Inflammatory Versus Proliferative Processes in Epidermis

Epidermal keratinocytes respond to injury by becoming activated, i.e. hyperproliferative, migratory, and proinflammatory. These processes are regulated by growth factors and cytokines. One of the markers of activated keratinocytes is keratin K6. We used a novel organ culture system to show that tumor necrosis factor α (TNFα) induces the expression of K6 protein and mRNA in human skin. Multiple isoforms of K6 are encoded by distinct genes and have distinct patterns of expression. By having shown previously that proliferative signals, such as epidermal growth factor (EGF), induce expression of the cytoskeletal protein keratin K6b, we here demonstrate that the same isoform, K6b, is also induced by TNFα, a proinflammatory cytokine. Specifically, TNFα induces the transcription of the K6b gene promoter. By using co-transfection, specific inhibitors, and antisense oligonucleotides, we have identified NFκB and C/EBPβ as the transcription factors that convey the TNFα signal. Both transcription factors are necessary for the induction of K6b by TNFα and act as a complex, although only C/EBPβ binds the K6b promoter DNA. By using transfection, site-directed mutagenesis, and footprinting, we have mapped the site that responds to TNFα, NFκB, and C/EBPβ. This site is separate from the one responsive to EGF and AP1. Our results show that the proinflammatory (TNFα) and the proliferative (EGF) signals in epidermis separately and independently regulate the expression of the same K6b keratin isoform. Thus, the cytoskeletal responses in epidermal cells can be precisely tuned by separate proliferative and inflammatory signals to fit the nature of the injuries that caused them.

Epidermis is our first line of defense from the environment and must often respond to various types of injury. Keratinocyte is the predominant cell type in the epidermis. When an injury occurs keratinocytes can become hyperproliferative, e.g. in wound healing, or they can become inflammatory, e.g. in contact dermatitis, or both, e.g. in psoriasis (1,2). These responses are coordinated and orchestrated by growth factors and cyto-kines that carry signals from cells to cells (3). Keratinocytes produce paracrine signals to alert fibroblasts, endothelial cells, melanocytes, and lymphocytes, as well as autocrine signals targeted at neighboring keratinocytes. In response to such signals keratinocytes produce and release additional signaling molecules, increase the levels of their cell surface receptors, and change their cytoskeleton.
The most prominent cytoskeletal proteins in keratinocytes are keratins. The basal layer of the epidermis produces keratins K5 and K14; the differentiating, suprabasal layers produce K1 and K10, and the activated keratinocytes express keratins K6, K16, and K17 (4,5). These three keratins are markers of inflammatory and hyperproliferative processes. Healthy interfollicular epidermis does not contain K6, K16, or K17; however, these keratins are constitutively expressed in the outer root sheet of the hair follicle, the nail bed, oral epithelia, and some other tissues (reviewed in Ref. 6). The expression of keratins K6 and K16 can be disconnected from proliferation per se, and cultured keratinocytes express K6 even when their proliferation is inhibited (7). Thus, the presence of K6 and K16 keratins could be a marker of inflammation.
In vitro, when grown in cell culture as monolayers, keratinocytes are already activated and hyperproliferative, expressing copious amounts of K6 keratin (8). However, convenient in vivo systems for analysis of the effects of TNF␣ and other growth factors and cytokines in human skin have not been described. Therefore, we developed an elegant and convenient "propior vivo" experimental system, using organ culture explants of human skin specimens otherwise discarded during surgery (9). We incubated such biopsies of human skin for relatively short times in minimal keratinocyte culture medium to which we added TNF␣. 1 In this system we demonstrated the specific induction of K6 keratin protein and mRNA expression by TNF␣, without the frequently concomitant proliferative signals.
The specific function of the K6 keratin is not fully understood, but curiously, there are multiple K6 genes in mammals, encoding virtually identical proteins (10,11). Mutations in K6a keratin lead to pachyonychia congenita Type I, whereas those in K6b lead to pachyonychia congenita Type II, reflecting the slightly different expression patterns of the two genes (12,13). The paralogous genes are differentially expressed in human and in murine tissues, although their expression is generally restricted to the suprabasal layers of stratified epithelia (11, 14 -16).
This poses an interesting conundrum: if both proliferative and proinflammatory signals induce K6 expression, is the same K6 protein induced by both, or does one of the K6 genes respond to the proliferative and the other to the proinflammatory stimuli? Our previous results have demonstrated that EGF or TGF␣, the proliferative signals, induce the expression of keratins K6b and K16 (17,18). Therefore, focusing on the K6b keratin, we decided to determine whether TNF␣, which is a strong proinflammatory but not a proliferative signal in keratinocytes (19), affects its expression.
TNF␣ is produced by a wide variety of cells in response to infection or injury, primarily macrophages and monocytes but also by epithelial cells including epidermal keratinocytes. A low level of TNF␣ is present in the upper layers of the healthy epidermis, but its synthesis and release from keratinocytes is greatly augmented in allergic and irritant contact dermatitis, infection, UV irradiation, etc. (20,21). In these pathological conditions, TNF␣ activates immune responses through inducing production of proteins such as amphiregulin, TGF␣, IL-1␣, IL-1 receptor antagonist, EGF receptor, and ICAM1 (22)(23)(24)(25)(26). Mice lacking TNF␣ develop normally but have delayed and prolonged inflammatory responses, confirming the role of TNF␣ in inflammation (27).
The signaling cascades mediating cellular responses to TNF␣ have been partly elucidated (28 -30). There are two TNF␣ receptors, but keratinocytes express mainly the 55-kDa receptor, type 1 (31)(32)(33). The most direct TNF␣ effect involves proteins TRADD and TRAF2 and activates transcription factors NFB and C/EBP␤. The NFB family includes the proteins p65, p50, and c-Rel, which both homo-and heterodimerize among themselves (34). These proteins are stored latent in the cytoplasm, bound to the inhibitory protein, IB. TNF␣ causes activation of IKKs, kinases that phosphorylate IB and induce its degradation, which results in activation and nuclear translocation of the NFB protein (28,(35)(36)(37). Knockout of IKK␣ has severe epidermal phenotype causing incomplete epidermal differentiation (38,39). On the other hand a knockout of IKK␤ is defective in signaling from TNF␣ to NFB (40,41). NFB proteins can interact with C/EBP␤, AP1, and other transcription factors to regulate gene expression (42,43). In keratinocytes in vitro overexpression of NFB inhibits proliferation. In epidermis in vivo NFB is present in all layers but is nuclear only in the suprabasal ones; this suggests a role for NFB in epidermal differentiation (44). On the other hand, constitutive activation of NFB in IB knockout mice results in normal epidermal development and differentiation but a widespread, lethal dermatitis in the first few days of life (45).
TNF␣ as well as other extracellular stimuli activate C/EBP␤ (46 -48). The mechanisms that activate C/EBP␤ have not been fully characterized. C/EBP␤, also known as NF-IL6 or LAP, interacts with many other transcription factors, such as the RB protein, the glucocorticoid receptor, Myc and, importantly for our studies, with AP1 and NFB (42, 49 -54). In epidermis the C/EBP proteins are differentially expressed during differentiation (55,56). Whereas knockout mice lacking C/EBP␤ have no cutaneous phenotype (57), overexpression of C/EBP␤ in keratinocytes causes growth arrest and induction of early differentiation markers (58).
To determine the roles of TNF␣-activated transcription factors in regulating K6b keratin gene expression, we have used the clone containing the promoter of the human K6b keratin gene (59,60). The promoter contains several sites that bind transcription factors responsive to extracellular stimuli (5, 59, 60). By using transfection experiments, gel shifts, and foot-printing, we have mapped the TNF␣-responsive element. We determined that both NFB and C/EBP␤ act through the same DNA sequence. Only C/EBP␤ binds this DNA directly and NFB does not. By using specific inhibitors and antisense oligonucleotides, we have shown that both NFB and C/EBP␤ are essential for the regulation by TNF␣, and we propose that a complex containing NFB and C/EBP␤ binds the K6b promoter through the C/EBP␤ DNA binding domain to convey the TNF␣ signal. Finally, we physically separated the DNA element responsive to TNF␣, NFB, and C/EBP␤ from the element responsive to EGF and AP1, thus showing that the inflammation and hyperproliferation in keratinocytes are distinct and independent processes.

EXPERIMENTAL PROCEDURES
Organ Culture Explants of Normal Human Skin-Pieces of normal human skin were obtained immediately after surgery. They were cut into pieces approximately 5 mm 3 and incubated in keratinocyte basal medium, KBM, (keratinocyte-SFM, Life Technologies, Inc.) with or without TNF␣ (50 ng/ml, Intergen), in a humidified incubator at 37°C for 24 h. Generally, we use 24-well culture dishes with up to 5 pieces in the same well and enough medium to just cover the explants. The explants were mounted in tissue Tec OCT compound (Sakura Finetek) and frozen. Sections, 4 -6 m thick, were obtained with a cryostat (Miles Laboratories), fixed with methanol/acetone for 10 min, incubated with anti-keratin K6 antibody (Progen Biotechnik GMBH) at 4°C overnight, treated with peroxidase-conjugated anti-mouse IgG secondary antibody (Vecstatin ABC-mouse IgG kit from Vector Laboratories), at room temperature for 1 h, incubated with ABC complex (Vector Laboratories) at room temperature for 1 h, and treated with 3,3Ј-diaminobenzidine-tetrahydrochloride (Dojindo Corp.) and 0.01% H 2 O 2 in Tris, pH 7.6, for 2 min. The samples were observed and photographed under the light microscope (Microphot-FXA, Nikon). Additional antibodies used were from Monosan, Uden Holland; antibodies specific for keratins K5, K8, K10, and K18 were from Progen, Heidelberg, Germany; antibodies specific for keratins K19 and K17 were from Neomarkers, Freemont, CA; and antibodies specific for K14 and for NFB and C/EBP␤ were from Santa Cruz Biotechnology.
RT-PCR from Explant Tissue-Explanted skin samples were incubated with or without TNF␣ for 16 h and harvested, and total RNAs were isolated utilizing RNeasy RNA extraction kit from Qiagen (Santa Clarita, CA). Between 1 and 15 g of RNA were subjected to RT-PCR, with Access RT-PCR system from Promega (Madison, WI). By optimizing the number of cycles (30 cycles) and application amount of total RNA (1-9 g), we achieved linear correlation between the amount of RNA added and the density of bands. We used commercial primers for glyceraldehyde-3-phosphate dehydrogenase (CLONTECH, Palo Alto, CA), and the K6 keratin primers are given in Table I. The PCR products were subjected to agarose-gel electrophoresis, visualized with ethidium bromide (Sigma) with a transilluminator from Ultraviolet Products (Upland, CA), and photographed with a photographing unit from Polaroid (Germany). The densities of bands were quantified by utilizing an image scanner (GT-9000 from Epson, Tokyo, Japan).
Immunofluorescence of Cultured Keratinocytes-Human epidermal keratinocytes in the fourth passage were plated on glass coverslips and grown for 24 h in KBM. The cells were then treated with TNF␣, washed twice with phosphate-buffered saline, and then fixed and methanol/ acetone (1:1) for 5 min. The coverslips were stained with NFB-and C/EBP␤-specific antibodies. As the secondary antibodies we used antimouse immunoglobulin G-fluorescein isothiocyanate conjugate absorbed with human serum proteins or anti-rabbit immunoglobulin Gfluorescence isothiocyanate conjugate absorbed with human serum proteins (both from Sigma).
Additional K6b promoter constructs were prepared by PCR with Thermus aquaticus DNA polymerase under conditions suggested by the manufacturer (Perkin-Elmer). All DNA primers, including the phosphorothioate-modified ones used in antisense experiments, were either synthesized on a Amersham Pharmacia Biotech Gene-Plus Synthesizer or provided by the Kaplan Comprehensive Cancer Center Core Facility.
They are listed in Table I. To create the deletions of the K6b promoter we used PCR with K6CAT as a template, a common proximal primer starting just upstream of the ATG translation initiation codon and a series of nested distal oligonucleotides (Table I). To create point mutations in the responsive element, we performed a two-round PCR mutagenesis procedure (65).
To clone the responsive element into a heterologous vector, we amplified the DNA using PCR and cloned it into the enhancer trap TK-CAT vector (Promega). The CAT activity of the resulting plasmid was compared with that of control plasmid TK-CAT (Promega). The sequences of the DNAs inserts were confirmed by the Kaplan Comprehensive Cancer Center Core Facility, using the dideoxy plasmid sequencing method. All DNAs used in transfections were purified using the Magic Megapreps DNA purification system (Promega).
Cell Growth and Transfection-Normal epidermal keratinocytes from human foreskin were a generous gift from Dr. M. Simon. The cultures were initiated using 3T3 feeder layers as described (66,67) and then frozen in liquid N 2 until used. Once thawed, the keratinocytes were grown without feeder cells in defined serum-free keratinocyte growth medium, KGM, supplemented with bovine pituitary extract epidermal growth factor, insulin, thyroid hormone, and hydrocortisone (keratinocyte-SFM, Life Technologies, Inc.). Cells were expanded through two 1:4 passages before transfection and transfected at ϳ80% confluence. Transfections using Polybrene with Me 2 SO shock were performed as described previously (68). Each transfection contained either 10 g/dish of K6CAT or 15 g/dish of deletion and mutant constructs as well as 3 g/dish of pRSVZ. The cells were incubated in KGM 18 h after the transfection, and then TNF␣ was added in combination with various other agents, as indicated. The cells were usually harvested 24 h later.
HeLa cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% calf serum and transfected using a modified calcium phosphate precipitation procedure (68). The medium was changed 14 -18 h after transfection, and cells were either left untreated for another 24 h or stimulated with TNF␣. The cell harvesting, disruption by repeated freeze-thaw cycles, and ␤-galactosidase assays have also been described (68). CAT protein concentration in the supernatant was measured utilizing either the functional assay (68) or a CAT enzyme-linked immunosorbent assay kit as suggested by the manufacturer (Roche Molecular Biochemicals). All CAT values were normalized for transfection efficiency by calculating the ratio of CAT activity to ␤-galactosidase in each transfected plate. Each transfection experiment was separately performed three or more times, with each data point resulting from duplicate or triplicate transfections.
Gel Shift Assay-Keratinocytes were grown to confluence in KGM; the medium was switched to KBM overnight, and then the cells were either left untreated or treated with 50 ng/ml TNF␣. At each time point cells from two 100-mm dishes were harvested by scraping, collected by centrifugation, washed with phosphate-buffered saline, and resus-pended in 100 l of buffer containing 20 mM HEPES, pH 7.8, 450 mM NaCl, 0.4 mM EDTA, 0.5 mM dithiothreitol, 25% glycerol, and 0.5 mM phenylmethylsulfonyl fluoride. The suspension was freeze-thawed in liquid N 2 three times and centrifuged at 4°C to remove debris. Approximately 5 g of protein from keratinocyte whole-cell extract was initially incubated for 15 min on ice with or without an excess of unlabeled competitor, in the presence of 1.5 g of poly(dI-dC)-nonspecific competitor (Stratagene) in a final volume of 25 l. The binding buffer contained 20 mM Tris-HCl, pH 7.6, 5 mM MgCl 2 , 100 mM NaCl, 10% glycerol, 1 mM dithiothreitol, 2% polyvinyl alcohol, and 0.1 mM EDTA. The probe was labeled using the Klenow fragment (Roche Molecular Biochemicals) and [␣-32 P]dCTP, 50 Ci (Amersham Pharmacia Biotech) per reaction, and purified by gel filtration using Sephadex G-50 columns (Chroma-Spin, CLONTECH). 32 P-Labeled oligonucleotide probe (80,000 cpm) was added, and the incubation was continued for an additional 30 min on ice. The free and the protein-bound DNAs were separated on 5% polyacrylamide gels (29:1 ϭ acrylamide:bisacrylamide). The gels were prerun for 30 min in 1ϫ TBE buffer and then run for 2-2.5 h at 125 V. The gels were transferred onto filter paper, dried, and exposed to x-ray film (X-Omat, Eastman Kodak Co.) at Ϫ70°C for 24 -48 h with screen intensifiers.
Antisense Strategies-Keratinocytes readily take up offered DNA. Specifically, short oligonucleotides can be introduced into these cells even in the absence of cationic lipid (69). We targeted the antisense oligonucleotides, in phosphorothioate form, to the sequences including and immediately upstream from the initiation codon. The oligonucleotides were stored at Ϫ70°C in water until use. Their sequences are given in Table I. Our approach was first to transfect HeLa cells with a well characterized responding construct, (NFB)3-CAT, adding the antisense oligonucleotides both to the transfected DNA and to the culture medium of transfected cells. The oligonucleotides were added to the medium immediately after the transfection and 18 h later; these are the times when we normally change the medium. The medium used with HeLa cells in antisense experiments contained 1% fetal calf serum. Usually we added 8 g of the oligonucleotide into 1 ml of Ca 3 (PO 4 ) 2 solution with the transfected DNA and, additionally, 15 mM of the oligonucleotide into the medium. Subsequently, the same concentrations and regimens were used with the K6CAT reporter.
Purification of NFB and C/EBP␤ Proteins-The plasmid expressing GST-tagged C/EBP␤ (61) was used to transform BL21(DE3) Escherichia coli (U. S Biochemical Corp.) which was grown in LB with ampicillin to A 600 of 0.8 and induced with 1 mM isopropyl-1-thio-␤-Dgalactopyranoside for 3 h. We used the GST-bulk purification kit that includes glutathione-Sepharose 4B and followed the procedures recommended by the manufacturer (Amersham Pharmacia Biotech). We prepared un-tagged protein using thrombin to remove the GST tag, but we found the tagged and the native proteins to have indistinguishable properties in gel shift and footprinting assays. The yield and purity of the proteins were assessed using standard SDS-polyacrylamide gels.
The plasmid expressing His 6 -NFB was transfected into the same bacterial host, and its expression was induced the same way. The tagged protein was purified using the Xpress purification system (Invitrogen).
We followed the manufacturer's recommendations for isolation of both the native and the denatured-renatured protein, and we found that the native protein protocol gave significantly higher yields of active protein.
The purified proteins were used in gel shift assays under same conditions used for the keratinocyte extracts, described above.
DNase I Footprinting Method-The oligonucleotide containing the C/EBP␤ binding sequence, 150 ng, was labeled in a kinase reaction using [␥-32 P]dATP. Then, 1.5 ϫ 10 6 cpm of the oligonucleotide was used in the primer extension reaction with Klenow DNA polymerase (Roche Molecular Biochemicals) and purified by elution from a 2.5% agarose gel into TE buffer, pH 8, at 4°C overnight. Two reactions were performed in parallel as follows: A ϩ G Maxam-Gilbert sequencing (using the reagents and protocols from the NEN Life Science Products sequencing kit) (70) and DNase I footprinting. For the footprinting, 25 l of the binding mix (see the gel shift protocol above), an amount of purified C/EBP␤ protein, usually 50 ng, and 50,000 cpm of the probe were incubated at 4°C. As a control, the C/EBP␤ protein is omitted from one of the samples. Then, 50 l of the solution containing 10 mM MgCl 2 and 5 mM CaCl 2 was added and incubated 1 min on ice. Next, 3 l of the 1:25 dilution of the DNase I (5 units/ml, Roche Molecular Biochemicals), which we have found optimal for our conditions, was added, and the incubation continued exactly for 1 min on ice. The reaction was stopped by adding 90 l of solution containing 20 mM EDTA, pH 8.0, 1% SDS, 0.2 M NaCl and 100 g/ml of yeast RNA. Next was phenol extraction, followed by ethanol precipitation. The pellet was resuspended in 1.4 l of 9 M urea, 1% Nonidet P-40 and after mixing, 4.6 l of formamide loading buffer (U. S. Biochemical Corp.) was added. All samples were heated at 90°C for 5 min, chilled on ice, and loaded onto a 12% sequencing polyacrylamide gel. Electrophoresis was run at 2,000 V for 2 h until the blue dye reached bottom of the gel. The gels were transferred onto filter paper, dried, and exposed to the x-ray film, as described above.

TNF␣ Activates NFB and Induces Expression of Keratin
K6b Protein in Normal Human Skin-Whereas healthy interfollicular epidermis does not contain K6 keratin, this protein is present in many inflammatory and hyperproliferative diseases, where it is induced by the growth factors and cytokines that orchestrate the inflammatory responses (14,59). EGF and TGF␣ cause keratinocytes to hyperproliferate and can induce the expression of K6b (18), but EGF and TGF␣ are not inherently proinflammatory. In contrast, TNF␣ does not cause keratinocytes to proliferate, although it is strongly proinflammatory (1). Therefore, we decided to determine whether TNF␣ could induce, in the absence of EGF/TGF␣, the expression of K6b keratin in human epidermis. Because convenient systems for analysis of the effects of TNF␣ and other growth factors and cytokines in human skin in vivo have not been described, we developed a new and elegant nearly in vivo experimental system that uses organ culture of human skin samples otherwise discarded during surgery (71)(72)(73)(74). We obtained 5-mm diameter biopsies of human skin and placed them in culture medium to which we added TNF␣. After 24 h the biopsies were frozen, sectioned, and the presence of K6 determined using specific antibodies. As a control we used a K17 keratin-specific antibody; the expression of K17 is induced by interferon ␥ but not by TNF␣ (17,75). We found that TNF␣ strongly and specifically induced the expression of K6 in the 24-h period, whereas K17 was not induced (Fig. 1). Without TNF␣, the presence of keratin K6 was detected in the perifollicular and eccrine epithelia, where K6 is normally seen (Fig. 1). There was no difference in expression of K8, K18, K5, K14, or K10 between the specimens incubated with or without TNF␣ (not shown).
We note that at the edges of the biopsy after 24 h a weak presence of K6 can be detected even in the absence of TNF␣ (not shown). This is presumably due to the release of the endogenous IL-1 by the peripheral keratinocytes damaged dur- ing the surgical procedure. IL-1 can also induce the expression of K6. 2 The expression of K6 at the edge of the biopsy is most prominent in the first suprabasal layer of keratinocytes. In contrast, K6 is present in all suprabasal layers of the TNF␣treated samples but most prominently in the granular layer, the layer of living cells most proximal to the medium that contains TNF␣.
In TNF␣-treated keratinocytes subtle phenotypic changes are observed; cells treated for 24 h with TNF␣ are flatter, swirled, less tightly packed, and whereas their nuclei are more prominent, the cell-cell boundaries are less distinct (Fig. 2).
We performed semi-quantitative RT-PCR with primers specific for K6 mRNA. Incubation of ex vivo skin samples with TNF␣ resulted in an approximately 3-fold increase of keratin K6 messenger RNA level (Fig. 3). The linearity of the assay was confirmed by quantification of the RT-PCR bands obtained with different amounts of input RNA.
We next examined the effects of TNF␣ on the activation of transcription factors in skin. Several transcription factors respond to TNF␣ and transduce its signal to the nucleus; these include NFB, AP1, and C/EBP␤ (52,53,76,77). Indeed, NFB, which when activated by TNF␣ enters the nucleus, was predominantly cytoplasmic in untreated skin samples but exclusively nuclear in the treated ones. On the other hand, C/EBP␤ was found in the nuclei of both treated and control skin explants. Both transcription factors behave similarly in culture and in vivo; C/EBP␤ is always nuclear, whereas NFB is cytoplasmic in unstimulated cells and enters the nucleus upon stimulation (not shown).
TNF␣ greatly increased the NFB DNA binding activity (Fig. 4). The activity is detectable 20 min after addition of TNF␣, peaks at 1 h, and then returns to the basal level in the next hour. This time course parallels closely the one of NFB nuclearization. Within the K6b promoter sequence we found a cluster of C/EBP sites (see below). By using this cluster as a probe in gel shift assays, we found a rapid activation of a DNA binding activity (Fig. 4). Enhanced DNA binding is observed 5 min after addition of TNF␣, peaks at 1 h, and then returns to basal level. These data suggest that addition of TNF␣ activates both NFB and C/EBP␤ transcription factors in human epidermal keratinocytes. The AP1 consensus binding activity was fairly high even in the absence of TNF␣, most likely due to the EGF in the medium, and did not change under the influence of TNF␣ (Fig. 4).
TNF␣ Activates the K6b Promoter through NFB and C/EBP␤ Transcription Factors-Because the regulation of keratin gene expression occurs at least partly at the level of transcription, we transfected keratinocytes and HeLa cells with a DNA construct that contains the K6b gene promoter driving the CAT reporter and then incubated the cells in the presence or absence of TNF␣. We found that in both cell types TNF␣ activates the K6b promoter dose-dependently (Fig. 5). The IL-8 gene promoter, used as a positive control, was similarly activated. The effect of TNF␣ is specific for K6b; promoters of several other keratin genes available were not activated by TNF␣. Although we cannot exclude the possibility that TNF␣responsive elements in the other keratin genes lie outside of the available sequences, with the exception of K17, these keratins are not associated with inflamed and proliferative conditions in skin. K17 is induced by interferon ␥ and not by TNF␣ ( Fig. 1 and Refs. 17 and 75). The results of the transfection experiments therefore confirm those obtained in vivo (Figs. 1 and 3); TNF␣ specifically and dose-dependently activates the promoter of the K6b keratin gene.
In the promoter of the K6b gene one finds consensus binding sites for NFB, C/EBP␤, and AP1 transcription factors. We have shown previously that the co-transfection of vectors expressing NFB and AP1 strongly induces the K6b promoter (60). Here we show that co-transfection of C/EBP␤ also strongly induces K6b promoter (Fig. 6). Furthermore, NFB, C/EBP␤, and AP1 synergize, providing several hundred-fold higher promoter activity when overexpressed together.
To confirm the involvement of NFB in the induction of keratin K6b promoter activity, we performed transfection assays using a vector expressing IB, an inhibitor of NFB. The co-transfection of IB suppressed the constitutive activity of the K6b promoter, the effect exactly opposite from that of NFB (see below, Fig. 7).
We used antisense oligonucleotides targeting NFB, IB, or C/EBP␤ as an alternative approach to examine the involvement of NFB and C/EBP␤ in the induction of K6b. The phos- FIG. 4. Gel shift assays of NFB, C/EBP␤, and AP1 proteins using extracts from TNF␣ treated keratinocytes. Cultured cells were treated with TNF␣ for the number of minutes, indicated above the lanes, harvested, and protein extracts prepared. These extracts were allowed to bind to the NFB consensus, a C/EBP␤-containing segment of K6b promoter or the AP1 consensus oligonucleotide (see Table I). The control lanes contained the 40-min time point with an excess of unlabeled oligonucleotide. The NFB and C/EBP␤ binding activities increase in the 1st h, whereas the AP1 binding activity does not. phorothioate oligonucleotides specific for NFB and IB mRNAs were designed to bind the initiation codon and the sequences immediately upstream, sites that commonly confer efficient antisense blocking (78). The antisense oligonucleotides were added to the transfected DNA mixture and subsequently to the medium of the transfected cells. Including the antisense DNA into the transfection mixture has the advantage of ensuring that the cells that received the transfected DNA also received the antisense oligonucleotides. A major advantage of our choice of sequences is the fact that NFB and IB have opposing effects on the reporter. The suppression of NFB synthesis, of course, should inhibit the NFB function, but the suppression of IB synthesis should enhance the NFB function because IB is an inhibitor of NFB. Therefore, the system is internally controlled. Nonspecific effects of the oligonucleotides, e.g. suppression of transcription commonly ob-served in most systems, will be equivalent in both transfected cultures, so that the two sequences serve as a control for each other. This approach has been used before in another cell type (78). As the reporter we initially used the NFB-responsive construct NFB-TK-CAT, which contains three tandem NFB sites linked to the TK-CAT responder. The AS-NFB oligonucleotide suppressed while AS-IB increased 3-fold the reporter activity (Fig. 7). As controls, we co-transfected DNA constructs overexpressing NFB and IB, which, respectively, increased and decreased the CAT levels, as expected.
When we tested the effects of the antisense oligonucleotides on the regulation of the K6b promoter we found that AS-NFB not only reduced the constitutive activity of K6CAT but also greatly inhibited its induction by TNF␣ (Fig. 7). In contrast, AS-IB allowed a substantial induction by TNF␣ and abolished the suppression by co-transfected IB. Thus the AS-IB oligonucleotide effects are antagonistic to those of AS-NFB, as expected.
Next we examined the effects of the AS-C/EBP␤ oligonucleotide and found that these were similar to the effects of AS-NFB; AS-C/EBP␤ reduced the constitutive activity of the K6b promoter and inhibited the induction by TNF␣ (Fig. 7). We note that antisense oligonucleotides have significant nonspecific effects; they can be toxic and inhibit the overall effect of TNF␣, as evidenced in our control sample, AS-NCoR. In the presence of AS-NCoR (and other unrelated oligonucleotides, not shown) both the constitutive and the TNF␣-induced activity of K6b promoter is reduced by approximately half. AS-NCoR oligonucleotide specifically blocked the effects of NCoR (79) in control experiments. 3 IB is a specific inhibitor of the NFB transcription factor, and CHOP is a specific inhibitor of the C/EBP family proteins (80,81). We therefore expected IB to inhibit specifically the 3  effects of NFB and CHOP to inhibit specifically the effects of C/EBP␤. To our surprise, co-transfecting IB abolished the activity of both NFB and C/EBP␤, and similarly, co-transfecting CHOP abolished both the C/EBP␤ and the NFB effects (Fig. 8). The effects are specific, because neither IB nor CHOP abolished the activity of AP1. IB and CHOP did, however, remove the synergistic effect between AP1 and either NFB or C/EBP␤. The simplest explanation of these results is that NFB and C/EBP␤ act in concert to activate the K6b promoter. Inhibiting either component abolishes the activation. Furthermore, both transcription factors are responsible for the induction by TNF␣ because co-transfection of either IB or CHOP abolished the induction by TNF␣ (Fig. 8).
Importantly, the AP1 transcription factor acts independently. The AP1-responsive element can be separated physically from the C/EBP␤ ϩ NFB-responsive one (60), and co-transfection of vectors expressing IB or CHOP does not abolish the AP1 effect (Fig. 8). This means that AP1, which is responsive to EGF, and the C/EBP␤ ϩ NFB, which are responsive to TNF␣, independently regulate the K6b keratin gene promoter.
Mapping the TNF␣-responsive DNA Element-In the promoter of the K6b keratin gene we found an NFB site, two AP1 sites, and a cluster of three C/EBP sites (Fig. 9A). To determine whether these sites constitute the TNF␣-responsive element, we prepared a series of deletion constructs leaving progressively shorter DNA sequences, and we transfected them into HeLa cells and into keratinocytes. The deletions that remove the NFB site and one or both of the AP1 sites were fully responsive to TNF␣ (Fig. 9B). This means that the NFB and AP1 sites do not play a role in the activation of the K6b promoter by TNF␣. In contrast, the deletion D172, which removes the most distal C/EBP site of the cluster, showed a reduced responsiveness to both TNF␣ and C/EBP␤, whereas D139, which removes all C/EBP sites, was completely nonresponsive. This suggests that the C/EBP sites are essential for TNF␣ signaling. Virtually identical responses were obtained with co-transfected vector expressing C/EBP␤ (Fig. 9B). This means that TNF␣ and C/EBP␤ work through the same DNA elements. We have shown before similar responses of the deletion constructs to NFB (60), which means that NFB works through the C/EBP sites as well (Fig. 9B). This finding supports our conclusion that both C/EBP␤ and NFB are necessary for the induction of K6 keratin expression by TNF␣.
In contrast, deletions D172 and D139 are fully responsive to AP1 (Fig. 9B (60)), which means that AP1 transcription factors play no role in the signaling by TNF␣. Parenthetically, the responsiveness to EGF parallels the responsiveness to AP1, suggesting that AP1 transcription factors convey the EGF signal (18). Thus, the TNF␣and the EGF-responsive elements in the K6b promoter DNA are distinct and independent.
To analyze the individual roles of the C/EBP sites, we performed site-directed mutagenesis. Mutations that disrupt either the proximal or the distal site reduced simultaneously the responsiveness to TNF␣, NFB, and C/EBP␤. Disrupting both sites abolished the responsiveness completely (Fig. 9C). This means that the C/EBP sites are necessary for regulation by TNF␣. To determine whether they are also sufficient for this regulation, we introduced them into a construct containing the minimal thymidine kinase gene promoter. Two constructs were prepared, the shorter, IS, received 61 bp, and the longer, IL, received 121 bp of the K6b promoter (Fig. 9A). The parent construct, TK-CAT, is not responsive to TNF␣, but both IL and IS were induced by approximately 2-3-fold (Fig. 9D). Note that the responsiveness to EGF has not been transferred with the K6b promoter sequences, which confirms that the TNF␣-re- sponsive element of the K6b keratin gene is distinct from the EGF-responsive element.
To determine which transcription factors bind to the responsive element, we prepared NFB-and C/EBP␤-tagged fusion proteins in E. coli, purified these proteins, and used them in gel shift assays. As the probe we used a synthetic 81-bp DNA oligonucleotide (K6 footprint long, Table I). We found robust binding of C/EBP␤ to the probe, but NFB did not bind (Fig.  10). The DNA binding of C/EBP␤ is specific because it could be inhibited by a C/EBP consensus oligonucleotide but not by an NFB-specific one (Fig. 10). Our current hypothesis is that NFB acts via protein-protein interaction with C/EBP␤. We tested this hypothesis using bacterially expressed, His-tagged NFB p65 protein in gel shift assays (63,82). The purified His-p65 did not bind to the K6b sequence. In the presence of both NFB and C/EBP␤ we expected to see a "supershift" of the C/EBP␤-generated band by p65, indicating a direct interaction of the two proteins bound to DNA, but we were unable to demonstrate it. Under the same conditions p65 bound to its consensus element (not shown).
To identify the exact DNA sequences where C/EBP␤ contacts keratin K6b promoter, we performed footprinting analysis. Addition of purified C/EBP␤ protein protected the C/EBP sites from cleavage by DNase I in a dose-dependent manner (Fig.  11A). Importantly, all C/EBP␤ sites are protected, which is congruent with the result from mutation analysis. The mutations were designed to affect only one potential C/EBP site, leaving the other two intact. It was therefore of interest to examine whether these mutants are still competent to bind C/EBP␤. Indeed, when we prepared the corresponding oligonucleotides and used then in footprinting experiments, we found that the M1 mutant bound C/EBP␤ in the downstream sequences, and M2 bound it in the upstream sequences. The binding to the double mutant was greatly reduced throughout the DNA (Fig. 11B).
In an attempt to correlate the in vitro DNA binding results with the in vivo effects of TNF␣, we grew HeLa cultures, starved them for 16 h, and then treated them with TNF␣, EGF, or serum. After 40 min extracts were prepared and used in footprinting experiments. This approach is significantly more difficult and less reproducible than the approach using purified proteins because the extracts contain many DNA-binding proteins, which may obscure the specific binding of C/EBP␤. However, we could show an increase of protein binding to the expected sites in the extracts prepared from TNF␣-treated cells. The effects of EGF, if any, were much weaker, whereas the addition of serum was without effect (Fig. 11C).

DISCUSSION
Cutaneous response to injury results in the release of cytokines and growth factors that are proinflammatory and cause hyperproliferation. Cytokines and growth factors often use overlapping signal transducing pathways, which results in shared effects. Here we show that the proinflammatory cytokine TNF␣ directly induces K6b keratin expression in normal human skin, describe the mechanism of this induction, and define the TNF␣-responsive regulatory element in the K6b gene promoter. The mechanism of TNF␣-dependent induction is completely separate and independent from the EGF-dependent induction of K6b expression described previously (18). Thus, the proinflammatory signals induce the expression of the very same keratin, K6b, that the usually concomitant hyperproliferative signals induce.
We demonstrated the effects of TNF␣, namely the induction of K6b expression and the activation of NFB transcription factor, both in cultured keratinocytes and in explants of human skin, a new experimental system designed to emulate in vivo conditions (9,(71)(72)(73)(74).
We have also shown that TNF␣ induces K6b at the transcriptional level, and we identified NFB and C/EBP␤ as the responsible transcription factors. Deletions and point mutations that show TNF␣, NFB, and C/EBP␤ all act at the same DNA site. The participation of both transcription factors is obligatory, neither C/EBP␤ nor NFB can act alone. This conclusion comes from experiments in which the specific inhibitors of NFB and C/EBP␤, IB or CHOP, respectively, inhibited both transcription factors and from the use of antisense oligonucleotides, which by depleting one of the transcription factor also inhibited the other. Particularly informative are the antisense oligonucleotide studies because in these overexpression of regulatory proteins is avoided.
NFB and C/EBP␤ are known to interact in regulating gene expression; however, usually both transcription factors bind DNA (83,84). The TNF␣-responsive element in the K6b gene element binds exclusively C/EBP␤. The interaction of the two FIG. 10. Gel shift of the TNF␣-responsive element with purified NFB and C/EBP␤ proteins. NFB and C/EBP␤, tagged with His6 and GST, respectively, were expressed in E. coli and columnpurified. They were then used in gel shift assays with an 81-bp oligonucleotide containing the responsive element (see Table I). Whereas C/EBP␤ binds the element, NFB does not. Only the C/EBP␤ consensus DNA competes for the binding, the NFB consensus does not. Fig. 9. Mapping the TNF␣-responsive site. A, the DNA sequence of the K6b promoter with the consensus binding sites for transcription factors marked. The end points of the deletions and of the insertions into TK-CAT are marked with triangles labeled D and I, respectively. The transcription start is at ϩ1, and the initiating methionine codon is underlined. Below is a line diagram of the promoter region, the deletion, and the insertion constructs. B, responses of the deletion constructs to TNF␣ and to co-transfected NFB, C/EBP␤, and AP1. Note the parallel between the responses to TNF␣ and C/EBP␤, and the very similar response to NFB, whereas the response to AP1 is distinct. C, the mutants of the responsive element. On the top M1 and M2 show the changes made in the K6b promoter sequence. The C/EBP sites were converted into restriction sites for easy identification. Both mutations severely diminish the responses to TNF␣ as well as to NFB and C/EBP␤. D, the responsive element can confer responsiveness to a heterologous promoter, in this case the TK promoter. Note that the EGF responsiveness has not been conferred with the K6b sequences. The lengths of the K6b DNA inserts into TK-CAT are indicated in A.
transcription factors seems to enhance the activity of promoters containing a C/EBP site and to suppress those with NFBbinding sites (43). Correspondingly, in our case NFB and C/EBP␤ synergize. One possible mechanism for this regulation invokes a complex containing NFB and C/EBP␤, a complex that has a different function from its individual components. Signals that activate only NFB, but not C/EBP␤ do not induce K6b gene. 4 NFB plays a role in preventing keratinocyte apoptosis (85), C/EBP␤ in keratinocyte differentiation, and both can be activated by proinflammatory signals. The strict reliance on both transcription factors ensures that only those proinflammatory signals that activate both NFB and C/EBP␤, such as TNF␣, induce K6b expression.
If NFB and C/EBP␤ act as a complex, we would expect synergistic induction when both are co-transfected, and indeed that is what we find. But when only one of the components is transfected we still see a very robust induction of the K6b promoter. We suggest that cells contain both the complex and the free partners and that their relative abundance is governed by the affinity coefficient and the mass action rules. We call this mechanism "abetting." The excess of the co-transfected factor recruits the endogenous partner into the complex and activates the K6b promoter. It may thus deplete the partner from the cell and make it unavailable for other promoters, those responsive to the other partner alone. This depletion is one of the explanations for the "squelching" effects seen in other systems (86,87). Abetting is, thus, the opposite side of squelching. Abetting may apply to other interacting transcription factors and to processes other that transcriptional regulation, e.g. those affecting the signal transduction pathways.
The DNA element responsive to TNF␣, NFB, and C/EBP␤ can be physically separated from the one responsive to EGF and AP1. Thus the inflammatory and the hyperproliferative signals that induce K6b expression are distinct and act independently. It is curious that the upstream consensus AP1 sites do not respond to AP1 and that the DNA sequences that do respond to AP1 have no resemblance to the consensus AP1binding site. But then, neither does NFB bind the NFBresponsive element in the K6b promoter; AP1 may also acts by abetting another transcription factor. This will be a subject for future studies.
We also describe a novel system using explants to analyze gene regulation in human skin. The explants have several advantages over other available systems. They use human 4 M. Blumenberg, manuscript in preparation.
FIG. 11. Footprinting the C/EBP␤-bound sites in the K6b promoter. A, purified C/EBP␤ protein was bound to the 81-bp fragment to protect the DNA from DNase. The two exterior lanes contain the unprotected DNA, and the three interior lanes contain increasing amounts (from left to right) of the C/EBP␤ protein. The same footprinting reactions were loaded onto the gel twice, 3 h apart, to resolve both the shorter (left) and longer fragments (right). The thin lines connect the same regions of the two loadings. Maxam-Gilbert sequencing reactions were run on the same gel to identify the protected region (not shown). The gray boxes on the sides of the gel indicate the footprints, the bases protected from DNase by C/EBP␤. On the top, the sequence of the K6b promoter is given, with the gray box indicating the protected region. B, DNA fragments corresponding to the mutants M1 and M2 were synthesized, labeled, and footprinted. The gray boxes on the sides of the gel indicate the protected areas. The bottom part of the footprint is retained in M1 and the top part in M2; both are attenuated in the double mutant M1ϩ2. WT, wild type. C, footprinting using cell extracts. HeLa cells were treated with TNF␣, EGF, serum, or left untreated (control). The left lane contained unprotected DNA. Note that similar intensities of the bands were obtained near the top of the gel, in the unprotected region, indicating similar loading efficiencies. The extract from TNF␣-treated cells binds to the DNA better than those from EGF-and serum-treated or untreated cells. material, the material that would be discarded otherwise. They are much easier and cheaper than grafts of human skin on nude mice (88 -90). The tissue maintains its architecture and differentiation. Most important, we have demonstrated that the explants respond normally to external stimuli and can be used to study and test the effects of various physical, chemical, or biological agents on skin. The system could be scaled up and used as a model to test pharmaceutical and skin care products, alleviating the need for animal experiments. A major current drawback is the time limit; the experiments must be completed within a few days, before the tissue deteriorates.
In summary, our results show that the inflammatory and the proliferative signals separately regulate the expression of K6b keratin. The inflammatory signals depends on NFB and C/EBP␤, using the C/EBP-binding sites in the K6 promoter DNA. The proliferative signals implicate AP1, which interacts with K6b DNA through an unknown mechanism.