Comparative Frequency of Fragile-X (FMR1) and Creatine Transporter (SLC6A8) Mutations in X-Linked Mental Retardation

To the Editor:

The study by Rosenberg et al. (2004), in the July 2004 issue of The American Journal of Human Genetics, and the previous work by Dr. Salomons's laboratory on the implications of the creatine transporter gene, SLC6A8, for X-linked mental retardation (XLMR) are very important contributions to the field (Salomons et al. 2001; Rosenberg et al. 2004). I wish, however, to qualify the concluding sentences in the abstract and the discussion section of the study by Rosenberg et al. (2004), which may lead readers to overestimate the incidence of mutations in the creatine transporter gene in mental retardation (MR). The authors write in the abstract that the “frequency of SLC6A8 mutations in XLMR is close to that of CGG expansions in FMR1” (Rosenberg et al. 2004, p. 97). This is certainly incorrect. Rosenberg et al. (2004) found a 2.2% prevalence of SLC6A8 mutations in families with proven or possible XLMR (the latter are families with at least two males affected by MR and compatible with X-linked inheritance). On the other hand, the FMR1 expansion mutation associated with fragile-X syndrome is found in ∼2%–3% of males with MR who were not selected for family history (these figures are based on cohorts with little clinical preselection apart from the exclusion of clearly chromosomal or syndromic forms of MR) (see de Vries et al. 1997; Hecimovic et al. 2002; Pandey et al. 2002; Grønskov et al. 2004; Biancalana et al., in press). In fact, when selection is based on possible X-linked inheritance, the proportion of individuals with fragile-X syndrome is much higher. For instance, in the study by Fishburn et al. (1983), fragile-X syndrome accounted for MR in 12 of 45 male sib pairs with “nonspecific” MR, a proportion (27%) that is thus >10 times higher than the reported incidence of SLC6A8 mutations in a cohort containing sib pairs such as these as well as families with even more obvious XLMR. In fact, we have proposed recently that, unless there are clear hotspots of mutations and/or a very large mutation target size (such as for Duchenne muscular dystrophy, Rett syndrome, and hemophilia A), the population incidence of X-linked diseases implicating genes of average size that lead to highly decreased reproductive fitness is 10–20 times lower than the incidence of fragile-X syndrome (1/50,000–1/100,000 for most X-linked diseases, compared with 1/∼5,000 males for the fragile-X syndrome) (Chelly and Mandel 2001). Thus, one expects that the contribution to XLMR of an average gene that does not present mutation hotspots would be 10–20 times lower than that of FMR1. SLC6A8 is such a gene (with 13 exons, a 635-aa coding sequence, and no indication of highly recurring mutations), and thus we predict that its incidence in “nonsyndromic” MR will be in the range of 0.1%–0.3%. Indeed, even for the ARX (X-linked Aristaless) gene, which has a clear mutation hotspot that accounts for ∼6.6% of families with X-linked “nonsyndromic” MR, the incidence of this ARX recurring mutation in cohorts of patients with MR is much lower (∼0.15%) than that of FMR1 mutations (Grønskov et al. 2004; Mandel and Chelly 2004).

I also suggest that, in reporting prevalence estimates that are based on relatively small numbers of positive cases, it would be useful to give confidence intervals (CIs). Thus, the observed prevalence, in the study by Rosenberg et al. (2004), of 2.2% may indeed be an underestimate, since some mutations may have been missed and some variants of uncertain significance at present may prove pathogenic, or it may be an overestimate of the true prevalence, since, for the reported data, the CI for the prevalence of proven mutations is 1.0%–4.4%.