RNA-seq analysis of ischemia stroke and normal brain in a tree shrew model with or without type 2 diabetes mellitus

Nowadays, similar strategies have been used for the treatment and prevention of acute stroke in both diabetes mellitus (DM) and non-DM populations. These strategies were analyzed to provide an experimental basis for the clinical prevention and treatment of stroke in patients both with and without DM. Tree shrews were randomly divided into control, DM, ischemic stroke (IS), and DMIS groups with 18 animals in each group. Serum biochemical indicators were used to assess metabolic status. Neural tissue damage was determined using triphenyl tetrazolium chloride staining, H-E staining, and electron microscopy. Differential gene expression of neural tissue between the DM and control groups and the IS and DMIS groups was measured using RNA-seq analysis. The serum glucose levels of the DM and DMIS groups were significantly higher than other groups. In the DMIS group, the infarct size was significantly larger than in the IS group (19.56 ± 1.25%), with a more obvious abnormal ultrastructure of neural cells. RNA-seq analysis showed that the expression of IL-8, C–C motif chemokine 2 (CCL2), and alpha-1-antichymotrypsin was significantly higher in the DM group than in the control group. The CCL7, ATP-binding cassette sub-family A member 12, and adhesion G protein-coupled receptor E2 levels were significantly higher in the DMIS group than in the IS group. For the prevention and treatment of stroke in patients with DM, reducing the inflammatory state of the nervous system may reduce the incidence of stroke and improve the prognosis of neurological function after IS.


Introduction
Stroke is the leading cause of adult disability and mortality worldwide (Collaborators GBDS 2019). Diabetes mellitus (DM) is an important risk factor for stroke, and patients with DM are 2-6 times more likely to experience a stroke (Jiang et al. 2018;Ma et al. 2018). In addition, approximately 20-30% of in-hospital patients with stroke also had DM, and 90% had type 2 DM (Hill 2014;Beckman et al. 2013) Stroke and DM have similar risk factors, including smoking, a sedentary lifestyle, obesity, a high-fat diet, and alcoholism (Hong et al. 2019). DM is also an important factor in determining post-stroke neural functional recovery. As a major concomitant disease, DM can aggravate brain tissue damage after ischemia, leading to worse neurological outcomes (Bhalla et al. 2013). DM is also correlated with larger infarct size and worse neurobehavioral outcomes (Ma et al. 2018;Hill 2014;Pintana et al. 2019). Patients with acute stroke with a blood glucose level greater than 8 mM have a nearly two-fold mortality rate and a greater likelihood of Ling Zhao and Shiying Huang are first authors and they contributed to this manuscript equally. complications than patients with a blood glucose level lower than 8 mM (Ma et al.a 2018;Romero et al. 2008). DM does not affect the severity of stroke at the initial stage, but is an independent influencing factor for neurological function in the long term (Liu et al, 2020).
DM may also affect the efficacy of drugs used in stroke therapy. For patients with DM, acute thrombolytic therapy was more likely to cause hemorrhagic transformations than for those without DM (Wen et al. 2020). Furthermore, thrombolytic therapy might not be the best treatment for stroke patients with hyperglycemia (Bruno et al 2010). Therefore, the presence of DM affects the therapeutic strategies used to treat stroke. Although current guidelines recommend strategies for treating hyperglycemia during a stroke, these are primarily based on expert opinion (Kruyt et al 2010). Therefore, the development of targeted effective treatment strategies for stroke patients with diabetes is the focus of the current study. High blood glucose affects the integrity of the white matter of the brain, the production of oligodendrocytes, and the reduction of neural plasticity during post-stroke recovery (Hill 2014;Pintana et al 2019). Therefore, high blood glucose levels may cause pro-inflammatory morphology in microglial cells and macrophages. However, the molecular mechanism for this phenomenon is not yet fully understood.
At present, similar strategies have been used for the treatment and prevention of acute stroke in both DM and non-DM populations. However, it is necessary to identify different treatments for stroke patients both with and without DM, as well as appropriate prevention techniques. In this study, tree shrews were used to establish DM and stroke models, and RNA-seq and bioinformatics analyses were performed to detect the gene expression characteristics of brain tissue in each group. Comparisons between the DM and control models and between the DM stroke and control stroke models were made to provide an experimental basis for the identification of clinical prevention techniques and treatments of stroke in patients with DM.

Animal models and groups
Adult male tree shrews weighing 130 ± 30 g were given free access to food and water. The cultivation environment was natural light for 12 h at 25 ± 3 °C. Adaptive feeding occurred one week before experimentation began. The tree shrews were randomly divided into four groups, namely, the control, DM, ischemia stroke (IS), and DMIS, with 18 animals per group (Fig. 1).
The DM model was established using the method described subsequently. High-fat feed (basic feed + 1% cholesterol + 0.1% sodium cholate + 10% lard + 5% egg yolk powder + 5% whole milk powder, Dashuo Chengdu, China) was fed to the tree shrews for eight weeks. After the injection of 2% streptozotocin (STZ, Sigma-Aldrich, USA) into the femoral vein at 100 mg/kg dose, blood glucose changes in the tree shrews were dynamically monitored on days 3, 7, and 14. When the blood glucose exceeded 16.7 mmol/L, the DM model was considered successful and was able to maintain hyperglycemia for four weeks. This model induced insulin resistance and established risk factors, such as high fat and high cholesterol levels, which are close to clinical DM characteristics. Animals in the control and IS groups were injected with sodium citrate.
The thrombotic cerebral IS model was induced using photochemical treatment. After anesthetizing, tree shrews received a 1-cm incision at the midpoint of the connection between the right tragus and the outer canthus of the right eye. The temporalis muscle was separated to expose the skull. A sterile aluminum sheet with a 0.5-cm diameter Fig. 1 The outline of animal grouping and experimental designs circular window hole was embedded in the incision. Then, 1.33 mL/kg Bengal red saline solution was injected into the femoral vein for 10 min. A SQ-III cerebral thrombosis device (Chinese patent number: ZL201420068737.2) was used to irradiate the surface of the skull through the sheet hole with a light intensity of 1.0 W/cm 2 (Fig. 2). This light beam with a 560 nm wavelength (λ) penetrated the skull to produce a photochemical reaction with Bengal red and oxygen in the cerebral blood vessels. This reaction damaged the vascular endothelium and induced thrombosis (Zhao et al. 2019). The regulation of ischemic postconditioning on metabolic disorders of type 2 DM combined with cerebral ischemia in tree shrews was investigated. The control and DM groups were subjected to sham operations without photochemical reaction. The experimental process was approved, supervised, and inspected by the Kunming Medical University Laboratory Animal Ethics Committee.

Serum biochemical indicators detection
After anesthetizing the animals in each group, blood was collected from the main abdominal vein and loaded into a coagulation tube. After centrifugation at 3000 rpm for 10 min, the sample was stored at −80 °C until further use. The biochemical indicators that were measured included glucose, lipids, C-reactive protein, total serum protein, bilirubin, albumin, globulin, alanine aminotransferase, aspartate aminotransferase, urea, creatinine, uric acid, triglycerides, total cholesterol, and high-and low-density lipoprotein cholesterol levels.

Triphenyl tetrazolium chloride (TTC) staining detection
Five models in each group were randomly selected. After 24 h of IS modeling, 10% formaldehyde was infused after anesthetizing. The brain was sectioned coronally, subjected to gradient dehydration, embedded in paraffin, and stained with H-E after sample slicing. In addition, five models were randomly selected in each group. After 24 h of IS modeling, the entire brain was retrieved after anesthesia and craniotomy and placed in 2% TTC solution at 37 °C for 30 min. Brain images were captured and the infarct size was measured.

Detection using transmission electron microscopy (TEM)
Five models in each group were randomly selected. After anesthesia, the brain tissue was quickly retrieved at a low temperature. The infarct tissue or corresponding area of the control group was washed, cut into 1 × 1 × 1 mm sections, and fixed with 3.5% glutaraldehyde at 4 °C. After rinsing with phosphate buffer, samples were placed in 1% tetraoxonium acid at 4 °C for 24 h, dehydrated using an alcoholacetone stepwise gradient and subjected to penetration, embedding, semi-thin sectioning, and positioning using light Fig. 2 The establishment of the cortical thrombotic cerebral ischemia model in tree shrews with an SQ-III cerebral thrombosis device microscopy. The area of interest was ultra-thinly sectioned followed by double staining with lead citrate-uranium acetate and observed using a transmission electron microscope (JEM-1011).

RNA isolation and cDNA preparation
Three models were randomly selected from each group. The infarct tissue was obtained as described above, placed in a pre-cooled 0.1% DEPC cryotube, and stored in a −80 °C refrigerator to prevent RNase contamination. RNA was isolated from the infarct tissue, the concentration was adjusted to 100 ng/μL, and 20 μL of complementary DNA was prepared in a reaction system. The mRNA was broken into short fragments and first-strand cDNA was synthesized using random primers. Buffer, dNTPs, and DNA polymerase I were added to the system to synthesize and purify the secondstrand cDNA.

RNA transcriptome sequencing analysis
Illumina Hi-seq2500 platform was used for RNA-seq detection. All samples were tested for cross-species contamination. The read samples were compared to the tree shrew genome. The spliced unigene of the sample was compared with the publicly available gene expression data and functional annotation was performed through homology. Differential expression analysis between two conditions/groups (two biological replicates per condition) was performed using the DESeq R package (1.10.1). DESeq provides statistical routines for determining differential expression in digital gene expression data using a model based on a negative binomial distribution. The resulting P-values were adjusted using Benjamini and Hochberg's approach for controlling the false discovery rate. Genes with an adjusted P-value < 0.05 found by DESeq were assigned as differentially expressed.

Gene ontology (GO) and KEGG pathway enrichment analysis
GO enrichment analysis of differentially expressed genes was performed using the GOseq R package, in which gene length bias was corrected. GO terms with a corrected P-value of less than 0.05 were considered significantly enriched by differentially expressed genes. KEGG is a database resource for understanding high-level functions and utilities of a biological system, such as a cell, organism, and ecosystem, from molecular-level information, especially large-scale molecular datasets generated using genome sequencing and other high-throughput experimental techniques (http:// www. genome. jp/ kegg/). We used KOBAS software to test the statistical enrichment of differentially expressed genes in KEGG pathways.

Rough comparisons
In the IS and control groups, tree shrews consumed and excreted normal amounts of drinking water and urine, respectively, as well as showed normal activated movement and displayed shiny body hair. In the DM and DMIS groups, the model weights were significantly reduced compared to those in the control and IS groups. The animals in the DM and DMIS groups also had significantly increased water intake and urine volume compared to the control and IS groups. There was no significant difference between the control and IS groups.

Detection of biochemical indicators
The serum glucose levels of the DM and DMIS groups were significantly higher than those of the control and IS groups. Compared to the control group, the DM group had significantly higher levels of total protein, glucose, LDH1, CK-MB, and HBDH. Compared to the IS group, the DMIS group had higher GLOB and glucose levels after DM modeling (Table 1).

TTC staining
After TTC staining, there was no obvious abnormality in the brain tissue of the control or DM groups. In the IS group, obvious infarcts were observed in the lighted area of the cerebral cortex with clear boundaries. The infarct area accounted for 16.16 ± 2.06% of the hemisphere. In the DMIS group, the infarct size was significantly larger than that in the IS group (19.56 ± 1.25%; Table 2; Figure S1A).

H-E staining
After H-E staining, brain tissue in the control group showed an intact structure and the neural cells had normal morphology, abundant cytoplasm, and no obvious nuclear shrinkage. The brain tissue of the DM group was not obviously different from that of the control group. Nerve cells had normal morphology without edema and were in a tight arrangement. In the IS group, neural cells in the cerebral cortex showed a disorderly arrangement, and pyramidal cells showed a triangular morphology and eosinophil staining. The nucleus appeared shrunken and the nucleolus was absent. A gap had formed around the cells and the boundaries of the nucleus and cytoplasm were unclear. The findings in the DMIS group were similar to those in the IS group in terms of structure. Most cells lacked an entire cell structure ( Figure S1B).

TEM detection
In the control group, the neural cells of the cortex were normal in terms of morphology and intact with clear nuclear membrane and nucleolus. The boundary of the neural cells was clear and the chromatin was abundant. The endoplasmic reticulum was arranged regularly and the ribosomes were also clear and intact. In the DM group, the perinuclear space of the neural cells was narrowed and the nucleoli were dense. The heterochromatin of cells was reduced and the mitochondria were swollen. The Golgi complex was expanded and the ribosomes of the rough endoplasmic reticulum were partially swollen and shed. Glial cells were observed around the thickening capillaries ( Figure S2).
In the IS group, the neural cells in the cerebral ischemic area showed pyknosis and some cells showed edema. The chromatin of the cells condensed into clumps and the organelle structure was indistinct. Swollen mitochondria and expanded endoplasmic reticulum were also observed. In the DMIS group, a large number of neural cells displayed pyknosis with irregular nuclei and increased heterochromatin. The cytoplasm was concentrated with an indistinct organelle structure and a large amount of debris. Mitochondria were swollen and the crests dissolved and absent. The endoplasmic reticula were also swollen and expanded ( Figure S2).

RNA-seq analysis
In this study, RNA sequencing was used to harvest transcriptomes of the control, IS, DM, and DMIS groups. A total of 12 sequencing libraries were prepared. The clean bases generated in each group were as follows: control group, 21.09G; IS group, 20.6G; DM group, 20.66G; and DMIS group, 21.81G. After removing the low-quality readings, a clean reading of greater than 95% was obtained on the Phred quality score Q20 level, and a clean reading greater than 90% was obtained on the Q30 level. The GC content was approximately 50% (Fig. 3). Analysis of differentially expressed genes between groups showed that there were 769 differentially expressed genes, including 508 upregulated and 261 downregulated genes between the DM and control groups. The top three upregulated and downregulated genes are listed in Table 3. There were 2206 differentially expressed genes, including 1118 upregulated and 1088 downregulated genes, between the IS and DMIS groups. The top three upregulated and downregulated genes are listed in Table 4.
There was one differential signaling pathway based on upregulated genes (padj < 0.05) between the DMIS and IS groups, which were proteoglycans in cancer. There were seven differential signaling pathways based on downregulated genes (padj < 0.05), including the glutamatergic synapse, axon guidance, and gastric acid secretion (Figs. 5C and D).

Discussion
At present, the prevention and treatment of stroke in patients with DM are similar to those in patients without DM. However, patients with DM have a higher risk of stroke and poorer neurological recovery prognosis than patients without DM. It is, thus, necessary to develop specific stroke prevention and treatment strategies for patients with DM. Although the mechanism is not yet completely understood, the hyperglycemia status may affect the integrity of white matter and the generation of oligodendrocytes after a stroke, as well as affect inflammation in the neural microenvironment and vascular permeability (Zuloaga et al. 2014;Mitome-Mishima et al. 2014;(Venkat et al. 2018). In this study, we innovatively implemented changes in the transcriptome of neural tissue in DM and control tree shrew models and analyzed the changes after stroke in the IS and DMIS models using RNA-seq analysis. This study aimed to provide a theoretical basis for the prevention and treatment of stroke in patients with DM.
Due to the depletion of non-human primate resources, this study adopted tree shrews with biological characteristics similar to those of rhesus monkeys. Tree shrews are the closest animals to humans in terms of anatomical and physiological characteristics compared to other rodents (Kaas et al. 2013). The structure of the brain and distribution of intracranial cerebral blood vessels of tree shrews are close to those of humans and most of the intracortical arteries are linearly distributed with a dense capillary network (Zhang at el. 2017). Therefore, the tree shrew IS model had pathological processes similar to those of clinical stroke patients. In the DM model, blood glucose increased significantly and affected body weight, water intake, and urine volume. After IS onset, the infarct size in the brains of the DMIS group significantly increased compared to that in the IS group. TEM results showed that DM increased the loss in the number of neural cells and the degree of damage.
Under acute stress, the stress-induced hyperglycemia is a common physiological response that usually refers to transient hyperglycemia (Yuan et al. 2021). In the population, stress-induced hyperglycemia occurs in approximately half of patients with acute ischemic stroke, and it is generally believed to be associated with an increased risk of death and poor neurological prognosis (Li et al. 2020). The potential mechanism included hypothalamic-pituitary-adrenal (HPA) axis activation, glucocorticoids release and the sympathetic nervous system activation, that leading to catecholamine-mediated hyperglycemia. But hyperglycemia has a dual role in brain injury. On the one hand, high glucose provides more energy supply for the brain, but it also exacerbates neuronal damage . In this study, the blood glucose of IS group was significantly higher than that of the control group. In further research, it is worth to conduct a study of RNA-seq analysis on whether there is a hyperglycemia after the ischemia stroke to provide clues for intervention of stress-induced hyperglycemia.
In addition to stress-induced hyperglycemia, clinical studies have found that hyperglycemia/ diabetes are also associated with poor prognosis after ischemia or hemorrhagic stroke that including higher mortality, poor neurological functional prognosis, longer hospital stay, readmission and recurrence rate (Lau et al. 2019). In a meta-analysis, diabetes was considered as an independent predictor of recurrence in stroke patients that recurrence rate of stroke patients with diabetes are significantly higher than that of patients without diabetes (HR: 1.45; 95%CI: 1.32-1.59) (Shou et al. 2015). And stroke patients with diabetes have more serious neural tissue damage and larger infarct size, and in-vitro study have confirmed that high glucose levels increase neural cells apoptosis (Bai et al. 2021). It is consistent with the results of our research that compared to IS group, DMIS group has larger infarct area, more neural cells loss, and more server neural tissue damage.
RNA-seq showed that by comparing the DM and control group, the expression of IL-8, CC motif chemokine 2, and alpha-1-antichymotrypsin was increased in the DM group. Two of these three genes were involved in the inflammatory response (Fig. 6A). BP of GO annotation also found that upregulated genes are involved in the immune response and immune system processes, and MF found that genes are mostly involved in cytokine and chemokine activities. IL-8 is a chemokine that has a chemotactic effect on inflammatory cells, such as neutrophils, basophils, and T cells, and is also involved in the activation of neutrophils. The increased expression of IL-8 in the DM model causes inflammatory cells, such as neutrophils, to accumulate in the nervous system, and induces a pre-inflammatory state in the nervous tissue. IL-8 has a potential correlation with the incidence of first stroke (Shishkova et al. 2018;Jenny et al. 2019). This may be one of the mechanisms by which patients with DM have a high risk of IS.
A recent study also found that the number of granulocytes and natural killer cells is related to brain damage after a stroke and that IL-8 could attract these two types of cells. Peripheral IL-8 levels are also considered to be related to infarct size and prognosis (He et al. 2018). In a mouse IS model, the expression of IL-8 in peripheral mononuclear cells could affect the occurrence of IS by regulating the PI3K/Akt/NF-kb signaling pathway. After reducing the expression of IL-8, it reduced the infarct size, activated the neurogliocytes, and improved neural function recovery (Lv et al. 2019). This study further suggests that IL-8 expression in the brain increases in patients with DM, as do the activities of the NF-κB, MAPK, TNF, and PI3K-Akt signaling pathways. Therefore, in patients with DM, the increased expression of IL-8 and the activities of the inflammatory status in the neural system may also be a cause of poor neurological function prognosis after IS onset. Therefore, for the prevention and treatment of stroke in patients with DM, not only blood glucose concentration but also the inflammatory status of the central nervous system must be reduced. This may allow the reduction of the incidence of stroke and improvement of the prognosis of neurological function in patients with DM.
In comparison with the IS and DMIS models, the expression of CCL7, ATP-binding cassette sub-family A member 12, and adhesion G protein-coupled receptor E2 increased (Fig. 6B). Adhesion G protein-coupled receptor E2 promotes chemotaxis and adhesion of granulocytes, promoting the release of inflammatory factors, such as IL-8 and TNFs, and increases the activity of the intracranial inflammatory response. CCL7 can also attract monocytes and eosinophils. In a mouse model, CCL7 expressed in meningeal mast cells could participate in the pathological process of stroke, including the infiltration of granulocytes and activation of macrophages, and further lead to an increase in brain swelling and infarct size (Arac et al. 2014). The CCL7 and adhesion G protein-coupled receptor E2-related high inflammatory response may also be a cause of worse neurological outcomes in patients with DM after IS.
Owing to the use of DM and IS models in tree shrews, there were differences in the actual pathogenesis of patients. However, RNA-seq studies on clinical patients with DM and IS are not yet available. Therefore, this study could be used as a reference for further clinical and basic research. This study was focus on the GO and  The differential signal pathways between the DM and control groups using KEGG analysis. (A: based on upregulated genes; B: based on downregulated genes). The differential signal pathways between the IS and DMIS groups using KEGG analysis. (C: based on upregulated genes; D: based on downregulated genes). The size of the spot reflects the number of genes involved and the color reflects the degree of statistical difference KEGG pathway enrichment analysis based on RNA-seq analysis in design, but not the single gene expression level. So, we ignored the design of RT-PCR verification of differentially expressed genes. This was a limitation of this work.
In conclusion, this study confirmed that in a tree shrew DM model, increased expression of IL-8 and activation of the inflammatory status may be causes of increased IS susceptibility, and high CCL7-related inflammatory response may also be a cause of poor neurological prognosis after IS. Therefore, in the prevention and treatment of stroke in patients with DM, not only reducing blood glucose levels but also reducing the inflammatory state of the nervous system may be a method of reducing the incidence of stroke and improving the prognosis of neurological function after IS.
Acknowledgements Thank Professor Jiejie Dai's laboratory. Thank Yuanyuan Han for providing suggestions and instructions in the process of modelling. Thank Ying Tang for helping us review pathology materials.
Author contributions Ling Zhao and Tingyu Ke: study concept and design, analysis and interpretation of data, drafting of the manuscript, funding acquisition. Ling Zhao, Shiying Huang, and Qiwei Liao: the acquisition of samples or data, statistical analysis, review of the manuscript. Ling Zhao, Xia Li, Shufen Tan, and Shuqing Li: statistical analysis, review of the manuscript, material supports. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Data availability
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.
Code availability Not applicable.

Conflicts of interest
The authors have no relevant financial or nonfinancial interests to disclose.