CAG repeat polymorphism within the KCNN3 gene is a significant contributor to susceptibility to anorexia nervosa: A case-control study of female patients and several ethnic groups in the Israeli Jewish population

INTRODUCTION

The relevance of human small-conductance Ca²⁺-activated potassium channel hSKCa3 to major psychoses, encoded by the KCNN3 gene, has been intensively studied. Three members of a gene family of small-conductance Ca²⁺-activated voltage-independent potassium channels SK1-3 have been identified by now [Kohler et al., 1996; Chandy et al., 1998; Desai et al., 2000]. All three are expressed in human brain, of which SK3 (KCNN3) is predominantly expressed in nigrostriatal and mesolimbic regions containing dopaminergic (DA) neurons [Dror et al., 1999; Rimini et al., 2000]. SK channels are fundamental in the regulation of neuron membrane potential through modulation of the after-hyperpolarization component and spike frequency [Sah, 1996; Xia et al., 1998]. SK3 specifically, has been recently shown to control pacemaker frequency and precision in DA neurons [Wolfart et al., 2001]. SK channels are implicated in memory and learning, as have been suggested on the basis of evidence of their involvement in synaptic plasticity and long-term potentiation [Messier et al., 1991; Behnisch and Reymann, 1998; Norris et al., 1998; Stackman et al., 2002]. The 1q21 chromosomal region that includes KCNN3 has been identified with susceptibility to familial schizophrenia in an earlier linkage study [Brzustowicz et al., 2000]. Chromosome 1q has been also reported by a recent family-based study as a potential localization of anorexia nervosa (AN) susceptibility locus [Develin et al., 2002]. Although another report has provided suggestive evidence for the presence of AN susceptibility locus on chromosome 1p, which does not contain KCNN3 [Grice et al., 2002].

The interest in KCNN3 has been furthermore prompted by the discovery of a multiallelic coding CAG repeat polymorphism that has been associated with schizophrenia [Chandy et al., 1998]. Case-control studies have shown that longer CAG repeats are over-represented in schizophrenia patients [Bowen et al., 1998; Chandy et al., 1998; Dror et al., 1999] and are associated with negative symptoms of schizophrenia [Cardno et al., 1999; Ritsner et al., 2002]. A rare finding of a schizophrenia patient with KCNN3 truncation mutation spanning the CAG repeat region [Bowen et al., 2001], has further suggested that the mutant protein, while mislocalized to the nucleus, confers a dominant suppression effect on SK channels conductance and firing pattern [Miller et al., 2001]. Moderately expanded CAG repeats within the KCNN3 have been also demonstrated in patients with ataxia [Figueroa et al., 2001]. In spite of the above-mentioned studies, which support KCNN3 contribution to major psychoses, numbers of attempts to reproduce these results in other populations have been inconclusive [Li et al., 1998; Antonarakis et al., 1999; Hawi et al., 1999; Rohrmeier et al., 1999; Tsai et al., 1999; Ujike et al., 2001; Laurent et al., 2003].

We have previously reported in a family-based study, that longer CAG repeats are preferentially transmitted to patients with AN [Koronyo-Hamaoui et al., 2002]. This is consistent with the view that KCNN3 might be a predisposing factor common to several psychopathologies. The population profile of KCNN3 alleles, in terms of the most prominent 2nd CAG repeat polymorphism, shows bimodal distribution with a major modal at 19 repeats that shifts towards longer (>19) repeats when population includes psychopathologies [Bowen et al., 1998; Chandy et al., 1998; Dror et al., 1999; Koronyo-Hamaoui et al., 2002]. Similar population profiles have been demonstrated in several other independent studies, although ethnic variation has not been sufficiently investigated under comparable conditions. We, therefore, presently pursue the analysis of this polymorphism in a larger series of AN patients and several control groups, while considering data on ethnicity and AN co-morbidity.

MATERIALS AND METHODS

RESULTS

Table I shows the 2nd CAG repeat polymorphism within the KCNN3 gene, in AN patients and control groups, the later consists of Jews of Ashkenazi, Iraqi, Iranian, Yemenite, North African, and Ethiopian origin. The distribution of KCNN3 alleles in various control groups is similar, that is evident from bimodal profile in the range of 11–22 repeats and a major modal at 19 repeats, previously described [Bowen et al., 1998; Chandy et al., 1998; Dror et al., 1999; Koronyo-Hamaoui et al., 2002]. In contrast, the allele distribution profile in the AN group shifts significantly towards longer CAG repeats in compare to the controls.

An alternative approach to the traditional analysis, which compares allele population pools, proposes an analysis that considers the two alleles in each individual separately, namely a major and a minor allele. Table II shows the length of the two individual alleles in various groups in terms of range, mean, modal, as well as length difference and sum. Comparison of all ethnic groups included in the control series, yields a significant variance in the mean length of the minor but not the major allele (P = 0.005 and P = 0.255, respectively, by Kruskal–Wallis non-parametric test). This is due to the Ethiopian and Iranian groups, which are distinct by the length of the minor allele in compare to the four other ethnic groups (Table II). Re-analysis of the control series, while excluding these two ethnic groups, results in non-significant difference in allele length (P = 0.103 and P = 0.470 for the minor and major allele, respectively). Since our AN patients are predominantly of Ashkenazi origin (70%) and do not include Iranian and Ethiopian ethnicities, we continue the case-control analysis including the Ashkenazi, Iraqi, North African, and Yemenite groups only, while corrected for the mean allele length according to the relative representation of these ethnicities in the AN patients group.

Table II shows the results of the case-control analysis by this approach. A shift towards longer alleles in the AN subset is apparent from the comparison of the major allele in the AN versus control group (mean repeat length = 19.57 ± 1.52 and 18.83 ± 1.39, respectively), yielding P < 0.001 (by two-tailed non-parametric Mann–Whitney test). Accordingly, the shift towards longer alleles is reproduced in the analysis by the sum of repeats in both alleles (36.90 ± 3.18 in AN vs. 35.72 ± 3.21 in controls), yielding P = 0.035 (by the same test). In contrast, the difference in the length of repeats between the major and minor alleles is similar in both groups (2.25 ± 1.98 vs. 1.95 ± 1.82, respectively, P = 0.189). These results propose that major KCNN3 allele is longer in AN patients. The assumption that sum of repeats could be a meaningful factor in predisposition to AN, stems from the previously proposed notion which argues that the total number of polyglutamine residues in the KCNN3 tetramer may have an additive effect on channel stability and function [Perutz, 1996]. However, no experimental evidences of the former have been reported, as yet. Another study has suggested that residues number asymmetry may affect channel activity [Saleem et al., 2000]. Nevertheless, in the present study the difference in number of CAG repeats between the two alleles has not been found a distinctive factor in AN.

We further perform the analysis under the hypothesis proposed by Chandy et al. [1998], implying that alleles longer than 19 repeats acquire dominant effect on disease susceptibility. Dichotomization at 19 repeats produces S/S, S/L, and L/L genotypes (S for short ≤19 and L for long >19 alleles). Table III shows frequencies of these genotypes in the AN and control groups that yields a significant difference with P = 0.0017 (by two-tailed Pearson's χ² test χ² = 12.73; DF = 2). An analysis which considers en masse hetero- and homo-zygous L allele genotypes (SL and LL), reveals a higher representation of these genotypes in the AN group with P < 0.001 (χ² = 11.48; OR = 3.53; DF = 1; 95% CI = 1.73–7.16). This odds ratio strongly suggests that the presence of the L allele is associated with a higher risk to develop AN condition. It is also consistent with other findings in schizophrenia and bipolar disorder [Bowen et al., 1998; Chandy et al., 1998; Dror et al., 1999].

In attempt to learn about possible genotype–phenotype correlation, 62 AN patients have been included in further association analysis, which considers AN co-morbidity categorized as depression, anxiety, and OCD disorders. Tentatively, patients with OCD have been found with longer CAG repeats, 10 out of 13 (77%) patients with OCD versus 25 out of 49 (51%) without OCD carry the L allele, yielding χ² = 2.804, DF = 1, and P = 0.049 (by one-tailed Pearson's χ² test). These results are expected to withstand verification in larger series of patients. None of the other co-morbidity categories have been found distinct by the frequency of the L allele.

DISCUSSION

We have previously shown in a family-based study that longer KCNN3 (hSKCa3) alleles are preferentially transmitted to AN [Koronyo-Hamaoui et al., 2002]. That has been replicated in the present case-control study, while working under the methodological approach that compares the two alleles of each individual rather than allele population pools. This distinction is particularly important if modus operandi of the gene is to be considered. The question, how longer CAG repeats of the KCNN3 gene confer pathogenicity, has been raised repeatedly. The hypothesis that the modal repeat length of 19 is a critical threshold for disease susceptibility [Chandy et al., 1998], has not been sufficiently supported by biophysical or biochemical evidence, so far. Another model that considers the total number of CAG repeats (glutamine residues) in the KCNN3 tetramer channel has been adopted in this study by the use of the analysis of repeat sum. However, the critical threshold at 19 repeats cannot be ruled out as a meaningful interpretation, since hetero- and homo-zygous carriers of the L allele (>19 repeats) have been found significantly more prevalent in AN group than in respective controls. The hypothesis suggesting that the tetramer channel asymmetry is associated with pathogenicity is interesting [Saleem et al., 2000], yet has not been supported in this study.

The bimodal profile characteristic of the KCNN3 allele distribution has been described in several population studies including European and USA Caucasian population [Bowen et al., 1998; Chandy et al., 1998], Ashkenazi Jews [Dror et al., 1999], as well as Japanese, Chinese, and Indian populations [Tsai et al., 1999; Saleem et al., 2000; Ujike et al., 2001]. This study, which includes an extended group of controls in order to obtain an authentic KCNN3 allele distribution in the general Israeli population, demonstrates that similar distribution exists in six well-characterized and relatively homogeneous populations of Jewish origin. We have found that four of these Jewish ethnic groups, namely Ashkenazi, North African, Iraqi, and Yemenite, do not differ significantly in distribution of KCNN3 alleles and thus could be considered, as a whole, a control group that is compatible with case-control studies of KCNN3 in the Israeli Jewish population.

The concluding part of this study argues for possible association of CAG repeat polymorphism with AN that includes co-morbid OCD phenotype, since a tendency towards longer CAG repeats has been observed in AN patients with OCD co-morbidity. The question whether KCNN3 is indeed associated with a specific AN endo-phenotype needs to be verified in a larger series of patients. To this end, a minimum of 40 AN patients with OCD and 80 AN patients with no OCD would be sufficient in order to obtain the power of 0.81 in the detection of any significant association (P < 0.05).

Two converging lines of evidence emerge from the recent literature that could foster our understanding of the functional relationship between SK channels and normal and abnormal CNS activity, as well as their relevance to mental disorders. As SK channels hyperpolarize the membrane potential [Moghaddam et al., 1997] thereby inactivate NMDA-receptor, hyperactive SK channels would therefore be expected to induce NMDA-receptors hypofunction resulting in enhanced susceptibility to neuropsychiatric disease. The association between longer CAG repeats and psychopathologies leads to the Perutz hypothesis suggesting that longer polyglutamine repeats might increase channel activity via gain of function effect [Perutz, 1996]. Yet, pharmacological blockade of SK channels have been reported to enhance neuronal firing and to lead to increased dopamine release, consistent with the pathogenesis of DA-related neuropsychiatric disorders [Shepard and Bunney, 1988, 1991]. In this context, the decreased functional expression of SK channels, like in the example of the dominant-negative suppression induced by the KCNN3 truncation mutant which lacks CAG repeats [Miller et al., 2001], is consistent with the consequences of pharmacological blockade. Although the specific manner by which KCNN3 repeats expansion predisposes to mental disorders is still unclear, findings of this polymorphism in the context of schizophrenia, bipolar disorder as well as AN, suggest that KCNN3 may be a key to the common mechanism underlying predisposition to psychoses, mood and eating disorders.