Evidence for two mechanisms by which tumor necrosis factor kills cells.

Tumor necrosis factor (TNF) can inhibit the differentiation of preadipocytes to adipocytes and will revert differentiated adipocytes to the preadipocyte state. TNF is not toxic to either adipocytes or preadipocytes when used alone but is highly toxic to these cells when used in conjunction with cycloheximide, yielding virtually 100% killing within 4-6 h of treatment. A cell line (TA1 R-6) was isolated which is resistant to the combined toxic effects of TNF and cycloheximide. This cell line is stable and, unlike the parental cell line, does not morphologically differentiate to adipocytes or express adipocyte-specific mRNAs. It has a more transformed appearance and growth pattern and, while resistant to the toxic effects of TNF and cycloheximide in a 6-h assay, has become sensitive to cytotoxicity induced by TNF used alone in a 3-day assay. The adipocyte differentiation-inducing agents, dexamethasone and indomethacin, block the cytotoxicity induced by TNF alone in the TA1 R-6 line but do not block the rapid cytotoxicity of TNF and cycloheximide in the parental line. These results provide both genetic and pharmacologic evidence that there are at least two distinct or overlapping pathways by which TNF mediates its effects.

perhaps independent biochemical pathways.
To further analyze the mechanisms by which TNF exerts its diverse functions, one would like to have variant cell lines that exhibit altered responses to TNF. We have recently focused our attention on the ability of TNF to inhibit differentiation of adipogenic cell lines, a reflection of its possible role in cachexia. During the course of these studies, TA1 cells were found to be resistant to the cytotoxic actions of TNF but were rapidly killed when treated simultaneously with TNF and cycloheximide. Cycloheximide has previously been shown to enhance TNF-mediated killing of other cells in culture (18). This observation formed the basis for selecting variant cell lines resistant to the cytotoxic actions of TNF seen in cycloheximide-treated TA1 cells. Among these resistant lines, one exhibited the unusual property of sensitivity to TNF alone in a standard cytotoxicity assay. In this manuscript we describe the characteristics of this variant line derived from TA1 cells and provide both genetic and pharmacological evidence that TNF cytotoxicity in the presence of cycloheximide entails activation of a pathway for cell killing distinct from that involved in standard (3 day) TNF-mediated cytotoxicity. Our results suggest that at least two different signaling pathways stimulated by TNF are associated with cytotoxicity, one of which may also play a role in inhibiting adipocyte differentiation.

MATERIALS AND METHODS
Cell Culture Conditions-TAl, TA1 R-6, and mouse L-929 cells were grown in Eagle's basal medium (GIBCO) supplemented with 10% heat inactivated (56 "C for 30 min) fetal calf serum. Cultures were grown at 37 "C in a humidified incubator in a 5% C02 atmosphere. The medium was changed every 2 days, To accelerate differentiation, confluent plates of cells were treated with indomethacin or dexamethasone as previously described (19,20). Cells were selected in cycloheximide (10 pg/ml) prepared in phosphate-buffered saline. TNF (Cetus) was reconstituted as directed and stored in frozen aliquots at -70 "C until use. For analysis of cytotoxicity, TA1 cells were plated in 96-well microtiter plates at 1 X 10' cells/well in 100 pl of complete media and allowed to reach confluence prior to treatment.
RNA Isolation and Analysis-Total RNA was prepared by a modification of the method described by Chirgwin et al. (21). Briefly, 10cm plates were washed with phosphate-buffered saline (4 "C) and drained. 4 ml of guanidium HC1, pH 5.2, 25 mM sodium citrate, was added, and the suspension was harvested by scraping and transferred to sterile 15-ml polypropylene tubes (Falcon). The DNA was sheared by passing the mixture through a 21-gauge needle 10 times. N-Lauryl sarcosine was then added to 0.5% and absolute ethanol was added to 50%. The tubes were placed at -20 "C overnight and then centrifuged at 7500 X g at 4 "C for 15 min to pellet the RNA. The RNA pellet was resuspended in 800 pl of guanidine HCI, transferred to a microcentrifuge tube, and 400 pl of cold (-20 'C) absolute ethanol was added. The tubes were placed at -20 "C for 4 h, and the RNA recovered by centrifugation in a microcentrifuge for 15 min. The pellet was washed two X in 70% ethanol, resuspended in 200-pl of proteinase-K buffer, and treated with proteinase-K (200 ng/ml) for 30 min at 37 "C (22). 600 p l of guanidium HCl and 400 pl of cold (-20 'C) absolute ethanol were added. The tubes were placed at -20 "C for 4 h, and the RNA was recovered by centrifugation for 15 min in a microcentrifuge. The RNA was washed with 70% ethanol and resuspended in H1O. 10 pg of RNA was then glyoxylated, electrophoresed on 1.2% agarose gel, stained with ethidium bromide for 30 min in 50 mM NaOH as described (22), then destained and neutralized by washing three times (10 min each) in 10 mM sodium phosphate buffer, pH 6.5, photographed and transferred to Nytran filters, baked, prehybridized, hybridized as described by the manufacturer, and probed with random primed cDNA probes as previously described (20).
Cytotoxic Assay-Cell cytotoxicity was measured by staining with crystal violet as previously described (23) except that cells were trypsinized and replated prior to staining so that only viable cells capable of reattaching to the plate were measured. Optical density at 590 was measured using a Dynatech microtiter plate reader.

Synergistic Cytotoxicity of TNF and Cycloheximide on TAl
Cells-We had noted in earlier experiments that cells treated concurrently with TNF and cycloheximide die within a few hours. To determine the time required for this cytotoxicity, TA1 cells were plated in 96-well microtiter plates and treated with TNF alone or in combination with cycloheximide for 0.5-6.0 h. The results of this experiment on preadipocytes and adipocytes is shown in Fig. 1, A and B, respectively. The results indicate that virtually complete lysis of TA1 preadipocytes and adipocytes occurs by 4-6 h after incubation with both agents together, and appreciable cytotoxicity was observed within 1-2 h of treatment. TNF alone is not toxic to these cells, and cycloheximide, at the concentration used, exhibits only 20-50% reduction in cell number over the 6-h pulse, which is mainly due to an inhibition in cell proliferation compared with the control cells rather than cytotoxicity. Even after 24 h of treatment with cycloheximide, the survival for the TA1, TA1 R-6, and mouse L-929 cells is 36, 38, and 40% of control, respectively.
Selection of Resistant Variants and Characterization of Resistant Clone R-6"Based on the above results, we developed a strategy for isolating variants resistant to the combined action of these agents. TNF (0, 0.1, 1.0, 10, and 100 ng/ml) and cycloheximide (0, 1, 10, and 100 pg/ml) were titered to determine the optimal concentrations for TNF-mediated killing of TA1 cells in the presence of cycloheximide. The optimal concentration of cycloheximide was 10 pg/ml. At this dose, the cytotoxicity due to cycloheximide alone was minimal, and the synergy with TNF yielded cytotoxicity sufficient to select rare, resistant variants. TNF-induced cytotoxicity could be readily detected at 0.1 ng/ml and appeared to plateau at 10-100 ng/ml (see Fig. 2 A ) . We therefore chose 10 ng/ml of TNF and 10 pg/ml of cycloheximide as appropriate concentrations for selection of TNF-resistant variants. 5 X lo7 cells were treated with TNF (10 ng/ml) and cycloheximide (10 pg/ml) for 6 h, then washed free of the selecting agents, and refed complete medium. No mutagenesis was used to select these variants. This schedule was repeated weekly for 3 weeks, and the surviving colonies were isolated by ring cloning, expanded, and analyzed for resistance to TNF and cycloheximide. The frequency of surviving cells from a series of these experiments was in the range of 1 X IOp6 to 1 x Among the several clones isolated, TA1 R-6 was chosen for detailed analysis of its response to TNF and cycloheximide because it was highly resistant to the cytotoxic effects of TNF and cycloheximide used together, but retained some TNF mediated functions. Other mutants with different properties will be described elsewhere. The relative degree of resistance of the TA1 R-6 cell line was determined by treating both TA1 tal rnf crystal violet. E , TA1-R6 treated with 10 ng/ml T N F and 1 p~ dexamethasone for 3 days, stained with crystal violet, and F, TA1-R6 treated with T N F 10 ng/ml and indomethacin (50 p~) for 3 days and stained with crystal violet. The magnification in A and B is greater than in C-F in order to show the accumulation of lipid droplets; lower magnification is more suitable for C-F to demonstrate focus formation and the effects of T N F on the TA1-R6 cells.
R-6 and the parental TA1 cells with cycloheximide and various concentrations of T N F for 8 h. The results, shown in Fig. 2A, indicate that in the presence of cycloheximide, as little as 0.1 ng/ml of T N F is required for nearly complete cytotoxicity of the parental cell line, while 100 ng/ml of T N F did not yield significant toxicity in the TA1 R-6 cell line, a greater than 1000-fold increase in resistance. To further char-acterize the response of these cells, cytotoxicity induced by T N F was determined at two concentrations of cycloheximide. Fig. 2B indicates that an 8-h treatment of these cells with either 10 or 100 pg/ml of cycloheximide results in approximately 20 and 60% cytotoxicity, respectively. The cytotoxicity is not appreciably increased by increasing concentrations of TNF. Since, the cytotoxicity of cycloheximide at 10 pg/ml at 24 h is not appreciably different for the parental TA1, the TA1 R-6, or TNF-sensitive mouse L-929 cells, we conclude that the relative resistance of TA1 R-6 cells to TNF/cycloheximide treatment is a consequence of an altered response to TNF, not cycloheximide.

Morphology and mRNA Expression of Mutant and Parental Cell
Lines-Morphologically, the mutant TA1 R-6 cell line appears more transformed than the TA1 parental line. Fig.  3A shows the normal morphology of the undifferentiated TA1 cells. These cells become contact inhibited, form a uniform monolayer, and exhibit well-defined nuclei and cytoplasm. Treatment of these cells with dexamethasone and indomethacin for 3 days induces them to differentiate into adipocytes with the accumulation of large lipid droplets in the cytoplasm (19,20). These droplets appear as the dark staining material in the perinuclear region of these cells when stained with oil red-0 (Fig. 3B). The TA1 R-6 mutant cells, in contrast, grow to much higher densities (1 x lo7 cells/lO-cm plate uersus 1 X lo6 cells/lO-cm plate for the TA1 cells), are not contact inhibited, and tend to pile on each other and form foci (Fig.  3C).
The TA1 R-6 cells cannot morphologically differentiate into adipocytes even following treatment with known inducers of differentiation such as indomethacin or dexamethasone.
To analyze the capability of the TA1, TA1 R-6, and mouse L-929 cell lines to induce adipocyte specific mRNAs, the cells were treated with indomethacin (125 p~) and dexamethasone (1 p~) . RNA was harvested at 3 days after treatment and analyzed for expression of two adipocyte-specific mRNAs, AP2 and clone 47. The results, shown in Fig. 4, indicate that TA1 cells demonstrate marked induction of these adipocyte specific mRNAs, while the TA1 R-6 and mouse L-929 cell lines exhibit no detectable expression of these adipocyte specific mRNAs, even after treatment with indomethacin and dexamethasone for 3 days. Response of TAl and TA1 R-6 Cells to TNF-TNF causes an inhibition of adipocyte-specific gene expression in TA1 adipocytes and reverts the phenotype of these cells to the fibroblast-like state. Furthermore, treatment of TA1 preadipocytes with TNF can block the conversion of preadipocytes to adipocytes (13). In our attempts to determine whether TNF would exert any effect upon the phenotype of the TNFresistant TA1 R-6 cell line (see above), we noted that while the TA1 R-6 cell line is fully resistant to the cytotoxic effects of TNF and cycloheximide in an 8-h assay, it exhibits a marked sensitivity to the cytotoxic effects of TNF when used alone for 2-4 days. Fig. 3 0 shows the appearance of TA1 R-6 cells after a 3-day exposure to TNF at 10 ng/ml. Only residual cellular debris, which stains darkly with crystal violet but lacks distinct nuclear and cytoplasmic structures, is observed. Viability was determined as a function of time of exposure to 10 ng/ml of TNF. The results, shown in indicate that cytotoxicity is evident by 2 days and is maximal at about 4 days of exposure in TA1 R-6 cells. This behavior is characteristic of the prototypical TNF-sensitive mouse L-929 cells in which cytotoxicity requires 2 days to become evident and 3-4 days to become maximal. The dramatic alteration in TNF sensitivity of this cell line relative to the parental TA1 cell line may be related to its apparent "transformed" growth behavior.
Inhibition of TNF Cytotoxicity with Dexamethasone and Indomethacin-Based on the knowledge that TNF is generally cytotoxic to transformed but not normal cells, and the fact that, in TA1 cells, dexamethasone and indomethacin induce the differentiated state (19,20), tests were performed to determine whether these differentiation-inducing agents altered the susceptibility of the parental or mutant TA1 cells to the actions of TNF. The mutant TA1 R-6 line was particularly interesting in this respect because of its transformed morphology and its unexpected cytotoxic response to TNF.
From the experiments shown in Fig. 1, A and B, we know that both preadipocytes and differentiated TA1 cells are rapidly killed by the combined action of TNF and cycloheximide. The time course for TNF and cycloheximide toxicity is slightly delayed in adipose cells compared to preadipose cells, but this difference is minimal. There is, therefore, little evidence that the state of differentiation appreciably alters the sensitivity of TA1 cells to the combination of TNF and cycloheximide. TA1 R-6 cells were treated with either dexamethasone (1 p~) or indomethacin (125 pM) concurrently with TNF for 3 days to determine whether either of these differentiationinducing drugs could alter the cytotoxicity of TNF. These results, shown in Fig. 3, E and F, indicate that dexamethasone and indomethacin can block the cytotoxic effects of TNF on these cells. Interestingly, the cells treated with indomethacin and TNF develop a unique spindle-shaped morphology that is not observed in the parental cell line (F). This morphology is not observed in TA1 R-6 cells treated with indomethacin alone (data not shown) and is not observed in cells treated with dexamethasone and TNF ( E ) . This suggests than an interaction between these two agents is responsible for the altered morphology. However, this altered phenotype is not required for resistance to TNF-mediated cytotoxicity since dexamethasone, which on its own does not affect morphology, still blocks cytotoxicity. To further verify the anti-cytotoxic effects of dexamethasone on these cells, a time course of treatment of TNF in the presence or absence of dexamethasone was performed (Fig. 5). These results indicate that dexamethasone can completely block the cytotoxic effects of TNF on these cells over the 4-day period of exposure to TNF. Since the TA1 R-6 cell line cannot establish a fully differentiated state, as measured by its inability to accumulate adipocyte-specific mRNA (Fig. 4), we surmise that these agents are not inhibiting TNF cytotoxicity by inducing the differentiated state. It is more likely that these drugs interact directly or indirectly with one or more of the pathways by which TNF elicits its toxic effects. To assess these possibilities, the concentration of dexamethasone and indomethacin required to inhibit TNF cytotoxicity were determined in mouse L-929 cells, which are sensitive to the cytotoxic effects of TNF, but do not differentiate to adipocytes. The results indicate that both dexamethasone (Fig. 6 A ) and indomethacin (Fig. 6 B ) block the cytotoxic effects of TNF in mouse L-929 cells. Since mouse L-929 cells do not develop a differentiated adipocyte phenotype upon treatment with indomethacin or dexamethasone, one must conclude that the induction of a specific differentiated state is not required for suppression of the cytotoxic effects of TNF. Moreover, the mechanism by which these drugs block TNF cytotoxicity is not exclusive to the TA1 R-6 mutant or cells of the adipocyte lineage. The concentration range over which dexamethasone blocks TNF cytotoxicity is broad, with 50% inhibition requiring about 10-20 nM of dexamethasone (Fig. 6 A ) . Indomethacin, in contrast, is effective over only a narrow range of concentrations, (30-200 p~) (Fig. 6 B ) , since above 300 p~ indomethacin is itself toxic. The effective concentration range for indomethacin is two to three orders of magnitude above that required to inhibit cyclooxygenase (prostaglandin synthetase), but overlaps the concentration range that is optimal for r TNF Cytotoxicity 4587 induction of the differentiated, adipocyte phenotype (20). This may indicate that the biochemical processes by which indomethacin induces differentiation are similar, if not identical, to those involved in blocking TNF cytotoxicity. Finally, dexamethasone and indomethacin may be blocking the cytotoxicity of TNF at different points in the pathway or by different mechanisms. Dexamethasone, while very potent at inhibiting the cytotoxic effects of TNF, only modestly stimulates TA1 cell differentiation, whereas indomethacin at 30-200 p~ markedly accelerates the conversion of preadipocytes to adipocytes (Fig. 4) (3-6 h) killing of TNF in the presence of cycloheximide, TA1 cells exposed to TNF and cycloheximide for 6 h were pretreated with indomethacin or dexamethasone. If the biochemical mechanisms by which TNF causes cytotoxicity are the same as those mechanisms by which TNF and cycloheximide cause cytotoxicity, we might expect dexamethasone or indomethacin to also inhibit toxicity caused by TNF and cycloheximide. However, indomethacin and dexamethasone, given concurrently with TNF/cycloheximide, did not provide measurable protection against cytotoxicity. Moreover, pretreatment of the TA1 cells with either indomethacin or dexamethasone for 0.5, 1, 3, 7, or 21 h prior to TNF/cycloheximide exposure did not provide measurable protection against cytotoxicity even if indomethacin and dexamethasone were kept in the media during the period of TNF/cycloheximide exposure (Table I) these pathways, variants were selected that were resistant to the synergistic cytotoxicity of TNF and cycloheximide. These variants were made in adipogenic TA1 cells in which TNF has been shown to inhibit and reverse differentiation (13). Individual variants might allow the analysis of common and independent pathways by which TNF induces cytotoxicity or inhibits differentiation. Preadipocytes and differentiation were analyzed to determine whether the state of differentiation altered the susceptibility of these cells to TNF plus cycloheximide. Toxicity is evident after only 1-2 h of exposure and causes virtually complete lysis of these cells by 4-6 h, with only a slight difference in the time required for killing preadipocytes compared to adipocytes. Mouse L-929 cells are also susceptible to this combination, but they require 8-10 h for complete lysis of the cells. Apart from the differences in the time course of killing, the cytotoxic effect of TNF and cycloheximide in combination does not appear to be specific for cell type or state of differentiation. The mechanism of killing is not known but apparently has no requirement for ongoing protein synthesis and is effective on a short time scale. These facts may be most simply accounted for by the accumulation of a toxic metabolite of a pathway induced by TNF. Such a metabolite may be a normal product of a TNF-induced signaltransduction pathway that is not metabolized due to the block in protein synthesis. This could in turn imply that a critical, but labile, enzyme may be involved in preventing the accumulation of a toxic metabolite.
The potent synergism between TNF and cycloheximide was utilized to select rare, resistant variants by repeated 6-h pulses of TNF and cycloheximide once a week over a 3-week period. One variant, TA1 R-6, was isolated that is completely resistant to the cytotoxicity induced by TNF in the presence of cycloheximide but is sensitive to TNF alone. This variant has a more transformed appearance, grows to higher densities, is not contact inhibited, and tends to form foci. Furthermore, this cell line does not morphologically differentiate into an adipocyte and is unable to express adipocyte-specific mRNAs, even following treatment with indomethacin and dexamethasone. Interestingly, the TA1 parental cell line only expresses differentiated functions when the cells have reached confluence and have become growth arrested (19,20). The failure of TA1 R-6 cells to become contact inhibited may be related to their failure to express adipocyte-specific RNAs and undergo differentiation to adipocytes. Alternatively, the genetic change that leads to the transformed phenotype may be responsible for the inability of TA1 R-6 cells to differentiate since introduction of known oncogenes into adipogenic cells prevents their differentiation (24).
Many transformed cell lines are susceptible to cytotoxicity by TNF alone whereas normal cells are not. This cytotoxicity is different from the toxicity induced in the presence of cycloheximide since it requires 2-4 days instead of 2-4 h. While the TA1 R-6 cell line is completely resistant to the combined cytotoxicity of TNF and cycloheximide, it is extremely sensitive to the cytotoxicity that occurs with TNF alone. In the parental TA1 cell line, TNF can block the conversion of preadipocytes to adipocytes and is not toxic to them during a 2-4-day exposure. Since transformed cells are susceptible to the cytotoxic effects of TNF alone and normal cells are not, it is possible that the changes that caused the T A 1 R-6 cell line to develop a more transformed phenotype also caused susceptibility to TNF toxicity. Since these cells are fully resistant to the toxicity induced by TNF and cycloheximide in 2-4 h, but are extremely sensitive to the cytotoxicity induced by TNF alone in 2-4 days, the mechanisms by which TNF induces toxicity in each case is likely to be independent.
Since tumor cells, but not normal cells, are susceptible to the cytotoxicity induced by TNF, it was of interest to determine whether the state of differentiation altered the cytotoxic effects of TNF, either alone or in combination with cycloheximide. To test this, mouse L-929 cells were analyzed. Mouse L-929 cells have no adipogenic potential, even when treated with indomethacin and dexamethasone, but are susceptible to TNF. Remarkably, indomethacin and dexamethasone, given concurrently with TNF, are capable of blocking the toxic effects of TNF in these cells. These results indicate the following two things: 1) the ability of these drugs to block TNF cytotoxicity is independent of the capacity of the cells to develop a specific differentiated state, and 2) the mechanism for protection by these drugs is not a phenomenon restricted to the TA1 R-6 cell line.
Dose-response analysis of these drugs in TA1 R-6 and mouse L-929 cells indicates that dexamethasone is active halfmaximally at approximately 10 nM; this is consistent with the protective effect against TNF cytotoxicity being mediated by the glucocorticoid receptor (25). Moreover, dexamethasone can block TNF-mediated repression of ferritin mRNA expression: and it is possible that dexamethasone may act to block the induction or repression of other genes regulated by TNF, genes that may be critical for inducing cytotoxicity. Interestingly, indomethacin is functional only in the range of 30-200 PM. Since complete inhibition of cyclooxygenase occurs with indomethacin at concentrations of 1 FM, these results would indicate that inhibition of cyclooxygenase (and, therefore, prostaglandin production) is not sufficient to block the cytotoxic effects of TNF. We have previously reported that induction of differentiation of TA1 preadipocytes to adipocytes by indomethacin is optimal at 30-200 PM (11). It, therefore, seems plausible that activation of pathways involved in the differentiation of these cells is also involved in the protection of these cells from the cytotoxic effects of TNF and that the mechanisms by which indomethacin and dexamethasone block TNF action are distinct from one another.
Kettelhut et al. (26) recently described the prevention of in uiuo toxic effects of TNF by cyclooxygenase inhibitors. Rats treated with TNF develop toxic shock and die within 2-4 h. Treatment with indomethacin or ibuprofen prior to treatment with TNF prevented the death of these animals. However, they also showed that indomethacin was unable to block the cytotoxic effects of TNF against HeLa, Mouse L-929, and Me 180 cells. These results are in contrast to our data, but may be due to the fact that the highest dose of indomethacin that they tested was 50 PM. This concentration is the lower limit for observable protection with indomethacin for mouse L-929 and TA1 R-6 cells in our experiments, and it is likely that they would not have observed a protective effect for this reason. In addition, there may be some differences in the sensitivity of their cell lines or in the culture conditions used which could account for the fact that they did not observe any degree of protection with 50 PM indomethacin.
Dexamethasone can inhibit the release of arachidonic acid from stimulated cells (25). The mechanism of this inhibition has been shown to require new protein synthesis and is thought to be due to the induction of a specific phospholipase AP inhibitor, lipocortin. Indomethacin, at high concentrations, has also been shown to inhibit phospholipase AZ activities (27). Thus, inhibition of phospholipase Az is potentially the principle mechanism by which these agents inhibit TNF * T. R. Reid and G. M. Ringold, unpublished observations. cytotoxicity; however, no data that is currently available C., Feinman, R., Hirai, M., and Tsujimoto, M. (1986) J. Exp. addresses this issue.

Med. 163,632-643
We have isolated a mutant cell line, TA1 R-6, which is