Bioinformatics analysis identified CDC20 as a potential drug target for cholangiocarcinoma

Cell culture

Human CCA cell lines derived from CCA patient tissues that included KKU-213A (poorly differentiated squamous cell carcinoma), KKU-213B (well differentiated squamous carcinoma) (Sripa et al., 2020), KKU-100 (poorly differentiated adenocarcinoma) (Sripa et al., 2005), and KKU-452 (poorly differentiated adenocarcinoma) (Saensa-ard et al., 2017) were cultured in Dulbecco’s Modified Eagle’s Medium (DMEM)(Gibco life technologies Corporation, Grand Island, NY) with 25 mM glucose supplemented with 1% antimycotic-antibiotic and 10% heat-inactivated fetal bovine serum at 37 °C, 5% CO₂.

For the 3D cell culture, the adherent CCA cells were trypsinized and seeded into a 96 well ultra-low attachment plates (Corning, Incorporated, Kennebunk, Maine) at the optimal starting cell numbers. CCA cells were then cultured at 37 °C, 5% CO₂ for 24–72 h for spheroid formation. Spheroid morphology was observed under an inverted light microscope. Spheroid integrity were analyzed by ImageJ NIH software (https://imagej.nih.gov/ij/download.html).

Microarray data sources

Three gene expression datasets of CCA and normal tissues were retrieved from the GEO database (https://www.ncbi.nlm.nih.gov/geo/) including GSE26566, GSE32225 and GSE76297. GSE26566 based on GPL6104 Illumina Humanref-8 v2.0 expression bead chip platforms, GSE32225 based on GPL8432 Illumina Humanref-8 WG-DASL v3.0 platforms and GSE76297 based on GPL17586 [HTA-2_0] Affymetrix Human Transcriptome Array 2.0 [transcript (gene) version].

Differentially expressed gene (DEGs) analysis

The DEGs between cancer and normal tissues were analyzed using the GEO2R online tool (https://www.ncbi.nlm.nih.gov/geo/geo2r/). Genes that met the cutoff criteria, adjusted P-value <0.05 and —logFC—≥ 0.5, were considered DEGs. A volcano plot was performed for data visualization using RStudio (https://rstudio.com/). Moreover, Venn diagrams were plotted (http://bioinformatics.psb.ugent.be/webtools/Venn/) for elucidating common DEGs among three microarray datasets.

Gene ontology, pathway enrichment, protein-protein interaction (PPI) and hub gene identification analysis

The common DEGs were inputted into STRING software version 11.0 (https://string-db.org/) for analyzing the interactions among various proteins to constitute a PPI network, gene ontology and pathway enrichment. Gene functions were classified into cellular components (CC), cellular processes (CP), molecular function (MF) and the Reactome pathway. GO results were visualized as a bubble plot by RStudio software. Screening and identification of hub genes were conducted with cytohubba (https://apps.cytoscape.org/apps/cytohubba) (Chin et al., 2014) of the Cytoscape (https://cytoscape.org/) plug-in MCC model. 

Prediction and prioritizing of novel candidate anti-cancer drugs

After hub gene identifications, the next step was further prioritizing and predicting a novel anti-cancer drug based on up-regulated hub genes using PanDrugs software (https://www.pandrugs.org) (Piñeiro Yáñez et al., 2018). Prioritization criteria for the selection of a novel anti-tumor drug was that the candidate drug had to have been used in clinical trials or is already approved for clinical use.

RNA extraction and real-time PCR

Cells were collected and total RNA samples were extracted from CCA cell lines using TRIzol^® reagent (Invitrogen; Thermo Fisher Scientific, Van Allen Way, California.) according to the manufacturer’s protocol. RNA (2 µg) was reverse transcribed to cDNA. The system conditions for real-time PCR contained (10 µl): 5 µl for 2X SYBR Green I Master LightCycler^®480 SYBR green I master (Roche Diagnostics GmbH, Mannheim, Germany), 2 µl of cDNA (25 ng/µl), 2 µl of forward and reverse primers and 1 µl DD H₂O PCR grade. The primers for CDC20 sequences were as follows: forward, 5′-CGGGTAGCAGAACACCATGT-3′reverse, 5′-ACTGGCCAAATGTCGTCCAT-3′. The PCR conditions were 95° for 5 min, followed by 50 cycles at 95° for 10 s and 62° for 10 s.

Western blot analysis

Cells were lysed with RIPA buffer (Thermo Fisher Scientific, Rockford, Illinois) containing Proteinase Inhibitor Cocktail (NACALAI TESQUE, INC., Kyoto, JAPAN). The lysates were shaken at 4 °C overnight. Protein content was determined using BCA assay kit (Thermo Fisher Scientific, Rockford, Illinois). Protein lysates were separated in a 12% (for actin) or 15% (for CDC20) sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and transferred on a polyvinylidene fluoride or polyvinylidene difluoride (PVDF) membrane (Merck KGaA, Darmstadt, Germany) using Towbin transfer buffer in a wet tank transfer blotting system (Bio-Rad Laboratories (Shanghai) Co., Ltd., Shanghai, China). After blocking with 5% skimmed milk for 1 h at room temperature (RT), membranes were incubated with each primary antibody, CDC20 (Cell Signaling Technology, Tokyo, Japan), or actin (Santa Cruz Biotechnology, Inc., Dallas, Texas) at 4 °C, overnight. Blots were then incubated with anti-rabbit IgG, HRP-linked Antibody (Cell Signaling Technology, Tokyo, Japan) or anti-mouse IgG, HRP-linked Antibody (Cell Signaling), at RT for 1 h. The western blot bands were detected using an ImageQuant LAS 4000 (GE Healthcare Bio-Sciences AB, Uppsala, Sweden).

The cytotoxicity test in two-dimensional (2D) and three-dimensional (3D) cell culture models

For the 2D cell culture model, CCA cell lines were seeded into 96-well plates (1,500 cells/well) overnight (16 h) and then treated with various concentrations of gemcitabine (Sigma-Aldrich, St. Louis, Missouri) or dinaciclib (Abcam plc., Cambridge, UK) for 72 h. After incubation, cell viability was measured by ATPlite 1step (PerkinElmer, inc, Waltham, Massachusetts) for luminescence detection of ATP from cultured cells under the 2D model using SpectraMax L Microplate Reader (Molecular Devices, St, San Jose, California).

For the 3D cell culture model, 2,500 cells/well of CCA cell lines were seeded into 96 well ultra-low attachment flat-bottom plates (Corning, Incorporated, Kennebunk, Maine), and incubated for 24 to 78 h depending on cell type. The formed spheroids were then exposed to gemcitabine or dinaciclib for 72 h. Cell viability was measured by the ATPlite3D kit (PerkinElmer, inc, Waltham, Massachusetts) using SpectraMax L Microplate Reader. Cells cultured in DMSO; 0.01% for 2D-culture and 0.4% for 3-D culture, were used as a vehicle control in all drug testing experiments. Three separated experiments with triplicate assays were performed.

Statistical analysis

Statistical analyses were conducted using Graphpad Prism software version 7.03 (GraphPad Software, Inc., San Diego, California). All experiments were performed in three independent experiments. The cytotoxicity results are expressed as mean ± SD. Statistical significance was determined by Students’ t test and P < 0.05 was noted as statistical significance.

Identification of common DEGs

Four GSE datasets of CCA were retrieved form GEO including GSE26566, GSE32225, GSE76297 and GSE89749. Of these, GSE89749 which comprised of a small number of normal cases and showed the abnormality of gene distribution was excluded from the study. The GSE26566 contained 104 cancer and 59 normal cases, while GSE32225 comprised of 149 cancer and 6 normal cases. The CCA cases that formed both datasets were non-Opisthorchis viverrini (OV)-associated CCA. GSE76297 datasets consisted of 91 OV-associated CCA and 92 surrounding liver tissues. The volcano plot elucidated DEGs of CCA and normal tissues (Figs. 1A–1C). Based on the selection criteria of an adjusted P value <0.05 and absolute log fold-changes ≥ 0.5, a total of 2,029, 4,153 and 4,212 DEGs were identified from GSE26566, GSE32225 and GSE76297. 1,152 and 877 genes were up- and down-regulated in the GSE26566 dataset. 2,165 and 1,988 genes were up- and down-regulated in GSE32225. 2,407 upregulated and 1,805 down-regulated genes were identified in GSE76297 dataset. Finally, Venn diagram analysis revealed that there are 226 common DEGs in the three microarray datasets, in which 124 were up-regulated and 102 down-regulated genes (Figs. 1D and 1E).

Gene ontology and pathway enrichment analysis

The GO and pathway enrichment analysis of common DEGs elucidated in cellular component (CC) terms (Fig. 2A), common DEGs were mainly enriched in cytoplasm, extracellular space, high-density lipoprotein particles and in the peroxisomal matrix. The common DEGs were also involved with several molecular functions (MF) (Fig. 2B) such as oxidoreductase activity, endopeptidase inhibitor activity and enzyme inhibitor activity. In cellular process terms (Fig. 2C), the results showed that common DEGs were significantly enriched in carboxylic acid catabolic processes, whereas small molecule metabolic processes, were responsive to inorganic substances and organic acid metabolic processes. Moreover, the Reactome pathway enrichment analysis indicated that common DEGs were mainly enriched in metabolism, metallothionein bound metals and cell cycles, especially mitosis (Fig. 2D).

PPI of common DEGs and hub gene identifications

The PPIs of 226 common DEGs were constructed using STRING software as shown in (Fig. 3A). The hub genes based on the PPI network were further prioritized (Figs. 3B and 3C). The results categorized the top 20 hub genes automatically into two groups. The up-regulated hub genes from common DEGs, including CCNB1, CDC20, MAD2L1, FANCI, CDKN3, RRM2, EZH2, MCM3, ANLN, MCM7 and HMMR were ranked from the degree of connectivity with other proteins (binding scores) (Table 1). Down-regulated hub genes included FGA, AHSG, SERPINA10, F9, F11, FETUB, C6, F12 and KNG1. Node colors are represented for the degree of connectivity. The pseudocolor scale from red to yellow represents the gene ranking from 1 to 20. The dark red color represents the highest degree while an orange color stands for the intermediate degree and yellow color is the lowest degree. Most of the up-regulated hub genes were involved with cell division and cell cycle control, nucleotide metabolism and cell movement.

To confirm the reliability of the analyzed results, the expression levels of all up-regulated hub genes using different datasets and different databases were further examined. RNA sequencing profiles were retrieved from The Cancer Genome Atlas (TCGA, https://www.cancer.gov/about-nci/organization/ccg/research/structural-genomics/tcga). Box plots were plotted for the elucidation expression level of the hub genes in patient CCA tissues compared with the normal tissues (Figs. 4 and 5A). The results elucidated strong evidence to support the analysis that all the identified hub genes were highly expressed in CCA when compared with normal controls even though the analysis was performed on different databases and different platforms.

Prioritizing and prediction of a novel targeted drug for CCA

Novel candidate drugs targeted for CCA were based on 11 up-regulated hub genes using PanDrugs, a bioinformatics platform, to prioritize anticancer drug treatments. PanDrugs not only provided drug status descriptions but it also predicted the possible drug response and the interaction between drugs. The drug prediction and prioritizing in Table 2 shows several potential novel drugs for 6 of the 11 upregulated hub genes (Fig. 3B). CDC20 was the candidate that fit to the selection criteria (Fig. S1) as CDC20 plays significant roles in cell cycle and mitosis, and was the high-scoring hub gene (Fig. 3B). There are several candidate drugs suggested for CDC20 (Table 2). Dinaciclib, however, was selected to be the drug of choice for CDC20 in this study as it has never been reported for CCA, and the safety and efficacy has been reported in the clinical trial phase III (Ghia et al., 2015; Sharp & Corp, 2017).

CDC20 is a candidate novel target for CCA

After the in-silico analysis, a series of experiments for investigating the antitumor activity of dinaciclib in CCA cell lines under 2D and 3D cell culture models was performed. CDC20 mRNA and protein were investigated in four CCA cell lines, KKU-100, KKU-213A, KKU-213B and KKU-452, using real-time PCR and western blot analysis. Expression of CDC20 in patient CCA tissues was significantly higher than those of normal counterpart (Fig. 5A), and all CCA cell lines differentially expressed CDC20 mRNA (Fig. S2 and File S1) and protein (Figs. 5B and 5C). The anti-tumor activity of dinaciclib in CCA cell lines was compared with that of gemcitabine, the standard chemotherapeutic agent for CCA treatment. In the 2D cell culture model (Figs. 5D and 5E), dinaciclib elucidated very effective anti-tumor activity against CCA cell lines at approximately 40 times of the 50% inhibitory concentrations (IC₅₀) lower than gemcitabine (Table 3). In the 3D cell culture model (Figs. 5F and 5G), KKU213A and KKU-452 spheroids were treated with various concentrations of dinaciclib or gemcitabine for 72 h. The results demonstrated that dinaciclib had highly effective antitumor activity better than a thousand times gemcitabine in both CCA cell lines (Table 3). The IC₅₀ of KKU-213A and KKU-452 spheroids to dinaciclib were 514.8 nM and 4.7 nM, whereas the IC₅₀ of gemcitabine against KKU-213A spheroid cannot be calculated because it was higher than 500 ×10³ nM, the maximum tested drug concentration. A similar trend was observed for KKU-452 spheroids with gemcitabine IC₅₀ is 25.78 ×10³ nM. The KKU-213 spheroid morphology at 72 h exposure time with both drugs and vehicle were demonstrated in Fig. 5H. It was then concluded that the CCA cells were more sensitive to dinaciclib than gemcitabine.