Assessment of Tumor Necrosis Factor Receptor and Fas Signaling Pathways by Transcriptional Profiling¹

TNF³-α and Fas ligand are two members of a growing superfamily of ligands that bind to their cognate receptors, TNF-R and Fas (CD95), on the cell surface and transmit downstream signals. Both ligands, upon binding to their cognate membrane receptor, cause recruitment of adapter molecules that in turn activate caspases and other proapoptotic signals,eventually leading to cell death (1). TNF-α is also able to elicit a pleiotropic response in most cell types through transcriptional activation of a wide variety of genes (2). Many of these target genes are implicated in immune and inflammatory responses. One of the downstream targets of TNF receptors is the transcription factor NF-κB. NF-κB is sequestered in the cytoplasm by IκB, and upon stimulation by TNF-α, IκB is phosphorylated,ubiquitinated, and targeted to the proteosome for degradation. The free NF-κB dimer, usually consisting of p65 and p50 subunits, then translocates to the nucleus and binds to consensus elements within the promoters of a variety of genes (3). Among the genes activated by NF-κB are genes thought to oppose the proapoptotic effects of TNF. IAPs, for example, are rapidly induced in response to TNF-α and are thought to delay apoptosis through direct inhibition of caspase activity (4). Previous reports have shown that Fas activation can induce secretion of the inflammatory cytokine IL-8, a known NF-κB-responsive gene (5). Furthermore, previous studies have reported that Fas ligation by an agonistic antibody activates NF-κB DNA binding in certain cell types (6)and that Fas ligation can activate the c-Jun NH₂kinase pathway in human colonic epithelial cells (7). Little is known, however, about transcriptional responses to activation of Fas. It is widely believed that the Fas receptor is a “pure killer” in that its activation of caspases and the resulting apoptosis require no transcriptional component. The ability of TNF-αto elicit dual responses of proliferation and cell death, depending on conditions, raises the question as to whether other members of the TNF family, like Fas ligand, are able to do the same.

In light of the evidence that there appear to be cell types and/or conditions under which Fas ligation is able to activate transcriptional pathways similar to TNF-α, we decided to examine the scope of this effect by comparing the downstream transcriptional profiles elicited by TNF-α and anti-Fas. We used cDNA microarrays, containing 4555 minimally redundant genes, to compare the expression profiles downstream of the TNF and Fas receptors in the colon adenocarcinoma cell line, HT29. Both agents were able to induce a largely overlapping set of genes, including a subset of known NF-κB targets like cIAP-2.

HT29 colon carcinoma cells obtained from American Type Culture Collection were grown in McCoy’s 5A medium supplemented with 10%fetal bovine serum, 1.0 mm sodium pyruvate, 2.0 mm l-glutamine, 50 units/ml penicillin G sodium, and 50 μg/ml streptomycin sulfate. Cells were grown at 37°C under 5% CO₂. Media and supplements were obtained from Life Technologies, Inc.

HT29 cells were seeded at 4.5 × 10⁵ cells/well in six-well culture dishes and pretreated with vehicle or 10 ng/ml IFN-γ (Life Technologies, Inc.)for 16 h. Cells were then treated with increasing concentrations of TNF-α (Life Technologies, Inc.) or anti-Fas monoclonal antibody clone CH-11 (Upstate Biotechnology, Inc.). After 24 h of incubation, YO-PRO-1 dye (Molecular Probes, Inc.) was added at a final concentration of 1 μm and incubated for 2 h at 37°C. Cells were analyzed on a fluorescent plate reader (Cytofluor II; Perceptive Biosystems) at excitation and emission wavelengths of 485 and 530 nm, respectively.

Microarray slides were produced as described previously(8). HT29 cells were treated with 10 ng/ml IFN-γ for 16 h and then with 100 ng/ml TNF-α or 100 ng/ml α-Fas antibody for 4 h. IFN-γ was maintained in the medium during the treatment period with each ligand. Total RNA was isolated using Trizol reagent(Life Technologies, Inc.), and poly(A) RNA was selected with Oligotex(Qiagen) according to the manufacturers’ protocols. Labeling and hybridization were performed as described previously (8). Briefly, first-strand cDNA probes were generated by incorporation of Cy3-dCTP or Cy5-dCTP (Amersham Pharmacia) during reverse transcription of 1 μg of purified mRNA. The resulting cDNA probes were purified by vacuum filtration, denatured at 94°C, and hybridized to an arrayed slide overnight at 42°C. Slides were washed in 1× SSC/0.2% SDS for 10 min and then in 0.1× SSC/0.2% SDS for 20 min. Slides were rinsed and dried, and fluorescence was captured using the Avalanche dual laser confocal scanner (Molecular Dynamics). Fluorescent intensities were quantified using Arrayvision 4.0 (Imaging Research).

The software program GeneSpring (version 3.2.2; Silicon Genetics) was used to analyze the array data as follows. The raw fluorescence units of genes represented more than once on each slide were averaged. The ratio was then taken of the signal (IFN-γ plus TNF-α orα-Fas):control (IFN-γ only). To normalize for variation among slides, the ratios (death ligand-treated:control) from each replicate slide were normalized to 1.0 (ratio of gene A in experiment X:median of all ratios measured in experiment X). A Student’s t test was performed to calculate whether the mean relative intensity for a gene was statistically different from 1.0. Control-only comparisons were used to establish significance values for determining induced or repressed genes. For up-regulated genes, we used a significance cutoff of 1.6 (average log ratio >0.204). For down-regulated genes, the significance value was set at 0.45 (average log ratio less than−0.347). The genes exceeding the set significance values, which also had P < 0.085, are presented.

HT29 cells were treated as above for microarray experiments. Total RNA was harvested at appropriate times after treatment, and 7 μg were run on a denaturing formaldehyde agarose gel. RNA was transferred to nylon membrane (Amersham Pharmacia) and cross-linked. Blots were prehybridized at 65°C for 1 h in Rapid-Hyb buffer (Amersham Pharmacia) and then hybridized with random-primed,[α-³²P]dCTP-labeled probes for 4–16 h at 65°C. Unbound probe was removed by washing twice in 2× SSC/0.1% SDS at room temperature and then twice in 0.5× SSC/0.1% SDS at 65°C. Blots were exposed to a phosphor screen and scanned using a Phosphorimager (Molecular Dynamics).

Treated HT29 cells were rinsed with cold PBS, scraped,transferred to tubes, and spun at 350 × gfor 5 min. Pellets were washed in 0.5 ml of buffer A [10 mm HEPES-NaOH (pH 7.8), 15 mm KCl, 2 mmMgCl₂, 0.1 mm EDTA, 1 mm DTT, and 1 mm PMSF] and resuspended in 0.2 ml of buffer A. After incubation on ice for 10 min,NP-40 (Igepal; Sigma) was added to a final concentration of 0.5%, and extracts were spun at 1330 × g for 15 min. Supernatants were removed, and nuclear pellets were resuspended in 30μl of buffer B [20 mm HEPES-NaOH (pH 7.8), 1.5 mm MgCl₂, 0.5 mm DTT, 420 mm NaCl, 0.2 mm EDTA, 25% glycerol, and 0.5 mm PMSF]. Suspended pellets were rocked vigorously at 4°C for 15 min and spun at 16,300 × g for 10 min. Supernatants were transferred to new tubes and brought to a final volume of 80 μl with buffer C [20 mm HEPES-NaOH (pH 7.9), 50 mm KCl, 0.2 mm EDTA, 0.5 mm DTT, and 0.5 mm PMSF]. Protein concentrations were determined using the Bio-Rad protein assay kit.

A 22-mer oligonucleotide (3.5 pmol) with a NF-κB consensus binding site (Promega) was 5′ end-labeled using T4 kinase (Promega) and purified over a Chromaspin-10 column (Clontech) to remove unincorporated [γ-³²P]ATP. For binding reactions, 6 μg nuclear extract were incubated with 1 μl(50,000–200,000 cpm) of ³²P-labeled probe in 10μl of final volume of 1× binding buffer [20% glycerol, 5 mm MgCl₂, 2.5 mm EDTA,2.5 mm DTT, 250 mm NaCl, 50 mmTris-HCl (pH 7.5), and 0.25 mg/ml poly(deoxyinosinic-deoxycytidylic acid)-poly(deoxyinosinic-deoxycytidylic acid)]. For competition assays, 50-fold molar excess of unlabeled oligo or 2 μg of anti-p65,anti-p50, or anti-c-Rel (Santa Cruz Biotechnology, Inc.) were added 15 min prior to addition of labeled oligonucleotide. Complexes were separated using SDS-PAGE on a 6% gel in 0.5× TBE (45 mmTris-borate, 1.0 mm EDTA) at 30 mA for 4 h. Gel was dried and exposed to phosphor screen and visualized on a Phosphorimager (Molecular Dynamics).

Whole-cell extracts were prepared from treated HT29 cells as follows. Cells were washed twice in cold PBS, incubated on ice for 10 min in lysis buffer [25 mm Tris-HCl (pH 7.4), 150 mmNaCl, 1 mm CaCl₂, 1% Triton X-100, 1 mm PMSF, 10 μg/ml leupeptin, and 10 μg/ml aprotinin],scraped from plate into tube, and spun at 14,000 rpm for 10 min at 4°C. Supernatants were removed and assayed for protein concentration using Bio-Rad protein assay kit. Protein extracts were analyzed using SDS-PAGE on 10% Tris-glycine gels (Novex) and transferred to polyvinylidene difluoride membranes (Gelman Sciences). Blots were incubated 1 h at room temperature in blocking buffer (5% powdered milk, PBS, and 0.1% Tween 20) and then with a 1:1000 dilution of either a polyclonal antibody to IκB-α or a phosphospecific antibody to serine 32 of IκB-α (New England Biolabs) in primary antibody buffer (5% BSA, PBS, and 0.1% Tween 20) overnight. Blots were washed three times in PBS/0.1% Tween 20 and then probed with a secondary antibody conjugated to horseradish peroxidase (Life Technologies, Inc.)for 30 min. After washing three times in PBS/0.1% Tween 20, protein was detected using a chemiluminescent substrate (DuPont NEN Life Sciences) according to manufacturer’s protocol.

IFN-γ in combination with TNF-α or α-Fas antibody act synergistically to induce apoptosis in a variety of cell types(5). To establish a concentration curve from which to select treatment conditions for microarray analyses, we preincubated HT29 colon adenocarcinoma cells with 10 ng/ml IFN-γ for 16 h and then treated these cells with increasing concentrations of TNF-α or the agonistic α-Fas CH-11 antibody. We then assayed for cell death by measuring uptake of YO-PRO-1 dye. As shown in Fig. 1, when cells were pretreated with IFN-γ, both TNF-α and α-Fas induced killing in a saturable, concentration-dependent manner. Because cells without IFN-γ treatment were relatively resistant to killing mediated by either receptor, our subsequent experiments included pretreatment of the cells with IFN-γ.

To examine the transcriptional pathways activated downstream of the TNF and Fas receptors, we used cDNA microarrays (9) to monitor transcriptional changes in HT29 cells upon treatment with TNF-α orα-Fas antibody. HT29 cells, preincubated with IFN-γ for 16 h,were treated with vehicle or a saturating concentration of either ligand for 4 h. Each culture, vehicle or ligand-treated, contained an equal concentration of IFN-γ during the treatment period. After incubation, mRNA was harvested and used to generate labeled cDNA probes(IFN-γ alone or IFN-γ plus either TNF-α or α-Fas). The probes were simultaneously hybridized to a microarray slide containing 4555 cDNAs in duplicate (Fig. 2,A). In addition, to assess the technical capabilities of our arrays, we generated probes labeled with each dye from untreated,control HT29 cells and cohybridized them in parallel with our experimental samples. Slides were analyzed by quantifying fluorescence intensities and plotting the fluorescence of control (IFN-γ only) versus treatment (IFN-γ plus TNF-α or α-Fas; Fig. 2,B). We then calculated the log of the relative intensity ratios (IFN-γ plus TNF-α or α-Fas:IFN-γ only) and averaged the four replicates. Control-only comparisons were used to establish significance values for determining induced or repressed genes. On the basis of this, we selected genes displaying expression ratios of >1.6(log value >0.204) or <0.45 (log value less than −0.347), which also had Ps of <0.085. These genes were sequence verified and are presented in Table 1.

Striking in our analysis was that TNF-α treatment resulted in an asymmetric distribution of induced versus suppressed genes. Treatment of HT29 cells with TNF-α caused the induction of 29 genes that exceeded our cutoff of 1.6 (log ≥0.204). In contrast, no genes exceeded the suppression levels established in our control only comparison. This indicates that the early transcriptional program elicited by TNF is aimed primarily at induction of target genes. It is possible that this distribution would change over time to generate a more symmetrical distribution; however, these data demonstrate the potential for asymmetry in the activation of transcriptional programs by specific ligands.

We categorized the up-regulated genes into three main groups:

(a) Fifteen of the 29 genes were known to be TNF-α or NF-κB regulated, including the cytokines IL-8 and GRO-2 (GRO-β) and the antiapoptosis gene apoptosis inhibitor 2 (cIAP-2; Ref.10). These genes served as positive internal controls and confirmed the validity of our microarray approach.

(b) As a second category, 10 of the 30 genes regulated by TNF-α were genes of known function that lacked previous description, implicating them downstream of the TNF receptor. Interestingly, genes in this category included genes that have been reported to be antiapoptotic. For example, one of the genes, 14-3-3η (tyrosine 3-monooxygenase/tryptophan 5-monooxygenase activation protein), belongs to a family of proteins that has been shown to inhibit cell death by binding to the apoptosis-promoting proteins BAD and forkhead transcription factor(FKHRL1; Ref. 11).

(c) Finally, we observed four TNF-α-regulated genes of unknown function that corresponded only to ESTs.

Several of the TNF-α-induced genes, including small inducible cytokine member 10, Humig (monokine induced by IFN-γ), and guanylate binding protein 2 are known targets of the IFN response pathway (8). It was possible, therefore, that the induction of these genes by TNF-α was occurring through increased cell sensitization to IFN-γ. We believe this unlikely in that many other known IFN response genes were absent for the TNF-α-inducible list (8). To address this directly, we performed array experiments wherein IFN-containing medium was removed and replaced with medium containing the death ligands only. Under these conditions, we saw the same TNF-α induction pattern, including those genes that are also targets of IFN. Further investigations will be required to fully characterize the integration of signaling pathways that emanate from the TNF receptors and how these pathways couple with specific transcription factor complexes.

Treatment of HT29 cells with α-Fas elicited a gene expression profile that almost directly overlapped with TNF-α. Table 1 shows the genes regulated by TNF-α and the response of these genes to α-Fas antibody. In this analysis, six genes were induced by Fas ligation to a level that exceeded the log expression ratio of 0.204. Five of the six genes were also induced by TNF-α, suggesting a common signaling pathway. There was only one gene (coagulation factor XIIIa)that was highly induced by α-Fas that failed to meet the criteria for induction by TNF-α. Our data indicate a general activation of NF-κB response genes by α-Fas and extend the previous results showing that Fas ligation stimulates IL-8 production (5).

We saw many more genes that were repressed, relative to the control comparison, in cells responding to α-Fas (54 genes) than were evident in the TNF-α-treated cells (0 genes). Because both TNF-α andα-Fas induce cell death in HT29 cells, we believe that many of these repressed changes could be attributable to the cells’ metabolic state,not attributable directly to repression. For example, basal levels of transcription or message stability could contribute to a nonspecific loss of certain mRNAs. Because we cannot distinguish between specific and nonspecific message loss, we have elected to present only the up-regulated targets of TNF-α andα-Fas.⁴

We then selected two genes, IL-8 and cIAP-2, with roles in inflammation and cell death regulation for further analysis. In addition, IL-8 and cIAP-2 were highly and moderately induced, respectively, by both ligands. We performed Northern blot analyses to monitor the temporal induction of each gene and to confirm the induction seen by microarray analysis (Fig. 2,C). HT29 cells pretreated with IFN-γ were treated with saturating concentrations of TNF-α and α-Fas, and RNA was harvested at various times after stimulation. TNF-α induced cIAP-2and IL-8 rapidly, with maximal expression of both transcripts occurring at 2 h. The kinetics of induction, however,differed for the two transcripts, with cIAP-2 message returning to basal level by 8 h and IL-8 remaining elevated. The induction of both genes by TNF-α was only slightly enhanced by pretreatment with IFN-γ. In contrast, α-Fas antibody caused an induction of both genes that was temporally delayed in comparison to TNF-α (Fig. 2 C). In the case of Fas ligation, maximal expression of both genes occurred at 6 h and was sustained for at least 8 h. Unlike TNF-α, pretreatment with IFN-γ was required for the induction of cIAP-2 and greatly enhanced the induction of IL-8 by α-Fas.

The numerous NF-κB-regulated genes induced in our microarray experiments led us to investigate whether α-Fas activated NF-κB DNA binding in HT29 cells. To determine whether NF-κB DNA binding activity was present in cells treated with α-Fas, we performed electrophoretic mobility shift assays on nuclear extracts from HT29 cells (Fig. 3,A). Binding to a NF-κB consensus oligonucleotide was evident in both TNF-α- and α-Fas-treated extracts. Antibody supershift experiments revealed that both p65 and p50 subunits of NF-κB bound to the DNA in response to either TNF-α or α-Fas (Fig. 3,A). Consistent with the transcriptional induction of cIAP-2 and IL-8 (Fig. 2,C), activation of NF-κB by TNF-α was only slightly augmented by pretreatment of the cells with IFN-γ. Induction of NF-κB by α-Fas, however, was greatly enhanced by IFN-γ. The binding induced by α-Fas, however,was weaker than that elicited by TNF-α, an observation that was consistent with the weaker induction of response genes by α-Fas as seen in the microarrays (Table 1).

We next examined the mechanisms through which Fas ligation leads to NF-κB activation. It was possible that α-Fas binds other cell surface receptors, distinct from Fas, which cause NF-κB activation. This was of particular concern, given the recent findings that several agonistic Fas antibodies show cross-reactivity with other cellular proteins (12). To exclude this possibility, we pretreated HT29 cells with a Fas-specific neutralizing antibody (ZB4; Upstate Biotechnology, Inc.) that recognizes an antigen distinct from the CH-11 activating antibody. The cells were treated with 500 ng/ml ZB4 for 1 h prior to stimulation with 100 ng/ml CH-11 for 4 h. RNA was harvested and analyzed by Northern blot. In these experiments,neutralizing antibody protected the cells from Fas-mediated apoptosis and inhibited Fas-mediated induction of IL-8 message by at least 70%. This indicated that the CH-11 antibody elicited death and transcriptional responses directly through Fas, not through nonspecific interactions.

Another possible mechanism for indirect activation of NF-κB after Fas ligation includes the processing of inflammatory cytokines by activation of caspases. For example, caspases can process pro-IL-1 and pro-IL-18 in response to Fas ligation and elicit an inflammatory response (13, 14). To address this possibility, we tested whether inhibition of caspases prevented induction of NF-κB response genes. We incubated HT29 cells with 10 μmZ-Val-Ala-Asp-fluoromethyl ketone, a general inhibitor of caspases, for 1 h prior to stimulation with α-Fas for 4 h and then measured the induction of cIAP-2 and IL-8 by Northern analysis. Although the Z-Val-Ala-Asp-fluoromethyl ketone blocked cell death, it failed to prevent induction of cIAP-2or IL-8 (data not shown). This eliminated caspases as generators of NF-κB activators and supported that Fas ligation was directly causing gene transcription in HT29 cells.

The most well-described method of NF-κB activation occurs through phosphorylation and subsequent degradation of IκB-α. We,therefore, analyzed the levels of IκB-α at various times after stimulation. Whole-cell lysates were prepared and analyzed by probing a Western blot with a polyclonal antibody to IκB-α. As expected,TNF-α caused a rapid degradation of IκB-α beginning at 5 min,with the protein being completely degraded by 15 min (Fig. 3 B). Surprisingly, α-Fas failed to cause any visible IκB-α degradation up to 20 min after stimulation. Phosphorylation of IκB-α by TNF-α was evident at 5 min, whereas there was no detectable phosphorylation by α-Fas for up to 20 min (data not shown). We then examined up to 4 h after stimulation by α-Fas and did not observe degradation of IκB-α (data not shown). Although lack of degradation was also true of IκB-β (data not shown), there are several additional isoforms of IκB, including ε, γ, Bcl3,p105, and p100 (10) through which α-Fas may activate NF-κB. The mechanism, however, appears distinct from that used by TNF-α.

Our data show that microarrays are capable of establishing gene expression profiles for members of the TNF family of death receptors. From such signature patterns, we were able to determine that ligation of the Fas receptor activates a NF-κB pathway, resulting in the transcriptional induction of genes thought to be part of inflammatory or survival pathways. Many human tumors overexpress Fas ligand(15, 16). This has led to speculation that increased levels of Fas ligand on the cell surface cause apoptosis of infiltrating lymphocytes, allowing the tumor cells to escape immune surveillance (15, 16, 17). In contrast, it is thought that NF-κB activation can elicit an antiapoptotic pathway by inducing the transcription of “survival” genes (18). Among these are apoptosis inhibitor 2 (cIAP-2) and the cytoprotective gene, manganese superoxide dismutase(SOD-2). We saw that each of these genes was up-regulated by either TNF-α or α-Fas under the same conditions that caused the cells to undergo apoptosis. Regulation of these genes by α-Fas has not been reported and contradicts the widespread notion that Fas signaling leads only to cell death. Fas-mediated activation of NF-κB genes, as demonstrated here, supports a role for Fas in inflammation. It is unclear, however, whether induction of genes like cIAP-2 alone is sufficient to block Fas-mediated apoptosis. Remacle-Bonnet et al.(19) demonstrated recently the inability of insulin-like growth factor I to protect HT29-D4 cells from Fas-mediated killing unless the cells were simultaneously exposed to TNF-α. This suggests that activation of pathways, like the mitogen-activated protein kinase/extracellular signal-regulated kinase cascade may be required as supplements to IAP-2 induction in blocking death receptor mediated apoptosis. Induction of cIAP-2 by α-Fas, however, is in agreement with studies demonstrating that transplanted Fas ligand-bearing tumors elicit an inflammatory response rather than conferring immune privilege (20). Finally, our data provide new insights that activation of Fas in some cancer cells can elicit a response similar to the inflammatory response seen with TNF-α. This suggests the possibility that overexpression of Fas ligand in tumors may induce a chronic inflammatory response, which could promote tumor development.