Age-related cortical shrinking
The UK Biobank large sample size enabled us to precisely estimate the mean of sulcal opening and grey matter (GM) thickness per age with a 1-year precision. Figure 1c underlines a strong correlation between the age and these two cortical features in the discovery sample. Between 47 and 73 years old, the sulcal opening increases on average of 0.025 mm/year, while the GM thickness decreases on average of 0.015 mm/year (Table S1). The Sylvian fissure and the subparietal sulcus show the maximum increase of opening and decrease of GM thickness, respectively, while the inferior temporal and left intraparietal have the minimum.
Heritability of sulcal opening and grey matter thickness around the sulci
In the discovery sample, we estimated the heritability (h2SNPs) of the sulcal opening and GM thickness using the GCTA method (Yang et al. 2011), i.e. the variance explained by all the SNPs (Fig. 2). Significant h2SNPs estimates for the sulcal opening range from 0.15 to 0.45, with a minimum in the left inferior temporal and a maximum in the left central sulcus (Table S2). Significant h2SNPs estimates for the GM thickness range from 0.15 to 0.37, with a minimum in the left superior frontal and a maximum in the left Sylvian fissure. Note that the sulcal opening heritability values are all higher than the ones of the GM thickness. The significant heritability estimates obtained support the fact that these two phenotypes are largely influenced by independent genetic factors and thus are good candidates for the following genome-wide association study.
Genome-wide association study (GWAS) of the cortical features
We performed a genome-wide association study on the genotyped data for the sulcal opening and GM thickness in the ten sulci. Manhattan and QQ plots are shown in the supplementary materials (Figures S4–11). Table 1 summarizes the 24 phenotype-SNP associations that were genome-wide significant in the discovery sample and nominally significant in the replication sample. Among these associations, 12 SNPs were unique at 5 different loci. The most represented locus, with 17 replicated association hits, is on chromosome 1, 27 kb before the transcription start site of the KCNK2. Within this locus, the two main associated SNPs are rs59084003 and rs864736 with 7 and 4 significant hits respectively. It should be noted that these two SNPs are not in strong linkage disequilibrium (LD) (r2 = 0.06) and other significant SNPs on this locus are either in LD with the first or the second one (Figure S12). The second most represented locus, with 4 significant associations, is on chromosome 16. This locus lies in the starting region of the non-coding RNA LOC101928708, in the vicinity of the protein coding gene C16orf95, followed by FBXO31. On this locus, the main associated SNP is rs9933149 with 3 hits. The three other loci each include a single significant replicated phenotype-SNP association. On chromosome 8, rs11774568, associated with the GM thickness in the left Sylvian fissure, is in a region with a high density of genes, between the genes DEFB136 and DEFB135. On chromosome 9, rs10980645, associated with the central sulcus opening, is an intronic variant of the LPAR1 gene. On chromosome 12, rs12146713, associated with the right STs opening, is an intronic variant of the NUAK1 gene.
Regional significance and direction of effect of the hit variants
DNA region upstream KCNK2 harbors significant SNP association with the sulcal opening (Fig. 3) and GM thickness (Figure S12) of the posterior cingulate, the central and intraparietal sulci. In addition, the implicated SNPs show near genome-wide significant influence on the sulcal features in the superior temporal, inferior frontal and subparietal sulci. Overall, this suggests a brain wide regulation for this genomic region. The sulcal opening is increased in carriers of the minor allele of rs864736 (maf = 46%, in our discovery sample), while it is decreased in carriers of the minor allele of rs59084003 (maf = 7%). The locus on chromosome 16 displays a more specific spatial control over the temporal and frontal lobes, with significant sulci including the inferior frontal, the superior and inferior temporal and the Sylvian fissure (Figures S13–14 a.). The sucal opening of these sulci is decreased in carriers of the minor allele rs9933149 (maf = 38%). It should be noted that the SNPs of these two loci have a significant pleiotropic influence on sulcal opening and GM thickness.
Regarding the three remaining loci, they preferentially influence one out of the two sulcal features. The intronic variant (rs12146713) of NUAK1 significantly influences the sulcal opening in the temporal region (Sylvian fissure, superior and inferior temporal sulci), as well as in the inferior frontal and intraparietal sulci (Figures S13-14 b.). The intronic variant (rs10980645) of LPAR1 significantly affects the sulcal opening of the central sulcus, and moderately affects the posterior cingulate (Figures S13-14 c.). The variant (rs11774568) chromosome 8, near DEFB136, appears to be linked with the GM thickness of the Sylvian fissure and superior temporal sulcus (Figures S7-8 d.).
Functional annotation of the loci
To obtain information on the role of the region upstream of KCNK2, we investigated the gene expression QTL using the GTEx database [GTEx Analysis Release V7 (dbGaP Accession phs000424.v7.p2)] (GTEx Consortium 2017). We found that the variants rs864736 and rs14526008, which are in LD, are significant eQTLs for the KCNK2 gene expression, with significant multi-tissue meta-analysis RE2 (random-effects model 2) (Han and Eskin 2011) p values of 9.7·10−6 and 5.7·10−9, respectively. Figure S15 presents the effect size in various tissues of rs864736 allelic configuration on KCNK2 gene expression. Even though the association barely reaches nominal significance in single brain tissue due to low sample sizes (~ 80–140 subjects), it overall emphasizes strong effects across brain tissues. The other significantly associated SNP near KCNK2 (rs59084003) was not reported as a significant eQTL in GTEx possibly for technical reasons because of its small minor allele frequency (7%). Indeed, the effect of allelic configurations cannot be well observed in the relative small GTEx sample. Table S3 summarizes the other eQTLs found in GTEx for the different loci.
In addition, rs864736 chromatin-state is annotated only for Brain Cingulate Gyrus (E069) tissue as being a recruitment site for PolyComb repressive complex (ReprPCWk, grey track, Fig. 4), which is an epigenetic system of gene silencing. This annotation in this specific tissue is particularly interesting because the strongest phenotypic associations with this variant are the ones with the posterior cingulate sulcus opening. We note also that the chromatin-state around rs504473 (in strong LD with rs864736) is annotated as enhancer (Enh, yellow track) specifically in brain tissues (Fig. 4) and thus this region seems to be crucial in the fine regulation of transcription of nearby genes, notably KCNK2, in brain tissues.