A misinterpretation of epithelial plasticity can be traced to the belief that an epithelial cell undergoing EMT should become a new fibroblast. However, the definition of EMT does not require that an epithelial cell becomes any specific type of a cell. EMT is defined by phenotypic and functional changes that are reminiscent of mesenchymal cells. This in no way means that an epithelial cell will become a fibroblast. Strictly speaking, the term 'mesenchyme' is synonymous with 'primitive connective tissue', that is, any tissue that 1) is neither epithelial nor muscular nor nervous (see Ross et al, pages 54 and 115 10); and 2) the function of which is to provide an extracellular matrix (ECM) used as a support for these other cell types to function properly. As was recently emphasized, the migration of an epithelial cell out of the tubular structure into the interstitium where it will produce matrix (and thereby act as a new fibroblast) is but the 'extreme' form of EMT, not the essential criterion for the diagnosis of EMT 911.

The term 'EMT' is by definition limited to the molecular events occurring in epithelial cells, but the presence of EMT reflects a more global process, affecting other cellular populations. Thus, the upregulation in the epithelium of the expression of mesenchymal genes is also seen in the renal interstitium. In IgA nephropathy, focal and segmental glomerulosclerosis or diabetic nephropathy, interstitial expression of α-smooth muscle actin (α-SMA) and FSP1 correlates with the progression of the disease, regardless of the origin of these FSP1-positive cells 121314. The reason why immunohistochemistry is of invaluable help and of high diagnostic value is that it is extremely sensitive in detecting the expression of mesenchymal proteins, specifically in highly differentiated epithelial cells subjected to some form of injury. The detection of structural markers is much easier when they are expressed de novo in cells, and more difficult when they are upregulated only in mesenchymal cells such as fibroblasts, which already have basal expression of such markers, in endothelial cells with a virtual cytoplasm, or in poorly defined cell types such as pericytes. Consistent with the idea that EMT is a part of a global fibrogenic process, the use of DNA microarrays has allowed quantitative measurement of the expression of mesenchymal genes in the total cortex from renal biopsies, and is 'generally' (that is, cortex-wide) upregulated in fibrosing allografts 151617. Similarly, preliminary results indicate that mRNA expression of EMT genes can be used as non-invasive surrogate markers for renal fibrosis in urine, even though the origin of the urinary cells that are being studied is not determined 18.

A milestone in the study of Type 2 EMT was the identification of FSP1 by Strutz et al., which has opened the door to the study of the epithelium as an active participant in tissue fibrosis4. Currently, the expression of FSP1 by the tubular epithelium is considered to be an incontrovertible criterion for the 'diagnosis' of EMT. We believe that this should be reconsidered for three reasons. First, the sensitivity of FSP1 as a fibroblast marker has been contested, as many authentic fibroblasts do not express it 14192021. Second, its specificity is a persistent matter of debate, as several studies have reported that FSP1 can be expressed by leukocytes, particularly macrophages 19212223242526. Even though the antibodies used to detect leukocytes in some of these studies have themselves overlapping specificities, a recent publication using a mouse model of liver injury showed that most FSP1-positive cells stained positively for F4/80, which is at present the most specific marker for macrophages 21. Even though the aberrant expression of FSP1 by renal epithelial cells is still highly specific to fibrogenic renal diseases, this does not necessarily imply that FSP1-positive epithelial cells will generate a neofibroblast (incidentally, this is true for all EMT markers). Thus, although FSP1 has been shown to induce EMT in vitro 42728, facilitating the metastasis of cancer cells, the consequences of its expression in the tubular epithelium during fibrogenesis remain to be determined 29. Last, its detection by immunostaining within the renal epithelium undergoing EMT is rare 41430 and sometimes technically difficult to obtain (for example, it will be negative if the fixative used is alcohol-formalin-acetic acid 30). Overall, these observations disqualify FSP1 as a usable biomarker in kidney biopsies.

It is important to use multiple markers simultaneously to detect the coordinated changes associated with EMT in the epithelial structures 31. In humans, where only the observation of renal biopsies is possible, the migration of tubular cells outside of the basal membrane is obviously impossible to 'see'. Furthermore, the evidence that tubular cells display a decrease in the expression of epithelial markers is always questionable because the various tubule segments differentially express the classic epithelial proteins such as cadherins, cytokeratins and others, even under physiological conditions 3233. For example, E-cadherin is minimally expressed by proximal tubular cells in humans, thus the assertion that E-cadherin is lost or decreased necessarily implies that a distal tubule is being studied 30.

By contrast with animal studies the effective clinical diagnosis of EMT means that, rather than the loss of an epithelial phenotype, it is the aberrant acquisition of mesenchymal properties that is being detected in epithelial cells. Two kinds of EMT markers can be used in humans, and can be classified according to their structural (for example, changes in the composition of the cytoskeleton) or functional (, for example, ECM production) relevance. Schematically, the expression of structural proteins (for example, vimentin) is a feature present in quiescent mesenchymal cells, whereas the expression of functional proteins constitutes a feature of activated mesenchymal cells. The heterogeneity of these markers also has several advantages, as it allows the detection of EMT-like changes at different stages, depending on the characteristics of the disease and the sensitivity and specificity needed for diagnosis (Table 1). For example, vimentin is an intermediate-filament constituent of the cytoskeleton of mesenchymal cells. Its expression is associated with the phenotype of cells, specifically shape, motility and adhesion 34. In tumors, it plays a role in apoptosis, invasion and metastasis, and is thus a new therapeutic target 353637. Vimentin knockout mice have delayed wound healing 38. Vimentin seems to play a role in some, but not all, models of renal injury and repair 3940. It is not expressed in the normal adult tubular epithelium, but it is strongly expressed by by interstitial cells and the foot processes of podocytes, and faintly by endothelial cells 4142. This opportunely serves as an internal positive control. This marker lacks specificity as a mesenchymal marker, but it is very sensitive in that vimentin staining is absolutely negative in normal renal tubular cells and positive at a very early stage after injury. Importantly, expression of vimentin does not preclude the possibility of a reversible alteration of the epithelium; the epithelial expression of vimentin was seen in a model of transient renal injury followed by complete repair of tissue 4344. Regardless, the sensitivity of its detection makes it an excellent candidate for early screening of tubular injury.

Conversely, α-SMA, a marker of activated fibroblasts, and heat-shock protein (HSP)47, a marker of collagen production, are less sensitive markers, but are more clearly linked to fibrosis 454647. β-catenin is a marker of particular interest here; depending on its localization in the cell, it is either a structural epithelial marker or a mesenchymal functional marker. Thus, whereas β-catenin connects E-cadherin molecules to the actin cytoskeleton in epithelial cells that behave like epithelial cells, it may translocate in the cytosol and subsequently into the nucleus to act as a mesenchymal transcription factor in epithelial cells that engage in EMT. β-catenin was actually shown to be a key activator of EMT, notably within the Wnt pathway, in various experimental models such as ureteral obstruction, the Fisher to Lewis allograft model of renal transplantation, and adriamycin nephropathy 48495051. Although any injury will probably trigger the expression of some structural proteins characteristic of fibroblasts, only an extended or repetitive injury will lead to a functional 'EMT' and thereby to the production of ECM by the surviving epithelium.

The expression of EMT markers in native kidneys was investigated by Jinde on 127 patients with IgA nephropathy, rapidly progressive glomerulonephritis and minimal-change disease (MCD) 45. One year later, another study evaluated 133 renal biopsies with nine different diagnoses 46. Impressively, these two studies found frequent expression of EMT markers in all the renal diseases with the exception of MCD, where it was rare or absent. More than just a collateral phenomenon associated with renal disease, EMT markers were shown to be associated with the severity of interstitial fibrosis and the impairment of renal function 4546. In patients with multiple myeloma and the highly fibrogenic condition cast nephropathy, the expression of EMT markers was highest, including in morphologically preserved tubules 52.

By contrast, EMT-marker expression was not found in patients with overt proteinuria due to MCD 454652, in keeping with the fact that the high proteinuria seen in MCD is rarely associated with interstitial fibrosis or renal failure 53. This strongly supports the use of EMT as a biomarker of renal fibrogenesis. Furthermore, in the study by Rastaldi et al., EMT markers were also correlated with renal function in a subgroup of 45 patients with little interstitial fibrosis or infiltration, emphasizing the excellent sensitivity of these markers 46. Other studies have focused on patients with one specific renal disease. In an ingenious study, Yamaguchi showed a correlation between EMT markers in urinary podocytes and the severity of diabetic nephropathy 54. Ultimately, it is in renal allografts that the association of EMT markers with severity of renal disease has been best documented. Vongwiwatana showed that EMT markers were associated with serum creatinine, proteinuria and a history of T-cell-mediated rejection 47. In another set of patients with stable renal function 3 months after transplantation, we found that EMT markers were associated with higher serum creatinine, cold-ischemia time and subclinical acute rejection 30.