Introduction

Expansions of the CGG repeat in FMR1 over about 200 copies are associated with methylation and subsequent gene inactivation, which causes the fragile X syndrome. Full mutations arise through maternal expansion from a premutation. Premutations have 50–200 CGGs and are unmethylated. Three phenotypes have been associated with FMR1 premutations: (1) learning difficulty;1 (2) premature ovarian failure (POF)2, 3 and (3) fragile-X-associated tremor/ataxia syndrome (FXTAS).4, 5 The underlying molecular mechanism is unknown, but full mutation females are not affected with POF, suggesting that reduced FMR1 protein levels do not affect ovarian function.3 There is a significant increase in FMR1 mRNA with increasing repeat number, leading to speculation of an RNA mediated gain-of-function mechanism, similar to that seen in myotonic dystrophy.6 If FMR1 expanded RNA is pathogenic, one might expect to see anticipation in affected families, such that as repeat number increased, phenotype became more severe or had an earlier age of onset. We have investigated the affect of CGG repeat number on menopause age.

Materials and methods

We tested 45 postmenopausal premutation carriers from 32 families, ascertained through a retarded proband with no knowledge of menopause history. All individuals had ceased menstruating for at least 6 months without known cause. Menstrual histories were obtained by interview, and data on a proportion of this cohort have been published previously.7 In the original study many premutations were not detected by PCR and an estimate of size was made from conventional Southern blots. We have reanalysed all premutations by two methods, on consecutive gels. Genomic DNA (100 ng) was amplified with primers c and f8 using the GC-rich PCR system (Roche Diagnostics, UK). Products were separated on 2% agarose gels and sizes were measured by comparison with 100 bp ladder using Syngene software. Ten μl of PCR product was separated by denaturing PAGE and transferred to nylon membrane. Blots were hybridized with a CGG5 digoxygenin (DIG) labeled oligonucleotide and visualized with chemiluminescent detection (Roche Diagnostics, UK). Bands were sized relative to a DIG labeled size marker.

The agarose gel method gave the most consistent and accurate measure of size for 41/45 samples. The remaining four were only detected following Southern blot and hybridization to the PCR product. A comparison of the two methods showed a significant difference between allele sizing (t=2.982, P=0.005). Further investigation of the relationship between allele measurements from the two methods using regression analysis, showed the relationship was principally linear (P<0.0001) with a smaller but significant quadratic effect (P=0.012). The Southern blot method underestimated allele length in smaller repeats and overestimated it in the larger repeats. For the four cases with only a Southern blot result, parameter estimates from this regression were used to predict the direct method allele size. These four predicted values plus the 41 successful direct PCR results were used to determine the relationship between repeat size and menopause age.

Results

Regression analysis showed a significant relationship between premutation size and menopause age using a quadratic model (Figure 1, R2=0.22, F=5.9197, P=0.0054, df=2,42) with a minimal value at 80 repeats. There was no evidence for a linear effect (R2=0.01, F=0.3978, P=0.5316, df=1,43). Three women had menopause ages less than 30 years and could possibly be construed as outliers. These women were removed from the analysis, the quadratic model retested and the relationship remained significant (R2=0.18, F=4.2476, P=0.0214, df=2,39). However, given that these women are not related and that they are unlikely to have made a significant error in their reported age of menopause given the acute consequences on their reproductive lives, we believe their data to be accurate and their inclusion in the model justified. The nonlinear association of menopause age with premutation size indicates that while premutations with less than 80 repeats demonstrate increasing severity of the phenotype with increasing repeat number, over 80 repeats there is a reduction in phenotype severity with increasing repeat number. We investigated this relationship further by dividing the data into those with FRAXA alleles <80 repeats and those with FRAXA alleles >80 repeats and examined the linear effects. In agreement with Sullivan et al9 there is suggestive evidence of a linear decline in menopause age with increasing repeat size in those individuals with alleles in the range of 50–80 repeats (R2=0.16, F=3.9519, P=0.0607, df=1,20), lack of formal significance is presumably a function of diminished sample size and power when the data are split (n=22). A significant positive linear association between repeat size and age of menopause is seen in the women with premutations in the 80–110 repeat range (R2=0.31, F=9.2686, P=0.0062, df=1,21) despite the small sample size (n=23). This linearity remained statistically significant (P=0.0494) even when the three women with very low menopause age were removed from the analysis.

Figure 1
figure 1

Relationship between menopause age and CGG repeat size.

Full size image

Discussion

Preliminary evidence for a nonlinear association was presented by our group in 2004 (Murray et al, pers comm). Sullivan et al9 also reported a positive association between repeat size and ovarian dysfunction in smaller premutations and while the data were suggestive of a plateau over 100 repeats they were not significant. We have used our method to determine the size of five premutations included in Sullivan et al9 and they were all within five repeats of the size calculated in the Atlanta study, making the two studies comparable with respect to sizing methods. Sullivan et al also suggested that carriers with >100 repeats may go through menopause around the same age as noncarriers, but analysis was limited by the relatively small sample size. We also had few women with >100 repeats and intend to collaborate with other groups to increase the number of individuals in this category.

Our data demonstrate a statistically significant nonlinear association between menopause age and premutation size, such that menopause age decreases with increasing repeat number until about 80 repeats, thereafter increasing repeat number is associated with increasing menopause age. There are several hypotheses to explain our observations:

  1. 1

    As premutation carriers are at risk of transmitting full mutations to their offspring, a proportion of oocytes in premutation carriers may in fact have full mutations, with the proportion increasing with increasing premutation size. Full mutations are not at increased risk of POF, therefore females with large premutations in their peripheral blood may be more like full mutations with respect to ovarian failure.

  2. 2

    In an RNA-mediated model, it is likely that secondary structure is an important factor and perhaps CGG repeats in the mid-premutation range adopt a particularly deleterious structure. This may allow inappropriate protein binding or affect localisation of the transcript within cells.

  3. 3

    Recent studies have demonstrated the presence of three alternative transcription start sites for FMR1 and the relative proportion of these varies with repeat number.10 It is possible that these qualitative differences in RNA transcripts play a role in the aetiology of premutation phenotypes.

  4. 4

    The positive association between menopause age and CGG repeat numbers for premutations >80 could be a result of ascertainment bias. Mothers and grandmothers of fragile X males have, by definition, had at least some reproductive success and may have larger FRAXA alleles because the premutation has expanded to a full mutation. POF often results in infertility, thus mothers and grandmothers of fragile X probands could have a low risk of POF. We interrogated our data and found no evidence that mothers or maternal grandmothers were associated with larger premutations (P=0.9).

These hypotheses are not mutually exclusive and may act in combination.

In summary, we have found that increasing repeat size is associated with increased risk of ovarian failure for premutations under 80 repeats, but larger premutations are in fact at reduced risk for POF. The data support an RNA-mediated gain-of-function mechanism, but disease severity may not simply be a function of RNA quantity. The presence of full mutation oocytes in the ovary may also explain our nonlinear association data: there may be a trade-off between increasing mRNA as repeat number increases and decreasing mRNA as large premutations expand to full mutations in oocytes.