Network Pharmacology Analysis and Molecular Characterization of the Herbal Medicine Formulation Qi-Fu-Yin for the Inhibition of the Neuroinflammatory Biomarker iNOS in Microglial BV-2 Cells: Implication for the Treatment of Alzheimer's Disease

Aberrant microglial activation drives neuroinflammation and neurodegeneration in Alzheimer's disease (AD). The present study is aimed at investigating whether the herbal formula Qi-Fu-Yin (QFY) could inhibit the inflammatory activation of cultured BV-2 microglia. A network pharmacology approach was employed to predict the active compounds of QFY, protein targets, and affected pathways. The representative pathways and molecular functions of the targets were analyzed by Gene Ontology (GO) and pathway enrichment. A total of 145 active compounds were selected from seven herbal ingredients of QFY. Targets (e.g., MAPT, APP, ACHE, iNOS, and COX-2) were predicted for the selected active compounds based on the relevance to AD and inflammation. As a validation, fractions were sequentially prepared by aqueous extraction, ethanolic precipitation, and HPLC separation, and assayed for downregulating two key proinflammatory biomarkers iNOS and COX-2 in lipopolysaccharide- (LPS-) challenged BV-2 cells by the Western blotting technique. Moreover, the compounds of QFY in 90% ethanol downregulated iNOS in BV-2 cells but showed no activity against COX-2 induction. Among the herbal ingredients of QFY, Angelicae Sinensis Radix and Ginseng Radix et Rhizoma contributed to the selective inhibition of iNOS induction. Furthermore, chemical analysis identified ginsenosides, especially Rg3, as antineuroinflammatory compounds. The herbal formula QFY may ameliorate neuroinflammation via downregulating iNOS in microglia.


Introduction
Alzheimer's disease (AD) is the major neurodegenerative cause of progressive dementia in the elderly [1]. The pathology of AD is hallmarked by the accumulation of extracellular β-amyloid (Aβ) and the formation of peptide plaques and intraneuronal tau lesions, resulting in impaired neurotransmission and neuronal death [2]. Among the existing pharmaco-logical anti-AD interventions, acetylcholinesterase inhibitors are known to restore cholinergic neurotransmission, whereas N-methyl-D-aspartate (NMDA) receptor antagonists suppress the neuronal excitability towards NMDA [3]. However, some controversial results showed that these agents barely prevented the progression of AD and could cause adverse effects [4]. Nevertheless, neurotoxic peptide Aβ activates microglia and exacerbates neuroinflammation, leading to the onset and progression of AD [5,6]. Indeed, the overexpression of proinflammatory enzymes such as inducible nitric oxide synthase (iNOS) and cyclooxygenase-2 (COX-2) jeopardizes the survival of neurons in brains [7]. Notably, excessive nitric oxide in the brain induced oxidative damage in neurons and led to the activation of apoptosis [8]. Therefore, effective inhibition of neuroinflammation represents a key strategy for the management of AD, but the in vivo efficacies of antineuroinflammatory and microglia-targeting agents remain uncertain [9].
The aim of the present study was to characterize the antineuroinflammatory activity of QFY. We employed a network pharmacology approach to analyze the compound-target interactions of QFY and the relevant signaling pathways. We further validated the antineuroinflammatory property in microglial BV-2 cell culture and identified the principal active compounds through a bioactivity-guided fractionation procedure.

Chemicals and Reagents.
The dried aqueous extracts of seven QFY ingredients were manufactured by Nong's Pharmaceutical Ltd., Hong Kong. Dulbecco's modified Eagle's medium (DMEM), fetal bovine serum (FBS), and penicillin/streptomycin solution were purchased from Invitrogen (Carlsbad, CA, USA). Protein assay dye reagent concentrate was purchased from Bio-Rad (Hercules, CA, USA). Lipopolysaccharide (LPS), RIPA assay buffer, protease inhibitor cocktail, and anti-rabbit horseradish peroxidase-(HRP-) conjugated IgG secondary antibody were purchased from Sigma-Aldrich (St. Louis, MO, USA). An antibody against iNOS was purchased from Abcam (Cambridge, UK). Antibodies against COX-2 and GAPDH were purchased from Cell Signaling Technology (Boston, MA, USA). Anti-mouse HRP-conjugated IgG was purchased from Santa Cruz Bio-technology (Santa Cruz, CA, USA). An enhanced chemiluminescence (ECL) detection reagent was purchased from GE Healthcare (Uppsala, Sweden). Ginsenoside Rg3 with the purity of >98% was purchased from Nanjing Spring and Autumn Biological Engineering Company (Nanjing, China).

2.2.
Identification of the Active Compounds in the Formulation QFY. The chemical compounds were selected from the herbs of QFY in the databases of TCMSP (https:// tcmspw.com/) based on the oral bioavailability ðOBÞ ≥ 30%, drug-likeness ðDLÞ ≥ 0:18, and blood-brain barrier ðBBBÞ ≥ −0:3. On the other hand, the compounds in the herbal ingredient not found in TCMSP were selected from the database of TCMID (http://www.megabionet.org/tcmid/) when the compounds possessed the blood-brain barrier (BBB) permeability of ≥-1 through checking pkCSM (http://biosig .unimelb.edu.au/pkcsm/) and did not violate more than one of the criteria stated in Lipinski's rule of five. In addition, the active compounds were collected by text mining. 2.4. Gene Ontology and Pathway Enrichment Analysis. The selected targets were analyzed for the enriched Gene Ontology (GO) terms and the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway using the online bioinformatics tool DAVID 6.8 at http://david.ncifcrf.gov. The annotations with adjusted p < 0:05 were considered significantly enriched.

Prediction of
2.5. Visualization of the Compound-Target Interaction and the Target-Pathway Relationship. The interaction networks were visualized using the open software package Cytoscape (http://www.cytoscape.org/). For the compound-target interactions, the active compounds in QFY and the corresponding protein targets were presented in a compound-target (C-T) network. For the target-pathway interactions, the targets and the related pathways were presented in a targetpathway (T-P) network.
2.6. Microglial BV-2 Cell Culture and Treatment. The murine microglial cell line BV-2 was purchased from the American Type Culture Collection (Manassas, VA, USA) and cultured in DMEM supplemented with 10% heat-inactivated FBS and 1% penicillin/streptomycin solution. The cells were incubated at an atmosphere with 5% CO 2 at 37°C. For the indicated treatments, BV-2 cells were plated in 6-well plates at a density of 2 × 10 5 /mL overnight. BV-2 cells were treated with herbal extracts as indicated and costimulated with 0.1 μg/mL LPS in DMEM supplemented with 3% FBS for 24 hours. 2.8. Western Blot Analysis. After drug treatment as indicated, BV-2 cells were washed twice with cold PBS and lysed with cold RIPA buffer to collect the cellular proteins. The cellular proteins (30 μg) from each sample were resolved by 12% SDS-polyacrylamide gel electrophoresis and transferred to a polyvinylidene difluoride (PVDF) membrane. After overnight blocking with 5% non-fat milk, the membranes were probed with primary antibodies against iNOS, COX-2, and GAPDH overnight. The bound antibodies were detected with the HRP conjugate of the corresponding anti-rabbit or antimouse secondary antibodies for 2 hours. The activity of HRP was detected with ECL detection reagents. The blots were quantified by a densitometric method using the NIH ImageJ software. MassHunter Quantitative Analysis B.06.00 from Agilent Technologies (Santa Clara, CA, USA).

Statistical
Analysis. Data was presented as mean ± SD of three replicates. The difference between two groups was analyzed by one-way ANOVA and Tukey's multiple comparison using the GraphPad Prism 7 software (La Jolla, CA, USA). A p value of <0.05 was considered statistically significant.

QFY Targets Key Proteins in Inflammation and AD
Pathogenesis. To characterize the biological targets of QFY in a comprehensive manner, we employed a network pharmacology approach to dissect the complex multicompound and multitarget interaction. A total of 812 compounds in QFY were firstly retrieved from TCMSP and TCMID. Based on the pharmacokinetic or physiochemical properties, 126 active compounds were obtained while another 19 compounds were added by searching the PubMed database for the implications in the treatment of AD. The nomenclature of the active compounds was guided by PubChem, and the detailed information is listed in Supplementary Tables 1 and 2. The active compounds were further classified by the Medical Subject Headings (MeSH) classification system and text mining. The candidate compounds were chemically categorized into flavonoids, terpenoids, alkaloids, xanthones, phenylpropanoids, lipids, glycosides, carboxylic acids, phenolic compound, and amino acids. The protein targets were fished out using various predictive models (e.g., SEA, STITCH, and SwissTargetPrediction). As shown in Figure 1(a), 25 protein targets formed a total of 213 interactions with 96 active compounds. Specifically, MAPT as an important pathological target in AD showed the highest connectivity with 56 of the active compounds. Several active compounds in QFY showed interactions with other AD-relevant mediators (e.g., BACE1, APP, and SNCA), cholinergic neurotransmission (e.g., ACHE, CHRM1, CHRNA4, and CHRNB2), and inflammation (e.g., COX-2, eNOS, and iNOS). For the target-pathway network as shown in Figure 1(b), the protein targets were mapped to 14 different biological pathways or diseases in KEGG. Six protein targets (i.e., MAPT, BACE1, APP, SNCA, MAOA, and MAOB) were categorized in "Alzheimer's disease," and five targets (i.e., ACHE, CHRM1, CHRNA4, CHRNB2, and FOS) were grouped to "cholinergic synapse," while another three targets (i.e., CASP8, FOS, and COX-2) were grouped to "TNF signaling pathway". To characterize the functional annotations of the protein targets, Gene Ontology (GO) and KEGG pathway enrichment analysis were carried out on the online bioinformatics tool DAVID. As shown in Figure 2, the inflammation-or AD-related GO terms appeared in the ten most significantly enriched GO terms from "biological process" and "molecular function," in particular, acetylcholinesterase activity (GO: 0003990), amyloid-beta binding (GO: 0001540), and response to lipopolysaccharide (GO: 0032496). The three most significantly enriched KEGG pathways were Alzheimer's disease (KEGG: hsa05010), cholinergic neurotransmission pathway 3 Oxidative Medicine and Cellular Longevity  Oxidative Medicine and Cellular Longevity (KEGG: hsa04725), and serotonergic neurotransmission pathway (KEGG: hsa04726).

QFY Differentially Affected iNOS and COX-2 Expression against LPS Stimulation.
To validate the antineuroinflammatory action of QFY and identify the corresponding active compounds, we developed a bioactivity-guided fractionation approach involving ethanolic precipitation, HPLC separation, and in vitro assays (Figure 3(a)). The aqueous extract of QFY was firstly precipitated in 50%, 75%, and 90% ethanol. The ethanol solutions and aqueous extract of QFY were assayed for the effects on iNOS and COX-2 in LPSstimulated BV-2 cells. As shown in Figure 3(b), LPSinduced iNOS expression was decreased by these ethanol solutions of QFY. The 90% ethanol solution of QFY (90% QFY) selectively suppressed LPS-induced iNOS expression in a concentration-dependent manner. In contrast, none of the four QFY preparations effectively suppressed LPSinduced COX-2 expression. To identify the specific compounds, 90% QFY preparation was separated into nine fractions by reverse-phase HPLC on a C18 column (Figure 3(c)). Based on the results of Western blot analysis in Figure 3(d), fractions 1, 2, and 9 showed strong activity to downregulate LPS-induced iNOS expression. Interestingly, these active fractions showed comparable activity with the parent 90% QFY preparation, whereas the other fractions did not show much activity.

Ginsenosides in RS Were Identified as the Principal Active
Compounds in QFY. To identify the active ingredients, seven herbal ingredients in QFY were processed into 90% ethanol solutions and assayed for downregulating iNOS in microglia following the procedure described for the whole QFY formulation. As shown in Figure 4(a), 90% ethanol extracts of DG (90% DG) and RS (90% RS) abolished LPS-induced iNOS expression but showed no effect against COX-2 induction. The 90% RS preparation appeared to be more potent than the 90% DG preparation, although less RS was used in the formula than DG. For this reason, the 90% RS solution was selected for further bioactivity-guided fractionation. The compounds in the 90% RS preparation were separated by HPLC under the same conditions as described for the 90% QFY preparation (Figure 4(b)). The resulting fractions were assayed for the activity towards iNOS induction. As shown in Figure 4(c), fraction 9 (90% RS-F9) effectively and selectively suppressed LPS-induced iNOS expression to a greater extent than the parent 90% RS solution.
To identify the specific ginsenosides, the 90% RS preparation and 90% RS-F9 were separated on a C18 HPLC column and detected with a MALDI-TOF mass spectrometer. As shown in Figure 5(a), the 90% RS-F9 fraction showed identical distribution of base peaks compared with the parent 90% RS solution within the elution time interval. As listed in Table 1, a total of 18 ginsenosides and one notoginsenoside were identified in the 90% RS-F9 fraction in a negative ion  To compare the migration pattern with pure compounds, commercial ginsenoside Rg3 was spiked in the 90% RS-F9 fraction prior to LC-MS/MS analysis. As shown in Figure 5(b), ginsenoside Rg3 was verified by using the identical elution time and mass fragmentation pattern (data not shown). In addition, ginsenoside Rg3 was quantified from the acquired scan data using a target ion of 829.50. The calibration curve for ginsenoside Rg3 (y = 62240:86x − 24991:72, R 2 = 0:9993) was optimized over the concentration range of 3.9-125 μg/mL. The calculated amount of ginseno-side Rg3 in 90% RS and 90% RS-F9 was 2:11 ± 0:02 mg/g and 2:05 ± 0:08 mg/g, respectively.
To examine the potency of different preparations to downregulate iNOS expression, ginsenoside Rg3 and the parent extracts (i.e., 90% QFY and 90% RS) were normalized based on quantification results in the assay of the in vitro effects. As shown in Figure 5(c), 90% QFY and 90% RS suppressed iNOS induction to an extent similar to 10 μM and 20 μM of ginsenoside Rg3, respectively.

Discussion
Microglial activation triggers neuroinflammation and neurodegeneration in the onset of AD [20]. Single-drug therapies such as nonsteroidal anti-inflammatory drugs (NSAIDs), TNF-α inhibitor, advanced glycation end product receptor  Figure 2: GO terms and KEGG pathway enrichment analysis of the protein targets. (a) The most significantly enriched GO terms were selected in terms of "biological process." (b) The most significantly enriched GO terms were selected in terms of "molecular function." (c) The most significantly enriched KEGG pathways were selected in terms of "KEGG pathways". 6 Oxidative Medicine and Cellular Longevity      10 Oxidative Medicine and Cellular Longevity  11 Oxidative Medicine and Cellular Longevity inhibitor, and PPARγ receptor agonist show varying efficacies in the treatment of AD [21]. Recent effort is directed to the discovery of multitarget drugs by a systems biology approach [22]. Many herbal formulas are well documented for the clinical treatment of dementia in traditional Chinese medicine, thereby representing a rich source for drug discovery [23]. The present study investigated the classical herbal formula QFY for the active compounds and protein targets and hence the molecular mechanisms for the management of AD. We firstly employed a network pharmacology approach to predict the biological targets of QFY and the relevant signaling pathways. We further identified the active compounds from QFY for targeting the neuroinflammatory biomarker iNOS.
Network pharmacology encompasses systems biology, pharmacology, and computational algorithms to robustly study the complex drug-target relationships [24,25]. Several herbal medicines including RS and YZ were previously studied for the molecular targets against AD by a network pharmacology approach [26]. In the present study, we employed a network pharmacology approach to study the action of the comprehensive anti-AD herbal formula QFY. Firstly, we shortlisted the active compounds with satisfactory bioavailability and BBB penetration based on the common pharmacokinetic or physiochemical parameters curated in databases. We secondly predicted 25 protein targets for these active compounds by target fishing. The C-T network in Figure 1(a) demonstrated that multiple compounds interacted with each protein, whereas the T-P network in Figure 1(b) also vividly depicted that the protein targets were involved in different pathways. Finally, the selected targets were subjected to GO and KEGG enrichment analysis. QFY may exert the pharmacological effect against AD by modulating Aβ aggregation, acetylcholinesterase activity, nitric oxide synthase activity, and cholinergic neurotransmission ( Figure 2).
Neuroinflammation is well known to cause extensive damage to the central nervous system, leading to the disruption of synaptic function and the exacerbation of Aβ pathology [27]. Along this line, the present study focused on the antineuroinflammatory activity of QFY and selected two key proinflammatory biomarkers iNOS and COX-2 from the prediction of network pharmacology. It was previously demonstrated that both iNOS and COX-2 were overexpressed in the activated microglia and contributed to neurodegeneration [28]. To investigate the effect of QFY on the expression of iNOS and COX-2, we adopted a bioactivityguided fractionation approach to separate QFY into different fractions by water extraction, ethanolic precipitation, and HPLC fractionation. We prepared the QFY formulation from the dried, concentrated granules of different herbal ingredients, which facilitated the extraction process and produced a similar chemical profile to regular decoction [29]. We assayed all fractions for the activity to suppress LPSinduced expression of iNOS and COX-2 in microglial BV-2 cells. Based on Western blot analysis, the aqueous QFY extract potentiated the stimulatory effect of LPS on the expression of iNOS and COX-2. The stimulatory effect on microglia could be attributed to the water-soluble active immunopolysaccharide in the herbal ingredients [30]. On the other hand, alcoholic precipitation removed charged small molecules and polysaccharides from the aqueous herbal extracts that could stimulate the cells [31,32]. Indeed, the 90% QFY preparation selectively suppressed LPSinduced iNOS expression in microglial BV-2 cells.
A bioactivity-guided fractionation approach is widely used to separate complex chemical mixture into different fractions, facilitating the identification of active compounds for specific biological entities [33][34][35]. In the present study, we firstly prepared a water extract and 50%, 75%, and 90% QFY ethanol extracts. Based on the bioassay results, the 90% QFY preparation appeared to be the active fraction. Therefore, the 90% QFY preparation was subsequently separated into 9 fractions by HPLC (Figure 3(c)). Western blot analysis suggested that fractions 1, 2, and 9 effectively suppressed LPS-induced iNOS expression, although different chemicals might contribute to the activity (Figure 3(d)). For the identification of the active compounds, we screened seven herbal ingredients in the formulation QFY and identified that 90% DG and 90% RS preparations selectively suppressed iNOS induction in LPS-challenged BV-2 cells (Figure 4(a)). We selected the 90% RS preparation for further identification of the active compounds since RS-derived active compounds better met the requirements for an orally active drug in network pharmacology analysis. Thus, the 90% RS preparation was separated into 9 fractions by HPLC using the same gradient and fractionation as described for the 90% QFY preparation. Western blot analysis suggested that fraction 9 derived from 90% RS best suppressed LPS-induced iNOS expression (Figures 4(b) and 4(c)). The 90% RS and 90% QFY preparations showed similar efficacy of inhibiting iNOS expression. In fact, previous studies demonstrated that RS exhibited excellent antineuroinflammatory and anti-Aβ potency [17,36]. Based on the LC-MS/MS profile, fraction 9 might contain several antineuroinflammatory ginsenosides such as Rh1, Rb2, Rd, Rg3, and Rg5 [37][38][39][40]. In particular, ginsenoside Rg3 could downregulate iNOS against LPS stimulation in macrophages and promote the phenotypic switch of macrophages towards a proresolving M2 subtype [41,42]. Indeed, LC-MS/MS analysis confirmed the presence of ginsenoside Rg3 in 90% RS-derived fraction 9 since ginsenoside Rg3 spike-in comigrated with the endogenous ginsenoside Rg3 (Figure 5(b)). Furthermore, 10 μM ginsenoside Rg3 appeared to achieve the same potency of 90% QFY and 90% RS for suppressing iNOS induction, although ginsenoside Rg3 failed to exhibit similar activity at the concentration in 90% RS as deduced by the UPLC quantification ( Figure 5(c)). Presumably, other compounds may act on iNOS induction in synergy with ginsenoside Rg3. As for the differential effects of QFY fractions on LPS-induced expression of iNOS and COX-2, we previously found that the concomitant activation of the antioxidant Nrf2/HO-1 pathway was a potential mechanism to support the selective suppression of iNOS over COX-2 in macrophages [43]. As elevation of iNOS resulted in oxidative damage in neurons, activation of the Nrf2/HO-1 pathway could effectively inhibit inflammation and oxidative stress, which restored cellular function to the physiological state [44]. Previous studies reported that ginseng and ginsenoside Rg3 could activate the Nrf2/HO-1 pathway [45,46]. We speculate that ginsenoside Rg3