“iAssembloids” Probe Glial Roles in Neuron Survival

To better model how neurons fire and interact with glia, some scientists place the cells into a brain or grow them in an organoid. Others have produced little spheres of human neural tissue by combining neurons, astrocytes, and microglia, all derived from induced pluripotent stem cells. Neurons in this system were more active than those in 2D cell culture. As reported January 14 in Neuron, scientists led by first author Emmy Li, now at Stanford University, California, and Martin Kampmann of the University of California, San Francisco, used this new “iAssembloid” system to run CRISPRi-based screens in search of genes important for neuron survival. They found that the kinase GSK3β in neurons, and ApoE4 in astrocytes, both exacerbate oxidative stress, causing neurons to die. Neither gene turned up in the same screen of 2D cultures, demonstrating the importance of a model system that keeps neurons active rather than letting them lie dormant.

  • iAssembloids combine hiPSC-derived neurons, microglia, and astrocytes in 3D spheres.
  • CRISPRi screening revealed genes that affect neuron survival in iAssembloids but not in 2D monoculture.
  • One gene, GSK3β, weakens neuron defenses against oxidative stress.
  • ApoE4 astrocytes cause hyperactivity and oxidative stress in neurons.

This isn’t the first 3D model system involving hiPSC-derived neurons and glia. For example, co-author Erik Ullian of UCSF previously led development of a similar system with neurons and astrocytes (Krencik et al., 2017). Besides adding a cell type, the new system is optimized for rapid production and flexibility, making it perfect for large screening studies, said Kampmann.

 “iAssembloids are a scalable platform bridging the gap between 2D cultures and organoids,” wrote Simon Schäfer at the Technical University of Munich in Germany, who was not involved in the study. “Its scalability for high-throughput CRISPR-based screens holds great potential for advancing research into neurodevelopmental and neurodegenerative disorders.” (Comment below.)

With the new protocols, hiPSC-derived neurons, microglia, and astrocytes self-assembled into stable spheres in about two weeks, each containing around 10,000 cells. Growing a brain organoid can take months. Compared to neurons in 2Dl monocultures, those in iAssembloids were slightly more mature, their gene expression comparable to that of a 17- to 18-week-old fetus. They were also far livelier. Neurons in iAssembloids spontaneously fired more than did their 2D counterparts. Curiously, neurons in spheres that contained more astrocytes were calmer, suggesting the glia soothe hyperactivity. Moreover, the astrocytes in iAssembloids adopted the characteristic star shape found in the brain, rather than the spindly form they typically take in 2D culture.

“It was quite stunning to see the beautiful complex morphology of cell types, such as astrocytes,” Kampmann told Alzforum (image below). “That's just morphology, but it also suggests to us that the system promotes a more physiological state for the different cell types.”

Sphere of Influence. In iAssembloids, microglia (red) form projections (white arrows), while astrocytes (green) assume star shapes typically found in the brain. [Courtesy of Kampmann et al., 2025.]

Scientists have been moving gradually toward 3D cultures that integrate different cell types, believing these models will be more realistic than 2D and monoculture systems. But evidence to support this contention has been lacking, noted Robert Krencik of Houston Methodist Research Institute, who was not involved in the work. Enter a CRISPRi-based screen Kampmann previously used to analyze 2D cultures of iPSC-derived neurons (Kampmann et al., 2019). The scientists ran this screen in the iAssembloids, suppressing thousands of genes to see which ones affect neuron survival. While some appeared to support or harm neurons in both models, others only did so in iAssembloids. The scientists selected 68 of these genes for further analysis.

“This study does provide direct and specific evidence that 3D multicellular cultures significantly differ from monolayer/monoculture,” wrote Krencik (comment below).

One gene that showed up only in the iAssembloid screen is glycogen synthase kinase 3b. The GSK3b enzyme phosphorylates tau and frequently has been linked to Alzheimer’s and other neurodegenerative diseases in years past. Knocking out GSK3B did not affect 2D monoculture neurons, but helped neurons survive in iAssembloids. To find out why, the scientists hunted for proteins that differed between the model systems.

“I just kept doing Western blots, and nothing was working,” said Li. “In my lab notebook, there were unhappy faces on everything.”

Eventually, Li and colleagues discovered that the answer lay in the activity of the neurons. Active neurons produce more reactive oxygen species, but GSK3β prevents the transcription factor nuclear factor erythroid 2-related factor 2 (NRF2) from entering the nucleus. Since NRF2 defends against oxidative stress by switching on a suite of protective genes, GSK3β hamstrings neurons’ ability to survive.

“Kampmann’s team demonstrated that unbiased CRISPRi screening using their 3D neuron-astrocyte-microglia assembled model, or iAssembloid, identified the GSK3β/NRF2 axis as a critical pathway regulating ROS-induced neuronal death,” wrote Doo Yeon Kim and Matthias Hebisch of Massachusetts General Hospital/Harvard Medical School, Boston. They were not involved in the work (comment below).

In the final part of the study, the scientists used their system to investigate cell-specific effects of ApoE4, the ApoE variant that greatly increases a person’s risk for Alzheimer’s disease. Several cell types produce ApoE, but in the brain most is made by astrocytes. To learn what astrocyte ApoE4 might be doing, the scientists compared iAssembloids in which astrocytes expressed ApoE4 with those containing astrocytes that produced ApoE3.

More CRISPRi screens showed that different genes in ApoE4 versus ApoE3 astrocytes modulated neuron survival. ApoE4 astrocytes boosted neuronal expression of genes involved in synaptic transmission and axon development—the machinery of action potentials. This led the scientists to again test for neuronal activity and oxidative stress. Both were higher in the ApoE4-astrocyte iAssembloids.

“We found it very intriguing that whether you have ApoE3 or E4 in astrocytes affects how excitable the neurons are, and also how much oxidative stress they experience in this system,” said Kampmann.

The findings illustrate iAssembloids’ potential as a mix-and-match system, said Kampmann. Hebisch and Kim think it could be useful in translational research. “An approach like this will help identify pathway activities in human disease models that directly reflect changes observed in patient brains,” they wrote.

Of course, there are limits. iAssembloids are uniform balls made to order, without the structural organization and developmental progression that organoids mimic. “If you are interested in those specific processes, then you would pick an organoid model,” said Li. “But if you're just looking at how cells interact on a fundamental level, and you want to do a high throughput screen, then I think the iAssembloid model is for you.”—Nala Rogers

Nala Rogers is a freelance science writer based in Silver Spring, Maryland.

Comments

  1. iAssembloids are a scalable platform bridging the gap between 2D cultures and organoids. In this study, Martin Kampmann’s team combined a 3D co-culture system with a CRISPR-based genetic perturbation screen, revealing GSK3B’s role in modulating the oxidative stress response triggered by neuronal activity. While the model doesn’t fully replicate the brain’s complexity—particularly with astrocytic and microglial cells showing limited diversity —it offers significant advantages over 2D systems. Its scalability for high-throughput CRISPR-based screens holds great potential for advancing research into neurodevelopmental and neurodegenerative disorders.

    View all comments by Simon Schafer

  2. This group has previously published similar studies conducting CRISPR interference-based screens individually on hPSC-derived induced neurons, microglia, and astrocytes. They unbiasedly identified functionally relevant gene targets of single guide RNAs, then followed up with mechanistic studies to validate the findings. For example, those studies identified genes whose expression is significantly correlated with neuronal survival, microglial survival/activation/phagocytosis, and astrocyte inflammation. In this new study, similar cell lines and strategies are utilized, yet the authors perform pioneering CRISPRi screens using multicellular free-floating sphere co-cultures termed ”iAssembloids.” What were the purposes of these studies and what did they reveal?

    One question this study addressed is whether cell co-culture versus monoculture—and whether “3D”/free-floating aggregate versus “2D”/monolayer conditions—identifies molecular and cellular differences that matter for experimental disease modeling. This is important as there is a recent trend in the field to state that 3D cultures are better and more relevant nervous system models in comparison to 2D conditions, however few specific examples support this claim. 

    Indeed, this study found numerous cell type-specific transcripts that are altered in the different conditions. These could be subdivided into categories, such as increased maturation of cells within the 3D co-culture conditions. This is expected as this phenomenon has already been described elsewhere, and cells are undoubtably influenced by each other as well as by distinct culture conditions. However, this study provides robust databases to specifically define these differences, and this will be useful across the scientific community.

    Does this mean researchers should stop using monolayer/monoculture conditions for experiments? No, as there are likely pros and cons for each. Yet this study does provide direct and specific evidence that 3D multicellular cultures significantly differ from monolayer/monoculture. Thus, results of some experiments will depend on the condition used, which could hamper successful translation of basic research findings into medical applications.

    A second question was whether this screening approach can be used to identify underlying mechanisms or pathways that explain differences between 2D monoculture and 3D multicellular culture conditions, focusing on neurons. As usual with this type of profiling experiment, it produced a lot of potential candidates to choose from for validation, so here they focused on a potential link between GSK3B knockdown and the NRF2 antioxidant response pathway. This link is already well known, but not yet investigated in the context of culture environment differences. The authors conclude that neurons in 3D conditions (in the absence of glia) are less healthy because there is also more neuronal activity in this condition, and this higher activity causes reactive oxygen species (ROS). They then provide evidence that neurons are more sensitive to this increased ROS in 3D conditions because there is higher GSK3B activity, and that GSK3B activity inhibits Nrf2-mediated protection from ROS. The conclusions are a bit of a reach, as these claims were not tested very rigorously, though they do suggest that reducing GSK3B (or enhancing Nrf2 activity) may be sufficient to replace the need for glia in these type of culture experiments.

    As another example of the utility of this screening approach, this study identified candidates that affect neuronal survival in co-culture with astrocytes of different APOE isoforms as a model of Alzheimer’s. Again, they found many candidates among different categories that will need to be validated among different models.

    Overall, this study demonstrates various strategies to use this type of unbiased screening approach in various conditions, mainly as proof of principle but also with interesting follow-up validation. Since these studies were conducted with a limited number of biological replicates and very specific methods, it remains to be seen if the conclusions are universal to other culture conditions, for example when using neural organoids that are temporally differentiated over extensive amounts of time, or with primary tissue, or in vivo, etc.

    View all comments by Robert Krencik

  3. This elegant study represents an exciting advance in using sophisticated human 3D cell culture models to investigate disease mechanisms, particularly those related to neuronal death and survival.

    Kampmann’s team demonstrated that unbiased CRISPRi screening using their 3D neuron-astrocyte-microglia assembled model, or iAssembloid, identified the GSK3β/NRF2 axis as a critical pathway regulating reactive oxygen species-induced neuronal death. Interestingly, they discovered that this axis is undetectable in 2D or even 3D neuron monoculture models, highlighting the need for more sophisticated models replicating physiological neural-glial networks.

    Notably, they detected and validated that a voltage-gated calcium channel beta subunit plays a role in APOE4 astrocyte-iAssembloids. A recent computational drug repositioning study identified Bumetanide, a loop diuretic, as a potential modifier of APOE4-associated Alzheimer’s disease pathology, possibly mitigating abnormal brain network activities. It would be exciting to see if these new APOE4 iAssembloid models could be applied to Alzheimer’s disease drug screening (Taubes et al., 2021; Boyarko et al., 2023). 

    While significant technological advancements have been achieved in this study, there are still limitations to these new iAssembloid models. Like many other 3D models, these models struggle to replicate key pathological markers of disease. For instance, APOE4 iAssembloids do not exhibit Ab plaques or neurofibrillary tangles, making it challenging to interpret APOE4 phenotypes within the context of comprehensive Alzheimer’s disease pathology. Furthermore, the acute nature of these models poses additional challenges in recapitulating disease pathology, which typically requires extended timeframes that are often unattainable even with the best chronic cellular models currently available. 

    In collaboration with Winston Hide’s group at Beth Israel Deaconess Medical Center in Boston, our lab recently published a novel bioinformatics platform, IPAA. This platform enables unbiased comparisons of human cellular models for diseased brains based on pathway activities (Naderi Yeganeh et al., 2025). An approach like this will help identify pathway activities in human disease models that directly reflect changes observed in patient brains.

    Nonetheless, many existing models share these technical hurdles. We congratulate Kampmann and colleagues for their achievement and look forward to potential collaborations in the future.

    References:

    Taubes A, Nova P, Zalocusky KA, Kosti I, Bicak M, Zilberter MY, Hao Y, Yoon SY, Oskotsky T, Pineda S, Chen B, Jones EA, Choudhary K, Grone B, Balestra ME, Chaudhry F, Paranjpe I, De Freitas J, Koutsodendris N, Chen N, Wang C, Chang W, An A, Glicksberg BS, Sirota M, Huang Y. Experimental and real-world evidence supporting the computational repurposing of bumetanide for APOE4-related Alzheimer's disease. Nat Aging. 2021 Oct;1(10):932-947. Epub 2021 Oct 11 PubMed.

    Boyarko B, Podvin S, Greenberg B, Momper JD, Huang Y, Gerwick WH, Bang AG, Quinti L, Griciuc A, Kim DY, Tanzi RE, Feldman HH, Hook V. Evaluation of bumetanide as a potential therapeutic agent for Alzheimer's disease. Front Pharmacol. 2023;14:1190402. Epub 2023 Aug 4 PubMed.

    Naderi Yeganeh P, Kwak SS, Jorfi M, Koler K, Kalatturu T, von Maydell D, Liu Z, Guo K, Choi Y, Park J, Abarca N, Bakiasi G, Cetinbas M, Sadreyev R, Griciuc A, Quinti L, Choi SH, Xia W, Tanzi RE, Hide W, Kim DY. Integrative pathway analysis across humans and 3D cellular models identifies the p38 MAPK-MK2 axis as a therapeutic target for Alzheimer's disease. Neuron. 2025 Jan 22;113(2):205-224.e8. Epub 2024 Nov 27 PubMed.

    View all comments by Matthias Hebisch View all comments by Doo Yeon Kim

  4. Summary of the work
    Li et al. developed a novel 3D co-culture model using iPSC-derived neurons, astrocytes, and microglia, and applied it for functional genomic screens to elucidate neuron-glia interactions. The authors first characterize some basic parameters of the model, showing advantages of the iAssembloids compared to 2D monoculture. They then perform a CRISPRi screen on neuronal survival in iAssembloids compared to 2D neuron monocultures, where they find large overlap but also iAssembloid-specific hits. One of those is GSK3β, which has been implicated in several neurodegenerative diseases, and whose knockdown increases survival in 3D iAssembloids, but not 2D monocultured neurons. They find that 3D culture and associated increased neuronal activity lead to an increase of reactive oxygen species in neurons, which in turn leads to increased activity of GSK3β. This causes decreased nuclear translocation of NRF2 and thus decreased expression of its anti-oxidative-stress target genes, eventually promoting neuronal death by ROS-induced ferroptosis. In their culture system, this cascade is mainly triggered by 3D culture, while presence of astrocytes may dampen it by decreasing neuronal activity.

    To investigate an example of neuron-astrocyte interactions in the context of neurodegenerative diseases, they perform another CRISPRi screen in iAssembloids containing neurons and ApoE3 or ApoE4 astrocytes, without microglia. They obtain several hits, which differentially regulate neuronal survival in ApoE3 vs E4 co-cultures, and focus on calcium voltage-gated channel auxiliary subunit beta 4 (CACNB4). Knockdown of CACNB4 leads to increased survival in ApoE3, but not ApoE4 co-cultures, which may be connected to increased neuronal activity and ROS levels in ApoE4 co-cultures. When comparing to a human AD transcriptomic dataset, they find that neuronal clusters with low expression of CACNB4 are less abundant in AD patients with E4/E4 genotype compared to E3/E3 genotype, potentially hinting toward a selective vulnerability of this subpopulation to neurodegeneration in ApoE4/E4 AD patients.

    Novel aspects and highlights
    This study provides a simple multicellular 3D system that can be used for genetic screening approaches. Due to the modularity of the model, cell-type-specific disease contributions can also be investigated, which is difficult in other systems such as brain organoids. To our knowledge, the study is the first one using such a 3D model for a CRISPRi screen, which is a great step forward as previous screens relied on 2D monocultures of different brain cell types, which are physiologically much less relevant.

    The authors use the results of their screen to elegantly confirm the connection between neuronal activity, ROS, GSK3β and NRF2, and propose the involvement of this pathway in neurodegenerative diseases. Moreover, by combining several brain cell types, the authors were able to investigate interactions between them, such as the effect of ApoE4 astrocytes on neuronal survival, a highly relevant aspect of AD research.

    Open questions and future perspectives
    It would be interesting to see more data on how reproducible the model is regarding cell type composition and distribution, and how mature and physiological the cell states are. Do the iAssembloids actually model central aspects of human brain tissue with complex and physiological cellular interactions, morphologies, matrix generation, etc.? Or are they just an advanced assembly of different single-cell types? Investigating this more deeply would not only further validate hits obtained from genetic screens, but also promote studies on neurodegenerative disease such as AD, as presence of non-physiological or activated cells in a model may mask disease-relevant changes. So far, while the co-culture seems to promote similarity with neuronal and glial phenotypes in the brain, neurons in 3D still seem to map quite closely to 2D neurons, and microglia–based on their non-ramified morphology–may still be relatively immature and/or activated, limiting studies on their interactions and reactions to stimuli.

    Future studies will also need to further elucidate whether the (hyper)activity-ROS-GSK3β-NRF2 axis is relevant for neurodegenerative diseases, or rather related to artificial oxidative stress caused by non-physiological cellular states in the 3D co-culture. In applications for disease research, the use of transcription-factor based differentiation may limit transferability, e.g.,, to mutant lines from patients, as the transgenes would need to be inserted. Nevertheless, the study provides first important concepts on how to apply such neuron-glia models to investigate disease—and investigating their interactions is indeed a highly relevant research field for AD, FTD, and beyond. Using lines with different ApoE genotypes provides a first demonstration how disease-relevant mutations can be investigated in this model, while the observed effects of ApoE4 astrocytes on neuronal activity as well as interaction with ApoE4 microglia call for more detailed analyses.

    Taken together, this is a very interesting study, expanding the toolset of the field to apply iPSC-derived brain cell types and their assemblies for screening purposes. It is setting the stage for future studies exploring mechanisms contributing to neurodegenerative disease in human neuron-glia models.

    View all comments by Dominik Paquet View all comments by Julien Klimmt

Make a Comment

To make a comment you must login or register.

References

Mutations Citations

  1. APOE C130R (ApoE4)

Paper Citations

  1. Krencik R, Seo K, van Asperen JV, Basu N, Cvetkovic C, Barlas S, Chen R, Ludwig C, Wang C, Ward ME, Gan L, Horner PJ, Rowitch DH, Ullian EM. Systematic Three-Dimensional Coculture Rapidly Recapitulates Interactions between Human Neurons and Astrocytes. Stem Cell Reports. 2017 Dec 12;9(6):1745-1753. Epub 2017 Nov 30 PubMed.
  2. Tian R, Gachechiladze MA, Ludwig CH, Laurie MT, Hong JY, Nathaniel D, Prabhu AV, Fernandopulle MS, Patel R, Abshari M, Ward ME, Kampmann M. CRISPR Interference-Based Platform for Multimodal Genetic Screens in Human iPSC-Derived Neurons. Neuron. 2019 Oct 23;104(2):239-255.e12. Epub 2019 Aug 15 PubMed.

Further Reading

No Available Further Reading

Primary Papers

  1. Li E, Benitez C, Boggess SC, Koontz M, Rose IV, Martinez D, Dräger N, Teter OM, Samelson AJ, Pierce N, Ullian EM, Kampmann M. CRISPRi-based screens in iAssembloids to elucidate neuron-glia interactions. Neuron. 2025 Jan 9; Epub 2025 Jan 9 PubMed.

Annotate

To make an annotation you must Login or Register.