Effects of Tumor Necrosis Factor- (cid:1) (TNF (cid:1) ) in Epidermal Keratinocytes Revealed Using Global Transcriptional Profiling*

Identification of tumor necrosis factor- (cid:1) (TNF (cid:1) ) as the key agent in inflammatory disorders, e.g. rheumatoid ar-thritis, Crohn’s disease, and psoriasis, led to TNF (cid:1) -target-ing therapies, which, although avoiding many of the side-effects of previous drugs, nonetheless causes other side-effects, including secondary infections and cancer. By controlling gene expression, TNF (cid:1) orchestrates the cutaneous responses to environmental damage and inflammation. To define TNF (cid:1) action in epidermis, we compared the transcriptional profiles of normal human keratinocytes untreated and treated with TNF (cid:1) for 1, 4, 24, and 48 h by using oligonucleotide microarrays. We found that TNF (cid:1) regulates not only immune and inflammatory responses but also tissue remodeling, cell motility, cell cycle, and apoptosis. Specifically, TNF (cid:1) regulates innate immunity and inflammation by inducing a characteristic large set of chemokines, including newly identified TNF (cid:1) targets, that attract neutrophils, macrophages, and skin-specific memory T-cells. Responding to tissue damage, TNF (cid:1) reg- ulates a set of genes for innate immunity, apoptosis, matrix proteoly- sis, etc., to eliminate pathogens and damaged cells, which is a well recognized de- structive process caused by TNF (cid:1) in inflammatory diseases and cancer. How- ever, TNF (cid:1) also orchestrates a well orga-nized biological process of recovery by reg- ulating tissue remodeling, repair, cell-cell interactions, cell-fate decision, and ECM production.

Tumor necrosis factor-␣ (TNF␣) 1 is a multifunctional cytokine that mediates inflammation, immune response, and apoptosis (1). Inappropriate production or persistent activation of TNF␣ participates in a wide spectrum of diseases, including septic shock, diabetes, cancer, graft rejection, rheumatoid ar-thritis, and Crohn's disease (2). Accumulating evidence indicates that TNF␣ also has a significant role in normal development and homeostasis of several organs. Mice deficient in TNF␣ lack germinal centers and show increased susceptibility to microbial pathogens due to incomplete inflammatory responses (3). TNFR1-mutant mice show similar abnormalities, in addition to defective formation of Peyer's patches (4). In skin, TNF␣ is the master cytokine regulator in inflammatory diseases, such as psoriasis, contact dermatitis, drug eruptions, cutaneous T-cell lymphoma, etc. (5)(6)(7). TNF␣ is found in skin after injury (8 -10) and is considered essential for angiogenesis during wound healing (11).
Agents targeting TNF␣ are now in clinical trials for treatment of inflammatory diseases, including psoriasis (12). However, our understanding of the TNF␣ action has not been sufficient to predict the full effects and side effects of such therapies because TNF␣ has a remarkable variety of functions and the anti-TNF␣ treatments can simultaneously affect many cellular processes in both abnormal and normal tissues (13).
TNF␣ is a homodimer of 157 amino acid subunits produced primarily by activated macrophages but also by other cell types, including epidermal keratinocytes (14). A low level of TNF␣ is present in the upper layer of the healthy epidermis, but its synthesis and release from keratinocytes are greatly augmented by injury, infection, UV irradiation, and contact sensitizers (8 -10, 15). Of the two distinct cell-surface receptors for TNF␣, TNFR1 and TNFR2, keratinocytes mainly express TNFR1 (16). The binding of TNF␣ to TNFR1 triggers a series of intracellular events resulting in the activation of transcription factors, including NFB, AP-1, CCAAT enhancer-binding protein ␤, and others (17), which are responsible for the induction of genes important for diverse biological processes, including cell growth and death, oncogenesis, and immune, inflammatory, and stress responses (14). TNF␣ activates the immune responses through inducing the production of additional signals, such as interleukin 1 (IL-1) and IL-8, transforming growth factor type ␤ (TGF-␤), intercellular adhesion molecule 1 (ICAM-1), etc. (18).
TNF␣ affects epidermis in cutaneous inflammatory diseases, and therefore we set out to identify and comprehensively analyze the TNF␣-regulated genes in epidermal keratinocytes by using DNA microarrays. Several reports of microarray analyses of TNF␣-regulated genes focused on cell lines or pathological conditions (19 -21). However, we decided to use primary cultures of healthy human epidermal keratinocytes and profile the transcriptional changes 1, 4, 24, and 48 h after TNF␣ treatment. The gene expression patterns were compared with the corresponding untreated control at each time point, which allowed us to describe novel responses not previously associated with TNF␣ action or cutaneous inflammation. We found that TNF␣ works fast, with many genes regulated already after 1 h.
Among the regulated genes, we found those involved in immune and inflammatory responses, and although TNF␣ was known to regulate some of these, many are newly identified targets. Unexpectedly, TNF␣ regulates genes that affect the cell motility and cytoskeleton, which activates keratinocyte migration. TNF␣ also regulates genes involved in cell cycle and apoptosis, as well as basement membrane components and matrix metalloproteases. Because these genes and processes comprise important elements of wound-healing, we predict that impaired wound-healing will be recognized as a side-effect of TNF␣-targeted therapies.
The results indicate that TNF␣ affects a wide range of processes to integrate the epidermal responses to injury and suggest that TNF␣ is responsible not only for the initiation of inflammation, but also for the subsequent repair and regenerative processes. This finding has significant implications for the development of TNF␣-targeted treatments.

EXPERIMENTAL PROCEDURES
Human Keratinocyte Cultures and Cytokine Treatment-We used the approach described before (22,23). Briefly, normal human neonatal foreskin epidermal keratinocytes were obtained from Dr. M. Simon (Living Skin Bank, Burn Unit SUNY, Stony Brook, NY). The cultures were initiated using 3T3 feeder layers and then frozen in liquid nitrogen until used. Once thawed, the keratinocytes were grown without feeder cells in a defined serum-free keratinocyte growth medium (KGM) supplemented with 2.5 ng/ml epidermal growth factor and 0.05 mg/ml bovine pituitary extract (keratinocyte-SFM, Invitrogen) at 37°C in 5% CO 2 . We avoid using serum because it can promote certain aspects of keratinocyte differentiation. The medium was replaced every 2 days, and the cells were expanded through three passages for the experiments. They were trypsinized with 0.025% trypsin, which was neutralized with 0.5 mg/ml of trypsin inhibitor. For all experiments, thirdpassage keratinocytes were used at 50 -70% confluence. These primary cultures provide a more appropriate target than immortalized, aneuploid cell lines, and by using a single large batch of cells, we avoided variability due to growth conditions. We changed the medium from KGM to keratinocyte basal medium without supplements 24 h before treatment to avoid the effects of the supplements in growth medium. Keratinocytes were treated with human recombinant TNF␣ (50 ng/ml, Sigma) and harvested by scraping at 1, 4, 24, and 48 h after treatment. To avoid effects of changes in keratinocyte physiology during their growth in cultures, each time point had a TNF␣-treated and a corresponding untreated, control sample. The entire experiment was performed twice, each time by a different experimenter.
Immunofluorescence Staining-Keratinocytes were grown on Lab-Tek chamber slides (Nunc, Rosklide, Denmark) and incubated with KGM. The medium was changed to keratinocyte basal medium 24 h before cytokine treatment, as described above. The cells were treated with TNF␣ (50 ng/ml). At several time points after treatment, the cells were rinsed with phosphate-buffered saline (PBS) and then immediately fixed with 70% methanol (NFB staining) for 10 min. After rinses with PBS, the cells were incubated with primary rabbit polyclonal antibody (Santa Cruz Biotechnology) diluted 1:200 in PBS containing 1% bovine serum albumin. After this, the cells were further incubated with a 1:200 dilution of FITC-labeled anti-IgG (Sigma) for 1 h at room temperature. The stained cells were observed under the microscope (Zeiss, Axiophot), and images were captured with a digital photo camera (DKC-5000, Sony). To visualize F-actin, we fixed the cells with 4% paraformaldehyde in PBS for 10 min at room temperature and used fluorescein-phalloidin (Molecular Probes, Eugene, OR) at the dilution of 1:40 after permeabilization of the cell membranes with 0.1% Triton X-100 in PBS.
To detect apoptotic cells, we used the DeadEnd fluorometric terminal deoxynucleotidyltransferase-mediated dUTP nick end labeling system (Promega), which measures the fragmented DNA by catalytically incorporating fluorescein-12-dUTP at 3Ј-OH DNA ends using the enzyme terminal deoxynucleotidyl transferase. Nucleus was counterstained with propidium iodide-containing mounting medium (Vector Labs). The stained cells were observed under the microscope and photographed. Time-lapse photographs were obtained at 30-min intervals with a digital camera (CoolPix, Nikon).
Preparation of Labeled cRNA and GeneChip TM Hybridization-We isolated total RNA from the cells with RNeasy kits (Qiagen) according to the manufacturer's instructions. Approximately 5-8 g of total RNA was reverse-transcribed, amplified, and labeled as described (22). Fifteen micrograms of labeled cRNA was hybridized to HGU95Av2 arrays (Affymetrix, Santa Clara, CA). Arrays were washed, stained with antibiotin streptavidin-phycoerythrin-labeled antibody, and scanned using the Agilent GeneArray TM scanner system (Hewlett-Packard).
Array Data Analysis-Generally, we used the same data analysis approach as before (22,23). The entire experiment was performed twice, which means that all microarray data points were independently obtained in duplicates. To compare data from multiple arrays, the signal of each probe array was scaled to the same target intensity value. We used Microsuite version 5.0 (Affymetrix) for data extraction. Differential expressions of transcripts were determined by calculating the fold change between the signal intensity values from the TNF␣-treated cells and corresponding control cells, using GeneChip TM data-mining software DMT3.0, Affymetrix). Duplicate experiments were compared using both a Student's t test and a Mann-Whitney non-parametric test. We did not consider either genes deemed absent from all 16 microarrays or genes expressed at levels sufficiently low that expression values remained below 50 arbitrary units in all 16 microarrays, because we consider these values insufficiently reliable. To improve reliability, we checked individually the absolute expression levels and p values among all four time points. Genes were considered regulated if both criteria were met: (i) at least one statistical test must find the gene regulated, and (ii) the average expression level difference in the duplicate experiments must be more than 2-fold relative to untreated control at any time point. 2 We developed an extensive gene annotation table describing the molecular function and biological category of the genes present on the chip. The table is based primarily on the Gene Ontology Consortium data 3 and the data by J. M. Rouillard. 4 Flow Cytometry-Approximately 5 ϫ 10 5 trypsinized keratinocytes were fixed in 70% ethanol in PBS for 2 h at 4°C. After washing in cold PBS, we stained the cells with 37.5 g/ml propidium iodide (Sigma) and 1 g/ml RNase (Ambion); cells were analyzed with FACScan (BD Biosciences).

RESULTS
The immediate effects of TNF␣ cause the activation of transcription factors, including NFB. As expected (24), the activation of NFB leads to its nuclear translocation (Fig. 1, A and B). The consequent transcriptional induction of the NFB-responsive genes results in profound morphological changes: the regular, polygonal "cobblestone" appearance of untreated keratinocytes is replaced by uneven arrangement, prominent cell-cell boundaries, and irregularly shaped cells 48 h after the TNF␣ treatment ( Fig. 1

, C and D). Few cells showed degenerative structures indicating cell death.
Microarray Data Mining-To obtain a comprehensive picture of the transcriptional changes caused by TNF␣, we grew primary cultures of untransformed, normal human epidermal keratinocytes and treated them with 50 ng/ml of TNF␣, which was followed by RNA isolation for microarray experiments. Of the ϳ12,000 genes present on the chips, some 7,300 were found expressed in keratinocytes; of these, some 230 genes were regulated by TNF␣, according to our criteria. Of the 230 genes, 183 were induced (80%), 43 were suppressed, and 4 were both up-regulated and down-regulated at different time points.
We used several well established data mining tools to analyze the microarray data. The scatter plots (in Fig. 2, A and B) representing log-transformed expression values of TNF␣treated keratinocytes (ordinate) versus untreated corresponding controls (abscissa) show the distribution of expressed genes during the experiment, and indicate that the number of regulated genes increased with the time lapsed. TNF␣ works quickly: more than 30 genes were induced at 1 h. Another feature of TNF␣ regulation is that the number of induced genes is always much larger than that of the suppressed ones. In this, TNF␣ is quite different from e.g. interferon-␥ or UV light, which induce and suppress roughly equal numbers of genes (22,23). Using a clustering algorithm, we compared the regulated genes at each time point after TNF␣ stimulation and found that the 24 and 48 h time points were very similar to each other, forming a "late wave" of regulated genes and suggesting a prolonged and sustained effect of TNF␣ (Fig. 2C). We compiled lists of all regulated genes at each time point and estimated the statistical probability, lod scores, that these lists contain overlapping genes. The matrix of lod scores indicates a significant overlap of the induced genes at all four time points and, similarly, an overlap of the repressed genes (Fig. 2D). Particularly similar are the lists of genes regulated at the 24-and 48-h time points, confirming the existence of the late wave.
To confirm independently the array results, we performed Northern blot analysis of eight representative genes (Fig. 3). We chose both up-and down-regulated genes and both earlyand late-regulated ones. To our satisfaction, all of the genes tested showed equivalent regulation in the two methods.
Functional Characterization of the TNF␣-regulated Genes-To understand the cellular processes affected in keratinocytes by TNF␣ stimulation, we arranged the regulated genes in a hierarchical table of their cellular functions (Fig. 4). The first level of hierarchy indicates the major function or category of each gene; the second integrates these functions into cellular tasks or activities; and the third assigns them to a global process affected by TNF␣. Below, we focus on the specific sets of TNF␣-regulated genes, those that are involved in the innate immunity, inflammation, tissue repair and remodeling, cytoskeletal rearrangements, epidermal differentiation, cell-cycle regulation, and apoptosis.
Effects upon Tissue Remodeling, Cytoskeleton, and Cell Migration-The effects of TNF␣ upon tissue remodeling, cytoskeleton, and cell migration have not been extensively studied. The reconstruction of the extracellular matrix (ECM) might be an important factor for tissue repair during and after inflammation. The induction of matrix metalloproteinase (MMP)-9 was particularly dramatic (see Fig. 3). MMP9 is the major gelatinase able to degrade collagen IV, and it is actively involved in tissue degradation, wound healing, and tumor metastasis (25). Surprisingly, it appears that MMP inhibitor TIMP3 was suppressed by TNF␣, which could result in a synergistic facilitation of tissue degradation by the MMPs in vivo. In addition, the induction of laminins LAMA3, LAMB3, and LAMC2, as well as HSPG2, a major heparan sulfate proteoglycan of basement membranes with adhesive and growth regulatory properties, strongly suggests an in vivo role for TNF␣ in the production of basement membrane (26). Fibril-associated collagen 16A1 and mucin-related membrane protein TIA-2 were also induced by TNF␣. TNF␣ regulates the coagulation system through the induction of plasminogen activator and its receptor (PLAU, PLAUR) and serum amyloid A (SAA1).
TNF␣ induced both adhesion and cytoskeleton-associated proteins in keratinocytes, suggesting a characteristic regulation of cell attachment and migration. Specifically, certain integrins (ITGA5, ITGAV, and ITGB6) and actin-related genes (EFNA1, CEP4, MLP, NCK1, and ARHE) were highly regulated by TNF␣. Integrins are transmembrane receptors that mediate the dynamic linkages between the cytoskeleton and the extracellular matrix, as well as transduce signals to and from the cell interior (27). Additionally, TNF␣ induced several adhesion molecules (NINJ1, HXB, and HEF1). This is the first report of NINJ1 induction by TNF␣; this induction peaks at 4 h after TNF␣ stimulation, indicating a rapid regulation (also shown in Fig. 3).
Interestingly, TNF␣ induced many proteinase inhibitors (HE4, CST6, PI3, SLPI, SPINK5, SERPINB1, SERPINB8, SERPINB2, SERPINB3), perhaps to protect keratinocytes from excess inflammation. Additional proteinase inhibitors, such as SPINK5, may have a role in epidermal barrier function (28) and protease inhibitor 3 (PI3) is an epithelial host-defense protein against microbial infections (29). Taken together, these TNF␣-induced changes in expression of integrins, adhesion molecules, and actin-related genes imply profound changes of the organization of the actin cytoskeleton. Indeed, we found major differences in the actin skeleton: in the treated cells, actin filaments assembled into perinuclear cages and extended into filopodia and lamellipodia. In contrast to the tight packing of control keratinocytes, the TNF␣-treated ones withdrew from their neighbors and spread out (Fig. 5A).
We expected that the cytoskeletal changes caused by TNF␣ would affect cell behavior, in particular, cell migration and cell-cell interactions. To examine the effect of TNF␣ on keratinocyte migration, we performed real-time recording of cell cultures. Fig. 5B shows the comparison of the growing pattern of keratinocyte cultures with or without TNF␣. Whereas the untreated keratinocytes simply fill out the empty spaces, the TNF␣-treated keratinocytes extensively migrate and completely change the shapes of the empty spaces by moving in and out of them. This suggests that TNF␣ stimulates the motility of keratinocytes. We traced the migration of keratinocytes using images from real-time recordings of cultured keratinocytes during the 8 h after TNF␣ treatment. We found that the TNF␣treated keratinocytes migrate much more than the untreated ones, they move in random directions, and they form many short-lived cell-to-cell contacts (Fig. 5C). Therefore, the TNF␣induced changes of expression of cytoskeleton-associated proteins significantly enhance keratinocyte motility, an essential component of the wound-healing process not previously associated with TNF␣.
Control of Cell Growth and Cell Death-Because TNF␣ only mildly affects the cell proliferation rate in culture (see below), its effects upon growth and apoptosis have remained largely unknown. However, our data clearly show that TNF␣ profoundly affects these processes. TNF␣ specifically affects the cell cycle of keratinocytes. The G 0 /G 1 switch gene (G 0 S 2 ), G 1 /S block genes (BTG2, BTG3), and GADD45A were induced, whereas the proliferation-associated proteins (CDC25C, RRM2, MCM3) were suppressed. This suggests that TNF␣ may maintain the keratinocyte cell-cycle in the G 1 phase. We confirmed the inhibitory effect of TNF␣ on cell proliferation by measuring it directly (Fig. 6A) and performing fluorescenceactivated cell sorter (FACS) analysis, which demonstrated a significant depletion of the cells in the S phase of the cell cycle (Fig. 6B).
TNF␣ induces cell death in some tumor cell lines (30); however, our cultures did not show prominent cell death after the TNF␣ treatment. Microarray analysis revealed that TNF␣ induced both pro-apoptotic (TNFSF10, BIK, BID) and anti-apoptotic genes (TRAF1, CFLAR, BIRC3, PSEN1), suggesting a charged balance for the upcoming cell-fate decision of survival or death. As expected, TNF␣ strongly induced two major TNF␣induced proteins, TNFAIP2 and TNFAIP3, which inhibit the NFB signaling without sensitizing the cells to apoptosis (31), implying an influence of these genes on the cell-fate control. Tumor suppressors may also affect the cell cycle and regulate apoptosis. TNF␣ induced the promyelocytic leukemia oncogene (PML) (32).
Taken together, these TNF␣-mediated changes of gene expression have profound effects upon the cell fate, i.e. differentiation, cell growth, and cell death, by inducing the expression of differentiation markers, proteins that block keratinocytes in the G 1 phase, pro-and anti-apoptotic proteins.
Transcription and Signaling Pathways-TNF␣ induces the expression of many transcription factors. Interestingly, TNF␣ seems to induce both the NFB subunits and the NFB inhibitory proteins. Nef-associated factor 1 (NAF1), NFKBIA (IB␣), and NFKBIE (IB⑀) have inhibitory effects upon NFB, whereas NFB1 (p105/p50), NFB2 (p49/p100), RELA (p65), and RELB are the components of NFB. Two interferon regulatory factor (IRF) family (IRF1, IRF5) and two Sry HMG box (SOX) family (SOX4, SOX9) proteins were induced, whereas MYC, a multifunctional transcriptional factor, was suppressed. The induction of IRF1 and SOX4 were especially remarkable, because IRF1 regulates cell growth and apoptosis (33), whereas SOX4 is required for the development of lymphocytes and thymocytes (34). Although SOX transcription factors perform a variety of important roles in vertebrate development, little is known about their function in adult tissues. We suspect that the induction of SOX4 and SOX9 by TNF␣ has an important role in epidermal tissue repair and morphogenesis.
Additional TNF␣-induced transcription factors and transcription regulators include MBOX1, HOXA9, MAFF, HDAC7B, HIVEP1, and ZNF267. MAFF is a stress-responsive transcription inducer (35). Histone deacetylase alters chromosome structure and affects transcription factor access to DNA (36). Thus, many transcription factors and regulators are mobilized for the immune response, cell-cell interaction, and cellfate control.
The Time Course of the TNF␣ Effects-The time course analysis of gene expression pattern reflects the sequence of the biological events responding to TNF␣ stimulation. We estimated the weighted earliest and median times of regulation for each functional category of regulated genes (Fig. 7). Judging from these results, the earliest effects of TNF␣ seem to assemble the immunocompetent cells at the site of the wound, and then initiate the immune and inflammatory responses. Subsequently, TNF␣ modulates adhesion, motility, and cell-cell interactions, which are followed by the induction of matrix metalloproteases that participate in ECM degradation. Finally, TNF␣ influences the cell fate, i.e. survival or cell death, and the rebuilding of the extracellular matrix. DISCUSSION TNF␣ is known as the principal regulator of diverse inflammatory and immune processes in human skin diseases, but the breadth of its action is not known. We have described here the global gene expression changes in normal human epidermal keratinocytes treated with TNF␣. We have identified comprehensively the set of TNF␣-regulated genes that play significant roles in skin innate immunity and inflammation. Furthermore, we have described novel functions of TNF␣ and discovered a wide spectrum of genes affecting many additional cellular processes that have not been associated previously with TNF␣ action or cutaneous inflammation. The result obtained with cultured keratinocytes may not be entirely reproduced in inflammatory skin diseases, although similar processes are likely under TNF␣ control in vivo as well. These processes include tissue repair and ECM remodeling, cytoskeletal changes, cell migration, keratinocyte differentiation, and cell-fate control. Our results describe fully an orchestrated program of the epidermal transcriptional responses controlled by TNF␣ in response to skin injury and inflammation. Although certain aspects of the TNF␣ responses can be damaging and destructive, others are clearly beneficial, and TNF␣-targeting therapies must take both into account.
Specifically, TNF␣, as the regulator of innate immunity and inflammation, induces specialized chemokines to attract neutrophils and macrophages to the skin in the early phase of inflammation. These results explain the pathological findings in acute bacterial infection, where the increased production of TNF␣ and massive infiltrations of phagocytic cells can be seen. Interestingly, TNF␣ also induced CCL27, a skin-specific memory T-cell attractant, suggesting the participation of TNF␣ in fixed drug eruption and other T-cell-mediated skin inflammation processes, such as psoriasis and atopic and allergic contact dermatitis (37,38). We show here that TNF␣ also induces the expression of antigen presentation-related molecules in keratinocytes.
In addition to initiating innate immunity and inflammation processes, TNF␣ also contributes to the repair of damaged skin. For example, TNF␣-induced angiogenesis is considered essential for wound healing, and TGF␤ has an important role in wound repair, whereas TGF␣ is mitogenic after skin injury (11,9,39,40). TNF␣ induces plasminogen, which is required for the normal repair of skin wounds and plays a central role in cancer invasion, and SAA1, an acute-phase reactant modulating platelet adhesion at injury site, which initiates coagulation via circulating micro-vesicles and platelets (41)(42)(43). This implies that, on the one hand, TNF␣ triggers the initiation of inflammation, but on the other hand, provides effective means for the attenuation of inflammation and its resolution as well.
MMP9 was among the molecules most dramatically induced by TNF␣, possibly causing tissue degradation. Excessive MMP9 activity is associated with non-healing chronic wounds (44), and deficient levels of MMP9 were found in hypertrophic scars where too much collagen is deposited (45). Additionally, MMP9 may be responsible for detaching the basal keratinocytes from the basement membrane, promoting their migration to cover the exposed connective tissue (46). Apparently, the TNF␣-regulated MMP9 plays a role both in tissue degradation and in repair from damage.
Activated macrophages and keratinocytes produce TNF␣ and subsequently, TNF␣ induces production of secondary proinflammatory cytokines. However, over-production of these inflammatory molecules may exaggerate the inflammation and potentially lead to life-threatening conditions, such as septic shock. Interestingly, we find that TNF␣ also provides mechanisms for negative feedback. For example, TNF␣ induced negative regulators of NFB pathway and anti-inflammatory molecules, including NAF1, NFKBIA, NFKBIE, SSI-1, TNFAIP3, SOD2, and many proteinase inhibitors.
We recognize that some of the TNF␣-regulated molecules have multiple functions, thus confounding simple interpretations. One example is ALOX: its metabolites are important in inflammatory skin diseases, such as psoriasis (47), but have also been involved in the establishment of the epidermal lipid barrier (48). TNF␣ regulates two antithetical processes, adhesion and migration, by inducing integrins and proteins that affect both actin polymerization and disassembly. The coordinated response is important because cell migration requires both a formation of cell-matrix adhesions at the leading edge and their dissolution at the rear end of the moving cell (49). In addition to integrins, TNF␣ induced actin-binding proteins, G-proteins, protein kinases, and phosphatases that affect the cytoskeletal organization (50). Thus, TNF␣ regulates both extracellular and cytoskeletal proteins, causing their rearrangements, which are essential for the keratinocyte attachment and migration.
TNF␣ induces NINJ1, which was originally identified as an adhesion molecule induced by nerve injury (51). CD47 initiates G-protein signaling and modulates cell adhesion and migration through its association with integrins (52). Ephrin A1 is a secreted ligand with a significant role in inflammatory angiogenesis and in actin re-organization through its receptor and Rho proteins as well as a role in the inflammatory angiogenesis Responding to tissue damage, TNF␣ regulates a set of genes for innate immunity, inflammation, apoptosis, matrix proteolysis, etc., to eliminate pathogens and damaged cells, which is a well recognized destructive process caused by TNF␣ in inflammatory diseases and cancer. However, TNF␣ also orchestrates a well organized biological process of recovery by regulating tissue remodeling, repair, cell-cell interactions, migration, cell-fate decision, and ECM production. induced by TNF␣ (53,54). CEP4 is a member of the CDC42binding protein family and regulates the organization of the actin cytoskeleton through Rho-GTPase, leading to cell shape changes (55). NCK1 has been suggested to function in the organization of the lamellipodia actin network (56). Smoothelin (SMTN) contains an actin localization sequence similar to troponin T (57). Little is known about the function of MLP, a protein that binds calmodulin and filamentous actin (58). The connection of some of the described proteins with neuron guidance (and, extrapolating, to keratinocyte migration path guidance) is very intriguing.
TNF␣ induces both pro-(BIK, BID) and anti-apoptotic proteins (TRAF1, CFLAR). We did not detect extensive apoptosis in the TNF␣-treated cells. We suggest that TNF␣ prepares keratinocytes to respond quickly to additional signals: if these promote apoptosis, the cells will die; if they promote survival, apoptosis will be inhibited.
Likewise, it was reported that TNF␣ suppresses the growth of human keratinocytes (59,60). We found that TNF␣ induced BTG2 protein, which suppresses the transcription of cyclin D1, leading to inhibition of the cell cycle at the G 1 phase (61), and GADD45A, which has also been implicated in the G 1 arrest (62). Thus, our microarray analysis suggests more precisely that the TNF␣-treated keratinocytes arrest in the G 1 phase of the cell cycle. We speculate that the G 1 arrest is required for the upcoming cell fate decision of keratinocytes, i.e. for a survival phase to be followed by differentiation, proliferation, or apoptosis.
There is a common thread in the effects of TNF␣ upon apoptosis and cell cycle: TNF␣ induces proteins that block keratinocytes in the G 1 phase, pro-and anti-apoptotic proteins, without fully blocking the cell cycle, initiating apoptosis, or preventing it. These effects have remained undetected before because, rather than causing full cell-cycle arrest or widespread apoptosis, TNF␣ in fact sensitizes the cells to additional stimuli that will direct the cells toward or away from these outcomes. We suggest that TNF␣ primes the keratinocytes to react quickly and efficiently to such additional extracellular stimuli. The additional stimuli, which could be Ca 2ϩ , growth factors, cytokines, hormones, or vitamins, are different in psoriasis, for example, than in atopic dermatitis or in wound healing, and they determine whether the fate of keratinocytes will be proliferation or apoptosis. This hypothesis will have to be proven directly in our future experiments.
Two curious aspects of TNF␣ regulation are a very rapid regulation of many genes and a preponderance of induced genes over the suppressed ones. In comparison, the responses of keratinocytes to UV light or interferon-␥ were much slower, with just a handful of genes regulated in the first hour (22,23). Apparently, keratinocytes must respond very quickly to TNF␣, because TNF␣ may signal a more grave and immediate danger, e.g. from infection, than either UV light, which principally damages DNA, or interferon-␥, which indicates that activated lymphocytes are present. The quickest response to TNF␣ is the induction of chemokines and cytokines that alert the organism to the damage, as well as of transcription factors that effect subsequent transcriptional changes. Cytoskeletal rearrangements occur later, while the cell fate determining changes are delayed even more (Fig. 7).
The signal transduction mechanisms responsive to TNF␣ activate AP1, NFB, CEBP/␤, and other transcription factors, all of which can both induce and suppress transcription of particular genes (63); it is unclear, therefore, why so many more genes are induced than repressed by TNF␣. UV light and TNF␣ activates an overlapping set of transcription factors, whereas interferon-␥ works primarily through signal transduc-ers and activators of transcription 1 (STAT1), which is, as a rule, a transcriptional activator and not a repressor (64). Therefore, additional comparative studies will be necessary to describe the details of the signal transduction and transcription factor activation in response to TNF␣ and other extracellular signals.
Anti-TNF␣ therapy has been proposed for the treatment of many inflammatory diseases (65,65). However, over-inhibiting the TNF␣ signals has a potential risk of causing delayed wound healing, secondary infections, and cancer due to suppression of the immune and inflammatory processes and defects of proper tissue repair. In fact, tuberculosis and carcinogenesis have been reported as side effects after long-term anti-TNF␣ therapy (66). Anti-TNF␣ treatments, therefore, must be finely tuned for each individual case. We believe that a comprehensive knowledge about the broad spectrum of gene expression regulated by TNF␣, which has been shown here, will contribute to make these treatments both more effective and safer.
In summary, this work presents a complete description of the transcriptional changes caused by TNF␣ in epidermal keratinocytes and we advance two significant and novel concepts: (i) TNF␣ is essential not just for the initial inflammation and response to injury, but also for the subsequent repair and recovery phase, and (ii) TNF␣ primes the keratinocytes to anticipate additional extracellular stimuli and then to react quickly and efficiently through proliferation or apoptosis. TNF␣ activates a series of immune responses and triggers inflammation, which is followed by tissue degradation and repair (Fig. 7). Keratinocytes migrate to repair damaged epidermis, contacting with and detaching from each other and the basement membrane. Some keratinocytes will die by apoptosis as a result of the inflammatory response, whereas some will survive on the newly regenerated basement membrane and proliferate, depending upon additional factors. Keratinocytes must also initiate differentiation to produce a stronger stratified epidermal shield against further tissue damage. These results demonstrate a unique aspect of TNF␣ as the key organizer of responses to skin injury, regulating a wide scope of biological processes ranging from immune response and inflammation to cell migration, epidermal differentiation, and tissue repair. We find it amazing that the regulation of gene expression for of all these processes can be achieved by a single cytokine, TNF␣, in a single cell type, the epidermal keratinocytes.