Abstract
Background: To investigate possible differences in the effects of soluble factors from oral squamous cell carcinoma (SCC) cells (UT-SCC-87) and normal oral keratinocytes (NOK) on fibroblast expression of genes involved in tumor stroma turnover. Materials and Methods: Transwell co-cultures with fibroblasts in collagen gels, and SCC cells or NOK in inserts were carried out. Fibroblast gene expression was measured with real-time polymerase chain reaction (PCR). Results: The expression of urokinase-type plasminogen activator (uPA) and plasminogen activator inhibitor-1 (PAI-1) was up-regulated in co-cultures with SCC cells but not with NOK. In contrast, both SCC cells and NOK regulated matrix metalloproteinase-1 (MMP1) and -3, and tissue inhibitor of metalloproteinases-2 (TIMP2) and -3 to a similar extent, while MMP2 and TIMP1 were largely unaffected. Interleukin 1 alpha (IL1α) up-regulated both MMP1 and MMP3 and down-regulated PAI-1, TIMP2 and -3. Conclusion: SCC and NOK regulate fibroblast expression of genes involved in tumor stroma turnover differentially in vitro. These observations may contribute to a better understanding of the mechanisms behind extracellular matrix turnover in tumors.
Tumor progression requires distinct interactions between malignant and stromal cells. Tumor growth relies on a coordinated growth and adaptation of the tumor stroma, characterized by activated fibroblast phenotypes and deposition and turnover of connective tissue matrix components. The net result is the formation of a dense, fibrotic extracellular matrix (ECM) (1-3). The malignant cells communicate with stromal cells by the release of soluble factors and exosomes, or by transferring signals in direct cell-cell adhesions, which stimulate the desmoplastic response (4). The desmoplasia, in turn, promotes the growth and invasiveness of tumor cells, and finally fibroblasts facilitate further migration of tumor cells in the ECM (5).
In healthy tissue, the basement membrane impedes cell-to-cell contact between the epithelium and the connective tissue stroma, and paracrine mechanisms exerted by secreted soluble factors constitute the main node of communication (6). These interactions regulate the turnover and adaptation of ECM during normal tissue maintenance and growth, and also during inflammation and tissue repair (7). In general, normal keratinocytes, reflecting the function of epithelia in general, maintain connective tissue homeostasis and counteract expression of genes that favor fibrotic responses (8, 9). We have previously shown that malignant and normal keratinocytes affect fibroblast expression of pro-collagens α1(I) and α1(III) differently (10).
As connective tissue turnover is controlled by a balance between synthesis of ECM components and their degradation, the above mentioned studies motivate a further evaluation of whether normal and malignant keratinocytes by virtue of soluble factors regulate the expression of proteases and their inhibitors from fibroblasts differently. So far, there have been searce reports with such direct comparisons between normal epithelial cell–fibroblast interactions, and carcinoma cell–fibroblast interactions. Differences between organotypic co-cultures with malignant and benign keratinocytes with respect to the regulation of the expression of MMP1 and MMP13 have, however, been found (11).
The urokinase-type plasminogen activator (uPA) is a serine protease expressed by several cell types, including fibroblasts (12). Many malignant cells express uPA and use this enzyme to migrate by direct paracrine action in the tumor cell vicinity. The best studied effect of uPA is its ability to convert plasminogen to plasmin which degrades fibrin, ECM proteins, and converts latent matrix metalloproteinases (MMPs) into their active forms (13). The activity of uPA is inhibited by plasminogen activator-1 (PAI-1), which is synthesized by a variety of cells, including vascular endothelial cells, adipocytes, macrophages and fibroblasts (14). PAI-1 also protects MMPs from plasmin-mediated activation and ECM proteins from proteolytic degradation (14).
The MMPs constitute a super family of zinc-dependent endopeptidases capable of degrading ECM proteins, and processing various extracellular bioactive molecules (15). They play a major role in cell adhesion, migration, proliferation and maturation, but also in the turnover and dynamics of the interstitium. There are currently 23 members of the human MMP family (16), at least 10 of which are secreted. Secreted MMPs are classified into three different functional subfamilies. Collagenases, e.g. MMP1, have the ability to degrade fibrillar collagens. Gelatinases, e.g. MMP2, are involved in degradation of collagen type IV in basal lamina, and have crucial roles in angiogenesis (17). Stromelysins, e.g. MMP3, have a broad specificity and degrade many ECM proteins, such as fibronectin, denatured collagens (gelatin), laminin and proteoglycans, but not fibrillar collagens. The expression of both MMP1 and MMP3 in fibroblasts is stimulated by keratinocyte-derived soluble factors (18-20). MMPs are inhibited by a family of tissue inhibitors of metalloproteinases (TIMPs). The inhibition is produced by the formation of a strong non-covalent complex (21).
The purpose of this investigation was to elucidate whether squamous cell carcinoma (SCC) cells and normal keratinocytes exert differential paracrine effects on fibroblast expression of genes important for ECM turnover. In analogy to our earlier studies (8, 10, 22, 23), the focus was placed on mechanisms that involve the release of soluble factors from epithelial cells and the potential role of epithelial-derived interleukin-1α.
Materials and Methods
Cells and reagents. A head and neck SCC cell line (UT-SCC-87) established at the University of Turku, was used in this study. The cell line was established from previously untreated primary tumor of the mobile tongue. UT-SCC-87 was established from a 29-year-old female patient presenting with a T3N1M0 grade 1 SCC. The methods used in establishing and characterizing the cell line have been previously described (24, 25). Normal oral keratinocytes (NOK) were obtained from the gingiva of a 28-year-old male, previously healthy and non-smoking. Human primary dermal fibroblasts were obtained from patients undergoing reconstructive breast surgery at the Department of Plastic Surgery. Fibroblasts from one healthy female were used for the experiments. Approval from the local Ethical Committee (Uppsala University) was obtained (#2005:332). This ethical approval entails written informed consent and written patient information.
PCR primers for uPA, PAI-1, MMP1, MMP2, MMP3, TIMP1, TIMP2, TIMP3 and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) came from Applied Biosystems (Cheshire, UK). Anakinra (Kineret®) came from Amgen (Breda, the Netherlands). For real-time PCR the following probes were used: uPA Hs00170182_m1; PAI-1 Hs00167155_m1; MMP1 Hs00233958_m1; MMP2 Hs00234422_m1; MMP3 Hs00233962_m1; TIMP1 Hs00171558_m1; TIMP2 Hs00234278_m1; TIMP3 Hs00165949_m1; and GAPDH Hs99999905_m1.
Cell isolation and culture. Skin samples were treated with dispase and epidermis mechanically separated from the underlying dermis. NOK were isolated as previously described (26). Following mechanical fragmentation, the epidermis was treated with trypsin and keratinocytes were propagated on irradiated 3T3 feeder cells (ATCC® CCL-92™; ATCC, Manassa, VA, USA) in Dulbeccos's Modified Eagle's Medium (DMEM):Ham's F12 (4:1) supplemented with 5 μg/ml Zn-free insulin (Sigma Chemical Co., St. Louis, MO, USA), 2 nM 3,3’,5-triido-D-thyronine (Sigma Chemical Co.), 0.4 μg/ml hydrocortisone (Sigma Chemical Co., St. Louis, MO, USA), 0.1 nM cholera toxin (Sigma Chemical Co.), 10 ng/ml epidermal growth factor (EGF; Austral Biologicals, San Ramon, CA, USA), 24 μg/ml adenine, 10% fetal bovine serum (FBS; HyClone, Logan, UT, USA), and 50 μg/ml gentamicin. NOK in passage 2-5 were used. Fibroblasts were isolated from dermal compartment by treatment with collagenase and subcultured in DMEM with 10% bovine calf serum (HyClone) and 50 μg/ml gentamicin. Subconfluent cells were washed with phosphate buffered saline (PBS) and detached with trypsin. Cells in passage 1-5 were used for the experiments.
Co-culture. Fibroblasts were cultured in collagen gels. Briefly, a cold solution of 1.6 ml collagen type I (Vitrogen), 0.15 ml (10×) Hank's balanced salt solution (10×HBSS) and 0.15 ml (FBS) with or without fibroblasts (2×105/well), was adjusted to pH 7.4 with 5 M NaOH and added to 6-well plates. After polymerization at 37°C, 2 ml DMEM with 10% bovine calf serum (HyClone) were added to each well. NOK, or malignant keratinocytes at passage 16 or 17, were seeded in Falcon polyurethane cell culture inserts (4.0-μm pore diameter). The inserts were pre-incubated with a mixture of 10 μg/ml bovine plasma fibronectin (Gibco BRL/Life Technology, Paisley, UK), 30 μg/ml bovine collagen (Vitrogen, Cohesion, Palo Alto, CA,USA) and 10 μg/ml bovine serum albumin (Sigma Chemical Co.) for two hours at 37°C. NOK and malignant oral keratinocytes were seeded in inserts, both in DMEM:Ham's F12 (4:1) supplemented with 5 μg/ml Zn-free insulin (Sigma Chemical Co.), 2nM 3,3’,5-triido-D-thyronine (Sigma Chemical Co.), 0.4 μg/ml hydrocortisone (Sigma Chemical Co.), 0.1 nM cholera toxin (Sigma Chemical Co.), 10 ng/ml EGF (Austral Biologicals), 24 μg/ml adenine, 10% fetal bovine serum (HyClone), and 50 μg/ml gentamicin. After 24 hours, the medium was changed in wells and in inserts to 2 ml DMEM/Ham's F12 (4:1) supplemented with 0.5% fetal calf serum (HyClone) in each. Inserts and wells with collagen gels were combined and propagated as cocultures for an additional 48 hours. As control, 0.15×106 fibroblasts were seeded in inserts instead of keratinocytes. Experiments were performed with different seeding densities of NOK and malignant keratinocytes, and the number of cells after 48 hours was assessed with a cell counter (Z2 Cell and Particle Counter; Beckman Coulter AB, Brea, CA, USA). The different seeding numbers of normal and malignant cells were to compensate for an observed higher proliferation of malignant cells. After titration with different seeding densities, a density of 0.30×106 for NOK and 0.16×106 for malignant cells was chosen for the experiments. In this way, the average cell number at the termination of co-cultures was 0.45×106 for NOK and 0.43×106 for malignant cells, and about 80% cell density was reached in all samples at the end of co-culture experiments.
RNA extraction. Collagen gels were dissolved and RNA was extracted using a modification of the one-step phenol-chloroform method (27). Gels were dissolved in TRIzol reagent (Life Technology) under extensive mixing (vortexing). The amount and quality of RNA was determined using an Agilent Bioanalyzer 2100 (Agilent Technology, Kista, Sweden).
Real-time PCR. One microgram total RNA per sample was used as template for synthesis of cDNA using a commercial kit [First Strand cDNA Synthesis Kit for RT-PCR (AMV), Roche Diagnostics Scandinavia AB, Bromma, Sweden]. Real-time semiquantitative PCR was performed on a LightCycler (Roche) using commercially available fluorescent probes from Applied Biosystems (Foster City, CA, USA). PCR reagents were obtained from Roche (LightCycler FastStart DNA Master Hybridization Probes Kit). Crossing point (Cp) values were calculated by the LightCycler software using the second derivative maximum method. Three cDNA samples for each gene product to be analyzed were diluted 10 fold prior to each round of amplification in the LightCycler. Cp values from these samples were subtracted from corresponding undiluted samples (ΔCp). PCR efficiency (E) was then calculated by the formula E=10(1/ΔCp). Relative levels of gene expression (RL) were then calculated using the formula RL=E(Cp ref-Cp test) where Cp ref is the Cp value for an arbitrary chosen reference sample, to which the others are compared. A ratio between the RL for the gene of interest and that for GAPDH was calculated for each sample.
Statistics. A total of 10 separate co-cultivation experiments were carried out. All comparisons between groups were made with paired Student's t-tests with Bonferroni-Holm correction for multiple tests (28). The number of experiments in each paired analysis varied between 5 and 10. A value of p≤0.05 was considered a significant difference.
Results
SCC cells and NOK differentially regulated the expression of plasmin regulators uPA and PAI-1 in fibroblasts (Figure 1 A and B). SCC cells did not affect the expression of uPA and PAI-1, while both genes were up-regulated by NOK. Both normal and malignant keratinocytes strongly up-regulated the expression of MMP1 and MMP3 (Figure 1 C and E). However, MMP1 was significantly less up-regulated by SCC cells compared to NOK. There was no effect of co-culture with either SCC cells or NOK on the expression of MMP2 mRNA levels (Figure 1 D). Furthermore, such co-cultures did not affect the expression of TIMP1 by fibroblasts (Figure 1 E). In contrast to this, both SCC cells and NOK caused a substantial down-regulation of the expression of TIMP2 and TIMP3.
The presence of the interleukin-1 (IL1) inhibitor, IL1ra Anakinra (Kineret®), was included in five out of the 10 experiments presented in Figure 1. This inhibition of IL1α had no effect on the increased expression of uPA by fibroblasts in co-cultures with NOK (Figure 1A). IL1ra partially abrogated the increased expression of PAI-1 in these cultures (Figure 1B). Although there was a numerical reduction of the increased expression of MMP1 and MMP3, this was not significant (Figure 1C and D). Finally, the inhibitory effect of SCC cells and NOK on expression of TIMP2 and TIMP3 was partially reversed, with the effects of SCC cells on TIMP2 and of NOK on TIMP3 reaching statistical significance (Figure 1 G and H).
Discussion
Here we show that normal but not malignant keratinocytes up-regulate the expression in of the genes for uPA and PAI-1 in fibroblasts, in a three-dimensional co-culture model only allowing for paracrine interaction without direct cell-to-cell contact.
This study has an important limitation in that only one malignant cell line, UT-SCC-87, was compared to primary oral keratinocytes from one patient. We have previously shown that the effects of UT-SCC-87 on fibroblast gene expression for ECM components, are similar to those of another oral SCC cell line, UT-SCC-30 (10). Nevertheless, further studies including additional carcinoma cell lines, as well as primary NOK from several patients, are required to intensify our observations. Importantly, in the present investigation, a relatively large (n=10) number of experiments were performed to cope with an expected large inter-assay heterogeneity in cultures with carcinoma cell lines. Thus, we feel that the main findings on differential uPA and PAI-1 regulation have been clearly established in the present cellular and experimental context, and constitute a firm base for further investigations.
We used non-contracted fibroblast-populated type I collagen gels in order for the cells to form cell matrix interactions and adopt an in vivo-like three-dimensional configuration. Fibroblasts in collagen gels have a lower, more controlled, metabolic and proliferative activity, compared to fibroblasts in monolayer cultures (29). Thus, fibroblast-populated non-contractile gels will contain cells with important phenotypic similarities to the myofibroblasts of tissue repair, and to the cancer-associated fibroblasts of tumors. Moreover, this model should constitute a system well-adapted for analyzing both stimulatory and inhibitory effects on gene expression in fibroblasts. The keratinocytes were seeded in a separate insert on a type I collagen-fibronectin substrate. The number of NOK was chosen to produce about 80-90% confluence at the end of the 48-hour co-culture period (data not shown). This experimental design maintained the NOK in an activated migratory and proliferative phenotype, corresponding to the process of re-epithelialization, throughout the experiment. Taken together, this model allowed us to explore fibroblast responses to keratinocyte-derived factors in a wound healing-like scenario and compare it to that of a malignant process.
Regulation of gene expression in fibroblasts by normal (NOK) and malignant keratinocytes (SCC-87). Keratinocytes were co-cultured with fibroblasts for 48 h as described in Materials and Methods. Fibroblast gene expression was measured with real-time PCR. Data are presented as gene expression relative to fibroblasts only, column F. The two right outermost columns in each graph represent experiments with the addition of 250 μg/ml of the interleukin-1 (IL1) receptor antagonist rIL1ra. Data are mean values±SEM of ten experiments. Paired t-test. Statistically significant at *p<0.05, **p<0.01 and ***p<0.001, respectively. Panel A: Urokinase-type plasminogen activator (uPA); B: plasminogen activator inhibitor-1 (PAI-1); C: matrix metalloproteinase-1 (MMP1); D: matrix metalloproteinase-2 (MMP2); E: matrix metalloproteinase-3 (MMP3); F: tissue inhibitor of metalloproteinase-1 (TIMP1); G: tissue inhibitor of metalloproteinase-2 (TIMP2); H: tissue inhibitor of metalloproteinase-3 (TIMP3).
It is well-established that IL1α is a major keratinocyte/malignant keratinocyte-derived factor, regulating fibroblast gene expression, in keratinocyte–fibroblast co-cultures. We have previously shown that both the SCC cells used in the present study as well as NOK release IL1α independently of fibroblasts, and also that the SCC cell line used here (UT-SCC-87) releases about three times as much IL1α as do NOK (10). Several of the effects observed in our system were indeed typical responses to IL1α, such as up-regulation of MMP1 and MMP3. In fact, all effects but the regulation of the plasmin regulators uPA and PAI-1, were at least partially reversed by IL1α inhibition.
There was an IL1α-mediated down-regulation of the MMP inhibitors TIMP2 and -3 in co-cultures of NOK and SCC cells, while the expression of TIMP1 was unaffected. The role of TIMPs in tumor progression has been shown to be highly complex. TIMPs have been found to inhibit invasion, metastasis and angiogenesis (30). Taken together, these data support the notion that keratinocytes have the capacity to activate an increase in fibroblast-mediated ECM turnover, through the secretion of IL1α, in wound healing as well as in the tumor environment.
The main purpose of this study was to identify differences between NOK and SCC cells on how they regulate the expression of genes in fibroblasts that play a role in the turn over of ECM. The main finding in this regard was the demonstration of a differential regulation of gene expression for the two plasmin regulators, uPA and PAI-1, which have been shown to be highly important during wound healing and in tumor progression. The expression of both uPA and PAI-1 was up-regulated by NOK and not affected by SCC cells. Moreover, PAI-1 expression was further accentuated in NOK co-cultures and up-regulated in SCC co-cultures by IL1 inhibition, suggesting that keratinocyte-derived IL1 down-regulates its expression, and that this mechanism is counter-regulated by other factors secreted in co-culture with keratinocytes. This PAI-1-up-regulating, IL1α-antagonistic mechanism appeared to be more pronounced in co-cultures with NOK, since total IL1α inhibition, as achieved by a 250 μg/ml concentration of Kineret®, produced stronger PAI-1 up-regulation in co-cultures with NOK than SCC cell. Taken together, these data suggest that the malignant keratinocytes used in this investigation were deficient in their ability to up-regulate the expression of both uPA and PAI-1 in fibroblasts through paracrine mechanisms. Pertaining to this, while the expressions of uPA and PAI-1 have been shown to be elevated in oral SCC carcinoma tissue, it has also been demonstrated that cell-to-cell contact is required for SCC, and pancreatic carcinoma cells, to induce uPA expression in fibroblasts (31-33).
The basement membrane normally separates the epithelium and underlying connective tissue compartment, which is why keratinocytes and fibroblasts mainly communicate through soluble factors. This normal tissue architecture is disrupted by tissue injury and by malignant processes, allowing for an increase in epithelial–mesenchymal communication through direct cell-to-cell contact. This is particularly true for malignant tumors, characterized by a disordered tissue architecture with a mixture of carcinoma cells and stromal cells in close apposition. In view of the above, the lack of up-regulation of uPA and PAI-1 in co-cultures with SCC cells could represent a phenotypic alteration that fits with a co-existence of cancer cells and fibroblasts in close proximity. Such fibroblast–carcinoma cell juxtapositioning has been shown to play a role in cancer cell invasion and migration. Possibly, abundant direct cell-to-cell interactions between fibroblasts and cancer cells in the tumor milieu could also be important for the build-up of a tumor-supporting stroma.
Importantly, given the heterogeneity of SCC of different stages and tissues, the results presented here clearly need further verification before general conclusions can be drawn. Nevertheless, these results point towards possible differences between healthy and malignant keratinocytes that warrant further investigation.
Acknowledgements
We thank Ms Marja Bostrom for excellent technical support. We thank the Thureus foundation and the Swedish Cancer Foundation for funding.
- Received April 28, 2013.
- Revision received June 4, 2013.
- Accepted June 6, 2013.
- Copyright© 2013 International Institute of Anticancer Research (Dr. John G. Delinassios), All rights reserved
