Characterization of the TBX5 binding site and analysis of mutations that cause Holt–Oram syndrome

Holt–Oram syndrome is caused by mutations in TBX5, a member of the T-box gene family. In order to identify DNA sequences to which the TBX5 protein binds, we have performed an in vitro binding site selection assay. We have identified an 8 bp core sequence that is part of the Brachyury consensus-binding site. We show that TBX5 binds to the full palindromic Brachyury binding site and to the half-palindrome, whereas Brachyury does not bind to the TBX5 site. Amino acids 1–237 of TBX5 are required for DNA binding. Analysis of the effects of specific substitution mutations that arise in Holt–Oram patients indicates that G80R and R237Q eliminate binding to the target site. DNA database analysis reveals that target sites are present in the upstream regions of several cardiac-expressed genes including cardiac α actin, atrial natriuretic factor, cardiac myosin heavy chain α, cardiac myosin heavy chain β, myosin light chain 1A, myosin light chain 1V and Nkx2.5. Cell transfection studies demonstrate that TBX5 activates the transcription of an atrial natriuretic factor reporter construct and this effect is significantly reduced by deletion of the TBX5 binding site.

INTRODUCTION

Holt–Oram syndrome (HOS, OMIM 142900) is a dominantly inherited disorder, which affects the development of the heart and upper limb (1). HOS is caused by mutations in TBX5 (2–5), one of a growing number of genes that make up the T-box gene family (6,7). These genes are related to the founder member, Brachyury (T), and to each other, by virtue of a conserved DNA sequence, the T-box, which encodes a DNA binding motif, the T-domain (8,9). Members of the T-box gene family play crucial roles during vertebrate and invertebrate development (10). Over the past 2 years, many new members of this gene family have been identified in a variety of species including human (11), mouse (12,13) and Caenorhabditis elegans (14). In humans, for example, 14 different T-box genes have been identified and most of these have also been described in the mouse (11).

Little is known about the function of most T-box genes. Experimental approaches and studies of mutants have shown that Brachyury and its Xenopus homologue (Xbra) are required for the formation of notochord and posterior mesoderm (15), and the functional inactivation of Tbx6 resulted in a mouse with three neural tubes (16). Detailed tissue-specific patterns of expression have been determined for several mouse T-box genes and it is clear that even the most closely related members of this gene family display distinct, though sometimes overlapping, spacial and temporal patterns of expression during development (17). Tbx2, Tbx3, Tbx4 and Tbx5, for example, are closely related having arisen from a common ancestor through two duplication events (18). Tbx2 is most closely related to Tbx3, and Tbx4 is most closely related to Tbx5. The evolutionary links between the members of each cognate pair, Tbx4/5 and Tbx2/3, are reflected in the overall similarity of their expression patterns (17). However, the difference in expression patterns between Tbx4 and Tbx5 in developing limbs is striking, and has been the subject of much recent interest. Whereas Tbx4 is expressed principally in the lower limbs and not in the upper limbs, Tbx5 is, for the most part, expressed in the upper limbs but not in the lower limbs (19–22). Moreover, recent work indicates that it is Tbx4 and Tbx5 that actually control fore and hind limb identity (23,24). Consistent with the role of TBX5 in the development of the upper limb, skeletal abnormalities in HOS range from triphalangial thumb or sloping shoulders in mildly affected patients to severe upper limb reduction defects in the most seriously affected (1).

In addition to its role in specifying forelimb identity, TBX5 is also involved in heart development. In situ hybridization studies in mouse, chick and humans show that TBX5 is expressed in the cardiac crescent early in the developing heart and subsequently in an asymmetric fashion in the left ventricle and on the left side of the ventricular septum. It is expressed strongly throughout the atrial walls, the atrial septa and the atrial aspect of the atrioventricular valves (2,25). This pattern of expression is consistent with the cardiac phenotype in HOS which includes, among other things, atrial septal defects (ASDs), ventricular septal defects (VSDs) and AV block (1).

The molecular basis of developmental cardiac abnormality is poorly understood, and TBX5 represents one of the few genes which, when mutated, is known to cause congenital heart disease. Another gene that is involved in the aetiology of congenital heart disease is NKX2.5 (26,27), a human homologue of tinman, the Drosophila homeobox-containing transcription factor (28) which is essential for the development of the dorsal vessel, the Drosophila equivalent of the vertebrate heart. Mutations in NKX2.5 cause dominantly inherited atrial septal defects (29,30). TBX5 encodes a protein that probably also acts as a transcription factor, which is likely to be the case for all members of the T-box family. However, despite the rapid increase in the number of characterized T-box genes, and detailed studies of their expression profiles, little is known about the downstream targets of these transcription factors. The best-studied T-box protein is Brachyury. Using an in vitro binding assay it was possible to identify a consensus DNA target site, for Brachyury, which consists of an almost palindromic 20 bp sequence (8). However, to date, it has been difficult to identify genes that are regulated through interaction with this 20 bp sequence. Casey et al. (31) have shown that in Xenopus, Xbra regulates the expression of eFGF by binding to a 10 bp sequence that is a half site of the Brachyury consensus site (31). Furthermore, Xbra and Veg T, a T-box gene whose mRNA is localized in the vegetal hemisphere of the egg and early embryo, have been shown to regulate Bix4, through binding to two separate Brachyury half sites, located 19 bases apart in the promoter of this gene (32). The only other T-box protein for which there is direct evidence of its target binding site is Tbx2. This protein has been shown to act as a repressor of the tyrosine-related protein 1 gene (TRP-1) through binding to a 6 bp sequence that is part of the Brachyury consensus (33).

Identification of the binding site for TBX5 is critical if we are fully to understand the molecular pathway in which it acts. Putative targets for TBX5 include HOX genes, other T-box genes, and members of the BMP, WNT and FGF gene families (23,24). Misexpression studies of Tbx5 in the chick produce effects on the expression of Hoxc9 and Hoxd9 (23) and on the expression of Tbx2, Tbx3 and Tbx4 (24). Recently Liberatore et al. (34) have generated transgenic mice that express Tbx5 from a β-myosin promoter throughout the primitive heart tube. Retardation of ventricular chamber morphogenesis and loss of ventricular-specific gene expression of Mlc2v were observed (34). In Xenopus, studies using a dominant negative version of XTbx5 have also resulted in a significant reduction or absence of myosin light chain 2 (Xmlc2), endogenous Tbx5 and Nkx2.5 (35). The effect on Nkx2.5 is particularly interesting in view of its role in causing atrial septal defects (29). Thus, indirect evidence from experiments in other species points to several potential targets for human TBX5.

In order to identify the DNA sequence(s) to which TBX5 binds, and ultimately the direct downstream target genes of this protein, we have performed an in vitro oligonucleotide selection procedure. We report that TBX5 binds to an octomer sequence that is part of the Brachyury consensus half site. This sequence is present in the upstream region of several important cardiac-expressed genes, including cardiac α actin (ACTC), atrial natriuretic factor (ANF), cardiac myosin heavy chain α (MYH6), cardiac myosin heavy chain β (MYH7), myosin light chain 1A (Mlc1a), myosin light chain 1V (Mlc1v) and Nkx2.5. Using cell transfection assays we show that TBX5 activates transcription of an ANF reporter construct, consistent with a role in the regulation of ANF expression. We have also characterized the DNA binding domain of TBX5 and show that some substitution mutations, which occur in HOS patients, eliminate binding to the DNA target site.

RESULTS

Identification of the TBX5 target binding site

In order to identify the DNA sequence to which TBX5 binds we have performed an in vitro selection procedure using a TBX5 protein with a C-terminal hexahistidine tag (TBX5-His6). The TBX5-His6 protein was produced in Escherichia coli as inclusion bodies. The protein was urea solubilized, affinity purified and renatured on a Ni-NTA agarose column. The protein was used directly for the in vitro selection of double-stranded random 26mer oligonucleotides (36). The oligonucleotides were subjected to rounds of selection followed by PCR amplification, and aliquots of the oligonucleotide pools were examined using an electromobility shift assay (EMSA) following zero, one, two and three rounds of selection. Specific complexes could be identified from the second and third rounds of selection. Double-stranded oligonucleotides obtained from the third round of selection were cloned and sequenced. Of 41 clones analysed, 35 could be aligned around a core containing the sequence (A/G)GGTGT, or its complement. Six clones contained sequences that differed from this core at a single base. Alignment of sequences along the core, allowed us to derive the consensus sequence (A/G)GGTGT(C/G/T)(A/G), which is part of the Brachyury binding site. The summary of this alignment is shown in Figure 1A. Seventeen of the clones contained two such core sequences, in tandem or in an inverted orientation, separated by 1–14 nucleotides. The alignment of sequences from the 17 secondary cores provided the consensus (A/G/T)GGTG(T/C)(T/G/C)(A/G/C) and is summarized in Figure 1B. Some of the bases in some of the cores were derived from the primer sequences flanking the random 26 bases of the oligonucleotides. The bases from the primers were not included when deriving the consensus core sequence.

Binding preferences of TBX5 to its target site

Binding studies were performed on fragments released from six of the 41 clones (Fig. 2A). TBX5 binds weakly to target sequence T-1, which contains a single core sequence (AGGTGTCC). Increased binding was observed with a fragment, T-26, containing a single site (AGGTGTTG), which differed from T-1 in the sequence at the 3′ end of the core. Fragments containing two target sites in each of the three possible orientations were analysed. Following a convention that assumes the Brachyury target palindrome is oriented head-to-head, the binding sites examined here are oriented tail-to-tail in T-11 and T-18, head-to-tail in T-33.2 and head-to-head in T-31 (Fig. 2B). Fast and slow mobility complexes were observed for the fragments containing two target sites (Fig. 2A). With fragment T-11, TBX5 predominantly formed a slow mobility complex whereas with T-18 and T-33.2 the proportions of slow and fast mobility complex were approximately equal. With fragment T-31, where the two halves of the palindrome are in the same orientation as on the Brachyury target site, the majority of the complex formed was of fast mobility with only a small proportion showing slow mobility. Competition with unlabelled oligos and anti-TBX5 antibody showed that these fast and slow mobility complexes were specific (data not shown).

The binding studies above indicate that TBX5 predominantly forms a slow mobility complex with fragment T-11. In order to determine whether the slow mobility complex represents a dimer, resulting from interaction of two full-length TBX5 molecules, we examined band shifts of fragment T-11 with full-length and truncated forms of TBX5. Whereas full-length TBX5 produces a slow mobility complex, the truncated TBX5 (mostly T-domain, with residues 1–237) produces two faster mobility complexes in which the lower of the two bands is more strongly represented than the upper. In the presence of mixed full-length and truncated TBX5 an intermediate complex was observed (Fig. 2C).

TBX5, Brachyury and their respective DNA binding sites

The binding site selection studies indicate that TBX5 recognizes a DNA sequence that is a part of the Brachyury consensus-binding site. In fact, the TBX5 target site contains the first 8 bases from one half of the 22 bp palindrome that is the Brachyury target. Therefore TBX5 should recognize both the full-length Brachyury palindrome (T-site) and the half-palindrome (T-site). This was tested on EMSA studies using the T-site and the T-site as targets. TBX5 produced a specific shifted complex with both the full-length T-site and the T-site and supershifts in both cases in the presence of anti-TBX5 antibody (Fig. 3A). No difference was detected in the pattern of complex formation on EMSA with full-length and T site targets. In order to check whether the mouse Brachyury protein recognizes the TBX5 consensus-binding site, we synthesized both TBX5 and Brachyury using in vitro coupled transcription and translation and performed EMSA studies with a 15 bp oligonucleotide probe containing the TBX5 consensus DNA target site defined in Figure 1. Although TBX5 gives rise to a complex, mouse Brachyury failed to produce any mobility shift on the oligonucleotide containing the TBX5 consensus-binding site (Fig. 3B).

Dissection of the TBX5 DNA binding domain and the effect of missense mutations

Amino acid sequence alignment with other T-box proteins putatively assigned the DNA binding domain of TBX5 to residues 55–237. To determine the DNA binding domain precisely, different truncated versions of TBX5 were generated by coupled in vitro transcription and translation, and analysed for DNA binding activity to fragment T-11 (Fig. 4A). Amino acid residues 1–237 were the minimum required for DNA binding and the presence of four additional amino acids (238–241) eliminated binding. Removal of the C-terminal 281 amino acids (238–518 inclusive) significantly enhanced the protein’s DNA binding affinity compared with the full-length TBX5. Removal of residues 1–54 from the full length TBX5 prevented its binding to the DNA target. In vitro generated N-terminally truncated proteins showed some degradation compared with full-length TBX5 when analysed on SDS–PAGE gels (data not shown).

Several substitution mutations of TBX5 have been described in HOS patients (4,5) and we have examined the effect of some of these on DNA binding. In addition to full-length TBX5, truncated versions of the protein were utilized in these studies because some, such as TBX5(l–279), show considerably increased binding to the DNA target site compared with the full-length protein. The effect of substitution mutations G80R, G169R, R237Q and S252I on the binding of TBX5(1–279) to target T-11 is shown in Figure 4B. Substitution mutations G80R and R237Q eliminate binding, whereas mutations G169R and S252I do not. The same results were observed for mutations G169R, R237Q and S252I on full-length TBX5 (data not shown). The effect of mutation G80R has not been tested on the full-length protein.

Putative TBX5 binding sites map upstream of several cardiac-expressed genes

In order to establish whether the DNA target sequence for TBX5 is present in the upstream region of characterized genes, we have analysed the available sequences in the EMBL database from the upstream regions of ANF, MYH6, MYH7, Mlc1a, Mlc1v and Nkx2.5. The sequences were searched using the ‘Findpatterns’ program with two different patterns, RGGTGTBR, as found in the primary TBX5 core consensus, and the less stringent DGGTGYBV that is the consensus of the secondary core. These results are summarized in Table 1. Assuming equal representation of each base, there are more putative sites in the upstream region of MYH6 than would be expected by chance (χ² test, P < 0.01). In the upstream regions of ANF, MYH7, Nkx2.5, Mlc1a and Mlc1v the number of putative binding sites is similar to that expected at random. However, the promoter regions of Mlc1a, MYH6 and MYH7 contain pairs of binding sites in extremely close proximity. In Mlc1a these are located 1250 bp upstream from the transcriptional start in a head-to-head orientation, separated by 9 bp. In the MYH6 upstream region, 1 kb from the transcriptional start, there is a 480 bp interval containing six sites, two of which are arranged in tandem, separated by 2 bp. Sixty-one base pairs upstream from the MYH7 transcriptional start site there are two TBX5 target sites in tandem, separated by 9 bp. The probability of finding three or more paired sites, separated by 10 bp or less, within the upstream regions of ANF, MYH6, MYH7, Nkx2.5, Mlc1a and Mlc1v is significant at P < 0.01 using a Poisson frequency distribution. One of the TBX5 target sites within the ANF promoter is located at positions 207–214 of sequence accession K02043, which overlaps with a potential enhancer sequence located at positions 203–213 (37). This site is conserved in the mouse.

TBX5 binds to the upstream region of ANF and activates reporter expression

As a putative TBX5 DNA binding site is present within the ANF promoter at a position that overlaps with a potential enhancer sequence (37) we set out to test whether TBX5 would form a complex with the ANF upstream region on EMSA. Bacterially expressed TBX5 was tested against double-stranded 24mer oligonucleotides, containing the wild-type and mutated ANF promoter sequence, shown in Figure 5A. The results of the EMSA studies show that the wild-type ANF sequence forms a complex with TBX5, whereas the mutated ANF sequence did not. The complex formed could be supershifted with anti-TBX5 antibody. Competition with wild-type and mutant oligos showed that TBX5 binding to the wild-type ANF sequence was specific (Fig. 5A).

In order to test whether TBX5 could transactivate a reporter construct containing the upstream region of ANF, co-transfection studies were performed with the TBX5 expression plasmid, pcDNA-TBX5, and two different ANF reporter constructs. pGL3-WtANF contained bases –3 to –337 of the ANF promoter cloned upstream of a luciferase reporter gene and pGL3-ΔANF contained the same fragment and reporter but with the TBX5 binding site deleted. Co-transfection of rat cardiomyocyte cell line H9c2 cells with pcDNA-TBX5 and pGL-3ANF resulted in a 6-fold increase in reporter activity over basal level (Fig. 5B). This TBX5 mediated activation was significantly reduced with pGL3-ΔANF in which the TBX5 site had been deleted. The difference in fold activation between the wild-type and TBX5 binding site-deleted constructs is significant using a student t-test (*P < 0.05). Similar transfection studies using COS-7 cells produced a 2–4-fold activation of the reporter which was not reduced following deletion of the binding site (data not shown).

Whole-mount in situ hybridization was performed on mouse embryos to compare the expression pattern of ANF with that of Tbx5. Expression of both genes is seen in the right and left atria and in the left ventricle (Fig. 5C).

DISCUSSION

If we are to develop a comprehensive understanding of the molecular basis of congenital heart disease it will be necessary to unravel the complex regulatory processes that control heart development. Some progress has been achieved in recent years with the identification of key cardiac genes, some of which have been implicated in inherited heart disorders (2,3,29). For example, we and others have shown that mutations in TBX5 cause HOS, which is characterized by a range of developmental cardiac and limb abnormalities (1,38). TBX5 contains a DNA binding domain that is conserved in other T-box family members, some of which have been shown to act as transcription factors (31,33).

In order to identify the target DNA sequence to which TBX5 binds we have performed an in vitro oligonucleotide enrichment procedure similar to that employed to identify the target site of Brachyury (8), and a consensus TBX5 binding site sequence, (A/G)GGTGT(C/T/G)(A/G), has been derived. This representspart of a half site from the Brachyury target sequence, which is a 20 bp near-palindrome consensus comprising of T(G/C)ACACCT/AGGTGTGAAATT. Although the binding site sequence was characterized in the mouse 6 years ago, the identification of downstream target genes regulated by Brachyury has proved elusive. However, two recent studies have identified putative targets for Xenopus Brachyury (Xbra) (31,32). Casey et al. (31) showed that Xbra regulates the expression of embryonic fibroblast growth factor (eFGF) by binding to an element, TTTCACACCT, that is located 1 kb upstream from the transcriptional start site. Furthermore, Tada et al. (32) identified two 10 bp elements in the Bix4 promoter to which Xbra and another T-box gene product, VegT, bind and induce luciferase expression from reporter constructs. The proximal element, ATTCACACGT, is located 66 bp from the Bix4 transcriptional start and the distal element, CTTCACACCT, is located 9 bp further upstream. In the ascidian, Ciona intestalis, Brachyury acts as a transcriptional enhancer of Ciona tropomyosin through binding to two sets of low-affinity half sites (39). In addition to the identification of targets for Brachyury, targets for Tbx2 have also been recognized (33). Tbx2 binds to two elements, MSEu and MSEi, in the promoter of the tyrosine-related protein 1 gene (TRP-1). MSEu and MSEi are located at –237 and at the transcription initiator site, respectively, of the TRP-1 gene. The core element recognized by Tbx2 in both MSEu and MSEi is a 6 bp sequence, GTGTGA (33). This sequence is present within the Brachyury binding site and it forms part of the TBX5 binding site. Two recent studies have described binding sites for other T-box genes (40,41). Szabo et al. (40) report putative T-box sites in the IL-2 and IFNγ genes. In each case the consensus match to the T binding site derives from 10 bases in the middle of the palindrome and transcription is activated by T-bet (40). Lamolet et al. (41) describe a near identical half site to that of the mouse T palindrome in the promoter of the POMC transcription factor. Their data suggest that transcription of POMC is activated by T-pit following binding to this site (41).

In the present study, binding site selection experiments have shown that TBX5 binds to single and paired non-palindromic sites in vitro. Binding to paired sites varies with spacing and orientation of the sites. With fragment T-11 in which the binding sites are oriented tail-to-tail, TBX5 predominantly forms a slow mobility complex. With the same fragment truncated TBX5 formed two complexes, both of which were of faster mobility than observed with the full-length protein. A mixture of full-length and truncated TBX5 produced an intermediate complex. The simplest interpretation of these data is that on target fragment T-11 full-length TBX5 binds mostly as a dimer, whereas truncated TBX5 binds preferentially as a monomer with reduced dimer formation. When both full-length and truncated TBX5 are present, the intermediate complex resulted from the binding of one full-length and one truncated molecule on the DNA fragment. It is noticeable for fragment T-31, in which the binding sites are oriented head-to-head, as for the two halves of the Brachyury palindrome, the majority of the complex formed with full-length TBX5 is of fast mobility, probably indicating binding by a monomer of TBX5. Whether this reflects a different preference in binding site orientation for TBX5 compared to Brachyury, or is simply due to problems of stoichiometry caused by the spacing of the two sites, requires clarification by further EMSA studies and X-ray crystallography. As TBX5 binds to both the full and half palindrome recognized by Brachyury and mouse Brachyury does not bind to the TBX5 consensus site it would appear that the sequence requirement for TBX5 binding, as for TBX2 binding, is less stringent than that for Brachyury binding.

In order to define the TBX5 DNA binding domain, protein truncation studies were performed. Although TBX5 N-terminal amino acid residues 1–54 are not fully conserved across species, removal of these residues from the full-length protein prevented binding to its DNA target. In view of the observed degradation, on SDS–PAGE, of in vitro generated N-terminally truncated TBX5 it is not possible to determine whether deletion of residues 1–54 prevents binding per se or whether the N-terminal deleted proteins suffer proteolysis. In mouse T, the removal of N-terminal non-conserved residues upstream of the T-domain also significantly reduced or eliminated its binding to the DNA target (8). Recently two missense mutations in HOS patients have been mapped to the N-terminal region of TBX5 (42), pointing to a functional role of this part of the molecule. Amino acid residues 1–237 were essential for DNA binding. Removal of the C-terminal 281 amino acids significantly enhanced the DNA binding affinity of TBX5, possibly indicating that the C-terminus functions to prevent or at least inhibit binding to DNA. Thus, TBX5 may require an interacting protein partner to fully affect transcription of target genes. Such an interaction has been described recently between two other cardiac transcription factors, GATA4 and NKX2.5, which synergistically activate the transcription of ANF and ACTC (43,44).

Twenty-four different HOS mutations have been described (2–5,42), the majority of which are nonsense mutations that would result in truncated proteins. Five missense mutations have been identified including G80R, G169R, R237Q, R237W and S252I. In order to gain insights to the mutational basis of HOS, we have generated TBX5 expression constructs containing nucleotide changes that correspond to mutations in HOS patients, for use in EMSA studies. Specifically, we have analysed missense mutations G80R, G169R, R237Q and S252I, and truncation mutation R279St. Mutant protein R279StTBX5 retained strong DNA binding activity (Fig. 2B). In fact, the truncated protein was required at approximately one-tenth the molar ratio of the full-length protein to produce an equivalent band-shift (data not shown). Although nonsense mutation R279St is predicted to create a polypeptide with an intact T-domain, which lacks most of the C-terminal region, it is not clear whether such a protein would be biologically active or whether its transcript would be subjected to nonsense-mediated decay (45). It is generally considered that truncation mutations would be non-functional resulting in haploinsufficiency of TBX5 (2,3). Proteins carrying the G80R and R237Q mutations completely lacked binding activity to the target DNA, whereas those with S252I and G169R retained binding activity. Individuals with mutation G80R show significant cardiac abnormalities but only minor skeletal problems whereas individuals with the R237Q mutation show significantly more severe limb abnormalities than are found in patients with nonsense mutations (4). This observation has led to the suggestion that TBX5 missense mutations may produce organ-specific dominant negative effects because the mutant proteins show different biophysical interactions with various target DNA sequences (4). The observed in vitro binding activities of the mutant proteins, however, would suggest otherwise. The complete absence of binding to the target site by TBX5 proteins carrying mutations G80R and R237Q would suggest that the differences in phenotype caused by these mutations are more likely due to differences in the way TBX5 interacts with distinct protein partners in developing limb and heart.

Analysis of mutation G169R also points to the relevance of protein–protein interaction in HOS missense mutations. The structure of the Xbra T-domain bound to target DNA has been resolved (46) and projection of TBX5 onto this structure (4) indicates that glycine 169 is not directly involved in DNA contact or in dimerization. Glycine 169 is part of a hydrophobic stretch within the T-domain (46) which is probably involved in protein–protein interaction. Replacement of glycine with arginine, a positively charged amino acid with a relatively bulky amino group, is likely to affect such protein–protein interaction in physiological conditions, but not prevent binding to the DNA target consistent with the retained binding activity of mutant protein G169RTBX5. Mutation G169R was identified in a Finnish family in which individuals show significant cardiac involvement with very mild skeletal findings (5).

Mutant protein S252ITBX5 showed similar binding to the DNA target to that observed for wild-type TBX5. This is consistent with the location of the S252I mutation, outside the DNA-binding domain. The phenotype of the patient with this mutation is not particularly severe. The only unusual feature is scoliosis, which is not commonly found in HOS. It is possible that scoliosis represents a severe clinical manifestation of the S252I mutation. However, as this mutation is present in a single sporadic case it is also possible that the observed scoliosis is co-incidental.

Searches of DNA sequence databases have revealed several putative cardiac-expressed targets for TBX5. One such gene with a TBX5 binding site in its upstream region is ANF. Two indirect lines of evidence support a role for TBX5 in the regulation of ANF expression. The TBX5 binding site at position –266 to –259 of the ANF promoter overlaps with an enhancer element (37), at position –270 to –260 (position 203–213 of sequence accession K02043). Also, this TBX5 binding site and its position relative to the transcriptional start site are conserved in the mouse. Furthermore, ANF and TBX5 show similar expression patterns throughout the development of the heart with TBX5 being expressed slightly earlier than ANF in the cardiac crescent at 7.5 days post conception (d.p.c.) (34,47).

In order to establish whether TBX5 could activate ANF expression we performed co-transfection studies with TBX5 expression construct pcDNA-TBX5, and two reporter constructs, pGL3-WtANF, containing the ANF promoter region –3 to –337 upstream of a luciferase reporter gene and pGL3-ΔANF, which is essentially the same construct as pGL3-WtANF but with the TBX5 binding site at position –266 to –259 deleted. The co-transfection studies show that deletion of the TBX5 binding site produces a significant reduction in ANF transcriptional activation, suggesting that part of its activity is due to the sequence specific interaction of TBX5 with its binding site. Deletion of the binding site did not entirely eliminate TBX5 activation of ANF, possibly indicating that TBX5 also plays a role in activating transcription at this promoter via interaction with other factors. The functional analysis of HOS mutant proteins should prove insightful in this respect. A dual role for TBX5 in transcriptional activation via direct binding to its target site and through protein–protein interaction with independently bound factors would mirror the situation for other cardiac transcription factors such as GATA4 (43,44). It is noteworthy that the results reported here with cell line H9c2 were not replicated with COS-7 cells. This suggests that H9c2, a cardiomyocyte cell line, may produce a co-factor which interacts with TBX5 possibly helping to recruit it to the binding site, or producing a conformational change once it is bound, to activate transcription. The increased binding affinity of the C-terminal deleted TBX5 may also point to the need for a co-factor to interact with the C-terminal domain of the full-length protein to enhance binding.

The presence of TBX5 binding sites in the upstream regions of other cardiac-expressed genes may suggest a role for this protein in their regulation. Where DNA sequences are available, we have looked for conservation of putative binding sites in the promoters of these genes. This analysis indicates that as for ANF putative binding sites upstream of MYH6 and of Mlc1v are conserved between mouse and human, whereas sites upstream of Nkx2.5 in humans are not conserved in the mouse (data not shown). Functional studies are now required to validate a role for TBX5 in regulating the expression of genes such as ACTC, Mlc1a and Mlc1v. It will be interesting to search for target sites in the upstream regions of other genes suggested by expression studies using dominant negative constructs in Xenopus (35) and by TBX5 misexpression studies in the chick (23,24) and mouse (34). The focus of the present study has been the identification of putative target genes for TBX5 in heart development, but it is equally important to identify similar target genes in the developing limb.

MATERIALS AND METHODS

TBX5 overexpression, purification and antibody production

The coding region of TBX5 was amplified from the cDNA clone pcT5.2 (2) using primer pairs TKGF1, 5′-GATATACATATGGCCGACGCAGACGAGGGCTTTGGC-3′ (sense) and TKGR5, 5′-TCTAGAGGATCCCTAGTGATGATGATGATGATGGCTATTGTCGCTCCACTCTGG-3′ to introduce a hexahistidine tag at the C-terminal end of the expressed protein. The PCR amplified product was cloned into pET11 at the NdeI and BamHI sites. Cloned products were sequenced and plasmid pDG983, which contains the correct sequence, was propagated into E.coli BL 21(DE3) cells and grown in luria broth (LB) in presence of 100 µg/ml ampicillin. The growing culture (OD600 = 0.6) was induced with 1 mM IPTG and allowed to grow for a further 3 h. Cells produced TBX5-His6 as inclusion bodies. IPTG induced cells were harvested by centrifugation and ruptured by sonication. Inclusion bodies were pelleted and solubilized with binding buffer B (20 mM Tris pH 7.4 , 0.5 M NaCl, 1 mM PMSF, 1 µg/ml leupeptin, 1 µg/ml pepstatin and 1 µg/ml aprotinin) containing 6 M urea at 4°C. The clear supernatant was subjected to 33% (NH₄)₂SO₄ fractionation. The pellet was redissolved in buffer B containing 6 M urea, and loaded directly onto a Ni-NTA-agarose column (Qiagen), pre-equilibrated with the same buffer. The column was washed with 10 vol of buffer B containing 20 mM imidazole and 6 M urea. The bound TBX5-His6 protein was then renatured using a gradient of 6–0 M urea, in 80 column volumes of buffer B containing 20% glycerol. Finally, the column was washed with 3 vol of buffer B, without denaturant, and eluted with buffer B containing 200 mM imidazole and 20% glycerol. Imidazole was removed from the eluate by dialysis and concentrated over Ultrafree-15 centrifugal filter device (Millipore) at 4°C. Antibodies against His-tag TBX5 protein were raised in rabbits and purified on a protein A-agarose column (Boehringer Mannheim) according to the manufacturer’s protocol.

For studies involving the DNA binding domain of TBX5 (amino acid residues 1–237) PCR amplification using primers 5′-CAT ATG GCT AGC ACC ATG GCC GAC GCA GAC GAG GGC-3′ and 5′-AGG TCC GAA TTC CTA CCG AAA TCC TTT GGC AAA GGG ATT ATT C-3′ was employed to generate the relevant fragment. The PCR product was digested with NheI and EcoRI and cloned into the same sites of pET28. The resulting construct encodes an N-terminal His-tagged product of 260 amino acids. Purification of the His-tagged DNA binding domain of TBX5 was performed according to the manufacturer’s protocol (Qiagen).

To generate mouse T protein we used an expression plasmid in which the Brachyury coding sequence was cloned downstream of the T7 promoter in pCRTM3 (a gift from Jane Sowden, Institute of Child Health, London, UK).

TBX5 binding site selection

Binding site selection was performed according to Pollock and Triesman (36). Briefly, a 26mer random oligonucleotide core flanked by defined primer sequences F, 5′-GCTGCAGTTGCACTGAATTCGCCTC-3′ and primer R, 5′-CAGGTCAGTTCAGCGGATCCTGTCG-3′ was rendered double-stranded by filling with Klenow using primer F. 1 pmol of the double-stranded oligonucleotide was mixed with 20 ng of TBX5-His6 in 20 µl of binding buffer BB (20 mM HEPES pH 7.9, 20 mM NaCl, 2 mM MgCl₂, 1 mM DTT, 1 µg poly dI-dC.poly dI-dC, as non-specific competitor, 1 mM PMSF, 1 µg/ml leupeptin, 1 µg/ml pepstatin and 10 µg/ml BSA), containing 1 µl of mouse monoclonal anti-His-antibody (Qiagen). The mixture was incubated at 30°C for 30 min. The DNA–protein complexes were adsorbed on 0.2 µg of anti-mouse Ig-Sepharose (Sigma) at 4°C overnight with gentle shaking. The solid agarose matrix was pelleted and washed four times with 500 µl binding buffer at room temperature. The bound DNA was eluted with extraction buffer (100 mM NaOAc, 50 mM Tris pH 8, 5 mM EDTA and 0.5% SDS) at 45°C for 10 min. The eluted DNA was purified by phenol/chloroform extraction and ethanol precipitated following the addition of 10 µg of glycogen. The precipitate was finally reconstituted in 50 µl 25 mM Tris buffer (pH 7.4). One-tenth of the eluate was PCR amplified (15 cycles of 94°C for 1 min, 62°C for 1 min, 72°C for 1 min) using Pfu polymerase (Stratagene) and amplified products were checked on a 2% agarose gel. An aliquot (0.5 pmol DNA) of the sample was subjected to two more rounds of selection and amplification. After the final round of selection, DNA from each selection round was labelled by PCR amplification (10 cycles) using [³²P]dCTP and checked for binding. The oligonucleotides from the third round of selection were cloned into the EcoRI and BamHI sites of pBluescript-SK(+). Following selection on IPTG/X-gal plates colonies were picked and DNA was prepared and sequenced on an ABI automated DNA sequencer.

EMSA

DNA from the oligonucleotide pool, prior to selection amplification, and oligonucleotides from the first, second and third rounds of selection were labelled by PCR amplification (10 cycles). To check the binding affinity of individual target clones, obtained from the binding site selection experiments, inserts were released by EcoRI and BamHI digestion, separated on 2% agarose gel and purified using a QIAquick gel extraction kit (Qiagen). These fragments were dephosphorylated using calf intestinal alkaline phosphatase (Gibco BRL) and end labelled. The TBX5 consensus-binding site was generated by annealing the oligonucleotide 5′-GGGAAGGTGTGACCC-3′ to the complementary sequence 5′-GGGTCACACCTTCCC-3′. The full-length Brachyury binding site was obtained by self-annealing the oligonucleotide 5′-GGGAATTTCACACCTAGGTGTGAAATTCCC-3′ and T1/2 site was generated by annealing the oligonucleotide 5′-GGGAGGTGTGAAATTCCC-3′ to its complementary oligonucleotide. To generate the putative wild-type TBX5 target corresponding to that present in the upstream of the ANF gene (–272 to –249) and a mutant version of the same sequence, the oligonucleotides 5′-CTCTTCTCACACCTTTGAAGTGGG-3′ and 5′-CTCTTCTCATCCCTTTGAAGTGGG-3′ were annealed to their complementary oligonucleotides, respectively. Annealed oligonucleotides were end-labelled using a ‘Ready To Go’ T4 polynucleotide kinase kit (Pharmacia) and the labelled products were purified on a Sephadex G-25 spin column (Pharmacia). To examine the binding of TBX5 to the various putative target sites, double-stranded DNA samples (0.1–0.2 pmol) were incubated with 20–200 ng of TBX5-His6 in binding buffer A (20 mM HEPES, 20% glycerol, 50 mM NaCl, 5 mM MgCl₂, 0.5 mM EDTA, 2 mM DTT and 1 mg/ml BSA) containing 1–2 µg of poly (dI-dC). poly (dI-dC) in a 20 µl volume. The reaction mixtures were incubated at 30°C for 30 min. For antibody supershift assays, 0.5–2.0 µl of monoclonal anti-His-antibody, or polyclonal TBX5 antibodies, were added to the tubes and further incubated for 10 min. For competition experiments, 10–100-fold excess of cold double-stranded oligonucleotides was added to the reaction mixture. All the samples were resolved on 5% polyacrylamide gels electrophoresed in 0.5× TBE at 10–15 V/cm for 2–3 h. After electrophoresis the gels were dried and subjected to autoradiography.

In vitro coupled transcription-translation

Reactions were performed using a transcription/translation coupled reticulocyte lysate system (Promega) according to the manufacturer’s protocol. 1.0–2.0 µg of template DNA was incubated at 30°C for 90 min in a 50 µl reaction volume containing 1 µl of T7 RNA polymerase and 80 U RNasin. For radioactive synthesis, 2 µl of [³⁵S]methionine (1000 Ci/mmol, at 10 mCi/ml, Amersham) was added along with amino acid mixture minus methionine. Immediately after incubation, PMSF (1 mM), leupeptin (1 µg/ml), pepstatin (1 µg/ml) and aprotinin (1 µg/ml) were added to the reaction mixture. Synthesis of appropriately sized proteins was analysed by SDS–PAGE and autoradiography. 1–5 µl of the reaction mixture was used in each assay.

Protein truncation and oligonucleotide directed mutagenesis

Various cDNAs encoding full-length and truncated forms of TBX5 were amplified from pcT5.2 DNA and cloned into pcDNA3.1 at the NheI and EcoRI sites. The primers used for these amplifications were TKGF7, 5′-CATATGGCTAGCACCATGGCCGACGCAGACGAGGGC-3′; TKGR9, 5′-AGGTCCGAATTCTCAAAATCCTTTGGCAAAGGGATTATTC-3′; TKGR7, 5′-AGGTCCGAATTCCTACCGAAATCCTTTGGCAAAGGGATTATTC-3′; TKGR8, 5′-AGGTCCGAATTCCTAGTCATCACTGCCCCGAAATCCTTTGGC-3′; TKGF8, 5′-CATATGGCTAGCACCATGATCAAAGTGTTTCTCCATGAAAGAGAACTG-3′; TKGR4, 5′-TCTAGAGGATCCTTAGCTATTGTCGCTCCACTCTGG-3′; TKGF9, 5′-CATATGGCTAGCACCATGCTGTGGCTAAAATTCCACGAAGTGGGC-3′; and TKGR10, 5′-AGGTCCGAATTCTCAAGACTCGCTGCTGAAAGGACTGTG-3′.

The combinations of primer pairs used to amplify the various products were TKGF7 and TKGR4 for full-length TBX5 (generating clone pcDNA-TBX5), TKGF7 and TKGR9 for TBX5(1–236), TKGF7 and TKGR7 for TBX5(1–237), TKGF7 and TKGR8 for TBX5(1–241), TKGF7 and TKGR10 [generating clone pcDNA-TBX5(1–279)] for TBX5(1–279), TKGF8 and TKGR4 for TBX5(54–518), and TKGF9 and TKGR4 for TBX5(63–518).

cDNAs encoding mutated forms of full-length TBX5 were generated using the ‘Quick Change’ site-directed mutagenesis kit (Stratagene). pcDNA-TBX5 or pcDNA-TBX5(1–279) was used as the template to generate TBX5G80R with primers G80R (forward and reverse), TBX5G169R with primers G169R (forward and reverse), TBX5R237Q with primers R237Q (forward and reverse), and TBX5S252I with primers S252I (forward and reverse). The oligonucleotide pairs used for these PCR amplifications were:

G80R forward, 5′-G ATC ATA ACC AAG GCT AGA AGG CGG ATG TTT CCC-3′;

G80R reverse, 5′-GGG AAA CAT CCG CCT TCT AGC CTT GGT TAT GAT C-3′;

G169R forward, 5′-CAC CTG GAC CCA TTT AGG CAT ATT ATT CTA AAT TCC-3′;

G169R reverse, 5′-GGA ATT TAG AAT AAT ATG CCT AAA TGG GTC CAG GTG-3′;

R237Q forward, 5′-C TTT GCC AAA GGA TTT CAG GGC AGT GAT GAC ATG GAG-3′;

R237Q reverse, 5′- CTC CAT GTC ATC ACT GCC CTG AAA TCC TTT GGC AAA G-3′;

S252I forward, 5′-GA ATG TCA AGA ATG CAA ATT AAA GAA TAT CCC GTG GTC-3′; and

S252I reverse, 5′-GAC CAC GGG ATA TTC TTT AAT TTG CAT TCT TGA CAT TC-3′.

The altered bases are underlined and encode the mutated amino acids. The temperature cycles used for these PCR amplifications were 95°C for 30 s followed by 15 cycles of 95°C for 30 s, 55°C for 1 min and 68°C for 13 min. After the final cycle all the tubes were kept on ice for 5 min and subjected to DpnI digestion, which specifically cleaved the parental strands. The digested DNAs were transformed into E.coli XL1-blue super competent cells and transformants were selected on LB ampicillin-agar plates. Plasmid DNA was prepared and sequenced using an ABI automated DNA sequencer to check that mutant sequences had been incorporated correctly and that reading frames had been maintained.

Reporter constructs and transient transfection assays

A DNA fragment (–3 to –337) from the ANF promoter was PCR amplified and inserted into the KpnI and SacI sites of pGL3 basic vector to generate the wild-type ANF reporter plasmid pGL3-WtANF. Site directed mutagenesis was employed to delete the TBX5 binding site at position –266 to –259 of the wild-type reporter with primer pairs 5′-CCA AGG ACT CTT TTT TAC TTT CGA AGT GGG AGC CTC TTG-3′ and 5′-CAA GAG GCT CCC ACT TCG AAA GTA AAA AAG AGT CCT TGG-3′ using a commercial kit (Stratagene).

Transfection studies were carried out in rat cardiomyocyte cell line H9c2. Cells were grown in 60 mm dishes containing Dulbecco’s modified Eagle’s medium (DMEM) with 10% fetal calf serum. Cells were transfected at 60–80% confluence with polyfect reagent (Qiagen) and 1.5 µg of reporter plasmid (pGL3-WtANF or pGL3-ΔANF), 1.0 µg of expression plasmid pcDNA-TBX5 or pcDNA3.1, and 4 ng of Renilla control, pRL-TK. The total amount of DNA (3.0 µg) was kept constant in each dish by adding an appropriate amount of pcDNA3.1, where necessary. Cells were harvested and lysed 24 h after transfection. Luciferase assays were performed using a Dual-Luciferase assay kit (Promega) according to the manufacturer’s protocol. Renilla luciferase was used to normalize transfection efficiency between dishes. Four separate transfection sessions were employed and at each session transfections were performed in triplicate and three luciferase/renilla readings were taken for each transfection.

In situ hybridization

Whole-mount in situ hybridization was performed according to the protocol described in Cheung et al. (48).

ACKNOWLEDGEMENTS

This work was supported by the British Heart Foundation.

+

To whom correspondence should be addressed. Tel: [phone]; Fax: [phone]; Email: [e-mail]

Figure 1. The TBX5 consensus-binding site. (A) Alignment of sequences from 41 clones to derive the primary consensus-binding site sequence. (B) Alignment of sequences from 17 clones to produce the secondary consensus-binding site. Bases present >10% of the time were considered to comprise the consensus. Some of the bases in some of the cores were derived from the primer sequences flanking the random 26 bases of the oligonucleotides. The bases from the primers were not included when deriving the consensus-binding site sequence.

Figure 2. Binding studies on TBX5. (A) EMSA study on the inserts from six clones; T-1, T-26, T-11, T-18, T-33.2 and T-31. S, slow mobility complex; F, fast mobility complex; asterisk, complex formed from degraded TBX5. (B) The sequences of the six clones analysed by EMSA. (C) Mixed full-length and truncated TBX5 form an intermediate complex on fragment T-11. Lane 1, probe fragment T-11 only; lane 2, probe with TBX5; lane 3, probe with truncated TBX5 (residues 1–237); lane 4, probe with both TBX5 and truncated TBX5; lane 5, as lane 4 with 2-fold increase in truncated TBX5. Arrows indicate various complexes: I, slow mobility complex; II, intermediate complex; III and IV, fast mobility complexes with truncated TBX5.

Figure 3. (A) Binding studies of bacterially expressed TBX5 on the mouse Brachyury consensus DNA binding site. BS-T is the full-length Brachyury palindrome probe and BS-T/2 is the half palindrome probe. Lane 1, probe only; lane 2, probe with TBX5; lane 3, probe with TBX5 and antibody to TBX5; and lane 4, probe with antibody to TBX5 only. (B) Binding studies on the TBX5 consensus-DNA binding site. A 15 bp oligo probe (BS-TBX5) was used for each lane. Lane 1, RRL, rabbit reticulocyte lysate control. Lane 2, INV-TBX5, in vitro translated TBX5 protein. Lane 3, BE-TBX5, bacterially expressed TBX5. Lane 4, M-T, in vitro translated mouse T protein.

Figure 4. (A) Dissection of the TBX5 DNA binding domain. Oligo T-11 was used as probe in each lane. Lane 1, RRL, rabbit reticulocyte lysate control. Lane 2, BE-TBX5, bacterially expressed TBX5. Lane 3, INV-TBX5, in vitro translated TBX5. Lanes 4–8, in vitro translated truncated forms of TBX5 protein as indicated on the top of each lane. (B) The effect of mutations in TBX5 on its DNA binding activity. Oligo T-11 was used as a probe in each lane and tested against in vitro translated versions of truncated (1–279)TBX5. Lane 1, RRL, rabbit reticulocyte lysate control. Lane 2, truncated (1–279)TBX5. Lane 3, truncated (1–279)TBX5 containing a G80R substitution. Lane 4, truncated (1–279)TBX5 with a G169R substitution. Lane 5, truncated (1–279)TBX5 with an R237Q substitution. Lane 6, truncated (1–279)TBX5 with an S252I substitution. The asterisk denotes a non-specific complex.

Figure 5. (A) TBX5 recognizes a putative binding site within the promoter region of ANF. Lane 1, mutant ANF probe only. Lanes 2–4, mutant ANF probe with increasing concentration of TBX5. Lane 5, wild-type ANF probe only. Lanes 6 and 7, wild-type ANF probe with increasing concentration of TBX5. Lane 8, wild-type ANF probe, TBX5 and competition with wild-type ANF oligo. Lane 9, wild-type ANF probe, TBX5 and competition with mutant ANF oligo. Lane 10, wild-type ANF probe, TBX5 and anti-TBX5 antibody. Arrow indicates supershift. (B) TBX5 activates transcription from the ANF promoter. pGL3-WtANF contains bases –337 to –3 of the wild-type ANF promoter in vector pGL3, and pGL3-ΔANF is the same construct but with the TBX5 site at position –266 to –259 deleted. pcDNA-TBX5 contains the TBX5 coding sequence in expression vector pcDNA3.1. Luciferase activity following transfections with reporter constructs (pGL3-WtANF or pGL3-ΔANF) and control vector (pcDNA3.1) is set at 1. Other values are shown relative to this. The bars represent an average of four independent experiments in which each transfection was carried out in triplicate. The values shown represent means ± SD. (C) Whole mount in situ hybridization showing expression of mANF and Tbx5 in both atria and the left ventricle. (i) 12.5 d.p.c. mouse heart probed with mANF. (ii) 13.5 d.p.c. mouse heart probed with Tbx5.

Table 1.

TBX5 binding sites are present in the upstream regions of cardiac-expressed genes

Gene	Accession no.	Length	Total sitesDGGTGYBVa	RGGTGTBRa	Double sites
ANF	K02043	473	2	1	0
MYH6	Z20656	4484	15	5	1
MYH7	M57965	3803	8	1	1
Nkx2.5	AF091351	3057	3	1	0
Mlc1a	X12971	1311	2	2	1
Mlc1v	X12972	1298	4	0	0

^a‘D’ represents A or G; ‘R’ represents A, G or T; ‘Y’ represents T or C; ‘B’ represents T, G or C; ‘V’ represents A, G or C.