Persistent Tumor Necrosis Factor Signaling in Normal Human Fibroblasts Prevents the Complete Resynthesis of IκB-α*

Transcription factor NF-κB is normally sequestered in the cytoplasm, complexed with IκB inhibitory proteins. Tumor necrosis factor (TNF) and interleukin-1 induce IκB-α phosphorylation, leading to IκB-α degradation and translocation of NF-κB to the nucleus where it activates genes important in inflammatory and immune responses. TNF and interleukin-1 actions are typically terminated by desensitization, and IκB-α reappearance normally occurs within 30–60 min. We found that in normal human FS-4 fibroblasts maintained in the presence of TNF, IκB-α protein failed to return to base-line levels for up to 15 h. Removal of TNF at any time during the 15-h period resulted in complete IκB-α resynthesis, suggesting that IκB-α reappearance was prevented by continued TNF signaling. Long term exposure of FS-4 fibroblasts to TNF led to a persistent presence of IκB-α mRNA, sustained IκB kinase activation, continuous proteasome-mediated degradation of IκB-α, and sustained nuclear localization of NF-κB. Continuous exposure of FS-4 cells to TNF did not lead to a sustained activation of p38 or ERK mitogen-activated protein kinases, suggesting that not all TNF-induced signaling pathways are persistently activated. These findings challenge the notion that all cytokine-mediated signals are rapidly terminated by desensitization and illustrate the need to elucidate the process of deactivation of TNF-induced signaling.

The transcription factor NF-B 1 is important in the regulation of genes involved in the immune and inflammatory responses, including genes encoding inflammatory cytokines (e.g. TNF, IL-1, IL-6, and IL-8), cell adhesion molecules (e.g. ICAM-1 and E-selectin), acute phase proteins, and many other proteins participating in host defenses, e.g. inducible nitric oxide synthase, major histocompatibility complex class I, and major histocompatibility complex class II (reviewed in Refs. [1][2][3]. The NF-B family is comprised of several proteins, including p65/RelA, RelB, c-Rel, p50/p105, and p52/p100. These Rel family members share an ϳ300 N-terminal amino acid sequence (the Rel homology domain), involved in subunit dimerization, DNA binding, and in the interaction of NF-B proteins with members of the inhibitor B (IB) family of proteins. NF-B is normally sequestered within the cytoplasm due to its interaction with IB proteins, which bind to the NF-B Rel homology domain and mask its nuclear localization sequence. Proteins comprising the IB family include IB-␣, IB-␤, IB-␥, IB-⑀, and Bcl-3.
A variety of stimuli, including the inflammatory cytokines TNF and IL-1, and bacterial lipopolysaccharide cause the inducible phosphorylation of N-terminal serines in IB (Ser-32 and Ser-36 in the IB-␣ isoform) leading to the subsequent ubiquitination of neighboring lysines and the proteasome-mediated degradation of IB-␣ (4 -6). TNF initiates a signaling cascade that leads to IB degradation by binding to cell surface TNF receptor 1 (TNFR1) and TNF receptor 2 (TNFR2). Subsequent signaling (reviewed in Refs. 7 and 8)) occurs through the recruitment of cytosolic signaling proteins including TNF receptor-associated death domain protein (TRADD), TNF receptor-associated factor 2 (TRAF2), and receptor interacting protein (RIP), eventually leading to the activation of the IB kinase complex (IKK). The 700 -900-kDa IKK complex, comprising the IKK␣, IKK␤, and IKK␥ subunits, directly phosphorylates IB (9 -11). Unmasking of the nuclear localization sequence following proteasome-mediated degradation of IB permits NF-B translocation to the nucleus, leading to transcriptional activation of a variety of genes. One of the genes transcriptionally activated by NF-B is the gene encoding IB-␣ (12), which upon its translation in the cytoplasm returns to the nucleus to dissociate NF-B-DNA ternary complexes (13), thereby turning off transcription of NF-B-driven genes. In addition, newly synthesized IB-␣ interacts with NF-B dimers in the cytoplasm and prevents NF-B translocation to the nucleus. This autoregulatory loop serves to terminate NF-B activation so as to prevent the protracted expression of mediators of host defense and inflammation that are regulated by NF-B.
There exist other mechanisms to prevent the deleterious effects that would likely result from persistent cytokine signaling. In view of the well known acute toxicity and chronic inflammatory disorders resulting from TNF overexpression (14 -16), it is not surprising that upon their extended exposure to TNF cells often become desensitized to TNF action. The most common mechanism of desensitization involves the down-modulation of cell surface TNFR expression, which can occur by receptor-mediated endocytosis (17)(18)(19), by TNFR shedding (20 -22), or by mechanisms that have not been fully characterized (23). Alternatively, TNF signaling may be blocked at some point along the intracellular signaling cascade by factors that inhibit the association or function of signaling intermediates in the TNF pathway, such as TRAF-interacting protein (TRIP (24)) and the recently identified silencer of death domain (SODD) protein (25).
In most cells, TNF induces IB-␣ degradation within 15 min, which is followed by IB-␣ resynthesis and complete reappearance of IB-␣ protein within approximately 30 min to 2 h (12,26). Complete reappearance of IB-␣ protein commonly occurs even if cells are maintained in the continuous presence of TNF (27)(28)(29), and this has been ascribed to the previously mentioned autoregulatory NF-B loop and to cellular desensitization to TNF action. We show here that in normal human diploid FS-4 fibroblasts stimulated with TNF, IB-␣ is rapidly degraded but IB-␣ reappearance is incomplete because cells do not become desensitized to TNF signaling, and newly synthesized IB-␣ continues to be phosphorylated and degraded. Continued signaling by TNF in these cells is evidenced by persistent activation of the IKK complex, ongoing proteasome-mediated IB-␣ degradation, continued nuclear localization of p65/RelA, and the persistent presence of IB-␣ mRNA. We also demonstrate that not all TNF signaling pathways are persistently activated in FS-4 cells, as TNF did not produce a sustained activation of the ERK and p38 MAP kinases. Our results challenge the paradigm that TNF signaling (and cytokine signaling in general) is always rapidly terminated by desensitization.

EXPERIMENTAL PROCEDURES
Cell Culture-Normal human FS-4 diploid fibroblasts were cultured in Dulbecco's modified Eagle's medium (DMEM) containing 5% heatinactivated fetal bovine serum (FBS). For use in experiments cells at passage 14 were split into 10-cm plates and serum-starved for 2-5 days in DMEM containing 0.25% FBS. COS-1 African Green Monkey kidney cells were maintained in DMEM containing 10% FBS and were serumstarved for 1-2 days in DMEM, 0.5% FBS prior to stimulation.
Materials--DMEM was purchased from Life Technologies, Inc. Recombinant human TNF-␣ was a gift from Masafumi Tsujimoto of the Suntory Institute for Biomedical Research (Osaka, Japan). Recombinant human IL-1␣ was obtained from the NCI, National Institutes of Health, Bethesda, MD. The TNFR1-specific TNF mutein (Trp-32/Thr-86) was a gift of Dr. Werner Lesslauer (Hoffmann-La Roche). The TNFR2-specific TNF mutein was a gift from Dr. B. Aggarwal (The University of Texas, MD Anderson Cancer Center, Houston). The proteasome inhibitor MG132 (benzyloxycarbonyl-Leu-Leu-Leu-CHO) was obtained from Biomol Research Laboratories (Plymouth Meeting, PA). Rabbit polyclonal antibodies against IB-␣, p38, ERK, p65/RelA, and IKK␣ were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). Anti-phospho-p38 (a rabbit polyclonal antibody that detects the phosphorylated Thr-180 and Tyr-182 residues in p38) and anti-phospho-IB-␣ (which detects the phosphorylated Ser-32 residue in IB-␣) were purchased from New England Biolabs (Beverly, MA). Anti-phospho-ERK was purchased from Promega (Madison, WI). Texas Redconjugated goat anti-rabbit IgG for use in immunolocalization studies was purchased from Vector Laboratories (Burlingame, CA). Protein A-Sepharose was purchased from Zymed Laboratories Inc. (South San Francisco, CA), and protein A/G-agarose from Santa Cruz Biotechnology. A GST-IB-␣ construct encoding the N-terminal 62 amino acids of the human IB-␣ sequence was provided by Deborah Alpert (New York University School of Medicine). Paraformaldehyde, dimethyl sulfoxide, Hoechst nuclear dye, and Triton X-100 were obtained from Sigma. All radioactive isotopes, including [␣-32 P]dCTP, [␥-32 P]dATP, and 35 S-Express Protein Labeling Mix, were purchased from NEN Life Science Products.
Immunoprecipitation of IKK and Kinase Assays-Whole cell lysates were prepared as described for immunoblotting. The IB kinase (IKK) complex was immunoprecipitated with an antibody to IKK␣ (3 g) for 1-2 h at 4°C, and immunocomplexes were collected with protein A/Gagarose (1-2 h, 4°C). The agarose beads were then washed three times with Nonidet P-40 lysis buffer and once with kinase buffer (20 mM HEPES, pH 7.6, 20 mM MgCl 2 , 20 mM ␤-glycerophosphate, 10 mM sodium fluoride, 0.2 mM sodium orthovanadate, and 0.2 mM dithiothreitol). Half of the immunoprecipitated fraction was separated via SDS-PAGE and immunoblotted with the IKK␣ antibody (1:500) to ascertain that equal amounts of the IKK complex were immunoprecipitated for all treatment conditions. The remainder of the immunoprecipitated IKK complexes was then incubated with a kinase buffer mixture containing 10 M ATP, 10 Ci of [␥-32 P]ATP/sample, and 5 g of GST-IB-␣/sample for 30 min at 30°C. The kinase reaction was terminated by boiling the samples in reducing sample buffer. Samples were separated on 12% SDS-PAGE; the gel was dried, and autoradiography was performed. To quantify their intensity, autoradiography bands were subjected to densitometric analysis (NIH Image 1.61).
p65/RelA NF-B Immunolocalization Studies-Serum-starved FS-4 fibroblasts were plated on glass coverslips precoated with 0.1% gelatin (150,000 cells/coverslip) in 24-well plates. After the appropriate treatments cells were washed with PBS, fixed with 4% paraformaldehyde for 20 min at room temperature, and permeabilized with PBS, 0.2% Triton X-100 for 20 min at room temperature. The permeabilized cells were blocked in TBS, 5% bovine serum albumin (1-2 h, room temperature), and then incubated with anti-p65/RelA antibody (1 g/ml final concentration) for 1-2 h at room temperature. Cells were then washed with PBS, 0.1% Triton X-100, and incubated with a Texas Red-conjugated goat anti-rabbit IgG secondary antibody (12 g/ml final concentration) for 1-2 h at room temperature while shielded from light. Cells were next washed with PBS, 0.1% Triton X-100 and with PBS, counterstained with Hoechst nuclear dye (1 g/ml) for 10 -15 min, and washed with PBS. Coverslips were then mounted on microscope slides and examined by fluorescence microscopy.
Pulse-Chase Metabolic Labeling Experiments-Serum-starved FS-4 cells were either left untreated or stimulated with TNF as indicated. The cells were then washed and incubated for 1 or 2 h in methionineand cysteine-free labeling DMEM (Cellgro) containing 0.5 mCi of 35 S-Express Protein Labeling Mix. The labeling medium was then removed, and the cells were either immediately lysed in Nonidet P-40 lysis buffer (for control and t ϭ 0 time points) or "chased" in complete DMEM, 0.25% FBS medium for the indicated times. Cells were then lysed, and supernatants were precleared with normal rabbit serum (Santa Cruz Biotechnology) and protein A-Sepharose. IB-␣ was then immunoprecipi-tated from supernatants with a rabbit polyclonal antibody against IB-␣ (4 g/sample, 1-2 h, 4°C) and collected with protein A-Sepharose (1-2 h, 4°C). The beads were washed three times with RIPA buffer (150 mM NaCl, 1% Nonidet P-40, 0.5% deoxycholic acid, 0.4% SDS, 50 mM Tris, pH 7.5). The beads were then boiled in the presence of reducing sample buffer and separated on 10% SDS-PAGE, and the gel was dried and exposed to film. The autoradiography bands were subjected to densitometric analysis, and all values were normalized to and expressed as a fraction of the t ϭ 0 value (designated as "mean normalized arbitrary densitometric units") and plotted versus time. Logarithmic or exponential curves were fit to the data points and used to calculate estimates of the IB-␣ half-life under different treatment conditions. A fraction of the immunoprecipitate was separated by SDS-PAGE and immunoblotted with the IB-␣ antibody to ascertain that equal amounts of lysate were loaded for all treatment conditions (not shown).

Degradation and Reappearance of IB-␣ Protein in Human
Fibroblasts Stimulated with TNF and IL-1-Normal human diploid FS-4 fibroblasts were treated with either TNF or IL-1 for periods ranging from 15 min to 15 h. The cells were then lysed, and the presence of IB-␣ was determined by immunoblot analysis (Fig. 1). A 15-min exposure to either TNF or IL-1 led to the complete disappearance of IB-␣. Upon further incubation of cells in the presence of TNF or IL-1, IB-␣ gradually reappeared. However, in cells incubated with TNF, IB-␣ protein remained below the level seen in unstimulated cells. In contrast, in cells treated continuously with IL-1, the reappearance of IB-␣ protein over time was more complete, with IB-␣ reaching a level similar to that seen in unstimulated cells by about 4 h.
To explain the difference in the extent of IB-␣ reappearance in cells treated with TNF and IL-1, we considered the possibility that IL-1 might be more rapidly depleted from the culture medium than TNF. We therefore treated cells with either TNF or IL-1 for 15 h and then restimulated cells with fresh TNF or IL-1 for 15 min (Fig. 2). When fibroblasts were restimulated with the homologous ligand, there was no significant decrease in IB-␣, whereas subsequent treatment with the heterologous ligand (i.e. addition of IL-1 to TNF-treated cultures or addition of TNF to IL-1-treated cultures) led to the complete disappearance of IB-␣ in 15 min. These findings indicate that following 15 h of continuous treatment with either TNF or IL-1, FS-4 cells are refractory to restimulation with the homologous ligand but are responsive to restimulation with the heterologous ligand. These results further suggested that the complete IB-␣ reappearance in IL-1-treated cells was not due to depletion of IL-1 from the culture medium or the loss of its biological activity. As restimulation of cells with the heterologous ligand led to a rapid and complete IB-␣ degradation, it is also reasonable to conclude that the apparatus necessary for the phosphorylation, ubiquitination, and degradation of IB-␣ is intact in cells treated for 15 h with TNF or IL-1. Analysis of IB-␣ levels in cells treated for various lengths of time with both TNF and IL-1 simultaneously showed that results at the early time points resembled those seen in TNF-treated cells, whereas the 15-h result was similar to that seen in cells treated with IL-1 alone (data not shown).
We next sought to determine whether the failure of complete IB-␣ reappearance in TNF-treated cells was reversible. We therefore treated FS-4 cells with TNF for periods ranging from 15 min to 15 h. In one group of cultures, TNF was removed after the initial 15 min treatment (i.e. after IB-␣ disappearance), and the cells were then incubated in TNF-free medium for the remainder of the indicated times, whereas other groups were maintained in the continuous presence of TNF until the end of the incubation period (Fig. 3A). This experiment showed that IB-␣ protein returned to levels seen in the control cultures by about 2 h following the removal of TNF. In contrast, in the continuous presence of TNF, IB-␣ levels remained below control levels for up to 15 h.
TNF is known to signal through two cell surface receptors, with TNFR1 being the main signaling receptor (7,8). To determine whether TNFR1 accounts for the failure of complete IB-␣ reappearance in TNF-treated FS-4 cells, we repeated the preceding experiment with a TNF mutein protein that can bind to TNFR1 but not to TNFR2 (32,33). The results of this experiment (Fig. 3B) were virtually identical to the findings obtained with wild-type TNF. A TNFR2-selective mutant TNF protein failed to induce IB-␣ degradation in the FS-4 cells (not shown). Together, these data suggested that TNFR1 plays a major, if not an exclusive, role in the ability of TNF to modulate the reappearance of IB-␣ following its TNF-driven degradation.
Evidence for Persistent TNF Signaling: Steady-state IB-␣ mRNA Levels and the Effect of Proteasome Inhibition on IB-␣ Protein Levels-Our next series of experiments was designed to investigate the mechanism whereby complete IB-␣ reappearance was curtailed in the continuous presence of TNF. In one group of experiments we used Northern blot analysis to study steady-state IB-␣ mRNA levels in FS-4 cells incubated for periods ranging from 30 min to 18 h, either in the continuous presence of TNF or with TNF removed from the cultures following an initial 15-min stimulation. Peak steady-state IB-␣ mRNA levels were reached at 1 h under both continuous and "pulse treatment" conditions (Fig. 4). In the continuous presence of TNF, IB-␣ mRNA was sustained near this peak level for up to 18 h. In contrast, IB-␣ mRNA levels decreased rapidly upon TNF removal from the culture medium. This result showed that the failure of IB-␣ protein to return to control levels in the continuous presence of TNF was not due to a decreased availability of IB-␣ mRNA. In fact, our observa-FIG. 1. IB-␣ degradation and reappearance in FS-4 fibroblasts treated with TNF or IL-1. Serum-starved FS-4 fibroblasts were treated for the indicated times with TNF (20 ng/ml) or IL-1 (4 ng/ml). At the end of the treatment period, the cells were lysed in a Nonidet P-40-based lysis buffer, and supernatants were boiled in denaturing sample buffer, resolved by 10% SDS-PAGE, transferred to Immobilon-P membranes, and immunoblotted with antibody to IB-␣. Blots were incubated with goat anti-rabbit IgG conjugated to horseradish peroxidase, and bands were visualized by enhanced chemiluminescence. Ctrl, control.
FIG. 2. Effect of long term stimulation with TNF or IL-1 on subsequent TNF-or IL-1-mediated IB-␣ degradation. Serumstarved FS-4 fibroblasts were treated for 15 min or 15 h with TNF (20 ng/ml) or IL-1 (4 ng/ml). At the end of the 15-h incubation, some groups were restimulated for 15 min with either TNF (20 ng/ml) or IL-1 (4 ng/ml), as indicated. Cells were then lysed, and lysates were resolved on 10% SDS-PAGE and immunoblotted with an anti-IB-␣ antibody as described in Fig. 1. Ctrl, control. tion that IB-␣ mRNA levels were sustained in the continuous presence of TNF, but not when TNF was removed after 15 min, is the opposite of what is seen at the protein level (Fig. 3A). Increased IB-␣ mRNA levels in cells treated continually with TNF probably reflect persistent TNF signaling, with ongoing TNF-mediated NF-B activation likely to be driving the transcription of IB-␣ under these conditions. We then considered the possibility that ongoing proteasomemediated degradation of IB-␣ in cells treated continually with TNF may explain the failure of complete IB-␣ reappearance, despite the presence of higher levels of IB-␣ mRNA in such cells. FS-4 cells were stimulated with TNF for periods ranging from 5 min to 15 h and treated with the proteasome inhibitor MG132 as indicated (Fig. 5). Cell lysates were subjected to immunoblot analysis with antibodies directed against IB-␣ and N-terminally phosphorylated IB-␣. Pretreatment of cells with MG132 decreased TNF-induced IB-␣ degradation (Fig. 5,  lanes 4 and 6, upper and middle panels). Importantly, in cells incubated with TNF for 1, 4, and 15 h, MG132 treatment resulted in the appearance of the phosphorylated IB-␣ species as well as in greatly increased levels of IB-␣. These findings indicate that proteasome-mediated degradation of inducibly phosphorylated IB-␣ is likely to be important in the failure of IB-␣ to return to control levels in cells maintained in the presence of TNF.

Persistent TNF Signaling: Activation of the IB Kinase Complex and Nuclear
Localization of p65/RelA-Seeking further evidence for persistent TNF signaling, we examined the activity of the IKK complex. The IKK complex includes the IKK␣ and IKK␤ subunits and is responsible for the inducible phosphorylation of IB-␣ (3, 9 -11). FS-4 cells were incubated for periods ranging from 15 min to 15 h, either in the continuous presence of TNF or with TNF present only during the initial 15 min. The IKK complex was immunoprecipitated, incubated with GST-IB-␣ in an in vitro kinase reaction, then resolved on SDS-PAGE, and subjected to autoradiography (Fig. 6). Results from this experiment indicated that IKK was very strongly activated at 15 min after TNF stimulation. In cells maintained in the continuous presence of TNF, IKK activity decreased gradually at later times but was still detectable after 15 h. Densitometric analysis revealed that the strength of the phosphorylated band at 15 h was roughly 6% of the level observed after a 15-min stimulation. IKK-mediated IB-␣ phosphorylation decreased faster when TNF was removed after a 15-min stimulation, with no kinase activity detectable by 15 h.
To investigate further the effect of TNF stimulation on NF-B activation, we examined the subcellular localization of p65/RelA in TNF-treated cells (Fig. 7). Cells were treated with TNF for the times shown, and where indicated TNF was removed after an initial 20-min stimulation. Fixed and permeabilized cells were incubated with an antibody against p65/ RelA and a Texas Red-conjugated secondary antibody and were examined by fluorescence microscopy. Whereas p65/RelA was present in the cytoplasm of untreated cells, cells treated with TNF for 20 min, 1 h, or 15 h showed p65/RelA exclusively in the nucleus, consistent with a persistent activation of NF-B. Removal of TNF following an initial 20-min stimulation produced a more heterogeneous p65/RelA staining pattern by 15 h, with p65/RelA detectable in both the nucleus and cytoplasm.
Stability of IB-␣ Protein in Unstimulated and TNF-stimulated Cells-To analyze the stability of the IB-␣ protein in unstimulated cells, as well as the half-life of IB-␣ protein in the presence or absence of TNF, we conducted pulse-chase metabolic labeling experiments. In control, unstimulated FS-4 cells we determined the half-life of basal IB-␣ to be 3.5 h (Fig.  8A, panel 1). Two other groups of cells were initially treated with TNF for 15 min in order to deplete cells of the IB-␣ protein. Cells in these groups were then metabolically labeled with [ 35 S]methionine and -cysteine in the presence (Fig. 8A,  panel 2) or absence (Fig. 8A, panel 3) of TNF and "chased" with   FIG. 3. Effect of treatment of FS-4 cells with TNF, or a TNFR1selective binding mutant, on the degradation and reappearance of IB-␣. Serum-starved FS-4 fibroblasts were treated for the indicated times with TNF (20 ng/ml, A) or a TNF mutant protein that binds exclusively to TNFR1 (20 ng/ml, B). Where indicated by wash, the cells were treated with TNF or the TNFR1-selective binding mutant for 15 min, washed twice with PBS, replenished with fresh serum-starvation medium (DMEM, 0.25% FBS), and incubated for the remaining time at 37°C. Cells were then lysed, and lysates were resolved on 10% SDS-PAGE and immunoblotted with an anti-IB-␣ antibody as described in Fig. 1. Ctrl, control.

FIG. 4. Steady-state IB-␣ mRNA levels in FS-4 fibroblasts treated with TNF.
Serum-starved FS-4 cells were treated with TNF (20 ng/ml) for the indicated times. Where indicated, TNF was removed after a 15-min stimulation, and the cells were washed and replenished with fresh medium without TNF. At the end of the incubation period the cells were lysed, and RNA was extracted. RNA was electrophoretically separated on 1% agarose gels, transferred onto nylon membranes, and hybridized with a 32 P-labeled IB-␣ cDNA probe. Membranes were then washed and exposed to film. Membranes were later stripped and rehybridized with a glyceraldehyde-3-phosphate dehydrogenase (GAPDH) cDNA probe to control for equal RNA loading. Ctrl, control.  4 and 6). For all other TNF time points (1, 4, and 15 h), MG132 (10 M) was added to the cells following an initial 15-min stimulation with TNF (20 ng/ml), and cells were maintained in the continuous presence of both MG132 and TNF for the remainder of the period (lanes 8, 10, and 12). Cells were then lysed, and culture supernatants were separated on 10% SDS-PAGE and immunoblotted with anti-IB-␣ (upper panels), anti-phospho-IB-␣ (middle panels), or anti-ERK (lower panels) as a loading control. Bands were visualized as described in Fig. 1. cold medium either in the presence (Fig. 8A, panel 2) or absence (Fig. 8A, panel 3) of TNF. This analysis showed that in the continuous presence of TNF the stability of newly synthesized IB-␣ protein was dramatically decreased, to a half-life of 5.2 min (Fig. 8A, panel 2). When, after the initial 15 min treatment with TNF, cells were labeled and chased in the absence of TNF, the half-life of newly synthesized IB-␣ was identical to that of the pre-existing IB-␣ protein seen in control, unstimulated cells, i.e. 3.5 h (Fig. 8A, panel 3). The latter result showed that removal of TNF led to a rapid and complete cessation of TNFinduced IB-␣ degradation. We also show that in FS-4 cells exposed to TNF for 15 h, the rate of IB-␣ protein synthesis was significantly higher than in unstimulated cells (Fig. 8B, lanes 1  and 2). Moreover, we demonstrate that the rate of IB-␣ degradation was higher when TNF was present during the chase period than in the absence of TNF (compare lane 1 to lane 3 and  lane 2 to lane 4). The results shown in Fig. 8B support the conclusion that there was active TNF signaling in FS-4 cells after their 15-h incubation in the presence of TNF. These data also show that the IB-␣ degradation apparatus was intact and functional in FS-4 cells subjected to long term TNF treatment.
Continuous TNF Signaling Does Not Lead to Persistent ERK and p38 MAP Kinase Activation-We have provided evidence that persistent signaling prevents the complete reappearance of IB-␣ in cells treated continuously with TNF for up to 15 h. It was of interest to determine whether other TNF-elicited signaling pathways were persistently active in TNF-treated FS-4 cells, especially in comparison to IL-1-induced responses. Toward this end we treated FS-4 fibroblasts with TNF or IL-1 for periods ranging from 15 min to 15 h, and we examined cell lysates for the presence of phosphorylated (i.e. activated) forms of the ERK and p38 MAP kinases by immunoblot analysis (Fig.  9). Both TNF and IL-1 induced a marked increase in the phosphorylation of the p44 and p42 ERK and the p38 MAP kinases at 15 and 30 min, which was no longer observed after 1 h. Desensitization of cells to the MAP kinase-inducing activity of TNF and IL-1 was ligand-specific, because restimulation of FS-4 cells treated with TNF or IL-1 for 15 h with the heterologous cytokine produced a rapid (within 15 min) and full activation of p38 MAP kinase (data not shown). These data indicate that neither TNF nor IL-1 induced a persistent activation of the ERK and p38 MAP kinases and suggest that continuous treatment with TNF does not induce a generalized and persistent activation of all TNF-responsive signaling pathways in FS-4 fibroblasts.

DISCUSSION
Cytokines are potent intercellular signaling molecules that serve important functions in host defenses. To prevent toxic side effects cytokines are generally produced only for limited  1 and 2) or chased in cysteine-and methioninecontaining medium in the absence or presence of TNF for 1 h as indicated (lanes 3 and 4). At the times indicated cells were lysed; IB-␣ was immunoprecipitated from the lysates, and immunocomplexes were collected by centrifugation with protein A-Sepharose beads. Samples were resolved by 10% SDS-PAGE and subjected to autoradiography. A fraction of the immunoprecipitate was retained and immunoblotted with antibody to IB-␣ to control for equal loading (not shown). Curves were fitted to the data points and used to estimate the half-life of IB-␣ protein under different treatment conditions (panels B1-3). Ctrl, control.
periods, usually in response to infection or trauma. Another feature that helps to guard against their deleterious side effects is that cytokines characteristically generate signals that are of short duration. Thus, long term exposure to a cytokine often leads to a selective desensitization of cells, which terminates signal propagation generated by this cytokine. One example of such desensitization is the resistance to TNF and IL-1 action that develops in cells chronically exposed to these cytokines. In most cells, TNF and IL-1 induce IB-␣ degradation within 15 min, which is rapidly followed by IB-␣ resynthesis and reappearance of the IB-␣ protein (12,26). Such complete IB-␣ reappearance is observed even if the cells are maintained in the continuous presence of TNF or IL-1 (28,29). In contrast, our present data show that in normal human FS-4 fibroblasts stimulated with TNF, IB-␣ is rapidly degraded but IB-␣ reappearance is incomplete. Analysis of the causes of incomplete IB-␣ reappearance revealed that human FS-4 fibroblasts do not become completely desensitized to TNF signaling, such that newly synthesized IB-␣ continues to be inducibly phosphorylated and degraded as a result of TNF action. Several mechanisms are known to limit the duration and extent of cytokine signaling. The duration of NF-B activation is limited by an autoregulatory process, in which NF-B drives the transcription of its own inhibitor, IB-␣ (27). Newly synthesized IB-␣ retains NF-B in the cytoplasm and, in addition, enters the nucleus to dissociate NF-B bound to DNA in ternary complexes (13,34,35). TNF signaling is known to be curtailed by a variety of mechanisms, including the shedding of soluble TNF receptors (19 -22), a regulated process that can be mediated by the action of the TNF-␣-converting enzyme whose primary function is to proteolytically process the transmembrane precursor form of TNF to its soluble form (36). TNFR shedding decreases the availability of cell surface TNF receptors. In addition, soluble TNFR generated by this process may inhibit TNF signaling by competing with cell surface TNFR for ligand binding (37,38). TNF signaling has also been shown to be regulated by the ligand-induced down-regulation of cell surface receptors via receptor-mediated endocytosis (17)(18)(19). Finally, cytokine signal duration may be limited by interactions between TNF signaling intermediates and intracellular inhibitors. Such signaling inhibitors include TRIP and I-TRAF, shown to prevent TRAF2-mediated NF-B activation (24,39), SODD that associates with TNFR1 and in the absence of TNF stimulation blocks TRADD and TRAF2 recruitment to the re-ceptor death domain (25), and A20 which inhibits TNF and IL-1 signaling (40).
Our results show that despite the existence of numerous safeguards against persistent TNF signal transduction, signaling is not completely blunted in normal human fibroblasts exposed chronically to TNF, as evidenced by the failure of a complete reappearance of IB-␣ for up to 15 h (Fig. 1), persistence of IB-␣ mRNA levels (Fig. 4), persistent proteasomemediated IB-␣ degradation (Fig. 5), long term activation of the IKK complex (Fig. 6), and persistence of p65/RelA in the nucleus (Fig. 7). In addition, metabolic labeling experiments showed that in FS-4 cells exposed to TNF for 15 h, TNF continued to drive increased synthesis of IB-␣ protein as well as its degradation (Fig. 8B). Earlier results obtained by electrophoretic mobility shift analysis also indicated that TNF-induced NF-B activation in FS-4 cells persisted for at least 9 h (41). Preliminary evidence suggests that the nature and degree of persistent signaling depend upon the cytokine and cell type involved. When COS-1 cells were employed in an experiment similar to that shown in Fig. 1, complete reappearance of IB-␣ protein was observed by 2 h in the presence of either TNF or IL-1 (data not shown). In addition, different signaling pathways activated by a single cytokine can differ in the degree of desensitization, as evidenced by the finding that both TNF-and IL-1-induced ERK and p38 MAP kinase activation in FS-4 cells were of a short duration (Fig. 9).
The molecular mechanisms that underlie the degree of desensitization to TNF and the level of persistent signaling have not yet been addressed. It is possible that these processes are affected by endogenous TNF signaling inhibitors, such as SODD (25). Human fibroblasts may express low levels of SODD, or failure of an efficient SODD-TNFR1 interaction in these cells may permit persistent TNF signaling. Similarly, the level or functional properties of other signaling inhibitory molecules, such as TRIP (24) or A20 (40), may affect the duration of TNF signaling. In addition, it is possible that TNF induces lower levels of receptor internalization or shedding in FS-4 fibroblasts compared with other cells, thus permitting longer signal duration. Diminished TNFR shedding has recently been described in hereditary inflammatory disorders involving mutations in TNFR1 that apparently limit the release of soluble TNFR and increase cellular responsiveness to TNF (42). One unanswered question is why TNF produces persistent NF-B activation but only transient activation of members of the MAP kinase family (Fig. 9). Whereas RIP has been shown to play an essential role in TNF-mediated NF-B activation and TRAF2 is required for TNF-driven JNK and p38 MAP kinase activation, both RIP and TRAF2 associate with TRADD (43,44). The bifurcation in these two TNF-signaling pathways that begins at TRADD might determine the relative persistence of signaling via the NF-B and the MAP kinase pathways.
Our results challenge the long held view that cytokine actions are terminated by rapid desensitization. The findings are biologically relevant because in many chronic inflammatory diseases persistent TNF signaling may lead to protracted NF-B activation that induces the expression of inflammatory mediators, perpetuating a cycle of cellular activation, leukocyte migration, and coincident inflammation and tissue damage. Persistent signaling by TNF and other cytokines may be relevant in the context of inflammatory conditions such as rheumatoid arthritis, inflammatory bowel disease, multiple sclerosis, and heart failure (15,16). In addition, since TNF exerts a mitogenic effect on normal fibroblasts (45), persistent TNF signaling might also play a role in certain beneficial host responses, such as wound healing. Mechanisms responsible for the termination of cytokine signaling (and especially the failure FIG. 9. Transient activation of ERK and p38 MAP kinases in TNF-or IL-1-treated FS-4 cells. Serum-starved FS-4 fibroblasts were treated for the indicated times with TNF (20 ng/ml) or IL-1 (4 ng/ml). Thereafter the cells were lysed, and lysates were processed as described in the legend for Fig. 1. Membranes were incubated with anti-phospho-ERK (pp44 and pp42), anti-ERK (p44 and p42), anti-phospho-p38 (pp38), or anti-p38 (p38). Blots were then incubated with goat antirabbit IgG conjugated to horseradish peroxidase and detected by chemiluminescence. thereof) have not been extensively studied. Our data show that much remains to be learned about these processes.