Assessing the Role of the T Cell Receptor β Gene Enhancer in Regulating Coding Joint Formation during V(D)J Recombination*

To assess the role of the T cell receptor (TCR) β gene enhancer (Eβ) in regulating the processing of VDJ recombinase-generated coding ends, we assayed TCRβ rearrangement of Eβ-deleted (ΔEβ) thymocytes in which cell death is inhibited via expression of a Bcl-2 transgene. Compared with ΔEβ, ΔEβ Bcl-2 thymocytes show a small accumulation of TCRβ standard recombination products, including coding ends, that involves the proximal Dβ-Jβ and Vβ14 loci but not the distal 5′ Vβ genes. These effects are detectable in double negative pro-T cells, predominate in double positive pre-T cells, and correlate with regional changes in chromosomal structure during double negative-to-double positive differentiation. We propose that Eβ, by driving long range nucleoprotein interactions and the control of locus expression and chromatin structure, indirectly contributes to the stabilization of coding ends within the recombination processing complexes. The results also illustrate Eβ-dependent and -independent changes in chromosomal structure, suggesting distinct modes of regulation of TCRβ allelic exclusion depending on the position within the locus.

assembly and for T and B lymphocyte differentiation. This process is mediated by an enzymatic complex (the VDJ recombinase) whose targets (the recombination signal sequences or RSSs) flank dispersed V, D, and J gene segments and consist of conserved seven-and nine-nucleotide sequences (the heptamer and nonamer) separated by a non-conserved 12-or 23-nucleotide spacer (1). The recombination-activating-gene (RAG)-1 and -2 products constitute the core components of the recombinase (2). The RAG genes (possibly together with the original RSSs) are thought to have been transferred, in the form of a composite transposon, from the prokaryotic world to the germline of a common ancestor of the jawed vertebrates (3).
V(D)J recombination has been divided into two phases, based on in vitro recombination studies and the biochemical characterization of rearrangement products (4). In the first phase, the RAG factors (assisted by architectural factors; i.e. the high mobility group-1 and -2 proteins) initiate recombination by binding to, and introducing DNA cleavage at, two RSSs with spacers of dissimilar lengths. Ensuing DNA double strand breaks (DSBs) yield two pairs of products that consist of 5Јphosphorylated, blunt-ended RSSs (called signal ends, SEs), and the hairpinned, adjacent coding sequences (called coding ends, CEs). In a second phase, which depends on the coordinated action of the RAG and DNA repair non-homologous endjoining (NHEJ) factors (including Ku70/86, DNA-PK/Artemis, XRCC4, and DNA ligase IV), the two CEs are rapidly processed (this involves opening of the hairpins and, often, deletion and/or addition of nucleotides) and ligated to form a coding joint (CJ). With slower kinetics, the SEs are precisely joined to form a signal joint (SJ). As in the case of various recombination systems in prokaryotes (5), synaptic complexes of V(D)J recombination have been characterized that contain all or part of the aforementioned nucleotide sequences and catalytic factors (reviewed in Ref. 6).
V(D)J recombination is confined to immature lymphocytes because of the restricted expression of the RAG genes. In addition, it is tightly controlled with respect to lymphoid cell lineage and within a given lineage to the developmental stage and possibly also the TCR/Ig allele used. For example, the TCR␤ and TCR␣ genes (assembled from V␤, D␤, and J␤ and from V␣ and J␣ gene segments, respectively) are, with a few exceptions, rearranged exclusively in the T cell lineage, with TCR␤ gene rearrangement in double negative (DN) pro-T cells preceding that of TCR␣ in double positive (DP) pre-T cells. Moreover, at the TCR␤ locus, D␤-to-J␤ rearrangement occurs first in CD44 ϩ CD25 ϩ DN thymocytes and, presumably, simultaneously on both alleles, followed by complete V␤-to-DJ␤ assembly in more mature CD44 Ϫ/lo CD25 ϩ cells, possibly with no allele synchronicity. Formation of a productive V␤-to-DJ␤ joint (i.e. that maintains an open reading frame within the TCR␤ gene) and expression of a TCR␤ chain (conditional for ␣␤ lineage normal development past the CD44 Ϫ/lo CD25 ϩ DN stage) result in the arrest of further V␤ rearrangement to mediate TCR␤ allelic exclusion (7). To a large extent, these controls are thought to involve the regulated modulation of RSS accessibility to the recombinase (8). The findings, using transgenic and knock-out mice, that transcriptional regulatory elements (enhancers/promoters) modulate cis-rearrangement and chromatin structure at TCR/Ig gene segments and/or loci gave a first hint toward an understanding of how recombinational accessibility is achieved (4,9).
In the mouse germline, the ϳ500-kb TCR␤ locus consists of ϳ35 distinct V␤ genes that, for the most part, are spread over a large DNA region extending from 200 -450 kb upstream of the duplicated D␤1-J␤1-C␤1/D␤2-J␤2-C␤2 clusters, except for one (V␤14), which lies, in opposite orientation, ϳ10 kb downstream (10). A single TCR␤ gene enhancer (E␤) has been described that is located within the C␤2-V␤14 intervening region (11). Targeted deletion of E␤ has revealed a striking phenotype. In the T cell lineage, D␤-to-J␤ CJs are drastically reduced at the targeted TCR␤ allele(s) (Ͼ50 -100-fold compared with the wild-type (wt)), with an even more severe defect in V␤-to-DJ␤ CJs. In homozygously deleted (E␤ Ϫ/Ϫ , hereafter ⌬E␤) mice, no TCR␤ chains are made, and no ␣␤ T cells can develop (12)(13)(14). Moreover, comparative analysis of molecular markers for chromatin structure in developmentally arrested DN, CD25 ϩ pro-T cells from either Rag Ϫ/Ϫ (hereafter Rag) or combinatorial (Rag Ϫ/Ϫ ϫ ⌬E␤; Rag ⌬E␤) mice provided compelling evidence for a primary function of E␤ in regulating chromatin opening within a limited (ϳ25 kb) upstream domain comprised of the D␤-J␤-C␤ clusters, with a minor effect on the 5Ј distal V␤ genes or 3Ј proximal V␤14 (15). However, RAG-mediated SEs at D␤ and J␤ gene segments (as well as the corresponding SJs) can readily be detected at E␤-deleted alleles, although at a level 10 -30-fold lower compared with the wt (16). The facts that TCR␤ rearrangement was initiated in ⌬E␤ thymocytes, but that formation of CJs may be more severely impaired, suggested an additional function for E␤ in CE processing. Here, using ⌬E␤ mice expressing an anti-apoptotic Bcl-2 transgene, we attempt to better delineate the actual impact of E␤ in enhancing DNA repair/CJ formation during TCR␤ locus recombination, relative to its effects on chromosomal accessibility. Our findings are consistent with a model in which E␤ impinges on the stabilization of CEs within the post-cleavage synaptic complex, in addition to its primary functions in regulating chromosomal access and locus expression.
Thymocyte Preparation and Cell Culture Conditions-Thymocyte preparation and cell culture have been described previously (15). 150 g of anti-CD3-⑀ (2C11; Pharmingen) monoclonal antibody were utilized for intraperitoneal injection of 4-week-old animals.
Flow Cytofluorometry Analyses and Cell Sorting-Cell-staining conditions, flow cytometric analyses, and cell purification by cell sorting were carried out as described by Leduc et al. (14).

Molecular Analyses of V(D)J Recombination Products and Chromatin
Structure at the TCR␤ Locus-Nucleic acid extraction, assays for SE, SJ, CJ, and hybrid joints (HJ) products, sequencing, and RT-PCR analyses, as well as ligation-mediated (LM)-PCR analysis of restriction enzyme accessibility and chromatin immunoprecipitation (ChIP)-PCR analysis of histone H3 acetylation were performed as described previously (13,15,16). Assays for CE products were performed according to Zhu et al. (19), with PCR products for CEs/SEs being separated through polyacrylamide gels (instead of agarose gels, as for the analysis of the amplified products in all other PCR assays). All PCR experiments were performed at least twice with consistent results. A list of oligonucleotide primers used in these experiments is available upon request.

RESULTS
Production of the ⌬E␤ B2 Mice and Characterization of Their TCR␤ Gene Recombination Profile-A defect in resolving RAGmediated DSBs that form at E␤-deleted (E␤ Ϫ ) alleles must result in cell death of the particular thymocytes. This could mask the actual levels of TCR␤ recombination products (e.g. SEs) in ⌬E␤ thymi and the role played by E␤ in promoting RSS accessibility versus post-cleavage assembly in vivo. To overcome this problem, we analyzed TCR␤ gene recombination at E␤ Ϫ alleles in the situation where cell death is inhibited; ⌬E␤ mice were bred with mice that express an anti-apoptotic human Bcl-2 transgene (Tg Bcl-2) in T lineage cells (B2 mice) (17). Cell counting and flow cytometric analysis indicated that constitutive Bcl-2 expression results in a slightly reduced proportion of DN cells in ⌬E␤ B2 versus ⌬E␤ thymi (from ϳ33 to 25.6%) and an increase in that of DP cells (from ϳ58 to 69%) although TCR␤ ϩ and genuine single positive cells are still missing (Table I). Moreover, Bcl-2 expression was found to prolong cell survival of E␤-deleted DN and DP thymocytes without rescuing the CD44 Ϫ/lo CD25 ϩ DN developmental block and accompanying cell proliferation defect (data not shown). These findings are in agreement with those from earlier studies demonstrating that Tg Bcl-2 expression inhibits cell death without substituting for major selection processes in the developing T cells such as pre-TCR-based ␤-selection or TCR␣␤based positive selection (17, 20 -22).
Overall, the effect of Bcl-2 in rescuing D␤-to-J␤ CJs may thus result from a small accumulation of intermediate SE/CE products combined to a specific, additional enhancement of CE processing (formally, because the CJ effect should be the product of SE/CE formation (or abundance) and CE resolution, the apparent effect on CE resolution in this case could be ϳ2.4ϫ (4.3/1.8) only). Consistent with this, the germline fragment containing D␤2/J␤2 gene segments was detected at high levels in both ⌬E␤ and ⌬E␤ B2 but not wt thymocytes (Fig. 1D). Also, as judged from similar SE and CJ assays, recombinase activity in ⌬E␤ B2 cells generally appears less marked within the D␤1-J␤1 cluster compared with D␤2-J␤2 (data not shown). The small accumulation of D␤2-to-J␤2 recombination products in ⌬E␤ B2 thymocytes prompted us to also check for the presence of V␤-to-DJ␤ CJs. Surprisingly, V␤14-to-DJ␤2 CJs were found in thymocytes from the ⌬␤ B2 mice (at levels varying from ϳ12 to 48% of those in the wt) whereas these products were routinely not detected in cells from ⌬E␤ littermates (e.g. Fig. 2). Analysis of SEs at V␤14 once again argued for a specific effect of Bcl-2 expression on the rescue of CEs compared with SEs (see below). However, V␤-to-DJ␤2 assembly in ⌬E␤ B2 thymocytes seems to be restricted to V␤14 only (i.e. the single V␤ gene located on the 3Ј end of the TCR␤ locus), as no accumulation of CJs was detected for several 5Ј V␤s, including members located in the proximal (V␤18), median (V␤20, V␤11, and V␤5), or distal (V␤4) parts of the 5Ј V␤ gene cluster ( Fig. 2) (data not shown).
Sequence analysis of D␤2-to-J␤2.5/2.6 CJs from ⌬E␤ B2 thymocytes generally showed the hallmarks of normal CE processing prior to joining, including occasional P and/or N nucleotide additions and short deletions; one D␤2-to-J␤2.5 CJ showed an unusually long (11-bp) N region (data not shown) (also, in Fig. 1C, SE resolution in ⌬E␤ B2 T cells appears to mostly result in standard SJs as they were digested by restriction enzyme ApaLI that cleaves the perfect fusion of two hep- The asterisk indicates a nonspecific band that was also amplified from kidney DNA. Serial dilution was as in B except that genomic (non-ligated) DNAs were used; dilution in lanes 2 and 3 was 1/4 and 1/8, respectively. E, quantitation of SE, SJ, and CJ products. PhosphorImager signals for the recombination products were quantified by densitometric scanning and corrected according to the signal from the C␤ control. Graphic representation for each product is shown, relative to a 100% value as defined from the corresponding signals in wt thymocytes. tamers). Parallel analysis of V␤14-to-DJ␤2 rearrangement similarly revealed canonical CJ/SJ features and, as expected, joining by DNA inversion.
Analysis of ⌬E␤ B2 Thymocytes for CE and HJ Products-The above data indicate that constitutive Bcl-2 expression effects a small accumulation of TCR␤ SE, SJ, and CJ products in E␤-deleted thymocytes that is confined to gene segments located proximal to the E␤ deletion, including the D␤-J␤ upstream segments and downstream V␤14 gene. We next investigated whether this effect also impacts on the accumulation of other forms of V(D)J recombination products within this domain, namely CEs and HJ products.
Although SEs 3Ј of D␤ gene segments can be found in thymocytes from the ⌬E␤ mice (e.g. Fig. 1), the corresponding CEs could not be detected by LM-PCR of genomic DNA treated with mung bean nuclease (to open the hairpin structures) and T4 DNA polymerase (to blunt occasional DNA overhang extremities) 2 (for details on this strategy, see Ref. 19). However, CEs are difficult to detect, even in wt thymocytes, because of their rapid processing and resolution into CJs. Conversely, CEs readily accumulate in developmentally arrested lymphocytes from NHEJ-deficient mice such as the Scid (DNA-PK-deficient) mice (19). Using the aforementioned strategy, CEs 3Ј of D␤2 were indeed observed in thymocytes from a Scid mouse (together with SEs 5Ј of D␤2), but not in those from a ⌬E␤ B2 mouse (Fig. 3, upper panels; in ⌬E␤ B2 DNA, the faint signal at a size close to that of CEs was not reproducibly observed), arguing that there is no accumulation of CEs at E␤ Ϫ alleles, even under the condition of constitutive Bcl-2 expression.
HJs are non-standard V(D)J rearranged products that result from the attack of a hairpinned CE by the SE liberated from the opposite gene segment participating to the recombination complex (schematized in Fig. 3). The reaction, which is mechanis-tically similar to a transposition, was partially reproduced in vitro using purified, truncated forms of RAG proteins (catalytically active core RAGs), in the absence of the NHEJ factors (23). In vivo, HJs are found at a low level in wt lymphocytes. In line with the in vitro results, it is widely accepted that HJs predominate in developing lymphocytes with NHEJ deficiencies. However, a recent study (24) suggests a more complex situation, as the non-core regions in full-length RAGs seem to down-modulate HJ levels in the absence of NHEJ. Intriguingly, HJs involving CEs 3Ј of D␤2 and J␤2 SEs could not be detected in ⌬E␤ or ⌬E␤ B2 thymocytes (Fig. 3, bottom panels). Thus, Tg Bcl-2 expression in E␤ Ϫ thymocytes results in an accumulation of SE/SJ/CJ products of D␤-to-J␤ rearrangement, but it has a negligible impact on parallel accumulation of CEs and HJs. Overall, these data do not support a role for E␤ in the recruitment of NHEJ factors to the recombination complex (see "Discussion"). However, they do not exclude an indirect effect of E␤ on CJ formation; e.g. the stabilization of CE intermediates within the PCS complex.
TCR␤ Gene Expression in E␤-Deleted Thymocytes-At TCR/Ig loci, activation of regional transcription and V(D)J recombination frequently (but not always) correlate (4). Earlier studies have shown that transcription of the unrearranged (germline) D␤-J␤ loci is strongly inhibited at E␤ Ϫ compared with E␤ ϩ alleles in early developing T cells, whereas that of V␤ genes is not significantly altered (15). We have analyzed TCR␤ gene expression in ⌬E␤ B2 versus ⌬E␤ thymocytes, using RT-PCR assays to study transcription through either germline J␤ or V␤ gene segments (J␤ Gl or V␤ Gl) or through partially (DJ␤) or completely (V␤DJ␤) rearranged products (DJ␤ Rg or V␤ Rg). We found no J␤ Gl The diagram schematizes the two assays and the region of the TCR␤ locus that was analyzed (outlined as in Fig. 1A), emphasizing the D␤2/J␤2 cleavage/recombination sites that were investigated. The vertical arrow indicates the position of 5Ј D␤2 SEs that, in the LM-PCR assay, can be amplified simultaneously with 3Ј D␤2 CEs. Horizontal arrows indicate the relative location of oligonucleotide primers used for PCR amplifications. The size of the amplified fragments is indicated. transcription in ⌬E␤ B2 thymocytes at the D␤1-J␤1 or D␤2-J␤2 loci; also, DJ␤ Rg transcription was negative in these cells (Fig. 4A, upper two panels) (data not shown). Therefore, despite the evidence of D␤-to-J␤ recombination at E␤ Ϫ alleles, there is no evidence of transcription through these loci (including in the Tg Bcl-2 expressing cells). This is yet another example of differential activation of the two processes. In addition, we found V␤ Gl transcription for V␤5, V␤11, and V␤14 at E␤ Ϫ alleles, but no V␤ Rg transcription, including for V␤14 and the ⌬E␤ B2 thymocytes (Fig. 4A) (data not shown), in agreement with the lack of TCR␤ ϩ cells (and ␤-selection) in the ⌬E␤/⌬E␤ B2 mice, as evidenced by flow cytometry. Whereas V␤ Gl transcription of V␤5 (and V␤11) appeared to be reduced in ⌬E␤ B2 compared with ⌬E␤ thymocytes, that of V␤14 was unchanged (or increased slightly; e.g. see Fig. 4A). These differential profiles likely result from a reduced DN/DP cell ratio in ⌬E␤ B2 versus ⌬E␤ thymi, coupled with E␤-independent developmental changes in V␤ Gl transcription (rather than from intrinsic differences between the two types of ⌬E␤ B2 and ⌬E␤ cells). Indeed, we found a dramatic down-regulation of V␤5 and V␤11 Gl transcripts in purified DP versus DN ⌬E␤ B2 thymocytes and steady-state (or slightly increased) levels of V␤14 Gl transcripts (Fig. 4B) (data not shown). Therefore, V␤ Gl transcription profiles in DP ⌬E␤ B2 cells (5Ј V␤ Gl Ϫ /V␤14 Gl ϩ ) correlate with those of V␤-to-DJ␤ CJs in ⌬E␤ B2 thymi. This lead us to investigate whether, in this situation, TCR␤ recombination also depends on DN-to-DP development and, potentially, E␤-independent changes in chromosomal access.
Developmentally Regulated Activity of the VDJ Recombinase May Be Compromised at E␤-Deleted Alleles-To investigate whether the predominance of TCR␤ SE, SJ, and CJ products in ⌬E␤ B2 versus ⌬E␤ thymi also correlates with their differences in cell distribution, we purified DN and DP thymocytes from both types of mice and analyzed the rearrangement of their TCR␤ locus, as described above. We first focused on D␤2-to-J␤2 rearrangement. Remarkably, we found SEs 3Ј of D␤2 and D␤2to-J␤2-6 SJs predominantly in DP thymocytes from both ⌬E␤ and ⌬E␤ B2 mice and at higher levels in Tg Bcl-2 expressing cells; however, the effect of Bcl-2 on the accumulation of rearrangement products was also visible in DN pro-T cells (Fig. 5,  A (lanes 3-6) and B (lanes 5-12)). Likewise, in E␤-deleted animals, we detected D␤2-to-J␤2 CJs predominantly in ⌬E␤ B2 DP thymocytes; Bcl-2 also had an effect on CJ accumulation in DN cells (Fig. 5C). Notably, CJs in ⌬E␤ B2 DN cells were detected at a slightly higher level (ϳ1.2ϫ) compared with those in ⌬E␤ DP cells despite a bias for SEs (ϳ2ϫ) in favor of the latter (Fig. 5, C and A, lanes 4 and 5, respectively) (data not shown). Further quantitation analysis (Fig. 5D) indicated that SEs without E␤ are reduced to ϳ7% (DN) and ϳ26% (DP) of those in the corresponding wt cells, whereas CJs are reduced to ϳ4% (DN) and ϳ10% (DP). By comparison, in ⌬E␤ B2 thymocytes, SEs increased slightly to, respectively, ϳ12% (DN; ϳ1.7ϫ) and ϳ42% (DP; ϳ1.6ϫ) whereas CJs increased to 24% (DN; ϳ6ϫ) and ϳ36% (DP; ϳ3.6ϫ). These data confirm our previous results of a preferential, although limited effect of Bcl-2 on D␤/J␤ CE resolution at E␤ Ϫ alleles, which is apparent in both DN (ϳ3.5ϫ (6/1.7)) and DP (ϳ2.2ϫ (3.6/1.6)) cells. They further suggest that, in most E␤-deleted thymocytes, recombinase activity at the D␤/J␤ RSSs is extended/delayed to cells that have developed to the DP stage. Significantly, we also found high levels of SEs 3Ј of D␤2 and D␤2-to-J␤2-6 SJs in both DN and DP wt thymocytes (Fig. 5, A (lanes 1 and 2) and B (lanes 1-4)), in agreement with similar findings at the D␤1-J␤1 gene cluster (25). Because unresolved DSBs generated in DN cells would arrest cell proliferation during DN-to-DP cell differentiation, the detection of SEs 3Ј of D␤ in DP cells from wt mice must reflect cell autonomous activity of the RAG factors at unrearranged D␤/J␤ loci (25) and/or at D␤-to-J␤ SJs within extrachromosomal circles (26).

E␤-Independent Changes in Chromatin Structure at the TCR␤ Locus during T Cell Development-
The extended lifespan conferred by a Bcl-2 transgene is likely to provide an extended time window per cell for V(D)J recombination (27), thus accounting for one aspect of our results (i.e. the rescue of CJ formation; see below). However, other mechanisms must account for the accumulation of SEs on a large scale in DP cells (including DP cells that develop in the absence of ␤-selection) (14) and for the profile of V␤ gene recombination in ⌬E␤ B2 thymocytes. One possibility would be that, upon DP development, accessibility to the RAG factors is established along the D␤/J␤ loci in an E␤-independent manner, whereas a repressive structure invades the 5Ј V␤ genes but not V␤14. This would impact on the controls of both TCR␤ gene recombination and allelic exclusion during T cell differentiation.
We have tested this model and analyzed chromatin structure at discrete TCR␤ regions using nuclei from Rag and Rag ⌬E␤ thymocytes (E␤ ϩ and E␤ Ϫ DN cells, respectively) and enzyme restriction/LM-PCR chromosomal accessibility assays, as described previously (15). Thymocytes from Rag and Rag ⌬E␤ mice treated by intraperitoneal injection of anti-CD3-⑀ monoclonal antibody (to mimic pre-TCR signaling) (28) were used as a source of DP-enriched nuclei. Fibroblasts were used as a non-lymphoid control. As predicted, we found that the J␤1 region is more likely to be cleaved in E␤ Ϫ DP (Rag ⌬E␤ CD3) and in E␤ Ϫ DN (Rag) or DP (Rag CD3) thymic nuclei, compared with E␤ Ϫ DN (Rag ⌬E␤) thymic or to fibroblastic nuclei (Fig.   6A, top panels; consistent results were also obtained at the J␤2 locus). Furthermore, we found V␤5 to be more resistant to cleavage in both E␤ ϩ and E␤ Ϫ DP (Rag CD3 and Rag ⌬E␤ CD3) nuclei relative to E␤ ϩ or E␤ Ϫ DN (Rag and Rag ⌬E␤) nuclei (Fig. 6A, middle upper panels; similar results were found at V␤11) (data not shown). Finally, we found V␤14 to be cleaved in T cell nuclei, independent of the developmental (DN or DP) stage and the presence of E␤ (Fig. 6A, middle lower panels).
Histone acetylation has emerged as an important regulator of chromatin structure (29) and of locus accessibility for V(D)J recombination in vivo (30). We used ChIP-PCR to compare histone H3 acetylation at D␤-J␤ and V␤ loci in Rag ⌬E␤ CD3 (DP) versus Rag ⌬E␤ (DN) thymocytes. We found that, during the course of anti-CD3-⑀-induced DN-to-DP differentiation, H3 acetylation of E␤ Ϫ thymocytes (i) increases slightly at J␤1 and D␤1 and, more readily, at D␤2; (ii) decreases (by ϳ2-fold) at V␤5; and (iii) is maintained at a steady state level at V␤14 (Fig.  6B). Overall, these data support our model of E␤-independent, DN-to-DP regulated changes in chromosomal organization (including histone H3 acetylation) at distinct regions throughout the TCR␤ locus.
E␤-Independent DSB Cleavage at V␤14 in DP Thymocytes-The drop in accessibility of 5Ј V␤ genes in CD3-⑀ triggered Rag thymocytes likely mimics a physiological mechanism involved in the feedback inhibition of V␤ gene rearrangement in response to pre-TCR-induced signaling (i.e. allelic exclusion). In this context, persistent accessibility of the D␤-J␤ and V␤14 locus regions potentially threatens allelic exclusion so that this process must be regulated differently at the 3Ј end of the TCR␤ locus. The finding of V␤14-to-DJ␤ CJs in ⌬E␤ B2 thymocytes also raises the question as to whether E␤, in conjunction with cell death control, could participate in this regulation. To address these issues, we tested E␤ ϩ and E␤ Ϫ thymocytes for the presence of SEs at both V␤5 and V␤14 and of V␤5-/V␤14-to-DJ␤2 CJs. As a source of E␤ ϩ or E␤ Ϫ thymocytes, we used wt FIG. 6. E␤-independent changes in TCR␤ chromatin structure in DP versus DN thymocytes. A, thymocyte nuclei from Rag Ϫ/Ϫ (Rag) and Rag Ϫ/Ϫ E␤ Ϫ/Ϫ (Rag ⌬E␤) mice and from littermates that have been injected with anti-CD3-⑀ monoclonal antibody (Rag CD3 and Rag ⌬E␤ CD3) or nuclei from 3T3 fibroblasts (fibro.) were treated with increasing amounts of the restriction enzyme RsaI and analyzed by LM-PCR for enzyme cleavage in various parts of the TCR␤ locus, as described previously (15). The bottom panels show C␤2 PCR amplifications (C␤) using RsaI-restricted, linker-ligated DNA samples. The J␤1 cleavage profile, with three visible bands (instead of the two in Ref. 15), is because of a longer gel migration and better separation of the RsaI restricted fragments. The results shown are representative of three independent experiments. B, histone H3 acetylation at J␤1-, D␤1-, D␤2-, V␤5-, and V␤14-associated sequences was investigated in Rag ⌬E␤ and Rag ⌬E␤ CD3 thymocytes by ChIP, in parallel with that at the hyperacetylated TCR␦ gene enhancer (E␦) and the hypoacetylated Oct-2 (Oct) gene. ChIP was performed using either an anti-AcH3 anti-serum (␣AcH3) or no antibody (control). Two concentrations (dilution 1/0 and 1/3) are shown for amplification of the ␣AcH3-precipitated fractions and input materials. The graphics represent the ␣AcH3-precipitated/input material ratios, determined after PhosphorImager scanning of the hybridizing images. The 100% value was attributed to the acetylated E␦ control. mice and TCR␤ transgenic mice (p14, a model for TCR␤ allelic exclusion (18)) or the ⌬E␤ and ⌬E␤ B2 mice, respectively.
As expected, we detected SEs at V␤5 predominantly in DN thymocytes from the E␤ ϩ animals and at a reduced level in p14 compared with wt cells (a Ͼ6-fold decrease as judged from densitometric analysis of PhosphorImager signals) whereas, in agreement with previous findings, these products were hardly visible in E␤ Ϫ (⌬E␤ and ⌬E␤ B2) thymocytes (Fig. 7A, top  panel). In contrast, we found V␤14 SEs to predominate in DP cells from the p14, ⌬E␤, and ⌬E␤ B2 mice (Fig. 7A, middle panel; V␤14 SEs were occasionally detected, at a lower level, in DN and/or DP cells from wt mice) (data not shown). Yet both V␤5 and V␤14 CJs were normally found in wt thymocytes and were strongly reduced in p14 cells (although, possibly, to a lesser extend for the V␤14 CJs); as expected, CJs were not detected at E␤ Ϫ alleles except for V␤14 and the ⌬E␤ B2 cells (Fig. 7B). The latter findings, coupled to those of V␤14 SEs in ⌬E␤ DP thymocytes (Fig. 7A, lane 7), are consistent with a specific effect of Bcl-2 on CE processing also at V␤14. Two other elements should also be considered. First, a 6-fold decrease of V␤5 SEs between wt and p14 DN cells (LM-PCR assays of Fig.  7A, lanes 2 and 4) can account for the drop of the corresponding CJs (PCR assays of Fig. 7B, lanes 4 and 6; also see lanes 2 and  3), implying that exclusion of V␤5 rearrangement is likely to be regulated primarily at the level of chromosomal access. In p14 thymocytes, V␤5 SEs may correspond to normally rearranging alleles; e.g. in a few DN cells that do not express the ␤ transgene. We cannot exclude, however, that inhibition of V␤5 rearrangement is regulated beyond the step of DSB cleavage in a small population of TCR␤ rearranging cells. Second, V␤5 and V␤14 CJs look similar in wt thymocytes (Fig. 7B, lanes 4 and 5). It is thus reasonable to assume that, similar to V␤5, most of the V␤14 CJs detected in wt DP cells are indeed generated in DN cells and then expanded by ␤-selection. The failure to detect V␤14 SEs in wt DN thymocytes may be because of a specific feature(s) in the processing of these products, linked to the mode of V␤14 rearrangement by DNA inversion and the constraint to preserve chromosome integrity at the site of SJ formation (31). Conversely, increased accumulation of V␤14 SEs in DP thymocytes from p14 and E␤-deleted mice may reveal a disorder of the latter control in these cells and, indeed, a unique mode of allelic exclusion at the 3Ј end of the TCR␤ locus, evidenced here by the high frequency of attempted V␤14 rearrangement. Given the accessibility of the V␤14/DJ␤ loci, the level at which V␤14 23-nucleotide RSS cleavage can be observed in total DP cells is predicted to depend on the proportion of complementary 5Ј D␤ 12-necleotide RSS left available for synapsis (4,32,33).
Processing of V␤14 SEs into CJs at E␤ Ϫ alleles can be rescued, to some extent, by constitutive Bcl-2 expression. To check whether this also occurs in the presence of E␤, we analyzed thymocytes from p14 B2 double transgenic mice. Indeed, we found increased levels (ϳ2.9ϫ) of V␤14 CJs in p14 B2 compared with p14 thymocytes, whereas V␤14 SEs were roughly equivalent in DP cells from both types of mice (Fig. 7C). Altogether, the above data strongly suggest that exclusion of V␤14 rearrangement proceeds through a unique mechanism, one aspect of which could be a specific defect in the resolution of discrete DNA DSBs and, most likely, the induction of cell death. E␤ does not appear to interfere with these processes, including the initial steps of the recombination reaction (synapsis and RAG-mediated DNA cleavage) during attempted V␤14-to-DJ␤ rearrangement. DISCUSSION A Critical Function for E␤ in Promoting Access to the D␤-J␤ Domains in pro-T Cells-We have analyzed TCR␤ gene rearrangement in ⌬E␤ B2 thymocytes to better assess E␤ function in the regulation of CJ formation during V(D)J recombination. Compared with ⌬E␤, ⌬E␤ B2 thymocytes show a small accumulation of TCR␤ standard recombination products, most notably CJs, that is confined to proximal D␤-J␤ and V␤14 loci. Although detectable in DN CD25 ϩ pro-T cells, these effects predominate in cells that have differentiated to the DP stage. FIG. 7. Analysis of V␤ SE and CJ products in DN and DP thymocytes from E␤ ؉ and E␤ ؊ mice. A and C, lower panels, genomic DNA from total and/or DN-and DP-sorted thymocytes in the indicated mouse lines was analyzed by LM-PCR for the presence of V␤5 or V␤14 SEs (V␤5 SE and V␤14 SE, respectively) and by long range PCR for the presence of V␤5-(D)J␤2 or V␤14-(D)J␤2 CJs (V␤5 CJ and V␤14 CJ, respectively; B and C, upper panels). In A, thymus DNA from a Rag mouse was used as a negative control. Serial dilution analyses of wt thymus DNA shown in B and C were as in Fig. 1D. C, PhosphorImager scanning analysis of recombination products in p14 B2 versus p14 thymocytes gave the following results (after normalization to C␤ controls): CJs, 2.53/0.88; SEs, 2.2/1.8, respectively. Because control C␤2 amplifications (C␤) used oligonucleotide primers located, respectively, upstream of (5Ј primer) and within (3Ј primer) exon I of C␤2, the p14 transgene was not detected in this assay.
Developmental cell selection is unlikely to account for these findings, as no TCR␤ rearranged products could be detected at either the RNA or protein levels. Instead, evidence for VDJ recombinase activity at these loci also in E␤ ϩ DP thymocytes, and the fact that the rearrangement profiles at distinct TCR␤ loci in DN and DP ⌬E␤ thymocytes correlate with their level of chromosomal accessibility in the particular cell subset, argue for delayed access of the recombinase to E␤ Ϫ alleles. Besides pointing to a possible, likely indirect effect of E␤ in assisting CJ formation, these data most notably emphasize this critical function of the enhancer for chromatin opening within the D␤-J␤ domains, which, unexpectedly, appears also limited to an early window of T cell development only (i.e. anterior to the CD44 Ϫ/lo CD25 ϩ DN cell stage).
A Mechanism of Enhancement of D␤-to-J␤ CJ Formation by E␤?-Resolution of Rag-mediated DNA breaks that can happen in E␤-deleted thymocytes is improved by Tg Bcl-2 expression, apparently with a slightly greater impact on CJ compared with SJ formation (e.g. Fig. 2). May this tell us something about a putative role of E␤ in enhancing CE processing? First, the effects conferred by Bcl-2 transgenes on developmental processes in lymphoid cells (including V(D)J recombination) have generally been attributed to cell extended lifespan (27,34), although a non-conventional role of Bcl2 and incidental effect(s) on the processing of injured DNA cannot be formally ruled out (35). Second, the resolution of CE and SE products is thought to proceed along two different pathways, involving distinct requirements and kinetics (4). Thus, although a deficiency in any of the factors of the NHEJ apparatus results in a CJ defect, some proceed with unaltered SJ formation (e.g. the SCID defect resulting from DNA-PK deficiency). Also, whereas SEs accumulate in lymphoid cells undergoing V(D)J recombination (and are eventually resolved after RAG expression is down-regulated), CEs are difficult to detect in the wild-type situation (they accumulate in DNA-PK-deficient lymphocytes), indicating that CJ formation must be tightly linked to RSS cleavage. In possible relation to these distinct behaviors, in vitro studies have suggested diametrically opposed stability of V(D)J recombination post-cleavage complexes depending on product content; complexes consisting of the two CEs and two SEs appear to be highly unstable whereas those consisting of the two SEs bound by the RAG factors are resistant to dissociation challenges (reviewed in Ref. 6). We believe that a direct effect of E␤ in mediating the recruitment of DNA-PK (or a DNA-PK/Artemis complex (36)) is unlikely, based on the CJ sequences from ⌬E␤ and ⌬E␤ B2 thymocytes that do not show the typical abnormalities associated with the SCID defect (i.e. extensive deletions, frequent usage of short sequence homologies, long palindromic junctional inserts), along with the seeming lack of CEs and HJs at E␤ Ϫ loci in Tg Bcl-2 expressing T cells (16) (this study). However, an effect of E␤ on the stabilization of post-cleavage synaptic complexes would be compatible with the small accumulation of CJs observed in ⌬E␤ B2 thymocytes; unstable CEs (especially at E␤ Ϫ alleles) would have more chances of being resolved when cell survival is extended. As for HJ formation (not found in ⌬E␤ or ⌬E␤ B2 thymocytes), it could be especially sensitive to suboptimal conformation and/or stability of the post-cleavage synaptic complexes (24).
We stress that the effect of E␤ on chromatin structure at the D␤-J␤ region and an incidental effect on CJ formation may not be mutually exclusive. First, the maintenance of an open structure may be required for NHEJ repair. Second, E␤ modulation of chromosomal accessibility almost certainly involves nucleoprotein interactions including additional cis-regulatory element(s) (e.g. the D␤1 upstream promoter (25)). Within such structures, recombination synaptic complexes (D␤-to-J␤ and, subsequently, V␤-to-DJ␤) could be optimally organized for RAG cleavage and/or the processing of the cleaved extremities. As an integral component, E␤ (and bound factors) may thereby contribute to optimizing V(D)J repair processes, for example through the anchorage/tethering of loose extremities within interacting distances or, in a more sophisticated way, by favoring catalytic transitions within the synaptic complexes. As the efficiency of CJ formation is likely influenced by the efficiency with which loose CEs are recaptured by the post-cleavage complex (37), DNA tethering should improve this process. A model of post-synaptic conformational isomerization of a recombination complex has also been described during phage Mu DNA transposition (38). Importantly, enhancers at other TCR/Ig loci may share the dual functions of E␤ on coupled regulation of chromatin modulation and CJ formation, which could be revealed once the two effects can be distinguished.
Implications for T Cell Differentiation and the Control of Allelic Exclusion at the TCR␤ Locus-Allelic exclusion at the TCR␤ locus is regulated at the level of V␤-to-DJ␤ joining. In this context, evidence of chromatin compaction at 5Ј V␤ genes during the DN-to-DP cell transition, including reduced histone acetylation, has accumulated (39 -41) (this study). This change in organization likely contributes to lock in allelic exclusion, at least within alleles carrying a germline 5Ј V␤ cluster (because we used Rag mice in our analyses, it is still unclear whether unrearranged V␤ genes upstream of a V␤DJ␤ CJ unit behave similarly). We now demonstrate that developmentally regulated re-organization of the 5Ј V␤ genes does not require E␤, in line with their previously reported E␤-independent transcription in earlier DN thymocytes. Although long range accessibility within antigen receptor loci has been inferred to depend on matrix attachment regions (42), targeted deletion of matrix attachment region ␤ (located ϳ400 bp upstream of E␤) does not change TCR␤ transcription and V(D)J recombination in developing T cells (43). Likewise, targeted deletion of J␤2-C␤2 intronic cis-elements does not alter these processes (44). Therefore, the re-organization at 5Ј V␤ genes could rather involve as yet undefined regulatory sequences within the upstream part of the TCR␤ locus. Assuming that the promoters of 5Ј V␤ genes and of V␤14 behave similarly (45), our finding that V␤14 is spared by the repressive effect makes the individual 5Ј V␤ promoters unlikely candidates. Indeed, a 5Ј V␤ gene was no longer under allelic exclusion control when inserted, together with associated promoter sequences, in the region upstream of D␤1 (46).
In DP thymocytes, the 3Ј end of the TCR␤ locus containing the D␤-J␤-C␤ and V␤14 domains maintains (or gains in the case of E␤ Ϫ alleles and the D␤-J␤-C␤ clusters) chromosomal accessibility. This might involve the regulated activation of additional cis elements that normally act redundantly with E␤ within these loci (40). At this stage of development and in the wild-type situation, fully derepressed chromatin may be required to ensure high levels of expression of a rearranged V␤DJ␤-C␤ unit. The ensuing drawback is cleavage by the RAG machinery past the ␤-selection checkpoint, with consequences such as the specific accumulation of D␤ and V␤14 SEs and DJ␤ CJs in DP thymocytes (see Fig. 5, A and C and see Fig. 7A) (25). Nevertheless, the fact that levels of V␤14 CJs remain extremely low in p14 DP cells (Fig. 7B) indicates that these consequences on TCR␤ allelic exclusion are minimized, possibly involving regulated cell death (Fig. 7C). Recent results (47) suggest that programmed cell death may be a parameter that also limits ongoing rearrangements along the TCRJ␣ locus in DP thymocytes. In this context, the particular situation of V␤14 and inversional mode of rearrangement could concur to the surprisingly opposite outcomes of V␤14-to-DJ␤ versus D␤-to-J␤ attempted recombination in DP cells. Notably, intrachromosomal V␤14-to-D␤ SJs within this accessible part of the locus could be ideal targets for DSB formation by a RAGmediated nick-nick mechanism (26). It is, however, important to stress that the actual levels of V(D)J recombination at D␤-J␤ and V␤14 loci in DP thymocytes is unclear and could be quite low given the large amounts of germline-sized D␤-J␤ containing fragments detected by PCR and Southern assays of DNA from E␤ Ϫ (⌬E␤ B2) and E␤ ϩ (p14) thymocytes. 3 The reasons for these paradoxical effects of combining chromosomal accessibility and low levels of recombination within RAG expressing cells (48) warrant further investigating efforts.
Based on in vitro studies, cleaved SEs are thought to form excellent substrates for RAG-mediated transposition, with the risk to compromise genomic stability in lymphocytes leading possibly to translocation and leukemia (26). One would predict that ⌬E␤ B2 mice may be subject to these effects. However, neoplasia does not significantly affect these animals (but translocation into an enhancerless TCR␤ locus may not result in oncogene-deregulated expression in the developing lymphocytes), and we did not find evidence of TCR␤ genomic instability using FISH analysis and PCR assays (including assays to search for TCR␤-to-TCR␣/␦ interlocus recombination). Specific pathways might exist to counteract such events in DP thymocytes involving, for example, a role of the recombination machinery in the prevention of transposition (49,50) and/or sensors of injured DNA (e.g. p53, ATM) acting to uncover and eliminate such damages and/or the damaged cells (51). The latter possibility is supported by the finding that introduction of an E␤ mutation onto a p53-deficient background accelerates tumor development in T cells. 4 Our findings indicate that E␤ is not involved in the sophisticated controls that, following ␤-selection, secure allelic exclusion at the TCR␤ locus. However, allele asynchronicity of V␤ gene assembly in earlier DN pro-T cells is a prerequisite for allelic exclusion, as this should leave cells enough time to test for V␤-to-DJ␤ productivity before V␤ gene rearrangement proceeds on the other allele (7). The molecular basis for this phenomenon are still unclear. It may involve a recombination machinery that incidentally operates at suboptimal efficiency or structural features to ensure that a single allele rearrange at a time, including specific properties of D␤-flanking RSS (4) and/or an epigenetic mark(s) established at antigen receptor genes early in development (52). We stress that E␤ could be involved in any of these controls (except may be for the latter), as evidenced by the differential effect exerted on activation of the D␤-J␤-versus the V␤-containing chromosomal domains and associated regulatory elements (15). Our current results of a rigorous control by E␤ in regulating chromosomal access in early pro-T cells further sustain this hypothesis. This may explain the evolutionary constrain to maintain a function of E␤ that is limited both spatially (the D␤-J␤ domains) and temporally (prior to the CD44 Ϫ/lo CD25 ϩ DN cell stage).