Searching for the Missing Heritability

The nature and purpose of sleep remain open scientific questions that continue to challenge sleep researchers. Twin studies show sleep traits, such as sleep duration and aspects of the sleep EEG, are heritable.^1–4 Low inter-individual similarity and high intra-individual similarity in sleep EEG spectra suggest trait-like characteristics of the sleep EEG and indicate the phenotype may be capable of facilitating identification of genes related to sleep physiology.⁵ This realization, along with recent technological developments in genotyping and microarray assessment of gene expression, has inspired researchers to search for mammalian sleep genes. Human genome wide association studies (GWAS) held initial promise, and still may reveal additional polymorphisms associated with sleep phenotypes^6,7; however, they have been limited by polygenic effects, epistasis, the complex nature of sleep traits, and the ever-present tension between the resources necessary for precise phenotyping and the statistical power generated by large sample sizes.^8,9 Also, most GWAS studies account for only a small proportion of familial clustering, even in highly heritable phenotypes like height,¹⁰ prompting researchers to wonder about the nature of this “missing heritability.” Although GWAS studies typically search for common genetic variants, it is unlikely that rare DNA sequence variants will close the missing heritability gap. These challenges necessitate creative statistical genetic approaches to genomic and transcriptomic sleep data. In this issue of SLEEP, Turek and colleagues¹¹ provide a novel method for assessing the association between the mouse genome, transcriptome, and sleep traits. In doing so, they provide a framework for future sleep gene discovery in mammals.

Sleep is a central nervous system function, and gene expression studies are tissue specific, so transcripts were assessed from three mouse brain regions important to sleep physiology—anterior cortex, thalamus/midbrain, and hypothalamus. Two novel systems genetics and statistical approaches were central to their analysis—the causal inference test (Figure 1) and weighted gene co-expression network analysis. The causal inference test assessed transcript abundance to identify candidate causal sleep genes (CCSG) related to the quantitative trait loci of interest, 24-hour REM sleep and wake, and created CCSG centered tissue specific local transcriptional regulatory networks. In parallel, the weighted gene co-expression network analysis assessed tissue specific gene transcript correlations to create co-expression modules hypothesized to be under similar regulatory control. Network modularity was further explored with a topological overlap matrix, which leveraged indirect gene interactions to reveal potential physiological relationships between network related genes. Using these methodologies Turek and colleagues uncovered 65 CCSGs and four expression modules enriched for these genes. Twenty genes from overlapping REM modules from three brain tissue types were also revealed, including genes involved in neurotransmission (Gabra2), protein function regulation (Art3), and cell signaling pathway regulation (Ubc). Sleep state regulatory genes, including a transcriptional silencer (Ncor2) and a synaptic vesicular protein (Amph), were also uncovered suggesting a REM sleep specific regulating module.

Figure 1.

Schematic of causal inference methodology utilized to reveal causal genes for complex traits such as 24-hour REM sleep and wake. In the causal model, the RNA transcript is directly controlled by the DNA locus, and acts upon the phenotypic trait. In the reactive model, the trait is directly controlled by the DNA locus, which is associated with the RNA transcript. In the independent model, both the RNA transcript and trait are controlled by the DNA locus, but independent of one another. This methodology allowed identification of the transcripts most likely to mediate the quantitative trait loci effects (e.g., the causal model) of REM sleep and wake.

Unraveling the genetic basis of sleep is a formidable challenge. The “missing heritability” conundrum necessitates novel research approaches to genomic and transcriptomic data. Turek and colleagues¹¹ present a novel integrative systems genetics approach to link DNA sequence variation, RNA transcript abundance, and sleep phenotyping to identify gene networks responsible for REM sleep and wake duration in mammals. The lack of overlap of REM related CCSG sets across brain tissues, along with the presence of CCSG related gene co-expression modules that were overlapped between tissues, speaks to the breadth and complexity of sleep neuronal networks and provides further evidence of the need for creative systems genetics approaches to these data sets. Further complicating matters is the epigenome, whereupon environmental and heritable factors influence gene expression through DNA methylation, histone modification, and non-coding RNA mediated gene silencing. Future sleep related gene expression research will need to account for these factors to truly understand the links between the sleep genome and transcriptome, and thereby bring us closer to a true understanding of the nature and purpose of sleep.

DISCLOSURE STATEMENT

Dr. Watson has indicated no financial conflicts of interest.

CITATION

Watson NF. Searching for the missing heritability. SLEEP 2011;34(11):1453-1454.