MicroRNA-21 induces resistance to 5-fluorouracil by down-regulating human DNA MutS homolog 2 (hMSH2)

Abstract

The overexpression of microRNA-21 (miR-21) is linked to a number of human tumors including colorectal cancer, where it appears to regulate the expression of tumor suppressor genes including p21, phosphatase and tensin homolog, TGFβ receptor II, and B-cell leukemia/lymphoma 2 -associated X protein. Here we demonstrate that miR-21 targets and down-regulates the core mismatch repair (MMR) recognition protein complex, human mutS homolog 2 (hMSH2) and 6 (hMSH6). Colorectal tumors that express a high level of miR-21 display reduced hMSH2 protein expression. Cells that overproduce miR-21 exhibit significantly reduced 5-fluorouracil (5-FU)-induced G2/M damage arrest and apoptosis that is characteristic of defects in the core MMR component. Moreover, xenograft studies demonstrate that miR-21 overexpression dramatically reduces the therapeutic efficacy of 5-FU. These studies suggest that the down-regulation of the MMR mutator gene associated with miR-21 overexpression may be an important clinical indicator of therapeutic efficacy in colorectal cancer.

Colorectal cancer (CRC) is one of the most frequently occurring cancers in United States with more than 140,000 new cases and about 50,000 deaths expected to occur in 2010 (1). 5-fluorouracil (5-FU)-based chemotherapy represents the gold standard for CRC treatment both in the adjuvant and metastatic setting (2). However, primary or acquired resistance to pyrimidine analog treatments is a common problem in the management of CRC patients (2). These observations highlight the need for a better understanding of resistance mechanisms and more effective therapies.

MicroRNAs are a class of small, noncoding RNAs that act as posttranscriptional regulators of gene expression and cell homeostasis (3). Overexpression of miR-21 is a common trait of many solid and hematological malignancies (4). miR-21 overexpression has been found in blood and stool samples from patients affected by CRC (5, 6). Moreover, miR-21 overexpression is associated with poor benefit from 5-FU adjuvant chemotherapy in stage-II and -III CRC (7).

The mismatch repair (MMR) system is involved in DNA damage recognition and repair. Human mutS homolog 2 (hMSH2) and human mutL homolog 1 (hMLH1) function as core MMR proteins and form heterodimers with protein homologs hMSH3 or hMSH6 and hMLH3 or hPMS2, respectively (8). Heterodimer formation is fundamental for the DNA damage recognition and represents a crucial step for the stability of the MMR protein homologs (9). Defects in MMR proteins have been associated with reduced or absent benefit from 5-FU adjuvant chemotherapy in clinical trials (10). MMR impairment appears to cause reduced incorporation of 5-FU metabolites into DNA, leading to reduced G2/M arrest and apoptosis after 5-FU treatment (11, 12).

Results

miR-21 Directly Targets hMSH2 and hMSH6 Protein Expression.

In silico analysis suggested that miR-21 might target hMSH2 [National Center for Biotechnology Information (NCBI) NM_000249.2] and hMSH6 (NCBI NM_000179.2) mRNA (TargetScan, Whitehead Institute, MIT; Fig. 1A) (13). We identified putative binding sites for miR-21 in both the hMSH2 and the hMSH6 3′ UTR. We examined the effect of miR-21 expression on endogenous hMSH2 and hMSH6 mRNA expression in CRC Colo-320DM and SW620 cells. Both cell lines display low basal miR-21 expression. We transfected these cell lines with miR-21 precursor or a scrambled miR precursor control (Fig. S1). Overexpression of an siRNA specific to hMSH2 (anti-MSH2) or hMSH6 (anti-MSH6) did not affect the levels of miR-21 (Fig. S1A). The mRNA levels of hMSH2 and hMSH6 were unaffected by overexpression of miR-21 (Fig. 1B). In contrast, anti-MSH2 and anti-MSH6 siRNA specifically reduced the expression of hMSH2 and hMSH6 mRNA, respectively (Fig. 1B). We note a consistent reduction in the expression of hMSH6 mRNA with the anti-MSH2 siRNA. This reduction could be a result of degenerate hybridization of the anti-MSH2 siRNA with the hMSH6 mRNA or of reduced hMSH6 mRNA stability resulting from the diminished heterodimeric protein partner hMSH2. These results suggest that miR-21 overexpression does not affect the mRNA levels of hMSH2 or hMSH6.

Fig. 1.

MSH2 and MSH6 are direct targets of miR-21. (A) miR-21 predicted seed regions in the hMSH2 and hMSH6 3′ UTR are shown. (B) Colo-320DM and SW620 cells were transiently transfected with miR-21, scrambled-miR, siRNA anti-MSH2, or anti-MSH6 for 48 h. hMSH2 and hMSH6 mRNA expression was analyzed by real-time PCR. (C) Western blotting analysis of miR-21–dependent down-regulation of both hMSH2 and hMSH6. (D) HCT-116, SW480, and RKO cells that contain high endogenous levels of miR-21 were transfected with an LNA anti-miR-21 or anti-miR control for 48 h followed by Western blotting analysis of hMSH2 and hMSH6 protein. (E) hMSH2 and hMSH6 3′ UTRs (Luc-WT) and hMSH2 and hMSH6 3′ UTRs containing a deletion of the miR-21 target site (Luc-mutant) were subcloned downstream of the luciferase genes and were cotransfected with miR-21 or scrambled miR. Luciferase activity was recorded after 24 h. Data represent the mean ± SD from at least three determinations from four independent transfections. *P < 0.01.

We examined the protein levels of hMSH2 and hMSH6 following transfection of miR-21 in Colo-320DM and SW620 cells by Western blot analysis (Fig. 1C and Fig. S1B). hMSH2 and hMSH6 proteins were significantly reduced in cells overexpressing miR-21 compared with cells overexpressing the scrambled miR. The anti-MSH2 and anti-MSH6 siRNA were transfected in these cell lines in parallel. We observed that cells transfected with miR-21 displayed a down-regulation of hMSH2 and hMSH6 that appeared comparable to levels in cells transfected with siRNAs. Conversely, we transfected CRC SW480, HCT-116, and RKO cells that contain high levels of endogenous miR-21 with a locked nucleic acid (LNA) against miR-21 (anti-miR-21) or a scrambled LNA (anti-miR control). We found that cells transfected with anti-miR-21 showed an increase in both hMSH2 and hMSH6 protein expression (Fig. 1D and Fig. S2 A and B), but no changes in mRNA levels were observed (Fig. S2C).

The entire 3′ UTR of hMSH2 or hMSH6 was subcloned downstream of the luciferase gene. The luciferase reporter construct along with a precursor miR-21 or scrambled miR then was transfected into the Colo-320DM cells. We observed a 50% and 37% reduction in the luciferase activity with constructs containing the miR-21 seed regions for hMSH2 or hMSH6, respectively (P < 0.001; Fig. 1E). Deletion of the miR-21 seed regions resulted in the restoration of luciferase activity for both vectors containing hMSH2 or hMSH6 (Fig. 1E). We transfected SW480 cells that displayed high levels of miR-21 expression with a luciferase reporter vector containing the wild-type or mutated hMSH2 and hMSH6 3′ UTR seed region (Fig. 1F). As expected, we found that ablation of the miR-21 binding site resulted in increased luciferase activity for both the hMSH2 and hMSH6 vector-transfected cells. To confirm these observations, SW480 cells were cotransfected with the hMSH2 and hMSH6 3′ UTR luciferase reporter plus the LNA anti-miR-21 or anti-miR control. LNA silencing of miR-21 induced an increase in luciferase activity (Fig.1F). Taken as a whole, these results suggest that miR-21 exerts a direct effect on the hMSH2 and hMSH6 3′ UTR that ultimately regulates hMSH2 and hMSH6 protein expression. Because hMSH2 protein status can affect hMSH6 protein stability and expression (9), we cannot exclude the possibility that miR-21 regulation and hMSH2 protein loss can contribute to hMSH6 down-regulation.

miR-21 Is Inversely Correlated with the MMR Core Protein hMSH2 in CRC Tissues.

We examined miR-21 and hMSH2 expression in two different CRC cohorts (Fig. 2). A tissue microarray containing 50 unselected cases of CRC and paired normal adjacent tissue was hybridized with an LNA anti-miR-21 or nonspecific LNA anti-miR control combined with immunohistochemical staining for hMSH2 protein (Fig. 2A) (14). A score for both miR-21 and hMSH2 protein expression was given according to the percentage of positive cells in the core. Forty-two of 50 cores were available for matched analysis of tumor and paired normal tissue. We found that miR-21 was up-regulated in 28 (66%) of these cases when tumor was compared with normal paired tissue. In 14 of 42 cases (33%), there was a strong down-regulation of hMSH2 in tumor compared with normal tissue. In all these cases miR-21 was found to be up-regulated. Pearson's correlation analysis in this subgroup of patients showed an r value of −0.82 (P < 0.001). Correlation analysis on the entire cohort of cases showed an r value of −0.63. CRC tissues that scored positive for both miR-21 and hMSH2 showed no coexpression in the same cancer nest (Fig. 2A, colabeling).

Fig. 2.

Expression of the MMR core protein hMSH2 is inversely correlated to miR-21 expression in CRC samples. (A) Paraffin-embedded, formalin-fixed CRC tissues were incubated with an LNA probe anti-miR-21, or scrambled probe and with immunohistochemical antibody against hMSH2. Representative photographs were captured with the Nuance system software. CRC samples in which staining was positive for both miR-21 and hMSH2 are shown. Blue and red staining identifies miR-21 and hMSH2 proteins, respectively. (B) RNA and proteins were extracted from fresh-frozen human colorectal tissues. miR-21 expression was assessed by Northern blotting, and MMR protein expression was assessed by Western blotting.

We examined fresh-frozen tumors from a second cohort of CRC samples for which cancer and normal adjacent tissues were available (Fig. 2B). miR-21 expression was determined by Northern analysis and RT-PCR, and hMSH2 protein expression was determined by Western analysis. In 26 cases, hMSH2 was down-regulated in tumors compared with normal adjacent tissue. In 24 of these cases (90%), miR-21 expression was increased in tumor as compared with adjacent normal tissues. Because we previously have reported that miR-155 can affect the expression of hMSH2 and other MMR proteins (15), we excluded from analysis 16 cases showing simultaneous overexpression of miR-155 and miR-21. An inverse correlation (r = −0.81; P < 0.02) still was evident in the remaining eight cases, highlighting the inverse correlation between miR-21 overexpression and hMSH2 down-regulation in CRC tumors (Fig. 2B and Fig. S3).

miR-21 Reduces G2/M Arrest and Apoptosis Following Exposure to 5-FU.

MMR-defective cell lines display resistance to a variety of therapeutic drugs, including 5-FU (16). These studies have demonstrated that resistance is the result of defective incorporation of 5-FU metabolites into DNA, leading to reduced damage-dependent G2/M arrest and subsequent apoptosis (11). We examined 5-FU–induced cell-cycle arrest and apoptosis in Colo-320DM and SW620 cells following transfection of miR-21. We used a scrambled miR as a control and compared our results with a similar transfection with an siRNA anti-MSH2 (Fig. 3). We found that miR-21 overexpression decreased the percentage of sub-G1 (apoptosis) and G2/M cells following treatment with 5-FU. Cells transfected with miR-21 displayed reduced G2/M arrest and apoptosis, similar to cells transfected with siRNA to hMSH2 (Fig. 3). The effect of miR-21 expression on 5-FU–mediated apoptosis was confirmed further in Colo-320DM and SW620 cells by Annexin V staining (Fig. S4). A similar response was observed in isogenic Lovo cells where the hMSH2 mutation [Lovo(MSH2⁻)] has been complemented with the introduction of chromosome 2 [Lovo(MSH2⁺)] (Fig. 4) (17). miR-21 overexpression, as well as siRNA to hMSH2, reduced sub-G1 and G2/M accumulation in Lovo(MSH2⁺) cells, whereas no effects were observed in Lovo(MSH2⁻) cells (Fig. 4). Cotransfection of Lovo(MSH2⁺) and Lovo(MSH2⁻) cells with a plasmid encoding the full-length hMSH2 cDNA promoted 5-FU–induced apoptosis and cell-cycle arrest (18). Cotransfection of the same plasmid along with miR-21 markedly reduced G2/M arrest and apoptosis (Fig. 4). Moreover, deletion of the target site in the hMSH2 cDNA rendered the message insensitive to miR-21 regulation, and cells retained a normal damage-induced G2/M arrest and apoptosis. Taken as a whole, these results are consistent with the conclusion that down-regulation of hMSH2 expression by miR-21 results in cellular resistance to 5-FU.

Fig. 3.

miR-21 inhibits 5-FU induced apoptosis in vitro. SW620 and Colo-320DM cells were synchronized at G0-G1 by serum starvation for 48 h. Cells then were trypsinized, counted, transfected with scrambled miR, miR-21, siRNA anti-MSH2, or siRNA-control, and replated in medium containing FBS. 5-FU was added 16 h after transfection, corresponding to a time just before entry into S phase but after the p53-mediated G1-S cell-cycle checkpoint. The cell cycle was analyzed 48 h after 5-FU administration. The percentage of G2/M-arrested and apoptotic (sub-G1) cells is shown. Data represent the mean ± SD from three determinations from three independent transfections. *P < 0.001.

Fig. 4.

miR-21–mediated 5-FU resistance is dependent on hMSH2 down-modulation. Lovo(MSH2⁺) and Lovo(MSH2⁻) cells were synchronized at G0-G1 by serum starvation for 48 h and transfected with miR-control, miR-21, siRNA anti-MSH2, or siRNA-control and vectors encoding full-length hMSH2 cDNA (with [WT] or without [MUT] the miR-21 seed region). The cell cycle was analyzed 48 h after 5-FU administration. The percentages of G2/M-arrested and apoptotic (sub-G1) cells in both Lovo(MSH2⁺) cells (blue bars) and Lovo(MSH2⁻) cells (pink bars) are shown and represent the mean ± SD from two determinations from three independent experiments. *P < 0.001.

Overexpression of miR-21 Induces 5-FU Resistance in a CRC Xenograft Model.

Our cellular studies suggested that miR-21 inhibits 5-FU–induced G2/M arrest and apoptosis by reducing the expression of hMSH2. We developed a xenograft colon cancer tumor model in which we generated stable clones of Lovo(MSH2⁺) cells that overexpressed miR-21 [Lovo(MSH2⁺)-miR-21] or an siRNA to hMSH2 [Lovo(MSH2⁺)-anti-MSH2] using a lentiviral expression system. Lovo(MSH2⁻) cells and Lovo(MSH2⁺) cells containing the stable insertion of an empty vector served as controls. We injected 5 × 10⁶ cells containing stable lentiviral expression into the flank of nude mice. When xenografts reached a palpable volume, 5-FU (50 mg·kg⁻¹·d⁻¹) was administered by i.p. injection for 5 consecutive days per week for 2 wk (19).

The 5-FU treatment proved more efficacious with Lovo(MSH2⁺) tumor xenografts than with Lovo(MSH2⁻) tumor xenografts (Fig. 5A and Table S1). These results are consistent with previous studies and suggest that MMR-proficient cells respond better to 5-FU therapy (11, 12). Importantly, stable overexpression of miR-21 [Lovo(MSH2⁺)-miR-21] resulted in a reduced response to 5-FU and caused a tumor growth rate comparable to that in Lovo(MSH2⁺) tumor cells infected with siRNA to hMSH2 [Lovo(MSH2⁺)-anti-MSH2] (Fig. 5A and Table S1). Furthermore, 2 wk after discontinuation of 5-FU, tumor growth appeared significantly greater in the cells infected with Lovo(MSH2⁺)-miR-21 than in controls, suggesting that miR-21 overexpression enhances cancer progression. We confirmed that the expression of hMSH2 was reduced dramatically in Lovo(MSH2⁺) tumor xenografts expressing miR-21 or the anti-MSH2 siRNA compared with the empty vector (Fig. 5 B and C). Taken together, our results support a central role for miR-21–dependent down-regulation of the hMSH2-hMSH6 heterodimer MMR protein in 5-FU resistance.

Fig. 5.

miR-21 causes resistance to 5-FU in vivo. Lovo(MSH2⁺) cells were stably infected with a lentiviral vector encoding for either miR-21 or siRNA anti-MSH2. As a control, Lovo(MSH2⁺) and Lovo(MSH2⁻) cells were infected with empty vectors. Nude mice were injected with Lovo(MSH2⁺)-Empty (n = 6), Lovo(MSH2⁺)-miR-21 (n = 6), Lovo(MSH2⁺)-anti-MSH2 (n = 6), or Lovo(MSH2⁻)-Empty (n = 6). When xenografts reached a palpable volume, 5-FU was administered by i.p. injection for 5 consecutive days per week for 2 wk (gray area). Tumor volume was measured before treatment and then once each week. The individual relative tumor volume (RTV) was calculated as follows RTV = V x /V1 where Vx is the volume in cubic millimeters at a given time and V1 is the volume at the start of treatment. Results are expressed as the mean percentage of change in tumor volume for each group of mice, ± SD. (A) Tumor growth during and after 5-FU treatment. (B) Representative tumor xenografts at week 6. (C) Western blotting analysis of hMSH2 protein expression in removed tumors.

Discussion

miR-21 is commonly overexpressed in a number of human tumors, including CRC (4, 7). In recent years several miR-21 tumor-suppressor targets have been identified that may accelerate the progression of cancer (20–24). We found an inverse relationship between colorectal tumor cells that overexpress miR-21 and those that express the hMSH2 tumor-suppressor protein. Moreover, we determined that miR-21 appears to target directly the 3′ UTR of both the hMSH2 and hMSH6 mRNA, resulting in significant down-regulation of protein expression. Because the hMSH2-hMSH6 heterodimer is the key initiation component, we expect that down-regulation of its core components is likely to affect mutation suppression by MMR and thereby enhance tumor progression. A clear prediction is that overexpression of miR-21 should enhance mutation rates via down-regulation of hMSH2 protein. Although these studies are ongoing, there are a number of other distinctive phenotypic effects that are indicators of MMR defects.

The state of the art therapeutic treatment of CRC includes 5-FU. 5-FU exerts its cytotoxic effects by misincorporation of fluoronucleotides into RNA and DNA and by inhibiting nucleotide synthesis by targeting the thymidylate synthetase enzyme. The overexpression of thymidylate synthetase, defects in 5-FU metabolism, TP53 mutations, and impairment of the MMR system are all hallmarks of 5-FU resistance and are predictors of clinical outcome (2). More recently both microRNA and gene expression analysis have revealed a higher level of complexity in predicting 5-FU benefit in stage-II and -III CRC patients who underwent adjuvant chemotherapy (7, 25). Indeed, a retrospective analysis of stage-II and -III CRC patients treated with 5-FU analogs showed reduced survival in patients with high miR-21 expression (7). The same findings were confirmed in the subgroup of stage-III CRC patients alone, whereas stage-II CRC patients showed no statistically significant correlation (7). The low number of patients may account for this latter result. Cells with genetic or epigenetic defects of the MMR machinery appear to tolerate 5-FU metabolites as a result of defects in G2/M arrest and apoptosis (11). Our studies have demonstrated that down-regulation of hMSH2 by miR-21 induces resistance to 5-FU in both a cellular model and a xenograft tumor model. Taken together, these results suggest that miR-21 tumor status is likely to be an important indicator of 5-FU therapeutic efficacy.

miR-21 appears to regulate a number of cell-cycle and tumor suppressor genes. It is possible that miR-21 could regulate other genes associated with cell-cycle arrest and/or drug resistance that might contribute to 5-FU resistance (22, 23). However, our studies suggest that down-regulation of hMSH2 plays a central role in the development of 5-FU resistance. Indeed, inhibition of 5-FU–induced apoptosis and G2/M arrest by miR-21 was comparable to that caused by siRNA-mediated selective inhibition of hMSH2. Moreover, transfection of Lovo(MSH2⁻) cells with miR-21 did not alter cell-cycle arrest or apoptosis, demonstrating that miR-21–induced effects are dependent on hMSH2 expression. Taken together, our results clearly suggest that inhibition of miR-21 might represent a synergic treatment to overcome 5-FU resistance. Indeed, it has shown recently that codelivery of antisense-miR-21 and 5-FU by a poly(amidoamine) dendrimer can sensitize cells to 5-FU by increasing cell-cycle arrest and apoptosis in glioma cell lines (26).

Our results suggest that miR-21–dependent down-regulation of hMSH2-hMSH6 might be responsible for both primary and acquired resistance to 5-FU. In clinical practice 5-FU usually is administered as a continuous infusion over a 48-h period (27). Interestingly, miR-21 expression appears to increase in cell lines continuously exposed to 5-FU (28). We speculate that this overexpression may be a secondary mechanism of resistance and that cells may acquire miR-21 overexpression to overcome 5-FU cytotoxicity. There is additional clinical relevance if one considers that frequently hMSH2 is down-regulated after primary chemotherapy including 5-FU or cisplatin in rectal and ovarian cancers (29, 30).

In summary, we have correlated 5-FU drug resistance in colorectal tumors to the overexpression of miR-21 that directly down-regulates the core MMR proteins hMSH2 and hMSH6, ultimately leading to a defect in damage-induced G2/M arrest and apoptosis. These studies suggest that miR-21 is likely to be a useful marker for therapeutic protocols in CRC.

Materials and Methods

Colo-320 DM, SW620, HCT-116, SW480, RKO, and cells were transfected using Lipofectamine 2000 (Invitrogen). Luciferase constructs were subcloned using the primers listed in Table S2. Lentiviral vectors for miR-21 overexpression and empty vectors were generated by using System Biosciences kit. Resutls of statistical analyses are expressed as mean ± SD unless indicated otherwise. Comparisons between groups were performed using the two-tailed Student's t test. A P value <0.05 was considered significant. Graphpad Prism version 5.0 was used for Pearson correlations. More detailed information is provided in SI Materials and Methods.