Mapping sleep’s phenotypic and genetic links to the brain and heart: a systematic analysis of multimodal brain and cardiac images in the UK Biobank

Zirui Fan, Yujue Li, Juan Shu, Xiaochen Yang, Bingxuan Li, Jinjie Lin, Qianwen Wang, Peristera Paschou, Tengfei Li, Hongtu Zhu, Bingxin Zhao
doi: https://doi.org/10.1101/2022.09.08.22279719

Abstract

Sleep is essential for the health of the brain and heart. Although sleep has been identified as a factor in a few specific clinical outcomes, a systematic analysis of the relationship between sleep and brain/heart and their genetic underpinnings is lacking. Medical images can provide useful clinical endophenotypes for organ structures and functions. Here we present a systematic genetic investigation of sleep-brain/heart connections using multi- modal brain and cardiac images from over 40,000 subjects in the UK Biobank. We identified novel phenotypic and genetic links between sleep and a wide range of imaging traits, such as brain structures, white matter integrity, brain activities, as well as cardiac structures and functions. We prioritized a number of imaging modalities and traits for specific sleep conditions, such as the resting brain function measures in the somatomotor network with narcolepsy. Sleep and imaging had overlapping genetic influences in 39 genomic loci, some of which showed evidence of shared causal genetic variants. In conclusion, large-scale imaging genetic data illuminate the implications of sleep on brain and cardiac health and their genetic links. An interactive web browser (www.ig4sleep.org) has been developed to facilitate exploring our results.

A growing body of evidence suggests that poor sleep is a risk factor for physical, cognitive, and mental health problems1. Sleep traits (such as sleep quality and duration) and disorders (such as insomnia and narcolepsy) are associated with various cardiovascular diseases and brain disorders. For example, rapid eye movement sleep behavior disorder was found to be a strong predictor for the development of neurodegenerative diseases such as dementia and Parkinson’s disease2,3. Sleep dysregulation and short sleep duration in midlife were consistently linked to higher risk of Alzheimer’s disease and other dementias4. In addition, both insufficient and excess sleep duration may cause a higher incidence of cardiovascular outcomes, including coronary heart disease5-7, hypertension8, atrial fibrillation9,10, and stroke11,12. Individuals with sleep disturbance were also more likely to have mental and psychiatric disorders13, such as anxiety14, post-traumatic stress disorder15, schizophrenia16, bipolar disorder, major depressive disorder (MDD), and attention-deficit/hyperactivity disorder (ADHD)17,18.

Magnetic resonance imaging (MRI) provides noninvasive and comprehensive measures of the structure and function of the human organs, including the brain and heart. Imaging traits generated from brain and cardiac MRI are well-established clinical endophenotypes and have been widely used in early prediction and detection of cardiovascular, neurological, and neuropsychiatric outcomes19-22. Using MRI data, several studies have examined the sleep-related structural and functional alterations in the brain and heart23-30. Two major limitations of most existing studies have been i) the limited study sample size, which was usually less than a few hundred; and ii) focusing on one single MRI modality (or trait) and/or one sleep trait, such as hippocampal atrophy25 and sleep duration30-32. It is known, however, that large sample sizes are needed for MRI studies to detect small effect sizes33 and produce reproducible findings, especially for brain functional MRI (fMRI) data34. In addition, distinct MRI modalities and sleep conditions may be relevant to different diseases and health-related complex characteristics21. Therefore, a systematic analysis of multi-modal MRI data and multiple sleep conditions in a large cohort would provide a more comprehensive understanding of sleep-related changes in brain and heart structures and functions.

Genome-wide association studies (GWAS) have shown that sleep disorders and traits are heritable and have a polygenic genetic architecture35-45. The sleep-associated genetic variants are enriched for genes expressed in the brain and for metabolic and psychiatric pathways44. Genetic correlations between sleep traits and brain-related disorders (such as depression and schizophrenia) have been discovered, suggesting their shared neurogenetic basis46. On the other hand, both brain and cardiac MRI traits are also heritable and hundreds of associated genetic loci have been identified in recent GWAS47- 58. However, few studies have ever integrated the multi-organ MRI measures and sleep conditions to explore the genetic interactions among sleep behavior, brain and cardiac structure/function, and related clinical endpoints.

To overcome these challenges, here we examined the phenotypic and genetic sleep- brain/heart connections using multi-modal cardiac and brain MRI data from more than 40,000 subjects in the UK Biobank (UKB) study59. We included three major brain MRI modalities: 1) brain anatomical and neuropathological structures from structural MRI, including regional brain volumes54 and cortical thickness traits60; 2) white matter microstructures from diffusion MRI, including multiple diffusion tensor imaging (DTI) parameters56; and 3) intrinsic and extrinsic functional organizations of the cerebral cortex from resting and task fMRI, including functional connectivity and amplitude traits58. For heart, we used 82 traits extracted from short-axis, long-axis, and aortic cine cardiac MRI61, including global and regional measures of 4 cardiac chambers (the left ventricle, right ventricle, left atrium, and right atrium), as well as 2 aortic sections (the ascending aorta and descending aorta). We mainly examined 7 self-reported sleep-related conditions in the UKB study: sleep duration, getting up in the morning, chronotype (morning/evening person), daytime nap, insomnia (sleeplessness), snoring, and narcolepsy (daytime sleepiness/dozing). Detailed information on these imaging and sleep data can be found in the Methods section. A series of phenotypic, genetic, and predictive analyses were conducted to understand the health implications of sleep. The overview of our study is presented in Figure 1 and our results can be explored via the interactive web browser at www.ig4sleep.org.

Fig. 1.
Fig. 1. Overview of our imaging genetics studies for sleep.

Multimodal brain and cardiac imaging were used to investigate the relationship between sleep and brain and heart health. We covered a full spectrum of imaging modalities, including T1-weighted structural MRI, diffusion MRI, resting-state and task-based fMRI, as well as short-axis, long-axis, and aortic cine images of cardiac MRI. Seven sleep conditions were examined, including sleep duration, getting up in the morning, chronotype, daytime nap, insomnia, snoring, and narcolepsy. Sleep-imaging connections can be explained in part by shared genetic factors. A variety of analyses were conducted throughout the paper in order to uncover the overlapping genetic influences, including polygenic risk scores, identification of shared genetic risk loci and colocalization, genome- wise and local genetic correlations, Mendelian randomizations, and integrated predictions.

RESULTS

An atlas of sleep associations with multi-modal brain and cardiac MRI traits

We examined the phenotypic associations between 7 sleep conditions and a wide range of brain and cardiac MRI traits, including 101 regional brain volumes54, 63 cortical thickness measures60, 110 DTI parameters56, 92 parcellation-based network-level traits in resting and task fMRI58, respectively, as well as 82 heart imaging traits60,61 (Table S1). We performed regression analysis with unrelated white British subjects (average n = 29,025, Methods). Among the 3,780 (7×540) association pairs, 395 associations were significant at the false discovery rate (FDR) level of 5% (by the Benjamini-Hochberg procedure) and 92 of them further passed the more stringent Bonferroni significance level (P < 1.32×10−5, 0.05/3780) (Figs. 2A, S1, and Table S2). Significant associations were observed for all the 7 sleep conditions and 6 groups of imaging traits. Below we highlight the associations survived the Bonferroni multiple testing adjustment.

Fig. 2.
Fig. 2. Phenotypic sleep-imaging associations.

(A) The -log10(p-value) of phenotypic correlations between 7 sleep traits and 6 groups of imaging, including 101 regional brain volumes, 63 cortical thickness measures, 110 DTI parameters, 92 resting fMRI traits, 92 task fMRI traits, and 82 heart imaging traits. See Table S1 for more information of these imaging traits. The red and black dashed lines indicate the Bonferroni significance level (P < 1.33 × 10−5) and the false discovery rate at 5%, respectively. Each sleep condition is labeled with a different color. (B) and (C) Significant correlations between area-level resting functional connectivity traits and narcolepsy (in B) and sleep duration (in C). We show the correlations that passed the Bonferroni significance level (P < 7.74×10−7). Correlation estimates are indicated by the color. Visual1, the primary visual network; and Visual2, the secondary visual network.

Resting fMRI showed the strongest sleep associations among all 6 groups of imaging traits, which were mostly related to sleep duration, narcolepsy, and getting up. Specifically, sleep duration had widespread negative associations with resting fMRI traits (including both functional connectivity and amplitude62 measures) in multiple functional networks (β range = [-0.062, -0.036], P < 1.02×10−5), with the strongest association in the somatomotor network (β = -0.062, P = 3.11×10−13). Somatomotor function was important in multiple sleep stages and poor sleep quality and shorter duration have been linked to increased somatomotor functional connectivity in young samples63-65. In addition, resting fMRI traits positively correlated with easy getting up (β range = [0.036, 0.048], P < 1.28×10−5) and were mostly positively associated with narcolepsy (β range = [-0.051, 0.064], P < 1.27×10−5). Narcolepsy patients have been found to have alerted functional organizations, although previous sample sizes were usually small, making conclusions inconclusive66,67. In our analysis, decreased functional connectivity was mainly found in the default mode network, as well as between the auditory and orbito-affective as well as ventral-multimodal networks68. Functional connectivity in most of other networks had positive associations with narcolepsy (Fig. S2). Compared to resting fMRI, task fMRI had much weaker signals, with only two associations with getting up remaining after Bonferroni correction (β range = [-0.039, 0.042], P < 1.07×10−5).

To uncover detailed spatial association patterns and pinpoint sleep-related brain regions, we further performed area-level resting fMRI analysis with 64,620 (360 × 359/2) functional connectivity measures among 360 brain functional areas in 12 networks68,69. These high-resolution fMRI traits provided fine-grained information on the functional organization of the cerebral cortex. At the Bonferroni significance level (P < 7.74×10−7, 0.05/64,620), sleep duration had negative associations with most functional areas in the somatomotor and auditory networks, as well as some areas of the visual (visual1 and visual2), cingulo-opercular, and language networks (Fig. 2B). In contrast, most of the positive associations with sleep duration were related to the areas of the default mode network. Furthermore, we partitioned the default mode network into seven clusters58, showing that sleep duration was predominantly related to the hippocampal and visual subclusters (Fig. S3A). As compared to long sleep duration, self-reported shorter sleep duration was associated with lower functional connectivity in the default mode network70-72. The default mode network is an active network that integrates information73 and plays an important role in the awake brain74. In addition to previous evidence that shorter sleep duration was related directly to higher levels of Amyloid-β75, our findings provided further evidence that reducing sleep duration had a significant impact on brain functioning. For narcolepsy, positive associations were mainly in the somatomotor network and the connections between the somatomotor and visual networks (Fig. 2C). In contrast, negative associations with narcolepsy were found in the default mode network, most of which were related to the visual subcluster (Fig. S3B). For getting up and daytime nap, most of their associations were related to the visual and somatomotor networks (Figs. S4A-S4B). Overall, these results showed that functional connectivity of the somatomotor and visual networks had substantial connections with sleep conditions. In the default model network, narcolepsy and sleep duration were associated with brain functional activity differently than in other networks.

Our area-level analysis also identified related functional areas for snoring, insomnia, and chronotype. Particularly, snoring was linked to resting functional connections in the frontoparietal network, especially the left and right 13l areas in the posterior orbital gyrus (Fig. S4C). The functional connectivity between the left/right 13L areas and frontoparietal, visual, cingulo-opercular, and dorsal attention networks was increased in snoring, while the functional connectivity with the default mode network decreased. Reduced functional connectivity was also observed within the default mode network (Fig. S3C). Associations with chronotype were mainly detected among cognitive networks, especially between the cingulo-opercular and default mode networks (Figs. S3D and S4D). In addition, insomnia was found to be negatively associated with resting functional connections within the somatomotor network (Fig. S4E). In summary, sleep-related brain functional variations were uncovered in specific brain areas, prioritizing related fMRI biomarkers for future sleep research.

Next, we observed significant associations between DTI parameters and multiple sleep conditions, especially chronotype (Figs. 2A and S5). Being an evening person was associated with higher mean diffusivity (MD) and axial diffusivity (AD) in multiple brain white matter tracts, such as the anterior corona radiata, body of corpus callosum, retrolenticular part of internal capsule, anterior limb of internal capsule, and superior corona radiata (β range = [0.042, 0.048], P < 1.23×10−5). Increased MD and AD may indicate abnormalities and healthy aging in glial tissue, in which the glial cells were involved in the regulation of circadian rhythms76. It has been reported that evening chronotype exhibited increased level of white matter changes77. In addition, we observed significant associations between snoring and the radial diffusiviy (RD) in the corticospinal tract (β = 0.040, P = 1.32×10−6), and between daytime nap and the AD of the body of corpus callosum tract (β = 0.042, P = 1.14×10−5). Compared to resting fMRI and DTI parameters, structural MRI traits (regional brain volumes and cortical thickness) had less phenotypic associations with sleep. Associations survived at the Bonferroni significance level were all related to daytime nap, which was associated with reduced right cerebellum white matter volume (β = -0.048, P = 1.30×10−5) and smaller cortical thickness of the left pars triangularis and right insula (β range = [-0.041, -0.039], P < 4.29×10−5).

Sleep also correlated significantly with cardiac MRI traits of the left ventricle, right ventricle, and right atrium (Figs. 2A and S6). Specifically, snoring was positively linked to both left and right heart structures, including the left/right ventricular end-diastolic volume (LVEDV and RVEDV), right atrium maximum and stroke volumes, and right ventricular end-systolic volume (RVESV) (β range = [-0.047, 0.064], P < 1.20×10−5). Snoring may affect the cardiac function and increase the risk of cardiovascular disease especially in women24,78. In addition, getting up had widespread associations with the LVEDV, left ventricular end-systolic volume (LVESV), left ventricular myocardial mass (LVM), RVEDV, RVESV, and right ventricular stroke volume (RVSV) (β range = [0.045, 0.072], P < 1.70×10− 6). Furthermore, negative associations with RVEDV were found for daytime nap and chronotype (β range = [-0.0613, -0.0610], P < 1.30×10−7). Human health is closely related to sleep chronotypes45, which are behavioral manifestations of circadian rhythms79. In perspective studies, early chronotype is often regarded as a sign of healthy sleep behavior80,81. Individuals with an early chronotype are typically at a reduced risk of developing cardiovascular diseases82 and heart failure81. Circadian and cardiovascular diseases, however, remain elusive in terms of their underlying mechanisms. The present study established a clear physiologic relationship between early chronotype and cardiac health. Overall, we mapped phenotypic links between the brain and cardiac MRI traits and sleep characteristics. These imaging endophenotypes may aid in future clinical research and applications to better understand sleep’s role in brain and heart-related clinical outcomes.

In the phenotypic analysis above, UKB subjects who had both imaging and sleep data were considered, whereas more than 80% UKB subjects had only sleep data and thus were excluded. In order to include these non-imaging UKB subjects in our study, we developed polygenic risk scores (PRS) by using GWAS summary statistics of imaging traits from previous studies54,56,58,60. These genetically predicted imaging traits allowed us to repeat the above sleep-imaging association analysis in non-imaging UKB unrelated white British subjects (average n = 217,254, removing relatives of imaging subjects, Methods). In this PRS analysis, 276 significant associations were identified at the FDR 5% level, covering all the 7 sleep conditions and 6 groups of imaging PRS. Among these 276 associations, 35 further passed the stringent Bonferroni significance level (P < 1.32×10−5) (Table S3 and Figs. S7-S8). The results of phenotypic and PRS analyses were largely consistent, indicating that the sleep-imaging associations discovered by a small proportion of UKB imaging subjects had good generalizability in the whole UKB cohort. For example, similar to the phenotypic analysis, resting fMRI traits also had the largest proportion of significant associations (97/276 at FDR 5% level and 25/35 at the Bonferroni level) in PRS analysis.

Furthermore, considering that imaging PRS were predicted with genetic information, genetic factors may play an important role in influencing sleep-imaging associations. In the following sections, we will examine these underlying genetic influences in more detail.

Characterizing sleep-imaging genetic overlaps in 39 genomic loci

To understand the genetic co-architecture underlying sleep-imaging associations, we characterized the genetic pleiotropy between sleep conditions and multi-modal brain and cardiac MRI traits at their jointly significant GWAS loci. Briefly, we searched for sleep- significant genetic variants reported in the NHGRI-EBI GWAS catalog83 (version 2022-07- 09). Then we identified those variants (and variants in linkage disequilibrium [LD] with them, r2 ≥ 0.6) that were also significant in GWAS of brain and cardiac MRI traits (Methods). We found that a total of 39 genomic loci showed shared genetic influences on both sleep and imaging traits, covering regional brain volumes54, DTI parameters56, whole brain independent component analysis (ICA)-based resting fMRI traits53,55,84, parcellation- based resting and task fMRI traits58, and cardiac MRI traits60 (Table S4). We tagged previous GWAS for a wide range of sleep conditions, including insomnia39, chronotype85,86, daytime nap39, depressive symptom (sleep problems)87, getting up39, hypersomnia88, sleep duration37,89, and snoring90 (Fig. 3 and Table S5). Additionally, we estimated the probability that sleep-imaging trait pairs shared causal genetic variants using the Bayesian colocalization test91. We considered pairs with a probability of shared causal variant (PPH4) greater than 0.8 to be colocalized91,92. Below we summarized the results for each imaging modality group.

Fig. 3.
Fig. 3. Jointly significant genomic loci for sleep conditions and imaging traits.

(A) In the NHGRI-EBI GWAS catalog (https://www.ebi.ac.uk/gwas/, version 2022-07-09), we found shared genetic influences between sleep and imaging traits in 39 genomic loci (names are in black color). That is, sleep and imaging GWAS reported significant variants in these loci and the index variants were in linkage disequilibrium (LD, r2 ≥ 0.6). Each imaging modality is labeled with a different color. We tagged a wide range of reported sleep traits and grouped them into chronotype (e.g., morningness and morning person), insomnia (e.g., insomnia and insomnia symptoms), sleep duration (e.g., sleep duration long sleep and short sleep), and others, such as snoring, hypersomnia, getting up, and daytime nap. (B) We further summarized the results into a table, where the x-axis represents the 39 genomic regions and y-axis displays the sleep traits. Each imaging modality is labeled with a different color and the orange color is used when more than one imaging modalities are observed in the locus.

There were 14 genomic loci where sleep-significant variants were associated with DTI parameters (LD r2≥ 0.6). For example, shared genetic components between DTI parameters and chronotype93 were found in 7 loci, including 1q25.2 (sleep index variant rs975025, tagged many DTI parameters, such as AD of the splenium of corpus callosum), 19p13.11 (rs9636202, mode of anisotropy (MO) of the external capsule), 3p12.3 (rs7429614, MO of the corticospinal tract), 6p22.2 (rs766406, MD of the inferior fronto- occipital fasciculus), 17q21.31 (rs72828815, MD of the anterior corona radiata), 11p15.5 (rs9795439, MO of the cingulum), 8p23.1 (rs2979256, FA of the inferior fronto-occipital fasciculus) (Figs. 4A-4B and S9-S12). The 1q25.2, 19p13.11, and 11p15.5 regions had strong evidence of shared causal genetic variants between chronotype and DTI parameters (PPH4 > 0.976). Multiple sleep index variants were expression quantitative trait loci (eQTLs) in brain tissues, such as rs975025 (for FAM163A and FAM20B), rs72828815 (for DBF4B and MEIOC), and rs9795439 (KRTAP5-4)94. For sleep duration37, we found shared genetic influences in three loci, including 10p12.31 (rs12246842, MD of the fornix), 6p22.1 (rs34556183, MD of the superior corona radiata), and 17q21.31 (rs1991556, MD of the anterior corona radiata) (Figs. S13-15). Colocalization was detected in 10p12.31 (PPH4 = 0.836), and the sleep index variants rs34556183 and rs1991556 were brain eQTLs of multiple genes such as ZSCAN31 and ARL17A.

Fig. 4.
Fig. 4. Selected genetic loci that were associated with both sleep conditions and imaging traits.

(A) In 1q25.2, we observed colocalization between chronotype (index variant rs975025) and the mean axial diffusivity of the splenium of corpus callosum tract (SCC AD, index variant rs755699529). The posterior probability of Bayesian colocalization analysis for the shared causal variant hypothesis (PPH4) is 0.976. (B) In 19p13.11, we observed colocalization between chronotype and the mean mode of anisotropy of the external capsule (EC MO, shared index variant rs9636202, PPH4 = 0.992). In this region, chronotype was also in LD (r2 ≥ 0.6) with reaction time, life satisfaction, smoking, hypertension, and coronary artery disease. (C) In 4q24, we observed colocalization between insomnia (index variant rs11097861) and functional connectivity among the triple networks of psychopathology (the default mode, central executive, and salience networks, index variant rs4417974, PPH4 = 0.952). (D) In 14q32.2, we observed colocalization between snoring and volume of the left thalamus proper (shared index variant rs2664299, PPH4 = 0.997). We also observed the shared associations (LD r2 ≥ 0.6) with self-reported math ability, cognitive performance, smoking initiation, and educational attainment.

In addition, DTI-insomnia42 genetic overlaps were found in 6 loci, including 3p21.31 (rs10865954, FA of the fornix-stria terminalis), 16q12.1 (rs1544637, AD across the whole brain), 15q26.1 (rs176647, FA of the posterior limb of internal capsule), 17q21.31 (rs2239923, MD of the anterior corona radiata), 9q22.31 (rs10156602, FA of the body of corpus callosum), and 10p12.31 (rs12246842, MD of the fornix) (Figs. S16-S19). The sleep- significant variant rs10865954 was a brain eQTL for multiple genes such as NCKIPSD, WDR6, and GMPPB. We also tagged daytime nap, snoring, and getting up in 17q21.31 (rs57222984, mean AD of the sperior fronto-occipital fasciculus). Hypersomnia had genetic overlaps with MO of the cingulum (hippocampus) in 2p16.1 (rs359268) (Figs. S20). Among these sleep-DTI regions, we also found shared genetic influences with well-being spectrum, neuroticism, depression, risk-taking tolerance, autism spectrum disorder, cognitive traits (e.g, intelligence, math ability, reaction time, and education), life satisfaction, smoking, hypertension, and coronary artery disease, which might be a cause or consequence of white matter changes in the brain. Overall, these results suggested that sleep conditions were genetically associated with white matter integrity across multiple genomic regions, where they also overlapped with neuropsychiatric disorders, cognitive traits, and cardiovascular risk factors.

For resting fMRI, we found the shared genetic influences with sleep in 9 genomic regions. Insomnia genetic loci were in LD (r2≥ 0.6) with resting fMRI traits in 4 regions, including 2q14.1 (rs62158170, such as the somatomotor network), 2p16.1 (rs12713372, the frontoparietal and posterior-multimodal networks), 4q24 (rs11097861, the default mode, central executive, and salience networks), as well as 19q13.32 (rs429358, the default mode network) (Figs. 4C and S21-S23). ICA-based84 and parcellation-based58 resting fMRI traits showed similar overlaps. Strong colocalizations between insomnia and resting fMRI were identified in 2q14.1, 4q24, and 19q13.32 (PPH4 > 0.951). The sleep index variant rs62158170 (2q14.1) was a brain eQTL of PAX8. As one of the two variants in the APOE ε4 locus, rs429358 (19q13.32) contributed to Alzheimer’s disease risk. In rs429358, insomnia also had shared genetic influences with task fMRI traits of the visual network (Fig. S24). Snoring was genetically linked to resting fMRI traits in 12q14.3 (rs10878269, PPH4 = 0.976), 3p11.1 (rs145367119), and 17q21.31 (rs57222984), all of which tagged the triple networks of psychopathology (the default mode, central executive, and salience networks)95 (Fig. S25-S27).

In 2q14.1, 2p16.1, and 17q21.31, we observed shared genetic influences with sleep duration, daytime nap, depressive symptom (sleep problems), and getting up (Figs. S28- S29). Chronotype-resting fMRI overlaps were found in 11q24.1 (rs3867239, the central executive, salience, default mode, and attention networks, PPH4 = 0.975) and 2p16.3 (rs17396357, the central executive, salience, default mode, and attention networks) (Figs. S30-S31). The rs17396357 was a brain eQTL of GTF2A1L, FOXN2, MSH6, STON1-GTF2A1L, STON1, and LHCGR. In these regions, we also found genetic links (LD r2≥ 0.6) with cognitive traits (e.g, cognitive decline, math ability and education), myocardial infarction, type 2 diabetes, mental health (e.g., depression, schizophrenia, autism spectrum disorder, and neuroticism), neurological disorders (e.g., Alzheimer’s disease and Parkinson’s disease), and blood pressure.

Sleep traits and regional brain volumes had shared genetic influences (LD r2≥ 0.6) in 20 genomic regions. Many brain regions were affected by 9 insomnia-related genomic loci (10q24.32, 12q24.31, 4q24, 2q33.3, 5q14.3, 11q14.1, 12q14.3, 10p12.31, and 18q21.2), such as the left/right basal forebrain (rs12411886), cerebellar vermal lobules I-V (rs28576953), left/right accumbens area (rs6855246 and rs13135092), cerebellar vermal lobules VIII-X (rs55772859), left/right putamen (rs16903122, rs667730, and rs375051009), total brain volume (rs61921611), and left/right lateral ventricle (rs12251016) (Figs. S32- S36). Colocalizations between insomnia and regional brain volumes were identified in 10q24.32 and 11q14.1 (PPH4 > 0.853). Insomnia has previously been associated with shrinking subcortical volumes and smaller cortical surface areas96,97. These findings may suggest the genetic basis of brain structure-insomnia connections. Chronotype also overlapped genetically with the left/right putamen in 5q14.3 (rs12657776), as well as other brain regions in 5 more loci, including 16q12.2 (rs4784256, such as brain stem), 9q22.2 (rs3138490, left amygdala), 11p15.5 (rs9795439, white matter), and 3q13.11 (rs34967119, right cuneus), and 2p11.2 (rs11681299, brain stem) (Figs. S37-S39). We observed colocalization between sleep and chronotype in 9p22.2, 11p15.5, and 2p11.2 (PPH4 > 0.856). Furthermore, we found genetic overlaps for daytime nap in 5 additional genomic loci: 1p31.1, 7p22.2, 17q21.31, 11q12.2, and 5q22.2. Daytime nap was genetically linked to the cerebellar vermal lobules VIII-X (rs10782582), left/right lateral ventricle (rs4557589), total brain volume (rs57222984), left/right cerebellum exterior (rs174541), and left/right cerebellum white matter (rs2099810) (Figs. S40-S42). Both 11q12.2 and 5q22.2 had strong evidence of shared causal genetic variants between daytime nap and regional brain volumes (PPH4 > 0.923). In addition, 17q21.31, 11q12.2, and 4q24 (rs13109404, left accumbens area) regions contributed to the genetic links between sleep duration and regional brain volumes. Similarly, 12q14.3, 14q32.2, and 17q21.31 were associated with both snoring and multiple regions, including the right inferior parietal (rs10878269), left thalamus proper (rs2664299), right fusiform (rs4792897), and total brain volume (rs57222984) (Figs. 4D and S43). We found colocalization between snoring and regional brain volumes in 12q14.3 and 14q32.2 (PPH4 > 0.956). There were also genetic overlaps with getting up in 3 of the above loci: 17q21.31, 4q24, and 11q12.2.

Genetic pleiotropy (r2≥ 0.6) was observed between sleep and cardiac MRI traits in 5 regions. Chronotype-heart overlaps were observed in 8p23.1 (rs11992186, regional peak circumferential strain), 22q13.1 (rs139911, ascending aorta maximum area), 11p11.2 (rs11605348, global myocardial-wall thickness), and 17q21.32 (rs11992186, RVEDV) (Figs. S44-S45). Shared genetic influences were found in 11p11.2 with insomnia (rs10838708, global myocardial-wall thickness) and in 17q21.31 with sleep duration, snoring, getting up, as well as daytime nap (rs1991556 and rs57222984, regional myocardial-wall thickness) (Figs. S46-S47). We also found genetic overlaps with mental health (e.g., neuroticism) and heart diseases (e.g., atrial fibrillation and coronary artery disease) in these loci. In summary, sleep had substantial genetic links with brain and heart structural and functional variations. Different sleep conditions, such as insomnia and daytime nap, may have genetic overlaps with the brain and heart in different genomic regions. Integrating imaging and sleep GWAS results may provide insight into sleep-related brain and heart health mechanisms.

Genome-wide and local genetic correlation patterns

We quantified the genome-wide genetic similarity between sleep conditions and imaging traits by genetic correlations (GC) estimated from cross-trait LD score regression98. We collected 19 sets of publicly available sleep GWAS summary statistics (Table S6) and screened them with GWAS summary statistics of multi-modal imaging traits (Methods). At 5% FDR level, significant genetic correlations were observed between sleep and all imaging modalities (Fig. 5A and Table S7). We reported below the top-ranking significant brain and cardiac MRI traits associated with sleep after applying the FDR correction (P < 6.18×10−4).

Fig. 5.
Fig. 5. Genetic correlations between sleep and imaging traits.

(A) We illustrated the pattern of significant genome-wide genetic correlation estimates (at the 5% false discovery rate level by LDSC, P < 6.18×10−4) between various groups of imaging traits (left upper panel) and sleep conditions (left lower panel). We listed the number of significant pairs for each imaging group and sleep condition on the right side.(B) We illustrated pattern of significant local genetic correlation estimates (at the 5% false discovery rate level, P < 1.87×10−4) between various groups of imaging traits (left upper panel) and sleep conditions (left lower panel) by LAVA. We listed the number of significant pairs for each imaging group and sleep condition on the right side.

There were many resting fMRI traits associated genetically with sleep conditions in agreement with our phenotypic association analysis. Particularly, sleep duration and narcolepsy were genetically associated with the functional connectivity of multiple brain networks, such as the somatomotor, visual, language, cingulo-opercular, and frontoparietal (GC range = [-0.472, 0.508], P < 6.18×10−4). Other sleep traits were also genetically correlated with more specific brain networks. For example, snoring was genetically associated with functional connectivity between the default mode and orbito- affective networks, as was daytime napping with functional connectivity between the cingulo-opercular and frontoparietal networks (GC range = [-0.286, 0.250], P < 6.00×10−4). Besides functional connectivity traits, resting amplitude traits62,99,100 were linked to sleep duration and narcolepsy, as well as snoring, with more area-level information revealed (Fig. S48). For example, amplitude results consistently demonstrated that narcolepsy was genetically associated with brain functional activity in the somatomotor and secondary visual areas (GC range = [0.183, 0.418], P < 6.00×10−4). In addition, sleep duration had negative genetic correlations with brain functional activity in the auditory, cingulo- opercular, dorsal-attention, somatomotor, and visual areas (GC range = [-0.240, -0.149], P < 6.00×10−4). Snoring was positively correlated with brain functional areas in the default mode, frontoparietal, and orbito-affective networks (GC range = [0.137, 0.234], P < 6.00×10−4). Genetic links between sleep and brain functions were also observed in task fMRI (GC range = [-0.315, 0.488], P < 6.00×10−4, Fig. S49). Overall, sleep and brain function were genetically correlated throughout brain functional networks, with specific patterns confined to certain areas.

For DTI parameters, significant negative genetic correlations were consistently observed between snoring and the multiple DTI-derived metrics (AD, MD, and RD) of the corticospinal tract (GC range = [-0.202, -0.123], P < 1.03×10−4, Fig. S50). The corticospinal tract is a major neuronal pathway carrying movement-related information from the cerebral cortex to the spinal cord. Impaired corticospinal tract was observed in poor cognitive performance and patients with obstructive sleep spnea101. A few genetic correlations were also observed between regional brain volumes and sleep conditions, especially insomnia (Fig. S51). Specifically, insomnia was genetically correlated with increased brain volume in the left and right caudate (GC range = [0.138, 0.170], P < 7.02×10−5) as well as decreased total brain volume (GC = -0.132, P < 2.15×10−5). The caudate is a well-known functional region for sleep and modulates multiple sleep stages102. Additionally, we found genetic correlations between left ventricle and chronotype, as well as between left atrium and snoring (GC range = [-0.192, 0.166], P < 6.00×10−4, Fig. S52). The left atrium has been reported to be associated with obstructive sleep apnea in coronary artery disease103.

In the above genome-wide genetic correlation analyses, sleep-imaging traits are assumed to be genetically similar throughout the genome. To examine local patterns of shared genetic basis between sleep and imaging and to localize coheritability in specific genomic regions, we partitioned the genome into thousands of regions and performed local genetic correlations analysis using LAVA104 and SUPERGNOVA105 pipelines. In both LAVA and SUPERGNOVA, local genetic correlations were found at the FDR 5% level (P < 1.87×10− 4 for LAVA, P < 5.89×10−5 for SUPERGNOVA) for all sleep conditions, and fMRI traits contributed the largest proportion of significant findings (Figs. 5B, S53, and Tables S8-S9).

In comparison with LDSC, both methods revealed more genetic correlations with task fMRI. For example, among the 777 significant local genetic correlations identified by LAVA, 46.98% (365/777) were related to task fMRI traits. More than 10% (39/365) of the task fMRI-sleep local genetic correlations were located in the 19q13.32 region, which was the locus identified in our association lookup analysis for task fMRI. In addition to 19q13.32, other top-ranking putative pleiotropic regions for task fMRI were 16q12.2, 14q11.2, 7p15.1, and 15q26.2. Moreover, sleep-imaging genetic connections can vary across genomic regions or even be opposites. For example, chronotype had a positive local genetic correlation with left MBelt area task amplitude in 1p13.2 (GC = 0.574, P = 1.10×10− 4), but negative correlation in 16q12.2 (GC = -0.466, P = 1.87×10−4). These results demonstrate the complex genetic co-architecture of sleep and imaging traits.

Causal genetic relationships detected by Mendelian randomization

To investigate the causal genetic links between sleep and imaging traits, we performed Mendelian randomization (MR) analyses with their GWAS summary statistics (Methods). At the Bonferroni significance level (P < 3.67×10−5), we found strong evidence of genetic causal effects from brain structures to sleep conditions (Table S10). For example, increased left ventral DC volume had genetic causal effects on being a morning person (β= -0.102, SE = 0.021, P = 1.71×10−6). All of the MR methods tested indicated the same causal genetic effects direction, and two of them were able to survive the Bonferroni multiple testing procedure. In addition, larger subcortical volumes were causally related to increased risk of daytime nap (β = 0.046, SE = 0.01, P = 3.81×10−6 for the left accumbens area, and β = 0.038, SE = 0.009, P = 1.14×10−5 for the right hippocampus). By balancing adenosine and dopamine activity, the accumbens played an important role in regulating sleep-wake patterns106,107. Furthermore, there was a significant genetic causal relationship between larger total brain volume and longer sleep duration (β = 0.055, SE = 0.01, P = 1.08×10−5). For DTI parameters, we found a causal link between higher MO of the external capsule tract and being an evening person (β = -0.085, SE = 0.02, P = 6.04×10− 6). All these results were summarized in Figure S54. There was no evidence of a causal relationship in another direction (in which sleep traits was the exposure and brain structural traits was the outcomes) after Bonferroni correction.

More genetic causal effects from brain structures to sleep were found at the FDR 5% level (P range = [1.71×10−6, 2.09×10−3]). In total, 73 imaging-sleep causal pairs were identified, most of which were related to DTI parameters (42/73) and regional brain volumes (29/73). DTI parameters from multiple white matter tracts showed genetic causal effects on sleep conditions, such as sleep duration, daytime nap, and chronotype. More volumetric traits were found to have causal effects on sleep. For example, several MR methods have consistently demonstrated that total brain volume has a causal genetic influence on insomnia (β < -0.044, P < 1.93×10−4). In summary, our MR analysis suggested a causal genetic pathway from brain structures to sleep conditions.

Sleep and mental health predictions by integrating multiple data types

According to our imaging genetics analyses, genetic data and multimodal brain and cardiac images could possibly be combined to predict sleep conditions. In clinical and epidemiological studies, sleep has also been reported to be affected by environmental factors, behaviors, biomarkers, and disease status, which makes it desirable to develop sleep prediction models that incorporate imaging genetics and other forms of data. In this section, we performed sleep predictions based on a variety of data types. Our goal is to understand the contributions made by these types of data and their relative performance in predicting different sleep disorders and traits. Our analysis focused on multi-modality brain and cardiac MRI traits, sleep polygenic risk score (PRS), biomarkers, disease status, and 6 categories of environmental variables, including lifestyle and environment, psychosocial factors, physical measures, local environments, and early life factors (Table S11). We used a training, validation, and testing design, where all model parameters were tuned on validation data, and prediction performance was assessed on independent testing data. Detailed information on model training and adjusted covariates can be found in the Methods section.

Figure 6A summarized the results of 7 different sleep conditions using different types of brain imaging data. All sleep conditions were significantly predicted by resting fMRI (prediction correlation β range = [0.040, 0.121], P range = [9.82×10−4, 4.05×10−23]). The ICA-based and parcellation-based resting fMRI traits, as well as the amplitude and functional connectivity traits, showed similar trends, although their relative performance may vary depending on the sleep characteristics being predicted. Each of the regional brain volumes, DTI parameters, and task fMRI traits can significantly predict at least five sleep conditions (Table S12). As for sleep duration, chronotype, and insomnia, adding these additional imaging types will not improve prediction performance over and above resting fMRI, suggesting that resting fMRI is largely responsible for the prediction power of other imaging modalities. As an example, both regional brain volumes and task-related fMRI traits were marginally significant in predicting sleep duration (β range = [0.030, 0.042], P range = [2.47×10−2, 7.28×10−4]). Using regional brain volumes, task fMRI traits, and resting fMRI traits together did not result in a better performance than using resting fMRI alone (β = 0.116 vs. 0.120). These results suggest that among the imaging modalities tested in this study, resting fMRI showed the strongest correlation with the three sleep traits, and it may be the most effective imaging modality to predict insomnia.

Fig. 6.
Fig. 6. Integrative prediction analysis for sleep conditions and mental traits.

(A) Predicting 7 sleep conditions using multimodal brain imaging traits, including DTI parameters, region brain volumes, ICA-based resting fMRI traits (resting ICA), parcellation-based network and area-level resting and task fMRI traits (resting g360 network, resting g360 area, task g360 network, and task g360 area), and all modalities (joint of all brain MRI traits). (B) Predicting 7 sleep conditions using multiple data types. PRS, sleep polygenic risk scores of genetic variants, Brain imaging, multimodal brain imaging traits. (C) The accuracy of neuroticism prediction analysis using different types of data. PRS, neuroticism polygenic risk scores of genetic variants, Brain imaging, multimodal brain imaging traits. (D) The accuracy of different data types in neuroticism prediction analysis before (marginal) and after controlling sleep traits (conditional on sleep).

In the case of getting up, snoring, daytime nap, and narcolepsy, it will be beneficial to integrate multiple imaging modalities rather than using one type of data. For example, task fMRI had the best prediction performance in predicting snoring across different imaging modalities (β range = [0.068, 0.085], P = [2.81×10−7, 1.55×10−10]). Snoring can also be significantly predicted by DTI parameters, regional brain volumes, and resting fMRI traits (β range = [0.040, 0.063], P range = [1.25×10−3, 2.23×10−7]). When all these imaging types were combined, the prediction accuracy increased to 0.103 (P = 1.68×10−14), suggesting that each imaging type contributes additionally to the prediction of snoring. In another example, when combining DTI parameters, regional brain volume, and resting fMRI traits, the prediction performance moved up to 0.109 (P = 1.14×10−18) for nap during day.

Next, we evaluated more data types in sleep prediction (Fig. 6B). Cardiac MRI traits were significantly predictive of 6 sleep traits, with the highest prediction accuracy observed for snoring (β = 0.067, P = 1.57×10−6). Insomnia can be more accurately predicted by using brain and cardiac MRI traits jointly (β = 0.087, P = 8.39×10−10, Fig. S55). In addition, sleep PRS were significant predictors for all sleep conditions (β range = [0.108, 0.201], P range = [2.46×10−15, 1.71×10−49]). Six different PRS methods were used, and all provided similar results (Fig. S56 and Table S12). The combination of PRS and brain imaging data increases prediction accuracy for all sleep traits, with narcolepsy showing the greatest improvement. PRS and imaging traits had a prediction accuracy of 0.108 (P = 2.46×10−15) and 0.124 (P = 2.68×10−20), respectively, for narcolepsy. The performance increased to 0.159 (P = 5.15×10−30) after combination.

Sleep conditions were also strongly predicted by biomarkers and environmental variables. The highest prediction accuracy of biomarker was observed on snoring (β = 0.081, P = 4.98×10−10). Environmental variables were much better at predicting insomnia and getting up than PRS and imaging traits. For example, the prediction accuracy of environmental variables was 0.226 (P = 5.25×10−68) for insomnia, which was largely contributed by the psychosocial factors (β = 0.202, P = 2.15×10−54, Fig. S57). More importantly, combining PRS, imaging, and environmental variables resulted in a higher level of accuracy when predicting all sleep traits. Narcolepsy and snoring were predicted similarly by the three types of data (β range = [0.108, 0.122], P range = [2.46×10−15, 4.58×10−19] for PRS; β range = [0.103, 0.124], P range = [1.68×10−14, 2.68×10−20] for brain imaging; β range = [0.113, 0.121], P range = [1.07×10−11, 1.01×10−17] for environmental variables), and the accuracy increased by more than 50% when they were combined (β range = [0.180, 0.198], P range = [1.17×10−19, 8.01×10−30]). In general, imaging, genetics, and environment all play key roles in predicting sleep variations. Integrating multiple types of data could benefit sleep prediction in clinical applications and research.

Mental health problems were highly comorbid with sleep disorders108,109. In order to gain a better understanding of how sleep contributes to mental health problems, we assessed the predictive power of sleep traits for three mental health traits: depression, neuroticism, and anxiety. We adjusted the same covariates as the above sleep prediction analysis and additionally examined other data types (Methods). For all three mental health traits, sleep traits and early life factors were among the top two strongest predictors (β range = [0.178, 0.316], P range = [9.24×10−43, 8.21×10−135], Figs. 6C, S58, and Table S13). These mental health traits were also significantly predicted by brain imaging and mental health PRS (β range = [0.034, 0.152], P range = [4.84×10−3, 8.43×10−29], Figs. 6C and S58). A combination of all predictors increased the prediction correlation to 0.413 for neuroticism, 0.353 for depression, and 0.277 for anxiety (P range = [5.65×10−91, 1.83×10−176]).

Furthermore, when sleep conditions were controlled as covariates, the prediction power of many traits was reduced, suggesting that sleep may partially explain their predictive power (Figs. 6D and S58). For example, after conditioning on sleep, electronic devices usage no longer had significant predictive power for neuroticism or depression; and alcohol, physical activity, and sun exposure did not have significant predictive power for anxiety. In addition, prediction performance of early life factors on neuroticism decreased from 0.178 (P = 9.24×10−43) to 0.147 (P = 3.80×10−32), suggesting 17.7% prediction power of early life factors on neuroticism could be mediated by sleep. The proportion of reduction was 16.1% for mental health PRS and 28.7% for disease status. Similar reduction was also found in prediction power of PRS and early life factors for depression and anxiety. In summary, sleep was an important predictor of mental health traits and may be able to mediate the influence of genetic and nongenetic risk factors on mental health.

DISCUSSION

Sleep is vital to both physical health and mental wellbeing. Understanding how sleep interacts with human health is of great interest. This study aimed to identify novel phenotypic and genetic connections between sleep-related conditions and human brain and heart health by using multimodal imaging data as endophenotypes. In our study, we found substantial links between sleep and the structure and function of the brain and the heart, some of which can be explained by shared genetic influences. We examined the genetic basis of sleep-imaging connections from a variety of perspectives, including the genetic loci that were jointly significant, the genetic covariance at the local level, and the genetic similarity across the genome. Based on MR analysis, sleep conditions have been found to be more a consequence than a cause of brain structural differences along the genetic pathway, matching the results in a recent study specially on sleep duration and brain structure32. Using prediction analysis, we assessed the relative contributions of genetic and imaging factors to sleep prediction and integrated them with additional environmental data to provide a more accurate prediction. Future epidemiological and clinical studies of sleep and comorbid conditions may benefit from these findings and insights.

In our study, we found that sleep and brain functions are strongly linked through multiple functional networks, including the default mode network. Several studies have shown that the default mode network activity is essential to conscious awareness74,110,111, which may provide an opportunity to investigate the relation between sleepiness and decreased functional connectivity of the default mode network. Daytime sleepiness is generally considered a clinical symptom of sleep disorders or a consequence of insufficient sleep. The symptoms of insomnia and sleep deprivation are usually accompanied by sleepiness112, which both have been linked to reduced connectivity in the brain’s default mode network70,113. Narcolepsy is a rare neurological disorder characterized by daytime sleepiness and the presence, or absence, of cataplexy-like episodes114,115. The cause of narcolepsy is determined by both genetic and environmental factors116, and it may lead to adverse cardiovascular events117. Our phenotypic and genetic results reveal possible mechanisms of default mode network brain activity that underlie these sleep disorders.

The present study has a few limitations. First, our analyses were mainly based on European subjects in the UKB study. In future studies, data from a broader range of sources could be incorporated into the imaging genetics study of sleep, and findings developed by UKB could be generalized to global populations. The current study focused on the human brain and heart, which are two important organs of the human body. Using abdominal images, such as those of the kidneys and liver118, it would be interesting to determine how sleep is connected to more organs. Finally, we focused primarily on linear relationships in our models. Several recent studies have demonstrated that certain sleep traits may also have nonlinear phenotypic and genetic links (such as sleep duration30,31). In the future, advanced statistical and machine learning methods may be used to identify more comprehensive relationships and improve prediction performance.

METHODS

Methods are available in the Methods section.

Note: One supplementary information pdf file and one supplementary table zip file are available.

Data Availability

GWAS summary statistics of brain and cardiac MRI traits can be freely downloaded at BIG-KP https://bigkp.org/ and Heart-KP https://heartkp.org/. GWAS summary statistics of sleep conditions used in this study are publicly available and the links can be found in Table S6. The individual level UK Biobank data used in this study can be obtained from https://www.ukbiobank.ac.uk/. We have built an interactive web browser to share our results at www.ig4sleep.org.

https://www.ig4sleep.org

AUTHOR CONTRIBUTIONS

B.Z. designed the study. Z.F., B.Z., J.S., X.Y., B.L., J.L., and T.L. analyzed the data. Y.L., Z.F., and B.Z. designed the website and developed online resources. Q.W., P.P., T.L., and H.Z. provide comments and interpret the results. Z.F. and B.Z. wrote the manuscript with feedback from all authors.

CORRESPINDENCE AND REQUESTS FOR MATERIALS should be addressed to B.Z

COMPETETING FINANCIAL INTERESTS

The authors declare no competing financial interests.

METHODS

Phenotypic sleep-imaging analyses

The UKB study (https://www.ukbiobank.ac.uk/) recruited approximately half a million participants aged between 40 and 69 years between 2006 and 201059. The ethics approval of the UKB study was from the North West Multicentre Research Ethics Committee (approval number: 11/NW/0382). The UKB imaging study started in 2014 and aimed to collect multi-modal imaging data from 100,000 subjects119. Detailed procedures to generate brain and cardiac MRI traits used phenotypic analysis can be found in previous papers54,56,58,60,61. All these imaging traits were generated from the raw images in Category 100003 (https://biobank.ndph.ox.ac.uk/showcase/label.cgi?id=100003). Overall, we used 101 regional brain volumes and 63 cortical thickness measures from T1- weighted structural MRI image54,60, 110 DTI parameters from diffusion MRI image56, 92 functional activity (amplitude) and connectivity traits from resting-state and task-based fMRI image58, respectively, as well as 82 cardiac MRI traits from the short-axis, long-axis, and aortic cine images61. These traits captured the structural and functional characteristics of the human brain, heart, and aorta. Briefly, the advanced normalization tools120 (ANTs) was used to generate regional brain volumes for 98 cortical and subcortical areas, as well as 3 global brain volume measures, including the total gray matter volume, total white matter volume, and total brain volume. Similarly, 63 global and regional cortical thickness measured were also generated by ANTs. In addition, we applied the ENIGMA-DTI pipeline121,122 to diffusion MRI, and generated 110 tract-averaged parameters, including the fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (AD), radial diffusivity (RD), and mode of anisotropy (MO) for 21 predefined major white matter tracts, as well as across the whole brain (5 × 22). For fMRI, we used a parcellation-based approach with the Glasser360 atlas69 and partitioned the cerebral cortex into 360 regions in 12 functional networks68, including the primary visual, secondary visual, auditory, somatomotor, cingulo-opercular, default mode, dorsal attention, frontoparietal, language, posterior multimodal, ventral multimodal, and orbito-affective networks. We calculated the mean amplitude of each network, the mean functional connectivity for each pair of networks (including within the same network), and the mean amplitude and mean functional connectivity of the whole cortex. As strong signals were identified in resting fMRI, we also considered the 64,620 (360 × 359/2) area- level high-resolution resting functional connectivity measures, which can provide more fine-grained details on functional organizations of cerebral cortex. The 82 cardiac MRI traits were from 6 categories, including 64 traits of left ventricle, 4 of left atrium, 4 of right ventricle, 4 of right atrium, 3 of ascending aorta, and 3 of descending aorta. See Table S1 for the complete ID list of these traits.

We studied the sleep-imaging phenotypic relationships with 7 self-reported sleep traits: the sleep duration (“About how many hours sleep do you get in every 24 hours? (please include naps)”, Data field 1160); getting up in morning (“On an average day, how easy do you find getting up in the morning? 1) not at all easy; 2) not very easy; 3) fairly easy; 4) very easy”, Data field 1170); morning/evening person (chronotype, “Do you consider yourself to be 1) definitely a ‘morning’ person; 2) more a ‘morning’ person than an ‘evening’ person; 3) more an ‘evening’ person than a ‘morning’ person; 4) definitely an ‘evening’ person?”, Data field 1180); nap during day (“Do you have a nap during the day? 1) never/rarely; 2) sometimes; 3) usually.”, Data field 1190); sleeplessness/insomnia (“Do you have trouble falling asleep at night or do you wake up in the middle of the night? 1) never/rarely; 2) sometimes; 3) usually.”, Data field 1200); snoring (“Does your partner or a close relative or friend complain about your snoring?”, Data field 1210); and daytime dozing/sleeping (narcolepsy, “How likely are you to doze off or fall asleep during the daytime when you don’t mean to? (e.g. when working, reading or driving) 1) never/rarely; 2) sometimes; 3) often; 4) all of the time.”, Data field 1220). We used the data coded by the UKB study and removed the subjects with responses “do not know” or “prefer not to answer”. In our phenotypic analysis, we used the white British imaging individuals in UKB phases 1 to 3 data release (released up through 2020, average n = 29,025, mean age range = (45,82), mean = 64.15, standard error = 7.67, and proportion of female was 51.6%). We fitted linear models for each pair of sleep and imaging traits, in which we adjusted for the effects of age (at imaging), age-squared, sex, age-sex interaction, age-squared-sex interaction, imaging site code, the top 40 genetic principal components (PCs)123, volumetric scaling, head motion, head motion-squared, brain position, and brain position- squared53,55. For regional brain volumes and regional cortical thickness measures, we additionally adjusted for the total brain volume and global mean thickness, respectively, to remove the global effects. The values greater than five times the median absolute deviation from the median were removed in each imaging trait and continuous covariate variable. The P values from two-sided t test were reported (R version 3.6.0).

We also performed the PRS-based association analysis for the same set of imaging traits. Specifically, we used the GWAS summary statistics of these imaging traits released by previous imaging GWAS studies54,56,58,60,61 and constructed PRS based on PRS-CS124. We used all default parameters in the PRS-CS software (https://github.com/getian107/PRScs) and generated the PRS for all non-imaging individuals in the UKB study (removing relatives of the UKB imaging individuals). We then used these genetically predicted imaging traits to repeat the above association analysis with the 7 sleep traits in non-imaging unrelated white British UKB subjects (average n = 217,254, removing relatives of imaging subjects, Methods). We adjusted for the effects of age (at baseline), age-squared, sex, age-sex interaction, age-squared-sex interaction, and the top 40 genetic PCs123.

Genetic sleep-imaging analyses

Our genetic analyses were mainly based on GWAS summary statistics of sleep-related traits and imaging traits from previous studies. First, we systematically looked for the genomic loci that were reported to be significant for both sleep and imaging traits. We looked up the reported sleep-significant genetic variants in the NHGRI-EBI GWAS catalog83 (https://www.ebi.ac.uk/gwas/, version 2022-07-09). Among these sleep- associated variants (and variants in LD with them, r2 ≥ 0.6), we searched for those that were also reported to be significant in previous GWAS of brain and cardiac MRI traits, including 101 regional brain volumes54, 110 DTI parameters56, 1,777 ICA-based resting fMRI traits84 (76 ICA-based amplitude traits and 1,701 ICA-based functional connectivity traits), 1,985 network-level parcellation-based resting and task fMRI traits58 (1,066 for resting and 919 for task), and 82 cardiac MRI traits60 (Table S4). For the sleep-significant genetic variants, we also searched to see whether they were reported brain eQTLs on MetaBrain (https://www.metabrain.nl/)94. Finally, we tested for whether the sleep and imaging traits had shared causal genetic variants using the Bayesian colocalization test91.

We examined genetic correlation analysis between sleep and imaging traits via LDSC125 (https://github.com/bulik/ldsc/, version 1.0.1). We collected and used 19 set of publicly available GWAS summary statistics for sleep traits (Table S6). For imaging traits, we screened the 101 regional brain volumes54, 110 DTI parameters56, 1,985 network-level parcellation-based resting and task fMRI traits58 (1,066 for resting and 919 for task), and 82 cardiac MRI traits60. To provide more details of the brain functional organization, we also examined the amplitude traits in each of the 360 brain regions for resting and task fMRI, respectively. The LD scores were calculated by LDSC and were based on the 1000 Genomes European data. We used the default setups of LDSC, in which the HapMap3126 variants and the variants in the major histocompatibility complex region were removed.

Local genetic correlation analyses between sleep and imaging traits were performed separately using LAVA104 (https://github.com/josefin-werme/LAVA) and SUPERGONA105 (https://github.com/qlu-lab/SUPERGNOVA). We tested the same sets of GWAS summary statistics as used in the above LDSC analysis. We followed the tutorials of the two methods and used the default setups. In SUPERGONA, we removed all single nucleotide polymorphisms (SNPs) with missing values in GWAS summary statistics and minor allele frequency (MAF) less than 0.05. We used the provided 1000 Genomes European reference panel and the genome partition file. In LAVA, we input the GWAS summary statistics, the 1000 Genomes European reference genotype data file, and the provided locus definition file. Within each locus, LAVA processed these input files and converted the marginal GWAS summary statistics to locus-specific joint effects. LAVA first performed univariate tests on all traits and loci to select those with sufficient local genetic signals. Bivariate tests were then performed on the selected loci and traits to examine their local genetic correlations.

We performed MR analyses using multiple methods in the two-sample MR package (https://mrcieu.github.io/TwoSampleMR/), including MR Egger, simple median, weighted median, panelized weighted median, and inverse variance weighted127. Bidirectional MR analyses were performed using GWAS summary statistics between each pair of sleep conditions and imaging traits, including 101 regional brain volumes54, 110 DTI parameters56, 92 network-level parcellation-based resting fMRI traits58 (the same traits as in the phenotypic analysis), and 82 cardiac MRI traits60. GWAS summary statistics of sleep conditions were obtained from the IEU GWAS database (https://gwas.mrcieu.ac.uk/). To select independent strong genetic instrumental variables, exposure GWAS summary statistics were clumped by using Plink128 (v1.9) with P-value significance threshold being 5×10−8 (p1) and the secondary significance threshold for clumped genetic variants was also set to be 5×10−8 (p2). Physical distance threshold for clumping was kb = 1,000, LD threshold for clumping was r2 = 0.01, and we used the 1000 Genomes European reference panel. Genetic variants of the exposure remained were extracted from the outcome dataset via the extract_outcome_data() function with default parameters. Harmonization was further performed to ensure the effects of each genetic variant on exposure and outcome were corresponding to each other, which was done by the harmonise_data function() with level of strictness in dealing with SNPs action = 2. Sleep-imaging pairs with number of SNPs less than 10 were discarded. MR was then conducted with the mr() function with the above methods.

Integrative prediction models for sleep and mental health traits

We developed prediction models for sleep by using genetic variants, imaging traits, and a wide range of other data types, such as biomarkers, disease status, and 6 categories of environmental variables, including lifestyle and environment, psychosocial factors, physical measures, local environments, early life factors, and cognitive functions (Table S11). First, we considered performance of the following imaging predictors: 101 regional brain volumes54, 110 DTI parameters56, 1,777 ICA-based resting fMRI traits84, 1,985 network-level parcellation-based resting and task fMRI traits58, 720 area-level parcellation-based resting and task fMRI amplitude traits (360 each), and 82 cardiac MRI traits60. We focused on the unrelated white British subjects and randomly divided the data into three independent parts: training (average n = 20270), validation (average n = 6790), and testing (average n = 6764). We estimated the effect sizes of these imaging traits by ridge regression in the glmnet129 package (R version 3.6.0). The same set of the covariates as in the phenotypic association analysis were controlled. Model parameters were tuned on validation data, and prediction performance was evaluated on testing data by calculating the correlation between predicted and observed sleep traits. Next, we examined the performance of genetic PRS for sleep traits. Data from all UKB white British subjects were used as training data, except for those in the above validation and testing data (and their relatives). We adjusted for the effects of age (at baseline), age-squared, sex, age-sex interaction, age-squared-sex interaction, and the top 40 genetic PCs. We developed using 6 different methods, including pruning and thresholding (P + T), lassosum130, LDpred2131, LDpred-funct132, DBSLMM133, and PRSCS124. The above validation dataset was used to tune the parameters. We also used other data types as predictors with ridge regressions. To help interpret and identify the top-ranking predictors from each category, we also performed phenotypic associations between these variables and sleep traits and shred the results in Table S14 and our website. The performance was evaluated on the subjects with all these data types in the testing dataset. Similar prediction models were also developed on three mental health traits, where sleep traits were added as predictors.

Code availability

We made use of publicly available software and tools. The codes are available upon reasonable request.

Data availability

GWAS summary statistics of brain and cardiac MRI traits can be freely downloaded at BIG- KP https://bigkp.org/ and Heart-KP https://heartkp.org/. GWAS summary statistics of sleep conditions used in this study are publicly available and the links can be found in Table S6. The individual level UK Biobank data used in this study can be obtained from https://www.ukbiobank.ac.uk/. We have built an interactive web browser to share our results at www.ig4sleep.org.

ACKNOWLEDGEMENTS

We thank the individuals represented in the UK Biobank for their participation and the research teams for their work in collecting, processing and disseminating these datasets for analysis. We would like to thank the University of North Carolina at Chapel Hill and Purdue University and their research computing groups for providing computational resources and support that have contributed to these research results. We gratefully acknowledge all the studies and databases that made GWAS and eQTL summary-level data publicly available. This research has been conducted using the UK Biobank resource (application number 22783), subject to a data transfer agreement.

REFERENCES

  1. 1.↵
    Tao, F. et al. Associations of sleep duration and quality with incident cardiovascular disease, cancer, and mortality: a prospective cohort study of 407,500 UK biobank participants. Sleep Medicine 81, 401–409 (2021).
    OpenUrl
  2. 2.↵
    Vendette, M. et al. REM sleep behavior disorder predicts cognitive impairment in Parkinson disease without dementia. Neurology 69, 1843–1849 (2007).
    OpenUrlCrossRefPubMed
  3. 3.↵
    Postuma, R.B. et al. Risk and predictors of dementia and parkinsonism in idiopathic REM sleep behaviour disorder: a multicentre study. Brain 142, 744–759 (2019).
    OpenUrlCrossRefPubMed
  4. 4.↵
    Sabia, S. et al. Association of sleep duration in middle and old age with incidence of dementia. Nature Communications 12, 1–10 (2021).
    OpenUrl
  5. 5.↵
    Cappuccio, F.P., Cooper, D., D’Elia, L., Strazzullo, P. & Miller, M.A. Sleep duration predicts cardiovascular outcomes: a systematic review and meta-analysis of prospective studies. European heart journal 32, 1484–1492 (2011).
    OpenUrlCrossRefPubMedWeb of Science
  6. 6.
    Lao, X.Q. et al. Sleep quality, sleep duration, and the risk of coronary heart disease: a prospective cohort study with 60,586 adults. Journal of Clinical Sleep Medicine 14, 109–117 (2018).
    OpenUrl
  7. 7.↵
    Jike, M., Itani, O., Watanabe, N., Buysse, D.J. & Kaneita, Y. Long sleep duration and health outcomes: a systematic review, meta-analysis and meta-regression. Sleep medicine reviews 39, 25–36 (2018).
    OpenUrlCrossRefPubMed
  8. 8.↵
    Nagai, M., Hoshide, S. & Kario, K. Sleep duration as a risk factor for cardiovascular disease-a review of the recent literature. Current cardiology reviews 6, 54–61 (2010).
    OpenUrl
  9. 9.↵
    Genuardi, M.V. et al. Association of short sleep duration and atrial fibrillation. Chest 156, 544–552 (2019).
    OpenUrl
  10. 10.↵
    Morovatdar, N. et al. Sleep duration and risk of atrial fibrillation: a systematic review. Journal of atrial Fibrillation 11(2019).
  11. 11.↵
    Titova, O.E., Michaëlsson, K. & Larsson, S.C. Sleep duration and stroke: prospective cohort study and mendelian randomization analysis. Stroke 51, 3279–3285 (2020).
    OpenUrl
  12. 12.↵
    Nikbakhtian, S. et al. Accelerometer-Derived Sleep Onset Timing and Cardiovascular Disease Incidence: A UK Biobank Cohort Study. Available at SSRN 3857637.
  13. 13.↵
    Freeman, D., Sheaves, B., Waite, F., Harvey, A.G. & Harrison, P.J. Sleep disturbance and psychiatric disorders. The Lancet Psychiatry 7, 628–637 (2020).
    OpenUrl
  14. 14.↵
    Alvaro, P.K., Roberts, R.M. & Harris, J.K. A systematic review assessing bidirectionality between sleep disturbances, anxiety, and depression. Sleep 36, 1059–1068 (2013).
    OpenUrlPubMedWeb of Science
  15. 15.↵
    Koffel, E., Khawaja, I.S. & Germain, A. Sleep disturbances in posttraumatic stress disorder: updated review and implications for treatment. Psychiatric annals 46, 173–176 (2016).
    OpenUrl
  16. 16.↵
    Kaskie, R.E., Graziano, B. & Ferrarelli, F. Schizophrenia and sleep disorders: links, risks, and management challenges. Nature and science of sleep 9, 227 (2017).
    OpenUrl
  17. 17.↵
    Lee, M.-J. et al. Comorbidity of narcolepsy and depressive disorders: a nationwide population-based study in Taiwan. Sleep medicine 39, 95–100 (2017).
    OpenUrl
  18. 18.↵
    Khurshid, K.A. Comorbid insomnia and psychiatric disorders: an update. Innovations in clinical neuroscience 15, 28 (2018).
    OpenUrl
  19. 19.↵
    Petersen, S.E. et al. UK Biobank’s cardiovascular magnetic resonance protocol. Journal of cardiovascular magnetic resonance 18, 1–7 (2015).
    OpenUrlCrossRef
  20. 20.
    Littlejohns, T.J., Sudlow, C., Allen, N.E. & Collins, R. UK Biobank: opportunities for cardiovascular research. European heart journal 40, 1158–1166 (2019).
    OpenUrlCrossRefPubMed
  21. 21.↵
    Miller, K.L. et al. Multimodal population brain imaging in the UK Biobank prospective epidemiological study. Nature Neuroscience 19, 1523–1536 (2016).
    OpenUrlCrossRefPubMed
  22. 22.↵
    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. Translational psychiatry 10, 1–28 (2020).
    OpenUrl
  23. 23.↵
    de Oliveira, F.G., Pinto, I., Valdigem, B., Senra, T. & Bertolami, A. Evaluation of late atrial enhancement by cardiac magnetic resonance imaging in patients with obstructive sleep apnea. Sleep Medicine 74, 204–210 (2020).
    OpenUrl
  24. 24.↵
    Curta, A. et al. Subclinical Changes in Cardiac Functional Parameters as Determined by Cardiovascular Magnetic Resonance (CMR) Imaging in Sleep Apnea and Snoring: Findings from UK Biobank. Medicina 57, 555 (2021).
    OpenUrl
  25. 25.↵
    Fjell, A.M. et al. Self-reported sleep relates to hippocampal atrophy across the adult lifespan: results from the Lifebrain consortium. Sleep 43, zsz280 (2020).
    OpenUrl
  26. 26.
    Sexton, C.E., Storsve, A.B., Walhovd, K.B., Johansen-Berg, H. & Fjell, A.M. Poor sleep quality is associated with increased cortical atrophy in community-dwelling adults. Neurology 83, 967–973 (2014).
    OpenUrlCrossRef
  27. 27.
    Zitser, J. et al. Sleep duration over 28 years, cognition, gray matter volume, and white matter microstructure: a prospective cohort study. Sleep 43, zsz290 (2020).
    OpenUrl
  28. 28.
    Amorim, L. et al. Poor sleep quality associates with decreased functional and structural brain connectivity in normative aging: a MRI multimodal approach. Frontiers in aging neuroscience 10, 375 (2018).
    OpenUrl
  29. 29.
    Li, G. et al. Magnetic resonance study on the brain structure and resting-state brain functional connectivity in primary insomnia patients. Medicine 97(2018).
  30. 30.↵
    Li, Y. et al. The brain structure and genetic mechanisms underlying the nonlinear association between sleep duration, cognition and mental health. Nature Aging 2, 425–437 (2022).
    OpenUrl
  31. 31.↵
    Tai, X.Y., Chen, C., Manohar, S. & Husain, M. Impact of sleep duration on executive function and brain structure. Communications biology 5, 1–10 (2022).
    OpenUrl
  32. 32.↵
    Fjell, A.M. et al. Sleep duration and brain structure – phenotypic associations and genotypic covariance. bioRxiv, 2022.02.15.480501 (2022).
  33. 33.↵
    Smith, S.M. & Nichols, T.E. Statistical challenges in “big data” human neuroimaging. Neuron 97, 263–268 (2018).
    OpenUrlCrossRefPubMed
  34. 34.↵
    Marek, S. et al. Towards reproducible brain-wide association studies. BioRxiv (2020).
  35. 35.↵
    Krohn, L. et al. Genome-wide association study of REM sleep behavior disorder identifies novel loci with distinct polygenic and brain expression effects. medRxiv (2021).
  36. 36.
    Dashti, H.S. et al. Genetic determinants of daytime napping and effects on cardiometabolic health. Nature communications 12, 1–15 (2021).
    OpenUrl
  37. 37.↵
    Dashti, H.S. et al. Genome-wide association study identifies genetic loci for self-reported habitual sleep duration supported by accelerometer-derived estimates. Nature communications 10, 1–12 (2019).
    OpenUrlCrossRef
  38. 38.
    Doherty, A. et al. GWAS identifies 10 loci for objectively-measured physical activity and sleep with causal roles in cardiometabolic disease. bioRxiv, 261719 (2018).
  39. 39.↵
    Jansen, P.R. et al. Genome-wide analysis of insomnia in 1,331,010 individuals identifies new risk loci and functional pathways. Nature Genetics 51, 394–403 (2019).
    OpenUrlCrossRefPubMed
  40. 40.
    Jones, S.E. et al. Genetic studies of accelerometer-based sleep measures yield new insights into human sleep behaviour. Nature communications 10, 1–12 (2019).
    OpenUrlCrossRef
  41. 41.
    Lane, J.M. et al. Genome-wide association analyses of sleep disturbance traits identify new loci and highlight shared genetics with neuropsychiatric and metabolic traits. Nature genetics 49, 274–281 (2017).
    OpenUrlCrossRefPubMed
  42. 42.↵
    Lane, J.M. et al. Biological and clinical insights from genetics of insomnia symptoms. Nature genetics 51, 387–393 (2019).
    OpenUrlPubMed
  43. 43.
    Wang, H. et al. Genome-wide association analysis of self-reported daytime sleepiness identifies 42 loci that suggest biological subtypes. Nature communications 10, 1–12 (2019).
    OpenUrlCrossRef
  44. 44.↵
    Watanabe, K. et al. Genome-wide meta-analysis of insomnia in over 2.3 million individuals indicates involvement of specific biological pathways through gene-prioritization. medRxiv (2020).
  45. 45.↵
    Lane, J.M. et al. Genetics of circadian rhythms and sleep in human health and disease. Nature Reviews Genetics, 1–17 (2022).
  46. 46.↵
    Hernandez, L.M. et al. Decoupling sleep and brain size in childhood: An investigation of genetic covariation in the ABCD study. medRxiv, 2020.10. 02.20204735 (2021).
  47. 47.↵
    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).
    OpenUrlCrossRef
  48. 48.
    Córdova-Palomera, A. et al. Cardiac Imaging of Aortic Valve Area From 34 287 UK Biobank Participants Reveals Novel Genetic Associations and Shared Genetic Comorbidity With Multiple Disease Phenotypes. Circulation: Genomic and Precision Medicine 13, e003014 (2020).
    OpenUrl
  49. 49.
    Meyer, H.V. et al. Genetic and functional insights into the fractal structure of the heart. Nature 584, 589–594 (2020).
    OpenUrlCrossRefPubMed
  50. 50.
    Pirruccello, J.P. et al. Analysis of cardiac magnetic resonance imaging in 36,000 individuals yields genetic insights into dilated cardiomyopathy. Nature communications 11, 1–10 (2020).
    OpenUrl
  51. 51.
    Pirruccello, J.P. et al. Genetic Analysis of Right Heart Structure and Function in 40,000 People. bioRxiv (2021).
  52. 52.
    Thanaj, M. et al. Genetic and environmental determinants of diastolic heart function. medRxiv (2021).
  53. 53.↵
    Elliott, L.T. et al. Genome-wide association studies of brain imaging phenotypes in UK Biobank. Nature 562, 210–216 (2018).
    OpenUrlCrossRefPubMed
  54. 54.↵
    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. Nature genetics 51, 1637–1644 (2019).
    OpenUrlCrossRefPubMed
  55. 55.↵
    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 (2021).
    OpenUrl
  56. 56.↵
    Zhao, B. et al. Common genetic variation influencing human white matter microstructure. Science 372(2021).
  57. 57.
    Grasby, K.L. et al. The genetic architecture of the human cerebral cortex. Science 367(2020).
  58. 58.↵
    Zhao, B. et al. Genetic influences on the intrinsic and extrinsic functional organizations of the cerebral cortex. medRxiv (2021).
  59. 59.↵
    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 medicine 12, e1001779 (2015).
    OpenUrl
  60. 60.↵
    Zhao, B. et al. Heart-brain connections: phenotypic and genetic insights from 40,000 cardiac and brain magnetic resonance images. medRxiv (2021).
  61. 61.↵
    Bai, W. et al. A population-based phenome-wide association study of cardiac and aortic structure and function. Nature Medicine 26, 1654–1662 (2020).
    OpenUrl
  62. 62.↵
    Bijsterbosch, J. et al. Investigations into within-and between-subject resting-state amplitude variations. Neuroimage 159, 57–69 (2017).
    OpenUrl
  63. 63.↵
    Bai, Y. et al. Resting-state functional connectivity of the sensory/somatomotor network associated with sleep quality: evidence from 202 young male samples. Brain Imaging and Behavior, 1–10 (2022).
  64. 64.
    Brooks, S.J., Katz, E.S. & Stamoulis, C. Shorter Duration and Lower Quality Sleep Have Widespread Detrimental Effects on Developing Functional Brain Networks in Early Adolescence. Cerebral cortex communications 3, tgab062 (2022).
    OpenUrl
  65. 65.↵
    Gorgoni, M. et al. Sleep deprivation affects somatosensory cortex excitability as tested through median nerve stimulation. Brain Stimul 7, 732–9 (2014).
    OpenUrl
  66. 66.↵
    Xiao, F. et al. Independent component analysis and graph theoretical analysis in patients with narcolepsy. Neuroscience bulletin 35, 743–755 (2019).
    OpenUrl
  67. 67.↵
    Engström, M., Hallböök, T., Szakacs, A., Karlsson, T. & Landtblom, A.-M. Functional magnetic resonance imaging in narcolepsy and the Kleine–Levin syndrome. Frontiers in neurology 5, 105 (2014).
    OpenUrl
  68. 68.↵
    Ji, J.L. et al. Mapping the human brain’s cortical-subcortical functional network organization. Neuroimage 185, 35–57 (2019).
    OpenUrl
  69. 69.↵
    Glasser, M.F. et al. A multi-modal parcellation of human cerebral cortex. Nature 536, 171–178 (2016).
    OpenUrlCrossRefPubMed
  70. 70.↵
    De Havas, J.A., Parimal, S., Soon, C.S. & Chee, M.W. Sleep deprivation reduces default mode network connectivity and anti-correlation during rest and task performance. Neuroimage 59, 1745–51 (2012).
    OpenUrlCrossRefPubMedWeb of Science
  71. 71.
    Robinson, J.L., Erath, S.A., Kana, R.K. & El-Sheikh, M. Neurophysiological differences in the adolescent brain following a single night of restricted sleep - A 7T fMRI study. Dev Cogn Neurosci 31, 1–10 (2018).
    OpenUrl
  72. 72.↵
    Samann, P.G. et al. Increased sleep pressure reduces resting state functional connectivity. MAGMA 23, 375–89 (2010).
    OpenUrlCrossRefPubMed
  73. 73.↵
    Yeshurun, Y., Nguyen, M. & Hasson, U. The default mode network: where the idiosyncratic self meets the shared social world. Nature Reviews Neuroscience 22, 181–192 (2021).
    OpenUrl
  74. 74.↵
    Ferrier, J., Tiran, E., Deffieux, T., Tanter, M. & Lenkei, Z. Functional imaging evidence for task-induced deactivation and disconnection of a major default mode network hub in the mouse brain. Proceedings of the National Academy of Sciences 117, 15270–15280 (2020).
    OpenUrlAbstract/FREE Full Text
  75. 75.↵
    Winer, J.R. et al. Association of short and long sleep duration with amyloid-β burden and cognition in aging. JAMA neurology 78, 1187–1196 (2021).
    OpenUrl
  76. 76.↵
    García-Marín, L.M. et al. Evidence of genetic overlap between circadian preference and brain white matter microstructure. Twin Research and Human Genetics 24, 1–6 (2021).
    OpenUrl
  77. 77.↵
    Rosenberg, J., Maximov, II, Reske, M., Grinberg, F. & Shah, N.J. “Early to bed, early to rise”: diffusion tensor imaging identifies chronotype-specificity. Neuroimage 84, 428–34 (2014).
    OpenUrlCrossRefPubMed
  78. 78.↵
    Hu, F.B. et al. Snoring and risk of cardiovascular disease in women. Journal of the American College of Cardiology 35, 308–313 (2000).
    OpenUrlFREE Full Text
  79. 79.↵
    Facer-Childs, E. & Brandstaetter, R. The impact of circadian phenotype and time since awakening on diurnal performance in athletes. Current biology 25, 518–522 (2015).
    OpenUrlCrossRefPubMed
  80. 80.↵
    Wang, Z. et al. Association of Sleep Duration, Napping, and Sleep Patterns With Risk of Cardiovascular Diseases: A Nationwide Twin Study. Journal of the American Heart Association 11, e025969 (2022).
    OpenUrl
  81. 81.↵
    Li, X. et al. Adherence to a healthy sleep pattern and incident heart failure: a prospective study of 408 802 UK Biobank participants. Circulation 143, 97–99 (2021).
    OpenUrl
  82. 82.↵
    Fan, M. et al. Sleep patterns, genetic susceptibility, and incident cardiovascular disease: a prospective study of 385 292 UK biobank participants. European heart journal 41, 1182–1189 (2020).
    OpenUrlPubMed
  83. 83.↵
    Buniello, A. et al. The NHGRI-EBI GWAS Catalog of published genome-wide association studies, targeted arrays and summary statistics 2019. Nucleic Acids Research 47, D1005–D1012 (2018).
    OpenUrlCrossRef
  84. 84.↵
    Zhao, B. et al. Common variants contribute to intrinsic human brain functional networks. Nature Genetics in press (2022).
  85. 85.↵
    Jones, S.E. et al. Genome-wide association analyses of chronotype in 697,828 individuals provides insights into circadian rhythms. Nature communications 10, 343 (2019).
    OpenUrl
  86. 86.↵
    Lane, J.M. et al. Genome-wide association analysis identifies novel loci for chronotype in 100,420 individuals from the UK Biobank. Nature communications 7, 1–10 (2016).
    OpenUrl
  87. 87.↵
    Thorp, J.G. et al. Genetic heterogeneity in self-reported depressive symptoms identified through genetic analyses of the PHQ-9. Psychological Medicine 50, 2385–2396 (2020).
    OpenUrlCrossRef
  88. 88.↵
    Khor, S.-S. et al. Genome-wide association study of HLA-DQB1* 06: 02 negative essential hypersomnia. PeerJ 1, e66 (2013).
    OpenUrl
  89. 89.↵
    Doherty, A. et al. GWAS identifies 14 loci for device-measured physical activity and sleep duration. Nature communications 9, 1–8 (2018).
    OpenUrl
  90. 90.↵
    Campos, A.I. et al. Insights into the aetiology of snoring from observational and genetic investigations in the UK Biobank. Nature communications 11, 1–12 (2020).
    OpenUrl
  91. 91.↵
    Giambartolomei, C. et al. Bayesian test for colocalisation between pairs of genetic association studies using summary statistics. PLoS genetics 10, e1004383 (2014).
    OpenUrlCrossRef
  92. 92.↵
    Kibinge, N.K., Relton, C.L., Gaunt, T.R. & Richardson, T.G. Characterizing the causal pathway for genetic variants associated with neurological phenotypes using human brain-derived proteome data. The American Journal of Human Genetics 106, 885–892 (2020).
    OpenUrlCrossRefPubMed
  93. 93.↵
    Jones, S.E. et al. Genome-wide association analyses of chronotype in 697,828 individuals provides insights into circadian rhythms. Nat Commun 10, 343 (2019).
    OpenUrlCrossRefPubMed
  94. 94.↵
    de Klein, N. et al. Brain expression quantitative trait locus and network analysis reveals downstream effects and putative drivers for brain-related diseases. bioRxiv (2021).
  95. 95.↵
    Menon, V. Large-scale brain networks and psychopathology: a unifying triple network model. Trends in cognitive sciences 15, 483–506 (2011).
    OpenUrlCrossRefPubMedWeb of Science
  96. 96.↵
    Emamian, F. et al. Alterations of subcortical brain structures in paradoxical and psychophysiological insomnia disorder. Frontiers in psychiatry, 611 (2021).
  97. 97.↵
    Leerssen, J. et al. Brain structural correlates of insomnia severity in 1053 individuals with major depressive disorder: results from the ENIGMA MDD Working Group. Translational psychiatry 10, 1–11 (2020).
    OpenUrl
  98. 98.↵
    Bulik-Sullivan, B. et al. An atlas of genetic correlations across human diseases and traits. Nat Genet 47, 1236–41 (2015).
    OpenUrlCrossRefPubMed
  99. 99.↵
    Alfaro-Almagro, F. et al. Image processing and Quality Control for the first 10,000 brain imaging datasets from UK Biobank. NeuroImage 166, 400–424 (2018).
    OpenUrlCrossRefPubMed
  100. 100.↵
    Zou, Q.-H. et al. An improved approach to detection of amplitude of low-frequency fluctuation (ALFF) for resting-state fMRI: fractional ALFF. Journal of neuroscience methods 172, 137–141 (2008).
    OpenUrlCrossRefPubMedWeb of Science
  101. 101.↵
    Chen, H.-L. et al. White matter damage and systemic inflammation in obstructive sleep apnea. Sleep 38, 361–370 (2015).
    OpenUrlCrossRef
  102. 102.↵
    Jones, N.F. & Ikuta, T. Sleep Duration is Associated with Caudate Volume and Executive Function. (2022).
  103. 103.↵
    Holtstrand Hjälm, H. et al. Association between left atrial enlargement and obstructive sleep apnea in a general population of 71-year-old men. Journal of Sleep Research 27, 254–260 (2018).
    OpenUrl
  104. 104.↵
    Werme, J., van der Sluis, S., Posthuma, D. & de Leeuw, C.A. An integrated framework for local genetic correlation analysis. Nature genetics 54, 274–282 (2022).
    OpenUrlCrossRef
  105. 105.↵
    Zhang, Y. et al. SUPERGNOVA: local genetic correlation analysis reveals heterogeneous etiologic sharing of complex traits. Genome Biol 22, 262 (2021).
    OpenUrlCrossRef
  106. 106.↵
    Oishi, Y. et al. Slow-wave sleep is controlled by a subset of nucleus accumbens core neurons in mice. Nature communications 8, 1–12 (2017).
    OpenUrlCrossRef
  107. 107.↵
    Sardi, N.F., Tobaldini, G., Morais, R.N. & Fischer, L. Nucleus accumbens mediates the pronociceptive effect of sleep deprivation: the role of adenosine A2A and dopamine D2 receptors. Pain 159, 75–84 (2018).
    OpenUrlCrossRefPubMed
  108. 108.↵
    Taylor, D.J. et al. Comorbidity of chronic insomnia with medical problems. Sleep 30, 213–218 (2007).
    OpenUrlPubMedWeb of Science
  109. 109.↵
    Ohayon, M.M. & Roth, T. What are the contributing factors for insomnia in the general population? Journal of psychosomatic research 51, 745–755 (2001).
    OpenUrlCrossRefPubMedWeb of Science
  110. 110.↵
    Luppi, A.I. et al. Consciousness-specific dynamic interactions of brain integration and functional diversity. Nat Commun 10, 4616 (2019).
    OpenUrlCrossRef
  111. 111.↵
    Horovitz, S.G. et al. Decoupling of the brain’s default mode network during deep sleep. Proc Natl Acad Sci U S A 106, 11376–81 (2009).
    OpenUrlAbstract/FREE Full Text
  112. 112.↵
    Stores, G. Clinical diagnosis and misdiagnosis of sleep disorders. J Neurol Neurosurg Psychiatry 78, 1293–7 (2007).
    OpenUrlAbstract/FREE Full Text
  113. 113.↵
    Nie, X. et al. Functional connectivity of paired default mode network subregions in primary insomnia. Neuropsychiatr Dis Treat 11, 3085–93 (2015).
    OpenUrl
  114. 114.↵
    Shan, L., Dauvilliers, Y. & Siegel, J.M. Interactions of the histamine and hypocretin systems in CNS disorders. Nat Rev Neurol 11, 401–13 (2015).
    OpenUrlCrossRefPubMed
  115. 115.↵
    Brown, J. & Makker, H.K. An approach to excessive daytime sleepiness in adults. Bmj 368, m1047 (2020).
    OpenUrlFREE Full Text
  116. 116.↵
    Sehgal, A. & Mignot, E. Genetics of sleep and sleep disorders. Cell 146, 194–207 (2011).
    OpenUrlCrossRefPubMedWeb of Science
  117. 117.↵
    Jennum, P.J., Plazzi, G., Silvani, A., Surkin, L.A. & Dauvilliers, Y. Cardiovascular disorders in narcolepsy: Review of associations and determinants. Sleep Med Rev 58, 101440 (2021).
    OpenUrlPubMed
  118. 118.↵
    Liu, Y. et al. Genetic architecture of 11 organ traits derived from abdominal MRI using deep learning. Elife 10, e65554 (2021).
    OpenUrl
  119. 119.↵
    Littlejohns, T.J. et al. The UK Biobank imaging enhancement of 100,000 participants: rationale, data collection, management and future directions. Nature communications 11, 1–12 (2020).
    OpenUrl
  120. 120.↵
    Avants, B.B. et al. A reproducible evaluation of ANTs similarity metric performance in brain image registration. Neuroimage 54, 2033–2044 (2011).
    OpenUrlCrossRefPubMedWeb of Science
  121. 121.↵
    Jahanshad, N. et al. Multi-site genetic analysis of diffusion images and voxelwise heritability analysis: A pilot project of the ENIGMA–DTI working group. Neuroimage 81, 455–469 (2013).
    OpenUrlCrossRefPubMedWeb of Science
  122. 122.↵
    Kochunov, P. et al. Multi-site study of additive genetic effects on fractional anisotropy of cerebral white matter: comparing meta and megaanalytical approaches for data pooling. Neuroimage 95, 136–150 (2014).
    OpenUrlCrossRefPubMed
  123. 123.↵
    Bycroft, C. et al. The UK Biobank resource with deep phenotyping and genomic data. Nature 562, 203–209 (2018).
    OpenUrlCrossRefPubMed
  124. 124.↵
    Ge, T., Chen, C.-Y., Ni, Y., Feng, Y.-C.A. & Smoller, J.W. Polygenic prediction via Bayesian regression and continuous shrinkage priors. Nature Communications 10, 1776 (2019).
    OpenUrl
  125. 125.↵
    Bulik-Sullivan, B. et al. An atlas of genetic correlations across human diseases and traits. Nature Genetics 47, 1236–1241 (2015).
    OpenUrlCrossRefPubMed
  126. 126.↵
    Consortium, I.H. Integrating common and rare genetic variation in diverse human populations. Nature 467, 52–58 (2010).
    OpenUrlCrossRefPubMedWeb of Science
  127. 127.↵
    Hemani, G. et al. The MR-Base platform supports systematic causal inference across the human phenome. Elife 7(2018).
  128. 128.↵
    Chang, C.C. et al. Second-generation PLINK: rising to the challenge of larger and richer datasets. Gigascience 4, 7 (2015).
    OpenUrlCrossRefPubMed
  129. 129.↵
    Friedman, J., Hastie, T. & Tibshirani, R. Regularization paths for generalized linear models via coordinate descent. Journal of statistical software 33, 1 (2010).
    OpenUrlCrossRef
  130. 130.↵
    Mak, T.S.H., Porsch, R.M., Choi, S.W., Zhou, X. & Sham, P.C. Polygenic scores via penalized regression on summary statistics. Genetic epidemiology 41, 469–480 (2017).
    OpenUrl
  131. 131.↵
    Privé, F., Arbel, J. & Vilhjálmsson, B.J. LDpred2: better, faster, stronger. Bioinformatics 36, 5424–5431 (2020).
    OpenUrl
  132. 132.↵
    Márquez-Luna, C. et al. Incorporating functional priors improves polygenic prediction accuracy in UK Biobank and 23andMe data sets. Nature Communications 12, 1–11 (2021).
    OpenUrl
  133. 133.↵
    Yang, S. & Zhou, X. Accurate and scalable construction of polygenic scores in large biobank data sets. The American Journal of Human Genetics 106, 679–693 (2020).
    OpenUrl
Back to top
Posted September 09, 2022.
Download PDF

Supplementary Material

Data/Code
Email
Mapping sleep’s phenotypic and genetic links to the brain and heart: a systematic analysis of multimodal brain and cardiac images in the UK Biobank
Zirui Fan, Yujue Li, Juan Shu, Xiaochen Yang, Bingxuan Li, Jinjie Lin, Qianwen Wang, Peristera Paschou, Tengfei Li, Hongtu Zhu, Bingxin Zhao
medRxiv 2022.09.08.22279719; doi: https://doi.org/10.1101/2022.09.08.22279719
Twitter logo Facebook logo LinkedIn logo Mendeley logo
Citation Tools

Subject Area