Background
As human economic development as progressed, both schizophrenia (SCZ) and type 2 diabetes mellitus (T2DM), complex polygenic inherited disorders, have become growing challenges that, to date, lack effective solutions [
1,
2]. Accumulating evidence from clinical samples demonstrates that the prevalence of T2DM in patients with SCZ is elevated 2 to 3 times compared with the general population, whereas the aetiology for the co-occurrence of SCZ and T2DM is multifactorial [
3]. Recent studies have shown that drug-naive patients with their first episode of SCZ have an increased risk of T2DM [
4,
5]. Moreover, the increased risk of T2DM is more apparent in young adults with SCZ [
3,
6]. Therefore, a better understanding of the genetic relationship between and common genetic basis of SCZ and T2DM is pivotal for providing insights into the treatment and prevention of these diseases.
Since inherited factors rarely correlate with confounders and exhibit no reverse causation, several studies with limited sample sizes have investigated the involved genes common to both SCZ and T2DM and have reported negligible genetic correlations between SCZ and T2DM [
7,
8]. This conflicts with a weak genome-wide negative correlation between SCZ and T2DM (rg = − 0.07 and P = 0.002) identified in a forthcoming article with a large-scale sample size of European (EUR) subjects[
9]. These inconsistent genetic analysis results may be because the use of limited sample sizes and certain analytical methods potentially result in underpowered correlation analyses, produce bias, and overestimate the results. Moreover, genome-wide association studies (GWASs) involving different population groups can provide samples from global populations to address some of the existing Eurocentric bias, which enhances the ability to identify disease associations and ensures that the findings are mostly relevant to all populations [
10]. Thus, a large-scale trans-ethnic genetic analysis can provide new and cross-validated evidence by employing a range of complementary approaches.
In this study, based on GWAS summary data from European (EUR) and East Asian (EAS) populations including a total of 1,466,906 subjects, multiple complementary genomic analysis approaches were utilized to explore the genetic basis for T2DM and SCZ at different levels, such as the whole-genome, autosomes, loci and causal variants. We aimed to provide more evidence of the genetic basis for the comorbidity of these two diseases. First, in addition to performing a linkage disequilibrium (LD) score regression analysis (LDSC) to estimate the genome-wide correlation of SCZ with T2DM, a stratified autosome-based LDSC was used to estimate autosome correlation. Second, Heritability Estimation from Summary Statistics (HESS) method was performed to estimate the locus-level genetic correlation. Third, based on the causal variants of each disease, polygenic overlap and Mendelian randomization (MR) analyses were performed to examine the genetic link between these two diseases. Furthermore, to identify the basic mechanisms underlying the comorbidity of SCZ and T2DM, a genome-wide cross-trait/ethnic meta-analysis was performed to identify the pleiotropic genes shared between SCZ and T2DM and to determine the common effective organs and blood cell types. Finally, a cross-trait/ethnic meta-analysis based on transcriptome-wide association study (TWAS) data was carried out to explore the canonical pathways in the effective organs (Figure S1).
Discussion
Here, we employed a series of genomic analysis approaches, which were complementary and leveraged GWAS summary data, to explore the genetic basis for the comorbidity of SCZ and T2DM. Moreover, large-scale trans-ethnic data were used to minimize the effects of potential confounding factors and some Eurocentric bias and to enhance the power of the analysis with regard to identifying trait associations. Finally, additional evidence was reported for the genetic correlation between SCZ and T2DM, and the pituitary gland was identified as a common effective organ for these two diseases. These results suggest that the comorbidity between SCZ and T2DM could be partially attributed to shared effective genes in particular organs rather than to genes in the whole body or to environmental factors alone.
Based on the heritability distributed over many variants with small effects and on the analysis using genome-wide variants instead of variants significantly associated with a disorder [
30], we found that SCZ had a significantly negative genetic correlation with T2DM in both the EUR and the EAS populations. Although the negative correlation remained significant only in the EUR samples after Bonferroni correction, there was no significant difference in the genetic correlation between the two populations. Similar results were obtained from the partitioned LDSC analysis of each autosome and the HESS analysis of different loci in each autosome. Next, based on the variants significantly associated with a disorder, which represent a substantial fraction of heritability in many diseases [
31], the MiXeR analysis results also supported the negative genetic correlation between the two diseases, and the MR analysis results suggested that in the EUR population, a genetic predisposition to SCZ tended to prevent T2DM. The effective gene analysis results suggested that more non-protein-coding genes, such as lncRNAs, were likely to take part in such prevention in the pituitary gland.
The reported Bonferroni-significant negative whole-genome correlation between SCZ and T2DM in Europeans was not observed in previous studies, in which nonsignificant negative whole-genome correlation was found with EUR sample sizes close to 140,000 [
7,
8]. However, a forthcoming article also found a significant negative whole-genome correlation on about 7 times more EUR subjects than those in previous studies using the same method of LDSC [
9]. This inconsistence may be because more recent GWAS data are used and a larger sample size may enhance the power of the correlation analysis. Furthermore, at the autosomes and loci level, we found multiple regions of positive and negative correlation between T2D and SCZ, a large sample size in the LDSC analysis may eliminate the biased estimates to null and provide more reliable overall analysis results[
32].
The reported genetic predisposition to SCZ with the tendency to prevent T2DM is incompatible with the epidemiological data indicating an elevated risk of T2DM in SCZ[
3,
4]. SCZ patients may have some lifestyle factors that predispose them to T2DM, such as an unusual diet and easy access to antipsychotics, which may outweigh the presence of the moderate genetic preventive factors and initiate T2DM by modulating non-protein-coding genes. It has been speculated that SCZ risk alleles should be naturally selected for elimination as patients with SCZ have reduced fertility and increased mortality. Conversely, the MiXeR analysis results demonstrated that the SCZ causal variants were of a larger number than those of T2DM and that the majority of T2DM causal variants overlapped with those for SCZ. Interestingly, in the EUR samples, single disorder-specific genomic regions rather than common regions were found to be more likely to cause the maximum negative genetic correlation than was the case in the EAS samples. According to the morphological evolution standpoint that negative genetic correlations between traits may result in a substantial change in morphology [
33], there are microevolutionary responses to the negative genetic correlation between SCZ and T2DM. These suggest that the impairment of molecular modulation in certain organs rather than in the whole body is responsible for the comorbidity of SCZ and T2DM.
Through genome-wide cross-trait/ethnic meta-analysis and positional gene mapping, the top six genes of CDKAL1, KCNQ1, HIST1H2AL, AUTS2, ARL6IP4 and HMGA1 are mainly located on chromosomes 6, 7, 11 and 12, while the top SNPs in chromosomes 9 and 10 are located on the non-protein-coding region. The top one gene CDKAL1 encoding a member of the methylthiotransferase family, a subfamily of the radical S-adenosylmethionine (SAM) superfamily, has been reported be involved in the susceptibility to T2DM in Europeans and Japanese [
34,
35] and to bipolar disorder in Europeans[
36]. Recent research demonstrates that Cdkal1 is necessary for normal mitochondrial morphology by regulating mitochondrial activity. HIST1H2AL located in a histone gene cluster region on chromosome 6p, encodes one of the core histone proteins, Histone H2A type. Histone variants involved in nucleosome composition and histone modification are important for neurodevelopment and are related to the susceptibility of psychiatric disorders[
37,
38]. HMGA1 encoding a nonhistone architectural transcription factor is involved in fundamental cellular processes by regulating chromatin structure and multiple gene expression including the insulin receptor (INSR) and Forkhead box protein O1 (FoxO1)[
39]. INSR is a master regulatory factor for insulin action and glucose homeostasis, and FoxO1 is a critical regulatory factor for gluconeogenesis and glycogenolysis[
40]. AUTS2, i.e. autism-susceptibility-gene-2, encodes an activator of transcription and regulates neurodevelopment. AUTS2 variants can cause a neurodevelopmental and somatic malformation with diverse phenotypes[
41]. KCNQ1 located on chromosome 11q, encodes a member of voltage-gated potassium channel subfamily that can affect cardiac and neuronal action potentials. KCNQ1 has been reported to be associated with SCZ and T2D[
42,
43]. ARL6IP4 encodes ADP ribosylation factor like GTPase 6 (ARL6) interacting protein 4 is predicted to be involved in RNA splicing and mRNA processing and has been reported to be associated with SCZ[
44]. Although the roles of ARL6IP4 is unknown, ARL6 is known to regulate intracellular protein traffic[
45].
Through genome-wide cross-trait/ethnic meta-analysis and integration analysis of tissue-specific gene expression profiles, the pituitary gland was identified as the effective organ for both SCZ and T2DM. Furthermore, the molecular metabolism pathways related to glutathione mediated arsenate detoxification and D-myo-inositol-trisphosphate metabolism were identified as potential basic molecular modulation mechanisms. The pituitary gland is a highly plastic system that can integrate the information from both external and internal environments and maintain homeostasis by the rhythmic secretion of key hormones, such as adrenocorticotrophic hormone and growth hormone, in pulse manner in all vertebrates[
46]. The pituitary is called the master gland since it controls the function of most other endocrine glands, such as the thyroid and adrenal cortex [
47]. In humans, the pituitary can be divided into three anatomically and developmentally distinct constituent parts, i.e. the neurohypophysis (posterior lobe), the adenohypophysis (anterior lobe) and the intermediate lobe[
48]. And pituitary cells have been found to organize in tightly wired networks in both homo and heterotypic manners and communicate with each other[
48]. Thus, the pituitary gland can quickly integrate the hypothalamic and systemic stimuli and optimize its function. It is a central part of the hypothalamic-pituitary gland-adrenal (HPA) axis, which is an important neuroendocrine system with a fundamental role in physiological adaptive responses to stressors [
49]. Cumulative psychiatric stress may induce allostatic load, and exert additional systemic and detrimental effects on neuroendocrine dysfunction, ultimately leading to the onset of T2DM.
Furthermore, it has been reported that chronic inorganic arsenic exposure can lead to neurobehavioral alterations and T2DM [
50,
51]. Although our previous study did not find a significant association of serum arsenic concentration with the risk of SCZ [
1], we found that the
GSTM1 (glutathione S-transferase Mu-1 (GSTM1) gene) null genotype had a risk ratio of 1.14 for SCZ [
52];
GSTM1 is involved in arsenic metabolism and detoxification in humans [
53]. D-myo-inositol-trisphosphate is a second messenger and mobilizes calcium (Ca2+). SCZ involves an abnormality in second messenger precursor availability [
54] and thus is characterized by reduced D-myo-inositol-trisphosphate levels. However, inositol supplementation is an effective and safe strategy for treating T2DM [
55].
CD19, i.e. Cluster of Differentiation 19, is expressed in all B lineage cells in humans and is a B-Lymphocyte surface biomarker[
56]. CD19 + B cells are so important that they can mediate immune response and regulation by the activation of T cells, the release of antibodies and the secretion of cytokines[
57]. Through producing antigen-specific antibody, CD19 + B cells can build the first line of defense against exogenous antigens and further facilitate phagocytosis for destruction and antigen-presentation. During the process, polyclonal B cells may be produced through a mechanism called bystander activation, which can enhance the excessive inflammatory response and self-destruct normal cells[
58]. B-cells have been suggested to be potentially therapeutic targets for SCZ and T2D[
57,
59]. However, no evidence is found for altered numbers of the CD19 + B cells in blood of patients with SCZ except the B-cell related cytokines and certain autoantibodies[
59]. And compared to obese subjects with non-T2D, obese subjects with T2D have been found to have several B cell defects in blood, including lower IL-10 production and ineffective antibody response to new antigens, but present much higher levels of polyclonal activation and antibody secretion[
60]. However, it is unclear if these findings result from T2D development or contribute to T2D pathogenesis in obesity. Although our study has identified CD19 + B cells as potentially trans-ethnic effective cells for SCZ and T2DM, a detailed mechanistic framework requires more actual laboratory work for explanation.
The main limitations of this study as follows: First, the lack of actual laboratory work validated the findings. Nevertheless, the current work strived to obtain the validated findings using a set of complementary methods to perform large-scale analyses at different levels. Second, the stringent Bonferroni correction led to some findings not reaching the threshold of significance. However, this is an effective method of controlling the risk of a type I statistical error. Third, the lack of individual data made it impossible to stratify the analysis by the severity of SCZ. Forth, only SNPs with MAF > 5% were included in these genetic analyses. These are common limitations to approaches based on summary GWAS data. Finally, no T2DM- or SCZ-related traits were analyzed here to provide more evidence. Our future research will investigate the corresponding traits.
The current study identified the pituitary gland as a common effective organ for both T2DM and SCZ, despite T2DM showing a negative genetic correlation with SCZ. Further research may consider T2DM-related glycaemic/lipid/blood pressure traits, including two-hour glucose, fasting glucose, fasting insulin, proinsulin, glycated hemoglobin A1c (HBA1c), low-density lipoprotein cholesterol (LDL), high-density lipoprotein cholesterol (HDL), total cholesterol, triglycerides, systolic blood pressure, diastolic blood pressure, and pulse pressure, and may also consider other mental disorders, such as depression. Future spatial transcriptomics studies using fresh samples may help to verify our findings and to provide new insights into the comorbidity of T2DM and SCZ.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.