Gene expression profiling of brain endothelial cells after experimental subarachnoid haemorrhage

Subarachnoid haemorrhage (SAH) is a type of hemorrhagic stroke that is associated with high morbidity and mortality. New effective treatments are needed to improve outcomes. The pathophysiology of SAH is complex and includes early brain injury and delayed cerebral ischemia, both of which are characterized by blood–brain barrier (BBB) impairment. We isolated brain endothelial cells (BECs) from mice subjected to SAH by injection of blood into the prechiasmatic cistern. We used gene expression profiling to identify 707 unique genes (2.8% of transcripts, 403 upregulated, 304 downregulated, 24,865 interrogated probe sets) that were significantly differentially expressed in mouse BECs after SAH. The pathway involving prostaglandin synthesis and regulation was significantly upregulated after SAH, including increased expression of the Ptgs2 gene and its corresponding COX-2 protein. Celecoxib, a selective COX-2 inhibitor, limited upregulation of Ptgs2 in BECs. In this study, we have defined the gene expression profiling of BECs after experimental SAH and provide further insight into BBB pathophysiology, which may be relevant to other neurological diseases such as traumatic brain injury, brain tumours, ischaemic stroke, multiple sclerosis, and neurodegenerative disorders.

RNA was successfully extracted and amplified from mouse brain endothelial cells. There was a relatively low amount of RNA extracted from CD45-CD31 + endothelial cells, so the RNA was amplified in order to perform gene expression profiling. This amplified RNA demonstrated enrichment of endothelial genes but relatively low levels of genes characteristic of other cell types ( Supplementary Fig. S3). In addition, the amplified RNA was enriched in genes characteristic of endothelial cells derived from the BBB (Slc2a1/Glut1, Abcb1a/ Mdr1a), but exhibited relatively low levels of genes characteristic of endothelial cells derived from arteries, veins, or lymphatic vessels (Fig. 3d). A fixed amount of exogenous mRNA plasmid was added to the endothelial cell suspension prior to RNA extraction, confirming similar first strand and amplification efficiencies between biological replicates ( Supplementary Fig. S4). Gene expression patterns pre-and post-amplification were similar ( Supplementary Fig. S4).
SAH and sham mice demonstrated distinct endothelial gene expression patterns. Gene expression profiling of CD45-CD31 + BECs derived from SAH and sham mice showed distinct gene expression patterns based on Pearson's correlation, principal component analysis (PCA), and unsupervised hierarchical clustering ( Fig. 4a-c). Among 24,865 interrogated probe sets, 707 genes (2.8%) showed significant differential expression after SAH with 403 genes upregulated and 304 genes downregulated (Fig. 4d). In particular, 236 genes were significantly upregulated by at least 1.5-fold and 200 genes were significantly downregulated to less than 75% of baseline value. Genes with the highest fold changes are shown in Fig. 5a,b.   Supplementary Fig. S9). Celecoxib did not affect Cox-2 protein expression in brain blood vessels but limited CD45-CD31 + BEC ptgs2 gene upregulation after SAH (Supplementary Fig. S9).

Discussion
Our study used a prechiasmatic injection model of SAH which demonstrated maximal BBB disruption at the 24 h time-point. We isolated BECs from mice with SAH. Using gene expression profiling with a microarray platform, we identified genes that were significantly differentially expressed in BECs after SAH. Pathways relevant to inflammation and prostaglandin signaling were upregulated in these cells. Ptgs2 (Cox-2) was upregulated at both the RNA and protein level and may represent a potential therapeutic target in SAH.
Our study revealed several biologically relevant targets in BECs after SAH (Fig. 5). There was significant upregulation of Ptgs2/COX-2 and we chose to investigate this target further due to overall upregulation of the prostaglandin synthesis pathways. COX-2, an inducible enzyme that generate various prostaglandins, has been well-studied and is druggable with the clinically available selective COX-2 inhibitor celecoxib. Unlike COX-1-derived prostaglandins which provide house-keeping functions and gastric cytoprotection, COX-2-derived www.nature.com/scientificreports/ prostaglandins are upregulated in states of inflammation and cancer in response to cytokines and mitogens 28 . It is not surprising that celecoxib did not affect the protein expression of COX-2 after SAH given that it is merely an inhibitor. Celecoxib may limit inflammation by limiting PGE2 production and signaling through proinflammatory EP2 receptors 29 . However, we were surprised that celecoxib downregulated the mRNA expression of Ptgs2/COX-2. The mechanism of this downregulation is unclear. COX-2 is also known to be constitutively expressed in kidney and brain tissue 30 . It is known that celecoxib can cross the BBB 31 . Patients with SAH have been treated with COX-2 inhibitors without clear adverse reactions 32 . There is concern for increased pro-thrombotic complications including myocardial infarction and ischaemic stroke in the setting of prolonged use of COX-2 inhibitors in patients with vascular risk factors 33 . The VIGOR (Vioxx Gastrointestinal Outcomes Research) clinical trial comparing rofecoxib, a select COX-2 inhibitor and naproxen, a non-selective non-steroidal anti-inflammatory drug (NSAID) in patients with rheumatoid arthritis, found a significant increase in cardiovascular events in patients treated with rofecoxib 34 . The mechanism for the increased pro-thrombotic events was thought to be suppression of prostacyclin (PGI2), a vascular-protective prostaglandin 28 . For celecoxib, the PRECISION trial found no significant increased risk in myocardial infarction or stroke compared with ibuprofen or naproxen 35 . The lack of increased cardiovascular risk associated with celecoxib was also seen in large retrospective clinical studies 36,37 .
COX-2 inhibitors have not been studied extensively in SAH patients. In preclinical studies, NS398, a COX-2 inhibitor, was used as treatment in a mouse endovascular perforation SAH model and was found to be neuroprotective 38 . COX-2 has been shown to be involved in the development of vasospasm via increased expression of endothelin-1 and activation of the JAK-STAT signalling cascade (Janus kinase-Signal transducer and activator of transcription) [39][40][41][42] . Celecoxib treatment was shown to attenuate vasospasm in a preclinical SAH model 39 . Celecoxib is known to antagonize L-type calcium channels, which can cause vasodilation 43 . EP4 is a downstream receptor for PGE2, which is produced by COX-2, and its activation was found to provide neuroprotection and BBB stability, and its antagonism was found to provide the reverse effect in a rat endovascular perforation SAH model 44 . However, EP4 is only one of 4 receptors for PGE2 (EP1, EP2, EP3, EP4), with each receptor having context-dependent proinflammatory versus anti-inflammatory signaling 28 .
COX-2 inhibitors may have additional benefits in SAH, including sodium retention to counteract cerebral salt wasting, and increased blood pressure, which may be a beneficial treatment for delayed cerebral ischemia after the ruptured aneurysm has been secured. A retrospective analysis of the clinical trial CONSCIOUS-1 comparing clazosentan to placebo in patients with SAH found that NSAID treatment during hospital admission was associated with improved clinical outcomes and decreased mortality 32 . Unfortunately, only a small number of patients were administered COX-2 inhibitors, preventing any conclusions regarding specific NSAIDs.
Angiopoietins are vascular growth factors which include Angpt1, an important vascular maintenance regulator through its signalling via the Tie2 receptor, and Angpt2, a vascular disruptive factor due to its partial antagonism of Tie2 signalling 45 . In many inflammatory diseases, Angpt2 is upregulated 45 . Prior studies have shown that hypoxia may upregulate Angpt2 in a COX-2-dependent manner 46 . However, in our study, COX-2 inhibitors did not significantly attenuate upregulation of Angpt2 after SAH. There are likely alternative pathways that lead to Angpt2 upregulation aside from prostanoid signaling.
Our study has several limitations. Although the BEC isolation methods were optimized to provide maximal efficiency, we cannot exclude ex vivo effects on gene expression changes secondary to the isolation process itself. For example, ex vivo exposure to thrombin can upregulate COX-2 expression in endothelial cells 47 . Our study used BECs derived from the left cerebral hemisphere, which had maximal BBB disruption in our SAH model. There may, however, be heterogeneity of gene expression in BECs from different brain compartments. Our immunofluorescence experiments looked at Cox-2 protein expression in vascular structures but without specific colocalization with endothlelial markers. Therefore, it is possible that the increased Cox-2 protein expression may not be related to or exclusive to endothelial cells, but rather vascular smooth muscle cells, pericytes or perivascular macrophages. Although we have focused on endothelial-derived COX-2 signaling as a target after SAH, we have not excluded the potential beneficial effects of COX-2 antagonism in other cell types.
Future experiments could fully address the time-frame of BBB disruption in this prechiasmatic model of SAH including earlier and later time-points. Also, it would be interesting to see if temporal changes in BBB disruption correlate with endothelial Ptgs2/COX-2 expression. Further studies may examine the COX-2 downstream pathways including the expression and activity of mPGES-1, PGE2 and its respective receptors EP1-4.
In conclusion, we isolated BECs in an experimental SAH model and identified COX-2 as a potential therapeutic target. BBB impairment is part of the pathophysiology of several neurological diseases aside from SAH including ischaemic stroke, traumatic brain injury, multiple sclerosis, brain tumours, epilepsy, and neurodegenerative disorders such as Parkinson's disease and Alzheimer's disease. Our hypothesis-generating results from gene expression profiling of BECs may have relevance to these other neurological diseases. The selective COX-2 inhibitor celecoxib is clinically available and may be studied further as a potential treatment for SAH. RNA extraction and microarray analysis. RNA         Decreased mRNA expression of tie2, vegfr2, and mfsd2a in Tie2 + Pdgfr-BECs after SAH. n = 4 per group. mRNA expression levels normalized to Actb (beta-actin). t-test with Holm-Sidak post-hoc correction *p < 0.05, **p < 0.01, ***p < 0.001. Data presented as means ± SEM. (e) Confocal microscopy images of coronal brain slices after SAH or sham procedure with neurons labeled with NeuN (green), Ptgs2 (Cox2) protein labelled in red, and nuclei labelled with DAPI (purple). n = 5 per group. (f) Protein expression of Tie2, Angpt1, and Angpt2 in left hemisphere brain homogenates after SAH or sham procedure using enzyme-linked immunosorbent assay (ELISA) kits. (g) Ratio of Angpt2 over Angpt1 in left hemisphere brain homogenates after SAH or shame procedure using ELISA kits. n = 5.