In-silico investigation of curcumin drug-likeness, gene-targets and prognostic relevance of the targets in panels of human cancer cohorts

Despite advancements in diagnostic and standard treatment modalities, cancer survival rate remains disappointing globally. It has however, been recognized that exploring the therapeutic properties of secondary metabolite from natural products may alleviate the problems of drug resistance and toxicity that besiege the conventional therapies, and hence improve the overall prognosis of cancer patient. To this end curcumin, a polyphenolic natural compound has been widely studied for it anticancer activities in in vitro and in vivo models. Computational technology has significantly improved the success rate of drug discovery and development, hence, it has become a widely explore tool in drug candidate identification. In this study we used computational approached to identify 12 genes that are potential druggable candidate for curcumin. The genes identified were found to be enriched in cancer and drug resistance associated signaling pathways. Interestingly, the top 3 identified genes; Microtubule-associated protein tau (MAPT), Toll-like receptor 9 (TLR9) and Tyrosyl-DNA phosphodiesterase 1 (TDP1) were observed to be over expressed in multiple cancer cohorts and were associated with poor prognoses of the patients. Curcumin has good physicochemical, bioavailability and ADMET properties. Importantly, it met the Lipinski's Rule of 5 for drug likeness and thus worthy of further in vitro and in vivo confirmation studies.


Introduction
Despite advancements in diagnostic and standard treatment modalities, cancer remain second leading cause of death worldwide [1]. Currently, there are over 18 million cancer cases with projection of 25 million cases by year 2030 [2,3]. The available treatment modalities for cancers including immunotherapy, radiation therapy and chemotherapy are disappointing, providing a minimal extension of the overall survival and, and general poor prognosis [1]. Furthermore, the available chemotherapy suffers from drug resistance and serious side effects. The dilemma is further compounded by activations of multiple pathways associated with cancer growth and metastasis, hence necessitating the need to explore natural products for alternative multi-target chemotherapeutic agent with minimal side effect [4].
Natural products are diverse in terms of structure and biological activities [5][6][7][8] and thus offer superior therapeutic options and safety than the conventional and synthetic compounds used in high throughput screening processes in drug discovery and development pipelines [9]. Thus, natural products, especially phytochemical extracts from medicinal flora have proven to be immensely valuable as a lead compound for drug discovery and development [10,11]. In fact, an approximated half of the conventional drugs are of phytochemical backbone [12].
Curcumin, is a polyphenolic natural compound isolated from Curcuma longa, a reputable medicinal plants commonly known as turmeric [13]. Curcumin has been reported for several therapeutic activities against a number of diseases; including antioxidants, hepato-protective [14], anti-inflammatory and anticancer activities. Several literatures have emphasized the anti-antineoplastic properties of curcumin against experimental cell line and chemically induced carcinogenesis in animal models of cancer [15][16][17][18].
Curcumin have also been reported to synergize and enhance the activities of conventional drugs in combination therapy; studies have revealed that curcumin can sensitize a variety of tumors including glioma, colon cancer, neuroblastoma, epidermal carcinoma, cervical carcinoma, and prostate cancer to chemo and radio therapies [19]. However, dysregulation of several signaling pathways have been implicated in the activities of curcumin [20]. Computational technology has significantly minimized the experimental drug trials and improve the success rate of drug discovery and development, hence, it has become a widely explore tool in drug candidate identification [21]. To this end, we conducted an in silico drug likeness and target predictions studies on curcumin with a view of unraveling the potential targets that could be implicated in it diverse anti-cancer activities.

Medicinal Chemistry, Drug likeness and in-silico ADMET Properties of curcumin
The medicinal chemistry, drug likeness and admet properties of curcumin were evaluated using SwissADME algorithm (http://www.swissadme.ch). The drug-likeness of the compounds were predicted in accordance to the Lipinski's Rule of 5 for drug candidate. The Rule of 5 accommodate a drug candidate with < 5 H-bond donors, < 10 H-bond acceptors, < 500 Da, and < 5 Log P (CLog P) as a good drug candidate [22].

Identification of curcumin targets, protein-protein interaction network and prognostic relevance of the target genes
The curcumin target genes were identified using SwissTarget algorithm. We used STRING servers (https://stringdb.org/) to predict protein-protein interactions (PPIs) and functional protein partners of the identified curcumin targets. We used the Tumor Immune Estimation Resource (TIMER2.0) algorithm (http://timer.cistrome.org/) for a differential expression analysis of MAPT, TLR9 and TDP1 in tumor vs. normal tissues from various cancers in The Cancer Genome Atlas (TCGA) database [23]. We used Kaplan-Meier (K-M) survival plots to analyze overall survival (OS) and disease-free survival (DFS) ratios between the cancer cohorts with high and low MAPT, TLR9 and TDP1expression profiles from the 9736 tumor samples across 33 cancer types in the Gene Expression Profiling Interactive Analysis (GEPIA) database (http://gepia.cancer-pku.cn/) [24].

Data analysis
The statistical significance of differentially expressed genes was evaluated using the Wilcoxon test. * p<0.05; ** p<0.01; *** p<0.001. The Kaplan-Meier curve was employ to present the patients' survival from different cancer cohorts.

Curcumin has acceptable physicochemical properties and met the requirement for drug likeness.
In-silico approach was recently found to aid identifying a safe and possible drug candidate and also to address the possible molecular mechanism of action of modest molecules with a big protein [21]. It is on this note that we conducted medicinal chemistry, pharmacokinetic (ADMET) studies and "drug-likeness" using Lipinski's rule of five on curcumin (1E,6E)-1,7-bis(4-hydroxy-3-methoxyphenyl) hepta-1,6-diene-3,5-dione) ( Figure 1A) to determine their activity within the human body. Our findings revealed that curcumin has acceptable physicochemical and bioavailability properties ( Figure 1C and Table 1). It is passively absorbed by the gastrointestinal tract but has poor blood brain barrier permeation ability ( Figure 1B). The 'Bioavailability Radar' (Fig. 1c) is designed based on prediction of six physicochemical properties; polarity (POLAR), molecular mass (SIZE), Lipophilicity (LIPO), flexibility (FLEX), saturation (INSATU) and water solubility (INSOLU). These properties of a molecule should fall within desired ranges for it to be suitable as a drug lead. These desired ranges are graphically represented by the pink shaded area of the bioavailability radar. Since all six properties for curcumin are within the pink range, according to the SwissADME calculations, it is suitable to be used as a drug. Also, it was predicted to have a high gastrointestinal absorbance and no violations of rules determining drug-likeness. It has good water solubility (Logs=-3.94) and synthetic availability (2.97%). The results of ADME and drug-likeness properties together with the bioavailability scores suggest that curcumin have good druglikeness properties, and thus have a good potential to be developed into oral drug for therapeutic application [25]

Curcumin are potential multi-targets agent
The past few years have witnessed significant advances in drug discovery technologies for the identification of novel lead compounds against a wide range of therapeutic targets including chemical matter to disrupt protein-protein interactions [26]. In the present study, our drug target analysis of curcumin identified membrane receptors (7%), kinases (20%), enzymes (33%) and others (40%) as the classes of the target genes ( figure 1D). Specifically, 12 genes including; Microtubule-associated protein tau, Toll-like receptor 9, Tyrosyl-DNA phosphodiesterase 1, Muscle blind-like protein, Lactoylglutathione lyase, RAC-alpha serine/threonine-protein kinase, Testosterone 17-beta-dehydrogenase 3, Gamma-secretase C-terminal fragment 59, Quinone oxidoreductase, Estradiol 17-beta-dehydrogenase 12 were identified as a druggable candidate for curcumin with probability range between 0.44-1 (Table 2).

Curcumin target are enriched in cancer associated signaling pathways
We conducted protein interaction networks of these identified curcumin target genes using the STRING algorithm ( Figure 2A) and found that, the enriched KEGG pathways in the gene network clustering include AMPK signaling pathway, mTOR signaling pathway, PI3K-Akt signaling pathway, Insulin resistance, Endometrial cancer, Insulin signaling pathway and EGFR tyrosine kinase inhibitor resistance with strong association strength range between 1.41 to 1.99 and false discovery range of between 1.03E-10 to 8.04E-12 (Table 3). The biological process enriched in the pathway include the intracellular signal transduction, cellular response to nerve growth factor stimulus, regulation of phosphate metabolic process, apoptotic process and cellular response to nerve growth factor stimulus ( Table 3). The second order clustering of the identified curcumin targets generated interactions of 34 and 182 nodes and edges respectively with average local clustering coefficient of 0.685 and PPI enrichment value of 1.0 x 10 -16 ( Figure 2B and accompanying table)

Curcumin targets are over-expressed in multiple cancers and are predictors of poor prognosis
We conducted a gene differential expression analysis of the three top ranked curcumin target genes (Microtubuleassociated protein tau (MAPT), Toll-like receptor 9 (TLR9) and Tyrosyl-DNA phosphodiesterase 1 (TDP1)), between cancer patient and normal patient and found that MAPT, TLR and TDP1 are overexpressed (p < 0.05) in multiple panels of cancer including; breast cancer, colon cancer, head and neck cancer, Kidney cancer, liver cancer, PAAD, melanoma, STAD, THCA and UECE ( figure 3). Furthermore, we also found that higher RNA expression profile of MAPT, TLR9 and TDP1 in all TCGA and the GTEx datasets correlate with low overall survival and disease free survival of cancer patients. (Figure 4)

Conclusion
In conclusion, this study identified 12 genes as a druggable candidate for curcumin. The genes identified were enriched in cancer and drug resistance associated signaling pathway. Interestingly, the top 3 identified genes; (Microtubuleassociated protein tau (MAPT), Toll-like receptor 9 (TLR9) and Tyrosyl-DNA phosphodiesterase 1 (TDP1) were observed to be over expressed in multiple cancer cohorts and were associated with poor prognoses of clinical cancer cohorts. Curcumin has good physicochemical, bioavailability and ADMET properties. Importantly, it met the Lipinski's Rule of 5 for drug likeness and thus worthy of further in vitro and in vivo confirmation studies on the targets identified in this study.