Silencing lncRNA HOTAIR improves the recovery of neurological function in ischemic stroke via the miR-148a-3p/KLF6 axis

Yiwen Huang Guangzhou First People's Hospital, School of Medicine, South China University of Technology Yuanyuan Wang Guangzhou First People's Hospital, School of Medicine, South China University of Technology Xiaobin Liu Guangzhou First People's Hospital, School of Medicine, South China University of Technology Yingjun Ouyang (  yingjunoy0728@163.com ) Guangzhou First People's Hospital, School of Medicine, South China University of Technology


Introduction
Stroke is one of the leading causes of mortality and disability around the world and ischemic stroke (IS) takes up about 87% of all strokes 1 . It occurs when brain blood ow stops and then leads to reduced supply of oxygen and glucose to cells in the brain 2 . The patients of IS can display neurological impairment such as paralysis, language disorders, facial and limb incoordination 3 . Various risk factors, including high blood pressure, diabetes mellitus, obesity, high blood cholesterol, smoking, atrial brillation, and previous transient ischemic attack, are involved in IS 4 . However, current measures to protect the brain against IS are still insu cient 5 . Therefore, an underlying mechanism of cerebral ischemic injury is urgently needed.
Long-non-coding RNA (LncRNA) is a transcript of RNA with a length of over 200 nucleotides but without an open reading framework 6 . LncRNAs play important roles in regulating the expression level of proteincoding genes and related signaling pathways involved in development of multiple diseases 7 . Study has reported that lncRNAs are involved in IS and mediate ischemic neuronal death in stroke 8 . LncRNAs homeobox antisense non-coding RNA (HOTAIR) has been reported to promote ischemic infarct induced by hypoxia 9 . Study has reported that HOTAIR aggravates H 2 O 2 -induced injury in cardiomyocytes and exacerbates myocardial ischemia-reperfusion (I/R) injury 10 . Evidence has con rmed that lncRNAs act as competing endogenous RNAs (ceRNAs) by competing for binding to microRNAs (miRs) 11 . It has been reported that HOTAIR acts as a ceRNA for miR-126-5p to modulate rab3a interacting protein expression in Parkinson's disease 12 . But the ceRNA network of HOTAIR in IS is elusive. miRs are small, endogenous, single-stranded, non-coding RNA molecules ranging in length from 18-25 nt 13 . Previous study has demonstrated that miRs contribute to IS pathogenesis, and miRs participate in the pathogenesis of ischemic brain injury in rodent stroke models 14 . In patients with acute IS, miR-148a was decreased 15 . miR-148a also protects murine liver from alleviates hepatic I/R injury 16 . Furthermore, miR-148a-3p participates in oxidative stress injury induced by I/R 17 . But, the role of miR-148a-3p in IS has not been investigated yet.
Whether HOTAIR can affect miR-148a-3p and participate in the neurological recovery in IS is largely unknown. And there is no report concerning this research topic at home and abroad. To address this gap in knowledge, in this work, we investigated the role of HOTAIR in the neurological recovery in IS and explored the potential molecular mechanism.

Ethic statement
All experiments were authorized by the Ethical Committee of Guangzhou First People's Hospital, School of Medicine, South China University of Technology.
The animal experiments were conducted following the guidelines for the care and use of laboratory animals.
All procedures were strictly implemented by the code of ethics.
Establishment and grouping of permanent middle cerebral artery occlusion (pMCAO) mice A total of 96 adult male C57BL/6 mice (6-8 weeks, 20-28 g) were purchased from Vital River Laboratory Animal Technology [SYXK (Beijing) 2017-0033, Beijing, China]. The mice were placed in separate cages at 22°C ± 2°C and 65 ± 5% relative humidity on a 12 h light/12 h dark cycle with free access to food and water. The pMCAO mouse model was established 18 . Mice were anesthetized with 2% pentobarbital sodium (50 mg/kg) (Sigma Chemical Co., St. Louis, MO, USA) by intraperitoneal injection. The right common carotid artery (CCA) was exposed and ligated using (5/0) surgical suture through a ventral midline neck incision.
Mice were placed in left lateral decubitus position. An incision was made from the lateral orbit to the external auditory meatus, the base the parotid gland and the upper part of the temporal muscle were isolated to expose the cortical branches of middle cerebral artery (MCA) under the right side of skull. A craniotomy was performed directly above the distal part of the MCA using a 0.8 mm high-speed micro drill. Coagulator was used to cauterize the exposed cortical branches of MCA on the right side of the brain. The operation should be performed with caution to avoid damage to the surrounding brain tissue.
The incision was sutured using conventional method. Mice should be kept warm during operation. The sham-operated mice only received the separation of cervical vessels and exposure of skull and drilling holes without ligation of right CCA and cauterization of corresponding arteries. The neurological function score and cerebral infarct area were analyzed 24 h after model establishment.
Mice were randomly assigned into 4 groups (N = 24 in each group): sham group, pMCAO group, pMCAO + sh-HOTAIR group, and pMCAO + sh-NC group. The low expression lentivirus shRNA HOTAIR (sh-HOTAIR) and its control (sh-NC) were constructed and packaged by GenePharma (Shanghai, China). Lentivirus processing was applied 2 weeks before pMCAO operation. Mice were anesthetized and xed in a stereotactic apparatus (RWD Life Science Co., Ltd, Shenzhen, China). The lentivirus was injected via the right lateral ventricle at the following coordinates: 0.3 mm posterior to bregma, 1 mm right lateral to sagittal suture, 2.2 mm deep to the dura. The 5 µL lentivirus was slowly injected at a rate of 1 µL/min using a Hamilton microinjector, and the needle was retained in place for 5 min following the injection. The titer for sh-HOTAIR and its control were 3 × 10 8 CFU/mL 19 .

Neurobehavioral evaluation
Mice were evaluated for neurological function after 24 h of pMCAO modeling based on the modi ed neurological severity score (mNSS) 20 in a double blind manner from perspectives of movement, sensation and re ex. Evaluation standards were: motor tests (1-9 points) including raising mice by the tail, placing mice on the oor and abnormal movement; sensory tests (1-2 points) including placing test (visual and tactile test) and proprioceptive test; re ex tests (1-3 points) including pupil re ex, corneal re ex and startle re ex. The mNSS score was divided into: minor injury (1-4 points), moderate injury (5-9 points) and severe injury (10-14 points). Higher score meant severer de cit.

Beam balance test
Beam balance test was applied to observe the ability of coordinating muscles and keeping balance of mice 20 . The wooden beam (length 1 m, width 5 mm, diameter 15 mm, horizontally positioned 1 m above the underlying surface) was supported by platforms at both sides. The mouse was placed at the start of the beam and scored based on its performance on the beam. The three-day trial was performed twice a day with a 5 min interval. The average value was taken as the experiment result. Performance on each trial was scored as follows: reach the platform or balance in steady posture, 0 point; grasp one side of the beam, 1 point; hang the beam and one limb falls off the beam, 2 points; hang the beam and two limbs fall off the beam or rotate on the beam for over 60 s, 3 points; try to keep balance on the beam for over 40 s but fall, 4 points; try to keep balance on the beam for over 20 s but fall, 5 points; fall with no intention to keep balance, 6 points. Higher score meant worse balance ability.

Morris water maze test
Water maze test was applied to evaluate the learning-memory, spatial location and orientation capability of mice 21 . Titanium dioxide (Shanghai Jianghu Titanium Dioxide Chemical Industry Co., Ltd, Shanghai, China) was added into the water at concentration of 125 g/L. Morris water maze system (Mobile Datum, Shanghai, China) was applied for animal behavior analysis and water temperature was kept at 25°C. The tests included two parts: orientation navigation experiment and space exploration experiment. In the orientation navigation experiment, mice were placed in the water and the time (s) to nd the platform was recorded (escape latent period). If the mice were unable to nd the platform within 60 s, the experimenter guided the mice to the platform and time was recorded as 60 s. Each mouse was trained for 4 times per day with 15-20 min interval for 4 days. In the space exploration experiment, on day 5, the platform was removed. Mice were placed in the water randomly and swam for 60 s. The time of the mice crossing the original platform was recorded. 2,3,5-triphenyltetrazolium chloride (TTC) staining Mice were euthanized by injection of 800 mg/kg pentobarbital sodium after beam balance test. Brain was taken out and made into continuous slices at 2 mm. The slices were incubated in 2% TTC solution (2530-85-0, Guidechem, Shanghai, China) at 37°C in the dark for 30 min. Normal brain tissues showed a pink or red color, and infarcted tissues were in light grey. TTC-stained slices were photographed using a digital camera. The cerebral infarcted area was calculated using Image J (National Institutes of Health, Bethesda, MD, USA) software. TdT-mediated dUTP Nick-End Labeling (TUNEL) The brain tissues were cut into sections at 5 µm and dewaxed by xylene and dehydrated with ethanol, and treated according to the instructions of TUNEL kit (C1089, Beyotime, Shanghai, China). Brie y, 20 µg/mL protease K working solution (Solarbio, Beijing, China) was added at room temperature for 15 min.
The sections were washed using phosphate buffered saline (PBS) for 3 times, and then added with 50 µL TUNEL detection solution and incubated in the dark for 60 min. After being washed by PBS for 3 times, the samples were added with primary antibody NeuN (ab177487, Abcam Inc., Cambridge, MA, USA) and the Alexa Fluor® 488-labeled secondary antibody (ab150077, Abcam). Then sections were counterstained using 4,6-diamidino-2-phenylin-dole (DAPI). The images were obtained by a uorescence microscope (Leica, Wetzlar, Germany). The apoptosis rate = apoptotic cells (red)/total cells (blue) * 100%. Oxygen-glucose deprivation (OGD) model establishment The in vitro OGD model was established as previously described 22 . Cells in good growth conditions were cultured in glucose, serum and oxygen free DMEM with 95% N 2 and 5% CO 2 at 37°C for 3 h. The cells were washed with PBS and then added with 10% FBS high glucose medium and cultured with 5% CO 2 at 37°C for the following experiments.

Cell counting kit-8 (CCK-8) assay
Cells from each group were extracted and made into single cell suspension and seeded in 96-well plates (1×10 4 cells/well) with 100 µL per well for 24 h. CCK-8 solution (Dojindo Molecular Technologies, Kumamoto, Japan) was added (10 µL/well) into each well, and the cells were incubated for 2 h at 37°C in a 5% CO 2 incubator. The absorbance was detected using an EnVision multimode plate reader (Plate Reader, PerkinElmer, Waltham, MA, USA) at 450 nm and relative cell viability was calculated.

Flow cytometry
The cells were detached by trypsin. A total of 1 × 10 5 cells/mL suspended cells were harvested and detected using Annexin V-FITC cell death detection kit (BD Bioscience, Franklin Lakes, NJ, USA). First, 100 µL 1× Annexin buffer was used to suspend cells. Then 5 µL Annexin V-FITC and 1 µL propidium iodide were added to stain the cells for 15 min at 4°C without light exposure. Apoptosis was then detected on a ow cytometer (FACScan, BD Biosciences).

Fractionation of nuclear and cytoplasmic RNA
The PARIS™ kit (#AM1921, Thermo Fisher Scienti c) was applied for fractionation of nuclear and cytoplasmic RNA to detect expression of lncRNA HOTAIR in nucleus and cytoplasm of mouse N2a cells.
The expression of HOTAIR was analyzed by RT-qPCR. Florescence in situ hybridization (FISH) Firstly, online software LncATLAS (http://lncatlas.crg.eu/) was applied to predict the distribution of HOTAIR. Then, FITC-labeled HOTAIR was designed and synthesized by GenePharma. Nuclei were stained with DAPI. All processes were conducted using the FISH kit (GenePharma) following the manufacturer's instructions. Images were obtained by a FV3000 microscope (Olympus, Tokyo, Japan).

Bioinformatics analyses
HOTAIR related miRs were retrieved using starBase database (starbase.sysu.edu.cn/) and RAID database (http://www.rna-society.org/raid2/index.html). IS expression chip GSE35338 of mouse samples was obtained through GEO database (https://www. ncbi.nlm.nih.gov/geo/), including 10 normal samples and 11 disease samples. The "limma" package in R language was used for differential analysis. False discovery rate (FDR) was used to rectify P value, and differentially expressed genes from IS mouse model were screened out with |logFC|>1 and adj.p.val < 0.05 as criteria. The miR-148q-3p target gene in human and mice was predicted using TargetScan database (http://www.targetscan.org/vert_71/) and binding sites were obtained. Interaction analysis of 11 candidate genes was performed using GeneMANIA database (http://genemania.org/), and then interaction scores were obtained. Dual-luciferase reporter assay The potential binding sites of lncRNA HOTAIR and miR-148a-3p were predicted using RNA22 v2 (https://cm.jefferson.edu/rna22/Interactive/) 24 . The potential binding sites of miR-148a-3p and kruppel-like factor 6 (KLF6) were predicted using starBase (http://starbase.sysu.edu.cn/index.php) 25 . The wild type (wt) and mutant type (mut) sequences of HOTAIR and KLF6 were cloned to pmiR-GLO luciferase vector (Promega, Madison, WI, USA) to construct HOTAIR-wt, KLF6-wt and HOTAIR-mut and KLF6-mut. HEK293T cells (Shanghai Institute of Biochemistry and Cell Biology, Chinese Academy of Sciences, Shanghai, China) were transferred with the constructed plasmids and mimic NC and miR-148a-3p mimic (GenePharma), respectively using lipofectamine 3000 (Thermo Fisher Scienti c). Luciferase activity was detected after 48 h by the Dual-Luciferase Reporter Assay kit (Vazyme, Nanjing, China) following the manufacturer's instructions. RNA pull-down The biotinylated mut or wt miR-148a-3p (GenePharma) or its negative control (NC-bio) was transfected into the cells. The cells were lysed using lysis buffer at 48 h post-transfection. Then, streptavidin agarose beads (Invitrogen, Carlsbad, CA, USA) were incubated with the cell lysates, and the RNA complex that bound to the beads was eluted and puri ed using TRIzol reagent (Takara, Shiga, Japan). Gene level in the RNA complex was analyzed by RT-qPCR 26 .

RT-qPCR
Total RNA and miRs were extracted separately from mice brain tissues and cells with TRIzol reagent and miRNeasy Mini Kit (Invitrogen). cDNA was synthetized using M-MLV reverse transcriptase (Invitrogen) or TaqMan MicroRNA Reverse Transcription Kit (Thermo Fisher Scienti c). RT-qPCR was performed using SYBR Green Real-Time PCR Master Mix (Thermo Fisher Scienti c) or TaqMan miRNA assay kit on Applied Biosystems7900 Real-Time PCR System. GAPDH and U6 were used as internal controls. Relative expression levels were calculated using the 2 −△△ct method. Ampli ed primer sequences are illustrated in Table 1.

Western blot
Total proteins from homogenate and cells were extracted with radio immunoprecipitation assay (RIPA) lysates (Beyotime). Proteins were quantitatively detected by a bicinchoninic acid kit (Thermo Fisher Scienti c). Extracted protein was denatured at 100°C for 5 min. A total of 50-100 µg protein was separated by sodium dodecyl sulfate-polyacrylamide gel (SDS-PAGE), and then proteins were transferred onto polyvinylidene uoride (PVDF) membranes and incubated with 5% skim milk for 2 h, followed by Tris-buffered saline Tween (TBST) washing for 3 times. The primary antibody p-STAT3 (1:20000, ab76315, Abcam) or GAPDH (1:1000, ab9485, Abcam) was added and the membranes were incubated overnight at 4°C. After washing in TBST, the secondary antibody Goat Anti-Rabbit IgG H&L (HRP) (1:5000, ab97051, Abcam) was added and the membranes were incubated for 2 h at 37°C and detected using chemiluminescent method. Grey value analysis was performed using Image J (National Institutes of Health). Statistical analysis Data analysis and drafting were performed using SPSS 21.0 software (IBM Corp. Armonk, NY, USA) and GraphPad Prism8.0.1 (GraphPad Software Inc., San Diego, CA, USA). Data were in normal distribution and presented as mean ± standard deviation (SD). Differences between groups were analyzed using oneway analysis of variance (ANOVA) or two-way ANOVA, followed by Tukey's multiple comparisons test or Sidak's multiple comparisons test. P value was obtained using a two-sided test. P < 0.05 was considered a statistically signi cant difference.

Results
LncRNA HOTAIR inhibited recovery of neurological function in pMCAO mice IS is attributed to an abrupt reduction of cerebral blood ow, resulting in hypoxia that triggers neuroin ammation, which could result in disability and even death 27 . LncRNAs play critical roles in occurrence and development of IS 1 . In order to explore the function of lncRNA in IS, lncRNAs related to IS were retrieved in LncDisease database (http://www.rnanut.net/lncrnadisease) 28 (Fig. 1A) and nally 2 candidate lncRNAs were obtained. The role of HOTAIR in the recovery of neurological function in IS is still not clear. To further investigate the role of HOTAIR, pMCAO mouse model was established rst, and neurological score was assessed using mNSS method. The pMCAO mice developed severe neurological de cit. The beam balance test suggested that pMCAO mice had poor ability to maintain balance. The escape latency in water maze test was obviously increased and the times crossing the platform were signi cantly decreased (Fig. 1B-D, P < 0.01). These results indicated the success in pMCAO mouse model.
Subsequently, the expression of HOTAIR in cerebral tissues at 24 h after pMCAO model establishment was detected by RT-qPCR. The result indicated that the expression of HOTAIR in pMCAO mice was increased signi cantly. Then, lentivirus was injected to the mice to silence HOTAIR expression. The result showed that the expression of HOTAIR was decreased signi cantly using the shRNA HOTAIR (Fig. 1E, P < 0.01). The scores of neurological de cits and motor balance were decreased in mice with silencing HOTAIR, along with decreased escape latency and increased platform crossing times ( Fig. 1B-D, all P < 0.01). In addition, the infarcted area in pMCAO mice was increased signi cantly, accompanied with increased apoptosis of neural cells, which were decreased after lowering HOTAIR expression (Fig. 1F-G, all P < 0.01). In ammation in cerebral tissues contributes to brain injury in IS 29,30 . Therefore, we examined the in uence of HOTAIR on in ammation in pMCAO mice and found that levels of TNF-α and IL-6 were increased while IL-10 level was decreased; but levels of TNF-α and IL-6 were decreased, and IL-10 level was increased in cerebral tissues with silencing HOTAIR (Fig. 1H, all P < 0.01). Above results demonstrated that silencing HOTAIR suppressed apoptosis and in ammation, attenuated neurobehavioral de cit and promoted the recovery of neurological function of pMCAO mice.

Silencing HOTAIR promoted viability of OGD-injured N2a cells and attenuated apoptosis and in ammation
We simulated IS in vitro using OGD-injured N2a cells to further investigate the role of HOTAIR in IS. The reduction of cell viability injured by OGD indicated the success in establishing OGD model ( Fig. 2A, P < 0.01). The expression of HOTAIR was detected by RT-qPCR and the result showed that HOTAIR expression in ODG cells increased signi cantly. Then, we lowered the expression of HOTAIR in OGD cells using shRNA HOTAIR (Fig. 2B, P < 0.01), and silencing HOTAIR increased remarkably the cell viability ( Fig. 2A, P < 0.01). Result of ow cytometry showed that apoptotic rate of OGD-induced cells was increased and silencing HOTAIR reduced apoptosis (Fig. 2C, P < 0.01). ELISA showed higher levels of TNF-α and IL-6 and lower level of IL-10 in OGD cells, and lower levels of TNF-α and IL-6 and higher level of IL-10 in cells with silencing HOTAIR expression (Fig. 2D, all P < 0.01). In conclusion, silencing HOTAIR promoted viability of OGD-induced neurons and attenuated apoptosis and in ammation HOTAIR regulated KLF6 expression via targeting miR-148a-3p Study has found that HOTAIR plays a regulatory role through ceRNA mechanism 31 . LncATLAS database (http://lncatlas.crg.eu/) predicted that HOTAIR was mainly located in the cytoplasm of N2a cells, which was also proved by the experiment results ( Fig. 3A-B). The downstream miRs of HOTAIR were further predicted by RAID and starBase databases (Fig. 3C) and miR-148a-3p and miR-326 were obtained. It has been found that miR-148a-3p expression was low in cerebral I/R 17 . IS chip GSE35338 was retrieved in GEO database and differential analysis revealed that most genes were highly expressed in IS samples (Fig. 3D). Because the samples of GSE35338 chip were mice, therefore, we predicted downstream regulatory target genes of miR-148a-3p in both human and mice. Meanwhile, the signi cant up-regulated genes in GSE35338 chip were intersected and nally 11 candidate genes were obtained (Fig. 3E).
miR-148a-3p downregulation partially reversed the effect of sh-HOTAIR on N2a cells post-ischemia To explore the effect of miR-148a-3p, we transferred miR-148a-3p inhibitor to N2a cells with silencing HOTAIR expression and succeeded in lowering miR-148a-3p expression (Fig. 4A, P < 0.01). Then the cells were induced by OGD. The result of CCK-8 demonstrated that the viability of N2a cells in sh-HOTAIR + inhibitor group was decreased (Fig. 4B, all P < 0.01). Flow cytometry showed increased apoptotic rate after inhibiting miR-148a-3p expression (Fig. 4C, P < 0.01). The levels of TNF-α and IL-6 were increased and IL-10 level was decreased in sh-HOTAIR + inhibitor group (Fig. 4D, all P < 0.01). These ndings revealed that miR-148a-3p downregulation reversed the effect of silencing HOTAIR on promoting N2a cell viability and attenuating apoptosis and in ammation.

KLF6 overexpression suppressed the effect of sh-HOTAIR on OGD-induced cell injury
To further verify the function of KLF6, we transferred pcDNA-KLF6 or pcDNA-NC into cells with silencing HOTAIR and then cells were treated by OGD to induce injury. RT-qPCR results demonstrated that the mRNA expression of KLF6 was upregulated (Fig. 5A, P < 0.01). Moreover, CCK-8 assay found that overexpression of KLF6 signi cantly repressed the viability of N2a cells (Fig. 5B, P < 0.01). Flow cytometry results revealed that overexpression of KLF6 led to the rising of apoptotic rate in N2a cells ( Fig. 5C, P < 0.01). ELISA showed that overexpression of KLF6 caused increases in TNF-α and IL-6 levels and decrease in IL-10 level (Fig. 5D, P < 0.01). Taken together, these results demonstrated that overexpression of KLF6 inhibited the viability of cells with silencing HOTAIR, increased apoptosis and aggravated in ammatory response.
HOTAIR blocked the activation of STAT3 signaling pathway via miR-148a-3p/KLF6 It has been reported that KLF6 inhibits the expression of anti-in ammatory genes by suppressing STAT3 signaling, which is involved in angiogenesis, neurogenesis and functional recovery after IS 32,33 . We hypothesized that STAT3 signaling pathway may be involved in IS. Western blot was performed to detect the expression of p-STAT3 protein. The results demonstrated that the expression of p-STAT3 was decreased in vivo and in vitro ischemic models and silencing HOTAIR expression increased its expression while miR-148a-3p downregulation or overexpression of KLF6 reduced p-STAT3 expression (Fig. 6A-B, all P < 0.01). These results suggested that HOTAIR may affect the neurological injury in IS via the miR-148a-3p/KLF6 axis and the STAT3 signaling pathway.

Discussion
IS can result in a series of sequences, including cognitive impairment and epilepsy 34 . It has been suggested that lncRNAs may become a potential target for therapeutic intervention in ischemic brain injury 35 . In present study, we investigated the role of lncRNA HOTAIR/miR-148a-3p/KLF6 in the recovery of neurological function in IS.
After bioinformatics analysis, HOTAIR related to IS was selected as the focus of this study. We found HOTAIR was elevated in pMCAO mice. Then mice were injected with shRNA HOTAIR to silence HOTAIR expression to further explore HOTAIR action in pMCAO. Our study demonstrated that pMCAO mice displayed severe neurologic impairments including high scores of neurological de cits and motor balance, poor balance ability, prolonged escape latency and decreased platform crossing times. However, these neurologic impairments were reversed after lentivirus injection of shRNA HOTAIR. In ammation and apoptosis play key roles in all stages of IS 34 and lncRNAs are suggested to be able to promote apoptosis and in ammation 1,36 . The infarcted area, in ammation and apoptosis in pMCAO mice were also decreased signi cantly after silencing HOTAIR. The result is consistent with the recent researches that HOTAIR promotes neuroin ammation 36 and apoptosis cells 9 . Therefore, silencing HOTAIR suppressed apoptosis and in ammation, attenuated neurobehavioral de cit and promoted the recovery of neurological function of pMCAO mice. Lin et al. have also demonstrated that knockdown of HOTAIR expression reduced the apoptosis rate among dopamine neurons 12 . We then simulated IS in vitro using OGD-injured N2a cells to further investigate the role of HOTAIR in IS. HOTAIR expression in ODG cells was increased signi cantly. Then, we lowered the expression of HOTAIR in OGD cells using shRNA HOTAIR.
Silencing HOTAIR remarkably promoted the cell viability, reduced apoptosis and levels of TNF-α and IL-6 and increased IL-10 level. Knocking down HOTAIR can facilitate cell viability and attenuate apoptosis 37 .
The anti-in ammatory ability of shRNA HOTAIR has been identi ed in several studies [38][39][40] . The results above revealed that silencing HOTAIR promoted viability of OGD-injured N2a cell, and attenuated apoptosis and in ammation.
Study suggested that HOTAIR may act as a ceRNA to impose post-transcriptional regulation 41 . The experiment results identi ed that HOTAIR was located in the cytoplasm of N2a cells. The downstream miRs of HOTAIR were further predicted and then miR-148a-3p and miR-326 were obtained. miR-148a-3p was selected because it was poorly expressed in cerebral I/R 17 . IS chip GSE35338 analysis revealed that most genes were highly expressed in IS samples. Meanwhile, the signi cant upregulated genes were intersected and interaction analysis was performed. The result showed the highest score in KLF6. Also, KLF6 was highly expressed in IS chip GSE35338. Above results implied that lncRNA-HOTAIR may regulate IS progress through the miR-148a-3p/KLF6 axis. Moreover, the binding relations of HOTAIR and miR-148a-3p and that of miR-148a-3p and KLF6 were con rmed by dual-luciferase assay. Thereby, our research demonstrated that HOTAIR regulated KLF6 expression via targeting miR-148a-3p.
To explore the effect of miR-148a-3p, we transferred miR-148a-3p inhibitor into N2a cells with silencing HOTAIR and then induced cells by OGD. Our data elicited that the viability of OGD-induced N2a cells was reduced remarkably after inhibiting miR-148a-3p, apoptotic rate was increased, and the levels of TNF-α and IL-6 were increased and IL-10 level was decreased. As a novel repressor of in ammatory gene expression, miR-148a-3p plays a vital role in many diseases 42 . Previous investigation showed that miR-148a can alleviate hepatic I/R injury in mice 43 . Collectively, our study demonstrated that miR-148a-3p knockdown reversed the effect of sh-HOTAIR on OGD-induced N2a cells.
KLF6 is important in modulating in ammation and immune responses 44 . To further investigate the effect of KLF6, we transferred pcDNA-KLF6 into cells with silencing HOTAIR and then cells were treated by OGD to induce injury. The viability of OGD-induced N2a cells was decreased after overexpressing KLF6, apoptotic rate was obviously increased and levels of TNF-α and IL-6 were increased and level of IL-10 was decreased. Zhang et al. found that KLF6 exacerbates renal tubular epithelial cell damage and aggravates the in ammatory response in renal I/R injury 45 . Furthermore, KLF6 promotes macrophage in ammatory and hypoxia response 46 . In short, KLF6 overexpression repressed the viability of sh-HOTAIR-treated cells, and aggravated apoptosis and in ammatory response.
Besides, functional synergy of KLF6 and STAT3 has been clari ed in central nervous system neurons 47 .
One recent study has found that the activation of STAT3 pathway promotes microglia/macrophage polarization toward anti-in ammatory phenotype 48 . Our experiment result suggested that the expression of p-STAT3 was decreased in vivo and in vitro ischemic model, while silencing HOTAIR increased p-STAT3 expression, and miR-148a-3p knockdown or overexpression of KLF6 reduced its expression. Activating JAK2/STAT3 pathway is involved in the protection against cerebral I/R injury, and enhances cerebral angiogenesis and promotes neurological recovery after IS 49 . Taken together, HOTAIR affected neurological injury in IS through inhibition of STAT3 pathway mediated by the miR-148a-3p/KLF6 axis.
In conclusion, our study found that lncRNA HOTAIR inhibited STAT3 pathway and the recovery of neurological function in IS via the miR-148a-3p/KLF6 axis. Limitations still exist in the present study.
Whether KLF6 can regulate neurological recovery through the STAT3 pathway, and whether HOTAIR and miR-148a-3p can become biomarkers and potential therapeutic targets for the diagnosis and prognosis of IS require further investigations. The focus of future researches can be placed on the regulating mechanism of HOTAIR on miR-148a-3p/KLF6 from the view of epigenetics. It's reported that HOTAIR epigenetically can suppress miR-122 expression in hepatocellular carcinoma via DNA methylation 50 . But the regulating mechanism and biological functions between HOTAIR and DNA methylation in IS remain unclear. Moreover, the studies are rare in the involvement of lncRNA CDKN2B-AS1 and miR-326 in neurological recovery in IS. Therefore, the next step can be taken from this perspective.

Declarations Data Availability
All the data generated or analyzed during this study are included in this published article.