hiPSC‐Derived Astrocytes From Individuals With Schizophrenia Induce a Dystrophic Phenotype in Microglial‐Like Cells

ABSTRACT

Neuroinflammation, particularly astrocyte reactivity, is increasingly linked to schizophrenia (SCZ). Yet, the crosstalk between astrocytes and microglia in SCZ, especially under pro‐inflammatory conditions, remains unclear. Here, we employed human induced‐pluripotent stem cells to compare how astrocytes from five age‐matched individuals with SCZ and five neurotypical controls, upon stimulation with TNF‐α, affected microglial biology. TNF‐α stimulation of SCZ astrocytes, relative to their control counterparts, triggered increased mRNA expression of pro‐inflammatory cytokines and CX3CL1. Interestingly, transcriptomic and gene set enrichment analyses revealed that reactive SCZ astrocytes promoted the downregulation of biological processes associated with immune cell proliferation and activation, phagocytosis, and cell migration in induced microglial‐like cells (iMGs). Under such conditions, iMGs assumed a dystrophic/senescent‐like phenotype, which was associated with accelerated transcriptional aging. Functional validations showed that TNF‐α‐stimulated SCZ astrocytes promoted reduced synaptoneurosomes phagocytosis by iMGs. Interestingly, while both reactive control and SCZ astrocytes were capable of inducing significant microglial migration in a CX3CR1‐dependent manner, TNF‐α‐stimulated SCZ astrocytes failed to promote greater iMG chemotaxis, compared with their stimulated control counterparts, despite secreting more than twice as much CX3CL1. This was likely due to SCZ astrocytes triggering reduction in CX3CR1 plasma membrane levels in iMGs. Altogether, these findings suggest that astrocytes contribute to SCZ pathology by altering normal microglial function and inducing a dystrophic phenotype.

• SCZ astrocytes display an increased pro‐inflammatory profile upon TNF‐α stimulation.

• Reactive SCZ astrocytes induce microglial‐like cells to assume a dystrophic phenotype.

• Reactive SCZ astrocytes limit iMGs' synaptoneurosome uptake and migration.

1. Introduction

Schizophrenia (SCZ) is a neurodevelopmental disorder with an estimated lifetime prevalence ranging between 0.3% and 0.7% worldwide (Solmi et al. 2023). Symptoms usually appear between late adolescence and early adulthood (Hafner et al. 1993) and can be divided into three broad categories: positive symptoms, such as hallucinations, psychosis, and delusions; negative symptoms, for example, depressive behavior, distraught thoughts, and social withdrawal; and cognitive deficits, including working memory impairments, learning disabilities, and attention deficits (Cannon 2015; Perez and Lodge 2014; Sakurai et al. 2015).

The causes of SCZ remain poorly understood, as both genetic and environmental risk factors interact to generate susceptible conditions towards the onset of this multifactorial disorder later in life (Hilker et al. 2018; Lipner et al. 2019; Muller et al. 2015; Nimgaonkar et al. 2017). Environmental risk factors, including childhood stress, obstetric complications, and maternal infection during pregnancy (Lipner et al. 2019; Nimgaonkar et al. 2017), trigger inflammatory signaling cascades, which affect the developing brain at prenatal and early postnatal stages (Hilker et al. 2018; Muller et al. 2015). Indeed, genetic, transcriptional, and serological evidence indicate increased inflammatory conditions in patients with SCZ, given the enhanced levels of multiple cytokines, such as IL‐1β, IL‐6, IL‐8, and TNF‐α (Frydecka et al. 2018; Muller et al. 2015; Rodrigues‐Amorim et al. 2018). Moreover, genetic variants associated with SCZ show dynamic regulatory activity in neural cells in response to cytokine exposure (Retallick‐Townsley et al. 2024).

Astrocytes provide metabolic support to neurons, facilitating the formation and modulation of synaptic connectivity, ionic buffering, while also uptaking and releasing neurotransmitters (Bernaus et al. 2020). Along with microglia, astrocytes perceive pro‐inflammatory stimulation through a wide range of immune receptors, ultimately leading to their activation (Linnerbauer et al. 2020). Astrocyte activation follows enhanced expression of reactivity markers, such as GFAP and Vimentin, as well as augmented secretion of a wide range of cytokines and chemokines (e.g., TNF‐α, IL‐6, GM‐CSF, TGF‐β, and CCL2) (Bernaus et al. 2020; Linnerbauer et al. 2020; Liu et al. 2011). Secretion of these immune factors can, in turn, act upon microglial cells, either boosting or suppressing their activation, greatly impacting their effector capabilities, such as phagocytosis and cell migration (Bernaus et al. 2020; Linnerbauer et al. 2020; Liu et al. 2011). Astrocytes also modulate microglial function under physiological conditions; for instance, astrocyte‐produced IL‐33 stimulates synaptic pruning by microglia, which is critical for proper neuronal connectivity (Vainchtein et al. 2018).

Human genetic, post‐mortem, and brain imaging analyses, together with mouse behavioral studies, link astrocytic dysfunction to SCZ (de Oliveira Figueiredo et al. 2022; Dietz et al. 2020). Astrocyte and interferon‐response gene expression modules were upregulated in patients with SCZ, while the microglia module was downregulated (Gandal et al. 2018). Likewise, PET scans reveal increased astrocyte reactivity in individuals with SCZ in the anterior cingulate cortex and left hippocampus (Kim et al. 2024). Finally, human‐induced pluripotent stem cells (hiPSCs)‐derived astrocytes from individuals with SCZ have shown impaired astrocytic differentiation and maturation (Windrem et al. 2017), abnormal calcium signaling (Szabo et al. 2021), decreased glutamate uptake (Szabo et al. 2021), and enhanced pro‐inflammatory profile (Koskuvi et al. 2022; Trindade et al. 2023).

Nonetheless, the crosstalk between astrocytes and microglia in SCZ, especially under pro‐inflammatory conditions, remains unclear. Here, we evaluated how reactive SCZ hiPSC‐derived astrocytes affected the microglial phenotype, both at the transcriptional and functional levels. We further assessed the contribution of the CX3CL1/CX3CR1 axis in this phenomenon and aimed at identifying putative transcriptional factors driving the observed alterations.

2. Methods

2.1. Ethics Statement

All experiments carried out throughout this work were approved by UFMG's Ethics Committee (COEP‐UFMG #90424518.3.1001.5149) and were performed according to the Helsinki Declaration and the Brazilian National Health Council Resolution 466/12.

2.2. Resources Table

A complete list of all materials and reagents used in this study can be found in Table S1.

2.3. Candidate Gene Screening

The list of transcripts or proteins identified in two transcriptomic (Akkouh et al. 2022; Szabo et al. 2021) and one proteomic (Trindade et al. 2023) studies comparing gene expression between hiPSC‐derived astrocytes sourced from individuals with schizophrenia and controls under basal conditions was obtained from each study's supplementary data. These lists were inputted into Metascape (Zhou et al. 2019) and up to 40 GO Biological Processes (GO BP) terms were retrieved per study (p < 0.01). Overlapping GO BP terms were identified, and genes associated with the chosen GO BP term were obtained from each study to shortlist potential candidates. A synaptic pruning network figure was created by merging GO:0098883 (“synaptic pruning”) and “Complement Signaling” (downloaded from SIGNOR 3.0) and editing the resulting network in Cytoscape (v. 3.10.1) (Shannon et al. 2003) for schematic visualization.

2.4. hiPSC‐Derived Neural Stem Cells

Neural stem cells (NSCs) were differentiated from hiPSCs from five neurotypical and five individuals with schizophrenia (Table S2) as previously described (de Lima et al. 2023; Trindade et al. 2020; Yan et al. 2013). Briefly, 80%–90% confluent hiPSCs were dissociated into single cells using Accutase solution and plated at 3 × 10⁴ cells/cm² in Geltrex‐coated tissue culture (TC)‐treated 6‐well plates in StemFlex Medium, supplemented with 1% antimycotic‐antibiotics solution (anti‐anti) and 10 μM Rho‐kinase inhibitor Y‐27632 (ROCKi). On the first day of differentiation (day 0), media was switched to PSC Neural Induction Medium (Neurobasal, 2% Neural Induction Supplement and 1% penicillin–streptomycin (P/S) solution) and changed every other day. On the 7th day, Neural Stem Cells (NSCs) were dissociated with Accutase and plated at 1 × 10⁵ cells/cm² in Geltrex‐coated TC 60‐mm petri dishes in NSC Expansion Medium (50% Advanced DMEM/F‐12, 50% Neurobasal, 2% Neural Induction Supplement, and 1% P/S) with 10 μM ROCKi. Media changes were performed three times a week. Cells were cultured in standard conditions (5% CO₂ atmosphere and 37°C in a humidified incubator). hiPSCs identity was confirmed by positive immunostaining for OCT4, SSEA‐1, TRA‐1‐60, and TRA‐1‐81 (Figure S1A–D). NSCs identity was confirmed by positive immunostaining for Nestin, PAX6, SOX1, and SOX9 (Figure S1E–H).

2.5. hiPSC‐Derived Astrocytes Differentiation

Astrocytes differentiation was performed as described elsewhere (de Lima et al. 2023; Trindade et al. 2020). Once hiPSC‐derived neural stem cells (NSCs) reached 90% confluency, cells were dissociated with Accutase and seeded at 5 × 10⁴ cells/cm² in Geltrex‐coated TC 25 cm² flasks in Astrocytes Differentiation Medium (DMEM/F‐12, 1× N‐2 supplement, 1% Heat‐inactivated Fetal Bovine Serum (FBS) and 1% antimycotic‐antibiotics solution (anti‐anti)). Media was replenished every other day for 21 days. Whenever differentiating astrocytes reached 90% confluency, cells were expanded to larger Geltrex‐coated TC flasks at a 1:3 ratio as described. By the end of this differentiation period (day 21), media was switched to Astrocytes Maturation Medium (DMEM/F‐12, 10% FBS, and 1% anti‐anti) to mature astrocytes for 5 weeks. Whenever astrocytes reached 90% confluency, they were split using Trypsin/EDTA 0.125% solution (diluted in PBS 1×^−/−) and expanded to 175 cm² TC flasks without Geltrex coating at a 1:2 ratio to negatively select undifferentiated non‐adhering cells. Cells were cultured in standard conditions. Astrocytes' identity was confirmed by positive immunostaining to EAAT1, Vimentin, GFAP, and Connexin 43 (Figure S2A,B).

2.6. Astrocytes Stimulation

After at least 5 weeks of maturation, astrocytes were detached using Trypsin/EDTA 0.125% solution, pelleted down by centrifugation (800g for 5 min), resuspended in Astrocytes Maturation Medium, and plated at 1.25 × 10⁴ cell/cm² density in either TC 6‐well plates or 25 cm² flasks. After 5 days in culture, cells were washed three times with PBS 1×^−/− and serum‐starved for 24 h in DMEM/F‐12 supplemented with 1% anti‐anti. Next, astrocytes were stimulated for 24 h with 10 ng/mL recombinant human TNF‐α or vehicle (0.1% BSA solution). Once stimulation finished, astrocytes conditioned media (A_CM) was collected, clarified by centrifugation (3000g, 4°C for 10 min), pooled together from all five donors (for each condition), flash frozen, and stored at −80°C until further use. In addition, astrocytes monolayers were harvested either in TRIzol reagent according to manufacturer instructions or in RIPA lysis buffer containing protease and phosphatase inhibitors for 1 h on ice and stored at −80°C.

2.7. Total RNA Extraction and RT‐qPCR

Total RNA was isolated using the TRIzol reagent as per manufacturer instructions and resuspended in 12 μL DEPC‐treated nuclease‐free water. RNA concentration and quality (260/230 and 260/280 ratios) were analyzed by spectrophotometry in the Multiskan GO (Thermo Scientific). Exactly 1 μg total RNA was reverse transcribed in a reaction mixture consisting of 15 ng/μL Random Primers, 50 mM Tris–HCl, 75 mM KCl, 3 mM MgCl₂, 625 μM dNTPs, 10 μM DTT, and M‐MLV reverse transcriptase. Reverse transcription conditions were as follows: (Random primers annealing) 70°C for 10 min and 4°C for 10 min; (cDNA extension) 42°C for 60 min and 70°C for 15 min. The resulting cDNA was diluted 1:10 in nuclease‐free water and subjected to qPCR using Power SYBR Green PCR Master Mix and 200 nM of forward and reverse primers mix (sequences available in Table S3) in the QuantStudio 7 System (Applied Biosystems). Thermal cycling conditions were: 50°C for 2 min, 95°C for 10 min; 40 cycles of 95°C for 15 s; and 60°C for 1 min; followed by melting curve analysis. All primer pairs had 90%–110% efficiency. Relative gene expression of target genes was normalized by the average of housekeeping genes (RPLP0 and IPO8) and calculated by the 2^−ΔCt method.

2.8. Donor Selection and Peripheral‐Blood Mononuclear Cells (PBMCs) Isolation

A male control individual was chosen (Table S2), after evaluation by a trained psychiatrist following the M.I.N.I. 7.0.0 (Sheehan et al. 1998) and fitting the following inclusion and exclusion criteria: less than 50 years of age; at least high school education; no previous history of neurological or neuropsychiatric conditions; absence of 1st degree relatives diagnosed with psychiatric disorders; absence of 2nd degree relatives diagnosed with schizophrenia, bipolar disorder, or autism spectrum disorder; non‐cannabis user; non‐smoker; capability to freely and voluntarily give informed consent to participate in this study. The study design was fully and carefully explained to the selected participant, who had the chance to ask questions afterwards to clarify any concerns regarding his study participation. Next, both the participant and the corresponding author signed two copies of the Informed Consent Form before proceeding with his blood harvest. Whole blood was drawn from this individual, diluted 1:1 with PBS 1×^−/−, carefully layered onto Histopaque 1077 density gradient, and centrifuged at 900g for 25 min (with acceleration and breaking settings turned off). After blood fractionation, the PBMCs layer was collected using a sterile plastic Pasteur pipette, resuspended in PBS 1×^−/− and centrifuged at 400g for 8 min. The supernatant was removed, cells resuspended once again in PBS 1×^−/− and pelleted by centrifugation at 400g for 8 min. PBMCs were resuspended in FBS, counted in a hemocytometer, and cryopreserved in FBS + 10% DMSO for later use.

2.9. Induced Microglial‐Like Cells (iMGs) Differentiation

The iMGs differentiation was carried out as described elsewhere (Ohgidani et al. 2014; Sellgren et al. 2019; Sellgren et al. 2017). PBMCs were plated in either 13‐mm acid‐etched glass coverslips or TC‐treated plasticware, both previously coated overnight with Geltrex, and cultured in RPMI 1640 medium, supplemented with 10% FBS and 1% penicillin–streptomycin (P/S) solution. Cell density was adjusted on an assay‐dependent basis. On the following day (day 0), media was completely switched to iMG differentiation media (MDM), comprised of RPMI 1640, 1% Glutamax, 100 ng/mL IL‐34, 10 ng/mL GM‐CSF, and 1% P/S. On the 6th day of differentiation, half of the media was changed and replenished with fresh MDM. On day 9, media was thoroughly removed, cells were washed twice with pre‐warmed RPMI 1640 medium to remove cell debris, and fresh MDM was added. iMGs were used for subsequent experiments between days 10 and 12 of differentiation. iMGs were kept in standard culture conditions. Microglial‐like cells identity was confirmed by their ramified morphology and positive immunostaining to CX3CR1, IBA1, and PU.1. Bright‐field images were captured using the FLoid Cell Imaging Station (Thermo Fisher).

2.10. hiPSC‐Derived Neuronal Differentiation

CF1 and CF2 NSCs were differentiated to neurons, according to Espuny‐Camacho et al. (2013) with modifications. 2.5 × 10⁴ or 5 × 10⁴ NSCs/cm² were seeded on 13‐mm acid‐etched sterile glass coverslips or TC‐treated 100‐mm petri dishes, respectively, coated with 0.1% poly‐ethylenimine (glass surface) or 0.001% poly‐ornithine (plastic surface) solution and 2% Geltrex, in Default Defined Medium (DDM), consisting of DMEM/F‐12, 2% B‐27 supplement, 1% N‐2 supplement, 1% Glutamax, 1% MEM non‐essential amino acids solution, 1 mM Sodium Pyruvate, 0.1 mM 2‐mercaptoethanol, 0.5 mg/mL BSA, 200 ng/mL l‐ascorbic acid, 1 μg/mL laminin, and 1% P/S, supplemented with 10 μM ROCKi. After 2 days in culture, media was thoroughly removed and replenished with fresh DDM without ROCKi. Three days later (day 0), half of the media was replaced with Neurobasal/B‐27 medium (Neurobasal, 2% B‐27 supplement, 1% Glutamax, 200 ng/mL l‐ascorbic acid, 1 μg/mL laminin, and 1% P/S). Half of the media was changed every 3–4 days by an isovolumetric mixture of DDM and Neurobasal/B‐27 media until day 60 of differentiation. At the end of differentiation, neuronal cultures were either stained for β‐tubulin III, MAP2, S100β, MBP, Synaptotagmin 2 (SYT 2), and PSD‐95 or used for synaptoneurosomes isolation. Cells were kept in standard culture conditions.

2.11. Synaptoneurosomes Isolation

Synaptoneurosomes isolation was performed as described by Sellgren et al. (2019) with minor modifications. After 60 days of differentiation, media was removed from neuronal cultures and replaced by non‐stimulated A_CM from their isogenic counterparts 24 h before cell harvest. Next, media was thoroughly removed, and cells were washed twice with ice‐cold PBS 1×. Exactly 1 mL per 100‐mm TC dish of ice‐cold Synaptoneurosomes Isolation Buffer (SIB; 10 mM HEPES, 1 mM EDTA, 2 mM EGTA, 0.5 mM DTT, and protease inhibitors; pH 7.0) (Villasana et al. 2006) was added and neurons were gently scrapped. This cell suspension was transferred to a sterile 1.5 mL conical tube and centrifuged at 1200g for 10 min at 4°C. The supernatant (homogenate) was transferred to a new sterile 1.5 mL conical tube and centrifuged at 15,000g for 20 min at 4°C. Finally, the supernatant (cytosolic fraction) was removed, pelleted synaptoneurosomes were resuspended in SIB with 5% DMSO and frozen at −80°C until further use. Aliquots of each fraction were collected at each step for further characterization by Western Blot and stored at −80°C. Freshly isolated synaptoneurosomes were also plated overnight on 13‐mm acid‐etched glass coverslips coated with 50 μg/mL poly‐d‐lysine solution and left to attach overnight under standard culture conditions for further immunostaining to Syntaxin 1 and Homer. Shortly before the phagocytosis assay, synaptoneurosomes were homogenized for up to 50 times in ice‐cold PBS 1× using a sterile micro‐capillary pipette tip, their concentration adjusted to 0.25 μg/μL and labeled with 2 μM Vybrant‐CM‐DiI, as per manufacturer instructions.

2.12. Synaptoneurosomes Phagocytosis Assay

Exactly 2.8 × 10⁵ PBMCs/cm² per well were seeded in 24‐well plates and differentiated as already described. On day 12, iMGs were incubated for 1 h with 2 ng/mL recombinant CX3CL1 or vehicle (0.1% BSA solution). Alternatively, iMGs were pre‐treated with 5 μM AZD8797 or vehicle (DMSO) for 1 h; then, half of the media was removed and replenished by an equal volume of pooled A_CM and cells incubated for another hour (AZD8797 was also added again to ensure its concentration was kept at 5 μM throughout the whole experiment). Next, 1.875 μg of CM‐DiI‐labeled synaptoneurosomes were added to each well, and the plates were gently shaken for thorough homogenization and incubated for 24 h in standard culture conditions in the Cytation 5 Cell Imaging Multi‐mode Reader (BioTek; CAPI—ICB, UFMG, Brazil). Four equally spaced image sets (bright field and RFP channels) per well were captured every 2 h. Experiments were conducted twice in duplicates to ensure reproducibility. After 24 h, images were retrieved and analyzed using the combination of the following open‐source softwares: (1) FIJI/ImageJ (v. 1.54f) (Schindelin et al. 2012) for brightness/contrast adjustments and denoising; (2) fastER (v. 1.3.5) (Hilsenbeck et al. 2017) for cell segmentation (bright field channel); (3) CellProfiler (v. 4.2.1) (Stirling et al. 2021) for single‐cell phagocytic index quantification. Phagocytic index was calculated as the ratio of red object area (μm²) inside a given cell to its total cell area (μm²). Orthogonal projection (z‐step: 2 μm; number of steps: 20) was generated using FIJI/ImageJ from vehicle‐treated (0.1% BSA) fixed samples (24 h time‐point) on glass coverslips immunostained for Alexa 633 Phalloidin and Hoechst 33342 as described.

2.13. Western Blot

Astrocytes' lysates (50 μg) and synaptoneurosome fractions (25 μg) protein concentration was quantified by the Bradford method. Samples were diluted in Laemmli Sample Buffer, heated at 95°C for 5 min, and later subjected to electrophoretic separation in 10% SDS‐PAGE. Proteins were transferred to 0.45 μm nitrocellulose membranes, blocked with 5% BSA solution diluted in TBS with 0.1% Tween‐20 (TBS‐T) for 1 h at room temperature, followed by overnight incubation at 4°C with primary antibodies diluted in 3% BSA solution in TBS‐T: anti‐CX3CL1 (1:500), anti‐β‐actin (1:5000), anti‐Syntaxin 1 (1:200), anti‐Homer (1:500), or anti‐Vinculin (1:10,000). The following day, primary antibodies were removed, membranes washed three times with TBS‐T for 5 min, and incubated for 1 h at room temperature with the following secondary antibodies diluted in 3% free‐fat milk solution in TBS‐T: HRP‐conjugated anti‐mouse IgG (1:2500), HRP‐conjugated anti‐rabbit IgG (1:2500), or HRP‐conjugated anti‐goat IgG (1:2500). Next, secondary antibodies were removed, membranes washed with TBS‐T as described, incubated for 5 min with the detection reagent for chemiluminescence reaction, and images acquired using the ImageQuant LAS 4000 platform. When appropriate, densitometric analysis was carried out using FIJI/ImageJ and expressed as the ratio of CX3CL1/β‐actin protein levels relative to non‐stimulated HCT astrocytes.

2.14. Immunofluorescence Staining

Cells and synaptoneurosomes were fixed for 15 min in 4% PFA solution (in PBS 1×). Next, samples were permeabilized for 10 min in 0.3% Triton X‐100 diluted in PBS 1× (PBS‐T) and blocked in 2% BSA solution (in PBS‐T) for 1 h. Afterwards, cells were incubated overnight at 4°C with primary antibodies diluted in blocking solution: anti‐OCT4 (1:200), anti‐SSEA‐4 (1:100), anti‐TRA‐1‐60 (1:100), anti‐TRA‐1‐81 (1:100), anti‐Nestin (1:400), anti‐PAX6 (1:100), anti‐SOX1 (1:200), anti‐SOX9 (1:100), anti‐GFAP (1:200), anti‐Vimentin (1:500), anti‐EAAT1 (1:100), anti‐Connexin 43 (1:100), anti‐MAP2 (1:500), anti‐β‐tubulin III (1:500), anti‐S100β (1:200), anti‐MBP (1:500), anti‐Synaptotagmin 2 (1:100), anti‐PSD95 (1:500), anti‐Syntaxin 1 (1:100), anti‐Homer (1:200), anti‐CX3CR1 (1:500), anti‐Iba1 (1:750), anti‐Pu.1 (1:200), and/or anti‐KI67 (1:100). The following day, the primary antibody solution was removed, samples washed three times for 5 min with PBS 1×, and incubated for 1 h, protected from light and at room temperature with the following secondary antibodies and dyes diluted in blocking solution: Alexa Fluor 488 anti‐Rat IgG (1:500), Alexa Fluor 488 anti‐Mouse IgG (1:400), Alexa Fluor 546 anti‐Rabbit IgG (1:300), Alexa Fluor 546 anti‐Rat IgG (1:300), Alexa Fluor 555 anti‐Goat IgG (1:500), Alexa Fluor 594 anti‐Mouse IgG (1:400), Alexa Fluor 633 anti‐Mouse IgG (1:500), Alexa Fluor 633 anti‐Rabbit IgG (1:500), Alexa Fluor 633 Phalloidin (1:1000), and/or Hoechst 33342 (1:500). After secondary antibody incubation, samples were washed as described and mounted on clean glass slides using the Hydromount mounting medium, overnight and protected from light. Coverslips were sealed with nail polish the following day. Images were acquired via confocal laser microscopy using either the Zeiss Apotome 3 (CAPI—ICB, UFMG, Brazil) or the Nikon A1 microscope (Centre for Gastrointestinal Biology, UFMG, Brazil) or via wide‐field microscopy using the Cytation 1 Cell Imaging Multimode Reader (Biotek; D'or Institute, Brazil). Images were analyzed on FIJI/ImageJ. CellProfiler software was used to automatically quantify KI67 puncta in iMGs.

2.15. Enzyme‐Linked Immunosorbent Assay (ELISA)

Soluble CX3CL1 was quantified in A_CM using the Human CX3CL1/Fractalkine DuoSet ELISA kit as per manufacturer instructions. Astrocyte monolayers of all cell lines were detached using Trypsin/EDTA 0.125% solution and counted in a hemocytometer immediately after A_CM harvest to normalize data by cell count.

2.16. cDNA Library Preparation and Sequencing

Exactly 4.0 × 10⁵ PBMCs/cm² per well were seeded in Geltrex‐coated TC 12‐well plates and differentiated as described. After 12 days of differentiation, half of the media was discarded and replaced by an equal volume of pooled A_CM. Exactly 24 h later, media was completely removed, and cells harvested in TRIzol reagent. Total RNA extraction was performed from samples generated from three independent iMGs' differentiation experiments as recommended by the manufacturer and RNA quality analyzed in the 2100 Bioanalyzer platform (Agilent). Two hundred nanograms of total RNA (RIN > 7) was subjected to rRNA depletion and cDNA library preparation using the NEBNext Ultra II Directional RNA Library Prep Kit, according to manufacturer instructions. cDNA library quality assessment and paired‐end sequencing (2 × 100 pb) using the NextSeq2000 platform (Illumina) was performed by the LaCTAD (UNICAMP, Brazil). Samples had an average sequencing coverage of 30 million reads/library.

2.17. Differential Gene Expression Analysis

RNA‐seq reads quality was assessed by FastQC (Andrews 2010). Adaptor sequences were trimmed out and reads with less than 50 bp or Phred < 20 were removed using Trimmomatic (v. 0.40) (Bolger et al. 2014). The STAR aligner (v. 2.7.11b) (Dobin et al. 2013) was used for mapping reads to the human reference genome (GRCh38), and featureCounts (v. 2.0.6) (Liao et al. 2014) was employed to count mapped reads. Features with fewer than 10 counts were excluded from further analysis. Batch effects were removed using the RUVseq package (residuals method; v. 1.34.0) (Risso et al. 2014). edgeR (v. 3.42.4) (Robinson et al. 2010) was employed for differential gene expression analysis (adjusted p value < 0.05; |log₂ FC|≥ 0.5). Principal component analysis was performed using the PCAtools package (v. 2.12.0) (Kevin Blighe 2024).

2.18. Gene Set Enrichment Analysis (GSEA)

For GSEA, lists of ranked genes were generated using the Signal‐to‐Noise ratio metric from the batch‐corrected gene count matrix. Next, these ranked lists were subjected to GSEA using the clusterProfiler package (v. 4.8.3) (Wu et al. 2021) to identify enriched GO Biological Processes and KEGG Pathways. The GSEA function was also used to identify microglial phenotypic signatures against the Prater et al. (2023) single‐nuclei RNA‐seq dataset. This dataset consists of human dorsolateral prefrontal cortex microglia isolated from post‐mortem brain tissue of individuals with Alzheimer's Disease; signature analysis was restricted to individuals bearing the APOE3/APOE3 genotype to prevent potential biases associated with the APOE4 allele. The GSEA function was also used to evaluate cellular senescence induction or inhibition using the CellAge database (Avelar et al. 2020). Adjusted p‐value threshold was set at 0.1.

2.19. Weighted Gene Co‐Expression Network Analysis (WGCNA)

For WGCNA, the batch‐corrected gene count matrix was used. Genes with fewer than 10 counts were excluded, and the top 10,000 genes with the highest variance were selected for network construction. A soft thresholding power of 11 was applied using the signed‐hybrid network type. The minimum module size was set to 30, and modules were merged if their Spearman correlation coefficient exceeded 0.8. Correlation p values were adjusted using the Benjamini and Hochberg method (adjusted p value < 0.05). The bioNERO package (v. 1.8.7) (Almeida‐Silva and Venancio 2022) was used alongside WGCNA (v. 1.72‐5) (Langfelder and Horvath 2008) for analysis and visualization. GO term enrichment analysis was performed using the biomaRt package (v. 2.56.1) (Durinck et al. 2005).

2.20. Master Regulator Analysis (MRA)

MRA analysis was carried out based on the pipeline published by Leng et al. (2022) with adaptations, using the RTN package (v. 2.24.0) (Fletcher et al. 2013). First, for co‐expression network reconstruction, 871 RNA‐seq datasets were downloaded from ARCHS4 (Lachmann et al. 2018), whose sample descriptors were “Microglia” and “Human.” Batch effects across datasets were removed using the Combat‐seq function of the sva package (v. 3.48.0) (Leek et al. 2024), gene counts log transformed using the vst function of the DESeq2 package (v. 1.40.2) (Love et al. 2014) and gene‐level count matrix subset to contain only the 16,954 genes identified in the RNA‐seq dataset of the present study. The list of human transcription factors published by Lambert et al. (2018) was input as regulatory elements for network reconstruction. After transcriptional regulatory network reconstruction finished, master regulators and their activity status for each experimental condition were calculated as described in the RTN package vignette (adjusted p value < 0.1). Associations between activated or repressed regulons and WGCNA modules were done using the tni.overlap.genesets() function.

2.21. Transcriptional Age Calculation

Transcriptional age was calculated from the batch‐corrected gene count matrix by the RNAAgeCalc package (v. 1.16) (Ren and Kuan 2020), using the “GTExAge” signature and “brain” as reference tissue.

2.22. iMGs Migration Assay

Exactly 3.8 × 10⁵ PBMCs/cm² were plated in Geltrex‐coated 60‐mm TC petri dishes and differentiated as already described. After 9 days of differentiation, iMGs were pre‐treated with 10 μM Rho‐kinase inhibitor Y‐27632 for 1 h, followed by incubation with hypertonic citrate saline solution (135 mM KCl and 15 mM sodium citrate diluted in PBS 1×) for 5 min at 37°C to gently loosen up attached cells. Next, iMGs were gently scraped using a cell scraper, collected into a 15 mL conical tube (pre‐coated with sterile 5% BSA solution, overnight at 4°C) and centrifuged at 400g for 8 min. Supernatant was discarded, iMGs resuspended in MDM, and 6.0 × 10⁴ cells/transwell plated into the upper compartment of each Geltrex‐coated transwell. Exactly 24 h after plating, iMGs were pre‐treated with AZD8797 (5 μM) or vehicle (DMSO) for 1 h, pooled A_CM was added to the lower compartment, and cells were incubated overnight (~16 h) in standard culture conditions. The next day, iMGs migration was stopped by fixing these cells for 15 min with 4% PFA. Transwells were washed three times with PBS 1×, non‐migrated cells (in the upper compartment) gently scraped off using a cotton swab, and migrated cells (in the lower compartment) stained with Hoechst 33342 (1:500; diluted in PBS 1×) for 10 min and protected from light. Samples were washed as described and subsequently imaged in the EVOS FL Color microscope (Thermo Fisher). Four to five image fields/transwell were randomly acquired and migrated cells quantified by a trained researcher, blind to the experimental conditions, using the CellCounter plugin on FIJI/ImageJ. Migrated cell count was expressed as a percentage relative to vehicle‐treated iMGs exposed to A_CM HCT (N.S.). This experiment was executed four times independently.

2.23. Cell‐Based ELISA

Exactly 5.26 × 10⁵ PBMCs/cm² per well were plated in Geltrex‐coated TC 96‐well plates and differentiated to iMGs as described. Once differentiation was completed, half of the media was removed, replenished by an equal volume of pooled A_CM and incubated for 24 h in standard culture conditions. Afterwards, media was carefully removed, and cells fixed with 4% PFA solution (in PBS 1×) for 15 min. Then, cell‐based ELISA was carried out based on Yang et al. (2021) with modifications. Samples were washed three times with PBS 1× for 5 min under gentle agitation to remove excessive fixative solution. Next, for total CX3CR1 measurements only, PBS 1× was removed and replaced by permeabilizing solution (0.1% Triton X‐100 in PBS 1×) for 10 min; for plasma membrane CX3CR1 measurements, samples were incubated with PBS 1× (without Triton X‐100) for an equal amount of time. Once this incubation was over, intracellular peroxidases were quenched by incubation with 1% H₂O₂ (in PBS 1×) for 20 min. Quenching solution was removed and samples blocked with 1% BSA (in PBS 1×) for 1 h at room temperature. Blocking solution was discarded and cells incubated with anti‐CX3CR1 (1:500; diluted in blocking solution) overnight at 4°C, and the plate sealed with parafilm to prevent evaporation. In the following day, primary antibody solution was removed, samples washed thrice as described and incubated with HRP‐conjugated anti‐rabbit IgG (1:2000) for 2 h at room temperature and protected from light exposure. Secondary antibody solution was removed, washed as described, and samples incubated with the substrate solution (0.4 mg/mL OPD, 0.8 μL/mL in citrate buffer (111.2 mM NaH₂PO₄, 27 mM citric acid, pH 5.0)) for 30 min at room temperature and protected from light. Chromogenic reaction was stopped by adding 5.5% H₂SO₄ solution to each well, and absorbance measured after 5 min at 490 nm in the Multiskan GO plate reader (Thermo Fisher). For whole cell staining, the chromogenic solution was removed, cells washed five times with PBS 1× as described and incubated with 1% crystal violet solution for 30 min under gentle agitation. Crystal violet solution was removed, samples washed extensively with ddH₂O to remove excessive amounts of dye, washed thrice with PBS 1× as described, and incubated with 1% SDS solution for 1 h under agitation to solubilize intracellular crystal violet. Absorbance was read at 595 nm in the plate reader, and this measurement was used to normalize CX3CR1 protein levels, according to the formula below:

$$\text{Relative Protein Levels}\left(\mathrm{RPL}\right)=\frac{{\mathrm{OD}}_{490}\text{Sample}-{\mathrm{OD}}_{490}\text{Blank}}{{\mathrm{OD}}_{595}\text{Sample}-{\mathrm{OD}}_{595}\text{Blank}}$$

RPL was expressed relative to iMGs exposed to A_CM HCT (N.S.). Experiments were conducted twice with five replicates each.

2.24. Mycoplasma Testing

All cell lines were routinely tested for Mycoplasma contamination as described elsewhere (Molla Kazemiha et al. 2009) (universal Mycoplasma primers can be found in Table S3). Only Mycoplasma‐free cells were used in this work.

2.25. Statistical Analysis

Data were tested for normal distribution using D'Agostino's K‐squared test. Except where otherwise specified, statistical analyses were carried out either by one‐way or two‐way ANOVA, followed by Holm–Sidak's multiple comparison test using the GraphPad Prism (v. 8.0.1) software. Synaptoneurosome phagocytosis by iMGs was analyzed by multilevel mixed‐effects linear regression with maximum likelihood estimation, followed by Sidak's multiple comparison test in R (v. 4.3.1). Bioinformatics methods' statistical analyses were conducted using each package's built‐in statistics in R. Unless otherwise stated, the significance level was set at 0.05 (α = 0.05).

3. Results

3.1. SCZ‐Sourced Astrocytes Display Increased Pro‐Inflammatory Response Following Stimulation With TNF‐α

Given the potential role for astrocytes in SCZ, we screened two transcriptomic (Akkouh et al. 2022; Szabo et al. 2021) and one proteomic (Trindade et al. 2023) studies that compared gene/protein expression changes between hiPSC‐derived astrocytes from individuals with SCZ and neurotypical controls (HCT), to identify common biological processes associated with all three datasets. Interestingly, only the “cellular homeostasis” (GO:0019725) term overlapped among all three (Figure S3A,B; Table S4A–C). Although no single candidate gene from the “cellular homeostasis” term overlapped among all three studies, three genes were present in at least two: CX3CL1 (the sole member of the CX3C chemokine family, also known as Fractalkine), NGFR (Nerve Growth Factor Receptor) and SLC24A2 (a calcium, potassium: sodium antiporter) (Figure S3C,D). CX3CL1 is associated with a wide range of biological processes that might be important in the context of SCZ, including microglial migration, immune response, and synaptic pruning, the latter having long been hypothesized as having a causative role in this disorder (Feinberg 1982; Howes and Onwordi 2023). Along with CX3CL1 (Paolicelli et al. 2011), classical complement components have also been implicated in synaptic elimination (Stevens et al. 2007; Yilmaz et al. 2021) (Figure S3E), while astrocyte‐produced IL‐33 has been shown to promote microglial‐mediated synaptic uptake (Vainchtein et al. 2018). In addition, reactive astrocytes display increased expression of a wide range of cytokines with pleiotropic activities (Bernaus et al. 2020; Linnerbauer et al. 2020; Liu et al. 2011; Moraga et al. 2014; Trindade et al. 2020), making them important targets for further investigation.

Taking in consideration the predicted involvement of inflammatory signaling in schizophrenia (Muller et al. 2015) and TNF‐α ability in inducing astrocyte reactivity (Trindade et al. 2020), we decided to assay the expression of CX3CL1, IL‐33, classical complement components, and pro‐ and anti‐inflammatory cytokines in hiPSC‐derived astrocytes from five neurotypical (HCT) and five SCZ cases upon stimulation with TNF‐α (Figure 1A). Independent of cell donor diagnostic, CX3CL1 mRNA was upregulated in astrocytes subjected to TNF‐α stimulation, while C3 mRNA trended in the same direction and no change was observed in IL‐33 and C4 expression (Figure 1B–E). Conversely, C1q mRNA expression was not detected in hiPSC‐derived astrocytes. Interestingly, CX3CL1 transcripts were significantly higher in stimulated SCZ astrocytes compared with stimulated controls (Figure 1D). Moreover, TNF‐α stimulation promoted the mRNA upregulation of the pro‐inflammatory cytokines CCL5, TNF‐α, IL‐1α, IL‐1β, IL‐6, and IL‐8, while the anti‐inflammatory cytokines IL‐4 and IL‐10 were not detected (Figure 1F–K). Noteworthy, stimulated SCZ astrocytes displayed significantly augmented CCL5, TNF‐α, IL‐1β and IL‐6 (and a trend towards increased IL‐1α) mRNA levels compared with stimulated HCT astrocytes (Figure 1F–J). These data confirm, in part, those previously observed by Trindade et al. (2023).

FIGURE 1.

SCZ astrocytes showed a stronger pro‐inflammatory response upon TNF‐α stimulation. (A) Experimental design of hiPSC‐derived astrocyte differentiation, stimulation with TNF‐α (10 ng/mL) and downstream assays (RT‐qPCR and conditioned media harvest for further applications). Schematic picture was drawn using Biorender. (B–K) mRNA expression of C3 (B), C4 (C), CX3CL1 (D), IL‐33 (E), CCL5 (F), TNF‐α (G), IL‐1α (H), IL‐1β (I), IL‐6 (J), and IL‐8 (K) in HCT and SCZ astrocytes stimulated with TNF‐α for 24 h. n = 5 (number of donor cell lines per group). (L–M) Representative Western blot of non‐secreted CX3CL1 in HCT and SCZ astrocytes stimulated with 10 ng/mL TNF‐α for 24 h (L) and its respective densitometric quantification (M). Representative blot out of three independent experiments. Data are expressed relative to non‐stimulated HCT astrocytes. CX3CL1 levels were normalized relative to β‐Actin and displayed in arbitrary units. n = 5 (number of donor cell lines per group). (N) Secreted CX3CL1 levels in HCT and SCZ astrocytes stimulated with TNF‐α for 24 h. n = 5 (number of donor cell lines per group). Data were analyzed by two‐way ANOVA, followed by Holm–Sidak's multiple comparison test. Bars represent Mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001.

Among these immune factors, CX3CL1 has been shown to be involved in a multitude of biological processes relevant in the context of SCZ. Thus, we sought to validate CX3CL1 mRNA expression findings (Figure 1D) at the protein level. TNF‐α stimulation increased the production and secretion of CX3CL1 in both HCT and SCZ astrocytes, while no change in expression was observed at the non‐secreted protein level (Figure 1L–N). Remarkably, stimulated SCZ astrocytes showed a 2.3‐fold augmentation in secreted CX3CL1 levels compared with stimulated control astrocytes (Figure 1N). Taken together, these data indicate that SCZ astrocytes responded more strongly to TNF‐α stimulation by producing greater levels of several pro‐inflammatory cytokines and the chemokine CX3CL1, relative to HCT astrocytes.

3.2. TNF‐α‐Stimulated SCZ Astrocytes Promote a Dysfunctional Transcriptional Response in Microglial‐Like Cells

Since many pro‐inflammatory cytokines and CX3CL1 display pleiotropic functions, being capable of inducing several distinct responses in immune cells (Mecca et al. 2018; Moraga et al. 2014), we queried potential biological processes that might be altered in microglial‐like cells by reactive SCZ astrocytes. We generated iMGs from a neurotypical individual (Figure S4), incubated these cells with astrocyte conditioned media (A_CM) harvested from HCT and SCZ astrocytes subjected to either vehicle or TNF‐α stimulation (Figure 2A), and applied bulk RNA‐seq to evaluate the transcriptional profile of these iMGs.

FIGURE 2.

iMGs displayed a dysfunctional transcriptional profile after incubation with A_CM from TNF‐α‐stimulated SCZ astrocytes. (A) Experimental design of healthy control iMGs' differentiation from PBMCs, incubation with A_CM at the indicated timepoints and downstream applications (synaptoneurosomes engulfment, cell migration and transcriptomic assays). Schematic picture was drawn using Biorender. (B) Principal component (PC) analysis of iMG samples incubated with all four distinct A_CM sources for 24 h. (left) PC1 × PC2; (middle) PC1 × PC3; (right) PC2 × PC3. (C) Venn diagrams showing shared upregulated (left) and downregulated (right) DEGs between iMGs exposed to A_CM HCT (+TNF‐α), A_CM SCZ (N.S.), or A_CM SCZ (+TNF‐α). (D) Volcano plot depicting DEGs in iMGs cultured with: (left) A_CM HCT (+TNF‐α) vs. A_CM HCT (N.S.), (middle) A_CM SCZ (N.S.) vs. A_CM HCT (N.S.), and (right) A_CM SCZ (+TNF‐α) vs. A_CM HCT (N.S.). Upregulated and downregulated DEGs are colored in red and blue, respectively. Circles and triangles indicate genes within and out of the plot axes range, respectively. (E) Gene‐set enrichment analysis (GSEA) plot results presenting the most relevant GO Biological Processes associated with each experimental condition. vs means that the indicated analysis is expressed relative to iMGs + A _CM HCT (N.S.). (F) Heatmaps indicating relative gene expression levels in iMGs after incubation with A_CM HCT (N.S.), A_CM HCT (+TNF‐α), A_CM SCZ (N.S.), or A_CM SCZ (+TNF‐α). Each plot indicates the top 25 non‐redundant genes associated with the following GO terms: Leukocyte Migration, Chemotaxis, Phagocytosis, and Receptor‐mediated endocytosis. Indicated genes may belong to more than one GO term. Data are displayed as mean‐centered values. Asterisks indicate differentially expressed genes (adjusted p value < 0.05; |log ₂ Fold‐change| ≥ 0.5).

Principal component (PC) analysis showed that iMG samples exposed to A_CM SCZ (+TNF‐α) clustered separately from iMGs cultured with any other A_CM in PC1; alternatively, PC2 segregated iMGs incubated with A_CM HCT (+TNF‐α) from all other samples. Only the third principal component effectively split samples based on A_CM sourced from either HCT or SCZ astrocytes (Figure 2B). Larger transcriptional differences were observed when iMGs were exposed to A_CM from TNF‐α‐stimulated astrocytes. Indeed, 414 differentially expressed genes (DEGs) (326 up‐ and 88 downregulated) were identified in iMGs + A_CM HCT (+TNF‐α), while 847 DEGs (388 up‐ and 459 downregulated) were found in iMGs + A_CM SCZ (+TNF‐α), and only 110 DEGs (83 up‐ and 27 downregulated) were detected in iMGs + A_CM SCZ (N.S.) (Figures 2C,D and S5A; Table S5A–D). Surprisingly, the iMGs + A_CM SCZ (N.S.) group shared more DEGs in common with iMGs + A_CM HCT (+TNF‐α) (51 DEGs) than with iMGs + A_CM SCZ (+TNF‐α) (14 DEGs) (Figure 2C). These data corroborate the PC analysis results, since the first two groups are clustered closer in PC1 compared with the latter (Figure 2B).

Gene set enrichment analysis (GSEA) further explored the biological processes (GO terms) associated with each experimental condition. Incubation with A_CM SCZ (+TNF‐α) led to the substantial downregulation of GO terms associated with several important microglial functions, including immune cell activation and proliferation, response to hypoxia, chemotaxis, cell migration, and phagocytosis. A limited amount of downregulated GO terms was also shared with iMGs + A_CM HCT (+TNF‐α), such as receptor‐mediated endocytosis, regulation of autophagy, and endosomal and lysosomal transport (Figure 2E; Table S6A). A closer inspection of the DEGs related to these downregulated GO terms unraveled that this molecular phenotype in iMGs + A_CM SCZ (+TNF‐α) was mainly driven by decreased expression of cytokines and chemokines (e.g., CCL2, CCL13, CCL18), and immune and phagocytic receptors (e.g., FPR1, TREM1, TREM2, MERTK, MARCO, IL2RA, IL2RB) (Figure 2F; Tables S5A–C and S6A). Conversely, both TNF‐α‐stimulated HCT and SCZ astrocytes induced the upregulation of GO terms associated with ribosome biogenesis, response to virus, and cellular response to interferon (IFN) in iMGs. Furthermore, TNF‐α stimulation begot a shift in the direction of iMGs' biological processes prompted by SCZ astrocytes. For example, while iMGs + A_CM SCZ (N.S.) displayed the upregulation of GO terms associated with chemotaxis, immune cell activation and proliferation, as well as the downregulation of processes related to mitochondrial metabolism, the contrary pattern was observed in iMGs + A_CM SCZ (+TNF‐α) (Figures 2E and S5B; Table S6A,B).

To confirm these results, we carried out GSEA using KEGG pathways. Indeed, we confirmed that both A_CM HCT (+TNF‐α) and A_CM SCZ (+TNF‐α) induced the downregulation of lysosomal and autophagy pathways, while only the latter promoted a significant reduction in the endocytic pathway (Figure S5C; Table S6C). Noteworthy, TNF‐α‐stimulated SCZ astrocytes also triggered the upregulation of the p53 signaling pathway, reactive oxygen species production, and multiple neurodegenerative disorders (e.g., Alzheimer's and Parkinson's Diseases) in iMGs, in marked contrast to iMGs + A_CM SCZ (N.S.) (Figure S5C,D; Table S6C,D).

To gain a better insight into how reactive SCZ astrocytes induce the downregulation of biological processes and pathways implicated in important microglial functions, we undertook a systems biology approach to establish whether their genes are organized into recognizable co‐expression modules. Weighted gene co‐expression network analysis resolved 35 modules (Figure S6A–C, Table S7). GO term overrepresentation analysis identified the biological processes associated with each module. The brown and blueviolet modules were largely enriched for terms involved in regulation of immune cell activation, autophagy, cell division and migration, chemotaxis, phagocytosis, and endocytosis (Figure S6D,E; Table S8). Notably, these two modules were negatively correlated with iMGs incubated with A_CM SCZ (+TNF‐α) and positively correlated with iMGs + A_CM SCZ (N.S.) (Figure S6C). Furthermore, the antiquewhite2 and lightgreen modules, positively correlated with iMGs + A_CM SCZ (+TNF‐α) samples, were enriched for GO terms linked to mitochondrial metabolism (Figure S6F,G; Table S8). On the other hand, the darkolivegreen2 and palevioletred modules were positively correlated with iMGs cultured in the presence of A_CM HCT (+TNF‐α) (Figure S6C). These two modules were mainly associated with biological processes related to response to cytokine and chemokine stimulation, as well as regulation of response to viruses and to IFN (Figure S6H,I; Table S8). Together, these findings corroborate the GSEA results (Figure 2E) and suggest that SCZ genetic background and pro‐inflammatory stimulation, such as by TNF‐α, are simultaneously required in astrocytes to make microglial‐like cells assume a dysfunctional molecular phenotype.

3.3. Master Regulator Analysis (MRA) Reveals Transcription Factors Involved in Establishing the Microglial Dysfunctional Response

Since co‐expressed genes often share common regulators, including upstream transcription factors (TFs), we performed in silico master regulator analysis (MRA) using a comprehensive list of human TFs (Lambert et al. 2018) to identify prospective regulators in our dataset. A total of 161 regulons were identified, of which 126 were shared between at least two groups, 5 were exclusive to iMGs + A_CM HCT (+TNF‐α) and 30 were only found in iMGs + A_CM SCZ (+TNF‐α) (Figure S7A; Table S9A–C). Next, we calculated the predicted activity status of these transcription factors and only kept those whose activity was either considered as activated or repressed in at least one experimental group (Figures S7B–E and S8; Table S9A–C). To assess putative TFs regulating the co‐expressed gene networks, we carefully cross‐examined the target genes regulated by each TF and the gene sets pertaining to each WGCNA module. IRF1, whose activity status was predicted to be activated and was the only TF exclusive to iMGs + A_CM HCT (+TNF‐α) (Figures S7B and S8; Table S9A), was associated with the darkolivegreen2 module (Figure S7E), pointing to its possible role in inducing the expression of genes involved in response to cytokine, such as IFN, and chemokine stimulation in this experimental group. Interestingly, we found four (HHEX, SOX4, MEF2C, and NFATC2) and two (CENPA and MXD3) TFs connected to the brown and blueviolet modules, respectively (Figure S7E). MEF2C was also the only TF associated with the antiquewhite2 module (Figure S7E). Except for SOX4, all these putative TFs had predicted repressed activity in the iMGs + A_CM SCZ (+TNF‐α) group (Figures S7D and S8, Table S9C). Among them, we observed that MEF2C mRNA expression was significantly reduced in iMGs cultured in the presence of A_CM SCZ (+TNF‐α), while NFATC2 and CENPA trended in the same direction (Figure 2D; Table S5C). These data point to a putative role of MEF2C, NFATC2, and CENPA in promoting the dysfunctional microglial phenotype triggered by reactive SCZ astrocytes.

3.4. Reactive SCZ Astrocytes Induce iMGs to Assume a Dystrophic/Senescent‐Like Phenotype

We were also puzzled by our results indicating that TNF‐α‐stimulated SCZ astrocytes promoted the upregulation of pathways associated with multiple neurodegenerative diseases, such as Alzheimer's (AD) and Parkinson's disease (Figure S5C,D and Table S6C,D), and decided to investigate whether reactive SCZ astrocytes could prompt these dysfunctional microglia to assume a molecular profile akin to those observed in such neurological disorders. Hence, our dataset was compared with microglial transcriptional signatures from human prefrontal cortex (PFC) brain samples obtained from patients with AD (Prater et al. 2023). iMGs cultured with A_CM SCZ (N.S.) showed a positive association with the “interferon ELN (endolysosomal network)” signature, while iMG + A_CM HCT (+TNF‐α) was associated with the “canonical inflammatory” and “cell cycle” ones. Remarkably, iMGs incubated with A_CM SCZ (+TNF‐α) were found to be positively enriched for the “dystrophic” microglial phenotype (Figure 3A).

FIGURE 3.

TNF‐α‐stimulated SCZ astrocytes promote a dystrophic microglial phenotype, leading to accelerated transcriptional age in iMGs. (A) iMGs' molecular phenotypes identified after exposure to the indicated A_CM for 24 h based on the human dorsolateral prefrontal cortex microglial transcriptional signatures described by Prater et al. (2023). (B) Microglial‐like cells signature was assessed against the CellAge database (Avelar et al. 2020), whose genes are associated with either induction (CellAge_Induces), inhibition (CellAge_Inhibits), or have a context‐dependent role (CellAge_Unclear) in promoting cellular senescence. Only significant (p < 0.1) associations are colored in (A) and (B). NES, net enrichment score. (C–H) mRNA expression of cellular senescence markers IL‐1α (C), CDKN1A (D), CDKN2B (E), MKI67 (F), MMP12 (G) and SERPINE1 (H) in iMGs cultured with A_CM HCT (N.S.), A_CM HCT (+TNF‐α), A_CM SCZ (N.S.) or A_CM SCZ (+TNF‐α). n = 3 (RNA‐seq replicates). Data are shown in TPM (transcripts per million reads). Data were analyzed by One‐way ANOVA, followed by Holm–Sidak's multiple comparison test. Bars represent Mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001. (I) Transcriptional age gap calculated from the gene expression profile of iMGs incubated with the indicated A_CM and expressed relative to the mean of iMGs + A_CM HCT (N.S.) samples (mean was set as 0). Data were analyzed by One‐way ANOVA, followed by Holm–Sidak's multiple comparison test. Compact‐letter display (a, b, and c) shows the statistically significant groups (p < 0.05).

Dystrophic microglia display morphological, molecular, and functional alterations characteristic of aged and senescent microglia (Angelova and Brown 2019; Lopes et al. 2008; Rim et al. 2024; Shahidehpour et al. 2021; Streit et al. 2004). Then, we evaluated whether A_CM SCZ (+TNF‐α) induced a senescent‐like state in iMGs by assessing their transcriptional profile against the CellAge database (Avelar et al. 2020). Interestingly, iMG + A_CM SCZ (+TNF‐α) displayed a positive association with genes that either induce or have a context‐dependent regulatory activity (“unclear” category) towards cellular senescence, whilst iMG + A_CM HCT (+TNF‐α) was also positively enriched for genes that display an inhibitory role (Figure 3B). To confirm these results, we analyzed the expression of some senescence markers (Suryadevara et al. 2024) in our RNA‐seq dataset. Curiously, reactive SCZ astrocytes induced the mRNA expression of more senescent markers (IL‐1α, CDKN1A, CDKN2B, SERPINE1, and MMP12) in iMGs than their reactive HCT counterparts (IL‐1α, SERPINE1, and MMP12), with the former displaying an even greater expression of MMP12 than the latter (Figure 3C–G). In addition, iMG + A_CM HCT (+TNF‐α) showed the augmented expression of MKI67, a well‐known proliferation marker, while the contrary was observed in iMG + A_CM SCZ (+TNF‐α) (Figure 3H), indicating a potential downregulation in cell proliferation, a cellular senescence hallmark (Suryadevara et al. 2024). To validate this data, we performed KI67 immunostaining in iMGs after incubation with all A_CM conditions for 24 h. Notably, only exposure to A_CM HCT (+TNF‐α) promoted an enhancement in the amount of KI67⁺ iMGs, while both A_CM SCZ (N.S.) and A_CM SCZ (+TNF‐α) failed to do so (Figure S9A,B). Furthermore, we did not observe changes in microglial‐like cells morphology and density after culturing them in the presence of any A_CM condition (Figure S9A,C).

Finally, iMGs' transcriptional age was calculated to verify whether reactive SCZ astrocytes promoted accelerated aging, thus potentially leading to this dystrophic microglial phenotype. Intriguingly, exposure to both A_CM HCT (+TNF‐α) and A_CM SCZ (+TNF‐α) significantly increased the transcriptional age of iMGs, while no difference was observed between A_CM SCZ (N.S.) and A_CM HCT (N.S.) (+0.275 vs. 0.000 years‐old). Nonetheless, this transcriptional age enhancement was almost two‐fold greater in iMGs incubated with A_CM SCZ (+TNF‐α) than A_CM HCT (+TNF‐α) (+7.593 vs. +3.831 years‐old) (Figure 3I). Taken together, these results suggest that reactive SCZ astrocytes trigger accelerated aging in iMGs, prompting them to assume a dystrophic/senescent‐like phenotype.

3.5. Reactive SCZ Astrocytes Impair iMGs' Synaptic Engulfment

Phagocytosis is among the key functions impaired in dystrophic microglia (Angelova and Brown 2019), whose GO term was downregulated in iMGs after exposure to A_CM SCZ (+TNF‐α) (Figure 2E). Given the strong association between synaptic uptake by microglial cells and SCZ development, we decided to assess whether reactive SCZ astrocytes could affect microglial engulfment of synaptic material (Figure 4A). To establish this assay, we first isolated synaptoneurosomes from mature hiPSC‐derived neurons (Figure S10A–E), fluorescently labeled them and added them to iMG cultures. As it can be seen in Figure S10F and Video S1, iMGs were able to successfully phagocytose labeled synaptoneurosomes. Next, we tested whether A_CM from reactive SCZ astrocytes affected iMGs' phagocytosis (Figure 4A). Interestingly, incubation with A_CM SCZ (+TNF‐α) was the only condition capable of diminishing synaptic engulfment by iMGs (Figures 4B–E, S11, and S12A).

FIGURE 4.

TNF‐α‐stimulated SCZ astrocytes impaired synaptoneurosomes engulfment by iMGs in a CX3CR1‐independent manner. (A) Experimental design of synaptoneurosomes phagocytosis assay by iMGs after pre‐treatment with AZD8797 (5 μM) and exposure to A_CM for 24 h. Schematic picture was drawn on Biorender. (B) iMGs (bright field) engulfing CM‐DiI‐labeled synaptoneurosomes (red) upon incubation with A_CM and after pre‐treatment with the CX3CR1 antagonist AZD8797. Panel depicting 16 h time‐point. Scale bar = 100 μm. (C) Quantification of synaptoneurosomes uptake by iMGs incubated with A_CM and pre‐treated with AZD8797 depicted in (B) and Figure S11. Blue symbols: A_CM from HCT astrocytes; red symbols: A_CM from SCZ astrocytes; open symbols: A_CM from non‐stimulated astrocytes; filled symbols: A_CM from TNF‐α‐stimulated astrocytes; circles and solid lines: Vehicle (DMSO)‐treated iMGs; squares and dashed lines: AZD8797‐treated iMGs. n = 355–744 (number of cells in 4 different fields of two independent experiments). Data were analyzed by Multilevel Mixed‐effects linear regression, followed by Sidak's multiple comparison test. Symbols represent Mean ± SEM. Pairwise significant comparisons are shown at the right side of the plot. (D, E) Schematic representation showing statistically significant differences depicting as reference groups vehicle‐treated iMGs + ACM HCT (N.S.) (D) and AZD8797‐treated iMGs + ACM HCT (N.S.) (E), respectively.

Considering this result, we revisited our initial screening to select a potential target that could modulate this process (Figure 1). We chose CX3CL1, as it has been previously shown to mediate synapse elimination (Paolicelli et al. 2011) and also displayed augmented secretion by stimulated SCZ astrocytes relative to stimulated controls (Figure 1N). Then, to further evaluate its role, iMGs were incubated in the presence of recombinant CX3CL1 (rCX3CL1), at approximately the same concentration secreted by reactive SCZ astrocytes (Figure S12B). Intriguingly, rCX3CL1 diminished synaptoneurosomes uptake by iMGs (Figure S12C,D). Next, we tested whether A_CM SCZ (+TNF‐α) required CX3CL1 signaling to lessen iMGs' phagocytosis (Figure 4A). Indeed, the pharmacological blockade of CX3CR1 by AZD8797 increased synaptoneurosomes engulfment relative to vehicle‐treated iMGs under control conditions (A_CM HCT (N.S.)) (Figures 4B–E, S11, and S12E). Nevertheless, CX3CR1 inhibition did not rescue the reduction in synaptoneurosomes phagocytosis by iMGs when incubated with A_CM SCZ (+TNF‐α), nor did it lead to enhanced synaptic material uptake in iMGs cultured with either A_CM HCT (+TNF‐α) or A_CM SCZ (N.S.), indicating that CX3CL1 was not responsible for promoting the observed phenotype (Figure 4B–E). Overall, these data suggest that reactive SCZ astrocytes hinder microglial synaptic material phagocytosis. However, these effects on iMGs were CX3CL1‐independent, not being reversed by the pharmacological inhibition of its receptor.

3.6. TNF‐α‐Stimulated SCZ Astrocytes Limit Microglial Migration

Cell migration is another phenomenon impaired in senescent microglia (Angelova and Brown 2019), whose CX3CL1 involvement has already been demonstrated. Hence, we next queried whether reactive SCZ astrocytes affected microglial chemotaxis. iMGs exposure to A_CM from both TNF‐α‐stimulated HCT and TNF‐α‐stimulated SCZ astrocytes enhanced microglial migration in a CX3CR1‐dependent manner (Figure 5A,B). Even though reactive SCZ astrocytes secreted more than twice as much CX3CL1 relative to their stimulated control counterparts, they failed to induce an even greater microglial migration (Figure 5A,B).

FIGURE 5.

TNF‐α‐stimulated HCT and SCZ astrocytes promoted similar microglial migration in a CX3CR1‐dependent manner. (A) Migrated iMGs after overnight incubation with A_CM and pre‐treatment with either vehicle (DMSO) or the CX3CR1 antagonist AZD8797 (5 μM). Nuclei were stained with Hoechst (grayscale). Scale bar = 100 μm. (B) Quantification of iMGs' migration shown in (A). n = 4 (number of replicates per group). Data were normalized relative to the percentage of migrated iMGs in vehicle‐treated cells cultured overnight in the presence of A _CM HCT (N.S.). Data were analyzed by Two‐way ANOVA, followed by Holm–Sidak's multiple comparison test. Bars represent Mean ± SEM. (C, D) CX3CR1 relative protein levels measured in the plasma membrane (C) and in the whole cell (D) in iMGs incubated with A_CM for 24 h. n = 9–10 (number of replicates in two independent experiments). Data were analyzed by One‐way ANOVA, followed by Holm–Sidak's multiple comparison test. Bars represent Mean ± SEM. *p < 0.05, **p < 0.01.

Intriguingly, an age‐dependent decrease in CX3CR1 surface levels has been observed in human circulating monocytes (Seidler et al. 2010). To assess whether this phenomenon could potentially limit iMGs' response to reactive SCZ astrocyte‐secreted CX3CL1, we measured the total and plasma membrane levels of CX3CR1 in iMGs exposed to A_CM for 24 h. SCZ A_CM, but not A_CM HCT (+TNF‐α), diminished CX3CR1 plasma membrane content in iMGs, while total CX3CR1 levels remained unchanged (Figure 5C,D). Taken together, these results indicate that the diminution of CX3CR1 plasma membrane levels prompted by reactive SCZ astrocytes acts to restrict microglial‐like cells migratory capabilities.

4. Discussion

Multiple lines of evidence link microglia and astrocytes to neuropathological mechanisms in SCZ (Gandal et al. 2018; Kim et al. 2024; Koskuvi et al. 2024; Szabo et al. 2021; Trindade et al. 2023; Uranova et al. 2021; Windrem et al. 2017), yet little is known about how these glial cells communicate and regulate each other's function in this disorder. Here, we applied a combination of functional assays and transcriptomics to determine how SCZ astrocytes modulate key aspects of microglial biology. Reactive SCZ astrocytes displayed a stronger response to pro‐inflammatory stimulation that, in turn, prompted microglial‐like cells to assume a dystrophic/senescent‐like phenotype. Indeed, we observed that these dysfunctional microglia display features often found in aged microglia, including the upregulation of senescence markers (e.g., CDKN1A, CDKN2B, and MMP12) and pathways (p53 signaling pathway), and the downregulation of several biological processes linked to cell proliferation, phagocytosis, and chemotaxis. Our functional assays confirmed the occurrence of these last two phenomena by showing impairments in synaptic phagocytosis and limited microglial migration in a CX3CR1‐independent and CX3CR1‐dependent fashion, respectively.

Supporting this hypothesis, post‐mortem findings reported augmented lipofuscin granules in microglia from younger and older individuals with SCZ, another characteristic cellular senescence feature (Uranova et al. 2021; Uranova et al. 2020). Furthermore, plenty of studies using anatomical, imaging, proteomic, metabolomic, and epigenomic techniques have suggested that individuals with SCZ display accelerated brain aging (Campeau et al. 2022; Caspi et al. 2024; Constantinides et al. 2023; Kaufmann et al. 2019; Lin et al. 2021; Ling et al. 2024; Stone et al. 2022; Tian et al. 2023). In addition, aged and senescent microglia, as well as other myeloid cells, have been shown to display impaired phagocytosis and cell migration (Angelova and Brown 2019; Caldeira et al. 2014; Hearps et al. 2012; Sharma 2021; Spittau 2017; Yanguas‐Casas et al. 2020).

A potential link between senescent immune cells and the CX3CL1/CX3CR1 axis was provided by Seidler et al. (2010). Indeed, the authors found that circulating monocytes harvested from older donors show diminished CX3CR1 surface expression, similar to what we observed in iMGs exposed to SCZ A_CM (Seidler et al. 2010). Given that CX3CR1 mRNA is downregulated in the blood and brain of patients with SCZ (Chamera et al. 2021; Gandal et al. 2018; Snijders et al. 2021), and knowing that the CX3CR1^A55T loss of function variant was identified as a SCZ risk factor (Ishizuka et al. 2017), we suggest that activated astrocytes contribute to the CX3CL1/CX3CR1 axis dysfunction in SCZ. Yet, one seeming contradiction in our results is that reactive SCZ astrocytes secreted twice the amount of CX3CL1 compared with HCT counterparts but did not induce a strong migratory response in iMGs. One possible explanation lies in the reduction in the CX3CR1 cell surface content in iMGs when incubated with ACM SCZ (+TNF‐α), along with the possible existence of alternate pathways leading to the transcriptional dysregulation of cell migration, as indicated by our results. Taken together, these results suggest that activated astrocytes disturb microglial CX3CL1/CX3CR1 signaling in SCZ.

We also employed computational methods to find putative transcription factors capable of promoting the microglial phenotypic transition observed here. We identified three potential candidates (MEF2C, NFATC2, and CENPA), whose predicted repressed activities in iMGs incubated with A_CM SCZ (+TNF‐α) may be responsible for downregulating the expression of genes associated with immune cell activation and proliferation, chemotaxis, and phagocytosis. Intriguingly, MEF2C and NFATC2 have already been implicated in SCZ (Cosgrove et al. 2021; Mitchell et al. 2018; Shimamoto‐Mitsuyama et al. 2021), while decreased expression of MEF2C and CENPA is associated with increased aging in microglia and p53‐dependent senescence induction in fibroblasts, respectively (Deczkowska et al. 2017; Maehara et al. 2010). These data suggest a potential mechanism explaining, at least in part, how reactive SCZ astrocytes prompt iMGs to assume a senescent‐like state. However, a thorough validation of how each TF affects microglia in the context of SCZ is beyond the scope of this work.

Despite the relevance of the data shown here, some limitations of this work remain. First, while we have provided strong evidence that reactive SCZ astrocytes induce a dystrophic phenotype in microglial‐like cells, the same set of experiments should be conducted using microglia obtained from patients with SCZ to have a more complete picture of the crosstalk between these two glial cell types in the context of this disorder. Second, it is not entirely clear what developmental time frame is modeled in this study (i.e., whether our findings recapitulate events that take place before or after symptom onset). Further research using more suitable developmental models, such as microglia‐containing organoids from individuals with SCZ, should be carried out to address this point. Moreover, since this study was conducted using iMGs from only one healthy control donor, caution should be exercised when generalizing the present findings. Finally, we only assayed the response of iMGs to soluble factors present in the astrocytes conditioned media; hence, it is important to acknowledge that potential effects driven by cell–cell interactions, which have not been investigated by this study, may also influence iMGs' response.

In summary, our results demonstrated that reactive SCZ astrocytes led to substantial alterations in microglial biology, both at functional and transcriptional levels. These findings should be viewed as part of a grander change in the immune response landscape of microglia. Together, they strengthen the hypothesis that SCZ development requires the simultaneous occurrence of a susceptible genetic background and external environmental stimuli (Muller et al. 2015).

Author Contributions

P.L.C. and F.M.R. conceived and designed the study. P.L.C. performed candidate gene screening, total RNA extraction, hiPSCs‐derived neurons characterization, synaptoneurosomal characterization, iMGs synaptoneurosomes engulfment analysis pipeline development and quantification, western blots, ELISA, cell‐based ELISA, iMGs' cell migration assay and immunostaining. G.V. differentiated hiPSCs to NSCs. G.V. and J.C.P.M. characterized hiPSCs and NSCs. P.L.C., J.P.S.L. and P.T. differentiated hiPSC‐derived astrocytes. P.T. characterized hiPSC‐derived astrocytes. P.L.C. and J.P.S.L. performed astrocytes stimulation, RT‐qPCR experiments, hiPSC‐derived neuronal differentiation, iMGs differentiation and characterization. P.L.C. and J.S.F. performed synaptoneurosomes isolation and synaptoneurosomal phagocytosis assay. J.P.S.L. carried out iMGs migration assay blind quantification. B.F.C. performed PBMC's donor psychiatric evaluation. R.N. supervised PBMC's donor candidate recruitment and helped to draft ethics committee protocol. N.C.S. drew whole blood from neurotypical donor. P.L.C., N.C.S. and R.C.C. performed PBMCs isolation from whole blood. R.C.C. and J.S.F. helped to standardize ELISA and cell‐based ELISA protocols. P.L.C., P.T. and L.C. established iMGs differentiation protocol. M.H.B. supervised iMGs differentiation protocol establishment. Y.M.H.T. performed RNA quality assessment in Bioanalyzer. P.L.C. and I.J.S.F. prepared cDNA libraries prior to RNA‐sequencing. P.L.C. and C.Y.L. conducted bioinformatic analyses and carried out statistical analyses. K.J.B. provided platforms for bioinformatic analysis. K.J.B. and F.M.R. supervised bioinformatic analysis. S.K.R. supervised hiPSC‐derived NSCs and astrocytes differentiations and donated NSCs and astrocytes cell lines, generated in his laboratory. P.L.C. and F.M.R. prepared the manuscript. L.B.V. provided research funding. F.M.R. supervised this study and obtained research grant funding. All researchers involved in this study read, made their contributions to the final version of this manuscript, and agreed to publish this work.

Ethics Statement

All experiments carried out throughout this work were approved by UFMG's Ethics Committee (COEP‐UFMG #90424518.3.1001.5149) and were performed according to the Helsinki Declaration and the Brazilian National Health Council Resolution 466/12.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: Supporting Information Figures.

Table S1: Supporting Information Tables.

Video S1: Time‐lapse video of iMGs engulfing fluorescently labeled synaptoneurosomes under basal conditions for up to 24 h. Scale bar = 100 μm.

Cardozo, P. L. , Lee C.‐Y., Lacerda J. P. S., et al. 2026. “ hiPSC‐Derived Astrocytes From Individuals With Schizophrenia Induce a Dystrophic Phenotype in Microglial‐Like Cells.” Glia 74, no. 1: e70085. 10.1002/glia.70085.

Data Availability Statement

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.