Working Memory-related Brain Imaging Phenotypes And Their Genetic Associations

Part 1: Working memory-related brain network phenotypes and their genetic associations Purpose:Genome-wide association studies(GWAS)have discovered genetic associations of brain imaging phenotypes,but most studies are based on European samples,leaving their genetic associations in other populations unknown.This study will investigate genetic associations of structural and functional phenotypes of brain network in healthy young Han Chinese using GWAS,and analyze the effects of genetic risk of brain network phenotypes on working memory.Materials and Methods:1.Genetic data collection and preprocessingThis study was based on the Chinese Imaging Genetics(CHIMGEN)project(n =7306).The high-throughput genotyping chip(ASA),designed for Asian populations,was used for genotyping,and IMPUTE2 was used for imputation,then 7,163 eligible subjects with 6,830,146 eligible single nucleotide polymorphism(SNPs)were included after strict quality control.2.Image data collection and preprocessingIn this study,9 MRI scanners were used in 30 centers,and three-dimensional T1 weighted imaging and resting-state functional MRI(rs-fMRI)data were obtained from all subjects who met the enrollment criteria.T1 images were preprocessed based on Freesurfer software,and the cortical surface area(CSA),cortical thickness(CTh)and cortical volume(CV)of each brain functional network were extracted based on the yeo network template.DPASFA software was used to pre-process the rs-fMRI,and the mean amplitude of low frequency fluctuations(ALFF)and regional homogeneity(ReHo)of each brain functional network were extracted.Since this study was a multicenter project and the brain image acquisition parameters were different across centers and scanners,the extracted metrics were harmonized by the Combat method.Among the subjects with qualified genetic data,7,067 subjects with qualified structural phenotypes and 6,881 subjects with qualified functional phenotypes were included.3.Working memory data collection and preprocessingWorking memory(WM)ability was measured by N-back tasks designed by the EPrime software.3-back accuracy was used as the WM performance in this study,based on genetically qualified subjects,the final number of subjects included for the 3-back accuracy was 6,427 after strict quality control.4.Genetic associations of structural and functional phenotypes of the cortical subnetworksA total of 170 structural and functional image-derived phenotypes(IDPs)from 17 networks were used to calculate their heritability using SUMHER software;GWAS was performed using BGENIE software,and to obtain reliable results,we categorized subjects acquired by the MR750 scanner as the discovery set,5,031 subjects for structural IDPs and 4,934 subjects for functional IDPs.And non-MR750 scannercollected subjects were grouped into the validation set with 2,036 subjects for structural IDPs and 1,947 subjects for functional IDPs.We used P<5E-8 as the significance threshold for the discovery set and nominally P<0.05 as the significance threshold for the validation set.The significant SNPs were selected by FUMA software.We used MAGMA software to perform gene-level association analysis to find more associated genes.5.Association between genetic risk of cortical subnetworks and WMThe cortical subnetwork IDPs that were significantly associated with 3-back accuracy were identified by partial correlation analysis,and then PRSice software was used to determine whether polygenic risk scores(PRS)of the significant IDPs were associated with 3-back accuracy.Results:1.Heritability of structural and functional IDPsIn this study,by calculating heritability,we found that genetic factors had a greater influence on the cortical subnetworks’ CV and CSA,while the heritability of CTh was smaller,and the heritability of structural IDPs ranged from 0.169-0.359;the heritability of cortical subnetworks’ ALFF and ReHo were both smaller,ranging from 0.175-0.200.2.Genetic associations of cortical subnetwork IDPsGWAS revealed that 38 IDPs were significantly associated with SNPs in the discovery set,and 15 IDPs could pass the significance threshold in the validation set.These IDPs included CSA and CV of the left salience network A,CV of the right visual network A,CV of the bilateral default network A/B,ALFF of the left default network C,ReHo of the right default network B,CV of the right somatomotor network B,ReHo of the left somatomotor network B,ALFF of the left dorsal attention network A,ReHo of the left limbic network B,ReHo and ALFF of the right limbic network B.Significant SNPs were located in the intergenic and intronic regions.MAGMA identified several genes associated with IDPs,such as the ALDH2 gene,which is associated with CV of the right visual network area A and the right dorsal attention network area A.This gene encodes a family of proteins aldehyde dehydrogenases,which are involved in glucose metabolism and neurotransmitter clearance in synaptic cleavage.CYFIP1 was associated with ALFF of the left default network area B.This gene encodes a protein that regulates cytoskeletal dynamics and protein translation,the deletion of which is associated with an increased risk of schizophrenia and epilepsy.3.Working memory-related cortical subnetwork IDPs and their genetic associationsWe found 35 IDPs that were significantly associated with 3-back accuracy,all of which could pass the correction for multiple comparisons(P<0.05/170 = 2.941E-4).Based on the significant SNPs found in the previous step,we found PRS for CSA of the left salience network A was significantly associated with 3-back accuracy.Conclusions:The CV and CSA heritability of 17 cortical sub-networks in healthy young Chinese Han was higher than that of CTh.GWAS identified 15 IDPs with significant associations with SNPs,all SNPs were located in intergenic and intronic regions;19genes were found to be associated with IDPs based on gene-level association analysis;these findings provide new evidence for the genetic basis of brain structure and function in the Chinese population.The PRS of the left-salience network CSA was significantly associated with working memory performance,which provided new idea for exploring the pathway of gene-brain-WM.Part 2: Working memory-related cortical regions and their genetic associations Purpose:Working memory is a basic human cognitive function.However,the genetic signatures and their biological pathways remain poorly understood.In the present study,we tried to clarify this issue by exploring the potential associations and pathways among genetic variants,brain morphometry and working memory performance.Materials and Methods:1.Data collectionThis part of the study was performed on two datasets,a Human Connectome Project(HCP)dataset,in which we included 1141 subjects with multiple races as well as relatives.The genetic data were imputed and quality-controlled.WM performance was evaluated by 2-back accuracy.The T1 images were preprocessed by Freesurfer and204 cortical IDPs were extracted according to the Desikan＿Killiany template.Another dataset was the UKB dataset.GWAS data with the same IDPs as extracted from HCP were downloaded.2.Association between IDPs and WMPartial correlation was performed between IDPs and 2-back accuracy in the HCP data using a linear mixed model(LMM),which was effective in correcting for population stratification.Multiple comparison was corrected for the number of independent tests.3.Identifying SNPs associated with IDPsBased on the identified WM-related IDPs,single nucleotide polymorphisms(SNPs)associated with these IDPs were screened for the following criteria: 1)SNPs with P<5E-8 in the UKB GWAS results;2)SNPs with nominal P<0.05 in the HCP GWAS statistics;and 3)SNPs with the same direction of effect in the HCP and UKB datasets.The independently significant SNPs were further obtained by FUMA software from the above filtered SNPs.4.Gene-IDPs-WM pathway analysisA mediation model with independently significant SNPs,left cuneus CV,and 2-back accuracy was constructed in the HCP data,and a mixed linear model was used to correct for population stratification.5.Validation in the white subset of the HCPSubjects whose race marked white were selected from the HCP and the above steps were repeated to validate the results.Results:1.We found a significant positive correlation between the left cuneus volume and 2-back accuracy(T=3.615,P = 3.150E-4,Cohen’s d = 0.226,corrected using family-wise error [FWE] method).2.Based on the LMM-based GWAS on the HCP dataset and UK Biobank 33 k GWAS summary statistics,we identified eight independent SNPs that were reliably associated with left cuneus volume in both UKB and HCP dataset.3.Within the eight SNPs,we found a negative correlation between the rs76119478 polymorphism and 2-back accuracy(T =-2.045,P = 0.041,Cohen’s d =-0.129).Finally,an LMM-based mediation analysis identified a significant effect of left cuneus volume in mediating rs76119478 polymorphism on the 2-back accuracy(indirect effect=-0.007,95% CI = [-0.045,-0.003]).4.These results were also replicated in a subgroup of Caucasians in the HCP population.Fine mapping demonstrated that rs76119478 maps on inter-gene regions of CTD-2315A10.2 adjacent to protein-encoding gene DAAM1,and is significantly associated with L3 HYPDH mRNA expression.Conclusions:Our study suggests that rs76119478 may regulate the working memory through exerting influence on the left cuneus volume.This part of the study provides new idea on the imaging genetic mechanisms of working memory.