Replicating studies of genetic modifiers in spinocerebellar ataxia type 3: can homogeneous cohorts aid?

Sir, We read with interest the paper by Tezenas du Montcel and colleagues (2014), which identified modifiers of the age at onset in spinocerebellar ataxias (SCAs). In polyglutamine (polyQ) SCAs, the size of the CAG expansion is incompletely correlated with the age at onset of the disease, suggesting the involvement of additional genetic factors. The identification of modifier genes for polyQ SCAs should not only contribute to more precise onset predictions, of relevance for genetic counselling, but also unveil aspects of the underlying pathogenesis, and pinpoint possible therapeutic targets (Génin et al., 2008). Proposed modifiers for SCAs have often failed to be validated in independent studies, a fact that has been mostly attributed to the lack of statistical power (as discussed, for example, in Pulst et al., 2005). The report by Tezenas du Montcel et al. (2014) represents therefore an important contribution to the problematic modifier genes in polyQ SCAs; this study includes the largest cohorts of patients for SCA1, 2, 3, 6, 7, 17 and dentato-rubro-pallydoluysian atrophy (DRPLA) analysed so far, which were recruited under the scope of the European Consortium on Spinocerebellar Ataxias (EUROSCA). Moreover, main findings from the EUROSCA series were tested in four independent smaller cohorts (three Caucasian cohorts from the USA, France and Italy, as well as one Asian from Japan). SCA3, also named Machado-Joseph disease, is the most frequent polyQ SCA, reaching in the Azores Islands (Portugal), the highest value of prevalence worldwide (Araújo, 2012). This particular epidemiological situation has allowed us to thoroughly follow a homogeneous SCA3 cohort, currently constituted by 93 patients (Table 1) belonging to extensively studied Azorean pedigrees. The Azorean population itself has been relatively isolated and is derived from a limited number of founders, factors which have shaped its genetic landscape (Pato et al., 1997; Santos et al., 2003, 2008; Montiel et al., 2005; Kristiansson et al., 2008). Moreover, these islands have a centralized health system with a regional programme specifically for SCA3, which has promoted the follow-up of all patients by the same neurologist, minimizing inter-observational error, and facilitating accurate annotation of clinical data, namely the age at onset. Characterization of the Azorean SCA3 cohort, and comparisons with EUROSCA and replication cohorts (Tezenas du Montcel et al., 2014) All quantitative variables were represented as mean ± standard deviation [minimum-maximum]. An independent t-test (two-tailed) was conducted to compare CAG repeats size in normal and expanded alleles, as well as age at onset for the different cohorts. Statistical analysis was performed in OpenEpi software (Dean et al., 2014) and IBM SPSS Statistics, version 22 (IBM Corp. Released 2013). A P-value lower than 0.05 was considered significant. Characterization of the Azorean SCA3 cohort, and comparisons with EUROSCA and replication cohorts (Tezenas du Montcel et al., 2014) All quantitative variables were represented as mean ± standard deviation [minimum-maximum]. An independent t-test (two-tailed) was conducted to compare CAG repeats size in normal and expanded alleles, as well as age at onset for the different cohorts. Statistical analysis was performed in OpenEpi software (Dean et al., 2014) and IBM SPSS Statistics, version 22 (IBM Corp. Released 2013). A P-value lower than 0.05 was considered significant. We tested the findings of Tezenas du Montcel et al. (2014) relative to SCA3. To do so, molecularly confirmed Azorean SCA3 patients were genotyped for SCA1, 2, 6, 7, DRPLA and Huntington’s disease loci (all patients were genotyped in the same laboratory; primers and conditions are available upon request). In the Azorean cohort, we observed a higher mean number of CAGs in the expanded allele with an accordingly earlier age at onset, when compared to the EUROSCA and to the French cohorts (Table 1). The opposite significant result is observed when we compare Azorean and Japanese SCA3 patients, the latter displaying a higher repeat length. Although average CAG length is higher in the Japanese patients, the onset is significantly delayed compared to the Azorean patients (Table 1). In their study, Tezenas du Montcel et al. (2014) found significant quadratic effects when estimating the variance in the explanation of onset provided by the expanded CAG tract size (improvement of 9% relative to a linear model). As Azorean SCA3 patients have a higher mean number of CAGs in the expanded allele, and larger alleles are more strongly correlated with onset than smaller expanded ones, we only obtain a modest improvement of 1.2% (Table 1). All the previous observations indicate the presence of differences on the profile and behaviour of the expanded alleles in SCA3, which seem to be population-dependent. In the EUROSCA cohort of SCA3, Tezenas du Montcel et al. (2014) failed to observe an association between onset and the size of the normal allele, which had previously been reported in other large studies (França et al., 2012; Zhou et al., 2014); this lack of association was also evidenced in their replication cohorts and is confirmed in ours. When testing additional effects of non-causative polymorphic (CAG)n-containing genes using the EUROSCA cohort, Tezenas du Montcel et al. (2014) proposed associations between the size of the ATXN2 (SCA2) and ATN1 (DRPLA) alleles and the age at onset of SCA3. These effects were not replicated in any of the alternative cohorts, or in our Azorean SCA3 cohort. An association was also found between shorter HTT alleles and SCA3 onset; this finding was not investigated in the replication cohorts and our Azorean SCA3 cohort does not support this association. For ATXN2, whose intermediate alleles (frequency of 7%) had been associated with an earlier SCA3 onset in the EUROSCA series, the lack of replication was interpreted by Tezenas du Montcel et al. (2014) as reflecting the rarity of ATXN2 intermediate alleles in the replication populations, a justification especially applicable to the French and Japanese SCA3 patients, which lack this class of alleles. Notwithstanding in our cohort, where intermediate alleles reach 5% (Fig. 1), patients with an intermediate ATXN2 allele also show a tendency for earlier estimated onset (34 ± 3 years, mean ± standard error) when compared with patients without an intermediate ATXN2 allele (38 ± 1 years). This result shows how a population-specific genetic profile at a candidate locus impacts the ability to confirm/dismiss a modifier effect. In this sense, replication studies should be performed in populations where the allelic profile for the candidate modifier locus/loci is similar; otherwise the ‘lack’ of replication becomes hard to interpret. Allelic distribution of CAG repeats in ATXN3, ATXN1, ATXN2, CACNA1A, ATXN7, ATN1 and HTT loci in Azorean SCA3 cohort. Polymorphic information content (PIC) and expected heterozygosity (H) are shown for each candidate loci. In the work of Tezenas du Montcel et al. (2014), when the ATXN3 locus was analysed as a potential modifier of other SCAs, it was reported as displaying a low level of polymorphism. A study of nearly 2000 Portuguese chromosomes determined the ATXN3 locus heterozygosity to be 85% (Lima et al., 2005). The high level of polymorphism in the ATXN3 locus was also described in the Japanese population [polymorphic information content (PIC) = 0.78; Sasaki et al., 1995]. Noteworthy, in our cohort, with the exception of ATXN2, all analysed candidate loci are highly polymorphic (Fig. 1), with values of PIC (Nagy et al., 2012) ranging from 63% (ATXN7) to 82% (HTT). As a positive finding, not highlighted in the EUROSCA series, we can further report that in our series, the longer allele at the ATXN1 is negatively correlated with onset, significantly increasing by 1.5% the explanation of SCA3 age at onset variability (part correlation coefficient; r2 = 0.701, P = 0.038). When compared to EUROSCA, we observe a higher frequency of longer normal ATXN1 alleles in our cohort. In summary, we believe that large multicentre collaborative studies, such as the EUROSCA initiative, are crucial in the analysis of genetic modifiers for SCAs. The ability to replicate findings from such large cohorts is dependent on how comparable the replications cohorts are relative to the discovery studies, both in the allelic profile of the causal mutation (CAG tracts size in polyQ SCAs) as well as in the allelic frequencies of the candidate modifier locus/loci. This study was supported by FEDER funds through the Operational Competitiveness Programme——COMPETE and by National Funds through FCT——Fundacão para a Ciência e a Tecnologia under the project FCOMP-01-0124-FEDER-028753 (PTDC/DTP/PIC/0370/2012). A PhD fellowship M3.1.2/F/006/2011 (M.R.) and postdoctoral fellowship M3.1.7/ F/031/2011 (A.R.) were supported by Fundo Regional para a Ciência (FRC), Governo dos Açores. C.B. is supported by the UK Medical Research Council (MRC).