In their letter to the editor, Gebhardt and Ueberham question the experimental evidence for our hypothesis that zonal gene expression in murine liver is controlled by two opposing signal gradients.1 Their arguments are focused on glutamine synthetase (GS), an established marker for perivenous hepatocytes, but ignore that our conclusions were based on the evaluation of about 20 different genes/proteins expressed in perivenous or periportal hepatocytes, respectively. In addition, several statements purely reflect the authors' opinion without giving reference to published literature data.
We agree that the mechanisms by which β-catenin signaling might influence GS expression are not fully understood. However, in contrast to the statement that GS expression “is somehow dependent on β-catenin signaling”, a fundamental role of β-catenin in regulation of GS expression has been clearly established by several groups: (i) GS is expressed in all hepatocytes of mice carrying an activated β-catenin transgene2 or a hepatocyte-targeted knockout of the β-catenin antagonist APC.3 (ii) By contrast, mice with hepatocyte-specific knockout of β-catenin or overexpression of the Wnt/β-catenin-signaling inhibitor DKK1 lack GS expression in liver.3, 4 Notably, livers of β-catenin knockout mice also lack expression of several cytochrome P450 isozymes,4 which are, like GS, overexpressed in liver tumors with mutated β-catenin.5 The fact that the Wnt antagonist DKK1 blocks expression of GS and other perivenous markers in mouse liver3 indicates that Wnt molecules are indeed involved in regulation of several perivenous markers as suggested in our hypothesis.1
Gebhardt and Ueberham criticize that only perivenous hepatocytes might have responded to β-catenin activation by Wnt or a GSK3β inhibitor.1 Earlier studies by Gebhardt's group have shown, however, that GS can be induced in periportal hepatocytes by co-cultivation with endothelial-like cells,6 demonstrating that periportal hepatocytes are capable of responding to external stimuli that induce “perivenous” mRNAs. In addition, preferential localization of activated β-catenin in perivenous hepatocytes has been demonstrated recently.3 Regarding Ras activation in periportal hepatocytes, we agree that no experimental data directly supporting this hypothesis are presented, but relevant indirect evidence has been cited in our recent paper.1
The statement that “many liver tumors with activated β-catenin do not express GS” is made without giving any references and does not reflect previous results showing an almost 100% congruity between mutational activation of β-catenin and GS expression, both in mouse and human liver tumors.2, 5, 7, 8 The notion that “hepatocellular tumors develop rarely in response to β-catenin mutation alone” does not apply to initiation/promotion experiments with phenobarbital which strongly selects for β-catenin, but not Ha-ras mutated liver tumors in mice.9 We routinely screen liver tumors for mutations in both genes and can exclude a coincidental occurrence of β-catenin and Ha-ras mutations, as suspected by Gebhardt and Ueberham.
Following appearance of our paper,1 other studies have been published,3, 4 supporting important parts of our hypothesis. We have also obtained new results, including serum effects on gene expression, which favor our original hypothesis. These data are described in a manuscript submitted for publication, which cannot be cited because it is presently undergoing the review process.