Inhibition of Epidermal Growth Factor Receptor Signaling Elevates 15-Hydroxyprostaglandin Dehydrogenase in Non–Small-Cell Lung Cancer

Non–small-cell lung cancer (NSCLC) is the leading cause of cancer-related death in the United States, with low 5-year survival rates of 12% to 15% (1). Prostaglandin E₂ (PGE₂), one of the major metabolites of cyclooxygenase-2 (COX-2), is overproduced in many human solid tumors, including lung cancer, and has been associated with increased tumor angiogenesis, metastasis, apoptosis, cell cycle regulation, and immune suppression (2). Intratumoral PGE₂ levels depend not only upon the rate of production but also on the rate of its degradation. The enzyme responsible for the degradation of PGE₂ is 15-hydroxyprostaglandin dehydrogenase (15-PGDH; ref. 3), and recent publications have shown that down-regulation of 15-PGDH is associated with multiple cancer types (4–7). Quantitative reverse transcription-PCR (RT-PCR) and Northern blots comparing matched normal and tumor tissues suggest that down-regulation is occurring at the transcriptional level (6, 7). Interestingly, treatment of colon carcinoma cells with epidermal growth factor (EGF) reduced the PGDH protein levels in some cell lines (6). In this same study, treatment with the EGF receptor (EGFR)–specific inhibitor erlotinib negated the effects of EGF and increased the protein levels of PGDH above control colon cancer cells.

In our examination of NSCLC, we show that 15-PGDH protein and mRNA are down-regulated in multiple subtypes of NSCLC. Notably, erlotinib treatment of NSCLC significantly elevates 15-PGDH protein levels in subset of NSCLC cell lines. We show that the mechanism for this up-regulation operates through the mitogen-activated protein kinase (MAPK) pathway. Interestingly, treatment with either erlotinib or a MAPK kinase (MEK) inhibitor decreases transcript levels of the transcriptional repressors ZEB1 and Slug in an erlotinib-responsive cell line but not in a cell line where 15-PGDH levels remain unchanged with erlotinib treatment. Although both ZEB1 and Slug expression is reduced, we were only able to show Slug binding to the PGDH promoter. In addition, only Slug had the ability to regulate transcriptional activity of the PGDH promoter. These results strongly imply that the transcriptional repressor Slug is important for the regulation of 15-PGDH in a subset of NSCLC.

Reagents. Erlotinib was generously provided by Genentech. U0126, trichostatin A (TSA), and anti-β-actin were purchased from Sigma-Aldrich. LY294002 was purchased from Calbiochem. Anti-COX-2 was purchased from Cayman Chemical, and anti-E-cadherin, anti–phosphorylated extracellular signal-regulated kinase (anti-pERK), and anti-ERK were from Cell Signaling Technologies.

Cell lines. NSCLC cell lines were maintained in RPMI 1640 containing 10% fetal bovine serum (FBS; Hyclone). Cos7 and NIH3T3 cell lines were grown in DMEM containing 10% FBS or calf serum (Hyclone), respectively. Medium was supplemented with 100 units/mL penicillin and 100 μg/mL streptomycin. Cells were maintained in a 5% CO₂ atmosphere at 37°C.

Constructs. The pCMV5 vector, pCMV5N vector, and pCMV5-murine ZEB1 (mZEB1) constructs were kindly provided by Scott Hiebert (Vanderbilt University, Nashville, TN). mZEB1 was released from pCMV5 with BamHI and subcloned into pFLAG-CMV-2 (Sigma) to generate FL-mZEB. pCMV-Sport6-human Slug and pPCR-BluntII-TOPO-human ZEB1 constructs were purchased from Open Biosystems. These constructs were sequence from the 5′ and 3′ ends to verify the presence of an insert and the identity of that insert. Human Slug was released from pCMV-Sport6 with BamHI and placed into pFLAG-CMV-2 to create FL-hSlug. A NotI/BamHI digest removed human ZEB1 from pPCR-BluntII-TOPO, which was subsequently subcloned into pCMV5N. pCMV5N is a modified pCMV5 vector in which a NotI linker was inserted into the MluI site. To create FL-hSlug, a PCR product was generated using pCMV-Sport6-hSlug as a template and the following primers: hSlug165F, 5′-GACGGATCCACCATGCCGCGCTCCTTCCTGGTCAAG-3′; hSlug971R, 5′-GACGGATCCTCAGTGTGCTACACAGCAGCCAGA. The human PGDH promoter region from bases −626 to +1 was amplified by PCR using the genomic DNA of NSCLC cell line H3255 with primers PGDHpr-626F (5′-CAGACGCGTCAGGTTTTCAGGACAGAGCAGAG-3′) and PGDHpr+1R (5′-CAGCTCGAGTGGTGCAGCCACTGCTGGG-3′). The resulting product was subcloned into the MluI and XhoI cut pGL3 basic (Promega) to create PGDH/pGL3. All PCR products were sequenced to verify that there were no PCR-generated mutations.

Microarray analysis. Primary lung cancers were obtained with informed consent from 49 patients (23 female and 26 male; median age, 64.8 years; range, 35–84 years) who underwent thoracotomy at Vanderbilt University Hospital. All clinical and pathologic information was obtained from medical records. Each tumor was classified by histologic subtype according WHO classification (8) by an expert pulmonary pathologist. All samples were frozen immediately after resection and stored in liquid nitrogen until extracting RNA with an RNeasy Micro kit according to manufacturer's instructions with Dnase I treatment. All the steps related to the microarray analysis were carried out in Vanderbilt Microarray Resource Core Lab with Affymetrix Human Genome U133 Plus2.0 chips (Affymetrix). After obtaining microarray data through GCOS software (Affymetrix), Gene traffic (Stratagene) was used to normalize data with the RMA method by taking average of a pooled, normal lung, total RNA control (BD Clontech) and nine normal tissues as the baseline.

Treatment of cell lines with EGF and inhibitors. For all cell culture experiments, erlotinib, U0126, and LY294002 were dissolved in 100% DMSO. Subconfluent cultures were serum starved overnight followed by addition of EGF (100 ng/mL). One hour after EGF addition, inhibitors were added. The cells were then harvested at various time points after treatment and lysed with modified radioimmunoprecipitation assay buffer (1× PBS, 0.5% sodium deoxycholate, 0.1% SDS, 1% NP40, and protease inhibitors). Erlotinib, U0126, and LY294002 treatments were allowed to proceed for 10 to 12 h before cells were harvested. For TSA treatment, A549 and H1435 cells at ∼70% to 80% confluence were treated with vehicle (DMSO) or 300 nmol/L TSA for 20 and 48 h. Cells were harvested, and PGDH protein levels were examined by Western blot as described.

Western blots. Proteins were separated on 4% to 20% or 10% SDS polyacrylamide gels and transferred to Hybond-ECL nitrocellulose (GE Healthcare Biosciences) or Immobilon P (Millipore). Membranes were then blocked in 5% milk in TBS, 0.1% Tween 20 and incubated with primary antibody (15-PGDH, 1:15,000; pERK, 1:1,000; ERK, 1:1,000; and E-cadherin, 1:1,000) overnight at 4°C. The β-actin antibody (1:5,000) was typically incubated for 1 h at room temperature. The membranes were then treated with horseradish peroxidase–conjugated secondary antibody and developed using chemiluminescent detection reagent (Pierce Chemical).

PGE₂ assay. PGE₂ was measured in culture medium by enzyme immunoassay per the manufacturer's instructions (Cayman Chemical). Briefly, 2.5 × 10⁵ cells were seeded per well into six-well dishes and allowed to grow overnight. Cells were serum starved for 16 to 20 h followed by addition of EGF (100 ng/mL). One hour after addition of EGF, erlotinib or vehicle was added. Ten hours after addition of EGF, arachidonic acid was added to a final concentration of 0.1 mmol/L. Two hours later, medium was harvested for PGE₂ assay, and cells were harvested for protein determination by detergent-compatible protein assay (Bio-Rad). The amount of PGE₂ in the medium was normalized to protein. Cell lysates were examined by Western blot as described to determine protein levels of COX-2 and PGDH.

Real-time RT-PCR. Total RNA was extracted from cell lines with a VersaGene RNA isolation kit (Gentra Systems). cDNA was synthesized using an Invitrogen Superscript III First-strand synthesis system (Invitrogen). For real-time PCR, relative gene expression was determined using QuantiTect SYBR Green PCR kit (Qiagen) on a Bio-Rad iCycler (Bio-Rad Laboratories). Primer pairs specific for ZEB, Slug, Snail, glyceraldehyde-3-phosphate dehydrogenase (GAPDH), and PCR reaction conditions are as follows: ZEB1F, 5′-AGCAGTGAAAGAGAAGGGAATGC-3′; ZEB1R, 5′-GGTCCTCTTCAGGTGCCTCAG-3′; SlugF, 5′-AATATGTGAGCCTGGGCGC-3′; SlugR, 5′-CTCTGTTGCAGTGAGGGCAAG-3′; SnailF, 5′-CGCGCTCTTTCCTCGTCAG-3′; SnailR, 5′-TCCCAGATGAGCATTGGCAG-3′; GAPDHF, 5′-TGCACCACCAACTGCTTAGC-3′; GAPDHR, 5′-GGCATGGACTGTGGTCATGAG-3′; 40 cycles at 95°C for 30 s, 60°C for 30 s, and 72°C for 30 s, and 75°C for 10 s. The quantitative RT-PCR for each treatment was done in triplicate, and the average C_t was determined. The relative gene expression, normalized to GAPDH and based on three separate experiments for H1435 and five experiments for A549, was calculated using the 2^-ΔCt method as described in Livak and Schmittgen (9).

Electrophoretic mobility shift assay. Cos7 cells were transfected with either pFLAG-CMV-2 (Sigma) or with FL-mZEB1 or FL-hSlug using LipofectAMINE (Invitrogen) following the manufacturer's instructions. Forty-eight hours after transfection, the nuclear lysates were prepared by the method of Schreiber et al. (10). Three oligonucleotides that span each of three E-boxes in the 15-PGDH promoter were made according to the sequences described in Mann et al. (5). Oligonucleotides were end-labeled using T4 polynucleotide kinase (New England Biolabs) and purified using the Nucleotide Removal kit (Qiagen) following the manufacturer's instructions, with the exception that the oligonucleotides were eluted from the spin column in 100 μL of 10 mmol/L Tris (pH 8). Sense and antisense labeled oligonucleotides were added together and allowed to anneal at room temperature for 20 to 30 min before adding to binding reactions. Unlabeled competitors were diluted in 10 mmol/L Tris and added together and allowed to anneal as described for the labeled oligonucleotides. Unlabeled competitor, specific or nonspecific at 50-fold molar excess, or antibodies, anti-FLAG (Sigma) or anti-HA (Covance, Denver, PA), 2 μg of each, were added to the binding reactions 5 min before the labeled oligonucleotides and allowed to incubate at room temperature. The 1× binding buffer was 20 mmol/L HEPES (pH 7.9), 60 mmol/L KCl, 0.1 mmol/L EDTA, 2 mmol/L DTT, 10% glycerol (Sigma), and 1 mg poly(deoxyinosinic-deoxycytidylic acid) (GE Healthcare). Following addition of the labeled oligonucleotides, the reactions remained at room temperature for 30 min. Reactions were added directly to a 5% non-denaturing acrylamide gel for separating complexes.

Transcription assay. PGDH/pGL3 was cotransfected into NIH3T3 cells with TK-pRL (Promega) and varying amounts of pFLAG-CMV-2 (Sigma), pCMV-Sport6-hSlug (Open Biosystems), FL-mZEB, or pCMV5-hZEB so that the total DNA amount remained constant at 2 μg. Transfections were done in six-well dishes (Becton Dickinson) using Superfect reagent (Qiagen) following the manufacturer's instructions. Cells were harvested 48 h after transfection and assayed for luciferase activity using the Dual Luciferase Reporter Assay System (Promega) following the manufacturer's protocol.

Statistical analysis. Data were analyzed by the Student's t test. All data were expressed as means ± SE, and differences were considered statistically significant when the P < 0.05.

The level of PGE₂ within a tissue is determined by a fine balance of synthesis and degradation (6). NAD⁺-dependent 15-PGDH is an enzyme that catabolizes prostaglandins and is the main enzyme responsible for the elimination of PGE₂ from tissues (3). To investigate whether 15-PGDH is down-regulated in NSCLC, we examined 15-PGDH protein levels in 39 lung cancer cell lines and 6 paired normal lung versus lung tumor tissues by Western blotting (Fig. 1A and B; Supplementary Table S1). Representing all NSCLC subtypes, we found that 36 of these 39 cell lines have dramatically decreased 15-PGDH protein, with only three cell lines expressing modest levels of 15-PGDH. In addition, the 15-PGDH protein level was clearly decreased in the tumor tissue of the six paired normal versus tumor patient samples independent of the histology (3 adenocarcinoma and 3 squamous cell carcinoma; Fig. 1B). Comparison of the PGDH transcript levels in a pooled normal lung RNA sample (BD Clontech) and nine normal lung tissues to 49 freshly resected NSCLC specimens using Affymetrix microarray analysis revealed that almost all of the tumors expressed lower levels of 15-PGDH mRNA (Fig. 1C). These data show that 15-PGDH is frequently and significantly down-regulated in all subtypes of lung cancers, suggesting that loss of 15-PGDH is an early and important event, and that this protein may have tumor suppressor activity (5–7).

Several lines of evidence are increasingly linking prostaglandins with modulation of EGFR signaling. Therapeutic targeting of the EGFR with EGFR tyrosine kinase inhibitors, such as erlotinib, has achieved beneficial results in unselected NSCLC patients (11). This has included dramatic responses, most often observed in patients with EGFR mutant tumors, as well as clinically significant stabilization of disease, often observed in patients without EGFR mutant tumors. Because of the previous work detailing the effects of EGF and erlotinib on 15-PGDH protein and message levels in colon cancer cell lines (6), we hypothesized that a portion of the benefit from erlotinib in lung cancer might come from modulation of PGE₂ levels as a consequence of elevation of 15-PGDH expression. We therefore examined the effects of erlotinib on 15-PGDH expression in seven human lung cancer cell lines. We found that treatment with erlotinib significantly increased 15-PGDH protein expression in a number of the treated cell lines, which included H1435, H1648, and H3255 cells (Fig. 2A,, top; Supplementary Fig. S1A). The A549 and H2122 cell lines are examples of NSCLC cell lines that do not up-regulate 15-PGDH in response to erlotinib treatment (Fig. 2A , bottom; Supplementary Fig. S1B). The average overall increase in 15-PGDH protein expression was 2.5-fold (Supplementary Fig. S1C). These data show that inhibition of EGFR pathway in NSCLC significantly elevated 15-PGDH protein levels in a subset of human lung cancer cells. Thus, alteration in the prostaglandin metabolic pathway may play a role in the benefits of erlotinib treatment.

We next determined the mechanism by which inhibition of the EGFR was increasing 15-PGDH levels. The downstream signaling pathways from the EGFR, including the phosphatidylinositol-3 kinase/protein kinase B (PI3K/AKT) and the Ras/ERK kinase signaling cascades, have been well characterized and are known to have prominent roles in cell survival and cell proliferation, respectively. By blocking each of these pathways with specific inhibitors, we hoped to determine which pathway was driving the increase of 15-PGDH expression. H1435 cells treated with EGF were incubated with specific inhibitors that block EGFR, AKT, or ERK signaling (Fig. 2B,, top). In cells treated with the PI3K inhibitor LY294002, the 15-PGDH protein level was similar to control, indicating that this pathway did not take part in the up-regulation of 15-PGDH. On the other hand, the small molecule inhibitor U0126, which blocks ERK activity by inhibiting the upstream MEK kinase, clearly caused an increase in the level of 15-PGDH protein similar to that seen with erlotinib. The treatment of A549 cells with the same inhibitors did not produce any changes in the levels of 15-PGDH (Fig. 2B , bottom). These data indicate that the decrease in the 15-PGDH protein expression level is associated with activation of the Ras/MEK/ERK kinase pathway.

The induction of 15-PGDH protein expression by erlotinib treatment should result in the reduction of PGE₂. However, despite the fact that 15-PGDH levels remain unchanged in A549 cells upon erlotinib treatment, both the H1435 and the A549 cell lines showed a reduction in the amount of PGE₂ in the culture medium (Fig. 2C). Because EGF stimulation has been shown to result in the induction of the COX-2 promoter (12), we examined the COX-2 protein levels in each cell line. For the A549 cells, COX-2 protein levels seem to decrease ∼4-fold. Thus, the lowered PGE₂ levels would seem to be solely due to decreased COX-2 levels (Fig. 2D), as 15-PGDH levels do not change. However, H1435 cells, which show a slight reduction in COX-2 protein levels (about 2-fold; Fig. 2D), have a greater overall reduction in PGE₂ in the medium (Fig. 2C). This difference is most likely due to the induction of 15-PGDH in the H1435 cells.

Although the use of the inhibitors clearly identified the ERK signaling pathway as a regulator of 15-PGDH protein levels, the molecular mechanism behind the down-regulation of 15-PGDH remained to be determined. Recently, it had been shown in a colon carcinoma cell line that ERK activation increased the intracellular protein levels of the transcriptional repressor Slug, which then down-regulated E-cadherin (13). E-cadherin protein levels were restored, and Slug protein levels reduced by treatment with the MEK inhibitors PD98059 and U0126. To test the possibility that this mechanism was operating in the H1435 cell line, we examined E-cadherin protein expression (Fig. 3A,, top). We found that both erlotinib and U0126 treatment increased the levels of E-cadherin in the H1435 cells, providing strong evidence that transcriptional repression was occurring through EGF stimulation of the MEK/ERK signaling pathway. This was not the case for A549 cells where E-cadherin protein levels remained unchanged with the addition of these inhibitors (Fig. 3A,, bottom). Examination of the 15-PGDH promoter revealed several E-boxes, which are sites bound by the Snail, Slug, and ZEB1 transcriptional repressors (14, 15). Indeed, the loss of PGDH expression in colon cancer is most likely due to the up-regulation and binding of the transcriptional repressor Snail to E-boxes located in the PGDH promoter (5). Quantitative RT-PCR analysis of H1435 and A549 cell lines revealed a reduction in ZEB1 and Slug transcripts following erlotinib and U0126 treatment in H1435 but not A549 cells (Fig. 3B and C). This strongly suggests that ZEB1 and Slug are downstream targets of the EGFR through activated ERK, and that one or both of these transcriptional repressors controls the expression of 15-PGDH in the H1435 cell line. We also examined the levels of Snail transcripts in H1435 and A549 cells and found that Snail transcripts were slightly, but significantly, decreased in both cell lines with erlotinib but not following U0126 treatment (Supplementary Fig. S2). The fold reduction in transcript levels was small when compared with ZEB1 and Slug, and the Snail transcript levels were not reduced significantly with U0126 treatment, suggesting that this repressor is not responsible for the repression of 15-PGDH in these cell lines.

Transcriptional repressors, such as ZEB1 and Slug, are known to repress transcription by recruiting histone deacetylases (HDAC) to promoters through an interaction with COOH-terminal binding protein (14, 15). Additional proof of the involvement of the transcriptional repressors ZEB1 and Slug in the regulation of 15-PGDH is provided by the use of a HDAC inhibitor (Fig. 3D). Addition of nanomolar amounts of TSA to H1435 and A549 cells significantly increases the protein levels of PGDH in these cell lines. Recently, similar results were observed by Tong et al. (16) using various HDAC inhibitors. Interestingly, they showed that HDAC inhibitors induced luciferase activity from a reporter construct containing the same 15-PGDH promoter region as Mann et al.

Our data suggest that ZEB1 and Slug are regulating the 15-PGDH promoter, but definitive proof would require that we show binding of these transcription factors to the promoter. To show that these transcriptional repressors bind to the 15-PGDH promoter, we synthesized three oligonucleotides for gel shift analysis that encompassed each of the three E-boxes located in the 15-PGDH 5′ flanking region (Fig. 4A). Nuclear lysates isolated from cells transfected with either vector or FLAG-tagged ZEB1 (FL-ZEB) or FLAG-tagged Slug (FL-Slug; Fig. 4B) were incubated with the labeled oligonucleotides. Surprisingly, only FL-Slug bound to one of the three oligonucleotides that encompassed an E-box just upstream of the putative TATA box and transcriptional start site (Fig. 4A). This E-box sequence, CAGGTG, present in the PGDHpr3 oligonucleotide is reported to be a perfect consensus-binding site for the Slug repressor (17). Figure 4C provides further proof that the complex formed in the gel shift contains Slug. A nonspecific oligo containing an E-box that is not the perfect Slug consensus (CAGCTG) does not compete with PGDHpr3 for binding. In addition, an anti-FLAG antibody, but not an anti-HA antibody, almost completely blocks complex formation and shifts the complex, providing further proof that Slug is binding this oligonucleotide derived from the PGDH promoter. The observation that only Slug, and not ZEB1, binds to the 15-PGDH promoter and regulates its activity is further bolstered by transcription assays using the first 626 bp of the 15-PGDH promoter (Fig. 4A). Slug clearly represses transcription by up to 10-fold in a dose-dependent manner. ZEB1 does not repress the 15-PGDH promoter at the lowest levels of plasmid and shows only weak and probably nonspecific inhibition with the highest amounts of FL-ZEB expression construct (Fig. 4D). These experiments were repeated with a human ZEB1 (hZEB1) containing expression construct with the same results (data not shown). Based on these observations, we suggest that Slug plays a key role in the repression of 15-PGDH in a subset of NSCLC.

The altered regulation of genes through the up-regulation of zinc-finger transcriptional repressors, such as ZEB1, Slug, and Snail, is being increasingly recognized as an important mechanism in cancer progression. We present evidence that the loss of 15-PGDH and the resulting increased PGE₂ level in a subset of NSCLC are most likely due to an increase in the amount of the Slug transcriptional repressor. Alternatively, the down-regulation of PGDH in colorectal cancer is probably due to the presence of higher levels of Snail (5). Several studies have shown that up-regulation of ZEB1 is responsible for a more mesenchymal state and is correlated with gefitinib resistance in NSCLC cell lines (18, 19). Resistant cell lines could be made more sensitive to gefitinib by pretreatment with an HDAC inhibitor. Investigation into the transcriptional regulation of Snail provides evidence that an activated MEK/ERK pathway is involved in its up-regulation (20). Our study and that of Mann et al. would indicate that Slug, ZEB1, and Snail are regulated by activated EGFR through the MEK/ERK pathway. Interestingly, two of the cell lines that did not show up-regulation of 15-PGDH upon erlotinib treatment (i.e., A549 and H2122) both have mutated K-Ras, whereas the responsive cell lines (i.e., H1435 and H1648) have wild-type K-Ras. This may be relevant as Slug seems to be regulated through the MEK/ERK pathway (21). It is in cell lines, and perhaps solid tumors, with wild-type K-Ras that erlotinib can down-regulate transcriptional repressors and up-regulate tumor suppressors such as 15-PGDH. The restoration of 15-PGDH protein levels and decrease in COX-2 levels through erlotinib treatment in a subset of NSCLC could provide one explanation for the survival benefit of erlotinib, particularly in those patients without an activating EGFR mutation. Understanding how these transcriptional repressors are regulated will be an important area for future research and will hopefully provide opportunities for more effective cancer treatments.

Grant support: National Cancer Institute Vanderbilt Lung Specialized Programs of Research Excellence grant CA90949 (D.P. Carbone). The work at the University of Kentucky was supported in part by NIH grant HL-46296 and the Kentucky Lung Cancer Research Program.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Dr. Kiyoshi Yanagisawa for technical assistance with RNA extraction from lung tumor tissue used in the microarray analysis and the work of the Vanderbilt Microarray Shared Resource.