Abstract

   Neuroinflammation triggered by microglia activation is hallmark of
   autism spectrum disorder (ASD), and this process includes crucial
   metabolic reprogramming from oxidative phosphorylation to glycolysis,
   which may cause neuron loss and functional impairment. The inhibitory
   immune checkpoint programmed cell death protein 1 (PD-1) on immune
   cells is an important target for tumor immunotherapy. However, the
   immunomodulatory effects of PD-1 in ASD remains to be elusive.
   Mesenchymal stem cell-derived extracellular vesicles (MSC-EVs) exhibit
   immunomodulatory capabilities in a range of neurological diseases. Our
   findings indicated the expression of PD-L1 on MSC-EVs, potentially
   facilitating signaling to PD-1-expressing microglia. Here, we showed
   how MSC-EVs activated of PD-L1/PD-1 axis and ameliorated glycolysis,
   neuroinflammation and autism-like behaviors. After first detecting
   elevated glycolysis and neuroinflammation in prefrontal cortex (PFC)
   tissue from the maternal immune activation (MIA) mice, we also
   demonstrated that PD-1 expression level was upregulated in microglia.
   Following given MSC-EVs carried PD-L1 into adult MIA offspring mice via
   intranasal administration, which bound with PD-1 on microglia and then
   the autism-like behaviors were alleviated as well. Further experiments
   verified that MSC-EVs could decreased the level of glycolysis and
   neuroinflammation by PD-1/ERK/HIF-1α pathway in the primary microglia
   in PFC of MIA offspring mice. Pharmacological blockade and genetic
   inhibition of PD-1 could weaken the effect of MSC-EVs and aggravate
   microglial dysfunction, glycolysis and autism-like behaviors in MIA
   offspring mice. Futhermore, PD-L1 deficient weakened the effect of
   MSC-EVs on neuroinflammation, glycolysis and autism-like behaviors in
   MIA offspring mice. Our research indicated the significant
   immunomodulatory capabilities of MSC-EVs, which play an important role
   in reprogramming microglial glucose metabolism and suppressing
   neuroinflammation in ASD. By activating the PD-L1/PD-1 axis and
   inhibiting the downstream ERK/HIF-1α pathway, MSC-EVs were found to
   alleviate autism-like behaviors, which revealing a novel pathological
   mechanism and offering promising therapeutic insights into ASD.

Graphical Abstract

   [46]graphic file with name 12951_2025_3250_Figa_HTML.jpg
   [47]Open in a new tab

Supplementary Information

   The online version contains supplementary material available at
   10.1186/s12951-025-03250-z.

   Keywords: Autism spectrum disorder, MSC-EVs, PD-L1/PD-1, Glycolysis,
   Neuroinflammation

Introduction

   Autism spectrum disorder (ASD) is currently one of the fastest-growing
   neurodevelopmental disorders characterized by high heritability and
   clinical heterogeneity, which leads to a huge socioeconomic burden
   [[48]1]. Core symptoms of ASD include social impairments, repetitive
   behaviors and cognitive impairments. However, the underlying
   pathogenesis of ASD is still elusive.

   Maternal immune activation (MIA) during pregnancy is a significant
   environmental risk factor to ASD [[49]2]. As a viral mimic compound,
   Poly (I:C) was capable of mimicking the pathological process of
   virus-induced MIA, stimulating the production and releasing of
   pro-inflammatory factors. It is worth noting that these cytokines can
   cross the placental barrier, and then accumulate in the fetus, thereby
   damaging the fetal brain, transmitting long-term changes in neural
   development, further leading to occurrence of ASD in offsprings
   [[50]2–[51]4]. Clinical evidence suggested that ASD individuals across
   a wide age range exhibited significant neuroinflammation in the
   postmortem brain. Additionally, neuroinflammation is a well-recognized
   pathogenesis in ASD, characterized by prominent microglia activation
   and pro-inflammatory cytokines increasing in the brain and cerebral
   spinal fluid [[52]5, [53]6].

   The abnormally activated microglia plays a significant role in
   neuroinflammation, with potential for conversion from mitochondrial
   oxidative phosphorylation to glycolysis [[54]7]. This metabolic shift
   triggers microglia to generate adenosine-triphosphate rapidly, and then
   enhancing its immune activation. Importantly, continuous glycolysis can
   produce lots of lactate which promotes plenty of pro-inflammatory
   cytokines release and accelerates microglial dysfunction in
   neurological diseases such as Alzheimer disease, ischemic stroke and
   multiple sclerosis [[55]8]. Therefore, reprogramming microglial glucose
   metabolism could serve as a promising treatment strategy to alleviate
   neuroinflammation.

   As an immune checkpoint hotspot, programmed cell death protein 1 (PD-1)
   is widely expressed on immune cells and a remarkable and high-efficient
   target in cancer immunotherapy. Upon binding with its ligand PD-L1,
   PD-1 can exert the immune inhibition capability [[56]9]. Accumulating
   evidences have implicated that PD-1 was involved in multiple
   physiological and pathological processes within the central nervous
   system (CNS) [[57]10]. Several studies have shown that PD-1 could
   express on the membrane of microglia, neuron and immune cells recruited
   from the periphery under physiological and pathological conditions
   [[58]11], highlighting a crucial role of the PD-L1/PD-1 axis in
   maintaining immune homeostasis. PD-1 pathway can promote the production
   of IL-10 and suppress inflammatory response in improving Alzheimer’s
   cognition [[59]12]. In addition, PD-L1/PD-1 axis exhibits notable
   effect on the prevention of excessive neuroinflammatory response of
   traumatic brain injury [[60]13].

   Extracellular vesicles derived from mesenchymal stem cells (MSC-EVs)
   exhibit powerful immunomodulatory property and lower immunogenicity,
   which can cross the blood-brain barrier because of the nano size
   [[61]14]. Interestingly, both MSCs and MSC-EVs highly express the
   immune checkpoint protein PD-L1. MSC could downregulate the activity of
   key enzyme HK2 by PD-L1/PD-1 axis, which further suppressed the
   glycolysis of T cell [[62]15]. Li et al. [[63]16] have found that the
   PD-L1 level was increased and then T cell activity was inhibited in
   patients after administrating the MSC-EVs in acute graft-versus-host
   disease Therefore, we aimed to elucidate the effect of MSC-EVs in
   immunomodulatory and glucose metabolic through PD-L1/PD-1 axis in ASD.

   Here, we showed there was abnormally activated in microglia and glucose
   metabolism reprogramming in MIA-ASD model. After intranasal
   administration of MSC-EVs, we observed that PD-L1 was increased in the
   brain, which could reduce microglia activation and glycolysis process,
   and then improved the autism-like behavior in MIA offspring mice. In
   addition, the neuroinflammation and microglia activation were also
   inhibited while blocking the glycolysis process by 2-DG. Moreover, both
   the in vivo and in vitro analysis highlighted the pivotal roles of
   MSC-EVs by regulating ERK/HIF-1α pathway in MIA-induced ASD mice. Taken
   together, our results elaborated the mechanism of immunomodulatory and
   glucose metabolism reprogramming underlying PD-L1/PD-1 axis, thereby
   providing new insight and strategy for precision treatment in ASD.

Methods

Cell culture

   The MSC derived from human umbilical cord used in this study were
   obtained from Nanjing Drum Tower Hospital. MSC were cultured in DMEM
   medium added 10% FBS and 1% penicillin-streptomycin. Primary mice
   microglia were isolated from postnatal (P1 to P3) mice as described in
   previous studies [[64]17]. The process involved digesting the separated
   membranes and blood vessels, followed by trypsin digestion of brain
   tissue for 5 min. The digested tissue was filtered through a 70 μm
   sieve, followed by centrifugation to collect the cells. The collected
   cells were then suspended in DMEM/F12 medium added FBS and
   penicillin-streptomycin. When the cells reached 90% confluence,
   microglia were harvested by shaking the flasks at 225 rpm for 5 h. The
   microglia were collected and seeded into 6-well plates.

   MSC cells were infected with a control lentivirus (MSC-EVs-Ctrl) or
   shCd274 lentivirus (MSC-EVs-shCd274), and the EVs were extracted using
   differential ultracentrifugation method.

Cell treatment

   To explore the effect of MSC-EVs on microglia, PBS or MSC-EVs were
   co-cultured with microglia of MIA offspring mice with a dose of 50 µg,
   and the following groups were performed: C57BL/6J + PBS,
   C57BL/6J + MSC-EVs, MIA + PBS and MIA + MSC-EVs. For PD-1 inhibition
   studies, microglia of MIA offspring mice received PD-1 Ab (RMP1-14, 1
   ug/10^6 cells, HY-[65]P99144, MedChemExpress) or control IgG for 24 h
   before MSC-EVs treatment. For the glycolysis inhibition studies,
   microglia were treated with 2-DG (5 mM, A07816, Innochem) for 24 h. To
   detected the effect of MSC-EVs on ERK activation, we applied the ERK
   activator honokiol (20 µM, HY-N0003, MedChemExpress) to the treatment
   with MSC-EVs. We also applied the ERK inhibitor SL327 (10 µM, HY-15437,
   MedChemExpress) to treat with MSC-EVs and PD-1 Ab.

   To explore the effect of PD-L1 of MSC-EVs on microglia in vitro, PBS,
   MSC-EVs-Ctrl or MSC-EVs-shCd274 were co-cultured with MIA primary
   microglia with a dose of 50 µg for 48 h, and the following groups were
   performed: PBS, MSC-EVs-Ctrl, MSC-EVs-shCd274.

Extraction and identification of MSC-EVs

   MSC-EVs were separated from the supernatant of MSC through a series of
   centrifugation steps.When MSC cells reached the 90% confluence, they
   were cultured in serum-free DMEM. After 48 h, the supernatants were
   collected to extract EVs. Initially, the supernatant was centrifuged at
   300 × g for 10 min, 2000 × g for 10 min, 10,000 × g for 30 min and
   10,000 × g for 60 min at 4℃, followed by ultracentrifugation at 100,000
   × g for 60 min using a CP100NX ultracentrifuge (Hitachi, Japan). The
   resulting pellet was resuspended in PBS and ultracentrifuged at 100,000
   × g for 60 min again. Finally, the pellet was resuspended using PBS for
   further analysis. The EVs were visualized and quantified using the
   Nanometer particle size meter (Zetasizer Nano ZS90, Malvern, UK). The
   transmission electron microscopy (TEM) (H-7650, Hitachi, Japan) was
   used to observe the morphology of the EVs. Western blotting analysis
   was conducted to confirm the EVs markers CD9, TSG101 and CD63.

Staining of MSC-EVs

   The extracellular vesicles (EVs) were stained and purified according to
   previously established protocols [[66]17]. The EVs protein
   concentration was adjusted to 100 µg/ml and co-incubated with Dil
   (C1036, Beyotime, China) for 15 min. Any excess dye was removed by
   incubating with 10% FBS to eliminate EVs. Subsequently, MSC-EVs samples
   were diluted using PBS and subjected to ultracentrifugation at 100,000
   × g for 1 h. The resulting pellet was resuspended in PBS for the
   subsequent experiments.

Uptake of MSC-EVs into the primary microglia

   The microglia (2 × 10^4 cells) were seeded in 35 mm dishes. After 24 h,
   50 µg of Dil-labeled MSC-EVs were administered to the cultured cells.
   Following the incubation at 37 °C with 5% CO[2]. Following incubation,
   the cells underwent three washes with PBS to eliminate the excess EVs.
   Then, the cells were stained with DAPI (Beyotime, China). The images
   were captured with Laser scanning confocal microscopes LSM 780 (Carl
   Zeiss, Germany).

Animal

   C57BL/6J mice were procured from the Beijing Vital River Laboratory
   Animal Technology Co., Ltd. The mice were randomly allocated into
   distinct groups and maintained in a controlled environment with a
   12-hour light/dark cycle, with a humidity of 55% ± 5%, a temperature of
   21 °C ± 1 °C and enough food and water.

MIA mouse modeling

   Pregnant C57BL/6J mice were administered intraperitoneally (i.p.) at a
   dose of 20 mg/kg poly(I:C) (Sigma-Aldrich) on embryonic day 12.5
   (E12.5) for experimental purposes. The control group mice received an
   equivalent volume of PBS without any active compounds, ensuring a
   comparable handling procedure. Indeed, the number of male ASD patients
   is significantly higher than that of females [[67]18, [68]19].
   Noteworthy, the similar gender differences in autism-like behaviors
   have been observed in the MIA offspring mice, as confirmed in previous
   studies [[69]20, [70]21]. Male offspring mice exhibit more pronounced
   autism-like behaviors compared to female offspring mice. Therefore, we
   chose male offspring mice for this study. After observing substantial
   gender disparities in the ASD-related traits of pups resulting from
   poly(I:C)-MIA offspring mice, we elected to focus solely on male
   offspring in our subsequent experiments.

Animal treatment

   To explore the effect of MSC-EVs on ASD, each mouse was administration
   intranasally with 100 µg of MSC-EVs or equal volume PBS every day for 7
   days, and the following groups were performed: C57BL/6J + PBS,
   C57BL/6J + MSC-EVs, MIA + PBS and MIA + MSC-EVs.

   PD-1 Ab (RMP1-14) or control IgG were injected into MIA offspring mice
   using brain stereotaxic apparatus. Then, MSC-EVs were delivered into
   brain via intranasal administration daily for 7 days. Next, behavioral
   tests and molecular experiment were performed to evaluate the impact of
   PD-1 Ab on MIA offspring mice.

   In brief, the anesthesia process was initiated in an induction chamber
   with a mixture of isoflurane and oxygen, initially administering 5 L
   per minute for induction, followed by approximately 3 L per minute to
   maintain anesthesia. Fur was removed from the skull’s surface using
   clippers, and the area was disinfected using iodophor.

   Place the mouse securely on a stereotactic stand and make a midline
   incision to expose the bregma. A burr hole was drilled precisely above
   the PFC, coordinates: AP: 1.90 mm from Bregma, DV: ±0.30 mm, ML:
   2.26 mm) using a high-precision Kopf micro-manipulator. PD-1 antibody
   (Ab) or IgG was administered slowly, at a rate of 60 nanoliters per
   minute, employing a pico-spritzer air puffer system. After the targeted
   volume of the reagent had been administered, the pipette was maintained
   in its position for a duration of 5 min before being gradually
   withdrawn from the brain. For glycolysis studies, male MIA offspring
   mice were intraperitoneal (i.p.) injection of 2-DG (250 mg/kg, A07816,
   Innochem) or PBS for 7 days.

   The pAAV-Iba-1-eGFP-Pdcd1-shRNA (AAV-Iba-1-shPdcd1, serotype MG1.2)
   (GenePharma, Shanghai, China) was used to target Iba-1^+ cells and
   interrupt PD-1 expression in microglia, and pAAV-Iba-1-eGFP-MCS used as
   a control vector (AAV-Ctrl). Briefly, the MIA offspring mice were
   placed in a stereotaxic apparatus, and was injected the viral vector
   (5 × 10^9 vg; 500 nl) into the PFC (coordinates: AP:1.90 mm,
   DV:0.30 mm, ML:2.26 mm) by a 10 µl syringe (Hamilton) connecting to a
   micro-syringe pump controller at a speed of 60 nl/min. Three weeks
   later, MSC-EVs were delivered into brain by intranasal administration
   once a day for one week. Next, behavioral tests and molecular
   experiment were performed to evaluate the impact of AAV-shPdcd1 on MIA
   offspring mice.

   To explore the effect of PD-L1 of MSC-EVs on neuroinflammation,
   glycolysis and autism-like behaviors in vivo, 100 µg MSC-EVs-shCd274,
   MSC-EVs-Ctrl or equal volume PBS was respectively administrated
   intranasally into MIA offspring mice daily for 7 days.

Behavioral experiments

   All behavioral assessments were conducted in a quiet setting, with a
   consistent scorer. The mice have adapted to the testing room for a
   minimum of three hours prior to the commencement of each behavioral
   examination to ensure minimal stress and accurate scoring. Animals
   completed the three-chamber test, Open field test (OFT), Elevated plus
   maze (EPM) test, Novel object recognition test (NOR), Grooming test and
   Marble burying test (MBT) as described in the Supplementary
   Information.

Cell viability

   In order to evaluate the cytotoxicity of MSC-EVs to primary mice
   microglia, the cells were cultured in a 96-well plate (1 × 10^4
   cells/well). After 6 h of culture, the cell survival level was analyzed
   using the CCK8 assay on a microplate reader acc ording to the
   instructions (HY-K0301, MedChemExpress).

qRT-PCR and other experiments

   The qRT-PCR, Western blot analysis, immunofluorescence and
   enzyme-linked immunosorbent assay (ELISA) methodologies were detailed
   in Supplementary Information. The primers and antibodies use in this
   study were showed in the Supplementary Tables [71]1 and Table [72]2,
   respectively.

Lactate level measurement

   Lactate level was measured in mice serum and microglia culture
   supernatants via the lactate testing kit (A019-2-1, Nanjing Jiancheng
   Bioengineering Institute) following the manufacturer’s protocol.

Data collection and preprocessing

   The public datasets include the single-cell sequencing data of
   postmortem PFC tissue from 10 ASD patients and 7 normal (PRJNA434002)
   [[73]22], microarray dataset of PFC tissue mRNA from 6 MIA offspring
   mice and 6 control mice. ([74]GSE77972) [[75]23] and the microarray
   dataset of postmortem FC tissue from 16 ASD patients and 16 normal
   developing children’s tissue ([76]GSE28521) [[77]24].

   For the chip data downloaded from the GEO database, probe-to-gene
   mapping was performed using the annotation files provided by the
   corresponding detection platform for each dataset. If a probe matches
   multiple genes simultaneously, the probe is removed; if multiple probes
   match the same gene, the mean detection abundance of these probes is
   taken as the final expression value for that gene.

   The raw mRNA-seq expression profile data underwent conversion to
   Fragments Per Kilobase of transcript per Million Mapped Reads (FPKM)
   values, a measure of gene expression intensity. The FPKM values were
   then subjected to a log2 transformation (by adding 1 to the expression
   value) for standardization, serving as the final gene expression
   measure.

   The data source is publicly available from the Sequence Read Archive
   (SRA, accession number PRJNA434002), with the sample were characterized
   as male and originating from the PFC. For the single-cell sequencing
   raw data (PRJNA434002), Cell Ranger software was employed for
   preprocessing, including FASTQ file generation, read alignment, and
   Unique Molecular Identifier (UMI) quantification. Cell Ranger was
   executed using default parameters, and the ENSEMBL GRCh38 reference
   genome was utilized to ensure comprehensive capture of intronic reads
   from the nuclear-enriched pre-mRNA transcriptome. The Seurat package in
   R was employed to generate Seurat objects for each specific cell type,
   housing scRNA-seq gene expression matrices. Batch correction and sample
   integration were subsequently executed through the IntegrateData
   function. To ensure data quality, cells were filtered according to the
   gene expression, retaining only those with a count of 200 to 4000 genes
   and less than 20% mitochondrial gene content. Normalization and scaling
   of the remaining scRNA-seq data were performed using NormalizeData and
   ScaleData. Principal component analysis (PCA) was conducted using
   RunPCA, and a Uniform Manifold Approximation and Projection (UMAP) was
   employed to condense the top principal components. Manual annotation of
   known gene markers was assigned to cell types. The Idents and DimPlot
   tools facilitated the annotation and visualization of both general cell
   types and their subtypes within the dataset.

   Scoring of the inflammation for microglia was performed using a scoring
   system based on the gene sets related to inflammation were obtained
   using the “HALLMARK INFLAMMATORY RESPONSE” of Molecular Signatures
   Database (MSIGDB) ([78]https://www.gsea-msigdb.org/gsea/msigdb).

Functional enrichment analysis

   Upload the differentially expressed genes to the DAVID software
   ([79]https://david.ncifcrf.gov/summary.jsp) [[80]25], select
   OFFICIAL_GENE_SYMBOL as the gene identifier, and set the species to
   Homo sapiens or Mus musculus. Then conduct KEGG pathway enrichment
   analysis. Using P < 0.05 as a threshold to screen KEGG pathway entries
   related to ASD as the main gene functional enrichment process and
   signaling pathways associated with ASD, predicting the pathological
   mechanisms of ASD. Subsequently, we conducted Gene Set Enrichment
   Analysis (GSEA) analysis [[81]26] related pathway enriched by KEGG
   using the cluster Profiler package.

Statistical analysis

   The GraphPad Prism 8.0.2 software ([82]https://www.graphpad.com) was
   used to conduct the statistical analyses. For pairwise comparisons
   between two groups, an unpaired Student’s t-test was employed. To
   evaluate multiple groups, the one/two-way ANOVA test were used. All
   data were analysed as mean ± SD, and statistical significance was
   showed as ^*P < 0.05, ^**P < 0.01 and ^***P < 0.001. Detailed
   information on the study design, replication numbers, and statistical
   tests employed are showed in figure legends.

Results

Construction of MIA model in pregnant and MIA offspring mice

   C57BL/6J mice were injected (i.p.) with 20 mg/kg of poly(I:C) at E12.5
   to construct the MIA model in pregnant stage (Fig. [83]1A). Then, after
   3 h of poly(I:C) injection, the level of IL-6 and TNF-α in the
   placenta, serum, spleen and mesenteric lymph nodes of MIA pregnant mice
   were all significantly elevated, whereas no differences were detected
   in IL-1β, IL-4 and IL-10 (Fig. [84]1B-E). In summary, these results
   showed that MIA pregnant mice experienced acute inflammatory response.

Fig. 1.

   [85]Fig. 1
   [86]Open in a new tab

   MIA offspring mice showed typical autism-like phenotypes. (A) Schematic
   diagram illustrating the experimental procedure of MIA offspring mice.
   (B-E) The inflammatory factors detection in placenta, spleen, serum and
   lymph node by qPCR and ELISA between PBS and MIA maternal groups. n=6.
   Two-way ANOVA. (F) Trajectory route and statistical chart of
   sociability and social preference in the three-chamber test of PBS and
   MIA offspring mice. n=8. Two-way ANOVA. (G) Trajectory route and
   statistical chart of entries times in center zone and time in center
   zone in the OFT of PBS and MIA offspring mice. n=8. Student’s t-test.
   (H) Trajectory route and statistical chart of entries in open arms and
   time in open arms in EPM test of PBS and MIA offspring mice. n=8.
   Student’s t-test. (I) Statistical chart of number in marbles buried of
   PBS and MIA offspring mice. n=8. Student’s t-test. (J) Statistical
   diagram of number of grooming times and duration grooming time of PBS
   and MIA offspring mice. n=8. Student’s t-test. (K) Statistical diagram
   of exploration time for each object in the NOR test of MIA offspring
   mice. n=8. Two-way ANOVA. ns P≥0.05, *P<0.05, **P<0.01, ***P<0.001

   When the MIA offspring mice grew to 8 weeks, we performed multiple
   behavioral tests to estimate the autism-like behaviors of mice. In the
   three-chamber test, social behaviors of mice were evaluated. During the
   first phase, PBS offspring mice prefered interacting with the stranger
   mouse 1 (S1) but not the empty cage. In contrast, MIA offspring mice
   did not exhibit a significant difference in the duration of exploring
   the empty cage compared to S1, suggesting a compromised social ability.
   In the second phase, PBS offspring mice spent longer time investigating
   the new stranger mouse 2 (S2) than S1, while MIA offspring mice spent
   no difference in contact time within S1 and S2, pointing to an impaired
   social preference pattern (Fig. [87]1F). Then, the OFT and EPM were
   employed to assess anxiety-related behaviors in mice. In the OFT test,
   a significant reduction in entry times in center zone and time spent in
   center zone were observed in MIA offspring mice compared to the PBS
   group (Fig. [88]1G). During the EPM assessment, the MIA offspring mice
   demonstrated less entries in open arms and shorter time in open arms
   (Fig. [89]1H), longer time in closed arms, while less zone transition
   times (Figure [90]S1). However, there were no significant difference in
   mean speed and total moving distance between the two groups in OFT and
   EPM (Figure [91]S1). These results indicated that MIA offspring mice
   spent less time exploring the central area or open arms due to anxiety
   and fear in unfamiliar environment. And this phenomenon was not caused
   by their basic motor ability.

   Then, we conducted an assessment of stereotyped behaviors through the
   MBT and grooming test. In the MBT, MIA offspring mice demonstrated a
   higher tendency, burying an average of approximately nine marbles, in
   contrast to the PBS group mice, who typically buried around four
   marbles (Fig. [92]1I). And the results of grooming test revealed that
   MIA offspring mice exhibited a significantly increased frequency and
   time of self-grooming compared to the PBS group (Fig. [93]1J). These
   results suggested that MIA offspring mice showed presence of repetitive
   stereotyped behaviors.

   To evaluate cognitive and memory performance, we employed the NOR test.
   Under normal circumstance, mice were exhibited a natural inclination
   for exploration of novel items, which shown as increased interest in
   touching or sniffing unfamiliar objects within a space. In the NOR
   test, we observed that during the initial familiarization phase, both
   MIA and PBS offspring mice displayed similar exploration duration for
   two identical objects, labeled a and b. However, during the testing
   phase, PBS offspring mice demonstrated a significantly higher level of
   curiosity towards a new object, c, which had replaced object b, while
   no difference in sniffing time between a and c of MIA offspring mice
   group. This difference in exploration time suggested that MIA offspring
   mice exhibited impaired cognitive and memory function (Fig. [94]1K). In
   summary, MIA offspring mice showed decreased social ability, impaired
   social preference, higher anxiety level, presence of repetitive
   stereotyped behaviors, and decreased cognitive ability. Therefore, MIA
   offspring mice could serve as a reliable model for subsequent ASD
   research.

The MIA offspring mice showed increased neuroinflammation

   Neuroinflammation is characterized by an inflammatory reaction in
   either the CNS or peripheral nervous system (PNS). When neural tissues
   are infected, injured, or subjected to other stimuli, immune cells and
   inflammatory factors gather in the affected area, leading to the
   occurrence of neuroinflammation. Firstly, we detected significantly
   elevated level of the inflammatory factors IL-6 and TNF-α in the PFC
   tissue in fetal brain, P0 (newborn), 4-week-old, 8-week-old and serum
   in 8-week-old MIA offspring mice (Fig. [95]2A-B), indicating
   significant neuroinflammatory response in MIA offspring mice and exist
   for a long time. Notably, the PFC plays a pivotal role in regulating
   behaviors, including social interaction, cognitive ability,
   decision-making and emotional regulation. Furthermore, impairments in
   the PFC have been closely linked to the onset and progression of ASD.
   Activated microglia stimulate immune response by secreting large
   amounts of cytokines and neurotoxic factors, thereby triggering
   neuroinflammatory response. The level of Iba-1 in the PFC tissue slices
   of MIA offspring mice was significantly higher than that in control
   mice (Fig. [96]2C). To investigate the correlation between
   neuroinflammation and ASD individuals, we performed bioinformatics
   analysis using the publicly available [97]GSE28521 dataset from the GEO
   database. This dataset contains transcriptomic data related to the PFC
   of individuals with ASD. Gene set enrichment analysis (GSEA) results
   showed that the inflammatory response, IL-6 production, tumor necrosis
   factor (TNF) production were up-regulated according to the expressed
   profile of [98]GSE28521 dataset (Fig. [99]2D). In this study, we also
   observed that the level of IL-6 and TNF-α was increased in primary
   microglia from MIA offspring mice (Fig. [100]2E, Figure S2).
   Furthermore, we performed single-cell sequencing analysis of the public
   PRJNA434002 dataset, which includes 12 cell types. The rain-cloud plot
   illustrated that the inflammation score of microglia in the PFC of ASD
   patients was higher compared to that of controls (Fig. [101]2F-G). GSEA
   indicated that signaling pathways associated with the immune response,
   inflammatory response, regulation of immune response, and cellular
   response to tumor necrosis factor were activated in ASD patients, as
   evidenced by the gene expression profiles of microglia in the
   PRJNA434002 dataset (Fig. [102]2H-I).

Fig. 2.

   [103]Fig. 2
   [104]Open in a new tab

   The MIA offspring mice showed the activation of neuroinflammation and
   microglia. (A-B) IL-6 and TNF-α level detection of PFC and serum by
   qPCR and ELISA respectively between PBS and MIA offspring mice. n=6.
   Two-way ANOVA. (C) Immunostaining analysis of Iba-1^+ of PFC in PBS and
   MIA offspring mice. Scale bar = 50 μm. Student’s t-test. (D) The
   results of GSEA enrichment in ASD and controls frontal cortex
   ([105]GSE28521). (E) The IL-6 and TNF-α level assessment in microglia
   between PBS and MIA offspring mice. n=3. Two-way ANOVA. (F-I) Analysis
   of scRNA-seq dataset (PRJNA434002) in ASD patients and controls PFC:
   (F) tSNE visualization revealed 12 cell clusters; (G) Raincloud plot
   showed the inflammation score in microglia of ASD patients and
   controls; (H-I) GSEA enrichment analysis of 4 pathways according to the
   expressed profile in microglia cluster. ns P≥0.05, *P<0.05, **P<0.01,
   ***P<0.001

MSC-EVs exhibit high expression of PD-L1, while MIA mouse brain tissue and
microglia show high level of PD-1

   The timelines and protocols were employed for isolation and
   characterization of MSC-EVs (Fig. [106]3A). NanoSight analysis results
   showed the size of MSC-EVs was ~ 180 nm in diameter (Fig. [107]3B).
   Furthermore, TEM demonstrated EVs with characteristic morphology
   enclosed by double membranesand exhibits a biconcave disk-like
   structure typical of EVs (Fig. [108]3C). Moreover, western blotting
   results showed the expression of the EVs-specific markers CD9, CD63 and
   TSG101 in the EVs formulation, but the cytoplasmic protein Calnexin was
   not expressed (Fig. [109]3D). The PD-L1 expression of surface of MSC
   and MSC-EVs were both significantly higher than the control group
   (Fig. [110]3E-F).

Fig. 3.

   [111]Fig. 3
   [112]Open in a new tab

   High expression of PD-L1 in MSC-EVs, and high expression of PD-1 in MIA
   offspring mice. (A) Schematic diagram illustrating the
   extraction–separation process of MSC-EVs. (B-D) Characterization of
   MSC-EVs: (B) Size distribution of MSC-EVs measured using a nanoparticle
   size meter; (C) TEM of MSCs-EVs isolated from MSCs. Scale bar = 200 nm;
   (D) Western blot analysis of CD9, CD63, TSG101 and calnexin in MSC-EVs.
   (E-F) Western blot analysis of the expression of PD-L1 in MSCs and
   MSC-EVs. n=3. Student’s t-test. (G) Western blot analysis of PD-1 and
   PD-L1 of PFC in PBS and MIA offspring mice. n=6. Two-way ANOVA. (H)
   Immunostaining analysis of co-localization of PD-1^+ and Iba-1^+ of PFC
   in MIA offspring mice. Scale bar = 50 μm. Student’s t-test. (I) Western
   blot analysis of PD-1 in microglia in PBS and MIA offspring mice. n=3.
   Student’s t-test. (J) Immunostaining analysis of co-localization of
   PD-1^+ and Iba-1^+ in microglia in PFC MIA offspring mice. Scale bar =
   20 μm. Student’s t-test. (K) KEGG analysis of the differentially
   expressed genes of PFC in MIA offspring mice and controls
   ([113]GSE77972). (L) KEGG analysis of the differentially expressed
   genes of FC in ASD patients and controls ([114]GSE28521).ns P≥0.05,
   *P<0.05, **P<0.01, ***P<0.001

   To further explore the effects of MSC-EVs on MIA offspring mice in
   vivo, we detected a significant increase in PD-1 level in the PFC of
   8-week-old MIA offspring mice, while there was no significant
   difference in PD-L1 level between the two groups (Fig. [115]3G, Figure
   S3). Double immunofluorescence staining results showed that the
   expression level of PD-1 on the surface of microglia in the PFC of MIA
   offspring mice were significantly higher than those in the control
   group mice (Fig. [116]3H, Figure S4). Moreover, the expression level of
   PD-1 in primary microglia from MIA offspring mice was significantly
   upregulated (Fig. [117]3I-J). These results indicated that when MSC-EVs
   intervened, their surface PD-L1 could bind to PD-1 on the surface of
   activated microglia in the PFC of MIA offspring mice, thereby exerting
   immunomodulatory and anti-inflammatory effects.

   In order to verify this, we conducted a biological functional analysis
   of ASD patients and MIA offspring mice using the public database. KEGG
   functional enrichment analysis based on the frontal cortex (FC) dataset
   [118]GSE77972 of MIA offspring mice revealed that differential genes
   were significantly enriched in pathways such as PD-L1/PD-1 pathway
   (Fig. [119]3K). Additionally, KEGG functional enrichment analysis of
   differential genes in the PFC dataset [120]GSE28521 of ASD patients
   showed potential abnormalities in pathways including PD-L1/PD-1 and
   TNF-α pathways (Fig. [121]3L). The collective evidence of substantial
   enrichment of the PD-L1/PD-1 pathway in both MIA offspring mice and
   patients with ASD also suggests a potential role for immune dysfunction
   in ASD, mediated through this axis.

The activation level of the ERK/HIF-1α pathway and Glycolysis were elevated
in the MIA offspring mice

   Increased glycolysis is a necessary condition for microglia activation
   and neuroinflammatory response. Compared to the control group, the
   expression level of HK2 and LDHA in the PFC tissue of MIA offspring
   mice as well as lactate level in the PFC and serum were elevated
   (Fig. [122]4A-C). And the expression level of HK2, LDHA and lactate
   level in the primary microglia of MIA offspring mice were all increased
   as well (Fig. [123]4I-J), indicating the glycolysis level was increased
   in the brain of MIA offspring mice.

Fig. 4.

   [124]Fig. 4
   [125]Open in a new tab

   The activation level of the ERK/HIF-1α pathway and glycolysis were
   elevated in the MIA offspring mice. (A-B) The expression of glycolysis
   key genes of PFC by qPCR (n=8) and western blot (n=6) between the PBS
   and MIA offspring mice. Two-way ANOVA. (C) Lactate level in the PFC and
   serum between the PBS and MIA offspring mice. n=6. Student’s t-test.
   (D) Schematic diagram illustrating the experimental procedure of 2-DG
   administration. (E) Western blot analysis of HK2 and LDHA in PFC in MIA
   offspring mice after 2-DG administration. n=6. Two-way ANOVA. (F)
   Lactate level of the PFC and serum in MIA offspring mice after 2-DG
   administration. n=6. Student’s t-test. (G) IL-6 and TNF-α level of PFC
   and serum in MIA offspring mice after 2-DG administration. n=6. Two-way
   ANOVA. (H) Immunostaining analysis of Iba-1^+ of PFC in MIA offspring
   mice after 2-DG administration. Scale bar = 50 μm. Student’s t-test.
   (I) Western blot analysis of HK2 and LDHA in microglia between the PBS
   and MIA offspring mice. n=3. Two-way ANOVA. (J) Lactate level of
   microglia supernatant between the PBS and MIA offspring mice. n=3.
   Student’s t-test. (K) Schematic diagram of 2-DG intervention for
   primary microglia from MIA offspring mice. (L-M) Western blot analysis
   of HK2 and LDHA in primary microglia with 2-DG intervention. n=3.
   Two-way ANOVA. (N) Lactate level of primary microglia supernatant with
   2-DG intervention. n=3. Student’s t-test. (O) IL-6 and TNF-α level of
   primary microglia with 2-DG intervention. n=3. Two-way ANOVA. (P-S)
   Western blot analysis of p-ERK/ERK and HIF-1α of PFC (n=6) and primary
   microglia (n=3) between PBS and MIA offspring mice. Student’s t-test.
   ns P≥0.05, *P<0.05, **P<0.01, ***P<0.001

   Intervention with the glycolysis inhibitor 2-deoxyglucose-6-phosphate
   (2-DG) in MIA offspring mice for one-week resulted in decreased
   expression of HK2 and LDHA compared to the PBS group, along with a
   reduction in the production and secretion of IL-6 and TNF-α
   (Fig. [126]4D-H). Similar effects of 2-DG on glycolysis and
   inflammation were also observed in microglia from MIA offspring mice
   (Fig. [127]4K-O), suggesting that inhibiting glycolysis could reduce
   microglial activation.

   Studies have shown that the activation of PD-L1/PD-1 axis could inhibit
   the phosphorylation of ERK in microglia [[128]27]. And it has been
   demonstrated that HIF-1α is a key transcription factor for
   glycolysis-related genes, and upregulation of HIF-1α could promote an
   increase in glycolysis. Compared to control mice, the expression level
   of p-ERK//ERK and HIF-1α were upregulated, manifesting an elevated
   activation level of the ERK/HIF-1α pathway in the PFC and primary
   microglia of MIA offspring mice (Fig. [129]4P-S). These results
   suggested that the activation of the ERK/HIF-1α pathway in microglia of
   MIA offspring mice may promote intracellular glycolysis, which could be
   an important factor in microglia activation.

   Moreover, the KEGG functional enrichment analysis of differential genes
   in the PFC dataset [130]GSE28521 of ASD patients showed potential
   abnormalities in pathways including HIF-1α and TNF-α pathways
   (Fig. [131]3L). Additionally, gene set enrichment analysis (GSEA) of
   the ERK1 and ERK2 cascade, postive regulation of ERK1 and ERK2 cascade,
   regulation of ERK1 and ERK2 cascade and HIF-1 signaling pathway were
   all upregulated according to the expressed profile of [132]GSE28521
   dataset (Figure S5).

MSC-EVs could alleviate autism-like behaviors of MIA offspring mice

   We administered 100 µg of MSC-EVs intranasally to each mouse daily for
   one week and conducted autism-related behavioral tests of mice after
   administration (Fig. [133]5A). Following the MSC-EVs intervention, the
   three-chamber test indicated an improvement in social ability and
   social preference in MIA offspring mice (Fig. [134]5B). In the OFT, the
   number of entries into the central zone and the time spent in the
   central zone were both increased in mice after MSC-EVs intervention
   (Fig. [135]5C, Figure S6). Results from the EPM test showed that
   following MSC-EVs intervention, MIA offspring mice exhibited increased
   activity distance, entries into the open arms and time spent in the
   open arms compared to the control group, while the time spent in the
   closed arms was significantly decreased (Fig. [136]5D, Figure S6).
   These results indicated that MSC-EVs intervention can improve anxiety
   level in MIA offspring mice. The MBT indicated a significant decrease
   in the number of marbles buried by mice after MSC-EVs intervention
   (Fig. [137]5E). The self-grooming test showed a significant decrease in
   both the frequency and duration of self-grooming in mice after MSC-EVs
   intervention, indicating an improvement in stereotyped behavior as well
   (Fig. [138]5F). In the NOR test, mice exhibited prolonged contact time
   with the novel object after MSC-EVs intervention, suggesting an
   improvement in cognitive ability (Fig. [139]5G).

Fig. 5.

   [140]Fig. 5
   [141]Open in a new tab

   MSC-EVs alleviated neuroinflammation and autism-like behaviors of MIA
   offspring mice. (A) Schematic diagram illustrating the experimental
   procedure of MSC-EVs administration. (B-G) Autism-behaviors analysis
   between PBS and MIA offspring mice after MSC-EVs administration. (B)
   Statistical chart of sociability and social preference in the
   three-chamber test. Two-way ANOVA; (C) Statistical chart of entries
   times in center zone and time in center zone in the OFT. One-way ANOVA;
   (D) Statistical chart of entries in open arms and time in open arms in
   EPM test. One-way ANOVA; (E) Statistical chart of number in marbles
   buried. One-way ANOVA; (F) Statistical chart of grooming times and
   duration of grooming. One-way ANOVA; (G) Statistical chart of
   exploration time for each object in the NOR test. Two-way ANOVA. n=8.
   (H) Co-localization of Dil-labeled MSC-EVs of PFC in MIA offspring
   mice. Scale bar = 20 μm. (I) Western blot analysis of PD-1 and PD-L1 of
   PFC in MIA offspring mice after MSC-EVs administration. n=6. Two-way
   ANOVA. (J) IL-6 and TNF-α level of PFC and serum between PBS and MIA
   offspring mice after MSC-EVs administration. n=6. Two-way ANOVA. (K)
   Immunostaining analysis of Iba-1^+ of PFC between PBS and MIA offspring
   mice after MSC-EVs administration. Scale bar = 50 μm. One-way ANOVA.
   (L) Co-localization of primary microglia with Dil-labeled MSC-EVs.
   Scale bar = 20 μm. (M) Western blot analysis of PD-1 and PD-L1 of
   primary microglia in MIA offspring mice with MSC-EVs intervention. n=3.
   Two-way ANOVA. (N) IL-6 and TNF-α level of primary microglia in MIA
   offspring mice with MSC-EVs intervention. n=3. Two-way ANOVA. ns
   P≥0.05, *P<0.05, **P<0.01, ***P<0.001

MSC-EVs could inhibit the activation of microglial and Glycolysis in MIA
offspring mice

   We administered DiI-labeled 100 µg MSC-EVs intranasally into the brains
   of MIA offspring mice. After 6 h, the DiI-labeled MSC-EVs were
   significantly enriched in the PFC of the mice, indicating that MSC-EVs
   could be accumulated in brain tissue via intranasal administration
   (Fig. [142]5H, Figure S7).

   The MSC-EVs could be successfully delivered to the mouse brain through
   intranasal administration, increasing the expression level of PD-L1 in
   the CNS, but without affecting PD-1 expression (Fig. [143]5I).
   Moreover, after one-week treatment of MSC-EVs, the IL-6 and TNF-α level
   in the PFC and serum were significantly reduced (Fig. [144]5J). As
   shown in Fig. [145]5K, in MIA + MSC-EVs group, the level of Iba-1^+,
   the size of the microglial soma, the number of branches, and the total
   branch length of microglia were all significantly decreased compared to
   MIA + PBS group. The results above indicated that MSC-EVs could inhibit
   neuroinflammatory response in the PFC of MIA offspring mice.

   Next, we co-cultured MSC-EVs labeled with DiI with primary microglia
   from the PFC of MIA offspring mice to explore the immunomodulatory
   effects and mechanisms of MSC-EVs on microglia. Initially, DiI-labeled
   MSC-EVs were co-cultured with primary microglia, and MSC-EVs labeled
   with orange fluorescence were successfully internalized by primary
   microglia (Fig. [146]5L).

   After 24 h of co-culture, as shown in Fig. [147]5M, compared to the PBS
   group, the level of PD-L1 were significantly increased in the MSC-EVs
   group, while the level of PD-1 showed no significant change. After 48 h
   of MSC-EVs intervention, the expression level of IL-6 and TNF-α in
   control group microglia remained unaffected, whereas MSC-EVs
   significantly reduced the expression of IL-6 and TNF-α in MIA offspring
   mice (Fig. [148]5N).

   Following the administration of MSC-EVs, the activity of the ERK/HIF-1α
   pathway was diminished. In the MSC-EVs treated group, the expression
   level of HK2 and LDHA in the PFC tissue of MIA offspring mice were
   significantly decreased, as were the level of lactate in both the PFC
   and serum (Fig. [149]6A-D). These findings suggested that MSC-EVs could
   effectively inhibit glycolysis in the PFC of MIA offspring mice.

Fig. 6.

   [150]Fig. 6
   [151]Open in a new tab

   MSC-EV repressed the glycolysis process in MIA offspring mice. (A-C)
   Western blot analysis of p-ERK/ERK, HIF-1α, HK2 and LDHA of PFC between
   PBS and MIA offspring mice after MSC-EVs administration. n=6. One-way
   ANOVA, Two-way ANOVA. (D) Lactate level of PFC and serum between PBS
   and MIA offspring mice after MSC-EVs administration. n=6. One-way
   ANOVA. (E-G) Western blot analysis of p-ERK/ERK, HIF-1α, HK2 and LDHA
   in primary microglia with MSC-EVs intervention. n=3.One-way ANOVA,
   Two-way ANOVA. (H) Lactate level in primary microglia supernatant with
   MSC-EVs intervention. n=3. One-way ANOVA. ns P≥0.05, *P<0.05, **P<0.01,
   ***P<0.001

   Following MSC-EVs administration, there was a significant decrease in
   the expression level of HK2 and LDHA in the PFC microglia of MIA
   offspring mice, as well as in lactate level in microglia culture
   supernatant. Moreover, MSC-EVs could inhibit the activity of the
   ERK/HIF-1α pathway (Fig. [152]6E-H), suggesting that MSC-EVs may
   suppress the activation of the ERK/HIF-1α pathway in microglia, thereby
   reduce cellular glycolysis level, subsequently inhibit the production
   of inflammatory factors.

MSC-EVs inhibit Glycolysis and neuroinflammation via the PD-L1/PD-1 pathway

   To investigate the effect of PD-L1 carried by MSC-EVs on MIA offspring
   mice, PD-1 Ab was injected into the PFC of MIA offspring mice using
   brain stereotactic apparatus. Then, MSC-EVs were delivered into brain
   by intranasal administration daily for 7 days (Fig. [153]7A).

Fig. 7.

   [154]Fig. 7
   [155]Open in a new tab

   Pharmacological blockade of PD-1 weakened the effect of MSC-EVs on
   improving autism-like behaviors in MIA offspring mice. (A) Schematic
   diagram illustrating the experiment procedure of MIA offspring mice
   after PD-1 Ab alone or PD-1 Ab + MSC-EVs administration. (B-G)
   Autism-behaviors analysis of MIA offspring mice after PD-1 Ab alone or
   PD-1 Ab + MSC-EVs administration. (B) Statistical chart of sociability
   and social preference in the three-chamber test. Two-way ANOVA; (C)
   Statistical chart of entries times in center zone and time in center
   zone in the OFT. One-way ANOVA; (D) Statistical chart of entries in
   open arms and time in open arms in EPM test. One-way ANOVA; (E)
   Statistical chart of number in marbles buried. One-way ANOVA; (F)
   Statistical chart of grooming times and duration grooming. One-way
   ANOVA; (G) Statistical diagram of exploration time for each object in
   the NOR test. Two-way ANOVA. n=8. ns P≥0.05, *P<0.05, **P<0.01,
   ***P<0.001

   In PD-1 Ab group, In the EPM test, duration in the open arms was
   significantly decreased, while, the duration in the closed arms was
   increased; and the number in marbles buried was increased; while social
   ability, social preference and cognitive level were not significant
   difference comparing with the IgG control group (Fig. [156]7B-G, Figure
   S8). Moreover, in PD-1 Ab group, the expression of IL-6, TNF-α,
   p-ERK/ERK, HIF-1α, HK2, LDHA and lactate were all elevated in the PFC
   and primary microglia, comparing with the IgG control group
   (Fig. [157]8A-N). Moreover, the level of Iba-1^+, the microglial soma
   size, branch number, and total branch length were significantly
   increased in PD-1 Ab group (Fig. [158]8B). Overall, pharmacological
   blockade of PD-1 enhanced neuroinflammation, glycolysis and anxiety,
   repetitive stereotype behaviors in MIA offspring mice.

Fig. 8.

   [159]Fig. 8
   [160]Open in a new tab

   Pharmacological blockade of PD-1 weakened the effect of MSC-EVs on
   neuroinflammation and glycolysis in MIA offspring mice. (A) IL-6 and
   TNF-α level of PFC and serum in MIA offspring mice after PD-1 Ab alone
   or PD-1 Ab + MSC-EVs administration. n=6. Two-way ANOVA. (B)
   Immunostaining analysis of Iba-1^+ of PFC in MIA offspring mice after
   PD-1 Ab alone or PD-1 Ab + MSC-EVs administration. Scale bar = 50 μm.
   One-way ANOVA. (C-F) Western blot analysis of PD-1, PD-L1, p-ERK/ERK,
   HIF-1α, HK2 and LDHA of PFC in MIA offspring mice after PD-1 Ab alone
   or PD-1 Ab + MSC-EVs administration. n=6. One-way ANOVA, Two-way ANOVA.
   (G) Lactate level of PFC and serum in MIA offspring mice after PD-1 Ab
   alone or PD-1 Ab + MSC-EVs administration. n=6. One-way ANOVA. (H)
   Schematic diagram illustrating the experiment procedure of with PD-1 Ab
   alone or PD-1 Ab + MSC-EVs intervention in primary microglia. (I) IL-6
   and TNF-α level of primary microglia with PD-1 Ab alone or PD-1 Ab +
   MSC-EVs intervention. n=3. Two-way ANOVA. (J-M) Western blot analysis
   of PD-1 and PD-L1. p-ERK/ERK, HIF-1α, HK2 and LDHA in primary microglia
   after PD-1 Ab alone or PD-1 Ab + MSC-EVs intervention. n=3. One-way
   ANOVA, Two-way ANOVA. (N) Lactate level in primary microglia
   supernatant after PD-1 Ab alone or PD-1 Ab + MSC-EVs intervention. n=3.
   One-way ANOVA. ns P≥0.05, *P<0.05, **P<0.01, ***P<0.001

   In addition, in PD-1 Ab + MSC-EVs group, the three-chamber test
   indicated that both social ability and social preference were
   decreased; the OFT test showed that entries times and duration in the
   center zone were significantly reduced; In the EPM test, entries times
   and duration into the open arms and the number of zone transition were
   significantly decreased, while the duration in the closed arms was
   increased; the number in marbles buried, grooming times and duration
   were all increased; the NOR test indicated cognitive level was
   decreased, comparing with the IgG + MSC-EVs group (Fig. [161]7B-G).
   Moreover, in PD-1 Ab + MSC-EVs group, the expression of IL-6, TNF-α,
   Iba-1, p-ERK/ERK, HIF-1α, HK2, LDHA and lactate were all elevated in
   the PFC and primary microglia, while the expression of PD-1 was no
   significant changed, comparing with the IgG + MSC-EVs group
   (Fig. [162]8A-N). Furthermore, comparing with the IgG + MSC-EVs control
   group, the level of Iba-1^+, the microglial soma size, branch number,
   and total branch length were all elevated in the PD-1 Ab + MSC-EVs
   group. Overall, pharmacological blockade of PD-L1/PD-1 weakened the
   effect of MSC-EVs on neuroinflammation, glycolysis and autism-like
   behaviors in MIA offspring mice.

   To further verify the effect of PD-L1 carried by MSC-EVs on MIA
   offspring mice, AAV-Iba-1-shPdcd1 was injected into the PFC of MIA
   offspring mice using brain stereotactic apparatus. Three weeks later,
   MSC-EVs were delivered into brain by intranasal administration daily
   for 7 days (Fig. [163]9A). In AAV-Iba-1-shPdcd1 + MSC-EVs group, the
   three-chamber test indicated that both social ability and social
   preference were decreased; the OFT test showed that entries times and
   duration in the center zone were significantly reduced; In the EPM
   test, entries times and duration into the open arms and the number of
   zone transition were significantly decreased, while the duration in the
   closed arms was increased; the number in marbles buried, grooming times
   and duration were all increased; the NOR test indicated cognitive level
   was decreased, comparing with the AAV-Ctrl + MSC-EVs group
   (Fig. [164]9B-G, Figure S9). Moreover, in AAV-Iba-1-shPdcd1 + MSC-EVs
   group, the expression of IL-6, TNF-α, Iba-1, p-ERK/ERK, HIF-1α, HK2,
   LDHA and lactate were all elevated in the PFC and serum, while the
   expression of PD-1 was decreased, comparing with the AAV-Ctrl + MSC-EVs
   group (Fig. [165]9H-O). Overall, microglia-specific PD-1 knockdown
   weakened the effect of MSC-EVs on neuroinflammation, glycolysis and
   autism-like behaviors in MIA offspring mice.

Fig. 9.

   [166]Fig. 9
   [167]Open in a new tab

   Microglia-specific PD-1 knockdown repressed the effect of MSC-EVs on
   inhibiting neuroinflammation, glycolysis and autism-like behaviors in
   MIA offspring mice. (A) Schematic diagram illustrating the experimental
   procedure of MIA offspring mice after Ctrl + MSC-EVs or
   AAV-Iba-1-shPdcd1 + MSC-EVs administration. (B-G) Autism-behaviors
   analysis of MIA offspring mice after Ctrl + MSC-EVs or
   AAV-Iba-1-shPdcd1 + MSC-EVs administration. (B) Statistical chart of
   sociability and social preference in the three-chamber test. Two-way
   ANOVA; (C) Statistical chart of entries times in center zone and time
   in center zone in the OFT. One-way ANOVA; (D) Statistical chart of
   entries in open arms and time in open arms in EPM test. One-way ANOVA;
   (E) Statistical chart of number in marbles buried. One-way ANOVA; (F)
   Statistical chart of grooming times and duration grooming. One-way
   ANOVA; (G) Statistical diagram of exploration time for each object in
   the NOR test. Two-way ANOVA. (H) Immunostaining analysis of GFP^+ cells
   stained with Iba-1^+. Scale bar: 20 μm. (I) Immunostaining analysis of
   Iba-1^+ of PFC in MIA offspring mice after Ctrl + MSC-EVs or
   AAV-Iba-1-shPdcd1 + MSC-EVs administration. Scale bar = 50 μm. One-way
   ANOVA. (J) IL-6 and TNF-α level of PFC and serum in MIA offspring mice
   after Ctrl + MSC-EVs or AAV-Iba-1-shPdcd1 + MSC-EVs administration.
   n=6. Two-way ANOVA. (K-N) Western blot analysis of PD-1, PD-L1,
   p-ERK/ERK, HIF-1α, HK2 and LDHA of PFC in MIA offspring mice after Ctrl
   + MSC-EVs or AAV-Iba-1-shPdcd1 + MSC-EVs administration. n=6. One-way
   ANOVA, Two-way ANOVA. (O) Lactate level of PFC and serum in MIA
   offspring mice after Ctrl + MSC-EVs or AAV-Iba-1-shPdcd1 + MSC-EVs
   administration. n=6. One-way ANOVA. ns P≥0.05, *P<0.05, **P<0.01,
   ***P<0.001

   Nextly, we reconfirmed whether PD-1 participates in the regulation on
   neuroinflammation and glycolysis in microglia by transfecting them with
   siRNA targeting Pdcd1 (siRNA-Pdcd1). Knocking down Pdcd1 in microglia
   inhibited expression of PD-1, p-ERK/ERK, HIF-1α, IL-6 and TNF-α, and
   glycolysis-related proteins HK2 and LDHA, which were similar to that
   showed upon PD-1 pharmacological inhibition (Fig. [168]10). In
   conclusion, our data indicated that MSC-EVs downregulates ERK/HIF-1α
   signaling to suppress neuroinflammation and elevated glycolysis in
   microglia.

Fig. 10.

   [169]Fig. 10
   [170]Open in a new tab

   Genetic suppression of PD-1 weakened the effect of MSC-EVs on
   neuroinflammation and glycolysis in MIA offspring mice. (A-D) Western
   blot analysis of PD-1, p-ERK/ERK, HIF-1α, HK2 and LDHA in primary
   microglia with MSC-EVs + si-Pdcd1 intervention. n=3. One-way ANOVA,
   Two-way ANOVA. (E) Lactate level in primary microglia supernatant with
   MSC-EVs + si-Pdcd1 intervention. n=3. One-way ANOVA. (F-G) IL-6 and
   TNF-α level of primary microglia with MSC-EVs + si-Pdcd1 intervention.
   n=3. Two-way ANOVA.ns P≥0.05, *P<0.05, **P<0.01, ***P<0.001

   To discern the effect of PD-L1 derived from MSC-EVs on
   neuroinflammation, glycolysis and autism-like behaviors in MIA
   offspring mice, we constructed engineering modified MSC-EVs with PD-L1
   deficient through transfecting shCd274 lentivirus (Figure S10). Then,
   MSC-EVs-Ctrl or MSC-EVs-shCd274 were delivered into MIA offspring brain
   by intranasal administration daily for 7 days (Fig. [171]11A). Behavior
   tests showed the social ability, social preference, anxiety and
   repetitive stereotyped behaviors were still presented in
   MSC-EVs-shCd274 group (Fig. [172]11B-G, Figure S11). In addition, the
   PD-L1 of PFC in MSC-EVs-shCd274 group were obviously less than
   MSC-EVs-Ctrl group. Comparing with MSC-EVs-Ctrl group, the expression
   of p-ERK/ERK, HIF-1α, HK2, LDHA, IL-6, TNF-α and lactate were all
   significantly increased in MSC-EVs-shCd274 group. Therefore, PD-L1
   deficient weakened the effect of MSC-EVs on neuroinflammation,
   glycolysis and autism-like behaviors in MIA offspring mice
   (Fig. [173]12A-G). In vitro, MSC-EVs-Ctrl or MSC-EVs-shCd274 were
   co-cultured with MIA primary microglia. As illustrated in
   Fig. [174]12H-N, in the MSC-EVs-shCd274 group, the expression of
   p-ERK/ERK, HIF-1α, HK2, and LDHA, IL-6, TNF-α and lactate were all
   significantly increased comparing with MSC-EVs-Ctrl group. In summary,
   PD-L1-deficient MSC-EVs diminished the protective effect of MSC-EVs on
   neuroinflammation and glycolysis in MIA primary microglia.

Fig. 11.

   [175]Fig. 11
   [176]Open in a new tab

   MSC-EVs with PD-L1 knockdown weakened the effect on autism-like
   behaviors in MIA offspring mice. (A) Schematic diagram illustrating the
   experiment procedure of MIA offspring mice after MSC-EVs-Ctrl or
   MSC-EVs-shCd274 administration. (B-G) Autism-behaviors analysis of MIA
   offspring mice after MSC-EVs-Ctrl or MSC-EVs-shCd274 administration.
   (B) Statistical chart of sociability and social preference in the
   three-chamber test. Two-way ANOVA; (C) Statistical chart of entries
   times in center zone and time in center zone in the OFT. One-way ANOVA;
   (D) Statistical chart of entries in open arms and time in open arms in
   EPM test. One-way ANOVA; (E) Statistical chart of number in marbles
   buried. One-way ANOVA; (F) Statistical chart of grooming times and
   duration grooming. One-way ANOVA; (G) Statistical diagram of
   exploration time for each object in the NOR test. Two-way ANOVA. n=8.
   ns P≥0.05, *P<0.05, **P<0.01, ***P<0.001

Fig. 12.

   Fig. 12
   [177]Open in a new tab

   MSC-EVs with PD-L1 knockdown weakened the effect on neuroinflammation
   and glycolysis in MIA offspring mice. (A) IL-6 and TNF-α level of PFC
   and serum in MIA offspring mice after MSC-EVs-Ctrl or MSC-EVs-shCd274
   administration. n=6. Two-way ANOVA. (B) Immunostaining analysis of
   Iba-1^+ of PFC in MIA offspring mice after MSC-EVs-Ctrl or
   MSC-EVs-shCd274 administration. n=6. Scale bar = 50 μm. One-way ANOVA.
   (C-F) Western blot analysis of PD-L1/PD-1, p-ERK/ERK, HIF-1α, HK2 and
   LDHA of PFC in MIA offspring mice after MSC-EVs-Ctrl or MSC-EVs-shCd274
   administration. n=6. One-way ANOVA, Two-way ANOVA. (G) Lactate level of
   PFC and serum in MIA offspring mice after MSC-EVs-Ctrl or
   MSC-EVs-shCd274 administration. n=6. One-way ANOVA. (H) Schematic
   diagram illustrating the experiment procedure of primary microglia with
   MSC-EVs-Ctrl or MSC-EVs-shCd274 intervention. (I) IL-6 and TNF-α level
   of primary microglia with MSC-EVs-Ctrl or MSC-EVs-shCd274 intervention.
   n=3. Two-way ANOVA. (J-M) Western blot analysis of PD-L1/PD-1,
   p-ERK/ERK, HIF-1α, HK2 and LDHA in primary microglia with MSC-EVs-Ctrl
   or MSC-EVs-shCd274 intervention. n=3. One-way ANOVA, Two-way ANOVA. (N)
   Lactate level of primary microglia supernatant with MSC-EVs-Ctrl or
   MSC-EVs-shCd274 intervention. n=3. One-way ANOVA. ns P≥0.05, *P<0.05,
   **P<0.01, ***P<0.001

   To futher investigated the association of ERK and HIF-1α in primary
   microglia of MIA offspring mice, we used honokiol, a specific activator
   of ERK, and SL327, a specific inhibitor of ERK. In contrast,
   pretreatment with 20 µM honokiol markedly induced MSC-EVs inhibited
   HIF-1α expression and glycolysis (Fig. [178]13A-F). Moreover,
   pretreatment with 10 µM SL327 markedly decreased MSC-EVs combined PD-1
   Ab-induced HIF-1α expression and glycolysis (Fig. [179]13G-L),
   indicating that HIF-1α-induced glycolysis were regulated by ERK
   activation in microglia.

Fig. 13.

   [180]Fig. 13
   [181]Open in a new tab

   MSC-EVs inhibited microglial activation and glycolysis via the
   PD-1/ERK/HIF-1α pathway. (A) Schematic diagram illustrating the
   experiment procedure of combining MSC-EVs with ERK activator honokiol
   intervention in primary microglia. (B-D) Western blot analysis of
   p-ERK, ERK, HIF-1α, HK2 and LDHA in primary microglia with combining
   MSC-EVs and ERK activator honokiol intervention. n=3. One-way ANOVA,
   Two-way ANOVA. (E) Lactate level in primary microglia supernatant with
   combining MSC-EVs with ERK activator honokiol intervention. n=3.
   One-way ANOVA. (F) IL-6 and TNF-α level of primary microglia with
   combining MSC-EVs with ERK activator honokiol intervention. n=3.
   Two-way ANOVA. (G) Schematic diagram illustrating the experiment
   procedure of MSC-EVs combining PD-1 Ab and ERK inhibitor SL327
   intervention in primary microglia. (H-J) Western blot analysis of
   p-ERK, ERK, HIF-1α, HK2 and LDHA in primary microglia of MSC-EVs
   combining PD-1 Ab and ERK inhibitor SL327 intervention. n=3. One-way
   ANOVA, Two-way ANOVA. (K) Lactate level of primary microglia
   supernatant of MSC-EVs combining PD-1 Ab and ERK inhibitor SL327
   intervention. n=3. One-way ANOVA. (L) IL-6 and TNF-α level of primary
   microglia of MSC-EVs combining PD-1 Ab and ERK inhibitor SL327
   intervention. n=3. Two-way ANOVA. ns P≥0.05, *P<0.05, **P<0.01,
   ***P<0.001

Discussion

   The neuroinflammatory response of microglia, which was triggered by
   stimulation and damage, could aggravate brain damage, affect brain
   development, and ultimately cause inflammatory CNS diseases
   [[182]28–[183]30]. More evidences support that MIA induced by maternal
   infections, allergies or autoimmune diseases, was one of the
   significant risk factors of ASD [[184]2]. It is worth emphasizing that
   the infectious agents can not injure the fetal brain directly, but they
   induce microglial activation and enhance immune response in the
   offspring by increasing pro-inflammatory cytokine level in both brain
   and peripheral serum in response to exogenous stimulation [[185]6,
   [186]31–[187]34]. More importantly, the cytokines and chemokines were
   elevated at different developmental stages and continued to develop
   into adulthood in MIA models, which eventually leading to
   neurodevelopmental disorders [[188]35]. In this study, we found that
   MIA offspring mice exhibited classic autism-like behaviors, and an
   increase in the number of microglia. Moreover, IL-6 and TNF-α were
   elevated significantly, not only in the fetal brain but also PFC and
   peripheral serum in adulthood offspring. All these results indicated
   that neuroinflammation was closely associated with the pathogenesis of
   ASD.

   Recent studies showed there was abnormally increasing lactate and LDHA
   in ASD patients [[189]36–[190]39]. Hideo et al. [[191]40] also reported
   that ASD patients exhibited high lactate level, consequently along with
   the brain pH decreased. Furthermore, this phenomenon was also observed
   in several kinds of ASD mice models. Licznerski et al. [[192]41] have
   found significant upregulation of HK1, HK2, PFKP, PKM2 and LDHA, which
   were all glycolysis-related key enzymes in Fmr1^−/− mice. Similarly,
   significantly increased glycolysis level were also observed in the PFC
   of Shank3B^−/− mice and VPA induced mice models. It has also been
   reported that the some ASD behaviors could be rescued when suppressnig
   the glycolysis by 2-DG in VPA-induced mice [[193]42]. Thus, the
   clinical reports and mice research above have inspired the importance
   of “Warburg effect hypothesis” in ASD pathology, indicating that rising
   glycolysis may be one risk factor for ASD [[194]43]. Our research
   showed that the lactate level were all elevated in the supernatant of
   primary microglia, PFC and peripheral serum from MIA offspring mice, at
   the same time, along with the increasing expression of glycolysis key
   enzymes (HK2 and LDHA).

   Furthermore, active microglia were typically involved in a metabolic
   reprogramming, mainly manifested in increased glycolysis, so as to
   provide energy rapidly. Microglia stimulated by LPS could induce fierce
   inflammtory response, with the glucose consumption and lactate
   production. While, supressing glycolysis could also eliminate
   LPS-induced microglial activation via NF-κB pathway [[195]44]. Thereby,
   these research findings have demonstrated that neuroinflammation can
   enhance glycolysis in the neurometabolism process. Conversely, it is an
   available way to downregulate inflammation activation by inhibiting the
   glycolysis. Here, IL-6, TNF-α and lactate level were all decreased in
   primary microglia and PFC of MIA offspring mice, upon interventing with
   the glycolysis inhibitor 2-DG. This study suggested that glucose
   metabolism may be a key regulatory factor in neurological inflammation,
   which broadening the avenue in treating inflammation-related diseases
   by manipulating the metabolism of immune cells.

   Remarkably, several evidences have showed that MSC-EVs exhibited
   significant therapeutic potential in various neurodevelopmental
   disorders through inhibiting neuroinflammation. Previous studies
   demonstrated that MSC-EVs pocessed the anti-inflammatory properties and
   facilitated neuroregeneration and angiogenesis when applied for
   ischemic stroke [[196]45]. Allaura et al. [[197]46] have reported that
   microglial activation was inhibited and release of IL-1β, TNF-α and
   IL-6 were reduced when intranasally delivering MSC-EVs to Alzheimer
   mice, futher to improve the recognization behavior of mice. In this
   study, we found that MSC-EVs could significantly decrease the
   expression of Iba-1^+, IL-6 and TNF-α in the PFC of MIA offspring mice.

   PD-1 is a co-inhibitory receptor that interacts with its ligand PD-L1
   to inhibit T cell receptor (TCR)-mediated signaling. It is prominently
   expressed on tumor infiltrating lymphocytes (TILs), while PD-L1 is
   highly expressed on tumor cells. Therefore, the immune checkpoint
   PD-L1/PD-1 is an effective method for tumor therapy, as tumor cells
   utilize regulatory defense mechanisms to evade attacks from immune
   cells. While the mechanisms of the PD-L1/PD-1 axis have primarily been
   studied in T cells, there are involved increasingly evidences that
   other cell types in the CNS. In addition, PD-L1/PD-1 axis has potential
   ability in reducing neuroinflammatory response and potentially
   regulating neuronal activity [[198]10]. Zhang et al. [[199]47] showed
   that PD-L1/PD-1 axis induced M2 polarization of microglia by reducing
   phosphorylation of STAT1, increasing phosphorylation of STAT6, and then
   downregulating the NF-κB signaling. Additionally, ERK pathway in
   microglia was a key regulatory factor in pro-inflammatory response,
   which could be modulated by PD-L1/PD-1 axis in neuropathic pain
   [[200]48]. Despite this, research on the relationship between the
   PD-1/PD-L1 axis and MIA induced ASD is currently limited. Here, we
   observed that PD-L1 was highly expressed in the MSC and MSC-EVs, the
   level of PD-1 was upregulated and the ERK pathway was activated in
   primary microglia and PFC of MIA mice model. MSC-EVs was performed to
   MIA offspring mice by intranasal delivering or co-cultured with primary
   microglia, which could significantly elevated PD-L1 level, then
   activated PD-L1/PD-1 axis, repressed the ERK pathway and futher
   inhibited the microglia activation and glycolysis. Therefore, we
   speculated that PD-L1/PD-1 axis has potential ability in decreasing
   neuroinflammatory response, implicating a possible method for ASD
   therapy.

   HIF-1α is a crucial transcription factor for glucose metabolism by
   promoting the transcription of the key enzymes in glycolysis (GLUT1,
   HK2, PKM2 and LDHA) [[201]49, [202]50]. It is known that the activation
   of ERK pathway could act as an upstream modulator of HIF-1α and
   increase the glycolysis in cancer [[203]51]. Our research indicated
   that ERK/HIF-1α pathway was inhibited by MSC-EVs due to PD-L1/PD-1
   axis. More importantly, the effects of MSC-EVs on the neuroinflammation
   and glycolysis could be weakend by brain stereotaxic injection of PD-1
   Ab or microglia-specific PD-1 knockdown via AAV-Iba-1-shPdcd1 in vivo,
   and trandfected with Pdcd1 siRNA in microglia in vitro. Notably,
   MSC-EVs with PD-L1 knockdown could also weaken the effect on
   neuroinflammation, glycolysis and autism-like behaviors of MIA
   offspring mice. What’s more, pharmacological inhibition of ERK could
   suppresse HIF-1α mediated glycolysis and neuroinflammation in primary
   microglia of MIA offspring mice. While, the ERK activator treatment
   exacerbate the level of pro-inflammatory factors and latate. All the
   above experiments have proved that MSC-EVs could maintain glycolysis
   and immune homeostasis through mediating ERK/HIF-1α and transcription
   of HK2 and LDHA by PD-L1/PD-1.

   MSC-EVs have a protective effect on nerve cells through their various
   bioactive substances, such as neurotrophic factors, immune regulatory
   and anti-inflammatory molecules, which can alleviate neuronal damage,
   inhibit neuronal apoptosis, and maintain the normal structure and
   function of neurons [[204]52, [205]53]. Zhao et al. have reported that
   PD-1 could regulate neuronal excitability in hippocampal neurons, and
   the selective deletion of PD-1 in excitatory neurons could enhance
   long-term potentiation (LTP) and memory of mice [[206]54]. Notably,
   PD-1 was expressed in microglia, astrocytes and neurons by
   immunostaining in our study, but only in microglia PD-1 was
   significantly higher than controls. Therefore, we explored that PD-L1
   carried by MSC-EVs attenuated neuroinflammation, glycolysis and
   autism-like behaviors in MIA offspring mice, which activated PD-L1/PD-1
   axis in microglia.

Conclusions

   In conclusion, based on the immunosuppressive properties of the
   PD-1/PD-L1 axis, our study explored the potential therapy on MSC-EVs in
   alleviating MIA-induced ASD in offspring. We identified a critical link
   between neuroinflammation and hyperactivated glycolysis, both in vivo
   and in vitro, highlighting their contribution to ASD pathogenesis.
   MSC-EVs effectively alleviated autism-like behaviors by suppressing
   hyperactivation of microglia and glycolysis through the PD-L1/PD-1 axis
   in MIA-induced offspring mice. Mechanistically, we found that MSC-EVs
   exerted a key role in glucose metabolism and neuroinflammation through
   PD-1/ERK/HIF-1α signaling in MIA offspring mice.

Electronic supplementary material

   Below is the link to the electronic supplementary material.
   [207]Supplementary Material 1^ (5.7MB, docx)

Acknowledgements