Identification of gene-phenotype connectivity associated with flavanone naringenin by functional network analysis

Search for direct protein targets of naringenin

The DrugBank database (https://www.drugbank.ca) is a web-based bioinformatic tool, including numerous drug information such as chemical structure, targets, transporters, biointeractions, and so on (Wishart et al., 2006, 2008). The comprehensive information could be used to facilitate construction of drug-target interactome in silico. Thus, the DrugBank database was employed to search for DPTs of naringenin. The analysis results were then used for further analysis.

Network visualization and KEGG pathway enrichment analysis

Direct protein targets obtained from the DrugBank database were in turn subject to retrieval in the STRING database (http://www.string-db.org/), setting the minimum required interaction score as 0.5 and max number of interactors as 20. The IPTs of naringenin were generated after removing repetitive proteins. Then interactions of all naringenin-mediated proteins were analyzed by the STRING web server, and their PPI network was visualized using Cytoscape software (version 3.6.1) (Su et al., 2014). Subsequently, all naringenin-mediated proteins were also carried on functional KEGG pathway enrichment analysis by DAVID with p-value < 0.05.

Exploring cancer genomics data linked to naringenin by cBioPortal

The cBioPortal for Cancer Genomics (http://cbioportal.org) is an open-access platform for cancer researchers to explore multidimensional cancer genomics data (Cerami et al., 2012; Gao et al., 2013). Biologic insights and clinical applications could be understood using these rich genomic data. In this study, screened protein genes (ESR1, PIK3CA, AKT1, MAPK1) from the investigation above were subject to genetic alteration analysis in all prostate cancer studies available in cBioPortal databases. Then, the prostate cancer study with the highest genetic alteration was chosen to analyze the connectivity of the screened protein genes (ESR1, PIK3CA, AKT1, MAPK1). Gao et al.’s (2013) study helped us to interpret the result of the query in cBioPortal.

Characterization of naringenin DPTs

Some small molecule chemical compounds can directly interact with proteins in organisms, which will result in alteration of the downstream pathway and various physiologic functions. In order to achieve DPTs of naringenin and relevant information, we entered Drugbank using naringenin as input. As a result, the output displayed DB03467 and total of eight primary DPTs. In the category of pharmacological activity, naringenin was classified as antiulcer agents, estrogen antagonists, BCRP/ABCG2 inhibitors, gastrointestinal agents, and hormone antagonists. Subsequent screening demonstrated seven DPTs belonging to human beings, which were presented in Table 1, ESR1, AKR1C1, CYP1B1, KANSL3, SHBG, CYP19A1, and ESR2. In addition, interactions between naringenin and seven DPTs were analyzed and illustrated in Fig. 2. It should be noted three direct target proteins (CYP19A1, SHBG, ESR2) showed a direct association with ESR1, suggesting that ESR1 played a critical role in targeted pathways controlled or mediated by naringenin.

Visualization and PPI network construction of naringenin-mediated proteins

In biological systems, physiologic functions rely on interactions among diverse proteins, which can be represented by a network consisting of nodes and edges. Thus, there should be of great importance to search for downstream targets of naringenin DPTs. To this end, STRING database was applied to identify DPTs-related proteins and the results were summarized in Table S1. In general, total of 109 naringenin-mediated proteins (Table S2) were generated including seven DPTs and 102 IPTs. Moreover, the interaction data of all naringenin-mediated proteins were obtained using STRING (Table S3) and the PPI network were then visualized by Cytoscape 3.6.1 (Su et al., 2014). As shown in Fig. 3, there were 912 PPI pairs in the network of naringenin-mediated proteins. The nodes indicated proteins and the edges indicated the interactions between these proteins. Next, node degree was evaluated to investigate the centrality of proteins and represented by size, low values to the small size. Ranked by degree value from large to small, top 10 proteins were screened out and showed in Table 2, including three DPTs (ESR1, CYP19A1, ESR2). These findings collectively indicated that biological effects mediated by naringenin are connected to ESR1.

KEGG enrichment pathway analysis of naringenin-mediated proteins

As a collection of molecular interaction, reaction, and relation networks, KEGG pathways offer a good understanding of high-level functions and the biological system. Thus, we performed the KEGG pathway enrichment analysis using DAVID to assess the functional feature of naringenin-mediated proteins (Table S4). As can be seen in Table 3, the top 15 KEGG pathways linked to naringenin-mediated proteins were identified and included steroid hormone biosynthesis, metabolism of xenobiotics by cytochrome P450, chemical carcinogenesis, drug metabolism-cytochrome P450, retinol metabolism, pentose and glucuronate interconversions, ascorbate and aldarate metabolism, ovarian steroidogenesis, drug metabolism-other enzymes, porphyrin and chlorophyl metabolism, prolactin signaling pathway, thyroid hormone signaling pathway, estrogen signaling pathway, metabolic pathways, and prostate cancer. The broad grouping of functional enrichment analysis results indicated that naringenin-mediated proteins were mainly linked to (1) estrogen-related signaling pathways, (2) basal metabolism pathways, (3) cancer-related pathways. Given the importance of ESR1 in all naringenin-mediated proteins, emphasis of biological functions was directed to pathway possessing ESR1. Thus, three ESR1-related pathways were identified: prolactin signaling pathway, thyroid hormone signaling pathway, and estrogen signaling pathway. Besides, because anticarcinogenic properties of naringenin have been reported in diverse malignant tumors as mentioned above, prostate cancer pathway consisting of eight proteins was found: AKT1, MAPK1, IGF1R, AR, CCND1, INS, PIK3CA, IGF1. This result suggested that prostate cancer might be as a phenotype connected to naringenin-mediated proteins.

Genetic alterations connected with naringenin-mediated protein genes in prostate cancer

Previous functional enrichment analysis uncovered the link between naringenin-mediated proteins and prostate cancer pathway. Further exploration was needed to validate this link. Since ESR1 enjoyed the highest centrality among all naringenin-mediated proteins as proved previously, and because three overlapping proteins (PIK3CA, AKT1, MAPK1) associated with ESR1 were found to be connected to prostate cancer pathway as well, the genomic alteration of these four protein genes (ESR1, PIK3CA, AKT1, MAPK1) were checked in prostate cancer using cBioPortal. Among 16 prostate cancer studies (Baca et al., 2013; Barbieri et al., 2012; Beltran et al., 2016; Fraser et al., 2017; Gao et al., 2014; Grasso et al., 2012; Hieronymus et al., 2014; Knudsen & Scher, 2009; Cancer Genome Atlas Research Network, 2015; Robinson et al., 2015; Taylor et al., 2010), gene alterations ranged from 0% to 31.58% as displayed in Fig. 4. Because of the most pronounced genetic alterations among all prostate cancer studies available in cBioPortal, the NEPC study (Beltran et al., 2016) was selected individually to view relevant genomic changes of four genes. A concise and compact summary of alterations in three queried genes (ESR1, PIK3CA, AKT1, MAPK1) was shown in Fig. 5 using OncoPrint. The results presented that most gene alterations belonged to amplification. Additional mutual exclusivity analysis indicated that every gene pair exhibited significant (p-value < 0.05) co-occurrence in prostate samples in the study of NEPC (Table 4). The co-occurrence of ESR1 and other three genes revealed a central axis function for ESR1 under naringenin control.

Subsequently, the Network tab embedded in cBioportal was exploited to explore the interactive relationship between four selected genes and genes that were altered in NEPC prostate cancer study. A query of ESR1, PIK3CA, AKT1, and MAPK1 automatically generated a network containing 50 neighbor genes of four query genes, and legends were available to explain network symbols (Fig. 6A). To manage network complexity, we filtered neighbors by 45% alteration, such that only AR gene with the highest alteration frequency remained in addition to four query genes (Fig. 6B). The pruned network revealed the potential interactions between naringenin-mediated genes and altered genes in prostate cancer samples. Moreover, specific cancer drugs acting on ESR1, PIK3CA, AKT1, and MAPK1 were displayed in Fig. 5. ESR1 was the main target of most FDA approved drugs (represented by yellow hexagon) in the network, providing a molecular basis for potential clinical applications of naringenin to treat prostate cancer targeting ESR1.