The NDE1 genomic locus can affect treatment of psychiatric illness through gene expression changes related to microRNA-484

Genetic studies of familial schizophrenia in Finland have observed significant associations with a group of biologically related genes, DISC1, NDE1, NDEL1, PDE4B and PDE4D, the ‘DISC1 network’. Here, we use gene expression and psychoactive medication use data to study their biological consequences and potential treatment implications. Gene expression levels were determined in 64 individuals from 18 families, while prescription medication information has been collected over a 10-year period for 931 affected individuals. We demonstrate that the NDE1 SNP rs2242549 associates with significant changes in gene expression for 2908 probes (2542 genes), of which 794 probes (719 genes) were replicable. A significant number of the genes altered were predicted targets of microRNA-484 (p = 3.0 × 10−8), located on a non-coding exon of NDE1. Variants within the NDE1 locus also displayed significant genotype by gender interaction to early cessation of psychoactive medications metabolized by CYP2C19. Furthermore, we demonstrate that miR-484 can affect the expression of CYP2C19 in a cell culture system. Thus, variation at the NDE1 locus may alter risk of mental illness, in part through modification of miR-484, and such modification alters treatment response to specific psychoactive medications, leading to the potential for use of this locus in targeting treatment.


FIGURES
: Graphical representation of the genetic effect of the NDE1 rs2242549 SNP on the gene expression levels in the discovery cohort of a) the largest GTEx replicated positive effect RAB24 (ILMN_2379718) (recessive model [GG and GT vs TT] p=3.5x10 -6 ) b) the largest GTEx replicated negative effect SCNN1D (ILMN_1754757) (recessive model [GG and GT vs TT] p=6.75x10 -4 ), c) the effect on TRIOBP (ILMN_1735788) (recessive model [GG and GT vs TT] p=2.99x10 -6 ) which replicates from our previous study but not in the GTEx database, and d) the non-significant effect on NDE1 (ILMN_1739805) which is highly significant in the GTEx database (p=2.3x10 -10 ). a) b) c) d) Figure S2: Power estimation of the discovery cohort to study the role of the NDE1 SNP rs2242549 with the gene expression data (n=39) with two effect sizes. For both estimations, the observed 90th percentile of the standard deviation for all genes from our data (σ= 0.527), and the value of estimated proportion of non-differentially expressed probes based on qvalue-calculations (pi0 = 0.485) were used. The effect size of the estimation a) was the maximum observed in our discovery cohort (Δ=0.52), and b) the minimum effect size able to give the power of 80 % (Δ= 0.250). For probes with smaller standard deviations, the power is underestimated. a) b)

Figure S3:
The linkage disequilibrium (LD) structure of the 7 SNPs previously genotyped on the larger Finnish family cohort for schizophrenia as illustrated by the Haploview (R1) program. a) Shows the LD as defined by D' and therefore represents the haplotypic structure of the region, whereas b) shows the R 2 LD and therefore represents which SNPs are directly correlated with each other and thus. Block structures are shown based on the respective prediction tools implemented in Haploview (58), "solid spine of LD" for D' blocks and "confidence intervals" (59) for R 2 blocks. a) b) Figure S4: Graphical representation of the gender by genetic interaction effect on medication groupings. Significance values can be found in Table 2.
All psychoactive drugs metabolised by CYP2C19 (Amitriptyline,Citalopram,Diazepam,Escitalopram,Fluoxetine,Mianserin,Sertraline) Table S1: Results of the genome wide gene expression analysis and replication. Sheet 1 (Table S1 a) p-value <0.05) lists all probes that were significantly altered for each of the five DISC1 network variants tested at the p-value threshold of p<0.05. Table includes effect size (β), standard error, t value and p-value for the three cohorts tested. In addition it lists those genes replicated from our previous study, and the q-value for the study in the family cohort.

TABLES
Furthermore, if the probe is to a gene predicted to be targeted by mir-484 by six or more programs as annotated by miRWalk, then in the mir-484 target column the word "Target" appears. Sheet 2 (Table S1 b) q-value <0.05) list all the same properties as in Sheet 1, but restricted to those probes with a q≤0.05 after applying the False Discovery Rate.
Separate Excel File: Table S1.xlsx Table S2: Results of the association analysis between individual psychoactive medications and the variants studied. a) p-values for the additive model controlling for gender, b) p-values for the additive model in interaction with gender, c) Odds ratios (and 95% CI) for the interaction terms that were significant at the uncorrected p-value ≤ 0.05 level. P-values, and their respective ORs, below 0.0021 are below the Bonferroni correction threshold for the 24 medications tested.