Loss of TRB3 Alters Dynamics of MLK3-JNK Signaling and Inhibits Cytokine-activated Pancreatic Beta Cell Death*

Background: MLK3 induces TRB3 to inhibit AKT and compromise beta cell survival. Results: AKT regulates novel phosphorylation, ubiquitination, and degradation of MLK3, explaining why TRB3−/− islets display attenuated MLK3/JNK activation and beta cell death. Conclusion: TRB3 functionally cooperates with the JNK module to fine-tune inflammatory signaling in beta cells. Significance: Even modest modulation of signaling dynamics can alter inflammatory outcomes. Disabling cellular defense mechanisms is essential for induction of apoptosis. We have previously shown that cytokine-mediated activation of the MAP3K MLK3 stabilizes TRB3 protein levels to inhibit AKT and compromise beta cell survival. Here, we show that genetic deletion of TRB3 results in basal activation of AKT, preserves mitochondrial integrity, and confers resistance against cytokine-induced pancreatic beta cell death. Mechanistically, we find that TRB3 stabilizes MLK3, most likely by suppressing AKT-directed phosphorylation, ubiquitination, and proteasomal degradation of MLK3. Accordingly, TRB3−/− islets show a decrease in both the amplitude and duration of cytokine-stimulated MLK3 induction and JNK activation. It is well known that JNK signaling is facilitated by a feed forward loop of sequential kinase phosphorylation and is reinforced by a mutual stabilization of the module components. The failure of TRB3−/− islets to mount an optimal JNK activation response, coupled with the ability of TRB3 to engage and maintain steady state levels of MLK3, recasts TRB3 as an integral functional component of the JNK module in pancreatic beta cells.

Cytokine-induced beta cell failure and death are crucial events that lead to type 1 as well as type 2 diabetes (1). Cytokines rapidly activate the stress kinase SAPK/JNK, and other proinflammatory kinases, setting in motion a series of events that unfold over several hours and cumulatively lead to dysfunction and demise of the beta cell (1)(2)(3). JNK activation has been clearly linked to mitochondrial stress and damage in multiple cell types, including the pancreatic beta cell (4 -6). We have reported that cytokine-induced MLK3-JNK pathway is instrumental in compromising mitochondrial integrity early in the apoptotic process, effectively reducing cellular defenses and increasing the potency of subsequent inflammatory events (6).
Signaling via the JNK module results from the sequential activation of a MAP3K, MAP2K, and JNK, complexed via a scaffold protein (7,8). Cytokine binding to cognate TNF and IL-1/TLR receptors triggers activation of the TRAF 2 family of atypical ubiquitin ligases and promotes Lys-63-linked ubiquitination and autophosphorylation of MAP3Ks (9 -11), including MLK3 (12)(13)(14). A series of reports have provided biological proof that cytokines favor JNK activation via the specific MLK3-MKK7-JNK-JIP1 signaling module (15)(16)(17). Furthermore, robust JNK signaling is facilitated by a self-amplifying, feed-forward loop of sequential kinase phosphorylation and is reinforced by mutual stabilization of the signaling module components (18).
We have previously shown that MLK3 activation and protein levels are both induced by IL-1␤ stimulation of mouse and human islets. Along with MLK3, IL-1␤ also stimulates tandem induction of TRB3 protein expression in pancreatic islets (6). Furthermore, in the beta cell, MLK3 engagement stabilizes TRB3 protein levels to suppress AKT survival kinase and facilitate downstream proinflammatory effects of IL-1␤ (6). Here, we report that AKT directly phosphorylates MLK3 at a novel site (Thr-477) located adjacent to the leucine zipper domain, triggering Lys-48-linked ubiquitination and proteasomal degradation of MLK3. Our results also show that the ubiquitin acceptor site instrumental for triggering degradation of MLK3, coincides with the acceptor site for TRAF6-dependent Lys-63linked ubiquitination and activation of MLK3. This site-specific convergence of functionally divergent ubiquitination events, shows that ubiquitination of MLK3 (Lys-48 versus Lys-63) plays a crucial role in driving cytokine-mediated beta cell death.
Consistent with the above mechanism, TRB3 null islets displayed higher levels of active AKT and attenuated cytokinestimulated MLK3 induction and JNK activation. More importantly, TRB3 Ϫ/Ϫ islets also displayed resistance to cytokinemediated beta cell death. We conclude that cytokines require a threshold of proinflammatory signal activation for inducing beta cell death and that TRB3 is required for optimal kinetics of MLK3-JNK activation to enable cell death. Taken together, these data suggest that factors responsible for fine-tuning the dynamics of cytokine signaling can be exploited for therapeutic intervention.

EXPERIMENTAL PROCEDURES
TRB3 Knock-out Mice-The mutant mouse strain was obtained from the European Mouse Mutant Archive (EMMA ID EM:02346), Helmholtz Zentrum Muenchen, Germany. Briefly, the strain was generated by retroviral insertion of a gene trap vector encoding a premature stop codon prior to the first coding exon and was originally developed by Lexicon Genetics, Inc., under the Wellcome Trust Knock-out Mouse Resource. Mice were housed in a 12-h light/12-h dark cycle at controlled temperature (25°C Ϯ 1°C). Genotyping was performed by PCR of genomic DNA according to above referenced protocols. In experimental procedures, we compared homozygous TRB3 knock-out mice with wild-type littermates (control mice). All experimental procedures were performed according to University of California San Diego Institutional Animal Care and Use Committee policies.
Cell Culture and Transfection-Min6 cells (passages 15-18 only) were grown in DMEM containing 25 mM glucose supplemented with 4% heat-inactivated FBS and 50 M ␤-mercaptoethanol. HEPG2 cells were grown in a 1:1 mix of DMEM and F12K and 5% heat-inactivated FBS. Transfections were carried out using Lipofectamine 2000 (Invitrogen) as per the manufacturer's instructions and were treated as described 48 h post transfection. For the TUNEL assay, Min6 cells were treated for 8 h with a mixture of 10 nM TNF-␣, 40 ng/ml IFN-␥, and 20 ng/ml IL-1␤, ϳ20 h post-transfection.
GST Pulldown Assay-Mammalian expression vectors encoding GST fusion proteins were expressed in Min6 or HepG2 cells followed by GST pulldowns with glutathione Sepharose (Amersham Biosciences) as described (6).
MLK3 Ubiquitination-Experiments were performed as described (14). Briefly, 40 -48 h post-transfection, cells were treated with insulin as described, harvested in 1% SDS, 50 mM Tris, pH 8.0, 1 mM EDTA containing phosphatase inhibitors, and 10 mM N-ethylmaleimide deubiquitinase inhibitor. Lysates were boiled, cooled on ice, sonicated, and precleared with Sepharose beads, and the supernatants diluted 5-fold with dilution buffer (150 mM NaCl, 20 mM Tris, pH 8.0, 1 mM EDTA) with a full complement of phosphatase and protease inhibitors and deubiquitinase inhibitor (N-ethylmaleimide), used for GST pulldowns, resolved by SDS-PAGE, and probed using anti-Myc antibodies. Blots were stripped and probed using anti-GST antibodies for GST-MLK3 input.
AKT-mediated in Vitro Phosphorylation of MLK3-GST-MLK3-WT, T477A, S674A, and double mutant (TS-AA), were expressed in HepG2 cells, and the cells were pretreated with AKT inhibitor for 4 h and CEP11004 for 1 h. GST-MLK3-WT and mutant proteins were isolated using GST pulldown, and expression levels were assessed using Coomassie staining, and anti-GST Western blots. Equal amounts of GST-bound MLK3 proteins (WT and mutants) were incubated with recombinant AKT1 (EMD Biosciences, San Diego, CA) for 40 min at 30°C in kinase assay buffer (200 M ATP, 20 mM Tris-HCl, pH 7.5, 10 mM MgCl 2 , 10 mM DTT, and 100 nM CEP11004) using [␥-32 P]ATP. Following SDS-PAGE, phosphoproteins were detected by autoradiography.
Splenocyte and Islet Co-Culture-Co-culture was performed as described previously with some modifications (6). Briefly, wild-type spleens were crushed and passed through a 70-m mesh. Red blood cells were lysed in 0.15 M NH 4 Cl and 1.5 ϫ 10 6 splenocytes per well were plated in 24-well dishes in RPMI supplemented with 10% heat-inactivated FBS. Splenocytes were stimulated with plate-bound anti-CD3 and exogenous anti-CD28 antibodies (10 g/ml and 1 g/ml, respectively; BD Biosciences) for 2 days. Isolated islets from WT or TRB3 Ϫ/Ϫ mice were rested overnight and co-cultured using Transwell filters in the presence of unstimulated or stimulated splenocytes.
Immunofluorescence and TUNEL Assays-Islets were fixed in zinc-buffered formalin (Anatech, Battle Creek, MI) for 1 h and processed for routine cryomicrotomy. Min6 cells were fixed in 2% paraformaldehyde for 10 min, and both the cells and frozen sections were permeabilized using 0.1% Triton X-100, and the TUNEL assay was carried out according to the manufacturer's instructions (Roche Diagnostics). Primary antibodies were visualized with species-specific secondary antibodies conjugated to fluorescent probes.
Microscopy and Image Acquisition-Fluorescent images were acquired on a Zeiss 710 confocal microscope. All images were assembled in Photoshop CS6, (Adobe Systems, Inc., San Jose, CA) and NIH ImageJ software. Quantification was performed on 10 randomly selected fields of view using the 40 or 60ϫ objective from three separate experiments. Total cell counts were obtained using DAPI (1 g/ml) nuclear counterstaining.
Statistics-Differences between means were examined using analysis of variance followed by a Bonferroni post hoc comparison. In all cases, p values Յ 0.05 were considered significant. Analysis was performed using GraphPad Prism software.

TRB3 Is Required for Optimal MLK3-JNK Activation-We
have previously reported that the MAP3K MLK3 and the pseudokinase TRB3 are both induced by IL-1␤ and function in concert to diminish survival mechanisms in cytokine-treated beta cells. Pseudokinases, although lacking in catalytic activity, function as allosteric modulators of other kinases, most likely in a context-dependent manner (19). The absence of a consistent and reliable readout for TRB3 action necessitates a more empirical exploration of TRB3 function. Published data suggest that TRB3 Ϫ/Ϫ mice display normal glucose homeostasis on a chow-fed regimen as assessed by glucose and insulin tolerance tests (20). To better understand the role of TRB3 in pancreatic beta cells, we obtained whole body TRB3 knock-out mice that were generated at source as described. The mice bred in normal Mendelian ratios, and based on growth, weight, and glucose tolerance tests, TRB3 Ϫ/Ϫ mice were indistinguishable from WT littermates, and we observed no discernable effect on pancreatic beta cell mass and size (data not shown). We verified TRB3 ablation using isolated islets pooled from multiple TRB3 Ϫ/Ϫ or WT mice. As seen in Fig. 1A, thapsigargin strongly induced TRB3 protein levels in WT islets, and no signal was observed in TRB3 Ϫ/Ϫ islets (compare lanes 2 and 4). Notably, TRB3 protein was not detected in the absence of stress in WT islets. Similar data were observed for TRB3 mRNA levels (supplemental Fig. S1).
Several studies have shown that TRB3 functions as a potent inhibitor of AKT kinase in multiple contexts (21)(22)(23)(24). Generally, pancreatic islets display high basal levels of AKT kinase activation, most likely a result of glucose-stimulated insulin secretion and autocrine insulin signaling (25,26). We observed no difference in the levels of total or phosphorylated AKT protein under normal culture conditions (data not shown) in islets of WT versus TRB3 Ϫ/Ϫ mice. Prior to assessing insulin-stimu-  2) and TRB3 knock-out (TRB3-KO, lanes 3 and 4) islets were treated with 1 M thapsigargin (THAP, lanes 2 and 4) for 16 h, and total cell lysates were assessed for TRB3 protein expression using SDS-PAGE followed by Western blotting using anti-TRB3 antibodies. ␤-Tubulin (␤-Tub) was used as a loading control. B, WT (lanes 1 and 2) and TRB3-KO (lanes 3 and 4) islets were treated with insulin (INS; lanes 2 and 4) for 10 min, and total cell lysates were analyzed by Western blotting using AKT antibodies specific for phospho-Ser4 -73 and Thr-308. Total AKT levels are shown. C, densitometric analysis of AKT phosphorylation in WT and TRB3-KO islets represent means Ϯ S.E. of three separate experiments. * and **, p Ͻ 0.05 versus untreated WT control. NS, not significant. D, WT (lanes 1-4) and TRB3-KO (lanes 5-8) islets were treated with 20 ng/ml of IL-1␤ for the indicated times, and lysates (60 g protein) were subjected to Western blotting using anti-MLK3 and phospho-JNK antibodies. Total JNK protein levels are shown. Expression under unstimulated conditions (NT) in WT and TRB3-KO islets is shown (lanes 9 and 10). E, densitometric analyses of time-dependent MLK3 induction and JNK activation from WT and TRB3-KO islets were graphed, and total kinase induction/activity over time was assessed by calculating area under the curve (AUC). Average values from three independent experiments are presented as a bar graph. *, p Ͻ 0.05 versus WT islets. F, Min6 cells incubated for 30 min (30Ј) in the presence (lanes 1 and 3) or absence (lane 2) of a synthetic inhibitor of AKT enzymatic activity were treated with IL-1␤ for 20 min (lanes 2 and 3), and total cell lysates were subjected to SDS-PAGE followed by Western blotting for endogenous MLK3 expression. ␤-Tubulin served as a loading control. G, densitometric quantitation of MLK3 protein levels normalized to ␤-tubulin shown as the mean Ϯ S.E. from three separate experiments. *, p Ͻ 0.05 versus IL-1␤ treatment in the absence of AKT inhibitor (inh). H, GST-MLK3 was expressed in HepG2 cells with KD (lane 1) or myristoylated (Myr) (lanes 2-4) AKT in the absence (lanes 1 and 2) or presence of increasing amounts of HA-tagged pseudokinase protein TRB3 (lanes 3 and 4). Cellular extracts were used for GST pulldowns and probed for total MLK3 levels using anti-GST antibodies. Lysate inputs were used for Western blotting of HA-TRB3 and HA-AKT. lated activation of AKT kinase, islets pooled from multiple WT or TRB3 Ϫ/Ϫ mice were cultured in KRBH buffer (without serum or glucose) for 2 h to minimize insulin secretion and autocrine insulin signaling. In the absence of stimulation, phospho-AKT levels drop rapidly, most likely due to net increase in phosphatase activity. Incubation in KRBH was followed by a short 10-min stimulation with insulin and Western blots of islet protein extracts were examined for AKT activation using phospho-AKT-Thr-308 and phospho-AKT-Ser-473 specific antibodies. Total AKT levels were used as control. WT islets showed a strong, insulin-stimulated induction of AKT phosphorylation at both Thr-308 and Ser-473 sites (Fig.  1B). TRB3 Ϫ/Ϫ islets displayed high basal AKT phosphorylation, even under fasting conditions. Insulin stimulation did not induce a significant change in AKT phosphorylation beyond that seen under basal conditions in TRB3 Ϫ/Ϫ islets (Fig. 1, B and C). These data suggest that in the absence of TRB3, dephosphorylation of AKT may be diminished.
We next examined how the absence of TRB3 impacts the proinflammatory signaling response in islets. To this end, we pooled islets from multiple mice within the WT and TRB3 knock-out groups, treated the islets with IL-1␤, and examined time-dependent MLK3 induction and JNK activation. Under unstimulated conditions, both groups showed little or no expression of MLK3 (Fig. 1D, lanes 9 and 10). Western blotting of cytokine-treated islets pooled from TRB3 null mice show a blunted and shorter induction of MLK3 protein levels compared with that in WT islets. Peak MLK3 induction in TRB3 Ϫ/Ϫ islets was ϳ3-4-fold lower than WT islets (compare lanes 1 and 2 with lane 6). MLK3 expression began to taper off by 60 min in TRB3 Ϫ/Ϫ islets (lanes 7 and 8), whereas WT islets showed persistence of MLK3 expression even at 5 h post-stimulus (lanes 3 and 4). Consistent with its role as a MAP3K that initiates cytokine-mediated JNK activation, the pattern of MLK3 induction was reflected in the time-dependent activation of phospho-JNK, with a similarly attenuated response in TRB3 Ϫ/Ϫ islets compared with WT islets. Again, JNK activation in WT islets persisted beyond that seen in TRB3 Ϫ/Ϫ islets. Densitometric analyses of Western blots from time courses of cytokine-treated islets were graphed and used to measure area under the curve to assess total MLK3 and JNK activation. Average area under the curve measurements from three experiments show sharply lower induction of both MLK3 and JNK in TRB3 Ϫ/Ϫ islets (Fig. 1E).
Results from Fig. 1, B-E, suggest that a loss of TRB3 expression and/or an associated increase in AKT activation determine the course of MLK3 induction. We therefore performed the converse experiment and asked whether AKT inhibition could augment IL-1␤-mediated MLK3 induction. Min6 insulinoma cells were treated with IL-1␤ in the presence and absence of an AKT-inhibiting compound and examined for endogenous MLK3 expression. As seen in Fig. 1F, a short (30-min) preinhibition of AKT did not induce a discernable basal expression of MLK3 protein, but prior inhibition of AKT was sufficient to potentiate IL-1␤-mediated induction of MLK3 (see densitometric analysis in Fig. 1G).
To examine whether TRB3 directly contributes to the dampening effect of AKT on MLK3 levels, we transfected Min6 cells with equal amounts of GST-MLK3 in the presence of kinase dead or constitutively active myristoylated AKT. Despite equal transfection, compared with AKT-KD (Fig. 1H, lane 1), expression of MLK3 was strongly reduced in the presence of myristoylated AKT (lane 2). Increasing expression of TRB3 restored and augmented MLK3 expression (lanes 3 and 4) nearly 2-fold, suggesting that MLK3 levels are sustained by TRB3-dependent inhibition of AKT.
AKT-mediated Phosphorylation Destablizes MLK3 Protein-We have previously reported that MLK3-TRB3 binding enhances levels of both proteins, indicating that TRB3 could both inhibit AKT and stabilize MLK3. To examine whether AKT could directly regulate MLK3 protein stability, GST-MLK3 fusion constructs were expressed in Min6 insulinoma cells and subjected to a cycloheximide chase in the absence or presence of AKT inhibitor. Experiments were conducted in regular growth medium, which maintains AKT activation due to autocrine effects of insulin. Western blots of GST pulldowns were assessed for total MLK3 protein levels using anti-GST antibodies. Results from these experiments showed increased, time-dependent loss of MLK3 in cells treated with insulin ( Fig A previous study has reported that AKT directly interacts with and inhibits MLK3 action possibly as a consequence of C-terminal serine 674 (Ser-674) phosphorylation (27). However, the biologic context and mechanism for phospho-Ser-674-mediated suppression of MLK3 activity remains to be defined. We therefore examined whether MLK3 was phosphorylated in response to insulin stimulation in beta cell lines. GST-MLK3-transfected Min6 cells were maintained in 3 mM glucose medium for 4 h and treated with insulin in the presence or absence of AKT inhibitor, and GST pulldowns were subjected to Western blotting using AKT, phosphosubstrate-specific antibodies. These antibodies recognize AKT kinase substrates with conserved arginine residues in the Ϫ3 and Ϫ5 positions (RXRXX(pS/pT)) of the phosphorylation site. Using these antibodies, we observed GST-MLK3 phosphorylation in response to insulin, which was lost upon treatment with AKT inhibitor (Fig. 2C).
Although the comparison of MLK3 protein sequences revealed 94 -99% identity across multiple mammalian species, the aforementioned AKT phosphorylation site at Ser-674 was poorly conserved (Fig. 2D). A closer examination identified an additional, broadly conserved consensus motif for AKT phosphorylation, located at threonine 477 (Thr-477), immediately adjacent to the leucine zipper homodimerization domain of MLK3 protein. We examined AKT-mediated MLK3 phosphorylation by an in vitro kinase assay using purified AKT kinase and GST fusion proteins of MLK3-WT or MLK3 bearing point mutations in each of the AKT consensus sites, individually or combined. As seen in Fig. 2E  Results from Fig. 2E prompted us to examine whether AKT also favored phosphorylation of Thr-477 in vivo. Again, the cells were starved in low glucose, and the samples were analyzed as described for Fig. 2C. Similar to results from in vitro kinase assays, we found that insulin stimulation resulted in phosphorylation of MLK3 at Thr-477 but not Ser-674 (Fig. 2F). Thus, under conditions used in this study, AKT preferentially phosphorylates MLK3 at a novel site both in vitro and in Min6 cells in vivo.
We next examined whether AKT-dependent loss of MLK3 protein stability required phosphorylation at Thr-477. Min6  -4), was performed using recombinant AKT kinase and ␥ 32 p-ATP, followed by SDS-PAGE and autoradiography. Equal levels of starting material were determined using anti-GST Western blotting. F, insulin-induced (20 mins) in vivo phosphorylation of MLK3 proteins was assessed using pulldowns of GST-MLK3 fusion proteins expressed in Min6 cells. AKT substrate-specific antibodies detected phosphorylation of GST-MLK3 WT (WT) (lanes 1 and 2) and S674A (lanes 5 and 6), but no signal was observed for T477A, or the TS-AA mutants (lanes 3 and 4 and lanes 7 and 8, respectively). G, stability of the GST-MLK3 WT and AKT consensus mutants T447A and S674A was assessed using GST pulldowns

TRB3 Modulates Cytokine-activated JNK Signaling Dynamics
cells expressing WT, T477A, or S674A mutants of GST-MLK3 were treated with cycloheximide and insulin in the presence or absence of AKT inhibitor. Cells were harvested prior to the insulin stimulus to mark basal levels of MLK3 protein expression at time 0 and at 100 min following insulin treatment. Western blots of the GST pulldowns showed that AKT inhibitor prevented the loss of MLK3 protein, regardless of the mutation (Fig. 2G, lanes 1 and 2). However, insulin treatment in the absence of AKT inhibitor markedly decreased (Ϸ60%) levels of both MLK3-WT and S674A mutant but had no effect on levels of T477A mutant of MLK3 (lane 3). Inputs from all samples were probed for the more stable ␤-tubulin protein used here as a control (half-life of 50 h (28)). A representative panel with comparable levels of ␤-tubulin from GST-MLK3-WT input samples was shown. Total protein levels at each time point were densitometrically assessed and an average of three experiments was graphically represented in Fig. 2H.
Thr-477 Phosphorylation Triggers Lys-48-linked Polyubiquitination of MLK3-Tandem phosphorylation and polyubiquitination are well characterized post-translational modifications for targeting the proteasomal clearance of cellular proteins. We examined whether MLK3 is targeted to the proteasome in response to insulin. We have previously shown that cytokine-stimulation activates MLK3 via TRAF6-dependent Lys-63-linked polyubiquitination (14). Lys-48-linked polyubiquitin chains target proteins toward proteasomal breakdown prompting us to examine insulin-induced Lys-63-independent polyubiquitination of MLK3 (Fig. 3A). GST-MLK3 was co-expressed with Lys-63 (Myc-Ub-K63R) mutant or Lys-48 (Myc-Ub-K48R) mutants of Myc-tagged ubiquitin, the cells briefly stimulated with insulin in the presence of the proteasome inhibitor MG132, and ubiquitination of GST-MLK3 pulldowns was assessed using anti-Myc antibodies. Insulin increased ubiquitination of MLK3 in the presence of Ub-K63R, but MLK3 polyubiquitination was abolished upon usage of ubiquitin-K48R mutant. To avoid confusion with Lys-63-linked ubiquitination, all subsequent experiments examining insulin-induced ubiquitination were conducted using Myc-Ub-K63R. We next examined whether insulin-mediated polyubiquitination of MLK3 requires AKT-dependent prephosphorylation. Min6 cells expressing GST-MLK3-WT or GST-MLK3-T477A were briefly treated with insulin in the presence of MG132, and Western blots of GST pulldowns were probed with anti-Myc antibodies. As seen in Fig. 3B, MLK3-WT was polyubiquitinated in response to insulin (lanes 1 and 2), and mutation of Thr-477 (lanes 3 and 4) resulted in a marked loss of the Mycubiquitin signal. Thus, AKT-mediated phosphorylation is required for insulin-stimulated MLK3 ubiquitination.
Insulin Destabilizes MLK3 via Lys-48-linked Ubiquitination of MLK3 at Lysine 264 -Using an experimental design similar to that in Fig. 3B, we localized insulin-dependent Lys-48-linked ubiquitination to the N-terminal half of MLK3, which harbors the catalytic domain of this kinase. As seen in Fig. 3C, compared with MLK3-WT, deletion of the first 400 amino acids (⌬400) induced a marked attenuation of insulin-stimulated MLK3 ubiquitination. Using identical experimental conditions to those in Fig. 3, B and C, we sought to map the ubiquitin acceptor site in MLK3 protein. Lysine to arginine (Lys-to-Arg) substitu-tion mutants of residues in the catalytic domain of MLK3 were assessed for insulin-induced polyubiquitination. In the absence of a clear known consensus site for Lys-48-linked ubiquitination, we first screened previously generated mutants used for mapping Lys-63-linked ubiquitination of MLK3 (14). Arginine substitution mutants of MLK3 at Lys-130, Lys-180, and Lys-264, were used in GST pulldown experiments followed by anti-Myc Western blots. Whereas insulin treatment induced ubiquitination of MLK3-WT, K130R, and K180R constructs, the K264R (lanes 7 and 8) mutation abolished insulin-stimulated ubiquitination of MLK3. Based on the widely accepted role of Lys-48-linked polyubiquitination in protein degradation, the K264R mutant of MLK3 would be expected to resist insulinstimulated degradation. As shown in Fig. 3E, we found that irrespective of the presence of insulin or an AKT inhibitor, MLK3-K264R did not show a significant change in protein levels (Fig. 3E, lanes 1-3). By contrast, insulin treatment resulted in Ͼ60% degradation of wild-type MLK3 protein, and this steep loss was prevented by the presence of an AKT inhibitor (lanes 4 -6). Inputs were probed for ␤-tubulin used here as control. A bar graph representing results from three independent experiments is shown in Fig. 3F. For direct comparison, cycloheximide chase experiments using MLK3-WT, T477A, S674A, phospho-mutants (Fig. 2G) were performed along with K264R ubiquitin mutant (Fig. 3E). However, for a clearer description of the results, these data have been represented in two independent figures. We conclude that AKT regulates MLK3 protein stability by targeting phosphorylation (Fig. 2G) and subsequent ubiquitination (Fig. 3E).
Although the ubiquitin acceptor mutant of MLK3 appears to be more stable, Fig. 3E revealed noticeably faster migration of MLK3-K264R. The significance of a faster migrating MLK3-K264R mutant can be explained by our previous report showing IL-1␤-stimulated, Lys-63-linked ubiquitination at Lys-264 (see schematic in Fig. 3G) (14). Because Lys-63-linked ubiquitination aids the dimerization, autophosphorylation, and activation of MLK3, mutation of Lys-264 also renders MLK3 inactive. As shown in Fig. 3H, we transfected GST-MLK3-WT, KD, or K264R mutants of MLK3 into Min6 cells along with GST-JNK (to distinguish from faster migrating endogenous JNK) and used the extracts for GST pulldowns followed by Western blotting for JNK phosphorylation. Consistent with the literature, and unlike MLK3-KD (lane 2), MLK3-WT expression was sufficient to induce autophosphorylation, as well as the phosphorylation of JNK (lane 3). A brief treatment of the cells with MLK inhibitor CEP11004 rapidly inhibited phosphorylation of MLK3 and GST-JNK (lane 4), whereas the K264R mutant of MLK3 was inactive and showed no JNK phosphorylation and minimal autophosphorylation. Taken together, these data clearly demonstrate that AKT regulates MLK3 by precluding activation-specific Lys-63-linked ubiquitination, while stimulating phosphorylation-dependent ubiquitination and proteasomal degradation.
Mutation of AKT Phospho-acceptor Site Augments Proapoptotic Properties of MLK3-We have previously demonstrated a key role for MLK3 in cytokine-induced beta cell apoptosis (6,14). Having found that AKT destabilizes MLK3 protein, we examined whether a point mutant of MLK3 (T477A) that OCTOBER 24, 2014 • VOLUME 289 • NUMBER 43  1 and 2) or Myc-ubiquitin-K48R (lanes 3 and 4). Cells were treated with for insulin for 30 min (lanes 2 and 4), and Western blots using anti-Myc antibodies were to detect ubiquitination of MLK3. Total levels of MLK3 in each lane are shown. B, GST pulldowns of MLK3-WT or T447A mutant from cells coexpressing Myc-Ub-K63R, treated with insulin for 30 min (lanes 2 and  4), show loss of ubiquitination for AKT phosphomutant of MLK3. C, insulin-mediated MLK3 ubiquitination was observed for GST-MLK3 WT but not for the N-terminal deletion of MLK3 (⌬400) using experimental conditions described in B. D, site-specific mutants of potential ubiquitin acceptor lysine residues in GST-MLK3 were examined for insulin-dependent ubiquitination as described in A and show specific loss of ubiquitination upon mutation of Lys-264 in MLK3 protein. E, autoradiogram showing time-dependent, insulin-mediated loss of GST-MLK3 WT (lanes 4 -6), but not GST-MLK3 K264R (lanes 1-3). Total MLK3 protein levels from untreated cells at time 0 are shown in lanes 1 and 4, whereas MLK3 levels from cycloheximide (Cx)-treated cells in the presence (lanes 2 and 4) or absence of (lanes 3 and 6) AKT inhibitor are shown. Note that for a direct comparison, experiments on stability of GST-MLK3 K264R were performed alongside corresponding MLK3 phospho-mutants T477 A and S674A (Fig. 2G). F, densitometric analysis MLK3 protein stability using conditions described in E are presented as a bar graph and represent means Ϯ S.E. of three independent experiments. *, p Ͻ 0.01 versus MLK3-WT treatment with insulin and **, p Ͻ 0.01 versus insulin (Ins) treatment of MLK3-WT in the presence of AKT inhibitor. G, schematic representation of MLK3 protein highlighting the acceptor lysine  3 and 4), and K264R mutant (lane 5), followed by Western blotting using phosho-JNK as well as phospho-MLK3-specific antibodies (middle panels). 500 nM MLK3 inhibitor Cep111004 was used as a control to demonstrate impact of MLK3 inhibition on JNK activity (lane 4). Total GST-MLK3 and GST-JNK input controls are shown. escaped AKT-dependent phosphorylation could increase its proapoptotic impact. Min6 insulinoma cells were transfected with GST control vector, GST-MLK3-WT, or GST-MLK3-T477A, and cell death was assessed using TUNEL assay in the presence or absence of cytokines (mixture of TNF-␣, IL-1␤, and IFN-␥). In Min6 cells, we generally observed cytokine-stimulated death as early as 16 h, peaking at 24 -36 h. Consistent with its role as a competence factor for beta cell apoptosis (6), MLK3-expressing cells are sensitized to the effect of cytokines, resulting in precocious cell death. To capture this accelerated time frame, MLK3-expressing cells were treated with cytokines for a sub-lethal period of 8 h. Cells were co-stained for MLK3 expression using anti-GST antibodies, and TUNELϩ cells were expressed as a percent of total transfected MLK3ϩ cells. In vector controls, cytokine treatment showed little change in cell death. MLK3-WT expression resulted in apoptosis of roughly 20 -30% of the transfected cells, which was further augmented (2-3-fold) in the presence of cytokines. Interestingly, MLK3-T477A expression strongly induced cell death, resulting in apoptosis of nearly 60% of the transfected cells, comparable with levels observed in cytokine-treated MLK3-WT expressing cells. More importantly, unlike MLK3-WT, addition of cytokines did not further augment MLK3-T477A-induced death of Min6 cells. These data clearly indicate that MLK3-T477A is already active at a maximal level and imply that AKT is likely the main negative regulator of MLK3 under prosurvival conditions.

TRB3 Modulates Cytokine-activated JNK Signaling Dynamics
Loss of TRB3 Attenuates Cytokine-induced Beta Cell Death-Results shown in the preceding figure would predict that primary beta cells from TRB3 Ϫ/Ϫ mice are resistant to cytokineinduced cell death. To examine this hypothesis, we used an ex vivo co-culture model intended to mimic the islet response to leukocyte invasion or insulitis associated with autoimmune type 1 diabetes. Unlike static culture of islets with cytokine cocktails, this system (depicted by a schematic in Fig. 4B) more authentically recapitulates the complex cytokine environment of insulitis. Splenocytes from WT mice were isolated and stimulated with plate bound anti-CD3 and anti-CD28 antibodies to trigger production of the full repertoire of cytokines as reported (6). For co-culture, isolated islets from WT or TRB3 Ϫ/Ϫ mice were suspended in Transwells and immersed in culture wells containing stimulated or unstimulated (naïve) splenocytes. Islets were thus co-cultured for 24 h, fixed, processed, and examined for activation of cell death. We have reported previously that cytokine induced TRB3 expression in islets and beta cell lines is highly correlated with conformational alteration in BAX, a proapoptotic BH3-only protein of the BCL family of proteins (6). Conformational change is required for and immediately precedes the insertion of BAX into the outer mitochondrial membrane, as detected by a conformation-specific antibody (BAX6A7) (29,30). BAX translocation, oligomerization, and insertion result in cytoplasmic leaching of cytochrome c, ultimately resulting in activation of executioner caspase-3 and induction of apoptosis (31,32). As seen in Fig. 4C, under unstimulated conditions, both WT and TRB3 Ϫ/Ϫ islets showed little or no BAX6A7 signal. Co-culture of WT islets with activated splenocytes resulted in BAX6A7 detection in ϳ8% of the beta cells (Fig. 4E). Consistent with our reported correlation between TRB3 induction and BAX conformational change, TRB3 Ϫ/Ϫ islets showed little signal beyond an occasional BAX6A7ϩ beta cell.
Finally, we examined whether loss of TRB3 expression could protect beta cells from cytokine-induced apoptosis. Again using the co-culture model, we observed strong induction of cell death in WT islets (Fig. 4D). Quantification of TUNELϩ cells showed induction of roughly 15-25% of total insulinϩ cells across multiple islets (Fig. 4F). In TRB3 Ϫ/Ϫ islets, we observed a virtual absence of beta cell death, indicating that loss of TRB3 strongly protected beta cells from apoptosis.

DISCUSSION
Accumulating evidence suggests that the MLK-JNK signaling pathway induces inflammation of metabolic tissues and contributes to impaired glucose homeostasis and decreased insulin sensitivity (33)(34)(35)(36)(37)(38). In the beta cell, cross-talk between MLK-JNK and insulin-AKT signaling determines the balance between dysfunction/death and survival, highlighting the importance of mechanisms that can fine tune and tip the balance toward JNK or AKT activity. We have previously shown that TRB3 is a downstream mediator of MLK3 and is required for induction of cytokine-dependent beta cell inflammation (6). Here, we show that TRB3 is instrumental in regulating the net amount of MLK3 available for participating in JNK signaling, most likely by inhibiting AKT and preventing AKT-mediated proteasomal degradation of MLK3. Thus, TRB3 functions as an allosteric modulator that finetunes the inverse balance between MLK3 and AKT in cytokineinduced beta cell inflammation.
In the beta cell, autocrine insulin action resulting from periodic glucose-stimulated insulin secretion (25,26) is sufficient to maintain the prosurvival effects of AKT. Mice with beta cellspecific deletion of both insulin and IGF1 receptors show profound loss of AKT activity, extensive beta cell death and severe diabetes (39). Similar phenotypes have been reported in mice with a global knock-out of AKT2 (40), and genetic ablation of IRS2 (insulin receptor substrate 2) (41). Additionally, incretins mediate their potent prosurvival effects at least in part by inducing IRS2 expression and subsequent AKT activation (42,43). AKT-dependent regulation of protein stability can be mediated by direct phosphorylation of substrates, as is the case with tuberin, FOXO3, and FOXO1 (44,45). However, AKT also indirectly regulates protein stability via GSK3 to impact beta cell function and survival. GSK3-mediated phosphorylation and proteasomal degradation is a prevalent mechanism (46) and reportedly regulates beta-cell specific transcription factors PDX1 (47,48) and MAF-A (49,50), as well as MCL-1 (51,52), an antiapoptotic BCL family protein.
In addition to the cross-talk between post-translational events of phosphorylation and ubiquitination, this study illustrates how cross-talk between two distinct types of ubiquitination events can control MLK3 function and stability. We have shown that IL-1␤/TRAF6 activates MLK3 by atypical Lys-63linked ubiquitination of Lys-264. Others have also reported both Lys-63-and Lys-48-linked ubiquitination of MLK3 (12,13), but the relevance to specific biological conditions has not been established. In this report, we clarify that competing Lys- OCTOBER 24, 2014 • VOLUME 289 • NUMBER 43 48-and Lys-63-linked ubiquitination events not only regulate the fate of the MLK3 molecule but likely also impact survival of the beta cell. The fact that both ubiquitination events can target the same lysine (Lys-264) residue in the kinase domain of MLK3 underscores the tight regulatory controls on MLK3

TRB3 Modulates Cytokine-activated JNK Signaling Dynamics
(depicted by a model in Fig. 4G) and thereby the importance of MLK3 in beta cell inflammation and death.
Despite the presence of multiple families of MAP3Ks, cytokines appear to signal via a fairly specific JNK signaling module that includes MLK3, MKK7/4, and JNK1, scaffolded by JIP1  Fig. 2.) B, schematic representation of splenocyte islet co-culture experimental model used for examination of beta cell death. C, BAX conformational change was examined in WT and TRB3-KO islets following 24 h of splenocyte and islet co-culture using unstimulated (Unstim) or stimulated (Stim) splenocytes. Immunofluorescence using insulin (pseudocolored blue) and BAX6A7 conformation-specific antibodies (pseudocolored white for clarity) show stimulation-dependent induction of BAX conformational change in WT islets but not in TRB3-KO islets. D, sections generated from the splenocyte and islet co-culture experiments in C were also processed and examined for insulin along with induction of death by TUNEL assay (pseudocolored white). Similar to data seen for BAX6A7, TRB3-KO islets show marked attenuation of cell death. Nuclei were counterstained with DAPI (shown in C only). E and F, quantification of BAX-6A7 and TUNEL-positive beta cells is shown in bar graphs and represent the mean Ϯ S.E. of three separate experiments. *, p Ͻ 0.01 versus WT islets cultured with unstimulated splenocytes. G and H, schematic representation of post-translational modifications of MLK3 (G) and a working model for the cross-talk between MLK3, TRB3, and AKT based on the findings shown in the current study (H). (15)(16)(17). We have reported that TRAF6-mediated, Lys-63linked ubiquitination induces MLK3 dissociation from JIP1 scaffold, enabling its dimerization and activation of the MLK3-JNK cascade (14). Furthermore, JNK activation is sustained by direct phosphorylation of JIP1, which inhibits re-association of MAP3Ks with JIP1 (53). Notably, Xu et al. (18) have demonstrated that upon cytokine treatment, each component of the JNK activation cascade contributes to the mutual stabilization of the entire module sustaining a feed-forward loop of MLK3-MKK7 and JNK activation. This report shows that TRB3 is required for optimal dynamics of JNK signaling such that loss of TRB3 decreases both the amplitude and duration of MLK3-JNK activation. In neurons, duration of MLK-JNK signaling is regulated by a negative feedback mechanism, in which dephosphorylation of JNK is induced by Jun/AP1 induced expression of MAP kinase phosphatase 1 (54). By contrast, a drop in total MLK3 protein levels could explain a decrease in both the amplitude and duration of JNK activation. As depicted in Fig. 4H, we propose that in the absence of TRB3, increased AKT activity could rapidly induce MLK3 degradation, decreasing total amounts of MLK3 available for JNK activation.
Results presented in this study show that a change in kinetics of MLK3-JNK activation can impact beta cell inflammation and survival. Downstream biological effects resulting from shifts in signaling dynamics may not be uniform (55,56) and could be likened to continuous analog versus discrete digital outputs. For example, increase or decrease in kinase signaling may result in a proportional shift in downstream effects. Alternately, a specific amplitude and duration of signaling may be required for attaining a signaling threshold, triggering a discrete switch-like "all or nothing" response (discussed in Ref. 57).
Thus, in addition to signal-dependent activation of distinct pathways and kinase isoforms, signaling agonists may elicit specific biologic responses via the information encoded within the dynamics of signal activation (57). It would follow, therefore, that in beta cells, the downstream impact of TRB3-dependent allosteric fine-tuning of signaling mechanisms could be complex and nuanced and merits further inquiry.