A Cross-Diagnostic GWAS of Longitudinal Executive Function Profile Scores

Urs Heilbronner1, Bernadette Wendel2, Monika Budde1, Katrin Gade1, Kristina Adorjan1, Janos L. Kalman1, Fanny Senner1, Till F. M. Andlauer3, Ashley L. Comes1, Eva C. Schulte1, Sergi Papiol1, Heike Bickeböller2 and Thomas G. Schulze1

1Institute of Psychiatric Phenomics and Genomics (IPPG), University Hospital, LMU Munich, Germany 2Department of Genetic Epidemiology, University Medical Center, Georg-August-University, Göttingen, Germany 3Department of Translational Psychiatry, Max Planck Institute of Psychiatry, Munich, Germany

IntroductionExecutive functions (EFs) concern the cognition involving con-trol and coordination of mental processes. They are impaired in severe mental disorder such as bipolar disorder and schizo-phrenia. Here, we study the genetic underpinnings of patient course over time in two core aspects of EFs: set-shifting and up-dating. We use data of the longitudinal PsyCourse study (Bud-de et al., 2018). In this multi-center study, patients from the af-fective-to-psychotic continuum (n=1047) were assessed at four measurement points across a period of 1.5 years.

MethodsPhenotypesThe Trail-Making-Test, Part B (TMT-B) and the Verbal Digit Span Back-wards (VDS) were used to assess set-shifting and updating compo-nents of EFs, respectively. All patients with at least three measure-ment points over time were analyzed, resulting in n=356 (TMT-B) and n=364 (VDS) individuals. We used a summary measurement, the area under the curve with respect to the increase (AUCI), to compute cognitive profiles scores that provide a measure of an indi-vidual’s longitudinal change over time. To impute missing scores at certain time points, we employed the R package Hmisc, using mul-tiple imputation, and creating multiple data sets (TMT-B: 10; VDS: 7). In each single data-set we calculated the AUCI for all individuals, thus yielding multiple phenotypes.

Genotyping, imputation and analysis of genetic variants Patients were genotyped with the Illumina PsychArray and common variants (MAF≥0.01) were imputed using the 1000 Genomes Phase 3 reference panel. We performed a GWAS for multiple phenotypes with PLINK 1.9, using linear regression with seven different covari-ates (age, sex, clinical center and four principal components from ancestry multidimensional scaling). We applied Rubin’s Rule to pool the multiple GWAS outputs.

DiscussionDespite finding no GWAS-significant hits, our results suggest longi-tudinal changes in EFs to be putatively associated with intracellu-lar function (TMT-B) and pathways affected by neurodegenerative disease (VDS). We will further research these phenotypes when more patients of the PsyCourse study are genotyped.

GrantsTGS: DFG Grants SCHU 1603/5-1 and SCHU 1603/7-1; BMBF Grants IntegraMent and BipoLife; Dr. Lisa-Oehler-Foundation (Kassel, Ger-many).HB: DFG Grants BI 576/5-1 and Research Training Group “Scaling Problems in Statistics” RTG 1644.

ResultsWe did not observe GWAS-significant hits. For the TMT-B, the top three variants were rs187822633 (chromosome (Chr) 2, p=5.95x10-7, β=738.1, 95% confidence interval (95% CI)=(722.9, 753.2)), rs79496814 (Chr 12, p=1.86x10-6, β=-236.6, 95% CI=(-241.7, -231.5)), and rs139931071 (Chr 10, p=2.75 x10-6, β=-1211.7, 95% CI=(-1238.2, -1185.2)). Rs187822633 is an intron variant in the “Leucine Rich Pen-tatricopeptide Repeat Containing” gene that is involved in cytoskele-tal organization, vesicular transport, and transcriptional regulation. Rs79496814 and rs139931071 are both intron variants in the genes for “Transmembrane and Tetratricopeptide Repeat Containing 2” and “Signal Transducing Adaptor Molecule”, respectively, and both proteins have functional roles as cellular adaptor molecules. The top three variants for VDS were rs200013838 (Chr 7, p=3.73x10-7, β=-42.3, 95% CI=(-43.2, -42.5)), rs61433316 (Chr 2, p=8.48x10-7, β=-24.7, 95% CI=(-25.2, -24.2)), and rs56280536 (Chr 2, p=8.48x10-7, β=-24.7, 95% CI=(-25.3, -24.2)). Rs200013838 is an intron variant in the “Family with Sequence Similarity 3 Member C” gene, which en-codes a secreted protein that has been associated with Alzheimer’s Disease (e.g. Hasegawa et al., 2014). Rs61433316 and rs56280536 are located near the gene for long intergenic non-protein coding RNA 1866.

ReferencesBudde et al. (2018), Am J Med Genet B Neuropsychiatr Genet doi: 10.1002/ajmg.b.32639. Eekhout et al. (2017), BMC Med Res Methodol 17:129.Hasegawa et al. (2014), Nat Commun 5:3917.Pruessner et al. (2003), Psychoneuroendocrinology 28:916-931.

Figure 1. Longitudinal profile scores (AUCIs; Pruessner et al., 2003, left) are created by imputing missing phenotype values using multi-ple imputation (Eekhout et al., 2017, middle). The AUCIs of each individual imputation are used as phenotypes in GWA studies (right). Results of each imputation are pooled afterwards to obtain a single p-value for each genetic variant.

Figure 2. QQ- and Manhattan plots of both GWAS.