In this study, we report that EZH2 was detected by immunohistochemistry in nearly all the investigated HCCs, CCs, hepatoblastomas, metastatic liver tumors and several other childhood cancers. However, none of the hepatocellular or biliary adenomas, high grade dysplastic or cirrhotic nodules was positive. The ductular reactions and biliary hamartomas were also consistently negative.
Knockdown of EZH2 reversed the tumorigenicity in experimental liver tumors, suggesting that EZH2 plays an important role in HCC tumorigenesis [23, 24]. Increased expression of EZH2 was correlated with unfavourable outcome of HCC  and metastatic capacity . Cai et al.  demonstrated convincingly that EZH2 is a highly sensitive diagnostic biomarker of HCC, which can be used to distinguish it from benign liver lesions such as hepatocellular adenomas, focal nodular hyperplasias, dysplastic and regenerative nodules. There is a complete agreement between the findings of Cai et al’s  and our study that all the investigated regenerative nodules and adenomas were negative. Our results also support that EZH2 is a sensitive marker of HCC. Although not reaching statistical significance, EZH2 appeared less able to recognize well differentiated HCCs in both studies. There seems to be a borderline or “grey zone” in highly differentiated hepatocellular tumors, in which EZH2 similarly to the other markers, is not absolutely reliable. It requires further direct comparative studies, which one of the recently applied markers (Glypican3, AFP, Hsp70, EZH2) is the most sensitive [3–7]. Most likely however, a panel of these antibodies would provide the most trustworthy information and EZH2 could be a useful member of this battery. All of the adenomas in our study were beta-catenin negative. The beta-catenin status of the adenomas in the study of Cai et al.  is not indicated. It would be of interest to examine EZH2 expression in beta-catenin positive hepatocellular adenomas, which have a higher tendency for malignant transformation .
In addition to hepatocellular lesions, there are other hepatic tumors, which may raise differential diagnostic problems. For this reason we investigated EZH2 expression in the most common other tumor types of liver. EZH2 staining was positive in 96% of cholangiocarcinomas, including 3 highly differentiated Klatskin tumors. As far as we know EZH2 expression has not previously been examined in this type of tumor. We have studied a few biliary cystadenomas and tumor mimicking ductal plate malformations, all of them were negative, as well as the ductular reactions. CC also must be differentiated from metastatic liver tumors. Practically all the metastatic adenocarcinomas, regardless of their origin, were positive for EZH2. Although our case number is low, the primary tumors of colon, pancreas, breast, lung [14–21] have already been described to express this tumor marker, so the few metastases we report probably do reflect reality. That is, EZH2 similarly to other CC markers e.g. CK7, CK19, claudin 4  does not provide major help in distinguishing cholangiocarcinomas from metastatic tumors, but it does seem to be able to differentiate reactive/hamartomatous biliary structures and benign biliary tumors from malignant ones. This is important because so far no such marker has been reported. Even the recently described highly specific cholangiocarcinoma marker, αvβ6 integrin  is positive in proliferating bile ducts. The combination of these two antibodies may facilitate the diagnosis of cholangiocarcinomas.
EZH2 expression has not yet been tested on hepatoblastomas and other primitive "blastic" tumors except Ewing’s sarcoma and rhabdoid tumors which were reported positive [29, 30]. The applied antibody recognizes these tumors with high sensitivity. All but two of the examined hepatoblastomas (n = 31) and the few other childhood tumors stained positively. This is not surprising, as it is considered that the major biological function of EZH2 is to maintain the undifferentiated stage of cells . Again, the number of non hepatoblastomas investigated is quite low, but considering the very consistent positive staining it is highly unlikely that EZH2 could be used to differentiate among these childhood tumors. However, if recent therapeutic approaches targeting EZH2  were successful, our observation in childhood tumors would gain significance.