Poor-Quality Vβ Recombination Signal Sequences and the DNA Damage Response ATM Kinase Collaborate to Establish TCRβ Gene Repertoire and Allelic Exclusion

This article is featured in Top Reads, p.

The ability of T and B lymphocyte populations to express a diverse array of Ag receptors is essential for adaptive immunity. This diversity is generated through the assembly of TCR and Ig genes by recombination of V, D, and J gene segments in developing T and B cells (1). In-frame V(D)J rearrangements and downstream C region exons create functional TCR and Ig genes. Semiconserved recombination signal sequences (RSSs) flank each V, D, and J gene segment and target the lymphocyte-specific RAG1/RAG2 (RAG) endonuclease (1). During V(D)J recombination, RAG binds to an RSS, captures another, and then induces DNA double-strand breaks (DSBs) between each synapsed RSS and its flanking segment, yielding two coding ends (CEs) and two signal ends (SEs) (1). RAG and the DSB response ATM protein kinase maintain RAG-cleaved DNA strands in postcleavage synaptic complexes, thereby facilitating the processing and repair of these DNA ends to produce a signal join (SJ) and a functional coding join (CJ) (2, 3). Most V(D)J rearrangements occur through deletion of intervening sequences to form a CJ within the chromosome and an SJ on an extrachromosomal excision circle (1). Others proceed via inversion of intervening sequences to create both a CJ and an SJ within the chromosome (1). The large combination of V(D)J rearrangements and the inherent imprecision in CJ formation together generate vast Ag receptor diversity. Gene segments are used nonrandomly during V(D)J recombination, implying that evolutionary pressure selected for factors that create beneficial Ag receptor gene repertoires. These factors likely include genetic variations of RSSs, epigenetic-based regulation of RSS accessibility to RAG complexes, and spatial proximity of gene segments. Indeed, for the mouse TCRβ (Tcrb) locus, differences in RSS qualities serve a fundamental role in establishing TCRβ gene repertoire in vivo (4–8).

Most T and B lymphocytes express TCR or Ig proteins from only a single allele. This allelic exclusion ensures identical specificity of every Ag receptor on an individual cell, thereby providing the molecular basis for highly specific immune responses (9). Allelic exclusion of TCRβ, IgH, and Igκ genes is achieved through protein from an in-frame V(D)J rearrangement on one allele signaling permanent feedback inhibition of V recombination on the other allele (9). As activation and enforcement of receptor-mediated feedback inhibition takes time, additional mechanisms must operate to prevent the second allele from initiating V recombination during this period. Although many potential factors have been proposed, including inefficient V rearrangements across large linear genomic distances, only two factors have been shown to dictate monoallelic recombination of V segments before and independent of receptor-enforced permanent feedback inhibition (4, 9, 10). One factor is the ubiquitous ATM kinase that is rapidly activated by DSBs and directs the cellular response to these genomic lesions. During V(D)J recombination, ATM facilitates CJ formation by maintaining CEs in postcleavage complexes and also signals transcriptional changes that regulate cellular survival or lymphocyte development (11, 12). How ATM helps enforce TCRβ, IgH, and Igκ allelic exclusion remains unknown (10, 13). However, at least for Igk loci, RAG DSBs induced during V recombination on one allele signal via ATM to repress RAG expression and block initiation of V rearrangement on the other allele until after the initial RAG DSBs are repaired (10, 14, 15). Although this transient feedback inhibition of V recombination likely helps mediate allelic exclusion, the ability of ATM to stabilize postcleavage complexes might increase the kinetics of CJ formation to shorten the interval between initiation of V recombination on one allele and resultant protein signaling to block V recombination on the other allele (10, 14, 15). The second factor is Vβ RSS quality. Weak Vβ RSSs enforce allelic exclusion by stochastically limiting the frequency that Vβ rearrangements initiate before establishment of TCRβ protein–signaled feedback inhibition, reducing the chance that functional TCRβ genes assemble on both alleles (4, 5). Elevating the rearrangement efficiency of a single Vβ segment through RSS replacement revealed that factors in addition to poor Vβ RSSs establish both interallelic competition for initiation of Vβ recombination and representation of Vβ gene segments in the αβ TCR repertoire, presumably to help impose TCRβ allelic exclusion (4). Notably, the potential functional relationship between monoallelic initiation of V segment rearrangements and ATM-mediated rapid DSB responses in enforcing allelic exclusion has not been addressed for TCRβ, IgH, or Igκ genes.

To determine whether ATM and poor-quality Vβ RSSs functionally interact to help establish TCRβ repertoire and allelic exclusion, we created and analyzed mice containing replacement of Vβ RSSs with the stronger 3′ Dβ1 RSS on ATM-sufficient and -deficient backgrounds. The mouse TCRβ locus (Tcrb) is comprised of 22 functional Vβ gene segments that reside upstream of two Dβ-Jβ-Cβ clusters (Dβ1-Jβ1-Cβ1 and Dβ2-Jβ2-Cβ2) and a single Vβ (V31) located just downstream of the Dβ-Jβ-Cβ clusters in the opposite transcriptional orientation as all other TCRβ coding sequences (Fig. 1A). V31 rearranges to either Dβ-Jβ-Cβ cluster only through inversion of intervening sequences (16). All other Vβ segments rearrange to either Dβ-Jβ-Cβ cluster by deletion, as well as to the Dβ2-Jβ2-Cβ2 cluster by inversion in rare instances following an out-of-frame inversional V31 rearrangement to the Dβ1-Jβ1-Cβ1 cluster (16, 17). Substitution of the weak RSSs of V2 and V31 with a better RSS elevates the frequencies that these Vβ segments initiate recombination, subsequently elevating both the representation of V2 and V31 in the TCRβ gene repertoire and the fraction of αβ T cells expressing TCRβ proteins from both alleles (4, 5). We demonstrate in this study that, when the same stronger RSS controls V2 and V31 rearrangements on opposite alleles, ATM deficiency increases the incidence of V2⁺ cells and reduces the proportion of V31⁺ and V2⁺V31⁺ cells. We show that ATM loss elevates the levels that V2 or V31 coding sequences generate a hybrid join (HJ) with a 5′ Dβ RSS to produce a nonfunctional rearrangement, with this effect greatest for V31^R rearrangements. We also find that substitution of the poor V1 and V2 RSSs with the stronger RSS raises the fractions of cells that express V1⁺, V2⁺, or both V1⁺ and V2⁺ TCRβ proteins. ATM deficiency further elevates these incidences by reducing competition between alleles for recombination of RSS-enhanced Vβ segments. Our data demonstrate that RSS-determined inefficient initiation of Vβ recombination and ATM-directed rapid DSB responses cooperate to establish the primary repertoire and allelic exclusion of TCRβ genes.

All mice assayed were 4 to 6 wk old, of mixed sex, and housed under specific pathogen-free conditions. All husbandry, breeding, and studies were performed in accordance with national guidelines and regulations and approved by the International Animal Care and Use Committee. V2^R/+/V31^+/R, Atm^+/−, and Eβ^−/− mice and genotyping strategies were reported previously (4, 18). The starting V2^R/+/V31^+/R mice were of a pure C57BL/6 background, whereas the starting Atm^+/− mice were of a mixed C57BL/6 and 129SvEv background. CRISPR was used to make V1^R/+ mice with the guide 5′-GACACAGTGGTAAACTCTGC-3′. The Transgenic Core microinjected C57BL/6 zygotes with a mix of CRISPR RNA, trans activating CRISPR RNA, Cas9 protein, and a single-stranded oligonucleotide DNA as described (19): single-stranded oligonucleotide DNA, 5′-GAGGCTGCAAGTGGCCAACATGAGCCAGGGCAGAACCTTGTACTGCACCTGCAGTGCAGACACGGTGATTCAATTCTATGGGAAGCCTTTACAAAAACCACACACAGACTACCCTGCCTTCCAAGCCTTGCTCCCTGCAAGCCCTTCTGAGCTTTCTT-3′ (Integrated DNA Technologies). For genotyping, primer 5′-TCGGCCACATTAGCTGTCTACATCC-3′ is used with primer 5′-CACGGTGATTCAATTCTATGGGAAGCCTT-3′ to identify V1^R alleles or primer 5′-CACAGTGGTAAACTCTGCAGGCG-3′ to identify wild-type (WT) alleles.

Single-cell suspensions were prepared and stained as described (4). Data were collected on an LSRFortessa and analyzed with FlowJo (Tree Star).

TaqMan quantitative PCR was performed to detect HJs using: 5′ Dβ1 RSS probe, 5′-TTCCAGCCCTCAAGGGGTAGAC-3′; 5′ Dβ1 RSS primer, 5′-GTCACCTTCCTTATCTTCAACTCCCCC-3′; 5′ Dβ2 RSS probe 5′-GGGTAGGCACCTGTGGGGAA-3′; 5′ Dβ2 RSS primer, 5′-TCCCAGCCCCTCTCAGTCAG-3′; V31 primer, 5′-AAATCAAGCCCTAACCTCTAC-3′; V2 primer, 5′-GTTCAAAGAAAAACCATTTAG-3′; and V1 primer, 5′-GCCACACGGGTCACTGATAC-3′. Rearrangements were normalized to the CD19 gene (4, 18).

Data are reported as mean ± SD. Statistical analyses were performed with Prism 8.

We first assayed effects of ATM deficiency on TCRβ expression in our mice containing V2 and V31 RSSs on opposite alleles replaced with the stronger 3′ Dβ1 RSS (4). In these V2^R/+/V31^+/R mice, the fractions of αβ T cells that express V2⁺, V31⁺, or both V2⁺ and V31⁺ TCRβ proteins are much greater than normal due to increased frequency of initiation of V2 and V31 recombination before TCRβ protein-mediated feedback inhibition (4). We bred V2^R/+/V31^+/R mice with Atm^+/− mice to create and analyze WT, Atm^−/−, V2^R/+/V31^+/R, and V2^R/+/V31^+/R:Atm^−/− mice using flow cytometry to determine the fractions of naive αβ T cells that express V2⁺, V31⁺, or V2⁺ and V31⁺ TCRβ proteins. We studied thymic αβ T cells to avoid influences from potential differential expansion or anatomical distribution of peripheral cells with altered Vβ repertoire or biallelic expression of TCRβ proteins. As the representation of a Vβ segment in the αβ TCR repertoire of thymic αβ T cells from C57BL/6 and 129SvEv mice mirrors the frequency that the Vβ segment recombines in double-negative (DN) thymocytes (4, 6, 20), this flow cytometry approach provides a surrogate for quantifying the relative frequencies that the native and RSS-replaced V2 and/or V31 segments recombine. This approach also facilitates comparison across all four genotypes because ATM deficiency decreases the numbers of developing and mature αβ T cells due to impaired V(D)J recombination and cellular proliferation (21). In line with our previous report (4), most αβ T cells in V2^R/+/V31^+/R mice express V2⁺ (33%) or V31⁺ (41%) TCRβ protein (Fig. 1B, 1D), and ∼3% expresses both V2⁺ and 31⁺ TCRβ proteins, each of which is dramatically higher than in WT mice (Fig. 1C, 1E). Although most V2^R/+/V31^+/R:Atm^−/− cells also express V2⁺ (47%) or V31⁺ (27%) TCRβ protein, these frequencies were increased by 42% or decreased by 34%, respectively, as compared with V2^R/+/V31^+/R mice (Fig. 1B, 1D). Reflecting our previous study (13), we observed similar frequencies of V2⁺ and V31⁺ cells in WT and Atm^−/− mice (Fig. 1B, 1D) and an increased proportion of V2⁺V31⁺ cells in Atm^−/− mice versus WT mice (Fig. 1C, 1E), although this difference was not significant with the sample sizes assayed in this study. In contrast, the frequency of V2⁺V31⁺ cells was decreased by 26.5% in V2^R/+/V31^+/R:Atm^−/− mice relative to V2^R/+/V31^+/R mice (2.0% versus 2.72%) (Fig. 1C, 1E). Although the numbers of thymocytes were lower in Atm^−/− and V2^R/+/V31^+/R:Atm^−/− mice relative to WT and V2^R/+/V31^+/R mice as expected, these numbers were equivalent within the ATM-deficient and -sufficient backgrounds (Fig. 1F). These results indicate that ATM deficiency elevates usage of V2 and decreases usage of V31 when the high-quality 3′ Dβ1 RSS controls rearrangement of these Vβ segments. Therefore, we conclude that ATM functions with the poor V2 and V31 RSSs to shape the primary TCRβ gene repertoire.

The opposite effect of ATM inactivation on V2 and V31 usage implies that ATM has a greater role in promoting successful recombination of V31 when the high-quality 3′ Dβ1 RSS controls recombination of these two Vβ segments. One difference between V31 and V2 rearrangements is that V31 recombines only by inversion, whereas V2 recombines predominantly by deletion (16, 17). ATM promotes CJ formation by stabilizing postcleavage complexes to maintain RAG-liberated DNA ends in proximity with each other (11, 22). During inversional rearrangements, ATM promotes CJ formation by suppressing nonfunctional deletions that result from an HJ between a CE and an SE (11, 23). ATM presumably inhibits HJs during inversional rearrangements by preventing the escape of the CE and SE on the genomic fragment between RAG cleavage sites (11). The escape of DNA between RAG cleavage sites would prevent CJ formation during rearrangement of V31 but not necessarily V2. In this scenario, the retained CEs of V31 and Dβ could resolve only into an HJ that deletes Dβ-Jβ-Cβ sequences (Fig. 2A), whereas retained V2 and Dβ CEs during V2 rearrangement through deletion could generate a functional CJ (Fig. 2B). For rare V2 rearrangements to the Dβ2-Jβ2-Cβ2 cluster by inversion, loss of genomic sequences between RAG cleavage sites would prevent CJ formation and could result in an HJ that deletes Dβ-Jβ-Cβ sequences (Fig. 2C). To our knowledge, a function for ATM in suppressing HJ formation during deletional recombination has not been reported. The formation of an HJ during V2 recombination by deletion would lead to a reciprocal HJ forming a nonfunctional inversion (Fig. 2D), aberrant repair generating a translocation, or persistence of a broken chromosome. Notably, none of these aberrant V2 rearrangements would produce a functional TCRβ gene. Considering these scenarios, the simplest explanation for the altered Vβ usage caused by ATM deficiency on the V2^R/+/V31^+/R background would be a higher frequency of aberrant, nonfunctional rearrangements for V31^R versus V2^R.

To measure nonfunctional rearrangements, we used TaqMan PCR to quantify V2 or V31 HJs with 5′ Dβ RSSs in total thymocytes of WT, Atm^−/−, V2^R/+/V31^+/R, and V2^R/+/V31^+/R:Atm^−/− mice. Consistent with published data (23), we detected HJs between V31 and each 5′ Dβ RSS at a higher level in Atm^−/− mice than WT mice (Fig. 2E). Notably, we also observed HJs between V2 and each 5′ Dβ RSS at a higher level in Atm^−/− mice than WT mice (Fig. 2E), which, to our knowledge, is the first evidence that ATM suppresses HJ formation during V(D)J recombination by deletion. Although V31/Dβ and V2/Dβ HJs are present at higher levels in Atm^−/− versus WT mice, these do not correlate with lower representation of V31⁺ or V2⁺ cells (13) (Fig. 1B, 1D), implying that ATM loss causes similar fractions of attempted V31 and V2 rearrangements to resolve as nonfunctional events. Consistent with this notion, the levels of V31/Dβ and V2/Dβ HJs each increased to a similar extent with ATM deficiency (Fig. 2E). When the 3′ Dβ1 RSS controls V2 and V31 rearrangements on opposite alleles in the V2^R/+/V31^+/R background, we also detected HJs between V31 or V2 and each 5′ Dβ RSS at a higher level with inactivation of ATM (Fig. 2E). The absolute levels and fold increases for each of these HJs were greater for V31^R than V2^R when ATM is inactivated (Fig. 2E). These data demonstrate that ATM has a greater role in suppressing nonfunctional rearrangement of V31 than V2 when the strong 3′ Dβ1 RSS directs recombination of these Vβ segments. Notably, this difference provides a molecular explanation for the increased usage of V2, decreased utilization of V31, and lower frequency of V2⁺V31⁺ αβ T cells caused by ATM deficiency in V2^R/+/V31^+/R mice.

The nonfunctional inversional V31^R rearrangements caused by ATM deficiency precludes the conclusion as to whether ATM and poor-quality Vβ RSSs functionally interact to help enforce TCRβ allelic exclusion. To address this issue, we decided to improve the RSS quality of V2 and another Vβ segment that recombines by deletion to avoid the greater increased level of nonfunctional V31 rearrangements by inversion when ATM is inactivated. We made mice carrying replacement of the V1 RSS with the 3′ Dβ1 RSS, which is stronger than the V1 RSS for recombination with Dβ RSSs (24). We first sought to determine whether this V1^R allele behaves like the V2^R allele. For this purpose, we generated and analyzed mice with the V1^R allele opposite a WT allele (V1^R/+) or an allele in which all Vβ rearrangements are blocked by deletion of the Tcrb locus enhancer (V1^R/Eβ⁻). We observed that the fraction of αβ T cells expressing V1⁺ TCRβ proteins was 2.8-fold higher in V1^R/+ mice relative to WT mice (Fig. 3A, 3B), indicating that poor V1 RSS quality determines V1 usage in the TCRβ gene repertoire by limiting the incidence of V1 rearrangement. The frequency of V1⁺ αβ T cells was 1.4-fold higher in V1^R/Eβ⁻ mice versus V1^R/+ mice (Fig. 3A, 3B), showing that V1^R and Vβ segments on the opposite allele compete for recombination and usage in the TCRβ gene repertoire. We previously reported that the frequency of V2⁺ αβ T cells is 6-fold greater than normal in V2^R/+ mice and increased an additional 1.5-fold higher in V2^R/Eβ⁻ mice (4). Accordingly, the V1^R allele behaves analogous to the V2^R allele, differing only in that the 3′ Dβ1 RSS increases the rearrangement of V1 to a lesser extent than V2.

After characterizing the V1^R allele, we bred V1^R/+ mice with V2^+/R and Atm^−/− mice to make and study WT, V1^R/+/V2^+/R, and V1^R/+/V2^+/R:Atm^−/− mice. We observed that the fractions of cells expressing V1⁺, V2⁺, or both V1⁺ and V2⁺ TCRβ proteins were increased in V1^R/+/V2^+/R mice relative to WT mice (Fig. 3D–G). These populations were increased by 2.0-fold for V1⁺ cells, 6.0-fold for V2⁺ cells, and 6.2-fold for V1⁺V2⁺ cells (Fig. 3D–G). Notably, V1⁺V2⁺ cells must express V1⁺ and V2⁺ TCRβ proteins from opposite alleles because a V1 rearrangement would delete a germline V2 segment or any V2 rearrangement on that allele. Thus, like the weak V2 and V31 RSSs (4), the poor-quality V1 RSS enforces TCRβ allelic exclusion by limiting the frequency of initiation of Vβ recombination before TCRβ protein–mediated silencing. We found that the proportions of V1⁺, V2⁺, and V1⁺V2⁺ αβ T cells were increased in V1^R/+/V2^+/R:Atm^−/− mice as compared with V1^R/+/V2^+/R mice, being elevated by 28.8% for V1⁺ cells, 4.2% for V2⁺ cells, and 57.1% for V1⁺V2⁺ cells (Fig. 3D–G). The numbers of thymocytes in V1^R/+/V2^+/R and V1^R/+/V2^+/R:Atm^−/− mice were equivalent to those in V2^R/+/V31^+/R and V2^R/+/V31^+/R:Atm^−/− mice, respectively (compare (Fig. 3H with 1F). Therefore, when the strong 3′ Dβ1 RSS controls V1 and V2 rearrangements on opposite alleles, ATM deficiency elevates both the usage of these Vβs in the TCRβ gene repertoire and the frequency of αβ T cells that express both V1⁺ and V2⁺ TCRβ proteins. Collectively, these data demonstrate that the poor-quality V1 and V2 RSSs functionally interact with each other and with ATM to help establish repertoire and allelic exclusion of TCRβ genes.

The increased frequencies of V1⁺, V2⁺, and V1⁺V2⁺ αβ T cells caused by ATM deficiency on the V1^R/+/V2^+/R background implies that ATM might help enforce TCRβ allelic exclusion by facilitating competition between alleles for initiation of Vβ recombination. We have shown that Vβ segments on a normal Tcrb allele compete with V1^R (Fig. 3A) or V2^R (4) on the opposite allele for rearrangement and usage in the TCRβ gene repertoire. The lower frequency of V1⁺ cells in V1^R/+/V2^+/R mice relative to V1^R/+ mice (12.5% versus 17.0%) (compare (Fig. 3A, 3B with 3D, 3E) indicates that V2^R is better than the normal V2 gene segment for competing with V1^R on the opposite allele. Accordingly, the greater increase in the frequency of V1⁺ cells as compared with V2⁺ cells upon ATM inactivation in V1^R/+/V2^+/R mice (Fig. 3D, 3E) is consistent with diminished interallelic competition for Vβ recombination.

To investigate further whether ATM might facilitate competition between alleles for Vβ recombination, we created and analyzed V1^R/R and V2^R/R mice on ATM-sufficient and -deficient backgrounds. Although the numbers of thymocytes were lower in V1^R/R:Atm^−/− and V2^R/R:Atm^−/− mice relative to V1^R/R and V2^R/R mice as expected, these numbers were equivalent within the ATM-deficient and -sufficient backgrounds (Fig. 4A). We used flow cytometry analysis of Vβ usage as a surrogate to measure in-frame CJs and TaqMan PCR to quantify HJs. We observed that ATM deficiency raised the incidence of V1⁺ or V2⁺ cells, with the effect more pronounced for V1⁺ cells (1.34-fold for V1⁺ cells versus 1.08-fold for V2⁺ cells) (Fig. 4B–E). We found that ATM inactivation increased the levels of HJs between V1^R or V2^R and each 5′ Dβ RSS (Fig. 4F, 4G), with the gains slightly higher for V1^R than V2^R. Although these differences might simply reflect the greater increase of V1^R rearrangement, they cannot account for the increased V1^R usage in the TCRβ gene repertoire because the HJs assayed preclude functional Vβ rearrangements. The frequency of V1⁺ cells in V1^R/R:Atm^−/− mice was 1.9-fold higher than in V1^R/+ mice (32.5% versus 17.0%) (compare (Fig. 4B, 4D with 3A, 3B). In the absence of interallelic competition for Vβ recombination, one would expect a slightly less than 2-fold increase in V1^R usage because some cells would assemble functional V1^R rearrangements on both alleles. Accordingly, the most likely explanation for the 1.9-fold increase in V1⁺ cells, especially when considered with increased HJs in V1^R/R:Atm^−/− mice relative to V1^R/R mice, is that ATM inactivation diminishes interallelic competition for V1 recombination and resulting usage in the TCRβ gene repertoire. Although we did not analyze V2^R/+ mice, the frequency of V2⁺ cells in V2^R/R:Atm^−/− mice was 1.7-fold higher than we previously reported for V2^R/+ mice (68.9% versus 40.0%) (4). This less pronounced effect of ATM deficiency at increasing V2⁺ cells versus V1⁺ cells most likely reflects that V2^R rearrangement cannot increase further from ATM loss due to competition from other Vβ segments on the same allele. However, we did not create genetic backgrounds necessary to distinguish contributions of ATM in mediating interallelic versus intra-allelic competition for rearrangement between V2^R and other Vβ segments. Nevertheless, we conclude that ATM facilitates interallelic competition for initiation of recombination of the V1^R and V2^R gene segments.

Finally, we sought to determine whether ATM also facilitates interallelic competition for V31 recombination despite ATM deficiency on the V2^R/+/V31^+/R background lowering the frequencies of V31⁺ and V2⁺V31⁺ cells. For this purpose, we created and analyzed V31^R/R mice on the ATM-sufficient and -deficient background. The numbers of thymocytes were lower in V31^R/R:Atm^−/− mice relative to V31^R/R mice as expected and equivalent to those within the other ATM-deficient and -sufficient backgrounds studied (Fig. 4A). We observed a slightly lower frequency of V31⁺ thymocytes in V31^R/R:Atm^−/− mice versus V31^R/R mice, but this difference was not significant with the numbers of animals analyzed (Fig. 4H, 4I). In contrast, we detected substantially higher levels of HJs between V31^R and each 5′ Dβ RSS in V31^R/R:Atm^−/− mice relative to V31^R/R mice (Fig. 4J). The only plausible explanation for these observations is that ATM deficiency increases both the frequency of initiation of V31 recombination and the fraction of attempted V31 rearrangements that resolve as HJs. In this context, ATM deficiency on the V31^R/R background has modest effect on V31 usage in the TCRβ gene repertoire because higher levels of CJ formation on one allele compensates for elevated levels of HJ formation on the opposite allele. Therefore, we conclude that ATM also facilitates competition between alleles for initiation of V31^R rearrangement.

Our study provides insights into how ATM and RSSs govern V(D)J recombination. It has been known for years that ATM promotes CJ formation during chromosomal V(D)J recombination by maintaining CEs in RAG postcleavage synaptic complexes (11). Presumably, it is the loss of this function that increases the levels that attempted inversional rearrangements of V31, Vδ5, and numerous Vκ segments yield HJs and resultant nonfunctional intralocus deletions in ATM-deficient mice (23). We now show that ATM deficiency elevates the levels that attempted deletional Vβ rearrangements produce HJs between Vβ coding sequences and 5′ Dβ RSSs. We did not determine whether the associated Vβ RSS SEs and Dβ CEs form HJs to generate inversions, repair with other DNA sequences to yield translocations, or remain unrepaired to maintain broken chromosomes. Nevertheless, these three outcomes all would preclude the assembly of functional TCRβ genes. To our knowledge, our results are the first demonstration that ATM promotes CJ formation during deletional V(D)J recombination by inhibiting HJ formation. ATM might do this by preventing the irreversible escape of one CE from RAG postcleavage complexes and/or helping RAG proteins orient CEs and SEs for ligation into a CJ and an SJ. We also discover that ATM deficiency and enhanced Vβ RSS strength have additive effects on increasing HJs during rearrangements of V1 or V2, but synergistic effects on increasing HJs during V31 rearrangements. This reveals that ATM has a more impactful role in promoting CJ formation by suppressing HJs when the strong 3′ Dβ1 RSS directs inversional V31 recombination versus deletional Vβ recombination. Although ATM exerts a dominant function for maintaining CEs in postcleavage complexes (11), it functions redundantly with the DNA-PKcs DSB response protein to do the same for SEs (22). Considering these points, we reason that the most likely explanation for why ATM is more important for suppressing HJ formation during rearrangement of V31^R than V1^R or V2^R is that properties of the 3′ Dβ1 RSS render it more likely than Vβ RSSs to irreversibly escape from postcleavage complexes in the absence of ATM. In this scenario, the maintained other SE and two CEs could form a CJ, an HJ, or an open-and-shut join between the Dβ CE and SE. For V1^R or V2^R recombination, CJs would produce an intact chromosome, be transcribed, and, if in-frame, make TCRβ proteins that drive DN thymocyte proliferation and differentiation. In contrast, for V31^R recombination, CJs would be on a broken chromosome and likely not transcribed due to DSB-induced transcriptional silencing (25). Yet, even if these CJs were transcribed, the ability of any resultant TCRβ protein to drive proliferation and differentiation would be countered by the broken chromosome ends signaling apoptosis. Such potential higher propensity of the 3′ Dβ1 RSS to escape postcleavage complexes in the absence of ATM should not impair Dβ-to-Jβ recombination but could increase the frequency that the excised SEs are not repaired into SJs and instead insert into other genomic locations. Regardless of the precise underlying mechanisms, our results demonstrate that ATM and the poor-quality V31 RSS collaborate to shape the primary TCRβ gene repertoire by hindering Vβ rearrangements from aberrantly resolving as nonfunctional HJs.

Our data indicate that RSS-determined inefficiency of Vβ rearrangements and ATM-mediated DSB responses cooperate to help enforce TCRβ allelic exclusion. We previously showed that improving the strength of the RSS fused to V2 or V31 increases the frequency that each Vβ initiates rearrangement and is represented in the TCRβ gene repertoire. These RSS enhancements revealed that Tcrb alleles compete for recombination of V2^R and V31^R segments, but these RSS-replaced Vβ gene segments on opposite alleles still increase the frequencies that αβ T cells expressing V2⁺ and V31⁺ TCRβ proteins from the same or distinct alleles (4). In this study, we have demonstrated that enhancing the strength of the RSS of V1 increases the: 1) frequencies that V1 rearranges and is represented in the TCRβ gene repertoire, 2) ability of its Tcrb allele to compete with the V2^R allele, and 3) percentage of αβ T cells that express V1⁺ and V2⁺ TCRβ proteins from opposite alleles. These results provide additional evidence that weak Vβ RSSs serve a fundamental role in helping enforce TCRβ allelic exclusion by stochastically decreasing the likelihood that both alleles initiate Vβ recombination before resulting protein from one allele can permanently halt Vβ recombination of the other allele. The lower representation of V2⁺ cells in V2^R/+/V31^+/R mice relative to V1^R/+/V2^+/R mice indicates that V31^R is better than V1^R for competing against V2^R for recombination and usage in the TCRβ gene repertoire. We previously demonstrated that V31^R outcompetes V2^R for recombination (4, 5). These findings are consistent with the possibility that Vβ segments located closer genomic distances to the Dβ-Jβ-Cβ clusters outcompete Vβ segments residing further upstream. However, RSS replacements of additional Vβ gene segments are necessary to determine the contribution of this factor versus other factors related to the chromosomal environment. We show in this study that ATM helps enforce TCRβ allelic exclusion by facilitating interallelic competition for Vβ rearrangement, even when the efficiency of Vβ recombination on both alleles is increased by elevated RSS quality. ATM does not enforce TCRβ allelic exclusion alone, as evidenced by the percentages of αβ T cells expressing V1⁺ and V2⁺ proteins in V1^R/+/V2^+/R:Atm^−/− mice (∼0.05%) or V2⁺ and V31⁺ proteins in V2^R/+/V31^+/R:Atm^−/− mice (∼2%), each of which is much lower than expected in the absence of other mechanisms. Without additional control, the expected percentages would be the products of the frequencies of αβ T cells expressing each Vβ alone, which would be ∼6% for V1^R/+/V2^+/R:Atm^−/− mice and ∼12% for V2^R/+/V31^+/R:Atm^−/− mice. Additional mechanisms that mediate TCRβ allelic exclusion include TCRβ protein signaled silencing of Vβ recombination and most likely association of Tcrb alleles with the nuclear lamina to repress Vβ accessibility and looping between Vβ and Dβ-Jβ segments (9, 26–28). Notably, despite the greater increase in HJs during recombination of V31^R than V1^R, a higher frequency of cells expresses V31⁺ and V2⁺ proteins in V2^R/+/V31^+/R:Atm^−/− mice than V1⁺ and V2⁺ proteins in V1^R/+/V2^+/R:Atm^−/− mice. This could be explained by V31^R initiating recombination and generating CJs at higher frequencies than V1^R even when ATM is inactivated. Furthermore, the greater increase in V1⁺V2⁺ cells versus V2⁺V31⁺ cells from ATM loss opens the possibility that V31 rearrangements are repressed less frequently by ATM and/or less effective at activating ATM signals that feedback inhibit Vβ recombination. The fact that ATM deficiency on the V31^R/R background has no significant effect on V31 representation in the TCRβ gene repertoire but causes a substantial increase in V31 HJs is consistent with these scenarios.

ATM activities in DSB repair, signaling from DSBs, or both could help enforce TCRβ allelic exclusion by facilitating competition between alleles for initiation of Vβ recombination. ATM-directed stabilization of RAG postcleavage complexes might accelerate CJ formation, reducing the time between onset of Vβ recombination on one allele and resultant TCRβ protein signaling to block Vβ recombination on the other allele. During this time period, weak Vβ RSSs would limit initiation of Vβ rearrangement on the other allele. Alternatively, rapid ATM signals from RAG DSBs could transiently feedback inhibit the start of Vβ recombination on the second allele before CJ formation on the first allele terminates ATM signaling. By the time or shortly after ATM signals cease, TCRβ protein expressed from the first allele would enforce permanent silencing of Vβ recombination on the second allele. For each of these scenarios, out-of-frame Vβ rearrangements incapable of making protein would not elicit feedback inhibition and allow initiation of Vβ recombination on the other allele. Although a role for ATM signaling remains speculative, observations from developing B cells illuminate possible mechanisms. In response to RAG DSBs induced during Igκ recombination in pre-B cells, ATM signals many transcriptional changes, including repression of Rag1/Rag2 expression and Igκ accessibility (10, 12, 29). The disruption of these two responses by inactivating an effector of ATM signaling relieves competition for RAG cleavage between Igκ alleles and increases the fraction of B cells that assemble and express Igκ genes on both alleles (15). ATM also appears to respond to RAG DSBs induced during V gene segment recombination on one IgH or Igκ allele by repositioning the other allele to repressive pericentric heterochromatin (30). It is possible that RAG DSBs introduced during Vβ recombination on one allele signal through ATM to repress RAG expression, accessibility of the other allele via changes in its nuclear location and/or transcriptional activity, chromosome looping between Vβ and Dβ-Jβ segments, or some combination of these and perhaps additional mechanisms. Determining the contributions of ATM functions in DSB repair versus DSB signaling at facilitating interallelic competition for Vβ recombination requires elucidating and then inactivating relevant ATM signaling pathways without impairing ATM DSB repair. We employed this approach in pre-B cells to show that ATM signaling from RAG DSBs, and not ATM-facilitated CJ formation, helps govern monoallelic initiation of Vκ recombination to enforce Igκ allelic exclusion (15). Yet, we cannot assume that the same ATM-dependent mechanisms operate in thymocytes to facilitate interallelic competition for Vβ rearrangement and even for different Vβ segments. In the latter context, the fact that ATM deficiency in V1^R/+/V2^+/R mice increases the usage of V1^R relatively more than V2^R could reflect that ATM accelerates the kinetics of CJ formation and/or silences Vβ accessibility more for V1^R. Nevertheless, our data demonstrate that ATM-directed DNA damage responses serve a fundamental role in functioning with poor-quality Vβ RSSs to help orchestrate monoallelic assembly and expression of TCRβ genes.

The mechanisms by which ATM facilitates intra-allelic competition for Vβ recombination also might control initiation of V(D)J recombination among Tcrb, Tcrg, and Tcrd loci. The DN thymocyte population assembles TCRβ, TCRγ, and/or TCRδ genes through V(D)J recombination. The resultant expression of γδ TCRs signals silencing of Rag1/Rag2 expression and differentiation of γδ T cells, whereas TCRβ protein expression creates pre-TCR complexes that signal repression of Rag1/Rag2 expression and differentiation of DP thymocytes (31). To our knowledge, there is no evidence for interallelic regulation of V recombination at Tcrg and Tcrd loci, which explains why ∼20% of γδ T cells contain in-frame Tcrd genes on both alleles (32). The frequency of DN thymocytes with DSB repair protein complexes present simultaneously on multiple TCR loci is 2- to 3-fold higher in Atm^−/− mice versus WT mice (33). The escape of CEs from RAG postcleavage complexes caused by ATM deficiency could account for this difference by decreasing the kinetics of CJ formation, which would extend the time before γδ TCRs or pre-TCRs could halt Rag1/Rag2 expression and thus RAG cleavage of other loci. Moreover, considering that DSBs from ionizing radiation repress Rag1/Rag2 expression in DN thymocytes (14), RAG DSBs induced on any Tcrb, Tcrg, or Tcrd allele could signal via ATM to transiently inhibit further RAG cleavage of any TCR locus by at least downregulating Rag1/Rag2 transcription. ATM-deficient mice invariably succumb to thymic lymphomas with oncogenic Tcrd locus translocations that arise when pre-TCR signals stimulate proliferation and differentiation of DN thymocytes with unrepaired RAG DSBs at Tcrd gene segments (34–36). Although the loss of ATM-dependent repair of CEs and activation of cell cycle checkpoints predispose to these translocations, additional ATM function in coordinating the initiation of RAG DSBs among TCR loci might contribute. Exploring these possibilities requires determining contributions of ATM functions in repair versus signaling to suppress DSB repair protein complexes simultaneously on multiple TCR loci and, if warranted, inactivating relevant ATM signaling pathways without impairing ATM DSB repair.

Our observations provide important new insights into factors that determine Ag receptor gene repertoires. Preselection repertoires are established through the relative recombination frequencies of individual gene segments. Computational analyses on the molecular features of TCRβ, IgH, and Igκ loci predict that V segment chromatin accessibility is the predominant factor that determines relative V rearrangement frequencies, whereas qualities of V RSSs and their degree of contact with (D)J segments play minor roles (37–39). A model based on computational analysis of epigenetic factors across TCRβ in DN thymocytes predicts that V1 and V2 have equivalent accessibility and predicts that placing V1 and V2 under control of the same RSS would lead to higher relative rearrangement of V1 because it has more contact with Dβ-Jβ segments (37). However, our data show increased relative rearrangements of V2 when the strong 3′ Dβ1 RSS controls V1 and V2 segments on opposite alleles. Two non–mutually exclusive possibilities could account for this difference between predicted and empirical results. One is that the algorithm used to calculate RSS quality is not accurate because it does not consider either the partner RSS or sequences surrounding either RSS (40), each of which can alter levels of recombination (24, 41–43). In this context, Tcrb sequences flanking the V1 and/or V2 RSSs could influence 3′ Dβ1 RSS activity, rendering V2^R with a higher intrinsic quality for recombination to Dβ-Jβ RSSs than V1^R. Another possibility is that the 11 parameters used for computational chromatin profiling are not sufficient for defining Vβ chromatin accessibility (37). Considering that nucleosome occupancy of RSSs inhibits their recombination and RSS variations alter nucleosome positioning (44), nucleosome mapping over RSSs and their flanking sequences should be included in computational models for V(D)J recombination. Finally, our finding that ATM cooperates with RSS identity to shape Vβ repertoire argues for the inclusion of activities for ATM and possibly other DSB response factors in computational-based models of Ag receptor gene repertoires. Regardless, our observations highlight the critical need to experimentally test in vivo models of V(D)J recombination formulated from computational analyses of correlative phenomena.

We thank Kyutae Lee for designing the V1^R allele and Adele Harman for electroporating reagents. We also thank Rebecca Glynn for helpful discussions of the work.

The authors have no financial conflicts of interest.