3.1 Analysis of nine cell types of TRG eQTL with mental illness SMR
We obtained 2,089 TRGs from TelNet. In the SMR analysis of sc-eQTL and PD, 40 TRG risk genes were identified in seven PD (AD, ADHD, BD, insomnia, MDD, neuroticism, and SC). They were mainly expressed in four cell types i.e., oligodendrocytes, excitatory neurons, dopaminergic neurons, and OPCs. The STX6 gene may confer the risk of AD by functioning in oligodendrocytes; the LSG1 gene may confer ADHD risk by functioning in Excitatory neurons; the RASGEF1C gene may function in OPCs to confer the risk of insomnia. We found a strong correlation between dopaminergic neurons and PD, where the NT5DC2 gene may reduce the risk of MDD, neuroticism, and SC due to high expression in dopaminergic cells (β MDD = -0.044; β Neuroticism = -0.098; β SC = -0.18) (Appendix Table 1). The greatest number of TRGs was found in SC, with the greatest number of SC risk genes identified in astrocytes (n = 10). Enrichment analysis of SC risk genes using g:Profiler revealed(40) that these genes were mainly associated with the pentose phosphate pathway (Padj=3.465×10− 2), and 22 risk genes may be regulated by the transcription factor PEA3 (Padj=3.828×10− 2) (Fig. 2).
3.2 Analysis of brain tissue eQTL and mental illness SMR of TRG
We identified more TRG with mental illness risk genes from the brain tissue than that in the nine cell types (n = 128) (Appendix Table 2), with 15 risk genes present in both the cells and brain tissue (Fig. 3A). The NSUN2 gene was present in seven cell types and in the brain tissue (Fig. 3B); however, we found the gene only in SC, and the gene was all of the above cells and tissues were negatively correlated with SC (βmean = -0.043). Interestingly, we found that the DNPH1 gene showed a positive correlation with neuroticism in dopaminergic neurons (β = 0.090), while it showed a negative correlation with neuroticism in the brain tissue (β = -0.053). We hypothesized that in dopaminergic neurons, the DNPH1 gene is positively correlated with neuroticism, which may be due to the involvement of the DNPH1 gene in biological processes related to emotional regulation and neurotransmitter signaling in dopaminergic neurons. In contrast, the DNPH1 gene was negatively correlated with neuroticism in brain tissues, possibly due to different functions and regulation of the gene in these tissues. This difference may be related to brain region-specific gene expression patterns, differences in cell type composition, and interaction of other genetic variants or environmental factors associated with neuroticism. We performed a protein-protein interaction network (PPI) (41) analysis of 128 TRG and mental illness risk genes (Appendix Fig. 1) and found significant interactions between these genes (PPI enrichment p-value = 6.33e− 15), and most of them were associated with Neuroticism. Furthermore, most of the genes were associated with metabolic processes such as nucleobase-containing compounds and heterocycle metabolic processes.
3.3 Analysis of brain tissue eQTL and mQTL of TRG with mental illness SMR
We identified 6,755 methylation sites in seven PD (AD, ADHD, BD, insomnia, MDD, neuroticism, and SC) (Appendix Table 3). Significant signals associated with diseases were screened according to the three-step SMR described in the Method section (2.4; Appendix Table 4). We found that ERCC2 was positively correlated with AD (β = 0.026). Thus low DNAm at the cg02288770 locus upregulated ERCC2 expression (β = -0.17), thereby increasing AD risk (β = -0.015) (Fig. 4A). In ADHD, we found four methylation sites regulating two genes (ELOVL1 and GIGYF2). Both the loci cg10842584 and cg16858125 exhibited positive correlations with ELOVL1 (β = 0.28;0.70), except for cg12739119, which was negatively correlated with ELOVL1 (β = -0.55) (Fig. 4B). Furthermore, high DNAm at the cg08698580 locus upregulated GIGYF2 expression (β = 0.75), thereby leading to an increased risk of ADHD (β mQTL = 0.096; β eQTL = 0.12) (Appendix Fig. 2). In BD, four genes i.e., ANP32E, PBRM1, NEK4, and GMIP are regulated by 41 methylation sites. We found that the methylation sites regulating PBRM1 had a high overlap with those regulating NEK4, but some of these methylation sites showed a correlation with both genes in both directions. For example, high DNAm at the cg02792780 locus downregulated PBRM1 expression (β = -0.51) and upregulated NEK4 expression (β = 1.50) to the point of increasing the risk of BD (β mQTL = 0.19; β PBRM1 = -0.32; βNEK4 = 0.13) (Appendix Figs. 3 and 4). We identified 38 methylation sites regulating 12 genes, thus influencing the risk of insomnia. Regarding MDD, 17 methylation sites regulated 4 genes, and XRCC3 was regulated by 12 methylation sites. All of them were positively associated with XRCC3, except for the cg07952815 and cg27170268 loci (Appendix Fig. 5). Two genes i.e., DNPH1 and IQCH, were present in both the brain tissue and cell type with neuroticism in the SMR analysis results; however, in the mQTL of brain tissue, we did not find methylation sites that fulfilled the three-step SMR with DNPH1. Notably, five methylation sites were positively correlated with IQCH (βmean = 0.58). Therefore, we concluded that high DNAm at these five methylation sites would upregulate IQCH expression, thereby reducing the risk of neuroticism (β eQTL = -0.10; β mQTL_mean = -0.076) (Appendix Fig. 6). Regarding SC, thirteen risk genes were identified using SMR for cell type as well as brain tissues—ten of these had associated methylation sites for regulation and increased or decreased the risk of SC.
3.4 Screening the proteome for multiple sclerosis-causing telomere length-associated proteins
At the Bonferroni significance ( P < 1.088 × 10 − 5), the MR analysis revealed 23 proteins associated with telomere length. One protein was found in the CSF and 22 proteins in plasma proteins (Fig. 5) (Appendix Table 5). At the Bonferroni significance ( P < 7.692 × 10− 4), MR analysis revealed seven telomere length-related proteins i.e., GUSB, PSG5, MDM4, SPDEF, TNS2, MSP, and PLK1 were associated with four PD i.e., BD, MDD, neuroticism, and SC (Appendix Table 6). Specifically, GUSB (OR = 2.79; 95% CI, 1.89–4.13) and PSG5 (OR = 1.09; 95% CI, 1.04–1.14) increased the risk of BD; GUSB (OR = 1.25; 95% CI, 1.10–1.41), MDM4 (OR = 1.32; 95% CI, 1.13–1.54), and SPDEF (OR = 1.17; 95% CI, 1.10–1.24) increased the risk of MDD; SPDEF (OR = 1.38; 95% CI, 1.24–1.53), TNS2 (OR = 1.39; 95% CI, 1.18–1.64), and MSP (OR = 1.13; 95% CI, 1.06–1.20) increased the neuroticism risk; and MDM4 (OR = 2.19; 95% CI, 1.84–2.61), PLK1 (OR = 1.97; 95% CI, 1.34–2.89), and SPDEF (OR = 1.28; 95% CI, 1.13–1.44) increased the SC risk.