Astrocyte induction of disease-associated microglia is suppressed by acute exposure to fAD neurons in human iPSC triple cultures

Abstract

Advancements in human induced pluripotent stem cell (hiPSC) technology have enabled co-culture models for disease modeling in physiologically relevant systems. However, co-culturing protocols face challenges in usability and consistency. Here, we introduce a robust, reproducible hiPSC-derived co-culture system integrating astrocytes, neurons, and microglia. This model leverages cryopreserved cells, enabling co-cultures within 20 days post-thaw. Comparing monocultures and tricultures, we demonstrate how cell-cell interactions shape transcriptional and functional states across all three cell types. Neurons in triculture exhibit increased spine density and activity, while astrocytes and microglia show altered responses to proinflammatory stimulation. Surprisingly, the presence of astrocytes induces upregulation of disease-associated microglia (DAM) genes, including TREM2, SPP1, APOE, and GPNMB in microglia. Additionally, while familial Alzheimer's disease neurons induce a prototypical inflammatory response in microglia, the DAM signature is significantly dampened. Collectively, this study establishes a versatile human triculture model as a valuable resource for dissecting neuron-glia interactions and their role in neurodegenerative disease.

Keywords: APOE; Alzheimer’s disease; CP: Neuroscience; CP: Stem cell research; TREM2; astrocytes; disease-associated microglia; fAD; iPSCs; microglial states; triculture.

Figure 1.. Selection of the optimal media for triculture of human iPSC-derived neurons, astrocytes, and microglia

(A) Schematic of the iPSC differentiation protocols for neurons (iNs), astrocytes (iAs), and microglia (iMGs). iNs and iAs were generated by lentiviral expression of lineage-specific transcription factors, while iMGs were derived via a hematopoietic precursor (HPC) stage using non-viral methods. Cryopreservation days are indicated; each cell type was fully differentiated before switching to triculture media (TCM). Key abbreviations: KSR, knockout serum replacement; N2B, neurobasal with N2/B27; EM, expansion medium; FGF, fibroblast growth factor medium. (B) Table summarizing composition of each cell-type-specific media and TCM (see Table S2 for full details). (C–E) Representative immunostaining of neurons (C), astrocytes (D), and microglia (E) maintained in either their respective media or TCM. Markers shown include TUJ1/NeuN for neurons, GFAP/S100B for astrocytes, and IBA1/INPP5D for microglia. Scale bars, 50 μm. (F) Timeline of the triculture workflow. Cryopreserved iNs, iAs, and iMGs were thawed and matured separately; astrocytes and microglia were sequentially plated onto neuron cultures on days 20 and 21 and then co-cultured for 3–6 days. The days in bold refer to the start of thawing the first stock of cryopreserved cells, and the non-bolded days refer to the day of differentiation for each individual cell type. (G) Bar plot showing the relative percentages of NeuN⁺ (neurons), IBA1⁺ (microglia), and CD44⁺ (astrocytes) cells at day 27, determined from immunostaining (n = 2 genetic backgrounds, 3 differentiations, and 2 wells per differentiation). Error bars represent standard error. See Figure S1E for individual well values. A representative field of view (FOV) is shown, with six FOVs per well analyzed by blinded quantification. Scale bar, 200 μm. (H) Representative triculture images (3–6 days of co-culture) labeled for CD44 (astrocytes), IBA1 (microglia), and TUJ1 (neurons). Scale bars, 100 μm. (I) Western blot of tricultures at days 3 and 6, probed for INPP5D, IBA1, TAU5, CD44, and GAPDH, confirming the presence of all three cell types.

Figure 2.. Triculture enhances neuronal spine density and mean firing rate

(A and B) scRNA-seq was performed on monocultures (MCs) of each cell type and on tricultures (TCs) after 6 days of co-culture (total 58,322 cells). UMAP plots show clustering by marker expression, with clusters labeled by cluster index in (A) and by cell/culture type in (B). For all scRNA-seq data, one genetic background was used. See Figures S2A–S2G for differentiation markers. (C) Proportional composition of each cluster, color coded to match (A). (D–F) Microglial cell, astrocyte, and neuron populations were isolated, re-clustered, and colored by culture condition (MC, orange; TC, green). (G) Dot plot of significantly enriched Gene Ontology (GO) pathways in iNs, iAs, and iMGs (DEGs from Figure S2H). Dot size reflects the number of DEGs in each pathway; color indicates adjusted p value. See Table S3 for additional details. (H) DiI(1,1’-dioctadecyl-3,3,3’,3’-tetramethylindocarbocyanine perchlorate) labeling of neuronal dendritic spines in MC and TC. Spine density was measured by reconstructing dendritic cables and counting spines per dendrite length; two genetic backgrounds for glial cells; n = 3 independent co-cultures and 3 wells per co-culture, with blinded analysis. Each dot represents the average of five dendrites per well; *p < 0.05 by mixed-effects model. Scale bar, 5 μm. (I) Synaptic vesicle (SV) release was assessed in MC and TCs using live-cell imaging of SypHy, a pH-sensitive fusion protein that fluoresces upon vesicle fusion and exposure to extracellular pH (7.4) after KCl stimulation. iNs were transduced with SypHy-expressing lentivirus on day 19, with iAs added on day 20 and iMGs on day 21. On day 27, iNs were imaged before and during KCl stimulation, followed by NH₄Cl treatment to unquench SypHy fluorescence and identify responsive puncta (≥2-fold fluorescence increase). Shown are representative traces from BR33 MCs and TCs depicting the average SypHy signal (±SEM) of all NH₄Cl-responsive puncta over time. Quantifications show the rate of SV release immediately after KCl stimulation (11–26 s) and the maximum SV pool released. N = 2 genetic backgrounds, 3 independent co-cultures, and 1–3 wells per co-culture. ***p < 0.001 and *p < 0.05, mixed-effects model. (J) Mean firing rate (spikes/s) of neurons from day 11 to day 49 in MC versus TC, recorded on Axion multielectrode arrays (MEAs) (64 electrodes/well, 4 wells/condition, and 1 differentiation). BR33 glia cells were used; see Figure S3A for BR24 glia. Mixed-effects analysis with Sidak’s multiple comparisons, ****p < 0.0001, ***p < 0.001, **p < 0.01, and *p < 0.05. (K) Representative immunostaining of MEA wells at iN day 50 (after 28 days of co-culture) showing CD44 (astrocytes), IBA1 (microglia), MAP2 (neurons), and DAPI (nuclei). Scale bar, 200 μm. (L) Mean edge weight (a measure of connectivity strength) on iN day 35 comparing TC to MC. BR33 glia cells were used; see Figure S3B for BR24 glia. N = 64 electrodes/well, 4 wells/condition, and 1 differentiation. Dots correspond to each individual well. Unpaired Student’s t test, **p < 0.01. (M) Representative network graphs on iN day 35 illustrating microscale network organization in MC and TC. Node strength (circle size) represents the influence of individual electrodes, edge weight (line thickness) indicates connection strength, and node color reflects average controllability, which quantifies the ability of individual nodes to facilitate transitions between different network states.

Figure 3.. The co-culture environment influences glial responses to inflammatory stimuli

(A) Dot plot of leading-edge DEGs (upregulated in iMGs from TC versus MC related to NF-κB signaling; see Figure 2G). Dot size indicates the fraction of cells expressing each gene; color indicates adjusted p value, with all genes shown having a p-adj. < 0.05. (B) Secreted levels of TNF and IL-1β in paired MCs and TCs were measured via ELISA. Since cells in MCs and TCs were plated in parallel at equal densities, the MC values represent the total cytokine levels detected in both iA and iMG MCs. N = 2 genetic backgrounds, 3 differentiations, and 2–3 wells/differentiation. Mixed-effects model. See Figures S3D and S3E for the measurements separated by iA and iMG MCs. (C) Schematic of culture conditions and inflammatory treatments (LPS or TNF + IL-1α + C1q). Cytokines were measured in the media using the MesoScale V-Plex panel; iN MC values were not in detection range, as expected, and data are not shown. (D and E) Heatmaps of cytokine responses (fold change) in iA/iMG MCs and TCs treated with LPS for 6 h (D) or TNF + IL-1α + C1q for 24 h (E). Data are shown as Z scores across culture conditions; N = 2 genetic backgrounds, 3 differentiations, and 3 wells/differentiation. Technical replicates were averaged for each differentiation. See Table S4 for full data. Two-way ANOVA with Dunnett’s multiple comparisons test. Diff, differentiation. (F) Schematic of secretome profiling. Conditioned media from MCs and TCs were collected in parallel and analyzed by tandem mass-tag (TMT) mass spectrometry (1 genetic background, 3 independent wells per condition). See also Table S5. (G) Heatmap of selected proteins identified in secretome profiling that are differentially expressed in Alzheimer’s disease (AD) versus not cognitively impaired (NCI) cerebral spinal fluid (CSF). (H) Expression of six AD-related proteins (NEFL, STMN1, CHI3L1, TAGLN, STAB1, and MSN) that are highly expressed in TCs. One-way ANOVA with Sidak’s multiple comparisons. Data are from the secretome experiment described in (F). Data are the mean ± SEM. ****p < 0.0001, ***p < 0.001, **p < 0.01, *p < 0.05, and ns, not significant.

Figure 4.. TC induces diverse transcriptional states in microglia

(A) Representative images of microglia (green, IBA1) in MC versus TC for two genetic backgrounds (BR24 and BR33). Scale bars, 50 μm. Insets show higher-magnification views; see Figure S3F for additional images. (B) UMAP plot of microglia, subclustered and labeled by cluster identity. Microglial data are subsetted from the scRNA-seq experiment shown in Figure 2; see Figures S3G and S3H for neuron and astrocyte subclusters, respectively. (C) Heatmap of DEGs across microglial cell clusters, organized by hierarchical similarity and microglial state. See Figures S3I and S3J for neuron and astrocyte marker data, respectively. See also Table S3. (D) UMAP feature expression plots of nine selected microglial cell cluster markers: AIF1 (general microglia), LYZ (cluster 0), SLC9A9 (1), S100A4 (2), SPP1/GPNMB/APOE/TREM2 (3), and NRG3 (4). See Figure S2I for cluster 3 marker expression in all cell types. (E) TMT-MS-based secretome profiling from paired MC/TC microglia (from experiment depicted in Figure 3F). Heatmap shows secreted levels of disease-associated microglia (DAM)-related proteins (SPP1, APOE, LIPA, PSAP, CD9, and ASAH1). The full dataset is in Table S5. See Figure S4A for bar graphs of each protein. (F–I) ELISA quantification of SPP1, APOE, GPNMB, and TREM2 in media from paired MCs and TCs, normalized to TC values within each experiment (N = 2 genetic backgrounds, 4–5 differentiations, and 2–3 wells/differentiation). Data are the mean ± SEM. Statistical comparisons used a mixed-effects model; ****p < 0.0001 and **p < 0.01.

Figure 5.. DAM signature upregulation in TC is dependent on astrocyte-microglial intercellular communication

(A) Representative western blot (WB) from one genetic background (technical replicates from a single differentiation) showing TREM2, IBA1, TUJ1, and CD44 across MC and TC. (B) WB quantification of TREM2 (normalized to IBA1) in microglial MC versus TC, expressed relative to microglial MC. N = 2 genetic backgrounds, 5 differentiations, and 2–3 wells/differentiation. (C) Ratio of soluble TREM2 (ELISA) to intracellular TREM2 (WB) in the same wells, normalized to microglial MC. N = 2 genetic backgrounds, 5 differentiations, and 2–3 wells/differentiation. (D) Immunofluorescence illustrating iMG-alone cultures (IBA1⁺, INPP5D⁺), iMG-iA co-cultures (IBA1⁺, GFAP⁺), iMG-iN co-cultures (IBA1⁺, TUJ1⁺), and iMG-iN-iA TCs (IBA1⁺, TUJ1⁺, GFAP⁺). Scale bars, 100 μm. (E) Representative WB (one genetic background, technical replicates from a single differentiation) probing TREM2, IBA1, TUJ1, and CD44 in the same co-culture conditions as in (D). (F–I) Quantification of intracellular TREM2 (WB) and soluble (ELISA) TREM2, SPP1, and GPNMB, normalized to TC within each experiment. N = 2 genetic backgrounds, 3 differentiations, and 3 wells/differentiation. (J) TCs treated with vehicle or TNF + IL-1α + C1q for 24 h, stained for neurons (TUJ1, gray), astrocytes (CD44, green), and microglia (IBA1, red). Scale bars, 50 μm. (K and L) Representative WB (one genetic background, technical replicates from a single differentiation) of TREM2 and IBA1 in TCs ± TNF + IL-1α + C1q, with quantification (TREM2/IBA1) normalized to vehicle controls. N = 2 genetic backgrounds, 3 differentiations, and 3–4 wells/differentiation. (M and N) Soluble TREM2 and secreted APOE (ELISA) in iMG MCs and TCs treated with vehicle or TNF + IL-1α + C1q. N = 2 genetic backgrounds, 2–3 differentiations, and 3 wells/differentiation. See Figures S4F and S4G for SPP1 and GPNMB. Data are the mean ± SEM; statistics by mixed-effects model. ****p < 0.0001, ***p < 0.001, **p < 0.01, *p < 0.05, and ns, not significant.

Figure 6.. TC with fAD neurons reduces spine density and exacerbates inflammatory signatures in microglia

(A–C) fAD or isogenic wild-type (WT) iNs were co-cultured with WT iAs and iMGs. (B and C) Representative immunostaining of iNs (TUJ1, magenta), iMGs (IBA1, cyan), iAs (CD44, green), and nuclei (DAPI, blue). Scale bars, 50 μm. See Figures S5A–S5F for cell-type proportions and additional images across WT and fAD TCs. (D and E) Secreted Aβ42 and Aβ40 (normalized to TUJ1) in WT versus fAD TC and iN MC, relative to WT TC (N = 2 genetic backgrounds for glia, 3 differentiations, and 2–4 wells/differentiation). (F and G) Representative WB (one genetic background, technical replicates from a single differentiation) of phosphorylated tau (PTAU^202/205) and total tau in WT and fAD TCs and MCs, with PTAU:TAU ratios normalized to WT TC (N = 2 genetic backgrounds for glia, 3 differentiations, and 2–4 wells/differentiation). (H) scRNA-seq of WT and fAD iNs in TCs with WT iAs/iMGs. UMAP plots of iMGs, iAs, and iNs are colored by WT versus fAD culture identity (one genetic background). See Figures S6A and S6B for labeling by cluster index and fraction of each cluster based upon culture identity. (I) Dot plot showing significantly enriched GO pathways in iAs and iMGs comparing fAD iN TC and WT iN TC within each cell type. Input data are DEGs identified in each cell type, comparing fAD to WT iN TCs. Dots are sized based on the number of DEGs in the indicated pathway and colored based on adjusted p value. Complete dataset can be found in Table S6. (J) Dendritic spine density in WT versus fAD iN TCs labeled with DiI. Data are normalized to WT TCs (N = 3 differentiations, 2 genetic backgrounds, and 2–3 wells/differentiation), with blinded analysis. Each dot represents the average of five dendrites per well. Scale bar, 5 μm. (K) Gene-concept network of leading-edge genes in the ‘‘inflammatory response’’ pathway, upregulated in microglia co-cultured with fAD iNs (from scRNA-seq in H and I). Data are the mean ± SEM; mixed-effects analysis. ****p < 0.0001, **p < 0.01, *p < 0.05, and ns, not significant.

Figure 7.. fAD neurons suppress astrocyte-mediated induction of DAM signatures

(A and B) UMAP plots of isolated microglial cells from WT and fAD TCs, re-clustered away from neuron/astrocyte expression vectors (from the experiment in Figure 6H). See Figures S6C–S6H for neuron and astrocyte clusters. (C) Proportional composition of each microglial cell subcluster by culture condition (WT versus fAD iNs). Colors match clusters in (B). Chi-squared test, ****p < 0.0001, and ns, not significant. (D) Heatmap of DEGs across the microglial cell subclusters; see Table S6 for the complete dataset. (E) Representative western blot (one genetic background, technical replicates from a single differentiation) of TREM2, IBA1, TUJ1, and CD44 in MC, WT TCs, and fAD TCs. (F–I) Secreted TREM2 and GPNMB (ELISA) in co-culture conditions, normalized to WT within each experiment. N = 2 genetic backgrounds, 4–5 differentiations, and 2–3 wells/differentiation. See Figure S7 for APOE and SPP1 data and an additional fAD line. (J) Representative immunofluorescence of WT TCs treated with fibrillar Aβ (fAβ) and labeled for astrocytes (CD44, magenta), neurons (TUJ1, gray), microglia (IBA1, red), and fibrillar Aβ (X34, green). Insets show higher magnification of each cell type’s X34 overlap. Scale bar, 50 μm. (K–N) Secreted TREM2 and GPNMB (ELISA) in TCs treated with fAβ (K and L) or the γ-secretase inhibitor DAPT (M and N). N = 2 genetic backgrounds, 2 differentiations, and 2–3 wells/differentiation. See Figure S7 for APOE and SPP1. Data are the mean ± SEM; mixed-effects analysis unless otherwise stated (C uses chi-squared). **p < 0.01, *p < 0.05, and ns, not significant.