A Role for NF-κB Binding Motifs in the Differential Induction of Chemokine Gene Expression in Human Corneal Epithelial Cells

Abstract

purpose. The interleukin (IL)-8 promoter possesses a NF-κB–binding site with affinity to p50p65 and p65p65 complexes while the monocyte chemoattractant protein (MCP)-1 promoter’s NF-κB–binding site has exclusive affinity to p50p65 heterodimers. The purpose of this study was to determine whether the two NF-κB sites play a role in the capacity of tumor necrosis factor (TNF)-α–stimulated human corneal epithelial cells (HCECs) to produce nanogram amounts of IL-8 in the absence of MCP-1 synthesis.

methods. IL-8 and MCP-1 promoters were cloned into luciferase reporter vectors. Site-directed mutagenesis of wild-type promoters was used to mutate the NF-κB–binding motif in the wild-type IL-8 reporter plasmid into a motif with exclusive affinity to p50p65 and to mutate the NF-κB binding motif in the wild-type MCP-1 reporter plasmid into a motif with affinity to p65p65. Luciferase activity was determined after transfection of reporter vector constructs into TNF-α–stimulated HCECs. The chromatin immunoprecipitation assay was used to confirm binding of NF-κB subunits to IL-8 and MCP-1 promoters in vivo.

results. Promoters with affinity to p65p65 homodimers were active in driving the expression of the reporter gene, whereas promoters with affinity to p50p65 heterodimers did not induce significant reporter gene expression. Incorporation of a CCAAT enhancer–binding protein (C/EBP)–binding site immediately upstream of p65p65-binding sites significantly enhanced promoter activity.

conclusions. The results suggest that the interaction of p65p65 homodimers and C/EBP transcriptional factors with IL-8 promoters and not MCP-1 promoters account for the capacity of HCECs to produce IL-8 selectively, in the absence of MCP-1 production.

Inflammatory chemokines are low-molecular-weight secreted proteins synthesized at sites of injury or disease in response to proinflammatory mediators such as interleukin (IL)-1α and tumor necrosis factor (TNF)-α. ^{1 2} Inflammatory chemokines contribute to host defenses by chemoattraction of leukocytes into inflamed tissue. ^{3 4} Both human corneal epithelial cells (HCECs) and human corneal keratocytes (HCKs) produce proinflammatory chemokines when exposed to appropriate stimuli. ^{5 6 7 8 9} However, the mix of chemokines produced by the two cell types in response to proinflammatory mediators is highly regulated. For example, TNF-α–stimulated HCKs synthesize interleukin (IL)-8, monocyte chemoattractant protein (MCP)-1, and regulated on activation normal T-cell expressed and secreted (RANTES). In contrast, TNF-α–stimulated HCECs synthesize IL-8 but not MCP-1 or RANTES. ^{10 11 12} The mechanisms responsible for the capacity of human corneal cells to synthesize unique combinations of chemokines selectively in response to inflammatory mediators is not known. This is an important question, because the repertoire of chemokines released by corneal cells at sites of injury or disease can dictate the types of inflammatory cells that accumulate within inflamed corneal tissue. ^{13 14}  

TNF-α is a global inducer of chemokine synthesis because of its capacity to initiate signal-transduction pathways that activate the transcriptional factors needed for induction of chemokine gene expression. ¹⁵ One of the important transcriptional factors activated by the proinflammatory mediator is a dimeric transcriptional factor called NF-κB. ^{16 17 18} The two NF-κB subunits forming an NF-κB complex belong to a family of transcriptional activators of which the two most important members are p50 and p65. NF-κB dimers can exist as homodimeric complexes consisting of two identical subunits (p50p50 or p65p65) or as heterodimeric complexes containing two nonidentical subunits (p50p65). Within cells, homodimers and heterodimers formed from p50 and p65 subunits are held in the cytoplasm by an associated inhibitor. ¹⁹ Stimulation with appropriate proinflammatory mediators induces a protein kinase cascade that activates NF-κB by initiating degradation of the inhibitor, in turn allowing translocation of NF-κB dimers into nuclei where they can interact with NF-κB–binding sites on chemokine promoters. ²⁰  

To begin to understand how human corneal cells can selectively express one chemokine gene and not another in response to proinflammatory stimuli, we chose to investigate the mechanisms whereby IL-8 but not MCP-1 is selectively expressed in TNF-α–stimulated HCECs. ^{10 11} NF-κB complexes are known to be the major activator of IL-8 and MCP-1 gene expression in human cells stimulated with TNF-α. ^{21 22 23 24 25 26} However, the promoter of the two genes differ, in that the IL-8 promoter is activated by p65p65 homodimers whereas the MCP-1 promoter has been shown to be activated by p50p65 heterodimers. ^{21 22 23} The finding that TNF-α–stimulated HCECs can synthesize IL-8 without synthesizing MCP-1 led us to test the hypothesis that HCECs have evolved the capacity to use differences in affinity of the IL-8 and MCP-1 promoters to p65p65 homodimers to differentially regulate synthesis of the two chemokines in response to TNF-α. ^{10 11} Our results suggest that the differences in the affinity of the two binding motifs to p65p65 homodimers can account for the reason that TNF-α–stimulated cells synthesize IL-8 and not MCP-1, as well as a possible mechanism to account for the reason that other select chemokine genes in addition to MCP-1 are not expressed in TNF-α–stimulated HCECs. 

Materials and Methods

Cell Culture

Human donor corneas were obtained through the National Disease Research Interchange (Philadelphia, PA). Pure primary cultures of human corneal epithelial cells were established in keratinocyte serum-free medium (K-SFM; Invitrogen-Gibco, Grand Island, NY) within 4 days of enucleation, as previously described. ^{10 11 12} When epithelial cell cultures were approximately 90% confluent, media were aspirated and replaced with 2 mL K-SFM medium (Invitrogen-Gibco) containing selected concentrations of human recombinant TNF-α (Genzyme, Cambridge, MA) or phorbol myristate acetate (PMA; Sigma-Aldrich, St. Louis, MO). Human keratocyte cells were harvested from corneal donors as described previously. ^{10 11 12} Keratocyte cultures were used for experiments when they were approximately 90% confluent. 

Analysis of Pre-mRNA Levels by RT-PCR

RT-PCR products were generated from total cellular RNA with a kit (GeneAmp RT-PCR kit; Applied Biosystems [ABI], Foster City, CA) according to the manufacturer’s instructions. Primers used to amplify pre-mRNA molecules were chosen so that forward primers were complementary to intron sequences, whereas reverse primers were complementary to exon sequences, as described previously. ^{12 27} The identity of all RT-PCR products was verified by size and by sequencing (model 373XL; ABI). Each primer pair and the predicted size of its RT-PCR product are as follows: IL-8 pre-mRNA (592 bp), sense 5′-CTC TCT TGG CAG CCT TCC TGA TT-3′, antisense 5′-AAC TTC TCC ACA ACC CTC TGC AC-3′; MCP-1 pre-mRNA (265 bp), sense 5′-CCT ATA GGA GCA GTT TGC CCT GGG-3′, antisense 5′-TG GAG TGA GTG TTC AAG TCT TCG-3′. Previously published primers were used to amplify glyceraldehyde 3-phosphate dehydrogenase (GAPDH) pre-mRNA molecules, as described. ^{12 28} All RT-PCR amplifications used thermocycles of 30 seconds at 95°C, 30 seconds at 65°C, and 2 minutes at 72°C. PCR products were analyzed on a 1.5% agarose gel stained with 1 μg/mL ethidium bromide and photographed (Digital Science SP700 camera; Kodak Scientific Imaging Systems, New Haven, CT). The digitized images of the PCR products were quantitated using the accompanying software (Kodak Scientific Imaging Systems). 

Western Blot Analysis

Twenty-microgram aliquots of nuclear extracts were boiled for 4 minutes in SDS sample buffer (Novex, San Diego, CA) plus β-mercaptoethanol and separated on a 10% tris-glycine gel (Novex) with SDS running buffer. The contents of the gel were transferred to nitrocellulose membranes using a commercial system (Western Transfer Apparatus system; Novex). Membranes were blocked with 5% nonfat dry milk in PBS-T (phosphate-buffered saline [PBS] with 0.05% Tween 20) overnight at 4°C, washed with PBS-T, and incubated with either anti-p50 or anti-p65 antibodies (Santa Cruz Biotechnology, Santa Cruz, CA) for 1 hour at room temperature. Blots were washed three times with PBS-T, incubated with horseradish peroxidase–conjugated secondary antibody (Cappel, Durham, NC) for 1 hour at room temperature, and then washed three times with PBS-T. Immune complexes were detected by enhanced chemiluminescence Western blot detection reagents, according to the manufacturer’s recommendations (Amersham Pharmacia Biotech, Piscataway, NJ). 

Preparation of Nuclear Extracts

Nuclear extracts were prepared by modification of the method of Han and Brasier. ²⁹ Briefly, cells were washed two times in ice-cold PBS followed by a brief wash in 3 mL lysis buffer (10 mM HEPES [pH 7.9], 10 mM KCl, 0.1 mM EDTA, 0.1 mM EGTA, 1 mM dithiothreitol [DTT], 0.5 mM phenylmethylsulfonyl fluoride [PMSF]), 1 μg/mL pepstatin A, 1 μg/mL leupeptin, 10 μg/mL soybean trypsin inhibitor, 10 μg/mL aprotinin, and 0.5% Nonidet P-40). Additional lysis buffer (1.5 mL) was then added, and the monolayers were incubated on ice for 20 minutes with occasional rocking. Cells were collected by scraping and then transferred to 2-mL tubes. The cells were pelleted, supernatants drawn off, and pellets resuspended in 60 μL nuclear extraction buffer (25% glycerol, 20 mM HEPES [pH 7.9], 0.4 M NaCl, 1 mM EDTA, 1 mM EGTA, 1 mM DTT, 1 mM PMSF, 1 μg/mL pepstatin A, 1 μg/mL leupeptin, 10 μg/mL soybean trypsin inhibitor, and 10 μg/mL aprotinin). The samples were then vortexed 10 seconds before incubation on ice for 30 minutes. Extraction mixtures were subsequently centrifuged at 10,000g for 10 minutes at 4°C. The supernatants, representing nuclear extracts, were aliquoted and stored at −80°C. Protein concentrations were determined by the Bradford assay (Bio-Rad, Hercules, CA). 

Electrophoretic Mobility Shift Assays

Sense and antisense oligonucleotides were synthesized (model 1000M DNA Synthesizer; Beckman, Fullerton, CA) and annealed to form double-stranded DNA probes. The probes were end labeled with [(γ-³²P] adenosine triphosphate (ATP) and T4 polynucleotide kinase and purified on microcentrifuge spin columns (5′-Prime→3′-Prime TE Midi Select D, G-25; Shelton Scientific, Shelton, CT) according to the manufacturer’s instructions. The sense strand sequences of the wild-type (wt) and mutant (mt) oligonucleotide probes used for mobility shift assays are as follows (binding sites for NF-κB are in bold and mutations are italic): IL-8 wt NF-κB–binding site, 5′-AATCG[b]TGGAATTTCCTCTGAC-3′; IL-8 mNF-κB–binding site, 5′-AATCG[b]T TAACTTTCCTCTGAC-3′; MCP-1 wt −88 to −79 NF-κB–binding site, 5′ CTCAT[b]GGAAGATCCCTCCTCCT-3′; MCP-1 −88 to −79 mNF-κB–binding site, 5′ CTCAT TGACGCTCCATCCTCCT-3′; MCP-1 wt p65p65 NF-κB–binding site, 5′-CAGAGT[b]GGGAATTTCCACTCA-3′; and MCP-1 p65p65 mNF-κB–binding site, 5′-CAGAGT TGAACTGTCCACTCA-3′. 

Protein-DNA binding reactions were performed with 7.5-μg aliquots of nuclear extract, labeled oligonucleotide, and 1 μg poly(dI-dC) in 10% glycerol, 100 mM NaCl, 5 mM MgCl, and 5 mM DTT for 15 minutes at room temperature. For competition assays, 50-fold excess unlabeled probe was incubated with nuclear extract at room temperature 10 minutes before the addition of labeled probe. For antibody supershift assays, 1 μL of antisera (Santa Cruz Biotechnology) was added 15 minutes after the addition of labeled probe and incubated on ice for 1 hour. Protein-DNA complexes were resolved by electrophoresis on a DNA retardation gel (6% TBE; Novex) in 0.5× TBE running buffer (89 mM Tris base, 89 mM boric acid, and 2 mM EDTA). The gels were dried and visualized by autoradiography. 

Plasmid Construction, Transient Transfections, and Luciferase Assays

The pMCP-1 expression vector possesses the human MCP-1 promoter and the upstream enhancer (nucleotides −2850 to 2051, relative to the transcription start site). ^{24 25} To construct the reporter vector, MCP-1 nucleotide sequences were amplified by from the human genome by PCR using the forward primer 5′-GGTACCGGTAACTGAGGATTCTGGACAGCATCAGAGC-3′ possessing a KpnI restriction site and a reverse primer 5′- CCCGGGGGATGTTT CTGGGTTAGTCTCAGCCT CTCGG-3′ possessing an XmaI site. The PCR products were cloned into a cloning vector (TOPO TA; Invitrogen, Carlsbad, CA) followed by subcloning 2901 bp of the hMCP-1 promoter (−2876 to +24) into the pGL3-basic luciferase reporter plasmid (Promega, Madison, WI). 

The IL-8 promoter-driven luciferase reporter vector pIL8f was constructed by synthesizing a 151-bp fragment (−110 to +41) of the IL-8 promoter and ligating it into the KpnI-HindIII site of the pGL3-basic luciferase reporter plasmid. The MCP-1 promoter-driven luciferase reporter vector pMCP1f was constructed by synthesizing a 160-bp fragment of the MCP-1 promoter (−136 to +24) and ligating it into the KpnI-XmaI site of pGL3-basic luciferase reporter plasmid. Site-directed mutageneses of CCAAT enhancer–binding protein (C/EBP) and NF-κB sites within the pIL8f and pMCP1f reporter plasmids were performed using the oligonucleotides described in the Results section. All mutations were introduced with a site-directed mutagenesis system, according to manufacturer’s instructions (Promega). All mutagenized plasmids were confirmed by sequencing (model 373XL sequencer; ABI). 

Subconfluent monolayers of HCECs in six-well plates were cotransfected with 1.0 μg of a pGL3 firefly luciferase reporter construct and 0.25 μg of a pRL-TK Renilla luciferase internal control plasmid (Promega, Madison, WI), by using a transfection reagent (FuGene; Roche Molecular Biochemicals, Indianapolis, IN) according to the manufacturer’s instructions. After a 48-hour incubation at 37°C, the cells were stimulated with 1 mL of medium containing 500 U/mL TNF-α and 100 μg/mL PMA or with medium alone. Fifteen hours later, firefly luciferase activity was determined using reagents in a dual-luciferase reporter assay system (Promega) and normalized for Renilla luciferase activity. The induction multiples were calculated by dividing the ratio of normalized firefly luciferase activity from stimulated cells by the ratio of normalized firefly luciferase activity from nonstimulated cells. 

Chromatin Immunoprecipitation Assays

Human corneal epithelial cells (10⁶) were either untreated or treated with TNF-α (500 U/mL) for 1 hour. The chromatin immunoprecipitation assay was then performed using a kit (Chromatin Immunoprecipitation [ChIP] Assay; Upstate Biotechnology, Lake Placid, NY) as directed by the manufacturer. Fifty microliters of the diluted, sonicated cell supernatant was saved and used to quantitate the amount of input DNA present in different samples before immunoprecipitation. For immunoprecipitation, 5 μg of anti-p65, anti-p50, or normal IgG control or no antibodies (Santa Cruz Biotechnologies) were used. The samples (including input DNA samples) were extracted once with phenol-chloroform, precipitated with ethanol, and resuspended in 50 μL H₂O (10 μL for input samples). The samples (excluding input DNA) were then processed on spin columns (QIAquick PCR Purification Kit; Qiagen, Inc., Valencia, CA) and recovered in the same volume as stated earlier. Ten microliters (1 μL of input material) of the samples were then used for PCR (Platinum Taq polymerase; Invitrogen) consisting of the following parameters: 95°C for 2 minutes and 40 cycles of 95°C for 45 seconds, 60°C for 45 seconds, and 72°C for 2 minutes. The primers used to amplify the MCP-1 and IL-8 NF-κB promoter elements are as follows: MCP-1 (210 bp) 5′-CCCATTTGCTCATTTGGTCTCAGC-3′ (forward) and 5′-GCTGCTGTCTCTGCCTCTTATTGA-3′ (reverse); IL-8 (231bp) 5′-GAAGTGTGATGACTCAGGTTTGCC-3′ (forward) and 5′-ACACAGTGCGAATGGTTCCTTCCG-3′ (reverse). PCR products were analyzed by agarose gel electrophoresis and 1 μg/mL ethidium bromide staining. Agarose gel images were then obtained (Digital Science SP700 camera; Kodak Scientific Imaging Systems). 

Results

Comparison of MCP-1 and IL-8 RNA Synthesis in TNF-α–Stimulated HCECs

To confirm that IL-8 and MCP-1 gene expression is differentially regulated at the transcriptional level in TNF-α–stimulated HCECs, we compared levels of IL-8 and MCP-1 pre-mRNA synthesis in TNF-α–stimulated HCECs by RT-PCR with primers designed to amplify either IL-8 or MCP-1 pre-mRNA molecules. It was found that TNF-α–stimulated cells possessed five times more IL-8 pre-mRNA than MCP-1 pre-mRNA at 3 hours after stimulation and seven times more IL-8 pre-RNA than MCP-1 pre-RNA at 6 hours after stimulation (Fig. 1) . Significant amounts of IL-8 pre-mRNA were also detected at 12 hours and 18 hours after stimulation, whereas no MCP-1 pre-mRNA was detectable at these time points. These results suggest therefore, that the IL-8 promoter is significantly more active than the MCP-1 promoter in stimulating chemokine gene expression in TNF-α–stimulated HCECs. 

Identification of the NF-κB Subunits Activated by TNF-α–Stimulated HCECs

The NF-κB site on the IL-8 promoter binds p65p65 complexes, whereas the NF-κB–binding site on the MCP-1 promoter binds p50p65 complexes. ^{21 22 23} It is possible therefore, that IL-8 but not MCP-1 is synthesized in TNF-α–stimulated HCECs, since the cells only produce p65 subunits thereby limiting the type of NF-κB complexes formed to p65p65 homodimers. 

To determine the composition of NF-κB complexes activated in TNF-α–stimulated cells, nuclear lysates prepared from stimulated cells were analyzed by electrophoretic mobility shift and supershift assays. It was found that two nucleoprotein complexes were generated with cell lysates from TNF-α–stimulated cells after incubation with the IL-8 NF-κB probe, whereas one nucleoprotein complex was generated when cell lysates were incubated with the MCP-1 NF-κB motif (Fig. 2 , lane c). NF-κB-DNA complexes were not detected in nuclear lysates prepared from nonstimulated cells (Fig. 2 , lane b). The DNA-binding activity was specific, because formation of the protein–DNA complexes could be competed out by 50-fold excess unlabeled wild-type probe (Fig. 2 , lane h) but not by competing mutant probe (Fig. 2 , lane i). 

In supershift assays, antibodies against both p50 (Fig. 2 , lane e) and p65 (Fig. 2 , lane g) supershifted the low-molecular-weight nucleoprotein complex generated with the IL-8 NF-κB probe. In contrast, the higher-molecular-weight nucleoprotein complex generated with this probe was supershifted by p65 antibodies (Fig. 2 , lane g) but not by p50 antibodies (Fig. 2 , lane e). When supershift experiments were performed with the nucleoprotein complex generated with the MCP-1 NF-κB probe, it was found that the complex was supershifted with antibodies against both p50 and p65. These results suggest therefore, that both p65p65 homodimers and p50p65 heterodimers are present in HCECs where they are capable of localizing to nuclei and of binding to their respective promoters after stimulation with TNF-α. 

Role of NF-κB–Binding Sites in MCP-1 and IL-8 Promoter Activity

To rule out the possibility that IL-8 but not MCP-1 gene expression is activated in TNF-α–stimulated cells because NF-κB complexes differentially bind IL-8 promoters in vivo, we performed ChIP assays on chromatin isolated from TNF-α–stimulated cells using antibody against either p50 or p65 subunits. ³⁰ In nonstimulated cells, neither p65 nor p50 bound to MCP-1 promoters (Fig. 3) . In contrast, p65 subunits were found associated with the IL-8 promoter, suggesting that p65p65 homodimers are responsible for constitutive IL-8 gene expression. However, after TNF-α stimulation, the promoters of the MCP-1 gene as well as those of the IL-8 gene were bound to NF-κB subunits. 

To determine whether IL-8 and MCP-1 NF-κB binding motifs are involved in differential regulation of the two genes, HCEC cultures were transfected with luciferase reporter vectors driven by either the wild-type IL-8 (Fig. 4a) or the wild-type MCP-1 promoter (Fig. 4d) . HCEC cultures were also transfected with luciferase reporter vectors driven by a recombinant MCP-1 promoter in which the wild-type NF-κB–binding site was replaced with an IL-8 NF-κB binding motif (Fig. 4e) or by a recombinant IL-8 promoter in which the IL-8 NF-κB–binding site was replaced by an MCP-1 NF-κB motif (Fig. 4b) . The cells were then stimulated with TNF-α to test the activity of the reporter vector constructs. In HCECs transfected with the IL-8 wild-type reporter vector (pIL8f), TNF-α stimulation enhanced promoter activity more than 11-fold (Fig. 5A , row a), whereas TNF-α stimulation of epithelial cells transfected with the MCP-1 wild-type reporter vector (pMCP1f) had little effect on luciferase activity (Fig 5A , row d). Even though the MCP-1 wild-type reporter vector was not activated in TNF-α–stimulated cells, there was a threefold activation in PMA-stimulated cells, indicating that the reporter construct was functionally active (Fig. 5A , row e). Replacing the NF-κB motif in the wild-type MCP-1 promoter plasmid with a IL-8 NF-κB motif induced a small but significant increase in its activity in TNF-α–stimulated cells (Fig. 5A , row c), whereas replacing the NF-κB motif in the wild-type IL-8 promoter vector with an MCP-1 NF-κB motif diminished promoter activity nearly sixfold (Fig. 5A , row b). 

In addition to the promoter, MCP-1 gene expression in human cells can be regulated by a distal enhancer element. ²⁵ To rule out the possibility that the MCP-1 promoter vector was not activated in TNF-α–stimulated cells because the vector was missing an upstream enhancer, we transfected HCECs with an MCP-1 reporter vector possessing the upstream enhancer element. When these cells were stimulated with TNF-α, reporter activity was not significantly higher than the reporter activity of MCP-1 reporter vectors missing the enhancer element (Fig. 5f) . This suggests that the failure of MCP-1 promoter vector to be activated by TNF-α stimulation does not occur because the reporter vector is missing an enhancer element. 

To provide further evidence that the activity of recombinant MCP-1 promoters was enhanced in TNF-α–stimulated HCECs because their NF-κB–binding motif has affinity to p65p65 homodimers, we constructed a second recombinant MCP-1 luciferase reporter vector in which the wild-type MCP-1 NF-κB–binding site was mutated into 5′-GGAATTTCC-3′ (Fig. 4g) . This NF-κB–binding site, like the NF-κB–binding site on the IL-8 gene, is reported to be capable of binding p65p65 homodimers. ²⁵ When this reporter plasmid was transfected into HCECs, luciferase activity was once again significantly enhanced (Fig. 5A , row g). The affinity of this site to p65p65 was confirmed by demonstrating that antibodies against p65 (Fig. 5B , lane g) but not against p50 (Fig. 5B , lane e) shifted both of the two nucleoprotein complexes formed by incubation of TNF-α–stimulated epithelial cell nuclear extracts with the 5′-GGAATTTCC-3′ probe (Fig. 5B , lane c). 

Enhancement of Binding of p65p65 Homodimers to Recombinant NF-κB Binding Motifs

Compared with the 11-fold increase in promoter activity demonstrated by the wild-type IL-8 promoter, the two- to threefold increase in promoter activity demonstrated by p65p65-binding recombinant MCP-1 promoters suggests that a p65p65-binding NF-κB motif is only marginally better than a p50p65-binding NF-κB motif in activating MCP-1 gene expression. However, it has been shown that a C/EBP-binding motif located upstream from the IL-8 NF-κB–binding motif enhances binding of p65p65 homodimers to the IL-8 promoter by interacting with the homodimer to facilitate its binding to DNA. ^{22 26} This suggests therefore that the low levels of promoter activity exhibited by recombinant MCP-1 promoters compared with the IL-8 promoter may occur because recombinant promoters are missing a C/EBP-binding motif and therefore bind very weakly to p65p65 homodimers. 

To increase the affinity of the p65p65-binding MCP-1 promoter to its ligand, we mutated the nucleotide sequences immediately upstream of the IL-8 NF-κB–binding motif in the recombinant MCP-1–driven reporter vector pMCP1/IL8NFκB into a C/EBP-binding motif to create the promoter construct MCP1/IL8CEBP-NF-κB (Fig. 4h) . The presence of both a C/EBP- and a p65p65-binding motif on the MCP-1 promoter enhanced promoter activity sixfold (Fig. 6d) compared with the two- to threefold increase in the activity of a recombinant MCP-1 promoter construct that was missing the C/EBP-binding site but otherwise identical (Fig. 6c) . The contribution of C/EBP-binding motifs to p65p65 binding was affirmed by the finding that deletion of the C/EBP-binding site from the IL-8 promoter reduced its activity fourfold (Fig. 6b) . In contrast to its effects on p65p65 binding, creation of a C/EBP-binding motif upstream of the p50p65 binding NF-κB motif on promoter vectors driven by the wild-type MCP-1 promoter had no effect on promoter activity (Fig. 6c) . This finding is consistent with reports that C/EBP transcriptional factors specifically enhance binding of p65p65 complexes to NF-κB–binding motifs. ^{22 26}  

Discussion

An important clue as to how IL-8 but not MCP-1 can be synthesized in TNF-α–stimulated HCECs was the discovery that reporter vectors driven by MCP-1 promoters were not activated in TNF-α–stimulated cells, even though their NF-κB–binding motif became filled with p50p65 heterodimers. This finding was surprising, because the binding of p50p65 heterodimers to MCP-1 promoters has been shown to activate MCP-1 gene expression in other cell types. ²³ The inability of p50p65 complexes to activate the MCP-1 promoter was not due to construction of a nonfunctional reporter vector, because MCP-1–driven reporter vectors were significantly activated in PMA-stimulated HCECs. Furthermore, the MCP-1 promoter was fully capable of stimulating significant levels of reporter activity in response to TNF-α stimulation when its affinity to p65p65 was enhanced. In addition, ChIP analysis revealed that NF-κB complexes bind to both MCP-1 and IL-8 promoters located within TNF-α–stimulated epithelial cell chromatin. Thus, failure of MCP-1 gene expression to be activated in TNF-α–stimulated HCECs is not due to selective in vivo binding of NF-κB complexes to the promoter of the IL-8 gene. These results suggest therefore, that the wild-type MCP-1 promoter is not activated in TNF-α–stimulated HCECs because DNA-bound p50p65 NF-κB isomers are not effective inducers of MCP-1 gene expression in response to this proinflammatory mediator. The fact that we detected small levels of MCP-1 pre-RNA synthesis in RT-PCR experiments even though significant MCP-1 promoter activity was not detected in luciferase reporter assays could be because the MCP-1 promoter retains some constitutive activity when positioned in its normal chromosomal location. 

In contrast to reporter vectors driven by MCP-1 promoters, reporter vectors driven by IL-8 promoters were readily activated in TNF-α–stimulated HCECs. Even though the IL-8 NF-κB motif bound both the p50p65 and p65p65 isomers in response to TNF-α stimulation, mutation of the IL-8 promoter into a p50p65-binding MCP-1 NF-κB motif significantly decreased promoter activity. This observation confirms results obtained in several other laboratories where it has been demonstrated that p65p65 homodimers are primarily responsible for activation of IL-8 gene expression in response to proinflammatory mediators. ^{21 22} Nevertheless, this finding along with the finding that p50p65-charged MCP-1 promoters are also not responsive to TNF-α stimulation suggests that the disparity in levels of IL-8 and MCP-1 pre-mRNA produced in TNF-α–stimulated HCECs is a result of the fact that the IL-8 promoter possesses an affinity to p65p65 homodimers missing in the MCP-1 promoter. 

We can propose at least two possible mechanisms that could explain how differences in affinity of the two promoters to p65p65 homodimers could differentially regulate their responsiveness to TNF-α stimulation. First, it has been discovered that DNA-bound NF-κB complexes need to interact with several transcriptional cofactors to stimulate transcription. ^{31 32 33 34 35 36 37} These cofactors mediate their effects by either bridging NF-κB complexes to the basal transcriptional machinery or by remodeling chromatin structure. ³⁸ One important cofactor required for NF-κB mediated activation of gene expression is TAF_II105. ^{33 34} This cofactor reacts with p65p65 homodimers much more efficiently than with p50p65 heterodimers. ³⁵ On the basis of this observation, it was predicted that any cell expressing TAF_II105 would be able to stimulate promoters regulated by p65p65 to greater levels of activity than promoters regulated by p50p65. Thus, the capacity of TNF-α–stimulated HCECs to synthesize IL-8 in the absence of MCP-1 production may reflect the fact that TNF-α induces synthesis of one or more cofactors necessary for transactivation of promoters possessing p65p65 homodimers, but not those possessing p50p65 heterodimers. Second, transcriptional repressor proteins specific for the MCP-1 promoter have been identified. ³⁹ Thus, we have not ruled out the possibility that a specific repressor of MCP-1 transcription is activated in TNF-α–stimulated cells. 

Another question that remains to be answered is whether p50p65 complexes formed in HCECs in response to TNF-α also lack the capacity to activate other chemokine promoters that bind the heterodimer. Although we have not directly investigated this question, it has been reported that, like MCP-1, the chemokine genes RANTES and growth-responsive gene-α (GRO-α) are not significantly upregulated in TNF-α–stimulated HCECs. ^{11 12} Of note, the promoters of these two chemokine genes share with the MCP-1 promoter a dependency on p50p65 for activation. ^{40 41} This raises the possibility therefore, that the same mechanisms responsible for the incapacity of p50p65 heterodimers to activate MCP-1 and IL-8 gene expression in TNF-α–stimulated HCECs may also be responsible for the lack of responsiveness of these genes to TNF-α stimulation. 

 Supported by National Institutes of Health Grant EY12713.

 Submitted for publication April 10, 2003; revised July 14, 2003 and February 13, 2004; accepted March 9, 2004.

 Disclosure: M.H. Ritchie, None; R.A. Fillmore, None; R.N. Lausch, None; J.E. Oakes, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

The effects of TNF-α stimulation on steady state levels of intracellular IL-8 and MCP-1 pre-mRNA. IL-8 pre-mRNAs (A) or MCP-1 pre-mRNAs (B) were amplified by RT-PCR from total RNA extracted from HCEC cultures stimulated with 500 U/mL TNF-α or fresh medium. The relative density represents the increase (-fold) in the density of pre-mRNA amplified from stimulated cells over that of pre-mRNA amplified from nonstimulated cells. Identical experiments were conducted with two additional donors and gave similar results.

Figure 2.

 

Identification of NF-κB complexes activated in TNF-α–stimulated cells. ³²P-labeled oligonucleotide probes were incubated with nuclear extracts harvested from nonstimulated cells or cells stimulated with 500 U/mL TNF-α for 1 hour. DNA–protein complexes were separated by electrophoresis and exposed to x-ray film. In supershift analyses, antibodies against either p50 or p65 (lanes e–g) were incubated with the nuclear extract–probe mixtures for an additional hour at 4°C.

Figure 3.

 

In vivo binding of NF-κB subunits to the IL-8 and MCP-1 promoters in TNF-α–stimulated cells. HCECs were exposed to either fresh medium or 500 U/mL TNF-α for 1 hour. Sonicated, in vivo cross-linked chromatin was then immunoprecipitated using p65 (anti-p65), p50 (anti-p50), control (Control Ab), or no antibodies (data not shown). The recovered immunoprecipitated DNA was then used in PCR with primers specific for either the IL-8 or MCP-1 promoters. Input DNA was also analyzed by PCR to ensure equivalent amounts of extracts were used for immunoprecipitation.

Figure 4.

 

Sequence comparison of IL-8 and MCP-1 luciferase reporter plasmid constructs. Sequences −110 to +41 and −136 to +24 relative to the transcription start site of the IL-8 and MCP-1 genes are illustrated. Binding sites for transcription factors are boxed. The IL-8 and MCP-1 wild-type nucleotide sequences were each cloned into a luciferase reporter plasmid to generate pIL8f and pMCP1f. Site-directed mutagenesis generated: (e) an MCP-1 reporter plasmid possessing an IL-8 NF-κB–binding site, (b) an IL-8 reporter plasmid possessing an MCP-1 NF-κB–binding site, (c) an IL-8 reporter plasmid containing a nonfunctional C/EBP-binding site, (f) an MCP-1 reporter plasmid possessing an IL-8 C/EBP-binding site upstream of its NF-κB–binding site, (h) an MCP-1 reporter plasmid possessing a IL-8 C/EBP-binding site and an IL-8 NF-κB–binding site, and (g) an MCP-1 reporter plasmid possessing a p65p65 NF-κB–binding site. The sequences of oligonucleotides used for mutagenesis of the IL-8 and MCP-1 wild-type promoters are shown with point mutations underlined.

Figure 5.

 

(A) The effects of NF-κB–binding site substitutions on MCP-1 promoter activity in TNF-α–stimulated corneal epithelial cells. HCECs were transiently cotransfected with select reporter constructs and a Renilla luciferase internal control reporter. After stimulation with either 500 U/mL TNF-α or 100 μg/mL PMA, the cells were lysed and firefly and Renilla luciferase activity determined. Induction (-fold) is expressed as the ratio of normalized firefly luciferase activity in extracts from stimulated cells relative to normalized firefly luciferase activity in extracts from nonstimulated cells. (Data are expressed as the mean ± SE, *P < 0.01.) (B) Identification of NF-κB complexes that bind to the MCP-1/MCP-1 probe. ³²P-labeled oligonucleotide probes were incubated with nuclear extracts of nonstimulated cells or cells stimulated with 500 U/mL TNF-α for 1 hour. DNA-protein complexes were separated by electrophoresis and exposed to x-ray film. In supershift analysis, antibodies against either p50 or p65 were incubated with the nuclear extract–probe mixtures for an additional 1 hour at 4°C.

Figure 6.

 

The effects of the IL-8 C/EBP-binding site and NF-κB–binding site substitutions on MCP-1 promoter activity. HCECs were transiently cotransfected with select reporter constructs and a Renilla luciferase internal control reporter. After stimulation with 500 U/mL TNF-α, cells were lysed, and firefly and Renilla luciferase activity determined. Induction (-fold) is expressed as the ratio of normalized firefly luciferase activity in extracts from stimulated cells relative to normalized firefly luciferase activity in extracts from nonstimulated cells. (Data expressed as the mean ± SE, *P < 0.01.)