Epithelial-mesenchymal transition in renal fibrosis – evidence for and against

Abstract

Epithelial to mesenchymal transition (EMT) is a well established biological process in metazoan embryological development. Over the past 15 years, investigators have sought to establish whether EMT also occurs in renal epithelial cells, following kidney injury, and to show that the mesenchymal cells formed could give rise to myofibroblasts which populate the renal interstitium, causing fibrosis within it. There is no doubt that proximal tubular epithelial cells (PTECs) can undergo EMT in vitro in response to TGFβ-1 and other inflammatory stimuli. Moreover, the results of experiments with animal models of renal fibrosis and examination of biopsies from patients with chronic kidney disease have lent support to the hypothesis that EMT occurs in vivo. This review discusses some of the key evidence underlying that idea and summarises recent advances in understanding the molecular mechanism underlying the process. Early experiments using mice which were genetically engineered to mark PTECs with the LacZ gene to trace their fate following kidney injury provided evidence supporting the occurrence of EMT. Recently, however, cell lineage tracking experiments using the red fluorescent protein (RFP) as a high-resolution marker for cells of renal epithelial origin did not replicate this result; the interstitial space following kidney injury was devoid of RFP expressing cells, leading the investigators to reject the renal EMT hypothesis.

Introduction

Irrespective of the primary cause, chronic kidney disease (CKD) is characterised by the development of progressive renal tubular interstitial fibrosis (TIF), loss of the tubules and loss of the peritubular capillaries (Eddy 2005). It has long been established that these changes correlate with progressive loss of renal function and end-stage renal failure (Risdon et al. 1968). Tubular interstitial fibrosis is because of excessive deposition of extracellular matrix in the tubular interstitium by activated fibroblasts called myofibroblasts. These cells express α-smooth muscle actin (α-SMA), are highly active in synthesising extracellular matrix proteins such as fibrillar collagens and fibronectin, and their numbers increase markedly in the interstitium in CKD. Their origin has been a topic of increasing interest over the last decade. To date, it has been reported that myofibroblasts can derive from the activation of renal interstitial fibroblasts (Strutz et al. 2000), perivascular fibroblasts and pericytes (Lin et al. 2008); from the differentiation of bone marrow-derived stem cells (Broekema et al. 2007); and by the transition of endothelial cells (Zeisberg et al. 2008) and tubular epithelial cells (Strutz et al. 1995; Iwano et al. 2002) to mesenchymal cells. The conversion of these various cells to myofibroblasts in CKD is driven by factors released from local resident cells responding to disease-induced stresses such as hypoxia, hyperglycaemia or proteinuria and from inflammatory and immune cells which infiltrate the interstitium in renal disease. They include a diverse range of proteins such as the growth factors, TGF β1 (Inazaki et al. 2004; Wang et al. 2010) and connective tissue growth factor (CTGF) (Yokoi et al. 2004), cytokines such as IL-1 (Vesey et al. 2002) and oncostatin M (Nightingale et al. 2004), angiotensin II (Ishidoya et al. 1995), proteases (Cheng et al. 2006; Zhang et al. 2007), plasminogen activator inhibitor-1 (Matsuo et al. 2005) and advanced glycation end products (Burns et al. 2006). The mechanism of epithelial-mesenchymal transition (EMT) has been studied intensively in vitro and in vivo. However, recently, its role in renal interstitial fibrosis in vivo has been challenged by new experimental data (Humphreys et al. 2010), which conflicts with earlier conclusions (Iwano et al. 2002). Our aim in this review is to examine the main features of EMT, the key evidence for and against its occurrence in CKD, and to summarise recent findings regarding the mechanisms involved. The reader is also referred to several authoritative reviews of EMT in CKD (Neilson 2006; Strutz 2009; Liu 2010).

Epithelial-mesenchymal transition

Epithelial cells are polarised with their basal surface attached to a basement membrane (bm) and their apical side facing the lumen of a tubular structure or body cavity. Laterally, they interact with adjacent epithelial cells through adherens junctions, desmosomes and tight junctions, forming a pavement of cells on the bm. Initially, it was established that epithelial cells undergo transition into mesenchymal cells during embryogenesis. Mesenchymal cells have two key characteristics, which are important in embryonic development: they are able to migrate through an extracellular matrix and subsequently to differentiate into the many different cell types of the mesenchymal cell lineage, including fibroblasts. The reverse process, mesenchymal-epithelial transition (MET), also occurs during embryonic development and from gastrulation following implantation of the blasocyst, several rounds of EMT/MET are required in metazoan embryogenesis to generate a multi-organ embryo with its specialised cells and tissues (Thiery et al. 2009). Epithelial mesenchymal transition occurring during embryogenesis has been classified as Type1 EMT (Kalluri & Weinberg 2009). Later studies proposed that EMT also occurs during tissue repair and fibrosis (Type 2) and during the progression and metastasis of cancers (Type 3) (Kalluri & Weinberg 2009). Types 2 and 3 EMT share many common features with Type 1, for example, loss of intercellular junctions between epithelial cells, their detachment from a basement membrane, loss of epithelial phenotype markers and acquisition of mesenchymal markers and cell migration. However, they also exhibit some different features. Thus, Type 1 EMT occurs at specific stages during embryogenesis, rather than over prolonged periods, and does not result in either fibrosis or metastasis, as do Types 2 and 3 EMT. Hence, some EMT mechanisms are likely to be Type specific, and this should be kept in mind when extrapolating findings from one context to another.

Epithelial-mesenchymal transition in kidney fibrosis

In the normal kidney, the renal tubules, comprised of epithelial cells sitting on a tubular basement membrane (tbm), are surrounded by a small amount of interstitial tissue, comprised of an extracellular matrix in which occasional fibroblasts reside. These originate during kidney development after EMT of epithelial cells lining the lumen of the tubules and migration of the mesenchymal cells into the interstitium (Neilson 2006). The fibroblasts derived from these are relatively quiescent cells, whose function is to maintain this matrix and kidney structure (Qi et al. 2006). One view of EMT in CKD in adult life is that it is a reiteration of this developmental EMT which, because it is triggered by prolonged release of stimulatory factors from resident renal cells and infiltrating inflammatory cells responding to pathological conditions, leads to large numbers of activated myofibroblasts appearing in the interstitium and thus fibrosis.

Liu proposed four key events occur in tubular EMT in renal fibrogenesis, based on the studies of tubulointerstitial changes following experimental unilateral ureteric obstruction (UUO) in vivo and on the response of epithelial cells in vitro to TGFβ1, an inducer of EMT. These are the following events: loss of epithelial adhesion, cytoskeletal reorganisation and de novo synthesis of α-SMA, disruption of the tubular basement membrane and finally enhanced cell migration and invasion of the interstitium (Liu 2004).

Loss of epithelial cell adhesion

An early event involves downregulation of expression of E-cadherin, a major component of adherens junctions, by transcriptional repressors such as Snail-1 that are induced by TGFβ (Yang & Liu 2001; Peinado et al. 2003). The ectodomains of E-cadherin proteins of adjacent cells bind to each other in adherens junctions. TGFβ also induces proteolytic shedding of these domains by matrix metalloproteinases (MMPs) (Zheng et al. 2009). The cytoplasmic domain of E-cadherin forms a complex with β-catenin and other proteins at the plasma membrane, which interacts with the cytoskeleton. Proteolytic shedding of the E-cadherin ectodomain results in the release of the β-catenin, which is also a component of the Wnt signalling pathway. Following translocation from the plasma membrane to the nucleus, β-catenin induces Slug, repressing transcription of E-cadherin and inducing EMT (Zheng et al. 2009). Other junctions between epithelial cells must also undergo disassembly for EMT to occur.

Cytoskeletal reorganisation and α-SMA induction

The epithelial cell cytoskeleton is comprised of a meshwork of β-actin microfilaments extending out from the perinuclear area through the cytoplasm to a dense peripheral ring of filaments under the plasma membrane (Meza et al. 1980). This connects via adaptor proteins to cell adhesion molecules in the plasma membrane and helps determine the shape and polarity of the cells. The shape and cytoskeleton of proximal tubular epithelial cells (PTECs) undergoing EMT in vitro in response to TGFβ undergo dramatic reorganisation (Tian et al. 2003). Proximal tubular epithelial cells lose their cobblestone morphology and become elongated and spindle shaped with front-end to back-end polarity, typical of fibroblasts. This is accompanied by the de novo expression of α-SMA and reorganisation of the actin microfibrillar network into longitudinal bundles of fibres running through the cytoplasm to the plasma membrane where they connect with focal adhesion sites which contain integrins anchoring the cell to the extracellular matrix. α-smooth muscle actin is incorporated into these fibres, which are referred to as stress fibres and are characteristic of myofibroblasts. However, α-SMA is expressed also in vascular smooth muscle cells and in the mesangial cells of injured glomeruli, so it is not a unique marker for myofibroblasts. Fibroblasts contain little or no α-SMA. Myofibroblasts are contractile and motile, properties that are important in wound healing and fibrosis, and the expression of α-SMA stress fibres in these cells increases their contractile activity at least 2-fold compared with α-SMA-negative fibroblasts (Hinz 2007). However, not all collagen-producing myofibroblasts, as defined by other criteria such as the expression of the markers HSP47 and FSP-1 (S100A4), express α-SMA (Okada et al. 2000). Thus, relying on α-SMA as a myofibroblast phenotype marker may underestimate the total number of collagen-producing cells in the tubulointerstitium. The assembly of stress fibres and focal adhesions occurs downstream of the activation of the small GTPase, RhoA (Hall & Nobes 2000). Activation of RhoA was shown to be a key step in renal epithelial cell EMT stimulated by an inflammatory milieu in vitro (Patel et al. 2005).

Induction of UUO in mice is followed by the expression of Wnt proteins and the accumulation of cytosolic and nuclear β-catenin in tubular epithelial cells, indicating activation of the Wnt canonical signalling pathway. Downregulating the Wnt co-receptors LRP-5/6, with the antagonist Dickkopf-1 (which promotes their endocytosis), prevents β-catenin accumulation, reduces the expression of α-SMA and FSP-1, and blocks the accumulation of interstitial collagen in the obstructed kidney (He et al. 2009). Treatment of UUO mice with secreted frizzled-related protein 4, a non-signalling competitor of frizzled proteins (i.e. Wnt receptors), had similar effects, supporting a major role for the Wnt/β-catenin pathway in renal fibrosis (Surendran et al. 2005). It was concluded that Wnt/β-catenin signalling may synergise with TGFβ/Smad signalling to induce transcriptional factors such as Snail, Slug and Twist which downregulate epithelial genes and induce mesenchymal genes in EMT. Moreover, it seems possible that non-canonical Wnt signalling (i.e. not involving β-catenin) could be involved in activating Rho-GTPases which are important for changes in cell shape and migration (Schlessinger et al. 2009) which, as noted earlier, are necessary for EMT.

Role of proteases in EMT and invasion of the interstitium

Epithelial cell dissociation is dependent on the activity of MMPs. Investigations into UUO in mice deficient in tissue-type plasminogen activator (tPA) provided support for the involvement of MMPs in EMT in vivo. Expression of matrix metalloproteinase-9 (MMP-9), an enzyme capable of degrading basement membrane-type IV collagen, is upregulated in obstructed kidneys of tPA^+/+ mice, largely in interstitial cells adjacent to tubules. However, tPA^−/− mice have reduced levels of MMP-9 after UUO, and structural changes and collagen deposition were attenuated in them compared with tPA^+/+ mice. After 7d UUO, double staining for α-SMA and a tubular cell marker (Tetragonolobus purpureas lectin) revealed disorganised-looking cells in, or adjacent to, tubules and were positive for both markers in tPA^+/+ mice but not in tPA^−/− mice. Such cells were interpreted as being in a transitional stage between epithelial cells and myofibroblasts. Tbm integrity showed signs of breakdown after 7d UUO in tPA^+/+ mice but not in null mice (Yang et al. 2002).

Plasminogen (plg), a plasma protein, is activated to plasmin by tPA or urokinase-type plasminogen activator. Plasmin being a serine protease might be expected to degrade extracellular matrix during tissue remodelling at sites of wound healing or fibrosis. However, comparative studies of UUO in plg^+/+ and plg^−/− mice showed a marked reduction in collagen deposition 21 days after surgery in the null mice. Prior to this, plg^−/− mice accumulated fewer α-SMA⁺/FSP-1⁺ interstitial myofibroblasts and their tubular epithelium maintained higher E-cadherin levels than were found in plg^+/+ mice. FSP-1 staining was also observed in tubular epithelium in plg^+/+ mice, suggesting the presence of cells in the process of transition. These results pointed to a role for plasminogen or plasmin in promoting EMT and interstitial fibrosis in UUO. Plasmin can interact with protease-activated receptor-1 (PAR-1), triggering signalling via p42/44 extracellular-related kinases (ERKs). After UUO, similarly reduced levels of PAR-1 were expressed on tubular epithelial cells in wild-type and null mice but phospho-ERK levels were significantly lower in the kidneys of null mice, suggesting that this pathway was stimulated via plasmin-ligating PAR-1. Levels of active TGFβ were also lower in plg^−/− mice after UUO (Zhang et al. 2007). These results, backed by in vitro experiments, supported the notion that plasmin plays a significant role in inducing EMT via ERK signalling in the UUO model. Latent-TGFβ activation by plasmin may occur by direct proteolytic cleavage and/or by ligation of PAR-1, enhancing an α(v)β6 integrin-dependent activation pathway (Jenkins et al. 2006).

An intact tbm may be necessary to maintain the epithelial phenotype, and its disruption by proteases such as MMP2 and MMP9, which are induced by TGFβ and EGF in renal fibrosis, is sufficient to trigger EMT in vitro (Zeisberg et al. 2001). To test this concept experimentally, Cheng et al. (2006) generated a transgenic mouse overexpressing constitutively active MMP2, driven by the renal proximal tubule-specific type 1 γGT promoter. Enzymatically active transgenic MMP2 protein was localised with a basolateral distribution around cells within proximal tubules. At 4 months of age, transgenic mice showed widespread translucent areas of tbm, suggesting proteolytic damage, together with a few areas of tbm disruption with columns of cells having a mesenchymal phenotype invading the interstitium. The latter was termed extratubular EMT. However, there was widespread loss of cytokeratin (an epithelial marker) and ZO-1 (an adherens junction protein) in tubules of transgenic mice, accompanied by de novo expression of FSP-1in epithelial cells, as well as in interstitial fibroblasts. Tubular epithelial cells also expressed HSP-47 (a chaperone for interstitial collagen secretion), normally a marker for fibroblasts. These changes were termed intratubular EMT. Interstitial collagen deposition was evident around tubules by 4 months, its pattern being consistent with synthesis by tubular cells. Only a few tubular cells expressed α-SMA and Vimentin, suggesting that full conversion to myofibroblasts was rare. Later, when many tubules had atrophied, collagen occupied the interstitium. Thus, it was concluded that targeted expression of MMP2 in the proximal tubule can elicit the whole range of EMT changes in vivo without any superimposed injury but, most importantly, a range of cellular phenotypes occur in EMT. The most prominent change was intratubular EMT.

Evidence of EMT in human renal disease

Rastaldi et al. (2002) investigated 133 biopsies from different human renal diseases for markers for epithelial (Cytokeratin, ZO-1) and mesenchymal cells (Vimentin, α-SMA) and for evidence of collagen synthesis in tubular cells by staining for prolyl 4-hydroxylase, HSP47, collagen I and collagen III and in situ hybridisation to detect mRNAs for these collagens. Tubular epithelial cells in normal kidneys were positive for cytokeratin and ZO-1 and did not synthesise fibrillar collagens (Rastaldi et al. 2002). However, there was consistent evidence for fibrillar collagen synthesis by epithelial cells in diseased kidneys, accompanied by the loss of cytokeratin and ZO-1 and acquisition of mesenchymal markers. Some tubules were positive for both epithelial and mesenchymal markers. Only occasional cells expressed α-SMA. Changes were often detected in cases where the tbm was intact and where there was only mild or no interstitial fibrosis. Collagen I and III synthesis was evident in areas where these proteins were detected only in the tbm. Overall, the number of tubular cells with EMT features correlated with the degree of interstitial fibrosis and the patient's serum creatinine level. These data suggest that epithelial cells undergo at least intratubular EMT in human CKD and contribute in this way to the fibrotic process.

Recent understanding of molecular mechanisms in EMT

A multitude of agents has been shown to activate EMT in vitro, including TGFβ, CTGF and receptor tyrosine kinases. Signals from these converge on several transcription factors including Snail, Slug, Zeb1/2 and Twist which transcriptionally induce EMT. Recently, a number of studies have focused on understanding the mechanisms for disruption of cell polarity. Par6 encodes a PDZ-domain-containing protein that is conserved in Drosophila and mammals and is a regulator of epithelial cell polarity and tight junctions. Par6 was shown to interact with TGFBRI and phosphorylation of Par6 by TGFBRI was required for TGFβ to initiate EMT cellular effects. Evidence was also presented that TGFβ promotes interaction of phospho-Par6 with the E3 ubiquitin ligase, Smurf1, a protein which has been shown to target RhoA for degradation, leading to disassembly of the tight junction complex (Ozdamar et al. 2005). Other investigations showed that Snail, a transcription factor involved in repression of target genes, antagonised expression of the Crumbs/Par6 polarity complex. The Crumbs complex, like the Par complexes, has been shown to control epithelial polarity and apical membrane formation. Snail abolished the localisation of Crumbs and Par from the cell membrane and additionally acted directly as a transcriptional repressor, suppressing Crumbs mRNA (Whiteman et al. 2008).

TGFβ-1 belongs to the TGFβ superfamily of proteins which are expressed in all mammalian cell types, as are their downstream signalling mediators, the Smad proteins. Recent studies have shown that prolonged stimulation of renal tubular epithelial cells with TGFβ-1 causes a decrease in the expression levels of the TGFβ receptor-associated protein SARA, accompanied by a decrease in Smad2 protein and increased expression of α-SMA. SARA did not affect Smad2 transcription but rather the absence of SARA enhanced the interaction of Smad2 with the E3 ubiquitin ligase, Smurf2, promoting Smad2 degradation. SARA knockdown in the absence of TGFβ also led to increased α-SMA expression and decreased expression of ZO-1 and E-cadherin (Runyan et al. 2009). This work highlights the importance of TGFβ in regulating EMT in a Smad2/SARA-dependent fashion.

TGFβ-1 also maintains gene silencing of epithelial phenotype genes such as E-cadherin by driving hypermethylation of the relevant promoters. This was confirmed by disrupting Smad signalling in cells undergoing EMT, which led to re-expression of previously hypermethylated genes. These findings are important because they show that TGFβ-1 not only drives transcription of EMT-associated genes but also works by blocking re-expression of epithelial markers, such as E-cadherin (Papageorgis et al. 2010).

Lefty proteins are novel TGFβ-ligands which function as antagonists of Nodal signalling. Nodal is a member of the TGFβ superfamily involved in cell differentiation during development and, together with Lefty proteins, is expressed in the left side of the developing organism (Bajoghli et al. 2007). Overexpression of Lefty A protein in human tubular cells blocked TGFβ-induced Smad2/3 phosphorylation, prevented E-cadherin loss, attenuated collagen and CTGF production and inhibited cell transdifferentiation. These interesting findings emphasise the role of TGFβ in EMT and identify a new mechanism for inhibiting EMT (Li et al. 2010). In contrast, an earlier study reported that a high level of Lefty protein expression was maintained in human embryonic stem cells by activation of Smad2/3. Moreover, it was shown that inhibition of the Smad pathway blocked expression of the Lefty proteins, suggesting that Lefty transcriptional activation depends on Smad2/3 activation (Besser 2004). These two different responses to the TGFβ/Smad pathway may reflect a different role for TGFβ in development compared with adulthood.

Heat shock protein 72 (HSP72) was shown to ameliorate renal tubulointerstitial fibrosis in UUO nephropathy, via inhibition of EMT progression and tubular apoptosis (Mao et al. 2008). A more recent study by the same authors showed that the peptide-binding domain of HSP72 was able to bind and block Smad3, indicating a molecular mechanism by which HSP72 attenuates TGFβ-induced EMT. Additionally, silencing of the HSP72 gene using siRNA enhanced TGFβ-induced phosphorylation of Smad3. These studies suggest HSP72 as an anti-fibrotic molecule which works by exerting domain-specific effects on Smad3 activation (Zhou et al. 2010).

Increasing evidence supports that connective tissue growth factor (CTGF, CCN2) is involved in mediating EMT and that it plays a key role in tissue fibrosis, including renal scarring (Cooker et al. 2007). Moreover, TGFβ upregulates the expression of CTGF (Riser et al. 2000). Thus, the idea that CTGF mediates some TGFβ-induced EMT effects has been explored. Commercially obtained C-terminal CTGF was used to stimulate human renal tubular cells, and EMT markers were investigated: α-SMA was upregulated significantly whilst E-cadherin expression was suppressed. Interestingly, these CTGF-induced effects were integrin-linked kinase (ILK)-dependent and ILK knockdown using siRNA reversed the CTGF-mediated α-SMA increase and E-cadherin loss (Liu et al. 2008). In a cell model of TGFβ-induced EMT, inhibition of ILK blocked fibronectin, Snail, plasminogen activator inhibitor 1, and MMP2 expression. Additionally, blockade of ILK in an in vivo model of obstructive nephropathy using a small molecule inhibitor potently blocked expression of α-SMA, fibrillar collagens, vimentin, β-catenin and Snail (Li et al. 2009). Integrins have been proposed to act as CTGF receptors (Gao & Brigstock 2005). A neutralising antibody against the C-terminus of CTGF was reported to abrogate CTGF-mediated-EMT effects. Interestingly, the authors reported that commercially obtained N-terminal CTGF (containing domains 1–3) failed to promote any morphological change or EMT marker changes in the cells, in accord with the notion that the different domains of CTGF drive distinct biological effects (Liu et al. 2006).

Preincubation of tubular cells with BMP-7 attenuates TGFβ-induced EMT effects such as induction of type I collagen, α-SMA and CTGF and reduction in E-cadherin expression (Xu et al. 2009). More recently, in vitro studies by Xu et al. (2010) revealed that cyclosporin A (cyclosporin), a profibrotic drug, induced TGFβ, α-SMA and type I collagen and suppressed E-cadherin in human tubular cells. Interestingly, it was demonstrated that silencing CTGF expression with siRNA attenuated the cyclosporine A-induced EMT changes. These data highlight the importance of CTGF in the pathogenesis of renal disease and provide further insight into how EMT may be induced in the kidney (Xu et al. 2010).

Tracing the fate of cells in EMT

Many investigations have utilised markers of cell phenotype to identify changes following induction of renal disease. One problem with such studies is the lack of markers identifying only cells of the fibroblast/myofibroblast phenotype. The lack of specificity of α-SMA as a myofibroblast marker has been referred to above. Strutz et al. (1995) used differential gene screening techniques to distinguish transcripts expressed in mouse tubulointertstitial fibroblasts, but not in PTECs, and identified a protein which they called fibroblast-specific protein-1 (FSP-1). This protein is more widely known as S100A4, a small Ca²⁺ binding protein which promotes tumour metastasis. S100A4 was first cloned and sequenced (as P9Ka) because it was implicated in the conversion of a cuboidal epithelial stem cell line, Rama 25, to elongated myoepithelial cells (Barraclough et al. 1987). Later studies showed that it is also expressed in several normal human cells such as monocytes, macrophages, neutrophil granulocytes and endothelial cells, as well as in tumour cells (Boye & Maelandsmo 2010), so it is not fibroblast specific. Immunohistochemical detection of FSP-1/S100A4 expression in kidneys undergoing fibrosis after inducing anti-tbm disease revealed staining in fibroblasts at the site of interstitial collagen deposition and in tubular epithelial cells at sites of interstitial inflammation. Based on this and in vitro experiments, the authors hypothesised that adult renal epithelial cells may sometimes convert to fibroblasts (Strutz et al.1995). Many subsequent investigations into EMT in renal fibrosis, some of which are cited earlier, have used FSP-1/S100A4 expression as a marker of the fibroblast phenotype, but results should be interpreted cautiously because of its lack of specificity.

Iwano et al. (2002) used Cre-recombinase technology to generate mice in which PTECs expressed LacZ as a marker. Mice expressing Cre driven by a γGT promoter were crossed with ROSA26R (R26R) mice carrying a floxed LacZ gene in front of a TATAA-less promoter. γGT expression is specific to PTECs and is only turned on about postpartum day 7. Thus, PTECs in unfloxed progeny became positive for LacZ staining at this time and their fate could be traced after subjecting these mice to UUO. After staining for LacZ and FSP-1/S100A4 at day 10 after UUO, confocal images showed double-stained epithelial cells associated with disaggregating tubules. The LacZ⁺ epithelial cells were also HSP47 positive indicating that they were synthesising fibrillar collagen. The interstitium contained both FSP1⁺/LacZ⁻ cells, which were attributed to endogenous interstitial fibroblasts formed before postpartum day 7, and FSP1⁺/LacZ⁺ fibroblasts which, it was concluded, were most likely derived via EMT.

In a more recent study, Cre/Lox technologies were coupled to both LacZ and red fluorescent protein (RFP) reporters and used to map the fate of renal epithelium in UUO nephropathy and in normal kidneys (Humphreys et al. 2010). R26R mice carrying a floxed LacZ gene, or Z/red mice carrying a floxed RFP gene, were crossed with knockin mice expressing Cre-recombinase driven by a Six2-GC promoter. This restricts expression of Lac Z or RFP to only nephron epithelial cells derived from the cap-mesenchyme which includes PTECs, but excludes collecting duct epithelia and renal interstitial cells. Additionally, the floxed mice were crossed with HoxB7-Cre mice to fate map all cells coming from the mesonephric duct and its derivatives, including collecting duct and ureteric epithelium. The authors studied the expression of LacZ and RFP 10 or 14 days after UUO injury. There was no evidence of RFP⁺ cells (i.e. RFP⁺ genetically labelled epithelial cells) in the interstitium in either the Six2-GC- or the HoxB7-driven reporter mice, showing that following injury, renal epithelial cells do not migrate to the interstitium and therefore do not become myofibroblasts in vivo. RFP⁺ cells were checked for expression of the myofibroblast markers α-SMA and S100A4 by immunohistochemical assays. There was no evidence for RFP⁺ cells co-expressing myofibroblast markers. Myofibroblasts were present at day 10 post-UUO injury as evident by α-SMA and S100A4 positively stained interstitial cells. However, these myofibroblasts were never seen to express RFP, suggesting that epithelial cells do not transdifferentiate to myofibroblasts. Interestingly, when epithelial cells obtained from the mouse model described earlier were cultured in dishes, they were initially α-SMA and S100A4 negative but, when stimulated with TGF-β, they expressed both markers, in accord with previous studies demonstrating the role of TGFβ as an EMT initiator in vitro.

Overall, Humphreys et al. (2010) showed that following UUO injury, epithelial cells do not migrate to the interstitium and do not express α-SMA and S100A4, in contrast to Iwano et al. (2002) who reported co-localisation of epithelial cells with S100A4. A significant difference between the studies of Humphreys et al. and Iwano et al. is in the use and detection of the bacterial LacZ gene as a reporter. Humphreys et al. tested LacZ genetically labelled mice to fate map epithelial cells and tested several anti-β-gal antibodies (including one from the company used by Iwano et al.) to detect LacZ expression, but did not find any antibody which gave a reliable signal. Thus, they detected LacZ expression by enzymatic activity and performed high-resolution cell fate tracking using RFP instead. Iwano et al. used only LacZ for fate mapping of the epithelium and, given the experience of Humphries et al. concerning anti-LacZ antibodies, one could speculate that the specificity of immunodetection might have been compromised in their system.

Concluding comments

It is well established that EMT occurs physiologically in embryogenesis. Such observations provide the proof that epithelial cells can convert to mesenchymal cells. Thus, it was reasonable to hypothesise that a similar process may occur in pathological states, and a role for EMT in cancer is widely accepted. It is also agreed that primary PTECs or cell lines such as HK2 can undergo EMT in vitro in response to TGFβ and other inflammatory stimuli. However, despite the results of previous experiments on animal models in vivo and studies of human renal disease biopsies giving credence to the notion of EMT in CKD, recent cell lineage tracking experiments do not support this hypothesis. Tracking cell lineage in vivo, using mice in which reliable and persistent markers of cell origin have been genetically engineered, provides the most rigorous method to test whether EMT occurs in renal fibrosis. It is difficult to reconcile the conflicting results of Iwano et al. (2002) and Humphreys et al. (2010), each of whom investigated EMT in the UUO model, other than by arguing that technical differences in detecting epithelial cell fate markers may account for the disparity in their results. Humphries et al. applied the same model to mice carrying RFP as a high-resolution marker of renal tubular cell origin and found no evidence for EMT. One could speculate that the promoters used by both investigations may be switched off when an epithelial cell undergoes EMT. This potential problem could be overcome by more than two epithelial promoters in conjunction with different fluorescent proteins to obtain a more accurate picture of an epithelial cell undergoing or not EMT.

Translating interesting in vitro findings to an in vivo situation is an imperative step for reaching a better understanding of human biology. However, one can question whether the murine UUO model, discussed widely in this review, is a good model to recapitulate human fibrosis which develops over a much longer time scale. On the other hand, UUO in mice is a quick and easy way to reproduce a model of renal injury and when used in genetically manipulated animals provides a powerful method to investigate mechanisms of renal fibrosis. Whilst the idea of EMT giving rise to interstitial fibroblasts in CKD has become less acceptable following the report of Humphreys et al. (2010), the idea of PTECs undergoing varying degrees of change to become local producers of fibrillar collagens, as part of a normal response to injury, should not be rejected. This was reported in mice over-expressing MMP2 in PTECs (Cheng et al. 2006) and is supported by observations on human CKD biopsies (Rastaldi et al. 2002). It may take months for such changes to occur in mice, or years in man, but if they do, they could conceivably make important contributions to fibrosis around tubules in the development of CKD.