Cancer stem cell–associated miRNAs serve as prognostic biomarkers in colorectal cancer

CD44v6 expression is upregulated in colorectal CSCs, and mediates increased chemoresistance.

One of the primary reasons for the poor survival in patients with CRC is the primary or secondary resistance to chemotherapeutic drugs; a process that is intimately linked to the existence of CSCs (5, 6). Because CD44v6 was recently proposed to be CRC stem cell marker (32), we examined whether there were differences in chemotherapeutic resistance in parental CRC cell lines with regard to enriched cells with a higher stem cell fraction. Accordingly, we cultured CRC cell lines as spheroids — a well-established approach for the enrichment of CSC-like cells (33, 34). To ensure representation from both microsatellite-stable (MSS) and microsatellite-unstable (MSI) phenotypes, we generated SDCSCs from HCT116 (MSI) and HT29 (MSS) cell lines (Figure 2A). The expression of the stem cell markers OCT4 and Nanog was significantly increased in SDCSCs versus respective parental cell lines (all P < 0.05, Figure 2B). Furthermore, the expression of 2 major stemness-suppressive miRNAs, miR-34a and miR-200c (35, 36), was significantly downregulated in SDCSCs (all P < 0.05), demonstrating that these cells have altered epigenetic profiles, which encompasses the inhibition of stemness-suppressive miRNAs (Figure 2C). Collectively, based on the expression of these molecular markers, compared with parental cells, SDCSCs appear to have an increased stem cell population.

Open in a separate window

Figure 2

SDCSCs have high CD44v6 expression and increased tolerance to chemotherapeutic drugs.

(A) Images of sphere-derived cancer stem cells (SDCSCs) and their respective parental cells. (B) Gene expression of pluripotency markers Oct4 and Nanog in SDCSCs and their parental cell lines (n = 3, Mann-Whitney’s U test). (C) Expression of the stemness-suppressive miRNAs miR-34a and miR-200c in SDCSCs and their parental cell lines (n = 3, Mann-Whitney U test). (D) Cytotoxicity from various doses of 5FU (top 2 panels) and oxaliplatin (bottom 2 panels) was assessed for SDCSCs and their respective parental cells lines, HCT116 and HT29. (E) Relative gene expression of CD44 variants in SDCSCs and their parental cell lines (n = 3, Mann-Whitney U test). (F) CD44v6 protein expression in SDCSCs and their parental cell lines, HCT116 and HT29. (G) Immunofluorescence assay of CD44v6 in SDCSCs from HCT116 cells with FITC staining (left) and DAPI nuclear staining (middle); a merged image is shown on the right. *P < 0.05, **P < 0.01, ***P < 0.001.

Next, we interrogated whether SDCSCs exhibit enhanced resistance to commonly used chemotherapeutic agents (5-fluorouracil [5FU] and oxaliplatin). The cytotoxicity of both chemotherapeutic agents was significantly attenuated in SDCSCs compared with that in the parental cells (all P < 0.05, Figure 2D), validating a higher tolerance of CSCs to cytotoxic chemotherapy. To examine the CD44 variants that are overexpressed in CSCs, we compared the expression of CD44 variants in SDCSCs concurrently with their respective parental cells. We observed that CD44v6 transcripts were consistently overexpressed in both cell types (both P < 0.05, Figure 2E), which was subsequently confirmed by corresponding changes in protein expression in Western immunoblotting and immunofluorescence experiments (Figure 2, F and G). Another key feature of CSCs is their high capacity to initiate tumor formation; hence, we used a xenotransplantation assay to evaluate their tumor-initiating capacity. While at higher concentrations of cells, differences between SDCSCs and parental cells were negligible, upon implantation of 1 × 10³ cells, SDCSCs generated a markedly greater number of xenograft tumors than the parental cells, confirming their higher tumor-initiating capacity (Supplemental Table 2; supplemental material available online with this article; https://doi.org/10.1172/jci.insight.125294DS1). Taken together, these findings highlight that CSCs express high levels of CD44v6, possess enhanced resistance to chemotherapeutic agents, and have an increased ability to form tumors.

While spheroids provide a useful model for studying CSCs, their inability to grow efficiently as a monolayer limits their use for functional evaluations. Therefore, to investigate the link between CSCs and chemoresistance, we generated 2 5FU-resistant (5FUR) cell lines from HCT116 and SW480 by continuously culturing these cells with increasing concentrations of 5FU over 9 months (37) (Supplemental Figure 1A). We found that 5FUR cells morphologically resembled cells with a mesenchymal origin (Supplemental Figure 1B) and had a higher tolerance to 5FU (Supplemental Figure 1C). Considering that previous studies have shown that 5FUR CRC cells have a higher expression of CSC markers such as CD44 and CD133 (38, 39), we examined whether CD44 variants were also overexpressed in our chemoresistant cell lines. While mRNA of most variants was overexpressed in the chemoresistant cells, we observed significantly higher overexpression of CD44 variants 4 and 6 in the 5FUR cells (both P < 0.05, Supplemental Figure 1D). When we examined expression of CD44 variants in both 5FUR cells and SDCSCs, we found that CD44v6 was the most consistently overexpressed variant in both cell types (Supplemental Figure 1E). To confirm that CD44v6 protein was overexpressed in the 5FUR cell lines as well, Western blotting and immunofluorescence analyses were undertaken, which substantiated our observation of increased expression of this variant in 5FUR cells (Supplemental Figure 1, F and G).

Downregulation of CD44v6 results in decreased stemness and enhanced sensitivity to chemotherapeutic drugs.

Considering that CD44v6 appears to be involved in the development of drug resistance in CSCs, we examined its functional role in CRC cell lines by inhibiting its expression (Figure 3A) using a previously validated siRNA (40). We wanted to ascertain whether suppression of CD44v6 alters key oncogenic traits in CSCs, such as cellular proliferation or colony-forming capacity. The inhibition of CD44v6 resulted in significantly reduced cellular proliferation and colony-forming capacity in both HCT116 and HT29 cell lines (all P < 0.05, Figure 3, B and C). To determine whether CD44v6 modulates the formation and proliferation of CSCs, we examined the spheroid-forming capacity of CD44v6-knockdown cells versus the parental cells. We found that the inhibition of CD44v6 suppresses spheroid-forming capacity, supporting the role of CD44v6 in tumor initiation and CSC expansion (both P < 0.001, Figure 3D). Next, we examined whether the inhibition of CD44v6 alters sensitivity to the chemotherapeutic drugs, and observed that the suppression of CD44v6 resulted in increased sensitivity to 5FU and oxaliplatin in both HCT116 and HT29 cell lines (P < 0.05, parental versus CD44v6-knockdown for doses from 2.5 to 10 μM; Figure 3E). Collectively, these findings indicate that CD44v6 regulates response to chemotherapeutic agents, in part through the regulation of cancer stemness.

Open in a separate window

Figure 3

Suppression of CD44v6 results in reduced oncogenicity and resistance to chemotherapeutic agents.

(A) CD44v6 protein expression following transient knockdown (KD) in HT29 and HCT116 cell lines. NC, negative control. (B) Cell proliferation of CD44v6 KD cells and parental cell lines (n = 8 per time point, Mann-Whitney U test). (C) Clonogenicity of HCT116 and HT29 cells with and without KD of CD44v6 (n = 6, Mann-Whitney U test). (D) Spheroid formation capacity of HCT116 and HT29 cells with and without KD of CD44v6 (HCT116: n = 20, HT29: n = 8, Mann-Whitney U test). (E) Cytotoxicity of various doses of 5FU (left) and oxaliplatin (right) was assessed for HCT116 cells with and without KD of CD44v6. *P < 0.05, ***P < 0.001.

CD44v6-positive SDCSCs demonstrate higher tumorigenicity and increased resistance to chemotherapeutic drugs.

Next, we investigated whether increased resistance to chemotherapeutic agents by CSCs is due to CD44v6 expression or is simply a result of their stem cell–like characteristics. Using a fluorescence-conjugated CD44v6 antibody, we sorted SDCSCs into CD44v6⁺ and those lacking the protein (CD44v6⁻, Figure 4A) and examined whether these 2 subpopulations of SDCSCs have distinct functional attributes. Interestingly, CD44v6⁺ SDCSCs showed markedly higher spheroid-forming capacity than CD44v6⁻ cells (both cell lines P < 0.001), indicating that SDCSCs with high expression of CD44v6 have more pronounced stem cell–like characteristics (Figure 4B). We thereafter assessed how the presence of 5FU or oxaliplatin might impact the spheroid-forming capacity of the 2 subpopulations of SDCSCs. We found that CD44v6⁺ SDCSCs derived from both cell lines formed markedly increased numbers of spheroids as well in the presence of 5FU or oxaliplatin than CD44v6⁻ SDCSCs (both cell lines, both drugs, P <0.05, Figure 4C). In addition, xenotransplantation assays showed that CD44v6⁺ SDCSCs had a markedly higher ability to generate tumors compared with CD44v6⁻ cells, indicating that CD44v6⁺ cells are CSCs with higher tumor initiation capacity (Supplemental Table 3). In summary, we demonstrate that CSCs that express CD44v6 are substantially less sensitive to chemotherapeutic drugs and show greater tumorigenicity than those that do not express CD44v6.

Open in a separate window

Figure 4

CD44v6⁺ compared with CD44v6^– sphere-derived cancer stem cells have higher stemness.

(A) FACS analysis of CD44v6-positive and -negative sphere-derived cancer stem cells (SDCSCs) derived from a HCT116 cell line. (B) Spheroid-forming capacity of HCT116- and HT29-derived CD44v6⁺ and CD44v6^– SDCSCs (n = 16, Mann-Whitney U test). (C) Spheroid-forming ability of HCT116 (left) and HT29 (right) CD44v6⁺ and CD44v6^– SDCSCs in the presence of 5FU (top) and oxaliplatin (bottom; n = 6 per time point, Mann-Whitney U test). *P < 0.05, **P < 0.01, ***P < 0.001.

CD44v6 SDCSCs possess a distinct miRNA expression profile.

Since CD44v6 expression is associated with a higher resistance to chemotherapeutic agents, we performed small RNA sequencing for CD44v6⁺ and CD44v6⁻ cells to examine whether any distinct molecular signatures can help explain these differences in chemotherapeutic tolerance (Figure 5A). In total, 42 differentially expressed miRNAs were identified, of which 10 were upregulated and 32 were downregulated (Figure 5B and Supplemental Table 4). Interestingly, several putative tumor-suppressor miRNAs that inhibit stemness and developmental pathways were identified in the list of downregulated miRNAs (28, 41, 42) (Figure 5B). These miRNAs included miR-101-3p, miR-34a-3p, let-7f-1-3p, and several miR-200 family members (miR-141-3p, miR-141-5p, miR-200a-3p, and miR429). Therefore, while expression levels of these stemness-suppressive miRNAs were generally downregulated in the total SDCSC population, their expression was significantly lower in CD44v6⁺ SDCSCs (all P < 0.05), indicating that these CSCs have an epigenetic profile indicative of their greater stemness. To establish their functional roles, we predicted putative target genes for the differentially expressed miRNAs using TargetScan and subsequently assessed the associated pathways through KEGG analysis (Figure 5C and Supplemental Figure 2). In addition, we analyzed associated pathways for both upregulated and downregulated miRNAs in the CD44v6⁺ population independently (Supplemental Figure 3, A and B). Interestingly, several transcription-associated pathways were found to be affected by miRNAs dysregulated in the CD44v6⁺ CSC population. Considering that miRNAs themselves are posttranscriptional regulators of gene expression, the identification of multiple transcription-associated pathways affected by this collection of miRNAs highlights the importance of these miRNAs in gene regulation within the cancer stem cells.

Open in a separate window

Figure 5

CD44v6⁺ sphere-derived cancer stem cells have a distinct miRNA profile.

(A) Heatmap of differentially expressed miRNAs between CD44v6⁺ and CD44v6^– sphere-derived cancer stem cells (SDCSCs). (B) List of upregulated and downregulated miRNAs ranked by P values. (C) List of pathways identified by KEGG analysis based on predicted downstream target genes for differentially expressed miRNAs between CD44v6⁺ and CD44v6^– SDCSCs. Down, downregulated; Up, upregulated.

Cell sorting.

SDCSCs derived from HCT116 and HT26 cell lines were stained with CD44v6-FITC monoclonal antibody (catalog MA5-16966; Invitrogen), and propidium iodide was used to eliminate dead cells. Cell sorting was performed using a BD FACSAria II (BD Biosciences) with a 100-μm nodule. The number of sorted cells was monitored with an antibody specific for CD44v6-FITC, and stained cells were deemed to be positive for CD44v6.

Cell viability, colony formation, and spheroid formation assays.

Cells were incubated with various concentrations of 5FU or oxaliplatin for 72 hours in 96-well plates, and cell proliferation was measured using an MTT assay, as described previously (52). For the comparison between parental cells and SDCSCs, we cultured the same number of parental cells and SDCSCs in stem cell medium in 96-well plates. For the spheroid formation assay, HCT116 and HT29 cells were dissociated into single cells using TrypLE (Gibco), filtered using a 40-μm cell strainer (Corning), and seeded in Ultra-Low attachment 96-well plates (Sigma-Aldrich) in serum-free stem cell medium. Spheroids were treated with 5FU or oxaliplatin 24 hours after seeding. Spheres were counted using a light microscope (Olympus) following 4-day incubation.

Gene expression, miRNA expression, and transfection analyses.

Total RNA was extracted using miRNeasy Mini Kits (QIAGEN) following the manufacturer’s instructions. For the analysis of mRNA expression, 1 μg of total RNA was reverse transcribed to complementary DNA using an Advantage RT-for-PCR kit (Clontech Laboratories Inc.). Power SYBR Green (Applied Biosystems) real-time PCR was performed using a StepOnePlus system (Applied Biosystems). The primers used were as follows: OCT4 (forward: 5′-ACATCAAAGCTCTGCAGAAAGAACT, reverse: 5′-CTGAATACCTTCCCAAATAGAACCC); Nanog (forward: 5′-CCGAAGAATAGCAATGGTGTGACG, reverse: 5′-AGGAGAATTTGGCTGGAACTGC); and GAPDH (forward: 5′-ACCCAGACTGTGGTATGG, reverse: 5′-CAGTGAGCTTCCCGTTCAG). qPCR for CD44 variants was performed in accordance with a previously described method (53). All qPCR gene expression was normalized to the expression of GAPDH and analyzed using the ΔΔCt method. For miRNA expression analysis, we used a TaqMan Real-Time PCR Assay Kit (Applied Biosystems) and TaqMan MicroRNA Reverse Transcription Kit (Applied Biosystems), as described previously (54). The expression of all miRNAs was analyzed using the ΔΔCt method and normalized using RNU6B for cell lines and miR-16 for clinical samples. All TaqMan primers were purchased from Ambion. To inhibit the expression of CD44v6, siRNA was used in HCT116 and HT29 cell lines using siRNA sequences described previously (40) with siPort NeoFX (Gibco) transfection reagent, in accordance with the manufacturer’s instructions.

Western immunoblotting.

Western immunoblotting experiments were performed as described previously (55). Cells were treated with various concentrations of 5FU or oxaliplatin for 24 hours and lysed using 100 μl of 1× SDS sample buffer containing β-mercaptoethanol. The primary antibodies used were a monoclonal mouse anti–human CD44v6 antibody (ab78960; Abcam) and a monoclonal rabbit anti–human BMI1 antibody (6964P; Cell Signaling Technology). Anti-mouse IgG or anti-rabbit IgG secondary antibodies were purchased from Santa Cruz Biotechnology Inc. A monoclonal mouse β-actin antibody (Sigma-Aldrich) was used as the loading control.

Xenotransplantation assay.

Five-week-old male athymic nude mice (Harlan Laboratories) were housed under 12-hour controlled conditions of light and fed ad libitum. The SDCSCs and CRC cell lines, at various concentrations (10² to 10⁵ cells), were injected subcutaneously into the abdominal flanks of the mice, and tumor development was assessed weekly, for up to 2 months.

Clinical specimens.

A cohort of 344 fresh frozen tissue specimens comprising 127 normal mucosa and 217 CRC tissues were included in this study. These tissues were collected from patients enrolled at Mie University Hospital, Tsu, Japan, and the Tokyo Medical Dental University, Tokyo, Japan. Detailed information on patient demographics and clinicopathological characteristics is provided in Supplemental Table 1.

Small RNA sequencing and miRNA analysis.

Next-generation sequencing library construction for miRNAs from tissue was performed using a TruSeq Small RNA Kit (Illumina) with up to 1 μg total RNA input, in accordance with the manufacturer’s protocol. The quality of individual libraries was assessed using a High Sensitivity DNA Kit (Agilent). Libraries were pooled together prior to size selection (~148 nt) by gel electrophoresis using a Pippin HT instrument (Sage Science Inc.). The efficiency of size selection was assessed using the High Sensitivity DNA Kit, and the pooled libraries were thereafter quantified by qPCR using a KAPA Library Quantification Kit, Universal (KAPA Biosystems), prior to sequencing on an Illumina HighSeq 2500 (Illumina) instrument, with single-end 35-base read lengths. Small RNA-Seq data were processed as described previously (56). In brief, the raw sequence image files from the Illumina HiSeq 2500 in the bcl format were converted to the fastq format using CASAVA v1.8.2, and the adapters from the 3′ end were clipped using cutadapt v.1.10 (http://cutadapt.readthedocs.io/en/stable/guide.html). Additionally, all of the reads shorter than 15 nt and 3′ bases below a quality score of 30 were discarded. sRNABench (http://bioinfo5.ugr.es/sRNAbench) was used to map the reads to RNA libraries, and alignments were performed with Bowtie v0.12.9. The algorithm used in sRNABench was based on mirAnalyzer, where the reads with the same sequence were collapsed and mapped to the human genome and the miRBase v21 miRNA database (http://www.mirbase.org/). Differentially expressed miRNAs were identified using DESeq2 (57) and selected based on adjusted P values (P < 0.05), absolute value of log₂ fold change >1, and a base mean value >20. Gene ontology (GO) biological process pathways (http://geneontology.org) for the network figure (Supplemental Figure 2) were selected in the following way: for each dysregulated miRNA, the list of target transcripts annotated by TargetScan version 7.1 (http://www.targetscan.org/vert_71/) was uploaded into the Database for Annotation, Visualization and Integrated Discovery (https://david.ncifcrf.gov) for functional annotation. A GO pathway was selected if it was significant (P ≤ 0.05) across all up- or downregulated miRNAs. These criteria resulted in the final selection of 15 GO pathways. The miRNA expression sequencing data from our study have been deposited in the GEO public domain with the accession number {"type":"entrez-geo","attrs":{"text":"GSE125904","term_id":"125904"}}GSE125904.

Statistical analysis.

Differences between groups were evaluated by the Mann-Whitney U test (2-tailed), Fisher’s exact test, and 1-way ANOVA, as appropriate. For time-to-event analyses, survival estimates were calculated using Kaplan-Meier analysis, and groups were compared with the log-rank test. The thresholds between low and high expression groups for CD44v6 and miR-1246 were defined by Youden’s index. OS was based on the time assessed from the date the patient underwent surgery until the date of death resulting from any cause, or last known follow-up for patients who were still alive. DFS was based on the time measured from the date the patient underwent curative surgery to the date of disease recurrence, death from any cause, or last contact with the patient. In univariate and multivariate analyses, a Cox proportional hazards model was used to estimate HRs and 95% CI for death or recurrence. Assumption of proportionality were confirmed for the Cox proportional hazards analyses by generating Kaplan-Meier survival curves (e.g., high- vs. low-expression groups) and by ensuring that the 2 curves did not intersect each other. All P values were 2-sided, and those less than 0.05 were considered statistically significant. All error bars in the figures represented mean ± SEM.

We thank Margaret M. Hinshelwood, Baylor Scott & White Research Institute, Charles A. Sammons Cancer Center at Dallas, for critical suggestions and editing, further improving the quality of this article. We also thank Tadanobu Shimura, Baylor Scott & White Research Institute, Charles A. Sammons Cancer Center at Dallas, for technical assistance. This work was supported by grants CA72851, CA184792, CA202797 and CA187956 from the National Cancer Institute, NIH; RP140784 from the Cancer Prevention Research Institute of Texas; and grants from the Baylor Foundation and Baylor Scott & White Research Institute.