Identification of miRNA and Their Regulatory Effects Induced by Total Flavonoids From Dracocephalum moldavica in the Treatment of Vascular Dementia

Vascular dementia (VaD) is a general term used to describe difficulties in memory, reasoning, judgment, and planning caused by a reduced blood flow to the brain and consequent brain damage, in which microRNAs (miRNAs) are involved. Dracocephalum moldavica L. (D. moldavica) is traditionally used in the treatment of cardiovascular diseases as well as VaD, but the biomolecular mechanisms underlying its therapeutic effect are obscure. In the present study, the molecular mechanisms involved in the treatment of VaD by the total flavonoids from Dracocephalum moldavica L. (TFDM) were explored by the identification of miRNA profiling using bioinformatics analysis and experimental verification. A total of 2,562 differentially expressed miRNAs (DEMs) and 3,522 differentially expressed genes (DEGs) were obtained from the GSE120584 and GSE122063 datasets, in which the gene functional enrichment and protein-protein interaction network of 93 core targets, originated from the intersection of the top DEM target genes and DEGs, were established for VaD gene profiling. One hundred and eighty-five targets interacting with 42 flavonoids in the TFDM were included in a compound-target network, subsequently found that they overlapped with potential targets for VaD. These 43 targets could be considered in the treatment of VaD by TFDM, and included CaMKII, MAPK, MAPT, PI3K, and KDR, closely associated with the vascular protective effect of TFDM, as well as anti-oxidative, anti-inflammatory, and anti-apoptotic properties. The subsequent analysis of the compound-target gene-miRNA network indicated that eight miRNAs that mediated 43 targets had a close interaction with TFDM, suggesting that the neuroprotective effects were principally due to kaempferol, apigenin, luteolin, and quercetin, which were mostly associated with the miR-3184-3p/ESR1, miR-6762-3p/CDK1, miR-6777-3p/ESRRA, and other related axes. Furthermore, the in vitro oxygen-glucose deprivation (OGD) model demonstrated that the dysregulation of miR-3184-3p and miR-6875-5p found by qRT-PCR was consistent with the changes in the bioinformatics analysis. TFDM and its active compounds involving tilianin, luteolin, and apigenin showed significant effects on the upregulation of miR-3184-3p and downregulation of miR-6875-5p in OGD-injured cells, in line with the improved cell viability. In conclusion, our findings revealed the underlying miRNA-target gene network and potential targets of TFDM in the treatment of VaD.


INTRODUCTION
Vascular dementia (VaD) represents a group of syndromes characterized by cognitive impairment resulting from the death of the hypoxic brain tissue caused by the reduced perfusion from diseased brain blood vessels (Iadecola, 2013), causing approximately 15-20% of all dementia cases (O'Brien and Thomas, 2015). The pathogenesis of VaD is complex and is the result of a variety of cardiovascular or genetic risk factors associated with Apolipoprotein E (APOE), angiotensin I converting enzyme (ACE), paraoxonase 1 (PON1), or presenilin 1 (PSEN1) genes, or due to stroke (O'Brien and Thomas, 2015). Thus, there is no authorized treatment for VaD in clinical practice. The current therapies for VaD are mostly performed using cholinesterase inhibitors or memantine due to the lack of a reliable biomarker or an established molecular mechanism, and these compounds are moderately effective. However, among all forms of dementia, VaD is considered unique because it can be prevented if the intervention is provided early enough (Thal et al., 2012). Therefore, an exploration of specific biomarkers or target genes for VaD might be of help in the development of mechanism-targeted prevention and therapy.
Although the pathologic mechanism of VaD is not yet fully elucidated, it has been reported that non-coding RNAs such as microRNAs (miRNAs), long non-coding RNAs, and circular RNAs are involved in the pathophysiology of VaD, including cognitive impairment, vascular dysfunction, neuroinflammatory response, blood-brain barrier disruption, and synaptic loss (Toyama et al., 2018;Liu et al., 2019b;Huang et al., 2020;Vijayan and Reddy, 2020;Wang et al., 2020). miRNAs are enriched in the brain and responsible for the regulation of post-transcription of a wide range of proteins through the degradation or inhibition of their target mRNAs. In a rat model of VaD, a reduction in miR-126 expression in endothelial cells revealed the pathophysiological process mediating the cognitive impairment in the rats accompanied by a decreased cerebral blood flow and vessel patency, causing the activation of astrocytes and microglia, and a reduction in synaptic plasticity (Yu et al., 2019). It has been reported that the miR-93mediated Toll-like receptor signaling pathway plays a functional role when cognition defects in VaD is treated by acupuncture . In addition, the involvement of miR-124 (Liu et al., 2019a), miR-210-5p (Ren et al., 2018), miR-132 (Yang et al., 2020a), and miR-7-5p (Xu et al., 2019) has been demonstrated in the development of VaD. Although it has been proposed that miRNAs represent therapeutic targets and diagnostic biomarkers for VaD, the varying expression profiles of miRNAs and disease characteristics of VaD are not fully understood.
D. moldavica is a traditional Uygur medicine in China, widely used to treat coronary heart disease, angina, and atherosclerosis as well as for calming the nerves (Dastmalchi et al., 2007;Maimaitiyiming et al., 2014;Miernisha et al., 2016). Modern neuropharmacological studies confirmed that D. moldavica has beneficial effects on the central nervous systems, as demonstrated by the promotion of prolonged pentobarbital-induced sedation in mice (Martinez-Vazquez et al., 2012), neuroprotection against cerebral ischemia/reperfusion injury in rats (Jia et al., 2017), and cognitive improvement against scopolamine-induced deficits in mice (Deepa et al., 2020). The total flavonoids from D. moldavica (TFDM) represent a standardized extract from D. moldavica and function as a novel treatment for VaD as established by our research team, obtaining the support of a number of major national projects in China and patent authorization. Our ongoing study demonstrated that TFDM provides neuroprotection and improvement of learning and memory capabilities against beta-amyloid-induced toxicity in vitro and in vivo . The principal flavonoids in TFDM, such as tilianin, luteolin, apigenin, and quercetin improve the cognitive function in various types of cognitive decline (Jiang et al., and, 2021Liu et al., 2009;Liu et al., 2013;Zhao et al., 2013). Furthermore, TFDM and its principal compounds reduce the cognitive deficit acting on different mechanisms, including antioxidative, anti-amyloidogenic, neurotrophic, anti-inflammatory, and anti-apoptotic pathways as a result of the regulation of the expression of a number of genes, such as amyloid-beta precursor protein (APP), beta-secretase 1 (BACE1), cAMP-responsive element-binding protein 1 (CREB), calcium/calmodulindependent protein kinase II gamma (CaMKII), mitogenactivated protein kinase (MAPK), and B-cell leukemia/ lymphoma-2 (Bcl-2) (Liu et al., 2009;Zhao et al., 2013;Jiang et al., 2018;Jiang et al., 2020;Liu et al., 2018). However, the effect of TFDM in the improvement of VaD through the regulation of multiple genes remains unclear. Thus, the discovery of the mechanism of action could allow its potential use in the treatment of VaD.
In the current study, bioinformatics approaches with biological verification were used to determine the miRNAs involved in the treatment of VaD by TFDM. Eight significant miRNAs that mediated 43 targets with a close interaction with TFDM were identified using the transcriptional profile network of VaD established from the Gene Expression Omnibus (GEO) database. The qRT-PCR assay revealed that the upregulation of miR-3184-3p and downregulation of miR-6875-5p might play critical roles in the treatment of VaD by TFDM. The workflow is displayed in Figure 1.

Dataset Acquisition
Two microarray datasets, GSE122063 (McKay et al., 2019) and GSE120584 (Shigemizu et al., 2019a), were selected from the GEO (https://www.ncbi.nlm.nih.gov/gds/) using "VaD" as the search term, with the species limited to "Homo sapiens". The GSE122063 dataset includes gene expression profiles of the frontal and temporal cortex of patients with VaD, Alzheimer's disease (AD), and non-demented controls (Control), and McKay et al. focused on the special changes in gene expression in samples of the frontal cortex of VaD patients, identifying the oxytocin receptors (OXTR) as significantly upregulated (McKay et al., 2019). In the present study, 11 samples from non-demented controls and 8 VaD samples were selected and subjected to analysis to obtain all the DEGs. Shigemizu et al. constructed risk prediction models for three subtypes of dementia, including AD, VaD, and dementia with Lewy bodies (DLB), using potential biomarkers from serum samples, and further analyzed the functional genes associated with DLB pathology (Shigemizu et al., 2019b). The samples of 91 VaD patients and normal subjects from the GSE120584 dataset were selected for miRNA analysis.
Data Processing and Differential Expression Analysis GEO2R (https://www.ncbi.nlm.nih.gov/geo/geo2r/) is an online analysis tool in the GEO database. It was used to screen differentially expressed genes (DEGs) between control and VaD tissue samples in the GSE122063 dataset, with a cut-off of |log Fold Change| > 0.5 and adjusted (adj.) p-value < 0.01. Volcano plots were generated to describe the distribution and expression of the DEGs that were identified. Differentially expressed miRNAs (DEMs) in VaD patients and normal subjects in the GSE120584 dataset were also analyzed using the GEO2R webtool. The six most upregulated and downregulated miRNAs were selected in ascending order by p-value.

Target Prediction of miRNAs
The target genes of the top DEMs were predicted using miRDB (http://mirdb.org/), miRWalk (http://mirwalk.umm.uniheidelberg.de/), and TargetScan (http://www.targetscan.org/ vert_72/) online analysis tools. Overlapping target genes were identified using the Venn diagrams and then used in additional analysis. TargetScan can predict the biological target of miRNA by searching for conserved sites that match the seed region of miRNA, which are ranked by cumulative weighted context++ scores of the sites and their probability of conserved targeting. In contrast, miRDB is designed to screen target genes with a target score ≥50 by analyzing thousands of miRNA target interactions from high-throughput sequencing experiments. The miRWalk database is based on the current miRBase, using miRNA names to search for miRNA-related target gene prediction data. In the above three databases, the "human" species was selected before the prediction of the target genes.

Gene Ontology and Kyoto Encyclopedia of Genes and Genomes Pathway Analysis
GO analysis was performed to gain insight into the molecular mechanisms of VaD by molecular function (MF), cellular component (CC), and biological process (BP) domains of the core targets derived from the overlap of target genes of the most differentially expressed miRNAs and DEGs. KEGG pathway enrichment analysis was performed to define the signaling pathways associated with these core targets. The analysis used a threshold of adj. p-values < 0.01.

Construction of a Protein-Protein Interaction Network
The STRING 11.0 database (https://string-db.org/cgi/input) was used to assess the PPI network information of the core targets to discover the network of the interacting genes. A combined score of more than 0.9 was the comprehensive protein interaction score selected as the cut-off criterion. Cytoscape 3.7.0 software was then used to visualize the resulting PPI network. The four algorithms Maximal Clique Centrality (MCC), Density of Maximum Neighborhood Component (DMNC), Maximum Neighborhood Component (MNC), and DEGREE in the cytohubba plug-in were used to identify standard and central targets with high connectivity in the gene expression network.

Construction of Compound-Target
Gene-miRNA Network for TFDM

Identification of Components and Target Screening
The flavonoids in D. moldavica were searched in the reported literature (Yang et al., 2013;Liu et al., 2019c;Wu et al., 2020). Then, query Canonical SMILES of flavonoids in Pubchem was used to predict potential target genes. The potential targets of these compounds were predicted with probability ≥0.05 using the online prediction tool SwissTargetPrediction (http://www. swisstargetprediction.ch/). This portal allows the estimation of the most likely macromolecular targets for small molecules considered biologically active. The prediction was based on the similarity of the 2D and 3D compounds.

Network Construction
Compound-target (C-T) and compound-target gene-miRNA (C-T-M) networks were constructed using Cytoscape. The eigenvector centrality parameter of the node topology was calculated using Gephi 0.9.2 software by the analysis of the importance of each node, which depends on the number of nodes of its neighbours and their importance. Therefore, it is considered an appropriate predictor in interactive networks. In the network, the nodes represent the compounds, the targets, and the miRNAs, while the edges represent the relationship between compounds and targets or between targets and miRNAs.

Extraction of TFDM and Isolation of Tilianin
After confirmation of the herb, the powder of the aboveground parts (90 kg) was airdried at room temperature (RT), then refluxed three times with 40% EtOH at 100°C. Next, 3.2 kg crude extract was collected after evaporation of the filtered ethanol solution under reduced pressure. The crude extract was partitioned by column chromatography with a HPD600 resin and eluted with water, 50% EtOH, and 70% EtOH. The eluent was concentrated and dried to obtain TFDM, which was evaluated by high-performance liquid chromatography (HPLC).
The HPLC analysis was performed on a Sthermos U3000 HPLC system (Thermo Fisher Scientific, PA, USA). All separations were performed on a Welch Ultimater XB-C18 column (250 × 4.6 mm, 5 μm), with an injection volume of 20 μL. The mobile phase was composed of acetonitrile A and 0.5% formic acid aqueous solution B using a specific gradient elution (0-60 min, 20% A; 60-69 min, 50% A; 69-70 min, 80% A; 70-80 min, 20% A). The flow rate of the mobile phase was 1.0 ml min −1 , and the temperature was maintained at 35°C. The components were quantified by the peak areas at 330 nm UV wavelength.
The purified product was collected by filtering the 70% EtOH eluate on a silica gel column (100-200 mesh, chloroform: methanol, 95:5-90:10-80:20) to remove the impurities. The structure of the compound was subsequently determined by the physicochemical and spectroscopic data, and the purity of the compound was 99% as revealed by HPLC analysis, yielding a total of 280 mg of tilianin, which was the same batch in our previous research .

Preparation of the Pharmaceutical Solution
Apigenin (4,5,7-trihydroxyflavone, 97%) and luteolin (3′,4′,5,7-tetrahydroxyflavone, 98%) used in this work were purchased from Acmec Biochemical (Shanghai, China). TFDM and other three active compounds were dissolved in dimethyl sulfoxide (DMSO, Sigma-Adrich, St. Louis, MO, USA) to a stock concentration of 100 mM and kept at a temperature of −20°C in the dark. The final dilution was freshly prepared at the time of the treatment by diluting the concentrated solution directly into the culture medium.

Cell Culture and Treatment
SH-SY5Y cells (ATCC; Manassas, VA, USA) were routinely cultured in Dulbecco's Modified Eagle Medium (DMEM; Gibco, Grand Island, NY, USA) supplemented with 10% fetal bovine serum (FBS; Gibco, Grand Island, NY, USA). Cells were seeded in 96-well plates at a density of 1 × 10 4 cells/well in sugar-free DMEM and subjected to the hypoxic condition of 93.7%N 2 / 1.3%O 2 /5%CO 2 for 24 h at 37°C in a tri-gas incubator (Wiggens, Straubenhardt, Germany) for the establishment of an in vitro OGD model. Cells were treated with different concentrations of TFDM (25-100 μg/ml), tilianin (8-32 μM), luteolin (2.5-10 μM), and apigenin (2.5-10 μM) in the standard medium in a normoxic condition for 12 h prior to OGD injury. Next, the medium was replaced with the same sugarfree one with the same compounds at the same concentration during the OGD process.

Cell Viability Assay
Cell survival was assessed using the cell counting kit-8 (CCK8; Vazyme Biotech, Nanjing, China) according to the manufacturer's instructions and measured at 450 nm using a Spark 20M multimode microplate reader (Tecan Group Ltd., Mannedorf, Switzerland).

Expression of DEMs by Quantitative Reverse Transcription-Polymerase Chain Reaction
The qRT-PCR was used to confirm the changes of the predicted DEMs in the OGD-injured SH-SY5Y cells treated with TFDM and principal compounds. Briefly, total RNA was extracted using Trizol reagent (Invitrogen). miRNA was reverse-transcribed using a miRNA 1st Strand cDNA Synthesis Kit (by stem-loop) (Vazyme Biotech). PCR was performed using miRNA Universal SYBR qPCR Master Mix (Vazyme Biotech). Relative miRNA expression was calculated using the 2 −ΔΔCT method. Results were normalized using small nuclear U6 RNA as the internal control. Primers are listed in Table 1.

Statistical Analysis
Statistical analysis was performed using GraphPad Prism Version 7.0 software (GraphPad Prism Software; La Jolla, CA, USA). Data were analyzed using one-way ANOVA, followed by a Tukey's multiple comparison test, or Student's t-test. The results were expressed as mean ± standard deviation (SD). p-values less than 0.05 were considered statistically significant.

Identification of DEGs in VaD
DEGs in VaD were analyzed in the GSE122063 dataset. Using the cut-off criteria |log Fold Change| > 0.5 and adj. p-value < 0.01, 1,478 significantly upregulated DEGs were identified and 2,044 were downregulated in the brains of patients with VaD compared with normal control brains, as presented in Figure 2A. The 15 most downregulated and upregulated genes whose expression changed significantly are listed in Table 2. The genes with a significantly increased expression included Nexin 31 (SNX31), Fc Fragment of IgG Binding Protein (FCGBP), SLAM Family Member 8 (SLAMF8), Maternal immune activation (MIA), and XLOC_001219, and those with a significantly decreased expression were RNA Binding Motif Protein 3 (RBM3), XLOC_006951, stathmin 1 (STMN1), LOC100506274, and chaperonin Containing TCP1 Subunit 6B (CCT6B).

Target Prediction, Core Target Analysis, and Functional Analysis
The target genes of the 12 most dysregulated DEMs were predicted using the online analysis tools miRDB, miRWalk, and TargetScan. A total of 2,495 targets were identified. The miRNA-gene regulatory network was mapped ( Figure 3A) using Cytoscape software. Next, the screening was performed using the parameters |log Fold Change| > 0.5 and adj. p-value < 0.01, and 3,522 DEGs were identified. A total of 311 common targets were identified by analyzing the association between DEGs and the target genes of miRNAs ( Figure 3B), considered core targets closely related to the principal pathophysiological process of VaD. STRING was then used to analyze the core targets, resulting in a total of 94 core targets with a combined_score ≥0.9, which were selected for further analysis. They were imported into Cytoscape 3.7.0 and the key associations were studied. The complex PPI network contained 94 nodes and 228 edges ( Figure 3C).
GO and KEGG pathway analyses were used to determine the enrichment of biological function and the pathways of these 94 core target genes. The GO annotation revealed that most significantly enriched pathways in BP were mostly enriched in post-translational protein modification, positive regulation of hydrolase activity, positive regulation of the MAPK cascade, and blood vessel development. The most significantly enriched pathways in MF were cytokine activity, receptor-ligand activity, ubiquitin-like protein ligase binding, and purine ribonucleoside binding. The CC terms were mainly represented by cytoplasmic vesicle membrane, secretory granule membrane, and cullin-RING ubiquitin ligase complex ( Figure 4A).
The signaling pathways enriched in KEGG analysis were ranked in ascending order according to the p-value. The  results revealed that the selected core targets were significantly enriched in the advanced glycation end products (AGE)receptor for AGE (RAGE) signaling pathway in diabetic complications, Ras signaling pathway, estrogen signaling pathway, Rap1 signaling pathway, MAPK signaling pathway, tumor necrosis factor (TNF) signaling pathway, and other signaling pathways ( Figure 4B).

Construction of a C-T-M Network for TFDM
D. moldavica contains 42 flavonoids, including kaempferol, chrysoeriol, diosmetin, 8-hydroxy-salvigenin, apigenin, luteolin, salvigenin, 3-hydroxyflavone, isorhamnetin, quercetin, acacetin, gardenin A/B, and thymonin, as shown in Table 4. A total of 185 targets for TFDM were screened by target prediction using SwissTargetPrediction (Figure 6). Prior to the prediction of potential VaD-related targets for TFDM, a combination of 2,495 targets predicted by the 12 most dysregulated miRNAs and 3,522 differentially expressed target genes resulted in 5,706 combined targets that were considered potential targets and comprised a library related to the pathophysiology of VaD. Subsequently, genes common to both the potential VaD target library and targets predicted for TFDM were analyzed and selected ( Figure 7A), yielding 43 common targets, classified as key targets for TFDM ( Figures 7A,B), and considered those effective for the treatment of VaD. Furthermore, the corresponding miRNAs that bind to the common targets were identified and analyzed, and the results indicated that 8 miRNAs mediated 43 targets that potentially contributed to the neuroprotective effects of TFDM (consisting of 42 flavonoid compounds) on VaD (Figure 8). The 15 most active compounds, 15 vital mRNAs, and 8 miRNAs were found from the list of eigenvector centrality in descending order, as listed in Table 5.

Neuroprotection and Verification of DEMs Regulated by TFDM
Considering our previous research on TFDM, this compound and its representative flavonoids, including tilianin, luteolin, and apigenin, were selected to investigate the effects on the expression of predicted miRNAs. The fingerprint (approximately 71 peaks) of TFDM and the structure of tilianin, luteolin, and apigenin are shown in Figures 9A-D.
Before the verification of the predicted miRNAs regulated by TFDM, the neuroprotective effects of TFDM and its representative flavonoids on OGD-injured SH-SY5Y cells were measured. TFDM significantly improved the cell viability of OGD-injured SH-SY5Y cells in a dose-dependent manner at 25 μg/ml, 50 μg/ml, and 100 μg/ml (all p < 0.001), as shown in Figure 9E. The three active compounds from TFDM, such as tilianin, luteolin, and apigenin, also increased cell viability at the selected concentrations (p < 0.01-0.001, Figures 9F-H). These results suggested that TFDM and its three active compounds had neuroprotective effects against OGD-induced injury.
According to the qRT-PCR results, the expression of miR-3184-3p was decreased, and miR-6875-5p was increased in response to the OGD injury (both p < 0.001, Figure 10A), consistent with the analysis of the C-T-M network. The exposure to OGD did not significantly changed the expression of miR-6784-3p compared with the control group ( Figure 10A), while the expression of miR-6777-3p and miR-6762-3p was increased compared with the control group (p < 0.05 and 0.001), which was an opposite result compared to the C-T-M network. Thus, these results indicated  Frontiers in Pharmacology | www.frontiersin.org December 2021 | Volume 12 | Article 796628 8 that miR-3184-3p and miR-6875-5p were confirmed as having a significant aberrant change in the process of cerebral ischemia, which was consistent with previous bioinformatics analysis.

DISCUSSION
To our knowledge, the present study is the first constructing a C-T-M network indicating the potential molecular mechanism of TFDM on the treatment of VaD. Subsequent confirmation provided evidence that miR-3184-3p and miR-6875-5p might be promising targets used by TFDM to exert therapeutic effects and synergistic action in the treatment of VaD.
In this study, 3,522 aberrantly expressed mRNAs were identified based on VaD microarray analysis, including 1,478 upregulated and 2,044 downregulated mRNAs. Among these genes with the most significant changes in expression, corticotropin-releasing hormone (CRH) and CD163 have a previously confirmed correlation with the occurrence and pathology of VaD (Heilig et al., 1995;Guo et al., 2018;Landis et al., 2018;Short et al., 2019). Several significantly upregulated genes in VaD were identified as first, including Heat shock 70 kDa protein 1A (HSPA1A), V-set and immunoglobulin domain containing 4 (VSIG4), and complement C1q B chain (C1QB), all defined as vascular disease risks (Dulin et al., 2010;Bos et al., 2017;Lyu et al., 2020).
A total of 2,562 miRNAs were significantly dysregulated, divided into 1,167 upregulated and 1,395 downregulated miRNAs. Based on a comprehensive study of data mining and bioinformatics analysis, the 12 most abnormally expressed miRNAs were acquired. Among these miRNAs, miR-6784-3p, miR-6762-3p, miR-4747-3p, and miR-4746-3p were reported for the first time. Other significantly expressed miRNAs, such as miR-6777-3p, miR-3184-3p, miR-3184-5p, miR-4667-5p, and miR-6875-5p, closely implicated with the occurrence and development of a variety of tumors (Miyamoto et al., 2015;Akazawa et al., 2019;Hindle et al., 2019;Zhou et al., 2020;Liu et al., 2021), were identified for the first time in VaD, suggesting a novel potential regulatory role in the pathology of this disease. miR-3675-3p is a unique biomarker representing a target in the inhibition of cell apoptosis through the MAPK signaling pathway . As a novel miRNA identified in our analysis, the downregulation of miR-3675-3p might participate in the induction of neuronal apoptosis in VaD. miR-4467 and miR-1268a showed an increased expression in VaD in the present study. The former has previously been described as a promising and robust candidate biomarker in the diagnosis of neurodegenerative diseases by its analysis in body fluids (Denk et al., 2015), while the latter is significantly associated with cerebrovascular disorders (Sonoda et al., 2019) and hypertension (Li et al., 2020), predicting the risk of cerebrovascular diseases prior to the onset of stroke.
An overlap between DEGs in these miRNA-predicted targets and the large number of DEGs in VaD was constructed to gain insight into the molecular functions of the most abnormally expressed miRNAs in VaD, then the related biological function and pathways were analyzed. GO analysis indicated that the dysregulated mRNAs performed a crucial role in transcription, post-transcription, and protein modification. The primary biological processes in which the target genes of these selected aberrant miRNAs were involved included a variety of binding processes, including the binding of nucleotides and chromatin, cell differentiation, vascular events involving blood vessel development, muscle structure development, branching Frontiers in Pharmacology | www.frontiersin.org December 2021 | Volume 12 | Article 796628 morphogenesis, and extensive regulation of receptor activity. The pathway enrichment analysis indicated the involvement of RAGE, Ras/MAPK, TNF, gonadotropin-releasing hormone (GnRH), and estrogen pathways, which are known as implicated in the pathogenesis of VaD (Emanuele et al., 2005;Belkhelfa et al., 2018;Wang et al., 2019;Yang et al., 2020b). Among these enriched pathways, the RAGE signaling pathway plays a role in inflammatory reactions through the proinflammatory cytokine overproduction mediated by the nuclear factor-κB (NF-κB), such as TNF-α (Vicente Miranda and Outeiro, 2010), both in the blood-brain barrier (BBB) and in the glial responses. MAPK is also a crucial signaling pathway in the pathological process of VaD, which destroys the BBB, activates inflammatory mediators, and induces programmed cell death (Mohamed et al., 2015). Thus, these predicted biological processes and signaling pathways indicate the correlation of pathological processes within miRNA-target genes, including neuroinflammation, vascular dysfunction, altered barrier permeability, and neuronal damage. A total of 13 central genes were obtained through the topological analysis of the PPI network. Among them, WNT5A, a Wingless-type (Wnt) family member, induces different aspects of synaptic differentiation and plasticity in the hippocampal neurons (Varela-Nallar et al., 2010). WNT5A was a predicted target of miR-4667-5p in the present work, and the protein Wnt5a that it encodes is known to induce the rapid activation of the CaMKII (Farias et al., 2009), thereby reinforcing the Wnt/Ca 2+ signaling pathway in the hippocampal neurons. Wnt5a also evokes cortical axon outgrowth and repression by calcium signaling pathways. It is required in the Wnt5a/CaMKII signaling pathway and is essential for the binding of tau that is  phosphorylated at Ser262 (Li et al., 2014). In addition, dysfunction of the Wnt signaling is linked to an increased Aβ formation, neuroinflammation, and neurotoxicity (Li et al., 2011;Tapia-Rojas et al., 2016). Therefore, a change in WNT5A expression represents a considerable potential as an indicator of VaD pathology.
Other central genes may additionally be involved in different pathological processes of neurodegeneration. For instance, AAK1, a predicted gene regulated by miR-4667-5p, encodes the adaptor-associated kinase 1 and is a modulatory enzyme of the adaptor protein complex 2 (AP-2) responsible for the regulation of clathrin-mediated endocytosis (Smythe and   Ayscough , 2003). In a murine model of AD periodic variation in AAK1 expression, it is closely correlated with cognitive decline (Fu et al., 2018). AAK1 inhibitors have been further developed to provide valuable treatments for disorders such as neuropathic pain, AD, and other neurodegenerative diseases (Abdel-Magid, 2017), indicating that AAK1 serves as a suitable biomarker of cognitive disorders, including VaD. By combining the analysis of various functionalities, the selected miRNAs and their central target genes were screened as biomarkers or potential targets for direct or indirect involvement in VaD. Notably, the present study indicates that TFDM has multiple targets and influences numerous pathways that prevent and treat VaD. On the one hand, TFDM contains 42 flavonoids, reflecting the multi-component action. On the other hand, a single flavonoid component had multiple targets. Thus, a molecular network was established based on the target prediction of each flavonoid contained in TFDM using Cytoscape, which revealed 43 key genes related by TFDM against the pathophysiology of VaD. Among the flavonoids of TFDM, common neuroprotective compounds such as kaempferol, chrysoeriol, apigenin, luteolin, quercetin, and salvigenin exhibited a favorable target docking score performance. Furthermore, eigenvector centrality was used to express the importance of the nodes; those present in a free state (aglycones) exhibited much stronger potential activity toward the targets than the sugar-bound flavone glycosides.
Kaempferol, one of the active compounds in TFDM, is also an essential ingredient in the extracts of Ginkgo biloba leaves, having antioxidant properties by acting as a free radical scavenger , and apigenin (H) after the exposure to OGD. Statistical analysis was performed using one-way ANOVA followed by a Tukey's multiple comparison test. n 3. ### p < 0.001 vs. control; **p < 0.01, ***p < 0.001 vs. OGD. Results are expressed as mean ± SD.
Frontiers in Pharmacology | www.frontiersin.org December 2021 | Volume 12 | Article 796628 (Ahlemeyer and Krieglstein, 2003). Kaempferol plays a role in the regulation of inflammation and apoptosis through the targeting of the PI3-kinase p110-gamma subunit (PIK3CG) (Shen et al., 2020). Luteolin, another active ingredient of TFDM, displays neuroprotective effects on the BBB by inhibiting fibrillary Aβ 1-40induced inflammation through the inhibition of the p38 MAPK signaling pathways, as reported by our previous work . Quercetin, also reported by our research group, provides neurovascular protection by reducing oxidative damage, inactivating RAGE-mediated pathways, and maintaining cholinergic neuron function, thereby providing an alternative medication for neuroprotection . Tilianin, an indicative reference of TFDM, ameliorates the cognitive dysfunction and neuronal damage in VaD by modulating p-CaMKII-mediated long-term memory, oxidation and inflammation associated with the oxidized-CaMKII (Jiang et al., 2021). Taken together, these active compounds might be considered ingredients effective in the treatment of VaD. Furthermore, the underlying mechanisms and potential targets of these active ingredients in TFDM were located into the purpose of the current bioinformatics analysis. This might support the TFDM-mediated neuroprotection on the attenuation of the Aβinduced cognitive decline and cardiocerebrovascular deficit by restoring the anti-amyloidogenic, neurotrophic, and MAPK pathways Zeng et al., 2018). The interactions between the targets and potential binding of miRNAs were analyzed to further explore the potential effective molecules and investigate the underlying mechanism used by these active compounds in TFDM. As demonstrated in the present study, miR-3184-3p might be a potential target gene FIGURE 10 | Regulation of miR-3184-3p and miR-6875-5p by TFDM and active compounds in OGD-injured SH-SY5Y cells. (A) Changes in the expression of miRNAs in OGD-injured SH-SY5Y cells compared with control cells. (B-E) Upregulation of miR-3184-3p in OGD-injured SH-SY5Y cells after the treatment with TFDM (B) and the active compounds tilianin (C), luteolin (D), and apigenin (E). (F-I) Downregulation of miR-6875-5p in OGD-injured SH-SY5Y cells after the treatment with TFDM (F) and the active compounds tilianin (G), luteolin (H), and apigenin (I). Statistical analysis was performed using Student's t-test or one-way ANOVA followed by a Tukey's multiple comparison test. n 3. ### p < 0.001 vs. control; *p < 0.05, **p < 0.01, ***p < 0.001 vs. OGD. Results were expressed as mean ± SD.
Frontiers in Pharmacology | www.frontiersin.org December 2021 | Volume 12 | Article 796628 of all the compounds in TFDM against VaD. miR-3184-3p targeting the PIK3CG was regulated by apigenin, luteolin, and quercetin against VaD. Another most enriched target miRNA was miR-6875-5p that could be regulated by 37 active components of TFDM. miR-6875-5p/G protein-coupled receptor kinase 6 (GRK6) and miR-6875-5p/RPS6KA3 might be responsible for the beneficial action of apigenin, luteolin, and tilianin. According to the results of this study, highly compound-enriched miR-6777-3p was targeted by apigenin, luteolin, quercetin, and kaempferol in the protection of blood vessels by regulating the expression of the steroid hormone receptor ERR1 (ESRRA) and CAMK2. In addition, other axes, including miR-3184-5p/ornithine decarboxylase (ODC1), miR-3675-3p/BDNF/NT-3 growth factors receptor (NTRK2), miR-4667-5p/Protein kinase C alpha type (PRKCA), miR-6784-3p/Lysosomal acid glucosylceramidase (GBA), and miR-6875-5p/MAPK14, are the most likely key mediators in the pathogenesis of VaD. Therefore, the discovery of novel miRNAs that mediate specific signaling pathways provides new insight into the function and molecular mechanisms of TFDM, which acts as a candidate for a multiple targeting therapy. The expression of 5 miRNAs that were most closely related to TFDM in the treatment of ischemic injury was measured by qRT-PCR. The expression of miR-3184-3p and miR-6875-5p in OGDinjured neuronal cells matched the bioinformatics results, suggesting that both miRNAs might play essential roles in the VaD process. In line with the neuroprotective effects of TFDM with respect to OGD injury, the abnormal expression of miR-3184-3p and miR-6875-5p was restored by TFDM, as well as by the active compounds tilianin, luteolin, and apigenin.
Although the present study described putative new biomarkers for VaD, identifying a number of promising DEMs, the results require a greater depth of experimental confirmation, such as the function and interaction assays of key genes in vitro and in vivo. Based on bioinformatic analysis and experimental verification, miRNA-mediated RAGE signaling and CaMKII-induced multitarget pathways should be confirmed, as well as the underlying molecular basis of key genes and miRNA molecules in turn to confirm the therapeutic effects of TFDM on VaD.

CONCLUSION
In conclusion, differentially expressed miRNAs, mRNAs, and their functional properties during the pathogenesis of VaD were identified as contributors through bioinformatics analysis, thus revealing the therapeutic efficacy and molecular mechanisms of TFDM. The qRT-PCR validation suggested that miR-3184-3p and miR-6875-5p might be potential targets of TFDM, which might mediate multiple signaling pathways involved in the pathology of VaD.

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/Supplementary Material.

AUTHOR CONTRIBUTIONS
ML conducted the miRNA profiling analysis, performed the qRT-PCR experiments, and drafted the manuscript. GS completed the sample preparation, identified compounds, and revised the manuscript. HJ, LZ, and KZ performed miRNA-mRNA network analysis. YL performed the HPLC profiling of TFDM. GA participated in the mRNA and miRNA profiling analysis and revised the manuscript. RL and ZL designed the study, interpreted the results, and revised the manuscript. All authors have read and approved the final manuscript.