The Nuclear Receptor Peroxisome Proliferator-activated Receptor-β/δ (PPARβ/δ) Promotes Oncogene-induced Cellular Senescence through Repression of Endoplasmic Reticulum Stress*

Background: It is unclear whether ER stress and associated unfolded protein response (UPR) can influence oncogene-induced senescence. Results: ER stress attenuated senescence by modulating kinases, and a positive feed forward loop was delineated where ER stress caused loss of senescence and promotion of tumorigenesis. Conclusion: A new role for ER stress and UPR that attenuates H-RAS-induced senescence was discovered. Significance: PPARβ/δ may suppress RAS-dependent tumorigenesis. Endoplasmic reticulum (ER) stress and ER stress-associated unfolded protein response (UPR) can promote cancer cell survival, but it remains unclear whether they can influence oncogene-induced senescence. The present study examined the role of ER stress in senescence using oncogene-dependent models. Increased ER stress attenuated senescence in part by up-regulating phosphorylated protein kinase B (p-AKT) and decreasing phosphorylated extracellular signal-regulated kinase (p-ERK). A positive feed forward loop between p-AKT, ER stress, and UPR was discovered whereby a transient increase of ER stress caused reduced senescence and promotion of tumorigenesis. Decreased ER stress was further correlated with increased senescence in both mouse and human tumors. Interestingly, H-RAS-expressing Pparβ/δ null cells and tumors having increased cell proliferation exhibited enhanced ER stress, decreased cellular senescence, and/or enhanced tumorigenicity. Collectively, these results demonstrate a new role for ER stress and UPR that attenuates H-RAS-induced senescence and suggest that PPARβ/δ can repress this oncogene-induced ER stress to promote senescence in accordance with its role as a tumor modifier that suppresses carcinogenesis.

to inhibit mTOR activity because this concentration was shown to specifically inhibit mTOR activity (15). To induce ER stress, thapsigargin (2.5 nM) or tunicamycin (25 ng/ml) were used at these concentrations which can increase ER stress without inducing apoptosis in mouse primary keratinocytes (data not shown). An anti-BiP antibody (N-20, Santa Cruz Biotechnology, Santa Cruz, CA) was used at a concentration of 3 g/ml to immunoneutralize cell-surface BiP as this concentration was previously shown to prevent ligand binding (16).
Virus Production and Keratinocyte Infection-The H-RAS retrovirus and lentivirus encoding shRNA against Atf4 and Xbp1 were generated as previously described (10).
Clonogenic Assay with ER Stress Inducers-Keratinocyte preparations from newborn mouse skin were placed into culture dishes for 1 h. After this "pre-attachment" period, the culture medium was removed and replated into new culture dishes. This approach minimizes the contamination of fibroblasts with primary keratinocytes. Primary keratinocytes were plated at high density and infected with the H-RAS retrovirus on day 2 of culture. For transient treatment, infected keratinocytes were treated with tunicamycin or thapsigargin for 4 days before culture in normal Eagle's minimal essential medium containing 0.05 mM calcium and 8% FBS (low calcium) or in Eagle's minimal essential medium containing 0.5 mM calcium medium and 8% FBS (high calcium) for 8 weeks. For prolonged treatment, infected keratinocytes were treated with tunicamycin or thapsigargin throughout the experiment. Colonies of keratinocytes were identified by staining the cells with 0.35% rhodamine, 10% formalin. Rhodamine-stained colonies were counted with a dissecting microscope.
Complete Carcinogenesis Bioassay-A complete carcinogenesis bioassay was performed as previously described (11).
Nude Mouse Grafting-For grafting experiments, H-RASinfected keratinocytes were treated with vehicle control, tunicamycin, or thapsigargin for 4 days before culture in normal low calcium medium for another 5 days. Approximately 1.5 ϫ 10 6 keratinocytes were mixed with 3 ϫ 10 6 mouse dermal fibroblasts (cultured for 1 week) and grafted onto the back of nude mice on a prepared skin graft site as previously described (20).
Flow Cytometry Analysis-Cells were stained with bromodeoxyuridine (BrdU) and analyzed for cell cycle progression as previously described (21). Cell distribution of H-RAS expression was analyzed as previously described (10).
Protein Synthesis Assay-The Click-iT AHA Alexa Fluor 488 protein synthesis HCS kit (Invitrogen) was used to measure protein synthesis in vitro. Briefly, H-RAS-expressing keratinocytes cultured in chamber slides were treated with different chemicals for 3 days before being pulsed with 50 M L-azidohomoalanine in methionine-free medium for 1 h. Cells were fixed with 2% formaldehyde and permeabilized by 0.5% Triton X-100 and then underwent a ligation reaction for 30 min in the dark. Representative photomicrographs were obtained with a light microscope (OLYMPUS BX61), and the mean intensity of Alexa-488 signal from at least 1000 cells was measured using Cell Profiler software (Broad Institute, Cambridge, MA).
Reactive Oxygen Species (ROS) Detection Assay-Image-iT live green reactive oxygen species detection kit (Invitrogen) was used to measure the level of ROS in H-RAS-expressing keratinocytes. Wild-type H-RAS-expressing keratinocytes were treated with thapsigargin or tunicamycin for 4 days. After treatment, cells were washed with Hank's balanced salt solution and incubated with 25 M 5-(and-6)-carboxy-2Ј,7Ј-dichlorodihydrofluorescein diacetate for 30 min at 37°C. Cells were washed three times with Hank's balanced salt solution before being subjected to imaging using a fluorescence microscope. The mean intensity of the fluorescent signal from at least 1000 cells was measured using Cell Profiler software (Broad Institute, Cambridge, MA).
Matrigel Assay-Basement membrane extract (Trevigen, Gaithersburg, MD) was diluted with low calcium medium at the final concentration of 1 mg/ml and plated on top of 8 M pore size cell culture inserts (BD Biosciences). An equal number of H-RAS-expressing keratinocytes transiently treated with either thapsigargin, tunicamycin, or vehicle control were plated on top of the half-gelled basement membrane extract and cultured with low calcium medium for 24 h. After 24 h, cells migrating across the membrane were identified by staining the insert membrane with 0.35% rhodamine, 10% formalin.
Human Adenoma Samples-Matched pairs of frozen normal colon tissue and colon adenomas were obtained from The Penn State Hershey Cancer Institute Tissue Bank.
Statistical Analysis-All data analysis was performed using GraphPad Prism Version 5.0 (GraphPad Software, La Jolla, CA). Statistical significance was assessed using one-tailed Student's t test or linear regression analysis. wild-type (ϩ/ϩ) or Ppar␤/␦-null (Ϫ/Ϫ) keratinocytes. Relative expression level of protein was normalized to that of ACTIN and is shown as the relative -fold change as compared with control. Values represent the mean Ϯ S.E. *, significantly different than (ϩ/ϩ) or siRNA control (p Յ 0.05).

PPAR␤/␦ Attenuates H-RAS-induced ER Stress-H-RAS
expression caused morphologic changes in keratinocytes characterized by a temporal increase in vacuolization (Fig. 1A). However, vacuolization occurred earlier, and the percentage of vacuolized cells was higher in H-RAS-expressing Ppar␤/␦-null keratinocytes as compared with wild-type keratinocytes (Fig. 1,  A and B). Because vacuolization can be indicative of an expanded ER compartment, ER content was quantified. The mean density staining of the ER tracker DPX (staining relative to cell size) was lower in H-RAS-expressing wild-type cells compared with Ppar␤/␦-null cells (Fig. 1C). Consistent with this observation, decreased expression of CALNEXIN (an ERresident protein) was found in H-RAS-expressing wild-type compared with Ppar␤/␦-null cells (Fig. 1D). The vacuolization and expansion of the ER compartment are consistent with previously observed H-RAS-induced ER stress previously observed in primary melanocytes (3).
In response to ER stress, UPR is activated to attenuate the ER stress by activating the three branches of the UPR pathway: 1) PERK, 2) ATF6, and 3) IRE1 (1, 2) (Fig. 1F). Microarray analysis from a previously published study (10) revealed increased expression of key UPR regulators in both wild-type and Ppar␤/ ␦-null cells in response to H-RAS activation, but the relative increase was markedly higher in Ppar␤/␦-null cells compared with wild type (Fig. 1, G and H). Western blot analysis confirmed decreased expression of the UPR sensor BiP and ATF6 in H-RAS-expressing wild-type cells compared with Ppar␤/␦null cells (Fig. 1D). Furthermore, although phosphorylated PERK (pPERK) was not detected in H-RAS-expressing wildtype cells, pPERK was found in H-RAS-expressing Ppar␤/␦null cells (Fig. 1D), an effect not due to a difference in H-RAS expression, which was expressed at similar levels in both cell types (Fig. 1D). Knockdown of PPAR␤/␦ in 308 keratinocytes, a cell line that expresses a constitutively active H-RAS, revealed similar results (Fig. 1E). Higher expression of ATF4, ATF6, and BiP were observed in 308 keratinocytes when PPAR␤/␦ was knocked down as compared with controls (Fig. 1E). These data indicate that PPAR␤/␦ attenuates two of the three branches of the UPR pathway activated by ER stress (Fig. 1F). Alternative splicing of Xbp1 mRNA was examined as an indirect measure of the third branch of the UPR pathway, IRE1 activity (Fig. 1F). Reduced splicing of Xbp1 mRNA was observed in H-RAS-expressing wild type compared with Ppar␤/␦-null cells ( Fig. 2A), suggesting that PPAR␤/␦ also attenuates the IRE1-XBP1 branch of the UPR response in H-RAS-expressing wild-type cells.
H-RAS-expressing wild-type cells were also less sensitive to tunicamycin (an ER stress inducer)-induced cell death as compared with Ppar␤/␦-null cells (Fig. 2B), consistent with the reduced ER stress observed in H-RAS-expressing wild-type cells (Fig. 1, A-C). Although the UPR initially promotes cell survival, in the presence of prolonged ER stress it can also activate pathways leading to apoptosis (1,2). Western blot analysis showed similar expression of pro-apoptotic BAD and BAX protein in H-RAS-expressing wild-type and Ppar␤/␦-null cells (Fig. 2C), consistent with propidium iodide staining revealing no change in the percentage of apoptotic cells in H-RAS-expressing wild-type and Ppar␤/␦-null cells (Fig. 2D). Additionally, overexpression of PPAR␤/␦ in human HaCaT keratinocytes decreased expression of the UPR markers ATF4, ATF6, and IRE1␣ (Fig. 2E). These data suggest that PPAR␤/␦ represses H-RAS-induced ER stress without triggering apoptosis.
H-RAS Promotes ER Stress through p-AKT/mTOR Signalingp-AKT activity can influence ER stress (3). This is consistent with relatively lower H-RAS-induced ER stress and lower p-AKT in H-RAS-expressing wild-type keratinocytes com-pared with H-RAS-expressing Ppar␤/␦-null cells (11). Similarly, this is consistent with relatively higher H-RAS-induced ER stress and higher p-AKT in H-RAS-expressing Ppar␤/␦null keratinocytes compared with H-RAS-expressing wild-type keratinocytes (11). These collective observations support the hypothesis that p-AKT (which is activated by PI3K) promotes ER stress in H-RAS-expressing cells. Inhibition of PI3K activity with LY294002 significantly decreased the percentage of vacuolized H-RAS-expressing cells in both genotypes (Fig. 3, A and  B). In addition, inhibition of PI3K with LY294002 decreased p-AKT and decreased expression of the ER stress and UPR mediators ATF6 and BiP in H-RAS-expressing cells (Fig. 3C). These data indicate that p-AKT activity promotes H-RAS-induced ER stress.
Mammalian target of rapamycin (mTOR) is a protein complex that could potentially mediate p-AKT-induced ER stress because 1) mTOR is a target of p-AKT (23) and 2) increased mTOR activity can cause increased protein synthesis (24), which could overload the ER, leading to ER stress. Inhibition of PI3K with LY294002 significantly decreased expression of the mTOR target genes phospho-p70S6K (p-p70S6K) and phospho-S6RP (p-S6RP) in H-RAS-expressing cells (Fig. 4A). This suggests that mTOR is indeed a target of p-AKT in H-RASexpressing keratinocytes. Consistent with a higher activity of mTOR in H-RAS-expressing Ppar␤/␦-null cells (Fig. 4A), H-RAS-expressing Ppar␤/␦-null cells also exhibited an ϳ4-fold higher protein synthesis rate compared with wild-type counterparts (Fig. 4C). Inhibition of PI3K and mTOR with LY294002 and rapamycin, respectively, reduced the rate of protein synthesis to a level similar to that of H-RAS-expressing wild-type cells. (Fig. 4, B and C) Although only an ϳ2-fold decrease was observed in H-RAS-expressing wild-type cells as compared with Ppar␤/␦-null cell (Fig. 4, B and C). Inhibition of mTOR with rapamycin also decreased the percentage of vacuolized cells and expression of ER stress and UPR markers IRE1␣ and BiP in H-RAS-expressing cells (Fig. 4, D-F), and the magnitude of these changes were greater in H-RAS-expressing Ppar␤/␦-null cells. These data collectively suggest that p-AKT activity increases H-RAS-induced ER stress by activating mTOR signaling leading to increased protein synthesis. Furthermore, these data also indicate that PPAR␤/␦ represses ER stress by attenuating p-AKT/mTOR signaling. Decreasing ER Stress-associated UPR Increases H-RAS-induced Senescence-PPAR␤/␦ promotes H-RAS-induced cellular senescence (11). The relatively lower level of ER stress observed in H-RAS-expressing wild-type keratinocytes in the present study correlates with relatively higher levels of senescence previously observed (11). Conversely, the relatively higher level of ER stress observed in H-RAS-expressing Ppar␤/ ␦-null keratinocytes in the present study correlates with relatively lower levels of senescence previously observed (11). Thus, H-RAS-expressing wild-type cells were used as a model of modest ER stress/high senescence, and the H-RAS-expressing Ppar␤/␦-null cells were used as a model of relatively high ER stress/low senescence.
Examination of H-RAS-expressing keratinocytes revealed a negative correlation between vacuolization (a marker of ER stress) and ␤-gal-positive staining (a marker of senescence) in both models (Fig. 5, A and B). Furthermore, rapidly proliferating H-RAS-expressing Ppar␤/␦-null cells that are resistant to calcium-induced cell cycle arrest as described previously (11) exhibited massive vacuolization (Fig. 5C). H-RAS-expressing keratinocytes resistant to calcium-induced differentiation are considered to have undergone malignant conversion (25) and were examined to demonstrate grossly exacerbated ER stress due to the absence of PPAR␤/␦ in a model of rapidly proliferating cells that are resistant to senescence (11). These observations support the hypothesis that ER stress and the UPR attenuate H-RAS-induced cellular senescence. To test this hypothesis, two of the three main UPR pathways (PERK and IRE1; Fig. 1E) were selectively inhibited by shRNA in H-RAS-expressing keratinocytes. Knockdown of ATF4 or XBP1 increased the number of senescent cells in both H-RAS-expressing wild-type and Ppar␤/␦-null keratinocytes (Fig.  6, A and B). Expression of the senescence marker p27 was also increased after ATF4 or XBP1 knockdown in both H-RAS-expressing wild-type and Ppar␤/␦-null cells (Fig. 6, C and D). Expression of p-AKT and the direct ATF4/XBP1 target gene (1) BiP were decreased after either ATF4 or XBP1 knockdown in both H-RASexpressing wild-type and Ppar␤/␦-null cells (Fig. 6, C and D). In contrast, no change of p-ERK expression was observed after either ATF4 or XBP-1 knockdown in either H-RAS-expressing wild-type or Ppar␤/␦-null cells (data not shown). Furthermore, knockdown of either ATF4 or XBP1 by shRNA partially restored cellular senescence in 3 of 6 previously described, calcium-resistant (i.e. malignant) H-RAS-expressing Ppar␤/␦-null clones (11) (Fig. 6E). These data suggest that ER stress and UPR attenuate H-RAS-induced senescence.
Increased ER Stress and UPR Decreases H-RAS-induced Senescence-Two common ER stress inducers, thapsigargin and tunicamycin (26), were used to further explore the relation-ship between UPR and senescence. Low concentrations of both chemicals were used to avoid extensive ER stress-induced cell death observed at higher concentrations. Treating both H-RAS-expressing wild-type and Ppar␤/␦-null keratinocytes with increasing concentrations of both reagents caused a dosedependent increase of vacuolized cells (data not shown) and a decreased number of senescent cells (Fig. 7, A and B). Because more marked effects were found in H-RAS-expressing wild- type cells due to lower basal ER stress compared with Ppar␤/␦null cells (Fig. 1), only wild-type H-RAS-expressing cells were used in the following studies. Thapsigargin or tunicamycin treatment caused an increase in BrdU-positive cells (Fig. 7C) and increased expression of PCNA (Fig. 8, A and B) compared with controls, which is indicative of higher cell proliferation and consistent with reduced senescence. In addition, increased expression of UPR markers IRE1␣, BiP, and PERK and decreased expression of the senescence markers p16 and DcR2 were observed in H-RAS-expressing cells treated with thapsigargin (Fig. 8A). Tunicamycin caused similar changes in the expression of UPR and senescence markers compared with controls (Fig. 8B). Combined, these data suggest that low concentrations of thapsigargin and tunicamycin are sufficient to induce ER stress and UPR, causing attenuation of H-RAS-induced senescence and increased cell proliferation.
Increased ER Stress and UPR Decreases H-RAS-induced Senescence by Increasing p-AKT-To examine the signaling pathways by which ER stress inhibits senescence and increases cell proliferation, downstream signaling pathways modulated by H-RAS, including p-AKT and p-ERK, were examined after thapsigargin and tunicamycin treatment. Interestingly, both ER stress inducers caused an increase in p-AKT in H-RAS-expressing keratinocytes (Fig. 8, A and B). In contrast, expression of H-RAS and p-ERK was modestly decreased after either thapsigargin or tunicamycin treatment (Fig. 8, A and B). The fact that increased ER stress caused both an increase of p-AKT and a decrease of p-ERK is important to note because a previous study suggested that both attenuation of p-ERK and potentiation of p-AKT are required to evade H-RAS-induced senescence (11). Flow cytometric analysis revealed that the modestly decreased expression of H-RAS by thapsigargin and tunicamycin was most likely due to a selection against higher H-RASexpressing cells (Fig. 8C), similar to that previously observed by ligand activation of PPAR␤/␦ (10), although more extensive studies are needed to confirm this idea. This could explain the decreased expression of H-RAS and p-ERK in response to an increase in ER stress induced by either thapsigargin or tunica- mycin. However, this selection does not explain how either thapsigargin or tunicamycin caused an increase in p-AKT.
To determine if ER stress increases cell proliferation and attenuates senescence through a p-AKT-dependent mechanism, H-RAS-expressing cells were co-treated with LY294002 and tunicamycin. Inhibition of PI3K activity with LY294002 prevented a tunicamycin-induced increase in BrdU labeling and PCNA expression (Figs. 7C and 8B) and partially abolished tunicamycin-induced repression of senescence (Fig. 8D). Moreover, co-treatment with LY294002 and tunicamycin restored expression of the senescence marker p27 but not p16 and DcR2 in H-RAS-expressing cells (Fig. 8B). These effects were not due to decreased ER stress (Fig. 8D). These findings suggest that although increased ER stress and a UPR increase cell proliferation through a p-AKT dependent mechanism, it partially attenuates senescence through a p-AKT dependent mechanism.

ER Stress-induced UPR Maintains Higher p-AKT Levels Partially through a Cell-surface BiP-dependent Mechanism-
The mechanisms of how ER-stress-induced UPR maintains higher p-AKT were examined next. The ER stress marker BiP was of particular interest because 1) expression of BiP correlated with that of p-AKT whether UPR was decreased (Fig. 6, C and D) or increased (Fig. 8, A and B), and 2) in addition to its role as an ER chaperone, BiP was also found on the cell surface of cancer cells and when coupled with other cell-surface protein partners can activate the PI3K-AKT pathway (27,28). To confirm that BiP is expressed on the cell surface in H-RAS-expressing keratinocytes, immunofluorescent studies were performed with antibodies against both BiP and E-cadherin (a cell-surface marker). Co-localization of E-cadherin and BiP would suggest that BiP is indeed expressed on the cell surface in H-RAS-expressing keratinocytes. Although expression of BiP was lacking on cell membranes in control keratinocytes (Fig. 9A), expression of BiP and E-cadherin on the cell-surface was detected in H-RAS-expressing keratinocytes (Fig. 9A). Neutralizing cell-surface BiP with an anti-BiP antibody (16) decreased expression of p-AKT and significantly decreased cell proliferation concomitant with decreased expression of p-RB expression in H-RAS-expressing cells (Fig. 9, B and C). However, no change in senescence was found after immunoneutralization of cell-surface BiP (Fig. 9D). Similarly, inducing ER stress with tunicamycin increased cell proliferation and immunoneutralizing cell-surface BiP repressed this effect (Fig. 9E) as well as tunicamycin-induced increased p-AKT and p-RB expression in H-RAS-expressing cells (Fig. 9F). In addition, immunoneutralizing cell-surface BiP had no effect on decreased senescence caused by tunicamycin treatment (Fig. 9G). That inhibition of cell proliferation by immunoneutralizing cell-surface BiP was more marked in tunicamycin-treated H-RAS-expressing keratinocytes (Fig. 9E) as compared with control H-RAS-expressing keratinocytes (Fig.  9B) suggests the effects of cell-surface BiP neutralization on cell proliferation are stronger when cell proliferation is more enhanced. More importantly, these data also indicate that cellsurface BiP can maintain high p-AKT in response to increased ER stress and promote cell proliferation but do not completely account for the attenuation of senescence caused by ER stress.  (Fig. 3), and ER stress-associated UPR in turn increases p-AKT expression (Fig.  8). This suggests the presence of a positive feed forward loop between p-AKT, ER stress, and UPR. If so, then transient stimulation of ER stress should be sufficient to establish the positive feed forward loop and maintain higher UPR and higher p-AKT activity, leading to evasion of senescence even after the initial stimuli is removed. To test this hypothesis, H-RAS-expressing keratinocytes were treated briefly with either thapsigargin or tunicamycin briefly, and senescence measured 8 days after treatment. Transient treatment with both thapsigargin and tunicamycin caused a decrease of senescence (Fig. 10A), concomitant with decreased expression of senescence marker p16, p27, and p21 in H-RAS-expressing keratinocytes compared with controls (Fig. 10B). In addition, increased expression of ER stress markers IRE1␣, BiP, and p-AKT and decreased expression of p-ERK were observed at this time in H-RAS-expressing keratinocytes after transient induction of ER stress by both thapsigargin and tunicamycin (Fig. 10B).

A Transient Increase of ER Stress Is Sufficient to Cause Evasion of Senescence and Malignant Conversion in Vitro-Enhanced p-AKT activity increases ER stress
Because decreased senescence is often associated with an increase in the extent of cancer malignancies (4), the effects of increased ER stress on malignant conversion of H-RAS-expressing keratinocytes were examined by clonogenic and Matrigel assays in vitro. After the introduction of oncogenic H-RAS, no colonies were observed after 4 weeks of culture in either low calcium or high calcium medium (Fig. 10, C and D), consistent with previous findings (11). In contrast, transiently treating H-RAS-expressing cells with thapsigargin resulted in the formation of colonies in both low calcium and high calcium medium (Fig. 10, C and D). Interestingly, prolonged treatment with thapsigargin resulted in no colonies cultured in low calcium and fewer colonies in high calcium medium compared with the transient treatment strategy (Fig. 10, C and D). Even though a number of colonies were initially observed at the early times of thapsigargin treatment, they eventually sloughed off (Fig. 10, C and D, data not shown). Additionally, transient treatment of H-RAS-expressing cells with either thapsigargin or tunicamycin increased the average number of cells migrating across a basal membrane in a Matrigel assay (Fig. 10E). ER stress can increase ROS levels (29) and thereby potentially increase the chance of secondary mutations, which may lead to escaping from senescence. To examine this possibility, the level of ROS was measured in H-RAS-expressing keratinocytes after thapsigargin or tunicamycin treatment. No change in ROS levels was found between groups (Fig. 10F). These data collectively suggest that a transient increase of ER stress is sufficient to cause evasion of senescence and malignant conversion in vitro.  (11) were used to confirm these effects in vivo. We utilized this model because 1) dimethylbenz[a]anthracene-induced benign skin tumors exhibit extensive cellular senescence (30), thus providing a valuable in vivo model to investigate the potential relationship between ER stress and cellular senescence, and 2) it was previously shown that PPAR␤/␦ promotes H-RAS-induced senescence in dimethylbenz[a]anthracene-induced skin tumors and that skin tumors from Ppar␤/␦-null mice exhibited increased cell proliferation and decreased cellular senescence in vivo (11). Skin tumors examined from this bioassay revealed that p-AKT expression positively correlated with that of ATF4 in both wild-type and Ppar␤/␦-null mouse skin tumors (Fig.  11A). Lower nuclear ATF4 expression was also found in skin tumors from wild type as compared with Ppar␤/␦-null mice (Fig. 11A). This is consistent with decreased p-AKT expression in skin tumors from wild-type mice as compared with Ppar␤/ ␦-null mice (11). XBP-1 and DcR2 were chosen as ER stress and senescence markers, respectively, to examine the relationship between ER stress and senescence in vivo. During increased ER stress, Xbp-1 mRNA is alternatively spliced, resulting in nuclear  JULY 18, 2014 • VOLUME 289 • NUMBER 29 translocation of active XBP-1 (29,31). Thus, an increased ratio of nuclear to cytoplasmic XBP-1 indicates increased ER stress (29,31). DcR2 is a senescence marker highly expressed in senescent skin papillomas but not in normal skin (30). An increased nuclear to cytoplasmic ratio of XBP-1 was found in skin tumors compared with normal skin in both wild-type and Ppar␤/␦-null mice, consistent with increased ER stress commonly observed in tumors. In addition, this ratio was significantly lower in wildtype tumors compared with Ppar␤/␦-null tumors, suggesting that PPAR␤/␦ also inhibits ER stress in vivo (Fig. 11, B and C). In addition, higher p-AKT expression positively correlated with increased XBP-1 expression in non-senescent regions (Fig.  11D). In contrast, senescent regions that expressed DcR2 (and also positive for ␤-Gal) expressed lower p-AKT and, more importantly, showed increased cytoplasmic XBP-1 distribution (Fig. 12, A-C). These effects were observed in skin tumors from both wild-type and Ppar␤/␦-null mice although a more striking correlation was observed in Ppar␤/␦-null mice (Fig. 12, A-C). A positive correlation between the expression of p-AKT and BiP in mouse skin tumors and a negative correlation between the expression of p16 and both BiP and ATF4 in mouse skin tumors were also confirmed by quantitative Western blot analysis (data not shown). These findings suggest a new pro-tumorigenic function of ER stress in attenuating senescence in vivo.

ER Stress Inhibits H-RAS-induced Senescence
To further test the hypothesis that increased ER stress can help bypass the senescence checkpoint and facilitate H-RAS-induced tumorigenesis of primary keratinocytes in vivo, 1.5 million H-RAS-expressing keratinocytes that were pretreated with vehicle control, tunicamycin, or thapsigargin were mixed with 3 million dermal fibroblasts and grafted onto the back of athymic nude mice. Interestingly, although no tumors were found in mice grafted with vehicle-treated H-RAS-expressing keratinocytes, 66 and 50% of mice grafted with the same number of H-RAS-expressing keratinocytes that were pretreated with tunicamycin and thapsigargin, respectively, developed skin tumors (Fig. 13, A and B). In addition, although mice grafted with vehicle-treated H-RAS-expressing keratinocytes only showed signs of mild hyperplasia (Fig. 13C), mice grafted with H-RAS-expressing keratinocytes that were pretreated with tunicamycin or thapsigargin developed both benign papillomas and malignant squamous cell carcinomas (Fig. 13C). This evidence implies that increased ER stress greatly facilitates H-RAS-induced tumorigenesis of primary keratinocytes in vivo.
ER Stress Is Negatively Correlated with Cellular Senescence in Human Benign Lesions-The relationship between cellular senescence and ER stress was also examined in human colon adenomas using analysis of a microarray database and exami- nation of human tissue samples. Human colon adenomas were examined because they contain cells with relatively high p16 and p53 expression and the absence of both mitotic and cell proliferation markers compared with normal tissue (32,33). Examination of a human colon adenoma microarray database (34) revealed that the average expression of p53 mRNA is higher in human colon adenomas as compared with normal colon, consistent with a relatively higher level of senescence in adenomas as compared with non-transformed tissue (Fig. 14A). By contrast, the average expression of ATF4 and BiP mRNA was higher in human colon adenomas as compared with normal colon, which is consistent with a relatively higher level of ER stress in adenomas as compared with non-transformed tissue (Fig. 14A). However, when tumors were examined, a negative correlation between mRNA expression of the senescence marker p53 and three UPR markers, ATF6, BiP, and ATF4, was found (Fig. 14B). In addition, a negative correlation was also found between p16 and BiP expression in three independent human colon adenoma tissue samples examined (Fig. 14C). These data strongly support the hypothesis that ER stress can also attenuate cellular senescence in human benign lesions.

DISCUSSION
The present study revealed a novel interaction between p-AKT signaling, ER stress, and cellular senescence as summarized in Fig. 15. ER stress and UPR can either promote or inhibit tumorigenesis given different cellular contexts (2). However, the relationship between ER stress and oncogene-induced cellular senescence is less clear. A previous study showed that ER stress promotes H-RAS-induced senescence in melanocytes (3). In this study attenuation of ER stress by knockdown of ATF4 or XBP-1 or overexpression of a dominant negative ATF6 all decreased H-RAS-induced cellular senescence in melanocytes (3). In contrast, in the present study, knockdown of ATF4 or XBP-1 increased H-RAS-induced senescence in both wild-type and Ppar␤/␦-null keratinocytes. Moreover, increasing ER stress by treatment with thapsigargin or tunicamycin attenuated cellular senescence in the present studies. The differences in the effects of ER stress on cellular senescence between melanocytes and mouse primary keratinocytes, mouse skin tumors, and human colon adenomas could reflect different effects of p-AKT signaling on cellular senescence in different tissues. Although p-AKT expression after ATF4 and XBP-1 knockdown in melanocytes was not changed in one study (3), decreased p-AKT signaling after both XBP-1 and ATF4 knockdown was observed in the present study. The fact that pharmacological inhibition of p-AKT signaling in melanocytes decreased cellular senescence suggests that p-AKT promotes cellular senescence in melanocytes (3). In contrast, pharmacological inhibition of p-AKT signaling in keratinocytes increased cellular senescence, suggesting that p-AKT inhibits cellular senescence in keratinocytes (11). The latter explanation is consistent with results from other studies (14,35). The mechanisms underlying these opposing effects of p-AKT signaling on H-RAS-induced senescence in different tissues remains unclear and requires additional experimentation.
H-RAS-induced higher p-AKT signaling activates mTOR, increases protein synthesis and ultimately enhances ER stress. Increased ER stress caused an increase in UPR, which in turn increased p-AKT signaling partially through an elevation in BiP expression. Interestingly, cell-surface BiP, which is often found in cancer cells, can induce the PI3K/AKT pathway when activated by ligand (27,28). However, the results indicated that cell-surface BIP activity only partially mediated the increased p-AKT induced by increasing ER stress but did not explain the attenuation of senescence caused by ER stress. Thus, it is of interest to note that increased expression of XBP1 and PERK by ER stress can also increase PI3K/AKT signaling (36,37). Furthermore, increased activity of intracellular BiP may also contribute to the increased p-AKT signaling through unknown mechanisms, and it remains possible that the increased p-AKT signaling in response to ER stress is mediated through multiple parallel pathways. Regardless of the precise mechanism, a positive feed forward loop between ER stress, UPR, and p-AKT was discovered in H-RAS-expressing keratinocytes. Intriguingly, a transient increase of ER stress induced by either thapsigargin or tunicamycin was sufficient to establish the positive feed forward loop, maintaining higher ER stress and UPR, leading to attenuation of senescence and promotion of cell proliferation and tumorigenesis long after the initial transient exposure. The effects of ER stress on promoting cell proliferation were dependent on the cell-surface BiP, as neutralizing cell-surface BiP with an anti-BiP antibody completely inhibited cell proliferation but didn't influence cellular senescence. One possible explanation is that the decrease of p-AKT expression by immu- FIGURE 13. Transiently increased ER stress facilitates H-RAS-induced tumorigenesis of primary keratinocytes in an allograft model. Athymic nude mice were grafted with a mixture of dermal fibroblasts and H-RAS-expressing keratinocytes transiently treated with thapsigargin (Thap), tunicamycin (Tu), or vehicle control (C). A, representative photographs of skin tumors resulting from these treatments. B, the percentage of mice that developed skin tumors from grafted mixtures of dermal fibroblasts and H-RASexpressing keratinocytes transiently treated with thapsigargin or tunicamycin. C, H&E staining of representative skin and tumors isolated from mice grafted with a mixture of vehicle control dermal fibroblasts and H-RAS-expressing keratinocytes or those transiently treated with tunicamycin or thapsigargin. SCC, squamous cell carcinoma. *, significantly different than control, p Յ 0.05. noneutralization of cell-surface BiP is relatively subtle compared with that seen by inhibition of PI3K by LY294002, and there may be different threshold levels of p-AKT required for promoting cell proliferation versus repressing senescence. It is also possible that immunoneutralizing cell-surface BiP triggers p-AKT-independent responses that offset the pro-senescence effects of decreased p-AKT signaling.
Higher p-AKT signaling was positively correlated with increased ER stress and UPR and negatively correlated with senescence in mouse skin tumors in vivo. In addition, a negative correlation between BiP and p16 was noted in human colon adenomas harboring KRAS mutations and between the expression of the senescence marker p53 mRNA and ER stress markers ATF4, ATF6, and BiP mRNAs in human colon adenomas. Interestingly, this negative correlation may also exist even in the absence of activated RAS/p-AKT signaling because ATF4 was shown to promote tumorigenesis by directly repressing p16 and p19 expression (38). Thus, ER stress may attenuate cellular senescence through both p-AKT-dependent and p-AKT-independent mechanisms.
In addition to activation of the p-AKT pathway, ER stress can also modulate p-ERK signaling and attenuate cellular senescence. Increasing ER stress by tunicamycin and thapsigargin decreases p-ERK signaling potentially by selecting against higher H-RAS-expressing cells. This selection could be due to increased sensitivity to ER stress-induced apoptosis in cells expressing relative higher level of H-RAS. Indeed, increased PARP cleavage was observed in H-RAS-expressing cells after tunicamycin and thapsigargin treatment (data not shown). Further experiments are needed to test this hypothesis. Several studies have reported that inhibition of p-ERK activity represses oncogene-induced senescence by down-regulating expression of p16 and p21 (11,39,40). Therefore, decreased p-ERK signaling could also mediate decreased cellular senescence after increased ER stress.
Interestingly, PPAR␤/␦ represses H-RAS-induced ER stress by repressing ILK/PDPK1/p-AKT/mTOR signaling pathway, consistent with repression of p-AKT signaling and decreased expression of ER stress markers including ATF4, ATF6, and BiP by overexpression of PPAR␤/␦. Previous studies showed that PPAR␤/␦ promotes H-RAS-induced cellular senescence by upregulating p-ERK and repressing p-AKT signaling by transcriptionally regulating the expression of RASGRP1 and ILK, respectively (11). Through the repression of ER stress, PPAR␤/␦ could reinforce the higher p-ERK and lower p-AKT signaling to promote cellular senescence (Fig. 15).