Our study has five main findings. First and foremost, we provided evidence that AD shared genetic basis with insomnia, snoring, and sleep duration. Second, cross-trait meta-analysis identified independent shared loci between AD and insomnia, snoring, or sleep duration and functional analysis highlighted that those shared loci were mainly enriched in liver tissue and lipid metabolic system as well as immune inflammatory system, and were involved in immunological disorders, very-low-density lipoprotein particle clearance, triglyceride-rich lipoprotein particle clearance, chylomicron remnant clearance and positive regulation of T cell mediated cytotoxicity pathways. Third, PPI analysis identified three potential drug target genes that interact with known FDA-approved drug target genes. Fourth, TWAS identified genes that were shared between AD and sleep phenotypes at tissue from immune system, cardiovascular system, endocrine system, digestive system and nervous system. Fifth, bi-directional MR suggested that higher risk of AD was causally related with shorter sleep duration. Our findings advance our understanding of the genetic contribution of AD and sleep pattern, provide insights into the potential regulatory role of shared inheritance whose function warrants follow-up, and elucidate the etiology and mechanisms underlying the co-morbidity of AD and sleep disorders.
Circadian rhythm disturbances have been suggested as biomarkers for clinical stage AD[43]. The findings of our genetic analyses were highly consistent, generally supporting the observational positive associations between AD with insomnia[3] and snoring[44], and the negative associations with sleep duration[45]. We also observed AD was positively associated with napping (rg = 0.16, p = 1.61×10− 2), but this significance disappeared after Bonferroni correction. However, recent longitudinal studies had shown that men with longer napping duration had greater cognitive decline and higher risk of cognitive impairment after adjustment for all covariates[46]. Mechanisms for this association between AD and napping were unknown; it might be partially explained by daytime napping is a result of the erosion of the area of the brain responsible for wakefulness by toxic tau proteins, the accumulation of which ultimately leads to AD, however such a putative causal mechanism needs further experimental validation[46].
Meanwhile, 31 independent SNPs from CPASSOC as well as 30 genes from independent TWAS signals of both AD and three SRPs suggested potential functions relevant to AD. The loci identified in both CPASSOC and TWAS analysis revealed potential shared biological mechanisms in AD progress and SRPs regulation involving immunological disorders, very-low-density lipoprotein particle clearance, triglyceride-rich lipoprotein particle clearance, chylomicron remnant clearance and positive regulation of T cell mediated cytotoxicity pathways. Consequently, we highlighted the potentially interesting functions of the novel associations for PRL(6p22.3) between AD with snoring, as well as focused PTPMT1(11p11.2) and KAT8(16p11.2) region shared between AD with insomnia or sleep duration.
Shared genes associated with AD and SRPs were enriched for expression in most liver and brain tissue, indicating that these disorders might be caused by malfunction of the endocrine system and nervous system. For example, PRL can influence sleep structure, and PRL deficient mice display less rapid eye movement (REM) sleep than wild-type mice[47]. Molina-Salinas et al. have shown that PRL can inhibit glutamate excitotoxicity through the AKT and STAT5 pathways, thereby protecting neuronal cells and decreasing the progression of Alzheimer's disease[22]. Evidence has suggested that somatostatin expression is down-regulated in early aging brains in snoring samples, leading to a progressive decrease in PRL and neprilysin activity and resulting in amyloid b (Ab) peptide accumulation in AD patients[48]. Additionally, PTPMT1 is localized to mitochondria via an N-terminal signaling sequence and is found anchored to the stromal surface of the inner membrane. Study shows that activation of protein tyrosine phosphatase (PTP) hastens the progression of AD[49]. W. Lutz et al. identified PTPMT1 as a common signal in AD and major depressive disorder, which showed consistent moderate expression in brain tissues[50]. A highly promising candidate gene is KAT8, as the dominant SNP at 16p11.2 is located within the third intron of KAT8 and multiple important variants within this locus affect the expression or methylation levels of KAT8 in multiple brain regions, including the hippocampus[51]. The chromatin modifier KAT8 is regulated by KANSL1, a gene associated with AD deficient in Apoε4. A study on Parkinson's disease reported that KAT8 is a potentially causal gene based on GWAS and differential gene expression, implying that KAT8 may have a common role in the neurodegeneration of AD and Parkinson's disease[52]. While previously reported information on gene function may be of great value, it is best to consider all implicated genes as putative causal factors to guide potential functional follow-up experiments.
Our MR analysis provided strong evidence that AD is associated with shorter sleep duration (βIVW = -0.056, PIVW = 1.03×10− 3). However, our findings don’t support a causal effect of sleep duration on AD risk. Notably, growing evidence suggests a J-shaped association between sleep duration and AD, suggesting that the causal effect in the long-sleeper group was larger than the short-sleeper group[53, 54]. Apparent inconsistences between our finding and previous MR studies may be partly due to different definitions, diagnostic criteria and forms of characterization of AD (such as cognitive impairment, memory loss, reaction time and so on)[53], different types of data (individual-level data or summary-level data)[53], different MR methods (linear MR or non-linear MR)[53], or different statistical analysis methods (genetic risk score)[54]. In addition, we found no causal relationship between AD and insomnia, whereas a recent MR study conducted by Huang et al. found that higher risk of AD was associated with lower risk of insomnia (OR: 0.99, PIVW = 7 × 10− 13)[55]. Given the consistency in population, sample size and statistical methods between the study by Huang et al. and the present study, we consider that the difference in results is due to Huang et al. adopted F-statistic > 10 to filter exposure-related SNPs to reduce the weak instrumental bias of using genotype data[55]. Finally, there was little evidence to support a causal effect of insomnia, sleep duration and snoring on AD risk, this finding is consistent with recent researches[55, 56]. Our findings add to previous evidence that AD pathology leads to increased wakefulness and high sleep fragmentation in transgenic mouse models[57], and results in neuronal loss of the suprachiasmatic nucleus (SCN), the master circadian clock of mammals, and the locus coeruleus that are essential for maintaining normal wakefulness[58]. Mechanisms underlie the causality between AD and SRPs remain to be elucidated.
Our study has notable strengths. Specifically, we used data from the largest GWAS available for each trait or disorder, and we explored a wide range of SPRs. Second, we leveraged SuperGnova and Gnova to assess the local genetic correlation and annotation-specific genetic correlation between AD and seven SRPs, respectively. SuperGnova has stronger statistical performance compared to HESS, and GNOVA provides more accurate genetic covariance estimates and
powerful statistical inference than LDSC. Third, the identification of potential target genes through PPI analysis provides a new perspective on shared structure. Fourth, we conducted cross-trait meta-analysis using CPASSOC, which is robust to heterogeneous effects and overlap samples between two phenotypes. Nevertheless, there are several potential limitations of our study. First, although TWAS increased the power to detect significant expression trait associations, the relatively smaller sample size for metabolic traits and GTEx reference panels in certain tissues may be inadequate in detect signals with small to moderate effect. Second, despite the large sample sizes of the consortium-based meta-analysis studies, there were differences in sample size and number of SNPs among different studies. Therefore, the enrichment of SNPs with potential common effects may be lower for traits with relatively few loci and samples in the source studies.
Third, some of the observed associations may not be due to independent effects of the same locus on AD and SRPs, but rather to correlations of traits in the causal pathway or through other unmeasured traits. Fourth, our study was limited to European ancestry, and the shared genetics in other ethnic groups are uncertain. Therefore, future research in other ethnic groups are encouraged. More work is needed to identify individual cell types and more detailed molecular mechanisms with the goal of developing potential therapeutic strategies.
In conclusion, our study provides strong evidence of genetic correlations and causality between AD and sleep pattern, identifies of genetic loci associated with both AD and SRPs risk, thus providing therapeutic opportunities to improve sleep quality and lower the risk of AD. Our results further advance our understanding of AD, and provide insight into the shared etiology for comorbid AD and sleep disorders.