Transforming growth factor-β signaling in cancer invasion and metastasis

Abstract

Transforming growth factor-β (TGF-β) family members are polypeptides with dual tumor suppressive and oncogenic effects. They signal through serine/threonine kinase receptor complexes, which phosphorylate cytoplasmic mediators, the Smads. Upon phosphorylation, Smads translocate to the nucleus and associate with transcriptional coactivators or corepressors, and regulate the transcriptional activation of various TGF-β responsive genes. In addition, TGF-β activates cellular mitogen-activated protein kinase signaling pathways, which crosstalk with Smad signaling and regulate growth, survival and motility of cells. During tumorigenesis, malignantly transformed cells often lose the response to the tumor suppressive effects of TGF-β, which, in turn, starts to act as an autocrine tumor promoting factor by enhancing cancer invasion and metastasis. In this review, we summarize current view on the role of TGF-β signaling in tumorigenesis, with emphasis on cancer invasion and metastasis. On the basis of these recent observations, we discuss new therapeutic strategies targeting TGF-β signaling at distinct levels as a basis for inhibiting tumor growth, angiogenesis, invasion and metastasis. © 2007 Wiley-Liss, Inc.

The transforming growth factor-β (TGF-β) superfamily of growth factors consists of more than 35 structurally related secreted polypeptides, including TGF-βs, activins and bone morphogenetic proteins (BMPs) (Table I). Members of the TGF-β family have diverse roles in regulating cellular functions, such as cell proliferation, differentiation, migration and extracellular matrix (ECM) deposition.1 TGF-β was originally discovered in cell culture media of sarcoma virus transformed mouse fibroblasts as an activity that induced anchorage-independent growth of nonmalignant cells.2 Later studies, however, showed that TGF-β is a potent tumor suppressor and inhibitor of cell proliferation.3 Because of the pleiotropic activities of TGF-β, dysregulation of TGF-β signaling has been implicated in pathological conditions, including cancer.

Table I. Members of the TGF-β Superfamily

TGF-βs	BMPs	Activins and inhibins	Others
AMH, anti-Müllerian hormone; BMP, bone morphogenetic protein; CDMP, cartilage-derived morphogenetic protein; GDF, growth differentiation factor; MIS, Müllerian inhibiting substance; OP, osteogenic protein.
TGF-β1	BMP-2	Activin A	AMH/MIS
TGF-β2	BMP-3/Osteogenin	Activin B	Lefty A
TGF-β3	BMP-3b/GDF-10	Activin AB	Lefty B
	BMP-4	Activin C
	BMP-5	Activin E
	BMP-6	Inhibin A
	BMP-7/OP-1	Inhibin B
	BMP-8a/OP-2	Inhibin C
	BMP-8b/OP-3
	BMP-9/GDF-2
	BMP-10
	BMP-11/GDF-11
	BMP-12/GDF-7/CDMP-3
	BMP-13/GDF-6/CDMP-2
	BMP-14/GDF-5/CDMP-1
	BMP-15/GDF-9b
	BMP-16/Nodal
	GDF-1
	GDF-3
	GDF-8 (Myostatin)
	GDF-9
	GDF-15

Three TGF-β isoforms, TGF-β1, TGF-β2 and TGF-β3, are ubiquitously expressed in mammalian tissues. All isoforms are secreted as biologically inactive precursors called latent TGF-βs (L-TGF-β). Cleavage or conformational change of the precursor protein latency-associated peptide from the L-TGF-β results in formation of biologically active TGF-β. Under physiological conditions, TGF-βs are activated by acidic cellular microenvironment and proteolysis, for instance by plasmin, and matrix metalloproteinases (MMPs).4 After activation, TGF-βs bind to their signaling receptors and initiate intracellular signaling via Smad and mitogen-activated protein kinase (MAPK) pathways (Fig. 1).

Figure 1. Major signaling pathways activated by TGF-β. TGF-β signals primarily via Smad and MAPK pathways. Smad and MAPK pathways are not distinct, but they functionally interact with each other. Co-Smad, common mediator Smad; ERK, extracellular signal-regulated kinase; I-Smad, inhibitory Smad; JNK, c-Jun N-terminal kinase; MEK, MAPK/ERK kinase; MKK, MAPK kinase; R-Smad, receptor-activated Smad; TAK, TGF-β-activated kinase.

TGF-β signal transduction

TGF-β receptors

There are 3 types of TGF-β receptors: type I, type II and type III receptors. Among them, the type I and type II receptors are the signaling receptors which form heteromeric complexes and initiate intracellular signaling following ligand binding. There are 7 different mammalian type I and 5 different type II receptors (Table II).5 The so-called accessory TGF-β type III receptors include betaglycan and endoglin (CD105).6 Recently, members of the glycosylphosphatidylinositol-anchored protein DRAGON family have been identified as coreceptors for BMPs.7, 8

Table II. TGF-β Signaling Receptors and their Ligands

	Ligand
ActR, activin receptor; ALK, activin-receptor like kinase; AMH, anti-Müllerian hormone; AMHR, AMH receptor; BMP, bone morphogenetic protein; BMPR, BMP receptor; TβR, TGF-β receptor; TSR, TGF-β superfamily receptor.
Type II receptors
TβRII	TGF-βs
ActRIIA	Activins, BMP-2, -4, -7
ActRIIB	Activins, BMP-2, -4, -7
BMPRII	BMPs
AMHR	AMH
Type I receptors
TβRI, ALK-5	TGF-βs, myostatin, GDF-11
TSR-1, ALK-1	Activins, TGF-βs, BMP-9, -10
ActRIB, ALK-4	Activins, myostatin
ActRI, ALK-2	BMPs
BMPRIA, ALK-3	BMPs
BMPRIB, ALK-6	BMPs
ALK-7	Activins, nodal

TGF-β type I and type II receptors are very similar transmembrane glycoproteins, with a glycosylated extracellular domain, a short transmembrane domain, and an intracellular serine/threonine kinase domain.5 Type II receptors are constitutively autophosphorylated on various serines. Type I receptors are characterized by a conserved region rich in glycine and serine (GS-region), which precedes the kinase domain. For initiating signal transduction, formation of a heteromeric complex between type I and II receptors is required. TGF-β binds type II receptors, which subsequently interact with type I receptor. This leads to crossphosphorylation of the GS domain in type I receptor by the type II receptor kinase, which results in activation of type I receptor kinase domain and initiation of downstream signaling and phosphorylation of the Smad signaling mediators (Fig. 2).5, 9

Figure 2. TGF-β signaling via the Smad pathway. Ligand binding to the type II receptor results in subsequent oligomerization of TGF-β receptors. Activated TGF-β receptor complex phosphorylates receptor-activated Smads (R-Smads) in their C-terminus. Activated R-Smads associate with a common-mediator Smad (Co-Smad), and this complex translocates into the nucleus, where Smads bind to DNA or associate with transcriptional coactivators or corepressors. Inhibitory Smads (I-Smads), the expression of which is induced by TGF-β, are capable of inhibiting the phosphorylation of R-Smads. I-Smads also recruit Smurf E3-ubiquitin ligases to the receptor complex, thus directing TGF-β receptors to degradation. Smad-anchor for receptor activation (SARA) facilitates the recruitment of R-Smads to the receptor complex.

Smad signaling

The mammalian Smad family consists of 8 members, which can be divided into 3 groups according to their function: receptor-activated Smads (R-Smads), common-mediator Smads (Co-Smads) and inhibitory Smads (I-Smads).10 Smad1, -2, -3, -5 and -8 belong to the group of receptor-activated Smads, Smad4 is the common-mediator Smad, and Smad6 and Smad7 belong to the inhibitory Smad group. Smad1, Smad5 and Smad8 mediate signals downstream of BMPs, whereas Smad2 and Smad3 mediate signals triggered by TGF-βs and activins. In general, Smads are widely expressed in most adult tissue and cell types, suggesting the importance of TGF-β signaling in tissue development and homeostasis.11 Accordingly, analyses of Smad knock-out mice have revealed that Smads are crucial in many processes during mammalian development, including gastrulation, angiogenesis, organogenesis, immune response, bone formation and wound healing. Hence, the majority of Smad-deficient mice die at embryonic stages.12

Smad proteins have in general 2 large conserved domains—the N-terminal mad homology 1 domain (MH1) and the C-terminal mad homology 2 domain (MH2)—which are separated by a less-conserved proline-rich linker region. The linker region participates in the crosstalk between Smads and the MAPK family.13, 14 In addition, the linker region of Smad3 has transcriptional activity.15 The MH1 domain mediates the DNA binding activity of Smads, whereas the actual functional domain is MH2 domain, which possesses transcriptional activity.16 MH2 domain is responsible for Smad–Smad interactions and interactions with other proteins.17 The MH2 domain of R-Smads contains an SSXS motif, which is the target of the ALK kinases. In the basal state, R-Smads are inactive, since the MH1 domain binds to the MH2 domain and inhibits it. The MH2 domain is liberated from the MH1 domain when ALK kinase phosphorylates the SSXS sequence. This leads to a conformational change and subsequent activation of the Smad signaling pathway.18

Smad4 and the I-Smads lack the SSXS motif, and thus cannot be phosphorylated by ALKs. SARA (Smad anchor for receptor activation), a FYVE domain membrane-bound protein interacts with Smad2 and Smad3 and facilitates their recruitment to the activated TGF-β receptor complex by controlling their subcellular localization.19 After activation, R-Smads associate with each others and Co-Smad4 via their MH2 domains. Subsequently, the hetero-oligomeric Smad complex translocates into the nucleus, where Smads can bind directly to DNA or associate with other transcription factors or transcriptional coactivators or corepressors, and regulate the transcription of various TGF-β responsive genes.10

Inhibitory Smads suppress R-Smad activation by competitively interacting with TGF-β receptors. After ligand stimulation, I-Smads translocate from the nucleus to the cytoplasm, where they associate with type I receptors and inhibit the phosphorylation of R-Smads.20 Smad7 is a general inhibitory Smad capable of inhibiting signals triggered by TGF-βs, activins and BMPs, whereas Smad6 inhibits only signals triggered by BMPs. In addition to inhibiting R-Smad phosphorylation, Smad7 can recruit E3 ubiquitin ligases Smurf1 (Smad ubiquitination regulatory factor) and Smurf2 to the receptor complex, thus directing TGF-β receptors to degradation.21 Smad7 also facilitates TβRI dephosphorylation by interacting with growth arrest and DNA-damage-inducible protein-34 (GADD34), a regulatory/targeting subunit of the protein phosphatase-1 holoenzyme,22 resulting in inhibition of TGF-β signaling. Smad6 may specifically compete with Smad4 for binding to BMPR-activated Smad1 by forming an inactive Smad1/Smad6 complex in the cytoplasm.23 In addition to their inhibitory effects in the cytoplasm, I-Smads function as repressor of transcription in the nucleus by recruiting histone deacetylases to TGF-β/Smad target genes.24

Crosstalk between Smad and MAPK signaling pathways

TGF-β signals also via MAPK pathways. Multicellular organisms have 3 well-characterized subfamilies of MAPKs that control a vast array of physiological processes. The MAPK pathways activated by TGF-β include the extracellular signal-regulated kinases, ERK1 and ERK2; the c-Jun N-terminal kinases, JNK1 and JNK2; and the 4 p38 isoforms, p38α, p38β, p38γ and p38δ (Fig. 1).

Several reports demonstrate that Smad and MAPK pathways have functional interactions.25 In certain situations, MAPKs have been implicated as positive regulators of Smad-dependent effects. For instance, MEKK-1, a component of the JNK pathway, can selectively activate Smad2-dependent transcription independently of TGF-β in endothelial cells.26 MEKK-1 phosphorylates Smad2 at a site distinct from the C-terminal SSXS motif usually phosphorylated by ALK5, probably in the linker region. Additionally, JNK can phosphorylate Smad3 in the linker region and facilitate both activation and nuclear accumulation of Smad3.27 ERK1/2 can phosphorylate Smad2 and Smad3 in their linker region, but not in the C-terminal SSXS motif.14, 28 Three ERK phosphorylation sites (T178, S203 and S207) have been identified and characterized within the linker region of Smad3.29 In addition, the linker region of Smad3 can be phosphorylated by p38.30 p38 MAPK regulates human prostate cell adhesion by directly phosphorylating Smad3 and enhancing its nuclear translocation. p38 positively modulates Smad signaling also in MCF10CA1h breast cancer cells by affecting the linker phosphorylation.31 This requires cooperation with Rho/ROCK (Rho-associated protein kinase) pathway. In gingival fibroblasts, p38 and Smad3 cooperate in regulating the TGF-β-induced MMP-13 gene expression, whereas ERK1/2 cooperates with Smad3 in regulating connective tissue growth factor (CTGF) gene expression.32, 33

Conversely, MAPKs can act as negative regulators of Smad signaling. Kretzschmar et al. demonstrated that ERK1/2 phosphorylation in the linker region of Smad1 inhibits its nuclear translocation, leading to suppression of Smad1-induced transcription.13 In a similar manner, activation of ERK1/2 by oncogenic Ras or activated forms of MEK1 can lead to the phosphorylation of Smad2 and Smad3 in the linker region, thereby inhibiting nuclear accumulation of Smad2 and Smad3, and reducing Smad-dependent transcription.34

Nevertheless, although the studies mentioned earlier imply that MAPKs function upstream of Smads, there is evidence that Smads can act upstream of MAPKs and mediate their activation. In head and neck SCC cells, activation of p38 MAPK is Smad-dependent, as expression of Smad7 and dominant-negative Smad3 efficiently prevents p38 activation.35 Accordingly, p38 activation by TGF-β can occur through a Smad-independent pathway via TAK1,36-38 as well as through Smad-dependent pathways, apparently via GADD45β.39-41 Smad6 can inhibit BMP/TAK1-induced phosphorylation of p38 in a mouse hybridoma cell line.42 ERK1/2 activation can also be mediated by Smads, since dominant negative Smad4 inhibits the ability of TGF-β to activate ERK1/2 in pancreatic acinar cells.43 However, Smad7 can mediate the TGF-β-induced activation on MKK3 and p38 in prostate cancer cells by acting as a scaffolding protein that facilitates TAK1- and MKK3-mediated activation of p38.44 Smad7 also has a role in regulating the JNK signaling cascade, as forced expression of Smad7 initiates a signaling cascade leading to JNK activation, and triggers cell death in multiple cell lines.45

TGF-β in tumorigenesis

The antiproliferative effects of TGF-β

TGF-β has a dual effect on tumor growth. As a multifunctional growth factor, it has both tumor suppressor and tumor promoting activities, depending on the stage of carcinogenesis and the responsiveness of the tumor cell (Fig. 3).46, 47 In early tumor stages and in normal epithelial cells, TGF-β functions as an antiproliferative factor inhibiting cell growth by induction of apoptosis and cell cycle arrest. Loss of the antiproliferative responsiveness to TGF-β is often considered as a major step in cancer progression.48 TGF-β upregulates the transcription of the cyclin-dependent kinase (CDK) inhibitors p21^(Cip1/Waf1) and p15^(Ink4b), which inhibit CDK phosphorylation of the retinoblastoma protein and thus halt cell-cycle progression in G1.49 Transcriptional cofactors C/EBPβ and FoxO, as well as the tumor protein p53, are essential for the induction of p15 and p21 expression and for the cytostatic functions of TGF-β.50-52 In addition, TGF-β downregulates the transcription of c-Myc in a Smad3-dependent manner.53 Downregulation of c-Myc expression appears to be a prerequisite to the growth-suppressive actions of TGF-β, since it is essential for p15 and p21 induction.54, 55 In human cancers, the C-MYC gene is often amplified or overexpressed, which can lead to TGF-β resistance. Furthermore, inactivation of the gene for p15, p21, Smads or TGF-β receptors can allow cells to escape from the growth control of TGF-β and lead to progression of tumor.56

Figure 3. Biphasic role of TGF-β in tumor progression. TGF-β exerts antiproliferative effects on normal epithelial cells and in early tumor development. However, as tumors develop, they start to produce TGF-β, and become refractory to the growth suppressive effects of TGF-β. EMT, epithelial-to-mesenchymal transition.

Proinvasive and metastatic effects of TGF-β

Although mutations in TGF-β signaling components in cancers occur, complete abrogation of TGF-β signaling is not a generalized phenomenon in cancers. In fact, in various tumor cells, the TGF-β signaling pathway is functional, and tumor cells can use TGF-β as a tumor-progression factor.57 Increased production of TGF-β occurs in different tumor types, and correlates with the severity of the tumor grade.49, 58 Tumor-derived TGF-β can affect several cell types in proximity of the tumor, thus producing a microenvironment that promotes tumor growth, invasion and metastasis59 (Fig. 4). TGF-β can also act on the tumor cells directly, and regulate their capacity to remodel the surrounding ECM. TGF-β enhances proteinase expression and plasmin generation by tumor cells, which in turn leads to enhanced activation of TGF-β and degradation of the ECM with a consequent release of stored TGF-β. In various cancer cells, TGF-β upregulates the expression of MMPs,35, 60-62 that are able to activate L-TGF-β. This provides a positive regulatory feedback loop leading to increased TGF-β activation and tumor progression (Fig. 4). In addition, expression of MMP-13 can promote survival of both SCC cells and stromal fibroblasts.63, 64 Invasion of head and neck SCC cells is potently stimulated by TGF-β and this is mediated via p38 MAPK pathway.60 Interestingly, the predominant p38 isoforms expressed by head and neck SCC cells are p38α and p38δ, both of which regulate invasion capacity and growth of head and neck SCC cells in culture and in vivo, suggesting these p38 isoforms as potential therapeutic targets in head and neck SCC.65

Figure 4. A proposed model for role of TGF-β in invasion and growth of head and neck SCCs. SCC cells produce TGF-β, which stimulates their MMP-13 production, which promotes invasion and activates latent TGF-β (L-TGF-β) from the ECM, providing a positive feedback loop. TGF-β also stimulates stromal fibroblasts to produce CTGF, tissue inhibitor of metalloproteinases (TIMP)-3, and MMP-13. MMP-13 can also promote survival of SCC cells and fibroblasts. TGF-β and CTGF both stimulate the proliferation of fibroblasts, which eventually undergo phenotypic change to myofibroblasts, associated with increased contraction and ECM production.

Evidence for the role of TGF-β signaling in the complex process of cancer metastasis has recently been documented specifically in breast cancer. In mouse models of breast cancer, TGF-β promotes bone metastasis mediated by secreted factors such as parathyroid hormone-related peptide, interleukin-11 and CTGF.66, 67 Recent studies demonstrate that Smad signaling is required for this TGF-β-induced bone metastasis of breast cancer cells.68, 69 RNA-interference-mediated depletion of Smad4 in MDA-MB-231 breast cancer cell line inhibited bone metastasis in a mouse xenograft model, and this could be restored by ectopic expression of Smad4. Azuma et al. demonstrated that adenoviral overexpression of Smad7 in cultured breast tumor cells and in an in vivo mouse model of breast cancer inhibited the invasive phenotype of tumor cells.70 Also Smad overexpression studies implicate Smad signaling in breast cancer. Overexpression of Smad3 had prometastatic effects on breast cancer cell lines,71 and a Smad-binding defective mutant of TβRI suppressed the metastatic capacity of breast cancer cell lines.72

Smad signaling has been implicated in mediating the tumor-promoting effects of TGF-β also in other models. Smad3 knock-out mice exhibited resistance to skin chemical carcinogenesis because of inhibition of keratinocyte proliferation and reduced inflammation.73 Inhibition of Smad signaling by Smad7 inhibited the tumorigenicity of human melanoma cells in vitro and in vivo, and this was associated with reduced MMP production.74 In addition, stable overexpression of Smad7 in human melanoma cells impaired the bone metastatic capacity of the cancer cells.75

We have recently noted that Smad signaling mediates the invasive phenotype of head and neck squamous carcinoma cells by regulating their collagenase production (Fig. 5).35 Inhibition of Smad signaling by Smad7 inhibited collagenase-1 (MMP-1) and collagenase-3 (MMP-13) production and invasion by SCC cells, and suppressed the initial growth of SCC xenografts in SCID mice. A recent study also demonstrates that Smad4 is a tumor suppressor in pancreatic adenocarcinomas, yet facilitating EMT and TGF-β-dependent growth of advanced tumors.76 Thus, Smad signaling contributes to the oncogenic activities of TGF-β in cancer cells that no longer respond to the antiproliferative signals of TGF-β, suggesting Smad pathway as a potential target to inhibit the invasion of malignant tumors.

Figure 5. TGF-β signaling pathways mediating head and neck SCC cell invasion. Smad3 and p38 MAPK mediate the TGF-β-elicited invasion of human head and neck SCC cells via induction of MMP-13 expression. TGF-β-induced activation of p38 is Smad-dependent. Smad7, dominant negative Smad3 (SmadDN) and kinase-defective ALK5 (KD-ALK5) can inhibit activation of Smad3, thus inhibiting MMP-13 expression and cell invasion. SB203580, specific chemical p38 inhibitor.

TGF-β signaling inhibitors for cancer treatment

Inhibition of TGF-β signal transduction would be beneficial in treating pathological conditions, e.g., tissue fibrosis and cancer, in which TGF-β plays a central role. Hence, development of specific inhibitors for TGF-β signaling has gained interest during the past years. Large molecular inhibitors, such as TGF-β-antibodies and antisense oligonucleotides, have been used quite widely in preclinical and clinical studies for cancer therapy.77, 78 AP-12009 (Antisense Pharma) is an antisense oligonucleotide against TGF-β2, which has been developed for the treatment of gliomas.79 In Phase I/II dose-escalation studies intratumoral delivery of AP-12009 resulted in significant increase in survival time. AP-11014, another antisense oligonucleotide targeted against TGF-β1, is being developed for treating human nonsmall-cell lung, colorectal and prostate cancers.

An attractive approach for drug development is the small molecular TGF-β inhibitors, which are selectively targeted to inhibit TGF-β receptor kinases. For instance, a selective inhibitor for ALK5 would inhibit ALK5-mediated Smad2/3 phosphorylation, and leave ALK5-independent pathways functional. The advantage of ALK kinase inhibitors is also that they inhibit all TGF-β isoforms, unlike for instance antisense oligonucleotides, which can only be targeted against a specific TGF-β isoform. SB-431542 and SB-505124 are compounds related to imidazole inhibitors of p38, and they potently inhibit ALK4, ALK5 and ALK7 activity.80, 81 However, as they resemble p38 inhibitors, they may also inhibit p38 activation, and thus are not entirely specific inhibitors for Smad2/3 signaling. This complicates their use in the treatment of diseases. LY364947, LY566578 and LY580276 are dihydropyrrolopyrazoles, which efficiently inhibit ALK5 kinase activity, and also inhibit EMT in normal murine mammary epithelial cells.82 Because of their specificity against ALK kinases, they would be potential candidates for drug development to treat cancer. Another promising ALK5 inhibitor is SD-208, which was shown to inhibit the growth and invasion of human and mouse glioma cells in vitro and in vivo.83 In addition, when mice were administered SD-208, TGF-β was no longer able to suppress the cells of the immune system, suggesting that SD-208 also enhances tumor immunogenicity. In another recent study, oral administration of SD-208 inhibited the metastatic efficiency and growth of mouse mammary carcinomas in vivo, while having no obvious signs of toxicity.84 Ehata et al. reported that oral administration of a TβRI-kinase inhibitor, Ki26894, significantly reduced the in vivo bone metastatic capacity of human breast cancer cells.85 Furthermore, another TβRI inhibitor, TβRI-I, also attenuated the bone metastasis of breast cancer cells.86

However, because of the multifunctional activities of TGF-β, complete loss of TGF-β signaling would eventually be detrimental. TGF-β is a potent immunosuppressive agent, and it plays a role in development of chronic inflammatory diseases,87 and so blocking TGF-β may promote inflammation and development of autoimmune diseases. In addition, it should be noted that TGF-β is an efficient suppressor of tumor growth in early stages of carcinogenesis, and inhibition of TGF-β signaling could enhance tumor growth in situations where cancer cells are still responsive to the antiproliferative effects of TGF-β.

Concluding remarks

Current observations provide evidence on the role of TGF-β in the complex process of carcinogenesis, tumor progression, invasion and metastasis. Unveiling the signaling pathways mediating the cellular effects of TGF-β has allowed development of novel therapeutic strategies targeting TGF-β signaling at distinct levels as a basis for inhibiting tumor growth, angiogenesis, invasion and metastasis. It is expected that these therapeutic approaches in combination with other effective anticancer therapies may provide new alternatives to treat patients with malignant tumors characterized by rapid progression, invasion and metastasis.