Lack of Phenobarbital-mediated Promotion of Hepatocarcinogenesis in Connexin32-Null Mice¹

Connexins are subunits of gap junctional channels, through which directly neighboring cells exchange low molecular weight molecules,such as ions, second messengers, and cellular metabolites. Intercellular communication mediated via gap junctions plays an important role in tissue homeostasis, embryonic development, and in cancer (for recent reviews, see Refs. 1, 2, 3, 4). The important role of gap junctional communication in multistage carcinogenesis is supported by several lines of evidence: Gap junctional proteins are often decreased in tumor tissue (5, 6, 7, 8, 9), and overexpression of connexins suppresses tumorigenicity of tumor cells(10). Moreover, targeted disruption of the Cx32³gene in mice is associated with enhanced occurrence of both spontaneous and chemically induced liver tumors (11, 12). The introduction of activated oncogenes into cells blocks GJIC (for review,see Ref. 4), and similar effects were seen if cells were exposed to tumor promoting agents in vitro. A large number of chemicals, including 12-tetradecanoyl-phorbol-13-acetate, PB,certain polyhalogenated biphenyls, and insecticides like endosulfane,chlordane, dieldrin or DDT, all known to promote carcinogenesis in skin or liver, have been tested, and many of these, albeit not all, were found to block GJIC (for reviews, see Refs. 13, 14, 15). Tumor promoters may affect GJIC by interference with cell signaling pathways that lead to direct blockage, e.g., by posttranslational modification of connexins or an increase in intracellular Ca²⁺ (for review, see Refs. 1 and3). Moreover, a decrease in expression of connexins has been reported for various liver tumor promoters (7, 9, 16, 17, 18).

The mechanisms by which GJIC affects the proliferation of transformed cells are not entirely understood, but two distinct types of cell-cell communication, termed homologous and heterologous communication, have to be considered (3), the latter describing communication between transformed and surrounding normal cells. Lack of heterologous GJIC between tumor and normal cells has been observed in vivo; e.g., rat liver preneoplastic foci failed to communicate with their surrounding normal hepatocytes (8). It has been suggested that loss of heterologous GJIC uncouples preneoplastic and neoplastic cells from growth-restraining signals of surrounding normal cells (19). Accordingly, a clonal expansion of tumor or tumor precursor cells in solid tissues would be expected to result from inhibition of heterologous GJIC by tumor promoters, thus explaining their enhancing activity on carcinogenesis. We have now performed an initiation-promotion experiment in Cx32-wild-type and Cx32-null mice and found that the tumor promoter PB enhances DEN-induced hepatocarcinogenesis in Cx32-competent, but not in Cx32-deficient mice. This observation demonstrates the important role of the Cx32 tumor suppressor protein during the promotional phase of liver carcinogenesis.

Cx32-deficient mice used in the present study were generated by standard methods of targeted homologous recombination leading to a mixed genetic background of C57BL/6 and 129Sv inbred strains(20). Cx32 heterozygous female C57BL/129Sv(Cx32^+/−) mice were crossed with male C3H/He (Cx32^Y/+) mice to yield male Cx32^Y/+ and Cx32^Y/−offspring. Tail tips were taken, and Cx32-genotyping was performed by standard PCR as recently described (12). Two separate carcinogenesis experiments were performed. In experiment 1, mice were i.p. injected with a single dose of DEN (90 μg/g of body weight) at 6 weeks of age, while 10 μg/g body weight of the carcinogen were injected into 12–15-day-old mice in the second experiment. After a treatment-free interval of 3 weeks, DEN-treated Cx32^Y/+ and Cx32^Y/− mice were randomly assigned to groups, which were either kept on a standard diet or on a diet containing 0.05% PB until sacrifice. The numbers of mice in the various treatment groups are listed in Tables 1,2,3. Mice were killed 39 weeks (experiment 1) or 25 weeks (experiment 2) after the start of PB treatment, and livers were inspected for macroscopically visible tumors. In experiment 1, where large tumors were apparent because of the longer duration of the experiment, the number and size of tumors were recorded, except for those animals where the tumor response was such that individual tumor counts were no longer possible. In both experiments, livers were frozen on blocks of dry ice and frozen sections were stained enzyme-histochemically for G6Pase activity (21). Number and volume fraction in liver of G6Pase-deficient lesions were quantitated by means of a computer-assisted digitizer system (22) and analyzed using standard stereological techniques (23).

Cx32 protein expression was analyzed by standard immunohistochemical protocols. In brief, frozen liver sections were fixed in ethanol(−20°C; 5 min) and air-dried, and endogenous peroxidase activities were blocked by treatment with methanol/H₂O₂. Sections were then incubated with polyclonal goat anti-Cx32 antibodies(SantaCruz Biotechnolgy, Santa Cruz, CA) at a 1:250 dilution, and antibody binding was visualized using horseradish peroxidaseconjugated antigoat-IgG antibodies (SantaCruz), followed by staining with 3-amino-9-ethylcarbazole/H₂O₂. Specimens were counterstained with hematoxylin.

For analysis of GJIC, groups (each n = 4) of adult Cx32^Y/+ mice of strain B6C3F1 (Charles River) were treated with a PB (0.05%)-containing diet for 2 weeks or were kept untreated. After killing, the lateral liver lobe was immediately transferred to ice-cold buffer (Krebs-Henseleit buffer supplemented with 25 mm glucose) and used for preparation of 250-μm-thick sections, using Leica microtome VT1000S. The sections were microinjected in a microscope submerging chamber at 37°C under constant flow of oxygenated buffer (Krebs-Henseleit buffer supplemented with 25 mm fructose) after an incubation period of 60–90 min at 37°C in an artificial oxygen atmosphere (95% O₂ + 5%CO₂). Microinjection of the fluorescent dye Alexa Fluor488 (Molecular Probes Europe BV, Leiden, the Netherlands) was performed by iontophoresis.⁴Before and after each microinjection, membrane potential was measured,and only those injections were considered where the potential was<−20 mV. Gap-junctional coupling was measured 10 min after injection with a special cross-field ocular under UV illumination with a Zeiss Axioskop FS determining the area of distribution, which was used for recalculation of the number of cells being affected.

For analysis of PB effects on liver weight and drug metabolizing enzymes, adult Cx32^Y/+ and Cx32^Y/− mice (three per group) were treated for 4 weeks with a PB (0.05%)-containing diet or were kept on control diet. Liver microsomes were prepared by standard procedures, and the contents of cytochrome P450 and cytochrome b₅ were determined by difference spectrometry as described (24).

Statistical analysis of data were performed using InStat for Windows NT(version 3.02) from GraphPad Software (San Diego, CA).

In this experiment, mice were given a single injection of DEN at the age of 6 weeks, and groups of mice were subsequently kept for 39 weeks on a PB-containing or control diet until sacrifice. At the end of the experiment, the tumor prevalences in Cx32^Y/+ mice kept on control or PB-containing diets were 76% (13 of 17) and 94%(17 of 18), respectively, whereas the corresponding values in Cx32^Y/− mice were 87% (13 of 15) and 100% (15 of 15), respectively. Macroscopically visible tumors were counted and classified according to size. In some of the animals, however,individual tumor counts were not possible because the entire liver appeared tumorigenic. The phenotype of livers from mice showing this extremely strong tumor response was therefore categorized as“polyadenomatosis.” Mice with polyadenomatosis were only observed in groups receiving PB, but the frequency of occurrence differed significantly (P < 0.01) between Cx32^Y/+ and Cx32^Y/− mice:whereas 50% of the PB-treated Cx32^Y/+ mice showed polyadenomatosis, only 7% of Cx32^Y/−mice showed this extreme tumor response (Fig. 1). Because the numbers and size of individual tumors could not be quantified accurately in mice with polyadenomatosis, these animals were excluded from the tumor analyses described below. When animals were grouped according to their largest tumor observed (Fig. 1), ∼50% of Cx32^Y/+ control mice not treated with PB showed tumors <2 mm in diameter, whereas only ∼10% had tumors of ≥10 mm. By contrast, an inverse relation was found in PB-treated Cx32^Y/+ mice, where ∼55% of mice had tumors≥10 mm, and only ∼20% had tumors <2 mm in diameter. This dramatic effect caused by PB treatment in Cx32-wild-type mice was clearly absent in mice of the Cx32-null genotype (Fig. 1). Similarly, the multiplicity of tumors stratified by size demonstrated a ∼6-fold increase by PB in the number of very large tumors (>7 mm in diameter) in Cx32^Y/+ mice, whereas no such increase was observed in Cx32^Y/− mice (Table 1).

The number and size of G6Pase-deficient neoplastic lesions in liver were subsequently quantified in 59 of 65 mice; six livers were not available for G6Pase staining. The numbers of lesions were increased by PB both in Cx32^Y/+ and Cx32^Y/− mice. There was, however, a highly significant ∼5-fold increase in the volume fraction occupied by G6Pase-deficient tissue in livers of Cx32^Y/+ mice treated with PB when compared with their counterparts not treated with the barbiturate (P = 0.0003). By contrast, no such increase was seen in Cx32^Y/− mice (Table 2).

In this experiment, 12–15-day-old mice were treated with DEN followed by 25 weeks of treatment with PB-containing or control diet until sacrifice. According to literature data (Ref. 25 and references cited therein) and our own observations made in other mouse strains,⁵we expected an inhibitory rather than a promotional effect of PB on hepatocarcinogenesis when using infant instead of adult mice. Because of the shorter duration of the experiment, analysis of the carcinogenic outcome in this experiment was based exclusively on the evaluation of G6Pase-deficient liver lesions. As summarized in Table 3, the number of G6Pase-deficient lesions was not significantly affected by PB treatment, neither in Cx32^Y/+ nor in Cx32^Y/− mice. By contrast, the mean volume fraction in liver occupied by G6Pase-deficient tissue was reduced by PB in both strains, although the effect was only significant in Cx32^Y/− mice (P = 0.0007).

A comparison of the effects of the Cx32 gene-knockout observed in animals not treated with PB in experiment 1 and 2 (see Tables 2 and 3)shows that G6Pase-deficient lesions were generally much less frequent in mice treated with DEN at adult age, resulting in an overall smaller volume fraction in liver, although a 9-fold higher carcinogen dose was used than in experiment 2, where DEN treatment was performed at infancy. This result was not unexpected because infant mice are much more susceptible to hepatocarcinogenesis, presumably because of the high rate of hepatocyte proliferation at this age (26).

The effect of PB on Cx32 expression and GJIC in Cx32^Y/+mice was investigated as follows. Sections from livers of mice from experiment 1 and 2 were stained with anti-Cx32 antibodies. In normal liver, there was no obvious effect of the barbiturate on the expression and intracellular localization of the Cx32 protein, as demonstrated by the representative examples shown in Fig. 2,A. Liver tumors from Cx32^Y/+ mice showed either unchanged or even slightly increased Cx32 staining (Fig. 3,B), whereas the staining reaction was decreased or totally absent in others (Fig. 3,A). In experiment 1, ∼60% of tumors showed positive or unchanged Cx32 staining in the absence of PB treatment, whereas ∼40% showed decreased or negative staining. In PB-treated Cx32^Y/+ mice, however, this relation was inversed: <20% of tumors were of the positive/unchanged phenotype, whereas the majority showed decreased or negative staining. It is remarkable that these effects were not seen in experiment 2,where an unchanged/positive Cx32 immunoreaction was detected in the vast majority of tumors, both in PB-treated and untreated mice (Fig. 3 C).

We then studied the effect of PB on gap junctional coupling between hepatocytes. Cx32^Y/+ and Cx32^Y/− mice were treated with PB-containing or control diet for 2 weeks, and liver thick-sections were prepared and microinjected with fluorescent dye Alexa Fluor 488 followed by measurement of dye spreading into adjacent cells. The results demonstrated clearly that PB treatment did not decrease GJIC in the normal liver, neither in Cx32^Y/+ nor in Cx32^Y/− mice (Fig. 2 B).

Finally, the effect of PB on liver weight and drug metabolizing enzymes in liver was investigated in adult mice kept for 4 weeks on a PB-containing diet. The results of this short-term study are summarized in Table 4. PB treatment led to a significant increase in the liver:body weight ratio both in Cx32^Y/+ and in Cx32^Y/− mice. Similarly, the liver microsomal contents of cytochromes P450 and b₅were significantly elevated by PB in both strains of mice, and the extent of the response to the microsomal inducer was in the same range. This indicates that the physiological response that triggers PB-mediated liver hypertrophy and microsomal enzyme changes is not affected by the connexin defect. Drug-related increases in liver weight and liver microsomal enzymes have been shown to be correlated with their tumor promotional activity (e.g., see Refs.27 and 28). For example, within a series of different barbiturates, including PB, those chemicals that produced liver hypertrophy and increases in cytochrome P450b were all potent liver tumor promoters while structural analogues that failed to induce microsomal enzymes also failed to display significant promotional effects (29). Induction of liver growth and liver promotional activity, however, do not seem to be strictly correlated but might diverge under certain conditions, as indicated by the data of the present study.

PB has been shown to enhance hepatocarcinogenesis in mice initiated by DEN at an adult age (25, 30, 31), but inhibited tumor formation when administered to mice initiated at infancy (Ref. 25 and references therein). Similar results were obtained in the present study in Cx32-wild-type mice in experiments 1 and 2, respectively. Although the suppressive effect of PB on the growth of G6Pase-deficient lesions observed in mice initiated by DEN at infancy was less dramatic than the one obtained by Lee et al. (25) and only significant in Cx32^Y/− mice, it was in the range of effects seen in a comparable study with male B6C3F1 mice performed earlier(32). The mechanisms of this paradoxical effect of the barbiturate are not entirely clear. It has been suggested(25) that PB shows opposing effects on basophilic and eosinophilic lesions in liver: in infant mice, the predominant types of lesions produced by DEN were basophilic hepatocellular adenomas, and the proliferation of these was suppressed by PB. In adult mice,however, eosinophilic lesions predominated in PB-treated mice,suggesting that the clonal outgrowth of this type of lesion was preferentially promoted by the barbiturate (25). In accordance with these results, most of the lesions in experiment 2 were basophilic, both in control and PB-treated mice, whereas the vast majority of lesions promoted by PB in experiment 1 were of a nonbasophilic/eosinophilic phenotype (data not shown).

We have previously demonstrated that mice deficient for Cx32 show a strong increase in growth of both spontaneous and DEN-induced preneoplastic and neoplastic lesions in liver, in comparison with their wild-type counterparts. This suggests that Cx32 inhibits the proliferation of transformed hepatocytes during hepatocarcinogenesis, i.e., it acts as a tumor suppressor protein (11, 12). This effect is mediated most likely through cell-cell communication via gap junctions. GJIC in livers of Cx32-null mice was found to be decreased by ∼80% estimated by electrophysiological measurements (33) and by direct dye injection technique into liver slices (see Fig. 2 B).⁴ We have now found that treatment of DEN-initiated adult Cx32-wild-type mice with PB results in tumor promotion, whereas this effect is lacking in Cx32-null mice. This demonstrates that functional Cx32 is required for tumor promotion by PB and suggests that the promoting agent counteracts tumor suppression by the connexin.

A decrease in GJIC is also produced by many tumor promoters (for review, see Refs. 14 and 15). These results have strengthened the hypothesis that suppression of GJIC is a key mechanism by which tumor promoters enhance the clonal expansion of potential cancer cells. Unexpectedly, PB did not affect dye transfer in acute slices of wild-type mouse liver, as shown in the present study nor was there an alteration in the expression or intercellular localization of Cx32 in liver from PB-treated wild-type mice. Similarly, there was no effect of PB on GJIC in Cx32-null mice that express only Cx26. This result is at variance with findings in rat liver where PB caused inhibition of dye transfer and concomitant decrease in Cx32-positive immunostaining (16, 34, 35) and with results from in vitro studies with primary mouse hepatocytes where PB led to partial suppression of GJIC(36). The reason for this discrepancy is not clear. In any case, however, normal hepatocytes are not the main target population for the tumor promoting effects of PB.

Although there was no detectable effect of PB on Cx32 expression and GJIC in normal liver from Cx32^Y/+ mice, the barbiturate led to a significant increase in the frequency of tumors with decreased or negative expression of Cx32 in experiment 1, where PB promoted hepatocarcinogenesis in Cx32-wild-type mice. Similar observations have been made earlier in a comparable study with B6C3F1 mice (13). By contrast, PB did not increase the frequency of Cx32-negative lesions in experiment 2, where no tumor promotion occurred. Potentially, there is a positive selection for tumor cells with less functional Cx32 protein because of lacking communication with their surrounding normal counterparts (heterologous GJIC). The signals exchanged between neighboring cells through GJIC for inhibition of tumor growth are presently unknown. In principle, these signals modulate cell division and/or death (apoptosis), both of which are critical for tumor development. Because Cx32-null mice were almost completely resistant to the tumor promoting activity of PB, our results clearly demonstrate that Cx32 represents an essential target for promotion of hepatocarcinogenesis by PB.

Cx32-null mice treated with the tumor initiating carcinogen DEN developed liver tumors much faster than identically treated Cx32-wild-type mice (11, 12). By contrast, our present data demonstrate an inverse situation for the carcinogenic response mediated by the tumor promoter PB. If this would apply to initiating and promoting hepatocarcinogens in general, the experimental system using Cx32-null and -wild-type mice would offer a very valuable toxicological tool to classify the carcinogens according to their predominant mode of action—tumor-initiating versustumor-promoting.

We thank Elke Zabinski for excellent technical assistance.