p38 and Activating Transcription Factor-2 Involvement in Osteoblast Osmotic Response to Elevated Extracellular Glucose*

Poorly controlled or untreated type I diabetes mellitus is characterized by hyperglycemia and is associated with decreased bone mass and osteoporosis. We have demonstrated that osteoblasts are sensitive to hyperglycemia-associated osmotic stress and respond to elevated extracellular glucose or mannitol by increasing c-jun and collagen I expression. To determine whether MAPKs are involved in this response, MC3T3-E1 osteoblasts were treated with 16.5 mm glucose, mannitol, or contrast dye for 1 h. Immunoblotting of phosphorylated p38 demonstrated activation of p38 MAPK by hyperosmotic stress in vitro andin vivo. Activation peaked at 20 min, remained detectable after 24 h, and was protein kinase C-independent. Activating transcription factor-2 (ATF-2) activation followed the same pattern as phospho-p38. Transactivation of cAMP response element (CRE)- and c-jun promoter (containing a CRE-like element)-reporter constructs increased following hyperosmotic treatment. SB 203580 (a p38 MAPK inhibitor) blocked ATF-2 phosphorylation, CRE transactivation, and c-jun promoter activation. Hyperosmotic activation of collagen I promoter activity was also inhibited by SB 203580, consistent with the involvement of c-jun in collagen I up-regulation. Therefore, we propose that hyperglycemia-induced increases in p38 MAPK activity and ATF-2 phosphorylation contribute to CRE activation and modulation of c-jun and collagen I expression in osteoblasts.

Diabetes is diagnosed by an elevated blood glucose level that is Ͼ200 mg/dl (11 mM) at any random testing (1). Cellular responses to elevated extracellular glucose are thought to contribute to the development of diabetic complications. Hyperglycemic responses can differ greatly between cell types and can involve osmotic stress pathways. Immediate effects of a hyperosmotic response include cell shrinkage, followed by modulation of ion transporter activity (2,3), leading to a cell regulatory volume increase (4). Subsequently, in mammalian cells, genes encoding proteins involved in the accumulation of intracellular "compatible osmolytes" are induced within hours to days after exposure to hyperglycemia (5)(6)(7). Mechanisms accounting for these responses include activation of intracellular signaling pathways (8), nonenzymatic glycosylation of DNA and proteins (9), and altered cellular metabolism (10). These responses are thought to ultimately contribute to tissue pathology secondary to diabetes.
In addition to well known complications such as neuropathy, nephropathy, and retinopathy, insulin-dependent diabetes mellitus (type I) is associated with decreased bone mass and increased risk of osteoporosis (11)(12)(13)(14)(15)(16)(17)(18)(19). Bone histomorphometry demonstrates a dysfunction in osteoblasts, the cells that form bone. Osteoblasts undergo three stages of development/differentiation: proliferation, extracellular matrix deposition, and extracellular matrix maturation to form bone (20,21). In diabetic bone, the number of fully mature osteoblasts is decreased (19,22). This suggests that hyperglycemia can suppress osteoblast differentiation. When osteoblasts are exposed to elevated extracellular glucose or mannitol in vitro, a similar effect occurs and is marked by decreased extracellular mineralization (23)(24)(25). Hyperglycemia-induced changes in osteoblast phenotype include up-regulation of collagen I and c-jun mRNA expression (associated with early stages of osteoblast maturation) and down-regulation of osteocalcin expression (a marker of differentiated osteoblasts) (23). Promoters of c-jun, collagen I, and osteocalcin contain AP-1 transcription factor-binding sites. Therefore, it is not surprising that expression of a dominantnegative AP-1 fusion protein in osteoblasts suppresses glucose and mannitol effects on gene expression. 1 This finding suggests that AP-1 member levels and activity are involved in osteoblast responsiveness to hyperglycemia and osmotic stress.
The mechanisms accounting for osmotic stress-induced c-jun expression and AP-1 activity in osteoblasts are not completely understood. Inhibitor studies have demonstrated that protein kinase C is involved in these osmotic stress-induced effects (23). In other cell types, MAPKs 2 have also been demonstrated to be involved in mediating cellular adaptive responses to osmotic stress (28 -34). Activation of MAPKs is known to modulate gene expression and, in particular, to activate and increase expression of c-Jun and AP-1 activities. This can occur through activation of JNK and p38, which phosphorylate c-Jun and ATF-2, respectively, and thereby increase transactivation at AP-1 and CRE-like (jun2TRE) (26) sequences in the c-jun promoter. The MAPK signaling pathways have also been implicated in the regulation of osteoblast phenotype and gene expression (27). Here, we demonstrate that hyperosmotic stress caused by a physiologically relevant increase in extracellular glucose results in the specific activation of p38 MAPK and its downstream transcription factor, ATF-2. This activation is associated with an increase in CRE and c-jun promoter transactivation. SB 203580, a p38 MAPK inhibitor, inhibits ATF-2 phosphorylation, CRE and c-jun promoter transactivation, and collagen I transactivation. These results demonstrate that p38 MAPK is involved in osmotic stress-induced osteoblast phenotypic changes.

MATERIALS AND METHODS
Cell Culture System-MC3T3-E1 cells (35), subcloned for maximal alkaline phosphatase staining and mineralization, were plated at 100,000 cells/100-mm dish and fed every 1-2 days with ␣-minimal essential medium (Invitrogen) containing 5.5 mM glucose (normal level) and supplemented with 10% fetal calf serum. Eleven days after plating and 24 h after the last feeding, glucose or mannitol (0.5 M stock solution) or diatrizoate meglumine (a contrast dye and osmotic agent that is not a sugar; Nycomed Inc., Princeton, NJ) was added directly to the medium in the tissue culture dish to yield the final concentrations of sugar noted for each experiment (ranging up to 22 mM or 22 mosM). For kinase inhibition studies, 30 min prior to addition of sugar, the cells were pretreated with an inhibitor of p38 activity (10 M SB 203580) (36,37), an inhibitor of protein kinase C activity (50 nM staurosporine, 100 nM Go 6976, or 200 nM Ro 31-8220) (23), or an inhibitor of protein kinase B (Akt) activity (Calbiochem) (38).
Whole Cell Protein Extraction-For whole cell protein isolation, MC3T3-E1 cells were washed with chilled 1ϫ phosphate-buffered saline, left on ice for 3 min, centrifuged at 800 ϫ g for 5 min at 4°C, and resuspended in lysis buffer (50 mM Tris (pH 7.4), 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, and 10% glycerol). A mixture of protease and phosphatase inhibitors (1 mM sodium orthovanadate, 2 mM phenylmethylsulfonyl fluoride, 5 g/ml aprotinin, 1 mM EGTA, 10 mM NaF, 1 mM sodium pyrophosphate, and 0.1 mM ␤-glycerophosphate) was added to lysis buffer. Samples were centrifuged at 14,000 rpm for 30 min at 4°C, and the supernatant protein concentration was quantitated by the Bio-Rad DC protein detection system.
Western Blot Analysis-Whole cell extracts were loaded (50 g/lane) on a mini-PAGE system. Following electrophoreses, proteins were transferred to polyvinylidene difluoride membranes (Bio-Rad) using a semidry transfer system. Protein transfer and size determinations were verified using prestained protein markers. Membranes were then blocked with 5% nonfat dry milk in TTBS (10 mM Tris-HCl, pH 8, 150 mM NaCl, 0.05% Tween-20) (39) and subsequently incubated with antibodies directed against different MAPK proteins as well as their active phosphorylated forms (Santa Cruz Biotechnology). Signals were detected using a horseradish peroxidase-conjugated secondary antibody and an enhanced chemiluminescence detection kit (ECL, Amersham Biosciences).
In Vivo Analysis-BALB/c mice (7-day-old littermates) were injected with 50 l of phosphate-buffered saline alone (vehicle) or containing 22 mM mannitol. Briefly, the needle was inserted at the base of the calvaria until it reached the central region of the calvaria, and the solution was injected subcutaneously over parietal bones. This volume of fluid was enough to visibly cover the entire calvaria. This method has been previously used successfully to examine osteoblast response in vivo to hormones such as parathyroid hormone (40), fibroblast growth factor (41), and transforming growth factor-␤ (42). Calvariae were dissected and flash-frozen in liquid nitrogen 1 h after injection. Samples were subsequently processed for whole cell protein and Western blot analyses as described above.
Measurement of Promoter Transactivation by Transient Transfection-Osteoblasts (MC3T3-E1 cells) were plated at a concentration of 100,000 cells/well of a six-well dish. Twenty-four hours later, cells were transfected with luciferase or chloramphenicol transferase reporter plasmids driven by the c-jun promoter (Ϫ1600 to ϩ170 bp; generously provide by Dr. P. Angel) (43), the collagen Ia1 promoter (Ϫ2296 to ϩ115 bp; generously provided by Dr. Lichtler) (26), three copies of the CREbinding sequence fused to a TATA-like promoter (p TAL ) region from the herpes simplex virus thymidine kinase promoter (CLONTECH), the same reporter vector without an inserted promoter, or an SV40 promoter-␤-galactosidase reporter construct (as a control). Five hours after transfection with a DNA/LipofectAMINE mixture (Invitrogen), 1 ml of 20% fetal bovine serum-supplemented ␣-minimal essential medium was added to the cells. Fifteen hours later, the medium was aspirated and replaced with fresh 10% fetal bovine serum-supplemented ␣-minimal essential medium. Twenty-four hours later, cells were subjected for 3 or 12 h to the indicated treatment and then harvested and lysed. Reporter activity was read immediately using a luciferase (Promega) or ␤-galactosidase (CLONTECH) assay system and a luminometer. Chloramphenicol transferase activity was measured as previously described (26).
Statistical Analysis-All statistical analyses were performed using the Microsoft Excel data analysis program for t test analysis or the SPSS statistical analysis program for analysis of variance with Bonferroni's test. Experiments were repeated at least three times unless otherwise stated. Values are expressed as means Ϯ S.E. except where indicated.

RESULTS
We previously demonstrated (23) that elevation of extracellular glucose or mannitol alters osteoblast gene expression, including c-jun and collagen I. Given the important role of MAPK in the regulation of transcription factor activities and gene expression, we examined the effect of treating osteoblasts with 16.5 mM extracellular sugar (to yield a final concentration of 22 mM sugar, as seen in uncontrolled diabetes) on activation of MAPK. Figs. 1 and 2 demonstrate that total levels of p38, ERK, and JNK were unaltered after 1 h of sugar treatment. However, examination of active phosphorylated MAPK forms identified a 3-4-fold increase in phosphorylated p38 in response to elevated extracellular glucose or mannitol (Fig. 1, A and B), suggesting that osmotic stress stimulates the activity of this kinase. Because both glucose and mannitol are sugars, we tested another hyperosmotic compound that is not a sugar, contrast dye (diatrizoate meglumine). Fig. 1C demonstrates that p38 was also activated by this additional osmotic agent. Activation of the p38 pathway was specific because increases in phosphorylation of JNK or ERK were not evident (Fig. 2). Under these conditions, positive controls (addition of 150 nM 12-O-tetradecanoylphorbol-13-acetate or 300 mM glucose) demonstrated that these kinases could be activated in osteoblasts.
Our previous study demonstrated that protein kinase C is involved in osteoblast responses to osmotic stress (23). Given that some protein kinase C isoforms are able to activate p38, we examined the role of protein kinase C in the osmotic activation of p38. As shown in Fig. 3, preincubation of osteoblasts with an inhibitor of protein kinase C (200 nM Ro 31-8220) did not block activation of p38. Similarly, preincubation with other protein kinase C inhibitors (50 M staurosporine and 100 nM Go 6976) did not block osmotic activation of p38 (data not shown). These findings demonstrate that protein kinase C activity is not required for osmotic activation of p38. However, treatment with these inhibitors is sufficient to suppress c-jun and collagen I mRNA induction by osmotic stress (23), suggesting that both pathways are activated in parallel and that both could be important for osteoblast responses to osmotic stress. Activated p38 is known to phosphorylate and activate ATF-2 (44). To determine whether elevation of extracellular glucose or mannitol enhances ATF-2 phosphorylation, Western blot analyses were performed using specific antibodies directed against ATF-2 and phosphorylated ATF-2. Fig. 4 demonstrates that although levels of ATF-2 remained constant, phosphorylation of ATF-2 (similar to p38) was clearly elevated 3-4-fold 1 h after hyperglycemic treatment.
To examine the characteristics of p38 and ATF-2 phosphorylation in response to osmotic stress, a complete time course was performed with osteoblasts being harvested at 0, 10, 15, 20, 30, and 60 min. Fig. 5A shows that phosphorylation of p38 was evident at 10 min, plateaued by 20 min, and remained activated 60 min after glucose treatment. Phosphorylation of ATF-2 was evident within 20 min after glucose treatment; this was slower than phosphorylation of p38 and is consistent with ATF-2 as a downstream target of p38. The response after 24 h was also examined. Although p38 is considered to be a fast and transient-activated kinase, p38 stayed activated even 24 h after the onset of osmotic stress (Fig. 5B). As expected, ATF-2 also remained activated at 24 h.
To determine whether the effect on p38 and ATF-2 phosphorylation was dose-dependent, osteoblasts were treated with 2, 4.5, 9.5, and 16.5 mM glucose (final concentrations equal to 7.5, 10, 15, and 22 mM glucose, respectively), and phosphorylated p38 and ATF-2 levels were examined 1 h after treatment by Western analysis. Remarkably, Fig. 6 demonstrates that addition of as little as 4.5 mM glucose was enough to increase p38 phosphorylation in osteoblasts. This level of glucose (10 mM final concentration, 180 mg/dl) is often seen in patients who are diagnosed with diabetes, and we have shown previously that c-jun expression is increased at this glucose concentration (23).
To examine whether p38 could also be activated by hyperosmotic stress in vivo, mice were injected subcutaneously over the calvariae with vehicle or mannitol. This method has previ-ously been used by others to examine osteoblast responsiveness to hormones (40 -42). Fig. 7 demonstrates that phosphorylation of p38 was increased in mannitol-injected calvariae compared with controls. Correspondingly, ATF-2 phosphorylation was also increased (Fig. 7). Thus, our in vitro response was reproduced in vivo.
To determine whether osmotic stress due to elevation of extracellular sugar can functionally influence transactivation of CRE, an enhancer element known to bind ATF-2, MC3T3-E1 cells were transfected with CRE-luciferase or cytomegalovirusluciferase reporter constructs and treated with glucose or mannitol. Fig. 8B demonstrates that increasing extracellular glucose or mannitol enhanced CRE transactivation Ͼ5-fold compared with the control cells. Cytomegalovirus-luciferase did not show any increase, demonstrating the specificity of the CRE induction (data not shown). To study the role of p38 in transactivation of CRE, we treated osteoblasts with 10 M SB 203580 30 min prior to glucose treatment and examined responsiveness. Fig. 8B demonstrates that inhibition of p38 was A, immunoblots were developed using specific antibodies directed against p38 or phosphorylated p38. The autoradiograph is representative of three experiments and contains extracts from osteoblasts treated with 5.5 mM glucose (control (C)), 22 mM glucose (G), and 5.5 mM glucose ϩ 16.5 mM mannitol (M). B, values were obtained from three separate experiments and represent mean -fold increase Ϯ S.E. in p38 phosphorylation relative to total p38 levels. Control values were set at 1. One way analysis of variance using Bonferroni's multiple comparison test revealed that the treated cellular response was significant compared with the untreated cellular response. *, p Յ 0.03. No significant differences were found between treated conditions. C, levels of p38 and phosphorylated p38 were examined by Western blot analysis using day 11 whole cell protein extracts from osteoblasts treated for 1 h with 5.5 mM glucose (control (C)) and 5.5 mM glucose ϩ 16.5 mM contrast dye (D). The autoradiograph is representative of three separate experiments.

FIG. 3. Activation of p38 is protein kinase C-independent.
Osteoblasts cultured for 11 days were pretreated for 30 min with 200 nM Ro 31-8220 (Ro). Cells (control (C)) were then treated with 16.5 mM glucose (G) or mannitol (M). After 1 h, cells were harvested for whole cell protein extracts. Extracts (50 g/lane) were separated by 10% SDS-PAGE. Immunoblots were developed using specific antibodies directed against p38 or phosphorylated p38. The autoradiograph is representative of three experiments. sufficient to abolish the CRE transactivation in response to elevated extracellular glucose. Inhibition of protein kinase B (Akt), a potential target of SB 203580, did not affect this response (data not shown). Because we have previously identified c-jun and collagen I as genes responsive to hyperglycemia, we also examined their promoter activities that contain AP-1 and CRE enhancer elements. As expected, glucose or mannitol treatment increased expression of both c-jun and collagen I reporters (Fig. 9). Furthermore, inhibition of p38 suppressed c-jun and collagen I promoter transactivation, but did not affect control cytomegalovirus-luciferase expression, demonstrating a critical role for p38 in this response. DISCUSSION We have shown previously that osteoblasts respond to an elevation in extracellular glucose through an osmotic response pathway (23). The response to hyperglycemia involves induction of c-jun and collagen I expression and suppression of osteocalcin, consistent with suppression of a differentiated osteoblast phenotype. Our findings that mannitol treatment completely mimics this effect (23) and that glucose uptake is unaffected during this time 1 suggest that the response is osmotic in nature. We have also demonstrated that this response involves protein kinase C activation (23). Here, we show that a MAPK pathway is also required for osteoblast responsiveness to hyperglycemia. Specifically, phosphorylation of p38 MAPK was increased 1 h after addition of 16.5 mM glucose or mannitol in vitro and in vivo. Activation of p38 following treatment of osteoblasts with contrast dye, a non-sugar osmotic agent, further demonstrated that this response is not sugar-related and results from increased hyperosmotic stress. This level of hyper-  Osteoblasts cultured for 11 days were treated with different levels of glucose to yield final concentrations of 5.5 (control (C)), 7.5, 10, 15, and 22 mM. Cells were harvested 1 h after addition of sugar for whole cell protein extraction. Extracts (50 g/lane) were separated by 10% SDS-PAGE. Immunoblots were developed using antibodies directed against ATF-2, p38, phosphorylated p38, or phosphorylated ATF-2.

FIG. 7. Hypertonicity stimulates p38 activation in vivo.
Mannitol (M; 22 mM) or phosphate-buffered saline (control (C)) was injected subcutaneously over 7-day-old mouse calvariae. After 1 h, whole cell protein extracts were made. Extracts (50 g/lane) were separated by 10% SDS-PAGE. Immunoblots were developed using antibodies directed against p38, phosphorylated p38, or phosphorylated ATF-2. A representative autoradiograph is shown of four separate experiments. osmolarity failed to activate the other MAPKs (ERK and JNK) in MC3T3-E1 cells. Osmotic activation of the p38 kinase pathway in osteoblasts is consistent with the role of this pathway in cellular responses to stress (45)(46)(47). It is clear that p38 is activated in response to hypertonicity (47); however, most studies have focused on the influence of high (500 -600 mM) osmolar stress (36,48,49). More recently, researchers have begun to examine the cellular influence of moderate osmolarity increases (22-50 mM). It is clear that cells can respond to such changes. For example, Park et al. (50) found that 30 mM glucose or mannitol stimulates ICAM-1 (intracellular adhesion molecule-1) expression in rat mesangial cells. Similar to our results, Duzgun et al. (51) have demonstrates that a 50 mM increase in glucose or mannitol is sufficient to activate p38 kinase in endothelial cells.
Hyperosmotic activation of p38 kinase promotes early cellular protective responses to extracellular hypertonicity through stimulating a cell regulatory volume increase (28). This function is consistent with the function of Hog1 (high osmolarity glycerol response-1), the yeast homolog of p38, which is critical for yeast adaptation to osmotic stress (52,53). These adaptive responses include modulation of membrane transporters (short-term) and/or metabolic pathways (long-term) that draw water back into the cell and partially restore cell volume and intracellular solute concentration (4, 54, 55). Inhibition of p38 activity has been demonstrated to block some of these re-sponses, including osmotic induction of osmolyte transporters in renal epithelial cells (29).
Osmotic stress activation of p38 signaling is suggested to involve cell shrinkage. This is supported by studies in rat kidney cells that exhibit activation of p38 in response to elevated extracellular sucrose, but not urea (28), which readily crosses the cell membrane and does not cause cell shrinkage. Maintaining intracellular osmolality during cell shrinkage does not inhibit activation of p38. This further suggests that cell shrinkage in itself, rather than intracellular hypertonicity, is required for activation of p38 kinase (28). Cell shrinkageinduced actin cytoskeleton remodeling is thought to be involved in the cellular response to osmotic stress as demonstrated by the suppression of neutrophil responses to osmotic stress through the use of actin remodeling inhibitors (56). The role of actin remodeling is further supported by the finding that Rac and Cdc42, regulators of the actin cytoskeleton and upstream activators of the p38 kinase pathway, are activated by osmotic stress in neutrophils (57). It is also known that some protein kinase C isoforms can activate p38. We have previously shown that protein kinase C is involved in osteoblast responsiveness to hyperosmotic stress (23). However, based on the data presented in Fig. 3, osmotic activation of p38 is protein kinase C-independent.
Once activated, MAPKs phosphorylate transcription factors, leading to increased activity and subsequent gene modulation in response to extracellular stimuli, including osmotic stress (32)(33)(34). We have shown that AP-1 activity, which can be modulated by MAPKs, is a crucial component in osteoblast responsiveness to osmotic stress and involves increased expression and activity of c-Jun (25). Here, we show that ATF-2 phosphorylation, mediated by p38 (based on our inhibitor studies), also increased dramatically after a 1-h treatment with 16.5 mM glucose or mannitol. Phosphorylation of ATF-2 has been demonstrated to increase its transcriptional activity (58 -60) and is known to increase in response to cellular stresses such as short-wavelength UV (59,61) and reperfusion after ischemia (26). ATF-2 target genes include important bone regulatory genes such as tumor necrosis factor-␣ (60), transforming growth factor-␤ (62), cyclin A (63), and c-jun (64). These genes are also known to play important roles in the stress response and cell growth and differentiation.
Because activation of p38 leads to ATF-2 phosphorylation (65), we expected the characteristics of ATF-2 phosphorylation to be similar to those of phosphorylated p38. Our results demonstrate that p38 was phosphorylated within 10 min after treatment with 16.5 mM glucose, but there was a lag period of 10 min before ATF-2 was phosphorylated. This is consistent with ATF-2 being a downstream target of p38. But what we found to be more interesting was the persistence of phosphorylation of both p38 and ATF-2 after 24 h. It is known that the ubiquitination of ATF-2 as well as c-Jun is phosphorylationdependent and that the early increase in ATF-2 phosphorylation should increase the degradation of this protein (66). However, it is possible that p38 stress-activated protein kinasemediated phosphorylation may stabilize ATF-2, as has been reported in the case of c-Jun (67,68). Our finding that phosphorylation of both p38 and ATF-2 occurred when extracellular glucose was between 10 and 15 mM is consistent with the diagnosis of diabetes at glucose levels between 10 and 15 mM and suggests that the response occurs at a pathological threshold level of hyperglycemia/osmotic stress.
SB 203580, an inhibitor of p38 activity at low concentrations (10 M), suppressed hyperglycemia/osmotic stress-induced ATF-2 phosphorylation, suggesting its p38 dependence. Inhibition of another kinase, protein kinase B (suggested to be affected by SB 203580) (69), did not influence the response, further supporting a role for p38. We have also demonstrated that inhibition of p38 could suppress a 5-fold hyperglycemic induction of CRE and c-jun reporter activation in response to glucose or mannitol treatment, consistent with increased ATF-2 phosphorylation and its ability to transactivate the expression of these genes (58 -60). In fibroblasts exposed to stress, CRE transactivation is also increased (59,64). Consistent with our results, Nadkarni et al. (36) demonstrated that hypertonic induction of aldose reductase mRNA and transactivation at an osmotic response element are inhibited by SB 203580 in hepatic cells, suggesting that p38 MAPK mediates activation of the transcription factors necessary for osmotic response element activation.
The combination of our results can be used to hypothesize a novel mechanism employed by osteoblasts in response to physiologically relevant hypertonicity. We have previously shown that hyperosmotic stress induces c-jun mRNA and protein levels (23). The major regulators of c-jun expression are ATF-2 and c-Jun (59). The promoter for c-jun contains a proximal AP-1 site (TGAGTCA) and a distal site called jun2TRE (TTAC-CTCA) (43). In vitro binding studies have revealed that the consensus AP-1 site is preferentially targeted by dimers composed of Jun and Fos members, whereas jun2TRE binds heterodimers composed of c-Jun and ATF proteins. Inhibition of either protein kinase C, as previously shown (23), or p38, as shown here, inhibits hyperglycemic induction of c-jun expression. This suggests that both signaling pathways are required. Although p38 has been shown to be important for activation of ATF-2, evidence suggests that protein kinase C is important for increasing the activity of c-Jun (70). Thus, these parallel pathways could converge at c-jun2TRE and cause hyperglycemic induction of c-jun expression. Our studies 1 incorporating a dominant-negative AP-1 construct (71) suggest that AP-1 activity is important for hyperglycemic up-regulation of collagen I expression, which is thought to involve AP-1 binding and transactivation (25,72,73). Therefore, suppression of c-jun expression by p38 inhibition should suppress collagen I expression, as we demonstrated in Fig. 9. Taken together, our findings demonstrate that activation of p38 is involved in a cascade of osteoblast responses to hyperglycemia and hyperosmotic stress, leading to phosphorylation of ATF-2 and increased c-jun and collagen I expression.