Tumor Suppression by IFN Regulatory Factor-1 Is Mediated by Transcriptional Down-regulation of Cyclin D1

IFN regulatory factors (IRF) are a family of transcription factors whose activity is implicated in a wide variety of biological processes including growth regulation, pathogen responses, immune activation, and inflammation (1, 2). IRF-1 is a transcriptional activator that binds to specific DNA sequences in promoters of genes that exert multiple effects, including cell growth inhibition (3, 4).

Interestingly, IRF-1–mediated inhibition of proliferation is much stronger in oncogenically transformed cells than in nontransformed cells. The inhibition of transformation by different oncogenes and the growth-suppressive effects in different kinds of cancer cells suggest that a key pathway is targeted by IRF-1 (5, 6).

IRF-1 has been identified as a tumor suppressor. It is lost, mutated, or rearranged in several cancers including hematopoietic, gastric, and breast cancers (7, 8). Consistent with its role as a tumor suppressor, it was previously shown that a variety of transformed cell types respond to forced expression of IRF-1 by inhibition of cell proliferation and reversion of the transformed phenotype (9). IRF-1 suppresses tumor growth by two mechanisms: an intrinsic effect and an enhanced immune cell recognition of the tumor (6, 10–12). However, loss of IRF-1 expression is not associated with increased rates of spontaneous tumor development in mice (13).

To resolve the mechanism by which IRF-1 inhibits oncogenic transformation, we created a cell line in which the expression of two cooperating oncogenes (c-myc and c-H-ras) can be transcriptionally regulated by doxycycline. Depending on the presence of doxycycline, these cells behave normal or oncogenically transformed. In these cells, IRF-1 is expressed as a β-estradiol–activatable IRF-1hER fusion protein. After IRF-1 activation, the oncogene-mediated acceleration of the cell cycle is reverted. A complete IRF-1–mediated reversion of the oncogenic phenotype is observed in soft agar growth assays, and inhibition of tumor growth is observed in nude mice as long as IRF-1 is active (6).

To identify the target genes of IRF-1 action, we used microarray analysis. Our data revealed that IRF-1 reverts 60% of all genes that are deregulated by the myc/ras–mediated transformation. Activation of IRF-1 resulted in a decreased expression of the central G₁-S phase regulator cyclin D1. This effect is mediated by inhibition of the upstream mitogen-activated protein kinase (MAPK) kinase/extracellular signal-regulated kinase (ERK) pathway. IRF-1–mediated effects on cell cycle progression were found to be eliminated by ectopic expression of cyclin D1. Consistent with this, decrease of cyclin D1 expression by RNA interference (RNAi) experiments prevents transformation by myc/ras expression, showing that cyclin D1 is the key to myc/ras–mediated transformation. Our findings show that IRF-1 mediates its tumor-suppressive effects by inhibiting oncogene-induced cyclin D1 expression.

Plasmids. Constructions were carried out by standard procedures (14). The constitutive expression vector for IRF-1-hER and the bicistronic construct for tetracycline-regulatable c-myc and H-ras expression (pHBTMR) was described earlier (6). pBabe-cyclin D1 was provided by B. Law (Vanderbilt University School of Medicine, Nashville, TN). Cyclin D1 promoter luciferase plasmid (−1745D1-luc) was obtained from R.G. Pestell (Thomas Jefferson, University, Philadelphia, PA). The constitutive expression vector for Renilla luciferase pBCRluc has been previously described (15). pSuperCd1a was obtained from M.F. Olson (The Beatson Institute for Cancer Research, Glasgow, United Kingdom) (16). Lentiviral plasmids pHR-SIN-SR and dcH1-GL4-SR for shRNA delivery were kindly provided by M. Scherr (Hannover Medical School, Hannover, Germany) (17). For generating dcH1-cd1a-SR, the H1-shRNA cassette from pSuperCd1a was released from pSuperCd1a plasmid by digestion with SmaI and HincII, and the resulting fragment was subsequently blunt-end ligated into the SnaBI site of pHR′-SIN-SR.

Cells and cell culture. The pHBTMR plasmid and the IRF-1 fusion protein encoding plasmid were stably engineered into NIH3T3 cells (myc/rasNIH3T3IH; refs. 6, 9). Cells were grown in DMEM (Sigma, Taufkirchen, Germany) plus 10% of estrogen-free FCS, antibiotics, glutamine, and doxycycline (2 mg/mL; Sigma) as indicated, IRF-1hER fusion protein was activated by β-estradiol (1 μmol/L; Merck, Darmstadt, Germany). For generation of stable cell lines, DNA was transfected using calcium-phosphate coprecipitation.

DNA microarray hybridization and data analysis. Total RNA was isolated from cells of two independent experiments by using TRIZOL (Invitrogen, Karlsruhe, Germany). The quality and integrity of the total RNA were confirmed by using the Agilent Technologies 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA). Biotin-labeled target synthesis was done by the manufacturer (Affymetrix, Memphis TN). About 12.5 μg of each biotinylated cRNA preparation were fragmented and placed in a hybridization cocktail that contained four biotinylated hybridization controls (BioB, BioC, BioD, and Cre) as recommended by the manufacturer. All samples were hybridized to the same lot of Affymetrix MGU74A for 16 h. The GeneChips were washed, stained with streptavidin-phycoerythrin, and read by using an Affymetrix GeneChip fluidic station and scanner.

Analysis of microarray data was done by using the Affymetrix Microarray Suite 5.0, Affymetrix MicroDB 3.0, and Affymetrix Data Mining Tool 3.0. For normalization, all array experiments were scaled to a target intensity of 150. Filtering of the results was done as follows: genes were considered as regulated when their fold change is ≥1.5 or less than or equal −1.5; the statistical variable for a significant expression level of each duplicate is >0.95. Original data files for all three arrays were uploaded in MIAME format for expression arrays at GEO accession no GSE 6505. Additionally, the signal difference of a certain gene should be >50. The cluster analysis was done with the TIGR Multiple Experiment Viewer (18). The signals were normalized before clustering. Experiments were clustered into a hierarchical tree that uses the Euclidean distance measurement and average linkage algorithm. Genes were clustered by using k-means algorithm with Euclidean distance measurement.

Preparation of recombinant lentiviral supernatants and lentiviral transduction. VSV.G-pseudo-typed lentiviral particles were generated by calcium phosphate cotransfection of 293T cells. Lentiviral preparations were titered in triplicate by serial dilutions of 4 × 10⁴ NIH3T3 cells in 24-well plates. The proportion of RFP-positive cells was determined 48 h after transduction by fluorescence-activated cell sorting (FACS) analysis (Becton Dickinson, Heidelberg, Germany). Lentiviral supernatants were used to transduce myc/rasNIH3T3IH cells. Cells were incubated for 16 h with the lentiviral supernatants and 2 μg/mL polybrene. Infected RFP-positive cells were isolated by FACS (Becton Dickinson).

Transfection and dual luciferase assay. −1745D1-luc were transfected using Metafectene (Biontex, Martinsried, Germany) together with pBCRluc. A total of 1 μg DNA was transfected. Forty-eight hours after transfection, cell lysates were prepared, and the activities of firefly and Renilla luciferase were assayed using the Dual Luciferase kit (Promega, Madison, WI). Signals were normalized for transfection efficiency to the internal Renilla luciferase controls. All experiments were done at a minimum of three times before calculating mean and SD as shown in the figures.

Colony formation assay. Anchorage-independent growth capacity was determined by assessing the colony formation efficiency of cells suspended in soft agar. Cells (1 × 10³) were seeded in 50 μL of 0.3% overlay agar in microtiter plates coated with 50 μL of 0.6% underlay agar. The induction medium was added to the top (50 μL per well). Colonies were counted 1 to 3 weeks after plating. Mean values of triplicates were plotted.

Western blot analysis and antibodies. Western blot analysis was accomplished according to standard procedures using enhanced chemiluminescence detection (Amersham, Munich, Germany). The following primary antibodies were used: estrogen receptor (HC-30), cyclin D1 (C-20), cyclin E (M-20), cyclin-dependent kinase 4 (CDK4; H-22) from Santa Cruz Biotechnology (Santa Cruz, CA); cyclin D3 (clone 1) from Becton Dickinson; MEK (#9122), phosphorylated Ser¹⁵ MEK (#9121), retinoblastoma protein (pRb; 4H1), phosphorylated pRb (Ser^807/811), ERK (#9102), anti–phosphorylated Ser¹⁵ ERK (#9101) from Cell Signaling (Danvers, MA); actin (Ab-1) from Oncogene (Cambridge, United Kingdom). Horseradish peroxidase–conjugated anti-rabbit and anti-mouse antibodies (Amersham) were used as secondary antibodies.

Cell proliferation and metabolic activity. Cell numbers were measured by counting cells 6, 12, 24, 48, 72, and 97 h after seeding. For determination of metabolic activity, 2 × 10³ cells per well were seeded into microtiter plates, and serial dilutions (1:1) were done allowing several independent measurement points. Cells were treated with the indicated concentration of β-estradiol. Metabolic activity was determined using the WST kit (Roche, Mannheim, Germany) following the manufacturer's instructions. For determination of metabolic activity and proliferation, mean values of triplicates were plotted.

CDK4 kinase assay and cell cycle analysis. Kinase assays were done as previously described (19, 20). CDK4 was immunoprecipitated with anti-CDK4 antibody (Santa Cruz Biotechnology) and 100 μL Protein A/G Agarose (Santa Cruz Biotechnology). Rb (amino acids 769–921; Santa Cruz Biotechnology) was used as substrate. All kinase assays were done twice.

For cell cycle analysis, cells were grown to 70% confluence, collected, and fixed with 80% methanol. The fixed cells were resuspended in PBS containing 50 mg/mL of each RNase A and propidium iodide (20 μg/mL). Stained cells were analyzed for relative DNA content by a FACScan analyzer (Becton Dickinson). The percentage of cells residing in the sub-G₁, G₀-G₁, S, and G₂-M phases was determined using the ModFit software.

Reverse transcription-PCR/RNA extraction and reverse transcription-PCR. Total RNA was extracted from myc/rasNIH3T3IH cells by using the RNeasy kit (Qiagen, Hilden, Germany). RNA was reverse transcribed using the SuperScript First-Strand Synthesis kit (Invitrogen). Levels of cyclin D1 and actin cDNA were detected by real-time PCR on the LightCycler (Roche). Expression levels were standardized for the housekeeping gene β-actin. To assess the specificity of the amplified PCR product, melting curve analysis was done. The following primers are used for cyclin D1 (5′-AGTGCGTGCAGAAGGAGATT-3′ and 5′-CACAACTTCTCGGCAGTCAA-3′) and actin (5′-TGGAATCCTGTGGCATCCATGAAAC-3′ and 5′-TAAAACGCAGCTCAGTAACAGTCCG-3′).

Tumor growth in nude mice. Male 8-week-old NMRI nude mice (Harlan, Borchen, Germany) were used. Mice were divided into six experimental groups (five mice for each group). 17β-Estradiol releasing pellets (50 mg E2, 60-day release; Innovative Research of America, Sarasota, FL) were implanted into the back of mice 3 days before cell injection. Cells (1 × 10⁶) in 0.1 mL PBS were injected s.c. into the flanks of the mice. Tumor sizes were measured and recorded twice a week using calipers. Data are presented as mean values of the tumor sizes.

Expression of IRF-1 reverts the expression of genes deregulated by transformation. IRF-1 mediates proliferation inhibition and reversion of the transformed phenotype of oncogenic cells. To uncover molecular alterations initiated by IRF-1 action, gene expression profiles of myc/ras–expressing NIH3T3IH cells were established. The cells for this analysis were engineered to express the oncogenes c-myc and c-Ha-ras under the control of the tetracycline-repressible promoter. IRF-1 is expressed as an estradiol-activatable IRF-1hER fusion protein. Treatment of these cells with β-estradiol leads to the activation of IRF-1 (6). The action of oncogenes and IRF-1 is shown in Fig. 1A. Although oncogene expression increased proliferation, concomitant activation of IRF-1 eliminated the proliferation-enhancing effect of oncogene activation.

We have shown earlier that expression of IRF-1 in transformed cells led to the accumulation of the cells in the G₁ phase of the cell cycle (6). In the nontransformed state, the majority of cells (78%) are in the G₁ phase. The transformed cells are characterized by a high portion of cells in S and G₂-M phases of the cell cycle and a reduced proportion in the G₁ phase (59%). Activation of IRF-1 in nontransformed cells had only little effects on cell cycle distribution. However, when IRF-1hER was activated in transformed cells, there was a significant reduction of cells in S and G₂-M phase and a corresponding increase in the G₁ phase of cell cycle from 59% to 84%. Twenty-four hours after IRF-1 activation, the cell cycle distribution of transformed cells was indistinguishable from nontransformed cells (Fig. 1B). As shown earlier, IRF-1 reverted the tumorigenic phenotype of c-myc– and c-H-ras–transformed cells (6). No colonies grew in doxycycline-containing plates when expression of the oncogenes was repressed (Fig. 1C). In the absence of doxycycline, oncogene expression was induced, leading to cell growth in soft agar and the formation of colonies. Additional treatment of these cells with β-estradiol to activate IRF-1 prevented soft agar growth. The data confirm the activity of IRF-1 as an efficient negative regulator of growth, cell cycle progression, and oncogenic transformation. Little effects of IRF-1 are detectable in the nontransformed state.

To identify genes that are involved in IRF-1–mediated effects, samples for expression profiling were taken before and 48 h after IRF-1 activation from myc/ras–expressing and nonexpressing cells. This late time point was chosen because short-term activation of IRF-1 does not revert the transformed phenotype. It is assumed that these profiles reflect direct as well as indirect effects of IRF-1; 1,347 genes were deregulated by oncogene expression >1.5-fold (Table S1). Surprisingly, 60% of these genes were reverted to the expression levels of nontransformed cells by activation of IRF-1 for 48 h (Fig. 1D, clusters 1 and 3). Among those different functional groups identified, one group contained cycle-regulating genes. Consistent with the cell cycle–inhibitory function of IRF-1 in transformed cells, nearly all cell cycle–regulating genes deregulated by transformation were found to be reverted by IRF-1 expression, suggesting that IRF-1 action influences the expression of a dominant factor. Because of the key role of cyclin D1 in cell cycle progression and neoplastic transformation, we examined its expression in relation to IRF-1 activity.

IRF-1 down-regulates cyclin D1 expression. Expression levels of proteins involved in the transition of the G₁ and S phase of the cell cycle were determined (Fig. 2A). Cyclin D1 was expressed at low levels in nontransformed cells. Expression of the oncogenes strikingly enhanced cyclin D1 levels. IRF-1 activation in the nontransformed state changed expression levels of cyclin D1 only marginally, but IRF-1 activation in myc/ras–expressing NIH3T3 cells resulted in a marked decrease in cyclin D1 protein (Fig. 2A). Down-modulation of cyclin D1 expression became apparent 24 to 48 h after IRF-1 activation and reached levels found in nontransformed cells. In addition, the expression of cyclin D3 is not influenced by IRF-1. To evaluate the specificity of cyclin D1 down-regulation by IRF-1, expression levels of cyclin E and CDK4 were determined because they also play an important role for G₁-S transition (Fig. 2B). Interestingly, Western blot analysis indicated that the cyclin E and CDK4 levels remained unaltered. In addition, the expression level of cdk2 is decreased 48 h after IRF-1 activation (data not shown). Because of the significant earlier down-regulation of cyclin D1 in comparison with cdk2, we conclude that cyclin D1 inhibition is responsible for IRF-1–mediated effects in transformed cells.

We determined whether down-regulation of cyclin D1 by IRF-1 is relevant for other transformed cell lines. Expression of IRF-1 in mouse liver carcinoma cell line BNL1ME led to proliferation inhibition, accumulation in the G₁ phase of the cell cycle, and inhibition of soft agar growth (Fig. S2B–D). In addition, expression of IRF-1 led to decreased cyclin D1 expression levels (Fig. S2A). These data confirm that, specifically, cyclin D1 is inhibited by IRF-1 in transformed cells.

IRF-1 inhibits promoter activity of the cyclin D1 gene. Down-regulation of cyclin D1 expression can occur at different levels (21). First, amounts of cyclin D1 mRNA in IRF-1–expressing cells were analyzed by quantitative PCR. Cells in the transformed state expressed 2-fold higher amounts of cyclin D1 mRNA compared with the same cells in the nontransformed state (Fig. 2C). No effects of IRF-1 activation on the cyclin D1 mRNA level were seen in the nontransformed cells (data not shown). In transformed cells, cyclin D1 mRNA levels were significantly reduced by IRF-1 in a time-dependent manner (i.e., the amount of cyclin D1 mRNA was similar to the level of nontransformed cells 48 h after IRF-1 activation). A reporter plasmid containing the cyclin D1 promoter upstream of the luciferase gene (−1745 CD1-Luc) was transiently transfected into myc/ras–transformed NIH 3T3 cells to test if IRF-1 activation down-regulates the activity of the cyclin D1 promoter. Cyclin D1 promoter activity was measured at different time points after transfection. Its activity was reduced to 40% 48 h after transfection (Fig. 2D,, left), whereas an ISRE promoter is activated by IRF-1 (Fig. 2D , right). Expression of a mutant IRF-1 that is defective in DNA binding (M6) did not inhibit cyclin D1 or activate the expression of the ISRE promoter. Hence, IRF-1 down-regulates cyclin D1 expression predominantly at the transcriptional level.

IRF-1 alters signaling pathways upstream of cyclin D1 induction. Down-regulation of cyclin D1 transcription occurs as late as 24 to 48 h after IRF-1 transfection, suggesting an indirect action of IRF-1. Accordingly, we did not find any IRF-E element in the sequence within a 3,000-bp region of the cyclin D1 promoter (data not shown). We therefore focused on signaling pathways promoting cyclin D1 expression that might be affected by the expression of IRF-1. Activation of the Ras/Raf/MAPK pathway is linked to the transcriptional induction of cyclin D1 (19, 22). Inhibition of cyclin D1 gene transcription could result from an inactivation of the MAPK pathway. In nontransformed cells, the activated forms of MEK and ERK were rarely present (Fig. 3A and B). Expression of the oncogenes c-myc and H-ras led to an activation of MEK and ERK. Activation of IRF-1 in myc/ras–transformed NIH3T3 for 48 h resulted in a marked decrease in the activated forms of MEK and ERK. To determine whether inhibition of MEK can lead to decreased cyclin D1 levels, the MEK inhibitor PD98059 was added to transformed cells. A reduction in cyclin D1 level similar to IRF-1–expressing cells was detected (Fig. S1). We further investigated whether the inhibition of MEK activation could inhibit proliferation, cell cycle progression, and transformation. Transformed cells in the presence of PD98059 showed inhibition of proliferation, an increase in G₁ population, and a reduced number of soft agar colonies compared with control cells. In contrast, cells treated with the p38 inhibitor SB202190 showed no differences to control cells. In addition, treatment of BNL1ME cells with the MEK inhibitor PD98059 led to down-regulation of cyclin D1, proliferation inhibition, G₁ phase accumulation, and inhibition of soft agar growth (Fig. S1A–D). Therefore, a reduction of MEK/ERK activation could be responsible for the decrease in D-type cyclin protein levels.

IRF-1 inhibits CDK4 activity and affects the phosphorylation status of pRb. Cyclin D1 is a regulatory subunit of the CDK4 holoenzyme that phosphorylates and thereby inactivates the tumor suppressor pRb. This is a critical event in G₁ phase to allow cells to pass through the cell cycle checkpoint and enter S phase. We examined whether down-regulation of cyclin D1 by IRF-1 was accompanied by inactivation of CDK4. D-type CDK activity was measured by an in vitro kinase assay. The activity of the immunoprecipitated CDK4 reached a strong inhibition after 24 and 48 h of IRF-1 activation (Fig. 3C). To confirm the effects of IRF-1 on CDK4 activity, we determined the phosphorylation status of pRb, one of its major targets. Mitogen-induced pRb phosphorylation was reduced 24 h and completely lost 48 h after IRF-1 activation (Fig. 3D). The decline of CDK4 activity and the inhibition of pRb phosphorylation showed a similar time course compared with IRF-1–mediated repression of cyclin D1. IRF-1 thus induces cell cycle arrest by down-regulating cyclin D1 expression that inhibits CDK4 activity and concomitant change in the level of hyperphosphorylated pRb.

Forced expression of cyclin D1 attenuates the effects of IRF-1 on cell cycle progression. The described results indicate that IRF-1 activity leads to the accumulation of cells in G₁ to repression of cyclin D1 transcription. Cyclin D1 overexpression should therefore override the growth-inhibiting effects of IRF-1. To test this hypothesis, stable cyclin D1 expression was established in the myc/rasNIH3T3IH cell line. Cyclin D1 was still strongly expressed after repression of oncogene expression (Fig. 4A). Interestingly, activation of IRF-1 led to a decrease of cyclin D1 protein expression, indicating that cyclin D1 expression was also regulated by protein stability. Nevertheless, cyclin D1 expression levels after IRF-1 activation was higher in nontransformed than in transformed cells. Compared with mock-transfected cells, the metabolic activity of cells constitutively expressing cyclin D1 cells was higher. Activation of IRF-1 in the nontransformed cells had only marginal effects on cell growth (Fig. 4B). IRF-1–mediated growth inhibition was partially but not completely overcome by overexpressing cyclin D1 in myc/ras–transformed cells: proliferation inhibition by IRF-1 was only 33% in cyclin D1–overexpressing transfectants compared with 64% in mock transfectants. Thus, down-regulation of cyclin D1 plays a critical role in IRF-1–mediated proliferation inhibition.

Forced expression of cyclin D1 on the cell cycle distribution was analyzed by FACS (Fig. 4C). Overexpression of cyclin D1 in nontransformed cells led to a cell cycle distribution typical for transformed cells. Activation of IRF-1 mediated a small increase of cells in the G₁ phase but not to the same extent as in nontransformed cells. In addition, IRF-1 failed to inhibit cell cycle distribution in c-myc/H-ras–transformed cells overexpressing cyclin D1. These results confirm that inhibition of cell cycle progression by IRF-1 is mainly mediated by down-regulation of cyclin D1. Soft agar assays with cyclin D1–overexpressing cells were done to investigate whether IRF-1 still can suppress the transformed phenotype. In agreement with earlier published data, overexpression of cyclin D1 in the nontransformed state led to the formation of soft agar colonies (Fig. 4D). However, fever cells were able to form colonies compared with myc/ras–expressing cells. Activation of IRF-1 decreased the number of colonies but was not able to revert this oncogenic phenotype completely. Although cyclin D1 overexpression enhanced colony formation in the transformed state, only a minor decrease was detected due to IRF-1 activation. Thus, cyclin D1 expression also counteracted this phenotype of IRF-1. The data confirm that down-regulation of cyclin D1 contributes to the antitumor activity of IRF-1.

Inhibition of endogenous cyclin D1 simulates IRF-1–like effects. To determine whether cyclin D1 expression is a key target of IRF-1, cyclin D1 protein levels were knocked down by RNAi. This was done by lentiviral transduction of short hairpin RNAs directed against cyclin D1 (shCycD1) or against unspecific (shGL4) target sequences. Lentiviral transduction of shGL4 had no effect on endogenous cyclin D1 expression. Expression of shCycD1 in transformed cells, however, drastically decreased the levels of cyclin D1 (Fig. 5A). Although in control cells activation of oncogene expression increased the metabolic activity 2.5-fold, knockdown of cyclin D1 mRNA strongly decreased the oncogenic effect to a 1.4-fold stimulation (Fig. 5B). Accordingly, the effect of IRF-1 activity in the cyclin D1 knockdown cells was significantly reduced. This was confirmed by cell cycle analysis data (Fig. 5C). These data confirm that cyclin D1 is an important mediator of myc/ras transformation and a major target molecule for IRF-1. Knockdown of cyclin D1 also inhibited soft agar growth in myc/ras–transformed cells (Fig. 5D). The number of colonies in these cultures was 10-fold lower than in mock-transfected cells. Concomitant activation of IRF-1 led to the complete elimination of colonies. Down-regulation of cyclin D1 in BNL1ME cells by shRNA led to proliferation inhibition, accumulation in the G₁ phase of the cell cycle, and inhibition of soft agar growth (Fig. S2B–D). The results from the knockdown experiments confirmed the role of cyclin D1 in some transformation and its reversion by IRF-1.

Cyclin D1 expression is essential for tumor formation. Tumor transplantation experiments in nude mice were done to determine whether the down-regulation of cyclin D1 was responsible for IRF-1–mediated inhibition of tumor formation. We have previously shown that IRF-1 expression prevents tumor growth of myc/ras–transformed cells in nude mice (6). Overexpression of cyclin D1 in tumor cells also led to tumor formation. The mice developed tumors with the same kinetics compared with mice injected with cells that were transformed by the expression of the oncogenes c-myc and H-ras (Fig. 6). Activation of IRF-1 in cyclin D1–overexpressing cells had no effect on their tumor growth, indicating that the loss of cyclin D1 down-regulation was essential for tumor growth. Down-regulation of cyclin D1 by shRNA expression prevented tumor growth in nude mice. In contrast, cells that expressed unspecific control shRNA grew to tumors in nude mice. These data indicate that cyclin D1 is a critical target for tumor formation, and that the tumor suppressor IRF-1 inhibits cyclin D1 expression, thereby preventing tumor formation.

Cancers are characterized by several genetic alterations that can contribute to the establishment of a transformed phenotype. Tumor cells acquire uncontrolled mitogen-independent progression through the cell cycle. D cyclins control cell cycle progression by binding and activating CDK4 and CDK6, which then initiate phosphorylation of pRb (reviewed by ref. 23). Cyclin E-CDK2 collaborates with cyclin D–dependent kinases to phosphorylate pRb. Phosphorylated pRb loses its transcriptional repressor function of E2F, which leads to transcriptional activation of E2F target genes, like cyclin E and cyclin A, that are required for S-phase entry.

Aberrant overexpression of D cyclins has been linked to loss of cell cycle control and a wide variety of malignancies. Genetic models in mice have further emphasized the importance of cyclin D1 in both development and tumorigenesis (24, 25). Several reports highlight the role of D type cyclins as critical downstream targets of other oncogenes. Cyclin D1 knockout mice are completely resistant to skin and mammary tumors induced by ras and ErbB-2 expression (26, 27).

Expression of IRF-1 in tumor cells results in a dramatic inhibition of their proliferation and tumorigenicity as shown earlier (6, 7). To define the molecular mechanism by which IRF-1 mediates its tumor-suppressive effects, gene expression profiling was done. Of the 1,347 genes that are differentially regulated by transformation, 60% were changed to the expression level of nontransformed cells by IRF-1. Among them, cyclin D1 was found. We showed that IRF-1 inhibits proliferation by negatively regulating the cell cycle at the level of G₁-S phase transition and thereby prevents tumor growth. IRF-1–induced cell cycle arrest is accompanied by a loss of cyclin D1 accumulation and can be attenuated by cyclin D1 overexpression. Loss of cyclin D1 by shRNA expression prevents the action of IRF-1. IRF-1 thus mediates its tumor-suppressive function by the down-regulating cyclin D1.

Down-regulation of cyclin D1 has clinical significance because many human neoplasias exhibit cyclin D1 deregulation. For example, in 50% of human breast cancer, overexpression of cyclin D1 is seen (28). Overexpression of cyclin D1 in transgenic mice results in increased proliferation, tissue hyperplasia, and tumorigenesis (29, 30). In contrast, decrease of cyclin D1 expression reduces tumorigenicity in nude mice (31). Our data stress the critical importance of cyclin D1 for tumor growth in vivo because down-regulation of cyclin D1 by shRNA prevents tumor growth. Thus, enhanced cyclin D1 expression is essential for myc/ras–induced transformation. IRF-1 mediated down-regulation of cyclin D1 prevents tumor growth in nude mice, showing that cyclin D1 is a target of IRF-1–mediated tumor suppression.

Overexpression of cyclin D1 overcomes the antitumor effect of IRF-1. Activation of IRF-1 in cyclin D1–overexpressing cells does not inhibit proliferation or transformation. Hence, IRF-1 inhibits transcription of the cyclin D1 gene but does not affect cyclin D1 function. This confirms down regulation of cyclin D1 transcription.

Cyclin D1 binds to CDK4/6 to inhibit the function of pRb by phosphorylation. Cyclin E/CDK2 collaborates with cyclin D–dependent kinases in repressing pRb. Activation of IRF-1 affects phosphorylation of pRb. Because the expression level of cyclin E and CDK4 is not affected by IRF-1, we conclude that the effect on pRb phosphorylation is mediated by the down-regulation of cyclin D1. The function of CDK4 must be inhibited by the low concentration of cyclin D1.

The IRF-1–induced accumulation of transformed cells in G₁ could also be mediated by high p21^Cip1 expression in tumor cells. In addition to regulating the kinase activity of CDK4 and CDK6, D cyclins have a number of activities that do not depend on the catalytic activity of their partner kinase. Cyclin D1-CDK4/6 complexes are able to activate cyclin E-CDK2 complexes by titration of the CDK inhibitors p21^Cip1 and p27^Kip1 (32). p21^Cip1 interacts with cyclin/CDK complexes (19) and may inhibit their activity (33). It has been reported that IRF-1, in cooperation with p53, induces p21^Cip1 leading to a G₁ arrest (34). IRF-1 binds to IRF-E elements in the human p21^Cip1 promoter (35). In breast cancer cells MDA-MB-468 and SK-BR-3, in which p53 is mutated, IRF-1 up-regulates p21^Cip1 (36). Furthermore, N-Ras–induced growth suppression in myeloid cells is mediated by IRF-1 (37). However, analysis of p21^Cip1 protein expression in our experimental model of c-myc/ras–transformed cells showed no alteration after IRF-1 activation (6). The transcription factor c-myc in association with Miz1 inhibits the expression of CDK inhibitors p21^Cip1 and p15^Ink4b (38). We therefore consider it unlikely that this cell cycle suppressor is involved in the growth-inhibitory effects of IRF-1 in the system described here.

Apart from its role as transcriptional activator, IRF-1 has also been implicated in repression of certain genes. For example, increased expression of IRF-1 leads to the repression of SLIPI. The effect seems to be mediated by ISRE-like sites in its promoter (39). IRF-1 also leads to the down-regulation of the Cdk2 promoter. However, this is mediated by down-regulation of SP1 protein levels (40). Recently, a repressor domain in the IRF-1 protein was identified that mediates repression of Cdk2 expression by direct action on its promoter (41). However, in our system, down-regulation of cdk2 occurs not until 48 h after IRF-1 activation (data not shown), which strongly argues for an indirect mechanism.

Accumulation of cyclin D1 is tightly regulated through multiple mechanisms, including promoter activation, mRNA stability, initiation of translation, and protein stability. Regulation at the level of mRNA accumulation can occur through destabilizing elements in its 3′ untranslated region. AU-rich elements on the distal region of the cyclin D1 mRNA are positively regulated by prostaglandin A₂ and negatively regulated by phosphatidylinositol 3-kinase (42, 43). Posttranslational control of cyclin D1 levels is mediated by phosphorylation-dependent polyubiquitination and degradation by the 26S proteosome (44). Here, we show by reverse transcription-PCR data that IRF-1 results in a decrease in cyclin D1 mRNA. Additionally, we found that IRF-1 decreases cyclin D1 promoter reporter activity (Fig. 2), indicating that IRF-1 plays a role in regulation cyclin D1 transcription. Interestingly, down-regulation of cyclin D1 transcription occurs as late as 24 h after the initiation of IRF-1 activation. Because estradiol activation of the constitutively expressed IRF-1 fusion protein occurs within minutes, the late reaction suggests that this inhibition is an indirect effect of IRF-1 on the cyclin D1 promoter, most probably mediated by thus far unknown IRF-1 target genes. In fact, we did not find any IRF-1–binding consensus sequence within a 3,000-bp region of the cyclin D1 promoter. It is known that ERK activity is critical for the transcriptional induction of the cyclin D1–encoding gene (45). The cyclin D1 promoter contains multiple regulatory elements, including activator protein-1, nuclear factor-κB, and others, that play a role in transcription of the gene (46, 47). Because activation of IRF-1 inhibits phosphorylation of MEK and ERK (Fig. 3), it is probable that down-regulation of the cyclin D1 expression involves the ERK signaling cascade. IRF-1 activation inhibits cell cycle progression preferentially in transformed cells with high cyclin D1 expression. Accordingly, only little effects of IRF-1 activity on cyclin D1 expression were detected in nontransformed cells. It is assumed that the low constitutive levels of cyclin D expression are independent of the activation of the MAPK pathway. This supports the view that IRF-1 is primarily acting on the MAPK pathway and through this on cyclin D1 expression.

Specific down-regulation of cyclin D1 is an interesting strategy for the therapy of tumors transformed by cells whose oncogenicity depends on cyclin D1. Down-regulation of cyclin D1 expression induces the inhibition of proliferation and the reversion of the transformed phenotype. Therefore, IRF-1 activation could be an attractive antitumor strategy. Our data indicate that IRF-1 can suppress tumor development only by down-regulation of cyclin D1. Earlier studies showed that IRF-1 can be a potent inducer of apoptosis (9, 34). Overexpression of IRF-1 prevents tumor growth of breast cancer cells in mice (12, 48). These findings strongly implicate IRF-1 as a tumor suppressor gene that acts independent of p53 to control apoptosis.

In addition to IRF-1–mediated effects on the growth and apoptosis of tumor cells, IRF-1 can also enhance immunogenicity of cells. Although the mechanism how IRF-1 mediates this effect is not fully understood, the up-regulation of MHC class I and II molecules is important. IRF-1 activation induces IFNs and cytokines that are involved in tumor-suppressive and immunomodulatory functions. Previously, we showed that IRF-1 induces significant therapeutic antitumoral immune responses and primes immunity against tumor-specific antigens (11). Therefore, IRF-1 is able to control tumor growth by two principle mechanisms, a direct antitumor growth effect mediated by down-regulation of cyclin D1 and an indirect one by enhancing recognition of the tumor cells.

Grant support: Deutsche Forschungsgemeinschaft grant SFB 566, Wilhelm-Sander-Stiftung grant 95.088.3, Deutsche Krebshilfe grant 10-2136, and Marie Curie Early-Stage Training of the European Community's Sixth Framework grant MEST-2004-504990 (J.E. Pulverer).

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 S. Kirchhoff for her critical discussion and M. Höxter for excellent technical support.