Editor—Holt-Oram syndrome (HOS) is a developmental disorder characterised by malformations of the radial ray of the forelimb and by congenital heart disease.1 The syndrome shows a marked variability in phenotype, with radial ray defects ranging from minor thumb abnormality through to severe reduction defect or phocomelia. The cardiac manifestations of HOS are similarly varied, and patients can present with a variety of structural heart abnormalities, atrial septal defects (ASDs) and ventricular septal defects (VSDs) being the most common, or conduction defects evident on ECG profiles. Previous studies have shown no correlation between the severity of a patient's cardiac and skeletal abnormalities.2 Intrafamilial variation can be wide.

HOS shows autosomal dominant inheritance and mutations in the T box transcription factor gene (TBX5) have been shown previously to be responsible for this disorder.3 4There is also evidence for genetic heterogeneity.5 The mechanism by which mutations in TBX5 cause a dominant phenotype is not understood at present, and it is anticipated that knowledge of the type of mutations causing HOS may shed light on this. Knowledge of a large number of mutations and the relation of a person's genotype to phenotype is also useful for genetic counselling. In the face of a growing demand for a molecular diagnostic test for HOS, it is also helpful to have a quantitative estimate of the ability of current methods to detect mutations inTBX5.

Twenty five cases with a clinical diagnosis of Holt-Oram syndrome have been tested for this study, bringing to 47 the total number of cases studied by us. Minimal diagnostic criteria were as described previously2: bilateral radial ray defect, plus either cardiac abnormality or family history of cardiac abnormality. Cases were referred by a variety of clinicians and underwent full clinical assessment including x ray, electrocardiography (ECG), and echocardiography. Information regarding the clinical features of these patients came from the referring doctor. The patients represented both sporadic (8) and familial (17) cases of Holt-Oram syndrome.

Mutational analysis was carried out using SSCP methods initially, followed by fluorescence sequencing of exons showing a non-standard SSCP banding pattern. The methods used were described in Liet al.3 Since that study, further analysis of the genomic structure ofTBX5 has recognised that the previously reported exon B is in fact two exons. Extensive resequencing of both genomic and cDNA forms of TBX5 has also been undertaken, leading to revisions in the previously reportedTBX5 sequence (the new sequence has accession number AF221714). This sequence is the same across the coding region as that produced by others.6 Of 17 new familial cases tested, eight showed linkage to chromosome 12 and the other families were too small to assess linkage. Table 1 shows that five mutations were identified in familial cases. Only one of these families was sufficiently large to show meaningful linkage to chromosome 12q markers. Of eight new sporadic cases studied, three have yielded mutations. Thus, in the 34 familial cases studied by us, eight mutations have been identified, and six mutations have been identified in 13 sporadic cases. The precise natures of all the new mutations identified are detailed in table 2.

The clinical features of HOS in all 47 cases are consistent with the previously described phenotype2 and show the wide spectrum of cardiac and skeletal abnormalities in this syndrome (see Bruneauet al 7 for details of the complexity of cardiac abnormalities in HOS patients). Most patients show at least one defect of cardiac septation (an atrial or ventricular septal defect, or atrioventricular block) and abnormalities of the thumb. Radial hypoplasias and aplasias are present in sporadic cases PpHs, H20s, H22s, and H16s and in the familial cases H6f and Ghf, although not in every affected member.

Twenty five non-translocation mutations have been reported, including those presented here. These mutations are of 19 distinct types, with six mutations being identical to previously described forms identified in unrelated subjects (this paper and Bassonet al 6). Of the 19 distinct non-translocation, disease causing mutations inTBX5 currently known, five are truncations, five amino acid substitutions, three splice site changes, and six reading frame shifts.

We have observed a significant difference in the proportion of cases in which a mutation was detected in our group of sporadic cases as opposed to our group of familial cases. Forty six percent (6/13) of sporadic cases studied in our laboratory have yielded mutations by SSCP screening, whereas only 24% (8/34) of familial cases have done so.

The overall mutation detection rate of 30% may be low for a variety of reasons. An inadequate mutation detection method would explain these results, yet the system in use by us is a standard one used on a variety of projects, all of which produce detection rates nearer the theoretical level for these techniques (approximately 95%). The largest PCR product used in the present analysis is only 326 bp long, well within the size acceptable for this kind of analysis. Mutation detection is also fully repeatable in our hands.

There are four other possible explanations for this low mutation detection rate: (1) clinical misdiagnosis in our patients, (2) genetic heterogeneity of Holt-Oram syndrome,5 (3) the presence of mutations in the untranslated and promoter regions ofTBX5, which have not been tested in this analysis, and (4) deletion of whole exons ofTBX5, which would not be recognised by SSCP.

Other studies of HOS have not published mutation detection rates. We can only speculate as to why the rate of mutation detection in sporadic cases should be higher than that in familial cases (particularly in those familial cases whose disease loci are known to be present on 12q). An obvious explanation for such a discrepancy is ascertainment bias. A sporadic case must have both detectable heart and limb symptoms to be diagnosed as having HOS, whereas a family need only show both these effects across the pedigree (rather than in one subject) along with an autosomal dominant inheritance pattern.

A group of five of the 25 mutations published so far are the same, a C to T transition at position 1500, generating a stop codon. This mutation is seen in both sporadic and familial HOS cases among patients who do not share common alleles at microsatellite loci closely linked to TBX5. This site is likely to represent, therefore, a true “mutation hotspot”. The residue is part of a CG duplet, and therefore is likely to be methylated, with a high frequency of mutation to a thymine residue. Studies across a variety of human diseases have found distributions of mutations skewed towards the mutation of CpG sites. The retinoblastoma gene,RB1, shows a distribution of mutations severely skewed towards a few C→T transition hotspots (see theRB1 mutation database8 for details).

Basson et al 6 argue, based on a collection of HOS families of varying symptom severity, that there is a relationship between the phenotype of patients and their specific mutation. It is proposed that truncation mutations inTBX5 lead to both limb and cardiac malformation, whereas single amino acid changes have different effects depending on their position in the T box. The set of HOS cases we have examined contains no large families of the type studied in this earlier analysis.

The five known cases (two familial and three sporadic) with the same mutation, a C→T transition at nucleotide 1500 in exon 8, present an opportunity to examine more closely the possibility of mutation specific genotype-phenotype correlation in cases with a truncation inTBX5. The clinical phenotypes of these cases are presented in table 3. Cardiac defects presented include isolated ASD and isolated VSD but no case with this mutation has a complex cardiac lesion. Syndactyly of the thumb and first finger is common within this group. There is much variation in symptom severity and the group as a whole shows no bias towards a particular severity of either cardiac or skeletal symptoms, in agreement with Bassonet al.6

Along with the truncation forms already described, we have identified two new mutations which each result in a single amino acid substitution. The change in family Fif inserts an arginine in place of a glycine at position 169, which is conserved across T box genes such as Xbra(Xenopus), T(mouse), TbxT (chick), andomb(Drosophila).9 This introduces a strongly basic residue into a non-polar region in the DNA binding T domain. Family Fif comprises eight affected subjects in two generations, who show significant cardiac involvement compared with very mild skeletal findings. One case has a complex lesion, ASD with VSD, and another case has pulmonary stenosis, a conotruncal malformation which is not typical in HOS. This is consistent with a role for TBX5 which extends beyond cardiac septation.10 Only one subject in family Fif has a demonstrable limb abnormality and this is stiffening of the thumbs. The phenotype of this family is therefore consistent with the suggestion that substitution mutations produce predominantly limb or predominantly cardiac features, depending upon their location withinTBX5. 6

The change of a serine to an isoleucine in patient Chas is outside the T domain of the protein and its biochemical effects are not known. The phenotype of this patient included a spinal scoliosis, which has not previously been observed in Holt-Oram syndrome, together with bilateral hypoplastic thumbs, syndactyly, and a ventricular septal defect. Currently available details on expression ofTBX5 during development of the mouse and chick give no evidence of expression in the developing spine which would account for such a phenotype being the result of mutation ofTBX5. 12 13

In summary, these new data expand our knowledge of the spectrum of mutations that cause Holt-Oram syndrome and also raise interesting questions about the genetic heterogeneity of this disease and its mutations. Clearly there is a need to improve the frequency of mutation detection in HOS and current analysis of untranslated and promoter regions, and screens for whole exon deletions should prove useful. A diagnostic service for TBX5 mutations is being set up.13