Cyclin-dependent Kinase 9 Is Required for Tumor Necrosis Factor-α-stimulated Matrix Metalloproteinase-9 Expression in Human Lung Adenocarcinoma Cells*

The proinflammatory cytokine tumor necrosis factor-α (TNF-α) promotes tumor progression through activation of matrix metalloproteinase (MMP) activity. MMP-9 is a gelatinase secreted by both cancer cells and surrounding stromal cells, and it contributes to TNF-α-stimulated tumor invasion and metastasis. Cyclin-dependent kinase 9 (CDK9), the catalytic component of positive transcription elongation factor-b, phosphorylates serine 2 residues in the C-terminal domain of RNA polymerase II for productive transcription elongation and is up-regulated upon exposure to various stresses. This study investigated roles of CDK9 in TNF-α-stimulated MMP-9 expression in human lung adenocarcinoma cells. CDK9 activity was inhibited using three different strategies, including the CDK9 pharmacological inhibitor 5,6-dichloro-1-β-d-ribofuranosylbenzimidazole (DRB), a dominant-negative CDK9, and a CDK9-specific small interfering RNA. All three approaches reduced TNF-α-mediated accumulation of MMP-9 in the conditioned media as demonstrated by gelatin zymography. In contrast, transforming growth factor-β1-induced accumulation of MMP-2 was unaffected by DRB. Expression of the MMP-9 gene was examined using reverse transcription real time PCR and using a transient transfection assay to evaluate MMP-9 promoter activity. DRB reduced the TNF-α-induced increase in MMP-9 mRNA levels but did not effect transforming growth factor-β1-induced MMP-2 mRNA expression. Consistently DRB and dominant-negative CDK9 completely abrogated TNF-α-stimulated human MMP-9 promoter activity. TNF-α did not regulate expression or localization of CDK9 or its regulatory partner Cyclin T. However, TNF-α stimulated CDK9 binding to Cyclin T and MMP-9 gene occupancy by both CDK9 and the serine 2-phosphorylated form of RNA polymerase II. Our findings indicate that CDK9 mediates TNF-α-induced MMP-9 transcription. Disruption of TNF-α signaling using CDK9 inhibitors could serve as a potential therapeutic strategy against tumor invasion and metastasis.

Metastasis is one of the six hallmarks of cancer (1), and the persistent inflammatory environment surrounding cancer cells promotes invasion and metastasis (2,3). The proinflammatory cytokine TNF-␣ 1 is a key contributor to the inflammatory network that promotes carcinoma progression (4). Genetic and antibody-based ablation of TNF-␣ expression inhibits development of experimental skin tumors in mice (5,6). Moreover TNF-␣ promotes carcinoma progression, at least in part, through stimulating malignant epithelial cell migration, invasion, and subsequent metastasis (7,8). Matrix metalloproteinase-9 (MMP-9) is a TNF-␣-responsive extracellular matrixdegrading enzyme that promotes carcinoma cell migration and invasion by degrading basement membrane (9). In addition, MMP-9 also promotes tumor angiogenesis through remodeling of the vascular structure and release of matrix-trapped angiogenic factors, such as vascular endothelial growth factor (10,11). TNF-␣ stimulates MMP-9 expression in a nuclear factor-B-dependent manner (12)(13)(14)(15), contributing to TNF-␣-induced malignant cell invasion and migration (8, 16 -19). Tremendous effort has been invested in understanding intracellular signaling pathways that mediate activation of expression of TNF-␣responsive genes. It is known that the signaling molecule protein kinase C and the transcription factors NF-B and activator protein-1 are essential for induction of MMP-9 by TNF-␣ (12,15,18,20). However, the molecular mechanisms responsible for activation of the MMP-9 promoter by TNF-␣ upon NF-B binding remain largely unknown.
Positive transcription elongation factor-b (P-TEFb), which is comprised of CDK9 and its regulatory partner Cyclin T, converts RNA polymerase II from a transcription initiation-competent form into a transcription elongation-competent form via phosphorylation of serine 2 residues in the heptapeptide repeats of the C-terminal domain (21). Two isoforms of CDK9 have been identified with the apparent molecular mass of 55 (CDK9 55 ) and 42 kDa (CDK9 42 ), respectively (22). CDK9 55 expression is regulated by a second promoter located upstream of the transcription start point used to generate mRNAs encoding CDK9 42 . Both isoforms are able to bind to Cyclin T1 and phosphorylate the C-terminal domain (22). Despite a well characterized role of P-TEFb in regulation of the human immuno-deficiency virus, type 1, long terminal repeat promoter by Tat (23), our understanding of the roles of P-TEFb in regulating transcription of cellular genes and cell behavior is still limited. Expression and activity of P-TEFb appears to be regulated by diverse physiological and pathological cues in a variety of cell types. Proinflammatory cytokines, including TNF-␣, increase levels of CDK9 and Cyclin T upon activation of macrophages and T lymphocytes (24,25). The relative abundance of the CDK9 55 and CDK9 42 isoforms varies according to cell type and experimental condition (22). The catalytic activity of CDK9 can be stimulated by stress (26,27) and cardiac hypertrophic signals (28) through release of P-TEFb from its inhibitory partners, 7SK RNA and HEXIM1. P-TEFb is critical for activation of the heat shock response genes involved in (29,30) and expression of gene products required for antigen processing and presentation (31,32). In addition to regulation of P-TEFb function at levels of expression and catalytic activity, selective promoter loading of P-TEFb by sequence-specific DNA binding transcription factors provides temporal and spatial regulation of P-TEFb function. For example, P-TEFb can be recruited by NF-B (33) and c-Myc (34,35) to the promoters harboring the cognate response elements and thereby enhances transcription elongation from NF-B-and c-Myc-responsive genes. P-TEFb is also loaded onto heat shock-responsive promoters upon heat shock (29,36).
In this study, we investigated the role of CDK9 in TNF-␣induced MMP-9 expression using a human lung adenocarcinoma cell line, A549 cells. Our results indicated that CDK9 was essential for the induction of MMP-9 by TNF-␣. Inhibition of CDK9 activity diminished the TNF-␣-stimulated increase in MMP-9 transcript and enzymatic activity in the conditioned media. Furthermore CDK9 was required for regulation of the MMP-9 promoter by TNF-␣ as demonstrated by a reduction in the TNF-␣-induced MMP-9 promoter activity upon introduction of 5,6-dichloro-1-␤-D-ribofuranosylbenzimidazole (DRB) and dominant-negative CDK9. Moreover TNF-␣ up-regulated CDK9 binding to Cyclin T. Lastly TNF-␣ stimulated CDK9 binding and phosphorylation of serine 2 of RNA Pol II on the MMP-9 gene in A549 cells. Taken together, our findings indicate that CDK9 mediates the MMP-9 response to TNF-␣ in human lung adenocarcinoma cells via a transcriptional regulatory mechanism.

EXPERIMENTAL PROCEDURES
Reagents and Antibodies-The pharmacological inhibitor of CDK9, DRB, and phorbol 12-myristate 13-acetate (PMA) were purchased from Sigma. DRB and PMA were dissolved in Me 2 SO and stored at Ϫ20°C. The pNF-B-Luc reporter construct (NF-B.Luc) harboring five repeats of the NF-B enhancer element was purchased from Stratagene (La Jolla, CA). The pRL-TK vector expressing Renilla luciferase driven by the herpes simplex virus thymidine kinase promoter was obtained from Promega (Madison, WI). The MMP-9pGL-3 reporter construct (MMP-9.Luc) in which expression of firefly luciferase is regulated by the human MMP-9 promoter (37) was kindly provided by Dr. D. Boyd at M. D. Anderson Cancer Center, Houston, TX. CMV2-FLAG (pFlag) and CMV2-FLAG-CDK9dn (CDK9dn), which express the epitope FLAG or FLAG-tagged dominant-negative CDK9 fusion protein (38) were gifts from Dr. A. P. Rice at Baylor College of Medicine, Houston, TX. Recombinant human TNF-␣ and TGF-␤1 were purchased from Peprotech (Rocky Hill, NJ) and stored at Ϫ80°C. The Silencer Validated siRNA targeting human CDK9 (CDK9si) and the negative control siRNA (CTLsi) were purchased from Ambion (Austin, TX). A rabbit polyclonal antibody specific for CDK9 (C-20), a goat polyclonal antibody specific for Cyclin T1 (T-18), and a goat polyclonal antibody specific for actin (I-19) were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). The mouse monoclonal antibodies specific for RNA Pol II with serine 2 or serine 5 phosphorylated in the C-terminal domain (H5 and H14) were purchased from Covance (Berkeley, CA).
Cell Culture-The human lung adenocarcinoma A549 cell line was obtained from ATCC and was cultured in Dulbecco's modified Eagle's medium (Sigma) with 10% fetal bovine serum (v/v), 100 g/ml penicil-lin, and 100 g/ml streptomycin in a humidified incubator with 5% CO 2 . To serum-starve cells, A549 cells were kept in Dulbecco's modified Eagle's medium containing 0.5% fetal bovine serum for 48 h prior to further treatments.
Transient Transfection and Luciferase Reporter Assays-A549 cells were split and seeded into a 24-well plate at a density of 5 ϫ 10 4 cells/well. On the next day, 50 ng of each luciferase reporter (firefly and Renilla) construct was co-transfected into A549 cells along with the indicated amount of either pFlag or CDK9dn plasmids using Lipofectamine Plus according to the manufacturer's instructions (Invitrogen). The cells were serum-starved for 24 h after transfection and then exposed to TNF-␣ (10 ng/ml) for 6 h. The treated cells were harvested and assayed for luciferase activity using a Lumat LB 9507 luminometer (Berthold) along with the Dual-Luciferase reporter assay system (Promega) according to the supplier's instructions. Variation in transfection efficiency was monitored and corrected with co-transfected pRL-TK. The ratio of firefly luciferase activity to Renilla luciferase activity in each sample served as a measurement of normalized luciferase activity. The -fold change of normalized luciferase activity was obtained by setting the values from the untreated cells at one. To determine the effect of DRB on the response of the MMP-9 promoter to TNF-␣, transient transfections were performed essentially the same as above except that increasing doses of DRB were added simultaneously with TNF-␣ to the transfected cells in place of the CDK9dn.
siRNA Transfection-A549 cells were seeded into 6-well culture dishes at a density of 2.5 ϫ 10 5 cells/well. On the following day the cells were transfected with 50 pmol of either CDK9si or CTLsi using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. One group of cells was exposed to TNF-␣ (10 ng/ml) 36 h after transfection. The transfected cells were lysed in 2ϫ Laemmli buffer at 36 h postexposure to TNF-␣ and boiled for 10 min. Equal amounts of protein from each sample were immunoblotted for CDK9 as described below. Conditioned medium was also collected for gelatin zymography.
Fractionation and Immunoblots-Serum-starved A549 cells were exposed to TNF-␣ (10 ng/ml) for the indicated times. The treated cells were lysed in 2ϫ Laemmli buffer. To fractionate the cells, the stimulated cells were lysed as described for the chromatin immunoprecipitation except for the cross-linking step. The nuclei were centrifuged for 5 min at 600 ϫ g, resuspended in the sonication buffer, and designated as the nuclear fraction. The supernatant was transferred to a fresh Eppendorf tube and designated as the cytosolic fraction. The concentration of each protein sample was determined using the Bio-Rad DC protein assay kit. Equal amounts of total protein from each sample were loaded into precast Novex Tris-glycine gels (Invitrogen) and immunoblotted for CDK9, Cyclin T1, RNA Pol II with serine 2 phosphorylated, and actin as described previously (39).
Real Time Reverse Transcription PCR-Real time reverse transcription-PCR was performed to determine mRNA levels of MMP-2 and MMP-9 in the A549 cells subjected to the indicated treatments. Total cellular RNA was isolated using the RNeasy isolation kit (Qiagen, Valencia, CA) according to the manufacturer's instructions. Reverse transcription was carried out with 2 g of total RNA and oligo(dT) 20 in a reaction volume of 20 l using a cloned avian myeloblastosis virus first strand cDNA synthesis kit (Invitrogen) following the instructions provided. Real time PCR was carried out in 20 l of PCR mixture consisting of 10 l of 2ϫ iQ SYBR Green Supermix and 2 l of each cDNA sample on an iCycler iQ real time detection system (Bio-Rad) in duplicates. The PCRs were recorded in real time and analyzed using the accompanying program, iCycler iQ real time PCR detection system software Version 3.0A (Bio-Rad). The mRNA level of the 36B4 housekeeping gene was also determined by real time reverse transcription-PCR in each cDNA sample to normalize the expression of genes of interest (GOI). All the PCR primers used were designed using Beacon Designer 2.0 (Premier Biosoft International, Palo Alto, CA). The primers used were as follows: human 36B4 (GenBank TM accession number NM001002), forward primer (nucleotides 97-116), 5Ј-CGACCTG-GAAGTCCAACTAC-3Ј, and reverse primer (nucleotides 205-188) 5Ј-ATCTGCTGCATCTGCTTG-3Ј; human MMP-2 (GenBank TM accession number NM004530), forward primer (nucleotides 2219 -2236) 5Ј-AGTCTGAAGAGCGTGAAG-3Ј, and reverse primer (nucleotides 2410 -2392) 5Ј-CCAGGTAGGAGTGAGAATG-3Ј; human MMP-9 (GenBank TM accession number NM004994), forward primer (nucleotides 1156 -1173), 5Ј-TGACAGCGACAAGAAGTG-3Ј, and reverse primer (nucleotides 1298 -1280) 5Ј-CAGTGAAGCGGTACATAGG-3Ј. The PCR condition for each particular amplicon was optimized using a series of one to eight dilutions of cDNA samples to establish a linear relationship between threshold cycle (Ct) values and the input amount of cDNA (a correlation coefficient Ͼ95%) and to achieve PCR efficiency around 100% (Ϯ5%). A one-cycle difference in the Ct values among samples can be converted into a 2-fold difference in the amount of cDNA when these two parameters are around 100%. The optimal PCR condition for MMP-9 was 95°C for 20 s followed by 60°C for 30 s. The optimal PCR condition for MMP-2 was 95°C for 20 s followed by 58°C for 30 s. The optimal PCR condition for 36B4 was 95°C for 20 s followed by 58°C for 20 s. Melt curve analysis was performed at the end of each PCR to confirm the specificity of the PCR product. To compare mRNA levels of genes of interest, Ct values of GOI among samples were compared after correction for 36B4 expression. The ratio of GOI versus the corresponding 36B4 of each sample was determined based on the equation GOI/ 36B4 ϭ 2 Ct(36B4) Ϫ Ct(GOI) (40,41). The ratio of MMP/36B4 was compared among samples, and the -fold change of GOI expression was obtained by setting the values from the untreated cells to one.
Gelatin Zymography-Serum-starved A549 cells were exposed to 10 ng/ml TNF-␣ or TGF-␤1 with or without increasing doses of DRB. Conditioned medium was collected 48 h after the specified treatments and processed for gelatin zymography. The medium was concentrated using Microcon Ym10 concentrators (Millipore). Twenty g of concentrated protein was used for gelatin zymography using precast Novex gelatin zymogram gels (Invitrogen) as described previously (42). Purified human MMP-2 and MMP-9 (Chemicon, Temecula, CA) were used as migration standards for the assays. The intensity of MMP-9/-2 gelatinolytic zones on the gels was quantified using the AlphaImager 2200 documentation analysis system with AlphaEase Version 5.04 stand alone software (Alpha Innotech Corp., San Leandro, CA).
Immunoprecipitation-Cell extracts prepared from A549 cells subjected to the indicated experimental conditions were used in immuno-precipitations with antibodies specific for Cyclin T1 as described elsewhere (43). Briefly the TNF-␣-stimulated A549 cells were lysed in EBC buffer (50 mM Tris, pH 8.0, 120 mM NaCl, 0.5% Nonidet P-40, 5 mM dithiothreitol). One milligram of each lysate was precleared with bovine serum albumin/Protein G-agarose slurry (Invitrogen) followed by immunoprecipitation with 1 g of Cyclin T1-specific antibody. The amount of CDK9 in the immune complexes was determined by immunoblotting as described above.
Chromatin Immunoprecipitation and Real Time PCR-Chromatin immunoprecipitation was performed as described elsewhere with minor modifications (44,45). Briefly A549 cells were serum-starved for 48 h followed by exposure to TNF-␣ and/or DRB for the indicated times. The treated A549 cells were then cross-linked in 1% formaldehyde at room temperature for 10 min. Sheared chromatin prepared from roughly 10 7 FIG. 1. Selective inhibition of TNF-␣-stimulated MMP-9 activity in the conditioned media by the CDK9 inhibitor DRB. A, serum-starved A549 cells were stimulated for with 10 ng/ml TNF-␣ for the indicated times. The conditioned media were collected and concentrated. An equal amount of protein from each sample was used for gelatin zymography. B, the experiments were performed similarly to that described in A with addition of the indicated amount of DRB for 48 h. C, results of an experiment similar to B except that 10 ng/ml TGF-␤1 was used in place of TNF-␣. D, results of an experiment similar to B except that 1 ng/ml PMA was used in place of TNF-␣. "S" refers to the solvent Me 2 SO used to dissolve DRB.
FIG. 2. Inhibition of CDK9 blocked TNF-␣-stimulated MMP-9 activity in the conditioned media. A, A549 cells were transfected with either dominant-negative CDK9 expression vector (DN) or the backbone vector pF. The transfected cells were then serum-starved for 24 h followed by exposure to TNF-␣ for 48 h. Gelatin zymography was performed on the conditioned media collected from treated cells as described in Fig. 1. B, -fold change of MMP-9 activity in the conditioned media by setting the values from the unstimulated cells transfected with pF control vector as one. The results shown are the average from three separate experiments. *, p value Ͻ 0.05 by Student's t test. C, CDK9si or CTLsi was introduced into the A549 cells followed by serum starvation and exposure to TNF-␣ for 48 h. Total cell lysates were collected from A549 cells transfected with either CDK9si or CTLsi and processed for immunoblots with an antibody specific for CDK9. Two isoforms of CDK9 (CDK9 55 and CDK9 42 ) were detected with apparent molecular masses around 42 and 55 kDa, respectively. The blot was stripped and probed with a ␤-actin-specific antibody to assess loading variation. D, MMP-9 activity was determined using gelatin zymography in the conditioned media from the cells treated as described in C. cells was immunoprecipitated with 1 g of the antibody specific for CDK9, 5 l of the antibodies for RNA Pol II with serine 2 or serine 5 phosphorylated, or a negative control antibody. The immunoprecipitated DNA was then recovered using a standard ethanol precipitation procedure with the addition of 20 g of yeast tRNA as a carrier. The precipitated DNA was air-dried and dissolved in 40 l of nuclease-free water. Each sample of 2 l was used for real time PCR in duplicates as described above. The primers for the human MMP-9 gene (GenBank TM accession number NM004994) were: forward primer (nucleotides 38 -55), 5Ј-CTGGTCCTGGTGCTCCTG-3Ј; reverse primer (nucleotides 148 -130), 5Ј-CTGCCTGTCGGTGAGATTG-3Ј. The primers for the human 36B4 gene are as described above. The PCR primers were designed to amplify the coding regions with fewer than 25 nucleotides downstream of the ATG site. The optimal PCR condition allowing quantitative analysis was established as described above. The optimal PCR condition for MMP-9 (nucleotides 38 -148) is 95°C for 20 s followed by 60°C for 30 s. The input of MMP-9 and 36B4 genes for each immunoprecipitation was determined by real time PCR. The ratios of immunoprecipitated MMP-9 or 36B4 gene versus the corresponding input for each sample were determined based on Ct values as described above and compared across the samples to assess change in abundance of CDK9 or phosphorylated RNA Pol II-occupied MMP-9 and 36B4 genes. The -fold change was established by setting the values in the unstimulated cells to one. The amount of immunoprecipitated MMP-9 and 36B4 by a specific antibody was 2-4 times more than that immunoprecipitated with a control antibody.

Requirement of CDK9 for TNF-␣-stimulated Expression of MMP-9 Human
Lung Adenocarcinoma Cells-To investigate a potential role of CDK9 in TNF-␣-stimulated expression, we examined MMP-9 activity in the conditioned media from TNF-␣-stimulated cells in the presence or absence of DRB, a pharmacological inhibitor of CDK9. A549 cells were used as our experimental system due to its intact response to TNF-␣ (46). Gelatin zymography was utilized to evaluate MMP-9 activity in the conditioned media collected from the treated cells. Serumstarved A549 cells exhibited a gradual increase in MMP-9 activity over time in response to exposure to TNF-␣ (10 ng/ml) with an appreciable increase at 24 h and a peak at 48 h postexposure (Fig. 1A). In contrast, MMP-2, the other member of the type IV collagenase family, was not affected by TNF-␣ (Fig. 1A). DRB substantially reduced TNF-␣-stimulated MMP-9 activity at 48 h in a dose-dependent manner with a maximal 80% decrease at a concentration of 20 M (Fig. 1B, top  panel, lanes 2 and 4 -6). The solvent Me 2 SO alone had little effect on MMP-9 activity (Fig. 1B, top panel, lane 3 versus lane  2). MMP-2 was not altered by either TNF-␣ or DRB (Fig. 1B,  bottom panel). These findings suggested that the inhibitory effect of DRB on MMP-9 response to TNF-␣ was not due to global inhibition of gene expression. To further address the concern of a global inhibitory effect of DRB, we examined the effect of DRB upon induction of MMP-2 in response to TGF-␤1. In contrast to the inhibitory effect of DRB on TNF-␣ induction of MMP-9, MMP-2 activity in the conditioned media was induced and remained elevated in the presence of DRB at a dose capable of blocking the MMP-9 response to TNF-␣ (Fig. 1C). These findings indicate that DRB selectively inhibits the MMP-9 response to TNF-␣ (Fig. 1, compare B and C). Similar observations were made in A549 cells cultured in serum-free media exposed to the treatments as above (data not shown), indicating that the components in the residual serum did not contribute to the observations. The tumor promoter PMA activates MMP-9 expression through activation of NF-B as does TNF-␣ (18). We then investigated whether DRB would inhibit activation of MMP-9 by PMA. PMA at a concentration of 1 ng/ml stimulated MMP-9 activity in the conditioned medium (Fig. 1D, top and bottom panels, lane 3 versus lane 1). DRB reduced PMA-induced MMP-9 activity (Fig. 1D, top and bottom  panels, lane 4 versus lane 3). These results suggest that inhibition of CDK9 antagonized TNF-␣-induced accumulation of MMP-9 activity in the conditioned media, and the antagonism involved inhibition of NF-B signaling. To address the concern of potential off-target effects of DRB, we used two CDK9-specific targeting approaches, namely introduction of dominantnegative CDK9 (CDK9dn) and of CDK9-specific siRNA (CDK9si). Expression of dominant-negative CDK9 by transient transfection partially reduced TNF-␣-induced MMP-9 activity in the conditioned media by 50% (Fig. 2A). The limited inhibitory effect was likely due to a limit of transfection efficiency. The difference in MMP-9 activity between TNF-␣-stimulated cells transfected with either pFlag or CDK9dn was statistically significant as determined with results obtained from three separate transfections (Fig. 2B). CDK9si was used to further confirm an essential role of CDK9 in the MMP-9 response to TNF-␣. Immunoblots with an antibody specific for CDK9 were used to assess protein levels of CDK9 in the cells transfected with either CDK9si or a control siRNA (CTLsi). Two CDK9sisensitive bands were detected with apparent molecular weights corresponding to the two identified isoforms of CDK9, CDK9 55 and CDK9 42 , respectively (22). Transient transfection of CDK9si, but not the control siRNA, reduced protein levels of the CDK9 isoforms by as much as 60% as determined by densitometry regardless of the presence of TNF-␣ (Fig. 2C). The apparent greater sensitivity of the CDK9 55 isoform to CDK9si can be attributed to the low level of expression and limits of detection in this low range. Other explanations for selective targeting of the CDK9 55 isoform by the CDK9si, such as greater accessibility of the target sequence or reduced stability of the CDK9 55 isoform, cannot be excluded at this time. In the conditioned media collected from the same transfected cells, we observed a decrease in TNF-␣-induced MMP-9 activity in cells transfected with CDK9si when compared with that of cells transfected with CTLsi (Fig. 2D). These results indicate an essential role of CDK9 in TNF-␣-induced MMP-9 activity.
Since an alteration in MMP-9 activity in the conditioned media can result from regulatory mechanisms other than regulation of MMP-9 expression in the cell, we questioned whether DRB directly inhibited TNF-␣-stimulated MMP-9 expression. mRNA levels of MMP-9 in A549 cells subject to the various combinations of TNF-␣ and/or DRB were examined by quantitative real time reverse transcription-PCR. The -fold change of mRNA levels of MMP-9 in TNF-␣-stimulated cells was established as described under "Experimental Procedures." TNF-␣ induced a substantial increase in mRNA levels of MMP-9 as early as 4 h with a 16-fold increase over that of unstimulated cells (Fig. 3A). The mRNA levels of MMP-9 increased gradually over time, reaching a peak of a 79-fold increase over control 24 h postexposure to TNF-␣ (Fig. 3A). DRB at a concentration of 20 M substantially diminished the TNF-␣-induced increase in MMP-9 mRNA at every time point tested with the induction reduced from 79-to 3-fold (Fig. 3A). On the contrary, mRNA levels of MMP-2 were not affected by either TNF-␣ or DRB (Fig.  3C). Moreover the induction of MMP-9 expression by TNF-␣ was inhibited by DRB in a dose-dependent manner with a maximal 90% inhibition at 20 M, whereas Me 2 SO, the solvent for DRB, alone displayed little effect on the induction (Fig. 3B). Consistent with the results from gelatin zymography, DRB at a concentration capable of antagonizing induction of MMP-9 by TNF did not effect the TGF-␤1-induced increase in MMP-2 mRNA (Fig. 3D). These data indicate that inhibition of CDK9

FIG. 4. Requirement for CDK9 in TNF-␣-induced expression of a luciferase reporter gene regulated by the human MMP-9 promoter.
A, A549 cells were transfected with a firefly luciferase reporter construct in which the reporter expression was regulated by the human MMP-9 promoter (MMP-9.Luc). The transfected cells were stimulated with TNF-␣ (10 ng/ml) with or without DRB for 6 h. The reporter activity was determined using the Stop & Glo Dual-Luciferase kit. The -fold change of relative reporter activity was recorded after correction using co-transfection with Renilla luciferase. B, experiments were performed similarly to that described in A except that the CDK9 activity was inhibited by expression of dominant-negative CDK9. C, the transfections and luciferase reporter assays were conducted as described in A. Expression of the luciferase reporter (NF-B.Luc) was regulated by a minimal promoter containing five repeats of a NF-B response element. Data shown are the average from at least three independent transfection experiments performed in duplicate. *, p value Ͻ 0.05 by Student's t test. DMSO, Me 2 SO. selectively antagonizes TNF-␣-stimulated MMP-9 expression.
Requirement of CDK9 for TNF-␣-stimulated Transcription from the Human MMP-9 Promoter-TNF-␣ stimulates transcription from the MMP-9 promoter through the NF-B response elements (15,18). We postulated that inhibition of CDK9 activity diminished the MMP-9 promoter response to TNF-␣ through mechanisms involving NF-B because DRB inhibited the MMP-9 response to PMA, a classic NF-B activator. Human MMP-9 promoter activity was analyzed in TNF-␣stimulated cells using transient transfection assays. TNF-␣ stimulated expression of firefly luciferase regulated by the human MMP-9 promoter by more than 2-fold over that in the unstimulated cells (Fig. 4A). Increasing concentrations of DRB abrogated activation of the MMP-9 promoter by TNF-␣, while the solvent Me 2 SO alone posed little effect on the promoter activity (Fig. 4A). Similarly TNF-␣-induced MMP-9 promoter activity was reduced to basal levels by expressing dominantnegative CDK9 but not the control pFlag (Fig. 4B). The human MMP-9 promoter contains two TNF-␣-responsive elements, the distal NF-B binding site and the proximal activator protein-1 site (15,18). We examined the effect of DRB using a luciferase reporter construct, NF-B.Luc in which transcription of the luciferase reporter is regulated by five repeats of NF-B binding sites, to test whether DRB can block NF-B response elementdependent activation of the NF-B.Luc reporter. DRB blocked activation of the NF-B.Luc reporter by TNF-␣ in a dose-dependent manner (Fig. 4C). A similar inhibitory effect of DRB and dominant-negative CDK9 was observed on activation of the MMP-9 promoter by PMA (data not shown). These findings indicate that CDK9 is essential for regulation of the human MMP-9 promoter by TNF-␣ and implicate NF-B as the responsive element.
Up-regulation of CDK9 Activity in the MMP-9 Gene by TNF-␣ in Vivo-CDK9 expression and its catalytic activity are upregulated in purified resting CD4 ϩ T cells by a combination of cytokines including TNF-␣ (25). We questioned whether TNF-␣ regulated CDK9 expression and activity in A549 cells. The protein levels of CDK9 and Cyclin T1 were examined in TNF-␣-stimulated cells. Immunoblots for CDK9 and Cyclin T1 with total cell lysates prepared from TNF-␣-stimulated A549 cells revealed little change in protein levels of CDK9 and Cyclin T1 throughout the 48 h postexposure (Fig. 5A). Moreover the relative abundance of the CDK9 55 and CDK9 42 isoforms remained unchanged upon exposure to TNF-␣. Similarly the serine 2-phosphorylated form of RNA Pol II, an indicator of global CDK9 activity, was not appreciably altered by TNF-␣, although a decline was observed with time after serum starvation (Fig.  5A). To study subcellular localization of CDK9 and Cyclin T1, the A549 cells were separated into nuclear and cytoplasmic fractions after lysis as described for the chromatin immunoprecipitation assay. Immunoblots were performed on both cytoplasmic and nuclear fractions to determine the subcellular distribution of CDK9 and Cyclin T1. No significant changes of the two CDK9 isoforms or Cyclin T1 protein expression were detected in the nuclear fractions or cytosolic fractions (Fig. 5, B  and C). These data indicate that TNF-␣ does not globally regulate CDK9 or Cyclin T1 expression and subcellular distribution. The association of CDK9 with Cyclin T1 in TNF-␣-stimulated cells was examined by co-immunoprecipitation. An increase of Cyclin T1-bound CDK9 emerged as early as 1 h postexposure to TNF-␣ and remained through 24 h in the cell lysates prepared from TNF-␣-stimulated A549 cells (Fig. 5D).
FIG. 5. TNF-␣ stimulated CDK9 binding to Cyclin T1. A, serumstarved A549 cells were stimulated with TNF-␣ (10 ng/ml) for the indicated time. Total protein was isolated from the treated cells, and an equal amount of protein from each sample was used for immunoblots to determine protein levels of CDK9, Cyclin T1, and RNA Pol II with serine 2 phosphorylated. The blot was stripped and reprobed with an anti-actin antibody to ensure equal loading. B, the nuclear and cytosolic fractions were obtained from TNF-␣-stimulated cells at the indicated times as described under "Experimental Procedures" and processed for immunoblots using a CDK9-specific antibody as in A. C, similar to B except that an antibody specific for Cyclin T1 was used in immunoblots. D, serum-starved A549 cells were stimulated with TNF-␣ for the indicated times. An equal amount of each cell lysate was immunoprecipitated with a Cyclin T1-specific antibody as described under "Experimental Procedures." The immune complexes were then fractionated and immunoblotted with a CDK9-specific antibody. The results shown are representative of three separate experiments. Cy, cytosol; NE, nuclear extract; Ser2-P, phosphoserine 2.
Results shown here were representative of three separate immunoprecipitation experiments. The extra band between CDK9 55 and CDK9 42 in Fig. 5D, lane 3, was not reproducible and likely an artifact. CDK9 did not immunoprecipitate with a nonspecific goat IgG (data not shown). CDK9 isoforms responded similarly in this assay (Fig. 5D). These results demonstrate that TNF-␣ stimulates binding of both CDK9 isoforms to Cyclin T1.
We then addressed whether TNF-␣ promoted CDK9 recruitment to the MMP-9 gene. Chromatin immunoprecipitation assays were performed with an antibody specific for CDK9, allowing us to examine occupancy of the MMP-9 gene by CDK9 following exposure of the cells to TNF-␣. The amount of CDK9bound MMP-9 gene fragment (MMP-9-(38 -148)) was quantified using real time PCR. TNF-␣ (10 ng/ml) stimulated the association of CDK9 with the MMP-9 gene through 24 h postexposure with a maximal 9-fold increase in binding over that from the unstimulated cells at 4 h (Fig. 6A). The MMP-9-(38 -148) fragment immunoprecipitated by a control antibody was less than half of that observed in the unstimulated cells (data not shown). The CDK9-bound 36B4 gene fragment (36B4-(93-143)) was also examined by ChIP assays. The amount of CDK9associated 36B4 gene was 4-fold higher than that of the control antibody (data not shown). However, the CDK9 binding to the 36B4 gene was not altered by TNF-␣ throughout 24 h (Fig. 6A). These findings indicate that TNF-␣ selectively stimulates loading of CDK9 onto the MMP-9 gene. We then questioned whether an increase in CDK9 occupancy resulted in elevated occupancy of the serine 2-phosphorylated form of RNA Pol II in the MMP-9 gene. Immunoprecipitation with an antibody specific for RNA Pol II with serine 2 phosphorylated was conducted with the sheared chromatin prepared from A549 cells stimulated with TNF-␣ for 4 h in the presence or absence of DRB at a concentration of 20 M. Exposure to TNF-␣ resulted in a more than 5-fold increase in the MMP-9 gene occupied by serine 2-phosphorylated RNA Pol II over that in the unstimulated cells, and the increase was reduced by 60% in the presence of DRB (Fig. 6B). In contrast, the MMP-9 gene bound by RNA Pol II with serine 5 phosphorylated, which is a modification by CDK7 and insensitive to DRB (28), displayed little change after the various treatments (Fig. 6B). Occupancy of the housekeeping gene 36B4 by the two phosphorylated forms of RNA Pol II remained unchanged in the presence of TNF-␣ and/or DRB (Fig. 6C). These results suggest that TNF-␣ stimulates CDK9 binding to the human MMP-9 gene in chromatin, which in turn leads to an increase in phosphorylation of serine 2 in the RNA Pol II bound to the MMP-9 gene.

DISCUSSION
The proinflammatory milieu surrounding tumors contributes to cancer development (2,3). Induction of MMP-9 by TNF-␣ mediates the tumor-promoting effect of TNF-␣ (9). Previous studies have identified NF-B and activator protein-1 as signal transduction factors involved in the induction of MMP-9 by TNF-␣ (15,18). In this study, we further explored the molecular mechanisms involved in transcriptional regulation of MMP-9 by TNF-␣. Our results indicated that CDK9, the catalytic subunit of P-TEFb, is required for TNF-␣-stimulated expression of MMP-9 in human lung adenocarcinoma cells. Compromising CDK9 function using chemical or molecular in-FIG. 6. TNF-␣ stimulated CDK9 binding and activity in the human MMP-9 gene in vivo. Serum-starved A549 cells were stimulated with TNF-␣ for the indicated times. Sheared chromatin was prepared and processed for ChIP assays with antibodies specific for CDK9 or phosphorylated forms of RNA Pol II. A, ChIP assays were performed with a CDK9-specific antibody. Real time PCR was utilized to quantify the coding regions of the human MMP-9 and 34B4 genes in the immunoprecipitated CDK9-DNA complexes with primers specific for the two genes. The -fold change of each gene was obtained by setting the values for the unstimulated cells as one. The data shown are representative of two separate experiments. B, ChIP assays were performed with antibodies specific for RNA Pol II with serine 2 phosphorylated or with serine 5 phosphorylated with the sheared chromatin prepared from A549 cells exposed to TNF-␣ and/or DRB (20 M) for 4 h. Real time PCR was used to quantify the coding regions of the human MMP-9 gene in the immunoprecipitated complexes with primers specific for the gene. The -fold change was obtained by setting the value for the unstimulated cells as one. The data shown are representative of two separate experiments. C, same as in B except the association of modified RNA Pol II with the 36B4 gene was quantified in the immune complexes using real time PCR with 36B4-specific primers. The -fold change was obtained by setting the value for the unstimulated cells as one. The data shown are representative of two separate experiments. Ser2-P, phosphoserine 2; Ser5-P, phosphoserine 5. hibitors diminished TNF-stimulated MMP-9 expression and expression of a reporter gene from the human MMP-9 promoter. Although TNF-␣ did not directly regulate expression and subcellular distribution of CDK9 or Cyclin T, TNF-␣ stimulated CDK9 binding to Cyclin T and loading of CDK9 onto the MMP-9 promoter and consequent phosphorylation of serine 2 in RNA Pol II. These molecular events mediate activation of MMP-9 expression by TNF-␣.
The proinflammatory cytokine TNF-␣ regulates physiological and pathological processes, at least in part, through activating the NF-B pathway (15,18). Regulation of gelatinase MMP-2 and MMP-9 expression by various cytokines including TNF-␣ and TGF-␤1 plays critical roles in important physiological and pathological processes, such as cancer invasion (11). In this study, we demonstrated distinct responses of gelatinases to TNF-␣ and TGF-␤1 in human lung adenocarcinoma cells. TNF-␣ induced a substantial increase in only MMP-9 activity in the conditioned media, whereas TGF-␤1 stimulated a robust increase in MMP-2 activity with little effect on MMP-9 activity (Fig. 1). Analyses of mRNA levels of MMP-2 and MMP-9 in TNF-␣-or TGF-␤1-stimulated A549 cells confirmed that the increase resulted from up-regulation of gelatinase expression by these two cytokines (Fig. 3). Although MMP-2 and MMP-9 belong to the same family of MMP and are type IV collagenases, which share substrate selectivity, their expression is regulated differentially and differs in various biological contexts (47). The distinct responses of MMP-2 and MMP-9 to the proinflammatory cytokine TNF-␣ and the anti-inflammatory cytokine TGF-␤1 in the human lung epithelial A549 cells are consistent with the finding of other investigators using other cell types and are likely involved in cellular responses to the two opposing cytokines during inflammation and cancer development. Interestingly the distinct gelatinase responses also displayed differential sensitivity to inhibition of CDK9 function. Inhibition of CDK9 activity by pharmacological and genetic approaches selectively blocked TNF-␣-induced MMP-9 expression in the cells and the conditioned media (Figs. 1, 2,  and 3). In contrast, the MMP-2 expression stimulated by TGF-␤1 exhibited resistance to the CDK9 inhibitor ( Figs. 1 and  3). TNF-␣ and TGF-␤1 can regulate MMP-2 and MMP-9 expression through both transcriptional and post-transcriptional mechanisms (48,49). It is unknown what accounts for the resistance of TGF-␤1 induction of MMP-2 in response to inhibition of CDK9 activity. These findings suggest TNF-␣ and TGF-␤1, the two opposing cytokines in the regulation of inflammation, utilize distinct signaling pathways based on dependence of CDK9 to regulate MMP-2 and MMP-9 expression.
Previous studies on CDK9 function in various cell types under different experimental conditions suggest that CDK9 modulates diverse cellular behavior through regulating expression of a subset of genes (50). For instance, CDK9 effects cellular responses to apoptotic signals in cancer cells through its interaction with NF-B and c-Myc and consequent regulation of transcription from the genes responsive to these transcription factors (33,35). In addition, CDK9 is essential for NF-B-dependent transcriptional elongation and is recruited to the interleukin-8 gene by NF-B upon exposure to TNF-␣ (33). An NF-B response element has been identified in the human MMP-9 promoter and is essential for the MMP-9 response to TNF-␣ (12)(13)(14). Our data generated with transient transfection reporter assays revealed that CDK9 was required for activation of the human MMP-9 promoter (MMP-9.Luc) and the NF-B-responsive promoter (NF-B.Luc) by TNF-␣ (Fig. 4). These findings also suggest that CDK9 regulates TNF-␣-induced MMP-9 expression through mechanisms involving NF-B similar to that observed in the interleukin-8 response to TNF-␣ (33). The discrepancy between the large increase in MMP-9 mRNA levels (Fig. 3A, 79-fold higher than control) and the modest increase in the MMP-9 luciferase activity (Fig. 4A, 2-fold higher than control) induced by TNF-␣ are consistent with the function of P-TEFb. The human immunodeficiency virus, type 1, long terminal repeat promoter possesses RNA polymerase pause sites that must be overcome by efficient stimulation of transcriptional elongation by Tat-recruited P-TEFb (51)(52)(53). Similarly the disparate regulation of the endogenous MMP-9 gene and MMP-9 luciferase by TNF-␣ likely reflects differences in transcription elongation efficiency. It is well established that MMP-9 cleaves extracellular matrix and non-matrix substrates and thereby promotes cancer metastasis and angiogenesis (11). In addition to the MMP-9 activity secreted by malignant cells, tumor-neighboring stromal cells, such as tumor-associated fibroblasts, are also stimulated by the tumor environment to secrete MMP-9 and thus facilitate cancer cell invasion and metastasis (2,3). The requirement of CDK9 for TNF-␣-stimulated MMP-9 expression broadens the cellular behaviors modulated by CDK9 and provides a potential therapeutic target to block metastasis and angiogenesis during cancer development.
CDK9 activity is regulated globally under certain experimental conditions. CDK9 expression can be up-regulated in T lymphocytes by a combination of cytokines that include TNF-␣ (25). The catalytic activity of CDK9 is elevated upon release from the inhibitors 7SK RNA and HEXIM1 in stressed cells (26,27) and in cardiocytes exposed to hypertrophic signals (28). We did not observe global alteration of either CDK9 or Cyclin T expression or serine 2 phosphorylation in A549 cells exposed to TNF-␣ (Fig. 5, A-C). However, TNF-␣ stimulated CDK9 binding to Cyclin T, which likely leads to an increase in a potentially active form of CDK9 (Fig. 5D). The two CDK9 isoforms appear to play similar roles in the response to TNF-␣ since they are regulated by TNF-␣ in a similar manner (Fig. 5, A-C). The discrepancy between increased Cyclin T-bound CDK9 and constant global serine 2 phosphorylation in the C-terminal domain of RNA Pol II can be explained by our results generated by ChIP assays (Fig. 6) and other previous studies that focus on CDK9 function in regulation of particular genes. One possibility is that the increased CDK9-Cyclin T1 complex is recruited to a subset of TNF-␣-responsive genes and results in focal phosphorylation of serine 2 within those genes instead of a global increase in phosphorylation of serine 2. CDK9 has been shown to be regulated within the loci of certain genes through recruitment by DNA sequence-specific transcription factors, such as NF-B and c-Myc, resulting in hyperphosphorylation of serine 2 in RNA Pol II bound to those genes (33,35). Our data generated with ChIP assays support this model. TNF-␣ increased the abundance of CDK9 and RNA Pol II (phosphoserine 2 form) occupation of MMP-9 gene by 8-and 5-fold over those of control, whereas the occupancy of the housekeeping gene 36B4 by these two proteins exhibited little change (Fig. 6). The recruitment of CDK9 to the MMP-9 gene is very likely accomplished through NF-B since the MMP-9 promoter contains NF-B response elements, and NF-B recruits CDK9 to another NF-B-responsive gene, interleukin-8 (33). In addition, only the phosphorylation of serine 2 on the MMP-9 gene, but not on the 36B4 gene, is sensitive to DRB (Fig. 6B), suggesting differential regulation of CDK9 activity bound to housekeeping genes and inducible genes. Our results indicate that TNF-␣stimulated CDK9 activity is an essential pathway for activation of MMP-9 expression. Interestingly stress-dependent recruitment of CDK9 to the p21/WAF1 gene appears to be a regulated step in activation of p21/WAF1 expression during the p53-dependent stress response (54). Our findings strengthen the importance of CDK9 activity to the expression of regulated genes in response to different biological cues.
In this study, we demonstrated that CDK9 is required for TNF-␣-stimulated MMP-9 expression in human lung adenocarcinoma cells. CDK9 mediates the MMP-9 response to TNF-␣ via a transcriptional regulatory mechanism that involves increased CDK9 binding and consequent transcriptional elongation along the MMP-9 gene. Blocking CDK9 activity provides an appealing therapeutic approach to curb cancer metastasis.