Critical Role of cAMP Response Element Binding Protein Expression in Hypoxia-elicited Induction of Epithelial Tumor Necrosis Factor-α*

Tissue hypoxia is intimately associated with a number of chronic inflammatory conditions of the intestine. In this study, we investigated the impact of hypoxia on the expression of a panel of inflammatory mediators by intestinal epithelia. Initial experiments revealed that epithelial (T84 cell) exposure to ambient hypoxia evoked a time-dependent induction of the proinflammatory markers tumor necrosis factor-α (TNF-α), interleukin-8 (IL-8), and major histocompatibility complex (MHC) class II (37 ± 6.1-, 7 ± 0.8-, and 9 ± 0.9-fold increase over normoxia, respectively, each p < 0.01). Since the gene regulatory elements for each of these molecules contains an NF-κB binding domain, we investigated the influence of hypoxia on NF-κB activation. Cellular hypoxia induced a time-dependent increase in nuclear p65, suggesting a dominant role for NF-κB in hypoxia-elicited induction of proinflammatory gene products. Further work, however, revealed that hypoxia does not influence epithelial intercellular adhesion molecule 1 (ICAM-1) or MHC class I, the promoters of which also contain NF-κB binding domains, suggesting differential responses to hypoxia. Importantly, the genes for TNF-α, IL-8, and MHC class II, but not ICAM-1 or MHC class I, contain cyclic AMP response element (CRE) consensus motifs. Thus, we examined the role of cAMP in the hypoxia-elicited phenotype. Hypoxia diminished CRE binding protein (CREB) expression. In parallel, T84 cell cAMP was diminished by hypoxia (83 ± 13.2% decrease, p < 0.001), and pharmacologic inhibition of protein kinase A induced TNF-α and protein release (9 ± 3.9-fold increase). Addback of cAMP resulted in reversal of hypoxia-elicited TNF-α release (86 ± 3.2% inhibition with 3 mm 8-bromo-cAMP). Furthermore, overexpression of CREB but not mutated CREB by retroviral-mediated gene transfer reversed hypoxia-elicited induction of TNF-α defining a causal relationship between hypoxia-elicited CREB reduction and TNF-α induction. Such data indicate a prominent role for CREB in the hypoxia-elicited epithelial phenotype and implicate intracellular cAMP as an important second messenger in differential induction of proinflammatory mediators.

Tissue hypoxia is intimately associated with a number of chronic inflammatory conditions of the intestine. In this study, we investigated the impact of hypoxia on the expression of a panel of inflammatory mediators by intestinal epithelia. Initial experiments revealed that epithelial (T84 cell) exposure to ambient hypoxia evoked a time-dependent induction of the proinflammatory markers tumor necrosis factor-␣ (TNF-␣), interleukin-8 (IL-8), and major histocompatibility complex (MHC) class II (37 ؎ 6.1-, 7 ؎ 0.8-, and 9 ؎ 0.9-fold increase over normoxia, respectively, each p < 0.01). Since the gene regulatory elements for each of these molecules contains an NF-B binding domain, we investigated the influence of hypoxia on NF-B activation. Cellular hypoxia induced a time-dependent increase in nuclear p65, suggesting a dominant role for NF-B in hypoxia-elicited induction of proinflammatory gene products. Further work, however, revealed that hypoxia does not influence epithelial intercellular adhesion molecule 1 (ICAM-1) or MHC class I, the promoters of which also contain NF-B binding domains, suggesting differential responses to hypoxia. Importantly, the genes for TNF-␣, IL-8, and MHC class II, but not ICAM-1 or MHC class I, contain cyclic AMP response element (CRE) consensus motifs. Thus, we examined the role of cAMP in the hypoxia-elicited phenotype. Hypoxia diminished CRE binding protein (CREB) expression. In parallel, T84 cell cAMP was diminished by hypoxia (83 ؎ 13.2% decrease, p < 0.001), and pharmacologic inhibition of protein kinase A induced TNF-␣ and protein release (9 ؎ 3.9-fold increase). Addback of cAMP resulted in reversal of hypoxia-elicited TNF-␣ release (86 ؎ 3.2% inhibition with 3 mM 8-bromo-cAMP). Furthermore, overexpression of CREB but not mutated CREB by retroviral-mediated gene transfer reversed hypoxia-elicited induction of TNF-␣ defining a causal relationship between hypoxiaelicited CREB reduction and TNF-␣ induction. Such data indicate a prominent role for CREB in the hypoxiaelicited epithelial phenotype and implicate intracellular cAMP as an important second messenger in differential induction of proinflammatory mediators.
The human intestine is lined with a single layer of protective epithelial cells that possess properties such as barrier and ion (and subsequent fluid) transport functions (1)(2)(3). Such epithelial functions are tightly controlled by an array of immunederived factors within the intestinal microenvironment, and regulation of such conditions can vary greatly during active episodes of inflammation. A number of previous studies have revealed specificity with regard to regulation of epithelial end point function, including tight junction permeability, electrogenic chloride secretion, and neutrophil transmigration (4 -8). This work utilizing intestinal epithelial cells in vitro has revealed that ligation of receptors for IFN-␥, 1 interleukin-4, or interleukin-13 results in increased macromolecular permeability (4,(7)(8)(9). On the whole, other cytokine responses (i.e. regulation of Cl Ϫ secretion, fluid transport, MHC, and MHC-like molecule expression, polymorphonuclear leukocyte transmigration, etc.) appear to reveal a unique "functional fingerprint" with respect to individual cytokines. Others have demonstrated the presence of functional receptors for IL-2 (10, 11), transforming growth factor-␤ (12), and hepatocyte growth factor (13) on intestinal epithelial cells. Based on this previous work, we have proposed that epithelial cells have the unique ability to "phenotype-switch," whereby epithelia lose classic qualities of epithelia (i.e. barrier function, ion transport properties, etc.) and assume features resembling immune-type cells (i.e. surface expression of MHC class I and II, antigen presentation properties, regulated polymorphonuclear leukocyte trafficking, etc.) (5).
Tissue ischemia/hypoxia often occurs concomitantly with other inflammatory processes, and thus, appropriate models to study mechanisms of inflammation should also account for conditions of cellular hypoxia. Surprisingly little is known about the mechanisms of hypoxic signal transduction. Several reports indicate that hypoxia down-regulates activity of cellular cAMP-generating machinery and that such diminished cAMP signaling can influence gene expression and end point cellular functions (14 -18). At the transcriptional level, a primary target for cAMP-mediated signaling is cAMP response element binding proteins, a family of 43-kDa leucine zipper transcription factors that share certain structural motifs, bind DNA as dimers, and regulate transcription of target genes (19). At present, it is not known whether hypoxia-mediated gene * This work was supported by National Institutes of Health Research Grants DK50189 and HL60569 (to S. P. C.) and DK09699 (to C. T. T.) and by a grant from the Crohn's and Colitis Foundation of America. 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  expression is directly related to CREB activity or CREB expression.
Recently, we have studied the role of epithelia in intestinal ischemia using a model epithelium, T84 cells. In response to conditions of ambient hypoxia, epithelia express a more classical immune-like phenotype, including release of proinflammatory cytokines and chemokines (20,21), regulated neutrophil transmigration (20), diminished ion transport (14), and induction of major histocompatibility complex class II expression (21). The basis for these changes in epithelial function are not well understood, and thus, in the studies defined here, we hypothesized a basic mechanism of hypoxia-induced phenotype switch. These studies revealed that hypoxia specifically induces a panel of epithelial proteins that bear cyclic AMP response elements (CRE) within the region important for regulation of gene expression. Further mechanistic insight was defined with one gene product, epithelial-derived TNF-␣, and revealed a critical role for CRE and CRE-binding proteins (CREB) in induction of TNF-␣. Such results define a primary role for CRE in the hypoxia-elicited epithelial phenotype and implicate CREB down-regulation as a mechanism of differential induction of proinflammatory mediators by cellular hypoxia.

Growth and Maintenance of T84 Intestinal Epithelial Cells-The T84
cell line is a human colonic carcinoma cell line (22) which, when plated on permeable membrane supports, forms polarized monolayers of columnar intestine-like epithelial cells. T84 cells are functionally well differentiated with regard to electrogenic Cl Ϫ secretion and ion transport and serve as excellent models of crypt columnar epithelial cells (22,23). T84 cells were grown as monolayers in a 1:1 mixture of Dulbecco/ Vogt modified Eagle's medium and Ham's F-12 medium, supplemented with 15 mM HEPES buffer (pH 7.5), 14 mM NaHCO 3 , 40 mg/ml penicillin, 8 mg/ml ampicillin, 90 mg/ml streptomycin, and 5% newborn calf serum. Monolayers were subcultured from flasks every 7-14 days by brief trypsin treatment (0.1% trypsin and 0.9 mM EDTA in Ca 2ϩ -and Mg 2ϩ -free phosphate-buffered saline).
Epithelial Exposure to Hypoxia-Epithelial exposure to hypoxia was performed as described previously (20). T84 cell growth media (1:1 Ham's F-12/Dulbecco's modified Eagle's medium) were replaced with fresh, pre-equilibrated hypoxic media, and cells were placed in a humidified environment within the hypoxia chamber (Coy Laboratory Products, Ann Arbor, MI) and maintained at 37°C. Standard hypoxic conditions, based on previous work (20), were pO 2 20 torr, pCO 2 35 torr, with the balance made up of nitrogen and water vapor. Normoxic controls were cells exposed to the same experimental protocols under conditions of atmospheric oxygen concentrations (pO 2 147 torr and pCO 2 35 torr within a tissue culture incubator).
Enzyme-linked Immunosorbent Assays (ELISAs)-Epithelial expression of indicated surface proteins was quantified using a cell-surface ELISA, as described before (5). Epithelial cells were grown and assayed for antibody binding following exposure to normoxia or hypoxia for the indicated periods. Cells were washed with ice-cold HBSS (Sigma), blocked with media for 30 min at 4°C. Anti-ICAM-1 monoclonal antibody (clone P2A4 (24) obtained from the Developmental Studies Hybridoma Bank, Iowa City, IA, used as undiluted cell culture supernatant), anti-MHC class I (25) (clone W6/32 obtained from the American Type Culture Collection, used as 1:100 diluted ascitic fluid), or anti-MHC class II (clone L243(5) obtained from the American Type Culture Collection, used as undiluted cell culture supernatant) were used to examine protein surface expression by ELISA. After washing with HBSS, a peroxidase-conjugated sheep anti-mouse secondary antibody (Cappel, West Chester, PA) was added. Secondary antibody (10 g/ml) was diluted in media containing 10% fetal bovine serum. After washing, plates were developed by addition of peroxidase substrate (2,2Ј-azinobis(3-ethylbenzthiazoline-6-sulfonic acid), 1 mM final concentration, Sigma) and read on a microtiter plate spectrophotometer at 405 nm (Molecular Devices). Controls consisted of media only and secondary antibody only. Data are presented as the mean Ϯ S.E. optical density at 405 nm (secondary antibody background subtracted). IFN-␥ (1000 units/ml, 48 h) served as a positive control for induction of these molecules, as described before (5).
Cytokine (TNF-␣) and chemokine (IL-8) levels were quantified by capture ELISA as described previously (20,21). Phorbol myristate pre-exposure (10 ng/ml, 12 h) served as a positive control for induction of IL-8 (20) and TNF-␣ (21). Western Blotting-Following experimental treatment of epithelial cells, whole cell extracts (for examination of CREB and phospho-CREB) were prepared as described previously (26). For analysis of nuclear extracts (NF-B activation), confluent monolayers of T84 cells on 100-mm Petri dishes were washed in ice-cold phosphate-buffered saline and lysed by incubation in 500 l of buffer A (10 mM HEPES (pH 8.0), 1.5 mM MgCl 2 , 10 mM KCl, 0.5 mM dithiothreitol, 200 mM sucrose, 0.5 mM phenylmethanesulfonyl fluoride, 1 g of both leupeptin and aprotinin per ml, 0.5% Nonidet P-40) for 5 min at 4°C. The crude nuclei released by lysis were collected by microcentrifugation (15 s). Nuclei were rinsed once in buffer A and resuspended in 100 l of buffer C (20 mM HEPES (pH 7.9), 1.5 mM MgCl 2 , 420 mM NaCl, 0.2 mM EDTA, 0.5 mM phenylmethanesulfonyl fluoride, 1.0 mM dithiothreitol, 1 g/ml of both leupeptin and aprotinin). Nuclei were incubated on a rocking platform at 4°C for 30 min and clarified by microcentrifugation for 5 min. Proteins were measured (DC protein assay, Bio-Rad). Samples (25 g/lane, as indicated) of T84 cell lysates were separated by non-reducing SDS-PAGE, transferred to nitrocellulose, and blocked overnight in blocking buffer (250 mM NaCl, 0.02% Tween 20, 5% goat serum, and 3% bovine serum albumin). For Western blotting, anti-NF-B (rabbit polyclonal antibody specific for p65 subunit of NF-B, Biomol, Plymouth Meeting, PA), anti-CREB (Upstate Biotechnology, Inc., Lake Placid, NY), or anti-phospho-CREB (Upstate Biotechnology, Inc., Lake Placid, NY) were added for 3 h; blots were washed, and species-matched peroxidase-conjugated secondary antibody was added, exactly as described previously (27). Labeled bands from washed blots were detected by ECL. Resulting bands were quantified from scanned images using NIH Image software (Bethesda).
Electrophoretic Mobility Shift Assay (EMSA)-Nuclear extracts of cells exposed to indicated experimental conditions were obtained as described above. The following synthetic oligonucleotide probes were synthesized (Genosys Biotechnologies, Inc.; The Woodlands, TX) and used as probes in EMSAs; the CRE-like motif (bold) lies at Ϫ115/Ϫ93 relative to the transcription start site in the TNF-␣ promoter, 5Ј-GTC-GACCTCCAGATGACGTCATGGGTTGTC-3Ј. Oligonucleotide probes for EMSA were digoxigenin-labeled according to manufacturer's instructions (Gel Shift kit, Roche Molecular Biochemicals). Labeled oligonucleotides were incubated with nuclear lysates for 10 min at 37°C and separated by electrophoresis on a 6% nondenaturing polyacrylamide gel. DNA-protein complexes were transblotted to nylon membrane, probed with anti-digoxigenin-peroxidase, and developed by ECL. Controls consisted of free probe alone and excess unlabeled probe. For supershift analysis, protein-DNA complexes were incubated with antiphospho-CREB antibodies (Upstate Biotechnology, Inc., Lake Placid, NY; 1:1000 dilution) for 1 h at 4°C prior to electrophoresis.
Measurement of cAMP-Confluent T84 monolayers on 6-well plates were exposed to indicated experimental conditions, and cAMP was quantified exactly as before (14). Briefly, cells were cooled to 4°C; cAMP was extracted from washed monolayers with extraction buffer (66% EtOH, 33% HBSS containing the phosphodiesterase inhibitor isobutylmethylxanthine, 5 mM, Sigma), and lysates were cleared by spinning at 10,000 ϫ g for 5 min and dried under vacuum to remove EtOH. Samples were reconstituted in assay buffer, and cAMP was quantified using displacement ELISAs (Amersham Pharmacia Biotech) according to manufacturer's instructions. Data were converted to concentration using a daily standard curve, and concentrations were expressed as cAMP per g of total protein.
Pharmacological Interference with Cellular cAMP Signaling-Epithelial cells grown on permeable supports were exposed basolaterally to Rp-cAMPs (0 -50 M Sigma), protein kinase A inhibitor amide (0 -30 M; Calbiochem), or 8-bromo-cAMP (0 -3 mM; Sigma) for 48 h prior to harvesting of cells for CREB determination or basolateral supernatants for TNF-␣ assay as described above.
Analysis of Messenger RNA Levels by PCR-Total RNA was purified according to the manufacturer's procedure from normoxic and hypoxic (6, 12 or 24 h) T84 cells using TRI Reagent (Molecular Research Center, Inc., Cincinnati, OH). 10 -50 g of each RNA sample was treated with DNase I (GenHunter Corp., Nashville, TN), and protein contaminants were then removed through a phenol/chloroform extraction step. RNA samples were ethanol-precipitated and resuspended in diethyl pyrocaronate-treated water. The reaction was set up according to Promega's Reverse Transcription System protocol (Promega Corp., Madison, WI). Briefly, 1 g of RNA was added to the reaction mixture consisting of 4 l of 25 mmol/liter MgCl 2 , 2 l of 10ϫ Reverse Transcriptase Buffer, 2 l of 10 mmol/liter of each dNTP, 0.5 l of rRNase in ribonuclease inhibitor (20 units total), 15 units of avian myeloblastosis virus reverse transcriptase, and 0.5 g of oligo(dT) 15 primer in a total volume of 20 l. The single-stranded cDNA was synthesized (MJ Research, Inc., Thermocycler Model PTC-200) using one cycle at 25°C for 10 min, one cycle at 42°C for 45 min, and one cycle at 95°C for 5 min followed by a final cycle at 4°C for 5 min. The Platinum TM Taq DNA Polymerase High Fidelity PCR System from Life Technologies, Inc., was used in the amplification step. The PCR reaction for human TNF-␣ contained 1 M each of the sense primer (5Ј-CGGGACGTGGAGCTGGCCGAGGAG-3Ј) and the antisense primer (5Ј-CACCAGCTGGTTATCTCTCAGCTC-3Ј), 5 l of Reverse Transcriptase reaction, 5 l of 10ϫ Platinum TM Taq high fidelity PCR buffer, 2 l of 50 mM MgSO 4 , 0.2 mmol of dNTP, and 2.5 units of Platinum TM Taq High Fidelity enzyme mix in a total volume of 50 l. The amplification reaction included a 5-min denaturation at 94°C and a 5-min annealing at 60°C, followed by 30 cycles at 72°C of 1.5 min, 94°C for 45 s, and 60°C for 45 s, with a final extension at 72°C for 10 min. The PCR reactions were then visualized on a 1% agarose gel containing 5 g/ml ethidium bromide. A 355-base pair fragment corresponding to TNF-␣ was observed. In order to ensure that an equal amount of template was used in each amplification reaction, 5 l of Reverse Transcriptase reaction was used as template with 1 M each of human ␤-actin sense primer (5Ј-TGACGGGGTCACCCACACTGTGC-CCATCTA-3Ј) and antisense primer (5Ј-CTAGAAGCATTTGCGGTG-GACGATGGAGGG-3Ј) in identical reactions; a 661-base pair amplified fragment with equal intensity was observed in all samples.
Generation of Retroviral CREB Overexpressing Cells-Retroviralmediated gene transfer of T84 cells with CREB and mutant CREB (serine to alanine mutated at site 133, the PKA phosphorylation site) containing vectors was performed using a previously described technique (28). Briefly, CREB or mutant CREB cDNA (a kind gift from Dr. Marc Montminy, Harvard Medical School) was expressed under the control of the CMV-IE promoter, and the cDNA for the dominant selectable marker (neomycin resistance) was expressed under control of the viral long terminal repeat. 10 6 epithelial cells were plated 24 h prior to infection. Cells were washed once, and then 3-4 ml of fresh, 0.45-m filtered, viral supernatant supplemented with 4 g/ml Polybrene were added to the adherent cells. After 8 h, 5 ml of fresh complete medium was added, and the cells were cultured for 48 h before drug selection.
Data Presentation-Cytokine, nucleotide, and cell-surface ELISA data were evaluated by analysis of variance and by Student's t test with p Ͻ 0.05 considered significant. All values are given as mean Ϯ S.E. for n experiments.

Hypoxia Selectively Induces Epithelial Proinflammatory Pro-
teins-Recently, we demonstrated that hypoxia enhances epithelial responses to the cytokine IFN-␥, and such responses were subsequently attributed to induction of epithelial TNF-␣ release from the basolateral surface (21). These studies, however, did not reveal significant insight into mechanisms that evoke induction of TNF-␣, and therefore, we sought to understand such findings at the mechanistic level. As an initial screen, a panel of epithelial proinflammatory markers were examined to determine whether such signaling by hypoxia was specific for TNF-␣ or is generalized to other molecules. As shown in Fig. 1, responses to hypoxia were specific. Indeed, epithelial exposure to hypoxia induced TNF-␣ release (maximal 37.0 Ϯ 6.1-fold increase, p Ͻ 0.01), IL-8 release (maximal 7.0 Ϯ 0.8-fold increase, p Ͻ 0.01), and induced surface expression of MHC class II (maximal 9.0 Ϯ 0.9-fold increase, p Ͻ 0.01). Hypoxia did not, however, influence epithelial ICAM-1 expression (p ϭ not significant compared with normoxia) or MHC class I (p ϭ not significant compared with normoxia) indicating a degree of specificity for signaling by hypoxia.
Hypoxia Induces Epithelial NF-B-In an attempt to gain insight into such specificity, we began by examining induction pathways of these pro-inflammatory genes. The regulatory regions of each of these genes contain a binding site for NF-B, a transcription factor important in induction of a number of proinflammatory genes (29). As shown in Fig. 2, Western blot analysis of nuclear extracts derived from epithelia exposed to hypoxia (measured pO 2 20 torr for 0 -48 h) revealed cytoplasmic-to-nuclear localization of the p65 subunit NF-B, a reliable readout of NF-B activation (27). Indeed, periods of hypoxia as short as 6 h revealed a significant cytoplasmic-to-nuclear localization of the p65 subunit of NF-B. Such responses to hypoxia were maximal by 24 h (densitometric measurement of 360.5 versus normoxic control value of 82.2 relative units) and were similar to our positive control PMA (497.9 relative units).
Hypoxia Induces the Expression of TNF-␣ Messenger RNA-We next concentrated on elucidating pathways of epithelial gene induction by hypoxia, specifically using TNF-␣ to define these principles. TNF-␣ gene induction and protein release can be regulated in a number of ways including transcriptional and post-translational pathways. Thus we examined whether hypoxia induces transcription of the epithelial TNF-␣ gene. As shown in Fig. 3, reverse transcriptase-PCR analysis revealed a time-dependent induction of TNF-␣. Such induction was detectable at 12 h, maximal at 24 h, and exceeded that of our positive control (PMA, 10 ng/ml for 12 h). These data reveal that hypoxia activates transcriptional pathways.
Role of cAMP Response Element in Induction of TNF-␣-In addition to a binding site for the transcription factor NF-B, the TNF-␣, IL-8, and MHC class II genes bear a cAMP responsive consensus binding site, termed a cAMP response element (CRE) (26,27). Importantly, the regulatory regions of those molecules not induced by hypoxia (MHC class I and ICAM-1, see Fig. 1) do not contain a CRE. These CRE DNA consensus motifs serve to regulate transcription of genes through protein kinase A and calcium-dependent pathways. In general, increases in intracellular cAMP are associated with decreased TNF-␣ release (30). Phosphorylation of nuclear proteins, termed cAMP response element binding proteins (CREB), positively or negatively regulate the activation of CRE-containing genes (e.g. cAMP can increase or decrease transcription) depending on the gene (31). Since a number of different consensus motifs serve as functional CREs, we examined the contribution of the CRE found within the TNF-␣ promoter. To do this, we developed an EMSA using oligonucleotides flanking the TNF-␣ CRE (32). As shown in Fig. 4, proteins from nuclear extracts of T84 cells bind to the CRE region of the TNF-␣ promoter. Furthermore, addition of antibody to the phosphorylated form of CREB demonstrate that the protein is indeed CREB. This so called "supershift" reaction indicates that the TNF-␣ gene bears a CRE which binds CREB of epithelial origin.
Hypoxia Down-regulates Epithelial CREB Expression-We next determined whether hypoxia influenced cellular CREB levels. As shown in Fig. 5A, Western blot analysis revealed that conditions which induce epithelial TNF-␣, IL-8, and MHC class II (hypoxia) resulted in attenuated levels of CREB (81.3, 90.7, and 91.1% decrease by densitometry at 24, 48 and 72 h hypoxia, respectively, compared with normoxic control). Addition of the PKA agonist forskolin (1 M, 24 h) partially protected the abrogation of CREB expression by hypoxia, particularly at the 24-h time point (117% of normoxic controls). Similarly, as shown in Fig. 5B, total levels of nuclear phospho-CREB were also influenced by hypoxia (47.4, 24.0, and 0.0% decrease by densitometry at 24, 48 and 72 h hypoxia, respectively, compared with normoxic control), and addition of forskolin partially reversed this response.
Role of Intracellular cAMP in Hypoxia-elicited Induction of TNF-␣-To futher evaluate the role of CREB in induction of epithelial TNF-␣, we determined the impact of elevating intracellular cAMP. We have recently demonstrated that conditions that liberate TNF-␣ production (i.e. cellular hypoxia) result in parallel diminutions in intracellular cAMP (14). Thus, we reasoned if TNF-␣ induction is negatively regulated by cAMP, then addback of intracellular cAMP (using the analog 8-bromo-cAMP) should diminish TNF-␣ release. The results shown in Fig. 6 demonstrate that decreased cAMP parallels increased epithelial TNF-␣ (Fig. 6A). Furthermore, the addition of 8-bromo-cAMP(0 -3mM)diminishedTNF-␣releaseinaconcentrationdependent manner (Fig. 6B; analysis of variance, p Ͻ 0.001). Addition of cAMP analogs to control cells did not activate cytoplasmic-to-nuclear localization of p65 and did not influence NF-B activation by either PMA or by cellular hypoxia (as determined by cytoplasmic-to-nuclear localization of p65, data not shown). Finally, addition of 8-bromo-cAMP to hypoxic cells (8 h) partially reversed the diminution in CREB levels (Fig.  6C). Interestingly, unlike our findings with forskolin (see  Nuclear lysates were generated and mixed with digoxigenin-labeled double-stranded oligonucleotide probes spanning Ϫ115 to Ϫ93 of the TNF-␣ promoter containing CRE consensus sequence (31). DNA-protein complexes were resolved on a 6% non-denaturing polyacrylamide gel, transblotted to nylon filters, probed with anti-digoxigenin-horseradish peroxidase, and developed by ECL. Lanes (left to right) represent: 1st, free probe without added protein; 2nd, lysates derived from control cells (note shifted band, black arrow); 3rd, lysates derived from control cells with added anti-phospho-CREB (note supershift, open arrow); 4th, same as 2nd lane with addition of 100ϫ unlabeled oligonucleotide. Data shown are representative of 3 experiments .   FIG. 2. Hypoxia activates NF-B translocation. T84 cells were grown on plastic cell culture plates and exposed to hypoxia for periods up to 48 h with the balance of the time in normoxia. PMA (1 g/ml)treated cells serve as a positive control. Following exposure, nuclear lysates were separated by SDS-PAGE and, following transfer to a nitrocellulose membrane, were detected with an antibody to the p65 subunit of NF-B. Integrated densitometry was carried out to quantify differential protein display. Data shown are a representative of 3 experiments.
Taken together, such data indicate the likelihood that the CREB-binding site is critical to induction of epithelial TNF-␣.
Protein Kinase A Inhibitors Induce TNF-␣ in Normoxic Epithelia-To directly examine the role of PKA in induction of epithelial TNF-␣ release, a specific PKA inhibitor (PKI) was used to diminish cAMP signaling in normoxic cells. TNF-␣ release and CREB expression were used as readouts for this response. As shown in Fig. 7A, PKI induced a significant increase in epithelial TNF-␣ release (9.0 Ϯ 3.9-fold increase over control) albeit to a lesser degree than epithelial exposure to hypoxia (see Fig. 1E). PKI also induced parallel decreases in total cellular CREB expression (Fig. 7B). Additionally, Rp-cAMPs, another PKA inhibitor, also induced TNF-␣ release (50 M; 2.4 Ϯ 0.6-fold increase over control). As a control for these experiments, we examined the influence of Rp-cAMPs on epithelial ICAM-1 expression. Rp-cAMPs failed to increase basal (0.6 Ϯ 0.02-fold over basal; p ϭ not significant compared with untreated controls) or IFN-␥-induced (0.82 Ϯ 0.06-fold over basal; p ϭ not significant compared with untreated controls) expression of ICAM-1. Such data indicate a direct role for PKA in induction of CRE-regulated genes and suggest that hypoxiaelicited diminutions in intracellular cAMP may explain induction of epithelial TNF-␣.
Overexpression of CREB Reverses the Hypoxia-elicited TNF-␣ Induction-The above described results define a prominent role for repression of CREB and phospho-CREB expression in induction of TNF-␣ by hypoxia. Thus, we reasoned that overexpression of CREB should, at least in part, diminish hypoxiaelicited induction of TNF-␣. T84 cells overexpressing CREB or mutant CREB (Ser-133/Ala-133) were generated by retroviral gene transfer (demonstrated by Western blot, Fig. 8A). Compared with control epithelia (12.21 Ϯ 2.26 pg of TNF-␣/mg of protein), cells overexpressing CREB (5.0 Ϯ 0.69 pg/mg protein; p Ͻ 0.02 compared with control cells), but not mutant CREB (10.38 Ϯ 1.74 pg/mg protein; p ϭ not significant compared with control cells), were significantly less responsive to hypoxia. These data provide direct evidence for CREB in hypoxia-elicited induction of TNF-␣. DISCUSSION Significant evidence supports a role for ischemia and resultant tissue hypoxia in development and maintenance of both chronic and acute inflammatory processes. For this reason, a detailed understanding of the mechanistic responses to hypoxia is crucial for the development of therapeutic strategies. In the studies outlined in this investigation, we sought to define pathways involved in the induction of proinflammatory gene products by cellular hypoxia, with an emphasis on TNF-␣. Given the critical role of epithelia in mucosal protection, a well defined intestinal epithelial model (T84 cells) was utilized to elucidate these pathways. Three primary observations are of note. First, these data indicate that cellular hypoxia results in transcriptional induction of genes that bear a CRE motif within the regions responsible for the regulation of gene expression. Second, down-regulation of protein kinase A activity, via cellular hypoxia, results in diminution of CREB expression, and overexpression of CREB in epithelial cells exposed to hypoxia protects the cells from transformation to the hypoxic phenotype. Third, these data reveal that maintenance of intracellular cAMP levels abrogates specific influences of epithelial hypoxia.
Recently, we undertook a series of studies aimed at identifying epithelia-derived factors that promote permeability through autocrine pathways. These studies revealed a soluble, transferable factor released from the epithelial surface in a polarized manner (similar to IL-8) (20) and bound to functional basolateral epithelial receptors (similar to IL-4, IL-13, and IFN-␥) (8). This factor was subsequently defined as soluble TNF-␣ (21). Herein, extensions of this work revealed induction of a number of epithelial proinflammatory markers (e.g. IL-8, MHC class II) and some degree of specificity in this regard (i.e. lack of induced MHC class I or ICAM-1). Examination of the promoter region of each these molecules revealed a binding site for NF-B, a transcription factor important in a number of inflammation-related pathways (29). Hypoxia induced NF-B cytoplasmic-to-nuclear localization; however, it was likely that additional factors were necessary, since both ICAM-1 and MHC FIG. 5. Hypoxia represses expression of CREB. T84 cells were grown to confluence on tissue culture-treated plastic wells and exposed to 0 -72 h hypoxia (Ϯ1 M forskolin, FSK and Forsk) with the balance of time in normoxia. Whole cell lysates were prepared and separated by SDS-PAGE before transfer to nitrocellulose membrane. CREB and phospho-CREB were detected with specific antibodies. Integrated densitometry was utilized to quantify differential protein display. Data shown are representative of 3 experiments. CTL, control (monolayers not exposed to forskolin). class I contain NF-B sites but are not induced by hypoxia. We have not determined the mechanism of epithelial NF-B activation by cellular hypoxia. However, we recently demonstrated a role for hypoxia-associated metabolic acidosis in proteasome activation and subsequent NF-B translocation in endotoxinactivated endothelial cells (27). Whether similar mechanisms play a role here are not known at the present time.
Closer examination of hypoxia-elicited gene sequences (IL-8, TNF-␣, and MHC class II) revealed the presence of CRE binding motifs, and thus, we determined the specific role of CRE and intracellular cAMP in induction of TNF-␣. A number of previous studies have suggested that proinflammatory gene products are responsive to alterations in intracellular cAMP. For instance, macrophage activation, and in particular TNF-␣ gene expression, is specifically down-regulated by agents that elevate cAMP (prostaglandin E 2 , dibutyryl cAMP, cholera toxin, and 8-bromo-cAMP) (30,(33)(34)(35)(36). Likewise, endothelial E-selectin, a CRE-bearing gene (37), is inversely related to intracellular cAMP (18). Consistent with previous reports, we FIG. 6. Exogenous cAMP reverses induction of epithelial TNF-␣. T84 epithelial cells were cultured on permeable supports and exposed to conditions that induce TNF-␣ release (hypoxia) for indicated periods (A) or to similar conditions in the presence of indicated concentrations of 8-bromo-cAMP (B). Monolayers were harvested and examined for intracellular cAMP (A) and polarized TNF-␣ release by ELISA (B) as described above; n, normal control; H, 48-h hypoxia. Data are expressed as mean Ϯ S.E. for n ϭ 4 -6 experiments. C, epithelia were exposed to normoxia (N) or hypoxia (H) for 8 h in the presence or absence of 8-bromo-cAMP (3 mM) and were examined for expression of CREB by Western blot (see "Materials and Methods"). demonstrate here that physiologic responses to cellular hypoxia are diminished intracellular cAMP and a parallel decrease in protein kinase A activity (14,15). Although the mechanisms underlying decreased cAMP and PKA activity remain to be clarified, the association of hypoxia with induction of CRE-containing genes may serve as a basic mechanism in regulating proinflammatory gene expression under these conditions.
To futher elucidate the role of CRE in induction of epithelial TNF-␣, we defined the impact of elevating intracellular cAMP. Addback of intracellular cAMP (using the analog 8-bromo-cAMP) diminished hypoxia-induced TNF-␣ release, and downregulation of PKA activity (using PKI and Rp-cAMPs) induced TNF-␣, albeit to a lesser extent than hypoxia. Importantly, epithelial exposure to 8-bromo-cAMP did not influence NF-B activation (data not shown). As a final point, it is important to note that CRE can also be regulated indirectly or directly by elevations in intracellular Ca 2ϩ , as would be the case with PMA exposure (31). Taken together, such data indicate the likelihood that the CREB-binding site is critical to induction of epithelial TNF-␣.
To define the causal relationship between hypoxia-elicited alterations in CREB expression and TNF-␣ release, we generated stable, T84 CREB-overexpressing cells. Exposure of such CREB-overexpressing cells to hypoxia resulted in significantly attenuate release of TNF-␣, thus indicating a direct relationship between hypoxia-elicited CREB decreases and TNF-␣ release. These studies also reveal a relative importance for phospho-CREB in the hypoxia response. Both CREB and phospho-CREB levels were significantly diminished by hypoxia; furthermore, overexpression of wild type CREB attenuated hypoxia-elicited TNF-␣ release. Cells overexpressing mutated CREB (which lacks a functional phosphorylation site) lost the capacity to attenuate responses to hypoxia.
Our results indicate that chronic PKA inactivation (via cellular hypoxia) results in diminished nuclear levels of CREB. Although it is known that CRE-related gene induction is tightly coupled through a series of specific events (PKA activation, CREB phosphorylation, and transcriptional induction/inhibition) (38), little is known about substrate (e.g. CREB) regulation. Our findings of diminished CREB expression by hypoxia suggest a negative feedback loop in which diminished PKA activity (via hypoxia) regulates expression of CREB, particularly unphosphorylated CREB. Evidence for such a pathway is twofold. First, epithelial phospho-CREB levels were only minimally influenced by hypoxia, and unphosphorylated CREB levels were depleted to nearly undetectable levels under similar conditions. These results define a primary role for regulation of CREB expression during hypoxia. Although little is known about regulated CREB expression, increasing evidence indicates a physiologic role for differential CREB levels in activated gene transcription. Indeed, CREB protein expression can be induced or repressed through a number of pathways, including elevated intracellular cAMP (39 -42). Moreover, significant tissue and cell specificity exist with regard to this issue. For instance, activation of Sertoli cell cAMP pathways activate CREB expression (41,43), and similar conditions result in long-lasting down-regulation in neurons (44). Additionally, in rat pheochromocytoma cells (PC12), it was recently demonstrated that hypoxia activates phosphorylation of CREB through an apparently novel pathway (45). Our data with epithelia suggest that chronically diminished PKA activity through hypoxia results in decreased CREB expression and subsequent activation of TNF-␣ transcription. As a second line of evidence for this pathway, in a similar fashion as hypoxia, inhibition of PKA in normoxic cells (with PKI-amide) resulted in a partial loss of unphosphorylated CREB, and activation of PKA (with forskolin) under hypoxic conditions partially salvaged the repression of CREB. Thus, the inhibition of transformation to the hypoxia-elicited phenotype by overexpression of CREB gives further evidence for a causal relationship between hypoxia-elicited decreases in CREB and the induction of TNF-␣ release by intestinal epithelial cells.
In summary, these results indicate that hypoxia-elicited diminution of CREB expression through diminished PKA activity may serve as a basic mechanism of activating negatively regulated, CRE-bearing genes in epithelia. Such findings have implications in the pathophysiology of a number of hypoxiarelated disorders and provide a potential target for the development of therapeutic strategies.