Disruption of Reproductive Aging in Female and Male Rats by Gestational Exposure to Estrogenic Endocrine Disruptors

Abstract

Polychlorinated biphenyls (PCBs) are industrial contaminants and known endocrine-disrupting chemicals. Previous work has shown that gestational exposure to PCBs cause changes in reproductive neuroendocrine processes. Here we extended work farther down the life spectrum and tested the hypothesis that early life exposure to Aroclor 1221 (A1221), a mixture of primarily estrogenic PCBs, results in sexually dimorphic aging-associated alterations to reproductive parameters in rats, and gene expression changes in hypothalamic nuclei that regulate reproductive function. Pregnant Sprague Dawley rats were injected on gestational days 16 and 18 with vehicle (dimethylsulfoxide), A1221 (1 mg/kg), or estradiol benzoate (50 μg/kg). Developmental parameters, estrous cyclicity (females), and timing of reproductive senescence were monitored in the offspring through 9 months of age. Expression of 48 genes was measured in 3 hypothalamic nuclei: the anteroventral periventricular nucleus (AVPV), arcuate nucleus (ARC), and median eminence (females only) by real-time RT-PCR. Serum LH, testosterone, and estradiol were assayed in the same animals. In males, A1221 had no effects; however, prenatal estradiol benzoate increased serum estradiol, gene expression in the AVPV (1 gene), and ARC (2 genes) compared with controls. In females, estrous cycles were longer in the A1221-exposed females throughout the life cycle. Gene expression was not affected in the AVPV, but significant changes were caused by A1221 in the ARC and median eminence as a function of cycling status. Bionetwork analysis demonstrated fundamental differences in physiology and gene expression between cycling and acyclic females independent of treatment. Thus, gestational exposure to biologically relevant levels of estrogenic endocrine-disrupting chemicals has sexually dimorphic effects, with an altered transition to reproductive aging in female rats but relatively little effect in males.

In mammals, reproductive aging is associated with the loss of reproductive capacity and dysregulation of the hypothalamic-pituitary-gonadal (HPG) axis. In the hypothalamus, there are transcriptional (1), translational (2–6), and ultrastructural morphological changes (7), culminating in altered expression and release of GnRH in a species-specific manner. Additionally, in rodents, feedback of gonadal sex steroid hormones on the hypothalamus and the pituitary gland becomes eroded during the process of reproductive aging. Although all levels of the HPG axis undergo age-related changes, mounting evidence suggests that changes in the neuronal/glial network regulating GnRH release from the hypothalamus plays a key role in reproductive senescence (8).

Recent data suggest that the process of reproductive aging is regulated by both genetic and environmental factors. Genetic studies reveal numerous genes associated with the menopausal transition in women (9). Other studies have focused on the role of the environment and especially endocrine-disrupting chemicals (EDCs) in the reduction of fertility and fecundity in both sexes. EDCs are an exogenous chemical, or mixture of chemicals, that interfere with any aspect of hormone action and include plastics and plasticizers (phthalates, bisphenol A), pharmaceuticals (diethylstilbestrol), pesticides (dichlorodiphenyl trichloroethane [DDT]), and industrial contaminants [polychlorinated biphenyls (PCBs), dioxins] among others. Because hormone systems regulate physiological processes such as metabolism, reproduction, and stress, these compounds undoubtedly have adverse effects on endocrine health in wildlife and humans (10, 11). Exposure to EDCs during critical periods of development (eg, the perinatal period) can result in lifelong alterations and an increased disease burden in adulthood, a concept referred to as the fetal basis of adult disease (12). Regarding reproductive senescence, emerging evidence suggests that exposure to EDCs during a critical period of development hastens the transition to acyclicity and causes other reproductive changes in female rodents (10, 13–18). However, to our knowledge, no study has investigated the effects of gestational exposure to EDCs on the aging male hypothalamus.

PCBs are industrial contaminants that are banned in many countries but that persist in the environment. Aroclor 1221 (A1221), the compound used in the present study, is a lightly (21%) chlorinated mixture of PCB congeners that were used as lubricants and adhesives in capacitors, turbines, and plasticizers (19, 20). Although PCBs have been banned for decades, exposure continues through contamination of soil, water, and food sources, especially fish. Nearly all humans and wildlife were exposed to PCBs (21, 22), with those exposed in utero 40-50 years ago at the peak of production entering reproductive senescence. As for the mechanism of action, no EDC is a pure hormone agonist or antagonist; in the case of A1221, weak estrogenic activity is detected in binding assays (23, 24), and it may also have thyroid (25) or androgen (26) activity and interfere with aromatase (27). A1221 is not believed to interact with aryl hydrocarbon receptor, as do more heavily chlorinated PCBs (28).

Here we investigated sexually dimorphic effects of perinatal exposure to A1221 on the aging hypothalamus of rats. The mixture of congeners in A1221 is more readily metabolized than most PCBs (29, 30) and is therefore less likely to bioaccumulate. This allows us to investigate the long-term consequences of a relatively acute exposure to EDCs. Previous studies of A1221's effects on reproductive aging in female rats (17, 18, 31) produced inconsistent results and did not investigate molecular end points. Although Gellert and Wilson (31) investigated gestational exposure to A1221 on male fertility at 6 months of age, to our knowledge, no study has investigated the long-term hypothalamic effects of A1221 on more aged male rodents.

Our goal was to determine whether gestational exposure to A1221 resulted in the following: 1) altered timing and progression of reproductive senescence (acyclicity) in female rats; 2) alterations in gene expression in key brain regions associated with the progression of reproductive aging in females; and 3) changes in male reproductive physiology and gene expression throughout the life cycle. Our overarching goal was to determine the molecular underpinnings of hypothalamic reproductive aging and how those processes might be disrupted by perinatal exposure to EDCs.

Materials and Methods

Animals

All animal protocols were conducted in accordance with the Guide for the Care and Use of Laboratory Animals and approved by The University of Texas at Austin Institutional Animal Care and Use Committee. Sprague Dawley rats (3-4 month virgin females) were purchased from Harlan Laboratories (Houston, Texas) and impregnated in-house as described previously (32, 33). Two to 3 weeks before conception, animals were switched to low phytoestrogen Harlan-Teklad 2019 Global Diet (Harlan-Teklad, Indianapolis, Indiana) ad libitum and housed individually under constant humidity and temperature (21-22°C) with a partially reversed 12-hour light, 12-hour dark cycle (lights on at 11:00 pm). The morning after successful mating (sperm positive vaginal smear) was termed embryonic day (E) 1. On E16 and E18, during the period of sexual differentiation of the rat brain (34–37), dams were weighed and randomly assigned to 1 of 3 treatment groups and injected with 0.1 mL vehicle [dimethylsulfoxide (DMSO) 99.5% Sigma number D4540; Sigma, St Louis, Missouri]; 50 μg/kg estradiol benzoate (EB; Sigma number E8515); or 1 mg/kg A1221 (AccuStandard, New Haven, Connecticut, number C221N).

Doses and routes of exposure were chosen based on previous work in our laboratory (32, 33, 38, 39) and others (40–44). The choice of EB was made to enable us to compare with our previous work and to allow us to compare the effects of A1221, which has weakly estrogenic properties as one of its potential mechanisms (23, 24). However, we acknowledge that A1221 is not a pure estrogen so the use of EB also enabled us to differentiate estrogenic from nonestrogenic effects. Although we did not measure body burden, we estimated that each pup is exposed to approximately 1:500 of the dose (ie, 2 μg/kg A1221 or 100 ng/kg EB) based on previous literature (45). In humans, concentrations of PCBs in maternal serum (46–49), cord blood (50), and milk fat (49) are between 1 and 9 ppb. Therefore, our dose is estimated to be within the range of human exposure to PCBs. Because of differences in metabolism of the estrogenic control (EB) (51, 52) and A1221 (30, 53–57) and to maintain consistency with other experiments in our laboratory (32, 33, 38, 39), EB was given as a sc injection and A1221 was given as an ip injection. To control for route, half of the DMSO controls were exposed via an sc injection and half were given an ip injection. Route was included as a covariate for analysis and was not found to significantly contribute to the outcomes observed.

On the day after birth [postnatal day (P) 1], litter composition, birth weights, and anogenital distance were recorded, and the litters were culled to equal sex ratios, with litter size ranging from 6 to 12 pups. Litter size was later considered as a covariate in our regression model, and we found no effects on any endpoints measured in this study. Body weights were monitored weekly throughout the life cycle, and anogenital distance was measured weekly until weaning (P21). Weaned pups were housed with same sex littermates (2-3 per cage) until P90. For a companion study, littermates were euthanized on P15, P30, P45 and P90. For the current study, 1 male and 1 female per litter were designated for aging to 9 months. Thus, the statistical unit of analysis is the litter. Rats were monitored for secondary sex characteristics of the onset of puberty daily (preputial separation in males, vaginal opening in females). Beginning on P90, animals were randomly assigned a new cage mate. Because of the large number of animals necessary for both studies, the animals were raised in 5 cohorts, with treatments equally distributed across each cohort.

Tissue collection and storage

Females were classified as reproductively cyclic (regular or irregular cycles) or as acyclic (persistent estrus; see below for details). Those females in persistent estrus were euthanized after at least 14 consecutive days of cornified smears and as close to 9 months of age as possible. Unless females were in persistent estrus, they were euthanized on the day of proestrus. Final sample sizes for females were as follows: cyclic DMSO, n = 13; acyclic DMSO, n = 8; cyclic EB, n = 11; acyclic EB, n = 6; cyclic A1221, n = 9; and acyclic A1221, n = 13. For males, 1 male/litter (DMSO, n = 22; EB, n = 18; and A1221, n = 22) was included in the analysis of the somatic and hormonal data, with a subset of hypothalamic punches used to measure gene expression (DMSO, n = 11; EB, n = 8; and A1221, n = 11). Rats were euthanized by rapid decapitation 1-3 hours before lights out. Brains were removed and sectioned in 1-mm coronal sections using an ice-cold stainless steel brain matrix. Sections were placed on an ice-cold microscope slide and snap frozen on dry ice. Frozen sections were placed on a freezing stage, allowed to equilibrate to −15°C to −20°C, and micropunches (0.98 mm diameter) were taken from each region of interest and placed in a frozen Eppendorf tube. Photographs were captured of sections before and after punching to ensure consistency across the cohorts. Trunk bloods were collected and allowed to clot, and serum was separated via centrifugation (1500 × g for 5 minutes). Tissues and serum were stored at −80°C until use.

RNA extraction

RNA was extracted from frozen anteroventral periventricular (AVPV), arcuate (ARC), and median eminence (ME; females only) punches of males and females using a Allprep DNA/RNA mini- (AVPV, ARC) or micro (ME)-kit (QIAGEN, Valencia, California), according to the manufacturer's protocols. RNA samples were eluted with nuclease-free water and treated with 1 U of TURBO deoxyribonuclease (Applied Biosystems Inc, Foster City, California) to rid samples of genomic DNA before ethanol precipitation. Resuspended samples were diluted to a concentration of 50 ng/μL (AVPV, ARC) or 1 ng/μL (ME). All samples were run on a Bioanalyzer 2100 (Agilent Technologies, Santa Clara, California) to assess RNA concentration, purity, and integrity.

Taqman microfluidic real-time PCR cards

Samples were run on custom-designed, microfluidic, 48-gene PCR cards (Applied Biosystems Inc, Foster City, California), with specific gene assays chosen based on a priori hypotheses and published reports on their importance in neuroendocrine function and sensitivity to disruption by EDCs (Supplemental Tables 1, 2, 4, and 5, published on The Endocrine Society's Journals Online web site at http://endo.endojournals.org). For technical reasons female ME samples were run on a separate Taqman low-density array (Supplemental Table 3), which was composed of a slightly modified panel of neuroendocrine genes. We did not have enough arrays from this lot to run the male MEs. mRNA (200 ng AVPV, ARC; 10 ng ME) was converted to cDNA using a high-capacity cDNA reverse transcription kit (Applied Biosystems) according to the manufacturer's protocol. Product was stored at −20°C until use.

Real-time RT-PCR was carried out on an ABI ViiA7 using Taqman universal master mix (Applied Biosystems) and the following run parameters: 50°C for 2 minutes, 95°C for 10 minutes, 45 cycles of 95°C for 15 seconds, and 60°C for 1 minute. Relative expression was determined for each sample using the comparative cycle threshold method (58–60). Samples were normalized to either 18s or Gapdh and calibrated to the median δ-cycle threshold of the DMSO females to determine the relative change in expression for each gene.

Serum hormone assays

Serum LH was measured in duplicate 50-μL samples in the laboratory of Dr Michael Woller (University of Wisconsin-Whitewater), by double-antibody competitive binding RIA. The assay was performed using the rat LH RP-3 standard from the National Hormone and Pituitary Program of the National Institute of Diabetes and Digestive and Kidney Diseases (kindly provided by Dr. A. F. Parlow, Torrance, California). All samples were measured in a single assay, and the intraassay variability was 3.85%.

Serum concentrations of sex steroid hormones were measured using a RIA for estradiol (Beckman Coulter, Webster, Texas) and testosterone (MP Biomedicals, Santa Ana, California) according to the manufacturer's recommended protocols. Samples were run in duplicate if possible. Samples having a coefficient of variation of 10% or greater were rerun when possible or dropped from the analysis.

Two estradiol assays were run with an assay sensitivity of 2.2 pg/mL, and the intraassay variability of 6.48% and 1.64%, respectively. The interassay coefficient of variation was 4.55%. One T assay was run with an assay sensitivity of 0.03 ng/mL and the intraassay variability of 1.40%.

Statistics

Data in females were analyzed for interactions of reproductive cycle status and treatment for all somatic markers recorded as well as gene expression data. Females were classified as either cyclic or acyclic using the vaginal smear data from the last 30 days of life. A female was cyclic if she had more than 4 nucleated (indicative of proestrus) days in the last 30 days (regular cyclers should have at least 5) or acyclic if there were fewer than 4 nucleated days during the last 30 days of life and/or the female was in persistent estrus for at least 14 days.

For gene expression data, statistics were performed using relative expression values for each sample. For females, multiple regression analysis was conducted using PASW software (IBM, Armonk, New York) to compare each end point (genes, hormones, and endocrine tissues) using cycle status and treatment as independent variables. For those end points in which a significant main effect of cycle status or treatment or an interaction was observed, data were split by independent variables, and a follow-up analysis was performed to determine specific differences between each group. For the males, a 1-way ANOVA was conducted using treatment as an independent variable. In both sexes, if data did not meet the assumptions for multiple regression analysis, data were transformed (natural log or square root) and reanalyzed. Data (transformed or not) that did not meet the assumptions for multiple regression were analyzed using a Kruskal-Wallis test followed by a Mann-Whitney test between each group. For hormone concentrations, an effect was considered significant at P < .05. For gene expression data, a Benjamini-Hochberg false-discovery rate correction (61, 62) was used to correct our P values to account for the large number of variables measured. Parametric data were tested for outliers using the z-score of the residuals from the initial regression. A data point was considered an outlier if the residual was greater than 2.5 SD from the initial line of best fit. Nonparametric data were tested for outliers using the Grubbs' outlier test. Confirmed outliers were excluded from final analysis.

Vaginal smear data were analyzed using rule-based pattern matching techniques to determine the number and length of each cycle for each animal. Briefly, cycles were identified by scanning the smear data from the day of vaginal opening until euthanasia for the diestrus 2 to proestrus transition. The make-up and length of each cycle was recorded throughout the life cycle. A repeated-measures ANOVA was conducted using the moving average of the cycle length (using 3 cycle lengths) for each animal to account for any missing data.

A bootstrap technique (63) as previously used by our laboratory (64) was used to examine possible relationships between gene expression, endocrine tissues, and serum hormones throughout development using Matlab software (The Mathworks, Natick, Massachusetts).

Hierarchical cluster analysis and heat maps were performed using a Multiple Experiment Viewer version 4.8.1 (MeV TM4.org), and clusters were validated using R statistical packages.

Results

Animals were monitored from birth to 9 months of age to determine whether perinatal exposure to estrogenic endocrine disruptors altered reproductive aging in males and females. We measured changes in reproductive tissues, tracked female estrous cycles, and measured 48 genes using a Taqman low-density array in 3 brain regions selected for their importance in reproductive function: AVPV and ARC (males and females) and ME (females only). Although no genes survived a Benjamini-Hochberg false discovery rate correction (61, 62), all genes had been specifically chosen based on our a priori hypotheses about their known neuroendocrine functions. Therefore, we have reported gene expression data as significant (P < .05) or as a trend (P < .1).

Effects of prenatal EDCs on females: somatic measurements, hormones, and estrous cycles

In females, few effects of treatment were observed on pituitary weight, gonadal somatic index (GSI; gonadal weight/body weight), or uterine weight (Figure 1). However, there were main effects of cycle status. In the acyclic animals, pituitary weights were increased, and gonadal somatic index and uterine weights were decreased compared with cyclic counterparts. Serum LH concentrations in cyclic control (DMSO) rats were substantially higher than in cyclic EB or A1221 rats. The lack of significance in this data set is likely attributable to the pulsatile nature of LH release and consequent high variability. Our goal was to euthanize the cycling females during their preovulatory LH surge. Therefore, further analysis was used to determine whether treated females were indeed having an LH surge. Each cycling female was categorized as surging or not surging based on their LH concentration (surging > 1 ng/mL) followed by a χ² test using the frequencies in the DMSO group as our expected values. This analysis revealed that there was a significant decrease in the number of surging animals treated with EB and A1221 in the cycling animals [DMSO, 61.5%; EB, 22% (P = .015); and A1221, 16.7% (P = .001)]. For serum estradiol, there were no significant effects of cycle status or treatment.

Figure 1.

Effects of prenatal A1221 or EB on hormones and somatic markers in aging female rats. Females were classified as cyclic or acyclic. Significant P values or trends for main effects of treatment or status and status × treatment interactions are indicated at the top of the graphs. Differences between cyclic and acyclic animals within groups are indicated by P values overlaid on relevant bar pairs. Cycle status had significant effects on pituitary weight, GSI, and uterine weights. Although not significant, serum LH (A) was decreased in acyclic females exposed to DMSO and EB, an effect not observed in the A1221 animals. Pituitary weights (B) significantly increased in acyclic animals, whereas GSI (D) and uterine weights (E) decreased in acyclic females when compared with cycling animals. No effects were observed on serum E₂ concentrations (C).

Repeated-measures ANOVA analysis of the moving average of the cycle length indicated there was a significant effect of treatment (P = .03) on cycle length throughout the lifecycle (Figure 2). Post hoc analysis revealed that females exposed to A1221, but not EB, had significantly longer cycles throughout their lifecycle when compared with the DMSO females and that the length of cycles increased with age (P < .001) in all treatment groups.

Figure 2.

Length of estrous cycles in females treated with A1221 or EB throughout the life cycle. The length of each cycle was determined beginning from vaginal opening through the end of life. Data were transformed to the moving average for each animal to account for any missing data and analyzed by repeated-measures ANOVA. Data are presented as the mean cycle length, in days, for each treatment group. The length of all cycles increased over time, independent of treatment. Additionally, treatment significantly altered the length of estrous cycles (P = .03). Post hoc analysis revealed that females treated with A1221 (dashed line) had longer cycles when compared with the DMSO controls (solid black line) throughout the life cycle (P = .03). EB exposure (gray line) had no effect on cycle length compared with DMSO. There was no interaction of treatment over time.

Effects of prenatal EDCs on females: gene expression in AVPV, ARC, and ME

Analysis of gene expression in hypothalamic regions of females revealed that cycle status (cyclic or acyclic) played a substantial role in results. Therefore, data are presented for effects of prenatal treatment, cycle status, and interactions of treatment by cycle status. Significant effects (P < .05) and trends (0.05 < P < .075) are graphed. Data for all the genes detected in each hypothalamic subregion are provided in the Supplemental Tables. In the female AVPV (Supplemental Table 1), only 1 gene was significantly altered by treatment: Grin2c (P = .046) was increased in EB treated females compared with DMSO controls.

In the ARC (Figure 3 and Supplemental Table 2), 2 genes were significantly affected by treatment: Lepr and Gnrhr. One gene, Oxt, was significantly altered by cycle status. A total of 14 genes displayed a significant status by treatment interaction (P < .05; Supplemental Table 2) and are also indicated in red on Figure 5B. Further analysis revealed that 8 of these latter mRNAs (Dnmt3a, Ar, Pdyn, Lepr, Mc3r, Oxt, Grin2b, and Grin2d; Figure 3) had lower expression in the acyclic A1221 females when compared with the acyclic DMSO counterparts. One gene, Esr1, was higher in the cyclic A1221 females when compared with cyclic DMSO and cyclic EB counterparts (Figure 3). As a whole, affected genes fell into 4 functional categories: epigenetic processes (Dnmt3a); steroid hormone receptors (Esr1 and Ar); neuropeptide signaling (Pdyn, Lepr, Mc3r, and Oxt); and glutamate signaling (Grin2b and Grin2d).

Figure 3.

Expression of 10 genes in the ARC are shown, selected based on significant effects of treatment, cycle status, or interactions of status × treatment. The most frequent finding was an interaction effect, with lower gene expression in the acyclic A1221 females when compared with acyclic DMSO counterparts. In addition, many genes showed increased expression in DMSO acyclic compared with DMSO cyclic females, a pattern that was reversed in A1221 females.

Figure 5.

Clustergrams of gene expression in females is shown for AVPV (A), ARC (B), and ME (C). Clusters were validated using R statistical packages and validated gene clusters are indicated by a red box. To maintain legibility of the figure, red boxes indicate the most external validated gene clusters (side), whereas P values are indicated for clusters by treatment and cycle status (top). Genes with a status × treatment interaction are indicated in red text (P < .1), and bolded (P < .05) to identify gene clusters associated with specific status/treatment-related expression patterns. This effect was observed in the ARC in which all significantly altered genes are in 1 cluster. The P value from the validation is listed for the clusters of the groups. DC, DMSO cyclic; DA, DMSO acyclic; EC, EB cyclic; EA, EB acyclic; AC, A1221 cyclic; AA, A1221 acyclic.

In the ME (Supplemental Table 3), there were no significant main effects of treatment or reproductive status. However, 4 genes displayed significant status by treatment interactions (Figure 4) in 2 functional categories: steroid hormone receptors (Esr1, Nr3c1) and neurotransmitter systems (Grin2b, Gad2). As shown in Figure 4, these 4 genes were expressed at higher levels in DMSO cycling than DMSO acyclic rats (significant for Nr3c1 and Grin2b; trend for Esr1). By contrast, this cycling status effect was not seen in EB or A1221 rats. In addition, DMSO cyclic females had higher Esr1 mRNA levels compared with EB and A1221 cyclic rats and a trend for higher Gad2 mRNA compared with EB cyclic rats.

Figure 4.

Expression of 4 genes in the female ME are shown, selected based on significant interactions of status × treatment. In general, there was lower expression in the cyclic A1221 and the cyclic EB females when compared with their cyclic DMSO counterparts. In DMSO females, all 4 genes had lower expression in acyclic than cyclic rats, an effect not seen in either A1221 or EB females.

Effects of prenatal EDCs on males

In an effort to determine whether the 9-month-old males were showing signs of reproductive aging, we compared their somatic data with the 3-month-old siblings from a companion study (not published). We found that regardless of treatment, the aged vehicle males had significantly lower serum LH and estradiol (E₂) (DMSO group) as well as significantly decreased GSI and pituitary weights.

When analyzing the treatment effects within the aged males, we found that treatment had no effect on body weight, pituitary weight, gonad weight, serum LH, or serum T (Table 1). A main effect of treatment was observed on serum estradiol (P = .022), although post hoc analysis revealed only trends for higher estradiol levels in the EB (P = .066) and A1221 (P = .053) males when compared with their DMSO counterparts. One gene was altered in the male AVPV: Avpr1a (P = .05; Table 1). Two genes were significantly altered in ARC of males: Bdnf (P = .008) and Ahr (P = .006), with trends for Slc17a6 and Oxt. The complete male data sets are shown in Supplemental Tables 4 (AVPV) and 5 (ARC).

Table 1.

Effects of Prenatal Treatment With DMSO, EB, or A1221 in Aging Male Rats

	P Value	DMSO		EB		A1221
		Mean	SEM	Mean	SEM	Mean	SEM
Body weight, g	.969	563	22	556	11	559	13
Pituitary weight, mg	.364	14	0.3	13	0.5	13	0.3
Adrenal weight, mg	.987	50	2	49	2	49	2
GSI, × 10−3	.961	7	0.2	8	0.1	8	0.2
Serum LH, ng/mL	.478	0.46	0.05	0.63	0.20	0.44	0.03
Serum T, ng/mL	.364	0.47	0.06	0.46	0.14	0.65	0.13
Serum E2, pg/mL	.022	2.36	0.26	7.72	2.97	4.13	0.59
AVPV Avpr1a	.05	0.75	0.10	1.25a	0.19	0.93	0.11
ARC Ahr	.006	1.18	0.14	1.49a	0.18	0.84	0.06
ARC Bdnf	.008	0.90	0.13	2.14a	0.35	1.42	0.26
ARC Slc17a6	.057	0.87	0.12	1.32	0.07	1.01	0.14
ARC Oxt	.091	0.76	0.14	1.24	0.25	0.52	0.07

P values in bold text indicate a significant main effect of treatment. All somatic and hormonal data are shown. Only those genes with significant differences or trends in AVPV or ARC are shown, with all other gene expression data on males presented in Supplemental Tables 4 and 5. No significant effects of prenatal treatment on somatic markers were detected. For serum hormones, a significant main effect of treatment on E₂ was found, although post hoc analysis indicated only trends for higher concentrations in EB and A1221 males when compared with DMSO. For gene expression, 3 genes had significant main effects of treatment. One gene in AVPV (Avpr1a) and 2 genes in ARC (Ahr, Bdnf) were significantly higher in EB compared with DMSO. Two genes in ARC (Slc17a6, Oxt) showed trends for such effects.

^aSignificant differences between treatment group and the control group (DMSO) after post hoc analysis.

Hierarchical cluster analysis in females

To determine relationships among gene expression patterns within each brain region, treatment, and cycle status in females, a hierarchical cluster analysis was conducted and clustergrams were generated for each brain region (64). Male cluster analysis was not performed because there were too few groups to be able to validate the clusters. Multiple experiment viewer (MeV 4.8.1) software was used to generate heat maps and perform analysis using the average linkage method and the correlation coefficients to express similarity. Clusters were validated using R statistical packages.

Clusters by cycle status and treatment revealed only treatment differences in the AVPV (Figure 5A, top) with EB females tending to cluster separately from DMSO and A1221 (P = .08). However, no validated gene clusters (Figure 5A, side) were observed in the AVPV, making it difficult to determine which genes might be driving the differences in the clustering of the EB animals.

In the ARC, DMSO cyclic females tended to cluster with the acyclic A1221 females (P = .07) and vice versa (A1221 cyclic with DMSO acyclic) (P = .08) (Figure 5B). This effect was also observed in the ME, although with a weak trend (P = .1 and P = .14, respectively). In the ARC, 2 validated gene clusters were identified: 1) the upper cluster composed of Gnrhr and Bdnf displayed increased expression in the EB females, and 2) the lower cluster consisting of decreased expression in the A1221 acyclic vs DMSO acyclic but similar expression to the cyclic DMSO females.

In the ME, 2 validated gene clusters were distinguished by increased expression in the A1221 cyclic females (upper cluster) and differences between the DMSO cyclic females and EDC-treated acyclic females (EB acyclic and A1221 acyclic) (lower cluster, Figure 5C).

Integration of genes, hormones, and somatic data networks in females

To examine the relationships among gonadal hormones, somatic changes, and hypothalamic gene expression relative to reproductive cycling status in females, we used the network analysis platform Cytoscape (1, 65–67) to generate networks based on significant Pearson correlation coefficients between hormones, endocrine tissues, and relative gene expression in each brain region. Data were first collapsed across the treatment groups in each region to identify common relationships that change during the transition from cyclicity to acyclicity. Genes with a significant interaction of treatment and cycling status in the ARC and ME are indicated with larger font (P < .1) and bolded (P < .05) to determine how/whether their relationships with other genes, hormones, and somatic markers might be changing in the transition to acyclicity (Figure 6).

Figure 6.

Cytoscape networks of genes, hormones, and somatic changes in 3 brain regions involved in reproductive function (AVPV, ARC, and ME) in cyclic vs acyclic females. Genes with a significant interaction of treatment × status are indicated by larger font (P < .1) and bolded (P < .05). Overall results show that the networks differ profoundly between cyclic and acylic rats in each brain region.

Relationships in the AVPV showed the greatest difference between cyclic and acyclic females, with the cyclic females displaying a number of positive and negative correlations (Figure 6). In the acyclic females, there were few correlations, mostly negative with serum LH level. Although there were not as many differences in the number of significant correlations in the ARC with reproductive aging, there were differences in end points that serve as hubs of regulation. In the cyclic ARC, Gnrh1 had numerous positive correlations; 13 of the 27 were genes that were significantly altered in the ARC, whereas the GSI and serum estradiol had mostly negative correlations. In the acyclic ARC, there were numerous positive correlations with Gnrhr (11 of 25 altered in the ARC) and pituitary weight and numerous negative correlations with uterine weight. Finally, in the ME, a similar number of correlations were observed in both networks, with most between genes rather than with hormone or somatic end points. Notably, the number of positive correlations with Nr3c1 (glucocorticoid receptor) almost doubled from 8 to 14 from cyclic to acyclic.

Discussion

Endocrine disrupting chemicals have become a prototypical model for the fetal basis of adult disease, with prenatal exposures during critical developmental periods associated with adult dysfunctions related to reproductive, metabolic, thyroid, and many other target systems. However, little of this research has extended the fetal basis model to aging populations and an even smaller subset on reproductive senescence. The current study sought to determine whether gestational exposure to a PCB mixture, A1221, would hasten reproductive aging in males and females through premature aging of neuroendocrine function, assayed here through a network of gene expression in three key hypothalamic regions. The basis for this work was the body of epidemiological evidence suggesting that there are correlations between exposures to some classes of EDCs, notably organochlorine pesticides, and advancements in the timing of menopause in humans (68–71). However, some of these results have been contradictory (68–73), possibly due to differences in exposure timing, dose, and type of EDC, as well as population differences. Although these studies did not focus on PCBs, a recent report correlated serum concentrations of estrogenic congeners of PCBs with increased length of menstrual cycles in women aged 18 to 44 years (74). Animal models may be more useful in determining the cause-and-effect links between early life EDCs and reproductive aging. Our laboratory recently found that perinatal exposure to the pesticide methoxychlor advanced reproductive senescence in Fischer female rats and altered gene expression and DNA methylation in the hypothalamus 16 months after the original exposure (15). Other rodent work shows that various EDCs, including A1221, can cause irregular estrous cyclicity in females and accelerate the transition to persistent estrus (10, 13, 14, 17, 18, 33). Our current study extends this work and adds in an analysis of molecular changes to the hypothalamic neural network that controls reproduction.

Based on the literature to date, our experimental design sought to address several novel points. First, we tested how PCBs at dosages consistent with human and wildlife exposures might affect reproductive physiology in both males and females. There is a paucity of data on sex differences caused by EDCs beyond the early parts of life. Second, we characterized physiological changes in the animals throughout their first 9 months of life, representing a significant portion of a rat's life cycle, and associated them with molecular end points in brain regions important for reproductive function and physiology. Third, we performed cluster and network analyses to gain insights into the complex relationships caused by prenatal exposure to A1221 or EB; this goes beyond the gene-by-gene or hormone-by-hormone approach and has been very informative about the big picture of bionetworks (64, 75).

Reproductive physiology is altered in females exposed to A1221 in utero

Although we did not see a significant advancement of the timing of reproductive senescence or in the number of acyclic females by 9 months of age, we did observe a significant increase in the length of estrous cycles in A1221 females (Figure 2) throughout the life cycle. Elongated cycles are a common characteristic of reproductive aging in both primates and rodents before transitioning to acyclicity (8), and the appearance of such cycles earlier in life may be indicative of an aging phenotype. Additionally, we noted a marked decreased (albeit nonsignificant) in serum LH in cycling females exposed perinatally to A1221 (Figure 1A) as well as a significant decrease in the number of cyclic females displaying a preovulatory surge in both the EB- and A1221-treated animals when compared with their control counterparts. The pulsatile nature of LH release probably led to high variability in serum LH concentrations between individuals. In rodents, age-related decreases in LH release on proestrus presage reproductive decline and are associated with a decreased GnRH drive (6, 76). Further studies are necessary to test whether this is the case in prenatally EDC-exposed female rats.

Gestational exposure to A1221 alters cycle status specific expression of neuroendocrine genes in the ARC and ME, but not the AVPV, of female rats

The ARC plays a role in the regulation of GnRH pulsatility (reviewed in Reference 77) and in the negative feedback of hormones on GnRH/LH release, processes that are subject to age-related decline (78). Of the 3 regions studied here, the ARC had the greatest number of genes that were affected by EDC treatment interacting with cycle status. Interestingly, gene expression tended to be higher in acyclic than cycling rats in the control DMSO groups, whereas opposite effects were seen for the A1221 rats (gene expression was lower in acyclic than cyclic). EB rats tended to fall in the middle (Figure 3). Those genes that exhibited a reversal in pattern were Dnmt3a, Esr1, Ar, Pdyn, Lepr, Mc3r, Oxt, Grin2b, and Grin2d. For all of these genes, acyclic rats had higher expression than cycling rats (DMSO), an effect that was lost or reversed in A1221 rats. It is notable that there is hypertrophy of subsets of cells in the aging ARC (79, 80); our finding that expression of most genes is higher in acyclic than cycling rats in the control group is consistent with this result.

The identified genes in the ARC (Figure 3) fell into 4 functional categories: epigenetic processes, steroid hormone receptors, neuropeptide signaling, and glutamate signaling. Although changes in hypothalamic sex steroid hormone feedback and glutamatergic signaling are thought to contribute to the transition to reproductive senescence (81), it should be noted that many of these genes are important for signaling cross talk among neuroendocrine systems regulating reproduction, stress, and metabolism (eg, Pdyn, Lepr, Mc3r, and Oxt) (82–85). All of these neuroendocrine axes undergo age related changes in rodents (86, 87) and nonhuman primates (88). Although further studies are necessary to determine the precise implication of these gene expression changes in the ARC, our data provide intriguing evidence that gestational exposure to A1221 cause alterations in gene expression in the ARC, which could affect numerous neuroendocrine axes. Furthermore, our data provide insight into target gene candidates that may play roles in the transition from cyclicity to acyclicity.

In the ME, we found that levels of steroid hormone receptor genes for estrogen receptor-α and the glucocorticoid receptor (Esr1, Nr3c1) and neurotransmitter systems in glutamate and γ-aminobutyric acid transmission (Grin2b, Gad2) were affected by prenatal EDCs (Figure 4). In DMSO rats, acyclic females had lower expression of all 4 of these genes compared with cycling DMSO rats (notably, this is the opposite gene pattern from the ARC). Again, A1221 rats had completely different gene expression patterns by cycle status from the DMSO group. The ME is mainly composed of glial cells and axons projecting to the portal capillary system to release neuropeptides (89); the low occurrence of neuronal cell bodies in this region suggests that the gene expression results in this experiment are probably mainly from glial cells or local mRNA in axons (90). The few studies on reproductive aging investigating changes specifically in the ME show that stimulation of GnRH release from terminals with glutamate or glutamate agonists was decreased in middle-aged females compared with young (6), and the ultrastructural organization of the ME, and the relationship between glia and GnRH terminals in this region, was disrupted with age (7, 91).

We were surprised that so few effects of aging or cycle status were found in the AVPV, considering its role in the preovulatory surge (92). Therefore, we hypothesize that the LH surge may be delayed or attenuated in the cycling females exposed to EB and A1221. In comparing results for the 3 hypothalamic regions, there are distinct regional differences in responses to prenatal EDCs and in the effects of cycle status. Overall, results from the ARC and ME indicate that patterns of gene expression between cyclic and acyclic rats seen in control treatments are obliterated or even reversed in A1221 rats. Although this requires further experimentation, we hypothesize that A1221 females begin the transition of reproductive aging earlier in life, possibly as a result of altered signaling between the 3 brain regions during this transition. This hypothesis was supported by our hierarchical cluster analysis, whereby acyclic A1221 females cluster with the cycling DMSO females and vice versa in both the ARC and ME. The elongated cycles and lower serum levels of LH in A1221 cycling rats further support this idea. Finally, expression patterns of genes in the ARC and ME in cycling females exposed to A1221 are similar to the acyclic, and dissimilar to the cyclic, control rats.

Our data must be interpreted in light of some limitations. All rats were euthanized at a single time of day, and it is possible that the treatment effects observed on gene expression are due to other influences such as a disruption of circadian rhythms or changes to other endocrine organs that feed back onto the hypothalamus. Although we were able to assay multiple genes, serum hormones, and somatic end points, several additional end points (eg, expression and localization of proteins in the brain; release of neurotransmitters/neuropeptides) would be extremely valuable complementary information. Finally, the age of euthanasia (9 months) is early in the aging process, and later ages would be very informative. Nevertheless, even at 9 months, it is clear that there are distinct differences between the prenatal treatment groups in our aging population.

Network analysis reveals region-specific changes between cycling and acyclic females

To generate new hypotheses about hypothalamic and peripheral changes during the transition of reproductive aging, we used Cytoscape (65–67) to generate networks of hormones, genes, and endocrine tissues in each of the regions of interest. The most striking difference between cyclic and acyclic networks was seen in the AVPV. This result was both interesting and surprising because the gene-by-gene analysis did not reveal specific targets, whereas the network analysis gave a more holistic perspective on patterns of change with aging. In the cycling animals there were 3 hubs of negative correlations: serum E₂, adrenal weights, and uterine weights. However, in the acyclic females, the number of significant correlations was substantially decreased, and virtually all negative relationships were with serum LH concentrations. Because the AVPV is necessary for the LH surge (93), it is not surprising that we saw stark differences between the cyclic and acyclic network. Moreover, it should be noted that the correlations observed in the cycling AVPV reveal some unexpected relationships. Notably, serum E₂ and Pgr (progesterone receptor) were negatively correlated, in contrast to other studies (94–96). Most work on the regulation of the progesterone receptor by E₂ is conducted on models of ovariectomy with or without estradiol treatment. Our current study on intact females provides new information about changes in this relationship with aging. Our recent publication (64) reported that during early postnatal development of intact female rats, the E₂-Pgr connection also differs from the OVX model, suggesting that models of natural development and aging with a bionetwork approach is important in understanding physiological regulatory processes.

In the ARC, one of the more surprising findings was that the cycling animals had a hub of positive correlations with Gnrh1 and a negative hub with gonadal somatic index (ovary) weights. GnRH cell bodies of rats are primarily located in the AVPV but not the ARC of the rodent (97). Perhaps there is local translation of Gnrh1 in the axons of the GnRH neurons that is reflected by our gene expression measures. Although local translation (to our knowledge) has not been observed in GnRH neurons specifically, other neuroendocrine cells, eg, oxytocin and vasopressin, use this routinely to maintain and replenish neuropeptide levels at the site of release (90). In the acyclic rat's ARC, the hubs of positive correlations were Gnrhr and pituitary and adrenal weights, whereas the negative hub was uterus weight. It is notable that many of the genes that were altered by treatment in the ARC are associated with the 2 positive hubs in the cyclic and acyclic networks, suggesting these genes may have differing roles in the cyclic and acyclic ARC. Interestingly, 2 of these somatic end points (pituitary and uterus) are the same ones that show significant effects of status independent of treatment, supporting the ARC's role in the central regulation of reproductive aging.

In the ME, there were few differences in the number of correlations between cyclic and acyclic end points. Additionally, there were not distinct hubs as observed in the AVPV and ARC. Many of the correlations were between genes and not peripheral measures, which was surprising because the ME is the region of final output of the GnRH system to the HPG axis. The reduced number of correlations and lack of hubs in the ME compared with the AVPV and ARC may signify a differential role in reproductive function. With the presence of numerous neurons in the AVPV and ARC, many of which modulate GnRH neuronal activation and release, the coordination of their activity may be aptly reflected in a network of gene expression. However, although changes in gene expression are still likely important at the site of the terminals in the ME, its major function is neuropeptide and transmitter release, a process that may be regulated quite differently from upstream biosynthesis.

Summary and conclusions

Gestational exposure to EDCs can disrupt reproductive function and result in premature reproductive aging. Our study added novel insight into these processes by characterizing the long-term changes in reproductive physiology and gene expression throughout the hypothalamus in aging males and females. We found few effects of treatment on expression or somatic markers of aging in males. In females, however, animals exposed prenatally to A1221 underwent a different reproductive aging process than their control counterparts. Additionally, we identified genes in both the ARC and ME as potential targets of EDCs as well as biomarkers for reproductive aging. Finally, we used bionetworks to identify relationships of genes, hormones, and somatic changes that provide insight into the process of reproductive aging and serve as hypothesis generating tools for future experiments on reproductive senescence.