PCB 153, a Non-dioxin–like Tumor Promoter, Selects for β-Catenin (Catnb)–Mutated Mouse Liver Tumors


Polychlorinated biphenyls (PCBs) are ubiquitous environmental toxicants which act as liver tumor promoters in rodents and can be classified as either dioxin-like or non-dioxin (phenobarbital [PB])–like inducers of cytochrome P-450. Since we have previously shown that tumor promotion by PB leads to clonal outgrowth of β-catenin (Catnb)–mutated but not Ha-ras–mutated mouse liver tumors, we were interested to know whether the non-dioxin–like tumor promoter 2,2′,4,4′,5,5′-hexachlorobiphenyl (PCB 153) shows the same selective pressure during tumor promotion. Male B6129SF2/J mice were given a single injection of N-nitrosodiethylamine (90 mg/kg body weight) at 9 weeks of age, followed by 39 weeks of treatment with PCB 153 (20 biweekly ip injections of 300 μmol/kg body weight) or corn oil as a control. Animals were killed 15 weeks after the last PCB 153 injection and liver tumors were identified by immunohistochemical staining of glutamine synthetase (GS) and analyzed for Catnb, Ha-ras, and B-raf mutations. Quantitative analyses revealed that GS-positive tumors were much larger and more frequent in livers from PCB 153–treated mice than in control animals, whereas GS-negative tumors were similar in both groups. Almost 90% (34/38) of all tumors from PCB 153–treated animals contained Catnb mutations, which compares to 45% (17/37) of tumors in the control group. Ha-ras and B-raf–mutated liver tumors were rare and not significantly different between treatment groups. These results clearly indicate that PCB 153 strongly selects for Catnb-mutated, GS-positive liver tumors, which is similar to the known action of PB, a prototypical tumor promoter in rodent liver.


Polychlorinated biphenyls (PCBs) are chemically stable, highly lipophilic, and persistent environmental toxicants. Although their production in Europe and in North America was banned more than two decades ago, PCBs are still being used in many parts of the world and are present in the human food chain worldwide (Fiedler, 2001; Hites et al., 2004). Depending on their chemical structure, congeneric PCBs can be classified as dioxin-like (more coplanar) PCBs that elicit their toxic effects primarily via binding to the aryl hydrocarbon receptor (AhR) (Bandiera et al., 1982), similar to halogenated dibenzo-dioxins and -furans. Introduction of ortho chlorines (in 2, 2′, 6, or 6′ positions) of biphenyl diminishes the avidity of AhR binding. PCBs with the backbone consisting of two ortho chlorines and two para chlorines resemble phenobarbital (PB) in their mode of cytochrome P-450 induction (Denomme et al., 1983), presumably acting through the constitutive androstane receptor (CAR) (Sanders et al., 2005; Waxman, 1999). PCBs with two or more ortho chlorine substituents may also bind to the pregnane X receptor and, like dexamethasone, induce cytochrome 3A proteins (Hurst and Waxman, 2005; Sanders et al., 2005; Schuetz et al., 1998). PCBs that are AhR agonists as well as those that activate CAR show tumor-promoting activity in rodent liver at doses that also induce cytochrome P-450 (Buchmann et al., 1986, 1991b).

PCBs are assumed to act as tumor promoters in rodent liver by stimulating the clonal outgrowth of initiated cells, and there is experimental evidence that different types of tumor-promoting agents may act on different initiated cell populations which are characterized by defined genotypes and their corresponding phenotypes. Experimentally induced liver tumors from mice are known to harbor activating mutations in largely three different genes, Catnb, Ha-ras, or B-raf, which is in contrast to the situation in rats where the genetic lesions are less well understood. The types of mutations observed in mouse liver tumors strongly depend on the treatment regimen used for their induction. In mice treated with a single injection of liver carcinogens, such as N-nitrosodiethylamine (DEN), a high percentage of liver tumors possess mutations in Ha-ras (Buchmann et al., 1991a; Maronpot et al., 1995) or B-raf (Jaworski et al., 2005), a downstream target in the Ras signal transduction pathway. By contrast, if liver tumors are induced in initiation-promotion experiments with PB, a prototypical tumor promoter in rodent liver, 80% of them contain activating mutations in the Catnb gene (coding for the oncoprotein β-catenin), indicating that PB strongly selects for this type of initiated liver cells (Aydinlik et al., 2001; Calvisi et al., 2004). Other liver tumor promoters such as the peroxisome proliferator WY-14,643, however, primarily select for cell populations with other types of gene mutations (Moennikes et al., 2003). Microarray analysis of global gene expression patterns in Ha-ras and Catnb-mutated liver tumors revealed characteristic differences which may be used for diagnostic purposes to identify either of the two tumor populations (Stahl et al., 2005). For example, Catnb-mutated tumors strongly overexpress glutamine synthetase (GS) and several cytochrome P-450 isoforms, which, by contrast, are almost entirely absent in Catnb wild-type, presumably Ha-ras or B-raf–mutated, tumors (Loeppen et al., 2002, 2005).

Human exposure scenarios generally include complex mixtures of different PCB congeners, dioxins and furans, raising the question of whether the individual constituents exert additive, synergistic, or possibly antagonistic effects (Schwarz and Appel, 2005). So far, most initiation-promotion experiments with individual PCBs or mixtures have been performed in rats and there is only a limited number of studies with mice (for a review, see Glauert et al., 2001). Mice would offer a valuable tool to address the question of whether dioxin-like and non-dioxin–like PCBs act on the same or different target cell populations since distinct tumor types with characteristic gene mutations can be discriminated. Previous studies with 2,3,7,8-tetrachloro-dibenzo-p-dioxin (TCDD) have shown that liver tumors from TCDD-treated mice contain Ha-ras mutations at a high frequency (Watson et al., 1995), whereas Catnb-mutated liver tumors are rarely detectable (Devereux et al., 1999). Since no comparable data are available with respect to PCBs, we analyzed the effects of 2,2′,4,4′,5,5′-hexachlorobiphenyl (PCB 153), a non-dioxin (PB)–like PCB, on Catnb, Ha-ras, and B-raf mutations in liver tumors that were induced in B6129SF2/J mice according to an initiation-promotion protocol. This study provides insight into the selection processes involved in liver tumor promotion by individual PCB congeners.



PCB 153 was synthesized, purified, and characterized as previously described (Schramm et al., 1985).

Induction of liver tumors.

Male B6129SF2/J mice were treated at 9 weeks of age with a single ip injection of DEN (90 mg/kg body weight). After 2 weeks of recovery, one group of mice received 20 biweekly ip injections of PCB 153 (dissolved in corn oil; 300 μmol/kg bodyweight per injection). Mice of the control group were given 27 biweekly ip injections of the solvent corn oil. At the age of 64 weeks (15 weeks after the last PCB 153 injection), animals were sacrificed and the livers were isolated and stored at 80°C.

Immunohistochemical analyses.

For quantification of liver tumors, frozen liver sections (10 μm) were stained immunohistochemically for GS, as recently described (Loeppen et al., 2002), using a primary anti-GS polyclonal antibody (1:1000; Sigma-Aldrich, Taufkirchen, Germany), a secondary peroxidase-conjugated anti-rabbit IgG antibody (1:20; Dako, Hamburg, Germany), and 3-amino-9-ethylcarbazole/H2O2 as substrates.

Quantification of GS-positive and GS-negative liver tumors was performed by projecting the GS-stained sections onto a digitizer screen (10- to 50-fold magnification), and the outlines of the sections and of GS-altered tumor transections were traced with a cursor and stored in a computer memory (Schwarz et al., 1989). The data were used to calculate the number of GS-positive and GS-negative tumors per square centimeter of liver tissue, the area fraction of liver tissue occupied by liver tumors, as well as the two-dimensional size class distribution of liver tumors of the two phenotypes.

Mutation analyses.

Serial liver sections were prepared and stained for the marker enzyme GS as described above. Thereafter, tissue samples were punched out with sharpened cannulas from GS-positive and GS-negative tumor transections, and the DNA regions containing the hotspot sites for mutations in the Ha-ras (codon 61), B-raf (codon 624), and Catnb gene (several codons in exon 3) were amplified by the polymerase chain reaction (PCR) using standard protocols (for PCR primers see Table 1). PCR products of the Ha-ras and B-raf gene were screened for mutations by restriction fragment length polymorphism (RFLP). The known hotspot mutations at Ha-ras codon 61 either generate new or delete available restriction enzyme recognition sites (for details, see Jaworski et al., 2005) which can be detected by the following enzymes (mutated codon 61 sequences in parentheses): Hpy188III (AAA), TaqI (CGA), XbaI (CTA), BspHI (CAT). Mutations at the hotspot site in codon 624 of the B-raf gene (mutated sequence GAG) lead to loss of a TspRI recognition site. PCR products were digested with the respective restriction enzymes (New England Biolabs, Frankfurt, Germany, or Fermentas, St Leon-Roth, Germany) and separated on 10% polyacrylamide gels. Each mutation detected was verified by classical dideoxynucleotide sequencing of the PCR products (custom sequencing by SeqLab, Goettingen, Germany). Mutations in the Catnb gene, which are located at various positions in exon 3, were analyzed by dideoxynucleotide sequencing of the PCR products (SeqLab), and the presence of mutations was confirmed by at least one independent PCR sequencing analysis. In all cases, PCR primers were used for the sequencing reactions.

Statistical analyses.

Data on quantification of GS-altered liver tumors were analyzed by the nonparametric Mann-Whitney test, and intergroup differences in mutation prevalences were analyzed by Fisher exact test using GraphPad InStat (V3.06) (GraphPad Software, Inc., San Diego, CA).


To analyze the tumor-promoting effect of PCB 153, liver sections from DEN-initiated mice that were subsequently treated with either PCB 153 or corn oil (control) were stained for the marker enzyme GS, and the total tumor response (GS-positive and GS-negative tumors, see Fig. 1A) was quantified by the use of a computer-assisted analysis system. As shown in Figure 1B, the area fraction occupied by GS-altered liver tumors was significantly higher in PCB 153–treated animals than in controls, and the number of tumor transections per square centimeter of liver tissue was also increased by PCB 153 treatment. Stratification of tumor transections into different diameter size classes revealed that PCB 153 treatment caused a marked shift toward higher size classes when compared to corn oil controls (Fig. 1C), demonstrating a strong tumor-promoting activity of PCB 153 in the present study.

We have previously shown that the liver tumor promoter PB strongly selects for Catnb-mutated, GS-positive liver lesions but not for Ha-ras–mutated liver cell populations (Aydinlik et al., 2001; Loeppen et al., 2002). To test if PCB 153 has a similar mode of action, we now analyzed liver tumors from the present study for mutations at the known hotspot sites in the Catnb, Ha-ras, and B-raf genes. In mouse liver tumors, Ha-ras mutations are located almost exclusively in codon 61 (for a review, see Maronpot et al., 1995), the hotspot mutation site in B-raf is codon 624 (Jaworski et al., 2005), whereas Catnb mutations occur at various codons (e.g., 33, 37, 41, 45, and others) located in exon 3 (formerly referred to as exon 2; Aydinlik et al., 2001; Calvisi et al., 2004; Devereux et al., 1999). For mutation analysis, tumor tissue samples were taken with punching cannulas from GS-stained liver sections and used for PCR amplification of Catnb, Ha-ras, and B-raf genomic regions containing the respective mutation sites. Thereafter, Catnb mutations were analyzed by direct sequencing of the PCR products to allow identification of all possible mutations at the relevant codons. Ha-ras and B-raf mutations were detected by RFLP analysis using restriction enzymes that are diagnostic for the known mutation types, and the presence of mutations was verified by direct sequencing (for typical examples, see Fig. 2). A summary of our data on the frequencies and types of Catnb, Ha-ras, and B-raf mutations detected in liver tumors from PCB 153–treated and control mice is given in Table 2. In both treatment groups, tumors with Ha-ras or B-raf mutations were relatively rare and their frequencies were not significantly different, albeit somewhat lower in PCB 153–treated animals. By contrast, the frequency of Catnb-mutated liver tumors was highly significantly increased in PCB 153–treated mice relative to control mice along with a decrease in the frequency of liver tumors with unknown types of gene mutations. Samples taken from normal parts of the liver did not contain any detectable mutations in the three genes analyzed.

Since tissue samples for mutation analysis were taken from GS-stained liver sections, we were able to directly correlate mutations with the GS-phenotype of tumors. In accordance with previous observations (Cadoret et al., 2002; Loeppen et al., 2002; Stahl et al., 2005), Catnb mutations were exclusively found in GS-positive liver tumors (51/51) but never in GS-negative liver tumors which contained Ha-ras, B-raf, or unknown gene mutations instead. Data on area fractions and numbers per square centimeter of liver tissue of the different genotypes show that Catnb-mutated, GS-positive tumors are much larger and more frequent in livers of PCB 153–treated mice than in corn oil–treated controls (Fig. 3), indicating that this cell population is the primary target for the tumor-promoting action of PCB 153.


PCB 153 is a non-dioxin–like PCB congener which is known to act as a tumor promoter in rodent liver and causes enzyme induction patterns similar to the prototypical tumor promoter PB (Buchmann et al., 1986). To address the question whether PCB 153 and PB have similar modes of tumor-promoting activity, we analyzed liver tumors from an initiation-promotion experiment with PCB 153 for mutations in the Catnb, Ha-ras, and B-raf genes. In addition, liver tumors were analyzed for expression of the marker enzyme GS which is known to be overexpressed in liver tumors with activated β-catenin (Cadoret et al., 2002; Loeppen et al., 2002; Stahl et al., 2005). Our results demonstrate that PCB 153 strongly stimulates the outgrowth of Catnb-mutated, GS-positive liver tumors, whereas the prevalence of Ha-ras or B-raf–mutated, GS-negative tumor populations is unaffected or even slightly reduced. These observations strongly resemble those obtained in previous studies with PB (Aydinlik et al., 2001; Calvisi et al., 2004). Overall, the types of mutations observed in either of the three genes correspond to those known to occur in mouse liver tumors under different experimental settings (e.g., see Aydinlik et al., 2001; Buchmann et al., 1991a; Devereux et al., 1999; Jaworski et al., 2005; Maronpot et al., 1995; Watson et al., 1995).

Many of the effects of non-dioxin–like PCBs, such as PCB 153, are assumed to be mediated via activation of the nuclear receptor CAR (Sanders et al., 2005; Waxman, 1999). It has recently been shown in a study with CAR null mice that this receptor is required for the tumor-promoting action of PB (Yamamoto et al., 2004), and CAR-mediated pleiotropic transcriptional responses may drive the processes underlying selection for Catnb-mutated, GS-positive liver tumors by PB (Stahl et al., 2005). Interestingly, CAR is expressed in Catnb-mutated liver tumors (Hailfinger et al., 2006) but downregulated in Ha-ras–mutated liver tumors which are refractory to promotion by PB (Stahl et al., 2005). In analogy to PB, CAR activation may also be causally involved in tumor promotion by PCB 153 and other non-dioxin–like congeners. Dioxin-like PCBs, by contrast, act primarily via binding to the AhR, similar to dioxins and furans, and thereby activate a different set of cellular responses. It is therefore reasonable to assume that tumor promotion by this class of PCBs relies on other cellular targets which drive selection for other types of initiated cell populations. This assumption is supported by recent observations showing that TCDD, the most potent AhR agonist, selects primarily for liver tumors containing Ha-ras mutations (Watson et al., 1995) but not for Catnb-mutated liver tumors (Devereux et al., 1999), which is in striking contrast to our present observations with regard to PCB 153.

The question of whether dioxin-like and non-dioxin–like PCBs act on the same or different target cell population is important for our understanding of interactive biological effects of PCBs and other organohalogen compounds present in complex mixtures (Schwarz and Appel, 2005). Recent studies have indicated that mixtures of PCBs may produce additive, synergistic, or even antagonistic tumor-promoting effects when compared to the individual compounds alone. For example, additive or synergistic effects have been shown for combinations of PCB 77 and PCB 52 (Sargent et al., 1991), PCB 153 and PCB 126 (Bager et al., 1995), PCB 105 and PCB 153 (Haag-Gronlund et al., 1998), or PCB 153 and mixtures containing PCB 126 and TCCD (Van der Plas et al., 1999). Other studies, however, have described antagonistic effects for combinations of PCB 153 and PCB 126 (Dean et al., 2002; Haag-Gronlund et al., 1998), PCB 105 and PCB 126 (Haag-Gronlund et al., 1998), or PCB 77 and PCB 153 (Berberian et al., 1995; Tharappel et al., 2002). Although there may be several different explanations for such interactive effects, an attractive hypothesis is that different congeners act on distinct subpopulations of initiated target cells, e.g., carrying either Catnb or Ha-ras mutations. Based on theoretical considerations one would expect that PCBs stimulating different populations of initiated cells produce additive effects when coadministered since they activate independent pathways in parallel. When acting on one and the same target cell population, however, mixtures of PCBs collectively feed into the same pathway, which would cause an increase in the effective total dose. In this case, synergistic effects might occur since liver tumor promoters often show nonlinear dose relationships with overproportional effects at higher doses of exposure (for review, see Schwarz, 1995, and references therein). On the other hand, non-dioxin–like PCBs may inhibit proliferation of the target cell population of dioxin-type compounds (e.g., see discussion in Schwarz and Appel, 2005), which would result in less than additive effects after coexposure. Such considerations are not only relevant for our understanding of basic mechanisms but may also have important implications for risk assessment of PCB mixtures present in the human environment. Studies addressing this issue are currently under way in our laboratory.

In summary, we have identified Catnb-mutated, GS-positive liver tumors as the primary target for the promoting action of PCB 153 in mice. The types of gene mutations selected for by dioxin-like PCBs are not yet known but may be similar to those observed in liver tumors from TCDD-treated mice which largely harbor Ha-ras mutations (Watson et al., 1995). Therefore, further analyses are required with regard to selection processes by dioxin-like PCBs to create a firm scientific basis for elaboration of possible interactive effects of complex mixtures of PCBs and other liver tumor–promoting agents.


We acknowledge the excellent technical assistance of Elke Zabinsky. This work was financially supported by the Federal Institute for Risk Assessment, Berlin, Germany, the National Institutes of Health (ES 07380, ES 013661), the Department of Defense (DAMD 17-02-1-0241), and the Kentucky Agricultural Experiment Station. The financial support of the Alexander von Humboldt Foundation is also gratefully acknowledged.