Plasminogen activator inhibitor type 2 inhibits tumor necrosis factor alpha-induced apoptosis. Evidence for an alternate biological function.

Plasminogen activator inhibitor type 2 (PAI-2) is a serine proteinase inhibitor or serpin that is a major product of macrophages in response to endotoxin and inflammatory cytokines. We have explored the role of PAI-2 in apoptotic cell death initiated by tumor necrosis factor alpha (TNF). HeLa cells stably transfected with PAI-2 cDNA were protected from TNF-induced apoptosis, whereas cells transfected with antisense PAI-2 cDNA, a control gene, or the plasmid vector alone remained susceptible. The level of PAI-2 expressed by different HeLa cell clones was inversely correlated with their sensitivity to TNF. Loss of TNF sensitivity was not a result of loss of TNF receptor binding. In contrast, PAI-2 expression did not confer protection against apoptosis induced by ultraviolet or ionizing radiation. The serine proteinase urokinase-type plasminogen activator was not demonstrated to be the target of PAI-2 action. The P1-Arg amino acid residue of PAI-2 was determined to be required for protection, because cells expressing PAI-2 with an Ala in this position were not protected from TNF-mediated cell death. The results suggest that intracellular PAI-2 might be an important factor in regulating cell death in TNF-mediated inflammatory processes through inhibition of a proteinase involved in TNF-induced apoptosis.

Programmed cell death is conserved throughout evolution as a strategy employed by multicellular organisms to regulate a wide range of physiological processes including embryo development, tissue remodelling, immune system development and cellular responses to infection, and tumorigenesis. Cell death can be triggered by a variety of stimuli that can ultimately lead to characteristic changes in the cell, frequently resulting in chromatin fragmentation and condensation, membrane blebbing, and collapse of the nucleus, a process called apoptosis (1,2). Within the immune system, tumor necrosis factor ␣ (TNF) 1 is an important effector of programmed cell death, playing a role in immune defense against viral, bacterial, and parasitic infections with the ability to target tumor cells and virusinfected cells (3,4). The initial events involved in TNF-induced cell death occur through the 55-kDa TNF receptor (5,6) with the subsequent signal transduction events being the subject of intense investigation (7,8). However, the specific cellular death pathway(s) triggered by TNF and their relationships to other effector-induced apoptotic cell death mechanisms are not well understood.
An evolutionary approach to the elucidation of mechanisms involved in cell death has led to the finding that cell death can be mediated, at least in part, by proteinases. The cell death gene, ced-3 from Caenorhabditis elegans, and its human homologue, the cysteine proteinase interleukin 1␤ converting enzyme (ICE), cause cell death when expressed in rodent fibroblasts (9). Serine proteinases are implicated in cell death induced by TNF, in that low molecular weight serine proteinase inhibitors protect tumor cells from the cytotoxic effects of TNF (10,11). More recently, the serine proteinase inhibitor (serpin), plasminogen activator inhibitor type-2 (PAI-2), was reported to confer protection from TNF-mediated cytolysis when overexpressed in HT1080 fibrosarcoma cells (12). PAI-2 is a major product of monocytes and macrophages in response to inflammatory mediators (13,14). PAI-2 was originally described as an inhibitor of the serine proteinase, urokinase-type plasminogen activator (uPA) (15,16). uPA specifically catalyzes the hydrolysis of a single Arg-Val bond in the widely distributed zymogen, plasminogen, to form plasmin, a broad spectrum serine proteinase that is capable of hydrolyzing a number of protein substrates (17). uPA has a well established role in cellular invasion and is involved in degradation of extracellular matrices and tissue remodelling (17,18). uPA has also been reported to function as a cellular growth factor (19,20) and to participate in the activation of hepatocyte growth factor (21), but its potential role in cell death is not known. In HT1080 tumor cells, uPA and PAI-2 are synthesized constitutively (12), and it is not possible to ascertain from this model whether the presence of uPA is intrinsic to the protective effect demonstrated by overexpression of PAI-2. In the present paper, we report that expression of PAI-2 protects from TNF-induced apoptosis in HeLa cells, a cell line that does not synthesize PAI-2 or significant levels of uPA and that is sensitive to the cytotoxic effects of TNF. Protection by PAI-2 is independent of extracellular uPA activity and is likely to involve an intracellular cell death proteinase.

Cells and Cell Culture-HeLa cells (ATCC-CLL 2) and
MonoMac6 cells (22) (kindly provided by Prof. H. W. L. Ziegler-Heitbrock, University of Munich) were incubated in a 5% CO 2 and 95% air atmosphere at 37°C and cultured in RPMI 1640 medium supplemented with 10% fetal calf serum, 2 mM glutamate, 25 mM HEPES buffer, 60 g/ml penicillin G, and 100 g/ml streptomycin sulfate. Cell viability was determined by trypan blue dye exclusion. All cultures were checked routinely and determined to be mycoplasma free.
cDNA Constructs-An 1867-bp DNA fragment encoding PAI-2 cDNA was excised from pJ7/PAI-2 (23) with EcoRI, end-filled, and inserted into the end-filled HindIII site of pRcCMV (Invitrogen). Both orientations were obtained to create sense and antisense PAI-2 expression constructs. The chloramphenicol acetyltransferase (CAT) cDNA gene was obtained by polymerase chain reaction amplification of a 785-bp region of pCATBasic (Promega) using the following polymerase chain reaction primers, and the resultant DNA fragment was cloned into the HindIII site of pRcCMV: CAT primer forward, 5Ј-GGCGAAGCTTCAG-GCGTTTA-3Ј, and CAT primer reverse, 5Ј-TACGCCAAGCTTGCAT-GCCT-3Ј.
The P 1 site mutant of PAI-2 was obtained by polymerase chain reaction amplification and overlap extension following the method of Ho et al. (24). The following synthesized oligonucleotides were used to alter the nucleotides at 1186 -1192, resulting in a mutation at Arg-380 (to Ala-380), and to generate a unique Nar1 restriction enzyme site (see Fig. 6A): Primer A, 5Ј-GGCTCAGATTCTAGAACTCC-3Ј; Primer B, 5Ј-GTATTTCTAGAAATGCACATAAC-3Ј; Primer C, 5Ј-GACAGGCGC-CACTGGACATGG-3Ј; and Primer D, 5Ј-CCATGTCCAGTGGCGC-CTGT-3Ј. Primers A and B and Primers C and D, respectively, were combined to generate DNA fragments containing overlapping mutations at nucleotides 1186 -1192 in the PAI-2 cDNA gene. These DNA fragments were then combined, and Primers A and D were used to generate a 988-bp DNA fragment extending between the XbaI restriction enzyme sites at positions 790 and 1778 of PAI-2 cDNA. This DNA fragment was cloned into the corresponding XbaI sites of pJ7/PAI-2 following excision of the 988-bp DNA fragment containing the native sequence at the P 1 site. The sense orientation was identified by restriction enzyme mapping, and the DNA sequence of the mutant construct was verified by DNA sequence analysis.
Transfections and Selection of Stable Cell Lines-Approximately 1 ϫ 10 7 HeLa cells in log phase were electroporated with 20 g of plasmid DNA at 250 V, 960 microfarads in RPMI 1640 containing 10% fetal calf serum. Following culture for 16 h, transfectants were selected with 0.8 mg/ml G418 (Life Technologies, Inc.). Individual clones were maintained routinely in 0.3 mg/ml G418. Expression of the transgene was determined by Northern and immunoblot analyses.
RNA Isolation and Northern Blot Analysis-Near confluent, adherent cultures were harvested by gentle scraping and washed by resuspension in phosphate-buffered saline (PBS). Total RNA was isolated from harvested cells using the guanidinium isothiocyanate method of Chomczynski and Sacchi (25). Each RNA sample (10 g) was electrophoresed on denaturing agarose gels containing 1.1% formaldehyde, transferred to Hybond N nylon membranes (Amersham Corp.) by capillary diffusion, and fixed by UV irradiation according to the manufacturer's instructions. The membranes were hybridized essentially as described (26) and washed to a final stringency of 0.1 ϫ SSC and 0.1% SDS at 65°C. The blots were probed with a 2.0-kb HindIII-EcoRI DNA fragment encoding PAI-2 cDNA (27), the 785-bp CAT DNA fragment, or pRcCMV (for the vector alone transfectants). Plasmids containing cDNA probes for the 55-and 70-kDa TNF receptors were obtained from Dr. David Lowen (Syntex USA, Palo Alto, CA), and manganese superoxide dismutase cDNA (28) was from Prof. N. Taniguchi (Osaka Medical School and Mitsui Toatsu Chemicals, Inc.). Each DNA fragment was radiolabeled by the random priming method (29). The blots were rehybridized with a labeled human 18 S rRNA oligonucleotide as a measure of total RNA loaded in each lane (26). The blots were exposed to Kodak XK-1 film between Dupont Cronex intensifying screens at Ϫ70°C for varying times and quantitated by scanning different exposures of autoradiographs using a scanning densitometer (Molecular Dynamics) driven by ImageQuant software.
Immunoblot and ELISA Analyses-For immunoblot analyses, cells were washed three times in PBS and harvested by gentle scraping into PBS. The cells were pelleted by centrifugation and lysed quickly on ice in cold PBS containing 0.5% Triton X-100, 5 mM EDTA, and 20 g/ml of the proteinase inhibitor, 4-(2-aminoethyl)benzenesulfonylfluoride, except where indicated in the figure legends. Cellular debris was removed by centrifugation at 12,000 ϫ g for 10 min, and the resulting cell lysate stored in aliquots at Ϫ70°C. The protein concentration of each sample was determined by Bio-Rad protein assay. The solubilized proteins (40 -80 g) were separated by SDS-polyacrylamide gel electrophoresis under nonreducing conditions using a 5-15% acrylamide gradient gel, and the proteins were electrophoretically blotted onto a nitrocellulose membrane (Bio-Rad) for 16 h at 30 V. Specific antigens were detected by incubation of the membrane at room temperature for 1 h with 1 g/ml of either the anti-catalytic, anti-uPA monoclonal antibody (American Diagnostica #394), the anti-PAI-2 monoclonal antibody (American Diagnostica #3750), the anti-CAT monoclonal antibody (5 Prime 3 3 Prime Inc. #5307-310127), or the anti-bcl-2 monoclonal antibody (Oncogene Sciences #OP60). Antibody binding was visualized by ECL detection (Amersham Corp.) according to the manufacturer's instructions. Membranes were exposed to Kodak XK-1 film for various times and quantitated using a scanning densitometer (Molecular Dynamics) as described above. Quantitation of PAI-2 levels by ELISA was carried out using the Tintelize PAI-2 ELISA kit (#220220) from American Diagnostica.
TNF Cytolysis Assays-Cytolysis was quantitated by two methods: 1) crystal violet staining (modified from Ref. 12), which measures cell survival as a function of cell adherence, and 2) the tetrazolium dyebased MTT assay (30), which provides a measure of cell viability. Cells were seeded at 10 4 cells/well in triplicate in 96-well plates for 16 h prior to treatment with the following agents for 8 h in the standard assay: TNF, 0 -50 ng/ml as indicated in the text, and cycloheximide, 10 g/ml. Cycloheximide was not cytotoxic at the concentrations used within the time frame of the experiments. For assay by crystal violet staining, the cells were washed in PBS and stained with 0.2% crystal violet in 10% ethanol or 10% formaldehyde. The dye was eluted with 33% acetic acid, and the absorbance was measured at 480 nm. For measurement by MTT assay, 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide (Sigma) was added to a final concentration of 0.4 g/ml/well and incubated for 4 h. Plates were centrifuged at 800 ϫ g for 5 min, and supernatant was removed. MTT crystals were dissolved in Me 2 SO, and the absorbance was measured at 570 nm.
For the experiments involving addition of anti-TNF or anti-uPA antibodies, the cells were washed three times in serum-free RPMI 1640, and the antibodies were added in RPMI 1640 containing 10% acidtreated fetal calf serum at the concentrations given in the figure legends. Cells were incubated for 30 min in the presence of antibody prior to treatment with TNF and/or cycloheximide. Competitive (#042) and noncompetitive (#047) anti-TNF monoclonal antibodies were obtained from Peptide Technology Ltd. (Sydney, Australia).
TNF Binding Assays-Cells were plated in triplicate at 2 ϫ 10 5 /well in 24-well culture plates (Linbro, Flow) and allowed to adhere overnight. The plates were then placed on ice, washed with binding medium (Hanks' balanced salt solution containing 5% heat-inactivated fetal calf serum), and incubated with 0.3 ml of cold binding medium containing 0.22 nM I 125 -labeled human recombinant TNF (10 5 dpm/well, 15-30 TBq/mmol, Amersham Australia). Selected wells also contained a 1000fold excess of unlabeled TNF (Pharma Biotechnologie, Hannover). After 1 h on ice, the monolayers were washed three times with binding medium prior to extraction with 0.5 M NaOH for 48 h prior to ␥ scintillation counting. Cell numbers were determined by trypsinization of identically treated monolayers run in parallel with the binding studies.
Fluorescent Nuclei Staining-Cells were seeded at 10 3 cells/well and cultured in RPMI 1640, 10% fetal calf serum for 16 h on sterile glass coverslips in 24-well plates. The media was removed and replaced with media alone or media containing 10 ng/ml TNF and/or 10 g/ml cycloheximide and incubated for 4 h. Staining was performed essentially as described by Chen (31) using Hoechst stain 33258 (Sigma) at 0.1 g/ml. Coverslips were mounted in 50% glycerol in citrate phosphate buffer (0.1 M citrate and 0.2 M Na 2 HPO 4 , pH 5.5), and the nuclei were visualized by fluorescent microscopy.
DNA Fragmentation-Cells were seeded at 2 ϫ 10 6 cells/well in a 24-well plate, allowed to adhere for 16 h, and where indicated treated with TNF (10 ng/ml) and/or cycloheximide (10 g/ml). Following incubation for 5 h, the cells were harvested and resuspended in buffer containing 20 mM Tris-HCl, pH 8.2, 800 mM NaCl, and 4 mM EDTA. SDS was added to a final concentration of 1%, proteinase K was added to a final concentration of 2 mg/ml, and the samples were incubated for 1 h at 37°C and centrifuged for 5 min at 10,000 ϫ g. The DNA was precipitated from the supernatants with 2.5 volumes of cold ethanol. The recovered DNA was analyzed by electrophoresis on a 1% agarose gel and visualized with ethidium bromide.
Electrophoretic Mobility Shift Assay-Binding reactions were carried out at for 10 min at 4°C with approximately 5 ϫ 10 3 cpm of radiolabeled NFB probe, 3 g of poly(dI⅐dC), and 1 g of nuclear extract under the following reaction conditions: 15 mM Tris-HCl, pH 7.5, 1.5 mM EDTA, 75 mM NaCl, 20 g bovine serum albumin, 7.5% glycerol, 0.3% Nonidet P-40, and 1.5 mM dithiothreitol). Synthesized single stranded oligonucleotides containing the NFB binding site were annealed and endfilled with the Klenow fragment of DNA polymerase in the presence of [␣-32 P]dCTP (1000 Ci/mmol). The primers used were as follows: NFBMain, 5Ј-ACTCACTTTCCGCTGCTCACTTTCC-3Ј, and NFBPri: 5Ј-GGAAAGTGAGCAGCGG-3Ј. For competition experiments, excess unlabeled double stranded NFB probe was incubated with nuclear extract in the presence of radiolabeled probe at 0°C for 10 min. DNAprotein complexes were separated on a 5% nondenaturing polyacrylamide gel run at 11 V/cm in TBE buffer (89 mM Tris borate and 2 mM EDTA, pH 8.2).

Generation and Characterization of Stable HeLa Cell Lines
That Express PAI-2-To generate sense and antisense PAI-2 cDNA, a DNA fragment containing the entire PAI-2 coding sequence and 3Ј-untranslated region was inserted in two orientations into the expression vector, pRcCMV, under control of the constitutive cytomegalovirus (CMV) promoter. As a positive control for an irrelevant gene, the coding sequence for the CAT gene was inserted into the same vector. HeLa cells were transfected with each of these constructs, as well as vector alone as a negative control, and stable transfectants were selected by resistance to G418. A range of stable clones were isolated from each transfection and analyzed for expression of the transgene by Northern and immunoblot analyses. Each clone was found to express varying levels of transgene mRNA. Northern blots of a selection of sense PAI-2, antisense PAI-2, and CAT clones are shown in Fig. 1A. In the PAI-2 sense clones, two mRNA transcripts are detected that correspond to the mobilities predicted as a result of alternate use of 3Ј-untranslated sequences, one being the PAI-2 3Ј-untranslated region (1.9 kb) and the second being the bovine growth hormone 3Ј-untranslated sequence contained within pRcCMV (2.1 kb). The CAT mRNA transcript is detected at approximately 1.0 kb. Analysis of PAI-2 expression by immunoblot analysis shows that PAI-2 mRNA levels correlate with the synthesis of immunoreactive PAI-2 protein in the sense clones, with none detected in the antisense PAI-2 or vector alone control cell lines (Fig. 1B). The PAI-2 protein synthesized has a molecular mass of 47 kDa and migrates with the same mobility as native PAI-2 induced by lipopolysaccharide in the macrophage-like MonoMac6 cell line. A second minor band is detected beneath the 47-kDa PAI-2 band in both Total RNA was isolated from approximately 1 ϫ 10 7 cells derived from each of the cloned transfectants and analyzed by Northern blot using a 32 P-labeled probe derived from PAI-2 cDNA. The lower panel demonstrates the total amount of RNA loaded in each lane following reprobing of the blots for 18S rRNA. The prefix S designates sense PAI-2 transfectants, and A designates antisense PAI-2 transfectants. CAT designates transfectants stably transfected with the CAT cDNA gene. B, immunoblot and ELISA analyses of PAI-2 protein expression. PAI-2 protein was detected by immunoblot analysis of cell lysates probed with anti-PAI-2 monoclonal antibody as described in the text. CON designates 5 l of lysate from lipopolysaccharide-treated MonoMac6 cells. The level of PAI-2 in each lysate sample was quantitated by ELISA using the PAI-2 Immunobind kit and are given below each lane. The concentration of PAI-2 is expressed as ng of PAI-2 (relative to the PAI-2 standard provided in the kit)/mg of cellular protein (determined by Bio-Rad protein assay). C, immunoblot analysis of CAT protein. CAT protein of 80 kDa was detected in cell lysates from CAT transfectants using an anti-CAT monoclonal antibody. D, uPA binding activity of PAI-2 synthesized by transfectants. Cell lysate from sense clone S1a was prepared as described under "Materials and Methods" except that 4-(2-aminoethyl)benzenesulfonylfluoride was omitted from the lysis buffer. To the lysate was added purified high molecular mass uPA (55 kDa) (Calbiochem), and the mixture was incubated for 1 h at 4°C prior to separation of lysate proteins by SDS-gel electrophoresis and immunoblotting. The 47-kDa PAI-2 and the characteristic SDS-resistant uPA-PAI-2 complex (approximately 92 kDa) were monitored using an anti-PAI-2 monoclonal antibody. Lane 1, control in the absence of added uPA; lane 2, S1a lysate with 2 ng of uPA; lane 3, S1a lysate with 4 ng of uPA. the sense PAI-2 clones and in the MonoMac6 cell lysates, which likely represents a PAI-2 proteolytic cleavage product. The amount of PAI-2 protein synthesized by each of the cell lines was quantitated by ELISA and was found to correlate with the levels detected by immunoblot analysis (Fig. 1B). Stable cell lines that had been transfected with the CAT gene produced the 80-kDa CAT protein (33), as shown in Fig. 1C.
The PAI-2 synthesized by the cells was tested for biological activity demonstrated by the ability to react with uPA, the serine proteinase target of PAI-2. Interaction of uPA and PAI-2 results in the formation of an SDS-resistant complex that demonstrates altered mobility on SDS-polyacrylamide gel electrophoresis or a characteristic "band shift" from 47 kDa to approximately 92 kDa (34). As demonstrated in Fig. 1D, addition of purified uPA to cell lysates derived from PAI-2 transfectants results in such a band shift, demonstrating that the recombinant PAI-2 is biologically active.
Resistance to TNF Cytotoxicity Correlates with PAI-2 Expression-As has been observed with a number of cell types in vitro, HeLa cells are susceptible to cell death by TNF when protein synthesis is inhibited (35). Each of the HeLa cell transfectants was treated with TNF in the presence of cycloheximide for 8 h, after which time the parental HeLa cell line showed evidence of extensive cell death. The amount of cell death was quantitated by two independent assays: 1) crystal violet staining, which monitors cell survival as a function of cell adherence, and 2) MTT assay, which provides a measure of viable cells. When assayed by either method, transfectants expressing PAI-2 (S1a, S1b) showed greater than 80% cell survival relative to antisense PAI-2 transfectants (A2/7, A2/17), cells transfected with vector alone (CMV), or cells transfected with CAT, each of which showed less than 30% survival over the 8-h time period (Fig. 2A). The cell death observed was not due to an effect of inhibition of protein synthesis, as treatment with cycloheximide alone had minimal effects on cell survival.
The effect of TNF concentration on the sensitivity of the HeLa transfectants to cell death is shown in Fig. 2C. The PAI-2-expressing cell lines show enhanced survival at concen-

FIG. 2. Sensitivity of HeLa transfectants to TNF-mediated cytolysis.
A, cell survival following treatment with TNF and cycloheximide measured by MTT assay. The HeLa transfectants, sense PAI-2 (S1a, S1b), antisense PAI-2 (A2/7, A2/17), vector alone (CMV), CAT, and parental HeLa cells were assayed untreated (dotted bar) or following treatment with 10 ng/ml TNF (white bar), 10 g/ml cycloheximide (hatched bar), or both of these agents (black bar) for 8 h. The data are representative of at least five independent experiments and are expressed as the percentages of survival relative to untreated cells. The error bars represent the standard error. Absence of error bars indicates that the error was too small to allow display of the error in those columns. B, cell survival following treatment with TNF and interferon-␥ measured by MTT assay. The HeLa transfectants, sense PAI-2 (S1a, S1b), antisense PAI-2 (A2/7, A2/17), vector alone (CMV), CAT, and parental HeLa cells were assayed untreated (dotted bar) or following treatment with 10 ng/ml TNF (white bar), 10 g/ml interferon-␥ (hatched bar), or both of these agents (black bar) for 8 h. As an increase in cell numbers was observed in untreated cells over the course of 72 h, the percentage of relative survival shown is a measure of cell growth of treated cells after 72 h relative to the growth of untreated cells. The data are representative of at least five independent experiments, and the error bars represent the standard error. C, relative survival of HeLa transfectants with increasing TNF concentration. Each transfectant was treated with increasing concentrations of TNF in the presence and the absence of 10 g/ml cycloheximide and with cycloheximide alone for 8 h and assayed for cell survival by crystal violet staining. The data are expressed as the percentages of survival relative to cells treated with cycloheximide alone. PAI-2-expressing sense transfectants, S1a (Ⅺ), S1b (E), S1n (छ), S1/4 ( ), S1/5 (µ), and S1/6 (Q), antisense transfectant A2/7 (f), the parental HeLa cell (ࡗ). and the CMV vector control transfectant ( ). D, plot of the log of the concentration of PAI-2 expressed by the PAI-2 sense transfectants versus the concentration of TNF required for 50% cell death. Values are derived from the data given in (C). The symbols for each sense clone are as in C, and the level of PAI-2 expressed by each transfectant was determined by ELISA relative to total cellular protein as described in Fig. 1. The line represents the best fit to the data points with a correlation coefficient r 2 ϭ 0.947.

PAI-2 Inhibits TNF-mediated Apoptosis
trations of TNF below 10 ng/ml but demonstrate less resistance to the cytolytic effects of TNF at higher concentrations of TNF. At 10 ng/ml the clones expressing the highest levels of PAI-2, S1a and S1b, were essentially resistant to TNF-induced cell death, as compared with antisense PAI-2, CAT, or CMV transfectants; whereas at 50 ng/ml TNF, S1a and S1b show approximately 50% survival compared with 25-30% for the control transfectants. Furthermore the resistance to TNF-induced cell death appears to correlate with the relative level of PAI-2 expressed by the transfected cells. Indeed, when the concentration of TNF required for 50% cell death for each of the sense PAI-2-expressing cell lines was plotted against the log of the concentration of PAI-2 expressed by each cell line (Fig. 2D), a linear relationship was obtained.
That the cell death observed was a direct consequence of TNF action was demonstrated by comparing the effects of non-FIG. 3. PAI-2 protects against cell death mediated by TNF, but protection is not correlated with TNF receptor expression. A, effect of competitive and noncompetitive anti-TNF antibodies on cytolysis by TNF of PAI-2 transfectants. The PAI-2 sense transfectant S1a (Ⅺ) and the antisense transfectant A2/7 (f) were preincubated with increasing concentrations of anti-TNF antibodies either competitive for TNF binding (solid line) or noncompetitive for TNF binding (dotted line) for 30 min. Cells were then treated with 10 ng/ml TNF and 10 g/ml cycloheximide as described under "Materials and Methods." Cell survival was measured by MTT assay and is represented as the percentage of survival relative to an identical sample treated with cycloheximide alone. Treatment of the cells with antibodies alone or in combination with TNF or cycloheximide had no significant effect on cell survival over the period of the assay. B, expression of 55-kDa TNF receptor in HeLa cell transfectants with differential sensitivity to TNF cytolysis measured by Northern blot analysis. Total RNA was isolated from approximately 1 ϫ 10 7 cells derived from each of the cloned transfectants and analyzed by Northern blot using a 32 P-labeled probe derived from 55-kDa TNF receptor cDNA. HeLa transfectants are as indicated, and CON designates mRNA isolated from MonoMac6 control cells. The lower panel demonstrates the total amount of RNA loaded in each lane following reprobing of the blots for 18 S rRNA. The bar graphs represent the levels of 55-kDa TNF receptor mRNA quantitated from scanned autoradiographs, normalized to the 18 S rRNA loading control and expressed relative to the MonoMac6 mRNA control (CON) loaded on each gel. C, TNF-induced expression of manganese superoxide dismutase mRNA in HeLa cell transfectants with differential sensitivity to TNF cytolysis. Total RNA was isolated from approximately 1 ϫ 10 7 cells derived from each of the HeLa transfectants and analyzed by Northern blot using a 32 P-labeled probe derived from manganese superoxide dismutase cDNA. The lower panel demonstrates the total amount of RNA loaded in each lane following reprobing of the blots for 18 S rRNA. D, electrophoretic mobility shift analysis of NFB activation in nuclear extracts from HeLa clones following treatment with TNF. The arrow indicates the retarded complex containing NFB, which is abolished in the presence of cold NFB competitor DNA. NS denotes nonspecific binding.
competitive and competitive anti-TNF antibodies (which interfere with TNF binding to its receptor) on cell death (Fig. 3A). The PAI-2-expressing clone, S1a, and the antisense clone, A2/7, were incubated with increasing concentrations of each of the antibodies and then subjected to challenge with TNF and cycloheximide. Incubation with increasing concentrations of the competitive anti-TNF antibody resulted in increased survival of both the sense and the antisense clones, whereas the noncompetitive antibody had no significant effect on the relative survival of either clone. Therefore, PAI-2 demonstrates protection against cell death mediated directly through TNF.

PAI-2 Protects from the Synergistic Cytotoxic Effects of TNF and Interferon-␥-
The inflammatory cytokine, interferon-␥, synergizes with TNF in the absence of cycloheximide to promote cell death (36,37). Treatment of HeLa cells with interferon-␥ alone has a cytostatic effect on HeLa cells, whereas treatment with TNF in the presence of interferon-␥ results in approximately 90% cell death after 72 h. However, the PAI-2expressing clones, S1a and S1b, were protected from TNF-and interferon-␥-mediated cell death, demonstrating 70 -80% cell survival, relative to the control clones, A2/7, A2/17, CMV, and CAT transfectants (10 -30% survival) as determined by both MTT (Fig. 2B) and crystal violet assays (data not shown). Interferon-␥ treatment alone had a cytostatic effect on antisense and control transfectants similar to that observed in HeLa cells, whereas interferon-␥-treated PAI-2-expressing cells continued to grow during this period.

Loss of TNF Sensitivity in PAI-2-expressing Clones Is Not a Result of Loss of TNF Receptor Binding-A mechanism by which cells can become resistant to cytolysis mediated by TNF
is through shedding of TNF receptors. Soluble TNF-binding proteins resulting from the proteolytic cleavage of TNF receptors can compete with the cell-associated receptors for TNF binding (38). Therefore, we examined whether the reduced sensitivity of PAI-2-expressing clones to TNF-induced cell death may be attributed to altered TNF receptor expression. Examination of TNF receptor mRNA levels in the sense and antisense transfectants by Northern blot analysis (Fig. 3B) showed that although each transfectant expressed variable levels of 55-kDa TNF receptor mRNA, TNF receptor expression could not be correlated with the sensitivity of the clones to cell death. No 70-kDa TNF receptor mRNA was detected in any of the transfectants or the parental HeLa cell line (data not shown). The TNF receptors present on the transfected clones were functional, as demonstrated by their ability to specifically bind 125 I-TNF. Only minor differences in binding were observed between PAI-2-expressing and control cells (Table I), which are unlikely to result in such a dramatic resistance to cell death. For example, S1b and ASM1 show similar levels of receptor binding, but their susceptibility to TNF cytolysis is very different. The PAI-2-expressing cells respond to TNF with an increase in the TNF-inducible gene, manganese superoxide dismutase (39) similar to control cells (Fig. 3C), and activation of the transcription factor NFB by TNF is also the same in PAI-2-expressing cells as in control cells (Fig. 3D), suggesting that signal transduction through the TNF receptor is not impaired. Therefore the resistance of these PAI-2 transfectants to TNF-induced cell death is not apparently due to alterations in functional TNF receptors.
Resistance to TNF Cytotoxicity Is Not Due to Changes in Manganese Superoxide Dismutase or bcl-2-The resistance to TNF-mediated cell death demonstrated by expression of PAI-2 is reminiscent of the protective effects reported by others for manganese superoxide dismutase (40) and bcl-2 (41). In order to investigate whether the protection conferred by PAI-2 may be a consequence of changes in expression of these other "protective" genes, the levels of these genes in the various transfected cell lines was examined. Each of the transfectants express two manganese superoxide dismutase mRNA transcripts (Fig. 4A), the levels of which do not correlate with PAI-2 expression or with the differential sensitivity of these cells to TNF cell death. Because alterations in manganese superoxide dismutase mRNA levels reflect the level of protection conferred by manganese superoxide dismutase (40), these results demonstrate that the protection observed is unlikely to be due to manganese superoxide dismutase. A similar result was obtained when the levels of bcl-2 were monitored in the cell lines by immunoblot analysis using an antibody directed against bcl-2 (Fig. 4B). Each cell line expressed levels of bcl-2 comparable with control cells, and the level of bcl-2 expressed by each transfectant could not be correlated with sensitivity to TNFinduced cell death.
PAI-2 Protects from TNF-induced Apoptosis-Cell death mediated by TNF can occur through programmed cell death or apoptosis in some cells but not in others (42). Apoptosis is characterized by morphological changes of cell shrinkage, chromatin condensation, and membrane blebbing (1) and usually involves internucleosomal DNA cleavage (43). To determine whether PAI-2 conferred protection against apoptosis triggered by TNF in these cells, the PAI-2 sense transfectant, S1a, and the antisense transfectant, A2/7, were examined by fluorescence microscopy following DNA staining. As shown in Fig. 5 (A  and B), the antisense transfectant, A2/7, undergoes about 87% apoptosis, as judged by morphology following TNF cytolytic challenge. In contrast, the TNF-resistant PAI-2-expressing cell line shows little evidence of apoptosis over a similar time period (Fig. 5, C and D), demonstrating that PAI-2 is acting to block apoptotic cell death. This is confirmed by the DNA fragmentation patterns observed for the each of the transfectants following TNF challenge (Fig. 5E). Although significant "DNA laddering" is detected in the TNF-sensitive transfectants, A2/7, ASM1, and the parental HeLa control, the TNF-resistant PAI-2 sense transfectant, S1a, shows little evidence of a similar pattern of DNA laddering.
The ability of PAI-2 to block TNF-induced apoptosis appears to be specific, in as much as the PAI-2 transfectants are susceptible to other stimuli that induce apoptotic cell death in HeLa cells, such as ultraviolet (UVC, 30 J/m Ϫ2 ) or gamma (1 kiloradian) irradiation (data not shown).
PAI-2 Protection Is Not Dependent on Urokinase-type Plasminogen Activator Activity-HeLa cells are sensitive to TNFmediated cell death but do not express significant levels of uPA; thus they appeared to be an ideal model to test the involvement of uPA in the apoptotic pathway blocked by PAI-2. However, although resting HeLa cells synthesize only barely detectable levels of uPA, treatment with cycloheximide stabilizes uPA a Specific binding was defined as the difference between total counts in the absence of unlabeled TNF and counts in the presence of unlabeled TNF (nonspecific binding). Nonspecific binding ranged from 5-10% of total binding. b Data taken from Figure 2. ϩϩϩ, strong protection; ϩ, weak protection; Ϫ, not protected.

PAI-2 Inhibits TNF-mediated Apoptosis
mRNA, and uPA protein is detected at more significant levels in lysates from cells treated with TNF and cycloheximide (Fig.  6B). However, no evidence of complex formation between the expressed PAI-2 and uPA resulting in a band shift on SDSpolyacrylamide gels was observed even following prolonged exposure of the immunoblot (Fig. 6A), suggesting that PAI-2 was not interacting with detectable amounts of cell-associated uPA. In order to block low levels of uPA activity, the transfectants were exposed to several inhibitors of uPA activity prior to and during challenge with TNF and cycloheximide. As shown in Fig. 6B, treatment with either the uPA inhibitor, amiloride (44), or the specific peptide inhibitor of uPA, glutamyl-glycylarginyl-chloromethyl ketone (45), had no effect on the relative survival of the PAI-2-expressing clone, S1a, or on the death of the antisense clone, A2/7. Incubation with a blocking monoclonal anti-uPA antibody, which inhibits uPA catalytic activity, also demonstrated no significant effect. Similarly, incubation of S1a or A2/7 with 2,7-bis(4-amidinobenzylidene)-cycloheptan-1-one dihydrochloride, a synthetic inhibitor of tissue-type plasminogen activator, uPA, trypsin, plasmin, and a selection of other trypsin-like serine proteinases, had no significant effect on the relative survival of the HeLa transfectants. These data demonstrate that uPA activity is not likely to play a role in the protective action of PAI-2 in TNF-induced cell death.
PAI-2 Protection Is Dependent upon the P 1 Amino Acid, Arg-380-PAI-2 is a member of the serine proteinase inhibitor, or serpin superfamily. Serpins are structurally conserved, single chain molecules containing a reactive site near the carboxyl terminus that acts as a pseudosubstrate for the cognate proteinase. The specificity of inhibition by serpins is largely governed by the nature of the P 1 reactive center amino acid residue, although secondary sites can also significantly modulate inhibition (46,47). In order to determine whether the P 1 residue of PAI-2, Arg-380 (48), was a determinant in mediating the observed protection against TNF-induced apoptosis, a mutant PAI-2 was engineered wherein Arg-380 was converted to Ala-380 by mutagenesis of the corresponding cDNA sequence FIG. 4. Expression of manganese superoxide dismutase and bcl-2 in HeLa cell transfectants with differential sensitivity to TNF cytolysis. A, Northern blot analysis of manganese superoxide dismutase mRNA levels in PAI-2 transfectants. Total RNA was isolated from approximately 1 ϫ 10 7 cells derived from each of the cloned transfectants and analyzed by Northern blot using a 32 P-labeled probe derived from manganese superoxide dismutase cDNA. The Northern blots are overexposed to reveal the constitutive low levels of manganese superoxide dismutase mRNA. The lower panel demonstrates the total amount of RNA loaded in each lane following reprobing of the blots for 18 S rRNA. The bar graphs represent the levels of 4-and 1-kb manganese superoxide dismutase mRNA quantitated from scanned autoradiographs, normalized to the 18S rRNA loading control and expressed relative to the MonoMac6 mRNA control (CON) loaded on each gel. B, immunoblot analysis of bcl-2 protein expression in PAI-2 transfectants. Cell lysates from each of the PAI-2 transfectants was probed with anti-bcl-2 monoclonal antibody as described in the text. For both A and B, the PAI-2 sense transfectants are S1a, S1b, S1/4, S1/5, and S1/6, and the antisense PAI-2 transfectants are A2/7, A2/12, A2/15, A2/16, and A2/17. HeLa is the parental cell; CON is lipopolysaccharide treated MonoMac6 cells. ASM1 is a transfectant containing Ala-380 PAI-2 (see Fig. 1).

FIG. 5. Expression of PAI-2 protects from apoptotic cell death.
Fluorescent micrographs of Hoechst (No. 33258) stained nuclei from the antisense A2/7 (A and B) and the PAI-2-expressing sense S1a (C and D) transfectants either untreated (A and C) or following treatment with 10 ng/ml TNF and 10 g/ml cycloheximide for 4 h (B and D). E, electrophoretic separation on agarose of DNA isolated from nuclei of PAI-2 sense (S1a), antisense (A2/7), HeLa parental cells, and a transfectant containing Ala-380 PAI-2 (ASM1) either untreated (Ϫ) or following treatment with TNF and cycloheximide for 6 h (ϩ). The gel was visualized by staining with ethidium bromide.
(illustrated in Fig. 7A). Documented mutations in the P 1 residues of other serpins, either naturally occurring or genetically engineered, alter the specificity of the inhibitor without influencing the overall architecture of the serpin (46). The rationale for this particular mutation was to convert the P 1 amino acid at the PAI-2 reactive site to that of chicken ovalbumin, a structurally related serpin that has no known proteinase inhibitor activity. Immunoblot analysis shows that the Ala-380 PAI-2 mutant is recognized by the anti-wild-type PAI-2 monoclonal antibody, indicating that the conformational epitope recognized by the monoclonal antibody is not altered in the mutant PAI-2 protein (Fig. 7B). Furthermore, in the presence of uPA there is no evidence of a typical band shift by SDS-polyacrylamide gel electrophoresis, demonstrating that the Ala-380 PAI-2 mutant has lost its uPA binding activity as predicted (Fig. 7B). However, Ala-380 PAI-2-expressing clones (e.g. ASM1) still display normal TNF receptor function as demonstrated by 125 I-TNF binding (Table I) and effective TNF-dependent signaling (Fig. 3C). Assay of the sensitivity to TNF of two clones expressing Ala-380 PAI-2 (ASM1 and ASM2) shows that these cells are as sensitive to TNF-induced apoptosis as the antisense PAI-2 mutant, A2/7, and wild-type HeLa cells (Fig. 7C). As the level of PAI-2 expressed by cells is related to the protection observed, the two wild-type PAI-2-expressing clones, S1/5 and S1n, which produce slightly higher and lower levels of PAI-2, respectively, as the ASM1 and ASM2 mutants, are provided in Fig. 7C for comparison. These results demonstrate that the P 1 amino acid residue of PAI-2 (Arg-380) is required for protection against TNF-mediated apoptosis. DISCUSSION TNF is a potent inducer of cell death in many systems, and some cells have developed mechanisms to protect against the cytotoxic effects of TNF. The results of this study show that expression of the macrophage product, PAI-2, protects HeLa tumor cells from apoptosis induced by TNF. The apoptotic death pathway blocked by PAI-2 is independent of bcl-2 and manganese superoxide dismutase. The mechanism of PAI-2 protection is independent of uPA but requires the P 1 -Arg amino acid residue of PAI-2.
The findings that serine proteinases may be involved in TNF-induced cell death (10,11) and that PAI-2 biosynthesis is part of an acute phase cellular response to TNF (49) has led to an investigation of the possible function of PAI-2 in TNFmediated cytotoxicity. Kumar and Baglioni (12) provided the first evidence that PAI-2 may function as a protective protein in TNF-induced cytotoxicity by demonstrating that overexpression of PAI-2 in the fibrosarcoma cell line HT1080 conferred resistance to these cells from the cytotoxic effects of TNF. We demonstrate that expression of PAI-2 in cells that do not synthesize endogenous PAI-2 confers protection from TNF-induced cell death and that PAI-2 protects against apoptosis. Protection from TNF-mediated apoptosis is observed in the presence of either the protein synthesis inhibitor cycloheximide or the inflammatory cytokine, interferon-␥. Several PAI-2-expressing clones were assayed for their level of PAI-2 expression, and this was correlated with their protection against TNF. Measurement of PAI-2 levels by ELISA shows that the levels of PAI-2 required for protection are well within the range produced by nontransfected cells. lipopolysaccharide-induced macrophage-like MonoMac6 cells synthesize a 4-fold greater amount of PAI-2 than the transfected cell line expressing the highest levels of PAI-2. A threshold level of PAI-2 appears to be required to protect cells against TNF-induced cell death with increases beyond this threshold having no additional TNF-binding proteins, which are soluble extracellular domains of TNF receptors produced by proteolytic cleavage, can be actively shed by cells and may function to prevent TNF from inducing continuous damage to the cell (38). The reduced sensitivity of PAI-2 transfectants to TNF is unlikely to be due to such altered TNF-receptor binding in these cells, because the sensitivity of the different transfectants to TNF could not be correlated to loss of TNF receptors or impaired ability of TNF to bind to receptors. Furthermore, signal transduction through the 55-kDa TNF receptor functions similarly in transfected cells as in control cells, evidenced by the TNF-mediated transcriptional activation of manganese superoxide dismutase, indicating that this TNF signaling pathway is functional in these cells.
Inhibition of protein synthesis in many tumor cells increases their sensitivity to TNF, an observation that has led to the suggestion that TNF itself may induce "protective proteins" in TNF-resistant cells. Overexpression of the oxygen radical scavenger manganese superoxide dismutase partially protects from TNF cytotoxicity, suggesting that other TNF-inducible proteins exist that are involved in conferring protection (40). The oncogene bcl-2 protects from apoptotic cell death but appears to have a cell-specific effect against TNF-induced cell death (41,50). The levels of manganese superoxide dismutase and bcl-2 were not found to correlate with expression of PAI-2 nor with the sensitivity of each of the transfectants to TNF-induced apoptosis. Given that manganese superoxide dismutase protects from cell death mechanisms mediated by oxygen radicals (51) and that PAI-2 does not protect from apoptosis induced by ultraviolet or ionizing radiation, both of which utilize oxygen radicals in their cytolytic mechanisms (52), it is likely that PAI-2 exerts its effect either upstream or independently of manganese superoxide dismutase. Thus, there may be several pathways leading to cell death, which possibly converge at different points, and it seems likely that several proteins may inhibit different arms of the cell death response.
The mechanism by which PAI-2 confers resistance to TNF cell death is not yet clear but is unlikely to be associated with uPA proteolytic activity. The data presented do not support a role for uPA in TNF-mediated cell death and further suggests that an intracellular proteinase target for PAI-2 might exist. Members of the serpin superfamily, such as PAI-2, commonly act as proteinase inhibitors, and there are precedents of serpins interacting with multiple proteinase targets. There is also indirect evidence for the involvement of proteinases in TNFmediated cell death (10,11). A characteristic of serine proteinases is the catalytic triad of Ser, His, and Asp residues that form a hydrogen bonding system responsible for peptide bond cleavage. The substrate specificity of serine proteinases is determined by interaction at the primary substrate binding site (S 1 ). Serine proteases can be divided into three major groups based on their S 1 specificity (53): trypsin-like enzymes, which cleave substrates with positively charged residues such as Arg or Lys at the P 1 site due to interaction with Asp in the S 1 pocket of these enzymes; chymotrypsin-like enzymes, which cleave substrates with aromatic or large hydrophobic residues at the P 1 position; and elastase-like proteinases, which prefer substrates with small aliphatic residues at the P 1 position. The essential requirement for an Arg in the serpin P 1 site of PAI-2 is evidence for the involvement of a proteinase and possibly a trypsin-like serine proteinase in the apoptotic death pathway blocked by PAI-2. A further implication of this finding is that the proteinase is likely to have a sequence preference for Arg adjacent and amino-terminal to the scissile peptide bond (P 1 residue) of its natural substrate. Identification of an intracellular proteinase that is inhibited by PAI-2 may provide a key to unravelling a novel pathway leading to cell death.
Alternatively, PAI-2 may inhibit an Arg-specific cysteine proteinase involved in TNF-induced apoptosis. A proteinase recently associated with the regulation of cell death and apoptosis is the cysteine proteinase, ICE. A specific inhibitor of ICE is the cowpox virus encoded serpin, crmA (54). PAI-2 and crmA show a high degree of structural similarity; residues 46 -338 of FIG. 7. Mutation of the PAI-2 serpin P 1 residue abolishes the protective effect of PAI-2. A, diagram indicating the mutations engineered into the PAI-2 cDNA sequence to convert the P 1 -arginine to an alanine in the ASM construct (Ala(380) PAI-2). B, relative levels of mutant PAI-2 in ASM1 and ASM2, compared with wild-type PAI-2 in S1a, S1n, and S1/5. HeLa cells were transfected with the ASM construct, stable transfectants were isolated, and PAI-2 expression was monitored in cell lysates by immunoblot analysis. The mutant PAI-2 does not demonstrate uPA binding activity as demonstrated by preincubation of ASM1 cell lysate with 2 ng of 55-kDa uPA for 1 h as described in the legend to Fig. 1D. Blots were probed with an anti-PAI-2 monoclonal antibody. C, relative survival of stable transfectants expressing wild-type and mutant PAI-2. Each of the transfectants was treated with an increasing concentration of TNF, in both the presence and the absence of 10 g/ml cycloheximide, treated with cycloheximide alone for 8 h, and assayed for cell survival by crystal violet staining. The data are represented as the percentage of survival relative to identical samples treated with cycloheximide alone. Treatment with either cycloheximide alone or TNF alone at the concentrations shown had no significant effect on cell survival over the period of the assay. S1a (Ⅺ), S1/5 ( ), ASM1 (i), ASM2 ( ), S1n (छ), and A2/7 (f) transfectants.
crmA share 35% sequence identity with residues 110 -415 of PAI-2. 2 However, PAI-2 is unlikely to represent the mammalian homologue of crmA, because crmA contains an Asp in the P 1 residue, and there is a requirement for a P 1 -Asp in the catalytic mechanism of ICE (55,56). ICE-like activity was found to be similar in PAI-2-expressing and -nonexpressing cells (see "Material and Methods"), indicating that PAI-2 does not inhibit ICE-like proteinases. Other serpins have been described that contain Arg in the P 1 position, which include maspin, a tumor suppressor gene implicated in breast cancer (57); the vaccinia virus-encoded protein K2L (58); the Limulus encoded intracellular coagulation inhibitor (59); and the myxoma virus-encoded SERP-1 (60). Further studies are required to determine the relationships between the functions and specific biological targets of these serpins and those of PAI-2.
The results presented here illustrate a function for intracellular PAI-2 in cell death, which may have important implications in the resolution of inflammatory and cell-mediated immune reactions that are mediated by TNF. Monocytes rely on uPA activity on their cell surface to affect invasion of tissues to access various targets, for example, inflammatory sites. PAI-2 has been shown to function extracellularly to limit uPA activity and thus regulate the invasive capabilities of monocytes (61). The vast majority of PAI-2 synthesized by monocytes, however, remains intracellular, probably due to an inefficient signal sequence (62). Activation of monocytes by inflammatory agents such as TNF induces a rapid accumulation of high levels of PAI-2 in the cytoplasm that does not appear to be correlated with uPA synthesis (63) and may play a functional role in regulating apoptotic cell death in these cells. It is of note that a PAI-2 proteolytic cleavage product has recently been linked with apoptosis in myeloid leukemic cells (64), suggesting that loss of the apoptosis inhibitory activity of PAI-2 may be associated with proteolytic cleavage of the PAI-2 molecule.
Monocytes at sites of inflammation may be exposed to concentrated doses of cytokines, including TNF, which may in combination be cytotoxic. Intracellular PAI-2 may be a part of a cellular response of monocytes/macrophages to delay cell death in order to ensure that they survive secretion of their own TNF, possibly to allow them to complete vital functions such as lymphocyte activation and antigen presentation prior to apoptosis. In support of the role of PAI-2 in cell death, PAI-2 is expressed in vivo in skin (65) in a zone associated with keratinocyte differentiation and apoptosis. Tumor cells have also been found to synthesize PAI-2 (66,67), and transformation by ras and src oncogenes has been shown to induce PAI-2 expression in vitro (68). Given the effect of PAI-2 on cell survival, acquisition of PAI-2 expression by tumor cells may reduce their susceptibility to TNF-mediated apoptosis and thereby allow a selective survival advantage to these tumor cells.