Preparation of human astrocytes with potent therapeutic functions from human pluripotent stem cells using ventral midbrain patterning

Graphical abstract

Highlights

• •Human forebrain & midbrain-type astrocytes were stably differentiated from hPSC-derived organoids.
• •hPSC-derived astrocytes mimic human brain counterparts in transcriptomics & physiology.
• •hPSC-derived astrocytes exhibited neurotrophic & anti-inflammatory activities.
• •hPSC-derived astrocytes alleviated PD & AD proteopathy in in-vitro & in-vivo settings.
• •Human midbrain-type astrocytes exhibit more potent therapeutics capacities.

Abstract

Introduction

Astrocytes are glial-type cells that protect neurons from toxic insults and support neuronal functions and metabolism in a healthy brain. Leveraging these physiological functions, transplantation of astrocytes or their derivatives has emerged as a potential therapeutic approach for neurodegenerative disorders.

Methods

To substantiate the clinical application of astrocyte-based therapy, we aimed to prepare human astrocytes with potent therapeutic capacities from human pluripotent stem cells (hPSCs). To that end, we used ventral midbrain patterning during the differentiation of hPSCs into astrocytes, based on the roles of midbrain-specific factors in potentiating glial neurotrophic/anti-inflammatory activity. To assess the therapeutic effects of human midbrain-type astrocytes, we transplanted them into mouse models of Parkinson's disease (PD) and Alzheimer's disease (AD).

Results

Through a comprehensive series of in-vitro and in-vivo experiments, we were able to establish that the midbrain-type astrocytes exhibited the abilities to effectively combat oxidative stress, counter excitotoxic glutamate, and manage pathological protein aggregates. Our strategy for preparing midbrain-type astrocytes yielded promising results, demonstrating the strong therapeutic potential of these cells in various neurotoxic contexts. Particularly noteworthy is their efficacy in PD and AD-specific proteopathic conditions, in which the midbrain-type astrocytes outperformed forebrain-type astrocytes derived by the same organoid-based method.

Conclusion

The enhanced functions of the midbrain-type astrocytes extended to their ability to release signaling molecules that inhibited neuronal deterioration and senescence while steering microglial cells away from a pro-inflammatory state. This success was evident in both in-vitro studies using human cells and in-vivo experiments conducted in mouse models of PD and AD. In the end, our human midbrain-type astrocytes demonstrated remarkable effectiveness in alleviating neurodegeneration, neuroinflammation, and the pathologies associated with the accumulation of α-synuclein and Amyloid β proteins.

Introduction

Astrocytes are the most abundant glial cells in the brain and have diverse neurotrophic and homeostatic functions. Their primary role is to safeguard neuronal cells against both extracellular and intracellular toxicity through various mechanisms including scavenging neuronal reactive oxygen species (ROS) [1] and clearing excessive excitotoxic glutamate [2], harmful lipids [3], and pathological protein aggregates [4], [5]. Astrocytes also release neurotrophic factors and anti-inflammatory cytokines [6] and transfer healthy mitochondria from astrocytes to damaged neuronal cells [7]. Moreover, they support neuronal energy and lipid metabolism by providing lactate and cholesterol to cells [8]. Additionally, astrocytes play a crucial role in regulating neuronal synaptic transmission through their involvement in the tripartite synapse [9].

Given their diverse neurotrophic functions, astrocytes have emerged as a prime focus for the development of disease-modifying therapies for neurodegenerative disorders that currently lack effective treatments. Previous studies have explored astrocyte cultures derived from rodent brains or through the differentiation of human pluripotent stem cells (hPSCs). Those studies have demonstrated therapeutic effects by transplanting cultured astrocytes and their derivatives, such as conditioned medium and exosomes, in various disease contexts (as reviewed in [10], [11], [12], [13]). However, further research to enhance the efficacy and safety of astrocyte-based therapies is crucial to their successful clinical translation. Continuing efforts are needed to achieve the desired outcomes in clinical applications.

Compared with other brain regions, the midbrain exhibits higher levels of ROS and toxic metabolites [14] due to the unique metabolism of dopaminergic neurons. However, the midbrain has a lower astrocyte-to-neuron ratio than other areas [15], indicating that each astrocyte in this region must possess heightened neuroprotective capabilities to mitigate the neurotoxic burden. Supporting that notion at the molecular level, orphan nuclear receptor NURR1 (NR4A2 or NURR1) and forkhead box A2 (FOXA2 or HNF3B), transcriptional and epigenetic factors specifically expressed in the midbrain [16], synergistically function as master regulators to enhance the neurotrophic and anti-inflammatory functions of glia [17], [18]. Studies have further validated the role of Nurr1 and Foxa2 in astrocytes by showing that forced expression of these midbrain-specific factors augments the therapeutic functions of astrocytes in cortical and hippocampal brain regions [19], [20]. Consistently, previous studies have also shown that astrocytes derived from the ventral midbrain of rodents exhibit more potent therapeutic capacities than those from other brain regions [5], [18]. Building on that knowledge, we adopted a ventral midbrain regional patterning approach to generate human astrocytes with robust therapeutic potential.

Successful hPSC-based cell therapy requires an hPSC differentiation protocol stable enough to yield donor cells with minimum cellular stress to ensure the survival and engraftment of transplanted cells. However, differentiation of hPSCs into midbrain-type cells necessitates close cell-to-cell contacts, often requiring extremely high cell densities and the use of combinatory patterning chemicals that can be toxic to the cells to some extent [21]. Consequently, in this challenging culture environment, patterned cells inevitably accumulate severe cellular stresses post-procedure. Recognizing this challenge, we hypothesized that inducing midbrain patterning in a 3D organoid culture environment, where physiological cell-to-cell interactions are established and maintained akin to in vivo midbrain development, could mitigate cellular stress. Building on this concept, we developed a protocol that we call the organoid-based method to prepare brain region–specific neural stem cells (NSCs) from hPSC-derived brain organoids [22]. Compared with conventional methods, the organoid-based method allows the preparation of donor cells with minimal mitochondrial and oxidative stress and cell senescence [22]. Along with their low cellular stress, donor cells prepared using the organoid method engrafted efficiently to host brain tissues in a stable and reproducible manner [22].

In this study, we prepared human astrocytes from hPSC-derived ventral midbrain–patterned organoids for therapeutic purposes. A series of transcriptome and functional analyses demonstrated that the astrocytes derived using our strategy exhibited excellent neurotrophic and regenerative capacities in various neurodegenerative conditions, including Parkinson’s and Alzheimer’s disease-specific contexts.

Methods

Cell culture

hESC culture

hESCs (H9; WA09; female; Wicell) were cultured on Matrigel (Corning) coated 60 mm dishes (Thermo Fisher, Waltham, MA) in feeder-free conditions with mTeSR^TM medium (Stemcell Tech. Inc., Vancouver, BC, Canada). The undifferentiated hESCs were maintained with a medium change every day and were passaged every 4–5 days (cell confluence > ∼70 %) using Accutase (Stemcell Technologies Inc.).

hESC differentiation into midbrain- or forebrain-type organoid-derived NSCs

A schematic protocol is presented in Fig 1A, and a detailed protocol is provided in Supplementary Fig. S1A. To obtain midbrain- and forebrain-NSCs, hESCs were differentiated using midbrain- and forebrain-like organoid protocols. hESCs were seeded at a density of 1ⅹ10⁴ cells/well in neural induction medium in a low-attachment 96-well plate (96 well Round-bottom plate; Corning, Corning, NY). The ROCK inhibitor Y27632 (20 μΜ, Sigma-Aldrich, St. Louis, MO) was added for the first 24 hr. The neuronal induction was performed using a medium with N2:Neurobasal (1:1, Gibco),B27 without vitamin A Invitrogen, Waltham, MA), GlutaMAX (1 %, Invitrogen Fisher Scientific), MEM-NEAA (1 %, minimum essential media-nonessential amino acid, Invitrogen Fisher Scientific), SB431542 (10 μM, Toris Bristol, UK), Noggin (100 ng/mL, Peprotech, Rocky Hill, NJ),insulin (25 mg/L; Gibco) and ascorbic acid (200 μΜ, Sigma-Aldrich). Midbrain-specific patterning of the differentiating hESCs was induced by adding sonic hedgehog (SHH, 100 ng/ mL, Peprotech) for days 1–14, purmorphamine (2 μΜ, Calbiochem, Millipore, Sigma, Burlington, MA) for days 1–17, CHIR99021 (Stemgent, Cambridge, MA) at 0.8 μM for days 2–10 and at 1.6 μM for days 11–17, and FGF8b (100 ng/ mL, Peprotech) for days 7–17. The medium was replenished every other day (except days 0–3). For forebrain patterning, XAV930 (2 μg; Stemgent) was administered for days 1–5. Floating spheres (organoids) were transferred to low-attachment 6-well plates (Corning, 8 spheres/well) on day 10.The organoids were cultured on shaking condition (orbital shaker,80 rpm) with medium changes every 2 days.

To enrich regional-specific (midbrain or forebrain)-type NSC populations in the organoids, thebasic fibroblast growth factor (bFGF;20 ng/ mL; Peprotech) and EGF (10 ng/mL; Peprotech) were applied. bFGF treatment began on day 10 (for the Forebrain (Fo)-type) or day 18 (for the Midbrain (Md)-type), and EGF treatment began on day 11 (for the Fo-type). For Md-type NSC enrichment, SHH (100 ng/ mL) and purmorphamine (2 μΜ) were administered during expansion. On day 18, the Md- and Fo-type organoids were separately chopped, dissociated with Accutase, and plated on 6-cm poly-L-ornithine (PLO)/fibronectin (FN)-coated dishes (Thermo Fisher Scientific) in N2 expansion medium (N2 base medium with 10 ng/mL BDNF, 10 ng/ mL GDNF, 10 ng/mL bFGF, and 200 μΜ ascorbic acid) supplemented with 5 μΜ Y27632. The NSCs were cultured in high density on 6-cm PLO/FN-coated dishes (Thermo Fisher Scientific) in NSC expansion medium in a humidified CO₂ incubator at 37℃ Cells passaged at 5–7 day intervals. Y27632 was added for 1 day after each cell passage.

Neuronal differentiation of the NSCs was performed between passages 2 and 7 and induced by N2 medium with 10 ng/mL GDNF, 10 ng/mL BDNF, 200 µM ascorbic acid, and 125 µM db-cAMP.

Derivation of astrocytes from NSCs

NSCs were maintained in PLO/FN-coated dishes in NSC medium, as detailed in Supplementary Fig. S1A. Astrocytes were used between passages 15 and 22 and were passaged every week. Late-stage NSCs (astrocyte progenitors) were seeded as dissociated single cells on PLO/FN-coated dishes at 1ⅹ10⁵ cells/cm² in astrocyte progenitor medium (N2 base medium containing 10 ng/mL EGF, 10 ng/mL bFGF, and 200 μΜ ascorbic acid). For midbrain astrocyte-progenitors (Md-AST-progenitor), the midbrain-specific cytokine purmorphamine (2 μΜ) was supplied continuously during expansion. For final astrocyte differentiation, astrocyte progenitors that had expanded for 5 days were spontaneously differentiated by withdrawing the mitogens (bFGF, EGF). All astrocytes were used 15 days after differentiation.

Derivation of human microglia from hESCs

Microglia were produced using an hESC-macrophage organoid-derived generation protocol. hESCs were seeded at 10,000 cells/well on a low-attachment 96-well round-bottom plate in mTESR1, hematopoietic cytokines (BMP4, VEGF-A, SCF), and the ROCK inhibitor Y276232 (10 µM) on day 0. Y276232 was used only for the first 24 hr, and the medium was changed every day until day 3. On day 4, 18–20 organoids were moved to one well of an ultra-low-binding 6-well plate with hematopoietic medium, Xvixo-15^TM (LONZA, Basel, Swiss) containing GlutaMAX and β-mercaptoethanol with cytokines (SCF, M−CSF, IL-3, FLT3). Macrophage progenitor cells were generated from macrophage organoids during days 11–17 and then harvested. The macrophage progenitor cells were collected and differentiated into microglia using N2, GlutaMAX, β-mercaptoethanol, M−CSF, and IL-34 for at least 2 weeks.

Conditioned medium preparation and treatment

Fresh N2 medium was added to cultured Md- and Fo-type astrocytes. The conditioned-medium (CM) was collected every other day for 10 days. The CMs were adjusted to a protein concentration of 0.1–0.15 mg/mL, filtered with a pore-size 0.45-μm, and stored at −80℃ until use. For experiments, the CMs were diluted with normalculture medium (1:1, v/v).

Passive conductance recording

Passive conductance recording was performed with samples continuously perfused with an artificial cerebral spinal fluid (ACSF) solution containing (in mM): 130 NaCl, 3.5 KCl, 24 NaHCO3, 1.25 NaH2PO4, 1.5 CaCl2, 1.5 MgCl2, and 10 d-(+)-glucose (pH 7.4, flow rate = 2 mL/min). The recording chamber was mounted on an upright Olympus microscope and viewed with a 40X water immersion objective lens. Cellular morphology was visualized with a camera and Imaging Workbench software (INDEC BioSystems, ver. 9.0.15.0, Los Altos, CA). Whole‐cell recordings were produced for the patterned astrocytes. The holding potential was −60 mV. Pipette resistance was typically 4–6 MΩ in voltage-clamp mode, and pipettes were filled with an internal solution containing (in mM) 150 KCl, 1 CaCl₂, 1 MgCl₂, 5 EGTA, and 10 HEPES (pH 7.3 was adjusted with KOH). Electrical signals were digitized and sampled at 10 kHz with a Digidata 1320A (Axon Instruments; Molecular Devices, San Jose, CA) and a Multiclamp 700B amplifier (Molecular Devices) using pCLAMP 10.2 software (Molecular Devices). Raw data collected at a sampling rate of 10 KHz (pClamp 10.2 software).

Assays for cellular ROS and cell senescence

Cellular ROS levels were estimated using CellROX Green (Thermo Fisher, Waltham, MA) following the manufacturer’s protocol. Cell senescence was measured using an SA-β-Gal staining kit (#9860; Cell Signaling, Danver, MA) according to the manufacturer’s instructions. Cells were plated at a density of 4ⅹ10⁴ cells/cm², and SA-β-Gal was measured after plating.

Metabolism assay

For the metabolic assays, astrocytes (1–2ⅹ10⁶) were harvested, and cell supernatants were prepared. Total cholesterol (EZ-total cholesterol assay kit, DG-TSC100; Dogenbio, Seoul, Korea), cholesterol efflux (cholesterol efflux assay kit, ab196985, Abcam), glutamate uptake activity (glutamate assay kit, ab83389; Abcam), and lactate metabolism (EZ-lactate assay kit, DG-LAC200; Dogenbio) were assessed according to the manufacturer’s protocol for each kit.

Immunostaining

Immunocytochemistry

Cells were fixed by 4 % paraformaldehyde (PFA) for 20 min treatment, washed 3 times with PBS. Add blocking solution (0.3 % Triton X-100 / 1 % BSA) for 40 min. Replace with primary antibodies and place at 4 °C overnight, followed by the treatment secondary antibodies for 1 hr at room temperature. The stained cells were mounted with mounting solution (Vectashield with DAPI mounting solution;Vector Lab., Burlingame, CA). An epifluorescence microscope (Leica Wetzlar, Germany) and a confocal microscope (Leica PCS SP5) were used for obtaining the photographs. Immunoreactive cells on coverslip cultures were quantified in randomly selected microscopic fields within a region of uniform cell growth using an eyepiece grid at a final magnification of 200x or 400x. On each coverslip, 10–20 microscopic fields were counted. Data are expressed as mean ± SEM of at least three independent cultures.

Immunohistochemistry

Perfused mouse brains were sinked in 30 % sucrose solution in PBS for 5 ∼ 6 days. The brains were embedded in O.C.T. (Sakura Finetek, Torrance, CA) for cryosection. Brain regions were sliced into 30 μm thicknesses using a freezing microtome (CM 1850; Leica Microsystems) and then subjected to immunohistochemistry and thioflavin S staining. The brain sections were blocked with 3 % BSA /0.7 % Triton X-100 in PBS for 1 hr at RT. For NURR1 staining of mouse brain sections from the SN, heat-induced epitope retrieval was conducted at 95℃ for 20 min in Tris-Cl, 1 mM EDTA buffer (pH 9.0), or 1X citrate buffer (pH 6.0, Sigma-Aldrich). Neurite length, soma size, the pS129- α-syn (α-synuclein) immunoreactivity of TH + Md-DA neurons in the SN, and TH + fiber intensity in the striatum (Md-astrocyte-grafted mice) were compared with those in the SN or striatum of their PBS-injected or Fo-astrocyte-grafted counterparts. The subsequent steps were consistent with those detailed above. Information about primary and secondary antibodies is provided in Supplemental Table S2. All images were captured using an epifluorescence (Leica) or confocal (Leica TCS SP5 confocal) microscope. The intensity of the immunofluorescence was measured using Leica Application Suite X (LAS X) or ImageJ (Image Processing and Analysis in Java, NIH) software.

Histological analysis

Brains were sliced (a thickness of 1 mm) using a mouse brain slice matrix (ZIVIX Instruments), and regions of the SN (PD model) or hippocampus (AD model) were dissected and subjected to WB analysis. In the experiment examining the host environments of the transplanted brains, the animals were sacrificed 10–12 weeks post-transplantation, and 8 to 10 regions of the graft–host interface (around 2 x 2 mm/graft) were dissected and analyzed using qPCR. Immunoreactive cells for pro-inflammatory markers were counted along the host-graft interfaces of cryosectioned brain slices. For immunohistochemistry, mice were anesthetized and perfused intracardially with 4 % PFA in PBS, followed by immunohistochemical analysis. The number of immunoreactive cells was quantified via unbiased stereological counting, with the Abercrombie correction applied to account for double counting in adjacent sections. The Abercrombie correction factor is (N = n × T/(T + D)), where N represents the actual cell number(n: the number of nuclear profiles, T:the section thickness (30 μm), D:the average diameter of the nuclei). The histological analyses were conducted every 8 sections throughout the midbrain (PD model) and hippocampus (AD model).

Thioflavin s staining

Cultured cells were stained in thioflavin S solution (500 μM in 50 % EtOH; Sigma-Aldrich, St. Louis, MO) for seven minutes and subsequently washed with EtOH and PBS in that order. Hoechst solution was added to the sample, which was mounted with CRYSTAL/MOUNT (Biomeda Corp., Foster City, CA).

Transcriptome analyses

Real-time PCR analysis

Total RNA from Md-astrocytes and Fo-astrocytes was isolated using Trizol^R (Invitrogen) according to the manufacturer’s instructions. Reverse transcriptase reactions were performed using a Superscript kit (Invitrogen). For each reaction, 1 μg of RNA was used. The amplified material was detected using a CFX96™ real-time system with iQ™ SYBR Green Supermix, and data were processed using a CFX connect real-time system (Bio-Rad Laboratories, Hercules, CA). All results were normalized to the GAPDH control. The primers used in this study are detailed in Supplementary Table S2.

RNA sequencing and analysis

Total RNA was extracted using Trizol according to the manufacturer’s instructions. Only samples with an RNA integrity number > 7.0 were processed for the RNA-seq analysis. Library construction was conducted using a TruSeq standard mRNA library prep kit (Illumina, San Diego, CA). Reads were mapped and counted against the GRCh38 human reference genome. Gene classification was performed through DAVID (david.ncifcrf.gov/) searches. Annotation information for DEGs, GO, and KEGG pathway analyses was acquired using the DAVID database and SRplot tool. Enriched categories in the GO and KEGG pathway analyses are presented with a p-value threshold of 0.05.

Phagocytic activity and clearance assay

The phagocytic activity of cultured Md-astrocytes and Fo-astrocytes was estimated using the uptake of fluorescent latex beads (F-13081, Thermo Fisher) or α-syn PFFs (α-syn Preformed Fibrils) or Aβ (Amyloid β) labeled with Alexa 488. Labeling the α-syn PFFs and Aβ was performed using a protein labeling kit (A30006, Invitrogen) in accordance with the manufacturer’s protocols. To estimate the phagocytic clearance activity of the cultured astrocytes, the cells were treated with labeled α-syn PFFs (1 μM) or Aβ (1.5 μM) for 24 hr and then washed. The intracellular fluorescent puncta remaining were counted at multiple time points up to 7 days.

Preparation of α-syn PFFs

Recombinant α-syn (human) and α-syn PFFs were prepared as described previously. Briefly, 5 mg/mL monomeric α-synuclein was incubated at 37 °C with continuous agitation at 1,000 rpm for 7 days, sonicated on ice for 3–5 s at 3 W (ultrasonic processor VC 505;Sonics & Materials, Inc., Newtown, CT), and stored at −80 °C until use as α-syn PFF. the α-syn fibrils were determined using a transmission electron microscope (Zeiss, Oberkochen, Germany) and the thioflavin T binding assay.

Preparation of Aβ aggregates

Human Aβ peptide (1–42) (Abcam, Cambridge, England) was dissolved in 5 mM NaOH/PBS solution and incubated at 37 °C for three days. The Aβ solution was centrifuged, and the resulting pellet (Aβ aggregates) was dissolved in 0.1 M sodium bicarbonate. The aggregates were stored at −80 °C until use. The Aβ aggregates and monomers were mixed, and the resulting Aβ mixture was used throughout this study.

Western blot analysis

To detect aggregate forms of α-syn, protein was extracted from cultured cells or animal brain tissues using a 1 % Triton X-100/PBS buffer containing a protease inhibitor (Roche) and phosphatase inhibitor cocktails (Sigma). After centrifugation (16,000 g, 10 min), the Triton X-100 supernatant was collected (Triton X-100-soluble protein). The pellet was dissolved in 1 % SDS sample buffer and then briefly sonicated (Triton X-100-insoluble protein). For Aβ aggregation analysis, protein was extracted from cultured cells or animal brain tissues using RIPA buffer with a protease inhibitor (Roche, Mannheim, Germany). Protein samples (15–40 μg of protein) were electrophoresed on SDS-PAGE gels (6–20 % gradient), transferred to a nitrocellulose membrane, blocked with 5 % BSA/PBS, and then incubated with primary antibodies (overnight, 4℃). The next day, the secondary antibodies treatment occurred for 1 h. Signals were captured with ChemiDoc (Bio-Rad Laboratories) and quantification was performed using ImageJ (NIH, Bethesda, MD).

Inflammasome analysis

Co-cultures of neuron + microglia + Md-astrocytes (Md-AST) or neuron + microglia + Fo-astrocytes (Fo-AST) were treated with LPS (1 µg/mL, 4 hr), followed by ATP treatment (2 mM, 30 min) before analysis. Inflammasome activation was assessed through WB analysis of proteins secreted in the CM. The cleaved/activated forms of caspase-1 and IL-1β were confirmed using specific antibodies. Secreted protein levels were measured in the media and normalized against Ponceau S-stained total protein levels.

Animal housing

Animal experiments were approved by the IACUC (Institutional Animal Care and Use Committee) of Hanyang College of Medicine (approval numbers 2021-0086A, 2021-0215A, and 2022-0085A). Experiments adhered to National Institutes of Health guidelines.

Preparation and cell transplantation for the α-syn PFF-induced PD model mice

Female ICR mice were anesthetized by Zoletil50 (0.1 mg/kg) with Rompun (93.2 g/kg). The α-syn-PFFs (1 μl, 5 mg/mL) were transplanted bilaterally into the Substantia Nigra (SNs) (±1.2 mm lateral to midline; 3.3 mm posterior to bregma; −4.6 mm ventral to dura) of 10-week-old mice. After 2 weeks, PBS (1.5 μl injection, sham control group), Md-AST (1.5 μl, 1ⅹ10⁵ cells/μl), or Fo-AST (1.5 μl, 1ⅹ10⁵ cells/μl) was transplanted bilaterally into the SNs of the mice. Mice received a daily injections of cyclosporine A (10 mg/kg, intraperitoneally) for 3–4 weeks starting 1 day before the grafting. Histological analyses were carried out 8–10 weeks post-transplantation.

Preparation and cell transplantation for the MPTP-induced PD model mice

Female ICR mice received intraperitoneally. injections of MPTP (30 mg/kg, Sigma-Aldrich) once daily for five consecutive days. Two to three days after the first MPTP injection, the mice were anesthetized by Zoletil50 (0.1 mg/kg) with Rompun (93.2–1 g/kg). PBS (3 μl injection, sham control group), Md-AST (3 μl, 1.5ⅹ10⁵ cells/μl), or Fo-AST (3 μl, 1.5ⅹ10⁵ cells/μl) was transplanted bilaterally into the SNs (±1.2 mm lateral to midline; 3.3 mm posterior to bregma; −4.6 mm ventral to dura) of the 10-week-old mice. The infusion was performed at a rate of 0.25 μl / min. The needle (26 g) was left in the injection site for 30 min after completion of each injection and removed slowly. The mice received daily injections of cyclosporine A (10 mg/kg, i.p.) for 3–4 weeks starting 1 day before the grafting. Histological analyses were conducted 12 weeks post-transplantation.

Preparation and cell transplantation for the Aβ-induced AD model mice

Aβ fibrils (1.5 μl, 100 μM) were injected bilaterally into the hippocampal dentate gyrus region (0.2 mm posterior to bregma; ± 0.1 mm lateral to the midline; −0.2 mm ventral to the dura) of 10-week-old female ICR mice under anesthesia induced by Zoletil50 (0.1 mg/kg) with Rompun (93.2 g/kg). Two weeks later, PBS (1.5 μl injection, sham control), Md-AST (1.5 μl injection, 1ⅹ10⁵ cells/μl), or Fo-AST (1.5 μl injection, 1ⅹ10⁵ cells/μl) was transplanted bilaterally into the hippocampus CA3 region (0.25 mm posterior to bregma; ±0.22 mm lateral to midline; −0.18 mm ventral to dura) of the mice. The infusion was performed at 0.25 μl/min. The needle (26 g) was left in the injection site for 30 min after each injection. The mice received daily injections of cyclosporine A (10 mg/kg, i.p. injection) for 3–4 weeks starting 1 day before the grafting. Histological analyses were carried out 8–10 weeks post-transplantation.

Animal behavior testing

To evaluate motor dysfunction, all behavioral experiments were performed using 3 trials after 2–3 days of training.

Rotarod test

Motor coordination and balance were measured by placing the animal on a rotating rod with accelerating rotation (4–44 rpm speed, 300 sec). The time the animal stayed on the rod was measured. If the animal succeeded for the full test, it was given the maximum score (300 sec, 300 score).

Pole test

Animals were placed on top of a vertical pole (50 cm long wooden pole, 1 cm in diameter) with their heads downward. The time taken to orient downward was measured in 3 trials. If the animal fell or slipped from the pole, it was given the highest value of the week.

Beam test

Motor coordination and balance were analyzed by placing the animal on a beam (square, 80 cm in length, 12 mm in diameter, 50 cm above the ground) and measuring the time required to traverse it. If the animal fell from the beam, it was given the highest value of the week.

Locomotor activity test

The locomotor activities of PD mice were monitored (HD C310, Logitech). The mice were placed in a cage and allowed to freely explore it for 20 min. The total distance traveled was measured automatically (Stoelting Co., IL, USA).

Statistical analysis

All data are expressed as the mean ± SEM (standard error of the mean). Statistical comparisons were performed using Student’s t-test, equal variance normal t-test, two-way or one-way ANOVA followed by Tukey’s multiple comparison test in GraphPad Prism 7.04. The statistical analysis methods, p-values, and n-values are indicated in the figure legends. A p-value less than 0.05 was considered significant.

Ethics statement

All experiments involving animals were conducted according to the ethical policies and procedures approved by the IACUC (Institutional Animal Care and Use Committee) of Hanyang College of Medicine (approval numbers 2021-0086A, 2021-0215A, and 2022-0085A).

Results

Generation of astrocytes from hPSC-derived midbrain organoids

To prepare human astrocytes with potent therapeutic capacity and the potential for efficient engraftment upon transplantation, we adopted the organoid-based method to prepare human midbrain-specific astrocytes from hPSCs (human ES cells, H9), as schematized in Fig. 1A and Supp. Fig. S1A. To that end, we first generated human midbrain-like organoids by differentiating hPSCs in 3D (3 dimensions) in the presence of the chemical combination for neural induction (SB431542, Noggin) and midbrain patterning (SHH/purmorphamine/FGF8b/CHIR999021), followed by the isolation of midbrain-specific NSCs from the organoids. The cultured midbrain NSCs proliferated stably through multiple cell passages. Consistent with the switch from neurogenic to astrogenic developmental propensity during brain development [23], the NSC culture in the early cell passages differentiated into neuronal cells, but the neurogenic potential of the midbrain NSCs was gradually switched toward GFAP+, S100B + astrocytes with increasing cell passages (Supp. Fig. S1B). Ultimately, the midbrain (Md)-NSC cultures that underwent > 15 passages (>130 days) uniformly differentiated into a homogeneous population of astrocytes expressing astrocyte-specific GFAP, S100B, aquaporin 4 (AQP4), and glutamate transporter (EAAT1) (Fig. 1B). As a control, forebrain (Fo)-type astrocytes were also prepared using the organoid-based method with forebrain patterning (Fig. 1A and Supp. Fig. S1A). Electrophysiologic analyses showed that both the Md- and Fo-astrocytes exhibited the linear current–voltage relationship, passive conductance, and low membrane resistance that endow astrocytes with their unique membrane properties (Fig. 1C). To further evaluate the transcriptome-based characteristics of the astrocytes derived from human brain organoids, we performed an RNA-sequencing (RNA-seq) analysis. A principal component analysis revealed that the transcriptomes of the Md- and Fo-astrocytes were similar to those of astrocytes primarily cultured from human brains, and their astrocytic maturities were positioned between those of adult and fetal human astrocytes (Fig. 1D). In DAVID-based analyses, genes commonly upregulated in the Md- and Fo-astrocytes (compared with undifferentiated hPSCs, >4 fold change (FC), FPKM > 1, p < 0.05) were enriched in the gene ontologies (GOs)/pathways associated with naïve astrocytic functions such as synaptic maturation, neurogenesis, brain extracellular matrix, negative regulation of neuronal apoptosis, response to toxic/oxidative/wounding stimuli, energy/glucose/lipid metabolism, microglial proliferation, and myelination (Supp. Fig. S2). Collectively, these findings indicate that both the Md- and Fo-astrocytes differentiated from hPSC-derived organoids acquired naïve astrocytic functions and characteristics.

Fig. 1.

Generation of astrocytes from human pluripotent stem cell (hPSC)-derived midbrain (Md)- and forebrain (Fo)-like organoids. (A) Schematic depicting the process of differentiating the Md and Fo organoid–based astrocytes. Human ESCs were induced to form region-specific Md- or Fo-patterned 3D organoids, followed by organoid dissociation and 2D plating. Neural stem cells (NSCs) from Md-like or Fo-like organoids were expanded through multiple passages (∼passage 15; ∼130 days) with basic fibroblast growth factor (bFGF). Detailed cytokine combinations are provided in Supplementary Information S1A. (B) Representative image of hPSC-derived regional astrocytes stained for astrocyte-specific markers (GFAP, AQP4, S100β, EAAT1). Scale bars are 50 μm. (C) Representative current traces of passive conductance (left) from −150 mV to 50 mV, recorded from Md- and Fo-derived astrocytes. Averaged traces of the I-V curve. Average resting membrane potential (mV) and current amplitude (pA) from −150 mV to 50 mV (right bar graphs). n = 3 biological replicates, 21–22 total cells per region patched. (D) Principle component analysis (PCA) of RNA-seq data from regional-patterned astrocytes (this study), human fetal- or adult-astrocytes [66], and hPSC Md-organoid NSCs (22). PCA scores were calculated using genes associated with astrocyte identity [67]. (E) Volcano plot showing the differentially expressed genes (DEGs) between the Md- and Fo-astrocytes. The plot displays the fold-change (x-axis) versus the significance (y-axis) of the genes identified in the RNA-seq data. The significance (p-value) and fold-change were converted to − Log10 and Log2, respectively. The vertical and horizontal dotted lines indicate the cut-off values for a fold-change >±3 and a p-value < 0.05, respectively. Blue dots are the genes increased in Md-astrocytes, and yellow dots are the downregulated genes. (F) Heatmap showing representative genes associated with Md and Fo development in Md- and Fo-astrocytes (n = 3 biological replicates; Supplementary Table S1). (G) Quantitative PCR analysis of genes associated with midbrain (FOXA2, LMX1A, NURR1, EN1) and forebrain (FOXG1, PAX6) development from Md-astrocytes (Md-AST) and Fo-astrocytes (Fo-AST). (H) Gene Ontology (blue bars) and KEGG pathway (green bars) analyses of the DEGs shown in (E). Significance at p < 0.05 with a DEG fold-change > 3. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Compared with Fo-astrocytes, 1046 genes were upregulated in the Md-astrocyte transcriptome, and 864 genes were downregulated (FPKM > 1, FC > 3, p < 0.05) (Fig. 1E). The brain region–specific patterning in the Md- and Fo-astrocytes was confirmed by their differential expression of region-specific genes in the transcriptome data (Fig. 1F), major factors of which were further validated by qPCR analysis (Fig. 1G). The GOs of the differentially expressed genes (DEGs) between the Md- and Fo-astrocytes were also categorized into naïve neuro-trophic/-supportive astrocytic functions of neuronal synapse (axonal guidance, synaptic plasticity), cellular homeostasis (cellular senescence, response to oxidative stress), and brain metabolism (cholesterol metabolism, fatty acid metabolism) (Fig. 1H). This indicates that, although the astrocytic functions were commonly acquired by the Md- and Fo-astrocytes, the levels of neurotrophic functions differed between region-specific astrocytes.

Cell-autonomous astrocytic functions rescue toxic environments and support neuronal glucose and cholesterol metabolism

The cell-intrinsic modes of astrocytic neurotrophic action include scavenging oxidative stress, extracellular toxic materials, and excitotoxic glutamate. When the astrocyte cultures were exposed to H₂O₂ (250 μM) for 24 hr (Fig. 2A), the ROS level, estimated using CellROX^TM, was lower in the cultured human Md-astrocytes than in the human Fo-astrocytes (Fig. 2B), indicating that the antioxidant capacity of the Md-astrocytes was superior to that of the Fo-astrocytes. Consequently, the Md-astrocytes exhibited greater resistance against ROS-induced cellular senescence than the Fo-astrocytes (Fig. 2C). The stronger ROS scavenging capacity of the Md-astrocytes was accompanied by higher expression of antioxidant genes, especially PARGC1A (PGC1α) and SOD2, which are powerful suppressors of ROS production, and lower expression of cell senescence–associated genes (Fig. 2D). We further observed that the cultured Md-astrocytes exhibited greater gene expression related to wound healing and greater wound healing capacity in a scratch injury model than the Fo-astrocytes (Supp. Fig. S3).

Fig. 2.

Midbrain (Md)-astrocytes have cell-autonomous functions to rescue toxic environments and support neuronal glucose and cholesterol metabolism to provide resistance against stressful conditions. (A) Schematic overview of the experimental procedure. (B–C) Md-astrocytes (Md-AST) and forebrain-astrocytes (Fo-AST) were exposed to H₂O₂ (250 μM) for 6 hr. Cellular ROS levels (B) were quantified by the intensity of CellROX staining, and cellular senescence (C) was estimated by quantifying the percentage of β-galactosidase (β-gal) + cells using the β-gal assay. Data are represented as the mean ± SEM. n = 4 culture coverslips. Scale bars are 25 or 50 μm. (D) Heatmaps displaying the expression of genes associated with anti-oxidation and cellular senescence in the RNA-seq data (Md-AST vs. Fo-AST). (E) Heatmap illustrating the upregulation of phagocytosis-related genes in Md-AST. (F–H) Phagocytic potential of cultured Md-AST. (F) General phagocytic activity was assessed using fluorescent latex beads (F-13081). Cultured astrocytes were incubated with the latex beads for 90 min and washed, and then the percentage of cells engulfing beads was quantified. Phagocytic clearance of α-syn (G) and amyloid β (Aβ) (H). Astrocytes were incubated with Alexa 488 labeled α-syn or Alexa 488-labeled Aβ for 1 day. The percentage of cells containing Alexa 488 was monitored for 6 days after withdrawal of the fluorescence-labeled α-syn or Aβ. Significant differences from Fo-AST are indicated at *P < 0.05, #P = 0.1. n = 4. One-way ANOVA. Scale bars are 25 or 50 μm. (I) Heatmaps showing gene expression from Md-AST and Fo-AST related to lactate metabolism, glutamate transport, and cholesterol synthesis & efflux. (J–M) Metabolic potentials were analyzed in Md-AST and Fo-AST. Md-AST released higher levels of lactate and cholesterol into the extracellular space and took up more glutamate. For the glutamate uptake assay (J) and total cholesterol assay (L), normally cultured Md-AST and Fo-AST were used. The lactate (K) and cholesterol efflux (M) assays were performed using astrocyte-conditioned media. Significant differences between the astrocyte groups p < 0.05*, 0.0001**, p = 0.15^#. t-test (B,C,F,J-M). Data are presented as the mean ± SEM. n = 3–5 independent experiments in duplicate or triplicate.

In the RNA-seq data, gene expression specific for phagocytosis was greater in the cultured Md-astrocytes than in the Fo-astrocytes (Fig. 2E). Consistently, compared with the Fo-astrocytes, the human Md-astrocytes exhibited greater phagocytic activity in an in vitro phagocytic assay using fluorescent latex beads (F-13081, Thermo Fisher) (Fig. 2F). Alpha-synuclein (α-syn) protein aggregation is a common pathological hallmark of Parkinson’s disease (PD), Lewy body dementia, and multiple system atrophy. Astrocytes have the capacity to eliminate extracellular α-syn aggregates via phagocytic clearance [5], [24], [25]. To assess the phagocytic clearance of aggregated α-syn proteins, the hPSC-derived astrocyte cultures were incubated with human α-syn preformed-fibrils (α-syn-PFFs) labeled with Alexa-488 for 24 hr. Both the Md- and Fo-astrocytes efficiently engulfed the fluorescence-labeled α-syn, peaking 2 days after removal of the α-syn, which was completely cleared over the next 5 days (Fig. 2G). However, the maximum amount of α-syn engulfed by the Md-astrocytes was greater than that by the Fo-astrocytes. Similarly, the Md-astrocytes exhibited a trend of higher phagocytic clearance capacity for FITC-labeled amyloid β (Aβ) aggregates, specific to Alzheimer’s disease (AD), although not reaching significance (p = 0.1 at day 2; Fig. 2H).

Astrocytes rescue neuronal cells from excitatory toxicity by scavenging excess excitatory glutamate neurotransmitters [26], [27]. Along with the upregulated expression of glutamate transporter genes (SLC1A1, SLC1A2, SLC6A6, SLC6A15, SLC6A16) (Fig. 2I), the Md-astrocytes exhibited more potent glutamate uptake capacity than the Fo-astrocytes (Fig. 2J). The other gene set that attracted our interest was involved in lactate metabolism because astrocytes physiologically support the high energy demands of neurons by providing lactates through the astrocyte-neuron lactate shuttle [28], [29]. Interestingly, along with higher expression of lactate metabolic genes (Fig. 2I), the lactate level secreted by cultured human Md-astrocytes was higher than that from the Fo-astrocytes (Fig. 2K), indicating the greater capacity of the Md-astrocytes to support neuronal glucose/energy metabolism. Astrocytes highly synthesize cholesterol, and neurons depend on cholesterol from astrocytes for their lipid metabolism [30] and myelin formation [31]. The capacity for cholesterol synthesis is regarded as a characteristic of young and healthy astrocytes with potent neurotrophic potential [32], [33]. Reduced cholesterol levels are found in the brains of neurological animal disease models, and there are some strategies that can increase cholesterol levels to restore neuronal physiology [34]. We noticed that cholesterol homeostasis was one of the significant GOs in the DEG analysis, with the cholesterol homeostatic genes (HMGCS1, SREBF1, APOE, ABCA1) upregulated in the Md-astrocytes (Fig. 2I). In the in vitro assay, the total intracellular cholesterol level in the Md-astrocyte cultures was significantly greater than in the Fo-astrocyte cultures (Fig. 2L), but the cholesterol efflux capacities did not differ significantly between the two types (Fig. 2M). Collectively, these findings suggest that the Md-astrocytes differentiated from hPSC-derived organoids acquired innate neurotropic and repair capacities that were more potent than those of the Fo-astrocytes.

Paracrine astrocytic functions to protect against neurodegeneration and pathologic microglial polarization

In addition to their cell-autonomous functions, astrocytes exert neurotrophic actions in a paracrine manner by releasing neurotrophic and anti-inflammatory cytokines [35]. To assess the paracrine neuroprotective activity of the cultured cells, medium was conditioned in the hPSC-derived Md - and Fo-astrocyte cultures for 10 days (collected every other day and combined), and the conditioned medium (CM) was added to hPSC-derived human neuronal cultures pretreated with the ROS-producing agent hydrogen peroxide (H₂O₂, 250 μM, 6 hr) (Fig. 3A). Six days after the astrocyte-CM (ACM) treatment, neuronal oxidative stress, estimated by CellROX (mean fluorescence intensity, MFI), TUJ1 + neuronal cells, and β-galactosidase + cellular senescence, was greatly reduced (Fig. 3B, C). Md-ACM had a greater antioxidant effect than Fo-ACM. In addition, oxidative stress–induced neuronal cell death, estimated by counting TUJ1 + neurons, was also greatly reduced in the presence of ACM (Fig. 3D). H₂O₂ treatment induced the shortening and fragmentation of TUJ1 + neurites, as well as the loss of synapsin + synaptic puncta. That neurite and synaptic degeneration were ameliorated by the ACM treatment and more significantly by the Md-ACM (Fig. 3D, E).

Fig. 3.

Factors secreted from midbrain (Md)-astrocytes protect neurons against toxic stimuli and prevent pathologic microglial polarization. (A) Schematic overview of the experimental procedure. To investigate paracrine effects, medium was conditioned in Md or forebrain-astrocyte (Md-AST or Fo-AST) culture for 10 days (collecting media every other day). The conditioned media (Md-ACM and Fo-ACM) were then added to neurons treated with H₂O₂ (250 μM for 6 hr) and cultured for 6 days (B–E) or to microglia treated with LPS (250 ng/mL) and cultured for 48 hr (F–I). (B–C) Paracrine astrocytic functions in protecting cellular ROS levels (B) and preventing cellular senescence (C) in neurons were analyzed. Scale bars are 25–50 μm. (D) Neuronal cell maintenance was quantified by counting the percentage of Tuj1 + cells and neuronal fiber. (E) Neuronal maturation was assessed as the density of synaptic puncta. Data are presented as the mean ± SEM. Significant differences from the non-ACM-treated control p < 0.01*, 0.001**, between the ACM-treated groups p < 0.01#, 0.001##. One-way ANOVA, followed by Tukey’s analysis. Scale bars are 25 μm. (F–I) Md-ACM had anti-inflammatory effects and prevented senescence through secreted factors. (F) Immunocytochemical analysis for microglial (Iba1 + ) immunoreactive pro-inflammatory/cytotoxic factors (IL-1β). (G) qPCR analysis of pro-inflammatory factor expression in LPS-activated microglia treated with Md-ACM or Fo-ACM. (H–I) The senescence level of the microglia was identified by the expression of P21, a senescence marker (H), and SA-β-gal activity (I). Data are indicated as the mean ± SEM. Significant differences from the WT #p < 0.001 and among the groups specified at p < 0.01*, 0.001**. One-way ANOVA, followed by Tukey’s analysis. Scale bars are 25 or 50 μm. (J) Gene expression analysis revealing the DEGs of candidate factors responsible for the differential paracrine actions of inflammatory cytokines and neurotrophic factors between human Md-AST and Fo-AST. The identified DEGs are listed in the heatmap.

Astrocytes interact closely with microglia, the resident brain immune cells, to regulate the brain environment. Astrocyte-microglia interactions are mainly mediated through the release of paracrine factors [35], [36]. To test the ability of the cultured astrocytes to modulate brain inflammation, we prepared human microglia cultures by differentiating hPSCs (described in ‘Materials & Methods’). LPS treatment efficiently polarized human microglia into their IL-1β-expressing pro-inflammatory form, upregulating pro-inflammatory cytokines (IL-1β, IL6, iNOS, TNFα) (Fig. 3F,G) and downregulating anti-inflammatory cytokine (IL-10) expression (Fig. 3G). The pro-inflammatory microglial polarization was greatly blunted in the presence of ACM (Fig. 3F,G). The Md-ACM had significantly greater effects than the Fo-ACM in rescuing the LPS-mediated changes of IL-1β + microglia (Fig. 3F). Furthermore, upon relatively long-term LPS treatment (250 ng for 2 days), microglial cell senescence was manifested by an increase of p21 + and β-galactosidase + senescent microglia 2 days after LPS removal (Fig. 3H, I). That microglial cell senescence was suppressed by Fo- and Md-ACM but more highly by Md-ACM (Fig. 3H, I). Microglial cell senescence is a key pathogenic mechanism and therapeutic target in neurodegenerative disorders [37], [38]. Candidate factors responsible for the differential paracrine actions of the human Md- and Fo-astrocytes are listed in Fig. 3J.

Therapeutic capacity of hPSC-derived Md-astrocytes in in vitro models of PD and AD

Based on the observed astrocytic cell-autonomous and paracrine activities, we next compared the roles of the Md- and Fo-astrocytes in in vitro human brain disease models containing hPSC-derived neurons, astrocytes, and microglia [39]. For the in vitro PD model, hPSCs were differentiated into midbrain-type dopamine (Md-DA) neurons, the neuronal type primarily affected in PD. Then, a tri-culture containing human Md-DA neurons, microglia, and Md-astrocytes was exposed to α-syn PFF (1.5 μg/mL, for 14 days) (Fig. 4A). At the same time, hPSCs were induced to differentiate into neurons specific for the forebrain (cortex), the brain region most affected in AD, and those neurons were co-cultured with hPSC-derived microglia and Fo-astrocytes. To induce AD-associated pathologies based on the amyloid cascade hypothesis [40], tri-cultures containing the forebrain cell components were treated with Aβ fibrils (1.5 μM, for 14 days) (Fig. 4I). We assessed the effects of replacing Fo-astrocytes with Md-astrocytes in the AD model (and Md-astrocytes with Fo-astrocytes in the PD model) on the pathologic manifestations of each disease. In the PD model, the substitution of Fo-astrocytes for Md-astrocytes resulted in more abundant α-syn aggregation, as shown by assays that used thioflavin S dye to visualize misfolded protein aggregates (Fig. 4B) and an immunocytochemical analysis against pS129-α-syn (phosphorylated α-syn at serine 129), a pathologically modified form of α-syn associated with toxic α-syn fibril formation [41](Fig. 4C). The greater α-syn aggregation in the Fo-astrocyte-containing tri-cultures was further confirmed by a Western blot (WB) analysis that displayed greater band intensities for α-syn aggregates in both the Triton X-100-soluble and -insoluble fractions of Fo-astrocyte-containing cultures (Fig. 4D). In contrast, when Md-astrocytes were included in the AD culture model instead of Fo-astrocytes, pathologic Aβ aggregation, estimated by thioflavin S staining (Fig. 4J) and WB analysis for Aβ proteins (Fig. 4K), was significantly decreased. These findings collectively indicate that Md-astrocytes have stronger protective functions against pathologic protein aggregation than Fo-astrocytes in both PD and AD tri-culture models.

Fig. 4.

Midbrain astrocytes (Md-AST) ameliorate PD- and AD-associated pathologies in an in-vitro culture model. (A) Schematic overview of the PD -associated experimental procedure. The effects of the cultured Md-AST on the levels of α-syn pathologies (B-H) were analyzed. Fourteen days after α-syn PFF (1.5 μg/mL) treatment, α-syn aggregation was detected as immunoreactivity against thioflavin-S (ThS) (B) and pS129-α-syn staining (C). Scale bars are 50 μm. *P < 0.05. t-test (D) Intracellular protein levels of α-syn monomers and oligomers were determined using a Western blot (WB) analysis. Data are represented as the mean ± SEM of protein levels relative to β-actin (n = 3). (E) Expression of neurotrophic and inflammatory genes in the presence of the α-syn PFF during co-culture with Md-AST or forebrain-astrocytes (Fo-AST), estimated by qPCR analyses. *P < 0.05. Student’s t-test. n = 3. (F) Immunoblot analysis to assess inflammasome activation. To activate the inflammasome pathway, PD-context model cells were treated with LPS (3 hr; 0.25 µg/mL) and then ATP (2.5 mM, 30 min) before conditioned-medium (CM) was collected. The levels of pro- and activated-IL-1β and caspase-1proteins released in the culture media were determined. The protein levels were measured in the media and normalized to Ponceau S-stained total protein levels. Significant differences from the LPS + ATP-treated(+)/Md-AST co-culture p < 0.05*. n = 6 independent cultures. t-test. (G, H) The effects of Md-AST on α-syn-induced neuronal degeneration were assessed by counting TH + DA neurons (G) and quantifying synaptic puncta per 100 μm (H). *p < 0.05, t-test. n = 3 independent experiments in triplicate. Scale bars are 25 or 50 μm. (I) Schematic overview of the AD -associated experimental procedure. (J-N) The effects of the cultured Md-AST on the levels of Aβ pathologies were analyzed. (J-K) Aβ aggregation was detected as immunoreactivity against thioflavin-S (ThS) (J) and WB analysis (K). *P < 0.05. t-test. (L) Expression of neurotrophic and inflammatory genes in the presence of Aβ fibrils during co-culture, estimated by qPCR analyses. *P < 0.05. Student’s t-test. n = 3. (M) The activated-inflammosome protein levels were measured in the AD-context model. The levels of proteins released in the culture media were determined. Significant differences from the LPS + ATP-treated(+)/Md-AST co-culture p < 0.05*. n = 6 independent cultures. Student’s t-test. (N) The effects of Md-AST on Aβ -induced neuronal degeneration were assessed by quantifying synaptic puncta per 100 μm. Scale bars are 25 or 50 μm. *P < 0.05. t-test.

Another common pathology in PD and AD is neuroinflammation, which positively interacts with α-syn and Aβ proteopathies in a feed-forward manner in ([42], [43], [44]). Consistently, administering α-syn-PFF or Aβ fibrils to hPSC-derived microglia cultures induced microglial polarization toward the expression of pro-inflammatory cytokines (Supp. Fig. S4). That microglial polarization was rescued in the presence of ACM, more highly by treatment with Md-ACM than Fo-ACM (Supp. Fig. S4). In addition to the astrocyte paracrine effect preventing microglial polarization, the expression of pro-inflammatory cytokines was lower in PD- and AD-tri-culture models harboring Md-astrocytes than in those containing Fo-astrocytes, and the levels of neurotrophic factors were significantly greater in the Md-astrocyte-containing models (Fig. 4E, L). Furthermore, weaker activation of the NLR family pyrin domain containing 3 (NLRP3) inflammasome, a central pathogenic contributor to neurodegenerative diseases[45], was found in the Md-astrocyte-containing PD- and AD-culture models, as shown by WB determination of the activated and released caspase-1 and IL-1β protein levels in the culture media (Fig. 4F,M).

Along with the observed astrocyte actions for treating α -syn proteopathy and neuroinflammation, the number of TH + Md-DA neurons and the synapsin + synaptic density on TuJ1 + neurites were significantly greater in the Md-astrocyte-containing PD model than in the Fo-astrocyte-PD model (Fig. 4G, H). Furthermore, greater synaptic density was observed in the AD model containing Md-astrocytes than in that containing Fo-astrocytes (Fig. 4N). Together, these findings indicate that, regardless of the disease context, hPSC-derived Md-astrocytes had greater therapeutic potency than Fo-astrocytes for treating α-syn- or Aβ-induced proteinopathy, inflammation, and neurodegeneration.

Therapeutic potential of hPSC-derived astrocytes in mouse models of PD

We next attempted to assess the therapeutic effects of hPSC-derived human astrocytes in vivo in PD animal models. Utilizing cultured astrocytes for transplantation represents a beneficial therapeutic approach for addressing neurodegenerative disorders. (reviewed in [13]), so we transplanted hPSC-derived Md- (or Fo-) astrocytes into MPTP-induced PD mice, the best-characterized PD animal model [46]. Intraperitoneal injection of MPTP (30 mg/kg, i.p., for 5 consecutive days; Fig. 5A) leads to a gradual but massive loss of Md-DA neurons in the midbrain: 70–80 % of TH + Md-DA neurons are lost by 2 months after MPTP injection. The midbrain neuron loss was hindered in the MPTP-PD mice that received bilateral intra-nigral transplantation of human astrocytes (Fig. 5B,C). Compared with the wild-type (WT) control mice, the TH + Md-DA neurons that remained in the substantia nigras (SNs) of MPTP-PD mice displayed smaller cell bodies (Fig. 5D) and blunted neurites (Fig. 5E), the morphologic characteristics of degenerating neurons (Fig. 5B). Furthermore, the Md-DA neurons in the MPTP-PD mice showed reduced expression of the midbrain-specific Nurr1, which is sensitively lost in a PD context prior to nigral DA neuron death [17], [47], [48], [49] and is used as a marker of early DA neuron degeneration (Fig. 5F, G). The degenerating Md-DA neuron indices were greatly rescued by astrocyte transplantation, more dramatically by Md-astrocytes than Fo-astrocytes (Fig. 5B–G). Accordingly, nigrostriatal DA innervation, estimated using TH + fiber intensity (MFI) in the striatum, was also significantly protected by the astrocyte transplantation, and more greatly in the mice grafted with Md-astrocytes (Supple Fig. S5). Along with those strong neuroprotective effects, bilateral intra-nigral transplantation of Md- or Fo-astrocytes greatly alleviated the motor deficits of PD mice, as shown by the pole (Fig. 5H), beam (Fig. 5I), rotarod (Fig. 5J), and locomotor (Fig. 5K) tests.

Fig. 5.

In vivo therapeutic functions of cultured midbrain astrocytes (Md-AST) in MPTP- (A–K) and α-synuclein (L–T)-induced PD model mice. (A) Schematic of the experimental procedure to generate MPTP-induced PD model mice. MPTP PD model mice were generated by MPTP i.p. injection (30 mg/kg) for 5 days. Two to three days after starting the MPTP injections, PBS (sham control), Md-AST, or forebrain-astrocytes (Fo-AST) were injected into both the left and right sides of the substantia nigra (SN). At 12 weeks post-transplantation, the recovery of DA neuron degeneration was assessed using immunohistochemistry and behaviors tests. (B) Representative images of transplantation effect of Md-AST or Fo-AST on PD pathologies in the SN region of the MPTP model. TH + neuronal loss (C), soma size of TH + Md-DA neurons (D), and neurite length (E) were analyzed. The effect of Md-AST transplantation on TH + Md-DA neurons in recovering the toxin-induced loss of Nurr1 was assessed by quantifying the percentage of Nurr1 + TH + cells (F) and the mean fluorescence intensity (MFI) of the Nurr1 signal, assessed by ImageJ (G). Significant differences from the WT #p < 0.001, and among the groups specified at p < 0.05*, 0.01**. n = 3–7 animals. One-way ANOVA, followed by Tukey’s analysis. Scale bars are 100 μm. (H–K) Behavior tests. The behaviors of the MPTP-induced PD model mice were assessed using pole (H), beam (I), rotarod (J), and locomotor activity (K) tests 12 weeks post-transplantation. Significant differences from the WT #p < 0.05 and between the groups p < 0.05*, p < 0.01**. n = 5–12 animals. One-way ANOVA, followed by Tukey’s analysis. Data are indicated as the mean ± SEM. (L) Schematic of the experimental procedure for the α–synuclein PD (α-syn-PD) model. The α-syn-PD model mice were generated by a bilateral injection of α-syn PFF (5 mg/mL) into the midbrain SNs. Two weeks later, cultured Md-AST or Fo-AST were transplanted into the midbrain SNs of the PD mice (or sham-control PBS injection into the SN). (M−N) Eight to ten weeks post-transplantation, α-synucleinopathy assessed as the percent of pS129-αsyn + TH + cells. The boxed areas in each image (1, 2, 3) are enlarged in the lower panel. Scale bars are 25 or 100 μm. (O–Q) DA neuron degeneration assessment. DA neuron degeneration was assessed by quantifying the numbers (O), soma sizes (P), and fiber lengths (Q) of TH + DA neurons (immunofluorescence stained). (R) Detection of α-syn aggregates by a Western blot (WB) analysis of α-syn (SN homogenates; n = 3–4 animals). Monomer and aggregate forms of α-syn were detected in the Triton X-100(Tx-100)–soluble and –insoluble (SDS soluble) fractions. The WB analysis was performed with 3 experimental replicates. (S-T) Assessment of local inflammation in the grafted SNs (host brain regions neighboring the grafts) compared with sham-operated (PBS-injected) areas of α-syn-PD mice and WT mice. (S) qPCR data showing Md-AST-mediated downregulation of pro-inflammatory cytokines and upregulation of neurotrophic factors at the graft–host interfaces of α-syn-PD-mouse SNs. (T) Immunohistochemical analyses of microglia (Iba1 + ) immunoreactive for pro-inflammatory/cytotoxic factors (CD11b and CD68). Immunoreactive cells along the host-graft interfaces were counted in 6 cryosectioned slices from three animals in each group. Data are expressed as percentages of immunoreactive cells in the Iba1 + microglial populations. Scale bars are 100 μm. α-syn-PD model experiments: n = 3–7 animals. Significant differences from the WT #p < 0.05, between the specified groups *p < 0.05, **p < 0.01. Two-way ANOVA, followed by Tukey’s analysis. Data are presented as the mean ± SEM.

We further assessed the in vivo capacity of the hPSC-derived astrocytes to treat PD-specific α-synucleinopathy in a PD mouse model generated by bilaterally injecting α-syn PFF into the SNs of mice [50](Fig. 5L). Alpha-syn pathology, neuroinflammation, and Md-DA neuron degeneration were manifested in the animal model (Fig. 5M–T). Two weeks after the α-syn injection, hPSC-derived Md- (or Fo-) astrocytes were transplanted into the SNs of the PD mice (Fig. 5L). When the astrocyte engraftment was assessed, the total numbers of human-specific GFAP (hGFAP) + astrocytes at 10 weeks post-transplantation were 7.2 ± 0.9 X 10⁴ cells/graft and 6.8 ± 1.1 X 10⁴ cells/graft in the mice transplanted with Md-astrocytes and Fo-astrocytes, respectively. Compared with phosphate-buffered saline (PBS) sham-operated controls, α-syn aggregates, detected by pS129-α-syn immunoreactivity, were significantly reduced in the SNs of PD mice grafted with Md-astrocytes (% of pS129-α-syn + TH + cells 10 weeks after transplantation: 10.42 ± 1.8 % (PBS-sham) vs. 4.09 ± 0.8 % (Md-AST) vs. 4.77 ± 0.2 % (Fo-AST); Fig. 5M,N). The astrocyte transplantation effect to treat α-syn pathology was further confirmed by WB analyses that showed significantly lower levels of α-syn aggregates (in Triton X-100 soluble fractions) and monomers (both in the soluble and insoluble fractions) in the SNs of PD mice treated with human astrocytes than in the sham-operated PD mice (Fig. 5R). In the SNs of the mice grafted with the astrocytes, the α-syn-induced loss of TH + DA neurons was prevented (Fig. 5O), and the surviving DA neurons looked healthier with greater soma sizes (Fig. 5P), and the neurite outgrowths (Fig. Q) than in the PBS sham-operated PD mice. Furthermore, the astrocyte transplantation corrected the neuroinflammatory environments in the SNs of the PD mice, which was evidenced by RT-PCR analyses exhibiting the downregulation of pro-inflammatory cytokines (TNFα, IL-1β, IL-6, iNOS) and the upregulation of neurotrophic factors (BDNF, GDNF, SHH) in the grafted SNs (Fig. 5S), as well as a significantly lower percentage of microglia expressing the CD11b and CD68 pro-inflammatory cytokines (Fig. 5T). The anti-inflammatory effects of human astrocyte transplantation also manifested in the MPTP-PD mice (Supp. Fig. S6). As in the in vitro assays, the Md-astrocytes tended to show greater therapeutic effects than the Fo-astrocytes in the in vivo PD mouse models, and significant differences were found in the TH + fiber length (Fig. 5E, Q), nigrostriatal DA neuronal innervation (TH intensity; Supple Fig. S5), and neuroinflammatory protection (Supp. Fig. S6 and Fig. 5S, T).

Md-astrocytes alleviate Aβ aggregation, inflammation, and synaptic loss in an AD mouse model

Finally, we tested the effects of astrocyte transplantation in an in vivo AD context. To this end, we generated an AD brain environment by bilaterally injecting Aβ fibrils into the hippocampi of mice [51], [52]. Two weeks later, cultured Md- (or Fo-) astrocytes were transplanted into the hippocampi of the AD mice (Fig. 6A), resulting in 7.0 ± 2.0 X 10⁴ cells/graft (Md-astrocytes) and 6.8 ± 1.5 X 10⁴ cells/graft (Fo-astrocytes) engrafted hGFAP + astrocytes at 10 weeks post-transplantation. Accumulation of aggregated Aβ proteins in the AD mouse hippocampi was significantly reduced by the astrocyte transplantation in the quantification of the areas and numbers of Aβ + and thioflavin + plaques (Fig. 6B) and in a WB analysis for Aβ (Fig. 6C) 10 weeks after transplantation (3 months after Aβ injection). Furthermore, pro-inflammatory cytokine expression and microglial activation in the Aβ-injected hippocampi were significantly ameliorated by astrocyte transplantation (Fig. 6D, E). Along with the aggregated Aβ deposition and neuroinflammation, SYNAPSIN 1 + puncta were lost in the neurons of the Aβ-treated mouse hippocampi (Fig. 6F). The synaptic degeneration was significantly prevented in the mouse grafted with the astrocytes (Fig. 6F). The transplantation of Md-astrocytes produced significantly greater therapeutic effects than Fo-astrocyte transplantation in Aβ aggregation (Fig. 6B, C) and neuroinflammatory protection (Fig. 6D, E) assays.

Fig. 6.

Alzheimer’s disease (AD)-associated Aβ pathology and inflammation ameliorated in AD mice transplanted with midbrain- or forebrain-astrocytes. (A) Experimental schematic for the midbrain- or forebrain-astrocytes (Md-AST or Fo-AST) transplanted in AD model mice. The AD pathologic model mice were generated by a bilateral injection of Aβ fibrils (100 μM) into the hippocampus. Two weeks later, cultured Md-AST or Fo-AST were transplanted into the hippocampal regions of the AD mice (or sham-control PBS injection into the hippocampus). At 8 to 10 weeks post-transplantation, Aβ pathology and inflammation in hippocampi with astrocyte transplants were compared with those in sham-operated hippocampi. (B) Therapeutic effects of Md-AST or Fo-AST on pathologic Aβ accumulation in the hippocampal regions of AD model mice. The Aβ aggregates were detected by counting Aβ- and thioflavin S–double positive (Aβ+/thioflavin S + ) puncta. (C) Detection of Aβ aggregates by Western blot analysis for Aβ (hippocampus homogenates; n = 3–4 animals). Monomer and aggregate forms of Aβ in the SDS soluble fractions. (D–E) Assessment of local inflammation in the grafted hippocampi (host brain regions neighboring the grafts) compared with sham-operated (PBS-injected) Aβ-AD and WT mice. (D) Immunohistochemical analyses of microglia (Iba1 + ) immunoreactive for pro-inflammatory/cytotoxic factors (CD11b and CD68). Immunoreactive cells along the host-graft interfaces were counted in 3 to 4 animals in each group. Data are expressed as percentages of the immunoreactive cells within the Iba1 + microglial populations. (E) qPCR-based determination of neurotrophic and pro-inflammatory cytokine expression in the graft–host interfaces of AD model mice transplanted with Md-AST or Fo-AST vs. the PBS-sham operated or WT-control. n = 5–8 samples. Significant differences from the WT mice #p < 0.001, between the specified groups *p < 0.05, **p < 0.0001. One-way ANOVA, followed by Turkey post hoc analysis. Scale bars are 100 μm. (F) Synaptic maturation of TUJ1 + neurons, estimated as synapsin + puncta density. Significant differences from the WT mice #p < 0.001, between the groups **p < 0.001. One-way ANOVA. Scale bars are 25 μm.

Discussion

In this study, our objective was to generate human astrocytes with robust therapeutic potential from hPSCs. To achieve that, we combined ventral midbrain regional patterning and organoid-based differentiation methods for hPSCs. Through a series of in vitro and in vivo assays, we demonstrated that the human Md-astrocytes exhibited cell-autonomous capabilities for treating oxidative insult, excitotoxic glutamate, and pathologic protein aggregates superior to those of the Fo-astrocytes used as controls. In addition, the enhanced capacities of the Md-astrocytes included paracrine functions to inhibit neuronal death/senescence and microglial polarization toward a pro-inflammatory state. Ultimately, our human astrocytes successfully ameliorated neurodegeneration, neuroinflammation, and pathologies associated with α-syn and Aβ protein accumulation in both in vitro in human cells and in vivo in mouse models of PD and AD.

The superior therapeutic functions of the Md-astrocytes are likely attributable to the actions of midbrain-specific factors known to suppress inflammation, oxidative stress, and cellular dyshomeostasis. One such factor is Nurr1, which has been demonstrated to inhibit the expression of pro-inflammatory genes through a process called trans-repression that is mediated by the Nurr1-CoREST epigenetic repressor complex[53]. Our research has revealed that another midbrain-specific factor, Foxa2, synergizes with Nurr1 to suppress glial inflammatory signals by downregulating serum/glucocorticoid-related kinase 1 [54], an intracellular signaling molecule involved in activating the NF-κB-mediated inflammatory pathway[17], [18], [55]. Additionally, the midbrain-specific Lmx1a and Lmx1b factors promote autophagic gene programs and enhance autophagic-lysosomal function[56]. Furthermore, LMX1A, EN1, and NURR1 have been shown to support mitochondrial biogenesis and function [57], [58]. These auto-lysosomal and mitochondrial functions are crucial for maintaining physiological neurotrophic glial functioning, and related impairments can lead to glial polarization toward pro-inflammatory and neurotoxic phenotypes [59], [60], [61].

Another crucial aspect to consider when preparing therapeutic cells derived from hPSCs is the stability of the differentiation protocol and its ability to minimize cellular stress during cell generation. This concern becomes particularly relevant when preparing midbrain-type cells because ventral midbrain patterning in differentiating hPSCs requires close cell–cell contact and a high cell density, along with the use of multiple patterning chemicals that can be more or less toxic to the cells [62]. In such challenging culture conditions, cellular stresses tend to accumulate in differentiating hPSCs, leading to compromised cell viability and loss of the desired pattern in the resultant cells [22], [63]. To address that issue, we used a 3D organoid culture system that emulates the cell–cell interactions, organization, and functional characteristics of the developing human midbrain. By patterning hPSCs in the organoid culture, we successfully generated ventral midbrain–patterned precursor cells from midbrain-like organoids [22]. This approach not only facilitated the efficient proliferation and differentiation of hPSC-derived astrocytic precursor cells into mature astrocytes, but also enabled their storage in liquid nitrogen for future use. Moreover, transplanting astrocytes generated with reduced cellular stress could significantly improve donor cell survival and engraftment.

In the absence of disease-modifying therapy for CNS disorders, a growing body of research supports the therapeutic benefits of implanting astrocytes and their derivatives (as reviewed in [13]). The present study is important because it focuses on the development of clinically relevant human astrocytes with potent therapeutic capabilities and verifies their efficacy across multiple disease scenarios. Our findings provide robust evidence supporting the clinical adoption of astrocyte-based strategies for treating brain disorders. Further investigations are warranted to explore the deeper therapeutic potential of astrocytes, particularly using disease models that more accurately recapitulate human neurodegenerative disorders. Given the substantial species-dependent differences in the immune system, which are major contributors to neurodegenerative disorders [64], [65], future assessments of astrocyte therapeutic potential should consider disease models with humanized immune systems.

CRediT author statement

Mi-Yoon Chang: Conceptualization, Funding acqusition, Data curation, Supervision, Writing and Resources. Visualization, Investigation and Validation. Sang-Hun Lee: Conceptualization, Funding acqusition, Data curation, Supervision, Writing and Resources. Minji Kang: Visualization, Investigation and Validation. Minji Kim: Visualization, Investigation and Validation. Min Jong Seok: Visualization, Investigation and Validation. Yunseon Yang: Visualization, Investigation and Validation. Young Eun Han: Visualization, Investigation and Validation. Do Gyeong Kim: Visualization, Investigation and Validation. Soo-Jin Oh: Formal analysis, Supervision and Resources. Hyeon Son: Formal analysis, Supervision and Resources.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following are the Supplementary data to this article:

Data Availability

All data in the paper are present in the paper and/or the Supplemental Information. Additional data related to this paper may be requested from the authors.