and C differed substantially between TEA-T and wild-type alleles, indicating that germline transcription through these nucleosomes causes both repositioning and loss of histone octamers. Other nucleosomes (e.g., nucleosome B) may be lost in a transcription-independent fashion.

Although accessibility of chromosomal RSSs is understood to be critical for the developmental regulation of V(D)J recombination, a detailed picture of nucleosome organization at accessible and inaccessible RSSs has been lacking. In this study, we have directly compared the organization of nucleosomes at accessible versus inaccessible Tcrb and Tcra alleles in primary thymocytes. We identified arrays of highly positioned nucleosomes and obtained evidence that this positioning is established at least in part by the regional DNA sequence. However, we found no consistent positioning of nucleosomes with respect to RSSs in vivo, a result that stands in sharp contrast to studies indicating that RSS nonamers function as dominant nucleosome positioning sequences in vitro (25). Notably, enhancer- and promoter-dependent accessibility was characterized by diminished abundance of certain nucleosomes, whereas others were repositioned. Moreover, in the case of Tcra, some of these changes were found to be a direct consequence of germline transcription through RSSs. Previous studies have emphasized the role of cis-acting regulatory elements and germline transcription in the covalent modification of nucleosomes at Ag receptor loci. Our data extend the description of chromatin disruption at RSSs, suggesting that RSS accessibility is characterized by substantial changes in nucleosome occupancy and organization.

Nucleosome positioning is influenced in part by intrinsic properties of the genomic DNA sequence (27, 29, 53). Efficient nucleosome assembly is promoted by sequence features that allow sharp bending of the DNA (e.g., AA/TT/TA dinucleotides spaced with 10 bp periodicity) and is inhibited by sequences that form rigid structures [e.g., extended poly(dA-dT) sequences]. However, the overall contribution of intrinsic DNA sequence features may be rather modest (53). In particular, the positioning of promoter-flanking nucleosomes seems to be established primarily by RNA polymerase II binding at the transcription start site, with neighboring nucleosomes then positioned as a downstream consequence (27, 29, 53). Consideration of these factors may explain why our in vivo data on nucleosome positioning relative to RSSs diverges so substantially from the in vitro data of Baumann et al. (25). One critical difference is that the in vitro nucleosome assembly was conducted with short DNA fragments that could only assemble a single nucleosome. Although the RSS nonamer may have some propensity to function as a nucleosome positioning sequence when tested in this context, its influence may not be strong enough to overcome additional regional sequence cues that are present in vivo or on larger substrates in vitro. Moreover, the in vitro data cannot account for the regional effects of RNA Pol II and other DNA binding factors on nucleosome organization in vivo. In any event, the notion that RSS nonamers could function as dominant nucleosome positioning sequences is at odds with the observation that the 5-bp sequence AAAAA, which characterizes many RSS nonamers, is incorporated into nucleosomes at the lowest frequency of all 5-bp sequences (29).

We found inaccessible Tcrb and Tcra alleles to be characterized by high occupancy of histone octamers. However, on at least a fraction of alleles, certain RSSs (3′D_β1, J_β1.2) were found to be internucleosomal even in their inaccessible configurations. This suggests that nucleosome occupancy at the RSS is not required to generate a RAG-inaccessible configuration. Rather, the large RAG1/2 complex may be sterically constrained from accessing internucleosomal RSSs because of the close packing of adjacent nucleosomes. Recombination may also be inhibited because adjacent nucleosomes in the enhancer- and promoter-deficient alleles lack critical histone modifications (e.g., H3K4me3) that are required for effective RAG1/2 binding and function.

We found accessible Tcrb and Tcra alleles to be characterized by low occupancy of histone octamers. However, it is formally possible that the regions studied are not nucleosome-depleted in vivo, but rather carry unstable nucleosomes that are lost during experimental manipulations in vitro. In this regard, the histone variant H3.3 is deposited across active transcription units (54, 55) and histone variant H2A.Z is deposited at positions flanking transcription start sites (56). Nucleosomes assembled with these histone variants are known to have diminished stability (57). Moreover, recent data indicate that a single H3.3- and H2A.Z-containing nucleosome deposited at the −1 position relative to the transcription start site is preferentially lost from standard mononucleosome preparations conducted under high salt conditions (150 mM) (58). Thus, our isolation conditions (100 mM NaCl) could have contributed to a loss of nucleosome A that protects the PD_β1 transcription initiation site. However, analysis of formaldehyde–cross-linked chromatin for deposition of H2A.Z revealed only very low levels of this histone variant from D_β1 to J_β1.3 and at J_α61 (H.D. Kondilis-Mangum and M.S. Krangel, unpublished observations). Thus H2A.Z deposition is unlikely to contribute significantly to nucleosome loss in these regions.

Our data for the Tcra locus demonstrate that nucleosome loss and repositioning can be a direct consequence of transcriptional elongation. Although we could not directly test the role of transcription at the Tcrb locus, we think it likely that many of the changes that we detect in the D_β1–J_β1 region are transcription-dependent as well. In this regard, transcription is known to be associated with transient eviction and redeposition of histones, because of the activities of chromatin remodeling complexes and histone chaperones that travel with RNA Pol II (59–61). Moreover, highly transcribed genes can display substantially reduced steady state nucleosome densities across their coding regions (62, 63). We conclude that transcription-dependent reorganization and eviction of histone octamers is likely to play a critical role in accessibility to RAG1/2 in vivo.

Transcription not only disrupts the organization of nucleosome arrays, but also promotes the covalent modification of histone tails (21). In particular, transcription across J_α61 is known to promote histone H3Ac, as well as the accumulation of H3K4me2, H3K4me3, and H3K36me3 marks (24). Histone H3Ac, H3K4me2, and H3K4me3 marks also characterize accessible D_β and J_β gene segments (15, 18, 49). Notably, the H3K4me3 modification, which we detect at high levels on every nucleosome in the D_β1–J_β1 and J_α61 regions (H.D. Kondilis-Mangum and M.S. Krangel, unpublished observations) has been linked to efficient RAG1/2 recruitment and enzymatic activity (18–20). This raises an obvious conundrum regarding the true chromatin state of accessible RSSs. Are nucleosomes absent or are they present and covalently modified? We suggest that in the wake of transcription across RSSs, individual alleles are mosaics characterized by intact nucleosomes with modified histone tails at some sites, and disassembled nucleosomes at others. Recombination would then be supported by a configuration in which one or more H3K4me3-modified nucleosomes flank a nucleosome-free RSS. Recent studies have provided an initial glimpse into the structure of a RAG1/2-mediated synaptic complex with a naked DNA substrate (64). A precise understanding of the analogous complex with a chromosomal DNA substrate will be an even greater challenge for future studies.

We thank Zanchun Huang and Steven Pierce for excellent technical assistance. We also thank Bingtao Hao and Juan Carabana for critical review of the manuscript.

Disclosures The authors have no financial conflicts of interest.