Phosphorylated Heat Shock Protein 27 Represses Growth of Hepatocellular Carcinoma via Inhibition of Extracellular Signal-regulated Kinase*

Heat shock protein 27, one of the low molecular weight stress proteins, is recognized as a molecular chaperone; however, other functions have not yet been well established. Phosphorylated heat shock protein 27 levels inversely correlate with the progression of human hepatocellular carcinoma. This study shows that phosphorylated heat shock protein 27 interferes with cell growth of the hepatocellular carcinoma-derived HuH7 cells in the presence of the proinflammatory cytokine, tumor necrosis factor-α, via inhibition of the sustained activation of the extracellular signal-regulated kinase signal pathway. The activities of Raf/extracellular signal-regulated kinase and subsequent activator protein-1 transactivation and the induction levels of cyclin D1 were lower in HuH7 cells transfected with phosphorylated heat shock protein 27 than those with unphosphorylated heat shock protein 27. Moreover, phosphorylated heat shock protein 27 up-regulated the levels of p38 mitogen-activated protein kinase and mitogen-activated protein kinase phosphatase-1, an inhibitory protein of extracellular signal-regulated kinase. These results indicate that phosphorylated heat shock protein 27 might suppress the extracellular signal-regulated kinase activity in the hepatocellular carcinoma cells via two separate pathways in an inflammatory state. The extracellular signal-regulated kinase activity is inversely correlated with phosphorylated heat shock protein 27 at serine 15 and also in human hepatocellular carcinoma tissues in vivo. Because the extracellular signal-regulated kinase signal pathway is a major proliferation signal of hepatocellular carcinoma, activator protein-1 activation is an early event in hepatocarcinogenesis. These findings strongly suggest that the control of the phosphorylated heat shock protein 27 levels could be a new therapeutic strategy especially to counter the recurrence of hepatocellular carcinoma.

activates MAPK that enhances phosphorylation of HSP27 (8). Phosphorylated HSP27 inhibits IB kinase (IKK) and reduces IB degradation, thus resulting in the suppression of the NF-B activation in HeLa cells (9). Accumulating evidence indicates that the phosphorylation of HSP27 holds the key to the TNF␣ related liver diseases.
Hepatocellular carcinoma (HCC) commonly arises in the liver with chronic inflammation and ranks fifth in frequency on a worldwide basis, thus causing more than 1 million deaths annually (10). The overall survival of patients with HCC even after resection is still unsatisfactory because of frequent recurrence. The recurrence rate at 5 years after the curative treatment may exceed 70% (10). This high recurrence rate is not because of local recurrence or metastasis from the original lesion but rather from second primary lesions (10). However, the most suitable prognostic factor that suggests patients with HCC are at high risk for early recurrence has not yet been identified. MAPKs are essential components of intracellular signal transduction and are activated by phosphorylation in response to various extracellular stimuli, including growth factors, cytokines, and environmental stress. Among the MAPK family, extracellular signal-regulated kinase (ERK) is a key molecule that transfers signals into the nuclei to induce proliferation and differentiation (11). In HCC, the ERK are activated, and they up-regulate cyclin D1 expression, which thus stimulates progression (12). Conversely, p38 MAPK negatively regulates cyclin D1 and antagonizes the c-Jun NH 2 -terminal kinase (JNK)-c-Jun pathway to suppress HCC development (13,14). Previous studies showed that the level of phosphorylated HSP27 in human HCC tissues inversely correlates with the tumor size and the TNM stage of HCC (15). In addition, a proapoptotic, tumor-suppressive molecule protein kinase C␦ regulates HSP27 phosphorylation at a point upstream of p38 MAPK in the human HCC-derived cell line, HuH7 cells (16). However, the exact role and regulatory mechanism of HSP27 in human HCC remain to be clarified.
This study aimed to clarify the role of phosphorylated HSP27 in HCC and to analyze the proliferation of the HCC cells transfected with unphosphorylatable or phospho-mimic mutants of human HSP27. The results showed that phosphorylated HSP27 repressed the HCC cell proliferation in the presence of proinflammatory cytokine, TNF␣, via inhibition of the Raf-ERK kinase (MEK)-ERK signaling pathway and the activation of p38 MAPK-MAPK phosphatase-1 (MKP-1) pathway.

EXPERIMENTAL PROCEDURES
Plasmids-Wild-type (WT) and mutant human HSP27s subcloned into pcDNA3.1 mammalian expression vector were kindly provided by Dr. C. Schäfer (Klinikum Grosshadern, Ludwig-Maximilians University Munich, Munich, Germany). For mutant HSP27 vectors, the cDNA of HSP27 had been mutated at serine residues 15, 78, and 82 to aspartate (3D) to imitate the phosphorylated HSP27 form or mutated at the same residues to alanine (3A) to prevent phosphorylation of HSP27 (17). A constitutively active MEK1 cDNA was the generous gift from Dr. N. G. Ahn (Howard Hughes Medical Institute, University of Colorado, Boulder) (18).
Cell Culture and Stable Transfections-Human HCC-derived HuH7 cells, which originated from well differentiated HCC tissues, were obtained from the Japanese Cancer Research Resources Bank. HuH7 cells were maintained in RPMI 1640 medium (Sigma) supplemented with 1% fetal calf serum. For stable transfections, 4 ϫ 10 5 HuH7 cells were cultured in 6-well dishes and then transfected with 2 g of the WT or mutant HSP27 plasmids that expresses geneticin (G418) resistance using 12 l of UniFECTOR transfection reagent (B-Bridge International, Mountain View, CA) in 1 ml of RPMI 1640 medium without fetal calf serum per well. One ml/well medium with 10% fetal calf serum was added 5 h after transfection. The cells were subcultured and grown in the presence of 1 mg/ml of G418 (EMD Chemicals, Inc., San Diego) 2 days later. After about 2 weeks, single G418-resistant colonies were obtained by serial dilution in 96-well dishes. The colonies then were maintained and analyzed individually for the expression of HSP27s.
Cell Growth Assay-Empty vector-transfected, WT, or mutant HSP27s stably expressing HuH7 cells were plated on 96-well dishes (1 ϫ 10 3 cells/well). Twenty four h after seeding, the cells were treated with or without 1 nM TNF␣ for the indicated time, and cell numbers were counted using the trypan blue dye exclusion method or using WST-1 reagent (Roche Diagnostics) according to the manufacturer's instructions. To investigate the influence of caspase 9 on the cell growth, WT, or the 3D HSP27, stably expressed HuH7 cells were treated with caspase-9 inhibitor I simultaneously with or without 1 nM TNF␣ for 6 days.
Western Blotting-The cultured cells, which overexpressed WT or mutant HSP27s, were stimulated with or without TNF␣ for the indicated time. The cells or the snap-frozen human HCC samples were lysed, homogenized, and sonicated in lysis buffer containing 62.5 mM Tris/HCl (pH 6.8), 2% SDS, 50 mM dithiothreitol, and 10% glycerol. A Western blot analysis was performed as described previously (16,19). Band intensities were visualized on x-ray film with the ECL Western blotting detection system (GE Healthcare). The protein band intensities were determined by integrating the optical density over the band area (band volume) using NIH image software. The samples from the cell cultures to be quantitatively compared by Western blots were run in the same gel. Values represent the amount of phospho-ERK or phospho-MEK divided by those of total ERK or total MEK, respectively. The values represent the amount of full-length and cleaved caspase 9, phospho-c-Raf, cyclin D1, phospho-p38 MAPK, and phospho-MKP-1 divided by those of GAPDH. To quantify the protein from the HCC tissue extracts, 0.25 l of MagicMark XP Western protein standard (Invitrogen), the marker protein, was run in every gel. Based on the intensity of the marker protein band on x-ray film, the proteins of the tissue samples were quantitatively compared. After being normalized by the intensity of the marker protein, values represent the amount of phospho-and total HSP27s or phospho-ERK divided by those of ␤-actin or total-ERK, respectively. The data of the normalized values of the protein bands were statistically analyzed as described under "Statistics." Luciferase Reporter Assay-A reporter plasmid, activator protein-1 (AP-1)-Luc was kindly provided by Dr. S. Kojima (RIKEN, Wako, Japan). The cells were stimulated with or without 1 nM TNF␣ for 48 h before transfection. At 5 h after transfection, another 24 h of stimulation of TNF␣ was performed. Transient transfection with the AP-1-Luc reporter (1 g/35-mm dish) and measurement of luciferase activity of cell lysates were performed using UniFECTOR transfection reagent and a dual luciferase reporter assay system (Promega Corp., Madison, WI) as described previously (20). The cells were cotransfected with pRL-CMV (Renilla luciferase; 100 ng/35-mm dish) as an internal standard to normalize transfection efficiency. To examine the involvement of MEK-ERK system in AP-1-mediated transactivation activity within the 3D HSP27 mutant overexpressed cells, active MEK1 was cotransfected with the reporter plasmid.
Tissue Specimens-HCC tissues were obtained by surgical resection from 44 patients infected with hepatitis viruses B (10 cases) or C (31 cases) and 3 patients with alcoholic cirrhosis at the Department of Surgery, Ogaki Municipal Hospital. No patient had previously undergone chemotherapy. The resected tissues were snap-frozen in liquid nitrogen and then stored at Ϫ80°C until used for the Western blot analysis. The resected HCC specimens were obtained according to protocol approved by the Committee for the Conduct of Human Research at Ogaki Municipal Hospital. Informed consent was obtained from all patients.
Statistics-Data are expressed as the means Ϯ S.D. Statistical significance of the data from the cell cultures was analyzed using one-way analysis of variance, followed by Dunnett's test, and the patient clinical data were analyzed using the Pearson correlation coefficient (r). All p values were derived from twotailed tests, and p Ͻ 0.05 was accepted as statistically significant. A Pearson correlation coefficient of ͉r͉ Ͼ 0.400 was accepted as a positive correlation.

RESULTS
Expression of HSP27 in Wild-type, Unphosphorylated Type, or Phospho-mimic Type HSP27-transfected HuH7 Cells-To investigate the effect of phosphorylated HSP27 on HCC cell growth, human HCC-derived HuH7 cells were stably transfected with cDNAs of mutant HSP27s with alanine 15, alanine 78, and alanine 82 (3A) that mimicked the unphosphorylated type or with aspartate 15, aspartate 78, and aspartate 82 (3D) that mimicked the phosphorylated type. For comparison purposes, HuH7 cells were also transfected with wild-type (WT) HSP27 cDNA or an empty pcDNA3.1 vector (empty). A Western blot analysis demonstrated that HSP27 was overexpressing in WT, 3A, or 3D HSP27 cDNA-transfected HuH7 cells (Fig. 1). The empty vector-transfected cells only expressed intact HSP27 proteins. Anti-phospho-Ser-15 HSP27 antibodies and anti-phospho-Ser-82 HSP27 antibodies reacted with the HSP27 protein that was overexpressed in both WT and 3D HSP27 vector-transfected cells (Fig. 1). The phosphorylated HSP27 protein level in WT HSP27 cDNA-transfected HuH7 cells was almost the same as that in 3D HSP27 cDNA-transfected cells. The antibodies for human-specific phospho-Ser-78 HSP27 also reacted with the HSP27 in WT or 3D HSP27 cDNA-transfected cells as the antibodies for other phosphorylated forms (data not shown). On the other hand, the overexpressed HSP27 protein in the 3A HSP27 cDNAtransfected cells did not react with the phospho-HSP27 antibodies ( Fig. 1).
Phosphorylated HSP27 Retarded the HCC Cell Growth in the Presence of TNF␣-To clarify the relationship between phosphorylation of HSP27 and HCC cell growth, we first studied whether the cell growth of phosphorylated HSP27-overexpressed HuH7 cells was suppressed compared with that of unphosphorylated HSP27-overexpressed cells. HCC commonly arises in the liver with chronic inflammation (10,21). In the liver, the levels of TNF␣, a proinflammatory stimuli, in patients with cirrhosis and HCC have been reported to be significantly higher than those in normal individuals (22). Therefore, the cell growth of phosphorylated HSP27-overexpressed HuH7 cells was examined both in the presence and in the absence of TNF␣. In the absence of TNF␣, all WT, 3A, or 3D HSP27-overexpressed cell lines and the empty vector-transfected cell line showed almost the same growth curve ( Fig. 2A, curves [1][2][3][4]. Even in the presence of 1 nM TNF␣, the empty vector or 3A HSP27 vector-transfected HuH7 cells also exhibited almost similar growth rate as in the absence of TNF␣ ( Fig.  2A, curves 5 and 7). However, the cell growth of WT and 3D HSP27-overexpressed HuH7 cells was remarkably delayed in comparison with that of 3A HSP27-overexpressed HuH7 cells in the presence of TNF␣ 6 days after incubation ( Fig. 2A, curves 6 and 8). Phosphorylated HSP27 therefore seems to inhibit the cell growth of HCC under inflammatory conditions. Nonphosphorylated HSP27 is an inhibitor for caspase-dependent apoptosis (1,5). It inhibits the interaction of cytochrome c, which is released from the permeabilized mitochondria, and pro-caspase 9. To study the relationship of apoptosis and the growth retardation of phosphorylated HSP27overexpressed HuH7 cells, the activities of caspase 9 were examined in WT, 3A, or 3D HSP27-transfected HuH7 cells in the presence of TNF␣. Regardless of the presence or the absence of TNF␣, the amount of the full-length caspase 9 increased in 3D HSP27 cDNA-transfected HuH7 cells (Fig. 2B, lanes 7 and 8). The cleaved and activated caspase 9 significantly increased in WT HSP27 cDNA-transfected HuH7 cells after 2 h of stimulation of TNF␣ (Fig. 2C, lane 4). However, the increased activation of caspase 9 was shown also in the 3D HSP27-overexpressed cells in the absence of TNF␣ (Fig. 2C, lane  7). The cleavage of caspase 9 in the 3D HSP27-overexpressed cells in the presence of TNF␣ was even weaker than in the absence of TNF␣ (Fig. 2C,  lanes 7 and 8). The similar tendency of the caspase 9 activities in the 2-h TNF␣-stimulated cells was also shown in the cells after 72 h of stimulation of TNF␣ (data not shown).
To confirm the cell growth retardation of WT and 3D HSP27-overexpressed HuH7 cells in the presence of TNF␣ is not related to caspase 9, we investigated whether caspase-9 inhibitor I, an irreversible inhibitor of caspase 9, restored the cell growth of WT and 3D HSP27 in the presence of TNF␣. As shown in Fig.  2D, caspase-9 inhibitor I, which alone had little effect on the cell number, did not affect the cell growth of both WT and 3D HSP27overexpressed HuH7 cells treated with 1 nM TNF␣ for 6 days (columns 4 and 8, in comparison with columns 3 and 7, respectively). Caspase-9 inhibitor I at a dose of 50 M was toxic (data not shown). Therefore, it seems unlikely that caspase-dependent apoptosis caused the growth retardation of the phosphorylated HSP27overexpressed cells in the presence of TNF␣.  JULY 4, 2008 • VOLUME 283 • NUMBER 27

JOURNAL OF BIOLOGICAL CHEMISTRY 18855
Phosphorylated HSP27 Inhibited Prolonged Activation of ERK Signal Transduction in the HCC Cells-ERK has been reported to act as a potent proliferative factor of HCC and be constitutively activated in the human HCC cells and tissues (12,23). Does the cell growth retardation of the phosphorylated HSP27-overexpressed cells correlate with the ERK activity? The basal levels of phosphorylated ERK were similar among all HSP27 cDNA-transfected cells (Fig. 3A). Although the ERK phosphorylation levels in all HSP27 cDNA-transfected cells were similarly increased after 2 h of stimulation with 1 nM TNF␣, phospho-ERK levels in both WT and the 3D HSP27overexpressed HuH7 cells significantly decreased in comparison with those in 3A HSP27 cDNA or empty vector-transfected HuH7 cells after 72 h of stimulation with TNF␣ (Fig. 3A). Total ERK proteins were expressed at almost the same levels among all HSP27 cDNA-transfected cells regardless of whether or not they were stimulated with TNF␣. The ERK activity is regulated by upstream kinases MEK and c-Raf. As shown in Fig. 3B, a significant decline of MEK activity in WT and the 3D HSP27overexpressed HCC cells was observed in comparison with that in 3A HSP27 cDNA or empty vector-transfected cells after 72 h of stimulation of TNF␣. Furthermore, significant attenuation of c-Raf activity was also shown in WT and the 3D HSP27overexpressed HCC cells following 48 h of TNF␣ stimulation (Fig. 3C). Therefore, phosphorylated HSP27 might act as a repressor for prolonged activation of ERK signaling pathway at a point upstream of c-Raf in the HCC cells.
Transactivation Activities of AP-1 and Cyclin D1 Expression Were Suppressed in the 3D HSP27-overexpressed HCC Cells-ERK contributes to the induction of AP-1 transcriptional activity, and AP-1 activates the cyclin D1 promoter to induce cell proliferation (13,24). Therefore, the effect of the phosphorylated HSP27 on AP-1 transactivation activity was assessed (Fig.  4A). After 72 h of stimulation with TNF␣, WT and the 3D HSP27overexpressed HuH7 cells expressed significantly less transactivation activity of AP-1 than 3A HSP27-introduced cells (Fig. 4A, columns 6 and 8, in comparison with column 7). A remarkable decrease of AP-1 transactivation activity was observed especially in 3D HSP27 cDNA-transfected cells (Fig. 4A, column 8). To confirm that the attenuation of this AP-1 transactivation activity occurred because of the ERK signaling pathway, constitutive active MEK1 cDNA was transfected into the 3D HSP27overexpressed HuH7 cells. The active MEK1 restored AP-1 transactivation activity of the 3D HSP27-overexpressed HuH7 cells to the similar level as the 3A HSP27-overexpressed or empty vector-transfected HuH7 cells (Fig. 4A, column 9 in comparison with columns 7 or 5). In the absence of TNF␣, no significant difference of the AP-1 transactivation activity among empty vector and all HSP27 cDNAs-transfected cells was shown (Fig. 4A, columns 1-4). Therefore, phosphorylated HSP27 presumably reduced AP-1mediated cell proliferation via ERK signaling pathway in the HCC tissues under inflammatory conditions. Next, cyclin D1 protein expression levels in empty vector and all HSP27 cDNA-transfected cells in the presence and absence of TNF␣ were examined. In WT and the 3D HSP27-transfected cells treated with TNF␣ for 72 h, cyclin D1 levels significantly decreased in comparison with those in the 3A HSP27-transfected cells (Fig. 4B, columns and lanes 6 and 8 in comparison  with column and lane 7). In the absence of TNF␣, no significant difference of cyclin D1 protein level among empty vector and all HSP27 cDNA-transfected cells was shown (Fig. 4B, columns  1-4). Therefore, phosphorylated HSP27 presumably reduced AP-1-mediated cell proliferation via ERK signaling pathway in the HCC tissues.
Increased Expression Level and Activation of p38 MAPK That Were Followed by the Induction of Active MKP1 Were Observed in the Phosphorylated HSP27-overexpressed HCC Cells-In eukaryotic cells, there are another two MAPKs, p38 MAPK and  1-4). Values represent the amount of phospho-ERK, phospho-MEK, or phospho-c-Raf divided by those of total ERK, total MEK, or GAPDH, respectively, and were plotted as fold induction in comparison with those in the empty vector-transfected cells (mean Ϯ S.D., n ϭ 3). **, p Ͻ 0.01 versus column 1; ϩ, p Ͻ 0.05; ϩϩ, p Ͻ 0.01.
JNK, in addition to ERK (11). Although p38 MAPK and JNK are less sensitive to growth signals than ERK, their activation is preferentially triggered by pro-inflammatory cytokines and environmental stresses (25). Therefore, p38 MAPK activities were next examined in the unphosphorylated and in the phosphorylated HSP27-overexpressed HCC cells. Because the maximum activity of ERK was observed after 2 h stimulation with TNF␣, the amount of p38 MAPK and phosphorylated p38 MAPK in the cells was also examined after 2 h of stimulation with TNF␣. Fig. 5A shows that the p38 MAPK level was increased in WT and especially in the 3D HSP27-overexpressed HuH7 cells in the presence of TNF␣ in comparison with that in the 3A HSP27 cDNA or empty vector-transfected cells. The increased levels of p38 MAPK were also observed even in the absence of TNF␣ in both WT and the 3D HSP27-overexpressed HuH7 cells (data not shown). Furthermore, TNF␣ stimulation significantly induced the p38 MAPK phosphorylation in both WT and the 3D HSP27 cDNA-transfected HuH7 cells in comparison with the empty vector and 3A HSP27 cDNA-transfected cells (Fig. 5A, lanes 2 and 4 in comparison  with lanes 1 or 3). The similar tendency of enhanced activation of p38 MAPK in WT and 3D HSP27-introduced HuH7 cells was also observed in 72-h TNF␣-stimulated cells, although the activity was less than after 2 h of stimulation (data not shown). HSP27 is phosphorylated by the p38 MAPK pathway (1). On the other hand, phosphorylated HSP27 enhanced p38 MAPK expression and activation in this experiment. This is probably the first report showing such a positive feedback from phosphorylated HSP27 to p38 MAPK. The activation of p38 MAPK is reported to induce MKP-1, a phosphatase that inactivates ERK (26). The effect of phosphorylated HSP27 was analyzed on an active form of MKP-1, phosphorylated MKP-1 (27). Phosphorylated MKP-1 was significantly induced in WT and the 3D HSP27-overexpressed cells after 72 h stimulation of TNF␣ in comparison with that in the 3A HSP27 cDNA-transfected cells (Fig. 5B, lanes 2 and 4 in comparison with lane 3). Phosphoryl-

Phosphorylated HSP27 Represses ERK-dependent HCC Cell Growth
ated MKP-1 was also induced in 3A HSP27 as well as WT and 3D HSP27 in HCC cells, although the induction was weaker (Fig. 5B, lane 3 in comparison with lane 1). The phosphorylated MKP-1 expression slightly increased in the WT and the 3D HSP27-transfected cells also in the absence of TNF␣ (data not shown). Not only the presence of the phosphorylated HSP27 but also the total amount of HSP27 might play some part in the induction of active MKP-1. The activity of JNK, the other MAPK, was not significantly changed in these experiments (data not shown).

The Levels of Phosphorylated HSP27 (Ser-15) Inversely Correlated with the Levels of ERK Activation in Human HCC
Tissues-The ERK activities were down-regulated by phosphorylated HSP27 in the human HCC cells as described above. Therefore, the in vivo correlation between the levels of ERK activity and the levels of phosphorylated HSP27 was investigated in HCC tissues. The levels of phosphorylated HSP27 (Ser-15) were significantly inversely correlated with the levels of activated ERK (r ϭ Ϫ0.468, p Ͻ 0.001; Fig. 6A). On the contrary, the levels of phosphorylated HSP27 (Ser-78), phosphorylated HSP27 (Ser-82), and total HSP27 did not correlate with those of the activated ERK (Fig. 6, B-D, respectively).

DISCUSSION
The influence of post-translational modification, such as phosphorylation on the function of HSP27, is precisely unknown. In this study, phosphorylated HSP27 reduced the cell growth rate of the HuH7 cells in the presence of TNF␣ (Fig. 2A). HSP27 reportedly inhibits the caspases to protect the cells from apoptosis, and the phosphorylation status of HSP27 influences that function (1,5). A transcription factor, NF-B, has been implicated in suppression of apoptosis, cell survival, proliferation, viral replication, inflammation, tumorigenesis, and metastasis, and all members of the TNF superfamily are known to activate it (28). The ability of HSP27 to interact with IKK␤ has been reported to be enhanced via the TNF␣-induced activation of MAPKdependent phosphorylation of HSP27, thus leading to the enhanced inhibition of IKK activity, reduced IB degradation, and consequent suppression of NF-B activation in HeLa cells (9). The overexpression of phosphorylated HSP27 increased caspase 9 protein levels and activity (Fig. 2, B and C). Phosphorylated HSP27 may therefore increase apoptosis in the HCC cells. However, the caspase activities in the phosphorylated HSP27-overexpressed cells were not enhanced by TNF␣ (Fig. 2C), and the caspase inhibitor did not restore the cell growth retardation of WT and phosphorylated HSP27-overexpressed cells in the presence of TNF␣ (Fig. 2D). These results suggest that the influence of some mechanisms other than apoptosis might play important roles in the control of the HCC cell proliferation by phosphorylated HSP27 in the presence of an inflammatory cytokine like TNF␣. However, the HCC cell growth retardation by phosphorylated HSP27 may be caused, in part, by the suppression of NF-B.
In human HCC, it is generally accepted that the activation of the ERK signal pathway leads to a mitogenic effect (12,20). This study presented novel evidence that phosphorylated HSP27 inhibits the sustained activation of the c-Raf-MEK-ERK pathway in an inflammatory environment (Fig. 3). The phosphorylation of HSP27 significantly correlated with the activity of ERK in not only HCC cells in vitro but also the specimens in patients with HCC in vivo (Fig. 6). There have so far been few reports addressing the influence of HSP27 on the ERK activation, except for the study by Lee and co-workers (29,30) where the overexpression of HSP25, the same species as human HSP27 in the mouse, was shown to down-regulate ERK expression, while also inhibiting their activation in mouse fibroblast L929 cells by a reduction in reactive oxygen species. Contrary to our results, HSP25 overexpression attenuated the H 2 O 2 -ERK pathway-mediated apoptosis in their experiments. The role of the HSP27-ERK pathway might be different in the mouse fibroblasts as compared with the human HCC cells. The attenuation of phosphorylated HSP27, especially phosphorylated at Ser-15, is correlated with HCC progression (15). In addition, the expression levels of HSP20, one of the low molecular weight HSP family (31), decreases in parallel with HCC progression and a significant correlation we observed between the levels of HSP20 and phosphorylated HSP27 at Ser-15 but not at Ser-78 and Ser-82 in human HCC tissues (19). This study revealed a significant inverse correlation between the levels of phosphorylated HSP27 (Ser-15) and ERK activity (Fig. 6). Based on these findings, it is probable that the phosphorylation at Ser-15 of HSP27 is important for repressing HCC cell growth activity. It has been shown that phosphorylation at Ser-15, but not Ser-78 and Ser-82, of HSP27 results in the conformational changes of HSP27 and the alteration of the direct interaction of HSP27 with other HSPs (32). Therefore, the alteration of direct interaction of HSP27 with HSP20 and/or other factors by the phosphorylation at Ser-15 may affect the ERK signal transduction pathway in HCC, thus resulting in the suppression of proliferation. Activated ERK is generally known to translocate into the nuclei and induce/activate transcription factors such as AP-1, which in turn increase the transcriptional activity of genes relevant for cell cycle progression, such as cyclin D1. In this study, the suppression of ERK activity by phosphorylated HSP27 attenuated the AP-1 transactivation activity and the expression of cyclin D1 (Fig. 4). ERK activation in human HCC is known to play an important role in multistep hepatocarcinogenesis, especially in the progression of HCC, at least in part through cyclin D1 up-regulation primarily induced by MAPK/ERK via c-Fos (12,33). In addition, active MEK1 restored the AP-1 activity levels suppressed by phosphorylated HSP27 to the control levels. Therefore, the suppression of AP-1-cyclin D1 signal transduction via the inhibition of ERK activity might be the significant mechanism for the proliferative control of the HCC cell by phosphorylated HSP27.
Although HSP27 phosphorylation is generally known to be a reversible process catalyzed by the MAPKAP2, a downstream substrate of p38 MAPK (1), phosphorylated HSP27 surprisingly induced p38 MAPK (Fig. 5A). The activation of the p38 MAPK cascade is preferentially triggered by pro-inflammatory cytokines, such as TNF␣ and environmental stress. The ␣-isoform of protein kinase C, an essential molecule of malignant cancer cells, has been reported to activate p38 MAPK while also stimulating cell migration and invasion in poorly differentiated human HCC cell lines (34). Transforming growth factor ␤ mediated the activation of p38 MAPK, and its downstream HSP27 may increase the invasive potential and matrix metalloprotease (MMP)-2 expression in human prostate cancer cells (35). On the contrary, p38 MAPK and p38 MAPK kinase (MKK3) have been shown to significantly inhibit mitogen-induced cyclin D1 expression in the constitutively active Raf-1 and estrogen receptor fusion protein stably expressed CCL39 cells (13). The association of human HCC with nearby normal tissues has been shown to reduce p38 MAPK and MKK6 activities especially in larger tumors (36). Moreover, it has been reported recently that p38 MAPK␣ suppresses liver cancer development by antagonizing the JNK-c-Jun pathway (14). There are at least four isoforms of p38 MAPK that have been identified and characterized, and the activation of p38 MAPK is mediated by MKK3, -4, and -6 (37). The role of p38 MAPK and its isoforms in HCC cell growth have not yet been established. It has been shown that p38 MAPK induces MKP-1, a major neg-ative regulator for ERK (26). This study also showed that phosphorylated HSP27 activated p38 MAPK and subsequently induced phosphorylated MKP-1 (Fig. 5B). These findings suggest that the cross-talk among phosphorylated HSP27, p38 MAPK, and MKP-1 might also regulate ERK activity in addition to the down-regulation of c-Raf-MEK-ERK signal transduction by phosphorylated HSP27, thus resulting in the suppression of HCC cell proliferation. The potential mechanism of phosphorylated HSP27 in HCC shown here is summarized in Fig. 7.
A number of studies have shown that the redox state plays a role in the regulation of TNF␣ intracellular signaling and ERK activation (38,39). HSP27 has been reported to regulate the intracellular reactive oxygen species and/or glutathione level (30,40). However, reactive oxygen species inhibitors do not suppress TNF-induced AP-1 activation (40). No change in the total glutathione levels is observed with the decrease in ERK phosphorylation in response to oxidative stress in primary cultured rat hepatocytes (41). It was recently reported that HSP27 is required for interleukin-1-activated transforming growth factor-␤-activated kinase-1 (TAK1) and downstream signaling by p38 MAPK, JNK, and their activator kinases and IKK␤ (42). TAK1 is the most upstream common signaling protein of both interleukin-1 and TNF␣. However, the activation of ERK

Phosphorylated HSP27 Represses ERK-dependent HCC Cell Growth
induced by interleukin-1 appears to be independent of HSP27 (42). Further studies should therefore investigate where and how phosphorylated HSP27 acts at a point upstream of c-Raf. The activation of AP-1 is known as an early event in HCC (43). In this experiment, phosphorylated HSP27 overexpression repressed prolonged activation of ERK signal transduction and down-regulated AP-1 activity, thus resulting in the suppression of HCC cell growth. These results suggest that phosphorylated HSP27 has a stronger influence in an earlier stage of the liver carcinogenesis. The remarkably high incidence of secondary liver cancer is actually responsible for the poor prognosis of liver cancer (10). At least one-third of post-therapeutic recurrence is because of the de novo cancer through multicentric carcinogenesis. The prevention of the recurrence of HCC at early stage is therefore urgently needed to enhance long term survival. Recently, an oral multikinase inhibitor, sorafenib, was developed, and the clinical trials against human HCC are in progress (44). Sorafenib blocks tumor cell proliferation by targeting Raf/MEK/ERK signaling at the level of Raf kinase (45). The HCC cell growth retardation via inhibition of ERK pathway by phosphorylated HSP27 shown here could therefore be a novel promising therapeutic strategy to prevent the recurrence of HCC.