Pathogenic myeloid phenotypes drive disease pathology in a novel human neurohistiocytosis model

Key Points

• •Patient-derived BRAF^V600E iPSCs produce microglia that are inflammatory and dysfunctional with neurohistiocytosis-specific gene alterations.
• •BRAF^V600E myeloid progenitors hyperproliferate and drive LCH-ND neuropathology including complex CNS lesions and motor deficits in mice.

Visual Abstract

Abstract

Innate immunity is increasingly recognized as a driver of neurodegeneration, although pathogenic mechanisms are incompletely understood. Langerhans cell histiocytosis (LCH) is an inflammatory myeloid neoplastic disorder caused by activating somatic mutations in MAPK pathway genes, most commonly BRAF^V600E, in myeloid precursors. A subset of patients with LCH develop progressive neurodegeneration (LCH-ND). We generated a human induced pluripotent stem cell (iPSC) model from patients with somatic hematologic mosaicism for BRAF^V600E. Brain macrophages/microglia from LCH iPSCs exhibit unique disease-specific pathogenic features. Stepwise differentiation identified hematopoietic progenitors as hyperproliferative, whereas brain macrophages were apoptosis resistant. Through application of cerebral organoids and a humanized murine xenotransplantation model, we identify marked heterogeneity of differentiation potential within clonal BRAF^V600E lines in vivo. This model phenocopied human-specific phenotypes, including dense basal ganglia foci of abnormal macrophages, marked neurodegeneration with astrogliosis, and progressive ataxia. This approach will allow for preclinical testing of therapeutics for LCH-ND.

Central nervous system Langerhans cell histiocytosis is a rare, devastating inflammatory myeloid neoplasm driven by somatic mutations in the MAPK pathway, most frequently the BRAF^V600E mutation. Rajamanickam et al established a high-fidelity in vivo model of its complex pathology using xenotransplantation of human BRAF^V600E-mutant hematopoietic progenitor cells into the brains of immunodeficient mice. They found that these progenitors are hyperproliferative and that the output of mature macrophages is highly inflammatory and apoptosis resistant in this environment, actively driving neuronal loss and reactive astrogliosis.

Introduction

Langerhans cell histiocytosis (LCH) is a heterogeneous myeloid neoplastic inflammatory disorder caused by somatic mutations in the MAPK pathway in hematopoietic lineage cells.1, 2, 3, 4 Chronic activation of the MAPK pathway leads to diverse inflammatory immune lesions and a highly variable clinical presentation. Approximately 20% of patients with LCH develop LCH-associated neurodegeneration (LCH-ND), a devastating form of neurodegeneration that can occur years after systemic remission.5, 6, 7 LCH-ND is characterized by ataxia, dysarthria, cognitive deficits, pituitary abnormalities, and white matter infiltrative lesions, ultimately leading to significant morbidity and mortality.8, 9, 10, 11

Systemic LCH lesions arise from myeloid cells carrying MAPK mutations and are canonically positive for CD207 (Langerin) and CD1a.¹ Pathogenic histiocytes recruit inflammatory infiltrating cells, resulting in the characteristic granulomatous lesions. LCH can present in the central nervous system (CNS) with similar granulomatous lesions or progressive diffuse neurodegeneration primarily affecting the brain stem and cerebellum. LCH-ND shows substantial heterogeneity in both neuroimaging and clinical presentations.⁷

LCH-ND historically was considered a paraneoplastic or autoimmune phenomenon because of the lack of characteristic CD207⁺ cells in the few reported biopsies.¹² More recently, LCH-ND has been shown to be associated with significant infiltration of areas of neurodegeneration with BRAF^V600E microglia-like/myeloid cells.⁶ The ability of BRAF^V600E hematopoietic precursors to drive LCH has been demonstrated in a mouse model.¹³ Unpredictability of patients at risk for LCH-ND, incomplete mechanistic understanding of pathogenesis and progression, and lack of effective therapies remain major unmet needs to improve outcomes for patients.

LCH-ND is associated with early-onset skull lesions and the BRAF^V600E genotype.^6,11,12,14 Induction of BRAF^V600E in microglia led to the hypothesis that LCH-ND may be prenatally seeded.¹⁵ However, perivascular localization of BRAF^V600E cells and a mouse model recapitulating LCH-ND with BRAF^V600E hematopoietic cells support a hematopoietic origin.^6,13 Furthermore, the etiology of heterogeneity in LCH-ND onset and progression is unclear, whether this arises from heterogeneity in immune cell infiltration, cell type of origin, or is intrinsic to the individual MAPK clone.

Several genetically inducible murine models of systemic LCH have been generated, including those driving enforced expression of the endogenous Braf^V600E locus under the Cd11c, Csf1r, and Scl promoters.^13,15, 16, 17, 18 Most of these models are not reported to develop neurological sequelae,16, 17, 18 other than the Scl mice^13,15 that developed diffuse infiltrative neurological lesions but not the substantial heterogeneity of human LCH-ND. Only a single report has applied human induced pluripotent stem cells (iPSCs) to LCH, primarily modeling in vitro early hematopoiesis to CD1a⁺/CD207⁺ lesional cells.¹⁹

Altogether, the limited animal models, cellular heterogeneity, and ontological uncertainties have considerable treatment ramifications for neurodegenerative risk. Several reports describe responses of LCH-ND to BRAF^V600E or MEK inhibition,20, 21, 22 but all with incomplete responses and high rates of treatment failure upon discontinuation. MAPK inhibition likely blunts the cellular pathogenic activity of the cell but may not eliminate precursor clones.²³ Improved risk-prediction and therapeutic strategies are needed.

In this study, we generated a human model of systemic LCH and LCH-ND, capitalizing on peripheral blood somatic mosaicism for the most common pathogenic LCH MAPK mutation,⁴ BRAF^V600E, to generate an iPSC cohort from patients with high-risk multisystem LCH. These models were developed to inform LCH-ND pathogenesis and create a platform for preclinical testing of novel therapeutic strategies. Application of this isogenic and controlled cohort to generate hematopoietic progenitors and differentiated brain macrophages/microglia identified a hyperproliferative progenitor state and a mature apoptosis-resistant phenotype, recapitulating key lesional cellular subtypes^17,24 and overlapping with LCH-ND biomarker and brain biopsy gene expression. Using both cerebral organoid cocultures and a murine xenotransplantation model of human microglia/brain macrophages, we identify the propensity of human MAPK-mutated myeloid clones to cause astrogliosis, neuronal loss, and neurodegeneration. Crucially, this model recapitulates key aspects of LCH-ND independent of vascular infiltration and not identified in current murine models, such as clonal heterogeneity, including LCH-ND nodules positive for CD1a and CD207 with coexisting infiltrative myeloid lesions negative for these markers. Altogether, these novel human myeloid models can be used for in vitro and in vivo preclinical testing of therapeutic single or combinatorial strategies to inform myeloid differentiation–specific sensitivity and future therapeutic approaches to the treatment of LCH-ND.

Methods

Samples/lines

Patient samples were obtained with informed consent under institutional review board approval from the University of California San Diego, or Baylor College of Medicine. Peripheral blood mononuclear cells were isolated and reprogrammed following the manufacturer’s protocol (Cytotune 2.0; Thermo Fisher Scientific). iPSCs were cultured as previously described,^25,26 and CRISPR gene editing was performed as previously described, using homology-directed repair.^27,28

Differentiations

Induced hematopoietic progenitor cells (iHPCs) and iPSC microglia were generated as previously described.^25,26 Morphology and phagocytosis live imaging were measured by Incucyte S3 (Essen BioScience) as previously described.²⁶ Myelin was isolated as previously described.²⁹ Cell cycle analysis used Vybrant DyeCycle (Thermo Fisher Scientific). Antibodies are listed in the supplemental Methods, available on the Blood website. Cytokine multiplexing was performed by Nomic, Inc. RNA sequencing library preparation and analysis were performed and analyzed as previously described.^30,31

Organoids

Organoids were generated and cocultured with microglia, and single cells were isolated and analyzed as previously described.^25,32

Xenotransplantation

Transplantation was neonatally performed as previously described,^25,33 and animals were perfused at 2 months of age. Microglial morphology,²⁶ myelin integrity,³⁴ and cortical thickness analyses were performed as previously described.³⁵

Patient peripheral blood samples were obtained with informed consent under either a protocol approved by the institutional review board of the University of California San Diego, or at Baylor College of Medicine. Informed consent was obtained from the parents of all study participants, and when applicable, assent was also obtained from patients. Bone marrow aspirates were remnants from consented clinically obtained samples from Baylor College of Medicine, also approved by the Baylor institutional review board.

Please see supplemental Methods for experimental details.

Results

Generation of BRAF^V600E iPSC lines and brain-relevant hematopoietic lineages identifies a proinflammatory BRAF^V600E state

We first sought to generate a human-specific model of the most common LCH MAPK mutation,⁴ BRAF^V600E. We generated iPSCs from patients clinically diagnosed with multisystem LCH and positive for BRAF^V600E within those hematopoietic samples. iPSCs were generated either from peripheral blood samples or remnants of clinically obtained bone marrow aspirates (Figure 1A-B). Isolated iPSC colonies were screened for BRAF^V600E (supplemental Figure 1A), with only 1 patient having identifiable BRAF^V600E iPSC colonies (Figure 1C). Reprogramming from somatic mosaic blood samples has only rarely been reported, so there is little precedent; potentially, the bias for control clones reflects the variability in cell types carrying the BRAF^V600E somatic mutation between individuals.³⁶ Therefore, we designed a homology-directed repair strategy with transient CRISPR/Cas9 transfection and a 200–base pair donor sequence to introduce the single–base pair V600E variant (BRAF c.1799T>A, p.Val600Glu) into patient-derived LCH iPSCs,^37,38 generating a cohort of isogenic lines (Figure 1C). Lines were karyotypically normal and, despite the BRAF^V600E variant, expressed both pluripotency markers and were capable of differentiation into all 3 primary germ layers (Figure 1B,D; supplemental Figure 1B).

Figure 1.

Establishment and validation of patient-derivediPSC carrying the LCH disease–causing BRAF^V600E mutation. (A) Schematic illustration of LCH pathology and establishment of BRAF^V600E and isogenic control stem cell lines from patient PBMCs. (B) Expression of pluripotency markers (NANOG, green channel; OCT4, red channel) by iPSCs. (C) Clinical details of patient-derived LCH samples used in developing the cell lines in this study. Variant allele frequency (VAF) of BRAF^V600E in the donor sample. (D) Differentiation of BRAF^V600E iPSC lines into all 3 primary germ layers and confirmation by immunofluorescence using layer specific markers. (E-F) BRAF^V600E and isogenic control iPSCs can differentiate equivalently into CD43⁺ hematopoietic stem cells as determined by flow cytometry. n = 3 cell lines per group, mean and standard error of the mean (SEM) are shown, unpaired 2-tailed t test. (G) Flow cytometry analysis of cell cycle stages of BRAF^V600E and control iHPCs. (H) Quantification of the percentage of cells in various cell cycle stages. (I) Microglial differentiation from iPSCs. (J) Expression of microglial marker genes by immunofluorescence (top) and flow cytometry (bottom). (K) High-throughput, automated image analysis of microglia (cell bodies [fuchsia] and ramifications [blue] pseudocolored overlays) showing reduced branch points and decreased process length in BRAF^V600E microglia (scale bar, 100 μm), consistent with an activated phenotype. Experiments were performed 3 times with 3 to 4 independent cell lines, and data represent mean and SEM, analyzed by 1-way analysis of variance with the Bonferroni multiple comparison test. ∗P < .05; ∗∗P < .01; ∗∗∗P < .001. CTRL, control; DAPI, 4′,6-diamidino-2-phenylindole; FITC, fluorescein isothiocyanate; ns, not significant; PBMCs, peripheral blood mononuclear cells. Figure created with BioRender.com. Coufal N. (2026). (A) https://BioRender.com/ovp5oi6; (D) https://BioRender.com/l58m2vr; (E,I) https://BioRender.com/1ypsn2b.

Clinical LCH lesion samples are cellularly heterogeneous,²⁴ such that isolated, defined BRAF^V600E carrying hematopoietic cells have not been available for study. We next queried the phenotype of hematopoietic progenitors (iHPCs) differentiated from iPSCs (Figure 1E), finding that all lines expressed the early multipotent hematopoietic marker³⁹ CD43 equivalently (Figure 1F). Given the clonal expansion in patients with LCH, we performed iHPC cell-cycle analysis, identifying an increased proportion of BRAF^V600E progenitors in S phase with a concomitant reduction in G0/G1 phase, suggesting early dysregulation in cell proliferation (Figure 1G-H).

Further differentiation of iHPCs with brain-specific small molecules and microglial growth factors has been shown to serve as a reasonable model of human microglia in vitro.⁴⁰ We next sought to query genotype-specific phenotypes (Figure 1I). We have previously shown in healthy controls that this differentiation closely recapitulates the yolk sac ontogeny of human microglia when iHPCs are subsequently xenotransplanted in the complex brain environment^25,33; however, BRAF^V600E may influence hematopoietic lineage determination. Resulting differentiated cells expressed canonical myeloid markers, including the transcription factor PU.1, which is critical for macrophage differentiation,^41,42 as well as protein ionized calcium-binding adapter protein 1 (IBA-1; Figure 1J). There was equivalent expression of CD45 and CX3CR1 across the cohort, with reduced CD11b in BRAF^V600E microglia (Figure 1J; supplemental Figure 2). Reduced CD11b has been associated with increased cytokine production (eg, tumor necrosis factor α [TNF-α]) and a proinflammatory state.⁴³ High-throughput automated imaging assessment of microglial morphology identified an ameboid morphology with a reduction in ramification complexity and process length in the BRAF^V600E genotype (Figure 1K).

Human microglia/myeloid cells from BRAF^V600E iPSCs share transcriptional characteristics with neurohistiocytosis

Next, we profiled both the patient-derived and CRISPR-edited iPSC-derived microglia (Figure 2A). A comparison to primary human microglial data sets,²⁵ healthy iPSC-derived microglia, as well as primary dendritic cell and monocyte populations⁴⁴ identified that BRAF^V600E microglia/brain myeloid cells did not more closely approximate dendritic cells or monocyte populations (Figure 2B). Using this combined cohort, we compared BRAF^V600E and control microglial/myeloid transcripts and identified 1020 differentially expressed genes (log₂ fold change of >2; adjusted P < .05; Figure 2C; supplemental Figure 3A) that differentiated the 2 isogenic genotypes. Gene ontology analysis of genes upregulated in BRAF^V600E microglia included hematopoietic lineage, cytokine production, especially TNF-mediated signaling, and, significantly, the positive regulation of the MAPK cascade (Figure 2D). Genes downregulated in BRAF^V600E microglia included a loss of inflammatory response regulation (Figure 2E), in keeping with the infiltrative and inflammatory nature of LCH-ND. BRAF^V600E genotype did not skew toward or away from a dendritic cell fate in this differentiation; overlap of differential genes, with gene sets enriched in microglia relative to dendritic cells (GSE29949), found no statistical overlap (logistic regression P = .67). Ingenuity analysis of upstream regulators identified key BRAF^V600E-upregulated regulators, including RIPK2 and NONO, both proinflammatory innate immune activators of the MAPK pathway^45,46 (supplemental Figure 3C-D). Canonical pathways upregulated in BRAF^V600E included Stathmin1, a MAPK target (supplemental Figure 3B,F). Overlay of differential genes with microglial state gene sets identified significant overlap specifically with interferon response microglia (supplemental Figure 3E) and, to a lesser extent, damage-associated human microglia.^30,33,47, 48, 49, 50, 51, 52, 53 Overarching themes included a loss of microglial homeostasis, elevated interferon response, and downregulation of lysosome-associated genes (Figure 2F). Finally, we performed a comparative analysis of previously published transcriptomic data from patient neurohistiocytosis brain lesions,¹⁵ finding significant gene overlap (Figure 2G) and a strong concordance between neurohistiocytosis brains and our differential genes using rank-rank hypergeometric overlap mapping (Figure 2H). Overall, these data recapitulate key transcriptional aspects of neurohistiocytosis among patients and define potential alternate upstream regulators and targets.

Figure 2.

BRAF^V600E microglia exhibit a unique transcriptional signature that recapitulates aspects of patients with neurohistiocytosis. (A) Transcriptomic study design, n = 1 to 3 biological replicates per line per clone. (B) PC analysis of microglial gene-expression patterns from this study (LCH^V600E and LCH control) with other control iPSC microglia, primary fetal microglia, and postnatal MG (Han et al²⁵) and monocytes and dendritic cells (Abud et al⁴⁴). (C) Volcano plot showing the relative difference in expression level between BRAF^V600E and control microglia (adjusted P < .05 and log₂ fold change [FC] > 2). (D) Pathway analysis of genes upregulated in BRAF^V600E. (E) Gene ontology of genes downregulated in BRAF^V600E MG. (F) Heat map of select differentially enriched genes in key pathways. FDR <0.05 and Log₂(FC) > 2. (G) Overlap of differential genes from this study and those of Mass et al¹⁵ from whole brain RNA sequencing of a patient with neurohistiocytosis, with some common genes identified. (H) Rank hypergeometric overlap of differential genes from BRAF^V600E microglia and those from a patient with neurohistiocytosis (Mass et al¹⁵) identifies strong concordance between the data sets, with 768 overlapping genes identified. GO, gene ontology; iMGs, iPSC microglia; MG, microglia; PC, principal component.

BRAF^V600E microglia/myeloid cells are MAPK-activated, apoptosis-resistant, and recapitulate neurohistiocytosis biomarkers and proinflammatory cytokines

The BRAF^V600E mutation results in a constitutively activated kinase,⁵⁴ leading to increased activation of the Ras/Raf/MEK/extracellular signal-regulated kinase (ERK) signaling ending in elevated phosphorylated ERK (phospho-ERK). We observed increased phospho-ERK in BRAF^V600E microglia (Figure 3A-B). RNA sequencing identified RIPK2 and NONO as possible upstream regulators in BRAF ^V600E microglia, both implicated in apoptosis resistance.^55,56 We queried cytochrome c levels in microglial media, finding elevated cytochrome c from BRAF^V600E microglia/myeloid cells, suggesting a basal level of apoptotic stress⁵⁷ (Figure 3C). We therefore queried, using a growth factor withdrawal caspase-3 assay, whether BRAF^V600E microglia/myeloid cells are apoptosis resistant (Figure 3D). Withdrawal of microglial survival factors resulted in increased apoptosis in healthy microglia within 24 hours; however, BRAF^V600E microglia/myeloid cells had no change in caspase-3 activity, suggesting apoptosis resistance (Figure 3E).

Figure 3.

Apoptotic priming and inflammation in neurohistiocytosis microglia. (A) BRAF^V600E microglia express higher phospho-ERK quantitated in panel B, n = 4 cell lines, repeated 3 times, mean and SEM are shown, unpaired 2-tailed t test. ∗P < .05; ∗∗P < .01; ∗∗∗∗P < .0001. (C) Cytochrome c is released in the media by BRAF^V600E microglia at baseline. (D) Schematic of the apoptotic priming assay. (E) Growth factor deprivation results in elevated control microglial apoptosis but no change in BRAF^V600E microglial apoptosis, expressed as percentage of caspase-3–positive microglia. Experiment performed 3 times, and the data represent the mean and SEM, analyzed by 2-way analysis of variance with Šidák multiple comparison correction. ∗∗∗∗P < .0001. (F) Live cell imaging of pHrodo-labeled myelin phagocytosis, brightfield, and pHrodo labeling (left). Time-lapse, high-content automated imaging analysis of pHrodo phagocytosis (right). (G) area under the curve, representing total cumulative myelin phagocytosis over 24 hours, 2-tailed t test. ∗∗P < .01. (H) Volcano plot of multiplexed cytokine enzyme-linked immunosorbent assay across 3 isogenic controls and BRAF^V600E microglia lines. Log2(FC) > 1 and −log10(P) > 1.3. (I) OPN and metalloprotease levels are elevated in media from BRAF^V600E microglia. (J) Elevated levels of TNF pathway inflammatory mediators, including resistin, ICAM-1, and TNFR1. Two-tailed t test. ∗P < .05. n = 4 lines per experiment. AUC, area under the curve; CTRL, control; DAPI, 4′,6-diamidino-2-phenylindole; EMMPRIN, extracellular matrix metalloproteinase inducer; ns, not significant; N.D., not detected; OPN, osteopontin; TNFR1, TNF receptor 1. Figure created with BioRender.com. Coufal N. (2026) https://BioRender.com/budzqyx.

In BRAF^V600E microglia/myeloid cells, reduced CD11b expression,⁵⁸ downregulated MAFB signaling in the ingenuity analysis,⁵⁹ and reduced phagosome formation in the canonical pathway analysis suggest phagocytic impairment. We found a significant reduction in myelin engulfment, as quantified by the area under the curve to capture total cumulative phagocytic activity over 24 hours (Figure 3F-G) as well as to other neurodegenerative stimuli (amyloid-β 1-42; supplemental Figure 4A), with preserved phagocytosis of bacterial stimuli (zymosan; supplemental Figure 4B). In an elevated cytokine proinflammatory environment with cellular death and myelin breakdown, reduced phagocytosis of myelin products is associated with impaired remyelination and worsening neurodegeneration.^60,61

Finally, we queried cytokine release by unstimulated BRAF^V600E microglia/myeloid cells. We focused especially on cytokines such as osteopontin, the only known cerebrospinal fluid biomarker for LCH-ND,⁶ and metalloproteases previously shown to be transcriptionally upregulated in LCH-ND brain lesions.^15,62,63 We used a highly multiplexed enzyme-linked immunosorbent assay for 187 protein targets (Figure 3H). We noted significantly higher levels of osteopontin (Figure 3I). We identified elevated release of extracellular matrix metalloproteinase (MMP) inducer (CD147), an upstream regulator of MMPs mediating leukocyte migration^64,65 and MMP induction66, 67, 68 (Figure 3I). Finally, we observed significantly higher levels of TNF mediators, although TNF trended without reaching statistical significance. This included resistin, a hormone that increases the production of proinflammatory cytokines, especially TNF⁶⁹; the TNF receptor 1, which is released by cells to bind and activate TNF; and ICAM-1, an adhesion molecule that is a downstream target of TNF⁷⁰ (Figure 3J). Altogether, the cytokine findings converge on osteopontin, metalloproteases, and TNF-mediated signaling, translating human LCH-ND findings.^6,15

Cerebral organoids identify the propensity of BRAF^V600E microglia to cause neurodegeneration

We hypothesized that BRAF^V600E microglia would cause neuronal death and potentially secondary astrogliosis. To study this interaction, we generated iPSC-derived 3-dimensional cerebral organoids.⁷¹ Cerebral organoids were cocultured with either control iPSC microglia or BRAF^V600E iPSC microglia differentiated from the same individual for 2 weeks to generate assembloids (Figure 4A). Single-cell RNA sequencing from assembloids^32,72 combined with molecular markers for cellular identity allowed for uniform manifold approximation and projection annotation (Figure 4B). This approach identified a combination of excitatory (SLC17A⁺) and inhibitory (GABRA2⁺) neuronal markers as well as radial glia at varying stages of differentiation. Finally, these organoids expressed markers for differentiating oligodendrocytes (PLP1⁺) and astrocytes (S100B⁺; Figure 4C).

Figure 4.

Neurodegeneration and reactive astrogliosis arise in organoids with BRAF^V600E microglia. (A) Schematic representation of the isogenic organoid-microglial assembloid experimental design. (B) UMAP showing cell annotations based on single-cell RNA sequencing data from assembloids consisting of 2 libraries of assembloids with control and 2 with isogenic controls cocultured with isogenic BRAF^V600E microglia (in total 10 787 cells). (C) Expression density of selected markers on UMAP. (D) Left: percentage distribution of cell types by condition. Center: log₂(FC) of percentage distribution based on the microglial genotype. Right: as on the left but showing only unique fractions by microglial genotype. The largest FC difference between the 2 microglial genotypes corresponds to the gain of astrocyte and immature cell types. (E) Expression of the reactive astrocyte signature. (F) Within the astrocyte cluster, elevated expression of reactive astrocyte markers. (G) Immunofluorescence imaging of assembloids, top neuronal (MAP2) and microglial (IBA-1) markers, bottom including caspase-3. (H) Quantitation of caspase by genotype, using 2 isogenic cell line assembloid cocultures and 3 to 5 organoids per cell line. IBA-1, ionized calcium-binding adapter protein 1; IPC, intermediate progenitor cell; MAP2, microtubule-associated protein 2; MG, microglia; MOL, myelinating oligodendrocyte; oRGC, outer radial glial cell; Pre-OPC, preoligodendrocyte precursor; RGC, radial glial cell; UMAP, uniform manifold approximation and projection. Figure created with BioRender.com. Coufal N. (2026) https://BioRender.com/fxxka10.

Next, we compared the abundance of each cell type stratified by genotype. With this approach, BRAF^V600E assembloids had more microglia/myeloid cells and astrocytes and a loss of myelinating oligodendrocytes and inhibitory neurons (Figure 4D). Given the astrocyte increase in BRAF^V600E assembloids, we queried the upregulation of reactive astrogliosis signature genes (38 genes extracted from the study by Matusova et al⁷³), finding that the BRAF^V600E assembloids had higher signature expression (Figure 4E). Specifically, within the astrocyte cluster, there were increases in astrogliosis genes such as CD44, GFAP, S100A10,⁷³ and SLC16A3 (Figure 4F). All organoids had microglia in proximity to neurons (Figure 4G), but there was an elevation of caspase-3, suggesting cell death in BRAF^V600E assembloids (Figure 4G-H). Altogether, these results suggest that microglial/myeloid expression of BRAF^V600E leads to astrogliosis and neuronal loss in a human brain model.

A humanized murine xenotransplantation model exhibits BRAF^V600E-dependent microgliopathy and human pathogenic features of neurohistiocytosis

Tissue-resident macrophages, especially microglia, are dependent on the brain environment for their cell type–specific gene expression programs.^30,74 Humanized murine xenotransplantation models for iPSC-derived hematopoietic progenitors have been shown to closely recapitulate the healthy human in vivo microglial gene signature.^25,33 We hypothesized that xenotransplantation of BRAF^V600E iHPCs would provide the complex brain environment necessary to phenocopy neurohistiocytosis pathology. Using a murine model amenable to human microglial xenotransplantation (Rag2^−/−Il2rg^−/−CSF1^h/hCsf1r^{ΔFIRE/ΔFIRE25,75}; Figure 5A), we found cells distributed throughout the brain, but a marked lack of microglial tiling with BRAF^V600E (Figure 5B; supplemental Figure 5A). Within the BRAF^V600E genotype, we noted striking microgliopathy and hyperproliferation with increased human myeloid cell density throughout numerous brain areas (Figure 5C; supplemental Figure 5A; supplemental Table 2). In keeping with the in vitro ameboid phenotype, BRAF^V600E myeloid cells exhibited reduced ramification complexity and increased soma size (Figure 5D-E).

Figure 5.

Microglial dystrophy in BRAF^V600E microglia in an in vivo model. (A) Schematic representation of the xenotransplantation approach. (B) Immunofluorescence for myeloid (IBA-1) marker and human nuclei (KU80). (C) Quantitation of microglial density across brain regions. One-way analysis of variance with Bonferroni multiple comparison test correction. (D) Assessment of microglial branch points (left) and the number of end points (right). (E) Quantitation of soma size (perimeter). (F) CD11c immunofluorescence, quantitated in panel G. (H) Quantitation of HLA-DR immunofluorescence, represented in panel I. (J) CD163 immunofluorescence. (K) Quantitation of CD163 intensity. (D-K) Two-tailed t test. Each point throughout represents an independent mouse brain, averaged over 3 to 5 sections imaged per mouse. ∗P < .05; ∗∗P < .01; ∗∗∗P < .001; ∗∗∗∗P < .0001. a.u., arbitrary unit; CTRL, control; DAPI, 4′,6-diamidino-2-phenylindole; HLA-DR, human leukocyte antigen-DR isotype; IBA-1, ionized calcium-binding adapter protein 1; KU80, human nuclei marker; PN, postnatal. Figure created with BioRender.com. Coufal N. (2026) https://BioRender.com/budzqyx.

We next queried the differentiation and activation state of BRAF^V600E myeloid cells in the brain environment. Unlike healthy control cells that follow a microglial ontogeny with this xenotransplantation approach,^25,33 BRAF^V600E myeloid cells exhibited neurohistiocytosis markers consistent with an inflamed myeloid state. This included extensive expression of CD11c throughout the brain (Figure 5F-G). CD11c is upregulated in peripheral LCH lesions and in bone marrow LCH dendritic cell precursors^1,16 and in the brain is associated with activation/inflammation in addition to specific developmental stages.⁷⁶ Similar to human neurohistiocytosis patient pathology, BRAF^V600E myeloid cells exhibited extensive upregulation of class II major histocompatibility complex (HLA-DR¹²; Figure 5H-I). Within the healthy brain, CD163 expression is generally restricted to border-associated macrophages, such as those within the meninges.⁷⁷ However, the scavenger receptor CD163 can be upregulated in inflamed microglia⁷⁸ and is extensively associated with LCH-ND pathological lesions.^2,6,79 We and others have shown that in this transplantation model, xenotransplanted cells within the meninges and border regions also express CD163, whereas differentiated microglia within the parenchyma do not.^25,33 Consistent with this, we found extensive global upregulation of CD163 in BRAF^V600E myeloid cells throughout the brain parenchyma (Figure 5J-K). Finally, expression of homeostatic microglial marker P2RY12 was markedly decreased in BRAF^V600E brain myeloid cells (supplemental Figure 5C). Overall, we found LCH-ND–specific upregulation of antigen presentation and inflammatory phenotypes that do not signify a specific myeloid ontogeny but recapitulate the inflammatory state of LCH-ND.

Macrophage heterogeneity leads to both proliferation and cellular senescence with brain region–specific foci in neurohistiocytosis

Given the marked myeloid pathology in animals that had received BRAF^V600E transplants, we hypothesized an underlying propensity for increased myeloid proliferation. Previous murine models have also found an increase in LCH-associated macrophage cellular senescence and a senescence-associated secretory profile, such that mice lacking Cdkn2a-p16 (INK-ATTAC) partially compensated for the peripheral LCH multisystem phenotype.¹³ We queried whether these cellular states are reflected in this xenotransplantation model. We first assessed brain macrophage cell division with colocalization of proliferative marker Ki-67 with human nuclei marker (Ku80) in adult animals that had received xenotransplants, finding a small but significant increase in animals that had received BRAF^V600E transplants (Figure 6A-B), suggesting a hyperproliferative state underlies an aspect of LCH-ND myeloid dysplasia. Second, we queried senescence markers, finding elevation of senescence-associated β-galactosidase in animals that had received BRAF^V600E xenotransplants (Figure 6C-D). We orthogonally validated this with p21, a canonical senescence marker⁸⁰ (Figure 6E-F).

Figure 6.

BRAF^V600E microglia exhibit marked heterogeneity in vivo with granuloma formation. (A) Proliferative human microglia (Ki67) are rarely noted. (B) Quantitation of rare microglial proliferation. (C) Immunohistochemistry for senescence-associated β-galactosidase (SAβGal). (D) Quantitation of β-galactosidase–positive cells. (E) Immunofluorescence for senescence marker p21. (F) Quantitation of p21⁺ human myeloid cells. (G) Immunofluorescence for pERK identifies rare positive cells within the BRAF^V600E-engrafted mice. (H) High-resolution images of a CD1a⁺ lesion in the ventral caudate nucleus, and hematoxylin and eosin (H&E) staining of hippocampal lesions of BRAF^V600E-engrafted murine brain. (I) Granuloma-like areas were markedly positive for CD1a and CD207, whereas most BRAF^V600E microglia were not and had higher expression of HLA-DR. (J) Immunohistochemistry from CNS biopsy with heterogeneous nodules seen on H&E staining, with nodular BRAF^V600E infiltrates, positive for CD207/Langerin as well as patchy, occasional CD1a positivity. Student t tests were performed for analysis of group means in panels with quantitation. Each point throughout represents an independent mouse brain, averaged over 3 to 5 sections imaged per mouse, 2-tailed t test. ∗P < .05; ∗∗∗∗P < .0001. CTRL, control; DAPI, 4′,6-diamidino-2-phenylindole; β-Gal, β-galactosidase; HLA-DR, human leukocyte antigen-DR isotype; IBA-1, ionized calcium-binding adapter protein 1; pERK, phosphorylated ERK.

Patients with LCH-ND exhibit marked heterogeneity in their lesions, with hyperdense foci and cellular accumulations, especially in the basal ganglia. Infiltrative and diffuse lesions are classically predominantly negative for CD1a and CD207, the canonical markers for peripheral LCH lesions, leading to confusion in LCH-ND’s etiology.^2,5,10 In querying the ability of clonal BRAF^V600E hematopoietic lineage cells to recapitulate this heterogeneity, we interestingly noted occasional rare cells positive for phospho-ERK, despite most xenotransplanted myeloid cells being negative (Figure 6G). We similarly noted foci positive for senescence markers, especially in the ventral basal ganglia (supplemental Figure 5B). High-resolution image of CD1a⁺ lesions in the mouse caudate nucleus showed intense cell-surface staining (3,3′-diaminodenzidine) and neoplastic histiocytic morphology strikingly similar to human brain biopsy lesions (Figure 6H-J). The number of lesions also varied among mice, highlighting heterogeneity, with rare foci in the basal ganglia and subiculum positive for CD1a and CD207 and highly HLA-DR positive and composed of dense human myeloid cell infiltrates (Figure 6I). These findings precisely phenocopy the lesional diversity of LCH-ND, which can include both diffuse infiltrates (CD1a and CD207⁻), whereas foci or rare CD1a⁺ cells are noted in some cases.^6,12,79 These foci are sometimes positive for BRAF^V600E by polymerase chain reaction or for phospho-ERK⁶ with diffuse infiltrates that do not express these peripheral LCH markers (Figure 6J).⁸¹ To further delineate the anatomical distribution of these BRAF^V600E agglomerations, we performed granular mapping of human myeloid infiltrative lesions (<50 μm²) across brain nuclei and substructures (supplemental Figure 5D). Although the forebrain and basal ganglia contained distinct, quantifiable focal aggregates, we observed a marked rostro-caudal increase in myeloid cell density. Of particular note, the density within the caudal midbrain tegmentum (including red nucleus and substantia nigra) and the hindbrain (pons and medulla) was extremely high. This robust increase in motor-critical regions resulted in a confluent pathological state that precluded the identification of individual nuclei or well-defined, isolated agglomerates. Such dense rostro-caudal infiltration mirrors the severe end of the pathological spectrum and the specific involvement of the brainstem and cerebellum observed in clinical neurohistiocytosis cases.

BRAF^V600E human myeloid cells cause astrogliosis, demyelination, and neuronal loss associated with behavioral deficits

We hypothesized that this inflammatory human myeloid state would lead to neurodegenerative phenotypes, using adult animals 2 months after xenotransplantation to assess for injury phenotypes (Figure 7A). First, we analyzed the degree of astrogliosis, using GFAP normalized to astrocyte number (SOX9⁺ nuclei). We noted a marked increase in GFAP⁺ murine astrocytes in the BRAF^V600E-transplanted brains (Figure 7B-C) across multiple brain areas. This was independent of astroglial density, which was increased but not statistically significant (Figure 7D). We subsequently interrogated myelination, assessing both the degree of cortical myelination and myelin basic protein immunointensity, which were decreased in animals that had undergone BRAF^V600E transplantation (Figure 7E-F). Finally, we assessed cortical layer thickness, finding a reduction specifically in the metabolically active large projection neurons of layer 5 (Figure 7G-H).⁸² To establish the translational relevance of these pathological findings, we performed functional assessments. Pathogenic mice displayed significant weight loss beginning at 5 weeks after transplantation (Figure 7K). Consistent with the marked loss of calbindin-positive Purkinje neurons in the cerebellum (Figure 7I-J), a hallmark often associated with clinical ataxia in patients with LCH-ND,¹² BRAF^V600E mice exhibited significant motor deficits. Mice that had undergone BRAF^V600E transplantation displayed reduced ability to balance on the rotarod (Figure 7L), along with significantly higher limb clasping scores and gait abnormalities, including reduced stride length in the right hind limb (Figure 7M-O). These neurological manifestations align with the ataxia reported in the Csf1r^MeriCreMer; Braf^V600E model.¹⁵ Finally, we assessed the impact of BRAF^V600E myeloid engraftment on survival, finding all control animals survived up to the terminal time point (Figure 7P). However, mice that had undergone BRAF^V600E transplantation exhibited severe deterioration characterized by progressive hindlimb paresis, malnutrition, and dehydration, resulting in euthanasia when reported moribund. In sum, BRAF^V600E myeloid cells caused marked astrogliosis with marked loss of neurons and myelin and motor abnormalities consistent with LCH-ND.

Figure 7.

BRAF^V600E microglia cause neurodegeneration and associated behavioral abnormalities. (A) Schematic. (B) Immunofluorescence for astrocyte markers GFAP and SOX9, and microglial IBA-1. (C) Quantification of astrogliosis as the number of GFAP SOX9 double-positive cells across brain regions. One-way analysis of variance (ANOVA) with Bonferroni correction. (D) Total astrocyte density as the number of SOX9-positive cells. Two-tailed t test. (E) Immunofluorescence for myelin basic protein (MBP). (F) Quantification of the percentage of myelinated cortex (left) and the MBP intensity (right) by a 2-tailed t test. (G) Cortical layer identification through immunofluorescence colabeling for NeuN and Vglut2. (H) Quantitation of cortical layer thickness, 1-way ANOVA with Bonferroni correction. (I) Cerebellar immunofluorescence for Purkinje neurons (Calbindin), myelin (MBP), and microglia (IBA-1). (J) Quantification of Purkinje neuronal density, 2-tailed t test. In all experiments, each point represents a separate animal that had undergone xenotransplantation, with 3 to 5 sections averaged per animal. (K) Behavioral assessment identifying progressive weight loss in BRAF^V600E mice, (L) impaired rotarod performance, (M) higher limb clasping scores, and (N-O) abnormal gait symmetry. (P) Kaplan-Meier curves of probability of survival of engrafted control (n = 6) and BRAF^V600E mice (n = 10); log-rank (Mantel-Cox) test. Weights were analyzed using a 2-way ANOVA mixed-effects model (interaction F(7, 82) = 9.962; P < .0001) followed by Šidák multiple comparison (panel K, group means at each weigh time point), and 2-tailed t test for panels L-M,O, each data point representing an individual animal). ∗P < .05; ∗∗P < .01; ∗∗∗P < .001. CTRL, control; IBA-1, ionized calcium-binding adapter protein 1; ICV, intracerebroventricular; NeuN, neuronal nuclei; Vglut2, vesicular glutamate transporter 2. Figure created with BioRender.com. Coufal N. (2026) https://BioRender.com/budzqyx.

Discussion

In this study, we developed a human iPSC model that replicates many of the critical mechanistic features of LCH-ND. To achieve this, we generated patient-derived iPSC lines carrying the pathogenic BRAF^V600E mutation through a combination of mixed population reprogramming and CRISPR/Cas9 gene editing. By studying the effects of this pathogenic variant that induces constitutive MAPK pathway activation on brain myeloid cell function and immune recruitment, we deepened our understanding of LCH-ND’s underlying pathophysiology. These approaches validate our model as a potential preclinical platform for testing therapeutic strategies, building on the only previous iPSC LCH model, which focused on early myeloid and peripheral lesional CD1a and CD207 modeling.

One hallmark of LCH is dysregulated histiocytic proliferation. We found that hematopoietic progenitors derived from BRAF^V600E iPSCs exhibited increased proliferation, consistent with this phenotype. However, our findings diverge from previous studies, which reported reduced proliferation in human bone marrow hematopoietic cells with enforced BRAF^V600E expression.¹⁷ This discrepancy may be attributed to compensatory mechanisms involving intact wild-type alleles in human hematopoietic cells with forced BRAF^V600E expression applied in previous studies, environment-specific interactions in transplantation, or related to the differentiation stage of iPSC hematopoietic progenitors, which may represent a more primitive yolk sac–derived state than bone marrow–derived CD34 progenitors.³⁹

To further investigate the impact of the BRAF^V600E mutation, we undertook xenotransplantation experiments. Engraftment of human hematopoietic progenitors into humanized immunodeficient brains resulted in extensive infiltration of BRAF^V600E microglia/myeloid cells in the CNS in adulthood. These brain myeloid cells were dysplastic and lacked normal brain microglial tiling. LCH-ND pathology is characterized by abnormal activation and clonal infiltration of BRAF^V600E cells predominantly affecting the white matter tracts, subcortical regions, and cerebellum. An increase in overall myeloid density was observed in a rostro-caudal fashion, in line with other models of BRAF^V600E-induced neurodegeneration.^13,15 Under homeostatic conditions, the myeloid capacity for self-renewal maintains a steady-state population without dependence on circulating monocytes.^83,84 Interestingly, although the number of actively proliferating BRAF^V600E myeloid cells was elevated, a significant proportion of the resulting brain myeloid cells exhibited signs of senescence. Senescent myeloid cells were more evenly distributed and abundant, in contrast to the proliferating cells that appeared sparse across the CNS. This suggests that the initial hyperproliferative phase is followed by a shift toward senescence, potentially contributing to the long-term consequences of LCH-ND.

The origin of LCH-ND has been debated,^13,15,85,86 with models demonstrating erythromyeloid and hematopoietic precursors with potential to develop into brain myeloid cells. This is complicated by patient and lesional heterogeneity, LCH-ND development years after systemic remission, and rarely CNS-limited LCH-ND. By xenografting human BRAF^V600E hematopoietic progenitors into a humanized neonatal murine model, we delineated the contribution of brain-resident intraparenchymal myeloid cells toward recapitulating LCH-ND pathology without the confounding effects of late peripheral BRAF^V600E myeloid cell or adaptive immune infiltration. A limitation of this approach is the inability to differentiate between these ontological questions, but this approach is powerful in identifying the effect of the brain environment on BRAF^V600E myeloid cells and especially myeloid heterogeneity. In this way, we recapitulated patient LCH-ND biomarkers, including osteopontin expression,⁶ pathological characteristics such as globally elevated CD163^6,79 and antigen presentation,^1,12 and elevated metalloproteases characteristic of peripheral LCH lesions.¹

Peripheral LCH infiltrative lesions, comprising CD1a⁺ and CD207⁺ myeloid cells, are a hallmark of LCH. In the brain, LCH-ND lesions are more heterogeneous, encompassing both diffuse and multifocal lesions that are only rarely CD1a⁺ or CD207⁺.^6,12,79 The nodular or foci identified in LCH-ND¹² are not recapitulated in any LCH models, so it has previously been unclear whether it is driven by the dysplastic myeloid clone or other immune cell infiltrates. We identified numerous agglomerations of engrafted human myeloid cells in our xenografted brains. These foci are sometimes densely packed with cells that are positive for CD1a and CD207, whose expression is otherwise absent throughout the xenotransplanted brain and infiltrating myeloid cells.

A significant finding of our study is the rapid onset of neurodegenerative pathology and associated functional deficits in the humanized BRAF^V600E model. Although behavior abnormalities were reported in other murine models, those manifestations typically did not emerge until ∼8 months of age because of the peripheral origin of the pathogenic mutation (bone marrow). In striking contrast, our model exhibited deficits as early as 2 months after transplantation, followed by progressive motor deficits. This markedly accelerated timeline highlights the potency of direct engraftment of BRAF^V600E myeloid progenitors into the neonatal brain environment and limits our experimental window to investigate therapeutic interventions. This limitation may be addressed in future mixed-engraftment or late BRAF^V600E introduction to more accurately mirror the somatic mosaicism and progressive evolution of clinical LCH-ND and provide the necessary experimental window to evaluate the efficacy of targeted therapies, such as MAPK pathway inhibitors or senolytics, in mitigating disease burden and preserving neurologic function.

Results from our xenografted mouse brain lesions share similarities with patients, consolidating the translational validity of our humanized mouse models of LCH-ND. LCH-ND remains difficult to treat and is often refractory and progressive. Our study strongly suggests this model may provide critical insights into treatment approaches to eradicate heterogeneous combinations of proliferative and senescent myeloid cells and the opportunity for mechanistic questions into myeloid niche–specific interactions.

Limitations

Engrafting a pure population of a single genotype of brain myeloid cell type limits analysis of immune-cell interactions. LCH represents a somatic BRAF^V600E mutation; therefore, true human disease replication would require mixed brain myeloid genotypes with genetic tracing, given the likelihood of myeloid cell responses to BRAF^V600E myeloid cells. BRAF^V600E brain myeloid cells exhibited early proliferation and senescence phenotypes with substantial heterogeneity; therefore, these studies do not discern which BRAF^V600E brain myeloid states define most of the disease pathology. Finally, these approaches cannot differentiate between infiltrating myeloid cells and yolk sac origin in LCH. Bone marrow transplant of iPSC hematopoietic progenitors was described recently⁸⁷ and may eventually allow exploration of these areas, although the murine genetics used herein lack the proper genetics for successful bone marrow transplant.

Conflict-of-interest disclosure: The authors declare no competing financial interests.