Novel insight into the etiology of ischemic stroke gained by integrative transcriptome-wide association study

Junghyun Jung, Zeyun Lu, Adam de Smith, Nicholas Mancuso
doi: https://doi.org/10.1101/2023.03.30.23287918

Abstract

Stroke, characterized by sudden neurological deficits, is the second leading cause of death worldwide. Although genome-wide association studies (GWAS) have successfully identified many genomic regions associated with ischemic stroke (IS), the genes underlying risk and their regulatory mechanisms remain elusive. Here, we integrate a large-scale GWAS (N=1,296,908) for IS together with mRNA, splicing, enhancer RNA (eRNA) and protein expression data (N=11,588) from 50 tissues. We identify 136 genes/eRNA/proteins associated with IS risk across 54 independent genomic regions and find IS risk is most enriched for eQTLs in arterial and brain-related tissues. Focusing on IS-relevant tissues, we prioritize 9 genes/proteins using probabilistic fine-mapping TWAS analyses. In addition, we discover that blood cell traits, particularly reticulocyte cells, have shared genetic contributions with IS using TWAS-based pheWAS and genetic correlation analysis. Lastly, we integrate our findings with a large-scale pharmacological database and identify a secondary bile acid, deoxycholic acid, as a potential therapeutic component. Our work highlights IS risk genes/splicing-sites/enhancer activity/proteins with their phenotypic consequences using relevant tissues as well as identify potential therapeutic candidates for IS.

Introduction

Stroke is a complex disease resulting from an interruption of blood flow to the brain [1]. A common type of stroke is ischemic stroke (IS), which is caused by cerebral infarction [2]. Diabetes, obesity, hypertension, and coronary artery disease are well-known risk factors for stroke [3–6], but the pathogenesis of IS is still largely unknown. Although genome-wide association studies (GWAS) have successfully identified genomic regions associated with IS outcomes, the genes underlying IS risk and their regulatory mechanisms remain elusive as the majority of associated variants are non-coding in nature [7,8].

Recently, the transcriptome-wide association study (TWAS) approach attempts to mitigate this gap in understanding by integrating GWAS associations together with molecular quantitative trait loci (molQTL) data [9,10]. Previous works have leveraged TWAS to identify candidate susceptibility genes for IS risk, however these analyses have three primary limitations. First, previous analyses were limited to integration of molQTLs measured in whole blood, adipose, and brain tissues [11,12], which may miss disease mechanisms in less understood or unknown disease-relevant tissues [13–15]. Second, prior works focused on integration of expression QTL (eQTL) and protein QTL (pQTL) [12,16], which may miss independent regulatory mechanisms important for IS risk. For example, an essential mechanism of gene regulation and a significant factor in genetic risk of disease is the genetic control of alternative splicing (i.e., sQTLs) [17]. Moreover, recent work demonstrated enhancers undergo activity-dependent transcription, resulting in the production of noncoding enhancer RNAs (eRNAs) which serve as a crucial hallmark of enhancer activation [18,19]. Third, while the GAGASTROKE prioritized relevant tissues for IS by leveraging eQTL data, it relied on the GTEx v7 study (N=388), [12], which had a smaller sample size than the European GTEx v8 study (N=588). Lastly, the L1000 Connectivity Map (CMap), a public database for large pharmalogical datasets, provides extensive gene expression profiles of thousands of compounds in various human cell lines [20]. Recent studies have successfully provided novel therapeutic candidates by utilizing TWAS results with the pharmalogical database [21,22].

Here, we integrate large-scale IS GWAS data (N=62,100 cases and 1,234,808 controls) with gene expression, alternative splicing, eRNA and protein abundance data (N=11,588) from 50 different tissues to identify IS susceptibility genes, tissues, and drug targets. We identify 136 genes/splicing sites/eRNA/proteins across 54 genomic regions whose genetically predicted activity is associated with IS risk using a multi-tissue mRNA/splicing/eRNA/protein transcriptome-wide association study (TWAS/spTWAS/eTWAS/PWAS). We leverage TWAS results to identify tissues relevant for IS risk and find arterial and brain most enriched for eQTL mediated heritability. Focusing on the IS-relevant tissues, we perform probabilistic fine-mapping analyses of TWAS results to prioritize 9 putative causal genes/proteins. Among them, only 3 genes/proteins (i.e., FOX2, F11, MMP12) were identified in previous GWAS or TWAS studies [7,12,16]. In addition, we conduct a TWAS-based pheWAS analysis to understand the phenotypic consequences of identified IS susceptibility genes, and the identified reticulocyte cell traits are significantly correlated with IS. In addition, we discover that reticulocyte traits have shared genetic contributions with IS using TWAS-based pheWAS and genetic correlation analysis. Lastly, to find therapeutic drug candidates for IS risk, we integrate our TWAS findings with 308,872 pairs of compound and compound-perturbed cell-type specific gene expression alterations from the L1000 Connectivity Map. Using this approach, we detect a secondary bile acid, deoxycholic-acid (DCA), as a potential therapeutic component. Overall, our results shed light on underlying molecular mechanisms and tissue contexts that cause ischemic stroke.

Materials and Methods

Ischemic stroke GWAS summary statistics

IS GWAS summary statistics from GIGASTROKE consortium [16] were downloaded from the GWAS Catalog (GCST90104540) (see URLs). We restricted summary statistics data to ischemic stroke (AIS) results from 1,296,908 individuals of predominantly European ancestry (62,100 cases and 1,234,808 controls). Next, we filtered summary statistics data to exclude SNPs with minor allele frequency (MAF) < 0.01 or any SNPs with strand-ambiguous variants (i.e. A/T or C/G; or vice-versa) using the focus munge tool [23], resulting in 6,335,571 bi-allelic SNPs for downstream analyses.

Reference functional data for predictive models of eQTL/spQTL/EeQLT/pQTL

To perform TWAS and spTWAS, we generated predictive models of gene and splicing expression using individuals from the Genotype-Tissue Expression Project (GTEx) v8 [24] using a modified FUSION script (see URLs). We downloaded genotype, phenotype, and covariate information from European-American subjects in the GTEx v8 study (48 tissues; N□=588). We defined the cis-mapping window as ±500kb around the transcription start site (TSS) after filtering based on minor allele frequency (MAF) < 0.005, Hardy-Weinberg Equilibrium (HWE) < 1 × 10−5. We included the reference set of covariates from the GTEx v8 eQTL/spQTL analyses, which included first five genotype principal components (PCs), 15 hidden covariate derived from Probabilistic Estimation of Expression Residuals (PEER) [25] factors, whole genome sequencing (WGS) platform (HiSeq 2000 or HiSeq X), WGS library preparation protocol (PCR-based or PCR-free), and donor gender. We estimated cis-SNP heritability Embedded Image of each model using REML as implemented in Genome-wide Complex Trait Analysis (GCTA) [26]. We focused on genes and splicing sites with nominally significant estimates of Embedded Image (p-value < 0.01), which resulted in 293,295 total tissue-gene pairs from 27,549 unique genes and 371,441 total splicing sites from 16,871 unique genes. To train and generate predictive models, we fitted LASSO [27], Elastic Net [28], and the Sum of Single Effects model (SuSiE) [29]. The best performing model for each gene/tissue (splice-site/tissue) context was selected by calculating cross-validation prediction accuracy, as implemented in the FUSION pipeline.

To perform enhancer TWAS analysis (eTWAS), we downloaded precalculated FUSION weights for enhancer eQTL (EeQTL) data from two brain tissues, the dorsolateral prefrontal cortex (DLPFC) (N=486) and the anterior cingulate cortex (ACC) (N=402), in CommonMind Consortium (CMC) [19]. Briefly, the LASSO [27], Elastic Net [28] were used to train and generate the eTWAS predictive models for the ACC (4,907 predictive models) and DLPFC (5,715 predictive models). In addition, we also downloaded fitted prediction models of gene expression in both ACC (8,021 predictive models) and DLPFC (8,946 predictive models) tissues. For detailed information of the eTWAS and TWAS predicted models, see the previously described ref [19].

To perform a proteome-wide association study (PWAS), we downloaded fitted prediction models of protein abundance trained using individuals from the INTERVAL (N=3,301) [30] and Atherosclerosis Risk in Communities (ARIC) [31] cohorts (see URLs). In the ARIC cohort, we focused only on the European ancestry dataset (N=7,213). Briefly, to train and generate the PWAS predictive models using FUSION script, the LASSO [27], Elastic Net [28], and SuSiE [29] were used for the INTERVAL (994 predictive models) and the Elastic Net [28] was used for the ARIC (1,305 predictive models). For detailed information of the PWAS predicted models, see the previously described INTERVAL [32] and the ARIC [33].

Transcriptome-wide association study analyses

We performed TWAS, spTWAS, eTWAS, or PWAS analyses using FUSION [9] with the trained GTExv8 [24], CMC [19], or INTERVAL [32] and ARIC [33] models, respectively. European LD reference data from the 1000G project [34] was used for TWAS analysis. We excluded TWAS associations from human leukocyte antigen (HLA) regions due to the complex LD patterns. The significance threshold of TWAS associations was determined using a per-tissue Bonferroni correction (Avg num tests = 6770; see Supplementary Table 1). We then carried out an adaptive permutation test using TWAS test statistics for each tissue panel. For this analysis, 106 maximum number of permutations was used, and the significance threshold was corrected with per-tissue Bonferroni correction.

To provide partial support for findings from our FUSION TWAS analyses, we leveraged S-PrediXcan [35] with pre-trained Multivariate Adaptive Shrinkage in R (MASHR)-based models of PredictDB [36,37]. In GTEx v8, we tested 648,028 and 1,681,295 models of expression and splicing site, respectively.

Colocalization analysis

We performed colocalization analysis to test whether the same causal variants were shared between stroke GWAS and gene/protein expression levels. We used coloc r package [29] together with stroke GWAS summary statistics and marginal molecular QTL (molQTL) results from FUSION prediction models. The evidence of colocalization was defined as PP3 + PP4 ≥ 0.8 and PP4/PP3 ≥ 2 where posterior probability (PP) was obtained from hypothesis H3 and H4. H3 is the PP of GWAS and eQTL signals are associated with different causal variants, and H4 is the PP of GWAS and eQTL signals are associated and share a single causal variant.

Mediated Expression Score regression analyses

To identify tissues relevant for IS risk, we used Mediated Expression Score Regression (MESC) to estimate the proportion of heritability mediated by assayed gene expression levels Embedded Image [38]. For MESC expression scores, the eQTL effect sizes using the SuSiE model [29] and expression cis-heritability using REML were imported from FUSION GTEx v8 weights. Only SNPs from the HapMap3 [39] were kept for this research.

Fine-mapping of TWAS associations

To differentiate between causal and tagging associations at TWAS risk regions, we performed probabilistic fine-mapping of TWAS results to prioritize genes using the tool FOCUS [23]. We generated an eQTL/spQTL/pQTL weight database for FOCUS by importing our trained GTEx v8 weights from FUSION. To determine approximately independent genomic regions, we used LD block architecture GRCh19 provided by Berisa and Pickrell et al. [40] (see URLs).

Phenome-wide association studies and genetic correlation analyses

To understand the phenotypic consequences of identified TWAS/spTWAS/PWAS associations, we performed a phenome-wide association study (pheWAS) for each identified gene using PhenomeXcan [41]. Using the TWAS/spTWAS/PWAS genes, phenotypes from the results of pheWAS analysis were reported based on transcriptome-wide significance (P-value < 2.25 × 10-6). In addition, the PheWAS result was filtered by traits involving at least five TWAS genes.

To determine the genetic relationship between IS and the phenotypes identified from PhenomeXcan, we performed genome-wide genetic correlation analyses using RHOGE [42] with publicly available GWAS summary statistics from the UK Biobank (see URLs). First, TWAS/spTWAS/PWAS analyses were performed using the same FUSION pipeline in the IS analysis. Next, we estimated the genome-wide genetic correlation between the IS and the PhenomeXcan traits derived from TWAS analysis using LD block architecture GRCh38 (see URLs) MacDonald et al. [43]. In addition, we performed SNP-based genetic correlation analysis using LD Score Regression (LDSC) [44] with the GWAS summary statistics data.

Mendelian randomization analysis

To assess causal effects of reticulocyte traits on IS risk, we performed Mendelian Randomization (MR) analysis. To extract instruments for use in MR from the 5 reticulocyte traits including “Reticulocyte percentage (ukb-d-30240_irnt)”, “Reticulocyte count (ukb-d-30250_irnt)”, “Immature reticulocyte fraction (ukb-d-30280_irnt)”, “High light scatter reticulocyte percentage (ukb-d-30290_irnt)”, and “High light scatter reticulocyte count (ukb-d-30300_irnt)”, we leveraged extract_instruments function in TwoSampleMR r package [45]. In addition, For the reverse relationship between reticulocyte traits and IS risk, the fine-mapping results of GIGASTROKE [16] were used as instruments. After harmonizing the effect of a SNP on the between IS and reticulocyte traits to the same allele, the MR analysis was conducted using 5 MR method (“MR Egger”, “Weighted median”, “Inverse variance weighted”, “Simple mode”, and “Weighted mode”).

Drug repositioning analysis

To identify potential drug candidates for IS, we used Trans-Phar [21] by comparing the inverse expression profiles between genetically regulated gene expression from TWAS with compound-induced gene expression profiles from a large-scale pharmacological database. Briefly, we obtained differentially expressed genes (DEGs) data from ref. [21] for each compound-induced gene expression data and used 13 tissue/cell-type categories assigned for 29 GTEx tissues and 77 L1000 Connectivity Map (CMap) cell types [21]. Then, we computed the rank correlations using Spearman’s rank correlation coefficient between the top 10% TWAS genes from each GTEx tissue and each CMAP expression level in the same group of tissues or cell types. In addition, we also calculated the Spearman’s rank correlation coefficient using the 134 TWAS/spTWAS/PWAS genes. Finally, a total of 308,872 P-values were collected for each correlation analysis. The significance threshold of the inverse correlation analysis was determined using a per-tissue or/and per-cell type Bonferroni correction (see Supplementary Table 2).

Results

Multi-tissue TWAS/spTWAS/eTWAS/PWAS identifies genes/proteins associated with IS risk

To identify susceptibility genes and proteins for IS risk, we conducted a multi-tissue TWAS, spTWAS, and PWAS analysis by integrating large-scale IS GWAS summary statistics (N=1,296,908) together with eQTL/spQTL data from GTEx v8 [24] in addition to plasma pQTL data from the INTERVAL [30] and the ARIC [33] studies. Moreover, we performed enhancer TWAS analysis (eTWAS) using brain EeQTL from CMC [19] to investigate the impact of genetic regulation of expressed enhancers on IS risk (see Methods). Focusing on predicted mRNA expression, we tested 309,826 panel-specific expression models across 28,073 genes and identified 268 TWAS associations. Significant associations represented 84 genes across 50 tissues and 41 independent 1 Mb genomic regions based on a per-panel Bonferroni correction threshold (see Fig. 1A and Supplementary Table 3). Next, to shed light on the role of alternative splicing for IS risk, we performed a multi-tissue splicing transcriptome-wide association study (spTWAS; see Methods). Of the 370,815 panel-specific splicing models with 16,848 tested genes, we identified 498 spTWAS associations. Associations represented 69 unique genes across 48 tissues at 29 independent 1Mb genomic regions (see Fig. 1B and Supplementary Table 4). In addition, we carried out an eTWAS to investigate expressed enhancer effects in two brain tissues using 10,622 tissue-specific eTWAS models with 8,397 eRNAs. We identified two enhancers, chr4:186651582:186652095 and chr16:87541308:87541807, in DLPFC and ACC tissues, respectively (see Fig. 1C and Supplementary Table 5). Lastly, we focused on the impact of genetically predicted protein abundance by performing a PWAS using 2,299 panel-specific protein abundance models with 1,556 proteins. We identified 7 PWAS associations across 6 proteins with 6 independent 1 Mb regions (see Fig. 1D and Supplementary Table 6). PWAS results from both pQTL panels showed strong positive correlation (R = 0.79; P-value < 2.2 × 10−16; see Fig. S1), suggesting that the genomic component of protein abundance is well-captured by fitted models. Comparing the analysis results of TWAS and spTWAS, a total of 21 genes were implicated by both approaches, with ABO and F11 representing the genes or proteins identified by all three approaches (see Fig. S2). In addition, we identified 54 genomic regions with a total of 136 susceptibility genes, splicing sites, eRNA, and proteins based on multi-tissue TWAS/spTWAS/PWAS approaches. Among the genomic regions, 20 genomic regions have been implicated in GWAS analysis [16].

Figure 1.
Figure 1. Multi-tissue TWAS/spTWAS/PWAS for IS risk.

Manhattan plots of multi-tissue (A) TWAS, (B) spTWAS, (C) eTWAS, and (D) PWAS. Each point corresponds to a P-value (y-axis) of TWAS/spTWAS/eTWAS/PWAS associations across reference panels and chromosomes (x-axis). The most significant associations of TWAS/spTWAS/eTWAS/PWAS genes among reference panels represent orange, light blue, black and green, respectively. The black dotted lines represent the maximum significant thresholds in this multi-tissue analysis (Supplementary Table 1). The thresholds of TWAS/spTWAS/eTWAS/PWAS indicate 2.23 × 10−5, 2.00 × 10−5, 1.02 × 10−5, and 5.03 × 10−5, respectively. The putative causal genes (PIP > 0.8) in Table 1 represented gene/proteins of TWAS/spTWAS/PWAS indicate red, navy, and dark green, respectively.

To provide additional support for genes identified using TWAS, we re-performed analyses using independent approaches and prediction models. First, we compared TWAS and spTWAS results with those computed by S-PrediXcan [35], an independent method to perform a TWAS analysis and fit predictive models. We found S-PrediXcan results were strongly correlated with FUSION-based TWAS (R = 0.80; P-value < 2.2 × 10−16) and spTWAS (R = 0.82; P-value < 2.2 × 10−16; see Fig. S3). Second, we carried out a co-localization analysis, which reports the posterior evidence that two phenotypes share a causal variant. Here, we identified 51.49% (138 out of 268) TWAS associations exhibited evidence of co-localization between GWAS signals and eQTLs (see Supplementary Table 3). Similarly, among the 498 spTWAS associations, we observed that 271 associations (54.41%) had evidence of co-localization between IS GWAS signal and spQTL association (see Supplementary Table 4). Moreover, we observed that the H4 posterior probability of co-localization, correlated with the TWAS (R = 0.52; P-value < 2.2 × 10−16) and spTWAS (R = 0.54; P-value < 2.2 × 10−16) (see Fig. S4). Repeating this analysis for PWAS signals, we observed ABO, F11, and MMP12 displayed evidence of colocalization between GWAS and pQTL signals (see Supplementary Table 6). Despite eTWAS identifying 2 eTWAS signals, we found little support for colocalization between eTWAS and IS GWAS (see Supplementary Table 5). The GIGASTROKE consortium recently performed IS TWAS analyses based on GTEx v7 prediction models and identified 17 genes across brain, artery, and heart tissues [16]. We sought to assess the stability of these associations by comparing them with our results which leveraged the larger GTEx v8 and INTERVAL/ARIC cohorts. Among the 17 TWAS genes identified in the previous analyses, we found 7 replicated in our TWAS results (average Bonferroni across tissues, P < 7.39 × 10−6). Similarly, we observed a significant correlation between the TWAS effect sizes in the original GIGASTROKE study with those computed using our prediction models (R = 0.73; see Fig. S5). In summary, our findings support a role for genetically regulated expression, splicing, and proteome levels contributing IS risk.

Relevant tissues for IS risk include brain and arterial tissues

Given the broad number of tissues exhibiting TWAS/spTWAS/PWAS associations, we sought to quantify which tissues are most relevant for IS risk for each molecular context. Specifically, we estimated the proportion of heritability mediated by cis-QTL of gene expression levels Embedded Image and alternative splicing levels Embedded Image using mediated expression score regression (MESC; see Methods) [38]. Using this approach, we identified 3 tissues which exhibited Embedded Image estimates greater than 0 at nominal significance (P-value < 0.05). We found “Artery - Aorta” exhibited the greatest Embedded Image value (Embedded Image, s.e= 0.038; P-value = 4.55 × 10−3), followed by “Esophagus_Muscularis” (Embedded Image, s.e= 0.040; P-value = 1.86 × 10−2) and “Brain_Amygdala” (Embedded Image, s.e= 0.040; P-value = 4.55 × 10−2) (see Fig. 2). Results across expression and splicing were consistent MESC across tissues (R = 0.33; P-value = 2.24 × 10−2; see Fig. S6). Previous studies reported that the risk of gastroesophageal reflux disease (GERD) is higher among stroke patients [46] and increases young adult stroke risk [47], suggesting that risk factors shared by individuals with GERD and stroke, such as diet, smoking, obesity, and metabolic syndrome, may account for this association. Our results are supportive of understood IS etiology in which IS occurs due to blood clotting or fatty deposits caused by atherosclerosis that obstruct an artery supplying blood to the brain [48]. Atherosclerosis is commonly thought of as a condition that affects the heart; however, it can also impact arteries located anywhere in the body [49]. In the splicing MESC, “Brain - Nucleus accumbens (basal ganglia)” (Embedded Image and s.e. = 0.056) was a high rank in the splicing MESC (see Fig. S7).

Figure 2.
Figure 2. Proportion of heritability mediated by gene expression levels.

The bar plots correspond the estimated expression MESC Embedded Image in each tissue panel in GTEx v8. Each error bar indicates jackknife standard errors. The red color represents significant tissue panels (P-value < 0.05).

Fine-mapping analysis identifies 9 causal genes/proteins for IS risk

Next, we performed TWAS fine-mapping analysis to identify putative causal genes with multiple signals in the TWAS region prioritizing IS-associated tissues obtained from the MESC analysis [23] (Fig. 2). Among the 84 TWAS significant genes, we identified 5 with posterior inclusion probability (PIP) > 0.8, which we denote as putative causal genes for IS (see Table 1). For example, ANO1 (PIP=0.89; also known as TMEM16A) is suggested to regulate calcium-activated chloride channels and has prior evidence in mouse ischemic stroke models where inhibition of ANO1 expression levels attenuated ischemic brain injury by neurological impairment [50]. We separately performed spTWAS fine-mapping analysis to identify splice variation that may be causally related to IS risk prioritizing brain tissues obtained from splicing MESC analyses. Of the 69 spTWAS significant genes, we identified only MAPKAPK5-AS1 with PIP > 0.8; see Table 1. The mitogen-activated protein kinase (MAPK)-activated protein kinase 5 (APK5), a member of the serine/threonine kinase family, is activated by cellular stress and proinflammatory cytokines [51]. The MAPKAPK5 Antisense RNA 1 (MAPKAPK5-AS1) prevents MAPKAPK5 from being translated into a protein and recent study showed that IS-like pathology was ameliorated by inhibiting the MAPK signaling pathway [52]. Lastly, we performed PWAS fine-mapping to identify protein levels causally relevant to IS risk. Of the 7 PWAS significant associations identified, we found 3 with PIP > 0.8 (see Table 1). Of these 3, F11 and MMP12 genes were identified in a previous TWAS of IS risk [12]. Together, we prioritize 9 putative causal genes/proteins based on relevant tissue for IS risk.

View this table:
Table 1. Putative causal genes/splicing site/proteins from TWAS/spTWAS/PWAS (PIP > 0.8).

*: All genomic locations are GRCh37.

†: FOCUS PIP values are obtained by tissue prioritization using “Artery” or “Brain” tissue.

IS susceptibility genes correlate with reticulocyte cell traits

To understand the phenotypic consequences of identified IS susceptibility genes, we performed a TWAS-based pheWAS analysis (see Methods). Analogous to the TWAS approach, the TWAS-based pheWAS approach has produced more biologically interpretable results by mapping the genome to the phenome using the transcriptome [41]. We identified 71 traits across a broad range of physical measures (47.9%) and blood cell traits (33.8%) (see Fig. S8). In the category of physical measures, an average of 25.4% of genes (11.7/46 genes) detected only by spTWAS analysis were enriched, whereas an average of 34.7% of genes (7.3/21 genes) associated with both TWAS and spTWAS were mostly enriched in blood cell traits (see Supplementary Table 7). Next, we applied genetic correlation analysis to test whether the PheWAS traits shared genetic contributions with IS at a transcriptome- or proteome-wide level (see Methods). The results show that the well-known risk factors for stroke, such as “Diastolic blood pressure. automated reading(4079_irnt)”, “Systolic blood pressure, automated reading (4080_irnt)”, “Non-cancer illness code, self-reported: hypertension (20002_1065)”, “Vascular/heart problems diagnosed by doctor: High blood pressure (6150_4)”, and “Non-cancer illness code, self-reported: high cholesterol (20002_1473)” [4,53,54], were significantly correlated with IS in the physical measures and medical conditions categories (FDR < 0.05) (see Fig. 3 and Supplementary Table 8). Notably, 5 reticulocyte traits, including “Reticulocyte count (30250_irnt)”, “Reticulocyte percentage (30240_irnt)”, “High light scatter reticulocyte percentage(30290_irnt)”, “High light scatter reticulocyte count(30300_irnt), and “Immature reticulocyte fraction (30280_irnt)”, were significantly positively correlated with IS in TWAS and spTWAS levels in the blood cell traits (FDR < 0.05). Reticulocytes are slightly immature red blood cells, and immature reticulocyte fraction levels are associated with acute infection, chronic renal insufficiency, and hematologic diseases [55]. However, we found little evidence for a causal relationship between IS risk and reticulocyte traits (see Fig. S9). Moreover, previous studies showed that high reticulocyte count or reticulocytosis are risk factors for stroke in children with sickle cell disease [56–58]. Additionally, we also used LD-score regression (LDSC) to measure shared genetic contributions between traits at the genome-wide level. We found significant correlations between IS and the reticulocyte traits and the well-known risk factors. Collectively, our results show that vascular traits and blood cell traits, especially reticulocyte cells, have shared genetic contributions with IS.

Figure 3.
Figure 3. Genetic correlation between IS and UKBB phenotypes.

Each point indicates a genetic correlation with standard errors from TWAS/spTWAS/eTWAS/PWAS. Significance levels are represented by red or orange color. We use the following abbreviations ρTWAS, transcriptomic correlation analysis using TWAS; ρspTWAS, transcriptomic correlation analysis using spTWAS; ρeTWAS, transcriptomic correlation analysis using eTWAS; ρPWAS, proteomic correlation analysis using PWAS; ρGWAS, genomic correlation analysis using GWAS. In the figure, we show results indicating significance in at least one TWAS/spTWAS/eTWAS/PWAS analysis.

TransPhar analysis identifies secondary bile acids as potential drug candidates for IS

Next, we sought to find potential drug candidates for IS treatment by evaluating an inverse expression relationship with compound-induced gene expression profiles from a large-scale pharmacological database, L1000 CMap [20]. We computed the rank correlations using Spearman’s rank correlation coefficient between the top 10% TWAS genes (or the 134 TWAS/spTWAS/PWAS genes) from each GTEx tissue (a total of 29 GTEx tissues) and each CMAP expression level in the same group of tissues or cell types. The top 10% TWAS genes of each tissue contained an average of about 22.26% of the TWAS/spTWAS/PWAS genes (mean 29.82 genes) among the 134 TWAS/spTWAS/PWAS genes or proteins (see Fig. S10). We obtained 308,872 relationships from all the tissue/cell type-compound pairs corresponding to each correlation analysis, and the results from the top 10% TWAS genes and the 134 TWAS/spTWAS/PWAS genes were significantly correlated in each tissue (see Fig. S11).

Finally, we found two significant TWAS-compound linkages based on false discovery rate (FDR) correction (see Table 2, Fig. S12, and S13). The two compounds were associated with brain tissue; among them, the mechanism of action has been elucidated for deoxycholic-acid (DCA). The DCA is classified as secondary bile acids, which are produced by gut microbiota [59]. Previous studies showed that other types of secondary bile acids, such as Ursodeoxycholic acid (UDCA) and tauroursodeoxycholic acid (TUDCA), have demonstrated neuroprotective effects in diverse models of neurodegenerative disorders [60–64] as well as stroke [65]. Notably, lower levels of total bile acid excretion is linked to an increased risk of stroke and death [66]. The GIGASTROKE consortium carried out the inverse relationship analysis between TWAS gene and compound expression levels using GTEx v7 data, but there were no significant results for IS [16]. In summary, our results suggest that a negative correlation between the TWAS genes and compound-induced gene expression levels provides the potential drug candidates for IS.

View this table:
Table 2. Potential drug candidates for IS.

Rho (ρ) represents Spearman’s Rank correlation coefficient, and FDR represents the false discovery rate.

Discussion

Stroke is the second leading cause of death worldwide, and the main cause of stroke is ischemic stroke (IS) due to cerebral infarction [2]. In this work, we identify 136 susceptibility genes, splicing sites, eRNA, and proteins spanning 54 genomic regions based on multi-tissue TWAS/spTWAS/PWAS analysis. Among them, 34 genomic regions were not discovered by the original GWAS study. Consequently, we highlight 9 potential causal genes/proteins using probabilistic fine-mapping analysis of the TWAS results with a focus on the IS-relevant tissues, including aorta artery and brain estimated by a total fraction of disease heritability mediated by gene expression levels. Moreover, we discovered that blood cell traits, particularly reticulocyte cells, have shared genetic contributions with IS. Lastly, we detected 2 potential therapeutic compounds for using inverse expression profiles between TWAS results with compound-induced gene expression profiles from a large-scale pharmacological database.

The genes underlying IS risk and their regulatory mechanisms are still unknown, despite the fact that GWAS have effectively identified numerous genomic regions related with IS, because around 90% of the GWAS loci are located in non-coding regions [7,8]. However, the TWAS methodology has produced more biologically interpretable results by integrating GWAS results with mQTL data, such as eQTL, spQTL, EnQTL, and pQTL [9,10,19]. Consistent with the TWAS approach, the PhenomeXcan has allowed us to identify the mediating role of gene expression in complex traits and convert variant-phenotype associations into gene-phenotype associations by providing biological hypotheses [9,36,67].

A notable finding was the significantly positive correlation between IS risk and reticulocyte traits (see Fig. 3). Reticulocytes are slightly immature red blood cells, and higher reticulocyte counts or percentages indicate higher hemolysis, which ultimately results in an increase in the amount of cell-free hemoglobin (CFH) in the blood [68]. The previous studies showed that the hemolysis and the CFH increases inflammation [69] and mediate vascular damage [70]. The findings suggested that a shared causal mechanism influence both the high level of reticulocytes with hemolysis and risk of IS.

After an initial ischemic stroke or transient ischemic attack (TIA) which is a short episode in which temporary blockage of blood flow to the brain, the annual risk of future ischemic stroke is 3 to 5% [71]. The TIA does not leave an impairment, but affected individuals are at increased risk of future ischemic events, especially in the days and weeks immediately after symptoms resolve [72]. The prompt initiation of a coordinated preventive strategy for IS is essential [73].

Recent studies have reported that the probability of success for clinical trials whose therapeutic targets are supported by human genetic information is approximately twice the probability of success for unsupported projects [74,75]. Taking advantage of the rapid growth of genomics fields, the utilization of human genetics information for new therapeutics have been performed in recent years [21,22,76,77]. Importantly, 66% (33 out of 50) of the FDA-approved new drugs were supported by human genetic information in 2021 [78]. In addition to genetic data, recent gene expression datasets have been leveraged for drug repositioning [79,80]. In particular, recent studies have investigated the utility of TWAS in drug discovery or repositioning [21,22]. One primary advantage of the TWAS-based drug repositioning strategy, which integrates large-scale GWAS together with gene expression data, compared with datasets solely comprised of expression is a vastly larger sample size. In general, the sample sizes of GWAS are many orders of magnitude larger (e.g., tens to hundreds of thousands) than those of gene expression datasets (e.g., tens to hundreds). In addition, the TWAS approach enables identifying potential therapeutic compounds that affect multiple tissues (e.g., heart, brain) or cell type-specific contexts by leveraging independent reference eQTL data. On the other hand, the expression data related to the tissue of interest are not easily accessible for many diseases, including IS.

To shed light on susceptibility genes and mechanisms for IS, we used the recently published GWAS data from the GIGASTROKE consortium. A TWAS analysis was also carried out by the GIGASTROKE group, which discovered 17 genes using prediction models based on GTEx v7 [16]. We found that 7/17 genes replicated with in our TWAS investigation in the brain, artery, and heart tissues with a strong correlation of TWAS effect-sizes (R = 0.73; see Fig. S5). Additionally, we identified 119 genes/proteins and demonstrated some strengths of our study. First, our reference panels for TWAS analysis were significantly larger than the GIGASTROKE study by leveraging GTEx v8 datasets [24], which allowed for more robust analysis and generalization of the findings. Second, our study used a diverse and extensive TWAS analysis, which included a multi-tissue mRNA/splicing/eRNA/protein transcriptome-wide association study (TWAS/spTWAS/eTWAS/PWAS). This diversity provides a broader range of biological mechanisms to understand IS etiology.

We note several limitations with our approach. First, our approach relies on using genetically predicted gene expression levels to identify genes whose expression associates with the genetic component of IS risk, which will require further functional studies for downstream validation. We note however, that our results were robust to choose TWAS approach (i.e., FUSION vs PrediXcan), statistical method (i.e., TWAS vs coloc), and eQTL reference panel (i.e., GTEx v7 vs GTEx v8). Second, we investigated the association between IS risk with diverse molecular contexts focusing on genetically predicted gene expression, protein abundance, splicing variation, and enhancer expression across multiple tissues when available. However, recent works have demonstrated improved association and colocalization power when investigating transcription factor binding, chromatin activity, and chromatin accessibility information [81,82], which may be potential mediators for IS risk, missed by our work. Despite these limitations, our results provide valuable insights into underlying molecular mechanisms and drug candidates for IS.

In conclusion, we highlight IS risk genes and proteins using the multi-tissue TWAS/spTWAS/eTWAS/PWAS approach. Moreover, we provided potential drug candidates for preventing IS. We believe that these findings could be used as valuable resources for understanding the underlying mechanisms and designing subsequent functional studies for IS treatment.

Data Availability

All data produced in the present work are contained in the manuscript.

Declaration of interests

The authors have no competing interests.

URLs

The full results of TWAS, spTWAS, eTWAS, PWAS: https://github.com/mancusolab/stroke_twas

Iscemic stroke summary statistics (GIGASTROKE): https://www.ebi.ac.uk/gwas/studies/GCST90104540

GTEx v8 predictive models: http://gusevlab.org/projects/fusion/#gtex-v8-multi-tissue-expression

FUSION script: https://github.com/gusevlab/fusion_twas

INTERVAL predictive models: https://www.mancusolab.com/pwas

ARIC predictive models: http://nilanjanchatterjeelab.org/pwas/

LD blocks (GRCh19): https://bitbucket.org/nygcresearch/ldetect-data/src/master/EUR/

LD blocks (GRCh38): https://github.com/jmacdon/LDblocks_GRCh38

GWAS summary statistics (UK Biobank): http://www.nealelab.is/uk-biobank

Acknowledgements

This work was funded in part by the National Institutes of Health (NIH) under awards R01HG012133 and R01GM140287.

References

  1. 1.↵
    Virani SS, Alonso A, Aparicio HJ, Benjamin EJ, Bittencourt MS, Callaway CW, et al. Heart disease and stroke statistics-2021 update: A report from the American Heart Association. Circulation. 2021;143: e254–e743.
    OpenUrlCrossRefPubMed
  2. 2.↵
    Feigin VL, Brainin M, Norrving B, Martins S, Sacco RL, Hacke W, et al. World Stroke Organization (WSO): Global Stroke Fact Sheet 2022. Int J Stroke. 2022;17: 18–29.
    OpenUrlCrossRefPubMed
  3. 3.↵
    Chen R, Ovbiagele B, Feng W. Diabetes and Stroke: Epidemiology, Pathophysiology, Pharmaceuticals and Outcomes. Am J Med Sci. 2016;351: 380–386.
    OpenUrlCrossRefPubMed
  4. 4.↵
    Wajngarten M, Silva GS. Hypertension and Stroke: Update on Treatment. Eur Cardiol. 2019;14: 111–115.
    OpenUrl
  5. 5.
    Horn JW, Feng T, Mørkedal B, Strand LB, Horn J, Mukamal K, et al. Obesity and Risk for First Ischemic Stroke Depends on Metabolic Syndrome: The HUNT Study. Stroke. 2021;52: 3555–3561.
    OpenUrl
  6. 6.↵
    Amarenco P, Lavallée PC, Labreuche J, Ducrocq G, Juliard J-M, Feldman L, et al. Coronary artery disease and risk of major vascular events after cerebral infarction. Stroke. 2013;44: 1505–1511.
    OpenUrlAbstract/FREE Full Text
  7. 7.↵
    Malik R, Chauhan G, Traylor M, Sargurupremraj M, Okada Y, Mishra A, et al. Multiancestry genome-wide association study of 520,000 subjects identifies 32 loci associated with stroke and stroke subtypes. Nat Genet. 2018;50: 524–537.
    OpenUrlCrossRefPubMed
  8. 8.↵
    Farh KK-H, Marson A, Zhu J, Kleinewietfeld M, Housley WJ, Beik S, et al. Genetic and epigenetic fine mapping of causal autoimmune disease variants. Nature. 2015;518: 337–343.
    OpenUrlCrossRefPubMed
  9. 9.↵
    Gusev A, Ko A, Shi H, Bhatia G, Chung W, Penninx BWJH, et al. Integrative approaches for large-scale transcriptome-wide association studies. Nat Genet. 2016;48: 245–252.
    OpenUrlCrossRefPubMed
  10. 10.↵
    Mancuso N, Gayther S, Gusev A, Zheng W, Penney KL, Kote-Jarai Z, et al. Large-scale transcriptome-wide association study identifies new prostate cancer risk regions.Nat Commun. 2018;9: 4079.
    OpenUrlPubMed
  11. 11.↵
    Yang J, Yan B, Fan Y, Yang L, Zhao B, He X, et al. Integrative analysis of transcriptome-wide association study and gene expression profiling identifies candidate genes associated with stroke. PeerJ. 2019. p. e7435. doi:10.7717/peerj.7435
    OpenUrlCrossRef
  12. 12.↵
    Wu B-S, Chen S-F, Huang S-Y, Ou Y-N, Deng Y-T, Chen S-D, et al. Identifying causal genes for stroke via integrating the proteome and transcriptome from brain and blood. J Transl Med. 2022;20: 181.
    OpenUrl
  13. 13.↵
    Li Y, Kellis M. Joint Bayesian inference of risk variants and tissue-specific epigenomic enrichments across multiple complex human diseases. Nucleic Acids Res. 2016;44: e144.
    OpenUrlCrossRefPubMed
  14. 14.
    Finucane HK, Reshef YA, Anttila V, Slowikowski K, Gusev A, Byrnes A, et al. Heritability enrichment of specifically expressed genes identifies disease-relevant tissues and cell types. Nat Genet. 2018;50: 621–629.
    OpenUrlCrossRefPubMed
  15. 15.↵
    Hao X, Zeng P, Zhang S, Zhou X. Identifying and exploiting trait-relevant tissues with multiple functional annotations in genome-wide association studies. PLoS Genet. 2018;14: e1007186.
    OpenUrlCrossRefPubMed
  16. 16.↵
    Mishra A, Malik R, Hachiya T, Jürgenson T, Namba S, Posner DC, et al. Stroke genetics informs drug discovery and risk prediction across ancestries. Nature. 2022;611: 115–123.
    OpenUrl
  17. 17.↵
    Li YI, van de Geijn B, Raj A, Knowles DA, Petti AA, Golan D, et al. RNA splicing is a primary link between genetic variation and disease. Science. 2016;352: 600–604.
    OpenUrlAbstract/FREE Full Text
  18. 18.↵
    Ding M, Liu Y, Liao X, Zhan H, Liu Y, Huang W. Enhancer RNAs (eRNAs): New Insights into Gene Transcription and Disease Treatment. J Cancer. 2018;9: 2334–2340.
    OpenUrl
  19. 19.↵
    Dong P, Hoffman GE, Apontes P, Bendl J, Rahman S, Fernando MB, et al. Population-level variation in enhancer expression identifies disease mechanisms in the human brain. Nature Genetics. 2022. pp. 1493–1503. doi:10.1038/s41588-022-01170-4
    OpenUrlCrossRef
  20. 20.↵
    Subramanian A, Narayan R, Corsello SM, Peck DD, Natoli TE, Lu X, et al. A Next Generation Connectivity Map: L1000 Platform and the First 1,000,000 Profiles.Cell. 2017;171: 1437-1452.e17.
    OpenUrlCrossRefPubMed
  21. 21.↵
    Konuma T, Ogawa K, Okada Y. Integration of genetically regulated gene expression and pharmacological library provides therapeutic drug candidates. Hum Mol Genet. 2021;30: 294–304.
    OpenUrlCrossRef
  22. 22.↵
    So H-C, Chau CK-L, Chiu W-T, Ho K-S, Lo C-P, Yim SH-Y, et al. Analysis of genome-wide association data highlights candidates for drug repositioning in psychiatry. Nat Neurosci. 2017;20: 1342–1349.
    OpenUrlCrossRef
  23. 23.↵
    Mancuso N, Freund MK, Johnson R, Shi H, Kichaev G, Gusev A, et al. Probabilistic fine-mapping of transcriptome-wide association studies. Nat Genet. 2019;51: 675–682.
    OpenUrlCrossRefPubMed
  24. 24.↵
    GTEx Consortium. The GTEx Consortium atlas of genetic regulatory effects across human tissues. Science. 2020;369: 1318–1330.
    OpenUrlAbstract/FREE Full Text
  25. 25.↵
    Stegle O, Parts L, Durbin R, Winn J. A Bayesian framework to account for complex non-genetic factors in gene expression levels greatly increases power in eQTL studies. PLoS Comput Biol. 2010;6: e1000770.
    OpenUrlCrossRefPubMed
  26. 26.↵
    Yang J, Lee SH, Goddard ME, Visscher PM. GCTA: a tool for genome-wide complex trait analysis. Am J Hum Genet. 2011;88: 76–82.
    OpenUrlCrossRefPubMed
  27. 27.↵
    Santosa F, Symes WW. Linear Inversion of Band-Limited Reflection Seismograms. SIAM J Sci and Stat Comput. 1986;7: 1307–1330.
    OpenUrl
  28. 28.↵
    Zou H, Hastie T. Regularization and variable selection via the elastic net. J R Stat Soc Series B Stat Methodol. 2005;67: 301–320.
    OpenUrlCrossRef
  29. 29.↵
    Wang G, Sarkar A, Carbonetto P, Stephens M. A simple new approach to variable selection in regression, with application to genetic fine mapping. J R Stat Soc Series B Stat Methodol. 2020;82: 1273–1300.
    OpenUrlCrossRef
  30. 30.↵
    Sun BB, Maranville JC, Peters JE, Stacey D, Staley JR, Blackshaw J, et al. Genomic atlas of the human plasma proteome. Nature. 2018;558: 73–79.
    OpenUrlCrossRefPubMed
  31. 31.↵
    Investigators TA, THE ARIC INVESTIGATORS. THE ATHEROSCLEROSIS RISK IN COMMUNIT (ARIC) STUDY: DESIGN AND OBJECTIVES. American Journal of Epidemiology. 1989. pp. 687–702. doi:10.1093/oxfordjournals.aje.a115184
    OpenUrlCrossRefPubMed
  32. 32.↵
    Pathak GA, Singh K, Miller-Fleming TW, Wendt FR, Ehsan N, Hou K, et al. Integrative genomic analyses identify susceptibility genes underlying COVID-19 hospitalization. Nat Commun. 2021;12: 4569.
    OpenUrl
  33. 33.↵
    Zhang J, Dutta D, Köttgen A, Tin A, Schlosser P, Grams ME, et al. Plasma proteome analyses in individuals of European and African ancestry identify cis-pQTLs and models for proteome-wide association studies. Nat Genet. 2022;54: 593–602.
    OpenUrlCrossRef
  34. 34.↵
    1000 Genomes Project Consortium, Auton A, Brooks LD, Durbin RM, Garrison EP, Kang HM, et al. A global reference for human genetic variation. Nature. 2015;526: 68–74.
    OpenUrlCrossRefPubMed
  35. 35.↵
    Barbeira AN, Dickinson SP, Bonazzola R, Zheng J, Wheeler HE, Torres JM, et al. Exploring the phenotypic consequences of tissue specific gene expression variation inferred from GWAS summary statistics. Nat Commun. 2018;9: 1825.
    OpenUrlCrossRefPubMed
  36. 36.↵
    Gamazon ER, Wheeler HE, Shah KP, Mozaffari SV, Aquino-Michaels K, Carroll RJ, et al. A gene-based association method for mapping traits using reference transcriptome data. Nat Genet. 2015;47: 1091–1098.
    OpenUrlCrossRefPubMed
  37. 37.↵
    Barbeira AN, Bonazzola R, Gamazon ER, Liang Y, Park Y, Kim-Hellmuth S, et al. Exploiting the GTEx resources to decipher the mechanisms at GWAS loci. bioRxiv. 2020. p. 814350. doi:10.1101/814350
    OpenUrlAbstract/FREE Full Text
  38. 38.↵
    Yao DW, O’Connor LJ, Price AL, Gusev A. Quantifying genetic effects on disease mediated by assayed gene expression levels. Nat Genet. 2020;52: 626–633.
    OpenUrl
  39. 39.↵
    International HapMap Consortium. The International HapMap Project. Nature. 2003;426: 789–796.
    OpenUrlCrossRefPubMedWeb of Science
  40. 40.↵
    Berisa T, Pickrell JK. Approximately independent linkage disequilibrium blocks in human populations. Bioinformatics. 2016;32: 283–285.
    OpenUrlCrossRefPubMed
  41. 41.↵
    Pividori M, Rajagopal PS, Barbeira A, Liang Y, Melia O, Bastarache L, et al. PhenomeXcan: Mapping the genome to the phenome through the transcriptome. Sci Adv. 2020;6. doi:10.1126/sciadv.aba2083
    OpenUrlFREE Full Text
  42. 42.↵
    Mancuso N, Shi H, Goddard P, Kichaev G, Gusev A, Pasaniuc B. Integrating Gene Expression with Summary Association Statistics to Identify Genes Associated with 30 Complex Traits. Am J Hum Genet. 2017;100: 473–487.
    OpenUrlCrossRefPubMed
  43. 43.↵
    MacDonald JW, Harrison T, Bammler TK, Mancuso N, Lindström S. An updated map of GRCh38 linkage disequilibrium blocks based on European ancestry data. bioRxiv. 2022. p. 2022.03.04.483057. doi:10.1101/2022.03.04.483057
    OpenUrlAbstract/FREE Full Text
  44. 44.↵
    Bulik-Sullivan BK, Loh P-R, Finucane HK, Ripke S, Yang J, Schizophrenia Working Group of the Psychiatric Genomics Consortium, et al. LD Score regression distinguishes confounding from polygenicity in genome-wide association studies. Nat Genet. 2015;47: 291–295.
    OpenUrlCrossRefPubMed
  45. 45.↵
    Bowden J, Davey Smith G, Burgess S. Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression. Int J Epidemiol. 2015;44: 512–525.
    OpenUrlCrossRefPubMed
  46. 46.↵
    Chang C-S, Chen H-J, Liao C-H. Patients with cerebral stroke have an increased risk of gastroesophageal reflux disease: A population-based cohort study. J Stroke Cerebrovasc Dis. 2018;27: 1267–1274.
    OpenUrl
  47. 47.↵
    Sheu J-J, Kang J-H, Lou H-Y, Lin H-C. Reflux esophagitis and the risk of stroke in young adults: a 1-year population-based follow-up study. Stroke. 2010;41: 2033–2037.
    OpenUrlAbstract/FREE Full Text
  48. 48.↵
    Arboix A. Cardiovascular risk factors for acute stroke: Risk profiles in the different subtypes of ischemic stroke. World J Clin Cases. 2015;3: 418–429.
    OpenUrlPubMed
  49. 49.↵
    Libby P, Buring JE, Badimon L, Hansson GK, Deanfield J, Bittencourt MS, et al. Atherosclerosis. Nat Rev Dis Primers. 2019;5: 56.
    OpenUrl
  50. 50.↵
    Liu P-Y, Zhang Z, Liu Y, Tang X-L, Shu S, Bao X-Y, et al. TMEM16A Inhibition Preserves Blood–Brain Barrier Integrity After Ischemic Stroke. Front Cell Neurosci. 2019;13. doi:10.3389/fncel.2019.00360
    OpenUrlCrossRef
  51. 51.↵
    New L, Jiang Y, Zhao M, Liu K, Zhu W, Flood LJ, et al. PRAK, a novel protein kinase regulated by the p38 MAP kinase. EMBO J. 1998;17: 3372–3384.
    OpenUrlAbstract/FREE Full Text
  52. 52.↵
    Islam MS, Shin H-Y, Yoo Y-J, Kim R, Jang Y-J, Akanda MR, et al. Olanzapine Ameliorates Ischemic Stroke-like Pathology in Gerbils and H2O2-Induced Neurotoxicity in SH-SY5Y Cells via Inhibiting the MAPK Signaling Pathway. Antioxidants (Basel). 2022;11. doi:10.3390/antiox11091697
    OpenUrlCrossRef
  53. 53.↵
    Prospective Studies Collaboration, Lewington S, Whitlock G, Clarke R, Sherliker P, Emberson J, et al. Blood cholesterol and vascular mortality by age, sex, and blood pressure: a meta-analysis of individual data from 61 prospective studies with 55,000 vascular deaths. Lancet. 2007;370: 1829–1839.
    OpenUrlCrossRefPubMedWeb of Science
  54. 54.↵
    Hörnsten C, Weidung B, Littbrand H, Carlberg B, Nordström P, Lövheim H, et al. High blood pressure as a risk factor for incident stroke among very old people: a population-based cohort study. J Hypertens. 2016;34: 2059–2065.
    OpenUrl
  55. 55.↵
    Chang CC, Kass L. Clinical significance of immature reticulocyte fraction determined by automated reticulocyte counting. Am J Clin Pathol. 1997;108: 69–73.
    OpenUrlCrossRef
  56. 56.↵
    Meier ER, Wright EC, Miller JL. Reticulocytosis and anemia are associated with an increased risk of death and stroke in the newborn cohort of the Cooperative Study of Sickle Cell Disease. Am J Hematol. 2014;89: 904–906.
    OpenUrlCrossRefPubMed
  57. 57.
    Silva CM, Giovani P, Viana MB. High reticulocyte count is an independent risk factor for cerebrovascular disease in children with sickle cell anemia. Pediatr Blood Cancer. 2011;56: 116–121.
    OpenUrlCrossRefPubMed
  58. 58.↵
    Belisário AR, Sales RR, Toledo NE, Muniz MB de SR, Velloso-Rodrigues C, Silva CM, et al. Reticulocyte count is the most important predictor of acute cerebral ischemia and high-risk transcranial Doppler in a newborn cohort of 395 children with sickle cell anemia. Ann Hematol. 2016;95: 1869–1880.
    OpenUrl
  59. 59.↵
    Monte MJ, Marin JJG, Antelo A, Vazquez-Tato J. Bile acids: chemistry, physiology, and pathophysiology. World J Gastroenterol. 2009;15: 804–816.
    OpenUrlCrossRefPubMedWeb of Science
  60. 60.↵
    Dionísio PA, Amaral JD, Ribeiro MF, Lo AC, D’Hooge R, Rodrigues CMP. Amyloid-β pathology is attenuated by tauroursodeoxycholic acid treatment in APP/PS1 mice after disease onset. Neurobiol Aging. 2015;36: 228–240.
    OpenUrlCrossRefPubMed
  61. 61.
    Lo AC, Callaerts-Vegh Z, Nunes AF, Rodrigues CMP, D’Hooge R. Tauroursodeoxycholic acid (TUDCA) supplementation prevents cognitive impairment and amyloid deposition in APP/PS1 mice. Neurobiol Dis. 2013;50: 21–29.
    OpenUrlCrossRef
  62. 62.
    Abdelkader NF, Safar MM, Salem HA. Ursodeoxycholic Acid Ameliorates Apoptotic Cascade in the Rotenone Model of Parkinson’s Disease: Modulation of Mitochondrial Perturbations. Mol Neurobiol. 2016;53: 810–817.
    OpenUrlCrossRefPubMed
  63. 63.
    Rosa AI, Fonseca I, Nunes MJ, Moreira S, Rodrigues E, Carvalho AN, et al. Novel insights into the antioxidant role of tauroursodeoxycholic acid in experimental models of Parkinson’s disease. Biochim Biophys Acta Mol Basis Dis. 2017;1863: 2171–2181.
    OpenUrl
  64. 64.↵
    Keene CD, Rodrigues CMP, Eich T, Chhabra MS, Steer CJ, Low WC. Tauroursodeoxycholic acid, a bile acid, is neuroprotective in a transgenic animal model of Huntington’s disease. Proc Natl Acad Sci U S A. 2002;99: 10671–10676.
    OpenUrlAbstract/FREE Full Text
  65. 65.↵
    Rodrigues CMP, Sola S, Nan Z, Castro RE, Ribeiro PS, Low WC, et al. Tauroursodeoxycholic acid reduces apoptosis and protects against neurological injury after acute hemorrhagic stroke in rats. Proc Natl Acad Sci U S A. 2003;100: 6087–6092.
    OpenUrlAbstract/FREE Full Text
  66. 66.↵
    Charach G, Karniel E, Novikov I, Galin L, Vons S, Grosskopf I, et al. Reduced bile acid excretion is an independent risk factor for stroke and mortality: A prospective follow-up study.Atherosclerosis. 2020;293: 79–85.
    OpenUrl
  67. 67.↵
    Zhu Z, Zhang F, Hu H, Bakshi A, Robinson MR, Powell JE, et al. Integration of summary data from GWAS and eQTL studies predicts complex trait gene targets. Nat Genet. 2016;48: 481–487.
    OpenUrlCrossRefPubMed
  68. 68.↵
    Astle WJ, Elding H, Jiang T, Allen D, Ruklisa D, Mann AL, et al. The Allelic landscape of human blood cell trait variation and links to common complex disease. Cell. 2016;167: 1415-1429.e19.
    OpenUrlCrossRefPubMed
  69. 69.↵
    Meegan JE, Shaver CM, Putz ND, Jesse JJ, Landstreet SR, Lee HNR, et al. Cell-free hemoglobin increases inflammation, lung apoptosis, and microvascular permeability in murine polymicrobial sepsis. PLoS One. 2020;15: e0228727.
    OpenUrl
  70. 70.↵
    Gladwin MT, Kanias T, Kim-Shapiro DB. Hemolysis and cell-free hemoglobin drive an intrinsic mechanism for human disease. The journal of clinical investigation. 2012. pp. 1205–1208.
  71. 71.↵
    Amarenco P, Lavallée PC, Labreuche J, Albers GW, Bornstein NM, Canhão P, et al. One-Year Risk of Stroke after Transient Ischemic Attack or Minor Stroke. N Engl J Med. 2016;374: 1533–1542.
    OpenUrlCrossRefPubMed
  72. 72.↵
    Giles MF, Rothwell PM. Risk of stroke early after transient ischaemic attack: a systematic review and meta-analysis. Lancet Neurol. 2007;6: 1063–1072.
    OpenUrlCrossRefPubMedWeb of Science
  73. 73.↵
    Rothwell PM, Giles MF, Chandratheva A, Marquardt L, Geraghty O, Redgrave JNE, et al. Effect of urgent treatment of transient ischaemic attack and minor stroke on early recurrent stroke (EXPRESS study): a prospective population-based sequential comparison. Lancet. 2007;370: 1432–1442.
    OpenUrlCrossRefPubMedWeb of Science
  74. 74.↵
    Nelson MR, Tipney H, Painter JL, Shen J, Nicoletti P, Shen Y, et al. The support of human genetic evidence for approved drug indications. Nat Genet. 2015;47: 856–860.
    OpenUrlCrossRefPubMed
  75. 75.↵
    King EA, Wade Davis J, Degner JF. Are drug targets with genetic support twice as likely to be approved? Revised estimates of the impact of genetic support for drug mechanisms on the probability of drug approval. PLoS Genet. 2019;15: e1008489.
    OpenUrlCrossRefPubMed
  76. 76.↵
    Okada Y, Wu D, Trynka G, Raj T, Terao C, Ikari K, et al. Genetics of rheumatoid arthritis contributes to biology and drug discovery. Nature. 2014;506: 376–381.
    OpenUrlCrossRefPubMedWeb of Science
  77. 77.↵
    Fang H, ULTRA-DD Consortium, De Wolf H, Knezevic B, Burnham KL, Osgood J, et al. A genetics-led approach defines the drug target landscape of 30 immune-related traits. Nat Genet. 2019;51: 1082–1091.
    OpenUrl
  78. 78.↵
    Ochoa D, Karim M, Ghoussaini maya, Hulcoop DG, McDonagh EM, Dunham I. Human genetics evidence supports two-thirds of the 2021 FDA-approved drugs. Nat Rev Drug Discov. 2022 [cited 8 Jul 2022]. doi:10.1038/d41573-022-00120-3
    OpenUrlCrossRef
  79. 79.↵
    Lamb J, Crawford ED, Peck D, Modell JW, Blat IC, Wrobel MJ, et al. The Connectivity Map: using gene-expression signatures to connect small molecules, genes, and disease. Science. 2006;313: 1929–1935.
    OpenUrlAbstract/FREE Full Text
  80. 80.↵
    Iorio F, Bosotti R, Scacheri E, Belcastro V, Mithbaokar P, Ferriero R, et al. Discovery of drug mode of action and drug repositioning from transcriptional responses. Proc Natl Acad Sci U S A. 2010;107: 14621–14626.
    OpenUrlAbstract/FREE Full Text
  81. 81.↵
    Grishin D, Gusev A. Allelic imbalance of chromatin accessibility in cancer identifies candidate causal risk variants and their mechanisms. Nat Genet. 2022;54: 837–849.
    OpenUrl
  82. 82.↵
    Baca SC, Singler C, Zacharia S, Seo J-H, Morova T, Hach F, et al. Genetic determinants of chromatin reveal prostate cancer risk mediated by context-dependent gene regulation. Nat Genet. 2022;54: 1364–1375.
    OpenUrl
Back to top
Posted March 31, 2023.
Download PDF

Supplementary Material

Data/Code
Email
Novel insight into the etiology of ischemic stroke gained by integrative transcriptome-wide association study
Junghyun Jung, Zeyun Lu, Adam de Smith, Nicholas Mancuso
medRxiv 2023.03.30.23287918; doi: https://doi.org/10.1101/2023.03.30.23287918
Twitter logo Facebook logo LinkedIn logo Mendeley logo
Citation Tools

Subject Area