Phospholipase Cϵ Activates Nuclear Factor-κB Signaling by Causing Cytoplasmic Localization of Ribosomal S6 Kinase and Facilitating Its Phosphorylation of Inhibitor κB in Colon Epithelial Cells*

Phospholipase Cϵ (PLCϵ), an effector of Ras and Rap small GTPases, plays a crucial role in inflammation by augmenting proinflammatory cytokine expression. This proinflammatory function of PLCϵ is implicated in its facilitative role in tumor promotion and progression during skin and colorectal carcinogenesis, although their direct link remains to be established. Moreover, the molecular mechanism underlying these functions of PLCϵ remains unknown except that PKD works downstream of PLCϵ. Here we show by employing the colitis-induced colorectal carcinogenesis model, where ApcMin/+ mice are administered with dextran sulfate sodium, that PLCϵ knock-out alleviates the colitis and suppresses the following tumorigenesis concomitant with marked attenuation of proinflammatory cytokine expression. In human colon epithelial Caco2 cells, TNF-α induces sustained expression of proinflammatory molecules and sustained activation of nuclear factor-κB (NF-κB) and PKD, the late phases of which are suppressed by not only siRNA-mediated PLCϵ knockdown but also treatment with a lysophosphatidic acid (LPA) receptor antagonist. Also, LPA stimulation induces these events in an early time course, suggesting that LPA mediates TNF-α signaling in an autocrine manner. Moreover, PLCϵ knockdown results in inhibition of phosphorylation of IκB by ribosomal S6 kinase (RSK) but not by IκB kinases. Subcellular fractionation suggests that enhanced phosphorylation of a scaffolding protein, PEA15 (phosphoprotein enriched in astrocytes 15), downstream of the PLCϵ-PKD axis causes sustained cytoplasmic localization of phosphorylated RSK, thereby facilitating IκB phosphorylation in the cytoplasm. These results suggest the crucial role of the TNF-α-LPA-LPA receptor-PLCϵ-PKD-PEA15-RSK-IκB-NF-κB pathway in facilitating inflammation and inflammation-associated carcinogenesis in the colon.

which identified PLC⑀ (PLCE1) as a predisposing gene for gastric and esophageal carcinomas (15,16). However, the direct link between the proinflammatory and tumor-promoting functions of PLC⑀ awaits further verification by using an animal model of inflammation-induced carcinogenesis.
Recent studies indicated that PLC⑀ activation augments the NF-B-dependent expression of inflammation-associated genes, particularly those of proinflammatory cytokines in cultured cells (17,18). In addition, it was reported that PKD, a serine/threonine kinase also called atypical PKC, which had been known to mediate NF-B activation triggered by various stimuli, such as LPA (19,20), was activated downstream of PLC⑀ (18,21,22), although underlying molecular mechanisms remain poorly understood. NF-B binds as a dimer of p65 RelA and p50 to a consensus DNA sequence on the promoters and enhancers of the target genes and regulates their transcription. NF-B activation is tightly controlled by several factors. Under a resting condition, NF-B is associated with IB and exists as an inactive form in the cytoplasm. Upon stimulation by extracellular ligands, such as TNF-␣, IB is phosphorylated by IB kinases at its specific serine residues, including Ser-32, and targeted for ubiquitination and subsequent proteasome degradation, thereby allowing NF-B to undergo nuclear translocation and activate the target gene expression (23,24). There are several proteins functioning as IB kinases, such as IKK complex, consisting of IKK␣, IKK␤, and IKK␥ (25); p90 RSK (26,27); and CK2 (28).
RSK family proteins are serine/threonine kinases and activated through phosphorylation by ERK1/2 and PDK1 (phosphoinositide-dependent kinase 1) in the cytoplasm. Thr-573 of RSK, located in the activation loop of the C-terminal kinase domain, is phosphorylated by ERK, leading to autophosphorylation of Ser-380 in the linker region (29). Subsequently, RSK is fully activated through phosphorylation by PDK1. Activated RSK is located in both the cytoplasm and nuclei and phosphorylates its substrates, such as IB (26,27), in the cytoplasm and c-Fos (30) and CREB (31) in the nuclei. The subcellular localization of RSK is reported to be regulated by a scaffold protein, PEA15 (phosphoprotein enriched in astrocytes 15), which directly associates with RSK and holds it in the cytoplasm, based on the observation that overexpression of PEA15 prevented nuclear translocation of RSK (32). Moreover, overexpression of a phosphorylation-mimicking mutant, S104D, of PEA15 failed to elevate the CREB promoter activity, suggesting that the phosphorylation of PEA15 might regulate the RSK localization (33).
In this study, we further investigate the relationship between the proinflammatory and tumor-promoting functions of PLC⑀ by employing the colitis-induced colorectal carcinogenesis model. Further, we analyze the molecular mechanism by which PLC⑀ augments the NF-B-dependent expression of inflammation-associated genes, demonstrating that the PLC⑀-PKD axis, activated by LPA receptor engagement, augments NF-B activation by causing sustained cytoplasmic localization of RSK via phosphorylation of PEA15 and thereby facilitating its phosphorylation of IB in the cytoplasm.
Plasmids-The full-length cDNA of human PLC⑀ was purchased as Flex ORF clones (Promega) and cloned into pFLAG-CMV2 (Sigma) for expression as FLAG fusions. The lipasedead human PLC⑀ mutant, PLC⑀⌬X, was generated by PCR-mediated mutagenesis to delete amino acids 1391-1541. pCMV-HA-PEA15 was constructed by inserting the human PEA15 cDNA, obtained by RT-PCR using Caco2 cell mRNA as a template, into pCMV-HA (Clontech). pCMV-HA-PEA15S104A and pCMV-HA-PEA15S104D were derived from it by site-directed mutagenesis. The primers used are listed in Table 1.
Cell Culture-A human colon cancer epithelial cell line Caco2 was purchased from the RIKEN cell bank (RBRC-RCB0988) and maintained in 5% CO 2 at 37°C in modified Eagle's minimum essential medium supplemented with 20% fetal bovine serum, non-essential amino acids, and 100 g/ml penicillin-streptomycin. Cells were washed with PBS and serum-starved for 3 h before subjecting to various experiments. Cells were pretreated with inhibitors and antagonists before stimulation by ligands, such as TNF-␣ and LPA.
Immunoblotting and Immunoprecipitation-Cells were lysed in a lysis buffer (20 mM Tris-HCl, pH 7.4, 250 mM NaCl, 3 mM EDTA, 0.5% (v/v) Nonidet P-40, and protease inhibitor mixture). The lysates were centrifuged at 21,500 ϫ g for 10 min, and the supernatants were used for immunoblotting and immunoprecipitation. Immunoblotting was performed by the standard procedure followed by acquisition of immunoreactive signals on the blots using ImageQuant LAS4000mini (GE Healthcare). The intensities of the immunoreactive signals were quantitated using ImageQuant TL (GE Healthcare).
RT-PCR and qRT-PCR-Total cellular RNA isolation, cDNA synthesis, RT-PCR, and qRT-PCR were performed as described previously (37). The relative mRNA level of each transcript was determined by the ⌬⌬Ct method with ␤-actin or GAPDH as a reference gene. The primers used are listed in Table 1.
Animals-Apc Min/ϩ mice, purchased from the Jackson Laboratory (Bar Harbor, ME), were bred with PLC⑀ ⌬X/⌬X mice (14), which had been back-crossed to C57BL/6JJcl mice (CLEA Japan, Tokyo, Japan) for at least 8 generations. The breeding strategy was ultimately designed to generate Apc Min/ϩ mice with the PLC⑀ ϩ/ϩ or PLC⑀ ⌬X/⌬X background. The genotypes of PLC⑀ (13) and Apc (see the Jackson Laboratory Web site) were determined by PCR of the tail DNAs. All of the animals were maintained in the animal facility of Kobe University Graduate School of Medicine, and the use and care of the animals were reviewed and approved by the Institutional Animal Care and Use Committee of Kobe University.
Induction of Colitis and Colitis-induced Colorectal Cancer-To induce colitis, drinking water containing 2.5% (w/v) DSS (molecular weight ϭ 36,000 -50,000; Wako Pure Chemical, Osaka, Japan), dissolved in tap water, was orally administered to 8-week-old mice for 5 days (38,39). After that, the drinking water was substituted by tap water, and the mice were maintained for 16 days to develop colon tumors.
Histochemical Analyses-After sacrificing, the colons were removed from mice, washed with PBS, serially sectioned in the longitudinal direction, and embedded in paraffin. After that, the sections were subjected to H&E staining (14). MPO staining was performed on frozen sections (10 m thick) as described (40).
Histopathological Classification of Tumors-Tumors in H&E-stained sections were observed under a microscope and classified into early and late adenomas and adenocarcinomas according to the histopathological criteria recommended for the study of mouse models of intestinal cancer as described (41). The classification was carried out by a pathologist blinded in regard to the mouse genotypes.
Immunohistochemistry-The whole colon was rolled up in a "Swiss roll" configuration and embedded in OCT compound for sectioning. After fixation with 4% paraformaldehyde, the sections (10 m thick) were subjected to immunostaining as described (11,14). Images were taken with an Olympus FSX100 microscope and Olympus FSX100-BSW software. Statistical Analysis-Data are expressed as the mean Ϯ S.D. An unpaired Student's t test was performed for determination of p values. In cases where the p values were Ͻ0.05, differences were considered to be statistically significant. All of the data were obtained from at least three independent experiments.

Results
Role of PLC⑀ in Inflammation-induced Carcinogenesis-We employed the Apc Min/ϩ mouse colitis-induced colorectal carcinogenesis model, where tumor development is dependent on the DSS-induced colon inflammation (38,39). At first, FIGURE 1. Roles of PLC⑀ in colitis and colitis-induced carcinogenesis. Acute colitis was induced in Apc Min/ϩ mice carrying various PLC⑀ backgrounds by oral administration of DSS for 5 days, and the colons were collected at day 0 and 5 (3 animals/group). A, results of the daily body weight measurements of PLC⑀ ϩ/ϩ and PLC⑀ ⌬X / ⌬X mice. B, the longitudinal sections of the colons of PLC⑀ ϩ/ϩ (top panels) and PLC⑀ ⌬X/⌬X mice (bottom panels) with (right) or without (left) DSS administration were subjected to H&E staining. C, infiltrating neutrophils (left panels) and macrophages were detected by MPO staining and immunostaining for CD68, respectively, in the colon sections of PLC⑀ ϩ/ϩ and PLC⑀ ⌬X/⌬X mice with or without DSS administration. The bar graphs on the bottom show the numbers of MPO-positive neutrophils (left) and CD68-positive macrophages (right) per inflamed region (mm 2 ). D, The mRNA levels of the indicated proteins in the colons of PLC⑀ ϩ/ϩ and PLC⑀ ⌬X/⌬X mice with or without DSS administration were quantified by qRT-PCR and normalized to the ␤-actin mRNA level. E, the colon sections of PLC⑀ ϩ/ϩ mice administered with DSS were subjected to immunofluorescence staining for CXCL2 or CCL2 (red) and PLC⑀ (green) and DAPI staining (blue). Magnified images of the boxed regions of the left panels are shown in the right panels. F, CCL2 and CXCL2 were detected by immunofluorescence staining with the anti-CCL2 and anti-CXCL2 antibodies, respectively (red), in the colon sections of PLC⑀ ϩ/ϩ and PLC⑀ ⌬X/⌬X mice with or without DSS administration. The sections were also stained with DAPI (blue). Representative data are shown in B, C, E, and F. G, colorectal tumors were observed on the 16th day after DSS administration (6 animals/group). Shown are the total numbers of tumors/mouse (left) and the numbers/mouse (middle) and the percentages (right) of tumors classified into low grade and high grade adenomas and adenocarcinomas in PLC⑀ ϩ/ϩ and PLC⑀ ⌬X/⌬X mice. The data are shown as the mean Ϯ S.D. (error bars). *, p Ͻ 0.05. Scale bars, 25 m. PLC⑀ ϩ/ϩ and PLC⑀ ⌬X/⌬X mice were orally administered with 2.5% DSS for 5 days to induce acute colitis, the severity of which was evaluated by the body weight loss and histological examination. Compared with PLC⑀ ϩ/ϩ mice, the reduction in the body weight was attenuated in PLC⑀ ⌬X/⌬X mice (Fig. 1A). Likewise, mucosal erosion and leukocyte infiltration were alleviated in PLC⑀ ⌬X/⌬X mice (Fig. 1B). In this model, infiltration of neutrophils expressing CXCR2 (CXC chemokine receptor 2) and macrophages expressing CCR2 (CC chemokine receptor 2) is known to play critical roles in the pathogenesis of colitis (42,43). Indeed, we observed marked increases in the numbers of leukocytes and macrophages in the lesional area as detected by MPO staining and immunostaining for CD68, respectively (Fig.  1C). Marked elevation of CXCR2 and CCR2 mRNAs was observed in the DSS-administered colons of PLC⑀ ϩ/ϩ mice (Fig. 1D). In contrast, the colons of PLC⑀ ⌬X/⌬X mice exhibited markedly attenuated responses to DSS administration. Moreover, the expression of other inflammation-associated genes, such as Tnf-␣, Cox-2, Cxcl1, Cxcl2, and Ccl2, showed a substantial increase after DSS administration in PLC⑀ ϩ/ϩ mice, which was also markedly suppressed in PLC⑀ ⌬X/⌬X mice (Fig. 1D). Intriguingly, CXCL2 and CCL2, the ligands for CXCR2 and CCR2, respectively, were abundantly expressed after DSS administration in the colon epithelial cells expressing PLC⑀, and this expression was markedly suppressed in PLC⑀ ⌬X/⌬X mice (Fig. 1, E and F). These results suggested that PLC⑀ might augment colon inflammation by facilitating the secretion from the colon epithelial cells of proinflammatory factors, such as CXCL2 and CCL2, which recruit neutrophils and macrophages, respectively.
We next analyzed the effects of the PLC⑀ genotypes on the colitis-induced colorectal carcinogenesis of Apc Min/ϩ mice (38,39). As a result, PLC⑀ ⌬X/⌬X mice exhibited substantial reduction in the total numbers of colon tumors observed on the 16th day after DSS administration, compared with PLC⑀ ϩ/ϩ mice (Fig. 1G). Histopathological classification of the tumors revealed that the ratios of low grade and high grade adenomas and adenocarcinomas were ϳ40, 30, and 30%, respectively, in PLC⑀ ϩ/ϩ mice. In contrast, Ͼ80% of the tumors were low grade adenomas and essentially no adenocarcinoma was formed in PLC⑀ ⌬X/⌬X mice. These results taken together gave further support for our notion that PLC⑀ facilitates tumor promotion and malignant progression by augmenting inflammation.
Role of PLC⑀ in NF-B-dependent Proinflammatory Gene Expression-The molecular mechanism underlying the action of PLC⑀ on the expression of proinflammatory factors was analyzed by using a human colon epithelial cell line, Caco2. Because PLC⑀ had been implicated in the induction of the inflammation-associated genes in response to TNF-␣ stimulation (8,11,17), we first examined the effects of the siRNA-mediated knockdown of PLC⑀ on TNF-␣-induced expression of these genes (Fig. 2). In Caco2 cells transfected with an siRNA control, the expression of CXCL8 and CCL2 was elevated in 1-2 h after TNF-␣ stimulation ( Fig. 2A). The transfection of two distinct siRNAs against PLC⑀, siRNA 1 and siRNA 2, which efficiently knocked down PLC⑀ expression (Fig.  2B), inhibited the expression of both CXCL8 and CCL2 significantly at the 2-h time point but not effectively at the 1-h time point ( Fig. 2A). In addition, the expression of COX-2 showed sustained elevation in 1-2 h, as was the case with CXCL8 (Fig.  2C). The expression of CCL20 was mainly elevated in 2 h only, whereas expression of CXCL1 was mainly elevated in 1 h only. In these cases, too, the transfection of siRNA 1 inhibited the expression at the 2-h time point of COX-2 and CCL20 (Fig. 2C).
TNF-␣-induced expression of the inflammation-associated genes, shown here to be regulated by PLC⑀, was known to be mediated by the NF-B pathway (44). This led us to examine the effects of PLC⑀ knockdown on the nuclear translocation of p65 (Fig. 3). Nuclear accumulation of p65 was detected in 30 and 90 min after TNF-␣ stimulation in Caco2 cells transfected with the control siRNA (Fig. 3A). Transfection of the two distinct PLC⑀ siRNAs inhibited the p65 nuclear accumulation at the time point of 90 min but not 30 min. The Ser-32 phosphorylation of IB, which is required for its degradation and subsequent p65 nuclear translocation (23,24), was observed in a biphasic manner, peaking at 10 and 90 min after TNF-␣ stim- ulation in cells transfected with the control siRNA (Fig. 3B). Transfection of the two PLC⑀ siRNAs lowered the peak at 90 min but not that at 10 min. The p65 nuclear localization observed at 30 min seemed to result from the IB phosphorylation at 10 min, taking account of the time lag between the two events. These results suggested that PLC⑀ might be involved in the late phase of the TNF-␣-induced NF-B activation.
Roles of PKD and LPA Receptor in PLC⑀-dependent NF-B Activation-We next examined the role of PKD downstream of PLC⑀. Treatment of Caco2 cells with the two distinct PKD inhibitors, CID755673 and CID2011756, inhibited the IB phosphorylation at 90 min but not at 10 min as well as the p65 nuclear translocation at 90 min but not at 30 min after TNF-␣ stimulation (Fig. 4, A and B). These results were very similar to those obtained with the siRNA-mediated knockdown of PLC⑀ (Fig. 3). Moreover, the Ser-916 phosphorylation of PKD induced by TNF-␣ stimula-tion was suppressed by transfection of the two PLC⑀ siRNAs (Fig.  4C), suggesting that PKD might function downstream of PLC⑀ for regulation of the NF-B activation.
The delayed action of PLC⑀ upon the TNF-␣-induced NF-B activation led us to think of the possibility that another humoral factor might mediate the activation of PLC⑀ induced by TNF-␣ stimulation. It was reported that TNF-␣ stimulation of cells induced the production of LPA and S1P (45,46), which were known to activate PLC⑀ (6,7). This prompted us to examine the effects of the antagonists of the receptors for LPA and S1P on the TNF-␣-induced cytokine expression and NF-B activation (Fig. 5). The expression of CXCL8 was abrogated by treatment of cells with the LPA receptor antagonist tetradecyl phosphonate (Fig. 5A). In contrast, treatment with the S1P receptor antagonist JTE013 seemed to have little effect on the CXCL8 expression at the 2-h time point, although it unexpectedly exhibited a stimulatory effect even in the absence of TNF-␣ (Fig. 5A). Presently, we do not have any idea for the mechanism underlying this stimulatory effect. Tetradecyl phosphonate inhibited the IB phosphorylation at 90 min but not at 10 min as well as the Ser-916 phosphorylation of PKD at 90 min after TNF-␣ stimulation (Fig. 5B). Furthermore, LPA stimulation of Caco2 cells induced both the IB phosphorylation and PKD phosphorylation in an early time course of 30 min, which was abrogated by the siRNA-mediated knockdown of PLC⑀ (Fig.  5C). These results strongly suggested that TNF-␣ stimulation induces LPA production and subsequent activation of LPA receptors in an autocrine manner, thereby leading to activation of the PLC⑀-PKD axis.
Role of RSK in PLC⑀-dependent NF-B Activation-We next analyzed the molecular mechanism by which PLC⑀ induced the IB phosphorylation, depending on the stimulation by TNF-␣ or LPA. To this end, we tested the involvement of three kinases, IKK, RSK, and CK2, which had been known to phosphorylate IB at Ser-32 (25)(26)(27)(28). The phosphorylation of IKK␣/␤ at Ser-176 and Ser-180, which is responsible for IKK activation (47,48), was induced in 10 min but not later after TNF-␣ stimulation (Fig. 6A). This early IKK phosphorylation was unaffected by the siRNA-mediated PLC⑀ knockdown and likely to account for the early phase of the IB phosphorylation directly caused by the TNF-␣ receptor signaling (Fig. 3B). Because no increase in phosphorylation was observed after 10 min, it is unlikely that IKK is involved in the PLC⑀-dependent IB phosphorylation. This led us to test RSK and CK2. The TNF-␣-induced CXCL8 expression was suppressed by treatment with the RSK inhibitor BI-D1870 but not the CK2 inhibitor TBB (Fig. 6B). In Caco2 cells transfected with the control siRNA, TNF-␣ stimulation induced elevation of RSK phosphorylation at Ser-380 in 90 min, which was diminished in cells transfected with the PLC⑀ siRNA (Fig. 6C) or pretreated with CID755673 and CID2011756 (Fig.  6D). Moreover, the RSK Ser-380 phosphorylation induced in 10 -30 min after LPA stimulation was abrogated by the siRNA-mediatedPLC⑀knockdown (Fig.6E).Ineithercase,phosphorylation of ERK1/2 at Thr-202 and Tyr-204 was elevated in a similar time course with the RSK phosphorylation and unaffected by the PLC⑀ knockdown. Furthermore, pretreatment with the MEK inhibitors, U0126 (49) and trametinib (50), or the RSK inhibitor BI-D1870 abrogated the IB phosphorylation at 90 min after TNF-␣ stimulation (Fig. 6, F and G). The two MEK inhibitors almost completely abolished the RSK phosphorylation, confirming that ERK activation is necessary for the Ser-380 phosphorylation and activation of RSK (29). These results taken together indicated that the PLC⑀-PKD axis facilitates the phosphorylation and activation of RSK by ERK, leading to the phosphorylation and activation of IB by RSK. The ERK activation did not seem to be regulated by it.
Molecular Mechanism for RSK Regulation by PLC⑀-RSK, activated through phosphorylation by ERK and PDK1 in the cytoplasm, was located in the cytoplasm and the nuclei, where it phosphorylates IB and CREB, respectively (26 -31), suggesting that blocking the nuclear translocation of RSK might enhance IB phosphorylation. Thus, we examined whether PLC⑀ affects the subcellular localization of RSK. In control Caco2 cells, the amount of the phosphorylated RSK located in the cytoplasm was increased in 90 min after TNF-␣ stimulation (Fig. 7A). Notably, transfection of the two distinct PLC⑀ siRNAs markedly reduced the amount of the phosphorylated RSK in the cytoplasm at 90 min, which was accompanied by its increase in the nuclei. Likewise, treatment with CID755673 or CID2011756 also resulted in disappearance from the cytoplasm and nuclear accumulation of the phosphorylated RSK at 90 min (Fig. 7B). Furthermore, overexpression of PLC⑀, but not the lipase-dead mutant, PLC⑀⌬X, increased the amount of the phosphorylated RSK located in the cytoplasm, decreased the amount of IB in the cytoplasm, and increased the nuclear translocation of p65 upon stimulation by a low dose of LPA (Fig.  7C), suggesting that the lipase activity of PLC⑀ was involved. These phenomena were suppressed by treatment with CID755673. These results taken together indicated that activation of the PLC⑀-PKD axis causes preferential localization of the phosphorylated RSK in the cytoplasm, leading to enhancement of the IB phosphorylation and nuclear translocation of p65.
We next analyzed the molecular mechanism by which PLC⑀ causes preferential cytoplasmic localization of RSK. We focused on PEA15, a scaffold protein for RSK in the cytoplasm, which had been reported to regulate subcellular localization of RSK, depending on the phosphorylation state (32,33). RSK could be co-immunoprecipitated with PEA15 in Caco2 cells stimulated with LPA, which was markedly enhanced by PLC⑀ overexpression (Fig. 7D). The association of RSK with PEA15 seemed independent of its Ser-380 phosphorylation. Moreover, LPA stimulation elevated the phosphorylation of PEA15 at Ser-104, a putative PKC target site implicated in enhancing the cytoplasmic localization of RSK, which was also enhanced by PLC⑀ overexpression (33) (Fig. 7E). In either case, the phenomenon was effectively inhibited by treatment with CID755673. These results suggested that the Ser-104 phosphorylation of PEA15, induced by the activated PLC⑀-PKD axis, might enhance its association with the phosphorylated RSK and facilitate its cytoplasmic localization. To test this hypothesis, we analyzed the effects of the mutations, S104A and S104D, of  p values (right). B, Caco2 cells were serum-starved for 3 h, treated with 10 g/ml BI-D1870 or 28 g/ml TBB for 30 min, and stimulated with 20 ng/ml TNF-␣ for the indicated times. The CXCL8 mRNA levels were measured as described in the legend to Fig. 5A. C, Caco2 cells were treated with the indicated siRNAs and stimulated with TNF-␣ as described in A. RSK phosphorylated at Ser-916 and total RSK in the cell lysates were detected by immunoblotting with the anti-phospho-RSK and anti-RSK Abs, respectively. ERK1/2 phosphorylated at Thr-202 and Tyr-204 and total ERK1/2 were detected by immunoblotting with the anti-phospho-ERK and anti-ERK Abs, respectively. PLC⑀ and actin were detected by immunoblotting with the anti-PLC⑀ and anti-actin Abs, respectively. The numbers below the immunoblots indicate -fold changes of the signals of the phosphorylated proteins divided by the signals of the total proteins over those at 0 min after TNF-␣ stimulation of the control cells. PEA15. The phosphorylation-mimicking mutant S104D efficiently associated with RSK, including the phosphorylated one, even in the absence of LPA stimulation, whereas the S104A mutant failed to associate with RSK (Fig. 7F). Moreover, overexpression of the S104D mutant enhanced the cytoplasmic localization of the phosphorylated RSK in the absence of LPA stimulation, whereas that of the S104A mutant showed inhibition of the LPA-induced cytoplasmic localization (Fig. 7G). These results further supported the crucial role of the Ser-104 phosphorylation of PEA15, mediated by the PLC⑀-PKD axis, in holding the phosphorylated RSK in the cytoplasm.

Discussion
In this study, we have shown that PLC⑀ augments inflammatory reactions by facilitating the proinflammatory cytokine expression in the colon administered with DSS. By employing the colitis-induced colorectal carcinogenesis model, we obtained further support for our hypothesis that this proinflammatory function of PLC⑀ is responsible for its crucial role in facilitating tumor promotion and malignant progression. PLC⑀ ⌬X/⌬X mice showed markedly attenuated responses to DSS administration, represented by reduced expression of the proinflammatory molecules, such as TNF-␣, COX-2, CXCL1, CXCL2, and CCL2 (Fig. 1D), all of which had been implicated in the pathogenesis of DSS-induced colitis and human inflammatory bowel diseases (42,43,(51)(52)(53)(54)(55)(56)(57)(58). Although this result indicated the involvement of PLC⑀ in augmenting the expression of the proinflammatory molecules, it was difficult to distinguish whether the observed increase of the proinflammatory molecules was the direct consequence of the PLC⑀ activation or the secondary phenomenon accompanying the infiltration of immune inflammatory cells. In this regard, it is of particular interest that CXCR2 ligands, CXCL1 and CXCL2, and a CCR2 ligand, CCL2, were overexpressed in the colon epithelial cells of DSS-administered PLC⑀ ϩ/ϩ mice (Fig. 1, E and F). Because PLC⑀ is not expressed in immune cells (8), it was likely that the epithelial expression of these chemokines represents one of the direct consequences of the PLC⑀ activation. Because mice carrying knock-out of CXCR2 or CCR2 had been shown to become less susceptible to DSS-induced colitis (42,43,51), these results suggested a mechanism whereby PLC⑀ might augment colon inflammation by facilitating the production from the colon epithelial cells of proinflammatory chemokines, such as CXCL2 and CCL2, which recruit neutrophils and macrophages, respectively, to the inflamed sites.
By taking our results obtained with the human colon epithelial cell line Caco2 together, we propose the following model for the molecular mechanism by which PLC⑀ augments the expression of proinflammatory molecules (Fig. 8). TNF-␣ stimulation induces the early phase of NF-B activation through phosphorylation and activation of IB by IKK, leading to the expression of the inflammation-associated genes, such as CXCL1, CXCL8, and COX-2, observed in 1 h (Fig. 2, A and C). Although this phase of the inflammation-associated gene expression seems to be partially abrogated by PLC⑀ knockdown, we presently have no mechanistic explanation except that the PKD-RSK-IB-NF-B pathway is not involved. Alternatively, the gene expression observed at the 1-h time point might be significantly contributed by that in the late phase, which will be explained below. In parallel with the early phase of the NF-B activation, TNF-␣ stimulation enhances the production and secretion of LPA, which induces activation of PLC⑀ through engagement of its receptors in an autocrine manner. PLC⑀ produces diacylglycerol and thereby induces PKC activation, leading to the activation of PKD through direct binding of diacylglycerol to its C1 domain and phosphorylation by PKC (59). Subsequently, the activated PKD induces phosphorylation of PEA15 at Ser-104. The phosphorylated PEA15 binds to RSK activated through phosphorylation by ERK, which is also activated by the LPA receptor engagement (Fig. 6, C-E). The activated RSK, held in the cytoplasm by association with the phosphorylated PEA15, FIGURE 7. Roles of PLC⑀ and PKD in cytoplasmic localization of phosphorylated RSK. A, Caco2 cells transfected with the indicated siRNAs were cultured for 38 -42 h, serum-starved for 3 h, and stimulated with 20 ng/ml TNF-␣ for the indicated times. The cell lysates were subjected to subcellular fractionation, and the resulting nuclear and cytoplasmic fractions were subjected to immunoblotting with the anti-phospho-RSK Ab. TBP and ␣-tubulin were used as markers for the nuclear and the cytoplasmic fractions, respectively. The averages of the intensities of the immunoreactive signals of phospho-RSK in the cytoplasmic and nuclear fractions at 90 min of three independent experiments are expressed as the mean Ϯ S.D. (error bars) in arbitrary units (AU) with p values (bottom). B, Caco2 cells were serum-starved for 3 h, pretreated with 10 g/ml CID755673 or 50 M CID2011756 for 30 min, and stimulated with 20 ng/ml TNF-␣ for the indicated times. Subcellular localization of phosphorylated RSK was examined as described in A. The averages of the intensities of the immunoreactive signals of phospho-RSK in the cytoplasmic and nuclear fractions at 90 min of three independent experiments are expressed as the mean Ϯ S.D. in arbitrary units with p values (bottom). C, Caco2 cells were transfected with pFLAG-CMV2-PLC⑀ or pFLAG-CMV2-PLC⑀⌬X and cultured for 14 -18 h. Subsequently, the cells were serum-starved for 14 -18 h, treated with or without 10 g/ml CID755673 for 15 min, and stimulated with 20 M LPA for 30 min. Subcellular localization of phosphorylated RSK was examined as described in A. Subcellular localization of p65 was measured as described in the legend to Fig. 3A. IB was detected by immunoblotting with the anti-IB Ab, and its levels are shown as -fold changes over those in the control cells. The averages of the intensities of the immunoreactive signals of phospho-RSK and IB in the cytoplasmic fractions and p65 in the nuclear fractions obtained from LPA-stimulated cells of three independent experiments are expressed as the means Ϯ S.D. in arbitrary units with p values (right). D, Caco2 cells were co-transfected with pFLAG-CMV2-PLC⑀ and pCMV-HA-PEA15, treated with or without CID755673, and stimulated with LPA as described in C. The cell lysates were immunoprecipitated (IP) with the anti-HA Ab and protein G-Sepharose. The resulting immunoprecipitates were subjected to immunoblotting with anti-phospho-RSK, anti-RSK, and anti-HA Abs. The numbers below the immunoblots show -fold changes of the levels of phosphorylated RSK and total RSK over those in the control cells. The averages of the intensities of the immunoreactive signals of phospho-RSK and total RSK of the immunoprecipitates obtained from LPA-stimulated cells of three independent experiments are expressed as the mean Ϯ S.D. in arbitrary units with p values (right). E, phosphorylation of PEA15 at Ser-104 was detected in cells described in D by immunoblotting with anti-phospho-PEA15 and anti-PEA15 antibodies. The numbers below the immunoblots indicate -fold changes of the phospho-PEA15 signals divided by the total PEA15 signals over that in the control cells. The averages of the intensities of the immunoreactive signals of phospho-PEA15 obtained from LPA-stimulated cells of three independent experiments are expressed as the means Ϯ S.D. in arbitrary units with p values (right). F and G, cells were transfected with either pCMV-HA-PEA15, pCMV-HA-PEA15S104A, or pCMV-HA-PEA15S104D in combination with pFLAG-CMV2-PLC⑀ and stimulated with LPA as described in C. F, immunoprecipitation with the anti-HA antibody and immunoblotting for detection of phosphorylated RSK, total RSK, and HA-PEA15 were performed as described in D. G, subcellular localization of phosphorylated RSK was examined as described in A. The averages of the intensities of the immunoreactive signals of phospho-RSK and total RSK in the immunoprecipitates (F) and phospho-RSK in the cytoplasmic fractions (G) of three independent experiments are expressed as the mean Ϯ S.D. in arbitrary units with p values (bottom). Three experiments performed independently yielded equivalent results.
phosphorylates IB and thereby causes the late phase of the TNF-␣-induced NF-B activation, leading to the expression of CXCL8, COX-2, CCL2, and CCL20. CXCL8, whose counterpart is nonexistent in mice, is a ligand of CXCR2, causes neutrophil recruitment, and hence is regarded as a functional homologue of mouse CXCL1 and CXCL2 (60 -62). Thus, the PLC⑀-dependent induction of CXCL8 observed in Caco2 cells may functionally substitute for CXCL2, which is induced in mouse colon epithelial cells in a PLC⑀-dependent manner after DSS administration. The delayed kinetics of the PLC⑀-dependent cytokine expression in response to TNF-␣ stimulation was observed before in dermal fibroblasts and epidermal keratinocytes, which led us to propose that unknown humoral factors might mediate activation of PLC⑀ in an autocrine manner (8). This is consistent with our present observation. The results taken together indicated that the differential actions of IKK and RSK on IB phosphorylation account for the sustained NF-B activation and the sustained expression of proinflammatory molecules, which seem to be involved in augmentation of inflammatory reactions and promotion of inflammation-associated carcinogenesis. Several questions, including the following, remain to be addressed experimentally: whether TNF-␣ stimulation of cells really induces the increased production of LPA as reported (45), by what mechanism PLC⑀ is activated downstream of the LPA receptors, and what kinase is directly responsible for the Ser-104 phosphorylation of PEA15.
The results of our present study may give a new mechanistic insight into not only the pathogenesis of human inflammatory bowel diseases and colorectal carcinogenesis but also of other inflammatory diseases and inflammation-associated carcinogenesis. Moreover, our study suggests that PLC⑀ may become a good candidate target for the development of anti-inflammatory and cancer-preventing drugs.
Author Contributions-M. W., H. E., and T. K. designed the study and analyzed the data. M. W. conducted most of the experiments. M. L. and A. E. conducted a part of the experiments. S. K. conducted the histopathological classification. T. K. conceived the idea for the project and wrote the paper with M. W. All of the authors reviewed the results and approved the final version of the manuscript.