Genome-wide association analysis of left ventricular imaging-derived phenotypes identifies 72 risk loci and yields genetic insights into hypertrophic cardiomyopathy

Ning, Caibo; Fan, Linyun; Jin, Meng; Wang, Wenji; Hu, Zhiqiang; Cai, Yimin; Chen, Liangkai; Lu, Zequn; Zhang, Ming; Chen, Can; Li, Yanmin; Zhang, Fuwei; Wang, Wenzhuo; Liu, Yizhuo; Chen, Shuoni; Jiang, Yuan; He, Chunyi; Wang, Zhuo; Chen, Xu; Li, Hanting; Li, Gaoyuan; Ma, Qianying; Geng, Hui; Tian, Wen; Zhang, Heng; Liu, Bo; Xia, Qing; Yang, Xiaojun; Liu, Zhongchun; Li, Bin; Zhu, Ying; Li, Xiangpan; Zhang, Shaoting; Tian, Jianbo; Miao, Xiaoping

doi:10.1038/s41467-023-43771-5

Download PDF

Article
Open access
Published: 30 November 2023

Genome-wide association analysis of left ventricular imaging-derived phenotypes identifies 72 risk loci and yields genetic insights into hypertrophic cardiomyopathy

Caibo Ning^1,2,3^na1,
Linyun Fan^1,2,3^na1,
Meng Jin⁴^na1,
Wenji Wang⁵^na1,
Zhiqiang Hu⁵^na1,
Yimin Cai ORCID: orcid.org/0000-0002-2458-1976¹^na1,
Liangkai Chen ORCID: orcid.org/0000-0002-5227-4236⁶^na1,
Zequn Lu¹,
Ming Zhang¹,
Can Chen¹,
Yanmin Li¹,
Fuwei Zhang¹,
Wenzhuo Wang¹,
Yizhuo Liu¹,
Shuoni Chen¹,
Yuan Jiang¹,
Chunyi He¹,
Zhuo Wang¹,
Xu Chen¹,
Hanting Li¹,
Gaoyuan Li¹,
Qianying Ma¹,
Hui Geng¹,
Wen Tian¹,
Heng Zhang¹,
Bo Liu ORCID: orcid.org/0000-0001-5118-1890⁴,
Qing Xia⁵,
Xiaojun Yang⁷,
Zhongchun Liu ORCID: orcid.org/0000-0001-5410-0312⁸,
Bin Li¹,
Ying Zhu^1,2,3,9,
Xiangpan Li²,
Shaoting Zhang ORCID: orcid.org/0000-0002-8719-448X^5,10,
Jianbo Tian ORCID: orcid.org/0000-0001-9493-694X^1,2,3,9 &
…
Xiaoping Miao ORCID: orcid.org/0000-0002-6818-9722^1,2,3,9

Nature Communications volume 14, Article number: 7900 (2023) Cite this article

3686 Accesses
Metrics details

Subjects

Abstract

Left ventricular regional wall thickness (LVRWT) is an independent predictor of morbidity and mortality in cardiovascular diseases (CVDs). To identify specific genetic influences on individual LVRWT, we established a novel deep learning algorithm to calculate 12 LVRWTs accurately in 42,194 individuals from the UK Biobank with cardiac magnetic resonance (CMR) imaging. Genome-wide association studies of CMR-derived 12 LVRWTs identified 72 significant genetic loci associated with at least one LVRWT phenotype (P < 5 × 10⁻⁸), which were revealed to actively participate in heart development and contraction pathways. Significant causal relationships were observed between the LVRWT traits and hypertrophic cardiomyopathy (HCM) using genetic correlation and Mendelian randomization analyses (P < 0.01). The polygenic risk score of inferoseptal LVRWT at end systole exhibited a notable association with incident HCM, facilitating the identification of high-risk individuals. The findings yield insights into the genetic determinants of LVRWT phenotypes and shed light on the biological basis for HCM etiology.

Inferring gene regulatory networks from single-cell multiome data using atlas-scale external data

Article Open access 12 April 2024

Genome-wide association studies

Article 26 August 2021

Genome-wide association analyses identify 95 risk loci and provide insights into the neurobiology of post-traumatic stress disorder

Article 18 April 2024

Introduction

The critical role of left ventricular (LV) structure and function in adversely affecting the prognosis of individuals is increasingly acknowledged. Changes in LV structure and function have been associated with acute myocardial infarction¹, heart failure² and mortality³. Likewise, the LV wall thickness is an independent predictor of morbidity and mortality of hypertrophic cardiomyopathy (HCM)⁴, diastolic dysfunction⁵, and ventricular arrhythmias⁶. Various LV regional wall thicknesses (LVRWT) occurred with different clinical implications^7,8, and its quantitative analysis is of great significance for the diagnosis of cardiovascular disease (CVD)⁹. Hence, LVRWT serving as intermediate phenotypes for CVD is needed for timely detection.

Cardiovascular magnetic resonance (CMR) is one of the most popular medical imaging modalities to quantify the LV phenotypes due to its noninvasive and versatile nature¹⁰. However, the highly variable cardiac structures across different subjects and quantification tasks of small targets pose the biggest challenge to accurate estimations of cardiac wall thicknesses in large investigations¹¹. As a result, such a large-scale genome-wide association (GWAS) study of the LVRWT imaging phenotype has not been performed to date. Previous research on the genetic architecture of CMR-derived traits concentrated primarily on LV volumetric and functional metrics such as stroke volume and ejection fraction^2,12. To solve this issue, we established a novel deep-learning framework named Myocardial Segmentation and Measurement Method (MSMM) to precisely estimate LVRWT. The UK Biobank (UKB), one of the largest population imaging studies, has acquired both high-quality standardized CMR examinations and genotype data¹³, offering a tremendous opportunity to investigate the unknown genetic determinants of LVRWT traits.

In this study, we aimed to investigate the genetic basis of 12 LVRWTs: anterior, anteroseptal, inferoseptal, inferior, inferolateral, and anterolateral LVRWT at end diastole and end systole. Using a deep learning-based method to segment LV structure and design an automatic measurement algorithm to quantify myocardial wall thickness, we performed GWASs on 12 LVRWT traits (all derived from the mid-cavity slice with half obtained during end-systole (ES) phase and half during end-diastole (ED) phase) in 42,194 UKB participants. We then evaluated the putative causal relationships between the LVRWT traits and CVDs using genetic correlation and Mendelian randomization (MR) analyses. Finally, we established the polygenic risk score (PRS) of each LVRWT trait, and validated whether the LVRWT-PRSs have the ability to identify HCM in 439,981 individuals without CMR imaging. Overall, this study substantially enhances our knowledge of the genetic basis of LVRWT and may provide valuable risk stratification guidance to identify high-risk sub-populations in individualized HCM prevention.

Results

Segmentation of left ventricle with deep learning

The MSMM procedure we established to measure the thicknesses were presented in Fig. 1. To achieve state-of-the-art performance in segmentation tasks, we trained the deep learning model using the Automated Cardiac Diagnosis Challenge (ACDC) dataset^14,15 annotated by one clinical expert, which has a total of 1902 annotated images. We randomly selected 1420 short axis images for training, 100 short axis images for validation and 382 short axis images for testing. Within the 382 testing images, 80 images belong to the two mid-cavity slices. Since only the mid-cavity slice quantification results were used for the subsequent analysis, we evaluated the methods on two mid-cavity slices of each case to ensure a consistent comparison. Moreover, we included 500 mid-cavity images from the UKB dataset for independent evaluation because that GWASs are performed on the UKB dataset. The Deep Layer Aggregation based deep learning model^16,17 as the network architecture (DLANet) was trained from 1420 ACDC images. The Deep Layer Aggregation (DLA)-based deep learning model as the network architecture was trained from these data. The trained deep learning model was then used to output LV segmentation results in the UKB imaging substudy of over 45,000 people¹⁸. Detailed information about model training, as well as quantitative and qualitative results, can be found in the “Methods“ section and Supplementary Methods.

**Fig. 1: Left ventricular wall thicknesses measurement with deep learning.**

Measurement of high-quality left ventricular regional wall thicknesses

We applied a measurement-based quantification method to postprocess the deep learning output to measure the LVRWT traits. A total of 45,353 participants with CMR imaging were identified for analyses in the UKB. For these individuals, 6 LVRWT phenotypes at end systole [end-systolic anterior (ES-A), end-systolic anteroseptal (ES-AS), end-systolic inferoseptal (ES-IS), end-systolic inferior (ES-I), end-systolic inferolateral (ES-IL), end-systolic anterolateral (ES-AL) LVRWT] and 6 LVRWT phenotypes at end diastole [end-diastolic anterior (ED-A), end-diastolic anteroseptal (ED-AS), end-diastolic inferoseptal (ED-IS), end-diastolic inferior (ED-I), end-diastolic inferolateral (ED-IL), end-diastolic anterolateral (ED-AL) LVRWT] were available. Following the exclusion of poor image quality and sample quality-control procedures, 42,194 individuals free from a diagnosis of myocardial infarction or heart failure remained (Supplementary Fig. 1). The average age of the cohort at the time of imaging visit was 64.1 years, and 47.3% were men at end systole and 47.2% at end diastole (Supplementary Table 1). The mean and standard deviation of 12 LVRWT phenotypes are reported in Supplementary Table 2. We next sought to estimate the phenotypic correlations of LVRWT traits. The strong positive phenotypic correlations were observed between regions of the same cardiac cycle phases (r² = 0.82 between ES-IL and ES-AL), and the end-diastolic LVRWT phenotypes and end-systolic LVRWT phenotypes were modestly positively correlated with one another (Supplementary Fig. 2a), which is not unexpected, given their firm physiological associations. Furthermore, to gain insight into the dimensionality of the data, we conducted a principal component analysis (PCA) and identified that utilizing five principal components explained over 90% of the variability in the LVRWT measurements (Supplementary Fig. 2b).

Genome-wide association studies of left ventricular regional wall thicknesses

To understand the common genetic basis for variation in myocardial wall thickness, we performed a series of GWAS on 12 LVRWT traits within each cohort. Firstly, due to the nonnormal distribution, we performed the rank-based inverse normal transformation of the residuals of LVRWT phenotypes (Supplementary Fig. 3). There was no evidence of confounding from population stratification or cryptic relatedness in our GWAS analyses, as demonstrated by low genomic inflation factor (λ = 1.050–1.099, Supplementary Fig. 4). Furthermore, we identified a total of 72 genome-wide significant variants associated with the 12 LVRWT traits (P < 5 × 10⁻⁸), including 16 variants for IS, 2 variants for I, 7 variants for IL, 10 variants for AL, 3 variants for A and 7 variants for AS at end systole stage; 5 variants for IS, 3 variants for I, 3 variants for IL, 4 variants for AL, 3 variants for A and 9 variants for AS at end diastole stage, respectively (Fig. 2 and Supplementary Data 1). Interestingly, there have 10 genetic variants associated with multiple LVRWT traits, resulting in 62 unique variants associated with the 12 LVRWT traits; for example, rs7928419 in SPI1 at 11p11.2 reached genome-wide significance in ES-IS, ED-IS and ED-AS (P_ES-IS = 1.0 × 10⁻¹², P_ED-IS = 1.3 × 10⁻⁹, P_ED-AS = 1.2 × 10⁻⁸, Supplementary Fig. 5). Furthermore, following the five PCs correction for multiple tests, we observed that 47 variants remained significant (P < 1.0 × 10⁻⁸, Supplementary Data 1).

**Fig. 2: Manhattan plots of genome-wide association studies results for 12 LVRWT phenotypes.**

Functional characterization for risk variants of 12 LVRWT traits

The LVRWT GWAS loci harbored a total of 6345 candidate variants in linkage disequilibrium (LD, r² > 0.6) with 62 lead variants. To characterize the features for risk variants of 12 LVRWT traits, we first generated the control variants set and functionally defined the genomic position distribution of variants via SnpEff¹⁹. Compared with control variants, we found risk variants were significant enriched within intron region, gene upstream and downstream regions, and intergenic region (Fig. 3a, b). We next investigated whether risk variants are enriched among genetic regulatory elements. We observed significant enrichments for risk variants of 12 LVRWT traits within the histone marks of promoters [H3K4 trimethylation markers (H3K4me3)], enhancers [H3K4 monomethylation marks (H3K4me1), H3K27 acetylation marks (H3K27ac) and H3K36 trimethylation markers (H3K36me3)], and transcriptional factor binding sites (TFBSs), while significant depletion among repressive transcription [H3K9 trimethylation markers (H3K9me3)] (Fig. 3c). Furthermore, to examine the characteristics for risk variants of 12 LVRWT traits in the context of their potential contribution to CVDs susceptibility, we conducted enrichment analyses using the data of GWAS summary statistics from 11 selected CVDs traits (Supplementary Data 2). As a result, risk variants of 12 LVRWTs were observed significantly enriched among CVDs GWASs loci (Fig. 3d). To sum up, these results indicated that risk variants may regulate the expression of genes by activating chromatin state, and might result in an increased CVDs incidence.

Identification and functional annotation of susceptible genes associated with left ventricular regional wall thicknesses

We further sought to identify candidate genes influencing LVRWT phenotypic variation using an integrative approach supported by multiple lines of evidence (Supplementary Fig. 6). Based on the downstream analysis of the discovery GWAS summary statistics, for end diastole and end systole, respectively, 31 genes and 21 genes were identified by position (±1 Mb of the lead variant) (Supplementary Data 1); 58 genes and 50 genes were discovered by expression quantitative trait locus (eQTL) mapping (Supplementary Data 3); and 45 genes and 25 genes were identified by transcriptome-wide analysis (TWAS) (Supplementary Data 4–9). Besides, Multi-marker Analysis of GenoMic Annotation (MAGMA) gene-based analyses yielded 55 and 44 significant genes for end diastole and end systole, respectively, (mean χ² statistics, P < 2.64 × 10⁻⁶) (Supplementary Data 10). We combined all of genes annotated using the four methods. As several loci were shared by multiple traits; counting each locus only once, we totally identified 127 candidate genes at end systole and 95 candidate genes at end diastole, respectively (Supplementary Data 11). Notably, ALPK3 was annotated by the four methods, and rs3803405 in proximity to ALPK3 was the most significant variants in this study. MYBPC3, which was identified by the MAGMA analysis for inferoseptal LVRWT at end systole, is one of the two most well-known HCM causal genes²⁰. ALPK3²¹, NMB²², and WNT3²³ loci, which are significant candidate genes that have been linked to inherited CVDs, were shared across most LVRWT traits (Supplementary Fig. 7).

To further dissert the potential function of the candidate genes, we characterized the biological pathways associated with the 127 genes at end systole and 95 genes at end diastole. The GO analyses demonstrated that the candidate genes are remarkably enriched in the biological pathways correlated with the heart development, heart contraction, and cardiac muscle cell development, which are essential for cardiac remodeling (Fig. 4a, b and Supplementary Data 12). The tissue expression analyses were performed separately for 54 specific tissue types, and tissues from the heart atrial appendage and heart left ventricle showed the most significant P value (P = 2.64 × 10⁻⁶) in most LVRWT traits (Fig. 4c, d and Supplementary Data 13). In summary, these findings significantly advance our understanding of the genetic basis of LVRWT phenotypes and suggest that candidate genes encode essential proteins involved in the construction and maintenance of sarcomeric infrastructure.

**Fig. 4: Pathway enrichment and tissue enrichment for susceptible genes associated with LVRWT.**

Heritability and genetic correlation of left ventricular regional wall thicknesses

We next estimated the proportion of LVRWT phenotypic variation attributable to the genotypes by the variance component analyses. The highest genome-wide variant heritability estimates were observed for inferoseptal LVRWT at end systole (ED-IS, at 28%), followed by inferoseptal LVRWT at end diastole (ED-IS, at 25%), while heritabilities were lower for anterior and anterolateral LVRWT at end diastole, which had a heritability of 17% (Fig. 5a). We observed significantly genetic correlations between LVRWT traits ranging from high (r_g = 0.97 between ES-IL and ES-A) to low (r_g = 0.42 between ES-A and ED-AS; Fig. 5a). Strong positive genetic correlations were generally found between regions of the same cardiac cycle phases and moderate correlations between regions of the different cardiac cycle phases, which may reflect genetic effects acting on the development of the cardiac wall.

**Fig. 5: SNP heritability and genetic correlations.**

Due to the interdependent nature of LVRWT and LV chambers and the importance of wall thickness remodeling, we first sought to quantify the strength of their genetic correlations by linkage disequilibrium score regression (LDSC) analyses using our LVRWT summary statistics and the summary data from a recently published LV GWAS¹². The LVRWT traits had highly positive genetic correlations with LV volumetric and functional traits such as LV ejection fraction (LVEF) and stroke volume (SV) (r_g range: 0.16–0.60), and negative genetic correlations with LVRWT traits were observed between LV end-diastolic volume (LVEDV), LV end-systolic volume (LVESV) and the body-surface-area indexed versions of these traits, including LVEDVi and LVESVi (r_g range: −0.58 to −0.13), which emphasized their strong physiological connections (Fig. 5b and Supplementary Data 14). Additionally, we also investigated the genetic correlations between the 12 LVRWT traits and cardiac measurements that could mediate the CVDs progression. The LVRWT traits were highly positively connected with traits such as heart rate, diastolic and systolic blood pressure (r_g range: 0.11–0.45, Fig. 5b and Supplementary Data 14). Collectively, these findings support the justification for investigating the genetics of LVRWT traits as a complementary gateway to understanding the drivers of cardiac remodeling.

Genetic correlation analyses between left ventricular regional wall thicknesses with cardiovascular diseases

To examine shared genetic effects between LVRWT traits and CVDs, we performed genetic correlation analyses with GWAS summary statistics from 11 selected CVDs traits. As expected, we observed statistically positive genetic correlations of 12 LVRWT traits with CVDs such as HCM, hypertension, pulmonary hypertension, chronic ischemic heart disease, and ischemic stroke. Anterior LVRWT at end diastole had the highest genetic correlation estimate with HCM than other LVRWT traits (ED-A: r_g = 0.65, P = 7.0 × 10⁻⁴). For inferoseptal LVRWT, we observed positive genetic correlations with HCM at both end systole and end diastole (ES-IS: r_g = 0.52, P = 1.4 × 10⁻³; ED-IS: r_g = 0.42, P = 6.8 × 10⁻³). Furthermore, HCM as the disease had the highest genetic correlation estimates with LVRWT than other CVDs (r_g range: 0.37–0.65) (Fig. 5c and Supplementary Data 15). In full, these findings of the genetic correlations between LVRWT traits with CVDs provide quantitative support for genetic overlap.

Mendelian randomization of left ventricular regional wall thicknesses exposures and cardiovascular diseases

The genetic correlations between 12 LVRWT traits and CVDs risk led us to the hypothesis that the increased LVRWT traits are causally associated with CVDs. We then tested such potential causality between 12 LVRWT exposures and CVDs outcomes using two-sample Mendelian randomization (MR). Although heterogeneity and the number of variants in the exposure-outcome effects are limitations, the findings of the inverse-variance-weighted method and sensitivity analyses support causal relations between 10 of 12 LVRWT traits and increased HCM risk (beta range: 0.45 to 3.10, P < 0.01) (Fig. 6 and Supplementary Data 16). We also found robust support for the causal effects of inferolateral, anteroseptal, and anterior LVRWT on cardiomyopathy at end systole and end diastole (P < 0.01). However, there is no statistically significant effect on hypertension, angina pectoris, myocardial infarction, chronic ischemic heart disease, and ischemic stroke (P > 0.05). Taken together, these findings suggest that the genetic relationships between LVRWT traits and HCM may partly reflect underlying causal processes.

**Fig. 6: Causal effects between LVRWT phenotypes and CVDs using Mendelian randomization.**

Polygenetic risk scores influence the risk for incident hypertrophic cardiomyopathy

For each of the 12 LVRWT traits, we derived PRSs weighting the genetic dosage by the effect size of independent genetic variants located at autosome (P < 1 × 10⁻⁵) from each LVRWT traits GWAS (Supplementary Data 17). We investigated whether PRSs could discriminate the 12 LVRWT traits. As expected, participants with a higher PRS tended to have thicker LVRWT for all 12 LVRWT traits, especially for inferoseptal LVRWT at end systole and end diastole (Fig. 7a, b and Supplementary Figs. 8a, b–12a, b), suggesting that the PRSs yielded favorable discrimination for the LVRWTs.

**Fig. 7: Distribution of PRS and cumulative incidence of HCM stratified by PRS.**

Having established the causal effects of LVRWT traits on HCM, we then sought to evaluate whether a genetic predisposition to LVRWT traits is associated with incident HCM in the remaining 439,981 individuals without CMR imaging data. We found that the PRS tertiles of inferoseptal LVRWT at end systole were associated with higher risk of HCM (hazard ratio (HR)_ES-IS = 1.69, 95% CI = 1.33–2.15, P = 2.31 × 10⁻⁵ for high tertile vs. low tertile, Supplementary Data 18). As expected, the PRS of inferoseptal LVRWT at end systole exhibited effective risk stratification within the HCM population (Fig. 7c). Moreover, we found that the PRS of inferoseptal LVRWT at end systole enabled a clear distinction between incident HCM cases and healthy controls (Fig. 7e). In addition, we found the similar results in inferolateral, anterolateral, anterior and anteroseptal LVRWT at end systole, and inferior, inferolateral and anterior LVRWT at end diastole (Supplementary Figs. 8d, f, 9c–f, 10c, e, 11c–f and 12c, e). Regrettably, the other 4 LVRWT traits (inferior LVRWT at end systole, and inferoseptal, anterolateral, and anteroseptal LVRWT at end diastole) did not exhibit strong performance in identifying individuals at elevated risk (Fig. 7d, f and Supplementary Figs. 8c, e, 10d, f and 12d, f). We also assessed the correlation between the constructed PRS for LVRWT traits and 10 other CVDs (Supplementary Data 18). We calculated the C-statistic for our HCM PRSs and noticed that it is not sufficiently satisfactory. However, a slight improvement is noted after further integrating the relevant PRS with clinical risk factors, as depicted in Supplementary Fig. 13. Despite the modest C-statistic values, this enhancement underscores the efficacy of our PRS in identifying individuals at an elevated risk for HCM. In total, higher genetically determined LVRWT was associated with higher HCM risk, which may provide valuable risk stratification guidance to identify high-risk individuals.

Discussion

To our knowledge, this study is the first largest individual-level GWAS to investigate the genetic architecture of LVRWT traits. We established a novel deep learning algorithm to assess 12 LVRWTs accurately using CMR imaging in a large population-based biobank. We identified 72 significant genetic loci associated with at least one LVRWT phenotype, and candidate genes were actively participating in heart development and heart contraction pathways. Furthermore, we evaluated the causal relationships between the LVRWT traits and CVDs with MR analyses and found supporting evidence of LVRWT being causal for increased HCM. We demonstrated that PRSs derived from LVRWT traits are associated with incident HCM. This suggests that the genetic basis of LVRWT traits may offer valuable information for the risk stratification of individuals, aiding in the early screening of HCM (Supplementary Fig. 14).

Due to various subjects and cardiac disorders having varied forms and structures, and the LV myocardium experiencing complicated regional deformation during the systole and diastole phases of the cardiac cycle¹¹, it is challenging to obtain an accurate estimation of LVRWT. Therefore, few studies were carried out on the quantification of LVRWT, and little was known about the heritability and genetic basis of LVRWT. Deep learning is a powerful tool for deriving quantitative phenotypes from raw signal data at a population level. In this paper, by using a deep learning-based method to segment LV structures, we were able to design an automatic measurement algorithm to quantify myocardial wall thicknesses accurately. Reassuringly, the mean absolute errors of LVRWT of our solutions are at most 1.04 mm, suggesting that the MSMM has the highest accuracy of LVRWT. Compared to the maximum wall thickness or mean wall thickness^24,25, we finely divided the wall thickness into 12 LVRWTs, permitted characterization of the genetic architecture of each region and provided new biological insights. According to heritability estimates ranging from 17% to 28%, our research revealed that a sizable portion of the LVRWT phenotypic variability is explained by the underlying genetics. Furthermore, the modest genetic correlations and limited locus overlap of the 12 LVRWT traits highlighted their distinct biology.

A total of 6345 candidate variants are identified in 62 unique loci. Characterizing the genomic characteristics for these risk variants of 12 LVRWT traits allowed us to identify their enrichment in intron region, gene upstream and downstream regions, and intergenic region, which sheds light on the regulatory functions of risk variants²⁶. This notion is also supported by the fact that candidate variants are enriched in TFBSs and histone modification markers like H3K4me3, H3K4me1, H3K27ac, and H3K36me3, which suggests a connection between chromatin status and regulatory properties of risk variants of LVRWT traits. Importantly, we further confirmed a significant enrichment of these risk variants in the CVDs GWAS loci. This finding strengthens the growing evidence that the majority of GWAS risk variants influence disease risk via their regulatory activities.

Extensive multilayered bioinformatic annotations identified candidate genes that have an important role in cardiac muscle cell development and heart contraction. In light of recent Khurshid’s study, it conducted a GWAS of CMR-derived left ventricular mass indexed (LVMI) in the UKB. By comparing the loci of the 13 genes from Khurshid’s study, we found an overlap with 5 genes also identified in our research²⁷. This overlap underscores the consistency of genetic associations related to cardiac traits and cardiomyopathy between our study and Khurshid’s study. In addition, many of candidate genes identified in this study have previously been linked to CVDs. Specifically, many of candidate genes overlapped with previously reported loci for the cardiac volumetric and functional phenotypes, such as ALPK3²⁸, NMB²⁹, CASQ2³⁰, CDKN1A³¹, ACTN2³² and GATA4³³. For example, the alpha kinase 3 (ALPK3) encodes a kinase previously implicated as the causal gene in HCM²¹. ALPK3 regulates the expression of transcription factors (like HEY and HAND proteins) to induce differentiation and maturation of cardiomyocytes at the beginning of the process³⁴. Mutations in ALPK3 cause familial cardiomyopathy and demonstrate loss of function as the underlying genetic mechanism³⁵. Additionally, HCM is a disease of sarcomere proteins, which are composed of thick and thin filaments and Z discs²⁰. In agreement with this notion, we also discovered loci like SYNPO2L, GBAP1, MYOZ1, CAPN9, DMTN, and MTSS that had not been identified in earlier research on cardiac volumetric and functional traits. Interestingly, these genes were revealed to actively participate in Z disc, myofilament, sarcomere, and actin pathways. As an illustration, the scaffolding and actin/myosin regulatory protein known as synaptopodin 2 Like (SYNPO2L), which is expressed in the cardiac muscle and localizes at the Z-disc and interacts with a number of other actin proteins, has been linked to cardiac arrhythmia³⁶. Altogether, several candidate genes are involved in heart development and contraction, and Z disc pathway activation, suggesting that candidate genes may contribute to the development of cardiomyopathies.

As the strong interdependence and genetic correlations between LVRWT and LV volumetric and functional traits, a significant fraction of the LVRWT loci were linked to previously identified loci for LV volumetric and functional imaging traits. This finding supports the justification for investigating the genetics of LVRWT traits as a complementary gateway to understanding the drivers of cardiac remodeling³⁷. Previous studies have indicated that LVRWT traits are independent predictors of a number of clinical outcomes such as stroke³⁸, ventricular arrhythmia³⁹, and hypertension⁴⁰. Although human measurement of ventricular wall thickness is prone to variability, a multi-center longitudinal cohort study has reported that LVRWT is a key imaging biomarker in HCM, guiding diagnosis, risk stratification, and clinical management²⁴. In consistent with these findings, we also systematically examined the genetic relations between LVRWT traits and CVDs obtained from published GWASs. We observed genetic associations between LVRWT traits and CVDs, such as HCM, hypertension, pulmonary hypertension, chronic ischemic heart disease, and ischemic stroke, which further reinforced the importance of LVRWT in maintaining efficient circulatory physiology. The results of the MR analysis also showed that LVRWT traits are potentially causally linked with CVDs. Notably, LVRWT traits have the strongest causative effects on HCM than other CVDs. Specifically, these findings point to the notion that HCM may represent the extreme of LVRWT phenotypic variations in certain people, which is helpful to demonstrate that LV wall thickening is the most consistent clinical marker of HCM⁴¹.

GWAS has identified genetic variations that contribute to complex traits and diseases, leading to the development of PRS. The effectiveness of the PRS is often evaluated by determining whether it can help screen the high-risk population to guide clinical or personal decision-making⁴². In agreement with this development, our analyses of PRSs shed light on the linkage between genetically determined LVRWT phenotypic variations and the incidence of HCM. Interestingly, the top-performing PRS, which included 115 variants linked to inferoseptal LVRWT at end systole, successfully identifies individuals at high risk of HCM, especially in those without CMR data. In accordance with recommendations, inferoseptal hypertrophy with a thickness greater than 15 mm was a major component of the diagnostic criteria for HCM^43,44. The PRSs might identify high-risk, asymptomatic individuals who would benefit from CMR to screen for HCM. Likewise, previous studies also have reported that the PRSs of LV structure and function traits predict heart failure events² and incident dilated cardiomyopathy¹². Collectively, PRS may provide complementary information within guideline-supported frameworks to better stratify different trajectories of HCM risk and inform clinical decision-making for primary prevention and early screening.

We acknowledge some limitations in our study. First, our study population largely consisted of European ancestry, limiting generalizability to other populations, and validation of our findings in more diverse cohorts is needed to assess their applicability to non-European populations. Second, further validation of PRS performance is required in independent cohorts. While PRSs derived from LVRWT traits showed associations with an increased risk of HCM, it’s important to recognize that the predictive performance of these PRSs were relatively limited. Our future research will focus on identifying additional genetic loci and molecular markers, with the goal of enhancing the predictive capabilities of our models. Furthermore, while our study provides strong statistical support for the loci that are highly specific for LVRWT phenotypic variation, future experimental studies using gene-editing techniques in cellular and animal models are warranted to elucidate the functional roles of the highlighted risk genes and the underlying mechanisms modulating LVRWT. Lastly, although our findings provide valuable insights into the genetic basis of LVRWT, the complexities of genetic regulation and functional mechanisms within specific cardiac cell types might require more advanced methodologies, such as emerging single-cell genomics techniques, to fully unravel the intricate genetic networks underlying LVRWT and related cardiac remodeling.

In conclusion, using a novel deep learning algorithm to quantify LVRWT measured by CMR, we uncover common variants at 72 significant genetic loci associated with LVRWT, and implicate candidate genes linked to HCM. Moreover, we demonstrate that PRSs derived from LVRWT traits are associated with incident HCM, even in individuals without CMR data. Altogether, these findings represent a substantial advance in our understanding of the genetic architecture of LVRWT phenotypes and shed light on the biological basis for HCM etiology, which may lead to potential novel therapeutic targets and personalized risk stratification strategies in the future.

Methods

Study population

The UKB is a large prospective cohort study of over 500,000 participants recruited at 22 assessment centers across the UK between 2006 and 2010. It has gathered a wealth of information on participants, including health and lifestyle data, physical measurements, biological samples, imputed genome-wide genotypes and imaging data. All participants provided informed consent. The ethical committees from the North West Multi-Center Research Ethics approved the study.

Sample selection

Disease information based on the UKB Field ID and date of first in-patient diagnosis is provided in Supplementary Data 19. In total, we analyzed 45,353 participants with CMR who had not withdrawn consent as of December 2020. After excluding participants with poor imaging, missing information on genetic data, previous myocardial infarction, diagnosis of heart failure as well as body mass index (BMI) < 16 or > 40 kg/m², 42,194 individuals were included in the analysis following the quality-control procedures outlined in Supplementary Fig. 1. Among the initial 456,937 individuals without CMR data, we excluded individuals without genotype data, resulting in a final sample size of 442,889 individuals. After accounting for missing information on smoking and alcohol consumption, the sample size was further reduced to 440,085 individuals. Additionally, in subsequent analyses focusing on specific diseases, we further excluded individuals who had the corresponding disease prior to their enrollment in the UKB. For instance, in the case of HCM, we excluded 104 individuals who had HCM before their enrollment, resulting in a final sample size of 439,981 individuals.

Definitions of covariates and phenotypes

All covariates recorded at the imaging visit were used in the analysis where possible. The UKB Data-Fields of covariates were listed in Supplementary Data 20. BMI was calculated as: weight(kg)/height(m)². The definitions were used for GWAS participant exclusion and PRS assessment.

Semantic segmentation and deep learning model training

CMR protocol had previously been described in detail¹³. In brief, CMR was performed with 1.5 Tesla scanners (MAGNETOM Aera, Syngo Platform VD13A, Siemens Healthcare, Erlangen, Germany) using electrocardiographic gating for cardiac synchronization. Long-axis cines and a complete short-axis stack of balanced steady-state free precession cines were acquired at one slice per breath hold.

Semantic segmentation aims to assign a label to each pixel in an image. In the proposed MSMM framework, the deep learning model for segmentation learns to extract MR image features and output one of three different classes for each pixel, including background, LV cavity, and LV myocardium. Labels of LV structures in the ACDC (https://www.creatis.insa-lyon.fr/Challenge/acdc/index.html) dataset are annotated by one clinical expert. There are 1420 images for training process, 100 images for validation and 382 images for testing process. Additionally, we resorted to a doctor who has 7 years of clinical experience to label the LV cavity and LV myocardium on the UKB dataset for a comprehensive evaluation. During the testing stage, 80 mid-cavity slices from the ACDC dataset and 500 mid-cavity slices from the UKB dataset were employed for evaluating the performance of the segmentation model.

We utilized the DLANet^16,17 architecture as a segmentation model, which uses two types of structures: the Hierarchical Deep Aggregation (HDA) and the Iterative Deep Aggregation (IDA). With the HDA and IDA, the DLA architecture can better merge information from multiple layers and scales, and achieve state-of-the-art performance in the segmentation task. DLANet with 166 hidden layers and 20,576,932 parameters was implemented in Pytorch. Training and inference processes were performed on NVIDIA GeForce GTX 1080 Ti GPUs. During the training process, the DLANet was trained from scratch for 1000 epochs. We employed the Adam⁴⁵ optimizer for minimizing the Dice Loss⁴⁶. The initial learning rate was set to 1 × 10⁻³ and decayed by 0.99 every epoch. The weight decay was 1 × 10⁻⁵. At the inference stage, one frame of a patient with all slices in the size of (8–12)*256*256 will take about 0.5 s on one GPU. Additionally, we compared the performance of the DLANet with state-of-the-art methods by re-implementing them and reporting the results in Supplementary Data 21 and 22. The comparative analysis across different datasets provides valuable insights into the efficacy of our proposed deep learning model.

Before the training process, we first performed data preprocessing operations. The in-plane resolution of the original cardiac MR images was resampled to 1.25 mm × 1.25 mm, and then the resampled images were cropped or padded to a size of 256 × 256. During training, data augmentations were employed including random shifting, rotation, scaling and so on. For images at the base or apex of the LV, we over-sampled them more than one times in an epoch to increase their frequency during the training data and improve the performance on these images for their difficulty. We ran validation every 20 epochs of training, and when the training process was complete, we selected the DLANet that performs best on the validation set for testing. Finally, the trained model output LV segmentation results on the UKB dataset and the results were used for subsequent quantification of the myocardium.

For quantification, we designed a measurement-based method to calculate wall thicknesses. Before the measurement, we normalize the MR images⁴⁷ by rotation to the fixed arrangement in order to perform subsequent measurements. The thicknesses rely on the calculation of the distance between the center and the contour of the mask, either epicardium or endocardium, for some direction. To this end, we first converted the Cartesian coordinate system into a polar coordinate system and then generated the distance based on the direction angle. Finally, we averaged ten measurements for wall thicknesses. We then used MSMM to calculate the myocardial thicknesses in the UKB. We obtained accurate estimations of 12 LVRWT traits which included 6 LVRWT phenotypes at end systole (ES-IS, ES-I, ES-IL, ES-AL, ES-A, and ES-AS) and 6 LVRWT phenotypes at end diastole (ED-IS, ED-I, ED-IL, ED-AL, ED-A, and ED-AS). Additionally, we assessed the DLANet’s performance across various training dataset sizes, as detailed in Supplementary Data 23 and 24, demonstrating the model’s robustness even with limited training data. The techniques used to postprocess the deep learning output to measure wall thickness is described in the Supplementary Methods.

Genotyping and imputation

Detailed information on genotyping and imputation in the UKB has been described previously⁴⁸. Briefly, participants were genotyped based on UK BiLEVE Axiom™ Array by Affymetrix (807,411 markers for 49,950 participants) and UKB Axiom Array by Affymetrix (825,927 markers for 438,427 participants). Genotype imputation was based on merged UK10K sequencing and 1000 Genomes phase3 reference panels with SHAPEIT3 and IMPUTE3⁴⁹. Variant positions were keyed to the GRCh37 human genome reference.

GWAS and heritability

We calculated the residuals by regressing LVRWT phenotypes on the covariates age, sex, BMI, and imaging center. We performed the GWAS of each normalized LVRWT trait using a linear mixed-model method using ~8.5 million well-imputed variants with minor allele frequency (MAF) ≥ 1% and imputation quality (INFO) score >0.3 by BOLT-LMM. We next estimated the heritability explained by the genotyped variants (h²_g variant) using BOLT-RELM. Both GWAS and heritability analysis models were adjusted for age, sex, BMI, and principal component (PC) 1–10.

Genetic correlation

Using summary statistics, we applied LDSC software⁵⁰ to estimate the genetic correlations (1) between 12 LVRWT traits; (2) between 12 LVRWT traits and 10 cardiac structure and function traits¹²: heart rate, diastolic blood pressure, systolic blood pressure, LV end-diastolic volume (LVEDV), LV end-systolic volume (LVESV), stroke volume (SV), the body-surface-area (BSA) indexed versions of these traits (LVEDVi, LVESVi, and SVi), and LV ejection fraction (LVEF); and (3) between 12 LVRWT traits and 11 CVDs: hypertension, pulmonary hypertension, dilated cardiomyopathy, HCM, cardiomyopathy, HEART FAILURE, atrial fibrillation, myocardial infarction, angina pectoris, chronic ischemic heart disease, ischemic stroke. Data sources of summary statistics for genetic correlation analyses were listed in Supplementary Data 2.

Given the unavailability of GWAS summary data for hypertrophic cardiomyopathy, pulmonary arterial hypertension, and dilated cardiomyopathies, we performed GWAS analyses utilizing ~8.5 million well-imputed variants, each with a MAF ≥ 1%, and an INFO score exceeding 0.3, using data from the UKB. Each GWAS analysis was adjusted for covariates including age, sex, BMI, smoking status, alcohol intake frequency, and the first ten principal components (PC1-10). The sample sizes for each disease were as follows: hypertrophic cardiomyopathy (552 cases and 2208 controls), pulmonary arterial hypertension (2047 cases and 8301 controls), and dilated cardiomyopathies (1303 cases and 5281 controls).

Functional follow-up with FUMA

The web tool Functional Mapping and Annotation of Genome-Wide Association Studies (FUMA) has previously been described in detail (https://fuma.ctglab.nl/)⁵¹. Genome-wide significant variants were defined by trait associations P < 5 × 10⁻⁸ and r² < 0.1 with correlated variants in 250 kb region using FUMA.

We utilized two main approaches to map genome-wide significant loci to genes via FUMA default settings and specialized datasets, as described follows: (1) positional mapping of variants, whereby variants within a 10kB window from known protein-coding genes in the human reference assembly (GRCh37/hg19) are mapped; (2) eQTL mapping whereby allelic variations at a variant is significantly linked to expression of a gene, where we considered eQTLs within heart atrial appendage and heart left ventricle from GTEx v8.

We also performed a generalized gene-set analysis using MAGMA within FUMA. variants within exonic, intronic, and untranslated regions were chosen for each gene. The 18,888 protein-coding genes were used in MAGMA. The mean of the summary statistic (χ²) of GWAS for the variants in a gene was used to determine the gene-based P value⁵². The Bonferroni method was used to calculate the P value significance threshold, which is 2.64 × 10⁻⁶ when 0.05 is divided by the total number of genes (18,888).

Transcriptome-wide association study

For each of the 12 LVRWT phenotypes, we performed a TWAS to identify the most strongly associated gene at each locus based on imputed cis-regulated gene expression. We used FUSION with eQTL data from GTEx v8. Precomputed transcript expression reference weights for the LV (5886 genes), the atrial appendage (6740 genes), and the artery coronary (3989 genes) were obtained from the FUSION authors’ website (http://gusevlab.org/projects/fusion/). A P < 8.49 × 10⁻⁶ was considered significant in LV tissue, a P < 7.42 × 10⁻⁶ was considered significant in atrial appendage tissue, a P < 1.28 × 10⁻⁵ was considered significant in artery coronary tissue. FUSION was then run with its default settings.

Enrichment and tissue expression analyses

Functional enrichment and pathway characterization of the 127 candidate genes at end systole and 95 candidate genes at end diastole were done in the website (http://kobas.cbi.pku.edu.cn/) to obtain Gene Ontology (GO) terms. Tissue expression analyses were obtained from GTEx which were also integrated in FUMA. Average gene expression per tissue type was utilized as a gene covariate to test for a positive link between gene expression in a given tissue type and genetic correlations.

Mendelian randomization

We used MR to estimate causal effects between the 12 LVRWT exposures and 11 CVDs. For two sample MR analyses, we used MR-Egger regression, weighted median and mode-based estimations as sensitivity analyses, along with the inverse-variance weighted (IVW) technique as our major model. To determine whether any discernible influence was mediated by outliers, the MR Pleiotropy RESidual Sum and Outlier (MR-PRESSO)⁵³ approach was applied. In brief, we selected variants that were genome-wide significant (P < 5 × 10⁻⁸) for each LVRWT trait, and remove variants in LD using default settings (r² < 0.001, kb = 10) to generate independent variations. A P < 0.05/5 traits = 0.01 was considered statistically significant. All analyses were performed using the R package TwoSampleMR.

Polygenic risk score

We used the C + T (clumping + thresholding) method⁵⁴ to construct the polygenic risk score (PRS) of each LVRWT trait based on the effect sizes derived from the LVRWT GWASs. The PRS was calculated through a weighted model, as shown below:

$${{{{{{{\rm{PRS}}}}}}}}_{j}=\mathop{\sum}\limits_{i=1}{\beta }_{i}{G}_{i,j}$$

where β values (the log of odds ratio) is the summary statistic for the effective allele and G is the number of the effective allele observed. We used variants with genome-wide significant (P < 1 × 10⁻⁵) and clumping window (r² < 0.1, kb = 250) to derive PRS. We repeated this procedure for each of the 12 traits, producing 12 PRSs. We categorized participants into three genetic risk levels: low (lowest tertile), intermediate (second tertile) and high (highest tertile).

Additional and detailed analyses are available in the Supplementary Methods.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

Data from the UKB (www.ukbiobank.ac.uk/register-apply) are available to all researchers upon making an application. This research has been conducted using the UKB Resource under Application 63454. Data sources of publicly available GWAS results were listed in Supplementary Data 2. The GWAS summary statistics generated in this study have been deposited in the Human Genome Research Institute GWAS Catalog under accession codes: GCST90278508 to GCST90278519 for the 12 LVRWTs, GCST90296096 for HCM, GCST90296097 for dilated cardiomyopathy, and GCST90296098 for pulmonary hypertension. Source data are provided with this paper.

Code availability

Data processing scripts used to perform the analyses described herein are available at https://github.com/goodluckncb/LVRWT-script.

References

Kelly, D. et al. Plasma tissue inhibitor of metalloproteinase-1 and matrix metalloproteinase-9: novel indicators of left ventricular remodelling and prognosis after acute myocardial infarction. Eur. Heart J. 29, 2116–2124 (2008).
Article PubMed Google Scholar
Aung, N. et al. Genome-wide analysis of left ventricular image-derived phenotypes identifies fourteen loci associated with cardiac morphogenesis and heart failure development. Circulation 140, 1318–1330 (2019).
Article CAS PubMed PubMed Central Google Scholar
Medvedofsky, D. et al. 2D and 3D echocardiography-derived indices of left ventricular function and shape: relationship with mortality. JACC Cardiovasc. Imaging 11, 1569–1579 (2018).
Article PubMed Google Scholar
Ho, C. Y. et al. The burden of early phenotypes and the influence of wall thickness in hypertrophic cardiomyopathy mutation carriers: findings from the HCMNet study. JAMA Cardiol. 2, 419–428 (2017).
Article PubMed PubMed Central Google Scholar
Selby, D. E., Palmer, B. M., LeWinter, M. M. & Meyer, M. Tachycardia-induced diastolic dysfunction and resting tone in myocardium from patients with a normal ejection fraction. J. Am. Coll. Cardiol. 58, 147–154 (2011).
Article PubMed PubMed Central Google Scholar
Ghali, J. K., Kadakia, S., Cooper, R. S. & Liao, Y. Impact of left ventricular hypertrophy on ventricular arrhythmias in the absence of coronary artery disease. J. Am. Coll. Cardiol. 17, 1277–1282 (1991).
Article CAS PubMed Google Scholar
de Marvao, A. et al. Precursors of hypertensive heart phenotype develop in healthy adults: a high-resolution 3D MRI study. JACC Cardiovasc. Imaging 8, 1260–1269 (2015).
Article PubMed PubMed Central Google Scholar
Cheng, A. et al. Heterogeneity of left ventricular wall thickening mechanisms. Circulation 118, 713–721 (2008).
Article PubMed PubMed Central Google Scholar
Park, K. et al. Prognostic effect of increased left ventricular wall thickness in severe aortic stenosis. Cardiovasc. Ultrasound 19, 5 (2021).
Article PubMed PubMed Central Google Scholar
Alfakih, K., Reid, S., Jones, T. & Sivananthan, M. Assessment of ventricular function and mass by cardiac magnetic resonance imaging. Eur. Radiol. 14, 1813–1822 (2004).
Article PubMed Google Scholar
Xue, W. et al. Direct estimation of regional wall thicknesses via residual recurrent neural network. in Information Processing in Medical Imaging (eds Niethammer, M. et al.) 505–516 (Springer International Publishing, 2017).
Pirruccello, J. P. et al. Analysis of cardiac magnetic resonance imaging in 36,000 individuals yields genetic insights into dilated cardiomyopathy. Nat. Commun. 11, 2254 (2020).
Article ADS CAS PubMed PubMed Central Google Scholar
Petersen, S. E. et al. UK Biobank’s cardiovascular magnetic resonance protocol. J. Cardiovasc. Magn. Reson. 18, 8 (2016).
Article PubMed PubMed Central Google Scholar
Bernard, O. et al. Deep learning techniques for automatic MRI cardiac multi-structures segmentation and diagnosis: is the problem solved? IEEE Trans. Med. Imaging 37, 2514–2525 (2018).
Article PubMed Google Scholar
Isensee, F., Jaeger, P. F., Kohl, S. A. A., Petersen, J. & Maier-Hein, K. H. nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation. Nat. Methods 18, 203–211 (2021).
Article CAS PubMed Google Scholar
Yu, F., Wang, D., Shelhamer, E. & Darrell, T. Deep layer aggregation. in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition 2403–2412 (2018).
Liu, D., Yan, Z., Chang, Q., Axel, L. & Metaxas, D. N. Refined deep layer aggregation for multi-disease, multi-view & multi-center cardiac MR segmentation. in International Workshop on Statistical Atlases and Computational Models of the Heart 315–322 (Springer, 2021).
Petersen, S. E. et al. Imaging in population science: cardiovascular magnetic resonance in 100,000 participants of UK Biobank—rationale, challenges and approaches. J. Cardiovasc. Magn. Reson. 15, 46 (2013).
Article PubMed PubMed Central Google Scholar
Cingolani, P. et al. A program for annotating and predicting the effects of single nucleotide polymorphisms, SnpEff: SNPs in the genome of Drosophila melanogaster strain w1118; iso-2; iso-3. Fly 6, 80–92 (2012).
Article CAS PubMed PubMed Central Google Scholar
Marian, A. J. & Braunwald, E. Hypertrophic cardiomyopathy: genetics, pathogenesis, clinical manifestations, diagnosis, and therapy. Circ. Res. 121, 749–770 (2017).
Article CAS PubMed PubMed Central Google Scholar
Li, L., Bainbridge, M. N., Tan, Y., Willerson, J. T. & Marian, A. J. A potential oligogenic etiology of hypertrophic cardiomyopathy: a classic single-gene disorder. Circ. Res. 120, 1084–1090 (2017).
Article CAS PubMed PubMed Central Google Scholar
Shah, S. et al. Four genetic loci influencing electrocardiographic indices of left ventricular hypertrophy. Circ. Cardiovasc. Genet. 4, 626–635 (2011).
Article CAS PubMed PubMed Central Google Scholar
Ren, H. et al. WNT3A rs752107(C > T) polymorphism is associated with an increased risk of essential hypertension and related cardiovascular diseases. Front. Cardiovasc. Med. 8, 675222 (2021).
Article CAS PubMed PubMed Central Google Scholar
O’Mahony, C. et al. A novel clinical risk prediction model for sudden cardiac death in hypertrophic cardiomyopathy (HCM risk-SCD). Eur. Heart J. 35, 2010–2020 (2014).
Article PubMed Google Scholar
Augusto, J. B. et al. Diagnosis and risk stratification in hypertrophic cardiomyopathy using machine learning wall thickness measurement: a comparison with human test-retest performance. Lancet Digit. Health 3, e20–e28 (2021).
Article CAS PubMed Google Scholar
Huan, T. et al. Genome-wide identification of DNA methylation QTLs in whole blood highlights pathways for cardiovascular disease. Nat. Commun. 10, 4267 (2019).
Article ADS PubMed PubMed Central Google Scholar
Khurshid, S. et al. Clinical and genetic associations of deep learning-derived cardiac magnetic resonance-based left ventricular mass. Nat. Commun. 14, 1558 (2023).
Article ADS CAS PubMed PubMed Central Google Scholar
Walsh, R. & Bezzina, C. R. ALPK3: a full spectrum cardiomyopathy gene? Eur. Heart J. 42, 3074–3077 (2021).
Article PubMed Google Scholar
Levin, M. G. et al. Genome-wide association and multi-trait analyses characterize the common genetic architecture of heart failure. Nat. Commun. 13, 6914 (2022).
Article ADS CAS PubMed PubMed Central Google Scholar
Refaat, M. M. et al. Association of CASQ2 polymorphisms with sudden cardiac arrest and heart failure in patients with coronary artery disease. Heart Rhythm 11, 646–652 (2014).
Article PubMed PubMed Central Google Scholar
Gyftopoulos, A. et al. Cyclin dependent kinase inhibitor A1 identified as a potential risk locus for hypertrophic cardiomyopathy. In Circulation Research, Vol. 127, A460–A460 (American Heart Association, 2020).
Prondzynski, M. et al. Disease modeling of a mutation in α‐actinin 2 guides clinical therapy in hypertrophic cardiomyopathy. EMBO Mol. Med. 11, e11115 (2019).
Svensson, E. C., Tufts, R. L., Polk, C. E. & Leiden, J. M. Molecular cloning of FOG-2: a modulator of transcription factor GATA-4 in cardiomyocytes. Proc. Natl Acad. Sci. USA 96, 956–961 (1999).
Article ADS CAS PubMed PubMed Central Google Scholar
Almomani, R. et al. Biallelic truncating mutations in ALPK3 cause severe pediatric cardiomyopathy. J. Am. Coll. Cardiol. 67, 515–525 (2016).
Article CAS PubMed Google Scholar
Phelan, D. G. et al. ALPK3-deficient cardiomyocytes generated from patient-derived induced pluripotent stem cells and mutant human embryonic stem cells display abnormal calcium handling and establish that ALPK3 deficiency underlies familial cardiomyopathy. Eur. Heart J. 37, 2586–2590 (2016).
Article CAS PubMed Google Scholar
Christophersen, I. E. et al. Large-scale analyses of common and rare variants identify 12 new loci associated with atrial fibrillation. Nat. Genet. 49, 946–952 (2017).
Article CAS PubMed PubMed Central Google Scholar
Tezuka, Y. et al. Association of relative wall thickness of left ventricle with incidence of thromboembolism in patients with non-valvular atrial fibrillation: the Fushimi AF Registry. Eur. Heart J. Qual. Care Clin. Outcomes 6, 273–283 (2020).
Article PubMed Google Scholar
Apostolakis, S., Sullivan, R. M., Olshansky, B. & Lip, G. Y. Left ventricular geometry and outcomes in patients with atrial fibrillation: the AFFIRM Trial. Int. J. Cardiol. 170, 303–308 (2014).
Article PubMed Google Scholar
Biton, Y. et al. Relative wall thickness and the risk for ventricular tachyarrhythmias in patients with left ventricular dysfunction. J. Am. Coll. Cardiol. 67, 303–312 (2016).
Article PubMed Google Scholar
Li, L., Shigematsu, Y., Hamada, M. & Hiwada, K. Relative wall thickness is an independent predictor of left ventricular systolic and diastolic dysfunctions in essential hypertension. Hypertens. Res. 24, 493–499 (2001).
Article PubMed Google Scholar
Maron, B. J., Casey, S. A., Hurrell, D. G. & Aeppli, D. M. Relation of left ventricular thickness to age and gender in hypertrophic cardiomyopathy. Am. J. Cardiol. 91, 1195–1198 (2003).
Article PubMed Google Scholar
Torkamani, A., Wineinger, N. E. & Topol, E. J. The personal and clinical utility of polygenic risk scores. Nat. Rev. Genet. 19, 581–590 (2018).
Article CAS PubMed Google Scholar
American College of Cardiology Foundation/American Heart Association Task Force on Practice Guidelines et al. 2011 ACCF/AHA guideline for the diagnosis and treatment of hypertrophic cardiomyopathy: a report of the American College of Cardiology Foundation/American Heart Association Task Force on Practice Guidelines. J. Thorac. Cardiovasc. Surg. 142, e153–e203 (2011).
Article Google Scholar
Authors/Task Force Members et al. 2014 ESC guidelines on diagnosis and management of hypertrophic cardiomyopathy: the Task Force for the Diagnosis and Management of Hypertrophic Cardiomyopathy of the European Society of Cardiology (ESC). Eur. Heart J. 35, 2733–2779 (2014).
Article Google Scholar
Kingma, D. P. & Ba, J. Adam: a method for stochastic optimization. (2014).
Milletari, F., Navab, N. & Ahmadi, S.-A. V-net: fully convolutional neural networks for volumetric medical image segmentation. in 2016 Fourth International Conference on 3D Vision (3DV) 565–571 (IEEE, 2016).
Wang, W. et al. Quantification of full left ventricular metrics via deep regression learning with contour-guidance. IEEE Access 7, 47918–47928 (2019).
Bycroft, C. et al. The UK Biobank resource with deep phenotyping and genomic data. Nature 562, 203–209 (2018).
Article ADS CAS PubMed PubMed Central Google Scholar
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 (2009).
Article PubMed PubMed Central Google Scholar
Bulik-Sullivan, B. K. et al. LD score regression distinguishes confounding from polygenicity in genome-wide association studies. Nat. Genet. 47, 291–295 (2015).
Article CAS PubMed PubMed Central Google Scholar
Watanabe, K., Taskesen, E., van Bochoven, A. & Posthuma, D. Functional mapping and annotation of genetic associations with FUMA. Nat. Commun. 8, 1826 (2017).
Article ADS PubMed PubMed Central Google Scholar
de Leeuw, C. A., Mooij, J. M., Heskes, T. & Posthuma, D. MAGMA: generalized gene-set analysis of GWAS data. PLoS Comput. Biol. 11, e1004219 (2015).
Article PubMed PubMed Central Google Scholar
Verbanck, M., Chen, C. Y., Neale, B. & Do, R. Publisher Correction: Detection of widespread horizontal pleiotropy in causal relationships inferred from Mendelian randomization between complex traits and diseases. Nat. Genet. 50, 1196 (2018).
Article CAS PubMed Google Scholar
Choi, S. W., Mak, T. S. & O’Reilly, P. F. Tutorial: a guide to performing polygenic risk score analyses. Nat. Protoc. 15, 2759–2772 (2020).
Article CAS PubMed PubMed Central Google Scholar

Download references

Acknowledgements

We are grateful to all members who participated in the study, as well as all individuals who helped us complete the research. This work was supported by National Science Fund for Excellent Young Scholars (NSFC-82322058), Program of National Natural Science Foundation of China (NSFC-82103929, NSFC-82273713), Young Elite Scientists Sponsorship Program by CAST (2022QNRC001), National Science Fund for Distinguished Young Scholars of Hubei Province of China (2023AFA046), Fundamental Research Funds for the Central Universities (WHU:2042022kf1205) and Knowledge Innovation Program of Wuhan (whkxjsj011, 2023020201010073) for J.T.; National Science Fund for Distinguished Young Scholars of China (NSFC-81925032), Key Program of National Natural Science Foundation of China (NSFC-82130098), the Leading Talent Program of the Health Commission of Hubei Province, Knowledge Innovation Program of Wuhan (2023020201010060) and Fundamental Research Funds for the Central Universities (2042022rc0026, 2042023kf1005) for X.M.; the Beijing Nova Program (Z201100006820064, Z211100002121165).

Author information

These authors contributed equally: Caibo Ning, Linyun Fan, Meng Jin, Wenji Wang, Zhiqiang Hu, Yimin Cai, Liangkai Chen.

Authors and Affiliations

Department of Epidemiology and Biostatistics, School of Public Health, Wuhan University, Wuhan, 430071, China
Caibo Ning, Linyun Fan, Yimin Cai, Zequn Lu, Ming Zhang, Can Chen, Yanmin Li, Fuwei Zhang, Wenzhuo Wang, Yizhuo Liu, Shuoni Chen, Yuan Jiang, Chunyi He, Zhuo Wang, Xu Chen, Hanting Li, Gaoyuan Li, Qianying Ma, Hui Geng, Wen Tian, Heng Zhang, Bin Li, Ying Zhu, Jianbo Tian & Xiaoping Miao
Department of Radiation Oncology, Renmin Hospital of Wuhan University, Wuhan, 430071, China
Caibo Ning, Linyun Fan, Ying Zhu, Xiangpan Li, Jianbo Tian & Xiaoping Miao
Department of Gastrointestinal Oncology, Zhongnan Hospital of Wuhan University, Wuhan, 430071, China
Caibo Ning, Linyun Fan, Ying Zhu, Jianbo Tian & Xiaoping Miao
Department of Oncology, Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, 430030, China
Meng Jin & Bo Liu
SenseTime Research, Shanghai, 201103, China
Wenji Wang, Zhiqiang Hu, Qing Xia & Shaoting Zhang
Department of Nutrition and Food Hygiene, Hubei Key Laboratory of Food Nutrition and Safety, School of Public Health, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, 430030, China
Liangkai Chen
Department of Gastrointestinal Surgery, Zhongnan Hospital of Wuhan University, Wuhan University, Wuhan, 430071, China
Xiaojun Yang
Department of Psychiatry, Renmin Hospital of Wuhan University, Wuhan, 430071, China
Zhongchun Liu
TaiKang Center for Life and Medical Sciences, Wuhan University, Wuhan, 430071, China
Ying Zhu, Jianbo Tian & Xiaoping Miao
Shanghai Artificial Intelligence Laboratory, Shanghai, 200232, China
Shaoting Zhang

Authors

Caibo Ning
View author publications
You can also search for this author in PubMed Google Scholar
Linyun Fan
View author publications
You can also search for this author in PubMed Google Scholar
Meng Jin
View author publications
You can also search for this author in PubMed Google Scholar
Wenji Wang
View author publications
You can also search for this author in PubMed Google Scholar
Zhiqiang Hu
View author publications
You can also search for this author in PubMed Google Scholar
Yimin Cai
View author publications
You can also search for this author in PubMed Google Scholar
Liangkai Chen
View author publications
You can also search for this author in PubMed Google Scholar
Zequn Lu
View author publications
You can also search for this author in PubMed Google Scholar
Ming Zhang
View author publications
You can also search for this author in PubMed Google Scholar
Can Chen
View author publications
You can also search for this author in PubMed Google Scholar
Yanmin Li
View author publications
You can also search for this author in PubMed Google Scholar
Fuwei Zhang
View author publications
You can also search for this author in PubMed Google Scholar
Wenzhuo Wang
View author publications
You can also search for this author in PubMed Google Scholar
Yizhuo Liu
View author publications
You can also search for this author in PubMed Google Scholar
Shuoni Chen
View author publications
You can also search for this author in PubMed Google Scholar
Yuan Jiang
View author publications
You can also search for this author in PubMed Google Scholar
Chunyi He
View author publications
You can also search for this author in PubMed Google Scholar
Zhuo Wang
View author publications
You can also search for this author in PubMed Google Scholar
Xu Chen
View author publications
You can also search for this author in PubMed Google Scholar
Hanting Li
View author publications
You can also search for this author in PubMed Google Scholar
Gaoyuan Li
View author publications
You can also search for this author in PubMed Google Scholar
Qianying Ma
View author publications
You can also search for this author in PubMed Google Scholar
Hui Geng
View author publications
You can also search for this author in PubMed Google Scholar
Wen Tian
View author publications
You can also search for this author in PubMed Google Scholar
Heng Zhang
View author publications
You can also search for this author in PubMed Google Scholar
Bo Liu
View author publications
You can also search for this author in PubMed Google Scholar
Qing Xia
View author publications
You can also search for this author in PubMed Google Scholar
Xiaojun Yang
View author publications
You can also search for this author in PubMed Google Scholar
Zhongchun Liu
View author publications
You can also search for this author in PubMed Google Scholar
Bin Li
View author publications
You can also search for this author in PubMed Google Scholar
Ying Zhu
View author publications
You can also search for this author in PubMed Google Scholar
Xiangpan Li
View author publications
You can also search for this author in PubMed Google Scholar
Shaoting Zhang
View author publications
You can also search for this author in PubMed Google Scholar
Jianbo Tian
View author publications
You can also search for this author in PubMed Google Scholar
Xiaoping Miao
View author publications
You can also search for this author in PubMed Google Scholar

Contributions

S.Z., J.T. and X.M. were the overall principal investigators in this study who conceived the study and obtained financial support. S.Z., J.T. and X.M. were responsible for the study design and supervised the entire study. C.N., L.F., M.J., W.J.W., Z.H., Y.C. and L.C. organized the data, carried out the statistical analysis and participated in writing the first draft of the manuscript. C.N. designed and drew the figures. L.F. and M.Z. critically revised the manuscript. Z.Q.L., C.C., Ya.L., F.Z., W.Z.W., Yi.L., S.C., Y.J., C.H., Z.W., X.C., H.L., G.L., Q.M., H.G., W.T., H.Z., B.L., Q.X., X.Y., Z.C.L., B.L., Y.Z. and X.L. organized and analyzed the data. These authors jointly supervised this work.

Corresponding authors

Correspondence to Shaoting Zhang, Jianbo Tian or Xiaoping Miao.

Ethics declarations

Competing interests

The authors declare no competing interests.

Peer review

Peer review information

Nature Communications thanks Nikhil Paliwal and the other, anonymous, reviewers for their contribution to the peer review of this work. A peer review file is available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Information

Peer Review File

Description of Additional Supplementary Files

Supplementary Data 1

Supplementary Data 2

Supplementary Data 3

Supplementary Data 4

Supplementary Data 5

Supplementary Data 6

Supplementary Data 7

Supplementary Data 8

Supplementary Data 9

Supplementary Data 10

Supplementary Data 11

Supplementary Data 12

Supplementary Data 13

Supplementary Data 14

Supplementary Data 15

Supplementary Data 16

Supplementary Data 17

Supplementary Data 18

Supplementary Data 19

Supplementary Data 20

Supplementary Data 21

Supplementary Data 22

Supplementary Data 23

Supplementary Data 24

Reporting Summary

Source data

Source Data

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Cite this article

Ning, C., Fan, L., Jin, M. et al. Genome-wide association analysis of left ventricular imaging-derived phenotypes identifies 72 risk loci and yields genetic insights into hypertrophic cardiomyopathy. Nat Commun 14, 7900 (2023). https://doi.org/10.1038/s41467-023-43771-5

Download citation

Received: 30 March 2023
Accepted: 18 November 2023
Published: 30 November 2023
DOI: https://doi.org/10.1038/s41467-023-43771-5

Comments

By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.