We read with great interest the article by Jeliazkova et al. The authors found that activation of Notch2 is able to reprogram both embryonic hepatoblasts and adult hepatocytes toward biliary cell differentiation. Also, the oncogenic potential of Notch2 is suggested by the development of premalignant lesions in Notch2-overexpressing livers. The latter results further substantiate our and others findings on the role of the Notch pathway in cholangiocarcinogenesis.[2, 3] Although our study was acknowledged by the authors, the interpretation of some of our findings was not entirely correct. Jeliazkova et al. speculated that Notch2 might be more important than Notch1 in hepatocarcinogenesis, since Notch1 would require the coexpression with AKT to be oncogenic. We instead showed (our supporting Fig. 1) that overexpression of Notch1 alone is sufficient for intrahepatic cholangiocarcinoma (ICC) development (Fig. 1A-C), which is tremendously accelerated by AKT coexpression. Concerning the origin of ICC from adult hepatocytes in AKT/Notch1 mice, both Jeliazkova et al. and Cardinale et al. questioned the specificity of the tracing model we used. We would like to emphasize here that we put much effort into establishing that using the capsid from adenoassociated virus 8 together with a transthyretin promoter afforded hepatocyte-specific marker gene activation in the mouse liver. Furthermore, we performed morphological studies, further explained in the present letter, that clearly demonstrate the hepatocellular origin of ICC in AKT/Notch1 mice. If we assume that ICC derive from progenitor cells in our experimental system, it would mean that the injected plasmids bypass the liver acinus against the sinusoidal blood flow to reach the utmost periportal area, then transfect progenitor cells without being incorporated by hepatocytes along the way. This hypothesis is highly unlikely since it contradicts the physiologic principle of hydrodynamic gene delivery, known to target nearly exclusively hepatocytes located in acinar zone 3 (i.e., close to the hepatic venule). Hypothetically, small amounts of plasmids might reach the canals of Hering and be incorporated by progenitor cells. However, in all models that we generated using this technique, transfected cells were fully differentiated hepatocytes located in zone 3 and, thus, preneoplastic lesions developed always in zone 3 vein proximity. The morphological demonstration, showing that affected single cells in AKT/Notch1 mice were never located in zone 1 but always in zone 3 (Fig. 1D-I) is a proof of the physiologic principle of the method, in line with all our models, and in absolute contradiction to the progenitor-cell hypothesis. Furthermore, electron microscopy showed the presence of tight junctions between transfected and normal hepatocytes (supporting Fig. 10), thus indicating their hepatocellular nature.
Matthias Evert, M.D.1Frank Dombrowski, M.D.1Biao Fan, M.D., Ph.D.2Silvia Ribback, M.D.1Xin Chen, Ph.D.2Diego F. Calvisi, M.D.1
1Institute of Pathology University of Greifswald Greifswald, Germany2Department of Bioengineering and Therapeutic Sciences University of CaliforniaSan Francisco, CA