SMAD signalling and TGF-β-dependent EMT

SMADs are a class of transcription factors that interact with and modulate the activity of the TGF-β receptor complex (Refs Reference Massague and Wotton23, Reference Bottinger and Bitzer24, Reference Schnaper29). Ligand-induced activation of the TGF-β receptor complex phosphorylates receptor-associated SMADs and promotes the recruitment of additional SMADs and other proteins. This activation also results in the translocation of a SMAD complex to the nucleus, where it activates the expression of some TGF-β target genes and represses the expression of others (Fig. 2). The specificity of different transcriptional responses is partly determined by the combination of different elements that make up the TGF-β-receptor–SMAD complex.

SMADs can be loosely divided into three groups. The first group are the receptor-regulated SMADs (or R-SMADs), which include SMAD2 and SMAD3 (which complex the type I TGF-β receptor and activin receptors, respectively) and SMAD1, 5 and 8 [which complex bone morphogenic protein (BMP) receptors]. Expression and activation of R-SMADs are necessary for the induction of EMT by TGF-β in renal tubular cell, as inhibitory RNA against SMAD2/3 is able to block TGF-β1-induced EMT in vitro (Ref. Reference Phanish30). In addition, SMAD3-knockout mice are protected from induction of EMT by TGF-β (Ref. Reference Zavadil31) and from tubulo-interstitial fibrosis induced following UUO, a model associated with profound renal scarring and EMT (Ref. Reference Sato32). Notably, in this model system, decreased expression of E-cadherin appears to be SMAD3-dependent, while increased expression of MMP2 appears to be SMAD2-dependent (Ref. Reference Phanish30). Importantly, such findings suggest that, at least at a signalling level, EMT can be differentially regulated (Ref. Reference Phanish30).

The second group of SMADs are the inhibitory SMADs (I-SMADs: SMAD6 and 7). SMAD7 inhibits the TGF-β/activin signalling pathway by binding to the TGF-β receptor, thereby preventing its phosphorylation and the subsequent receptor-dependent activation of the R-SMADs (Refs 33, 34). SMAD7 is also able to promote the inactivation or degradation of the TGF-β receptor, by recruiting protein phosphatase 1 (PP1) and ubiquitin ligases, respectively. In addition, I-SMADs can disrupt the formation of the functional SMAD–DNA complex at the SMAD-recognition elements, and thus prevent TGF-β-regulated transcription. SMAD6 is also a regulator of BMP, itself an antagonist of TGF-β-dependent signalling (Ref. Reference Itoh35) (see below).

Disruption of SMAD7 function increases TGF-β/SMAD2/3-dependent renal fibrosis and inflammation, indicating that SMAD7 is a pivotal negative regulator of TGF-β/SMAD- dependent signalling (Ref. Reference Chung36). Overexpression of SMAD7 has been shown to antagonise TGF-β-mediated changes in various models of EMT. For example, overexpression of SMAD7 prevents SMAD3-dependent EMT following UUO (Ref. Reference Lan37) and may also prevent TGF-β1-induced EMT (Refs 38, Reference Li39) and angiotensin-II-induced EMT in vitro (Ref. Reference Carvajal40). Although reduced expression of SMAD7 has been observed in models of renal fibrosis (Ref. Reference Fukasawa41), overexpression of SMAD7 may also induce apoptosis in some systems, making its therapeutic potential problematic (Ref. Reference Schiffer42).

The final category of SMADs contains a single member, SMAD4, which is a common-partner SMAD (or Co-SMAD). SMAD4 oligomerises with R-SMADs (SMAD2 and/or SMAD3) after their phosphorylation by the TGF-β receptor, and therein facilitates the translocation of the SMAD complex to the nucleus (Ref. Reference Massague43). The role of SMAD4 in tubular EMT and renal fibrosis is less clear than for the other SMADs. However, increased cytoplasmic expression of SMAD4 may be involved in augmenting TGF-β1 signalling in the renal tubulo-interstitial cells associated with experimental renal fibrosis (Ref. Reference Goto44).

Connective tissue growth factor and EMT

CTGF is a matrix-associated, heparin-binding protein that plays an important role in the regulation of cell proliferation, apoptosis, embryogenesis, differentiation, and wound healing (Refs Reference Oemar and Luscher45, Reference Grotendorst46). In addition, CTGF has been shown to directly induce EMT. For example, infection of a tubular epithelial cell line (NRK52-E) with an adenovirus encoding CTGF causes a dose-dependent transition in cellular morphology (Ref. Reference Burns, Kantharidis and Thomas47). Similar changes were also seen in NRK52-E cells with the addition of exogenous CTGF (Ref. Reference Burns48). However, when compared with the complete transition to myofibroblasts induced by TGF-β1, EMT induced by CTGF, delivered as a recombinant protein or in an adenovirus, was only partial (Ref. Reference Burns48). Rather than observing a complete loss of E-cadherin expression, there was a rapid reorganisation of the pattern of expression, from a membrane-associated distribution characteristic of epithelial cells to a perinuclear pattern. This change may represent trafficking of E-cadherin to endosomes and/or lysosomes where it may be internalised and degraded. The functional significance of modified localisation remains to be established. However, given the key role of E-cadherin in maintaining the structural integrity and polarity of epithelial cells, this redistribution of E-cadherin may modify adherens junction formation and cell polarity, and contribute to the loss of cytoskeletal integrity and the acquisition of a mesenchymal phenotype (Refs Reference Liu2, Reference Kalluri and Neilson6, Reference Strutz, Muller and Neilson11). A similar change in E-cadherin distribution has been previously reported in response to TGF-β1, at an earlier time point (Ref. Reference Peinado, Quintanilla and Cano49), suggesting that CTGF is predominantly an early, rather than late, player in the EMT pathway.

The expression of CTGF is induced by TGF-β1 (Ref. Reference Okada50), and other stimuli that trigger the expression of TGF-β1 (e.g. angiotensin II, AGEs) (Ref. Reference Burns48) – contexts in which it is broadly regarded as a key downstream mediator. For example, blockade of CTGF is able to attenuate TGF-β-dependent EMT in vitro and in vivo. However, although blockade of CTGF reduced the changes associated with EMT, it did not completely ameliorate the phenotypic changes caused by TGF-β1 treatment (Refs Reference Burns, Kantharidis and Thomas47, Reference Burns48) (Fig. 3). Nonetheless, this partial blockade may be sufficient to reduce renal fibrosis in vivo. In a separate study, CTGF blockade attenuated angiotensin-II-induced EMT (Ref. Reference Chen51), indicating that angiotensin II induces EMT partly through a CTGF-dependent pathway. In rodent models of chronic kidney fibrosis, CTGF appears to be required in addition to TGF-β1 to induce persistent injury (Ref. Reference Mori52). In addition, blockade of CTGF using CTGF antisense oligonucleotides attenuated increases in CTGF, fibronectin (FN1), collagen I (COL1) and αSMA gene expression and reduced the interstitial fibrotic areas normally seen in the UUO model or renal fibrosis, despite the sustained overexpression of TGF-β being unaffected (Ref. Reference Yokoi53). This supports the role of CTGF as a (partial) downstream mediator of the actions of TGF-β. Finally, diabetes-induced glomerular basement membrane thickening is attenuated in heterozygous Ctgf ^+/− mice (Ref. Reference van Nieuwenhoven54).

Figure 3.

CTGF blockade attenuates TGF-β1-induced αSMA expression in tubular epithelial cells. NRK52-E cells were preincubated in transfection reagent (oligoF) in the presence or absence of small interfering RNA (siRNA) targeting CTGF (CTGF-286) (1nm) for 24 h prior to the addition of TGF-β1 (10 ng/ml) for 3 days. The altered protein expression of αSMA was demonstrated by confocal microscopy and western blotting. (a) Cells were immunostained for αSMA (green staining). The cell nuclei were stained with propidium idodide (red). All images are taken at magnification ×640. (b) Western blot demonstrating the expression of CTGF, αSMA and β-tubulin. (c) Quantitation of western blot data for αSMA protein expression (versus β-tubulin expression). The oligoF + control group was arbitrarily assigned a value of 1. Each bar represents the mean ± SEM for three (protein) samples per group (*P < 0.01 versus control + oligoF; ^#P < 0.01 versus TGF-β1 + oligoF). The figure is compiled from experiments and data reported in Ref. Reference Burns48. Abbreviations: CTGF, connective tissue growth factor; αSMA, α smooth muscle actin; TGF-β1, transforming growth factor β1.

The mechanism by which CTGF mediates its effects remains to be fully established. Certainly CTGF has a number of actions that may be considered independent to TGF-β (Refs Reference Yang and Liu1, Reference Lan18, Reference Okada50), as TGF-β1-neutralising antibodies do not prevent the induction of EMT by CTGF in tubular cells (Ref. Reference Burns48). As well as mediating some of the actions of TGF-β, CTGF may also enhance the effects of TGF-β by inhibiting the negative feedback loop of the TGF-β–SMAD signalling pathway (Ref. Reference Abdel Wahab and Mason55). For example, treatment of mesangial cells with CTGF rapidly inhibits the expression of the I-SMAD SMAD7 (Ref. Reference Wahab, Weston and Mason56), possibly via a tyrosine receptor kinase-A (TrkA/TRK1)- and TGF-β immediate early gene (TIEG/KLF10)-mediated pathway (Ref. Reference Wahab, Weston and Mason57). In addition, studies in a human kidney cell line (HK-2 cells) have shown that CTGF-induced EMT is also able to activate the mitogen-activated protein kinase (MAPK) and phosphoinositide 3-kinase (PI3K) signalling pathways implicated in the development and progression of EMT (Ref. Reference Liu58) (see below).

Several studies have investigated possible receptors for CTGF. However, a definitive CTGF-specific receptor remains to be identified. A crosslinking study in a human chondrocytic cell line (HCS-2/8) identified a ¹²⁵I-labelled-CTGF–receptor complex with an apparent molecular mass of 280 kDa as a potential CTGF receptor (Ref. Reference Nishida59). CTGF has been shown to bind to the multiligand receptor LRP2 (low-density lipoprotein receptor-related protein 2) in fibroblasts (Ref. Reference Segarini60). However, a separate study has shown LRP2 plays a role in the endocytosis of CTGF, which leads to its degradation by endosomes (Ref. Reference Chen61). Therefore, it is possible that binding of CTGF to this putative receptor may facilitate clearance rather than lead to signal transduction. More recently, the receptor TrkA and the pan-neurotrophin receptor (NGFR/p75^NTR) were shown to interact with CTGF in primary human mesangial cells (HMCs) (Ref. Reference Wahab, Weston and Mason57). These receptors are activated by nerve growth factor (NGF) and mediate the biological effects of neurotrophins. In further support of the finding that CTGF activates the receptor TrkA in HMCs, the TrkA-selective inhibitor K252a blocked the internalisation and sorting of the receptor–ligand complex and the CTGF-dependent induction of TIEG (Ref. Reference Wahab, Weston and Mason57); however, unlike blockade of CTGF, the blockade of TrkA does not prevent TGF-β1-mediated EMT in proximal tubular cells in vitro [Burns, W.C. (2009) PhD thesis. University of Melbourne, Australia], suggesting the actions of CTGF in this model are probably through alternative and as yet unknown signalling pathways.

ILK and Snail and TGF-β-dependent EMT

The induction of an intracellular serine/threonine kinase – integrin-linked kinase (ILK) – also plays a role in mediating TGF-β1- and CTGF-dependent EMT (Ref. Reference Liu62). TGF-β1 is a potent inducer of ILK in renal tubular epithelial cells, via activation of SMAD2/3 (Ref. Reference Li38) and CTGF (Ref. Reference Liu62). In experimental studies, activation of ILK is able to independently induce EMT (Ref. Reference Li38), while forced expression of a ‘kinase-dead’ ILK is able to block TGF-β1-induced tubular EMT (Ref. Reference Li38), suggesting that activation of ILK may be an important committed step for the induction of EMT. Consistent with this hypothesis, hepatocyte growth factor (HGF), an endogenous inhibitor of EMT (see section below), decreases the expression of ILK, paralleling its antifibrotic actions (Ref. Reference Li38).

ILK is thought to act by coupling integrins and growth factor receptors at the cell surface with intracellular signalling mediators, including protein kinase B (AKT3) and glycogen synthase kinase-3 (GSK3) through phosphorylation (Refs Reference Liu2, 38). Phosphorylation of GSK3 inhibits its functions, leading to the stabilisation of β-catenin and the zinc finger transcriptional repressor SNAI1 (Snail). Induction of SNAI1 significantly impacts on the development and progression of EMT (Refs Reference Acloque10, Reference Thiery63), including the transcriptional repression of E-cadherin and the tight-junction proteins claudins and ZO-1 (Ref. Reference Ikenouchi64). The expression of SNAI1 is partly regulated by other proteins with which it complexes, including SMAD3 and SMAD4 and the nuclear cofactor HMGA2 (high-mobility-group A2 protein) (Refs Reference Thuault65, Reference Thuault66, Reference Vincent67). SNAI1 expression also can be induced by TGF-β1 via MAPK- and PI3K-dependent pathways (see below) (Ref. Reference Peinado, Quintanilla and Cano49). Consistent with the important role of SNAI1 in TGF-β-dependent signalling, in experimental models overexpression of SNAI1 results in renal fibrosis and increased expression of MMP2 (Ref. Reference Yokoyama68) and MMP9 (Ref. Reference Jorda69).

Additional and downstream signalling pathways for TGF-β-dependent EMT

Activation of TGF-β receptors is also able to stimulate a range of pathways independent of and/or synergistic to classical SMAD-dependent gene activation (Refs Reference Zavadil and Bottinger20, Reference Derynck and Zhang70, Reference Kokudo71). Although most data indicate that SMADs are necessary to complete TGF-β-dependent EMT in tubular cells, activation of so-called non-SMAD pathways (Ref. Reference Zhang72) – including RhoA (RHOA; Ras homologue gene family, member A), MAPK and PI3K signalling cascades (Ref. Reference Derynck and Zhang70) – also contributes to the development and progression of the phenotypic transition (Refs Reference Zavadil and Bottinger20, Reference Kokudo71). Calling them ‘SMAD-independent’ is somewhat of a misnomer, as each can regulate TGF-β and SMAD expression and activity through feedback pathways and vice versa. Moroever, their actions with respect to EMT appear to be more cooperative than independent or alternative.

Endogenous inhibitors of EMT

Although most attention has focused on the positive signals that trigger EMT, this transition also requires repression of signals that prevent phenotypic transition and maintain homeostasis. These signals may also act to stimulate uncommitted cells to restore their epithelial phenotype (mesenchymal–epithelial transition, MET) (Ref. Reference Hay and Zuk9). The most well known of these homeostatic signals is BMP7 (Ref. Reference Wang, Lapage and Hirschberg93). BMP7 is a member of the TGF-β superfamily, which, like its other members, interacts with (and phosphorylates) R-SMADs and other associated proteins in heterogeneous complexes. In renal cell culture, BMP7 is able to prevent TGF-β1-induced EMT (Refs Reference Postlethwaite, Shigemitsu and Kanangat94, Reference Zeisberg95), possibly by inducing the expression of ID2 (Refs Reference Kondo96, Reference Kowanetz97), an antagonist of SMAD2/3 signalling (Ref. Reference Saika98) and a critical regulator of αSMA expression during EMT (Ref. Reference Valcourt99). In addition, BMP7 is able to directly induce MET in adult renal fibroblasts and thus potentially assist in the regeneration of an epithelial phenotype, particularly in uncommitted cells (Ref. Reference Zeisberg, Shah and Kalluri100). These beneficial actions are paralleled by experimental studies showing that BMP7 is also able to prevent renal fibrosis in a rat model of UUO or even partly reverse fibrosis when provided as a late intervention (Ref. Reference Zeisberg95). Similarly, treatment of diabetic mice with BMP7 is able to partially restore tubular function and normalise urinary albumin excretion (Ref. Reference Wang101). In support of these beneficial actions, the expression of gremlin, a secreted BMP antagonist that acts by binding BMP7 and neutralising its activity, is increased in cultured mouse proximal tubule cells treated with TGF-β1 (Ref. Reference Wang, Lapage and Hirschberg93). Similarly, in diabetic nephropathy, the renal expression of gremlin is upregulated as BMP7 levels are also reduced (Ref. Reference Wang, Lapage and Hirschberg93).

HGF has also been shown to inhibit TGF-β1-induced EMT in the diseased kidney (Refs Reference Yang and Liu21, Reference Yang, Dai and Liu102), possibly by preventing TGF-β1-induced ILK expression (Ref. Reference Li38) (see above), or via the induction of the SMAD transcriptional corepressor SnoN, which results in the formation of a transcriptionally inactive SnoN–SMAD complex (Ref. Reference Yang, Dai and Liu102). Overexpression of HGF is able to attenuate renal injury in experimental diabetes (Ref. Reference Dai103) or even induce regression of mesangial expansion and glomerulosclerosis in advanced disease (Ref. Reference Cruzado104). Notably, these beneficial effects have been associated with suppression of TGF-β1 and CTGF expression, paralleling a reduction in renal interstitial myofibroblasts (Ref. Reference Cruzado104). The effects of HGF in other systems and other types of EMT may not be the same as those observed in the kidney. In the original descriptions, HGF was known as scatter factor and was the first cytokine shown to cause cell scattering or EMT in different cancer types (Ref. Reference Brinkmann105), including renal cancer (Ref. Reference Weidner106).

A range of other transcriptional repressors involved in the regulation of TGF-β1-dependent signalling and EMT have been described, including SNAI2 (Slug) (Refs Reference Savagner, Yamada and Thiery107, Reference Zavadil108), FOS, TWIST 1, members of the Hairy/Enhancer-of-split family (Refs Reference Zavadil and Bottinger20, Reference Zavadil31), and the δEF family members δEF1 (DEFA1) and SIP1 (SFRS2IP) (Refs Reference Eger109, Reference Comijn110). These have largely been demonstrated in nonrenal models and are beyond the scope of the current review. However, it is likely that each will ultimately prove to play some role in EMT and renal fibrogenesis.