OPEN The Pharmacogenomics Journal (2018) 18, 23–28 www.nature.com/tpj ORIGINAL ARTICLE Epigenomics alternations and dynamic transcriptional changes in responses to 5-fluorouracil stimulation reveal mechanisms of acquired drug resistance of colorectal cancer cells Y Shen1,3, M Tong1,3, Q Liang1, Y Guo2, HQ Sun1, W Zheng1,LAo1, Z Guo1 and F She1 A drug-induced resistant cancer cell is different from its parent cell in transcriptional response to drug treatment. The distinct transcriptional response pattern of a drug-induced resistant cancer cell to drug treatment might be introduced by acquired DNA methylation aberration in the cell exposing to sustained drug stimulation. In this study, we performed both transcriptional and DNA methylation profiles of the HCT-8 wild-type cells (HCT-8/WT) for human colorectal cancer (CRC) and the 5-fluorouracil (5-FU)- induced resistant cells (HCT-8/5-FU) after treatment with 5-FU for 0, 24 and 48 h. Integrated analysis of transcriptional and DNA methylation profiles showed that genes with promoter hypermethylation and concordant expression silencing in the HCT-8/5-FU cells are mainly involved in pathways of pyrimidine metabolism and drug metabolism-cytochrome P450. Transcriptional analysis confirmed that genes with transcriptional differences between a drug-induced resistant cell and its parent cell after drug treatment for a certain time, rather than their primary transcriptional differences, are more likely to be involved in drug resistance. Specifically, transcriptional differences between the drug-induced resistant cells and parental cells after drug treatment for 24 h were significantly consistent with the differentially expressed genes (termed as CRG5-FU) between the tissues of nonresponders and responders of CRCs to 5-FU-based therapy and the consistence increased after drug treatment for 48 h (binomial test, P-value = 1.88E − 06). This study reveals a major epigenetic mechanism inducing the HCT-8/WT cells to acquire resistance to 5-FU and suggests an appropriate time interval (24–48 h) of 5-FU exposure for identifying clinically relevant drug resistance signatures from drug-induced resistant cell models. The Pharmacogenomics Journal (2018) 18, 23–28; doi:10.1038/tpj.2016.91; published online 3 January 2017 INTRODUCTION cells synchronously exposing to drug for a certain time, are more Colon and rectal cancer (CRC) is a fatal cancer with low overall likely to be involved in drug resistance. This result suggests a 5-year survival rate.1 Until now, 5-fluorouracil (5-FU)-based novel strategy to identify clinically relevant drug resistance genes 10 chemotherapy is the first-line clinical treatment of CRC,2 but the from drug-induced resistant cell models. However, the cells response rate is only ∼ 50%,3 and most of patients acquire analyzed in our previous study were treated with 5-FU for only 6, resistance during chemotherapy.4,5,6 The process of a tumor 12 and 24 h. Thus, it is necessary to prolong the time of drug treatment to investigate the characteristics of sustained acquiring drug resistance during chemotherapy could be 13,14 mimicked partially by the common process of establishing drug- responses. Another problem remains to be addressed is to explain the underlying molecular mechanism for the acquired resistant cells by exposing parental cells to drug treatment in vitro distinct transcriptional response characteristics of a drug-induced over a period of time. Therefore, it should be helpful to explore resistant cancer cell to drug treatment. It has been reported that drug-induced resistant cell models to understand both innate and aberrant DNA hypermethylation within gene promoters and acquired drug resistance mechanisms of cancers. consequent gene silencing might play a major role in generating Based on a drug-induced resistant cell model, researchers often 15–17 fi drug-resistant phenotypes. Therefore, we could assume that rst identify differentially expressed genes (DEGs) between the distinct transcriptional response characteristics of a drug- parental and resistant cells, termed basally deregulated (BD) induced resistant cancer cell to drug treatment could be 7–9 fi genes, and de ned these DEGs as drug resistance genes. introduced by acquired DNA methylation aberration in the cell However, it has been recognized that most of these BD genes exposing to sustained drug stimulation. might simply represent drug-induced transcriptional changes In this work, we performed both transcriptional and DNA 10–12 irrelevant to drug resistance. Recently, taking the DEGs methylational profiles of HCT-8 wild type cells (HCT-8/WT) for between the nonresponders and responders of CRC patients human CRC and its 5-FU-induced resistant cell line (HCT-8/5-FU). receiving 5-FU-based therapy (termed CRG5-FU) as reference, we We first uncovered that the genes with both promoter hyper- reported that inducible difference (ID) genes, which represent the methylation and expression silencing during 5-FU treatment are transcriptional differences between resistant cells and parental mainly involved in pathways of pyrimidine metabolism and drug 1Department of Bioinformatics, Key Laboratory of Ministry of Education for Gastrointestinal Cancer, Fujian Medical University, Fuzhou, China and 2Department of Preventive Medicine, School of Basic Medicine Sciences, Gannan Medical University, Ganzhou, China. Correspondence: Professor Z Guo or Professor F She, Department of Bioinformatic, Key Laboratory of Ministry of Education for Gastrointestinal Cancer, Fujian Medical University, Fuzhou, Fujian 350108, China. E-mail: [email protected] or [email protected] 3These two authors contributed equally to this work. Received 20 July 2016; revised 6 November 2016; accepted 14 November 2016; published online 3 January 2017 Acquired resistance mechanism of 5-fluorouracil Y Shen et al 24 metabolism-cytochrome P450 that might be a major epigenetic CA 92122, USA). A total amount of 1 mg genomic DNA from one mechanism of a cancer cell to acquire 5FU resistance. Then, we experiment was bisulfite modified using the EZ DNA Methylation Kit (Zymo confirmed that ID genes detected at the 24 h time point were Research, Irvine, CA, USA). In all, 200 ng of converted DNA was further significantly consistent with CRG and the consistency became processed to run BeadArrays according to the manufacturer’s instructions. 5-FU fl more significant after drug treatment for 48 h (binomial test, Each locus is represented by uorescent signals from two bead types corresponding to the methylated (M) and unmethylated (U) alleles, P-value = 1.88E − 06). This result suggested an appropriate time – respectively. interval (24 48 h) of drug treatment for identifying genes related Methylation and expression profiling data were available in NCBI to drug resistance. (National Center for Biotechnology Information) Gene Expression Omnibus (GEO accession number: GSE81008). MATERIALS AND METHODS Data processing Cell lines and reagents Gene expression profiling data were preprocessed with the Robust HCT-8/WT and HCT-8/5-FU cells were purchased from Keygentec (Nanjing, Microarray Average algorithm. Each probe set ID was mapped to its China). Complete growth medium RPMI-1640 that contains 10% fetal calf Entrez gene ID with the corresponding custom CDF files. If multiple probe – serum and 1% penicillin streptomycin solution was used for cell culture. Both sets were mapped to the same gene, the expression value for the gene fi cell lines were maintained at 37 °C in a humidi ed atmosphere containing was defined as the arithmetic mean of the values of the multiple probe 5% CO . MTT assay was performed to determine the 5-FU resistance in both 2 sets (on the log2 scale). of HCT-8/WT and HCT-8/5-FU cell lines. 5-FU and penicillin–streptomycin The raw signals of unmethylated (U) and methylated (M) bead types solution (100 × ) were purchased from Sigma-Aldrich Corp. St. Louis, MO, USA. were background corrected and computed into β-values using the RPMI-1640 and fetal calf serum were purchased from Thermo Fisher GenomeStudio software Methylation Module (5200 Illumina Way, San Scientific, 168 Third Avenue, Waltham, MA, USA. Cell culture plates and other Diego, CA 92122, USA). A total of 25 943 CpG loci located within the consumables were purchased from Corning and Axygen, 836 North Street, proximal promoter regions of the transcription start sites of 14 109 genes Building 300 Suite 3401, Tewksbury, MA 01876, USA. were used as background CpG loci in this study. The β-value was used to estimate the methylation level of the CpG locus using the ratio of Gene expression and methylation profiling intensities between methylated and unmethylated alleles: ðÞ; HCT-8/WT and HCT-8/5-FU cells were cultured in 5-FU free medium for β ¼ max M 0 48 h, and then both cell lines were transferred to the medium for 0, 24 and ðmaxðÞþM; 0 maxðÞþU; 0 100Þ 48 h with HCT-8/WT half-maximal inhibitory concentration doses of 5-FU (Figure 1). Untreated cells (0 h) and treated cells (24 and 48 h) were subject to extract RNA to be subjected to RNA microarray analysis. Total RNA was isolated from three independent experiments using TRizol (Thermofisher) Concordance score ’ according to the manufacturer s instructions. Before the following For DEGs from two independent data sets sharing k DEGs, of which s genes experiments, total RNA was quality controlled by formaldehyde denatured have the same up- or down-regulation directions between the treated cell – agarose gel electrophoresis and ultraviolet visible spectrophotometer. type and the control cell type, the concordance score was calculated as s/k. Total RNA (100 ng) from one experiment was sent to commercial kits for If k genes have both methylation and expression changes, among which s complementary DNA synthesis, complementary RNA synthesis, antisense genes are hypermethylated (or hypomethylated) and correspondingly RNA synthesis and its fragmentation (1792, Ambion, Thermo Fisher downregulated (or upregulated), then the concordance score was Scientific, 168 Third Avenue, Waltham, MA, USA).