From Dish to Living Brain: How Lab-Grown Human Microglia Are Rewriting Neuroimmune Research

Deep inside the human brain, a population of immune cells known as microglia stands guard over every synapse, pruning neural connections during development, clearing cellular debris and responding to the earliest molecular signs of disease. When these cells malfunction, the consequences reach far beyond inflammation: genetic studies have repeatedly shown that risk variants for Alzheimer’s disease, autism spectrum disorder and schizophrenia are strikingly enriched in microglial genes. Yet for decades, scientists attempting to study these human cells have been trapped between two unsatisfying options, relying on rodent models whose microglia diverge substantially from our own, or extracting primary human microglia that lose their defining characteristics almost the moment they leave the brain. A comprehensive new review published in Experimental & Molecular Medicine by Jina Lee and Yongjin Yoo of Korea University College of Medicine maps how the field is escaping this dilemma through an escalating series of human pluripotent stem cell-based platforms, from simple cultures to three-dimensional organoids and, most recently, chimeric mice carrying human immune cells throughout their brains.

The case against conventional rodent models is now starkly quantitative. According to the review, approximately 41 percent of putative Alzheimer’s disease risk genes lack direct mouse orthologues, and even when a murine counterpart exists, protein sequence identity frequently falls below 60 percent. The APOE ε4 polymorphism, the single largest genetic risk factor for sporadic Alzheimer’s disease, is entirely absent in rodents, while genes such as CR1, CD33, TREM2 and members of the MS4A cluster either have multiple orthologues or poor sequence conservation. Human microglia also display greater transcriptional heterogeneity than their murine equivalents, expressing disease risk genes that are poorly conserved across species. For a disease field built on polygenic architecture, these discrepancies mean that findings in mice often cannot be translated to human biology, no matter how elegant the experiment.

The second problem is environmental. Microglia arise during early embryogenesis from yolk-sac-derived erythromyeloid progenitors and persist as a self-renewing population governed by CSF1R signalling and transcription factors such as PU.1 and IRF8. In the healthy brain they maintain homeostatic functions including complement-mediated synaptic pruning, apoptotic-cell clearance and support of oligodendrocyte maturation through IGF1 secretion, underpinned by signature genes like P2RY12, TMEM119 and CX3CR1. But when microglia are isolated from brain tissue and placed in standard culture dishes, thousands of genes change their expression. Cultured cells abandon their homeostatic surveillance programme and instead upregulate inflammatory cytokines such as IL1B, CXCL8 and CCL20, chemokine receptors, antigen presentation factors and NADPH oxidase components. In effect, removing a microglial cell from the brain pushes it into a chronic activation state that bears little resemblance to its in vivo identity, contaminating nearly every experiment performed on isolated cells.

The solution begins with human pluripotent stem cells, which can be steered into microglia through two complementary methodologies. Cytokine-based differentiation recapitulates embryonic ontogeny, sequentially exposing cells to morphogens such as BMP4, VEGF and stem-cell factor to generate primitive myeloid precursors, followed by terminal survival factors including TGFβ, M-CSF and IL-34. Modern iterations use chemically defined, serum-free conditions and even hypoxic incubation to mimic the oxygen environment of the developing embryonic niche, eliminating the need for manual sorting and feeder layers while improving reproducibility and scalability. Because these protocols mimic natural development, they yield cells whose homeostatic profiles align closely with primary human microglia, making them robust models for developmental and neuroimmune studies.

The alternative approach trades developmental fidelity for speed. Direct lineage conversion uses forced expression of transcription factor cocktails, including PU.1, CEBPα, CEBPβ, IRF5, IRF8 and either MAFB or FLI1, to activate the microglial transcriptional network within days rather than weeks. Recent protocols have integrated these factor cassettes into genomic safe harbour loci such as AAVS1 and CLYBL, ensuring stable, uniform expression without random viral integration. The rapid induction kinetics and homogeneity of this strategy make it particularly well suited to pooled CRISPR screening, allowing researchers to systematically identify regulators of microglial disease states and validate polygenic risk variants at scale. High-resolution mapping of stage-specific regulatory networks, in which persistent core regulators like PU.1 act alongside fetal-state drivers and late-stage IRF, KLF and STAT families, now offers a conceptual route to overcoming the fetal-like limitations that currently constrain both approaches.

Whatever their origin, isolated microglia drift without company, so the field has progressively rebuilt the brain’s cellular ecosystem around them. Two-dimensional co-cultures and tri-cultures combining hPSC-derived neurons, astrocytes and microglia have already yielded fundamental insights: astrocytes sustain microglial survival by secreting CSF1, microfluidic devices reveal microglial migration toward amyloid-β and phosphorylated tau aggregates, and acute exposure to familial Alzheimer’s neurons suppresses the disease-associated microglial signatures otherwise induced by astrocytes. Yet monolayer systems impose artificial planar interactions and fail to reproduce the brain’s viscoelastic extracellular matrix, capping the physiological maturity the cells can achieve.

Three-dimensional brain organoids have pushed the modelling considerably further. When microglial progenitors are embedded within a neuroectodermal matrix containing diverse neural progenitors and glia, either by exogenous addition or by endogenous co-differentiation through targeted PU.1 overexpression, single-cell transcriptomics shows that the resulting cells preserve homeostatic states far better than in planar culture. The relationship is reciprocal: integrated microglia attenuate IRF gene expression and reduce stress markers such as double-stranded DNA breaks in radial glia, modulating the developmental trajectories of their neural neighbours. One striking mechanism involves lipid crosstalk, in which organoid-integrated microglia accumulate PLIN2-positive lipid droplets and export cholesterol that neural progenitor cells then take up, triggering metabolic reprogramming that limits excessive proliferation while promoting axonogenesis and neuronal maturation. In neurodegenerative contexts, microglia within organoids protect neural circuits from amyloid-β toxicity by mitigating ferroptosis, and these systems now support pooled CRISPR interference screening in situ. Multi-regional assembloids extend the paradigm further, showing that microglia adopt niche-specific identities, such as GABA(B) receptor enrichment in striatal regions, and that autism-associated SCN2A mutations trigger GABA(B) receptor-dependent microglial pruning of inhibitory synapses, a pathology reversible by targeted receptor inhibition.

Even organoids cannot fully reconstruct the intact central nervous system, which is why the review’s most consequential section concerns xenotransplantation. Human microglial progenitors can be injected into the brains of immunodeficient mice engineered to express human CSF1, since murine CSF1 binds poorly to the human receptor. Pre-treating neonates with a CSF1R inhibitor creates a permissive niche, and human cells come to constitute roughly 30 percent of all microglia within three weeks, expanding to 80 percent chimerism over three to six months as they outcompete residual murine cells. Within the mouse brain, the transplanted human cells transform: they restore canonical homeostatic genes including TMEM119, P2RY12, CX3CR1 and CSF1R, downregulate artefactually elevated activation markers such as SPP1, CTSB, CD9 and GPNMB, and acquire ramified morphology and surveillance motility matching endogenous cells. A newer refinement engineers a G795A point mutation into CSF1R, rendering the human cells resistant to pharmacological depletion and enabling near-complete replacement of the murine microglial compartment across the entire brain.

The disease applications of these chimeras are already substantial. In an Alzheimer’s model, xenografted human microglia adopt five distinct transcriptional programmes, including homeostatic, disease-associated, HLA antigen-presenting, cytokine-responsive and interferon-responsive states. In a frontotemporal dementia model driven by progranulin deficiency, wild-type human microglia restore progranulin levels, reduce lipofuscinosis and rescue thalamic circuit hyperexcitability and social behaviour deficits. In C9orf72-associated amyotrophic lateral sclerosis chimeras, hexanucleotide repeat expansions impair the microglial transition to disease-associated states and induce endolysosomal dysfunction. Chimeric platforms incorporating human neurons alongside human microglia reveal that autism-related neuronal defects drive abnormal microglial states in a cell-non-autonomous manner, while patient-derived microglia in Down syndrome models show aberrant synaptic pruning. Most strikingly for translational medicine, the FDA-approved antibody lecanemab significantly reduced amyloid plaques in chimeric mice only through Fc-mediated activation of human microglia, with Fc deletion abolishing the effect, a mechanism invisible in conventional mice because the antibody lacks cross-reactivity with murine targets.

Significant challenges remain before these platforms reach their full potential. Current protocols yield microglia retaining fetal or early postnatal signatures, necessitating epigenetic ageing or senescence induction to capture the dystrophic, lipid-accumulating phenotypes that drive late-onset neurodegeneration. Organoid systems suffer from batch variability, inconsistent microglial distribution and survival losses during extended culture, while immunodeficient hosts preclude analysis of adaptive immune contributions to neurodegeneration. The rodent brain may also impose species-specific constraints on human microglial behaviour. Nevertheless, the trajectory is clear: monolayer cultures provide the tractability for genome-wide screening, organoids restore the neuroglial crosstalk that reverses transcriptomic drift, and chimeric models serve as a translational bridge embedding human cells in living systems with systemic circulation and progressive ageing. Together they form a continuous pipeline from high-throughput discovery to in vivo preclinical validation, one that Lee and Yoo argue will ultimately guide clinical trial design and biomarker discovery for neuroinflammation, and bring precision therapies for microglia-driven neurological diseases within realistic reach.

Subject of Research: Human pluripotent stem cell-based modelling of microglia for neurological disease research

Article Title: Translational modelling of human microglia from cultures to chimeras

Article References: Translational modelling of human microglia from cultures to chimeras. (n.d.). https://doi.org/10.1038/s12276-026-01866-1

Image Credits: AI Generated

DOI: 10.1038/s12276-026-01866-1

Keywords: microglia, human pluripotent stem cells, brain organoids, xenotransplantation, Alzheimer's disease, neuroinflammation, CRISPR screening, chimeric mouse models, CSF1R, neurodegeneration, synaptic pruning, disease-associated microglia