Up-regulation of S100P Expression by Non-steroidal Anti-inflammatory Drugs and Its Role in Anti-tumorigenic Effects*

Epidemiological studies have revealed that prolonged use of non-steroidal anti-inflammatory drugs (NSAIDs) reduces the risk of cancer. Various mechanisms, including induction of apoptosis and inhibition of the growth and invasion of cancer cells, have been implicated in this anti-tumorigenic activity. In this study we focused on S100P, which is known to be overexpressed in clinically isolated tumors and which functions through both intracellular and extracellular mechanisms. We showed the up-regulation of S100P expression in human gastric carcinoma cells treated with various NSAIDs, including celecoxib. The celecoxib-mediated up-regulation of S100P was suppressed by the transfection of cells with small interfering RNA for activating transcription factor 4 (ATF4), a transcription factor involved in the endoplasmic reticulum stress response. Furthermore, deletion of ATF4 binding consensus sequence located in the promoter of the S100P gene resulted in inhibition of celecoxibmediated transcriptional activation of the gene. These results suggest that celecoxib up-regulates the expression of S100P through an ATF4-mediated endoplasmic reticulum stress response. Celecoxib inhibited the growth and induced apoptosis, and these actions could be either suppressed or stimulated by transfection of cells with S100P overexpression plasmid or small interfering RNA, respectively. Celecoxib also inhibited the invasive activity of the cells. Cromolyn, which inhibits the binding of S100P to its receptor, enhanced the celecoxib-mediated inhibition of cell invasion and growth but did not affect apoptosis. These results suggest that S100P affects apoptosis, cell growth, and invasion through either an intracellular or an extracellular mechanism and that the up-regulation of S100P expression by NSAIDs reduces their anti-tumorigenic activity.

matory effects, recent epidemiological studies have revealed that prolonged NSAID use reduces the risk of cancer, whereas preclinical and clinical studies have indicated that some NSAIDs, in particular celecoxib, are effective in the treatment and prevention of cancer (1). The anti-tumorigenic activity of NSAIDs is believed to involve various mechanisms, including induction of apoptosis, cell growth suppression, inhibition of angiogenesis, and inhibition of metastasis (cell invasion suppression) (2,3).
NSAIDs act as inhibitors of cyclooxygenase (COX), an enzyme essential for the synthesis of prostaglandins (PGs). PGs, such as PGE 2 , inhibit apoptosis of cancer cells and stimulate their growth and invasion as well as promote angiogenesis (4 -6). Thus, it is certain that the anti-tumorigenic effect of NSAIDs was mediated mainly through the inhibition of COX. However, several lines of evidence now suggest that a COXindependent mechanism may also be involved (7,8). To investigate this COX-independent mechanism, we used DNA microarray techniques to systematically search for genes in human gastric carcinoma (AGS) cells whose expression was altered by the NSAID indomethacin in a COX-independent manner (9,10). This analysis revealed that NSAIDs induce an endoplasmic reticulum (ER) stress response (11). ER stress response is induced through transcription factors, such as activating transcription factor 6 (ATF6) 6 and ATF4 (12)(13)(14), and we have previously reported that both ATF4 and ATF6 are activated by various NSAIDs, including indomethacin and celecoxib (15,16). In this study we focused our attention on S100P, the expression of which appears to be induced by indomethacin based on results from DNA microarray analysis (10). S100P is a member of the S100 family of EF-hand Ca 2ϩbinding proteins (17). Overexpression of S100P has been observed in tumors clinically isolated from various tissue types, with the extent of the overexpression being positively correlated to the degree of malignancy (18 -23). Overproduction of S100P appears to stimulate tumor malignancy through both intracellular and extracellular mechanisms (24). Secreted S100P binds to its receptor, the receptor for activated glycation end products (RAGE), thereby stimulating the invasion and growth of cancer cells or inhibiting their apoptosis through activation of extracellular-regulated kinase (ERK) and nuclear factor-B (NF-B) (20,(25)(26)(27). Furthermore, S100P was suggested to function also in cells through its binding to ezrin and Casy/SIP (28 -30). It has recently been reported that S100P induces the expression of cathepsin D; however, the mechanism responsible for this effect is still to be identified (31).
It was recently reported that the expression of S100P is altered by some anti-tumor drugs (18,32), although the underlying regulatory mechanism remains unknown. In this study we report that various NSAIDs up-regulate the expression of S100P through an ATF4-mediated ER stress response. Furthermore, our results suggest that up-regulation of S100P expression by NSAIDs negatively affects their anti-tumorigenic activity through inhibition of apoptosis, stimulation of cancer cell growth and invasion.

EXPERIMENTAL PROCEDURES
Materials and Animals-RPMI 1640 medium was obtained from Nissui Pharmaceutical Co. Fetal bovine serum was purchased from Invitrogen. 1,2-Bis(2-aminophenoxy)ethane-N,N,NЈNЈ-tetraacetic acid (BAPTA-AM) was obtained from Dojindo Co. Cromolyn sodium salt, tunicamycin, normal mouse IgG, 1,4-diamino-2,3-dicyano-1,4-bis (o-aminophenylmercapto) butadiene ethanolate (U0126), N-p-tosyl-L-phenylalanine chloromethyl ketone (TPCK), and staurosporine were purchased from Sigma-Aldrich. Indomethacin, diclofenac, and thapsigargin were obtained from Wako Co. Celecoxib and meloxicam were from LKT Laboratories Inc. An antibody against actin, IB, or RAGE was purchased from Santa Cruz Biotechnology Inc., antibody against ERK was from Cell Signaling, and antibody against S100P was from R&D Systems Inc. The RNeasy kit, siRNAs, HiPerFect, and RNAiFect were from Qiagen. A first-strand cDNA synthesis kit was purchased from Amersham Biosciences. Lipofectamine (TM2000), zymogram developing buffer, and pcDNA3.1 plasmid were obtained from Invitrogen. The iQ SYBR Green Supermix was from Bio-Rad. S100P enzyme-linked immunosorbent assay kits were purchased from CircuLex. Matrigel was from BD Biosciences, and the 24-well transwells were from Costar. The Dual Luciferase Assay System was from Promega. Male nonobese diabetes/ severe combined immunodeficiency mice (5 weeks of age) were obtained from the Charles River. The experiments and procedures described here were carried out in accordance with the Guide for the Care and Use of Laboratory Animals as adopted and promulgated by the National Institute of Health and were approved by the Animal Care Committee of Kumamoto University.
Cell Culture and Overexpression of S100P-AGS, MKN45, and Kato III are human carcinoma cell lines derived from stomach. Cells were cultured in RPMI 1640 medium supplemented with 10% fetal bovine serum, 100 units/ml penicillin, and 100 g/ml streptomycin in a humidified atmosphere of 95% air with 5% CO 2 at 37°C. NSAIDs were dissolved in DMSO, and control experiments were performed in the same concentrations of DMSO alone. Cells were exposed to NSAIDs and other chemicals by changing the medium. Unless otherwise noted, cells were cultured for 24 h before use in experiments. The overex-pression plasmid for S100P was constructed by insertion of S100P-encoding DNA fragments from the plasmid (S100P wild type (His tag), a gift from Dr. Gerke, University of Muenster (28)) into pcDNA3.1. Transfection of AGS cells with the plasmid was then carried out using Lipofectamine (TM2000) according to the manufacturer's protocols. Stable transfectants expressing S100P were selected by immunoblotting analysis. Positive clones were maintained in the presence of 800 g/ml G418.
Real-time RT-PCR Analysis-Total RNA was extracted from cells using an RNeasy kit according to the manufacturer's instructions. Samples were reverse-transcribed using a firststrand cDNA synthesis kit. Synthesized cDNA was applied in real-time RT-PCR (Chromo 4 instrument (Bio-Rad)) experiments using iQ SYBR GREEN Supermix and analyzed with Opticon Monitor Software. The cycle conditions were 2 min at 50°C followed by 10 min at 90°C and finally 45 cycles each at 95°C for 30 s and 63°C for 60 s. Specificity was confirmed by electrophoretic analysis of the reaction products and by inclusion of template-or reverse transcriptase-free controls. To normalize the amount of total RNA present in each reaction, the actin gene was used as an internal standard. Primers were designed using the Primer3 Web site. Primers are listed as name, forward primer, and reverse primer: S100P, 5Ј-GATGC-CGTGGATAAATTGCT-3Ј, 5Ј-ACTTGTGACAGGCAGAC-GTG-3Ј; cathepsin D, 5Ј-GACACAGGCACTTCCCTCAT-3Ј, 5Ј-CCTCCCAGCTTCAGTGTGAT-3Ј; actin, 5Ј-GGACTTC-GAGCAAGAGATGG-3Ј, 5Ј-AGCACTGTGTTGGCGTA-CAG-3Ј.
Luciferase Assay-DNA fragments of the S100P promoter (from Ϫ1200 to Ϫ1) were amplified by PCR and ligated into the XhoI-HindIII site of the Photinus pyralis luciferase reporter plasmid, pGL3, to generate pS100Pluc. A plasmid with deletion of the amino acid-responsible element (AARE) sequence (33,34) (from Ϫ84 to Ϫ76) was generated by PCR. All plasmids were sequenced to exclude unexpected mutations.
The luciferase assay was performed as described previously (11,35). Cells were transfected with 1 g of each of the P. pyralis luciferase reporter plasmids (pS100Pluc or its derivative) and 0.125 g of the internal standard plasmid bearing the Renilla reniformis luciferase reporter (pRL-SV40). P. pyralis luciferase activity in cell extracts was measured using the Dual Luciferase Assay System and then normalized for R. reniformis luciferase activity.
Gelatin Zymography-The proteolytic activity of matrix metalloproteinase (MMP)-9 was assessed by SDS-PAGE using zymogram gels containing 0.1% (m/v) gelatin, as described previously (36). The culture medium was concentrated, and the protein concentration was determined according to the Bradford method (37). Following electrophoresis at 4°C, the gels were washed with 2.5% Triton X-100 for 30 min at room temperature and incubated with zymogram development buffer for 2 days at 37°C. Bands were visualized by staining with Coomassie Brilliant Blue.
Immunoblotting Analysis-Whole-cell extracts were prepared as described previously (38). The protein concentration of the samples was determined by the Bradford method (37). Samples were applied to polyacrylamide SDS gels and subjected S100P and NSAIDs FEBRUARY 13, 2009 • VOLUME 284 • NUMBER 7 to electrophoresis, and the resultant proteins were then immunoblotted with their respective antibodies.
Analysis of Apoptosis by Fluorescence-activated Cell Sorting-Apoptosis was monitored by fluorescence-activated cell sorting analysis as described previously (39). Briefly, cells were collected by centrifugation. The pellets were fixed with 70% ethanol for 4 h at Ϫ20°C, centrifuged, and re-suspended in phosphate-citrate buffer (0.2 M Na 2 HPO 4 and 4 mM citric acid), then incubated for 20 min at room temperature. After centrifugation, the pellets were re-suspended in DNA staining solution (50 g/ml propidium iodide and 10 g/ml RNase A) and incubated for 20 min at room temperature. Samples were scanned with a FACSCalibur (BD Biosciences) cell sorter under conditions designed to measure only specific propidium iodide-mediated fluorescence. Apoptotic cells appeared as a hypodiploid peak due to nuclear fragmentation and loss of DNA.
Invasion Assay-The invasion assay was performed as previously described (40) with some modifications. Serum-free RPMI 1640 medium containing 5 mg/ml Matrigel was applied to the upper chamber of a 24-well transwell and incubated at 37°C for 4 h. The cell suspension was applied to the Matrigel, and the lower chamber of the transwell was filled with culture medium containing 10% fetal bovine serum and 5 g/ml fibronectin. The plate was incubated at 37°C for 24 h. Cells were removed from the upper surface of the membrane, and the lower surface of the membrane was stained for 10 min with 0.5% crystal violet in 25% methanol, rinsed with distilled water, and air-dried overnight. The crystal violet was then extracted with 0.1 M sodium citrate in 50% ethanol, and the absorbance was measured at 585 nm.
Xenograft Tumor Growth-This assay was done as described (15,16) with some modification. Briefly, each nonobese diabetes/severe combined immunodeficiency mouse was inoculated subcutaneously in the both flanks with 1 ϫ 10 7 AGS cells. After 3 weeks, the mice began to receive a single daily intraperitoneal administration of cromolym, a protocol that continued for the duration of the study. Tumors were measured every 7 days, and their volumes were calculated.
Statistical Analysis-All values are expressed as the mean Ϯ S.D. Two-way analysis of variance followed by the Tukey test or the Student's t test for unpaired results was used to evaluate differences between more than three groups or between two groups. Differences were considered to be significant for values of p Ͻ 0.05.

RESULTS
NSAIDs Up-regulate S100P Expression-In a previous study we used DNA microarray analysis to search for genes whose expression is altered by indomethacin and found that S100P mRNA expression was up-regulated (10). In the present study we confirmed this result using a real-time RT-PCR technique. As shown in Fig. 1A, indomethacin up-regulated S100P mRNA expression in a dose-dependent manner. A similar result was obtained with the NSAID celecoxib, a finding that is particularly significant given its importance as an anti-cancer drug (Fig. 1B). Immunoblotting experiments revealed that celecoxib also up-regulates the expression of S100P at the protein level in both AGS cells and in another gastric cancer-derived cell line, MKN45 cells (Fig. 2, A and E). A similar response was observed in AGS cells with a number of other NSAIDs (indomethacin, meloxicam, and diclofenac; Fig. 2

, B-D).
COX exists as two subtypes, COX-1 and COX-2. Given that celecoxib and meloxicam are COX-2 selective, the results shown in Fig. 2, A-D, suggest that NSAIDs up-regulate S100P expression irrespective of COX selectivity. We next examined the celecoxib-mediated up-regulation of S100P expression in Kato III cells, in which COX-1 but not COX-2 mRNA is expressed (41). This phenotype was confirmed by RT-PCR (data not shown). As shown in Fig. 2F, celecoxib up-regulated the expression of S100P even in the Kato III cells; in other words, a COX-2-selective NSAID up-regulated S100P expression in cells lacking COX-2 expression, suggesting that this occurs independently of COX inhibition. For further confirmation of this point, we examined the effect of PGE 2 , revealing that it had no effect on the expression of S100P in the presence or absence of celecoxib (data not shown).
As described above, secreted S100P regulates various cell functions through its binding to RAGE (20,(25)(26)(27)(28). We monitored the level of S100P in the culture medium by enzymelinked immunosorbent assay and found that it increased in a dose-dependent manner in response to celecoxib treatment (Fig. 2G).
Mechanism for Up-regulation of S100P Expression by Celecoxib-As outlined above, the mechanism underlying the regulation of S100P expression is unknown. We have recently revealed that various NSAIDs including celecoxib induce an ER stress response, and the concentration of each NSAID required to mediate this response (15) is similar to that which induces S100P expression (Fig. 2). We investigated the role of ER stress response in NSAID-induced S100P expression by examining the effect of other ER stress-inducing chemicals (thapsigargin and tunicamycin) on the expression of S100P mRNA. As shown in Fig. 3, A and B, both of these chemicals increased S100P mRNA. We also confirmed that both agents up-regulated the expression of glucose-regulated protein (GRP) 78 mRNA, a representative marker of the ER stress response at the concentration specified in Fig. 3, A and B (data not shown). In contrast, exposure to staurosporine, which does not produce such a response (15,16), had no significant effect on S100P mRNA expression (Fig. 3C), suggesting that the expression of S100P is indeed linked to the ER stress response.
To confirm this we examined the effect of BAPTA-AM, an intracellular Ca 2ϩ chelator, on celecoxib-mediated up-regulation of S100P. We have previously shown that NSAIDs increase the intracellular Ca 2ϩ concentration and that this increase is required for the NSAID-induced ER stress response (15,42). BAPTA-AM significantly inhibited the celecoxib-mediated up-regulation of S100P mRNA (Fig. 4A) and GRP78 mRNA (data not shown) expression. At the concentration used, BAPTA-AM did not affect cell viability (data not shown). These results suggest that an increase in intracellular Ca 2ϩ and the resultant induction of the ER stress response are somehow involved in the up-regulation of S100P. We next used siRNA for ATF4 and ATF6 to examine the contribution of these ER stress    FEBRUARY 13, 2009 • VOLUME 284 • NUMBER 7 response-related transcription factors to the celecoxib-dependent up-regulation of S100P mRNA expression. Transfection of a given siRNA decreased mRNA and protein levels of its target gene but had no effect on those of the other gene in either the absence or presence of celecoxib (data not shown). As illustrated in Fig. 4B, the transfection of siRNA for ATF4 suppressed the celecoxib-mediated up-regulation of S100P mRNA expression but did not affect its basal expression (Fig. 4B). In contrast, ATF6 siRNA had no significant effect on S100P mRNA expression (Fig. 4C), suggesting that ATF4 rather than ATF6 is responsible for the celecoxib-mediated transcriptional activation of the S100P gene.

S100P and NSAIDs
AARE is the consensus sequence to which ATF4 binds when stimulating the transcription of downstream genes (33,34). We identified the AARE sequence in the promoter of the S100P gene (Fig. 5A), then tested the contribution of this sequence to the celecoxib-mediated transcriptional activation of the gene by examining the effect of its deletion on the promoter activity of S100P gene using a reporter plasmid where the promoter of the S100P gene was inserted upstream of the luciferase gene (Fig. 5A). As shown in Fig. 5B, treatment of cells not only with celecoxib but also with thapsigargin stimulated the luciferase activity in the cells, suggesting that up-regulation of S100P expression by celecoxib is achieved at the level of transcription through the ER stress response. Furthermore, the deletion of AARE significantly decreased the luciferase activity in the presence of celecoxib or thapsigargin but not in their absence (Fig.  5B), indicating that ATF4 binding to AARE plays an important role in celecoxib-mediated transcriptional activation of the S100P gene.
Role of Up-regulation of S100P Expression in the in Vitro Anti-tumorigenic Activity of Celecoxib-As described above, various mechanisms have been proposed for the chemopreventive and chemotherapeutic action of NSAIDs; these include inhibition of cell growth and invasion and stimulation of apo-ptosis (2, 3). On the other hand, expression of S100P has been shown to stimulate the aggressiveness of cancer cells through stimulation of their growth and invasiveness and inhibition of apoptosis via both intracellular and extracellular mechanisms (20,(25)(26)(27)(28)(29)(30). Here, we examined the role of celecoxib-mediated up-regulation of S100P expression in its anti-tumorigenic activity in vitro. This was achieved by constructing stable transfectants of AGS cells that continuously overexpressed S100P (Clone 8), this being confirmed at both the mRNA and protein levels ( Fig. 6A and see Fig. 8B). Treatment of cells with celecoxib (40 M) up-regulated the expression of S100P mRNA even in the S100P-overexpressing cells (Fig. 6A). Fig. 6B shows the cell growth curve for S100P-overexpressing cells and mock transfectant control cells in the presence or absence of celecoxib; S100P-overexpressing cells had a faster growth rate than control cells in not only the absence but also the presence of celecoxib. We also examined the effect of S100P siRNA on celecoxib-mediated inhibition of cell growth after  . Effect of S100P expression on celecoxib-dependent inhibition of cell growth. S100P-overexpressing AGS cells (Clone 8) and mock transfectant-expressing control cells (Mock) were incubated with the indicated concentrations of celecoxib and/or cromolyn for either 24 h (A) or the indicated periods (B and D). C, AGS cells were transfected with siRNA for S100P (siS100P) or non-silencing siRNA (ns) with the total amount of siRNA fixed at 1 g. After 24 h, cells were incubated with the indicated concentrations of celecoxib for the indicated periods. A, S100P mRNA expression was monitored and expressed as described in the legend of Fig. 1. B-D, cell numbers were determined by direct cell counting. Each mouse was inoculated with S100P-overexpressing AGS cells (Clone 8) and mock transfectant-expressing control cells (Mock), leading to tumor development. Cromolyn was then administered intraperitoneally as a single daily dose (5 mg/kg) for the duration of the study. Tumors were measured every 7 days, and their volumes are calculated (E). Values are the mean Ϯ S.D. (n ϭ 3 (A-D) or n ϭ 6 (E)). *, p Ͻ 0.05; **, p Ͻ 0.01; n.s., not significant. first confirming that this siRNA, but not nonspecific siRNA, suppresses S100P expression at both the mRNA and protein levels (see Fig. 8, E and F). As shown in Fig. 6C, cell growth was significantly suppressed by S100P siRNA transfection in both the absence and the presence of celecoxib. These results suggest that celecoxib-mediated up-regulation of S100P expression weakens the inhibitory effect of the drug on AGS cell growth.
As described above, S100P functions extracellularly via its binding to RAGE (20,(25)(26)(27). To test the contribution of this extracellular mechanism, we examined the effect of cromolyn, an antiallergy drug that has recently been shown to act as an inhibitor of S100P binding to RAGE (27), on the celecoxibmediated inhibition of cell growth. Cromolyn did not affect the expression of S100P mRNA (Fig. 6A). However, it slightly enhanced the inhibitory effect of celecoxib on cell growth without significantly affecting growth in the absence of the drug (Fig. 6D), suggesting that celecoxib-induced S100P may function extracellularly (via its binding to RAGE) to weaken the inhibitory effect of celecoxib on cell growth. In S100P-overexpressing cells, cromolyn slightly inhibited cell growth even in the absence of celecoxib, suggesting that the extracellular S100P signaling may generally (rather than specifically in the presence of celecoxib) play an important role in the regulation of cell growth.
We also examined the effect of overexpression of S100P and cromolyn on growth of xenograft tumors in immunodeficient nonobese diabetes/severe combined immunodeficiency mice characterized by T cell, B cell, and natural killer cell deficiency and lack of macrophage function. Tumors were developed in mice by inoculation subcutaneously of AGS cells (S100P-overexpressing (Clone 8) and mock transfectant control cells). Growth of xenograft tumors that overexpress S100P was faster than that of control (Fig. 6E). Intraperitoneal administration of cromolyn clearly inhibited the growth of xenograft tumors that overexpress S100P but not that of the control (Fig. 6E), suggesting that S100P stimulates the growth of AGS cells also in vivo extracellularly (via its binding to RAGE). Results are basically similar to those observed previously (27).
The invasive capacity of cancer cells is also important for the progression of tumors, especially in relation to metastasis, and we have recently reported that NSAIDs inhibit the invasiveness of AGS cells (9). Thus, we tested the contribution of up-regulation of S100P expression to celecoxib-mediated inhibition of cell invasiveness. As shown in Fig. 7A, celecoxib inhibited the invasive activity of AGS cells, an effect that was further stimulated in the presence of cromolyn. S100P-overexpressing cells displayed greater invasiveness than control cells in both the absence and the presence of celecoxib, with cromolyn reducing the invasive activity of the S100P-overexpressing cells but not that of control cells (Fig. 7A). Furthermore, even in the presence of celecoxib, S100P-overexpressing cells displayed greater invasiveness than control cells. We also examined the effect of neutralizing antibodies against S100P or RAGE on the S100P-overexpression-dependent stimulation of cell invasion activity. As shown in Fig. 7B, each neutralizing antibodies suppressed the S100P overexpression-dependent stimulation of cell invasion activity. All these results suggest that celecoxib-induced S100P decreases the inhibitory effect of the drug on cell invasion via its extracellular binding to RAGE.
It has recently been reported that S100P up-regulates the expression of cathepsin D, which stimulates the invasion of cancer cells (31). As shown in Fig. 7C, cathepsin D mRNA FIGURE 7. Effect of S100P expression on celecoxib-dependent inhibition of cell invasion. A and B, the invasive activity of S100P-overexpressing AGS cells (Clone 8) and mock transfectant control cells (Mock) was measured in the presence of the indicated concentrations of celecoxib and cromolyn or antibodies against S100P (20 g/ml) or RAGE (2 g/ml) as described under "Experimental Procedures" and is expressed relative to the control. C and D, S100P-overexpressing AGS cells (Clone 8) and mock transfectant-expressing control cells (Mock) were incubated with the indicated concentrations of celecoxib and/or cromolyn for 24 h. Both types of cells were preincubated with U0126 (20 M) or TPCK (20 M) for 1 h and further incubated for 24 h without the drug (E). The mRNA expression of cathepsin D was monitored and expressed as described in the legend of Fig. 1 (C). MMP-9 activity in the culture medium was measured as described under "Experimental Procedures." The clear band intensity was determined (D and E). The presence of phosphorylated ERK (P-ERK) and IB was monitored by immunoblotting (E). Values are the mean Ϯ S.D. (n ϭ 3 (A-C and E) or 6 (D)). *, p Ͻ 0.05; **, p Ͻ 0.01; n.s., not significant (A-D). FEBRUARY 13, 2009 • VOLUME 284 • NUMBER 7 expression was higher in S100P-overexpressing cells than in control cells, an effect that was suppressed in the presence of cromolyn, suggesting that S100P up-regulates the expression of cathepsin D through the extracellular mechanism. However, celecoxib and/or cromolyn had no effect on cathepsin D mRNA expression in control cells (Fig. 7C). Therefore, expression of cathepsin D does not seem to be involved in the inhibitory effect of S100P on celecoxib-mediated suppression of cell invasiveness. MMPs, especially MMP-9, also play an important role in cell invasion (43,44). We, therefore, next examined the effect of celecoxib and/or cromolyn on MMP-9 activity using gelatin zymography. The band intensity, indicative of MMP-9 activity, was decreased by treatment of cells with celecoxib, and this effect was further stimulated in the presence of cromolyn (Fig.  7D). S100P-overexpressing cells showed higher MMP-9 activity than control cells in both the presence and absence of celecoxib, and cromolyn inhibited MMP-9 activity in S100P-overexpressing cells (Fig. 7D). These results suggest that expression of S100P stimulates MMP-9 activity through the extracellular mechanism and that this mechanism is involved in the inhibitory effect of S100P on celecoxib-mediated suppression of cell invasion.

S100P and NSAIDs
It was reported that S100P activates ERK and NF-B (25,26), both of which are known to activate MMP-9 (45)(46)(47), suggesting that the S100P-dependent activation of MMP-9 is mediated by ERK and NF-B. To test this idea, we examined the effect of an inhibitor for ERK or NF-B on S100P-dependent activation of MMP-9. We confirmed that overexpression of S100P caused phosphorylation (activation) of ERK and decrease in the amount of IB (an inhibitor for NF-B) (Fig.  7E), both of which were inhibited by cromolyn (data not shown). As shown in Fig. 7E, treatment of cells with U0126, an inhibitor of ERK phosphorylation (activation), inhibited not only S100P-dependent phosphorylation of ERK but also activation of MMP-9. Treatment of S100P-overexpressing cells with TPCK (an inhibitor of proteasome system that depredates IB) recovered the level of IB in S100P-overexpressing cells (Fig. 7E), suggesting that this agent inhibited the activity of NF-B. Treatment of cells with TPCK inhibited S100P-dependent activation of MMP-9. Results in Fig.  7E suggest that the S100P-dependent activation of MMP-9 is mediated by ERK and NF-B.
We have recently reported that celecoxib induces apoptosis through induction of the ER stress response, particularly the induction of CHOP (11,15). Here we examined the role of up-regulation of S100P expression in apoptosis induced by celecoxib. Celecoxib at a concentration of 60 M up-regulated the expression of both S100P mRNA and protein not only in control cells but also in S100P-overexpressing cells (Fig. 8, A and B). Treatment of AGS cells with celecoxib at the same concentration clearly induced apoptosis, as described above (16), and this apoptosis was significantly suppressed in S100P-overexpressing cells (Fig. 8C). Celecoxib-induced expression of CHOP mRNA was also suppressed in S100Poverexpressing cells (Fig. 8D). These results suggest that S100P induction inhibits celecoxib-induced apoptosis through suppression of CHOP expression. We further tested this idea using siRNA for S100P and an 80 M concentration of celecoxib. Transfection of S100P siRNA decreased the expression of S100P mRNA (Fig.  8E) and S100P protein (Fig. 8F) in the presence or absence of celecoxib (80 M). This transfection stimulated celecoxib-induced apoptosis but not its basal level (Fig. 8G). . AGS cells were transfected with siRNA for S100P (siS100P) or non-silencing siRNA (ns), with the total amount of siRNA fixed at 1 g. After 24 h the cells were incubated with the indicated concentrations of celecoxib or staurosporine for a further 6 h (E), 12 h (F), or 8 h (G). S100P or CHOP mRNA expression was monitored and expressed as described in the legend of Fig.  1 (A, D, and E). S100P expression was monitored as described in the legend of Fig. 2 (B and F). Apoptosis was determined by fluorescence activated cell sorting, as described under "Experimental Procedures" (C, G, and H). Values are the mean Ϯ S.D. (n ϭ 3). *, p Ͻ 0.05; **, p Ͻ 0.01; n.s., not significant.
To examine the specificity of this anti-apoptotic effect of S100P, the apoptosis induced by staurosporine was compared between S100P-overexpressing cells and the mock transfectant control cells. As shown in Fig. 8C, staurosporine-induced apoptosis was indistinguishable between S100P-overexpressing cells and control cells. Furthermore, transfection of S100P siRNA did not affect this outcome (Fig. 8G). These results suggest that the suppression of apoptosis by overexpression of S100P is specific for apoptosis induced by ER stress responseinducing drugs.
Finally, we also tested the contribution of the extracellular S100P signaling to its inhibitory effect on celecoxib-induced apoptosis using cromolyn. Cromolyn did not affect the expression of S100P mRNA in the presence or absence of celecoxib (Fig. 8A). As shown in Fig. 8H, treatment of cells with cromolyn had no effect on celecoxib-induced apoptosis, suggesting that extracellular mechanisms are not involved in the inhibitory effect of S100P in this situation.

DISCUSSION
In this study we have shown that various NSAIDs, including celecoxib, a drug that is clinically used for cancer therapy, upregulate the expression of S100P in cultured cancer cells. Given that celecoxib (a COX-2 selective NSAID) up-regulated the expression of S100P in Kato III cells that lack COX-2 mRNA expression and that exogenous application of PGE 2 did not affect this up-regulation, it appears to occur independent of COX inhibition, as does the up-regulation of GRP78 and oxygen-regulated protein 150 by the same drug (15,16). S100P has previously been reported to be overexpressed in tumors clinically isolated from various tissue types, and expression of S100P has been shown to stimulate the aggressiveness of cancer cells (18 -23, 25, 26, 28, 48, 49). Thus, up-regulation of S100P could play an important role in the anti-tumorigenic activity of NSAIDs (see below).
Although the expression of S100P in clinically isolated tumors has been well described, little is known about the effect of anti-tumor drugs on the expression of S100P. Only bifunctional alkylating agents and retinoic acid have been shown to up-regulate S100P expression, although the mechanism governing this up-regulation remains unknown (18,32). Based on the following combination of results from this study, we strongly suggest that celecoxib up-regulates the expression of S100PthroughanATF4-mediatedERstressresponse;celecoxibdependent up-regulation of S100P expression was suppressed by siRNA for ATF4, other ER stress response-inducing chemicals (thapsigargin and tunicamycin) also up-regulated the expression of S100P, and deletion of the ATF4 binding consensus sequence (AARE) in the S100P gene promoter resulted in inhibition of celecoxib-dependent activation of its promoter activity. ER stressors phosphorylate protein kinase R-like ER kinase located in the ER membrane, which in turn phosphorylates eukaryotic initiation factor-2␣, leading to induction of ATF4 expression (50). We have previously reported that celecoxib stimulates the phosphorylation of both protein kinase R-like ER kinase and eukaryotic initiation factor-2␣ and induces the expression of ATF4, all of which are inhibited in the presence of the intracellular Ca 2ϩ chelator, BAPTA-AM (15).
In this study we have shown that BAPTA-AM also inhibits the celecoxib-mediated up-regulation of S100P expression. We have also demonstrated that NSAIDs could cause permeabilization of cytoplasmic membranes, resulting in an increase in intracellular Ca 2ϩ due to stimulation of the influx of extracellular Ca 2ϩ (42,51). Therefore, it seems that celecoxib up-regulates the expression of S100P mRNA through permeabilization of cytoplasmic membranes, an increase in the intracellular Ca 2ϩ level, phosphorylation of both protein kinase R-like ER kinase and eukaryotic initiation factor-2␣, and induction of ATF4 expression. As far as we are aware, this is the first indication of the mechanism underlying transcriptional regulation of S100P expression by anti-tumor drugs, information which should prove valuable in increasing our understanding of the modes of action of other anti-tumor drugs, and the role of S100P overexpression in tumors in vivo. For example, because it has been suggested that retinoic acid induces the ER stress response (52), retinoic acid may induce the expression of S100P via ER stress response.
It is well known that S100P stimulates the aggressiveness of cancer cells in various ways (including inhibition of apoptosis and stimulation of growth and invasion of cancer cells) and that NSAIDs, especially celecoxib, display their anti-tumor activity by exerting the opposite effects. Therefore, understanding the role of up-regulation of S100P expression in the anti-tumor activity of NSAIDs is important if we are to apply these drugs to cancer therapy. Here we have demonstrated that celecoxib-mediated inhibition of growth and invasion and induction of apoptosis are suppressed in S100P-overexpressing cells, and celecoxib-mediated inhibition of growth and induction of apoptosis were stimulated by transfection of cells with S100P siRNA. These results suggest that S100P induced by celecoxib decreases the anti-tumor activity of the drug. In the case of apoptosis, we have also shown that S100P suppresses the celecoxib-induced expression of CHOP, an ER stress-induced transcriptional factor with apoptosis-inducing ability. Furthermore, expression of S100P did not affect staurosporine-induced apoptosis. These findings suggest that S100P exerts a protective effect against accumulation of unfolded proteins in the ER, resulting in suppression of the ER stress response and apoptosis. Because some proteins of the S100 family have been reported to function as molecular chaperones, it is possible that S100P acts as ER chaperones to protect cells from ER stressors. We also found that expression of S100P activates MMP-9 activity and suppresses the inhibitory effect of celecoxib on this activity, suggesting that MMP-9 is involved in the effect of S100P on celecoxib-mediated inhibition of cell invasion. This is likely due to the fact that S100P activates ERK and NF-B (25,26), both of which are known to activate MMP-9 (45)(46)(47). S100P functions through both intracellular (example, Ca 2ϩdependent binding to ezrin) and extracellular (binding to RAGE and resulting activation of ERK and NF-B) mechanisms. Here we used cromolyn to test which pathway is dominant for the inhibitory effect of S100P on the anti-tumorigenic potential of celecoxib. Cromolyn not only inhibited the growth and invasion of S100P-overexpressing cells (but not control cells) but also stimulated the celecoxib-mediated inhibition of cell growth and invasion, suggesting that celecoxib-induced S100P and NSAIDs FEBRUARY 13, 2009 • VOLUME 284 • NUMBER 7 S100P stimulates cell growth and invasion through the extracellular mechanisms. However, cromolyn had no effect on celecoxib-induced apoptosis, suggesting that the protective effect of S100P in this situation is mediated through the intracellular mechanism. In contrast, it has recently been reported that cromolyn stimulates gemcitabine-induced apoptosis (27). Thus, the mechanism governing the inhibitory effect of S100P on apoptosis appears to differ depending on whether celecoxib or gemcitabine is the inducing agent.
Resistance to chemotherapy is one of the major obstacles facing effective cancer therapy. From this point of view, overexpression of S100P in tumors is a significant problem, particularly as a correlation has been reported between the expression level of S100P and chemoresistance (53). Because of poor vascularization, solid tumors usually exist under conditions of glucose starvation and hypoxia, which causes induction of the ER stress response, with overexpression of ER chaperones being reported in various types of tumors (54 -56). In this study we have shown that S100P can be induced through the ER stress response. Therefore, overexpression of S100P in tumors in vivo may be mediated via this mechanism in addition to the previously proposed mechanism, hypomethylation of the S100P gene (57,58). Furthermore, our finding that overproduction of S100P makes cancer cells resistant to celecoxib is of considerable importance if considering the use of this drug as a chemotherapeutic agent; it seems that not only constitutive overproduction of S100P in tumors but also S100P induced by celecoxib can render them chemoresistant to the drug. We, therefore, propose that an inhibitor of S100P may prove to be clinically efficacious by making cancer cells more responsive to celecoxib and other anti-tumor agents with the ability to induce ER stress response.