|Home | About | Journals | Submit | Contact Us | Français|
Histone 3 lysine 4 trimethylation (H3K4me3) is associated with promoters of active genes and found at hot spots for DNA recombination. Here we have shown that PAXIP1, a protein associated with MLL3 and MLL4 methyltransferase and the DNA damage response, regulates RAG-mediated cleavage and repair during V(D)J recombination in CD4+ CD8+ DP thymocytes. Loss of PAXIP1 in developing thymocytes diminished Jα H3K4me3 and germline transcription, suppressed double strand break formation at 3’ Jα segments but resulted in accumulation of unresolved T cell receptor a-chain gene (Tcra) breaks. Moreover, PAXIP1 was essential for release of mature single positive (SP) αβT cells from the thymus through transcriptional activation of sphingosine-1-phosphate receptor S1pr1 as well as for natural killer T cell development. Thus, in addition to maintaining genome integrity during Tcra rearrangements, PAXIP1 controls distinct transcriptional programs during DP differentiation necessary for Tcra locus accessibility, licensing mature thymocytes for trafficking and natural killer T cell development.
DNA-mediated processes including transcription and DNA repair are regulated by post-translational modifications of the histone components of chromatin (Taverna et al., 2007). Post-translational modifications are carried out by histone methyltransferases and demethylases, histone acetyltransferases, protein kinases, and ubiquitin-and SUMO-protein ligases. Specific histone modifications are recognized by so called “reader-effector” modules in non-histone proteins, Such reader proteins frequently carry enzymatic activity on the chromatin template which can render the DNA more accessible to the transcription and/or DNA repair machineries (Taverna et al., 2007).
The Recombination Activating Genes (Rag1 and Rag2) are lymphoid specific proteins that are essential for V(D)J recombination, a process that generates diversity in the antigen receptor repertoire (Gellert, 2002; Schatz and Swanson, 2011). RAG1 and RAG2 (herein referred to as RAG) bind and cleave DNA at specific recombination signal sequences (RSSs) that flank each V, D and J gene segment. Thereafter the liberated DNA ends are recognized and repaired by the non-homologous end joining (NHEJ) pathway. The ability of RAG to initiate V(D)J recombination is dictated by the accessibility of RSSs within chromatin (Krangel, 2007; Osipovich and Oltz, 2010).
The first clue about the mechanisms regulating RAG activity was the discovery that germline transcripts of antigen receptor gene segments correlate with recombinationally active open chromatin configuration susceptible to double strand break (DSB) cleavage (Yancopoulos and Alt, 1985). The link between transcription, chromatin structure and recombination was strengthened by studies documenting the effects of transcriptional blockade on downstream histone modifications and RAG-dependent DSBs (Abarrategui and Krangel, 2006). A further link between germline transcription and increased accessibility was the discovery that RAG is targeted to histone 3 lysine 4 trimethylation (H3K4me3) (Liu et al., 2007; Matthews et al., 2007). The RAG2 protein contains a PHD finger at its C terminus, which interacts with H3K4me3. H3K4me3 is highly enriched in the 5’ end of all transcription units including chromatin encompassing recombinationally active genes in the immunoglobulin and T cell receptor loci. A recent study reported that RAG2 binds to virtually all genomic regions containing H3K4me3 (Ji et al., 2010). Moreover, RAG2 association with H3K4me3 is dependent on the RAG2 PHD domain (Ji et al., 2010), and lymphocyte development is partially impaired by PHD mutations that affect RAG2’s recognition of H3K4me3 (Matthews et al., 2007). Finally, biochemical studies revealed that H3K4me3 stimulates the catalytic function of RAG (Grundy et al., 2010; Shimazaki et al., 2009). Altogether, these studies suggest a model in which RAG2 association with H3K4me3 via its PHD-finger domain enhances DNA binding and cleavage. In this way beyond the recognition of the RSS in the DNA, the PHD chromatin reader in RAG recombinase increases the efficiency of V(D)J recombination.
Trimethylation of H3K4 is mediated by H3K4 methylases, which include at least 6 members (Set1A, Set1B, MLL1, MLL2, MLL3 and MLL4), each of which contains a conserved SET domain carrying the methyltransferase activity. Each H3K4 methylase exists in a large protein complex which shares four common subunits ASH2L, RBBP5, WDR5 and DPY30. Deletion of WDR5 results in a marked reduction of the recombination activity of RAG2 (Matthews et al., 2007), consistent with a role for H3K4me3 in enhancing V(D)J recombination. In addition to shared subunit composition, methylase complexes contain unique components which may provide target specificity to the complex. For example, PAXIP1 (Pax interaction with transcription-activation domain protein-1), is a subunit of the MLL3 and MLL4 complex (Cho et al., 2007; Issaeva et al., 2007; Patel et al., 2007) which regulates H3K4me3 and germline transcription initiation at IgG3 and IgG1 switch regions during class switch recombination (Daniel et al., 2010; Schwab et al., 2011) (MLL3 is also known as KMT2C, and MLL4 is also known as ALR, MLL2, or KMT2D)(Cho et al., 2012). Transcription of switch regions is thought to render them in an accessible configuration that allows DNA cleavage by the enzyme, activation induced deaminase (AID) (Stavnezer et al., 2008). Nevertheless, loss of Paxip1 does not affect amounts of H3K4me3 or transcription of Sμ and Sε, indicating that the PAXIP1-MLL3-MLL4 complex promotes accessibility of some but not all switch loci (Daniel et al., 2010).
RAG and AID-dependent DNA breaks at antigen receptor locus are marked by the phosphorylation of the histone variant H2AX (γ-H2AX) (Chen et al., 2000; Petersen et al., 2001). In addition to promoting germline transcription at the IgH locus, PAXIP1 exists in a separate complex that accumulates at sites of DNA damage in a manner dependent on the γ-H2AX DNA damage response pathway (Gong et al., 2009). Beyond promoting AID accessibility for IgG3 and IgG1 class switch recombination, PAXIP1 appears to function subsequently in the repair of a subset of AID dependent DSBs (Daniel et al., 2010).
Because H3K4me3 is thought to influence the accessibility of chromatin at V(D)J gene segments, it is of interest to determine the methylase activities that target the recombinase to specific loci. Here we have analyzed thymic development in mice deficient in PAXIP1. We have found that PAXIP1controls RAG mediated DSB formation and repair during Tcra recombination. Moreover, the development of natural killer T (NKT) cells, which require recombination and expression of a very limited and invariant TCRα chain, was ablated in the absence of PAXIP1. Surprisingly, mature Paxip1−/− thymocytes that successfully completed V(D)J recombination failed to exit the thymus due to impaired expression of the G protein-coupled sphingosine-1 phosphate receptor 1 (S1PR1). Thus, PAXIP1 regulates both recombinase accessibility and DNA repair in double positive (DP) T cells as well as the exit of mature single positive (SP) T cells from the thymus. We propose that PAXIP1 serves as a gatekeeper that ensures that cells with faulty rearrangements are not exported to the periphery.
To determine the physiologic consequence of Paxip1 loss during T cell development, we crossed Paxip1f/f mice with mice expressing Cre recombinase driven by the Lck proximal promoter (Hennet et al., 1995), which is active in immature thymocytes. Expression of Paxip1 in mutant thymocytes was barely detectable by quantitative reverse transcriptase-polymerase chain reaction (RT-PCR), confirming the efficiency of deletion (Fig. S1). In Paxip1f/f × Lck Cre (referred to as Paxip1−/−) mice, total thymocyte recovery was reduced to 50% of Paxip1f/f (referred to as WT) mice (Figures 1A and 1B). Consistent with this, staining for CD4 and CD8 co-receptors in 6- to 10- week old mice revealed a 25% decrease in the percentage and 50% decrease in the number of double positive (DP) thymocytes (Figures 1A and 1B), whereas the percentage of double negative (DN) thymocytes was similar (Figure S2A). The DN-DP thymocyte transition is associated with an approximately 100-fold expansion in thymus cellularity. We considered the possibility that reduced number of DP T cells might reflect inefficient proliferation. To test this, we examined the expression of CD25 which gets diluted during the DN-DP T cell transition (Crompton et al., 1994)(Figure S2B). Consistent with this, we observed a 3-fold increase in the percentage of Paxip1−/− DP T cells expressing CD25, indicative of decreased proliferation (Figure S2B).
Unlike the reduction in DP thymoctyes, we observed an unexpected increase in the number of CD4+ and CD8+ SP T cells in Paxip1−/− thymi (Figures 1A and 1B). SP thymocytes that accumulate in the absence of Paxip1 were mature T cells, evidenced by a greater frequency of Paxip1−/− SP T cells that express high amounts of TCRβ, CD69, and the maturation marker Qa2, and low amounts of heat stable antigen (HSA) in both CD4+ and CD8+ populations (Figures 1A and 1C; Figure S2C;). Associated with the accumulation of SP T cells in the thymus, there was a marked depletion in the number of mature Paxip1 deficient T cells in the periphery (Figure 2). Whereas 78% of cells in WT lymph nodes were T cells, defined as TCRβ+, only 42% of Paxip1−/− cells expressed TCRβ; moreover, there was 80% reduction in the number of CD4+ and CD8+ T cells in this compartment (n=16 mice analyzed) (Figure 2A). Finally, Paxip1 deletion resulted in a 75% reduction in the number of splenic T cells (n=9) as well as a reduction of T cells in blood (Fig. 2B). The paucity of peripheral T cells (Figure 2) concomitant with the accumulation of mature T cells in thymus (Figure 1) suggested the possibility that PAXIP1 was required for T cell migration from the thymus to the periphery (see below).
The Tcra locus becomes accessible at the DP stage which correlates with Jα chromatin marked by H3K4me3 (Krangel, 2007). We used chromatin immunoprecipitation (ChIP) coupled with Illumina sequencing to examine genome wide H3K4me3 profiles of DP T cells (CD4+CD8+TCRlo/med), CD4+ SP T cells (CD4+CD8−TCRhi) and CD8+ SP T cells (CD4−CD8+TCRhi) populations from sorted WT and Paxip1−/− thymocytes. Primary rearrangements utilize Jα segments at the 5’ end of the 65-kb Jα array and secondary Vα-Jα rearrangements make use of additional Jα segments located 3’ of the primary VαJα (Krangel, 2007). In WT DP T cells, we found that the sequence tag distribution of H3K4me3 was evenly spread along Jα locus as previously reported (Ji et al., 2010) (Figure 3A). In contrast, Paxip1−/− DP T cells displayed a skewed H3K4me3 profile with the highest H3K4me3 signals at the 5’ Jα segments and a gradual decay toward the 3’ end of the Jα array (Figure 3A). For example, in WT DP thymocytes, 41% of the total tag counts within the 65 kb Jα array were 3’ of Jα30, whereas tag counts at the 3’ end of the locus were reduced to 13% in Paxip1−/− DPs (Figure 3A). Consistent with the results in DP T cells, the distribution of H3K4me3 in Paxip1−/− CD4+ SP and Paxip1−/− CD8+ SP T cells also showed skewing towards 5’ relative to 3’ segments (Figure S3A). Thus, loss of Paxip1 results in decreased H3K4me3 at 3’ Jα segments in both DP and SP T cells.
To determine whether decreased H3K4me3 at 3’ Jα segments correlated with impairment in Vα to Jα rearrangements, we examined the Jα repertoire in WT and Paxip1−/− thymocytes. TCRβlo/med DP cells were sorted and TCR transcripts were reverse transcribed and PCR amplified with a TRAV12 (Vα8) family primer and a Cα primer, as previously described (Guo et al., 2002; Hawwari et al., 2005). Jα usage was then assessed by Southern blotting with a series of Jα-specific oligonucleotide probes. This analysis of cDNA revealed greater utilization of 5’ segments and less usage of 3’ Jα segments in Paxip1−/− thymocytes relative to WT (Figure 3B). Moreover, DP thymocytes from Paxip1−/− mice showed decreased TCR Cα mRNA as measured by quantitative RT-PCR (Figure 3C).
To determine whether skewed Jα usage is due to biased Vα to Jα recombination, we utilized DNA from sorted TCRβlo/med DP thymocytes to amplify genomic fragments generated by TRAV12 (Vα8) to Jα rearrangements by real-time PCR. Recombination from TRAV12 to 5’ Jα segments were detected at a similar amount in Paxip1−/− and WT thymocytes (Figure 3D). However, DNA rearrangements to more distal 3’Jα segments (eg. Jα37-Jα2) were reduced in Paxip1 mutant cells relative to WT thymocytes (Figure 3D). Thus, decreased 3’ Jα usage and transcription (Figures 3B and 3C) correlates with defective Vα to Jα rearrangements (Figure 3D) in Paxip1−/− thymocytes.
Altered 3’ Jα usage and recombination in Paxip1 deficient thymocytes could result from increased programmed cell death of DP thymocytes as seen in Rorc-deficient mice (Guo et al., 2002), deregulation of Rag expression in DP thymocytes (Yannoutsos et al., 2001), or decreased cleavage of Tcra segments. We found that the viability of Paxip1−/− thymocytes (62%) was comparable to WT (67%) after overnight culture (Figure S3B). Moreover, RAG expression was not decreased in the absence of Paxip1 (Figure S3C), indicating that lack of 3’ Jα usage was not due to shortened DP T cells lifespan or deregulated recombinase expression.
Although RAG expression was intact in Paxip1−/− thymocytes (Figure S3C), impaired RAG-mediated cleavage of 3’ Jα segments could account for altered Jα usage and recombination. To examine the efficiency of DNA double strand break formation, we performed ligation-mediated PCR to amplify Jα signal ends as described (McMurry et al., 1997). We found that DSBs formed efficiently in Paxip1−/− thymocytes at 5’ Jα segments (eg. Jα61, Jα58 and Jα49) (Figure 3E). However, there was a dramatic decrease in DSB formation at 3’ end of the Jα array (e.g. at Jα27, Jα13 and Jα6) (Figure 3E). Thus, deficiency in PAXIP1-dependent H3K4me3 at 3’ Jα genes correlates with impaired cleavage of those segments.
Primary Tcra rearrangements are dependent on transcription mediated by T early α (TEA) and Jα49 promoters at the 5’ end of the array, which limits recombination to the 5’ Jα61-Jα52 gene segments. After initial Vα-Jα rearrangements, Vα promoters drive subsequent rounds of secondary rearrangements to Jα segments that are more 3’ (Hawwari and Krangel, 2007). During primary and secondary rearrangements, the accessibility to limited sets of Jα segments correlates with activating histone marks such as H3K4 tri-methylation (Abarrategui and Krangel, 2006).
One possible explanation for the alteration in TCRα repertoire is that Paxip1 deficiency reduces H3K4me3 and germline transcription throughout Jα rearrangements. If this were the case, the most profound recombination deficit might appear at the 3’ end of the locus, as observed, because 3’ recombination is dependent on prior 5’ rearrangements (Krangel, 2007), and each step would be less efficient in the absence of Paxip1. To examine the effects of Paxip1 on H3K4me3 and transcription at the 5’end of the locus without the confounding effects of ongoing rearrangements, we crossed Paxip1−/− mice with Rag2−/− mice expressing a rearranged Tcrb transgene. In the absence of Rag2, all recombination is eliminated. However, the presence of the Tcrb transgene allowed the Tcra locus to be transcriptionally active, and in a configuration whereby only 5’ Jα segments were accessible. Comparison of H3K4me3 in the 5’ end of the Jα array in Rag2−/− × TCRββ × Paxip1f/f × Lck Cre+ (Rag2−/− Paxip1−/−) vs. Rag2−/− × TCRββ × Paxip1f/f × Lck Cre − (Rag2−/−) DP T cells revealed that PAXIP1 deficiency resulted in a 44% reduction in the number of sequence tags focused at Jα61-Jα52 (Figure 3F). In addition, there was a reduction in germline transcripts at all accessible Jα elements and an overall reduction in Tcra transcripts in Rag2−/− Paxip1−/− DP T cells relative to Rag2−/− (Figure 3G). These data suggest that PAXIP1 contributes to the accessibility of the Tcra locus by promoting H3K4me3 and germline transcription throughout the Jα cluster.
In addition to its role in transcription, PAXIP1 is recruited to sites of DNA damage where it promotes DSB repair by homologous recombination (Wang et al., 2010b). We observed that 5’ Jα segments are cleaved in the absence of Paxip1 (Figure 3E), and therefore we could assess whether DSBs are efficiently resolved. To ascertain whether Tcra and Tcrd locus integrity is compromised, we assayed for TCRα and δ associated chromosomal aberrations in metaphases spreads prepared from thymocytes. We performed FISH analysis with a Tcra and Tcrd locus probe which hybridizes to the middle of chromosome 14 together with a telomeric probe, which marks the ends of all chromosomes. Whereas TCRα and δ associated chromosomal aberrations were undetectable in WT thymocytes, 2.2% of Paxip1−/− thymocytes showed a broken chromosome 14 evident by the Tcra probe hybridizing to the end of a shortened chromosome 14 (Figure 3H). This amount of TCRα and δ associated breaks is similar to that observed in mice lacking the NHEJ factor 53BP1 (Difilippantonio et al., 2008). In addition to TCRα and δ associated aberrations, there was a 4-fold increase in spontaneous DNA damage in Paxip1−/− thymoctyes (Figure 3H), consistent with a role for PAXIP1 in repairing replication-associated DSBs (Wang et al., 2010b), and with the observed decreased cellularity of Paxip1−/− thymocytes (Figure 1B and Figure S2B). We conclude that Jα rearrangements are less efficient in the absence of PAXIP1 because of impaired accessibility, cleavage, and DNA repair.
Like conventional αβ T cell lineages, NK T cells arise from DP precursors in the thymus through secondary Tcra rearrangement, TCR-mediated signaling and selection (Godfrey and Berzins, 2007). NK T cells which express an invariant Vα14-Jα18 T cell receptor recognize lipid antigen presented by the molecule CD1d. Whereas Paxip1 deficiency resulted in a 50% decrease in the total number of thymocytes, there was 90–95% decrease in the number of NK T cells stained by specific CD1d tetramers in the mutant thymus and spleen (n=4 mice analyzed) (Figures. 4A–C). To determine the stage of NK T cell development affected by the absence of Paxip1 we analyzed CD1d tetramer positive thymocytes for their expression of HSA, CD44 and NK1.1. The few NKT cells in the Paxip1−/− thymus accumulated in the early immature populations (CD1dtetramer+HSA+CD44loNK1.1−) (Figure 4A), suggesting that the block in NKT cell development is at early stages of positive selection.
To directly examine Vα14-Jα18 rearrangements, we performed quantitative RT-PCR with Vα and Jα specific primers (Gapin et al., 2001). Consistent with the fact that Jα18 is localized at the 3’ end of the Jα cluster, which is under-utilized in the absence of PAXIP1 (Figures. 3A–E), we found that Vα14-Jα18 recombination was reduced 14 times in Paxip1−/− thymocytes (Figure 4D). Failure to generate the invariant Vα14-Jα18 T cell receptor provides an explanation for why NKT cell development is arrested at early stages in the absence of PAXIP1.
Next, we attempted to identify genes with deregulated H3K4me3 during the DP to SP T cell transition. To this end, we compared the ChIP-Seq of H3K4me3 in DP, CD4+ SP, and CD8+ SP T cell populations at gene promoters. In WT, the vast majority (27494; 98%) of H3K4me3 enriched genes were those that were shared amongst DP and SP T cells (Figure 5A, Table S1 and Figure S4A). Only 20 out of 27494 genes lost H3K4me3 during the DP to SP T cell transition (Figure 5A; inset, row S1 and Table S1) and 192 promoters gained H3K4me3 in both SP (CD4+ and CD8+) T cell populations (Figure 5A, inset, row S6 and Table S1). Nevertheless, a very small subset of H3K4me3 enriched genes were found uniquely in CD8+ and CD4+ T cells (121 and 166 genes respectively, Figure 5A, inset, row S4 and S5 and Table S1). For example Runx3, which is up-regulated during CD8+ T cell differentiation (Wang et al., 2010a), showed H3K4me3 in CD8+ but not in CD4+ SP T cells; Zbtb7b, which is upregulated in CD4+ SP T cells (He et al., 2010), exhibited H3K4me3 binding in CD4+ but not in CD8+ SP T cells. Nevertheless, very few genes (<2% of total) acquire H3K4me3 during the DP to SP T cell transition.
We next determined whether PAXIP1 contributed to H3K4 trimethylation by comparing the distribution of H3K4me3 in WT and Paxip1−/− sorted populations. In DP and SP T cells, the vast majority of promoters (>99.95%) showed no dependence of H3K4me3 on Paxip1 (Figure 5B). For example, when utilizing a 3-fold threshold, only 10 genes out of 21,902 unique RefSeq gene symbols showed altered H3K4me3 binding in Paxip1−/− DP cells (Figure 5B), and only 93 genes showed at least a 2-fold change (Fig. S4B and Table S2). The majority of H3K4me3-bound genes (3 genes in DPs, 36 genes in CD4+ and 32 genes in CD8+ T cells) that were regulated by PAXIP1 showed an increase in H3K4me3 in WT relative to Paxip1−/− (Figure 5B, Tables S2–4 and Figure S4B). Notable among those genes that exhibited major deficits in H3K4me3 in Paxip1−/− thymocytes was Paxip1 itself (likely because the construct deletes the Paxip1 promoter, transcriptional start and exon 1 (Patel et al., 2007)) and the sphingosine-1 phosphate receptor 1 S1pr1 (Figures 5A, 5C and 5D).
To determine whether there is a correlation between the genes that showed large deficits in H3K4me3 in the absence of Paxip1 (Figure 5B) and genes showing large alterations in H3K4me3 in WT cells during the DP to SP T cell transition (Figure 5A), we compared changes of H3K4me3 in SP Paxip1−/− relative to SP WT vs. DP WT relative to SP WT T cells (Figures 5C and 5D). We found that 20 out of the 36 (56%) Paxip1-dependent genes that were deregulated in CD4+ SP T cells (Figure 5B) altered their amounts of H3K4me3 by more than 3 times during the DP-SP T cell transition (Figure 5C). Similarly, 19 of 32 (60%) Paxip1-dependent genes in CD8+ SP T cells (Figure 5B) exhibited a major increase in H3K4me3 during the DP-SP T cell transition (Figure 5D). Overall, there was a greater effect from Paxip1 deficiency on H3K4me3 for those genes that gained H3K4me3, and thereby become transcriptionally activated during DP-SP T cell differentiation, relative to those genes that did not change H3K4me3 notably (Figures 5 C and 5D, genes in quadrants I and II compared to III and IV, P(Chi-square test) < 1e-16 for both CD4+ and CD8+ SP T cells). Thus, despite the fact that PAXIP1 was essential for H3K4 trimethylation in less than 0.16% of promoters in SP T cells (36 out of 21902) (Figure 5B), PAXIP1 contributed disproportionately to promoters in which H3K4me3 increases during the DP-SP T cell transition (27 out of 313 DP-SP promoters or 8.6%).
To determine why only a subset of genes is a target for PAXIP1-dependent H3K4me3, we performed genome-wide localization studies using ChIP-seq for PAXIP1 in WT and Paxip1−/− thymocytes. Because PAXIP1-binding showed some background signals in Paxip1−/− thymocytes, we considered only those differentially enriched regions at gene promoters (n=20,808) in which PAXIP1-binding in WT exceeded Paxip1 deficient by greater than 3-fold with False Discovery Rates less than 0.01 (see methods). To more rigorously assess the PAXIP1-bound DNA sequences, we examined them for the enrichment of consensus motifs identified from publicly available ChIP-Seq databases (see methods). We identified 25 motifs with more than 2-fold enrichment and log likelihood ratio greater than 3 (Table S5; see extended methods). Using this method, we identified the ZBTB16 and PAX family motifs as among the highest scoring DNA sequences (Figure 6A and Table S5). Overall, 77% of the PAXIP1-binding sites scored positive for at least one ZBTB16 consensus motif and more than 56% were positive for at least one consensus paired domain or homeobox motif. It is known that PAXIP1 and PAX1 physically associate and colocalize in active chromatin (Lechner et al., 2000), and ZBTB16 has also been reported to interact with PAXIP1 (Rual et al., 2005). These observations support the specificity of our PAXIP1 binding sequences.
The list of identified PAXIP1 bindings was then used for Gene Ontology (GO) enrichment analysis, which enabled the identification of biological processes that are affected by PAXIP1. Interestingly, the top 11 enriched processes associated with the strongest PAXIP1 targets were those involved in T cell/lymphocyte differentiation, activation and V(D)J recombination (Figure 6B and Table S6), suggestive of T-cell specific roles for PAXIP1 in thymocytes .
PAXIP1 preferentially bound to promoters and overlapped with the closest nucleosome-free region to the transcription start site, as determined by the normalized distribution of PAXIP1 and H3K4me3 around transcription start sites (TSSs) (from +/−2 kb relative to TSS) (Figure 6C). On average, 91% of promoters occupied by H3K4me3 had PAXIP1 binding, and conversely 90% of PAXIP1-occupied promoters were also bound by H3K4me3. Given the strong positive effect of PAXIP1 on differential H3K4me3 and expression of several genes that were up-regulated during the DP-SP T cell transition (Figure 5), we asked whether we could discern the global effect of Paxip1 on cognate sites of H3K4me3. By comparing the genome-wide localization of PAXIP1 to H3K4me3, we found that increased PAXIP1 binding intensity correlated with the extent of H3K4 trimethylation (Figure S5A and S5B). Specifically, those genes in SP lineages which showed the greatest deficits in H3K4me3 in the absence of PAXIP1 were associated with significant PAXIP1binding in WT T cells (Figures S5A and S5B).
To further determine the frequency of PAXIP1-dependent genes that exhibited direct PAXIP1 binding, we identified the number of PAXIP1-occupied promoters that had at least 2-fold increase in PAXIP1 dependent H3K4 tri-methylation (as determined in Figure S4B). We compared this to control groups consisting of the same number of genes taken from the next most relevant PAXIP1-dependent promoters with the highest increase in H3K4me3 not exceeding 2-fold. We observed 83%, 63% and 25% increase in direct PAXIP1 association with the most significant PAXIP1-dependent promoters in CD8+ SP, CD4+ SP, and DP T cell populations respectively relative to the controls (Figure 6D). Therefore, PAXIP1-dependent increases in H3K4me3 frequently result from direct binding of the PAXIP1 associated methyltransferase complex near the promoter.
Since CD4+ and CD8+ SP T cells accumulate in PAXIP1-deficient thymocytes (Figure 1), we hypothesized that genes involved in cell migration might be deregulated in both populations. We therefore compared the distribution of H3K4me3 in sorted CD4+ SP and CD8+ SP T cells from WT and Paxip1−/− thymocytes. Relative to Paxip1−/− CD4+ T cells, 36 genes from WT CD4+ T cells had increased expression of H3K4me3 near their promoters (Figure 5B and Figure 7A). 32 genes in WT CD8+ T cells also had increased H3K4me3 deposition relative to Paxip1−/− CD8+ thymocytes (Figure 5B and Figure 7A). Among these genes with deregulated H3K4me3 at their promoters, only 9 exhibited an H3K4me3 deficit in both Paxip1−/− CD4+ and CD8+ T lineages which accumulate in the mutant thymus (Figure 7A).
Examination of the list of common (CD4+ and CD8+ T cell) down-regulated genes in Paxip1−/− SP T cells revealed that the most important alteration, after Paxip1 itself, was in the sphingosine-1-phosphate receptor S1pr1 (Figure 7B and Table S7), which is essential for both αβ and γδ thymocyte and peripheral T cell emigration (Allende et al., 2004; Matloubian et al., 2004; Odumade et al., 2010). We observed abundant signal for H3K4me3 near S1pr1 for both CD4+ and CD8+ T cell lineages, whereas H3K4me3 was barely detectable in the absence of Paxip1 (Figure S6A). Another factor implicated in T cell homeostatis was IL6st (Figure 7B), which encodes the IL-6 receptor subunit, gp130. Surface staining for gp130 revealed that Paxip1−/− thymocytes were severely impaired in gp130 expression (Figure S6B).
In addition to S1PR1, the only other factors that have been implicated in thymocyte emigration are FOXO1 and KLF2 DNA binding proteins (Freitas and Rocha, 2009). Expression of all three factors (Figure 7C) and H3K4me3 association near their promoters (Figure S4A) is up-regulated during the DP-SP T cell transition; moreover, these gene products appeared to function in the same pathway, since FOXO1 is critical for KLF2 expression, which in turn drives S1PR1 expression (Freitas and Rocha, 2009; Kerdiles et al., 2009). To determine whether PAXIP1 promotes the expression of transcription factors critical for thymocyte emigration, we directly assayed Foxo1, Klf2 and S1pr1 expression in DP and SP cells by quantitative RT-PCR. As predicted, all three genes were up-regulated during the DP-SP transition in WT cells. Similarly, Foxo1 and Klf2 were also up-regulated in PAXIP1-deficient cells; however, S1PR1 RNA and protein failed to be induced in Paxip1−/− SP thymocytes (Figures 7C and 7D).
We conclude that amongst the factors known to drive thymocyte egress and the size of the peripheral pool, S1PR1 appears to be uniquely regulated by PAXIP1. Moreover, PAXIP1 bound directly to the S1PR1 promoter (Figure S5C). S1PR1 deficiency is characterized by a greater number of SP cells expressing β7 integrin and intermediate amount of CD69 in SP thymocytes (Matloubian et al., 2004), and Paxip1−/− SP thymocytes mirrored this phenotype (Figure 7E). Additionally, we found a profound accumulation of HSAlo mature γδ T cells in the thymus which also require PAXIP1 for thymic export (Odumade et al., 2010)(Figure S2D). Together, these data demonstrate that among several genes up-regulated in mature SP thymocytes, PAXIP1 is essential for induction of S1pr1 during the DP to SP T cell transition.
In addition to S1pr1 and Paxip1, there are 7 genes in which H3K4me3 is mis-regulated by more than 3-fold and 74 genes decreased by at least 50% in Paxip1−/− mature SP thymocytes (Figures 7A and 7B; Table S7). To determine whether failure to up-regulate S1pr1 contributes to the accumulation of mature thymoctyes, we crossed Paxip1-deficient mice with mice expressing high amounts of an S1PR1 transgene under control of the human Cd2 promoter (Liu et al., 2009). In contrast to Paxip1 deficiency, SP thymocytes no longer accumulated in S1PR1Tg LckCre Paxip1f/f (Paxip1−/−S1PR1Tg) mice (Figure 7F), and the frequency of CD4+ and CD8+ SP T cells in Paxip1−/−S1PR1Tg was even lower than in WT. Moreover, S1PR1 overexpression normalized Paxip1−/− SP maturation, evidenced by a decrease in frequency of Paxip1−/−S1PR1Tg cells expressing CD69 and HSA, and an increase in SP cells expressing β7-integrin relative to Paxip1−/− thymocytes (Figure 7F and Figure 1C). We conclude that among various deregulated genes in SP thymocytes, defective S1PR1 expression is a primary factor influencing the accumulation of T cells in the Paxip1 deficient thymus.
Here we have shown that the MLL3 and 4 specific co-factor PAXIP1 regulates H3K4me3 in a subset of genes during thymocyte development. PAXIP1 controlled accessibility in DP thymocytes by promoting H3K4me3, transcription and cleavage within Jα chromatin. Based on these findings, it is likely that RAG2 binding to the Jα locus is also reduced in the absence of Paxip1; this could be the result of defective transcription and associated chromatin opening as well as from diminished interaction between the RAG2 PHD domain and H3K4me3. In any case, the PAXIP1-associated activity cannot be the only methyltransferase that targets RAG to the Tcra locus since the reduction in H3K4me3 and transcription was only 2-times lower in Paxip1−/− thymocytes relative to WT. A consequence of this redundancy is that Jα rearrangements are not eliminated in the absence of PAXIP1 but instead are focused to the 5’ end of the Jα locus. Nevertheless, NK T cell development is blocked due to failure to recombine to the Jα18 segment which lies at the 3’ end of the Tcra locus.
While the 5’ end of the Jα locus is efficiently broken, it is inefficiently repaired, evidenced by the accumulation of TCRα and δ associated chromosomal aberrations in Paxip1−/− thymocytes. RAG mediated breaks are repaired by the NHEJ pathway. Interestingly, PAXIP1 has been implicated in promoting homologous recombination rather than NHEJ (Wang et al., 2010b). Our ChIP analysis of genomic regions bound by PAXIP1 revealed that PAXIP1 associates near promoters (within 2kb of TSSs) of known NHEJ factors including XRCC4, DNA-PKcs and XRCC6 (Ku80). This suggests the possibility that PAXIP1’s role in Tcra recombination might be linked to transcriptional regulation of NHEJ factors. However, we have found no difference in the expression of NHEJ factors in WT and Paxip1−/− thymocytes as measured by quantitative RT-PCR (not shown). Since PAXIP1 is recruited to DSBs in a MLL4-independent but γ-H2AX-dependent manner (Gong et al., 2009), it therefore seems more likely that PAXIP1’s role in repair of RAG mediated DSBs is directly mediated by recruitment to DSBs rather than transcription of NHEJ proteins.
PAXIP1 regulates H3K4me3 in a small fraction of promoters in developing T cells. Among the genes that are positively regulated by PAXIP1 in CD4+ or CD8+ SP T cells, approximately 50% are those that changed H3K4me3 notably during the DP-SP T cell transition. Moreover, many of the genes that showed PAXIP1 dependence for H3K4me3 also exhibited PAXIP1 binding near their promoters. This indicates that PAXIP1 activity is most important for a subset of T cell specific genes whose expression is induced during DP to SP T cell differentiation. This includes S1pr1, but does not include other up-regulated genes such as Bcl2, Sell and Il7r.
It has been reported that PAX2 and PAX5 recruit PAXIP1 and the H3K4 methyltransferase complex to PAX-dependent genes (Patel et al., 2007; Schwab et al., 2011). If this mode of recruitment is generalizable, it suggests that complexes containing PAXIP1 and methyltransferases are targeted to DNA through interactions with sequence specific transcription factors. According to this hypothesis, transcription factors such as FOXO1 or KLF2 might facilitate association of PAXIP1 with the S1pr1 promoter, which in turn recruits RNA polymerase II through ensuing H3K4me3. Other genes such as Bcl2, Sell and Il7r would be induced in a PAXIP1-independent manner because transcription factors that activate these promoters would not interact with PAXIP1.
Is there a connection between PAXIP1 functions in Tcra rearrangement and thymoctye export? One possibility is that PAXIP1 serves as a gatekeeper that ensures that developing thymocytes with faulty rearrangements are not exported to the periphery. Previously we have shown that in addition to its role in DNA double strand break repair, the ATM kinase prevents the transmission of DNA breaks to daughter cells (Callen et al., 2007). Impairment of both these repair and checkpoint functions may account for the high amount of antigen receptor associated breaks in mature peripheral Atm−/− lymphocytes, some of which are generated in earlier development as a result of failed V(D)J recombination (Callen et al., 2007). By promoting phosphorylation of H2AX, ATM is also required for the recruitment of PAXIP1 to DSB sites (Gong et al., 2009). PAXIP1 accumulation at DNA damage sites occurs independently of the MLL3 and 4 methyltransferase complex (Gong et al., 2009). Nevertheless, PAXIP1 association with sites of DNA damage during Tcra rearrangements might effectively deplete MLL3 and 4 of its transcription targeting co-factor PAXIP1. In this case, there would be less efficient induction of PAXIP1-dependent genes such as S1PR1 during the DNA damage response, which in turn would reduce thymic export. These two specialized roles of PAXIP1 in DNA repair and transcription might therefore have evolved as part of a checkpoint that prevents the propagation of potentially oncogenic DNA damage out of the thymus.
Paxip1f/f (Kim et al., 2007), TCRβ transgenic (Shinkai et al., 1993), and Lck Cre transgenic (Hennet et al., 1995) and S1PR1 transgenic (Liu et al., 2009) mice were generated as previously described. Rag2−/− mice were obtained from Taconic Laboratories. All experiments were performed in compliance with NIH Intramural Animal Care and Use program.
Analysis of Jα usage in sorted TCRβlo/med DP thymocytes was performed by Southern blot and probing of TRAV12 (Vα8)-Cα RT-PCR products with Jα-specific probes as described (Guo et al 2002; Hawwari et al. 2005). TCRα transcripts were analyzed by quantitative real-time PCR using a QuantiFast SYBR Green PCR kit (Qiagen). TCRα coding joints were analyzed in genomic DNA of sorted DP thymocytes by quantitative real-time PCR as above, with normalization to B2m. Primer sequences are provided in Table S8. Jα DSBs were quantified in genomic DNA of sorted DP thymocytes by ligation-mediated PCR (McMurry et al 1997) using primers and probes described (Seitan et al., 2011). Metaphases were obtained from thymocytes stimulated with anti-TCR (H57, 2 µg/ml; PharMingen) and anti-CD28 antibodies (5 µg/ml; PharMingen) and analyzed with BAC probes containing the TCRα locus (TCR Cα-232F18) and telomere-repeat specific peptide nucleic acid (PNA) probes (Applied Biosystems).
Total RNA was extracted with TRIzol (Invitrogen) or RNeasy mini kit (Qiagen) and was reverse transcribed with Superscript III cDNA synth (Invitrogen). Quantitative RT-PCR was performed using SYBR Green (Perkin Elmer) with the 7900HT Fast Real-time PCR system (Applied Biosystems). Primers used are listed in Extended Experimental Procedures.
Single cell suspensions from thymocytes, lymph nodes and spleen were isolated from 6–10 week old mice. Antibodies used for flow cytometric analysis are listed in Extended Experimental Procedures.
At least 10 million sorted DP (CD4+CD8+TCRlo/med), CD4+ SP (CD4+CD8-TCRhi) and CD8+ SP (CD4−CD8+TCRhi) populations was used to make cross-linked chromatin. Chromatin was processed for H3K4me3 or PAXIP1 ChIP-Seq with Illumina sequencing as described (Daniel et al., 2010).
Detailed description is available in the Extended Experimental Procedures.
We thank Gustavo Gutierrez-Cruz for technical assistance, Susan Sharrow, Larry Granger and Tony Adams for flow cytometry and Remy Bosselut for discussions. This work was supported by the Intramural Research Program of the NIH, National Cancer Institute, and Center for Cancer Research to A.N. and J-H.P, and a Senior Scholar award from the Ellison Medical Foundation to A.N. M.S.K. was supported by NIH grant R37 GM41052.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.