Herpesvirus latent infection promotes stroke via activating the OTUD1/NF-κB signaling pathway

Database search strategy

We independently searched three English databases, including Pubmed, Web of Science, and EMBASE, by two researchers from their inception to March 2023. Taking Pubmed as an example, a combination of subject headings and free text terms was used for the search. The subject headings mainly included stroke and herpesviruses (including herpes simplex virus types 1 and 2 (HSV-1 and HSV-2), varicella-zoster virus (VZV) or herpes zoster virus (HZ), Epstein-Barr virus (EBV), cytomegalovirus (CMV), herpesviruses 6, 7, and 8) [20].

Inclusion and exclusion criteria

Cross-sectional research, ecological studies, case series, case reports, and reviews were all eliminated. Studies must report or allow for the derivation of impact estimates. We were not restricted by time, publishing status, language, geographical location, or medical setting. Studies were included if stroke (first or subsequent) was the outcome, either clinically diagnosed or self-reported.

The following are the inclusion criteria: case-control, cohort, or self-controlled case series (SCCS) studies that evaluate the relationship between herpesvirus infection, and the risk of stroke are eligible for inclusion; humans make up the study’s population; the stroke is the outcome; multivariate-adjusted relative risk, hazard ratio, or odds ratio with 95% confidence interval (CI) are provided. Exclusion criteria: (1) duplicate publications; (2) inadequate data; (3) not giving effect sizes (HR, RR, or OR); (3) case studies, analyses, thorough reviews, or abstracts [20, 21].

Data extraction

A standardized data collection form was used to independently extract data from included studies by two researchers based on the PICOS (population, exposure, comparator, outcome, and study design) framework. If any disputes arose during the data extraction process, they were resolved through discussion and negotiation with multiple research personnel until a consensus was reached. As this was an etiological study, “exposure” replaced “intervention,” and “study characteristics’ were expanded to “study design.” We also recorded the most fully adjusted effect estimate (HR, RR, or OR) for the association between exposure and stroke.

Evidence-based medicine software and data analysis

The meta-analysis used Stata 12.0 (StataCorp, College Station, TX, USA). We initially used the random-effects inverse variance model to construct the summary OR and its accompanying 95% CI to assess the relationship between herpesvirus exposure and stroke risk. The heterogeneity across studies was measured using the I2 statistic, with larger I2 values suggesting more heterogeneity. After removing trials with significant heterogeneity, the Galbraith plot was used to determine the source of heterogeneity. Using the Metafunnel and Metabias commands, the funnel plot was utilized to examine publication bias subjectively and statistically. The durability of the meta-analysis findings was tested using sensitivity analysis employing the trim-and-fill approach in the presence of publication bias. Additionally, the Metaninf command was used to assess the impact of individual studies on the outcomes of the meta-analysis.

Gene differential expression analysis

The Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/gds) was used to obtain the GSE51365 (Mus musculus) latent herpesvirus infection-related dataset and the GSE22255 (Homo sapiens) blood gene expression dataset for ischemic stroke (IS). The GSE51365 dataset contains brain tissue samples from three control (MOCK), wild-type (MHV-68), and mutant (ORF73) herpesvirus infection groups, respectively. The GSE22255 dataset contains peripheral blood mononuclear cell (PBMC) samples from 20 IS patients and 20 age- and gender-matched controls. The “limma” package in R language filters differentially expressed genes, with a p-value < 0.05 as the threshold. The “ggplot2” package was used to plot the gene differential expression volcano plot and box plot, while the “heatmap” package was used to plot the heatmap of differential gene expression. R version 4.2.1 was used for all analyses [22].

Weighted gene co-expression network analysis (WGCNA)

The R software package “WGCNA” was used to perform WGCNA analysis on the GSE22255 dataset. First, each gene’s median absolute deviation (MAD) was determined, and the bottom 30% of genes with the smallest MAD were removed. Second, the goodSamplesGenes function filters the DEG expression matrix to remove unqualified genes and establish a scale-free co-expression network. Third, a suitable soft threshold was chosen using the “pickSoftThreshold” function; the adjacency matrix was then transformed; the topological overlap matrix (TOM) was then calculated; and finally, a hierarchical clustering dendrogram was built to separate similar gene expression into different modules, with a minimum of 200 genes per module. To merge possible similar modules, a threshold of 0.2 was defined as the cutting height. Finally, the module eigengenes (MEs) were used to summarize the expression profiles of each module, and the correlation between MEs and traits was calculated to select the most relevant modules for further analysis [23–25].

ROC curve analysis

The significant module genes from the WGCNA analysis were intersected with the differentially expressed genes in the wild-type and mutant strains of murine gammaherpesvirus-infected brain tissues from dataset GSE51365. The results were visualized using the R ggplot2 and VennDiagram packages. The intersection genes were analyzed using the pROC package in R, and the ROC (receiver operating characteristic) curve was plotted using the ggplot2 package [26].

GO and KEGG enrichment analysis

Pearson correlation analysis was used to screen for genes related to OTUD1 in dataset GSE22255, and the intersection of differentially expressed genes obtained from dataset GSE22255 and the genes related to OTUD1 were visualized using a co-expression heatmap. Further, GO (Gene Ontology) and KEGG (Kyoto Encyclopedia of Genes and Genomes) enrichment analysis was performed using R packages such as “clusterProfiler,” “org.Hs.eg.db,” “enrichplot,” and “GOplot.” The GO and KEGG enrichment analysis results were visualized using a bubble plot [27].

Statistical analysis

Using STATA 12.0 software, a meta-analysis in evidence-based medicine was carried out. The statistical tests were two-sided, and P < 0.05 was considered statistically significant. All statistical analyses in the bioinformatics part were conducted using R software. A non-paired t-test was used to compare the two groups, and Pearson correlation analysis was used to observe the correlation between the primary genes and other genes. P < 0.05 was considered statistically significant.

The meta-analysis of evidence-based medicine shows that there is a correlation between herpesvirus infection and an increased risk of stroke

The meta-analysis included 13 studies (4 case-control studies, 8 cohort studies, and 1 SCCS study) with publication dates ranging from 2007 to 2023. Of these, 8 studies evaluated the relationship between VZV or HZ and stroke, while the other 5 evaluated the relationship between multiple herpesviruses (such as HSV1/2, VZV, EBV, or CMV) infection and stroke. Baseline characteristics of the included literature are presented in the Supplementary File 1.

First, we conducted a meta-analysis evaluating the association between various herpesvirus infections and stroke using a random-effects model. The findings revealed that whereas HSV infection was linked to a higher incidence of stroke (OR: 1.18, 95% CI 0.84–1.64, I2 = 48.5%), VZV infection was not (OR: 1.20, 95% CI 1.11–1.30, I2 = 77.9%). While EBV infection was linked to a higher incidence of stroke (OR: 1.29, 95% CI 0.89–1.87, I2 = 0.0%), CMV infection was not (OR: 1.45, 95% CI 0.84–2.49, I2 = 69%) (Figure 1). As shown in Figure 1, there was significant heterogeneity in the VZV group model analysis. Therefore, we further performed a subgroup analysis based on follow-up time, which showed that VZV infection was associated with stroke risk in all follow-up periods, including over 1 year, up to 12 weeks, and up to 26 weeks, and that VZV infection was associated with an increased risk of stroke (Figure 2).

We then conducted further analyses to assess heterogeneity, publication bias, and sensitivity. The Galbraith plot indicated that several studies were outside the confidence interval, suggesting that these studies had larger heterogeneity and may have had a greater impact on the combined effect size (Figure 3A). The funnel plot showed that several studies were not scattered within the funnel (Figure 3B). The Egger’s test indicated a significant asymmetry in the funnel plot with statistical significance (P = 0.019) (Figure 3C), indicating the presence of publication bias in the analysis results. Further sensitivity analysis using the trim and fill method showed that the effect sizes (before: fixed: 0.096; random: 0.171; after: fixed: 0.083; random: 0.106) and P values (both less than 0.05) changed little, indicating that the meta-analysis results were stable and reliable (Figure 3D). Finally, we examined the influence of individual studies on the results, and the combined effect sizes of each study were within the confidence interval range, indicating that there was little difference among the studies (Supplementary Figure 1).

In conclusion, the evidence-based medicine meta-analysis revealed that herpesvirus infection is linked to a higher risk of stroke, with the strongest associations between VZV, EBV, and HSV infection and stroke risk.

Identification of differentially expressed genes in wild-type and mutant herpesvirus latent infection-induced brain tissues through GEO chip differential analysis

To investigate the potential molecular mechanisms underlying herpesvirus latent infection-induced cerebral infarction, we first retrieved the dataset GSE51365 from the GEO database, which contains transcriptome changes induced by latent herpesvirus infection in different organs. We selected brain tissue samples from the control (MOCK) group, wild-type (MHV-68) group, and mutant-type (ORF73) group infected with gamma herpesvirus for differential analysis. We identified 900 differentially expressed genes (449 upregulated and 451 downregulated) in the MHV-68 group (Figure 4A, 4B) and 2104 differentially expressed genes (1142 upregulated and 962 downregulated) in the ORF73 group (Figure 4C, 4D).

WGCNA co-expression analysis identified 165 stroke-associated hub genes

We further performed a WGCNA co-expression analysis on the GEO stroke-related expression profile dataset. After the hierarchical clustering of the samples (Figure 5A), a soft-thresholding power of β = 12 (scale-free R2 = 0.9) was selected to construct a scale-free network (Figure 5B). Nine gene co-expression modules were identified in the GSE22255 dataset (Figure 5C), among which only the pink module was significantly associated with stroke occurrence with a correlation level of 0.4 (P = 0.01) (Figure 5D). The module gene importance was also calculated and revealed that the pink module genes had the highest gene significance score, and their gene importance was significantly positively correlated (Figure 5E, 5F). Therefore, we identified 165 genes in the pink module as stroke-associated hub genes.

OTUD1 may be a key gene in the mechanism of herpesvirus latent infection-induced stroke

To further identify key genes in the mechanism of herpesvirus latent infection-induced stroke, we took the intersection of the hub genes identified by WGCNA and the differentially expressed genes in wild-type and mutant herpesvirus-infected brain tissues, resulting in three overlapping genes: OTUD1, NFIL3, and OSM (Figure 6A).

ROC curve analysis indicated that OTUD1 and NFIL3 could predict stroke occurrence with a certain degree of accuracy, while OSM could not (Figure 6B). The differential expression of OTUD1, NFIL3, and OSM in the stroke dataset samples is shown in Supplementary Table 1, among which only OTUD1 and OSM were significantly upregulated in stroke. Additionally, differential expression analysis of the GSE51365 dataset revealed that OTUD1 was downregulated in brain tissues infected with wild-type herpesvirus, while it was significantly upregulated in brain tissues infected with mutant herpesvirus (Figure 6C).

Previous studies have reported that after viral infections such as herpes simplex virus 1 (HSV-1) infection, OTUD1 is upregulated, and RNA viruses specifically promote the expression of the deubiquitinating enzyme OTUD1 through an NF-κB-dependent mechanism during the early stages of viral infection [13, 28]. Therefore, we speculate that OTUD1 may be a key gene in the mechanism of herpesvirus latent infection-induced stroke.

OTUD1 may mediate the occurrence of stroke by activating the NF-κB signaling pathway in herpes simplex virus latent infection

To further explore the mechanism by which OTUD1 regulates stroke in herpes simplex virus latent infection, we performed single-gene correlation analysis on the stroke samples in the GSE22255 dataset and obtained 4848 genes significantly positively or negatively correlated with OTUD1. We then intersected the top 50 significantly correlated positive and negative genes with the differentially expressed genes in the stroke samples of the GSE22255 dataset to obtain 13 intersection genes (Figure 7A). The co-expression heatmap of OTUD1 and the 13 intersection genes is shown in Figure 7B, where OTUD1 is significantly positively correlated with PTGS2, CXCL2, YTHDF3-AS1, BCL10, EIF1, and AC079305.10, and significantly negatively correlated with DPF2, POLK, LOC388692, SP1, STK38, SUPT20H, and FAM192A.

Further, GO function and KEGG pathway enrichment analyses of these 14 genes showed that they are mainly involved in the biological processes of forming complexes or complexes with other proteins, participating in lipopolysaccharide reactions, and responding to environmental, mechanical, or non-biological stimuli (Figure 7C). The KEGG analysis showed that these 14 genes are mainly enriched in the NF-kappa B signaling pathway, human cytomegalovirus infection, Kaposi sarcoma-associated herpesvirus infection, IL-17 signaling pathway, and TNF signaling pathway (Figure 7D), indicating that they are closely related to herpes virus infection as well as inflammation and immune response. The most notable enrichment among them is in the NF-B signaling pathway, including the genes PTGS2, CXCL2, and BCL10, which positively correlate with OTUD1. Therefore, we speculate that OTUD1 may mediate the occurrence of stroke in herpes simplex virus latent infection by activating the NF-κB signaling pathway.