Interaction between head and neck squamous cell carcinoma cells and fibroblasts in the biosynthesis of PGE₂

Abstract

Prostaglandin (PG)E₂ is relevant in tumor biology, and interactions between tumor and stroma cells dramatically influence tumor progression. We tested the hypothesis that cross-talk between head and neck squamous cell carcinoma (HNSCC) cells and fibroblasts could substantially enhance PGE₂ biosynthesis. We observed an enhanced production of PGE₂ in cocultures of HNSCC cell lines and fibroblasts, which was consistent with an upregulation of COX-2 and microsomal PGE-synthase-1 (mPGES-1) in fibroblasts. In cultured endothelial cells, medium from fibroblasts treated with tumor cell-conditioned medium induced in vitro angiogenesis, and in tumor cell induced migration and proliferation, these effects were sensitive to PGs inhibition. Proteomic analysis shows that tumor cells released IL-1, and tumor cell-induced COX-2 and mPGES-1 were suppressed by the IL-1-receptor antagonist. IL-1α levels were higher than those of IL-1β in the tumor cell-conditioning medium and in the secretion from samples obtained from 20 patients with HNSCC. Fractionation of tumor cell-conditioning media indicated that tumor cells secreted mature and unprocessed forms of IL-1. Our results support the concept that tumor-associated fibroblasts are a relevant source of PGE₂ in the tumor mass. Because mPGES-1 seems to be essential for a substantial biosynthesis of PGE₂, these findings also strengthen the concept that mPGES-1 may be \a target for therapeutic intervention in patients with HNSCC.

Eicosanoids derived from polyunsaturated fatty acids are soluble mediators that exert a key role in the physiopathology of many disorders, including inflammation, thrombosis, and cancer. Prostanoids derived from arachidonic acid (AAc) through the cyclooxygenase (COX) pathway are particularly relevant. The increasing interest in the role of prostanoids in the context of cancer originates in the large epidemiological trials that showed that COX-inhibiting nonsteroidal anti-inflammatory drugs could be beneficial against the development and growth of malignancies (1).

Prostaglandin (PG)H₂ is the common cyclic-peroxide intermediate in the biosynthesis of prostanoids derived from AAc. The other prostanoids are formed in reactions catalyzed by specific synthases acting on PGH₂ (2). In contrast with the ubiquitous expression of COXs, expression of downstream synthases confers a cell-specific prostanoid profile. COX-2 receives the most attention because, unlike COX-1, which is widely expressed, its expression is restricted in nonpathologic settings to a few cell types and tissues, but it is over-expressed in a wide range of cell types in tumors and inflamed tissues. COX-2 is transiently and selectively induced by pro-inflammatory cytokines, tumor promoters, growth factors, and hormones (2, 3). These observations generated the simplistic paradigm that COX-2 is associated with pathological situations such as cancer, including HNSCC (4, 5). Structural differences between the two COX isoenzymes have important pharmacological and biological consequences. Unlike COX-2, COX-1 activity exhibits a cooperative dependence on the substrate concentration [reviewed in Refs. (3) and (6)]. This difference is relevant for the relative prominence of both isoenzymes in vivo; it may permit COX-2 to compete more effectively for the substrate when both the isoenzymes are expressed in the same cells. These properties, and the evidence accumulated from coexpression studies, have led to the concept that COX-2 is functionally coupled with downstream synthases such as microsomal PGE-synthase-1 (mPGES-1) (7). Nevertheless, COX-2 selective inhibitors entail a cardiovascular hazard (8), partly attributable to a reduced production of PGI₂ by vascular cells (9).

Among the prostanoids, PGE₂ has received the most attention because its levels are elevated in many tumors, including HNSCC (10–12), and because it has several biological activities that are compatible with tumor progression. PGE₂ promotes cancer cell growth and survival by several mechanisms, including increased proliferation, inhibition of apoptosis, increased migration and invasiveness, angiogenesis, suppression of immune attack, and chronic inflammation [reviewed in Refs. (11) and (12)]. Conversion of PGH₂ to PGE₂ is catalyzed by PGE-synthases (PGESs). Three PGES isoenzymes have been characterized: two microsomal isoforms (mPGES-1 and mPGES-2) and one cytosolic isoform (cPGES) (7). mPGES-1 is inducible by pro-inflammatory cytokines, and it seems to be the essential PGES isoenzyme involved in PGE₂ biosynthesis under inflammatory conditions (13–16). Additionally, mPGES-1 has been reported to be over-expressed in several types of tumors (17–20). COX-2/mPGES-1 is widely regarded as the main contributing enzymatic tandem for PGE₂ biosynthesis under pathological conditions. Genetic inactivation or pharmacological inhibition of COX-2 reduces tumor-induced neovascularization (11, 12, 21). Furthermore, mPGES-1 appears to be critical for tumor angiogenesis (22).

Accumulated evidence indicates that interactions between tumor cells and stromal cells may dramatically influence tumor progression (23, 24). Fibroblasts seem to be particularly relevant among stromal cells, and carcinoma-associated fibroblasts are frequently observed in the stroma of human carcinomas. Their presence in large numbers is often associated with the development of high-grade malignancies and poor prognoses. The molecular mechanisms underlying tumor-promoting capabilities of carcinoma-associated fibroblasts include the release of cytokines and growth factors, which promote tumor cells growth and migration, angiogenesis, and evasion of the immune response (reviewed in References 25 and 26). Tumor cell-fibroblast cross talk causes mutual influence, which results in a promotion of tumor progression. Tumor cells release soluble factors that could act on the neighboring stroma cells, inducing expression of genes, for example chemokines, which amplify inflammation (27, 28). The present work was conducted to test the hypothesis that the cross-talk between HNSCC and fibroblast could substantially enhance PGE₂ biosynthesis through COX-2/mPGES-1 induction in fibroblasts.

MATERIALS AND METHODS

Tumor cell culture and obtaining the conditioned medium

FaDu and SSC-25, permanent pharynx, and tongue squamous cell carcinoma cell lines were obtained from American Type Culture Collection (ATCC HTB-43 and CRL-1628). Cells were grown in DMEM containing 10% FBS and supplemented with 2 mM/l L-glutamine, 1 mM/l sodium pyruvate, 100 U/ml penicillin, and 100 μg/ml streptomycin (all from Biological Industries, Kibbutz Beit Haemek, Israel). When the cells reached early confluence, the medium was replaced with fresh medium containing 1% FBS. Conditioned medium was collected from cultures 48 h later. The conditioned medium was centrifuged and stored at –80°C until fibroblast stimulation.

Fibroblast culture and treatment

Human dermal fibroblasts were isolated and cultured as described (29). Cells in the 3-4 passage were seeded in 6-well plates and cultured in DMEM containing 10% FBS. Confluent fibroblasts were stimulated by replacing the medium with DMEM containing 1% FBS (control) or conditioned medium from tumor cells. Fibroblasts were kept in the culture chamber for different periods of time before prostanoids and enzyme expression were analyzed. COX-2 and mPGES-1 protein expression induced by tumor cell-conditioned medium were also analyzed in a tumor-derived fibroblast cell line CCD-18Co. The human colon carcinoma-derived fibroblast cell line CCD-18Co was obtained from ATCC (ATCC-CRL-1459) and was cultured as indicated in the product information sheet from ATCC. This cell line was treated as described for dermal fibroblasts.

Co-culture assays

Dermal fibroblasts were cultured in 12-well culture plates, and tumor cells were seeded in 0.4 μm pore cell culture inserts (Becton Dickinson Labware, Franklin Lakes, NJ). When cells reached confluence, inserts were placed in 12-well culture plates containing the fibroblasts, and plates were incubated for 24 h in the culture chamber. Thereafter, culture media and cells were recovered for analysis.

Obtaining the conditioned media from fibroblasts

Dermal fibroblasts were cultivated in 75 cm² culture flasks. When the fibroblasts reached early confluence, the medium was replaced with fresh medium (DMEM) containing 1% FBS or FaDu-conditioned medium (FaDu-CM). Conditioned media were collected from cultures 48 h later. The conditioned media were then centrifuged and stored at –80°C until the angiogenesis, migration, and proliferation assays. To obtain conditioned media free of prostanoids, all procedures, including those to obtain FaDu-CM, were performed in the presence of 10 µM/l indomethacin.

Determination of the AAc profile

Cells were incubated with 25 μM/l of [¹⁴C]AAc, (55–58 mCi/mmol; GE Healthcare, Buckinghamshire, UK) for 10 min as described (30). Prostanoids were analyzed by HPLC as previously described (29).

Analysis of prostanoids

PGE₂ and 6-oxo-PGF_1α (stable hydrolysis product of PGI₂) were analyzed by specific enzyme immunoassay (Cayman Chemical, Ann Arbor, MI) following the manufacturer's instructions.

Analysis of mRNA levels of COX-2 and mPGES-1

Total RNA was extracted by chloroform isopropanol precipitation using Ultraspec (Biotecx Laboratories, Inc., Houston, TX) according to the manufacturer's instructions. Reverse transcription was performed with 0.5 μg of RNA per 10 μl reaction mixture, and enzyme mRNA levels were studied by real-time PCR as previously described (14). Gene expression data were normalized to β-actin as endogenous control, and RNA of untreated cells was used as a calibrator sample.

Analysis of COX-1, COX-2, mPGES-1, mPGES-2, and cPGES protein

Fibroblast protein extracts were analyzed by Western blot as described (14). Briefly, cell cultures were washed twice with PBS and lysed with lysis buffer 20 mM/l Tris-HCl (pH 7.4) containing protease inhibitor cocktail (Roche Diagnostics GmbH), 1 mM/l EDTA, and 0.1% Triton X-100. Protein concentration was determined by the Bradford method. Total protein equivalents were resolved by SDS-PAGE and electrotransferred onto polyvinylidene difluoride membranes (Immobilon-P; Millipore Millipore Ibérica, Spain). Membranes were incubated with antibodies against the human enzymes (Cayman Chemical). Bound antibody was detected using the appropriate horseradish peroxidase-conjugated antibody (Dako, Glostrup, Denmark) and a chemiluminescent detection system (Amersham ECL Plus Western Blotting Detection Reagents, GE Healthcare). Results were normalized by β-actin (Sigma) used as a loading control.

Cell cycle and apoptosis studies

For the cell cycle experiments, after the required treatment, cells were fixed in ice-cold 70% ethanol and stored at −20°C for at least 18 h. Cells were stained with 0.5 ml propidium iodide/Rnase staining buffer (BD Biosciences, San Diego, CA) for 15 min at room temperature and analyzed by flow cytometry (FACS Calibur; Beckton Dickinson) and Cell Quest Pro software. Apoptosis was determined by washing harvested cells with cold PBS and staining them with an annexin V-FITC apoptosis detection kit (Bender MedSystems GmbH, Viena, Austria) according to the manufacturer's instructions. Samples were then subjected to flow cytometry analysis.

In vitro angiogenesis assay

Human umbilical vein endothelial cells (HUVECs) were isolated and cultured as previously described (31). In vitro angiogenesis assays were performed as described (32). Briefly, 12.5 to 15 × 10³ HUVECs were seeded onto growth factor-reduced Matrigel (Becton Dickinson Matrigel Basement Membrane Matrix) in 96-well plates and exposed to treatments as required. After 4.5 h of treatment, photographs were taken with a Nikon Digital Sight DS-2MBW camera mounted on a Nikon Eclipse TS100 microscope using a 10×/0.40 objective, and the number of closed polygons in the endothelial cell mesh was determined.

Invasion assay

The Radius™ 96-Well Cell Migration Assay Kit (Cell Biolabs, Inc., San Diego, CA) was used following the manufacturer's instructions. FaDu cells were seeded in the assay plate and subjected to the required treatment for 6.5 h. Afterward, photographs were taken with a Olympus C5050 digital camera attached to an Olympus BX50 microscope. Standardized size photographs were printed in a high-quality photograph paper, and the cell-free area was outlined and cut out. The percentage of initial cell-free area invaded was determined by gravimetry.

Microscopic characterization of fibroblast phenotype

For immunostaining studies, dermal fibroblasts or tumor-derived fibroblast cell line CCD-18Co were grown in 12 mm cover glasses (Marlenfeld GmbH and Co. KG, Lauda-Königshofen, Germany), treated as required, and fixed with methanol:acetone 1:1 at −20°C. Cells were stained using a Monoclonal Anti-Vimentin-Cy3 conjugate (C9080 Sigma-Aldrich Madrid, Spain). Nuclei were counterstained with Hoechst 33342 (Sigma-Aldrich) 1 μg/ml diluted in the antibody solution. Fluorescent images were recorded in a contrast (bright-field) microscopy and fluorescence microscope (Axiovert-200M Zeiss, Jena, Germany).

Signaling pathways involved in tumor cell-induced expression of COX-2 and mPGES-1

Dermal fibroblasts were incubated with conditioned medium of tumor cells for 48 h in the absence and in the presence of the indicated concentrations of GF109203× (a general inhibitor of protein kinase C), U0126 or PD98059 (inhibitors of mitogen-activated protein kinase kinase 1/2), SB203580 (a p38 mitogen-activated protein kinase [p38-MAPK] inhibitor), LY294002 (phosphoinoside 3-kinase inhibitor), rapamycin (mammalian target of rapamycin inhibitor) (all from Sigma) and Gö6976 (a Ca²⁺-dependent protein kinase C inhibitor), and Akt-inhibitor (1L6-Hydroxymethyl-chiro-inosito-2(R)-2-O-methyl-3-O-octadecyl-sn-glycerocarbonate; protein kinase B inhibitor), both from Calbiochem, Darmstad, Germany. Cells were incubated for 30 min with the inhibitors before the addition of the tumor cell-conditioned medium. Thereafter, COX-2 and mPGES-1 protein expression was evaluated as described above.

Phosphorylation of p38-MAPK was determined by immunoblotting using a p38-MAPK polyclonal antibody and ant phosphor-p38-MAPK (Tyr180/Tyr182) mouse monoclonal antibody (Cell Signaling Technology, Inc., Beverly, MA) as described above.

Thermal stability and conditioned medium fractionation

Tumor cell-conditioned media were heated for 30 min at 50, 75, and 100°C. Media were then stored at −80°C until their use to stimulate fibroblasts. Another set of experiments was performed using tumor cell-conditioned media filtered through exclusion molecular weight (MW) membranes of 10, 30, and 50 kDa (Amicon Ultra; Millipore, Carrigtwohill, Co. Cork, Ireland) following the manufacturer's instructions.

Protein analysis of tumor cell-conditioned media

The protein array test was performed using the Human Cytokines Antibody Array 3 (RayBiotech, Inc., Norcross, GA), and samples were processed following the manufacturer's protocol. Chemiluminiscent detection was performed with Amersham ECL Plus Western Blotting Detection Reagents. The density of the blots was measured in a GelDoc 2000 using Quantity One software (Bio-Rad Laboratories, Hercules, CA).

Two-D nano-liquid chromatography mass spectrometry (MS) and database searching analysis of tumor cell conditioned media was performed in a liquid chromatograph (Ultimate 3000; LC Packings, Amsterdam, The Netherlands) coupled to an nESI-LTQ mass spectrometer (Thermo Electron Corp., Bremen, Germany). A detailed description of the method is supplied as supplementary material.

Quantitative IL-1α, IL-1β, and IL-1 receptor antagonist analysis

Quantitative analysis of IL-1α, IL-1β, and IL-1 receptor antagonist (IL-1ra) in the culture media was performed by specific ELISA following the manufacturer's instructions (IL-1α and IL-1ra were from R&D Systems, Minneapolis, MN; IL-1β was from eBioscience, San Diego, CA).

Effect of IL-1 receptor blocking on tumor cell induction of COX-2 and mPGES-1

To explore the possible involvement of IL-1 on tumor cell-stimulation of dermal fibroblasts, cells were exposed to human recombinant IL-1β or tumor cell-conditioned medium for the indicated period of time in the presence of recombinant human IL-1 receptor antagonist (IL-1ra) (PeproTech, London, UK).

Determination of IL-1α, IL-1β, and IL-1ra released by nontumoral mucosa and HNSCC tumor samples

The present study was approved by the HSCSP Ethics Committee, and informed consent was obtained from each subject. A tissue sample of tumor and nontumoral mucosa (contralateral to the site of the primary tumor) was obtained at the pretreatment period from 20 patients with a HNSCC diagnosed and treated in our center (see Supplementary Table I for characteristics of patients). Immediately after extraction, tumor and nontumoral mucosa pieces of HNSCC were cut into small fragments with a surgical blade (≈1 mm) and used for experiments without further manipulation. Fragments of 100–200 mg were placed in 1 ml DMEM supplemented with 2 mM/l L-glutamine, 1 mM/l sodium pyruvate, 100 U/ml penicillin, and 100 μg/ml streptomycin and incubated for 48 h in the culture chamber. The culture medium was later recovered and kept at −80°C until IL-1α, IL-1β, and IL-1ra were analyzed.

Statistics

Sigma-Plot 11 software was used for statistical analysis. Statistical significance between pairs of groups was assessed using the paired samples t-test; multiple comparisons were performed by ANOVA test. A p value <0.05 was considered significant.

RESULTS

Fig. 1A shows illustrative HPLC chromatograms of the AAc profile of FaDu and dermal fibroblasts regarding prostanoids after incubation with [¹⁴C]labeled-AAc. Both cells produced PGE₂ as the major prostanoid, and fibroblasts additionally produced PGI₂ (determined as its stable hydrolysis product 6-oxo-PGF_1α). When we analyzed the release of PGE₂ by cocultures of FaDu and fibroblasts, we found that production of PGE₂ was significantly higher than the production by cells incubated alone and was higher than the sum of the individual production of FaDu plus fibroblasts. This indicated that coculture of both cell types causes a synergistic effect on PGE₂ biosynthesis (Fig. 1B). We then explored the effect of the coculture on the expression of the enzymes involved in PGE₂ biosynthesis. We examined the mutual influence of each cell type in the expression of COX-1, COX-2, mPGES-1, mPGES-2, and cPGES. Both cell types expressed all the enzymes, but COX-1, mPGES-2, and cPGES were not modified in the coculture samples when compared with cells incubated individually (not shown). In contrast, COX-2 was upregulated in FaDu and fibroblasts after coculture, whereas mPGES-1 was only appreciably upregulated in the fibroblasts (Fig. 1C). However, coculture caused much more COX-2 upregulation in fibroblasts than in FaDu.

Fig. 1.

A: Illustrative HPLC chromatograms of the AAc profile of FaDu and fibroblasts regarding prostanoids after incubation with ¹⁴C-labeled-AAc. Cells were incubated with 25 μM/l of [¹⁴C]AAc for 10 min, and prostanoids were analyzed by HPLC. B: Release of PGE₂ by cocultures of FaDu and fibroblasts. Fibroblasts were placed in culture plates and in the coculture experiments tumor cells were placed in 0.4 μm pore cell culture inserts that were located over the wells containing the fibroblasts. Plates were then incubated for 24 h. Thereafter, PGE₂ was analyzed in the media. Values are mean ± SEM (n = 5). *P < 0.05 when compared with the sum of the production of FaDu plus fibroblasts incubated separately (FaDu+Fibro.). C: Representative immunoblotting analysis of COX-2, mPGES-1, and β-actin proteins after the incubations described above (n = 2).

We next explored the effect of FaDu-CM, and human recombinant IL-1β as positive control, on the release of prostanoids by fibroblasts. FaDu-CM, like IL-1β, induced the release of PGE₂ and PGI₂ in a time-dependent manner (Fig. 2). PGE₂ accumulated in the culture medium was about 5-fold the amount of 6-oxo-PGF_1α.

Fig. 2.

Time course of prostanoid release by fibroblasts incubated without (control) and with 10 U/ml of human recombinant IL-1β or FaDu-CM. Values are mean ± SEM (n = 4). *P < 0.05 when compared with controls.

Fig. 3 shows the effect of conditioned medium from two head and neck tumor cell lines on the expression of COX-2 and mPGES-1 in dermal fibroblasts analyzed in terms of mRNA and protein. FaDu-CM and SCC-25-CM time dependently induced mPGES-1 mRNA levels in the fibroblasts, whereas only FaDu-CM was able to significantly induce COX-2 mRNA levels (Fig. 3A). Analysis of the proteins showed similar results. Only FaDu-CM increased COX-2 protein levels, whereas both FaDu-CM and SCC-25-CM time-dependently upregulated mPGES-1 in terms of protein. Tumor cell-induced mPGES-1 expression was delayed compared with COX-2 (Fig. 3). In addition, the transcription inhibitor actinomycin-D totally suppressed the effect of tumor cells on COX-2 and mPGES-1 expression (not shown).

Fig. 3.

Effect of FaDu (FaDu-CM) and SCC-25 (SCC-25-CM)-conditioned media on COX-2 and mPGES-1 expression in cultured dermal fibroblasts. A: COX-2 and mPGES-1 mRNA levels were analyzed by real-time PCR. Results are expressed relative to untreated cells (n = 4). B: COX-2 and mPGES-1 protein levels were analyzed by Western blot; the graph represents computer-assisted densitometry values normalized to untreated fibroblasts (n = 5). Representative immunoblots are also shown. Values in all graphs are mean ± SEM. *P < 0.05, ** P < 0.01, and *** P < 0.001 versus untreated cells.

To explore other activities potentially related to factors released by tumor cells, we performed cell cycle and apoptosis analysis of fibroblasts exposed to FaDu-CM. Fig. 4A shows the results of flow cytometry analysis. FaDu-CM slightly, but significantly, modified cell cycle in cultured dermal fibroblasts. The percentage of cells in G₀/G₁ decreased, and the percentage in synthesis (S) and G_2/mitosis (M) phases increased. AnnexinV-FITC apoptosis study showed that FaDu-CM did not cause apoptosis of fibroblasts (not shown). Microscope observation of cultured fibroblasts showed that dermal fibroblasts treated with FaDu-CM modified fibroblasts phenotype toward contracted phenotype (Fig. 4B). The same was observed with SCC-25-CM (not shown). This contracted phenotype was also observed after treatment of the tumor-derived fibroblast cell line CCD-18Co with tumor-cell conditioned medium (Fig. 4C). To confirm that PGE₂ biosynthetic machinery can also be modified in tumor-derived fibroblasts by tumor cells, the CCD-18Co cell line was incubated in the same conditions as dermal fibroblasts, and COX-2 and mPGES-1 protein were then analyzed. Fig. 4D shows observations similar to those from the experiments performed with dermal fibroblasts depicted in Fig. 3B.

Fig. 4.

A: Effect of FaDu-CM on dermal fibroblast cell cycle distribution. Fibroblasts were incubated with FaDu-CM for the indicated period of time (FaDu-CM). Cells were then stained with 0.5 ml propidium iodide/Rnase staining buffer and analyzed by flow cytometry. Bars represent mean ± SEM (n = 4). *P < 0.05 and ** P < 0.01 versus untreated cells (controls). B: Effect of FaDu-CM on dermal fibroblast phenotype. Dermal fibroblasts were incubated with fresh culture medium with 1% FBS (a) and FaDu-CM (b) for 24 h. Cells were then stained with anti-vimentin antibody as described in the Materials and Methods section; photographs are representative of two independent experiments. C: The same experiment as in B performed with tumor-derived fibroblast cell line CCD-18Co. D: Effect of FaDu-CM and SCC-25-CM on COX-2 and mPGES-1 protein expression in CCD-18Co cell line. Representative immunoblottings out of two independent experiments are shown.

We next explored the biological significance of PGE₂ secreted by tumor cell-stimulated fibroblasts in in vitro angiogenesis assays and in tumor cell migration and proliferation experiments. First, we examined the effect of conditioned media of nonstimulated dermal fibroblasts, FaDu-CM and FaDu-CM-stimulated fibroblasts in the absence or presence of indomethacin, on the ability of HUVEC to form capillary-like structures on Matrigel. As a control of indomethacin activity, we measured the concentration of PGE₂ in the conditioned media (fibroblasts, 2.33 ± 2.26; fibroblast-indomethacin, 0.15 ± 0.09; FaDu, 0.56 ± 0.17; FaDu-indomethacin, 0.09 ± 0.008; FaDu-CM-stimulated fibroblasts, 35.36 ± 15.29; and FaDu-CM-stimulated fibroblasts-indomethacin, 0.28 ± 0.18; nM/l [mean ± SEM; n = 3]). Conditioned media from nonstimulated fibroblasts did not induce significant in vitro angiogenesis, whereas FaDu-CM and FaDu-CM-stimulated fibroblasts conditioned media significantly induced formation of capillary-like structures (P < 0.05 and P < 0.001, respectively). Similar to PGE₂ synthesis, the effect of conditioned media of fibroblasts exposed to FaDu-CM was significantly higher than that of FaDu-CM alone (P < 0.05). Moreover, indomethacin significantly decreased the effect of FaDu-CM-stimulated fibroblasts on the formation of capillary-like structures (P < 0.01); however, the differences observed between nonstimulated fibroblast conditioned media, FaDu-CM and FaDu-CM-stimulated fibroblasts conditioned media collected in the presence of indomethacin, were not significant (Fig. 5).

Fig. 5.

Effect of conditioned medium of FaDu-CM-estimulated fibroblast on in vitro angiogenesis. HUVECs were seeded onto growth factor reduced Matrigel in 96-well plates and exposed to fresh medium containing 1% FBS (controls) in the absence or presence of 10 µM/l indomethacin (controls indomethacin), nonstimulated fibroblast-conditioned media collected without (fibroblasts) or with indomethacin (Fib-indo), FaDu-CM collected without (FaDu) or with indomethacin (FaDu-indo), and FaDu-CM-stimulated fibroblasts conditioned media collected without (Fib-FaDu) or with indomethacin (Fib-FaDu-indo). After 4.5 h of treatment, standardized field photographs were taken and the number of closed polygons in the endothelial cell mesh was counted. Bars are mean ± SEM (expressed as relative to the corresponding controls with or without indomethacin) of three independent experiments each of them performed in nineplicate. *P < 0.05, **P < 0.01, and *** P < 0.001; ^#P < 0.05 and ^###P < 0.001 when compared with controls. Representative photographs are also shown (lower panels).

Conditioned media of fibroblast and fibroblasts treated with FaDu-CM obtained in the absence and in the presence of indomethacin as described previously were also used to evaluate the effect on tumor cell migration and proliferation. To measure 2-D cell migration, the ability of FaDu cells to invade a cell-free region was evaluated by determining the occupied free area after 6.5 h. PGE₂ effectively induced FaDu migration (Fig. 6A). The presence of indomethacin in the controls and in the nonstimulated fibroblasts significantly increased the ability of FaDu to invade the cell-free region (P < 0.05), whereas in the case of the FaDu-CM-stimulated fibroblasts, the opposite occurred, and the rate of closing the cell-free area was significantly (P < 0.05) higher in the samples without indomethacin. Fig. 6B shows the results regarding the ability of FaDu cell line exposed to fresh media (controls), conditioned media of nonstimulated fibroblasts, and conditioned media of fibroblasts stimulated with FaDu-CM to invade the cell-free area. The results are expressed as relative to the corresponding conditioned media collected in the presence of indomethacin. Fibroblast-conditioned media did not modify the relative rate of invasion, whereas conditioned media of FaDu-CM-stimulated fibroblasts significantly increased the relative invasion rate when compared with controls. Fig. 6C shows the results of flow cytometry analysis regarding cell cycle of FaDu exposed to the conditioned media. Like in the migration experiments, nonstimulated fibroblast conditioned media did not modify the proportion of cells in G₀/G₁ or in S plus G₂/M phases when expressed relative to the same media obtained in the presence of indomethacin, whereas FaDu-CM-stimulated fibroblast conditioned media significantly modified FaDu cell cycling. The percentage of cells in G₀/G₁ decreased, and concomitantly the percentage in S+G₂/M phases increased. AnnexinV-FITC apoptosis study showed that none of the conditioned media tested modified the percentage of apoptotic FaDu cells (not shown).

Fig. 6.

A: Effect of PGE₂ on migration of FaDu. FaDu cells were seeded in the assay plate and subjected to the treatment with fresh medium containing 1% FBS (controls) or the same medium containing 100 nM/l of PGE₂ (Cayman Chemicals) for 6.5 h. Standardized sized photographs were printed and the percentage of initial cell-free area invaded was determined by gravimetry. Bars are mean ± SEM (n = 9). ***P < 0.001 when compared with controls. Illustrative photographs are also shown. B: Effect of conditioned media of FaDu-CM-stimulated fibroblast on migration of FaDu. After FaDu seeding, cells were subjected to the treatment with fresh medium containing 1% FBS (controls), nonstimulated fibroblasts conditioned media (fibroblasts), and FaDu-CM-stimulated fibroblasts conditioned media (Fib-FaDu); all cells were collected in the absence or presence of 10 µM/l indomethacin. Bars are mean ± SEM; results are expressed as relative to the initial cell-free area occupied by cells stimulated by the corresponding conditioned media obtained in the presence of indomethacin ns, nonsignificant; n = 3 performed in triplicate. **P < 0.01. C: Effect of conditioned medium of FaDu-CM-stimulated fibroblasts medium on FaDu cell cycle distribution. FaDu were incubated for 24 h with the same conditioned media that were used for the migration experiments, and flow cytometry analysis was then performed. Bars are mean ± SEM. Results are expressed as relative to cells stimulated by the corresponding conditioned media obtained in the presence of indomethacin. ns, nonsignificant; n = 4 independent experiments. **P < 0.01.

We then tested the thermal stability of the active component in the culture-conditioned medium from tumor cells. We determined expression of COX-2 and mPGES-1 in dermal fibroblasts after heating the tumor cell-conditioned media at 50, 75, and 100°C for 30 min. The media heated at 75 and 100°C lost its ability to induce enzyme expression (not shown). We then exposed dermal fibroblasts to the FaDu-CM and SCC-25-CM previously filtered by MW-exclusion membranes (<10, <30, and <50 kDa). The filtrate containing molecules of MW <10 kDa was not active in inducing COX-2 or mPGES-1, whereas filtrates containing molecules of MW <30 or <50 kDa were active in inducing COX-2 and mPGES-1. However, filtrates with MW <50 kDa were more active than those with MW <30 kDa (Fig. 7A). These results suggest that the active agents were a mixture of proteins with MW between 10–30 and 30–50 kDa.

Fig. 7.

A: Representative immunoblots of COX-2 and mPGES-1 from fibroblasts incubated 24 h with culture medium (control), tumor cell-conditioned media (whole medium), or tumor cell-conditioned media filtered through MW-exclusion membranes of 10, 30, and 50 kDa (<10 kDa, <30 KDa, <50 kDa) (n = 3). B: Representative immunoblots of COX-2 from fibroblasts incubated 24 h with culture medium (control) or FaDu-CM plus the indicated concentration of SB203580 (n = 2). C: Representative immunoblots of the time course of total and phosphorylated p38-MAPK in response to the exposure of dermal fibroblasts to FaDu-CM (n = 2). D: Representative immunoblots of COX-2 and mPGES-1 from fibroblasts incubated 24 h with culture medium alone (control) or with 10 U/ml IL-1β or tumor cell-conditioned media plus the indicated concentration of human recombinant IL-1ra (n = 4).

To characterize signaling pathways involved in the tumor cell-induced COX-2 and mPGES-1 expression, we incubated dermal fibroblasts with several inhibitors during the stimulation. As a first approximation, we analyzed COX-2 and mPGES-1 expression in cells stimulated with FaDu-CM- and SCC-25-CM-conditioned medium in the presence of a constant concentration of inhibitor as follows: GF109203× (5 μM/l), Gö6976 (1 μM/l), rapamycin (1 μM/l), SB203580 (10 μM/l), U0126 (10 μM/l), PD98059 (50 μM/l), LY294002 (50 μM/l), and Akt-inhibitor (25 μM/l). Densitometric evaluation of the immunoblottings showed that the p38-MAPK inhibitor SB203580 significantly reduced the effect of the tumor cell-conditioned medium on COX-2 induction but not on mPGES-1. The other inhibitors did not significantly inhibit COX-2 or mPGES-1 expression induced by tumor cells (not shown). Because these results suggested that p38-MAPK was involved in the FaDu-CM-induced COX-2, further experiments were carried out to study the concentration-dependent effect of SB203580 and therefore to rule out the possibility that the effects observed in the preliminary experiments were unspecific. The Western blot in Fig. 7B shows that the inhibitor suppressed the effect of the tumor cell-conditioned media on COX-2 protein expression in a concentration-dependent manner. Fig. 7C shows that FaDu-CM effectively induced p38-MAPK phophorilation in fibroblasts after 10 min of incubation.

We identified proteins by nano-liquid chromatography MS/MS and database searching and by protein array tests to find cytokines, growth factors, and proteases in the conditioned media from FaDu and SCC-25. We identified 589 different proteins released by tumor cells, 169 of which were present in FaDu- and SCC-25-conditioned medium (a table showing the list of cytokines, growth factors, and proteases is supplied as supplementary material). From among them, IL-1α and IL-1β were identified, these being clear candidates for the COX-2- and mPGES-1-inducing activities. Using ELISA, we evaluated the concentration of IL-1α, IL-1β, and IL-1ra in the conditioned media from FaDu and SCC-25 (FaDu, IL-1α: 742.5 ± 42.8, IL-1β: 22.4 ± 1.2 and IL-1ra: 320.9 ± 42.9; SCC-25, IL-1α: 67.4 ± 2.6 and IL-1β: 1.3 ± 0.1; IL-1ra: 3166.3 ± 172.6; pg/ml, n = 3, mean ± SEM). Because IL-1 released by tumor cells was a candidate to induce COX-2 and mPGES-1, we incubated dermal fibroblasts with the tumor cell-conditioned media, or IL-1β, in the presence of human recombinant IL-1ra. IL-1ra concentration-dependently suppressed IL-1β- and tumor cell-conditioned media-induced expression of COX-2 and mPGES-1 (Fig. 7D).

We also evaluated levels of IL-1α, IL-1β, and IL-1ra in the culture medium after 48 h of incubation of nontumoral mucosa and tumor-paired samples from 20 patients with a HNSCC. The characteristics of patients included in the study are shown in Table 1. Tumors had upregulated levels of IL-1α and IL-1β when compared with paired nontumoral mucosa (Fig. 8). In contrast, levels of IL-1ra were similar in nontumoral mucosa and tumors samples. Like tumor cell lines, IL-1α levels were higher than those of IL-1β.

TABLE 1.

Characteristics of the patients included in the study of IL-1 and PGE₂ release

	IL-1α, IL-1β, and IL-1ra tumor and nontumoral mucosa (n = 20)	PGE2 in tumor and adjacent mucosa (n = 5)
Location
Oral cavity	4 (20%)	1 (20%)
Oropharynx	1 (5%)	1 (20%)
Hypopharynx	3 (15%)
Larynx	12 (60%)	3 (60%)
T classification
T1	—	1 (20%)
T2	6 (30%)	1 (20%)
T3	6 (30%)	3 (60%)
T4	8 (40%)
N classification
N0	9 (45%)	5 (100%)
N1	1 (5%)
N2	10 (50%)
N3	—
Tumor differentiation
Good	2 (10%)	1 (20%)
Moderate	17 (85%)	4 (80%)
Poor	1 (5%)

Fig. 8.

IL-1α, IL-1β, and IL-1ra released by fresh paired nontumoral mucosa (Mucosa) and HNSCC samples (Tumor) in 48 h. In the Y-axis values are represented as Log₁₀, and text boxes depict mean ± SEM of actual values (n = 20). ***P < 0.001 when compared with mucosa samples.

To explore if HNSCC influenced PGE₂ production in the tumor adjacent mucosa, tumor and nontumoral adjacent mucosa, placed at about 1 cm from the tumor edge, were collected from another set of five patients with HNSCC. The characteristics of these patients are shown in Table 1. PGE₂ levels of in the medium were then analyzed after incubation for 48 h. Levels of PGE₂ released by tumor-adjacent mucosa were higher in than those released by tumor samples in all patients, and data comparison showed that the difference was statistically significant (P < 0.05; tumor, 8.3 ± 3.2; adjacent mucosa: 26.5 ± 7.8 pM/mg tissue, mean ± SEM).

DISCUSSION

In this study, we found that HNSCC cells and fibroblasts produce substantially more PGE₂ when cocultured than when incubated individually. This finding was consistent with the upregulation of COX-2 and mPGES-1 in the fibroblasts after coculture with tumor cells. Tumor cell-conditioned medium induced the release of PGE₂ and PGI₂, indicating that mediators released by tumor cells mobilized AAc from intracellular stores. In addition, COX-2 and mPGES-1 were upregulated by tumor cell-conditioned medium in terms of mRNA and protein. The fact that actinomycin D suppressed the effect indicates that regulation was at a transcriptional level. Time course kinetics showed that induced expression of mPGES-1 was delayed with respect to COX-2. This was consistent with our previous data regarding IL-1β-induced expression of these enzymes in vascular smooth muscle cells (14). In addition, COX-2 and mPGES-1 upregulation by tumor cell-conditioned media was not only observed in dermal fibroblasts but also in the tumor-derived fibroblast cell line CCD-18Co, which suggests this is a general effect.

Tumor cell-conditioned medium also exerted a slight mitogen activity on fibroblasts that was consistent with the presence of many growth factors and cytokines in the tumor cell-conditioned media (Supplementary Table II). The constricted phenotype that we observed in dermal and tumor-derived fibroblasts after treatment with tumor cell-conditioned media seems to be related with the action of cytokines such as TGF-β (33, 34). From among these cytokines, IL-1α and IL-1β were candidates to account for the COX-2- and mPGES-1-inducing activity of tumor cell-conditioned media. Quantitative analysis showed that IL-1α was more abundant than IL-1β in FaDu- and SCC-25-conditioned media. The use of MW-exclusion filters showed that there were two active fractions (10–30 and 30–50 kDa). Despite the fact that tumor cells released many cytokines and growth factors, IL-1ra suppressed the effect of tumor cells on COX-2 and mPGES-1, which indicates that IL-1 was essential for the enzyme-inducing activity. Altogether, these results suggest that IL-1 was released by tumor cells in mature (10- to 30-kDa fraction) and unprocessed form (30- to 50-kDa fraction). The fact that the concentration of IL-1α and IL-1β was one order of magnitude higher and that IL-1ra levels were one order of magnitude lower in FaDu than in SCC-25 could explain that the former was more active in inducing COX-2 in fibroblasts. Indeed, we did not observe induction of COX-2 by SCC-25 in terms of protein. Nevertheless, SCC-25 produced IL-1 in sufficient quantity to induce mPGES-1 in the fibroblasts. Another difference between COX-2 and mPGES-1 induction was that SB203580 (a p38MAPK inhibitor) concentration-dependently inhibited FaDu-CM-induced expression of COX-2 but not mPGES-1. In addition, FaDu-CM caused p38-MAPK phosphorylation in fibroblasts, which was consistent with an effect mediated by IL-1. These results are also consistent with the notion that the expression of COX-2 and that of mPGES-1 are differently regulated.

IL-1β needs to be processed to be active, whereas IL-1α is active in the both mature and unprocessed form (35, 36). Increased levels of IL-1 are commonly found in human tumors and tumor cell lines (35, 37). Consistent with previous reports (38), we found that tumor samples from patients with HNSCC produce 10-fold more IL-1α and IL-1β than paired nontumoral mucosa, IL-1α also being the most abundant isoform, whereas IL-1ra was not up-regulated in HNSCC. Additionally, FaDu and SCC-25 cell lines produced more IL-1α than IL-1β. These results highlight the possible relevance of unprocessed forms of IL-1 in vivo, particularly in the case of IL-1α.

We have found that conditioned medium of fibroblasts stimulated with FaDu-CM had a higher effect on in vitro angiogenesis than those of nonstimulated fibroblasts or FaDu-CM alone. In addition, differences between the effects observed with the different conditioned media were abolished when PGE₂ production was suppressed by the presence of indomethacin. PGE₂ was most likely the prostanoid responsible for the possible effects observed because PGI₂ synthesized by fibroblasts could only be found in its inactive form 6-oxo-PGF_1α when conditioned media were assayed. This was consistent with previous reports showing PGE₂-induced in vitro angiogenesis (32). Conditioned media obtained in the presence of indomethacin still contained the drug when these media were used for the biological assays because it was not possible to extract it from the medium without the risk of altering the contents of rest of components (PGE₂, active proteins, etc.). It is known that indomethacin undergoes hydrolysis in aqueous solutions and that the hydrolysis kinetics depends on the pH, with a maximum stability at pH between 4 and 5 and the hydrolysis being faster at pH >8. Therefore, cells used for the biological activity tests exposed to conditioned media obtained in the presence of indomethacin were likely also exposed to certain concentrations of indomethacin and its hydrolysis products. To minimize this problem, adequate controls with indomethacin were performed, and, before the biological assays, controls with fresh medium with indomethacin were allowed to stand with indomethacin in the culture chamber the same period of time as used to obtain the conditioned media. The contribution of tumor cells or fibroblasts on indomethacin stability during the obtaining of conditioned media could cause some differences in the final concentration of indomethacin between the different conditioned media to which cells used for the biological tests were exposed. These putative differences could represent a limitation of the study.

We observed that the presence of indomethacin in fresh media and in conditioned media from nonstimulated fibroblasts increased the ability of tumor cells to invade the cell-free area in the tumor cell migration tests. PGE₂ concentration in the conditioned media from FaDu-CM-stimulated fibroblasts was about 15-fold higher than that of fibroblasts alone and much higher than that of fresh media, which was undetectable. Therefore, the inhibitory effect of indomethacin on tumor cell migration attributable to PGE₂ inhibition should be more evident when conditioned media from FaDu-CM-stimulated fibroblasts than when conditioned media from nonstimulated fibroblasts or controls was used. The enhancing effect of indomethacin per se on tumor cell migration was unlikely due to the inhibition of PGs in the tumor cells during the assay because PGE₂ was the only PG detectable produced by tumor cells and because it has been shown that PGE₂ exhibited a stimulatory effect regarding tumor cell migration. The exact mechanism of the enhancing effect of indomethacin per se (and/or its hydrolysis products) on tumor cell migration remains unknown. To clarify the exact mechanism of the effect of indomethacin per se is beyond the scope of this study. The effect on tumor cell migration attributable to the PGE₂ was expressed as the ratio of closure without indomethacin to that with indomethacin. The effect of PGE₂ was indeed under-evaluated due to the enhancing effect of indomethacin per se. Despite this fact, we found in the case of conditioned media from fibroblasts stimulated with FaDu-CM that samples without indomethacin showed a higher effect on tumor cell migration than that with the drug, which occurs in the assays of controls and conditioned media from nonstimulated fibroblasts. Collectively, these results strongly suggest that PGE₂ was the responsible for the strong differences between the activity on tumor cell migration of conditioned media from FaDu-CM-stimulated fibroblasts and those from controls or nonstimulated fibroblasts. We also observed that indomethacin-sensitive proliferation of FaDu cell line was substantially increased by conditioned medium from fibroblasts stimulated with FaDu-CM.

The present data indicate that HNSCC could induce PGE₂ biosynthesis in the neighboring stroma cells. This concept was supported by data showing that samples from surrounding nontumoral mucosa produced more PGE₂ than samples from tumor. This is in agreement with data reported by Schuon et al. (39), who found high levels of PGE₂ in the tumor-surrounding nonmalignant mucosa in patients with HNSCC. PGE₂-dependent tumor angiogenesis has been attributed to COX-2 and mPGES-1 expression in tumors such as HNSCC (5, 40–42). The accumulative evidence indicates that cancer development is regulated by interactions between tumor cells and activated stroma cells (23, 24, 43, 44). Indeed, stroma cell-derived PGE₂ plays an important role in intestinal tumor angiogenesis by inducing vascular endothelial growth factor and basic fibroblast growth factor (45). Our results support the idea that induced expression of COX-2/mPGES-1 in the stroma cells could be in many cases a more relevant source of PGE₂ than tumor cells themselves. Collectively, these data suggest that PGE₂ derived from tumor-stimulated fibroblasts could play a major role in HNSCC development and progression.

In conclusion, our results show that prostanoid biosynthesis in the tumor could occur through direct interaction between fibroblasts and tumor cells. Our results show that IL-1 released by tumor cells plays a key role in inducing the expression of COX-2 and mPGES-1 in fibroblasts. They also support the notion that the contribution of fibroblasts to the tumor pool of PGE₂ could be relevant. Because mPGES-1 seems to be essential for a substantial biosynthesis of PGE₂ (13–16, 22), these findings also strengthen the concept that mPGES-1 may be a target for therapeutic intervention in patients with HNSCC.