Introduction

Alzheimer’s disease (AD), a putative neurodegenerative disease, is the most common type of dementia among elderly people, which affects more than 30 million people in the world and the number will be tripled by 2050 [1]. While AD is primarily recognized for cognitive decline, particularly amnesia [2], psychotic symptoms like hallucinations and delusions are observed in about 40–60% patients with AD [3, 4]. These symptoms may lead to more pronounced cognitive impairment, accelerated cognitive deterioration, and ultimately, a poorer prognosis [3,4,5,6,7]. Notably, psychotic and cognitive symptoms are also the common symptoms of schizophrenia (SCZ) [8], a well-known debilitating psychiatric disorder with a lifetime prevalence of approximately 1% [9], although AD typically shows more frequent non-bizarre delusions and visual hallucinations than SCZ [4, 10, 11]. The investigation of the mechanisms underlying the shared symptoms between AD and SCZ may inform the therapeutic targets shared by both disorders, and therefore accelerate the development of treatments that may benefit both AD and SCZ.

AD and SCZ are highly heritable with an estimated heritability of 60–80% for AD [2, 12] and 80% for SCZ [12, 13]. Several genome-wide association studies (GWASs) have been conducted on AD [14, 15] and SCZ [16,17,18], respectively. The largest GWAS for AD identified 38 risk loci, indicating the contribution of microglia, immune cells, and protein catabolism to AD pathology [14]. Meanwhile, the largest GWAS for SCZ reported 287 risk loci, suggesting the involvement of neuronal function (especially the synaptic biological processes) in SCZ pathology [17]. However, these genetic risk loci can only account for a small proportion of the single nucleotide polymorphism (SNP)-based heritability of each disorder. Moreover, these studies only reported one genome-wide significant SNP (rs6504163) shared by the two disorders, which cannot fully explain for their shared symptoms. Given that the current sample sizes are insufficient for identifying all genetic risk SNPs associated with both disorders, more informative methodologies should be employed to discover additional risk SNPs for AD and SCZ as well as those shared by both disorders. For example, the conditional (condFDR) and conjunctional (conjFDR) FDR method [19] have been used to discover the more genetic associations for phenotypes and the shared genetic SNPs between phenotypes [20,21,22,23,24]. The discovery of additional risk SNPs shared by AD and SCZ may provide new insights into the pathophysiology (biological processes) shared by these two disorders, potentially informing future therapeutic strategies, such as the repurposing of existing antipsychotic drugs that target the shared pathophysiology to treat AD-related psychosis. This is extremely important because few pharmacological agents have been approved to treat AD-related psychosis [25, 26].

Two prior linkage disequilibrium (LD) score regression (LDSC) analyses failed to find a significant genetic correlation between AD and SCZ [27, 28], which may be caused by factors, such as the insufficient sample sizes of GWASs. In this study, we aimed to detect the risk SNPs between AD and SCZ by conducting the cond/conjFDR analyses based on the summary statistics of the currently largest GWASs for AD (n = 455,258) [15] and SCZ (n = 130,644) [17] conducted in individuals of European (EUR) ancestry. In addition to find the risk SNPs shared by AD and SCZ, we also aimed to identify additional risk SNPs associated with AD and SCZ using condFDR by exploiting the joint statistical power of these two GWASs through an empirical Bayesian framework [19].

Materials and methods

GWAS summary statistics for AD and SCZ

The EUR-GWAS summary statistics of AD was SCZ were obtained from 455,258 individuals, including 71,880 patients with clinically diagnosed AD or AD-by-proxy and 383,378 healthy controls [15], while the EUR-GWAS summary statistics of SCZ obtained from 130,644 individuals, including 53,386 patients with SCZ and 77,258 healthy controls [17]. The details about the diagnostic criteria, sample sizes, and sources of the two datasets are provided in Supplementary Table S1.

Genetic correlation analysis

To avoid the bias due to the complex regional LD patterns, we cut out SNPs within the major histocompatibility complex (MHC, hg19 location chr6: 25119106-33854733) and chromosome 8p23.1(hg19 location chr8: 7242715-12483982). These regions are characterized by strong SNP associations and long-range LD, which can obscure the effects of individual loci [19, 29]. The high gene density, extreme polymorphism, and functional clustering of related genes within the MHC region, combined with strong LD, make it particularly challenging to disentangle the contributions of specific loci [30, 31]. By excluding these regions, we aimed to reduce confounding effects and improve the clarity and reliability of genetic loci identification in our study [31, 32]. Then, LDSC [32] was conducted to roughly calculate SNP-based genetic correlations between SCZ and AD based on the Python package (https://github.com/bulik/ldsc).

Cond/conjFDR analysis

By leveraging genetic pleiotropy of complex traits including disorders with polygenic architecture, the cond/conjFDR method (https://github.com/alexploner/cfdr.pleio) utilizes an empirical Bayes statistical framework to improve the discovery of SNPs associated with each trait and to identify SNPs shared by two traits based on the GWAS summary statistics of both traits [19]. The condFDR analysis uses the overlapping SNP-trait associations from GWAS for another trait to re-rank the test statistics for a primary trait, conditioned on the associations of the secondary trait [33, 34]. By inverting the role of the two traits, the condFDR analysis identifies SNPs associated with the other trait. The conjFDR analysis identifies SNPs associated with both traits [34, 35], by defining the conjFDR value for each SNP as the maximum of the two condFDR values for this SNP, providing a conservative estimate of FDR for SNPs jointly associated with both traits [36]. The cond/conjFDR method can be applied to the cross-trait analysis between AD and SCZ because it has been successfully used to identify the shared SNPs between two traits even without overall genetic correlation [37]. Before the cond/conjFDR analyses, the conditional quantile-quantile (Q-Q) plot [19] was utilized to visualize cross-trait enrichment by illustrating the distribution of P values for the primary trait while conditioning on the significance of its association with the secondary trait at thresholds of P ≤ 1, 0.01, 0.001, and 0.0001 [20, 21, 35, 38]. The significant enrichment with both traits was confirmed when SNPs with lower P values for the primary trait increased as a function of the significance of the association with the secondary trait. In the condFDR analyses, we initially defined AD as the primary trait and SCZ as the second trait, and then we shifted the role of the two traits. Therefore, we could discover SNPs associated with SCZ and AD at condFDR <0.01, respectively. We defined conjFDR < 0.05 as the significance threshold for SNPs shared by SCZ and AD and condFDR <0.01 as the significance threshold for the conditional analyses. The two significance thresholds were selected because they were consistently applied in almost all previous cond/conjFDR studies [20, 21, 35]. In addition, we also assessed the effective direction of the common SNPs by observing the beta values. We excluded the extended major histocompatibility complex (hg19 location chr6: 25119106–33854733) and chromosome 8p23.1 (hg19 location chr8: 7242715–12483982) genomic regions from the FDR fitting procedures because complex correlations in regions with intricate LD can bias FDR estimation.

Lead SNPs definition

Functional mapping and annotation (FUMA) [39] (http://fuma.ctglab.nl/) was used to define lead SNPs with an EUR-LD reference panel of the 1000 Genomes Project [40]. For each conditional (FDR < 0.01) or conjunctional (FDR < 0.05) analysis, we defined all significant SNPs with LD r2 < 0.6 as independent significant SNPs, from which we selected SNPs with LD r2 < 0.1 as lead SNPs. To define distinct genomic loci, we merged loci within 250 kb into one locus and selected the most significant SNP as the lead SNP of the merged locus. Of all SNPs identified by the conjunctional analysis (conjFDR < 0.05), we defined those with an LD r2 < 0.1 as lead SNPs and named them as SNPs shared by SCZ and AD. Similarly, we defined lead SNPs identified by SCZ conditioned on AD (condFDR <0.01) as SCZ-related SNPs and those identified by AD conditioned on SCZ (condFDR <0.01) as AD-related SNPs. Of the SCZ-related SNPs, those showing an LD r2 < 0.1 with AD-related SNPs and shared SNPs were defined as SCZ-only SNPs. Of the AD-related SNPs, those showing an LD r2 < 0.1 with SCZ-related SNPs and shared SNPs were defined as AD-only SNPs. Lastly, the genetic signatures shared by SCZ and AD were defined as conjFDR < 0.05 and exhibiting the same effect direction between patients with SCZ and AD.

Functional annotation

The three categories of lead SNPs (SCZ-only SNPs, AD-only SNPs, shared SNPs with the same effect direction) were annotated using FUMA v1.4.1 (https://fuma.ctglab.nl), an online platform for functional annotations of genetic variants [39]. With FUMA, we used annotate variation (ANNOVAR) to determine the functional consequences of the identified SNPs by mapping them to genes and regulatory elements [41]. We calculated combined annotation dependent depletion (CADD) scores to assess deleteriousness and pathogenicity of each SNP [42]. We assessed the regulatory potential of the identified SNPs using RegulomeDB, which integrates the available data to prioritize SNPs that may involve in gene regulation by affecting regulatory elements [43]. We also assessed the minimum chromatin state to predict the consequence of each SNP [44].

All the candidate SNPs in the genomic loci having an LD r2 ≥ 0.6 with one of the shared lead SNPs were aligned to genes by three gene-mapping methods. The first method involved positional gene mapping, where SNPs were linked to genes according to the physical proximity around a 10-kb window in the human reference assembly (GRCh37/hg19). The second method was the expression quantitative trait locus (eQTL) mapping, where SNPs were mapped to genes based on the expression level of a gene is affected by one SNP across 54 human tissue types, utilizing data from the Genotype-Tissue Expression (GTEx) dataset [45]. Significant eQTLs were determined at FDR P < 0.05. The third strategy was chromatin interaction mapping, establishing connections between SNPs and genes through three-dimensional DNA-DNA interactions (FDR P < 1 × 10−6) [39]. Based on the genes identified by either of three strategies, we conducted gene enrichment analyses using ToppGene (https://toppgene.cchmc.org/), which is a reliable tool to perform comprehensive biological pathway analyses [46]. In this analysis, all genes encoding protein were considered as the background genes, and the final results were adjusted for multiple testing (FDR P < 0.05).

Colocalization with gene expression in brain tissue

The coloc package in R [47] (https://chr1swallace.github.io/coloc) was used to identify the Bayesian colocalizations of the lead SNPs shared by SCZ and AD with eQTLs of six brain tissues (cortex, anterior cingulate cortex, frontal cortex, hippocampus, caudate, and putamen) from GTEx [45]. For each identified lead SNP shared by SCZ and AD, we selected SNPs within 250 kb centered on the lead SNP from the original GWASs for SCZ or AD. For each pair of disorders and brain tissues, we calculated the posterior probability (PP) for a SNP being associated with one or both traits by evaluating five possible hypotheses: H0, no association with either trait; H1 or H2, association with only one trait; H3, association with both traits driven by distinct causal variants; and H4, association with both traits driven by a shared causal variant. A higher PP for H4 (PP.H4) means a higher probability of a shared causal variant between traits. Here, we considered evidence for colocalization significant when PP.H4 > 0.80. Shared colocalization should meet this criterion in both SCZ-GWAS and AD-GWAS.

Results

Genetic overlap and correlation between SCZ and AD

Based on the GWAS summary statistics for SCZ and AD, we conducted LDSC analysis and found a modest but statistically significant genetic correlation between SCZ and AD (rg = 0.097; P = 0.026). The conditional Q-Q plots showed pleiotropic enrichment between SCZ and AD. The conditional Q-Q figures of SCZ showed consistent leftward deflection for subsets of variants with higher significance in AD when SCZ was conditioned on AD and vice versa (Fig. 1), indicating substantial polygenic overlap between SCZ and AD.

Fig. 1: Polygenic overlap between SCZ and AD.
Fig. 1: Polygenic overlap between SCZ and AD.The alternative text for this image may have been generated using AI.
Full size image

Conditional Q-Q figures of nominal versus empirical -log10 P values in AD below the conventional GWAS significant threshold of P < 5 × 10−8 with SCZ A and vice versa B, at different levels of significance P ≤ 1, P ≤ 0.1, P ≤ 0.01, P ≤ 0.001, P ≤ 0.0001, separately, are shown. The purple lines indicate all SNPs. The gray lines indicate the null hypothesis. Increasing leftward deflection from the gray line for SNPs with higher significance in the conditional phenotype indicates polygenic overlap. AD Alzheimer’s disease, SCZ schizophrenia.

The shared and distinct genetic signatures between SCZ and AD

At conjFDR < 0.05, we identified 16 lead SNPs associated with both SCZ and AD (Table 1), and the Manhattan plots of conjunctional analyses of SCZ and AD are shown in Fig. 2A. Considering the direction of effect (Beta value), 8 of 16 lead SNPs (50.00%) with concordant association directions in both patients with SCZ and AD were defined as the genetic signatures shared by SCZ and AD, and their genomic distributions are shown in Fig. 3A. All the 8 shared genetic signatures by SCZ and AD were not reached the genome-wide significance (P < 5 × 10−8) in the previous GWASs for both disorders [15, 17]. For example, rs12979278 (chr19) in the MAMSTER gene was identified as a SNP shared by SCZ and AD, but it cannot survive at the genome-wide significance threshold (P < 5 × 10−8) in both SCZ-GWAS (P = 1.07 × 10−5) and AD-GWAS (P = 5.75 × 10−5) (Fig. 3D).

Table 1 Lead SNPs jointly associated with SCZ and AD at conjFDR < 0.05.
Full size table
Fig. 2: The genetic variants jointly and conditionally associated with SCZ and AD.
Fig. 2: The genetic variants jointly and conditionally associated with SCZ and AD.The alternative text for this image may have been generated using AI.
Full size image

A Manhattan plot shows the -log10 transformed conjFDR values for each SNP on the y axis and chromosomal positions along the x axis. The lilac horizontal line represents the threshold for significant shared SNPs (conjFDR < 0.05) between SCZ and AD. B, C Manhattan plots shows the -log10 transformed condFDR values for SCZ conditioned on AD (B), and vice versa (C). The y-axis shows the -log10 transformed condFDR values for each SNP, while the x-axis represents the position of chromosomes. The lilac horizontal lines represent the threshold for conditionally significant SNPs (conjFDR < 0.05). Abbreviations: AD Alzheimer’s disease, FDR false discovery rate, SCZ schizophrenia.

Fig. 3: Shared and distinct genetic architectures between SCZ and AD.
Fig. 3: Shared and distinct genetic architectures between SCZ and AD.The alternative text for this image may have been generated using AI.
Full size image

A The ideogram shows the chromosomal distribution of 268 SCZ-only, 125 AD-only, and 8 shared genetic signatures, visualizing genomic regions of overlap or divergence. B-D. B. The regional plots show an example of the SCZ-only SNPs. The rs2867655 is significantly associated with SCZ conditioned on AD (condFDR = 0.008), but it is not significant (P = 1.12 × 10−7) in SCZ-GWAS. C The regional plots show an example of the AD-only SNPs. The rs17783630 is significantly associated with AD conditioned on SCZ (condFDR = 0.004), but it is not significant (P = 6.38 × 10−7) in AD-GWAS. D The regional plots show an example (rs12979278, conjFDR = 0.024) of SNPs shared by SCZ and AD in the conjFDR analysis, but it is significant in neither SCZ-GWAS (P = 1.07 × 10−5) nor AD-GWAS (P = 5.75 × 10−5). AD Alzheimer’s disease, condFDR conditional false discovery rate, conjFDR conjunctional false discovery rate, GWAS genome-wide association study, SCZ schizophrenia, SNP single nucleotide polymorphism.

We performed condFDR analyses to improve the discovery of SNPs associated with SCZ or AD. At condFDR <0.01, we found 279 lead SNPs associated with SCZ conditioned on AD (Supplementary Table S2), and 133 lead SNPs associated with AD conditioned on SCZ (Supplementary Table S3). The results of the conditional analyses for SCZ and AD were shown using Manhattan plots Fig. 2B, C. Of the 279 condFDR-identified SNPs for SCZ, 268 SNPs were defined as SCZ-only genetic signatures after excluding SNPs showing LD (r2 ≥ 0.1) with any of the 133 condFDR-identified SNPs for AD and 16 conjFDR-identified shared SNPs. Of these 268 SCZ-only lead SNPs, 78 were considered as novel because they could not survive at P < 5 × 10−8 in the original GWAS for SCZ and did not show LD (r2 ≥ 0.6) with any SNPs in the reported 287 distinct loci of SCZ [17]. For example, rs2867655 (chr20) was identified as a SCZ-only SNP that could not survive at the genome-wide significance threshold (P < 5 × 10−8) in SCZ-GWAS (Fig. 3B).With the same strategy we also defined 125 SNPs as AD-only genetic signatures and 55 novel AD-only SNPs [15]. For example, rs17783630 (chr14) was identified as AD-only SNP (Fig. 3C). The distinct genetic signatures for SCZ and AD are provided in Supplementary Tables S2–3 and their genomic distributions are shown in Fig. 3A.

Functional annotations

Functional annotations were conducted for 268 SCZ-only, 125 AD-only, and 8 shared genetic signatures using FUMA. Most SNPs were located in the intron (49% in SCZ-only, 49% in AD-only, and 25% in shared groups) and intergenic (28% in SCZ-only, 22% in AD-only, and 25% in shared groups) regions of protein-coding genes (Fig. 4A). Employing a cutoff of a minimum chromatin state < 8, we identified that 90% of SCZ-only SNPs, 97% of AD-only SNPs, and 100% of shared SNPs were situated within open chromatin regions (Fig. 4B). Among these SNPs, 6 SCZ-only SNPs (rs3098974, rs12492391, rs12671676, rs9668929, rs6499508, rs133367) exhibited RegulomeDB scores of 1f, while 4 AD-only SNPs displayed RegulomeDB scores of 1b (rs13230744), 1d (rs17783630, rs2455069), and 1f (rs6557994) (Fig. 4C), suggesting that they may exert regulatory control over gene expression by affecting transcription factor binding [43]. Furthermore, CADD scores above 12.37 were observed for 21 SCZ-only SNPs and 2 AD-only SNPs (Fig. 4D), indicating their potential deleterious effects [42]. For example, rs2867655 (chr20; CADD = 22.1; Fig. 3B) was one of the most deleterious SCZ-only SNPs. This SNP acts as an eQTL for PTPRT, playing an essential role in neuronal development and the regulation of synaptic formation [48]. The novel SNP rs17783630 (chr14; minimum chromatin state = 2; RegulomeDB = 1d; Fig. 3C) was a regulatory AD-only SNP, which is an eQTL and mQTL of SLC24A4, playing a crucial role in neuronal development and being associated with pathological diagnosis of AD [49]. Additionally, the shared novel SNP rs12979278 (chr19; minimum chromatin state = 1; RegulomeDB = 4; Fig. 3D) is another regulatory SNP, an eQTL and mQTL of MAMSTR in prefrontal cortex. Dysfunction of MAMSTR may lead to increased microglia sensitivity to immune stimulation, a common feature in the pathological AD and SCZ [50, 51].

Fig. 4: Functional annotations of genetic signatures.
Fig. 4: Functional annotations of genetic signatures.The alternative text for this image may have been generated using AI.
Full size image

A The functional consequences of the 268 SCZ-only (green), 125 AD-only (blue), and 8 shared (red) SNPs. B The minimum chromatin state of the three groups of SNPs. The scores of 1–7 suggest open chromatin states. C The RegulomeDB scores of the three groups of SNPs. The lower scores mean higher likelihood of having regulatory functions. D The distribution of CADD scores of the three groups of SNPs. The scores above 12.37 indicate deleterious. AD Alzheimer’s disease, CADD combined annotation dependent depletion, SCZ schizophrenia, SNP single nucleotide polymorphism.

We used three different gene-mapping strategies to map all the candidate SNPs in the genomic loci having an LD r2 ≥ 0.6 with the 8 genetic signatures shared by SCZ and AD, and obtained 124 unique genes (Supplementary Table S4). And we found there were mainly enriched for the biological process of cell-cell adhesion and molecular function of calcium ion binding (Supplementary Table S5).

Shared colocalizations with eQTLs of brain tissues

Coloc was utilized to detect the shared genetic colocalizations between SCZ and AD with eQTLs from 6 brain tissues at the shared SNPs. The shared colocalizations with brain eQTLs were defined as PP.H4 > 0.8 in both SCZ and AD. We found that SCZ and AD had a shared colocalization (SCZ: PP.H4 = 0.900; AD: PP.H4 = 0.973) at rs6494036 (chr15) with the SAFB like transcription modulator (SLTM) expression in putamen (Supplementary Table S6).

Discussion

In this study, we conducted the cond/conjFDR analysis to identify SNPs associated with SCZ or AD and to discover SNPs shared by both disorders. We found 268 SCZ-only SNPs, 125 AD-only SNPs, and 8 SNPs shared by SCZ and AD. The shared genetic signatures between SCZ and AD may explain for the clinical symptoms shared by both disorders and hold the potential to generate novel hypotheses regarding the relationship between these two disorders and may offer valuable insights for clinical practices.

Although GWASs have identified many common genetic variants associated with SCZ or AD [14,15,16,17, 52], only 1 genome-wide significant SNP (rs6504163) was shared by both disorders, which may be biased by the fact that the sample sizes of current SCZ-GWASs and AD-GWASs are far from enough to discover all SNPs associated with each disorder. Consistent with the marginally significant positive genetic correlation (rg = 0.08, P = 0.044) between SCZ and AD [27], we also found a modest genetic correlation (rg = 0.097; P = 0.026) between these two disorders. However, the LDSC-based genetic correlation analysis may underestimate the genetic sharing between two disorders when some shared SNPs have opposite directions of effects. These indicate that more suitable approaches may discover more SNPs shared by SCZ and AD.

By conducting the condFDR analysis to enhance the ability of genetic discovery, one contribution of this study is the identification of 78 novel loci for SCZ and 55 novel loci for AD compared to their original GWASs [15, 17], improving our understanding of the genetic architectures of these two disorders. For example, rs2867655 (chr20) was identified as a new risk SNP for SCZ, a nonsynonymous exonic variant in PTPRT. This gene encodes a member of the type IIB receptor-type protein tyrosine phosphatase family, which is exclusively expressed in the central nervous system [53, 54]. PTPRT plays a crucial role in the regulation of synaptic formation and neuronal development [48], providing additional evidence for the involvement of the two biological processes in schizophrenia [55]. In line with the reported risk SNPs in MS4A6E and CD33 in previous AD-GWASs [14, 56, 57], we found two novel exonic nonsynonymous variants for AD: rs12798157 (chr11) in MS4A6E and rs2455069 (chr19) in CD33. CD33 encodes a sialic-acid-binding immunoglobulin-like lectin family, which is implicated in cell adhesion, endocytosis, and immune regulation [58]. MS4A6E encodes a member of the membrane-spanning 4A gene family, which also affects the immune system [59]. These findings provide further evidence for the role of immune dysfunction in AD.

The most important contribution of this study is the discovery of eight lead SNPs shared by SCZ and AD, which may explain for the psychotic symptoms and cognitive decline shared by both disorders and inform therapeutic strategies that are potentially effective for both disorders. For instance, rs6504163 in ACE has been identified as the only shared SNP between GWASs for SCZ and AD [15, 17]. We confirmed the shared association of ACE with these two disorders by identifying a new shared SNP (rs4308) in this gene. The angiotensin-converting enzyme (ACE), a critical component of renin–angiotensin system (RAS), has been associated with both SCZ and AD [60,61,62,63,64,65,66]. In the postmortem parietal cortex samples, the levels of ACE2 protein and mRNA were higher in patients with a neuropathological diagnosis of AD compared to healthy controls, and brain soluble ACE2 levels were associated with cognitive decline and the accumulation of soluble amyloid-β peptides (Aβ) and insoluble phospho-tau [66]. Similarly, patients with SCZ also showed increased ACE levels in cerebrospinal fluid [63] and serum [64] as well as increased plasmatic ACE activity [65]. The increased plasmatic ACE activity was also correlated with cognitive decline in SCZ patients [65]. These findings indicate that the increased ACE levels may account for the shared cognitive decline between SCZ and AD, and that drugs such as ACE inhibitors or angiotensin receptor antagonists (ARAs) may benefit for cognitive improvement in both disorders [62]. The rs12979278 in MAMSTR was another lead SNP shared by SCZ and AD, and MAMSTR stimulates the expression of Myocyte Enhancer Factor-2C (MEF2C). MEF2C plays a crucial role in neurogenesis and synapse development [67] and has been associated with cognitive decline and risk for neuropsychiatric disorders including SCZ and AD [68,69,70]. These findings indicate that the MEF2C deficiency may also account for the shared cognitive decline between SCZ and AD, informing the therapeutic potential of targeting MEF2C to improve cognitive function in both disorders. However, the mechanisms underlying the associations between SNPs and psychotic symptoms shared by SCZ and AD remain largely unknown, deserving further investigation.

We also found that SCZ and AD showed a shared colocalization with expression of SAFB-like transcription modulator (SLTM) in the putamen at the shared locus with rs6494036 (chr15) as the lead SNP. SLTM induces apoptosis [71] and regulates GLI factor binding to chromatin and contributes to the transcriptional outcomes of Sonic hedgehog (SHH) signaling [71]. The SHH pathway affects neurogenesis and neural patterning during the development of the central nervous system, and the dysregulation of SHH signaling in the brain contributes to neuropsychiatric disorders including SCZ and AD [72,73,74]. These findings indicate that the SLTM expression in the putamen may account for the shared symptoms between SCZ and AD, and targeting SLTM expression and its downstream SHH pathway are potentially effective for both disorders.

Several limitations should be considered when interpreting our findings. First, the potential confounders related to population stratification may influence our results. Second, the use of summary-level data rather than individual-level data may reduce the precise of our analyses. Third, the sample sizes of SCZ-GWAS and AD-GWAS are far from enough to discover all SNPs associated with each disorder, which may limit the ability to identify SNPs shared by SCZ and AD. Lastly, the SCZ-GWAS and AD-GWAS summary statistics were derived from meta-analysis of many original studies, and thus the inconsistency in diagnostic criteria for disorders across studies may influence our findings.

In sum, our study provides novel insight into the pleiotropic enrichment between SCZ and AD, discovering eight shared genetic architectures with a concordant allelic effect direction. Future studies should elucidate whether and how the shared genetic variants causally lead to the comorbidity and shared symptoms between SCZ and AD. Through targeting the shared pathophysiological pathways, new therapeutic strategies should be developed to improve clinical outcomes of both disorders, especially for AD patients with psychotic symptoms.