Gene expression profiles of stroked brain tissue across sex, species, stroke model, and time

In total, 338 studies were screened that examined gene-expression differences in stroke. Of these, 213 studies analyzed brain tissue and used non-stroked brain tissue controls. Datasets with missing information, duplicates, pooled samples, poor quality controls, and no clear separation between stroked and non-stroked groups were further excluded. In total, we included 34 datasets from mice (12), rats (15), humans (2) and primates (3) (Fig. 1A). The datasets included in this meta-analysis compared gene expression in stroked brain tissue to control brain tissue (either from intact, sham operated or contralesional brain regions). Detailed information of each dataset is provided in Suppl. Table 1 describing the GEO ID, sample type, control type, stroke model, time point, sex, and further information.

The majority of datasets were acquired either in the acute, (< 24 h: 49%) or subacute (> 24 h and < 7d: 36%) phase after stroke, only few datasets investigated long-term gene expression changes after stroke (> 7d: 10%). The most frequently applied model of stroke was the tMCAo (67%), followed by the pMCAo model (31%) only 2% of the studies used the photothrombotic stroke model. Surprisingly, the vast majority of datasets only used the male sex (74%) (Fig. 1B). The majority of datasets had data from 10 to 25k genes, only four micro-array studies showed data for less than 10k hits (all of them were derived from human and primate samples) (Fig. 1C). Principal component analysis (PCA) revealed the widest separation across the datasets for the acute, subacute, and long-term periods after stroke indicating the importance of time in the transcriptomic signatures (Fig. 1D).

Fig. 1

Overview of studies dissecting stroke-related gene expression differences in the brain. (A) (left) Excluded and included stroke studies from NCBI GEO. (Right) Exclusion of datasets subdivided by species (mouse, rat, human and primates) with missing information, duplicates, pooled samples, poor quality controls, and no clear separation between stroked and non-stroked groups were further excluded. (B) Distribution of individual studies by time point, stroke model and sex. (C) Number of gene hits in individual studies. More information about the IDs can be found in Suppl Table 1. (D) Principal component analysis (PCA) of all stroke studies across different species and time points

Gene expression differences in stroked brain tissue between mice and rats are pronounced in acute phase after injury

The vast majority of stroke studies are performed in either rats or mice, hence the focus of the subsequent analysis was on the more detailed rodent datasets.

To analyze the effect of the species, the transcriptomes were compared at the acute, subacute, and long-term timepoints after MCAo stroke induction. Analysis of differentially expressed genes (DEGs) revealed that there was a greater overlap between upregulated genes (~ 25%) in the mouse and rat transcriptome to all time points (Fig. 2A, Suppl. Figure 1). Most upregulated genes acutely after stroke were as expected early-response and inflammation associated genes such as Hspa1a, Cxcl1, Ccl3, and Fos. For downregulated genes the overlap was higher at the subacute and long-term time point (shared downregulated genes: acute: 1%, subacute: 12%, long-term: 5%, Fig. 2A). A list of the 60 most DEG can be found in Suppl. Tables 2–4.

Gene ontology (GO) analysis was carried out to identify the biological function of the DEGs in the mouse and rat stroke transcriptome. Interestingly, most significantly enriched GO terms were distinct at the different time points. While only 10% of the top30 GO terms in the acute phase were stroke-related, the majority of enriched GO terms in the subacute and long-term time point were directly related to the stroke pathology. For instance, inflammation related GO terms (e.g., immune system response, regulation of leukocyte activation) were differentially enriched in the subacute and long-term phase between mouse and rat. In the long-term, differences in the synaptic signaling and synaptic organization were among the top enriched GO terms (Fig. 2B, Suppl. Tables 8–10).

Different stroke-models have distinct gene expression profiles in stroked brain tissue

Acutely after stroke transient MCAo (tMCAo) and permanent MCAo (pMCAo) are the most suitable models to mimic acute human stroke cascade [7]. These methods enable either permanent occlusion of the blood vessels or transient ischemia with reperfusion.

First gene expression datasets from stroked tissue were compared in rodents using tMCAo to pMCAo procedure at the acute (< 24 h) time period. The total number of up- and downregulated DEG was higher following tMCAo acutely after stroke. While 22% of common genes were upregulated in both models only 1% of common genes were downregulated acutely after stroke indicating a unique stroke signature between tMCAo and pMCAo models (Fig. 2C, Suppl. Table 5). Most significantly enriched pathways in pMCAo included detection of chemical stimulus and innate defense responses whereas most enriched pathway in tMCAo included e.g. regulators of vascular and smooth muscle cell responses (Fig. 2D, Suppl. Table 11).

Apart from MCAo models, long-term recovery after stroke can additionally be evaluated using a photothrombotic stroke model (PT). Although the PT model does cause a vasogenic edema acutely after stroke (that is uncharacteristic for human stroke), the method is minimally invasive and results in well-characterized long-term sensory-motor deficits and gradual incomplete recovery [10, 15]. DEG were compared in stroked tissue of pMCAo, tMCAo and PT models at long-term (7–28 d) period. More genes were differentially expressed in pMCAo and tMCAo at a long-term period compared to the acute phase after stroke. A higher overlap could be observed for downregulated and upregulated genes (18%) in tMCAo and PT model, whereas long-term pMCAo gene expression had a highly unique molecular signature compared to the other stroke models (tMCAo: 3%; PT: 3%) (Fig. 2C, Suppl. Table 6). Most significantly enriched pathways in the tMCAo model compared to pMCAo and PT were related to regulators of immune responses e.g., positive regulation of immune system responses, regulation of cytokine production, leukocyte activation indicating a considerably stronger immune activity in tMCAo model. Apart from immune system related pathways, the PT stroke model showed reduced enrichment in remodeling related pathways including, regulation of cell adhesion, actin cytoskeleton organization and tissue and blood vessel morphogenesis compared to the tMCAo model (Fig. 2D Suppl. Tables 12–14).

Stroke gene expression signature is unique in the acute, subacute, and long-term post-injury time periods

As the highest deviation in the transcriptome signature after stroke appears to be dependent on the time point after stroke (Figs. 1D and 2E, Suppl. Table 7), I investigated which pathways were enriched to the different time points.

Most (adaptive) immune-related pathways were strongly upregulated at the long-term time point compared to the acute phase after stroke such as lymphocyte activation, positive regulation to immune response, and adaptive immune response (Fig. 2F, Suppl. Figure 2). However, in the subacute phase many processes involving synapse activation and signaling as well as neurotransmitter secretion were downregulated compared to the acute phase. Interestingly, these pathways e.g., synaptic signaling and synaptic plasticity appear to be re-activated at the long-term phase after stroke potentially indicating endogenous repair and remodeling processes (Fig. 2F, Suppl. Figure 2, Suppl Tables 15–17).

Fig. 2

Gene expression profiles after stroke differ across species. (A) Venn Diagram representing number of shared and differentially expressed genes between mice and rats to different time points. (B) Gene ontology analysis of a subset of most significantly enriched biological processes in rats and mice. (C) Venn Diagram representing number of shared and differentially expressed genes between permanent MCAo and transient MCAo and the photothrombotic stroke model at different time points. (D) Gene ontology analysis of a subset of most significantly enriched biological processes in different stroke models. (E) Venn Diagram representing number of shared and differentially expressed genes between different time periods after stroke. (F) Gene ontology analysis of a subset of most significantly enriched biological processes at different time points