Deciphering the Pharmacological Mechanisms of Guizhi-Fuling Capsule on Primary Dysmenorrhea Through Network Pharmacology

The dosages and ratios of GZFLC compounds could influence their intervention intensity on targets. Considering the influence of GZFLC compounds’ dosages and ratios, we selected major phytochemistry compounds and absorbed compounds of GZFLC as representative compounds to analyze GZFLC’s pharmacological mechanisms through network pharmacology. We have investigated phytochemistry compounds of GZFLC and absorbed compounds of GZFLC in plasma previously (Zhong et al., 2016a; Yin et al., 2016; Zhang et al., 2017; Yang et al., 2020). In total, 51 compounds were selected from our previous investigations of phytochemistry and absorbed compounds of GZFLC based on the following criteria: 1) the characteristic compounds were accounted for the major proportion and verified with reference substances; 2) the compounds were quantificationally recorded in each herb from Chinese Pharmacopoeia (2015) (Committee 2015); 3) each compound had specified chemical structure and Pubchem CID.

GZFLC was provided by the company Jiangsu Kanion Pharmaceutical Co., Ltd., composed of Cinnamomum cassia (L.) J. Presl (20%), Poria cocos (20%), Paeonia × suffruticosa Andrews (20%), Paeonia lactiflora Pall (20%), Prunus persica (L.). Batsch (20%). The full taxonomic names of all herbs included have been fully validated using the Kew Medicinal Plant Names Service. Guizhi-Fuling capsules have established fingerprints (Wang et al., 2009; Li et al., 2012) from herb to preparation to make sure that the quality of the preparation is uniform and stable (Li et al., 2015). Guizhi-Fuling capsule is a Chinese patent medicine approved and supervised by CFDA under the same strict quality control standards. Meanwhile, Guizhi-Fuling capsules in our study have come from the same preparation. The chemical constituents analysis of GZFLC has been reported in the literature (Yang et al., 2020). The UPLC analysis of GZFLC (Jiangsu Kanion Pharmaceutical Co., Ltd. Jiangsu, China, No. 20160319) was performed on Agilent SB-RRHD C18 (100 mm × 2.1 mm, 1.8 μm) column at 30°C with a mobile phase consisting of acetonitrile and 0.02% trifluoroacetic acid for gradient elution and the flow rate of 0.2 ml/min. The detection wavelength was set at 230 nm. The representative UPLC fingerprints of GZFLC were shown in Supplementary Figure S1 in the Supplementary Materials.

To explore the relativity of representative GZFLC compounds and targets, the potential target profiles of compounds in GZFLC were predicted by drugCIPHER-CS (Zhao and Li, 2010), a high-accuracy algorithm developed for global prediction of compound targets. In principle, drugCIPHER-CS used a linear regression model which related chemical similarity vector between compounds in GZFLC and drugs in DrugBank (Wishart et al., 2008) to the drug-protein closeness vector based on a protein–protein interaction network to calculate the concordance score of each compound-protein pair. The concordance score was treated as the likelihood of the compound targeting the protein. Thus, the candidate proteins with high concordance scores were prioritized to be potential targets of the compound. Top 100 ranking predicted proteins of each compound were kept as potential target profiles.

To validate the target prediction of compounds, literature mining method was used to obtain the co-occurrence results of biomolecules and each compound. The biomolecules related to each compound from literature mining were manually examined to delete the false positive results and then considered as the reported targets. Comparing the reported targets with the predicted targets via direct mapping and indirect link, the reliability of target prediction was measured by. | | The predicted targets related to the reported targets | | The predicted targets ∗ % 100

To identify potential targets regulated by GZFLC or herbs, predicted proteins ranking top or appearing in the target profiles of many compounds in GZFLC or herbs were hypothesized to be important in the pharmacological effects of GZFLC or herbs. Based on this hypothesis, the potential targets of GZFLC or herbs were considered as significantly frequently occurring targets (p < 0.05) by comparing the number of occurrences of a target protein in the target profiles of all compounds in GZFLC or herbs to that in random background (Liang et al., 2014). A Poisson binomial statistical model was used to obtain the random background distribution and represented as:P(K=k)=∑A∈Fk∏j∈Api∏j∈Ac(1−pj).P(K=k) is the probability of a certain target in k compounds’ target profiles, Fk is the set including all sets containing k compounds, A is an element of Fk, and Ac is the complement of A. pi and pj are the probabilities of a certain target in iand jcompound’s target profiles respectively.

To investigate the potential signaling pathways or biological processes regulated by GZFLC and each herb, functional enrichment analysis was carried out in Comparative Toxicogenomics database (Davis et al., 2019) to obtain the over represented Gene Ontology (GO) biological processes (BP)/KEGG signaling pathways based on the potential targets of GZFLC and each herb. The significantly enriched GO BP/KEGG signaling pathways with p < 0.05 after Benjamin’s correction were selected for further study. All significantly enriched GO BP/KEGG signaling pathways were categorized by means of key words mapping.

The herbs-compounds-biological functional modules-biological molecules network of GZFLC was constructed to elucidate its putative anti-primary dysmenorrhea mechanisms by considering protein–protein interactions and crosstalk among pathways. First, herbs were connected to their representative compounds. Second, the potential targets of GZFLC were mapped into protein–protein interaction network and the targets were annotated by its biological functions. Third, an edge was added between herbs and biological functional modules if potential targets of a certain herb were significantly enriched in biological functional modules related GO BP or KEGG signaling pathways (p < 0.05). The herbs-compounds-biological functional modules-biological molecules network was visualized using Cytoscape v3.7.2 (Shannon et al., 2003).

We hypothesized that the drugs with similar target profiles predicted by drugCIPHER-CS may have similar biological activities. The resemblance of compounds’ and drugs’ biological activities could be further defined on the basis of the similarity between concordance scores of their target profiles. Hierarchical clustering analysis of target profiles was applied to measure the biological activity resemblance of compounds and drugs. Thus, to identify anti-primary dysmenorrhea candidate compounds, hierarchical clustering analysis was conducted using R statistic software (R version 3.6.3) based on target profiles of FDA-approved anti-primary dysmenorrhea drugs and 51 compounds of GZFLC.

As GZFLC had multiple compounds and targets, we first qualitatively explored the relativity of representative GZFLC compounds and targets. And then we selected predicted compounds and a specific target for preliminary analysis of their dosages and ratios as a starting point. In vitro enzyme activity assays were used to analyze the relationship between predicted GZFLC compounds’ different dosages and their intervention intensity on a specific target COX2, and to investigate GZFLC’s intervention intensity on COX2 under the specific ratio of GZFLC compounds which is relatively fixed in GZFLC. The content in the GZFLC (Jiangsu Kanion Pharmaceutical Co., Ltd. Jiangsu, China, No. 170601) was weighed and extracted with dimethyl sulfoxide (DMSO, 32.308:1, W/V, mg/ml) at 37°C for 30 min. The liquid was collected by centrifugation, and the supernatant was filtered through a 0.22 μ filter. The effect of GZFLC or pure chemicals including 1,2,3,4,6-penta-O-galloyl-beta-d-glucopyranose (purity (%) ≥98%, CAS: 14,937–32–7, Pubchem CID: 65238, Nanjing SenBeiJia Biological Technology Co., Ltd.), galloylpaeoniflorin (purity (%) ≥98%, CAS: 122,965–41–7, Pubchem CID: 46882879, Nanjing SenBeiJia Biological Technology Co., Ltd.), ethyl gallate (purity (%)≥98%, CAS: 831–61–8, Pubchem CID: 13250, Nanjing SenBeiJia Biological Technology Co., Ltd.) and gallic acid (purity (%)≥98%, CAS: 149–91–7, Pubchem CID: 370, Nanjing SenBeiJia Biological Technology Co., Ltd.) on COX2 activity was measured using cyclooxygenase-2 inhibitor screening kit (Beyotime, Cat. No. #S0168). Briefly, the recombinant human COX2 assay mixture (total volume = 90μL/well) consisting of assay buffer, cofactor, COX2 enzyme, and test compound (diluted in DMSO, 5.56% DMSO finally in assay) was incubated at 25°C for 5 min. Detection of the product was performed by adding probe (5μL/well). Arachidonic acid was added and the reaction was allowed to proceed for 15 min at 37°C. The fluorescence of the resulting solution was measured using SpectraMax M2e multilabel plate reader (Molecular Devices, United Kingdom). COX2 specific inhibitor ibuprofen (98.82% purity, AbMole, Cat. No. M3359) was included as a positive control. Data was represented as X¯±SD of duplicated samples. The effect of the samples on COX2 was calculated using the following equation:Relative activity of COX2(%)=(RFUsample−RFUBWRFUA−RFUBW)×100%.

Female ICR mice (18–22 g of weight, SPF, license number: SCXK (SU) 2012–0004) were provided by the comparative medicine center of Yangzhou University. All animal experimental procedures were performed in accordance with the Guide for the Care and Use of Laboratory Animals and were approved by the animal ethics committee of the Institutional Animal Care and Use Committee (IACUC) of Kanion Pharmaceutical Co., Ltd.

Isolation and culture of primary myometrial cells were prepared as previously reported (Mosher et al., 2013; Sun et al., 2017). Briefly, Uterine were removed from adult female ICR mice (7–8 weeks old, purchased from Nanjing Branch of Beijing Weitong Lihua Laboratory Animal Technology Co., Ltd., license number: SCXK (SU) 2016–0003). Uterine were cut longitudinally, and endometria were scraped gently. The myometrial cells were digested in 2 ml collagenase Ⅱ (GIBCO, Cat. No. 1797319) (0.025 g L⁻¹), then cells were plated onto 25 cm² culture bottle (Costar, US) and were cultured in complete DMEM (GIBCO, Cat. No. 11330032) at 37°C with 5%CO₂.

Myometrial cells were seeded in 96 well black-walled clear-bottom plates (Costar, US) (4×10⁴/ml density), and incubated overnight at 37°C in air containing 5%CO₂. The blank group and model group were added with 100 μL serum-free medium, and the GZFLC group was added with GZFLC (Jiangsu Kanion Pharmaceutical Co., Ltd. Jiangsu, China, No. 20160319) solution prepared with serum-free medium (25 g⋅L⁻¹) for 1 h. The supernatant was discarded, 100 μL serum-free medium was added to the control group, and 100 μL oxytocin (Cayman, 11,799) (final concentration of 28 μM) was added to the model group and GZFLC group for 15 min respectively. The supernatant was discarded, fixed with paraformaldehyde (Amresco, Cat. No. J531) (0.4 g⋅L⁻¹) for 15 min, washed with PBS for 5 min 3 times, and blocked with 5% goat serum (Boster, Cat. No. 10E12B09)/0.3% tritonx-100 for 1 h. Then cells were incubated with primary antibody against p-SAPK/JNK(CST), p-p44/42 MAPK(Erk1/2) (CST), p-p38(CST) (1:150, 1:150, 1:2000) at 4°C overnight. The next day, it was equilibrated to room temperature, followed by 1 h incubation with Alexa Fluor^® 488 Conjugate (1:1000, Invitrogen, United States) at room temperature, washed 3 times with PBS, counterstained with DRAQ5^® for 5 min, washed once with PBS, and imaged with a high-content instrument (Thermo, ArrayScanⅦ).

Oxytocin-induced dysmenorrhea in mice was performed according to the method previously reported (Sun et al., 2002). Fifty female ICR mice (license number: SCXK (SU) 2012–0004) were evenly divided into five groups as follows: normal group, model group, ibuprofen group and the GZFLC groups of different doses (0.54 g/kg and 1.08 g/kg). Except for the normal group (subcutaneous injection of saline), the other groups were administrated by subcutaneous estradiol benzoate injection (0.05 ml/mouse on the first and 10th day, and 0.025 ml/mouse on the remaining days) for ten consecutive days. The model group, GZFLC groups, and Ibuprofen group were orally administrated on the fourth day with 0.5% CMC-Na of the same volume (20 ml/kg), GZFLC (0.54 g/kg and 1.08 g/kg) and ibuprofen (0.1 g/kg) for seven days (once a day) in the period, respectively. On the 10th day, oxytocin (0.2 ml/2 IU/mouse) was administered by peritoneal injection 1 h after the last administration. The number of animal writhing times was observed and recorded within 30 min after injection of oxytocin. The number of writhing inhibition rate (IR %) was calculated according to the formula (inhibition rate% = (number of writhing times in the model group-number of writhing times in the medication group)/number of writhing times in the model group) × 100%). The mice were sacrificed immediately. The uterus tissues were fixed with 10% formalin for further study.

The uterus tissues were fixed in 10% formalin and stained with hematoxylin and eosin (H&E). Histopathological changes were observed by optical microscope (OLYMPUS DX45, Japan) including degeneration and necrosis in the endometrial epithelial cells, congestion, edema, inflammatory cell infiltration in the lamina propria, the number of glands (increased or decreased), secretion in the gland cavity, and lesions in the muscle layer and serosal layer. According to the severity of the lesion, it was recorded as one point (mild), two points (moderate), three points (severe), four points (extremely severe), no lesion was recorded as 0 point. All scores were accumulated and the average score () of each animal in each group was calculated. X ¯ ± S D

The Fisher’s exact test was used to identify significantly enriched KEGG signaling pathways and GO biological processes. The immunofluorescence assay results were analyzed by SPSS 13.0 software. One-way and two-way analysis of variance (ANOVA) were used to calculate statistical levels between groups. The enzyme activity assay results were analyzed using Graphpad Prism software (Graph Pad, United States) presented as X¯±SD of duplicated samples. The animal assay results were analyzed using Graphpad Prism software (Graph Pad, United States) presented as X¯±SEM or X¯±SD of duplicated samples. p-value less than 0.05 was considered to be statistically significant.

We selected 51 representative compounds of GZFLC according to our previous investigations to further study. It has been known that bioactive natural compounds may have multiple targets to exert certain therapeutic effects. To reveal the pharmacological mechanisms of GZFLC for treating primary dysmenorrhea, potential targets of 51 representative compounds were predicted by our state-of-art network-based algorithm drugCIPHER-CS due to the lack of these compounds’ known target records. The top 100 predicted targets of each compound were selected as its potential targets for high precision in this algorithm, called target profiles of each compound. Moreover, the reliability of the predicted targets was verified making use of literature mining method based on text searching. The predicted targets were validated by the reported targets with literature evidence via direct mapping or indirect link as shown in Figure 2A. The precision of predicted targets was used to measure the reliability, which stands for the percentage of the predicted targets supported by literatures. Greater than 73% of the predicted targets of compounds could be supported with literature evidence in Figure 2B, while the others could be novel targets which would be investigated in the future study. These results indicated the reliability of our predicted targets of GZFLC compounds used for further investigation of pharmacological mechanisms.

At first, there was a hypothesis that if a biological molecule was an important target in the pharmacological effects of GZFLC, it may rank top or appear in the target profiles of many compounds in GZFLC. Then, based on this hypothesis, we selected the most representative targets of GZFLC compared to the null model with Poisson binomial statistics as potential targets regulated by GZFLC to uncover its therapeutic mechanisms for primary dysmenorrhea. Finally, we got 240 potential targets as GZFLC network target and made functional enrichment analysis on these targets. The significantly enriched biological processes and signaling pathways were achieved as listed in Table 1. According to key words mapping, the network target of GZFLC significantly enriched in some pain, inflammation, endocrine, blood circulation and energy metabolism related biological processes and signaling pathways (p < 0.05). Hence, potential functional modules regulated by GZFLC on primary dysmenorrhea could be related to pain, inflammation, endocrine, blood circulation and energy metabolism.

We also got the potential targets of each herb in GZFLC respectively and conducted the functional enrichment analysis for these targets of each herb with the significantly enriched biological processes and signaling pathways of GZFLC as shown in Figure 3A. Those results could help us to better elucidate effects of GZFLC for treating primary dysmenorrhea. Cinnamomum cassia (L.) J. Presl, Prunus persica (L.) Batsch mainly regulated pain, inflammation, endocrine, blood circulation modules. Paeonia lactiflora Pall. and Paeonia × suffruticosa Andrews mainly regulated pain, inflammation, endocrine, blood circulation and energy metabolism modules. Poria cocos mainly regulated inflammation, endocrine and energy metabolism modules. To further elucidate the underlying molecular mechanisms of GZFLC, the biological molecular network targeted by GZFLC was constructed to uncover the relationships among herbs, compounds, biological functional modules and targets in Figure 3B. These results suggested that GZFLC could reduce pain and inflammation, and improve endocrine, blood circulation and energy metabolism, which was supported by evidences from the following in vitro and in vivo experiments.

Moreover, the hierarchical clustering analysis of GZFLC compounds and first-line drugs for primary dysmenorrhea was performed to further investigate the anti-primary dysmenorrhea activities of GZFLC compounds. The therapeutic effects of first-line drugs might suggest hypothetical effects for several compounds in GZFLC within the same cluster. The target profiles of GZFLC compounds and nine FDA-approved anti-primary dysmenorrhea drugs including non-steroidal anti-inflammatory drugs (NSAIDs: Ibuprofen, Naproxen sodium, Mefenamic acid and celecoxib) and oral contraceptives (OCs: Norgestimate, Ethynodiol diacetate, Drospirenone, Mestranol and Norethisterone) were predicted by drugCIPHER-CS firstly. Then, hierarchical clustering analysis was performed based on target profiles of GZFLC compounds, NSAIDs and OCs. These two kinds of first-line drugs have different mechanisms on primary dysmenorrhea. First, NSAIDs are often taken to exert analgesic effects via inhibition of cyclooxygenase (Brune and Patrignani, 2015). The compounds of GZFLC in the same cluster with the NSAIDs might be potential active compounds to exert analgesic effects. As shown in Figure 3C, the four NSAIDs were clustered into two major groups including 36 compounds of GZFLC, indicating that these compounds might have a similar mechanism of analgesic effects to relieve primary dysmenorrhea with NSAIDs. Second, the OCs were used to regulate hormone levels to suppress ovulation, thereby reducing dysmenorrhea (Proctor et al., 2001). The compounds in GZFLC in the same cluster with OCs were considered to have a similar therapeutic mechanism on primary dysmenorrhea with OCs. As shown in Figure 3C, the five oral contraceptives were clustered with 23 compounds from GZFLC, which mainly derived from Paeonia lactiflora Pall. and Paeonia × suffruticosa Andrews, suggesting that these compounds might have the bioactivity of regulating hormone levels to improve endocrine, thereby relieving primary dysmenorrhea.

Recent studies have demonstrated that hyper-secretion of prostaglandins and an increased uterine contractility are two causes of pain associated with dysmenorrhea (Bernardi et al., 2017). NSAIDs exert analgesia effects to effectively alleviate primary dysmenorrhea through decreasing prostaglandin levels via inhibition of cyclooxygenase-mediated production (Marjoribanks et al., 2010). COX2 is an important cyclooxygenase that catalyzes arachidonic acid conversion to prostaglandins, which is a major target of NSAIDs. Meanwhile, COX2 was predicted as an important biomolecule regulated by GZFLC from our network target analysis. We therefore examined whether GZFLC and its potential bioactive compounds above could affect COX2 activity. We found that GZFLC and four predicted main compounds, including 1,2,3,4,6-penta-O-galloyl-beta-d-glucopyranose, galloylpaeoniflorin, ethyl gallate and gallic acid, could inhibit the activity of COX2 in vitro. As shown in Figure 4A, GZFLC inhibited the activity of COX2 in vitro with an IC50 value of 5.92 μg/ml 1,2,3,4,6-penta-O-galloyl-beta-d-glucopyranose, galloylpaeoniflorin, ethyl gallate and gallic acid inhibited the activity of COX2 in vitro with IC50 values of 0.38 μM (0.36 μg/ml), 0.74 μM (0.47 μg/ml), 1.11 μM (0.22 μg/ml) and 1.95 μM (0.33 μg/ml) respectively. Ibuprofen was selected as the positive control in our study and the results showed that ibuprofen potently inhibited the activity of COX2 with an IC50 value of 50.09 μM (10.33 μg/ml).

As a result, the network target of GZFLC was significantly enriched in MAPK signaling pathway. MAPKs could increase cPLA2 activity and hence result in prostaglandin production and myometrial contraction (Lin et al., 1993). Therefore, we used the immunofluorescence assay to determine whether GZFLC could affect MAPK signaling pathway to reduce myometrial contraction. We found that uterine smooth muscle cells in oxytocin-induced dysmenorrhea murine model displayed the increase of p-SAPK/JNK, p-p44/42 MAPK (Erk1/2) and p-p38 protein levels, and this effect was attenuated in cells pre-treated with GZFLC (25 g⋅L⁻¹) as shown in Figure 4B, which suggested that GZFLC could downregulate the MAPK signaling pathway.

Integrating network target analysis and in vitro experiments, as shown in Figure 4C, GZFLC could inhibit the activity of COX2 and downregulate MAPK signaling pathway to reduce the production of PGs and myometrial contraction to exert analgesia effects for treating primary dysmenorrhea. In addition, GZFLC could regulate cytosolic calcium ion concentration, intervene oxytocin receptor (OTR) to reduce myometiral contraction and decrease inflammation, which was validated in literatures (Sun et al., 2016a; Zhong et al., 2016b) and subsequent in vivo experiments.

Primary dysmenorrhea is also related to imbalances in women’s endocrine system during the menstrual cycle (Britannica, The Editors of Encyclopedia, 2016). And a color Doppler study (Altunyurt et al., 2005) has found that there was increased impedance to blood flow within the uterus of patients with primary dysmenorrhea on the first day of the menstrual cycle, suggesting the disturbance of blood circulation in primary dysmenorrhea patients. Meanwhile, a previous study (Xiong et al., 2019) has found that there were disorders of energy metabolism in primary dysmenorrhea rats. Thus, Blood circulation, endocrine and energy metabolism may be three important functional modules to be regulated to gain sustained beneficial effects. According to the network target analysis, GZFLC was predicted to regulate blood circulation, endocrine and energy metabolism-related biological processes and signaling pathways, which may contribute to the sustained beneficial effects of GZFLC on primary dysmenorrhea. The effect of GZFLC on blood circulation, endocrine and energy metabolism have been verified in literatures (Su et al., 2015; Sun et al., 2016a; Xiong et al., 2019).

We have investigated the effect of GZFLC on the writhing response in oxytocin-induced dysmenorrhea murine model to study its analgesic activity. Analgesic activity was determined by observed decreases of writhing times in oxytocin-induced murine model, which was a method for quantitative evaluation of pain. As shown in Figure 5, GZFLC significantly inhibited the oxytocin-induced writhing response. The inhibition percentages of writhing times were 43.3% and 52.8% after oral administration of GZFLC (0.54 g/kg and 1.08 g/kg) respectively while 85.5% after oral administration of ibuprofen. The numbers of writhing times at GZFLC groups (0.54 g/kg and 1.08 g/kg) and ibuprofen group (0.1 g/kg) were 18.5 ± 4.5, 15.4 ± 4.6 and 5.9 ± 2.1 respectively, significantly lower than that of the model group (32.6 ± 8.5). Each herbal medicine in GZFLC was of the same amount and their dosages and ratios were the same as the traditional dosages and ratios in Guizhi Fuling Wan (Committee 2015). Next, we have investigated effect of GZFLC on uterine histopathology in oxytocin-induced dysmenorrhea murine model. As shown in Figures 6A,B, the histopathological results showed that compared with the normal group, the model group mainly showed degeneration and necrosis of endometrial epithelial cells, lamina propria edema, a small amount of inflammatory cell infiltration, reduced number of glands in the lamina propria, secretions in the gland cavity, and the disordered arrangement of smooth muscle cells in the muscle layer. There was a significant difference between the model group and the normal group, indicating that the oxytocin-induced dysmenorrhea murine model was successfully constructed. Meanwhile, ibuprofen and GZFLC had the effect on reducing the degree of uterine lesions as shown in Figures 6C–E. Moreover, there was a significant difference in 1.08 g/kg of GZFLC group compared with the model group (p < 0.05) as shown in Figure 6F.