The meninges consist of three layers—the pia mater hugging the brain surface, the arachnoid mater above it, and the dura mater outermost. Together they anchor the brain to the skull, carry its blood supply, house immune defences, and provide the channels through which cerebrospinal fluid flows. Despite their clinical importance, no one had ever produced a single-cell atlas of the human meninges during development. That gap mattered, because tumours are known to hijack developmental programs, and identifying a tumour’s cell of origin often requires comparing it against the normal cell types present during organ formation.

To close that gap, the researchers collected meningeal tissue from 16 samples spanning postconceptional weeks 5 through 13, obtained from elective abortions with informed consent under approved ethical protocols. Using droplet-based single-cell RNA sequencing, they profiled 156,726 high-quality cells, which they organised into 245 clusters annotated across four hierarchical levels: 115 cell types, 105 subclasses, 45 classes and 8 superclasses. Fibroblasts marked by the transcription factor FOXC1 made up roughly half the dataset, alongside PECAM1-positive endothelial cells, ABCC9-positive perivascular cells, PTPRC-positive immune cells and even anucleated erythropoietic cells. Interacting neural, neural crest and choroid plexus epithelial cells were also captured, giving the atlas an unusually complete view of the brain’s outer environment.

Crucially, the team paired the sequencing with spatial transcriptomics using the Xenium platform, cryosectioning entire embryonic heads at weeks 6 and 9.5 so that the meninges remained anatomically intact. A custom probe panel targeting 389 genes, chosen from the sequencing data, allowed them to locate each cell type in situ. At week 6 they identified the primary meninx, the earliest meningeal structure described in the literature, a thin fibroblast sheet wrapping the entire brain together with FOXQ1-positive endothelial cells. By week 9.5, the spatial data revealed a remarkably complete stack of tissue layers, from the ventricular zone of the developing cortex up through Cajal–Retzius cells, meningeal fibroblasts, the future skull, periosteum and skin.

One of the atlas’s most consequential findings concerns how the three meningeal layers form. Rather than branching off from distinct precursor lineages the way blood cells do in haematopoiesis, the pia, arachnoid and dura emerged as a continuum of cell states. Layer-specific markers—LAMC3 for pia, SLC22A6 for arachnoid and COL8A1 for dura—showed smooth gene expression gradients at every age examined, and maturation scores built from adult layer-enriched genes revealed that all three layers mature concurrently, their adult gene programs already active in the first trimester and steadily increasing through week 13. Cycling cells were present in every layer, further supporting a model of parallel, gradual refinement rather than branched lineage commitment.

Within the dura, the spatial data exposed an unexpected subdivision. Beyond the thick outer dura, which expressed MSX2 and blended into osteogenic skull and periosteal tissues, the researchers found a distinct inner dura: a single sheet of cells apposed directly onto the arachnoid, marked by SLC47A1 and, surprisingly, the epithelial marker CDH1. This inner layer expressed a full complement of tight-junction genes—Occludin, several Claudins, junctional adhesion molecules and their intracellular scaffolds—plus adherens junction components and even tricellular junction proteins. In other words, the inner dura looks like a second barrier layer, histologically resembling arachnoid barrier cells rather than ordinary dura. Because the same CDH1-positive inner dura appears in adult human and mouse meninges, the team proposes that the complete meningeal barrier may be built from two cell sheets—the arachnoid barrier and the inner dura—and suggests renaming it the arachnoid–dura barrier.

The atlas also delivered surprises well beyond the dura. The fetal meninges harboured 19 immune cell types as early as week 5, including macrophages tiled sparsely across the primary meninx, microglia confined to the brain side of the leptomeningeal boundary, and—remarkably—B-lineage cells. CD34-positive, IL7R-positive, CD19-low early lymphoid progenitors appeared in the meninges at week 5, a full week earlier than such cells had been reported in fetal liver, and matured through pre-pro, pro, pre and immature B cell stages by weeks 8 to 13. This positions the meninges as an unexpected site of early human B cell differentiation. The vascular analysis was equally revealing: venous endothelial cells expressing the lymphatic transcription factor PROX1 suggested that subarachnoid lymphangiogenesis may begin in early fetal life, whereas in mice meningeal lymphatics develop only after birth.

The pivotal question, however, was what all this says about meningiomas. The team reanalysed published single-cell and bulk RNA sequencing data from human meningiomas and generated new Xenium spatial data from seven tumours, including one grade I and six grade III tumours. Using latent Dirichlet allocation topic modelling—a computational method that identifies co-expressed gene modules corresponding to cell identities—they defined 35 topics from the fetal fibroblast data, including topics for pia, arachnoid, inner dura, outer dura and a barrier topic shared between arachnoid and inner dura cells. When these topics were transferred to the tumour data, the result was unambiguous: pia and arachnoid topics were nearly absent, while inner dura, dura and barrier topics dominated. Tumour cells expressed CDH1 and SLC47A1 but showed no detectable SLC22A6, the arachnoid marker. The same pattern held in bulk RNA sequencing of 185 meningiomas across grades I to III, with dural topics dominating in merlin-intact, immune-enriched and hypermitotic methylation subtypes alike.

Genetic evidence pointed the same way. Because both tumour suppressors and oncogenes must normally be expressed in a cell before transformation can occur, the researchers compiled the 20 most frequently mutated genes in meningiomas from the COSMIC Cancer database and measured their collective expression across the fetal and adult meningeal cell types. The dural lineage—particularly committed dura precursors and inner dura cells—showed by far the highest expression of these putative cancer drivers, a result that persisted even after removing genes commonly mutated in glioblastoma, intestinal or pancreatic cancers. The authors are careful to frame this as a hypothesis rather than a settled fact: it remains possible that tumours arising from arachnoid barrier cells later convert to a dura-like state, and previous mouse models that generated meningiomas from PTGDS-expressing cells did not distinguish arachnoid barrier from inner dura, since PTGDS is expressed in both. Still, the convergence of spatial, single-cell, bulk and genetic lines of evidence makes dural cells a serious new candidate for the meningioma cell of origin.

The clinical implications could be substantial. Higher-grade atypical and anaplastic meningiomas currently have no effective drug therapies, a failure partly attributed to decades of uncertainty about what these tumours actually are. If meningiomas are, at a transcriptional level, dural tumours, then therapeutic strategies, animal models and drug screens designed around arachnoid biology may have been aiming at the wrong target. The atlas itself—freely available through public repositories including the European Genome-Phenome Archive and GitHub—now provides the reference framework the field has lacked, charting not only the fibroblast layers but the immune, vascular and choroid plexus cells that co-develop with them. Whether dura-lineage cells can truly give rise to meningiomas will require lineage-tracing and functional experiments, but this study has, for the first time, given the question a molecular foundation—and in doing so, it has reopened a debate that medicine thought it had closed more than 150 years ago.

Subject of Research: Single-cell and spatial transcriptomic mapping of human fetal meninges development and its implications for the cellular origin of meningiomas

Article Title: Cell atlas of the developing human meninges reveals a dura-like nature of meningiomas

Article References: Vinsland, E., Marco Salas, S., Kapustová, I., Hu, L., Webb, S., Li, X., He, X., Nilsson, M., Haniffa, M., Barker, R. A., Persson, O., Raleigh, D. R., Sundström, E., Lönnerberg, P., & Linnarsson, S. (2026). Cell atlas of the developing human meninges reveals a dura-like nature of meningiomas. Nature Cell Biology. https://doi.org/10.1038/s41556-026-02074-9

Image Credits: AI Generated

DOI: 10.1038/s41556-026-02074-9

Keywords: meninges, meningioma, dura mater, arachnoid, single-cell RNA sequencing, spatial transcriptomics, fetal development, cell atlas, brain tumours, fibroblasts, tight junctions, cell of origin