The T-box Repressors TBX2 and TBX3 Specifically Regulate the Tumor Suppressor Gene p14 ARF via a Variant T-site in the Initiator*

The murine tumor suppressor p19 ARF ( p14 ARF in humans) is thought to fulfill an important protective role in preventing primary cells from oncogenic transformation via its action in the p53 pathway. Several disease-implicated regulators of p19 ARF are known to date, among which are the T-box genes TBX2 , which resides on an amplicon in primary breast tumors, and TBX3 , which is mutated in the human developmental disorder Ulnar-Mammary syndrome. Here we identify a variant T-site, matching 13 of 20 nucleotides of a consensus T-site, as the essential TBX2/TBX3-binding element in the human p14 ARF promoter. Mutant analysis indicates that both the consensus T-box and a C-terminal conserved repression domain are essential for p14 ARF repression. Whereas the core nucleotides required for interaction of the archetypal T-box protein Brachyury with a consensus T-site are conserved in the variant site, additional flanking nucleotides contribute to the specificity of TBX2 binding. This is illustrated by the inability of TBX1A or Xbra to activate via the variant p14 ARF T-site. Importantly, this suggests a hitherto unsuspected level of specificity associated with T-box factors and corre-sponding recognition sites in regulating their target genes in vivo .

The murine tumor suppressor p19 ARF (p14 ARF in humans) is thought to fulfill an important protective role in preventing primary cells from oncogenic transformation via its action in the p53 pathway. Several diseaseimplicated regulators of p19 ARF are known to date, among which are the T-box genes TBX2, which resides on an amplicon in primary breast tumors, and TBX3, which is mutated in the human developmental disorder Ulnar-Mammary syndrome. Here we identify a variant T-site, matching 13 of 20 nucleotides of a consensus Tsite, as the essential TBX2/TBX3-binding element in the human p14 ARF promoter. Mutant analysis indicates that both the consensus T-box and a C-terminal conserved repression domain are essential for p14 ARF repression. Whereas the core nucleotides required for interaction of the archetypal T-box protein Brachyury with a consensus T-site are conserved in the variant site, additional flanking nucleotides contribute to the specificity of TBX2 binding. This is illustrated by the inability of TBX1A or Xbra to activate via the variant p14 ARF T-site. Importantly, this suggests a hitherto unsuspected level of specificity associated with T-box factors and corresponding recognition sites in regulating their target genes in vivo.
T-box genes are a relatively recently discovered gene family characterized by a 180 -200-amino acid DNA-binding T-box domain (reviewed in Refs. [1][2][3]. Genetic studies in species ranging from Xenopus to human point to a crucial role for T-box genes in controlling development in a gene dose-dependent manner. This is clearly illustrated by two human syndromes that lead to multiple developmental abnormalities and that are characterized by both patterning defects and hypoplasia: Holt-Oram syndrome and Ulnar-Mammary syndrome. These syndromes are associated with haplo-insufficiency of TBX5 and TBX3, respectively (4,5). Although data on the functions of the steadily growing family of T-box genes are now emerging, relatively little is known about T-box gene targets and the molecular mechanism underlying specific gene regulation by individual T-box genes. In vitro site selection experiments with Brachyury, the founder of the T-box gene family, revealed the preference for a palindromic sequence (6) and the crystal structure showed that the Brachyury T-box binds this so-called T-site as a dimer (7). Of all T-box genes only three, TBX2, TBX3 and Xenopus ET, are currently known to be transcriptional repressors, probably acting via a putative repression domain (8,9). TBX2 was shown to act on the melanocyte-specific TRP-1 promoter, not by a regular T-site but by two T-half-sites located 200 bp apart (8). Also for other in vivo T-box targets a picture is now emerging in which they are controlled by separate T-half-sites.
We recently found TBX2 in a genetic senescence bypass screen to potently down-regulate the p19 ARF tumor suppressor, thereby causing efficient immortalization of primary fibroblasts (10). Additionally, in an independent screen, TBX3 was found to have that same activity (11). This placed T-box gene family members for the first time in the regulation of cell proliferation. P19 ARF in mice or p14 ARF in humans is the alternative transcript encoded by the unusual INK4a/ARF locus coding also for p16 INK4a (reviewed in Ref. 12). Both of these cell cycle inhibitors and tumor suppressors are implicated in cancer-relevant pathways; p16 INK4a acts to inhibit CDK4 and CDK6, thereby preventing inactivating phosphorylation on the Rb tumor suppressor protein (13,14), whereas p19 ARF acts in the p53 pathway (reviewed in Refs. 15 and 16). Induction of p19 ARF upon serial passaging of rodent cells leads via MDM2 to a stable and transcriptionally active p53, which by activating its targets can lead to a growth arrest called replicative senescence. P19 ARF is also induced upon hyperproliferative signaling by oncogenes such as c-myc and adenoviruses E1A, E2F-1, and Ras V12 (17)(18)(19)(20). Therefore, p19 ARF activation is thought to work as an important fail-safe mechanism, because efficient oncogenic transformation by these factors can occur only when the ARF-Mdm2-p53 pathway is inactivated (15,16).
Several regulators of the p19 ARF promoter are known to date. Of these a number are most likely indirect regulators, such as c-Myc (17), Twist (21), JunD (22), DAP kinase (23), c-Abl (24), E1A (18), and Ras V12 (20). However, some of them potentially are direct regulators, for example DMP-1 (25,26), BMI-1 (27), p53 (28), and E2F1 (19). In accordance with the important cancer-preventing role of p19 ARF , deregulation of many of these p19 ARF transcriptional regulators has been shown to play a role in tumorigenesis. For instance overexpression of Bmi-1, a member of the Polycomb-group of repressor proteins, results in lymphomagenesis and likely contributes to several human malignancies (29,30). For TBX2 a potential role in breast cancer development has been suggested because the locus resides on an amplicon present in a subset of primary breast cancers (10). Here we have studied the mechanism and cis-requirements for repression of p14 ARF /p19 ARF as a bona fide TBX2 and TBX3 target and found that TBX2 and TBX3 are direct and specific regulators acting via a variant T-site that is present close to the p14 ARF transcriptional start site. * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ Supported by a Pioneer grant from the Dutch Organization for Scientific Research.
§ Supported by grants from the Dutch Cancer Society.
Repression Assays-COS-7 cells were co-transfected at 40% confluency by calcium phosphate precipitation with 20 g of reporter plasmid and 0 -1 g of expression plasmid. The TBX2 constructs, TBX2, TBX2RD, TBX2TB(R122E, R123E), and TBX2TB(A272E), were expressed from pCDNA3.1 (Invitrogen), TBX1A-HA from pCMV, Xbra from pEGFP-C1 (32), and E2F1 from pCMV. For the p14 ARF promoter, the original CAT reporter constructs described by Robertson and Jones (28) were used. In addition we used the pGL3-basic vector (Promega) in which we recloned a HindIII/SalI fragment containing the wild-type (Ϫ19/ϩ54) p14 ARF promoter insert. To circumvent problems with low reporter activity, p14 ARF reporters with double (Ϫ19/ϩ54) inserts were used because these constructs are more active than those with single inserts (data not shown). Mutants of the T-site were generated by PCR and sequence-verified. The HSV tk reporter was constructed as described (32). Into the SmaI site of this construct the same doublestranded oligos as used for electrophoretic mobility shift assays (EMSAs) 1 (see below) were cloned, resulting in a construct with two Brachyury consensus binding sites (B), two variant T-sites (ϩ9/ϩ29), two mutant variant T-sites (ϩ9/ϩ29mut), or one variant T-site including flanking sequences (Ϫ14/ϩ35). All transfections contained equal amounts of pCDNA3.1 and included 1 g of pSV ␤-galactosidase plasmid (Promega) as an internal transfection control. CAT, luciferase, and ␤-galactosidase activities were measured 48 h after transfection by standard methods. All transfections were performed in duplicate, and at least two independent experiments were done to confirm reproducibility.

Both T-box Mutations and Deletion of the Repression Domain of TBX2
Abrogate p19 ARF Repression-To investigate the role of the T-box domain of TBX2 in mediating p19 ARF repression, we made two point mutants, TBX2TB(R122E,R123E) and TBX2TB(A272E). These mutations were designed based on the crystal structure of the T-box domain of Brachyury in complex with a consensus T-site (7) and were predicted to disrupt DNA binding. In previous work we showed that TBX2 rescues MEFs from senescence and leads to their immortalization (10). In addition, the high p19 ARF levels in BmiϪ/Ϫ MEFs due to abrogation of Polycomb repression of the INK4a/ARF locus is efficiently reduced by even low levels of TBX2, resulting in immortalization of these prematurely senescent fibroblasts (27). In contrast, mock-infected wild-type and BmiϪ/Ϫ MEFs as well as cells infected with either one of the T-box mutants enter senescence at the same passage (Fig. 1A). Western blot analysis revealed that both T-box mutants are expressed but show the same high levels of p19 ARF as senescent cells, in sharp contrast to the severely down-regulated p19 ARF levels in TBX2overexpressing cells (Fig. 1B). Mutants of the T-box of TBX3 were also incapable of rescuing MEFs from senescence, as opposed to the readily immortalizing wild-type protein (11). This indicates an essential role for the conserved T-box domain in p19 ARF repression.
In our previous study a deletion mutant of TBX2 lacking the repression domain (TBX2RD) proved to be impaired in repressing p19 ARF , although it still can to some extent down-regulate p19 ARF , probably because of relatively higher expression levels (Ref. 10 and Fig. 1D; also see discussion below). Here we show that this residual activity also induces a prolonged life span of BmiϪ/Ϫ MEFs, which have high p19 ARF levels. However these cells still arrest at later passages. This indicates that TBX2RD is not sufficient to fully bypass senescence ( Fig. 1C) and demonstrates that the repression domain also is a critical contributor to p19 ARF repression.
Deletion constructs of the human p14 ARF promoter transfected to COS-7 cells revealed that TBX2 and TBX3 act as strong dose-dependent repressors through an element located in the region of -19 to ϩ54 (10,11). This repression was found to be dependent on the repression domain of TBX2 (10). In agreement with the data described above, TBX2TB(R122E,R123E) is incapable of and TBX2TBA272E is severely impaired in repressing p14 ARF , whereas both proteins could be detected readily by Western blot (Fig. 1, E and F). Although both the repression domain and the T-box domain are involved in mediating p19 ARF or p14 ARF repression, neither of the mutants appears to work as a dominant-negative in transient repression assays, i.e. we observed no competition with wild-type TBX2 or TBX3 for p14 ARF promoter repression (data not shown).
A Newly Identified Variant T-site in the p14 ARF Promoter Is Bound by TBX2 and TBX3-To test whether the inactivating mutations of the T-box reflect impaired binding to the p14 ARF promoter we performed electrophoretic mobility shift assays. Incubation of probe Ϫ14/ϩ35, spanning part of the repressed region, with either TBX2-or TBX3-overexpressing COS-7 cell extracts resulted in a shifted complex that was not observed using mock-transfected extracts ( Fig. 2A, lanes 1, 2, 15, and  16). This complex does contain TBX2 because it is not affected by pre-immune serum but can be supershifted by rabbit polyclonal ␣-TBX2 antibody to a complex barely capable of entering the gel ( Fig. 2A, lanes 3 and 4). The same supershifts were observed with an independently generated mouse monoclonal ␣-TBX2 antibody (data not shown and Fig. 2E). The presence of such large aggregates upon antibody addition was noted before in case of TBX2 (32). The complex is specific because cold -14/ϩ35 probe is able to compete the complex, whereas an unrelated E2F binding site oligo cannot do so ( Fig. 2A, lanes 5  and 7). A probe containing a Brachyury consensus binding site that has been described to bind TBX2 (Ref. 8 and our own unpublished observations), is also able to compete the complex ( Fig. 2A, lane 6). When overexpressing the repression domain mutant the complex with probe -14/ϩ35 could still be formed and supershifted with ␣-TBX2 antibody, although less efficiently ( Fig. 2A, lanes 8 -10). In contrast, both T-box mutants of 1 The abbreviations used are: EMSA, electrophoretic mobility shift assay; MEF, mouse embryo fibroblast; HA, hemagglutinin; CAT, chloramphenicol acetyltransferase; HSV, herpes simplex virus; tk, thymidine kinase; GFP, green fluorescent protein; MSCV, mouse stem cell virus; Luc, luciferase; oligo, oligonucleotide. The asterisk indicates a band that appears because of non-specific antibody recognition. C, proliferation of BmiϪ/Ϫ MEFs infected with control, TBX2, and TBX2RD LZRS-iresGFP retroviruses on a 3T3 schedule. Note that TBX2RD is capable of extending MEF life span although it will not immortalize cells as TBX2 can do. D, Western blot analysis of TBX2 and p19 ARF levels of BmiϪ/Ϫ MEFs at passage 4. Note that TBX2RD is less effective in down-regulating p19 ARF than TBX2, despite its higher expression levels. E, dose-dependent repression of the (Ϫ19/ϩ54) p14 ARF promoter CAT reporter by TBX2 is impaired in the case of TBX2RD, TBX2TB(R122E,R123E), and TBX2TB(A272E) mutants. F, expression of TBX2 wild-type and mutant proteins in COS-7 cells.
TBX2 and TBX3 were unable to form a complex with probe -14/ϩ35 either in the presence or absence of antibody ( Fig. 2A, lanes 11-14 and 17-18). For TBX2TB(R122E, R123E) this is in agreement with the inability of a TBX2R122A mutant to bind a consensus T-site (32).
The only elements present in the small promoter fragment of p14 ARF that is still repressed by TBX2 are an initiator element and an inverse E2F site (28). Although no consensus T-site is present, after a closer inspection of region -19 to ϩ54 we identified a variant T-site matching 13 of 20 of the nucleotides of a consensus palindromic T-site (Fig. 2B). In this variant T-site the T-half-sites are spaced one nucleotide apart but are orientated similarly as in the original consensus site. The possibility that this site is a functional T-site is strengthened by the fact that all four nucleotides marked as important DNA specificity determinants in the crystal structure of Brachyury in complex with a T-site are retained (Ref. 7, Fig. 2B), as are most of the nucleotides (9 of 12) that are selected in 85% of the cases of in vitro binding site selection experiments with Brachyury (Ref. 6, Fig. 2B).
Using EMSA, a TBX2-specific complex could be seen with probe ϩ9/ϩ29 (spanning this variant T-site) that could be supershifted with the rabbit polyclonal ␣-TBX2 antibody but was absent from mock-transfected lysates (Fig. 2C, lanes 1-4). The complex with probe ϩ9/ϩ29 appears to be formed less efficiently than with probe -14/ϩ35, indicating that sequences outside of the ϩ9/ϩ29 probe might also contribute to DNA binding. The addition of ␣-TBX2 antibody appears to stabilize the complex (Fig. 2C, lane 4). Others have seen such a Tprotein-DNA complex stabilization by antibodies as well in the case of Brachyury and Xbra (6,32). In contrast to probe -14/ ϩ35, the supershifted complex with probe ϩ9/ϩ29 migrates faster and yields a sharper band. The complex could be competed with cold ϩ9/ϩ29 oligo and a Brachyury consensus binding site but not with ϩ9/ϩ29mut in which critical nucleotides were mutated (Fig. 2C, lanes 5-7). Moreover, when ϩ9/ϩ29mut was used (as probe), no complex in TBX2 overexpression extracts could be seen either in the presence or absence of ␣-TBX2 antibody, indicating that the mutated nucleotides are indeed involved in DNA binding (data not shown). The repression domain mutant was able to bind ϩ9/ϩ29 in the presence of ␣-TBX2 antibody, again indicating that DNA binding ability of this mutant apparently is weaker and that ␣-TBX2 antibody stabilizes this complex (Fig. 2C, lanes 8 and 9). As expected, the T-box mutant proteins are incapable of binding the variant T-site (Fig. 2C, lanes 10 -13).
To get a better insight into the requirements for DNA binding by TBX2, we performed a more detailed mutagenesis of the variant T-site. We made a set of four mutants: two (14/ϩ35CC and -14/ϩ35GG) in which we mutated two nucleotides in either T-half-site, thereby leaving the other T-half-site intact; and two (Ϫ14/ϩ35CCGG and -14/ϩ35mut) in which we mutated both T-half-sites, respectively, by a four-or seven-nucleotide change (Fig. 2B). When introduced into probe ϩ9/ϩ29 all mutations including those that disrupt only one half-site disrupted DNA binding (data not shown). However when introduced into the larger -14/ϩ35 probe, only mutations disrupting both T-halfsites completely abolished DNA binding (Fig. 2D). In agreement with TBX2 being able to bind consensus T-half-sites (32), we could still detect binding of TBX2 to mutants of either T-half-site, albeit with a lower efficiency. Moreover mutating the left T-half-site appears to disrupt DNA binding more severely than mutating the right T-half-site. Together, these results indicate that the core structure of T-sites is functionally conserved in the variant T-site and that additional sequences in probe -14/ϩ35 contribute to the stability of the complex.
In extracts of MEFs infected with TBX2 we were unable to detect a complex with p14 ARF promoter probes (Fig. 2E, lanes 1  and 2, 4 and 5). Similar TBX2 binding detection problems have been observed previously (Ref. 8). The detection problems are most likely due to masking by the multiple complexes that apparently are formed with the probe in these extracts, as in strong support of direct binding we could detect a large supershifted complex in TBX2-infected MEF cell extracts with both the rabbit polyclonal and mouse monoclonal ␣-TBX2 antibody (Fig. 2E, lane 3). Likewise, a somewhat faster migrating complex was observed with the repression domain mutant (Fig. 2E,  lane 6). The complex was absent when a probe containing the incomplete T-site was used (probe -12/ϩ23, data not shown) or when non-TBX2-overexpressing MEFs were used (Fig. 2E,  lanes 9 and 10). In conclusion these EMSA experiments prove that TBX2 and TBX3 can directly bind p14 ARF promoter oligos that span a variant T-site.
The Variant T-site of p14 ARF Is Required for p14 ARF Promoter Repression-To prove that the variant T-site is the relevant site via which TBX2 and TBX3 mediate repression we mutated the site in (Ϫ19/ϩ54) p14 ARF promoter-reporter constructs at exactly the same positions as those used in EMSAs. Because these mutations reside in the core promoter region, it was not surprising that substantial and progressive activity loss was observed. However, in agreement with the binding data obtained in EMSA, T-half-site mutants (Ϫ19/ϩ54CC and Ϫ19/ ϩ54GG) could still be repressed by TBX2, whereas mutants in which both T-half-sites were disrupted (Ϫ19/ϩ54CCGG and -19/ϩ54mut) were insensitive to TBX2 even at a very high dose (Fig. 3A). From this finding we not only conclude that mutating the variant T-site abolishes the repressive effect of TBX2, thereby proving that this is the relevant site, but also that TBX2 is able to bind and repress when only one half-site is intact.
To further test whether TBX2 can mediate repression through a variant T-site, we made chimeric promoters consisting of the HSV tk promoter driving the luciferase gene, upstream, of which we cloned both the consensus and variant T-sites (Fig. 3B). These constructs have been described previously as only being responsive to TBX2 when four T-half-sites are present (32). However, in our hands the sole HSV tk promoter was repressed 4-fold by TBX2 (Fig. 3C). This might be explained by the cell type used (COS-7 and Phoenix cells by us versus 293 cells in the other study). Nevertheless, when either two consensus T-sites, two variant T-sites of the p14 ARF promoter (sequence ϩ9/ϩ29) or a somewhat larger part of the p14 ARF promoter (sequence -14/ϩ35), are upstream of the HSV tk promoter, these constructs were repressed up to 10-fold by TBX2 (Fig. 3C). In contrast, when the mutant T-site (ϩ9/ ϩ29mut), which was incapable of binding TBX2 in EMSAs, was cloned upstream, the level of repression was comparable with that of HSV tk alone (ϳ4-fold, Fig. 3C). In conclusion, these results demonstrate that TBX2 is able to mediate repression through the variant T-site in the p14 ARF promoter.
TBX1A and Xbra Are Not Able to Act on the p14 ARF Promoter-To test whether the variant T-site in the p14 ARF promoter can also potentially be regulated by other T-box factors or might be specific to TBX2 and TBX3, we performed co-transfection assays in COS-7 cells with the p14 ARF reporter constructs and two T-box-containing transcriptional activators, TBX1A and Xbra. Xbra had no effect on the small (Ϫ19/ϩ54) p14 ARF promoter (Fig. 4A) and could only activate the long (Ϫ2465/ϩ54) p14 ARF promoter 2-fold at high dose (Fig. 4B). Expression of the protein could clearly be detected by its eGFP tag (data not shown). In agreement with previous results, we could potently activate the HSV tk promoter with two up -FIG. 2. TBX2 and TBX3 bind to a variant T-site present in the p14 ARF promoter. A, EMSA with probe -14/ϩ35 and extracts of COS-7 cells overexpressing TBX2, TBX2RD, TBX2TB(R122E, R123E), TBX2TB(A272E), TBX3, TBX3 L143P, or TBX3 Y149S. Competitor oligos (Ϫ14/ϩ35, Brachyury consensus binding site B, and E2F binding site) and antibodies (pr, pre-immune serum; r, rabbit polyclonal ␣-TBX2 antibody) are added as indicated. Note that TBX2, TBX2RD, and TBX3 can bind probe Ϫ14/ϩ35 in both the absence and presence of ␣-TBX2 antibody as opposed to the DNA binding inability of T-box mutants of TBX2 and TBX3. B, schematic outline of sequences -19 to ϩ54 of the p14 ARF promoter with the stream consensus T-sites with Xbra ( Fig. 4C and Ref. 32). In contrast, Xbra did not activate the HSV tk promoter with two upstream variant T-sites (Fig. 4C). Both experiments show that Xbra is not able to regulate via the p14 ARF variant T-site.
TBX1A can reduce p14 ARF promoter activity of the short (Ϫ19/ϩ54) construct when added at very high concentrations, but in comparison, TBX2 is much more active in repressing, also at much lower concentrations (Fig. 4A). Moreover, on the long and more active (Ϫ2465/ϩ54) p14 ARF promoter, we did not see an effect of TBX1A (Fig. 4B). Although we could clearly detect expression of TBX1A (Fig. 4D), we could not check the activity of the protein because TBX1A is not active on the HSV tk promoter with the two upstream T-sites (Ref. 32) and no other TBX1A-responsive promoters have been described to our knowledge. In conclusion, we show that at least the T-box factors TBX1A and Xbra are not able to act via the variant T-site located in the p14 ARF promoter, suggesting a level of specificity for T-box family proteins in target gene recognition.

DISCUSSION
In this report we demonstrate that TBX2 and TBX3 bind a variant T-site in the p14 ARF promoter, thereby down-regulating its gene expression. TBX2/TBX3 can bind this variant T-site, whereas point mutants of the DNA-binding T-box domain, both at the C-terminal and N-terminal parts, can not do so. The point mutants are to a variable extent impaired in repressing p14 ARF in transient repression assays, i.e. TBX2TB(R122E,R123E) is completely inactive and TBX2TB(A272E) weakly active at high concentrations. Nevertheless, both mutants are incapable of down-regulating endogenous p19 ARF levels in MEFs and because of this are incapable of bypassing senescence arrest. Although in vitro site selection experiments with Brachyury in the past (6) and with Xbra, VegT, and eomesodermin more recently (34) appear to select for a repeat (palindrome) of Thalf-sites, thus far described in vivo T-box targets seem to be regulated by T-half-sites. Some of these promoters contain multiple, separate T-half-sites, such as in the case of Ci-trop regulation by Ci-Bra in Ciona intestinalis (35), Bix4 by VegT (36) or eFGF by Xbra in Xenopus (37), orthopedia regulation by Brachyenteron (Byn) in Drosophila (38), and the melanocytespecific TRP-1 promoter by Tbx2 (8). Interestingly, for Bix4 and TRP-1, the T-half-sites map in the vicinity of the transcriptional start site or even within the initiator element (TRP-1), in analogy to the here described variant T-site in the p14 ARF initiator (inr) element and the inverse E2F site underlined. An alignment is made with a consensus T-site. Nucleotides marked as important DNA specificity determinants in the crystal structure of Brachyury in complex with a T-site are shown in capital letters (7), and nucleotides that are selected in 85% of the cases of in vitro binding site selection experiments with Brachyury are underlined in bold (6). Underneath the p14 ARF promoter sequence, the probes used in the EMSAs are depicted. For variant T-site mutations the differences with wild-type are indicated in bold. C, EMSA with probe ϩ9/ϩ29 and extracts of COS-7 cells. In analogy to panel A, TBX2 and TBX2RD are able to bind probe ϩ9/ϩ29, spanning the variant T-site in p14 ARF . D, EMSA with both wild-type and mutant -14/ϩ35 probes demonstrating the TBX2 is able to bind when one of the T-half-sites is disrupted but not when both T-half-sites are disrupted. Extracts used are from mock-transfected (Ϫ) or TBX2-overexpressing (T) COS-7 cells. E, EMSA with probe -19/ϩ54 and extracts of ARFϪ/Ϫ MEFs noninfected (Ϫ), TBX2-infected (T), or TBX2RD-infected (RD). Pre-immune serum (pm) and mouse polyclonal ␣-TBX2 antibody (m) were added as indicated. In strong support of direct binding, a supershifted complex is detected, upon addition of antibody, that migrates slightly faster for TBX2RD compared with wild-type TBX2.

FIG. 3. The variant T-site is necessary for the repression of the p14 ARF promoter by TBX2. A, dose-dependent repression of wild-type
and T-half-site mutant (Ϫ19/ϩ54) p14 ARF promoter-Luc reporters (p14 ARF wt, p14 ARF (CC), and p14 ARF (GG)) by TBX2 but not of (Ϫ19/ ϩ54) p14 ARF promoter-Luc reporter constructs in which both T-halfsites are disrupted (p14 ARF (CCGG) and p14 ARF (mut)). initiator. Other T-box targets are regulated by cooperation of a T-box protein and a cofactor that bind to contiguous sites within the same regulatory element. Good examples are Tbx5 and Nkx2-5, which synergistically activate the cardiac-specific Nppa, ANF, and cx40 promoters. Moreover, these two proteins bind these promoters in tandem, on an element containing both binding sites, but can also interact in the absence of DNA (39,40). In addition, TBX2 and Nkx2-5 simultaneously interact on a double site of the ANF promoter, leading to its repression (41). In pituitary cells, Tpit or TBX19 and Pitx cooperate to activate POMC (pro-opiomelanocortin) gene transcription also by binding such a double site (42,43). Notwithstanding these recent discoveries, relatively little is known about the sequence requirements that determine individual target gene specificity for the large family of T-box domain-containing transcription factors.
Here, we demonstrate for the first time that an important T-box target, the p14 ARF tumor suppressor gene, is regulated by a variant palindromic T-site. Whereas the core CACC-NNNGGTG nucleotides of the variant site, which form an imperfect palindrome, are well conserved in the "consensus" T-site, flanking nucleotides are divergent (Fig. 2B). The imperfect palindromic site is reminiscent of the preference for a repeat of T-half-sites for Brachyury, Xbra, VegT, and eomesodermin found in in vitro binding studies (6,34). In addition, our mutational analysis of the conserved "core" sequence clearly demonstrates its requirement for the binding of TBX2 and TBX3 to the variant site, likely reflecting the highly conserved structure of the T-box domain (7). Although TBX2 in vitro has previously been shown able to bind to T-half-sites, a preference for (palindromic) T-sites appears to exist (32). In addition, it was shown that TBX2 can bind variants of a consensus T-halfsite that were generated by mutagenesis of the consensus Tsite (8). However such sequences are not present in the small -19/ϩ54 p14 ARF promoter fragment, and their biological relevance for in vivo target gene regulation remains to be demonstrated. Interestingly, it has just been reported that orthopedia is regulated by Byn via 15 binding sites, which all differ in at least two nucleotides from a consensus T-half-site (38). Therefore, in agreement with our data, Byn can regulate a promoter through variants of a T-half-site. However, the orthopedia promoter does not contain a palindromic T-site as seen for the p14 ARF promoter. Close inspection of more upstream sequence of the p14 ARF promoter revealed no other variant T-(half) sites. Presumably the p14 ARF palindromic variant T-site is a high affinity site in vivo and might therefore be capable of conferring repression on its own. Importantly, the divergent flanking sequences of the variant p14 ARF T-site contribute to determining the specificity for binding to TBX2 and TBX3, as the related T-box-containing proteins TBX1A and Xbra are not able to activate transcription of reporter constructs harboring this variant site. In contrast, the orthopedia promoter could also be activated by mouse and Xenopus Brachyury, suggesting that these proteins can also recognize Byn sites (38). However the activation by Brachyury via the multiple Byn sites is much lower than in case of Byn, which could reflect intrinsic Byn preference/specificity associated with these sites. The occurrence of variant T-sites in relevant target genes clearly illustrates the in vivo selected specificity in target gene recognition associated with individual T-box protein family members.
A further indication of such divergence in target gene recognition is illustrated by the fact that TBX2 alone can potently repress p14 ARF , in contrast to the situation for TBX5 and its co-factor Nkx2-5, which need to bind and act in cooperation to strongly activate their target genes (39,40). This fact and the apparent inability of the C-terminally deleted TBX2 mutant to compete with wild type TBX2 for p14 ARF binding suggest that the existence of a co-factor for TBX2 may not be a prerequisite. However, our EMSAs do point to a role for flanking sequences outside of the variant T-site to promote stable complex formation, which in turn could point to the requirement for binding of such a co-factor, although the sequences of the p14 ARF promoter, aside from an imperfect E2F site, do not indicate the presence of consensus binding sites for established or potential co-factors such as Pitx or Nkx2-5.
In addition to the T-box we also found a C-terminal conserved domain of TBX2 to be essential for repression. A deletion mutant of this repression domain could still bind the variant T-site, although somewhat less efficiently. As the affinity of the repression domain mutant for the variant T-site is impaired, this might explain why this mutant could not compete with wild-type TBX2 for repression. In contrast, previously others (32) have demonstrated comparable DNA binding activity of wild-type and repression domain mutant proteins; however, in these studies a probe was used that contained four T-half-sites, which may well explain the different outcome. The repression domain was first mapped within Xenopus ET and its human ortholog, TBX3, and has subsequently been found to be highly conserved in TBX2 (amino acids 535-629 in TBX2, Ref. 9). Others (32) claim the existence of an additional repression domain in TBX2 at position 407-561, although they do not acknowledge the small overlap between these two regions. Our inactivating deletion encompassed amino acids 501-618, thus affecting both proposed domains. Although the repression domain deletion mutant is incapable of fully rescuing MEFs from senescence, the protein does appear to have some residual ability to down-regulate p19 ARF , contributing to an extended MEF life span. Likewise, others (44) have recently noticed a low efficiency of MEF immortalization by TBX3 repression domain deletion mutants as well. This may well be explained by the fact that the repression domain mutants retain the ability to bind to the variant T-site, which is embedded in the initiator sequence. Conceivably, such binding may interfere to some extent with promoter function by obstruction. Nevertheless, our results clearly indicate that active repression by TBX2 via the repression domain is required for p14 ARF down-regulation. We hypothesize that the repression domain is involved in recruiting other proteins, such as co-repressors, to the p14 ARF promoter, thereby mediating repression and creating the observed slow migrating EMSA complexes. Remarkably, so far only one protein, in addition to the two above described transcription co-factors, has been shown to bind a T-box protein family member. This factor is CASK (calcium/calmodulin-dependent serine protein kinase), a membrane-associated guanylate kinase and component of cell junctions, which binds Tbr-1 and then translocates to the nucleus (45). In vitro this complex can bind T-sites and activate transcription of the T-site-containing reelin promoter. Whether this reflects a brain-specific or more general mechanism remains to be seen.
In conclusion we identified a variant T-site, composed of two inverted, imperfect T-half-sites, as the essential TBX2/TBX3 binding element in an important in vivo relevant TBX2/TBX3 target, the p14 ARF tumor suppressor promoter. The core structure of T-box/T-site is conserved, as disrupting mutations could be made in either the binding site DNA or in critical T-box amino acids, based on knowledge of the complex of Brachyury and a consensus T-site. The existence of a variant palindromic T-site described herein is, to our knowledge, unprecedented and offers a possible explanation for the selection of a repeat of T-half-sites in the in vitro binding site selection experiments performed with T-box proteins. Importantly, our study points to a hitherto unsuspected level of specificity for individual T-box factors in recognizing their respective target genes, which opens new avenues of research in the further exploration of T-box target gene regulation.