The Peroxisome Proliferator-Activated Receptor Gamma System Regulates Ultraviolet B-Induced Prostaglandin E2 Production in Human Epidermal Keratinocytes

Studies using PPARγ agonists in mouse skin have suggested that peroxisome proliferator-activated receptor gamma (PPARγ) is irrelevant to cutaneous photobiology. However, in several epithelial cell lines, ultraviolet B (UVB) has been shown to induce the nonenzymatic production of oxidized phospholipids that act as PPARγ agonists. UVB is also a potent inducer of prostaglandin E2 (PGE2) production and COX-2 expression in keratinocytes and PPARγ is coupled to increased PGE2 production in other cell lines. In this current study, we demonstrate that PPARγ agonists, but not PPARα or PPARβ/δ agonists, induce PGE2 production and COX-2 expression in primary human keratinocytes (PHKs). Importantly, PPARγ agonist-induced COX-2 expression and PGE2 production were partially inhibited by the PPARγ antagonist, GW9662, indicating that both PPARγ-dependent and -independent pathways are likely involved. GW9662 also suppressed UVB and tert-butylhydroperoxide- (TBH-) induced PGE2 production in PHKs and intact human epidermis and partially inhibited UVB-induced COX-2 expression in PHKs. These findings provide evidence that PPARγ is relevant to cutaneous photobiology in human epidermis.


Introduction
Peroxisome proliferator-activated receptors (PPARs) are ligand-activated nuclear transcription factors that were initially identified as being crucial for regulating the formation of intracellular organelles, called peroxisomes, that are involved in lipid metabolism (reviewed in [1]). Three different PPARs subtypes have been cloned (α, β/δ, and γ) that differ in ligand specificity, tissue expression, and transcriptional targets. All three PPARs require heterodimerization with the retinoid X receptors (RXRs) for transcriptional activity. PPAR:RXR heterodimers induce target gene transcription by binding to specific peroxisome proliferators response elements (PPREs) in the promoter region of target genes [1]. Importantly, PPARα, PPARβ/δ, and PPARγ are all expressed in keratinocytes and in human and rodent epidermis [1][2][3]. We have recently demonstrated that PPARγ is expressed in adult primary human keratinocytes and three different immortalized or malignant human keratinocyte cell lines (A431, HaCaT, and KB cells) [4,5]. In KB epidermoid carcinoma cells and SZ95 sebocytes cells, we have also demonstrated that oxidative stress, including ultraviolet B irradiation, results in the production of oxidized lipid species with potent PPARγ ligand activity [4,5].
While it is clear that UVB and other oxidative stressors can result in PPARγ ligand production, a previous study using exogenous PPARγ agonists failed to demonstrate any effect on either chemical carcinogenesis or UVB-induced skin cancer formation [12]. These negative findings raise doubts concerning the relevance of PPARγ to cutaneous photobiology. Yet, it should be noted that mice with heterozygous germline deletion of PPARγ or mice with epidermal-specific loss of PPARγ exhibit an increase in chemical carcinogen-induced skin tumors [13,14]. This suggests the possibility that loss of function models, such as the use of PPARγ antagonists rather than agonists, might be more informative for studies designed to examine the role of PPARγ in photobiology. Given that we have already demonstrated that UVB induces PPARγ ligand formation, we hypothesized that the lack of effect of exogenous PPARγ ligands on photocarcinogenesis could be explained by the fact that PPARγ is already engaged by UVB-induced ligand production. Inasmuch as our previous studies using cell lines indicate that PPARγ is coupled to epithelial COX-2 expression and PGE 2 production, we therefore utilized the PPARγ antagonist, GW9662, to determine whether PPARγ is involved in regulating UVB-induced COX-2 expression and PGE 2 production in primary human keratinocytes and intact human epidermal explants. The results of the present studies indicate that PPARγ is functionally coupled to a readily measured photobiological response in human primary epidermal keratinocytes and is therefore relevant to cutaneous photobiology.

Cell
Culture. Adult primary human keratinocytes (PHKs) were prepared from discarded epidermis that was obtained from reductive mammoplasties and panniculectomies as previously described [15]. Telomeraseimmortalized primary human keratinocytes (N/TERT-1) were obtained from Dr. Rheinwald (Department of Medicine and Harvard Skin Disease Research Center, Brigham, and Women's Hospital, Boston, MA) [16]. PHKs and N/TERT-1 cells were cultured on tissue culture plastic or wells that were precoated with type I collagen. PHKs and N/TERT-1 cells were grown in serum-free media (Keratinocyte serumfree media, K-SFM; Gibco Invitrogen, Carlsbad, CA). Media were supplemented with 40 IU per mL penicillin, 40 μg per mL streptomycin, and 0.1 μg per mL amphotericin B. The cells were cultured in 95% air and 5% CO 2 at 37 • C. All studies were done using PHKs or N/TERT-1 cells that were plated at sufficient density to achieve approximately 70%-80% confluence prior to experimental manipulation. All studies with human skin have been approved by the Indiana University-Purdue University at Indianapolis Institutional Review Boards using the Declaration of Helsinki Principles.

Immunoblotting.
Immunoblotting for PPARγ was done using mouse monoclonal anti-PPARγ antibody (clone E8; Santa Cruz Biotechnology, Santa Cruz, CA), essentially as described in [4]. The specificity of this antibody for PPARγ has previously been demonstrated in wild-type versus tissuespecific PPARγ knockout mice [17].

Ultraviolet B Irradiation of Cultured Keratinocytes.
For UVB irradiation studies, a Philips F20T12/UV-B lamp (270-390 nm), containing 2.6% UVC, 43.6% UVB, and 53.8% UVA, was utilized. The UVB dose was measured using an IL1700 radiometer and a SED240 UVB detector (International Light, Newburyport, MA). All irradiations were performed at a distance of 8 cm from the UVB light source. All irradiations were done on PHKs grown in 24-well plates. For GW9662-and NS398-treated cells, media containing vehicle, 1 μM GW9662, or 10 μM NS398 were added either 1 hour (GW9662) or 30 minutes (NS398) prior to UVB irradiation. The cells were then washed twice with PBS, and then the cells were irradiated with 600 J/m 2 UVB. DMEM containing 10% FBS was then added containing vehicle, GW9662, or NS398. After eight hours at 37 • C, the culture supernatants were removed for PGE 2 quantitation, the cell monolayer was trypsinized, and the cells were counted. PGE 2 was then normalized to cell count.

Tert-Butylhydroperoxide (TBH) Studies in PHKs In Vitro.
For TBH studies, cells were plated onto 24-well plates in high-calcium DMEM containing 10% fetal bovine serum. The cells were then pretreated with GW9662 or NS398 as detailed in the previous section prior to TBH addition. Culture supernatants were then collected after an eight-hour incubation at 37 • C.

PPAR Agonist Studies.
For PPAR agonist studies, agonists were added in low-calcium K-SFM media or in K-SFM media supplemented with 1 mM calcium and incubated for 24 hours at 37 • C. To assess PPARγ-dependent induction of PGE 2 , GW9662 (1 μM) was added 1 hour prior to agonist addition. PGE 2 was then quantitated in tissue culture media by EIA and normalized to either cell count or total cellular protein (BCA assay; Pierce Biotechnology, Rockford, IL).

RNA Isolation and COX-2 Quantitative RT-PCR.
PHKs were treated with vehicle and ciglitazone (with and without GW9662) as detailed in the previous section. At 2, 4, 8, and 24 hours after the addition of reagents, the cell monolayers were processed for RNA isolation using an RNeasy kit (Qiagen) according to the manufacturer's protocol. Following first-strand DNA synthesis, quantitative RT-PCR (qRT-PCR) was performed using primers specific to human COX-2 and 18S rRNA with a Cepheid Smart Cycler real-time PCR instrument (Fisher Scientific, Pittsburgh, PA). COX-2 and 18S qRT-PCR were performed as previously described [5]. COX-2 results were then normalized to 18S using the ΔΔCt method [18].

Human Epidermal Explant Preparation and UVB Irradiation.
Adult human epidermis obtained from panniculectomies was obtained postoperatively. Subdermal fat and a portion of the lower dermis were immediately removed using a scalpel blade. The epidermis was then cut into small (approx. 8 × 8 mm) sections, weighed, and then cultured in 12-well plates submerged in Keratinocyte-SFM media for 48-72 hours. The explants were then pretreated with vehicle, 1 μM GW9662 (1 hour), or 10 μM NS398 (30 minutes) prior to UV treatment. After pretreatment, the explants were washed twice with PBS and a minimal amount of PBS was added to cover the explants to maintain hydration. The explants were then irradiated with 1,800 J/m 2 of UVB light. Control cells received no UVB light. The explants were then submerged in serum-free DMEM containing vehicle, 1 μM GW9662, or 10 μM NS398 and cultured for 8 hours at 37 • C. At this time the media were removed for PGE 2 quantitation. All PGE 2 levels were then normalized to tissue weight.

COX-2 Immunoblot.
Second-passage PHKs grown on 6-well plates to near confluence were treated with 1 μM GW9662 or vehicle 1 hour prior to washing the wells with HBSS, then irradiating the cells with 300 J/m 2 of UVB. The HBSS was then replaced with media containing vehicle or 1 μM GW9662, and the cells were processed 20 hours later in RIPA buffer supplemented 1 : 100 with a protease inhibitor cocktail (Sigma-Aldrich). Protein was quantitated using a DC protein assay, and equal amounts of total protein (50 μg/lane) were then separated on 7.5% SDS-PAGE gels. Rabbit polyclonal anti-COX-2 (Cayman Chemical) diluted 1 : 1000 in TBS with 0.1% Tween 20 and 2% Blotto was added and incubated overnight at 4 • C in 2% Blotto. Goat anti-rabbit HRP conjugate (1 : 10,000; Source) in 5% Blotto was applied and the immunoreactive bands were detected enhanced chemiluminescence. The exposed films were developed and scanned and band intensity was determined using NIH Image J software. The band intensity of the COX-2  Figure 1: PPARγ expression by immunoblot in PHKs. 40 μg cellular protein isolated from PHKs from two different individuals was grown on tissue culture plastic coated with (Coll) and without (Plas) type I rat tail collagen. Protein from telomerase-immortalized primary human keratinocytes (N/TERT-1) was also utilized. In all cases, the proteins were separated on a 10% SDS-PAGE and PPARγ protein was detected using a monoclonal anti-PPARγ antibody. Loading was verified by stripping the blots and performing an immunoblot for β-Actin.
band at 72 kDa was normalized to that of an invariant nonspecific high-molecular-weight band (>180 kDa) observed in all lanes.

Primary Human Keratinocytes Express
PPARγ. Studies in mouse skin have shown that PPARγ is expressed at low levels and may be functionally irrelevant in mouse keratinocytes [8,12]. We therefore first verified that PPARγ is expressed in the PHKs from several different individuals under several different culture conditions (culture on plastic or type I collagen) (Figure 1). To exclude the possibility that the PPARγ seen in these primary cultures was derived from other cell types, immunoblotting for PPARγ was also done on telomerase-immortalized human keratinocytes (N/TERT-1) [16]. This data supports previous reports that PHKs express mRNA and protein for PPARγ [19][20][21].

PPARγ Agonists, but not PPARα and PPARβ/δ Agonists, Stimulate PGE 2 Formation in PHKs through Both PPARγ-Dependent and -Independent Mechanisms.
In SZ95 sebocytes, we have previously demonstrated that GW9662 was effective in blocking PGE 2 production elicited by the synthetic PPARγ agonist ciglitazone [5]. We therefore examined whether ciglitazone could also induce PGE 2 formation in PHKs. Given that a TZD type of PPARγ agonist has been reported to induce COX-2 expression independent of PPARγ in a mouse keratinocyte cell line [22], we also examined whether the endogenous PPARγ ligand, azPC, could also induce PGE 2 production in PHKs. Finally, the ability of PPARα or PPARβ/δ to regulate PGE 2 production in PHKs was also examined. In Figure 2 azPC results in a significant increase in PGE 2 production. The PPARα agonist, WY-14,643, resulted in only a slight, but reproducible, increase in PGE 2 production. Treatment with a PPARβ/δ agonist (GW501516, 500 nM) did not alter PGE 2 production. It should be noted that NS398 significantly blocked both basal and PPARγ agonist-induced changes in PGE 2 production. As a percent of the vehicle control-treated cells, PGE 2 levels were 8.1 ± 7.4% (n = 4), 11.2 ± 12.5% (n = 4), and 12.2 ± 8.8% (n = 3) for NS398 treatment alone, ciglitazone + NS398, and azPC + NS398-treated cells, respectively. The above studies suggest that PPARγ, but not PPARα or PPARβ/δ, is coupled to PGE 2 production in PHKs. To determine whether ciglitazone-and azPC-dependent PGE 2 production was specific to PPARγ, we next examined the ability of GW9662 to block PPARγ agonist-induced PGE 2 production. As shown in Figure 2(b), pretreatment of PHKs with 1 μM GW9662 resulted in a significant inhibition of both ciglitazone-and azPC-induced PGE 2 production. However, this inhibition was not complete: GW9662 inhibited ciglitazone-and azPC-induced PGE 2 production by 59% and 62%, respectively. In contrast, GW9662 had no significant effect on vehicle-treated keratinocytes. Inasmuch as this same dose of GW9662 completely inhibited ciglitazone-induced PGE 2 formation in SZ95 sebocytes [5], this data suggests that PGE 2 production induced by azPC and ciglitazone occurs through both PPARγ-dependent and -independent mechanisms in PHKs. As noted above, the ability of NS398 to inhibit basal and PPARγ agonist-induced PGE 2 production suggests that PPARγ agonists induce expression of COX-2. We have previously demonstrated that PPARγ is coupled to UVB-and tertbutylhydroperoxide-mediated COX-2 expression in SZ95 sebocytes. This idea is supported by studies in SZ95 sebocytes in which the PPARγ antagonist GW9662 completely blocks PGE 2 formation induced by ciglitazone, UVB, and TBH [5]. Moreover, the COX-2 promoter is known to contain putative peroxisome proliferators response elements (PPREs) [23]. Thus, we next examined the ability of ciglitazone to induce COX-2 expression in PHKs. In Figure 2(c), we show that COX-2 mRNA expression is induced from 2 to 24 hours after ciglitizone addition. Moreover, GW9662 treatment suppressed COX-2 induction.

PPARγ Mediates UVB and Oxidative Stressor-Induced
We have previously demonstrated that UVB exposure activates PPARγ via the production of oxidized lipid species in KB epidermal carcinoma cells and SZ95 sebocytes [4,5]. Moreover, UVB-induced PPARγ activation is necessary for optimal UVB-mediated COX-2 expression or PGE 2 production in both KB cells and SZ95 sebocytes [4,5]. Finally, the lipid soluble oxidant, TBH, also induces COX-2 and PGE 2 production in SZ95 sebocytes through a PPARγ-dependent mechanism [5]. We therefore sought to establish the importance of PPARγ in photobiology by demonstrating that PPARγ antagonism could alter UVB-mediated induction of PGE 2 production in PHKs. As shown in Figure 3, both UVB and the lipid soluble oxidant TBH were capable of inducing a marked increase in PGE 2 production in PHKs. Importantly, pretreatment with the PPARγ-specific antagonist, GW9662, blocked 68% of the TBH and 80% of the UVB-mediated increases in PGE 2 . The ability of the selective COX-2 inhibitor, NS398, to inhibit PGE 2 formation was utilized as a negative control. As expected, NS398 markedly suppressed basal and UVBinduced PGE 2 production. The calcium ionophore, A23187, which is a potent activator of both COX-2 and upstream phospholipases, was utilized as an irrelevant positive control. As expected, GW9662 had no affect on ionophore-induced PGE 2 .
Finally, we examined the ability of GW9662 to inhibit UVB-induced COX-2 expression. In Figure 3(b), we show a representative immunoblot demonstrating that GW9662 strongly inhibits UVB-induced COX-2 expression (mean suppression of 61.5% in two separate experiments).

UVB-Induced PGE 2 Production Occurs via PPARγ in
Human Epidermal Explants. The above studies indicate that PPARγ antagonism suppresses UVB-and TBH-induced PGE 2 production and COX-2 expression in cultured PHKs in vitro. We next sought to determine whether PPARγ antagonism could alter UVB-induced PGE 2 formation in intact human skin ex vivo. In Figure 4(a), we show that skin explants exposed to 1,800 J/m2 of UVB exhibited a significant increase in PGE 2 production at 8 hours. Importantly, UVB-induced PGE 2 production was nearly completely blocked by pretreatment with 1 μM GW9662. A similar induction of PGE 2 production was observed 24 hours after UVB irradiation, which was also blocked by pretreatment with GW9662 (Figure 4(b)). Finally, as a negative control, we show that the COX-2 inhibitor, NS398, suppresses both control and irradiated PGE 2 levels. These studies show that GW9662 is effective topically and also verifies the importance of PPARγ in mediating a photobiological response in intact human skin.

Conclusions
Inasmuch as the epidermis is constantly exposed to ultraviolet light and other oxidative stressors, it is not surprising that epidermal cells would have developed intracellular mechanisms to detect and respond to these stresses. Among the many cellular signaling pathways that are induced by UV light, COX-2 induction and PGE 2 production have been well described [5,23,24]. We have recently demonstrated that UVB induces PGE 2 synthesis via the production of oxidativestress-induced PPARγ-specific ligands in other cell types [4,5]. Thus, PPARγ may serve as a cellular "signal transducer" that converts oxidative stress into cellular responses. In this study, we demonstrate that PPARγ agonists, including ciglitazone and azPC, are capable of inducing PGE 2 production in cultured PHKs. We then demonstrate that UVB-and TBH-induced PGE 2 production in human epidermis ex vivo and cultured PHKs in vitro occurs via a PPARγ-dependent mechanism. The role of COX-2 in PGE 2 production in PHKs was verified by demonstrating that GW9662 inhibits ciglitazone-induced COX-2 mRNA expression as well as UVB-induced COX-2 protein expression. Collectively, these findings strongly suggest that PPARγ acts as a photosensor (and likely an oxidative stress sensor) that acts to translate the insult into a biochemical signaling cascade. Previous work has generated conflicting data regarding the functional relevance of PPARγ in epidermal and keratinocyte biology. In mice, one group failed to observe PPARγ expression in cultured SKH-1 hairless mouse keratinocytes by RT-PCR [8]. In yet another study, PPARγ was observed to be expressed at the mRNA level in human primary keratinocytes, yet functional PPARγ activity was not observed using a PPRE-luciferase reporter assay [19]. However, another group demonstrated that PPARγ was expressed in keratinocytes isolated from SENCAR mice using immunoblot analysis [25]. Yet another study using PPARγ conditional knockout mice demonstrated that PPARγ is involved in epidermal differentiation [26]. In our studies, we demonstrate that PPARγ is expressed at the protein level in primary human keratinocytes and that PPARγ is coupled to UVB-and TBH-induced PGE 2 production. Our findings that PPARγ mediates UVB-and TBH-induced PGE 2 production provide additional evidence that PPARγ is functionally relevant to human cutaneous biology and, in particular, is relevant to human keratinocyte photobiology. This is particularly important given previous work by our group and others demonstrating that endogenous PPARγ ligand formation is induced by oxidative stress, including UVB [4,5,27,28].
While our studies indicate a clear role for PPARγ in regulating UVB-and TBH-induced PGE 2 production in PHKs, previous studies have shown that PPARα and PPARβ/δ are also expressed in human epidermis and cultured keratinocytes [1,20]. Though a PPARα agonist induced a small amount of PGE 2 formation in PHKs (Figure 2(a)), our data does not support a role for either PPARα or PPARβ/δ as major mediators of increased PGE 2 production in PHKs.
Recent studies have demonstrated that mice with hemizygous germline loss of PPARγ and epidermal-specific loss of PPARγ exhibit increased susceptibility to cutaneous chemical carcinogenesis [13,14]. This data indicates a possible role for PPARγ as a tumor-suppressing agent in skin. Studies are currently underway to examine the potential role of PPARγ in cutaneous photocarcinogenesis. It should be noted that recent studies by He et al. (2005) indicate that topical or systemic administration of TZD-type PPARγ agonists have no significant effect on chemical (DMBA/phorbol ester) or UVB-mediated carcinogenesis [12]. There could be several explanations for this discrepancy. First, we and others have previously shown that UVB or other oxidative stressors induce PPARγ ligand production from cellular glycerophosphocholines [4,7]. Moreover, phorbol esters act to induce oxidative stress, including the production of oxidized lipids in epidermis [29,30]. Thus, it is quite possible that PPARγ receptors are already engaged by endogenous PPARγ ligands produced through phospholipid oxidation, thus mitigating the effects of exogenous ligand. Alternatively, a followup study by He et al. (2006) indicated that TZD-type PPARγ agonists induce COX-2 expression in a keratinocyte cell line lacking PPARγ [22]. The authors speculate that this PPARγ-independent production of protumorigenic COX-2 may counteract any tumor-suppressing function of PPARγ. This PPARγ-independent induction of COX-2 would also explain the results seen in Figures 2 and 3, in which GW9662 was unable to completely abolish PPARγ agonist or UVBinduced PGE 2 production and COX-2 expression. Finally, the differences could potentially be explained by variations in the murine genetic background or carcinogenesis protocol.
In addition to a potential role for PPARγ in cutaneous photocarcinogenesis, independent lines of evidence suggest that PPARγ should be a focus for future studies examining its potential role as a mediator of phototherapeutic responses. Recently, PPARγ agonists have been shown to be effective antipsoriatic agents by reducing the keratinocyte hyperplasia associated with psoriasis [31]. This is significant as UVB exposure is a well-known treatment option for patients with psoriasis. Since psoriasis is thought to occur as a result of deregulated T-cell function [32,33], it is interesting that both PPARγ [34][35][36] and PGE 2 [37][38][39] act as T-cell and dendritic cell immunosuppressants and that topical application of PGE 2 has been shown to result in clinical improvement of psoriatic lesions [40]. Given our findings that UVB activates keratinocyte PPARγ, it is possible that some of the therapeutic benefits of UVB treatment may be due to PPARγ activation.
In conclusion, we provide evidence that PPARγ stimulates PGE 2 formation in PHKs. Moreover, PPARγ mediates UVB-and TBH-induced PGE 2 production in human epidermis and primary human keratinocytes. This suggests that PPARγ may play an important role in UVB-and oxidativestress-induced changes in epidermal biology.