The αvβ6 integrin is up-regulated on epithelial malignancies and has been implicated in various aspects of cancer progression. Immunohistochemical analysis of αvβ6 expression in 10 human tumor types showed increased expression relative to normal tissues. Squamous carcinomas of the cervix, skin, esophagus, and head and neck exhibited the highest frequency of expression, with positive immunostaining in 92% (n = 46), 84% (n = 49), 68% (n = 56), and 64% (n = 100) of cases, respectively. We studied the role of αvβ6 in Detroit 562 human pharyngeal carcinoma cells in vitro and in vivo. Prominent αvβ6 expression was detected on tumor xenografts at the tumor-stroma interface resembling the expression on human head and neck carcinomas. Nonetheless, coculturing cells in vitro with matrix proteins did not up-regulate αvβ6 expression. Detroit 562 cells showed αvβ6-dependent adhesion and activation of transforming growth factor-β (TGF-β) that was inhibited >90% with an αvβ6 blocking antibody, 6.3G9. Although both recombinant soluble TGF-β receptor type-II (rsTGF-βRII-Fc) and 6.3G9 inhibited TGF-β–mediated Smad2/3 phosphorylation in vitro, there was no effect on proliferation. Conversely, in vivo, 6.3G9 and rsTGF-βRII-Fc inhibited xenograft tumor growth by 50% (n = 10, P < 0.05) and >90% (n = 10, P < 0.001), respectively, suggesting a role for the microenvironment in this response. However, stromal collagen and smooth muscle actin content in xenograft sections were unchanged with treatments. Although further studies are required to consolidate in vitro and in vivo results and define the mechanisms of tumor inhibition by αvβ6 antibodies, our findings support a role for αvβ6 in human cancer and underscore the therapeutic potential of function blocking αvβ6 antibodies. [Cancer Res 2008;68(2):561–70]
The αvβ6 receptor is one member of the large family of integrin proteins that are expressed in heterodimeric form on the surface of essentially all cells (1, 2). Integrins are involved in regulating a variety of cellular processes, including adhesion, migration, proliferation, differentiation, and death (3–5). The expression of αvβ6 is restricted primarily to epithelial cells where it is expressed at relatively low levels in healthy tissue and significantly up-regulated during development, injury, wound healing, and in epithelial tumors (6–11). The αvβ6 integrin may have multiple regulatory functions in tumor cell biology. The extracellular and transmembrane domains have been shown to mediate transforming growth factor-β (TGF-β) activation and adhesion (12, 13). The cytoplasmic domain of the β6 subunit contains a unique 11-amino-acid sequence that has been implicated in mediating cell proliferation, matrix metalloproteinase production, invasion, and survival (14–18).
αvβ6 can bind to several ligands, including fibronectin, tenascin, and the latency-associated peptide-1 and latency-associated peptide-3 (LAP1 and LAP3), the NH2-terminal region of the latent precursor form of TGF-β1 and TGF-β3, through a direct interaction with an RGD motif (2, 13, 19). αvβ6 binding to LAP1 or LAP3 leads to activation of the latent precursor form of TGF-β1 and TGF-β3 (13). Thus, up-regulated expression of αvβ6 can lead to local activation of TGF-β, which then has the potential to affect neighboring cells in a paracrine fashion. Up-regulated αvβ6 expression has been shown to be associated with increased epithelial to mesenchymal transition, increased tumor cell invasion, and increased metastatic potential in vivo (9, 16, 20, 21). The overexpression of αvβ6 on tumors, particularly at the tumor-stromal interface, may reflect a unique mechanism for local activation of TGF-β1 and the ability to promote tumor progression, invasion, and metastasis.
The TGF-β1 cytokine is a pleiotropic growth factor that can regulate cell proliferation, differentiation, and immune responses (22, 23). Although TGF-β is recognized as an inhibitor of normal epithelial cell proliferation and an early-stage tumor suppressor, tumor cells often acquire a resistance to the growth-suppressive activities of TGF-β1 during tumor progression (24–26). In some tumors, TGF-β1 expression and activity has been implicated in promoting survival, progression, and metastases (25, 27, 28). This is postulated to be mediated by both paracrine effects in the local tumor-stromal environment, including the effects of TGF-β1 on immune surveillance, angiogenesis, and increased tumor interstitial pressure, as well as autocrine effects inducing survival signals directly in tumor cells (29, 30). Several studies have now shown the antitumor effects of inhibiting the TGF-β1 pathway (24, 27, 28, 31, 32).
We have previously described the generation of potent and selective function blocking αvβ6 monoclonal antibodies (mAbs; ref. 33). In this article, we further describe the identification of mAbs that specifically bind to denatured β6, allowing for the analysis of αvβ6 expression on paraffin-embedded tissues. Previously generated αvβ6 antibodies used for immunohistochemistry have not clearly shown selective binding to the αvβ6 integrin. A murine mAb described here, 6.2A1, represents the first αvβ6 antibody to be fully characterized, and that shows specific binding to both intact and denatured forms of both human and mouse αvβ6 integrin. This is a major advantage in that most human tissues that are available for immunohistochemical analysis are only accessible as paraffin-embedded tissue. The 6.2A1 mAb was used to analyze the expression of αvβ6 in a variety of human primary and metastatic cancers, and to compare this expression with that in normal tissue. Further, we generated a human/mouse chimeric version of this mAb (ch6.2A1) to allow immunohistochemical detection of αvβ6 on tissue isolated from mouse models. The human/mouse chimeric form of the mAb contains a human IgG1 Fc region, allowing the use of an antihuman secondary antibody. In our studies, squamous carcinomas showed a particularly high frequency of intense αvβ6 expression, consistent with a potentially important role for this integrin in both the progression and invasion of oral cancers (34). A function blocking αvβ6 mAb, 6.3G9, and rsTGF-βRII-Fc were used to further define the role of αvβ6 in squamous cell carcinomas. Both methods of inhibiting TGF-β led to inhibition of Detroit 562 human pharyngeal carcinoma xenograft growth in vivo, without a direct inhibitory effect on in vitro cell growth. The results described suggest that the in vivo growth inhibitory effects may be mediated through both direct autocrine effects on the tumor cells as well as indirect paracrine interactions with the tumor stromal environment.
Materials and Methods
Antibodies and proteins. Murine αvβ6 mAbs, 6.3G9 and 6.4B4, used for cell-based assays and in vivo efficacy studies, and 6.2A1, used for immunohistochemistry studies, were generated as previously described (33). A human/mouse chimeric form of 6.2A1 (ch6.2A1), used for immunostaining tumor xenograft sections, was generated as described below. Murine rsTGF-βRII-Fc and human soluble αvβ6 (hsαvβ6) were generated as previously described (33, 35). Human recombinant soluble TGF-β receptor type II (rshuTGF-βRII), LAP1, TGF-β1, and latent TGF-β1 were purchased from R&D Systems. Human plasma fibronectin was purchased from Chemicon. Antibodies were purchased from the following vendors: phycoerythrin-conjugated anti-αv (Chemicon); goat anti-mouse IgG phycoerythrin- and horse radish peroxidase (HRP)–conjugated anti-rabbit antibody (Jackson ImmunoResearch); donkey anti-goat IgG Alexa 488 and goat anti-rabbit Alexa Fluor 594 (Molecular Probes); goat anti-human TGF-βRII (R&D Systems); biotinylated anti-mouse, anti-goat, and anti-human IgG (Vector Laboratories); rabbit anti-phosphorylated Smad2/3 (p-Smad2/3; Cell Signaling Technology, Inc.); and rabbit anti–TGF-β (Santa Cruz Biotechnology).
Tissue collections. Paraffin-embedded human tissues were obtained from Phenopath, Inc. Tissue arrays were purchased from Imgenex Histo-arrays and from Petagen.
Cell lines and cell culture. Detroit 562 human pharyngeal cancer cells were purchased from American Type Tissue Collection Repository and cultured in MEM (Eagle) with Earle's (Invitrogen), 0.1 mmol/L nonessential amino acids (Biowhittaker), 1.0 mmol/L sodium pyruvate (Sigma), 10% fetal bovine serum (FBS; JRH Biosciences), and 2 mmol/L l-glutamine (Mediatech). SW480β6 cells and mink lung epithelial cells transfected with a PAI-1-luciferase reporter gene construct (TMLC) were generated and cultured as previously described (12, 36).
High-performance liquid chromatography purification of β6 integrin subunit. Purified hsαvβ6 was loaded onto a 4.6 mm i.d. × 100 mm POROS R2/H reverse phase high-performance liquid chromatography (HPLC) column (Perseptive Biosystems), using a gradient from 35% to 50% acetonitrile in 0.1% trifluoroacetic acid. Peak fractions containing β6 were pooled, lyophilized, and reconstituted in PBS. Proteins were analyzed using electrophoresis on a 4% to 20% SDS-polyacrylamide gel (Invitrogen) under nonreducing conditions and visualized using Coomassie blue protein stain.
Generation of human/mouse chimeric, IgG2a, and aglycosylated IgG2a mAbs. ch6.2A1 cDNA was generated from parent hybridoma RNA using a First Strand cDNA synthesis kit (Amersham/Pharmacia). Heavy and light chain variable regions were amplified by PCR using 3′ primers used for cDNA synthesis and pools of degenerate primers and Pfu DNA polymerase (Stratagene). Variable regions were ligated into expression vectors with human IgG1 constant regions.
To generate the IgG2a and aglycosylated IgG2a forms of 6.3G9, DNA encoding the signal sequence and variable region of 6.3G9 was amplified by PCR with oligonucleotide primers containing a 5′ NotI and 3′ BstE2 site. Constant regions of wild-type and N300Q (aglycosylated) variant forms of murine IgG2a were PCR amplified from murine IgG2a cDNA with primers containing a 5′ BstE2 and 3′ NotI site. Variable and constant regions were cloned into the NotI site of the pKJS157 plasmid (Invitrogen). Murine κ chain was generated by PCR amplification of 6.3G9 cDNA with primers containing a 5′ XbaI and 3′ HindIII site. The murine κ constant region was PCR amplified from murine κ cDNA with primers containing a 5′ HindIII and 3′ XbaI site. Light chain variable and constant regions were cloned into the XbaI site of the pKJS157 plasmid.
Flow cytometry analysis. Cells were resuspended in fluorescence-activated cell sorting (FACS) buffer (1× PBS, 2% FBS, 0.1% NaN3, 1 mmol/L CaCl2, and 1 mmol/L MgCl2) containing 6.3G9 or anti–TGF-βRII antibodies, at 10 μg/mL. Where indicated, rshuTGF-βRII was added in 10-fold excess of the TGF-βRII antibody. After incubation for 30 min on ice, cells were washed and resuspended in FACS buffer containing 5 μg/mL phycoerythrin-conjugated secondary antibodies and incubated on ice for 30 min. After washing, binding of secondary antibody was monitored by flow cytometry.
Protein binding assays and cell-based adhesion assays. These assays were carried out as previously described (33).
Proliferation assays. Detroit 562 cells were seeded at 2.5 × 105 per well in 96-well plates in culture medium containing 2% charcoal-stripped serum and allowed to adhere overnight. Active or latent TGF-β1 was added to quadruplicate wells and cells were incubated at 37°C for 24 h for TGF-β1 titration experiments and for 48 h for blocking studies. Latent TGF-β1 was added at 10 ng/mL, and 6.3G9 was added at 10 μg/mL. Methyl-[3H]thymidine (1 μCi; Amersham/GE Healthcare Biosciences) was added per well for the last 4 h of the incubation period. After trypsinization, cells were harvested using a Brandel MXR96TI Harvester (Brandel) on Wallac 1450 filtermats for subsequent measurement of [3H]thymidine incorporation using a Wallac 1450 Microbeta scintillation counter (Wallac/Perkin-Elmer).
TGF-β1 ELISA. Detroit 562 cells were plated at 5 × 106 per well in six-well plates and incubated overnight. Cells were washed with PBS, and culture medium containing 1% FBS medium was added and incubated at 37°C for 24 h. Culture supernatants were collected and analyzed for both active and total TGF-β1 levels using a Quantikine human TGF-β1 immunoassay kit following the protocol provided (R&D Systems).
TGF-β1 bioassay (TMLC coculture assay). This assay was carried out as previously described (33) using Detroit 562 cells and TMLC, mink lung epithelial cells transfected with a PAI-1-luciferase reporter gene construct. mAbs and rsTGF-βRII were assayed at a final concentration of 10 μg/mL.
Immunohistochemistry. Sections were deparaffinized in xylene and ethanol, rehydrated in water, and immersed in methanol containing 0.45% H2O. Tissues were then incubated with pepsin (Zymed) and blocked with avidin and biotin (Vector Laboratories). For immunostaining β6 on tumor xenograft tissues, sections were incubated with ch6.2A1 and an antihuman biotinylated secondary antibody. For immunostaining β6 on human tissue, sections were incubated with 6.2A1 and an antimurine biotinylated secondary antibody. For immunostaining TGF-βRII, sections were incubated with primary antibody, with or without a 10-fold molar excess of rshuTGF-βRII. Washed sections were incubated with ABC reagent (Vector Laboratories), for 30 min at room temperature, and 3,3′-diaminobenzidine substrate (Vector Laboratories). Frozen tissue sections embedded in OCT compound (Sakura Tokyo) were fixed in acetone and blocked with 0.5% casein/0.05% thimerosal in PBS. For immunostaining, TGF-β sections were incubated with primary antibody and an anti-rabbit Alexa Fluor 488 secondary antibody. Analysis of αvβ6 expression used an H-score, which represents a composite score reflecting both the intensity of staining and the percentage of positively stained cells. The H-score represents the sum of the products of the percentage of positive cells (0–100) and the intensity of staining scored on a four-point scale, where 0 represents no staining above background, 1+ represents weak staining, 2+ represents moderate staining, and 3+ represents strong staining. A maximum H-score of 300 represents 100% of cells exhibiting 3+ staining.
Analysis of integrin β6 and p-Smad2/3 protein expression. For detecting p-Smad2/3, cells were lysed in 150 mmol/L NaCl, 20 mmol/L Tris-HCl (pH 8.0), 2 mmol/L EGTA, 2% Triton X-100, and for detecting β6, cells were lysed in 10 mmol/L HEPES, 0.5 mmol/L MgCl2 (pH 7.5), 1% Triton X-100 each containing Complete Mini protease inhibitor (Roche Diagnostics). After electrophoresis and transfer, membranes were blocked and incubated with 1:10,000 anti-pSmad2/3 or 1 μg/mL 6.2A1, washed, and then incubated with HRP-conjugated secondary antibody. Bands were detected using SuperSignal West Pico chemiluminescent substrate (Pierce Biotechnology, Inc.).
Tumor xenograft studies. Female athymic nude mice, 6 to 8 weeks old (Harlan Sprague-Dawley, Inc.) were used for all studies. Mice were maintained in accordance with the Biogen Idec Institutional Animal Care and Use Committee, and city, state, and federal guidelines for the humane treatment and care of laboratory animals. Animals were s.c. implanted (day 0) on the right flank with 5 × 106 Detroit 562 cells, suspended in medium without serum. Tumors were measured with calipers and tumor size (mm) was converted to mg using the formula mg = (length × width 2) / 2. Body weights were monitored throughout the course of each study. Treatments were initiated when the tumor size reached a minimum weight of 80 mg. Studies were run as randomized, double-blinded trials with a vehicle control (n = 30) and treatment groups (n = 10). Antibodies and rsTGF-βRII-Fc were administered i.p. thrice weekly (dosing every other day). No significant effect on body weight was measured for any of the treatment groups. Statistical comparisons were made comparing control groups and test article using the Student's t test.
Expression of αvβ6 in squamous cell carcinomas. To characterize the expression of αvβ6 across a wide spectrum of human malignant and nonmalignant tissue types, a panel of hybridoma clones was screened for mAbs capable of recognizing the β6 subunit in immunohistochemical staining of paraffin-embedded tissue sections. To this end, the β6 subunit was separated chromatographically from hsαvβ6 (12) and analyzed by SDS-PAGE and Coomassie blue staining (Fig. 1A,, left). The resultant isolated and denatured β6 was used for screening hybridoma supernatants for antibody binding using a solid-phase ELISA (data not shown). Murine mAb, 6.2A1, was identified and chosen for further characterization. 6.2A1 specifically recognized β6 in immunoblots, comparing β6-transfected SW40 cells with untransfected SW480 cells (Fig. 1A,, right), and also immunostained sections of paraffin-embedded β6-transfected SW40 cells but not untransfected SW480 cells (Fig. 1B). 6.2A1 bound to human αvβ6 with a ED50 equal to 62.9 ng/mL, and in contrast to our previously described mAbs, 6.2A1 did not inhibit ligand binding to αvβ6 (33).
Generation of 6.2A1 allowed for analysis of αvβ6 expression across several tumor types by paraffin immunohistochemistry. Although αvβ6 was clearly expressed (H-score >50) in a wide spectrum of epithelial malignancies, expression was particularly notable in carcinomas predominantly of squamous cell origin of the cervix, skin, esophagus, and head and neck, which were positive in 92% (n = 46), 84% (n = 49), 68% (n = 56), and 64% (n = 100) of cases, respectively (Table 1). Strong expression of αvβ6, defined as an H-score >150, was observed in 63% of cervical, 50% of esophageal, 41% of skin, and 35% of head and neck squamous cell carcinomas. A statistically significant correlation between the level of αvβ6 expression and tumor grade and stage was not observed in our studies. An analysis of a larger number of cases, for each tumor type, would be required to determine if a statistically significance relationship exists. In our studies, the n value for most of the individual tumor types was ∼50 cases.
|Tumor site .||Positive expression (H-score >50), % .||N .|
|Head and neck||64||100|
|Tumor site .||Positive expression (H-score >50), % .||N .|
|Head and neck||64||100|
There was no significant difference in the frequency of positive expression in cases of squamous cell carcinoma of the lung compared with nonsquamous cell carcinoma.
The strong expression of αvβ6 in malignant tissues, particularly squamous cell carcinoma, sharply contrasts with the absent or minimal expression observed in nonmalignant tissues (Fig. 1C, and data not shown). In normal tissue, αvβ6 expression was limited to the basal cells of occasional samples of squamous epithelium, in some cases with associated reactive epidermal changes. There was variable but limited expression along the colonic columnar epithelium, by the syncytiotrophoblast cells of placental villi, and in placental membranes (data not shown). αvβ6 was not detected on any connective or neural tissue examined (data not shown). Consistent with its restricted expression in epithelium and absence in normal connective tissue, αvβ6 was not detected on any sarcomas examined (data not shown).
αvβ6 exhibited a unique pattern of expression in a significant percentage of squamous cell carcinomas that was characterized by strong staining of malignant cells located along the periphery of nests of infiltrating tumor tissue (e.g., see Fig. 1C). These cells form the interface with the adjacent tumor stroma. Malignant cells located within the tumor focus exhibited reduced, and occasionally absent, staining. This pattern was particularly prominent among squamous cell carcinoma of the head and neck, where ∼50% of tumors that were positive for αvβ6 expression exhibited this tumor-stroma interface pattern of expression.
Analysis of tumor metastases derived from a wide spectrum of primary epithelial malignancies revealed αvβ6 expression in 77% of cases (n = 52; Fig. 1D). Among nine cases of primary carcinoma (five cervical, two ovarian, one endometrial, and one thyroid) for which a matched metastasis was available for analysis, all cases exhibited an equivalent or higher level of expression (H-score) in the metastatic lesion. Moreover, the metastatic tumor lesions often exhibited the tumor-stroma interface pattern of αvβ6 expression that was commonly observed in primary tumor sections. Together, these results indicate that αvβ6 expression persists through a metastatic transition.
In vitro characterization of αvβ6 function in Detroit 562 cells. We evaluated the functional properties of αvβ6 on the Detroit 562 cell line, derived from a human pharyngeal squamous carcinoma. As determined by flow cytometry, Detroit 562 cells show significant surface expression of αvβ6 (Fig. 2A). Cell adhesion assays were carried out to determine if these cells adhere to the αvβ6 ligands, LAP and fibronectin, and would be inhibited by the function blocking αvβ6 mAb, 6.3G9. This mAb has been previously shown to bind selectively to αvβ6 with an ED50 equal to 14.5 ng/mL (33). Cells bound to both ligands (Fig. 2B) and 10 μg/mL of 6.3G9 inhibited adhesion to LAP by >90% and completely inhibited adhesion to fibronectin. Because the expression of αvβ6 is not always sufficient to confer the ability to activate TGF-β (13), we tested Detroit 562 cells in a coculture system previously described as a method to monitor αvβ6-mediated activation of TGF-β (13). Coculture of Detroit 562 cells with TMLC led to a significant increase in luciferase activity relative to coculture with negative control SW480 cells (Fig. 2C). 6.3G9 and rsTGF-βRII-Fc were tested at 10 μg/mL for their ability to block Detroit 562–mediated TGF-β activation. 6.3G9 reduced activation to background level, whereas rsTGF-βRII-Fc inhibited activation by 44% (Fig. 2C). The cells were assayed for their ability to secrete TGF-β1. Active TGF-β1 was barely detectable whereas ∼300 pg/mL of acid activated latent TGF-β1 were produced in culture after 24 h as measured by ELISA (Fig. 2D). These findings suggest that Detroit 562 cells produce latent TGF-β1 that can be activated in an αvβ6–dependent manner.
Effects of TGF-β, αvβ6 mAbs, and rsTGF-βRII-Fc on Detroit 562 cellular proliferation and Smad phosphorylation in vitro. To determine whether the TGF-β signaling pathway is intact in Detroit 562 cells, we confirmed the expression of TGF-βRI, TGF-βRII, and Smad4 protein, and evaluated the phosphorylation status of Smad2/3 before and after treatment with latent and active forms of TGF-β1. TGF-βRI and Smad4 were expressed by Detroit 562 cells as determined by Western blot analysis (data not shown). TGF-βRII was also expressed as determined by flow cytometry and immunohistochemical analysis of xenograft sections (Fig. 3A and B). The binding of anti–TGF-βRII antibody was completely inhibited by the coincubation of rshuTGF-βRII (Fig. 3A and B). TGF-βRI and TGF-βRII contain no mutations in Detroit 562 cells (data not shown and ref. 37). To confirm that the TGF-β1 receptors are capable of mediating intracellular signaling, we analyzed p-Smad2/3 in the cells in response to the addition of TGF-β1 (Fig. 3C). Addition of either the active or latent forms of TGF-β1 induced p-Smad2/3. In keeping with the ability of αvβ6 to activate TGF-β, we analyzed the effect of 6.3G9 on the phosphorylation of Smad2/3 in Detroit 562 cells alone or in combination with latent or active TGF-β1. The addition of 6.3G9 significantly reduced both the basal level of p-Smad2/3 and the phosphorylation induced by the addition of 100 ng/mL latent TGF-β1 (Fig. 3C). As expected, 6.3G9 had no effect on p-Smad2/3 induced by the addition of active TGF-β1 because the LAP portion of the molecule is not present. The addition of rsTGF-βRII-Fc inhibited both the basal level of p-Smad2/3 and that induced by the addition of active TGF-β1.
We also tested the effects of TGF-β1 (38) on the proliferation of Detroit 562 cells in vitro. The addition of either the active (Fig. 4A) or latent form of TGF-β1 (Fig. 4B) led to a modest inhibition of growth. A maximum inhibition of 20% was achieved by active TGF-β1 that was statistically significant at concentrations ≥74 pg/mL (P < 0.05). A maximum inhibition of 25% was achieved with latent TGF-β1 that was statistically significant at concentrations ≥195 pg/mL (P < 0.05). The addition of 6.3G9 or rsTGF-βRII-Fc had no direct effects on the proliferation of Detroit 562 cells (Fig. 4C) nor did 6.3G9 block the inhibition of growth induced by the addition of latent TGF-β1 (Fig. 4D). In similar studies, Detroit 562 cells were cultured in collagen or Matrigel and no significant effects of 6.3G9 on cell proliferation were observed (data not shown). Collectively, these findings suggest that whereas αvβ6 can mediate TGF-β–induced intracellular signaling, it does not seem to regulate Detroit 562 proliferation in an autocrine manner as assessed in this in vitro system.
Effect of anti-αvβ6 mAbs and rsTGF-βRII-Fc on growth of Detroit 562 cells in vivo. Detroit 562 cells grown as xenograft tumors were evaluated for the expression of αvβ6 and TGF-β using immunohistochemistry. Although the majority of cells express αvβ6 by flow cytometry, the xenograft tumors were found to express αvβ6 predominantly at the tumor-stroma interface (Fig. 5A). This suggests that there is further regulation of αvβ6 expression in vivo leading to increased expression at the tumor-stromal interface. Immunohistochemical analysis of TGF-β in the xenograft sections showed significant TGF-β in the surrounding stroma (Fig. 5A). We investigated whether treatment with αvβ6 mAbs would inhibit tumor growth and also tested the effect of rsTGF-βRII-Fc as an inhibitor of the TGF-β pathway. Treatments included function blocking αvβ6 mAb, 6.3G9; non–function blocking αvβ6 mAb, 6.4B4; rsTGF-βRII-Fc; or vehicle control (Fig. 5B–D). In a dose ranging study, 6.3G9 inhibited tumor growth in a dose-dependent manner resulting in 50% inhibition at 10 mg/kg compared with vehicle control–treated mice (n = 10, P < 0.05; Fig. 5B). Equivalent doses of 6.4B4, an antibody that binds with high affinity to αvβ6 but shows minimal ligand blocking activity (33), had no effect on tumor growth (Fig. 5C). rsTGF-βRII-Fc, at 2 mg/kg, resulted in a nearly complete inhibition of tumor growth (n = 10, P < 0.001; Fig. 5B and C). In four separate experiments, 6.3G9 produced a statistically significant inhibition of tumor growth with activity ranging from 40% to 70% compared with vehicle control. Inhibition in vivo by a function blocking αvβ6 mAb and rsTGF-βRII-Fc, and not by the nonblocking αvβ6 mAb, suggests that blocking TGF-β activity is the relevant mechanism of action of these molecules.
Additional xenograft studies were conducted to evaluate the importance of antibody-mediated effector function for 6.3G9. An IgG2a version of 6.3G9 was generated to provide optimal effector function in mice. In addition, an IgG2a construct with an N300Q mutation in the CH2 domain of the mAb was generated to eliminate glycosylation of the mAb and to inhibit Fcγ receptor binding as previously reported (39, 40). Similar approaches have been taken with other mAbs to evaluate the role of effector function in vivo (41). Wild-type and mutant aglycosylated IgG2a versions of 6.3G9 showed similar binding affinities for αvβ6 and blocked ligand binding with similar potencies (data not shown). All forms of the mAb had a similar serum half-life in vivo (data not shown). Wild-type and aglycosylated IgG2a forms of 6.3G9 produced similar inhibitory effects, which was comparable with the inhibition of growth detected with the parent IgG1 version of 6.3G9 (Fig. 5D). These findings indicate that the in vivo efficacy detected with 6.3G9 is not mediated by effector function.
The expression of αvβ6 in a variety of epithelial malignancies was analyzed using the 6.2A1 antibody. These studies showed that αvβ6 was most highly up-regulated in squamous cell carcinomas, where it was found to be expressed in more than 60% of cases evaluated of the head and neck, esophagus, cervix, and skin, thus further validating and extending previous reports of αvβ6 in human carcinomas (6, 9–11, 42). In addition, 77% of metastatic lesions, representing a range of tumor types that had metastasized to different tissues, were found to overexpress αvβ6. An association of αvβ6 expression with clinical variables such as disease progression, metastases, and survival has been described (9, 11, 42, 43). However, although a correlation between αvβ6 expression and tumor grade and stage has been reported in cervical squamous cell carcinomas (11), no correlation was found to exist in other studies analyzing different tumor types (42, 44, 45). These differences may relate in part to the sample size evaluated or reflect real differences among the various tumor types. Collectively, these findings support the hypothesis that αvβ6 may facilitate tumor progression, invasion, and metastasis, and may represent a prognostic marker in certain tumor types.
We observed increased expression of αvβ6 on head and neck carcinomas particularly at the invasive edge where tumor cells are in direct contact with the tumor-associated stroma. Similar leading edge expression was found in the Detroit 562 xenograft tumors, thus offering an attractive model to study functional effects of blocking αvβ6 in vivo. In vitro, the function blocking αvβ6 mAb, 6.3G9, inhibited latent TGF-β–induced Smad2/3 phosphorylation in Detroit 562 cells, but had no direct effect on proliferation. However, in vivo, both 6.3G9 and rsTGF-βRII-Fc significantly inhibited Detroit 562 tumor growth. Because the inhibitory effects of mAbs on tumor growth in vivo can be mediated at least in part by antibody-dependent cellular cytotoxicity (41, 46), we generated an IgG2a version of 6.3G9 with an N300Q mutation in the CH2 domain of the mAb. Glycosylation of this domain mediates Fcγ receptor binding and is inhibited by this mutation, subsequently attenuating antibody-dependent cellular cytotoxicity (39, 40). These studies revealed that the ability of 6.3G9 to inhibit tumor growth in vivo was not related to effector function as the wild-type and mutant forms of the antibody had equivalent effects on tumor growth. The difference in the degree of tumor growth inhibition observed comparing 6.3G9 and rsTGF-βRII-Fc may reflect the extent of TGF-β inhibition by the different therapeutics. rsTGF-βRII has the potential to inhibit all active TGF-β present in the stromal compartment, whereas 6.3G9 is limited to inhibition of TGF-β activation at the epithelial tumor-stromal interface.
The αvβ6 integrin not only mediates TGF-β activation, but has also been reported to regulate intracellular signaling upon binding to extracellular matrix ligands. In vitro and in vivo studies have suggested that αvβ6 binding to fibronectin can promote migration, invasion, matrix metalloproteinase production, and metastases through fyn and extracellular signal-regulated kinase/mitogen-activated protein kinase (MAPK) signaling pathways (16, 34, 47). Our studies evaluating the role of αvβ6 mediating intracellular signaling through MAPK and AKT did not reveal direct effects of blocking αvβ6 with mAbs on these downstream events in Detroit 562 cells in vitro (data not shown). Our findings suggest that the in vivo tumor inhibitory effects of blocking αvβ6 or TGF-β may be mediated in part through paracrine effects involving the tumor microenvironment. The predominant expression of αvβ6 on the leading edge of tumors suggests that expression may be mediated by matrix interactions. In vitro culture of Detroit 562 cells in the presence of collagen I, collagen IV, laminin, or fibronectin resulted in no difference in αvβ6 expression as determined by flow cytometry (data not shown). However, immunostaining of Detroit 562 xenograft tissue for smooth muscle actin and collagen (Sirius red stain) consistently exhibited a more reticular pattern of expression along the periphery of the xenograft tumor (data not shown), similar to the localized expression of αvβ6 on the leading edges of the same tumors. These observations are consistent with a correlation between αvβ6 expression and spatial characteristics of the associated matrix. Similarly, staining of human oral carcinoma revealed smooth muscle actin staining on the invasive front of the tumors at the same locations where αvβ6 expression was detected. Although we observed this pattern of expression in both xenografts and human primary tumors, no clear differences in collagen and smooth muscle actin staining was observed in the xenograft studies comparing the different treatment groups. Additionally, no difference in angiogenesis, as determined by CD31 immunostaining, was detected (data not shown). It is possible that αvβ6 plays a role in tumor progression in vivo through its proinvasion properties (17, 20, 21). The ability of αvβ6 to promote tumor cell invasion into the surrounding stroma may induce a growth-permissive environment. We found that Detroit 562 cells invade in vitro and this activity is blocked by the addition of αvβ6 mAbs (data not shown). It is possible that in vivo, the αvβ6 blocking mAb is affecting tumor progression through direct effects on the invasive properties of the tumor cells. This would be consistent with the predominant expression of αvβ6 on the leading invasive edge of the Detroit 562 tumor xenografts. It has been shown that TGF-β can promote increased expression of αvβ6 (7). Consequently, αvβ6 expressed on tumors provides a mechanism for local activation of TGF-β, which can, in turn, further drive increased αvβ6 expression in vivo. This could create a self-sustaining positive-feedback loop leading to maintenance of TGF-β activation in a localized manner. In support of this concept, we found that expression of αvβ6 on Detroit 562 xenograft tumors was nearly completely down-modulated by in vivo treatment with rsTGF-βRII-Fc (data not shown). In the xenograft model, it is possible that 6.3G9 disrupts a feedback loop by inhibiting TGF-β activation, and rsTGF-βRII-Fc could do so by blocking TGF-β directly.
αvβ6 expression is minimal to negligible in most normal tissues but highly up-regulated on epithelial tumors and metastases. Thus, αvβ6 mAbs provide an approach for localized suppression of TGF-β, which could otherwise be a major limitation to the use of systemic inhibitors of the TGF-β pathway. The importance of this is underscored by the known homeostatic roles of the TGF-β pathway in regulating inflammation, immune tolerance, and tumor suppression (22, 23, 48, 49). A comparison of the genetic null phenotypes for αvβ6 and TGF-β, and conditional TGF-β pathway null mice, further supports the hypothesis that blocking αvβ6-mediated TGF-β activation may be preferable to global targeting of the TGF-β pathway in disease. Mice completely deficient for αvβ6 function develop normally and have a typical lifespan, with mild inflammation that is limited to the lung and skin, and with late-onset emphysema (50). In contrast, TGF-β–deficient mice that do not die in utero show severe inflammation in multiple organ systems, resulting in death at 3 to 4 weeks of age, and conditional TGF-β pathway null mice have been associated with significantly increased tumorigenesis (23, 48, 49, 51, 52). Thus, αvβ6 blocking mAbs provide an opportunity to inhibit tumor growth and minimize the potential for tumor promoting and pro-inflammatory activity that could be associated with systemic blockade of TGF-β.
Note: L.A. Koopman, D.R. Leone, and S. Ho contributed equally to this work.
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.
We thank Joe Amatucci and Konrad Miatkowski for experimental assistance, and Ellen Garber for her guidance and advice.