A Human Suppressor of c-Jun N-terminal Kinase 1 Activation by Tumor Necrosis Factor α*

Tumor necrosis factor α (TNFα) has pleiotropic effects on cellular metabolism. One of the signaling paths from the TNFα receptor induces a stress-activated protein kinase cascade. Components within this TNFα kinase cascade include mitogen-activated protein kinase/extracellular signal-regulated kinase kinase kinase 1 (MEKK1) and stress-activated protein kinase/extracellular signal-regulated kinase kinase (SEK), which regulate the activity of c-Jun N-terminal kinase 1 (JNK1). Currently, molecules upstream of MEKK1 that link TNFα receptor to downstream kinases are not well understood. Besides TNFα, many other stimuli including several oncoproteins can activate JNK1. In most cases, the signaling cascade(s) leading from oncoproteins to JNK1 is poorly elucidated. We report here that the human T-cell lymphotrophic virus, type I (HTLV-I) oncoprotein, Tax, can activate JNK1. We isolated a novel human cell factor, G-protein pathway suppressor 2 (GPS2), by its ability to bind the HTLV-I oncoprotein, and we show that this factor can potently suppress Tax activation of JNK1. In trying to understand the mechanism of GPS2 activity, we found that it also suppressed TNFα activation of JNK1 but not TNFα activation of p38 kinase nor phorbol activation of extracellular signal-regulated kinase 2. Because GPS2 has minimal effect on MEKK1- or SEK-regulated JNK1 activity, it could act at a point between the TNFα receptor and MEKK1 in the initial step(s) of this kinase cascade. Alternatively, it is not excluded that GPS2 could work in a parallel pathway that leads from TNFα to JNK1. GPS2 represents a new molecule that could contribute important insights toward how cytokine- and oncoprotein-mediated signal transduction might converge.

We have been interested in how viral proteins influence cellular signaling pathways. Because viruses are obligatory cellular parasites, it is reasonable that viral gene regulators mimic cellular counterparts. Accordingly, viruses ranging from oncogenic retroviruses (3) to hepatitis virus (15) to herpesvirus (16) have been documented to affect Ras-dependent kinase cascades. In human T-cell lymphotrophic virus (HTLV-I), a viral oncoprotein, Tax, is known to transcriptionally modulate the expression of cellular immediate-early genes including cjun (17,18). The mechanism through which Tax affects this class of cellular genes is unclear and is an area of intense investigation. Because it cannot bind DNA directly but can pleiotropically activate many promoters that do not share a common responsive motif (19 -23), Tax likely functions through protein-protein contacts and/or signaling cascades (reviewed in Ref. 24). Some cellular proteins that bind Tax have been described (Refs. [25][26][27]reviewed in Ref. 24). However, how these factors might mediate the various activities ascribed to Tax is incompletely understood.
To better understand functional mechanisms, we searched for cellular factors that might bind Tax and subserve its transcriptional activity. We also asked if Tax could activate a * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) U28963.
STE4 Hpl Suppression Assay-Suppression of STE4 Hpl lethality was performed essentially as described (14). Briefly, transformed DBC yeasts were plated onto SD-leucine-arginine ϩ canavanine medium to select for the suppression plasmid and against the pU␣2C maintenance plasmid. Viable cells were counted after 3 days of cultivation. Experiments were in triplicate.
Reporter Assays-FUS2-LacZ reporter assay was performed in YM4271 yeasts harboring either pEB39 or pEB39 ϩ GPS2 after growing in selective media to an A 600 of 0.5-0.8. Cells were induced with 5 g/ml ␣-factor. Aliquots were collected at the indicated times and assayed for ␤-galactosidase activity using chlorophenol red-␤-D-galactopyranoside (CPRG) as substrate. Yeast cells were disrupted by freezethaw and by additional vortexing with acid-washed glass beads. ␤-Galactosidase was assayed using the method of Miller (33), except that CPRG was used for color development and the reaction buffer consisted of 100 mM HEPES, 150 mM NaCl, 2 mM MgCl 2 , 0.5 mM L-aspartate, 1% bovine serum albumin, and 2 mM CPRG. ␤-Galactosidase activity was expressed as relative CPRG units, defined in the same manner as Miller units (33). Results are triplicate determinations of duplicate transformants. Kinase assays using HA-tagged kinases were performed as described previously (13).
Northern Blot Analysis-Human multiple tissue Northern blots (CLONTECH) were probed with a 32 P-labeled 1.2-kilobase GPS2 and/or a 2-kilobase human ␤-actin cDNA probe as per manufacturer's protocol.
Protein Affinity Chromatography -E. coli-produced proteins GST, GST-GPS2, and His-Tax were purified according to manufacturers' protocols (Pharmacia Biotech Inc. and Invitrogen). For binding studies, proteins were incubated either with glutathione-Sepharose (Pharmacia) or Probond resin (Invitrogen). GST and GST-GPS2-bound resins were washed extensively with phosphate-buffered saline, whereas the His-Tax-containing resin was washed with buffer A (50 mM sodium phosphate, pH 8.0, 300 mM NaCl, 0.5 mM ␤-mercaptoethanol, and 1 mM phenylmethylsulfonyl fluoride) containing 40 mM imidazole. GST and GST-GPS2 resins were equilibrated with buffer B (20 mM HEPES-KOH, pH 7.9, 20 mM KCl, 1 mM MgCl 2 , 17% glycerol, and 2 mM dithiothreitol). His-Tax was eluted from Probond resin with buffer A containing 300 mM imidazole. The His-Tax-containing eluate (ϳ5 ng/l His-Tax) was diluted 10-fold by the addition of buffer B and incubated with the GST and GST-GPS2 resins. A flow-through fraction was collected. The resins were washed with buffer B and then eluted with buffer B containing 0.1, 0.25, and 0.5 M KCl. 1 ml of the last buffer B wash and each eluate was then trichloroacetic acid-precipitated to concentrate the proteins. The samples were then separated by 10% SDS-PAGE, transferred to membrane, probed with rabbit anti-Tax serum, and visualized by chemiluminescence (Western-Light; Tropix Inc.).
Immunoprecipitations and Western Blot Analysis-Protein-protein complexes were immunoprecipitated under native conditions as described previously (25). Immunoprecipitates and proteins from total cell extracts were solubilized in SDS gel loading buffer (60 mM Tris base, 2% SDS, 10% glycerol, 5% 2-mercaptoethanol). Samples containing equal amounts (20 g) of protein were separated by 12% SDS-PAGE and electroblotted onto Immobilon-P membranes (Millipore Corp.) using a Millipore semi-dry blotting apparatus. Immunoblotting was performed with rabbit antisera raised against Tax (25) or against synthetic GPS2 peptides (amino acids 44 -69 and 307-327). Blots were visualized by chemiluminescence using goat anti-rabbit alkaline phosphatase-conjugated antibody. The primary and secondary antibodies were used at 1:1,000 and 1:10,000 dilutions, respectively.

RESULTS AND DISCUSSION
In searching for human proteins that might bind HTLV-I Tax, we screened 1 ϫ 10 8 individual transformants from a HeLa S3 cDNA library using the yeast two-hybrid approach

FIG. 1. Delineation of the interaction between HTLV-I Tax and
GPS2 by the yeast two-hybrid assay. A, representative ␤-galactosidase filter assays performed on HF7c yeast expressing the indicated Gal4bd (BD) or Gal4ad (AD) fusion proteins. Blue colonies on the filters, after 6 h of incubation with chromophore, photograph as black dots. BD-p53 ϩ AD-SV40/T antigen was used as a positive control. BD-GPS2 ϩ AD-SV40/T and BD-p53 ϩ AD-Tax were used as negative controls. B, a summary of qualitative and quantitative readouts from three different reporter assays. A (ϩ) filter assay was based qualitatively on the criterion that colonies turn blue within 6 h of incubation with X-gal. Separately, ␤-galactosidase activity was quantified based on relative CPRG units, defined in the same manner as Miller units (see "Materials and Methods"). Lastly, for the His3p assay, colonies that survived histidine dropout in the presence of 10 mM 3-aminotriazole (3AT) were scored as (ϩ), whereas colonies that failed to survive after 60 h of observation were scored as (Ϫ). (34). Five reproducibly reactive independent transformants were identified. One contained a 1,165-base pair human cDNA which encoded a protein that bound Tax but not SV40 T antigen (Fig. 1A) nor p53 (Fig. 1B).
We sequenced this cDNA completely on both strands and found it to encode an open reading frame of 327 amino acids with a predicted molecular size of 37 kDa ( Fig. 2A). A homology search was performed using the BLAST server from the National Center for Biotechnology Information. 2 Our initial search uncovered no relatives in the data bases. However, at time of our sequence deposition, another cDNA differing minimally in nucleotides but extensively in amino acids was deposited concurrently from an independent group (J. Colicelli, UCLA). Based on our sequence data, the amino acid errors in that deposition were corrected, revealing that the two independently isolated cDNAs encoded for identical proteins (Gen-Bank TM accession number of the correct sequence is U28963). By mutual agreement, this protein was named GPS2 (G-protein pathway supprressor 2).
Although GPS2 is a human protein, it has motifs and homologies 3 that suggest relatedness to previously described suppressors of a G ␣ null allele in yeast (14). Heterotrimeric G ␣,␤,␥ proteins participate in yeast pheromone-regulated metabolism. In principle, loss of G ␣ manifests as lethality resulting from unchecked signaling through G ␤,␥ . We asked whether GPS2 plays a role in regulating this signaling cascade in yeast, with the thought that a positive finding would lead us to explore an analogous function for this protein in mammalian cascade(s). In Fig. 3A, we expressed full-length GPS2 protein (GPS2 (1-327)) from an ADH1 promoter on a high copy plasmid in yeasts that have a constitutively lethal STE4 Hpl mutation. Yeasts transformed with vector alone showed poor viability (Fig. 3A). On the other hand, GPS2 expression increased yeast viability (GPS2 (1-327); Fig. 3A) by more than 12-fold over that seen with vector alone. Deletion analyses narrowed the portion necessary for suppressive activity to be within residues 23-223 (Fig. 3A). How this domain functions remains to be clarified; however, one limited interpretation of STE4 Hpl suppression is that GPS2 interrupts signaling at a step downstream of the Ste4-encoded G ␤ molecule.
The above finding suggests that GPS2 can interdict a cascade in yeast that leads to activation of pheromone-responsive genes. To test this hypothesis directly, expression of a pheromone-responsive gene, FUS2, was measured in the presence or absence of GPS2 (Fig. 3B). Yeasts were first transformed with either FUS2-LacZ reporter plasmid (pEB39) alone or the pEB39 reporter together with GPS2 and then treated with ␣-factor. Induction of LacZ was monitored (Fig. 3B) over time. Comparedwithcontrol,GPS2significantlyinhibitedpheromonedependent expression of FUS2. This supports that GPS2 functions at a step in the cascade between pheromone receptor and pheromone-responsive genes. (Since this work was completed and first submitted, these findings have been independently confirmed by Colicelli and colleagues; Ref. 35).
Yeast heterotrimeric G-protein signaling and mammalian MAPK pathways are closely conserved (9). The observations on GPS2 in yeast suggest that this human protein might serve a related, and perhaps more authentic, function in primate cells (i.e. a general role in a conserved signal transduction pathway).
To check this hypothesis, we first wished to verify that GPS2 mRNA and protein are indeed generally expressed in human tissues and cells. In Fig. 4, mRNAs from 16 different human tissues and 8 different established human cell lines were analyzed using a GPS2-specific probe. We found that GPS2 mRNA was expressed abundantly in all the tested tissues and cells (Fig. 4A). A smaller survey of primate and murine cell lines showed that the GPS2 protein expression profile is consistent with its observed RNA pattern (Fig. 4B). On Western blotting, we noted that a GPS2-specific 37-kDa band is commonly present in human (Fig. 4B, lanes 3, 4, and 6, Jurkat, C81, HeLa) and monkey (Fig. 4B, lane 1, COS) cells. (Interestingly, at this exposure, GPS2-specific signal was not identified in mouse cells (Fig. 4B, lanes 2 and 5, F9, NIH 3T3). However, in other experiments a more weakly reactive band of 43 kDa was seen in murine cells using the rabbit antisera raised against human GPS2 (data not shown), suggesting that a nonidentical, but related, protein is conserved in mouse cells.) We did find that transfection of a GPS2 expression plasmid into HeLa cells could enforce overexpression of protein (Fig. 4C). The ubiquitous and relatively abundant expression of GPS2 in primate tissues/cells is compatible with its serving a general function in a conserved signal transduction (i.e. MAPK) pathway.
Previously, it was shown that Tax induces the expression (17,18) and the activity (36) of c-Jun and that c-Jun plays an important role in regulating transcription from the HTLV-I long terminal repeat (25,36). Furthermore, it is well established that one of the mammalian MAPK pathways that regulates c-Jun activity is mediated through JNK1 (37-40; reviewed in Ref. 41). These facts, coupled with the suggested role for GPS2 in MAPK signaling and its ability to bind Tax (Figs. 1 and 2), provided clues that logically directed us to explore a link between the HTLV-I oncoprotein and JNK1. The ability of Tax to activate JNK1 was checked using a transient transfection in vitro kinase assay (Fig. 5). We co-introduced into HeLa cells an expression plasmid for a hemagglutinin epitope-tagged JNK1 protein (HA-JNK1) and pUC19 carrier (Fig. 5A, lane 1), a GPS2 expression plasmid (Fig. 5A, lane 2), a Tax expression plasmid (Fig. 5A, lane 3), a plasmid encoding a constitutively activated MEKK1 mutant protein (Fig. 5A, lane 4 MEKK⌬; Ref. 31), or pMEKK⌬ ϩ GPS2 (Fig. 5A, lane 5). We checked for activation of HA-JNK1 by immunoprecipitating cell lysates 3 J. Colicelli, personal communication. with a monoclonal antibody directed against the HA epitope followed by assaying for GST-cJun1-79-phosphorylating activity in the immunoprecipitates (Fig. 5A). After normalizing for protein recovery by Western blotting for HA-JNK1 (Fig. 5A,  bottom panel), we verified that expression of MEKK⌬, as expected (31) ). Base-line JNK1 activity in these assays was low (Fig. 5A,  lane 1). In this setting, GPS2 overexpression had no inhibitory effects on the basal HA-JNK1 function (Fig. 5A, lane 2), even upon deliberate overexposures of film. GPS2 overexpression also failed to affect the activity of MEKK⌬ (Fig. 5A, lane 5). One interpretation consistent with the latter observation is that GPS2 functions at a step upstream of MEKK1.
Because HTLV-I is a human T-cell virus, in principle a more physiological test for Tax-JNK1 activation would be one performed in human Jurkat T-cells. Hence, we repeated the experiment coelectroporating HA-JNK1 with effector plasmids into Jurkat cells (Fig. 5B). In Jurkat cells, the magnitude of JNK1 activation by MEKK⌬ (Fig. 5B, lane 4) and by Tax (Fig.  5B, lane 2) was similar to that in HeLa cells. Notably, we observed that simultaneous overexpression of GPS2 reproducibly suppressed Tax activation of JNK1 (Fig. 5B; compare lane 3 with lane 2). A similar finding has also been documented for adherent cells (data not shown), thus providing a functional correlate to the physical contact between Tax and GPS2. We also noted that, as in HeLa cells (Fig. 5A, lane 5), GPS2 overexpression did not affect the activity of MEKK⌬ (Fig. 5B,  lane 5).
Many stimuli that trigger mammalian MAPKs affect either the c-Jun or the c-Fos component of AP-1 (37-40; reviewed in Ref. 41). Intracellular AP-1 activity can be assayed using a responsive reporter plasmid (e.g. pAP-1CAT; Ref. 30). To validate the biology of the above in vitro biochemical results, we wished to understand the intracellular effects of GPS2 on stimuli that activate JNK1 (Fig. 6). Hence, we transfected HeLa cells with pAP-1CAT (Fig. 6A, lanes 1 and 2) or cotransfected cells with pAP-1CAT ϩ MEKK⌬ plasmid ϩ escalating amounts  11 and 14) g of pGPS2 plasmid was induced by treatment with TNF␣ (Calbiochem; 10 ng/ml final concentration, lanes 9 -11), or TPA (Sigma; 25 ng/ml final concentration, lanes [12][13][14]. Total amounts of transfected DNAs were normalized to 7 g in all samples with addition of carrier plasmid. TNF␣ or TPA was added to cells 24 h after transfection, and CAT activities were assayed 48 h after transfection. Left panel, a representative assay; right panel, graphic representation of averages from three separate experiments. B, comparison of the effect of GPS2 on TNF␣ or UV light-activated expression. pAP-1CAT was transfected with increasing amounts of cotransfected pGPS2 into HeLa cells as in panel A. The samples in lanes 3-5 were treated with TNF␣ (10 ng/ml final concentration), whereas those in lanes 6 -8 were treated with 40 J/m 2 in a manner previously described (42). CAT activities were assayed 48 h after transfection. Left panel, a representative assay; right panel, graphic representation of averages from three separate experiments. TNFa, TNF␣. of GPS2 expression plasmid (Fig. 6A, lanes 3-5) or pAP-1CAT ϩ SEK plasmid ϩ escalating amounts of GPS2 expression plasmid (Fig. 6A, lanes 6 -8). Alternatively, HeLa cells were transfected with pAP-1CAT ϩ escalating amounts of GPS2 plasmid and treated with either TNF␣ (Fig. 6A, lanes 9 -11) or TPA (Fig. 6A, lanes 12-14). CAT activities were assayed 48 h after transfection, the experiments were repeated three times, and the data were tabulated (Fig. 6A, right graph). We found that on average MEKK⌬ (Fig. 6A, lanes 3-5), SEK (Fig. 6A,  lanes 6 -8), TNF␣ (Fig. 6A, lanes 9 -11), and TPA (Fig. 6A, lanes  12-14) activated pAP-1CAT expression by 8.7-, 4.2-, 10.2-, and 15-fold, respectively. Notably, coexpression of GPS2 suppressed TNF␣-mediated (Fig. 6A, lanes 9 -11) activation of pAP-1CAT but did not affect MEKK⌬- (Fig. 6A, lanes 3-5), SEK- (Fig. 6A, lanes 6 -8), or TPA- (Fig. 6A, lanes 12-14) responsive expression. In a parallel series of transfections, we also found that UV induction of pAP-1CAT activity (Fig. 6B, lanes 6 -8) was also unaffected by GPS2 overexpression. TNF␣, MEKK1 and SEK are sequential components of the same signal cascade that leads to JNK1 (reviewed in Ref. 4); TPA is known to activate AP-1 through ERK (40), and parts of the inductive activity of UV have been attributed to the activation of CSBP/p38 MAPK and or JNK1 (42). Thus a compatible interpretation (that does not exclude other interpretations) of the pAP-1CAT results is that GPS2 affects the JNK1 cascade at a step after TNF␣ and before MEKK1 and that GPS2 does not affect activation of ERK nor CSBP/p38 MAPK. To the extent that UV treatment can lead to JNK1 activation, our results suggest that GPS2 does not interdict this path.
Findings up to this point suggest that GPS2 plays specific roles in the linkage between JNK1-Tax and JNK1-TNF␣. For the former, GPS2 is a direct binding protein whose expression represses the ability of Tax to activate JNK1 (Fig. 5). Because Tax is a potent activator of c-Jun expression and activity (17,18), a question arises as to how this might occur despite GPS2 activity. One possibility is that Tax might regulate GPS2 through a loop-back mechanism. We investigated this hypothesis using a Jurkat cell line that can inducibly express Tax upon treatment with metals such as CdCl 2 (26). We treated Jurkat (Fig. 9, lanes 1-3) and Jurkat-Tax (Fig. 9, lanes 4 -6) cells with CdCl 2 and followed the time course of Tax ( Fig. 9, left blot) and GPS2 (Fig. 9, right graph) expression. (Quantitation of GPS2 expression was based upon the average scanning results from Western blotting signals from three independent experiments. For normalizations, each film exposure also contained an escalating series of known amounts of Tax protein that was reacted against anti-Tax serum.) Whereas Jurkat cells were essentially unperturbed by metal treatment (Fig. 9, right, Jurkat), we noted that at 12 h after treatment (Fig. 9, right, Jurkat-Tax) maximal Tax expression in Jurkat Tax cells (Fig. 9, lane 6) was accompanied by significantly reduced steady-state amounts of GPS2. This finding was replicated in three separate assays, and the results showed an average of a 4-fold drop in GPS2 expression at 12 h after metal treatment (Fig. 9, right graph).
Here, we have described the isolation and characterization of a human protein, GPS2, that binds the HTLV-I Tax oncoprotein and suppresses Tax-and TNF␣-mediated activation of JNK1. Genetic analysis in yeast indicates that GPS2 functions downstream of Ste4-encoded G ␤ subunit. One interpretation suggests that GPS2 acts after the TNF␣ receptor and upstream of the MEKK1 kinase. Our data do not exclude that GPS2 could work to interdict a pathway from TNF␣ to JNK1 that might not involve MEKK1. Taken together, the findings here are compatible with a possible locale for GPS2 function close to the cyto-solic side of the TNF␣ cell surface membrane receptor. We note that GPS2 is specific to this route of signaling, since it does not affect CSBP/p38 nor ERK activation (Figs. 5 and 7). At the same time, to the extent that UV activation might lead to JNK1 activity through a path slightly different from TNF␣, GPS2 does not affect the former route (Fig. 5).