Resveratrol ameliorates prenatal progestin exposure-induced autism-like behavior through ERβ activation

17β-estradiol (E2, #E2758); progesterone (P4, #P0130); levonorgestrel (LNG, #1362602); medroxyprogesterone acetate (MPA, #1378001); nestorone (NES, # SML0550); norethindrone (NET, #1469005); norethindrone acetate (NETA, #1470004); norgestimate (NGM, # 1471914); hydroxyprogesterone caproate (OHPC, #1329006), and resveratrol (RSV, #R5010) were obtained from Sigma. Norethynodrel (NEN, #E4600–000) was obtained from Steraloids.

The animal protocol conformed to the US NIH guidelines (Guide for the Care and Use of Laboratory Animals, No. 85–23, revised 1996) and was reviewed and approved by the Institutional Animal Care and Use Committee from Wuhan University [9].

The animal behavior test of offspring was carried out at 10 weeks of age. Female offspring were tested in the diestrus phase, which was confirmed by vaginal smears. Autism-like behavior was evaluated using the marble burying test (MBT) and social interaction (SI) test.

The amygdala neurons were isolated for in vitro primary cell culture analysis [9, 35]. The DNA methylation on the rat ERβ promoter was evaluated by a real-time PCR-based methylation-specific PCR (MSP) analysis as described previously with some modifications [9, 36‐38]. The lentivirus for rat ERβ shRNA and empty control (EMP) were prepared previously in our lab [9]. The mRNA was measured by real-time quantitative PCR using primers provided in Additional file 1: Table S1, and the protein was measured by western blotting [9]. The SIRT1 activity assay was evaluated in nuclear extract using a SIRT1 Fluorometric Drug Discovery Kit (Cat #: BML-AK555, Enzo Life Sciences) [39]. Oxidative stress was evaluated by in vivo superoxide anion (O₂⁻) release [40] and 3-nitrotyrosine formation. DNA damage was evaluated by 8-OHdG and γH2AX formation [9]. Mitochondrial function was evaluated by mitochondrial DNA copies [9] and intracellular ATP level [40]. The histone methylation was evaluated by chromatin immunoprecipitation (ChIP) analysis [9]. Fatty acid metabolism was evaluated by in vitro lipid transport assay [41] and fatty acid oxidation assay [42, 43]. Statistical analysis was conducted using SPSS 22 software, and a P value of < 0.05 was considered significant, and the data was given as mean ± SEM [9].

We first investigated the potential effect of prenatal progestin exposure on ERβ expression in the amygdala and autism-like behavior in 10-week-old male offspring, and then evaluated whether postnatal resveratrol treatment could reverse this effect. In Table 1 (left panel of first row), several steroids, 17β-estradiol (E2), progesterone (P4), and eight different kinds of clinically relevant progestins, including levonorgestrel (LNG), medroxyprogesterone acetate (MPA), nestorone (NES), norethindrone (NET), norethindrone acetate (NETA), norethynodrel (NEN), norgestimate (NGM), and hydroxyprogesterone caproate (OHPC), were used for prenatal exposure to different 3-month-old pregnant dams for 21 days (n = 8). The 5-week-old male offspring were then treated by either control (CTL) or resveratrol (RSV) for 4 weeks, and the ERβ mRNA in the amygdala was evaluated by real-time PCR and the autism-like behavior was evaluated by social interaction time. Our results showed that E2 had no effect, while P4 slightly decreased ERβ expression but had no significant effect on social interaction time, and almost all the clinically relevant progestins significantly decreased ERβ expression (except the NGM), and decreased social interaction time (except the NEN) compared to the vehicle group. Furthermore, resveratrol treatment almost completely reversed the suppression effect on ERβ expression and social interaction time (partly for the NES and NGM). Our results indicate that postnatal resveratrol treatment reverses prenatal progestin exposure-induced ERβ suppression and autism-like behavior.

We then evaluated the potential effect of prenatal resveratrol treatment on prenatal progestin exposure-induced ERβ suppression and autism-like behavior. In Table 1 (right panel of first row), prenatal exposure of P4 and all the eight different kinds of clinically relevant progestins significantly decreased ERβ expression and social interaction time (except the NEN), while prenatal resveratrol treatment (oral administration of 20 mg/kg RSV for 21 days) almost completely reversed prenatal progestin exposure-induced suppression effect on ERβ expression (partly for NEN) and social interaction time. Our results indicate that prenatal resveratrol treatment prevents prenatal progestin exposure-induced ERβ suppression and autism-like behavior.

In Table 1, the prenatal exposure of norethindrone (NET) in pregnant dams showed the most dramatic suppression effect on ERβ expression and social interaction time in offspring compared to other progestins. The NET was chosen for the prenatal progestin exposure in the later experiments to investigate the detailed mechanism and effect of resveratrol on autism-like behavior. The animal was treated as shown in Fig. 1b. We first measured the gene expression of ERβ, SOD2, and ERRα in the hypothalamus and hippocampus (see Additional file 1: Figure S1). We found that prenatal NET exposure showed no effect on gene expression. Also, RSV treatment showed no effect on the expression of ERβ and SIRT1, while the expression of SOD2 and ERRα was significantly increased in both the hypothalamus and hippocampus by RSV treatment compared to the control (CTL) group. We then measured the mRNA expression of ERβ (see Fig. 2a), SOD2 (see Fig. 2b), and ERRα (see Fig. 2c) in the amygdala. We found that RSV treatment completely reversed prenatal NET exposure-induced suppression of ERβ, SOD2, and ERRα in both male and female offspring. Furthermore, in female offspring, RSV significantly increased expression of SOD2 and ERRα in both the VEH and NET groups, while in male offspring there was much less gene activation (see Fig. 2b, c), indicating that female offspring seem more responsive to RSV treatment compared to male offspring. We also measured the protein levels in the amygdala (see Fig. 2d–g). The results showed that RSV completely reversed the NET-mediated suppression effect on the expression of ERβ, SOD2, and ERRα. Furthermore, RSV increased ERRα expression in VEH treatment compared to the CTL group in male offspring (see Fig. 2d, f), and in female offspring, RSV significantly increased ERRα expression in both VEH and NET treatment compared to CTL group (see Fig. 2e, g). We also measured the SIRT1 mRNA expression (see Figure S2a) and SIRT1 activity (see Figure S2b), which showed no difference in any of the treatments, indicating that SIRT1 may not be involved in RSV-mediated effect (see detailed statistical information in Additional file 1: Data S1). Our results suggest that RSV reverses prenatal NET exposure-induced suppression of ERβ and its target genes, and female offspring seem more responsive than male offspring.

In Fig. 3a, b, the DNA methylation on the ERβ promoter was significantly increased by prenatal NET exposure, and RSV treatment completely diminished this effect compared to the CTL group. We then measured the histone methylation on the ERβ promoter by ChIP techniques. In male offspring (see Fig. 3c), prenatal NET exposure increased H3K9 di-methylation (H3K9me2) by 1.80-fold and H3K27 tri-methylation (H3K27me3) by 2.13-fold, while in female offspring (see Fig. 3d), it increased H3K9 di-methylation (H3K9me2) by 2.14-fold and H3K27 tri-methylation (H3K27me3) by 1.56-fold, but had no effect on H3K9 tri-methylation (H3K9me3). RSV treatment completely diminished the effect on H3K9me2, but partly diminished the effect on H3K27me3 in male offspring compared to the CTL group, while in female offspring, RSV partly diminished the effect on H3K9 di-methylation, but completely diminished the effect on H3K27 tri-methylation (see detailed statistical information in Additional file 1: Data S2). Our results indicate that postnatal RSV administration diminishes prenatal levonorgestrel exposure-induced methylation of DNA and histone on the ERβ promoter in the amygdala.

We evaluated the effect of RSV treatment on prenatal NET exposure-induced ERβ suppression and the subsequent molecular consequences in the amygdala, including ROS generation, DNA damage, mitochondrial function and lipid metabolism. We found that prenatal NET exposure significantly increased superoxide anion (O₂⁻) release by 2.27-fold (in male) and 1.88-fold (in female) respectively (see Fig. 4a); it increased 3-nitrotyrosine formation by 2.03-fold (in male) and 1.54-fold (in female) respectively (see Fig. 4b); it increased 8-OHdG formation by 2.65-fold (in male) and 2.27-fold (in female) respectively (see Fig. 4c); it increased γH2AX formation by 1.94-fold (in male) and 1.48-fold (in female) respectively (see Fig. 4d, e); it decreased mitochondrial DNA copies by 38% (in male) and 24% (in female) respectively (see Fig. 4f); it decreased intracellular ATP level by 45% (in male) and 25% (in female) respectively (see Fig. 4g). Furthermore, it decreased in vivo fatty acid oxidation by 47% (in male) and 18% (in female) respectively (see Fig. 4h), and it decreased in vitro fatty acid uptake by 42% (in male) and 26% (in female) respectively (see Fig. 4i). RSV treatment partly reversed this effect in male offspring, while in female offspring, RSV completely diminished this effect. Furthermore, in VEH group, RSV treatment significantly increased mitochondrial function (see Fig. 4f, g) and in vivo fatty acid oxidation metabolism (see Fig. 4h) in both male and female offspring. On the other hand, RSV treatment increased in vitro fatty acid uptake (see Fig. 4i) in female offspring but not in male offspring from VEH group (see detailed statistical information in Additional file 1: Data S3). Our results indicate that RSV ameliorates prenatal NET exposure-induced oxidative stress and dysfunction of mitochondria and lipid metabolism, and that female offspring were less responsive to NET exposure than male offspring.

We measured the effect of RSV on prenatal NET exposure-induced autism-like behavior. In male offspring, prenatal NET exposure decreased buried marbles by 52% (see Fig. 5a), while there was no difference in female offspring. We then evaluated the social interaction time. In male offspring, the NET treatment decreased by 32% in sniffing, 48% in mounting, no difference in grooming partner, and 33% in total social interaction (see Fig. 5b), while RSV treatment completely diminished this effect. In female offspring, the NET treatment decreased by 20% in sniffing, no difference in mounting or grooming partner, and 22% in total social interaction (see Fig. 5c). Furthermore, the RSV treatment completely diminished this effect in both male and female offspring (see detailed statistical information in Additional file 1: Data S4). Our results indicate that RSV completely reverses prenatal NET exposure-induced autism-like behavior and that male offspring seem more sensitive to NET-induced effect compared to female offspring.

Prenatal NET exposure treated male offspring received lentivirus infusion of shERβ shown in Fig. 1b. We first measured the gene expression of ERβ, SOD2, and ERRα. The results showed that prenatal NET exposure significantly decreased mRNA levels (see Fig. 6a) and protein levels (see Fig. 6b, c) of those genes. We then measured prenatal NET exposure-mediated molecular consequences in the amygdala. The results showed that postnatal resveratrol treatment ameliorates prenatal NET exposure-induced oxidative stress (see Additional file 1: Figure S3a, b), DNA damage (see Additional file 1: Figure S3c, e), dysfunction of mitochondria (see Additional file 1: Figure S3f, g), and lipid metabolism (see Additional file 1: Figure S3 h, i) through ERβ activation. RSV treatment completely reversed this suppression effect, while ERβ knockdown (shERβ) significantly diminished the effect of RSV (see Additional file 1: Figure S3). Next, we investigated the potential effect of RSV treatment and ERβ expression on prenatal NET exposure-induced autism-like behavior in male offspring. We found that prenatal NET exposure (NET/EMP/CTL) decreased buried marbles by 32% (see Fig. 6d) compared to the control group. RSV treatment completely reversed prenatal NET exposure-induced autism-like behavior, while ERβ knockdown (shERβ) completely diminished the effect of RSV. We also measured the social interaction time (see Fig. 6e). The results showed that prenatal NET exposure resulted in a 38% decrease in sniffing and a 37% decrease in total social interaction, while it showed no difference in mounting and grooming partner. RSV treatment partly diminished this effect, while ERβ knockdown (shERβ) completely diminished the effect of RSV (see detailed statistical information in Additional file 1: Data S5). Our results indicate that RSV ameliorates prenatal norethindrone exposure-induced autism-like behavior in offspring through ERβ activation.

Three-month-old pregnant dams were exposed to either NET (20 μg norethindrone) or VEH (vehicle) for ~ 21 days during the whole pregnancy, and they also received either control (PreCTL) or resveratrol (PreRSV) prenatal treatment by oral administration. Both male and female offspring at 10 weeks of age were used for biomedical analysis and autism-like behavior testing. We first measured the ERβ gene expression (see Fig. 7a) and found that prenatal RSV treatment completely reversed prenatal NET exposure-induced ERβ suppression. We also measured the epigenetic changes with histone methylation on the ERβ promoter in both male (see Additional file 1: Figure S4a) and female offspring (see Additional file 1: Figure S4b). We found that prenatal RSV treatment completely normalized prenatal NET exposure-induced histone methylation, including H3K9me2 and H3K27me3. We also measured the oxidative stress, including superoxide anion (see Additional file 1: Figure S4c) and 8-OHdG formation (see Additional file 1: Figure S4d); mitochondria functions, including mitochondrial DNA copies (see Additional file 1: Figure S4e) and intracellular ATP level (see Additional file 1: Figure S4f); as well as lipid metabolism, including palmitate oxidation (see Additional file 1: Figure S4 g) and fatty acid uptake (see Additional file 1: Figure S4 h). It showed that prenatal RSV treatment completely (in female offspring) or partly (in male offspring) reversed NET-induced effect compared to control group. Finally, we measured autism-like behavior, including buried marble testing (see Fig. 7b) and social interaction time testing for both male (see Fig. 7c) and female offspring (see Fig. 7d). We found that prenatal RSV treatment completely reversed prenatal NET exposure-induced autism-like behavior in both male and female offspring (see detailed statistical information in Additional file 1: Data S6). Our results indicate that prenatal RSV treatment prevents prenatal NET exposure-induced autism-like behavior.

Natural progesterone or synthetic progestins, such as hydroxyprogesterone caproate (OHPC), have been used to reduce the incidence of preterm birth [44], where there is a signal for embryo-fetal toxicity associated with OHPC in the two largest clinical trials [45]. Progestin is the major component in hormone replacement therapy (HRT) and combined oral contraceptive (COC), and taking oral contraceptives when pregnancy begins is a strong risk factor for ASD development [8]. Also, progestin concentrations found in water can suppress ERβ expression and affect brain functions [21, 46]. Our current data show that prenatal progestin exposure suppresses ERβ expression in the amygdala [47] and subsequently contributes to ASD development [9]. This consideration may potentially apply to humans, as prenatal progestin exposure may be associated with ASD development via directly or indirectly taking progestin compounds, which includes contraceptive pills, progesterone/progestin pills to prevent threatened abortion, and COC-contaminated drinking water and seafood during pregnancy [21].

It has been reported that RSV inhibits DNA methyltransferases (DNMTs) [48] and subsequently decreases DNA methylation [49] and histone methylation [50, 51]. We show that RSV in vivo treatment significantly decreases both DNA methylation and histone methylation on the ERβ promoter, and subsequently increases ERβ expression. This is the first time we report the potential neuroprotective mechanism of RSV through RSV-mediated demethylation with subsequent gene activation. Furthermore, we show that RSV treatment can only demethylate prenatal NET exposure-mediated hypermethylation on both DNA and histone, while it has little effect on basal methylation in the prenatal VEH exposure group. This suggests that RSV-mediated demethylation is not only due to RSV-mediated DNMT inhibition, but rather that some other indirect factors may also be involved [52].

We have previously reported that ERβ expression is regulated by SIRT1 through complexes of SIRT1-PPARγ/RXR-p300 on the ERβ promoter [14], and RSV has been reported to be able to activate SIRT1 either directly [53] or indirectly through AMPK activation [54]. We suppose that resveratrol treatment should be able to activate SIRT1 and subsequently upregulate ERβ expression, while our results show that RSV treatment significantly increases ERβ expression, but neither SIRT1 expression nor activity changes in the amygdala, suggesting that a SIRT1-independent ERβ activating mechanism is required for RSV-mediated ERβ activation. Our further investigation shows that RSV treatment in vivo significantly alters DNA methylation and epigenetic changes with demethylation on the ERβ promoter in the amygdala, which subsequently activates ERβ expression with ameliorated autism-like behavior. This is the first time we report the potential mechanism for RSV-mediated gene activation through demethylation of either DNA or histone in the nervous system, which sheds a light on rescuing ASD symptoms through resveratrol treatment.

Our results show that resveratrol (RSV) treatment reverses prenatal NET exposure-induced methylation of DNA and histone on the ERβ promoter and subsequently reverses prenatal NET exposure-induced ERβ suppression, but it does not change the basal DNA and histone methylation level and subsequent basal ERβ expression. On the other hand, the basal expression level of SOD2 and ERRα increases significantly in response to RSV treatment in not only the amygdala, but also in the hypothalamus and hippocampus; this kind of gene activation is not regulated by ERβ as reported previously [11, 12] since the basal ERβ expression has no change in response to RSV treatment. Also, the in vitro cell culture experiments show that RSV treatment significantly increases the expression of SOD2 and ERRα, and female offspring seem more responsive to RSV treatment compared to male offspring. This can be explained because the female offspring have higher basal ERβ expression compared to male offspring, or because RSV can directly activate SOD2 and ERRα expression through another pathway, such as PGC1α [53]. Furthermore, in female rats, RSV increased expression of both SOD2 and ERRα, even in rats that were not exposed to NET, suggesting a more general effect. Our results indicate that another factor other than ERβ may also be involved in regulating the gene activation of SOD2 and ERRα. This can be explained by RSV-mediated PGC-1α activation, as it has been reported that RSV activates PGC-1α either by SIRT1 [53] or AMPK (54)-mediated indirect activation or by direct deacetylation of PGC-1α [55], and the activated PGC-1α then upregulates the expression of ERRα [56, 57] and SOD2 [58].

This study has several limitations. In animal behavioral tests, the marble burying test and the social interaction time were used for the evaluation of autism-like behavior, while these tests do not really reflect adequate tests of this syndrome. The marble burying test has been used to describe anxiety and repetitive behavior, but not clearly presented as a test for autism. A more common test for autism should be the social preference test and the ultrasonic vocalization test. In addition, during the animal treatment, prenatal exposure of either progestin or resveratrol was used throughout the whole pregnancy period (21 days), while this rarely happens in the human being. Usually, the first trimester during the pregnancy is considered as the most sensitive period for the prenatal exposure of risk factors; in this case, prenatal exposure of progestin or resveratrol during the first 7 days of pregnancy in dams may be more reasonable for this treatment. On the other hand, we were using higher doses of resveratrol and progestin for the treatment of pregnant dams in this study in order to achieve a significant effect in animals to mimic the long-term exposure of lower doses in humans, and this may be a potential limitation for human application.