METHODS

Patients

Our initial study group (P1) consisted of 78 white patients with 22q11.2 deletion, of whom 66 had congenital heart defects, and who were recruited from the paediatric cardiology centres of Erlangen and Tuebingen (southern Germany) (HP1); and 12 patients with normal hearts or with clinically non-relevant persistent arterial duct or persistent foramen ovale only, who were recruited from the Erlangen genetic clinic (NP1). Our confirmatory study group comprised 96 patients with 22q11.2 deletion recruited from the medical centre of Leuven, Belgium (P2), of whom 57 had a congenital heart defect (HP2) and 39 did not (NP2) (table 1). Our normal control group consisted of 96 healthy, dizygous individuals of north European origin.

The study was approved by the University of Erlangen-Nuremberg ethics review board and appropriate informed consent was obtained from human subjects.

Procedures

Sequencing was undertaken after polymerase chain reaction (PCR) amplification with intronic primers of all coding exons from the three different splice variants of TBX1, and in addition of the 5′UTR exon 1 on capillary sequencers (ABI 3100, 3730). Primers for exons 2 (“1F, 1R”) and 9a (“7F, 8R”) were used, as reported by Chieffo et al.¹⁹ Primer sequences for exons 4, 5, 6, 7/8, and 9c were kindly provided by Peter Scambler. Primers for exons 1, 3, and 9b were generated with primer 3²⁰ and are as follows: tbx1aF-GAGCAGATGTCTCAGCCCAG, tbx1aR-CCACACTCCTCTTCACCTGC, tbx3bF-CACGCAGCTCTCGCATTTCT, tbx3bR-GATAGGTGTTAGGAGGGGAG, tbx9bF-AGGCCACAAACACTTTGACC, and tbx9bR-TTCCATCACAGCCTCTTCAC. For P2 and normal controls, exons 2, 4, and 9a were co-amplified with a multiplex PCR kit (Qiagen, Hilden, Germany). Sequencing results were compared with the sequence from cosmid clone 91c (AC000091), now included within the RefSeq sequence NM_005992. In dizygous controls, haplotype structure was determined using PHASE.²¹ Linkage disequilibrium (LD) was calculated with ld_max using an expectation-maximisation algorithm²² and visualised with GOLD.²³ Frequencies were compared by χ² testing (Pearson) or by Fisher’s exact test.

RESULTS

In study group P1 we detected 14 different sequence variants, seven of which had a frequency of 0.10 and above (table 2). All but the ΔAGG379–381 and V107V variants were detected in similar frequencies in normal controls. The ΔAGG379–381 variant was detected in a baby girl with a large subpulmonary ventricular septal defect (VSD), who inherited the 9 base pair (bp) deletion from her father, who also had a VSD but no 22q11.2 deletion. The V107V variant was found in a boy with pulmonary atresia-VSD; he inherited the deletion 22q11.2 from his mother, who had neither a congenital heart defect nor a TBX1 variant. The father of this patient was not available. The V107V variant is located four nucleotides 5′ to the T-box and turned out to be the only one that affects a conserved nucleotide and amino acid (conserved in Mus musculus and Xenopus laevis, but not in Danio rerio or Drosophila melanogaster). The boy with the V107V variant is of local Franconian origin, as were the majority of the other patients in study group 1 and the normal controls.

As the haplotypes were naturally provided in these hemizygous individuals, we initially simply determined haplotype frequencies, grouping all seven common SNPs together. This showed that only five common haplotypes accounted for 93% of the chromosomes within our patient group P1 (fig 1), which would be within the commonly observed numbers for single haplotype blocks.²⁴ To enhance the statistical power we analysed a further 96 hemizygous patients in whom we detected two further SNPs, and 96 dizygous healthy controls. Increasing the haplotype information from 78 to 174 hemizygous individuals reduced the incidence of the five common haplotypes to 85.6% of chromosomes. Reassessment of LD structure revealed a total of three LD blocks within the coding region and one starting with the 5′UTR region of exon 1 (table 3, figs 2 and 3).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Position of rare variants on background haplotypes constructed from all observed single nucleotide polymorphisms (SNPs). The order of the SNPs is according to table 2 from exon 1 to 9a. The frequency of the common haplotypes within different patient groups is given in per cent, alleles of common SNPs (>0.09) are written with bold capitals, wild type alleles of rare variants are written with small letters in the haplotype line. Numbers at the right side represent the number of patients in each group in whom a certain rare variant was observed on the respective background haplotype. As control persons are dizygous, both background haplotypes of each individual are given in parentheses.*Same individual. v, Variant of background haplotype.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

GOLD graphical overview of linkage disequilibrium for seven common TBX1 single nucleotide polymorphisms with naturally given haplotype data from (left) 78 and (right) 174 hemizygous patients.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Schematic overview of linkage disequilibrium (LD) organisation of TBX1. LD blocks are separated by black crosses. Common SNPs (frequency >0.09) are in bold. Exon 1 (5′UTR) belongs to LD block I of unknown size, exon 2 (5′UTR) belongs to an LD block with maximum size of about 3 kb, exons 3 to 9c—covering at least 6 kb—constitute the core LD block III, with three common haplotypes accounting for 98.4% of 366 chromosomes; and exon 9a represents LD block IV of unknown size with at least three common haplotypes accounting for 100% of 366 chromosomes. Haplotypes of defined blocks were each observed in combination with every haplotype from neighbouring blocks, but most frequent combinations are equally coloured. Block II haplotype 2 was only observed once in 366 chromosomes in combination with haplotype 3 from block III.*Haplotype tag SNP.

Comparison of the frequencies of single SNPs and common haplotypes between groups P1 and P2, showed no significant difference; we therefore concluded that we can merge P1 and P2 without bias from ethnic background. There was no significant difference in frequencies of variants or common haplotypes between group HP (123 combined patients with 22q11.2 deletion and congenital heart defects) and group NP (51 combined patients with 22q11.2 deletion without relevant congenital heart defects) (tables 2 and 4). As the majority of patients were young children recruited through cardiology, the frequency of other malformations such as cleft palate was too low to undertake association studies. The same holds true for later onset symptoms such as psychiatric disease, which could not be assessed in this patient group.

DISCUSSION

We detected 16 sequence variants within the remaining TBX1 copy in 174 deleted patients, of which only three were recorded in dbSNP; thus seven turned out to be common, with a frequency of 0.10 and above (table 2). None of the seven common or the nine rare variants with a frequency of <0.10 showed a significant difference between deletion patients with and without congenital heart defects or the normal control group, respectively. However, two of the rare variants—1135–1143delgccggcggc (ΔAGG379–381) and 321g-a (V107V)—were not detected in patients without congenital heart defects or in healthy controls. Thus it is likely that the 9 bp deletion at least contributes to the heart phenotype, as it segregated with VSD in one family. The 321g-a (V107V) variant turned out to be the only one that affects a conserved nucleotide and might therefore also represent a relevant mutation, as synonymous changes can affect splicing. As TBX1 is not expressed in blood, splicing could not be tested in this patient. Nevertheless, even if the 9 bp deletion and V107V, respectively, are considered relevant contributors to the expression of congenital heart defects, TBX1 variants would still account for only 1.6% of cases of congenital heart disease in 22q11.2 deletion patients. However, functional studies are necessary to confirm the impact of these rare variants.

To further evaluate whether there were haplotypes or common non-coding sequence variants not detected in our scan that could act as modifiers, we investigated whether certain haplotypes might influence the expression of the heart phenotype. As the haplotype structure of the TBX1 locus had not been analysed before, we first had to determine LD blocks. Definition of LD blocks aims to identify genomic regions where a few unique haplotypes are distributed widely throughout the population, but the methods are still subjective; some approaches make use of known or inferred haplotypes, others use pairwise LD measures, and yet others use a combination of the two.^25,26 In contrast to analysis of haplotype structure in dizygous individuals, haplotype construction in our patients is not hampered by likelihood assumptions as they are hemizygous, thus providing naturally given haplotype information. Although in our initial study group, grouping of all seven common SNPs together into one LD block resulted in only five common haplotypes, accounting for 93% of chromosomes, expanding the patient group from 78 to 174 hemizygous individuals reduced the incidence of the five common haplotypes to 85.6% of chromosomes, which suggests the existence of more than one LD block.²⁴ This observation is in line with the recommendation that at least 100 chromosomes should be used in haplotype block studies.²⁷ Moreover, with the “one LD block only” definition, several recurrent rare variants would have occurred on different haplotypes (fig 1), which we assumed was unlikely—especially for variants IVS7+31c-t and A353A, which always occurred together. When we postulated that the combination of these two SNPs appeared only once, the core ancestral haplotype block would only consist of four common SNPs (fig 3). Under this assumption, all but one rare variant would have occurred only on one haplotype each, and D′ values would be 1.0 for each SNP pair comparison within the blocks (table 3). By this approach only three different haplotypes occur within each LD block, so our definition is in line with the four gamete test definition.²⁷ Despite the complete pairwise LD for block III, SNP A99A occurred eight times on haplotype 1 and once on haplotype 2. However, as this SNP affects an CpG doublet, an independent origin is likely. As D′ values confining haplotype blocks are still arbitrary and differ widely in different studies,^28,29 it is noteworthy that common SNP 6 (T350M), which does not belong to the core block III under our definition, is also in complete LD with all but one SNP (3 = L222L) from this block (table 3). Our study also shows that PHASE is able to produce the same haplotype distributions by likelihood calculation in dizygous individuals as in our naturally phased hemizygous patients.

As haplotype distribution was equal for all groups (table 4), we were able to exclude an influence of undetected intronic variants residing on common haplotypes of TBX1 on expression of the heart phenotype in patients with 22q11.2 deletion. Thus our results show that variants within the TBX1 region are not common modifiers of heart phenotype in patients with deletion 22q11.2. However, as we found that the 5′ promotor region is located on a different haplotype block, it is still possible that common variants within the TBX1 promotor region, recently shown to span about 15 kb,³⁰ could influence the manifestation of congenital heart defects in DGS/VCFS patients.

As a strong candidate for a major contributor to the DGS/VCFS phenotype,^{15–18,31–33}TBX1 has been analysed previously for mutations in non-deleted DGS/VCFS patients. Splice variant A (including alternative exon 9a) has been sequenced in 16 patients, but only five of seven common polymorphisms were identified.¹⁹ Splice variant C (including alternative exon 9c) was analysed in 105 non-deleted patients, of whom 40 had clinical features of DGS/VCFS and 65 had isolated cardiac lesions frequently seen in DGS/VCFS.³⁴ In the latter study, eight common polymorphisms were reported, of which six are identical with four common and two less common variants, with the frequencies of 0.04 to 0.05 seen in our study. However, it is likely that the two further variants reported as −75g-c and IVS7+46 are probably also identical with our common SNPs −85g-c and IVS7+31c-t, respectively. This assumption is supported by the fact that in our study IVS7+31c-t always occurred together with A353A, which also seems to be the case for A353A and “IVS7+46”, because they appeared with exactly the same incidence. Gong et al also detected 10 rare variants, including amino acid exchanges, deletions, and duplications—nine of which were not seen in 101–139 controls.³⁴ However, the clinical relevance of these rare variants is not proven as all patients whose parents were available inherited the rare variant from a healthy parent. Conti et al analysed all three different TBX1 splice variants in 41 patients with non-syndromic cono-truncal heart defects for mutations by SSCP analysis, and detected five of our common SNPs as well as two novel variants (exon 9c 1022c-a, normal population frequency 1%; exon 10B 3′UTR mRNA position 1486c-a, normal population frequency 10.5%).³⁵ These investigators concluded that with a multifactorial model for congenital heart defects they cannot rule out the possibility that these rare or common variants are somehow involved in the pathogenesis of the patients’ heart malformations. However, they made no attempt to carry out an association study.

Our nine rare variants were not seen in 162 non-deleted patients analysed by the other groups, and we did not see any of the 10 rare variants reported by them³⁴ in our 174 hemizygous patients. Although this might just reflect the low incidence of these variants, we cannot exclude the possibility that, in particular, the frameshift variants detected in non-deleted DGS/VCFS/congenital heart defects patients could be lethal in our hemizygous patients, as early lethality has been shown for double mutant mice.^15–17

With respect to further association studies our results show that exon 3 to 9c, covering at least 6 kb, constitute a single core LD block, with three common haplotypes accounting for 98.36% of 366 chromosomes (block III), while exons 1, 2, and 9a are parts of different LD blocks (I, II, and IV, respectively). Thus typing of at least five haplotype tag SNPs is necessary to cover the common haplotype structure of the 27 kb region of TBX1, which is in line with the assumption from a first chromosome 22 LD map with 15 kb SNP spacing that the DGS/VCFS region shows low LD, and thus fine mapping for any association might require a high density of markers.²⁸