A new susceptibility locus for bipolar affective disorder in PAR1 on Xp22.3/Yp11.3^†

Bipolar affective disorder (BPAD [MIM 125480]) is a psychiatric disorder characterized by severe episodes of mania and depression with a prevalence of ∼1% in the world population. It is associated with a significant morbidity and mortality [World Health Organization, 2002]. The inheritance pattern is complex, involving both multiple genes and environmental factors. Occasional families may exist in which a single gene plays the major role in determining susceptibility, but the majority of bipolar disorder involves the interaction of multiple genes [Abou Jamra et al., 2007] or more complex genetic mechanisms such as imprinting [McMahon et al., 1995]. Genetic studies have provided strong evidence of a major genetic contribution to BPAD [Craddock and Jones, 1999]. Systematic genome screens have been reported on a variety of sample sets, ranging from large affected pedigrees in genetic isolates to large numbers of affected sibling pairs. The pattern of findings is consistent with there being no gene of major effect to explain the majority of cases of BPAD, and several regions have been implicated by individual studies. Table I (supplementary material) lists the most promising chromosomal regions suggested for BPAD in linkage studies. There have also been conflicting results, using linkage and association analysis, regarding the involvement of the pseudoautosomal regions (PAR1 and PAR2) in harboring susceptibility genes for BPAD [Müller et al., 2002]. PAR1 and PAR2 are two small regions on the tips of the X and Y chromosomes where pairing and crossover take place during male meiosis. Their physical lengths are ∼2.7 Mb for PAR1 and ∼0.33 Mb for PAR2. For more detailed information about the human PARs, we refer to the review by Flaquer et al., 2008.

Here, we present a linkage analysis using the largest set of genetic markers in the PAR1 so far and a substantial subset of families (96 of 108 families with still available blood to genotype markers in the PAR1) from a large collection of BPAD-affected families, with which a genomewide linkage analysis was previously performed where PAR1 was poorly covered [Schumacher et al., 2005]. In total, we genotyped 96 families of Spanish and German descent consisting of 720 subjects, of whom 346 were affected according to the broad phenotype definition (described below). The study conformed with all ethical guidelines of the institutions involved. The characteristics of these families are summarized in Table II (supplementary material). The ascertainment scheme has been described in detail by Schumacher et al. 2005. In summary, the phenotype evaluation was based on DSM-IV criteria [American Psychiatric Association, 1994]. The inclusion criteria for families with BPAD were the presence of a proband with bipolar I (BPI) disorder and a secondary affected sibling with either BPI, bipolar II (BPII), schizoaffective disorder bipolar type (SA/BP), or unipolar recurrent depression (UPR). Three definitions of affection status (ASD) were given, ASDI (narrow) includes individuals with BPI only; ASDII (medium) includes all individuals who received a diagnosis of BPI, BPII, or SA/BP; and ASDIII (broad) also includes individuals with UPR.We genotyped five polymorphic microsatellite markers (STR) described in public databases. To have a good coverage of the PAR1 region, we further identified two STR markers (XY_078 and XY_230). Genotyping for the eight STR markers was conducted using the procedures described by Cichon et al. [Cichon et al., 2001a,Cichon et al., 2001b]. The genetic intermarker distance was on average 5.16 cM in males and 0.43 cM in females. An average heterozygosity of 70% was observed.

Genetic markers in the PARs follow the same inheritance pattern as autosomal markers, becoming progressively more sex-linked (i.e., recombination events are more rare between X and Y) as they approach to the pseudoautosomal boundary. When using multipoint linkage analysis in the PARs, it is crucial to use sex-specific maps because there exists a marked difference in recombination frequencies between males and females [Flaquer et al., 2009]. In nonparametric multipoint analysis, the standard tools with sex-specific maps are valid for analyzing the PARs as long as there does not exist an excess of sex-concordant affected sib pairs. That is because even the transmission of markers in the PARs is similar to that of markers in the autosomal regions, males are more likely to receive the allele linked to the Y chromosome from their father, whereas females are more likely to receive the allele linked to their father's X chromosome. This behavior is more extreme for markers nearer to the sex-specific region. Therefore, one would expect an increased identity-by-descent sharing among same-sex-pairs regardless of whether a disease-susceptibility gene is present or not [Dupuis and Van Eerdewegh, 2000]. So, we first checked our data to see whether an excess of same-sex pairs was present. German families did not show differences in the number of concordant and discordant pairs. However, in Spanish families an excess of female-pairs was observed versus male-pairs (ratio of 72:22). The chi-squared test shows this difference in female-pairs and male-pairs to be statistically significant (P < 0.0001). After checking the Spanish families, we realized that in seven families all affected individuals were females. Consequently, to avoid false results, we decided to exclude these families from the analysis, obtaining nonsignificant differences between female-pairs and male-pairs for the remaining sample (ratio: 26:22). In the end, a total of 89 pedigrees (54 German, 35 Spanish) were used in linkage analysis. ALLEGRO, version 2.0f [Gudbjartsson et al., 2000] was applied for multipoint nonparametric analysis. The S_all statistic [Whittemore and Halpern, 1994] was used with the exponential model as the core statistic to test for linkage. This statistic is transformed into a statistic, Z_lr, that is robust against the incompleteness of the marker data, following the method of Kong and Cox 1997. It is asymptotically normally distributed with mean 0 and variance 1 under the hypothesis of no linkage. For compatibility with other studies, we also report the classic NPL scores [Kruglyak et al., 1996]. Indeed, when the IBD information is complete, there is a one-to-one correspondence between the NPL score and the Z_lr score for the exponential model. However, normal approximation tends to work better with Z_lr than NPL score [Nicolae et al., 1998]. All the analyses were performed using the most recent pseudoautosomal sex-specific genetic coordinates [Flaquer et al., 2009] implemented in the Rutgers map. In addition to the eight STR markers, sex as a phenotypic marker was added at the end of PAR1 to represent the SRY (sex determiner) gene. For this marker, females were denoted as homozygous with alleles 1/1 and males as heterozygous 1/2. To determine the inferential validity of the maximum Z_lr scores that were obtained, we performed systematic simulations under the null hypothesis of no linkage. We generated 10,000 replicates using MERLIN [Abecasis et al., 2002]. Each replicate was analyzed in the same way as was the original data. The empirical P-value was calculated as the proportion of all replicates showing a Z_lr score equal to or higher than the one observed in the real dataset. In addition, the Bonferroni correction was applied to the empirical P-values to account for the testing of three affection status definitions. Following the suggestion from McMahon et al. 1995 about a possible imprinting mechanism in BPAD, we performed a new analysis to allow for a parent-of-origin effect. Imprinting was investigated with the imprinting-based score function [Karason et al., 2003] that considers separately the paternal and maternal allele sharing of two affected relatives as implemented in ALLEGRO. In addition to the Z_lr analysis, we performed multipoint parametric (LOD) linkage analysis. In parametric analysis, the genetic parameters of the inheritance model, i.e., the penetrances and disease allele frequency, must be specified prior to the analysis. It is well known that the power to detect linkage decreases when the specified model is not sufficiently close to the true one. Consequently, for each of the subsamples, we calculated multipoint MOD scores, i.e., LOD scores that were maximized over genetic model parameters using the program GENEHUNTER-MODSCORE version 3.0 [Mattheisen et al., 2008]. This program calculates MOD scores automatically by varying the disease-allele frequency and three penetrances. We used the option “modcalc single” to perform a separate maximization for each genetic position assumed for the putative disease locus. This procedure yields a MOD score, in conjunction with the penetrances and disease allele frequency of the best-fitting trait model, for every genetic position.

Detailed results of the best Z_lr scores for both, the combined sample and for each of the subsamples (German and Spanish) are presented in Table I and Supplementary Figure 1. NPL scores are also shown for compatibility with other studies. In the combined sample, significant Z_lr scores (P_empirical < 0.05 after Bonferroni correction for multiple testing) were obtained under the ASDIII phenotype at male/female positions 31/3 cM (Z_lr = 2.44, P_empirical = 0.0192), at 33/3 cM (Z_lr = 2.87, P_empirical = 0.0051), at 53/6 cM (Z_lr = 3.11, P_empirical = 0.0018) and at 53.14/6 cM (Z_lr = 2.09, P_empirical = 0.0453). When analyzing the two subsamples separately, the highest scores were obtained in the Spanish subsample for the three phenotypes at 33/3 cM (ASDIII: Z_lr = 3.54, P_empirical = 0.0009. ASDII: Z_lr = 2.63, P_empirical = 0.0129. ASDI: Z_lr = 2.12, P_empirical = 0.0429). The German subsample did not show any significant score for any of the three phenotype definitions.

No significant results were obtained when testing for an excess of maternal or paternal sharing in this region, resulting in no evidence of genomic imprinting for any of the phenotype definitions nor any of the subsamples. No power was gained with the MOD score analysis. The highest MOD score value was 1.75 in the combined sample at 53/6 cM for ASDIII with penetrances {f_+/+ = 0.02, f_Het = 0.08, f_m/m = 0.08} and disease-allele frequency 0.06. Hence, this MOD score points to a nonzero phenocopy rate, a strongly reduced penetrance, an observation that is consistent with a genetically complex trait. There is an indication of a dominant mode of inheritance, although the MOD score is not very high.

The Spanish subsample shows a consistent peak for the three phenotype definitions at marker DXS1071 whose position in the genome is 1.63 Mbp from the X/Y telomere. In this region, very close to this marker the genes ASMTL and ASMT have been mapped at 1.59 Mbp and 1.79 Mbp respectively [Ross et al., 2005]. ASMTL is an ubiquitously expressed homologue of ASMT, whose precise biochemical function is not known [Ried et al., 1998]. ASMT encodes acetylserotonin methyltransferase, which catalyzes the final step in the synthesis of melatonin and is abundant in the pineal gland and retina and was proposed as a candidate gene for psychiatric disorders [Yi et al., 1993].

The aim of this study was the identification of chromosomal regions in PAR1 containing genes that confer susceptibility to this psychiatric disorder. To achieve this goal, linkage analysis was performed. The highest linkage peak provides a consistent picture for the three phenotypes in the Spanish subsample, obtaining the best score for the ASDIII phenotype (Z_lr score of 3.54 at DXS1071, P_empirical = 0.0009). The different results obtained in this study between German and Spanish subsamples is to be expected with a genetically complex disorder [Suarez et al., 1994], because factors such as different size and structure of the family samples and locus heterogeneity could influence the linkage results and lead to differences between the subsamples [Terwilliger et al., 1997]. Also the fact that MOD score analysis pointed to a nonzero phenocopy rate and strongly reduced penetrance for this region reflects and supports the opinion that BPAD has a complex mode of inheritance that is caused by multiple genes, and with different combinations of predisposing genes segregating in different families and populations. PAR1 has not attracted particular attention in previous linkage analysis of BPAD, despite the fact that it hosts the ASMT gene, which was previously suggested as a candidate gene for psychiatric disorders. Given the results presented here, this genome region in PAR1 definitely should be taken into consideration in further genetic analysis of BPAD.

WEB RESOURCES

Online Mendelian Inheritance in Man: http://www.ncbi.nlm.nih.gov/entrez/dispomim.cgi?id=125480.

World Health Organization: http://www.who.int/whr/2002/whr2002_annex3.pdf (for World Health Report 2002).

Rutgers genetic map: http://compgen.rutgers.edu/maps.