PGA1-induced apoptosis involves specific activation of H-Ras and N-Ras in cellular endomembranes

The cyclopentenone prostaglandin A1 (PGA1) is an inducer of cell death in cancer cells. However, the mechanism that initiates this cytotoxic response remains elusive. Here we report that PGA1 triggers apoptosis by a process that entails the specific activation of H- and N-Ras isoforms, leading to caspase activation. Cells without H- and N-Ras did not undergo apoptosis upon PGA1 treatment; in these cells, the cellular demise was rescued by overexpression of either H-Ras or N-Ras. Consistently, the mutant H-Ras-C118S, defective for binding PGA1, did not produce cell death. Molecular analysis revealed a key role for the RAF-MEK-ERK signaling pathway in the apoptotic process through the induction of calpain activity and caspase-12 cleavage. We propose that PGA1 evokes a specific physiological cell death program, through H- and N-Ras, but not K-Ras, activation at endomembranes. Our results highlight a novel mechanism that may be of potential interest for tumor treatment.

The mammalian genome contains three ras genes that encode the 21-kDa proteins H-Ras, N-Ras, and K-Ras with their two isoforms, K-Ras4A and K-Ras4B, which are generated from two alternative fourth exons. Ras proteins are small GTPases that act as molecular switches connecting a wide spectrum of extracellular signals from cell-surface receptors to intracellular pathways to control cell proliferation, differentiation, senescence, and death. 1 Ras proteins are activated by guanine nucleotide-exchange factors, which promote the exchange of guanosine diphosphate (GDP) for guanosine triphosphate (GTP), resulting in a conformational change of the tertiary structure of Ras and exposing its effector loop to interacting partners. The intrinsic GTPase activity of Ras proteins stimulated by GTPase-activating proteins restores the GDP-bound state and terminates Ras signaling. Active GTP-bound Ras interacts with effector proteins that modulate different signaling pathways to generate specific biological outcomes. As Ras proteins are ubiquitously expressed (except K-Ras4A) 2 and share a high degree of sequence homology and a large number of molecular activators, it was long assumed that their role was redundant. However, functional redundancy is not complete, as demonstrated by embryonic lethality in K-Ras knockouts 3,4 and the demonstration that the three Ras proteins have specific roles according to Ras isoform-dependent subcellular compartmentalization. 5 This observation implies that biochemical and biophysical aspects of specific subcellular sites determine both the Ras isoform and the set of effectors that could be recruited, thus generating different molecular and biological outputs. 6 Cyclopentenone prostaglandins (CyPGs) are eicosanoids with a varied spectrum of biological activity, including anti-inflammatory and antitumor effects, induction of oxidative stress, modulation of heat-shock response (HSP), and antiviral activity. 7,8 They are thought to originate from the free radical-induced peroxidation of arachidonic acid (isoprostane pathway) 9 and the dehydration of prostaglandins. 10 CyPG contain an α,β-unsaturated carbonyl group in the cyclopentane ring that favors the formation of Michael adducts with sulfhydryl groups of proteins. 11 This mechanism is responsible for many of the biological properties of these compounds.
However, the variety of cellular responses to CyPG appears to be cell type-specific and concentration-dependent; indeed, cytoprotective and cytotoxic responses have been associated with low and high concentrations of CyPG, respectively. 12,13 In this regard, we reported that H-Ras is a target for the addition of a typical CyPG, the 15-deoxy-12,14-prostaglandin J 2 (15d-PGJ 2 ), and that this effect is associated with specific activation of H-Ras-dependent pathways and increased proliferation and inhibition of apoptosis in NIH3T3 fibroblasts 14 and MCA3D keratinocytes. 15 Nevertheless, members of the prostaglandin A and J series act as potent inhibitors of some human tumors both in vitro and in vivo, by inducing cell cycle arrest and apoptosis. 7,10 In endothelial cells, activation of peroxisome proliferator-activated receptor (PPAR) receptors by 15d-PGJ 2 induces nuclear localization of receptors and caspase-mediated apoptosis, 16 and both 15d-PGJ 2 and cyclopentenone prostaglandin A1 (PGA 1 ) induce apoptosis in AGS cells by a PPAR-independent mechanism. 17 Although the role of CyPG as inducers of apoptosis is well documented, the relationship between Ras activation by CyPG and triggering of apoptosis, or proliferation, is not fully understood. We have explored the molecular mechanisms underlying PGA 1 -induced apoptosis. We found that PGA 1 induces apoptosis in mouse fibroblasts by specific binding to and activation of H-and N-Ras at endomembranes, and that this process requires activation of extracellular signal-regulated kinases (ERKs). A detailed analysis revealed that the mechanism of cell death induced by PGA 1 requires ERKmediated activation of calpain, an endoplasmic reticulum (ER) protease, and leads to a caspase-dependent apoptosis. Finally, we present evidence that the mechanism of apoptosis is shared by human cancer cells.

Results
PGA 1 induces apoptosis in MEFs. The CyPGs display potent antiproliferative activity in various cellular models. These compounds induce cell cycle arrest or apoptosis depending on the cell type and treatment conditions. 18,19 Synchronously starved NIH3T3 cells were incubated in the presence of PGA 1 to assess antiproliferative activity (according to the scheme shown in Supplementary Figure S1A). Cell number decreased sharply: no viable cells remained after 48 h of treatment, even in the presence of strong mitogenic signals such as fibroblast growth factor (FGF) (Figure 1a). Flow cytometry analysis of propidium iodide-stained cells revealed that~75% of PGA 1 -treated cells showed an increase in cells with hypodiploid DNA content after 24 h (Supplementary Figure S1B). In addition, DAPI (4′, 6-diamidino-2-phenylindole) staining also revealed apoptotic nuclear morphology after 24 h (Supplementary Figure S1C). Caspase cleavage was examined to verify that the effects of PGA 1 reflected apoptosis. Western blotting using antibodies that detect both procaspase and activated cleaved forms revealed the presence of cleaved caspase-12, -9, and -3 in cell lysates from PGA 1 -treated cells, as well as cleaved poly(ADP-ribose) polymerase (PARP), a typical target for activated caspase-3 ( Figure 1b). Moreover, both annexin V-binding analysis ( Figure 1c) and DNA laddering (Supple-mentary Figure S1D) indicated that this cell death was due to an apoptotic process. In agreement with these data, both caspase-3 activation and the increase in annexin V-positive cells were completely abolished by the pancaspase inhibitor Z-Asp-CH2-DCB (Figure 1d). Taken together, these results demonstrate that treatment of mouse fibroblast cells with PGA 1 leads to apoptosis through activation of a caspasedependent pathway.
PGA 1 -induced apoptosis requires expression of H-and N-Ras. PGA 1 -induced apoptosis has been reported in several cellular models, including cancer cells. 10 However, the precise molecular mechanism by which PGA 1 elicits cell elimination remains unclear. We have previously reported that PGA 1 binds to and activates H-Ras, N-Ras, and K-Ras to a similar extent. 20 As Ras proteins promote activation of the effector cascades that trigger apoptosis, we used a knockout approach to evaluate the potential role of these proteins in mediating PGA 1 -dependent apoptosis (Supplementary Figure S2A). Analysis of relevant apoptosis markers upon PGA 1 incubation in mouse embryonic fibroblasts (MEFs) for H-Ras − / − or N-Ras − / − did not reveal differences in caspase activation or in PARP proteolysis compared with wt MEFs. Interestingly, in H-Ras − / − /N-Ras − / − doubleknockout MEFs, PGA 1 failed to induce caspase activation and apoptosis (Figure 2a). Ectopic restitution of H-Ras or N-Ras expression in double-knockout cells caused cell death increase, activation of PARP, and caspase-3 by PGA 1 (Figure 2b and Supplementary Figure S2B). Given that PGA 1 -induced apoptosis requires expression of H-Ras or N-Ras, we reasoned that overexpression of these isoforms could synergize with the PGA 1 effects. Indeed, PGA 1 -induced dose-dependent increases in cleaved caspase-3 levels were much greater in NIH3T3 cell lines stably overexpressing ectopic AU5-H-Ras-wt or AU5-N-Ras-wt than in control cell lines ( Figure 2c, upper panel). In agreement with these data, treatment of Ras-overexpressing fibroblasts with PGA 1 markedly increased the population of annexin V-positive cells ( Figure 2c, lower panel). Moreover, ectopic expression of a dominant-negative mutant of H-Ras (H-Ras-N17) reduced caspase-3 activation levels upon stimulation with PGA 1 (Figure 2d). Taken together, these results show that signaling by H-Ras and N-Ras, but not K-Ras, are required to promote PGA 1 -induced apoptosis.
Ras-Cys118 residue is required for PGA 1 -induced apoptosis. Our previous studies showed that PGA 1 binds preferably to Cys118 in all three Ras proteins, and Cys184 and Cys181 in H-Ras. However, rather than Cys184 or Cys181, PGA 1 activates H-Ras by binding to Cys118. 20 As Cys181 is shared by N-Ras and H-Ras, whereas Cys184 is unique to H-Ras, we speculated that Ras-Cys118 was the main target for PGA 1 in mediating apoptosis. To determine the relative contribution of Cys118, double-knockout MEFs were transfected with H-Ras-wt, H-Ras-C118S, or H-Ras-C184S. As shown in Figure 3a and Supplementary Figure S2B, expression of H-Ras-wt or H-Ras-C184S induced cell death increase and caspase-3 activation in response to PGA 1 , whereas expression of H-Ras-C118S did not. Furthermore, analysis of transfected PGA 1 -stimulated MEFs showed that the C118S mutant did not increase the annexin V-positive cell population ( Figure 3b). These data provide strong evidence that Cys118 of H-and N-Ras is necessary for PGA 1 -induced apoptosis.

PGA 1 induces Ras activation in endomembranes.
To examine whether activation of Ras by PGA 1 could be affected by intracellular localization, we measured the Ras-GTP levels induced by PGA 1 in HeLa cells transiently transfected with H-Ras targeted either to the plasma membrane or to endomembrane compartments. We found that M1-H-Ras-SS and, to a lesser extent, KDEL-H-Ras-SS, H-Ras constructs targeted to the ER and the Golgi apparatus, respectively, 21 were activated after stimulation by PGA 1 (Figure 4a). In contrast, LCK-H-Ras-SS and CD8-H-Ras-SS, which directed H-Ras specifically towards the plasma membrane (lipid rafts and bulk membrane, respectively 21 ) were not activated by PGA 1 . Moreover, as shown in Figure  To confirm that PGA 1 preferentially activated Ras on endomembranes, we studied the subcellular localization of PGA 1 -induced Ras activation in vivo using double fluorescence confocal microscopy analysis of CH7C17 Jurkat cells, a human T-cell line that does not express H-Ras, but activates endogenous N-Ras and ERK in response to PGA 1 or CD3 and apoptosis in response to PGA 1 (Supplementary Figures  S3B-D). For this purpose, we co-transfected CH7C17 cells with enhanced cyan fluorescent protein (CFP)-H-Ras, which distributes between the plasma membrane and the ER/Golgi complex, and YFP-RBD-Raf-1, a yellow fluorescent tracker of H-Ras activation in live cells. Whereas YFP-RBD-Raf-1 was localized throughout the cytoplasm of non-stimulated cells, stimulation with CD3 or PGA 1 increased colocalization of YFP-RBD-Raf-1 with CFP-H-Ras on endomembranes ( Figure 4c) and was higher than on the plasma membrane ( Figure 4d). Thus, these results strongly suggest that PGA 1 activates Ras proteins in endomembrane compartments.

PGA 1 induces calpain activation in fibroblasts.
Caspase-12 has been localized on the cytoplasmic side of the ER, and is activated by alterations of the ER homeostasis such as mobilization of intracellular calcium. As caspase-12 is activated in response to PGA 1 stimulation and the activation of caspase-12 may occur through ER stress-induced calpains, 22 we tested the possible involvement of these endopeptidases on the PGA 1 -induced apoptosis. Figure 5a shows a weak but sustained increase of intracellular calcium concentration upon PGA 1 treatment in wt MEFs; in contrast, the calcium levels displayed in H-Ras − / − /N-Ras − / − doubleknockout MEF cells were not significantly elevated. Western blotting analysis of PGA 1 -treated wt MEFs revealed that the levels of autoproteolytically cleaved calpain-1 were much greater than in H-Ras − / − /N-Ras − / − double-knockout MEFs, whereas the levels of calpain-2 remained unaltered ( Figure 5b). Furthermore, ALLN ((2S)-2-acetamido-4-methyl-N-((2S)-4-methyl-1-oxo-1-(((2S)-1-oxohexan-2-yl)amino)pentan-2-yl)pentanamide), a calpain specific inhibitor, completely blocked the calpain activity induced by PGA 1 in wt MEFs ( Figure 5c) and totally inhibited the proteolysis of both caspase-12 and calpain-1 and, in a lower extent, the activation of caspase-9 and -3 ( Figure 5d). Taken together, these data indicate that calpain activity is required for cleavage of caspase-12 induced by PGA 1 .
ERK activation is required for PGA 1 -dependent apoptosis. One of the most important signaling pathways downstream of Ras is the RAF-MEK-ERK cascade, and previous work has demonstrated that calpain is activated by ERK signaling. 23 As PGA 1 induced both activation of Ras and calpain proteins, we assessed the activation of these kinases downstream of Ras. Treatment of wt MEFs with PGA 1 induced activation of RAF, MEK, and ERK, although to a lesser extent than FGF (Figure 6a). Consistent with these results, PGA 1 treatment promoted significant activation of an Elk-1 luciferase reporter, although this activation was once again weaker than that induced by FGF (Figure 6b). Despite the weak and short duration of phospho-ERK levels (Supplementary Figures S4A and B), we speculated that this activation could be responsible for the apoptosis observed. To examine whether ERK activation was required for PGA 1induced apoptosis, the MEK1/2 inhibitor U0126 was added before PGA 1 treatment (~50 min) and left until the end of PGA 1 treatment. As shown in Figure 6c (and Supplementary Figure S4C), in wt MEFs the inhibitor completely blocked calpain-1 and caspase-12, -9 and -3 activation induced by PGA 1 . Moreover, both the calpain activity and Annexin V-positive cell number were markedly reduced in the presence of the MEK inhibitor upon PGA 1 stimulation (Figures 6d and e). Taken together, these data strongly suggest that PGA 1 promotes apoptosis through the activation of the RAF-MEK-ERK pathway.

PGA 1 also induces apoptosis in human cancer cells.
Apoptosis is the leading mechanism proposed to account for the antitumoral and antiproliferative effects of PGA 1 in cancer cells. 24 Indeed, we assayed PGA 1 -induced apoptosis in a panel of 14 human cell lines from different cancer types (Supplementary Figures S5A and S6). Although the sensitivity to PGA 1 showed differences between the distinct cancer cell lines, it remained very similar among lines of the same tumor type, with the exception of the non-small-cell lung cancer (NSCLC) lines. Thus, in this case, most H358 and H23 cells underwent apoptosis upon PGA 1 treatment, whereas A549, H522, and H2126 cells showed higher resistance to PGA 1 -induced cell death. Next, we examined whether PGA 1 treatment activated in human cancer cells the same mechanism than in MEFs. Western blotting and analysis of apoptotic markers in H358 and A549, K-RasV12mutated NSCLC lines with different sensitivity to PGA (Figure 7a), showed activation of calpain-1, caspase-9, and caspase-3 in H358 cells but not in A549 cells. Altogether, these data suggest that in human cancer cells the mechanism of PGA 1 -induced apoptosis may be mediated by a similar set of signaling proteins than in mouse embryo fibroblast cells.

Discussion
Although PGA 1 displays potent antiproliferative activity, a fundamental yet unanswered question is the underlined mechanism. The cyclopentane ring of CyPG posseses an α-, β-unsaturated carbonyl group that reacts with sulfhydryl groups in cysteine residues of cellular proteins by means of Michael's addition. 25 During recent years, several protein targets have been identified for CyPG, including NF-κB, AP1, Keap-1, and LKB1. [26][27][28] We have reported that Ras proteins can also be modified by the addition of 15d-PGJ 2 and PGA 1 ; in fact, the three Ras isoforms (H-Ras, N-Ras, and K-Ras4B) bind to and are activated by PGA 1.
14,20 Induction of apoptosis by PGA 1 occurs in different cell types, such as in endothelial cells, gastric epithelial cells, lung, prostate, and colon cancer cells. 16,17,29 In contrast, some cellular models suggest that PGA 1 protects against apoptosis through a mechanism that involves inhibition of NF-κB activation. 30 Our results show that PGA 1 induces apoptosis by a caspase-dependent mechanism that requires either H-Ras or N-Ras, but not K-Ras. PGA 1 -induced apoptosis is likely initiated by an ER/ Golgi complex pool of Ras and requires activation of ERK.
Indeed, the only MEFs that showed activation of caspase-3 after stimulation with PGA 1 were those that harbor the H-ras and/or the N-ras genes. Consistent with this observation, the ectopic overexpression of H-Ras or N-Ras in double-knockout MEFs was sufficient to rescue the apoptotic levels induced by PGA 1 . Moreover, NIH3T3 cells constitutively overexpressing ectopic H-Ras or N-Ras displayed synergistic enhancement of PGA 1 apoptotic effects, even at low concentrations. In addition, overexpression of a dominant-negative mutant of H-Ras (H-Ras-N17) in fibroblasts caused a significant reduction in caspase-3 activation, indicating that functional Ras signaling is essential for PGA 1 -induced apoptosis.
In this study, we have provided evidence that the PGA 1 activates Ras proteins in the ER and in the Golgi complex, but not at the plasma membrane (lipid rafts or bulk membrane). CFP-H-Ras in CH7C17 cells colocalized in vivo with YFP-RBD-Raf-1 in endomembrane systems upon stimulation with PGA 1 . A large body of evidence supports the possibility that H-Ras/N-Ras activation can occur in these organelles; nevertheless, the specific outcome of this activation remains debated. Thus, activated H-Ras targeting the Golgi apparatus has been reported to elicit strong activation of ERK and AKT, but weak activation of JNK, whereas the opposite effects were described when H-Ras targeted the ER. 31 Other studies, 32 however, which used the Golgi apparatus-tethering signal KDELr containing the mutation N193D that fixes it permanently to the Golgi apparatus, reported that activated H-Ras tethered to the Golgi apparatus (KDEL-H-RasV12-SS) only   Figure S4D).
Further studies are necessary to elucidate the transduction pathway downstream of ERK that is associated with caspase-3 cleavage upon stimulation with PGA 1 . Nevertheless, it has been reported that growth factors activate calpains downstream of ERK. 34 Our results further support a link between PGA 1 , ERK, and calpain activation, as the MEK inhibitor U0126 abolished ERK, calpain, and caspase activation and therefore PGA 1 -induced apoptosis, suggesting that calpains have an essential role in the apoptotic pathway induced by PGA 1 downstream of ERK activation. Although activation of caspase-12, a marker of ER stress, may occur through ER stress-induced calpains, 35 we do not know if the Ras activation induced by PGA 1 in endomembranes could trigger an ER stress response. CyPGs exert a variety of biological actions, most remarkably among them the stress response. PGA 1 is an important inducer of stress response by increasing the synthesis of HSPs and activating heat-shock transcription factors. 36 In rats with acute liver damage, pretreatment with PGE 1 and somatostatin inhibits apoptosis and alleviates ER stress by induction of HSP70 and BiP/ GRP78 (immunoglobulin-binding protein/), but suppression of CHOP (C/EBP homologous protein). 37 Indeed, Δ 12 -PGJ 2 and PGA 1 induce BiP gene expression through unfolded protein response (UPR). 38 BiP is a molecular chaperone that is synthesized constitutively, although accumulation of unfolded proteins in the ER changes in Ca 2+ levels, reducing environment and block of glycosylation induces transcription of the BiP gene. 39 Furthermore, ER stress can be induced by a Recent studies have demonstrated that the activation of K-Ras on the ER surface cooperates with Nox4 to initiate the UPR and autophagy and therefore prevent cell death and promote differentiation in HUVEC cells. 43 We found evidences supporting the existence of a PGA 1 -induced UPR. Indeed, upon prostaglandin stimulation, we observed an elevation in the levels of the proteins BiP and CHOP in wt MEFs and H358 cells ( Supplementary Figures S5B and S5C). These data strongly suggest that PGA 1 could trigger an ER stress response through H-and N-Ras activation. In summary, our study provides new insights into the molecular mechanism underlying the induction of apoptosis by PGA 1 . Our findings reveal a novel mechanism by which PGA 1 stimulates cell death as a consequence of specific activation of H-Ras and N-Ras in endomembranes (ER and the Golgi complex), whereas K-Ras was dispensable for this function (summarized in Figure 7b). These data also have implications for signal transduction and cancer, as the RAF-MEK-ERK cascade is the signaling pathway responsible for this cell death through ERK-induced calpain activity and subsequent caspase activation even in cells with deletion of the TP53 gene and that harbor oncogenic K-Ras as H358 cells. Our results highlight the relevance for physiological outcomes of the intensity, duration, and specific subcellular location of Ras activation. PGA 1 and some analogs received great attention in the 1990 s; although some analogs showed promising results in preclinical studies, their use in the clinical practice has not been put forward. 24 More recently, other potential clinical uses of PGA 1 or its analogs have been explored, including the potential beneficial effects of A-type cyclopentenone prostaglandins in atherosclerosis 44 and its potential use in the inhibition of the aldo-keto reductase enzymes, which are involved in cancer resistance to chemical treatment and in the development of diabetic complications. 45,46 Here our data highlight the possibility that cyclopentenone prostaglandins, because of their specific cellular effects, can be useful in the design of new therapeutic drugs and these studies are under way. Cell proliferation and cell cycle analysis. Cells (2.5 × 10 5 ) were seeded in triplicate in 60 mm dishes and synchronized as described above. They were then treated with FGF, PGA 1 , or vehicle (dimethyl sulfoxide (DMSO)) for the specified times, trypsinized at different intervals, and counted with a hemocytometer. Images were collected with a DS-L1 Digital Sight Camera System (Nikon, Nikon Instruments Europe B.V., L´Hospitalet de Llobregat, Spain) coupled to a Eclipse TS100 inverted microscope, × 40 objective lens (Nikon). Cell cycle profiles were determined by propidium iodide staining (0.1 mg/ml) and flow cytometry.
Annexin V-FITC staining. Cell lines were grown and treated as described above. Apoptosis was determined using fluorescein isothiocyanate-conjugated Annexin V (Annexin V-FITC)/Propidium Iodide Apoptosis Detection Kit (R&D systems, Minneapolis, MN, USA) according to the manufacturer's instructions. Cells that were positive for Annexin V-FITC and/or propidium iodide were analyzed using a BD FACS flow cytometer (SpainCustomer Service, San Agustin de Guadalix, Spain).
Apoptosis assays. Cells were trypsinized, mounted on glass slides, and then fixed with 70% ethanol. Morphological changes in chromatin structure were assessed after staining with DAPI. Apoptosis was characterized according to chromatin condensation and fragmentation using fluorescence microscopy. The incidence of apoptosis in each preparation was analyzed by counting 500 cells and determining the percentage of apoptotic cells. Images were collected with an HCX PL APO × 40 NA 1.32 oil-immersion objective lens (Leica, Leica Microsistemas S.L.U., Barcelona, Spain).