Malignant transformation of tissue stem cells (SC) may be the root of most cancer. Accordingly, we identified miRNA expression patterns in the normal human colonic SC niche to understand how cancer stem cells (CSC) may arise. In profiling miRNA expression in SC-enriched crypt subsections isolated from fresh, normal surgical specimens, we identified 16 miRNAs that were differentially expressed in the crypt bottom, creating an SC signature for normal colonic epithelia (NCE). A parallel analysis of colorectal cancer tissues showed differential expression of 83 miRNAs relative to NCE. Within the 16 miRNA signature for the normal SC niche, we found that miR-206, miR-007-3, and miR-23b individually could distinguish colorectal cancer from NCE. Notably, miR-23b, which was increased in colorectal cancer, was predicted to target the SC-expressed G protein-coupled receptor LGR5. Cell biology investigations showed that miR-23b regulated CSC phenotypes globally at the level of proliferation, cell cycle, self-renewal, epithelial–mesenchymal transition, invasion, and resistance to the colorectal cancer chemotherapeutic agent 5-fluorouracil. In mechanistic experiments, we found that miR-23b decreased LGR5 expression and increased ALDH+ CSCs. CSC analyses confirmed that levels of LGR5 and miR-23b are inversely correlated in ALDH+ CSCs and that distinct subpopulations of LGR5+ and ALDH+ CSCs exist. Overall, our results define a critical function for miR-23b, which, by targeting LGR5, contributes to overpopulation of ALDH+ CSCs and colorectal cancer. Cancer Res; 77(14); 3778–90. ©2017 AACR.

Mounting evidence indicates that (i) stem cells (SC) are the cells of origin of cancer (1, 2), (ii) SC overpopulation drives tumor initiation and progression (3–5), and (iii) SCs are resistant to conventional anticancer therapies. We found that ALDH1 is a marker for normal and malignant human colonic SC and tracks SC overpopulation during colon tumorigenesis. Although this finding and others indicate that cancer SC (CSC) overpopulation drives tumor growth, it is incompletely understood which dysregulated mechanisms cause the SC overpopulation. Because evidence points to an important role for miRNAs in the pathogenesis of various diseases, we studied miRNAs as a possible mechanism in colorectal cancer.

Aberrantly expressed miRNAs lead to widespread transcriptional dysregulation and cancer (6–10). In colorectal cancer, differential miRNA expression has been related to stage and site of the disease (11). Mounting evidence also indicates a role for miRNAs in the maintenance of the CSC phenotype (12–16).

We investigated dysregulated mechanisms in colonic SCs in colorectal cancer that are due to changes in miRNA expression. Our initial goal was to identify the set of miRNAs and their target genes that are specific to the normal colonic SC niche, and then identify the subset of these miRNAs that are aberrantly expressed in colorectal cancers compared with normal colonic epithelium (NCE). Our second goal was to see if some miRNAs are key to regulation of normal colonic SC populations, and when dysregulated contribute to SC overpopulation and colon tumorigenesis.

Accordingly, we devised an innovative strategy for miRNA profiling of human colonic SCs. Because colonic SCs have unique functional properties, the crypt bottom, which contains most colonic SCs, should have a unique gene expression profile that should be discernable by microarray analysis. Therefore, in the current study, we: (i) isolated pure crypts from surrounding stromal elements, (ii) isolated crypt subsections (bottom 1/10 and top 9/10), and then (iii) used microarray-based miRNA expression profiling to compare the SC-enriched crypt with the crypt top. We hypothesized that specific miRNAs are selectively expressed in the normal crypt SC niche and that some or all of these miRNAs contribute to SC mechanisms involved in colorectal cancer growth and development.

Isolation of colonic epithelium

The use of human clinical samples in this study was approved by the Thomas Jefferson University and Christiana Care Health System Institutional Review Boards. Informed written consent was obtained from the subjects (wherever necessary). The patient studies were conducted in accordance with the following ethical guidelines: Declaration of Helsinki, International Ethical Guidelines for Biomedical Research Involving Human Subjects, Belmont Report, and U.S. Common Rule. Colon tissues removed from patients during surgery were collected during or immediately after surgery and crypts were isolated as we previously described (17–19). Two different crypt subregions (SC enriched bottom 1/10 vs. top 9/10) were then isolated using microdissection under a dissecting microscope. All tissue samples (n = 5) were patient-matched (i.e., tumor and normal tissues were derived from the same patient).

Isolation of total RNA and microarray analysis

RNA was isolated using the TRIzol method as described by the manufacturer (GIBCO BRL) and purity determined by absorbance ratio at 260/280 nm. The total RNA was transcribed into first-strand cDNA, labeled, and hybridized to 368 miRNA-specific oligonucleotide (including 161 human, 84 murine, and 3 arabidopsis) probes by the Kimmel Cancer Center Microarray Facility at Thomas Jefferson University (Philadelphia, PA) as described (20).

TaqMan miRNA assay

Quantitative RT-PCR was performed using TaqMan MicroRNA Reverse Transcription (RT), TaqMan MicroRNA Assay and Taqman Universal PCR Master Mix kits according to the manufacturer's instruction. Specific primers for miR-10a, miR-198, miR-25, miR-146, miR-206, miR-218, miR-23b, and U6 RNA were obtained in the specific TaqMan MicroRNA Assay kits. PCR reactions were run on an ABI 7000 Real-Time PCR System (cDNA generated from 20 ng total RNA/reaction) using four replicates per tumor/normal sample pair. The relative amount of each miRNA in colorectal cancer versus NCE was determined using U6 to normalize for technical variability. Statistical analysis was done by the Student t test: Paired two sample for means.

Prediction and analysis of gene targets for miRNAs showing differential expression in the crypt bottom

The predicted target transcripts of miRNAs differentially expressed in the SC-enriched crypt bottom versus top were used to identify associations between them. Genome-wide miRNA–mRNA target prediction was done using rna22 by computationally seeking "hits" on the 3′UTRs of all known transcripts as described (21). Hierarchical clustering was also used to identify shared targeting and to visualize if the increased and decreased miRNAs could coregulate a subset of the transcripts.

Cell lines

HT29, SW480, and Caco2 colorectal cancer cell lines were purchased from the ATCC (year 2013; authenticated by the provider by cytogenetic analysis) and were grown in McCoy and L-15 media (GIBCO), respectively, containing 10% FBS and 1% penicillin/streptomycin. All the experiments were carried out within 10 passages after being thawed. The cells were routinely tested for mycoplasma by a Universal mycoplasma detection kit by the ATCC. The cells were grown in 5% CO2 and 95% air at 37°C.

Analysis and sorting of ALDH-high cells

The Aldefluor assay was performed according to the manufacturer's protocol (Stem Cell Technologies, Inc.) and samples were run on a BD FACS Aria II using the FACSDiva Software as described previously (5, 22).

Transient transfection of anti-miRNA and precursor of miR-23b

miR-23b anti-miRNA, precursor siRNA, and respective controls (Life Technologies) were used to transfect cell lines SW480, CaCo2, and HT29 in a 12-well or a 6-well format, and media were changed 24 hours posttransfection. Increase or decrease of miRNA levels was evaluated using a miRNA TaqMan Assay. The cells were analyzed 24–48 hours posttransfection.

Proliferation assays

Cells were plated and grown in serum-free media in 24-well plates. Forty-eight hours after treatment, cells were trypsinized, counted, and analyzed for viability by Trypan blue dye exclusion using a Countess Cell counter (Invitrogen). WST1 (Cayman Chemical) and MTT assays (Life Technologies, Inc.) were performed according to the manufacturer's protocol.

Cell-cycle analysis

Propidium iodide cell-cycle analysis was done as described previously (22).

Soft agar assay

Soft agar assay was carried out as previously described (23).

Colonosphere assay

A colonosphere assay was carried out as previously described (22, 24).

Invasion assay

Cells were plated onto an 8-μm pore-sized 24-well insert containing Matrigel (300 μg/mL) in serum-free DMEM media. Inserts were placed in 24-well plates with 10% serum-containing media. Cells were incubated for 48 hours, after which, invasive cells were fixed in 4% paraformaldehyde and stained with crystal violet. Images were taken using a NIKON light microscope, and number of cells was counted for each replicate per treatment.

Immunocytochemistry and FACS

Immunocytochemistry was done on SW480 cells as previously described (18, 19) using mouse anti-human E-cadherin (Abcam) and rabbit anti-human vimentin (Abcam) primary antibodies. LGR5/ALDH1 double staining: LGR5 staining on cells was done using PE-conjugated antibody (Miltenyi Biotec) according to the manufacturer's protocol. This was immediately followed by ALDEFLUOR assay. The quadrants were drawn using proper controls (PE-conjugated IgG) and ALDH+ DEAB sample.

Western blotting

Western blotting was performed as previously described (18, 19). Briefly, samples (30 μg) were run on precast gels (Lonza, Inc.) using the Bio-Rad system. Membranes were incubated with primary anti-vimentin and E-cadherin antibodies overnight. Blots were developed using West Dura Supersignal ECL (Thermo Scientific) and viewed using SYNGENE G-Box. The band intensities of the blot were measured using GeneSYS software.

Generation of stable clones

HEK 293T cells were transfected with miR-23b precursor expressing lentiviral vector, miR-23b miRZip lentiviral inhibition vector, or the control vector (System Biosciences, Inc.) together with helper plasmid (pCMVΔR8.2) and envelope plasmid (pMD.G). Viral particles collected were used for transducing HT29 cells along with polybrene solution (10 μg/mL). Media were changed after 24 hours, and puromycin (5.0 μg/mL) was added 48 hours after transfection. Colonies obtained after selection were expanded, maintained, and used for further experiments.

Luciferase assay

Plasmids expressing luciferase and the 3′UTR of the predicted target genes, LGR5 and LRIG1, were purchased from GeneCopoeia. Thirty-thousand cells were plated in a 96-well plate and 24 hours later cotransfected with the luciferase vector and miR-23b precursor (catalog no. AM17100; Life Technologies) and precursor control molecules using Lipofectamine 2000. The media were changed 24 hours posttransfection, and the cells were used for Dual-Color Luciferase Assay (Promega) and luminescence was recorded using the TECAN plate reader.

5-Fluorouracil treatment

The IC50 was determined for 5-fluorouracil (5-FU) against HT29 cell lines using varying concentrations of the drug (20–60 μmol/L). The drug at the IC50 concentration was added to the cells and the cells were counted 24 hours posttreatment.

RNA-Seq analysis

HT29 cells transfected with miR-23b anti-miRNA, precursor, and their respective controls were used for mRNA-Seq. The RNA was isolated 48 hours after transfection using TRIzol LS reagent (Invitrogen) according to the manufacturer's protocol. RNA quality was determined using a Fragment Analyzer (Advanced Analytical Technologies, Inc.), and Poly(A)+ RNA was isolated with the Absolutely mRNA Purification Kit (Agilent). RNA-Seq libraries were constructed using the TruSeq RNA sample prep Kit (Illumina). The indexed libraries were sequenced as a pool using Illumina HiSeq technology. The reads were then matched to human genome hg19 using Gencode V11 annotation with 1 mismatch allowed. Differentially expressed transcripts were identified using Cuffdiff (Cufflinks V2.0.2) with a threshold of 2-fold change with a significant P value in the treated samples as compared with the control. The accession number for our microarray data is GSE59290. The gene list was narrowed down further using the Gene Ontology (GO) term-enrichment analysis tools such as DAVID and results from the miRNA target prediction tools TARGETSCAN, miRANDA, and rna22.

Statistical analysis

An unpaired t test was used to determine significance between the treatment group and the appropriate controls. All the values obtained with a P value less than 0.05 were considered to be statistically significant.

The first sections of the results involved a discovery approach that identified candidate miRNAs for colonic SCs, which then allowed us to do further experiments that generated novel biological findings on SCs as described in the latter sections of the results.

miRNAs that are selectively expressed in the colonic SC niche

We profiled the colonic crypt subsections (i.e., bottom vs. top subsections of crypts isolated from the same patient sample) and discovered in the crypt bottom a distinct signature involving 16 differentially expressed miRNAs (Fig. 1A). This signature included miRNA species that were either significantly over-represented (n = 11) or significantly under-represented (n = 5) in the bottom relative to the crypt top (Student t test; P < 0.05). Among the miRNAs analyzed, miR-23b showed the greatest downregulation (1.42-fold, P = 0.015) and miR-25 the greatest upregulation (1.11-fold, P = 0.0487).

Figure 1.

A, Heat map of miRNA expression in the bottom 1/10th (leftmost three columns) versus the upper 9/10th (rightmost three columns) of normal human colonic crypts. Note that miRNAs that are overexpressed in the bottom of the crypt (red) relative to the top are underexpressed in the top of the crypt (green) relative to the bottom. The key shows how various colors represent particular relative levels of miRNAs. This signature includes miRNA species that were either significantly over-represented (miR-025, miR-007-3, miR-142, miR-009-2, miR-143, miR-010a, miR-153, miR-125b-2, miR-031, miR-151, miR-172a) or significantly under-represented (miR-023b, miR-206, miR-198, miR-206, miR-218-2) in the bottom relative to the crypt top. B, miRNA expression, as detected by microarray analysis in colorectal cancer tissues versus purified colonic epithelium. It shows that 83 miRNAs are differentially expressed in tissues from colorectal cancer patients (rightmost six columns) versus normal colonic epithelium (leftmost five columns). This figure illustrates the patterns that are seen when a large set of miRNAs is surveyed; it is not meant to show details, which is why the vertical axis is not legible. However, the quantitative data are given in Supplementary Table S1. C, miRNA expression profiles in purified colonic epithelium (leftmost four columns) versus colorectal cancer (rightmost five columns) using our 16 miRNA signature (see A) for the crypt bottom. The difference between expression of these 16 miRNAs in normal versus colorectal cancer was statistically significant (P < 0.05) and could distinguish malignant from normal colon tissues. Three of the 16 miRNAs (miR-206, miR-007, and miR-23b), when used individually, can also distinguish malignant from normal colon tissues. *, P < 0.05; **, P < 0.01; error bars, SEM.

Figure 1.

A, Heat map of miRNA expression in the bottom 1/10th (leftmost three columns) versus the upper 9/10th (rightmost three columns) of normal human colonic crypts. Note that miRNAs that are overexpressed in the bottom of the crypt (red) relative to the top are underexpressed in the top of the crypt (green) relative to the bottom. The key shows how various colors represent particular relative levels of miRNAs. This signature includes miRNA species that were either significantly over-represented (miR-025, miR-007-3, miR-142, miR-009-2, miR-143, miR-010a, miR-153, miR-125b-2, miR-031, miR-151, miR-172a) or significantly under-represented (miR-023b, miR-206, miR-198, miR-206, miR-218-2) in the bottom relative to the crypt top. B, miRNA expression, as detected by microarray analysis in colorectal cancer tissues versus purified colonic epithelium. It shows that 83 miRNAs are differentially expressed in tissues from colorectal cancer patients (rightmost six columns) versus normal colonic epithelium (leftmost five columns). This figure illustrates the patterns that are seen when a large set of miRNAs is surveyed; it is not meant to show details, which is why the vertical axis is not legible. However, the quantitative data are given in Supplementary Table S1. C, miRNA expression profiles in purified colonic epithelium (leftmost four columns) versus colorectal cancer (rightmost five columns) using our 16 miRNA signature (see A) for the crypt bottom. The difference between expression of these 16 miRNAs in normal versus colorectal cancer was statistically significant (P < 0.05) and could distinguish malignant from normal colon tissues. Three of the 16 miRNAs (miR-206, miR-007, and miR-23b), when used individually, can also distinguish malignant from normal colon tissues. *, P < 0.05; **, P < 0.01; error bars, SEM.

Close modal

miRNAs that are aberrantly expressed in human colorectal cancers

Profiling miRNA expression in colorectal cancer tissues versus purified NCE using the same microarray showed that 83 miRNAs were differentially expressed (P < 0.05; Fig. 1B; Supplementary Table S1). Of these, 26 showed highly significant differences (P < 0.01). We then determined whether any or all of the miRNAs in the 16 miRNA signature could distinguish colorectal cancer from purified NCE. A hierarchical clustering of the samples by miRNA expression data for the 16 miRNA set in colorectal cancer versus NCE is shown in Fig. 1C. Differences in expression of 3 of the 16 (miR-206, miR-007-3, and miR-23b) each could distinguish colorectal cancer from NCE (P < 0.05) as could the signature itself (P < 0.01). The differential expression was independently analyzed in additional NCE and colorectal cancer samples for a select group (n = 7) of these 16 miRNAs (Supplementary Fig. S1). Differential expression was confirmed in the majority of miRNAs tested.

Number of mRNAs that are targeted by miRNAs in the SC-enriched signature

To identify the number of predicted gene targets, we used human data and the rna22 algorithm (21) in an unsupervised approach, which does not penalize nonconserved interactions. The number of genes whose 3′ UTR regions are predicted to be targets of 1 or more of the 16 miRNAs of interest was high (>5,000). The miRNAs were organized into three main clusters, each containing both increased and decreased miRNAs (Supplementary Fig. S2).

The number of target genes predicted to have multiple hits in the 3′ region involving miRNAs from the 16 miRNA set was then determined (Supplementary Table S2). It is clear that different pairs of increased and decreased miRNAs are predicted to both cotarget and individually target different subsets of transcripts. Furthermore, complex coregulation by several of the miRNAs appears to be the norm rather than the exception. These computational results provide additional evidence that the differentially expressed miRNAs are biologically related and at least some of the transcripts predicted to be targeted by multiple miRNAs in the SC-enriched profile are likely to have biological relevance.

Specific gene targets for the miRNAs in the SC-enriched signature

Using the rna22 method, two lists of candidate targets were generated: one for miRNAs over-represented in the crypt bottom (11 miRNAs, Fig. 1A) and one for miRNAs over-represented in the upper crypt (5 miRNAs, Fig. 1A). Predicted target mRNAs were ranked by the number of differentially expressed miRNAs predicted to target them. Evaluation of the top 200 predicted targets from rna22 analysis for the two classes of miRNAs (using PubMed searches and DAVID Analysis) showed that a substantial number of these predicted targets were reported to be involved in colorectal cancer development. This is consistent with the fact that our preliminary data show that the set of miRNAs for the crypt bottom is able to distinguish colorectal cancer from NCE. Selected candidates (n = 31) from the top 200 predicted targets for miRNAs over-represented in the crypt bottom and in the upper crypt are listed in Supplementary Table S3. Predicted targets for miRNAs over-represented in the crypt bottom were consistent with the types of genes, such as apoptosis-related genes like APAF1 (ENSG00000120868) and cytochrome c (ENSG00000172115; ref. 25), which should be down regulated in SC populations. Conversely, predicted targets for miRNAs over-represented in the upper crypt were consistent with genes involved in regulation of cell growth, such as MeCP2 (ENSG00000169057; ref. 26), PAK2 (ENSG00000180370; refs. 27, 28), and c-MYB (ENSG00000118513; ref. 29), which should be down regulated in nonproliferating cells. Thus, target gene prediction for miRNAs in the bottom and upper crypts was consistent with anticipated expression of genes based on cell phenotypes along the crypt axis.

Classification of gene categories for the gene targets

Gene categories for the 200 highest scoring target genes for miRNAs enriched in the bottom and upper crypts are shown in Supplementary Table S3 based on GO term-enrichment analysis. Here, we found categories of predicted target genes to be consistent with anticipated gene expression patterns based on cell phenotypes in the bottom and upper crypts where, respectively, stem/proliferative cells and differentiated cells reside.

Predicted targets common to the differentially expressed miRNAs

We then focused on specific miRNAs to analyze if they target the predicted genes. We first identified predicted common targets of miR-007-3, miR-25, and miR-206 using rna22. We surveyed the targets and their biological roles and chose four interesting targets for further study: (i) methyl-CpG–binding protein 2 (MECP2), (ii) transcription factor SOX-11, (iii) lecithin retinol acyltransferase (LRAT), and (iv) disintegrin and metalloproteinase domain-containing protein 12 Precursor (ADAM12). To test for a possible relationship between miRNA and mRNA, the levels of these miRNAs were compared with the level of expression of mRNAs of these four common predicted targets in five colorectal cancer cell lines (COLO320, SW480, HCT116, HT29, and LoVo). Supplementary Fig. S3 suggests an inverse relationship between miRNA and mRNA expression levels, suggesting that some of the miRNAs are downregulating the mRNA of the predicted common target. However, this relationship did not hold for all of the predicted targets, suggesting that some might not be actual targets. For example, LRAT was predicted to be the common target for the three miRNAs, but LRAT expression was up regulated in several cell lines. That predicted targets from rna22 computational analysis do not turn out to be functional targets might also be explained by different confounding factors such as extent of Watson crick base pairing, proximity to secondary structures in mRNA strand, and AU-rich composition.

To further test if these miRNAs change mRNA expression levels of their predicted targets, HT29 and SW480 cells were transfected with the anti-miRNA for the respective miRNAs, and real-time PCR was used to analyze mRNA levels of predicted targets (SOX11, ADAM12, and MeCP2). We found that miR-25 anti-miRNA and cotransfection of miR-25 and miR-007-3 anti-miRNA induced an increase in SOX11 and ADAM12 mRNAs in SW480 cells but not HT29 cells (Supplementary Fig. S4). This might be explained by the fact that among the cell lines analyzed (Supplementary Fig. S3), we found SW480 cells had the lowest endogenous levels of these predicted target genes and had inversely highest levels of miR-25 and miR-007-3. Variability in the other miRNA anti-miRNA transfection results indicated that these candidate targets might not be the best ones to analyze and that more likely targets might involve specific SC genes. Consequently, we focused on expression of these miRNAs in SCs, their ability to modulate expression of predicted SC gene targets, and their ability to affect various SC properties.

Candidate miRNAs expressed in colonic CSCs

Using an initial screen, the relative expression of four miRNAs (miR-25, miR-007-3, miR-206, and miR-23b) was measured in CSCs versus non-CSCs that were isolated using the ALDEFLUOR assay. miR-23b and miR-206 levels were increased and miR-25 and miR-007-3 levels decreased in ALDH+ versus ALDH cancer cells (Supplementary Fig. S5).

Because miR-23b showed the highest level of differential expression, we identified targets of miR-23b for known SC-classified genes (using miRNA prediction tools rna22, targetscan, and MiRANDA). ALDH1A1, LRIG1, and LGR5 were predicted targets. Two approaches were then used for validation: (i) luciferase assay and (ii) changes in expression of predicted targets following transfection with precursor and anti-miRNA constructs. miR-23b precursor–treated HT29 and SW480 cells had reduced luciferase activity for LGR5 and LRIG1 but not ALDH1 (Supplementary Fig. S6), which was consonant with LGR5 and LRIG1 being targets of miR-23b. However, a second approach provided stronger support for LGR5 as being a prime target. In this second approach, LGR5 (but not LRIG1) mRNA levels were inversely correlated to the levels of miR-23b in HT29 cells transfected with miR-23b precursor and anti-miRNA (Supplementary Fig. S6).

miR-23b regulation of colonic SCs

Because miR-23b was overexpressed (>2-fold) in ALDH+ cells (Fig. 2A and B), we determined whether miR-23b precursor or anti-miRNA transfection modulates SC proportion. The transient increase in miR-23b increased the proportion of ALDH+ cells; a decrease of miR-23b showed the opposite effect in both SW480 and HT29 lines (Fig. 2C and D). We observed a similar trend in response to miR-23b modulation in controls double-transfected with TagRFP-expressing vector to confirm that ALDH+ cells are indeed transfected (Fig. 2D).

Figure 2.

miR-23b regulates proportion of ALDH+ cells. A, Normalized miR-23b expression in ALDH+ cells versus ALDH cells in colorectal cancer cell lines SW480 and HT29. B, ALDEFLUOR assay on SW480 cells showing two plots, with (left) and without (right) DEAB inhibitor, which identifies 12.5% of ALDH+ cells. C, Fold change in ALDEFLUOR+ cells transfected with miR-23b precursor or anti-miRNA versus controls in SW480 and HT29 cells. D, Fold change in ALDEFLUOR+/TagRFP copositive cells transfected with miR-23b precursor or anti-miRNA versus controls in HT29 cells.

Figure 2.

miR-23b regulates proportion of ALDH+ cells. A, Normalized miR-23b expression in ALDH+ cells versus ALDH cells in colorectal cancer cell lines SW480 and HT29. B, ALDEFLUOR assay on SW480 cells showing two plots, with (left) and without (right) DEAB inhibitor, which identifies 12.5% of ALDH+ cells. C, Fold change in ALDEFLUOR+ cells transfected with miR-23b precursor or anti-miRNA versus controls in SW480 and HT29 cells. D, Fold change in ALDEFLUOR+/TagRFP copositive cells transfected with miR-23b precursor or anti-miRNA versus controls in HT29 cells.

Close modal

We then found that LGR5 and ALDH1 identify different SC subpopulations of HT29 cells. LGR5 mRNA levels were down regulated in ALDH+ cells, and FACS staining showed distinct populations of LGR5+ and ALDH+ cells (Fig. 3A and B).

Figure 3.

LGR5 and ALDH1 expressions do not overlap in cancer cells. A, FACS results for HT29 cells for LGR5 staining and ALDELFLUOR+ activity. The gates were drawn based on the DEAB controls. The percent positive cells for the gated results is shown B, Normalized LGR5 mRNA levels of ALDH+ cells versus ALDH cells for SW480 and HT29 cell lines.

Figure 3.

LGR5 and ALDH1 expressions do not overlap in cancer cells. A, FACS results for HT29 cells for LGR5 staining and ALDELFLUOR+ activity. The gates were drawn based on the DEAB controls. The percent positive cells for the gated results is shown B, Normalized LGR5 mRNA levels of ALDH+ cells versus ALDH cells for SW480 and HT29 cell lines.

Close modal

Effect of miR-23b on cell proliferation

Transiently increasing miR-23b decreased proliferation of both SW480 and HT29 cells (Supplementary Fig. S7A). When miR-23b anti-miRNA was transfected, proliferation increased (Supplementary Fig. S7B). These results were validated using an MTT assay (Supplementary Fig. S7C) that showed that miR-23b precursor decreased proliferation of SW480 and HT29 cells (17% and 13%, respectively) and miR-23b precursor increased proliferation (14% for each). Because miRNA23b is known to cause cells to accumulate in G0–G1 (30–32), we tested for this effect in colorectal cancer cell lines. Transfection with miR-23b precursor increased the number of cells in G0–G1; the number in S-phase decreased (Supplementary Fig. S7D and S7E).

Effect of miR-23b on self-renewal

Soft-agar growth assay showed that transient knockdown of miR-23b inhibits anchorage-independent growth as reflected by decreasing colony formation of HT29 cells (Fig. 4A and B). Sphere-formation assay using nonadherent cultures was also done following transient miR-23b transfection of HT29 cells. Transfection with miR-23b anti-miRNA resulted in smaller spheres, and miR-23b precursor resulted in larger spheres (Fig. 4C and D). Because SW480 cells demonstrate poor colony formation, they were not assayed. Alternatively, we transfected CaCo2 cells and analyzed sphere formation (Supplementary Fig. S8), which showed very similar results to HT29 cells. In independent experiments, stably transfected cells having miR-23b overexpression and knockdown showed similar effects on proliferation (Fig. 4E and F) and on self-renewal (Fig. 4G and H) as did transiently transfected cells (Supplementary Fig. S7A–S7D; Fig. 4A and B). To explain these results, we conjecture that an increase in 23b increases the proportion of ALDH1+ SCs. As sphere assays are a week-long assay, the immediate effect of miR-23b is to increase the number of ALDH+ cells in this assay capable of self-renewal and sphere formation, and that is why we see a greater number of spheres over the period of 7 days as compared with the control.

Figure 4.

miR-23b affects self-renewal. A, Soft-agar assay for HT29 cells after transfection with miR-23b anti-miRNA, control, and untreated samples. B, Quantification of colonies in soft agar. C, Colonosphere assays for HT29 cells after transfection with miR-23b precursor, anti-miRNA, and control samples. D, Quantification of colonospheres in suspension cultures. E, Proliferation assay results on the miR-23b-overexpressing, knockdown, and the control clones. F, The bar graph represents percentage of cells in various phases of the cell cycle of the overexpressing and the knockdown clone as compared with the control. G, Bright-field images of colonospheres generated at day 8 from the miR-23b–overexpressing, knockdown, and the control clone. H, Size of the colonospheres.

Figure 4.

miR-23b affects self-renewal. A, Soft-agar assay for HT29 cells after transfection with miR-23b anti-miRNA, control, and untreated samples. B, Quantification of colonies in soft agar. C, Colonosphere assays for HT29 cells after transfection with miR-23b precursor, anti-miRNA, and control samples. D, Quantification of colonospheres in suspension cultures. E, Proliferation assay results on the miR-23b-overexpressing, knockdown, and the control clones. F, The bar graph represents percentage of cells in various phases of the cell cycle of the overexpressing and the knockdown clone as compared with the control. G, Bright-field images of colonospheres generated at day 8 from the miR-23b–overexpressing, knockdown, and the control clone. H, Size of the colonospheres.

Close modal

Effect of miR-23b on epithelial–mesenchymal transition and invasion

Immunocytochemistry for epithelial mesenchymal transition (EMT) markers vimentin and E-cadherin showed that transfection with the precursor promoted EMT as indicated by increased vimentin and decreased E-cadherin staining in SW480 cells (Fig. 5A and B). The opposite effect on vimentin and E-cadherin was observed with miR-23b anti-miRNA. Immunostaining results were validated by Western blotting (Fig. 5C and D). Because HT29 cells did not express vimentin, they were not assayed. In Matrigel invasion assays, miR-23b overexpression in SW480 cells promoted cell invasion versus controls (Fig. 5E and F).

Figure 5.

miR-23b promotes EMT and invasive characteristics. A, Immunocytochemistry for E-cadherin and vimentin for SW480 cells 48 hours after transfection with miR-23b anti-miRNA and precursor. B, Quantification of ICC results in A. C, Western blots for E-cadherin and vimentin expression in SW480 cells 48 hours after transfection by miR-23b precursor (P), anti-miRNA (A), and controls (PN and AN). D, Quantification of Western blot results in C. E, Invasion assay on SW480 cells treated with miR-23b precursor, anti-miRNA, and control. F, Quantification of results in E.

Figure 5.

miR-23b promotes EMT and invasive characteristics. A, Immunocytochemistry for E-cadherin and vimentin for SW480 cells 48 hours after transfection with miR-23b anti-miRNA and precursor. B, Quantification of ICC results in A. C, Western blots for E-cadherin and vimentin expression in SW480 cells 48 hours after transfection by miR-23b precursor (P), anti-miRNA (A), and controls (PN and AN). D, Quantification of Western blot results in C. E, Invasion assay on SW480 cells treated with miR-23b precursor, anti-miRNA, and control. F, Quantification of results in E.

Close modal

Effect of miR-23b on chemosensitivity

The expression of miR-23b in HT29 cells was decreased and the levels of its target LGR5 increased after 5-FU treatment versus controls (Fig. 6A and D). Treatment of overexpressing and knockdown miR-23b clones with 5-FU was done to further investigate this effect. 5-FU treatment of the miR-23b knockdown clones demonstrated a lower 5-FU–induced cytotoxicity compared with control clones (Fig. 6B and C). We also investigated the effect of 5-FU on ALDH+ and LGR5+ populations. Treatment of HT29 cells with 5-FU increased the proportion of ALDH+/LGR5+ copositive cells as well LGR5+ cells relative to ALDH+ cells (Fig. 6E). Similar results were obtained with 5-FU treatment of CaCo2 cells (Supplementary Fig. S8). These results indicate that 5-FU treatment induces the conversion of ALDH+ cells to LGR5+ and copositive cells.

Figure 6.

Inhibition by miR-23b promotes chemoresistance to 5-FU. A, expression of miR-23b in HT29 cells treated with 5-FU versus DMSO. B, Proliferation assay results 24 hours after 5-FU treatment of the miR-23b–overexpressing and knockdown clones of HT29. Values represent average of normalized cell count as compared with the DMSO control. C, WST1 assays results showing viability of miR-23b–overexpressing and knockdown clones versus controls 24 hours after 5-FU treatment. Values represent average relative absorbance of drug-treated cells as compared with DMSO control. D, Normalized mRNA levels of ALDH1 and LGR5 in 5-FU–treated HT29 cells versus control. E, Fold change in proportion of cells expressing SC markers ALDH1, LGR5, or both in response to 5-FU treatment of HT29 cells as compared with control.

Figure 6.

Inhibition by miR-23b promotes chemoresistance to 5-FU. A, expression of miR-23b in HT29 cells treated with 5-FU versus DMSO. B, Proliferation assay results 24 hours after 5-FU treatment of the miR-23b–overexpressing and knockdown clones of HT29. Values represent average of normalized cell count as compared with the DMSO control. C, WST1 assays results showing viability of miR-23b–overexpressing and knockdown clones versus controls 24 hours after 5-FU treatment. Values represent average relative absorbance of drug-treated cells as compared with DMSO control. D, Normalized mRNA levels of ALDH1 and LGR5 in 5-FU–treated HT29 cells versus control. E, Fold change in proportion of cells expressing SC markers ALDH1, LGR5, or both in response to 5-FU treatment of HT29 cells as compared with control.

Close modal

Novel targets of miR-23b identified by global mRNA-Seq analysis

RNA-Seq analysis across samples was done to identify genes that were differentially expressed with varying levels of miRNA23b in HT29 cells (GEO accession no: GSE59290). Around 3,000 genes were found to be up regulated (>2-fold) in the anti-miRNA-treated cells and down regulated in the precursor-treated cells (Fig. 7A). Of these 3,000 genes, 72 were identified as predicted targets of miR-23b using multiple miRNA target prediction tools such as rna22, miRANDA, and TARGETSCAN (Fig. 7B). GO term analysis of the 3,000 genes revealed majority of them are involved in regulation of cell cycle, response to stress, and protein transport and localization (Fig. 7C).

Figure 7.

RNA-Seq analysis to identify novel targets of miR-23b. A, Log (fold change) values of genes that are upregulated in miR-23b anti-miRNA-treated (red) cells and downregulated in miR-23b precursor–treated (blue) cells versus controls. B, Heatmap showing the differential expression (red, overexpressed; green, underexpressed) of gene transcripts that are predicted targets of miR-23b in cells treated with miR-23b precursor (P), anti-miRNA (A), and their respective controls (PN, AN). The candidate genes that are targets of miR-23b (from the heatmap) and that are classified as playing a role in self-renewal include APLP2, BCL2L13, ATF3, AKT2, RND3, ITGB1, TGFBI, ING4, ABCC1, DICER1, and ATF2. C, GO term analysis of the 3,000 genes using DAVID software.

Figure 7.

RNA-Seq analysis to identify novel targets of miR-23b. A, Log (fold change) values of genes that are upregulated in miR-23b anti-miRNA-treated (red) cells and downregulated in miR-23b precursor–treated (blue) cells versus controls. B, Heatmap showing the differential expression (red, overexpressed; green, underexpressed) of gene transcripts that are predicted targets of miR-23b in cells treated with miR-23b precursor (P), anti-miRNA (A), and their respective controls (PN, AN). The candidate genes that are targets of miR-23b (from the heatmap) and that are classified as playing a role in self-renewal include APLP2, BCL2L13, ATF3, AKT2, RND3, ITGB1, TGFBI, ING4, ABCC1, DICER1, and ATF2. C, GO term analysis of the 3,000 genes using DAVID software.

Close modal

This study was designed to identify novel miRNAs that are selectively expressed in the human crypt SC niche and that contribute to dysregulation of colonic SCs in colorectal cancer. For the 16-miRNA signature for the SC niche, we found categories of predicted target genes to be consistent with anticipated gene expression patterns based on cell phenotypes in the bottom and upper crypts where, respectively, stem/proliferative cells or differentiated cells reside. We then made several interesting discoveries: many of the miRNAs in the 16-miRNA signature are predicted to have common targets; differences in expression of three of the miRNAs (miR-206, miR-007-3, and miR-23b) could individually distinguish colorectal cancer from NCE, as could differences in the 16-miRNA panel itself. Because this study was not designed as a biomarker study, we are not making any claims that these miRNA sets have any clinical utility, rather we are interested in understanding miRNA mechanisms that contribute to the SC origin of colorectal cancer. In the future, however, it may be worthwhile to explore the value of these miRNAs as biomarkers in studies that include the pathology and molecular characteristics of the cancers.

So why does an miRNA signature for an SC population predict cancer? It could be that SCs are overproduced in colorectal cancer (2, 3, 5), and so, there would be more SC-associated markers in the malignant colon, including miRNAs that comprise the SC signature. Alternatively, some miRNAs in the signature might be critical to regulation of the colonic SC population, and when dysregulated, these miRNAs contribute to colorectal cancer development.

Accordingly, we investigated whether four candidate miRNAs (miR-25, miR-007-3, miR-206, and miR-23b) might be regulators of colonic CSCs. Among them, we found miR-23b expression was increased in primary colorectal cancers and in ALDH+ SC population from colorectal cancer cell lines. In several studies, miR-23b has been shown to play a role in the growth of a different cancer types (33–36) and other miRNAs have been implicated in CSC biology (15, 37–39). Thus, we conjectured that miR-23b has a role in maintenance of the colonic CSC function.

To test this conjecture, we examined the effect of miRNA23b on the SC properties of self-renewal, EMT, and invasion. Our findings show that increasing miR-23b promotes colony and sphere formation, and reducing miR-23b levels had the opposite effect. Moreover, we found that miR-23b enhanced the invasive characteristics of SW480 cells, which could be explained by our finding that miR-23b overexpression decreases E-cadherin and increases vimentin, which should promote EMT and invasion. EMT has been indicated to be associated with CSC phenotype (40), and our data suggest that miR-23b levels promote CSC phenotype via EMT.

We also found that miR-23b overexpression decreased cell proliferation and induced accumulation of cells in the G0–G1 phase of the cell cycle. This G0–G1 accumulation has been reported for other cancers and endothelial cells (30–32, 41, 42). These observations are consistent with our finding that miR-23b is overexpressed in ALDH+ SCs and that increasing miR-23b levels led to an increased number of ALDH+ SCs. Thus, it appears that miR-23b overexpression increases the proportion of ALDH+ CSCs, which are relatively quiescent or slowly cycling.

This raises the question: How might miR-23b increase the number of ALDH+ SCs? To find an answer, we looked for SC markers that might be targeted by miR-23b and discovered that LGR5 was a predicted target. We then found that LGR5 was down regulated in ALDH+ SCs, which is concordant with miR-23b being overexpressed in ALDH+ cells. Moreover, FACS analysis showed that distinct subpopulations of LGR5+ and ALDH+ SCs exist in colorectal cancer cell lines. Thus, it appears that multiple CSC populations coexist and that miR-23b regulates the sizes of these CSC subpopulations. In this view, ALDH+ cells have high miR-23b levels and are slow-cycling, self-renewing SCs. LGR5+ cells have low levels of miR-23b and are rapidly-cycling SCs with some degree of self-renewal and resistance to anticancer drugs. This model suggests a hierarchy of cells with varying degrees of stemness within a colon tumor. This possibility is consistent with reports on normal murine intestine, where the BMI+ SCs can give rise to LGR5+ SCs upon specific depletion of LGR5 cells (43). For example, LGR5+ cells, when exposed to irinotecan, convert to LGR5 cells, which can revert back to LGR5+ cells in the absence of the drug (44).

So how might miR-23b be involved in the mechanism by which CSCs are resistant to chemotherapy? Recently, miR-23b was found to contribute to chemoresistance of CSCs (45). Our results indicated that 5-FU treatment reduces miR-23b expression in colorectal cancer cells, suggesting that low levels of this miRNA may have a role in chemoresistance. This is possible because LGR5 mRNA levels are inversely correlated with miR-23a/b expression in colorectal cancer (46). Indeed, LGR5 levels have been shown to be a predictor of response to 5-FU treatment in cancer patients where high LGR5 expression is associated with poor outcomes after chemotherapy with 5-FU (47). We demonstrated using multiple functional assays that LGR5 expression is regulated by miR-23b. An increase in LGR5, which is known to mark cycling SCs, could explain the increase in proliferation of cells with reduced miR-23b levels. LGR5 protein has also been reported to enhance Wnt signaling via R-Spondin (48). This mechanism suggests that LGR5 expression in turn regulates miR-23b expression in a negative feedback loop mechanism mediated by Wnt target genes such as Myc (49, 50).

Because each miRNA in a cell is predicted to influence the expression of as many as 3,000 mRNA targets, it becomes essential to identify other targets regulated by miR-23b that could possibly play a role in self-renewal mechanisms of CSCs. Indeed, RNA-Seq analysis has already identified global mRNA changes due to miR-23b for breast cancer (51). Our RNA-Seq analysis of HT29 cells with altered levels of miR-23b revealed new gene targets such as ATF2 and AKT2, which are important components of Wnt and AKT/PI3Kinase signaling (30, 52–54). We predict that via the regulation of multiple targets, miR-23b plays an important role in driving tumor growth.

Our identification of a unique miRNA expression signature for the human colonic SC niche, one that distinguishes malignant from normal epithelium, led to our discovery of miR-23b as a regulator of human colon CSC subpopulations. Our findings have additional implications. For instance, our study suggests that among the miRNAs that signify cells at the crypt bottom, some of the other miRNAs in the set (other than miR-23b) may also be key to regulation of normal colonic SC population dynamics, and when dysregulated, they contribute to SC overpopulation and colon tumorigenesis. Given that most mRNAs are targeted by several different miRNAs, identification of such miRNAs and prediction of their respective mRNA targets could lead to the discovery of new CSC-based pathways involved in colorectal cancer. For example, if the 16-miRNA signature is aberrantly expressed in malignant colon, then there will likely be dysregulation of the expression of tens or hundreds of genes, ones that are targeted by the 16 miRNAs. Indeed, recent genomic scale gene expression analyses have found hundreds of mRNAs that are differentially expressed during colorectal cancer development (18). Our studies are a first step toward achieving our long-term goal, which is identifying targetable mechanisms involved in the SC etiology of colorectal cancer.

No potential conflicts of interest were disclosed.

Conception and design: V. Viswanathan, B.M. Boman

Development of methodology: V. Viswanathan, I. Rigoutsos, B.M. Boman

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): V. Viswanathan, L. Opdenaker, S. Modarai, P. Green, B.M. Boman

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): V. Viswanathan, L. Opdenaker, S. Modarai, J. Palazzo, J. Fields, I. Rigoutsos, G. Gonye, B.M. Boman

Writing, review, and/or revision of the manuscript: V. Viswanathan, L. Opdenaker, S. Modarai, P. Green, D. Galileo, J. Palazzo, J. Fields, I. Rigoutsos, G. Gonye, B.M. Boman

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): V. Viswanathan, M. Accerbi

Study supervision: D. Galileo

We thank Dr. Nicholas Petrelli for his generous support through the CTCR at the Helen F. Graham Cancer Center and Research Institute, RoseMarie Pope and Patricia Swanson for administrative support, Michael Samedio for technical assistance, Brenda Rabino, Lori Huelsenbeck-Dill, Marlee Goins, Drs. Mary Iacocca, Scott Goldstein, Gerald Isenberg, Frederick Dentsman, and Joseph D'Amico for assistance with tissue procurement, the CTCR Core Facility at the Helen F. Graham Cancer Center and Research Institute for use of the BD FACSAria II Flow Cytometer, Drs. Jennifer Ma, Massimo Negrini, Manuela Ferracin, and Chang-Gong Liu in the Kimmel Cancer Center Microarray Facility for assistance with the miRNA profiling, and Dr. Carlo Croce for supplying the miRNA chip.

This study was supported in part by NIH P20RR01 6472-04 to B.M. Boman, the Bioscience Center for Advanced Technology (B.M. Boman), Cancer B*Ware Foundation (all authors), and the Center for Translational Cancer Research (all authors). Additional support was provided by NIH R01GM096471 to P. Green. Use of the BioHen computer cluster in the Center for Bioinformatics and Computational Biology Bioinformatics Core was made possible through funding from Delaware INBRE NIH/NIGMS GM103446 (all authors) and the Delaware Biotechnology Institute (all authors).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Boman
BM
,
Wicha
MS
. 
Cancer stem cells: a step toward the cure
.
J Clin Oncol
2008
;
26
:
2795
9
.
2.
Boman
BM
,
Huang
E
. 
Human colon cancer stem cells: a new paradigm in gastrointestinal oncology
.
J Clin Oncol
2008
;
26
:
2828
38
.
3.
Boman
BM
,
Fields
JZ
,
Bonham-Carter
O
,
Runquist
OA
. 
Computer modeling implicates stem cell overproduction in colon cancer initiation
.
Cancer Res
2001
;
61
:
8408
11
.
4.
Boman
BM
,
Walters
R
,
Fields
JZ
,
Kovatich
AJ
,
Zhang
T
,
Isenberg
GA
, et al
Colonic crypt changes during adenoma development in familial adenomatous polyposis: immunohistochemical evidence for expansion of the crypt base cell population
.
Am J Pathol
2004
;
165
:
1489
98
.
5.
Huang
EH
,
Hynes
MJ
,
Zhang
T
,
Ginestier
C
,
Dontu
G
,
Appelman
H
, et al
Aldehyde dehydrogenase 1 is a marker for normal and malignant human colonic stem cells (SC) and tracks SC overpopulation during colon tumorigenesis
.
Cancer Res
2009
;
69
:
3382
9
.
6.
Bino
J
,
Enright
AJ
,
Aravin
A
,
Tuschl
T
,
Sander
C
,
Marks
DS
. 
Human MicroRNA targets
.
PLoS Biol
2004
;
2
:
e363
.
7.
Calin
GA
,
Sevignani
C
,
Dumitru
CD
,
Hyslop
T
,
Noch
E
,
Yendamuri
S
, et al
Human microRNA genes are frequently located at fragile sites and genomic regions involved in cancers
.
Proc Natl Acad Sci U S A
2004
;
101
:
2999
3004
.
8.
Croce
CM
. 
Causes and consequences of microRNA dysregulation in cancer
.
Nat Rev Genet
2009
;
10
:
704
14
.
9.
Calin
GA
,
Croce
CM
. 
MicroRNA signatures in human cancers
.
Nat Rev Cancer
2006
;
6
:
857
66
.
10.
Iorio
MV
,
Croce
CM
. 
MicroRNA dysregulation in cancer: diagnostics, monitoring and therapeutics. A comprehensive review
.
EMBO Mol Med
2012
;
4
:
143
59
.
11.
Michael
MZ
,
SM
OC
,
van Holst Pellekaan
NG
,
Young
GP
,
James
RJ
. 
Reduced accumulation of specific microRNAs in colorectal neoplasia
.
Mol Cancer Res
2003
;
1
:
882
91
.
12.
Ivey
KN
,
Srivastava
D
,
Ivey
KN
,
Srivastava
D
. 
MicroRNAs as regulators of differentiation and cell fate decisions
.
Cell Stem Cell
2010
;
7
:
36
41
.
13.
Mathieu
J
,
Ruohola-Baker
H
. 
Regulation of stem cell populations by microRNAs
.
Adv Exp Med Biol
2013
;
786
:
329
51
.
14.
Takahashi
RU
,
Miyazaki
H
,
Ochiya
T
. 
The role of microRNAs in the regulation of cancer stem cells
.
Front Genet
2014
;
4
:
295
.
15.
Shimono
Y
,
Zabala
M
,
Cho
RW
,
Lobo
N
,
Dalerba
P
,
Qian
D
, et al
Downregulation of miRNA-200c links breast cancer stem cells with normal stem cells
.
Cell
2009
;
138
:
592
603
.
16.
Gangaraju
VK
,
Lin
H
. 
MicroRNAs: key regulators of stem cells
.
Nat Rev Mol Cell Biol
2009
;
10
:
116
25
.
17.
Wildrick
DM
,
Lointier
P
,
Nichols
DH
,
Roll
R
,
Quintanilla
B
,
Boman
BM
. 
Isolation of normal human colonic mucosa: comparison of methods
.
In Vitro Cell Dev Biol Anim
1997
;
33
:
18
27
.
18.
Bhatlekar
S
,
Addya
S
,
Salunek
M
,
Orr
C
,
Surrey
S
,
McKenzie
S
, et al
Identification of a developmental gene expression signature, including HOX genes, for the normal human colonic crypt stem cell niche – Overexpression of the signature parallels stem cell overpopulation during colon tumorigenesis
.
Stem Cells Dev
2014
;
23
:
167
79
.
19.
Zhang
T
,
Fields
JZ
,
Otevrel
T
,
Masuda
E
,
Palazzo
JP
,
Isenberg
GA
, et al
Survivin-induced Aurora-B kinase activation - A mechanism by which APC mutations contribute to increased mitoses during colon cancer development
.
Am J Pathol
2010
;
177
:
2816
26
.
20.
Liu
CG
,
Calin
GA
,
Meloon
B
,
Gamliel
N
,
Sevignani
C
,
Ferracin
M
, et al
An oligonucleotide microchip for genome-wide microRNA profiling in human and mouse tissues
.
Proc Natl Acad Sci U S A
2004
;
101
:
9740
4
.
21.
Miranda
KC
,
Huynh
T
,
Tay
Y
,
Ang
YS
,
Tam
WL
,
Thomson
AM
, et al
A pattern-based method for the identification of MicroRNA binding sites and their corresponding heteroduplexes
.
Cell
2006
;
126
:
1203
17
.
22.
Opdenaker
LM
,
Modarai
SR
,
Boman
BM
. 
The proportion of ALDEFLUOR-positive cancer stem cells changes with cell culture density due to the expression of different ALDH isoforms
.
Cancer Stud Mol Med
2015
;
2
:
87
95
.
23.
Olive
M
,
Untawale
S
,
Coffey
RJ
,
Siciliano
MJ
,
Wildrick
DM
,
Fritsche
H
, et al
Characterization of the DiFi rectal carcinoma cell line derived from a familial adenomatous polyposis patient
.
In Vitro Cell Dev Biol
1993
;
29
:
239
48
.
24.
Kanwar
SS
,
Yu
Y
,
Nautiyal
J
,
Patel
BB
,
Majumdar
AP
. 
The Wnt/beta-catenin pathway regulates growth and maintenance of colonospheres
.
Mol Cancer
2010
;
9
:
212
.
25.
Li
P
,
Nijhawan
D
,
Budihardjo
I
,
Srinivasula
SM
,
Ahmad
M
,
Alnemri
ES
, et al
Cytochrome c and dATP-dependent formation of Apaf-1/caspase-9 complex initiates an apoptotic protease cascade
.
Cell
1997
;
91
:
479
89
.
26.
Lee
SH
,
Kim
IJ
,
Kim
JG
,
Park
JS
,
Kim
YS
,
Yamaguchi
M
, et al
Regulation of intestinal stem cell proliferation by human methyl-CpG-binding protein-2 in Drosophila
.
Cell Struct Funct
2011
;
36
:
197
208
.
27.
Deng
WW
,
Wu
L
,
Bu
LL
,
Liu
JF
,
Li
YC
,
Ma
SR
, et al
PAK2 promotes migration and proliferation of salivary gland adenoid cystic carcinoma
.
Am J Transl Res
2016
;
8
:
3387
97
.
28.
Siu
MK
,
Yeung
MC
,
Zhang
H
,
Kong
DS
,
Ho
JW
,
Ngan
HY
, et al
p21-Activated kinase-1 promotes aggressive phenotype, cell proliferation, and invasion in gestational trophoblastic disease
.
Am J Pathol
2010
;
176
:
3015
22
.
29.
Gewirtz
AM
,
Calabretta
B
. 
A c-myb antisense oligodeoxynucleotide inhibits normal human hematopoiesis in vitro
.
Science
1988
;
242
:
1303
6
.
30.
Zhang
H
,
Hao
Y
,
Yang
J
,
Zhou
Y
,
Li
J
,
Yin
S
, et al
Genome-wide functional screening of miR-23b as a pleiotropic modulator suppressing cancer metastasis
.
Nat Commun
2011
;
2
:
554
.
31.
He
HC
,
Zhu
JG
,
Chen
XB
,
Chen
SM
,
Han
ZD
,
Dai
QS
, et al
MicroRNA-23b downregulates peroxiredoxin III in human prostate cancer
.
FEBS Lett
2012
;
586
:
2451
8
.
32.
Majid
S
,
Dar
AA
,
Saini
S
,
Deng
G
,
Chang
I
,
Greene
K
, et al
MicroRNA-23b functions as a tumor suppressor by regulating Zeb1 in bladder cancer
.
PLoS One
2013
;
8
:
e67686
.
33.
Viswanathan
V
,
Fields
J
,
Boman
BM
. 
The miRNA23b-regulated signaling network as a key to cancer development – implications for translational research and therapeutics
.
J Mol Med
2014
;
92
:
1129
38
.
34.
Wang
P
,
Zhang
J
,
Zhang
L
,
Zhu
Z
,
Fan
J
,
Chen
L
, et al
MicroRNA 23b regulates autophagy associated with radioresistance of pancreatic cancer cells
.
Gastroenterology
2013
;
145
:
1133
43
.
35.
Begum
S
,
Hayashi
M
,
Ogawa
T
,
Jabboure
FJ
,
Brait
M
,
Izumchenko
E
, et al
An integrated genome-wide approach to discover deregulated microRNAs in non-small cell lung cancer: clinical significance of miR-23b-3p deregulation
.
Sci Rep
2015
;
5
:
13236
.
36.
Zhou
X
,
Xu
X
,
Wang
J
,
Lin
J
,
Chen
W
. 
Identifying miRNA/mRNA negative regulation pairs in colorectal cancer
.
Sci Rep
2015
;
5
:
12995
.
37.
Yu
XF
,
Zou
J
,
Bao
ZJ
,
Dong
J
. 
miR-93 suppresses proliferation and colony formation of human colon cancer stem cells
.
World J Gastroenterol
2011
;
7
:
4711
7
.
38.
Geng
J
,
Luo
H
,
Pu
Y
,
Zhou
Z
,
Wu
X
,
Xu
W
, et al
Methylation mediated silencing of miR-23b expression and its role in glioma stem cells
.
Neurosci Lett
2012
;
528
:
185
9
.
39.
Saini
S
,
Majid
S
,
Shahryari
V
,
Arora
S
,
Yamamura
S
,
Chang
I
, et al
miRNA-708 control of CD44(+) prostate cancer-initiating cells
.
Cancer Res
2012
;
72
:
3618
30
.
40.
Liu
X
,
Fan
D
. 
The epithelial-mesenchymal transition and cancer stem cells: functional and mechanistic links
.
Curr Pharm Des
2015
;
21
:
1279
91
.
41.
Liu
W
,
Zabirnyk
O
,
Wang
H
,
Shiao
YH
,
Nickerson
ML
,
Khalil
S
et al 
miR-23b targets proline oxidase, a novel tumor suppressor protein in renal cancer
.
Oncogene
2010
;
29
:
4914
24
.
42.
Wang
KC
,
Garmire
LX
,
Young
A
,
Nguyen
P
,
Trinh
A
,
Subramaniam
S
, et al
Role of microRNA-23b in flow-regulation of Rb phosphorylation and endothelial cell growth
.
Proc Natl Acad Sci U S A
2010
;
107
:
3234
9
.
43.
Tian
H
,
Biehs
B
,
Warming
S
,
Leong
KG
,
Rangell
L
,
Klein
OD
, et al
A reserve stem cell population in small intestine renders Lgr5-positive cells dispensable
.
Nature
2011
;
478
:
255
9
.
44.
Kobayashi
S
,
Yamada-Okabe
H
,
Suzuki
M
,
Natori
O
,
Kato
A
,
Matsubara
K
, et al
LGR5-positive colon cancer stem cells interconvert with drug-resistant LGR5-negative cells and are capable of tumor reconstitution
.
Stem Cells
2012
;
30
:
2631
44
.
45.
Park
YT
,
Jeong
JY
,
Lee
MJ
,
Kim
KI
,
Kim
TH
,
Kwon
YD
, et al
MicroRNAs overexpressed in ovarian ALDH1-positive cells are associated with chemoresistance
.
J Ovarian Res
2013
;
6
:
18
46.
Takahashi
H
,
Ishii
H
,
Nishida
N
,
Takemasa
I
,
Mizushima
T
,
Ikeda
M
, et al
Significance of Lgr5(+ve) cancer stem cells in the colon and rectum
.
Ann Surg Oncol
2011
;
18
:
1166
74
.
47.
Hsu
HC
,
Liu
YS
,
Tseng
KC
,
Hsu
CL
,
Liang
Y
,
Yang
TS
, et al
Overexpression of Lgr5 correlates with resistance to 5-FU-based chemotherapy in colorectal cancer
.
Int J Colorectal Dis
2013
;
28
:
1535
46
.
48.
de Lau
W
,
Barker
N
,
Low
TY
,
Koo
BK
,
Li
VS
,
Teunissen
H
, et al
Lgr5 homologues associate with Wnt receptors and mediate R-spondin signalling
.
Nature
2011
;
476
:
293
7
.
49.
Gao
P
,
Tchernyshyov
I
,
Chang
TC
,
Lee
YS
,
Kita
K
,
Ochi
T
, et al
c-Myc suppression of miR-23a/b enhances mitochondrial glutaminase expression and glutamine metabolism
.
Nature
2009
;
458
:
762
5
.
50.
Pinto
D
,
Gregorieff
A
,
Begthel
H
,
Clevers
H
. 
Canonical Wnt signals are essential for homeostasis of the intestinal epithelium
.
Genes Dev
2003
;
17
:
1709
13
.
51.
Pellegrino
L
,
Stebbing
J
,
Braga
VM
,
Frampton
AE
,
Jacob
J
,
Buluwela
L
, et al
miR-23b regulates cytoskeletal remodeling, motility and metastasis by directly targeting multiple transcripts
.
Nucleic Acids Res
2013
;
41
:
5400
12
.
52.
Zaman
MS
,
Thamminana
S
,
Shahryari
V
,
Chiyomaru
T
,
Deng
G
,
Saini
S
, et al
Inhibition of PTEN gene expression by oncogenic miR-23b-3p in renal cancer
.
PLoS One
2012
;
7
:
e50203
.
53.
Saoncella
S
,
Tassone
B
,
Deklic
E
,
Avolio
F
,
Jon
C
,
Tornillo
G
, et al
Nuclear Akt2 opposes limbal keratinocyte stem cell selfrenewal by repressing a FOXO-mTORC1 signaling pathway
.
Stem Cells
2014
;
32
:
754
69
.
54.
Grumolato
L
,
Liu
G
,
Haremaki
T
,
Mungamuri
SK
,
Mong
P
,
Akiri
G
, et al
beta-Catenin-independent activation of TCF1/LEF1 in human hematopoietic tumor cells through interaction with ATF2 transcription factors
.
PLoS Genet
2013
;
9
:
e1003603
.

Supplementary data