Inhibition of TGF-β Enhances the In Vivo Antitumor Efficacy of EGF Receptor–Targeted Therapy

This article is featured in Highlights of This Issue, p. 2309

Concurrent chemoradiation for locally advanced head and neck squamous cell carcinoma (HNSCC) is limited by its toxicity and the development of recurrent disease in 30% to 40% of patients (1, 2). Efforts to improve the treatment of HNSCC have targeted the EGF receptor (EGFR), a receptor tyrosine kinase that is overexpressed and aberrantly activated in almost all such neoplasms (3–5). Activation of EGFR signaling promotes tumor cell proliferation and survival, and facilitates tumor angiogenesis (6, 7). Strategies to target EGFR have focused on either EGFR tyrosine kinase inhibitors (TKI) or monoclonal antibodies (mAb) that specifically bind the extracellular domain of the receptor, such as the human–mouse chimeric IgG1 mAb, cetuximab (8, 9). The direct effect of EGFR-targeted mAbs on tumor cells involves specific blockade of EGFR signaling via interference with binding of EGFR ligands to the extracellular domain of the receptor (10–12). In addition, the interaction of the Fc region of an antibody to Fcγ receptors on immune effector cells also induces antibody-dependent cellular cytotoxicity (ADCC; refs. 12–16).

Treatment of patients with locoregionally advanced HNSCC with a combination of cetuximab and radiation improved overall survival compared with radiation alone (17). With a median follow-up of 54.0 months, the median duration of overall survival was 49.0 months among patients treated with combined therapy and 29.3 months among those treated with radiotherapy alone. However, the survival benefit from cetuximab was not uniformly observed across all patients. The beneficial effect of cetuximab seemed to be preferentially evident in a subset of patients with the typical characteristics of human papillomavirus (HPV)-positive head and neck cancer (those with oropharyngeal cancer who were males and less than 65 years). After cetuximab and radiation therapy, patients with HPV-positive tumors showed a 60% 2-year progression-free survival (PFS) compared with only 23% PFS for patients with HPV-negative tumors. Identification of the molecular determinants of resistance to EGFR-targeted mAbs is crucial for improving their clinical benefit against HNSCC.

In this study, we find that patients with HPV-negative HNSCC exhibit an abnormal elevation of serum levels of TGF-β, a multifunctional cytokine that regulates cell growth and differentiation (18, 19). We show that TGF-β exerts an extrinsic inhibition of the cytotoxic function of immune effectors while simultaneously providing an intrinsic EGFR-independent survival signal that protects tumor cells from immune cell–mediated ADCC. Although the autonomous expression of TGF-β enables tumor cells to evade ADCC and resist the antitumor activity of cetuximab in vivo, combinatorial treatment with cetuximab and a TGF-β–blocking antibody prevents the emergence of such resistant tumor cells and improves the regression of HNSCC tumor xenografts. These results show that tumor cell expression of TGF-β is a key determinant of resistance to EGFR-targeted antibodies, and provide a rationale for combinatorial targeting of TGF-β to improve the response of HNSCC and other EGFR-expressing cancers to cetuximab-based therapy.

Plasma samples were obtained from 47 patients with HNSCC and 10 patients with pleomorphic adenoma (as non-HNSCC controls) who consented for the Institutional Research Board-approved tissue banking protocol at Johns Hopkins University (Baltimore, MD). The serum levels of TGF-β1, TGF-β2, TGF-β3, IL-10, IL-6, VEGF, IFN-α, IFN-γ, IL-2, IL-12p70, IL-12p40/p70, IL-15, TNF-α, eotaxin, osteopontin, GCSF, SDF-1α, IL-4, GRO-α, and IL-8 were measured by multiplex bead assays using reagents from Millipore and the Luminex 100 system in accordance with the manufacturer recommended protocols (20). The HPV status was determined by in situ hybridization using high-risk HPV-specific probes and p16 immunohistochemical staining as described (21).

The human HNSCC cell line UM-SCC-1 (designated SCC-1 or SCC1) and its isogenic derivative 1CC8 (designated SCC-1CC8 or 1CC8) were a kind gift from the University of Michigan and maintained in 10% FBS in Dulbecco's Modified Eagle's Medium (DMEM) supplemented with 1% hydrocortisone. The erlotinib-resistant isogenic derivative of SCC-1 cells (designated SCC-1T) was generated via continuous exposure of SCC-1 cells to escalating concentrations of erlotinib over 6 months (22), and maintained in culture medium supplemented with erlotinib. The human HNSCC cell lines (HACAT, HO1u1, 93VU147T, CAL27, SKN3, SCC25, HN5, SCC47, UNC7, SQ20B, HSC3, SQ9G, SCC6, HMS001, SCC61, FADU, SCC9, JHU22, SCC090, HSC2, JSQ3, SCC15, UNC10) and H358 lung carcinoma cells were cultured in their respective recommended media. Cells were maintained in a humidified atmosphere containing 5% CO₂ at 37°C. Each cell line was authenticated using fingerprinting at the Johns Hopkins Genetic Resources Core Facility, and used within 6 months of authentication.

Tumor xenografts were generated by implanting 3 × 10⁶ tumor cells into the right flank of 5- to 6-week-old female athymic mice (nu/nu; National Cancer Institute, Bethesda, MD). Once the tumors reached a size of approximately 100 mm³, the mice were randomized (8–10 mice per group), and treated with cetuximab [Imclone; 5 mg/kg intraperitoneally (i.p.) twice weekly for 4 weeks) and/or TGF-β antibody (Bioxcell; 5mg/kg i.p., once weekly for 4 weeks). Tumor size was measured weekly and tumor volume was calculated using the formula (length × width × height). The animals were maintained in accordance with guidelines of the American Association of Laboratory Animal Care and a research protocol approved by the Johns Hopkins University Animal Use and Care Committee.

The night before imaging, mice were fasted for 12 hours. Water was provided ad libitum. On the day of imaging, each mouse was injected with 250 μCi of [18^F]FDG via the tail vein and imaged 45 minutes post injection, using the Mosaic HP (Philips) small animal PET imagers with 15-minute static acquisition. A CT scan was done at the same time using the CT component NanoSPECT/CT (Bioscan) in vivo animal imagers. The standard uptake values were computed by normalizing the PET activity for each mouse to the injected dose and animal weight and coregistered with CT images using Amira 5.2.2 (Visage Imaging, Inc.).

Serum was collected from mice by tail bleeding for measurement of TGF-β using ELISA (R&D Systems). Tumor cells were extracted from xenografts using collagenase digestion followed by RBC lysis, and cultured for 48 hours in DMEM containing 0.1% FBS. Tumor cell supernatants were evaluated by ELISA to determine the amount of TGF-β expressed by 1 × 10⁶ cells per 24 hours.

Human peripheral blood mononuclear cells (PBMC) from normal donors were stimulated with recombinant human interleukin-2 (rh IL-2; 200 IU/mL; Chiron) in the presence or absence of rhTGFβ1 (5 ng/mL; R&D Systems) in Adoptive Immunotherapy Media V (AIM V) containing low Ig FBS (Invitrogen) for 48 hours and used as effectors. Tumor cells (5,000 cells/well) were cultured in a 96-well U-bottomed plate in growth medium containing low Ig serum for 24 hours before being used as targets. Target cells were washed and coated with cetuximab 2 μg/mL and incubated for an hour in a CO₂ incubator. PBMCs were washed, counted, and plated on target cells in a range of effector:target (E:T) ratios. Plates were incubated for 3 to 4 hours and analyzed for ADCC using an aCella-Tox kit (Cell Technology Inc.).

PBMCs were stimulated as described earlier (ADCC assay) for 48 hours and RNA was extracted using TRIzol (Invitrogen) followed by RNAeasy kit cleanup (Qiagen). RNA was reverse transcribed to cDNA using Superscript III (Invitrogen) that then used as a template for real-time PCR. PCR primers were designed to amplify cDNA fragments approximately 150 to 250 bp in length using primer3 software. Gene products were amplified using iTaq SYBR green Supermix with Rox dye (Bio-Rad Laboratories) with the following amplification program for 40 cycles: 95°C for 15 minutes, 60°C for 1 minute. All reactions were conducted in triplicate, with water controls, and relative quantity was calculated after normalizing to GAPDH expression. The following primer sequences were used: (i) NKG2D: 5′-GGCTCCATTCTCTCACCCA-3′ (forward) and 5′-TAAAGCTCGAGGCATAGAGTGC-3′ (reverse); (ii) granzyme A: 5′-TCCTATAGATTTCTGGCATCCTCTC-3′ (forward) and 5′-TTCCTCCAATAATTTTTTCACAGACA-3′ (reverse); (iii) granzyme B: 5′-TCCTAAGAACTTCTCCAACGACATC-3′ (forward) and 5′-GCACAGCTCTGGTCCGCT-3′ (reverse); (iv) Apo2L/TRAIL: 5′-CCCCTGCTGGCAAGTCAA-3′ (forward) and 5′-TGAACTGTAGAAATGGTTTCCTCAGA-3′ (reverse); (v) IFN-γ: 5′-GAAAAGCTGACTAATTATTCGGTAACTG-3′ (forward) and 5′-GTTCAGCCATCACTTGGATGAG-3′ (reverse); and (vi) GAPDH: 5′-CAACTACATGGTTTACATGTTC-3′ (forward) and 5′-GCCAGTGGACTCCACGAC (reverse). Expression of each specific mRNA (gene) relative to GAPDH was calculated based on the threshold cycle (C_t) as 2^−Δ(ΔCt), where |$\Delta (\Delta C_{\rm t}) = \Delta C_{{\rm t}_{{\rm gene}}} - \Delta C_{{\rm t}_{{\rm GAPDH}}}$|⁠.

PBMCs were stimulated as described earlier (ADCC assay) for 3 days and stained extracellularly with anti-CD56 APC (BD Biosciences), anti-Apo2L/TRAIL PE, or anti-FAS-L PE (eBioscience) for 10 to 15 minutes at 4°C. The cells were washed twice and then run on the FACS Calibur (BD Biosciences) and evaluated with Cell Quest software.

Tumor cells were plated at a density of 3,000 per well in 96-well plates. The following day, cells were treated with 0 to 10 μmol/L erlotinib for an additional 72 hours. Cell viability was subsequently assayed using the MTT colorimetric assay. The plates were read using a SpectraMax plate reader (Molecular Devices Corp.) at 570 nm with a reference wavelength of 650 nm. A minimum of 6 wells were tested for each erlotinib dose. In some experiments, tumor cells were pretreated with either rhTGFβ1 or rhTGFβ3 (5 ng/mL; R&D Systems) to evaluate the effect of TGF-β on their sensitivity to erlotinib. The effect of TGF-β on the response of SCC47 cells to erlotinib was evaluated after transfection with either 100 nmol/L AKT siRNA or Control siRNA (Cell Signaling Technology) for 48 to 72 hours.

Xenograft samples were homogenized in radioimmunoprecipitation assay (RIPA) buffer followed by sonication. In some experiments, cells were plated at a density of 500 per well in 6-well plates and cultured with either rhTGFβ1 (5 ng/mL; R&D Systems) or TGFβRII-Fc (500 ng/mL; refs. 23, 24) for 1 to 7 days before immunoblot analysis. Cells were lysed in RIPA lysis buffer containing protease inhibitors (Roche Diagnostic Systems) and phosphatase inhibitor cocktail (Sigma-Aldrich). Lysate protein concentrations were determined by the Lowry protein assay (Bio-Rad Laboratories). Equal amounts of protein were mixed with Laemmli sample buffer (62.5 mmol/L Tris-HCl, pH 6.8, 2% SDS, 10% glycerol, 0.1 mol/L DTT, and 0.01% bromophenol blue), run on 4% to 12% NuPAGE gels and transferred to nitrocellulose membrane (Bio-Rad Laboratories). The membrane was blocked with PBS supplemented with 0.1% Tween 20 and 5% nonfat milk for 1 hour at room temperature, and probed with primary antibody for 1 hour at room temperature or overnight at 4°C followed by HRP-conjugated appropriate secondary antibodies (Santa Cruz Biotechnology). The following primary antibodies were used: antibodies against EGFR, phospho-EGFR, AKT, phospho-AKT (Ser473), E-cadherin, vimentin (Cell Signaling Technology), and anti–β-actin antibody (Sigma-Aldrich). Signals from immunoreactive bands were detected by enhanced chemiluminescence (GE Healthcare).

Differentially expressed cytokines were determined comparing non-HNSCC controls versus HNSCC, HPV-positive versus HPV-negative, and collection at the time of new diagnosis versus recurrence diagnosis. Student unpaired t tests were used for statistical analysis to compare each of the 2 groups. When comparing across cytokines, results with Benjamini–Hotchberg adjusted P values (25) below 0.05 were considered significant. When making comparisons between classes for single cytokines, results with unadjusted P < 0.05 were considered significant. All statistical analyses of cytokines were conducted using R according to the script in Supplementary Information (Supplementary Methods). Student unpaired t test was used for all other statistical analyses between 2 groups. Results with P < 0.05 were considered significant. Statistical analyses were conducted using GraphPad InStat version 3.0a for Macintosh (GraphPad Software).

Plasma samples were procured from 47 patients with HNSCC (Supplementary Table S1) and 10 patients with pleomorphic adenoma (as benign non-HNSCC controls). Thirty-six samples were obtained at the time of new diagnosis of HNSCC and 11 samples were obtained at the time of recurrence. Measurement of serum levels of 20 cytokines showed that TGF-β1 was significantly elevated in patients with HNSCC compared with non-HNSCC controls (P = 0.0085; Fig. 1A and Supplementary Fig. S1). A striking variation in serum levels of TGF-β1 was also noted among patients with HNSCC (Fig. 1A and Supplementary Table S2). Univariate analysis of samples from newly diagnosed patients showed that TGF-β1 levels were significantly higher in HPV-negative HNSCC compared with non-HNSCC controls (P < 0.0003; Fig. 1B). Whereas patients with HPV-negative tumors showed an abnormal elevation of serum TGF-β1, patients with HPV-positive tumors exhibited relatively lower levels of TGF-β1 that were not significantly higher than non-HNSCC controls (Fig. 1B).

To determine whether HNSCC cells express and secrete TGF-β into the tumor microenvironment, the amount of TGF-β produced by 25 different human HNSCC cell lines in supernatants of tumor cells was measured by ELISA. Akin to the variable serum levels of TGF-β observed in patients with HNSCC, HNSCC cell lines produced differing amounts of TGF-β in vitro (Fig. 2A). Significantly, isogenic variants of the same cell line (SCC-1) expressed strikingly different levels of TGF-β. Although parental SCC-1 cells expressed low levels of TGF-β (37.8 pg/10⁶ cells/24 h), derivatives of SCC-1 cells selected for resistance to cetuximab (SCC-1CC8) or the EGFR TKI, erlotinib (SCC-1T), expressed significantly higher levels of TGF-β [SCC-1CC8, 124.5 pg/10⁶ cells/24 h (P = 0.004); SCC-1T, 550 pg/10⁶ cells/24 h (P < 0.001); Fig. 2A]. To determine whether tumor cell production of TGF-β influences the ability of immune effector cells to induce cetuximab-mediated ADCC, cetuximab-coated tumor cells were exposed to normal human PBMCs that were pretreated with rhIL-2 in the presence or absence of rhTGF-β1 for 48 hours. HNSCC cell lines exhibited variable susceptibility to cetuximab-mediated ADCC (Fig. 2B). Exposure of PBMCs to rhTGF-β1 resulted in a marked decline in their ability to induce cetuximab-mediated ADCC of all HNSCC cell lines, including HPV-positive cell lines (SCC47, SCC90) and HPV-negative cell lines (SCC1, 1CC8, SCC-1T, SCC6, UNC10; Fig. 2B). Exposing PBMCs to rhTGF-β resulted in a significant diminution in their expression of several cytotoxic effector molecules, including granzyme B, Apo2L/TRAIL, CD95L/FasL, and IFN-γ (Fig. 2C and D). Real-time PCR showed that exposure to rhTGF-β resulted in approximately 6-fold reduction in the mRNA levels of granzyme B and Apo2L/TRAIL, as well as approximately 4-fold decrease in levels of IFN-γ mRNA (Fig. 2C). Similarly, flow cytometry analysis showed a significant reduction of surface expression of Apo2L/TRAIL and CD95L/FasL on CD56⁺ NK cells after exposure to rhTGF-β (Fig. 2D). These data indicate that the autonomous production of TGF-β by tumor cells can exert a tumor cell-extrinsic inhibition of the cytotoxic function of immune effectors that induce cetuximab-mediated ADCC.

The SCC-1T cell line was derived by culturing SCC-1 cells in the presence of escalating concentrations of the EGFR TKI, erlotinib. Although SCC-1 cells exhibited a dose-dependent reduction of viability in response to erlotinib, SCC-1T cells were relatively resistant to treatment (Fig. 3A). SCC-1T cells expressed more than 10-fold higher levels of TGF-β compared with SCC-1 cells (Fig. 3B). The elevation in autonomous expression of TGF-β in SCC-1T cells was associated with increased phosphorylation of AKT and epithelial–mesenchymal transition (EMT), as showed by the loss of expression of E-cadherin and appearance of vimentin (Fig. 3C). Because tumor cell acquisition of an erlotinib-resistant phenotype was coselected with an increase in autonomous expression of TGF-β, we investigated whether exposure to TGF-β directly enables EGFR-independent survival of tumor cells. Immunoblot analyses showed that treatment of EGFR-expressing H358 tumor cells with either rhTGF-β1 or rhTGF-β3 resulted in increased expression of phosphorylated AKT (p-AKT) that was apparent after 7 days of exposure (Fig. 3D). The constitutive activation of AKT was associated with TGF-β—induced EMT, as evidenced by the appearance of vimentin and concurrent reduction in the expression of E-cadherin (Fig. 3D). The acquisition of an EMT phenotype by tumor cells after exposure to TGF-β was attended with a concurrent decrease in the expression of phosphorylated EGFR (p-EGFR; Fig. 3D). These effects of TGF-β were also confirmed in 3 HNSCC cell lines [SCC6 (HPV-negative), SCC47 (HPV-positive), and UNC10 (HPV-negative)]. Exposure of HNSCC cell lines to TGF-β resulted in increased expression of p-AKT, EMT, and decreased p-EGFR expression (Fig. 3E).

To determine whether the TGF-β–induced switch of tumor cells from an active EGFR/low p-AKT phenotype to an inactive EGFR/high p-AKT phenotype renders them independent of EGFR-activated survival signals, HNSCC cell lines (SCC6, SCC47, and UNC10) or H358 cells that were pretreated with either TGF-β1 or TGF-β3 and their untreated counterparts (control) were exposed to graded concentrations of erlotinib (26). Although SCC6, SCC47, and H358 cells exhibited a dose-dependent reduction of viability after treatment with erlotinib, pretreatment with TGF-β rendered all 3 lines relatively resistant to erlotinib (Fig. 3F). We further established the importance of AKT activation in TGF-β–mediated resistance to EGFR-targeted therapy using siRNA-mediated knockdown of AKT in SCC47 cells (Fig. 3F and G). Treatment of SCC47 cells with siRNA against AKT inhibited TGF-β–induced elevation of p-AKT (Fig. 3G). siRNA-mediated loss of AKT sensitized SCC47 cells to erlotinib and counteracted the ability of TGF-β to confer resistance to erlotinib (Fig. 3F).

To investigate whether autonomous expression of TGF-β modulates the response of tumor cells to EGFR-targeted mAbs in vivo, we examined the effect of cetuximab on tumor xenografts comprising either SCC-1 cells or its derivative, SCC-1CC8 (1CC8). Although the in vivo growth of SCC-1 tumors was arrested by cetuximab (untreated vs. treated, P < 0.05), 1CC8 tumors were relatively less responsive to the same treatment (untreated vs. treated, P = 0.09; Fig. 4A). However, cetuximab alone was not sufficient to induce regression of either SCC-1 or 1CC8 tumor xenografts. Because TGF-β inhibited cetuximab-mediated ADCC of tumor cells in vitro, we investigated whether treatment of tumor-bearing mice with cetuximab results in the in vivo selection of resistant tumor cells that express higher levels of TGF-β. Tumor cells extracted from residual tumors in untreated or cetuximab-treated mice were cultured ex vivo, and the production of TGF-β in tumor cell supernatants was measured by ELISA. Explanted SCC-1 tumor cells derived from residual nonregressed tumors after cetuximab treatment expressed significantly higher levels of TGF-β (135 pg/10⁶ cells/24 h) compared with either SCC-1 cells derived from untreated tumor-bearing animals (40 pg/10⁶ cells/24 h; P = 0.01) or SCC-1 cells before inoculation into mice (37.8 pg/10⁶ cells/24 h; P = 0.01; Fig. 4B). Indeed, the elevated level of expression of TGF-β in residual SCC-1 tumor cells derived from cetuximab-treated mice approximated that of residual 1CC8 cells after cetuximab therapy (P = 0.6; Fig. 4B). Consistent with the increase in the autonomous expression of TGF-β, tumor-derived SCC-1 cells from cetuximab-treated animals exhibited increased phosphorylation of AKT and reduced expression of E-cadherin, a marker of EMT (Fig. 4C). Treatment of residual SCC1 or 1CC8 tumor cells with a TGF-β antagonist (TGFβRII-Fc) reversed the phosphorylation of AKT, indicating that tumor cell–derived TGF-β was responsible for EGFR-independent activation of AKT (Fig. 4D). In accordance with the in vivo selection of TGF-β–overexpressing resistant tumor cells in cetuximab-treated animals, serum levels of TGF-β were equally elevated in SCC-1- and 1CC8–tumor-bearing mice after cetuximab therapy, but were restored to that of naïve mice after treatment with an antagonist TGF-β antibody (TGF-β antibody; Fig. 4E).

Because treatment with cetuximab resulted in the selection of TGF-β–overexpressing resistant tumor cells, we investigated whether inhibition of TGF-β with TGF-β antibody can improve the in vivo antitumor efficacy of cetuximab against tumor xenografts of HNSCC cells. Consistent with the selection of TGF-β–overexpressing variants of SCC-1 tumor cells in cetuximab-treated animals, cetuximab alone was not sufficient to induce regression of either SCC-1 or 1CC8 tumor xenografts (Fig. 5A and B). However, combinatorial treatment of mice with TGF-β antibody improved the antitumor efficacy of cetuximab against tumor xenografts of SCC-1 cells as well as 1CC8 cells (Fig. 5A and B). Unlike cetuximab alone, combined treatment with cetuximab and TGF-β antibody resulted in regression of both SCC-1 and 1CC8 tumor xenografts (Fig. 5A and B). In animals bearing 1CC8 tumor xenografts, treatment for 30 days with the combination of cetuximab and TGF-β antibody resulted in significantly smaller tumors (mean ± SEM = 15 ± 6 mm³) compared with cetuximab alone (mean ± SEM = 109 ± 11 mm³; P < 0.0001). Analyses of tumor-free survival by in vivo PET-CT imaging at day 50 showed that cetuximab alone failed to induce complete tumor regression in any animal bearing either SCC-1 tumors (0/8) or 1CC8 tumors (0/10; Fig. 5C). In contrast, combined treatment with cetuximab and TGF-β antibody resulted in complete regression of SCC-1 tumors in 7 of 8 mice (87.5%) and 1CC8 tumors in 9 of 10 mice (90%; Fig. 5C). Treatment with TGF-β antibody alone did not have any appreciable effect on the growth of either SCC-1 or 1CC8 tumor xenografts. These data showed that combinatorial treatment with cetuximab and TGF-β antibody significantly improved tumor-free survival in comparison to either TGF-β alone or cetuximab alone in mice bearing either SCC1 tumors (P < 0.0001) or 1CC8 tumors (P < 0.0001). Therefore, inhibition of TGF-β not only negated the relative de novo resistance of 1CC8 tumors to cetuximab, but also augmented the antitumor efficacy of cetuximab against SCC-1 tumors. Moreover, there was no evidence of toxicity in animals treated with either cetuximab alone or the combination of cetuximab and TGF-β.

The mechanisms by which anti-EGFR antibodies, such as cetuximab, execute their antitumor effect in vivo include blockade of receptor–ligand interactions that stimulate tumor cell survival and growth and engagement of Fcγ receptors on immune effector cells that trigger ADCC of EGFR-expressing tumor cells (11, 12, 15, 27). The importance of both these mechanisms to the in vivo efficacy of anti-EGFR mAbs is supported by previous studies that showed that the F(ab′)₂ fragment of the anti-EGFR antibody 225 (on which cetuximab is based) was able to reduce the in vivo growth of A431 tumor cells with only 50% of the activity shown by the intact antibody comprising a functional Fc domain (11). The crucial contribution of ADCC in mediating the in vivo antitumor activity of therapeutic mAbs is also evident from studies showing that the ability of these agents to arrest growth of tumors in normal mice is impaired in FcRγ^−/− mice that are deficient in activating Fc receptors (27). Although the antibody Fc region triggers the activation of immune effector cells, the concomitant blockade of EGFR-mediated survival signals in tumor cells by the EGFR-specific antibody may also serve to render tumor cells more susceptible to immune effector cell-mediated death (28). The data presented here indicate that tumors play an active role in counteracting both these actions of anti-EGFR mAbs by increasing tumor cell-autonomous expression of TGF-β (29).

Our data show that tumor cell expression of TGF-β can exert a 2-pronged inhibitory effect on cetuximab-mediated ADCC of tumor cells. We find that TGF-β suppresses the expression of several key molecular effectors of immune cell–mediated cytotoxicity, including Apo2L/TRAIL, CD95L/FasL, granzyme B, and IFN-γ. In addition to exerting a tumor cell extrinsic inhibition of the cytotoxic function of immune effectors, TGF-β also promotes the activation of AKT (30), thereby providing an intrinsic EGFR-independent survival signal that protects tumor cells from the direct- and immune cell–mediated cytotoxic effects of cetuximab or erlotinib. Our data indicate that treatment with cetuximab results in the selection of tumor cells that express higher levels of TGF-β, thereby counteracting ADCC and limiting the efficacy of treatment. Conversely, combinatorial treatment with TGF-β–blocking antibody counteracts the selection of TGF-β–overexpressing tumor cells and immune suppression in tumor-bearing animals treated with cetuximab, thereby restoring ADCC and enhancing the antitumor efficacy of cetuximab in vivo. These results identify TGF-β as a key molecular determinant of the de novo or acquired resistance of cancers to EGFR-targeted mAbs, and provide a rationale for combinatorial targeting of TGF-β to improve the antitumor efficacy of EGFR-specific antibody therapy.

Although TGF-β exerts a tumor-suppressive effect on normal epithelial cells, tumor cells frequently become refractory to the growth inhibitory effect of TGF-β and acquire an ability to increase expression and secretion of TGF-β (31–34). This switch enables tumor cells to leverage the tumor promoting effects of TGF-β in the tumor microenvironment to facilitate tumor progression, invasion, and metastasis via promotion of EMT in carcinoma cells and suppression of immune responses (31–39). Many human cancers, including a majority of HNSCC, overexpress TGF-β, and the elevation of TGF-β is correlated with tumor progression, invasion, metastases, and poor prognosis. As such, the production TGF-β by tumor cells may be a frequent mechanism by which cancers induce immune tolerance in the tumor microenvironment and evade elimination by cetuximab. TGF-β–mediated resistance to cetuximab is distinct from previously described mechanisms associated with reduced therapeutic efficacy of cetuximab, such as mutated EGFR variants, redundant autocrine, or paracrine signaling by other EGFR receptors or insulin-like growth factor receptor (IGF-1R), or constitutive activation of either PI3K/AKT or Ras (40–45). The abnormal elevation of serum TGF-β in patients with HPV-negative HNSCC may underlie the relatively lower benefit of cetuximab-based therapy in these patients compared with that observed in patients with HPV-positive HNSCC who exhibit relatively lower levels of TGF-β. Our studies suggest that the clinical efficacy of anti-EGFR antibodies against HNSCC and other EGFR-expressing cancers could be enhanced by strategies to simultaneously sequester and counteract TGF-β in the tumor microenvironment (46–48).

Dr. Califano is the Director of Research of the Milton J. Dance Head and Neck Endowment. The terms of this arrangement are being managed by the Johns Hopkins University in accordance with its conflict of interest policies. Dr. Christine Chung received research funding from AstraZeneca, Lilly Oncology and Bayer, and honoraria from Bristol-Myers Squibb, Amgen, Boehringer Ingelheim and Merck for educational lectures and serving on ad hoc scientific advisory boards. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A. Bedi, X. Chang, K. Noonan, J. A. Califano, I. Borrello, D. Sidransky, R. Ravi

Development of methodology: A. Bedi, X. Chang, K. Noonan, E. J. Fertig, C. H. Chung, R. Ravi

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Bedi, X. Chang, K. Noonan, V. Pham, C. H. Chung, R. Ravi

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Bedi, K. Noonan, R. Bedi, E. J. Fertig, M. Considine, J. A. Califano, I. Borrello, C. H. Chung, R. Ravi

Writing, review, and/or revision of the manuscript: A. Bedi, K. Noonan, R. Bedi, E. J. Fertig, M. Considine, J. A. Califano, C. H. Chung, R. Ravi

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Bedi, V. Pham, R. Bedi, E. J. Fertig, M. Considine, C. H. Chung, R. Ravi

Study supervision: A. Bedi, R. Ravi

The authors thank the Center for Infection and Inflammation Imaging Research (CI3R—www.hopkinsmedicine.org/ci3r) and Dr. Sanjay Jain for conducting PET/CT imaging analysis of mice.

This work was supported by Grant NIH R01CA123277 from the National Cancer Institute, NIH to A. Bedi and R. Ravi; National Cancer Institute SPORE (5P50CA096784-05) to J.A. Califano; NIH R01 DE017982 to C.H. Chung; and NIH K25CA141053 to E.J. Fertig.

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.