Introduction

Fragile X syndrome is the most common cause of inherited mental retardation, with an incidence of about 1 in 1,500 males and 1 in 2,500 females. It is associated with a rare, fragile site at Xq27.3 (FRAXA). In males, the syndrome is associated with moderate to severe mental retardation, typical facies, post-pubertal macroorchidism and a folate-sensitive fragile site. Phenotypic expression has been linked to abnormal cytosine methylation of a single CpG island [1–3]. This region contains a repetitive sequence (CGG)n that appears to lengthen dramatically in fragile X patients and has been identified as the first exon of the FMR1 gene [4]. Analysis of length variation in the (CGG) repeat in normal individuals has shown a range of allele sizes extending from 6 to 54 repeats [5, 6]. These normal alleles are stable when segregating in families. Premutations showing no phenotypic effect in fragile X families range in size from approximately 54 to over 200 repeats, and are unstable when segregating in families. Alleles with more than 200 repeats correspond to the full mutation with meiotic and mitotic instability [5]. Expansion of premutations to full mutations only occurs in female meiotic transmission, accounting for the observation that daughters of normal transmitting males (NTMs) never show the fragile X phenotype [2].

Three CA repeats have been described flanking the CGG locus: DXS548 located 150 kb from the CGG repeat on the centromeric side [7], and FRAXAC1 and FRAX-AC2, located 10 kb on the centromeric and telomeric sides, respectively [8]. Analysis of these CA repeats in normal and fragile X chromosomes showed significant differences in allele and haplotype distributions, indicating that a limited number of primary events may have been at the origin of most present-day fragile X chromosomes in Caucasian populations [9–12]. In a recent report [13], the FRAXAC2 locus was shown to be a hypermutable DNA sequence composed of three variable subregions. Inheritance studies showed that this microsatellite was stable in normal families but unstable in fragile-X-derived meioses. However, the mutation rate of this locus does not seem high enough to blur completely the original allele combinations, as was shown by analysis of a BanI RFLP [10] and by the strong linkage disequilibria found on both normal and fragile X chromosomes [9, 10, 12, and this work].

Among the founder chromosomes at the origin of fragile X chromosomes, two are strongly associated with the fragile X syndrome: DX204-AC155, a haplotype composed of the 204-bp allele at the DXS548 locus and of the 155-bp allele at the FRAXAC2 locus, and DX196-AC151, a haplotype composed of the 196-bp allele at DXS548 and of the 151-bp allele at FRAXAC2. To explain the existence of such chromosomes associated with fragile X syndrome, it has been suggested that the stability of normal CGGs may depend on their length or on possible differential interspersion of AGG triplets, as uninterrupted exact repeats are expected to be more unstable [10]. High-risk haplotypes would carry longer CGG and/or more perfect repeats. In order to investigate such hypotheses, we compared the size of the CGG repeat in normal chromosomes carrying high-risk haplotypes with that in other chromsomes. Our findings show a clear-cut difference in the distribution of the CGG repeat in DX204-AC155 chromosomes, for which the larger size of the CGG accounts for the observed intermediate 36–45 repeats range. This result suggests that the transition from the normal (<36) to the abnormal (>54) range occurred by a multistep process, for instance unequal crossing-over followed by DNA polymerase slippage.

Methods

Subjects

Typing of the DXS548 CA repeat was performed on a sample of 554 unrelated French individuals from the general population. Typing of FRAXAC2 and sizing the CGG repeat were performed on selected individuals carrying the expected alleles and haplotypes.

Analysis of CA Repeats

Polymerase chain reaction (PCR) amplifications were performed using the oligonucleotide primers described by Verkerk et al. [4] for DXS548, and by Richards et al. [8], for the FRAXAC2 locus. PCR products were migrated on a sequencing gel, transferred onto a nylon membrane, and hybridized with a 32P-5′-end-labeled oligonucleotide (CA)10. Hybridization was performed at 50 °C for 2 h and the membranes were washed in 2 × SSC/0.1 % SDS for 10 min at room temperature and 5 min at 50 °C.

Sizing of CGG Repeats

PCR of the CGG repeats was performed according to Fu et al. [5], except that we used 200 µM 7-deaza-dGTP instead of 50 µM dGTP plus 150 µM 7-deaza-dGTP. PCR products were comigrated with a sequencing reaction of exon 11 of the CFTR gene [14] as size marker. In some cases, PCR of the CGG repeats was performed according to Erster et al. [15]: PCR products were migrated on a sequencing gel, transferred onto a nylon membrane and hybridized with a 32P-end-labeled oligonucleotide (CGG)7. Hybridization was performed at 55°C for 2 h and the membranes were washed in 2 × SSC/0.1% SDS for 10 min at room temperature and 10 min at 55 °C.

Statistical Methods

Analysis of variance (ANOVA) including all three groups was performed and the significance was tested by the F test. The mean sizes in the different groups were compared by a Student t test, after a common estimation of the overall variance.

Results

Of the 554 chromosomes tested with DXS548, 52 (9%) carried the DX204 allele, 66 (12%) carried the DX196 allele and 436 (79%) carried other alleles, mostly DX194. These proportions are similar to those found in a previous study of 188 unrelated normal chromosomes: 9, 14 and 77%, respectively [10]. The 118 chromosomes carrying DX204 or DX196 were analyzed at the FRAXAC2 locus. Finally, 31 chromosomes were identified carrying one of the two selected high-risk haplotypes, DX204-AC155 (3%) or DX196-AC151 (3%), and were subsequently analysed at the CGG locus. These percentages were also not significantly different from those found in a previous study of 153 unrelated normal chromosomes [10].

The sizes of the CGG repeats in these 31 high-risk haplotypes are shown in table 1. Thirty-two chromosomes carrying haplotypes other than DX204-AC155 or DX196-AC151 were also analyzed at the CGG locus.

Table 1 Distribution of the number of CGG repeats in the 63 individuals tested
Full size table

Comparison of the CGG in the three groups by analysis of variance showed strong heterogeneity (F[2,61] = 12.6; p ≪ 10−3). In the 17 DX204-AC155 haplotypes, the mean size of the CGG locus was 38.4 ± 7.6 repeats. This value is significantly higher than the mean size in the 14 DX196-AC151 haplotypes (28.3 ± 5.1 repeats; p < 10−3) or the 32 control haplotypes (29.4 ± 6.7 repeats; p < 10−3). There was no difference between DX196-AC151 and control haplotypes (p ≈ 0.4).

In three chromosomes (individuals 7, 9 and 51), we found CGGs of 50, 55 and 59 repeats, respectively. These values are very close to 54, the normal premutation limit defined by Fu et al. [5]. Two of these chromosomes carried a DX204-AC155 high-risk haplotype, but the other was found in the third group and carried the DX196-AC155 haplotype. This haplotype is rare in the general and fragile X populations (its frequency was found to be approximately 1 and 3%, respectively) and may derive by crossing-over from a DX204-AC155 haplotype previously enlarged at the CGG repeat. However, this haplotype may also derive from another at-risk haplotype by mutation at FRAXAC2 or DXS548.

When possible, the stability of the CGG alleles was examined in siblings of individuals carrying high-risk haplotypes. The CGG alleles were stably inherited except in individual 9, a male carrying a premutated allele (55 repeats) transmitted to his daughter with one more repeat (56 repeats).

Discussion

The existence of a founder effect in fragile X syndrome suggests that an initial genetic event has occurred on a few occasions in human history and has been responsible for the risk of developing further expansion of the CGG repeat. Among the genetic events able to generate instability of CGG and consequently to increase the risk of expansion, the increased size of the repeat was expected to be very likely. Our results show that this is the case in DX204-AC155 haplotypes, where the mean CGG is approximately 10 repeats longer than in other haplotypes. A DX196-AC155 chromosome has been found with a CGG of 59 repeats, a value in the premutation range. This chromosome may be derived by crossing-over from a DX204-AC155 haplotype previously expanded at the CGG repeat. If we assume that this chromosome was a former DX204-AC155 haplotype, 100% of the CGG >35 repeats were found associated with this haplotype. This suggests that a primary genetic event occurred on the DX204-AC155 chromosome and generated a new enlarged CGG allele. Such a genetic event differed from the normal production of a polymorphism (for instance by DNA polymerase slippage), because it disrupted the distribution of alleles by generating a new mode with a dramatically high variance, reflecting great instability (fig. 1). Unequal crossing-over, creating the DX204-AC155 at-risk haplotype, would account for this disruption that has generated a reservoir of large CGG repeats associated with the DX204-AC155 haplotype, from which DNA polymerase slippage could lead to recurrent premutations as well as shorter alleles ranging in the normal size. This intermediate range, from which fragile chromosomes may be derived, was previously postulated on the basis of population data, to reconcile the high incidence of fragile X syndrome with the absence of de novo mutations [16,17].

Fig. 1
figure 1

Distribution of the number of repeats in DX204-AC155 haplotypes and others, including DX196-AC151, in the 63 individuals tested. The clear departure (p < 10−3) between DX204-AC155 (mean size 38.4 ± 7.6 repeats) and other haplotypes (mean size 29.1 ± 6.2 repeats) suggests that a primary event, probably unequal crossing-over, disrupted the bimodal distribution of normal haplotypes. It is also clear that this primary event increased the instability of the repeats, as shown by the spread of the variance.

Full size image

Because DX204-AC155 haplotypes account for only 14% of fragile X chromosomes [10], they cannot be considered to be the reservoir of all the fragile X mutations, as has been shown in myotonic dystrophy (MD) for the alleles carrying CTG repeats ranging from 19 to 30 triplets that represent 78% of MD chromosomes [18]. Moreover, at least six founder chromosomes must be hypothesized to account for present-day fragile X chromosomes [10]. It is thus clear that most of the fragile X chromosomes are derived from ancestral haplotypes other than DX204-AC155.

We observed no CGG with more than 35 repeats in the high-risk DX196-AC151 haplotype or any other haplotype. Two hypotheses may account for this observation. Firstly, the fragile X mutations not associated with DX204-AC155 may result from a one-step process, i.e. an unequal crossing-over leading directly from the normal (<35) to the premutation range. The absence of interspersed AGG triplets in these founder chromosomes could make them so unstable that the size jumps straight from the normal 6–35 range to more than 54 repeats, as has been suggested by the observation that exact repeats are less stable than interrupted repeats [2–4]. Secondly, it is possible that the intermediate range was not observed because of the small number of chromosomes analyzed, especially if the unequal crossing-over generated the high-risk haplotype in a pool of DX196-AC151 haplotypes that preexisted in the normal population. In this case, the high-risk DX196-AC151 chromosomes would be blended in low-risk DX196-AC151 chromosomes, resulting in the nondetection of increased CGGs. In contrast, we were able to document an increase in the CGG in DX204-AC155 because the crossing-over probably created this founder chromosome.

Mutations caused by expanding trinucleotide repeats have also been discovered in MD [19, 20], in Kennedy’s disease [21] and more recently in Huntington’s disease [22] and spinocerebellar ataxia type 1 [23]. Analysis of an insertion/deletion polymorphism close to the MD locus showed a clear-cut difference in the distribution of an allele for which the larger size of the CTG accounts for the observed intermediate 19–30 repeats range [18]. As in fragile X syndrome, this finding also suggests that the transition from the normal (< 18) to the abnormal (>36) range occurred by a multistep process, for instance unequal crossing-over followed by DNA polymerase slippage. In these two repeat-expansion disorders, there is an intermediate status that constitutes a reservoir of further pathological chromosomes. Analyses of linkage disequilibria in Huntington’s chorea, Kennedy’s syndrome and spinocerebellar ataxia I are not yet documented but perhaps they will lead to similar observations. If confirmed, our finding of an intermediate range, subject to further expansion, may lead to the development of more appropriate screening and counseling of the general population.

The DX204-AC155 chromosomes carry a CGG that corresponds to the S allele described by Morton and Macpherson [17], an allele differing from the unstable premutated Z allele and from the normal stable N allele, by its moderate instability. The absence of an S allele in fragile X families may be due to the time (in generations) needed by the CGG to expand and reach a size ≥ 54 repeats.