Trans-ancestry genome-wide association meta-analyses of hippocampal and subfield volumes

doi:10.21203/rs.3.rs-1430156/v1

Download PDF

Article

Trans-ancestry genome-wide association meta-analyses of hippocampal and subfield volumes

https://doi.org/10.21203/rs.3.rs-1430156/v1

This work is licensed under a CC BY 4.0 License

Journal Publication

published 19 Jun, 2023

Read the published version in Nature Genetics →

Version 1

posted

You are reading this latest preprint version

The hippocampus is critical for memory and cognition and neuropsychiatric disorders, and its subfields differ in cytoarchitecture, function and disease vulnerability. The genetic architectures of hippocampal and subfield volumes have been studied in individuals of European ancestry; however, the results cannot explain for inter-ancestry differences in their associations with cognitive abilities and brain disorders. Here, we conduct trans-ancestry genome-wide association meta-analyses in 65,821 individuals for hippocampal volume and 39,007 for subfield volumes, including 7,039 individuals of East Asian ancestry. We identify 539 locus-trait associations at P < 5×10^− 8 for 44 hippocampal volumetric traits, including 161 novel associations and 41 novel loci. Trans-ancestry analysis improves the precision of fine-mapping mainly caused by the inter-ancestry difference in linkage disequilibrium and can enhance the prediction performance of polygenic scores in under-represented population by including hundreds of individuals from this population. Genetic variants associated with hippocampal and subfield volumes are enriched for neurogenesis and Wnt signaling and also affect cognition, emotion and neuropsychiatric disorders, such as Alzheimer's disease, schizophrenia, and depression. Our results highlight the value of trans-ancestry analysis in the investigation of genetic architectures of human traits.

The hippocampus embedded in the medial temporal lobe of the human brain is critical for episodic memory, spatial navigation and other higher-level cognitive functions^1–3, and particularly sensitive to various stressful events⁴. The structural and functional impairments of the hippocampus have been observed in numerous neuropsychiatric disorders with different pathogenic mechanisms, such as Alzheimer’s disease (AD)^5,6, Parkinson’s disease⁷, schizophrenia⁸, major depressive disorder (MDD)⁹, anxiety¹⁰, addiction¹¹, and posttraumatic stress disorder¹². The hippocampus is a heterogeneous structure consisting of subfields with distinct cytoarchitecture and function and vulnerability to neuropsychiatric disorders^13,14. For instance, the vulnerable subfield is CA1 in schizophrenia¹⁵ and dentate gyrus in MDD⁹. The subiculum and CA1 are more prominently affected by AD than the dentate gyrus and CA3¹⁶.

As a recognized neuroimaging biomarker for AD, hippocampal volume calculated from structural magnetic resonance imaging (MRI) is highly heritable with an estimated twin-based heritability of > 75%¹⁷, which motivates the investigation of the genetic architecture of hippocampal volume. Prior genome-wide association studies (GWASs) have identified a number of genetic variants for hippocampal and subfield volumes in individuals of European ancestry (EUR)^17–24, but non-EUR populations, such as East Asian ancestry (EAS), are severely under-represented. The results found in EUR cannot explain for the inter-ancestry difference in the association of hippocampal volume with memory²⁵. The variant-level differences in allele frequency, linkage disequilibrium (LD), and effect size across populations also prevent us from extrapolating genetic associations identified in EUR to non-EUR populations. For example, the genetic risk variants in CREB1 shared by bipolar disorder and MDD are associated with hippocampal volume in EUR, but their risk alleles are completely absent in EAS²⁶. Moreover, the genetic associations of hippocampal volume identified in EUR cannot be replicated in EAS²⁷. These results highlight the need of conducting GWASs of hippocampal volume in ancestrally diverse populations to uncover the ancestry-shared and ancestry-specific genetic associations of hippocampal volume and to ensure the resulting GWAS findings to be broadly applicable²⁸. Furthermore, trans-ancestry GWASs have found to benefit for discovering new genetic associations, improving fine-mapping of causal variants, estimating polygenic score (PGS), and addressing inter-ancestry health disparities^29–33. However, no trans-ancestry GWASs have been conducted for hippocampal and subfield volumes so far mainly due to the lack of neuroimaging genetics data of non-EUR populations.

The Chinese Imaging Genetics (CHIMGEN)³⁴ project has collected genomic and neuroimaging data from 7,306 healthy Chinese Han participants, which provides a unique opportunity to investigate the genetic architectures of hippocampal and subfield volumes using trans-ancestry GWAS meta-analyses by incorporating genetic data from EUR. In this study, we aim to clarify the following questions: (a) whether and why trans-ancestry analysis can identify novel associations; (b) whether and why trans-ancestry analysis can improve the precision of fine-mapping; and (c) whether and how trans-ancestry analysis can promote the prediction ability of PGS in under-represented populations. We are also interested in functional and clinical associations of the identified genetic variants associated with hippocampal volumetric traits.

Trans-ancestry genetic associations of hippocampal and subfield volumes

In this study, we aimed to identify genetic loci associated with hippocampal and subfield volumes, including 44 traits symmetrically distributed in the left and right hippocampus (Fig. 1a). To obtain GWAS summary statistics of these traits in EAS, we carried out GWASs for the 44 hippocampal volumetric traits at 6,830,146 single nucleotide polymorphisms (SNPs) in 7,039 EAS participants (Supplementary Table 1) from CHIMGEN. We identified 21 significant locus-trait associations at P < 5 × 10^− 8 (Extended Data Fig. 1), including 12 novel associations not reported in EUR-specific studies (Supplementary Table 2). The obtained GWAS summary statistics of hippocampal and subfield volumes in EAS were used for the following trans-ancestry meta-analyses.

We first performed trans-ancestry GWAS meta-analyses for the left and right hippocampal volumes in 65,821 participants (EAS: 7,039 CHIMGEN participants; EUR: n = 58,782, 31,968 from UKBB and 26,814 from ENIGMA), and identified 77 locus-trait associations at P < 5 × 10^− 8, including 46 novel associations (Fig. 1b) that have not been reported in previous EUR-specific GWASs. Then we conducted trans-ancestry GWAS meta-analyses for the 42 hippocampal subfield volumes in 39,007 participants (EAS: n = 7,039 from CHIMGEN; EUR: n = 31,968 from UKBB), and identified 462 locus-trait associations at P < 5 × 10^− 8, including 115 novel associations (Fig. 1b).

In the trans-ancestry GWAS meta-analyses for all 44 phenotypes of hippocampal and subfield volumes, we identified 539 locus-trait associations (Extended Data Fig. 2, Supplementary Table 3) at P < 5 × 10^− 8, including 161 novel associations (Fig. 1b) not reported in EUR-specific GWASs. The 539 locus-trait associations belonged to 118 loci, and 59 loci were correlated with more than one trait. The 161 novel associations belonged to 83 loci, including 41 novel loci that have not been reported in EUR-specific GWASs. For example, the locus with rs62430477 as the lead SNP was one of the novel loci identified by trans-ancestry meta-analyses for the right CA4 body volume (Fig. 1d). Since 80 of the 539 locus-trait associations showed nominal allelic effect heterogeneity (Cochran’s Q-test: P < 0.05) between EAS and EUR (Supplementary Table 3), we also validated these findings using Han and Eskin’s random effects model³⁵, which models the inter-study heterogeneity better. We found that 79 lead SNPs of the 80 associations reached genome-wide significance (P < 5 × 10^− 8) and the remaining one was also closed to the significant level (P = 5.40 × 10^− 8) (Supplementary Table 3), indicating that the results of trans-ancestry meta-analyses are reliable.

It is an open question whether trans-ancestry meta-analysis itself can provide novel biological insight since the identified new associations by trans-ancestry meta-analyses may be attributed to the increase of the sample size. To answer the question, we compared the results derived from trans-ancestry meta-analysis and those from EUR-specific GWAS with comparable sample size. For hippocampal volumes, we compared trans-ancestry GWAS meta-analyses in 55,135 participants (7,039 from CHIMGEN, 21,282 from UKBB discovery and 26,814 from ENIGMA) and EUR-specific GWAS meta-analyses in 58,782 participants (31,968 from UKBB and 26,814 from ENIGMA). For hippocampal subfield volumes, we compared trans-ancestry GWAS meta-analyses in 28,321 participants (7,039 from CHIMGEN and 21,282 from UKBB discovery) and EUR-specific GWASs in 31,968 participants from UKBB. We found 60 locus-trait associations (P < 5 × 10^− 8) (Fig. 1c) that were discovered only through trans-ancestry meta-analyses (Extended Data Fig. 3 and Supplementary Table 4), indicating that ancestrally diverse populations with even smaller sample size than a single population can yield unique discoveries possibly due to the difference in genetic architectures between populations. As an example, the locus with rs2274692 as the lead SNP for the left CA1 body volume was identified in trans-ancestry meta-analysis rather than in EUR-specific GWAS with larger sample size (Fig. 1e).

To determine potential factors accounting for the improved power of trans-ancestry meta-analysis, we compared the LD, allele frequencies and effect sizes of the lead SNPs of the 60 locus-trait associations between EAS (CHIMGEN, n = 7,039) and EUR (UKBB, n = 31,968). Overall, the 60 associations had larger effects (Wilcoxon rank-sum test: P = 5.42 × 10^− 4 for positive effects and P = 7.39 × 10^− 8 for negative effects, Fig. 1f and g) on hippocampal and subfield volumes in EAS than in EUR. The 60 associations contained 50 SNPs, whose minor allele frequencies (MAFs) had no significant difference (Wilcoxon singed rank test, P = 0.367) between EAS and EUR (Fig. 1h). EAS-specific and EUR-specific LD values were measured by r² using their respective 1000 Genomes Phase 3 reference datasets³⁶. For each lead SNP of a given locus, we separately calculated the r² of the SNP to all other SNPs in the locus in EAS and EUR, and compared the r² values greater than 0.01 in both ancestral populations. Overall, the r² of the 50 lead SNPs had no significant difference between EAS and EUR (Wilcoxon rank-sum test: P = 0.23, Extended Data Fig. 4), and the relationships between LD and genomic distance of EAS and EUR are shown in Fig. 1i. In this study, we found 21 associations whose lead SNPs had a similar effect allele frequency (EAF, difference within 5%) between EUR and EAS, 19/21 had larger effects on the hippocampal and subfield volumes in EAS than those in EUR (grey spheres in Fig. 1f). For instance, rs823385 had almost the same EAF values (EAF_EAS = 0.581, EAF_EUR = 0.558) in EAS and EUR, but its effect size on the right hippocampal volume was twice larger in EAS (beta = 0.0597) than that in EUR (beta = 0.0290). In another example, rs12495978 had similar EAF values (EAF_EAS = 0.193, EAF_EUR = 0.151) in the two populations, it was much more strongly associated with the left GC-ML-DG head volume (beta = 0.102) in EAS than that (beta = 0.0351) in EUR. Of the 39 associations whose lead SNPs had different EAFs between EAS and EUR, 35 associations had larger effect sizes on hippocampal and subfield volumes in EAS than those in EUR (Fig. 1f). For example, the effect allele of rs2235933 was more common in EAS than in EUR (EAF_EUR = 0.271, EAF_EAS = 0.578), and its effect on the right presubiculum body volume was larger in EAS (beta = -0.0647) than that in EUR (beta = -0.0312). Although the effect allele of rs1518085 was more common in EUR than in EAS (EAF_EUR = 0.851, EAF_EAS = 0.417), its effect on the left subiculum head volume in EAS (beta = -0.0866) was still larger than that in EUR (beta = -0.0401). Taken together, these findings indicate that the difference in effect size between ancestral populations, especially the relatively greater effect size in the under-represented population, is an important cause for novel discoveries in trans-ancestry meta-analysis.

Ancestry-shared and ancestry-specific genetic associations of hippocampal and subfield volumes

Since no large-scale GWAS has been conducted for hippocampal and subfield volumes in non-EUR populations, the ancestry-shared and ancestry-specific genetic associations of hippocampal and subfield volumes between EUR and EAS remain unknown. To compare the SNP’s effect sizes on hippocampal and subfield volumes between EUR and EAS, we chose the SNPs with genome-wide significance (P < 5 × 10^− 8) in any of the four GWASs including trans-ancestry meta-analyses (n = 65,821), CHIMGEN (n = 7,039), UKBB (n = 31,968), or ENIGMA (n = 26,814) for hippocampal volume, and those in any of the three GWASs including trans-ancestry meta-analyses (n = 39,007), CHIMGEN (n = 7,039) or UKBB (n = 31,968) for hippocampal subfield volumes. After excluding duplicate associations and those with loci genotyped only in one population, we obtained 581 locus-trait associations (Supplementary Table 5) and compared the direction of effect of the lead SNPs between EAS (7,039 CHIMGEN participants) and EUR (31,968 UKBB participants). We observed 86.92% of these associations (505/581) with consistent direction of effect between EAS and EUR (Fig. 2a). In the tests for allelic effect heterogeneity, we found 462/581 (79.52%) ancestry-shared genetic associations with no evidence for allelic effect heterogeneity (Cochran’s Q-test: P ≥ 0.05) between EAS and EUR (Fig. 2b). For instance, two loci with rs2363340 (chr10) and rs6496265 (chr15) as the lead SNPs showed significant effects on several phenotypes of hippocampal and subfield volumes in both EAS and EUR, such as the right hippocampal tail (Fig. 2c left) and left hippocampal body volumes (Fig. 2c right). We also found 119/581 (20.48%) potential ancestry-specific genetic associations with nominally significant (Cochran’s Q-test: P < 0.05) in the allelic effect heterogeneity tests (Fig. 2b). Using a threshold of P < 8.61 × 10^− 5 to additionally correct for 581 associations, we found 23/581 (3.96%) reliable ancestry-specific genetic associations (Fig. 2b). The distribution of different types of genetic associations across the 44 traits of hippocampal and subfield volumes is shown in Fig. 2e. The left CA4 head volume had the largest number (n = 3) and ratio (23.08%) of reliable ancestry-specific genetic associations (Fig. 2e). For example, the locus with rs7315280 as the lead SNP, a previously identified locus for hippocampal volumetric traits in chromosome 12 in EUR-specific GWAS, showed strong allelic effect heterogeneity on the right hippocampal tail volume between EAS and EUR (Cochran’s Q-test: P = 2.14 × 10^− 8). The locus affected the trait only in EUR (Fig. 2d left). In addition, there were loci that had a greater effect on hippocampal traits in EAS than in EUR, such as the locus with rs6463429 as the lead SNP, showed an ancestry-specific genetic association for left GC-ML-DG head volume (Cochran’s Q-test: P = 1.06 × 10^− 6; Fig. 2d right), left CA4 head volume (Cochran’s Q-test: P = 1.48 × 10^− 6) and left hippocampal head (Cochran’s Q-test: P = 4.35 × 10^− 6). This was a novel locus identified in EAS for the three hippocampal traits, however, the locus had no impact for these traits in EUR.

Trans-ancestry analysis improves the resolution of fine-mapping

To test whether the resolution of fine-mapping based on trans-ancestry GWAS summary statistics is higher than that based on ancestry-specific GWAS summary statistics, we compared the fine-mapping results derived from trans-ancestry meta-analyses in 28,321 participants (7,039 from CHIMGEN and 21,282 from UKBB discovery) and those derived from EUR-specific GWASs in 21,282 individuals from UKBB discovery. After excluding 36 loci not present in CHIMGEN genetic data, the remaining 231 genome-wide significant locus-trait associations identified in UKBB discovery were used for comparing the fine-mapping performance based on GWAS summary statistics from trans-ancestry and single-ancestry analyses. For each locus, the range (± 500kb) and SNPs were kept consistent between trans-ancestry and ancestry-specific datasets, and fine-mapping was performed using PAINTOR³⁷. The fine-mapping results are demonstrated in Fig. 3a and Supplementary Table 6. The fine-mapping resolution was improved in 121/231 (52.38%) locus-trait associations, and the number of the most precise fine-mapping (one variant in 95% credible set) was increased from 2 in single-ancestry analysis to 10 in trans-ancestry analysis. Since the fine-mapping resolution is increased with the sample size in GWASs³¹, we also compared the fine-mapping results derived from trans-ancestry meta-analyses in 28,321 participants (7,039 from CHIMGEN and 21,282 from UKBB discovery) and those derived from single-ancestry GWASs in 31,968 EUR individuals from UKBB discovery (n = 21,282) and replication (n = 10,686). In the 121 locus-trait associations with improved fine-mapping resolution in trans-ancestry analysis, the fine-mapping resolution of 82 loci (67.8%) was still higher in trans-ancestry analysis than in single-ancestry analysis with 12.9% larger sample size (Fig. 3a). Moreover, the number of the most precise fine-mapping was still 2 in single-ancestry analysis with a greater sample size, which was much smaller than 10 in trans-ancestry analysis. These findings indicate that the improved resolution of fine-mapping in trans-ancestry analysis cannot be simply explained by the increase in sample size. One example is the locus with rs1370938 as the lead SNP (Fig. 3b), the number of causal SNPs in 95% credible set shifted from three in EUR-specific analysis to one in trans-ancestry fine-mapping. The improved resolution of fine-mapping seems to be related to the difference in LD pattern of the locus between EAS and EUR (Fig. 3c), rs1370938 was highly linked to the other two SNPs included in 95% credible set in EUR, but it was not correlated with these two SNPs in EAS. In addition, the fine-mapping results from PAINTOR were further confirmed by approximate Bayesian approach³⁸. The sizes (the number of SNPs) of 95% credible sets were significantly correlated between the two fine-mapping methods (Pearson’s r = 0.995), and the 95% credible sets with one causal SNP were identical (Supplementary Table 6).

To further explore which genetic features are associated with the improved fine-mapping in trans-ancestry analysis, we compared the inter-ancestry differences in LD, allele frequencies, and effect sizes of the peak SNPs of 82 locus-trait associations with improved fine-mapping resolution in trans-ancestry analysis relative to EUR-specific analyses. The 82 locus-trait associations contained 47 distinct lead SNPs, and their LD (r²) values were calculated according to the above-mentioned approach. Overall, the r² values of 47 lead SNPs were much higher (Wilcoxon rank-sum test: P = 3.78 × 10^− 133) in EUR than in EAS (Fig. 3d). We further compared the r² grouped by distance (50kb as interval) and found consistently higher LD in EUR than in EAS within the range of 500kb (Fig. 3e). Despite the MAFs of these lead SNPs in EAS were less frequent (Wilcoxon singed rank test: P = 2.51 × 10^− 4) than that in EUR (Fig. 3f), their effect sizes did not show significant difference between trans-ancestry and EUR-specific analyses (Wilcoxon rank-sum test, P = 0.88) (Fig. 3g). These findings suggest that the LD difference between ancestries is a major contributor for the improved fine-mapping resolution in trans-ancestry analysis, and the contribution of the MAF difference needs to be further confirmed due to the moderate significance and large within-ancestry variance.

Transferability for PGS of trans-ancestry meta-analysis

So far, all large-scale GWASs for hippocampal and subfield volumes have been conducted in EUR individuals; however, the PGSs constructed only based on EUR individuals tend to underperform when predicting for individuals from non-EUR populations^39,40. In this study, we systematically assessed the applicability of PGSs constructed based on different combinations of individuals with distinct ancestries to predict traits of individuals from the under-represented population (EAS). In these analyses, CHIMGEN participants were divided into two groups according to the types of MRI scanners: 5,023 participants (CHIMGEN discovery) whose MRI data were acquired by the same type of scanner (GE MR750) and parameters were regarded as the base dataset in the PGS analysis; and the remaining 2,016 participants whose MRI data were acquired by other nine types of MR scanners were regarded as the target dataset in this analysis (Extended Data Fig. 5). We compared the prediction performance of PGS constructed based on three types of GWAS summary statistics: (a) trans-ancestry meta-analysis in 26,305 participants (5,023 from CHIMGEN discovery and 21,282 from UKBB discovery); (b) EUR-specific GWAS in 31,968 participants from UKBB discovery (n = 21,282) and replication (n = 10,686); and (c) EAS-specific GWAS in 5,023 participants from CHIMGEN discovery. Because we were interested in the prediction performance of PGSs for traits of individuals from the under-represented population, we only included 2,016 EAS participants from CHIMGEN as the target dataset in all PGS analyses (Supplementary Table 7). We found that the prediction performance (R²_average = 0.011) of PGSs derived from the trans-ancestry meta-analysis outperformed than those derived from EAS-specific GWAS (R²_average = 0.0056; Wilcoxon rank-sum test: P = 4.08 \(\times\) 10^− 11) and EUR-specific GWAS (R²_average = 0.0069; Wilcoxon rank-sum test: P = 6.09\(\times\) 10^− 8). The trans-ancestry meta-analysis could improve the predictive performance in 42/44 traits relative to EAS-specific GWAS and in 40/44 traits than EUR-specific GWAS (Fig. 4a and b). The improved prediction performance of PGSs derived from trans-ancestry meta-analysis cannot simply attribute to the sample size because the sample size of the EUR-specific GWAS (n = 31,968) was 21.5% greater than that of the trans-ancestry meta-analysis (n = 26,305).

Consistent with previous studies⁴¹, we found that PGSs derived from the same ancestral population (EAS, 5,023 from CHIMGEN discovery) could achieve similar prediction performance (Wilcoxon rank-sum test of R², P = 0.46) relative to that from different ancestral population with much larger sample size (EUR, 21,282 from UKBB discovery) (Fig. 4b). We also found that PGS derived from EUR (n = 31,968) performed even better (Wilcoxon rank-sum test of R², P = 0.012) than that derived from the same under-represented population (EAS, n = 5,023) (Fig. 4b), indicating that PGS derived from large-sample EUR population can be used for predicting traits of individuals from the under-represented population when there was no GWAS summary statistics of the population with enough large sample size.

We further assessed the prediction performance of the PGSs derived from the trans-ancestry meta-analyses that included different numbers of EAS participants in the base datasets, which were extracted according to random sampling within the 5,023 participants from CHIMGEN discovery, and found that the inclusion of limited sample size of EAS participants can improve the prediction performance of PGS (Fig. 4c). When we constructed PGS based on trans-ancestry GWAS results of 500 EAS participants from CHIMGEN discovery and 21,282 EUR participants from UKBB discovery, its prediction performance was improved compared with that of PGS constructed using UKBB discovery participants (Wilcoxon rank-sum test of R², P = 0.02) and was comparable with the prediction performance of PGS constructed using all UKBB participants (n = 31,968) (Wilcoxon rank-sum test of R², P = 0.42), which was better than that constructed using UKBB discovery due to the increased sample size (Wilcoxon rank-sum test of R², P = 0.0031). Compared with PGSs constructed using all UKBB participants (n = 31,968), the prediction performance was improved by constructing PGS using 3,000 EAS from CHIMGEN discovery and 21,282 EUR participants from UKBB discovery (Wilcoxon rank-sum test of R², P = 3.55 \(\times\) 10^− 4) (Fig. 4d). Taken together, these results suggest that the construction of PGSs based on trans-ancestry GWAS results could improve the prediction performance for traits of individuals from the under-represented population by adding a moderate sample size of individuals from the population.

Functional annotation of genetic loci associated with hippocampal and subfield volumes

To understand functional consequences of genetic loci associated with hippocampal and subfield volumes, we performed functional annotation of significant SNPs (P < 5 × 10^− 8) derived from trans-ancestry analysis using an integrative web-based platform (FUMA)⁴². Gene functional consequences of these SNPs were investigated using ANNOVAR⁴³ and combined annotation-dependent depletion (CADD) scores⁴⁴, and effects of SNPs on gene expression were explored by expression quantitative trait loci (eQTL) mapping of hippocampal tissue from GTEx⁴⁵. Of the 4,844 SNPs, 4,833 were annotated by ANNOVAR, and they were mainly located in intronic (56.76%) and intergenic (27.21%) areas (Fig. 5a). We also identified 20 SNPs of non-synonymous mutations in exons (Supplementary Table 8) including four novel mutations in genes (NR1H3, rs2279238; MADD, rs1051006 and rs2290148; APOE, rs429358). NR1H3 encodes liver X receptor alpha and is linked to neurogenesis in the hippocampus⁴⁶ and AD⁴⁷. The MADD expression reduces in the hippocampus with AD pathology, which contributes to neuronal cell death⁴⁸. The rs429358 (CADD score = 12.64) is a missense mutation in APOE, which has been associated with hippocampal volume⁴⁹ and AD⁵⁰. The rs4076700 (a missense mutation in FBXW8) had the highest CADD score (26.8), and FBXW8 knockdown can lead to abnormal morphology of Golgi apparatus and reduced dendrite length in hippocampal neurons⁵¹.

In the hippocampal tissue, 835 distinct SNPs were eQTLs of 21 protein-coding genes (1,412 SNP-gene pairs) at a 5% false discovery rate (FDR) threshold (Fig. 5b and Supplementary Table 9). The rs11158037 located in a novel locus for the left CA1 body volume (P = 1.40\(\times\) 10^− 8, beta = 0.043) was an eQTL of ATG14 (P = 1.11\(\times\) 10^− 7, normalized effect size = 0.29), an autophagy-specific regulator⁵² associated with neural differentiation, axon growth, and neuronal signaling⁵³. The rs2311528 located in another novel locus for the right presubiculum body volume (P = 3.83\(\times\) 10^− 8, beta = 0.043) was correlated with the expression of CROCC (P = 4.07 \(\times\) 10^− 7, normalized effect size = 0.29), an encoding gene of the major structural component of the ciliary rootlet and being linked to brain development⁵⁴. Ancestry-shared significant SNPs in chr10:126427037–126562724 for hippocampal and subfield volumes were eQTLs of LHPP (P = 1.21 \(\times\) 10^− 6, normalized effect size = 0.19) and METTL10 (P = 2.41 \(\times\) 10^− 11, normalized effect size = -0.42). The former is a risk gene of MDD⁵⁵ and the latter is related to the process of translation⁵⁶.

All 118 loci (4,844 SNPs) were mapped to 166 protein-coding genes based on location, among which the 41 novel loci were aligned to 49 protein-coding genes (Fig. 5b). We performed the statistical over-representation analysis with WebGestalt⁵⁷ online tool to identify enriched gene ontology (GO) terms of biological process for these genes. The 166 genes were enriched for 71 GO biological process terms (q_c < 0.05, Benjamini-Hochberg FDR corrected) (Fig. 5c and Supplementary Table 10), mainly for the neuron-related biological processes, such as neurogenesis (P = 1.13\(\times\) 10^− 4), neuron development (P = 3.20 \(\times\) 10^− 5) and differentiation (P = 3.59 \(\times\) 10^− 5), axonogenesis (P = 4.39 \(\times\) 10^− 5), and axon extension (P = 7.50 \(\times\) 10^− 5) and guidance (P = 1.42 \(\times\) 10^− 4). Of these over-represented genes, nine (RAP1A, PTCH1, STRN, SPTBN1, RIPOR2, VEGFA, SFRP1, PARD3, and APOE) were mapped by the novel loci. We also found significant enrichment for Wnt signaling (e.g. negative regulation of the Wnt signaling, P = 1.37 \(\times\) 10^− 6), and the enriched genes included those (STRN, IGFBP2, SFRP1, WWOX, and APOE) mapped by the novel loci.

Phenome-wide association studies of genetic loci associated with hippocampal and subfield volumes

We queried the 4,844 genome-wide significant (P < 5 × 10^− 8) SNPs associated with hippocampal and subfield volumes in GWAS Catalog⁵⁸ by FUMA⁴². We only focused on brain-related traits and neuropsychiatric disorders. We found that 66 SNPs from 19 genomic regions associated with hippocampal and subfield volumes also affected the focused phenotypes at P < 5 × 10^− 8, including neuropsychiatric disorders (AD and schizophrenia), cognitive-related traits (educational attainment, math ability, cognitive performance, verbal declarative memory, intelligence, and reaction time), mental health (neuroticism, worry, depressive symptoms, life satisfaction, and general risk tolerance), and smoking/drinking (smoking and alcohol consumption) (Fig. 5d and Supplementary Table 11). For example, 13 SNPs in 2q24.2 (SLC4A10 and DPP4) associated with hippocampal volume and 30 subfield volumes were also correlated with cognitive-related traits and smoking/drinking traits. SLC4A10 encodes a Na⁺-coupled HCO₃⁻ exchanger, which is abundant in CA3 pyramidal cells and regulates neuronal excitability and synaptic short-term plasticity^59,60. DPP4 is highly expressed in the hippocampus and is correlated with cognitive impairment⁶¹. Another example, two SNPs located in 10q26.13 associated with hippocampal volume and 22 subfield volumes that regulate METTL10 expression in the hippocampus were also correlated with cognitive-related traits. The rs73405293 (chr12) marked the most significant locus for hippocampal volume was also associated with educational attainment. This SNP is located in the intron of TESC encoding tescalcin with a neuroprotective effect⁶². In the novel loci identified by trans-ancestry analysis, three SNPs (PTCH1, a gene correlated with neurogenesis in the hippocampus⁶³) associated with the left CA1 body volume were also correlated with cognitive-related traits. Notably, we found that 5 SNPs located in 17q21.31 associated with four hippocampal subfield volumes but not with the whole hippocampal volume were also correlated with cognitive-related traits and mental health, indicating that the subfield-specific genetic architecture may contribute to the inter-subfield differences of the hippocampus in cytoarchitecture and function and vulnerability to brain disorders^13,64.

In this study, we conducted the first trans-ancestry GWAS meta-analyses for 44 traits of hippocampal and subfield volumes in up to 65,821 individuals. The results improve our understanding of the genetic architecture of the whole and regional hippocampal volumes by providing 161 novel associations and 41 novel loci. Furthermore, we proved that trans-ancestry analysis with even smaller samples relative to ancestry-specific analysis could still yield new genetic discoveries, which mainly benefited from the relatively greater effect size in under-represented population rather than the inter-ancestry difference in allele frequency or LD. Taking together, these results highlight the value of a more global representation of populations in genetic studies. Despite most of the identified genetic associations with hippocampal volumetric traits were shared by ancestries, we also found ancestry-specific genetic associations, which may account for the inter-ancestry discrepancy in hippocampus-related findings.

In addition to confirming the previously identified advantages of trans-ancestry analysis in fine-mapping and genetic risk prediction^29,30, we provided additional insight into these points. In terms of fine-mapping, we found that the LD difference between ancestries contributed more to the improved resolution of fine-mapping by trans-ancestry analysis than the inter-ancestry difference in allele frequency and effect size. In genetic risk prediction, we proved that the prediction performance could be significantly improved by including only hundreds of individuals from the under-represented population in trans-ancestry analysis. This finding is exciting in the field of genetic risk prediction because small sample-size genetic data of the under-represented population are already available for a large number of human traits and disorders.

Beyond the discovery of novel genetic associations for hippocampal and subfield volumes, we provided new biological insight into the relation of hippocampus with neuropsychiatric disorders. On the one hand, we provided the first evidence for the effects of several AD-related genes (NR1H3, MADD and APOE) on the hippocampus based on the discovered associations of non-synonymous mutations in exons of these genes with hippocampal and subfield volumes. On the other hand, we identified two inter-related biological convergence points (neurogenesis and Wnt signaling) between hippocampus and neuropsychiatric disorders since they were main enrichment terms for genetic loci associated with hippocampal and subfield volumes and critical mechanisms in neuropsychiatric disorders including AD, MDD, schizophrenia, and autism spectrum disorder⁶⁵. The shared biological substrates support hippocampal and subfield volumes to be used as biomarkers for these disorders, and novel agents targeting the aberrant Wnt signaling in the hippocampus may benefit for patients with neuropsychiatric disorders with similar abnormality in the signal pathway.

Funding

This work was supported by the National Key Research and Development Program of China (2018YFC1314300) to C.Y., the National Natural Science Foundation of China to CY (82030053, 81425013). The authors report no biomedical financial interests or potential conflicts of interest. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Competing interests

The authors have declared that no competing interests exist.

Acknowledgements

We are grateful to the UKBB and ENIGMA for providing GWAS summary statistics of hippocampal and subfield volumes.

Author contributions

Chunshui Yu and Nana Liu designed the study and wrote the manuscript. All authors critically reviewed the manuscript. Nana Liu, Longjiang Zhang, Tian Tian, Jingliang Cheng, Bing Zhang, Shijun Qiu, Mulin Jun Li, Kai Xu, Xi-Nian Zuo, Meiyun Wang, Zhaoxiang Ye, Wen Qin, Feng Liu, Meng Liang, Qiang Xu, Jilian Fu, Jiayuan Xu and Chunshui Yu were the principal investigators. Longjiang Zhang, Tian Tian, Jingliang Cheng, Bing Zhang, Shijun Qiu, Zuojun Geng, Guangbin Cui, Quan Zhang, Weihua Liao, Yongqiang Yu, Hui Zhang, Bo Gao, Xiaojun Xu, Tong Han, Zhenwei Yao, Wenzhen Zhu, Peng Zhang, Wei Li, Dapeng Shi, Caihong Wang, Su Lui, Zhihan Yan, Feng Chen, Jiance Li, Jing Zhang, Dawei Wang, Wen Shen, Yanwei Miao, Junfang Xian, Jia-Hong Gao, Xiaochu Zhang, Kai Xu, Xi-Nian Zuo, Meiyun Wang, Zhaoxiang Ye and Chunshui Yuacquired the data. Chunshui Yu, Zhaoxiang Ye and Meiyun Wang supervised this work. Nana Liu analyzed the data.

Additional information

Supplementary Information is available for this paper.

Correspondence and requests for materials should be addressed to Chunshui Yu, Zhaoxiang Ye and Meiyun Wang.

Data availability

The raw brain imaging and genotype data are available from the authors upon reasonable request and with permissions of the CHIMGEN consortia. The GWAS summary statistics used in this work from two publicly available datasets: the UKBB study (https://open.win.ox.ac.uk/ukbiobank/big40/) and the ENIGMA study (http://enigma.ini.usc.edu/). All trans-ancestry meta-analysis GWAS summary statistics of 44 hippocampal and subfield volumes can be found at http://www.mulinlab.org/Hipp_img_TA_meta/.

Code availability

We made use of publicly available software and tools. All codes used to generate results that are reported in this paper are available upon request.

The Full List of CHIMGEN Consortium Authors

Department of Radiology and Tianjin Key Laboratory of Functional Imaging, Tianjin Medical University General Hospital

Chunshui Yu; Quan Zhang; Wen Qin; Feng Liu; Junping Wang; Qiang Xu; Jiayuan Xu; Xue Zhang; Xinjun Suo; Jilian Fu; Congcong Yuan; Yuan Ji; Hui Xue; Tianying Gao; Junpeng Liu; Yanjun Li; Xi Guo; Lixue Xu; Jiajia Zhu; Huaigui Liu; Fangshi Zhao; Jie Sun; Yongjie Xu; Huanhuan Cai; Jie Tang; Yaodan Zhang; Yongqin Xiong; Xianting Sun; Nannan Pan; Xue Zhang (Junior); Jiayang Yang; Nana Liu; Ya Wen; Dan Zhu; Bingjie Wu; Wenshuang Zhu; Qingqing Diao; Yujuan Cao; Bingbing Yang; Lining Guo; Yingying Xie; Jiahui Lin; Zhimin Li; Yan Zhang; Kaizhong Xue; Zirui Wang; Junlin Shen

School of Medical Imaging, Tianjin Medical University

Meng Liang; Xuejun Zhang; Hao Ding; Qian Su

Department of Bioinformatics, The Province and Ministry Co-sponsored Collaborative Innovation Center for Medical Epigenetics, School of Basic Medical Sciences, Tianjin Medical University

Mulin Jun Li; Shijie Zhang

Department of Radiology, Tianjin Medical University Cancer Institute and Hospital

Zhaoxiang Ye; Peng Zhang; Wei Li

Department of Radiology, Henan Provincial People's Hospital & Zhengzhou University People's Hospital

Meiyun Wang; Dapeng Shi; Lun Ma; Yan Bai; Min Guan; Wei Wei

Department of Magnetic Resonance Imaging, The First Affiliated Hospital of Zhengzhou University

Jingliang Cheng; Caihong Wang; Peifang Miao; Fuhong Duan; Yafei Guo; Weijian Wang

Department of Medical Imaging, Jinling Hospital, Medical School of Nanjing University

Longjiang Zhang; Lijuan Zheng; Lilin; Yunfei Wang; Han Zhang; Xinyuan Zhang

Department of Radiology, Drum Tower Hospital, Medical School of Nanjing University

Bing Zhang; Zhao Qing; Sichu Wu; Junxia Wang; Yi Sun; Yang He

Institute of Psychology, Chinese Academy of Sciences

Xi-Nian Zuo; Zhe Zhang; Yin-Shan Wang; Quan Zhou

Department of Radiology, Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology

Wenzhen Zhu; Tian Tian

Department of Medical Imaging, The First Affiliated Hospital of Guangzhou University of Traditional Chinese Medicine

Shijun Qiu; Yi Liang; Yujie Liu; Hui Zeng; Jingxian Chen

Department of Radiology, The Affiliated Hospital of Xuzhou Medical University

Kai Xu; Haitao Ge; Peng Xu; Cailuan Lu; Chen Wu; Xiaoying Yang

Department of Medical Imaging, The Second Hospital of Hebei Medical University

Zuojun Geng; Yuzhao Wang; Yaikai Wu; Xuran Feng; Ling Li; Duo Gao

Division of Life Science and Medicine, University of Science & Technology of China

Xiaochu Zhang; Rujing Zha; Ying Li; Lizhuang Yang; Ying Chen; Ling Zuo

Center for MRI Research, Academy for Advanced Interdisciplinary Studies, Peking University

Jia-Hong Gao; Jianqiao Ge; Guoyuan Yang

Functional and Molecular Imaging Key Lab of Shaanxi Province & Department of Radiology, Tangdu Hospital, Air Force Medical University

Guangbin Cui; Wen Wang; Linfeng Yan; Yang Yang; Jin Zhang

Department of Radiology, Beijing Tongren Hospital, Capital Medical University

Junfang Xian; Qian Wang; Xiaoxia Qu; Ying Wang

Department of Radiology, Characteristic Medical Center of Chinese People's Armed Police Force

Quan Zhang; Fei Yuan

Department of Radiology, Qilu Hospital of Shandong University

Dawei Wang; Li Hu; Jizhen Li

Department of Radiology, The First Affiliated Hospital of Dalian Medical University

Yanwei Miao; Weiwei Wang; Yujing Zhou

Department of Radiology, Tianjin First Center Hospital

Wen Shen; Miaomiao Long; Lihua Liu

Department of Radiology, The First Affiliated Hospital of Anhui Medical University

Yongqiang Yu; Xiaohu Li; Xiaoshu Li

Department of Radiology, The First Affiliated Hospital of Wenzhou Medical University

Jiance Li; Yunjun Yang; Nengzhi Xia

Department of Radiology, Xiangya Hospital, Central South University

Weihua Liao; Shuai Yang; Youming Zhang

Department of Magnetic Resonance, Lanzhou University Second Hospital

Jing Zhang; Guangyao Liu; Laiyang Ma

Department of Radiology, The First Hospital of Shanxi Medical University

Hui Zhang; Xiaochun Wang; Ying Lei

Department of Radiology, Yantai Yuhuangding Hospital

Bo Gao; Gang Zhang; Kang Yuan

Department of Radiology, The Second Affiliated Hospital of Zhejiang University, School of Medicine

Xiaojun Xu; Jingjing Xu; Xiaojun Guan

Department of Radiology, Hainan General Hospital

Feng Chen; Yuankai Lin; Hui Juan Chen

Department of Radiology, The Second Affiliated Hospital and Yuying Children's Hospital of Wenzhou Medical University

Zhihan Yan; Yuchuan Fu; Yi Lu

Department of Radiology, Tianjin Huanhu Hospital

Tong Han; Jun Guo; Hao Lu

Department of Radiology, Huashan Hospital, Fudan University

Zhenwei Yao; Yue Wu

Department of Radiology, the Center for Medical Imaging, West China Hospital of Sichuan University

Su Lui

1 Scoville, W. B. & Milner, B. Loss of recent memory after bilateral hippocampal lesions. J Neurol Neurosurg Psychiatry 20, 11-21, doi:10.1136/jnnp.20.1.11 (1957).

2 Morris, R. G., Garrud, P., Rawlins, J. N. & O'Keefe, J. Place navigation impaired in rats with hippocampal lesions. Nature 297, 681-683, doi:10.1038/297681a0 (1982).

3 Lisman, J. et al. Viewpoints: how the hippocampus contributes to memory, navigation and cognition. Nat Neurosci 20, 1434-1447, doi:10.1038/nn.4661 (2017).

4 Kim, J. J. & Diamond, D. M. The stressed hippocampus, synaptic plasticity and lost memories. Nat Rev Neurosci 3, 453-462, doi:10.1038/nrn849 (2002).

5 Moreno-Jiménez, E. P. et al. Adult hippocampal neurogenesis is abundant in neurologically healthy subjects and drops sharply in patients with Alzheimer's disease. Nat Med 25, 554-560, doi:10.1038/s41591-019-0375-9 (2019).

6 Li, J. Q. et al. Risk factors for predicting progression from mild cognitive impairment to Alzheimer's disease: a systematic review and meta-analysis of cohort studies. J Neurol Neurosurg Psychiatry 87, 476-484, doi:10.1136/jnnp-2014-310095 (2016).

7 Calabresi, P., Castrioto, A., Di Filippo, M. & Picconi, B. New experimental and clinical links between the hippocampus and the dopaminergic system in Parkinson's disease. Lancet Neurol 12, 811-821, doi:10.1016/s1474-4422(13)70118-2 (2013).

8 Mattai, A. et al. Hippocampal volume development in healthy siblings of childhood-onset schizophrenia patients. Am J Psychiatry 168, 427-435, doi:10.1176/appi.ajp.2010.10050681 (2011).

9 Treadway, M. T. et al. Illness progression, recent stress, and morphometry of hippocampal subfields and medial prefrontal cortex in major depression. Biol Psychiatry 77, 285-294, doi:10.1016/j.biopsych.2014.06.018 (2015).

10 Gonçalves, J. T., Schafer, S. T. & Gage, F. H. Adult Neurogenesis in the Hippocampus: From Stem Cells to Behavior. Cell 167, 897-914, doi:10.1016/j.cell.2016.10.021 (2016).

11 Kühn, S. et al. Plasticity of hippocampal subfield volume cornu ammonis 2+3 over the course of withdrawal in patients with alcohol dependence. JAMA Psychiatry 71, 806-811, doi:10.1001/jamapsychiatry.2014.352 (2014).

12 Pitman, R. K. et al. Biological studies of post-traumatic stress disorder. Nat Rev Neurosci 13, 769-787, doi:10.1038/nrn3339 (2012).

13 Fanselow, M. S. & Dong, H. W. Are the dorsal and ventral hippocampus functionally distinct structures? Neuron 65, 7-19, doi:10.1016/j.neuron.2009.11.031 (2010).

14 Iglesias, J. E. et al. A computational atlas of the hippocampal formation using ex vivo, ultra-high resolution MRI: Application to adaptive segmentation of in vivo MRI. Neuroimage 115, 117-137, doi:10.1016/j.neuroimage.2015.04.042 (2015).

15 Ho, N. F. et al. Progression from selective to general involvement of hippocampal subfields in schizophrenia. Mol Psychiatry 22, 142-152, doi:10.1038/mp.2016.4 (2017).

16 Small, S. A., Schobel, S. A., Buxton, R. B., Witter, M. P. & Barnes, C. A. A pathophysiological framework of hippocampal dysfunction in ageing and disease. Nat Rev Neurosci 12, 585-601, doi:10.1038/nrn3085 (2011).

17 Hibar, D. P. et al. Common genetic variants influence human subcortical brain structures. Nature 520, 224-229, doi:10.1038/nature14101 (2015).

18 Bis, J. C. et al. Common variants at 12q14 and 12q24 are associated with hippocampal volume. Nat Genet 44, 545-551, doi:10.1038/ng.2237 (2012).

19 Stein, J. L. et al. Identification of common variants associated with human hippocampal and intracranial volumes. Nat Genet 44, 552-561, doi:10.1038/ng.2250 (2012).

20 Hibar, D. P. et al. Novel genetic loci associated with hippocampal volume. Nat Commun 8, 13624, doi:10.1038/ncomms13624 (2017).

21 Elliott, L. T. et al. Genome-wide association studies of brain imaging phenotypes in UK Biobank. Nature 562, 210-216, doi:10.1038/s41586-018-0571-7 (2018).

22 Zhao, B. et al. Genome-wide association analysis of 19,629 individuals identifies variants influencing regional brain volumes and refines their genetic co-architecture with cognitive and mental health traits. Nat Genet 51, 1637-1644, doi:10.1038/s41588-019-0516-6 (2019).

23 Smith, S. M. et al. An expanded set of genome-wide association studies of brain imaging phenotypes in UK Biobank. Nature Neuroscience 24, 737-745, doi:10.1038/s41593-021-00826-4 (2021).

24 van der Meer, D. et al. Brain scans from 21,297 individuals reveal the genetic architecture of hippocampal subfield volumes. Mol Psychiatry 25, 3053-3065, doi:10.1038/s41380-018-0262-7 (2020).

25 Zahodne, L. B. et al. Structural MRI Predictors of Late-Life Cognition Differ Across African Americans, Hispanics, and Whites. Curr Alzheimer Res 12, 632-639, doi:10.2174/1567205012666150530203214 (2015).

26 Li, M. et al. Allelic differences between Europeans and Chinese for CREB1 SNPs and their implications in gene expression regulation, hippocampal structure and function, and bipolar disorder susceptibility. Mol Psychiatry 19, 452-461, doi:10.1038/mp.2013.37 (2014).

27 Li, M. et al. Failure of replicating the association between hippocampal volume and 3 single-nucleotide polymorphisms identified from the European genome-wide association study in Asian populations. Neurobiol Aging 35, 2883.e2881-2883.e2882, doi:10.1016/j.neurobiolaging.2014.07.015 (2014).

28 Peterson, R. E. et al. Genome-wide Association Studies in Ancestrally Diverse Populations: Opportunities, Methods, Pitfalls, and Recommendations. Cell 179, 589-603, doi:10.1016/j.cell.2019.08.051 (2019).

29 Li, Y. R. & Keating, B. J. Trans-ethnic genome-wide association studies: advantages and challenges of mapping in diverse populations. Genome Med 6, 91, doi:10.1186/s13073-014-0091-5 (2014).

30 Chen, M. H. et al. Trans-ethnic and Ancestry-Specific Blood-Cell Genetics in 746,667 Individuals from 5 Global Populations. Cell 182, 1198-1213.e1114, doi:10.1016/j.cell.2020.06.045 (2020).

31 Schaid, D. J., Chen, W. & Larson, N. B. From genome-wide associations to candidate causal variants by statistical fine-mapping. Nat Rev Genet 19, 491-504, doi:10.1038/s41576-018-0016-z (2018).

32 Chen, J. et al. The trans-ancestral genomic architecture of glycemic traits. Nat Genet 53, 840-860, doi:10.1038/s41588-021-00852-9 (2021).

33 Wojcik, G. L. et al. Genetic analyses of diverse populations improves discovery for complex traits. Nature 570, 514-518, doi:10.1038/s41586-019-1310-4 (2019).

34 Xu, Q. et al. CHIMGEN: a Chinese imaging genetics cohort to enhance cross-ethnic and cross-geographic brain research. Mol Psychiatry 25, 517-529, doi:10.1038/s41380-019-0627-6 (2020).

35 Han, B. & Eskin, E. Random-effects model aimed at discovering associations in meta-analysis of genome-wide association studies. Am J Hum Genet 88, 586-598, doi:10.1016/j.ajhg.2011.04.014 (2011).

36 Auton, A. et al. A global reference for human genetic variation. Nature 526, 68-74, doi:10.1038/nature15393 (2015).

37 Kichaev, G. & Pasaniuc, B. Leveraging Functional-Annotation Data in Trans-ethnic Fine-Mapping Studies. Am J Hum Genet 97, 260-271, doi:10.1016/j.ajhg.2015.06.007 (2015).

38 Maller, J. B. et al. Bayesian refinement of association signals for 14 loci in 3 common diseases. Nat Genet 44, 1294-1301, doi:10.1038/ng.2435 (2012).

39 Duncan, L. et al. Analysis of polygenic risk score usage and performance in diverse human populations. Nat Commun 10, 3328, doi:10.1038/s41467-019-11112-0 (2019).

40 Martin, A. R. et al. Clinical use of current polygenic risk scores may exacerbate health disparities. Nature Genetics 51, 584-591, doi:10.1038/s41588-019-0379-x (2019).

41 Koyama, S. et al. Population-specific and trans-ancestry genome-wide analyses identify distinct and shared genetic risk loci for coronary artery disease. Nat Genet 52, 1169-1177, doi:10.1038/s41588-020-0705-3 (2020).

42 Watanabe, K., Taskesen, E., van Bochoven, A. & Posthuma, D. Functional mapping and annotation of genetic associations with FUMA. Nat Commun 8, 1826, doi:10.1038/s41467-017-01261-5 (2017).

43 Wang, K., Li, M. & Hakonarson, H. ANNOVAR: functional annotation of genetic variants from high-throughput sequencing data. Nucleic Acids Res 38, e164, doi:10.1093/nar/gkq603 (2010).

44 Kircher, M. et al. A general framework for estimating the relative pathogenicity of human genetic variants. Nat Genet 46, 310-315, doi:10.1038/ng.2892 (2014).

45 Battle, A., Brown, C. D., Engelhardt, B. E. & Montgomery, S. B. Genetic effects on gene expression across human tissues. Nature 550, 204-213, doi:10.1038/nature24277 (2017).

46 Chen, L., Song, D., Chen, B., Yang, X. & Cheng, O. Activation of liver X receptor promotes hippocampal neurogenesis and improves long-term cognitive function recovery in acute cerebral ischemia-reperfusion mice. J Neurochem 154, 205-217, doi:10.1111/jnc.14890 (2020).

47 Zhao, L. et al. The roles of liver X receptor α in inflammation and inflammation-associated diseases. J Cell Physiol 236, 4807-4828, doi:10.1002/jcp.30204 (2021).

48 Del Villar, K. & Miller, C. A. Down-regulation of DENN/MADD, a TNF receptor binding protein, correlates with neuronal cell death in Alzheimer's disease brain and hippocampal neurons. Proc Natl Acad Sci U S A 101, 4210-4215, doi:10.1073/pnas.0307349101 (2004).

49 Liu, Y. et al. APOE genotype and neuroimaging markers of Alzheimer's disease: systematic review and meta-analysis. J Neurol Neurosurg Psychiatry 86, 127-134, doi:10.1136/jnnp-2014-307719 (2015).

50 Belloy, M. E., Napolioni, V. & Greicius, M. D. A Quarter Century of APOE and Alzheimer's Disease: Progress to Date and the Path Forward. Neuron 101, 820-838, doi:10.1016/j.neuron.2019.01.056 (2019).

51 Litterman, N. et al. An OBSL1-Cul7Fbxw8 ubiquitin ligase signaling mechanism regulates Golgi morphology and dendrite patterning. PLoS Biol 9, e1001060, doi:10.1371/journal.pbio.1001060 (2011).

52 Diao, J. et al. ATG14 promotes membrane tethering and fusion of autophagosomes to endolysosomes. Nature 520, 563-566, doi:10.1038/nature14147 (2015).

53 Lee, K. M., Hwang, S. K. & Lee, J. A. Neuronal autophagy and neurodevelopmental disorders. Exp Neurobiol 22, 133-142, doi:10.5607/en.2013.22.3.133 (2013).

54 Artuso, R. et al. Investigation of modifier genes within copy number variations in Rett syndrome. J Hum Genet 56, 508-515, doi:10.1038/jhg.2011.50 (2011).

55 Sparse whole-genome sequencing identifies two loci for major depressive disorder. Nature 523, 588-591, doi:10.1038/nature14659 (2015).

56 Shimazu, T., Barjau, J., Sohtome, Y., Sodeoka, M. & Shinkai, Y. Selenium-based S-adenosylmethionine analog reveals the mammalian seven-beta-strand methyltransferase METTL10 to be an EF1A1 lysine methyltransferase. PLoS One 9, e105394, doi:10.1371/journal.pone.0105394 (2014).

57 Liao, Y., Wang, J., Jaehnig, E. J., Shi, Z. & Zhang, B. WebGestalt 2019: gene set analysis toolkit with revamped UIs and APIs. Nucleic Acids Res 47, W199-w205, doi:10.1093/nar/gkz401 (2019).

58 Buniello, A. et al. The NHGRI-EBI GWAS Catalog of published genome-wide association studies, targeted arrays and summary statistics 2019. Nucleic Acids Res 47, D1005-d1012, doi:10.1093/nar/gky1120 (2019).

59 Jacobs, S. et al. Mice with targeted Slc4a10 gene disruption have small brain ventricles and show reduced neuronal excitability. Proc Natl Acad Sci U S A 105, 311-316, doi:10.1073/pnas.0705487105 (2008).

60 Sinning, A., Liebmann, L. & Hübner, C. A. Disruption of Slc4a10 augments neuronal excitability and modulates synaptic short-term plasticity. Front Cell Neurosci 9, 223, doi:10.3389/fncel.2015.00223 (2015).

61 Sun, C. et al. Nonenzymatic function of DPP4 in diabetes-associated mitochondrial dysfunction and cognitive impairment. Alzheimers Dement, doi:10.1002/alz.12437 (2021).

62 Takamatsu, G. et al. Tescalcin is a potential target of class I histone deacetylase inhibitors in neurons. Biochem Biophys Res Commun 482, 1327-1333, doi:10.1016/j.bbrc.2016.12.036 (2017).

63 Antonelli, F., Casciati, A. & Pazzaglia, S. Sonic hedgehog signaling controls dentate gyrus patterning and adult neurogenesis in the hippocampus. Neural Regen Res 14, 59-61, doi:10.4103/1673-5374.243703 (2019).

64 Haukvik, U. K. et al. In vivo hippocampal subfield volumes in schizophrenia and bipolar disorder. Biol Psychiatry 77, 581-588, doi:10.1016/j.biopsych.2014.06.020 (2015).

65 Hussaini, S. M. et al. Wnt signaling in neuropsychiatric disorders: ties with adult hippocampal neurogenesis and behavior. Neurosci Biobehav Rev 47, 369-383, doi:10.1016/j.neubiorev.2014.09.005 (2014).

Study populations

CHIMGEN

The Chinese Imaging Genetics (CHIMGEN) study (http://chimgen.tmu.edu.cn/) has collected genomic, neuroimaging, environmental, and behavioral data from 7,306 healthy Chinese Han participants aged 18-30 years³⁴. Participants were recruited from 32 centers and neuroimaging data were acquired by 30 MR scanners. (Supplementary Table 12). The inclusion and exclusion criteria are presented in Supplementary Table 13, and the recruitment processes, standard operating procedures, and data quality control can be found in the website (http://chimgen.tmu.edu.cn/)³⁴. Informed consent was obtained from all participants and the study was approved by local ethical committees of all research institutions.

Genetic data processing

In the 7,306 CHIMGEN participants, 7,195 were genotyped by Illumina ASA-750K (Asian Screening Array), which was specially designed for EAS individuals. The genotyped data were aligned to the human reference genome (GRCh37/hg19). The quality control of the CHIMGEN genetic data was performed by PLINK⁶⁶ (http://zzz.bwh.harvard.edu/plink/). In the variant-level quality control, we excluded duplicated variants and variants with call rate < 95%, minor allele frequency (MAF) < 0.001, and Hardy-Weinberg equilibrium (HWE) P < 1 × 10^-6. In the sample-level quality control, we excluded participants with sex mismatch between self-reported and genotyping data, close relations estimated by identity by descent (IBD > 0.1875), excess heterozygosity (> mean ± 5SD), missing genotypes > 3%, and outliers relative to EAS identified by principal components analysis (PCA). The numbers of variants and participants removed in each step are reported in Extended Data Fig. 6. After the variant- and sample-level quality control, a total of 7,163 individuals and 549,309 variants were included in the further imputation. The obtained SNPs were pre-phased by SHAPEIT2⁶⁷ and imputed by IMPUTE2⁶⁸ with the merged reference panel from 1000 Genomes Project (1KGP)³⁶ and SG10K project⁶⁹. Based on the criteria of MAF 0.01 and IMPUTE2 info quality score 0.9, 6,830,146 imputed SNPs were finally included in GWAS.

MRI data processing

Structural MRI data were acquired by 10 types of 3.0-Tesla MRI scanners, and scanning parameters are presented in Supplementary Table 14. The image quality of each scanner and consistency across scanners were assessed³⁴, and quality control was performed to exclude participants with unqualified imaging data, details please see the website (http://chimgen.tmu.edu.cn/en/). Of the 7,163 participants with qualified genetic data, 23 participants without structural MRI data and 79 participants with unqualified structural MRI data were excluded. Hippocampal and subfield volumes were estimated by a subfield segmentation algorithm provided by FreeSurfer v7.0⁷⁰ (https://surfer.nmr.mgh.harvard.edu/). This algorithm can automatically segment out hippocampal subfields based on a probabilistic atlas, which was built with ultra-high-resolution ex vivo MRI data¹⁴. According to the atlas, the left or right hippocampus can be segmented into 19 subfields, including the parasubiculum, hippocampus-amygdala-transition-area (HATA), fimbria, hippocampal tail, hippocampal fissure, and heads and bodies of the presubiculum, subiculum, CA1, CA3 (including the CA2 and CA3 subfields), CA4, granule cell and molecular layer of the dentate gyrus (GC-ML-DG), and molecular layer of hippocampus proper (molecular layer HP). We also included the volumes of the entire hippocampus, hippocampal head, and hippocampal body as three additional phenotypes, and thus we finally generated 44 traits for hippocampal and subfield volumes. After excluding 22 participants with segmentation failure, we included 7,039 participants with qualified data of the hippocampal and subfield volumes obtained from the 30 MR scanners (Supplementary Fig. 5). The ComBat method was used to harmonize hippocampal and subfield volumes data across scanners, which can remove between-scanner variation and preserve biological variability⁷¹. Then the rank-based inverse Gaussian transformation was applied to enhance Gaussianity and to reduce the influence of outliers.

UKBB

The UK Biobank (UKBB) cohort is a prospective study of 500,000 individuals aged between 40-69 years old at recruitment⁷². UKBB has collected a wide variety of phenotypes and biological samples, a portion of the participants have performed brain MRI examinations⁷³. With UKBB data, two GWASs have included hippocampal volume or hippocampal subfield volumes as the studied traits, and provided GWAS summary statistics²¹^,²³. One study included 8,411 EUR participants as the discovery dataset and provided GWAS summary statistics for hippocampal volume²¹. Another study included 31,968 EUR participants (21,282 for discovery) and provided GWAS summary statistics for both hippocampal volume and hippocampal subfield volumes²³ (https://open.win.ox.ac.uk/ukbiobank/big40/). In the latter study, the hippocampus and hippocampal subfields were also segmented by FreeSurfer v7.0, and thus GWAS summary statistics of this study²³ were used for the trans-ancestry meta-analysis.

ENIGMA

The enhancing neuroimaging genetics through meta-analysis (ENIGMA) consortia (http://enigma.ini.usc.edu/) focuses on neuroimaging genetics by pooling worldwide data⁷⁴. The consortia have conducted a collaborative large-scale GWAS of hippocampal volume to identify genetic associations²⁰. The GWAS summary statistics of the mean hippocampus volume was derived from a meta-analysis of 26,814 EUR participants collected from the ENIGMA consortium and the cohort for heart and aging research in genomic epidemiology (CHARGE). The hippocampal volume was segmented using FMRIB Software Library (FSL)⁷⁵ and FreeSurfer, details about genotyping, quality control, and genome-wide meta-analysis can be found in the original paper²⁰.

GWAS of hippocampal and subfield volumes in CHIMGEN

We used BGENIE v1.3^21,76 (https://jmarchini.org/bgenie/) to perform GWASs with an additive model, which tested linear associations between qualified genetic variants and hippocampal volumetric traits (n = 44) in the 7,039 CHIMGEN participants, while controlling for the effects of sex, age, age², age × sex, total intracranial volume, and top 3 genetic principal components from PCA.

Trans-ancestry meta-analysis of GWAS summary statistics

Trans-ancestry meta-analysis was performed based on GWAS summary statistics of hippocampal and subfield volumes in EUR participants from UKBB and ENIGMA and in EAS participants from CHIMGEN. All the base positions of these datasets were based on GRCh37/hg19 human reference genome. For GWAS summary data from UKBB, we removed genetic variants with MAF < 0.01. Indel variants were removed because GWAS summary statistics of ENIGMA did not have specific allele information. To handle tri-allelic variants, we matched each variant across the three datasets based on the chromosome, base position (hg19), and alleles. We applied the METASOFT v.2³⁵ (http://genetics.cs.ucla.edu/meta/) to perform the meta-analysis. METASOFT can perform meta-analysis based on either the fixed-effects model or the random-effects model. The former estimates the inverse variance weighted effect size, and the latter detects genetic associations under heterogeneity by Han and Eskin’s Random Effects model. Because most of the downstream analyses in this study needed beta value and standard error that can only be obtained from the fixed-effects model, we reported the results derived from the fixed-effects model as the main results and validated these results by the Han and Eskin’s Random Effects model. Cochran’s Q test was used to assess the degree of heterogeneity between EAS and EUR GWAS by estimating the effect size for each genetic variant. Cochran’s Q is based on the test with a null hypothesis that homogeneity existed between studies and an alternative hypothesis that heterogeneity existed between studies. The genome-wide significance threshold was set at P < 5 × 10^-8. In a given genomic region, a lead SNP was defined as the SNP with the lowest p-value, and its locus was defined as the ±500 kb region centered at the lead SNP. A locus was considered novel if the lead SNP was ±500 kb away from any known loci for the hippocampal volumetric traits of interest.

Comparing linkage disequilibrium between ancestries

LD was measured using r² between SNPs by PLINK with 1000 Genomes Phase 3 reference dataset. For each locus, we calculated the r² between the lead SNP and all remaining SNPs in corresponding locus in EAS and EUR. The most commonly used threshold of r² is 0.1 in estimating LD and LD decay⁷⁷. We compared the r² between SNPs that both greater than 0.01 in the two populations.

Statistical fine-mapping

Statistical fine-mapping was carried out using the probabilistic annotation integrator (PAINTOR)³⁷ (https://github.com/gkichaev/PAINTOR_V3.0/wiki). For each included SNP, PAINTOR makes a causal inference by generating posterior probability based on the Bayes theorem with the effect of associations and the LD pattern. For each locus, we set a window of 1 Mb centered at the lead SNP and one causal variant was assumed. For GWAS summary data from a single ancestral population (EUR or EAS), the LD value of each SNP was estimated based on 1000 genomes phase 3 reference data of EUR or EAS. For GWAS summary data from the trans-ancestry analysis, we first estimated the LD values with the reference data for each ancestry, and then calculated a weighted LD value for each SNP by including the sample size of each population in trans-ancestry meta-analysis as respective weight³³. The 95% credible set of a locus was calculated based on the posterior probabilities of SNPs within the locus. Specifically, the posterior probabilities of the SNPs within one locus were ordered from the largest to the smallest, and then we accumulated the posterior probabilities from the largest one until the joint probability ≥ 95%, the list of SNPs formed the 95% credible set. To validate the fine-mapping results derived from PAINTOR, we also calculated approximate Bayes factors (aBF)³⁸ for each SNP within the locus using its beta and SE values, and then divided aBF of the SNP by the accumulated aBF for all SNPs within the locus to calculate the posterior probability of the SNP. The 95% credible set for each locus was determined by the joint probability ≥ 95% based on the ordering of the posterior probabilities of SNPs within the locus.

Polygenic score (PGS)

The PGS was calculated using PRSice v.1.25 software⁷⁸ (https://www.prsice.info/). PRSice removed the ambiguous variants and matched SNPs between base and target datasets, applied LD-pruning on the target dataset (clump-kb 250kb, clump-r² 0.1). PRSice performed optimization of the p-value threshold for the most accurate prediction, as measured by r-squared (r²), the explained variance from a multivariate regression model. We used the r² to estimate the predictive accuracy. The effects of sex, age, age², age × sex, total intracranial volume, and top 3 genetic principal components from PCA were controlled during the analysis.

Annotation of the identified SNPs

FUMA⁴² (https://fuma.ctglab.nl/) was used to perform functional annotation of SNPs identified in the trans-ancestry meta-analysis. FUMA is an online platform for annotating and prioritizing genetic variants by integrating comprehensive annotation data sources. We included all SNPs with genome-wide significance (P < 5 × 10^-8) in functional annotation and did not use the LD information. The SNPs were mapped to the databases of functional annotations, including ANNOVAR categories⁴³, combined annotation-dependent depletion (CADD) scores⁴⁴ and eQTL dataset⁴⁵ of hippocampal tissue provided by GTEx. ANNOVAR annotates the SNPs based on the SNP’s genic position, such as exon, intron, UTR3, and intergenic region. CADD scores prioritize deleterious and pathogenic variants. A SNP is considered potentially pathogenic when the CADD score was above 12.37⁷⁹. The eQTL dataset describes the relationships between SNPs and gene expression in hippocampal tissue, an FDR of 0.05 was used to define the genome-wide significant eQTL associations. The SNPs were also mapped to GWAS catalog⁵⁸ (https://www.ebi.ac.uk/gwas/) of brain-related phenotypes and neuropsychiatric diseases by FUMA.

The identified genetic variants by the trans-ancestry meta-analysis were mapped to genes based on physical distance; each SNP was mapped to the nearest gene in the human reference assembly (GRCh37/hg19). The generated genes were investigated by performing a statistical over representation analysis using WebGestalt⁵⁷ online tool (http://www.webgestalt.org/) based on GO biological processes. We used the BH-FDR correction for multiple testing (q_c < 0.05).

References of Method

66 Purcell, S. et al. PLINK: a tool set for whole-genome association and population-based linkage analyses. Am J Hum Genet 81, 559-575, doi:10.1086/519795 (2007).

67 Delaneau, O., Zagury, J. F. & Marchini, J. Improved whole-chromosome phasing for disease and population genetic studies. Nat Methods 10, 5-6, doi:10.1038/nmeth.2307 (2013).

68 Howie, B. N., Donnelly, P. & Marchini, J. A flexible and accurate genotype imputation method for the next generation of genome-wide association studies. PLoS Genet 5, e1000529, doi:10.1371/journal.pgen.1000529 (2009).

69 Wu, D. et al. Large-Scale Whole-Genome Sequencing of Three Diverse Asian Populations in Singapore. Cell 179, 736-749.e715, doi:10.1016/j.cell.2019.09.019 (2019).

70 Fischl, B. et al. Whole brain segmentation: automated labeling of neuroanatomical structures in the human brain. Neuron 33, 341-355, doi:10.1016/s0896-6273(02)00569-x (2002).

71 Fortin, J. P. et al. Harmonization of multi-site diffusion tensor imaging data. Neuroimage 161, 149-170, doi:10.1016/j.neuroimage.2017.08.047 (2017).

72 Sudlow, C. et al. UK biobank: an open access resource for identifying the causes of a wide range of complex diseases of middle and old age. PLoS Med 12, e1001779, doi:10.1371/journal.pmed.1001779 (2015).

73 Miller, K. L. et al. Multimodal population brain imaging in the UK Biobank prospective epidemiological study. Nat Neurosci 19, 1523-1536, doi:10.1038/nn.4393 (2016).

74 Thompson, P. M. et al. ENIGMA and global neuroscience: A decade of large-scale studies of the brain in health and disease across more than 40 countries. Transl Psychiatry 10, 100, doi:10.1038/s41398-020-0705-1 (2020).

75 Jenkinson, M., Beckmann, C. F., Behrens, T. E., Woolrich, M. W. & Smith, S. M. FSL. Neuroimage 62, 782-790, doi:10.1016/j.neuroimage.2011.09.015 (2012).

76 Bycroft, C. et al. The UK Biobank resource with deep phenotyping and genomic data. Nature 562, 203-209, doi:10.1038/s41586-018-0579-z (2018).

77 Vos, P. G. et al. Evaluation of LD decay and various LD-decay estimators in simulated and SNP-array data of tetraploid potato. Theor Appl Genet 130, 123-135, doi:10.1007/s00122-016-2798-8 (2017).

78 Choi, S. W., Mak, T. S. & O'Reilly, P. F. Tutorial: a guide to performing polygenic risk score analyses. Nat Protoc 15, 2759-2772, doi:10.1038/s41596-020-0353-1 (2020).

79 Amendola, L. M. et al. Actionable exomic incidental findings in 6503 participants: challenges of variant classification. Genome Res 25, 305-315, doi:10.1101/gr.183483.114 (2015).

There is NO Competing Interest.

ExtendedDataFigures.docx
Extended Data Figures 1-6
SupplementaryTables.xlsx
Supplementary Tables 1-14

Download PDF

Journal Publication

published 19 Jun, 2023

Read the published version in Nature Genetics →

Version 1

posted

You are reading this latest preprint version

Trans-ancestry genome-wide association meta-analyses of hippocampal and subfield volumes

Status:

Journal Publication

Version 1

Abstract

Figures

Introduction

Results

Trans-ancestry genetic associations of hippocampal and subfield volumes

Ancestry-shared and ancestry-specific genetic associations of hippocampal and subfield volumes

Trans-ancestry analysis improves the resolution of fine-mapping

Transferability for PGS of trans-ancestry meta-analysis

Functional annotation of genetic loci associated with hippocampal and subfield volumes

Phenome-wide association studies of genetic loci associated with hippocampal and subfield volumes

Conclusions

Declarations

References

method

Additional Declarations

Supplementary Files

Status:

Journal Publication

Version 1