The tumor suppressive effect of TGF-β was supported by several animal models with defected TGF-β signaling. For example, TβRI/phosphatase and tensin homolog (PTEN) knockout mice developed full-penetrance HNSCC while mice with PTEN deletion presented with hyperproliferation in the head and neck epithelium (30). Similarly, Smad4 deletion in head and neck epithelium also exhibited spontaneous HNSCC in mice (26). In normal epithelial cells, TGF-β may maintain homeostasis through the regulation of proliferation and apoptosis. TGF-β signaling arrests the cell cycle in phase G1, and mechanisms underlying normal cell growth inhibition include upregulation of cyclin-dependent kinase (CDK) inhibitors and downregulation of Myc expression (31). Smad3/Smad4 complexes interact with forkhead box O (FoxO) to increase the expression of CDK inhibitors, namely p15 and p21 (1,3,32). In the presence of co-repressors, Smad3/Smad4 complex also interact with regulatory elements of the Myc promoter, a cell cycle regulator gene (33,34). As a consequence, the mRNA and protein expression of Myc are reduced. Smads induce apoptosis in epithelial cells through the activation of P53, Bcl-2-like protein 11 and death-associated protein kinase, and the repression of Akt. TβRI also induces apoptosis in normal epithelial cells through TGF-β-activated kinase 1 (TAK1)-p38/c-Jun N-terminal kinase, independent of Smads (Fig. 1) (1,35–40). However, there is no direct evidence to support that the same molecular mechanisms are involved in HNSCC. Inhibitor of DNA-binding/differentiation (ID) proteins are known to negatively regulate cell differentiation by interfering with basic helix-loop-helix transcription factors. It has been reported that in keratinocytes TGF-β upregulated the expression of cyclic AMP-dependent transcription factor-3, which served as a cofactor assisting the binding of Smad3/Smad4 complexes to the ID1 promoter; consequently, ID1 expression was downregulated and cell differentiation was promoted in vitro (41,42).

Although TGF-β signaling is widely known to mediate cell cycle arrest and enhance apoptosis in normal epithelium or in the early stage of tumor formation, it also induces epithelial cell overproliferation and inhibits apoptosis at a later stage of oncogenesis. For instance, Lu et al (27) demonstrated that tumor cells and epithelial cells from adjacent tissues expressed increased levels of TGF-β1, as compared with those in epithelial cells from normal control human tissues. However, in transgenic mice, overexpression of TGF-β1 was reported to result in the hyperproliferation of cells at the head and neck epithelium, and to enhance inflammation and angiogenesis (27). This suggested that TGF-β1 promoted cell proliferation by the formation of an extracellular microenvironment in favor of tumor formation, even at early stage of carcinogenesis (Fig. 1) (27). Although, TGF-β1 functions as a potent chemotactic molecule for leukocytes, it has been reported that inflammatory cytokines and growth factors secreted by infiltrated leukocytes may counteract the negative effects of TGF-β1 on the cell cycle (27). In fact, loss of TβRI or TβRII partly subverts TGF-β1-induced cell cycle arrest, and this effect along with the increased production of TGF-β1 may result in its accumulation in the extracellular microenvironment. The loss of TβRI or TβRII has also been proven to lead to increased cell proliferation and inhibit the apoptosis of HNSCC cells, respectively (29,30). Additionally, a previous study identified that improved proliferation and inhibited apoptosis due to decreased TβRI levels were alternatively associated with the activation of the PI3K/Akt signaling pathway (43).

TGF-β signaling disruptions are associated with poor prognosis partly due to the induction of epithelial-mesenchymal transition (EMT). EMT is a cellular process during which a cell with epithelial characteristics, such as cell polarity and cell-cell conjunction, translates to a cell with mesenchymal characteristics, such as motility (44). Decreased E-cadherin and increased vimentin levels are hallmarks of the EMT, while Snail and Twist are important factors negatively regulating E-cadherin (45). In OSCC, TGF-β signaling has been implicated in EMT through Snail and upregulation of matrix metalloprotease 9 (MMP-9) levels (46,47). Yu et al (48) demonstrated that TGF-β1 expression in HNSCC was correlated with decreased E-cadherin level through the phosphorylation of Smad2/3 and subsequent involvement of Smad4, which bound to the Snail promoter (49). Independently of Smads, TGF-β1 also regulates the Snail family proteins via the extracellular signal-regulated kinase (ERK)1/2 pathway in HNSCC (45). In addition, MMP-9 degrades the extracellular matrix components and basement membrane, and is regulated by TGF-β1 through Smad2/3 and myosin light chain kinase in human HNSCC cell lines (Fig. 1) (50). Notably, Sun et al (46) demonstrated a reciprocal interaction between MMP-9 and Snail regulation. MMP-9 induced EMT partly through the expression of Snail, while Snail was involved in TGF-β1-modulated MMP-9 expression by increasing Ets-1 (46). Another study indicated that TGF-β1 promoted MMP-9 expression by Slug (Snail2) (51). Furthermore, TGF-β1 may enhance EMT in cooperation with other growth factors in HNSCC. Compared with TGF-β1 or epithelial growth factor (EGF) alone, long-term co-stimulation with TGF-β1 and EGF in an OSCC cell culture model induced a phenotype transition, displaying upregulation of vimentin and downregulation of E-cadherin at the protein level, as well as markedly enhanced invasiveness (52). It was also reported that these observations may be associated with TGF-β1/EGF causing extracellular matrix remodeling by a plasmin/MMP-10/MMP-1-dependent collagen remodeling axis (52).

The concept of tumor immune surveillance was initially described in 1967 by Burnet (59). This refers to an organism discerning and eliminating cancer cells via the innate and adaptive immune system. At the same time, cancer cells are able to evade discernment and attack from the immune system, thus promoting tumor progression. Over the last 50 years, TGF-β has been demonstrated to abrogate tumor-suppressing immune cell functions and to support tumor-promoting functions in certain types of cancer (23,24,60). This section outlines how TGF-β signaling promotes tumor progression in HNSCC by affecting several types of immune cells.

Dendritic cells (DCs), first identified by Steinman in 1973 (61), are the most efficient antigen-presenting cells (61,62). Researchers have identified DC inactivation in HNSCC bearing hosts (63). Exogenous TGF-β has been demonstrated to hamper DC maturation by compromising the expression levels of major histocompatibility complex class II and costimulatory molecules (64). In addition, TGF-β immobilizes DCs in order to prevent the migration of DCs and tumor antigens to lymph nodes (17,65). Upon exposure to TGF-β, DCs facilitate immune tolerance and the differentiation of CD4⁺ T cells to T regulatory cells (Tregs) in the TME of HNSCC (66). These results suggested that TGF-β induced DC dysfunction via regulating the maturation and mobility of DCs in peripheral organs.

Previous studies have indicated that tumor-associated macrophages are mainly of the M2 phenotype and are positively correlated with the histopathological grades of HNSCC bearing hosts (67–69). Exposure to enhanced expression of TGF-β results in an M2 macrophage phenotype, which typically expressed CD206 (70). Mechanisms of TGF-β inducing M2 include the negative regulation of Toll-like receptor (TLR) signaling that causes induction of anti-tumor cytokines, such as tumor necrosis factor (TNF)-α, interleukin (IL)-12 and interferon (IFN)-γ, in order to participate in macrophage responses (71,72). Furthermore, Standiford et al (73) demonstrated that activated TGF-β signaling is required in IL receptor-associated kinase induction, which is a critically negative regulator of TLR signaling. Besides participating in macrophages polarization, TGF-β recruits macrophages to the TME, where they further produce TGF-β and thus a vicious cycle is formed (60).

Natural killer (NK) cells are frequently incompetent in HNSCC (74,75), and have been demonstrated to be suppressed by TGF-β, partly due to the stable expression of TβRII on NK cells (31,76). TGF-β binds to TβRII and then activates downstream components, resulting in the transcriptional repression of NK group 2 member D (NKG2D) tumor cell recognition receptors, in turn suppressing the cytotoxic effects of NK cells (76,77). Similarly, Klöss et al (78) reported that enhanced TGF-β1 plasma levels were correlated with a decreased NKG2D-dependent cytotoxic ability in relapsed HNSCC patients. Furthermore, Ghiringhelli et al (79) indicated that membrane-bound TGF-β expressed on Tregs negatively regulated NKG2D and the type I transmembrane protein NKp30 on NK cells, thus suppressing NK cell functions through Treg-NK cell interaction.

Myeloid-derived suppressor cells (MDSCs), first detected in cancer patients in 1984 (80), consist of a heterogeneous population of immature myeloid cells, including the precursors of DCs and macrophages. One of the most notable characteristics of MDSCs is that they suppress the activity of T cells (81,82). Previous studies have reported the presence of MDSCs in the peripheral blood of HNSCC patients, and that high infiltration of MDSCs promoted an immune-suppressive TME (83–85). Bian et al (30) demonstrated that decreased TGF-β signaling upregulated C-X-C motif chemokine ligand (CXCL)1, CXCL5, prostaglandin-endoperoxide synthase 2 and CXC receptor 3, contributing to the recruitment of MDSCs in an HNSCC model in TβRI/phosphatase and tensin homolog (PTEN) 2cKO mice. However, further research is required to support whether the recruitment of MDSCs in an HNSCC model is associated with a TGF-β signaling defect alone, or simultaneous TGF-β signaling defect and PTEN loss.

T lymphocytes can be classified into four subtypes, including cytotoxic T, helper T (Th), regulatory T (Treg) and memory T (Tm) cells. With the exception of Tm cells, all these subtypes have been verified to serve an important role in tumor escape from immune surveillance (86). Similarly to epithelial cells, TGF-β induces primary T lymphocyte growth inhibition in G1 phase by downregulating CDK4 (87). In addition, the trigger of the conversion of local primary T cells to Treg or Th cells in mice is partly associated with the transcription factor FoxP3 and retinoic acid receptor-related orphan receptor γt, respectively, whose transcription activity can be modulated by TGF-β (88,89).

Treg cells, a subtype of CD4⁺ T lymphocytes, suppress tumor immune reaction and contribute to the immune tolerance to tumor antigens. Research has verified an enrichment of Treg cells in the blood of HNSCC patients (90,91), and these cells are recruited and activated by TGF-β1 (92). Among them, the Treg type 1 cells produce TGF-β1 and IL-10, rather than having direct contact with responder cells to induce inhibition, and thus favor local immune suppression (92,93). TGF-β skews CD4⁺ T cells differentiating away from antitumor effector Th1 and towards the Th17 phenotype. Li et al (94) suggested that the proportion of Th17 cells in the peripheral blood increased with cancer progression and metastasis. This proportion was higher in the blood of HNSCC patients, with evidently enhanced levels observed in patients with advanced tumors and/or lymph node metastasis. Furthermore, Th17 cells are characterized by the production of IL-17, whose secretion is associated with TGF-β in HNSCC patients (94,95). IL-17 acts to induce tumor cell proliferation and, notably, enhances angiogenesis in immune-deficient hosts, contributing to a tumor friendly microenvironment (2,17). In a study by Laad et al (96), low CD8⁺ cytotoxic T lymphocytes (CTL) frequency was reported in the peripheral blood and tumor tissues of oral cancer patients. CTLs are often inactivated, while the expression levels of their effector products for cytolysis, namely IFN-γ, granzyme A, granzyme B, perforin and Fas ligand, are repressed by TGF-β (65). Taken together, TGF-β and its downstream signaling are able to interfere with CD4⁺ and CD8⁺ T cell differentiation and effector functions.