Genetic insights into drug targets for sporadic Creutzfeldt-Jakob disease: Integrative multi-omics analysis

Objective: Sporadic Creutzfeldt-Jakob disease (sCJD) is a fatal rapidly progressive neurodegenerative disorder with no effective therapeutic interventions. We aimed to identify potential genetically-supported drug targets for sCJD by integrating multi-omics data. Methods: Multi-omics-wide association studies, Mendelian randomization, and colocalization analyses were employed to explore potential therapeutic targets using expression, single-cell expression, DNA methylation, and protein quantitative trait locus data from blood and brain tissues. Outcome data was from a case-control genome-wide association study, which included 4110 sCJD patients and 13,569 controls. Further investigations encompassed druggability, potential side effects, and associated biological pathways of the identified targets. Results: Integrative multi-omics analysis identified 23 potential therapeutic targets for sCJD, with five targets ( STX6, XYLT2, PDIA4, FUCA2, KIAA1614 ) having higher levels of evidence. One target ( XYLT2 ) shows promise for repurposing, two targets ( XYLT2 , PDIA4 ) are druggable, and three ( STX6 , KIAA1614 , and FUCA2 ) targets represent potential future breakthrough points. The expression level of STX6 and XYLT2 in neurons and oligo-dendrocytes was closely associated with an increased risk of sCJD. Brain regions with high expression of STX6 or causal links to sCJD were often those areas commonly affected by sCJD. Conclusions: Our study identified five potential therapeutic targets for sCJD. Further investigations are warranted to elucidate the mechanisms underlying the new targets for developing disease therapies or initiate clinical trials.


Introduction
Human prion diseases are a group of fatal, rapidly progressive neurodegenerative disorders.The pathogenesis involves misfolding cellular prion protein (PrP C ) into its abnormal conformers (PrP Sc ), which aggregate and spread in the brain, ultimately leading to neuronal loss and spongiform degeneration (Collinge, 2001).The most common form of human prion disease is sporadic Creutzfeldt-Jakob disease (sCJD), which accounts for 80% of all human prion diseases (Uttley et al., 2020).Patients with sCJD may show high signal on diffusion-weighted magnetic resonance imaging in the basal ganglia (especially the caudate nucleus and putamen) and cortex (such as the temporal, parietal, and occipital lobes) and a significant elevation of 14-3-3 protein in the cerebrospinal fluid (CSF) (Zerr et al., 2009).The average duration from diagnosis to death in sCJD ranges from 5 to 6 months, though it can vary from weeks to several years (Llorens et al., 2020).However, due to the unknown etiology of sCJD, there are currently no effective strategies for its prevention and treatment.
One strategy that has been proposed to impede disease progression is the PrP-targeted drug strategy.Several repurposed drugs, including doxycycline, quinacrine, and flupirtine, have been evaluated to treat patients with CJD in clinical trials (Appleby et al., 2019;Forloni et al., 2019).However, none of these treatments have demonstrated significant effects on disease progression or mortality rates.Promisingly, a new drug candidate, the PRN100 antibody designed to bind and stabilize PrP C , didn't halt the progression of the disease in six patients, but it appears to have stabilized in three patients for periods (Mead et al., 2022).Additionally, antisense oligonucleotides designed to target the human PRNP RNA sequence are now in the clinical trial stage (NCT06153966).Due to the unknown etiology of sCJD and its nonassociation with PRNP mutations, the development of therapeutic strategies targeting the upstream aspects of PrP or alternative targets apart from PrP is highly necessary.Notably, researchers have explored non-PrP-targeted drug strategy, such as targeting the SERPINA3 protein, which is upregulated in the human frontal cortex affected by prion diseases (Colini Baldeschi et al., 2022).
Large biobank-based genetic studies provide a valuable platform for identifying and validating numerous new drug targets that are more likely to succeed in clinical trials (Nelson et al., 2015;Trajanoska et al., 2023).Potential therapeutic targets have been identified in several diseases such as Alzheimer's disease, Parkinson's disease, and neuropsychiatric disorders, by integrating data from expression, single-cell expression, DNA methylation and protein quantitative trait loci (eQTLs, sc-eQTLs, mQTLs, and pQTLs) in blood and brain using Mendelian randomization (MR) analysis and transcriptome-, epigenome-, and proteome-wide association studies (TWAS, EWAS, and PWAS) (Ou et al., 2021;Liu et al., 2021;Ge et al., 2023a;Storm et al., 2021).TWAS/ EWAS/PWAS is a powerful approach that integrates the gene expression, DNA methylation, and protein abundance reference panel with genome-wide associations from large-scale GWASs to predict and prioritize cis-regulated targets associated with human diseases (Ou et al., 2021;Liu et al., 2021).The performance is primarily influenced by the accuracy of single nucleotide polymorphisms (SNPs)-based prediction model.By leveraging genetic variants (typically SNPs) as instrumental variables (IVs), MR analysis provides a unique advantage in causal inference compared to associative analyses.Genetic alleles are randomly distributed during conception, thereby reducing susceptibility to confounding factors that could distort observational outcomes.Additionally, the absence of disease influence on genotype helps prevent reverse causation bias.MR is increasingly employed to explore diverse Fig. 1.Overview of the study design.Multi-omics wide association analysis was conducted on brain and blood tissues from various studies, followed by Two-sample Mendelian Randomization and colonization analysis.Based on these results, the identified targets were categorized into four tiers.Finally, we performed further analyses on the identified targets, including cell-type and gene expression pattern analysis, enrichment pathway and protein interaction network analysis, side-effect prediction, and druggability analysis.treatment strategies, identifying causal associations between risk factors/molecular target and disease to guide therapeutic targeting effectively (Ge et al., 2023a;Storm et al., 2021).Although the methods differ, both associative analysis and MR analysis are used for drug target identification.Integrating both approaches facilitates a comprehensive identification of potential drug targets and enhances mutual validation.Additionally, identifying potential drug targets from the entirety of genetic expression, including transcription, epigenetic modifications, and protein expression, may also reveal the underlying upstream pathogenesis.
In this study, we employed MWAS, MR, and colocalization analyses to explore potential therapeutic targets for sCJD using multi-omics data from both blood and brain tissues.Furthermore, we conducted additional investigations into the druggability and possible side effects, brain gene pattern and cell-type specificity analysis, and biological pathways associated with the identified potential targets (Fig. 1).

Outcome data
We used the largest publicly available summary statistics data from a 2020 case-control genome-wide association study (GWAS) on sCJD.The data included 4110 diagnosed probable or definite sCJD patients and 13,569 controls from countries with populations of predominantly European ancestries (Jones et al., 2020).

Exposure data
Multi-omics quantitative trait locus (xQTLs) data was obtained from multiple studies, including blood samples and brain tissues from various brain regions.The xQTLs data encompass cis-expression, cis-single-cell expression, cis-DNA methylation, and cis-protein QTLs (cis-eQTLs, cis-sc-eQTLs, cis-mQTLs, and cis-pQTLs).We obtained the brain cis-eQTL data from the BrainMeta (Qi et al., 2022), Wingo 2023(Wingo et al., 2023), PsychENCODE (Wang et al., 2018), and GTEx study (GTEx Consortium, 2020); the blood cis-eQTL data from the eQTLGen study and GTEx study; the brain cis-mQTL data from the Brain-mMeta mQTL study; the blood cis-mQTL data from a meta-analysis of the Brisbane Systems Genetics Study and the Lothian Birth Cohorts (Wu et al., 2018); the brain cis-pQTL data from the Religious Orders Study and Memory and Aging Project (ROSMAP) study, the Banner Sun Health Research Institute study, Wingo 2023, and Yang 2021study (Yang et al., 2021); and the blood cis-pQTL data from the UK Biobank (Sun et al., 2023), Zheng 2020 (Zheng et al., 2020), and Yang 2021 study.
To explore potential therapeutic targets at the cellular level, we included single-cell eQTL data.Cell-type-specific cis-eQTL data were obtained from two studies.Bryois 2022 (Bryois et al., 2022) included data from 192 individuals across eight brain cell types derived from the prefrontal cortex, temporal cortex, and white matter.These data were collected from five different datasets, and some of the patients included in the study were from the ROSMAP cohort (approximately 47.4%).In another study (Fujita et al., 2022), single-nucleus RNA sequencing (snRNAseq) data from the dorsolateral prefrontal cortex of 424 individuals from the ROSMAP cohort were used to generate cell-typespecific cis-eQTLs.This analysis identified cis-eQTLs specific to seven cell types and 81 cell subtypes, with 64 of these subtypes publicly available.
Table S1 provides detailed descriptions and sources of the xQTL and GWAS data used in this study.The original publications provide complete information about the data.

Multi-omics wide association analysis (MWAS)
We performed transcriptome-wide association study (TWAS), proteome-wide association study (PWAS), and epigenome-wide association study (EWAS) using the FUSION method with default settings and parameters (Wingo et al., 2021).The external expression reference panel was constructed based on mRNA expression levels for TWAS (eQTL), DNA methylation levels for EWAS (mQTL), and protein abundances for PWAS (pQTL).The specific information regarding the reference panels can be found in Table S1.TWAS, EWAS, and PWAS results were corrected using the Bonferroni method (0.05/the number of genes, CpG, or proteins included in the analysis).The FUSION code for MWAS can be available from http://gusevlab.org/projects/fusion/.

Mendelian randomization and colocalization
The validity of an MR study relies on three essential assumptions: (1) a strong correlation between genetic variants and the exposure; (2) no association between genetic variants and confounding factors; and (3) genetic variants solely influence the outcome through the exposure and not through alternative pathways.We performed the following steps to select instruments according to previous studies (Zheng et al., 2020).First, candidate IVs for xQTL were chosen at the genome-wide significance threshold of P < 5 × 10 − 8 (Storm et al., 2021;Zheng et al., 2020).Next, SNPs associated with each feature were clumped at an R 2 < 0.001, using a clumping window of 10,000 kb.This clumping process was performed using European samples from the 1000 Genomes Project.To reduce the confounding effects caused by indirect pathways, we only selected cis-xQTLs for MR analysis.Cis-xQTLs are more likely to directly influence gene expression or protein levels through their effects on transcription or translation.Additionally, we removed SNPs located within the major histocompatibility complex region (chromosome 6, 26-34 Mb).All genetic instruments used in this study can be available at https://doi.org/10.5281/zenodo.10635391.To ensure the correct causal direction, after harmonizing the exposures and outcome alleles, we employed Steiger filtering to eliminate SNPs that explained a greater proportion of variation in the outcome (sCJD) than in the exposure (xQTL).
The primary analyses estimated effects using the Wald ratio when only one SNP was available.In cases where two or more genetic instruments were available, the main analysis primarily used the inverse variance weighted (IVW) method.Additionally, several complementary sensitivity analyses were conducted using the weighted median, weighted mode, and MR-Egger regression with bootstrapped standard error method.Furthermore, heterogeneity analysis based on Cochran's Q method and pleiotropy analysis based on the MR-Egger regression were performed.
We conducted separate false discovery rate (FDR) corrections for different MR analyses between the outcome and exposures in different studies.We applied FDR correction by defining significance at an FDR P value of 0.05 to account for multiple testing (Storm et al., 2021).
To strengthen the evidence for causal relationships, we performed Bayesian colocalization analysis using the "coloc" function in the R package.Specifically, we calculated the posterior probability of the same causal variant being shared between the xQTL and sCJD for significant MR results.The Bayesian colocalization analysis was conducted within a 1-Mb window on either side of the sentinel variant.Default prior probabilities were used, with p1 = 1E− 4, p2 = 1E− 4, and p12 = 1E− 5.The colocalization analysis yielded posterior probabilities corresponding to five hypotheses.Of particular interest is the posterior probability of hypothesis 4 (PPH4), which indicates an association between the sCJD and the gene, DNA methylation, or protein with a shared causal variant.We defined PPH4 > 50% but <80% as weak colocalization and PPH4 > 80% as strong colocalization (Ge et al., 2023b).
Afterwards, we classified the targets obtained from the MWAS and MR into 4 evidence tiers, denoted as Tier 1, Tier 2, Tier 3, and Tier 4. The classification was based on whether they passed MWAS, MR, colocalization analysis, and whether they were replicated.The specific criteria and overview of the study design can be found in Fig. 1.
MR and colocalization analysis were completed using R software

Gene expression pattern and cell-type specificity analysis
We used the JuGEx tool and the Allen Human Brain Atlas to investigate gene expression patterns of identified risk genes (Tier 1-2) across various brain regions and compare gene expression differences between two specific regions (Bludau et al., 2018).The regions of interest were determined using the Automated Anatomical Labeling atlas (Rolls et al., 2015).The Gene Expression distribution image was obtained by analyzing microarray data from the Allen Human Brain Atlas.To visualize the enriched regions of gene expression, the mRNA expression data of the genes of interest was downloaded from www.meduniwien.ac.at/n euroimaging/mRNA.htmland was applied a threshold to highlight the top 1% expression.
To further explore whether the identified risk genes (Tier 1-2) are enriched in different cell types across various brain regions, the CELLEX (CELL-type EXpression-specificity) approach was employed to calculate cell-type expression specificity (ES) (Zhang et al., 2022).We selected snRNAseq data from brain regions commonly affected in sCJD, including multiple cortical areas, the caudate nucleus, putamen, thalamus, Fig. 2. Manhattan plots for the sCJD MWASs (TWAS, EWAS and TWAS) in the human brain and blood tissues.Significant genes were labeled and highlighted in red.The black line represents the Bonferroni correction threshold of 0.05/the number of genes in the analysis.(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)cerebellum, and spinal cord.The snRNAseq data was obtained from the Human Brain Cell Atlas study and profiled using either the SMART-Seq v4 or the 10× Genomics Chromium Single Cell 3′ v3 RNA-seq platforms (Jorstad et al., 2023).The sources of the RNA-seq data can be found in Table S1.

Enrichment analysis and protein interaction network
Functional Enrichment analysis was conducted using Metascape, an  automated meta-analysis tool to understand biological pathways (Zhou et al., 2019).The identified sCJD-related genes (Tier 1-4) and the risk loci-associated genes (PRNP and GAL3ST1) identified through GWAS were used as the input gene sets, and the enrichment analysis were carried out with GO Biological Processes and KEGG Pathway (Jones et al., 2020).All genes in the genome have been used as the enrichment background, and P values were calculated based on the cumulative hypergeometric distribution with the Benjamini-Hochberg FDR correction for multiple testing (P < 0.01).Protein-protein interaction (PPI) enrichment analysis was performed using the following databases based on Metascape: STRING6, BioGrid7, OmniPath8, InWeb_IM9.The Molecular Complex Detection (MCODE) algorithm was applied to identify densely connected network components on the PPI network.

Phenome-wide association analysis and druggability evaluation
To explore possible side effects of sCJD-related genes (Tier 1-4), the AstraZeneca PheWAS Portal was used to conduct a gene-level phenomewide association study (PheWAS) (Cao et al., 2023).We employed the default threshold values provided by the research, considering a P-value <1 × 10 − 6 as suggestive and a P-value <1 × 10 − 8 as significant.Finally, we further evaluated the identified genes as potential therapeutic targets by using the DGIdb database to search for gene-drug interactions (Freshour et al., 2021).

MWAS identifies ten genes that are associated with sCJD
We first performed tissue-specific TWASs by using five reference panel data (three for the brain and two for blood).Eight genes were identified as associated with sCJD in brain tissues.Specifically, STX6, RASSF2, TBC1D10A, ACDB6, and KIAA1614 were found to be associated with an increased risk of sCJD, while RETSAT, DAG1, and PRSS53 were associated with a decreased risk of sCJD.Only STX6 showed consistent associations across multiple studies and brain regions, including the cortex, striatum, and spinal cord.PWASs revealed that both STX6 and PDIA4 were associated with an increased risk of sCJD.However, only the association of STX6 was replicated in two studies.Notably, STX6 was the only target that consistently showed positive associations in both TWASs and PWASs studies.In the blood analyses, TWASs identified only one gene (ARIH2) associated with an increased risk of sCJD in one study, while no genes related to sCJD were identified in EWASs and PWASs (Fig. 2; Table 1).
In the MR analysis of sc-eQTL, we discovered that STX6 in excitatory neurons and oligodendrocytes was associated with an increased risk of sCJD.To note, the STX6 in oligodendrocytes showed consistent findings across two studies and cell subcluster analyses and exhibited strong evidence of colocalization with sCJD (PPH4 > 0.9).SERPINB6 located on microglia was also found to be associated with an increased risk of sCJD in two studies and show strong evidence of colocalization with sCJD (PPH4 > 0.8) in one study.Additionally, XYLT2 was identified to be associated with an increased risk of sCJD, although it was found in different cell types across two studies.PDIA4 in excitatory neurons were also associated with an increased risk of sCJD, but this finding was not replicated (Fig. 3; Tables S2 and S3).
A total of four CpG sites was found to have a significant effect on the risk of sCJD.Among them, ch.1.3532527F(KIAA1614) from the brain and cg07977490 and cg07464859 (RP11.1021N.1) from blood were associated with a decreased risk of sCJD, and showed strong evidence of colocalization with sCJD (PPH4 > 0.9).However, cg02413285 (PRNP) from blood was associated with an increased risk of sCJD but showed no evidence of colocalization (PPH4 = 0.18; Table 2; Tables S2 and S3).
After applying FDR correction, 8 proteins were found to be causally associated with the risk of sCJD.These proteins include STX6, PDIA4, FXN, TEX264, FUCA2, MANF, GPNMB from the brain, and MMP1 from blood.All of these proteins showed evidence of colocalization.Specifically, STX6, PDIA4, TEX264, FUCA2, GPNMB, and MMP1 were associated with an increased risk of sCJD, while FXN and MANF were associated with a decreased risk of sCJD.Of note, PDIA4 and FUCA2 were replicated in two studies and showed evidence of colocalization with sCJD (PPH4 > 0.5; Table 2; Tables S2 and S3).
Among these significant associations, the IVW method was only performed as the primary analysis method in the FXN protein (2 SNPs).No heterogeneity was detected (P-value of Cochran's Q value = 0.91).
In summary, combining MWAS, MR, and colocalization analyses, a total of 23 genes (20 in brain and 3 in blood) was identified as being associated with sCJD.These genes were further categorized based on the strength of evidence.STX6 was categorized as Tier 1 (passing MWAS, MR, and PPH4 > 0.8 in ≥ two studies), while PDIA4, KIAA1614, XYLT2, and FUCA2 were categorized as Tier 2 (passing MWAS, MR, and PPH4 > 0.8 in one study; passing MR and PPH4 > 0.8 in ≥ two studies).Detailed results can be found in Table 3.

Cell-type and gene expression pattern in the brain
We investigated the brain gene expression patterns of five genes mostly associated with sCJD (Tier 1 and Tier 2) and conducted cell-type specificity analysis across different brain regions.We found that STX6 was predominantly expressed in the striatum, hippocampus, and parieto-occipital lobe, while PDIA4, XYLT2, and FUCA2 were highly expressed in the thalamus, and KIAA1614 showed high expression in the cerebellum (Fig. 4A, B).The comparative P-values for gene expression differences between these brain regions can be found in Table S4.These enriched regions corresponded to commonly affected areas in sCJD.Further analysis revealed that STX6 was enriched in glutamatergic neurons in the cortex.In the lingual gyrus and head of the caudate nucleus, STX6 was also enriched in endothelial cells and fibroblasts.In the posterior nuclear complex of the thalamus, PDIA4 was primarily enriched in oligodendrocytes and pericytes.XYLT2 showed enrichment in neurons and leukocytes, while FUCA2 was enriched in neurons and choroid plexus epithelial cells.In the lateral hemisphere of the cerebellum, KIAA1614 was mainly enriched in astrocytes (Fig. 4C).

PPI network and enrichment result
The protein-protein interaction analysis revealed two interactive networks, with one centered around PRNP and the other centered around STX6.Additionally, MCODE analysis identified a protein functional module shown in Fig. 4D.In this module, STX6, KIAA1614, and MR1 are all located adjacently on chromosome 1, but GO enrichment analysis did not reveal any functional enrichment.The significant pathways in the GO biological processes category were mainly associated with protein maturation, protein phosphorylation, regulation of proteolysis, and the regulation of the MAPK cascade (Fig. 4E; Table S5).

Safety and druggability
We performed gene-level phenome-wide association studies (phe-WAS) on the identified sCJD-associated genes using the AstraZeneca PheWAS Portal database to evaluate their potential safety as therapeutic targets.Ultimately, we identified pheWAS results for 18 genes.Among them, MMP1 was found to be associated with MMP3, MMP8, and MMP12 proteins.RETSAT was associated with CAPG and TGOLN2 proteins, and FUCA2 was found to be associated with FUCA1 (Tables S6  and S7).No gene-disease phenotype associations were discovered.To assess the druggability of the identified genes, we conducted drug-gene interaction analysis.DGIdb identified six potential therapeutic targets, including XYLT2, MR1, GPNMB, FXN, PRSS53, and MMP1.Among these targets, five antineoplastic agents were identified, including carboplatin and gemcitabine for targeting XYLT2, Glembatumumab vedotin for targeting GPNMB and Leuprorelin acetate and Marimastat for targeting MMP1 (Table S8).

Discussion
In this study, we explored potential therapeutic targets for sCJD by integrating data from human brain and blood xQTLs.We identified and prioritized 23 targets (Table 3) associated with sCJD risk by integrating MWAS, MR, and colocalization analyses.Five targets (STX6, XYLT2, PDIA4, FUCA2, KIAA1614) had higher levels of evidence, with STX6 being the highest-ranked target.STX6, PDIA4, and KIAA1614 have been validated in both MWAS analysis and MR analysis.Importantly, our study provides novel evidence at the cellular level based on causal inference analysis, confirming STX6 and XYLT2 as risk genes for sCJD.We also revealed the differential expressions of these targets across brain regions and cell types.Druggability evaluation revealed antineoplastic drugs (carboplatin and gemcitabine) targeting XYLT2 may hold the potentials of being repurposed as therapeutic drugs for sCJD.
Our study provided the evidence of the causal role of STX6 expression level and protein abundances in elevating sCJD risk.Furthermore, we discovered that the increased expression of STX6 only in the cortex, hippocampus and striatum was associated with an increased risk of sCJD.The brain regions commonly affected by sCJD include the cortex, striatum, and thalamus (Venkatraghavan et al., 2023).Interestingly, the brain regions with high expression of the STX6 in healthy individuals match well with the commonly affected regions in sCJD (Fig. 4B).Consistent with our findings, Jones et al. (Jones et al., 2020) demonstrated a correlation between risk variants in STX6 and an increased risk of sCJD while showing no correlation with other forms of CJD, including the acquired form such as variant CJD, iatrogenic CJD, and kuru disease.They employed eCAVIAR13 to identify a strong correlation between sCJD risk and increased expression of STX6 mRNA in the caudate and putamen regions (Jones et al., 2020).Additionally, our study expanded the evidence on the cell level.We found that STX6 was predominantly expressed in neurons in the cortex and striatum, particularly in excitatory neurons (Fig. 4A-C).Notably, we observed that increased expression of STX6 only in excitatory neurons and oligodendrocytes was associated with an elevated risk of sCJD.Consistent with our findings, previous studies performed immunohistology of frontal cortex in sCJD and controls and revealed syntaxin 6 is expressed in neurons and oligodendrocytes (Jones et al., 2020).A study observed the deposition of PrP Sc in oligodendrocytes within the white matter tissue of sCJD (Gelpi et al., 2021).Syntaxin 6 primarily involves vesicle fusion during retrograde transport between early endosomes and the trans-Golgi network.
Recent studies have shed light on the capacity of syntaxin 6 to mediate the release of tau from neuronal cells and facilitate the propagation of pathological tau aggregates (Lee et al., 2021).Interestingly, the CSF total (t)-Tau has been considered another critical biomarker for diagnosing CJD in addition to the 14-3-3 protein.It is unknown whether syntaxin 6, expressed in brain regions abundant in STX6, contributes to the release and spreading of PrP Sc in neurons or oligodendrocytes.Although there are currently no drugs targeting STX6, pheWAS results indicate that the development of drugs targeting STX6 may be safe.Furthermore, a recent animal study showed that mice with genetic depletion of STX6 exhibited prolonged disease incubation periods after inoculation with ME7 prions, suggesting the effectiveness of targeting STX6 (Jones et al., 2023).
Our study revealed a causal association between increased XYLT2 gene expression at the cellular level and an elevated risk of sCJD.Notably, the approved drugs carboplatin and gemcitabine interact with XYLT2.A study showed that platinum complexes cisplatin or carboplatin can inhibit the aggregation of PrP106-126, which shares some physicochemical and biological properties with PrP Sc (Wang et al., 2014).It is essential to investigate potential overlapping mechanisms between sCJD and tumors, as it would be highly promising if drugs commonly employed in tumor treatment, such as carboplatin, could be repurposed to manage sCJD.However, caution is needed in interpreting this result.On one hand, our findings merely suggest a simple association between the drug and the gene; on the other hand, it is crucial to prioritize the safety of these drugs, considering their toxic effects and potential adverse reactions in patients.XYLT2 encodes the enzyme xylosyltransferase 2 (XylT2), which plays a role in synthesizing glycosaminoglycan (GAG) chains in sulfated proteoglycans (Pönighaus et al., 2007).Decreased expression of XylT2 leads to reduced synthesis of heparan sulfate proteoglycans (HSPGs) (Kim et al., 2018).In the brain tissue of individuals with CJD, HSPGs accumulation has been observed in PrP amyloid plaques and the neurons and astrocytes adjacent to them (Snow et al., 1990).XYLT2 is predominantly expressed in astrocytes and neurons (Fig. 4C).Additionally, GAGs have been found in PrP Sc deposits and demonstrated to enhance the misfolding of PrP in vitro (Ellett et al., 2015).We hypothesize that high XylT2 expression would increase HSPG synthesis and GAG levels, affecting prion misfolding and plaque formation.Further experimental studies are warranted to explore the role of XYLT2 in the pathogenesis of CJD.
Besides the relationship between XYLT2 and sCJD suggesting the therapeutic potential of antineoplastic drugs for sCJD, our study also provides additional evidence.We have revealed for the first time that FUCA2 was associated with the elevated sCJD risk.FUCA2 (alpha-Lfucosidase 2) is a gene that codes for a secreted non-lysosomal enzyme.Previous studies have observed that FUCA2 is overexpressed in 24 types of tumors, suggesting it is a potential target for tumor therapy (Zhong et al., 2021).Additionally, besides XYLT2, GPNMB and MMP1, which are associated with an increased risk of sCJD, they also interact with antineoplastic drugs (Table S8).Many previous studies also support the relationship between tumors and sCJD.Interestingly, the PRNP gene and PrP C have a close relationship with tumors.PrP C is a potential therapeutic target for tumors (Costa et al., 2016).Furthermore, elevated levels of SERPINA3 have been consistently observed in individuals diagnosed with CJD, while its expression has also been found to be increased in various tumors (de Mezer et al., 2023).In summary, conducting clinical trials using antineoplastic drugs to treat CJD in the future is necessary.However, prior to that, a thorough assessment of the potential risks posed by these drugs to CJD patients must be carried out.PDIA4 (protein disulfide isomerase family A member 4) is a novel endoplasmic reticulum chaperone.It is involved in the pathogenesis of pancreatic β-cells in the periphery (Tseng et al., 2023).Our study revealed, for the first time, that the gene expression and protein levels of PDIA4 in brain tissue are associated with an increased risk of sCJD.A study explored the global gene expression profile in samples from the frontal cortex of 10 sCJD cases and 10 controls.Results unveiled several genes with significant expression variations in sCJD.Notably, five genes (PDIA4, SERPINB6, DAG1, GPNMB, RASSF2) showed upregulation and one gene (NCKIPSD) showed downregulation, aligning with our identified therapeutic targets (Bartoletti-Stella et al., 2019).A PDIA4 antagonist has been developed and is in the preclinical stages for treating diabetes (Tseng et al., 2023).Exploring the effectiveness of targeting PDIA4 in brain tissue to treat sCJD would be worth investigating in future studies.
There are some limitations to our study.First, the unavailability of additional sCJD GWAS hinders the validation of our findings.However, we included the most extensive and publicly available sCJD GWAS data.We used multiple study-specific exposure datasets to reduce the risk of false positives.We employed various methods, including MWAS, MR, and colocalization, to confirm our findings, with some results consistent with previous studies.Second, the current analysis was predominantly restricted to European populations, and validation of our findings in other ancestries is needed.Third, the stringent statistical analysis and significance thresholds may filter out some potential disease-related targets, resulting in some exposures without eligible instruments not being included in the final analysis.The PRNP-related instruments were retained only in the blood eQTL of eQTLGen and the sc-eQTL of ROSMAP.
In conclusion, our study explored potential genetically-supported drug targets for sCJD by integrating multi-omics data.We identified five targets with higher levels of evidence.Among them, one target (XYLT2) shows promise for repurposing, two targets (XYLT2, PDIA4) are druggable, and three (STX6, KIAA1614, and FUCA2) targets represent potential future breakthrough points.The evidence of STX6 as the risk gene of sCJD is supported by cellular-level data and co-localization in disease-specific brain regions.Future experimental studies are necessary to evaluate the underlying biological mechanisms of the identified targets and conduct clinical trials on targets that could be repurposed for the management of sCJD.

Fig. 3 .
Fig. 3. MR results of single-cell eQTL on the risk of sCJD.Significant genes (FDR-corrected P-value <0.05) are labeled and highlighted in red.A) results from the ROSMAP cohort, B) results from the Bryois2022 study, and C) results from cell subtypes within the ROSMAP cohort.(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 4 .
Fig. 4. Follow-up analyses.A) Expression levels of identified sCJD-related genes in different brain regions.B) Visualization of gene expression distribution for STX6, PDIA4, and KIAA1614 based on the Allen Human Brain Atlas (thresholded at the highest 1% expression).C) Cell-type specificity enrichment of identified sCJDrelated genes in different cell types across various brain regions.D) Two protein-protein interaction (PPI) networks were identified using Metascape.MCODE analysis revealed a protein functional module.E) Gene ontology enrichment results.

Table 1
TWAS and PWAS in brain and blood tissues associated with sCJD.

Table 2
MR and colocalization using brain and blood e/m/pQTL.

Table 3
Prioritization of drug targets for sCJD.