DanQi Pill protects against heart failure through the arachidonic acid metabolism pathway by attenuating different cyclooxygenases and leukotrienes B4

We used DrugCIPHER-CS to predict the drug targets of the main compounds of DQP, following the procedure described in [17]. The herbal compounds were first identified in the Modern Chinese materia medica database [18]. The identified compounds were then typed into the durgbank (http://​www.​drugbank.​ca/​) to search their structural information including their canonical and isomeric SMILES. Then we applied DrugCIPHER-CS to predict the potential drug targets of these compounds. DrugCIPHER-CS achieved good prediction performance in our previous study and can infer drug-targets in the genome-wide scale [17]. This method is based on the hypotheses that i) drugs with similar chemical structure usually bind functionally related proteins, and ii) functional relationship between the proteins can be measured by their distance in the protein interaction network. Given a set of known drug (drug space)-target (target space) interactions, for a query drug and a candidate target-gene, drugCIPHER-CS will measure the likelihood of their interaction based on the correlation between the query drug’s structure similarity vector with the drug space and the candidate gene’s functional similarity vector with the target space. For a query compound, drugCIPHER-CS will prioritize the proteins in the protein interaction network (i.e. candidate proteins) according to the order of the decreasing drug-target interaction likelihood, and the candidate proteins with high likelihood will be hypothesized as the potential drug-targets [16].

Here known drug-target interactions are obtained from DrugBank database [19]. We only use those drug-target interactions whose drugs are FDA-approved and have InChI identifiers [20]. The chemical structure similarity is calculated based on compounds’ MOLPRINT 2D descriptors and Tanimoto coefficient [21].

After obtaining the potential drug targets, we analyzed the significantly enriched KEGG biological pathways and GO biological processes of these potential targets using the hypergeometric cumulative distribution test [22]. A pathway was significantly enriched with candidate target-genes when its corresponding upper-tailed P-value of hypergeometric cumulative distribution was smaller than 0.05. The pathways were ranked according to the order of the increasing P-values. Then, GO annotations of human proteins were obtained from the GO project Web site (http://​www.​geneontology.​org/​) [23]. The top 0.1% candidate target-genes were significantly enriched with genes annotated with a GO term when its corresponding upper-tailed P-value of hypergeometric cumulative distribution was smaller than 0.05. These GO terms were ranked according to the order of the increasing P-values. KEGG biological pathway data were downloaded from the KEGG database [24].

Studies were performed in accordance with the China Physiological Society’s “Guiding Principles in the Care and Use of Animals” and with approval of the Animal Care Committee of Beijing Medical Center (SCXK: 2006-0009). A total of 80 male Specific Pathogen Free (SPF)-grade Sprague–Dawley (SD) rats weighing 240 ± 10 g were selected (purchased from Beijing Vital River Laboratory Animal Technology Co. Ltd.). HF was induced by direct coronary ligation as previously described [25]. Briefly, sixty SD rats were anaesthetized with pentobarbital sodium (1%, 50 mg kg − 1 intraperitoneally). The left anterior descending coronary artery (LAD) was ligated with a 5–0 polypropylene suture (Surgipro, CT, USA) directly proximal to its main branching point. After electrocardiograph testing, rats that averaged QT-interval prolongation in three pre-cordial leads were included in the study. Sham-operated groups of the rest twenty rats were prepared following an identical procedure, but without the actual tying of the polypropylene suture. The overall mortality of rats that underwent induction of HF during the entire experimental period (up to 28 days after operation) was 30%. The majority of death occurred on the day of or the day after the surgery, probably because of acute pump failure or lethal arrhythmias. The operated rats were then randomly divided into two groups: 20 in the model group, 20 in the DQP group. Meanwhile, 20 in sham-operated group were investigated together. The rats were fed with a standard diet and water, and were maintained on a 12 h light and dark cycle. The DQP group was treated for 28 days, with the total daily dosage of 1.5 g/kg of the concentrated DQP (Tongren Tang, Beijing, China) dissolved in water. The sham and model groups received the same volume of water through oral gavage as the DQP vehicle. At the end of the study, all animals were anaesthetized using pentobarbital sodium following an overnight fast. Blood samples were collected through abdominal aorta puncture and were placing on ice. After centrifugation, plasma was collected, aliquoted, and stored at -80°C until analysis of each indicator. The heart was excised and incubated in ice-cold PBS to wash out blood. Each left ventricle was then carefully dissected to remove all the necrotic/scarred zones to keep only the viable myocardium in the marginal zone of the infarct region in model animals. The left ventricular myocardial below ligation bit in sham animals were also dissected.

Echocardiography was used to detect left ventricle (LV) end-systolic diameter (LVEDs), LV end-diastolic diameter (LVEDd), ejection fraction (EF) and fractional shortening (FS). A sector scanner (PST 65A, Toshiba Company, Japan), which generates two-dimensional images at a frame rate of 300 frames/s to 500 frames/s, was used. Fractional shortening (FS%) was calculated by the following equation: FS% = [( LVEDd- LVEDs)/LVEDd]*100%.

The DQP (Series: 6128006) used in this study was manufactured by Tongrentang company (Beijing, China) with the roots of 150 g S. miltiorrhiza and 150 g P. notoginseng. Briefly, the residue of P. notoginseng was mixed with all S. miltiorrhiza bunge, followed by extraction with hot water (twice at 2 h each). The water extract was then concentrated to form a paste, and ethanol was added. After 24 h, the filtrate, which was the final product, was collected. Based on the recommendation of daily human dosage (20 g/d) and the equivalent conversion between animal and people by body surface area and body weight, dosage of 1.5 g/kg was chosen in the present study.

Levels of plasma indicators were quantified in duplicate using commercial ELISA kits (Abcam Inc., Cambridge, MA, USA). Each assay was performed following the kits’ instructions. Standards at a series of concentrations were run parallel to the samples. The concentrations of the samples were calculated in reference to their corresponding standard curves and expressed as ng/mL.

The heart tissue was homogenized in RIPALYSIS buffer (i.e., 50 mM Tris-HCl pH7.4, 150 mM NaCl, 2 mM EDTA, 1% NP-40, and 0.1% SDS), and total protein was extracted from this homogenate. The protein concentration in each extract was measured using a protein assay kit (Pierce; Rockford, IL) and was then adjusted to the same value in all samples with 2× 4% Sodium dodecyl sulfate (SDS) sample buffer. The samples were boiled for 5 min, followed by loading on a 12.5% SDS-Polyacrylamide gel electrophoresis (PAGE) gel (30 mg protein/10 ml per well) for electrophoresis using a Bio-Rad mini gel apparatus at 100 V for 2 h. The fractionated protein on the gel was transferred onto an NC membrane (Millipore) and electrophoresed at 300 mA for 90 min. The membrane was first probed with COX1 primary antibody (anti-COX Type 1 antibody, ab18801, Abcam, 1:500) and secondary antibody (donkey polyclonal secondary antibody to rabbit IgG-HRP,ab97064, Abcam, 1:5000), and then treated with ECL (ECL plus Western blotting detection reagent, GE Healthcare) for 1 min at room temperature. The bands in the membrane were visualized and analyzed using UVP BioImaging Systems. After obtaining the COX1 blot density, the membrane was then treated using Restore Western Blot Stripping Buffer (Thermo Scientific) to remove the COX1signal, followed by probing with glyceraldehyde-3-phosphate dehydrogenase (GAPDH) primary antibodies (GAPDH mouse monoclonal IgG, ab8245, Abcam, 1:2000). The COX2, prostaglandin E2 receptor 4 (EP4), and leukotrienes B4 receptor (LTB4R) antibody blot densities were determined using the same process. The final reported data were the normalized COX1 band densities by GAPDH.

ANOVA using SAS 9.2 statistical software (SAS Institute, NC, USA) was applied to evaluate between-group differences in the outcome variables, follow-up least significant differences (LSD) analysis verified these differences were significant. P < 0.05 was considered statistically significant. Results were presented as mean values with their corresponding standard deviations.

Through a comprehensive search, we collected 92 compounds of DQP with 50 and 42 in Danshen and Sanqi respectively. The DrugCIPHER-CS method [17] was used to infer the potential targets of the compounds (see Methods). For each of the compound of DQP, DrugCIPHER-CS ranked its candidate targets according to the order of decreasing possibility of being targeted by the compound. We chose the top 0.1% candidate targets of each compound and consequently obtained 232 candidate target genes. These candidate targets were significantly enriched as known cardiovascular disease-related genes (i.e., the known targets of drugs whose ACT code uses “C” as the first level) in DrugBank [16], consistent with the curative effect of DQP on cardiovascular disease.

Furthermore, we analyzed the enriched KEGG biological pathways among these potential targets. Seven significantly enriched pathways among the top 0.1% candidate targets were analyzed, including the pathways of neuroactive ligand–receptor interaction, AA metabolism, calcium-signaling pathway, aminoacyl-tRNA biosynthesis and renin-angiotensin system (RAAS) et al. (Table 1). These significantly enriched biological pathways provided important information on the mechanism of DQP. Many of these enriched pathways have been associated with HF, e.g., the importance of neuroactive ligand–receptor interaction in the development and progress of cardiovascular diseases [26]. The key proteins in this pathway, such as adrenergic, angiotensin, and calcitonin receptor-like neurotensin receptors, are closely related to cardiac function [27‐29]. Aminoacyl-tRNA biosynthesis pathway is important in cardiovascular angiogenesis [30]. The relationship between calcium-signaling pathway and HF was also identified in the present study. Calcium antagonists have been widely used to inhibit extracellular calcium influx, reduce the concentration of intracellular calcium, and lower myocardial contractility [31]. Upregulation of the renin–angiotensin system is crucial in the deterioration of cardiovascular function [32]. However, the AA metabolism was rarely investigated both in HF and in DQP’s efficacy, it showed a significant difference as well as the pathway coverage in our study.

We also analyzed the functional distribution of these candidate targets (Table 2). The significantly enriched GO biological processes of these targets are cellular amino acid metabolic process, biosynthetic process, small molecule metabolic process, cellular nitrogen compound metabolic process, and circulatory system process, which suggest that the DQP may intervene in these progresses. These enriched pathways and GO functional annotations provided important clues to the molecular mechanism of DQP.

Twenty-eight days after surgery, the EF and FS values of rats in the model group were significantly lower than those of the sham group, as indicated by echocardiography (P < 0.05, Figure 1). EF values of ligation rats in model group dropped compared with those of the sham group, which suggested that a steady HF model was established. Twenty-eight days after treatment with DQP, EF values in the treatment group increased compared with those in the model group (Figure 1, Table 3).

The MASSON images of the left ventricular tissue were shown in Figure 2. Cardiomyocytes in the sham group were orderly arranged. Thickening and lengthening of myocardial fibers were observed in the model group. The nuclei were stained dark, which indicated local tissue fibrosis in the model group, while DQP had the inhibitory effect on ventricular hypertrophy.

Neurohormonal activation is important in the development of cardiovascular disease such as HF. However, the role of AA metabolism in HF development is still not clear [33]. Some studies show that inflammatory cells can release AA, which may be further metabolized into active products mediating cardiac fibrosis and HF. Some of the metabolic products, such as prostacyclin, are considered potential therapeutic targets of HF [34]. In the current study, the alteration of the AA metabolism pathway in HF was identified by DrugCIPHER-CS prediction and validated in animal models.

Western blot results showed that COX1 level in the model group (0.79 ± 0.096) increased 28 days after operation compared with that in the sham group (0.45 ± 0.01) (P = 0.031, Figure 3). After treatment with DQP, COX1 level in the DQP group (0.39 ± 0.08) decreased compared with that in the model group (P = 0.017). COX1 level in DQP-treated group showed no significant difference compared with that in the sham group (P = 0.896). COX2 level in the model group (1.213 ± 0.095) increased compared with that the sham group (0.409 ± 0.027) (P = 0.000). The level of COX2 in the DQP group (0.58 ± 0.057) decreased compared with that in the model group (P = 0.011) 28 days after treatment, recovering almost to sham group level.

Furthermore, LTB4 receptor (LTB4R) level in the model group (0.89 ± 0.113) was upregulated compared with that in the sham group (0.271 ± 0.098) (P = 0.000). Twenty-eight days after treatment with DQP, LTB4R level in the DQP-treated group (0.41 ± 0.106) decreased significantly compared with that in the model group (P =0.003, Figure 4).

Prostaglandin E2 (PGE2) and its receptors are considered better targets with protective effects in curing HF. Especially receptor 4 of PGE2, E-series prostanoid receptor 4 (EP4), is currently emerging as most versatile and promising among PGE2 receptors, and the PGE2/EP4 signalling is thought to be the potential targets to HF and myocardial ischemia [35]. In our study, EP4 of model group (0.14 ± 0.058) was down-regulated compared with that in the sham group (0.28 ± 0.052) (P = 0.023). In the DQP group, the EP4 level (0.26 ± 0.061) increased compared with that in the model group (P = 0.031), which had no significant difference with the sham group (Figure 4).

Prostacyclin I2 (PGI2) and thromboxane A2 (TXA2) are the products of AA metabolism through the COXs pathway. PGI2 has a cardioprotective effect by inhibiting platelet aggregation, whereas TXA2, which is quickly metabolized to TXB2, increases the risk of cardiovascular events [36]. In our study, ELISA results showed that levels of 6-Keto-PGF1α, which is the stable metabolite of PGI2, in the model group decreased compared with that in the sham group (P = 0.036). DQP upregulated 6-Keto-PGF1α levels in the model group almost back to normal conditions 28 days after administration. TXB2 is the product of TXA2 and can therefore reflect the level of TXA2 in plasma. The level of TXB2 increased in the model group. Although DQP showed no significant effect on TXB2, DQP increased the 6-Keto-PGF1α-to-TXB2 (P/T) ratio (Table 4).

There are some limitations in our study. For example, the search for DQP compounds has been done through the Chinese Materia Medica only, which is far from comprehensive. We presume all components of herbal formulation compounds are absorbed and utilized; improvement should be made in our future work.