Autism‐like behavior of murine offspring induced by prenatal exposure to progestin is associated with gastrointestinal dysfunction due to claudin‐1 suppression

Autism spectrum disorders (ASD) are associated with the contribution of many prenatal risk factors; in particular, the sex hormone progestin and vitamin D receptor (VDR) are associated with gastrointestinal (GI) symptoms in ASD development, although the related mechanism remains unclear. We investigated the possible role and mechanism of progestin 17‐hydroxyprogesterone caproate (17‐OHPC) exposure‐induced GI dysfunction and autism‐like behaviours (ALB) in mouse offspring. An intestine‐specific VDR‐deficient mouse model was established for prenatal treatment, while transplantation of haematopoietic stem cells (HSCT) with related gene manipulation was used for postnatal treatment for 17‐OHPC exposure‐induced GI dysfunction and ALB in mouse offspring. The in vivo mouse experiments found that VDR deficiency mimics prenatal 17‐OHPC exposure‐mediated GI dysfunction, but has no effect on 17‐OHPC‐mediated autism‐like behaviours (ALB) in mouse offspring. Furthermore, prenatal 17‐OHPC exposure induces CLDN1 suppression in intestine epithelial cells, and transplantation of HSCT with CLDN1 expression ameliorates prenatal 17‐OHPC exposure‐mediated GI dysfunction, but has no effect on 17‐OHPC‐mediated ALB in offspring. In conclusion, prenatal 17‐OHPC exposure triggers GI dysfunction in autism‐like mouse offspring via CLDN1 suppression, providing a possible explanation for the involvement of CLDN1 and VDR in prenatal 17‐OHPC exposure‐mediated GI dysfunction with ASD.

Autism spectrum disorders (ASD) are associated with the contribution of many prenatal risk factors; in particular, the sex hormone progestin and vitamin D receptor (VDR) are associated with gastrointestinal (GI) symptoms in ASD development, although the related mechanism remains unclear. We investigated the possible role and mechanism of progestin 17hydroxyprogesterone caproate (17-OHPC) exposure-induced GI dysfunction and autism-like behaviours (ALB) in mouse offspring. An intestinespecific VDR-deficient mouse model was established for prenatal treatment, while transplantation of haematopoietic stem cells (HSCT) with related gene manipulation was used for postnatal treatment for 17-OHPC exposure-induced GI dysfunction and ALB in mouse offspring. The in vivo mouse experiments found that VDR deficiency mimics prenatal 17-OHPC exposure-mediated GI dysfunction, but has no effect on 17-OHPCmediated autism-like behaviours (ALB) in mouse offspring. Furthermore, prenatal 17-OHPC exposure induces CLDN1 suppression in intestine epithelial cells, and transplantation of HSCT with CLDN1 expression ameliorates prenatal 17-OHPC exposure-mediated GI dysfunction, but has no effect on 17-OHPC-mediated ALB in offspring. In conclusion, prenatal 17-OHPC exposure triggers GI dysfunction in autism-like mouse offspring via CLDN1 suppression, providing a possible explanation for the involvement of CLDN1 and VDR in prenatal 17-OHPC exposure-mediated GI dysfunction with ASD.

Introduction
Autism spectrum disorder (ASD) development has been associated with the contribution of many prenatal factors [1,2], including sex hormones in particular, although the mechanism remains to be fully addressed [3][4][5]. Progestins, the main components of oral contraceptive hormones, were originally used for birth control or prevention of preterm birth; later on, they also began to be widely misused in agriculture, resulting in contamination of the environment. For instance, high doses of oral contraceptives were either sprayed on the surface of fruits (e.g., grapes and strawberries) or fed to commercial seafood (e.g., fish and lobsters) to speed upgrowth by preventing female reproduction [6,7]. Thus, humans may also be exposed to exogenous progestins through these food sources. Our epidemiological study showed that prenatal hormone exposure, which includes the use of progestin during pregnancy and progestin-contaminated food intake, is coincided with significantly increased ASD prevalence [8]. However, there is no evidence to show that higher levels of endogenous progestin (e.g., progesterone) are associated with increased risk for ASD. In addition, prenatal progestin exposure triggers autism-like behaviours (ALB) and social deficits in rodent offspring [9][10][11]. This has been supported by evidence from a variety of epidemiological studies showing that prenatal progestin exposure may be associated with autism development [6,12].
Gastrointestinal (GI) symptoms may be associated with autism development [13], and gut microbiota may affect brain function and animal behaviours by gutbrain axis [14,15]. Furthermore, sex hormones can trigger GI symptoms, including inflammatory bowel disease [16] and alterations of gut microbiota [17][18][19][20], although the potential mechanism for hormonemediated GI symptoms remains largely unknown. Tight junction proteins, which primarily include claudins (CLDN), occludin, junction adhesion molecule, and zonula occludens, play a critical role in GI barrier permeability and selectivity [21]. Additionally, sex hormones regulate the expression of tight junctions, subsequently contributing to GI symptoms [22,23]; moreover, we have recently shown that maternal diabetes suppresses CLDN1 in mouse offspring [24]. Thus, we hypothesize that prenatal progestin exposure may modulate GI dysfunction through suppression of tight junctions in autism-like offspring.
Vitamin D receptor (VDR) becomes active after heterodimer formation with retinoid X receptor in the presence of vitamin D. It then translocates into the nucleus and binds with vitamin D response element, subsequently initiating target gene expression [25]. Both VDR and vitamin D have been reported to modulate brain function and gut microbiota [26], which are associated with ASD development [27,28]. Additionally, VDR is involved in modification of gut microbiota, while the detailed mechanism remains unclear [29,30].
We investigated the possible effect and mechanism for prenatal progestin exposure-induced GI dysfunction and ALB in offspring. 17-hydroxyprogesterone caproate (17-OHPC) was selected for this study due to its wide usage in the prevention of preterm birth [7,31]. The preliminary study found that 17-OHPC exposure triggers claudin 1 (CLDN1) suppression through epigenetic changes and decreased binding ability of VDR on the CLDN1 promoter, suggesting that VDR plays a role in 17-OHPC-induced GI dysfunction through regulation of CLDN1 expression. The intestine-specific VDR knockdown mice were established to mimic CLDN1 deficiency for investigating the potential contribution of CLDN1 on GI dysfunction in autism-like mouse offspring. In addition, haematopoietic stem cell (HSC) transplantation (HSCT) with CLDN1 expression [24] was employed to study the possible rescuing effect on 17-OHPC exposuremediated GI dysfunction. Also, prenatal exposure of either maternal diabetes or progestin triggers gene suppression of oestrogen receptor b (ERb) [9][10][11], superoxide dismutase (SOD2) [32], and synaptophysin (SYP) [33] in the amygdala, contributing to autism-like behaviours in offspring, while other regions (e.g., hypothalamus and hippocampus) showed little effect. Thus, we chose the amygdala as the region of interest for this study, and other regions such as the hypothalamus and hippocampus, which may also have potential effects [33,34] were used as controls to eliminate potential bias [33]. The expression of SYP, ERb, and SOD2 in the above regions was chosen as the indicator to evaluate potential autism-like behaviours in mouse offspring. Moreover, we have previously reported that male offspring seem to be more susceptible to prenatal progestin exposure-mediated ALB compared to female offspring [10]; thus, male offspring were selected for the experimental analysis to exclude the possible confounding effects of oestrogen and oestrogen receptors [10,35].

Results
Intestine-specific VDR deficiency shows no effect on 17-OHPC exposure-induced ALB The wild type (WT) or intestine-specific VDR knockdown (Vil-shVDR) dams received either CTL or 17-OHPC treatment, and the mouse offspring were employed for assays. Gene expression was measured in the amygdala, and it was found that prenatal 17-OHPC treatment significantly reduced mRNA expression of SOD2, SYP, and ERb, while VDR deficiency had no effect (see Fig. 1A). Measurements were also taken for the related proteins, and they were consistent with mRNA levels (see Fig. 1B,C and Fig. S1A). We also determined oxidative stress in the amygdala and found that 17-OHPC exposure potentiated O 2 ÁÀ release (see Fig. 1D) and 8-OHdG generation (see Fig. 1E), while VDR deficiency did not show any effect. In addition, we determined mRNA levels in other brain tissues and found that 17-OHPC exposure reduced SYP levels only in the hypothalamus (see Fig. 1F), but not in the hippocampus (see Fig. 1G). Levels of other genes, including SOD2 and ERb, did not change significantly, and again, VDR deficiency did not show any effect. Finally, measurements of ALB exhibited by offspring were taken, and we found that prenatal 17-OHPC treatment significantly reduced ultrasonic vocalization (see Fig. 2A), social interaction time (see Fig. 2B), as well as sociability (see Fig. 2C) and social novelty (see Fig. 2D); again, VDR deficiency had no significant effect on animal behaviours.
Intestine-specific VDR deficiency mimics prenatal 17-OHPC exposure-mediated gastrointestinal dysfunction The wild type (WT) or intestine-specific VDR knockdown (Vil-shVDR) dams were injected with either control (CTL) or 17-OHPC, and the related offspring were employed for analysis. Gene expression in intestine epithelial cells (IEC) was measured, and it was found that 17-OHPC treatment reduced CLDN1 mRNA levels, but did not affect VDR mRNA (see Fig. 3A). Protein levels were determined, and the expression was similar to that of mRNA levels (see Fig. 3B,C and Fig. S1B). We then evaluated GI function, and it was determined that 17-OHPC exposure significantly potentiated intestinal permeability compared with the WT/CTL treatment; VDR deficiency mimicked the effect (see Fig. 3D). We also determined microbial composition through 16S rRNA sequencing techniques. It was found that species richness (see Fig. 3E) and diversity (see Fig. 3F) did not differ significantly among treatments. Furthermore, 17-OHPC exposure triggered significant changes in microbial composition; the phyla Firmicutes, Proteobacteria, and Verrucomicrobia dominated the microbiome in WT/ CTL offspring, but a shift toward Firmicutes, Deferribacteres, and Actinobacteria occurred as a result of 17-OHPC treatment (WT/OHPC); VDR deficiency mimicked the effect of 17-OHPC (see Fig. 3G). In addition, the relative abundance of g_Mucispirillum was reduced significantly (see Fig. 3H), and the relative abundance of p_Proteobacteria and p_Tenericutes increased, while abundance of p_Deferribacteres decreased significantly (see Fig. 3I) in WT/OHPC mice compared with WT/CTL mice; VDR deficiency   mimicked the effect. We conclude that VDR deficiency mimics 17-OHPC exposure-mediated GI dysfunction through CLDN1 suppression.
Haematopoietic stem cell transplantation (HSCT) of ERb ameliorates prenatal 17-OHPC exposuremediated autism-like behaviours, while CLDN1 transplantation shows no effect Dams were injected with either CTL or 17-OHPC, and the related offspring received HSCT with the expression of empty (EMP), CLDN1 (↑CLDN1), or ERb (↑ERb). The treated animals were used for analysis after 12 weeks of HSCT. Gene expression was measured, and it was found that 17-OHPC exposure reduced mRNA levels of SOD2, SYP, and ERb in the amygdala; HSCT of CLDN1 had no effect, while HSCT of ERb partly reversed the effect (see Fig. 4A). Moreover, protein levels were measured and found to be consistent with related mRNA levels (see Fig. 4B, C and Fig. S1C). In addition, mRNA levels were determined in other brain tissues, and it was found that 17-OHPC exposure reduced SYP mRNA in the hypothalamus (see Fig. 4D), but not in the hippocampus (see Fig. 4E). mRNA levels of ERb and SOD2 did not change significantly; additionally, HSCT of either CLDN1 or ERb did not show any effect. We then evaluated redox balance in serum, and it was found that 17-OHPC treatment (OHPC/HSC-EMP) significantly reduced the GSH/GSSG ratio compared with the CTL/HSC-EMP group; HSCT of either CLDN1 or ERb partly reversed the effect (see Fig. 5A). We also determined the proinflammatory cytokines in serum and found that 17-OHPC exposure potentiated cytokine levels in serum compared with the CTL/HSC-EMP group. HSCT of CLDN1 (OHPC/HSCT-↑CLDN1) had little effect, while HSCT of ERb (OHPC/HSCT-↑ERb) partly reversed the effect (see Fig. 5B-E). Autism-like behaviours in the mouse offspring was evaluated, and it was found that 17-OHPC treatment significantly reduced ultrasonic vocalization (see Fig. 6A) and social interaction time (see Fig. 6B), as well as sociability (see Fig. 6C) and social novelty (see Fig. 6D). HSCT of CLDN1 showed no effect on behaviour, while HSCT of ERb completely reversed the 17-OHPC-induced effect on ultrasonic vocalization and partly reversed the effect on social interaction, but showed no effect on the three-chambered social test. In conclusion, HSCT of ERb ameliorates 17-OHPC exposure-mediated autism- like behaviours, while HSCT of CLDN1 had no effect in offspring.
Haematopoietic stem cell transplantation (HSCT) of CLDN1 or ERb ameliorates 17-OHPC exposuremediated oxidative stress and proinflammation in IEC cells Dams were injected with CTL or 17-OHPC, and the offspring received HSCT with the expression of empty (EMP), CLDN1 (↑CLDN1), or ERb (↑ERb), and the IEC were purified for biological assays. Gene expression in IEC cells was determined, and it was found that 17-OHPC treatment significantly reduced CLDN1 mRNA expression; HSCT of either CLDN1 or ERb completely reversed this effect, and gene manipulation of CLDN1 and ERb through lentivirus in HSCT was successful (see Fig. 7A). We also measured the proteins, and they were consistent with mRNA levels (see Fig. 7B,C and Fig. S1D). In addition, the CLDN1  protein was quantified by immunostaining, and the expression was similar to mRNA expression (see Fig. 7D,E). We then determined redox balance and inflammation, and it was found that 17-OHPC exposure significantly reduced the GSH/GSSG ratio (see Fig. 7F) and increased the expression of proinflammatory cytokines (see Fig. 7G); HSCT of either CLDN1 or ERb partly reversed this effect.
Haematopoietic stem cell transplantation (HSCT) of either CLDN1 or ERb ameliorates prenatal 17-OHPC exposure-mediated gastrointestinal dysfunction GI dysfunction was evaluated in the offspring, and we found that 17-OHPC exposure potentiated intestinal permeability; HSCT of CLDN1 had no effect, while HSCT of ERRb partly reversed the effect (see Fig. 8A).
The microbial populations were also determined, and we found that species richness (see Fig. 8B) and diversity (see Fig. 8C) had no difference among treatments. Additionally, 17-OHPC exposure achieved a significant difference in gut microbial composition. The phyla Firmicutes, Proteobacteria, and Verrucomicrobia dominated the microbiome in CTL/HSC-EMP offspring, but with 17-OHPC treatment (OHPC/HSC-EMP), a shift toward Firmicutes, Deferribacteres, and Actinobacteria occurred, and HSCT of either CLDN1 or ERb partly ameliorated the effect (see Fig. 8D). Additionally, the relative abundance of g_Mucispirillum reduced significantly (see Fig. 8E), and the relative abundance of p_Proteobacteria and p_Tenericutes increased, while abundance of p_Deferribacteres decreased significantly (see Fig. 8F) in OHPC/HSC-EMP mice compared with CTL/HSC-EMP mice; again, HSCT of either CLDN1 or ERb partly reversed this effect. We conclude that HSCT of either CLDN1 or ERb ameliorates prenatal 17-OHPC exposure-mediated GI dysfunction.

Schematic model for progestin 17-OHPC exposure-mediated GI dysfunction and postnatal treatment in autism-like offspring
Prenatal exposure of progestin 17-OHPC induces epigenetic changes both in the brain and in tight junction proteins, subsequently triggering ALB and GI dysfunction in mouse offspring [10,11,24,36]. Prenatal treatment with intestine-specific VDR deficiency mimics 17-OHPC exposure-mediated GI dysfunction, indicating that VDR-mediated CLDN1 suppression in IEC plays a critical role in GI dysfunction. Postnatal treatment of HSCT with CLDN1 expression ameliorates 17-OHPC-mediated GI dysfunction, but has no effect on 17-OHPC-mediated ALB. Meanwhile, HSCT treatment with ERb expression ameliorates both 17-OHPCmediated GI dysfunction and ALB, indicating that 17-OHPC-mediated epigenetic changes and CLDN1 suppression contributes to GI dysfunction in mouse offspring (see Fig. 9).

Discussion
We showed that intestine-specific VDR deficiency mimicked prenatal 17-OHPC exposure-induced GI dysfunction and ALB in mouse offspring. HSCT with CLDN1 expression partly reversed 17-OHPC exposure-mediated GI dysfunction, but had no effect on ALB in offspring.

17-OHPC-mediated autism-like behaviours
Sex hormones are associated with ASD development [8], and prenatal exposure to either progestins or androgens triggers autism-like behaviours and social deficits in rodent offspring [9][10][11]37]. 17-OHPC was initially used as an oral contraceptive hormone; in more recent years, it has become more widely used for prevention of preterm birth [7,38]. We have previously studied the effects of 20 mgÁkg À1 body weight of medroxyprogesterone acetate (MPA) as a prenatal treatment in rodents to mimic high doses of MPA exposure in humans [11]; in this study, 10 mgÁkg À1 body weight of 17-OHPC were used to mimic medium doses of progestin exposure to evaluate the possible effect on ALB in mouse offspring. Moreover, it has been reported that prenatal 17-OHPC treatment decreased cognition and altered prefrontal cortex anatomy in rat offspring [34], and a national cohort study in France reported that prenatal exposure to synthetic progestins is associated with neurodevelopmental disorders in children [12]. The present study reveals that prenatal 17-OHPC treatment induces significant gene suppression of SOD2 [32], ERb [10], and SYP [33] in brain tissues with obvious ALB in offspring, suggesting that prenatal progestin exposure may contribute to ASD prevalence [6]; thus, the usage of progestins for prevention of preterm birth is a risk factor for ASD development [31,38].

17-OHPC-mediated CLDN1 suppression and GI dysfunction
Our results showed that intestine-specific VDR deficiency mimics, while HSCT of CLDN1 partly reverses, 17-OHPC exposure-induced CLDN1 suppression and GI dysfunction in mouse offspring, suggesting that VDR may potentially contribute to hormone-mediated GI dysfunction by regulation of CLDN1 expression. This reveals a possible mechanism for VDR-mediated GI dysfunction [26,29,30] and provides a potential target for treatment of hormone-mediated GI dysfunction through regulation of CLDN1 expression [18,19]. This is consistent with our recent finding that CLDN1 is suppressed in IEC with GI dysfunction in maternal diabetes-induced offspring [24]. Interestingly, it has been reported that VDR is associated with ASD development [39][40][41][42], leading us to hypothesize that VDR may contribute to brain function modulation by regulation of gene expression of tight junction components; this potential mechanism is still under our investigation.

HSCT for GI dysfunction treatment
We have previously reported that transplantation of HSC with SOD2 expression can partly reverse maternal diabetes-induced ALB and GI dysfunction, indicating that HSCT with related gene manipulation may be a potential choice for treatment [24]. In this study, we found that HSCT with CLDN1 expression can partly reverse prenatal 17-OHPC exposuremediated GI dysfunction, but has no effect on ALB, indicating that CLDN1 suppression in IEC may partly contribute to prenatal 17-OHPC exposure-mediated GI dysfunction. In addition, HSCT with ERb expression partly reverses 17-OHPC-mediated ALB and GI dysfunction, while transplantation of healthy HSC without any gene manipulation shows no effect, indicating that ERb may contribute to 17-OHPC-mediated epigenetic modifications as evidenced by previous reports [24]. This can help explain why transfusion of healthy stem cells appears to have very limited clinical therapeutical effects for people with ASD [43,44]. In order to achieve better therapeutic effects through HSCT, gene manipulation on stem cells may be an option, while gene manipulation through lentiviruscarried expression may introduce extra risks during treatment. In this case, gene manipulation through cytokine-mediated epigenetic modification becomes an interesting alternative for improving outcomes for people for ASD, and this process is still under our investigation [45,46].

Conclusions
Prenatal 17-OHPC exposure triggers ALB and GI dysfunction in mouse offspring. Intestine-specific VDR deficiency partly mimics, while HSCT of CLDN1 expression partly reverses, prenatal 17-OHPC exposure-induced GI dysfunction. We conclude that prenatal 17-OHPC exposure induces GI dysfunction by CLDN1 suppression in autism-like mouse offspring.

Reagents and materials
The

Analysis of gene expression
RNA was purified and reverse transcribed using the commercial kit from Qiagen (Shanghai, China), and real-time qPCR was conducted on iCycler iQ (Bio-Rad, Shanghai, China), and the primers are shown in Table 1. The results were calculated using the DD CT method provided by Qiagen using b-actin as the housekeeping gene. Protein levels were analysed by Western blotting, and the proteins were extracted from treated cells, run, isolated and transferred to membrane for staining by related primary antibodies followed by species-specific secondary antibodies. The protein density was quantified by ODYSSEY System (LI-COR, Hongkong, China). Gene expression was also determined by immunostaining. In brief, treated cells were fixed by 4% of paraformaldehyde, then incubated by primary antibody using either 8-oxo-dG or CLDN1, followed by incubation of secondary antibody with the addition of DAPI staining for nuclei. The representative photos were taken by confocal microscope and quantified by IMAGE J. [33].

Isolation and characterization of HSC
Bone marrow cells were collected from wild type with vehicle (WT/CTL) treated mice, and the HSC were then isolated and characterized as described previously by our lab [24]. The verified HSC were then infected by the related lentivirus for overexpression of empty (EMP), CLDN1 (↑CLDN1), or ERb (↑ERb) in order to develop HSC-EMP, HSC-↑CLDN1, or HSC-↑ERb cells. which there is very small variation in the results, four samples were randomly picked from the nine samples in each of the eight exposure groups (n = 4). For other biological assays, such as western blotting, five samples were picked randomly from the nine offspring in each group (n = 5).

Mouse protocol 1: Preparation of intestine-specific VDR deficiency mouse
The VDR fl/fl mouse (loxP flanking insert on first exon of VDR gene) was kindly provided by Dr Haimou Zhang. The Villin-Cre (Vil1-Cre, #021504) transgenic mouse, which expresses Cre recombinase in intestinal epithelial cells, was purchased from Jackson Laboratories. Intestine-specific VDR-deficient mice (named as Vil-shVDR) were bred from VDR fl/fl mice by cross-breeding with Vil1-Cre mouse for more than four generations in a C57BL/6J background, and the positive generation was verified by genotyping using primers provided in Table 1 to confirm the existence of loxP and Cre recombinase [29]. All the experimental animals were fed with mouse chew fortified with vitamin D3 (+VD) [28].

Mouse protocol 2: Prenatal treatment by 17-OHPC or VDR deficiency
Three months old female dams were verified by pregnancy, and then received treatment of either 17-OHPC (10 mgÁkg À1 body weight) or control (CTL) vehicle containing 0.1% ethanol in organic sesame oil [10]. The drugs (0.1 mL, diluted by vehicle) were injected peritoneally every 2 days until offspring delivery. The treated dams (n = 9 for each treatment group) were then assigned into the following four groups: Group 1: VDR wild type (WT) with CTL injection (WT/CTL); Group 2: VDR WT with 17-OHPC injection (WT/OHPC); Group 3: VDR deficiency with CTL injection (Vil-shVDR/CTL); Group 4: VDR deficiency with 17-OHPC injection (Vil-shVDR/OHPC). Amygdala neurons were isolated on embryonic day 18 (E18) for later biological analysis [10,33]. Only one male offspring was randomly separated from each dam and fed until 7-8 weeks old for the measurement of ALB and GI dysfunction. GI dysfunction evaluation was conducted through intestinal permeability assay and faecal microbiome analysis as described in a previous article [24]. The offspring were then sacrificed, and the serum was used for the measurement of GSH/GSSG ratio and cytokine levels of MCP1, IL17a, IL6, and IL1b [24]. Various brain regions, including amygdala, hippocampus, and hypothalamus, were dissected and frozen for later measurement of gene expression and redox balance, and the IEC were isolated and verified for biological assays [24].

Mouse protocol 3: Postnatal treatment by HSCT with gene manipulation
Six weeks old offspring from either the CTL or 17-OHPC treated groups were employed as recipients for HSCT operation. The recipient animals received irradiation by 2 doses of 6 Gy with 3 h apart [48], and 4 h after irradiation, 2 9 10 6 of HSC cells were transplanted through tail vein injection. The animals (n = 9 for each group) were randomly assigned into the following four groups: Group 1: CTL offspring receiving HSCT with HSC cells infected by EMP lentivirus (CTL/HSC-EMP); Group 2: 17-OHPC offspring receiving HSCT with HSC cells infected by EMP lentivirus (OHPC/HSC-EMP); Group 3: 17-OHPC offspring receiving HSCT with HSC cells infected by CLDN1 lentivirus (OHPC/HSC-↑CLDN1); and Group 4: 17-OHPC offspring receiving HSCT with HSC cells infected by ERb lentivirus (OHPC/HSC-↑ERb). Twelve weeks were allowed for HSCT-recipient animals to recover for bone marrow reconstitution before the measurement of ALB and GI dysfunction at 18-19 weeks old, and the same biological assays were conducted as described in Protocol 2.

Evaluation of animal behaviour
Animal behaviours were evaluated in the male offspring (n = 9 for each group) at 7-8 weeks of age for Mouse Protocol 2 and at 18-19 weeks of age for Mouse Protocol 3. Autism-like behaviours was evaluated by ultrasonic vocalization, social interaction test and a three-chambered social test as described previously [33].

Statistical analysis
The data were presented as mean AE SD (standard deviation). For the analysis of animal behaviours and GI dysfunction, nine male offspring were used from each treatment group, and each offspring came from a different dam (n = 9), for a total of 72 offspring across the eight treatment groups. For the mRNA analysis, four samples were randomly selected from the nine samples (n = 4) for each of the eight treatment groups, and for other kinds of molecular analysis, five samples were randomly selected from the nine samples (n = 5) for each treatment group. One-way ANOVA (analysis of variance) together with Tur-keyÀKramer test was employed to measure statistical differences across the different treatments, and two-way ANOVA together with Bonferroni post hoc test was employed to measure the differences of two factors (e.g. VDR deficiency and 17-OHPC treatment) by SPSS 22 (IBM, New York, NY, USA), with a P < 0.05 being considered significant.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article.