Genetics Awakens the Complex Relationship Between Sleep and Psychiatric Disorders

The complex relationship between mental health and sleep disturbances was recognized by Hippocrates, who argued that sleeplessness is a sign of suffering and may lead to mental illness, while daytime sleepiness is an indication of malady (Hippocratic Corpus, about 400 bc). However, with sleep disturbances common among patients with psychiatric disorders, the prevailing modern view has been that psychiatric disorders have a major impact on sleep; indeed, sleep problems are listed as symptoms and diagnostic criteria for many psychiatric conditions (1).

Research in the last decades has fostered an emerging, in fact rather Hippocratic, awareness that the relationships between psychiatric disorders and sleep are complex and characterized by bidirectional causality. Evidence that sleep alterations are associated with a higher risk for psychiatric disorders has been accumulating since the 1980s, and it has been shown that sleep deprivation can have opposite effects on psychiatric disorders (1). In addition, while some treatments are used to address either psychiatric and sleep disorders, other treatments for psychiatric disorders may disrupt sleep, and some treatments for sleep disorders may affect psychiatric symptoms (1). Alteration of the circadian system and the evening chronotype have also been linked to an increased risk for psychopathology [cited in (2)], and nocturnal wakefulness has been indicated as risk factor for suicide, capable of predicting next-day suicidal ideation; this may be explained by the fact that the circadian rhythm of negative affect peaks at night (3) and/or by the overlap with depression.

The advent of genome-wide association studies (GWASs), supported by molecular studies, has added to the notion of common mechanisms underlying sleep disturbances and psychiatric disorders. Genetic correlations have been identified for chronotype with major depressive disorder (MDD) and schizophrenia (SCZ), for sleep duration with bipolar disorder (BIP) and SCZ, and between insomnia and MDD [cited in (2)]. Moreover, a possible causal relationship of reduced sleep duration, evening-type chronotype and insomnia on SCZ, and bidirectional effects between insomnia and MDD are suggested by Mendelian randomization analysis [cited in (2)]. While genetic correlation and Mendelian randomization suggest the existence of a shared genetic architecture of sleep disturbances and psychiatric disorders, they are limited in detailing such genetic overlap because they are based on aggregate measures that sum the same and opposite effects of genetic variants on a given pair of traits. Indeed, they cannot identify genetic loci whose effects on sleep disturbances and psychiatric disorders have opposite directionality. The identification of such loci may help in understanding the complex relationship between these phenotypes and identifying sub-groups of patients who may need different treatments. These approaches, as with all GWASs of complex syndromes, are dependent on the fidelity and “purity” of the diagnostic categories that are being compared.

In the current issue of Biological Psychiatry, O’Connell et al. (2) use MiXeR and condFDR/conjFDR to deeply analyze the potential common genetic basis of sleep-related phenotypes and psychiatric disorders, beyond genetic correlation. MiXeR and condFDR/conjFDR are two novel genetic statistical methodologies which, by leveraging GWAS-derived summary statistics, can identify polygenic overlap and shared loci between traits or diseases, regardless of effect direction. O’Connell et al. (2) focus on three psychiatric disorders (MDD, BIP, and SCZ) and three sleep-related phenotypes (insomnia, chronotype, and sleep duration). Of note, summary statistics derived from GWASs on sleep were based on self-reports, which can be biased by psychiatric symptoms (4); for this reason, shared loci with effect directions that are opposite between sleep and psychiatric phenotypes may more reliably denote genetic overlap. Also, the various samples may contain individuals with primary disorders represented in the other samples—for example, the sleep phenotype cohort may contain some subjects with MDD and BIP, and the psychiatric cohorts may contain some subjects with sleep disorders. This may render potential genetic overlap confounded by mixed phenotypes in the samples. However, to limit sample overlap with the sleep-related phenotypes, the authors excluded the UK Biobank contribution to the MDD sample from all analyses (2).

Using MiXeR, O’Connell et al. (2) find evidence of large genetic overlap between all psychiatric and sleep-related traits. For example, of the total number of single nucleotide polymorphisms (SNPs) linked with both traits, 80% is the proportion of SNPs associated with both SCZ and chronotype, and both SCZ and BIP show a genetic overlap with all sleep-related phenotypes. Moreover, nearly all SNPs associated with sleep duration and chronotype are also associated with MDD. Notably, when analyzing the concordance of the effect direction of the shared SNPs with the psychiatric and sleep phenotypes, the authors find a moderate congruency, with values ranging from 47% to 58% [the only exception is the congruency between MDD and insomnia, where the measurement of genetic overlap is less accurate, likely for the low heritability and high polygenicity estimates (2)]. These data indicate that the shared genetic architecture between psychiatric and sleep phenotypes is composed not only by SNPs associated with the two phenotypes with consistent directionality, but also by SNPs associated with the two phenotypes with opposite directionality. Commendably, the authors perform a negative control analysis, detecting a very small genetic overlap with height. The association of psychiatric disorders with SNPs linked with opposite sleep phenotypes is consistent with evidence showing that these phenotypes correlate also with worse health outcomes [cited in (4)].

Using condFDR/conjFDR, O’Connell et al. (2) then identify the specific shared loci significantly associated with each pair of traits, and they prioritize 70 credible genes that may be regulated by genetic variants associated with sleep and psychiatric phenotypes. Remarkably, the predicted effect of lead SNPs of shared loci of psychiatric and sleep phenotypes converge on an uncharacterized long noncoding RNA, LOC100127955. LOC100127955 lies in a locus also harboring a microRNA, MIR465, and a protein-coding gene, MAD1L1: one is rather tempted to speculate that the interplay between these different RNA species, which could act as competitive endogenous RNAs, or microRNA sponges (5), may in part explain the different outcomes that may be associated with LOC100127955 expression. Another credible gene associated with SCZ, chronotype, and insomnia is AS3MT, which has been shown to be up-regulated in early neuronal differentiation (6). This is consistent with the link between sleep and brain development and the hypothesis that some of the shared genes contribute to mediate the relationship between genetic risk factors, early brain development, sleep, and psychiatric phenotypes.

Searching for loci linked with both psychiatric and sleep phenotypes, the condFDR/conjFDR analysis also led to the detection of 9 novel loci associated with MDD, 6 novel loci associated with SCZ, and 46 novel loci associated with sleep-related phenotypes (2). Such findings support the notion that the study of the genetic architecture of a complex disorder may benefit from attention to clinical phenotyping instead of being limited to diagnostic grouping.

In validating a genetic relationship between psychiatric disorders and sleep-related phenotypes, the findings by O’Connell et al. (2) are compatible with the existence of different etiological mechanisms linking genetic risk factors with these conditions. Indeed, risk loci, through the same genetic variant or through variants in linkage disequilibrium, could affect 1) sleep phenotypes, which may in turn affect risk for psychiatric disorders, 2) the risk for psychiatric disorders, which may in turn affect sleep phenotypes, or 3) sleep and psychiatric phenotypes independently. The complexity of this interplay is increased by the fact that sleep and psychiatric phenotypes may affect each other bidirectionally and also through the effect of medications. However, the effect size of these genetic risk variants is small, indicating that they could contribute to slightly alter developmental programs of the brain that are linked to many phenotypes, not only those that are psychiatric and sleep-related.

Moreover, we cannot exclude the possibility that the genetic overlap shown by O’Connell et al. (2) is partly driven by other factors. In this regard, minority and socioeconomic status seem to affect sleep phenotypes (7), and social stratification and socioeconomic status may drive the genetic relationship between health and psychiatry-related traits (8). Being MiXeR- and condFDR/conjFDR-limited to bivariate analyses, further adjustments of these methods may be needed to examine the role of other factors, e.g., social stratification, socioeconomic status, and stressors. Molecular studies may help explore the mechanisms through which genes associated with sleep and psychiatric disorders may affect brain function, shedding light on the path from genes to disease (9).

Once again, the study by O’Connell et al. (2) shows that genes do not code for brain disorders; rather, they interact to affect complex phenotypes branching out in trajectories that may lead to disease (10). The configuration of such diseases can vary between individuals, thus making it desirable to design personalized treatments informed by the knowledge of the genetic factors and the trajectories implicated. Perhaps these data could rekindle the interest for studying intermediate phenotypes located in the path from genes to disease.