Glutathione-Deficient Mice Have Increased Sensitivity to Transplacental Benzo[a]pyrene-Induced Premature Ovarian Failure and Ovarian Tumorigenesis

Sixty percent of women diagnosed with ovarian cancer will die of the disease; it is the leading cause of death from gynecologic cancers. Because ovarian cancer tends to be asymptomatic until it has reached an advanced stage, treatment is often ineffective. Understanding the causes of ovarian cancer is critical for the development of preventive strategies and methods for early diagnosis, yet the underlying cause of most ovarian cancers remains elusive. Ninety percent of malignant ovarian cancers in women are histologically classified as epithelial, whereas about 7% are classified as stromal (1). Epithelial ovarian cancers are thought to originate from the ovarian surface epithelium, and stromal ovarian cancers are thought to originate from granulosa cells, theca cells, other stromal cells, or their counterparts in the testicular sex cords (Sertoli cells and Leydig cells; ref. 1).

Polycyclic aromatic hydrocarbons (PAH) are ubiquitous environmental pollutants found in air pollution, cooked foods, and tobacco smoke. Human biomonitoring data show that essentially all Americans are exposed to PAHs (2). It has been known for decades that PAHs like benzo[a]pyrene (BaP) are toxic to the ovary, destroying ovarian follicles and causing ovarian tumors in rodents. The PAHs 7,12-dimethylbenzanthracene (DMBA), 3-methylcholanthrene, and BaP all destroyed follicles in pre- and peripubertal mice after both single high doses (3, 4) and repeated low doses (5). The significant PAH content of tobacco smoke is thought to play a role in the known ovarian toxicity of smoking. Women who smoke have decreased per menstrual cycle probability of pregnancy compared with women who do not smoke (6, 7), and the onset of menopause occurs several years earlier in women who smoke (8). Treatment of adult mice with various PAHs by several routes induced ovarian tumors (9–11), and women who smoke have increased risk of epithelial ovarian cancer (12, 13).

There is increasing evidence that environmental exposures that occur during embryonic and fetal development modify the risk for developing diseases, such as cancer, in adulthood. In utero exposure to PAHs causes ovarian toxicity to the female offspring. Treatment of pregnant mice with BaP dose dependently decreased fertility and resulted in atrophic ovaries with few follicles in female offspring (14). Transplacental ovarian carcinogenesis has been shown for the PAHs DMBA (15) and dibenz[def,p]chrysene (16), but to the best of our knowledge, has not been reported for BaP. BaP, like dibenz[def,p]chrysene, is relevant to humans because environmental exposure to these 2 PAHs is ubiquitous, whereas DMBA is not present in the environment.

To exert their toxicity, PAHs generally require metabolic activation to toxic metabolites by the phase I metabolizing enzymes, cytochromes P450 (CYP) and microsomal epoxide hydrolase (17). Of the major CYPs involved in PAH metabolism, Cyp1b1 is expressed in mouse embryos from embryonic day (E) 11 to E17, whereas Cyp1a1 is expressed on E7, but not later, and Cyp1a2 is not expressed (18, 19). To our knowledge, no information is available on fetal ovary expression of any of these CYPs in mice. Prostaglandin–endoperoxide synthases are also key enzymes involved in bioactivation of PAHs in the developing embryo (20, 21). Prostaglandin–endoperoxide synthases oxidize PAHs to free radical intermediates, which can initiate reactive oxygen species (ROS) generation (21). Phase II metabolism or detoxification of toxic BaP metabolites occurs via glutathione-S-transferase–catalyzed conjugation with glutathione (GSH; refs. 22, 23). During mouse development, embryos express GSH synthetic enzymes and synthesize GSH beginning at the blastocyst stage (24). Expression of Gsta4 is present in the whole embryo on E7.5 and E8.5 and in the ovary on E14.5, and Gstp1, Gstp2, Gstm1, and Gstm2 are expressed throughout the embryo on E14.5 (25).

GSH is the most abundant intracellular nonprotein thiol and one of the most important intracellular antioxidants. It is present in cells at millimolar concentrations. GSH has numerous intracellular functions, including reduction of hydrogen peroxide and lipid peroxides as a cofactor for GSH peroxidases, detoxification of electrophilic toxicants as a cofactor for glutathione-S-transferases, regulation of protein function, and regulation of nucleotide metabolism. GSH is synthesized in 2 ATP-dependent reactions (26). The first, rate-limiting reaction is catalyzed by glutamate cysteine ligase (GCL), a heterodimer composed of a catalytic (GCLC) and a modifier (GCLM) subunit. Mice that lack Gclc die during embryonic development (27, 28). Mice that lack Gclm survive and reproduce but have greatly reduced tissue levels of GSH (29, 30). We recently reported that female Gclm^−/− mice have compromised fertility due to early preimplantation embryo mortality (31).

In view of the importance of GSH in maintaining cellular redox status and in detoxifying reactive metabolites of BaP, we hypothesized that Gclm-null mice have greater sensitivity to the ovarian toxicity of prenatal exposure to BaP. We further hypothesized that BaP is a transplacental ovarian carcinogen and that female Gclm-null mice are more susceptible than wild-type littermates to ovarian tumorigenesis resulting from in utero BaP exposure.

All chemicals and reagents were purchased from Fisher Scientific or Sigma Aldrich unless otherwise noted.

Gclm-null mice were generated by disrupting the Gclm gene by replacing exon 1 with a β-galactosidase/neomycin phospho-transferase fusion minigene (30, 32). The mice were backcrossed 8 times onto a C57BL/6J genetic background (B6.129-Gclm^tm1Tjka; hereafter referred to as Gclm^−/−). Mice for these experiments were generated at the University of California, Irvine (UC Irvine). Offspring were genotyped by PCR using primers for both the native Gclm sequence and the β-Geo sequence on DNA extracted from tail or toe snips as previously described (32). The mice were housed in an American Association for the Accreditation of Laboratory Animal Care–accredited facility, with free access to deionized water and soy-free laboratory chow, on a 14:10 hour light–dark cycle. Temperature was maintained at 21° to 23°C. The experimental protocols were carried out in accordance with the Guide for the Care and Use of Laboratory Animals and were approved by the Institutional Animal Care and Use Committee at UC Irvine.

Estrous cycle stage in individually housed adult female mice was evaluated every morning by microscopic examination of fresh vaginal lavage fluid obtained in 0.9% sodium chloride.

Gclm^+/− female mice were mated with Gclm^+/− or Gclm^−/− male mice on the afternoon of proestrus based on vaginal cytology. Females were checked for vaginal plugs the following morning. The day of vaginal plug detection in the female was designated gestational day (GD) 1. Dams were treated by oral gavage with 10 mg/kg BaP in sesame oil daily from GD7 to GD16 (experiment 1) or 2 mg/kg/d from GD7 to GD16 (experiment 2). Control animals were gavaged with the same volume of sesame oil alone in both experiments. The dosing regimen in experiment 1 was based on a previous study in mice, which showed that offspring of dams treated with this dose had reduced fertility compared with controls but were not completely infertile (14). We used the lower dose in the second study because of an apparently increased intrauterine mortality of Gclm^−/− female fetuses in experiment 1. Of 8 litters born to dams treated with 10 mg/kg BaP, only 2 Gclm^−/− females (both from the same litter) were born, compared with 7 Gclm^+/+ females from 5 litters. Dams were allowed to deliver and care for their litters. Litters were weaned on postnatal day (PND) 21, and dams were euthanized with CO₂ on the day of weaning. Maternal ovaries were processed for histomorphometry.

Fertility of Gclm^−/− and Gclm^+/+ female offspring was tested in a continuous breeding assay (see below). At the end of the breeding assay, female mice were removed from their mates. After the last litter was delivered, estrous cycles were monitored for at least 14 days and mice were killed by CO₂ asphyxiation on the next morning of estrus or on day 15 if they were not cycling. Skin, mammary glands, livers, kidneys, lungs, uteri, and ovaries were carefully examined for gross evidence of tumors at necropsy. In experiment 1, both ovaries were fixed in Bouin fixative, processed for histology, serially sectioned at 5 μm, stained with hematoxylin and eosin, and used for histomorphometry, including diagnosis of ovarian tumors. In experiment 2, both ovaries were examined for ovarian tumors. One ovary was processed as for experiment 1 and used for histomorphometry, and the other was fixed in 4% paraformaldehyde, cryopreserved in 15% sucrose in PBS, embedded in optimal cutting temperature (OCT) embedding compound, and every fourth section was used for hematoxylin and eosin staining. Gclm^+/− female offspring were sacrificed by CO₂ euthanasia at 35 days of age, and ovaries were processed as for experiment 1 for ovarian histomorphometry.

Beginning at 2 months of age, control or BaP-treated Gclm^+/+ or Gclm^−/− female mice were mated to wild-type, 10-week-old C57BL/6J male mice (Jackson Laboratories) for a continuous breeding assay. Females were checked daily for pups. On the day of birth, live and dead pups were counted and sexed, live pups were weighed, and pups were humanely euthanized.

Complete serial histologic sections from each animal were evaluated by 2 of the investigators (U. Luderer and J. Lim) without knowledge of genotype or treatment using an Olympus BX60 microscope equipped with Plan Fluor ×10, ×20, and ×40 objectives, Plan Achromat ×4 objective (Olympus America), and a Retiga 2000R cooled CCD mono digital camera (QImaging). Ovarian follicles were classified as primordial (single layer of flattened granulosa cells), primary (single layer of cuboidal or mixed cuboidal/flattened granulosa cells), secondary (more than one layer of granulosa cells), or antral (possessing an antral cavity or several fluid filled vesicles in the case of early antral follicles) and were further classified as healthy or atretic. Atretic secondary and antral follicles were identified by the presence of 3 or more pyknotic granulosa cells per largest cross section and separation of the oocyte from the granulosa cells (33). Atretic primordial and primary follicles were identified by eosinophilic oocytes (34). Primordial, primary, and secondary follicles were counted in every fifth serial section. The counts were multiplied 5 times to obtain estimates of the total number of follicles per ovary (35). Antral follicles were followed through every serial section, taking care to count each of these structures only once. Each section of the ovaries from 7.5-month-old mice was also evaluated for the presence of ovarian tumors or cysts by a board-certified veterinary pathologist (G.W. Lawson) blind to genotype and treatment. Perpendicular tumor diameters were measured in the largest tumor cross section. The average tumor diameter was calculated for each animal. Several tumors contained very large cysts. For these cystic tumors, the largest diameters with and without the cyst were measured, and average diameters were calculated with and without the cyst. Sensitivity analyses compared the effect of genotype on average tumor diameter when cysts were included versus not included in the measurements.

One slide was selected from each ovarian tumor for destaining with 1% hydrochloric acid and 1.5% ammonium hydroxide in 70% ethanol and subsequent immunostaining with an antibody directed against mouse pancytokeratin (Sigma C1801, Clone PCK-25), a marker for epithelial ovarian tumors (36). Immunostaining was done using the Mouse on Mouse (MOM) Kit according to the manufacturer's instructions (Vector Laboratories). Briefly, slides were subjected to antigen retrieval in 10 mmol/L sodium citrate with 0.05% Tween-20 at 95°C, blocked with mouse MOM immunoglobulin G blocking reagent, washed, blocked with avidin and biotin blocking reagents, washed, incubated with 1:50 dilution of primary antibody in MOM diluent, washed, incubated with 3% hydrogen peroxide in PBS to block endogenous peroxidases, incubated with biotinylated secondary antibody, washed, incubated with ABC reagent, washed, incubated with diaminobenzidine substrate in peroxide buffer, and counterstained with hematoxylin. Negative control slides (without primary antibody, primary antibody replaced with mouse immunoglobulin, without secondary antibody) and an untreated ovary control were run simultaneously with the experimental slides. All negative control slides showed no nonspecific staining. The untreated ovary showed staining only in the ovarian surface epithelium.

Because experiments 1 and 2 were conducted 2 years apart, the effect of the experiment on various endpoints was examined for Gclm^+/+ females exposed prenatally to 0 mg/kg BaP. As there were no significant effects of the experiment on any of the endpoints, the data were combined for subsequent analyses. The effects of genotype and BaP dose on continuous outcome variables were analyzed using generalized estimating equations, a form of generalized linear models, with BaP dose, Gclm genotype, and BaP times genotype interaction modeled as fixed effects. To adjust for litter effects, litter numbers were entered into the model as a subject effect. An unstructured working correlation matrix was used for the litter effects, except in the rare cases when the model did not converge, in which case an exchangeable working correlation matrix was used (assumes homogeneous correlation among littermates). When the outcome variable was a fraction (e.g., fraction of cornified or leukocytic vaginal cytology), the data were first arcsine square root-transformed before they were entered into the models; for these analyses only, the unadjusted means and SEMs were used for presentation of the data. Statistical analyses were conducted using SPSS 20.0 for MacIntosh.

Data on pregnancy outcomes are shown in Table 1. The percentage of F0 females with vaginal plugs that delivered litters did not differ by BaP dose. Litter size did not differ by BaP dose. Full litter mortality occurred for 5 of thirteen 10 mg/kg BaP-treated dams, 4 of nineteen 2 mg/kg BaP-treated dams, and 6 of 26 control dams; the effect of BaP dose was not statistically significant. Although fewer Gclm^−/− female offspring were born to 10 mg/kg BaP-treated females than to 2 mg/kg BaP-treated or control females, the effect of BaP dose on offspring genotype distributions was not statistically significant (Table 1).

Ovarian follicle numbers in the dams were not affected by the 10 mg/kg BaP dose (1,354 ± 242 healthy follicles per ovary in 0 mg/kg/d BaP-treated dams and 1,363 ± 320 in 10 mg/kg/d BaP-treated dams).

The results of the continuous breeding studies of Gclm^−/− and Gclm^+/+ female offspring are shown in Fig. 1. Two mice had prolapsed uteri and had to be euthanized before the end of the 20-week breeding study in experiment 1. Both were wild type, one in the control group and one in the 10 mg/kg/d BaP-treated group. Among the remaining mice, in utero treatment with 10 mg/kg/d BaP profoundly decreased fertility of female mice of both genotypes, whereas treatment with 2 mg/kg/d BaP decreased fertility to a much greater extent in Gclm^−/− females than in Gclm^+/+ females. Two of two 10 mg/kg/d BaP-treated Gclm^−/− mice and 4 of 6 Gclm^+/+ 10 mg/kg/d BaP-treated mice delivered no litters and showed no signs of pregnancy during the study period. The other 2 surviving Gclm^+/+ 10 mg/kg/d BaP-treated mice had 1 and 3 litters, and the BaP-treated mouse with uterine prolapse had delivered the second litter, when she was euthanized at week 11 of the study. The effects of genotype (P < 0.001) and BaP dose (P < 0.001) on the number of litters produced in 20 weeks were statistically significant (Fig. 1A). In addition, to having fewer litters, BaP-treated females produced fewer offspring in 20 weeks (P < 0.001, effect of BaP dose); this effect of prenatal BaP treatment was more pronounced in the Gclm^−/− females than in the Gclm^+/+ females (P<0.001, BaP dose × genotype interaction; Fig. 1B). Gclm^−/− females produced significantly fewer pups compared with Gclm^+/+ littermates (P < 0.001, effect of genotype), as we have previously reported (31).

Estrous cycle data collected after completion of the breeding study are shown in Table 2. Two of 13 Gclm^+/+ and 2 of 12 Gclm^−/− oil-treated female offspring were not cycling at 7.5 months of age. None of the 10 mg/kg/d BaP-treated offspring displayed estrous cycles (P < 0.05, effect of BaP on presence or absence of cycles). Most of the 10 mg/kg/d BaP-treated mice had persistently cornified vaginal cytology, whereas most of the irregularly cycling control and 2 mg/kg/d BaP mice had predominantly leukocytic vaginal cytology (P < 0.005, effect of BaP on percentage of cornified and percentage of leukocytic vaginal smears). Predominantly cornified vaginal cytology, termed “persistent vaginal estrus,” is uncommon in normal aged, reproductively senescent mice, whereas predominantly leukocytic vaginal cytology is commonly observed in aged mice. Persistent cornified vaginal histology has been reported following neonatal treatment with various estrogenic chemicals. The number of transitions between cornified and leukocytic cytology, consistent with an estrus to metestrus transition, in 14 days declined significantly with BaP dose (P < 0.001, effect of BaP). There were no effects of BaP dose or genotype on estrous cycle length in the mice that were cycling. Gclm^−/− mice had fewer leukocytic vaginal smears than Gclm^+/+ females (P < 0.001, effect of genotype).

To assess the effect of gestational treatment with BaP on prepubertal ovarian follicle numbers, ovarian histomorphometry was conducted on a subset of the Gclm heterozygous F1 female offspring from experiments 1 and 2 at 35 days of age. Exposure to BaP from GD 7 to 16 dose dependently decreased the total number of healthy ovarian follicles from 2,250 ± 348 in 0 mg/kg BaP-exposed, to 1,036 ± 338 in 2 mg/kg BaP-exposed, and 34 ± 16 in 10 mg/kg BaP-exposed (P < 0.001, effect of BaP dose). The effect of BaP dose was statistically significant for numbers of healthy primordial follicles, primary follicles, secondary follicles, and antral follicles (P = 0.001, P = 0.007, P < 0.001, P = 0.041), as well as for numbers of atretic secondary and antral follicles (P = 0.005; P = 0.031; Table 3). Very few atretic primordial or primary follicles were observed in any BaP dose group, and there were no apparent effects of BaP dose on these outcomes (data not shown).

Ovarian follicle counts were also conducted on both ovaries per mouse in the breeding study from experiment 1 and 1 ovary per mouse from the breeding study in experiment 2, both at about 7.5 months of age (Fig. 2). Prenatal BaP exposure dose dependently decreased the total number of healthy follicles, as well as the numbers of healthy primordial, primary, secondary, and antral follicles (P < 0.001, effects of BaP dose). Consistent with the lack of litters in the 10 mg/kg/d BaP-treated mice after week 12 of the breeding study, there were no antral or secondary follicles in any of the 10 mg/kg/d BaP-exposed ovaries. Deletion of Gclm was associated with greater sensitivity to follicle depletion due to prenatal BaP exposure. The interaction between BaP dose and Gclm genotype was statistically significant for total follicle count (P < 0.001; Fig. 2A), primordial follicle count (P < 0.001; Fig. 2B), primary follicle count (P = 0.025; Fig. 2C), and secondary follicle count (P = 0.032, Fig. 2D). In addition, deletion of Gclm was independently associated with fewer primordial, primary, and secondary follicles and total healthy follicles (P < 0.001, P = 0.007, P = 0.032, P < 0.001, effect of genotype, respectively).

Serial sections of both ovaries were evaluated for ovarian tumors. The results are shown in Table 4, Fig. 3, and Supplementary Fig. S1. Two of two 10 mg/kg/d BaP-treated Gclm^−/− female offspring had bilateral ovarian tumors that replaced the entire ovarian parenchyma and were invading the periovarian fat. Three of six 10 mg/kg/d BaP-treated Gclm^+/+ offspring had bilateral ovarian tumors, 2 of 6 had unilateral ovarian tumors, and 1 of 6 had no ovarian tumors; 4 of the 8 tumors were invading the periovarian fat (Supplementary Fig. S1B). Four of six 10 mg/kg/d BaP-treated Gclm^+/+ offspring had very large, complex, fluid-filled cysts within the ovaries; 2 of these were in tumors (Supplementary Fig. S1C and Fig. 3C). One of nine 2 mg/kg/d BaP-treated Gclm^−/− females had a unilateral ovarian tumor (Supplementary Fig. S1F and Fig. 3D). None of the ovaries from 2 mg/kg/d BaP-treated Gclm^+/+ females or from 0 mg/kg/d BaP-treated females of either genotype had any ovarian tumors. The adjusted average number of ovarian tumors per mouse was 0 ± 0 for both 0 mg/kg/d BaP-treated groups and the Gclm^+/+ 2 mg/kg/d BaP-treated group, 0.11 ± 0.09 in the Gclm^−/− 2 mg/kg/d BaP-treated group, 1.3 ± 0.2 in the Gclm^+/+ 10 mg/kg/d BaP-treated group, and 2.0 ± 0 in the Gclm^−/− 10 mg/kg/d BaP-treated group (P < 0.001, effects of BaP dose, genotype, and dose × genotype interaction). In addition to a larger percentage of the tumors showing invasion of the periovarian fat in the Gclm^−/− females, their tumors also tended to be larger. The average largest tumor diameter, excluding cystic structures in 2 tumors, was 1,894 ± 203 μm in the 10 mg/kg/d BaP-treated Gclm^−/− females and 1,239 ± 252 μm in the 10 mg/kg/d BaP-treated Gclm^+/+ females (P = 0.02 by t test). When cysts were included in the diameters, the tumors were still larger in the Gclm^−/− females (1,894 ± 203 μm vs. 1,396 ± 444 μm in the Gclm^+/+ females), but the difference was no longer statistically significant (P = 0.20).

All of the ovarian tumors were initially classified as stromal tumors based on their histologic appearance (Supplementary Fig. S1). Although invasion of the periovarian fat was noted for several tumors, the relative paucity of mitotic figures and lack of metastases was consistent with benign tumors. Immunostaining for the epithelial marker cytokeratin showed that all but 2 of the tumors strongly expressed cytokeratin in the cells lining the abundant tubule-like structures in these tumors (Fig. 3A–C). One of these cytokeratin-negative tumors, in a Gclm^+/+ 10 mg/kg/d BaP-exposed female, contained a large cyst, and only the cells lining the cyst and the ovarian surface epithelium displayed cytokeratin immunostaining (not shown). The other cytokeratin-negative tumor was the only tumor found in a 2 mg/kg/d BaP-exposed female (Supplementary Fig. S1F and Fig. 3D); only the ovarian surface epithelium and a small epithelial inclusion showed cytokeratin immunostaining. Cytokeratin is expressed only in the ovarian surface epithelium in normal ovaries (36). In view of this pattern of cytokeratin immunostaining, we concluded that the majority of the tumors induced by prenatal BaP exposure were benign epithelial tumors, most likely tubular adenomas. The 2 cytokeratin-negative tumors were diagnosed as granulosa theca cell tumors based on their histologic appearance.

We did not observe gross tumors in any organ besides ovaries at necropsy. Histologic examination of the livers and kidneys of the 10 mg/kg/d BaP-exposed mice also did not reveal any tumors.

Our results show for the first time that the ubiquitous PAH BaP is a transplacental ovarian tumorigen at doses that do not produce tumors in other organs. The majority of the ovarian tumors induced by prenatal exposure to BaP were positive for the epithelial marker cytokeratin. Our study confirms earlier findings that prenatal oral exposure to the PAH BaP causes premature reproductive senescence in F1 female mice (14), and our data show that this is due to premature depletion of ovarian follicles. Importantly, our data show that GSH deficiency due to deletion of Gclm sensitizes female embryos to the ovotoxicity and the transplacental ovarian tumorigenicity of BaP.

Transplacental exposure to the PAHs DMBA and dibenzo[def,p]chrysene has been reported to cause sex-cord stromal ovarian tumors (15, 16), and our current results show that BaP is also a transplacental ovarian carcinogen. Importantly, immunostaining showed that most of the BaP-induced ovarian tumors were strongly positive for cytokeratin, an epithelial cell marker, indicating that they were of epithelial and not stromal origin (36). Adult exposure to PAHs, including BaP, by several routes has long been known to induce ovarian tumors in mice (9–11). Although the doses used in those studies were high compared with those used in the present study, the incidences of females with ovarian tumors tended to be lower than we observed in the 10 mg/kg/d BaP-treated group. Taken together, these results suggest that the period of ovarian development is a sensitive window for ovarian carcinogenesis. Our data also show for the first time that embryonic GSH deficiency increases the sensitivity to transplacental ovarian tumorigenesis by BaP. Our findings that transplacental PAH exposure causes epithelial ovarian cancers in mice and that GSH-deficient embryos have increased sensitivity to this effect have relevance to humans. Mainstream and sidestream tobacco smoke contains dozens of carcinogenic PAHs, including BaP, and women who smoke have increased risk of epithelial ovarian cancer (12, 13). Polymorphisms in GCLM and GCLC, which affect GSH synthesis, exist in humans (37–40) and may modulate sensitivity to transplacental ovarian toxicity and tumorigenesis of PAHs.

Decreased ability to detoxify reactive metabolites of BaP and/or ROS produced as a result of BaP metabolism likely causes the increased sensitivity of the Gclm^−/− embryonic ovary to BaP. GSH conjugation is an important phase II detoxification mechanism for the diol epoxide metabolites (22, 23), which are thought to be the ovotoxic metabolites in peripubertal mice (41). GSH conjugation is also important in phase II metabolism of arene oxide and quinone metabolites of PAHs. CYPs 1A1, 1A2, and 1B1 are the major P450 enzymes involved in the metabolic activation of BaP in vitro, although in vivo metabolism of BaP by CYP1A1 seems to be more important in detoxification (17, 42). Expression of most CYPs remains low until after birth. Exceptions include transient constitutive expression of Cyp1a1 on GD 7 in the mouse, induction of Cyp1a1 expression in the liver and lung as early as GD 12.5 by exposure to PAHs in the mouse, and constitutive expression of Cyp1b1 in the mouse embryo from GD 11 onward and in the human fetal thymus and kidney (18–20, 43). Previous work has shown that embryos constitutively express prostaglandin–endoperoxide synthase 1 and 2, that prostaglandin–endoperoxide synthases can bioactivate BaP to free radical intermediates that initiate ROS formation, that gestational treatment with BaP increases embryonic oxidative protein and DNA damage, and that antioxidants are protective against teratogenesis caused by BaP (21, 44, 45). Although these previous studies did not examine developmental toxicity to the reproductive system, our findings of increased sensitivity of GSH-deficient female embryos to transplacental ovarian toxicity of BaP in the present study and our observation of increased sensitivity of GSH-deficient male embryos to transplacental testicular toxicity of BaP (46) provide further support for a role for oxidative stress in the developmental toxicity of BaP.

The mechanism by which transplacental exposure to BaP and other PAHs causes ovarian tumors is uncertain. It is well known that metabolism of PAHs leads to the formation of mutagenic metabolites, including diol epoxides and ROS (17). Although we are not aware of any published studies describing PAH-induced ovarian mutations, BaP treatment caused testicular DNA adducts and germline mutations in male mice (47), and BaP treatment resulted in the formation of ovarian DNA adducts in adult mice and rats (48, 49). Different histopathologic types of ovarian tumors have been reported in Cyp1b1-null mice (cystadenomas) versus wild-type mice (granulosa cell tumors) treated with DMBA (48). The authors postulated that this could be due to the absence of Cyp1a1 expression in stromal cells, versus its presence in ovarian surface epithelial cells where it could compensate for the lack of Cyp1b1. It is also possible that different PAH metabolites have different mechanisms of ovarian tumorigenesis. Oocyte depletion has been proposed to play a role in ovarian tumorigenesis by PAHs (10, 11). Although most of the tumors that we observed occurred in ovaries that were devoid of oocytes and follicles, the tumor that occurred in the 2 mg/kg/d BaP-treated Gclm^−/− female occurred in an ovary that still contained hundreds of healthy follicles. This supports the contention that oocyte depletion and the resultant high levels of gonadotropin hormone secretion may promote ovarian tumor growth but may not be required for ovarian tumor initiation.

The ED₅₀ for primordial follicle destruction by 15 daily intraperitoneal doses of BaP in peripubertal mice was 3 mg/kg/d (5). In contrast, we observed no effect on follicle numbers in the F0 dams and an ED₅₀ of about 2 mg/kg/d in the Gclm wild-type and heterozygous F1 offspring with 10 days of oral dosing to the pregnant dam in the present study. Although comparisons with earlier studies of BaP on ovarian follicle destruction are complicated by their use of intraperitoneal dosing (3–5), taken together, the results suggest that the period of in utero development is the most sensitive to the ovotoxicity of BaP, followed by the peripubertal period, with adults being least sensitive. In a previous study, oral gavage of CD-1 dams with 10 mg/kg/d BaP during the same dosing window that we used in the present study caused about a 50% decrease in pup production by female offspring (14). In contrast, we observed a 96% reduction in pup production in the wild-type 10 mg/kg/d BaP-treated F1 females. This is consistent with earlier reports that C57BL/6 mice are more sensitive to the peripubertal ovarian toxicity of BaP than other strains (3, 4, 41). An alternate explanation is that the Gclm heterozygosity of the mothers may have contributed to the increased ovarian toxicity of the transplacental exposure compared with the earlier study. In the future, this could be tested by transferring embryos from heterozygous dams mated with heterozygous males to pseudopregnant wild-type recipients, then treating the recipient dams with BaP. Treatment with the PAH DMBA destroyed small follicles in a dose- and time-dependent manner in human ovarian explants that were implanted subcutaneously in mice (50), showing that human ovarian follicles are susceptible to PAH-induced destruction.

Our findings show that the PAH BaP is a transplacental ovarian tumorigen, which causes mainly epithelial ovarian tumors in mice. Furthermore, deficiency of GSH due to deletion of Gclm increases sensitivity to premature ovarian failure and ovarian tumorigenesis induced by in utero exposure to BaP. These findings have broader implications because human exposure to BaP and other PAHs is ubiquitous and known human polymorphisms in GCLC and GCLM may define a sensitive subpopulation of women at greater risk for ovarian toxicity and tumorigenesis from early life exposures to PAHs.

No potential conflicts of interest were disclosed.

Conception and design: U. Luderer

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J. Lim, B.N. Nakamura, L. Ortiz, J.A. Hur, U. Luderer

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): G.W. Lawson, U. Luderer

Writing, review, and/or revision of the manuscript: J. Lim, G.W. Lawson, T.J. Kavanagh, U. Luderer

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Lim

Study supervision: U. Luderer

The authors thank Dr. Christine McLaren of the UC Irvine Chao Family Comprehensive Cancer Center Statistical Core for advice regarding statistical analyses.

This work was supported by the NIH (R01ES020454, and R21AG032087 to U. Luderer; P30CA062203, the University of California Irvine (UC Irvine) Chao Family Comprehensive Cancer Center; R01ES010849 and P50ES015915 to T.J. Kavanagh; P30ES007033, the University of Washington Center for Ecogenetics and Environmental Health); the UC Irvine Office of Research; and the UC Irvine Center for Occupational and Environmental Health.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.