Diabetes mellitus (DM) has drawn much attention of researchers due to its increasing mortality and complex complications. According to the International Diabetes Federation, DM has become a major threat to the health and the third biggest killer after cardiovascular diseases and cancer¹. Now DM has terrible influences on both high- and low-income countries, with bearing the majority of the burden in low-income countries. According to the latest statistics of WHO, DM directly caused 1.5 million deaths in 2012. There were approximately 422 million (1 in every 11 people) diabetics worldwide in 2014, which was 14-fold more than those in 1980². And the cases will reach 600 million by 2030 based on the developing trend (Kokil et al., 2015).

Diabetes mellitus is a group of chronic and metabolic diseases, in which there are high blood glucose levels over a long period. It is caused by either the body can’t produce enough insulin or the body can’t effectively respond to the produced insulin. Based on different pathogenesis, DM can be categorized into three types. Type 1 diabetes mellitus (T1DM), namely insulin-dependent DM or childhood-onset diabetes, is caused by insulin deficiency, which is related to dysfunction β cells (Enk and Mandelboim, 2014). Type 2 diabetes mellitus (T2DM), namely non-insulin-dependent DM or adult-onset diabetes, is characterized by insulin resistance (IR) and relatively insulin secretion reducing. There is another type of the diabetes, gestational diabetes mellitus (GDM), which is characterized by high blood glucose levels in pregnant women without previously diagnosed diabetes (Keller et al., 2008; Puri and Hebrok, 2012). According to clinical and non-clinical studies, T2DM is the main category accounting for about 90% of diabetes cases, which is also the focus of this review. The risk factors for T2DM are relatively clear including age, obesity, lifestyle, dietary patterns gene-environment interactions and epigenomics, etc. (Herman and Zimmet, 2012; Karter et al., 2013; Maruthur, 2013). It is reported that multiple genes are involved in genetic susceptibility. The pathogenesis of the DM diseases are closely related to the alterations in multiple signal pathways such as JAK-STAT, AMPK, and PI3K (Richard and Stephens, 2011; Samuel and Shulman, 2012). And the changes of microbial environment might affect the DM pathogenesis (Wen et al., 2008). Study of Qin showed that gut microbial dysbiosis and pathogenic bacteria increasing were observed in patients with T2DM (Qin et al., 2012). If left untreated, DM can cause many complications, including kidneys damage, eyes damage, heart disease, stroke, bone fracture risks, and function failure (Jason et al., 2016). So many recent studies are focusing on the anti-diabetic drugs discovery.

A series of medicines, including analogs, sulfonylureas, thiazolidinedione, biguanides, α-glucosidase inhibitors, and DDP-4 inhibitors, have been used as commercial products for therapy of DM (Morral, 2003; Tian N. et al., 2013). However, because of the “one target, one drug” mechanism, they might have similar side effects, such as weight gain, cardiovascular disease, glycopenia, and gastrointestinal effects (Shah and Mudaliar, 2010). Considering the long-term treatment, side effects and high price of the current antidiabetic drugs, there is a huge demand for effective, low-toxic or non-toxic and affordable drugs for DM.

Compared with the synthetic drugs, natural products have played an important role in DM treatment. The treatment benefits are attributed to multi-components and multi-targets to produce combined or synergistic effects. With the rapid development of novel technologies, increasing studies have been focused on antidiabetic natural products. Nowadays, multi-components and multi-targets therapies have been proved to be more effective and less toxicity than traditional single-target drugs (Espinoza-Fonseca, 2006). With the rapid progress of bioinformatics, network pharmacology, integrating network biology and pharmacology, is considered as a promising approach toward more cost-effective drug development (Khan and Khan, 2016). It can reveal the underlying complex relationship between multi-components and multi-targets. Moreover, network pharmacology provides a system-level approach to understand the pathogenesis of diseases, and can be used for lead discovery, target identification and indication prediction, which coincides with the holistic and systematic concepts in traditional Chinese medicine (TCM) theory (Hopkins, 2008; Tang and Aittokallio, 2014). Compared with conventional “one gene, one drug, one disease” mode, network pharmacology focuses on “multi-targets, multi-components” treatment to diseases. And signal network analysis is a highly attractive tool to investigate complex relationships. What’s more, based on network pharmacology, feasible “drug-target-disease” network models can be established through high-throughput screening and bioinformatics (Poornima et al., 2016). It was reported that network pharmacology had various applications, such as herbal ingredient target/disease gene prediction, network balance regulation, elucidation of synergistic ingredient pairs and active ingredient groups (Tao et al., 2013; Zhang X. et al., 2014; Zintzaras et al., 2014). Thus, strategies based on network including network pharmacology and network medicine are increasingly developed and applied. And several databases have been established to assist network pharmacology analysis including PhIN (pharmacology interaction network database) (Wang Z. et al., 2015), C²Maps (computational connectivity maps) (Huang et al., 2012), databases on TCM (Li and Zhang, 2013), CVDHD (The cardiovascular disease herbal database) (Gu et al., 2013b) and anticancer drugs database (Azmi et al., 2010).

In drug discovery, combined multi-components and multi-targets, natural products based on network pharmacology can afford bright perspectives for treating DM in a systematic manner. It is an efficient and time-saving method to find the potential application in other therapeutic categories of drugs by predicting their targets. Therefore, the network pharmacology approach is developed to find effective drugs for the treatment of DM and related metabolic disorders such as obesity and metabolic syndrome. Based on network pharmacology, the antidiabetic compounds were detected from Ge-Gen-Qin-Lian decoction formula, and the mechanisms were elucidated (Li et al., 2014). In the study of Zhang S. et al. (2014), the “drug-target-pathway-disease” network of green tea was established. And neurotrophin signal pathway, p53 signal pathway, and T2DM pathway related to DM were screened by network pharmacology (Zhang S. et al., 2014). It was reported that the network pharmacology analysis was developed to illustrate the action mechanism of Tangminling Pills for T2DM therapy (Gu et al., 2011).

The aim of this review is to provide a summary of network pharmacology studies on the bioactive compounds and action mechanisms of natural products for the treatment of DM. The recent progresses on the different active constituents, databases and applications in the treatment of DM were highlighted. The mechanisms related to T2DM, including α-amylase and α-glucosidase inhibitory, targeting β cell dysfunction, AMPK signal pathway, PI3K/Akt signal pathway and modulation of gut microbiota were summarized. And the perspectives on their reasonable applications for therapy of DM and related drug discovery were put forward.

Network pharmacology, as a powerful tool to elucidate complex relationships from a systematic and biological level, has been more and more popular in recent years. It has been applied to various complex chronic diseases, such as cardio-cerebral ischemic diseases (Gu et al., 2013b), cancer (Azmi et al., 2010; Zhang et al., 2012; Poornima et al., 2016), pancreatitis (Hong et al., 2016), gout (Zhao et al., 2015), Alzheimer’s disease and DM and other metabolic disease (Loscalzo, 2011). For DM therapy, network pharmacology has been widely used in the mechanism studies.

According to the study of Liu et al. (2007), a gene network enrichment analysis (GNEA) was used to identify two sets of genes, IS-HD and NR-HD gene sets, which was associated with insulin signal and a network of nuclear receptors. They identified a network of PPI between members from IS-HD and NR-HD gene sets that might facilitate signal. By using GNEA, it was found that insulin-signal process was obviously transcriptionally altered in IR and T2DM. Their results also indicated that integrating high-throughput microarray studies together with PPI networks was an effective method to prove the existing biological processes associated with a complex disorder (Liu et al., 2007).

In the study of Li et al. (2014), the network pharmacology method was used to assist detecting the potential antidiabetic ingredients from the GQD. They collected 287 available chemical components, with 42 of them found in Ge-Gen, 57 of them found in Huang-Qin, 22 of them found in Huang-Lian, and 166 of them found in Gan-Cao. In this study, network target analysis was applied to elucidate the relationships between herbal ingredients and bioactivities. And the network pharmacology method also contributed to better understand the action mechanisms of GQD formula. It was also found that 4-Hydroxymephenytoin, a novel compound from Puerariae Lobatae radix (Ge-Gen), could increase the insulin secretion and improve IR (Li et al., 2014). The network pharmacology approach had been an effective strategy for the modern research on herbal drugs on the DM treatment, and the multi-active compounds could be identified and the multi-targets action mechanisms could be elucidated.

As the second most consumed beverage in the world next to water, Tea (Camellia sinensis L.) had been used as antidiabetic candidate for a long time according to folk remedies (Chen et al., 2007). Tea polyphenols (GTPs) was one of the main components of tea, mainly including EGCG, EGC, ECG, and EC. Different GTPs could act on different targets in the signal network of complex diseases (diabetes, cancer, cardiovascular, etc.), resulting in a synergistic effect, which was in accordance with the theory of network pharmacology. Study of network pharmacology approach on GTPs identified three pathways, p53 signal pathway, neurotrophin signal pathway, and T2DM pathway, which related to DM from all 147 pathways. And they established a diabetes-related pathway by integrating these three pathways (Zhang S. et al., 2014).

In order to find the potential inhibitors from TCM for the T2DM-related targets, Tian S. et al. (2013) developed an integrated approach that combined molecular docking and pharmacophore mapping, and they established the compound-target interaction network. A total of 2,479 non-duplicated compounds from these 32 Chinese herbs of 52 classical TCM formulas were found from two databases. The Bayesian classifiers with satisfactory discrimination capabilities for 15 targets were employed to screen the 2,479 compounds. The results showed that molecular docking or pharmacophore mapping could give satisfactory predictions for most targets. It was proved that some herbal ingredients could directly interact with T2DM related targets, while some ingredients relieved T2DM via antioxidant effects or other supplementary pharmacological effects (Tian S. et al., 2013).

Zhao H.L. et al. (2012) reported that male Zucker diabetic fatty rats were treated by JCU, a kind of TCM, which contained three herbs and berberine. The results showed that JCU could improve hyperglycemia, ameliorate IR, normalize liver enzyme and hepatocyte ballooning degeneration. On three levels including miRNA level, protein level and mRNA level, the study elucidated the action mechanisms of JCU, regulating the critical cytokines on the related pathways including improving the quality and quantity of cell survival, increasing glucose uptake, and ameliorating insulin sensitivity (Zhao H.L. et al., 2012). It in-depth validated that there was of great significance of multi-components TCM to treat DM.

In the work of Gu et al. (2011), the combination of molecular docking and network pharmacology analysis was developed to illustrate the action mechanism of Tangminling Pills for T2DM therapy. The drug–drug network and drug-target network illustrated that more than 100 active compounds of 667 compounds in the formula would target 37 target proteins related to T2DM. That is, they played important roles in the biological network. As well as the critical ingredients in Tangminling Pills were predicted and some of them had been reported in literatures. In addition, several compounds including Rheidin A, Sennoside C, Rheidin C, procyanidin C1, and Dihydrobaicalin were of importance as the antidiabetic candidate due to its pharmacological effects (Gu et al., 2011).

Diabetes mellitus, a complicated metabolic disease, is a currently health problem causing significant mobility and mortality worldwide, which can cause various complications such as kidneys dysfunction, eyes damage, and cardiovascular diseases. Current antidiabetic drugs are still limited due to single-compound, single-target, side-effects, drug tolerance and low efficacy and they can’t cure DM completely. And it is significant that the world populations are heterogeneous and genetic polymorphisms in pharmacologically relevant genes varying across geographical region (Suarez-Kurtz et al., 2014). As a novel perspective, network pharmacology is becoming more and more popular in drug discovery. Compared with the single target of the current antidiabetic drugs, it is an promising approach which can combat the complex problems (Tao et al., 2014). It may also contribute to better explain the low clinical efficacy, side effects and toxicity of the current clinical drugs. At the same time, natural products have vast structural diversity and various bioactivities, which can provide the opportunities to find different lead compounds, different targets for DM. Hence, limited studies result in an urgent need to develop novel network pharmacology approaches to change current theories of drug discovery and to improve our understanding of action mechanisms. Although network pharmacology of natural products is still in its infancy, the appropriate use of network pharmacology approaches may initiate new directions, overcome the disadvantages of current antidiabetic therapies as well as contribute new insights into the discovery of novel antidiabetic drugs. Specific researches can be done to better elucidate the usage of natural products, and explore the possible action mechanisms.

WL drafted and revised the manuscript. GY, YP, and CW helped in literature survey and manuscript preparation. HC contributed to the conception and design of the work, revised and improved the manuscript. All authors approved on the finally submitted version of the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.