Identification of secondary microglial formation centers in the human fetal brain

Originating from the yolk sac (YS), microglia serve as the primary resident phagocytes in the brain, maintaining structural and functional homeostasis throughout development and adulthood (Askew et al., 2017; Bennett et al., 2018; Monier et al., 2006; Nimmerjahn et al., 2005). In humans, microglia enter the developing brain at around 4.5−5.5 gestational weeks (gw) and eventually account for ∼15% of all adult brain cells (Herculano-Houzel, 2012; Verney et al., 2010). The human fetal brain has a large outer subventricular zone (oSVZ), which drives the large size and complexity of the human cortex (Lui et al., 2011). The oSVZ emerges at ∼13 gw, rapidly expands following its formation, and resolves around 24 gw (Baburamani et al., 2020; Hansen et al., 2010; Lui et al., 2011). However, the mechanisms by which the microglial pool robustly expands to meet the cellular demand during this oSVZ period remain poorly understood.

YS-derived microglia enter the fetal brains prior to the generation of astrocytes and oligodendrocytes and the formation of the oSVZ (Lui et al., 2011). Microglia reach the brain through circulation-dependent or circulation-independent routes in murine and zebrafish models (Ginhoux et al., 2010; Xu et al., 2016). CSF-1 receptor (CSF-1R) activation by CSF-1 and IL-34 dictates microglial homeostasis and regional distribution in the murine fetal brain (Freuchet et al., 2021; Kana et al., 2019; Wang et al., 2012). Anatomically, IL-34 regulates microglial maintenance in the murine forebrain and retina, while CSF-1 governs the murine cerebellum; interestingly, IL-34 has also been shown to inhibit microglial maturation and phagocytosis (Devlin et al., 2024, Preprint; Kana et al., 2019; O'Koren et al., 2019). Notably, human IL-34, an evolutionarily conserved protein, does not cross-react with murine CSF-1R, highlighting species-specific differences (Baghdadi et al., 2018; Freuchet et al., 2021). Clinically, microglial dysregulation is implicated in a wide spectrum of neurodevelopmental and adult disorders (Wilson et al., 2023), including Down’s syndrome (DS) (Jin et al., 2022), obsessive–compulsive disorder (Frick and Pittenger, 2016), Rett syndrome (Maezawa and Jin, 2010), and Huntington’s disease (Yang et al., 2017). However, the precise roles of IL-34 and CSF-1 in human brain microglia remain elusive due to the scarcity of primary human samples and the lack of robustly validated in vitro models.

Brain organoids derived from human induced pluripotent stem cells (hiPSCs) recapitulate key anatomical features of the early human fetal brain, especially the oSVZ, serving as an invaluable tool for studying human-specific neurodevelopment (Di Lullo and Kriegstein, 2017; Lancaster et al., 2013; Zhou et al., 2024). However, conventional brain organoids lack YS-derived microglia (Lancaster et al., 2013). Incorporating human microglia into these organoids partially mimics their in vivo physiological roles and complex morphology, while simultaneously enhancing overall organoid maturation (Ormel et al., 2018; Park et al., 2023; Zhang et al., 2023). Therefore, microglia-integrated chimeric human brain organoids offer a robust platform to study the developmental dynamics of microglia in the human fetal brain.

Herein, we combined immunostaining with multiple antibodies and state-of-the-art high-resolution scanning of thick (50 µm) sections to map the precise structural and spatial information of human fetal brain microglia, including healthy and diseased brains. We subsequently utilized chimeric microglia–brain organoids to replicate the precise colonization patterns observed in vivo. In the 12.5–22-gw fetal brains, we identified a large microglial expansion center with distinct morphology, specific anatomical locations, high proliferative microglia, and high-density IL-34⁺ cells, a structure we term the secondary microglial formation center (SMFC). Notably, the emergence of the SMFC coincides with oSVZ formation, and this structure is significantly expanded in the fetal brains of DS and Edwards syndrome (ES). Furthermore, we successfully recapitulated the SMFC in chimeric microglia–human cortical organoids (hCOs). Together, these findings highlight the critical developmental and clinical significance of the SMFC in human brain expansion.

The size of human microglia is ∼50 µm (Torres-Platas et al., 2014). Therefore, immunostaining in thin brain sections (<20 µm) makes it impossible to visualize intact microglia. To see the spatiotemporal status of intact microglia in the human fetal brain, we used IBA-1, CD34, and Ki67, or IBA-1, Ki67, and vimentin antibody immunostaining in 50-µm human fetal brain sections (5−22 gw) (Table S1). We scanned the sections using high-resolution confocal microscopy, including structured illumination microscopy (SIM) with a resolution of near 64 nm. Consistent with previous reports (Menassa et al., 2022; Rezaie and Male, 1999; Verney et al., 2010), a limited number of microglia with amoeboid or bipolar morphology were distributed in the early cortex of 7.5 gw fetal brain, and this density increased with advancing gestational age (Fig. S1, A and B). In SIM images, microglia were amoeboid at the early stage (7.5 gw); subsequently, microglial processes became more complex and branched, and morphological complexity scores increased from 7.5 to 15 gw (Fig. S1, C and D).

Microglia in the ventricular zone (VZ) and subventricular zone (SVZ), including the inner SVZ (iSVZ) and oSVZ, nested between the vimentin⁺ radial glial (RG) fibers, and the major body axis of the microglia aligned parallel to the orientation of the RG fibers (Fig. 1, A–C); microglial processes interacted with RG fibers (Fig. 1, A and C). SIM images revealed that the microglia, which interact with GFAP⁺/vimentin⁺ fibers, had multiple bulbous endings (Fig. 1 D i–iii), a microglial phagocytic unit (Davalos et al., 2005; Nimmerjahn et al., 2005). Notably, some large microglia (>80 µm) exhibited several thick projections with bulbous endings, some of which contain phagocytic GFAP⁺ fragments, and enormous thin projections (<200 nm in diameter), which represent 17.6% (96/547; cells = 58) of all projections, and cross with vimentin⁺ fibers (Fig. 1 D i–iii). The cortical plate (CP) microglia display more bulbous endings than the RG-attaching or ventricular microglia (Fig. 1 E), and the projections of RG-attaching microglia had more crossing points with the RG than those in the CP regions (Fig. 1 F). Approximately 4% of the microglia in the iSVZ and oSVZ interact with microvasculature (Fig. 1, G and H). Together, this approach provides precise structural, physiological, and spatial details about intact microglia, especially the structures smaller than 200 nm and physiology-related signatures, such as bulbous endings and scaffolding.

While screening the 50-µm sections, we identified a large and dense Ki67⁺ microglial aggregate near the caudate in the fetal brains at 12.5–16 gw (n = 10). All microglial aggregates contain rich Ki67⁺/IBA-1⁺ cells, whereas cortical, lateral ganglionic eminence (LGE), and ventricular regions lack such high densities of Ki67⁺ microglia (Fig. 2, A–D; and Fig. S2, A–E). Ki67⁺ microglia in those aggregates account for 26.65% (4,625/17,352) of all aggregate microglia, and 6.32% (209/3,306) of all scattered microglia in other regions (Table S2). The average Ki67⁺ microglial percentage is around 6.5–7.5% in the SVZ, CP, and ventricle, and 24.65% in microglial aggregates, remaining stable from 13 to 16 gw (Fig. S2 D and Fig. 2 E). By estimating microglial aggregate size using nine consecutive sections from 15-gw fetal brains, the largest aggregate measured 1.736 mm² (total area of nine sections: 6.88 mm²) (Fig. 2 F). Observably, Ki67⁺ round microglia are richer and denser in the center of microglial aggregates, whereas ramified or bipolar microglia dominate in the margins (Fig. 2 G). The large proliferative microglial aggregates coincide with the emergence of the oSVZ in human fetal brains, which contributes to the large and complex human brain (Hansen et al., 2010). Extended bipolar morphology is a signature of migrating microglia (Dormann and Weijer, 2006), and the long axis angle of bipolar microglia indicates the orientation (Wei et al., 2009). Notably, the marginal microglia of the aggregate are oriented toward deep brain structures (Fig. 2, H and I). The microvasculature-attaching marginal microglia in aggregate also oriented the deep caudate (Fig. 2 J). The acute angle of the marginal microglia toward the caudate and the superficial marginal zone is 20° and 40°, respectively (Fig. 2, K and L), suggesting directional alignment in the marginal microglia (Lois and Alvarez-Buylla, 1994; Menassa and Gomez-Nicola, 2018). Anatomically, the two ends of the proliferative microglial aggregate in 12 gw + 4 days (12.5)−16 gw fetal brains are located in the CTIP2⁺ cellular bridge of the internal capsule (IC) distal to the LGE, and the middle fraction appeared in the cortical layer next to the CTIP2⁺ LGE (Fig. 3, A and B). Therefore, the proliferative microglial aggregate might be a specific microglial expansion center in the human fetal brain cortex during the oSVZ stage.

The large proliferative microglial aggregate anatomically overlaps with a reported 3b microglial cluster, which is speculated to be a “waiting site” that removes lavish axons, promotes neuroaxonal growth, and mediates axonal pathways (Verney et al., 2010). The Ki67⁺ or Ki67⁻ round microglia in the aggregate center lacked membrane ruffles of Ki67⁺ or Ki67⁻ ventricular amoeboid microglia and Ki67⁺ microglial long projections with bulbous endings in SIM 3D images (Fig. 4 A). These cells resided within the vimentin⁺ fibers, which lack an orientation, and their short processes rarely interacted with vimentin⁺ fibers compared with RG-interacting or CP microglia (Fig. 4, B and C). A reported SPP1⁺/galectin-3⁺ microglial cluster in 9–14-gw human fetal brains, especially those spanning the IC in 14-gw fetal brains (Lawrence et al., 2024), also partially overlapped with the large proliferative microglial aggregate in coronal sections (>12.5 gw). SPP1, also known as osteopontin, is widely expressed across many cell types, including tumor cells and multiple immune cells (Gu and Muller, 2025), and enhances cell migration (Wu et al., 2022; Yi et al., 2022). We used SPP1 and galectin-3 antibodies to evaluate their expression patterns of SPP1 and galectin-3 in the sagittal and coronal sections of human fetal brains at 10−13 gw. SPP1⁺ microglia in the smaller microglial cluster of fetal brains at 10 gw + 6 days and 11 gw + 2 days (Fig. 4, D and E), or the larger microglial aggregates of fetal brains at 12 gw + 4 days and 13 gw show ramified structure with multiple long processes or bipolar morphology (Fig. S2, F–H; and Fig. 4 F). SPP1⁺ microglia are mainly distributed in the margin or the two ends of the larger microglial aggregate (>12.5 gw) (Fig. S2, F and G; and Fig. 4 F). Strikingly, the round microglia, including Ki67⁺ and Ki67⁻ microglia, within the larger microglial aggregate (>12.5 gw) rarely express SPP1 (Fig. S2, F and G; and Fig. 4 F). Microglia in microglial clusters (<12 gw) exhibit more complex processes than those in the SMFC (Fig. 4 G). In the cortical layer or the striatum, long bipolar microglia and microvasculature-interacting bipolar microglia highly expressed SPP1, whereas cortical ramified resident microglia with bulbous endings did not express SPP1 (Fig. S2 I). The SPP1 expression levels in the margin of microglial aggregate (>12.5 gw) are higher than those in the center (Fig. S2 J). Most microglia, including round, bipolar, and ramified, in the smaller microglial clusters (<12 gw) or the larger microglial clusters (>12.5 gw) expressed galectin-3, with particularly strong expression in the microglial cluster r (<12 gw) (Fig. 4, D, E, and G; and Fig. S2, G, H, K, and L). In the scattered microglia of fetal brains, galectin-3 was not present in resident ramified microglia with multiple bulbous endings, SPP1⁺ long bipolar microglia, or some SPP1⁺ microvasculature-interacting microglia (Fig. S2, G, H, and I). The large proliferative microglial aggregates in fetal brains (>12.5 gw) are a novel structure with distinct cellular morphology, anatomical location, and molecular signatures, and differ from previously reported microglial clusters. Here, we refer to it as the SMFC.

The CSF-1R pathway, mediated by CSF-1 and IL-34, controls microglial proliferation and survival in the fetal brain (Elmore et al., 2014). To evaluate the CSF-1R pathway status in the SMFC, we immunostained 5−16-gw fetal brain sections, including coronal and sagittal sections (50 µm), with IL-34, IBA-1, CSF-1R, or CSF-1 antibodies. Notably, the SMFCs harbor rich IL-34⁺/IBA1⁻ cells, and their density in the SMFC region is much higher than that in other brain regions, including VZ, SVZ, subplate, CP, and ganglionic eminence (GE) (Fig. 5, A and B; and Fig. S2 M). CSF-1 is weakly expressed in the microglia of the SMFC, and a few CSF-1⁺/IBA-1⁺ cells are present in the margin of the SMFC (Fig. S2 N). SMFC microglia, especially in the center, expressed higher levels of CSF-1R compared with the scattered microglia in other brain regions (Fig. 5, A and B). The microglial clusters in the coronal and sagittal sections of fetal brains at 10 gw + 6 days or 11 gw + 2 days harbor sporadic IL-34⁺ cells, some of which express Ki67 (Fig. 5 C and Fig. S2 L). In the 12 gw + 4 day fetal brain, the SMFCs harbor a moderate density of IL-34⁺/IBA-1⁻ cells compared with the high IL-34 density in the SMFCs of fetal brains at 13−14 gw, and 13% (113/875) of IL-34⁺/IBA-1⁻ cells are Ki67⁺ (Fig. 5, A and D–F; and Fig. S2 I). IL-34⁺/IBA1⁻ cells are also rare in the cortical and GE regions of 5–9.5-gw fetal brain (Fig. S2 O). Significantly, the percentage of Ki67⁺ microglia in the SMFC with a high IL-34⁺ cell density (>12.5 gw) is identical to that in microglial clusters with few IL-34⁺ cells (<12 gw) (Fig. 5 G). These findings establish that high-density IL-34–expressing cells constitute another distinct signature of the SMFC, suggesting that the IL-34-CSF-1R pathway may regulate the SMFC formation.

We found that 27.2% (560/2,056) of IL-34⁺/IBA-1⁻ cells in the SMFC expressed NeuN, a mature neuronal marker (Fig. 6 A). Strikingly, massive IL-34⁺/NeuN⁺ cells were also present in our hCOs (Liu et al., 2024), which secreted IL-34 and CSF-1 (Fig. 6 B and Fig. S3 A). To determine whether integrating microglia into IL-34–expressing hCO replicates the SMFC, we induced H9-GFP, a human embryonic stem cell expressing GFP, into iMicroglia, and introduced the iMicroglia into hCO at day 28 (Fig. 6 C; and Fig. S3, B and C). Longitudinal tracking revealed that a prominent microglial aggregate appeared on the surface of hCO at 12 days after fusion (daf 12) and lasted to daf 15 (Fig. 6 D). At daf 15, 80% (32/40) and 20% (8/40) of the iMicroglia in aggregates are Ki67⁺ and EdU⁺, respectively (Fig. 6 D). The density of EdU⁺ and Ki67⁺ cells in the surface microglial aggregates is higher than that in other regions (Fig. 6, E and F). A limited number of microglia with amoeboid morphology and small processes are inside iMicroglia-hCO at daf 15 (Fig. S3 D). At daf 30, abundant scattered iMicroglia with long processes, large body size, phagocytic nuclear debris, and lower proliferative signatures appeared inside hCOs and were preferentially enriched in the necrotic region of hCOs (Fig. 6 G and Fig. S3, D–I), indicating that the microglial aggregate emerges before the appearance of phagocytic ramified microglia in chimeric iMicroglia-hCOs.

The coculturing medium of chimeric iMicroglia-hCOs contains 20 ng/ml IL-34 and 100 ng/ml CSF-1. Based on the secretion of CSF-1 and IL-34 by hCOs, we constructed another chimeric SV40-immortalized microglia-hCOs (SV40-hCO) without IL-34 and CSF-1 supplements (Fig. 6 H). Like iMicroglia in hCOs, intra-organoids SV40 microglia phagocytosed nuclear fragments or MAP2⁺ fragments/PSD95⁺ puncta in SV40-hCO (Fig. S3 J). SV40-hCOs exhibited upregulated expression of genes related to synaptic and dendrite development, including BNIP3L, MAP2K1, PGK1, ENO1, and IGFBP2 genes, and microglia-specific genes (e.g., CXCL16, NUPR1, TNFAIP3, TNFRSF19, CLEC11A, and CLEC4A), compared with hCOs (Fig. 6 I). SV40-hCOs also showed increased IL-1β, IL-8, CCL2, and CX3CL1 secretion and reduced TGF-β secretion (Fig. 6 J). Compared with hCOs, SV40-hCOs increased TBR2⁺ cells in the SVZ, and Ki67⁺ progenitor cells in the neural tubes, but not PAX6⁺ or CTIP2⁺ neuronal progenitors (Fig. S3, K–M). Electrophysiological profiling via longitudinal high-density microelectrode arrays (HD-MEAs) in SV40-hCOs showed that the mean firing rate, spike amplitude, and interspike intervals (ISIs) were unchanged (Fig. S3, N and P). However, at the network level, network dynamics was significantly remodeled in SV40-hCOs, exhibiting biphasic evolution of network burst strength and synchronization: a transient amplification peaking at 48 h after integration, followed by attenuation at 72 h, in contrast to the progressive intensification observed in control groups (Fig. S3 P, left). Moreover, significant shifts in interburst intervals and intraburst ISI variability underscore a reorganization of network-level bursting dynamics in chimeric models (Fig. S3 P, right). Together, SV40 microglia drive the secretion of proinflammatory cytokines, enhance neuronal maturation, and selectively reconfigure network synchronization in SV40-hCOs, indicating that SV40 microglia are suitable for studying microglial behaviors.

We noticed that microglia also formed multiple aggregates on the SV40-hCO surface on daf 5, and that inhibiting IL-34 expression in the hCO with shRNA (lentiviral particles) reduced the number of surface microglial aggregates (Fig. 6, K and L; and Fig. S3 Q). 46.0% (192/417) of surface aggregate SV40 microglia, which expressed CSF-1R, were Ki67⁺, and 12.5% (9/72) of surface SV40 microglia in the aggregate were labeled by EdU (Fig. 6 M and Fig. S3 R). Furthermore, exogenous administration of 20 ng/ml IL-34 to SV40-hCO increased the size and density of SV40 microglial aggregates, and inhibiting CSF-1R activity with PLX5622 (pretreated), a classical CSF-1R inhibitor (Wen et al., 2023), decreased their size and density (Fig. 6, N–P). Taken together, the IL-34-CSF-1R pathway regulates the SMFC formation in hCOs.

To determine whether IL-34 drives the small, round microglia in the SMFC, we tested its effects on microglia, using CSF-1 and IFN-γ as controls. 20 ng/ml and 100 ng/ml IL-34 treatment produced round cells with fewer processes and significantly reduced the major axis in SV40 microglia, whereas 20 and 100 ng/ml CSF-1 produced complex microglia with multiple processes and increased the major axis without affecting the minor axis (Fig. 7, A–C). Compared with CSF-1 and IL-34, 20 ng/ml IFN-γ, a proinflammatory cytokine (Ji et al., 2021), reduced the minor axis of SV40 microglia without affecting the major axis (Fig. 7, A–C). Both 20 and 100 ng/ml IL-34 or CSF-1 treatments increased the number of Ki67⁺ and EdU⁺ SV40 microglia (Fig. 7, D–G). RNA sequencing (RNA-seq) data reveal that both 20 ng/ml and 100 ng/ml CFS-1 primarily increased the transcription of differentiation-, developmental-, and morphogenesis-related genes compared with IL-34 treatment in SV40 microglia (Fig. 7 H). However, IL-34 treatment increases the transcripts of genes involved in protein trafficking/transport, mitochondrial function, ribosome function, and RNA processing compared with CSF-1 treatment (Fig. 7 I). IFN-γ treatment, but not IL-34 and CSF-1 treatment, increased immune response-related genes, including ICAM1, SERPING1, IFI35, CCL2, CXCL10, IL15RA, and IL32 (Fig. 7 J). Consistent with findings in SV40 microglia, IL-34 treatment increased Ki67⁺ and EdU⁺ cells and reduced the major axis of iMicroglia (Fig. 7, K–P). Those in vitro findings substantiate the hypothesis that IL-34, but not CSF-1, may be the major driver of the SMFC in the human fetal brain (Fig. 7 Q).

To see the clinical significance of the SMFC in brain development disorders, we analyzed 14–16-gw DS fetal brains (n = 5). In DS fetal brains, we identified larger SMFCs with 39.0% (7,787/19,952) Ki67⁺ microglia compared with 5.58% (173/3,099) in scattered microglia (Fig. 8, A–C; Fig. S4 A; and Table S2). The DS SMFC region uniformly harbors a high density of IL-34⁺/IBA-1⁻ cells (Fig. 8, A and D; and Fig. S4, B–D), some of which express NeuN (Fig. S4 E). The microglia of the DS SMFC uniformly expressed higher CSF-1R compared with the scattered microglia (Fig. 8, E and F; and Fig. S4, B–D), and sporadically expressed CSF-1 (Fig. S4 F). Multiple SPP1⁺ bipolar and microvessel-interacting microglia exist in the SMFC margin, whereas round microglia in the SMFC and scattered ramified microglia are with weakly or not expressed SPP1 (Fig. S4, G and H). Galectin-3 is uniformly present in the microglia of the DS SMFC, including round and bipolar microglia (Fig. S4 H). Anatomically, the two ends of the DS SMFC extended into the CTIP2⁺ cellular bridge of IC, and the middle section did not overlap with the DARPP32⁺ region (Fig. 8 E; and Fig. S4, I and J). The morphology of SMFC microglia is identical to that of healthy SMFC microglia and lacks long processes, bulbous endings, and membrane ruffles (Fig. 8, G and H). Consistent with the impaired brain development of DS, more complicated processes and LAMP1⁺ bulbous endings were present in the scattered microglia (Fig. 8, I and J; and Fig. S4, K–N), and more caspase-3⁺ (cas3) cellular projections and bodies were present in the region of scattered microglia but not in the expanded DS SMFC (Fig. 8, K and L). Similar to the healthy SMFC, cas3⁺ cells, CD8⁺ cells, CD177⁺ neutrophils, iNOS⁺, and CD206⁺ microglia were absent in the DS/healthy SMFC (Fig. 8, M and N; and Fig. S4, O and P), excluding an inflammatory role of the SMFC. Together, the DS SMFC and the healthy SMFC are identical except for size (Fig. 8 O).

We analyzed fetal brains from individuals with ES (n = 2) and Turner syndrome (TS, n = 2) to examine SMFC changes in these conditions. ES and TS fetal brains (14–15 gw) harbor SMFCs with round microglia possessing few membrane ruffles, bulbous endings, higher CSF-1R, and a rich population of IL-34⁺/IBA-1⁻ cells (ES, 3.437 mm²; TS, 1.683 mm²) (Fig. 9, A–I; and Fig. S5 A). Ki67⁺ microglia accounted 60.5% (5,894/9,741) and 45.1% (1,251/2,771) in the TS/ES SMFCs, and 5.2% (84/1,627) and 5.4% (109/2,009) in the TS/ES scattered microglia, respectively (Fig. 9 F and Table S2). The anatomical location of the SMFC in the TS and ES brain completely overlaps with that of the SMFC in the healthy and DS fetal brain (Fig. S5, A–E). Similar to other SMFCs, the two ends of the TS/ES SMFC extended into the CTIP2⁺ cellular bridge of IC but did not overlap with the DARPP32⁺ region, and the middle region of the SMFC is located in the cortical layer next to IC (Fig. 9 D; and Fig. S5, B and D). In the ES and TS fetal brain, more complex processes with bulbous endings were present in scattered microglia (Fig. 9, J and K). Notably, identifiable cas3⁺ cells and fibers in ES but not the TS fetal brain samples are significantly higher than in the healthy brain (Fig. 9, L and M). The morphology and distribution patterns of SPP1⁺ and galectin-3⁺ cells in the ES and TS SMFC and other brain regions are identical to those of the DS and healthy fetal brains (Fig. S5, C and E). Collectively, these data highlight the SMFC as a clinically relevant structure responsive to neurodevelopmental perturbations.

Comprehensive structural and spatial profiling of intact microglia at super-resolution has been lacking in studies of the human fetal brain. Using super-resolution scanning of 50-µm sections, we identified novel large microglial aggregates in 12.5−22-gw human fetal brains, referred to as the SMFC. This structure is characterized by over 25% Ki67⁺ microglia, high-density IL-34⁺ cells, and evaluated CSF-1R expression and significant expansion in neurodevelopmentally impaired fetal brains, such as those with DS and ES. The microglia within the SMFC are morphologically round, lack phagocytic units and projections, and differ from well-described ramified, bipolar, and amoeboid microglia. Markedly, the appearance of the SMFC coincides with the emergence of the oSVZ in human fetal brains. While the percentage of Ki67⁺ microglia in the scattered microglia of human fetal brain is nearly identical to that of mouse fetal brain microglia (5–10%) (Barry-Carroll et al., 2023; Menassa et al., 2022), the proliferative fraction within the SMFC is over 25%. Given that the SMFC covers a substantial area of ∼7 mm² in the fetal brain and exhibits robust proliferation, this structure may compensate for microglial demand during the rapid cortical expansion period.

IL-34 controls the population of murine forebrain microglia and inhibits microglial maturation (Devlin et al., 2024, Preprint; Kana et al., 2019). Our data—including high-density IL-34⁺ cells in the SMFC, cellular and molecular shifts in IL-34– and CSF-1–treated microglia, and IL-34 knockdown and stimulation in chimeric microglia-hCOs—support the notion that IL-34 secretion drives the SMFC formation. CSF-1R inhibition in chimeric microglia-hCOs further confirmed the necessity of the IL-34-CSF-1R pathway in this process. CSF-1R inhibition in chimeric microglia-hCOs further confirmed the necessity of the IL-34-CSF-1R pathway in this process. Neuronal death activates microglia and enhances microglial proliferation (Dang et al., 2018). Although SV40-immortalized microglia in hCOs secrete multiple cytokines, phagocytose synapses and cellular debris, and express IBA-1 and CSF-1R, their morphological complexity is significantly lower than that of iMicroglia. However, SV40 microglia form surface proliferative aggregates more quickly than iMicroglia, suggesting that they functionally resemble immature rather than mature microglia.

Similar to other observations (Seidl et al., 2001), we noticed increased neuronal death in the DS and ES fetal brains, which concomitantly present with an expanded SMFC. Trisomy of chromosome 21 in DS aberrantly activates microglia, induces neuroinflammation, and increases synaptic pruning (Jin et al., 2022; Zheng et al., 2021). However, the absence of obvious neuronal apoptosis, activated microglia with phagocytic morphology, CD8⁺ cells, and CD177⁺ neutrophils in the DS SMFC regions indicates that the expanded SMFC functions primarily as a microglial proliferation center rather than an immune response center. This hypothesis is corroborated by the finding that IL-34 enhances microglial proliferation without promoting proinflammatory or phagocytic activities. Furthermore, we observed an increased number of bulbous endings in scattered microglia across the cortex and in the LGE in 13–15-gw fetal brains. We observed an increased number of bulbous endings in scattered microglia across the cortex and in LGE in 13−15-gw fetal brains. Because the SMFC appears before the onset of high-density synaptogenesis in the fetal brain, the primary role of microglia with bulbous endings at this stage might be scavenging apoptotic cells or aberrant fibers rather than synaptic pruning.

The precise mechanisms underlying IL-34’s role in the SMFC formation are not fully addressed in this study due to the lack of an IL-34–null ES/iPSC line to generate chimeric microglia-hCOs. Additionally, the direct relationship between the oSVZ emergence and SMFC formation remains hypothetical. Current in vitro brain organoid models do not fully recapitulate the extensive architecture of the human fetal brain, particularly the large oSVZ. Thus, validating the direct interplay between the oSVZ and the SMFC requires the development of next-generation human brain organoids with well-developed oSVZ regions. Because the oSVZ is uniquely expanded in human brain development (Lui et al., 2011), the evolutionary uniqueness of the SMFC needs to be addressed by investigating the SMFC or SMFC-like structures across different species using our approach, despite no identical structures being reported in lab species (Barry-Carroll et al., 2023; Geirsdottir et al., 2019; Lawrence et al., 2024; Yu et al., 2024).

The use of hiPSCs was approved by the Ethics Committee of the Institutes of Biomedical Sciences at Fudan University, Shanghai, China (no. 28). The human embryonic stem cell line H9 (WA09) was provided by Prof. Su-Chun Zhang, Waisman Center, Madison, USA. Human pluripotent stem cells (hPSCs) were cultured under feeder-free conditions on Matrigel (354234; Corning)- or vitronectin XF–coated (07180; STEMCELL Technologies) 6-well plates in Essential 8 (E8) medium (A1517001; Gibco).

This study included two cohorts of fetal brain samples: a historical healthy cohort collected from 2006 to 2008, which was cryosectioned immediately and stored at −80°C (Table S1), and a prospective cohort collected from 2023 to 2024 (Table S1). All fetal brain tissues were obtained with informed maternal consent and institutional ethical approval (Ethics Committees of Pingdingshan Maternal and Child Health Hospital [2025-006] and Fudan University, Shanghai, China [2009-13]). Clinical diagnoses were confirmed via an Affymetrix CytoScan 750K array and karyotype analysis. Inclusion and exclusion criteria are detailed in Table S1; briefly, specimens exhibiting autolysis, internal hemorrhage, or inflammation were excluded. Gross pathology was evaluated by a senior pathologist, and tissue histopathology was assessed after immunofluorescence imaging.

hPSCs were dissociated into single cells and seeded into ultra-low-attachment V-bottom 96-well plates to form spheroids (day 0). On day 1, spheroids were transferred to suspension culture. From days 1 to 6, the medium consisted of 48% DMEM/F12, 48% E8 medium, 1% NEAA, 1% GlutaMAX, 1% N2 Supplement, 1% penicillin–streptomycin (P/S), 2 μM SB-431542, and 0.3 mM LDN-193189. Optionally, the WNT inhibitor XAV-939 (2.5 μM) and two SMAD inhibitors were added. From days 7 to 14, the medium was replaced with 96% Neurobasal, 1% NEAA, 1% GlutaMAX, 1% N2 Supplement, 1% P/S, 20 ng/ml EGF, and 20 ng/ml bFGF. From day 15 onward, small molecules were withdrawn, and 2% B27 was supplemented into the basal medium.

The H9-GFP line was maintained on vitronectin in E8 medium. To generate embryoid bodies (EBs), cells were dissociated using Accutase (5 min, 37°C) and seeded into ultra-low-attachment 60-mm dishes in E8 medium supplemented with 10 nM Y27632 for 24 h (day 0). Hematopoietic and myeloid differentiation was induced across six stages in APEL II (05270; STEMCELL Technologies) or StemPro-34 (10639011; Gibco) media. Differentiation stages: Stage 1 (day 1, mesoderm induction): APEL II with 10 ng/ml BMP-4 and 5 ng/ml bFGF. Stage 2 (days 2–7, hematopoietic progenitors): APEL II with 10 ng/ml BMP-4, 5 ng/ml bFGF, 50 ng/ml VEGF, and 100 ng/ml SCF. Stage 3 (days 8 and 9, myeloid progenitors): APEL II with 10 ng/ml bFGF, 50 ng/ml VEGF, 50 ng/ml SCF, 10 ng/ml IGF-1, 25 ng/ml IL-3, 50 ng/ml M-CSF, and 50 ng/ml GM-CSF. Stage 4 (days 10–20, iMac precursors): 40–50 EBs were plated per well in Matrigel-coated 6-well plates in StemPro-34 supplemented with the Stage 3 cytokine cocktail. Stage 5 (days 21 and 22): suspension cells were harvested and cultured in StemPro-34 with M-CSF and GM-CSF increased to 100 ng/ml. Stage 6 (days 23–26): StemPro-34 with 5 ng/ml bFGF, 50 ng/ml VEGF, 50 ng/ml SCF, 10 ng/ml IGF-1, 100 ng/ml M-CSF, and 100 ng/ml GM-CSF.

We adapted a previously established protocol to generate microglia–cortical organoids. Briefly, 1 × 10⁵ day-26 iMacs were cocultured with one day-26 hCO in 1 ml of organoid medium supplemented with 100 ng/ml M-CSF and 20 ng/ml IL-34 in an ultra-low-attachment 3.5-cm dish. On day 2, an additional 1 ml of supplemented organoid medium was added. From day 5, half of the medium was changed every 3 days for 18 days.

SV40-immortalized microglia were purchased from Applied Biological Materials Inc. (T3961). Cells were maintained in Prigrow III (TM003) or DMEM/F12 with 10% FBS on Applied Cell Extracellular Matrix–coated plates, and passaged at 80% confluence using 0.25% trypsin. To generate SV40-hCOs, SV40 microglia were introduced into the hCO culture medium on days 30 and 60 at a density of 1 × 10⁴ cells/ml. The coculture was incubated overnight on an orbital shaker to facilitate microglial integration. The medium was subsequently replaced with microglia-free medium and maintained for 7 days.

HD-MEA chips (MaxLab) were coated with 0.07% (vol/vol) poly(ethylenimine) (Sigma-Aldrich) and laminin. Whole hCOs were mounted onto the chips and secured with a droplet of Matrigel. Following a 30-min gelation period at 37°C, the chips were filled with BrainPhys Neuronal Medium (STEMCELL Technologies) supplemented with 2% NeuroCult SM1, 1% N2 Supplement-A, 1× GlutaMAX, 20 ng/ml BDNF, 20 ng/ml GDNF, 20 ng/ml NT-3, 1 mM dibutyryl-cAMP, and 1× Antibiotic–Antimycotic. Recordings commenced at least 7 days after plating. Spontaneous activity was mapped sequentially across the 1,020-electrode array (30 s/configuration) at a 20-kHz sampling rate. Spikes were detected utilizing a threshold of 5× the root-mean-square noise of the band-pass–filtered signal. Data reflect recordings from 3 hCOs and three SV40-hCOs.

Day-37 hCOs were embedded in 3% low-melting agarose and sectioned into 300-μm slices using a vibratome (WPI). Slices were recovered in ultra-low-attachment dishes for 5 days. For IL-34 knockdown, slices were incubated for 24 h in 96-well plates containing IL-34 lentivirus (1:10 and 1:100 dilutions) and 5 μg/ml Polybrene. After transduction, slices were recovered in fresh medium for 48 h before 2 × 10⁴ SV40 microglia were seeded per well (day 0). Cocultures were maintained on an orbital shaker, with medium replaced on days 1 and 3, and fixed in 4% paraformaldehyde (PFA) on day 5.

For cytokine stimulation, SV40 microglia or hCOs were treated with 20 or 100 ng/ml of human recombinant IL-34 (HumanKine, HZ-1316), CSF-1, or IFN-γ for up to 120 h. For proliferation assays, EdU (10 µM; Invitrogen) was added to the culture medium for 12 h prior to fixation. Cells/organoids were permeabilized and processed utilizing the Click-iT reaction kit according to the manufacturer’s protocol (Chen et al., 2022).

Organoids and fetal brain samples were fixed in 4% PFA for 30 min or 24 h, respectively, cryoprotected in 30% sucrose, and embedded in OCT. Sections (15 µm for organoids; 50 µm for fetal brains) were blocked and permeabilized in 0.3% Triton X-100/10% normal donkey serum for 1 h, followed by overnight incubation at 4°C with primary antibodies (Tables S3 and S4). Slices were washed and incubated with fluorophore-conjugated secondary antibodies (Table S5) and DAPI for 1 h at room temperature. High-resolution imaging was performed using an Olympus SpinSR10 spinning disk confocal microscope, a Nikon Structured Illumination Microscope(∼64-nm resolution), a Leica wide-field confocal microscope, or a Zeiss 880 laser scanning confocal microscope.

Bulk RNA-seq was performed on SV40 microglia treated with 20 ng/ml IFN-γ, 20 ng/ml IL-34, 20 ng/ml CSF-1, or 100 ng/ml CSF-1 (n = 3 biological replicates per condition), alongside unstimulated controls (n = 6). Libraries were sequenced on an Illumina NovaSeq 6000 platform (PE150 strategy). Raw reads were quality-filtered, and differential gene expression analysis was executed using the DESeq2 package in R (version 4.4.1). Differentially expressed genes (DEGs) were defined by an absolute log₂ fold change ≥1.0 and a Benjamini–Hochberg adjusted P <0.05. Gene Ontology functional enrichment analysis for DEGs was performed using the clusterProfiler package, assessing Biological Process, Molecular Function, and Cellular Component categories.

Cell quantification, fluorescence intensity, and spatial distributions were analyzed using ImageJ (Fiji) and Imaris 9.8. For microglial morphometry, ImageJ was used to binarize and skeletonize single-cell images, followed by the “Analyze Skeleton” plugin. FracLac analysis was employed to determine cellular complexity (box-counting method). Statistical analyses were performed using GraphPad Prism. Data are presented as the mean ± SD or SEM, as indicated. Two-group comparisons were assessed using the Mann–Whitney U test or Student’s t test, while multiple comparisons were analyzed via one-way ANOVA followed by Tukey’s or Holm–Sidak’s post hoc tests. Significance was defined as P ≤ 0.05.

Table S1 shows baseline characteristics of fetal brain samples. Table S2 shows characteristics of the SMFC in healthy and diseased fetal brain samples. Table S3 shows antibodies for immunostaining with their purpose. Table S4 shows the list of primary antibodies. Table S5 shows the list of secondary antibodies, dyes, and reagents. Fig. S1 shows the changes of microglial density and complexity in the fetal brains (related to Fig. 1). Fig. S2 shows the proliferative microglial aggregates are present in the same location next to the LGE, absent in the LGE and ventricle of human fetal brains with different gw, and harbor SPP1⁺ bipolar and microvasculature-interacting microglia on the border region (related to Figs. 2, 3, 4, and 5). Fig. S3 shows characterization of hCOs and chimeric iMicroglia-hCOs (related to Fig. 6). Fig. S4 shows characterization of the SMFCs in DS fetal brains (related to Fig. 8). Fig. S5 shows characterization of the SMFCs in ES and TS fetal brain (related to Fig. 9).

All relevant data are included in the paper and are available upon reasonable request. The bulk RNA-seq and proteomic datasets generated and/or analyzed during this study are available in the Sequence Read Archive under the following accession numbers: PRJNA1152524, PRJNA1156058; and the National Genomics Data Center BioProject database under the following accession number: PRJCA057062.

This work was supported by the National Key Research and Development Program of China (grant no. 2021YFA1101302 to L. Ma) and the National Natural Science Foundation of China (grant no. 32370852 and U24A2014 to L. Ma).

Author contributions: Chenyun Song: conceptualization, data curation, formal analysis, investigation, methodology, software, validation, visualization, and writing—original draft, review, and editing. Xinyu Chen: data curation, formal analysis, investigation, and methodology. Rong Ji: investigation. Yang Liu: investigation, methodology, and writing—review and editing. Yawen Han: formal analysis, investigation, validation, and writing—review and editing. Fangzhou Ye: software and writing—review and editing. Ling Zhang: investigation. Li Li: methodology, resources, software, and validation. Lu Gao: resources. Qizhi He: conceptualization, data curation, investigation, methodology, resources, validation, visualization, and writing—original draft, review, and editing. Lixiang Ma: funding acquisition, investigation, project administration, supervision, and writing—review and editing. Hexige Saiyin: conceptualization, data curation, formal analysis, investigation, methodology, project administration, software, supervision, validation, visualization, and writing—original draft, review, and editing.