Bioinformatics screening the novel and promising targets of curcumin in hepatocellular carcinoma chemotherapy and prognosis

Background The aim of present study was to screen the novel and promising targets of curcumin in hepatocellular carcinoma diagnosis and chemotherapy. Methods Potential targets of curcumin were screened from SwissTargetPrediction, ParmMapper and drugbank databases. Potential aberrant genes of hepatocellular carcinoma were screened from Genecards databases. Fifty paired hepatocellular carcinoma patients’ gene expression profiles from the GEO database were used to test potential targets of curcumin. Besides, GO analysis, KEGG pathway enrichment analysis and PPI network construction were used to explore the underlying mechanism of candidate hub genes. ROC analysis and Kaplan-Meier analysis were used to evaluate the diagnostic and prognostic value of candidate hub genes, respectively. Real-time PCR was used to verify the results of bioinformatics analysis. Results Bioinformatics analysis results suggested that AURKA, CDK1, CCNB1, TOP2A, CYP2B6, CYP2C9, and CYP3A4 genes served as candidate hub genes. AURKA, CDK1, CCNB1 and TOP2A were significantly upregulated and correlated with poor prognosis in hepatocellular carcinoma, AUC values of which were 95.7, 96.9, 98.1 and 96.1% respectively. There was not significant correlation between the expression of CYP2B6 and prognosis of hepatocellular carcinoma, while CYP2C9 and CYP3A4 genes were significantly downregulated and correlated with poor prognosis in hepatocellular carcinoma. AUC values of CYP2B6, CYP2C9, and CYP3A4 were 96.0, 97.0 and 88.0% respectively. In vitro, we further confirmed that curcumin significantly downregulated the expression of AURKA, CDK1, and TOP2A genes, while significantly upregulated the expression of CYP2B6, CYP2C9, and CYP3A4 genes. Conclusions Our results provided a novel panel of AURKA, CDK1, TOP2A, CYP2C9, and CYP3A4 candidate genes for curcumin related chemotherapy of hepatocellular carcinoma. Supplementary Information The online version contains supplementary material available at 10.1186/s12906-021-03487-9.

common type of primary liver cancer, with its high-risk factors at least including metabolic liver disease, hepatitis virus infection, and alcohol abuse [3,4]. Long-term exposure of hepatitis B/C virus would develop chronic viral hepatitis, then followed by liver cirrhosis and hepatocellular carcinoma [3,5]. Nonalcoholic fatty liver disease or metabolic associated fatty liver disease, one typical metabolic liver disease, would develop nonalcoholic steatohepatitis, then followed liver cirrhosis and hepatocellular carcinoma [6,7]. Alcohol abuse would induce chronic liver injury, which developed liver fibrosis and eventually progressed to hepatocellular carcinoma [8].
Earlier-stage clinical symptoms of hepatocellular carcinoma were vague or nonspecific, thus hepatocellular carcinoma patients usually were diagnosis at an intermediate and advanced stage. Surgical resection was the ideal option for earlier-stage hepatocellular carcinoma patients without cirrhosis, while transplantation was the best option for those earlier-stage hepatocellular carcinoma patients with cirrhosis [9,10]. Systemic therapies at least include chemotherapy, immunotherapy and radiotherapy, which were strongly recommended for hepatocellular carcinoma patients at intermediate and advanced stage. However, up to date, chemoprevention and adjuvant therapy regarded as much less efficient interventions in advanced hepatocellular carcinoma treatments [9,11].
Natural herb or peptide have exhibited antioxidant, anti-inflammatory, and anti-proliferative effects on disease treatment [12][13][14][15][16]. Accumulating evidence indicated that curcumin was a promising natural compound, which has been extensively investigated and shown multiply therapeutic activities, at least including anticancer, antivirus, anti-arthritis, anti-amyloid, anti-oxidation, and anti-inflammatory [17]. In the molecular events of liver disease, curcumin inhibits HBV gene expression and replication via down-regulation of PGC-1α [18]. Curcumin had no effect on HCV RNA replication or viral assembly/ release, but impaired virus binding and entry into human liver cells [19]. Randomized Controlled Trials showed that curcumin significantly ameliorated nonalcoholic fatty liver disease [20,21]. Curcumin increased PPARγ to inhibit the expression of SREBP-2 and low-density lipoprotein receptor, which subsequently inactivated hepatic stellate cells, curcumin also increased SREBP-1c to promote lipid storage [22]. Thus, these findings indicated the potential therapeutic value of curcumin in protecting against liver steatosis and fibrosis. Curcumin would decrease the stemness of liver cancer stem cells by attenuating NF-κB/HDAC signaling [23]. Interestingly, curcumin suppressed stromal cell-derived factor-1/CXCR4 signaling to reduce the incidence of circulating gastric cancer cells, and subsequently decreased the risk of secondary liver cancer [24]. Thus, these previous studies highlighted that curcumin might exhibit multifunction in the initiation and progression of liver cancer.
In the present study, we performed bioinformatic analysis to screen the targets of curcumin, which would contribute to initiation and progression of hepatocellular carcinoma. Then we further validated these candidates with gene profiles of hepatocellular carcinoma patients, and tried to answer the underlying mechanism.

Screening the potential curcumin-related targets for hepatocellular carcinoma therapy
Simplified Molecular Input Line Entry Specification (SMILES) structure of curcumin was obtained from Pubchem website (https:// pubch em. ncbi. nlm. nih. gov/). The SMILES structure of curcumin was used to predict the potential Homo Sapiens target from SwissTargetPrediction database and PharmMapper database (version 2017) [25,26]. We used "curcumin" to screen its verified targets from DrugBank database [27]. Curcumin relative targets were then integrated with the above three databases and removed the repeats. We used "Hepatocellular Carcinoma" and "Hepatocellular Cancer" items to acquire the potential hepatocellular carcinoma relative target from GeneCards database [28], then we transformed their gene ID from Uniprot database (https:// www. unipr ot. org/) for further analysis. We next merged the curcumin relative target and the hepatocellular carcinoma relative target and picked up the overlapped candidate for further analysis.

Clinical data collection and processing
The hepatocellular carcinoma relative microarray data (GSE14520) was based on GPL3921 platform and downloaded from Gene Expression Omnibus (GEO) database by using GEOquery package (version 2.56.0) [29], the data was firstly normalized by using normalizeBe-tweenArrays function of limma package (version 3.44.3) [30][31][32]. We then obtained 50 fully paired gene expression matrix of normal adjacent liver tissues and hepatocellular carcinoma tissues from GSE14520. The principal component analysis (PCA) was performed to visually present the data by FactoMineR package (version 2.3) and factoextra package (version 1.0.7) [33]. The gene expression matrix and differential expression analysis were carried out by using pheatmap package (version 1.0.12).

Protein-protein interaction (PPI) network construction
The selected candidate gene was uploaded onto Search Tool for the Retrieval of Interacting Genes (STRING, version 11.0) databaseto predict and construct a potential PPI network [38], then network data integration, analysis, and visualization were performed by using Cytoscape software (version 3.7.1).

Expression, prognosis, and diagnosis analysis of selected genes
In order to track the dynamic expression of selected genes in initiation and progression of hepatocellular carcinoma, we downloaded the mRNA expression profile and the corresponding clinical information of hepatocellular carcinoma from The Cancer Genome Atlas (TCGA) database (http:// cance rgeno me. nih. gov/) by TCGAbiolinks package (version 2.16.3), including normal samples (n = 44), stage I samples (n = 175), stage II samples (n = 86), stage III samples (n = 84), and stage IV samples (n = 5). Expression data of selected genes were processed by GraphPad Prism 9, and analyzed by Student's t-test. The Kaplan-Meier survival analysis was constructed by Gene Expression Profiling Interactive Analysis (GEPIA) on-line tool [39]. The receiver operating characteristic (ROC) curve was plotted by using dplyr package (version 1.0.3) and pROC package (1.16.2) [40]. P-value less than 0.05 was regarded as the cutoff value of statistical significance.

Cell culture and treatment
HepG2.2.15 cell line was derived from HepG2 cell which was transfected full length DNA of Hepatitis B Virus, purchased from China Center for Type Culture Collection, and which was a gift from Dr. Liufeng Mao. For cell culture, HepG2.2.15 cells were maintained in complete growth Dulbecco's Modified Eagle Medium (DMEM) containing 4.5 g/L glucose (Life Technologies, Inc., Carlsbad, CA), 10% fetal bovine serum, 100 U/ml penicillin, and 100 μg/ml streptomycin. All cells were incubated at 37 °C with 5% CO2. For cell treatment, HepG2.2.15 cells were seeded into 12-well plates at 2 × 10 5 cells/well overnight, then cells were treated with 1, 4, 10 μM curcumin (CSNpharm, China) for further 24 h, dimethyl sulfoxide (Beyotime, China) was used as the vehicle control.

RNA isolation and quantitative RT-PCR
Total RNA was extracted using the RNAeasy ™ Animal RNA Isolation Kit with Spin Column (Beyotime, China) according to manufacturer's instructions. Total RNA was reversely transcribed amplified using the Beyo-Fast ™ SYBR Green One-Step qRT-PCR Kit (Beyotime, China). All real-time PCR primers were listed as following: AURKA, 5′-CTA ACG GCT GAG CTC TTG GA-3′ and 5′-GAA CCG ACA GGG GAC TTG AC-3′. Relative mRNA expression levels of target genes were normalized by comparing to GAPDH, and calculated using the 2 − ΔΔCt method [41]. Expression data of selected genes were processed by GraphPad Prism 9, and analyzed by Student's t-test. P-value less than 0.05 was regarded as the cutoff value of statistical significance.

Cell counting kit-8 assay
Cells were seeded into 96-well plates at 10000 cells/well, after culture for 24 h, cells were treated with indicated concentrations of curcumin (CSNpharm, China) for further 48 h, and then the cell vitality was measured by the cell counting kit-8 (Beyotime, China) kit. The cellular viability (%) = OD value (treated cell) / OD value (control cell) × 100%.

Potential targets of curcumin in hepatocellular carcinoma therapy
OEChem 2.1.5, PubChem release). 3D conformer of curcumin was shown in Fig. 1A. Thirteen candidates of curcumin were acquired by Drugbank, 299 candidates of curcumin were acquired by PhamMapper, 100 candidates (Probability> 0) of curcumin were acquired by Swisstarget (Fig. 1B). After deleting the repeats candidates, we obtained 144 candidates of curcumin (Fig. 1B). Five thousand two hundred eighty-five aberrant genes were acquired by GeneCards, when they overlapped with candidates of curcumin, we here obtained 109 potential targets of curcumin in hepatocellular carcinoma therapy (Fig. 1B). KEGG pathway analysis showed that aberrant expression of these 109 potential targets significantly and mainly correlated with viral-induced carcinoma, at least including prostate cancer, hepatocellular carcinoma, pancreatic cancer, and leukemia. They also highly correlated with cytochrome P450-related drug metabolism (Fig. 1C). Biological process (BP) of GO analysis showed that these 109 potential targets mainly regulated xenobiotic-induced protein phosphorylation in long-chain fatty acid metabolic process (Fig. 1D). Cellular component (CC) of GO analysis showed that these 109 potential targets mainly regulated activities of protein kinases in the transcriptional process (Fig. 1D). Molecular function (MF) of GO analysis showed that these 109 potential targets mainly responded to activities of the transmembrane receptor protein kinase and the histone kinase (Fig. 1D).

Validating the 109 selected potential targets in GSE14520 dataset
Gene profiles of 50 paired normal adjacent liver tissues and hepatocellular carcinoma tissues were extracted from GSE14520 dataset. Principal component analysis (PCA) showed that there was a clear distinction between normal adjacent liver tissues and hepatocellular carcinoma tissues ( Fig. 2A). Correlation coefficient heatmap analysis also showed that there was a significant difference of gene profiles between the two groups. Besides, Genetic variation in normal adjacent liver tissues was smaller than those in hepatocellular carcinoma tissues (Fig. 2B). Among these 109 selected potential targets, 3 selected potential targets (BRAF, TLR9, CDK3) were missing in the GSE14520 dataset. 9 (8.5%) selected potential targets were significantly upregulated in GSE14520 array, 14 (13.2%) selected potential targets were significantly downregulated in GSE14520 dataset, the rest had no significant difference (Fig. 2C).

PPI network analysis and functional annotations of the selected potential targets
According to PPI network analysis, 105 nodes were found, and the average number of neighbors (degree distribution) was 11.505 (Fig. 3A, Supplemental Table 1). Besides, the top list of hub nodes (Degree distribu-tion>11) were showed in Supplemental Table 2. When we merged these hub genes with differentially expressed genes in GSE14520 dataset and differentially expressed genes of hepatocellular carcinoma in GEPIA database, seven communal genes were found, they were AURKA, CDK1, CCNB1, TOP2A, CYP3A4, CYP2C9, and CYP2B6 (Fig. 3B). We then constructed the core-network of these seven hub genes and their neighbor genes ( Fig. 3C and D). Our results further showed that AURKA, CCNB1 and CDK1 involved in Oocyte meiosis and maturation via regulated FoxO signaling and/or p53 signaling. Besides, CCNB1 and CDK1 were also involved in virus-related carcinogenesis. TOP2A was mainly involved in platinum drug resistance (Fig. 3E). CYP3A4, CYP2C9, and CYP2B6 mainly involved in cytochrome P450 related metabolism under physiological or pathological situations (Fig. 3F).

Clinical correlation of selected genes in hepatocellular carcinoma
Transcriptional expression profile of hepatocellular carcinoma patients' tumor tissues and adjacent normal tissues was extracted from TCGA database. The expression of AURKA, CCNB1, CDK1, or TOP2A in each stage samples was higher than the expression of its normal samples. There was not significantly difference of expression of AURKA, CCNB1, CDK1, or TOP2A between any two different stage samples, except for the expression of CDK1 and TOP2A in stage III samples were slightly and significantly decreased while compared with stage I samples ( Fig. 4A-D). The expression of CYP2B6, CYP2C9, or CYP3A4 in each stage samples were less than the expression of its normal samples. There was not significantly difference of expression of CYP2B6, CYP2C9, or CYP3A4 between any two different stage samples (Fig. 4E-G). Besides, AURKA, CCNB1, CDK1 and TOP2A positively and significantly correlated with poor prognosis of hepatocellular carcinoma, while CYP2C9 and CYP3A4 negatively and significantly correlated with poor prognosis of hepatocellular carcinoma, but there was not significant correlation between the expression of CYP2B6 and prognosis of hepatocellular carcinoma (Fig. 4H-N). ROC curves showed that the AUC of AURKA, CCNB1, CDK1, TOP2A, CYP2B6, CYP2C9 and CYP3A4 were 95.7, 98.1, 96.9, 96.1, 96.0, 97.0 and 88.0%, respectively (Fig. 4O).

The effect of curcumin on regulating the target genes
We next investigated the effect of curcumin on regulating the target genes. Our results showed that curcumin dosedependently decreased cellular viability in HepG2.2.15 cells, and the cellular viability remained at least 95% when the concentration of curcumin was under or equal to 10 μM (Fig. 5A). The lower dose (from 0 μM to 10 μM) of curcumin was then used for the subsequent experiments. Compared to the vehicle treatment, curcumin significantly decreased mRNA expression of AURKA, CDK1, and TOP2A (Fig. 5B-D). Curcumin had no effect on transcription of CCNB1 (Fig. 5E). Besides, higher concentrations of curcumin significantly increased mRNA expression of CYP2B6, CYP2C9, and CYP3A4 ( Fig. 5F-H).

Discussion
Liver take predominantly advantage in multi-fundamental physiological processes, at least including material metabolism, digestion, detoxification, and coagulation. Carcinoma in liver not only defected its bio-functions, but also attenuated clinical treatments and amplified the toxic and side-effects [9]. As described above, accumulating in vitro and in vivo studies have indicated that curcumin acted as a promising compound to exhibit multifunction in the prevention and treatment of liver cancer. We herein mainly focused and predicted the potential clinical effects of curcumin in hepatocellular carcinoma treatment. In the present study, we then screened and verified several novel and promising curcumin-target genes in hepatocellular carcinoma therapy via bioinformatics analysis approach.
Previous studies have showed that natural herb, at least including saffron, safranal, salvadora persica, ginger and their metabolites suppressed inflammatory, proliferative, and oxidative pathways, while triggered caspases activities and DNA instability to induced cytotoxicity and apoptosis in the liver cancer cells [42][43][44][45][46]. In the present study, seven common target genes were selected when we overlapped curcumin predicted targets, aberrant genes in initiation and progression of hepatocellular carcinoma, and differentially expressed genes of GSE14520 dataset. Previous studies showed that AURKA promoted proliferation and metastasis of hepatocellular carcinoma cells [47][48][49]. Besides, AURKA also involved in formation of secondary liver cancer [50,51]. Blocking CDK1/PDK1/β-Catenin signaling would inhibit proliferation and EMT of hepatocellular carcinoma cells [52]. Previous studies showed that CCNB1 was upregulated to promote proliferation of hepatocellular carcinoma via decreased its negative regulators, at least including RNA-binding motif protein 43 (RBM43), miR-199a-3p, or miR-144 [53][54][55]. So far as we know, only one study showed that synthetic resveratrol-curcumin hybrid compound 4c significantly decreased the expression AURKA, AURKB, and CCNB1 in MCF-7 cells, it also significantly inhibited proliferation of MCF-7 cells, A549 cells, and HepG2 cells [56]. Otherwise, accumulating studies have shown that curcumin and its analogs significantly downregulated the expression of CDK1 to induce cell cycle arrest in various human cancers, but not in hepatocellular carcinoma [57][58][59].
In the present study, our results showed that curcumin significantly decreased the transcription of AURKA and CDK1 in HBV-transfected HepG2.2.15 cells, but not CCNB1. Otherwise, previous studies reported the dose of curcumin-induced cytotoxicity in HepG2 cells seemed to be contradictory. Lee and his college reported that no cytotoxicity was observed in HepG2 cells when curcumin was not more than 20 μM for 48 h treatment [60]. But Soni and his college reported that 10 μM curcumin sufficiently and significantly decreased the cell viability of HepG2 cells [61]. In line with Lee' s result, our results showed that the lower dose (from 0 μM to 10 μM) curcumin has no significant effect on decreasing cellular viability of HepG2.2 cells. Taken these together, our results suggested that the lower dose curcumin treatment might not induced sufficient downregulation of AURKA or CDK1 to inhibit proliferation of hepatocellular carcinoma cells. Alternatively, curcumin-induced downregulation of AURKA and CDK1 might confer the other antitumor effects, but not inhibiting proliferation..
Previous studies showed that inhibition of AURKA would activate NF-κB signaling pathway to confer radio-resistance or chemoresistance in hepatocellular carcinoma and acute myeloid leukemia [62][63][64]. Blocking CDK1/PDK1/β-Catenin signaling would decrease stemness of cancer stem cells of hepatocellular carcinoma cells, and reverse sorafenib chemoresistance [52]. It's reported that TOP2A was elevated in doxorubicinresistant hepatocellular carcinoma cells, in line with this finding, the other group showed that TOP2A was elevated in doxorubicin-resistant hepatocellular carcinoma patients, TOP2A inhibitor etoposide would facilitate doxorubicin-induced cytotoxicity in primary cancer cells of hepatocellular carcinoma [65,66]. TOP2A also conferred platinum resistance in several human cancers [67,68]. In the present study, our results also showed that curcumin significantly decreased the transcription of TOP2A in HBV-transfected HepG2.2.15 cells. Accumulating studies have shown that curcumin exhibited therapeutic roles by facilitating the cytotoxicity of chemotherapeutic drugs and reversing their chemoresistance [69][70][71][72][73]. Thus, our results indicated that curcumin might decrease the expression of AURKA, CDK1 and TOP2A to reverse chemoresistance in hepatocellular carcinoma treatment. Cytochrome P450 (CYP) enzymes play important roles in endogenous and xenobiotic metabolism of liver, but their roles in tumorigenesis and progression remain as a complex context [74]. Previous studies have highlighted that most CYP members, such as CYP2C9 and CYP3A4, were defected in hepatocellular carcinoma initiation and progression [75,76]. CYP2C9 was mainly activated in the metabolism drugs, at least including the activation of cyclophosphamide and tamoxifen, and the clearance of idarubicin [74]. CYP3A4 was activated in the metabolism of procarcinogens, and contributed to the clearance of several chemotherapeutic agents, at least including cisplatin, etoposide or doxorubicin [74,77]. CYP2B6 also correlated with metabolism of procarcinogens, but it tended to involved in the metabolic activation of anticancer prodrugs [74]. In the present study, our results further showed that curcumin significantly upregulated the expression of CYP2B6, CYP2C9 and CYP3A4. Otherwise, Previous studies have highlighted that curcumin ameliorated side-effects of chemotherapeutic drugs and exhibited hepatoprotective effects [20][21][22]73]. Thus, the finding of us and the other group together suggested that it should pay more attention to curcumin-induced P450s on regulating the drug activation, inactivation or clearance while consumed with other certain anticancer drugs.

Conclusion
In summary, so far as we known, our results first showed that CDK1, TOP2A, CYP2C9, and CYP3A4 genes correlated to curcumin-related chemotherapy of hepatocellular carcinoma, more than just correlated to diagnosis and prognosis of hepatocellular carcinoma which had highlighted by the previous studies [76,78]. Our result also showed that AURKA served as a new diagnosis and prognosis of hepatocellular carcinoma, even as also a potential and novel curcumin-related therapeutic target for hepatocellular carcinoma. Taken these together, the present study provided a new insight and proposal for curcumin-related hepatocellular carcinoma therapies.