Effect of Gastrodin on Cognitive Dysfunction in Diabetes by Inhibiting PAK2 Phosphorylation

Diabetes and cognitive dysfunction are highly prevalent disorders, while the underlying mechanism is still elusive. The effects of Gastrodin on central nervous system have been emphasized recently. In this study, we aim to explore the potential mechanism leading to cognitive dysfunction in diabetes and the therapeutic effect of Gastrodin. Diabetes was induced by a single injection of streptozotocin. RNA sequencing technique was used to identify the potential factors involved. Western blot and immunouorescence were applied to detect the protein expression. Our results have shown that spatial learning was impaired and hippocampal pyramidal neurons were damaged in diabetic rats, which could be ameliorated by Gastrodin intervention. Transcriptional analysis identied differential expression genes, which were conrmed by qPCR and western blot. Furthermore, p21 activated kinase 2 (PAK2) was selected and its inhibitor could promote the survival of primary hippocampal neurons. It suggested that PAK2 pathway may be involved in cognitive dysfunction in diabetes and a therapeutic target for Gastrodin intervention. signaling pathway for further investigation. We compared the expression of PAK2 between normal controls, diabetic rats and diabetic rats with Gastrodin intervention. We found that PI3K/AKT signaling pathway was downregulated in diabetic rats, while p-PAK2 expression was increased and GLUT4 expression decreased. After Gastrodin intervention, PI3K/AKT signaling pathway was restored, coupled with decreased p-PAK2 expression and increased GLUT4 expression. This indicated that PAK2 activation in diabetes contributed to decreased GLUT4 expression through PI3K/AKT pathway. As a result, it reduced neuronal glucose uptake and led to decreased hippocampal neuronal energy and neuronal injury. Suppressing PAK2 activity by Gastrodin intervention or PAK2 inhibitors can restore neuronal glucose uptake and protect hippocampal neurons.


Introduction
Cognitive dysfunction has been increasingly recognized as a common complication and comorbidity of both type 1 and type 2 diabetes [1]. Up to 20% of type 2 diabetic patients older than 60 years could have dementia [2]. Cognitive dysfunction in diabetes (CID) includes dysfunction in memory, executive function, language and spatial ability [3]. However, the underlying mechanisms between diabetes and cognitive dysfunction are largely unknown.
The hippocampus plays a key role in cognitive functions and is affected in many diseases. The hippocampal CA1 pyramidal neurons are selectively attacked during the progression of Alzheimer's disease [4]. Diabetes could cause neuronal loss in the hippocampal CA1 region and hippocampal insulin resistance is a potential mediator of CID [5]. The decline of hippocampal neuronal function is closely related to cognitive dysfunction caused by degenerative diseases [6]. To search for the potential factors involved in diabetes-related hippocampal damage, we used RNA-seq to reveal the expression changes of signi cant differential genes between the hippocampus of diabetic rats and normal controls.
Gastrodia elata is a traditional Chinese herbal medicine, which has been used for the treatment of headache, dizziness, epilepsy, stroke and amnesia for a long time [7]. Gastrodin is a phenolphthalein compound and the main active ingredient of Gastrodia elata, whose therapeutical effects on central nervous system diseases has been frequently reported recently [8]. It works through regulating neurotransmitters [9], restoring vascular function [10], anti-oxidation and anti-in ammatory effects [11]. Since there is no effective treatment of CID, we aim to test the effects of Gastrodin on protecting hippocampal neurons and improving cognitive function in diabetes.

Primary Neuron Culture
Primary neurons were prepared from the hippocampus of P0 C57 mouse. Brie y, dissociated hippocampal cells were spread on 6-well plates coated with poly-D-lysine (Sigma-Aldrich, St. Louis, MO, USA) and cultured in DMEM at a density of 25 X 10 5 cells/well. After 3.5 h of seeding, the medium was changed to Neurobasal medium (Gibco, Carlsbad, NM, USA) supplemented with B-27 (Gibco). Cells were cultured in a humidi ed incubator at 37℃ with 5% CO 2 . Cultures were used for experiments 3 to 7 days after seeding, which were divided into NC, DM (50 mM high glucose for 48 hours), DM+G (50 mM high glucose and 30 µM Gastrodin for 24 hours) and DM+I (50 mM high glucose and 5 µM FRAX597 for 24 hours) groups.

Morris Water Maze Test
Morris water maze was performed as previously described [12]. All trials were performed in a quiet room with indirect lighting and the animals would be dried under a heater after each experiment. The apparatus was a circular tank with 190 cm in diameter as a swimming pool and contained water at approximately 22 ± 1 ℃. Spatial learning was assessed across repeated trials for 7 days. On the rst day, the animal was placed into the water for 2 min. On the second day, a circular platform of 15 cm in diameter was positioned 2 cm below the water, and the rats were rst put on the platform for 30 s and then guided into the water. If the rats found the platform within 30 s, they would be remaining on it for 30 s. If not, they would be guided to the platform and staying on it for 30 s. Afterwards, the rats were put into the water facing the sidewalls subsequently from one of the four separate quadrants of the pool. If the animal failed to nd the platform within 30 s, it would be guided to it and stay there for 30 s. Since the third day, the rats began to be released into the water facing the sidewalls of the pool from one of the quadrants and should nd the platform within 120 s. If it succeeds, it would be allowed to stay for 30 s. And if it failed, it would be guided to the platform and stay for 30 s. On the last day, the platform was removed and the rats were released into the water facing the wall of the pool from the farthest point from the platform. In this test, if the rat recalled the position of the platform, it would swim along a shorter path to the platform on the second trial. The rats not managed to remember the platform's position in the previous days would not be able to locate it easily. Therefore, we compared the differences among these groups to evaluate their spatial learning ability.

Hematoxylin and Eosin (H&E) staining
To observe the histological changes, the sections were rst incubated with hematoxylin (Beyotime Institute of Biotechnology, Shanghai, China) for 5 min and then washed with 1% ethanol hydrochloride for 3 s. After rinsing with water, the sections were stained with eosin. After this, the sections were captured under a light microscope at the magni cation ×400 in a blinded manner.

RNA-sequencing
The mRNA libraries were sequenced on the Illumina sequencing platform by Genedenovo Biotechnology Co., Ltd (Guangzhou, China). After the total RNA is extracted from the sample, for eukaryotes, the magnetic beads with Oligo (dT) are used to enrich the mRNA, and the fragmentation buffer is added to the obtained mRNA to make the fragment into a short fragment, and then the fragmented mRNA is used as a template. The rst strand of cDNA was synthesized by random hexamers, and the second strand of cDNA was synthesized by adding buffer, dNTPs, RNase H and DNA polymerase I. It was puri ed by QiaQuick PCR kit and eluted with EB buffer. Add base A, add sequencing linker, and then recover the target size fragment by agarose gel electrophoresis, and carry out PCR ampli cation to complete the whole library preparation work. The constructed library was sequenced with Illumina HiSeqTM 2500. The SRA data have been uploaded to NCBI (BioProject: PRJNA759189).

Bioinformatics analysis
The bioinformatics analysis is mainly divided into three modules: First, the TopHat comparison is performed separately for the reads and reference genomes of each sample, and the comparison results of each sample are obtained. Then the cu inks are used to assemble the transcripts, and the assembly of each sample is obtained. Second, multiple samples were grouped and combined using cuffmerge according to different treatments, the results of different groups were also combined by cuffmerge, and nally the expression levels of the genes in different groups were obtained. Finally, the predicted gene is analyzed by using edgeR for difference analysis, functional annotation of the differential genes and annotation of new genes. qPCR The hippocampal tissues were isolated and collected to examine the mRNA level. RNA extraction was conducted by RNApre pure Tissue Kit reagent (TIANGEN, Beijing, China). After measurement of RNA concentration by spectrophotometer (Bio-Rad, Hercules, CA, USA), the reverse transcription reaction was performed by a RevertAid First Strand cDNA Synthesis Kit (Thermo Fisher Scienti c, Waltham, MA, USA) according to the manufacturer's instruction. Then cDNA was used to amplify and quantify by SYBR Green Realtime PCR Master (Solarbio, Beijign, China). The ampli cation parameters were 95• C for 2 min followed by 45 cycles of 95• C for15 s and 58• C for 15 s. The measurement was conducted in triplicate. The mRNA level was normalized to the reference gene GAPDH and shown as relative expression of mRNA by 2 −Δ ct method.

Western Blotting Analysis
The rats were anesthetized with 10% chloral hydrate administered intraperitoneally. The hippocampal tissues were rapidly dissected and immediately frozen in liquid nitrogen and stored in −80 ℃. Proteins were extracted from the hippocampal tissues by RIPA buffer (1:1; Beyotime) containing a 1% protease inhibitor cocktail (1:100; Cell Signaling Technology, Danvers, MA, USA) and 1% phosphatase inhibitor cocktails (1:100; Cell Signaling Technology) at 4 ℃. Homogenates were centrifuged at 12,000 g for 10 min, and the supernatant was collected. Protein concentration was measured using a BCA protein assay kit. The proteins (30 µg) were loaded unto SDS-PAGE gel. The gels were electrophoresed and then transferred to PVDF membranes.
After that, the membranes were blocked with a blocking buffer using 5% non-fat milk for 120 min and probed with primary antibodies overnight at 4 ℃. They were then incubated for 2 h at room temperature with appropriate secondary mouse antibodies (1:1,000; Thermo Fisher Scienti c). The following primary antibodies were used for this study: rabbit anti-

Double Immuno uorescence
The hippocampus was dissected, immersed in 4% formaldehyde, dehydrated, cleared with xylene, and embedded in para n blocks. Para n sections of 4 µm thickness were depara nized and hydrated through a series of graded alcohol. The tissues were incubated in citrate buffer for antigen retrieval and the slices were incubated with 5% normal goat serum. The following primary antibodies were used: rabbit anti-PI3K antibody (1:1,00 dilution; Abcam), rabbit anti-p-AKT antibody (1:25 dilution; CST) and mouse anti-GLUT4 antibody (1:20 dilution; Santa Cruz). Primary antibodies were added in a fresh blocking solution and incubated overnight at 4 ℃. The staining was visualized with anti-mouse and anti-rabbit Alexa Fluor 488 and 568 2nd antibodies (1:500; A11010, A11001, Invitrogen, Waltham, MA, USA). After washing in PBS, secondary antibodies were added in PBS containing 0.1% Triton X-100 to prevent nonspeci c antigen binding for 2 h at room temperature. Tissue sections were viewed, and images captured on an Olympus FV1000 microscope.

CCK8 Assay
Cells (4 X 10 4 /well) were seeded in 96-well culture plates and exposed to different conditions as above. Cell viability was then determined by Cell Counting Kit-8 (CCK8) (Beyotime, China) assay. After the exposure, 10 µl of CCK8 solution was added to each well, and the plates were incubated for an additional 2 h at 37°C. Cell viability was measured as the absorbance at 450 nm with a microplate reader (Bio-Rad, USA) and expressed as a percentage of the control level. The mean optical density values from six wells for each treatment were used as the index of cell viability.

Statistical processing
Statistical analysis was performed using the SPSS 17.0 statistical software package. The data of each group were expressed as mean±standard deviation (), and the one-way ANOVA was used to compare the sample groups.

Gastrodin Intervention Improved Cognitive Function and Ameliorated Pathological Changes in Diabetic Rats
To evaluate the effects of Gastrodin intervention on cognitive function, Morris water maze test was performed. Diabetic rats exhibited signi cantly higher escape latency than the controls (p<0.01), which was signi cantly decreased after 60 mg/kg of Gastrodin intervention ( Figure 1A). In the space exploration stage, the number of platform crossing and the target quadrant dwelling time were signi cantly decreased in the DM9w+S group than that of the NC9w group (p<0.01). 60 mg/kg of Gastrodin intervention signi cantly increased the target quadrant dwelling time of diabetic rats, while it had no effect on the number of platform crossings ( Figure 1B-C). Interestingly, 120 mg/kg of Gastrodin intervention showed no therapeutic effect on the performance of Morris water maze test.
H&E staining showed that neurons in the CA1 pyramidal region of the hippocampus were with intact structure and clear nuclei in the NC9w group. In addition, the axon of the radiatum layer was extended and arranged neatly. In the DM9w+S group, the number of pyramidal neurons in the CA1 region of the hippocampus was noticeably decreased. Furthermore, the cytoplasm was reduced, the nucleus exhibited edema, and the axon extension of the radiatum layer was decreased and disarranged.  (Figure 2A-B).
The properties of DEGs were described according to the internationally standardized gene function classi cation system Gene Ontology (GO). The molecular function, cellular component and biological progress of the genes were respectively described.
GO analysis showed that the main modules of these DEGs are cellular process, transmitter transport and extracellular component changes. Pathway enrichment analysis with KEGG Pathway was applied to nd signi cant enrichment pathways in sub-differential expression compared with the entire genetic background. The rst 20 signi cantly enriched pathways between different groups were found, which included cell adhesion molecule pathway, phagocytic pathway, extracellular matrix receptor interaction pathway, PI3K-AKT signaling pathway, focal adhesion pathway, etc ( Figure 2C). Furthermore, DEGs with fold change ≥ 5 were selected according to the behavioral changes assessed by the Morris water maze test (Table 1). PAK2, Mrpl1, RagA/B and Sp8 were selected for further con rmation by qPCR test. The result showed that the expression of Mrp11, RagA and RagB in DM9w+S group was not signi cantly higher than that of the NC9w group. By contrast, the expression of PAK2 and Sp8 was signi cantly increased after diabetes induction. In addition, their expression in the DM9w+G60 and DM9w+G120 group was signi cantly lower than that of the DM9w+S group, which was consistent with transcriptome changes ( Figure 2D-H). Because diabetic rats exhibited impaired learning ability and Gastrodin had a therapeutic effect, we reviewed the reports on cognition and found that PI3K-AKT signaling pathway was associated with insulin-mediated cognitive function. Moreover, PAK2 was closely related with PI3K-AKT insulin signaling, which downwardly resulted in reduced glucose uptake under pathological conditions through glucose transduction receptors GLUT4. It would lead to the damage of insulin-dependent neurons and compromised cognitive function. Therefore, PAK2 and PI3K/AKT/GLUT4 signaling pathway was selected for further investigation.

Gastrodin Intervention Inhibited the Phosphorylation of PAK2 Through Regulating PI3K/AKT Pathway in Diabetic Rats
The expression of PI3K, p-AKT and GLUT4 in the hippocampus of DM9w+S group was signi cantly lower than that of the The expression of PI3K, p-AKT, p-PAK2 and GLUT4 in the primary neurons has shown similar changes compared with the hippocampus. The protein expression of PI3K, p-AKT and GLUT4 was signi cantly decreased after high glucose intervention, while p-PAK2 was signi cantly increased. After both 30 µM of Gastrodin and 5 µM of FRAX597 treatment, the expression of PI3K, p-AKT and GLUT4 was signi cantly increased and p-PAK2 was signi cantly decreased ( Figure 5).
Double immuno uorescent staining has shown that PI3K, p-AKT and GLUT4 were distributed mainly in the cell body of the primary hippocampal neurons ( Figure 6). The diminution of the above factors after high glucose exposure could be restored both by Gastrodin and PAK2 inhibitor intervention. Moreover, the number of primary hippocampal neurons exhibited a drastic reduction in the DM group. Gastrodin and PAK2 inhibition showed a protective effect on the survival of neurons.

Discussion
This present study has shown that diabetes could cause hippocampal neuronal damage and impairment of spatial learning.
Our study is the rst to screen for the key factors for underlying the therapeutic effects of Gastrodin on diabetes-induced hippocampal injury using RNA-seq. After screening the relevant enrichment pathways, we selected PAK2 and PI3K/AKT signaling pathway for further investigation. We compared the expression of PAK2 between normal controls, diabetic rats and diabetic rats with Gastrodin intervention. We found that PI3K/AKT signaling pathway was downregulated in diabetic rats, while p-PAK2 expression was increased and GLUT4 expression decreased. After Gastrodin intervention, PI3K/AKT signaling pathway was restored, coupled with decreased p-PAK2 expression and increased GLUT4 expression. This indicated that PAK2 activation in diabetes contributed to decreased GLUT4 expression through PI3K/AKT pathway. As a result, it reduced neuronal glucose uptake and led to decreased hippocampal neuronal energy and neuronal injury. Suppressing PAK2 activity by Gastrodin intervention or PAK2 inhibitors can restore neuronal glucose uptake and protect hippocampal neurons.
CID is one of the typical central nervous system complications of DM, but its pathogenesis is far from clear [13]. In the brain, insulin signaling is involved in processes including neurogenesis, cognition, eating behavior, and glucose metabolism [14].
Previous studies have shown that insulin signaling dysfunction is an important mechanism of diabetic complications, including cognitive dysfunction [15]. However, the underlying mechanism of insulin signaling dysfunction is still unknown. The result of our bioinformatic analysis suggested that P21-activated kinase 2 (PAK2) could be an important factor of CID, which may regulate the insulin signaling function and glucose metabolism in the hippocampus.
P21-activated kinase (PAK) is an important component of glucose homeostasis in the muscle, pancreas and liver by mediating insulin signaling [16,17]. Previous studies have found that PAK2 is involved in neuronal insulin signaling, glucose uptake and insulin resistance [18]. However, its role in neuronal insulin signaling is still unknown [19]. We found that the expression of PAK2 was elevated after diabetes, which suppressed the expression of GLUT4 through PI3K/AKT pathway. Other independent studies have also indicated that PI3K/AKT is involved in the regulation of glucose uptake, which were regulated by PAK2 in the neurons [17]. In Alzheimer's disease, impaired insulin signaling could inhibit the PI3K/AKT pathway and lead to neurodegeneration by increasing oxidative stress, apoptosis, mitochondrial dysfunction and necrosis [20].
The activity of PI3K/AKT pathway is related to the phosphorylation of AKT. Phosphorylation of AKT, a key downstream protein of PI3K, would be downregulated at high glucose level [18]. This signal transduction dysregulation caused by inhibition of AKT phosphorylation is the key to hyperglycemia induced cognitive de cits [21,22]. Furthermore, inhibition of AKT and PI3K expression by treatment with AKTi-1/2 and wortmannin reduced insulin regulation of PAK2 and glucose uptake [23].
GLUT4, an insulin-dependent glucose transporter, is involved in the uptake of glucose by neurons [24]. This process is associated with normal cognitive formation [25]. In the neurons, an increase in neurotransmitter activity is also considered as an indirect effect of insulin-dependent glucose uptake, by which insulin regulates cognitive activity [26]. We have found that increased expression of PAK2 in the hippocampus of diabetic rats suppressed GLUT4 expression, which further reduced glucose uptake, supported by a study indicating that overexpression of PAK2 could reduce GLUT4 expression and glucose uptake [27]. Moreover, inhibition PAK2 expression restored the expression of GLUT4 and protect hippocampal neurons.
Finally, we found that Gastrodin intervention had a similar effect on CID as PAK2 inhibitor, suggesting that it could be a potential treatment for CID [28,29]. Recently, the therapeutic effects of Gastrodin have been widely investigated. Some clinical trials have found the bene cial outcome of Gastrodin on patients with vascular dementia [8,30]. Though many studies have found that Gastrodin treated diabetes through PI3K/AKT pathway [31,32], our study was the rst to report that the PAK2 is important for the effect of Gastrodin. Our result also found that 60 mg/kg of Gastrodin intervention had a better outcome than 120 mg/kg. This could be partly explained by the lower expression of PI3K/AKT/GLUT4 signaling pathway.

Consent for publication
The authors declare that no human subjects were involved in the study.

Data availability statement
The sequencing data has been deposited into the Sequence Read Archive (accession: PRJNA759189). Other datasets generated during the current study are available from the corresponding author on reasonable request.

Con ict of interest
Author Yi-Dan Liu was employed by the company Kunming Pharmaceutical Corporation. The remaining authors declare that the research was conducted in the absence of any commercial or nancial relationships that could be construed as a potential con ict of interest.    Gastrodin intervention suppressed diabetes-induced PAK2 phosphorylation and activated PI3K-AKT pathway. Western blot analysis of PI3K (A), p-PAK2 (B), p-AKT (C) and GLUT4 (D) protein expression in the hippocampus of the NC9W, DM9W+S, DM9W+G60 and DM9W+G120 groups. Bar graphs represented optical density of these factors normalized with β-actin, while p-PAK2 and p-AKT were further normalized with total-PAK2 and total-AKT respectively. * p < 0.05 and * * p < 0.01.