Human Pluripotent Stem Cell-Derived Astrocyte Functionality Compares Favorably with Primary Rat Astrocytes

Astrocytes are essential for the formation and maintenance of neural networks. However, a major technical challenge for investigating astrocyte function and disease-related pathophysiology has been the limited ability to obtain functional human astrocytes. Despite recent advances in human pluripotent stem cell (hPSC) techniques, primary rodent astrocytes remain the gold standard in coculture with human neurons. We demonstrate that a combination of leukemia inhibitory factor (LIF) and bone morphogenetic protein-4 (BMP4) directs hPSC-derived neural precursor cells to a highly pure population of astroglia in 28 d. Using single-cell RNA sequencing, we confirm the astroglial identity of these cells and highlight profound transcriptional adaptations in cocultured hPSC-derived astrocytes and neurons, consistent with their further maturation. In coculture with human neurons, multielectrode array recordings revealed robust network activity of human neurons in a coculture with hPSC-derived or rat astrocytes [3.63 ± 0.44 min−1 (hPSC-derived), 2.86 ± 0.64 min−1 (rat); p = 0.19]. In comparison, we found increased spike frequency within network bursts of human neurons cocultured with hPSC-derived astrocytes [56.31 ± 8.56 Hz (hPSC-derived), 24.77 ± 4.04 Hz (rat); p < 0.01], and whole-cell patch-clamp recordings revealed an increase of postsynaptic currents [2.76 ± 0.39 Hz (hPSC-derived), 1.07 ± 0.14 Hz (rat); p < 0.001], consistent with a corresponding increase in synapse density [14.90 ± 1.27/100 μm2 (hPSC-derived), 8.39 ± 0.63/100 μm2 (rat); p < 0.001]. Taken together, we show that hPSC-derived astrocytes compare favorably with rat astrocytes in supporting human neural network activity and maturation, providing a fully human platform for investigating astrocyte function and neuronal-glial interactions.


Introduction
Astrocytes are required for microcircuit function and no longer considered to merely provide structural support for neurons (Allen and Barres, 2009).Astrocytes provide neurons with a critical source of metabolites (Bélanger et al., 2011), regulate blood flow (Macvicar and Newman, 2015), maintain the blood-brain barrier (Abbott et al., 2006), regulate inflammation in the central nervous system (Giovannoni and Quintana, 2020), facilitate synapse formation (Allen and Eroglu, 2017), modulate neuronal network activity (Deemyad et al., 2018), and contribute to memory encoding (Kol et al., 2020;Sun et al., 2024).Astrocyte morphological complexity is one of the distinguishing features between the human and rodent brain (Oberheim et al., 2006).Moreover, increasing evidence has highlighted robust functional differences between rodent and human astrocytes.Compared with their rodent counterparts, human astrocytes exhibit distinct calcium responses (Zhang et al., 2016) and enhanced synaptogenesis (Diniz et al., 2012).
Guided differentiation of human pluripotent stem cells (hPSCs) provides the opportunity to study the development of the human brain in vitro (Astick and Vanderhaeghen, 2018).Multiple protocols have been reported for differentiation of human stem cells into neurons (Marchetto et al., 2010;Miller et al., 2013;Gunhanlar et al., 2018).A widely adopted method that yields a pure culture of excitatory neurons through forced Ngn2-overexpression requires coculture with astrocytes to ensure neuronal survival and maturation (Zhang et al., 2013).The currently available sources of astrocytes for coculture with hPSC-derived neurons include primary rodent astrocytes (the current gold standard in the field; Hulme et al., 2022;S. Wang et al., 2022;Bullmann et al., 2024) or human pluripotent stem cell-derived astrocytes (Krencik et al., 2017;Jovanovic et al., 2023).Using rodent astrocytes can be problematic for many experimental designs, due to the genomic and functional differences between human and rodent cells.Two recent studies (Garcia et al., 2019;Hedegaard et al., 2020) have demonstrated that supplementing additional iPSC-derived astrocytes to a neural culture containing neurons, astrocytes, and NPCs improves the electrophysiological properties of neurons.Other studies describe increased synaptogenesis or enhanced neuronal electrophysiological maturation; however, this is often compared with a neuronal culture without astrocytes (Krencik et al., 2017;Jovanovic et al., 2023) or they were unable to fully recapitulate neuronal electrophysiological properties in a coculture with hPSC-derived astrocytes compared with primary murine astrocytes (Shih et al., 2021).Generally speaking, readouts to validate hPSC-derived astrocyte protocols are focused on astrocyte characterization by comparing their transcriptomic profile to primary human or rodent astrocytes (Caiazzo et al., 2015;Kondo et al., 2016;Sloan et al., 2017;Tcw et al., 2017;di Domenico et al., 2019).
Here, we present a functional comparison of human and rodent astrocytes in coculture with human neurons using a modified protocol to differentiate hPSC-derived neural progenitor cells (NPCs) into functional cortical astrocytes.We have systematically compared the electrophysiological properties of hPSC-derived neurons in coculture with primary rat astrocytes or hPSC-derived astrocytes and show that hPSC-derived astrocytes are able to support a higher level of neuronal activity compared with primary rat astrocytes.Using immunocytochemistry, (single-cell) RNA sequencing, and flow cytometry, we demonstrate that hPSC-derived astroglia express the canonical astrocytic markers and at similar levels when compared with primary rat astrocytes.Morphological analysis shows that hPSC-derived astroglia display hominid morphological features.Both in a mono-and in a coculture with human neurons, hPSC-derived astrocytes are larger compared with rat astrocytes.Following xenotransplantation and integration into the local microenvironment of a mouse brain, hPSC-derived astrocytes maintain this hominid morphological characteristic and are larger when compared with resident mouse astrocytes.We show that synapse development and spontaneous excitatory postsynaptic currents are increased in cocultures of human neurons and hPSC-derived astrocytes compared with cocultures of human neurons with rat astrocytes.Taken together, our data demonstrate that hPSC-derived astrocytes promote neuronal maturation and synaptic function, thus eliminating the need for their rodent counterparts in human neural coculture systems.

Materials and Methods
Resource availability.Further information and requests for resources and reagents should be directed to the corresponding authors, Steven A. Kushner (sk2602@cumc.columbia.edu)or Femke M.S. de Vrij (f.devrij@erasmusmc.nl).
All mouse experiments were approved by the local animal welfare committee, and mice were kept under standard housing conditions with ad libitum access to food and water.Both male and female immunodeficient Rag2 −/− BALB/c (Jackson, 014593) mice were used for transplantation studies and killed between 4 and 40 weeks of age.
All procedures with human tissue were performed with the approval of the Medical Ethical Committee of the Erasmus MC Rotterdam, including written informed consent of all subjects for brain donation in accordance with Dutch license procedures and the Declaration of Helsinki.Fresh-frozen postmortem tissue blocks containing the middle frontal gyrus (BA9) from three donors [61 (female), 79 (male), and 81 (female) years old] were obtained from the Erasmus MC Department of Pathology.
Experimental design and statistical analysis.For each experiment, human astroglia were derived from cryopreserved NPCs that had previously been established from four different hPSC lines and specific lines, and total n is indicated in the figure legends.Cultures containing hPSC-derived or rat astrocytes were grown in parallel to control for batch effects.Bulk RNA sequencing data was analyzed using Fisher's exact test and FDR corrected for multiple testing.Single-cell RNA sequencing data was analyzed using Seurat's implementation of DESeq2.The parameters logfc.thresholdand min.pctwere set to 0. p values adjusted for multiple testing error were used for thresholding significance at p < 0.05.For functional studies, statistical comparisons were performed using Fisher's exact test, two-tailed t test, or analysis of variance (ANOVA), as indicated.Data are expressed as mean ± SEM, unless otherwise specified.The threshold for significance was set at p < 0.05 for all statistical comparisons.Astroglia differentiation.We validated our protocol using four independent human pluripotent stem cell (hPSC) lines: induced pluripotent stem cell lines [WTC-11 provided by Bruce R. Conklin (The Gladstone Institute and UCSF ; Miyaoka et al., 2014), RRID:CVCL_Y803 (iPS1), two in-house previously established control lines (de Vrij et al., 2019;Erasmus MC iPS Core Facility EMC13i955-3, male, age 57, iPS2;EMC14i96-6, female, age 54, iPS3)] and an embryonic stem cell (ES) line [SA001 (Adewumi et al., 2007), RRID:CVCL_B347, male].All hPSC lines and their derivatives underwent regular microarray-based screening for structural genomic variation and were screened for mycoplasma every other month.Pluripotent stem cells were differentiated to NPCs in 40 d and cryopreserved as previously described (Gunhanlar et al., 2018) with slight modifications (Extended Data Fig. 1-1).All cells were maintained in an incubator at 37°C/5% CO 2 .Human PSCs were expanded in hES medium (Table 1) on a feeder layer of mouse embryonic fibroblasts.A 60-70% confluent 6-well plate of undifferentiated hPSC colonies was lifted from the feeder layer using collagenase (Thermo Fisher Scientific, 17104019).Colonies were transferred to a 10 cm dish containing hES medium without fibroblast growth factor on a shaker (+/− 50 rpm).On Day 3, the medium was changed to neural induction medium (Table 1) and refreshed every other day.After 7 d in suspension, EBs were collected and seeded on laminin-coated dishes [20 μg/ml (Sigma, L2020)].On Day 15, cells were switched to NPC medium (Table 1).Cells were passaged 1:4 every week thereafter using collagenase and a cell lifter.After Passage 3, NPC cultures were purified using fluorescence-activated cell sorting (FACS; Yuan et al., 2011;Bauer et al., 2021).NPCs were detached from the culture plate using Accutase (StemCell Technologies, 07920) and resuspended into single cell solution.CD184 + /CD44 − /CD271 − /CD24 + cells were collected using a FACSaria III (BD Biosciences) and expanded.To obtain hPSC-derived astroglia, NPCs (Passages 5-10) were passaged 1:4 using Accutase when confluent and subsequently grown in Astrocyte medium containing leukemia inhibitory factor (LIF; 10 ng/ml) and bone morphogenetic protein 4 (BMP4; 10 ng/ml) for 4 weeks (Table 1, Fig. 1A).Cells were grown on laminin-coated dishes [20 μg/ml (Sigma, L2020)] and passaged 1:4 when confluent.The passaging ratio was adapted to the proliferation rate that gradually slows down at later stages of differentiation.Following the 4 week differentiation period, astroglia could be maintained for at least an additional 6 weeks.
Flow cytometry quantification.Cells were detached from the culture dish using Accutase (StemCell Technologies, 07920), washed, spun down, resuspended in PBS/2% FBS, and stained with a primary antibody (Table 2) on ice for 30 min.Next, cells were washed and stained using a secondary antibody (Jackson ImmunoResearch, 1:400), kept on ice for 30 min, and washed two more times.Samples were analyzed on an LSRFortessa (BD Biosciences).Secondary antibody alone was used as an isotype control.
Bulk RNA sequencing.Total RNA was isolated from NPCs and their derived astroglia (iPS1, iPS2, iPS3, and ES cell lines, n = 3 per cell type) using an RNeasy mini kit (Qiagen, 74104).RNA samples were prepped using TruSeq Stranded mRNA Library kit (Illumina, 20020594).The resulting DNA libraries were sequenced according to the Illumina TruSeq Rapid v2 protocol on an Illumina HiSeq2500 sequencer.A total of 50 bp reads were generated, trimmed, and mapped against GRCh38 using HiSat2 (version 2.1.0).Gene expression values were called using htseq-count (version 0.9.1).Sequencing resulted in at least 21.1 M reads per sample, with at least 16.7 M counts in the expression profile and 22.7-25.2thousand expressed genes.Analysis was performed using a custom R script.
Culture dissociation and single-cell RNA sequencing.Single-cell RNA sequencing experiments were performed using hPSC-derived astroglia and Ngn2-induced neurons derived from the iPS1 line.For each sample, four cultures were pooled and dissociated using the Papain Dissociation System (Worthington Biochemicals, LK003150) according to manufacturer's instructions.Single-cell solutions were run on a Chromium Controller, and final libraries were generated with the Chromium Next GEM Single cell 3' reagents kit v3.1 (dual index; PN-1000268, PN-1000120 PN-1000242, 10x Genomics) according to the manufacturer's protocol.Libraries were sequenced on an Illumina Novaseq6000 system (28-10-10-90 cycles) with a target of 25,000 reads/cell.
Astroglia engraftment in neonatal Rag2 −/− mice.Human iPSC-derived astroglia were xenotransplanted into immunodeficient neonatal (P1) Rag2 −/− BALB/c mice under cryoanesthesia.Roughly 5-10 × 10 4 astroglia were delivered in a 1 μl of PBS-drop via a 1-mm-diameter pulled glass pipette into five different sites: posterior and anterior anlagen of the corpus callosum bilaterally and in the cerebellar peduncle dorsally (Windrem et al., 2008).Mice were killed between 4 and 40 weeks of age by transcardiac perfusion with saline, followed by 4% PFA.Brains were removed; left in 4% PFA for 2 h at room temperature; transferred to a 10% sucrose/phosphate buffer (PB 0.1 M), pH 7.3; and stored overnight at 4°C.Brains were embedded in 12% gelatin/10% sucrose blocks.Fixation was performed for 2 h at room temperature in a 10% PFA/ 30% sucrose solution.Embedded brains were stored at 4°C before being sectioned into 40 μm slices on a freezing microtome (Leica; SM2000R).Brain sections were preincubated with a blocking buffer [0.5% Triton X-100 (Sigma, T8787) and 10% normal horse serum (NHS; Thermo Fisher, 16050122) in PBS] for 1 h at room temperature.Primary antibody incubation was done for 48 h at 4°C.Secondary antibody incubation was performed for 2 h at room temperature.Both primary and secondary antibody incubations were performed in a staining buffer (2% NHS and 0.5% Triton X-100 in PBS).Samples were mounted using Mowiol 4-88 (Sigma-Aldrich, 81381) and imaged using a Zeiss LSM 800 confocal microscope.
Human brain immunocytochemistry.All procedures with human tissue were performed with the approval of the Medical Ethical Committee of the Erasmus MC Rotterdam, including written informed consent of all subjects for brain donation in accordance with Dutch license procedures and the Declaration of Helsinki.Fresh-frozen postmortem tissue blocks containing the middle frontal gyrus (BA9) from three donors [61 (female), 79 (male), and 81 (female) years old] were obtained from the Erasmus MC Department of Pathology.Donors were confirmed to have no past medical history of any known psychiatric or neurologic illness, with additional confirmation of the absence of clinical neuropathology by autopsy examination (Amin et al., 2018).Tissue blocks were postfixed for 7 d in 4% paraformaldehyde (0.1 M phosphate buffer), pH 7.3, at 4°C.Tissue was subsequently transferred to 10% sucrose (0.1 M phosphate buffer), pH 7.3, and stored overnight at 4°C.Embedding was performed in 12% gelatin/10% sucrose, with fixation in 10% paraformaldehyde/30% sucrose solution for 4 h at room temperature and overnight immersion in 30% sucrose at 4°C.Serial 40 μm sections were collected along the rostro-caudal axis using a freezing microtome (Leica; SM2000R) and stored at −20°C in a solution containing 37.5% ethylene glycol (Avantor, 9300), 37.5% glycerol (VWR Chemicals, 24 386.298) and 25% 0.1 M phosphate buffer.Free-floating sections were washed thoroughly with PBS before being incubated in sodium citrate (10 mM) at 80°C for 45 min and rinsed with PBS.Sections were preincubated with a blocking PBS buffer containing 1% Triton X-100 and 5% bovine serum albumin for 1 h at room temperature.Primary antibody labeling was performed in PBS buffer containing 1% Triton X-100 and 1% BSA for 72 h at 4°C.Following primary antibody labeling, sections were washed with PBS and then incubated with corresponding Alexa-conjugated secondary antibodies and cyanine dyes (1:400, Braunschweig Chemicals) in PBS buffer containing 1% Triton X-100 and 1% BSA for 4 h at room temperature.Nuclear staining was performed using DAPI (1:10,000, Thermo Fisher Scientific).Images were acquired using a Zeiss LSM 800 confocal microscope.
Astrocyte size quantifications.Maximum projection images of 40 μm brain slices or in vitro cultures were used to determine cell size by drawing regions of interest (ROIs) around typical protoplasmic astrocytes and calculating maximum diameter with the Fiji module of NIH ImageJ.Following xenotransplantation into the murine brain, astrocytes were identified from tissue sections using a combination of antibodies (Table 2): (human) GFAP, STEM121, and human nuclear antigen (hNA).Neighboring mouse astrocytes and xenotransplanted human iPSC-derived astrocytes were analyzed in mice of different ages (4-40 weeks).Human astrocyte size was also quantified in the middle frontal gyrus of postmortem tissue from all three donors [age 61 (n = 9), 79 (n = 8), and 81 (n = 11)].In vitro astrocyte size was quantified in cultures containing either primary rat astrocytes or human iPSC-derived astrocytes, from either pure mono-cultures or in coculture with Ngn2-induced neurons, as indicated.Astrocyte surface area was determined by thresholding for GFAP using the Fiji module of NIH ImageJ and calculating the percentage GFAP-positive cell surface of the total surface area.
Coculture with Ngn2 neurons in independent laboratory.On Day 0, hiPSCs containing an integrated Ngn2 cassette under a TET-controlled promotor were passaged on a 1:100 Matrigel coating (Sigma-Aldrich, CLS356231) and grown in E8 medium (A1517001, Thermo Fisher Scientific) containing 4 µg/ml doxycycline (D5207, Sigma-Aldrich) and 10 µM ROCK inhibitor (Y0503, Sigma-Aldrich).The following day medium was switched to Ngn2 Day 1 medium (Table 1).To support neuronal maturation, hPSC-derived astrocytes or freshly prepared rat astrocytes were added to the culture on Day 2 in a 1:1 ratio.On Day 3, the medium was changed to Ngn2 medium (Table 1) and cytosine β-D-arabinofuranoside (Ara-C; 2 µM; Sigma-Aldrich, C1768) was added once to remove proliferating cells from the culture, facilitating long-term recordings of the cultures.From Day 5 onward, half of the medium was refreshed three times per week.Culture medium was additionally supplemented with 2.5% FBS (Sigma-Aldrich, F2442) to support astrocyte viability from Day 10 onward.
Cocultures were kept at 37°C/5% CO 2 throughout the entire differentiation process.
Microelectrode array recordings.Spontaneous electrophysiological activity of iCell GlutaNeurons with or without astrocytes [hPSC-derived or primary rat (ScienCell, SCCR1800)] was recorded using a multiwell MEA system (Multi Channel Systems).Plates were kept at 37°C and maintained at 5% CO 2 .Plates were equilibrated to the chamber for 10 min and recorded for an additional 10 min.The signal was sampled at 10 kHz, filtered with a high-pass Butterworth filter with 100 Hz cutoff, and a low-pass fourth-order Butterworth filter with 3,500 Hz cutoff.The noise threshold was set at +/−4.5 standard deviations.Recordings were analyzed off-line using Multiwell-Analyzer (Multi Channel Systems) and SPYCODE (Bologna et al., 2010).
Electrophysiological recordings.iCell GlutaNeurons (Cellular Dynamics, R1034) were grown in a coculture with primary rat (ScienCell, SCCR1800) or hPSC-derived astrocytes for 1-2 weeks, after which culture slides were transferred to the recording chamber and whole-cell patch-clamp recordings were performed as previously described (Gunhanlar et al., 2018).Briefly, cultures were equilibrated to artificial cerebrospinal fluid (ACSF).In the recording chamber, slides were continuously perfused with ACSF at 1.5-2 ml/min, saturated with 95% O 2 /5% CO 2 and maintained at 20-22°C.Recordings were performed with borosilicate glass recording micropipettes (3-6 MΩ).Data were acquired at 10 kHz using an Axon MultiClamp 700B amplifier (Molecular Devices), filtered at 3 kHz, and analyzed using pClamp 10.1 (Molecular Devices).Current-clamp recordings were performed at a holding potential of −70 mV.Intrinsic membrane properties were analyzed using a series of hyperpolarizing and depolarizing square wave currents (500 ms duration, 1 s interstimulus interval) in 5 pA steps, ranging from −30 to +30 pA.Data analysis was performed using a custom-designed script in Igor Pro-8.0 (WaveMetrics).Input resistance was calculated from the first two hyperpolarizing steps.Active properties were extracted from the first depolarizing step resulting in AP firing.AP threshold was defined by the moment at which the second derivative of the voltage exceeded the baseline.AP amplitude was measured from threshold.Neurons were categorized as "firing" if they were capable of firing three or more mature APs without significant accommodation during a depolarizing current step.Voltage-clamp recordings were performed at a holding potential of −80 mV.Synaptic events were detected using Mini Analysis Program (Synaptosoft).Bursting activity was defined as a period longer than 5 s with a frequency of >20 Hz, followed by a return to baseline.
To further characterize the hPSC-derived astroglia, and compare their expression profile to the parental NPCs, we performed bulk RNA sequencing of three independent differentiation batches of each of the four hPSC lines.Hierarchical clustering based on Euclidean distance using variance stabilizing transformed (VST) counts for established cell typespecific markers confirmed the distinction between NPC and astroglia samples and batch-to-batch reproducibility (Fig. 1F).PCA reveals that PCA 1 separates the samples on cell type (Extended Data Fig. 1-3).Canonical astrocyte genes, e.g., GFAP, AQP4, and S100B, were robustly upregulated in astroglia samples compared with NPC samples, while neuronal genes, e.g., MAP2, TUBB3, and DCX, were downregulated.
Flow cytometry based on GFAP, CD44, and SOX9 immunolabeling exhibited a similar geometric mean fluorescence intensity and cellular purity between hPSC-derived and primary rat astrocytes (Fig. 1G-I, Extended Data Fig. 1-4).These results together demonstrate that 4 week differentiation of NPCs in medium containing LIF and BMP4 is sufficient to obtain a relatively pure and homogenous population of hPSC-derived astroglia.Single-cell RNA sequencing confirms astroglial identity and reveals further specification of astrocytes in neuronal coculture During brain development, astrocyte maturation is associated with profound transcriptional changes and mutually coregulated with neuronal maturation (Zhang et al., 2016;Shan et al., 2021).In order to gain insight into the transcriptional profile and cell type identity of hPSC-derived astroglia and the transcriptomic impact of neuronal coculture on these cells, we performed single-cell RNA sequencing (scRNA seq) in three independent conditions: (1) astroglia mono-culture, (2) neuronal mono-culture, and (3) coculture of astrocytes and neurons.As expected, the neuron and astroglia samples showed robust expression of their corresponding cell type-specific markers [e.g., VIM and FABP7 (astroglia); MAP2 and NEUROG2 (neurons); Extended Data Fig. 2-1A,B].
For unbiased assignment of cell type identity to clusters in our samples, we calculated Pearson's correlation coefficient between the individual clusters in our samples and the cell type-annotated clusters at different developmental stages (gestational weeks 14-25) from a recently published scRNAseq dataset of the developing human brain (Bhaduri et al., 2021).Using this dataset, we transferred cell type and developmental age labels to clusters in our scRNA sequencing samples.The transcriptional profile of both astroglia and neuronal mono-cultures most closely resembled the cerebral cortex during gestational week 18 (Extended Data Fig. 2-1C,D).Mono-culture hPSC-derived astroglia included "radial glia" (38%), "dividing" (52%), and "endothelial" (10%) cells (Extended Data Fig. 2-1C,E).Mono-culture neurons were most strongly correlated with excitatory neurons (90%), while the remaining 10% of cells had transcriptional profiles of dividing or intermediate progenitor cells (IPCs; Extended Data Fig. 2-1D,F).In the astrocyte-neuronal cocultures, we observed 60% "excitatory neurons," 20% "radial glia," 13% "astrocytes," and 12% "dividing" cells (Fig. 2A,B).Interestingly, the cocultures yielded neurons with transcriptional profiles most strongly associated with gestational week 19-slightly older than neurons from the mono-culture condition-whereas mono-and coculture astrocytes are both most strongly associated with gestational week 18 (Fig. 2A, Extended Data Fig. 2-1C,D).Neurons most closely resembled "excitatory neurons," both in mono-and coculture, consistent with their glutamatergic identity.In contrast, hPSC-derived astroglia underwent a profound transcriptional adaptation in neuronal coculture.Compared with astrocyte mono-culture, we observed fewer "dividing" cells, more "radial glia," and a cluster observed exclusively in coculture that was most strongly correlated with "astrocytes."Notably, in the coculture sample, cluster 10 ("radial glia") showed a strong increase in APOE expression (log2fold = 4.8; p < 0.01) compared with all other clusters (Extended Data Fig. 2-1H), while in the astrocyte mono-culture we did not observe a defined cluster with APOE expression (Extended Data Fig. 2-1I).
In order to further characterize the maturation and transcriptional changes of hPSC-derived astrocytes during neuronal coculture, we next sought to identify genes that were differentially expressed in astrocytes from their corresponding mono-culture and coculture conditions.We selected clusters in which >20% of cells expressed VIM, S100B, and SOX9 to create an integrated sample with mono-and cocultured astrocytes.Similarly, we used RBFOX3, MAP2, and NEUROG2 as markers to select neuronal clusters from mono-culture and coculture conditions to generate a corresponding integrated neuron sample (Fig. 2C).By retaining the original sample identity of individual cells in these integrated samples, we were able to directly compare their gene expression profiles (Fig. 2E, Extended Data Fig. 2-1G).
In the integrated astrocyte sample, some clusters showed a skewed distribution in the original sample identity.For example, clusters labeled "dividing" contained cells mainly from the mono-culture (mono-culture, 75.8%; coculture, 24.2%), and in the cluster labeled "astrocyte" most cells originated from the coculture sample (mono-culture, 16.8%; coculture, 37.2%; Fig. 2D,E; Table 3).No such skewing of clusters was observed in the integrated neuron sample (Extended Data Fig. 2-1G, Table 4).
Next, we performed differential expression analysis on the integrated samples to gain insight into the adaptations in astrocytes and neurons during coculture.Figure 2F displays a volcano plot of the differentially expressed genes (DEGs) of the integrated astrocyte sample with the most highly regulated genes shown in Table 5.As expected, many wellestablished markers of mature astrocytes were highly upregulated in hPSC-derived astrocytes from the coculture sample, e.g., S100B and SLC1A3 (Fig. 2E,H).Moreover, multiple genes were found that are known to be specifically upregulated during astrocyte-neuron interactions and coordinated maturation, e.g., SPARCL1 (Singh et al., 2022), METRN (Nishino et al., 2004;Jørgensen et al., 2009), and CROC4 (Jeffrey et al., 2000), as well as known disease-associated genes such as CRYAB (Guo et al., 2019) and APOE (Lanfranco et al., 2021; Fig. 2F).DEG analysis in the integrated neuron sample revealed few differentially expressed genes with a log2fold > 1, demonstrating the efficacy of Ngn2 overexpression to induce neuronal differentiation (Table 6).
Cross-referencing the DEGs from the integrated astrocyte sample with those of a study (Zhang et al., 2016) that investigated transcriptional differences between primary human astrocyte progenitors and mature astrocytes revealed a high correspondence (Fisher's exact; p < 0.001; Table 7) between datasets, suggesting that astrocytes undergo additional maturation when grown with neurons.Pseudotime analysis of the integrated astrocyte sample confirmed this observation, in which "dividing" cells from the mono-culture sample are positioned at the beginning of the developmental trajectory and cocultured astrocytes enriched at the end (Fig. 2D,E,G).
In order to gain insight into the biological processes associated with DEGs in mono-or cocultured astrocytes, we performed a gene ontology (GO) pathway analysis based on genes that showed a log2fold change >1 or smaller than −1 when comparing astrocytes from the mono-to the coculture condition.GO term analysis revealed distinct transcriptional  iPS1) and primary fetal brain tissue (Bhaduri et al., 2021).See Extended Data Figure 2-2 for heatmaps and corresponding data for astroglial and neuronal mono-cultures.B, UMAP projection of coculture sample with transferred cell type labels with the highest correlation from primary brain tissue.C, Expression of genes used for selection of astrocytes (top) and neurons (bottom) from samples in the coculture sample.D, UMAP projection of integrated astrocyte sample with cell type label transfer.E, UMAP projection of integrated astrocyte sample with original sample identity indicated in green (coculture) or red (mono-culture).F, Volcano plot of significantly up-(red) and down-regulated profiles active in mono-(Fig.2I) or cocultured astrocytes (Fig. 2J), suggesting a switch in the primary cellular functioning of astrocytes depending on culture conditions.The top 25 (p value) GO terms based on the DEGs in the mono-culture astrocytes were associated with angiogenesis, migration and cell motility, cytoskeleton organization, and cellular component biogenesis.After coculture, the top 25 (p value) GO terms based on DEGs in astrocytes were related to neuronal development, anatomical and cellular development, and antigen presentation.Based on our GO term analysis, the transcriptional profile of astrocytes in a pure mono-culture suggests a prominent role in angiogenesis and migration, whereas the transcriptional profile of astrocytes in a coculture with neurons is associated with neuronal development and maturation and immune response.

hPSC-derived astrocytes retain hominid morphological characteristics in vitro and following xenotransplantation into the murine brain
Pluripotent stem cells have provided a unique opportunity to study the development and physiology of human brain cell lineages.This is especially notable for astrocytes, the most highly dimorphic cell type between higher-order primates and other mammals (Oberheim et al., 2006).Compared with their rodent counterpart, human astrocytes are larger and have a more complex morphology.We sought to investigate whether our hPSC-derived astrocytes preserve this feature.In vitro, hPSC-derived astroglia are larger than primary rat astrocytes in both a pure astrocyte culture (hPSC, 119.60 μm ± 4.74; rat, 87.25 μm ± 5.37; two-tailed t test; p < 0.001; Fig. 3A-C) and when grown in a coculture with neurons (hPSC, 208.91 μm ± 12.25; rat, 129.54 μm ± 7.69; two-tailed t test; p < 0.001; Fig. 3D-F).In correspondence with this, the percentage of astrocytic GFAP-positive surface area is also increased in fully human cocultures when seeding identical cell numbers (hPSC, 23.71% ± 3.07; rat, 14.59 ± 2.44; two-tailed t test; p < 0.05; Extended Data Fig. 3-1).Interestingly, hPSC-derived astrocytes and primary rat astrocytes both adopt a larger protoplasmic morphology when grown in coculture with neurons compared with astrocytes from a mono-culture (Fig. 3E,F).
(blue) genes in coculture astrocytes versus mono-culture astrocytes.G, Pseudotime trajectory of the integrated astrocyte sample suggests further maturation of astrocytes in the coculture sample, pseudotime indicated from blue (early) to yellow (late).H, Expression of mature astrocyte markers, S100B and SLC1A3, is skewed toward cells from the coculture sample.I, J, Top 25 enriched GO-terms visualized in a treeplot based on DEGs in mono-(I) or coculture (J) astrocytes.Cell size quantification in an in vitro environment can be limited by the 2D environment in which cells are often maintained.It has previously been established that upon xenotransplantation, human astrocytes are larger compared with neighboring murine host astrocytes (Preman et al., 2021;Voronkov et al., 2022;Baranes et al., 2023).In order to  investigate whether we could replicate this, hPSC-derived astroglia obtained using our protocol were xenotransplanted into the brains of neonatal immunodeficient Rag2 −/− mice.One week after xenotransplantation, cells were mainly found near the subventricular zone (SVZ) of the lateral ventricles (Extended Data Fig. 3-2A).By 4 weeks after xenotransplantation, cells had migrated away via the rostral migratory stream from the SVZ and could be found in the olfactory bulb (Extended Data Fig. 3-2B).Eight weeks after xenotransplantation, human cells were more globally distributed throughout the brain (Extended Data Fig. 3-2C).Eight months after xenotransplantation, cells remained present in the midbrain, olfactory bulb, hippocampus, and cortex (Extended Data Fig. 3-1D).Following xenotransplantation hPSC-derived astrocytes evenly spread out through the host brain, reminiscent of astrocytic domains typically observed for endogenous astrocytes in vivo (Oberheim et al., 2006;Extended Data Fig. 3-1C).Moreover, we observed that xenotransplanted hPSC-derived astrocytes maintained their increased hominid cellular diameter (Oberheim et al., 2006) in the murine brain.Xenotransplanted hPSC-derived astrocytes had a larger maximum diameter (137.70 μm ± 10.3) compared with their murine counterparts (29.28 μm ± 2.26; one-way ANOVA; p < 0.001) and highly similar to those in the human postmortem brain (137.66μm ± 5.48; one-way ANOVA, p = 0.99; Extended Data Fig. 3-2E,F).
hPSC-derived astrocytes promote the formation of functional synapses Astrocytes are essential for proper neuronal network maturation and actively involved in the formation of synapses and establishment of network activity (Mederos et al., 2018).In order to compare species-specific astrocyte support of neuronal network maturation, we quantified synapse formation and performed MEA and whole-cell patch-clamp recordings from mono-cultured human neurons, neurons cocultured with primary rat astrocytes, or cocultured with hPSC-derived astrocytes (Figs. 4,5).

Discussion
Astrocytes are essential for neuronal network development, survival, and electrophysiological maturation (Pyka et al., 2011;Clarke and Barres, 2013;Tang et al., 2013).Compared with their rodent counterpart, human astrocytes are larger and have a more complex morphology.A specific subtype of astrocyte, interlaminar astrocytes, is found exclusively in higher-order primates (Colombo et al., 1998).Primary human astrocytes have previously been shown to have functional differences with rodent astrocytes, as well (Diniz et al., 2012;Zhang et al., 2016).Accordingly, there is a widely acknowledged need for protocols to obtain functional astrocytes from human pluripotent stem cells, demonstrated by the many protocols that currently exist (Caiazzo et al., 2015;Kondo et al., 2016;Krencik et al., 2017;Sloan et al., 2017;Tcw et al., 2017;di Domenico et al., 2019;Garcia et al., 2019;Hedegaard et al., 2020;Shih et al., 2021;Jovanovic et al., 2023).By adopting existing protocols and supplementing culture media with BMP4 and LIF, we were able to efficiently establish a pure culture of functional hPSC-derived astroglia from an intermediate cryopreserved NPC stage in 28 d.By making use of an intermediate stage of NPCs that can be expanded and survives cryopreservation, human astrocyte cultures can be rapidly established while maintaining the genomic integrity of the parental hPSC line.Immunocytochemistry and RNA sequencing reveals a relatively homogenous population of cells with widespread expression of canonical astrocyte markers such as GFAP, S100B, SOX9, and CD44 (Fig. 1).Single-cell RNA sequencing shows that the transcriptional profile of these cells is most similar to primary human "radial glia," "astrocytes," and a group of "dividing" cells (Fig. 2A,B, Extended Data Fig. 2-1C).
BMP4 and LIF have previously been demonstrated to promote differentiation of neural stem cells to astrocytes (Mabie et al., 1997;Koblar et al., 1998;Bonaguidi et al., 2005).Canonically, growth factors from the BMP family signal via SMAD-dependent pathways (R. N. Wang et al., 2014), whereas LIF activates the JAK/STAT pathway (Bonni et al., 1997).During late embryogenesis, the binding of the STAT3-SMAD1 complex to astrocytic promotors induces further maturation of glial progenitors toward astrocytes (Nakashima et al., 1999).In vitro, treatment of murine embryonic subventricular zone progenitor cells with LIF generates proliferating GFAP + astrocyte progenitor cells, after which subsequent BMP4 exposure further differentiates these cells to mature astrocytes (Bonaguidi et al., 2005).Our approach for deriving astrocytes from human pluripotent stem cells (hPSCs) leverages this signalling mechanism by first establishing hPSC-derived NPCs and then subsequently inducing astrocyte differentiation through addition of BMP4 and LIF, thereby resulting in a pure culture of proliferating hPSC-derived astroglia.
A widely adopted method that rapidly yields a pure population of hPSC-derived neurons through forcible overexpression of Ngn2 (Zhang et al., 2013) was developed using supplementation with primary rodent astrocytes (Frega et al., 2015).Here, we demonstrate the ability to establish a fully human PSC-derived neural coculture system (Figs. 4,5).This provides the opportunity to precisely control the cellular composition, making it possible to study the effects of cell type-specific genotypes or targeted genetic manipulations of cocultured astrocytes and/or neurons.In addition, we show that hPSC-derived astrocytes can be efficiently integrated into the existing workflow of an independent laboratory (Extended Data Fig. 5-1), emphasizing the robustness of the method.
Another advantage of the method is the relative ease of producing hPSC-derived NPCs and astroglia compared with using rodents for obtaining primary astrocytes and thereby also contributes to reducing the use of laboratory animals.Primary murine astrocyte cultures can be established within 30 d following dissociating the brain of a P0 pup (Güler et al., 2021), provided one has access to an active laboratory animal facility.Obtaining hPSC-derived NPCs can take up to 60 d depending on the protocol used; we utilize an embryoid body approach (Gunhanlar et al., 2018), while a different laboratory has used our protocol to successfully establish hPSC-derived astroglia from NPCs differentiated using a dual-SMAD inhibition approach (Gordillo-Sampedro et al., 2024).Once hPSC-derived NPCs have been established, these can be expanded and cryopreserved.By making use of an intermediate stage of NPCs, human astrocyte cultures can be rapidly established while maintaining the genomic integrity of the parental hPSC line.From the NPC stage, obtaining hPSC-derived astroglia requires a similar time investment as primary murine astrocytes, without the need for an active animal colony.If a laboratory is equipped to perform in vitro experiments, no specialized equipment is required to integrate our protocol into their workflow.
Through scRNA sequencing, we gained insight into the transcriptional profile of different populations of astroglia and observed profound changes in their profile when cocultured with neurons.Even though we detected a relatively uniform expression of GFAP, S100B, and SOX9 in pure hPSC-derived astroglia, cross-referencing our scRNAseq data to a human fetal brain database (Bhaduri et al., 2021) revealed distinct astroglial subtypes within this in vitro astrocyte population, especially when cocultured with neurons (Fig. 2A, Extended Data Fig. 2-1C).This finding highlights the importance of using an unbiased approach when assigning cell type identity to scRNA sequencing data, as relying on a small set of genes to verify cell type identity would have masked the cellular diversity in our hPSC-astroglia populations.
Human PSC-derived astroglia seem to undergo additional developmental specification when cocultured with neurons.We demonstrate that this adaptation changes the transcriptional profile of astroglia and their associated cellular subtype diversity (Fig. 2D).In vivo, astrocytes are traditionally divided into subtypes based on their morphology (Oberheim et al., 2012).Recent attempts have been made using scRNA sequencing techniques to further specify astrocyte subtypes (Batiuk et al., 2020;Qian et al., 2023).Here we show that it is also possible to make a distinction in vitro based on their transcriptomic profile, e.g., cluster 10 (radial glia 2, Fig. 2B) displays a distinct change in APOE expression (Extended Data Fig. 2-1H,I).The identification of such a specific cluster of cells provides future opportunities for studies focused on the etiology of Alzheimer's disease and highlights the importance of establishing adequate culture conditions when using hPSC-derived cultures to study genes of interest.
DEG analysis in the integrated neuron sample revealed few differentially expressed genes with a log2fold > 1 (Table 6).It has previously been established that continuous Ngn2 overexpression is sufficient to induce transcriptionally mature neurons (Lin et al., 2021).Interestingly, we did observe significant upregulation in cocultured neurons of FOS (log2fold = 1.30; p < 0.01), an immediate early gene regulated by neuronal activity (Curran and Morgan, 1995).The increased synaptic maturation we observed in our cocultures, despite the lack of increased synaptic gene expression in the DEG analysis, could originate from the local and transient transcriptomic changes required for synaptic plasticity that might not be reflected in a culture that has reached a homeostatic equilibrium.Furthermore, the procedure of dissociation of the neural cultures for scRNA will likely result in a negative selection against local synaptic transcripts.
We demonstrate that hPSC-derived astroglia maintain morphological hominid characteristics (Oberheim et al., 2006), both in vitro and in vivo.When comparing the morphology of primary rat and hPSC-derived astroglia in vitro, we found that hPSC-astroglia are larger (Fig. 3A), in line with a previous report (Zhang et al., 2016) comparing primary cultures of human and rat astrocytes.Interestingly we also observe a change of in vitro morphology upon coculture, in which astrocytes are larger and more complex when grown in coculture with neurons (Fig. 3D).When xenotransplanted into the murine brain, this species difference persists.Human PSC-derived astrocytes are larger compared with neighboring mouse astrocytes (Extended Data Fig. 3-1F).These findings confirm the bidirectional interaction between neurons and astrocytes.This is in line with a previous report that tissue microenvironment is a critical driver of astrocyte diversity in vivo (Endo et al., 2022).
Moreover, we demonstrate that hPSC-derived astrocytes are able to more efficiently support the development of synapses compared with primary rat astrocytes (Fig. 4) and the maturation of neural network activity (Fig. 5).Our findings show that hiPSC-derived neurons are more active and receive more synaptic input in a coculture with hPSC-derived versus rodent astrocytes.Synapse formation is accelerated in a fully human coculture system, resulting in an increased detection of sEPSCs through whole-cell electrophysiology (Fig. 4).Using MEA recordings, we observed a higher firing frequency in cocultures with hPSC-derived astrocytes compared with primary rat astrocytes, both for individual events and within (network) bursts (Fig. 5).We were unable to detect any substantial synapse formation or neuronal activity in cultures without astrocytes and had great difficulty maintaining these cultures for prolonged periods of time, illustrating the crucial role astrocytes play in neuronal maturation and network formation.
In a coculture with human excitatory neurons, we show that both rat and hPSC-derived astrocytes promote the formation of functional synapses.Whole-cell electrophysiological recordings suggest that the increased activity of the neuronastrocyte cocultures observed using MEAs (Fig. 5) is not the result of changes in intrinsic neuronal properties (Extended Data Fig. 4-1), but more likely due to an increased frequency of EPSCs in the absence of a change in EPSC amplitude (Fig. 4C,D).Furthermore, our findings of an increased density of synapses in cocultures with hPSC-derived astrocytes (Fig. 4B) are likely to underlie at least a substantial proportion of the increase in EPSC frequency.The role of astrocytes in the formation and proper functioning of synapses has long been established (Allen and Eroglu, 2017).In correspondence with previous literature (Diniz et al., 2012), this process is accelerated in a fully human coculture system.As we demonstrate that hPSC-derived astrocytes are larger and cover more surface area compared with rat astrocytes in a coculture with neurons (Extended Data Fig. 3-1), it could be that the observed increases in neuronal activity are due to an increase in the total astrocyte cellular content in a fully human coculture system.Another possibility is that hPSC-derived astrocytes express an increased number of ion channels on their membrane or secrete different amounts of synaptogenic proteins, as has been described before for primary human astrocytes (Diniz et al., 2012).Importantly, however, we acknowledge that our experimental design precludes our ability to draw any meaningful conclusions about possible evolutionary differences between human and rat astrocytes.
The astroglia in this study were established from hPSC-derived NPCs in 28 d using a combination of BMP4 and LIF.We show that the functionality of these astrocytes compares favorably with rat astrocytes in a coculture system with human neurons.Human astrocyte-neuron cocultures are more active and mature more rapidly compared with a coculture of human neurons and rat astrocytes.Taken together, our data highlight the functional advantage of using hPSC-derived versus rat astrocytes for neuronal culture.Moreover, a fully human neuron-astrocyte coculture system provides a platform with a human genomic background for investigating astrocyte function and neuronal-glial interactions.

Figure 1 .
Figure 1.Astrocyte differentiation leads to a pure population of human astroglia in four weeks.A, Schematic representation of the differentiation protocol with representative differential interference contrast images of the different stages (scale bars, 200 μm).B, Staining for astrocytic markers 4 weeks after NPC stage (GFAP, green, and S100β, red) and an early neuronal marker (β-tubulin, cyan; scale bar, 200 μm).Immunofluorescent labeling of astrocytic markers of all cell lines used in this study is displayed in Extended Data Figure 1-1.Differentiation efficiency of astrocyte medium containing BMP4 and LIF, only BMP4, only LIF, or no additional growth factors is quantified in Extended Data Figure 1-2.C, Staining for glial membrane marker CD44 (red) and GFAP (green; scale bar, 50 μm).D, Nuclear astrocytic marker SOX9 is shown in red, together with CD44 in green (scale bar, 50 μm).E, A subset of astroglia (GFAP, green) stains positive for ALDH1L1 (red) a mature astrocyte marker (scale bar, 100 μm).F, Clustering of RNA sequencing data of three differentiation batches of NPCs to astroglia from four different human pluripotent stem cell lines (iPS1, iPS2, iPS3, and ES).Genes of interest are depicted on the left, on the top 3 neuronal genes and below 26 astrocyte genes.See Extended Data Figure 1-3 for a Principal Component Analysis plot of bulk RNA sequencing results.G-I, Flow cytometry data from iPS1, iPS2, and ES lines shows that 5-week-old astroglia cultures show comparable geometric mean fluorescence intensity (MFI) to primary rat astrocytes for GFAP (G), CD44 (H), and SOX9 (I).Fluorescence intensity histogram plots of these data are depicted in Extended Data Figure 1-4.

Figure 2 .
Figure 2. Single-cell RNA sequencing data confirms astrocyte identity and reveals transcriptional changes in coculture conditions.A, Heatmap showing Pearson's correlation between scRNA-seq cell clusters of the coculture sample (iPS1) and primary fetal brain tissue (Bhaduri et al., 2021).See Extended Data Figure 2-2 for heatmaps and corresponding data for astroglial and neuronal mono-cultures.B, UMAP projection of coculture sample with transferred cell type labels with the highest correlation from primary brain tissue.C, Expression of genes used for selection of astrocytes (top) and neurons (bottom) from samples in the coculture sample.D, UMAP projection of integrated astrocyte sample with cell type label transfer.E, UMAP projection of integrated astrocyte sample with original sample identity indicated in green (coculture) or red (mono-culture).F, Volcano plot of significantly up-(red) and down-regulated genes are significantly (Bonferroni corrected) up-(green) or downregulated (red) in iPSC-derived astrocytes when grown in a coculture with Ngn2 neurons.

Figure 5 .
Figure 5. MEA recordings of (co)cultures consisting of iPSC-derived neurons with or without astrocytes.A, Experimental setup of neuronal cocultures, NPCs (blue) are first differentiated to astroglia (green) and plated together with neurons (red) in a coculture [human n (individual wells) = 44 (iPS1 n = 24, iPS2 n = 8, iPS3 n = 6, ES n = 6), primary rat n = 18, neurons alone n = 8].B, Representative raster plots of spontaneous electrophysiological activity of individual wells in different coculture conditions, with human astrocytes (black) from different batches, rat astrocytes (blue), or neurons alone (yellow).C, Mean firing frequency over time.D, E, Activity of all electrodes over multiple experiments presented in frequency bins 10 (D) and 20 (E) days after plating.F-H, individual burst per minute (F), duration (G), and frequency (H).I-K, Network burst per minute (I) and duration (J) are similar between rat and human astrocyte cocultures; firing frequency (K) is increased in human astrocyte cocultures.MEA data from an independent laboratory using hPSC-derived astrocytes is presented in Extended Data Figure5-1.

Table 1 .
Overview of media and reagents used

Table 2 .
List of antibodies used

Table 3 .
Cluster composition of integrated astrocyte sample

Table 7 .
Top differentially expressed genes in human astrocyte progenitor cells and mature astrocytes as previously reported byZhang et al.