Cancer-Associated Fibroblasts Induce Matrix Metalloproteinase–Mediated Cetuximab Resistance in Head and Neck Squamous Cell Carcinoma Cells

Cancer of the head and neck is the sixth most common form of cancer worldwide with 650,000 new cases each year. The majority of all head and neck cancers are squamous cell carcinomas (HNSCC) with origins in the mucosa of the oral cavity, throat, nose or sinuses. Platinum-based chemoradiotherapy is considered to be the standard treatment for locally advanced HNSCC. However, clinical drug resistance remains a major problem, and the overall survival rate for patients with HNSCC has not increased over the past decades. This treatment deficiency stresses the need for new therapeutic strategies and biomarkers that are predictive of the treatment response.

The EGF receptor (EGFR; also called ErbB1 or HER1) is a transmembrane receptor tyrosine kinase that belongs to the HER family of receptors. Upon ligand binding, the receptor dimerizes, which leads to the autophosphorylation and activation of the intracellular tyrosine kinase domain (1). Kinase activation triggers multiple downstream signaling pathways involving MAPK, PI3K/Akt, STATs, Src, and PLCγ, which regulate key cellular processes including proliferation and survival. EGFR is frequently overexpressed in a number of human solid tumors, including HNSCC, where the high expression of EGFR correlates with a poor clinical outcome (2–3).

Due to the relationship between the overexpression of EGFR and the aggressive behavior of tumor cells, therapies aiming to prevent EGFR signaling, including monoclonal antibodies and small molecule tyrosine kinase inhibitors, have been developed. Cetuximab (Erbitux) is a monoclonal antibody that inhibits tumor growth by binding to the extracellular domain of EGFR to prevent ligand binding (4–6). In addition, cetuximab may also stimulate the internalization of EGFR, which leads to the downregulation of its cell surface expression, and may also trigger antibody-dependent, cell-mediated cytotoxicity (7–8). Cetuximab treatment has been proven to be effective in recurrent or metastatic HNSCC both as first-line treatment in combination with platinum-based chemotherapy (9–12), and as second-line treatment in patients with platinum-refractory disease (13–15). Cetuximab was also shown to improve locoregional control and overall survival when combined with radiotherapy in the first-line treatment of patients with locoregionally advanced HNSCC (16). However, the cetuximab response rate is generally not higher than 20%, highlighting the need for predictive markers of cetuximab treatment response. Increased knowledge regarding the molecular mechanisms by which tumor cells resist cetuximab treatment would enable the selection of patients that would benefit from treatment. Ideally, these resistance mechanisms would also be targetable and would thus provide a possibility to increase the efficacy of cetuximab treatment.

For many years, cancer research has focused mainly on the cancer cell and its molecular changes; however, the stromal microenvironment is now generally accepted to contribute to tumorigenesis in cancers of an epithelial origin (17). Already in premalignant dysplasia, fibroblasts, endothelial cells, and immune cells are recruited to the stroma. The progression from dysplasia into invasive cancer involves the disruption of the basement membrane barrier to enable direct contact between the tumor cells, stromal cells, and extracellular matrix (ECM). During this progression, the stroma undergoes a transformation into a reactive tumor stroma that supports tumor growth and metastasis. Cancer-associated fibroblasts (CAF), the major cellular component of the tumor stroma, may derive from different subsets of cells and likely display functional differences depending on their cancer type of origin. CAFs provide protumorigenic signals that lead to increased tumor growth via the stimulation of tumor cell proliferation and angiogenesis as well as the formation of metastases by the enhancement of the migratory and invasive potential of the tumor cells (18–25). As a number of publications suggest that CAFs could also modulate the drug sensitivity of cancer cells (26–34), this study was undertaken to evaluate the influence of CAFs on the cetuximab response in HNSCC cell lines.

Head and neck squamous cell carcinomas cell lines (UT-SCC-9, UT-SCC-24A, UT-SCC-19A, and UT-SCC-76A used in passages 28–39, 16–29, 12–15, and 15–18, respectively) were established at the Grénman laboratory in Turku and cultured in Dulbecco's Modified Eagle's Medium supplemented with 2 mmol/L glutamine, 1% nonessential amino acids, 50 IU/mL penicillin, 50 μg/mL streptomycin, and 5% fetal bovine serum (FBS; all from GIBCO). The cells were incubated in humidified air with 5% CO₂ at 37°C and subcultured twice a week using 0.25% trypsin with 0.02% EDTA.

Cancer-associated fibroblasts were isolated from 7 patients with HNSCC that had given their informed consent (approved by the Linköping University Ethical Committee). Small pieces of tumor tissue were distributed at the bottom of 25 cm² cell culture flasks that had been precoated with serum, and 2 mL of keratinocyte serum-free media (GIBCO) supplemented with 50 IU/mL penicillin, 50 μg/mL streptomycin, 2.5 μg/mL amphotericine B (Fungizone; GIBCO), and 20% FBS (Integro B.V.) was added. The tissue cultures were incubated at 37°C in humidified air with 5% CO₂, and the media were changed every 3 to 4 days. CAFs were detached and separated from the epithelial cells by differential trypsinization and subsequently cultured in Dulbecco's Modified Eagle's Medium supplemented with 2 mmol/L glutamine, 1% nonessential amino acids, 50 IU/mL penicillin, 50 μg/mL streptomycin, and 5% FBS. All CAF cultures showed homogenous, positive immunofluorescent staining for vimentin and negative staining for cytokeratin (data not shown). Seven different CAF cultures originating from tumors of the tongue, larynx, tonsil, gingiva, and buccal mucosa were used in this study. For all experiments, CAFs were used at passages 2 to 7.

CAF-conditioned media were collected from confluent cultures 72 hours after the media were changed, precleared by a 5-minute centrifugation at 2,060 × g, and stored at −20°C until use. For size-fractionation of the conditioned media, centrifugal filter units (Amicon ultra-4, Millipore, Billerica) were used with molecular weight cutoffs of 50 and 100 kDa (3,000 × g for 45 minutes).

The tumor cells were seeded in 12-well plates (BD Falcon, Franklin Lakes) at a density of 200 (UT-SCC-24A), 625 (UT-SCC-76A), or 750 (UT-SCC-9 and -19A) cells/cm² depending on the plating efficiency of each cell line. After 48 hours, cetuximab (30 nmol/L; Erbitux) or gefitinib (50 nmol/L; Iressa; AstraZeneca) was added, and the cells were incubated for another 8 days before the cytostatic/cytotoxic effect of the drug was evaluated by crystal violet staining. Where indicated, the cultures were pretreated with 1 μmol/L of MMP inhibitor III (Calbiochem), 100 nmol/L Met inhibitor PHA-665752 (Sigma-Aldrich), or 0.1 to 5 ng/mL of recombinant human hepatocyte growth factor (HGF; Sigma).

The effect of CAFs on the sensitivity of the tumor cells to cetuximab and gefitinib treatment was evaluated by coculture using transwell polycarbonate membrane filters with a 0.4 μm pore size (Falcon, BD Biosciences) or by the addition of fibroblast conditioned media (25% v/v). The tumor cells were allowed to attach before the CAFs were seeded in the filters at a density of 3,300 cells/cm², which resulted in approximately a 1:1 ratio of CAFs to tumor cells.

The cells were fixed in 4% paraformaldehyde (PFA) for 20 minutes at room temperature (RT) and stained with 0.04% crystal violet in 1% ethanol (20 minutes, RT). The plates were then washed extensively under running tap water and air dried. After solubilization of the samples in 1% SDS, the optical density values were read using a Victor plate reader (EG & G Wallac) at 550 nm.

Fibroblasts grown on glass coverslips were fixed with 4% PFA in phosphate-buffered saline (PBS) for 20 minutes at 4°C and subsequently incubated with 0.1% saponin and 5% FBS in PBS for 20 minutes at RT. The cells were then incubated with a monoclonal mouse antihuman vimentin antibody (1:50; Santa Cruz Biotechnology) or a monoclonal mouse antihuman multicytokeratin antibody (1:50; Novacastra) followed by incubation with an Alexa Fluor 488 monoclonal goat antimouse antibody (1:400; Invitrogen). Finally, cells were rinsed in PBS and distilled water, mounted in Vectashield Mounting Medium (Vector Laboratories), and examined with an Olympus photomicroscope (Olympus) using a blue excitation light and a green barrier filter. The controls incubated without the antivimentin or anticytokeratin antibodies did not show any staining.

The cells were washed in PBS and lysed in 63 mmol/L Tris-HCl buffer (pH 6.8) containing 10% glycerol, 2% SDS, 20 μmol/L dithiotreitol, 1% bromphenol blue (all from Sigma-Aldrich), Protease arrest (G-Biosciences), and phosSTOP (Roche Diagnostics GmbH). The protein concentration was determined as described by Lowry and colleagues (35), and the samples were subjected to gel electrophoresis. The proteins were then transferred to a nitrocellulose membrane using an iBlot system (Invitrogen). The proteins were detected using rabbit anti-EGFR (Santa Cruz Biotechnology) and mouse antiphospho-EGFR antibodies (Tyr1173; Millipore) and visualized by HRP-conjugated goat antirabbit and goat antimouse antibodies (Santa Cruz Biotechnology). Equal loading was verified using a HRP-conjugated goat antiactin antibody (Santa Cruz Biotechnology).

The quantitative determination of HGF and MMP-1 levels in conditioned media was carried out according to the manufacturer's instructions using the Quantikine (human HGF) and Flourokine E Enzyme Activity Assay (human active MMP1; both from RnD Systems). The optical density for HGF and the relative fluorescence for MMP-1 were measured using the Versa Max (Molecular Devices Corp) and Victor (EG & G Wallac) microplate readers, respectively. All analyses were carried out twice in duplicate, and the mean value was used for further investigations.

Total RNA was extracted with the RNeasy Mini Kit (Qiagen), and cDNA was obtained using the High Capacity RNA-to-cDNA Kit (Applied Biosystems). The mRNA expression of HGF and MMP-1 was analyzed using a 7,500 Fast Real-Time PCR system and FAM/MGB probes (Applied Biosystems). All reactions and conditions were carried out according to the manufacturer's instructions.

Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was amplified as an internal standard. The data were calculated using the comparative Ct method to present the data as a fold difference in the expression level relative to a control sample (36).

The expression of 87 genes (Supplementary Table S1) primarily involved in the epithelial to mesenchymal transition, EGFR signaling, and treatment resistance was analyzed using a custom-made PCR array plate from SABiosciences. The PCR reaction was carried out according to the manufacturer's protocol.

Total RNA was extracted as described above, and cDNA was acquired using the RT² First Strand Kit (SABiosciences). DNA and RNA contamination was assessed using a genomic DNA control and a positive PCR control, respectively. The average of 5 housekeeping genes [beta-2-mictoglobulin (B2M), hypoxanthine phosphribosyltransferase 1 (HPRT1), ribosomal protein L13a (RPL13A), GAPDH, and beta-actin (ACTB)] was used to normalize the Ct-values. The relative change in mRNA expression was calculated by the comparative Ct method.

Cells were seeded at a density of 6,700 cells/cm² in 6-well plates and transfected 24 hours later with siRNA targeting HGF or MMP-1 or a nontargeting siRNA with no homology to any known human gene (AllStars Negative Control siRNA; all from Qiagen) using the HiPerFect transfection reagent (Qiagen). The final concentration of siRNA in the culture medium was 10 nmol/L. Gene silencing was verified by quantitative real-time PCR (qPCR) and ELISA at 48 and 96 hours after transfection, respectively. A 70% knockdown was considered acceptable for further studies.

For cetuximab sensitivity studies, the cells were reseeded in 12-well plates, as described above, 24 hours after transfection. To obtain media conditioned by siRNA-treated cultures, the media were replaced 24 hours after transfection, and the conditioned media were collected after an additional 72 hours.

Cetuximab treatment (30 nmol/L) resulted in a 40% and 60% reduction of cell growth in the HNSCC UT-SCC-9 (larynx) and UT-SCC-24A (tongue) cell lines, respectively, suggesting that these cells, in part, rely on EGFR signaling to maintain a high proliferation rate. However, when the tumor cells were cocultured with HNSCC-derived CAFs in a transwell system, they were less susceptible to cetuximab treatment. All investigated CAFs completely abolished the cetuximab-induced growth inhibition in the UT-SCC-9 cell line (Fig. 1A and Supplementary Fig. 1A). Interestingly, during the coculture of UT-SCC-9 cells with LK0826Fib, LK0850Fib, or LK0862Fib, cetuximab treatment even stimulated cell growth compared with the respective untreated cocultures. A more modest effect on the treatment response was observed in the UT-SCC-24A cell line where 4 of 7 CAFs significantly reduced the cetuximab sensitivity, but a complete resistance to treatment was not achieved (Fig. 1B and Supplementary Fig. 1B). The results from the UT-SCC-19A and UT-SCC-76A cell lines, which are derived from the larynx and tongue, respectively, showed that the magnitude of the protection conferred by CAFs was not dependent on the tumor location (Fig. 1C and D). Interestingly, CAF cultures were completely resistant to cetuximab treatment, which indicates that mitogenic pathways other than EGFR are dominant in these cells (Fig. 1E). Collectively, these data suggest that under the influence of CAFs tumor cells are less susceptible or even resistant to the growth inhibition caused by cetuximab treatment.

In UT-SCC-9 cultures, which showed the most pronounced CAF-mediated induction of cetuximab resistance, CAFs did not alter the proliferation rate in the absence of cetuximab (Fig. 2A). However, in the other 3 cell lines that showed a more modest protection from cetuximab, the coculture of the tumor cells with CAFs resulted in a significant stimulation of cell growth (Fig. 2B–D). These data indicate that the cetuximab resistance conferred by CAFs is not dependent on the ability of the fibroblasts to stimulate tumor cell proliferation.

To establish whether CAFs specifically interfere with the function of cetuximab or rather if they suppress the effect of EGFR inhibition in general, the cells were treated with the EGFR-selective tyrosine kinase inhibitor gefitinib (50 nmol/L). Gefitinib reduced cell growth by 50% and 75%, in UT-SCC-9 and UT-SCC-24A monocultures, respectively (Fig. 3A and B). Similar to the results observed for cetuximab, CAFs provided almost complete protection against the effects of gefitinib treatment in UT-SCC-9 cultures but showed a less pronounced effect in UT-SCC-24A cells. As expected, CAFs did not respond to gefitinib treatment (Fig. 3C). These data suggest that CAFs do not interfere with the interaction between cetuximab and EGFR but rather trigger signaling events that counteract the effects of general EGFR inhibition.

Media conditioned by CAFs induced cetuximab resistance in UT-SCC-9 and UT-SCC-24A monocultures, suggesting that CAF-derived soluble factors mediate the protective effect of these cells (Fig. 4A and B and Supplementary Fig. 2). The cetuximab resistance in UT-SCC-9 increased with increasing amounts of media conditioned by LK0862Fib (Fig. 4C). When tumor cell cultures were treated with cetuximab in the presence of size-fractioned conditioned media, the protective effect was observed in the fraction containing factors larger than 50 kDa (Fig. 4D). Interestingly, both the less than 100 kDa and more than 100 kDa fractions conferred cetuximab resistance albeit at a lower level than the unfractionated conditioned media. These results suggest that several resistance factors may cooperate to induce the cetuximab resistance in tumor cells.

To further examine the protective mechanisms underlying the CAF-induced cetuximab resistance, the mRNA and protein levels of EGFR and its downstream effectors were examined. For these studies, UT-SSC-9 and LK0862Fib cultures were used as the LK0862Fib offered the UT-SCC-9 cells a high level of protection from cetuximab treatment in both coculture and conditioned media experiments.

No changes in the EGFR mRNA or protein levels were found in the UT-SSC-9 cells upon coculture with CAFs (Fig. 5A and B). The amount of phosphorylated EGFR, which was very low in untreated cultures, increased in response to EGF stimulation. In cetuximab-treated cultures, the EGF-induced increase in phosphorylated EGFR was prevented. Importantly, cetuximab also blocked EGFR phosphorylation in the coculture, showing that coculture did not confer resistance by preventing the EGFR-antagonistic effect of cetuximab. Furthermore, no changes in the expression or phosphorylation status of AKT or ERK 1/2 were seen upon coculture with CAFs that could explain the treatment resistance observed (data not shown).

As CAFs were previously shown to induce resistance to EGFR tyrosine kinase inhibitors by secretion of HGF (29, 37), the possible involvement of this growth factor in the CAF-mediated cetuximab resistance was investigated. Tumor cells, which lacked HGF expression, stimulated the secretion of HGF when cocultured with CAFs (Fig. 6A). Adding recombinant HGF confirmed that HGF could act as a resistance factor at relevant concentrations (Fig. 6B). However, the media collected from CAFs transfected with siRNA targeting HGF still protected tumor cells from cetuximab treatment (Fig. 6C). Moreover, a Met inhibitor also failed to prevent treatment resistance, thus, the possibility that HGF contributes to the CAF-induced cetuximab resistance was excluded (Fig. 6D).

In an attempt to identify the cause of the reduced sensitivity of tumor cells to cetuximab treatment, a PCR array analyzing the expression of 87 genes was carried out in mono- and cocultured tumor cells (Supplementary Table S1). The expression levels of 2 proteins, matrix metalloproteinase-1 (MMP-1) and the EGFR ligand amphiregulin, were found to increase by more than 2-fold in cells cocultured with CAFs. Recombinant amphiregulin failed to protect tumor cells from cetuximab treatment and was therefore ruled out as a possible resistance factor (data not shown). Using quantitative PCR, the increased expression of MMP-1 in tumor cells cocultured with CAFs was confirmed (Fig. 7A). Furthermore, an even greater upregulation of MMP-1 mRNA expression was observed in cocultured CAFs. We failed to detect any active MMP-1 secreted by the tumor cells; however, in accordance with the mRNA data, a higher level of active MMP-1 was found in the media collected from the cocultures compared with the media conditioned by CAFs alone (Fig. 7B). An inhibitor of MMP-1 significantly reduced the protective effect of CAF-conditioned media (Fig. 7C); however, media collected from CAFs transfected with siRNA targeting MMP-1was as effective in inducing resistance as media from MMP-1 expressing CAFs (Fig. 7D). The downregulation of MMP-1 expression in tumor cells also failed to increase their cetuximab sensitivity but rather resulted in the opposite effect (Fig. 7E).

Because the MMP-1 inhibitor also blocks several other MMPs (MMP-2, -3, -7, and -13), these results collectively indicate that one of the other MMPs may confer protection to cetuximab treatment or that several CAF-regulated MMPs cooperate in the induction of drug resistance.

The importance of the tumor microenvironment for tumor growth and metastasis has been recognized for many years. In 1990, Teicher and colleagues proposed that something other than the inherent properties of tumor cells must affect the ability of these cells to resist cytotoxic drugs (38). Since then, several studies have shown that fibroblasts can induce resistance to various anticancer treatments, such as tamoxifen in the treatment of breast cancer (28, 39), gemcitabine and radiotherapy in the treatment of pancreatic adenocarcinoma (26), and paclitaxel in the treatment of non–small cell lung cancer (30). Conversely, fibroblasts may also sensitize tumor cells to therapy as was shown for RAF and MEK1 inhibitors in breast cancer (33). However, the influence of CAFs on the cetuximab treatment response in HNSCC has not previously been investigated.

Here, we show that CAFs isolated from patients with HNSCC desensitize HNSCC cell lines to the EGFR-targeted therapies cetuximab and gefitinib by the secretion of unidentified soluble factors. The protective effect was partially dependent on MMP activity, which suggests that CAFs may release MMPs or factors that influence the expression or activation of tumor cell-derived MMPs. To our knowledge, this is the first paper that shows a CAF-dependent induction of tumor cell cetuximab resistance. In agreement with our findings regarding the outcome of gefitinib treatment, it was recently reported that human lung embryonic fibroblasts and CAFs from patients with lung cancer reduce the effect of this drug in lung cancer cell lines harboring EGFR-activating mutations (29). These mutations are very rare in HNSCC (40–41). Thus, our data provide evidence that CAF-mediated resistance might also operate in tumors lacking EGFR-activating mutations.

Full protection from cetuximab-induced growth inhibition was observed in UT-SCC-9 cells, whereas the influence of CAFs on the treatment response was less pronounced in the 3 other cell lines tested. This discrepancy in the magnitude of the protection conferred by CAFs was independent of the location of the tumor from which the cell lines originate. As similar results were obtained using CAF-conditioned media, it is unlikely that the particular sensitivity of UT-SCC-9 involves a greater capacity of these tumor cells to induce the secretion of CAF-derived resistance factors. Thus, other inherent properties of the tumor cells must determine the extent to which CAFs influence the treatment response. When identified, these properties could be useful for the prediction of the cetuximab treatment outcome.

Our data suggest that factors associated with both CAFs and tumor cells determine the outcome of EGFR-targeted therapy. This indicates that successful individualized treatment of patients with HNSCC is possible only if predictive factors of the treatment response are identified both in tumor cells and the tumor microenvironment. The identification of stromal resistance factors that hamper the clinical response of EGFR-targeted therapy may not only provide useful biomarkers, but also open new treatment possibilities. If one could target the crosstalk between cancer cells and CAFs, the efficacy of cetuximab treatment in patients with HNSCC could be greatly improved.

Importantly, in the presence of 3 (LK0826Fib, LK0850Fib, and LK0862Fib) of 7 CAFs tested, the growth inhibition caused by cetuximab treatment in the UT-SCC-9 cell line was not only abolished, but also replaced by a growth stimulation. This result further highlights the need for biomarkers to predict the cetuximab treatment response. These biomarkers would enable not only the identification of patients that would benefit, but also those that could potentially be harmed by cetuximab treatment.

Much effort has been put into the development of MMP inhibitors to treat cancer. Although an impressive effect was observed for MMP inhibitors in in vitro and mouse models, the results from clinical trials have been disappointing (42–43). The failure of these drugs may in part be explained by the fact that many of the patients recruited in these trials were at the most advanced, metastatic stages of cancer. As is the case for most anticancer drugs, MMP inhibitors may be more effective when given at earlier stages of cancer progression. Our data suggest that MMPs are involved in the development of stroma-induced resistance to EGFR-targeted therapies. Furthermore, previous studies have indicated a correlation between MMPs and the resistance to cytotoxic agents (44–46). Thus, when combined with other anticancer therapies at earlier stages of cancer, MMP inhibitors may improve the clinical outcome.

In summary, we have shown that CAFs induce cetuximab resistance in HNSCC cells in an MMP-dependent manner. These results suggest that targeting the crosstalk between CAFs and tumor cells may circumvent the resistance to EGFR-targeted therapy in patients with HNSCC.

Studies in the A. Östman laboratory was partially funded by an unrestricted research grant from Merck Serono, Stockholm, Sweden. A. Östman is on the advisory board of Pfizer, ImClone, and Merck Serono. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A.C. Johansson, A. Ansell, F. Jerhammar, R. Grénman, A. Östman, K. Roberg

Development of methodology: M. Bradic Lindh, R. Grénman, K.B. Roberg

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.C. Johansson, A. Ansell, Eva Munck-Wikland, R. Grénman

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.C. Johansson, A. Ansell, M. Bradic Lindh, A. Östman

Writing, review, and/or revision of the manuscript: A.C. Johansson, A. Ansell, F. Jerhammar, Eva Munck-Wikland, R. Grénman, A. Östman, K. Roberg

Study supervision: A. Östman

In loving memory of Dr. Cathrine Nilsson. The authors thank Camilla Janefjord for technical assistance.

A.C. Johansson received financial support from the Johan and Jakob Söderberg Foundation, the Foundation Olle Engqvist Byggmästare, the Swedish Laryng Foundation, Borgholm Rotary Club, the Swedish National Board of Health and Welfare, Lions Research Foundation, the Lars Hierta Memorial Foundation, the Tore Nilsson Foundation for Medical Research, the Swedish Society for medical research, the County Council of Östergötland, and the Cancer Foundation of Östergötland. A. Östman received support from Merck Serono, the Swedish Research Council (349-2008-6578), and the Swedish Cancer Society (CAN 2009/1136). K. Roberg received support from the Swedish Cancer Society (CAN 2010/545).

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.