Both rare and common genetic variants contribute to autism in the Faroe Islands

Abstract

The number of genes associated with autism is increasing, but few studies have been performed on epidemiological cohorts and in isolated populations. Here, we investigated 357 individuals from the Faroe Islands including 36 individuals with autism, 136 of their relatives and 185 non-autism controls. Data from SNP array and whole exome sequencing revealed that individuals with autism had a higher burden of rare exonic copy-number variants altering autism associated genes (deletions (p = 0.0352) or duplications (p = 0.0352)), higher inbreeding status (p = 0.023) and a higher load of rare homozygous deleterious variants (p = 0.011) compared to controls. Our analysis supports the role of several genes/loci associated with autism (e.g., NRXN1, ADNP, 22q11 deletion) and identified new truncating (e.g., GRIK2, ROBO1, NINL, and IMMP2L) or recessive deleterious variants (e.g., KIRREL3 and CNTNAP2) affecting autism-associated genes. It also revealed three genes involved in synaptic plasticity, RIMS4, KALRN, and PLA2G4A, carrying de novo deleterious variants in individuals with autism without intellectual disability. In summary, our analysis provides a better understanding of the genetic architecture of autism in isolated populations by highlighting the role of both common and rare gene variants and pointing at new autism-risk genes. It also indicates that more knowledge about how multiple genetic hits affect neuronal function will be necessary to fully understand the genetic architecture of autism.

Fig. 1

Genetic background of the Faroese population. a Geographic localization of the Faroe Islands. b Multidimensional scaling plots (MDS) of genome-wide identity by state (IBS) pairwise distances between 1000 Genomes and Faroese populations. Each dot represents an individual and the distance between two dots corresponds to genetic distance based on genome-wide pairwise IBS calculations. c Degree of inbreeding across 1000 Genomes and Faroese populations. The inbreeding coefficients of the Faroe non-autism control individuals (n = 176) were compared to the 1000 Genomes populations. The 26 populations from 1000 Genomes project are described in S1 Appendix. Logarithmic scale was used for y-axis

Fig. 2

Rare CNVs in Faroese individuals. Rare copy-number variant (CNV) analysis among gene-set lists within Faroe individuals (XHMM CNV calling from WES and CNV frequency < 0.01 in controls). The number of exonic CNV carriers altering any gene or gene-set lists (SFARI genes, pLI > 0.9 genes and Brain genes, see Materials and Methods section) were compared between individuals with autism, siblings and controls (one-sided Fisher’s exact test: n_autism= 36, n_sib = 28, n_controls= 107, p_{CNV_loss_SFARI}= 0.035, OR_{CNV_loss_All_SFARI}= 6.56; p_{CNV_loss_pLI>0.9}= 0.014, OR_{CNV_loss_pLI>0.9}= 13.25; p_{CNV_gain_All_genes}= 0.005, OR_{CNV_gain_All_genes}= 3.09; p_{CNV_gain_SFARI}= 0.035, OR_{CNV_gain_All_SFARI}= 6.56; p_{CNV_gain_Brain}= 0.003*, OR_{CNV_gain_All_Brain}= 5.61; *indicates the one withstanding Bonferroni correction for 12 tests; for families with multiple siblings, only one sibling was kept (closest on age)). Error bars represent confidence interval

Fig. 3

Genetic recessive mutations in Faroese individuals with autism. a Distribution of the inbreeding coefficient in Faroese individuals (one-sided Mann Whitney U-test: n_autism= 36, n_control= 176, n_sibling= 30; U_{control.vs.autism}= 2500, p_{control.vs.autism}= 0.023; U_{control.vs.sibling}= 1995, p_{control.vs. sibling}= 0.016; sibling inbreeding coefficients were averaged out by family; P-values were adjusted for principal component ancestry (3 principal components); *indicates the one withstanding Bonferroni correction for two paired comparisons). b. Number of rare LGD + MIS30 homozygous mutations carried per individual (one-sided Mann Whitney U-test: n_autism= 36, n_control= 107, n_sibling= 28; U_{control.vs.autism}= 1305, p_{controls.vs.autisms}= 0.011; sibling number of rare SNV were averaged out by family; P-values were adjusted for inbreeding; * indicates the one withstanding Bonferroni correction for two paired comparisons). c Venn diagram of the genes carrying the variants from b. Genes names are in bold and annotated when they are part of our gene-set lists (SFARI genes, pLI > 0.9 genes and Brain genes, see methods section). The plot on the right shows the proportion of individuals in each category carrying at least one mutated gene in our gene-sets lists (Fisher’s exact test: p_{controls.vs.autisms}= 0.03; p_{controls.vs. siblings}= 0.03). Error bars represent standard error. d and e are describing two specific families carrying multiple variants. “0” and “1” refer to wildtype or mutated allele, respectively. The localizations of the variants are indicated along the proteins and alignments throughout species showed the strong conservation of the altered amino acids. Nb Number, SNV Single Nucleotide Variant, LGD likely gene disruptive, MIS30 missense variants with CADD score ≥30, IgD immunoglobulin domain, TIL Trypsin Inhibitor-like, FA5/8 C Coagulation factor 5/8 type C domain, LamG Laminin G domain, EGF epidermal growth factor like domain, Fibr. Fibrinogen, alpha/beta/gamma chain, C-terminal globular domain, AAA ATPases associated domains

Fig. 4

Distribution of the genome-wide polygenic score for autism in Faroese individuals. a Distribution of the genome-wide polygenic score for autism (GPS-autism) of controls, autisms and siblings (one-sided Mann Whitney U-test: n_autism= 36, n_control= 176, n_sibling= 55; U_{control.vs.autism}= 2460, p_{control.vs.autism}= 0.017; Q1_autism = 0.0009, Q2_autism = 0.0011; Q3_autism = 0.0012, Q1_control = 0.0009, Q2_control = 0.0010, Q3_control = 0.0011, Q1_sibling = 0.0009, Q2_sibling = 0.0011 and Q3_sibling = 0.0012). b Distribution of the GPS-autism for the cases without intellectual disability (ID) and the cases with ID (one-sided Mann Whitney U-test: n_{autism-with-ID}= 12, n_{autism-without-ID}= 24; U_ID.vs.no-ID= 91, p_ID.vs.no-ID= 0.039; Q1_no-ID = 0.0010, Q2_no-ID = 0.0012; Q3_no-ID = 0.0013, Q1_ID = 0.0008, Q2_ID = 0.0009, Q3_ID = 0.0012). The GPS was calculated using PRSice-2 (see methods section, P-value threshold of 0.2 and R2 of 0.036). P-values were computed on data adjusted for principal component ancestry and for inbreeding

Fig. 5

Stratification of autism in Faroese individuals. On the left, the stratification was built using hierarchical clustering on the number of genes carrying rare deleterious variants altering SFARI genes (MIS30, LGD, or CNV) and on the genome-wide polygenic score for autism (GPS-autism). The other columns were not used for the clustering. The genetic profile contains variants with a predicted impact on the condition of the individual with autism. The clinical profile gives a subset of relevant information for each individual with autism. ID intellectual disability, M male, F female, del deletion, dup duplication