The expression of ischemia-related genes was determined by comparing the infarct-induced expression (combined samples from infarcted and peri-infarcted areas) to that in the contralateral hemisphere: 77, 92 and 15 genes were de-regulated in stroke-affected regions in the 3 patient survival groups respectively, while 9, 51, 48, 166, 253, 117 and 261 genes were altered at the 7 different time-points in the animal model compared to the controls (Figure 1). The combined number of differentially expressed transcripts in stroke patients represented 6.5%, 7.8% and 1.3% respectively in each survival group of the total number of the genes on the microarray. These findings compare with 0.8%, 4.3%, 4%, 14.1%, 21.5%, 10% and 22.2% of genes respectively at each time-point in rats.

In total, 126 genes were deregulated after stroke in humans and 335 in the rat MCAO model. However, these data are not directly comparable since many transcripts in the human array were not present in the rat array and vice versa. Out of a total of 393 genes present in both arrays, 31, 49 and 5 showed deregulated expression in the 3 patient groups respectively, whilst 7, 27, 26, 62, 107, 34 and 107 genes were deregulated at each of the 7 time-points respectively following rat MCAO (Table 3, Figure 1). Of the 393 overlapping transcripts, the expression of 36 was changed only in the human study, compared with 184 that were altered only in the animal model, while only 41 deregulated genes were shared between the two studies. Interestingly, the mean fold changes in the human data were much higher than in the rat.

Amongst these genes we examined in more detail a small subset with no prior report of a role in stroke (PAK1, MMP11 and INI1). PAK1 was only induced in the human study although present in both microarray sets, MMP11 was induced in both cases, while INI1 was induced in the human but was not present in the rat microarray set. To confirm the microarray data, RT-PCR was carried out on selected deregulated genes. The temporal expression patterns of these genes following RT-PCR showed good agreement with the corresponding expression profiles obtained from the microarray analysis, supporting the validity of the data obtained from the microarrays. Using Western blotting and immunohistochemistry, PAK1, INI1 and MMP11 protein expression and localization was determined in the contralateral and ipsilateral brain areas of individual stroke patients and rats subjected to MCAO, and in HBMEC and HFN exposed to OGD and reperfusion.

In agreement with the microarray data, RT-PCR demonstrated an increase in ini1 mRNA levels in peri-infarcted and infarcted areas of patients who survived between 2 and 6 days following stroke (Figure 2A). Analysis of INI1 protein expression in samples from individual stroke patients showed that protein levels were increased in peri-infarcted and infarcted regions in 8 of 12 samples (Table 4; Figure 2Bi and 2Bii). Only one patient who survived for 3 days after stroke showed decreased protein expression. Cells from contralateral white matter were not stained for INI1 but some weak neuronal cytoplasmic staining was seen in grey matter (Figure 2Ci). An increase in its expression was observed in the cytoplasm of cells with the morphological appearance of glia and microvessels from peri-infarcted and infarcted areas of patients surviving for 3 to 29 days after stroke (Figure 2Cii and 2Ciii). In the rat, RT-PCR and Western blotting demonstrated no notable changes in INI1 mRNA and protein expression respectively following MCAO. Weak cytoplasmic staining was observed in contralateral neurons but no differences in the level of INI1 neuronal expression occurred following MCAO (data not included). Finally, HFN and HBMEC exposed to OGD and/or reperfusion showed no difference in mRNA and protein levels for INI1 when compared with untreated cells.

For MMP11, RT-PCR data agreed with the findings from the microarray study, showing increased mRNA levels in infarcted and peri-infarcted tissue from patients surviving 2–20 days following stroke (Figure 3Ai). Western blotting in individual patient samples demonstrated that 6 of 12 patients had elevated MMP11 protein levels (Table 4; Figure 3Bi and 3Bii). The majority of cells from contralateral grey and white matter were not stained for MMP11 (Figure 3Ci). In patients surviving from 3 days to 4 weeks, endothelial cells and neurons from both infarcted and peri-infarcted tissue were stained positive for MMP11 (Figure 3Cii and 3Ciii). In the rat model, RT-PCR confirmed the microarray data for some of the time-points, showing no notable change in mRNA levels at 1 and 12 h but a prolonged upregulation at 3 days following MCAO (Figure 3Aii). Protein levels were elevated at 1 h, 24 h and 3 days, after which they returned to control levels. No staining for MMP11 was seen in contralateral areas (Figure 3Di), but an increase in its expression occurred in neurons following MCAO, in particular at 12 and 24 h (Figure 3Dii). MMP11 mRNA and protein levels remained unchanged in HFN and HBMEC exposed to conditions of oxygen-glucose deprivation.

RT-PCR confirmed the upregulation of pak1 determined by the microarrays in pooled samples from stroke patients who survived between 2 and 6 days following stroke (Figure 4A). Western blotting showed an upregulation in the protein levels of PAK1 in 6 of 12 patients (Table 4; Figure 4Bi and 4Bii). No staining was seen in contralateral white matter, while, in grey matter, PAK1 stained weakly the cytoplasm of some neurons (Figure 4Di). In patients surviving for 3 days to 4 weeks after stroke, increased PAK1 nuclear staining was seen in neurons in both peri-infarcted and infarcted regions (Figure 4Dii). In the rat, RT-PCR showed no significant changes in the mRNA levels for pak1 at most of the time-points examined. However, Western blotting showed an upregulation in protein levels 1, 12 and 24 h after MCAO, returning to control levels at 3 days, and becoming downregulated at 7 days following MCAO (Figure 4Ci and 4Cii). Weak staining was observed in neurons from the contralateral hemisphere (Figure 4Ei), but an increase in cytoplasmic and nuclear staining in neurons occurred following MCAO, in particular at 1 h and 24 h (Figure 4Eii). Finally, an increase in PAK1 expression was also seen in human foetal neurons following oxygen-glucose deprivation (Figure 4Fi and 4Fii).

Human brain tissue samples were obtained from 12 patients who died from acute ischaemic stroke, with the approval of the local Ethics Committee and Brain Bank at the Department of Neuropathology, Collegium Medicum, Jagiellonian University, Krakow, Poland. All patients were admitted with large middle cerebral artery strokes confirmed by CT-scan or MRI. The patients, 10 male and 2 female, were aged between 51 and 86 years and had survived between 2–37 days following ischaemic stroke (Table 2). Routine blood parameters were determined on admission. Full clinical examinations, including NIH Stroke Scale, were also carried out on admission. Excluded from the study were patients with recent history of head trauma, major cardiac, renal, hepatic or cancerous disease and obvious signs of infection after admission. Immediately after death the body was put in a cold chamber and tissue was collected within 6 h of death. Tissue samples were taken from infarct and peri-infarcted zones while controls were obtained from the contralateral hemisphere at the same time. The peri-infarcted areas were defined in tissue sections as the tissue immediately surrounding the infarcted core which contained some necrotic cells and showed evidence of tissue disorganisation confirmed by histology. Sections were stained with 2,3,5-triphenyltetrazolium chloride which stains active mitochondria pink; therefore, non stained areas represented stroke affected cortical regions (data not included). Tissue specimens were immediately frozen in liquid nitrogen, kept at -70°C and a portion of each sample was processed for histology and stained with haematoxylin and eosin to determine tissue morphology [51].

Stroke experiments were performed on female Sprague-Dawley rats (weight: 230–270 g) as they suffer less than male during ischaemia. Cerebral ischaemia was produced using a modified method of Baron [52] by distal, permanent occlusion of the MCA by electrocautery as described elsewhere [53,54]. The mortality in this model is very low. Sets of six animals (3 for morphological studies and 3 pooled together) for each time-point were sacrificed at 1 h, 4 h, 12 h, 24 h and 3, 7 and 21 days.

Human brain microvascular endothelial cells (HBMEC) were obtained from TCS CellWorks (Buckingham, UK) and cultured according to the supplier's instructions. Human foetal (cerebral cortical) neurons (HFN) were extracted and cultured with permission from the Local Ethics Committee. Brain tissue from foetus specimens of 14–19 weeks gestational age, legally aborted and with the appropriate written consent, were collected in cold preservation medium and cells were isolated and cultured as described elsewhere [55]. For OGD experiments, the culture medium was replaced by glucose-free medium containing 2% foetal bovine serum (TCS CellWorks, Buckingham, UK) and cells were cultured at 37°C in a humidified chamber with 94% N₂, 1% O₂, and 5% CO₂ for 6 h (HBMEC) or 95% N₂ and 5% CO₂ for 14 h (HFN) followed by 24 h reperfusion in fresh medium containing 4.5 g/l glucose. This resulted in approximately 30% of cells undergoing apoptosis after OGD and 60% following reoxygenation, as determined from our pilot studies. Cells cultured in normoxic conditions without glucose deprivation were used as controls. In some experiments, propidium iodide (10 μg/ml) was added to the cultures 1 h before the end of the experiment to stain dead and dying cells.

We established mRNA expression profiles of the damaged brain tissues between 2 to 6 days, 9 to 20 days, and 26 to 37 days after stroke in human patients and 1, 4, 12, 24 hours and 3, 7 and 21 days after the ischaemic insult in rats. The corresponding samples from the non-ischemic control hemisphere were used to measure the normal mRNA abundance of the modulated genes in each tissue at each time point. RNA from three stroke patients was pooled for each patient survival group while RNA from three MCAO rats was also pooled at each time-point to improve yields in preparation of poly A⁺ RNA. Although pooling was previously thought to affect data quality, Kendziorski et al. [56] have recently shown that inference was not adversely affected by pooling. The different patient groups were selected to match the three physiological stages following stroke i.e. the inflammatory (lasting up to a maximum of 5–6 days), the proliferative (lasting up to three weeks following stroke) and the remodelling/maturation (starting during the third or fourth week).

RNA was extracted according to the manufacturer's protocols (BD Biosciences, Oxfordshire, UK) and its quality was measured spectrophotometically. The protocol recommended by Clontech in their Atlas 1.2 microarray kit was used without any modification. Briefly, RNA was reverse-transcribed to cDNA, ³²P-labelled and applied to the array for overnight hybridisation at 68°C. Following washing, the array was exposed to a phosphorimaging plate for 12–72 hours and data analysis was performed using the AtlasImage 1.5 software. The results were normalized using two housekeeping genes, ubiquitin and glyceraldehyde 3-phosphate dehydrogenase (GAPDH). As in the majority of microarray studies mentioned before, only those genes upregulated > 2-fold or downregulated < 0.5-fold were counted as deregulated and taken into consideration. The microarray data are available in Gene Expression Omnibus under the accession number GSE9391.

Gene expression was examined by semi-quantitative RT-PCR with standard reaction conditions of a 10 min denaturation at 94°C, followed by 35 cycles of 1 min at 94°C, 1 min at primer-specific annealing temperatures (Table 5) and 1 min at 72°C and a final 10 min extension step at 72°C. Samples without cDNA were used as negative controls and the products were visualized by agarose gel electrophoresis (1.5% w/v) and DNA stained with ethidium bromide (10 mg/ml). All experiments were carried out at 25, 30, 35 and 40 cycles to ensure the semi-quantitative nature of the results. The results were normalized using housekeeping gene GAPDH and semi-quantitavely analyzed using Scion Imaging Software version 4.02 (Scion Corporation, Maryland, USA). Sense and antisense oligonucleotide primers containing 18–27 nucleotides based on previously reported mRNA sequences in the GenBank depository were designed with the aid of the Primer3 Output Program (Version 0.2). InVitrogen plc. (Paisley, UK) synthesized the primer sets (Table 5).

Proteins were extracted from tissues and the protein concentration of each sample was determined using the BioRad assay. For Western blotting, 10 μg of protein were separated by SDS-PAGE (13% w/v) and the proteins were electro-blotted onto nitrocellulose filters as described previously [57]. Filters were blocked in 1% w/v bovine serum albumin (BSA) in Tris-buffered saline Tween (TBS Tween) and stained overnight at 4°C with antibodies to the following proteins (obtained from Autogen Bioclear, Wiltshire, UK, unless stated otherwise) diluted in 1% BSA: MMP11 (CalBiochem; 1:500), PAK1 (1:500), INI1 (1:500), and α-actin (Sigma, 1:1000) used as a loading control. Membranes were washed in TBS-Tween before staining with the appropriate peroxidase-conjugated secondary antibody, diluted 1:1000 in 5% w/v milk in TBS-Tween for l h. Blots were developed with the ECL detection system (Amersham, UK). The relative intensities of the bands were measured in an LKB densitometer. Results are semi-quantitative and are given as a numerical (fold) change compared to the control (contralateral tissue) which was given an arbitrary value of 1.0. All experiments were performed twice and a representative example of patient(s) showing an increase in protein expression is given.

Paraffin-embedded tissue samples were processed and serial 5 μm sections were cut. The Avidin-Biotin Peroxidase (ABC Vectastain kit, Vector Laboratories, Peterborough, UK) method was used and antibodies to MMP11, PAK1 and INI1 were used at a dilution of 1:50. Paraffin-embedded sections were deparaffinized, rehydrated and boiled for 10 min in an antigen unmasking solution of concentrated citric acid pH 6.0 as described elsewhere [57]. Slides were incubated in 0.5% v/v H₂O₂ in methanol for 30 min, with normal serum for 20 min and then with a primary antibody (diluted in normal serum) for 30 min, followed by 30-min incubation with biotinylated secondary antibody (diluted 1:50) and finally with ABC complex (diluted 1:50) for 30 min at RT. Staining was completed after incubation with DAB substrate chromogen solution for 3–10 min. Slides were counterstained with haematoxylin, dehydrated, cleared and mounted in DPX. For immunofluorescence, cultured cells were fixed in 4% paraformaldehyde for 20 min, permeabilized with 0.2% Triton ×100 for 10 min, blocked with normal serum and stained with the primary antibody as above, followed by 1 h incubation with Alexa-fluor conjugated dye at RT. Negative control slides were performed in parallel, where primary antibody was replaced with washing buffer and processed as above (data not included).