|Home | About | Journals | Submit | Contact Us | Français|
Mice treated neonatally with diethylnitrosamine (DEN) develop liver tumors in a male-dominant manner, reflecting the male bias in human hepatocellular carcinoma. Evidence suggests that estrogen, androgen, prolactin (PRL) and growth hormone (GH) modify liver tumorigenesis. We determined the roles of estrogen receptor-α (ERα) and prolactin receptor (PRLR) using receptor null mice, ERαKO (C57Bl/6J) and PRLR-KO (129Ola-X-C57BL/6), in the neonatal-DEN model of liver tumorigenesis. In both mouse strains, females had reduced tumorigenesis compared with males (P < 0.01), regardless of ERα or PRLR status. Tumorigenesis was not affected by ovariectomy in C57Bl/6J mice but it was increased by ovariectomy in the mixed strain, 129Ola-X-C57BL/6, regardless of PRLR status. ERαKO males had 47% fewer tumors than ERα wild-type males (P < 0.01). On the other hand, estradiol treatment protected against tumorigenesis in males only in the presence of ERα. As evidenced by liver gene expression, lack of ERα did not alter the pattern of GH secretion in males but resulted in the male GH pattern in females. These observations indicate that ERα is not required for lower tumorigenesis in females, but it is required for the protective effects of exogenously delivered estradiol. Unexpectedly, the results indicate that ERα plays a role in promotion of liver tumors in males. In addition, it can be concluded that sex differences in liver tumorigenesis cannot be explained by the sexually dimorphic pattern of GH secretion. The results also rule out PRL as the mediator of the protective effect of the ovaries.
In the period of 30 years from 1975 to 2005, there was an 89% increase in the rate of liver cancer in the USA (http://seer.cancer.gov/statistics/). Liver cancer is mainly a male disease, with approximately three times higher incidence and mortality among men than women (1,2). Although the sex differences in liver cancer may be attributed to differences in lifestyle (3–7), animal models suggest that sex hormones play a pivotal role in tumor progression. When the potent carcinogen, N,N-diethylnitrosamine (DEN) is administered to young mice [postnatal day (PND) 12—14], there is a dramatic effect of sex on the number of tumors that develop as the animals age (8–10). From endocrine ablation experiments, it was shown that ovarian estrogens protect against tumor progression, whereas testicular androgens promote tumorigenesis (8–12). Evidence also suggests that the protective effect of estrogen against chemically induced liver tumors is mediated by pituitary prolactin (PRL) (13), acting through liver prolactin receptors (PRLR) (14). On the other hand, because the sexual dichotomy in tumorigenesis was absent in growth hormone (GH)-deficient mice, it has been suggested that the male pattern of GH secretion mediates the effect of testicular androgen (15). Thus, it appears that pituitary hormones are required for steroid hormone regulation of liver tumorigenesis. However, these studies do not rule out the possibility that steroid hormones have both direct effects on the hepatocyte and indirect effects via pituitary hormones, nor do they differentiate between permissive and causative roles of the pituitary hormones on steroid action in the liver.
Considering the human epidemiologic evidence for sex differences in hepatocellular carcinoma (HCC) and experimental evidence that sex steroids influence growth and progression of tumors and tumor cells, clinical studies were undertaken to determine if endocrine manipulation could affect the course of advanced disease. Trials using the progestin, megestrol acetate, the gonadotropin releasing hormone (GnRH) agonist, leuprorelin, the anti-androgen, flutamide and the anti-estrogen, tamoxifen have been uniformly disappointing (16–19). However, a lack of effect in advanced disease may not be indicative of efficacy of a drug designed to target the tumor promotion component of the pathology. The rational design of an endocrine therapy requires a more complete understanding of the role of hormones in the tumorigenic process and how specific drugs might affect those processes. For example, the anti-estrogens, tamoxifen and raloxifene, are known to behave very differently in different tissues, producing estrogen-agonist activity in one tissue while behaving as an antagonist in another (20,21). For example, in the liver cell line, Hep3B, raloxifene induces insulin-like growth factor (IGF-I) gene transcription, whereas estradiol or tamoxifen inhibits it (22). Thus, in order to rationally approach the endocrine aspects of liver cancer, we must further our knowledge of how hormones and organ systems interact during the process of tumorigenesis.
Androgens and estrogens work through their cognate receptors and associated cofactors to control gene expression. The androgen receptor (AR) is expressed from a gene located on the X-chromosome (23). There are two isoforms of estrogen receptor, ERα and ERβ, that are derived from genes on separate autosomes (24). The mouse liver expresses AR (11) and ERα but not ERβ (25). The major estrogen secreted by the ovaries is estradiol. The major androgen secreted by the testes is testosterone.
The objectives of the experiments described in this report were: determine the roles of ERα and PRLR in the protective effects of female sex against DEN-induced tumors and determine if ERα mediates the inhibition of tumorigenesis in males by exogenous estrogen treatment. We examined liver tumorigenesis in DEN-treated wild-type (WT) mice or in mice that were null for ERα (ERαKO) or PRLR (PRLR-KO). We found that the typically high level of tumorigenesis in males was reduced by ERα deficiency, but the lack of ERα did not alter the low tumorigenic response in the female. Examination of sexually dimorphic liver gene expression indicated that the lack of ERα had no effect on the male pattern of GH secretion, whereas in the female ERαKO, the male pattern of GH secretion developed. In addition, the protective effects of the ovaries were independent of PRLR. These unexpected findings suggest a new paradigm for hormonal influence on HCC that complicates any prospective endocrine therapy approach without further elucidation of the mechanisms involved.
ERα+/− mice were provided by D.Lubahn, University of Missouri, Columbus (26). The mice had been backcrossed into a C57Bl/6J background. Heterozygous males and females were bred and offspring were injected with 20 mg/kg DEN; Sigma, St Louis, MO) at PND12–14; at this time, the animals were ear tagged and a specimen of their tail was used in genotype analysis as described previously (26). At weaning, heterozygous offspring (ERα+/−) were culled and the WT and ERα−/− offspring were maintained for experimental groups.
The 52-week tumorigenesis assay was set up with the following treatment groups: (i) ERα WT females, ovary intact (WT Intact); (ii) ERαKO females, ovary intact (KO Intact); (iii) WT females, ovariectomized (WT Ovxd); (iv) ERαKO females, ovariectomized (KO Ovxd); (v) WT males, vehicle control (WT control); (vi) WT males, E2 treated (WT + E2); (vii) ERαKO males, vehicle control (KO control); (viii) ERαKO males, E2 treated (KO + E2). Ovariectomies were performed on mice at 4 weeks of age. E2 treatments were achieved via subcutaneous Silastic implants. The capsules were prepared as described elsewhere (27,28). Briefly, each capsule was prepared from a piece of Silastic tubing (0.062” ID 0.125” OD; Cole-Palmer, Vernon Hills, IL) 11 mm in length overall. The ends of the tubing were plugged with pieces of wooden dowel and covered with Silastic cement; the center portion, 5 mm in length, was filled with ~10 mg crystalline E2. At 52 weeks, the time of termination, crystalline hormone was still visibly present in the capsule. In long-term experiments, these capsules were shown to stimulate estrogen-responsive tissue beyond 1 year after implantation (29). Livers were harvested at 52 weeks of age, fixed in formalin, held in 70% ethanol and surface tumor nodules on all lobes were counted.
PRLR heterozygous mice (PRLR+/−) were obtained from Paul Kelly (Faculte de Medecine Necker, Paris, France) (30); these mice were reported to be in the mixed (129Ola-X-C57Bl/6J) background (31). Female heterozygotes (PRLR+/−) were bred to homozygous null males (PRLR−/−). The heterozygous dams produced insufficient milk to maintain the pups and therefore, the pups were fostered off to Balb/c dams (Harlan, Indianapolis, IN) that had recently given birth. All offsprings were treated with DEN at PND12–14. Since the breeding was set up as PRLR+/− females plus PRLR−/− males, the offspring were either PRLR heterozygous (Het) or homozygous null (KO); groups of male and female Het and KO mice were left intact and untreated. In addition, groups of Het and KO females were ovariectomized at 4 weeks of age. Livers were harvested at 52 weeks and tumor nodules were counted on the entire liver.
A set of DEN-treated WT and ERαKO mice were killed at 12 weeks of age to determine liver gene expression in males and females. RNA was extracted with Trizol reagent from ~100 mg tissue. Final concentration of RNA was estimated by determination of OD260 and 1 μg of each preparation was subjected to reverse transcription using an oligo-deoxythymidine primer. Three aliquots of each complementary DNA were subjected polymerase chain reaction (PCR) for Gapdh and the resulting product was separated by gel electrophoresis, stained with ethidium bromide and quantified by image analysis (Scanalytics, Scion Corporation, Vienna, VA). These quantitative results were used to adjust the amount of each sample subjected to PCR amplification of liver specific sexually dimorphic genes. Genes tested and the primers used in the analyses are listed in Table I.
PCR was carried out with cycles of 1 min at 94°C, 45 s at 60°C and 1.5 min at 72°C, followed by an 8 min extension period at 72°C. Each PCR reaction was carried out for 20 and 25 cycles to allow comparison of sex and genotype effects on the target gene expression. The amplification products were separated by electrophoresis, stained with ethidium bromide and photographed.
Tumor multiplicity was calculated from the number of tumor nodules for each animal and this was averaged within a group. Data are expressed as means ± standard error of the mean. Means were compared by one-way analysis of variance using the Kruskal–Wallis test to determine significance followed by the Dunns test to compare individual group means against each other. Tumor incidence refers to the percentage of animals displaying identifiable surface tumor nodules and treatment incidence was tested against incidence in the corresponding control group using Fisher’s exact test. In all cases, treatment means or incidences were considered different if P <0.05.
Tumor multiplicity was much reduced in females compared with males, but tumorigenesis was unaffected by ovariectomy in this strain of mice (Table II). Nor did ERα status affect tumorigenesis in the females. Tumor incidence was 90–100% in females and did not differ among the four treatment groups. Furthermore, under the experimental conditions used in this study, tumor incidence of ovary-intact females did not differ from WT control males.
In male mice, E2 reduced tumorigenesis and this effect was dependent on ERα status (Table II). Both tumor multiplicity and incidence were reduced by E2 treatment. In addition, ERαKO males had a reduced tumor multiplicity compared with WT males and tumor numbers were unchanged by estradiol treatment in the ERαKO males.
We analyzed several genes that are expressed in a sexually dimorphic manner in the liver as a result of male or female GH secretion patterns (Figure 1). Livers were derived from animals that had been treated with DEN at PND12–14 and harvested at 12 weeks of age; there were no tumor nodules apparent on the livers at this time point. In WT animals, Elov3, Cyp4a12 and Hsd3b5 were expressed to a much larger extent in male than female, whereas Cyp2b13 and Cype3a41 were expressed exclusively in females. Cyp3a16 was expressed equally in male and female. The pattern of enzyme expression was unchanged in the ERα null male mice, but in the female, the patterns were reversed by the lack of ERα. These observations confirm those of Seyoushi et al. (32), who used other liver genes to determine the role of ERα in establishing the sexually dimorphic GH secretory pattern.
In this experiment, all animals, except for 1 of the 15 intact female knockouts, had liver tumors. Tumor multiplicity was lower in females compared with males, regardless of PRLR status (Table III). Ovariectomy resulted in a much higher tumor count in females, one that was comparable with that of males and this too was unaffected by PRLR status.
This is the first study to directly examine the role of ERα in DEN-induced liver tumorigenesis. In the context of earlier literature indicating that ovarian estrogen inhibited tumorigenesis and male androgen enhanced it (8–12), the lowered tumorigenic response in male ERαKO and the ERα- and ovary-independent protective effects in the female were unexpected findings.
Our results point to a role for ERα in mediating enhanced tumorigenesis in males. The lower tumorigenesis in the ERαKO males was not due to a decrease in androgen nor to a disruption in the male pattern of GH secretion. It has been shown that ERαKO males have serum testosterone concentrations that are several folds higher than those of WT males (33). Although it is difficult to determine the GH secretion rates in mice, expression of sex-specific genes in the liver reflects male or female GH secretory patterns (34–38). Our results on liver gene expression extend and confirm those reported by Sueyoshi et al. (32), indicating that imprinting of the male GH secretion does not require ERα, whereas female ERαKO mice exhibited male patterns of gene expression, indicating that pubertal estrogen acts through this receptor to override the male pattern. Since tumorigenesis was decreased in male ERαKO mice in the face of continued male GH secretion and since there was no increase in tumorigenesis in females exhibiting the male pattern of GH secretion, these results suggest that the sexually dimorphic pattern of GH secretion is not responsible for sex differences in liver tumorigenesis.
It may be that the interplay between GH secretory patterns and activated ERα within the liver determine whether estrogen promotes or retards progression of preneoplastic foci to tumors. It has been suggested that the small decrement in body mass of male ERαKO mice is due to attenuation of the GH/IGF-I axis (39,40). However, it is not known which part of that axis, GH secretion or IGF-I expression, is affected. It has been shown that estrogen can stimulate IGF-I expression in the liver and the decreased serum IGF-I concentrations seen in ERαKO mice may simply be due to the decrement in that stimulus (40). Furthermore, the reduced levels of IGF-I in ERαKO males are comparable with levels seen in WT females (39,40). It may be that both GH and estrogen are required for the male level of liver IGF-I expression and that the characteristic male tumorigenic response is related to this high level of IGF-I. Alternatively, even in the absence of estrogen ERα may play a role in IGF-I action (41).
The cytokine, interleukin (IL)-6, is another growth factor implicated in liver tumorigenesis and this too may be modified by estrogen. In acute toxicological analyses performed in adult mice, it has been shown that DEN causes hepatocyte cell death, the debris from which induces production of IL-6 from Kupffer cells; the cytokine in turn stimulates a compensatory cell proliferation in the surviving hepatocytes (42–44). In IL-6-null mice, the male bias for DEN-induced liver tumorigenesis is eliminated (44). It was proposed that ERα mediates estrogen protection against hepatocarcinogenesis by reducing the production of IL-6 by Kupffer cells (44). This hypothesis has been enthusiastically received for its clinical implications (45–47). However, it is well established that the gender differences in the DEN-induced tumorigenesis model occur at the promotion stage, that is, preneoplastic foci develop equally in males and females over the course of 10–12 weeks after DEN administration, whereas fully developed tumor nodules exhibit sex bias only after several months (10). Furthermore, the acute responses to DEN in the juvenile mouse tumorigenesis model would occur at a time when there is no ovarian production of estrogen, that is, in the prepubescent mouse. Thus, the very transient IL-6 response demonstrated following a single injection of DEN to adult mice may not be related to the gender effect that is evident several months after carcinogenic insult in neonates. Our data further question the relevance of the acute protective effect of estrogen as demonstrated by acute responses to high doses of DEN treatment in adult mice. Although full protection by estrogen against DEN-induced cell death required ERα in male mice (44), we found that lack of ERα did not result in greater tumorigenesis in females and it actually reduced tumorigenesis in males. Thus, our results do not support the hypothesis that the protective effect of female factors is related to estrogen-induced reduction in IL-6 expression.
Estrogens are formed from androgens through the action of Cyp19a1, steroid aromatase. Cyp19a1 is expressed in the testis, brain and peripheral fat (48). It is controversial whether Cyp19a1 is expressed in the normal adult human liver but it has been clearly demonstrated in diseased liver and in liver cancer cell lines (49–51). Although circulating levels of estrogens are expected to be lower in males than females, it is not known whether localized liver tissue concentrations of estrogen differ according to sex. Thus, it is possible that androgens affect liver tumorigenesis through aromatization and subsequent activation of ERα to directly affect hepatocytes. Using the testicular feminization mutant (Tfm) mouse Kemp et al. (11) found that the high rate of tumorigenesis in male mice was AR-dependent; in addition, observations on mice that were mosaic for AR in the liver suggested that the effect of androgen was mediated by secondary paracrine or endocrine factors. Furthermore, although the absolute numbers were lower in the animals lacking AR, testosterone treatment did increase tumor incidence and multiplicity in the AR-null males (11). Notably, the aromatizable androgen, testosterone, was used in that experiment, thereby allowing for the possibility that the tumor promoting effect of testosterone in the absence of its cognate receptor might be mediated by estrogen metabolites. Our own data indicate that ERα is required for ~47% of the male effect, suggesting that estrogen derived from aromatization of testicular androgen plays a major role. Thus, estrogen may be the secondary secreted growth factor hypothesized to mediate the androgen effect in Tfm, AR- mosaic mice.
The role of ERα in male liver tumorigenesis demonstrated in the mouse may have bearing on the course of HCC in humans. Liver disease associated with hepatitis virus is aggravated by alcohol, indeed, alcohol and viral infection synergize to increase the risk of HCC (1,5–7,52). Alcohol has also been shown to increase expression of aromatase in the diseased liver (53,54). In light of our results, it might be hypothesized that this increase in aromatase is at least partially responsible for the increased risk in males due to alcohol intake.
ERα does mediate a protective effect of exogenously delivered estrogen as evidenced by the reduction in tumor counts in WT males treated with E2 but not in E2-treated ERαKO males. Our E2 treatment capsules produce a constant high physiological level of hormone (55). Thus, ERα can mediate two opposing effects: a tumor-enhancing effect of low levels of endogenous estrogen and a tumor-inhibiting effect of high estrogen levels produced by exogenously delivered hormone. This type of biphasic dose–response is not uncommon for estrogen (56–60).
On the other hand, in WT C57Bl/6J females, we did not observe an effect of ovariectomy nor was there an effect of ERα deficiency. The lack of effect of ovariectomy compared with the increase in tumorigenesis reported by others is probably due to two aspects of the study design: the strain of mice used and the dose of DEN administered. It has been shown earlier that the tumorigenic response to DEN and the effects of gonadal hormones is strain dependent, with C57Bl/6J female mice exhibiting a much lower tumor incidence and multiplicity compared with other strains (10,15). In those earlier studies, the dose of DEN was ~5 mg/kg (0.05 μmol/g body wt) compared with the 20 mg/kg dose administered in our study; the higher dose produced a nearly 100% incidence in both intact and ovariectomized animals, compared with incidences of 7 and 45% for intact and ovariectomized female C57Bl/6J mice receiving the low dose of DEN (15). Thus, in those earlier studies, the low multiplicity reported for intact females was mainly due to a low incidence of mice with tumors, whereas in our study, multiplicity more closely reflected the number of tumors per animal. Thus, when the dose of DEN is sufficient to overcome the resistance to tumor induction in C57Bl/6J mice, the protective effect of the female factor(s) is still evident but it is not dependent on ovarian hormone(s) or ERα.
Comparing the effects of ovariectomy in the ERαKO and the PRLR studies also points to a dramatic influence of genetic background on the role of ovarian factors in liver tumorigenesis. The PRLR-knockout mice are in a mixed genetic background (129Ola-X-C57Bl/6J); ovariectomy did result in heightened tumorigenesis in these animals. On the other hand, ovariectomy had no effect on tumorigenesis in the C57Bl/6J inbred animals. Thus, genetic components of the female C57Bl/6J mouse provide protection even in the estrogen-deficient state. Genetic studies have been performed to identify genes that are responsible for heightened sensitivity of female C57BR/cdJ to DEN-induced liver tumorigenesis compared with C57Bl/6J females (61–63), however, it may be equally important to identify the genes that confer ovary-independent resistance to tumorigenesis in female C57Bl/6J mice.
This is also the first study to directly examine the role of PRLR in DEN-induced liver tumorigenesis. Since manipulation of endogenous PRL secretion indicates that it protects against tumorigenesis in the liver (13) and since estrogen does increase PRL secretion by the pituitary (64), it is an appealing candidate as a mediator of the protective action of the ovaries. However, the results of this study indicate that the PRLR does not mediate the protective effects of endogenous female factor(s).
In summary, exogenous E2 protects against carcinogen-induced liver tumorigenesis in an ERα-dependent manner, but ERα is not required for the low tumorigenic response in the female. In the male, the full tumorigenic response is partially dependent on ERα, implicating either a role for aromatization of endogenous androgens in the process or for a role of unliganded ERα. Although PRL may be a candidate as a mediator of protective effects of exogenously delivered estrogen, the unaltered response in PRLR-KO mouse to DEN indicates that it is not responsible for the protective effects of the ovarian factor(s).
National Institutes of Health (ES014367).
Technical assistance of Tasha Brame and John Doty was invaluable in completion of the reported experiments.
Conflict of Interest Statement: None declared.