Salvianolic Acid B and Ginsenoside Re Synergistically Protect Against Ox-LDL-Induced Endothelial Apoptosis Through the Antioxidative and Antiinflammatory Mechanisms

Salvianolic acid B (SalB) and ginsenoside Re (Re) protect endotheliocytes against apoptosis through different mechanisms. However, whether both compounds could synergistically protect endothelial cells against oxidized low-density lipoprotein (Ox-LDL)-induced apoptosis is unclear. This study aimed to assess the protective effect of combined SalB and Re (SR) treatment on Ox-LDL-induced endothelial apoptosis and to explore the mechanism underlying this effect. Results showed that SalB, Re, or SR could protect against Ox-LDL-induced endothelial apoptosis. Furthermore, the composition of SR was optimized through central composite design with response surface methodology. SR with a composition of 60 μg/mL of SalB and 120 μg/mL of Re exerted the optimal protective effect. Network pharmacology research revealed that SalB and Re in SR synergistically protect against Ox-LDL-induced endothelial apoptosis by regulating oxidative stress and phlogistic pathways. In vitro experiments confirmed these results. Compared with the same dose of SalB or Re alone, SR significantly decreased the contents of inflammatory mediators and increased the activities of antioxidant enzymes. SR could synergistically restore the balanced redox state of the cells and inhibit the activation of nuclear transcription factor kappa B and the caspase cascade by activating the phosphatidylinositol 3 kinase/protein kinase B pathway and inhibiting the phosphorylation of p38 mitogen-activated protein kinase. These pathways are regulated by down-regulating the expression of lectin-like Ox-LDL receptor-1 and NADPH oxidase and up-regulating the expression of estrogen receptor alpha. Therefore, SR effectively prevents Ox-LDL-induced endothelial apoptosis through antioxidative and antiinflammatory mechanisms.


INTRODUCTION
Atherosclerosis (AS), the main pathological basis of coronary artery disease and stroke, is a chronic inflammatory and multifactorial disease (Ross, 1999;Peters et al., 2012). The apoptosis of endothelial cells initiates and promotes the development of atherosclerosis (Asai et al., 2000;Chen et al., 2005;Hansson, 2005). Therefore, treatment with compounds that protect endothelial cells against apoptosis is an effective strategy for preventing and treating atherosclerosis. Among the many factors that induce endothelial apoptosis, oxidized lowdensity lipoprotein (Ox-LDL) is the most important (Martinet and Kockx, 2001;Chen et al., 2007). Ox-LDL first binds to lectinlike Ox-LDL receptor-1 (LOX-1) and then activates NADPH oxidase to produce numerous types of reactive oxygen species (ROS) that can initiate the caspase cascade (Chen et al., 2004); activate the p38 mitogen-activated protein kinase (p38-MAPK)mediated signaling pathway, or inhibit the phosphatidylinositol 3 kinase/protein kinase B (PI3K/Akt) pathway. These effects accelerate the activation of nuclear transcription factor kappa B (NF-κB) and up-regulate the expression of inflammatory mediators, thus further inducing endothelial apoptosis and apoptosis-related protein expression (Li and Mehta, 2009;Lee et al., 2010;Ahsan et al., 2015). Therefore, Ox-LDL-induced endothelial apoptosis should be effectively controlled through treatment with a combination of compounds with antioxidant and antiinflammatory activities.
The commercially available Guanxin Danshen formulation which is composed mainly of Salviae miltiorrhizae Bunge and Panax notoginseng (Burkill) F. H. Chen and is widely used to treat atherosclerotic cardiovascular diseases. The detailed mechanism underlying the pharmaceutical activity of S. miltiorrhizae Bunge and P. notoginseng (Burkill) F. H. Chen has never been thoroughly elucidated. Nevertheless, their components undoubtedly exert their curative effects through multiple compounds and targets (Han et al., 2008;Li et al., 2016). The main active ingredients formula in these two herbs possess the characteristic of formulae and simplify the research of complex formulae. Therefore, the active components formula is a requisite cut-point for developing and explicating the traditional chinese medicine (Li et al., 2010;Isgut et al., 2017). Salvianolic acid B (SalB) is the highest amount of the hydrophilic phenolic acid from S. miltiorrhizae Bunge, and ginsenoside Re (Re) is one of the major water-soluble triol ginsenoside derived from P. notoginseng (Burkill) F. H. Chen. SalB exhibits antiapoptotic, antioxidant, and antiinflammatory activities (Lee et al., 2014;Ren et al., 2016;Sun et al., 2016). Meanwhile, Re exerts pharmacological effects on the cardiovascular system (Peng et al., 2012). Furthermore, evidence suggests that combining SalB or Re with other active components promotes the pharmacological activities of these compounds (Li et al., 2008;Zhou et al., 2016). However, reports on the effect of SalB or Re on Ox-LDL-induced endothelial apoptosis remain limited, and the synergistic effects of SalB and Re against Ox-LDL-induced endothelial apoptosis remain unknown.
This study aimed to test and verify the effect of SalB or Re alone on Ox-LDL-induced endothelial apoptosis and to identify the combination of SalB and Re (SR) with the optimal protective effect through response surface methodology (RSM). The dose-effect curves of the combined drug treatments were assessed through the median-effect method. Moreover, the mechanism underlying the pharmaceutical action of SR was analyzed through network pharmacology and validated in vitro.

Reagent and Materials
SalB and Re (purity > 98%) were purchased from Shanghai Winherb Medical S&T Development (Shanghai, China). Dimethyl sulfoxide (DMSO) and collagenase I were acquired from Sigma-Aldrich (St. Louis, MO, United States). All cellculture materials were obtained from Lifeline (Randallstown, MD, United States

Preparation and Cultivation of Cells
According to previously described (Sun et al., 2013), HUVECs were isolated from fresh human umbilical veins by using 0.1% collagenase I. After 13 min enzymatic dissociation, HUVECs were collected by centrifugation at 3600 rpm for 15 min and cultured on 1% gelatin-coated plastic dishes by VascuLife R Medium (CID: LL0003) supplemented with 1% penicillin/streptomycin (v/v). The cells were cultured in a humidified incubator with 95% air and 5% CO 2 at 37 • C. Cells from passages 3-7 were used for subsequent experiments, and media were refreshed every 2 days. All experiments using HUVEC cells were approved by the Ethics Committee of Peking Union Medical College (SYXK-2013-0023, Beijing, China) and was administrated in accordance with the Declaration of Helsinki. The neonate cords were donated by the Maternal and Child Care Service Centre in Beijing, China. The study protocol was explained and all participating donors were given written informed consents.

Assay for Cell Viability
To detect the protective effects of SalB or Re dissolved in DMSO on HUVECs, HUVECs were cultured in 96-well plates at a density of 10 4 cells/well and grown for 24 h. Then, the HUVECs were pretreated with a concentration gradient of SalB (3.125,6.25,12.5,25,50,100 µg/mL) or Re (6.25,12.5,25,50,100,200 µg/mL) for 12 h in serum-free medium. Next, the HUVECs were incubated with 100 µg/mL of Ox-LDL for 24 h which did not have SalB or Re. Subsequently, 5 mg/mL of MTT in fresh medium was added to each well, and the HUVECs were incubated for an additional 4 h. Finally, 150 µL DMSO was added to dissolve the formazan crystals after removing the supernatant. Absorbance was measured by a plate reader (Infinite M1000, Tecan, Sunrise, Austria) at 570 nm.

The 2 k Factor Design Experiment
The two factors were investigated: SalB; Re. The influence of variations in these factors on the cell viability was evaluated using a 2 k factorial design (Table 1). A total of four experimental runs were conducted (randomly) for which three replicates were performed. According to the result of the single-factor experiment, the concentration range of SalB is 50 -100 µg/mL and Re is 70 -140 µg/mL. This gave the same conditions in each experiment. The outputs of the experimental design were analyzed using Design Expert 8.0.6 software (Stat-Ease, Inc.). The magnitude and direction of factor effect were determined via OD calculations and the significance of the factor effect was evaluated using ANOVA ( Table 2).

Steepest Ascent Test
In order to determine the suitable ranges which tend to approach the optimal condition, the steepest ascent test was designed based on significant factors. The path began at the center point of the 2 k factorial design, serving as the origin for the steepest ascent experiment, that's mean that the initial concentration of SalB is 75 µg/mL and Re is 105 µg/mL. The direction of steepest ascent was decided by the first-order mathematical linear equation predicted from the 2 k factorial design, so the direction of SalB was -4 and the Re was +5 ( Table 2). Four runs were evaluated over one plate with replicated three, the same conditions were given in each experiment. Finally, the cell viability was evaluated. Statistical analysis and model prediction were performed using Design Expert 8.0.6 software (Stat-Ease, Inc.).

Response Surface Methodology
Response surface methodology with central composite design (CCD) was performed to optimize the combination of SalB and Re (SR). A two-factor, five-level center CCD was finished. CCD was created by Design Expert 8.0.6 software (Stat-Ease, Inc.), and thirteen combinations, including six replicates at central point was chosen randomly according to CCD which are listed in Table 3. Based on the result of the steepest ascent test, the initial concentration of SalB is 63 µg/mL and Re is 120 µg/mL. The data analysis, model construction, and 3D response surface graph generation were performed using Design Expert 8.0.6 software (Stat-Ease, Inc.).

Analysis of the Median-Effect Equation
To further research SalB-Re interactions in SR, a range of SR was analyzed through the median-effect method. According to the previous description (Chou, 2006), experiment scheme of SR was designed by CalcuSyn 2.1 (Cambridge, United Kingdom), and six combinations listed in Table 4 were evaluated over one plate with replicated three, the same conditions were given in each experiment. Finally, SalB-Re interactions in SR were assessed on the basis of the CI value derived by CalcuSyn 2.1  (Cambridge, United Kingdom) and were defined as follows: CI > 1 antagonistic, CI = 1 additive, and CI < 1 synergistic.

Validation With Cell Models
To further verify the protective effects of SalB, Re, or SR, other two models were tested using H9c2 cells obtained from the Chinese Academy of Sciences Cell Bank (Shanghai, China). The first model was H 2 O 2 -induced apoptosis of H9c2 cardiomyoblasts (Sun et al., 2011), in a nutshell, H9c2 cells were cultured on high glucose DMEM supplemented with supplemented with 1% penicillin/streptomycin (v/v). The cells were cultured in a humidified incubator with 95% air and 5% CO 2 at 37 • C for 24 h. When cells reached 70 -80% confluence, cells were incubated with 60 µg/mL of SalB, 120 µg/mL of Re and SR (60 µg/mL -120 µg/mL) for 12 h, and then removed the supernatant. Next, these cells were incubated with 150 mM H 2 O 2 for 6 h. At last cell viability was measured by MTT. The second model was hypoxia-reoxygenation-induced apoptosis of H9c2 (Liao et al., 2016). Briefly, the cells were cultured on high glucose DMEM supplemented with supplemented with 1% penicillin/streptomycin (v/v). The cells were cultured in a humidified incubator with 95% air and 5% CO 2 at 37 • C for 24 h. When cells reached 70 -80% confluence, cells were incubated with 60 µg/mL of SalB, 120 µg/mL of Re and SR (60 µg/mL -120 µg/mL) for 12 h, and then removed the supernatant. Next, these cells were cultured in 37 • C for 6 h in an anaerobic glove box, where normal air was removed by a combination of 5% CO 2 , 5% H 2 and 90% N 2 , and then they were removed to the regular incubator for 12 h with the medium replaced by high glucose medium to mimic reperfusion. At last cell viability was measured by MTT. In addition, all experiments were accomplished on the HUVECs except for the part of validation with cell models, which used H9c2 cells to prove the effect of SR.

Construction of Pathways or Disease Networks Targeted by SalB and Re
First, information on SalB and Re targets were first collected from the literature (Web of Science, Pubmed, ScienceDirect, SpringerLink, CNKI). The cut-off date for entries is June 17, 2016. Second, changes in gene expression following the Ox-LDL activation of LOX-1 in endothelial cells was studied by analyzing GSE13139. Third, the Ox-LDL-induced endothelial apoptosis database was constructed with a protein chip (Qin et al., 2015). Furthermore, atherosclerosis-related genes were retrieved from GeneCards 1 . Finally, molecular docking was performed by idTarget 2 and PharmMapper 3 . The framework for the target genes, pathways, or disease networks of SalB and Re was thus generated.
Gene expression profiles were analyzed using the bioconductor packet. DAVID 6.7 was used for GO enrichment analysis 4 . KEGG pathways were analyzed using Kobas 2.0. 5 A network was constructed to better understand the complex relationships between SalB and Re, their target pathways and associated diseases. The network was structured and analyzed using Cytoscape 3.4.0. The interactions of target genes were analyzed using STRING 6 .

Flow Cytometric Detection of Apoptosis
Early apoptosis and necrosis were measured using the Annexin V-FITC Apoptosis Detection Kit. Briefly, HUVECs were treated with or without the concentrations of 60 µg/mL of SalB, 120 µg/mL of Re and SR (60 µg/mL -120 µg/mL) for 12 h, followed by 100 µg/mL Ox-LDL for another 24, and then the treated cells were harvested and washed twice with cold PBS. Subsequently, 10 6 cells were incubated with 10 µL of Annexin-V-FITC and 1 µL of PI working solution (100 µg/mL) in the dark for 15 min at 37 • C. Finally, cell culture plates were filled with 500 µL of binding buffer. Cellular fluorescence was measured through flow cytometry (BD, San Jose, CA, United States).

Measurement of Mitochondrial Membrane Potential
The change in mitochondrial membrane potential was detected through JC-1 staining. HUVECs were pre-incubated with or without the concentrations of 60 µg/mL of SalB, 120 µg/mL of Re and SR (60 µg/mL -120 µg/mL) for 12 h, followed by 100 µg/mL Ox-LDL for another 24, and then the treated cells were incubated with JC-1 (10 µM final concentration) at 37 • C in the dark for 30 min. Next, the treated cells were washed three times with warm PBS, harvested and dissolved in 500 µL PBS. Finally, the intensity of green fluorescence was determined at an excitation wavelength of 485 nm and an emission wavelength of 535 nm. The intensity of red fluorescence was determined at an excitation wavelength of 585 nm and an emission wavelength of 590 nm. The fluorescence intensity of the cell was analyzed by using a microplate reader (Infinite M1000, Tecan).

Measurement of LDH Levels and CAT, GSH-Px, and SOD Activities
HUVECs were cultured at 10 4 cells/well in 6 -well plates for 24 h. The cells were pre-incubated with or without the concentrations of 60 µg/mL of SalB, 120 µg/mL of Re and SR (60 µg/mL -120 µg/mL) for 12 h, followed by 100 µg/mL Ox-LDL for another 24, and then the supernatant was collected for the following experiments. The activities of GSH-Px, SOD, and CAT, and the release of LDH were quantified using commercial kits in accordance with the manufacturer's instructions.

Analysis of ROS Activity
HUVECs were cultured at 10 4 cells/well in 6 -well plates for 24 h. The cells were pre-incubated with or without the concentrations of 60 µg/mL of SalB, 120 µg/mL of Re and SR (60 µg/mL -120 µg/mL) for 12 h, followed by 100 µg/mL Ox-LDL for another 24, and then the supernatant was removed. After treatment, cells were collected and incubated with carboxy-H2DCFDA (25 µM) in the dark for 30 min at 37 • C, and then washed thrice with PBS, and quantitatively analyzed through flow cytometry (BD, San Jose, CA, United States).

Determination of Inflammatory Mediator Contents
HUVECs were cultured at 10 4 cells/well in 6 -well plates for 24 h. The cells were pre-incubated with or without the concentrations of 60 µg/mL of SalB, 120 µg/mL of Re and SR (60 µg/mL -120 µg/mL) for 12 h, followed by 100 µg/mL ox-LDL for another 24, and then the supernatant was collected for the following experiments. Standard ELISA kits were used to measure IL-6, TNF-α, and MCP-1 contents in accordance with the manufacturer's instructions.

Analysis of Caspase-3 Activation
Caspase-3 activation was measured using a fluorescein-active caspase-3 staining kit in accordance with the manufacturer's instructions. Briefly, HUVECs were cultured at 10 4 cells/well in 6 -well plates for 24 h. The cells were pre-incubated with or without the concentrations of 60 µg/mL of SalB, 120 µg/mL of Re and SR (60 µg/mL -120 µg/mL) for 12 h, followed by 100 µg/mL Ox-LDL for another 24, and the supernatant was removed. Then, 10 6 cells were harvested and incubated with 1 ml of the substrate FITC-DEVD-FMK in the dark for 1 h at 37 • C. Subsequently, the cells were centrifuged for 5 min at 3,000 rpm, and the supernatant was removed. Finally, the cells were washed twice with PBS and resuspended in 300 ml of wash buffer on ice. The samples were tested through flow cytometry (BD, San Jose, CA, United States).

Statistical Analysis
All experiments were repeated at least thrice. All quantitative data were presented as mean ± standard deviation and analyzed through Student's t-test or ANOVA by SPSS (version 20.0, Chicago, IL, United States). P < 0.05 was considered significant.

Effect of Single Factors on the Protective Effects of SR Against Ox-LDL-Induced Endothelial Injury
We assessed the protective effects of SalB or Re on endothelial cells by performing 3-(4,5-dimethylthiazol-2yl)-2,5-diphenyltetrazolium bromide (MTT) assay. We found that treatment for different durations with different doses of SalB or Re alone could attenuate Ox-LDL-induced endothelial injury (Supplementary Figures 1, 2). Furthermore, we observed the ideal protective against Ox-LDL-induced endothelial injury when we treated cells with 50 µg/mL of SalB and 100 µg/mL of Re. We identified the optimal concentration of SalB or Re by refining the concentration range of these components on the basis of the above conclusions. As shown in Figure 1A, 12 h of treatment with 59.05-90 µg/mL SalB or 102.4-160 µg/mL Re provided the optimal protective effect against Ox-LDL-induced endothelial injury ( Figure 1B): 81 µg/mL SalB could make the survival rate of HUVECs reach 75.33% and 128 µg/mL Re was 83.25%. Therefore, we used 50-100 µg/mL of SalB and 70-140 µg/mL of Re in further experiments.

Formulation Building and Optimization by RSM
After the 2 k factor design experiment and the steepest ascent test, we applied RSM with CCD to identify the optimal formulation of SR. The 2 k factor design experiment showed that the optimization direction and step length of SalB are −4 and those of Re is 5, and the initial SR formulation is 75 -105 µg/mL (SalB -Re, SR) (Supplementary Material 2). The result of the steepest ascent test indicated that the ideal dosage of SR is 63-120 µg/mL (Supplementary Material 3). We then used CCD to optimize SR formulation, as shown in Supplementary Material 4. Thirteen sets of experimental data were analyzed by a quadratic model and subjected to analysis of variance (ANOVA). Subjecting the quadratic model to ANOVA revealed that the quadratic regression equation fitted the response surface shown in Supplementary Material 4 (R 2 = 0.9481, adjusted R 2 = 0.9111). The response surface plot of cell viability versus SalB and Re concentration is shown in Figures 1C,D, where SR has an optimal formulation of 57.8 µg/mL SalB and 125.2 µg/mL Re. Thus, we used the combination of 60 µg/mL of SalB and 120 µg/mL of Re in the follow-up study.
We performed the median-effect method to explore the interaction of SalB and Re. The value of the combination index (CI) is between 0 and 1, indicating that SalB and Re act cooperatively (Figures 1E,F). Meanwhile, statistical analysis of the quadratic regression equation also implied that the components of SalB and Re were significant interactions (Supplementary Material 4).

Validation of the Protective Effect of SR
To test the reliability of the optimization method, we used three different cell models to confirm that SR exerts protective effects. As shown in Figure 2, the protective effect of 120 µg/mL of Re on Ox-LDL-induced endothelial injury and H/R-induced H9c2 injury is stronger than that of 60 µg/mL of SalB (Figures 2A,B). However, the effect of SalB is stronger than that of Re in the H 2 O 2 -induced H9c2 injury model ( Figure 2C). In general,

Construction and Analysis of Networks
To elucidate the combinatorial principle and the molecular mechanism underlying the synergetic protective effects SalB and Re against Ox-LDL-induced endothelial injury (Figure 3), we first designed an integrated strategy for exploring potential targets on the basis of the network pharmacology framework. We identified 111 candidate targets of SalB, of which 74 are specific targets of SalB and 37 are common targets of SalB and Re. Given that Re has 32 specific targets, we obtained 69 candidate targets of Re.
To further investigate the potential pharmacological effects of SR against AS, we constructed compound-target pathways and disease networks on the basis of component-target interaction. As shown in Figure 4, the 37 common targets of SalB and Re mainly regulate the signaling pathways of oxidative stress, inflammation, cell proliferation, and glycolipid metabolism. The analysis of compound-target-disease networks indicated that the physiological effects of these 37 targets are mainly related to CVD or glycolipid metabolism diseases, such as metabolic disease and diabetes type II.
We found that the specific targets of SalB mainly regulate the signal pathways of apoptosis, inflammation, and energy metabolism, whereas those of Re mainly participate in the pathways of lipid metabolism and blood coagulation. Furthermore, the signal pathways of insulin resistance and leukocyte transendothelial migration are regulated by the specific targets of SalB and Re (Figure 5). Next, the analysis of compound-target-disease networks revealed that the specific targets of SalB are related to cardiovascular diseases, particularly to atherosclerosis, whereas those of Re mainly regulate inflammatory bowel diseases. Most interestingly, the specific targets of SalB and Re can regulate immune diseases ( Figure 5).
To explore the key genes regulated by SalB and Re, we analyzed gene interactions through STRING. The key genes of the common targets of SalB and Re are SOD1, SOD2, SOD3, CAT, GPX, TLR4, SRC, and MAPKS (Supplementary Figure 3). The key genes of the specific targets of SalB are PI3K, MAPKs, MMPs, CD40, RHOA, FOS, and END1 (Supplementary Figure 4). The key genes of the specific targets of Re are ESR1, ESR2, and PPARD (Supplementary Figure 5).

Effects of SR on the Apoptosis of Endothelial Cells
First, we validated the effects of SR on Ox-LDL-induced endothelial apoptosis. We examined the apoptosis rates of endothelial cells by using Annexin V-FITC/PI double dyes. As shown in Figures 6A,B, the apoptosis rates of SalB-, Re-, or SR-treated human umbilical vein endothelial cells (HUVECs) markedly decreased relative to those of Ox-LDLtreated HUVECs. The protective effect of SR on Ox-LDLinduced apoptosis in HUVECs is the strongest among that of other monotherapies, and the viability of cells in the Re treatment group is stronger than that in the SalB treatment group (Figures 6A,B).
To test the antiapoptotic effect of SR, we measured the effect of SR on mitochondrial permeability, a key indicator of early apoptosis. The SR-treated group exhibited the highest red/green fluorescence intensity ratio. The red/green fluorescence intensity ratio of the Re group is lower than that of the SR group, and that of the SalB is the lowest (Figure 6C). Therefore, the intensity of the antiapoptotic effect exerted by the compounds follows the order of SR > Re > SalB.

Effects of SR on the ROS Generation and Antioxidant Enzyme Activity
Then, we validated the effects of SR on ROS generation and antioxidant enzyme activity in accordance with the results of network pharmacology. As shown in Figures 7A,B, incubating HUVECs with SalB, Re, or SR for 12 h prior  to Ox-LDL exposure markedly reduced ROS generation by these cells. The inhibitory effect of SR is the strongest, and that of SalB is stronger than that of Re. Moreover, analyzing antioxidant enzyme activity revealed that the levels of superoxide dismutase (SOD), glutathione peroxidase (GSH-Px), and catalase (CAT) in the treated groups are significantly higher than those in the Ox-LDL group (Figures 7C-E). The ability of the components to regulate antioxidant enzyme activity is consistent with their ability to inhibit ROS production.

Effects of SR on the Secretion of Adhesion Molecules and Proinflammatory Mediators
The results of network pharmacology showed that SR can regulate the secretion of inflammatory factors, cytokines, and adhesion molecules. We further validated the effects of SR on the activity or expression of related factors by using enzymelinked immunosorbent assay kits (ELISA) or by performing Western blot analysis. As shown in Figures 8A-C, the activities of interleukin (IL)-6, monocyte chemotactic protein (MCP)-1, and tumor necrosis factor (TNF)-α are significantly up-regulated in the Ox-LDL group relative to those in the control group. However, compared with Ox-LDL treatment, SalB, Re, or SR treatment down-regulated IL-6, TNF-α, and MCP-1 expression in HUVECs undergoing Ox-LDL-induced apoptosis. SR more effectively decreased IL-6, TNF-α, and MCP-1 activities than SalB and Re, and SalB is more effective than Re.
Notably, under SalB, Re, or SR treatment, the expression patterns of ICAM-1 and VCAM-1 were different from those of IL-6, TNF-α, and MCP-1. The expression levels of ICAM-1 and VCAM-1 were more effectively inhibited by SR than by SalB and were unaffected by Re (Figures 8D-F).

Mediatory Effects of SR on the Expression of LOX-1, NOX4, and ERα
To study the protective mechanism of SR, we first explored the expression patterns of LOX-1, NOX4, and ERα through gene interaction analysis. As shown in Figures 9A,B, Ox-LDL treatment up-regulated the protein levels of LOX-1 and NOX4. As expected, pretreatment with SalB, Re, or SR could reverse this phenomenon, and SR is more effectively reversed the expression of these proteins than SalB or Re. SalB more effectively regulated NOX4 expression than Re. However, the regulatory effects of SalB and Re on LOX-1 expression did not significantly differ. In addition, SalB and Re activated ERα expression ( Figure 9C). The effect of SR is strongest among those of the three treatments, and the effect of Re is stronger than that of SalB.

SR Inhibits the Ox-LDL-Induced Apoptosis of HUVECs via p38MAPK and PI3K/Akt
We continued to study the downstream signaling pathways of the key gene targets of SalB and Re by analyzing receptor protein and gene expression. We found that Ox-LDL treatment could remarkably up-regulate the phosphorylation of p38MAPK compared with the control treatment. Pretreatment with SalB, Re, or SR could significantly down-regulate the Ox-LDL-induced phosphorylation of p38MAPK. SR exhibited the most intense inhibitory action among all treatments. The inhibitory action of SalB is stronger than that of the Re groups ( Figure 10A). The addition of a p38MAPK agonist markedly reduced the protective effect of SR against Ox-LDL-induced apoptosis in HUVECs ( Figure 10A). However, FIGURE 6 | Protective effects of SalB, Re, and SR on HUVECs undergoing Ox-LDL-induced early apoptosis. (A) Annexin V/PI staining was used to detect the protective effects of SalB, Re, and SR on HUVECs undergoing Ox-LDL-induced apoptosis. (B) Quantitative analysis of the apoptosis rates of HUVECs. (C) JC-1 staining was used to detect mitochondrial membrane potential changes in SalB-, Re-, or SR-treated HUVECs during Ox-LDL-induced apoptosis. * * * P < 0.001, * P < 0.05 vs. Ox-LDL; ### P < 0.001 vs. Control; $$ P < 0.01, $ P < 0.05 vs. SalB; &&& P < 0.001, & P < 0.05 vs. SR.
compared with the phosphorylation of p38MAPK, the effect of SalB, Re, or SR on PI3K/Akt was in the opposite direction ( Figure 10B). Compared with the control treatment, Ox-LDL treatment could markedly down-regulate the phosphorylation level of Akt (S473). However, preincubation with SalB, Re, or SR significantly increased the phosphorylation level of Akt compared with preincubation with Ox-LDL. SR exhibited the strongest activating effect among all three treatments, and Re exhibited a stronger effect than SalB. To determine the role of the SR in PI3K/Akt activation, we examined the effects of LY294002 (12.5 nM) treatment on the activation of PI3K/Akt pathways and found that SR treatment continued to protect HUVECs against Ox-LDL-induced apoptosis ( Figure 10B). These results indicated that SR protected against Ox-LDL-induced apoptosis in HUVECs by inhibiting p38MAPK and activating PI3K/Akt.

Inhibitory Effects of SR on the Nuclear Translocation of NF-κB
Given that nuclear transcription factor kappa B (NF-κB) is a key downstream factor of p38MAPK and PI3K/Akt, we performed Western blot analysis to study the nuclear translocation of NF-κB. As shown in Figure 11, Ox-LDL treatment could markedly activate the nuclear translocation of NF-κB from the cytoplasm to the nucleus relative. However, SalB, Re, or SR pretreatment could significantly reverse this effect. The inhibitory action of SR is the strongest among those of all three treatments, and that of Re is stronger than that of SalB. However, that addition (C-E) Commercial kits were used to determine the effects of SalB, Re, and SR on the levels of SOD, GSH-Px, and CAT activities in HUVECs undergoing Ox-LDL-induced apoptosis. ### P < 0.001 vs. Control; * * * P < 0.001, * * P < 0.01, * P < 0.05 vs. Ox-LDL; $$$ P < 0.001, $$ P < 0.01, $ P < 0.05 vs. SalB; &&& P < 0.001, && P < 0.01, & P < 0.05 vs. SR.
of anisomycin (an activator of p38MAPK, 1 nM) or LY294002 (an inhibitor of PI3K, 12.5 nM) weakened the inhibitory action of SR. These results indicated that SR inhibits the nuclear translocation of NF-κB through the p38MAPK and PI3K/Akt pathways.

SR Restores the Balanced Expression of Proapoptotic and Antiapoptotic Proteins
Given the important role of NF-κB in modulating cell survival, we investigated the antiapoptosis effect of SR. Compared with the control treatment, Ox-LDL treatment up-regulated the expression of the proapoptotic proteins Bax and Smac and simultaneously down-regulated the expression of the antiapoptotic proteins Bcl2 and cIAP-2. Preincubation with SalB, Re, or SR reversed these effects. SR exhibited the strongest regulatory activity among all three treatments, and Re exhibited stronger regulatory activity than SalB. However, treatment with anisomycin or LY294002 significantly weakened the regulatory ability of SR (Figures 12A-D).
We used a flow cytometer to test the activity of cleaved caspase-3 ( Figure 12E). We found that the activity of cleaved caspase-3 increased under Ox-LDL treatment relative to that under the control treatment. This effect, however, was reversed through incubation with SalB, Re, or SR. The regulatory effects of SalB, Re, and SR on the activity of cleaved caspase-3 are similar to their regulatory effects on the expression of proapoptotic proteins. These results indicated that SR protected HUVECs against Ox-LDL-induced apoptosis through the p38MAPK/NF-κB and PI3K/Akt/NF-κB pathways.

DISCUSSION
Atherosclerosis is a chronic inflammatory and multifactorial disease. Its treatment through fixed-dose combination therapy has received increased attention in Western medicine (Anderson et al., 2017;Bahiru et al., 2017). For 100s of years, TCM formulations, such as Guanxin Danshen formula, have been used to treat atherosclerosis; these formulations are prepared by combining medicinal herbs in accordance with the rule of Jun-Chen-Zuo-Shi . These formulations exert antiatherogenic effects through the synergistic effects of multiple components on multiple targets. However, understanding and elaborating the mechanism underlying the antiatherogenic effects of TCM formulations remain difficult. Studies on TCM conducted during the past decade have focused on the isolation, purification, structural identification, and pharmacological activity of TCM formulations. However, the simple quantitative analysis of one or several active components of TCM formulations is not in line with the principle of TCM treatment. Thus, a systematic method that could not only validate the efficacy of TCM formulations but also reflect the principle of FIGURE 8 | SalB, Re, and SR inhibit the Ox-LDL-induced secretion of inflammatory cytokines in HUVECs. (A-C) ELISA kits were used to determine the protective effects of SalB, Re, and SR on the secretion of IL-6, TNF-α, and MCP-1 in HUVECs undergoing Ox-LDL-induced apoptosis. (D-F) Pretreatment with anisomycin or LY294002 attenuated the protective effects of SalB, Re, and SR on the expression levels of ICAM-1 and VCAM-1 in HUVECs undergoing Ox-LDL-induced apoptosis. ### P < 0.001 vs. Control; * * * P < 0.001, * * P < 0.01, * P < 0.05 vs. Ox-LDL; $$$ P < 0.001, $$ P < 0.01, $ P < 0.05 vs. SalB; &&& P < 0.001, && P < 0.01, & P < 0.05 vs. SR. TCM treatment should be established. In this study, we first used HUVECs to construct an Ox-LDL-induced apoptosis model, which reflects the pathophysiology of atherosclerosis (Orekhov and Ivanova, 2016). Then, we used this model to screen for antiapoptotic components from S. miltiorrhizae and Notoginseng Radix, the two pharmaceutically active herbs in the Guanxin Danshen formulation. Next, we optimized the combination of the active ingredients of SalB and Re through RSM and used the value of the CI to explore interactions among the identified active components. Furthermore, we analyzed the antiapoptotic mechanism of the compounds through network pharmacology and utilized an in vitro model to validate our conclusion.
Guanxin Danshen formula is a TCM preparation used to treat cardiovascular diseases. It is included in the 1995 edition of the Chinese Pharmacopeia. The main therapeutic components of this formula is from S. miltiorrhizae and Notoginseng Radix; these components can be used to treat atherosclerosis (Sedighi et al., 2017), but it is not clearly what kind of the active substances produce therapeutic effects on anti-atherosclerosis? How make up a prescription can exert most effective protection Control; * * * P < 0.001, * * P < 0.01, * P < 0.05 vs. Ox-LDL; $ P < 0.05 vs. SalB; &&& P < 0.001, && P < 0.01 vs. SR. SalB and 102.4-160 µg/mL Re provide optimal protective effects against apoptosis (Huang et al., 2016;Chen et al., 2017). However, previous research has focused solely on the simple quantitative analysis of one component to identify the fundamental mechanisms of the pharmaceutical effects of Guanxin Danshen formula. Therefore, we further adopted RSM, which has a rich history in non-linear combinations (Fang et al., 2008;Hather et al., 2013), to systematically study the combination of the active components of SalB and Re. The generated threedimensional response surfaces and two-dimensional contour plots indicated that SalB and Re mutually interact and exert a quadratic effect on responses (Minto et al., 2000). The generated response surface suggested that SalB and Re in SR synergistically protect HUVECs from Ox-LDL-induced apoptosis. This finding is consistent with the ANOVA of the fitted quadratic model and was validated on the basis of the CI value (Chou, 2006). Furthermore, the antiapoptic effect of SR is stronger than that of SalB or Re alone, as validated in different cell models. Moreover, the synergistic effects of SalB and Re are supported by the combination of S. miltiorrhiza and Notoginseng Radix . RSM results and the CI value indicated that the effect of SR is more intense than that of SalB or Re alone. Nevertheless, its therapeutic mechanism remained unclear. Therefore, we performed network pharmacology analysis to investigate the molecular mechanisms of action of SR (Hopkins, 2008). We found that SalB and Re have 111 and 69 target candidate genes, respectively, in the endotheliocyte model of Ox-LDL-induced apoptosis. Interestingly, SalB and Re share 37 common targets. Target enrichment analysis and gene-gene interaction networks showed that the main signal pathways of oxidative stress and inflammation are regulated by the shared target genes of SalB and Re. These genes include GPX1, GPX4, GPX3, SOD1, SOD3, SOD2, and MAPKs and are implicated in the Ox-LDL-induced apoptosis of HUVECs (Li and Mehta, 2009;Lee et al., 2010;Ahsan et al., 2015). Through the analysis of target pathways and disease networks, we found that SalB or Re mediates the same signaling pathways or the same disease network through their specific targets. This finding again proved that SalB and Re in SR synergistically act to protect HUVECs against Ox-LDL-induced apoptosis. The specific target genes of SalB mainly regulate the signal pathways of oxidative stress and inflammation through PI3K, MAPKs, and MMPs. PI3K is the key gene of the PI3K/Akt pathway, and p38MAPK is a type of MAPK. The PI3K/Akt pathway could regulate apoptosis and antioxidative responses Yang et al., 2017). The p38MAPK pathway can adjust the secretion of inflammatory mediators (Li and Mehta, 2009), such as MMPs, ICAM-1, VACM-1, and MCP-1. Moreover, bioinformatics analysis revealed that the specific target genes of Re, such as AR, ESRs, and PPAR, mainly regulate hormone-related pathways. Re is a phytoestrogen, and evidence indicates that phytoestrogens exert cardiovascular protective and antiapoptotic effects through hormone receptors (Cos et al., 2003;Gencel et al., 2012), such as AR and ESRs FIGURE 13 | Molecular mechanism underlying the antiapoptotic effect of SR. T-bars represent inhibition, and arrows represent activation. (Chakrabarti et al., 2014), which was in good agreement with the network analysis of Re.
Network pharmacology analysis revealed that SR protects endotheliocytes against Ox-LDL-induced apoptosis by regulating the signaling pathways of oxidative stress and inflammation. We proved that SR can markedly down-regulate the expression levels of LOX-1 and NOX4; up-regulate the expression of ERα; reduce the generation of ROS; promote the activities of the antioxidant enzymes SOD, GSH-Px, and CAT; inhibit the expression levels of ICAM-1 and VCAM-1; decrease the expression levels of IL-6, TNF-α, and MCP-1; inhibit the phosphorylation of p38MAPK and the activation of the PI3K/Akt pathway; accelerate the activation of NF-κB; decrease the expression levels of apoptosis proteins and inflammatory mediators; increase the expression levels of antiapoptotic proteins; and finally prevent the Ox-LDLinduced apoptosis of HUVECs. These results are consistent with the network pharmacology analysis and correspond with the pathological changes associated with atherosclerosis that Ox-LDL-induced endotheliocyte apoptosis is the initial factor of atherosclerosis (Asai et al., 2000;Chen et al., 2005;Hansson, 2005), and oxidative stress and inflammation are two critical factors in the Ox-LDL-induced apoptosis of HUVECs (Li and Mehta, 2009;Lee et al., 2010;Ahsan et al., 2015). Pretreatment with anisomycin or LY294002 weakened the protective effect of SR against the Ox-LDL-induced apoptosis of HUVECs. Data indicated that SR protects HUVECs against Ox-LDL-induced apoptosis through the PI3K/Akt and p-p38MAPK/p38MAPK pathways. Nevertheless, this study also has disadvantages. For example, the effect of SalB on the expression of LOX-1 and ERα should be further investigated. Similarly, the effect of Re on the expression of LOX-1 and NOX4 should be tested further, and the protective effect of SR should be validated in vivo.

CONCLUSION
To summarize (Figure 13), we demonstrated that SalB and Re in SR could synergistically protect HUVECs against Ox-LDL-induced apoptosis. SalB and Re in SR exert their synergistic antiapoptotic effects by up-regulating the activities of antioxidant enzymes, inhibiting the secretion of inflammatory factors, and restoring the balanced expression of proapoptotic and antiapoptotic proteins through the p38MAPK/NF-κB and PI3K/Akt/NF-κB pathways, which are mediated by the receptors of LOX-1, NOX4, and ERα. Our findings indicated that SR may be an effective treatment against endothelial cell apoptosis.

AUTHOR CONTRIBUTIONS
KY performed the major experiments and wrote this paper. YL checked the results of the experiments. SL revised the manuscript. RH was responsible for network pharmacology. YD and PL were in charge of screening active compounds. GS and XS answered for designing experiments.