Navigating Human Astrocyte Differentiation: Direct and Rapid One‐Step Differentiation of Induced Pluripotent Stem Cells to Functional Astrocytes Supporting Neuronal Network Development

ABSTRACT

Astrocytes play a pivotal role in neuronal network development. Despite the well‐known role of astrocytes in the pathophysiology of neurologic disorders, the utilization of induced pluripotent stem cell (iPSC)‐derived astrocytes in neuronal networks remains limited. Here, we present a streamlined one‐step protocol for the differentiation of iPSCs directly into functional astrocytes without the need for ectopic gene expression or neural progenitor cell generation. We found that culturing iPSCs directly in commercial astrocyte medium, was sufficient to differentiate iPSCs into functional astrocytes within 5 weeks. More than 60 iPSC lines were successfully differentiated into astrocytes by independent researchers across 10 independent laboratories. Validation of the iPSC‐astrocyte cultures demonstrated consistent astrocyte differentiation with minimal batch‐to‐batch variability. In dept. characterization of a subset of iPSC lines confirmed astrocyte identity and functionality of the iPSC‐astrocyte monocultures by immunofluorescence, flow cytometry, RNA sequencing, glutamate uptake assays and calcium signaling recordings. Optimization of the protocol enabled co‐culture of iPSC‐astrocytes with Ngn2 iPSC‐derived neurons (iNeurons), promoting neuronal differentiation and synapse formation. Lastly, we used single‐cell electrophysiology and multi‐electrode arrays, by four independent researchers, to confirm robust neuronal network development in 5‐week‐old iPSC‐astrocyte and iNeuron co‐cultures. This protocol offers a rapid and efficient method to establish all‐human astrocyte‐neuron co‐cultures, facilitating the investigation of cell‐type‐specific contributions to disease pathogenesis. Its validation across numerous iPSC lines in 10 independent laboratories highlights the reproducibility of the protocol and positions it as a platform for advancing disease modeling in human neural networks.

• One‐step protocol enabling direct differentiation of human iPSCs into functional astrocytes within 5 weeks.

• Validated in > 60 lines across 10 labs.

• iPSC‐astrocytes support robust neuronal network development in co‐culture.

1. Introduction

Glia cells were originally thought to only provide physical and structural support to the neurons, and therefore were referred to as the glue of the brain (Somjen 1988). Astrocytes are the most abundant glial cell type in the brain. Even though structural support of astrocytes to neurons is still considered essential, over the years, there has been increasing recognition of the functional relevance of the astrocytes in the brain (He and Sun 2007). Nowadays astrocytes are known to support neurons at many different levels: neurogenesis, synaptogenesis including both formation and maturation of the synapses, metabolic support, synaptic plasticity, neurotransmission including uptake of neurotransmitters, as well as support of the blood–brain barrier and regulation of the blood flow (Volterra and Meldolesi 2005; Sidoryk‐Wegrzynowicz et al. 2011; Siracusa et al. 2019; Kim et al. 2019). The involvement of astrocytes in the pathogenesis of neurodevelopmental and neurodegenerative disorders is becoming increasingly recognized (Pietilainen et al. 2023; Koskuvi et al. 2022; Jantti et al. 2022; Di Giorgio et al. 2007).

Induced pluripotent stem cells (iPSCs) have paved the road to studying human brain development in a dish (Astick and Vanderhaeghen 2018), also allowing the use of patient‐derived material. One of the frequently used protocols to differentiate iPSCs into glutamatergic neurons makes use of Ngn2 overexpression, also known as induced neurons (iNeurons) (Frega et al. 2017; Lin et al. 2021; Zhang et al. 2013). To ensure survival and proper maturation of the iNeurons, co‐culturing with functional astrocytes is required, for which often primary embryonic rodent astrocytes are used (Frega et al. 2017; Zhang et al. 2013). The use of rodent astrocytes in these co‐cultures has significant limitations including genomic and functional differences (Vasile et al. 2017). Additionally, often wildtype rodent astrocytes are used, which potentially biases the development of disease‐linked neurons. Healthy astrocytes could potentially buffer neuronal effects, especially in the context of metabolic disorders. For these reasons iNeurons should preferentially be co‐cultured with iPSC‐derived astrocytes from the same genotype, which will not only circumvent the above‐mentioned limitations, but also allow for studying the cell‐type specific effect of the disease, especially for disorders in which the affected genes are enriched in astrocytes.

Many different iPSC‐astrocyte differentiation protocols have been described in literature (Koskuvi et al. 2022; Perriot et al. 2021, 2018; Tcw et al. 2017; Mulica et al. 2023; Chandrasekaran et al. 2016; Lendemeijer et al. 2024; Voulgaris et al. 2022; de Leeuw et al. 2020; Shaltouki et al. 2013; Jovanovic et al. 2023; Li et al. 2018; Neyrinck et al. 2021; Soubannier et al. 2020; Yeon et al. 2021; Giordano et al. 2023; Szeky et al. 2024; Dittlau et al. 2024; Ryan et al. 2020; Roybon et al. 2013; Jiang et al. 2013; Krencik et al. 2011; Brennand and Gage 2011; Qiu and Caiazzo 2025; Bosco et al. 2024; Berryer et al. 2023; Gupta et al. 2012; Santos et al. 2022; Tchieu et al. 2019). Generally, these protocols can be categorized into two main approaches: differentiation using ectopic gene expression of transcription factor(s) such as SOX9, NFIA and NFIB, or differentiation using an intermediate differentiation step towards neural progenitor cells (NPCs). Direct differentiation using gene overexpression often results in homogenous cultures of astrocytes (Li et al. 2018; Neyrinck et al. 2021; Yeon et al. 2021). In line with the large scale of astrocytic functions, increasing evidence shows a wide molecular, morphological, and functional astrocytic diversity in vivo (Westergard and Rothstein 2020; Ben Haim and Rowitch 2017; Khakh and Sofroniew 2015), which is also considered to be important for in vitro neuronal cultures (Gottipati et al. 2020; Stifani 2021). Astrocytes can also be derived from NPCs, resulting in more heterogeneous astrocyte populations. However, these protocols are often more time‐consuming and not always result in pure astrocyte populations (Li et al. 2018; Neyrinck et al. 2021; Yeon et al. 2021). Overall, functional readouts to validate new protocols to create iPSC‐derived astrocytes are mostly focused on only astrocyte characterization in monoculture, but their ability to support neurons in co‐culture is often not or very poorly described.

Given the need for iPSC‐derived astrocytes that can be used for astrocyte disease modeling and support iNeurons when co‐cultured together, we aimed to develop an efficient and robust differentiation protocol. We were inspired by the principle “we are what we eat,” previously described by Tcw et al. (2017), where they applied primary astrocytic media on NPCs to induce an astrocytic phenotype (Soubannier et al. 2020). In line with this principle, we have developed a fast and simple one‐step protocol to create functional iPSC‐astrocytes. In this protocol, iPSCs are directly cultured in commercial astrocyte medium, resulting in functional iPSC‐astrocytes within only 5 weeks of differentiation, without the use of forced overexpression (~1 month of stable expression of constructs) or an NPC intermediate step (~1 month of NPC differentiation). The astrocyte differentiation trajectory for each iPSC‐line across different batches is consistent demonstrating minimal batch‐to‐batch variability, in terms of morphology and astrocytic markers expression. Identity and functionality of the iPSC‐astrocytes were confirmed using a variety of assays, including RNA sequencing, glutamate uptake assays and calcium signaling recordings. We aimed to develop a robust and reliable protocol to co‐culture the iPSC‐astrocytes together with the iNeurons. After extensive testing of various adjusted co‐culture protocols, the iPSC‐astrocyte and iNeuron co‐cultures exhibited consistent neuronal network development when assessed, by four independent researchers from two different labs, with multi‐electrode arrays (MEAs). Our manuscript offers a thorough, step‐by‐step description of the optimization process, including negative data.

Validation of this protocol has been established in more than 60 iPSC lines across multiple independent labs, underscoring its robustness and broad applicability. To our knowledge, 12 independent labs have tested the protocol. Among these, one laboratory did not obtain astrocytes, while two others encountered previously described technical issues, specifically early proliferation arrest and residual iPSC colonies, which we consider solvable with further optimization. In addition, six laboratories have employed this protocol to establish co‐cultures of iPSC‐derived astrocytes with iPSC‐derived neurons. Two of these reported unsuccessful co‐cultures, one of which involved a modified version of the original procedure. Building on this collaborative foundation, we further invite researchers to apply the protocol and share their observations to help guide its continued refinement.

2. Methods

2.1. iPSC Cell Lines and Maintenance

In this study, iPSC astrocyte differentiation was performed for 64 different iPSC‐lines, which include commercial control lines as well as patient‐derived and isogenic lines (Table S1) (Schuurmans et al. 2023a, 2023b; Dyke et al. 2023). A subset of these lines has been used for detailed validation of the protocol, which is described in this study. iPSCs were cultured in Essential 8 Flex Basal Medium (Gibco; A2858501) supplemented with Primocin (0.1 mg/mL; Invivogen; ANT‐PM‐2) on Geltrex‐coated (Gibco; A1413301) 6‐well plates (Corning; 353046) at 37°C/5% CO₂. At around 80% confluency, the iPSCs were passaged with ReLeSR (Stemcell Technologies; 100‐0483) upon washing in Dulbecco's phosphate‐buffered saline (dPBS; Thermo Fisher; 14190169).

2.2. iPSC‐Astrocyte Differentiation

Prior to plating of the iPSCs, 6‐well plates were coated with 1 mL of 5 μg/mL (0.53 μg Biolaminin per cm²). Human recombinant laminin‐521 (Biolaminin; Biolamina; LN521‐05) in 1× dPBS⁺⁺ (Gibco; 14040117) and incubated overnight at 4°C. To start astrocyte differentiation, iPSCs were washed with dPBS and subsequently dissociated by incubating with TrypLE (Gibco; 12604021) for 3–5 min at 37°C/5% CO₂. Afterwards the iPSC were collected and washed by spinning down in DMEM/F12 (Gibco; 11320074). Upon removal of the supernatant, the pellet was resuspended in Astrocyte medium (AM; ScienCell; 1801) supplemented with RevitaCell (Gibco; A2644501) and Primocin. Accordingly, depending on the cell line, 150.000–300.000 (5.800–31.600 cells/cm²) cells were plated per well on a pre‐coated 6‐well plate and cultured at 37°C/5% CO₂. The next day AM supplemented with Primocin was refreshed to withdraw RevitaCell from the medium and to remove dead cells. AM supplemented with Primocin was fully refreshed every other day (e.g., three times a week). At 100% confluency (see Figure S1 for ideal splitting density), after around 5 days, the astrocytes were split by dissociation using TrypLE and washed by spinning down in DMEM/F12. The pellet was resuspended in AM supplemented with RevitaCell and Primocin and all cells were transferred to a T25 flask (Corning; 430372) (please note that this has been changed in the updated protocol, see Supporting Information for updated step‐by‐step protocol). The next day, AM was supplemented with Primocin was fully refreshed to withdraw RevitaCell from the medium. When the astrocytes reached again 90%–100% confluency the cells were split as described previously, and were completely transferred to a T75 flask (Corning; 430641U) followed by full change of AM supplemented with Primocin the day after. Hereafter astrocytes were split at 90%–100% confluency at 1:3 onto new T75 flasks and cultured in AM supplemented with Primocin in the absence of RevitaCell. After around 5 weeks of differentiation, medium changes were downscaled to twice a week, as the astrocytes matured and therefore the proliferation rate decreased. Throughout the differentiation process the cells were imaged using the Invitrogen EVOS XL Core Configured Cell Imager. An additional overview of all materials required for the astrocyte differentiation is provided in Table S2.

The protocol described above applies to the majority of the iPSC‐lines used in this study. However, for a subset of iPSC‐lines, adjustments to the differentiation protocol had to be made to improve astrocyte differentiation, which are described below. In addition, important notes are described, which have been shown to greatly improve astrocyte differentiation and decrease batch‐to‐batch variability and therefore are highly recommended to take into account.

Note A

To ensure full recovery upon thawing, iPSCs need to be cultured for at least 2 weeks before starting astrocyte differentiation. iPSCs are preferably cultured on Matrigel, Geltrex, or Biolaminin rather than on Cultrex or Vitronectin.

Note B

The number of iPSCs plated on day 0 (days in vitro, DIV 0) may vary from 150.000–300.000 cells per well of a 6‐well plate, generally depending on the proliferation rate of that specific iPSC‐line. For example, if a line is growing very slowly, for example, it needs to be split less than once a week (usually every 4–5 days), we would recommend a total of 300.000 iPSCs per well. In contrast, if an iPSC line shows a high proliferation rate, for example, has to be split twice a week, we recommend a total of 150.000 iPSCs per well. This is to prevent splitting within the first 5 days of differentiation, which we find to be less optimal for the astrocyte differentiation process. Ideally, the astrocytes are 100% confluent (very packed) around DIV 5–7 to ensure successful differentiation.

The seeding density on DIV 0 should be optimized for each iPSC line. We recommend testing several initial plating densities in parallel and then continuing differentiation with the one or two densities that perform best. In case of a “too low” number of iPSCs at DIV 0, the astrocytes might stop proliferating due to insufficient cell–cell contact. In contrast, a “too high” number of iPSCs at DIV 0 might result in a too high astrocyte density at DIV 5, often resulting in increased cell death upon splitting or iPSC contamination due to insufficient astrocyte differentiation (Figure S1A).

Note C

The standard procedure should be to split with RevitaCell during the first ~3 weeks of differentiation and to change medium the day after. After 3 weeks, the differentiating astrocytes can be split without RevitaCell. However, if the astrocytes have a very low proliferation rate (e.g., have to be split less than once a week), RevitaCell treatment can be prolonged until week five or six, or for the entire differentiation period if needed. If the astrocytes grow very fast (e.g., have to be split twice a week), RevitaCell treatment can be stopped by week two. During the initial differentiation of a new iPSC line, when culture parameters are still to be optimized, we advise for the first split from T75 to T75, testing one T75 split with RevitaCell and one without. This parallel approach helps prevent complete loss of the differentiation if cell survival is suboptimal without RevitaCell.

Note D

In case the astrocytes have not yet reached 100% confluency, but the cells do not seem to spread out, but instead grow on top of each other, the astrocytes can be split 1:1 or 1:2 (Figure S1B) in a new flask (Figure S1C). This can also be done in case the astrocytes have completely stopped proliferating (Figure S1D). Astrocytes may stop proliferating early (e.g., before or soon after the first split) if plated too sparsely, often leading to cell death. To avoid this, it is recommended to increase the initial plating density and split at a lower ratio (e.g., 1:2 rather than 1:3). If unsure whether to split 1:2 or 1:3, it is advised to test both in parallel; in general, a higher density (1:2) is safer.

Note E

Some lines show a very rapid increase in size and a drastic decrease in proliferation rate (Figure S1D), within the first week of astrocyte differentiation. From the 64 iPSC‐lines, the astrocytes derived from four iPSC‐lines did not proliferate at all upon plating in AM at DIV 0 or quickly thereafter. Therefore, the upscaling of the number of astrocytes was limited or even impossible. In addition, due to their relative fast increase in astrocyte volume and decreased proliferation rate, these cells are assumed to be more sensitive to the splitting procedure compared to the other astrocyte cultures, as increased cell death was observed upon splitting of these lines. To optimize the astrocyte differentiation for these iPSC lines, several adjustments have been tested. The following adjustments are currently considered most optimal to improve astrocyte differentiation and survival:

• −In order to ensure sufficient cell–cell contact, which promotes astrocyte proliferation, a higher number of iPSCs can be plated at DIV 0 (300.000–500.000 iPSCs per well).
• −The iPSC cultures of the respective lines also show decreased proliferation (e.g., splitting the iPSC line once every 2 weeks). Therefore, iPSC culturing on human Biolaminin‐coated plates instead of Geltrex‐coated plates (2 weeks prior to the start of astrocyte differentiation), has been shown to enhance the proliferation rate of these iPSC lines. We observed this to also result in increased proliferation of the astrocytes of these iPSC lines.
• −The most effective adjustment for improving astrocyte differentiation was complete withdrawal of FBS from the medium the day after the first split. These cell lines appeared to be sensitive to FBS, which likely induced a stress response that complicated further astrocyte maturation.

Note F

Some iPSC lines show relatively delayed astrocyte differentiation compared to the majority of the lines (often iPSC lines being reprogrammed using Sendai vectors). This so‐called delayed astrocyte differentiation is defined by iPSC‐like morphology during the first one‐ or two‐weeks of differentiation. This also includes some iPSC colonies remaining within the astrocyte cultures for the first 1 or 2 weeks of differentiation (Figure S1E), indicating less efficient astrocyte differentiation for these lines. In order to improve astrocyte differentiation, it is recommended to plate a relatively low number of iPSCs at DIV 0 (150.000 cells per well). In addition, to lose the remaining iPSC colonies within the astrocyte cultures, it is recommended to split the astrocyte cultures without RevitaCell.

Note G

We tested cryopreservation of DIV 23 astrocytes and subsequent thawing. Although the cells survived freezing, their proliferation rate decreased after recovery. Therefore, whenever possible, it is preferable to use freshly differentiated astrocytes rather than frozen stocks.

2.3. Neuronal Progenitor Cells Differentiation

The protocol for the generation of NPCs from iPSCs was based on a previously published protocol described by Shi et al. (2012). Briefly, iPSC colonies were dissociated into single cells using Accutase (Sigma; A6964) and 200.000–300.000 cells (depending on the iPSC‐line) were plated in a Matrigel (Corning; 734‐1101) coated 12‐well plate (Corning; 353043) and cultured in E8 Flex medium supplemented with RevitaCell. At 95% cell confluency, which was achieved 4 days after plating, the medium was switched to neural induction medium (NIM). NIM contains 1:1 mixture of Medium A and Medium B (referred to as neural maintenance medium (NMM)) supplemented with 1 μM dorsomorphin (Sigma Aldrich; P5499) and 10 μM SB431542 (Stemcell Technologies; 72234) to inhibit SMAD signaling. Medium A medium consists of DMEM/F12, 1× N2 (Gibco; 17502001), 5 μg/mL Insulin (Gibco; 13105002), 2 mM GlutaMAX (Gibco; 35050061), 100 μM non‐essential amino acids (Thermo Fisher Scientific; 11140‐035), 100 μM β‐mercaptoethanol (Thermo Fisher Scientific; 21985‐023) and Primocin. Medium B is made of neurobasal medium (Gibco; 21103049), 1× B‐27 (Gibco; 17504044), 2 mM GlutaMAX and Primocin. For the first 10–12 days medium was changed every day and the cells were closely monitored for neuroepithelial induction which was based on morphological changes and was expected to start between eight to 12 days after plating. When neuroepithelium was induced (D11), cells were detached from the Matrigel coated plate using ReLeSR and transferred onto poly‐L‐ornithine (PLO; Sigma; P3655)/Biolaminin coated 12‐well plates and cultured in NIM supplemented with RevitaCell. After 24 h, NIM was changed one more time, which is the last day of incubation with dual SMAD inhibitors (D12). Then medium was switched to NMM supplemented with 10 ng/mL EGF (Peprotech; AF‐100‐15) and 10 ng/mL FGF (R&D systems; 235‐F4‐025) and was refreshed every 3 days. At confluency, after approximately 1 week of culture, the cells were detached with ReLeSR and split onto PLO/Biolaminin coated 6‐well plates (P0 of the NPCs). After overnight incubation, the medium was changed to NMM supplemented with 20 ng/mL EGF and 20 ng/mL FGF. From this point on, the cells were passaged at confluency, which was usually 4 days after passage, in a 1:2 or 1:3 ratio on PLO/Biolaminin pre‐coated wells of a 6‐well plate in NMM medium supplemented with 20 ng/mL EGF and 20 ng/mL FGF and RevitaCell. An additional overview of all materials required for the NPC differentiation is provided in Table S3. Accordingly, NPC‐astrocytes were generated according to the iPSC‐astrocyte protocol.

2.4. Rat Astrocytes

The rat astrocytes used in this study were obtained from embryonic E18 rat brains, as described previously (Frega et al. 2017). The pregnant rat (Crl:WI[Han], Charles River) underwent anesthesia with isoflurane and was euthanized by cervical dislocation; accordingly the pups were collected and decapitated. The astrocytes were isolated from the brains and pooled. Rats were housed individually since they were euthanized immediately upon arrival at the animal facility. All animal procedures were conducted in accordance with the Animal Care Committee of the Radboudumc, The Netherlands, and adhered to the guidelines of the Dutch Council for Animal Care and the European Communities Council Directive 2010/63/EU (ethics approval for project W P 2 0 17‐0 0 4 8‐0 0 3).

2.5. Co‐Culture of iNeurons and iPSC‐Derived Astrocytes

iPSCs were differentiated towards excitatory neurons using overexpression of NGN2 as described previously, with minor adjustments (Frega et al. 2017, 2019). Briefly, prior to differentiation, iPSCs were infected with lentiviral constructs encoding Ngn2 combined with rtTA. In order to obtain a pure population of doxycycline‐inducible excitatory neurons, iPSCs were selected using increasing concentrations of G418 (100–250 μg/mL; Sigma‐Aldrich; #G8168) for the rtTA‐construct and puromycin (1–2 μg/mL; Sigma; P9620) for the Ngn2‐construct.

Forty‐eight‐CytoView (Axion Biosystems) MEA plates or nitric‐acid treated glass coverslips (Epredia; 631‐0713) were pre‐coated with 50 μg/mL PLO diluted in borate buffer for 3 h followed by overnight incubation with 10 μg/mL Biolaminin in 1× dPBS⁺⁺ at 4°C. To start neuronal differentiation (DIV 0), iPSC colonies were dissociated into single cells using TrypLE and plated onto the pre‐coated MEA plates or coverslips in E8 basal medium (Gibco; A1517001) supplemented with RevitaCell, doxycycline (4 μg/mL; Sigma; D5207), and Primocin. The next day, medium was refreshed to DMEM/F12 supplemented with N2, 10 ng/mL NT3 (Peprotech; 450‐03), 10 ng/mL BDNF (Peprotech; 450‐02), MEM non‐essential amino acid solution (NEAA; Sigma‐Aldrich; M7145), 4 μg/μL doxycycline, and Primocin. Three days after plating (DIV 3), all medium was refreshed to neurobasal medium supplemented with B‐27, GlutaMAX, Primocin, NT3, BDNF, and 4 μg/μL doxycycline. In addition, Cytosine β‐D‐arabinofuranoside (Ara‐C; 0.5 μM; Sigma‐Aldrich; C1768) was added to remove proliferating iPSCs from the culture.

At DIV 6 of the neuronal differentiation, iPSC‐derived astrocytes were added to the neurons. For this the same procedure as described for splitting is applied. Briefly, the iPSC‐derived astrocytes were dissociated using TrypLE and washed by spinning down in DMEM/F12. The pellet was resuspended in neurobasal medium supplemented with B‐27, GlutaMAX, Primocin, NT3, BDNF, doxycycline (1 μg/mL; note: lower concentration), RevitaCell and 2.5% AM FCS (ScienCell; 0010) (please note that this is not added in the updated protocol, see Supporting Information). Then medium of the neurons was refreshed with neurobasal medium supplemented with B‐27, GlutaMAX, Primocin, NT3, BDNF, doxycycline (1 μg/mL), RevitaCell and 2.5% AM FCS. Accordingly, the iPSC‐derived astrocytes were added to the neurons in a neuron: astrocyte ratio 2:1, for example, 6.750 astrocytes to 13.500 neurons plated on a 48‐well CytoView MEA plate. Starting at DIV 8, half of the medium was changed three times a week. To reduce stress for both the iPSC‐derived neurons as well as the iPSC‐astrocytes, doxycycline was withdrawn from the medium starting at DIV 10. The co‐cultures were kept at 37°C/5% CO₂ throughout the whole differentiation process. An additional overview of all materials required for the co‐culture is provided in Table S4.

Note H

It is known that the astrocytes derived from some iPSC‐lines are more sensitive to dissociation than others, leading to an increased cell death upon splitting. Therefore, it is recommended to optimize the number of cells to be used for the co‐culture for each iPSC astrocyte culture.

Note I

It is crucial to pretreat the coverslips with nitric acid and subsequent extensive washing in ultrapure water. Figure S2 shows the (long‐term) effects of this treatment on the cultures, indicating that no treatment results in severe clustering of the neurons and astrocytes.

2.6. Immunocytochemistry

Cells were fixed with 4% paraformaldehyde (Sigma‐Aldrich; 441244)/4% sucrose (Sigma; S7903) for 15 min, washed three times with PBS (1×; Sigma; P5493) and blocked for 1 h with blocking buffer (BB; 5% normal donkey serum [Jackson Immuno Research; 017‐000‐121]), 5% normal horse serum (Gibco; 26050070), 5% normal goat serum (Invitrogen; 10189722), 1% BSA (Sigma‐Aldrich; A‐6003), 0.1% D‐lysine (Sigma‐Aldrich; L802), 1% glycine (Sigma‐Aldrich; L802) and 0.4% Triton X‐100 (Sigma‐Aldrich; T8787) in PBS at room temperature (RT). Primary and secondary antibodies were diluted in BB and incubated overnight at 4°C or for 1 h at RT, respectively. To stain the nuclei, cells were incubated with Hoechst (Thermo Scientific; 62249) diluted in PBS for 10 min at RT and accordingly mounted using DAKO fluorescent mounting medium (DAKO; S3023). Zeiss Axio Imager Z1 was used to image the samples. The following antibodies were used: Rabbit anti‐Vimentin (1:300; Abcam; ab92547), Mouse anti‐Tuj1 (1:300; BioLegend; 801201), Mouse anti‐Nestin (1:200; Invitrogen; MA1110), Rabbit anti‐EAAT1 (1:100; Abcam; ab416), Rabbit anti‐GFAP (1:300; Sigma‐Aldrich; AB5804), Mouse anti‐CD44 (1:300; Invitrogen; MA5‐13890), Mouse anti‐ALDH1L1 (1:300; Novus Biologicals; NBP2‐50045), Rabbit anti‐GLUD1 (1:300; Invitrogen; PA5‐28301), Mouse anti‐AQP4 (1:300; Biorbyt; ORB323095), Rabbit anti‐PDLIM7 (1:300; Novus Biologicals; NBP2‐58734), Guinea pig anti‐MAP2 (1:1000; Synaptic Systems; 188004), Mouse anti‐Synapsin I (1:1000; Synaptic Systems; 106001), Goat anti‐Rabbit Alexa Fluor 568 (1:1000; Molecular probes; A11011), Goat anti‐Mouse Alexa Fluor 488 (1:1000; Invitrogen; A11029), Goat anti‐Guinea Pig Alexa Fluor 488 (1:1000; Invitrogen; A11073), Goat anti‐Guinea Pig Alexa Fluor 647 (1:1000; Molecular probes; A21450), Goat anti‐Mouse Alexa Fluor 647 (1:1000; Molecular probes; A21237), Streptavidin, Alexa Fluor 568 conjugate (1:1000; Invitrogen; S11226). Zeiss Axio Imager Z1 was used to image the samples (see Supporting Information S6 for details regarding laser and exposure settings).

2.7. Flowcytometry Analysis

At DIV 35, astrocytes were dissociated into a single‐cell suspension using TrypLE. The cells were dissolved in Viability Dye eFluor 780 (Invitrogen; 65‐0865‐14) and incubated at 4°C for 30 min. Subsequently, cells were washed in flow buffer by centrifugation at 300× g for 5 min (1% BSA in PBS) and then fixed for 15 min at RT with 2% paraformaldehyde followed by another wash. To prevent a specific binding to Fc‐receptors, the cells were blocked with 2.5% Human TruStain FcX blocking solution (BioLegend; 422302) for 10 min. Subsequently, the cells were incubated with anti‐CD44 Brilliant Violet 421 (Biolegend; 338809) and CD49f PE/Cyanine7 (Biolegend; 313621) for 30 min at 4°C. Afterwards, the cells were washed two times and finally dissolved in 200 μL flow buffer for flow cytometry measurement. Gating of side scatter (SS) area versus forward scatter (FS) area was used to define a cell population based on size and granularity. To remove duplicates, gating between FS height and FS Area was used. Further, dead cells were negatively gated by viability dye intensity (Invitrogen; 17324361). This dye reacts with free amines and gives an intense fluorescent signal when the membrane is ruptured. In live cells, the reactivity of the dye is restricted to the cell membrane, which results in distinctively different fluorescence. An unstained sample was used to determine the auto‐fluorescence levels of cells in the channels corresponding to the antibodies used. This was used to set the threshold for the CD49f‐ and CD44‐positive populations. Analysis was performed using Kaluza C Analysis software (Beckman Coulter).

2.8. RT‐qPCR

Total RNA was isolated from the cells with NucleoSpin RNA Mini Kit (Macherey‐Nagel; MN 740955.250), according to the manufacturer's instructions, with the following adjustments: no use of β‐mercaptoethanol, and 30 min instead of 15 min incubation with rDNase. Subsequently, RNA was reverse‐transcribed into cDNA using iScript cDNA Synthesis Kit (Bio‐Rad; 1708891). GoTaq quantitative PCR (qPCR) Master Mix (Promega; A6002) was used to conduct the qPCR. The following program was used in the 7500 Fast Real Time PCR System apparatus (Applied Biosystems): a denaturation at 95°C for 2 min, 40 cycles of 30 s 95°C and 30 s 60°C followed by a melting curve stage of 15 s 95°C, 30 s 60°C and 15 s 95°C. Samples were always prepared in triplicate. Outliers were determined if a value differed more than 0.5 Ct from the other two values in the technical triplicate. The relative mRNA expression was calculated using the 2^−ΔΔCt method with normalization against the housekeeping genes GUSB, ERCC6 and PPIA. Primers are listed in Table S5.

2.9. RNA Sequencing

2.9.1. RNA‐Seq Library Preparation

Five different iPSC‐derived astrocyte cultures (iPSC‐line 1, iPSC‐line 11, iPSC‐line 4, iPSC‐line 6 and iPSC‐line 7) and three different NPC‐derived astrocyte cultures (iPSC‐line 1, iPSC‐line 2 and iPSC‐line 15) were prepared for RNA sequencing analysis. Biological triplicates of all astrocyte cultures were seeded onto 6‐well plates in AM. Upon confluency, the astrocytes were washed twice with ice‐cold PBS and collected using DNA/RNA shield (Zymo 20 Research; #ZY‐R1200‐125). Accordingly, RNA was isolated using Quick‐RNA Microprep kit (Zymo 20 Research, #ZY‐R1051), according to the manufacturer's instructions. The quality of RNA was assessed using Agilent's Tapestation system and the RNA Integrity Number (RIN) values ranged between 7.3 and 9.6. Subsequently, cDNA libraries were prepared using the NEBNext Ultra II Directional RNA Library Prep Kit, and paired‐end reads were sequenced on an Illumina NovaSeq 6000 platform at GenomeScan B.V. Leiden.

2.9.2. RNA‐Seq Data Processing

We used Fastp to eliminate PolyG artifacts and clip adapters, including a list of adapter sequences currently used by Illumina (Chen et al. 2018). Subsequently, we mapped the reads to the GRCh38 human reference genome using HISAT2. As the library was reversely stranded, we set ‐‐rna‐strandness = RF. The resulting SAM files were sorted and indexed using SAMtools (Li et al. 2009). Next, we performed UMI deduplication with UMI‐tools (Smith et al. 2017), and feature counting was done with Subread's featureCounts (Liao et al. 2014).

Paired‐end RNA‐seq data of post‐mortem human astrocytes were downloaded from the GSE73721 GEO repository (Zhang et al. 2016). Fastp was used to trim the reads, while HISAT2 was employed to align the reads to the GRCh38. SAM files were sorted and indexed with SAMtools, and featureCounts for feature counting.

2.9.3. RNA‐Seq Analysis

Raw count matrices were loaded in R version 4.2.1. When Ensemble IDs mapped to the same gene symbols, we considered only the IDs with the highest expression per sample. We filtered out genes with low expression by retaining those with expression values exceeding one transcript per million reads in at least six samples. Subsequently, counts were normalized using voom function from the limma R package (Ritchie et al. 2015). Genes were then ranked based on their variance, and Principal component analysis (PCA) was performed on scaled expression data of the 2500 most variable genes, using prcomp function. Given that the first principal component predominantly explained the variability between human post‐mortem and in vitro samples, the Euclidean distance between each pair of samples was calculated based on the values of the principal components, excluding the first component. These distances were then used for hierarchical clustering with pheatmap function of pheatmap R package. To create heatmaps illustrating the expression patterns of astrocyte marker genes across the samples, normalized counts were scaled between genes. All plots were generated using custom code built on ggplot2 v.3.4.0 functions in R.

2.10. Glutamate Uptake Analysis

To prepare samples for glutamate uptake analysis, 20.000 iPSC‐astrocytes (DIV 34) or rat astrocytes per well were seeded onto 24‐well plates (Corning; 353047) and cultured in AM overnight at 37°C/5% CO₂. Three hours prior to harvest, AM was supplemented with 100 μM glutamic acid (Sigma; G1251) and Primocin was refreshed. Upon incubation, the medium was collected and stored at −20°C until further processed. To measure uptake of glutamic acid by the iPSC‐derived astrocytes, samples were diluted 50 times and measured using the Amplex Red Glutamic Acid/Glutamate Oxidase Assay Kit (Thermo Fisher Scientific; A12221) according to the manufacturer's instructions. Fluorescent signal was measured at 540 nm excitation and 590 nm emission using a microplate reader (Tecan's Spark 20 M multimode reader). The percentage of glutamic acid taken up by the astrocytes relative to the negative control was calculated.

2.11. Biocytin Filling Using Patch‐Clamping

Coverslips with iPSC‐astrocyte monocultures were placed in a recording chamber on the stage of an Olympus BX51WI upright microscope (Olympus Life Science), equipped with infrared differential interference contrast optics, an Olympus LUMPlanFL N 60× water‐immersion objective (Olympus Life Science), and a kappa MXC 200 camera system (Kappa optronics GmbH) for visualization. Through the recording chamber there was a continuous flow of carbongenated (95% O₂/5% CO₂) artificial cerebrospinal fluid (aCSF) (124 mM NaCl (Sigma; S5886), 1.25 mM NaH₂PO₄ (Sigma; S0751), 3 mM KCl (Sigma; P9333), 26 mM NaHCO₃ (Sigma; S5761), 11 mM Glucose (Sigma; G5767), 2 mM CaCl₂ (Merck; 1023780500), 1 mM MgCl₂ (Merck; M2670)) at 32°C. Patch pipettes (ID 0.86 mm, OD 1.05 mm, resistance 5–8 MΩ) were pulled from borosilicate glass with filament and fire‐polished ends (Science Products GmbH) using the Narishige PC‐10 micropipette puller. Pipettes were filled with a potassium‐based solution with 0.5% biocytin (Sigma‐Aldrich; B4261) containing: 130 mM K‐Gluconate, 5 mM KCl, 10 mM HEPES, 2.5 mM MgCl₂, 4 mM Na₂‐ATP, 0.4 mM Na₂‐GTP, 10 mM Na‐phosphocreatine and 0.6 mM EGTA (with pH adjusted to 7.2 and osmolarity to 290 mOsmol). Passive properties were continuously monitored using a Digidata 1140A digitizer and a Multiclamp 700B amplifier (Molecular Devices). After reaching whole‐cell configuration, astrocytes were kept at a holding potential of −60 mV for 15 min to allow the biocytin to diffuse out of the pipette into the attached astrocytes. After 15 min the pipette was slowly retracted to reseal the plasma membrane. Coverslips were left in the recording chamber for another 15 min to wash off excess biocytin.

2.12. Calcium Imaging

To prepare the cells for calcium imaging analysis, iPSC‐derived astrocytes, grown on glass coverslips, were incubated with Fluo‐8‐AM calcium binding dye (Abcam; ab142773) diluted in HBSS buffer (Sigma; H6648) supplemented with 20 mM HEPES (Sigma‐Aldrich; H4034) for 30 min at 37°C. To remove excessive Fluo‐8‐AM, the cells were washed three times with HBSS buffer and subsequently allowed to recover in AM without FBS for 15 min. To image the iPSC‐derived astrocytes, the OLYMPUS BX51WI Upright Microscope using a 20× magnification was used combined with the Hamamatsu camera (ORCA Flash2.8). During recordings, the cells were illuminated at 471 nm by LED light (KSL470, Rapp OptoElectronic). The cells were perfused continuously with oxygenated aCSF (119 mM NaCl, 26.2 mM NaHCO3, 2.5 mM KCl, 1 mM NaH2PO4, 21 mM glucose, 2 mM CaCl2, 2.3 mM MgCl2) and measured in aCSF supplemented with or without 50 μM glutamic acid. Fluorescent signals were recorded for 10 min at 250 msec intervals (4 Hz). During evoked recordings, the cells were recorded for 2 min to obtain the baseline measurements. Accordingly, while recording, the original aCSF was replaced by aCSF supplemented with 50 μM glutamic acid, which was slowly perfused towards the cells.

Data were analyzed using a home‐made script in MATLAB (The Math Works Inc. MATLAB. Version 2020b). Regions of interest (ROIs) were selected manually per measurement to represent single iPSC‐derived astrocytes. Subsequently, the mean signal of each ROI was plotted per frame. The calcium response signal was calculated as the change in fluorescent signal of the ROI according to the formula ΔF/F = (F − F0)/F0. F represents the fluorescent signal for each ROI over time, while F0 represents the baseline fluorescence. To correct for the decay of the baseline intensity due to bleaching, exponential fitting was used. To represent the mean intensity for each ROI, the area under the curve was calculated.

2.13. Olink Proteomics

To assess the astrocyte secretome, DIV 35 iPSC‐line 1 astrocytes were sampled for Olink proteomics using the Target 96 inflammation panel. The Olink assay is based on Proximity Extension Assay (PEA) technology. In short, 96 different oligonucleotide antibody‐pairs are designed which can bind to the proteins in the sample. These antibody‐pairs contain unique DNA sequences, which allow for hybridization to each other upon specific binding to the protein. Subsequent proximity extension will create 96 unique DNA reporter sequences which are amplified by real‐time PCR.

Astrocytes were seeded onto 6‐well plates several days prior to sample collection to allow the cells to grow confluent. The day before sample collection, the medium was refreshed. At DIV 35, the medium of the astrocytes was collected and stored at −20°C until further processing. In addition, as baseline control, fresh medium was collected and stored. The Olink assay was performed according to the manufacturer's instructions. Several controls were included. The inter‐plate Control (IPC) is a pool of 92 antibodies, each with one pair of unique DNA‐tags positioned in fixed proximity. In addition, a negative control was included, consisting of running buffer, as well as a sample control consisting of pooled samples to assess potential variation between runs. The data were first normalized for the IPC. Afterwards, the fold change for each protein was calculated relative to the baseline medium sample.

2.14. Synapsin Quantification

To quantify the number of Synapsin puncta, iPSC‐astrocyte and iNeuron co‐cultures were stained for MAP2, Vimentin, and Synapsin. Ten images per coverslip were made using Zeiss Axio Imager Z1. Accordingly, the number of Synapsin puncta per X μm dendrite was quantified using FIJI. Synapsin puncta for two neurons (one or two dendrites) per image were calculated. Accordingly, the number of Synapsin puncta per 10 μm dendrite was calculated.

2.15. MEA Recordings and Analysis

To record spontaneous network activity of human co‐cultures using MEAs, iPSC‐derived neurons and iPSC‐astrocytes were plated onto the CytoView 48‐well MEA plates (Axion Biosystems) as described previously. At least four wells per culture condition were included for each experiment. In brief, each well contains 16 electrodes with a diameter of 50 μm arranged in a 4 × 4 grid spaced 350 μm apart. Recordings were performed weekly on a Maestro Pro MEA system with AxIS Navigator software (Axion Biosystems), starting at DIV 21 until DIV 58 of the differentiation. Spontaneous network activity was recorded for 5 min, after a 15 min period of acclimatization, in a recording chamber which was maintained at 37°C/5% CO₂. An adaptive spike threshold at which spikes were detected was set at ±6 STD.

Analysis of the MEA recordings was performed using the Envelope network burst detection algorithm of the Axion Biosystems Neural Metrics Tool. Threshold factor was set at 3.0, minimum inter‐burst interval (ms) at 100 ms, and both the minimum number of active electrodes and burst inclusion were set at 50%. From the analysis, the following parameters were extracted: mean firing rate, the number of bursts, the number of network bursts (NBs), and the NB duration.

2.16. Patch Clamp

Single cell recordings in whole cell patch clamp configuration were performed at DIV35 neurons and as previously described (Frega et al. 2019; van Hugte et al. 2023). Coverslips were placed in a recording chamber on the stage of an Olympus BX51WI upright microscope (Olympus Life Science), equipped with infrared differential interference contrast optics, an Olympus LUMPlanFL N 60× water‐immersion objective (Olympus Life Science), and a kappa MXC 200 camera system (Kappa optronics GmbH) for visualization. The recording chamber was continuously perfused with carbongenated (95% O₂/5% CO₂) aCSF (124 mM NaCl, 1.25 mM NaH₂PO₄, 3 mM KCl, 26 mM NaHCO₃, 11 mM Glucose, 2 mM CaCl₂, 1 mM MgCl₂) at 32°C. Patch pipettes (ID 0.86 mm, OD 1.05 mm, resistance 5–8 MΩ) were pulled from borosilicate glass with filament and fire‐polished ends (Science Products GmbH) using the Narishige PC‐10 micropipette puller. For recordings of intrinsic properties in current clamp mode and spontaneous excitatory postsynaptic currents (sEPSCs) in voltage clamp mode, pipettes were filled with a potassium‐based solution containing: 130 mM K‐Gluconate, 5 mM KCl, 10 mM HEPES, 2.5 mM MgCl₂, 4 mM Na₂‐ATP, 0.4 mM Na₃‐GTP, 10 mM Na‐phosphocreatine, 0.6 mM EGTA (with pH adjusted to 7.2 and osmolarity to 290 mOsmol). All recordings were acquired using a Digidata 1140A digitizer and a Multiclamp 700B amplifier (Molecular Devices), with a sampling rate set at 20 kHz and a lowpass 1 kHz filter during recording. Recordings were not corrected for liquid junction potential (±17 mV), which was calculated according to the stationary Nernst–Planck equation (Marino et al. 2014) using LJPcalc software (https://swharden.com/LJPcalc). Recordings were not analyzed if series resistance was above 25 MΩ or when the recording reached below a 10:1 ratio of membrane to series resistance. Resting membrane potential (Vrmp) was determined directly after reaching whole‐cell configuration. Further analysis of active and passive membrane properties was conducted at a holding potential of −60 mV. sEPSCs were measured by 10 min continuous recording in aCSF at a holding potential of −60 mV. Amplitude and frequency of individual events within sEPSC recordings were not analyzed using Mini Analysis.

2.17. Meta‐Analysis of MEA Data for FBS Effect

2.17.1. Preprocessing and Merging of Neural Metrics

MEA recordings collected across four researchers were batch‐processed using the Axion Biosystems NeuralMetric Tool. Single‐electrode bursts were detected when a minimum of 5 consecutive spikes each had an inter‐spike interval not exceeding 100 ms. Network bursts were identified using the envelope algorithm, with a detection threshold of 1.5, a minimum inter‐burst interval of 100 ms, a minimum of 70% active electrodes required, and a burst inclusion criterion of 75%. Per‐well neural metrics were exported as CSV files for downstream analysis.

Files per researcher were imported into R (version ≥ 4.0) using the readxl and tidyverse packages. The first 29 rows of each file, which contained instrument metadata rather than measurements, were skipped during import, and the subsequent 82 rows of neural metrics were retained. Data were transposed so that each row represented one well and each column one electrophysiological parameter. A well identifier (wellID) was derived from the row names and appended as an explicit column. Individual recording files were combined into a single data frame, retaining the source filename as a reference variable. Wells corresponding to neuronal networks with rodent astrocyte cultures were excluded from further processing. Experimental covariates were extracted from the source filename and the plate map. These included: researcher identity, plate identifier, plating month, presence of inhibitory neurons, microglia co‐culture status (determined by parsing +MG/−MG annotations in the treatment label), cytokine treatment, FBS usage (assigned per plate ID via a predefined lookup table), recording date, days in vitro (DIV), genotype, and well border position (border wells were identified as the outermost row and column positions of the 48‐well plate layout). Wells that had received pharmacological treatments during the recording trajectory were excluded. Electrophysiological parameters (columns 13–88) were coerced to numeric.

Preprocessed data files from four researchers were imported and combined into a single data frame using rbind. Recordings were restricted to timepoints with a minimum of 100 wells represented across the merged dataset. Timepoints falling below this threshold were excluded, yielding an effective analysis window of DIV14 to DIV70. Columns that were entirely NA across all samples were dropped. Columns with no observed variance (i.e., ≤ 1 unique non‐missing value) were identified and removed, as they carry no information for downstream analysis. Missing values were imputed with zero (reflecting the absence of detectable activity). Two outlier wells identified during exploratory PCA were excluded prior to all analyses. The cleaned, merged data frame was exported for analysis.

2.17.2. Dimensionality Reduction, Variance Partitioning and FBS Effect Analysis

Neural metrics parameters were z‐score scaled and submitted to principal component analysis (PCA) using prcomp. PCA scores were annotated with experimental covariates (DIV, researcher, FBS condition, genotype, microglia, border position) and MEA parameters (such as synchrony, burst percentage, burst duration, random spikes) and visualized using ggplot2.

To quantify the contribution of each experimental covariate to the overall variability in the MEA parameter space, variance partition analysis was performed using the variancePartition package. All electrophysiological parameters were z‐score scaled and transposed to a features × samples matrix. A linear mixed model formula was specified including DIV as a fixed continuous covariate and researcher, plate ID, plating month, inhibitory neuron status, microglia status, FBS condition, genotype, and border position as random effects. Variance fractions were estimated across all parameters using fitExtractVarPartModel and summarized with plotVarPart.

To isolate the effect of FBS exposure while controlling for known confounders, differential analysis across all electrophysiological parameters was performed using the dream function from variancePartition, fitting a linear mixed model per parameter. The model included FBS condition, genotype, microglia status, DIV, and inhibitory neuron status as fixed effects, and researcher, plate ID, and border position as random effects. Differential parameters associated with FBS exposure were extracted using topTable (coefficient FBSyes) with a false discovery rate (FDR) threshold of 1% (Benjamini–Hochberg adjusted p value < 0.01). Results were displayed as a volcano plot of log‐fold change versus −log₁₀(FDR). For visualization of individual parameters, group means ± standard error of the mean (SEM) were plotted as bar graphs with t‐test p values.

2.18. Statistical Analysis

Statistical analysis was performed using Graphpad Prism (version 10 for Windows, GraphPad Software). For comparisons between two conditions at one time point, an unpaired t‐test was performed. All values are reported as mean ± standard deviation (SD). Results with p values lower than 0.05 were considered as significantly different (*), p < 0.01 (**), p < 0.001 (***), p < 0.0001 (****).

3. Results

3.1. Direct Differentiation From iPSCs Towards Morphologically Mature Astrocytes

Astrocyte differentiation was initiated by plating the iPSCs as single cells on human Biolaminin‐coated plates and culturing them in commercial AM (ScienCell) supplemented with RevitaCell (Figure 1A, methods). After 48 h of plating, we already observed the first changes in morphology as the cells exhibited a more flat and less condensed appearance compared to iPSCs. Over the course of the first week of differentiation, these morphological alterations persisted, with cells gradually increasing in size, flattening, and demonstrating reduced proliferation rates relative to iPSCs. By the end of the initial 3 weeks, the astrocytes formed a predominantly homogenous culture. However, the astrocytes changed into a more heterogeneous culture, characterized by differences in astrocyte size and morphology, starting after approximately 3 weeks (depending on the iPSC‐line) (Figure 1A). To further characterize morphological maturation over time, the iPSC‐astrocytes were immunostained for vimentin at multiple time points (Figure 1B), confirming increased morphological heterogeneity over time, characterized by increased astrocyte size as well as morphological complexity. This was most strikingly observed by the increasing number of astrocytic processes during maturation. In this study, a total of 64 iPSC‐lines (Table S1) were differentiated towards astrocytes by independent researchers across 10 independent labs (Figure 2 and Figure S3) (Radboudumc (Nadif Kasri lab/Garanto lab/van Bokhoven lab/de Witte lab), Ellis lab, Jung‐Klawitter lab, Parent lab, Weckhuysen lab, Elgersma lab, Lefeber lab, de Vries lab, Polymenidou lab, and Dolga lab). For a subset of these iPSC‐astrocyte cultures, we performed in‐depth characterization to validate the protocol and to confirm astrocyte identity (Figure 1C).

FIGURE 1.

Direct differentiation from iPSCs towards astrocytes results in morphologically mature astrocytes. (A) Schematic representation of the protocol to directly differentiate iPSCs towards astrocytes, including representative brightfield images over time during differentiation. All pictures were taken at the same magnification (scale bar = 100 μm). (B) Representative images of vimentin (white) immunostaining of iPSC‐astrocytes (iPSC‐line 4) at DIV 7, DIV 21, DIV 35, DIV 85 and DIV 111. All pictures were taken at the same magnification (scale bar = 50 μm or 25 μm in zoom image). (C) Representative brightfield images of 10 different DIV 35 iPSC‐astrocyte cultures of one batch. All pictures were taken at the same magnification (scale bar = 100 μm).

FIGURE 2.

Astrocyte differentiation of 58 different iPSC lines across 10 different labs. Representative images of vimentin (red) immunostaining of 57 of the 64 different iPSC‐astrocyte cultures, which were cultured across different batches, by several different researchers and across 10 different labs. The astrocytes are around 4–6 weeks old. Scale bar represents 50 μm.

Among all the differentiated iPSC lines, we found line‐to‐line variability, defined as differences in morphology, size, and proliferation rate, across different batches (Figures 1C and 2; Figure S3). Importantly, the astrocyte differentiation trajectory for each iPSC line was consistent across different batches, as shown by similar morphological characteristics across different astrocyte batches using brightfield and immunofluorescent imaging (Figure S4). These results demonstrate minimal batch‐to‐batch (morphological) variability. We found no direct correlation between morphological differences or proliferation rate and the quality or functionality of the iPSC‐astrocytes. Notably, we observed a delayed astrocyte differentiation (defined by iPSC‐like morphology during the first 1 or 2 weeks during differentiation) in iPSC lines reprogrammed using Sendai vectors compared to those reprogrammed using lentiviral or episomal vectors. However, additional experiments featuring systematic side‐by‐side comparisons would be necessary to further validate this observation.

3.2. Expression of Functional Astrocyte Markers Confirms Maturation of iPSC‐Derived Astrocytes

To assess the purity and cell‐lineage specificity of the iPSC‐derived astrocyte cultures, we performed flow cytometry on DIV 35 iPSC‐astrocytes (Figure 3A). Flow cytometry analysis showed differences in astrocyte size, accounting for morphological astrocyte heterogeneity as previously described (Khakh and Deneen 2019; Endo et al. 2022). In addition, the astrocytes were highly viable (92%) and showed expression of CD49f, indicating successful differentiation towards the astrocyte lineage (Barbar et al. 2020). We observed a subpopulation that was negative for CD44 (29%), which further demonstrates cellular heterogeneity. Seventy percent of the iPSC‐astrocytes co‐expressed CD44 and CD49f at DIV 35.

FIGURE 3.

Molecular characterization of iPSC‐astrocytes confirms astrocyte maturation and identity. (A) Flow Cytometry data, including viability and unstained control, showing co‐expression of CD44 and CD49f by 70% of the iPSC‐astrocyte population. (B) Relative expression of NFIA, VIM, CD44, GFAP and AQP4 by qPCR in DIV 20, DIV34, DIV 42 and DIV 111 iPSC‐astrocytes (iPSC‐line 1). Expression of genes was normalized for the average expression of GUSB, ERCC1 and PPIA (n = 6). (C) Representative images of immunostaining for astrocyte markers EAAT1 (red), Nestin (white), PDLIM7 (red), ALDH1L1 (white), GLUD1 (red), AQP4 (white), GFAP (red) and CD44 (white) in DIV 45 and DIV 111 iPSC‐astrocytes. All pictures were taken at the same magnification (scale bar = 50 or 10 μm in the zoomed image).

To track the development of iPSC‐astrocytes over time, we performed qPCR for NFIA, VIM, CD44, GFAP, and AQP4 at various time points: DIV 20, DIV 34, DIV 42, and DIV 111 (Figure 3B). As expected, we observed increased expression of NFIA and VIM over time. Expression of CD44 peaked at DIV 34, after which it decreased. GFAP expression increased for the first three timepoints but decreased at DIV 111. Interestingly, we find that AQP4 expression decreased over time. This reduction in RNA is accompanied by more abundant AQP4 protein at early timepoints localized throughout the whole cell, while during maturation and therefore at later timepoints, AQP4 levels are decreased as it is localized and limited to the astrocytic end feet (Nielsen et al. 1997; Frigeri et al. 1995; Mader and Brimberg 2019), as observed by immunofluorescence (Figure S5).

We further characterized DIV 45 and DIV 111 iPSC‐astrocytes using immunofluorescence of multiple astrocyte specific markers (Figure 3C). We observed the astrocytes to be positive for Nestin and ALDH1L1 at both time points. EAAT1, GLUD1, and PDLIM7 levels increased over time. PDLIM7 also specifically localized to the actin filaments at DIV 111. Similar to what we observed by qPCR, GFAP and CD44 levels were higher at DIV 45 compared to DIV 111. As mentioned before, AQP4 was present throughout the whole cell body in most DIV 45 astrocytes, while at DIV 111 AQP4 was mostly limited to the astrocytic end feet (Figure S5).

3.3. RNA Sequencing Confirms Astrocytic Identity of iPSC‐Astrocytes

As far as we know, all in vitro iPSC‐astrocyte differentiation protocols in the literature either use ectopic gene overexpression or first differentiate the iPSCs towards NPCs or progenitor‐like cells e.g., by dual SMAD inhibition, which are differentiated towards astrocytes in a second step. Our protocol did not include such an intermediate differentiation step, instead we directly differentiated iPSCs to astrocytes. To assess the molecular similarity between iPSC‐derived astrocytes, NPC‐derived astrocytes, and primary human astrocytes, as well as other brain cell types, we performed bulk RNA sequencing for biological triplicates of five different iPSC‐astrocyte cultures and three different NPC‐astrocyte cultures (Lendemeijer et al. 2024; Shi et al. 2012). Additionally, we included previously published RNA sequencing data of post‐mortem fetal and adult astrocytes, along with other CNS cell types (Zhang et al. 2016). Principal component (PC) analysis showed that the iPSC‐astrocytes and NPC‐astrocytes clustered closely together (Figure 4A,B), demonstrating minimal differences between direct differentiation from iPSCs towards astrocytes and astrocyte differentiation including an intermediate NPC stage. While the first PC separated the samples based on their origin (in vitro or post mortem) (Figure 4A), the second and third PCs segregated the samples bason on their cell type. So all astrocytes, irrespective of their origin, clustered more closely together compared to the other brain cell types (Figure 4B). Hierarchical clustering of the samples based on PCs‐derived distances further supported this observation, grouping post‐mortem and in vitro astrocytes closer together in comparison to neurons, oligodendrocytes and myeloid cells (Figure 4C). We observed that astrocyte‐specific genes such as VIM, GJA1, CD44, PDLIM7, TNC, APOE, GLUD1, GLUL and NFIX are highly expressed by both the iPSC‐astrocytes and NPC‐astrocytes, while neuronal genes (DCX, NGN2 and RBFOX3) and oligodendrocyte genes (SOX10 and MBP) are lowly expressed or not detected (Figure 4D). The expression of iPSC genes (OCT4, SSEA4) were not detected. However, other astrocyte‐specific genes, such as GFAP and S100B and glutamate transporter genes such as SLC4A4 and SLC1A2 were expressed at relatively low levels by the iPSC‐astrocyte monocultures. Together with the relatively high expression of genes associated with gliogenic commitment and early astrocyte states (e.g., VIM, CD44 and NFIX), these transcriptional patterns suggest that the DIV 35 iPSC‐derived astrocytes generated using this protocol resemble developing or immature astrocytes rather than fully mature astrocytes. This interpretation is further supported by their clustering with fetal astrocyte reference datasets in the transcriptomic analysis.

FIGURE 4.

RNA sequencing analysis of iPSC‐astrocytes and NPC‐astrocytes compared to post‐mortem astrocyte samples. Bulk RNA sequencing of triplicates of five different iPSC‐astrocyte cultures and three different NPC‐astrocyte cultures compared to post‐mortem astrocyte gene expression datasets. Principle component (PC) analysis was performed of which (A) PC1 and PC2 showed separation based on origin and (B) PC2 and PC3 showed separation based on cell type identity, showing clustering of iPSC‐astrocytes and NPC‐astrocytes with post‐mortem astrocyte samples. (C) Hierarchical clustering based on PCs‐derived sample distances. (D) Heatmap gene expression by the iPSC‐astrocytes and NPC‐astrocytes, indicating high expression of astrocyte genes and low expression of neuronal and oligodendrocyte genes.

3.4. Functional Characterization of iPSC‐Astrocytes Confirms Their Ability to Take Up Glutamate and Form Gap Junctions in Monoculture

One of the main functions of astrocytes is to take up excessive glutamate from the synaptic cleft. To assess this functionality, we measured the percentage of glutamic acid uptake by DIV 35 monoculture iPSC‐astrocytes 3 h after adding glutamic acid to the medium. We assessed 10 different iPSC‐astrocyte cultures and one rat astrocyte culture (Figure S6). Three iPSC‐astrocyte cultures showed 20%–30% uptake of the additional glutamic acid, which is 2–3 times higher compared to the rat astrocytes. Four other iPSC‐astrocyte cultures showed 5%–10% uptake, while three iPSC‐astrocyte cultures showed only a minimal uptake of glutamic acid. We continued further functional characterization with iPSC‐line 1 astrocytes, as they showed the most effective uptake of glutamic acid (Figure 5A).

FIGURE 5.

Functional characterization of iPSC‐astrocytes confirms the ability to take up glutamate and communicate. (A) The percentage of glutamic acid uptake by monoculture iPSC‐line 1 astrocytes (n = 4) was measured and compared to one rat astrocyte culture (n = 6). (B) Astrocyte connectivity was assessed by injection of Biocytin (indicated by arrow) using single‐cell patch‐clamping of one iPSC‐astrocyte in monoculture, followed by immunolabeling with a Streptavidin 568 conjugated antibody. (C) Astrocytic Ca²⁺ signaling was measured upon incubation with the Fluo‐8‐AM dye, by 10 min recording of which the first 5 min were baseline (n = 20) followed by a 50 μM glutamic acid stimulus (n = 20). Astrocytic Ca²⁺ signaling was calculated as area under the curve of the calcium waves. Paired t‐test showed significantly increased activity upon glutamic acid stimulus with p < 0.0001, (scale bar = 50 μm). (D) The astrocyte secretome was analyzed using Olink proteomics in triplicate of medium derived from astrocytes (iPSC‐line 1). Data were first normalized for the interplate control, and accordingly fold change for each protein was calculated relative to baseline medium. Proteins with > 1 log fold change are shown.

To facilitate rapid communication among astrocytes, they are interconnected through gap junctions, forming what is known as the astrocytic syncytium or astrocyte connectome. Previous in vivo studies have delineated the connectivity of astrocytes in the brain (Boal et al. 2021; Cooper et al. 2020). To evaluate the capacity of our iPSC‐astrocytes in monoculture to establish gap junctions and interconnect, we utilized biocytin, a cellular tracer, which was injected into one astrocyte via single‐cell path‐clamping. The wide spreading of the biocytin dye demonstrated the ability of iPSC‐astrocytes to establish connections (Figure 5B).

Astrocytic communication often involves Ca²⁺ signaling. Various stimuli such as ATP or glutamic acid can trigger astrocytic Ca²⁺ signaling, leading to the release of gliotransmitters (Wang et al. 2009). To assess the astrocytic Ca²⁺ signaling ability of our iPSC‐astrocytes in monoculture, we employed Fluo‐8‐AM labeling. We first recorded baseline activity followed by stimulation with glutamic acid (Figure 5C). At baseline, most astrocytes were active, of which some astrocytes demonstrated synchronized calcium waves. Upon stimulation with glutamic acid, astrocytic Ca²⁺ signaling exhibited a significant increase, confirming their ability to rapidly respond to stimuli.

3.5. iPSC‐Astrocytes Secrete Growth Factors Involved in Neuronal and Glial Development

The importance of astrocyte‐secreted proteins on the development, maturation and function of neurons and synapses is well described in the literature (Chung et al. 2015; Jha et al. 2018). These proteins support neuronal function and development, but are also crucial for the development and maturation of astrocytes and other glial cell types. To assess the astrocyte secretome we performed Olink proteomics on the supernatant of iPSC‐astrocytes in monoculture (Figure 5D). We observed a relatively high secretion of proteins affecting neuronal function and development, such as MCP‐1, IL‐6, HGF, TWEAK, and VEGFA (Okabe et al. 2020; Yepes 2007; Desole et al. 2021; Erta et al. 2012; Yao and Tsirka 2014). In addition, some of the highly secreted proteins such as IL‐6 and CSF‐1 are known to be important for both microglia and oligodendrocyte development (Filipovic and Zecevic 2008; Easley‐Neal et al. 2019). LIF, an important growth factor known to be involved in astrocyte development (Rose‐John 2018; Bonni et al. 1997), was also highly secreted by the iPSC‐astrocytes. Interestingly, in some astrocyte differentiation protocols, LIF is described as one of the main drivers of astrocyte differentiation (Lendemeijer et al. 2024; Fukuda et al. 2007).

3.6. Optimization of the Protocol to Co‐Culture iPSC‐Astrocytes Together With iNeurons

Our primary motivation to develop a new astrocyte differentiation protocol was to establish an all‐human astrocyte‐neuron co‐culture, in which astrocytes functionally support neuronal development. We cultured iNeurons according to our previously published protocol (Frega et al. 2017) and assessed the ability of the iPSC‐astrocytes to support neuronal development. Our key criteria for evaluating the co‐culture protocol included reproducibility as well as high‐quality neurons, defined by fine dendritic outgrowth and no neuronal clustering. Immunofluorescent imaging and MEAs served as our primary measures for assessing co‐culture viability and functionality.

Unfortunately, utilizing the standard co‐culture protocol (Frega et al. 2017) did not result in viable and reproducible co‐cultures (Figure S7). In order to improve the viability of the co‐cultures, the following adjustments to the protocol were tested: varying concentrations of doxycycline; different timing of adding FCS as well as different types of FCS; various types of medium during plating of the astrocytes; addition of RevitaCell during plating of the astrocytes; withdrawal of AraC. The effect of these adjustments on the viability of both cell types was assessed using immunofluorescent imaging. Upon implementation of the adjustments into the protocol (addition of RevitaCell while plating of the astrocytes; lower doxycycline concentration when astrocytes are added; standard FBS replaced by AM FBS to lower the endotoxin concentrations), immunofluorescent imaging confirmed significantly improved viability of both the iNeurons and iPSC‐astrocytes in the co‐cultures (Figure S7).

To investigate whether these iNeurons are also active and form synchronized neuronal networks when co‐cultured with iPSC‐astrocytes, we measured spontaneous neuronal activity using MEAs. Synchronized neuronal network activity, which is defined by NBs, is generally considered a characteristic of mature in vitro neuronal cultures (Cerina et al. 2023; Mijdam et al. 2023; Klein Gunnewiek et al. 2020; Linda et al. 2022; Mossink et al. 2021). Therefore, while assessing the co‐culture protocol, the number of NBs was considered the most important parameter. We measured neural activity of the iPSC‐line 1 co‐cultures on MEAs including four biological replicates. While the appearance of NB activity was considered positive, the NB frequency was highly variable across the four different replicates (Figure S7). We therefore implemented additional adjustments to the protocol (neuron: astrocyte ratio 2:1 instead of 1:1; DIV 42 instead of DIV 35 iPSC‐astrocytes; addition of FBS on DIV 3 instead of DIV 10 to promote astrocyte survival) and assessed these co‐cultures on the MEA. Upon the new adjustments, only 50% of the biological replicates showed NB activity, which also showed varying NB frequency and NB duration (Figure S7). Upon implementation of the final adjustments (adding iPSC‐astrocytes at DIV 6 instead of DIV 3; more details in Methods), all four biological replicates showed a consistent number of NBs and NB duration (Figure S7).

3.7. iPSC‐Astrocytes Support Neuronal Development, Maturation and Synapse Formation When Co‐Cultured With iNeurons

Upon validating the iPSC‐astrocytes in monoculture, we assessed whether the iPSC‐astrocytes also functionally support neuronal development when co‐cultured with iNeurons. Our key criteria for evaluating the co‐culture protocol included reproducibility as well as high‐quality neurons, defined by fine dendritic outgrowth and no neuronal clustering. To confirm the reproducibility of the iPSC‐astrocytes within the co‐cultures, we tested multiple iPSC‐lines across different batches and assessed these co‐cultures using immunocytochemistry (Figure 6A). The overall viability and quality of both the iNeurons and iPSC‐astrocytes in all co‐cultures were good. The neuronal morphology was comparable to neurons co‐cultured with rodent astrocytes (Frega et al. 2017; Mijdam et al. 2023; Klein Gunnewiek et al. 2020; Linda et al. 2022; Mossink et al. 2021), showing fine dendritic outgrowth and relatively small nuclei, with no neuronal clustering. Although we observed healthy astrocytes for all different co‐cultures, some morphological differences of the astrocytes were observed between iPSC‐lines and batches. For example, the astrocytes of iPSC‐line 1 in batch 2 showed a heterogenous mix of naïve and potentially reactive (star‐shaped) astrocytes, which we therefore consider a co‐culture of excellent quality. In contrast to the astrocytes of iPSC‐line 1 in batch 1, which are relatively flat and morphologically similar to most (DIV 42) monoculture astrocytes, the astrocytes in the second batch clearly show increased star‐shaped morphology (as indicated by an increased number of astrocytic protrusions) compared to monocultures. However, the differences in astrocytic morphology did not seem to affect neuronal morphology between the two batches. Interestingly, a lower survival rate of astrocytes also did not necessarily result in decreased neuronal survival or quality, as shown by the co‐cultures with astrocytes of iPSC‐line 7. iPSC‐line 7 showed lower astrocyte survival in both batches indicating that this could be line specific, and possibly solved by plating a higher number of astrocytes to compensate.

FIGURE 6.

Immunofluorescence of iPSC‐astrocyte and iNeuron co‐cultures. (A) Representative images of immunostaining for Vimentin (red) and MAP2 (white) of two independent batches of DIV 21 co‐cultures with iNeurons (iPSC‐line 1) together with iPSC‐astrocytes generated from iPSC‐line 1, iPSC‐line 4, iPSC‐line 6 and iPSC‐line 7. All pictures were taken at the same magnification (scale bar = 50 μm). (B) Representative images of immunostaining for Vimentin (red) and MAP2 (white) of DIV 21 co‐cultures with iNeurons (iPSC‐line 1) together with iPSC‐astrocytes generated from iPSC‐line 1, showing interaction of the astrocytic end feet with the neuronal dendrites. All pictures were taken at the same magnification (scale bar = 10 μm). (C) Representative images of immunostaining for Vimentin (red), MAP2 (white) and Synapsin (green) of two independent batches of DIV 21 co‐cultures with iNeurons (iPSC‐line 1) together with iPSC‐astrocytes generated from iPSC‐line 4. All pictures were taken at the same magnification (scale bar = 20 μm). White arrows indicate synapsin puncta. One‐way ANOVA of the number of synapsin puncta per 10 μm neuronal dendrite was not significantly different between the two independent co‐culture batches (n = 15 neurons).

In addition, we observed arborizations of the astrocytic end feet, interacting with their surrounding neurons (Figure 6B). To investigate whether the iPSC‐astrocytes also support synapse formation, we quantified the number of synapses within the co‐cultures using immunofluorescent labelling of Synapsin (Figure 6C). We observed no significant differences in the number of Synapsin puncta between two different co‐culture batches. The number of Synapsin puncta (~3 puncta/10 μm) is in the same range as observed for the standard co‐cultures with rodent astrocytes (Mijdam et al. 2023; Klein Gunnewiek et al. 2020; Linda et al. 2022; Mossink et al. 2021).

3.8. iPSC‐Astrocytes Support Robust Synchronization of the Neuronal Activity of iNeurons on MEA

Immunofluorescent imaging was used to confirm that the iPSC‐astrocytes support neuronal development, maturation, as well as synapse formation when co‐cultured with iNeurons. As described previously, to assess whether the iNeurons in these co‐cultures also form synchronized neuronal networks, MEAs were used to measure spontaneous neuronal activity. Extensive testing of numerous adjustments to the protocol (Figure S7), assessed using the MEAs, ultimately led to the development of a successful co‐culture approach (see Methods section for details). In summary, Ngn2 stable iPSCs were differentiated towards neurons by addition of doxycycline. In contrast to the protocol using rodent astrocytes, iPSC‐astrocytes were added at DIV 6 (of the neuronal differentiation), in a 2:1 ratio instead of 1:1. Additionally, at DIV 6 the cells were cultured in the presence of RevitaCell to promote iPSC‐astrocyte survival and with AM FCS instead of the standard FCS used previously (Figure 7A).

FIGURE 7.

MEA analysis confirming neuronal network activity in iPSC‐astrocyte and iNeuron co‐cultures. (A) Schematic representation of the protocol to co‐culture iPSC‐astrocytes with iNeurons. (B) Raster plots of the four different wells of the DIV 23 and DIV 37 co‐culture with iNeurons (iPSC‐line 1) together with iPSC‐astrocytes (scale bar = 50 s). All four wells show consistent network burst frequency at DIV 37. In addition, mean firing rate, number of bursts, number of network bursts, and network burst duration of the four wells at DIV 21, 30, 37, 44, and 51 are shown. (C) Representative 60‐s whole‐cell voltage‐clamp recording of spontaneous excitatory postsynaptic currents (sEPSCs) of DIV 35 co‐culture with iNeurons (iPSC‐line 1) together with iPSC‐astrocytes, with 90 ms zoom‐in of an sEPSC. Quantification of sEPSC burst frequency, sEPSC amplitude, and sEPSC frequency (n = 8) is shown.

Using the final co‐culture protocol, the iNeurons in all four biological replicates showed a consistent number of NBs and NB duration (Figure 7B). Although the NB frequency changed over time, similarly as described for co‐cultures with rodent astrocytes, we observed that the number of NBs within the biological replicates for each time point is consistent. Other parameters such as the number of bursts and the mean firing rate show higher variability within biological replicates. However, some of these parameters are generally considered to be more variable across biological replicates (Mossink et al. 2021). Interestingly, the network activity of the co‐cultures with iPSC‐astrocytes is delayed (starting at DIV 30) compared to co‐cultures with rodent astrocytes (starting at DIV 16/23). This is in line with studies showing delayed neuronal development in humans compared to rodents and the great apes (Zeiss 2021; Marchetto et al. 2019; Benito‐Kwiecinski et al. 2021; Gomez‐Robles et al. 2024; Charvet and Finlay 2018). Moreover, using whole‐cell voltage‐clamp recordings we recorded spontaneous excitatory synaptic currents at DIV35. All recorded neurons received synaptic input and showed burst activity (Figure 7C).

Finally, similar to what we did for the monocultures, we assessed the consistency of neuronal co‐cultures on MEA across four independent researchers (B1–B4) from two different laboratories (Figure 8). All cultures were generated from control neurons and control astrocytes, although each researcher used different control iPSC lines. Network activity was detected as early as DIV 35 in B2 (iPSC‐line 48), whereas synchronization emerged at DIV 42 in B1 (iPSC‐line 35) and B3 (iPSC‐line 1) and even later at DIV 58 in B4 (iPSC‐line 53). Across all cultures, the number of network bursts generally increased during differentiation. Over time, network burst duration increased in B1, decreased in B2, and remained stable in B3 and B4, likely reflecting differences in glutamate receptor organization within these cultures. Importantly, despite some wells in B1 and B4 showing limited network bursting, recordings were consistent across cultures within each batch. Thus, while some variability in individual parameters was observed, we consider the current protocol reproducible and suitable for future disease‐phenotyping studies.

FIGURE 8.

MEA analysis of iPSC‐astrocyte and iNeuron co‐cultures by four independent researchers across two different labs. Representative raster plots of the four different batches (referring to neuronal cultures of different control iPSC lines cultured by independent researchers) of the DIV 35, DIV 42 and DIV 58 co‐culture with iNeurons together with iPSC‐astrocytes (scale bar = 50 s). In addition, mean firing rate, number of bursts, number of network bursts and network burst duration of the four wells at DIV 35, DIV 42 and DIV 58 are shown. n = 12 for B1; n = 8 for B2; n = 8 for B3; and n = 24 for B4.

3.9. Removal of FBS From iPSC‐Astrocyte iNeuron Co‐Cultures Reduces Variability Across and Within Neuronal Networks

As previously noted by Liddelow et al. (2017) and White et al. (2024), the use of FBS in culture medium can induce a reactive astrocyte phenotype. We therefore explored strategies to eliminate FBS from the differentiation medium. Complete removal of FBS from DIV 0 onward resulted in poor survival, but withdrawing FBS the day after the first split (approximately DIV 6) improved astrocyte differentiation. To confirm that iPSC‐astrocytes retained their identity following FBS withdrawal, we assessed expression of key astrocyte markers by RT‐qPCR. We found that VIM and NFIA were expressed at comparable levels in −FBS and +FBS iPSC‐astrocytes, and both markers were enriched relative to undifferentiated iPSCs or iPSC‐derived microglia, confirming that astrocyte identity is preserved in the absence of FBS (Figure S8A). Notably, we observed higher expression of the glutamate transporter SLC1A3 in −FBS iPSC‐astrocytes, which may reflect a shift towards a more homeostatic astrocyte state (Figure S8A). Further, we show that FBS withdrawal did not impair the formation or survival of neuron–astrocyte co‐cultures (Figure S8B).

To systematically evaluate whether removal of FBS affects output measures such as neuronal network activity, we conducted a meta‐analysis across 20 MEA plates from four independent researchers, yielding over 3000 recordings with a near‐equal distribution of FBS conditions (55.5% −FBS and 44.5% +FBS) (Figure 9A). Variance partition analysis revealed that genotype was the dominant source of variance in electrophysiological parameters, exceeding the contribution of researcher identity (Figure 9B). Factors such as well position (border or not) and FBS use explained comparatively little variance in neuronal network parameters, suggesting that FBS removal does not broadly disrupt network activity.

FIGURE 9.

Removal of FBS from iPSC‐astrocyte and iNeuron co‐cultures does not alter neuronal network development and function, but reduces variance across wells and within recordings. (A) Schematic overview of meta‐analysis covering 3298 individual multi‐electrode array (MEA) recordings across 20 MEA plates, 14 developmental timepoints and 4 independent researchers. (B) Variance partitioning analysis showing the contribution of each covariate to MEA parameter variability, border = wells at the outside of the 48‐well MEA plates, DIV = days in vitro, inhibitory = presence of inhibitory neurons in culture, MG = presence of microglia in culture. (C) Principal component (PC) plots showing days in vitro (DIV), Network synchrony measured by Area under normalized cross‐correlation, Researcher, and Genotype, N = 20 batches, n = 3298 recordings. (D) PC plot of FBS condition (left) and mean squared distance as a measure of distance in PC space, unpaired t‐test. (E) Volcano plot showing the effect of FBS use on MEA parameters estimated using a mixed‐effects regression model controlling for genotype, MG, DIV, inhibitory status, researcher, plate ID, and border effects. (F) Quantification of the standard deviation of the number of spikes per burst and the mean firing rate (Hz) per FBS condition, N = 20, n = 1830 recordings in −FBS, n = 1468 recordings for + FBS, unpaired t‐test.

Principal component (PC) analysis demonstrated that recordings follow a time‐dependent trajectory in electrophysiological space, driven by a progressive increase in network synchronization over days in vitro (Figure 9C). Consistent with the variance partition results, clustering in PC space was predominantly governed by genotype rather than by researcher or FBS condition, indicating that networks matured along a comparable developmental trajectory regardless of FBS use. Notably, however, the spread of recordings in PC space was significantly lower in the −FBS condition (Figure 9D), suggesting that removal of FBS reduces well‐to‐well variability in electrophysiological profiles.

To isolate the effect of FBS while correcting for all remaining covariates, we applied a linear mixed‐effects regression model across all 73 MEA output parameters (Figure 9E). Only three out of 73 parameters reached statistical significance (FDR‐adjusted p value < 0.01), and all three were measures of within‐recording variability rather than general activity levels. Specifically, the standard deviation of the number of spikes per burst was significantly reduced in −FBS networks (Figure 9F), indicating that FBS removal decreases the variability of bursting behavior within individual recordings. Core network parameters, such as mean firing rate, were not significantly altered (Figure 9F), reinforcing the conclusion that FBS removal does not impair neuronal network function but instead contributes to reduced variability both across and within wells.

3.10. Final Adjustments to the Protocol

The results presented in this study were obtained with astrocytes differentiated according to the methods described earlier. However, during protocol optimization and through collaboration with several other laboratories, we identified additional adjustments that improve both the efficiency and quality of astrocyte differentiation and improve the quality of co‐cultures with iNeurons.

We additionally improved astrocyte survival by shortening the astrocyte splitting procedure. Instead of dissociating astrocytes to single cells with TrypLE, washing in DMEM/F12 by centrifugation, and resuspending in medium before replating, we now use a simplified approach: a reduced volume of TrypLE (e.g., ~300 μL for a six‐well plate well and ~1 mL for a T75 flask) is added, incubated at room temperature for several minutes, and then directly diluted with astrocyte medium before replating into a larger volume of the same medium.

Finally, we addressed frequent attachment problems encountered during the first split, specifically onto T25 flasks. Instead of transferring cells directly into a T25 flask, we now transfer all cells from one full six‐well plate well into three wells of a new six‐well plate at a 1:3 ratio. Once these wells reach full confluency, they are dissociated and combined into a single T75 flask. This modification has improved attachment and overall culture stability during the early stages of differentiation.

4. Discussion

We developed a protocol for iPSC‐derived astrocytes that can support neuronal development when co‐cultured together with iNeurons. In contrast to other protocols, which either use ectopic gene expression or intermediate NPC stages, our protocol directly differentiates iPSCs towards functional astrocytes using only commercial AM. The protocol has been validated in a total of 64 different iPSC lines across 10 independent laboratories, to varying extents, including lines carrying disease‐linked genetic variants highlighting its relevance for disease modeling studies. Here we have described a detailed characterization of a subset of these iPSC‐astrocytes cultures both in monoculture as well as in co‐culture together with iNeurons.

4.1. iPSC‐Astrocytes in Monoculture

Morphological differences defined by astrocyte volume, morphological complexity and proliferation rate were observed between the different iPSC‐lines, both across different differentiation batches as well as within batches (Figure 1C). Nevertheless, these morphological differences are consistent for each iPSC‐line across different batches, demonstrating minimal batch‐to‐batch variability. So the morphological differences between iPSC‐lines most likely originate from genetic heterogeneity rather than environmental factors. As mentioned before, a potential factor influencing differentiation efficiency of the astrocytes (as immature astrocytes tend to grow faster than mature astrocytes) could be the method used for reprogramming of the iPSC‐lines. Differences in reprogramming efficiency between the different vectors (Sendai, episomal and retroviral vectors) are known (Cieslar‐Pobuda et al. 2017; Schlaeger et al. 2015; Hu 2014). For example, reprogramming efficiency was shown to be less efficient using retroviral vectors compared to episomal and Sendai virus vectors (Trevisan et al. 2017), which could potentially also affect astrocyte differentiation efficiency. However, it is also possible that the observed differences are more closely related to the quality of the resulting iPSC lines, specifically, how completely they reprogram epigenetically to a pluripotent state, rather than to the reprogramming method itself. Additional research is needed to further investigate these possibilities.

Due to the significance of astrocytic diversity in the brain (Oberheim et al. 2006), it was anticipated that heterogenous iPSC‐astrocyte cultures would yield superior astrocytes compared to more homogeneous iPSC‐astrocyte cultures. This superiority could manifest in various ways, such as the expression of a wider array of astrocytic genes or the secretion of more astrocyte‐specific proteins. Surprisingly, no discernible correlation between the heterogeneity of the astrocyte monocultures (or other morphological differences) and the functionality or quality of the co‐cultures was observed. Despite the acknowledged importance of astrocytic diversity in in vivo brain development (Oberheim et al. 2006), the limited effect of the astrocyte monoculture heterogeneity on the co‐culture quality suggests that such diversity may not be as crucial for in vitro neuronal network development. However, in this study the differences were mainly based on morphological heterogeneity and proliferation rate. Conducting more comprehensive analyses, such as single‐cell RNA sequencing, could potentially elucidate the correlation between these morphological disparities and astrocytic function.

While our analyses consistently demonstrate efficient astrocyte specification, some data suggests that the generated cells likely represent astrocytes at an early or intermediate stage of maturation. For example, the presence of CD49f⁺ and CD44⁺ populations, sustained expression of vimentin, and detectable Nestin expression are features that are associated with immature or developing astrocytes. It should also be noted that most of the functional assays in this study were performed at relatively early differentiation stages (approximately DIV35–DIV45), which therefore further reflect an immature astrocyte phenotype. Overall, the generation of fully mature astrocytes in vitro remains challenging, particularly in monoculture systems where key environmental signals provided by neuron–astrocyte and glia–glia interactions are absent. Despite this limitation, immature iPSC‐derived astrocytes are widely used and remain highly valuable models for studying early astrocyte biology and disease‐associated phenotypes, as well as for compound or therapeutic screening due to their relatively short differentiation timelines and scalability. For studies requiring more advanced maturation stages, extended culture periods or co‐culture with neurons may further promote astrocyte maturation.

4.2. iPSC‐Astrocytes in Co‐Culture With iNeurons

We also assessed the ability of the iPSC‐astrocytes to support neuronal (network) development when co‐cultured with iNeurons. Considering that these protocols are developed for disease modeling and phenotyping, our most important criteria was reproducibility. As a starting point our “standard” co‐culture protocol (Frega et al. 2017) was tested, using immunofluorescent imaging and MEA as the main readouts to assess the quality and functionality of the co‐cultures. Upon testing multiple adjusted versions of the protocol, the final co‐culture protocol described here is currently the most optimal for both the iPSC‐astrocytes and iNeurons and considered reproducible. Some of these adjustments significantly improved the quality of the co‐cultures. One of them was the use of FBS from the AM kit instead of regular FBS, most likely due to the lower concentrations of endotoxins. Moreover, addition of FBS to the medium at the moment of astrocyte plating instead of 7 days after astrocyte plating, significantly improved the viability of the iPSC‐astrocytes in co‐culture. However, our latest adjustment suggests that complete withdrawal of FBS from the astrocyte differentiation and the co‐culture further reduces variability between and within networks. DIV 42 was also shown to be the most optimal time point for the iPSC‐astrocytes to co‐mature with the iNeurons. However, given the differences between the iPSC‐lines during astrocyte differentiation, the most optimal timepoint of plating might vary between the different iPSC‐lines. Lastly, due to the increased cell size of the iPSC‐astrocytes compared to rodent astrocytes in monoculture, possibly resulting in increased stress due to limited space, different ratios of neurons to astrocytes were assessed. Changing the 1:1 neuron to astrocyte ratio to 2:1 ratio greatly improved the quality of the co‐cultures.

In contrast to other in vitro studies (Vasile et al. 2017; Lendemeijer et al. 2024), the number of synapses in co‐cultures with iPSC‐astrocytes is similar to co‐cultures with rodent astrocytes. We also demonstrated that when co‐cultured with iPSC‐astrocytes, neuronal network activity started around DIV 30, while previous research has shown that when co‐cultured with ex vivo rodent astrocytes, neuronal network activity starts around DIV 14 (Mossink et al. 2021). In line with literature, this may suggest that human brain development compared to rodent brain development is delayed. However, in order to investigate the effect of astrocytes on the number of synapses or neuronal network activity in rodents versus humans (which is beyond the scope of this study), ex vivo primary astrocytes should not be compared to in vitro iPSC‐derived astrocytes. Especially since neurons and other glial cell types highly influence astrocyte development, both via direct contact (Drago et al. 2017; Matejuk and Ransohoff 2020) but also indirectly by the secretion of metabolites and proteins (Drago et al. 2017; Matejuk and Ransohoff 2020; Nakanishi et al. 2007; Nemes‐Baran et al. 2020; McNamara et al. 2023). Therefore, to assess potential differences between human and rodent astrocytes, either human and rodent iPSC‐astrocytes, or ex vivo primary human and rodent astrocytes should be used.

In addition to the astrocyte‐neuron co‐culture system described here, we very recently developed and validated a complementary protocol for generating iPSC‐derived microglia, both in monoculture and in triculture with iNeurons and the iPSC‐astrocytes (Mordelt et al. 2025). This study demonstrates that our iPSC‐astrocytes promote long‐term maintenance and functional integration of microglia in vitro, thereby enabling more comprehensive human model systems that include neuron‐astrocyte‐microglia interactions.

Although the morphological differences between different batches of the astrocyte monocultures were minimal, significant differences in astrocytic morphology were observed between different batches of co‐cultures using the same iPSC‐line. In one batch the iPSC‐astrocytes co‐matured into a heterogenous mix of naïve and potentially reactive (star‐shaped) astrocytes, while in the other batch the iPSC‐astrocytes appeared to fail to co‐mature and therefore maintained their flat monoculture‐like morphology. It is still unclear how this difference can be explained, but it most likely originates from variables during the iPSC‐astrocyte differentiation, including the procedure of splitting the astrocytes during differentiation. Although the procedure might seem consistent, the difference between splitting a 95% confluent flask at 1:3 ratio compared to splitting a 100% confluent flask in 1:3 ratio, might have had more impact than previously anticipated. Especially given that confluency, cell density and cell–cell contact greatly impact glial cell development (Joseph et al. 2021). In order to improve consistency; instead of splitting at 1:3 ratio, a specific number of cells, for example, one million cells per T75 flask could be re‐seeded. Therefore, the cell number might be more consistent during differentiation, and across different iPSC‐astrocyte batches. Surprisingly, the flat monoculture‐like astrocytes seemed to support the neurons as well as the heterogeneous astrocyte cultures including the star‐shaped astrocytes. This indicates that in co‐culture, the differences in astrocyte morphology defined by, for example, the number of astrocytic protrusions, does not greatly affect neuronal growth and maturation which may suggest that so‐called naïve astrocytes are of more importance in in vitro co‐cultures compared to the reactive star‐shaped astrocytes (Gottipati et al. 2020). However, here the quality of the neurons was solely assessed based on morphology, so potential differences, for example, at gene expression or electrophysiological level should be identified in additional studies.

Relatively little is currently known about the long‐term development of neuronal network dynamics in co‐culture systems consisting of human iPSC‐derived neurons and astrocytes on MEA. Therefore the mechanisms underlying the activity patterns observed in such systems remain incompletely understood and likely require further investigation. In the present study, neuronal network activity increased during the early stages of co‐culture and peaked around DIV37, after which several activity parameters, including mean firing rate and the number of network bursts, gradually declined. Similar developmental trajectories have been reported in other in vitro neuronal culture systems, where neuronal activity increases during early stages of network formation as synaptic connectivity is established and neurons progressively integrate into functional circuits (Chiappalone et al. 2006). As networks mature, activity patterns often undergo refinement, which can lead to stabilization or even reduction of overall firing rates or burst frequency despite increasing network organization. These dynamics have been linked to processes such as synaptic pruning, homeostatic plasticity mechanisms that regulate network excitability, and the progressive maturation of inhibitory signaling within neuronal circuits. In addition, astrocytes are known to play an important role in shaping neuronal network activity through mechanisms including glutamate uptake, potassium buffering, and the modulation of synapse formation and elimination (Allen and Eroglu 2017). The progressive development of neuron–astrocyte interactions in the present co‐culture system may therefore also contribute to the refinement of network activity over time. Finally, long‐term in vitro culture conditions may influence network dynamics through factors such as metabolic constraints, synaptic scaling processes, or shifts in the balance between excitatory and inhibitory signaling. Together, these factors may potentially contribute to the observed post‐DIV37 decline in activity metrics while still reflecting ongoing maturation and stabilization of the neuronal network.

In addition, astrocytes play a critical role in regulating extracellular glutamate levels through the expression of glutamate transporters such as EAAT1 and EAAT2, thereby preventing excitotoxicity and modulating neuronal activity. However, the efficiency of glutamate uptake can vary between astrocyte populations. Such variability may influence the extent to which astrocytes support neuronal maturation, synaptic development, and neuronal network activity in co‐culture systems. This is particularly relevant in the context of disease modeling, where astrocyte dysfunction in glutamate homeostasis has been implicated in several neurological disorders. While the astrocyte differentiation protocol described here generates astrocytes expressing canonical markers and key functional properties, differences in glutamate uptake efficiency between lines may still arise either due to intrinsic genetic background contributing to the phenotype of disease lines, or due to differentiation variability. Therefore, when applying this protocol in neuron–astrocyte co‐culture systems, the glutamate uptake capacity (in monoculture) should be considered as a potential contributor to variability in neuronal maturation and activity.

4.3. The Effect of Skipping Directed NPC Differentiation During Astrocyte Differentiation

As mentioned before, most in vitro iPSC‐astrocyte differentiation protocols first differentiate the iPSCs towards NPCs, which are later differentiated towards astrocytes. Since this protocol does not include an intermediate NPC differentiation step, we compared the iPSC‐astrocytes to astrocytes derived from NPCs (NPC‐astrocytes) using bulk RNA sequencing. iPSC‐astrocytes and NPC‐astrocytes clustered closely together, confirming minimal differences between these two protocols. Recently, Räsänen et al. (2024) also demonstrated neuronal network activity in iPSC‐astrocyte (NPC‐derived) and iNeuron co‐cultures, showing a similar developmental trajectory of the human co‐culture over time as was shown in this study. This further indicates that skipping directed differentiation into the NPC intermediate stage has minimal effect on in vitro astrocyte development, and subsequent neuronal network development. Although we did not specifically differentiate the iPSCs towards NPCs for expansion using, for example, dual SMAD inhibition, it could still be that using the commercial AM (of which the exact composition is unknown) the iPSCs transit rapidly through an NPC‐like stage or astrocyte precursor cell type and later towards functional astrocytes. Since the astrocyte differentiation protocol used in this study relies on a commercial astrocyte medium (AM) kit with proprietary composition, the precise contribution of individual components cannot be determined and potential changes introduced by the manufacturer could influence the differentiation outcome. However, the formulation is serum‐free, thereby eliminating FBS, which represents one of the major sources of batch‐to‐batch variability in cell culture systems. Defined differentiation media may provide additional flexibility to modulate specific signaling pathways involved in astrocyte specification. As discussed previously, numerous astrocyte differentiation protocols based on defined conditions have been described in the literature (Koskuvi et al. 2022; Perriot et al. 2021, 2018; Tcw et al. 2017; Mulica et al. 2023; Chandrasekaran et al. 2016; Lendemeijer et al. 2024; Voulgaris et al. 2022; de Leeuw et al. 2020; Shaltouki et al. 2013; Jovanovic et al. 2023; Li et al. 2018; Neyrinck et al. 2021; Soubannier et al. 2020; Yeon et al. 2021; Giordano et al. 2023; Szeky et al. 2024; Dittlau et al. 2024; Ryan et al. 2020; Roybon et al. 2013; Jiang et al. 2013; Krencik et al. 2011; Brennand and Gage 2011; Qiu and Caiazzo 2025; Bosco et al. 2024; Berryer et al. 2023; Gupta et al. 2012; Santos et al. 2022; Tchieu et al. 2019), including approaches relying on ectopic expression of transcription factors such as SOX9, NFIA, or NFIB, as well as protocols that generate astrocytes via a NPCs. These systems may complement the present protocol for studies aiming to further dissect the molecular mechanisms underlying astrocyte differentiation.

In conclusion, our results show that direct differentiation of iPSCs into astrocytes using commercial AM results consistently in mature and functional astrocytes. These iPSC‐astrocytes can functionally support neuronal network formation when co‐cultured with iNeurons, showing consistent neuronal network patterns. Although variability between different iPSC lines is acknowledged, batch‐to‐batch variability is minimal. The removal of FBS from the co‐cultures reduced variance in neuronal network activity across and within wells. The results from over 60 iPSC lines have shown that both the iPSC‐astrocyte differentiation protocol as well as the co‐culture protocol are robust and consistent, allowing it to be used for cell‐type specific disease modeling and phenotyping.

Author Contributions

Conceptualization and experimental design: Imke M.E. Schuurmans, Annika Mordelt, and Nael Nadif Kasri. Experiments for optimisation of astrocyte protocol: Imke M.E. Schuurmans, Annika Mordelt, Katrin Linda, Denise Duineveld, Marina P. Hommersom, Lisa Rahm, Emma Dyke, Gijs‐Jan Scholten, Caroline Knorz, Carlos Gonzalez Jimenez, Noelle van Egmond, Astrid Oudakker, Chantal Bijnagte‐Schoenmaker, Eline van de Ven. RNAseq analysis: Sofia Puvogel. Cross‐validation of astrocyte protocol in independent labs: Marta Guevara‐Ferrer, Wei Wei, Samuel Hofmann, Sabine Jung‐Klawitter, Sandra Mojica‐Perez, Jack Parent, Lidia Carotenuto, Sarah Weckhuysen, Nadine Maas, Ype Elgersma, Anita Lygeroudi, Helga E. de Vries, Luc Jordi, Magdalini Polymenidou, Teresa Mitchell‐Garcia, Iulia Dragan, Amalia Dolga, Lot D. De Witte, James Ellis. Contribution of MEA data: Denise Duineveld, Carlos Gonzalez Jimenez, Marina P. Hommersom, Nicky Scheefhals. Writing, original draft: Imke M.E. Schuurmans, Annika Mordelt. Review and editing: Nael Nadif Kasri. Project supervision: Clara D.M. van Karnebeek, Alejandro Garanto, Nael Nadif Kasri.

Funding

This work was supported by the European Joint Programme on Rare Diseases (EJPRD grant no: 825575 awarded to C.D.M.K. and N.N.K.), and a Stem Cell Network Impact Grant (to J.E.). I.M.E.S. is supported by an internal Radboudumc PhD grant provided by the Radboud Institute for Molecular Life Sciences to C.D.M.K. and A.G. M.G.‐F. is supported by a SickKids Restracomp Fellowship.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: (A) Brightfield images of different hiPSC‐astrocyte cultures at DIV 5, ranging from (too) low to (too) high astrocyte densities at DIV 5. Color bar is indicative of acceptable and unacceptable astrocyte densities. All pictures were taken at the same magnification (scale bar = 100 μm). (B) Brightfield images showing astrocytes cultures that are recommended to be split either 1:1 or 1:2 ratio in order to rescue the differentiation and/or proliferation of these cultures. All pictures were taken at the same magnification (scale bar = 100 μm). (C) Brightfield images highlighting the importance of not reusing T75 flasks for splitting, but instead using new flask to prevent insufficient astrocyte growth and the culture of hiPSC‐like colonies (indicated by white arrows). All pictures were taken at the same magnification (scale bar = 100 μm). (D) Brightfield images of two example astrocyte cultures that showed a rapidly decreased proliferation rate within the first week(s). All pictures were taken at the same magnification (scale bar = 100 μm). (E) The left brightfield image shows an astrocyte culture with so‐called delayed astrocyte differentiation, indicated be a remaining hiPSC‐colony, before it has been split. The right brightfield image shows to same astrocyte culture after splitting without Revitacell, which improves astrocyte differentiation and removes the remaining hiPSC‐colonies. All pictures were taken at the same magnification (scale bar = 100 μm).

Figure S2: Brightfield images of DIV 8, DIV 15 and DIV 20 co‐cultures of iNeurons with iPSC‐derived astrocytes (hiPSC line 1), which were cultured either on untreated coverslips (left) or nitric acid pretreated coverslips. Scale bar represents 100 μm.

Figure S3: Brightfield images of 46 (from all 64) different hiPSC‐astrocyte cultures, which were cultured across different batches, by several different researchers across 11 different labs. The astrocytes are around 2.5–6 weeks old. Scale bar represents 100 μm.

Figure S4: (A) Brightfield images of DIV 21 astrocytes derived from hiPSC‐line 1 and hiPSC‐line 3 across four different batches. All pictures were taken at the same magnification (scale bar = 100 μm). (B) Immunostaining of Vimentin (red) by DIV 35 astrocytes derived from hiPSC‐line 1 across four different batches. All pictures were taken at the same magnification (scale bar = 50 μm).

Figure S5: Expression of AQP4 (white) by immunolabeling in astrocytes from hiPSC‐line 1 at DIV 35, 45, 82 and DIV 111. The images show that AQP4 is expressed throughout the whole astrocyte cell body at early stages, while later during differentiation the expression of AQP4 decreases as it localizes towards the astrocytic end feet (indicated by the white arrows). All pictures were taken at the same magnification (scale bar = 50 μm or 10 μm in the zoomed image).

Figure S6: The percentage of glutamic acid uptake by monoculture hiPSC‐astrocytes was measured for 10 different hiPSC‐astrocyte cultures and one rat astrocyte culture.

Figure S7: Schematic representation of the optimization of the co‐culture is depicted. As starting point the “standard” co‐culture protocol was tested. Brightfield imaging, immunofluorescent imaging and/or MEA were used as the main readout to assess the quality of the co‐cultures. Upon testing many adjusted versions of the protocol, the bottom protocol is considered most optimal and reproducible. All pictures were taken at the same magnification (scale bar = 50 μm).

Figure S8: (A) Relative gene expression of VIM, NFIA, SLC1A3, and CSF1R measured by qPCR in iPSC‐astrocytes differentiated with our without FBS and iPSCs and iPSC‐microglia as controls, DIV of astrocyte = 42–50., N = 2–3 differentiations per cell type. (B) Brightfield images of DIV 30 co‐cultures of iNeurons with iPSC‐derived astrocytes (hiPSC line 1), of which the astrocytes were differentiated with FBS (left) or without FBS the day upon the first split of differentiation and during the co‐culture (right). Scale bar represents 100 μm.

Table S1: Overview cell lines.

Table S2: Astrocyte differentiation.

Table S3: NPC differentiation.

Table S4: Co‐culture.

Table S5: Primers.

Table S6: Microscopy setting.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.