Flexible Stereospecific Interactions and Composition within Nucleoprotein Complexes Assembled on the TCRα Gene Enhancer¹

The generation of T lymphocytes requires a precise orchestration of TCR loci (Tcr) assembly in the context of a highly ordered program of cellular differentiation (1, 2, 3) (Fig. 1 A). The most immature thymocytes are known as double-negative (DN)⁸ thymocytes based in their no expression of CD4 and CD8. Four different DN subpopulations can be distinguished: DN1 (CD25⁻CD44⁺), DN2 (CD25⁺CD44⁺), DN3 (CD25⁺CD44⁻), and DN4 (CD25⁻CD44⁻), based on the expression of the surface markers CD25 and CD44 (4, 5). Tcrd becomes transcriptionally active at late DN1 thymocytes (6, 7). Initial completed rearrangements at Tcrg and Tcrd are detected in DN2 thymocytes, whereas extensive rearrangement at these loci and active rearrangements at Tcrb are detected in DN3 thymocytes (8, 9, 10, 11, 12). DN3 thymocytes that successfully rearrange their Tcrg and Tcrd express a γδ TCR and normally differentiate along the γδ T cell pathway via the DN4 stage (7, 13). This is a consequence of intracellular signaling directed by expression of a γδ TCR at DN3 cells, in a process known as γδ-selection (7). DN3 thymocytes that have not rearranged their Tcrg or Tcrd in frame, but have successfully rearranged Tcrb, normally differentiate along the αβ T cell pathway via the DN4 and double positive (DP) CD4⁺CD8⁺ stages, in a process known as β-selection (13). This is a consequence of intracellular signaling directed by expression of a pre-TCR and by Notch at DN3 thymocytes (14). Locus accessibility and induction of Tcra transcription and recombination depends on pre-TCR signaling at DN3 thymocytes, but not on Notch signaling (14). Pre-TCR signals are induced in two phases, one phase of rapid proliferation and a nonproliferative phase that it is compatible with activation of gene rearrangement at Tcra (15). At DN4 thymocytes rearrangements at Tcra are first detected, and occur extensively at the DP stage, coincidently with the second phase of pre-TCR signaling (4, 11, 15, 16, 17, 18).

The Tcrad locus has an unique genomic structure, which includes both Tcra and Tcrd, that determines Tcrd to be deleted from the chromosome upon Vα to Jα recombination at the DP stage (19). This process irreversibly commits DP thymocytes to the αβ-T cell lineage. Hence, developmental ordering of Tcrd and Tcra rearrangements is a critical component of αβ vs γδ T cell lineage commitment. The orchestration of the distinct developmental programs for V(D)J recombination and transcription of Tcra and Tcrd is controlled by the activity of two enhancers present at the Tcrad locus, the Tcra enhancer (Eα), and the Tcrd enhancer (Eδ) (18, 20, 21, 22, 23, 24). Eα is positioned downstream of Cα, while Eδ is situated upstream of Cδ, between the Vα and Jα clusters (25). Eα and Eδ work as a “developmental switch” with Eα “off” and Eδ “on” at the DN stages, and Eα “on” and Eδ “off” at the DP stage during thymocyte development (18, 25). These transcriptional enhancers are critical regulators of rearrangement and gene expression of Tcrad locus through promoters located at considerable distances by promoting developmental stage specific changes in chromatin structure (2, 22, 23, 26, 27, 28). Germline deletion of Eα results in severe reduction of germline Jα and Vα transcription and VαJα rearrangement at Tcra, but it does not alter Tcrd recombination and attenuates transcription of rearranged Tcrd (22, 27). In contrast, elimination of Eδ severely impairs recombination at Tcrd, without affecting transcription and recombination at Tcra (23). Locus control by Eα and Eδ cannot be explained simply by enhancer genomic location because replacement of Eα with Eδ in knock-in experiments failed to restore recombination (29). At present, the molecular mechanisms by which Eα and Eδ function are regulated during T cell development are unknown.

Eδ is regulated by critical functional collaboration between c-Myb and Runx factors bound mostly independently to a 30-bp element, denoted as δE3 (30, 31, 32, 33, 34). However, Eα activity is controlled by drastic functional synergy and cooperative binding among several transcription factors to a 116-bp enhancer fragment, denoted as Tα1-Tα2, creating a compact and functional nucleoprotein structure known as enhanceosome (35, 36, 37, 38, 39). Proteins that are thought form a functional Tα1-Tα2 enhanceosome include activation transcription factor (ATF)/cAMP response element binding protein (CREB), T cell factor-1 (TCF-1)/lymphocyte enhancer-binding factor-1 (LEF-1), Runx1, and Ets-1 (38, 39, 40, 41, 42, 43, 44, 45). The high level of structural organization among Tα1-Tα2 binding proteins is thought to ensure the high level functional cooperativity of the various enhanceosome components (36, 37, 38, 39). In vivo studies of Tα1-Tα2 occupancy revealed that TCF-1/LEF-1 factors must bind to DNA in conjunction with ets site binding factors for enhancer function, being both intact Tα2 TCF/LEF and ets sites required for enhanceosome assembly and function (36, 37). These results confirmed that Tα1-Tα2 constitutes a true enhanceosome because factor binding at Tα2 occurs in a strict cooperative fashion and no single Tα1-Tα2 binding factor can access chromatin to play a unique initiating role in its assembly (36). Hence, simultaneous availability of multiple enhancer binding proteins is required for chromatin disruption and stable binding site occupancy at Eα, as well as the activation of transcription and V(D)J recombination.

We have now created a mutant version of 1.4-kb Eα, EαMC, enhanceosome, in which Eδ myb and runx binding sites have been substituted for Tα2 runx and ets binding sites, which argues against the notion of a requirement for a very strict Eα enhanceosome structure for function. Transcriptional activity of EαMC was very potent and stronger to that of intact Eα in transient transfection experiments. These results indicate that the stereospecific interactions among proteins that form an Eα enhanceosome are rather flexible because we have created a new functional enhanceosome by substituting a critical site for other transcription factor binding site (ets site for myb site) and modifying the position of the runx binding site within the enhancer. Analysis of EαMC mouse thymocyte populations revealed a premature onset of minilocus V(D)J recombination during thymocyte development that was indistinguishable from that of Tα1-Tα2 (37, 46). Previous experiments demonstrated that a 275-bp enhancer fragment, including the Tα1-Tα4 elements, contains all the binding sites required for the correct developmental activation of Eα (46), indicating that protein binding to Tα3-Tα4 is critical to holding in check the potential for premature activation of V(D)J recombination by Tα1-Tα2. Our results show that Tα3-Tα4 cannot enforce a developmental delay to a chimeric “early” enhancer, Tα1-Tα2MC, indicating the need for a specific collaboration between Tα2 runx/ets sites binding proteins and proteins bound to Tα3-Tα4 for proper developmental activation of the enhancer in vivo.

ChIP was conducted with modifications of the protocol outlined in Upstate Biotechnology ChIP Assay Kit. Mouse total thymocytes were centrifuged, washed, and resuspended in cold PBS at 10⁷ cells/ml. Cells were cross-linked by adding formaldehyde to 1% (w/v) and incubated for 10 min at room temperature (RT). The reaction was stopped by adding glycine to 0.125 M and incubated for 5 min at RT. Cells were washed twice in PBS. Cells (5 × 10⁷) were resuspended in 500 μl of 1% SDS, 10 mM EDTA, 50 mM Tris-HCl (pH 8.1), and a mixture of protease inhibitors (Roche), and incubated for 10 min on ice. Cell suspension was sonicated using a Branson sonicator at 55% of duty cycle alternating 30 s on and 1–2 min off for 10 cycles while the sample was immersed in an ice/water bath. Chromosomal DNA was reduced to an average size of around 200–500 bp as determined by agarose gel analysis. Lysate was then centrifuged for 1 h at 13000 × g at 4°C, and supernatant diluted 10-fold by adding 4.5 ml of 0.01% SDS, 1.1% (v/v) Triton X-100, 1.2 mM EDTA, 16.7 mM Tris-HCl (pH 8.1), 167 mM NaCl and a mixture of protease inhibitors. One percent of starting material (50 μl) was saved as input for PCR detection. Chromatin was precleared by incubation for 2–3 h at 4°C with 20 μl of BSA 50 mg/ml, 75 μg of salmon sperm DNA, 25 μg of isotype-match control Ab, and 150 μl of 50% salmon sperm DNA/protein A-agarose slurry (Upstate Biotechnology). Lysate was centrifuged to eliminate beads with control Ab and unspecific DNA. Precleared chromatin corresponding to 4–5 × 10⁶ thymocytes (∼400–500 μl) was used for immunoprecipitation by incubating for 16 h at 4°C with 10 μg of specific or isotype-matched control Abs, followed by 70 μl of protein A-agarose slurry for an additional 1 h incubation at 4°C. Immunoprecipitates were washed by rocking for 5 min at RT once each with the following ice-cold buffers (containing protease inhibitors): 1) 0.01% SDS, 1.1% Triton X-100, 1.2 mM EDTA, 20 mM Tris-HCl (pH 8.0), 170 mM NaCl; 2) 0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris-HCl (pH 8.0), 500 mM NaCl; 3) 1% Nonidet P-40, 1% deoxycholic acid, 100 mM Tris-HCl (pH 8.0), 500 mM LiCl; and twice with 4) 10 mM Tris-HCl (pH 8.0), 1 mM EDTA. DNA/Protein/Ab complexes were eluted from protein A-agarose by rocking twice for 15 min at RT in 250 μl of 50 mM NaHCO3, 1% SDS. After reversal of cross-links and deproteination, purified DNA from input and bound fractions were resuspended in 50 μl of water. Presence of Eα and Oct2 sequences in the bound and input fractions was assessed by PCR using primers previously described (26). Radioactive PCRs were conducted by incorporation of [³²P]-dCTP to the PCR products, using as templates 2 μl of 1/10 diluted input DNA (equivalent to 0.004% of the taken input that corresponds to 0.0004% of the starting material used at each immunoprecipitation) or 2 μl of Ab-bound DNA (equivalent to 4% of total Ab-bound material). PCR conditions were: 5 min at 95°C, followed by 35 cycles of 25 s at 95°C, 25 s at 60°C, 30 s at 72°C, and a final extension step of 2 min at 72°C. PCR products were resolved by 8% PAGE.

Genomic footprinting was performed as previously described (18). RAG-2^−/−, RAG-2^−/− × TCRβ, HEB^−/−Id3^−/−, and HEB^−/−Id3^−/− mice were previously described (47, 48, 49, 50, 51, 52). RAG-2^−/− and RAG-2^−/− × TCRβ were obtained from Taconic Farms. Oligonucleotides for human 1.4-kb Eα and EαMC top strand analysis in transgenic thymocytes were CTT AATCAGGTGAGATCAAGGTCTGAT, TCTGATCTTGGCTGGTTTG GAGG, and GATCTTGGCTGGTTTGGAGGGGCAG. Those for human 1.4-kb Eα and EαMC bottom strand analysis were GCAATTAATAAACAGGTCAGACATTGAG, AAGTCTCCCACTTCCCTCCAGGTGT, and CTCCCACTTCCCTCCAGGTGTTTGGG.

EαMC minilocus: A cassette containing the myb and runx binding sites of δE3 were substituted for the runx and ets binding sites of Tα2. For this purpose, PCR overlap extension (35) was performed using as template a 1.4-kb Eα (KpnI-BamHI fragment) inserted in a version of pBluescript KS⁺ that had previously been modified to eliminate plasmid backbone BstXI sites. Mutagenic oligonucleotides CM-2A (TGTGGTTTCCAACCGTTACTGCGGGAGAGCTTCAAAGG) and CM-2B (TAACGGTTGGAAACCACAAGAAGAGTTTAAAATACTGA) were used along with the −40 and reverse primers in PCR. The final PCR product was digested with BstXI and the resulting EαMC BstXI fragment was inserted into BstXI digested 1.4-kb Eα plasmid from which the wild-type Eα BstXI fragment had been removed. The 1.4-kb EαMC was sequenced to confirm its structure, and was then excised from the plasmid with BamHI, blunted with Klenow, and ligated into XbaI-digested, blunted, and phosphatase-treated pBluescript carrying the enhancerless Tcrd minilocus (53). Minilocus structure was confirmed by dideoxynucleotide sequence analysis.

Tcrd minilocus DNA was purified as described previously (53), and was microinjected into fertilized C57BL/6 × SJL F2 eggs by the Duke University Comprehensive Cancer Center Transgenic Mouse Shared Resource. Progeny tail DNA samples were analyzed on Southern blots as previously described (53). Transgenes were maintained on a mixed C57BL/6 × SJL background. Copy number was determined by analysis of tail DNA on a slot blot (Schleicher and Schuell) using radiolabeled Cδ cDNA probe. Hybridization signals were quantified relative to previously identified single copy integrants using a PhosphorImager (Molecular Dynamics).

Plasmid Vδ1-CAT was described previously and versions of this plasmid carrying either the monomeric 35-bp δE3, 60-bp δE3-δE4, 34-bp Tα1, or 33-bp Tα2 enhancer fragments, or the intact 1.4-kb Eα, were described previously (30, 54, 55). To generate plasmids Tα1-Tα2-Vδ1-CAT and Tα1-Tα2MC-Vδ1-CAT, enhancer fragments were excised by BstXI and DraI digestion from the Eα and EαMC plasmids, and were blunted with T4-polymerase and cloned into XbaI-digested, Klenow-blunted, and phosphatase-treated Vδ1-CAT plasmid. To generate Vδ1-luciferase plasmid, 1.6-kb Vδ1 fragment was excised from Vδ1-CAT plasmid by SacI and HindIII digestion, and cloned into SacI and HindIII digested pXPG-firefly luciferase plasmid (56). To generate Eα and EαMC-Vδ1-luciferase plasmids, 1.4-kb Eα and 1.4-kb EαMC were excised from pBS BstXI plasmid by BamHI digestion, and cloned into BamHI digested and phosphatase-treated Vδ1-luciferase plasmid. The 5′ to 3′ enhancer orientation was confirmed by a PCR strategy.

The human leukemia T cell line Jurkat was cultured and transfected as described previously (31). For CAT assays, the acetylation of [¹⁴C]-chloramphenicol (Dupont-New England Nuclear was assayed as described previously (30) and quantified with a PhosphorImager. For luciferase assays, Jurkat cells (4 × 10⁶) were washed with PBS, resuspended in 300 μl of RPMI 1640, and transferred to 0.4-cm gap BTX cuvettes. Five micrograms of purified (with Qiagen Plasmid Maxi Kit) pXPG derivatives plus 10 ng of pRL-TK (renilla luciferase plasmid; Promega) were added, and cells were electroporated at 800 mF and 260 V in a BTX electroporator. Five hundred microliters of ice-cold inactivated FBS were added immediately to the cuvette and cells were kept for 15 min on ice. Cells were then diluted with 5 ml of RPMI 1640 containing 10% FCS (Life Technologies), and penicillin-streptomycin (Sigma-Aldrich), and incubated at 37°C with 5% CO2 for 24 h. Luciferase activity (firefly/renilla) was measured with a Dual Luciferase Reporter Kit (Promega).

Genomic DNA was prepared from thymi of 4- to 6-wk old mice by standard techniques. For single copy transgenic lines, 12 ng of genomic DNA was used as a template for PCR. For multicopy integrants, the quantity of DNA used was reduced to account for copy number to insure that the PCR amplifications were in the linear range (53). All PCR, gel electrophoresis, blotting, and hybridization with ³²P-labeled probes were performed as previously described (53). Hybridization signals were quantified using a PhosphorImager, and reported values for VDJ recombination were normalized to the Cδ signal for each PCR template.

For cell sorting, thymocyte populations were obtained from total mouse litters obtained from heterozygous crossings as described (18, 46). Preparation of cell templates for PCR analysis, conditions for PCR, and primers and probes used to analyze V(D)J recombination of both murine endogenous Tcra and the human Tcrd minilocus were previously described (18, 53). For quantitative real-time PCR analysis, same cell templates and primers were assessed with iQ SYBR Green Supermix (Bio-Rad) at Bio-Rad iCycler thermocycler.

Eα consists on four regions for protein binding, called Tα1 to Tα4 (35) (Fig. 1,B). Previous genomic footprinting analysis using dimethylsulfate at the endogenous murine Eα revealed that Tα1-Tα4 elements and flanking areas are extensively occupied both at DN3 and DP thymocytes from RAG2^−/− and RAG2^−/− × TCRβ mice, respectively, as well as at DP thymocytes from wild-type mice (18, 57) (Fig. 2). The minimal Eα, Tα1-Tα2, consists on a compact nucleoprotein structure or enhanceosome created by stereospecific interactions between transactivators bound to their cognate sites within the enhancer (36, 38, 39), that contains critical binding sites for ATF/CREB, TCF-1/LEF-1, Runx1, and Ets-1 proteins (38, 39, 40, 41, 42, 43, 44, 45). Eα enhanceosome structure extends the limits of Tα1-Tα2, because Tα1-Tα2 and Tα1-Tα4-bound factors interact physically to form a stable multiprotein complex in vitro, and to precisely regulate enhancer function in vivo (46). Outside of Tα1-Tα2, genomic footprinting experiments revealed occupancy of an AP1/GATA site, an ets site, an E-box (E-box-I) and a GC-box (GC-I), which can bind Sp1 in vitro, upstream of Tα1, a site for GATA-3 and an E-box (E-box-II) in Tα3, an Sp1 site between Tα3 and Tα4, a runx site and another E-box (E-box-III) in Tα4, and two CACC-boxes and a GC-box (GC-III) downstream of Tα4 (18, 36, 57). Previous in vitro experiments with specific Abs revealed presence of the Tα1-, Tα2-, and Tα3-binding factors TCF-1/LEF-1, Ets-1, Fli-1, GATA-3, Sp1, and CREB-1 on Tα1-Tα4 enhanceosome formed in DP cells (46).

In an attempt to characterize the identity of the proteins that are present at the endogenous Eα enhanceosome in vivo, we have now combined ChIP and genomic footprinting analyses in mouse thymocytes (Fig. 2). Consistent with previous ChIP experiments, our analysis confirmed in vivo binding of ATF-2, CREB-1, GATA-3, and Ets-1 to Eα (45, 58, 59) (Fig. 2, A and B). In addition, we have observed in vivo binding of Fli-1, Sp1, HEB, and E2A to Eα sequences (Fig. 2 C). These experiments indicate that Eα enhanceosome is formed by the coordinated assembly of multiple transcription factors including CREB-1 presumably bound to Tα1 CRE-site (39, 46, 57); ATF-2 that could bind as an AP-1 complex with c-Jun to the 5′Tα1 GATA/AP1 site (60); Fli-1 or Ets-1 that could bind indistinguishably to an isolated ets site at 5′Tα1 (45); Runx-1 and Ets-1 bound cooperatively to Tα2 runx and ets-binding sites (45); and HEB and E2A likely bound to the three E-boxes present within the enhancer (61).

Detection of E2A and HEB proteins as part of Eα enhanceosome (Fig. 2,C) is in agreement with the reported ability of E2A proteins to bind in vitro to Tα3 E-box-II and Tα4 E-box-III (61). HEB/E2A heterodimers have been shown to be the major form of these proteins present in T cells, whereas E2A/E2A homodimers is the form present in B-cells (62, 63). To map the specific E-boxes occupied in vivo by HEB at Eα, we have analyzed in vivo occupancy of Eα E-boxes in thymocytes from HEB^−/− mice (47). We analyzed the effect of deletion of HEB in absence of Id3 (in the context of Id3^−/− mice), because it has been reported that disruption of Id3 increase the viability of HEB^−/− mice without affecting their thymocyte maturation phenotype (50). In agreement with a no obvious developmental defect of Id3^−/− thymocytes (49), we did not see any difference between analysis of in vivo Eα occupancy in Id3^+/+ and Id3^−/− mice (Fig. 2,D). Comparing the analysis obtained in HEB^+/+Id3^−/− and HEB^−/−Id3^−/− mice, we found that Tα4 E-box-III footprints were diminished in HEB^−/−Id3^−/− thymocytes (Fig. 2, D and E). This result suggests specific binding of HEB to Tα4 E-box-III, but not to the 5′Tα1 E-box-I and the Tα3 E-box-II, which are likely occupied by the E2A homodimers in vivo. Our results do not distinguish whether HEB binds to Tα4 E-box-III as a heterodimer with E2A proteins or as a HEB homodimer. Our data supports the possibility that different E-boxes might be occupied by different combinations of E-proteins and presents a possible scenario to explain the unique roles provided by E2A and HEB proteins in early thymopoiesis (47, 50, 63, 64, 65, 66). It is also interesting to note that the pattern of footprints found at the Tα4 E-box-III observed in HEB^−/− thymocytes resembles to that found in thymocytes from RAG2^−/− mice, which are arrested in a DN3 stage, but not in thymocytes from RAG2^−/− × TCRβ mice, which develop to a DP stage (18, 51, 52). These data suggest that binding of HEB to Tα4 E-box-III might be involved in Eα activation during DN3 to DP transition.

Eα function in activation of both transcription and V(D)J recombination during thymocyte development relies on its ability to recruit histone modifying activities through specific interactions with enhancer-bound transcription factors (26). In this context, Eα functions by promoting developmental stage-specific changes in histone acetylation at long distances through entire Tcra, including both Jα and Vα gene segment clusters (26, 27). CREB-binding protein (CBP) and p300 are histone acetyltransferases (HATs) that contain different domains for interaction with transcription factors, and act as integrators for the assembly of other well studied enhanceosomes, as it is the case for IFN-β and TNF-α enhanceosomes (67, 68, 69). To analyze the presence of CBP and p300, as well as the histone deacetylase (HDAC) 1 and its corepressor mSin3 at the Eα enhanceosome in vivo, we performed ChIP experiments in total mouse thymocytes (Fig. 2 F). In agreement with an important role for Eα-dependent recruitment of HATs for enhancer function at the endogenous locus in DP cells (26), we found that CBP and p300 were present at Eα enhanceosome in total mouse thymocytes, whereas the corepressor mSin3 and its associated HDAC1 were absent. These experiments are consistent with long-range histone acetylation directed by Eα function in DP thymocytes (26), and suggest that CBP and p300 recruitment through an active Eα enhanceosome is involved in enhancer function.

Tα1-Tα2 has been considered a paradigm for enhanceosome structure (38, 39). Previous in vivo studies of Tα1-Tα2 occupancy revealed that none of enhancer binding factors can bind to enhancer DNA at physiological concentrations, and that factor binding at Tα2, including TCF-1/LEF-1 factors, Runx1 and Ets-1, occurs in a highly cooperative all-or-none fashion (36, 45). These results explained why intact Tα2 TCF/LEF and ets binding sites are both essential for enhancer activity (36, 37). These results indicated that Eα has an inherent great stability due to strong cooperative binding among their binding-proteins. This high level of structural organization would ensure a high level functional cooperativity of the various enhanceosome components. We have created a mutant version of Eα enhanceosome that argues against the notion of a requirement for a very strict Eα enhanceosome structure for function (Fig. 3,A). This mutant Eα does not contain the essential Tα2 ets binding site (36, 37, 40), and consists on a chimeric Eδ/Eα, EαMC, in which minimal Eδ, δE3, myb, and runx binding sites were substituted for the Tα2 runx and ets binding sites in the context of the entire 1.4-Eα (Fig. 3Α). Myb and runx binding sites are essential for the activation of both transcription and V(D)J recombination by Eδ (30, 33, 34, 55). Preliminary characterization of the Tα1-Tα2 fragment of the chimeric enhancer, Tα1-Tα2MC, confirmed efficient in vitro binding of c-Myb and Runx1 to it (data not shown). We next analyzed the ability of Tα1-Tα2MC to activate transcription from the Vδ1 promoter in transient transfected Jurkat cells, as compared with a series of wild-type fragments of Eδ and Eα (Fig. 3,B). In these experiments, enhancer fragments were subcloned upstream of the Vδ1 promoter in the enhancer-dependent test construct Vδ1-CAT, and enhancer activity was measured as previously reported (30, 54, 55). As shown previously (30, 54, 55), both δE3 and δE3-δE4 efficiently activated transcription from the Vδ1 promoter (12.9 and 25.7-fold induction, respectively). As expected (70), a single copy of Tα1-Tα2 was a potent transcriptional enhancer (92.4-fold induction), whereas Tα1 alone and Tα2 alone were ineffective (1.4- and 0.5-fold induction, respectively). Strikingly, Tα1-Tα2MC displayed an activity that was greater than Tα1-Tα2 and comparable to that of the entire 1.4-kb Eα (169.3 and 132.5-fold induction for Tα1Tα2MC and Eα, respectively). Hence, the chimeric Tα1-Tα2MC enhancer is an extremely potent transcriptional activator as judged by transient transfection. The strong transcriptional activity of Tα1-Tα2MC cannot be explained by the weak activity of δE3 itself, but by a dramatic functional synergism among the proteins bound to the chimeric enhancer. Tα1-Tα2MC activity has to be a consequence from strong functional cooperativity between Tα1 and Tα2MC binding proteins. In agreement with the stronger transcriptional enhancer activity of Tα1-Tα2MC compared with that of Tα1-Tα2, 1.4-kb EαMC activity was 2.6-fold more potent than that of wild-type 1.4-kb Eα, suggesting the existence of an additional level of synergism among proteins bound inside and outside Tα1-Tα2MC at EαMC (Fig. 3 C). Hence, in contrast with previous data that supported a model for very specific stereospecific interactions between the different proteins bound to Eα to ensure a high level of cooperativity among them, our data demonstrate that the stereospecific interactions among the different Eα enhanceosome components could be rather flexible.

Previous experiments studies using a reporter construct consisting on an unrearranged human Tcrd minilocus in transgenic mice have demonstrated enhancer dependency of developmental regulation of V(D)J recombination at the Tcrad locus (18, 21, 37, 46). The Tcrd minilocus used in these experiments contains Vδ1, Vδ2, Dδ3, Jδ1, and Jδ3 gene segments, Cδ, and an enhancer within the Jδ3-Cδ intron (53) (Fig. 4,A). V gene segments within the construct carry a mutation to avoid expression of a transgene-derived TCRδ-chain that could interfere with normal T cell development (53). Analysis of rearrangements at the transgenic minilocus driven by 1.4-kb Eδ, 1.4-kb Eα, or no enhancer revealed that the Vδ to Dδ rearrangement step is enhancer-independent, whereas the VδDδ to Jδ rearrangement step is enhancer-dependent (53). To analyze the activation of V(D)J recombination by 1.4-kb EαMC, we generated new lines of transgenic mice, JS and JU, carrying a minilocus containing the chimeric enhancer (Figs. 3,A and 4,A). JS and JU carry 8–10 and 25–30 copies of the EαMC minilocus, respectively. Analysis of minilocus V(D)J recombination was performed by PCR from thymus genomic DNA templates using Vδ1 or Vδ2 primers in conjunction with Jδ1 primers as described previously (53) (Fig. 4,B). This PCR strategy amplifies 0.3-kb fragments corresponding to complete VδDδJδ rearrangements and 1.2-kb fragments corresponding to VδDδ rearrangements. PCR were also performed in parallel with a pair of Cδ primers as an internal control. All PCR experiments were performed under conditions previously established to yield linear amplification (18, 21, 33, 34, 37, 46, 53). PCR products were electrophoresed through agarose gels, blotted, and detected by hybridization with appropriate ³²P-labeled Vδ1, Vδ2, or Cδ probes. We detected high levels of Vδ1Dδ3Jδ1 and Vδ2-Dδ3-Jδ1 rearrangements (similar to 1.4-kb Eα line J) in each line (Fig. 5). Hence, the chimeric enhancer is an extremely efficient activator of enhancer-dependent V(D)J recombination at judged by analysis of our recombination reporter construct in transgenic mouse thymocytes.

Previous experiments using a transgenic reporter substrate revealed that 116-bp Tα1-Tα2 led to premature activation of V(D)J recombination compared with that observed for the entire 1.4-kb Eα, whereas 275-bp Tα1-Tα4, led to a correct developmental activation (37, 46). These experiments indicated that binding sites within Tα3-Tα4 collaborate with factors bound to Tα1-Tα2 for the strict developmental regulation of Tcra rearrangement. To better understand the mechanisms responsible for appropriate developmental regulation and dramatic functional synergy between transcription factors that bind to Eα, and specifically address whether Eα regions outside of Tα1-Tα2 collaborate functionally with Tα2 runx/ets-bound proteins, we evaluated developmental control by EαMC, and compared it with that regulated by Eδ, Eα, Tα1-Tα2, and Tα1-Tα4 through analysis of activation of Vδ1Dδ3 to Jδ1 minilocus rearrangement in sorted thymocyte populations from 4 to 6 wk old transgenic mice (Fig. 6 and supplemental Table I)⁹ (18, 46).

To evaluate these rearrangements, we used a PCR strategy that allows a semiquantitative analysis of these rearrangements: 2 μl of each cell lysate, corresponding to 2,000 cells, were analyzed with 25 cycles of amplification and Southern blots (Fig. 6). To validate this PCR strategy for quantification of Vδ1Dδ3Jδ1 rearrangements, we have analyzed Cδ and Vδ1Dδ3Jδ1 amplicons using serial dilutions of DP cell templates from the transgenic line T2, which is one of the transgenic lines with the maximum number of copies analyzed, 30 copies, and adjusted the amplified signals for transgene copy number (Fig. 7). Perfect linearity was observed at these PCR analyses indicating that experimental conditions used in experiments shown in Fig. 6 allows for correct quantification of transgenic rearrangements and comparison among different cell templates. Estimated amount of transgenic DNA analyzed in our experiments is within the range of linearity previously established for these type of PCR experiments (53, 71). Accordingly, analysis of 2 μl of undiluted samples of DP templates from other transgenic lines correlated perfectly with their copy number (Fig. 7), indicating that results obtained from this type of analysis at all the transgenic cell templates can be compared and correctly interpreted.

In agreement with previous analysis of the Eδ transgenic line A (18, 21, 46), we detected high levels of Vδ1Dδ3Jδ1 rearrangement in both αβ and γδ T cells, and relative to rearrangement in αβ T cells (arbitrarily set to 100%), rearrangement was essentially complete in DN3 thymocytes. In agreement also with the previously described critical role for Eα in the developmental activation of Tcra rearrangement (18, 22), the pattern of minilocus rearrangement in Eα transgenic lines (as it is the case for the Eα transgenic line L) paralleled that of endogenous Tcra with no rearrangements observed in γδ T cells and not being completed until de DP stage and subsequent αβ T cells (18, 21, 46). These results are consistent with previous analyses of endogenous murine Tcrd and Tcra rearrangements (8, 9). Hence, this minilocus reporter construct provides a useful approach to evaluate how changes in enhancer structure impact its ability to developmentally regulate V(D)J recombination in vivo. Consistent also with previous experiments (37, 46), Tα1-Tα2 directed a premature activation of enhancer-dependent VδDδ to Jδ rearrangement, on the basis of elevated rearrangement in DN3 thymocytes, γδ T cells, and DN4 thymocytes, as it is seen in the analysis of the Tα1-Tα2 transgenic line T2 (26, 19, and 36%, respectively) (Fig. 6 and supplemental Table I), whereas Tα1-Tα4 enhancer fragment is sufficient for proper developmental regulation of Eα minilocus since its developmental activation of V(D)J recombination driven was essentially indistinguishable from that driven by the entire 1.4-kb Eα, as it is illustrated in the analysis of the Tα1-Tα4 transgenic line C3: Vδ1Dδ3Jδ1 rearrangement was detected at relatively low levels in DN4 cells and at high levels in DP and αβ T cells, but was only barely detected in DN3 and γδ T cells. All these data together indicated that Eα sequences in the Tα3-Tα4 region function to prevent the premature activation of V(D)J recombination that would otherwise be directed by Tα1-Tα2. As an internal control for cell purity, we always examined endogenous murine Vα to Jα rearrangements in the same-sorted cells (Fig. 6 and supplemental Table I). As expected, endogenous VαJα rearrangement was undetectable in DN1, DN2, and DN3 thymocytes and γδ T cells; was detected at low levels in DN4 thymocytes; and was completed in both DP thymocytes and αβ T cells.

In the analysis of EαMC lines (Fig. 6 and supplemental Table I), we observed extensive Vδ1Dδ3Jδ1 rearrangements as early as DN3 (39% for JS and 25% for JU). High level of rearrangements were also detected in γδ T cells (37% for JS and 19% for JU) and DN4 (50% for JS and 25% for JU). In fact, developmental activation of V(D)J recombination by EαMC is indistinguishable from that driven by Tα1-Tα2 (46). In agreement with a correct evaluation and quantification of data, quantitative real-time PCR analyses of Vδ1Dδ3Jδ1 rearrangements in transgenic sorted cells correlate perfectly with data shown in Fig. 6 (Fig. 8). Vδ1Dδ3Jδ1 rearrangements in JS, T2, and M lines were 23, 27, and 0.41%, respectively, in DN3 thymocytes; 40, 39, and 4%, respectively, in DN4 thymocytes; and 23, 29, and 0.6% in γδ-T cells.

These results indicate that proteins bound to the Tα3-Tα4 region cannot delay the activation of a chimeric “early” enhancer, but rather, specifically collaborate with Tα2 runx/ets sites binding proteins to effect appropriate developmental control. These experiments demonstrate that although stereospecific interactions among proteins that form a Tα1-Tα2 enhanceosome could be rather flexible, structural constraints must exist for functional collaboration between Tα1-Tα2 and Tα3-Tα4 binding proteins. Hence, specific interactions between Tα1-Tα2 and Tα3-Tα4-bound factors are essential to developmentally control of enhancer activation.

Tα1-Tα2 consists on a compact nucleoprotein structure created by stereospecific interactions between transactivators bound to their cognate sites within the enhancer, and it is considered as a paradigm for enhanceosomes. From the general definition of enhanceosomes, both δE3 and Tα1-Tα2 can be considered as true enhanceosomes because they contain very tightly clustered sites for transcription factors in a compact genomic sequence (30 bp for δE3 and 116 bp for Tα1-Tα2) that operate as a scaffold to assemble a unified multiprotein complex to form a higher-order three-dimensional transcription factor/enhancer DNA complex (30, 31, 33, 34, 36, 37, 38, 39, 45, 72, 73), as it is the case for other well characterized enhanceosomes, such as the IFN-β enhancer and the TNF-α promoter (74, 75, 76). All these enhanceosomes show a high degree of cooperativity in their assembly and a very robust functional synergy among their components, because alterations in individual binding sites have drastic effects on enhancer function, as it is also the case for Eα and Eδ (30, 31, 32, 33, 34, 36, 37, 38, 77). The clustered sites that define δE3 and Tα1-Tα2 reflect the synergy required for weak, but critical, interactions among the proteins bound to these enhanceosomes. In this way, the function of these enhanceosomes is much more than the sum of individual factor contributions, but it becomes from a network of stereospecific interactions among enhancer-bound proteins to ensure that a specific gene would be activated only if all the enhanceosome components are all simultaneously present in the same nucleus (36). This high level of structural organization ensures a higher level of functional cooperativity among the various enhanceosome components.

In general, architectural proteins are believed to orchestrate the formation of enhanceosomes by facilitating interactions between distantly bound factors, as it is the case for TCF-1/LEF-1 factors at Tα1-Tα2 (38, 39) or for high mobility group-I(Y) at the IFN-β enhancer (78). TCF-1/LEF-1-induced DNA bending was thought to play an organizing role that promotes specific interactions between ATF/CREB factors and Ets-1 bound to distal ends of the enhancer (38), and that helped to recruit non-DNA binding proteins that provide additional bridges among the various DNA-bound factors (77, 79). However, TCF-1/LEF-1 factors have very low binding specificity to their specific sequences (80), and in vivo studies of Tα1-Tα2 occupancy revealed that, under physiological conditions, none of the Eα activators, including TCF-1/LEF-1 factors, can bind Eα DNA in their own and access chromatin to play a unique initiating role in enhanceosome assembly (36). These experiments indicated that TCF-1/LEF-1 factors must bind to this enhancer in vivo in conjunction with Ets-1 (and presumably Runx1, which binds in a highly cooperative fashion with Ets-1 in vitro) in an all or none fashion (36, 38, 45, 81) that explained why intact Tα2 TCF/LEF and ets sites are both essential for in vivo enhancer activity (36, 37). Hence, simultaneous availability of multiple enhancer binding proteins is required for chromatin disruption and stable binding site occupancy at Eα, as well as the activation of transcription and V(D)J recombination.

Our present ChIP and genomic footprinting experiments have identified many of the proteins that are part of the endogenous Eα enhanceosome in vivo in mouse thymocytes (Fig. 2). Our experiments confirmed the presence of ATF-2, CREB-1, GATA-3, and Ets-1 (45, 58, 59), and demonstrated the presence of Fli-1, Sp1, HEB, and E2A at Eα sequences in vivo. Furthermore, our data identify the Tα4 E-box-III as a possibly specific in vivo binding site for HEB at this enhancer. In addition to these DNA-binding factors, our ChIP experiments have revealed the presence of the HATs CBP and p300, and the absence of the corepressor mSin3 and its associated HDAC1 at Eα in mouse thymocytes. These experiments suggest that CBP and p300 are recruited to create an active Eα enhanceosome through the coordinated assembly of multiple transcription factors. These factors include CREB-1, presumably bound to Tα1 CRE site (39, 46, 58); ATF-2 that could bind as an AP-1 complex with c-Jun to the 5′Tα1 GATA/AP1 site (58, 60); Fli-1 or Ets-1 that could bind indistinguishably to an isolated ets site at 5′Tα1 (45); Runx-1 and Ets-1 bound cooperatively to Tα2 runx and ets-binding sites (38, 45, 81); HEB that could bind specifically to the Tα4 E-box-III; and E2A (likely bound to the three E-boxes present within the enhancer) (61) are involved in enhancer function. These data are consistent with long-range histone acetylation directed by Eα function in DP thymocytes (26).

Our mutant version of Eα enhanceosome has argued against the notion of a requirement for a strict Eα enhanceosome structure for function (EαMC). This mutant Eα does not contain the essential Tα2-ets binding site (37, 40), and consists on the substitution δE3 myb and runx binding sites for Tα2 runx and ets binding sites in the context of the entire 1.4-Eα. Our initial goal in designing the EαMC chimera was to substitute a myb site in place of the ets site in Tα2. However, a direct replacement would have oriented the myb and runx sites in a manner distinct from that found in Eδ. Because previous studies revealed that a precise alignment of these binding sites is required for functional synergism between c-Myb and Runx factors in Eδ (30), we deliberately conserved the Eδ arrangement of myb and runx sites to insure minimal Eδ function within the chimera. The chimeric enhancer retains two runx sites in the Tα2 region that are now separated by the myb site rather than being adjacent to one another. It is important to note that although δE3 myb and runx binding sites are essential (30, 33, 34, 55), they are not sufficient for the activation of transcription and V(D)J recombination in the context of the Tcrd minilocus (34), hence these sites cannot function as a dominant element on its own. Although Runx factors are context-dependent transcription factors that can functionally collaborate with Ets-1 and with c-Myb (30, 38, 45, 81), the requirement for formation of the compact nucleoprotein structure that it is created at Tα1-Tα2 by stereospecific interactions between transactivators bound to their cognate sites within the enhancer makes rather surprising that we could replace the Tα2 runx and ets sites with minimal Eδ myb and runx sites to generate a chimeric, reorganized version of Tα1-Tα2 that is more potent in transient transfection experiments than either of the parent enhancers (Tα1-Tα2 and δE3). Furthermore, an intact ets site was demonstrated to be essential for the assembly and function of the Tα1-Tα2 enhanceosome in vivo (36, 37). Despite a wealth of data describing the stereospecific assembly of the Tα1-Tα2 enhanceosome (38, 39), our data indicate that the structural constraints on enhanceosome assembly and function are rather relaxed. Hence, there appears to be substantial flexibility regarding the specific components that can be effectively assembled into a functional enhanceosome.

Transcriptional enhancers are critical regulators of rearrangement and gene expression of Ag receptor genes through promoters located at considerable distances (2). Precise developmental activation of these enhancers must therefore be orchestrated by the specific array of transcription factors and coactivators recruited to the region. In this way, the Eα is supposed to expose a unique activating surface, which can initiate very precisely and efficiently a developmental program of events that counteracts the repressive chromatin environment, leading to the downloading of the basal transcriptional machinery to its specific promoters and activation of V(D)J recombination. We previously found that Tα1-Tα2 is subjected to premature activation in vivo, and Tα3-Tα4 functions to enforce a developmentally appropriate activation to Tα1-Tα2 (37, 46). Our present data indicate that the functional interaction between these two regions appears to be highly specific and dependent on Tα2 runx/ets binding sites, because Tα3-Tα4 could not enforce a similar developmental delay on an active chimeric “early” enhancer containing Eδ myb/runx binding sites. Deregulation in the onset of V(D)J recombination directed by the chimeric enhancer might depend on the loss of Tα2 ets site occupancy, the different reorganization of the runx sites, or both. At present, the identity of the Tα3-Tα4 factors that could collaborate with factors bound to Tα2 runx/ets binding sites in the developmental activation of Eα is unknown. Because δE3 myb/runx sites cannot functionally interact with the Tα3-Tα4 elements to inhibit rearrangement in DN3 thymocytes, our data suggest that Tα2-runx/ets site binding factors are involved in specific functional interactions with Tα3-Tα4 binding proteins to maintain a repressed state until the DN4 stage of thymocyte development. The fact that δE3-δE4 elements are unable to activate recombination at this system also support our hypothesis, because δE3 cannot act as a dominant “early” element by itself (34). In conclusion, our data strongly suggest a requirement for functional interactions between Tα2-runx and/or ets sites binding factors with Tα3-Tα4 binding proteins to ensure onset of rearrangement at the DN4/DP stages. In an attempt to better understand how functional interactions between Tα1-Tα2 and Tα3-Tα4 might affect enhancer function, we have conducted genomic footprinting analyses comparing Eα and EαMC thymocyte samples from M, JS, and JU transgenic lines (data not shown). These experiments indicate that factor binding to Tα2MC does not affect occupancy of Tα1 and Tα3-Tα4 binding sites and surrounding areas, suggesting that functional collaboration between Tα2 runx/ets binding sites and Tα3-Tα4 bound factors occurs at a level of regulation beyond accessibility of other transcription factors to their binding sites within chromatin or cooperative factor loading to enhancer DNA.

Gene targeting experiments in mice have established that Ets-1 is important for development of the αβ T cell lineage by regulating functional integrity of the pre-TCR signaling and allelic exclusion at Tcrb (82, 83, 84), and for development of NK cells (85, 86). However, no effect on expression at Tcra has been detected in T cells from Ets-1-deficient mice. Although in vivo Ets-1 binding at Tα2 has been detected in Jurkat cells (45), it is possible that other factors different from Ets-1 could occupy the Tα2 ets site in DN3 thymocytes or during β-selection, and be responsible for the correct activation of Eα in the transition of DN3 to DP. This latter scenario implies that Eα can assemble different enhanceosomes during T cell differentiation. This flexibility in the assembly of different types of nucleoprotein complexes among Eα-binding proteins could explain in part our results of creating a very potent EαMC enhanceosome with different components assembled in the complex. Previous in vivo footprinting experiments indicated that occupancy of Eα is indistinguishable in DN3 and DP thymocytes suggesting that the identity of the majority of proteins must be identical or very closely related in the enhanceosomes assembled in both cases (18, 57). However the high flexibility that we observe in the assembly of functional multiprotein complexes at this sequence might be indicative of a possible assembly of distinct complexes at different moments of development. The flexibility that permits the arrival and departure of some transcription factors at a given time has been observed in the assembly of TNF-α enhanceosome (75, 76). This enhancer responds to a large variety of signals such as Ag receptor engagement, virus, LPS, and ionophorus and is equipped with a collection of binding sites that are differentially used depending on the signal. For example, this enhanceosome contains NFAT and ATF-2/c-Jun after Ag receptor engagement or ionophorus stimulation; NFAT, ATF-2/c-Jun, and Sp1 after virus infection; and Sp1, Egr-1, Ets/Elk, and ATF-2/c-Jun after LPS stimulation. Similarly to this, the possibility of assembly of distinct sets of proteins on Eα might represent a more flexible form of information processing during thymocyte development. Characterization of the nature of the different enhanceosome complexes assembled on Eα during DN3 to DP transition is a goal for our future investigations.

We thank Michael S. Krangel for support and helpful contributions during the course of this work, Cheryl Bock for generation of transgenic mice at the Duke University Shared Transgenic Mouse Facility, Dr. Yuan Zhuang at Duke University Medical Center for providing HEB^+/+Id3^−/− and HEB^−/−Id3^−/− mice, Hubertus Kohler and Tracy Hayden for their help with sorting experiments, Allison Dwileski for help with statistics and graphical representation of data, and Michael S. Krangel and Carles Suñé for critical reading of the manuscript.

The authors have no financial conflict of interest.