Genetic influences on the shape of brain ventricular and subcortical structures

1 Brain ventricular and subcortical structures are heritable both in size and shape . Genetic 2 influences on brain region size have been studied using conventional volumetric 3 measures, but little is known about the genetic basis of ventricular and subcortical shapes. 4 Here we developed pipelines to extract seven complementary shape measures for lateral 5 ventricles, subcortical structures, and hippocampal subfields . Based on over 45,000 6 subjects in the UK Biobank and ABCD studies, 60 genetic loci were identified to be 7 associated with brain shape features (P < 1.09 × 10-10), 19 of which were not detectable 8 by volumetric measures of these brain structures . Ventricular and subcortical shape 9 features were genetically related to cognitive functions, mental health traits, and multiple 10 brain disorders, such as the attention-deficit/hyperactivity disorder. Vertex-based shape 11 analysis was performed to precisely localize the brain regions with these shared genetic 12 influences. Mendelian randomization suggests brain shape cau sally contributes to 13 neurological and neuropsychiatric disorders , including Alzheimer's disease and 14 schizophrenia. Our results uncover the genetic architecture of brain shape for ventricular 15 and subcortical structures and prioritize the genetic factors underlying disease-related 16 shape variations. 17 18

ABCD; Brain disorders; GWAS; Hippocampus subfields; Mental health; 19 Subcortical and ventricular shapes; UK Biobank. 20 21 22 23 24 25 26 27 28 29 30 31 32 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint 3 Human brain ventricular and subcortical structures are involved in complex brain 1 activities and have important roles in the regulation of cognitive, emotional, and motor 2 functions1-7. Morphometric variations of these brain structures can be quantified in-vivo 3 by structural magnetic resonance imaging (MRI) . MRI-based volumetric measures (such 4 as regional brain volumes) can estimate a region’s overall size, providing a conventional 5 measure of the gross variation of the structure. However, such aggregate measures may 6 not be sensitive to within-region local changes and may not fully capture the complexity 7 of structural deformations. Shape analysis has gained increasing attention to overcome 8 these limitations and characterize brain morphometry beyond simple volumetric traits8-9 11. Recent studies have found that shape features can precisely localize shape 10 deformations in brain structures, providing finer-grained information of the location and 11 pattern of morphological variations, which may not be detectable in traditional volume 12 analysis12,13. For example, shape analysis of the ventricular and subcortical structures has 13 provided sensitive biomarkers for healthy aging14 and the onset and progression of a wide 14 range of brain diseases, including Alzheimer’s disease (AD)15,16, schizophrenia 17,18, 15 epilepsy19, major depressive disorder (MDD) 20,21, 22q11.2 deletion syndrome 22, and 16 bipolar disorder23. 17 18 Both size (volume) and shape of brain ventricular and subcortical structures have been 19 found to be heritable in family studies24,25 and general populations13,26-28. For example, 20 the narrow sense single -nucleotide polymorphism (SNP) heritability estimates for the 21 volume of ventricular and subcortical structures were all higher or close to 40%26,27 in the 22 UK Biobank29 (UKB) studies, and the highest SNP heritability estimates for shape features 23 ranged from 32.7% to 53.3% across structures in the Rotterdam Study28. Genome-wide 24 association studies (GWAS) have been conducted to uncover the genetic basis of 25 ventricular and subcortical volumes30-40, yielding hundreds of associated genetic variants 26 and shared genetic influences with brain disorders and complex traits. However, there is 27 no large-scale GWAS on ventricular and subcortical shape features and their genetic 28 architecture has yet to be determined. 29 30 Using raw MRI from over 45,000 subjects in the UKB and Adolescent Brain Cognitive 31 Development41 (ABCD) studies, we developed pipelines to extract ventricular and 32 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint 4 subcortical shape features and characterized their genetic architectures. We identified 60 1 novel genetic loci that contributing to the shape variations , 19 of which cannot be 2 identified in previous GWAS of volumetric measures of these brain structures using the 3 same datasets. We found ventricular and subcortical shape features had shared genetic 4 influences with many cognitive traits and major brain disorders. We further revealed the 5 localized pattern of genetic effects in vertex -wise analysis and identified causal genetic 6 links between brain shape and disorders using Mendelian randomization analysis. The 7

of this shape study demonstrated genetic effects on ventricular and subcortical 8 structures at a finer spatial resolution than that of traditional volumetric analysis . Our 9 GWAS results will be available through the Brain Imaging Genetics Knowledge Portal (BIG-10 KP) https://bigkp.org/. 11 12

13 Generating reproducible ventricular and subcortical shape features 14 We developed pipelines to extract shape features from raw structural MRI images for 8 15 ventricular and subcortical structures, including the left/right lateral ventricles, nucleus 16 accumbens, amygdala, caudate, hippocampus, pallidum, putamen, and thalamus. 17 Furthermore, we studied 7 subfields of the hippocampus, namely the left/right cornu 18 ammonis 1 (CA1), CA3, fimbria, hippocampus-amygdala-transition-area (HATA), 19 hippocampal tail, presubiculum, and subiculum (Fig.1A). An overview of our workflow can 20 be found in Fig. S1. Shape deformations are usually decomposed into two components: 21 one within the surfaces, and the other along the normal axis of the surfaces. For each 22 vertex in the shape image, we calculated 7 complementary shape statistics, including the 23 radial distance from the medial model14 (referred to as the radial distance); the (log) 24 determinant and two eigenvalues of the Jacobian matrix from the surface tensor-based 25 morphometry (TBM) model42 (referred to as the determinant, eigenvalue1, and 26 eigenvalue2); and three features from the multivariate surface TBM (mTBM) model10,11 27 (referred to as the mTBM1, mTBM2, and mTBM3) (Figs.1B and S2). Briefly, the radial 28 distance describes morphometric changes along the surface normal direction11 and is also 29 called the radial thickness10. On the other hand, the 6 surface TBM and mTBM model 30 features capture surface deformations perpendicular to the surface normal axis (such as 31 rotation, dilation, and shear within the surfaces). For example, the determinant of 32 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint 5 Jacobian matrix is analogous to a surface area 43, measuring local area dilation or 1 contraction. It quantifies the surface dilation ratio between the given template and the 2 study subject by matching a small surface patch around a particular point of the subject 3 surface to the corresponding point on the template . The three advanced surface mTBM 4 features analyze the full surface tensor using log-Euclidean metrics and can capture more 5 complicated surface deformations10,11. 6 7 After extracting vertex-wise maps, we aggregated them and generated region-specific 8 summary-level features for downstream genetic analyses. For each shape statistics, we 9 have two groups of features . The first group includes 210 structure-averaged shape 10 features in regions or subfields by taking the mean across all the vertices within the 11 structure (7 shape measurements × (8 shape structures + 7 hippocampal subfields) × 2 12 hemispheres). In the second group, we applied principal component analysis to extract 13 1,120 region-specific principal components (PCs) by taking the top 10 PCs of the vertex-14 wise map for each of the 7 statistics in the 8 ventricular and subcortical structures 15 (left/right, 7 × 8 × 2 × 10) (Methods). Principal component analysis is a well-established 16

for dimension reduction with a wide range of neuroimaging applications. In shape 17 analysis, the top -ranked PCs can characterize the strongest variation components of 18 shape statistics within each structure, which can provide more microstructural details 19 about shape deformations omitted by structure-averaged measures, while alleviating 20 multiple testing burdens (Fig.1C). Clinically, variations represented by these PCs may 21 localize shape changes that are more relevant to specific brain-related complex traits or 22 diseases. 23 24 We evaluated the intra -subject reproducibility of the above region or subfield -specific 25 shape features using the repeat scans from the UKB repeat imaging visit ( average time 26 between visits = 2 years , average n = 2,788). Specifically, we quantified individual-level 27 differences between the two visits by calculating the intraclass correlation coefficient 28 (ICC) of each shape feature between two observations from all revisited individuals. The 29 average ICC was 0. 443 (standard error = 0. 234) across the 1,330 (210 + 1,120) shape 30 features (Table S1). There were 457 shape features with ICC > 0.5, including 117 features 31 for lateral ventricles, 246 for subcortical structures, and 94 for hippocampal subfields (Fig. 32 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint 6 S3A, mean ICC = 0.722, standard error = 0.132). The average ICC of the 7 shape statistics 1 ranged from 0.684 to 0.752 (Fig. S3B), and the lateral ventricles had the highest mean ICC 2 across all the 8 structures (Fig. S3C, mean ICC = 0.816, standard error = 0.129). Our later 3 genetic analyses focused primarily on these 457 reproducible shapes features (ICC > 0.5, 4 363 region-level and 94 subfield-level traits) (Table S2). 5 6 Heritability and associated genetic loci of ventricular and subcortical shape features 7 We estimated SNP heritability (h2) for these 457 reproducible shapes features using the 8 UKB individuals of white British ancestry via GCTA 44 (average n = 32,631, phases 1 -3 9 release). Most heritability estimates (456/457) were significant after adjusting for 10 multiple comparisons using the Benjamini -Hochberg procedure to control the false 11 discovery rate (FDR) at 0.05 level ( Fig. S4A and Table S3). The mean heritability ranged 12 from 22.1% to 27.3% (standard error = 1.85%) across the 7 groups of shape statistics , 13 suggesting different shape deformation measures were under comparable genetic 14 controls (Fig. S4B). The highest heritability reported in each structure ranged from 51.2% 15 (for lateral ventricles) to 19.3% (for amygdala) (Fig. S2C). On average, lateral ventricles 16 had higher heritability than that of subcortical structures (32.2% vs. 21.6%, P < 2.2 × 10-17 16). Subfield analysis provided more information for genetic influences on different parts 18 of the hippocampus. For example, we found the CA3 subfield had the highest heritability 19 among the 7 subfields, while the lowest heritability was observed on the HATA subfield 20 (Fig. S4D). In addition, the heritability estimates were largely consistent in females and 21 males (Fig. S5, correlation = 0.83). 22 23 It is known that the volum etric measures of these structures were also heritable26-28. To 24 quantify the genetic effects additionally contributing on shape measures, we compare d 25 the estimates of genetic variance for 363 region-level shape features before and after 26 adjusting for their corresponding regional volumes as covariates (Table S4). For the 129 27 mean and first PC ( PC1) shape features, we found that the average genetic variance 28 reduced from 0.241 to 0.145, indicating that 39.8% (0.096/0.241) genetic variations were 29 shared by regional brain volumes and shape features (Fig. 2A ). In this group of shape 30 features, the proportion varied greatly from region to region. For example, the largest 31 proportion was observed in the lateral ventricles (79.2%) and the least was observed in 32 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint 7 the amygdala (3.4%). As both mean and PC1 features captured the major variations in the 1 brain region45, these results quantified the overlapping genetic influences between shape 2 and volumetric measures. On the other hand, for other PCs (other than the PC1s), the 3 genetic variance estimates were much more consistent before and after adjusting for 4 regional volumes (Fig. 2B). It was common for these PCs to capture more local variations 5 that are not captured by the mean or PC1 s45. Specifically, t he average proportion of 6 reduction was 11.9% and the reductions were small for the majority of shape features. 7 There was a substantial reduction of shape features in a few second PCs (PC2) on the 8 lateral ventricles (70.9%), which can be explained by the fact that the ventricles were large 9 and therefore the second PC still mainly reflected global variations. Overall, these results 10 suggest that local PCs of shape features can detect genetic influences which are largely 11 independent of those found in regional volumes or aggregated shape measures. 12 13 We performed GWAS for the 457 shape features using the UKB individuals of white British 14 ancestry (average n = 32,631, Methods). The average intercept in linkage disequilibrium 15 score regression (LDSC) 46 was 1.0072 (range = (0.982, 1.034)) , indicating no genomic 16 inflation of summary statistics because of confounding factors. At a stringent significance 17 level 1.09 × 10-10 (5 × 10-8/457, additionally adjusted for the number of shape features), 18 622 independent ( linkage disequilibrium [LD] r2 < 0.1) significant shape-variant 19 associations47 were identified , which were distributed across 60 genomic regions 20 (cytogenetic bands). There were 38 regions associated with the lateral ventricles, 10 with 21 hippocampal subfields, 9 with hippocampus, 7 with putamen, 6 with nucleus accumbens, 22 6 with caudate, 4 with pallidum, and 4 with thalamus (Fig. 2C). Table S5 summarizes the 23 list of index genetic variants and their associated shape features. Among the 60 regions, 24 19 were not identified by ventricular and subcortical regional volumes (at the 5 × 10-8/30 25 significance level ) in the same dataset. The genetic effects were highly consistent 26 between males and females in the sex-specific GWAS (Fig. S6, correlation = 0.966), where 27 analysis was conducted for females and males separately. 28 29 Using 5 independent European and non -European datasets, we replicated the genomic 30 loci identified in our discovery GWAS. First, we performed GWAS on a UKB European 31 dataset, which includes European individuals in the new UKB phase 4 data (early 2021 32 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint 8 release) and individuals of white but non-British ancestry in the UKB phases 1-3 data 1 (UKBE, removed the relatives of the discovery sample , average n = 4,596). Of the 622 2 identified independent (LD r2 < 0.1) shape-variant associations, 410 (65.9%) from 37 3 (61.7%, 37/60 ) loci passed the 0.05 nominal significance level in UKBE. Their genetic 4 effects all had concordant directions in the UKBE and original discovery GWAS and were 5 highly similar (correlation = 0.980). Second, we repeated GWAS on a validation dataset 6 with UKB non-European subjects (UKBNE, average n = 1,224). In this dataset, 19 loci can 7 be validated, and most of their associations (83/86) had the same directions with those 8 in the discovery GWAS and UKBNE. The validated genetic effects were highly consistent 9 between the white British discovery GWAS and non -European GWAS (correlation = 10 0.931). These results suggest similar genetic effects on subjects from different ancestries 11 in the same cohort. 12 13 Next, we carried out GWAS on 3 ABCD validation datasets: the ABCD European (ABCDE, 14 average n = 3,177), ABCD Hispanic ( ABCDH, average n = 662), and ABCD Black (ABCDB, 15 average n = 1,002). In ABCDE, 23 of the 60 genomic loci were significant at nominal 16 significance level and had the same effect direction as in the UKB discovery GWAS . 17 The ABCDH and ABCDB had 13 and 15 validated loci, respectively. Interestingly, we found 18 that the genetic effects of 3q28 locus on lateral ventricles were much larger in the three 19 ABCD datasets than those of the UKB discovery sample , especially for the non-European 20 subjects in ABCDH (mean absolute genetic effects 0.082 vs. 0.3367, P = 1.35 × 10-7) and 21 ABCDB (mean absolute genetic effects 0.082 vs. 0.401, P = 1.67 × 10-4). The 3q28 locus 22 was reported to have the strongest associations with lateral ventricular volume31 and was 23 widely associated with Alzheimer’s disease risk and biomarkers48. Larger 3q28 effects in 24 ABCD may suggest that the genetic effects on lateral ventricles were stronger for younger 25 subjects and/or non-European subjects. Overall, 43 of the 60 loci can be validated in at 26 least one of the 5 datasets, 30 loci can be validated in more than one dataset, and 5 loci 27 can be consistently validated in all datasets, including 3q28, 17q24.1, 14q32.11, 12q14.3, 28 and 10q26.13. The validated loci (such as 3q28) may have higher genetic effect sizes in 29 ABCD than in UKB discovery GWAS . These validation results were summarized in Figure 30 S7 and Table S5. 31 32 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint 9 Finally, we detected gene-level associations using MAGMA49 and FUMA47. MAGMA 1 reported 127 significant genes with 1,343 associations (P < 5.82 × 10-9, adjusted for 457 2 phenotypes), covering all the ventricular and subcortical structures (Fig. S8 and Table S6). 3 Among the 127 significant genes, 59 were identified by regional brain volumes35, 53 were 4 observed in DTI parameters45, and 12 overlapped with functional MRI (fMRI) traits50. Eight 5 genes were ass ociated with all the 4 brain imaging modalities, including FAM175B, 6 FAM53B, METTL10, and RP11-12J10.3 (METTL10-FAM53B readthrough) in the 10q26.13 7 region, as well as EPHA3 (3p11.1), ZIC1 (3q24), ZIC4 (3q24), and DAAM1 (14q23.1). The 8 FAM175B, FAM53B, and METTL10 genes had important functions in ribosomal translation 9 and cell regeneration51, and have been mapped to cocaine dependence52 and subjective 10 well-being53. The EPHA3 was involved in axon guidance 54 and was highly expressed in 11 mesenchymal subtype glioblastoma55. The DAAM1 was an important part of the planar 12 cell polarity signaling in neural development 56 and was highly expressed in human 13 cerebral cortex57. In addition, t he ZIC genes were important components in patterning 14 the cerebellum 58. Overall, these 127 MAGMA-significant genes showed gene ontology 15 enrichments59 in “cell morphogenesis involved in differe ntiation (GO:0000904)” and “T 16 cell receptor signaling pathway (GO:0050852)” biological processes at FDR 0.05 level (P < 17 2.76 × 10-6). We also used FUMA47 to map significant variants (P < 1.09 × 10-10) to genes 18 through a combination of their base pair location, gene expression, and 3D chromatin (Hi-19 C) interaction. FUMA reported 383 associated genes, 313 of which were not discovered 20 in MAGMA ( Table S7). These results demonstrate the polygenic genetic architecture of 21 shape features and prioritize important genes involved in the biological pathways of brain 22 functions and diseases. 23 24 The shared genetic influences with complex brain traits and disorders. 25 We took a further look into the 43 validated regions , providing variant annotations and 26 details of shared genetic influences with other complex traits and diseases. For all the 27 independent (LD r2 < 0.1) significant variants (and variants in LD, r2 ≥ 0.6) detected in these 28 validated regions, we searched for their GWAS signals reported in the NHGRI-EBI GWAS 29 catalog60. Shape deformation of ventricular and subcortica l structures has substantial 30 regional and local genetic overlaps with complex traits and clinical endpoints. The full 31 information was presented in Table S8, and below we highlighted some regions and their 32 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint 10 reported variants and genes reported for brain structures/functions, neurological 1 disorders, psychiatric disorders, psychological traits, migraine, cognitive traits, 2 educational attainment, sleep/physical activity, osteoarthrosis/pain, Alzheimer's Disease 3 biomarkers, diabetes/kidney diseases, blood traits, blood pressure, smoking/drinking, 4 lung/liver, and lipoprotein cholesterol. 5 6 Our results were concordant with previous GWAS results for regional brain volumes and 7 cortical thickness in many genomic loci, including 3q24 (index variant rs2279829, the 8 nearest gene ZIC4), 8q24.12 (rs10283100, ENPP2), 9q31.3 (rs734250, LPAR1), 9q33.1 9 (rs10983205, ASTN2), 11q14.3 (rs1531249, FAT3), 11q23.1 (rs34077344, LINC02550), 10 11q23.3 (rs10892133, DSCAML1), 12q14.3 (rs61921502, MSRB3), 12q23.3 (rs12369969, 11 NUAK1), 12q24.22 (rs7132910, HRK), 14q22.3 (rs945270, KTN1), and 16q22.3 (rs7193665, 12 ZFHX3) (Figs. S 9-S20). For example, rs7132910 was identified to be associated with 13 hippocampal volume 61. It the current study, we found rs7132910 -associations with 14 multiple shape features of the hippocampus, particularly the subiculum subfield. 15 Additionally, rs2279829 and rs7132910 were expression quantitative trait loci (eQTLs) of 16 ZIC4 and TESC in human brain tissues 62, suggesting that these shape -associated variants 17 were known to affect gene expression levels in human brain. Among these regions, we 18 also tagged genetic variants (LD r2 ≥ 0.6) reported for risk-taking63, alcohol consumption, 19 smoking initiation 64, lung function 65, chronic obstructive pulmonary disease (COPD)66, 20 blood pressure65, and Alzheimer's disease pathologies67. 21 22 Our shape GWAS results frequently tagged regions reported for white matter 23 microstructure, including 17q24.1 (rs62072157, GNA13), 2p13.2 (rs34754475, DYSF), 24 3q28 (3:190672426_CT_C, GMNC), 5q14.2 (rs12187334, ATP6AP1L), 7p21.1 (rs4329170, 25 TWISTNB), 7p22.2 (rs1183079, GNA12), 7p22.3 (rs368699386, AMZ1), 14q32.12 26 (rs529889896, CCDC88C), 16q24.2 (rs56023709, C16orf95), and 16q24.3 (rs8404, CDK10) 27 (Figs. 3A and S21-S30). For example, rs62072157 was associated with (the ninth PC of) 28 the radial distance of the right lateral ventricle, and it also had significant associations 29 with the fornix (column and body) and anterior corona radiata white matter tracts (P < 4 30 × 10-9). Rs62072157 was an eQTL of GNA13 and RGS9 in brain tissues62. The GNA13 was a 31 core gene involved in early brain development regulations68 and has been implicated in 32 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint 11 brain diseases such as schizophrenia69. In addition, rs1183079 was a brain eQTL of AMZ1, 1 which was associated with mTBM2 and radial distance of the right lateral ventricle, as 2 well as multiple white matter tracts, such as the body of corpus callosum, anterior corona 3 radiata, retrolenticular part of internal capsule, posterior corona radiata, and superior 4 corona radiata (P < 5.5 × 10-11). Furthermore, rs8404 was a brain eQTL of CDK10, SPATA33, 5 VPS9D1, MC1R, and ACSF3. Rs8404 was associated with multiple shape features of the 6 left hippocampus and affected the integrity of the retrolenticular part of internal capsule 7 and superior longitudinal fasciculus tracts (P < 2 × 10 -8). The CDK10 was important for 8 neural development70. In addition to Alzheimer's disease biomarkers 48 and COPD66, we 9 found shared genetic influence s with type 2 diabetes71 and blood traits (such as plasma 10 homocysteine levels 72 and p latelet distribution width 73) on these white matte r-11 overlapping genomic loci. 12 13 In 6p22.1 and 18q21.2 regions, we observed the shared genetic influences between shape 14 features and multiple psychiatric disorders and psychological traits . For example, we 15 tagged rs7766356 (nearest gene ZSCAN23, 6p22.1) and rs11665242 (DCC, 18q21.2), which 16 have been implicated with schizophrenia 74,75 (Figs. S31-S32). Rs7766356 was an eQTL of 17 ZSCAN23, ZSCAN31, ZKSCAN3, and ZSCAN26, which might represent drug targets for 18 schizophrenia76. We also tagged risk variants for bipolar disorder77 (e.g., rs144447022) 19 and MDD78 (e.g., rs926552) in these two regions and for neuroticism79 in 5q14.3 (e.g., 20 rs16902900, TMEM161B) (Fig. S33). In 2q24.2, 6q22.32, 8p11.21 regions (Fig. 3B and Figs. 21 S34-S36), as well as the 6p22.1 and 18q21.2 regions related to brain disorders, we found 22 shared genetic influences with a wide range of cognitive and education al traits, such as 23 intelligence80 (e.g., rs 2268894, DPP4, 2q24.2), educational attainment 81 (e.g., 24 rs11759026, CENPW, 6q22.32; rs2974312, SMIM19, 8p11.21), self-reported math ability81 25 (e.g., rs71559051, H2BC15, 6p22.1), and verbal-numerical reasoning82 (e.g., rs62100026, 26 DCC, 18q21.2). Finally, we tagged sleep-related variants in 10p12.31 and 11q14.1 regions, 27 such as insomnia83 (e.g., rs12251016, MLLT10, 10p12.31; and rs667730, DLG2, 11q14.1, 28 Figs. S37 -S38). These results suggest the shape features could be used as imaging 29 biomarkers to study the etiologic study of brain-related diseases and complex traits. 30 31 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint 12 Our results also help us to better understand the genetic links between brain atrophy and 1 the health of other organs. For example, the index rs10901814 (FAM53B, 10q26.13) was 2 associated with the shape features in the hippocampus and lateral ventricles and it was a 3 brain eQTL for EEF1AKMT2 and LHPP genes. In this region, we tagged risk variant (index 4 variant rs4962691 for estimated glomerular filtration rate (eGFR) 84, which was a clinical 5 biomarker for kidney function and disease (Fig. 3C ). Brain and kidney had similar 6 hemodynamic mechanisms and shared physiological links 85. Cognitive impairment and 7 accompanied brain structural changes (such as hippocampus volume) have been 8 frequently reported in chronic kidney disease 86,87. To precisely localize the pattern of 9 genetic effects on brain shapes, we took the eGFR lead index rs4962691 and performed 10 vertex-wise analysis on spatial maps of hippocampus and lateral ventricles. We found that 11 the rs4962691-related shape deformation was mainly localized to specific areas of the 12 hippocampus and lateral ventricles, such as the hippocampal tail and CA1, as well as the 13 atrium and posterior horn of lateral ventricles (Fig. 3D). These genetic overlaps and local 14 structural variations may represent the mediated brain changes related to the cognitive 15 impairment in chronic kidney disease. 16 17 Genetic correlations with complex traits and clinical outcomes 18 We explored genetic correlations (GC) between shape features and a wide range of other 19 complex traits. First, we used LDSC88 to examine pairwise genetic correlation between the 20 457 shape features and 211 brain structural traits, including 101 regional brain volumes35 21 and 110 diffusion tensor imaging (DTI) parameters45. Among the 96,427 (457 × 211) tests, 22 16.22% were significant at the FDR 5% level (Fig. S39 and Table S9). Both regional brain 23 volumes and DTI parameters had significant genetic correlations with ventricular and 24 subcortical shape features. For DTI parameters, the strongest associations were observed 25 on the lateral ventricles, and the top 5 associated white matter tracts included the fornix, 26 body of corpus callosum, superior corona radiata, posterior corona radiata, and posterior 27 limb of internal capsule (P < 1.49 × 10-55). Meanwhile, either the fornix or body of corpus 28 callosum tracts consistently had the strongest associations with the subcortical structures, 29 such as the hippocampus (P = 9.67 × 10-25), nucleus accumbens (P = 2.18 × 10-23), amygdala 30 (P = 4.51 × 10-8), caudate (P = 1.32 × 10-19), pallidum (P = 9.36 × 10-16), putamen (P = 2.73 31 × 10-15), and thalamus (P = 8.95 × 10-38). Our subfield analysis further revealed that the 32 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint 13 fornix and body of corpus callosum associations were mainly localized in the 1 presubiculum subfield of hippocampus (P < 3.68 × 10 -14). The fornix located below the 2 corpus callosum and connected the hippocampus to subcortical structures89. Our results 3 suggested that the fornix integrity and shape deformation s had strong genetic overlaps. 4 For regional brain volumes, we found associations for volumes of both cortical and 5 subcortical structures. As expected, the ventricular and subcortical structures had strong 6 genetic correlations with subcortical volumes. In addition, genetic correlations were also 7 widely observed between cortical structures and ventricular and subcortical shapes. Top-8 ranked cortical volumes included the left/right insula ( with putamen, P < 7.61 × 10-18), 9 left/right isthmus cingulate (with ventricle and hippocampus presubiculum, P < 9.77 × 10-10 10), left/right lingual (with ventricle, P < 2.70 × 10 -14), left/right pericalcarine (with 11 ventricle, P < 3.41 × 10-12), and left/right cuneus (with ventricle, P < 2.37 × 10-10). Overall, 12 these results suggest that ventricular and subcortical shape features are genetically 13 related to white matter integrity and structural variations of cortical regions. 14 15 Next, we examined genetic correlations between the 457 shape features and 48 complex 16 traits and diseases. At the FDR 5% level (457 × 48 tests), we found the shape features were 17 associated with brain disorders (such as attention-deficit/hyperactivity disorder (ADHD), 18 schizophrenia, and anorexia nervosa ), cognitive traits (such as cognitive function, 19 intelligence, and reaction time ), sleep traits (such as snoring, insomnia, extreme 20 chronotype), neuroticism, risk-taking, metabolic traits, and cardiovascular diseases (such 21 as hypertension and coronary artery disease (CAD)) (Fig. S40 and Table S10). For example, 22 ADHD was positively correlated with shape features of the left hippocampus and lateral 23 ventricles (|GC| > 0.184, P < 5.69 × 10-4, Figs. 4A-4B). In ADHD, there have been reports 24 of abnormalities in hippocampal structures, possibly as a result of the brain's efforts to 25 compensate for disruptions of time perception and a tendency to avoid waiting 90. The 26 Rotterdam Study reported that ADHD -related genetic variants were associated with 27 structural brain changes in the lateral ventricles 91. In addition, schizophrenia was 28 genetically associated with shape features of the thalamus and hippocampus (| GC| > 29 0.133, P < 6.14 × 10-4). Both the thalamus and hippocampus had crucial roles in functional 30 and structural pathways related to schizophrenia92 and smaller volume s in the two 31 structures were frequently reported in schizophrenia patients93,94. An earlier study using 32 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint 14 subcortical brain volumes was not able to detect genetic overlap between schizophrenia 1 risk and subcortical structures95. Furthermore, most significant genetic correlations with 2 cognitive function, intelligence, and education were with the lateral ventricles (|GC| > 3 0.113, P 0.115, P < 7.90 × 10-4, Fig. 4C). Ventricular enlargement was strongly 5 correlated with cognitive performance decline96. The t halamus passed information 6 between the brain and body and anticipatory thalamic activity can predict reaction time97. 7 We also observed specific genetic correlations for other traits, such as between insomnia 8 and the caudate (|GC| > 0.141, P 0.283, P 0.148, 10 P 0.260, P < 6.40 × 10-4, Fig. 4D). CAD had long-12 term negative impact on brain health and reduced neural connectivity changes in the 13 hippocampus were observed in CAD and may contribute to cognitive impairment 98. A 14 significant genetic correlation between AD and the hippocampus was not observed. 15 16 Phenome-wide association study using shape polygenic risk scores 17 We tested associations between the 457 shape features and more complex traits and 18 clinical outcomes using polygenic risk scores (PRS) of shape features in the UKB non -19 imaging cohort (Methods). A total of 276 complex traits and clinical outcomes were 20 selected from a variety of categories (Table S11). Briefly, we constructed PRS using PRS-21 CS99 for shape features on unrelated UKB subjects without brain MRIs (n = 379,860, also 22 removing relatives of imaging subjects). We then focused on the UKB white British 23 subjects and randomly selected 70% individuals (average n = 202,405) as discovery 24 sample to test pairwise associations (457 × 276 tests) and validated these results in a hold-25 out dataset consisting of the rest of 30% white British subjects (average n = 86,736), UKB 26 white non-British subjects (average n = 20,746), and UKB non-white subjects (average n = 27 21,587). Detailed information on the adjusted covariates can be found in the Methods 28 section. We prioritized significant associations in the validation sample (at the nominal 29 significance level) with concordant regression coefficients when they passed the 30 Bonferroni significance level in the discovery sample. Among the 457 × 276 tests, 2,907 31 were significant at Bonferroni significance level in the discovery sample, among which 32 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint 15 94.60% ( 2,750) were validated (Fig. S 41 and Table S 12). We highlighted below the 1 association patterns with clinical outcomes, mental health, cognitive function, physical 2 activity, lifestyle, and biomarkers. 3 4 We observed significant associations between shape PRS and multiple diseases, including 5 diabetes, hyperthyroidism, hypothyroidism, multiple sclerosis , psoriasis, and vascular 6 heart problems. For example, diabetes was significantly associated with PRS for multiple 7 shape features of the left hippocampus and its subfields (|b | > 0.0109, P < 2.47 × 10-7). 8 These findings were consistent with previous studies showing patients with long diabetes 9 had higher risk of hippocampal atrophy, and loss of hippocampal neuroplasticity and 10 neurogenesis100,101. In addition, both hyperthyroidism and hypothyroidism were mostly 11 associated with shape PRS of the lateral ventricles and hippocampus (and hippocampal 12 subfields) (|b | > 0.0107, P < 3.87 × 10-7). Adults with hypothyroidism were reported to 13 have decreased hippocampal volume 102 and people with hyperthyroidism ha d smaller 14 grey matter volume in bilateral hippocampus 103. Hyperthyroidism and hypothyroidism 15 had also been linked to changes of brain ventricle size104. Multiple sclerosis was mostly 16 associated with shape features of the putamen and hippocampus (|b | > 0.0108, P < 3.42 17 × 10-7). Decreased putamen and hippocampal volume s have been reported in multiple 18 sclerosis patients105,106. We also found associations between vascular heart problems and 19 shape PRS of the hippocampus (|b | > 0.0105, P < 3.71 × 10-7), consistent with previous 20 studies showing decreased hippocampal volume among patients with vascular heart 21 problems107. 22 23 There were significant associations between shape PRS and multiple mental health traits 24 related to anxiety, depression, and neuroticism (|b | > 0.0106, P < 1.09 × 10-11). Most of 25 these mental health traits were associated with shape features of the hippocampus and 26 its subfields. These observed associations were consistent with recent findings on 27 reduced hippocampal volume in subiculum 108 and fimbria109 among patients with MDD. 28 Significant associations were also found between mental health and shape PRS of other 29 brain structures, such as between nervous feelings and the nucleus accumbens, 30 neuroticism and the putamen, as well as tiredness/lethargy and the caudate. 31 32 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint 16 The blood biochemistry biomarkers were widely associated with the shape PRS of all 1 structures, with the majority of associations involving the hippocampus, hippocampal 2 subfields, and lateral ventricle s. Examples of associated biomarkers included 3 apolipoprotein A, aspartate aminotransferase , glycated haemoglobin (HbA1c) , high-4 density lipoprotein ( HDL) cholesterol, insulin-like growth factor 1 ( IGF-1), total protein , 5 and urate (|b | > 0.009, P < 4.28 × 10 -42). It is known that urate cause d hippocampal 6 infection, which in turn induce d cognitive dysfunction 110. HbA1c measure d the blood 7 sugar level and was used in diagnosis of diabetes. Our results were consistent with a 8 recent study that higher level of HbA1c was associated with smaller hippocampal 9 volume111. In summary, shape PRS uncovered the links between shape features and a 10 wide range of complex traits and disease s. As the shape PRS were genetically predicted 11 traits, the se observed associations also indicate the widespread underlying shared 12 genetic influences. 13 14 Causal relationships with clinical endpoints detected by Mendelian randomization. 15 To explore the causes and consequences of shape deformation s, Mendelian 16 randomization (MR) was used to identify potential causal relationships between the 457 17 shape features and 288 clinical endpoints collected by FinnGen 112 and the Psychiatric 18 Genomics Consortium113 (Table S13). We tested 14 different MR methods114-117, and the 19 detailed implementation information can be found in the Methods section. 20 21 At the Bonferroni significance level (P < 3.92 × 10-8), we found strong evidence of genetic 22 causal effects from shape features to brain disorders, including AD, schizophrenia, and 23 cross disorders (five major psychiatric disorders118) (Fig. 5A and Table S14). For example, 24 multiple ventricular shape statistics had significant genetic causal effects on Alzheimer’s 25 disease (|b | > 0.474, P < 2.73 × 10-8), most of which were from the left lateral ventricle. 26 AD can affect gray and white matter structures surrounding the ventricles, and ventricular 27 enlargement and expansion have been frequently identified in AD9,119-121. Our findings 28 provide further evidence for the causal genetic pathway underlying brain structural 29 variations and Alzheimer’s disease. There was a consistent sign of causal genetic effects 30 across different MR methods. In addition, shape features of the putamen and 31 hippocampus were causally linked to schizophrenia (|b| > 0.339, P < 1.97 × 10-8). In 32 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint 17 schizophrenia, neuropsychological impairments have been associated with hippocampal 1 structures94. It is well known that the putamen is associated with both increased 2 dopamine synthesis capacity and frontostriatal dysconnectivity in schizophrenia, as well 3 as with antipsychotic treatment effects122. Additionally, we did not detect causal genetic 4 links to schizophrenia by analyzing the volumes of putamen and hippocampus structures 5 (|b| 0.0011). Furthermore, there were significant genetic causal relationships 6 between shape features and other diseases of the nervous system, such as carpal tunnel 7 syndrome (|b | > 0.249, P 0.256, P < 7.61 × 10-14). All 8 of the above results passed the MR -Egger intercept test, indicating the absence of 9 horizontal pleiotropy. 10 11 Additionally, we identified causal relationships where sleep traits were the exposure and 12 brain structural traits were the outcome at Bonferroni significance level (P < 1.12×10-8). A 13 large proportion of the significant findings (26/42) were related to the diseases of the 14 circulatory system, such as aortic aneurysm, calcific aortic valvular stenosis, heart failure, 15 and hypertensive heart disease ( Fig. 5B and Table S14 ). For example, calcific aortic 16 valvular stenosis had causal genetic links to the shape of the pallidum (|b | > 0.059, P 0.075, P < 4.28 × 10 -9). Multiple brain disorders, such as stroke 123 and 19 dementia124, can be caused by diseases of the circulatory system. We also found that 20 COPD was causally related to the shape of hippocampus and lateral ventricles (| b | > 21 0.064, P < 9.87 × 10-9). Cognitive impairments and related alterations of hippocampus had 22 been found in COPD patients 125. Overall, our MR results suggest that genetic causal 23 pathways may exist between brain shapes and brain disorders, such as Alzheimer’s 24 disease. These findings also reveal possible genetic mechanisms of non -brain diseases 25 (e.g., heart diseases) that underlie brain health. 26 27

28 Brain volumes are commonly used to analyze brain ventricular and subcortical structures. 29 Increasing evidence, however, indicates that volumetric measurements can only partially 30 capture structural complexity. As a result, shape analysis has been performed to uncover 31 deformations that are not visible in volumetric analysis. Using over 45,000 MRI scans in 32 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint 18 the UKB and ABCD studies , our study uncovered the genetic architecture of shape 1 features within the ventricular and subcortical regions. We identified genetic connections 2 between shape features and a wide range of complex traits and clinical outcomes. 3 Leveraging shape features identified new loci that could not be identified by brain 4 volumes and provided fine details for localizing the genetic effect patterns within brain 5 structures. The results of our study improved the spatial resolution for identifying 6 genetically important brain areas that influence d clinical outcomes. For example, 7 rs4962691, a risk variant for eGFR, ha d genetic effects in specific parts of the 8 hippocampus and lateral ventricles. These shape features may be used as 9 endophenotypes of the cognitive impairment in kidney disease . In summary, as one of 10 the first large-scale studies to examine the genetic architecture of ventricular and 11 subcortical shape features, our results provide specific shape biomarkers that can be used 12 in clinical research questions. 13 14 The present study has a few limitations. First, the current analysis mainly used data from 15 the European UKB subjects, which may limit the generalizability of our research findings. 16 In the validation analysis, we have observed that European -significant genetic variants 17 tended to have larger effects in non -European cohorts of the ABCD study. The inclusion 18 of more global samples and identification of cross -population components of genetic 19 effects on brain shape will be of great interest in future studies. Second, we used PCA to 20 extract low -rank features from ve rtex-wise maps, which can capture local shape 21 deformations while mitigating multiple testing burden for genome-wise testing. PCA is a 22 powerful statistical tool to extract linear structures in the data. Nevertheless, PCA may 23 not be the most efficient method for reducing dimensions because complicated shape 24 variations can be non -linear. It might be possible to generate more powerful shape 25 features using variational autoencoder 126 and transfer learning 127. Lastly, we studied 26 hippocampal subfields, which may reflect specific biological processes and cognitive 27 functions128,129. Other subcortical structures (such as amygdala130) and lateral ventricles12 28 can also be divided into different portions or subregions with distinct functions . The 29 development and application of automated segmentation methods in large -scale MRI 30 datasets will enable the discovery of genetic influences at the subfield level for more brain 31 structures. 32 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint 19 1

2

are available in the Methods section. 3 Note: One supplementary information pdf file, one supplementary figure pdf file, and one 4 supplementary table zip file are available. 5 6

7 This research was partially supported by U.S. NIH grants MH086633 (HT.Z.), MH116527 8 (TF.L.), and U01HG011720 (Y.L.). We thank the individuals represented in the UK Biobank 9 and ABCD studies for their participation and the research teams for their work in 10 collecting, processing and disseminatin g these datasets for analysis. We thank Doug 11 Crabill for helpful conversations on computing. We would like to thank the University of 12 North Carolina at Chapel Hill and Purdue University and their research computing group 13 for providing computational resources and support that have contributed to these 14 research results. We gratefully acknowledge all the studies and databases that made 15 GWAS summary data available. This research has been conducted using the UK Biobank 16 resource (application number 22783), subject to a data transfer agreement. Part of the 17 data used in the preparation of this article were obtained from the Adolescent Brain 18 Cognitive Development (ABCD) Study (https://abcdstudy.org), held in the NIMH Data 19 Archive (NDA). This is a multisite, longitudinal study designed to recruit more than 10,000 20 children age 9-10 and follow them over 10 years into early adulthood. The ABCD Study is 21 supported by the National Institutes of Health and additional federal partners under 22 award numbers U01DA041022, U01DA041028, U01DA041048, U01DA041089, 23 U01DA041106, U01DA041117, U01DA041120, U01DA041134, U01DA041148, 24 U01DA041156, U01DA041174, U24DA041123, U24DA041147, U01DA041093, and 25 U01DA041025. A full list of supporters is available at https://abcdstudy.org/federal -26 partners.html. A listing of participati ng sites and a complete listing of the study 27 investigators can be found at https://abcdstudy.org/scientists/workgroups/. ABCD 28 consortium investigators designed and implemented the study and/or provided data but 29 did not necessarily participate in analysis or writing of this report. This manuscript reflects 30 the views of the authors and may not reflect the opinions or views of the NIH or ABCD 31 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint 20 consortium investigators. Assistance for this project was provided by the UNC Intellectual 1 and Developmental Disabilities Research Center (NICHD; P50 HD103573). 2 3 AUTHOR CONTRIBUTIONS 4 B.Z. and H.Z. designed the study. B.Z., TF.L., X.Y., J.S., X.W., TY. L, Y.Y., Z.W., Z.F., and Z. J. 5 analyzed the data. TF. L., X.W., TY. L, Y.Y., J. C., Y.S., J.T., D.X, Z.Z., M.G., W.G., and C.T. 6 processed the MRI data. Y.W. and Q.D. helped on the shape feature pipelines. Y.L. and 7 J.L.S. provided feedback on study design and results interpretation s. B.Z. wrote the 8 manuscript with feedback from all authors. 9 10 CORRESPINDENCE AND REQUESTS FOR MATERIALS should be addressed to H.Z. 11 12 COMPETETING FINANCIAL INTERESTS 13 The authors declare no competing financial interests. 14 15

16 1. Scelsi, C. et al. The lateral ventricles: a detailed review of anatomy, development, 17 and anatomic variations. American Journal of Neuroradiology 41, 566-572 18 (2020). 19 2. Crosson, B. Subcortical Functions in Cognition. Vol. 31 419-421 (Springer, 2021). 20 3. Johnson, M.H. Subcortical face processing. Nature Reviews Neuroscience 6, 766-21 774 (2005). 22 4. Utter, A.A. & Basso, M.A. The basal ganglia: an overview of circuits and function. 23 Neuroscience & Biobehavioral Reviews 32, 333-342 (2008). 24 5. Berridge, K.C. & Kringelbach, M.L. Pleasure systems in the brain. Neuron 86, 646-25 664 (2015). 26 6. Bressler, S.L. & Menon, V. Large-scale brain networks in cognition: emerging 27

and principles. Trends in cognitive sciences 14, 277-290 (2010). 28 7. Doyon, J. & Benali, H. Reorganization and plasticity in the adult brain during 29 learning of motor skills. Current opinion in neurobiology 15, 161-167 (2005). 30 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint 21 8. Tang, X. et al. A fully-automated subcortical and ventricular shape generation 1 pipeline preserving smoothness and anatomical topology. Frontiers in 2 neuroscience 12, 321 (2018). 3 9. Thompson, P.M. et al. Mapping hippocampal and ventricular change in 4 Alzheimer disease. Neuroimage 22, 1754-1766 (2004). 5 10. Wang, Y., Chan, T.F., Toga, A.W. & Thompson, P.M. Multivariate tensor-based 6 brain anatomical surface morphometry via holomorphic one-forms. in 7 International conference on medical image computing and computer-assisted 8 intervention 337-344 (Springer, 2009). 9 11. Wang, Y. et al. Surface-based TBM boosts power to detect disease effects on the 10 brain: an N= 804 ADNI study. Neuroimage 56, 1993-2010 (2011). 11 12. Thompson, P.M. et al. ENIGMA and global neuroscience: A decade of large-scale 12 studies of the brain in health and disease across more than 40 countries. 13 Translational psychiatry 10, 1-28 (2020). 14 13. Ge, T. et al. Multidimensional heritability analysis of neuroanatomical shape. 15 Nature communications 7, 13291 (2016). 16 14. Ching, C.R. et al. Sex-dependent age trajectories of subcortical brain structures: 17 analysis of large-scale percentile models and shape morphometry. in 16th 18 International Symposium on Medical Information Processing and Analysis Vol. 19 11583 44-54 (SPIE, 2020). 20 15. Tang, X. et al. Shape abnormalities of subcortical and ventricular structures in 21 mild cognitive impairment and Alzheimer's disease: detecting, quantifying, and 22 predicting. Human brain mapping 35, 3701-3725 (2014). 23 16. Cong, S. et al. Surface-based morphometric analysis of hippocampal subfields in 24 mild cognitive impairment and Alzheimer's disease. in 2015 IEEE 58th 25 International Midwest Symposium on Circuits and Systems (MWSCAS) 1-4 (IEEE, 26 2015). 27 17. Jamea, A.A. et al. Volumetric and Shape Analysis of the Subcortical Regions in 28 Schizophrenia Patients: A Pilot Study. Journal of Clinical Imaging Science 9(2019). 29 18. Gutman, B.A. et al. A meta-analysis of deep brain structural shape and 30 asymmetry abnormalities in 2,833 individuals with schizophrenia compared with 31 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint 22 3,929 healthy volunteers via the ENIGMA Consortium. Human brain mapping 43, 1 352-372 (2022). 2 19. Kim, J.B., Suh, S.-I. & Kim, J.H. Volumetric and shape analysis of hippocampal 3 subfields in unilateral mesial temporal lobe epilepsy with hippocampal atrophy. 4 Epilepsy research 117, 74-81 (2015). 5 20. Lu, Y. et al. The volumetric and shape changes of the putamen and thalamus in 6 first episode, untreated major depressive disorder. NeuroImage: Clinical 11, 658-7 666 (2016). 8 21. Ho, T.C. et al. Subcortical shape alterations in major depressive disorder: 9 Findings from the ENIGMA major depressive disorder working group. Human 10 brain mapping 43, 341-351 (2022). 11 22. Ching, C.R. et al. Mapping subcortical brain alterations in 22q11. 2 deletion 12 syndrome: Effects of deletion size and convergence with idiopathic 13 neuropsychiatric illness. American Journal of Psychiatry 177, 589-600 (2020). 14 23. Quigley, S.J. et al. Volume and shape analysis of subcortical brain structures and 15 ventricles in euthymic bipolar I disorder. Psychiatry Research: Neuroimaging 233, 16 324-330 (2015). 17 24. Roalf, D.R. et al. Heritability of subcortical and limbic brain volume and shape in 18 multiplex-multigenerational families with schizophrenia. Biological psychiatry 77, 19 137-146 (2015). 20 25. Styner, M. et al. Morphometric analysis of lateral ventricles in schizophrenia and 21 healthy controls regarding genetic and disease-specific factors. Proceedings of 22 the National Academy of Sciences 102, 4872-4877 (2005). 23 26. Biton, A. et al. Polygenic architecture of human neuroanatomical diversity. 24 bioRxiv, 592337 (2019). 25 27. Zhao, B. et al. Heritability of regional brain volumes in large-scale neuroimaging 26 and genetic studies. Cerebral Cortex 29, 2904-2914 (2018). 27 28. Roshchupkin, G.V. et al. Heritability of the shape of subcortical brain structures 28 in the general population. Nature communications 7(2016). 29 29. Miller, K.L. et al. Multimodal population brain imaging in the UK Biobank 30 prospective epidemiological study. Nature Neuroscience 19, 1523-1536 (2016). 31 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint 23 30. Satizabal, C.L. et al. Genetic architecture of subcortical brain structures in 38,851 1 individuals. Nature genetics 51, 1624-1636 (2019). 2 31. Vojinovic, D. et al. Genome-wide association study of 23,500 individuals 3 identifies 7 loci associated with brain ventricular volume. Nature 4 communications 9, 1-11 (2018). 5 32. Hibar, D.P. et al. Novel genetic loci associated with hippocampal volume. Nature 6 Communications 8, 13624 (2017). 7 33. Hibar, D.P. et al. Common genetic variants influence human subcortical brain 8 structures. Nature 520, 224-229 (2015). 9 34. Van der Meer, D. et al. Brain scans from 21,297 individuals reveal the genetic 10 architecture of hippocampal subfield volumes. Molecular psychiatry 25, 3053-11 3065 (2020). 12 35. Zhao, B. et al. Genome-wide association analysis of 19,629 individuals identifies 13 variants influencing regional brain volumes and refines their genetic co-14 architecture with cognitive and mental health traits. Nature genetics 51, 1637-15 1644 (2019). 16 36. Medland, S.E. et al. Ten years of enhancing neuro-imaging genetics through 17 meta-analysis: An overview from the ENIGMA Genetics Working Group. Human 18 Brain Mapping 43, 292-299 (2022). 19 37. Elliott, L.T. et al. Genome-wide association studies of brain imaging phenotypes 20 in UK Biobank. Nature 562, 210-216 (2018). 21 38. van der Meer, D. et al. Understanding the genetic determinants of the brain with 22 MOSTest. Nature communications 11, 1-9 (2020). 23 39. Jansen, P.R. et al. Genome-wide meta-analysis of brain volume identifies 24 genomic loci and genes shared with intelligence. Nature communications 11, 1-25 12 (2020). 26 40. Smith, S.M. et al. An expanded set of genome-wide association studies of brain 27 imaging phenotypes in UK Biobank. Nature neuroscience 24, 737-745 (2021). 28 41. Chaarani, B. et al. Baseline brain function in the preadolescents of the ABCD 29 Study. Nature Neuroscience, 1-11 (2021). 30 42. Gutman, B.A. et al. Medial demons registration localizes the degree of genetic 31 influence over subcortical shape variability: an N= 1480 meta-analysis. in 2015 32 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint 24 IEEE 12th International Symposium on Biomedical Imaging (ISBI) 1402-1406 1 (IEEE, 2015). 2 43. Couvy-Duchesne, B. et al. Parsimonious model for mass-univariate vertexwise 3 analysis. Journal of Medical Imaging 9, 052404 (2022). 4 44. Yang, J., Lee, S.H., Goddard, M.E. & Visscher, P.M. GCTA: a tool for genome-wide 5 complex trait analysis. The American Journal of Human Genetics 88, 76-82 6 (2011). 7 45. Zhao, B. et al. Common genetic variation influencing human white matter 8 microstructure. Science 372(2021). 9 46. Bulik-Sullivan, B.K. et al. LD Score regression distinguishes confounding from 10 polygenicity in genome-wide association studies. Nature genetics 47, 291-295 11 (2015). 12 47. Watanabe, K., Taskesen, E., Bochoven, A. & Posthuma, D. Functional mapping 13 and annotation of genetic associations with FUMA. Nature Communications 8, 14 1826 (2017). 15 48. Cruchaga, C. et al. GWAS of cerebrospinal fluid tau levels identifies risk variants 16 for Alzheimer’s disease. Neuron 78, 256-268 (2013). 17 49. de Leeuw, C.A., Mooij, J.M., Heskes, T. & Posthuma, D. MAGMA: generalized 18 gene-set analysis of GWAS data. PLoS Computational Biology 11, e1004219 19 (2015). 20 50. Zhao, B. et al. Genetic influences on the intrinsic and extrinsic functional 21 organizations of the cerebral cortex. medRxiv (2021). 22 51. Tissink, E. et al. Specificity and overlap in the genetic architectures of functional 23 and structural connectivity within cerebral resting-state networks. bioRxiv 24 (2022). 25 52. Gelernter, J. et al. Genome-wide association study of cocaine dependence and 26 related traits: FAM53B identified as a risk gene. Molecular psychiatry 19, 717-27 723 (2014). 28 53. Okbay, A. et al. Genetic variants associated with subjective well-being, 29 depressive symptoms, and neuroticism identified through genome-wide 30 analyses. Nature Genetics 48, 624–633 (2016). 31 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint 25 54. Nishikimi, M., Oishi, K., Tabata, H., Torii, K. & Nakajima, K. Segregation and 1 pathfinding of callosal axons through EphA3 signaling. Journal of Neuroscience 2 31, 16251-16260 (2011). 3 55. Day, B.W. et al. EphA3 maintains tumorigenicity and is a therapeutic target in 4 glioblastoma multiforme. Cancer cell 23, 238-248 (2013). 5 56. Tissir, F. & Goffinet, A.M. Planar cell polarity signaling in neural development. 6 Current opinion in neurobiology 20, 572-577 (2010). 7 57. Shin, J. et al. Global and regional development of the human cerebral cortex: 8 Molecular architecture and occupational aptitudes. Cerebral Cortex 30, 4121-9 4139 (2020). 10 58. Gaston-Massuet, C., Henderson, D.J., Greene, N.D. & Copp, A.J. Zic4, a zinc-finger 11 transcription factor, is expressed in the developing mouse nervous system. 12 Developmental dynamics: an official publication of the American Association of 13 Anatomists 233, 1110-1115 (2005). 14 59. Mi, H., Muruganujan, A., Ebert, D., Huang, X. & Thomas, P.D. PANTHER version 15 14: more genomes, a new PANTHER GO-slim and improvements in enrichment 16 analysis tools. Nucleic acids research 47, D419-D426 (2019). 17 60. Buniello, A. et al. The NHGRI-EBI GWAS Catalog of published genome-wide 18 association studies, targeted arrays and summary statistics 2019. Nucleic Acids 19 Research 47, D1005-D1012 (2018). 20 61. Stein, J.L. et al. Identification of common variants associated with human 21 hippocampal and intracranial volumes. Nature Genetics 44, 552-561 (2012). 22 62. de Klein, N. et al. Brain expression quantitative trait locus and network analysis 23 reveals downstream effects and putative drivers for brain-related diseases. 24 bioRxiv (2021). 25 63. Linnér, R.K. et al. Genome-wide association analyses of risk tolerance and risky 26 behaviors in over 1 million individuals identify hundreds of loci and shared 27 genetic influences. Nature Genetics 51, 245-257 (2019). 28 64. Liu, M. et al. Association studies of up to 1.2 million individuals yield new insights 29 into the genetic etiology of tobacco and alcohol use. Nature genetics 51, 237-244 30 (2019). 31 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint 26 65. Kichaev, G. et al. Leveraging polygenic functional enrichment to improve GWAS 1 power. The American Journal of Human Genetics 104, 65-75 (2019). 2 66. Sakornsakolpat, P. et al. Genetic landscape of chronic obstructive pulmonary 3 disease identifies heterogeneous cell-type and phenotype associations. Nature 4 genetics 51, 494-505 (2019). 5 67. Wang, H. et al. Genome-wide interaction analysis of pathological hallmarks in 6 Alzheimer's disease. Neurobiology of aging 93, 61-68 (2020). 7 68. Xiang, B. et al. The role of genes affected by human evolution marker GNA13 in 8 schizophrenia. Progress in Neuro-Psychopharmacology and Biological Psychiatry 9 98, 109764 (2020). 10 69. Hirayama-Kurogi, M. et al. Downregulation of GNA13-ERK network in prefrontal 11 cortex of schizophrenia brain identified by combined focused and targeted 12 quantitative proteomics. Journal of Proteomics 158, 31-42 (2017). 13 70. Düster, R., Ji, Y., Pan, K.-T., Urlaub, H. & Geyer, M. Functional characterization of 14 the human Cdk10/Cyclin Q complex. Open Biology 12, 210381 (2022). 15 71. Vujkovic, M. et al. Discovery of 318 new risk loci for type 2 diabetes and related 16 vascular outcomes among 1.4 million participants in a multi-ancestry meta-17 analysis. Nature genetics 52, 680-691 (2020). 18 72. Shane, B. et al. The 677C→ T variant of MTHFR is the major genetic modifier of 19 biomarkers of folate status in a young, healthy Irish population. The American 20 journal of clinical nutrition 108, 1334-1341 (2018). 21 73. Vuckovic, D. et al. The polygenic and monogenic basis of blood traits and 22 diseases. Cell 182, 1214-1231. e11 (2020). 23 74. Bigdeli, T.B. et al. Genome-wide association studies of schizophrenia and bipolar 24 disorder in a diverse cohort of US veterans. Schizophrenia bulletin 47, 517-529 25 (2021). 26 75. Periyasamy, S. et al. Association of schizophrenia risk with disordered niacin 27 metabolism in an Indian genome-wide association study. JAMA psychiatry 76, 28 1026-1034 (2019). 29 76. Lago, S.G. & Bahn, S. The druggable schizophrenia genome: from repurposing 30 opportunities to unexplored drug targets. NPJ genomic medicine 7, 1-13 (2022). 31 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint 27 77. Wu, Y. et al. Multi-trait analysis for genome-wide association study of five 1 psychiatric disorders. Translational psychiatry 10, 1-11 (2020). 2 78. Cai, N. et al. Minimal phenotyping yields genome-wide association signals of low 3 specificity for major depression. Nature Genetics 52, 437-447 (2020). 4 79. Nagel, M. et al. Meta-analysis of genome-wide association studies for 5 neuroticism in 449,484 individuals identifies novel genetic loci and pathways. 6 Nature Genetics 50, 920 (2018). 7 80. Savage, J.E. et al. Genome-wide association meta-analysis in 269,867 individuals 8 identifies new genetic and functional links to intelligence. Nature Genetics 50, 9 912-919 (2018). 10 81. Lee, J.J. et al. Gene discovery and polygenic prediction from a genome-wide 11 association study of educational attainment in 1.1 million individuals. Nature 12 Genetics 50, 1112–1121 (2018). 13 82. de la Fuente, J., Davies, G., Grotzinger, A.D., Tucker-Drob, E.M. & Deary, I.J. A 14 general dimension of genetic sharing across diverse cognitive traits inferred from 15 molecular data. Nature Human Behaviour 5, 49-58 (2021). 16 83. Jansen, P.R. et al. Genome-wide analysis of insomnia in 1,331,010 individuals 17 identifies new risk loci and functional pathways. Nature Genetics 51, 394-403 18 (2019). 19 84. Morris, A.P. et al. Trans-ethnic kidney function association study reveals putative 20 causal genes and effects on kidney-specific disease aetiologies. Nature 21 communications 10, 1-14 (2019). 22 85. Afsar, B. et al. Brain-kidney cross-talk: Definition and emerging evidence. 23 European journal of internal medicine 36, 7-12 (2016). 24 86. Chang, C.Y. et al. Cognitive impairment and hippocampal atrophy in chronic 25 kidney disease. Acta Neurologica Scandinavica 136, 477-485 (2017). 26 87. Vemuri, P. et al. Association of kidney function biomarkers with brain MRI 27 findings: the BRINK study. Journal of Alzheimer's Disease 55, 1069-1082 (2017). 28 88. Bulik-Sullivan, B. et al. An atlas of genetic correlations across human diseases 29 and traits. Nature Genetics 47, 1236-1241 (2015). 30 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint 28 89. Choi, Y.J., Lee, E.J. & Lee, J.E. The fornix: functional anatomy, normal 1 neuroimaging, and various pathological conditions. Investigative Magnetic 2 Resonance Imaging 25, 59-75 (2021). 3 90. Plessen, K.J. et al. Hippocampus and amygdala morphology in attention-4 deficit/hyperactivity disorder. Archives of general psychiatry 63, 795-807 (2006). 5 91. Vilor-Tejedor, N. et al. Aging-Dependent Genetic Effects Associated to ADHD 6 Predict Longitudinal Changes of Ventricular Volumes in Adulthood. Frontiers in 7 psychiatry 11, 574 (2020). 8 92. Pergola, G., Selvaggi, P., Trizio, S., Bertolino, A. & Blasi, G. The role of the 9 thalamus in schizophrenia from a neuroimaging perspective. Neuroscience & 10 Biobehavioral Reviews 54, 57-75 (2015). 11 93. Csernansky, J.G. et al. Abnormalities of thalamic volume and shape in 12 schizophrenia. American Journal of Psychiatry 161, 896-902 (2004). 13 94. Harrison, P.J. The hippocampus in schizophrenia: a review of the 14 neuropathological evidence and its pathophysiological implications. 15 Psychopharmacology 174, 151-162 (2004). 16 95. Franke, B. et al. Genetic influences on schizophrenia and subcortical brain 17 volumes: large-scale proof of concept. Nature Neuroscience 19, 420-431 (2016). 18 96. Apostolova, L.G. et al. Hippocampal atrophy and ventricular enlargement in 19 normal aging, mild cognitive impairment and Alzheimer's disease. Alzheimer 20 disease and associated disorders 26, 17 (2012). 21 97. Nikulin, V.V. et al. Anticipatory activity in the human thalamus is predictive of 22 reaction times. Neuroscience 155, 1275-1283 (2008). 23 98. Niu, J. et al. Neural Dysconnectivity in the Hippocampus Correlates With White 24 Matter Lesions and Cognitive Measures in Patients With Coronary Artery 25 Disease. Frontiers in Aging Neuroscience 14(2022). 26 99. Ge, T., Chen, C.-Y., Ni, Y., Feng, Y.-C.A. & Smoller, J.W. Polygenic prediction via 27 Bayesian regression and continuous shrinkage priors. Nature Communications 28 10, 1776 (2019). 29 100. Hirabayashi, N. et al. Association between diabetes and hippocampal atrophy in 30 elderly Japanese: the Hisayama study. Diabetes care 39, 1543-1549 (2016). 31 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint 29 101. Ho, N., Sommers, M.S. & Lucki, I. Effects of diabetes on hippocampal 1 neurogenesis: links to cognition and depression. Neuroscience & Biobehavioral 2 Reviews 37, 1346-1362 (2013). 3 102. Cooke, G.E., Mullally, S., Correia, N., O'Mara, S.M. & Gibney, J. Hippocampal 4 volume is decreased in adults with hypothyroidism. Thyroid 24, 433-440 (2014). 5 103. Zhang, W. et al. Grey matter abnormalities in untreated hyperthyroidism: a 6 voxel-based morphometry study using the DARTEL approach. European journal 7 of radiology 83, e43-e48 (2014). 8 104. Oatridge, A. et al. Changes in brain size with treatment in patients with hyper-or 9 hypothyroidism. American journal of neuroradiology 23, 1539-1544 (2002). 10 105. Krämer, J. et al. Early and degressive putamen atrophy in multiple sclerosis. 11 International journal of molecular sciences 16, 23195-23209 (2015). 12 106. Boscheron, J. et al. Insights on the Relationship Between Hippocampal 13 Connectivity and Memory Performances at the Early Stage of Multiple Sclerosis. 14 Frontiers in neurology 12, 667531 (2021). 15 107. Barekatain, M. et al. The relationship between regional brain volumes and the 16 extent of coronary artery disease in mild cognitive impairment. Journal of 17 Research in Medical Sciences: The Official Journal of Isfahan University of 18 Medical Sciences 19, 739 (2014). 19 108. Roddy, D.W. et al. The hippocampus in depression: more than the sum of its 20 parts? Advanced hippocampal substructure segmentation in depression. 21 Biological Psychiatry 85, 487-497 (2019). 22 109. Xu, J. et al. Left fimbria atrophy is associated with hippocampal metabolism in 23 female major depressive disorder patients. in 2018 40th Annual International 24 Conference of the IEEE Engineering in Medicine and Biology Society (EMBC) 1136-25 1139 (IEEE, 2018). 26 110. Shao, X. et al. Uric acid induces cognitive dysfunction through hippocampal 27 inflammation in rodents and humans. Journal of Neuroscience 36, 10990-11005 28 (2016). 29 111. Garfield, V. et al. HbA1c and brain health across the entire glycaemic spectrum. 30 Diabetes, Obesity and Metabolism 23, 1140-1149 (2021). 31 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint 30 112. Kurki, M.I. et al. FinnGen: Unique genetic insights from combining isolated 1 population and national health register data. medRxiv (2022). 2 113. Sullivan, P.F. et al. Psychiatric genomics: an update and an agenda. American 3 Journal of Psychiatry 175, 15-27 (2017). 4 114. Hemani, G. et al. The MR-Base platform supports systematic causal inference 5 across the human phenome. elife 7(2018). 6 115. Zhao, Q., Wang, J., Hemani, G., Bowden, J. & Small, D.S. Statistical inference in 7 two-sample summary-data Mendelian randomization using robust adjusted 8 profile score. The Annals of Statistics 48, 1742-1769 (2020). 9 116. Wang, J. et al. Causal inference for heritable phenotypic risk factors using 10 heterogeneous genetic instruments. PLoS genetics 17, e1009575 (2021). 11 117. Ye, T., Shao, J. & Kang, H. Debiased inverse-variance weighted estimator in two-12 sample summary-data Mendelian randomization. The Annals of statistics 49, 13 2079-2100 (2021). 14 118. Consortium, C.-D.G.o.t.P.G. Identification of risk loci with shared effects on five 15 major psychiatric disorders: a genome-wide analysis. The Lancet 381, 1371-1379 16 (2013). 17 119. Nestor, S.M. et al. Ventricular enlargement as a possible measure of Alzheimer's 18 disease progression validated using the Alzheimer's disease neuroimaging 19 initiative database. Brain 131, 2443-2454 (2008). 20 120. Adamo, S. et al. Ventricular expansion, white matter hyperintensities, and global 21 cognition in Alzheimer’s disease and normal aging. medRxiv (2020). 22 121. Ferrarini, L. et al. Shape differences of the brain ventricles in Alzheimer's disease. 23 Neuroimage 32, 1060-1069 (2006). 24 122. Luo, Q. et al. Association of a schizophrenia-risk nonsynonymous variant with 25 putamen volume in adolescents: a voxelwise and genome-wide association 26 study. JAMA psychiatry 76, 435-445 (2019). 27 123. Gardener, H., Wright, C.B., Rundek, T. & Sacco, R.L. Brain health and shared risk 28 factors for dementia and stroke. Nature Reviews Neurology 11, 651-657 (2015). 29 124. Broce, I.J. et al. Dissecting the genetic relationship between cardiovascular risk 30 factors and Alzheimer’s disease. Acta neuropathologica 137, 209-226 (2019). 31 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint 31 125. Li, J. & Fei, G.-H. The unique alterations of hippocampus and cognitive 1 impairment in chronic obstructive pulmonary disease. Respiratory research 14, 2 1-9 (2013). 3 126. Wang, D. & Gu, J. VASC: dimension reduction and visualization of single-cell RNA-4 seq data by deep variational autoencoder. Genomics, proteomics & 5 bioinformatics 16, 320-331 (2018). 6 127. Kirchler, M. et al. transferGWAS: GWAS of images using deep transfer learning. 7 Bioinformatics 38, 3621-3628 (2022). 8 128. Vilor-Tejedor, N. et al. Genetic Influences on Hippocampal Subfields: An 9 Emerging Area of Neuroscience Research. Neurology Genetics 7(2021). 10 129. Sämann, P.G. et al. FreeSurfer-based segmentation of hippocampal subfields: A 11 review of methods and applications, with a novel quality control procedure for 12 ENIGMA studies and other collaborative efforts. Human brain mapping 43, 207-13 233 (2022). 14 130. Nakamura, Y. et al. The association between amygdala subfield-related 15 functional connectivity and stigma reduction 12 months after social contacts: a 16 functional neuroimaging study in a subgroup of a randomized controlled trial. 17 Frontiers in human neuroscience 14, 356 (2020). 18 131. Alfaro-Almagro, F. et al. Image processing and Quality Control for the first 10,000 19 brain imaging datasets from UK Biobank. NeuroImage 166, 400-424 (2018). 20 132. Casey, B.J. et al. The adolescent brain cognitive development (ABCD) study: 21 imaging acquisition across 21 sites. Developmental cognitive neuroscience 32, 22 43-54 (2018). 23 133. Bycroft, C. et al. The UK Biobank resource with deep phenotyping and genomic 24 data. Nature 562, 203-209 (2018). 25 134. Chen, C.-Y. et al. Improved ancestry inference using weights from external 26

panels. Bioinformatics 29, 1399-1406 (2013). 27 135. Avants, B.B. et al. A reproducible evaluation of ANTs similarity metric 28 performance in brain image registration. Neuroimage 54, 2033-2044 (2011). 29 136. Jiang, L. et al. A resource-efficient tool for mixed model association analysis of 30 large-scale data. Nature genetics 51, 1749 (2019). 31 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint 32 137. Dey, R. et al. An efficient and accurate frailty model approach for genome-wide 1 survival association analysis controlling for population structure and relatedness 2 in large-scale biobanks. bioRxiv (2020). 3 4

5 Shape features and imaging datasets. 6 The raw structural MRI data from the UKB and ABCD studies were used in this study. The 7 UKB study ’s ethics approval was obtained from the North West Multicentre Research 8 Ethics Committee (approval number: 11/NW/0382). The procedures of the ABCD study 9 were approved by the institutional review boards at ABCD collec tion sites (approval 10 numbers: 201708123 and 160091). The image collection and processing procedures can 11 be found in Alfaro-Almagro, et al. 131 for the UKB study and Casey, et al. 132 for the ABCD 12 study. 13 14 The shape feature generation pipeline was detailed in the Supplementary Note and an 15 overview of the procedures and examples were virtualized in Figures S1 and S42. Briefly, 16 we focused on 8 ventricular and subcortical structures, including the left/right lateral 17 ventricles, nucleus accumbens, amygdala, caudate, hippocampus, pallidum, putamen, 18 and thalamus. We constructed the vertex -wise map for 7 different shape statistics , 19 including radial distance, mTBM1, mTBM2, mTBM 3, determinant , eigenvalue1, and 20 eigenvalue2. For each shape statistics, we aggregated these vertex-wise data by 1) taking 21 the mean across all vertices for each structure; and 2) generating the top-ranked PCs for 22 each structure (left and right hemispheres separately). Intuitively, the purpose of these 23 PCs was to capture global and local variations within the vertex -wise representation of 24 brain structures. Typically, the first one or two PCs represented the global patterns, which 25 were similar to the mean values. Other PCs, however, captured local variations that mean 26 or top-ranked PCs missed. See Figure 1C for an illustration. Additionally, we segmented 27 the hippocampus and calculated the mean of shape statistics for each of the following 28 subfields: the left/right CA1, CA3, fimbria, HATA, hippocampal tail, presubiculum, and 29 subiculum. In total, we had 1,330 (210 mean values and 1,120 PCs) shape features. Based 30 on the UKB revisit data, we calculated the reproducibility (ICC) and selected those with 31 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint 33 ICC > 0.5, resulting in a final set of 457 shape features (363 at the structure -level and 94 1 at the subfield-level) for genetic analysis. See Table S1 for their names and descriptions. 2 3 We analyzed the above 457 reproducible traits in the following datasets: 1) the white 4 British discovery dataset, where the data were from white British subjects in UKB phases 5 1 to 3 imaging data (average n = 32,631, released up through 2020); 2) the UKB European 6 validation dataset, which included White individuals in the newly released UKB phase 4 7 data and the UKB non-British white individuals in phases 1 to 3 data (UKBE, removed the 8 relatives of the discovery sample, average n = 4,596); 3) the UKB non-European validation 9 dataset that consisting of non -White subjects in the UKB phases 1 to 4 data (UKBNE, 10 average n = 1,224); 4) the UKB first revisit data set (average n = 2,788); and 5) the ABCD 11 dataset (average n = 8,496). The average age (at imaging) of all UKB subjects was 64.2 12 (standard error = 7.73), 51.6% were females; the average age for all ABCD students was 13 9.93 (standard error = 0.62), 48.2% were females. Self-reported ethnicity (Data-Field 14 21000) was used to assign ancestry in UKB, whose accuracy was verified in Bycroft, et al. 15 133. We assigned ancestry to the ABCD participants based on self -reported ethni c 16

combined with SNPweights 134 inferences, see Zhao, et al. 45 for more 17 information. 18 19 Heritability and GWAS analysis. We downloaded UKB imputed genetics data133 (Data-20 Category 263) and locally imputed the ABCD genetic data using the Michigan Imputation 21 Server (https://imputationserver.sph.umich.edu/) with the 1000 Genomes Phase 3 22 (Version 5) reference panel45. In both UKB and ABCD, the following quality controls were 23 performed on imaging subjects with genetics data: 1) removing subjects with > 10% 24 missing genotypes; 2) removing genetic variants with minor allele frequency (MAF) 10%; 4) removing variants that 26 failed the Hardy-Weinberg test at 1 × 10-7 significance level; and 5) removing genetic 27 variants with imputation INFO score < 0.8. We used GCTA44 to estimate SNP-based 28 heritability with all autosomal SNPs in the white British discovery dataset (average n = 29 32,631). The adjusted covariates include age (at imaging), age-squared, sex, age-sex 30 interaction, age-squared-sex interaction, imaging site, the top 40 genetic PCs133, 31 volumetric scaling, head motion, head motion-squared, brain position, and brain position-32 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint 34 squared37,40. For the 363 structure -level shape features, t he genetic variance estimates 1 were also extracted from the GCTA and were compared before and after additionally 2 adjusting for the regional volum etric measurements. Specifically, for each of the 7 3 subcortical structures, we additionally controlled the corresponding FIRST subcortical 4 volumes (Category 1102). For the lateral ventricles, we adjusted for the regional volumes 5 estimated from ANTs 135. The genome -wide association analysis was conducte d with a 6 linear mixed effect model using fastGWA 136, adjusted for the same covariates as the 7 GCTA. We also conducted GWAS separately in validation datasets and adjusted for only 8 the top 10 genetic PCs rather than the top 40. In the ABCD dataset, we performed 9 validation GWAS separately for African American, European, and Hispanic subjects , 10 removing one subject randomly from each twin pair45. In all analyses, we removed values 11 greater than five times the median absolute deviation from the median for each 12 continuous phenotype or covariate variable. 13 14 We used FUMA47 (version v1.3.8) to characterize genomic loci with European LD files from 15 the 1000 Genomes . To define the LD boundaries , FUMA used independent significant 16 variants, which were genetic variants whose P-value smaller than the predefined 17 threshold (here was 1.09 × 10-10, 5 × 10-8/457) and were independent of other significant 18 variants (LD r2 < 0.6). FUMA then constructed LD blocks for these independent significant 19 variants by tagging all variants in LD ( r2 ≥ 0.6) with at least one independent significant 20 variant with a MAF ≥ 0.0005. There may have been variants from the 1000 Genomes 21

panel that were not included in the GWAS. Moreover, within these significant 22 variants, we defined independent lead va riants as those that were independent from 23 each other ( LD r2 < 0.1). In the case of close LD blocks (<250 kb based on the closest 24 boundary variants of LD blocks), they were merged into one genomic locus. Independent 25 significant variants and all the variants in LD with them (r2 ≥ 0.6) were looked up on the 26 NHGRI-EBI GWAS catalog (version e104_2021-09-15) to search for associations ( P < 9 × 27 10-6) reported for any traits. For selected colocalized index variants, we also performed 28 association analysis in vertex -wise data to illustrate local association patterns. The 29 significance threshold was set to be 0.05/number of vertices in each structure. We 30 adjusted for the same set of covariates as in the above genome-wise analysis. 31 32 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint 35 Gene-based testing was performed using UKB white British discovery GWAS summary 1 statistics for 18,796 protein-coding genes via MAGMA49 (version 1.08). We used default 2 MAGMA settings with zero window size around each gene. We also conducted functional 3 annotation and mapping analysis in FUMA, where genetic variants were annotated with 4 their genomic functionality and then were mapped to 35,808 candidate genes using 5 positional, eQTL, and 3D chromatin interaction information. Brain -related tissues/cells 6 were selected in all options and the default values were used for all other parameters in 7 FUMA. LDSC88 (version 1.0.1) was used to infer genetic correlations. LD scores were 8 computed using 1000 Genomes European data provided by LDSC. The major 9 histocompatibility complex (MHC) region was removed from the HapMap3 variants. 10 11 Polygenic risk scores on UKB non-imaging subjects. As a first step, we constructed a PRS 12 based on PRS-CS99 for each shape feature. We input GWAS summary stati stics from the 13 UKB white British discovery dataset (average n = 32,631), and randomly selected 1,500 14 subjects from the UKB European validation dataset as validation. We used all default 15 parameters in the PRS-CS software (https://github.com/getian107/PRScs) and generated 16 the PRS for all non-imaging individuals in the UKB study (removing relatives of the UKB 17 imaging individuals). The second step was to explore the associations with 276 18 phenotypes across various trait domains using these non-imaging UKB individuals, 19 including 24 mental health traits (Category 100060), 5 cognitive traits (Category 100026), 20 12 physical activity traits (Category 100054), 6 electronic device use traits (Category 21 100053), 8 sun exposure traits (Category 100055), 3 sexual factor traits (Category 22 100056), 3 social support traits (Category 100061), 12 family history of diseases (Category 23 100034), 21 diet traits (Category 100052), 9 alcohol drinking traits (Category 100051), 6 24 smoking traits (Category 100058), 34 blood biochemistry biomarkers (Category 17518), 3 25 blood pressure traits (Category 100011), 3 spirometry traits (Category 100020), 32 early 26 life factors (Categories 135, 100033, 100034, and 100072), 9 greenspace and coastal 27 proximity (Category 151), 2 hand grip strength (Category 100019), 13 residential air 28 pollution traits (Category 114), 5 residential noise pollution traits (Category 115), 2 body 29 composition traits by impedance (Category 100009), 4 health and medical history traits 30 (Category 100036), 3 female specific factors (Category 100069), 1 education trait 31 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint 36 (Category 100063), and 57 curated disease phenotypes based on Dey, et al. 137 (Table 1 S11). 2 3 We used a discovery -validation design and repeated our analysis in two independent 4 samples: 1) the discovery sample, which consisted of 70% randomly selected independent 5 UKB non -imaging subjects of white British ancestry (average n = 202,405) and 2) the 6 validation sample, including the left 30% independent UKB white British non -imaging 7 subjects (average n = 86,736), white non -British non -imaging subjects (average n = 8 20,746), and non -white non -imaging subjects ( average n = 21,587). The adjusted 9 covariates included age , age -squared, sex, age -sex interaction, age -squared-sex 10 interaction, as well as 40 genetic PCs. We reported P values from the two-sided t test and 11 prioritized on the results that were 1) significant after Bonferroni correction in the 12 discovery dataset, 2) significant at nominal significance level (0.05) in the validation 13 dataset; and 3) the regression coefficients had matched directions in the discovery and 14 validation datasets. 15 16 MR analysis with clinical endpoints. We examined the genetic causal relationships 17 between the 457 shape features and 288 clinical endpoints, where 275 of them were from 18 FinnGen (https://www.finngen.fi/en/access_results) and 13 were from the PGC 19 (https://pgc.unc.edu/). For FinnGen, we selected 275 clinical traits from the latest release 20 (R7) and with more than 5,000 cases. To reduce the potential influence of sample overlap, 21 we avoid PGC studies that have been using data solely from the UKB study. More detailed 22 information can be found in Table S13. 23 24 Before running MR methods, we performed standard preprocessing steps on GWAS data. 25 The genetic variants were first selected with significance threshold 5 × 10!" in the 26 exposure GWAS. To ensure the genetic variants used in the MR were independent, LD 27 clumping was implemented using 𝑟# = 0.01 , window size = 10,000, and the 1000 28 Genomes European ancestry data being the reference panel. The harmonization 29 procedure in the TwoSampleMR package (https://mrcieu.github.io/TwoSampleMR/) 30 helped us infer the correct allele alignment, therefore the selected variants on the 31 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint 37 exposure and the reported effect of the same variant on the outcome corresponded to 1 the same allele. 2 3 We tested 14 MR methods 114-117, including the IVW, IVW multiple random effect model, 4 IVW fixed effect model, MR-Egger, Simple Median, Weighted Median, Penalized 5 Weighted Median, Simple Mode, Simple Mode (NOME), Weighed Mode, Weighted Mode 6 (NOME), DIVW, GRAPPLE, and MR-RAPS. To ensure the reliability of our results, we 7 compared estimates from different methods. After running MR analysis on all pairs 8 between brain shape features and clinical endpoints, we used two steps to select the 9 significant causal results. We first removed all the estimated causal associations with less 10 than 6 variants used in the MR analysis. Then for the remaining estimates, we performed 11 Bonferroni adjustments for multiple testing. Besides comparing estimates across 12 different MR methods, we also tested potential violations in MR analysis to make sure 13 our results were reliable. For example, a significant intercept of MR Egger regression 14 indicated the presence of horizontal pleiotropy. All reported results have passed these 15 tests. 16 17 Code availability 18 We made use of publicly available software and tools. The pipelines used in shape feature 19 extractions can be found at https://www.nitrc.org/frs/?group_id=1461. The codes used 20 in other parts of the paper are available upon reasonable request. 21 22 Data availability 23 The individual-level data used in the present study can be applied from the UKB 24 (https://www.ukbiobank.ac.uk/) and ABCD (https://abcdstudy.org/) studies. Our GWAS 25 summary statistics will be shared on Zenodo and at the BIG-KP https://bigkp.org/. Our 26 GWAS results will also be available via the interactive web browser at 27 http://165.227.78.169:443/. 28 29 Figure Legends 30 Fig. 1 Illustrations of brain structures and their shape features. 31 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint 38 (A) In the left panel, we illustrate the 8 ventricular and subcortical structures, including 1 the lateral ventricles, nucleus accumbens, amygdala, caudate, hippocampus, pallium, 2 putamen, and thalamus. In the right panel, we present the 7 hippocampal subfields, 3 including the cornu ammonis 1 (CA1), CA3, fimbria, hippocampus -amygdala-transition-4 area (HATA), hippocampal tail, presubiculum, and subiculum. (B) We illustrate the spatial 5 pattern of radial distance in the vertex-wise map s of 8 ventricular and subcortical 6 structures. These maps were generated by averaging the data from 500 randomly 7 selected UKB subjects. See Figure S2 for additional maps of other 6 shape statistics 8 (mTBM1, mTBM2, mTBM3, determinant, eigenvalue1, and eigenvalue2). L, left; R, right. 9 (C) Comparison between the mean radial distance (RD) and structure-specific RD principal 10 components (PCs) on the lateral ventricles. (I) illustrates an example vertex-wise RD map 11 within the lateral ventricles after inter-subject centralization; (II) shows the residual RD 12 map after removing the within-subject mean RD; In (III) and (IV), instead of removing the 13 within-subject mean as in (II), we removed the top one and 10 RD PCs, respectively. (V) 14 illustrates the standard deviation across the vertices in residual RD map for each subject 15 in the UKB (n = 32746). Comparing (II) with (III), the top one PC can capture more spatial 16 variations than the mean RD and thus reduce the standard deviations of residuals in ( V). 17 (V) also shows that t he standard deviations are further reduced after removing the top 18 10 RD PCs (in (I V)), suggesting that additional PCs can account for more local spatial 19 variations that are ignored by the mean RD (in (III)) or top one RD PC (in (II)). 20 21 Fig. 2 Genetic variance and the associated genomic regions of shape features. 22 (A-B) The dots represent the genetic variance estimates of shape features. We compare 23 the original (marginal) genetic variance estimates before adjusting for corresponding 24 volumetric measurements (x axis) and the conditional genetic variance estimates after 25 adjusting for volumes (y axis). The results for the mean values and top one PCs (PC1s) are 26 displayed in the left panel (A), and results for the other PCs are displayed in the right panel 27 (B). We show the significant estimates after controlling the false discovery rate of multiple 28 testing at 5% level. Based on these results, we find that volumes can partially capture the 29 genetic influences on PC1s and mean features (in A) , while they cannot capture the 30 majority of genetic influences on other PCs (in B). (C) Ideogram of 60 genomic regions 31 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint 39 influencing shape features (P < 1.09 × 10 -10), 19 of which were not identified by the 1 corresponding volumetric measurements . The colors of dots represent the different 2 structures (and hippocampal subfields). Each signal dot indicates that at least one of the 3 shape features of this brain structure is associated with the genomic region. The name of 4 genomic regions replicated in more than one validation datasets or in one validation 5 dataset at the nominal significance level were highlighted in red and brown labels, 6 respectively. 7 8 Fig. 3 Genetic loci associated with both shape features and other complex traits. 9 (A) In the 17q24.1 region, we observed shared genetic influences (LD 𝑟# ≥ 0.6) between 10 shape features (e.g., Vent_Right_RD_PC9, index variant rs62072157) and brain white 11 matter microstructure (e.g., FX_MD, index variant rs35122942). Vent_Right_RD_PC9, the 12 ninth PC of the radial distance in right lateral ventricle; FX_MD, the mean diffusivity in the 13 fornix tract (column and body of fornix) . (B) In the 2q24.2 region, we observed shared 14 genetic influences (LD 𝑟# ≥ 0.6) between shape features (e.g., Hipp_Right_Eigen1_PC6, 15 index variant rs1014445) and cognitive traits (e.g., intelligence, index variant rs2268894). 16 Hipp_Right_Eigen1_PC6, the sixth PC of the eigenvalue1 in right hippocampus. (C) In the 17 10q26.13 region, we observed shared genetic influences (LD 𝑟# ≥ 0.6) between shape 18 features (e.g., Sub_CA3_Right_RD_Mean, index variant rs10901814) and kidney function 19 biomarker (eGRF, index variant rs4962691). Sub_CA3_Right_RD_Mean, the mean radial 20 distance in the CA3 subfield of right hippocampus; eGRF, estimated glomerular filtration 21 rate. In (D), we illustrate the signed -log10(P-value) of associations between the 22 rs4962691 variant and vertex-wise data of the hippocampus and lateral ventricles for 4 23 shape statistics, including radial distance, mTBM1, mTBM2, and mTBM3. It was observed 24 that the pattern of genetic effects varied by subregion within each structure. 25 26 Fig. 4 Selected genetic correlations with complex traits and diseases. 27 (A) We illustrate pairwise genetic correlations between shape features (x axis) and other 28 complex traits and diseases (y axis) estimated by LDSC. The asterisks highlight significant 29 pairs after controlling the FDR at 5% level . The colors represent the genetic correlation 30 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint 40 estimates. See Table S1 for descriptions of the shape features. (B) We illustrate brain 1 structures whose shape features are geneti cally related to brain disorders, such as 2 schizophrenia, attention-deficit/hyperactivity disorder (ADHD), and anorexia nervosa. (C) 3 We illustrate brain structures whose shape features are genetically related to cognitive 4 traits, education, and behavioral traits. (D) We illustrate brain structures whose shape 5 features are genetically related to cardiovascular diseases, such as coronary artery 6 disease and hypertension. 7 8 Fig. 5 Mendelian randomization analysis with clinical outcomes. 9 (A) Significant causal genetic links from brain shape features to clinical endpoints (P < 3.92 10 × 10-8). (B) Significant causal genetic links from clinical endpoints to brain shape features 11 (P < 1.12×10-8). In both A and B, the chord plot in the middle display each causal pair. The 12 first level (out) circle indicates each disease or brain structure. The second level circle 13 indicates the specific diseases with each disease category, or shape features within each 14 brain structure. The third level circle on the shape feature side indicates which disease s 15 the shape feature links to. In addition, the circle plots on the left - and right-hand sides 16 display the number of significant pairs for each exposure and outcome variable, 17 respectively. FG, FinnGen; PGC, Psychiatric Genomics Consortium ; COPD, chronic 18 obstructive pulmonary disease; Sub, subfield of hippocampus; HP, hippocampal tail; and 19 CA, cornu ammonis. See Table S1 for descriptions of the shape features. 20 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint Lateral ventricles Hippocampus AmygdalaCaudate PalliumPutamenThalamus Hippocampal subfields CA1HATASubiculum PresubiculumCA3 FimbriaHippocampal tail Ventricular and subcortical structures Nucleus accumbens A BNucleus accumbensL R AmygdalaCaudateHippocampus Pallium PutamenThalamus Lateral ventricles L R LR LR Higher LowerRadial distance C Remove top 1 RD PC Theradialdistance(RD)map(afterinter-subjectcentralization) Remove RD meanRemove top 10 RD PC -0.45 0.45Residual RD 0.0 0.5 1.0 1.5 2.0 2.5 Spatial standard deviation Remove mean RDRemove top 1 RD PCRemove top 10 RD PCs (I) (II) (III) (IV) (V) Figure 1 1 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint 0.0 0.1 0.2 0.3 0.4 0.5 0.0 0.1 0.2 0.3 0.4 0.5 Marginal genetic variance Conditional genetic variance Mean and PC1 Nucleus accumbens Amygdala Caudate Hippocampus Pallidum Putamen Thalamus Ventricles 0.0 0.1 0.2 0.3 0.4 0.5 0.0 0.1 0.2 0.3 0.4 0.5 Marginal genetic variance Conditional genetic variance Other PCsA B 1p32.2 10p12.3110q11.23 10q26.13 11q13.411q14.111q14.311q23.111q23.3 12p11.2212q14.312q23.312q24.2213p31.114q22.114q22.314q23.114q32.1114q32.1215q26.216q22.316q24.216q24.3 17q21.3117q24.118q21.2 1p331p36.12 1q23.1 22q13.1 2p13.23p11.1 2p212p24.2 2q24.22q33.3 2p16.1 3q24 3q28 4q12 4q32.3 4q245q14.25q14.35q31.15q34 6p25.16p22.16p22.2 6q22.32 7p22.37p22.27p22.17p21.18p11.21 8q24.129q22.329q31.39q33.19q34.3 C Figure 2 2 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint (task fMRI) chr17, Region: 17q24.1 62.8 mb 62.9 mb 63 mb 63.1 mb 0 2 4 6 8 10 Vent Right RD PC9 0 5 10 FX MD Gene Model AMZ2P1 ARHGAP27P1 ARHGAP27P1−BPTFP1−KPNA2P3 GNA13 LOC100507002 LRRC37A3 MIR6080 PLEKHM1P1 RGS9 FX MD GWAS Vent_Right_RD_PC9 GWAS rs62072157 rs35122942 629720006297400062976000629780006298000062982000 Subcortical volumeWhite matter microstructure Brain morphology Shape MRI SNP(s)GWAScatalog A chr10, Region: 10q26.13 126.3 mb 126.4 mb 126.5 mb 126.6 mb 0 5 10 15 20 25 Sub CA3 Right RD Mean 0 5 10 eGFR Gene Model ABRAXAS2 CTBP2 EEF1AKMT2 FAM53B FAM53B−AS1 LHPP ZRANB1 eGRFGWAS Sub_CA3_Right_RD_Mean GWAS rs10901814 rs4962691 126420000126440000126460000126480000126500000126520000126540000 Estimated glomerular filtration rateEstimated glomerular filtration rateBrain region volumesCortical thickness Cortical thicknessSubcortical volume Dentate gyrus molecular layer volume Estimated glomerular filtration rate Estimated glomerular filtration rate Hippocampal tail volume Dentate gyrus granule cell layer volumeHippocampal subfield CA4 volumeHippocampal volume Cortical surface area Cortical surface area Educational Attainment Shape MRI SNP(s)GWAScatalog C Hippocampus Lateral ventricles RRadial distancemTBM1 mTBM2mTBM3L 15 -15–log10(p-value) ×Sign(β) chr2, Region: 2q24.2 162.7 mb 162.8 mb 162.9 mb 163 mb 0 2 4 6 8 10 12 Hipp Right Eigen1 PC6 0 5 10 IntelligenceGene Model DPP4 DPP4−DT FAP GCG LOC101929532 Intelligence GWAS Hipp_Right_Eigen1_PC6 GWAS rs1014445 rs2268894 162800000162820000162840000162860000162880000 Cognitive abilitySmoking initiationCognitive abilityCognitive performanceSmoking status Smoking initiation Alcohol consumptionVerbal−numerical reasoningCognitive performanceEducational AttainmentGeneral cognitive abilityHighest math class taken Cortical surface area Cognitive ability IntelligenceSmoking status Subiculum volumeDentate gyrus molecular layer volumeHippocampal tail volumeHippocampal volumeSubcortical volume Hippocampal tail volume Cognitive abilityIntelligenceSelf−reported math abilityHippocampal volume Smoking initiationBrain morphologySubcortical volume Shape MRI SNP(s)GWAScatalog B DRs4962691 Figure 3 3 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint −0.3 −0.2 −0.1 0.0 0.1 0.2 0.3 Sub_HP_tail_Left_mTBM3_Mean Amyg_Left_Eigen2_PC7 Caud_Right_mTBM2_Mean Caud_Right_RD_PC4 Hipp_Right_mTBM3_PC3 Hipp_Left_mTBM3_Mean Hipp_Left_mTBM3_PC2 Hipp_Left_mTBM3_PC7 Hipp_Left_Dete_PC7 Hipp_Left_Eigen1_PC10 Hipp_Right_mTBM1_PC4 Puta_Left_RD_PC5 Puta_Left_mTBM3_PC9 Thal_Left_RD_Mean Thal_Right_RD_PC3 Vent_Right_mTBM2_PC4 Vent_Right_Dete_PC7 Vent_Right_Eigen2_PC2 Vent_Left_RD_PC4 Vent_Left_RD_PC8 Vent_Left_mTBM3_PC8 Vent_Left_Eigen1_PC5 Vent_Right_RD_PC3 Vent_Right_RD_PC7 ADHD ALS Anorexia Nervosa Automobile speeding Coronary artery disease Cognitive function Cross disorder Education Extreme chronotypeHigh blood pressure Insomnia IntelligenceNeuroticism Reaction timeRisk toleranceRisky behaviorsSchizophrenia Snoring *** * *** **** ************ *** ** ** ** ** * A Lateral ventricles HippocampusAmygdalaCaudatePalliumPutamenThalamus Nucleus accumbens BBrain disorders ADHD(hippocampus and lateral ventricle)Schizophrenia(thalamus and hippocampus) CCognitive traits, education, and behavioral traits Cognitive function, intelligence, and education (lateral ventricle)Reac>on >me (thalamus and lateral ventricle) Insomnia (caudate) Risk tolerance (hippocampus) Neuroticism(lateral ventricle) DCardiovascular diseases Coronary artery disease (hippocampus and lateral ventricle) Hypertension (hippocampus, putamen, and lateral ventricle) Anorexia nervosa (lateral ventricle) Figure 4 4 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint 1 1 1 111 22 3 3 11 111 1 1 2 2 21 111111 2 111 1 2 1 4 4 4 1 2 2 1 3 4 1 1 1 5 3 1 2 AccuAccu_Right_Eigen2_PC3AmygAmyg_Right_RD_PC3 HippHipp_Right_Eigen1_PC7Hipp_Left_mTBM3_PC5Hipp_Left_Dete_PC8Hipp_Left_mTBM2_PC5 PallPall_Left_Eigen1_MeanPall_Left_Eigen2_MeanPall_Left_RD_MeanPall_Left_RD_PC2Pall_Left_Dete_MeanPall_Left_mTBM1_MeanPutaPuta_Left_RD_PC4Puta_Left_Dete_MeanPuta_Left_mTBM1_MeanPuta_Left_mTBM3_MeanSub−CA3Sub_CA3_Right_mTBM2_MeanSub−fimbriaSub_fimbria_Right_Dete_MeanSub_fimbria_Right_Eigen1_MeanSub_fimbria_Right_mTBM3_MeanSub_fimbria_Right_mTBM1_Mean Sub−HPSub_HP_tail_Right_Dete_MeanSub_HP_tail_Right_mTBM3_Mean Sub−presubiculumSub_presubiculum_Right_Eigen1_MeanSub_presubiculum_Right_Eigen2_MeanSub_presubiculum_Left_Eigen1_MeanVentVent_Right_Eigen1_PC9Vent_Left_Dete_PC6Vent_Left_Eigen1_PC5Vent_Left_Eigen2_PC4Vent_Right_Eigen2_PC6Vent_Left_RD_PC10 COPD and related endpointsCOPD Diseases of the circulatory systemAortic aneurysmCalcific aortic valvular stenosisCalcific aortic valvular stenosis, including rheumatic feverCalcific aortic valvular stenosis, operatedHeart failure, not strictNon−rheumatic valve diseasesPeripheral artery diseaseValvular operationsHypertensive Heart DiseaseHypertensive heart and/or renal disease Diseases of the genitourinary systemAll dysplastic lesions of the cervix uteriDiseases of male genital organs Diseases of the eye and adnexaOther retinal disorders Mental and behavioural disordersDementia, including avohilmo Rheuma endpointsOther systemic involvement of connective tissue (FG)Primary gonarthrosis, bilateral Diseases of the ear and mastoid processDiseases of middle ear and mastoid 1 3 3 3 6 123 1 1 4 1 1 4 2 4 11 1 1 HippHipp_Right_mTBM3_Mean PutaPuta_Left_Eigen1_MeanPuta_Left_Eigen2_MeanPuta_Left_RD_Mean Sub−CA1Sub_CA1_Left_Dete_MeanSub_CA1_Left_Eigen1_MeanSub_CA1_Left_Eigen2_MeanSub_CA1_Left_mTBM1_MeanSub_CA1_Left_mTBM3_Mean COPD and related endpointsCOPDDiseases marked as autimmune originAutoimmune diseasesAutoimmune diseases excluding thyroid diseasesAutoimmune diseases excluding thyroid diseases, strict definition Diseases of the eye and adnexaDiabetic retinopathyGlaucoma suspect Diseases of the nervous systemAlzheimer's disease, including avohilmoAlzheimer's disease, wide definitionAlzheimer's disease, wide definition (more control exclusions)Migraine with auraCarpal tunnel syndrome Mental and behavioural disordersDementiaSpecific personality disorders VentVent_Left_Dete_PC6Vent_Left_Eigen1_PC5Vent_Left_Eigen2_PC1Vent_Left_RD_MeanVent_Left_RD_PC1Vent_Left_RD_PC3Vent_Left_mTBM2_PC2Vent_Left_mTBM3_PC6Vent_Right_RD_PC1Vent_Right_mTBM1_PC6Vent_Right_mTBM2_MeanVent_Right_mTBM2_PC1Vent_Right_mTBM3_Mean Rheuma endpointsOther systemic involvement of connective tissue (FG)Other/unspecified seropositiverheumatoid arthritisSeropositive rheumatoid arthritisSeropositive rheumatoid arthritis, wide Neuropsychiatric disorders (PGC)Schizophrenia (Corvin 2014)Schizophrenia (Pardinas 2018)Cross disorder 2 2 1 1 2 2 2 2 4 1 5 2 1 5 3 1 3 1 1 1 11 A B Figure 5 5 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted September 27, 2022. ; https://doi.org/10.1101/2022.09.26.22279691doi: medRxiv preprint