Potent Prostaglandin A1 Analogs That Suppress Tumor Cell Growth through Induction of p21 and Reduction of Cyclin E*

Although the cyclopentenone prostaglandin A1 (PGA1) is known to arrest the cell cycle at the G1 phase in vitro and to suppress tumor growth in vivo, its relatively weak activity limits its usefulness in cancer chemotherapy. In an attempt to develop antitumor drugs of greater potency and conspicuous biological specificity, we synthesized novel analogs based on the structure of PGA1. Of the newly synthesized analogs, 15-epi-Δ7-PGA1 methyl ester (NAG-0092), 12-iso-Δ7-PGA1 methyl ester (NAG-0093), and ent-Δ7-PGA1methyl ester (NAG-0022) possess a cross-conjugated dienone structure around the five-member ring with unnatural configurations at C(12) and/or C(15) and were found to be far more potent than native PGA1 in inhibiting cell growth and causing G1arrest in A172 human glioma cells. These three analogs induced the expression of p21 at both RNA and protein levels in a time- and dose-dependent fashion. Kinase assays with A172 cells treated with these analogs revealed that both cyclin A- and E-dependent kinase activities were markedly reduced, although cyclin D1-dependent kinase activity was unaffected. Immunoprecipitation-Western blot analysis showed that the decrease in cyclin A-dependent kinase activity was due to an increased association of p21 with cyclin A-cyclin-dependent kinase 2 complexes, whereas the decrease in cyclin E-dependent activity was due to a combined mechanism involving reduction in cyclin E protein itself and increased association of p21. Thus, these newly synthesized PGA1analogs may prove to be powerful tools in cancer chemotherapy as well as in investigations of the structural basis of the antiproliferative activity of A series prostaglandins.

Prostaglandins of the A series (PGAs) 1 contain an ␣,␤-unsaturated carbonyl group at the five-member ring and are derived from E type prostaglandins by dehydration in plasma or aqueous solution (1). PGAs have been reported to have antiprolif-erative activity in vitro (2)(3)(4)(5)(6)(7)(8) as well as antitumor activity in vivo (2, 7, 9 -11), and although these properties suggest that PGAs would be potentially useful in chemotherapy of malignant tumors, their relatively weak activity and extreme instability in plasma limit their clinical application.
In order to develop more stable and potent PGA analogs, it appeared to be important to decipher the structural features contributing to the stability and antiproliferative activity of A type PGs. A type PGs are conjugated with glutathione in the cells, and the resultant PG-glutathiones are immediately exported from the cells through the ATP-dependent glutathione S conjugate export pump (MRP/GS-X pump) (12), thereby reducing the intracellular accumulation of PGs. The GS-X pump was first reported to eliminate cytotoxic drugs from tumor cells and to play an important role in developing drug resistance in tumor cells (13). Thus, it is necessary to develop PGA 1 analogs that are not readily conjugated with glutathione.
We previously demonstrated that the cross-conjugated dienone PGs, including native ⌬ 12 -PGJ 2 and an artificial analog ⌬ 7 -PGA 1 methyl ester, reveal much more potent antiproliferative effects than simple enone-type PGs, such as PGA 1 (12). Although PGs containing a cross-conjugated dienone structure still react reversibly with glutathione to form equilibrium mixtures of free PG and PG-glutathione conjugate in the cell, this reaction biased the equilibrium reaction with intracellular glutathione to the formation of enone-conjugated forms, accumulating free PGs in the cell (12). Thus, the eminent potency of dienone PGs over simple enone PGs may be explained by the difference in the extrusion mechanism. In accordance with this concept, we have demonstrated that the antiproliferative effect of ⌬ 7 -PGA 1 methyl ester is significantly enhanced in the presence of an MRP/GS-X pump inhibitor. 2 Cell cycle progression in eukaryotes is controlled by specific activation and subsequent inactivation of cyclin-dependent kinases (Cdks) (14). Two distinct families of Cdk inhibitors (CKIs) have been described in mammalian cells (15). One of these is the Cip/Kip family, which includes the structurally related proteins p21, p27, and p57. Members of this family inhibit a variety of Cdk activities in vitro (16 -24). The other family of CKIs is the INK4 family, the members of which specifically inhibit cyclin D-dependent kinases in vitro (25)(26)(27)(28)(29). It has recently been reported that treatment of breast carcinoma cells with PGA 2 leads to G 1 arrest in association with increased expression of p21 (30), suggesting that PGA 2 inhibits tumor cell growth through the induction of p21. We also found that ⌬ 7 -PGA 1 methyl ester induces p21 via a p53-independent * This work was supported in part by Grants-in-Aid for Scientific Research on Priority Areas 09273104 (to M. N.) and 09273102 (to M. S.) and by Grant-in-Aid for Scientific Research (A) 08408023 (to M. S.) from the Ministry of Education, Science, Sports and Culture of Japan. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
In this study, we examined the ability of PGA 1 analogs that contain a cross-conjugated dienone unit to inhibit the proliferation of A172 human glioma cells at molecular level. The results indicate that three analogs, 15-epi-⌬ 7 -PGA 1 methyl ester (NAG-0092), 12-iso-⌬ 7 -PGA 1 methyl ester (NAG-0093), and ent-⌬ 7 -PGA 1 methyl ester (NAG-0022), are far more potent than native PGA 1 in arresting cells at the G 1 phase through a combined mechanism involving the induction of p21 and the suppression of cyclin E.

MATERIALS AND METHODS
PGs and Their Analogs-PGA 1 , PGA 2 (Sigma), PGJ 2 , 15-deoxy-⌬ 12,14 -PGJ 2 (Cayman Chemical Company), and PGA 1 analogs were dissolved in absolute ethanol, and dilutions were made in Dulbecco's modified essential medium. The final concentration of ethanol did not exceed 0.1% or affect cell growth. All manipulations of these compounds were performed under low lighting conditions. The code names and chemical structures of these compounds are shown in Fig. 1. Detailed methods for the design and synthesis of PGA 1 analogs are described elsewhere. 3 Cell Culture-Human glioma A172 cells were cultured in Dulbecco's modified essential medium (Life Technologies, Inc.) supplemented with 10% heat-inactivated fetal bovine serum (Life Technologies, Inc.) and penicillin-streptomycin (Life Technologies, Inc.) under standard tissue culture conditions. Cell numbers were counted with a cell counter (Coulter).
Cell Cycle Analysis-Cell cycle distribution was analyzed by flow cytometry. Briefly, 2 ϫ 10 6 cells were harvested in PBS containing 2 mM EDTA, washed once with PBS, and fixed in PBS containing 0.2% paraformaldehyde for 30 min. Fixed cells were washed once in PBS and permeabilized for 30 min with 0.2% Tween 20 and 1 mg/ml RNase A in PBS. Cells were then washed once in PBS and stained with 50 g/ml of propidium iodide (Boehringer Mannheim). The stained cells were analyzed by flow cytometry (Coulter).
Northern and Western Blot Analysis-Total cellular RNA was extracted from A172 cells using RNA isolation reagent (Isogen, Nippon Gene Co., Tokyo, Japan) according to the manufacture's instructions. RNA (20 g of each) was denatured, separated by electrophoresis on 1.0% agarose-formaldehyde gels, and transferred onto nylon membranes (Micron Separations, MA). p21 or cyclin E cDNA was labeled with [␣-32 P]dCTP using a random primer labeling kit (Amersham Pharmacia Biotech), and the hybridization signal was quantified by means of a Fuji Imaging Analyzer BAS-1500. GAPDH mRNA was used as an internal control for variations in loading and transfer efficiency among samples.
Statistical Analysis-Statistical significance was determined by the analysis of variance. Scheffe's F test procedure for multiple pairwise post-hoc comparisons of means was implemented, and a p value of less than 0.05 was considered statistically significant.

RESULTS
Chemical Structure of the Synthetic PGA 1 Analogs- Fig. 1 shows the chemical structures of the synthetic PGA 1 analogs identified in this study as more potent antitumor compounds than native PGs. All of the analogs possess a cross-conjugated dienone unit and a methyl ester form at C(1), whereas native PGA 1 and PGA 2 are simple enone compounds and contain no methyl ester. Some of the synthetic PGs have unnatural configurations at C(12) and C (15).
Potent Antiproliferative Activity of PGA 1 Analogs-The synthetic PGA 1 analogs were first tested for their ability to inhibit cell proliferation, using PGA 1 , PGA 2 , and PGJ 2 as references. A172 cells (from a glioblastoma cell line) were treated with each of the test compounds at a 5 M concentration, and changes in the cell number were examined over a period of 72 h. As shown in Fig. 2, NAG-0022, NAG-0092, and NAG-0093 were most potent in suppressing the proliferation of A172 cells at 48 and 72 h (p Ͻ 0.05). NAG-0026 also displayed a substantial antiproliferative activity (p Ͻ 0.05). PGA 1 , PGA 2 , and PGJ 2 had almost no effect on the cell growth at 5 M ( Fig. 2 and data not shown). G 1 Arrest by PGA 1 Analogs-In order to examine the mechanism(s) of growth suppression by PGA 1 analogs and also to rule out the possibility of nonspecific cytotoxicity, we next determined the cell cycle distribution of A172 cells after treatment of these cells with the various PGA 1 analogs, using a fluorescence-activated cell sorter (Fig. 3). At a concentration of 5 M, NAG-0022, NAG-0092, and NAG-0093 were most potent in arresting cells at the G 1 phase (66.7, 74.9, and 71.0%, respectively) as compared with vehicle-treated cells (42.2%), whereas the other compounds showed only a modest effect on cell cycle distribution ( Fig. 3 and data not shown). The growth inhibitory activity of the PGA 1 analogs (Fig. 2) correlated well with their ability to cause G 1 arrest (Fig. 3), suggesting that most of these analogs inhibit cell growth in this manner.
Induction of p21 Expression by PGA 1 Analogs-It has been reported that PGA 2 arrests the cell cycle at the G 1 phase through the induction of p21 CKI at high concentrations (25-36 M) (30,42). To determine whether this is also the case with PGA 1 analogs, we examined changes in p21 expression after using the analogs to treat the A172 cells. As shown in Fig. 4, the baseline level of p21 mRNA expression of these cells was low, and NAG-0022, NAG-0092, and NAG-0093 caused a dosedependent increase in the steady-state mRNA level, with a clear-cut effect being observed at 1.0 M and a maximal response at 5.0 -10 M (approximately 7-fold as compared with vehicle-treated cells). Only a modest increase in p21 mRNA was observed with NAG-0026, and PGA 1 , PGA 2 , and ⌬ 12 -PGJ 2 had little or no effect at 5 M (Fig. 4).
The time course of the effect of NAG-0092 on p21 mRNA is shown in Fig. 5. Induction of p21 mRNA expression occurred as early as 2 h and reached a maximal level at 8 h after treatment with 5 M NAG-0092. The induction of p21 protein was also examined by Western blot analysis. As shown in Fig. 6, NAG-0092 and NAG-0093 at 5 M markedly increased the protein level of p21, whereas PGA 1 and PGA 2 had only a modest effect. These results correlated well with those of Northern blot analysis (Fig. 4). Taken together with the findings that PGA 1 analogs did not induce the expression of other known CKIs, including p27, p57, p15, and p16 (data not shown), it is suggested that PGA 1 analogs arrest the cell cycle at the G 1 phase, at least in part through the induction of p21 CKI. PGA 1 Analogs Suppress Cyclin A-and Cyclin E-but not Cyclin D1-dependent Kinase Activities-To further elucidate the mechanism(s) by which PGA 1 analogs arrested the cell cycle at G 1 , we examined cyclin E-, A-and D1-dependent kinase activities, all of which are necessary for G 1 progression and G 1 /S transition. As shown in Fig. 7A, cyclin A-and E-dependent kinase activities were almost completely inhibited at 24 h after treatment with 5 M NAG-0092, whereas cyclin D1-dependent kinase activity was not affected (data not shown).
To determine whether the suppression of cyclin A-and E-dependent kinase activities by the PGA 1 analog was due to the induction of p21 and its association with the Cdk complexes, we analyzed cyclin A-Cdk2 and cyclin E-Cdk2 complexes by IP-Western analysis. As shown in Fig. 7B, the amount of Cdk2 complexed with cyclin A did not change after treatment with PGA 1 or NAG-0092. In contrast, the amount of p21 in the cyclin A-Cdk2 complex was markedly increased after treatment with NAG-0092, but only modest increases were noted with PGA 1 (Fig. 7B). Taken together, the increased association of p21 with cyclin A-Cdk2 complexes appeared to be important for the inhibition of kinase activity. On the other hand, analysis of cyclin E-Cdk2 complexes revealed that the amount of Cdk2 complexed with cyclin E was markedly reduced after treatment with NAG-0092 (Fig. 7B). Northern blot analysis demonstrated that NAG-0092 at 5 M caused a marked decline in the expression of cyclin E mRNA (Fig. 7C), and Western blotting confirmed that NAG-0092 induced a decrease in the cyclin E protein level (Fig. 7D). Although the level of cyclin E protein was markedly reduced after treatment with NAG-0092, the amount of p21 in the anti-cyclin E immunoprecipitates did not decrease (Fig. 7B), suggesting that the relative amount of p21 complexed with cyclin E increased. It is likely, therefore, that the decrease in cyclin E-dependent kinase activity is due to a combined mechanism involving decline in cyclin E protein itself and increased p21 level in cyclin E complexes.
Unexpectedly, cyclin E-but not cyclin A-dependent kinase activity was enhanced rather than suppressed by PGA 1 treatment (Fig. 7A). Taken together with our observations that the amount of Cdk2 associated with cyclin E decreased after PGA 1 treatment whereas that of p21 in the complexes remained constant (Fig. 7B), it is conceivable that the relative amount of cyclin E-Cdk2-p21 ternary complexes was increased by PGA 1 . In light of the reported findings that p21 can promote the association of Cdk with cyclins and thus stabilize cyclin-Cdk complexes (43), the apparent stimulation of cyclin E-dependent kinase activity by PGA 1 may be explained by this novel function of p21 as an assembly factor rather than as a Cdk inhibitor.
Interestingly, although cyclin E mRNA did not change after treatment with PGA 1 (Fig. 7C), the level of cyclin E protein was significantly reduced (Fig. 7D), which may also be related to enhanced cyclin E-dependent kinase activity after PGA 1 treatment. Recently, it was reported that cyclin E could be phosphorylated by Cdk2 and that phosphorylated cyclin E is readily degraded by ubiquitine-proteosome pathway (44). Therefore, enhanced cyclin E-Cdk2 activity may result in phosphorylation of cyclin E and its rapid degradation, leading to the reduction of cyclin E protein without changes in its mRNA level. Total RNA was isolated from A172 cells after treatment for 24 h with increasing concentrations (0, 1, 2.5, 5, and 10 M) of the indicated compounds and was then subjected to electrophoresis on a 1% formaldehyde-agarose gel. Northern blotting was performed as described under "Materials and Methods" using human p21 cDNA as a probe. GAPDH mRNA served as an internal control.

DISCUSSION
In the present study, we tested a series of PGA 1 analogs for their activities in inhibiting the growth of A172 glioblastoma cells and found that three of these were far more potent than native PGA 1 , PGA 2 , and PGJ 2 . Although it has been reported that some PGs of the A and J series are capable of suppressing tumor cell growth both in vitro (2)(3)(4)(5)(6)(7)(8) and in vivo (2, 7, 9 -11), their activities are rather modest, presumably due to their intracellular instability, necessitating the use of high doses to obtain substantial antitumor effects. We set out to develop PG analogs of the A and J series that show increased stability, because we considered that some of these compounds might prove to be more potent in inhibiting tumor cell growth than the native compounds. In parallel with this task, we attempted to elucidate the molecular mechanism(s) by which our novel analogs inhibit cell growth.
There is accumulating evidence that active export of anticancer drugs from cells is one of the major mechanisms of drug resistance. Recent studies of the multidrug resistance phenotype of tumor cells have led to the discovery of P-glycoprotein, which mediates the efflux of anticancer drugs, such as doxorubicin, vincristine, and taxol (45,46). More recently, another type of drug transporter (GS-X/MRP pump) has been identified as a mediator of glutathione-associated drug resistance. It has been reported that A type PGs are effectively conjugated with glutathione in the cells (12), raising the possibility that GS-X/ MRP pump may play an important role in excluding the PGs from the cells. This concept was further supported by the observation that HL-60 cells overexpressing functional GS-X/ MRP pump reveal far more resistance to the antiproliferative effect of ⌬ 7 -PGA 1 methyl ester than control HL-60 cells.
We have previously shown that the cross-conjugated dienone unit, as compared with the simple enone unit in native PGA, protects PGs against removal from the cytoplasm by GS-X/ MRP pump (12) and allows PGs to bind more stably to target proteins in the nucleus or cytosol, suggesting that the crossconjugated dienone unit stabilizes PGs within the cell. However, in terms of clinical application, it was required to develop more stable and potent PGA 1 analogs.
Of the newly synthesized analogs that we synthesized, NAG-0092, NAG-0093 and NAG-0022, which showed the higher antiproliferative activity for A172 cells than NAG-0026 (⌬ 7 -PGA 1 methyl ester) (Figs. 1 and 2), possess both a hydroxy group at C(15) and a double bond at C(13)-C (14). Thus, the double bond and hydroxy group in the -chain are essential for the potent antiproliferative activities of PGs for these cells. The double bond is thought to fix the conformation of PGs, and the hydroxy group constitutes a hydrogen bonding, allowing sufficient interaction with the target molecule. Although native PGA 2 has been reported to inhibit cell growth at 25-36 M (30, 42, 47), almost no growth inhibitory effect was observed at the lower concentrations (up to 10 M) used in the current study.
FACScan analysis revealed that PGA 1 analogs arrested the cell cycle at the G 1 phase (Fig. 3). Interestingly, although NAG-0026 substantially inhibited cell growth, it did not cause accumulation of cells at G 1 (data not shown), raising the possibility that growth inhibition by this particular analog may be due to its nonspecific toxicity.
It has recently been reported that PGA 2 inhibits the G 1 cyclin-dependent kinase activities through the induction of p21 Cell lysates in IP-kinase buffer were immunoprecipitated with anti-cyclin A or anti-cyclin E antibodies as described under "Materials and Methods." The resultant immunoprecipitates were assayed for their kinase activity using histone H1 as a substrate. None indicates a control experiment in which cells were left untreated. B, analysis of cyclin A and E complexes by IP-Western blotting. The lysates of A172 cells in IP-kinase buffer were prepared under the same conditions as described in A. After immunoprecipitation with anti-cyclin A or anti-cyclin E antibodies, the immunoprecipitates were subjected to Western blotting using anti-Cdk2 or anti-p21 antibodies. C, reduction in cyclin E mRNA expression by the PGA 1 analog NAG-0092. Total RNA was isolated from A172 cells treated for 24 h with increasing concentrations of either PGA 1 or NAG-0092. Total RNA (20 g/lane) was subjected to electrophoresis on a 1% formaldehydeagarose gel, and Northern blotting was performed as described under "Materials and Methods" using human cyclin E cDNA as a probe. The expression of GAPDH mRNA was used as an internal control. D, reduction in cyclin E proteins by the PGA 1 analog NAG-0092. A172 cells were treated with 5 M of PGA 1 or NAG-0092 for 24 h and then were lysed in IP-kinase buffer. The cell lysates (100 g/lane) were separated by SDS-polyacrylamide gel electrophoresis (10% polyacrylamide gel) and subjected to Western blotting using anti-cyclin A or anti-cyclin E antibody. (30). We showed by both Northern and Western blot analyses that the PGA 1 analogs with the greatest ability to inhibit cell growth markedly induced the expression of p21 (Figs. 4 and 6). Because p21 expression was induced within 2 h after treatment with PGA 1 analogs (Fig. 5) and because the induction was enhanced in the presence of cycloheximide (data not shown), it is plausible that the induction of p21 expression resulted from the direct genomic action of the PGA 1 analogs.
Based on our analysis of cyclin-Cdk complexes and their activities, PGA 1 analogs seem to inhibit tumor cell growth through two distinct mechanisms: the induction of p21 and the suppression of cyclin E expression (Fig. 7). Inactivation of cyclin A-Cdk2 activity by PGA 1 analogs appears to occur mainly through the former mechanism, whereas suppression of cyclin E-Cdk2 activity appears to involve both increased p21 and reduced cyclin E expressions. Taken together with the findings that PGA 1 analogs did not induce the expression of other known Cdk inhibitors or cyclins (data not shown), it is concluded that the combined effects on the expression of p21 and cyclin E are responsible for the growth suppressive function of PGA 1 analogs. Although the molecular and cellular mechanisms by which PGA 1 analogs induce p21 and suppress cyclin E expression remain elusive, the induction of p21 and the suppression of cyclin E by PGA 1 analogs may involve a transcriptional mechanism because the PGA 1 analogs are incorporated into nuclei without further metabolism (12).
Members of the J 2 series of PGs have been reported to have a unique spectrum of biological effects, including the inhibition of cell cycle progression, the suppression of viral replication, and the stimulation of osteogenesis (11). Results of previous studies have indicated that the activation of peroxisome proliferator-activated receptor ␥, a member of the nuclear receptor superfamily, is mediated by metabolites of J type PGs (48,49). Although it has been demonstrated that PGA 2 can also activate peroxisome proliferator-activated receptor ␥, its potency appears to be relatively weak (50). In view of our present results that both PGJ 2 and 15-deoxy-⌬ 12,14 -PGJ 2 , putative ligands for peroxisome proliferator-activated receptor ␥, are not effective in the induction of p21 expression in A172 cells, PGs of the A series are likely to serve as ligands for an as yet unidentified member(s) of the nuclear receptor superfamily distinct from peroxisome proliferator-activated receptor ␥.
In conclusion, we have identified novel PGA 1 analogs that strongly suppress cellular proliferation through their combined effects on p21 CKI and cyclin E. Our results may help to decipher the structural features essential to the antiproliferative activity of the A and J series of PGs. Although it remains to be determined whether the increased potency of our analogs reflects an increased stability within cells or an increased binding affinity to proteins, these analogs may prove to be powerful tools for elucidating the molecular functions of the A series of PGs in both in vitro and in vivo studies.