Introduction
Recent studies have revealed that cancer stem cells (CSCs) are a source of therapy resistance, disease recurrence, and metastasis to other organs (1–3). At least two types of CSCs, dormant (dCSC) and activated (aCSC), are involved in tumor homeostasis, which are in contrast to two types of stem cells, dormant and activated types in normal skin, intestine and the hematopoietic system (4). Our previous study indicated that CD13+CD90− dCSCs of hepatocellular carcinoma survive in hypoxic areas in marginal regions in liver after therapy (5). In CD13+/CD90− dCSCs, the occurrence of double-strand breaks (DSBs) in genomic DNA, a deleterious cellular event, and damage-induced repairs that are necessary for cellular survival (6), reduce after therapy presumably due to CD13/aminopeptidase N functioning as a scavenger of reactive oxygen species (ROS) (5) and partially due to error-prone repair such as non-homologous end-joining (NHEJ) (6,7). In a sharp contrast, CD13−/CD90+/− dCSCs are sensitive to therapeutic insults from chemotherapeutic agents, which is associated with ROS-induced cell death after chemotherapy (8); however, damage is typically repaired though high-fidelity, error-free homologous recombination (HR) (6,7). Thus, dCSCs may be a cause of accumulation of deleterious mutations and should be targeted in therapy in terms of complete eradication of malignant cells, although hibernation therapy (the induction of dormancy) may be a viable option dependent on the patient’s condition (7). Chemotherapy results in a shift from aCSCs to dCSCs and accumulation of dCSCs after treatment, and dormancy may function as a type of refuge for the survival of malignant cells. CD13 cells play a role in the inhibition of increase in ROS and the resultant suppression of cell death during the process of epithelial mesenchymal transition (EMT) of metastatic CSCs (9). The exposure to a CSC-specific inhibitor, ubenimex, resulted in considerable eradication of malignant cells in vivo, indicating an apparent benefit in the combination of conventional chemotherapy and a CSC-specific inhibitor.
Here, we performed transcriptome analysis for coding mRNAs and non-coding microRNAs (miRs) in CD13+/CD90− dCSCs. This study allowed us to identify several pathways, which play a role in fundamental mechanisms in above-mentioned potentially malignant phenotype, and provided further clues for identification of molecular targets in therapeutic approaches for dCSCs.
Materials and methods
Cell cultures
Cell lines were maintained in Dulbecco’s modified Eagle’s medium (DMEM; Nacalai Tesque, Kyoto, Japan) supplemented with 10% fetal bovine serum (FBS) at 37°C in a 5% humidified CO2 atmosphere.
Flow cytometric analysis and cell sorting
The antibodies used were purchased from BD Biosciences (Tokyo, Japan). In brief, cells were harvested with trypsin and EDTA. Doublet cells were eliminated using FSC-A/FSC-H and SSC-A/SSC-H. Dead and dying cells were eliminated with 7-AAD (BD Pharmingen, San Jose, CA, USA). Isotype controls (BD Biosciences) were used. FcR blocking was performed using an FcR blocking reagent (Miltenyi Biotec, Bergisch Gladbach, Germany).
RNA
Total-RNA was extracted using TRIzol reagent (Invitrogen/Life Technologies Japan, Tokyo, Japan). Reverse transcription was performed with SuperScript III reverse transcription kit (Invitrogen). qPCR was performed using the LightCycler TaqMan Master Kit (Roche Diagnostics, Tokyo, Japan) for cDNA amplification of target-specific genes. Purified cDNA from mouse ES cells was used as a positive control for target genes. The expression of mRNA copies was normalized to GAPDH (for mRNA) or RNU48 (for miR) expression, as indicated. The RNA samples were analyzed using SurePrint G3 Human GE 8×60K Microarray and the Human miRNA Microarray 8×15K Rel.12.0 (Takara, Kyoto, Japan).
Statistical analysis
For continuous variables, results are expressed as means ± SE. The relationship between the gene expression level and cell count was analyzed by chi-square and Wilcoxon rank tests. All data were analyzed using JMP software (SAS Institute, Cary, NC, USA). P-values of <0.05 were considered statistically significant.
Results and Discussion
A study of CD13<sup>+</sup>/CD90<sup>−</sup> as dCSCs
CD13+/aminopeptidase N is expressed in liver CSCs (5), where it is involved in the reduction of ROS through the glutathione reductase pathway. Considering that another independent study has shown CD90+ as a candidate stem marker, critical to tumorigenicity in mice in vivo and clinical outcomes of patients (10), we indicated that CD13+/CD90− cells exist in dormant phase of cell cycle, whereas CD13+/CD90+ cells are predominantly in the S phase and CD13−/CD90+ cells are in the G2/M phase (5). In previous studies, we have elucidated that following exposure to genotoxic insults, such as chemotherapy or radiation therapy, CD13− cells shift to the CD13+ fraction in dormant phase of cell cycle. In dCSCs, double stranded breaks (DSBs) are repaired predominantly through the error-prone NHEJ mechanism (6,7). In sharp contrast, the high-fidelity HR-type repair proteins are increased in non-dormant CSCs compared with NHEJ proteins, of which cells are usually sensitized through chemoradiation therapy (6,7). Thus, after chemoradiation therapy, NHEJ supposedly contributes to the generation of misrepair after DSBs, which may cause chromosomal deletions, insertions, or translocations, and subsequent genomic instability (11). Such genomic alterations lead to the inactivation of tumor suppressor genes and activation of oncogenes, which become more apparent during tumor development of primary lesions, recurrence and metastasis (12). Nevertheless, there remains an important issue to be addressed, i.e., the identification of molecular mechanisms fundamental for initiation and development of tumor tissues composed of stem cell hierarchy, and moreover, the type of mechanism involved in CSC-based heterogeneous tumors. We began with transcriptome assessment, i.e., the expression of mRNAs and miRs and their association with CD13+/CD90− cells in supporting or maintaining CSC survival in the absence of genotoxic stimuli, which may be beneficial in the study of the basal situation and may help understand the differences between therapy-resistant clones and de novo tumor-initiating cells.
As shown in Fig. 1, we separated liver cancer cells by FACS sorting into CD13+CD90− dCSCs from other non-CSCs. Considering that miRs play a role in the regulation of mRNA in its stability and translation as an inhibitory regulation system, we focused on increased expression of mRNA clones and decreased expression of miR clones. The data of high-density array screening indicated 17 clones of increased expression in CD13+CD90− populations compared with unsorted cells with more than 2-fold significant increase. The data were almost consistent in CD13+/CD90− populations compared with non-CD13+/CD90− cells (Table I). Next, we analyzed miR expression and successfully isolated nine miRs in CD13+/CD90− population compared with unsorted cells with more than 4-fold significant decrease. The data were almost consistent in CD13+/CD90− populations compared with non-CD13+/CD90− cells (Table II).
Identification of regulatory networks
By assessment of pairs of miRs and its putative target mRNAs using prediction software (http://www.targetscan.org/; http://www.microrna.org/microrna/home.do), we confirmed the data of the array by quantitative PCR. As shown in the representative data, the expression of miR-101 was downregulated in CD13+/CD90− cells compared with non-CD13+/CD90− cells or CD13−/CD90− cells; in sharp contrast, the expression of putative targets, EZH2 (enhancer of zeste homolog 2; http://www.genecards.org/cgi-bin/carddisp.pl?gene=EZH2&search=EZH2), JARID1A; (http://www.genecards.org/cgi-bin/carddisp.pl?gene=KDM5A&search=JARID1A), and JMJD1B (http://www.genecards.org/cgi-bin/carddisp.pl?gene=KDM3B&search=JMJD1B) were increased (Fig. 2; summarized in Fig. 3). CD13 mRNA expression was increased in CD13+/CD90− cells, but not in CD13−/CD90+, CD13−/CD90−, or unsorted cells. The expression of interleukin-8 (IL-8; http://www.genecards.org/cgi-bin/carddisp.pl?gene=IL8&search=Interleukin+8) and endothelin 1 (EDN1; http://www.genecards.org/cgi-bin/carddisp.pl?gene=EDN1&search=Endothelin+1) was increased in CD13+/CD90− cells, but not in others (Fig. 3; summarized in Fig. 3). Through this study, we were able to find at least three core regulatory networks for the maintenance of dormant CD13+/CD90− cells, but not in others, i.e., miR-182 pathways, miR-101 pathways, cytokines/chemokines pathways (IL-8, CXCL5 [http://www.genecards.org/cgi-bin/carddisp.pl?gene=CXCL5&search=CXCL5], CCL-20 [http://www.genecards.org/cgi-bin/carddisp.pl?gene=CCL20&search=CCL-20]), growth factors and their receptor pathways [EDN1, Wnt2 (http://www.genecards.org/cgi-bin/carddisp.pl?gene=WNT2&search=Wnt2), Wnt7B (http://www.gene-cards.org/cgi-bin/carddisp.pl?gene=WNT7B&search=Wnt7B), and TNF (http://www.genecards.org/cgi-bin/carddisp.pl?gene=TNF&search=TNF)]. As noted in this study, we did not detect any apparent involvement in DNA damage response machineries, except for an association underlined by TP53INP1 (http://www.genecards.org/cgi-bin/carddisp.pl?gene=TP53INP1&search=TP53INP1) and BRCA1 (http://www.genecards.org/cgi-bin/carddisp.pl?gene=BRCA1&search=BRCA1), suggesting that (1) the damage response in dCSCs is characteristic after exposure to genotoxic stimuli, whereas in the absence of damage insults, they are not apparent and (2) the DNA damage response was regulated mainly by the modification of proteins such as phosphorylation or ubiquitination, and the expression array technology was less sensitive to pathway detection and other networks may have been missed.
Significance of regulatory networks (<xref rid="f4-ijo-41-03-0979" ref-type="fig">Fig. 4</xref>)
The EZH2 gene encodes a member of the Polycomb group (PcG) family, which forms multimeric protein complexes involved in maintaining a transcriptionally repressive state of genes over successive cellular generations (13). Reportedly, the genomic loss of miR-101, a putative tumor suppressor, leads to overexpression of histone methyltransferase EZH2 in cancer (14,15), hypoxia, and androgen-dependent conditions (16) as well as in gastric cancer (17), pancreatic cancer (18), lung cancer (19) glioblastoma (20) and invasive squamous cell carcinoma (21). Thus, PcG proteins are critical epigenetic mediators of stem cell pluripotency and CSC functions, which may be implicated in human cancer pathogenesis, probably indicating candidacy for novel pharmacological targets of cancer therapy. Recently, it was reported that the administration of diflourinated-curcumin (CDF), a novel analogue of the turmeric spice component curcumin, has antioxidant properties and inhibits tumor growth through reduced expression of EZH2, Notch-1, CD44, EpCAM, and Nanog and increased expression of let-7, miR-26a, and miR-101 (18). These findings indicated that CDF inhibited CSC growth by targeting an EZH2-miRNA regulatory circuit for epigenetically controlled gene expression. In the present study, we identified various miR-101 targets, including JARID1A, JMJD1B, TP53INP1, and EZH2, suggesting that these target molecules act together to maintain CSC dormancy, and proposed the possible significance of the miR-101 pathway. We also identified the miR-182 pathway, in which miR-182-mediated downregulation of BRCA1 impacts DNA repair and sensitivity to poly-ADP ribose polymerase (PARP) inhibitors (22). The overexpression of miR-182 was reported in high-grade ovarian papillary serous carcinoma (23). Reportedly, aberrant miR-182 expression promotes melanoma metastasis by repressing FOXO3 (24). Taken together with our study, the miR-182 pathway may be involved in the maintenance of CSC function in a similar manner. As summarized in Fig. 4, we identified other cytokines, chemokines, growth factors, receptors and pathways. These findings may facilitate further studies on the regulatory mechanisms of the dormant phase of gastrointestinal CSCs.