Presence of Membrane-Bound TGF-β1 and Its Regulation by IL-2-Activated Immune Cell-Derived IFN-γ in Head and Neck Squamous Cell Carcinoma Cell Lines¹

Transforming growth factor β1 is a well-characterized immune suppressive cytokine and, in the tumor microenvironment, is the most predominantly expressed of all members of the TGF-β superfamily. TGF-β1 is secreted by tumor cells and regulatory immune cells, such as regulatory T cells (Treg),³ tumor-associated macrophages, immature dendritic cells (1), and CD14⁺ HLA-DR^low/− monocytes of cancer patients (2). We and others have reported that elevated serum levels of TGF-β are correlated with poor prognosis for cancer patients and the suppression of antitumor immunity (3, 4, 5). TGF-β1 diversely suppresses host innate and adaptive immune systems by inhibiting the growth and activation of antitumor effector cells and down-regulating NK receptors (NKG2D, NKp30) (6) and cytotoxic molecules (granzyme, perforin) (7, 8). In addition to inhibiting immune suppression, TGF-β1 is also involved in several mechanisms of tumor promotion, such as metastasis (9), angiogenesis, and the epithelial to mesenchymal transition (10).

TGF-β acts as growth inhibitor of normal cells and cells in the early stages of tumors. However, during the development of various types of cancer, tumor cells become resistant to TGF-β by acquiring mutations in TGF-β receptors and/or the core Smad components (11, 12). In addition to gaining responsiveness to TGF-β signaling, malignant tumor cells actively secrete TGF-β (13). Overexpression of TGF-β1 and suppression of antitumor immunity have been reported in head and neck squamous cell carcinoma (HNSCC) cells of both humans and mice (14, 15, 16). However, little is known about the mechanisms responsible for the overexpression of TGF-β in tumor cells. Aside from cancer, TGF-β is a major factor in the regulation of inflammation, wound healing, and several immunological diseases, such as rheumatoid arthritis, fibrosis, and asthma (17). Therefore, induction of TGF-β in these diseases implies that inflammation-associated factors may be involved in the overexpression of TGF-β. We hypothesized that TGF-β1 overexpression in tumor cells occurs through activated immune cell-derived factors rather than through an oncogenic mechanism. To test this, we investigated whether activated immune cells are able to regulate the levels of membrane-bound TGF-β1 (mTGF-β1) in HNSCC cells when the two cell types are cultured together.

Five HNSCC cell lines, SNU-1041, SNU-1066, SNU-1076, PCI-1, and PCI-50, were used in this study. SNU cell lines were obtained from the Korean Cell Line Bank (18) (Seoul National University, Seoul, Korea) and PCI cell lines were obtained from the Pittsburgh Cancer Institute (19) (University of Pittsburgh, Pittsburgh, PA). All cells were cultured in RPMI 1640 (for SNU cell lines) or DMEM (for PCI cell lines) supplemented with 10% FBS and 10 mg/ml gentamicin.

Human PBLs were purified from whole blood of healthy donors using Ficoll-Hypaque (GE Healthcare) density gradient centrifugation followed by depletion of monocytes by adhesion. For activation, PBLs were cultured in RPMI 1640 medium with 500 U/ml recombinant human (rh) IL-2 and 10% FBS for 3 days. NK cells were purified by adding RosetteSep NK enrichment mixture (StemCell Technologies) before density gradient centrifugation and cultured with 200 U/ml rhIL-2 in RPMI 1640 with 10% FBS. After 3 days, cell culture supernatants were harvested by centrifugation, filtered, and cryopreserved at −20°C.

FITC- or PE-conjugated Abs against human, phospho-STAT1 (pY701), STAT3 (pY705), and NF-κB p65 (pS529) were from BD Pharmingen. Nonconjugated Abs against TGF-β1 (clone 141322; R&D Systems), integrin α_vβ₆ (Chemicon International) and control IgG (R&D Systems) were used as primary Abs followed by secondary Ab PE-conjugated donkey anti-mouse IgG (Jackson ImmunoResearch Laboratories) for indirect flow cytometry analysis. Recombinant human latency-associated peptide (LAP), TGF-β1, IL-1β, IL-2, IL-6, IFN-γ, and TNF-α were purchased from R&D Systems. Actinomycin D and cycloheximide were purchased from Calbiochem. Membrane translocation inhibitor, brefeldin A, was used with GolgiPlug (BD Biosciences).

Cytokine (IL-1β, IL-6, IFN-γ, and TNF-α) concentrations in culture supernatants were measured by the sandwich ELISA method. All reagents and Abs were purchased from Endogen, except IL-6 (R&D Systems). The minimum detectable dose was 10 pg/ml. Calculations of results were done using Softmax Pro software (Molecular Devices).

HNSCC cell lines were seeded in a 12- or 24-well plate overnight to obtain a cell monolayer of 90% confluence. IL-2-activated PBLs or NK cells were cultured for 3 days and then harvested. Cells were resuspended in new medium and cocultured with HNSCC cell lines at a ratio of 1:1. To block direct cell contact, when coculturing, we seeded immune cells on Transwell inserts (pore size, 0.4 μm; Corning) of the upper wells.

To avoid loss of the adhesion molecules, HNSCC cells were detached using cell dissociation buffer (Invitrogen). Cells were indirectly bound to mouse anti-TGF-β1 Ab followed by PE-conjugated donkey anti-mouse IgG. In some experiments, dissociated cells were incubated with recombinant human LAP at 4°C for 2 h (20). Stained cells were analyzed on a FACSCalibur flow cytometer using CellQuest software (BD Biosciences).

To detect phosphorylated proteins (p-STAT1, p-STAT3, p-NF-κB p65), harvested cells were fixed with 4% paraformaldehyde (Cytofix fixation buffer; BD Pharmingen) for 10 min at 37°C and then permeabilized with ice-cold 90% methanol (Perm III buffer; BD Pharmingen) for 30 min on ice. Cells were then stained and analyzed.

Total RNA from SNU-1066 cells was purified using the RNAqueous-4PCR kit (Ambion) following the manufacturer’s instructions. DNase-treated mRNA (1 μg) was reverse-transcribed with Moloney murine leukemia virus reverse transcriptase (RETROscript kit; Ambion). PCR were performed with a PE Applied Biosystems Prism 7000 sequence detector with the TaqMan Universal PCR Master Mix and Primer/Probe Mixture (predesigned TaqMan Gene Expression Assays; PE Applied Biosystems). Data were analyzed according to the comparative threshold cycle method using the internal control (GAPDH) transcript levels to normalize differences in sample loading and preparation. Results are represented as the n-fold difference of transcript levels between different samples and are expressed as the mean ± SD of triplicate reaction wells.

We cocultured IL-2-activated PBLs with five HNSCC cell lines and evaluated changes in the level of mTGF-β1 expression in the cancer cell lines by FACS analysis. After coculturing, the amount of mTGF-β1 in the cancer cells was significantly higher than in the control group, even though the basal level and degree of change among the cell lines varied (Fig. 1,A). We also obtained similar results from coculturing IL-2-activated NK cells with the cancer cell lines (data not shown). To determine whether the induction of mTGF-β1 occurred via a soluble factor or through direct cell-cell contact, we seeded IL-2-activated PBLs in Transwell inserts. We also treated HNSCC cells with the CM of activated PBLs or NK cells. mTGF-β1 levels in the cancer cell lines increased even without direct cell contact (Fig. 1, B and C). We concluded that soluble factors derived from IL-2-activated PBLs or NK cells were responsible for the significant induction of mTGF-β1 expression in the cancer cell lines. In some experiments, we observed that direct cell contact induced even higher levels of mTGF-β1 expression than what was induced in the Transwell group. We interpreted these results to mean that an induction of immune cell-derived factors was occurring in response to cell contact. When immune cells contact cancer cells, they secrete higher levels of IFN-γ (see above) and we confirmed this phenomenon using ELISA and intracellular cytokine staining (data not shown). Because CM contains rhIL-2 used for immune cell activation, we treated the cancer cells with various concentrations of rhIL-2 (10–2000 U/ml). We found the induction of mTGF-β1 to be IL-2 independent because treatment of IL-2 did not increase mTGF-β1 levels in the treated cancer cells (data not shown).

To determine which soluble factors were responsible for the induction of mTGF-β1 in the cancer cells, we analyzed the transcription factor profiles of SNU-1066 cells after treatment with CM. We detected phosphorylation of STAT3 and STAT1 in response to CM treatment, whereas NF-κB p65 was not phosphorylated (Fig. 2,A). We then analyzed the cytokine profile of CM by ELISA. IL-2-activated PBLs secreted high levels of IL-1β (1486 ± 221.2 pg/ml), IL-6 (998 ± 23.8 pg/ml), and IFN-γ (707 ± 8.7 pg/ml), while NK cells secreted IFN-γ (703 ± 5.6 pg/ml) only. Both PBLs and NK cells secreted very low levels of TNF-α (below 50 pg/ml) by our activation method. Because it is well known that TGF-β1 is overexpressed during inflammation, we hypothesized that proinflammatory cytokines, such as IL-1β, IL-6, and IFN-γ, may be responsible for induction of TGF-β1 in the cocultured cancer cells. To verify this, we treated SNU-1066 cells with various concentrations of these cytokines. After 24 h of exogenous cytokine treatment, we detected an induction of mTGF-β1 only in IFN-γ-treated cells and not in IL-1β- or IL-6-treated cells (Fig. 2,B). To test whether the IFN-γ in CM is indeed the TGF-β1 inducer, we incubated CM with an anti-IFN-γ neutralization Ab before treating the SNU-1066 cells. The induction of mTGF-β1 by CM was almost completely abolished when IFN-γ was first blocked with neutralizing Ab (Fig. 2,C). We confirmed the IFN-γ-mediated mTGF-β1 induction in all five HNSCC cell lines (Fig. 2 D).

Although the IL-2-activated PBLs and NK cells secreted the minimum level of TNF-α or less in our initial experiments, proinflammatory cytokine TNF-α is an important factor in the interaction between tumor and immune cells, such as tumor-associated macrophages (21). For that reason, we examined the effect of exogenous TNF-α on mTGF-β1 expression. Addition of exogenous rhTNF-α increased mTGF-β1 expression in SNU-1066 cells in a dose-dependent manner (Fig. 3,A). Moreover, when IFN-γ was combined with TNF-α, the effect was synergistically enhanced (Fig. 3,B). Although all five HNSCC cell lines showed significant induction of mTGF-β1 when cocultured with IL-2-activated PBLs and NK cells, two of the five cell lines did not respond to TNF-α (Fig. 3,C). A previous study has shown that several tumor cells express endogenous TNF-α, which is up-regulated by exogenous cytokines (22). As shown in Fig. 3 D, addition of exogenous TNF-α induced endogenous TNF-α mRNA levels in SNU-1066 cells, whereas we were unable to detect IFN-γ mRNA regardless of cytokine treatment. This suggests that TNF-α up-regulates both mTGF-β1 and endogenous TNF-α expression in this cell line and that TNF-α may induce mTGF-β1. It has been reported that IFN-γ induces TNF-α-regulated signaling by inducing TNF-α expression in mouse macrophages (23, 24). However, in our model, IFN-γ alone is not sufficient to induce endogenous TNF-α expression. Therefore, IFN-γ-mediated mTGF-β1 induction is not a secondary effect mediated by induced endogenous TNF-α. In addition, we also measured the change in IFN/TNF receptors in response to exogenous cytokines. However, we were unable to detect significant changes in any of the five cancer cell lines (data now shown).

TGF-β1 is secreted in a complex with LAP. After proteolytic cleavage of LAP, mature TGF-β1 is liberated and binds the TGF-β receptor (25). We tested several possibilities to determine the molecular nature of the membrane-bound form of TGF-β1. First, because it has been reported that the membrane integrin α_vβ₆ can bind to the LAP-TGF-β1 complex and activated it, we measured the change in integrin α_vβ₆ levels in the cancer cell lines (26). By flow cytometry, the changes of integrin α_vβ₆ expression levels by the addition of exogenous IFN/TNF are not correlated with mTGF-β1 (data not shown). Therefore, mTGF-β1 induction by IFN/TNF is integrin α_vβ₆ independent. In addition, we incubated each cell line with exogenous recombinant TGF-β1 protein at concentrations of up to 50 ng/ml, but still detected no increase in mTGF-β1 (data now shown). Therefore, the mTGF-β1 we detected by flow cytometry is not a rebound form of once-secreted mTGF-β1. Next, we tested the possibility that IFN-γ or TNF-α can cleave LAP and liberate mTGF-β1. If this was the case, we expect that in CM-treated cells surface levels of LAP expression would be decreased. However, with the addition of CM, LAP expression levels were higher than in the control group (Fig. 4). Moreover, CM-treated cells had higher rhLAP-binding ability. Because rhLAP has the ability to bind bioactive TGF-β1 at the cell membrane (1, 27), these findings suggest that the mTGF-β1 up-regulated by IFN-γ is newly synthesized and bioactive.

To confirm the de novo synthesis of mTGF-β1, we treated cells with the inhibitors of several mechanisms that may be involved in this process. First, we blocked the membrane translocation of TGF-β1 with brefeldin A (BFA). Because BFA interferes with protein transport between the endoplasmic reticulum and the Golgi apparatus, it would specifically block the membrane localization of newly synthesized TGF-β1. As shown in Fig. 5,A, mTGF-β1 expression levels in BFA-pretreated SNU-1066 cells remained unchanged by the addition of IFN-γ. In addition, we preincubated cells with either the transcription inhibitor actinomycin D (ActD) or translation inhibitor cycloheximide (CHX) before IFN-γ treatment. ActD effectively inhibited mTGF-β1 induction by IFN-γ for 12 h (Fig. 5,B). Although the effect was only short term, CHX also blocked mTGF-β1 induction for 4–8 h. Finally, we investigated the effect of IFN-γ and TNF-α on TGF-β1 mRNA regulation using real-time RT-PCR analysis. As shown in Fig. 5,C, both IFN-γ and TNF-α increased TGF-β1 mRNA expression in SNU-1066 cells, peaking at 9 h after treatment. Although the increase in mRNA expression was not as significant as the increase in protein levels, because mTGF-β1 induction was initiated soon after cytokine treatment (Fig. 5 C), we conclude that elevated TGF-β1 mRNA is rapidly translated. Altogether, these data demonstrate that IFN-γ- or TNF-α-mediated mTGF-β1 induction is the result of de novo synthesis.

The overexpression of TGF-β1 in cancer cells and the down-regulation of tumor immunity are well-known phenomena (3, 4, 28). In this study, we present the novel finding that tumor cells, like Treg, express an active membrane-bound form of TGF-β1, which is up-regulated by extrinsic factors such as IFN/TNF. Recently, several reports have shown that Treg use mTGF-β1 without secretion to suppress non-Treg through direct cell contacts (29, 30, 31). TGF-β1 is secreted in an inactive form in a complex with LAP and it is generally accepted that TGF-β1activation (liberated from LAP) can occur outside the cell. Several mechanisms for TGF-β1 activation have been proposed, but they all focus on cellular TGF-β1 activation rather than newly synthesized, LAP-free TGF-β1.

To our knowledge, only one study has been reported on the role of mTGF-β1 in cancer cells (32). In that report, Baker et al. (32) use FACS analysis to show that colorectal cancer cell lines express functional TGF-β1 at the cell membrane and demonstrate by the low pH buffer elution method that the FACS-detected mTGF-β1 is not a population of secreted TGF-β1 that has reassociated with the membrane. Pulsing cells with low pH solution to elute membrane-rebound TGF-β1 was also done by Broderick et al.(33) in human memory T cells. In contrast to the results of Broderick et al. (33), Baker et al. (32) detected increased expression of mTGF-β1 after elution, which is what we observed in the present study (data now shown). Our results suggest that the increase in detected TGF-β1 is endogenous. Increased mTGF-β1 expression after low pH treatment implies that acidic conditions may liberate active TGF-β1 from LAP. No correlation between integrin α_vβ₆ and mTGF-β1 expression was observed after IFN/TNF treatment and no increases in mTGF-β1 were detected by treatment with exogenous rhTGF-β1, which also suggests that the mTGF-β1 is endogenous. In addition, we observed that CM-treated SNU-1066 cells are able to bind more exogenous rhLAP than control cells. As reported, because rhLAP has a strong affinity for active TGF-β1 (1, 20, 27), an increase in the intensity of LAP staining means that more active TGF-β1 is present at the cell membrane.

We have reported measuring elevated TGF-β1 plasma levels in cancer patients (5). However, elevated levels of systemic TGF-β1 are not necessarily derived from tumor cells because the half-life of circulating TGF-β is very short (34). TGF-β1, not bound to LAP, has a high affinity for cell membranes and is rapidly diminished from culture medium in vitro (35). Moreover, different mechanisms seem to be involved in generating the membrane-bound and secreted forms of TGF-β1. We analyzed the levels of secreted active or latent TGF-β1 by ELISA and compared them with mTGF-β1 levels, but the correlation varied between different cell lines (our unpublished data). Therefore, the effect of mTGF-β1 may be limited to the infiltrated immune cells in the tumor microenvironment.

We speculate that expression of active TGF-β1 in cancer cells is regulated by tumor-stroma interactions. Our findings are supported by those of other investigators: Dasgupta et al. (16) have shown that expression of active TGF-β1 occurs only in vivo, while cultured HNSCC cells do not express TGF-β1 before inoculation in vitro. Lim et al. (36) have also shown that inoculated tumor cells express higher levels of TGF-β1 in vivo compared with cells cultured in vitro). Using a mouse model, these authors found that the in vivo overexpression of TGF-β may present a limitation to dendritic cell-based therapy. Although the mechanism for this has not been elucidated, these two studies have led to our interpretation that active TGF-β1 expression in cancer cells is regulated by tumor-stroma interactions. As mentioned above, TGF-β is generally induced during inflammation. Many investigators agree that some types of cancer are related to infection or chronic inflammation. Without infection during tumor development, tissue destruction occurs, which can cause local inflammation. Inflammatory factors induce TGF-β overexpression in tumor cells and accelerate genetic instability. In this process, tumor cells gaining TGF-β resistance through mutation of TGF-β receptors or Smad molecules can be selected and survive. Although we used TGF-β-resistant cell lines (18), even TGF-β-sensitive cells secrete TGF-β. Indeed, we treated the TGF-β-sensitive gastric cell lines, SNU-16 and SNU-620, with IFN/TNF and observed an induction in mTGF-β1 (our unpublished data).

In the tumor microenvironment, tumor cells have many tools to evade antitumor immunity. Some use intrinsic mechanisms through transformation processes, others may be extrinsic, i.e., immunoselection. For example, local tissue destruction is accompanied by tumorigenesis, leading to inflammation and an abundant infiltration of immune cells. Less immunogenic tumor cells may survive in this process; on the hand, tumor cells have to detect immune activating factors and evade antitumor immunity (tumor count attack). Many types of tumor cells can use proinflammatory cytokines as growth factors or recognize them as danger signals. For example, the representative proinflammatory cytokine TNF-α activates antiapoptotic and metastatic molecules through the NF-κB pathway (37). IL-6, another proinflammatory cytokine, is also used by tumor cells as a growth-promoting and antiapoptotic factor (38). In the tumor microenvironment, IL-6 phosphorylates STAT3 in stromal and tumor cells (39). Through this mechanism, induced STAT3-regulated factors, such as vascular endothelial growth factor and IL-10, reversibly attenuate antitumor immunity (40, 41). Many studies have reported that tumor cells can act as immune cells. For example, just like immune cells, tumor cells use cytokines as growth factors and express the chemokine receptors (42). The cytokines expressed by tumor, including TGF-β1, are used to “educate” infiltrated immune cells to help the survival and invasion of the tumor (21). Although some cytokines derived from immune cells inhibit tumor growth or induce apoptosis, tumor cells effectively detect and block them. That is, tumor cells may detect immune activating factors and responses.

Our intention here is not to insist that IFN-γ is a harmful cytokine regulator of tumor immunity. IFN-γ is a crucial factor for immune surveillance and antitumor responses. Moreover, IFN-γ directly inhibits tumor cell growth. Indeed, we observed that IFN-γ decreases cell growth in three of the SNU cell lines used in this study by MTT assay (data now shown). However, tumor cells have several defense mechanisms against IFN-γ, both intrinsically and extrinsically. It has been reported that STAT1 is essential for docetaxel resistance in prostate cancer (43). In tumor immunity, IFN-γ decreases NK cell sensitivity of mouse MCA sarcoma by reducing NKG2D ligand H60 expression (44). Hanada et al. (48) have reported the spontaneous development of colorectal carcinomas in SOCS1-deficient mice. The authors imply that this is due to hyperactivation of STAT1, a key factor of IFN-γ signaling. In this model, inducible NO synthases and cyclooxygenase 2 induced by IFN-γ are major tumor-promoting factors. Moreover, expression of the immune regulatory factor IDO is also induced by IFN-γ and TNF-α, as is mTGF-β1 in our model (45). Indeed, we have observed that cyclooxygenase 2 and IDO are up-regulated by the addition of IFN-γ and TNF-α to SNU-1066 cells (our unpublished data). It has been assumed that myeloid-derived suppressor cells use arginase I and inducible NO synthase for immune suppression. Toward this end, IFN-γ may be involved in IL-4/13 signaling-mediated initiation of immune tolerance (46). Tumor-associated B7-H1, one of the B7 family inhibitory receptors, is up-regulated by IFN and induces T cell apoptosis (47). These studies show that tumor cells or regulatory immune cells use several mechanisms of suppression in response to IFN-γ.

We have shown that IFN-γ- and TNF-α-induced mTGF-β1 overexpression is a result of de novo synthesis, but the molecular mechanism responsible for this must be further elucidated. There are no known binding sites for IFN-γ-mediated transcription factors, such as STAT1, in the TGF-β1 promoter. Our future studies will look at the role of STAT1 in IFN-γ-mediated TGF-β1 overexpression.

In summary, we have described a novel immune evasion mechanism of tumor cells that responds to inflammatory cytokines via TGF-β1 induction. These findings provide significant insight into tumor immunity and may ultimately enhance the development of immunotherapy.

We are grateful to Dr. Chung-Gyu Park, Dr. Jae-Jung Lee, and Dr. Seok-Woo Park for helpful discussion.

The authors have no financial conflict of interest.