2.1. Database Construction

Candidate bioactive ingredients of 12 herbs in WZESD were screened from the Traditional Chinese Medicine Systems Pharmacology database (TCMSP, http://lsp.nwu.edu.cn/tcmsp.php) [25], the BATMAN-TCM [26, 27] (http://bionet.ncpsb.org/batman-tcm/) database, and literature searching in the PubMed database. TCMSP database is a unique systems pharmacology platform designed for herbal medicines. The newly developed TCMSP provides up-to-date, quantitative, and system information about TCM ingredients, ADME-related properties, targets, and diseases [28]. The newest version of TCMSP comprises 510 effective herbal entries registered in the Chinese Pharmacopoeia with more than 31,000 ingredients, which spread over 18 different drug classes [29]. BATMAN-TCM database is the first online bioinformatics analysis tool specially designed for the research of molecular mechanisms of TCM based on TCM ingredients’ target prediction and subsequent network pharmacology analyses [26]. The structures of these ingredients were saved as a mol2 format for further analysis. Chem3D Pro 14.0 was employed to optimize these molecules and minimize the energy.

Active ingredient screening including absorption, distribution, metabolism, and excretion (ADME) evaluations of drugs is crucial in drug development and discovery [30]. Because biological experiments are time-consuming and of high cost, the identification of ADME properties by computational methods has now become an inevitable choice in pharmaceutical research. Here, 2 ADME-related models, including the evaluation of oral bioavailability (OB) and drug-likeness (DL), were calculated to screen the potential bioactive ingredients of WZESD. OB represents the percentage of an orally administered dose of unchanged drug that reaches the systemic circulation and is one of the most important pharmacokinetic parameters [31]. Poor OB is usually the main reason for the failure of drug discovery, especially for TCM which in most cases are oral administration. DL index is a qualitative concept used in drug design for estimating the “druggability” of a substance calculated using the Tanimoto coefficient [32, 33]. The evaluation of DL can facilitate screening for excellent compounds and increase the hit rate for the candidate drugs. For screening in the TCMSP database, OB ≥ 30% (a suggested criterion by the TCMSP database) and DL ≥ 0.18 (mean DL value for all DrugBank compounds) were regarded as the threshold for screening possible candidate drugs presently. At the same time, we also screened the BATMAN-TCM database and pieces of literature in the PubMed database with the Latin name of each herb for the candidate compounds.

2.2. Target Prediction

Drug-target mappings were obtained from three sources. Experimental validated drug-target pairs were retrieved from the HIT database [34]. For those ingredients without validated targets, an in-house developed model SysDT was used which efficiently integrated the chemical, genomic, and pharmacological information for drug targeting and discovery with the random forest (RF) and support vector machine (SVM) algorithm [35]. This method displays the incredible performance of prediction for drug-target interactions, with a concordance of 82.83%, a sensitivity of 81.33%, and a specificity of 93.62%, respectively [28]. The information for the targets based on the above 2 methods was collected from the TCMSP database. A similarity-based method on the basis of the BATMAN-TCM database was employed to predict potential targets of TCM ingredients, the core idea of which was to rank potential drug-target interactions based on their similarity to the known drug-target interactions [26]. UniProt [36] (http://www.uniprot.org/) was applied to obtain the official name of the predicted targets.

2.3. Oligoasthenozoospermia-Specific Protein Collection

Information on oligoasthenozoospermia-associated target genes was collected from GeneCards: the human gene database [37] (http://www.genecards.org/, ver. 4.8.2), and only “Homo sapiens” proteins linked to oligoasthenozoospermia were selected. GeneCards is a searchable, integrative database that automatically integrates gene-centric data from ∼125 web sources, including genomic, transcriptomic, proteomic, clinical, and functional information. Furthermore, these target genes were submitted to STRING [38, 39] (https://string-db.org/) to generate the protein interacting network. The STRING database aims to collect, score, and integrate all publicly available sources of protein-protein interaction information and to complement these with computational predictions and then achieve a comprehensive and objective global network, including direct (physical) as well as indirect (functional) interactions [40].

2.4. Molecular Docking

AutoDock Vina [41] was used in this study to evaluate the potential binding mode between bioactive compounds and putative targets; Discovery Studio 4.5 Client, BIOVIA [42], was applied to analyze the docked structures. The crystal structure of the target proteins of WZESD for oligoasthenozoospermia therapy was downloaded from the RCSB Protein Data Bank [43] (http://www.rcsb.org). The 3D chemical structures of bioactive ingredients were downloaded from the PubChem Compound database [44] or TCMSP database and submitted to minimize the energy by using the molecular mechanics-2 (MM2) force field in Chem3D Pro. The protein-ligand docking active site was defined by the location of the original ligand. Dimensions of the grids were set at 30 × 30 × 30 Ǻ in the x-, y-, and z-directions, with a spacing of 0.375 Ǻ between the grid points and the center placed at the active site of the original ligand crystallographic structures. All other docking and consequent scoring parameters used were kept at their default settings. The compound was regarded to be an efficient drug if the binding affinity was higher than that of the original ligand.

2.5. Network Construction and Analysis

All networks were generated and analyzed by an open source of bioinformatics package for biological network analysis and visualization, Cytoscape 3.5.1 [45]. Two topological parameters, degree and betweenness centrality, were calculated for the obtained networks which imply the significance of a node. The top 10 compounds and targets based on the 2 parameters were considered as key components and core proteins.

2.6. Bioinformatics Analysis

Gene Ontology (GO) analysis is a major bioinformatics tool for annotating genes and gene products, and it aims to unify the representation of gene and gene product attributes across all species [46]. The Kyoto Encyclopedia of Genes and Genomes (KEGG) [47] is a collection of databases containing advanced functional information for the systematic analysis of gene functions, biological pathways, diseases, drugs, and chemical substances. For GO and KEGG pathway analyses, the Metascape online tool [48] (http://metascape.org) and “Enrichr” platform [49] (http://amp.pharm.mssm.edu/Enrichr/) were used to identify the predominant biological processes (BP) and KEGG pathways regulated by WZESD.

3.1. Bioactive Compound Screening and Target Prediction for WZESD

A total of 276 candidate compounds were identified in WZESD, including 66 in SC, 10 in PS, 69 in LF, 15 in CS, 6 in RF, 77 in RS, 11 in FR, 3 in DRR, 5 in RRP, 14 in PR, 6 in OJ, and 23 in EH, respectively (Table S1). The 276 candidate compounds yield 681 candidate targets (Table S2), and the connections between them reach up to 5055. The candidate compounds and candidate targets were submitted to generate the cC-cT network (Figure 1), and details are depicted in the following part.

3.2. Candidate Compound-Candidate Target (cC-cT) Network Construction and Analysis

As shown in Figure 1, the network consists of 998 nodes and 5055 edges, including 12 herbs (squares), 276 compounds (triangles), and 681 targets (squares). The numbers of candidate targets in SC, PS, LF, RF, CS, RS, FR, DRR, RRP, PR, OJ, and EH were 204, 162, 388, 215, 178, 238, 178, 20, 20, 134, 41, and 216, respectively (Table S3). Although the number of targets in each herb was different, they overlapped dramatically in the 12 herbs indicating that different ingredients in WZESD shared common or similar targets with synergistic effects. The duplicated target proteins were removed, and 681 unique terms were retained. Much of the compounds possessed more than one target, and a target was anchored by more than one compound, revealing the “multicompound”, “multitarget” therapeutic mechanism of WZESD. Two centrality indicators, degree and betweenness, were calculated in this network. Different centralities reflect different importance of nodes in a network from different angles. The top 10 compounds and targets through network analysis are shown in Tables 1 and 2. Quercetin, L-asparagine, kaempferol, and stigmasterol were predicted as the major active compounds of WZESD. The proteins including PTGS2, NCOA2, PGR, PTGS1, and HSP90 were predicted as essential pharmacological proteins for the therapeutic effects of WZESD.

3.3. Analyses on the Oligoasthenozoospermia-Based Specific Protein Interaction Network

446 target genes related to oligoasthenozoospermia were searched in the GeneCards database (Table S4) and further submitted to the STRING database (34 proteins were not recognized) to construct the protein-protein interaction (PPI) network (Figure 2(a)). This network contains 412 nodes and 4297 edges. The size and color of the node are proportional to the value of betweenness and degree (Table S5), respectively. The core proteins through network analysis were ALB, TP53, INS, AKT1, and MAPK1 (Table 3) which indicated the essential role of these proteins in the pathophysiology process of oligoasthenozoospermia. Seventy-nine proteins in this network were anchored by WZESD and further discussed in the following part (Figure 2(b)).

3.4. Co-Bioinformatics Analysis for Targets of WZESD and Oligoasthenozoospermia-Specific Proteins

To explore the pharmacology mechanism of WZESD for oligoasthenozoospermia treatment, the relationship between WZESD targets and oligoasthenozoospermia-specific proteins was analyzed (Figure 3). 79 proteins were overlapped in the 2 groups of protein lists (Figure 3(a)). All the selected targets of WZESD and oligoasthenospermia-specific proteins were collected for enrichment network analysis, and most of the targets of WZESD could fall into GO items with the same statistical significance as oligoasthenozoospermia-specific proteins (Figures 3(b)–3(d)), indicating the strong function association between the two comparison cohorts. Gamete generation (GO: 0007276), the developmental process involved in reproduction (GO: 0003006), regulation of hormone levels (GO: 0010817), etc., were the top 10 enriched GO clusters.

3.5. Potential Compound-Potential Target (pC-pT) Network Construction and GO and KEGG Analyses

To further investigate the therapeutic mechanism of WZESD for oligoasthenozoospermia treatment, the 79 overlapped target proteins were clarified in detail, and a pC-pT network was constructed (Figure 4). This network consists of 330 nodes (12 herbs, 213 potential compounds, and 79 potential targets) and 1390 edges. Compounds were manually divided into 2 groups of lists: compounds originated from WZYZ pill (LightSlateBlue) and compounds from other herbs of WZESD (red). The Wuzi Yanzong (WZYZ) pill is one of the most commonly prescribed Chinese herbal formulas for male infertility treatment. It was first recorded in the book called “Xuan Jie Lu” in AD 733 and spread around Northeast Asia over the next thousand years. Cuscutae Semen (CS, Tusizi), Lycii Fructus (LF, Gouqizi), Rubi Fructus (RF, Fupenzi), Schisandra Chinensis (SC, Wuweizi), and Plantaginis Semen (PS, Cheqianzi) are the main components of WZYZ pill. WZESD is originated from WZYZ pill. Based on WZYZ pill, Radix Salviae (RS, Danshen), Figwort Root (FR, Xuanshen), Rehmanniae Radix Praeparata (RRP, Shudihuang), Dried Radix Rehmanniae (DRR, Shengdihuang), Epimrdii Herba (EH, Yinyanghuo), Ophiopogon japonicus (OJ, Maidong), and Polygonati Rhizoma (PR, Huangjing) were added to enhance the efficacy. So, we compared the compounds and targets between the two groups. Seven compounds, including quercetin, campesterol, kaempferol, beta-sitosterol, sitosterol, gamma-sitosterol, and stigmasterol, were overlapped between the 2 groups (DeepSkyBlue triangles). 108 specific compounds (LightSlateBlue triangles) and 7 overlapped compounds (DeepSkyBlue triangles) from WZYZ pill anchored 70 potential targets (including 12 specific). However, 98 components (7 overlapped included) in the additional group targeted 67 potential proteins (including 9 specific proteins: EDN1, PCNA, DNMT1, SLC6A14, VDAC1, VDAC3, VDAC2, BGLAP, and NFKB1). Quercetin, kaempferol, luteolin, beta-sitosterol, and stigmasterol were the crucial potential active compounds, while PTGS1/2, PGR, NCOA2, ESR1, and AR were the core targets of WZESD for oligoasthenozoospermia therapy through network analysis (Tables 4 and 5).

For example, quercetin is a flavonoid that has been reported to possess strong antioxidant properties and is an effective free radical scavenger, and it showed positive effects on multiple functional parameters of spermatozoa, including viability and motility [53]. Kaempferol is another kind of flavonoid, and both quercetin and kaempferol can reduce the DNA damaging in the human sperm caused by reactive oxygen species (ROS) [54]. Luteolin can protect mice from severe acute pancreatitis through anti-inflammatory and antioxidant effects [55]. For target analysis, prostaglandin G/H synthase 1/2 (PTGS1/2), namely, cyclooxygenase-1/2 (COX-1/2), is the key enzyme in the conversion of polyunsaturated fatty acids and arachidonic acid to prostaglandin (PG). COX-2 has been demonstrated to be upregulated in infertility men [56], and treatment with the COX-2 inhibitor can improve sperm motility and morphology as well as increase the pregnancy rate in infertile males [57]. Estrogen receptor 1 (ESR1) is one of the main nuclear hormone receptors for estrogen and plays an important role in the regulation of spermatogenesis [58]. Efferent ductules and epididymal functions are dependent on estrogen signaling through ESR1, whose loss impaired ion transport and water reabsorption, resulting in an abnormal sperm [59]. Androgen receptor (AR) is essential for normal male reproductive development and function. AR signaling is required for the maintenance of spermatogonial numbers, blood-testis barrier integrity, completion of meiosis, adhesion of spermatids, and spermiation [60].

GO and KEGG analyses were conducted via the online “Enrichr” platform to further explain the pharmacological mechanisms [61] (Figure 5). The 79 potential targets mainly participate in nitric oxide biosynthetic and metabolic process, apoptosis, and regulation of reactive oxygen biosynthetic process, which were strongly associated with oligoasthenozoospermia (Figure 5(a)). Reactive oxygen species (ROS) are an integral component of sperm developmental physiology, capacitation, and function. Elevated ROS levels, from processes such as infection or inflammation, can be associated with aberrations of sperm development, function, and fertilizing capacity [62]. The oxidative damage targets all cell components, reducing sperm motility and mitochondrial activity [63]. Therapy with ROS scavengers can significantly improve the in vitro function of human spermatozoa and may be beneficial in infertility patients [64, 65]. Apoptosis plays an important role in regulating spermatogenesis of various mammalian species, including humans [66]. This process of regulated cell death serves several important functions in the testis, a few of which include maintaining appropriate germ cells to Sertoli cell ratio, removing defective germ cells, and maintenance of overall quality control in sperm production [67]. However, high rates of apoptosis have been reported in testicular biopsies from infertile men [68]. Inflammatory disease has been established to affect male reproductive function and fertility. Male accessory gland infections account for almost 15% of all cases of male infertility seen in infertility clinics [69].

The top 10 enriched KEGG pathways were ranked in the descending order using a combined score as follows: pathways in cancer, prostate cancer, HIF-1 signaling pathway, etc. (Figure 5(b)). We found that pathways in cancer was the most significant enrichment term for the 79 target proteins; others such as hepatitis B, toxoplasmosis, and apoptosis share several same target genes with cancer-related pathways. This implies the crosstalk among different pathways in the complex molecular biological network. Many pathways such as the Wnt signaling pathway, PI3K-AKT signaling pathway, and MAPK signaling pathway are the main part of the cancer signaling pathway, which results in sustained angiogenesis, apoptosis evading, and cell proliferation. To a certain degree, there are several common pathways between the two different pathophysiological processes of spermatogenesis and tumorigenesis. The regulation of these signaling pathways by WZESD may be the therapeutic mechanism for oligoasthenozoospermia treatment.

3.6. Molecular Docking Validation

To evaluate the reliability of the above drug-target interactions and to further predict the accurate binding modes, a molecular docking method with AutoDock Vina [41] was employed. Molecular docking is a computational method that attempts to predict the noncovalent binding of macromolecules or, more frequently, of a macromolecule (receptor) and a small molecule (ligand) efficiently. The crystal structures of the potential targets were downloaded from the RCSB Protein Data Bank (http://www.rcsb.org). The original ligand was split and redocked to the proteins, the same as the potential compounds. A compound was considered to be an efficient drug only when the binding affinity was higher than that of the original ligand. 39 potential targets and 180 potential compounds were employed for molecular docking (Table S6). Finally, 61 potential compounds showed better binding affinity with 21 potential targets and formed 120 pairs of compound-target interactions (Figure 6). AR, NOS2/3, PIK3CG, and PTGS2 were the top 5 targets bound by active compounds through molecular docking validation. The detailed docking result is shown in Table S7. Three proteins, including PTGS2, XDH, and NOS2 with corresponding compounds that have the highest binding affinity, were used to display the exact binding mode (Figure 7). 1,2,5,6-Tetrahydrotanshinone was located in the binding pocket of PTGS2 and formed one conventional hydrogen bond by interacting with the key amino acid SER-530. Additionally, van der Waals with TYR-348, PHE-381, and ARG-120, pi-sigma with VAL-349, LEU-352, and ALA-527, and amide-pi stacked with GLY-526 were found in the active site which helped in the stabilization of the compound at the binding site (Figure 7(a), the global view; Figure 7(b), 3D partial view; Figure 7(c), 2D view for the interaction). In the active site of XDH, SER-876, ARG-880, SER-1008, and THR1010 formed 4 solid conventional hydrogen bonds with luteolin. Besides, other noncovalent bonds such as van der Waals, carbon-hydrogen bond, and amide-pi stacking were also formed in the binding site and helped in the stability of the complex (Figure 7(d), the global view; Figure 7(e), 3D partial view; Figure 7(f), 2D view for the interaction). However, there were no hydrogen bonds found in the interaction between NOS2 and hypolaetin, but van der Waals with ILE-195, LEU-203, SER-236, etc., and amide-pi stacked with TRP-188 and PHE-363 were recognized in the active site and mediated the pharmacologic action (Figure 7(g), the global view; Figure 7(h), 3D partial view; Figure 7(i), 2D view for the interaction). Based on these findings, we consider that noncovalent bonds such as hydrogen bond, van der Waals, carbon-hydrogen bond, and amide-pi stacking play key roles in the protein-ligand recognition and stability, which may help determine the pharmacology activities.