Novel Phosphorylation-dependent Ubiquitination of Tristetraprolin by Mitogen-activated Protein Kinase/Extracellular Signal-regulated Kinase Kinase Kinase 1 (MEKK1) and Tumor Necrosis Factor Receptor-associated Factor 2 (TRAF2)*

Background: During inflammation, tristetraprolin displays a dual function as mRNA destabilizer and regulator of NF-κB signaling. Results: Tristetraprolin is inducibly hypermodified by phosphorylation and ubiquitination, affecting cell viability. Conclusion: Tristetraprolin hypermodification provokes its functional shift toward the activation of JNK signaling. Significance: The NF-κB/JNK cross-talk is central for our understanding of chronic inflammatory diseases and cancer development. Acute versus chronic inflammation is controlled by the accurate activation and regulation of interdependent signaling cascades. TNF-receptor 1 engagement concomitantly activates NF-κB and JNK signaling. The correctly timed activation of these pathways is the key to account for the balance between NF-κB-mediated cell survival and cell death, the latter fostered by prolonged JNK activation. Tristetraprolin (TTP), initially described as an mRNA destabilizing protein, acts as negative feedback regulator of the inflammatory response: it destabilizes cytokine-mRNAs but also acts as an NF-κB inhibitor by interfering with the p65/RelA nuclear import pathway. Our biochemical studies provide evidence that TTP contributes to the NF-κB/JNK balance. We find that the MAP 3-kinase MEKK1 acts as a novel TTP kinase that, together with the TNF receptor-associated factor 2 (TRAF2), constitutes not only a main determinate of the NF-κB-JNK cross-talk but also facilitates “TTP hypermodification”: MEKK1 triggers TTP phosphorylation as prerequisite for its Lys-63-linked, TRAF2-mediated ubiquitination. Consequently, TTP no longer affects NF-κB activity but promotes the activation of JNK. Based on our data, we suggest a model where upon TNFα induction, TTP transits a hypo- to hypermodified state, thereby contributing to the molecular regulation of NF-κB versus JNK signaling cascades.

Inflammation constitutes a complex process involving different cell types that release proinflammatory mediators that in turn engage cell surface receptors, leading to activation of intracellular signaling cascades and diverse cellular responses. TNF␣ is a pleiotropic proinflammatory cytokine that elicits proliferation, differentiation, or cell death. Binding of TNF␣ to its receptor, TNFR1, 2 initiates the recruitment of various adaptors including TRAF proteins, which trigger the onset of receptor specific proximal cascades that finally converge at the inhibitor of B kinase (IKK) signalosome. The multiprotein complex consisting of the inhibitor of B kinases ␣ and ␤ (IKK ␣/␤) and the regulatory subunit IKK␥ (NEMO) phosphorylates the inhibitor of B ␣ (IB␣) protein, leading to its proteosomal degradation. This results in nuclear translocation of the transcription factor NF-B/p65 and the expression of genes encoding pro-as well as anti-inflammatory cytokines, adhesion molecules, enzymes, and antiapoptotic mediators (1). In addition to proinflammatory signaling and cell survival, NF-B also controls the timed cross-talk with proapoptotic pathways, e.g. activation of the MAPK JNK.
One important aspect during TNFR1 signaling is the kinetic of IKK and MAPK activation. Mechanistically, the TRAF proteins, constituting E3 ubiquitin ligases, mediate the ubiquitination of other TNFR1 complex members, thereby facilitating both NF-B and JNK activation (2). Although these cascades are concomitantly activated and share common regulatory proteins up to the level of MAP3Ks, the onset of IKKs proceeds faster than the activation of MAPKs, resulting in a temporal * This work was funded by Austrian Science Foundation Project T511-B20 (to separation of these signaling events (3). Accordingly, immediate early expressed NF-B target genes constitute essential inhibitory feedback mediators as exemplified by GADD45␤, which blocks the activated JNK cascade at the level of the MAP2Ks MKK4/7, thereby allowing an unhindered first wave of NF-B activity (4,5).
The main MAP3K that accounts for JNK activation in this context is MEKK1, a 196-kDa protein with dual function: it displays phospho-kinase activity, which is attributed to the kinase domain at its C terminus, and E3 ligase activity, because of the PHD domain at its N terminus. The PHD domain of MEKK1 mediates not only its Lys-63-linked autoubiquitination, leading to a stabilized interaction with other proteins (6), but also ERK1/2 polyubiquitination and degradation in response to stress stimuli (7). Autoubiquitination of MEKK1 strongly depends on its kinase activity, as well as the presence of TRAF2. Importantly, it is the kinase domain that accounts for TNF␣-induced signaling to JNK as demonstrated by the analysis of cells obtained from MEKK1 Ϫ/Ϫ mice (8,9).
Inflammation is in principal a beneficial reaction that only becomes detrimental when it occurs in an excessive manner or is not appropriately terminated. Therefore, its resolution has to be correctly timed and controlled. Importantly, NF-B is not only a potent initiator of inflammation but also plays an essential role during its resolution (10,11), because it controls the expression of a set of terminators acting at different levels: at the receptor level, affecting all downstream events (12), at the level of signal transduction (13,14) and individual transcription factors (15)(16)(17), as well as at the post-transcriptional level, preventing translation and promoting degradation of transcripts. One example for the latter is TTP, an mRNA destabilizing protein, which is involved in the degradation of proinflammatory mediators including TNF␣, various interleukins, as well as enzymes, phosphatases, and adhesion molecules. It belongs to the 12-O-tetradecanoylphorbol-13-acetate-inducible sequence 11 (TIS11) family of proteins also including TIS11b, TIS11d, and ZFP36L3. All of these members have been found to contribute to mRNA destabilization. They act by binding to specific AU-rich elements in the 3Ј-UTR of the target mRNAs through a tandem zinc finger structure, which appears highly conserved among all TIS proteins (18,19). TTP itself is induced by a variety of stimuli including proinflammatory as well as anti-inflammatory cytokines, serum, insulin, and growth factors and appears abundant in many different cell types, including fibroblasts, monocytes, macrophages, neutrophils, T-cells, and endothelial cells (20 -23). TTP knock-out mice exhibit a classical inflammatory phenotype (24), which is mainly caused by the extensive overproduction of the cytokine TNF␣. However, beside TNF␣, the turnover of several other mRNAs is controlled by TTP in a stimulus a cell type-specific manner, including inflammatory mediators as interleukin-1␤, -2, -3, and -6, Ccl2 and -3, Cox-2, as well as IFN␥ or IL-10 (25)(26)(27)(28)(29)(30). Removal of these "inflammatory mRNAs" occurs with a specific time delay indicating that TTP activity is specifically regulated. In this regard, an important aspect appears to be TTP hyperphosphorylation; however, its functional relevance remained largely elusive up to now (31). The most extensively studied TTP kinase is p38␣: p38-MAPK-MK2 signaling is of particular interest in terms of cytokine stabilization, because MK2 was shown to directly phosphorylate TTP after LPS stimulation in mouse macrophages, reducing its ability to destabilize mRNAs (32)(33)(34)(35)(36). Vice versa, dephosphorylation by protein phosphatase 2A results in reactivation of its destabilizing ability (37). In addition to p38␣, various other kinases have been shown to phosphorylate TTP in vitro and in vivo (38). These include the MAPK ERK1/2, JNK, the protein kinase family members PKA and PKC, as well as IKK1 and glykogen synthase kinase-3 (39). However, not all TTP kinases alter the function of TTP as a destabilizing protein as indicated by a study using a series of TTP mutants, wherein the single mutations of serines and/or threonines showed that phosphorylation does not necessarily impact its mRNA-degrading function (40).
Recently, we and others (41) have shown that TTP does not solely act as an mRNA destabilizing protein, it also impairs NF-B transcriptional activity by specifically interfering with the p65 nuclear import pathway and p65 transactivation, respectively. Targeting the NF-B cascade at the level of the transcription factor itself extends its activity to selectively influence subsets of NF-B target genes that do not contain an AUrich element, most notably the JNK blocker GADD45␤ (42). Because our earlier results indicated that TTP also contributes to the onset of signals leading to JNK activation, we aimed to gain deeper insight into its role in the NF-B-JNK signaling cross-talk. We found that the ability of TTP to differentially affect both cascades depends on specific post-translational modifications: phosphorylation by MEKK1 and subsequent TRAF2-mediated, Lys-63-linked ubiquitination resulted in a TTP functional switch away from NF-B inhibition toward the onset of JNK signaling. This finding provides novel insight into the regulated balance between NF-B and JNK signaling cascades on the molecular level.

EXPERIMENTAL PROCEDURES
Plasmids and Cloning-Expression vectors for FLAG-TTP, -ZnN, and -ZnC were generated using the Expand High Fidelity PLUS PCR system (Roche Applied Science). PCR fragments were subcloned into pcDNA3.1-FLAG. For other plasmids, general vector information, and primer sequences, see supplemental Tables S1 and S2.
Generation and Propagation of TTP Adenovirus-For TTP adenovirus generation, the human TTP coding sequence was amplified from pCMV-MycTTP using the primers BglIIMycTTP fwd/rev. The PCR fragment was subcloned into the BamHI site of pShuttleTripTetlac-BamHI. To produce the recombinant AdEasyTetOFF-MycTTP plasmid, pShut-tleTripTetlac-MycTTP was linearized (PmeI) and cotransformed with pAdEasyTetOFFrev into BJ5183 bacterial cells (43). The resulting AdEasyTetOFF-MycTTP was, after quality control, linearized with PacI, transfected into HEK 293 lac cells, and grown in the presence of 4 nM doxycycline until plaque formation. A high titer adenoviral stock was obtained as described previously (44).
Reporter Gene Assay-Cells were transfected with the indicated reporter and/or expression plasmids using 4 g of total DNA/6-well. The experiments were performed in triplicate, and luciferase values were normalized to cotransfected ␤-galac-tosidase, illustrated as mean fold induction. The error bars represent standard deviations of the mean.
Cell Culture and Transfections-HEK 293 cells and HeLa cells were obtained from ATCC; WT, TTP KO, and MEKK1 ⌬KD MEF kindly provided by P. J. Blackshear and Ying Xia (45), respectively. The cells were cultured in DMEM (Bio-Whittaker) supplemented with 10% FCS (Sigma), 2 mM L-glutamine (Sigma), penicillin (100units/ml), and streptomycin (100 g/ml). HUVEC were isolated from umbilical cords as described (46) and maintained in M199 medium (Lonza) supplemented with 20% FCS (Sigma), 2 mM L-glutamine (Sigma), penicillin (100 units/ml), streptomycin (100 g/ml), 5 units/ml heparin, and 25 g/ml ECGS (Promocell). HEK 293 cells were transfected using calcium phosphate (47), MEF using polyethylenimine (48). For TTP adenoviral transduction, the cells were Cell Extracts, Western Blotting, and Densitometry-For cell lysates of reporter assays, calf intestine alkaline phosphatase treatment (Roche Applied Science; 20 units, 15 min, 37°C), as well as analysis of molecular TTP patterning, cells were lysed in 1ϫ passive lysis buffer (Promega). For detection of endogenous (p)-JNK levels in MEF, TTP in HUVEC, and virus-infected cells, the pellets were resuspended in 2ϫ Laemmli buffer. The proteins were separated by either 10% or 5-12% gradient SDS-PAGE, transferred to nitrocellulose membranes (Hybond-C, Amersham Biosciences), and blocked in 5% nonfat dry milk in PBS and 0.1% Tween 20 (TBS used for phosphoantibodies) for 1 h at room temperature before incubation with the first antibody overnight (4°C). Antibodies were used as indicated within the figures (see also supplemental Table S3). Densitometric analysis of MEKK1 knockdown and p-JNK1/2 levels were done using the background-corrected integrated densities of either MEKK1 or the p-JNK bands normalized to the respective loading controls (ImageJ, National Institutes of Health).
In Vivo Ubiquitination Assays-HEK 293 cells were transfected with a total amount of 4 g of DNA/6-well. 24 h later, the cells were harvested, and either ubiquitinated proteins were precipitated using nickel-nitrilotriacetic acid-agarose (Qiagen), according to the Tansey lab protocol, or tagged TTP was immunoprecipitated using FLAG matrix (anti-FLAG M2-agarose; Sigma), and ubiquitin chains were detected by specific antibodies (supplemental Table S3). Lysates were loaded on 5-12% gradient SDS-polyacrylamide gels. For ubiquitination analysis of endogenous TTP, the cells were resuspended in lysis buffer (20 mM Tris, pH 7.5, 150 mM NaCl, 1 mM EDTA, 10% glycerol, 1% Triton; protease/phosphatase inhibitors, 1 mM NEM) containing 6 M urea, RNase A (3 g/l; Qiagen), Benzonase (1 units/l; Merck), and 1.5 mM MgCl 2 (denaturing lysis buffer) for 15 min at 4°C. The lysates were diluted with lysis buffer 1/25 and precleared (1 h, protein A/G-agarose) prior to immunoprecipitation with TTP antibody prebound to protein A/G-agarose for 2 h at 4°C. Precipitates were washed three times with lysis buffer, resuspended in denaturing buffer again, and finally diluted with Laemmli buffer prior to loading on 7.5% SDS-polyacrylamide gels.
Data Analysis-Statistical differences between samples were analyzed using the paired Student's t test. Two-tailed probability values of Ͻ0.05, Ͻ0.01, and Ͻ0.001 were considered significant and highly significant, respectively. The p values are given within the figure legends.

MEKK1
Constitutes a Novel TTP Kinase-As we have shown previously, TTP impairs NF-B activity in an AU-rich elementmediated decay-independent manner (42). The MAPK p38␣ induces TTP phosphorylation, traceable by a "molecular weight shift" of the protein in Western blots, inactivating TTP mRNA binding and degrading ability, but in contrast, p38␣ did not block TTP function toward NF-B (supplemental Fig. S1A, left panel). When testing a set of other kinases, we found that coexpression of MEKK1, which has not been described as TTP kinase yet, could even counteract TTP inhibition of p65-induced NF-B promoter activity (Fig. 1A, left panel, and supplemental Fig. S1A, right panel). As shown in Fig. 1B, the kinase domain of MEKK1 appeared sufficient for the rescue, whereas a kinase-deficient MEKK1 mutant led to the complete reconstitution of TTP-mediated NF-B inhibition. Mutation of the PHD domain slightly attenuated TTP activity. Of note, p38␣ even counteracted MEKK1 function toward TTP ( Fig. 1A and supplemental Fig. S1A), and we noticed a substantial difference in TTP protein "patterning" on Western blots after coexpression of either p38␣ or MEKK1. Whereas p38␣ expression resulted in the molecular weight shift of TTP already described earlier (49,50), coexpression of MEKK1 led to an increase of "higher molecular weight species" (50 -72 kDa, termed HMW) of TTP, accompanied by a different patterning (Fig. 1A, right panel). In line with that, the kinase-dead MEKK1 mutant inhib-ited the accumulation of this HMW-TTP (Fig. 1B, right panel). Importantly, we observed a similar patterning of endogenous TTP after TNF␣ stimulation in human umbilical vein endothelial cells (HUVEC): "lower molecular weight" species (ϳ36 -50 kDa, termed LMW), for example as produced by p38␣-mediated phosphorylation appeared early upon TNF␣ treatment, whereas HMW-TTP occurred at later time points (Fig. 1C). Pharmacological inhibition of p38␣, JNK, or ERK activity did not block the MEKK1-mediated TTP protein shift (supplemental Fig. S1B), suggesting that MEKK1 phosphorylates TTP at different or additional sites.
Consequently, we analyzed the physical interaction of MEKK1 with TTP. Immunoprecipitation of HA-tagged MEKK1 led to coprecipitation of HMW-TTP as shown in Fig.  2A. Interestingly, and in line with our earlier observations, p38␣ coprecipitated with LMW-TTP (Fig. 2B). On the endogenous level, an interaction of TTP with full-length MEKK1 was barely detectable. Instead, a strong interaction with a ϳ95-kDa fragment was detected 90 min after TNF␣ stimulation, concomitantly with the appearance of HMW-TTP in the input control (Fig. 2C). In this context it is important to note that cleavage of MEKK1 occurs during the induction of apoptosis, resulting in the release of a 91-kDa activated kinase domain from insoluble structures to the soluble cytoplasmic fraction (51,52). It is therefore conceivable that TTP predominantly interacts with and becomes phosphorylated by the cytosolic kinase fragment of MEKK1. To further confirm that the observed HMW species are the result of MEKK1-mediated phosphorylation, we utilized phosphatase treatment, resulting in a complete reversal of the molecular shift and reappearance of TTP comparable with the control sample (Fig. 2D). To confine the localization of the MEKK1-mediated TTP phosphorylation, we used TTP truncation constructs (scheme in Fig. 2D). The molecular weight of a ZnC-TTP-mutant was not influenced by MEKK1 coexpression, whereas the ZnN-TTP mutant had similar "shifting" characteristics as the full-length protein (Fig. 2D, middle and bottom panels, respectively). The differences in band strength and distribution can be attributed to the varying composition of the gradient polyacrylamide gels. Together, our data suggest that MEKK1 physically interacts with TTP to phosphorylate its N terminus, thereby abrogating TTP function toward NF-B.

MEKK1-mediated Phosphorylation Serves as Basis for TRAF2-directed, Lys-63-linked TTP Ubiquitination-MEKK1
is a key player of the TNFR1-associated cytoplasmic signaling complex and acts in concert with TRAF2 to trigger activation of the JNK signaling cascade (53). Thereby, TRAF2-mediated ubiquitination processes provide the basis for stabilization and functionality of the JNK-activating complex (54). Using in vivo ubiquitin assays, we investigated whether TTP was also part of these ubiquitination events. We observed that not only TTP, but also its mRNA binding-deficient mutant (C124R), as well as the ZnN-TTP mutant were ubiquitinated when phosphorylated by MEKK1, whereas the ZnC-TTP mutant was not (Fig.  3A). Moreover, TTP ubiquitination was confirmed endogenously in MEF and HUVEC (Fig. 3B). In line with the observed MEKK1-TTP interaction, TTP polyubiquitination was detected only after TNF␣ stimulation in the presence of the deubiquitinase inhibitor NEM.
Because the MEKK1 mutant M1 Kin exhibits no E3 ligase activity, we investigated whether the MEKK1 interacting adap- tor protein TRAF2 could function as E3 ligase in this context. TRAF2 coexpression indeed resulted in a strongly enhanced TTP ubiquitination that was independent of proteasome inhibition by MG132 (Fig. 3C). This suggested that TRAF2 attached ubiquitin chains were Lys-63-linked, and moreover, phosphorylation by MEKK1 was important for effective ubiquitination.
These observations were further supported by the use of dominant negative MEKK1, as well as TRAF2 mutants: mutation of either MEKK1 or TRAF2 impaired whereas mutation of both completely abolished TTP ubiquitination (Fig. 3D). The expression of dominant negative MEKK1 could not be detected in the input when dominant negative TRAF2 was coexpressed (Fig. 3D, bottom panels, M1dnKin/T2dn), explained by the fact that MEKK1 stabilization depends on an intact kinase domain as well as an intact TRAF2 E3 ligase domain (3,6,53). The use of ubiquitin mutants additionally validated the TRAF2-mediated, Lys-63-linked TTP ubiquitination (Fig. 4, A and B). Over and above, we also found Lys-63-linked ubiquitin chains attached to endogenous TTP after TNF␣ stimulation in HUVEC. In line with the above results, MG132 treatment had no effect, whereas the addition of NEM strongly enhanced Lys-63-linked TTP ubiquitination (Fig. 4C). The latter can be attributed to the zinc finger domain of TTP, because required ubiquitin target lysine residues occur only in this region. Further, the loss of MEKK1 function, achieved either by lentiviral, shRNA-mediated MEKK1 knockdown in HUVEC (Fig. 4D, left panel) or by the Note the presence of HMW, as well as LMW-TTP in the input controls of both experiments (bottom panels). C, primary HUVECs were stimulated with TNF␣ for the indicated time periods, and endogenous TTP was immunoprecipitated using a TTP antibody described earlier (66). Bound MEKK1 mainly appears as a cleaved fragment at ϳ95 kDa rather than as full-length fragment (LE * , long exposure) after 90 and 180 min of stimulation (left panels). Note that endogenous HMW-TTP initially appears after 90 min as shown in the input control (right panels). HC * , heavy chain. D, HEK 293 cells were transfected with full-length TTP or TTP mutant constructs (ZnC, ZnN; see scheme) only (Ϫ) or in combination with M1 Kin as indicated. Cell lysates containing TTP/M1 Kin were divided, and one half was treated with calf intestine alkaline phosphatase (P * ). Western blot (WB) analysis of TTP and its mutants demonstrated that MEKK1 shifted the molecular weight of the full-length protein as well as of the N-terminal TTP-mutant (top and bottom panels) but had no effect on the C-terminal mutant (middle panel). P in the scheme represents a hypothetical phosphorylation site. CCCH, cysteine-cysteine-cysteine-histidine-tandem zinc finger.
use of MEF deficient in the MEKK1 kinase domain (Fig. 4D, right panel) confirmed that endogenous TTP ubiquitination depended on phosphorylation by MEKK1. Importantly, the E3 ligase activity of MEKK1 was negligible in terms of TTP ubiquitination, proven by the respective MEKK1 mutant (data not shown), as well as the use of ⌬KD MEF (Fig. 4D, right panel).
Our results indicate that TTP is covalently modified by ubiquitin in vivo (although we cannot formally exclude that other proteins bound to TTP complexes are not dissociated by 6 M urea) and that attached polyubiquitin chains are predominantly Lys-63-linked. In summary, MEKK1 mediated, N-terminal TTP phosphorylation appears to be prerequisite for TRAF2mediated, Lys-63-linked TTP polyubiquitination.
TTP Promotes JNK and AP-1 Activity and Affects Cell Viability-The molecular cross-talk between NF-B and JNK signaling cascades is the key to account for the timed regulation of biological processes as cell growth, differentiation, and apoptosis. TNFR1-induced prolonged JNK activation results not only from blocking NF-B activity but depends on TRAF2-MEKK1 complex formation leading to JNK and subsequent AP-1 activation (8,53,55). When we analyzed TTP KO MEF for TNF␣-induced JNK activation, we observed that the lack of TTP caused an impairment of p-JNK levels under stimulated conditions, comparable with the situation observed in MEKK1 ⌬KD MEF, whereas total JNK levels were not affected (Fig. 5A). When analyzing TTP expression in WT MEF (Fig. 5A, middle  panel), we noticed that forced, prolonged JNK activation (achieved through specific IKK2 inhibition by BMS; Fig. 5A, 90 ϩ -180 ϩ ) goes in line with the appearance of HMW forms of TTP (see Fig. 7, top blot). In contrast, TTP was generally less abundant in ⌬KD MEF, and although HMW-TTP was somewhat detectable, which can also be attributed to phosphoryla-  NOVEMBER 4, 2011 • VOLUME 286 • NUMBER 44

JOURNAL OF BIOLOGICAL CHEMISTRY 38471
tion by other kinases, it appeared visibly different. Notably, mainly JNK2 activation was impaired but restored after reconstitution of TTP expression in KO MEF (Fig. 5B). Accordingly, AP-1 activity was also affected by TTP: whereas TTP increased AP-1 promoter activity in a dose-dependent manner in HEK 293 cells (Fig. 5C, top panel), AP-1 activity was diminished in KO MEF and could be restored to WT levels after TTP reconstitution (Fig. 5C, bottom panel). This is in line with our previous findings where TTP down-regulated NF-B promoter activity dose-dependently in HEK 293 cells, whereas MEF lacking TTP displayed a much higher basal NF-B promoter activity (42).
In light of the results above, we analyzed the potential of TTP to influence cell growth and/or apoptosis. We generated a recombinant adenovirus for expression of Myc-tagged TTP under the control of a tetracycline responsive promoter (Tet-OFF-system). This allowed the controlled expression of TTP in terms of time and expression levels. Via titration experiments, we assured that viral infection does not affect cell viability (data not shown). Under the established conditions, we tested the impact of TTP on cell viability: HeLa cells ectopically expressing TTP (Fig. 6A, verified by Western blotting) proliferated significantly slower as compared with non-TTP-expressing control cells over a time period of 72 h. Because doxycycline itself affected the cells between 48 and 72 h of treatment (data not shown) only effects within the first 48 h were further studied (Fig. 6A, bottom graphs). Uninfected and control adenovirusinfected cells without or treated with doxycycline exhibited a doubling in cell number (100%) during the first 24 h. The cells that ectopically expressed TTP reached a cell number of ϳ60% compared with the above mentioned controls after 24 h (graph 1). Remarkably, in one experiment, where we erroneously added a lower doxycycline concentration to control AdTTPinfected cells, we observed a decline in cell number (graph 2, 48 h, #), which could be correlated to the respective TTP protein levels assessed by Western blotting (#). Obviously, upcoming TTP expression, attributed to insufficient doxycycline treatment, resulted in an immediate convergency of total cell numbers from control infected and TTP-expressing cells. Besides, the coexpression of TTP or its N-terminal mutant ZnN with their modifiers MEKK1 (kinase domain), TRAF2, ubiquitin, or Lys-63-linked ubiquitin led to an alteration of cell viability in HEK 293 cells (Fig. 6, B and C). In agreement with the results in Fig. 3D, cells that coexpressed dominant negative versions of either MEKK1 or TRAF2 remained unaffected (Fig.  6B). Together, our data provide evidence that TTP is hypermodified by the combined action of MEKK1 and the E3 ligase TRAF2 (Fig. 6D). Thus we hypothesize that the ability of TTP to affect cell viability is strongly dependent on an accurately timed, post-translational control of the protein through p38␣ on the one hand and MEKK1 on the other hand. The expression of TTP in combination with appropriately directed hypermodification events differentially regulate its involvement in the cross-talk between NF-B and JNK signaling. Therefore, we propose that TTP acts as "balancer" in the coordinated control of survivalversus death-promoting signaling cascades (Fig. 7).

DISCUSSION
NF-B constitutes the key transcription factor in the regulation of inflammatory processes. It concomitantly controls cell survival by initiating the transcription of antiapoptotic and antioxidative genes, as well as the timed cross-talk with "proapoptotic" pathways, represented mainly by the MAPK cascade leading to the activation of JNK. In this context, our novel finding that MEKK1, an upstream JNK-activating kinase, counteracts TTP-mediated inhibition of NF-B promoter activity, provides further insight into the interdependent regulation of NF-B and JNK signaling cascades. In vitro, the observed reversion occurred with canonical 5ϫ NF-B promoters, as well as with the more physiologic human IL-8 or tissue factor promoters (data not shown), both containing NF-B-binding sites, and was dependent on MEKK1 kinase activity.
Remarkably, p38␣, which has already been shown to abolish TTP-mediated mRNA degradation, had no effect on TTP-mediated NF-B inhibition, it rather counteracted MEKK1 function in our settings. Additionally, we found distinct "TTP forms" bound to these two proteins: p38␣ coprecipitated with LMW-TTP, whereas MEKK1 interacted with the HMW species. Interesting enough, these different TTP species resembled the appearance of endogenous TTP in HUVEC and MEF, where TTP appears in a LMW form in the beginning of TNF␣ stimulation and gradually adopts the HMW form later on.
After TNFR1 ligation, MEKK1 acts in concert with TRAF2 to activate JNK signals, and our observation of TRAF2-MEKK1mediated ubiquitination suggested a role for TTP in this scenario. TTP indeed promoted JNK activity mainly affecting p-JNK2 levels. Generally, JNK1 and 2 differentially bind to and regulate c-Jun stability, phosphorylation, and subsequent AP-1 complex activation, thereby affecting cell proliferation and transformation or inducing apoptosis in a cell type-and stimulus-dependent manner. JNK2 exerts higher binding affinities for c-Jun than JNK1, and fibroblasts lacking JNK2 exhibit a proliferation advantage (56,57), which is in line with the fact that TTP KO MEF grow faster than WT MEF.
Strikingly, also AP-1 activity was influenced by TTP: low doses of TTP inhibited, whereas increasing amounts enhanced AP-1 activity in reporter gene assays. Interestingly, the opposite is true for the effect of TTP on NF-B promoters: increasing TTP levels successively decrease its activity (42). ; at very early time points low TTP levels appear, which increase upon time to reach a maximum and finally rapidly disappear. Based on our results, we postulate that lower doses of TTP can only be accessible by p38␣, for example, to produce LMW-species that act on the inhibition of NF-B activity. During this time frame, other MAPKs, such as MEKK1 for example, would be blocked until TTP has reached a specific level and not until then generate the observed HMW species that would subsequently be ubiquitinated by TRAF2. This would be in agreement with our finding that p38␣ overrules MEKK1 function toward TTP and allow the correct timing for JNK activation; consequently, "hypermodified" TTP would no longer act on NF-B but actively contribute to JNK activation instead and therefore constitute a balancer in the cross-talk between these signaling cascades.
Moreover, MEKK1-mediated phosphorylation occurs at the N terminus of TTP, a region that is not conserved among the members of the well described TIS family of proteins, all of which exhibit the ability to destabilize mRNAs. This might be an indication for a TTP-specific function, acquired in the course of evolution where TTP may have structurally and functionally diverged from an initially common ancestor. This is supported by the described functions of homologous proteins: in Saccharomyces pombe, a TTP-related protein accounts for the transmission of a pheromone-induced Ras/mitogen-activated protein kinase signal (58). In Saccharomyces cerevisiae, a similar cysteine-cysteine-cysteine-histidine-type tandem zinc finger protein is involved in metabolism and retards cell growth when overexpressed. In addition, a TIS11 homologue was found implicated in cell cycle control in Drosophila melanogaster (59). Interestingly, in Caenorhabditis elegans, a related cysteine-cysteine-cysteine-histidine tandem zinc finger protein termed PIE-1 was described as bifunctional protein, involved in the control of mRNA stability as well as in the inhibition of transcription (60).
Additionally, it is important to note that also the expression of human TTP is not restricted to pro-and anti-inflammatory mediators and their mRNA degradation: TTP is induced in response to the breast cancer susceptibility protein BRCA1 (61), as well as upon TGF␤ stimulation in human T-cells. Interestingly, TGF␤ controls cell proliferation and differentiation and signals through the SMAD-dependent pathway to induce apoptosis by the activation of JNK. Additionally, TGF␤ acts on the control of NF-B: TNF␣ stimulation of TGF␤ pretreated cells results in restricted NF-B target gene expression caused by impaired p65 nuclear translocation (62), reminiscent of our previous findings. Furthermore, TGF␤ stimulated T-cells upregulate TTP via heterodimeric SMAD3/4 transcription factors, and TGF␤ as well as TTP-deficient mice exhibit overlapping phenotypes (63). Thus, it is tempting to implicate TTP into TGF␤-triggered, JNK-mediated apoptosis.
Moreover, a study by Johnson and colleagues (64,65) unraveled that continuous expression of TTP/TIS11 proteins induces various primary and transformed cell types to undergo apoptosis; however, the mechanism remained elusive. In their study, moderate expression levels of TTP were insufficient to cause widespread cell death, but they induced cell death upon concomitant TNF␣ treatment. It is striking, therefore, that neither TIS11b nor TIS11d acted in synergy with TNF␣, which is the only functional difference that has been identified among TTP/TIS11 proteins so far. In addition, they determine the ZnN region of TTP as being responsible for the onset of cell death, reflecting a specific functional divergence between TTP and other TIS11 proteins (64,65). This supports our hypothesis that TTP, once hypermodified in its ZnN region, actively contributes to cell death-promoting signaling cascades.
In summary, it is well established that TTP regulates a variety of biological processes. Its described role in inflammatory as well as tumorigenic diseases clearly shows that TTP is functionally much more complex than previously appreciated. Our findings suggest a model wherein TTP contributes to the balance of NF-B versus JNK signaling, which depends on a series of timedependent, alternating post-translational hypermodifications, including phosphorylation and regulatory ubiquitination. These contribute to a functional switch of TTP: upon the course of a TNF␣ induced response the "early TTP" (probably LMW species) might interfere with NF-B signaling, whereas the time-dependent accumulation of hypermodified HMW-TTP is associated with the onset of JNK signaling (model in Fig.  7). Future studies will provide more mechanistic insight into TTP-mediated JNK activation and include analyses of these modifications on the endogenous level, investigation of MEKK1-specific target phospho-sites as well as lysine residues, which are targeted for ubiquitination.