EGFR Inhibition Induces Proinflammatory Cytokines via NOX4 in HNSCC

Epidermal growth factor receptor (EGFR) is expressed at high levels in head and neck squamous cell carcinomas (HNSCCs) (1); however, EGFR-based chemotherapy has limited results in HNSCC patients (2–6). Numerous studies have proposed mechanisms (i.e., alternate signaling pathways, mutations, EMT) that may be responsible for poor responses and treatment failures because of EGFRI treatment (2). However, these discoveries have not yet led to significant improvements in response rates to EGFRIs in clinical trials. Therefore, the identification and understanding of molecular mechanisms associated with HNSCC tumor response to EGFR inhibition could lead to improvements in drug efficacy and HNSCC patient survival.

In an effort to further investigate the mechanism of action of clinical EGFRIs and possibly identify novel signaling pathways affected by EGFR inhibition, we used gene expression profiling to study the EGFR tyrosine kinase inhibitor (TKI) erlotinib. Erlotinib (marketed as Tarceva) is used to treat non–small cell lung cancer and pancreatic cancer (7). It is also being studied in clinical trials in other cancer disease sites including HNSCC (6, 8, 9). Pathway analysis of the resultant gene expression data (using MetaCore Genego software) indicated a significant upregulation in proinflammatory immune response pathways in 3 HNSCC tumor cell lines treated with erlotinib (unpublished data). These results led to the reasoning that a proinflammatory response may be responsible in part for the reduced efficacy of EGFRIs because of the well-known role of inflammation in tumor progression (10–12). This study determines if clinical EGFRIs induce proinflammatory cytokine expression and if interleukin (IL)-6 signaling in particular reduces the efficacy of erlotinib in HNSCC cells. In addition, because we have previously shown that erlotinib increased oxidative stress in HNSCC cells via NADPH oxidase 4 (NOX4; ref. 13), and proinflammatory pathways may be induced by redox activation, this study will also investigate the role of NOX4 in proinflammatory cytokine expression.

FaDu and Cal-27 human HNSCC cells were obtained from the American Type Culture Collection (ATCC). SQ20B HNSCC cells (14) were a gift from Dr. Anjali Gupta (Department of Radiation Oncology, The University of Iowa, Iowa). FaDu, Cal-27, and SQ20B cells were maintained in Dulbecco's modified Eagle's medium containing 4 mmol/L l-glutamine, 1 mmol/L sodium pyruvate, 1.5 g/L sodium bicarbonate, and 4.5 g/L glucose with 10% FBS (Hyclone). All HNSCC cell lines are EGFR positive and are sensitive to EGFR inhibitors. All cell lines were authenticated by the ATCC for viability (before freezing and after thawing), growth, morphology, and isoenzymology. Cells were stored according to the supplier's instructions and used over a course of no more than 3 months after resuscitation of frozen aliquots. Cultures were maintained in 5% CO₂ and air humidified in a 37°C incubator.

Erlotinib (ERL; Tarceva), cetuximab (CET; Erbitux), lapatinib (LAP; Tykerb), panitumumab (VEC; Vectibix), and tocilizumab (TOC; Actemra) were obtained from the inpatient pharmacy at the University of Iowa Hospitals and Clinics. Human immunoglobulin G (IgG) and dimethyl sulfoxide (DMSO) was obtained from Sigma-Aldrich. Inhibitors of JNK (SP6; SP600125), p38 (SB2; SB203580), and STAT3 (AG4; AG490) were obtained from Calbiochem. The NF-κB inhibitor BAY117092 (BAY) was obtained from Santa Cruz Biotechnology and recombinant human IL-6 was obtained from R&D Systems. Drugs were added to cells at final concentrations of 5 μmol/L ERL, 17 nmol/L CET, 1 μmol/L LAP, 100 nmol/L VEC, 1 μmol/L TOC, 5 μmol/L SP6, 10 μmol/L SB2, 10 μmol/L BAY, and 100 ng/mL IL-6. The required volume of each drug was added directly to complete cell culture media on cells to achieve the indicated final concentrations.

FaDu, Cal-27, and SQ20B HNSCC cells were treated with DMSO or erlotinib (5 μmol/L, 48 hours). Total RNA was extracted from treated cells using the manufacturer's protocol RNeasy mini kit (Qiagen): DNA microarray sample processing. RNA sample preparation for hybridization and the subsequent hybridization to the Illumina beadchips were performed at the University of Iowa DNA Facility using the manufacturer's recommended protocol. Briefly, 100 ng total RNA was converted to amplified biotin-cRNA using the Ambion TotalPrep RNA Amplification Kit for Illumina Expression BeadChip (Ambion, Inc.; Cat. No. AMIL1791) according to the manufacturer's recommended protocol. Seven hundred and fifty nanograms of this product were mixed with Illumina hybridization buffer, placed onto Illumina HumanHT-12v4 BeadChips (Part No. BD-103-0204), and incubated at 58°C for 17 hours, with rocking, in an Illumina Hybridization Oven. Following hybridization, the arrays were washed, blocked, and then stained with streptavidin-Cy3 (Amersham/GE Healthcare) according to the Illumina Whole-Genome Gene Expression Direct Hybridization Assay protocol. Beadchips were scanned with the Illumina iScan System (ID #N054) and data collected using the GenomeStudio software v2011.1.

Tissue culture supernatants from drug-treated and/or adenoviral-transfected cells were collected and assayed for secreted cytokines using the Bio-Plex Pro Human cytokine 8-plex assay (Bio-Rad Laboratories) according to the manufacturer's recommended protocol.

Levels of IL-6 of treated cells were determined by ELISA. The culture media of drug-treated and/or adenoviral-transfected cells were harvested and IL-6 was detected according to the manufacturer's protocol using the Human IL-6 Quantikine ELISA Kit (R&D Systems).

Cell lysates were standardized for protein content, resolved on 4% to 12% SDS polyacrylamide gels, and blotted onto nitrocellulose membranes. Membranes were probed with rabbit anti-STAT3, anti-pSTAT3, anti-β-actin (Cell Signaling), and anti-NOX4 (Abcam) antibodies. Antibody binding was detected by using an ECL Chemiluminescence Kit (Amersham).

Total RNA was extracted from treated cells after indicated time points using the RNeasy mini kit (Qiagen). The cDNA was amplified from 800 ng of total RNA using iScript cDNA synthesis kit (Bio-Rad). Thermocycler conditions included a 5-minute incubation at 25°C, a 30-minute incubation at 42°C, and a 5-minute incubation at 85°C. The cDNAs were subjected to qPCR analysis with the following 5′→3′ primers (sense and antisense, respectively): human IL-6: AGGACTGGAGATGTCTGAGGCTC and GCGCTTGTGGAGAAGGAGTTC, NOX4: CTCAGCGAATCAATCAGCTGTG and AGAGGAACACGACAATCAGCCTTAG, and human 18S: CCTTGGATGTGGTAGCCGTTT and AACTTTCGATGGTAGTCGCCG. The assay was performed in a 96-well optical plate, with a final reaction volume of 20 μL, including synthesized cDNA (20 ng), oligonucleotide primers (100 μmol/L each), and 2× SYBR Green/ROX PCR master mix (Bio-Rad Laboratories). Samples were run on an ABI PRISM Sequence Detection System (model 7000; Applied Biosystems). PCR conditions were 50°C for 2 minutes, 95°C for 2 minutes and 30 seconds, 95°C for 15 seconds, and 60°C for 1 minute for 40 cycles. Results were analyzed using ABI PRISM 7000 SDS software. The denaturation and annealing steps were carried out for 40 cycles to determine the threshold cycle (CT) values for all of the genes analyzed. Samples were checked for nonspecific products or primer/dimer amplification by melting curve analysis. The CT values for the target genes in all of the samples (analyzed in duplicate or triplicate) were normalized on the basis of the abundance of the 18S transcript, and the fold difference (relative abundance) was calculated using the formula 2−ΔΔCT and was plotted as the mean.

Clonogenic survival was determined as previously described (15). Individual assays were performed with multiple dilutions with at least 4 cloning dishes per data point, repeated in at least 3 separate experiments.

Attached cells were labeled with 5-(and-6)-carboxy-2′,7′-dichlorodihydrofluorescein diacetate (DCFH, 10 μg/mL) dissolved in 0.1% DMSO for 15 minutes at 37°C. Culture plates were placed on ice, trypsinized, resuspended in ice cold PBS, and analyzed by using a FACScan flow cytometer (excitation 488 nm, emission 530 nm bandpass filter). The mean fluorescence intensity of 10,000 cells was analyzed in each sample and corrected for autofluorescence from unlabeled cells.

Wild-type human NOX4 cDNA (BC040105.1) was obtained from Open Biosystems (MHS1010-9204787) in the pCMV-SPORT6 vector. cDNA for the dominant-negative form of NOX4 lacking the C-terminal NADPH binding domain was generated by introduction of a stop codon at amino acid 384 as described by Mahadev and colleagues (16) using Stratagene Quickchange XL site-directed mutagenesis with forward CTG GAC AGA ACG ATT TCG AGA TTA ACT ACT GCC TCC and reverse GGA GGC AGT AGT TAA TCT CGA AAT CGT TCT GTC CAG primers (nucleotide change underlined). The entire cDNA for the Wild-type and dominant-negative clones were subsequently sequence verified and subcloned into the pacAd5.CMV.K-NpA shuttle vector (U Iowa Gene Transfer Vector Core, GTVC) using Kpn1/Not1 restriction. Viruses (AdNOX4wt and AdNOX4dn) were purified by the U Iowa GTVC using CsCl gradient centrifugation and dialyzed against buffer containing 3% sucrose/PBS and stored at −80°C. Each virus was tested for wild-type revertants and for titer by PCR and A549 plaque assay as described in Anderson and colleagues (17). An empty vector lacking the NOX4 construct was used as a control. Construction and characterization of recombinant E1-deleted adenoviral vectors encoding siRNA targeted against eGFP (AdsiCON) or Nox4 (AdsiNox4) have each been described previously (18). All vectors were obtained from the University of Iowa Gene Vector Core. HNSCC cells in serum-free media were infected with 150 multiplicity of infection of the above-described adenoviral vectors for 24 hours. Biochemical analyses were performed 72 to 96 hours after transfection.

Female 4- to 5-week-old athymic-nu/nu nude mice were purchased from Harlan Laboratories. Mice were housed in a pathogen-free barrier room in the Animal Care Facility at the University of Iowa and handled using aseptic procedures. All procedures were approved by the IACUC committee of the University of Iowa and conformed to the guidelines established by the NIH. Mice were allowed at least 3 days to acclimate before beginning experimentation, and food and water were made freely available. Tumor cells were inoculated into nude mice by subcutaneous injection of 0.1 mL aliquots of saline containing 4 × 10⁶ FaDu or SQ20B cells into the right flank using 26-gauge needles.

Mice started drug treatment 1 week after tumor inoculation. Mice were evaluated daily and tumor measurements were taken 3 times per week using Vernier calipers. Tumor volumes were calculated using the formula: tumor volume = (length × width²)/2, where the length was the longest dimension and width was the dimension perpendicular to length.

Mice were divided into 4 groups (n = 8–9 mice/group). ERL group: ERL was suspended in water and administered orally 12.5 mg/kg every day for 10 days. TOC group: TOC was administered intraperitoneally (i.p.) 1 mg/kg every other day for 10 days. ERL+TOC group: mice were administered ERL orally 12.5 mg/kg every day and 1 mg/kg TOC i.p. every other day for 10 days. Control group: Mice were administered orally 100 μL water every day and 1 mg/kg IgG i.p. every other day for 10 days. Mice were euthanized via CO₂ gas asphyxiation when tumor diameter exceeded 1.5 cm in any dimension.

Statistical analysis was done using GraphPad Prism version 5 for Windows (GraphPad Software). Differences between 3 or more means were determined by one-way ANOVA with Tukey posttests. Linear mixed effects regression models were used to estimate and compare the group-specific change in tumor growth curves. All statistical analyses were performed at the P < 0.05 level of significance.

The gene expression profiles of FaDu, Cal-27, and SQ20B HNSCC cells exposed to erlotinib (5 μmol/L, 48 hours) versus DMSO were analyzed by high-throughput microarray. Genetic network analysis of the resultant gene expression data for all 3 cell lines (n = 3 experiments per cell line) was carried out using Metacore (GeneGo). Thirty networks were identified using the GeneGo tool (Supplementary Fig. S1) that identified functional relationships between gene products based on known interactions in the scientific literature. Of these networks, we focused on the first scored (by the number of pathways) network with a P-value of 7.3 × 10⁻²¹ and z-score of 9.89 (Supplementary Table S1 and Fig. 1A). The genes in this network were related to positive regulation of immune response processes, response to stimulus, and NF-κB transcription factor activity. In addition, signaling pathways including toll-like receptor, IL-17, and TNF-α pathways were implicated in the activation of NF-κB (Fig. 1A). According to the network shown in Fig. 1A, NF-κB activation resulted in the expression of cytokines involved in proinflammatory pathways such as IL-1β, IL-4, IL-6, IL-12β, CCL20 (MIP3A), granulocyte macrophage colony-stimulating factor (GM-CSF), IP10, and IFN-γ. Of these cytokines, IL-6 seemed to be of importance because the IL-6/JAK/STAT3 pathway was also identified in this network (Fig. 1A). Altogether, these results suggest that the induction of proinflammatory pathways may play a role in the mechanism of action of erlotinib.

To confirm that erlotinib may induce the expression of proinflammatory cytokines, levels of 8 cytokines (IL-2, IL-4, IL-6, IL-8, IL-10, IL-12, IFN-γ, and GM-CSF) were measured using a Human Cytokine 8-Plex Panel from the media of FaDu, Cal-27, and SQ20B cells treated 48 hours with DMSO or 5 μmol/L Erlotinib. Of these 8 cytokines, erlotinib increased levels of IL-6 and GM-CSF from FaDu cells; IL-6, GM-CSF, and IFN-γ from SQ20B cells; and IL-6, IL-8, GM-CSF, IFN-γ, and TNF-α from Cal-27 cells compared with control-treated cells (Fig. 1B), which supports the network analysis shown in Fig. 1. IL-2 and IL-10 were below the limit of detection. Using SQ20B cells, we additionally observed that lapatinib and panitumumab increased levels of IL-4, IL-6, IL-8, GM-CSF, and IFN-γ whereas cetuximab increased only IL-4, IL-6, and IFNγ (Fig. 1C). IL-2, IL-10, and TNF-α were below the limit of detection. These results suggest that all clinical EGFR inhibitors may induce the secretion of proinflammatory cytokines.

Given that the IL-6 signaling pathway was identified in the microarray network analysis as important in the mechanism of action of erlotinib (Fig. 1A), we analyzed the expression of IL-6 mRNA (using reverse transcriptase-PCR [RT-PCR]) and protein (using ELISA) over the course of 48 hours in erlotinib-treated HNSCC cells. IL-6 mRNA expression increased over time in erlotinib-treated Cal-27 and SQ20B cells (Fig. 2A). However, in FaDu cells, IL-6 mRNA levels initially peaked at 30 minutes then peaked again at 16 hours before declining at 24 and 48 hours (Fig. 2A). We observed that erlotinib significantly increased IL-6 protein secretion (compared with control-treated cells) beginning at 6 hours in FaDu cells (Fig. 2B), 8 hours in Cal-27 cells (Fig. 2C), and 24 hours in SQ20B cells (Fig. 2D) and increased gradually more than a 24-hour period. Erlotinib induced IL-6 secretion in other EGFR-positive cell lines such as A549 (lung cancer) and MiaPaca2 (pancreatic cancer) cells (Supplementary Fig. S1), suggesting that erlotinib-induced IL-6 is not specific to HNSCC cell lines. Knockdown of EGFR using siRNA induced IL-6 in a similar manner to erlotinib and other EGFRIs in FaDu cells (Supplementary Fig. S2), suggesting that the effects we see with erlotinib are likely because of EGFR inhibition. However, in siEGFR-transfected FaDu cells those were treated with erlotinib, we observed an additional increase in IL-6 secretion compared with siEGFR-transfected FaDu cells treated with DMSO, suggesting that some off-target effects of erlotinib may contribute to IL-6 production in this cell line (Supplementary Fig. S2). In addition, erlotinib increased STAT3 tyrosine phosphorylation in FaDu, Cal-27, and SQ20B cells after 48 hours as observed by increased pSTAT3β (79 Kd) isoform expression but not pSTAT3α isoform expression (86 Kd). In FaDu cells, erlotinib increased both pSTAT3 isoforms. Total STAT3 expression was not affected by erlotinib in all cell lines (Fig. 2E). These results suggest that EGFR pathway inhibition with erlotinib induces IL-6 mRNA and protein expression and possibly downstream IL-6 signaling in HNSCC cells and this phenomenon may be observed in other EGFR-positive tumor cell models.

To confirm that erlotinib-induced IL-6 was a result of NF-κB activation, we pretreated FaDu cells with the NF-κB inhibitor BAY117082 before treatment with erlotinib then analyzed IL-6 expression by ELISA. We observed that NF-κB inhibition completely reversed erlotinib-induced IL-6 expression confirming that NF-κB activation was involved in erlotinib-induced IL-6 activation (Fig. 2F). Given that both NF-κB and AP-1 binding sites are present on the promoter region of IL-6 (19) and MAPK signaling pathways have been shown to affect AP-1 activity (20), we determined if MAPK activation was also involved in erlotinib-induced IL-6 expression. We observed that pretreatment of cells with inhibitors of p38 (SB203580) and JNK (SP600125) partially suppressed erlotinib-induced IL-6 expression, suggesting that MAPK activation and thus AP-1 activation was also involved in IL-6 expression (Fig. 2F).

To determine the significance of erlotinib-induced IL-6 expression, we pretreated FaDu cells with 100 ng/mL of human recombinant IL-6 for 2 hours before treatment with erlotinib in serum-free media then analyzed for cell survival (via clonogenic assay). We found that exogenous IL-6 was able to completely protect against erlotinib-induced cytotoxicity (Fig. 3A). Conversely, to determine if inhibition of the IL-6 pathway enhances HNSCC tumor cell response to erlotinib, we treated FaDu cells with the IL-6 receptor antagonist tocilizumab with or without erlotinib in vitro. We observed that tocilizumab was able to sensitize FaDu cells to erlotinib in vitro (Fig. 3B). The combination of erlotinib and tocilizumab was tested in FaDu and SQ20B xenografts (Fig. 3C and D). The results indicated that tocilizumab was able to increase the efficacy of erlotinib after 10 days of treatment in both xenograft models (Fig. 3C and D). These results suggest that IL-6 pathway inhibition may be a viable strategy to enhance HNSCC tumor response to EGFRIs.

Oxidative stress has been implicated in inflammatory responses through NF-κB activation in various chronic diseases (21, 22). Our previous work has shown that erlotinib induced metabolic oxidative stress in HNSCC cells via NOX4 (13). To determine if NOX4 contributed to the increase in proinflammatory cytokines induced by erlotinib, we knocked down NOX4 using adenoviral siRNA targeted to NOX4 before treatment with erlotinib. Erlotinib increased NOX4 mRNA (Fig. 4A and B) and protein expression (Fig. 4A) in all 3 HNSCC cell lines, and transfection with AdsiNOX4 was effective at suppressing the NOX4 mRNA (Fig. 4B), IL-6 mRNA (Fig. 4C), and IL-6 protein (Fig. 4D). In addition, knockdown of NOX4 expression was able to suppress erlotinib-induced expression of IL-2, IL-4, IFN-γ, and TNF-α in Cal-27 cells and IFN-γ in SQ20B cells (Table 1). Even though AdsiNOX4 was able to consistently decrease erlotinib-induced NOX4 mRNA expression, we were unable to detect similar changes in NOX4 protein expression. We did observe a partial decrease in erlotinib-induced NOX4 protein expression with AdsiNOX4 in FaDu cells; however, we were unable to detect decreased NOX4 protein expression in Cal-27 and SQ20B cells (data not shown) even though the effect of AdsiNOX4 was clearly observed by mRNA levels (Fig. 4A) in these cell lines. This suggests that NOX4 may have a long half-life in our HNSCC cells and may be resistant to proteolytic degradation as observed in THP-1 monocytes (23) and human monocyte-derived macrophages (24). Altogether, these results suggest that NOX4-mediated oxidative stress may be involved in the increase in proinflammatory cytokines induced by EGFR inhibitors.

To confirm that NOX4 was involved in reactive oxygen species (ROS) and cytokine production, we transfected HNSCC cells with adenoviral vectors containing wild-type NOX4 (AdNOX4wt), dominant negative NOX4 (AdNOX4dn), and empty (AdEMPTY) before measurement of DCFH oxidation and cytokine production. Overexpression of NOX4 resulted in an increase in DCFH oxidation in all 3 cell lines (Fig. 5A) and also increased IL-6, IL-8, and GM-CSF in FaDu cells; IL-2, IL-6, GM-CSF, and TNF-α in Cal-27 cells; and IL-6 and IFN-γ in SQ20B cells (Table 2), suggesting that NOX4-induced ROS was capable of inducing proinflammatory cytokine expression. However, overexpression of NOX4dn either did not affect or suppressed ROS (Fig. 5A) and proinflammatory cytokine levels (Table 2).

To investigate how NOX4 expression affects cell survival and HNSCC cell sensitivity to erlotinib, FaDu, Cal-27, and SQ20B cells were transfected with AdEMPTY, AdNOX4wt, and AdNOX4dn as mentioned above before treatment with DMSO and erlotinib then analyzed by clonogenic assay. We observed that overexpression of NOX4wt significantly increased cytotoxicity of all 3 cell lines and increased the sensitivity of FaDu and SQ20B cells to erlotinib (Fig. 5B). Overexpression of NOX4dn increased cell survival in FaDu and Cal-27 cells and significantly protected all 3 cell lines from erlotinib-induced cytotoxicity (Fig. 5B). These results suggest that NOX4 expression may affect HNSCC response to erlotinib.

These studies presented here indicate that EGFRIs increase the expression of proinflammatory mediators through a mechanism involving NOX4-mediated oxidative stress. Chronic inflammation may play a significant role in tumor promotion, migration, and invasion in some cancers (25), and many of the cytokines found to be increased by EGFRIs (e.g., IL-2, IL-4, IL-6, IL-8, GM-CSF, IFN, and TNF [Figs. 1 and 2]) are involved in immune/inflammatory responses, angiogenesis, metastasis, and tumor progression (25). This implies that the pressure of EGFRI treatment itself may drive the reduction in drug efficacy by inducing oxidative stress and inflammation in the tumor microenvironment.

The observation that EGFRIs induce immune and inflammatory responses is not unheard of because EGFRIs are known to promote skin inflammation in a majority of patients (26). What is less known, however, is how these inflammatory pathways affect the antitumor activity of EGFRIs. For example, we show that exogenous IL-6 was able to completely protect FaDu cells from the antitumor effect of erlotinib (Fig. 3A), which supports prior work demonstrating the ability of IL-6 to protect lung cancer cells against erlotinib in lung cancer cells (27). IL-6 is a pleotropic cytokine with a wide range of biological activities and is well known for its role in inflammation, tumor progression, and chemoresistance in HNSCC (28–33). Increases in IL-6 expression correlate with poor prognosis in HNSCC patients and patients resistant to chemotherapy have shown significantly higher serum IL-6 levels than those who did respond (28, 29, 31).

Crosstalk exists between the EGFR and IL-6 signaling pathways because both pathways are able to phosphorylate STAT3 (33). STAT3 activation has been shown to mediate cancer cell proliferation and carcinogenesis (34). This crosstalk is initiated by the secretion of IL-6, resulting from NF-κB activation, and the subsequent phosphorylation of STAT3 leading to STAT3-dependent gene expression (33). EGFR activation also leads to phosphorylation of STAT3 (35). Therefore, in the presence of high levels of IL-6, STAT3 expression is sustained in the presence of EGFRIs and the antitumor effect of EGFRIs may be reduced. In our studies, the ability of erlotinib to not change or surprisingly increase STAT3 tyrosine phosphorylation (Fig. 2E) support the idea of crosstalk between the IL-6 and EGFR signaling systems. Given the crosstalk between the EGFR and the IL-6 pathways, IL-6 pathway inhibitors should be considered a promising strategy to enhance tumor response to EGFRIs. The IL-6 pathway inhibitor tocilizumab is already approved for use in rheumatoid arthritis and systemic juvenile idiopathic arthritis as an anti-inflammatory agent in combination with methotrexate (36). Tocilizumab is a humanized monoclonal antibody against the interleukin-6 receptor (IL-6R) that blocks IL-6 from binding to its receptor (36–38). Tocilizumab has shown antitumor and antiangiogenic activity as a single agent in glioma cells and oral squamous carcinoma cells in vitro and in vivo (37, 38). We have now shown that tocilizumab was able to effectively sensitize FaDu and SQ20B HNSCC xenografts to erlotinib in vivo (Fig. 3C and D), suggesting that the combination of EGFRIs and IL-6 pathway inhibitors may be a viable strategy for the treatment of EGFR-positive tumors.

The second major finding of these studies is the role that NOX4 plays in inflammation in HNSCC cells. NOX enzymes are a family of transmembrane enzymes (NOX1-5, DUOX1, and DUOX2) that produce ROS in response to stimuli including growth factors (39). We have previously shown that erlotinib induces oxidative stress via NOX4 in FaDu HNSCC cells and that NOX4 expression was necessary for the antitumor activity of erlotinib (13). In support of this, we have now duplicated these results in other HNSCC cell lines (Fig. 4A) and additionally have shown that overexpression of NOX4 increased the antitumor effect of erlotinib in all 3 HNSCC cell lines (Fig. 5B). Furthermore, overexpression of a dominant negative form of NOX4 (NOX4dn), lacking the NADPH binding domain and thus ROS producing capacity (Fig. 5A), significantly protected cells from erlotinib-induced toxicity (Fig. 5B) suggesting that NOX4 plays a vital role in erlotinib-induced cytotoxicity.

Given that NF-κB is redox regulated (40), we proposed that NOX4-mediated oxidative stress may activate NF-κB leading to cytokine expression. Suppression of NOX4 expression was indeed effective at suppressing erlotinib-induced IL-6 expression in all 3 cell lines (Fig. 4C and D) suggesting that suppression of NOX4 would also suppress NF-κB activation. These studies are currently underway. Interestingly decreased NOX4 expression also suppressed many of the other proinflammatory cytokines induced by erlotinib in Cal-27 and SQ20B cells, but this effect was cell line specific (Table 1). In addition, in the absence of erlotinib, NOX4 overexpression in itself was able to increase the expression of proinflammatory cytokines (Table 2), further supporting the role of NOX4 in inflammation. Overall, these results suggest a dual role of NOX4 in the mechanism of action of erlotinib and possibly other EGFRIs. NOX4-induced oxidative stress has already been established by our laboratory as being important for the antitumor activity of erlotinib (Fig. 5B; ref. 13). However, NOX4 seems to also induce a cytoprotective mechanism via the induction of proinflammatory cytokines such as IL-6, which may reduce the antitumor effect of erlotinib (Fig. 5C).

Overall these studies will shed some light on the mechanism of action of EGFRIs and how we can improve targeted therapy for HNSCC patients. Cisplatin and radiation have long been the backbone and standard of care of HNSCC patients presenting with locally advanced disease. Cisplatin-based chemotherapy and radiation have already been shown to induce IL-6 secretion (41, 42), implying that EGFRI-induced IL-6 may not be unique to EGFR inhibitors. It is possible that the cell death caused by standard chemo-/radiotherapy in HNSCC may be activating pathways leading to IL-6 and other proinflammatory cytokines. Nevertheless, the use of IL-6 inhibitors may be an important adjuvant for not only EGFRIs but for other chemo-/radiotherapy regimens used for HNSCC treatment.

No potential conflicts of interest were disclosed.

Conception and design: E.V.M. Fletcher, A. Goel, A.L. Simons

Development of methodology: E.V.M. Fletcher, L. Love-Homan, A.L. Simons

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E.V.M. Fletcher, L. Love-Homan, A. Sobhakumari, C.R. Feddersen, A.T. Koch, A.L. Simons

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E.V.M. Fletcher, L. Love-Homan, A. Sobhakumari, A. Goel, A.L. Simons

Writing, review, and/or revision of the manuscript: E.V.M. Fletcher, L. Love-Homan, A. Goel, A.L. Simons

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L. Love-Homan

Study supervision: A. Goel, A.L. Simons

The authors thank Dr. A. Shepard for providing the NOX4wt and NOX4dn adenoviral vectors, and Dr. Robin Davisson for the siNOX4 adenoviral vector. The authors thank Dr. T. Bair in the Bioinformatics Division at the University of Iowa for his assistance in running the microarray studies. The authors also acknowledge the Flow Cytometry Facility, which is a Carver College of Medicine Core Research Facilities/Holden Comprehensive Cancer Center Core Laboratory at the University of Iowa. The Facility is funded through user fees and the generous financial support of the Carver College of Medicine, Holden Comprehensive Cancer Center, and Iowa City Veteran's Administration Medical Center.

This work was supported by the Department of Pathology and grants NIH K01CA134941 (A.L. Simons), R01CA127958 (A. Goel), and IRG-77-004-34 (A.L. Simons) from the American Cancer Society, administered through the Holden Comprehensive Cancer Center at the University of Iowa.

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.