Neural organoids incorporating microglia to assess neuroinflammation and toxicities induced by known developmental neurotoxins

Graphical abstract

Highlights

• •Complex neuroimmune cerebral organoid model for use in mechanistic studies.
• •Neurotoxins evaluated via RNA-seq, cytokine and cell injury biomarkers.
• •Pathway and correlation analysis performed on RNA-seq results.
• •Findings suggest microglia may play a role in the neurotoxicity of lead (Pb) exposure.

Abstract

The use of iPSC-derived complex in vitro 3D cellular constructs is a promising avenue to more accurately predict human neural toxicities and reduce the use of animal models. We have generated a neural organoid model which incorporates iPSC-derived microglia and enables interrogation of neuroinflammation induced by pre-clinical drug candidates of varying modalities and chemical compounds in industrial use. Herein we describe the generation and characterization of this model system and its utility in assessing toxicity. We exposed the neuroimmune organoids to a variety of developmental neurotoxins and measured cellular damage by release of LDH, GFAP, and NF-L into the cell culture supernatants. Additionally, to determine whether the compounds led to activation of microglia-mediated inflammation, we measured IL-8 secretion and assessed microglia-specific gene transcriptional analysis using bulk RNA sequencing. Spearman correlation matrices using both differentially expressed genes in the RNA sequencing data and pathway analysis using Gene Ontology Enrichment revealed that microglia may play a role in the toxicity of these compounds which has been widely overlooked in standardized neurotoxicity tests. Treatment of the organoids with lead acetate demonstrates a dose–response curve of IL-8 secretion and alterations in the microglial morphology. Our findings suggest that both direct neurotoxicity and indirect neuroinflammatory mechanisms contribute to the potentially harmful effects of these compounds in the developing central nervous system.

Introduction

Testing of chemicals and preclinical pharmaceuticals for developmental neurotoxicity is important as the prenatal brain is often more sensitive to chemical assault and can result in structural or functional perturbations of the brain (Grandjean and Landrigan, 2006, Rice and Barone, 2000). Traditional regulatory approaches assess developmental neurotoxicity using in vivo tests on pregnant rats, which involve prenatal and postnatal exposure followed by the evaluation of physical, neurodevelopmental, behavioral, and neuropathological outcomes in offspring (EPA, 1998, OECD, 2007). These tests are resource-intensive and often lack clear chemical alerts (Behl, 2019, Blum, 2023, Paparella et al., 2020, Rovida and Hartung, 2009, Tsuji and Crofton, 2012). As a result, there has been a shift toward in vitro testing and integrated approaches to assess molecular and cellular disruptions associated with neurodevelopmental toxicity. New approach methodologies (NAMs), including in vitro assays and computational models, aim to reduce animal use and enhance safety testing. International efforts have led to the establishment of the Developmental Neurotoxicity In Vitro Battery (DNT IVB) (Blum, 2023, Harrill, 2018, Koch, 2022, OECD, 2023, Shafer, 2019), which identifies neurodevelopmental changes at the cellular level. There are currently 17 assays in the DNT IVB which use animal- and human-based cell cultures that can measure changes in proliferation, differentiation, apoptosis, migration, neurite formation, synaptogenesis, and neural network formation (Blum, 2023, Bal-Price, 2018, Carstens, 2022, Fritsche, 2017, Debad, 2025). Notably, these assays do not address endocrine and immune system perturbations and may not predict neuronal damage leading to the activation of microglia or microglia activation leading to neuronal damage (Gao, 2023, Qin, 2023, Woodburn et al., 2021).

The use of iPSC-derived complex in vitro 3D cellular constructs is a promising avenue to more accurately predict human neural toxicities and reduce the use of animal models (Cao, 2022, El Din, et al., 2024, Fan, 2022, Joshi, 2025, Matsui and Shinozawa, 2021, Wu, 2025). First reported by Lancaster et al. in 2013 (Lancaster, 2013), neural organoids have been used to interrogate neural development, viral infections, disease mechanisms, and neural toxicities (reviewed in (Birtele et al., 2024, Eichmuller and Knoblich, 2022). Schwartz et al originally described neuroimmune organoids generated from neural precursor cells (NPCs), mesenchymal stem cells (MSCs), endothelial cells (ECs) and microglia/macrophage progenitor cells derived from embryonic stem cells. In this work, we have developed iPSC-derived neuroimmune organoids and utilized them to screen for toxicities including microglia activation, astrocyte and neuronal damage, and to assess differentially expressed genes through transcriptional analysis.

Recently, interest in the role of microglia in both the healthy and diseased brain has risen as their function during development (Lukens and Eyo, 2022) and throughout aging (Gao, 2023) has been implicated in a variety of disorders. As the innate immune cells of the central nervous system, microglia are responsible for maintaining brain homeostasis by phagocytosing debris and pathogens, pruning synapses, and responding to assaults by regulating inflammation (reviewed in (Pallares-Moratalla and Bergers, 2024)). Microglia taken out of the physiological context of the brain and cultured on two-dimensional surfaces are transcriptionally and morphologically different than in the in-vivo setting (Cadiz, 2022). However, microglia incorporated into the organoids described here display an in vivo-like ramified morphology and demonstrate functional reactivity with appropriate cytokine secretion and gene expression signatures. Here we describe the characterization of the responsiveness of these neuroimmune organoids using single-cell RNA sequencing (scRNA-seq) and a 48-plex cytokine panel in response to LPS and IFNγ stimulation. To determine the functionality of the model in characterizing mechanisms of neurotoxicity, we subjected the organoids to a panel of known neurotoxins and controls to measure general toxicity by lactate dehydrogenase (LDH) release, cell specific damage using glial fibrillary acidic protein (GFAP) release as a marker for astrocytic damage or neurofilament light chain (NF-L) release as a marker for neuronal damage. Additionally, interleukin-8 (IL-8) secretion was measured as a marker of microglial activation. Finally, cumulative cell states in the organoid system were assessed using transcriptional changes through bulk RNA-sequencing.

The compounds used in this screen included five negative compounds recommended in guidance documents published by the Organization for Economic Co-operation and Development (OECD) (OECD, 2023), including vehicle controls of water and DMSO, glycerol, saccharin and a well-tolerated molecule in clinical use, pravastatin (Aschner, 2017). For assessing toxicity, we selected molecules that are, or have been, in clinical use and are known to cause developmental abnormalities or neurotoxicities, including thalidomide (Hallene, 2006, Qin, 2012, Vorhees et al., 2001); valproic acid (Chaudhary and Parvez, 2012, Gassowska-Dobrowolska, 2023, Martin and Manzoni, 2014, Rinaldi et al., 2008, Taleb, 2021, Zimmer, 2012) and colchicine (Joy, 2019, Krug, 2013, Mundy and Tilson, 1990, Pitts, 1991, Zaniani et al., 2013); polycyclic aromatic hydrocarbons benzo(a)pyrene and anthracene (Olasehinde and Olaniran, 2022); pesticides rotenone (Krug, 2013, Cannon, 2009, Emmrich, 2013, Gao, 2002, Gao et al., 2003, Wu and Johnson, 2007) and chlorpyrifos (Hernandez, 2015, Howard, 2005, Slotkin et al., 2012, Wu, 2017, Yen, 2011); the flame retardant tricresyl phosphate (TCP) (Knoll-Gellida, 2021); and industrial chemicals acrylamide (Erkekoglu and Baydar, 2014, Exon, 2006, Kopanska, 2022, Mundy, 2015, Pennisi, 2013, Zhao et al., 2022), bisphenol A (Al-Shami, 2024, Costa and Cairrao, 2024) and lead acetate (lead) (Zimmer, 2012) (Beaudin, 2007, Canfield, 2003). We demonstrate that the human neuroimmune organoid cultures can be assessed with multiple complementary assays, and outputs can then be deconvoluted for cell-specific contributions. This approach allows for more in-depth understanding of the mechanisms of action underlying neurotoxicity and has the potential to aid in the development of safer chemicals or novel therapeutics by identifying toxic liabilities, off-target effects, or undesirable responses.

Results and discussion

Characterization of neuroimmune organoids

Schwartz et al originally described neuroimmune organoids generated from NPCs, MSCs, ECs and microglia/macrophage progenitor cells derived from embryonic stem cells (Barry, 2017, Schwartz, 2015). This protocol has been adapted to the generation of planar neuroimmune organoids generated from cells derived from iPSCs (Majumder, 2024). The organoids are built upon a synthetic poly-ethylene glycol-based hydrogel that has been optimized for its mechanical, adhesive, and degradative properties. The resulting organoids are five to 10-cell layers thick, and remain adherent to the hydrogel, which provides a three-dimensional environment for the cells to interact with each other and deposit their own extracellular matrix environments (Majumder, 2024). The organoids are highly reproducible (Barry, 2017) and are absent of a necrotic core, common with more morphologically sphere-like organoids and spheroids (De Paola, 2023) Fig. 1.

Fig. 1.

Generation and immunofluorescence characterization of planar neuroimmune organoid. Organoids are generated in 96-well plates by the addition of iPSC-derived cells to the top of a 1) PEG-based hydrogel. Neuroimmune organoids were created by sequentially adding 2) neural progenitor cells (NPCs), 3) endothelial cells (ECs) and mesenchymal cells (MSCs), and finally 4) microglia (MGs). Organoids are cultured 21–28 days from initial NPC plating prior to treatments. For the neurotoxicity assessment herein, organoids were treated on day 23 and dosing occurred for 4 days. Cell culture supernatants were collected each day. On day 4, organoids were either harvested for transcriptional analysis or fixed for immunocytochemistry analysis. Representative images show the glial components of the neuroimmune organoids. Astrocytes and microglia were stained using antibodies to detect GFAP (green, left) or IBA1 (orange, right), respectively. Neuronal staining was performed using β3 tubulin (blue, both panels). The images shown are maximum intensity projections of 20X confocal images.

Neuroimmune organoids were cultured to day 22 and either maintained in Neural Maintenance Medium (NMM) for 24 h or in NMM containing 50 ng/mL LPS and IFNγ. Organoids were harvested for single cell RNA-Seq analysis utilizing a modified papain-based digestion, and the singularized cells were fixed with paraformaldehyde and frozen prior to being subjected to a 10X Genomics flex library generation protocol and sequenced. Fig. 2A represents UMAP visualization of scRNA-seq data including both the vehicle control and stimulated population (n = 29,508 cells total, n = 15,255 and 14,253, respectively). An average of 6,937 genes were identified in each cell. Cell types were clustered and annotated based on canonical genes described in the figure legend. We observed the maintenance of neural progenitor (12.4 %) and dividing cells (6.9 %) in the organoids and differentiation from the NPCs into radial glia (19 %), astrocytes (17 %), immature excitatory (4.1 %) and inhibitory neurons (18.6 %), and excitatory (12.7 %) and inhibitory neurons (4.7 %). We observed a population of vascular related-like cells (1 %) (Sato et al., 2023), MSCs (1.7 %), and microglia (1.9 %) (See also Supplementary Fig. S1). Microglia seeding density was assessed experimentally by bulk RNA sequencing and cytokine analysis to optimize for a homeostatic phenotype in the un-treated condition and for response changes induced with LPS + IFNγ. Each organoid contains approximately 200,000 cells. Broadly, NSG1 transcript defines the neuronal components (Fig. 2, B), SLC1A3 the glial components (Fig. 2, C),and CSF1R the microglial component (Fig. 2, D). When comparing the stimulated organoids to control, STAT1 is broadly upregulated in the LPS + IFNγ-treated organoids, while CXCL10 is upregulated in the glial populations (Fig. 2, E and F).

Fig. 2.

Transcriptomic and Proteomic Characterization of Cell Types in the Neuroimmune Organoid Model Reveals Distinct CNS-type Cell populations which are Responsive to Proinflammatory Activation. (A-F) Mapping of scRNA-seq data from Stem Pharm’s Neuroinflammatory organoid model illustrates distinct CNS-type cell types. A) Annotated UMAP representation of all cells (n = 29,508) from scRNA-seq after quality control and batch correction. Cells colored by cell type annotation. Assigned cell types as follows: Excitatory (BDNF, SLC17A6, TLX3), Immature Excitatory (NEUROG2, PHOX2B), Immature Inhibitory (LHX5), Inhibitory (SLC32A1, DLX5, GAD1), Neural Progenitors (NEUROG1, NEUROG4), Radial Glia (GBX2), Dividing Cells (MKI67), MSCs (COL2A1, COL9A3), Vascular-Related Like (FMOD, HAND2), Astrocytes (AQP4), and Microglia (AIF1, CD68). B-D) UMAP visualization of the whole scRNA-seq data set colored by B) neuronal (NSG1), C) Glial (SLC1A3), and D) Microglial (CSF1R) cell type marker expression levels per cell. E-F) UMAP visualization of the whole scRNA-seq data divided into the vehicle control vs stimulated with 50 ng/mL of LPS/IFNγ (n = 15255 vs 14253, respectively) using induction of E) STAT1 and F) CXCL10 as markers of the immune response elicited by LPS/IFNγ. (G-I) Cytokine production by Stem Pharm’s Neuroinflammatory organoid model. Data represents three biological replicates per condition with one organoid per replicate. G) Human cytokine 48-Plex Discovery array was performed by Eve Technologies and data was visualized as a heatmap colored by mean value from 3 biological replicates. Organoids were exposed to either vehicle control, 50 ng/mL of LPS/IFNγ, 10 ng/mL of TGFβ/IL-10, or 10 ng/mL of IL-4/IL-13 and supernatant was collected after 24hrs. For visualization purposes, cytokine measurements were normalized to the highest and measurements below the limit of detection were entered as zero. H) Release of IP-10 into the supernatant after stimulation occurs in both the presence and absence of microglia with gene expression of CXCL10 is detected in the glial and microglial populations. I) LPS/IFNγ stimulated IL-8 production in the organoid supernatant occurs only in the organoids in which microglia are integrated with gene expression of CXCL8 is detected only in the microglial population. One-way ANOVA and Tukey post hoc test were performed; p-value significance is as follows: *** p ≤ 0.001, **** p ≤ 0.0001. Error bars visualize standard deviation from the mean. J) Sub-clustering of the annotated microglia population from scRNA-seq reveals 4 distinct homeostatic populations distinguished by expression of K) CX3CR1 and an activated LPS/IFNγ responding cluster identified by expression of L) IFIT3. Clustering and visualization performed using Seurat. Populations Responsive to Proinflammatory Activation.

A multi-plexed cytokine panel was utilized to assess secretion of 48 cytokines and chemokines in response to treatment of LPS + IFNγ (50 ng/mL each), TGF-β + IL-10 (10 ng/mL each), and IL-4 + IL-13 (10 ng/mL each) at a 24-hour timepoint (Fig. 2, G). Notably, LPS + IFNγ stimulates a broad pro-inflammatory response while combination treatments of either TGF-β + IL-10 or IL-4 + IL-13 attenuates the VEGF-A, TNFβ, M−CSF, IL-22, and GM-CSF signal. An increase in sCD40L, typically an indicator of activated CD4 + T lymphocytes, is detectable with LPS + IFNγ treatment. Of the cell types introduced into our system, only ECs have been previously reported to have been able to be stimulated to produce CD40 ligand or sCD40L through the presence of activated CD40-expressing monocytes (Wagner, 2004). Here, CD40-positive microglial activation likely stimulates expression and secretion of sCD40L from the ECs, verifying the presence and functionality of the introduced EC population despite absence of a discrete EC population as assessed by absence of an annotated EC cluster defined by marker genes PECAM1 and CLDN5 via scRNA-seq. However, the observed vascular related-like cell population expresses high levels of both CSPG4 and PDGFRB (Supplemental Fig. S2), suggesting a pericyte-like transcriptional pattern, likely arising from MSCs in response to the introduced ECs.

We investigated the induction of IP-10 (CXCL10) and IL-8 (CXCL8) by LPS + IFNγ in more detail by analyzing stimulated organoids that either contained or did not contain microglia by immunoassays for these analytes. Cell culture supernatant levels of IP-10 are below the detectable range in the absence of LPS + IFNγ stimulation, but are greater than 20 ng/mL when stimulated in the presence or absence of microglia. UMAP plots of the single cell data suggest that the astrocyte population is contributing to this protein expression level (Fig. 2, H). When concentrations of IL-8 are measured in the supernatant, it was found that microglia were required for the induction of IL-8 by LPS + IFNγ as is confirmed by the single cell transcriptional analysis (Fig. 2, I).

Immunofluorescence analysis and confocal microscopy demonstrate uniform distribution of neurons, astrocytes, and microglia in the organoids (Fig. 1 and Fig. 6). Microglia morphology is ramified with elongated processes as is typical of homeostatic microglia. This homeostatic state is confirmed through gene expression of CX3CR1 (Fig. 2, K), which is reduced after activation with LPS + IFNγ. Likewise, increased expression of IFIT3 (Fig. 2, L) confirms that microglia within the system retain their ability to transcriptionally react to stimuli, while functionally capable of secreting a variety of cytokines (Fig. 2, G).

Fig. 6.

Exposure of neuroimmune organoids to a series of concentrations of lead indicates a time- and concentration- dependent response. Analytes were measured in supernatant collected over 4 days of continuous exposure (n = 2 biological replicates per concentration) for A) IL-8 production, B) NF-L release, C) GFAP release, and D) LDH release. Dose-response curves were fitted and EC₅₀ values were calculated with GraphPad Prism software using non-linear regression variable slope four parameters with best-fit values for EC₅₀, values were normalized to the maximum and minimum response on day 1. Error bars visualize standard deviation from the mean. E) Microglial morphological analysis on organoids fixed and stained for IBA1 and imaged at 10X using confocal microscopy demonstrates an increase in sphericity, reflecting an increase in microglial activation (n = 1–2 images with 264–386 microglia present per image). F) Representative images of the lead exposure IBA1 + microglia in the organoid imaged at 20X using confocal microscopy.

Neurotoxicity screening in neuroimmune organoids

To determine the utility of the neuroimmune organoids for toxicity testing, we subjected organoids to a panel of control compounds and known neurotoxins. Dosing concentrations were selected from literature values, and for this initial screen, concentrations on the high end of literature values were selected, with the intent to categorize appropriate differential responses to the compounds and not assess safety levels. A summary of the compounds and dosing are provided in Table 1. Interpretation of the data from this screen must be measured considering that a single, high dose of each compound was used and transcriptional analysis was performed only after four days of treatment. Follow-on studies including dose response curves and analysis at additional timepoints would be important next steps. We dosed compounds on day 23 organoids and performed half-media changes daily over four days with renewal of fresh treatment-containing media. Each day, half of the cell culture supernatant was harvested and frozen for later analysis. On day 4 post treatment, all supernatant was collected, and RNA was extracted from the organoids. Day 1 and combined supernatants from day 3 and day 4 were analyzed for IL-8 protein in the cell culture supernatant (Fig. 3, A). At day 1, colchicine, benzo(a)pyrene, permethrin, acrylamide, and lead demonstrated statistically upregulated IL-8 levels over their respective vehicle controls (DMSO or water). This finding corresponds with previously published works which demonstrated similar induction of IL-8 by colchicine in patients or in cell culture applications (Ozen et al., 2017, Lopez-Aceves, 2021, Monteiro-Riviere et al., 2003, Hung, 2021, Harshitha et al., 2024, Metryka, 2018). Day 3 and day 4 supernatants were combined to ensure adequate supernatant volume (day 3–4) and were consistent with these results, except for acrylamide, which was no longer statistically different. Day 1 and day 3–4 cell culture supernatants were also analyzed for GFAP and NF-L protein (Fig. 3, B and C) to demonstrate astrocyte and neural-specific damage, respectively. Colchicine, permethrin, TCP, and acrylamide demonstrated a statistically significant increase in GFAP compared to water and DMSO controls on day 1. However, only colchicine and rotenone demonstrated a statistically significant increase in the day 3–4 samples. Glycerol, saccharin, colchicine, permethrin, TCP, acrylamide, and bisphenol A demonstrated a statistically significant increase in NF-L on day 1 compared to water or DMSO controls. While by day 3–4, only colchicine and rotenone provoked statistically significant release of NF-L. Interestingly, permethrin treatment over time resulted in a significant decrease in NF-L levels in the supernatant. It should be noted that for the NF-L measurements, both glycerol and saccharin caused an unexpected yet statistically significant increase in NF-L in the cell culture supernatant. This result could be due to changes in the osmolarity of the media with these compounds and requires future studies to understand the mechanisms involved. When assayed in the combined day 3–4 supernatant, this increase is no longer observed with glycerol, but remains increased with saccharin. These results combined with the transcriptional analysis would suggest this may be due in part to saccharin-mediated effects on lipid homeostasis that has been previously reported in the literature (Erbas, 2018). Transcriptional analysis of glycerol and saccharin treatments (Fig. 4, B), as well as the gene ontology enrichment analysis (Fig. 5, A-B) suggest that these two treatment controls had the highest correlation with one another, suggesting that the changes observed may have similar underlying mechanisms.

Table 1.

Compounds analyzed in this study, including their common name, general class, CAS number, abbreviated rationale for their inclusion, references and dose assessed. For transcriptional analysis, compounds were dosed for 4 days.

Treatment/Toxin	Class	CAS number	Rationale	References	Dose Tested
H2O	(−) Control	7732–18-5	Solvent for hydrophilic compounds	(Aschner, 2017)	0.80 %
DMSO	(−) Control	67–68-5	Solvent for Lipophilic compounds	(Aschner, 2017)	0.50 %
Glycerol	(−) Control	56–81-5	Suggested control for MPS	(Aschner, 2017)	100 µM
Saccharin sodium salt	(−) Control	81–07-2	Non-metabolic, nontoxic sweetener	(Aschner, 2017)	100 µM
Pravastatin	Drug/ (−) Control	81131–70-6	Well-tolerated statin	(Aschner, 2017)	1.2 µM
Thalidomide	Drug	50–35-1	Known teratogen	(Hallene, 2006, Qin, 2012, Vorhees et al., 2001)	100 µM
Valproic Acid	Drug	1069–66-5	Linked to neurodevelopment disorders	(Chaudhary and Parvez, 2012, Martin and Manzoni, 2014, Rinaldi et al., 2008, Taleb, 2021, Zimmer, 2012, Gassowska-Dobrowolska, 2020)	100 µM
Colchicine	Drug	64–86-8	Microtubule polymerization inhibitor, anti-mitotic	(Joy, 2019, Krug, 2013, Mundy and Tilson, 1990, Pitts, 1991, Zaniani et al., 2013)	50 µM
Benzo(a)pyrene	Polycyclic aromatic hydrocarbon	50–32-8	Known carcinogen, potential DNT		100 µM
Anthracene	Polycyclic aromatic hydrocarbon	120–12-7	ROS generator, potential DNT	(Olasehinde and Olaniran, 2022)	100 µM
Rotenone	Pesticide	83–79-4	Inhibits NADPH, linked to Parkinson’s disease	(Krug, 2013, Cannon, 2009, Emmrich, 2013, Gao, 2002, Gao et al., 2003, Wu and Johnson, 2007)	50 µM
Chlorpyrifos	Pesticide	2921–88-2	Established cholinergic neurotoxin	(Hernandez, 2015, Howard, 2005, Slotkin et al., 2012, Wu, 2017, Yen, 2011)	100 µM
Permethrin	Pesticide	52645–53-1	Disrupts sodium channels, leads to hyperexcitability	(Lopez-Aceves, 2021, Naughton, 2024, Carloni, 2013, Harrill, 2008, Shafer et al., 2008)	100 µM
Tricresyl phosphate	Flame Retardant	1330–78-5	Known anticholinesterase activity	(Knoll-Gellida, 2021)	100 µM
Acrylamide	Industrial Toxin	79–06-1	Drives apoptosis and inhibits protective autophagy	(Erkekoglu and Baydar, 2014, Exon, 2006, Kopanska, 2022, Mundy, 2015, Pennisi, 2013, Zhao et al., 2022)	100 µM
Bisphenol A	Industrial Toxin	80–05-7	Linked to impaired brain development	(Al-Shami, 2024, Costa and Cairrao, 2024)	100 µM
Lead (II) acetate trihydrate	Industrial Toxin	6080–56-4	Linked to impaired brain development	(Zimmer, 2012, Beaudin, 2007, Canfield, 2003)	100 µM

Fig. 3.

Cytokine and Biomarker Analysis of Supernatant Reveals Cell and Time-specific Effects of Toxins in the neuroimmune organoid model A) IL-8 production measured in supernatant collected after a 24hr exposure and after 3–4 days (combined) of chronic exposure to either vehicle controls or toxin treatment conditions (n = 3 biological replicates with one organoid per replicate, supernatants from day 3 and 4 combined due to low sample volume). B) GFAP and C) NF-L release after 24hr exposure and after 3–4 days of chronic exposure to either vehicle controls or toxins (n = 3 biological replicates, supernatants from day 3 and 4 combined due to low sample volume). D) LDH measurement in the supernatant of the organoids over 4 days of continuous exposure to toxins. (6 biological replicates with one organoid per replicate). One-way ANOVA and Šidák post hoc test were performed; p-value significance shown are of compounds to their respective control (either H₂O or DMSO) and is as follows: * p ≤ 0.05, ** p ≤ 0.01, *** p ≤ 0.001, **** p ≤ 0.0001. Error bars visualize standard deviation from the mean.

Fig. 4.

Transcriptomic Characterization of Neuroimmune Organoid Model using Bulk-RNA Sequencing A) PCA plot showing replicate organoids exposed to toxins and clustered according to their variance in two dimensions. B) Spearman’s rank correlation coefficient r matrix of treatment conditions. Log2FC of all significantly modulated genes (p_adj ≤ 0.05, n = 7825 genes) were input to perform correlation analysis. Heatmap colored by Spearman’s correlation coefficients (ρ), exact values shown inset in heatmap squares. C) Representative subset (n = 70 of 7825 significantly modulated genes, p_adj ≤ 0.05) of cell specific genes altered by treatment conditions, visualized in a heatmap colored by Log2FC to their respective vehicle control (H₂O or DMSO).

Fig. 5.

Pathway Analysis by Gene Ontology Enrichment of the Significantly Modulated Gene Transcriptional Changes in the Treatment Conditions Indicates Both Colchicine and Lead Treatments Result in Unique Alterations in the Immune Response in the Neuroimmune Organoid Model. A) Spearman’s rank correlation coefficient r matrix of the overlap between Gene Ontology enrichment analysis (p ≤ 0.05, n = 4611 terms) was performed using binary input of whether a GO term was significantly modulated in the gene expression data (1) or absent (0). B) Spearman’s rank correlation coefficient r matrix of the overlap between the upregulation or downregulation of the Gene Ontology enrichment analysis (p ≤ 0.05, n = 4611 terms) was performed using binary input of whether a GO term was significantly upregulated (1) or downregulated (−1). Data visualized as a heatmap colored by Spearman coefficient (ρ) using GraphPad Prism 10. ‘X’ indicates where no correlation values could be calculated. C) Representative subset (n = 69 of 4611) significantly modulated pathways, p ≤ 0.05) altered by treatment conditions, visualized in a heatmap colored by Log10 p-value to their respective vehicle control (H₂O or DMSO), only statistically significant p-values are colored and those that did not reach statistical significance are left blank.

Detection of LDH release into the cell culture supernatant, a general indication of cell toxicity and damage, was performed with a luciferase-based substrate assay on samples from each of the four days of the experiment. Eight of the fifteen treatment groups demonstrated a statistically significant release after 24hr of exposure. However, only three of these were statistically higher over controls on day 2 (colchicine, anthracene, and rotenone) and two on days 3 and 4 (colchicine and rotenone) (Fig. 3, D). Of note, LDH levels in rotenone-treated samples were not statistically increased until day 2 yet reached statistically significant levels by day 2 which were sustained until the end of the experiment. Similarly, the release of GFAP and NF-L by rotenone was also not observed in day 1 supernatants but was very high in the day 3–4 samples. Indeed, the rotenone-treated organoids demonstrated significant cytotoxicity and RNA isolated from day 4 rotenone-treated cultures failed quality control standards for transcriptional measurements as the RNA was too degraded. Rotenone is an inhibitor of the mitochondrial complex I which leads to ATP depletion, the formation of reactive oxygen species, oxidative stress and apoptosis (Heinz, 2017). Given its potent effects on cell viability on day 2–4, it is surprising that damage was not observed in day 1 cultures by GFAP or NF-L release. This demonstrates that assay timing is crucial when conducting screens with a variety of assay readouts and mechanisms of action.

RNA was harvested from day 4 treatment groups and analyzed by bulk RNA-Seq utilizing the Illumina RNA Prep with Enrichment, (L) Tagmentation protocol. Principal component analysis (PCA) demonstrates that among the treated samples analyzed, colchicine caused the largest transcriptional changes between the samples (Fig. 4, A). All the other treatments clustered similarly on the PC1 axis and demonstrated variability on the PC2 axis. We analyzed differentially expressed genes between each treatment group compared to respective vehicle control and found that of the 60,609 transcripts measured, 7,825 reached statistical significance in at least one of the treatment conditions. Correlation analysis using Spearman’s rank correlation coefficient matrix of the treatment conditions indicates that at the individual gene transcription level, anthracene and chlorpyrifos (ρ = 0.79), along with permethrin and TCP (ρ = 0.73) have the highest correlation coefficients (Fig. 4, B). Of the significantly modulated genes, a representative subset of canonically expressed CNS cell type-specific genes was used to generate a heatmap, with genes clustered by cell type or function including microglial, pan-glial (glia), neuronal, extracellular matrix (ECM), or cell death-related genes (Fig. 4, C). As observed by PCA analysis and correlation matrix, the effect of colchicine is unique among the treatments and is apparent in the gene transcription signature of microglia, pan-glia, ECM, and neurons. Within the microglial and glial-specific genes, lead provoked a pattern of neuroinflammation similar to colchicine, exhibiting loss of homeostatic CX3CR1 and increases in cytokines CXCL9 and CCL13.

Pathway analysis was performed using gene ontology (GO) enrichment of the significantly modulated gene transcriptional changes in the treatment conditions, yielding a total of 4611 unique GO terms. A Spearman’s rank correlation matrix of the modulated GO terms indicates that many of the compounds modulate the same pathways (Fig. 5, A). Another Spearman’s correlation matrix on the directionality of the overlapping GO terms shows a high correlation in the up or down-regulation of GO term pathways among benzo(a)pyrene, anthracene, and chlorpyrifos (Fig. 5, B). This high correlation coefficient is due primarily to the ability of these compounds to modulate the cell cycle, as illustrated with the representative GO term heatmap (Fig. 5, C). As expected, most of the neurotoxic compounds significantly modulate developmental pathways, leading to disruptions in synapses and in cell signaling. Additionally, as observed previously in the gene expression data (Fig. 4, C), colchicine and lead also modulate similar GO terms involved in the immune response, with lead showing specificity in modulating the interferon pathway (Fig. 5, C).

As a microtubule inhibitor used to prevent or treat attacks of gout, colchicine is known for having anti-inflammatory effects and has been proposed as having applications in treating other chronic inflammatory diseases (Dalbeth et al., 2014). However, our findings, like others (Golpour, 2021, Schwarz, 2016), suggests that colchicine is capable of inducing an acute pro-inflammatory state (Fig. 3, A), which may, consequentially, exacerbate cell injury to astrocytes and neurons (Fig. 3, B-C). Colchicine is reported to have very low blood–brain barrier (BBB) penetration (Matsuyama, 1989), however, the toxicity and pro-inflammatory effect of the compound would suggest caution be taken in its use among individuals with a compromised BBB or pre-existing neuroinflammatory condition.

Interestingly, in contrast to colchicine which led to significant upregulation of all metrics of cell injury and were not cell-type specific, lead evoked a neuroinflammatory response prior to any cell injury biomarkers. Treatment with lead resulted in similar IL-8 levels as colchicine both acutely and after 3–4 days of continuous exposure. Though the proteomic measurements after 4 days did not result in measurable increase in GFAP or NF-L, pathway analysis point to significant transcriptional changes in cell death and development. The effect of lead on the immune system of individuals with acute or chronic exposure has shown that the heavy metal can result in significant increases in TNF-α with eventual dysregulation of the immune response (Harshitha et al., 2024). The high induction of IFIT1; ISG15, and MX2 (Fig. 4, C) corresponds with increased interferon signaling in the organoids. There is evidence that the effects of continuous lead exposure results in alterations of the interferon antiviral response (Gainer, 1974) with studies in both occupational exposure (Dobrakowski, 2016) and in early life exposure (Zheng, 2021) finding that higher blood lead concentrations were positively correlated with serum levels of IFNγ. As lead readily crosses the blood brain barrier and can accumulate in cells as a calcium analog, lead poses a particular threat to neurological health from in utero development through adulthood (Collin, 2022). Our findings suggest that one of the earliest observed effects of lead exposure in the brain is an activation of the immune response in microglia, which occurs foremost in our model over markers of proteomic cell injury (Fig. 3). This lead-induced neuroinflammatory phenotype has been previously reported in rodent models (Shvachiy, 2023, Su, 2021) and has important implications in understanding the mechanisms by which lead-mediated neurotoxicity occurs and how it might be prevented or slowed.

To further characterize the observed response to lead, we performed a 6-point dose–response experiment over time, assessing IL-8, NF-L, GFAP and LDH secretion daily over four days (Fig. 6). Unlike the screening experiment, here we performed a complete media change daily to ensure we had enough supernatant for downstream analysis at each data point and to assess secretion within a defined window without carry-over from the previous day. IL-8 secretion at the highest concentrations of lead (100 µM) was similar on days 1 and 2-post exposure but decreased daily on days 3 and 4 (Fig. 6, A). IL-8 secretion was dose-responsive, with an observed EC₅₀ value 3.5–9 µM across the four days. NF-L, GFAP and LDH release was also dose-dependent at day 1, with EC₅₀ values ranging from 1.2 – 6.5 µM. With lead treatment, the observed neuronal and astrocyte toxicity is notably acute, with little detectable NF-L or GFAP release observed after day 1. To further characterize the effects of lead on the microglial population, we performed immunofluorescence analysis and microglia sphericity measurements on organoids that had been treated with the various lead concentrations for 4 days. Mean microglia sphericity values increase over vehicle control at each lead concentration evaluated (Fig. 6, E and F). A wide range of microglia morphologies have been described in the literature in response to an injury or diseased state (Doorn, 2014, Green and Rowe, 2024, Morrison, 2017, Savage et al., 2019). Loss of microglia ramifications and changes in microglia sphericity measurements can be attributed to microglia in a more active inflammatory, (Adrian, 2023, Stence et al., 2001, Torres-Platas, 2014, Uhlemann, 2016) phagocytic, (Doorn, 2014, Badanjak, 2021) or motile (Marin-Teva, 2011, Wicks, 2022) state and can be used with other analytical methods to describe effects on the microglial population. Here, concurrent with transcriptional changes and release of IL-8 into the supernatant, the change in morphology suggests treatment with lead acetate drives a neuroinflammatory response.

In contrast to our neuroinflammatory findings for lead, injury inflicted by TCP does not appear to have a microglia-mediated component. Unlike colchicine and lead, TCP did not elicit IL-8 production (Fig. 3, A), yet treatment on day 1 incurred significant increases in GFAP, NF-L, and LDH release (Fig. 3, B-D). In this case, later upregulation of CCL18 and CCL13 (Fig. 4, C) would appear to be the result of indirect microglial response to neuronal and astrocytic cell injury, rather than to the direct response from compound treatment itself. Interestingly, both permethrin and acrylamide exhibited significant increase in all markers of cell injury and activation at day 1, however these subsided by measurements performed at later time points (Fig. 3). Further transcriptional analysis at earlier time points would be of interest in these cases. Of note, thalidomide, a known teratogen that is not considered a developmental neurotoxin (Kim and Scialli, 2011) did not generate significant changes in IL-8, GFAP, NF-L or LDH signals and transcriptionally correlated most closely to the control conditions in our screen. This is consistent with negative results observed in DNT IVB assays (Harrill, 2018, Frank, 2017).

The contribution by environmental pollutants brought by anthropogenic influence on the quality of soil, water, and atmosphere has great potential in affecting human health (Ismail et al., 2005). Identification and limiting the use of these toxins is of critical importance in preserving quality of life and neuronal health for both current and future generations. Microglia play an essential role in all stages in life, sculpting synapses and activating an immune response to protect the brain from infection and injury. However, persistent inflammation in the brain has been implicated as a driver of neurodegenerative disorders (Muzio et al., 2021). The contribution of neuroinflammation to neurotoxicity has been widely overlooked in the literature. As recent advances in neuroscience have revealed, chronic neuroinflammation can directly and indirectly impact neuronal health and exacerbate neurodegeneration. The inflammatory component of neurotoxic screening should be taken into consideration in assessing long-term exposure of the population and its effect on neurological health.

Materials and methods

Cell culture

iPSC-derived human NPCs from a male donor, (FCDI, Madison, WI) were maintained in complete STEMdiff Neural Progenitor Medium (NPM), prepared by adding 1 mL of the Neural Progenitor Supplement A (STEMCELL Technologies 05836) and 50 μL of the Neural Progenitor Supplement B (STEMCELL Technologies 05837) to 50 mL of Neural Progenitor Basal Medium (StemCell Technologies 05834). When thawing or passaging, 10 μM Y27632 (Rock inhibitor; Chemdea NC0407157) was included in the medium and removed day 1 post thawing/passaging. NPCs were cultured on Geltrex (Gibco A1413302) diluted 1:100 with DMEM/F12 (Gibco 11330032). Cells were cultured at 37 °C, 5 % CO₂ atmosphere and passaged at 95 % confluent with StemPro Accutase (Gibco A1110501) diluted 1:1 with DPBS (Gibco 14190094). NPCs were assessed for expression of NPC markers SOX2, Ki67 and Nestin by immunofluorescence analysis and are used within 5 passages of the original cell banking.

iPSC-derived human iCell ECs (FCDI C1114) were cultured according to the supplier’s recommendations in VascuLife Basal Medium (LifeLine LL-0003) with VascuLife VEGF LifeFactors. Heat Inactivated FBS (HyClone SH30071.03HI) was added to the complete media at 10 % in place of the supplied FBS. The supplied antimicrobial supplement was not added to the complete media. ECs were maintained on a coating of 3 μg/cm² Fibronectin (Sigma FC010). Cells were cultured at 37 °C, 5 % CO₂ atmosphere and passaged at 80 % confluent with TrypLE Express (Gibco 12605028). Cells are used within 5 passages of receipt from the supplier.

iPSC-derived human iCell MSCs (FCDI R1098) were cultured in complete MSC medium, 0.25X Ham’s F-12 (Gibco 11765054), 0.75X IMDM (Gibco 12440046), 50 μg/mL Ascorbic Acid (Sigma A8960), 0.5X B-27 Supplement Minus Vitamin A (Gibco 12587010), 50 ng/mL bFGF (R&D Systems 233-FB), 0.05 % BSA (Gibco 15260037), 1X GlutaMAX (Gibco 35050061), 450 μM MTG (Sigma M6145), 0.5X N-2 Supplement (Gibco 17502048), 50 ng/mL PDGF-BB (PeproTech 100-14B). MSCs were cultured on a coating of 5 μg/mL Fibronectin (Sigma FC010) and 10 μg/mL Collagen I, (Gibco A1048301). Cells were cultured at 37 °C, 5 % CO₂ atmosphere and passaged at 80 % confluent with TrypLE Express (Gibco 12605028). Cells are used within 5 passages of receipt from the supplier.

Organoid generation and Culturing

Human neuroimmune organoids were generated from NPCs, microglia (MG) (FCDI C1110), ECs, and MSCs. Except for the microglia, cells were cultured per the cell supplier’s instructions prior to plating onto a PEG-based hydrogel substrate optimized for this application (proprietary formulation, Stem Pharm, similar to that described in (Barry, 2017, Schwartz, 2015, Majumder, 2024). iPSC-derived Microglia were added directly to organoids from cryopreservation. Generation of planar neural organoids have previously been described (Barry, 2017, Schwartz, 2015, Majumder, 2024). The PEG-based hydrogel was polymerized in μ-Plate 96 Well 3D plates (Ibidi 89646), and equilibrated overnight in PBS and then in NMM prior to plating NPCs. NPCs, MSCs, ECs and MG were plated in serum free medium at the following time points; NPCs day 0 (25 K/well), MSCs and ECs day 3 (18 K and 1.8 K/well respectively), MG day 14 (12.5 K/well). Organoids were maintained in Neural Maintenance Media (NMM): DF3S (DMEM/F12 (Gibco 11330032) supplemented with 64 μg/mL L-ascorbic acid-2-phosphate magnesium (Sigma A8960), 14 ng/mL Sodium Selenium (Sigma S5261), 543 μg/mL Sodium Bicarbonate (Gibco 25080–094)) supplemented with 1X B-27 Supplement (Gibco 17504044), 1X N-2 Supplement (Gibco, 17502048), 1X GlutaMAX (Gibco 35050061), 1X MEM NEAA (Gibco 11140050), 1X Penicillin-Streptomycin (Gibco 15140122). NMM was supplemented with 5 ng/mL Heat Stable bFGF (Gibco PHG0367) day 0–5 and 100 ng/mL VEGF (R&D Systems 293-VE) day 3–13. Cultures were fed daily with 50 % of medium removed and added back with fresh medium. The organoids were cultured until day 21–23 post plating with NMM prior to stimulation or dosing of compounds.

Stimulation of organoids

For cytokine analysis on day 21 post plating, neuroimmune organoids were stimulated with 50 ng/mL of each LPS (Sigma L6529) and IFNγ (R&D Systems 285-IF) or 10 ng/mL of each TGF-β (PeproTech 100–21) and IL-10 (R&D Systems 217-IL) or 10 ng/mL of each IL-4 (R&D Systems 204-IL) and IL-13 (R&D Systems 213-IL). Organoids were cultured for 24 h at 37 °C and 5 % CO₂ prior to collecting cell supernatant. The collected supernatant was frozen at −80 °C prior to cytokine analysis. Supernatant was analyzed with the Human Cytokine/Chemokine Panel A 48-Plex Discovery Assay® Array (HD48A) (Eve Technologies, Calgary, AB, Canada). Three biological replicates were assessed for each condition, one organoid per biological replicate.

Immunofluorescence analysis

Untreated day 28 organoids were fixed with 4 % paraformaldehyde in PBS for 1 h and stored in PBS at 4 °C. Organoid samples were permeabilized and blocked with PBS containing 10 % Donkey Serum (Sigma D9663), 0.2 % TritonX-100 (Sigma T9284) and 0.02 % Sodium Azide (Fisher S227I) (Antibody Incubation Solution, AIS) for 1 h. Anti-GFAP (1:250, AbCam AB33922), or Anti-IBA1 (1:250, Wako 019–19741), were incubated overnight at 4 °C and washed three times in PBS. Donkey anti-rabbit AF555 antibody (1:500, Thermo A31572), and AF-647 conjugated anti β3 tubulin antibody (1:250, RnD Systems IC1195R), was incubated overnight at 4 °C and washed three times in PBS. Images were captured on a Nikon AX R confocal microscope with a 20X objective at the University of Wisconsin-Madison Optical-Imaging Core. Image processing was performed in Nikon NIS-Elements and Fiji(Schindelin, 2012). Maximum intensity projection images are presented. 3D microglia morphology analysis was run on images taken on a Nikon AX R confocal microscope with a 10X objective. Images were acquired from the center of the well with a 3 µm step interval. Analysis was performed using Fiji (Schindelin, 2012), 3D Objects Counter (Bolte and Cordelieres, 2006), and 3D ImageJ Suite (Ollion, 2013). Microglia objects were isolated from the IBA1 channel using 3D Objects Counter. Objects were added to the 3D ImageJ Suite 3D Manager to measure their volume and sphericity. Microglia objects were filtered for objects > 4,000 and < 60,000 vol in calibrated unit. Sphericity in calibrated unit was graphed and analyzed by one-way ANOVA with multiple comparisons. Each treatment analysis used 1–2 images, with 264–386 microglia present per image.

Toxicology screen dosing and Harvest

On day 23 post plating, neuroimmune organoids were dosed with known neurological toxins and controls (Table 1). Three biological replicates were assessed for each condition, one organoid per biological replicate. Organoids were treated for 4 days with 50 % media changes containing compound each day. Supernatant was collected and stored at −80 °C each day. On day 4, supernatants were collected, and organoids were harvested for RNA using TRIzol Reagent (Invitrogen 15596018). Lysates were processed through a QIAshredder (Qiagen 79656). Samples were frozen at −80 °C prior to RNA purification. RNA was purified using a chloroform extraction protocol utilizing Phasemaker Tubes (Invitrogen A33248), and RNA was purified using the RNA Clean and Concentrator-5 with DNase treatment kit (Zymo Research R1014). Samples were eluted in 15 μL of DNase/RNase free water and stored at −80 °C.

RNA Quantification and Qualification

RNA was quantified following the manufacturers protocol for the Quant-iT RiboGreen RNA Assay Kit (Invitrogen R11490). RNA was qualified using a gel electrophoresis protocol as described below. 20–30 ng of RNA was combined with 2X RNA loading Dye (NEB B0363S) and denatured at 70 °C for 10 min then placed on ice for 2 min. The ssRNA Ladder (NEB N0362S) was also denatured at 70 °C for 10 min then placed on ice for 2 min. RNA samples and Ladder were loaded on a 1.5 % Agarose (Fisher BP160) Gel. The gel was run at 80–100 V for about 60–90 min. The gel was stained using SYBR Gold (Invitrogen S11494) and imaged to determine the integrity of the 28S (∼4.8 kb) and 18S (∼2 kb) rRNA bands.

RNA library Prep and characterization

The Illumina RNA Prep with Enrichment, (L) Tagmentation protocol (Document # 1,000,000,124,435 v03) was followed for the RNA Library Prep. 50 ng of RNA was inputted per sample. The index plate used was the IDT for Illumina – DNA/RNA UD Indexes Set B (Illumina 20025080). Throughout the RNA Library Prep protocol, DNA was quantified using the Quant-iT dsDNA Assay Kit, High Sensitivity (Invitrogen Q33120). DNA was qualified using a gel electrophoresis protocol as described below. 20–30 ng of DNA was combined with 6X TriTrack DNA loading Dye (Thermo Scientific R1161) and denatured at 90 °C for 2.5 min then placed on ice for 2 min. The GeneRuler 100 bp DNA Ladder (Thermo Scientific SM0243) was not denatured prior to loading onto the gel. DNA samples and Ladder were loaded on a 1.5 % Agarose (Fisher BP160) Gel. The gel was run at 80–100 V for 60–90 min. The gel was stained using SYBR Gold (Invitrogen S11494) and imaged to determine the size of the library fragments and to confirm lack of genomic DNA or primer contaminants.

Bulk RNA-Seq and analysis

Pooled library samples were sent to the University of Wisconsin Biotechnology Center for bulk RNA-seq. Samples were qualified using a Bioanalyzer (Agilent Technologies) and analyzed using the TapeStation Analysis Software 5.1 (Agilent Technologies). Pooled samples were sequenced on the NovaSeq X Plus (Illumina). Sequencing data was analyzed through Illumina BaseSpace Sequence Hub (v7.28.0). Samples were concatenated with DRAGEN FASTQ Toolkit (v1.3.1) and aligned using DRAGEN RNA (v4.3.13) with the human reference genome GRCh38. Differential expression analysis was performed using DRAGEN Differential Expression (v4.3.7). Results from DRAGEN Differential Expression were then run through an R pipeline using gprofiler2 (v0.2.2) to perform gene enrichment analysis (Kolberg, 2020, Sievert, 2020).

Neural organoid dissociation for scRNAseq

On day 23, six neuroimmune organoids from control or LPS + IFNγ stimulated (50 ng/mL each, 24 h) conditions were dissociated using a modified protocol from the Papain Digestion Kit (Worthington Biochemical Corp. LK003150) and fixation protocols from 10x Genomics (Pleasanton, CA). Organoids were washed with LINC wash buffer (1 % Pluronic F-68 (Gibco 24040032), 0.04 % BSA (Gibco 15260037) in PBS) in the culture well. The organoid was transferred to a round bottom tube containing the dissociation solution (63 % TrypLE Express (Gibco 12605010), 27 % Papain (Worthington LK003150), 4.5 % DNase (Worthington LK003150), 2 U/μL Protector RNAse (Roche 3335399001), 5 μg/mL actinomycin D (Sigma A1410), 10 μM triptolide (Sigma T3652), 27 μg/mL anisomycin (Sigma A9789)). Samples were placed at 37 °C for 40 min with titration every 10 min to dissociate. The samples were centrifuged at 300 x g for 5 min. The dissociation solution was aspirated, and samples were washed with LINC wash buffer. Samples were passed through a 100 μM cell strainer and centrifuged at 500 x g for 5 min at 4 °C. The wash buffer was aspirated, and cells resuspended in LINC wash buffer. Samples were counted and checked for viability. Cells were fixed with 2 % paraformaldehyde (EMS 15710) for 1 h at room temperature. Equal parts LINC quenching solution (0.1 % TRIS (Fisher BP2471500) in LINC wash buffer) was added to the fixed cells. Samples were centrifuged at 500 x g for 5 min at 4 °C. Solution was aspirated and samples resuspended in Wash buffer then counted. Samples were centrifuged at 500 x g for 5 min at 4 °C. Wash buffer was aspirated, and samples resuspended in LINC freezing buffer (5 % DMSO (Sigma D2650), 10 % Glycerol (Sigma G5516), 0.2 U/μL SUPERaseIN RNase Inhibitor (Invitrogen AM2694) in LINC wash buffer). Samples were frozen at −80 °C for 24 h prior to moving to liquid nitrogen for long term storage.

scRNA-Seq

scRNA-seq was performed at the University of Wisconsin Biotechnology Gene Expression Center. Seurat (v5.1.0) (Hao, 2024) was used for quality control, normalization, sample integration, cell clustering, and visualization. Cells were filtered for features > 4,000 and < 10,000, transcripts < 60,000, and mitochondrial percent < 10 %. Clustering was performed using UMAP. Cell clusters were annotated based on the expression of key marker genes identified for each target cell type. The scRNA-seq data for each cluster was first collapsed using AggregateExpression() function in Seurat (Hao, 2024), generating bulk-like profiles that represented the pooled expression of all cells in each cluster. The median expression of marker genes associated with each target cell type was then calculated from these bulk-like profiles. Each cluster was subsequently assigned to the cell type with the highest median expression score for its corresponding marker genes. The list of target cell types and their associated marker genes included Astrocytes (AQP4), Dividing Cells (MKI67), Excitatory Neurons (BDNF, SLC17A6, TLX3), Immature Excitatory Neurons (NEUROG2, PHOX2B), Immature Inhibitory Neurons (LHX5), Inhibitory Neurons (SLC32A1, DLX5, GAD1), MSCs (COL2A1, COL9A3), Microglia (AIF1, CD68), Radial Glia (GBX2), Vascular-Related Like (FMOD, HAND2), and Neural Progenitors (NEUROG1, NEUROG4).

10x Genomics Loupe Browser v8.0.0 was used for visualization and differential expression analysis to assist with marker gene identification.

Biochemical assays

Changes in Lactose Dehydrogenase in the cell culture supernatants were assessed with the LDH-Glo Cytotoxicity Assay (Promega J2381) per the manufacturer’s protocol in 384-well format in Corning 3574 plates. Samples were diluted 1:200 prior to being added to the assay plate. NF-L and GFAP concentrations in the cell culture supernatants were quantified using the Human NEFL SimpleStep ELISA Kit (Abcam ab288171) and the Human GFAP SimpleStep ELISA Kit (Abcam ab288175) per the manufacturer’s protocol, in 384-well format. Samples were diluted 1:100 for the GFAP assay prior to being added to the assay plate. Samples from day 1 were diluted 1:5 and day 3–4 were diluted 1:2 for the NEFL assay prior to being added to the assay plate. IL-8 and IP-10 concentrations in the cell culture supernatants were assessed with either the Lumit IL-8 Immunoassay (Promega CS2032C02) following manufacturer’s protocol in a 384 well plate (Corning 3574) or with an IL-8 ELISA (RayBiotech ELH-IL8) or IP-10 ELISA (RayBiotech ELH-IP10) following the manufacturer’s protocol. Samples were diluted 1:10 for the IL-8 ELISA and 1:40 for the IP-10 ELISA. Data was analyzed by one-way ANOVA with multiple comparisons and a post-hoc Dunnett’s test, comparing the vehicle controls (either H₂0 or DMSO) to the means of the samples with that vehicle control. Measurement of IL-8 induction by LPS + IFNγ stimulation in the organoids was employed as a lot-to-lot quality control metric to demonstrate microglia incorporation and responsiveness to stimulation prior to use in screening. Error bars reflect the standard deviation from the mean.

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Nina Y. Yuan, William D. Richards, Kailyn T. Parham, Sophia G. Clark, and Connie S. Lebakken are employees of Stem Pharm, Inc. and have exercisable stock options. Connie S. Lebakken is a stock holder of Stem Pharm. Steven W. Smith is a consultant to Stem Pharm.

Appendix A. Supplementary data

The following are the Supplementary data to this article:

Data availability

Data will be made available on request.