Early Life Exposure to Endocrine-Disrupting Chemicals Causes Lifelong Molecular Reprogramming of the Hypothalamus and Premature Reproductive Aging

Abstract

Gestational exposure to the estrogenic endocrine disruptor methoxychlor (MXC) disrupts the female reproductive system at the molecular, physiological, and behavioral levels in adulthood. The current study addressed whether perinatal exposure to endocrine disruptors reprograms expression of a suite of genes expressed in the hypothalamus that control reproductive function and related these molecular changes to premature reproductive aging. Fischer rats were exposed daily for 12 consecutive days to vehicle (dimethylsulfoxide), estradiol benzoate (EB) (1 mg/kg), and MXC (low dose, 20 μg/kg or high dose, 100 mg/kg), beginning on embryonic d 19 through postnatal d 7. The perinatally exposed females were aged to 16–17 months and monitored for reproductive senescence. After euthanasia, hypothalamic regions [preoptic area (POA) and medial basal hypothalamus] were dissected for real-time PCR of gene expression or pyrosequencing to assess DNA methylation of the Esr1 gene. Using a 48-gene PCR platform, two genes (Kiss1 and Esr1) were significantly different in the POA of endocrine-disrupting chemical-exposed rats compared with vehicle-exposed rats after Bonferroni correction. Fifteen POA genes were up-regulated by at least 50% in EB or high-dose MXC compared with vehicle. To understand the epigenetic basis of the increased Esr1 gene expression, we performed bisulfite conversion and pyrosequencing of the Esr1 promoter. EB-treated rats had significantly higher percentage of methylation at three CpG sites in the Esr1 promoter compared with control rats. Together with these molecular effects, perinatal MXC and EB altered estrous cyclicity and advanced reproductive senescence. Thus, early life exposure to endocrine disruptors has lifelong effects on neuroendocrine gene expression and DNA methylation, together with causing the advancement of reproductive senescence.

The control of reproductive aging in female mammals involves a complex interplay of the hypothalamus, pituitary, and ovary. The hypothalamus plays an important role in this life transition, with hypothalamic GnRH neurons changing the synthesis and release of the decapeptide with aging in a species-specific manner (1, 2). Throughout the life cycle, regulatory inputs to GnRH neurons from neuropeptides (e.g. kisspeptin), neurotransmitters [e.g. glutamate and γ-aminobutyric acid (GABA)], and neurotrophic factors (e.g. TGF and IGF) control GnRH gene expression and release, and the balance of these inputs determine the final output of GnRH release from nerve terminals into the portal capillary system that vascularizes the anterior pituitary gland (3–6). In addition, feedback from peripheral hormones, such as sex steroids (7, 8), regulates hypothalamic output via actions of estradiol, progesterone, and androgens on their respective steroid hormone receptors that are abundant in the hypothalamus. This complex network of neurons, glia, and steroid regulation changes substantially during reproductive aging (9–12). Nevertheless, relatively little is known about how the hypothalamus changes with age, and its cause-and-effect role on reproductive senescence.

The timing and progression of reproductive senescence is determined by a combination of genetic and environmental factors. To date, most research has focused on genetic predispositions (13). Nevertheless, the environment is likely to play a role, and there is speculation that endocrine-disrupting chemicals (EDC) may hasten reproductive aging, with the potential outcome of shortening an individual's reproductive lifespan. EDC are compounds in the environment that act upon the body's hormonal systems and include industrial contaminants, plastics/plasticizers, pesticides, and other compounds (14). Recent studies show that exposures to EDC during key developmental periods, especially prenatal/early postnatal life, can cause molecular/cellular changes that affect the function of the affected tissues later in life, a concept referred to as the fetal/developmental basis of adult disease (15). The mechanisms for these effects are diverse and probably involve epigenetic molecular changes, including DNA methylation and histone modifications, the consequences of which are manifested later in life. Intriguingly, the latency between exposure and adult dysfunction extends into the realm of reproductive aging. Animal studies show that reproductive aging is accelerated by EDC [methoxychlor (MXC) (16), bisphenol A (17), and dioxins (18)], and recent epidemiological evidence links developmental EDC exposures to accelerated menopause [diethylstilbestrol (19) and perfluorocarbons (20)].

The current study sought to draw mechanistic connections between early life developmental exposure to an estrogenic endocrine disruptor, MXC, its actions on the hypothalamus, and the consequences on reproductive senescence. MXC is an organochlorine pesticide with estrogenic and antiestrogenic properties (16, 21, 22). Some studies have investigated effects of MXC on the female hypothalamus (23), but few have taken a mechanistic approach to understanding how MXC may affect hypothalamic gene expression (24, 25), and none has linked early life exposures to reproductive aging. Amid concerns that infertility is on the rise (26, 27), understanding the links between EDC exposures and reproductive aging are critical. Furthermore, such work extends observations on the fetal basis of adult disease beyond the young adult and into the realm of aging. Finally, this study focused on the hypothalamic mechanisms by which EDC may reprogram the neuroendocrine circuit regulating reproduction. Therefore, as a whole, this project's goal was to determine early life programming of the entire life cycle's reproductive capacity and the underlying molecular mechanisms.

Results

Loss of estrous cyclicity is accelerated in perinatal endocrine-disrupted rats

Rats were monitored by daily vaginal smears for a minimum of 12 consecutive days per month, or longer (up to 20 d) if more time was required to establish cyclicity status, during the 5-month period leading up to euthanasia. Table 1 shows that most dimethylsulfoxide (DMSO)-treated rats were still exhibiting regular cycles at 16–17 months of age. The low-dose MXC group had some diminution in regular cycles with aging, but this was only significantly different from the vehicle group at 14 months. By contrast, the high-dose MXC and the estradiol benzoate (EB)-treated groups had few (if any) rats still exhibiting cycles by 13 months of age. Fisher's exact test revealed that these latter groups were significantly different from the control group from 13 to 17 months of age (P < 0.001–0.0001). Thus, perinatal EDC treatment was associated with premature reproductive failure.

Table 1.

Estrous cycle classification for aging rats treated between E19 and P7

	13 Months	14 Months	15 Months	16 Months	17 Months
DMSO	83% Reg	100% Reg	83% Reg	80% Reg	100% Reg
	n = 6	n = 6	n = 6	n = 5	n = 5
20 μg of MXC	100% Reg	78% Reg*	78% Reg	66% Reg	66% Reg
	n = 9	n = 9	n = 9	n = 9	n = 9
100 mg of MXC	0% Reg**	22% Reg**	11% Reg**	11% Reg**	0% Reg**
	n = 9	n = 9	n = 9	n = 9	n = 9
1 mg of EB	0% Reg**	0% Reg**	0% Reg**	0% Reg**	0% Reg**
	n = 5	n = 5	n = 5	n = 5	n = 5

Fisher's exact test was conducted with significance at *, P < 0.001; **, P < 0.0001 vs. control (DMSO) at the same age. Reg, Regular estrous cycles of 4–5 d; n, sample size at each age.

Serum estradiol and progesterone concentrations in endocrine-disrupted rats

Serum hormone levels were measured in terminal blood samples. Figure 1 shows that serum estradiol concentrations in the aged females were affected by perinatal treatment (P < 0.005, Kruskal-Wallis). Levels of estradiol were significantly lower in the EB group than all the other groups (Mann-Whitney post hoc, P < 0.005). Both the low-dose and high-dose MXC groups also differed from the control (P < 0.05). For progesterone, no differences were determined among the groups (P = 0.375).

Fig. 1.

Serum concentrations of estradiol (A) and progesterone (B) in aging female rats at 16–17 months of age. Levels of estradiol were significantly lower in the EB group compared with all others. Levels were also lower in the two MXC groups compared with DMSO. Serum progesterone concentrations were unaffected by perinatal exposure to EDC. #, P < 0.05 vs. control; *, P < 0.005 vs. all other groups.

Hypothalamic gene expression is reprogrammed in aging, endocrine-disrupted rats

The low-density PCR array was used to measure expression of 48 genes in the preoptic area (POA) and medial basal hypothalamus (MBH) of perinatally exposed rats euthanized at aging endpoints. Figure 2 shows expression of those genes in the POA that differed in the aging rats by 50% or more between the perinatal vehicle- and EDC-treated groups and based on raw P values of less than 0.05 before Bonferroni correction. Seventeen genes were differentially regulated by perinatal EDC treatments according to these criteria, including steroid hormone receptors/coregulatory factors/binding partners (Esr1, AR, Pgr, Srd5a1, Sts, and Arnt); glutamate/GABA receptor subunits (Gria3, Grin2b, Grik2, and Gabbr1); neurotrophic factors/receptors (Tgfa, Tgfb1, and Igf1r); neuropeptides/receptors (Kiss1, Kiss1r, and Gnrhr); and the transcription factor, Stat5b. Of these 17 genes, 15 showed a similar pattern of expression in the aging rats, with mRNA levels higher in the EB and/or MXC high-dose groups compared with vehicle. Two genes had unique expression patterns. Kiss1 mRNA levels were significantly lower in the EB group compared with all other groups (P < 0.001). Gnrhr mRNA levels were higher in the high-dose MXC group compared with all others (P < 0.05, nonsignificant after Bonferroni correction). After Bonferroni correction, the two POA genes significantly affected by perinatal EDC treatment were: Esr1 (P ≤ 0.001) and Kiss1 (P < 0.001) (Fig. 2).

Fig. 2.

Low-density PCR analysis identified 48 transcripts in the POA, and data shown for those transcripts with raw P < 0.05 grouped into 5 classes: A) steroid hormone signaling; B) neurotransmitter signaling; C) cytokine signaline; D) neuropeptide signaling; and E) transcription factors. 17 genes met these criteria, and of those, 15 had expression patterns with levels higher in the EB and/or MXC high-dose group compared with DMSO. For this group, Esr1 gene expression met significance at P < 0.001 (DMSO and MXC low dose were both significantly lower than EB and MXC high dose). Two genes had unique expression patterns. Kiss1 gene expression was significant at P < 0.001, with levels significantly lower in the EB group compared with all others. Gnrhr expression had highest levels in the MXC high-dose group. Raw P values are shown for each gene. Gene expression data for the remaining POA genes (P > 0.05) are presented in Supplemental Table 1.

In the MBH, gene expression of only one transcript (solute carrier 17a6, vesicular glutamate transporter 2) was higher in the MXC high-dose group compared with all others (P < 0.05, nonsignificant after Bonferroni correction). Other MBH genes were not altered by perinatal EDC treatments. Supplemental Tables 1 and 2, published on The Endocrine Society's Journals Online web site at http://mend.endojournals.org, show relative gene expression of those genes that were not affected in either POA (Supplemental Table 1) and all genes in MBH (Supplemental Table 2).

DNA methylation of estrogen receptor (ER)α (Esr1) gene by perinatal EDC

Because the Esr1 gene was significantly up-regulated in the brain of aging females, we performed bisulfite conversion and pyrosequencing of two sites of the Esr1 regulatory regions in exon 1b and intron 1 to determine whether long-term epigenetic programming of the Esr1 gene had occurred. These regions were chosen because DNA methylation of CpG sites in these two gene regions is regulated by maternal behavior or perinatal estrogen when assessed in young adults (28–30). Fourteen CpG sites in exon 1b, and 11 sites in intron 1, were analyzed by pyrosequencing for percentage of methylation. The total percentage of methylation of all the CpG sites within the two regions of analysis was unaffected by perinatal EDC treatment (Fig. 3). The sequences and location of CpG sites of the regions of analysis are also shown in Fig. 3. When individual CpG sites were examined for percentage of methylation, one site in exon 1b (+46 from the transcription start site) and two sites in intron 1 (+1733 and +1786 from the transcription start site) were up-regulated in the perinatally EB-treated rats compared with controls (Fig. 4).

Fig. 3.

Methylation of CpG sites in the Esr1 gene was assayed by bisulfite reaction and pyrosequencing. For exon 1b (A), total methylation across the 14 sites analyzed was not significantly altered by the perinatal EDC exposure. Similarly, total methylation of 11 sites in intron 1 (B) was unaffected by treatment. The regions of analysis are shown in C (exon 1b) and D (intron 1). In C and D, CpG sites are highlighted in gray. The transcription start site in exon 1b is indicated by bold underlined text. Numbers to the left of the sequence (B and C) are nucleotide positions relative to the transcription start site in exon 1b.

Fig. 4.

Methylation of individual CpG sites in exon 1b (A) and intron 1 (B) of Esr1 is shown for the four perinatal treatment groups. In exon 1b, one site located at +46 relative to the transcription start site in exon 1 differed by perinatal treatment, with levels significantly higher in the EB group compared with DMSO. In intron 1, two sites located at +1733 and +1786 were also affected, again with the EB group having significantly higher percentage of methylation compared with DMSO.

Putative transcription factors in exon 1b and intron 1 of Esr1

MatInspector analysis of the exon 1b and intron 1 sequences of the ERα gene revealed numerous putative transcription factor sites on the forward (+) and reverse (−) strands. Table 2 shows the list of transcription factors with more than 80% similarity to the prototypical sequences in the 221-nucleotide sequence assayed for exon 1b. Forty putative transcription factors with homology to the rat prototypical sequences were identified, with similar distribution on the forward (+) and reverse (−) DNA strand. There were seven potential binding sites for six transcription factors (one transcription factor had two sites) in the hypermethylated +46 region.

Table 2.

Transcription factors in the exon 1b sequence with more than 80% homology to the rat prototypical sequences

Family	Description	Position (from)	Position (to)	Anchor	Strand	Matrix similarity
YBXF	Y-box binding transcription factor	−143	−131	−137	(−)	0.889
ETSF	Human/murine ETS factors	−134	−114	−124	(−)	0.855
NF1F	Nuclear factor 1	−130	−110	−120	(+)	0.929
SORY	SOX/SRY-sex/testis-determining factors and related HMG box factors	−128	−104	−116	(−)	0.925
MYOD	Myoblast determining factors	−127	−111	−119	(−)	0.911
SORY	SOX/SRY-sex/testis-determining factors and related HMG box factors	−125	−101	−113	(−)	0.984
HOXF	Paralog hox genes 1–8 from the four hox clusters A, B, C, D	−123	−105	−114	(−)	0.966
YY1F	Activator/repressor binding to transcription initiation site	−121	−101	−111	(+)	0.877
WHNF	Winged helix binding sites	−109	−99	−104	(−)	0.974
AP4R	AP4 and related proteins	−94	−78	−86	(+)	0.851
MYOD	Myoblast determining factors	−94	−78	−86	(−)	0.928
NEUR	NeuroD, Beta2, HLH domain pancreas transcription factor 1	−93	−81	−87	(+)	0.964
CTCF	CTCF and BORIS gene family transcrip-tional regulators	−82	−56	−69	(+)	0.843
EGRF	EGR/NGF induced protein C and related factors	−78	−62	−70	(+)	0.888
SP1F	GC-box factors SP1/GC	−77	−61	−69	(+)	0.910
MAXF	Myc associated zinc fingers	−74	−62	−68	(+)	0.913
KLFS	Krueppel like transcription factors	−74	−58	−66	(+)	0.824
HESF	Vertebrate homologs of enhancer of split complex	−65	−51	−58	(−)	0.870
EBOX	E-box binding factors	−64	−52	−58	(+)	0.912
EBOX	E-box binding factors	−63	−51	−57	(−)	0.928
LTFM	Lactotransferrin motif	−61	−53	−57	(−)	0.984
CP2F	CP2-erythrocyte factor related to Drosophila Elf1	−47	−29	−38	(+)	0.881
NOLF	Neuron-specific olfactory factor	−42	−20	−31	(+)	0.888
PRDM	PRD1-BF1 and RIZ homologous (PR) domain proteins (PRDM)	−42	−13	−27	(−)	0.802
NFKB	Nuclear factor κ B/c-rel	−36	−22	−29	(−)	0.824
FAST	FAST-1 SMAD interacting proteins	−23	−7	−15	(−)	0.851
FKHD	Forkhead domain factors	−15	2	−7	(+)	0.926
EGRF	EGR/NGF induced protein C and related factors	−3	14	6	(−)	0.808
SP1F	GC-box factors SP1/GC	19	35	27	(−)	0.879
CDEF	Cell cycle regulators: Cell cycle dependent element	24	36	30	(−)	0.949
E2FF	E2F-myc activator/cell cycle regulator	25	41	33	(+)	0.864
TF2B	RNA polymerase II transcription factor IIB	28	34	31	(+)	1.000
HOMF	Homeodomain transcription factors	31	49	40	(+)	0.899
ZFO3	C2H2 zinc finger transcription factors 3	34	46	40	(−)	0.916
HEAT	Heat shock factors	35	59	47	(−)	0.953
HEAT	Heat shock factors	36	60	48	(+)	0.886
INRE	Core promotor initiator elements	39	49	44	(+)	0.945
SORY	SOX/SRY-sex/testis-determining factors and related HMG box factors	39	63	51	(−)	0.872
HOXF	Paralog hox genes 1–8 from the four hox clusters A, B, C, D	44	62	53	(−)	0.942
TEAF	TEA/ATTS DNA binding domain factors	49	61	55	(+)	0.967

Positions are shown relative to the transcription start site in exon 1b. (+), Forward; (−), reverse.

In the 174 nucleotide sequence of intron 1 (Table 3), 33 transcription factors were found that met the 80% homology cut-off. A cluster of 13 overlapping transcription factors distributed across the (+) and (−) DNA strand was found in the 20-nucleotide region from 1827 to 1846 nucleotides downstream from the transcription start site in exon 1b, a region characterized by four adjacent repeating CpG sites (see Fig. 3). Within that region, five putative ZF5 POZ domain zinc finger, four E2F-myc activator/cell cycle regulator, two nuclear respiratory factor 1 (NRF1), one cell cycle-dependent element, and one signal transducer and activator of transcription (STAT) sites were identified. The intron 1 sequence also contained a total of five STAT sites in the region of analysis.

Table 3.

Transcription factors in the intron 1 sequence with more than 80% homology to the rat prototypical sequences

Family	Description	Position (from)	Position (to)	Anchor	Strand	Matrix similarity
E2FF	E2F-myc activator/cell cycle regulator	1746	1762	1754	(−)	0.815
PLAG	Pleomorphic adenoma gene	1747	1769	1758	(−)	1.000
KLFS	Krueppel like transcription factors	1757	1773	1765	(−)	0.928
INSM	Insulinoma associated factors	1759	1771	1765	(−)	0.917
STAT	Signal transducer and activator of transcription	1784	1802	1793	(−)	0.952
HNF1	Hepatic nuclear factor 1	1792	1808	1800	(+)	0.852
NFAT	Nuclear factor of activated T-cells	1792	1810	1801	(+)	0.872
MYT1	MYT1 C2HC zinc finger protein	1795	1807	1801	(+)	0.967
MYT1	MYT1 C2HC zinc finger protein	1798	1810	1804	(−)	0.817
STAT	Signal transducer and activator of transcription	1800	1818	1809	(−)	0.910
STAT	Signal transducer and activator of transcription	1802	1820	1811	(+)	0.961
EVI1	EVI1-myleoid transforming protein	1808	1824	1816	(+)	0.821
GATA	GATA binding factors	1811	1823	1817	(+)	0.982
OVOL	OVO homolog-like transcription factors	1812	1826	1819	(−)	0.831
ZF5F	ZF5 POZ domain zinc finger	1827	1841	1834	(+)	0.966
E2FF	E2F-myc activator/cell cycle regulator	1827	1843	1835	(−)	0.897
NRF1	Nuclear respiratory factor 1	1827	1843	1835	(−)	0.830
ZF5F	ZF5 POZ domain zinc finger	1828	1845	1835	(−)	0.952
E2FF	E2F-myc activator/cell cycle regulator	1828	1845	1836	(+)	0.894
ZF5F	ZF5 POZ domain zinc finger	1829	1844	1836	(+)	1.000
NRF1	Nuclear respiratory factor 1	1829	1845	1837	(−)	0.802
E2FF	E2F-myc activator/cell cycle regulator	1829	1845	1837	(−)	0.946
ZF5F	ZF5 POZ domain zinc finger	1830	1844	1837	(−)	0.919
E2FF	E2F-myc activator/cell cycle regulator	1830	1846	1838	(+)	0.908
ZF5F	ZF5 POZ domain zinc finger	1831	1845	1838	(+)	0.967
CDEF	Cell cycle dependent element	1834	1846	1840	(+)	0.906
STAT	Signal transducer and activator of Transcription	1844	1862	1853	(−)	0.934
VTBP	Vertebrate TATA binding protein factor	1851	1867	1859	(+)	0.850
NKXH	NKX homeodomain factors	1851	1869	1860	(+)	0.964
RUSH	SWI/SNF related nucleophosphoproteins with a RING finger DNA binding motif	1854	1864	1859	(−)	0.993
STAT	Signal transducer and activator of transcription	1854	1872	1863	(−)	0.980
LTFM	Lactotransferrin motif	1855	1863	1859	(−)	0.910
KLFS	Krueppel like transcription factors	1869	1885	1877	(−)	0.925

Positions are shown relative to the transcription start site in exon 1b. (+), Forward; (−), reverse.

Body weight is affected by perinatal EDC

Rats were weighed on the day of euthanasia. As shown in Fig. 5, the EB group had a significantly higher body weight than rats in all of the other treatment groups (P < 0.005). Post hoc analysis (Fisher's least significant difference) showed that the EB group was significantly heavier than all other groups (P < 0.01 vs. all).

Fig. 5.

Body weight (g) is shown for rats on the day of euthanasia at 16–17 months of age. The EB group had significantly heavier body weights than all other groups. *, P < 0.01 vs. all other treatments.

Discussion

The current data show that early life exposures to environmental endocrine disruptors hasten premature reproductive aging, thereby diminishing an individual's lifetime reproductive capacity. Whether this is the case in women remains to be proven, although epidemiological data suggest that certain EDCs are associated with earlier age at menopause. Women who had been exposed to high levels of dioxin due to a chemical plant explosion in Seveso, Italy, showed increased risk for early menopause (31). Body burden of DDE (dichlorodiphenyldichloroethylene), a metabolite of the organochlorine pesticide DDT (dichlorodiphenyltrichloroethane), was associated with earlier age at menopause in a population of women studied in North Carolina (32). A cross-sectional study conducted on Hispanic women relating their body burdens of organochlorine pesticides to age at menopause showed that menopause occurred at earlier ages in women exposed to DDT, DDE, β-hexachlorocyclohexane, and trans-nonachlor (33). These results are consistent with the current study's results.

By contrast, other human studies show no relationship between serum EDC and age at menopause (polychlorinated biphenyl, Refs. 32, 34), and one study indicated that on average, menopause occurred 3 months later in women handling pesticides than in controls (35). To our knowledge, links between MXC and menopause have not been studied in humans.

There are plausible explanations for these different outcomes across the human data, not the least of which are differences in the nature of the EDC studied. Humans (and wildlife) are exposed to complex mixtures of environmental contaminants, and exposures occur throughout the life cycles. Although epidemiological studies are valuable, the ability to draw causal connections between exposure and outcome is not possible. Notably, authors on both sides of this literature make the point that there is not a linear dose-response relationship between body burden and biological outcome. Thus, differences among studies may also result from making inferences across different exposure levels that may not be comparable from one population to another. Last, it is becoming more widely accepted that exposures to EDCs during critical developmental windows may have latent effects (14). However, to draw this kind of connection in humans, it is necessary to relate secular trends in the timing of menopause to exposures received 40–50 yr ago. Thus, animal studies are necessary to understand causal relationships and to discern the underlying mechanisms.

Developmental EDC exposures hasten reproductive senescence

Our current results and others show that the reproductive life cycle, which includes puberty, adult reproductive function, and aging, is significantly altered by perinatal EDC exposures. Estrogenic EDCs, such as MXC, cause advancements in the timing of puberty in females (16; reviewed in Ref. 36). This earlier timing of puberty, however, is accompanied by a diminution of fertility in perinatally treated animals when those rats were examined at d 50–60 (16). Together with results showing earlier reproductive senescence, these results as a whole indicate that the reproductive life cycle is substantially shorter in EDC-exposed animals. This finding is consistent with results for bisphenol A (17), dioxins (18), and others (reviewed in Ref. 35) but not heretofore studied from the hypothalamic molecular approach.

The control of natural reproductive aging, and disruption of these processes by EDC resulting in premature reproductive senescence, involves coordinated physiological processes regulated by the hypothalamic-pituitary-gonadal (HPG) axis. At least some of the reproductive aging process caused by MXC is due to actions on the ovary. Perinatal exposure disrupted folliculogenesis, leading to an increased number of preantral and early antral follicles that failed to achieve maturation and ovulation, and fewer/no corpora lutea were detectable (16). This was accompanied by cellular/molecular changes to the ovary. Other EDCs have been shown to have ovarian follicular effects, including polychlorinated biphenyls (37) and bisphenol A (17, 38). However, these animals still have a follicular pool, making it likely that the hypothalamus is not only a target of perinatal MXC but that its reprogramming hastens reproductive aging.

Developmental reprogramming of the hypothalamus

Hypothalamic function is vulnerable to perinatal EDC actions (reviewed in Ref. 39), and our current data add novel information about the molecular mechanisms by which this occurs. We show lifelong changes in hypothalamic gene expression associated with premature reproductive aging, in a region-specific manner. In the POA of rats, a complex regulatory neural/glial network controlling HPG reproductive function was up-regulated by 50% or more by high-dose MXC or EB treatments, relative to control. The three major steroid hormone receptors (ERα, androgen receptor, and progesterone receptor), glutamate/GABA receptors, growth factors (TGFα and TGFβ), the Kiss1 receptor, and the transcription factor Stat5b showed this pattern of expression. In addition, a key neuropeptide regulating GnRH release, kisspeptin (Kiss1 gene), was down-regulated in the EB group, and the GnRH receptor was up-regulated in the MXC high-dose group. The sum of these changes would be integrated at the level of GnRH output from the hypothalamus to pituitary and subsequently to the gonad. Although GnRH gene expression itself was not altered, this is not surprising in light of evidence that it is probably GnRH release and not gene expression that is most highly regulated, including during reproductive aging (reviewed in Ref. 4).

The two genes significantly affected in the POA by perinatal EDC exposure were Kiss1 (down-regulated in the EB group) and Esr1 (ERα; up-regulated in the EB and MXC high-dose groups). The importance of these two genes on female reproduction, both independently as well as in combination, cannot be understated. Kiss1 neurons in the POA play a crucial role in the capacity for rodents and other species to undergo ovulatory processes (40; reviewed in Ref. 41). This role extends to reproductive aging, which is accompanied by changes in the Kiss1 population as shown in rats, monkeys, and humans, with a concurrent decline in ovulatory capacity (9, 10, 42). Furthermore, the Kiss1 system is vulnerable to EDC exposures (43–45). Together with the current work, the hypothalamic Kiss1 system is an important target for future mechanistic studies on how developmental EDC exposures alter hypothalamic function.

The Esr1 gene was significantly up-regulated by EB and high-dose MXC, a finding consistent with previous work showing that estrogenic EDCs can affect hypothalamic Esr1 gene or ERα protein expression (45; reviewed in Ref. 36). The same hypothalamic cells may coexpress Kiss 1 and ERα (46). Recent work has shown that positive feedback regulation of the HPG axis requires coexpression of ERα within Kiss1 neurons (40, 47). These estrogen-sensitive kisspeptinergic neurons are a logical target for further research on endocrine disruptors and are potentially vulnerable to long-term regulation by perinatal exposure.

Lifelong epigenetic changes in Esr1 induced by perinatal EDC

Based on our observation of long-lasting effects of MXC and EB on Esr1 gene expression, together with current literature suggesting that expression of Esr1 is epigenetically regulated in the hypothalamus (28, 30, 48, 49), we found that methylation at three sites (+46 site of exon 1b, +1733 and +1786 sites of intron 1) was increased in the EB group. The MXC rats were not significantly different from control in their DNA methylation pattern, possibly because their reproductive function is still maintained (albeit diminished) with aging. In the EB rats, by contrast, the high dosage given perinatally had profound consequences on reproductive aging, gene expression, and DNA methylation pattern. EB rats also showed higher body weight at death, suggesting that adipogenesis and/or the hypothalamic control of energy balance was dysregulated by the perinatal treatment.

It was surprising to us that both Esr1 mRNA levels and CpG methylation of the Esr1 promoter were increased in the EB group, because the relationship between gene expression and CpG methylation is predicted to be inverse. This has been shown in the brain for Esr1 and for other genes, such as Bdnf (28, 30, 48, 49). Several factors may explain the lack of such a relationship for the EB-treated rats in the current study. First, there are numerous other CpG sites on the Esr1 gene whose methylation pattern was not measured herein but which may play different regulatory roles in Esr1 gene expression. Second, other epigenetic and nonepigenetic regulatory factors (e.g. coregulatory factors and transcription factors) may come into play in the final measure of Esr1 mRNA levels. Third, histone modifications, not measured in this study, can contribute substantially to the epigenetic regulation of gene expression. Fourth, the POA is a highly heterogeneous tissue, with different cell types that may be differentially affected at the gene and protein level by perinatal endocrine disruptors. Fifth, the timing of exposure and the long lag until gene expression measurements may come into play. Indeed, DNA methylation is not fixed throughout the life cycle but may undergo dynamic change postnatally. This has even been shown for Esr1, whose DNA methylation profile undergoes shifts from postnatal d 1 to 20 to 60 in rats (29). Similarly, this concept applies to developmental exposures to EDCs affecting DNA methylation with profiles that undergo dynamic change well into aging in the prostate gland (50). In our rats, this concept could be tested by comparing hypothalamic Esr1 methylation patterns of neonatally endocrine-disrupted rats throughout postnatal life into aging, an area that we are currently investigating.

We scanned the Esr1 sequence for the presence of putative transcription factors to guide future analyses of gene regulatory regions that may be affected by developmental estrogenic EDC exposures. In exon 1b, the hypermethylated CpG at +46 is within the binding sites for several transcription factors, such as SORY, HEAT, and ZFO3, that were shown to play roles as transcriptional repressors (51–55), which may relate to the lack of expected inverse correlation between the methylation and gene expression. In intron 1, there was a cluster of transcription factors (ZF5 POZ domain zinc finger, E2F-myc activator/cell cycle regulator, NRF1, cell cycle-dependent element, and STAT) overlying the GC-rich region from +1827 to 1846. The ZF5 transcription factor is best studied for its role as a transcriptional repressor (56). E2F is better associated with transcriptional activation and, more recently, with histone H3K56 acetylation (57). NRF1 is well known for its role in mitochondrial function, but recently has been associated with the mediation of estrogen effects on ERα gene transcription (58), something that may be relevant to the current finding for specific effects of EB on DNA methylation at that site in intron 1. Finally, the intron 1 sequence also contained five STAT sites in our region of analysis, a finding that highlights the potential importance of this transcription factor in Esr1 regulation.

Conclusions

The current findings extend the ever-growing body of research collectively termed the fetal/developmental basis of adult disease to the end of the life spectrum. Our results show that not only do perinatal EDC reprogram hypothalamic gene expression but that they cause long-term changes in epigenetic properties of the Esr1 gene and possibly others not studied herein. These findings have substantial implications for humans. By hastening senescence, EDC may eliminate the possibility of biological children for women who may postpone childbirth for personal or professional reasons. Considering the important roles of estrogens on targets in body and brain, early reproductive senescence may accelerate some disease-related states associated with menopause and affect quality of life in the aging population of women.

Materials and Methods

Animals and husbandry

Fischer (CDF) inbred rats were obtained from Charles River Laboratories (Wilmington, MA) to generate timed-pregnant females. Animals in this study were siblings of those described in Ref. 16 and had identical husbandry in the labs at Rutgers University. Rats were fed a soy-free scientific diet 5V01 rat chow (Lab Diet manufactured by PMI Nutrition International LLC, Brentwood, MO) and tap water ad libitum. They were allowed to develop to 16–17 months of age. During this time, they were subjected to periodic monitoring of estrous cycles by daily vaginal smears (see Ref. 16 for details). The cycles were categorized as regular (4–5 d), prolonged (6+ d), persistent estrus, or persistent diestrus (59). All procedures in the present study were carried out in accordance with the guidelines of the Rutgers University Animal Care and Facilities Committee.

Treatments

Experimental subjects were exposed to one of four treatments (below) for a total of 12 consecutive days. Daily injections began with the pregnant dam (ip) on gestational d 19 through parturition on d 22 (four prenatal days). Then, injections continued with the pups (sc) from postnatal d 0 (birth) to d 7 (eight postnatal days). Treatments were: 1) vehicle control [DMSO sesame oil at 1:2 (three litters, n = 5 individuals)]; 2) EB (1 mg/kg · d) as a positive estrogenic control (three litters, n = 5 individuals); 3) MXC (Sigma, St. Louis, MO), 20 μg/kg · d (low-dose MXC; five litters, n = 9 individuals); and 4) 100 mg/kg · d (high-dose MXC; five litters, n = 9 individuals) in 1 ml of vehicle per kilogram of body weight. The choice of doses of MXC and EB was based on the companion study, in which all animals were treated to cause ovarian disruption (16). The 20 μg/kg · d was selected as an environmentally relevant, low dose of MXC (60, 61). The 100 mg/kg · d MXC is considered a midrange dose, chosen for comparison with published studies (23, 62, 63). The EB dose was selected to be high so as to ensure disruption of both ovary and hypothalamus by an estrogenic compound.

Euthanasia and tissue/blood collection

At 16–17 months of age, female rats were euthanized by decapitation between 1400 and 1600 h. Rats with regular estrous cycles were euthanized on the day of proestrus; on this light cycle, rats are in the preovulatory period of the cycle, when hypothalamic activity is increasing and serum LH and estradiol concentrations are rising. Noncycling rats were euthanized on persistent diestrus or persistent estrus. Trunk blood was collected and serum separated by centrifugation and stored at −80 C. Brains were rapidly removed, and the POA and the MBH were blocked as described previously (64), snap frozen in liquid nitrogen, and stored at −80 C.

RNA extraction and real-time PCR

After shipment to the University of Texas, RNA was extracted and purified, treated with deoxyribonuclease (TURBO DNA kit; Applied Biosystems, Inc., Foster City, CA), and purity and integrity were assessed on the bioanalyzer 2100 (Agilent, Cedar Creek, TX). cDNA conversion was carried out on 2 μg of cytoplasmic RNA (High-capacity cDNA reverse transcription kit; Applied Biosystems, Inc.), aliquoted, and processed for real-time PCR as published (45, 65). A 48-gene TaqMan PCR-based array (Applied Biosystems, Inc.) was designed and used for real-time PCR reactions, which were run on the ABI 7900 real-time PCR machine using parameters described elsewhere (45, 65, 66). Data were analyzed using the normalized Ct (δ-Ct) for each sample before transformation to fold change.

Hormone RIA

Serum 17β-estradiol and progesterone were measured in duplicate 100 μl of serum samples using commercially available RIA kits, according to the manufacturer's instructions (Coat-A-Count RIA kits; Siemens Medical Solutions, Malvern, PA). The assay sensitivity, intra- and interassay coefficients of variance were 8 pg/ml, 5.5 and 6.4% (respectively) for estradiol, and 0.2 ng/ml, 4.6 and 5.9% for progesterone.

DNA methylation analysis of the ERα promoter

During the extraction of RNA described above, the genomic DNA fraction was isolated from the nuclear fraction of the POA and MBH samples used for cytoplasmic RNA isolation. Briefly, nuclear pellets were resuspended in high salt buffer and treated with proteinase K followed by a short (10 min) ribonuclease A treatment. Samples were then extracted with phenol/chloroform and precipitated in 100% isopropanol at −20 C overnight. Pelleted DNA was resuspended in nuclease-free H₂O and aliquoted at 20 ng/μl. A 50-μl aliquot for each DNA sample was sent to EpigenDX (Worcester, MA) for bisulfite reaction and pyrosequencing of the ERα regulatory regions in exon 1b and intron 1. Sequences of these regions are shown in Fig. 3.

Transcription factor analysis

Using MatInspector software (Genomatix, Inc., Munich, Germany), the intron 1 and exon 1b sequences used by EpigenDX for DNA methylation analysis were scanned to determine which putative rat transcription factor binding sites were localized in these regions.

Analysis and statistics

Experimental subjects derived from three to five litters per treatment and a total of five to nine rats were used per group. No significant effects were found when birth litter was used as a covariate. Furthermore, intra- and interlitter variability were comparable. Therefore, individual rats were used as the unit of analysis for statistical purposes. ANOVA was used to compare differences among treatment groups. Post hoc analysis was performed (Fisher's projected least significant difference) when a main treatment effect at P < 0.05 was identified. When appropriate, Bonferroni corrections were made for multiple analyses and data reported using Tukey post hoc tests. In a few cases, data did not meet assumptions of homogeneity of variance based on Levene's test and were instead analyzed by the nonparametric Kruskal-Wallis test. For estrous cyclicity, Fisher's exact test was used to compare differences between treated animals with controls.