Accumulated common variants in the broader fragile X gene family modulate autistic phenotypes

Abstract

Fragile X syndrome (FXS) is mostly caused by a CGG triplet expansion in the fragile X mental retardation 1 gene (FMR1). Up to 60% of affected males fulfill criteria for autism spectrum disorder (ASD), making FXS the most frequent monogenetic cause of syndromic ASD. It is unknown, however, whether normal variants (independent of mutations) in the fragile X gene family (FMR1, FXR1, FXR2) and in FMR2 modulate autistic features. Here, we report an accumulation model of 8 SNPs in these genes, associated with autistic traits in a discovery sample of male patients with schizophrenia (N = 692) and three independent replicate samples: patients with schizophrenia (N = 626), patients with other psychiatric diagnoses (N = 111) and a general population sample (N = 2005). For first mechanistic insight, we contrasted microRNA expression in peripheral blood mononuclear cells of selected extreme group subjects with high- versus low-risk constellation regarding the accumulation model. Thereby, the brain-expressed miR-181 species emerged as potential "umbrella regulator", with several seed matches across the fragile X gene family and FMR2. To conclude, normal variation in these genes contributes to the continuum of autistic phenotypes.

Keywords: FMR1; FMR2; FXR1; FXR2; PGAS; miR‐181.

Figure 1. Positions of SNPs in FMR1,FXR1,FXR2, and FMR2, forming the 8‐SNP model as well as FMR1 and FMR2 repeat polymorphism length distribution in the discovery sample

1. Schematic overview of FMR1,FXR1,FXR2,FMR2, and position of the 8 selected single nucleotide polymorphisms (SNPs). Line represents introns, gray box at the beginning and end of each gene stands for UTR region, and red boxes represent exons. Gene structure plots generated using FancyGene (Rambaldi & Ciccarelli, 2009).

2. Distribution of repeat polymorphism lengths in FMR1 of male GRAS schizophrenia patients and healthy controls.

3. Distribution of repeat polymorphism lengths in FMR2 of male GRAS schizophrenia patients and healthy controls.

Figure 2. SNP overview and unbiased selection according to the standard operating procedure (SOP) developed for phenotype‐based genetic association study (PGAS) approaches

1. Genes from the “broader fragile X family” and their chromosomal position.

2. SNP numbers available through direct genotyping in the here used semicustom genotyping array (Affymetrix, Santa Clara, CA, USA).

3. SNPs fulfilling some of the first round of selection criteria (“functional” = SNPs, i.e. located in promoter region, 3′UTR or coding sequence; MAF = MAF ≥ 0.2; LD = SNPs that “survive” after linkage disequilibrium pruning: r² < 0.8). Underlined are the 13 SNPs selected for the PGAS approach using the PAUSS (selection requirements: fulfilled 2 of the above criteria or were functional). Not more than 3 SNPs per gene are selected to avoid overrepresentation of one gene.

4. SNPs with a tendency in PGAS (see Fig 3A) at single SNP basis went into the final accumulation model.

Figure 3. Criteria of final SNP selection in the phenotype‐based genetic association study (PGAS) approach—Standard operating procedure (SOP)

1. A, B

   A total of 13 SNPs preselected according to the SOP presented in Fig 2 underwent PGAS screening as exemplified here: (A) PAUSS association pattern of an exemplary fictitious autosomal (upper panel) and a sex‐chromosomal (lower panel) SNP which are eligible for the accumulation model. The genotype associated with the highest average PAUSS (in this example CC) is the “proautistic genotype” (indicated by the black arrow) and is assigned a score of 1. The heterozygous genotype is assigned a score of 0.5 and the homozygous genotype associated with the lowest PAUSS receives a score of 0. Please note that the difference between genotypes does not have to be statistically significant. (B) PAUSS association pattern of an exemplary fictitious autosomal (upper panel) and a sex‐chromosomal (lower panel) SNP which would not be selected for the accumulation model because of unclear phenotypical/biological relevance.

2. C, D

   The specificity of the association of the 8‐SNP accumulation model with an autistic phenotype (as determined using PAUSS; compare Fig 4B) is controlled by applying an unrelated (or “non‐sense”) phenotype, for example, delusional‐depression: (C) Intercorrelation pattern of single items included in the delusional‐depression composite score, used here as example control phenotype. Cronbach's alpha is presented as measure of internal consistency. (D) The delusional‐depression composite score is not associated with the number of proautistic genotypes of the 8‐SNP risk model in the discovery sample.

Data information: Error bars represent SEM.

Figure 4. Association of autism severity readouts in the discovery and 3 independent replication samples with the number of proautistic genotypes in the 8‐SNP risk model derived from the broader fragile X gene family

1. PAUSS (PANSS autism severity score) composition and item intercorrelation pattern in the GRAS sample of male schizophrenic individuals (discovery sample). Cronbach's alpha is presented as measure of internal consistency and also provided for the male replication samples I and II.

2. Association of PAUSS with the number of proautistic genotypes of the 8‐SNP risk model in the discovery sample; mean ± SEM.

3. PAUSS comparison of extreme groups with high and low numbers of accumulated proautistic genotypes in the discovery sample; binary logistic regression analysis with non‐z‐standardized PAUSS as dependent variable; mean ± SEM.

4. PAUSS comparison of extreme groups with high and low numbers of accumulated proautistic genotypes in replication sample I of male schizophrenia patients; binary logistic regression analysis; mean ± SEM.

5. PAUSS comparison of extreme groups with high and low numbers of accumulated proautistic genotypes in replication sample II of male disease control patients; Mann–Whitney U‐test; mean ± SEM.

6. The highly significant correlation of PAUSS and social support underlines the validity of social support as an autism proxy phenotype; mean ± SEM.

7. Comparison of the extreme groups with high and low numbers of accumulated proautistic genotypes for the autism proxy phenotype social support in the discovery sample; binary logistic regression analysis; mean ± SEM.

8. Comparison of the extreme groups with high and low numbers of accumulated proautistic genotypes for the autism proxy phenotype social support in the male replication sample III from general population; binary logistic regression analysis; mean ± SEM. For all replications, P‐values for one‐sided tests are shown.

Figure 5. Pilot experiments toward first mechanistic insight

1. FXR1 expression in PBMC of individuals carrying the rs2601 GG (low risk; N = 16) versus AA genotype (high risk; N = 27). Data represent mean ± SEM.

2. PBMC microRNA expression was normalized and data are plotted from all 4 miR‐181‐5p members (miR‐181a, miR‐181b, miR‐181c, d miR‐181‐5p). The bottom and top of the box are the first and third quartiles; the band inside the box is the median; the ends of the whiskers represent minimum and maximum values of the data.

3. Sequence homology of all four human mature miR‐181 species shown together with sequences in the broader fragile X gene family containing binding sites for the miR‐181 family (seed matches identified using Target Scan Human and SFOLD). The red letters specify seed sequences and seed matches, respectively. Chromosome positions for each seed sequence and seed match are denoted (human genome assembly GRCh38/hg38).

4. Denoted are ∆∆G values for binding of each of the miR‐181 family members to different 3′UTR positions. Positions were identified using Target Scan Human and SFOLD and then processed using PITA algorithm to yield the denoted ∆∆G values. ∆∆G is an energetic score, and the more negative its value, the stronger is the expected binding of the microRNA to the given site (Kertesz et al, 2007).