Is it the end of the line for the EMT?


  • Tatiana Kisseleva M.D., Ph.D.,

    1. Division of Gastroenterology Department of Medicine University of California San Diego School of Medicine San Diego, CA
    Search for more papers by this author
  • David A. Brenner M.D.

    Corresponding author
    1. Division of Gastroenterology Department of Medicine University of California San Diego School of Medicine San Diego, CA
    • Division of Gastroenterology, Department of Medicine, University of California, San Diego, School of Medicine, 9500 Gilman Drive #0602, Biomedical Sciences Building 1318, La Jolla, CA, 92093-0602
    Search for more papers by this author
    • fax: 858-822-0084

  • Potential conflict of interest: Nothing to report.

See Article on Page 1685

Epithelial-to-mesenchymal transition (EMT) is an important biological concept that describes the reversible transition of differentiated epithelial cells into mesenchymal cells with increased motility and changes in gene expression. Three types of EMT have been proposed: Type 1 EMT describes the invasion of transition cells into the mesenchyme during development. Type 2 EMT describes the transition into mesenchymal cells, potentially including myofibroblasts during wound healing and repair. Type 3 EMT describes invasion of cancer cells, including those migrating into the circulation to generate distant metastases.1


AFP, alpha-fetoprotein; α-SMA, α-smooth muscle actin; EMT, epithelial-to-mesenchymal transition; FSP1, fibroblast-specific protein 1; HSC, hepatic stellate cell; TGF-β, transforming growth factor β.

Fibrosis studies conducted in different organs including the liver have demonstrated that myofibroblasts are the primary source of extracellular matrix.2 Myofibroblasts are immunophenotypically characterized by a stellate shape, expression of abundant pericellular matrix, and fibrotic genes (α-smooth muscle actin [α-SMA], nonmuscle myosin, fibronectin, vimentin).3 Ultrastructurally, myofibroblasts are defined by prominent rough endoplasmic reticulum, a Golgi apparatus producing collagen, peripheral myofilaments, fibronexus (no lamina), and gap junctions.3 Myofibroblasts are implicated in wound healing and fibroproliferative disorders.4-6 In response to fibrogenic stimuli, such as transforming growth factor β1 (TGF-β1), myofibroblasts express α-SMA, secrete extracellular matrix (fibronectin, collagen type I and III), obtain high contractility, and change phenotype (production of the stress fibers).7

But where do these myofibroblasts come from? Hepatic stellate cells (HSCs) are considered to be a major source of fibrogenic myofibroblasts in the injured liver.8 Hepatic myofibroblasts may also originate from portal fibroblasts,2 bone marrow–derived mesenchymal cells, and fibrocytes,9, 10 and by the transition of epithelial cells11 and endothelial cells12 to mesenchymal cells.

What is the evidence for type 2 EMT being the etiology of the myofibroblasts in liver fibrosis? First, multiple studies in the kidney and in the lung were interpreted as showing EMT during fibrosis in those organs, and this concept was extrapolated to liver fibrosis (discussed in Zeisberg and Duffield13). Second, primary cell culture studies have clearly demonstrated that cholangiocytes and hepatocytes undergo a change in the phenotype and gene expression toward a mesenchymal cell, especially after incubation with TGF-beta, which is the cytokine most closely associated with EMT.14 Third, immunohistochemical studies both in experimental and clinical liver fibrosis have reported the coexpression of mesenchymal markers (fibroblast-specific protein 1 [FSP1], α-SMA, vimentin, desmin) with the original epithelial markers (cytokeratin-19 for cholangiocytes and albumin for hepatocytes).15 These types of studies have been recently questioned, because the allegedly EMT-specific marker FSP1 (also called S100A4) has now been shown to be expressed by nonfibroblast cells in the liver, including by a subset of monocytes.16 Fourth, at least one study11 used lineage tracing in mouse liver (see description below) to demonstrate that cells that were originally hepatocytes (i.e., at one point expressed albumin) expressed FSP1 after a fibrogenic liver injury. However, a potential artifact in these studies is the specific lack of the anti–β-galactosidase antibodies used for cell fate mapping.

The current report by the Wells laboratory in HEPATOLOGY17 provides the strongest evidence against EMT in the liver as a source of myofibroblasts. This study uses lineage tracing generated by crossing the alpha-fetoprotein (AFP) cre mouse with the ROSA26YFP stop mouse to trace the fate of any cell ever expressing AFP (Fig. 1A). As expected, all the cholangiocytes and all the hepatocytes were genetically labeled, because they are derived from AFP-expressing precursor cells. Furthermore, AFP+ progenitor cells were also irreversibly genetically marked. The critical result was that after inducing liver fibrosis by a variety of methods, none of the resulting myofibroblasts originated from the genetically marked epithelial (AFP+) cells.

Figure 1.

Analysis of EMT. (A) It is proposed that in type 2 EMT the myofibroblasts in liver fibrosis originate from hepatic epithelial cells, consisting of hepatocytes (albumin+ [Alb+] cells), cholangiocytes (cytokeratin-19+ [K19+] cells), or progenitor cells (AFP+ cells). (B) Determining the origin of myofibroblasts using cell fate mapping. If a cell expressed AFP, it will be irreversibly genetically labeled. AFP-driven CRE labeled epithelial progenitor cells, cholangiocytes, and hepatocytes, but failed to label any HSCs or myofibroblasts. YFP, yellow fluorescent protein.

This important article corroborates and extends two previous studies in assessing the contribution of epithelial cells to myofibroblasts in liver fibrosis. The first article used the robust albumin cre mouse to mark all the hepatocytes.16 The second study used a recently developed inducible cytokeratin-19 cre mouse to mark cholangiocytes.18 Both studies failed to detect any myofibroblasts in the fibrotic liver that originated from the epithelial cells. Thus, using three independent strains of cre mice as well as independent experimental methods (fluorescence-activated cell sorting, immunofluorescence to detect myofibroblast markers, and β-galactosidase enzymatic activity), these combined studies demonstrate that hepatic epithelial cells do not contribute to experimental liver fibrosis (Fig. 1B).

So what are the caveats? The most obvious is that that these powerful techniques should be applied to assess EMT in additional experimental models in which there is severe injury to epithelial cells, such as in alcoholic liver disease. Furthermore, the short length of experimental liver fibrosis (3 weeks to 3 months) may not reflect the cellular pathophysiology that occurs in chronic human liver disease, including the reactive ductile proliferation seen in patients. Therefore, the coexpression of epithelial and mesenchymal markers in cells from biopsies from patients with primary biliary cirrhosis or primary sclerosing cholangitis requires careful analysis and follow-up. Finally, studies consistently demonstrate that injured epithelial cells change their gene expression to produce fibrogenic agonists. Injured hepatocytes produce hedgehog ligands that activate stellate cells,19 and injured cholangiocytes produce the fibrogenic cytokine TGFβ.20 Thus, although there is no current evidence that myofibroblasts originate from hepatic epithelial cells, the original concept of type 2 EMT as injury-induced changes in epithelial cells continues to provide insight into liver fibrosis.