Peroxisome Proliferator-activated Receptor γ-independent Activation of p38 MAPK by Thiazolidinediones Involves Calcium/Calmodulin-dependent Protein Kinase II and Protein Kinase R

The thiazolidinediones (TZDs) are synthetic peroxisome proliferator-activated receptor γ (PPARγ) ligands that promote increased insulin sensitivity in type II diabetic patients. In addition to their ability to improve glucose homeostasis, TZDs also exert anti-proliferative effects by a mechanism that is unclear. Our laboratory has shown that two TZDs, ciglitazone and troglitazone, rapidly induce calcium-dependent p38 mitogen-activated protein kinase (MAPK) phosphorylation in liver epithelial cells. Here, we further characterize the mechanism responsible for p38 MAPK activation by PPARγ ligands and correlate this with the induction of endoplasmic reticulum (ER) stress. Specifically, we show that TZDs rapidly activate the ER stress-responsive pancreatic eukaryotic initiation factor 2α (eIF2α) kinase or PKR (double-stranded RNA-activated protein kinase)-like endoplasmic reticulum kinase/pancreatic eIF2α kinase, and that activation of these kinases is correlated with subsequent eIF2α phosphorylation. Interestingly, PPARγ ligands not only activated calcium/calmodulin-dependent kinase II (CaMKII) 2-fold over control, but the selective CaMKII inhibitor, KN-93, attenuated MKK3/6 and p38 as well as PKR and eIF2α phosphorylation. Although CaMKII was not affected by inhibition of PKR with 2-aminopurine, phosphorylation of MKK3/6 and p38 as well as eIF2α were significantly reduced. Collectively, these data provide evidence that CaMKII is a regulator of PKR-dependent p38 and eIF2α phosphorylation in response to ER calcium depletion by TZDs. Furthermore, using structural derivatives of TZDs that lack PPARγ ligand-binding activity as well as a PPARγ antagonist, we show that activation of these kinase signaling pathways is PPARγ-independent.

The thiazolidinedione (TZD) 1 drug class was created over 20 years ago with the synthesis of ciglitazone, an analog of the hypolipidemic agent clofibrate (1). The TZDs have since grown to include three other members, troglitazone, rosiglitazone, and pioglitazone, which function to improve metabolic control in patients with type II diabetes by increasing insulin sensitivity in adipose tissue, muscle, and liver (2). Following their discovery, it was learned that TZDs are ligands for the gamma isoform of the peroxisome proliferator-activated receptor (PPAR␥) (3). From this, it was understood that these agents exert their insulin-sensitizing effects primarily through PPAR␥dependent transcription of genes involved in glucose and lipid metabolism and energy balance.
In addition to their ability to promote effective glycemic control, TZDs have also been shown to exert growth inhibitory effects in multiple cell and animal models (4 -7). This is consistent with the observation that PPAR␥ is both necessary and sufficient to promote adipocyte differentiation (8). However, the sensitivity of various cell lines to TZD-induced growth inhibition does not correlate with levels of PPAR␥ expression (9). This suggests that TZDs have distinct PPAR␥-independent effects that are important for these additional, unanticipated mechanisms of action. In support of this hypothesis, troglitazone was shown to equally inhibit proliferation of both PPAR␥ Ϫ/Ϫ and PPAR␥ ϩ/ϩ mouse embryonic stem cells (10). In addition, structural derivatives of ciglitazone and troglitazone that lack PPAR␥ ligand-binding activity were demonstrated to prevent growth of prostate cancer cells. 2 Interestingly, these derivatives exhibited more potent growth inhibitory effects * This work was supported in part by the National Institute of Health with Public Health Service Grants and the Environmental Protection Agency Science to Achieve Results Award R-82921401-0. 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.
Both our laboratory and others have shown that TZDs influence mitogen-activated protein kinase (MAPK) activity (11)(12)(13)(14). The three best characterized mammalian MAPKs, extracellular signal-regulated kinase (Erk), p38, and c-Jun N-terminal kinase (JNK), are known to play important roles in coordinating a variety of cellular processes, including growth, differentiation, and, in some cases, apoptosis (15). MAPKs are activated via a three-tiered, phospho-relay mechanism whereby a MAPK kinase kinase (MKKK or MEKK) phosphorylates an MAPK kinase (MKK or MEK), which subsequently phosphorylates a MAPK. For example, p38 is selectively phosphorylated by MKK3 and MKK6 (16,17); on the other hand, multiple MEKKs are known to converge on a particular MKK (18). This increases the complexity and selectivity of these signaling pathways such that distinct MAPK cascades can be activated both independently and simultaneously in a stimulus-specific manner. In general, the Erk pathway is activated primarily by mitogens, whereas JNK and p38 are preferentially activated by environmental stresses and inflammatory cytokines. Activation of MAPKs results in phosphorylation of transcription factors that increase the expression of target genes. Thus, in addition to their PPAR␥ ligand-binding activity, the ability of TZDs to induce MAPK phosphorylation may represent an additional pathway by which these agents affect cell growth and differentiation. It is therefore important to understand how these agents induce MAPK signaling as well as to determine the role of PPAR␥ in this process.
We previously characterized the mechanism by which ciglitazone and troglitazone acutely activate members of the MAPK family in GN4 rat liver epithelial cells (11). Interestingly, our data suggest that PPAR␥ is not required for TZD-induced MAPK phosphorylation. In the current study, we have utilized the aforementioned structural derivatives of ciglitazone and troglitazone that are devoid of PPAR␥ ligand-binding activity in addition to a PPAR␥ antagonist to provide definitive evidence that MAPK activation by TZDs in GN4 cells is indeed PPAR␥-independent. Additionally, we have further characterized the mechanism responsible for p38 activation by TZDs and provide evidence suggesting a link between p38 activation and induction of endoplasmic reticulum stress.
Cell Culture-GN4 rat liver epithelial cells were grown in Richter's minimal essential medium supplemented with 10% heat-inactivated FBS and penicillin/streptomycin as detailed earlier (23). A549 human lung carcinoma cells were purchased from ATCC and similarly propagated in Dulbecco's modified Eagle's medium/Nutrient mixture F-12 (1/1) supplemented with serum and antibiotics as above as well as 1.5 g/L sodium bicarbonate and 2 mM L-glutamine. Prior to experiments, cells at 70 -80% confluency were serum-starved overnight in the appropriate medium containing 0.1% FBS.
Immunoblotting-In a typical experiment, 10 g of cell lysate was resuspended in SDS-PAGE sample buffer (0.5 M Tris, pH 6.8, 4% SDS, 20% glycerol, 10% ␤-mercaptoethanol, 0.1% bromphenol blue) and heated at 95°C for 5 min to denature proteins. Lysates were then resolved by SDS-PAGE on Novex pre-cast 10% Tris-glycine gels (Invitrogen) and transferred to polyvinylidene fluoride (Immobilon-P, Millipore). Immunoblots were incubated with the appropriate primary antibody overnight at 4°C, washed 3ϫ with TBST, and probed with horseradish peroxidase-conjugated secondary antibodies for 1 h at room temperature. Immunoblots were then developed with ECL (Amersham Pharmacia Biotech) according to the manufacturer's instructions and visualized by autoradiography (X-Omat Blue film, Eastman Kodak). In certain instances, membranes were stripped in buffer (62.5 mM Tris, pH 6.7, 2% SDS, 100 mM ␤-mercaptoethanol) at 55°C for 30 min and reprobed with another antibody.
Immunoprecipitation-Following stimulation, cells were rinsed as described above and scraped into ice-cold radioimmune precipitation assay buffer without SDS. The lysates were cleared by centrifugation. 300 g of cell lysate was immunoprecipitated by incubation with the appropriate primary antibody overnight at 4°C under slight agitation. Fifteen microliters of Protein A-agarose beads (Santa Cruz Biotechnology) were added to each sample, which were then incubated an additional hour at 4°C. Immune complexes were collected by brief centrifugation and washed four times in ice-cold lysis buffer. Remaining wash buffer was carefully removed with a Hamilton syringe. Immune complexes were then resuspended in SDS-PAGE sample buffer and resolved by SDS-PAGE as described above.
In Vitro CaMKII Kinase Assay-CaMKII kinase activity in stimulated GN4 cells was measured using a commercial CaMKII assay kit (Upstate Biotechnology) according to the manufacturer's instructions with slight modification. Briefly, CaMKII was immunoprecipitated from 400 g of cell lysate in HEPES binding buffer (20 mM HEPES, pH 7.7, 50 mM NaCl, 2.5 mM MgCl 2 , 0.1 mM EDTA, 0.5% Triton X-100) by overnight incubation with an anti-CaMKII antibody (H-300, Santa Cruz Biotechnology). CaMKII activity present in immune complexes was assessed by measuring the transfer of the ␥-phosphate of [␥-32 P]ATP to a specific CaMKII substrate peptide for 10 min at 30°C. Phosphorylated substrate was then separated from residual [␥-32 P]ATP using P-81 phosphocellulose paper (Whatman) and quantified with a scintillation counter.

Src-dependent Activation of EGFR and Erk by Ciglitazone
Does Not Require PPAR␥-Previously, we have shown that ciglitazone-induced Erk phosphorylation in rat liver epithelial (GN4) cells required Src-dependent EGFR transactivation (11). Interestingly, Src kinase activity was unaffected by a PPAR␥ antagonist suggesting that ciglitazone activates this signaling pathway via a non-genomic mechanism. We recently obtained derivatives of two thiazolidinediones, ⌬2-ciglitazone and ⌬2troglitazone ( Fig. 1), which completely lack PPAR␥ ligandbinding activity. 2 Because pharmacological inhibitors often have nonspecific effects, we tested whether ⌬2-ciglitazone retained the ability to activate EGFR and Erk in GN4 cells. Similar to ciglitazone, treatment of GN4 cells with ⌬2-ciglita-zone induced rapid, robust EGFR phosphorylation that correlated with Erk activation (Fig. 2, A and B). In contrast, ⌬2troglitazone failed to significantly activate EGFR and Erk, consistent with our previous finding that troglitazone was a much weaker Erk activator than ciglitazone in GN4 cells (11). These data demonstrate that the ⌬2-derivatives act similarly to their parent compounds with respect to EGFR and Erk phosphorylation.
Because Src is required for EGFR transactivation and downstream Erk phosphorylation by ciglitazone (11), we tested whether this kinase played a similar, necessary role in signaling events following exposure to ⌬2-ciglitazone. Using an antibody that recognizes the EGFR when phosphorylated at Tyr 845 , a putative Src-specific phosphorylation site (24), we observed that ⌬2-ciglitazone, like ciglitazone, induced Tyr 845 phosphorylation (Fig. 3). Pretreatment of cells with PP2, a selective Src kinase inhibitor, completely blocked ⌬2-ciglitazone-induced EGFR phosphorylation at Tyr 845 as well as downstream Erk activation. In contrast, pretreatment of cells with PP3, a pharmacologically inactive analog of PP2, did not affect the ability of ⌬2-ciglitazone to induce either EGFR or Erk phosphorylation. Thus, these data show that EGFR transactivation as well as Erk phosphorylation by ⌬2-ciglitazone requires Src. Collectively, these findings corroborate our earlier studies using a PPAR␥ antagonist and, importantly, provide additional evidence that TZDs activate PPAR␥-independent kinase signaling pathways in GN4 cells.
⌬2-Thiazolidinediones Are Less Effective Activators of the p38 MAPK Pathway-In addition to their effects on the EGFR and Erk, we previously showed that both ciglitazone and troglitazone rapidly activate p38 in GN4 cells (11). PPAR␥ ligandinduced p38 phosphorylation occurred in an EGFR-independent manner and was instead correlated with increases in intracellular calcium and activation of the calcium-dependent tyrosine kinase Pyk2. Here, we tested whether ⌬2-ciglitazone and ⌬2-troglitazone retained the ability to similarly activate Pyk2 and p38. Interestingly, both of the ⌬2-derivatives showed significantly reduced efficacy in activating p38 when compared with the parent compounds (Fig. 4A). Examination of MKK3/6, the MAP kinase kinase immediately upstream of p38, and Pyk2 phosphorylation (Fig. 4B) revealed similar findings; ⌬2ciglitazone and ⌬2-troglitazone failed to activate these kinases as strongly as ciglitazone and troglitazone, respectively.
To further evaluate the role of PPAR␥ in TZD-dependent p38 activation, we pretreated GN4 cells with the PPAR␥ antagonist GW9662 prior to stimulation with ciglitazone, troglitazone, or their respective ⌬2-derivatives. GW9662 failed to inhibit the ability of ciglitazone or troglitazone to activate MKK3/6 and p38 (Fig. 4C). Importantly, co-treatment of GN4 cells with GW9662 and either of the ⌬2-TZDs did not potentiate the lack of effect on MKK3/6 and p38 phosphorylation in response to these derivatives alone. Together, these findings suggest that TZD-induced p38 phosphorylation, like Erk activation, is also PPAR␥-independent. In support of this hypothesis, structur- ally related TZDs and more potent PPAR␥ agonists, rosiglitazone and pioglitazone, were unable to activate p38 MAPK in GN4 cells (data not shown).
Role of Pyk2 as an Upstream Activator of p38 -Because our previous studies suggested a link between Pyk2 and p38 MAPK, we further examined if activation of Pyk2 was necessary for the effects of PPAR␥ ligands on p38 phosphorylation. Specifically, GN4 cells were infected with a recombinant adenovirus encoding either bacterial ␤-galactosidase (Ad.lacZ) or a C-terminal inhibitory form of Pyk2 (Ad.CRNK) prior to stimulation with PPAR␥ ligands. CRNK represents an alternative potential splice variant of Pyk2 (21), which functions to nega-tively regulate endogenous Pyk2 autophosphorylation (25). Angiotensin II-dependent Pyk2 phosphorylation in GN4 cells was reduced to near basal levels following infection with increasing amounts of Ad.CRNK with maximal inhibition observed using 4 ϫ 10 6 plaque-forming units/ml, a titer that also produced significant CRNK expression (data not shown). Therefore, this dose was used in subsequent studies. Although Ad.lacZ had no affect on PPAR␥ ligand-dependent Pyk2 phosphorylation, overexpression of CRNK significantly blunted Pyk2 activation by both ciglitazone and troglitazone (Fig. 5A). Immunoblots from the same lysates revealed that, in contrast to Pyk2, CRNK failed to prevent activation of either MKK3/6 or p38 MAPK (Fig. 5B). These results suggest that independent but parallel, calcium-regulated signaling pathways activate Pyk2 and p38 in this cell type.
CaMKII but Not PKC Is Required for MKK3/6 and p38 Activation by PPAR␥ Ligands-Our earlier work with BAPTA-AM showed that intracellular calcium was required for both Pyk2 and p38 activation in response to PPAR␥ ligands (11). Yet, PPAR␥ agonist-induced MKK3/6 and p38 phosphorylations are unaffected by adenoviral-mediated overexpression of a Pyk2 dominant negative, conditions that prevented Pyk2 activation in response to these drugs. Together, these findings suggest that an additional calcium-activated kinase is induced by PPAR␥ ligands, ultimately leading to p38 phosphorylation. Similar to Pyk2, calcium/calmodulin-dependent kinase II (CaMKII) and the classic protein kinase Cs (PKC␣, -␤, and -␥) represent other calcium-regulated kinases that could be acti- FIG. 4. ⌬2-TZDs are weaker activators of p38 MAPK and Pyk2 in GN4 cells. Serum-starved GN4 cells were stimulated with ciglitazone, troglitazone, or their respective ⌬2-derivatives (50 M) for 10 min. Me 2 SO (DMSO, 0.1%) served as the vehicle control. A, 10 g of cell lysate was resolved by SDS-PAGE. Activation of MKK3/6 and p38 were determined by immunoblotting (IB) with anti-phospho MKK3/6 (p-MKK3/6) and anti-phospho p38 (p-p38) antibodies, respectively. Blots were stripped and reprobed for total p38 as described. B, Pyk2 was immunoprecipitated (IP) from 300 g of cell lysate, and immune complexes were resolved by SDS-PAGE. Immunoblots were probed with a pan anti-phosphotyrosine, PY99, antibody to detect changes in Pyk2 phosphorylation. Total Pyk2 was assessed using an anti-Pyk2 antibody. C, prior to stimulation, some cells were preincubated with the PPAR␥ antagonist, GW9662 (1 M, 1 h). vated by PPAR␥ ligands. To test the involvement of PKC in p38 activation by PPAR␥ agonists, classic PKCs were depleted by chronic exposure of GN4 cells to TPA. Although PKC depletion completely prevented transient, TPA-dependent Erk activation (data not shown), ciglitazone and troglitazone-induced p38 phosphorylation was not affected (Fig. 6A). This result suggests that classic PKCs are not involved in PPAR␥ ligand-dependent p38 activation.
To determine whether CaMKII is necessary for p38 phosphorylation in response to PPAR␥ agonists, the ability of ciglitazone and troglitazone to activate p38 in the presence of the selective CaMKII inhibitor, , was evaluated. Interestingly, pretreatment of GN4 cells with KN-93 blocked both MKK3/6 as well as p38 activation by PPAR␥ ligands (Fig. 6B). These effects were specific for KN-93, because KN-92, a pharmacologically inactive analog of KN-93, did not prevent activation of these kinases.
Further, the ability of ciglitazone and troglitazone to directly stimulate CaMKII activity was examined. CaMKII was immunoprecipitated from cell lysates, and in vitro kinase assays were performed using a specific substrate peptide. Ciglitazone and troglitazone significantly increased CaMKII activity 2-fold over vehicle control within 10 min, a time that coincides with maximal p38 phosphorylation (Table I). Consistent with its documented ability to directly inhibit CaMKII (26), KN-93 significantly blunted PPAR␥ ligand-induced CaMKII activation. In contrast to its effects on CaMKII, MKK3/6, and p38, KN-93 failed to prevent TZD-induced Pyk2 phosphorylation (Fig. 6C). Collectively, these findings support our hypothesis that CaMKII, and not Pyk2, acts upstream of MKK3/6 and p38 in this signaling pathway.
⌬2-Ciglitazone and ⌬2-troglitazone were less effective MKK3/6 and p38 activators than their parent TZDs in GN4 cells (Fig. 4). Because MKK3/6 and p38 phosphorylation in response to ciglitazone and troglitazone was KN-93-sensitive, we assessed whether the ⌬2-derivatives failed to potently activate CaMKII. Indeed, CaMKII kinase assays revealed that ⌬2-ciglitazone and ⌬2-troglitazone were significantly weaker CaMKII activators than their parent compounds (Table I). These data further support our finding that CaMKII is necessary for p38 activation in response to PPAR␥ ligands and is consistent with the inability of ⌬2-TZDs to effectively induce MKK3/6 and p38 phosphorylation.
PPAR␥ Ligands Induce ER Stress and eIF2␣ Phosphorylation in GN4 Cells-Ciglitazone and troglitazone were reported to rapidly increase cytosolic calcium by inhibiting capacitative calcium entry into the endoplasmic reticulum (ER) (10). Calcium is actively involved in proper protein folding and transport within the ER such that its depletion leads to the accumulation of misfolded proteins and subsequent ER stress (27,28). To test whether PPAR␥ ligands induce ER stress in GN4 cells, we assessed the ability of these compounds to rapidly phosphorylate PERK, a recently discovered ER-resident serine/ threonine kinase whose activity is increased selectively in response to chemical-induced ER stress (29,30). Immunoblotting with a phospho-PERK antibody revealed that treatment of GN4 cells with ciglitazone and troglitazone induced a time-dependent increase in PERK activation (Fig. 7A). We also ob-  a Denotes statistical significant differences from ciglitazone or troglitazone (10-min treatments) as appropriate (p Ͻ 0.01; one-way ANOVA).
b Denotes statistical significant differences from ciglitazone or troglitazone (10-min treatments) as appropriate (p Ͻ 0.05; one-way ANOVA).
served PERK activation in response to thapsigargin, an agent known to induce ER stress through calcium store depletion (data not shown). PERK phosphorylation was first observed ϳ5-10 min following exposure to PPAR␥ ligands and occurred within the same time frame as activation of other calciumregulated signaling pathways (e.g. Pyk2 and CaMKII, MKK3/6, and p38). Similar to our earlier data, ⌬2-ciglitazone and ⌬2troglitazone were unable to activate PERK to the same extent as the parent compounds (Fig. 7B).
An early response to ER stress is to inhibit protein synthesis by blocking translation initiation (31). Phosphorylation of the ␣ subunit of eukaryotic initiation factor 2 (eIF2␣) at Ser 51 by several protein kinases, including PERK results in an inhibition of mRNA translation (30,(32)(33)(34)(35). Because PPAR␥ ligands induced PERK activation, we determined whether eIF2␣ was also phosphorylated. Using a phospho-specific antibody that detects eIF2␣ when phosphorylated at Ser 51 , we observed that ciglitazone and troglitazone induced eIF2␣ phosphorylation at this site as early as 15 min after treatment and that phosphorylation was sustained for at least 1 h (Fig. 8A). Phosphorylation of eIF2␣ was also observed in response to thapsigargin (data not shown). Similar to our observations with PERK, ⌬2-ciglitazone and ⌬2-troglitazone were unable to induce eIF2␣ phosphorylation to the same extent as ciglitazone and troglitazone, respectively (Fig. 8B).
Role of PKR in PPAR␥ Ligand-dependent p38 Activation-Depletion of ER calcium can also lead to activation of doublestranded RNA-activated protein kinase or PKR (36). Interestingly, PKR is not only involved in ER stress signaling, but was also recently shown to contribute to activation of the p38 pathway in response to both endotoxin and pro-inflammatory cytokines (37). Because PPAR␥ ligand-induced p38 phosphorylation is accompanied by activation of ER stress signaling pathways, we tested whether PKR played a role in TZD-dependent MKK3/6 and p38 phosphorylation. Pretreatment of GN4 cells with 2-aminopurine, an adenine analog inhibitor of PKR activity (38), reduced MKK3/6 and p38 phosphorylation in response to PPAR␥ ligands (Fig. 9A). This inhibitory effect was selective for the p38 pathway, because the EGFR and Erk were still activated by ciglitazone in the presence of 2-aminopurine (data not shown). In addition to MKK3/6 and p38, inhibition of PKR with 2-aminopurine also blunted PPAR␥ agonist-dependent eIF2␣ phosphorylation (Fig. 9B), a finding consistent with the established role of PKR as an ER stress-induced eIF2␣ kinase (39).
Because a PKR inhibitor abolished TZD-dependent activation of p38 MAPK and ER stress signaling pathways, we assessed whether these compounds affected PKR activity. Because commercially available phospho-specific PKR antibodies are marketed for detection of human or mouse PKR, and GN4 cells are not amenable to transfection, we determined the effect of TZDs on PKR activity in A549 human lung carcinoma cells. PPAR␥ ligands were recently reported to rapidly induce ER stress in these cells (40), suggesting that pathways similar to those we have demonstrated here in GN4 are also activated in A549. Indeed, stimulation of serum-starved A549 cells with ciglitazone and troglitazone led to a time-dependent increase in MKK3/6 and p38 phosphorylation (Fig. 9C). Interestingly, both TZDs also increased PKR phosphorylation at times that paralleled MAPK activation. Similar to our findings in GN4 cells, troglitazone was a stronger activator of MKK3/6 and p38 than ciglitazone. Consistent with a role for PKR as an upstream mediator of TZD-dependent MAPK phosphorylation, troglitazone also induced greater PKR activation than ciglitazone. Furthermore, pretreatment of A549 cells with 2-aminopurine prior to stimulation with PPAR␥ ligands not only blunted PKR phosphorylation, as expected, but also decreased TZD-induced MKK3/6 and p38 activation (Fig. 9D).
CaMKII Is Correlated with PKR Activation in Response to PPAR␥ Ligands-Collectively, the data presented here suggest a role for both CaMKII and PKR as upstream regulators of PPAR␥ ligand-dependent p38 activation. To determine if activation of these two kinases are correlated in response to PPAR␥ agonists, we tested whether PKR inhibition could influence TZD-induced CaMKII activation. In the presence of 2-aminopurine, PPAR␥ ligand-dependent CaMKII activity was not affected (Table I), suggesting that CaMKII is activated independently of PKR. We then performed the converse of this experiment and assessed whether CaMKII inhibition influenced PKR activity. Interestingly, the CaMKII inhibitor KN-93, but not the structurally inactive analog KN-92, significantly blunted PPAR␥ ligand-dependent PKR phosphorylation (Fig. 10A). Similar results were observed for eIF2␣; eIF2␣ phosphorylation in response to ciglitazone and troglitazone was blunted only in the presence of KN-93 (Fig. 10B). Because activation of p38 in response to ciglitazone and troglitazone was also KN-93-and 2-aminopurine-sensitive, we determined whether p38 was necessary for eIF2␣ phosphorylation. In contrast to KN-93, inhibition of p38 with SB203580 had no affect on PPAR␥ ligand-induced eIF2␣ phosphorylation (data not shown). Because PERK is not only known to phosphorylate eIF2␣ in response to ER stress but is also activated in GN4 cells by ciglitazone and troglitazone, we tested whether KN-93 affected PPAR␥ ligand-induced PERK activation. Pretreatment of cells with KN-93 was unable to prevent PERK phosphorylation in response to TZDs (data not shown). Collectively, these data suggest that CaMKIIdependent activation of PKR is critical for both p38 and eIF2␣ phosphorylation by PPAR␥ ligands. DISCUSSION The observation that PPAR␥ ligands rapidly activate MAPKs by our laboratory and others (11)(12)(13)(14) raises three important questions: 1) What is the mechanism for activation of MAPKs by these compounds? 2) Is PPAR␥ ligand-binding activity required for MAPK phosphorylation? and 3) Does PPAR␥ Lysates were prepared as described under "Experimental Procedures." Anti-phospho PKR and anti-phospho eIF2␣ antibodies, respectively, were used to determine changes in PKR (A) and eIF2␣ (B) phosphorylation. Blots were stripped and reprobed with anti-PKR or anti-actin antibodies.
ligand-induced MAPK phosphorylation contribute to the pharmacological effects of these agents? Our previous work, which focused on how PPAR␥ agonists activated MAPKs in a rat liver epithelial cell line, showed that Erk phosphorylation required Src-dependent EGFR transactivation (11). Moreover, we found that p38 activation occurred independently of the EGFR and instead was correlated with Pyk2 phosphorylation and increases in intracellular calcium. In the current study, we have elaborated on the mechanism of p38 activation by PPAR␥ ligands and also determined the role of PPAR␥ in this process.
To directly address the question of whether PPAR␥ is necessary for TZD-induced MAPK activation in GN4 cells, we have utilized chemical derivatives of these agents that are devoid of ligand-binding activity. The rationale for synthesizing the ⌬2derivatives came from the observation that insertion of a double bond adjoining the terminal thiazolidine-2,4-dione ring abolished PPAR␥ ligand-binding activity (41), presumably due to increased structural rigidity surrounding the heterocyclic system. Here, we provide evidence that ⌬2-ciglitazone retained the ability to induce Src-dependent EGFR phosphorylation and downstream Erk activation. This clearly shows that PPAR␥ is not required for TZD-induced Erk phosphorylation in GN4 cells and supports our earlier studies with a PPAR␥ antagonist (11).
Interestingly, ⌬2-ciglitazone and ⌬2-troglitazone failed to activate CaMKII, Pyk2, and p38 with the same efficacy as their parent compounds. Based on our earlier data examining the effects of ciglitazone and ⌬2-ciglitazone on EGFR/Erk signaling, these findings were both unexpected and intriguing. Although this observation suggests a role for PPAR␥ in these signaling events, the PPAR␥ antagonist GW9662 had no affect on TZD-dependent p38 phosphorylation. Moreover, structurally related TZDs with higher PPAR␥ binding affinity do not activate p38 in this cell model. Together, these observations suggest that activation of the p38 MAPK pathway in GN4 cells by ciglitazone and troglitazone is non-genomic. As opposed to a requirement for PPAR␥, we propose that the failure of the ⌬2-TZDs to strongly activate p38 results instead from an inability of these compounds to efficiently mobilize calcium from the ER. This hypothesis is discussed in more detail below.
While investigating the mechanism of PPAR␥ ligand-induced p38 phosphorylation, we observed that p38 and Pyk2 are activated by parallel but independent pathways (Fig. 11). Although the role of Pyk2 as an upstream activator of p38 MAPK has been reported in response to several different stimuli (11,22,42), the lack of a highly selective Pyk2 inhibitor has made this difficult to substantiate. CRNK (a C-terminal inhibitory form of Pyk2) is a Pyk2-targeted inhibitor thought to compete with the full-length protein for binding partners necessary for activation (25). Adenovirus expression of CRNK significantly reduced Pyk2 activation by ciglitazone and troglitazone as expected, yet had no affect on either MKK3/6 or p38 phosphorylation by PPAR␥ ligands. Although these data do not refute our earlier finding that Pyk2 and p38 activations are correlated, it further defines this mechanism. Specifically, the current studies illustrate that Pyk2 and p38 phosphorylations in response to PPAR␥ ligands occur via calcium-dependent pathways that may be parallel but independent.
The data presented here identify a role for CaMKII in TZDinduced p38 phosphorylation. CaMKII has previously been implicated as an activator of the p38 MAPK pathway in response to calcium signals in neurons (43). In the current study, both ciglitazone and troglitazone time dependently increased CaMKII kinase activity ϳ2-fold over vehicle control at times that coincide with maximal p38 phosphorylation. Moreover, the selective CaMKII inhibitor KN-93, but not its structurally inactive analog KN-92, reduced PPAR␥ ligand-dependent MKK3/6 and p38 activation to near basal levels yet had no affect on Pyk2 activation. Collectively, these data not only support our findings with CRNK, but also further suggest that Pyk2 and p38 are activated by separate pathways. Instead, CaMKII is a critical upstream activator of p38 phosphorylation in GN4 cells in response to PPAR␥ ligands. Although PKC is also activated by increases in intracellular calcium, we found no evidence (by TPA down-regulation) for involvement of this kinase family in PPAR␥ agonist-dependent p38 activation. Similar to our previous findings with BAPTA-AM (11), neither CRNK overexpression nor KN-93 prevented EGFR and Erk phosphorylation in response to these compounds (data not shown). Importantly, this provides further evidence that PPAR␥ ligands activate two distinct kinase pathways in GN4 cells.
We hypothesize that the intracellular calcium required for activation of these kinases in GN4 cells is derived from the endoplasmic reticulum (ER). Ciglitazone and troglitazone were previously shown to promote calcium release from the ER leading to PKR-dependent phosphorylation of eIF2␣, translation inhibition, and ultimately growth arrest (10). Together, these observations suggest that PPAR␥ ligands cause ER stress. In support of this idea, we show here that TZDs not only induce eIF2␣ phosphorylation but also activate PERK and PKR. Although TZD-dependent PKR activation was previously reported (10), this is the first evidence that these compounds also influence PERK activity; our finding is significant because PERK phosphorylation was previously observed only in response to ER and not cytoplasmic stress (30). Moreover, PERK and PKR are important for ER stress-induced eIF2␣ phospho- FIG. 11. Schematic representation of the proposed mechanism responsible for MAPK activation by PPAR␥ ligands in GN4 cells. Stimulation of rat liver epithelial cells with PPAR␥ ligands activates two distinct, PPAR␥-independent kinase-mediated signaling cascades resulting in MAPK phosphorylation. Ciglitazone induces Src-dependent EGFR transactivation, leading to the recruitment of Grb2 and SOS, which in turn promote Ras activation and subsequent Erk phosphorylation. Simultaneously, both ciglitazone and troglitazone induce EGFRindependent activation of p38 MAPK that is correlated with increases in intracellular calcium and Pyk2 activation. Specifically, TZDs promote calcium release from the ER-activating CaMKII and the ER stress-responsive kinase PERK. CaMKII, most likely via serine phosphorylation of PACT, promotes activation of PKR. PKR then facilitates downstream MKK3/6 and p38 as well as eIF2␣ phosphorylation. Collectively, modulation of MAPK activity and eIF2␣ phosphorylation is a potential mechanism for how TZDs affect cell growth in a PPAR␥independent manner. rylation (39,44). Indeed, a PKR-selective inhibitor attenuated PPAR␥ ligand-dependent eIF2␣ phosphorylation in GN4 cells. Collectively, these data suggest that intracellular calcium mobilization by PPAR␥ ligands in GN4 cells interferes with proper protein folding in the ER thus promoting ER stress.
These findings have led us to hypothesize that disruptions in ER calcium homeostasis are not only important for ER stress, but also provide a mechanistic link identifying the intracellular source of calcium required for PPAR␥ ligand-induced CaMKII, Pyk2, and p38 activation (Fig. 11). Evidence for p38 as a transducer of ER stress has been documented previously, because this MAPK was shown to phosphorylate multiple ER stressinduced transcription factors (e.g. activating transcription factor 2, activating transcription factor 6, and growth arrest and DNA damage-inducible gene 153), leading to increases in transcriptional activity (45)(46)(47). In addition, we provide evidence here that ER stress-responsive eIF2␣ kinases are not only activated by PPAR␥ ligands, but that PKR contributes to PPAR␥ agonist-dependent MKK3/6 and p38 activation.
Although PKR has been implicated as an upstream activator of p38 in the innate immune response (37), this is the first evidence that PKR can influence ER stress-induced p38 phosphorylation. Although the mechanism by which PKR activates p38 remains unclear, it was recently shown that PKR physically interacts with MKK6 in response to double-stranded RNA, forming a catalytic complex and thus facilitating p38 phosphorylation (48). These previous studies along with the current data using 2-aminopurine suggest that PKR acts as an important upstream mediator of TZD-dependent p38 activation. However, our findings do not discount the potential role of additional, unknown MKK kinases that could also contribute to p38 phosphorylation in this model. Specifically, 2-aminopurine completely blunts PKR activation in response to PPAR␥ ligands, but modest MKK3/6 and p38 phosphorylation is still observed. Furthermore, the time course for PKR activation in A549 cells closely parallels rather than precedes that of MKK3/6 and p38. Although this could be due to a lack of sensitivity in the immunoblot assay, additional studies in the current model are necessary to clarify a direct relationship between PKR and p38 as previously reported in different cell types. Importantly, these data identify the ER as a potential key mediator in TZD-induced signaling in GN4 cells (Fig. 11); however, future studies are needed to determine a causal role for the ER in this mechanism.
Interestingly, the data presented here support a role for both CaMKII and PKR as regulators of the p38 MAPK and eIF2␣ pathways in response to PPAR␥ ligands. Although doublestranded RNA is the classic PKR stimulus, the activity of this kinase is also increased in double-stranded RNA-independent conditions by a novel protein activator PKR-activating protein (also known as RAX) (49,50). In response to cellular stresses such as thapsigargin, PACT is phosphorylated on serine residues and associates with PKR leading to increases in kinase activity and eIF2␣ phosphorylation (50,51). Although serine phosphorylation of PACT occurs at putative CaMKII consensus sites, the kinase responsible for PACT phosphorylation has not been identified. Based on our current findings, we propose that CaMKII via PACT activates PKR in response to PPAR␥ ligands; PKR then facilitates downstream MKK3/6 and p38 as well as eIF2␣ phosphorylation (Fig. 11).
Consistent with their inability to potently activate CaMKII, Pyk2, and p38, ⌬2-TZDs also failed to induce PERK phosphorylation to the same extent as their parent compounds. Thus, in the pathways we examined in GN4 cells, ⌬2-ciglitazone and ⌬2-troglitazone differ from ciglitazone and troglitazone only in their ability to activate the "calcium-regulated" kinases. Al-though we have shown that PPAR␥ ligand-induced Pyk2 and p38 phosphorylation in GN4 cells is calcium-dependent (11), activation of PERK, PKR, and CaMKII in response to intracellular calcium release has been documented extensively in the literature. On the other hand, BAPTA-AM, Ad.CRNK, KN-93, and 2-aminopurine failed to significantly inhibit PPAR␥ liganddependent EGFR and Erk phosphorylation (data not shown). This observation suggests that activation of this latter signaling pathway is calcium-independent and may therefore explain why ⌬2-ciglitazone, which does not potently activate p38, induces EGFR and Erk phosphorylation to the same extent as ciglitazone. We therefore hypothesize that the failure of the ⌬2-derivatives to potently induce phosphorylation of certain kinases in GN4 cells is related to their inability to promote ER calcium release. Consistent with this hypothesis, ⌬2-ciglitazone and ⌬2-troglitazone were unable to induce eIF2␣ phosphorylation to the same extent as the parent compounds.
In summary, we provide evidence that the PPAR␥ ligands ciglitazone and troglitazone activate CaMKII and that this calcium-activated kinase, as opposed to Pyk2, is essential for downstream MKK3/6 and p38 phosphorylation. Activation of MAPKs is correlated with phosphorylation of the ER stresssensitive kinase PERK and subsequent eIF2␣ phosphorylation. We identify CaMKII as a potential novel regulator of PKR in response to ER stress and demonstrate that PKR plays a necessary role in activation of the p38 pathway in response to ER calcium depletion. The ability of ciglitazone and troglitazone to induce ER stress as well as activation of MAPKs may play an important role in their effects on cell growth and maturation. Consistent with earlier studies, we found here that both ciglitazone and troglitazone significantly reduced GN4 cell viability (data not shown). In contrast, rosiglitazone and pioglitazone not only failed to induce MAPKs and eIF2␣ phosphorylation in this cell model, but also were unable to significantly affect GN4 cell viability suggesting that these kinases play a necessary role in PPAR␥ ligand-induced cell inhibition. Collectively, these observations suggest that TZDs induce PPAR␥-independent signaling events that have potential relevance to the mechanism responsible for their antitumor activity in a variety of cancers.