miR-15a and miR-16 Are Implicated in Cell Cycle Regulation in a Rb-Dependent Manner and Are Frequently Deleted or Down-regulated in Non–Small Cell Lung Cancer

Lung cancer is the leading cause of cancer-associated death and is responsible for more deaths than the next three most common tumors combined (breast, prostate, and colon; ref. 1). The high mortality rate of this disease entity is primarily due to the fact that it is only diagnosed at an advanced stage. Lung cancer comprises several histologic types including small cell lung cancer and non–small cell lung cancer (NSCLC); the latter can be further subdivided into two major types, squamous cell carcinoma and adenocarcinoma (2). Squamous cell carcinomas usually arise from the major bronchi, whereas adenocarcinomas arise from distant airway bronchioles and alveoli.

Lung cancer is characterized by multiple genetic changes affecting different oncogenes or tumor suppressor genes involved in cell cycle control, DNA repair, and apoptosis (2). The most common alterations are overexpression of cyclin D1 (CCND1) and Bcl2, mutations of KRAS or members of the ERBB family, and mutations or inactivation of Rb, p16^INK4a, and TP53 (2). Altered expression of marker genes in neoplastic tissues may be due to genetic alterations or promoter methylation, but in many cases the mechanism of dysregulation is unknown.

MicroRNAs (miRNA) constitute a novel class of regulatory molecules at the posttranscriptional level, which are involved in tumorigenesis (3). These short RNAs of 19 to 25 nucleotides play key roles in a wide variety of biological processes including cell fate specification, proliferation, cell death, and energy metabolism (4). Several miRNAs have been identified, which are directly involved in tumorigenesis of lung cancer and are considered as prognostic markers for overall survival or predictive markers for chemotherapy (5, 6).

miR-15a and miR-16-1 are located at chromosome 13q14, a region that is deleted in 68% of chronic lymphocytic leukemia (7). Cimmino and colleagues showed that miR-15a/miR-16 expression is inversely correlated to Bcl2 expression and that Bcl2 repression by these miRNAs induces spontaneous apoptosis in a leukemic cell line model (8). In a colon carcinoma cell line, however, the same miRNAs coordinately regulate different mRNA targets, including CDK6, CARD10, CDC27, and C10orf46, which act in concert to control cell cycle progression (9), suggesting that these miRNAs may have cell type-specific functions. The spectrum of potential targets of miR-16 was further expanded by the recent work of Chen and colleagues (10), Liu and colleagues (11), and Bonci and colleagues (12), who showed that G1 cyclins are also regulated by this miRNA.

To investigate if miR-15a/miR-16 are important for the tumorigenesis of lung cancer, we collected tissues from squamous cell carcinomas and adenocarcinomas and corresponding normal tissues by laser capture microdissection and compared these tissues for miRNA expression. We report that miR-15a/miR-16 are deleted or down-regulated in the majority of NSCLC. In addition, we investigated the molecular mechanism underlying miR-15a/miR-16 regulation in different NSCLC cell lines. We show that miR-15a/miR-16 is able to induce cell cycle arrest in G₁-G₀ in a Rb-dependent manner and that G₁ cyclins are physiologic targets of these miRNAs. We propose two alternative pathways by which NSCLC cells escape miR-15a/miR-16–induced cell cycle arrest: (a) down-regulation of miR-15a/miR-16 and (b) inactivation of the Rb gene.

Cell lines and culture conditions. The NSCLC cell lines Calu-1, Calu-6, A549, H2009, H1299, and H358 were obtained from the American Type Culture Collection. All cell lines were cultured in Iscove's modified Dulbecco's medium supplemented with 2 mmol/L l-alanyl-l-glutamine, 1% penicillin/streptomycin, and 5% fetal bovine serum (Sigma) at 37°C and 5% CO₂.

Constructs. Luciferase reporter constructs were generated containing a firefly luciferase gene cloned between HindIII and XbaI sites of pcDNA3.0 (Invitrogen). DNA fragments encompassing the 3′-untranslated region (UTR) of CCND1, cyclin D2 (CCND2), and cyclin E1 (CCNE1), respectively, were amplified from genomic DNA and cloned into the XbaI site of the luciferase construct. Luciferase plasmids containing one or three copies of a miR-15a/miR-16 target site from CCND1, CCND2, or CCNE1, respectively, were obtained by cloning double-stranded oligonucleotides into the XbaI site of the luciferase plasmid (Supplementary Table S1; Supplementary Fig. S1). pcDNA3 3xmiR-15a/miR-16 was constructed by cloning three copies of the miR-15a/miR-16 locus in tandem into pcDNA3.0. Rc/CMV cyclinD1 and Rc/CMV cyclin E expression constructs (13) were obtained from Addgene. The miR-15a/miR-16 target sites were deleted from Rc/CMV CCNE1 by PCR amplification, giving rise to Rc/CMV cyclin EΔ3′-UTR. Primers used for amplification are indicated in Supplementary Table S1.

Transfection. Cells were seeded in culture flasks 24 h before transfection. Cotransfections with plasmid DNA were done using Effectene reagent (Qiagen); all other transfections were done using HiPerFect (Qiagen). Transfection was done using 10 to 20 nmol/L of a mixture of equal amounts of hsa pre-miR-15a and hsa pre-miR-16 precursors or pre-miR miRNA precursor control 1 (Ambion), 60 nmol/L Rb-kd small interfering RNA (siGENOME SMARTpool; Dharmacon), 60 nmol/L siCONTROL nontargeting Pool 2 (Dharmacon), 100 nmol/L of a mixture of equal amounts of anti-miR-15a and anti-miR-16 inhibitors or anti-miRNA inhibitor control (Ambion), pCMV-Rb (14) or empty control plasmid, firefly luciferase constructs, and Renilla luciferase reporter plasmid pGl4.74 (Promega) or expression constructs as described above. Transfection efficiency of short RNAs and plasmid DNA was monitored using siGloGreen transfection indicator (Dharmacon) or a RFP expression plasmid, respectively.

Luciferase activity assay, cell cycle analysis, and apoptosis assay. Luciferase activity assays were done 48 h after transfection using a dual-luciferase reporter assay system (Promega) and an Infinite 200 reader (Tecan).

Cell cycle analysis was done essentially as described (9). Cells were analyzed using a LSR II flow cytometer (BD Biosciences) and FlowJo 8.8.4 software (Tree Star).

For apoptosis assays, floating and adherent cells were harvested 24 to 72 h after transfection, combined, and washed with PBS. Annexin V (1:100) either alone or in combination with 10 μg/mL propidium iodide (Sigma) was added to the cells and samples were analyzed within 30 min after staining. Quantification of fluorescence was done by flow cytometry as described above.

RNA isolation and real-time PCR. Total RNA was extracted from cultured cells using the miRVana RNA isolation kit according to the manufacturer's instructions (Ambion). TaqMan miRNA assays (Applied Biosystems) were used to quantify the expression of mature hsa miR-15a/miR-16. Reverse transcription was done using the TaqMan miRNA reverse transcription kit. Reverse transcription of mRNAs was done using TaqMan reverse transcription reagents (Applied Biosystems). Quantitative PCR of CCND2 was done using a TaqMan assay (Applied Biosystems); all other amplifications were done using Quantitect Primer assays (Qiagen). Quantitative PCR was done in a Real-time PCR System 7500 (Applied Biosystems). Gene expression of miRNAs was calculated relative to RNU48 or RNU6B; mRNA levels were normalized to the level obtained for GAPDH, hb2m, HPRT-1, or RPL13A, respectively. Changes in expression were calculated using the ΔΔCt method.

Western blot analysis and immunohistochemistry. Proteins were separated by SDS-PAGE and Western blotting was done using polyvinylidene difluoride membrane (Millipore). Unspecific sites were blocked in 5% nonfat dry milk or 5% bovine serum albumin at room temperature for 1 h. Antibodies directed against CCND1 (clone DCS-6; DAKO; diluted 1:100), CCNE1 (clone 13A3; Neomarkers; 1:200), Rb (clone 3C8; QED Bioscience; 1:1,000), phospho-Rb (Ser⁸⁰⁷/Ser⁸¹¹; Cell Signaling Technology; 1:1,000), and α-tubulin (clone B512; Sigma; 1:5,000) were used. Secondary antibody was goat anti-mouse horseradish peroxidase or goat anti-rabbit horseradish peroxidase (Bio-Rad) used at 1:5,000 or 1:7,000, respectively.

For immunohistochemistry, 4 μm formalin-fixed, paraffin-embedded sections were treated essentially as described (15). Anti-Bcl2 (clone 124; DAKO), anti-CCND1, anti-CCNE1, and anti-Rb antibodies were used at 1:30, 1:25, 1:25, and 1:100 dilutions, respectively. Mouse IgG1 (DAKO; 1:20) was used as a negative control. Sections were incubated with EnVision+ system (labeled polymer horseradish peroxidase anti-mouse; DAKO) for 30 min at room temperature, developed in 3,3′-diaminobenzidine (Sigma) for 8 min, and counterstained with hematoxylin.

Laser capture microdissection and loss of heterozygosity analysis. Formalin-fixed, paraffin-embedded sections were deparaffinized and stained with methyl green (Merck). Approximately 1,000 tumor cells were captured onto an adhesive cap using a PALM Microbeam (PALM; Zeiss). The dissected material was resuspended in digestion buffer (RecoverAll; Ambion). tRNA (120 ng) was added to the samples and incubated for 10 min at 90°C. All subsequent purification steps were done following the manufacturer's instructions, except that the DNase treatment step was omitted. Loss of heterozygosity (LOH) analysis was done as described (7) using a genetic analyzer (ABI Prism 3100 Avant; Applied Biosystems) and the GeneMapper software 4.0.

All experiments using human specimens were done according to the ethical guidelines of the Institute of Pathology at the University of Bern and were reviewed by the institutional review board.

Statistics. Statistical differences were calculated using unpaired two-tailed Student's t test. A probability of P ≤ 0.05 was considered statistically significant. Statistical significance of inverse correlation was calculated by the N-1 χ² test (2 × 2 table Pearson's analysis; ref. 16).

miR-15a and miR-16 are frequently down-regulated in lung cancer. It has been shown previously that miR-15a/miR-16 can induce apoptosis or cell cycle arrest depending on the cell line (8–11). To address their role in NSCLC, we analyzed 11 adenocarcinomas and 12 squamous cell carcinomas of the lung from the archive of the Institute of Pathology at the University of Bern. Quantification of RNAs is challenged by extensive fragmentation and modification of nucleic acids during formalin fixation (17). To address this problem, nucleic acids were subjected to a heat treatment before RNA extraction, which reverses methylol groups introduced during formalin fixation. Under these conditions, fresh and formalin-fixed, paraffin-embedded tissues gave rise to similar miR-16 levels as indicated by quantitative PCR (Supplementary Fig. S2A). The quality of miRNAs may also depend on the postoperative period before fixation. Experimental tissue samples were left at room temperature for up to 5 h before fixation, which, however, did not affect miR-16 quantification (Supplementary Fig. S2A). In addition, detection of miR-16 was linear over a wide concentration range (Supplementary Fig. S2B), indicating that miR-16 can be quantified accurately from formalin-fixed laser capture microdissected material.

Tumor tissues from adenocarcinomas and squamous cell carcinomas were compared with matched normal tissue from alveolar or bronchial epithelium, respectively, which had been microdissected from the same slide. miR-16 was down-regulated in 82% (9 of 11) of adenocarcinomas (P < 0.001) and 67% (8 of 12) of squamous cell carcinomas (in 5 of which down-regulation was statistically significant; P < 0.05; Fig. 1A). Interestingly, miR-16 was significantly up-regulated in 26% of tumors. miR-15a gave rise to a similar expression pattern as miR-16 (data not shown) but was more difficult to detect due to its low abundance. Both miRNAs were also significantly down-regulated in the NSCLC cell lines Calu-1, Calu-6, A549, H2009, H1299, and H358 (Fig. 1B).

In the majority of cases, down-regulation of miR-15a/miR-16 was associated with a deletion of one allele of the miR-15a/miR-16 locus. This was based on the finding that two microsatellites, D13S273 and D13S272, flanking the miR-15a/miR-16 locus revealed a LOH in 73% (8 of 11) of adenocarcinomas and 83% (10 of 12) squamous cell carcinomas (Fig. 1A).

miR-16 expression inversely correlates to CCND1 in NSCLC. To investigate if dysregulation of miR-16 may affect the expression of potential targets, CCND1, CCNE1, and Bcl2 were analyzed by immunohistochemistry. Analysis of the same tumor area on serial tissue sections revealed an inverse correlation in the expression of miR-16 and CCND1 in 19 of 23 samples (Table 1; Supplementary Fig. S3), suggesting that CCND1 is a physiologic target. In contrast, no inverse correlation was observed for CCNE1 or Bcl2. Based on these results, we cannot exclude that the latter genes are targets, because additional layers of gene regulation may exist, which contribute to CCNE1 and Bcl2 expression.

Cellular phenotypes triggered by miR-15a and miR-16 in NSCLC cell lines. To investigate the cellular phenotypes triggered by these miRNAs, NSCLC cell lines were cotransfected with miR-15a/miR-16 precursors and analyzed for apoptosis and cell cycle arrest. None of these cell lines underwent spontaneous apoptosis as indicated by propidium iodide staining (Fig. 2A), Annexin V staining, or morphologic changes (data not shown). However, in 5 of 6 cell lines, miR-15a/miR-16 induced cell cycle arrest in G₁-G₀ in a significant percentage of the cell population (P < 0.01; Fig. 2B). This phenotype was more pronounced when cells were treated with nocodazole 24 h after transfection with miRNA precursors (P < 0.001). In the case of A549, 33% of the cell population had undergone arrest in G₁-G₀, whereas only 8% of the control transfected with an unrelated RNA were in this phase of the cell cycle (Fig. 2B and C). Comparable results were obtained for the cell lines Calu-1, Calu-6, H1299, and H358 (Fig. 2B; Supplementary Fig. S4). In marked contrast, only 2% of H2009 cells were in the G₁-G₀ phase on cotransfection with miR-15a/miR-16 precursors (Fig. 2B and C). It can be excluded that the low percentage of cells in G₁-G₀ was due to a low transfection efficiency, because ∼90% of H2009 cells were transfected with siGloGreen (fluorescently labeled nonfunctional small interfering RNA). Thus, the cell cycle regulatory activity of miR-15a/miR-16 is cell line-specific.

Down-regulation of endogenous G₁ cyclin mRNAs by miR-15a and miR-16. A search of the TargetScan database allowed the identification of CCND1, CCND2, CCND3, and CCNE1 as predicted targets of miR-15a/miR-16 (Supplementary Fig. S5). To analyze if miR-15a/miR-16 are able to regulate the steady-state level of mRNA of these potential target genes, cells were cotransfected with miR-15a/miR-16 precursors and analyzed for the expression of G₁ cyclin mRNAs by quantitative PCR. CCND2 did not yield any PCR product in most of the cell lines analyzed. The results of the other three G₁ cyclins are shown in Fig. 3A. All cell lines transfected with miR-15a/miR-16 precursors gave rise to reduced levels of CCND1, CCND3, and CCNE1 mRNAs relative to the control transfected with an unrelated RNA.

Cotransfection of A549 and H2009 cells with miR-15a/miR-16 precursors also resulted in 50% to 70% less CCND1 and CCNE1 proteins relative to the control (Fig. 3B). Thus, both cell lines are equally able to down-regulate G₁ cyclins in response to miR-15a/miR-16. In agreement with these results, the level of phospho-Rb was significantly reduced (Fig. 3B).

To confirm that cell cycle arrest was due to down-regulation of G₁ cyclins, A549 was cotransfected with miR-15a/miR-16–refractory expression constructs and miR-15a/miR-16 precursors. Overexpression of CCNE1 and CCND1 was confirmed by Western blotting (Supplementary Fig. S6). As shown in Fig. 3C, cotransfection with miR-15a/miR-16 precursors in combination with Rc/CMV cyclinD1 (P = 0.003) or Rc/CMV cyclin EΔ3′-UTR (P = 0.04) in each case leads to a significant decrease in G₁-G₀ arrest compared with cells cotransfected with miR-15a/miR-16 precursors in combination with a control construct. Thus, both CCND1 and CCNE1 can partially rescue miR-15a/miR-16–induced cell cycle arrest.

G₁ cyclins are physiologic targets of miR-15a and miR-16 in NSCLC cell lines. To investigate if G₁ cyclins are regulated directly by miR-15a/miR-16, a series of luciferase constructs were made containing predicted target sequences from the different G₁ cyclins cloned downstream of the luciferase reporter gene (Fig. 4A; Supplementary Fig. S1) and cotransfected into H2009 cells together with miR-15a/miR-16 precursors. CCND1 contains two potential target sites for miR-15a/miR-16 (Supplementary Fig. S5). A construct containing the first target site within the 3′-UTR of CCND1 (Luc 1xTS) resulted in 35% luciferase activity relative to the parental Luc construct (Fig. 4B). The activity obtained with this construct was comparable with a construct containing a part of the 3′-UTR of CCND1 encompassing both target sites (Luc 3′-UTR). A third construct containing three copies of a single target site in tandem (Luc 3xTS) gave rise to 25% activity. To confirm the specificity of the assay, mutations were introduced into the target sequence of miR-15a/miR-16 giving rise to Luc mTS (Fig. 4A; Supplementary Fig. S1). The activity obtained with this construct was almost restored to the activity obtained with the Luc construct. CCND1 and CCNE1 constructs gave rise to comparable luciferase activities (Fig. 4B). In contrast, CCND2 seems to be less responsive to miR-15a/miR-16. Insertion of a single target site (TS-1 or TS-2) or three copies of one target site (TS-2) in tandem into the Luc construct did not significantly reduce luciferase activity (Fig. 4B). However, a region of the 3′-UTR encompassing two target sites conferred efficient down-regulation of luciferase activity. Thus, one target site may not be sufficient for down-regulation of luciferase activity. Again, comparable results were obtained in H2009 and A549 cells (Fig. 4B).

We next investigated if G₁ cyclins are regulated by endogenous miR-15a/miR-16. H2009 cells were transfected with the Luc 3xTS construct containing three target sites from the CCND1 or CCNE1 genes, respectively, or with the Luc 3′-UTR construct containing CCND2-specific sequences. All three constructs gave rise to luciferase activities, which were significantly lower than the activity obtained with the Luc construct (Fig. 4C). Relative activities were 47% (P < 0.00001) for the CCND1 Luc 3xTS construct, 33% (P = 0.003) for the CCNE1 Luc 3xTS construct, and 62% (P = 0.001) for the CCND2 Luc 3′-UTR construct. These results clearly indicate that endogenous levels of miR-15a/miR-16 are sufficient to down-regulate G₁ cyclins. In agreement with these results, anti-miRs specific for miR-15a/miR-16 were able to completely restore the activity obtained with the Luc 3xTS constructs from CCND1 and CCNE1 (P < 0.002; Fig. 4C) to the level obtained with the Luc construct.

These results indicate that although miR-15a/miR-16 are significantly reduced (Fig. 1B), they are still able to control G₁ cyclins in NSCLC cell lines. To investigate if down-regulation of miR-15a/miR-16 is a mechanism by which NSCLC cells induce overexpression of target genes, H2009 cells were cotransfected with an expression construct containing three copies of the miR-15a/miR-16 locus and CCND1 Luc 3xTS. In these cells, the level of miRNAs was restored almost to the level obtained in normal lung tissue. These cells exhibited two times less luciferase activity relative to the control (Supplementary Fig. S7), showing that a reduction in miR-15a/miR-16 activity as we observed in our lung cancer cell lines leads to overexpression of CCND1.

miR-15a– and miR-16–induced cell cycle arrest depends on Rb expression. Based on our results, we may conclude that miR-15a/miR-16–induced cell cycle arrest is due to down-regulation of G₁ cyclins, but this may not apply for H2009, which is resistant. H2009 cells differ from the other NSCLC cell lines by the fact that they lack Rb. Cyclin D in complexes with CDK4 and CDK6 and cyclin E in a complex with CDK2 regulate progression through the G₁-S boundary of the cell cycle. These complexes phosphorylate and thereby prevent Rb from binding to E2F, which, on release, drives cells from G₁ into S phase (reviewed by ref. 18). Thus, Rb-deficient cells no longer depend on cyclin D (19, 20) and therefore may not respond to miR-15a/miR-16. To test this hypothesis, Rb was expressed in H2009 cells and analyzed for miR-15a/miR-16–induced cell cycle arrest. Cotransfection with miR-15a/miR-16 precursors and an empty control plasmid induced cell cycle arrest in 7.4 ± 1% of the population (Fig. 5A). The percentage of cells in the G₁-G₀ phase of the cell cycle increased to 23 ± 1% on cotransfection with a Rb expression plasmid and control precursor RNA (P = 0.001), indicating that Rb per se is able to induce cell cycle arrest in a relatively high number of cells. However, Rb in combination with miR-15a/miR-16 precursors induced cell cycle arrest in significantly more cells (30 ± 3%; P = 0.01) than Rb in combination with the control precursor (Fig. 5A). Thus, miR-15a/miR-16 depend on Rb to exert their phenotype. The expression of Rb protein was not affected by coexpression with miR-15a/miR-16, but phospho-Rb was reduced by 50% (Fig. 5B).

To verify our results, the complementary experiment was done in A549 cells in which the Rb gene was knocked down by RNA interference. These cells expressed three times less Rb protein than the control (Fig. 5C). As expected, A549 cells transfected with small interfering RNAs against Rb were significantly more resistant to miR-15a/miR-16–induced cell cycle arrest (11 ± 1% in G₁-G₀) than cells transfected with control small interfering RNA (23 ± 2% in G₁-G₀; P < 0.001; Fig. 5D). In contrast, Rb small interfering RNA in combination with precursor control RNA had no effect (3 ± 1%). In conclusion, loss of Rb confers resistance to miR-15a/miR-16–induced cell cycle arrest.

Deregulated cell proliferation is a key mechanism for neoplastic progression (21). This study shows that miR-15a/miR-16 are negative regulators of cell cycle progression in NSCLC. We propose two mechanisms by which tumor cells can escape miR-15a/miR-16–induced cell cycle arrest either by down-regulation of miR-15a/miR-16 or by inactivation of Rb.

miR-15a/miR-16 are expressed at reduced levels in ∼70% of adenocarcinomas or squamous cell carcinomas and in 6 NSCLC cell lines; in the majority of cases, this is associated with a LOH of the miRNA locus. However, other mechanisms may contribute to miRNA down-regulation. For example, miRNA processing is compromised, owing to reduced expression of Dicer in a significant number of lung cancers (22). Transcription of the miR-15a/miR-16 locus may also be affected. Chung and colleagues showed that the promoter activity was reduced by PAX5 in B-cell neoplasm (23), and it is likely that a similar mechanism may exist in lung cancer. In addition, a point mutation in the miR-16 primary transcript has been linked to diminished expression of the mature miRNA (24). Our results suggest that down-regulation of miR-15a/miR-16 may contribute to tumor growth because this is directly coupled to increased levels of G₁ cyclins.

That miR-15a/miR-16–induced cell cycle arrest depends on Rb was evidenced by our findings that H2009 cells lacking Rb are completely resistant to miR-15a/miR-16, whereas introduction of a functional copy of Rb into these cells renders them more sensitive. Consistent with these results, down-regulation of Rb in A549 by RNA interference confers resistance to these miRNAs. This mechanism may apply for up to 14% of squamous cell carcinomas and 33% of adenocarcinomas, which are Rb-negative (25). We found that Rb was down-regulated in 4 of 11 adenocarcinomas (tumors 5-7 and 9) and 3 of 12 squamous cell carcinomas (tumors 1, 4, and 12; see Fig. 1). It is unlikely that two Rb-related proteins, p107 and p130, can compensate for a loss of Rb, because their functions are distinct from that of Rb (26). In addition, these proteins are normally expressed in H2009 cells, which are refractory to miR-15a/miR-16 (27). In conclusion, 87% (20 of 23) of NSCLC had either down-regulated miR-15a/miR-16 or inactivated Rb.

Hemizygous or homozygous loss of 13q14 has been associated with numerous malignancies including chronic lymphocytic leukemia (7), mantle cell lymphoma (28), multiple myeloma (29), breast cancer (30), and high-grade carcinoma of the prostate (31), suggesting that these deletions are of pathogenetic significance. miR-15a/miR-16 are presumably the long-sought candidate tumor suppressors of this region and are proven to be important regulators of Bcl2 expression in chronic lymphocytic leukemia (7, 8). miR-15a/miR-16 are also frequently down-regulated in chronic lymphocytic leukemia and pituitary adenomas (32). Interestingly, both miRNAs are frequently up-regulated in cervical cancer (33). Because these cells normally express an inactive form of Rb, we may conclude that cell cycle progression of cervical cancer cells no longer depends on miR-15a/miR-16 activity. Because the level of miR-15a/miR-16 is normally high in these tumors, this may suggest a dual role of miR-15a/miR-16 as tumor-suppressing and oncogenic miRNAs. A similar mechanism may also exist in NSCLC, because both miRNAs were up-regulated in 28% (5 of 18) of the tumor samples, in 3 of which Rb was down-regulated.

What are the targets that are responsible for miR-15a/miR-16–induced cell cycle arrest in NSCLC? miRNAs including miR-15a/miR-16 can affect hundreds of mRNAs (9), which renders it difficult to identify the biologically relevant targets. Although it is believed that miR-15a/miR-16-induced cell cycle arrest is mediated by targeting G₁ cyclins (10, 11), it cannot be excluded that this is induced by an indirect mechanism, for example, by targeting mRNAs from essential genes. Based on our findings, however, that miR-15a/miR-16-induced cell cycle arrest depends on Rb, we can conclude that this is due to down-regulation of proteins directly involved in cell cycle control.

It was shown recently that CCND1, CCND3, CCNE1, CDK4, and CDK6 are direct targets of miR-15a/miR-16 (9–12). We confirm that CCND1, CCND3, and CCNE1 are also targets in NSCLC cell lines and provide evidence that CCND2 is an additional target of these miRNAs. For most of these targets, it was unclear if they are also regulated by physiologic concentrations of miR-15a/miR-16. We show that CCND1, CCND2, and CCNE1 are directly regulated by miR-15a/miR-16 under physiologic conditions. Using luciferase reporter constructs containing target sequences from the different cyclin genes, we showed that endogenous miRNAs are able to down-regulate luciferase activity and that this effect is reversed by cotransfection with anti-miRs against miR-15a/miR-16. miR-16 inversely correlates with CCND1 protein in NSCLC, which is in line with the finding that this is a physiologic target. In addition, miR-15a/miR-16–induced cell cycle arrest can be partially restored by overexpression of CCND1 or CCNE1.

In conclusion, we show that miR-15a/miR-16 induce cell cycle arrest by down-regulation of G₁ cyclins, whereas NSCLC cells escape these growth-inhibitory signals either by down-regulation or loss of function of Rb, by down-regulation of miR-15a/miR-16, or by other means. This may constitute a general mechanism implicated in tumorigenesis, because LOH 13q14 and Rb inactivation are frequent events in various tumors. Furthermore, our data suggest that miR-15a/miR-16 might be used as therapeutic agents in Rb-proficient NSCLC.

No potential conflicts of interest were disclosed.

Grant support: Swiss National Science Foundation (E. Vassella).

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

We thank Mario Tschan for plasmids and cell lines, Bernadette Wyder for introduction into flow cytometry, and Sabine Jakob and Silvia Rihs for protocols and helpful discussions.