TGFβ signaling in head and neck squamous cell carcinoma

Abstract

Transforming growth factor beta (TGFβ) is a key regulator of epithelial cell proliferation, immune function and angiogenesis. Because TGFβ signaling maintains epithelial homeostasis, dysregulated TGFβ signaling is common in many malignancies, including head and neck squamous cell carcinoma (HNSCC). Defective TGFβ signaling in epithelial cells causes hyperproliferation, reduced apoptosis and increased genomic instability, and the compensatory increase in TGFβ production by tumor epithelial cells with TGFβ signaling defects further promotes tumor growth and metastases by increasing angiogenesis and inflammation in tumor stromal cells. Here, we review the mouse models that we used to study TGFβ signaling in HNSCC.

Head and neck squamous cell carcinoma (HNSCC) includes tumors of the nasal cavity, paranasal sinuses, oral cavity, pharynx and larynx and is the sixth most common cancer worldwide, with ~650 000 new cases and ~350 000 deaths each year (Parkin et al., 2005). Despite advances in surgery, radiotherapy and chemotherapy, there has been little improvement in HNSCC survival in the last 30 years (Edwards et al., 2010). HNSCC patients commonly present with locally advanced disease involving both vital structures and regional lymph nodes and ~10% of patients present with distant metastases. Even with successful local control, ~50% of the patients relapse with local or distant disease progression (Argiris et al., 2008).

The major risk factors for HNSCC are alcohol, tobacco and betel (areca) nut use (Curado and Hashibe, 2009). While tobacco use increases the HNSCC risk 3–9-fold, the combination of alcohol and tobacco use synergistically increases the HNSCC risk ~100-fold (Hecht, 2003). In addition, human papillomavirus causes a subset of HNSCCs located primarily in the hypopharynx (Licitra et al., 2006). A number of syndromes, including Fanconi’s anemia, xeroderma pigmentosum, ataxia telangiectasia and Bloom syndrome, also increase the HNSCC risk and are characterized by DNA repair defects and genomic instability (Prime et al., 2001).

HNSCC results from an accumulation of genetic and epigenetic aberrations affecting numerous cellular processes, including proliferation, apoptosis, inflammation, angiogenesis and DNA repair (Forastiere et al., 2001; Mao et al., 2004). Genetic alterations at 17p13 (encoding p53) and 9p21 (encoding p14 and p16) have been reported in HNSCC (McCaul et al., 2002; Hunter et al., 2005) and both epidermal growth factor receptor (EGFR)/signal transducer and activator of transcription 3 (STAT3) and PI3 kinase/PTEN/Akt pathways are activated in HNSCC. EGFR regulates cell growth, migration and survival. EGFR is overexpressed in 80–90% of HNSCCs (Grandis and Tweardy, 1993) and signals through the transcription factor, STAT3 (Grandis et al., 2000). EGFR overexpression correlates with aggressive tumor behavior and poor clinical outcome (Shin et al., 1994; van Oijen et al., 1998) and anti-EGFR therapy (cetuximab) is used to treat HNSCC (Karamouzis et al., 2007). In the Western world, Ras mutations occur in 5–10% of HNSCCs (Chang et al., 1991; Lu et al., 2006), but Ras overexpression occurs in the majority of HNSCCs (Lu et al., 2006). In contrast, Ras mutations occur in ~35% of oral cancers associated with chewing tobacco in southern Asia (Saranath et al., 1991).

Mouse models of skin and head and neck cancer demonstrate the importance of transforming growth factor (TGF)β signaling in both tumor initiation and progression (Cui et al., 1996; Li et al., 2004; Yang et al., 2005; Lu et al., 2006; Qiao et al., 2006; Bornstein et al., 2009). As TGFβ signaling disruption typically causes embryonic lethality, conditional, tissue-specific genetic manipulation is required for genetic studies of TGFβ in vivo. Common approaches include tetracycline to de-repress or activate the tet operon (tet-off or tet-on), tamoxifen to activate an estrogen receptor fusion protein or RU486 to activate a progesterone receptor fusion protein (Lewandoski, 2001; Jonkers and Berns, 2002) (Figure 1). Targeting is accomplished by placing the transactivator or inducible Cre recombinase downstream of a keratin promoter expressed in basal keratinocytes of stratified epithelial tissues (Byrne et al., 1994; Wang et al., 1997b). By placing a floxed stop codon upstream of a target (for example, Kras^G12D), this system can also be used to ‘knock-in’ oncogenes (Jackson et al., 2001).

Figure 1.

Conditional mouse models facilitate tissue-specific manipulation of TGFβ signaling. (a) The inducible, epithelial PR gene-switch system requires transactivator and target transgenes. The transactivator is a fusion protein containing a Gal4 DNA-binding domain, an RU486-responsive PR domain and the NF-κB p65 transactivation domain. The transactivator is under the control of a keratin promoter that restricts expression to basal keratinocytes. The target consists of Gal4 binding sites and a minimal tata promoter upstream of the gene of interest. Mice with the transactivator are crossed with animals harboring the target. Local RU486 treatment causes translocation of the transactivator into the nucleus, where it binds the Gal4 sites and induces expression of the target gene. (b) The inducible knockout system uses Cre/LoxP technology, in which target genes or sequences are flanked by LoxP sites (‘floxed’) that can be excised by Cre recombinase. Tissue specificity is achieved by placing the Cre recombinase under the control of a keratin promoter; temporal control is achieved by the RU486-inducible Cre recombinase–PR fusion protein. RU486 causes nuclear translocation and activation of the Cre recombinase, which then excises floxed sequences.

Although single genetic alterations do not typically induce tumor formation, chemical carcinogenesis can be used to determine whether genetic modifications affect susceptibility to tumor development. For example, a single dose of 7,12-dimethylbenz[a]anthracene to the skin or oral epithelium induces tumor-initiating H-Ras mutations (Balmain and Pragnell, 1983). Other carcinogens such as 4-nitroquinoline-1 oxide have been used to induce HNSCC through DNA adduct formation and H-ras mutations (Hawkins et al., 1994; Yuan et al., 1994). In this review, we will discuss the roles of individual TGFβ signaling components in HNSCC development and progression and highlight what we have learned about these key regulators of epithelial homeostasis from both human HNSCC samples and mouse models of HNSCC.

TGFβ signaling components

The TGFβ superfamily includes the TGFβ ligands (TGFβ1, 2 and 3), bone morphogenic proteins (BMPs) and activins/inhibins (Shi and Massague, 2003). TGFβ ligands signal through two types of transmembrane serine/threonine kinase receptors: TGFβ receptor I (TGFβRI), also known as activin receptor-like kinase 5, and TGFβ receptor II (TGFβRII) (Figure 2) A total of five TGFβRII and seven TGFβRI family members have been described (Massague, 2008). Canonical TGFβ signaling is mediated through Smad family members; there are eight Smads divided into three groups: receptor-activated Smads (R-Smads), inhibitory Smads and the common Smad (Smad4) (Massague, 2008). After ligand binding, TGFβRII phosphorylates TGFβRI, which then phosphorylates R-Smads. TGFβ ligands signal through R-Smads 2 and 3, while BMPs employ R-Smads 1, 5 and 8 (Massague, 2000). Phosphorylated R-Smads heterotrimize with Smad4, translocate to the nucleus and regulate gene expression at Smad binding elements (Massague et al., 2005). Interactions between Smad complexes and other transcription factors, co-activators and co-repressors provide additional tissue and gene specificity (Massague, 2000). Inhibitory Smads attenuate TGFβ signaling by competing with R-Smads for binding at TGFβRI and by recruiting ubiquitin ligases to degrade TGFβRI and R-Smads (Ebisawa et al., 2001; Massague et al., 2005). In non-canonical (Smad-independent) TGFβ signaling, ligand-bound TGFβ receptors activate other signaling pathways such as the MAP kinases, phosphatidylinositol-3-kinase/AKT and Rho GTPase, which likely enhance tumor growth after disruption of canonical TGFβ-Smad signaling (Zhang, 2009).

Figure 2.

TGFβ/BMP signaling regulates tissue homeostasis. TGFβ and BMPs signal through different R-Smads. Phosphorylated R-Smads heterotrimerize with Smad4, translocate into the nucleus and mediate gene transcription by binding Smad binding elements (SBEs) in the promoters of TGFβ and BMP responsive genes. Interactions with transcriptional coactivators and corepressors provide an additional layer of regulation.

TGFβ as a regulator of epithelial homeostasis

TGFβ signaling regulates tissue homeostasis by modulating cell growth, differentiation, apoptosis, migration, inflammation and angiogenesis (Ten Dijke et al., 2002; Bertolino et al., 2005; Pardali et al., 2005; Massague, 2008). In epithelial cells, TGFβ inhibits proliferation and promotes both apoptosis and differentiation (Massague, 2008). By inducing expression of the cell cycle regulators, p15Ink4b and p21Cip1, TGFβ inhibits cell cycle progression (Seoane et al., 2004; Gomis et al., 2006). In addition, TGFβ downregulates c-Myc expression through interactions between Smads and transcriptional co-repressors (Chen et al., 2002). TGFβ promotes differentiation by downregulating inhibitor of differentiation/DNA binding proteins (Kang et al., 2003). TGFβ promotes apoptosis through both Smad-dependent and -independent mechanisms (Yanagisawa et al., 1998; Pardali and Moustakas, 2007). TGFβRI directly interacts with the E3 ubiquitin ligase, TRAF6, causing apoptosis through TAK1-p38/ JNK (Sorrentino et al., 2008; Yamashita et al., 2008), while Smads activate apoptosis through activation of ATM, p53 and BIM, and repression of AKT (Zhang et al., 2006; Wang et al., 2008). Given the numerous actions of TGFβ on epithelial homeostasis, it is not surprising that TGFβ signaling disruption is common in many cancers, including HNSCC (Levy and Hill, 2006). Studies in mouse models and human HNSCC samples illustrate that different TGFβ signaling disruptions promote epithelial carcinogenesis through both unique and overlapping mechanisms.

Defective TGFβ signaling in HNSCC

While downregulation of TGFβRII, Smad4 and Smad2 is common in human HNSCCs, alterations of TGFβRI and Smad3 are relatively rare (Chen et al., 2001; Muro-Cacho et al., 2001), suggesting discrete roles for these molecules in carcinogenesis. In contrast, TGFβ1 ligand overexpression is common in human HNSCC and is likely related to defective TGFβ signaling in tumor epithelial cells (Lu et al., 2004). As summarized in Table 1, we have generated a number of mouse models to illustrate how these TGFβ signaling defects promote HNSCC development in vivo.

Table 1.

Mouse models used to study TGFβ in stratified epithelia

Gene	Sites	Initiator	Phenotype	References
TGFβ1	Conditional overexpression in oral epithelium	None or DMBA/TPA	Epithelial hyperplasia Inflammation Increased angiogenesis	Weeks et al. (2001); Lu et al. (2004)
TGFβRII	Conditional deletion in oral epithelium	DMBA or KrasG12D mutation	HNSCC formation Increased TGFβ ligand Increased inflammation Increased angiogenesis	Lu et al. (2006)
Smad2	Conditional deletion in skin	Topical DMBA/TPA	Increased tumor kinetics Increased EMT: Snail activation and loss of E-cadherin	Hoot et al. (2008)
Smad3	Germline deletion	Topical DMBA/TPA	Reduced papillomas; no SCCs Increased apoptosis Reduced leukocyte and macrophage infiltration	Li et al. (2004)
Smad4	Conditional deletion in oral epithelium or skin	None	Spontaneous SCC formation Increased genomic instability Increased TGFβ ligand Increased inflammation	Qiao et al. (2006); Bornstein et al. (2009); Owens et al. (2010)
Smad3/Smad4	Conditional deletion in oral epithelium	None	Abrogation of Smad4 loss-induced inflammation	Bornstein et al. (2009)

Abbreviations: DMBA, 7,12-dimethylbenz[a]anthracene; EMT, epithelial-to-mesenchymal transition; HNSCC, head and neck squamous cell carcinoma; SCC, squamous cell carcinoma; TGF, transforming growth factor.

The role of the TGFβ1 ligand

While canonical TGFβ1 signaling inhibits epithelial proliferation, TGFβ overexpression promotes tumorigenesis through paracrine effects on the tumor stroma leading to increased inflammation and angiogenesis (Lu et al., 2004). This dual role of TGFβ1 was demonstrated in mouse models with conditional epithelial TGFβ1 over-expression (Cui et al., 1996; Weeks et al., 2001; Lu et al., 2004). While TGFβ1 overexpression initially inhibits the growth of chemically induced skin tumors, it paradoxically increases the malignant conversion of papillomas to carcinomas (Cui et al., 1996). Similarly, TGFβ1 overexpression early in skin carcinogenesis inhibits tumor formation, while overexpression in established tumors increases metastasis (Weeks et al., 2001).

Increased TGFβ1 expression is seen in ~80% of human HNSCCs and correlates with more advanced disease and reduced survival (Lu et al., 2004; Levy and Hill, 2006). Non-malignant tissue adjacent to HNSCCs also frequently exhibits increased TGFβ1, suggesting that TGFβ1 overexpression is an early event in HNSCC development (Lu et al., 2004). To study TGFβ1 over-expression, our lab developed a conditional gene switch model in which TGFβ1 overexpression in oral keratinocytes can be induced by RU486 (Lu et al., 2004) (Figure 1a). Although TGFβ1 expression at levels similar to those seen in human HNSCC did not cause tumor formation, these mice exhibit increased angiogenesis and inflammation in the stroma, which subsequently caused marked epithelial hyperplasia (Lu et al., 2004). In this model, increased angiogenesis in the oral mucosa was mediated by the endothelial TGFβRI receptor, ALK1, and TGFβ-dependent inflammation was comprised of CD4 + T cells, granulocytes and macrophages, likely stimulated by increased interleukin-1β and tumor necrosis factor-α (Lu et al., 2004).

TGFβ1 can be either an immunosuppressor or a pro-inflammatory cytokine (Teicher, 2007; Wrzesinski et al., 2007). TGFβ1 suppresses tumor immunity by suppressing the activity of dendritic cells and natural killer cells that mediate tumor immunity but are functionally inactivated in HNSCC (Wrzesinski et al., 2007; Moutsopoulos et al., 2008). TGFβ1 also recruits and activates T regulatory cells that suppress tumor immune responses and are elevated in HNSCC patients (Moutsopoulos et al., 2008; Boucek et al., 2009). In contrast, TGFβ1 is important for T-cell differentiation and the maturation and activation of Th17 cells (van Vlasselaer et al., 1992; Mangan et al., 2006; Veldhoen et al., 2006) and contributes to the chronic inflammation associated with carcinogenesis by attracting both neutrophils and monocytes (Bierie and Moses, 2006b). In summary, TGFβ1 inhibits anti-tumor immunity while activating tumor-promoting inflammation. Loss of functional TGFβ signaling causes a compensatory increase in TGFβ1 production that paradoxically stimulates tumor growth by affecting the tumor stroma, increasing invasion, inflammation and angiogenesis (Lu et al., 2006; Bian et al., 2009; Bornstein et al., 2009).

The role of TGFβRII

Decreased expression of TGFβRII has been well documented in a variety of malignancies, including HNSCC, and occurs as a result of genetic alteration or epigenetic silencing (Wang et al., 1997a; Muro-Cacho et al., 1999). In oral squamous cell carcinomas (SCCs) TGFβRII mutation occurs in 21% of tumors and reduced copy number is seen in 40–50% of tumors (Snijders et al., 2005; Sparano et al., 2006). Studies in HNSCC cell lines have also found TGFβRII down-regulation by both promoter methylation and promoter mutation (Garrigue-Antar et al., 1995; Worsham et al., 2006). In sum, TGFβRII expression may be reduced in >70% of HNSCCs (Lu et al., 2006) and is associated with reduced tumor differentiation and more aggressive clinical behavior (Muro-Cacho et al., 1999; Fukai et al., 2003).

To determine whether TGFβRII loss directly caused HNSCC, we developed a mouse model with conditional TGFβRII deletion in oral keratinocytes (Muro-Cacho et al., 2001; Lu et al., 2006). While TGFβRII deletion alone did not result in spontaneous tumor formation, when combined with a tumor-initiating Ras mutation, animals developed full-penetrance HNSCC (Lu et al., 2006). These tumors were pathologically similar to human HNSCC and harbor many of the same molecular aberrations as human HNSCC, including reduced expression of the cyclin-dependent kinase inhibitors, p15 and p21, and increased expression of c-myc, cyclin-D and EGFR (Lu et al., 2006). This model demonstrates that while TGFβRII loss does not initiate tumor formation, it promotes malignant progression of Ras-initiated tumors (Lu et al., 2006). This cooperation between TGFβRII and Ras activation has subsequently been supported in mouse models of TGFβRII-deleted anogenital tumors in which keratinocyte-specific TGFβRII deletion causes spontaneous tumor formation in the transitional epithelial zone between the mucosa and external squamous epithelium, a region with increased Ras signaling (Guasch et al., 2007). TGFβRII-null HNSCC had increased inflammation and angiogenesis mediated by increased TGFβ production, illustrating the carcinogenic role of TGFβ in the tumor microenvironment after loss of TGFβ signaling competency in tumor epithelial cells (Lu et al., 2006).

The role of TGFβRI

Although genetic alteration of TGFβRI by mutation or deletion is rare in human HNSCC, transcriptional repression of TGFβRI by promoter hypermethylation has been described (Kang et al., 1999) and reduced TGFβRI immunostaining is associated with increased invasion and metastasis in human esophageal SCC (Fukai et al., 2003). In HNSCC, less is known about the relationship between TGFβRI loss and clinical tumor behavior; however, TGFβRI mutations have been detected in 19% of HNSCC metastases (Chen et al., 2001). When TGFβRI is deleted using a neurofilament promoter, mice develop SCCs in both periorbital and perianal areas (Honjo et al., 2007). However, like TGFβRII, TGFβRI deletion in oral epithelia does not initiate tumor formation but allows progression to HNSCC after oral treatment with the Ras mutagen, 7,12-dimethylbenz[a]anthracene (Bian et al., 2009). These tumors have increased TGFβ1 expression and activated phosphatidylinositol-3-kinase/Akt signaling, as is frequently seen after disruption of canonical TGFβ signaling (Bian et al., 2009).

The role of Smad4

Smad4 is a potent tumor suppressor in the oral cavity

Smad4 was initially identified as a tumor suppressor in the pancreas (Hahn et al., 1996) and was later characterized as a key mediator of TGFβ signaling (Zhang et al., 1996). As Smad4 mediates both TGFβ and BMP signaling, Smad4 loss abrogates signaling of both pathways. Smad4 loss or inactivation is common in many malignancies, including HNSCC (Bierie and Moses, 2006a). Loss of chromosome 18q, a region encoding both Smad2 and Smad4, is common in human HNSCC (Kim et al., 1996; Papadimitrakopoulou et al., 1998; Takebayashi et al., 2000; Snijders et al., 2005) and loss of heterozygosity at the Smad4 locus also occurs in ~50% of human HNSCCs (Kim et al., 1996; Takebayashi et al., 2000). Although loss of 18q is associated with reduced survival in human HNSCC (Pearlstein et al., 1998), it is unknown whether this is related strictly to Smad4 loss, the simultaneous loss of both Smad2 and Smad4, or loss of other genes within this region. Somatic inactivation of Smad4 has been reported in 22–40% of human HNSCCs (Xie et al., 2003; Iamaroon et al., 2006). Although point mutations are relatively infrequent in human HNSCCs (Kim et al., 1996), loss of a single Smad4 allele occurs in 30–50% of human HNSCCs (Kim et al., 1996; Takebayashi et al., 2000; Bornstein et al., 2009). In cases where a single copy of Smad4 is lost, additional mutation on the other allele is rare (Kim et al., 1996). However, studies using Smad4 knockout mouse models have shown that loss of a single copy of the Smad4 gene confers a 50% reduction in the Smad4 mRNA and protein levels (Sirard et al., 1998; Yang et al., 1998), and this increases HNSCC susceptibility (Bornstein et al., 2009). We examined Smad4 expression in human HNSCC, and found that 86% of tumors and 67% of adjacent non-malignant mucosa had >50% Smad4 reduction, suggesting that Smad4 down-regulation is an early event in HNSCC development (Bornstein et al., 2009). Although we found loss of heterozygosity at the Smad4 locus in 33% of HNSCCs, the higher frequency of Smad4 loss at the mRNA and protein levels suggests that other epigenetic, posttranscriptional or posttranslational modifications contribute to reduced Smad4 expression in HNSCC (Bornstein et al., 2009).

As Smad4 loss or inactivation is common in tumorigenesis, a number of mouse models have been created to assess the role of Smad4. In combination with other oncogenic alterations, Smad4 deletion causes tumors in the colon (Takaku et al., 1998; Kitamura et al., 2007), pancreas (Bardeesy et al., 2006; Izeradjene et al., 2007), forestomach (Teng et al., 2006) and liver (Xu et al., 2006), indicating that Smad4 loss promotes tumor progression. In addition, Smad4 deletion causes spontaneous tumors in the stomach (Xu et al., 2000), skin (Yang et al., 2005; Qiao et al., 2006) and mammary gland (Li et al., 2003), demonstrating that Smad4 can also initiate tumor formation in certain tissues. Epidermal Smad4 deletion blocks growth inhibition by TGFβ, leading first to hyperproliferation and subsequently to skin SCC (Yang et al., 2005; Qiao et al., 2006). We developed a conditional mouse model to delete Smad4 in oral keratinocytes and found that these mice developed HNSCC starting at 29 weeks of age (Bornstein et al., 2009). These tumors resembled human HNSCC histologically and also metastasized to regional lymph nodes (Bornstein et al., 2009). As oral epithelial TGFβRII deletion required an initiating event (for example, Ras activation), we investigated whether spontaneous Ras activation was responsible for tumor initiation after Smad4 deletion and found that while only a small number of tumors exhibited Ras mutations, most of the Smad4-deficient tumors had increased Ras protein levels (Bornstein et al., 2009), suggesting that Ras activation has a role in tumor initiation after Smad4 deletion.

Smad4 loss increases genomic instability by inhibiting DNA repair

As cancers are usually the result of an accumulation of oncogenic changes, we hypothesized that Smad4 deletion might affect genetic stability and facilitate acquisition of additional genetic alterations. Indeed, Smad4 knockout tumors had increased centrosome numbers and reduced expression of several Brca/ Fanc family members responsible for double-stranded DNA repair (Bornstein et al., 2009). This was particularly interesting because germline mutations in the Fanc/Brca pathway lead to Fanconi anemia in humans, a syndrome characterized by bone marrow failure and markedly increased susceptibility to HNSCC (Kutler et al., 2003). Other studies have shown downregulation of Fanc/Brca genes in sporadic HNSCC (Marsit et al., 2004; Sparano et al., 2006; Weber et al., 2007; Wreesmann et al., 2007) and mice with epithelial-specific deletion of Brca1 develop oral SCCs (Berton et al., 2003), suggesting that defective DNA repair contributes greatly to Smad4 loss-induced HNSCC.

Smad4 loss increases inflammation

Inflammation associated with Smad4 loss is important for tumorigenesis. Smad4-deficient intestinal cells recruit immature myeloid cells expressing MMP2, MMP9 and the chemokine receptor CCR1 to the tumor invasion front (Kitamura et al., 2007), and Smad4 deletion in T cells causes intestinal tumors, indicating that Smad4 loss in both epithelial and inflammatory cells can promote cancer development (Kim et al., 2006). Smad4 loss in the oral mucosa causes leukocyte infiltration of macrophages, granulocytes, T lymphocytes and proinflammatory Th17 cells prior to tumor formation and this is likely mediated by increased TGFβ1 expression (Bornstein et al., 2009). Indeed, these mice have significantly increased TGFβ1 ligand in both the HNSCC and the adjacent mucosa within 1 week of Smad4 deletion (Bornstein et al., 2009). Consistent with reports that abrogation of TGFβ signaling and elevated TGFβ1 increases tumor invasion and migration via increased BMP signaling through Smad1 and Smad5 (Bharathy et al., 2008; Liu et al., 2009), Smad4-deleted oral epithelia also has augmented Smads 1/5/8 (Bornstein et al., 2009).

The role of Smad3

Smad3 loss blocks inflammation induced by Smad4 loss

While reduced expression of Smad2 and Smad4 is common in human cancer, this is not the case for Smad3 (Sjoblom et al., 2006; Bierie and Moses, 2006b). In fact, Smad3 knockout mice are resistant to chemical skin carcinogenesis and exhibit both decreased hyperplasia and proliferation, and increased apoptosis (Li et al., 2004; Tannehill-Gregg et al., 2004). Interestingly, chemically induced skin tumors generated in Smad3 knockout animals have reduced inflammation and fewer tumor-associated macrophages (Li et al., 2004; Tannehill-Gregg et al., 2004). We observed increased Smad3 expression in Smad4-loss induced HNSCC and that Smad3 haploinsufficiency attenuated inflammation resulting from Smad4 deletion in the oral cavity (Bornstein et al., 2009). However, further studies are needed to determine whether Smad3 loss attenuates HNSCC formation after Smad4 deletion.

The role of Smad2

Smad2 lies near Smad4 on chromosome 18 and is frequently lost in HNSCCs (Papadimitrakopoulou et al., 1998; Takebayashi et al., 2000; Snijders et al., 2005). Complete loss of the Smad2 protein in HNSCC ranges from 1 to 38% (Muro-Cacho et al., 2001; Xie et al., 2003; Iamaroon et al., 2006) and it is more commonly lost in poorly differentiated tumors (Muro-Cacho et al., 2001). However, similar to Smad4, mice with a single-allele deletion of Smad2 showed an ~50% Smad2 protein reduction and exhibited haploinsufficiency, that is, increased susceptibility to SCC formation (Hoot et al., 2008). We found that 94% of human skin SCC had reduced Smad2 mRNA expression and 70% had a >50% reduction in Smad2 protein by immunostaining (Hoot et al., 2008). Among these tumors, 67% had Smad2 loss of heterozygosity (Hoot et al., 2008). In HNSCC, reduced phospho-Smad2 immunostaining is associated with lymph node and distal metastasis and a worse prognosis (Muro-Cacho et al., 2001). While Smad2 deletion in the oral cavity or skin does not cause spontaneous tumor formation (Hoot et al., 2008), mice heterozygous for Smad2 have increased susceptibility to chemical skin carcinogenesis and develop more poorly differentiated SCCs than wild-type mice (Tannehill-Gregg et al., 2004; Hoot et al., 2008).

Smad2 loss promotes epithelial-to-mesenchymal transition

Chemically induced SCCs in keratinocyte-specific Smad2 knockout mice were less differentiated and exhibited increased epithelial-to-mesenchymal transition (EMT) compared with chemically induced tumors generated in wild-type mice (Hoot et al., 2008). Interestingly, both tumors and premalignant skin from Smad2 knockout mice have markedly reduced E-cadherin expression secondary to overexpression of Snail family transcriptional repressors (Hoot et al., 2008). TGFβ functions as a well-known regulator of EMT through both Smad-dependent and -independent mechanisms (Zavadil and Bottinger, 2005). Unlike Smad4 or TGFβRII deletion, Smad2 deletion did not increase TGFβ ligand expression (Hoot et al., 2008). Hence, increased EMT after Smad2 deletion could not be explained by Smad-independent TGFβ signaling (Hoot et al., 2008). In fact, Smad2 loss increased Smad4 binding at and transcription from the Snail promoter that presumably promoted EMT (Hoot et al., 2008) (Figure 3b). Interestingly, late-stage Smad2-deleted tumors also had reduced Smad4 expression that could have potentially increased EMT through increased TGFβ1 elaboration and Smad-independent TGFβ signaling (Hoot et al., 2008).

Figure 3.

Differential roles of Smad4 and Smad2 in epithelial carcinogenesis in vivo. (a) Smad4 loss inhibits Fanc/Brca-mediated DNA repair, causing increased genomic instability, leading to increased proliferation and reduced apoptosis in epithelial cells. Additionally, Smad4 deletion leads to a compensatory increase in TGFβ ligand expression, causing inflammation and angiogenesis in the surrounding tumor stroma. (b) Smad2 loss increases EMT. In the absence of Smad2, increased Smad3/4 binding at the Snail promoter increases Snail transcription and causes EMT.

Interactions between TGFβ and other signaling pathways in HNSCC

Although HNSCC models illustrate the importance of TGFβ signaling, the phenotypes of these animals cannot be explained solely by loss of TGFβ-mediated growth inhibition. A recurring theme in HNSCC animal models is Ras activation. While Ras activation in the oral mucosa causes hyperproliferation and benign papilloma formation (Lu et al., 2006), concomitant TGFβRII deletion causes degeneration of these benign lesions into carcinomas; hence Ras activation appears to be required for tumor initiation in a TGFβRII-null background (Lu et al., 2006; Guasch et al., 2007). Interestingly, although Smad4 deletion in the oral mucosa causes spontaneous HNSCCs, these tumors also exhibit Ras activation, suggesting that Ras and TGFβ pathways cooperate in HNSCC development (Bornstein et al., 2009). Activated EGFR-STAT3 signaling is also common in HNSCC (Grandis and Tweardy, 1993; Song and Grandis, 2000) and, similar to their human counterparts, TGFβRII knockout HNSCCs have increased EGFR expression and increased pSTAT immunostaining, highlighting the activated EGFR-STAT3 in these tumors (Lu et al., 2006). Cancerous inhibitor of PP2A stabilizes c-myc and causes malignant transformation because it is transcriptionally repressed by TGFβ signaling; cancerous inhibitor of PP2A is also overexpressed in TGFβRII-null tumors (Junttila et al., 2007).

TGFβ-mediated stimulation of proinflammatory pathways also creates a tumor microenvironment conducive to HNSCC development. For example, either TGFβRII deletion in the oral mucosa or TGFβ1 overexpression by the oral mucosa causes NF-κB activation in the adjacent stroma (Lu et al., 2004; Cohen et al., 2009). Interestingly, there is a direct correlation between p53 status and TGFβRII expression; cell lines with mutant p53 have reduced TGFβRII expression, which can be reversed by restoring wild-type p53 expression (Cohen et al., 2009). These studies highlight the interactions between TGFβ and other signaling pathways in HNSCC formation.

Summary

Defects in TGFβ signaling promote tumor development by modulating a variety of processes (proliferation, apoptosis and DNA repair). In addition, compensatory TGFβ1 overexpression by epithelial cells with defects in TGFβ signaling can further augment tumor growth and metastases by creating a supportive microenvironment (Figure 3a). In human cancers, reduced expression of TGFβRII, Smad2 and Smad4 is common, occurs by a variety of mechanisms, and is usually associated with more aggressive tumor behavior and a worse clinical outcome. In mouse models, loss of TGFβRII or Smad4 abrogates TGFβ-mediated tumor suppression but also causes a compensatory increase in TGFβ ligand expression that promotes inflammation and angiogenesis in the tumor microenvironment (Figure 3a). Although they share some common features, loss of TGFβRII and Smad4 are not equivalent in HNSCC development. TGFβRII loss promotes malignant conversion but cannot initiate HNSCC development, while Smad4 loss initiates and promotes tumor formation, presumably through inhibition of DNA repair mediated by the Brca/ Fanc family. Smad2 deletion does not cause spontaneous HNSCC formation, but causes increased EMT through hyperactivation of Smad4-mediated Snail activation. In contrast to the tumor-suppressor activities of other TGFβ signaling components, Smad3 appears to have an important role in TGFβ-induced inflammation. Well-defined mouse models have helped elucidate how specific TGFβ signaling defects promote HNSCC growth and development and will likely be pivotal in developing biomarkers and testing targeted therapies for HNSCC in the future.