Micro RNA 145 Targets the Insulin Receptor Substrate-1 and Inhibits the Growth of Colon Cancer Cells*

The insulin receptor substrate-1 (IRS-1), a docking protein for both the type 1 insulin-like growth factor receptor (IGF-IR) and the insulin receptor, is known to send a mitogenic, anti-apoptotic, and anti-differentiation signal. Several micro RNAs (miRs) are suggested by the data base as possible candidates for targeting IRS-1. We show here that one of the miRs predicted by the data base, miR145, whether transfected as a synthetic oligonucleotide or expressed from a plasmid, causes down-regulation of IRS-1 in human colon cancer cells. IRS-1 mRNA is not decreased by miR145, while it is down-regulated by an siRNA targeting IRS-1. Targeting of the IRS-1 3′-untranslated region (UTR) by miR145 was confirmed using a reporter gene (luciferase) expressing the miR145 binding sites of the IRS-1 3′-UTR. In agreement with the role of IRS-1 in cell proliferation, we show that treatment of human colon cancer cells with miR145 causes growth arrest comparable to the use of an siRNA against IRS-1. Taken together, these results identify miR145 as a micro RNA that down-regulates the IRS-1 protein, and inhibits the growth of human cancer cells.

The insulin receptor substrate-1 (IRS-1) 2 is one of the major substrates of both the type 1 insulin-like growth factor receptor (IGF-IR) and the insulin receptor (InR). IRS-1 plays an important role in cell growth and cell proliferation (1). IRS-1, especially when activated by the IGF-IR, sends an unambiguous mitogenic, anti-apoptotic, and anti-differentiation signal (2,3). IRS-1 levels are often increased in human cancer (4), and they are low or even absent in differentiating cells (1,5,6). Overexpression of IRS-1 causes cell transformation, including the ability to form colonies in soft agar and tumors in mice (7,8). Transgenic expression of IRS-1 in the mammary gland of mice causes mammary hyperplasia, tumorigenicity, and metastases (9). Conversely, down-regulation of IRS-1 (by antisense or siRNA procedures) reverses the transformed phenotype (10 -12). The IRS proteins are conserved during evolution, and a gene described in Drosophila, called chico, is the equivalent of IRS-1 to IRS-4 in mammalian cells. IRS proteins play an important role in cell size. Deletion of chico reduces fly weight by 65% in females and 55% in males (13). Mice with a targeted disruption of the IRS-1 genes are also smaller than their wild-type littermates (14), and ectopic expression of IRS-1 increases rRNA synthesis and doubles cell size in cells in culture (7,15). Thus, IRS-1 seems to play important roles in cell growth (cell size), cell proliferation, and differentiation.
Micro RNA (miRs) are RNAs of ϳ22-nucleotides long, that arise from one arm of longer endogenous hairpin transcripts. The characteristics of miRs have been summarized in several reviews (16 -19). Briefly, miRs are cleaved from one arm of a longer endogenous double-stranded precursor (70 -100 nt in length) by Drosha and Dicer enzymes (RNase III family). They are transcribed by RNA polymerase II (20) as long primary transcripts (pri-miRNAs), which are cropped and cleaved to produce the pre-miR and the mature miR (21). They are complementary to genomic regions, and one of their modes of action is to bind to the 3Ј-untranslated regions of mRNA (3Ј-UTR), inhibiting translation (the target mRNA levels remain unchanged). They can function also by cleaving a target mRNA, in which case the miR may target sequences outside the 3Ј-UTR (18). miRs play crucial roles in eukaryotic gene regulation, especially in development and differentiation (22)(23)(24)(25). A few reports have tied miRs to cancer (26 -30). Targets of miRs can be obtained from the data base (see below), although it is understood that the presumed targets have to be validated experimentally.
We first screened some data base candidates for IRS-1 targeting, and we then focused on a single miR (miR145) testing it for its ability to down-regulate IRS-1 in cells in culture. We show that synthetic oligonucleotides (oligos) of miR145 cause down-regulation of the IRS-1 protein in human colon cancer cells and that its effect is slightly less pronounced than the effect of an siRNA against IRS-1. Whereas the siRNA causes a downregulation of IRS-1 mRNA, miR145 does not, indicating that the effect is probably on translation. A reporter gene carrying the 3Ј-UTR or the miR145 binding sites of IRS-1 is also downregulated by miR145, while an IRS-1 cDNA without its 3Ј-UTR is not affected. Finally, an expression plasmid expressing a hairpin precursor miR145 also down-regulates IRS-1 when transfected into colon cancer cells. Although siRNA is more effective than miR145 in down-regulating IRS-1 levels, miR145 and siRNA have similar inhibitory effects on the growth of colon cancer cells in culture; in fact, in some experiments miR145 was more potent than siRNA in inhibiting cell proliferation. This is probably because miRs target multiple proteins along the same pathway (31,32). Indeed, miR145 targets also the IGF-IR (see below). Taken together, these results demonstrate that miR145 targets the 3Ј-UTR of IRS-1 mRNA, and that the targeting has a profound effect on the growth of human colon cancer cells. This is the first demonstration of a specific miR targeting a transduction molecule of the IGF-IR/insulin receptor signaling pathway. Its inhibition of growth in human cancer cells in culture is compatible with the well known ability of IRS-1 to stimulate cell proliferation and transformation (see above).

EXPERIMENTAL PROCEDURES
Cells-Colorectal cancer cell lines HCT116-Dicer-KO 2 and DLD1-Dicer-KO 4 were kindly provided by Dr. Bert Vogelstein (33), and the parental cells HCT116 and DLD1 were from ATCC (Manassas, VA). Both are human colorectal adenocarcinomas cell lines. All cells were cultured in McCoy's 5A medium supplemented with 10% fetal bovine serum and penicillin/ streptomycin. In the Dicer-KO cells, the exon 5 of the Dicer gene encoding helicase is replaced by a neoR gene. BT-20, a human breast cancer cell line, was from ATCC and grown in Dulbecco's modified Eagle's medium/F12 medium supplemented with 10% calf serum, L-glutamine, and penicillin/streptomycin. Rϩ and R12 cells (34) were generated from R-cells, which are 3T3-like mouse embryonic fibroblasts (MEFs) with a targeted disruption of endogenous IGF-IR genes. Rϩ and R12 cells are RϪ cells stably transfected with a plasmid expressing human IGF1R. Rϩ has 300-fold more IGF1R (9 ϫ 10 5 receptors) than R12 (3 ϫ 10 3 receptors). Rϩ cells grow in serum-free medium supplemented solely with IGF-I, whereas R12 do not. Both cell lines were cultured in Dulbecco's modified Eagle's medium ϩ 10% fetal bovine serum ϩ penicillin/streptomycin medium.
TaqMan Real Time RT-PCR-Messenger RNAs of IRS-1 were extracted using RNeasy Mini kit (Qiagen, Valencia, CA). miRNAs were extracted using Micro RNA Isolation kit (Stratagene, La Jolla, CA) or mirVana miRNA Isolation kit (Ambion, Austin, TX). Primers and probes specific for human IRS-1 and internal control 18S rRNA were purchased from Applied Biosystems (ABI, Framingham, MA). TaqMan One-step RT-PCR Master Mix Reagents kit (ABI, Roche, Branchburg, NJ) was used to detect IRS-1 mRNA. Amplification and detection was performed using 7900HT Sequence Detection System (ABI), using 40 cycles of denaturation at 95°C (15 s) and annealing/ extension at 60°C (60 s). This was preceded by reverse transcription at 50°C for 30 min and denaturation at 95°C for 10 min. To quantitate mature miRNA, TaqMan MicroRNA Assays kits were purchased from ABI to detect miR145 (cat. 4373133) and a control miR (RNU6B, cat. 4373381). It is a twostep protocol requiring reverse transcription (cat. 4366596) with a miRNA-specific primer, followed by real time PCR with TaqMan probes (cat. 4324018). The assays target only mature microRNAs, not their precursors, ensuring biologically relevant results. The fold change of target gene in treatment groups relative to mock-treated samples were calculated according to the ABI Relative Quantification Methodology. The absolute miR145 levels in parental HCT116 and HCT116-Dicer-KO cells were also calculated according to a standard curve of miR145 (Dharmacon hsa-miR145 ds-oligo served as the standard). For details, refer to the ABI user's bulletin: Relative Quantitation of Gene Expression: ABI PRISM 7700 Sequence Detection System: User Bulletin 2: Rev B.
Luciferase Assay-Dual luciferase vector psiCHECK2 was purchased from Promega (Madison, WI). HCT116-Dicer-KO 2 cells were seeded in 96-well plate. The cells were transfected with different psiCHECK2 constructs containing 3Ј-UTR of human IRS-1 or miR145 potential binding sites (see supplemental data), in the presence or absence of miR145 (Dharmacon). 48 h later, the firefly and Renilla luciferase activities were assayed using Dual-Glo Luciferase assay system (Promega) in Tecan Safire Microplate Reader II. Because all the miR potential binding sequences were cloned at the 3Ј of Renilla luciferase gene, the ratio of the luminescent signals from Renilla versus firefly represents the target specificity of miRs. All experiments were performed in triplicate, i.e. 3 wells for each condition.
The potential binding sites of miR145 on the 3Ј-UTR of human IRS-1 were cloned into multiclonal sites (MCS) of a dual luciferase vector psiCHECK2 (Promega). Double strand oligos (listed under supplemental materials) were generated by annealing sense and antisense strands, and further ligated into psiCHECK2 digested with XhoI and NotI.
Plasmids-The pSuper.retro.neo.GFP plasmid (abbreviated pSuper) was purchased from Oligoengine. It is controlled by a 5Ј-LTR, has a variety of restriction sites for insertion, and the transfected cells can be selected either by neomycin or GFP (FACS sorter). It has been tested by Cimmino et al. (35). Double strand-oligo inserts, ϳ70-nt hairpin stem loop pre-miR145 plus 20, 40, 80, or promoter ϩ 160 nt flanking sequences at each side of hairpin, were PCR-amplified from human genomic DNA (Promega, G3041) and cloned into BglII and HindIII sites of pSuper. The resulting constructs were called pSuper-hairpin145_20nt (clone 26), pSuper-hairpin145_40nt (clone 28), pSuper-hairpin145_80nt (clone 30), and pSuper-hairpin145_160nt (clone 32). The sequences of the inserts were confirmed by DNA sequencing using primers suggested by OligoEngine (Seattle, WA). The mature miR145 (24 nt) was also directly cloned into pSuper. The resulting clones are called pSuper-mature145_24nt (clone 18). All the above primers for PCR and cloning are listed under supplemental materials.
Data Base-miR target genes were screened with the Target Scan program, the miRanda program, miRBase, and miRNAMap. The targets were confirmed by BLAST alignment with the corresponding NCBI DNA data base for homologies between miRs and their targets.

Potential IRS-1-specific miRs-
The data base identified several miRs as targeting IRS-1, and selected candidates are listed in Fig. 1, panel A. The structure of the IRS-1 pre-mRNA is unusual and relevant to the experiments described below. The pre-mRNA structure (NCBI for NM_010570, GeneID: 16367. Locus tag: MGI:99454) is given in panel B. In terms of 3Ј-UTR, the IRS-1 mRNA has an exon of 4,640 bp, with the coding region extending from residue 924 to residue 4619. Then the 3Ј-UTR begins (21 bp), interrupted by an intron of 49,172 bp, and completed by an additional 995 bp of 3Ј-UTR. Fig. 1, panel C, gives a more detailed presentation of the 3Ј-UTR of the IRS-1 cDNA, with the two putative binding sites for miR145. One binding site is in the 21-bp sequence immediately after the stop codon, while the 2nd binding site is separated from the first (in the genome) by almost 50,000 bp.
HCT116 and DLD1 Cells-Parental HCT116 and DLD1 cells are both colon carcinoma cell lines frequently used in research. HCT116-Dicer-KO and DLD1-KO cells, both kind gifts from Dr. Bert Vogelstein (Johns Hopkins University), are HCT116 and DLD1 cells in which exon 5 of the Dicer gene (the helicase domain) has been disrupted (33). Because Dicer is required for proper processing of mature miRs, these cells have markedly reduced amounts of mature miRs and display accumulation of miR precursors. The low levels of mature miR145 in HCT116-KO cells, in comparison to parental cells, are shown in Fig. 2, panel A. Actually, miR145 cannot be detected in either parental or HCT116-KO cells by Northern blot (data not shown), and can only be detected by TaqMan. Fig. 2, panel B, shows IRS-1 protein levels in selected cell lines. IRS-1 levels are slightly higher in HCT116-KO cells than in parental HCT116 cells (GAPDH levels monitor protein loading). We have included in this Western blot lysates of Rϩ and R12 cells, mouse embryo fibroblasts known to have substantial levels of IRS-1 (34) as well as BT-20 mammary cancer cells, that do not express IRS-1 at all (8) and serve as the negative control. Fig. 2  transfected efficiently with a synthetic oligo of miR145 (Dharmacon). We also transfected the miR148a oligos into both HCT116-KO and DLD1-KO cells as negative controls for the Northern blot with labeled miR145 probe (transfected miR148a could be detected after transfection, with the appropriate miR148a probe, data not shown). For subsequent studies, we used mostly HCT116-KO cells, designated from now on as KO cells, to screen several synthetic oligos for their ability to decrease IRS-1 levels. miR145 Down-regulates the IRS-1 Protein in KO Cells-In early experiments, we determined the effects of miR oligos at 24 and 48 h after transfection. The results were inconclusive (not shown), and we decided to try longer time points, given the half-life of the IRS-1 protein (12). Fig. 3 shows that miR145 decreases IRS-1 levels 96 h after transfection. There are other miRs (predicted from the data base) that down-regulate IRS-1, i.e. miR154 and miR207. miR148a was also predicted to target IRS-1, but, in these experiments, failed to decrease IRS-1 levels. miR145 and miR207 oligos down-regulated IRS-1 in parental HCT116 cells, KO HCT cells and KO DLD1 cells. miR154 was effective only on the first 2 cell lines. An siRNA against IRS-1 (see "Experimental Procedures") was more effective than miR145 in down-regulating IRS-1 protein levels, especially in parental cells. Fig.  4, panel A, shows repeated experiments in which KO HCT116 cells were transfected with the miR145 oligo, in four separate experiments. In all of them, 96 h after transfection, the levels of IRS-1 protein were significantly lower than in untreated or mock-transfected KO cells. miR145 also down-regulates the IGF-IR (Fig. 4, panel A), although to a lesser extent than IRS-1. IRS-2 expression is not decreased (supplemental data), as expected from the data base, indicating that IRS-2 is not targeted by miR145. In this communication, we have focused on miR145 and its targeting of IRS-1.
We compared the effect of miR145 to the effect of siRNA against IRS-1 on IRS-1 levels. This is shown in Fig. 4, panel B, where we measured IRS-1 protein levels 24 h and 5 days after transfection. siRNA is more efficient than miR145 in down-regulating IRS-1 protein levels. Notice that at 24 h after transfection, even the siRNA has only a moderate effect on IRS-1 levels (left panel of panel B).
IRS-1 mRNA Levels Are Not Down-regulated by miR145-It is generally agreed that in the majority of cases, miRs act by inhibiting translation, although in some cases, they may cause breakdown of the mRNA (see Introduction). We tested the levels of IRS-1 mRNA in KO cells transfected with either miR145 oligos or siRNA/IRS-1, and compared it to control cells (untreated or mock-transfected). The results (by TaqMan real-time RT-PCR) of repeated experiments are summarized in Fig. 5. As expected, siRNA markedly decreases IRS-1 levels, but these remain constant in cells transfected with miR145 synthetic oligos. In fact, there was a small increase of IRS-1 mRNA in cells transfected with the miR oligo. The increase at 96 h was actually statistically significant. These experiments show that miR145 down-regulates the IRS-1 protein, but not the mRNA.
miR145 Down-regulates a Reporter Gene with Sequences from the 3Ј-UTR of IRS-1-To confirm that miR145 targets the 3Ј-UTR of IRS-1, we carried out experiments used by other investigators to determine the specificity of the 3Ј-UTR targeting (36 -38). The general approach has been to insert the 3Ј-UTR in question at the 3Ј-end of a reporter gene, often luciferase (36). We followed the same protocol, and made four dif-  ferent constructs in which luciferase was expressed with the 3Ј-UTR of IRS-1 (full-length) or with the presumed binding sites of miR145 to the 3Ј-UTR of IRS-1 cDNA (see Fig. 1). One construct had the 1st putative binding site for miR145, a 2nd construct had the 2nd putative binding site, and the final construct had both binding sites (see Fig. 1). The constructs were then co-transfected with miR145 oligos, with cells transfected only with the constructs serving as the controls. The results of a typical experiment are shown in Fig.  6. There was a significant decrease in the expression of the luciferase reporter in cells co-transfected with miR145 oligos and the luciferase carrying the full-length 3Ј-UTR of IRS-1 or a 3Ј-UTR containing the two presumed binding sites of 3Ј-UTR for miR145. The constructs in which luciferase had only one binding site for miR145 were also slightly decreased, but to a lesser extent. The results were subjected to a two-tailed t test. The decrease in luciferase in the cells with the 3Ј-UTR or the two miR145 binding sites were significant (p Ͻ 0.05). The decreases for the constructs with only one binding site were p ϭ 0.6 and therefore not significant. The data in Fig. 6 come from a single experiment (n ϭ 3), although the experiment was repeated three times, with the same results. IRS-1 without a 3Ј-UTR-Another way of confirming that a given 3Ј-UTR is targeted by a miR is to ask whether the miR no longer down-regulates a protein, whose cDNA has been deprived of its 3Ј-UTR. We used for this purpose a truncated IRS-1 cDNA lacking its 3Ј-UTR, and coding for a shorter protein, distinguishable from the wild-type endogenous IRS-1 in HCT116 cells. The truncated IRS-1 was transfected with miR145 oligos into Dicer-KO cells and the results (96 h after transfection) are summarized in Fig. 7. The blot was quantified by densitometry, because of the slight variability in GAPDH levels. The endogenous wild-type IRS-1 is again down-regulated (about 50%) by miR145, but the truncated IRS-1 is not. The siRNA was used as a control to show that both full-length and truncated proteins are down-regulated by the siRNA against IRS-1. This experiment was repeated.
Expression of miR145 in pSuper-We next attempted to express miR145 in the pSuper plasmid. Preliminary experiments indicated that cloning of the mature miR145 straight into pSuper did not result in detectable expression. We therefore investigated whether the addition of flanking sequences to the precursor miR could increase the levels of expression. The flanking sequences used were the genomic sequences flanking the hairpin precursor miR145 (see data base). We tried 20, 40, and 80 nucleotides on each side. The results (Fig. 8, panel A) show TaqMan RT-PCR determinations of mature miR145 levels in parental HCT116 cells, KO cells, and 293FT cells transfected for 48 h with the different pSuper constructs. 20 and 40 nucleotides of flanking sequences improved the expression of miR145 cloned in pSuper, with the 20 nucleotides being the obvious first choice. The experiments were repeated using Northern blots to measure the levels of mature miR145 (Fig. 8,  panel B). By these methods, 20 nucleotides of flanking sequences are the optimal condition for miR expression, although some expression is detectable also with 40 and 80 flanking nucleotides. This is more evident in 293FT cells than in parental HCT116 cells. Our results are at variance with the report of Chen et al. (24), who found that the general strategy for miR expression required 270 nucleotides (22 nucleotides of mature miR plus 125 nucleotides of genomic sequences on each side). The discrepancy may be due to the miR or the pSuper. We tested the effect of pSuper/miR145/20 nucleotides on IRS-1 levels in parental HC116 cells. IRS-1 is down-regulated in cells transfected with this pSuper construct (clone 26) 96 h after transfection (Fig. 8).

Effect of miR145 and siRNA/IRS-1 on the Growth and Morphology of KO Cells-Whereas the previous experiments
focused on the precise targeting of IRS-1 by miR145, the next experiment focuses on the effect of miR145 on cell proliferation. Thus, the results of Fig. 9 are not just the effect of IRS-1 down-regulation, but the effect of miR145 on multiple targets. KO cells were transfected with oligos of four different miRs and with siRNA/IRS-1 and examined 4 days after transfection. Whether using the plates or the microscopic pictures of the plates (Fig. 9, 1st and 2nd rows, respectively), it is clear that miR145, miR154, and miR207 inhibit the growth of KO cells as effectively as siRNA. Transfection with miR148a gave the same picture as in mock-transfected cells. In the experiments of Fig.  9, siRNA gave a better down-regulation of IRS-1 than the miRs, while the biological effects seem to be similar (in some experiments, miR145 was even better than the siRNA in inhibiting cell proliferation). Because IRS-2 expression levels were not decreased by miR145, we were unable to demonstrate a reduction in IGF-1 activation of downstream targets like Akt and ERKs (supplemental data). However, an IRS-1-specific target, c-Myc, was clearly reduced in these cells (supplemental data). These findings suggest that other miR145 targets play an important role in inhibiting cellular proliferation.
A careful observation of the treated cells suggests that they may be larger, with somewhat more cytoplasm than the mocktransfected cells. Because IRS-1 is a strong inhibitor of differentiation (7,39,40), we asked whether treatment with the anti-IRS-1 strategies could have induced differentiation of colon cells. We could not detect differentiation in miR145-treated cells (data not shown). We also subjected the treated cells to tests for apoptosis, but they also came out negative (not shown). The growth retardation of cells treated with miR145 was simply a growth arrest. However, cells transfected with miR145 had a tendency to accumulate in the G 2 phase of the cell cycle (technically the G 2 /M phase, but since we did not see an increase in mitoses, it is fair to say G 2 phase). For instance, at 96 h after transfection with miR145, 23.7% of the KO cells were in G 2 against 11% in mock-transfected or miR-negative-transfected cells. The biological effects of miR145 were not limited to HCT116 Dicer-KO cells. We also observed a dramatic inhibition of cell growth in DLD1 KO cells and in a line of mouse embryo fibroblasts transformed by v-Src (data not shown, but available on request).

DISCUSSION
Although the data base may suggest miRs that target a given protein, it is necessary to confirm the target experimentally. The experimental verification (36 -38, 41) is usually based on demonstrating that: 1) the target protein is down-regulated by the predicted miR; 2) a reporter gene expressing the 3Ј-UTR or the miR-binding sites of the targeted mRNA is also down-regulated by the predicted miR; 3) the targeted protein is not down-regulated when the 3Ј-UTR is missing; 4) the miR has a biological function predicted by the biological function of the  1 ϩ 2)) or the entire 3Ј-UTR of IRS-1 (see Fig. 1). The five plasmids were transfected with or without the miR145 oligos, and luciferase levels were determined after 48 h (for corrections caused by transfection efficiency, see "Experimental Procedures"). Data were generated from three repeated experiments.  miR145 was predicted to target IRS-1 mRNA by the data base (see Fig. 1). The use of the KO HCT116 cells was crucial in these experiments, because it allowed us to screen quickly the more promising miRs and to test various constructs. The KO cells produce very little amounts of mature miRs (strict Dicer-KO cells are not viable, and for this reason Vogelstein and co-workers generated a cell line with a hypomorphic phenotype. Dicer is ineffective and mature miRs levels are very low). Other miRs (miR154 and miR207) predicted by the data base down-regulated IRS-1, except for miR148a, which confirms the requirement for experimental validation. miR145 also down-regulated the IGF-IR, not surprisingly as miRs are said to target often multiple mRNAs in the same signaling pathway (31). We have not further pursued the IGF-IR target in the present communication, nor have we pursued further the other miRs that seem to down-regulate IRS-1, miR154, and miR207. In this communication, we have focused on miR145 and its effect on IRS-1 expression.
Our results show that IRS-1 mRNA levels are not downregulated by miR145 (while they are strongly affected by an siRNA against IRS-1). miR145 is presumably acting on the translation of IRS-1, which is believed to be the most common mechanism of miR targeting (16). The targeting of the IRS-1 3Ј-UTR was confirmed using a reporter gene, luciferase, carrying the 3Ј-UTR of IRS-1 or its two putative binding sites for miR145. It would have been desirable to further confirm the targeting by using the genomic sequences that are 3Ј to the IRS-1 coding sequences, as done by Wu and Belasco (36) for miR125 a and b. However, although this may be feasible, it presents considerable difficulties in our case because the genomic sequences of the 3Ј-UTR of IRS-1 are separated by almost 50,000 base pairs (see Fig. 1). We used instead a different construct, with the whole 3Ј-UTR of IRS-1 or its putative binding sites for miR145. Significantly, the best results for downregulation of luciferase were obtained with the whole 3Ј-UTR of IRS-1 or with the construct expressing both binding sites.
It is not surprising that the best results on IRS-1 protein down-regulation are obtained 72 or 96 h after transfection with miR145 oligos. miR145 does not decrease IRS-1 mRNA levels, and it takes some time for the IRS-1 protein to turnover (according to Cesarone et al.,Ref. 12, the half-life of the IRS-1 protein is at least 48 h).
The effect of miR145 on the growth and morphology of HCT116 KO cells is dramatic. However, as already pointed out, this is not the results of IRS-1 targeting by miR45, but the sum of the effects that miR145 has on multiple targets. At first, we speculated that inhibition of cell proliferation and morphological changes of HCT116 cells by miR145 treatment might involve induction of a differentiation pathway. We tested several differentiation markers for colon cells, but we could not detect any change compared with mock or miR-negative control-treated cells (not shown, available on request). It seems that the inhibition of HCT116 cells proliferation by miR145 does not involve induction of cell differentiation, and it might be instead the consequence of cell cycle arrest. Further characterization is underway and will be extended to other cell lines whose growth was inhibited by miR145 (DLD1 colon cancer cells, and v-Src-transformed mouse embryo fibroblasts). miR145 down-regulated IRS-1 protein levels but did not decrease the level of IRS-1 mRNA. Instead, we observed a slight increase of IRS-1 transcripts. We believe this is a case of feedback activity, when miR145 suppresses the translation of IRS-1, cells see less IRS-1 protein and to compensate, the transcription of irs1 gene is accelerated. We detected this compensatory increase of IRS-1 mRNA level by TaqMan real-time RT-PCR (Fig. 5).
A number of reports have suggested a role of miRs in cancer, see the Introduction and the reviews by Hwang and Mendell (42), or by Esquela-Kerscher and Slack (29). There are three reports that miR145 is down-regulated in cancer cells (26,27,43), and Kent and Mendell (44), in their review, list miR145 as a tumor suppressor miR. In none of these cases, however, were the targeted genes identified. Interestingly, in two of those three references, the down-regulation of miR145 was observed in colon cancer cells (28,43). Another miR reported to inhibit colon cancer cell growth is let-7 (45). Our results, showing that miR145 inhibits colon cancer cells growth, are compatible with those observations. Indeed, this is the first demonstration of a miR that specifically targets a signal transduction molecule of the IGF-IR/insulin receptor axis and inhibits growth of cancer cells. The effect of miR145 on IRS-1 levels and cancer cell growth is in agreement with the frequent observation that IRS-1 is a strong mitogen and an inhibitor of differentiation (3). Ectopic expression of IRS-1 causes tumor growth in mice of myeloid cells that, without IRS-1, undergo differentiation (7,46). IRS-1 is a strong inducer of the ID proteins that inhibit differentiation (39,40) and DeAngelis et al. (8) have shown that transformation by the SV40 T antigen requires tyrosyl phosphorylation of IRS-1. Dalmay and Edwards (47) suggested that the anticancer effect of miR145 may be due to the fact that it targets paxillin. Whereas paxillin is a potential target, we propose that IRS-1 may be an even better one. miR145 has 1093 predicted targets in human and 890 in mouse according to miRBase (December 2006). The 5Ј seed region, positions 2-8 of mature miRNA, is conserved in metazoan and plays a key role in target recognition. The large number of target mRNAs down-regulated by miRs has been studied by Lim et al. (31) using microarray analysis. A similar concept has been adapted to the off-target effects of siRNA. Although siRNA is designed to be perfectly matched with the on-target mRNA, it can also mediate knockdown of dozens to hundreds of other genes via perfect matches between the hexamer or heptamer seed (positions 2-7 or 2-8 of the antisense strand) of an siRNA and the 3Ј-UTR (but not the 5Ј-UTR or open reading frame) of these off-target genes. Because proteome screens of miR targets and/or siRNA off-targets are not as advanced or extensively available as mRNA microarray analysis, identifying potential targets of translational inhibition is still challenging. A more convincing way to prove a phenotype as the consequence of the knockdown of a specific target by RNAi is to design and apply several siRNAs targeting different regions of the same target mRNA. Even though different siRNAs have different offtarget effects, if they all produce the same phenotype by targeting the common "on-target" mRNA, one can draw the conclusion of the correlation of the phenotype with the target gene. In our study, we hypothesized that the malignant growth of colon cell HCT116 is related to the expression of IRS-1 in these cells. By identifying and introducing miR145, a down-regulated miR in colorectal cancers, into HCT116 cells, suppression of translation of IRS-1 by miR145 leads to the inhibition of cancer cell proliferation. As a proof-of-concept control, we also transfected siRNA against IRS-1 into HCT116 cells. Cleavage of miR145 and IRS-1 IRS-1 mRNA by siIRS-1 and knockdown of the IRS-1 protein again inhibited cell growth. Therefore, the relationship between the phenotype (inhibition of cell proliferation) and the target specificity of miR145 on IRS-1 is confirmed. Furthermore, miR-negative control or miR148a-treated cells neither change the levels of IRS-1 protein nor suppress cell proliferation, which again strongly supported our hypothesis that phenotypic inhibition of colon cancer cell proliferation is correlated with the down-regulation of the target protein IRS-1 via IRS-1-specific miR145 or siRNA. However, as already mentioned, this does not exclude the (very strong) possibility that miR145 targets also other mRNAs of the same signaling pathway.
Although expression of miRs by pSuper has been reported several times in the literature, we have had some problems in obtaining a good expression. In fact, in our hands, the expression of a mature miR145 in pSuper was substantial only in the presence of 20 genomic flanking nucleotides on each side of the precursor miR145 sequence. We have no explanation for the discrepancy with the literature, except that it may be a peculiarity of miR145. miR148a is another computer-predicted miR targeting IRS-1. However, in our study, it can neither down-regulate IRS-1 translation nor inhibit colon cancer cell proliferation. Interestingly, a review published recently by Cummins et al. (33) listed differentially expressed miRNAs in colorectal cancer. In this list, while miR145 and miR143 are down-regulated, miR148a is up-regulated in colorectal adenocarcinomas compared with matched normal colonic epithelia. This coincidence sheds light on the potential usage of miR145 as an anti-colon cancer therapeutic by targeting IRS-1. In conclusion, we have demonstrated rigorously that miR145 does indeed target IRS-1 and has a profound biological effect on human colon cancer cells.