Abstract
The skin-targeted overexpression of the glucocorticoid receptor (GR) in transgenic mice dramatically impairs the inflammatory responses to tumor promoter agents and suppresses skin tumor development. The antiinflammatory, rapid effects of corticosteroids are partially exerted through interference of GR with the phosphatidylinositol 3-kinase (PI3K)/Akt signaling pathway in several tissues, a highly relevant pathway in the mouse skin tumor progression process. In this work, we aimed to elucidate whether a cross-talk mechanism between GR and PI3K/Akt occurred in intact skin as well as the biological relevance of this interaction during skin tumorigenesis. We report that, in transgenic mice overexpressing the receptor, GR physically associated with p85α/PI3K in skin, resulting in decreased Akt and IκB kinase activity. GR activation by dexamethasone in normal mouse skin also decreased Akt activity within minutes, whereas cotreatment with the GR antagonist RU486 abolished dexamethasone action. Indeed, GR exerted a nongenomic action because keratinocyte transfection with a transcriptionally defective receptor mutant still decreased PI3K and Akt activity. Moreover, GR coexpression greatly reduced the accelerated growth of malignant tumors and increased Akt activity induced by Akt-transfected keratinocytes, as shown by in vivo tumorigenic assays. Overall, our data strongly indicate that GR/PI3K-Akt cross-talk constitutes a major mechanism underlying the antitumor effect of glucocorticoids in skin.
GLUCOCORTICOID HORMONES (GCS) exert pleiotropic effects by acting through the glucocorticoid receptor (GR), a ligand-induced transcription factor (reviewed in Ref. 1). GR can regulate gene expression through DNA binding-dependent and -independent mechanisms (reviewed in Ref. 2). The former requires ligand-induced dimerization of GR and binding to specific DNA palindromic sequences called glucocorticoid response elements, whereas the latter does not require DNA binding of GR, but rather is mediated through interference with other transcription factors, such as nuclear factor-κB (NF-κB) or activator protein 1 (reviewed in Ref. 3). In addition, it has been recently demonstrated that some rapid, antiinflammatory effects of corticosteroids are exerted through GR interference with the phosphatidylinositol 3-kinase (PI3K)/Akt signaling pathway (4). This cross-talk is mediated by hormone-bound GR association with the regulatory subunit of PI3K, p85α, and does not require DNA binding (4, 5).
GCs are potent inhibitors of keratinocyte proliferation and effective antiinflammatory compounds, which have been widely used for the treatment of a broad range of hyperproliferative and inflammatory skin disorders (6). Using a transgenic mouse model that overexpresses GR under the control of the keratin 5 promoter (K5-GR mice), we have previously shown that GR overexpression inhibits epidermal proliferation and abolishes the induction of proinflammatory cytokines in vivo (7). Furthermore, we demonstrated that GR plays a tumor suppressor role during mouse skin tumorigenesis, because K5-GR mice are highly resistant to tumor development (8). We have also reported that the PI3K and its downstream target Akt/protein kinase B are sequentially activated during mouse skin tumor progression induced by the two-stage carcinogenesis protocol (9). In this work, our goal was to elucidate whether a cross-talk mechanism between GR and PI3K/Akt occurred in skin as well as its biological relevance in skin tumorigenesis. We have used the K5-GR transgenic mouse model as a setting in which GR is constitutively overexpressed in skin (7, 8) and found physical association between GR and p85/PI3K, which resulted in decreased Akt activity. The activity of the downstream IκB kinase (IKK) was reduced in GR-overexpressing skin, and this could be due, at least partially, to decreased levels of the regulatory subunit IKK-γ.
The effect induced by GR overexpression was specific because topical treatment of normal mouse skin with the synthetic analog dexamethasone (Dex) also decreased Akt activity, whereas cotreatment with the GR antagonist RU486 abolished Dex action. This effect was nontranscriptional because down-regulation of the PI3K/Akt activity could be achieved by keratinocyte transfection with a transcriptionally defective GR mutant. Moreover, we have shown, by in vivo tumorigenic assays, that coexpression of GR along with Akt in PB keratinocytes was able to greatly reduce the increased rate of tumor growth induced by Akt alone, the reduction of Akt activity being a major mechanism underlying the antitumor effect of glucocorticoids.
RESULTS
K5-GR transgenic mouse skin was used as a model system with which to analyze the cross-talk between overexpressed GR and the PI3K/Akt signaling pathway. When we examined whole-cell skin extracts from K5-GR mice, we found that the overexpression of GR protein correlated inversely with phosphorylation of Akt (ser-473) without altering total Akt protein levels (Fig. 1A). To further confirm that decreased phosphorylated Akt (p-Akt) levels reflected decreased Akt activity, we performed in vitro kinase assays. Our results clearly demonstrated that increased GR expression in mouse skin reduced Akt activity by approximately 60% (Fig. 1A). We next examined whether GR interacted with p85α/PI3K in skin, as has been described previously for other tissues (4). By coimmunoprecipitation experiments, we showed that GR physically associated with the p85α subunit of PI3K in both transgenic and nontransgenic skin (Fig. 1B, left panel). In transgenic skin, we found higher levels of GR associated with p85α, thus indicating that GR activation, elicited through GR-targeted overexpression, favors GR/p85 association. Total p85α levels were similar in nontransgenic and transgenic skin (Fig. 1B, right panel). Because Akt is an upstream activator of the IKK, we attempted to investigate whether GR interference with PI3K/Akt signaling was able to modulate IKK activity. In vitro kinase assays showed decreased IKK activity in transgenic vs. nontransgenic skin, which correlated inversely with the amount of overexpressed GR (Fig. 2A, IVK). Furthermore, IKKα, IKKβ, and IKKγ protein levels were analyzed by immunoblotting showing decreased IKKγ protein levels in transgenic skin whereas IKKα and -β protein levels were unchanged (Fig. 2A, compare lanes 1–3 and lanes 4 and 5). Reduced IKK activity was in agreement with the observed decrease in the κB-binding activity of transgenic skin extracts as compared with nontransgenic samples in both newborn and adult skin (Fig. 2B, lanes 1–3 and 4 and 5). The κB-containing complexes were formed by p50/p65 and p50/p50 dimers, as determined by supershift experiments (Ref. 7 and data not shown).
GR Associates with p85/PI3K in Skin and Reduces Akt Activity A, Whole-cell extracts from nontransgenic (Co) and K5-GR transgenic (Tg) adult skin were analyzed by immunoblotting with specific antibodies against GR, Akt, and p-Akt (ser-473). In vitro kinase assays (IVK) were performed after immunoprecipitation of protein extracts with anti-Akt antibody and using histone H2B as a sustrate. Percentage of activity is shown. B, Coimmunoprecipitation assays were performed by incubating whole-cell extracts (WCE) from nontransgenic (Co) and K5-GR transgenic (Tg) skin with an anti-p85 antibody followed by immunoblotting with anti-GR. The amount of p85 in the immunoprecipitated (left panel) and the total amount of GR and p85 in WCE before immunoprecipitation is also shown (right panel).
IKK Activity and NF-κB Function Is Reduced in K5-GR Transgenic Skin A, Lanes 1–5 represent whole-cell extracts from nontransgenic (Co) and K5-GR transgenic (Tg) newborn (1, 2, 3 ) and adult (4, 5 ) dorsal skin. Immunoblotting was performed by using specific antibodies against GR, IKKα, IKKβ, and IKKγ. In vitro kinase IKK assay was performed with the extracts corresponding to lanes 1–5 immunoprecipitated with a mixture of anti-IKKα antibodies and using IκBα as substrate. B, EMSA was performed by incubating whole-cell extracts from newborn and adult mouse skin obtained from nontransgenic and transgenic mice (identical to extracts in lanes 1–5, panel A) with a labeled κB oligonucleotide. The composition of the retarded complexes is indicated at the left.
We next addressed whether ligand-induced activation of the receptor could specifically recapitulate the effect of overexpressed GR on PI3K/Akt signaling (Fig. 3). We topically applied 0.5 μg Dex to adult mice for 30 min and analyzed the amount of total and phosphorylated Akt by immunoblotting (Fig. 3A). Our results showed that, for the indicated doses, Dex was able to dramatically reduce p-Akt levels as compared with untreated skin (Fig. 3A). We observed a maximal effect on p-Akt with 1.6 μg Dex, and this effect was completely abolished by cotreatment with the GR antagonist, RU486 (Fig. 3A). Both p85 and GR protein levels remained unchanged upon Dex treatment (Fig. 3A). The kinetics of Dex treatment showed a pronounced decrease on Akt phosphorylation at 15 min, and it was maximal at 30 min (Fig. 3B). Our data indicate that GR interferes with the PI3K signaling pathway, as previously shown in endothelial cells; however, the outcome of such interaction in epithelial cells is a reduction in the Akt activity, suggesting that GR/PI3K cross-talk is inhibitory in skin.
Dex Topical Treatment Decreases Akt Activity in Skin A, Wild-type dorsal skin was treated with Dex (0.5 μg and 1.6 μg), combined Dex (1.6 μg) + RU486 (16.0 μg) for 30 min or left untreated. Skin whole-cell extracts were prepared and analyzed by immunoblotting by using the indicated antibodies. B, Time course of Dex effect on Akt phosphorylation. Wild-type skin was treated with Dex (1.6 μg) for the indicated times or Dex+RU486 (D+R) for 30 min. A representative experiment demonstrating Akt and p-Akt protein levels is shown.
Because the responses elicited by activated nuclear hormone receptors are rather complex, we tried to confirm that the observed interaction between GR and p85α-PI3K/Akt in skin in vivo could be recapitulated in the keratinocyte cell line PB. We performed transfection studies using PB cells and expression vectors for wild-type GR or a C-terminal zinc finger GR mutant, defective in transcriptional activation (10) along with the reporter plasmid mouse mammary tumor virus-chloramphenicol acetyltransferase (MMTV-CAT) (Fig. 4). We have used a GR dimerization mutant (i.e. unable to bind DNA and transcriptionally inactive, although competent in transrepression), previously reported by Helmberg et al. (10). As expected, GR overexpression induced CAT activity upon Dex treatment by approximately 5-fold whereas the mutant GR was unable to transcriptionally regulate the MMTV-CAT reporter (Fig. 4A). Expression levels for the wild-type and mutant GR were assayed by immunoblotting upon keratinocyte transfection (Fig. 4A). We next analyzed whether GR and p85α physically interacted in the absence/presence of the ligand in GR-transfected keratinocytes (Fig. 4B). We performed coimmunoprecipitation assays by using an anti-p85α antibody followed by immunoblotting with GR and observed that GR/p85 association was clearly increased upon Dex addition without changes in p85 protein levels (Fig. 4B). It appears as a feasible mechanism that activated GR favors the formation of GR/p85 complexes, thus excluding p85 from bona fide p85/p110 active complexes. We examined PI3K activity in GR-transfected keratinocytes by measuring the formation of phosphatidylinositol 3,4,5-triphosphate (PIP3), catalyzed by this enzyme, by in vitro kinase assays (Fig. 4C). Dex treatment elicited a reduction in PI3K activity, and this effect was independent of GR-regulated transcription because actinomicin D did not change the effect of ligand-activated receptor (Fig. 4C, lanes 1–3). In addition, the transcriptionally defective mutant GR was still able to reduce PI3K activity in response to Dex (Fig. 4C, lanes 4 and 5). We further examined the effect of GC treatment in GR-transfected keratinocytes on Akt activity and found reduced p-Akt levels (Fig. 4D, lanes 1 and 2). Of note, the transcriptional inhibitor actinomicin D did not change the Dex effect (Fig. 4D, lanes 1–3), and keratinocyte transfection with the transcriptionally defective mutant GR still decreased p-Akt levels upon Dex treatment (Fig. 4D, lanes 4 and 5). Altogether, our data indicate that activation of GR in keratinocytes, either in vivo or by keratinocyte transfection, interferes with PI3K/Akt signaling through a nontranscriptional action of the GR.
Ligand-Activated GR Decreases PI3K/Akt Signaling in Keratinocytes through a Nontranscriptional Action of the Receptor A, PB keratinocytes were transiently transfected with MMTV-CAT reporter along with expression vectors encoding for empty vector (pcDNA3), GR, or a transcriptionally defective mutant GR (mutGR). After depletion of endogenous GCs, cells were treated with 1 μm Dex for 6 h and CAT activity was determined. Expression levels for the wild-type and mutant GR upon transfection were assayed by immunoblotting (GR). B, GR-transfected keratinocytes were used for coimmunoprecipitation assays to show GR/p85 interaction in the absence/presence of Dex. Total levels of p85 and GR before immunoprecipitation are also shown. C, PI3K activity was determined in PB keratinocytes transfected with wild-type GR (lanes 1–3) or a transcriptionally defective mutant GR (lanes 4 and 5), in the absence/presence of Dex for 30 min, as indicated. Cells were preincubated with 1 μm actinomicin D (ActD) for 30 min before Dex treatment (lane 3). After thin layer chromatography, the product of the kinase reaction PIP3 was visualized by autoradiography and quantitated. Percentage of activity is shown. D, Whole-cell extracts as in panel C were analyzed by immunoblotting with antibodies against Akt and p-Akt.
We have recently demonstrated that GR acts as a tumor suppressor in skin (8), whereas Akt-dependent signaling is a major contributor to tumor progression and conversion in mouse skin carcinogenesis (9). In this context, we attempted to determine the consequences of GR/Akt interference in epidermal carcinogenesis by stable transfection of PB keratinocytes with the corresponding expression vectors followed by sc injection of transfectants into nude mice. We firstly assessed transfection efficiency by immunoblotting using specific antibodies against GR and Akt (Fig. 5A and data not shown) and then evaluated the consequences of GR and Akt overexpression on Akt phosphorylation (Fig. 5A). Akt-transfected keratinocytes showed increased p-Akt levels as compared with empty vector (Fig. 5A, compare lanes 1 and 4), in agreement with our previous observations (9). Remarkably, Dex treatment reduced the amount of p-Akt and this effect was ligand specific as assessed by cotreatment with RU486 (Fig. 5A, compare lanes 4–6). We did not detect any changes in p85α protein levels in any transfectant (Fig. 5A, lanes 1–6).
GR Overexpression in Keratinocytes Counteracts Akt-Induced Tumorigenesis through Reduced Akt Activity A, PB keratinocytes were transfected with expression vectors encoding for empty vector pcDNA3 alone (lane 1), GR (lanes 2 and 3) or Akt (lanes 4–6). After depletion of endogenous GCs, cells were treated with Dex, RU486, or combined Dex+RU486 for 1 h. Whole-cell extracts were prepared and analyzed by immunoblotting with the indicated antibodies. B, Kinetics of tumor growth. Transfected PB keratinocytes were injected sc into nude mice, and tumor volume was evaluated from d 28 to d 56 after injection. Tumor volume variations are shown for tumors originated from the indicated transfected keratinocytes from d 28 to d 39. C, Average tumor volume per genotype at harvesting. Data are shown as mean ± sd. Asterisks indicate statistically significant differences relative to Akt-transfected cells (P < 0.05). D, In vivo BrdU labeling was performed to determine the proliferation rate of all tumor groups, and percentage of BrdU-positive keratinocytes is indicated. Mean values ± sd are shown. Asterisks indicate statistically significant differences relative to Akt-transfected cells (*, P < 0.05; **, P < 0.01). E, Tumors arising from keratinocytes transfected with pcDNA3 alone (lane 1), GR (lane 2), Akt (lanes 3–5), or GR plus Akt (lanes 6–8) were collected and whole-cell extracts were prepared. Total Akt, p-Akt, GR, and p85 protein levels were determined by immunoblotting using specific antibodies.
To analyze the functional consequences of GR/Akt interference in carcinogenesis, PB keratinocytes transfected with pcDNA3, GR, Akt, or GR plus Akt were injected sc into nude mice. We subsequently evaluated the rate of tumor appearance as well as tumor growth and histopathological features (Fig. 5 and data not shown). The percentage of mice bearing tumors was determined by counting tumors larger than 3 mm at 28 d after injection. In agreement with our previous studies, vector-transfected control PB keratinocytes produced tumors in a minor percentage of the injected animals (22%, n = 18) whereas Akt-transfected keratinocytes gave rise to an increased tumor number (79%, n = 19). GR-transfected keratinocytes produced an average tumor number similar to vector-transfected cells (17%, n = 12). Remarkably, cotransfection of GR plus Akt in PB keratinocytes caused tumor growth only in a low percentage of the animals (23%, n = 13), demonstrating that GR could reverse to a great extent the tumor growth induced by Akt. Tumor volume was determined for all groups from d 28 until the day of harvesting, and kinetics of tumor volume growth (Fig. 5B) as well as average tumor volume per genotype at harvesting (Fig. 5C) is illustrated. Remarkably, tumor growth in Akt-transformed keratinocytes was delayed by approximately 1 wk as a consequence of GR overexpresssion, this difference being statistically significant (P < 0.05) at this stage (Fig. 5B). The average tumor volume in Akt-transformed keratinocytes increased almost 3-fold as compared with Akt/GR-transformed keratinocytes (P < 0.01) (Fig. 5C). Overall, our data indicate that, throughout all tumor stages evaluated, GR overexpression produced 2- to 4-fold reduction on tumor volume as compared with those tumors originated from Akt-transfected keratinocytes, and these differences were found to be statistically significant. Given that GR is an effective inhibitor of epidermal proliferation whereas Akt has been described as a positive modulator of proliferation (6, 11), we examined whether GR could counteract the proliferative effect induced by Akt (Fig. 5D). In vivo bromodeoxyuridine (BrdU) labeling experiments demonstrated that Akt increased tumor proliferation rate by 232% (P < 0.01) as compared with empty vector. Remarkably, GR was able to partially reverse the proliferative effect of Akt (152%, P < 0.05). We also used a three-tier system (12) to score the differentiation grade of tumors from each genotype and found no major differences among Akt- and Akt/GR-transformed keratinocytes (data not shown). Altogether, our data indicate that overexpression of GR exerts an antiproliferative role on mouse skin tumorigenesis induced by Akt overexpression, without altering differentiation of tumor cells.
To elucidate the molecular mechanisms underlying the antitumor role of GR on Akt-induced tumorigenesis, we analyzed the Akt activity in all tumors arising at different time points after cell injection (Fig. 5E and data not shown). Our data show that 93% of tumors from Akt-transfected cells exhibited increased p-Akt levels as compared with controls (Fig. 5E, lanes 3–5). Coexpression of GR along with Akt produced a marked decrease in p-Akt levels in 66.6% of tumors examined, and these results were consistently found throughout all time points of tumor collection (Fig. 5E, lanes 6–8). Tumors that arose from vector- and GR-transfected keratinocytes showed undetectable levels of p-Akt at any stage (Fig. 5E, lanes 1 and 2). We did not detect any changes in p85α protein levels in any experimental group of tumors (Fig. 5E). Collectively, our data provide evidence that GR greatly decreased tumor growth elicited by Akt in skin, mechanistically acting through interference with the PI3K/Akt signaling pathway. Our results thus highlight GR/PI3K-Akt interference as a major player mediating the antitumor action of GCs in transformed keratinocytes.
DISCUSSION
It is currently accepted that steroids exert their pleiotropic effects through genomic and nongenomic actions (13). According to the classic model of steroid action, hormones bind to intracellular receptors and subsequently modulate transcription and protein synthesis, thus triggering genomic events finally responsible for delayed effects. However, increasing evidence shows that very rapid effects of steroids are clearly incompatible with the genomic model (13). In the case of glucocorticoids, there is an enormous therapeutic relevance of these rapid effects, as recently demonstrated by the cardiovascular and neuroprotective effects of pharmacological doses of GCs (4, 5). This rapid action of GCs is exerted through cross-talk between GR and PI3K/Akt signaling pathways. Understanding the role of these interactions in skin carcinogenesis is key to developing therapies that might interrupt or reverse tumor onset and progression.
PI3K is a heterodimer composed of the 85-kDa regulatory subunit (p85) and the catalytic 110-kDa (p110) α- and β-subunits. PI3K plays a pivotal role in several cell signaling networks, including cell cycle progression, differentiation, survival, invasion, metastasis, and angiogenesis (for a review, see Ref. 14). The mechanisms underlying p85/p110 PI3K activity are not fully known but involve the recruitment of proteins containing Src-homology 2 and pleckstrin homology domains to the cell membrane by the lipid product of PI3K, PIP3, where they are activated (reviewed in Ref. 15). Recently, the involvement of PI3K signaling in carcinogenesis has been highlighted not only by the well characterized activity of PTEN [phosphatase and tensin homolog (mutated in multiple advanced cancers 1)] as a tumor suppressor, but also because mutated forms of p85 have been found in human cancers (16).
Several biological effects of PI3K are mediated through activation of its downstream target Akt/protein kinase B. The serine/threonine kinase Akt is implicated in cancer progression because it stimulates proliferation and suppresses apoptosis. Our previous work has shown that PI3K and Akt are sequentially activated during mouse skin tumor progression induced by the two-stage carcinogenesis protocol (9). Akt can activate the transcription factor NF-κB (17) and, accordingly, we have also demonstrated that NF-κB is constitutively activated during mouse skin carcinogenesis, starting at the middle stage of tumor promotion (18). It is assumed that the antiinflammatory action of GCs is mediated mostly through GR interference with NF-κB and activator protein 1 transcription factors (19, 20). We have recently provided evidence of the tumor suppressor role of GR in skin tumor development by using an experimental model of ras-mediated tumorigenesis in a K5-GR transgenic background (8). One of the molecular mechanisms underlying the antiinflammatory and antitumor effects of GR involves cross-talk between this hormone receptor and NF-κB, resulting in blockage of NF-κB activity (7, 8). Remarkably, a number of reports demonstrated that Akt activates IKK, leading to IκBα degradation and NF-κB nuclear translocation (21), but also eliciting an alternative pathway that results in phosphorylation and activation of the NF-κB subunits p50 and p65 (22). In this context, we have shown that GR interference with PI3K/Akt signaling is able to modulate IKK activity, at least partially, through down-regulation of IKKγ, concomitantly with reduced NF-κB activity (Fig. 2). It is worth noting that we have recently shown in the K5K10 transgenic mouse model, which is also resistant to skin-induced tumorigenesis, that impaired Akt activation leads to decreased NF-κB signaling, partly through down-regulation of IKKβ and IKKγ (23).
The corticosteroid dose required to achieve therapeutic effects is probably dependent on the cell type, as demonstrated by recent reports (5). These studies have shown that only pharmacological concentration of corticosteroids stimulated PI3K through physical association between the steroid hormone receptor and p85α (4) and had neuroprotective effects, as reported for the estrogen receptor (24). When more physiologically relevant doses of Dex were used in similar experiments, they did not alter p-Akt levels and concomitantly had no protective effects (4, 5). It remains unclear how this nontranscriptional action of GR, although specific, is exclusively achieved through pharmacological doses of corticosteroids (5).
In the in vivo experiments reported here, we show that treatment with physiological doses of Dex reduced Akt-phosphorylated levels and overall Akt activity in skin, and this effect could be reversed by cotreatment with the GR antagonist RU486 (Fig. 2), thus indicating the specificity of low-dose GC action. We have also analyzed the consequences of topical skin treatment by using high doses of Dex (40 μg) and found no effect on p-Akt levels as compared with untreated skin (data not shown). Our data strongly suggest that the corticosteroid doses required for GR interference with the PI3K/Akt pathway may depend on tissue-specific factors. In skeletal muscle, Dex blocked IGF-I-induced phosphorylation of Akt, a PI3K-dependent process, through transcriptional up-regulation and increased cellular abundance of the p85α subunit (25). Overabundant p85α monomer competes for IGF-receptor sustrate-1 binding with p85/p110 PI3K heterodimers, inhibiting approximation of the p110 catalytic subunit to the membrane surface and thus retarding PI3K activity (26). In our studies, we have not observed increased levels of p85 protein as a consequence of Dex treatment within the observed time course (up to 3 h) (Fig. 2 and data not shown). A potential drawback of this study is that GR overexpression may exceed the capacity of the chaperones and force the interaction with other molecules, such as PI3K. However, our results indicate that the association between GR and PI3K is physiologically relevant because it normally occurs in mouse skin and mouse keratinocytes. Moreover, either the skin-targeted overexpression of GR in K5-GR mice or the addition of exogenous hormone increased GR/PI3K association, as shown by coimmunoprecipitation assays (Figs. 1 and 4).
Importantly, our experiments demonstrate that ligand-activated GR exerts nontranscriptional regulation on the PI3K/Akt pathway as shown by the time course and specificity of physiological doses of Dex on Akt function (Fig. 3) as well as the outcome of the experiments using actinomicin D and the transcriptionally defective GR mutant (Fig. 4), as recently reported (4, 5).
GCs have broad and often controversial effects, depending on the cell type and environmental context (26). In a number of cell types, GCs have been described as potent antitumor agents. In rat glioma cells, treatment with GCs leads to complete reversion of their transformed phenotype and loss of their tumorigenic potential (27). Moreover, the administration of Dex is a standard procedure for glioblastoma patients, and recent studies of retroviral-mediated gene transfer into experimental gliomas highlight the relevance of balanced GC levels for gene therapy protocols (28). In other endocrine types of cancer, such as prostate cancer, it has been reported that GCs potentiate the antitumor effect of 1,25-dihydroxycholecalciferol (vitamin D) (29). Recently, phase I and II clinical trials of vitamin D in combination with Dex have been initiated in patients with prostate cancer (30). At the molecular level, vitamin D targets several potential markers of the apoptotic pathway in prostate cell models and elicits the reduction of phosphorylated levels of Akt (29).
GCs are the most potent inhibitors of tumor promotion in mouse skin, when applied with a promoting agent at the early stages of promotion. By performing in vivo tumorigenic assays, we have addressed the specific interplay between GR and PI3K/Akt in a pathological setting. Our results unequivocally showed that GR overexpression was able to counteract the proliferative effects induced by PI3K/Akt, thus highlighting the importance of this biological antagonism in skin carcinogenesis. PI3K has been shown to be the target of various antitumor drugs, and specifically inhibiting Akt activity may be a valid approach to treat cancer and increase the efficacy of chemotherapy.
MATERIALS AND METHODS
Animals and in Vivo Tumorigenic Assays
K5-GR hemizygous mice of line 285 (B6D2 mixed genetic background) were previously described (7). K5-GR and wild type (B6D2/F2) female mice (8 to 10 wk old) were used for experiments (n =12), animals were killed, and skin was collected. For topical treatments, mouse dorsal skin was shaved 2 d before and then treated with the indicated doses of Dex (Sigma Chemical Co., St. Louis, MO), RU486 (BIOMOL Research Laboratories, Inc., Plymouth Meeting, PA), or acetone as a vehicle. Experiments were performed in duplicate and at least four animals per point were used.
For in vivo tumorigenic experiments, pooled PB keratinocyte clones (>100 clones) were grown, trypsinized, and injected sc into outbred Hsd:Athymic 6-wk-old female nu/nu mice (Harlan Interfauna Iberica, Barcelona, Spain). Each mouse was injected in both flanks with PB/pcDNA3 (right) and PB/pcDNA3-GR (left), PB/Akt (right) and PB/GR-Akt (left) or PB/Akt (left) and PB/pcDNA3 (right). The number of mice injected was 12–18 per transfection group, and the incidence and size of tumors were recorded twice weekly. Animals were killed at 6–9 wk (depending on tumor size). Tumors were harvested at 39–64 d after animals were injected, and tumor volumes were estimated by the formula π/6 ab2, a and b being major and minor axis of the tumor, respectively.
All animal experimentation was conducted in accordance with accepted standards of humane animal care and complies with international guidelines.
Cell Culture, Transfection, and Treatments
PB keratinocytes were grown in DMEM (BioWhittaker, Inc., Walkersville, MD) supplemented with 10% fetal calf serum (BioWhittaker, Inc.). Transfections were performed by the calcium phosphate method by using the indicated plasmids (9). The C-terminal zinc finger mutant GR has been previously reported as a transcriptionally defective receptor by Helmberg et al. (10). For transient transfection, PB keratinocytes were transfected for 6 h with 5 μg MMTV-CAT along with 5 μg pcDNA3, 5 μg GR, or 5 μg mutGR (10). Cells were charcoal stripped for an overnight incubation followed by Dex treatment (1 μm) for 30 min or incubation with actinomicin D (1 μm) for 30 min before Dex treatment. Cells were washed, and whole-cell extracts were prepared and CAT activity was assayed by the CAT ELISA kit (Roche Molecular Biochemicals, Indianapolis, IN) according to the manufacturer’s recommendations. Transfection experiments were performed in triplicate, and mean value ± sd is indicated.
For stable transfectants, PB/pcDNA3 and PB/Akt clones were selected by using G418 (0.15 mg/ml). These cell populations were subsequently cotransfected with K5-GR and pGL-VP(H) plasmids, and transfectants were selected by using G418 (0.15 mg/ml) and hygromicin (0.3 mg/ml). Steroid hormones and growth factors were removed from fetal bovine serum by charcoal stripping. Cells were treated with vehicle, 1 μm Dex, or 10 μm RU486 for 1 h.
Immunoblotting
Whole-cell protein extracts were prepared as previously described (7), separated by SDS-PAGE, and transferred to Hybond membrane (Amersham Pharmacia Biotech, Arlington Heights, IL). Membranes were stained with Ponceau S (Sigma Chemical Co.) to verify equal protein loading and transfer. Bands were visualized using Amersham enhanced chemiluminescent reagent and Hyperfilm (Amersham Pharmacia Biotech). The antibodies against GR (sc-1004), Akt (N-19), IKKα (sc-7182), and IKKγ (sc-8330) were purchased from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA); p-Akt (ser-473) was obtained from Cell Signaling Technology, Inc. (Beverley, MA); p85 antibody was from Upstate Biotechnology, Inc. (Lake Placid, NY), and IKKβ antibody was obtained from Imgenex (San Diego, CA) (IMG-19). Secondary rabbit, goat, or mouse peroxidase-conjugated antibodies were purchased from Amersham Pharmacia Biotech.
In Vitro Kinase Assays
Akt activity was determined in mouse skin after immunoprecipitation with anti-Akt antibody (1 μl/25 μg protein) using histone 2B as sustrate (H2B; Roche Molecular Biochemicals) as previously described (9). The Akt activity was quantitated using a phosphor imager (Bio-Rad Laboratories, Inc., Hercules, CA) and expressed as percentage of the activity. The phosphor imager values for control samples (normal skin) were arbitrarily set as 1, and the other experimental groups were expressed as relative to control values. IKK activity was assayed in mouse skin after immunoprecipitation with a mixture of different IKKα antibodies (Santa Cruz Biotechnology, Inc.) and using full-length IκBα (Santa Cruz Biotechnology, Inc.) as substrate, as previously reported (31). PI3K activity was determined in PB keratinocytes (150 μg whole-cell extracts) using l-α-phosphatidylinositol (Sigma Chemical Co.) as substrate, as previously described (9). After thin layer chromatography on Silica gel plates (Merck & Co., Inc., Rahway, NJ), 32P-labeled phospholipids were visualized by autoradiography, and PIP3 was quantitated using a phosphor imager and expressed as percentage of activity. The phosphor imager values for control samples (untreated mouse PB keratinocytes) were arbitrarily set as 1, and the other experimental groups were expressed as relative to control values.
EMSAs
EMSA were performed by incubating whole-cell extracts from newborn and adult mouse skin obtained from nontransgenic and transgenic mice with a labeled oligonucleotide corresponding to a palindromic κB site, as previously described (7). The sequence of the κB oligonucleotide coding strand was: 5′-GATCCAACGGCAGGGGAATTCCCCTCTCCTTA-3′. The composition of the retarded complexes was determined by supershift experiments, as reported previously (7).
BrdU Immunostaining
BrdU (Roche Molecular Biochemicals, 6 mg/kg of body weight) was injected ip 1 h before animals were killed. Tumors and surrounding skin were fixed in 70% ethanol before processing. Sections (4 μm) were deparaffinized and incubated with anti-BrdU monoclonal antibody (Roche Molecular Biochemicals) followed by antimouse peroxidase-conjugated antibody (Amersham Pharmacia Biotech). Immunostaining was visualized with the Avidin-Biotin-Complex (ABC) kit (Vectastain Elite, Vector Laboratories, Inc., Burlingame, CA) as previously described (7). Sections were counterstained with hematoxylin and analyzed by light microscopy. The number of BrdU-positive keratinocytes was calculated as percentage of more than 2000 cells counted per sample. The number of tumors analyzed for each transfection group was: six (pcDNA3), five (pcDNA/GR), 13 (Akt), and 11 (GR/Akt). The differences between values obtained for Akt and GR/Akt were statistically significant (P < 0.05) when compared by a Mann-Whitney (Wilcoxon) W test.
Acknowledgments
We thank C. Caelles for the kind gift of the plasmid mutGR. We acknowledge Ana Llopis for expert technical assistance, Juan Luis Fernández Masip and Eva Donet for help with tumor processing, Jesús Martinez for help with the generation of tumors, and personnel at the Animal Facility of the Instituto de Biomedicina de Valencia and Centro de Investigaciones Energéticas, Medioambientales y Tecnológicas for animal care.
This work was supported by Grants SAF2002-04368-C02-01, SAF2002-04368-C02-02, and SAF2002-01037 from the Spanish Ministry of Science and Technology and Grant 08.1/0054/2001 from Comunidad Autonoma de Madrid.
H.L. and A.P. contributed equally to this work and should both be considered first authors.
Abbreviations:
- BrdU,
Bromodeoxyuridine;
- Dex,
dexamethasone;
- GCs,
glucocorticoid hormones;
- GR,
glucocorticoid receptor;
- IKK,
IκB kinase;
- MMTV-CAT,
mouse mammary tumor virus-chloramphenicol acetyltransferase;
- p-Akt,
phosphorylated Akt;
- NF-κB,
nuclear factor κB;
- PI3K,
phosphatidylinositol 3-kinase;
- PIP3,
phosphatidylinositol 3,4,5-triphosphate.
