Hematopoietic stem cell transplantation ameliorates maternal diabetes–mediated gastrointestinal symptoms and autism‐like behavior in mouse offspring

Abstract Epidemiological studies have shown that maternal diabetes is associated with autism spectrum disorder development, although the detailed mechanism remains unclear. We have previously found that maternal diabetes induces persistent epigenetic changes and gene suppression in neurons, subsequently triggering autism‐like behavior (ALB). In this study, we investigated the potential role and effect of hematopoietic stem cells (HSCs) on maternal diabetes–mediated gastrointestinal (GI) dysfunction and ALB in a mouse model. We show in vitro that transient hyperglycemia induced persistent epigenetic changes and gene suppression of tight junction proteins. In vivo, maternal diabetes–mediated oxidative stress induced gene suppression and inflammation in both peripheral blood mononuclear cells and intestine epithelial cells, subsequently triggering GI dysfunction with increased intestinal permeability and altered microbiota compositions, as well as suppressed gene expression in neurons and subsequent ALB in offspring; HSC transplantation (HSCT) ameliorates this effect by systematically reversing maternal diabetes–mediated oxidative stress. We conclude that HSCT can ameliorate maternal diabetes–mediated GI symptoms and autism‐like behavior in mouse offspring.


Introduction
Autism spectrum disorders (ASDs) are a group of neurodevelopmental disorders primarily triggered by prenatal risk factor exposure and characterized by impaired social skills and restricted behaviors. 1 The prevalence rate of ASD has increased to 1:54, 2 a These authors contributed equally to this manuscript. with a male to female ratio of > 4:1. 3 A variety of factors, including maternal diabetes [4][5][6] and prenatal exposure to hormones (e.g., progestin and androgen), [7][8][9][10] have been reported to contribute to ASD development, while the detailed mechanism of their cause and related therapies remains largely unknown and unavailable, respectively.
Gastrointestinal (GI) symptoms have been reported to be associated with ASD, 11 and the gut microbiota has been associated with brain dysfunction and ASD pathogenesis through the gut-brain axis. Furthermore, microbiota transfer therapy has been reported to ameliorate GI and ASD symptoms in autistic patients, 12 while the cause-effect relationship between ASD and GI dysfunction (and related mechanisms) remains unclear. [13][14][15][16] Tight junction proteins, including claudin 1 (CLDN1), occludin (OCLN), and zonula occludens-1 (ZO1), are the principal determinants of GI barrier function. 17 Disruption of tight junctions is involved in GI symptoms in ASD patients, who show increased intestinal permeability; 18,19 increased diamine oxidase (DAO) activity 20 and zonulin levels in serum 21 are considered to be intestinal permeability markers in autistic subjects. Hematopoietic stem cells (HSCs) located within the bone marrow are responsible for longterm maintenance and generation of blood cell lineages. 22,23 We have previously reported that maternal diabetes induces epigenetic changes and gene suppression in neurons that, subsequently, contribute to autism-like behaviors (ALBs) in offspring. [4][5][6] Furthermore, maternal diabetesmediated epigenetic modifications in HSCs can be inherited by peripheral blood mononuclear cells (PBMCs) during differentiation, 24 subsequently inducing immune dysfunction in autismlike offspring. 25,26 While HSC-based therapeutic treatment has been shown to ameliorate ASD symptoms, 27,28 the underlying mechanism remains unclear and the strategy needs improvement. 29 It has been early reported that transient hyperglycemia can induce persistent epigenetic changes and alter gene expression during subsequent normoglycemia. 30 We have recently found that maternal diabetes induces persistent gene suppression of superoxide dismutase 2 (SOD2) 5 and oxytocin receptor 4 in brain regions, and these changes contribute to ALB in rodent offspring. 25 Furthermore, in those studies, which used a maternal diabetes model of streptozotocin injectioninduced type I diabetes, the subsequent mouse offspring showed ALBs in almost every individual, which included suppressed ultrasonic vocalization (USV), as well as impaired social recognition and interaction. 4,6,25 In the present study, we evaluated the potential effect of maternal diabetes-mediated epigenetic changes and gene expression in the GI system of mouse offspring. We investigated the effects of HSC transfer on GI dysfunction and ALB with the use of a maternal diabetes-mediated autismlike mouse model. In vitro, using human colon stem cells, we show that transient hyperglycemia induces persistent suppression of CLDN1 and ZO1 through epigenetic modifications, and that SOD2 expression reverses, while short-hairpin (sh)SOD2 expression mimics, this effect. In vivo, we show that HSC transplantation (HSCT) ameliorates maternal diabetes-mediated oxidative stress and inflammation in PBMCs and intestinal epithelial cells (IECs), and it reverses tight junction protein suppression in IECs, GI dysfunction, and altered gut microbiota composition. We also show that HSCT ameliorates maternal diabetes-mediated ALB by reversing maternal diabetes-mediated systemic oxidative stress. We conclude that HSCT ameliorates maternal diabetes-mediated GI symptoms and ALB in autism-like offspring. The results offer a possible therapeutic target, and potential mechanism of HSC-mediated treatment, for individuals with ASD.

Materials and methods
Detailed Materials and Methods section is described in Supplementary Materials (see File S1, online only), and primers for RT-qPCR are provided in Table S1 (online only).

Preparation of lentiviruses
The lentivirus mouse plasmids for shSod2 (sc-41656-SH) and nontarget control (sc-108060) were obtained from Santa Cruz Biotechnology. The human shSOD2-expressing lentivirus was prepared previously in our lab. 5 Lentiviruses were expressed using Lenti-X TM Lentiviral Expression Systems (from Clontech) according to the manufacturer's instructions.
In vivo mouse experiments Protocol 1. Generation of chronic diabetic mice. Diabetic female mice were generated by injection of 35 mg/kg streptozocin (STZ) after 8-h fasting. Blood glucose with > 250 mg/dL was considered diabetes positive. Females from either control (CTL) or STZ group were caged with males to generate subsequent offspring, and one offspring from each verified pregnant dam (n = 9 for each group) was used for the subsequent treatment in protocol 3.

Protocol 2. Preparation of hematopoietic stem cells with the manipulation of Sod2 expression.
Mouse HSCs were isolated and identified from 4-to 6-week-old male offspring donor mice from either CTL or STZ group, as described below. 33,34 The cells were then purified by density centrifugation using Histopaque 1083 R (#−1083-1, Sigma) and resuspended in 10 mL of RPMI 1640 supplemented with 10% FBS and 2 mM EDTA. The isolated HSCs were infected by lentivirus for the expression of empty (EMP), Sod2 expression (↑Sod2), or Sod2 knockdown (shSod2) to develop HSC/EMP, HSC/↑Sod2, or HSC/shSod2 cells. Protocol 3. Postnatal treatment by hematopoietic stem cell transplantation. Male offspring (6 weeks old) from either CTL or STZ group were used as recipients for HSCT. The recipient male mice were lethally irradiated with two doses of 6 Gy 3 h apart; 35 and after 4 h of irradiation, 2 × 10 6 HSCs prepared from protocol 2 were systemically transplanted by tail vein injection. The experimental mice were randomly separated into the following four groups. Group 1: offspring from CTL group received HSCT with HSCs from the CTL group infected with EMP lentivirus (CTL-HSCT/CTL/EMP); group 2: offspring from STZ group received HSCT with HSCs from the STZ group infected with EMP lentivirus (STZ-HSCT/STZ/EMP); group 3: offspring from STZ group received HSCT with HSCs from the STZ group infected with ↑Sod2 lentivirus (STZ-HSCT/STZ/↑Sod2); and group 4: offspring from CTL group received HSCT with HSCs from the CTL group infected with shSod2 lentivirus (CTL-HSCT/CTL/shSod2). All transplant recipient mice were set aside for a minimum of 12 weeks to allow for complete reconstitution of the bone marrow before they were used for evaluation of ALB and GI symptoms; brain tissues, including amygdala, hypothalamus, and hippocampus, were then collected. Whole blood was collected by heart puncture and the serum, and PBMCs were isolated. IECs were isolated as described below for further biological assays.

Animal behavior test
The ALB test from mouse offspring was evaluated using USV, social interaction (SI) tests, and a three-chambered social test as described in Supplementary Information (see Data S1, online only).

Isolation and characterization of HSCs
The HSC preparation was a minor modification from previously described method. 33 In brief, the whole bone marrow cells were collected from tibias in treated mice. Bone marrow cells were stained with antibodies for the identification of HSCs (c-Kit + Sca1 + Lin -), and the following antibodies were used: c-Kit-PE (#12-1171-82), Sca-1-FITC (#11-5981-82), and anti-Lineage Antibody Cocktail comprises a mixture of PE-Cy5-conjugated antibodies, including anti-B220, anti-CD4, anti-CD8, anti-Gr-1, anti-Mac-1, and anti-TER119 (from eBioscience). For HSC sorting, debris, dead, and clumped cells were removed to obtain single and viable cells; then, the c-Kit + Sca1 + Lincell population was isolated by HSC sorting; FACS analysis was performed on BD FACSMelody TM Cell Sorter. 34

Isolation of mouse IECs
The protocol for isolation of IECs was based on the previously described method with minor modifications. In brief, the small and large intestines were harvested individually from treated mice and rinsed extensively with RPMI-1640 media (from Lonza) after Peyer's patches were removed (for small intestine). The rinsed intestines were opened longitudinally and macerated; the tissue pieces were shaken gently in RPMI-1640 containing 2 mM EDTA and 10% fetal calf serum. The tissue preparations were passed through 70-μm mesh filters, and the resulting single-cell suspensions were applied to Percoll (from Sigma) density gradients of 25%, 40%, and 75%. After centrifugation at 2000× g for 20 min, the interface between the 25% and 40% layers was collected to obtain IECs. The cells were stained using antibodies for either epithelial cell adhesion molecule (EpCAM, from Biolegend) or CD45 (from Biolegend) and nucleic acid dye (Via-Probe, from BD Biosciences). The Via-Probe -CD45 -EpCAM + IECs were sorted using the BD FACSMelody TM Cell Sorter (BD Biosciences) for further biological assays. 36,37 Intestinal permeability assay An intestinal permeability assay was followed based on the previously described method with minor modifications. In brief, treated mice were fasted for 4 h before the experiment and then, the FITCdextran (50 mg/mL, Cat# 46944 from Sigma) was gavaged (600 mg/kg). After 4 h, whole blood was collected by cardiac puncture and placed at room temperature for 1 h before being centrifuged at a speed of 3000 rpm for 10 minutes. The supernatant was then transferred to a new tube for further centrifugation at a speed of 12,000 rpm for 10 min at 4°C. The subsequent supernatant (serum) was diluted with equal volume of PBS, and then 100 μL of the diluted serum was added to a 96cell microplate. The concentration of FITC in the serum was determined at excitation/emission wavelengths of 485/530 nm using an FL×800 microplate fluorescence reader (Bio-Tek). Serial diluted FITCdextran (0, 0.5, 1, 2, 4, 6, 8, and 10 μg/μL) was used as standards. Serum of mice administered PBS was used as a negative control. 14,38 Fecal microbiome analysis Fecal samples of the experimental mice were collected and stored at −80°C before being processed. Microbial DNA was extracted using a QIAamp Fast DNA Stool Mini Kit (from Qiagen) according to the manufacturer's protocol. 39 The purity and concentration of the extracted DNA were evaluated using agarose gel electrophoresis. The fecal microbiota was studied by performing V3-V4 16S rDNA amplicon sequencing to obtain operational taxonomic units (OTU) defining the bacterial communities. 40 Sequencing samples from frozen fecal pellets were prepared, sequenced, and subsequently processed using the MiSeq Pe300 Sequencing Platform (from Illumina) by Shanghai OE Biotech Inc. The raw data were treated and processed using a QIIME 2 TM software package, and the subsequent sequences of OTU were blasted in the Silva database (version 138). The alpha and beta diversities were analyzed using QIIME 2 TM software package. 14

Serum biochemical analysis
The whole blood was collected from experimental mice by heart puncture and the serum was prepared by centrifugation at 2000× g for 15 minutes. The specimens were then stored at −80°C until analysis. The GSH/GSSG ratio was determined using a GSH/GSSG Ratio Detection Assay Kit (Fluorometric-Green) (#ab138881) from Abcam, while DAO activity was measured by DAO ELISA Kit (#MBS160374) and the zonulin levels were determined by Mouse Zonulin ELISA Kit (#NC1314884) from MyBioSource according to the manufacturers' instructions. 21  Protein Assay Kit (Pierce Biotechnology) according to the manufacturer's instructions; lysates were adjusted to 200 mg/mL with extraction buffer; 50 mL lysates were diluted two-fold in sample dilution buffer and analyzed in duplicates. Analytes were quantified in each sample against a calibration curve of known concentrations. 42

Transient hyperglycemia triggers persistent gene suppression during subsequent normoglycemia through consistent oxidative stress
We evaluated the potential effect of transient hyperglycemia on the expression of tight junction proteins ZO1, CLDN1, and OCLN. Human colon stem cells were treated with either 5 mM low glucose (LG) or 25 mM high glucose (HG) for 4 days (4d) and then infected by lentiviruses control (CTL), SOD2 overexpression (↑SOD2), or SOD2 knockdown (shSOD2) for 1 day before they were further treated by LG for another 4 days. The results showed that 4-day hyperglycemia plus 4-day normoglycemia (HG(4d) + LG(4d)/CTL) treatment significantly suppressed mRNA levels of ZO1 and CLDN1, compared with the LG plus 4-day normoglycemia (LG(4d) + LG(4d)/CTL) treatment. Cells with SOD2 overexpression under hyperglycemia (HG(4d) + LG(4d)/↑SOD2) did not show suppressed mRNA levels of ZO1 and CLDN1, while knockdown of SOD2 under LG (LG(4d) + LG(4d)/shSOD2) mimicked the HG effect. In contrast to ZO1 and CLDN1, OCLN mRNA did not differ significantly among the treatment groups (Fig. 1A). We also measured protein levels, and an expression pattern consistent with the mRNA levels was observed ( We then evaluated the possible effect of epigenetic modifications on the ZO1 and CLDN1 promoters. The results showed that HG(4d) + LG(4d)/CTL treatment significantly increased DNA methylation on the CLDN1 promoter ( Fig. 1D) but had no effect on H3K9 and H3K27 modifications (Fig. 1E). Treatment did not change DNA methylation on the ZO1 promoter (Fig. 1F), but significantly increased H3K9me3 and H3K27me2 modifications (Fig. 1G), compared with LG (LG(4d)  Furthermore, we evaluated the possible effect of hyperglycemia and SOD2 expression on H4 methylation and histone acetylation on the CLDN1 (Fig.  S2, online only) and ZO1 (Fig. S3, online only) promoters; neither showed any changes. We concluded that transient hyperglycemia triggers epigenetic changes and gene suppression during subsequent normoglycemia.  Fig. 2A and B), indicating that HSCs can differentiate into PBMCs after HSCT. We then evaluated the potential effect of HSCT on endogenous Sod2 expression in PBMCs in the context of induced maternal diabetes, including effects on Sod2 mRNA (Fig. 2C), SOD protein level (Fig. 2D and E; and Fig. S1B, online only), and SOD2 activity (Fig. 2F). The data showed that maternal diabetes induction (STZ-HSCT/STZ/EMP) significantly decreased endogenous Sod2 expression, compared with the control (CTL-HSCT/CTL/EMP) group. STZ-HSCT/STZ/↑Sod2 treatment reversed, while CTL-HSCT/CTL/shSod2 treatment mimicked, this effect.
Together, the above results indicated that HSCT with increased Sod2 expression ameliorates maternal diabetes-mediated oxidative stress in PBMCs.

Potential effect of HSCT with increased Sod2 expression on maternal diabetes-mediated gene regulation in IECs
Next, we evaluated the potential effect on gene expression of Sod2 and the tight junction protein genes Zo1, Cldn1, and Ocln. We found that maternal diabetes induction (STZ-HSCT/STZ/EMP) significantly decreased mRNA levels of Sod2, Zo1, and Cldn1, compared with the control (CTL-HSCT/CTL/EMP) group. STZ-HSCT/STZ/↑Sod2 treatment reversed, while CTL-HSCT/CTL/shSod2 treatment mimicked, the effect on Cldn1 mRNA levels but had no effect on Sod2 and Zo1 mRNAs. In addition, none of the treatments had any effect on Ocln mRNA levels (Fig. 4A). We also measured the protein levels and found expression patterns consistent with those of the mRNA levels ( Fig. 4B  and C; Fig. S1D, online only). We then evaluated SOD2 activity in IECs, finding an activity pattern consistent with its mRNA level. This further confirmed that HSCT with increased Sod2 expression did not affect maternal diabetes-mediated SOD2 suppression in IECs (Fig. 4D). We also measured CLDN1 protein level by immunostaining in IECs and found a pattern consistent with its mRNA level, indicating that STZ-HSCT/STZ/↑Sod2 treatment reverses, while CTL-HSCT/CTL/shSod2 treatment mimics, maternal diabetes-mediated CLDN1 suppression in IECs ( Fig. 4E and F).
Schematic model for the potential effect of HSCT with increased Sod2 expression on maternal diabetes-mediated GI symptoms and ALB Maternal diabetes-mediated oxidative stress triggers epigenetic modifications in subsequent off-spring tissues. Epigenetic modification in IECs and HSCs/PBMCs results in gene suppression, oxidative stress, and inflammation, subsequently contributing to GI symptoms, which include increased intestinal permeability and altered gut microbiota compositions. Furthermore, epigenetic modification in the brain results in gene suppression and oxidative stress, triggering ASD development and ALB. On the other hand, HSCT operation and the manipulation of Sod2 expression diminish maternal diabetes-mediated systematic oxidative stress (e.g., increased GSH/GSSG ratio), subsequently reversing maternal diabetesmediated GI symptoms and ASD development (Fig. 8).

Discussion
In this study, we found that transient hyperglycemia induces persistent suppression of tight junction proteins by epigenetic modifications. Maternal diabetes induces gene suppression, oxidative stress, and inflammation in PBMCs and IECs, subsequently contributing to GI dysfunction in autism-like animal models, while HSCT with increased Sod2 expression partly reverses this effect and subsequently ameliorates maternal diabetesmediated ALB.

Potential effect of HSCT on maternal diabetes-mediated IECs and GI dysfunction
It has been reported that hyperglycemia-mediated ROS overgeneration is the potential driving force for diabetic complications, 43 and transient hyperglycemia triggers persistent epigenetic changes by ROS generation. 30 We have recently reported that maternal diabetes induces ALB through hyperglycemia-mediated persistent oxidative stress and Sod2 mRNA suppression, indicating that SOD2 may play an important role in maternal diabetesmediated ALB and GI dysfunction. 5   compositions. On the other hand, transplantation of Sod2-expressing HSCs partly reversed, while transplantation of shSod2-expressing HSCs partly mimicked this effect, indicating that maternal diabetes-mediated oxidative stress and subsequent epigenetic changes and gene suppression could be the potential driving force for GI dysfunction in autism-like offspring of female mice with maternal diabetes. Also, interestingly, we found that transplantation of Sod2-expressing HSCs could partly reverse maternal diabetes-mediated Cldn1 suppression in IECs in offspring, which can be explained by Sod2-expressing HSCs reducing maternal diabetesmediated ROS generation, subsequently reversing oxidative stress-mediated epigenetic changes on the Cldn1 promoter. 4,6 In addition, the antioxidant and SOD mimetic MnTBAP 5 was also used for the treatment and produced a similar effect as SOD2 on IECs, PBMCs, and GI dysfunction, but had little effect on maternal diabetes-mediated ALB; one explanation for this is that antioxidants may be effective transiently, while Sod2 expression is not. 43 HSCs can differentiate into PBMCs after HSCT We have previously found that maternal diabetesmediated oxidative stress triggers epigenetic changes and gene suppression in both HSCs and PBMCs. PBMCs can inherit the similar properties from HSCs during differentiation after HSCT operation, and these properties may include epigenetic changes and subsequent gene suppression. 25 In this study, we showed that GFP lentivirus-infected HSCs from donor mice were found in PBMCs from recipient mice with ∼16.3% GFP + after HSCT; this further shows that HSCs can differentiate into PBMCs after HSCT, even though the number of transduced cells is relatively low. This can be explained by the following: (1) GFP lentivirus is lost during HSC differentiation in the absence of selective antibiotics; in fact, most of PBMCs differentiate after HSCT and are not labeled by GFP; (2) GFP + HSCs are partly removed by the recipient immune system after HSCT; and/or (3) GFP + HSCs partly lost their function owing to the presence of GFP protein interference. We assume that HSCs used for HSCT in autism-like offspring may have a higher chance of survival and function because of the absence of GFP interference. 34 Potential effect of HSCT on maternal diabetes-mediated ALB We have previously reported that amygdala neurons play a dominant role in maternal diabetes-mediated ALB through oxidative stress-mediated epigenetic changes and subsequent gene suppression in autism-like offspring. [4][5][6] In our study here, we found that transplantation of Sod2-expressing HSCs partly reversed both maternal diabetes-mediated Sod2 suppression in amygdala tissues as well as subsequent oxidative stress and ALB in offspring. A remaining question is how HSCT affects maternal diabetes-mediated ALB in the presence of the blood brain barrier. 18 It has been reported that increased oxidative stress in the plasma, especially increased GSH/GSSG ratio, can be the potential biomarker for ASD subjects; 44,45 our results above have shown that transplantation of Sod2-expressing HSCs partly reversed the maternal diabetes-mediated GSH/GSSG ratio increase in serum. We hypothesize that HSCT with increased Sod2 expression diminishes maternal diabetes-mediated systematic oxidative stress and epigenetic changes in neurons, subsequently ameliorating ALB in autism-like offspring.

Conclusions
Maternal diabetes induces persistent epigenetic changes and gene suppression in PBMCs, IECs, and neurons, while transplantation of HSCs ameliorates maternal diabetes-mediated GI dysfunction and ALB by systematically reversing maternal diabetes-mediated oxidative stress. HSCs may play a potential therapeutic role for clinical treatment of Individuals with ASD.

Acknowledgments
This study was financially supported by Guangdong Basic and Applied Basic Research Foundation: 2020A1515110644; Shenzhen Science and Technology Project: JCYJ20200109150700942; and Foshan Science and Technology Advanced Project: 1920001000541.

Supporting information
Additional supporting information may be found in the online version of this article.    Figure S4. Transplantation of Sod2-expressing HSCs ameliorates, while transplantation of shSOD2-expressing HSCs mimics, maternal diabetes-mediated inflammation in IEC. Figure S5. Potential effect of HSCT on maternal diabetes-mediated gene expression in the brain. Table S1. Sequences of primers for real-time quantitative PCR (qPCR) Supplementary File S1. Materials and methods