Interferon-γ-mediated Inhibition of Cyclin A Gene Transcription Is Independent of Individual cis-Acting Elements in the Cyclin A Promoter*

Interferons (IFNs) affect cellular functions by altering gene expression. The eukaryotic cell cycle is governed in part by the periodic transcription of cyclin genes, whose protein products associate with and positively regulate the cyclin-dependent kinases. To understand better the growth inhibitory effect of IFN-γ on vascular smooth muscle cells (VSMCs), we compared the expression and activity of G1 and S phase cyclins in control and IFN-γ-treated VSMCs. IFN-γ treatment did not inhibit the G1 cyclins but did decrease cyclin A protein, mRNA, and associated kinase activity by 85, 90, and 90%, respectively. Nuclear run-on and mRNA stability determinations indicated that this decrease was the result of transcriptional inhibition. To investigate the molecular basis of this inhibition, we examined protein-DNA interactions involving the cyclin A promoter. Electromobility shift assays showed little change with IFN-γ treatment in the binding of nuclear proteins to isolated ATF, NF-Y, and CDE elements. In vivo genomic footprinting indicated that IFN-γ treatment changed the occupancy of chromosomal NF-Y and CDE sites slightly and did not affect occupancy of the ATF site. In a previous study of transforming growth factor-β1-mediated inhibition of the cyclin A promoter, we mapped the inhibitory effect to the ATF site; in the present study of IFN-γ treatment, functional analysis by transient transfection showed that inhibition of the cyclin A promoter persisted despite mutation of the ATF, NF-Y, or CDE elements. We hypothesize that IFN-γ inhibits cyclin A transcription by modifying co-activators or general transcription factors within the complex that drives transcription of the cyclin A gene.

The pleiotropic cellular effects of the interferons (IFNs) 1 are thought to result from induction of gene expression. However, because IFNs increase expression of 50 -100 different proteins, many without known function (1), the changes in cellular activity mediated by IFNs, including suppression of growth, are at best partially understood. Normal cellular replication is governed in part by periodic transcription of the cyclin genes, whose protein products associate with and positively regulate the cyclin-dependent kinases (cdks). The cdks are negatively regulated by phosphorylation or by association with a number of distinct proteins known as cdk inhibitors (reviewed in Refs. [2][3][4][5]. Although growth inhibitory properties have been associated with several of the proteins induced by IFNs (6 -8), there is little information about how genes induced by IFN signaling, particularly those affected by IFN-␥, interact directly with the machinery that regulates progression of the cell cycle.
IFN-␣/␤ and IFN-␥ are structurally dissimilar and bind to distinct cell-surface receptors. The pathways the IFN receptors use to signal the nucleus to induce gene expression share some components, and the various IFN types regulate expression of partially overlapping sets of genes (9). Although induction of gene expression by IFN-␣/␤ is typically direct, requiring no new protein synthesis, that effected by IFN-␥, which targets a more diverse set of genes, can be direct or indirect (1). The precise subset of genes affected differs not only with the type of IFN but also with the type of cell (1).
In neoplastic hematopoietic cell lines, IFN-␣ treatment suppresses growth and is linked to decreased phosphorylation of retinoblastoma protein (10,11) and CDC2 (12); decreased expression of c-Myc (11), cyclin D3 (13), cyclin A (11,12), and CDC25A (13); and altered DNA binding by E2F (14). The independence of some of these effects may indicate the activation of parallel signaling pathways (11,13). In primary bone marrow-derived macrophages, IFN-␥ treatment inhibits phosphorylation of retinoblastoma protein (15) and expression of cyclin D1 coincident with its ability to inhibit DNA synthesis (16). In irreversibly arrested mammary epithelial cells, IFN-␥ may mediate inhibition of cdk2 and cyclin E-associated kinase activity through p27 kip1 , which is up-regulated by a post-translational mechanism (17).
Because overlap among targets of IFN-␥ and IFN-␣/␤ is only partial, and because the profile of genes affected by individual IFNs varies by cell type, extrapolations about mechanisms among IFNs and cell types are uncertain. For example, cell cycle arrest due to IFN-␣ treatment in hematopoietic cells corresponds to the G 0 (13) or G 0 -G 1 phase (11), whereas arrest due to IFN-␥ treatment corresponds to the mid-G 1 phase in mammary epithelial cells (18) and to the late G 1 or early S phase in primary macrophages (15). These differences in the point of cell cycle arrest indicate functional differences in the underlying growth inhibitory mechanisms.
IFN-␥ inhibits the growth of vascular smooth muscle cells (VSMCs). Because this effect occurs in vivo (19) as well as in vitro (20), IFN-␥ may be clinically useful in vascular pathologies associated with high levels of VSMC proliferation (19). The growth inhibitory effect of IFN-␥ on VSMCs has been linked with induction of 2Ј-5Ј-oligoadenylate synthetase (21), the first enzyme in the 2Ј-5Ј-adenylate IFN-regulated pathway of RNA degradation (22), and with suppression of c-myc mRNA and protein (23,24). Little is known, however, about the effect of these changes in gene expression on components of the cell cycle machinery in VSMCs.
We evaluated the effect of IFN-␥ on the expression of cdks and cdk inhibitors in VSMCs and the expression and activity of cyclins in VSMCs. IFN-␥ treatment induced both proliferative (CDK6) and antiproliferative (P15) genes but did not inhibit kinase activity associated with the G 1 cyclins. In contrast, cyclin A-associated kinase activity decreased markedly after IFN-␥ treatment, in proportion to decreases in the expression of cyclin A mRNA and protein. Although this inhibition of cyclin A expression appeared to be caused by a decrease in transcription, electromobility shift assays, in vivo genomic footprinting, and functional analysis by transient transfection indicated that the decrease did not depend on individual cisacting elements in the cyclin A promoter.

MATERIALS AND METHODS
Cell Culture-VSMCs were harvested from the aortas of male Sprague-Dawley rats (200 -250 g) by enzymatic dissociation according to the method of Gunther et al. (25). The growth medium for these cells was Dulbecco's modified Eagle's medium (Life Technologies, Inc.) supplemented with 10% fetal bovine serum (HyClone), penicillin (100 units/ml), streptomycin (100 g/ml), and 10 mM HEPES (pH 7.4) (Sigma); cells were made quiescent by culture for 72 h in medium containing 0.4% calf serum. Rat VSMCs were used for experimentation within 4 -6 passages from primary culture. Growth medium for human VSMCs (aortic; Clonetics) consisted of medium 199 (Life Technologies, Inc.) supplemented with 20% fetal bovine serum, penicillin (100 units/ ml), streptomycin (100 g/ml), and amphotericin B (250 ng/ml). These cells were used within 8 passages from primary culture. Recombinant rat IFN-␥ (Life Technologies, Inc.) and human IFN-␥ (Genzyme) were reconstituted in sterile water and stored in aliquots at Ϫ80°C until use. For some experiments, VSMCs were synchronized by incubation with medium containing 0.4% calf serum (HyClone) for 72 h before stimulation with growth medium (with or without IFN-␥). For in vivo footprinting experiments, human VSMCs were synchronized according to a two-step protocol as follows: cells in growth medium were first blocked in S phase for 20 h with hydroxyurea (1 mM) and then released and blocked in M phase for 20 h by adding new medium containing nocodazole (500 ng/ml). Synchronized monolayers were washed with phosphate-buffered saline, stimulated with fresh growth medium with or without IFN-␥, and harvested at defined intervals for DNA or RNA extraction.
Protein Immunoblot and Cyclin-associated Kinase Assays-Nuclear proteins were extracted as described (26), and protein concentration was determined by the Lowry method (DC protein assay, Bio-Rad). Total nuclear protein (30 g) from control and IFN-␥-treated cells was fractionated by electrophoresis through 8% SDS-polyacrylamide gels and transferred to a polyvinylidene difluoride membrane (Immobilon-P, Millipore). Nonspecific binding sites were blocked by incubation in 4% skim milk in Tris-buffered saline for 40 min, followed by incubation overnight at 4°C with primary antibodies. Anti-cyclin antibodies and working concentrations were rabbit anti-cyclin D1 (PharMingen 14726E), 1:5000; rabbit anti-cyclin E (Santa Cruz Biotechnology sc-481), 1:2500; and rabbit anti-human cyclin A (gift of R. Schlegel, Millennium Pharmaceuticals, Cambridge, MA), 1:2000. Immune complexes were visualized with the ECL detection system (Amersham Pharmacia Biotech). Cyclin-dependent kinase assays were performed as described (27), with modifications. In brief, total cellular protein (50 g) from control and IFN-␥-treated rat VSMCs was immunoprecipitated with antiserum (2.5 g) against cyclin A, D1, or E at 4°C for 2 h, followed by incubation at 4°C for 1 h with protein A-Sepharose beads (16 mg/ml; Amersham Pharmacia Biotech). The beads were incubated for 20 min at 30°C in 40 l of kinase buffer containing [␥-32 P]ATP (6,000 cpm/pmol; 100 mM), with glutathione S-transferase-retinoblastoma protein (Rb) (40 g/ml; Rb amino acids 769 -921, Santa Cruz Biotechnology) used as phosphorylation substrate for anti-cyclin D1 immunoprecipitates, and histone H1 (200 g/ml; Boehringer Mannheim) used as phosphorylation substrate for anti-cyclin A or E immunoprecipitates. Reaction products were resolved on 12% SDS-polyacrylamide gels and detected by autoradiography and phosphorimaging. Signals were quantified with the ImageQuant version 1.1 software analysis program (Molecular Dynamics).
RNA Half-life Determination-Quiescent rat VSMCs were stimulated with growth medium with or without IFN-␥ for 24 h before addition of actinomycin D (5 g/ml, Boehringer Mannheim). Cells were harvested for RNA extraction at intervals over the ensuing 8 h.
Nuclear Run-on Analysis-Quiescent rat VSMCs were stimulated for 24 h with growth medium with or without IFN-␥. The cells were lysed and nuclei were isolated as described (26). Nuclear suspension samples (200 l) were incubated with 0.5 mM each of CTP, ATP, and GTP and 250 Ci of 32 P-labeled UTP (3000 Ci/mmol, NEN Life Science Products) for 30 min. The samples were extracted with phenol/chloroform, precipitated, and resuspended in water for scintillation counting; equivalent amounts of radioactive probe were added to nitrocellulose filters bearing target cyclin A and control ␤-actin cDNAs. Filters were hybridized for 72 h at 40°C in the presence of formamide, washed, and subjected to autoradiography. Radioactivity hybridizing to cyclin A target cDNA was normalized to the activity of the corresponding ␤-actin control.
In Vivo DNA Footprinting by Ligation-mediated (LM) PCR-In vivo dimethyl sulfate (DMS) treatment was performed as described by Mueller and Wold (37) with some modification (38). In brief, for in vivo DNA methylation, intact human VSMC monolayers were exposed to 0.1% DMS at room temperature for 2.5 min. Cells were washed and lysed, and methylated DNA was recovered and deproteinized overnight at 37°C. For in vitro methylation, naked genomic DNA was treated with 1% DMS for 4 or 6 min at room temperature. Methylated DNA was cleaved by treatment with 1 M piperidine. Human VSMCs were treated in vivo with DNase I according to the method of Pfeifer and Riggs (39) with some modification. Cells in a monolayer were permeabilized with 0.05% lysolecithin (Sigma) and then exposed to DNase I (Boehringer Mannheim, 150 Kunitz units/ml) for 4 min at room temperature. Genomic DNA was treated in vitro with DNase I (0.03 Kunitz units/ml) for 8 min at room temperature.
LM-PCR was performed according to the method of Garrity and Wold as described by Patterson et al. (38). The 5Ј set of primers was derived from the published human cyclin A promoter sequence (40) and designed according to the criteria of Mueller et al. as described by Ausubel et al. (28). Primers 5Ј-1, 5Ј-2, and 5Ј-3, with their respective positions (40) and calculated T m s were as follows: 5Ј-CCCTAAATCCTAC- The 3Ј set of primers was described by Zwicker et al. (41). For first-strand synthesis, DNA was denatured at 95.5°C for 5 min, annealed at 55°C for 30 min, and extended at 76°C for 10 min. Ligation to asymmetric linkers was performed for 6 h at 17°C. PCR cycling conditions for the 5Ј primer set were 95.5°C for 1 min, 65°C for 2 min, and 76°C for 3 min, for a total of 21 cycles. PCR cycling conditions for the 3Ј primer set were 95.5°C for 1 min, 68°C for 2 min, and 76°C for 3 min, for a total of 21 cycles. Two cycles of primer extension were performed. Primers were annealed at 68.5°C with 32 P-end-labeled primer 5Ј-3 and at 72°C with 32 P-end-labeled primer 3 derived from Zwicker et al. (41). Primer extension products were separated on 6% denaturing polyacrylamide gels and visualized by autoradiography.
Transient Transfection Analysis of Promoter Activity-Reporter constructs were made by ligating wild-type (42) or mutated cyclin A promoter fragments into the promoterless pCAT3 basic vector (Promega). Deletion fragments and mutants were generated by PCR with Pfu polymerase (Stratagene) by using standard cycling conditions. Primers for the Ϫ83 to ϩ5 fragment were forward 5Ј-GAATGCACGTCAAGGC-CGCA-3Ј and reverse 5Ј-CCGCTGGAGCGCGGCTGTTC-3Ј. Site-directed mutagenesis of the reverse NF-Y site (ATTGG) was performed by PCR from a wild-type template with the forward primer 5Ј-GAGCGCTTTCcggttTCCATTTCAATAG-3Ј. The CDE-CHR sequence (CGCGGGATACTTGAA) was mutated by PCR with the primer 5Ј-TCCATTTCAATAGTatattGATcaggtAACTGCAAGAACAGCC-3Ј. (Lowercase denotes mutant bases.) The ATF mutant was described elsewhere (42). Mutations were confirmed by dideoxy-sequencing with Thermosequenase (Amersham Pharmacia Biotech). Rat VSMCs in 6-well plates were transfected with reporter plasmid (2 g/well) by the DEAE-dextran method (43). After 4 h, cells were treated with growth medium with vehicle (control) or rat IFN-␥ for 24 h until harvest. Chloramphenicol acetyltransferase activity (CAT) was assayed as described (44). To correct for variability in transfection efficiency, we co-transfected cells with 0.5 g/well of pCMV-␤-gal (CLONTECH) and determined ␤-galactosidase activity by solution assay. Transfections were performed three times, each time in triplicate. Corrected CAT activities (mean Ϯ S.E.) were normalized in relation to a parallel transfection with pCAT3 control (Promega), the activity of which was set as 100%.
Data Analysis-Comparisons among groups were made by factorial analysis of variance followed by a Bonferroni/Dunn post hoc analysis. Statistical significance was accepted for a p value Ͻ 0.05.

IFN-␥ Has Varied Effects on Expression of Cdks and Their
Inhibitors-After confirming the growth inhibitory effect of IFN-␥ treatment (19 -21) on VSMC DNA synthesis and cell replication, we determined how IFN-␥ affected the expression of genes involved in progression and arrest of the cell cycle. Northern analysis was performed with RNA derived from rat VSMCs synchronized in a quiescent state and then stimulated to enter the cell cycle by the addition of 10% fetal bovine serum with or without IFN-␥. A 4 -5.5-fold induction of Stat 1a mRNA in IFN-␥-treated cells (Fig. 1) confirmed their responsiveness to IFN-␥ (45). Although we postulated that this growth inhibition mediated by IFN-␥ might be reflected in a decrease in the expression of cdks or an increase in that of cdk inhibitors, we found little change in the level of CDK2, CDK4, p21, or p27 mRNA during the 48 h after stimulation. Note that CDK6 and p15, which contribute to and inhibit, respectively, activity associated with D-type cyclins (46), were both induced by approximately 2-fold in cells treated with IFN-␥ in comparison with time-matched controls.
IFN-␥ Inhibits Expression and Activity of Cyclin A but Not Those of Cyclin D1 or E-IFN-␥ treatment is associated with arrest of the cell cycle in the G 1 phase in fibroblasts (47,48) and mammary epithelial cells (18). In myeloid cells, growth arrest associated with IFN-␥ is linked to a decrease in expression of a principal G 1 cyclin, cyclin D1, even when the cytokine is added late in the G 1 phase (16). In light of these observations and the induction of CDK6 and p15 mRNAs in VSMCs that we observed ( Fig. 1), we next assessed the effect of IFN-␥ on G 1 and G 1 -S phase cyclin expression and activity.
Cyclin D1 expression in rat VSMCs was also induced by IFN-␥ at both the mRNA (data not shown) and protein levels, with immunoblot analysis showing a 3.8-fold increase in the amount of cyclin D1 ( Fig. 2A). IFN-␥ treatment had no significant effect on cyclin E protein levels, whereas it decreased cyclin A expression by 85% ( Fig. 2A); this decrease was present also at the mRNA level (data not shown).
We then assayed cyclin-associated kinase activity to determine how these changes affected net cell cycle kinase activity. Despite the IFN-␥-mediated induction of p15, we found that cyclin D1-associated kinase activity increased 1.7-fold and that cyclin E-associated kinase activity changed little. Cyclin A-associated kinase activity, on the other hand, was inhibited by 90% (Fig. 2B).
These results are consistent with an IFN-␥-mediated arrest of the VSMC cycle that involves late G 1 or G 1 -S phase regulatory mechanisms. Because cyclin A-associated kinase activity was inhibited in proportion to the decrease in cyclin A expression, and cyclins D1 and E were not inhibited, we investigated the means by which IFN-␥ decreased cyclin A expression.
IFN-␥ Inhibits Transcription of the Cyclin A Gene Without Affecting mRNA Stability-IFNs can inhibit gene expression by induction and activation of RNA degradation pathways (49) FIG. 1. Northern analysis of cdks and cdk inhibitors in control and IFN-␥-treated VSMCs. Rat VSMCs were incubated for 72 h under low serum conditions and then stimulated with growth medium with or without IFN-␥ (300 units/ml). Total cellular RNA was extracted at the indicated time points after stimulation. Samples (10 g) were resolved by electrophoresis through denaturing agarose gels and transferred to nitrocellulose filters, which were hybridized with 32 P-labeled cDNA probes as indicated. Filters were washed as described under "Materials and Methods" and analyzed by exposure to radiography film and phosphor screens. and by post-transcriptional mechanisms (50). To determine the underlying mechanism of the decrease in cyclin A mRNA mediated by IFN-␥, we directly assessed the mRNA half-life and transcriptional rate. The stability of cyclin A mRNA in the presence or absence of IFN-␥ was evaluated by inhibiting transcription in rat VSMCs with actinomycin D. The calculated cyclin A mRNA half-life of 2.7 h was not affected by IFN-␥ treatment (Fig. 3A). In nuclear run-on experiments, IFN-␥ did not change transcription of the control ␤-actin gene, but it inhibited radioactive signal hybridizing to cyclin A sequences by 67% (Fig. 3B). These studies indicate that the decrease in expression of cyclin A resulting from IFN-␥ treatment is due to inhibition of transcription.
Transcriptional Activity of the Proximal Cyclin A Promoter Is Inhibited by IFN-␥-Our finding that IFN-␥ treatment decreased cyclin A transcription in VSMCs suggested that we could map the inhibitory effect to cis-acting elements in the cyclin A promoter and so identify downstream signaling pathways affected by the cytokine. In growing rat VSMCs transiently transfected with cyclin A promoter constructs driving expression of a CAT reporter gene (Fig. 4), the activity of the Ϫ406/ϩ205 fragment was similar to that of the pCAT3 control plasmid. Deletion of 5Ј sequences to yield the fragment Ϫ266/ ϩ205 had relatively little effect on overall activity, despite the removal of a consensus AP1-binding site. The further removal of 5Ј sequences including three consensus Sp1 sites and 3Ј sequences including two consensus E2F sites, one consensus p53 site, and a potential composite Yi/Sp1 element (40), to yield the Ϫ83/ϩ5 fragment, decreased activity to approximately 30% of the activity of the reference plasmid. Inhibition of cyclin A promoter activity by IFN-␥ was retained in all constructs tested. The degree of inhibition by IFN-␥, ranging from 55 to 66%, was similar to that found in the nuclear run-on analysis (Fig. 3B). On the basis of these findings, we narrowed our search for an IFN-␥-responsive element to the Ϫ83/ϩ5 fragment; further deletions within this fragment severely impaired promoter activity (data not shown).
Protein Occupancy of Proximal Cyclin A Promoter Elements Shows Little Change with IFN-␥ Treatment-To determine whether IFN-␥ treatment changed protein binding to elements in the proximal cyclin A promoter, we performed in vivo foot-

FIG. 2. Effect of IFN-␥ on cyclin expression and activity in VSMCs.
A, rat VSMCs in growth medium were treated with vehicle (control) or IFN-␥ (300 units/ml). After 24 h, nuclear protein was extracted, fractionated through 8% SDS-polyacrylamide gels, and transferred to membranes for immunoblotting. Primary antibody hybridization was performed at 4°C overnight, and immune complexes were visualized by chemiluminescence autoradiography. Bands correspond to cyclin D1 (34 kDa), cyclin E (51 kDa), and cyclin A (58 kDa). B, cyclin-associated kinase activity was assayed in total cellular protein (50 g) from control and IFN-␥-treated rat VSMCs, with antibodies directed against cyclins D1, E, and A used for immunoprecipitation. Glutathione S-transferase-Rb, containing Rb amino acids 769-921, was used as kinase substrate for cyclin D1 immunoprecipitates, and histone H1 was used as kinase substrate for cyclin E and A immunoprecipitates. 32 P-Labeled reaction products were resolved on 12% SDS-polyacrylamide gels and detected by autoradiography and phosphorimaging.

FIG. 3. Effect of IFN-␥ on cyclin A mRNA stability and rate of transcription in rat VSMCs.
A, 24 h after quiescent rat VSMCs had been stimulated with growth medium with or without IFN-␥ (300 units/ ml), the cells were treated with actinomycin D (5 g/ml). Total RNA was harvested over the next 8 h and used in Northern analysis for cyclin A mRNA. Cyclin A signals were corrected for differences in loading against the activity of a radiolabeled oligonucleotide probe for the 18 S ribosome. Corrected values are plotted in log scale as a percentage of the 0-h value. B, the rate of cyclin A mRNA transcription was evaluated by nuclear run-on analysis. Subconfluent rat VSMCs were incubated in medium containing 0.4% calf serum for 72 h and stimulated by addition of growth medium with or without IFN-␥ (300 units/ml). Nuclei were isolated after 28 h of stimulation, and in vitro transcription was allowed to resume in the presence of [␣-32 P]UTP. Equivalent amounts of 32 Plabeled, in vitro transcribed RNA probes from control and treated cells were hybridized to 1 g of denatured cyclin A cDNA and ␤-actin cDNA on nitrocellulose filters. The filters were washed, and signal activities were assessed by autoradiography.

FIG. 4. Transient transfection analysis of cyclin A promoter activity in rat VSMCs.
Constructs bearing fragments of the human cyclin A promoter and driving expression of a CAT reporter gene were transiently transfected into rat VSMCs. Transfected cells were treated with growth medium with or without IFN-␥ (300 units/ml). After 24 h, cellular lysates were harvested, and CAT activity was determined in a solution assay. CAT activities were divided by corresponding (co-transfected) ␤-galactosidase activities to correct for variation in transfection efficiency and normalized against the activity of the pCAT3 control plasmid, which was given the arbitrary value of 100. Open bars, control; filled bars, IFN-␥. *, p Ͻ 0.05. printing with DNase I (Fig. 5), electromobility shift assays (Fig.  5), and in vivo footprinting with dimethyl sulfate (Fig. 6). In comparison with DNA treated in vitro, in vivo DNase I footprinting in control (growing) VSMC cultures showed protection of the native cyclin A promoter in three distinct regions between Ϫ83 and ϩ5, and at a fourth site near Ϫ150 (Fig. 5A, heavy bars). Each protected region encompassed an identified consensus binding site (ATF, NF-Y, CDE-CHR, and Sp1) (Fig.  5A, thin bars). The footprinted area containing the ATF site extended in the 5Ј direction more than 20 bases beyond the consensus ATF sequence itself. The protected regions were demarcated by hypersensitive sites (Fig. 5A, asterisks). The hypersensitive site at Ϫ60, between the ATF and NF-Y core sites, maps to the area between the NF-Y core and putative flanking regions of the cyclin A promoter described by Zwicker et al. (41,51). Although there were no obvious differences between the footprints derived from control or IFN-␥-treated VSMCs, our finding (Fig. 5A) confirmed that the ATF, NF-Y, and CDE-CHR consensus binding sites were actually occupied by protein in the native chromosomal context. We did not find additional protected elements between bases Ϫ83 and ϩ5.
To obtain an independent means of evaluating DNA-protein binding, we also performed electromobility shift assays with nuclear proteins extracted from control and IFN-␥-treated VSMCs. Consistent with the absence of change in protein binding to these sequences revealed by in vivo DNase I footprinting, treatment with IFN-␥ did not affect the intensity of specific binding of nuclear proteins to oligonucleotides bearing ATF, NF-Y, or CDE-CHR sites (Fig. 5B).
To look for IFN-␥-mediated changes in protein interactions with the proximal cyclin A promoter not detectable by DNase I footprinting or electromobility shift assay, we performed in vivo footprinting with DMS. Synchronized VSMCs were released into growth medium with or without IFN-␥, and DNA was harvested after 18, 24, and 30 h. These time points corresponded to early, mid, and late S phase on the basis of [ 3 H]thymidine incorporation studies (data not shown). To confirm the effect of IFN-␥ on cyclin A expression, we harvested RNA from cells cultured and treated with IFN-␥ in parallel. Northern analysis showed that in comparison with control samples, IFN-␥ treatment decreased cyclin A mRNA by 95% at 18 h after release, by 90% at 24 h, and by 69% at 30 h (Fig. 6A).
The results of DMS footprinting in vivo are shown in Fig. 6, B (coding strand) and C (noncoding strand). DNA ladders from both the coding and the noncoding strands indicated protein interactions with the ATF, NF-Y, and CDE sites, consistent with previous reports (41,52). The ATF site was protected in all in vivo samples and showed little change with IFN-␥ treatment. On the coding strand (Fig. 6B), guanines Ϫ76 and Ϫ79 were visible in samples treated in vitro but not in vivo. The same was true for noncoding strand guanines Ϫ74 and Ϫ77 (Fig. 6C). Protein binding to the ATF site of the proximal cyclin A promoter was also indicated by a hypersensitive coding strand guanine at Ϫ70 (Fig. 6B, diamond a), whereas on the noncoding strand an adenine at Ϫ75 was faintly visible in all in vivo samples (Fig. 6C, asterisk b). Like guanine, adenine can be methylated by DMS but is less susceptible to cleavage than is methylated guanine; its hypersensitivity may reflect torsional strain on the DNA (37).
In comparison with the in vitro samples, the CDE site was partially protected at all in vivo time points. However, there was some variability in the extent of protection as follows: on the coding strand, guanines Ϫ32 and Ϫ33 (Fig. 6B, diamond e) were more protected at time 0 and in the IFN-␥-treated samples and less protected in the control samples. Thus, maximal protection of the CDE site corresponded to a decrease in transcription of cyclin A, as reflected in the Northern analysis (Fig.  6A). The decrease in transcription is consistent with the model proposed by Zwicker et al. (41) in which occupancy of this site mediates repression of positive upstream elements. In the CHR site, guanine Ϫ24 (Fig. 6B, coding strand) showed a slight but consistent protection at all in vivo time points.
In the NF-Y core site (41) (Fig. 6B, NF-Y C ), guanines Ϫ51 and Ϫ52 on the coding strand were relatively protected throughout the time course, but protection appeared greatest in control samples. In the NF-Y flanking region (41) (Fig. 6B, NF-Y F ), there was some variability of the footprint corresponding to six guanines located between Ϫ71 and Ϫ61 (between the ATF and NF-Y sites), but there was no consistent pattern that correlated with IFN-␥ treatment. On the noncoding strand FIG. 5. Protein interactions with the proximal cyclin A promoter defined by in vivo DNase I footprinting and electromobility shift assay. Quiescent human VSMCs were stimulated with growth medium with vehicle or human IFN-␥ at 10 ng/ml for 24 h before harvesting. A, cellular monolayers were permeabilized with lysolecithin and treated with DNase I in vivo. Purified human VSMC DNA that had been treated in vitro was used as a control. Protection of the coding strand from the activity of DNase I was analyzed by LM-PCR with the 3Ј set of primers described under "Materials and Methods." The scale shows nucleotide positions within the human cyclin A promoter. Protected regions are indicated by heavy bars and consensus binding sites by thin bars. Asterisks correspond to hypersensitive residues. B, nuclear proteins were extracted from human VSMCs and used for electromobility shift analysis of protein binding to the ATF, NF-Y, and CDE-CHR elements in the proximal human cyclin A promoter. Competition assays with 50-fold excess of unlabeled probe were performed to verify specificity. Arrows indicate specific protein-DNA complexes. (Fig. 6C), guanine Ϫ56 was always protected in vivo (diamond c). A hypersensitive residue 5Ј to the core NF-Y site (corresponding to adenine Ϫ58) appeared in all in vivo samples; this residue was most prominent in the G 2 -M (time 0) sample (asterisk c). Signal from the guanine at Ϫ60 on the noncoding strand (diamond d) was prominent in the G 2 -M (time 0) sample but was largely protected in both the growing and the IFN-␥treated samples.

Inhibition of Cyclin A Promoter Activity by IFN-␥ Is Not Affected by Mutation of Protein-bound Elements-
The electromobility shift assays and in vivo footprinting with DNase I indicated little apparent change with IFN-␥ treatment in protein binding to cis-acting elements in the proximal cyclin A promoter. In vivo footprinting with DMS showed complex and relatively subtle changes in binding through a more extended time course. Whereas protection of the ATF site was relatively invariant, the changes in footprint involving the CDE and NF-Y elements suggested that they might be involved in mediating the inhibitory effect of IFN-␥. To determine the functional importance of these sites for promoter inhibition by IFN-␥, we performed further transient transfection experiments with mutated cyclin A promoter constructs. In tran-siently transfected growing cells, mutation of the ATF, NF-Y, and CDE-CHR sequences led in each case to a decrease in cyclin A promoter activity (Fig. 7) in comparison with the activity of the wild-type (parent) construct (Ϫ266/ϩ205, Fig. 4). The greatest reduction in activity occurred with the ATF site mutant, and the activity of the NF-Y mutant was closest to that of the wild type. Nevertheless, in transiently transfected cells treated with IFN-␥, inhibition of promoter activity was not affected significantly by any of these mutations; with all three constructs, activity fell by 55-61% with IFN-␥ treatment. DISCUSSION We have investigated the molecular basis of VSMC growth inhibition by IFN-␥. First we found that whereas most of the cdks and cdk inhibitors were not affected by exposure to IFN-␥, CDK6 and p15 were both induced at the mRNA level (Fig. 1). Despite these changes, however, cyclins D1 and E and the activities of their associated kinases were not inhibited (Fig. 2,  A and B), although cyclin A protein and cyclin A-associated kinase activity decreased substantially. These findings suggest that IFN-␥ treatment of VSMCs results in a relatively late G 1 or G 1 -S junction cell cycle block. The simplest explanation for these findings is that the decrease in cyclin A protein leads to the decrease in cyclin A-associated kinase activity; this observation suggested that regulation of cyclin A expression might be a target of IFN-␥-mediated signaling in VSMCs. Cyclin A mRNA stability and nuclear run-on studies indicated in turn that the decrease in cyclin A expression resulting from IFN-␥ treatment was due to inhibition of cyclin A gene transcription.
Cyclin A promotes cell cycle progression by associating with and stimulating cyclin-dependent kinases CDC2 and CDK2 and is a critical regulator of the cell cycle at both the G 1 -S junction (53,54) and during the G 2 -M transition (53). Like that of other cell cycle molecules, the level of cyclin A protein is controlled primarily by cyclical changes in mRNA expression. The human cyclin A promoter has been cloned (40) and characterized in a number of growth-related contexts such as contact inhibition (42), adhesion dependence (55), and TGF-␤1 treatment (56,57). Inhibition of cyclin A promoter activity by TGF-␤1 requires an intact ATF site, as we (57) and others (56) have reported previously; this effect involves decreased phosphorylation of CREB and ATF-1 (57). We hypothesized that it would be possible to identify the cis-acting elements in the cyclin A promoter required for its inhibition by IFN-␥. In turn, this identification would point to downstream targets of IFN-␥ signaling that are critical to regulation of cell cycle progression,

FIG. 6. In vivo DMS footprinting analysis of protein interactions with the proximal cyclin A promoter in human VSMCs, with and without inhibition by IFN-␥.
Human VSMCs blocked in G 2 were released into growth medium containing vehicle (control) or human IFN-␥ at 10 ng/ml and harvested at the indicated time points. A, Northern analysis was performed on RNAs from cells treated in parallel. Two cyclin A mRNA isoforms, 2.7 kilobases and 1.8 kilobases, are indicated. B and C, DNA in cellular monolayers was methylated with DMS in vivo for 2.5 min; purified DNA was methylated with DMS in vitro for 4 or 6 min (control). Methylated DNA was cleaved and used in LM-PCR. The coding strand (B) was analyzed with the 3Ј primer set, and the noncoding strand (C) was analyzed with the 5Ј primer set. In the vicinity of the ATF-binding site, diamond a indicates a hypersensitive guanine residue at Ϫ70, and asterisk b indicates a methylated adenine at Ϫ75 seen only in vivo. In the vicinity of the NF-Y site, diamond c and asterisk c indicate a protected guanine at Ϫ56 and a hypersensitive adenine at Ϫ58, respectively, and diamond d indicates a protected guanine at Ϫ60. In the region of the CDE, diamond e denotes guanines at Ϫ32 and Ϫ33 that show some variation in protection through the time course. FIG. 7. Transient transfection analysis of mutated cyclin A promoter activity in rat VSMCs. Cyclin A promoter fragments (Ϫ266/ϩ205) with mutations of proximal cis-acting elements ATF, NF-Y, and CDE-CHR were transiently transfected into rat VSMCs. Transfected cells were treated with growth medium supplemented with vehicle (control) or rat IFN-␥ (300 units/ml). After 24 h, cellular lysates were harvested, and CAT activity was determined by a solution assay, as described in the legend to Fig. 4. Open bars, control; filled bars, IFN-␥. *, p Ͻ 0.05. as our analogous identification had pointed to downstream targets of TGF-␤1 signaling.
Our initial transient transfections involved cyclin A promoter constructs driving expression of a luciferase reporter gene. With this system, however, we encountered a problematic nonspecific effect of IFN-␥ on luciferase activity that extended even to control promoters. Others have also reported difficulties with using luciferase as a reporter for analysis of IFN-␥mediated effects (58). The CAT system, in contrast, did not show nonspecific effects with IFN-␥, and we used it for all subsequent transfections.
Transient transfections with a CAT reporter showed that the inhibitory effect of IFN-␥ was retained in an 88-base pair fragment of the proximal promoter. This fragment contained consensus binding sites for ATF and NF-Y transcription factors, as well as the CDE and CHR elements (40,41,52). One model of regulation of cyclin A promoter activity during the cell cycle involves phase-specific protein binding to the CDE and CHR elements to periodically repress transcriptional activation mediated through the upstream positive regulatory elements NF-Y, ATF, and Sp1 (41,52). We anticipated that the inhibitory effect of IFN-␥ would be mediated through one of the elements between bases Ϫ83 and ϩ5.
However, complementary lines of evidence from in vivo DNase I footprinting (Fig. 5A) and electromobility shift assays (Fig. 5B) argued against a substantial change with IFN-␥ treatment in the level of protein binding to cis-acting elements in the Ϫ85/ϩ5 cyclin A fragment. Further investigation by in vivo DMS footprinting (Fig. 6, B and C) showed only limited changes in the protection pattern with IFN-␥ treatment, despite confirmation of inhibition of cyclin A mRNA expression by as much as 95% (Fig. 6A). We then evaluated the functional significance of protein binding to these sites by transiently transfecting mutated promoter constructs. These experiments indicated the functional importance of the ATF, NF-Y, and CDE-CHR binding sites in the overall activity of the cyclin A promoter but showed further that inhibition by IFN-␥ occurs regardless of mutations at these specific sites. The mechanism of down-regulation of cyclin A promoter activity by IFN-␥ is thus not a simple extension of the repressive mechanism that normally operates in G 1 (41,52) nor does it involve the ATF site in a manner analogous to down-regulation by TGF-␤1 (56,57).
Change in protein-DNA binding is by no means required for altered transcriptional activity. For example, in vivo footprinting indicated no significant difference in protein interaction with the essential serum response element and flanking c-fos promoter sequences after A431 cells had been stimulated with epidermal growth factor, a treatment that resulted in dramatic induction and then repression of c-fos mRNA as determined by Northern analysis (59). In the case of the cyclin A promoter, we previously found that TGF-␤1 treatment led to changes in transcription factor activity through modifications not reflected in protein abundance or DNA binding per se (57). These findings suggested that for transcription of certain genes, specific protein-DNA interactions are maintained and do not have to be disrupted to effect either activation or repression of transcriptional activity. In our analysis of TGF-␤1 signaling, however, we identified a functional role for the ATF site in the cyclin A promoter, because its mutation eliminated the inhibitory effect of TGF-␤1. With IFN-␥, we found that change in protein-DNA binding involving the proximal cyclin A promoter was quite limited and that mutation of the protein-bound DNA elements in the proximal promoter did not reduce the inhibitory effect of IFN-␥.
Indeed, although IFN-␥ inhibits transcription of many genes (60 -66), no IFN-␥-specific inhibitory elements have been iden-tified. One proposed mechanism for inhibition of gene expression by IFN-␥, derived from studies of the macrophage scavenger receptor gene promoter, involves competition between positive regulatory factors (AP-1, ETS) and a direct target of IFN-␥ signaling (activated Stat 1a). AP-1, ETS, and Stat 1a all depend on the essential co-activators CBP and p300 for full transactivation, so when cellular levels of these co-activators are limiting, the increase in activated Stat 1a due to IFN-␥ treatment limits AP-1-and ETS-mediated transactivation by titrating away the co-activators (67). Such a mechanism, however, requires DNA binding by individual positive regulators, and thus inhibition should be susceptible to mutation of these individual cis-acting elements; our transfections with site-directed mutations of these elements showed that cyclin A promoter activity was still fully inhibited by IFN-␥. Moreover, co-transfection of a p300 expression plasmid in transient transfection analysis did not relieve IFN-␥-mediated inhibition of cyclin A promoter activity (data not shown).
In contrast to our limited understanding of how IFN-␥ inhibits gene expression, functionally significant cis-acting elements mediating transactivation in response to IFN-␥ signaling have been precisely identified in a number of genes (68,69). Nevertheless, despite the directness of signaling through the JAK-Stat pathway, alternative mechanisms for gene induction by IFN-␥ that do not require changes in cis element binding do exist. Class II MHC genes are strongly induced by IFN-␥, but expression of the transcription factors that bind directly to class II MHC promoters is not affected by IFN-␥ (70). In vivo footprinting of class II MHC gene promoters showed that promoter occupancy in the class II MHC-negative mutant cell line RJ2.2.5 was no different from that in the class II MHC-positive Raji line, implicating a defect in a co-activator protein (71). This co-activator was subsequently identified as class II transactivator (CIITA) (72), an IFN-␥-inducible protein whose expression is strongly correlated with that of the class II MHC genes (73). CIITA has separable functional domains: the Nterminal domain provides transcriptional activation and the C-terminal domain confers promoter specificity (74). Mutation of any one of five distinct elements in the proximal promoter of the class II MHC DR␣ gene resulted in impaired CIITA-dependent transactivation (74), consistent with a model of CIITA as a second tier co-activator that does not bind DNA directly but activates transcription of specific promoters by a mechanism that involves multiple cis-acting elements.
IFN-␥ can induce gene expression by direct and indirect mechanisms and can also inhibit gene expression, presumably through indirect pathways. Despite the opposite natures of these effects, induction of class II MHC transcription and inhibition of cyclin A transcription resulting from IFN-␥ treatment both share a similar dissociation of promoter activity from individual protein-DNA interactions in the proximal promoter region. RNA polymerase II-mediated transcription involves a complex of factors that probably varies in composition from one promoter to the next (75). Evidence from yeast indicates that some cell cycle-regulated genes have unique requirements for particular components of the general transcription factor TFIID complex (76). Indeed, the complex associated with cyclin A expression is known to be functionally distinctive, as a specific mutation in TAF(II)250, the largest subunit of TFIID, disrupts transcription from the cyclin A but not the c-fos or c-myc promoter (77). This suggests that transcription complexes assembling on growth-related genes may be selectively modified in response to changes in cellular growth status. Our findings about IFN-␥ support the existence of a means of regulating transcriptional activity of the cyclin A promoter that does not involve direct DNA-protein interactions. More likely it involves alteration of co-activators or other components of the transcription complex that assembles on the cyclin A gene. Given the array of cellular targets affected by IFN-␥, and its known growth inhibitory effects on a number of distinct cell types, we speculate that its regulatory capacity may extend to components of the transcriptional apparatus that drives expression of the cyclin A gene.