Sleep disorders are not an uncommon phenomenon in society. Insomnia, marked by an inadequate amount of sleep or quality of sleep, affects about 17% of Canadians (1). The opposite of insomnia, hypersomnia, is characterized by excessive sleepiness or daytime sleep, affects about 8% of individuals (2, 3). Sleep disturbances are known risk factors for diseases such as coronary heart disease, diabetes and arthritis, and an overall poorer quality of life (4).

A close relationship also exists between sleep disturbances and psychiatric disorders (5). More specifically, research suggests that sleep disturbances can precede the onset of depression (6, 7). Prevalence rates for insomnia and hypersomnia are much higher among depressed individuals population; about 44–88% suffer from insomnia while 25% experience hypersomnia (8). These concurrent sleep problems lead to poorer patient outcomes, including a decreased remission rate, and a greater risk of suicide and suicidal ideation (9–12).

Although the full biology behind sleep remains to be fully elucidated, we do know sleep is, at least partially, regulated by genes via circadian rhythms. Circadian rhythms are believed to be one component of the sleep-wake cycle. They are an internal timing mechanism that uses external stimuli to entrain/maintain a 24-h cycle. This leads to external expression of this internal mechanism. The most prevalent and persistent stimulus for mammals is the light-dark cycle (13, 14). There is a group of genes, known as the core circadian or clock genes, which are part of a series of transcriptional-translational feedback loops that help regulate the circadian rhythmicity (15). Some of the core circadian genes have been implicated in both major depression and accompanying sleep problems, albeit with mixed findings (16).

A recent, genome-wide association study (GWAS) (17) in over 1.3 million individuals from two large studies, the UK Biobank and 23 and Me. In this study, they reported numerous significant loci for insomnia, dozing and napping (characteristic of hypersomnia). Here we use a polygenic risk score (PRS) analysis to determine if hits for these three phenotypes have a shared genetic architecture with insomnia and hypersomnia in individuals diagnosed with major depression.

The prevalence of insomnia in our sample was 77.8% while hypersomnia was only present in 29.7%.

N = 4,583 (68%) of individuals were female in the insomnia sample and N = 4,358 (68%) in the hypersomnia sample. Of the females, 77.3% reported insomnia while 22.7% did not; 31.9% of females reported hypersomnia while 68.1% did not. In males, 79% reported insomnia while 21% did not; 25% reported hypersomnia while 75% did not.

Simply, polygenic risk scores are the sum of alleles per locus weighted by an estimate of the alleles' effect sizes based on GWAS summary statistics (38, 39). The most popular method, and arguably the most straight-forward is the pruning + thresholding method, which is available using PRSice (v2) software (37). First, SNPs are pruned based on linkage disequilibrium so that one is left with a smaller number of independent SNPs. Second, eight p-value cut-off scores are used to construct the eight thresholds (37–39). In order to account for our covariates, and maximize power, we performed a linear regression on the PRS scores for each study, standardized the residuals, and then used the standardized residuals to perform a logistic regression on our phenotypes of interest, insomnia and hypersomnia.

First, we found that the prevalence rates for insomnia were representative of the prevalence rates from other reports (8, 40). The prevalence of hypersomnia in our sample was slightly higher than others have reported (8, 40). Interestingly, the prevalence of insomnia between sexes was similar, but the prevalence of hypersomnia in females was 7% higher than in males. Our results are opposite to prior literature, that suggested depressed males are more likely to experience hypersomnia than depressed females (41). However, their sample was significantly smaller (N < 500), which may explain the difference in prevalence rates.

In insomnia, we found that PRS scores for SNPs with a p-value threshold of ≤ 0.05 in the insomnia GWAS marginally predicted insomnia in our PGC MDD cohort. However, <1% of variance was explained by these markers. This truly is not surprising given that insomnia is already known to be a polygenic disorder with an estimated SNP-based heritability to be 7% in the UK Biobank sample whose summary statistics we used (17). This is much lower than the heritability of insomnia estimated using twin studies, which suggests 38–59% (17, 42). The substantial range in heritability was due to sex differences, as the heritability of insomnia was much higher in females (59%) than males (38%), which is why we opted to include sex as a covariate (42). Furthermore, there are many sleep-related changes that differ between the sexes, including sex differences in the homeostatic response to sleep as well as slow wave amplitude during non-rapid-eye movement sleep (43, 44).

The other point we would like to note about the insomnia findings comes from our leave-one-out analysis. We noticed that our results became more significant when QI6C was removed from the analysis. Both the p-value became more significant (p = 0.002) and the variance explained increased (R² = 2.36⁻⁰³), which suggests that this dataset is more heterogenous. Indeed, QI6C is a community sample with diagnoses based on self-report questionnaires. Interestingly, QIO2 is also a community sample, recruited in an identical manner to QI6C; but when this sample is removed, significance is decreased. As with all studies, there are a few other limitations to this work. First, the phenotype we are using to characterize insomnia and hypersomnia are based on self-report data. Some studies suggest that self-reported sleep measures differ from objectively measured sleep data and therefore interpretation of data may be more limited (45). Secondly, there are numerous lifestyle factors known to affect sleep for which we were unable to covary. For example, we were unable to include medications in our analyses. Arguably, almost all of the medications on the market today can affect sleep (46). One of the most well-known classes to influence sleep is sleep medications, such as sedative/hypnotics, which are commonly used to treat insomnia (47). Benzodiazepines are another class of medication used to induce sleep (47). Interestingly, in 2010, 20.8 million prescriptions for sleep medication were written in the USA, a 293% increase from the number written in 1999, which confirms how common these sleep medications are used (48). There are also substances that promote insomnia (47). For example, there is extensive evidence that proves that caffeine reduces sleep quality and duration (49). This would likely bias one's subjective opinions of their sleep. Other lifestyle factor that can negatively affect restorative sleep are regular stress and a lack of routine exercise (50). Third, the unique aspect regarding the PGC-MDD sample is that it is comprised of many smaller samples, including variations such as different genotyping platforms. This means, to control for potential sample effects, the sample origin should be included in the analysis, which negatively affects the power of the analyses. A recent study has also suggested seven subtypes of insomnia (51). Although it would have been interesting to investigate these subtypes in our sample, to determine if one has a stronger genetic basis than others, we were unable to do so as we were limited by the phenotype data. In future, we hope to be able to extend upon our results by incorporating these various subtypes into a model. Finally, despite the fact that the sample sizes were similar for both phenotypes of interest (insomnia and hypersomnia), we were only able to provide evidence for genetic overlap with insomnia. We hypothesize our lack of findings for hypersomnia was because we were using GWAS summary statistics for napping and dozing to represent hypersomnia. This does not fully capture hypersomnia, a broad term used for a variety of sleep disorders ranging from narcolepsy to idiopathic hypersomnia.

To the best of our knowledge, this study is the first of its kind to investigate and find genetic overlap between self-reported insomnia in the general population and depressed individuals. This could have important implications on treating MDD and insomnia going forward. Although the results need to be taken with caution as this is a preliminary study and very little variance can be accounted for by the genes implicated in insomnia in the UK Biobank, suggesting that the genetics overlap is minimal. Further research should focus on a similar study design, but using objectively measured sleep data, which should provide more precision for sleep measures. Overall, our results may make helpful contributions toward more accurate diagnosis and personalized treatment.

The data analyzed in this study is subject to the following licenses/restrictions: The data that support the findings of this study are available from the Psychiatric Genomics Consortium-Major Depressive Disorder (PGC MDD) Working group. Some restrictions apply to the availability of these data. Information on data access is available online at: https://pgcdataaccess.formstack.com/forms/pgc_data_access. Requests to access these datasets should be directed to https://pgcdataaccess.formstack.com/forms/pgc_data_access.

The studies involving human participants were reviewed and approved by the Research Ethics Board at the Centre for Addiction and Mental Health. The patients/participants provided their written informed consent at each site of collection.

LM conceptualized the idea, performed the analyses, and prepared the first draft of the manuscript. AT assisted with statistical analysis, results interpretation, and editing of the manuscript. VG helped with the conceptualization of the idea and data analysis. CZ assisted with analysis interpretation and preparation of the manuscript. VM assisted with data analysis. CL helped with data acquisition, statistical analysis, interpretation of results, and manuscript edits. NM helped with data acquisition, statistical analysis, and manuscript edits. AM helped with statistical analyses and manuscript editing. SA and SR helped with data analysis. MA, BB, DL, DB, BP, GB, SH, LJ, IJ, EB, IH, JP, JS, MW, YM, SS, EG, and GW each contributed to data acquisition and provided edits to the manuscript. GMB assisted with results interpretation and manuscript edits. JK helped with the conceptualization of the idea, data analysis, results interpretation, and editing of the manuscript. All authors contributed to the article and approved the submitted version.

This project was supported by the Frederick Banting and Charles Best Canada Graduate Scholarship (LM), the Granville Nickerson Fellowship in Pharmacogenetics (AT), Brain and Behavior Research Foundation: NARSAD (AT), McLaughlin Centre Accelerator Grant (2019-2020) (AT), CAMH Foundation (VG), Brain and Behavior Research Foundation (NARSAD Young Investigator) (VG), McLaughlin Centre Accelerator Grant (VG), Larry and Judy Tanenbaum Foundation (JK). The NTR/NESDA dataset was funded by the: Netherlands Organization for Scientific Research (NWO) and MagW/ZonMW grants 904-61-090, 985-10-002, 912-10-020, 904-61-193,480-04-004, 463-06-001, 451-04-034, 400-05-717, Addiction-31160008, Middelgroot-911-09-032, Spinozapremie 56-464-14192, Center for Medical Systems Biology (CSMB, NWO Genomics), Biobanking and Biomolecular Resources Research Infrastructure (BBMRI–NL, 184.021.007); the European Science Foundation (ESF, EU/QLRT-2001-01254), the European Community's Seventh Framework Program (FP7/2007-2013), ENGAGE (HEALTH-F4-2007-201413); the European Science Council (ERC Advanced, 230374), Rutgers University Cell and DNA Repository (NIMH U24 MH068457-06), the Avera Institute, Sioux Falls, South Dakota (USA) and the National Institutes of Health (NIH, R01D0042157-01A, MH081802, Grand Opportunity grants 1RC2 MH089951 and 1RC2 MH089995), the Netherlands Organization for Scientific Research (Geestkracht program grant 10-000-1002); the Center for Medical Systems Biology (CSMB, NWO Genomics), Biobanking and Biomolecular Resources Research Infrastructure (BBMRI-NL), VU University's Institutes for Health and Care Research (EMGO+) and Neuroscience Campus Amsterdam, University Medical Center Groningen, Leiden University Medical Center, National Institutes of Health (NIH, R01D0042157-01A, MH081802, Grand Opportunity grants 1RC2 MH089951 and 1RC2 MH089995), Genetic Association Information Network (GAIN) of the Foundation for the National Institutes of Health. Computing was supported by BiG Grid, the Dutch e-Science Grid, which was financially supported by NWO.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.