Author contributions: N.B.: conception and design, collection and assembly of data, data analysis and interpretation, and manuscript writing; V.B. and R.Z.: data analysis and interpretation; M.G. and I.L.: collection and assembly of data; J.R. and J.J.S.: provision of study material or patients; M.M.A.: data analysis and interpretation and writing manuscript; E.B., P.F., and J.G.-F.: conception and design, collection and assembly of data, data analysis and interpretation, manuscript writing, and final approval of manuscript.
Disclosure of potential conflicts of interest is found at the end of this article.
First published online in STEM CELLSEXPRESS September 21, 2011.
Many antitumor therapies affect rapidly dividing cells. However, tumor proliferation may be driven by cancer stem cells (CSCs), which divide slowly and are relatively resistant to cytotoxic drugs. Thus, many tumors may progress because CSCs are not sensitive to the treatment. In this work, we searched for target genes whose expression is involved in proliferation and chemoresistance of CSCs. Both of these processes could be controlled simultaneously by cell regulators such as microRNAs (miRNAs). Therefore, colonospheres with properties of CSCs were obtained from different colon carcinoma cells, and miRNA profiling was performed. The results showed that miR-451 was downregulated in colonspheres versus parental cells. Surprisingly, expression of miR-451 caused a decrease in self-renewal, tumorigenicity, and chemoresistance to irinotecan of colonspheres. We identified cyclooxygenase-2 (COX-2) as an indirect miR-451 target gene involved in sphere growth. Our results indicate that miR-451 downregulation allows the expression of the direct target gene macrophage migration inhibitory factor, involved in the expression of COX-2. In turn, COX-2 allows Wnt activation, which is essential for CSC growth. Furthermore, miR-451 restoration decreases expression of the ATP-binding cassette drug transporter ABCB1 and results in irinotecan sensitization. These findings correlate well with the lower expression of miR-451 observed in patients who did not respond to irinotecan-based first-line therapy compared with patients who did. Our data suggest that miR-451 is a novel candidate to circumvent recurrence and drug resistance in colorectal cancer and could be used as a marker to predict response to irinotecan in patients with colon carcinoma. STEM CELLS 2011;1661–1671
Solid tumors are composed of a mixture of cells with different characteristics that fulfil different roles in tumor growth. Currently, it is believed that only a small proportion of the cells within a tumor are capable of driving tumor proliferation . These cells are called tumor initiating stem cells or cancer stem cells (CSCs). CSCs can divide to yield a more differentiated cell and a daughter cell that maintains the same properties as the parental cell. This ability to self-renew enables CSCs to perpetuate the tumor. This novel view of cancer growth may have tremendous implications in the treatment of all kinds of tumors. Anti-tumor therapies that do not target CSCs may lead to a reduction of the tumor mass but not the regression of the tumor. In fact, therapies are generally based on drugs that affect rapidly dividing cells while CSC division is slow . Furthermore, CSCs are relatively resistant to cytotoxic systemic therapies . Thus, many tumors may progress because CSCs are not sensitive to the treatment. Novel therapies should be developed that target CSCs . These include induction of CSC differentiation to cells sensitive to conventional therapy or targeting of genes expressed only in CSCs that allow self-renewal and chemoresistance. Ideal targets could be cell regulators that simultaneously control the stemness and the resistance of CSCs.
microRNAs (miRNAs) are cell regulators capable of controlling the expression of several genes at the same time . Generally, miRNA precursors are transcribed as long primary transcripts named pri-miRNAs which can contain one or more immature miRNAs. The pri-miRNA is processed sequentially to a double-stranded miRNA that is loaded into the RNA-induced silencing complex (RISC). One of the strands, called the guide strand, remains in RISC as a mature miRNA, whereas the other strand is degraded. Once this happens, the miRNA complex is functional for regulation. The miRNA guides RISC to the target RNA inducing gene silencing by translational repression and/or deadenylation and degradation of mRNA targets. miRNA regulation of gene expression contributes to essential processes for the organism such as differentiation, apoptosis, or proliferation . In recent years, several miRNAs have been shown to play a relevant role in the biology of human cancer . Moreover, recent findings suggest that miRNAs might be involved in the tumorogenicity and resistance to chemotherapy of CSCs [8, 9]. For instance, it has been suggested that overexpression of miR-215 and miR-140 could control the slow proliferation and the chemoresistance of colon cancer CSCs [8, 9].
Colorectal cancer (CRC) is the second cause of cancer-related death in most industrialized countries. It has been recently shown that CSCs can be isolated from colorectal tumors and colon cancer cell lines . Colon cancer CSCs can be propagated in suspension as colon spheres, which express several stem cell markers and, as expected, have greater resistance to chemotherapy and show the ability to increase tumor initiation in vivo [3, 11].
In this study, we searched for miRNAs that control the ability of colon cancer CSCs to self-renew and resist drugs. We found that both functions are controlled by miR-451. Upregulation of miR-451 resulted in reduction of colon sphere formation and growth, inhibition of tumorigenicity in vivo, and sensitization to the irinotecan active metabolite, SN38. We show that the decrease of colorectal CSC formation and growth could result from miR-451-mediated downregulation of cyclooxygenase-2 (COX-2) and Wnt pathway, which seems essential to maintain cell stemness. Furthermore, we show that the targeting of ATP-binding cassette (ABCB1) by miR-451 could decrease SN38 resistance in CSCs. Our findings should be valuable in the development of novel therapies that target CSCs in CRCs.
MATERIALS AND METHODS
Cells and Cell Culture Conditions
The human CRC cell lines were obtained from American Type Culture Collection, Middlesex, UK, http://www.lgcstandards-atcc.org/. All cells were cultured in RPMI-1640 (Gibco, Invitrogen Corporation, Grand Island, NY, http://www.invitrogen.com/site/us/en/home/brands/Gibco.html) supplemented with 10% fetal bovine serum and 1% penicillin–streptomycin at 37°C in a 5% CO2 atmosphere. The human CSCs and the parental cells obtained from human colon cancer tissue were purchased from CelProgen, San Pedro, CA, http://www.celprogen.com/. Human primary CSCs express CD133, CD44, CD34, CD 10SSEA3/4, Oct4, Alkaline Phosphatase, Aldehyde Dehydrogenase, Telomerase, Sox2, cKit, and Lin28 and less than 1,000 cells are sufficient to form tumors. Human parental cells express Vimentin, Variable S100, CEA, Galactosyl Transferase II, CK-7, CK-20, smooth muscle actin, Bcl2, Ki-67, P504S, and Mucin (MUC-1 and MUC-3). For the generation of colon spheres, 2,000 or 200 cells were plated per 24- or 96-well, respectively, on ultra-low attachment plates (Corning, Corning, NY, http://www.corning.com/index.aspx). CRC cells were grown in serum-free medium Dulbecco's modified eagle medium (DMEM)/F12+GlutMAX-I (Gibco) containing 1% N2 (Gibco), 2% B27 (Gibco), 20 ng/ml human fibroblast growth factor (FGF)-2 (Sigma, St. Louis, MO, http://www.sigmaaldrich.com), and 50 ng/ml epidermal growth factor (EGF) (Sigma) and supplemented with medium after 3 days of incubation. After 7 days, plates were analyzed for colon sphere formation. When indicated, the colon spheres were dissociated to single cells and then reseeded to yield the second generation of colon spheres. For quantification of the number of cells per colon sphere, colon spheres were collected with a 40-μm sieve (BD Biosciences, San Jose, CA, http://www.bdbiosciences.com) and dissociated with trypsin to give single cell suspensions. Viable cells were counted in a Neubauer chamber using trypan blue exclusion.
Microscopically confirmed CRC tumor samples were obtained from 35 patients homogeneously treated with an irinotecan-based first-line therapy for metastatic colorectal cancer (mCRC) . The characteristics of the patients are shown in Supporting Information Table 1. Total RNA from paraffin-embedded CRC tissues was extracted with the RecoverAll Total Nucleic Acid Isolation Kit (Ambion).
Fifty pmol of pre-miR-451, pre-miR-144, pre-miR-144*, or the scrambled control (Ambion, Austin, TX, http://www.ambion.com/) were transfected into CRC cell lines grown in six-well dishes. Small interfering RNAs (siRNAs) specific to human ABCB1 (ON-TARGETplus SMART pool L-003868-00-0005) and control siRNA (ON-TARGETplus Non-targeting pool D-001810-10-05) were designed and synthesized by Dharmacon, Lafayette, CO, http://www.dharmacon.com/Home.aspx?id=642. Transfection was accomplished with Lipofectamine 2000 (Invitrogen, Grand Island, NY, http://www.invitrogen.com/site/us/en/home.html). Transfection efficiency (>95%) was confirmed with the use of the Silencer 6-carboxy-fluo-rescine (FAM)-labeled Negative Control (Ambion). Retroviral vector containing pre-miR-144/451 cluster was obtained from Source Bioscience GeneService (Cambridge, UK, http://www.lifesciences.sourcebioscience.com/). To obtain viral particles, packaging cells were transfected using the calcium phosphate method. Empty vector was used to obtain mock infective viral particles. For infection, DLD1 cells were incubated with viral supernatants. After 48 hours of infection, cell populations were incubated in medium containing the appropriate antibiotic, and antibiotic-resistant pools were expanded.
To evaluate Wnt activity, a T-cell factor (TCF) luciferase activity assay was performed. In brief, cells were transiently transfected with a control plasmid expressing Renilla (pRL-CMV, Promega, Madison, WI, http://www.promega.com/) and 5 μg of either plasmid TOPflash or FOPflash TCF reporters (Upstate Biotechnology, Billerica, MA, http://www.millipore.com/company/cp1/redirect-ab) using Lipofectamine 2000 (Invitrogen). After 24 or 48 hours of transfection, cell extracts in lysis buffer were monitored for luciferase activity using Luciferase Assay System (Promega). Light units were recorded using a luminometer. Transfection efficiency was normalized by Renilla luciferase. All experiments were performed in triplicate and repeated independently thrice.
Cell Cytotoxicity Assay
5-Fluoruracil (FU), oxaliplatin (LOHP), SN38, and celecoxib were obtained from Sigma and were dissolved according to the manufacturer's instructions. Cells and colon spheres were plated in 96-well plates (10,000 cells per well) and treated with serial dilutions of the agents, in triplicates. Cell viability was measured with a CellTiter 96 Aqueous Assay Kit (3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium, inner salt (MTS); Promega) 72 hours after treatment with 5-FU, LOHP, or SN38 or 7 days postincubation with celecoxib. The percentage of cell survival in treated samples was normalized with untreated controls. IC50 was calculated with GraphPad Prism 5.0 (San Diego, CA).
Total RNA was extracted from the cells with TRIzol (Invitrogen). miRNA profiling was carried out using Taqman probes and primer sets in an Applied Biosystems PRISM 7900HT Fast Real-Time PCR System (Applied Biosystems, Carlsbad, CA, http://www.appliedbiosystems.com/absite/us/en/home.html) as described . Individual expression of miR-451, miR-144, and miR-144* was measured using the qRT-PCR assay and calculated using the comparative Ct method (2−ΔΔCT) using RNAU6B as the reference gene. Taqman PCR reactions were performed with ABI PRISM 7300 Sequence Detection System (Applied Biosystems) using primers and probes purchased from Applied Biosystems Hs01009259_m1 (CD133), Hs00174139_m1 (CD44), Hs00158980_m1 (EpCam), Hs00233455_m1 (CD166), Hs02379687_s1 (CD24), Hs00236988_g1 (macrophage inhibitory factor [MIF]), Hs01573476_g1 (COX-2), Hs00184500_m1 (ABCB1), and Hs99999901_s1 (RNA18S). Gene expression of CD133, CD44, EpCam, CD24, CD166, MIF, COX-2, and ABCB1 was calculated using the comparative Ct method (2−ΔΔCt) normalized to the endogenous RNA18S control.
Hematoxylin–eosin stained sections were dehydrated and mounted in synthetic resin. Inhibition of endogenous peroxidase was performed with 3% H2O2 (5 minutes) and slides were heated in citrate (pH = 6) or pepsin. After washes in tris buffered saline (TBS), slides were incubated with specific antibodies against CD133 (C24B9 rabbit IgG, Cell Signaling, Danvers, MA, http://www.cellsignal.com/), CD44 (Ab-4 mouse IgG2a, LabVision, Fremont, CA, http://www.labvision.com/), EpCam (VU1D9 mouse IgG1, Cell Signaling), CD24 (Ab-2 mouse IgM, LabVision), and CD166 (MOG/O7 mouse IgG2a, Novocastra, Barcelona, Spain, http://www.leica-microsystems.com/products/histology-systems/novocastra-reagents/), or isotype-matched controls at appropriate dilutions overnight at 4°C. Then sections were incubated with Dako Advance horseradish peroxidase (HRP) Detection System (DakoCytomation, Glostrup, Denmark, http://www.dako.com/es/index.htm). Staining was revealed using DAB substrate and counterstained with hematoxylin.
Cells were harvested 48 hours after transfection with plasmids expressing miR-451 precursor, scrambled control, siABCB1, or control siRNA. Cells were washed with phosphate buffered saline (PBS) and lysed for 30 minutes at 4°C in a buffer containing 1% Triton X-100 and complete protease inhibitor cocktail (Roche, Barcelona, Spain, http://www.roche-applied-science.com/index.jsp). Lysates were cleared by centrifugation, the amount of protein was quantified by the Bradford assay, and equal amounts were separated by 10% sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) gels (Invitrogen) and transferred onto nitrocellulose membranes (Bio-Rad, Hercules, CA, http://www.bio-rad.com/). Inmunoblotting was performed and antibodies specific for ABCB1 (D-11 mouse IgG2b, Santa Cruz Biotechnology, Santa Cruz, CA, http://www.scbt.com/) and β-tubulin (Tub 2.1, mouse IgG1, Sigma) were detected using HRP-conjugated anti-mouse antibody (Sigma) and visualized by chemoluminescence detection systems (PerkinElmer LAS, Waltham, MA, http://www.perkinelmer.com/).
In Vivo Studies
Animal studies were performed according to institutional guidelines. Viable cells were injected subcutaneously into both flanks of 6–8-week-old female BALB/cA-Rag2−/−γc−/− mice. Tumors were measured every 2–3 days. Tumor volumes were calculated by using the equation, V (in mm3) = (A × B2)/2, where A is the largest diameter and B is the diameter perpendicular to A. Tumor samples were paraffin-embedded, and tissue sections were stained with hematoxylin and eosin.
All experiments were carried out at least three times. Two-way analysis of variance (ANOVA) followed by a Bonferroni multiple comparison test and two-tailed t tests were used to analyze the in vitro and in vivo data. The statistical analysis was performed with the GraphPad Prism software and with the SPSS v. 13. p < .05 was defined as statistically significant and is indicated with an asterisk.
Colon Spheres Exhibit Properties of CSCs
Initially, we examined the ability of CRC cell lines to form colon spheres in serum-free media supplemented with growth factors. After 7 days under these conditions, DLD1, HT29, SW620, and LS513 cells lost adherence to the plate and clustered in multicellular spheroids that closely resembled colon spheres. The spheres observed ranged from 100 to 200 μm of diameter and contained around 50 cells if they were derived from LS513 cells or 107–140 cells if they were derived from DLD1, HT29, or SW620 (Fig. 1A and Supporting Information Fig. 1). LoVo and RKO did not form these structures but rather groups of individual cells with a diameter larger than 200 μm. Colon spheres could be dissociated to single cells and reseeded to yield the second generation of colon spheres, which had morphological characteristics similar to the first generation of colon spheres (Supporting Information Fig. 1). To determine whether colon spheres were enriched in CSCs, we analyzed the expression of several putative CSC markers by qRT-PCR in parental and colon sphere cells. As shown in Figure 1B, DLD1, HT29, LS513, and SW620 colon spheres expressed higher levels of stem cell markers such as CD133, CD44, EpCam, CD166, and CD24 than the parental cells. In line with the morphology studies, we could not detect changes in the expression levels of these markers in LoVo and RKO colon sphere cells (Fig. 1B). We used immunohistochemistry to determine that most of the cells in the colon spheres expressed the CSCs markers (Fig. 1C and Supporting Information Fig. 2). Similar results were obtained with colon spheres derived from human colon CSCs (Supporting Information Fig. 3).
To determine whether colon spheres have a higher tumor initiation capacity than parental cells, we inoculated 1 × 104 viable parental DLD1 cells or DLD1 colon sphere cells into both flanks of BALB/cA-Rag2−/−γc−/− mice (n = 7 mice per group; Fig. 1D). Colon sphere cells used were obtained from a first or a second generation. We found that colon sphere cells generated significantly larger tumors than parental cells and that injection of the second generation of colon spheres resulted in the largest tumors. Histopathological analysis of tumor tissues indicated that xenografts formed by colon spheres resemble primary human CRCs (Supporting Information Fig. 2D).
Finally, we examined the chemoresistance of the colon spheres and parental CRC cell lines to 5-FU, LOHP, and SN38. As shown in Supporting Information Table 2, colon spheres derived from the first and second generation showed a marked increase in resistance to the three drugs tested when compared with the parental cell line. Similarly to what has been already shown , we found that resistance to irinotecan-mediated inhibition of cell growth was at least partly due to the decreased apoptosis observed in colon tumor spheres (data not shown). Together, these results demonstrated that colon spheres exhibit the main characteristics of CSCs.
Expression Profiling Identifies MiR-144/451 Cluster Differentially Expressed in Colon Spheres
To identify miRNAs differentially expressed in colon spheres, we compared the expression profiles of 754 individual miRNAs between DLD1-derived colon spheres and the parental DLD1 cell line using the Taqman real-time multiplex RT-PCR method. Twelve abundant miRNAs, with Cts ≤30, were differentially expressed at ≥4-fold (Supporting Information Table 3). Two of these, miR-451 and miR-144* were downregulated in colon spheres versus parental cell lines and are expressed from the same cluster. The miR-144/451 cluster has been previously related with differentiation  and cancer . In addition, low miR-451 expression has been associated with higher risk of disease relapse in gastric cancer patients and is involved in radiation response . Therefore, we decided to further investigate the role of the miR-144/451 cluster in the phenotype of CSCs. The cluster expresses miR-144* and miR-451, but also miR-144, which normally behaves as the guide strand of miR-144*. However, miR-144 was not present in version 2.0 of the miRNA panel. Therefore, using qRT-PCR, we validated the expression of miR-451, miR-144, and miR-144* in parental and colon spheres generated from DLD1, HT29, LS513, and SW620 CRC cell lines or from human colon CSCs. A significant decrease in miR-144/451 cluster expression levels in colon spheres of all CRC cells tested in relation to parental cells was found (Fig. 2A and Supporting Information Fig. 4). As expected, there was no variation in miR-144/451 cluster expression in LoVo and RKO parental cells compared with the aggregates that do not express CSC markers formed by these cells after incubation in colon sphere conditions (data not shown).
A major feature of CSCs is their strong ability to establish themselves as tumor xenografts. To determine whether expression of the miR-144/451 cluster was able to reverse the high capacity of colon spheres to initiate tumor formation, we used a retroviral system that expresses miR-144/451 cluster over the long term. Thus, we isolated DLD1 colon spheres transduced with the retrovirus, and we verified that they overexpressed miR-144, miR-144*, and miR-451 (Supporting Information Fig. 5A). We then inoculated 1 × 104 cells obtained from the first (Fig. 2B) or second (Fig. 2C) generation of DLD1 colon spheres or from DLD1 colon spheres overexpressing miR-144/451 into both flanks of BALB/cA-Rag2−/−γc−/− mice (n = 7 mice per group). The DLD1 colon spheres that overexpressed miRNA-144/451 generated significantly smaller tumors than DLD1 colon spheres from the first or second generation. Histopathological analysis indicated that the tumors were poorly differentiated and glandular lumens were seen only occasionally (Supporting Information Fig. 5B). These results suggest that the miR-144/451 cluster may play a pivotal role in colon sphere tumorigenicity.
We were excited at the possibility that all miRNAs from the miR-144/451 cluster collaborate to inhibit colon sphere growth. Alternatively, growth inhibition could be exerted by only one of the miRNAs from the cluster. To distinguish between these two possibilities, we transfected CRC cell lines with pre-miRNAs expressing miR-144, miR-144*, miR-451, or a scrambled control, and then verified miRNA overexpression (Supporting Information Fig. 5C) and evaluated colon sphere formation. Overexpression of miR-451 significantly reduced DLD1 colon sphere formation (Fig. 3A). Moreover, these colon spheres were smaller than those transfected with the scrambled miRNA (Fig. 3B). These results were confirmed by a MTS assay plating a limited number of cells in a low-attachment 96-well plate (Fig. 3C). Similar results were observed with a second generation of DLD1-derived colon spheres (Fig. 3A–3C) or using HT29, LS513, and SW620 CRC cell lines (Fig. 3D–3F). However, these changes were not observed in cells expressing miR-144 or miR-144*, indicating that only miR-451 is involved in colon sphere formation and self-renewal of colon sphere cells.
COX-2 Is Involved in the Maintenance of Colon Spheres and Is Regulated Indirectly by MiR-451 Via MIF
Wnt activity defines colon CSCs . Therefore, we hypothesized that miR-451 could downregulate Wnt to decrease self-renewal and growth of colon spheres. We failed to identify canonical binding sites for miR-451 in Wnt genes. Nevertheless, we decided to evaluate Wnt activity in parental cells or colon spheres transfected with a pre-miRNA that expresses miR-451 or a scrambled control. Interestingly, miR-451 expression significantly reduced Wnt signaling compared to control cells (Fig. 4A and Supporting Information Fig. 6A). To decipher the molecular mechanism that allows miR-451 to affect Wnt, we searched for miR-451-validated direct targets (Supporting Information Table 3). One of the validated direct targets of miR-451 is the macrophage migration inhibitory factor (MIF) . Overexpression of miR-451 downregulates MIF mRNA and protein in DLD1 cells . Furthermore, miR-451 inhibits expression of renilla reporters with the 3′ untranslated region (UTR) of MIF but does not affect expression of the same reporter with a mutation in the miR-451 seed-binding sequence . MIF has been involved in the regulation of COX-2 transcription [18–20] and COX-2 activates Wnt signaling by inhibiting β-catenin degradation . Therefore, we analyzed MIF mRNA and protein levels in 'parental and colon sphere cells, and we determined if these levels fluctuated when miR-451 was overexpressed. We found that MIF expression was higher in colon spheres than in parental cells and that miR-451 overexpression downregulated MIF (Fig. 4B and data not shown). Similar results were observed when COX-2 mRNA expression was evaluated (Fig. 4B). Thus, decreased levels of miR-451 in colon spheres allow overexpression of MIF and COX-2 and this correlated with increased Wnt activity. To determine whether MIF and COX-2 affect Wnt signaling also in colon spheres, we compared TCF activity in parental cells, control spheres, or colon spheres of the first or second generation expressing a siRNA that inhibits MIF expression or spheres incubated with the COX-2-specific inhibitor celecoxib. The results confirmed that blockade of MIF or COX-2 results in a drastic inhibition of Wnt signaling also in colon spheres (Supporting Information Fig. 6B, 6C).
To determine whether COX-2 function is important in colon sphere maintenance, we analyzed the effect of the COX-2 inhibitor celecoxib on colon spheres. Treatment with celecoxib inhibited the growth of DLD1- and LS513-derived colon spheres in a concentration-dependent manner, as measured by the number of colon spheres obtained (Fig. 4C) and the number of cells per colon sphere (Fig. 4D). Moreover, this celecoxib effect was higher in colon spheres than in parental cell lines or in spheres expressing miR-451 (Supporting Information Fig. 7 and data not shown). Similar results were obtained when MIF and COX-2 expression was evaluated in colon spheres derived from human colon CSCs or when these cells were incubated with celecoxib (Supporting Information Fig. 8). These results indicate that COX-2 may play an important role in colon sphere maintenance and can be targeted indirectly by miR-451 via MIF.
Finally, we wondered whether miR-451 expression has any role in CSC chemoresistance. To this aim, we determined the IC50 of 5-FU, LOHP, and SN38 in control or miR-451-expressing DLD1 colon spheres. We detected a significant decrease in the chemoresistance to SN38 of DLD-1 colon spheres overexpressing miR-451 compared to control cells. This effect was observed with the first (1,286 nM vs. 392 nM [p = .031]) and second (1,526 nM vs. 590 nM [p = .004]) generation of colon spheres (Fig. 5A). These results were confirmed in LS513 or HT29-derived colon spheres (Fig. 5B and data not shown). In contrast, no changes were detected in the resistance of colon spheres to 5-FU and LOHP despite miR-451 overexpression (Supporting Information Fig. 9).
ABCB1 Is Involved in Increasing SN38 Resistance and Is Regulated by MiR-451
It has been reported that ABCB1 is involved in the active efflux of SN38  and that miR-451 directly targets ABCB1 . Consequently, we hypothesized that miR-451 may contribute to increase the sensitivity to SN38 treatment by reducing ABCB1 expression. To verify these results in our biological system, we first determined whether miR-451 expression affects ABCB1 accumulation in CRCs. ABCB1 levels were quantified in parental cells or colon spheres transfected with pre-miRNAs expressing miR-451 or a scrambled control. Overexpression of miR-451 significantly decreased the expression of ABCB1 mRNA and protein levels (Fig. 6A, 6B and Supporting Information Fig. 10A). To confirm the role of ABCB1 in the regulation of chemosensitivity to SN38 in colon spheres and CRC cell lines, we used a siRNA specific to ABCB1. Downregulation of ABCB1 by siRNA (Supporting Information Fig. 10B) resulted in an increased sensitivity to SN38 as compared to control cells (1,329 nM vs. 464 nM, p = .003 for DLD1 and 95.5 nM vs. 45.9 nM, p = .03 for LS513; Fig. 6C and Supporting Information Fig. 10C). Our results demonstrate that upregulation of ABCB1 transporter could account for SN38 resistance in colon spheres and, importantly, ABCB1 expression can be downregulated by miR-451.
Correlation of CD44 and MiR-451 Expression with Response to Irinotecan-Based Chemotherapy in Metastatic CRC Patients
To translate our findings into the clinical setting, we measured the expression of cancer stem cell markers (CD133 and CD44) and miR-451 in a cohort of CRC tumor samples homogeneously treated with irinotecan-based first-line therapy for metastatic CRC  (Supporting Information Table 1). As shown in Figure 7, CD44 mRNA levels were significantly higher in tumor samples from patients who did not respond to therapy than in responders (Mann–Whitney U test, p = .012). The same trend was observed for CD133 mRNA but the differences did not reach statistical significance (Mann–Whitney U test, p = .08). Importantly, miR-451 expression was significantly lower in tumor samples from patients who did not respond to therapy as compared to those from responders (Mann–Whitney U test, p = .029).
CSCs can be studied in tissue culture in search of novel markers and therapeutic targets that improve diagnosis, prognosis, and treatment of tumors such as CRC. In this study, we obtained CRC-derived colon spheres to show that they share key characteristics of CSCs (Fig. 1, Supporting Information Figs. 1–3 and Supporting Information Table 2): (i) they express higher levels of several CSC markers, as detected by quantitative real time polymerase chain reaction (qRT-PCR) and immunohistochemistry (IHC); (ii) they have increased resistance to the chemotherapy most widely used in CRC treatment; (iii) they have the capacity to self-renew to yield a second generation of colon spheres; and (iv) they have higher potential to form tumors in mice indicating that they can both self-renew and aberrantly differentiate to generate the bulk of cells within the tumor. In CRC, stem cell renewal and differentiation have been linked to several signaling pathways such as Wnt and tumor growth factor (TGFb) . We hypothesized that miRNA misregulation could affect these pathways and/or other molecular mechanisms that result in the CSC phenotype.
Expression profiling showed that four abundant miRNAs are upregulated and eight downregulated in colon spheres versus parental cells (Supporting Information Table 3). Of the upregulated miRNAs the only validated target is Rb, whose expression should decrease to allow proliferation of CSCs. Validated targets of downregulated miRNAs are related to the Wnt, TGFb, and nuclear factor kappa-light-chain-enhancer of activated B cells (NFkb) pathways. Target genes are also involved in genome stability, growth, and epigenetic regulation. All of these genes could be important for CSC viability.
miRNA expression has been analyzed by other authors in CRC CSCs [9, 25, 26]. When we compare our results with those published, we find some differences and some similarities. We failed to detect reproducible overexpression of miR-215 or miR-140 . However, we also found overexpression of miR455-3p and miR455-5p [25, 26]. In fact, we have determined that miR-455-5p levels increase twofold in spheres derived from DLD1, LS513 or SW620 cell lines compared to parental cells (data not shown). We believe that differences in the expression pattern of particular miRNAs may represent differences in the CSCs of origin. Theoretically, different combinations of miRNAs could lead to a CSC phenotype. However, we wanted to study in more detail a miRNA whose expression is relevant for different CSCs. We chose miR-451 for different reasons: (i) all miRNAs expressed from the miR144/451 cluster were downregulated; (ii) downregulation of all members of the cluster was validated in colon spheres derived from human colon CSCs or from four different cell lines and downregulation of miR451 was observed in CRC nonresponding patients (Figs. 2, 7 and Supporting Information Fig. 4); (iii) upregulation of the cluster decreased tumorigenicity (Fig. 2B). We believe that downregulation of miR-451 and not that of the other members of the cluster, is relevant for the phenotype of CSCs. Overexpression of miR-451 decreases formation and growth of colon spheres and resistance to SN38 (Figs. 3 and 5). Moreover, miR-451 has been implicated in several tumors [27, 28] and miR-451 expression is also downregulated in glioma stem cells  and in relapsed patients with head and neck squamous cell carcinoma . Low expression of miR-451 is also significantly associated with a worse outcome in resected gastric cancer patients treated with adjuvant chemoradiation . In fact, miR-451 has been associated with radioresistance in lung , doxorubicin resistance in breast cancer , and to imatinib treatment in glioma CSCs . Furthermore, as miR-451 controls the balance of cell proliferation and migration in response to metabolic stress , it has been postulated to play a prominent role in the differentiation of the erythroid compartment [31–33] and contributes to the formation of basolateral polarity in epithelial cells .
We have addressed the molecular mechanisms involved in miR-451-mediated effects. Our results show that downregulation of miR-451 increases ABCB1 and MIF (Fig. 4B, 6A, 6B and Supporting Information Fig. 10A). Both genes are validated targets of miR-451 [16, 23]. MIF has been shown to increase expression of COX-2 , which is essential for formation and growth of colon spheres (Fig. 4C–4D, Supporting Information Figs. 7 and 8). COX-2 could have a direct impact on colon spheres, as shown recently for breast CSCs . This would provide an explanation for the chemopreventive and anticancer activities of selective COX-2 inhibitors . In fact, COX-2 is overexpressed in several tumors. Furthermore, COX-2 can prevent β-catenin degradation and, as a result, activate Wnt signaling, which is strongly associated with stemness and self-renewal of CSCs [17, 21] (Fig. 4A and Supporting Information Fig. 6).
We show that expression of miR-451 leads to a decrease in ABCB1 (Fig. 6A, 6B and Supporting Information Fig. 10A). This is in agreement with a previous report that shows that ABCB1 is a direct target of miR-451 in breast cancer cells [16, 23]. miR-451 has also been shown to activate ABCB1 expression in ovarian cancer by an unknown mechanism, suggesting that the miR-451 effect on ABCB1 may be different in different cells . miR-451 downregulation in colon cancer cells allows high levels of ABCB1 and therefore, a more efficient efflux of anticancer drugs such as irinotecan-derived SN38 metabolite  (Figs. 5, 6 and Supporting Information Fig. 10). During the process of review of our work, a complementary study was published that also shows that colon sphere cultures obtained from human colorectal tumors are resistant to irinotecan and that inhibition of ABCB1 allows irinotecan eradication of tumor initiating cells . The authors describe that ABCB1 is expressed in a subpopulation of more differentiated proliferation-competent tumor cells. These cells are able to protect from irinotecan neighbor ABCB1 negative tumor initiation cells. Further experiments will be required to determine the levels of miR-451 in the different subpopulations.
The impact of miR-451 on ABCB1 and irinotecan resistance may have clinical relevance. In fact, we detected an association between miR-451 levels and response to treatment in a cohort of human CRC tumors (Fig. 7). Patients who responded to irinotecan-based therapy harbor higher expression levels of miR-451 than patients who did not respond to the treatment. Given the widespread use of irinotecan in the management of CRC, our results may have a clear impact on the identification of patients with a higher likelihood of responding to this agent. Evidently, before reaching definitive conclusions, we need to confirm these results in a larger cohort of patients. However, miR-451 has already been proven to be a marker for relapse in gastric cancer after curative surgery in patients homogeneously treated with adjuvant chemoradiation and for increased radiation response in gastrointestinal cancer cells lines . These findings suggest that miR-451 might be a novel molecular marker that could prove useful in the selection of patients that respond to treatment with irinotecan and/or radiation.
miR-451 could not only be useful as a marker but also serve as a target for the development of novel therapeutic strategies to overcome drug resistance and tumor growth. Given that low expression of miR-451 is important for CRC chemosensitivity and maintenance of critical properties of the CSC phenotype, novel therapies could be developed that increase miR-451 expression in all tumor cells or in CSCs. Putative options could be the use of gene therapy vectors that express miR-451, drugs that reactivate expression of the miR-144/451 cluster within the tumor or methods that induce capture of artificial exosomes loaded with miR-451 by tumor cells. In conclusion, our results suggest that miR-451 influences the efficiency of irinotecan by a dual mechanism involving ABCB1 expression and the maintenance of the crucial CSC phenotype. Therefore, miR-451 may be a promising candidate as a prognostic marker and as a target for the development of new therapies for the treatment of CRC.
Combination of irinotecan with fluorouracil is one of the gold standards for the treatment of metastatic colon carcinoma. However, the success of irinotecan and other antitumor therapies may be limited when they do not target CSCs, which may govern tumor proliferation and harbor resistance to chemotherapy. In this work, we show that expression of miR-451 causes a decrease in self-renewal, tumorigenicity, and chemoresistance to irinotecan of CSCs. In fact, expression of miR-451 is downregulated in patients who did not respond to irinotecan-based first-line therapy compared with responding patients and in colon spheres with CSC properties versus parental cells. We also analyzed the molecular mechanisms involved in the miR-451 effect. miR-451 targets COX-2 and ABCB1. COX-2 is involved in colon sphere growth, because celecoxib, a specific COX-2 inhibitor, decreases growth of colon spheres. ABCB1 drug transporter is a direct target of miR-451 essential for irinotecan resistance. Further development of the results described in this work could have a great impact on clinical practice. miR-451 and COX-2 seem promising targets to develop therapies aimed at halting CRC progression and chemoresistance due to their implication in CSC function. In fact, celecoxib could be beneficial to avoid recurrence after CRC resection. Moreover, miR-451 could be used as a marker to predict response to irinotecan in patients with colon carcinoma.
We thank Dr. Paul Miller for English editorial work. This work was supported by the Fondo de Investigacion Sanitaria (PI10/02518) through the “UTE project CIMA” and by the project RNAREG (CSD2009-00080), funded by the Ministry of Science and Innovation under the program CONSOLIDER INGENIO 2010.