Role of clonal inflammatory microglia in histiocytosis-associated neurodegeneration

Summary:

Langerhans cell Histiocytosis (LCH) and Erdheim-Chester disease (ECD) are clonal myeloid disorders, associated with MAP-Kinase activating mutations and increased risk of neurodegeneration. We found microglia mutant clones in LCH and ECD patients, whether or not they presented with clinical symptoms of neurodegeneration, associated with microgliosis, astrocytosis, and neuronal loss, predominantly in the rhombencephalon grey nuclei. Neurological symptoms were associated with PU.1⁺ clones size (p= 0.0003), in patients with the longest evolution of the disease, indicating a phase of subclinical incipient neurodegeneration. Genetic bar-coding analysis suggest that clones may originate from definitive or yolk sac hematopoiesis depending on patients. In a mouse model, disease topography was attributable to a local clonal proliferative advantage, and microglia depletion by a CSF1R-inhibitor limited neuronal loss and improved survival. These studies characterize a neurodegenerative disease associated with clonal proliferation of inflammatory microglia. The long preclinical stage represents a therapeutic window before irreversible neuronal depletion.

eTOC Blurb

Vicario et al., detected microglia mosaicism in Histiocytosis patients, predominant in the rhombencephalon grey nuclei and associated with astrocytosis, neuronal loss, and neurological symptoms in patients with the larger clones. These studies characterize a clonal microglia-associated neurodegenerative disease with a long preclinical stage / therapeutic window before irreversible neuronal depletion.

Graphical Abstract

Introduction

Langerhans cell histiocytosis (LCH)¹ and Erdheim-Chester disease (ECD) ^2,3, are rare clonal myeloid disorders, with an estimated incidence of ~5 cases per million people per year ⁴. They are sometimes co-diagnosed in the same patients⁵, and both associated with a high prevalence of the BRAF^V600E mutation^1–3,6–8 as well as an increased risk of neurodegenerative disease (neuro-histiocytosis)^1,3,9–11 and other hematologic and solid malignancies^12–14. LCH and ECD were grouped under the name of “L-histiocytosis”⁵, but their clinical presentation typically differs and LCH is predominantly a pediatric disease while ECD is diagnosed in adults⁵. In both LCH and ECD however, neuro-histiocytosis is characterized by progressive symmetric cerebellar syndrome, tetra-pyramidal syndrome with or without motor deficits, pseudobulbar palsy, and cognitive and behavioral impairment ^{10,11,15–17}. Neurodegeneration can occur decades after the original diagnosis of histiocytosis, in patients considered « in remission » ^1,10,15, albeit it can also be the initial presentation of the disease ³. Brain MRI abnormalities, frequently non-specific, include T1 hypersignal of deep pons and cerebellar grey nuclei, demyelination and atrophy located preferentially to posterior fossa ^3,15–19. Full brain pathological examination was only reported in 3 cases in the medical literature which reported neuronal loss, gliosis, and demyelination²⁰. The mechanism of histiocytosis-associated neurodegeneration is poorly understood^15,16,20, and its prognosis is poor, in the absence of a curative treatment ^3,4,9. The cell of origin is also an outstanding question. Mutant bone marrow hematopoietic progenitors can invade the brain, giving rise to mutant microglia-like cells ^21,22 but a mutation in yolk sac progenitors for pro-definitive hematopoiesis, which generate self-renewing tissue macrophages including microglia ^23–25, also gives rise to mutant microglia and neurodegeneration in mice, in the absence of bone marrow-derived clones²⁶.

We reasoned that understanding of the natural history of the disease at the molecular level would help in the identification of novel therapeutic strategies and targets. Therefore, we undertook a comprehensive and systematic molecular and pathological analysis of the brains from a series of LCH and ECD patients. We report here that histiocytosis-associated neurodegeneration is linked with the clonal proliferation of inflammatory microglia and neuronal death, and predominates in the brainstem grey nuclei, cerebellum and hippocampus. A genetic bar-coding analysis suggest that mutant clones originate from either definitive or possibly from yolk sac hematopoiesis depending on patients. Importantly, molecular, and histological signs of neurodegeneration can be detected within 2 years of the initial diagnosis of histiocytosis and in the absence of clinical signs of neurodegeneration, which can appear 8 to 25 years after the initial diagnosis and is associated with a larger size of the microglial clones, irrespective of their cellular origin. These results identify a stage of incipient neurodegeneration which may represent a therapeutic window. Accordingly, we found in mouse models that microglia depletion during this pre-clinical phase limits neuro-inflammation and neuronal death and improves survival.

Results

Pervasive BRAF^V600E PU.1⁺ clones in the brain of LCH and ECD patients

We studied 8 consecutive patients (Table 1) diagnosed with pediatric onset LCH (n=2), adult onset mixed LCH/ED (n=2), or adult onset ECD (n=4) based on BRAF^V600E positive lesions, for whom post-mortem whole brain (in 7 cases) or brain biopsy (n=1, Patient#2) and blood or bone marrow samples were obtained after informed consent for the purpose of this study (Table S1). Four patients were diagnosed with neuro-histiocytosis in the course of the disease and the other 4 were free of neurological symptoms (Table 1 and Extended Data Patients). Brain and blood samples from 35 neuro-typical individuals without histiocytosis (Table S1) were also studied as controls. Blood or Bone Marrow (BM) samples and nuclei suspensions from frozen brain tissue sampled from the frontal cortex to the spinal cord, and FACS sorted into PU.1⁺ myeloid nuclei, NeuN⁺ neuronal nuclei²⁷, and PU.1⁻ NeuN⁻ (DN) stromal/immune cell nuclei (Figure 1A and Figure S1A), were subjected to targeted deep sequencing (²⁸, see Methods, Table S2, S3) at an average depth of ~1100X (Figure S1B), and to droplet digital PCR (ddPCR) (Table S3). BRAFc.1799T>A (corresponding to BRAF^V600E) was the variant most frequently detected, present in multiple PU.1⁺ brain samples, from all histiocytosis patients (Figure 1B and Table S3). BRAF^V600E detection was confirmed by ddPCR (Table S3). In contrast, BRAF^V600E was not detected in NeuN⁺ and DN samples from histiocytosis patients and in NeuN⁺, PU.1⁺ and DN samples from controls (Figure 1C, Figure S1C and Table S3). In a mixed-effects linear regression model with donor as random effects and age, sex, sequencing depth, anatomical locations of the sample, and neurological status incorporated as fixed effects, the diagnosis of histiocytosis was the only factor associated with BRAF^V600E in PU1 cells (P < 0.0167), with a positive association between BRAF^V600E PU1 nuclei and histiocytosis (p=0.002).

Table 1. Characteristics of patients.

	DX	Sex	BRAF V600E	Tissues analyzed	Age at DX LCH/ECD	Age at DX of neuro-histio	Age at death	Delay between DX of LCH/ECD and neuro-histio * or death from other cause ✞	Brain MRI	Extra-neurological symptoms of histiocytosis	Other diseases	Cause of death
#1	LCH	M	+	Brain, blood	0.5	13	21	12 y *	+	Skin, bone	None	Neuro-histio
#2	LCH	F	+	Cerebellar biopsy, blood	0.5	16	Alive at 26	15 y *	+	Skin, bone	None	Alive with severe neuro-histio
#3	LCH/ECD	M	+	Brain, blood	52	77	78	25 y *	+	Skin, bone, heart, lung	CH	Neuro-histio
#8	ECD	F	+	Brain, bone marrow	58	66	69	8 y *	+	Bone, Retro-Peritoneal, heart	Metastatic breast cancer, CH	Neuro-histio sepsis
#4	ECD	M	+	Brain, bone marrow	59	-	66	7 y ✞	+$	Skin, bone, heart, lung, Retro-Peritoneal	Pancreatic cancer, CH	Pancreatic cancer
#5	LCH/ECD	F	+	Brain, bone marrow	75	-	78	3 y ✞	+$	Sclerosing cholangitis, Skin, bone, heart, Lung.	None	Hepatic insufficiency
#6	ECD	M	+	Brain, bone marrow	59	-	61	2 y ✞	+$	Skin, bone	CH T cell lymphoma	T cell lymphoma
#7	ECD	M	+	Brain, bone marrow	85	-	89	4 y ✞	-	Bone, Retro-Peritoneal	Kidney cancer, Prostate cancer, CH	Sepsis

DX: Diagnosis (LCH: Langerhans Cell Histiocytosis; ECD: Erdheim-Chester Disease); CH: Clonal hematopoiesis; nd: not determined; y: years.

^*Delay between diagnosis of LCH/ECD and diagnosis of Neuro-histio.

^✞Delay between diagnosis of LCH/ECD and death from other causes.

^$:non-specific hyperintensities in the pons / cerebellum identified in posteriori review of MRI.

Figure 1. Detection of mutations in brain and matching blood or bone marrow from histiocytosis patients.

(A) Left, schematic of post-mortem brain samples obtained from patients. Right, representative flow cytometry dot-plots of brain nuclei from patient #1 and labeled with anti-NeuN and anti-PU.1 antibodies (% of total). (B) Oncoplot represents mutated genes (with 4 or more mutant reads), number of mutations per sample and % of samples carrying mutations in PU.1+ samples (n=71) and matching blood or bone marrow samples (BM) (n=12) from Histiocytosis patients (n=8). (C) Variant allelic frequency (VAF, %, HemePACT) for BRAFc.1799T>A (V600E) in PU.1+, NeuN+, DN (PU.1-,NeuN-) and blood or bone marrow samples from Histiocytosis patients (71 brain samples from 8 patients), and controls (104 samples from n=35). Each dot represents a sample. Statistics: p-value was calculated using a mixed-effects linear regression model (see methods).

These results indicated that the BRAF^V600E mutation was specific for patients in comparison to controls, and -surprisingly-pervasive among PU.1+ nuclei samples across the patients’ brain, even in the absence of clinical neurological symptoms.

Overt and incipient neurodegeneration in histiocytosis patients

Neuropathological analysis of corresponding brain samples (see Extended Data Patients) indeed revealed the presence of histological lesions in all patients, including the 4 patients without clinical neurodegeneration (Figure 2A, red and purple arrows). Histological lesions were found in the anatomical areas where the BRAF^V600E mutation was identified (Figure 2A, red arrows) and consisted in focal, non-systematized, areas of microglial (IBA1+) and astrocyte (GFAP) activation, and neuronal loss in the grey matter (Figure 2B) and axonal spheroids in the white matter. The pons, cerebellum, and hippocampus were the most frequently affected, while the cortex was rarely involved (Figure 2A, B and Figure S2A). Differential gene expression and pathway analyses of whole-tissue RNAseq from patients and control brains showed upregulation of inflammatory and phagocytic signatures including complement, IL1, and phagocytic receptors, mainly driven by the patient’s brainstem and cerebellum samples, and down-regulation of genes associated with neuronal and synaptic activity (Figure 2C, D and Table S4). Of note, a posteriori review of available brain MRIs showed the presence of nonspecific hyperintense signals in the dentate nuclei of 3 of the 4 patients without clinical symptoms, reminiscent of nonspecific hyperintense signals found in the pons, dentate nuclei, and cerebellum of patients with neurological symptoms (Figure S2B).

Figure 2. Histological and molecular analysis of the brain of histiocytosis patients.

(A) Schematic of the brains of the 8 patients annotated for the detection of BRAFc.1799T>A (V600E) and/or of histological signs of neurodegeneration (Histo+). Bar graphs represents the proportion of tested brain samples positive for BRAFc.1799T>A (V600E) by HemePACT and/or histological signs of neurodegeneration among patients with (left) or without (right) neurological symptoms. (B). Left, representative H&E, IBA1 (microglia marker) and GFAP (astrocyte marker) of pons (top) and cerebellum (bottom) from patient #1 and an age-match control for comparison. Right, representative IBA1 and GFAP of pons from patient #8, #4, #5, #7 and an age-match control for comparison. Arrows indicate neuronal nuclei. (C) Pathway enrichment among differentially expressed genes (DEG, red: upregulated genes, blue: downregulated genes) by RNAseq analysis (FDR <0.05, log2FC >= or <= 1.5/−1.5) of brain tissue from Histiocytosis patients (n=13) and controls (n=11) using g:profiler webtool. Pathways are selected based on FDR <= 0.05 and ordered by significance. (D) Hierarchical clustering of DEG (log2FC >= 1.5, log2FC <= −1.5, FDR <0.05) between brain samples from Histiocytoses (n=13) and controls (n=11). Expression values are Z score transformed. (E) Bar graphs represents the proportion of tested brain samples positive for BRAFc.1799T>A (V600E) by HemePACT and/or histological signs of neurodegeneration among patients with (left) or without (right) neurological symptoms. (F) Variant allelic frequency (VAF, %, HemePACT)) for BRAFc.1799T>A (V600E) in PU.1+, samples from patients with (red) and without (blue) neurological symptoms. Each symbol represents a patient. Statistics: p-value is calculated using a mixed-effects linear regression model (see methods, and Fig.S2E).

These data indicated that the patients with pediatric onset LCH and adult onset ECD presented with qualitatively similar molecular and cellular signs of microglial activation and neurodegeneration, whether they presented or not with neurological symptoms, and therefore that patients without clinical symptoms presented with ‘incipient’ BRAF^V600E-associated neurodegeneration.

Preclinical neurodegeneration characterized by BRAF^V600E-mediated clonal expansion.

As shown above, areas of histological neuroinflammation and neuronal loss overlapped with molecular detection of the BRAF^V600E variant in the same regions (red arrows, Figure 2A), however, this overlap was more consistent in patients with clinical neuro-histiocytosis than in patients without neurological symptoms (Figure 2A, E and Figure S2C). In addition, analysis by 2 neuropathologists determined that histological damage in the pons was more intense in patients with clinical symptoms (Figure S2D). More quantitatively, a mixed-effects linear regression model analysis with donor as random effects and age, sex, sequencing depth, anatomical locations of the sample, and neurological status, incorporated as fixed effects showed that only neurological status and age were significant predictors of the size of the BRAFV600E clone (P = 0.007 and 0.041, respectively) (Figure 2F and Figure S2E). Of note, the delay between the initial diagnosis of LCH or ECD and the diagnosis of neuro-histiocytosis was of 8 to 25 years, while the 4 patients without clinical neurological symptoms died from other causes 2 to 7 years after diagnosis of ECD (Table 1).

These results altogether demonstrate that patients develop molecular and histological features of incipient neurodegeneration within a few years from the initial diagnosis of histiocytosis and possibly decade(s) before development of clinical symptoms, while clinical symptoms and the severity of histological changes, correlate with the size of the BRAF clone, suggesting progressive damage as mutant clones expand.

Preferential proliferation of BRAF^V600E microglial clones in the mammalian rhombencephalon

As shown above, detection of BRAF^V600E clones, neuronal death, astrogliosis, and microgliosis, and the neuroinflammatory signature predominated in the patients’ hippocampus, brainstem and cerebellum (rhombencephalon) (Figure 2A), in accordance with the classical cerebellar syndrome, pseudobulbar palsy, and cognitive and behavioral impairment reported in neuro-histiocytosis ^3,11,15,16. To explore the mechanisms involved in the anatomical topography of the clonal process, we performed an analysis of the allelic frequency (AF) distribution of the BRAF^V600E variant as a function of the location of samples, from the frontal cortex to the medulla oblongata. Results showed that the AF of BRAF^V600E clones by HemePACT as well as by ddPCR increases along a rostro-caudal gradient (r =0.75, p=0.005, and r= 0.7, p= 0.0089 respectively, Figure 3A a Figure S3A). This increased AF of BRAF^V600E clones along a rostro-caudal gradient was observed in patients whether they presented with neurological symptoms or not (r= 0.67, p: 0.01, r= 0.8, p= 0.002, Figure 3A).

Figure 3. Analysis of mutant microglia across brain regions in human and mouse models of neuro-histiocytosis.

(A) Variant allelic frequency (VAF, %, by HemePACT) of BRAFc.1799T>A (V600E) in PU.1+ nuclei from histiocytosis patients across brain regions (n=8, patients with neuro-histiocytosis are color-coded in red, patients without a diagnosis of neuro-histiocytosis are color-coded in blue). Statistics: the fitted line, R-squared and corresponding p value were calculated by simple linear regression by assigning numbers from 1–8 to each brain region from along a rostro caudal axis. Gray line: all patients. Red line: patients with neuro-histiocytosis. Blue line: patients without neuro-histiocytosis. (B) Representative mouse sagittal midline brain sections from 6 months old Csf1r^MerCreMer; Braf^LSL-V600E mice (pulsed with OH-TAM at E8.5), Cx3cr1^CreERt2 ; Braf^LSL-V600E (pulsed with OH-TAM at E9.5) and littermate controls stained with anti-IBA1 or anti-GFAP, Scale bar 1000uM. (C) Allelic frequency of the Braf^V600E allele in microglia purified from dissected brain regions from 2 months old, and at 6–12 months-old analyzed by droplet digital PCR (ddPCR). Dots and colored lines represent individual mice, boxes represent variance, with line at mean. (D) RNAseq analysis performed in FACS-isolated microglia from cortex and brainstem from 2-month-old Cx3cr1^CreERT2 Braf^LSL-V600E mice (n=3), and littermate controls (n=3) pulsed with OH-TAM at E9.5. Top, principal component analysis (PCA). Bottom, pathway analysis of significantly upregulated genes (FDR <0.05, log2FC >= 1.5) in microglia from old Cx3cr1^CreERT2 Braf^LSL-V600E versus littermate control using g:profiler webtool. Pathways are selected based on FDR <= 0.05 and ordered by significance. (E) Hierarchical clustering of DEG from ‘mitotic cell cycle process’ (GO:1903047, left) from analysis in D.

This anatomical distribution of microglial clones and brain damage could correspond in theory to preferential engraftment of circulating clones and/or to a local survival or proliferative advantage of mutant clones in the hindbrain. Mouse models of neuro-histiocytosis, generated by mosaic targeting of a Braf^V600E allele ²⁹ in embryonic resident macrophages ^24,30, result in focal proliferation and activation of macrophages across several organs, including microglia, followed by paralysis and neurodegeneration after ~6 month of life (²⁶, Figure 3B and Figure S3B-D). We therefore measured Braf^V600E allelic frequency in microglia from experimental mice and control littermates, along the brain rostro-caudal axis of over time. Interestingly, Braf^V600E microglia was randomly distributed throughout the brain in young mice, at various allelic frequencies but later selectively accumulated in the rhombencephalon (Figure 3C) phenocopying the results from patients and suggesting a local survival or proliferative advantage of the mutant clones.

Pathway analysis of differentially expressed genes in RNAseq from FACS-sorted microglia from the cortex and brainstem of mice and control littermates showed a vigorous and brainstem-specific microglia proliferative response associated with an inflammatory signature and the absence of cellular senescence (Figure 3D, E and Figure S4). As previously shown³¹ the proliferative activity of wild-type microglia is higher in brainstem than the cortex however, this effect was exacerbated in Braf^V600E microglia (Figure 3E). It is of note that enforced expression of BrafF^V600E at high allelic frequency in both cortex and rhombencephalon resulted instead in microglia and astrocyte activation in both cortex and rhombencephalon (Figure S5), suggesting that cortex microglia may not per se be refractory to activation by Braf^V600E. Altogether, analysis of mouse models strongly suggests that the preferential accumulation of BRAF^V600E microglial clones in the mammalian rhombencephalon is driven, at least in part, by a local proliferative advantage of microglia amplified by the BRAF^V600E mutation.

Cellular origin of the patients’ PU.1⁺ BRAF^V600E clones

Single nuclei (sn)-RNAseq indicated that ~93% of PU.1⁺ nuclei are annotated as microglia (³² see Methods), but this does not allow to distinguish a resident or bone marrow origin of these microglia-like cells³². We therefore used a genetic bar-coding approach to investigate the putative origins of the patients PU.1⁺ BRAF^V600E clones. Analysis of all single nucleotide variants (SNV) identified by deep sequencing showed that the diagnosis of histiocytosis was the only factor significantly associated with SNV burden in PU.1 nuclei in a mixed-effects linear regression model (P < 0.001) (Figure 4A). In addition, the clonal diversity of brain PU.1+ nuclei did not reflect the clonal diversity of blood cells, i.e. brain PU.1⁺ SNV were rarely detectable in corresponding blood or bone marrow samples, and vice-versa blood/BM SNVs were rarely detected in the brain (Figure 1B and Figure 4B), consistent with the local maintenance and diversification of brain resident microglia^{24,26,33–35}.

Figure 4. Natural history of BRAF^V600E clone.

(A) Mutational load in NeuN, DN (double negative, NeuN-, PU.1-), PU.1 and Blood/Bone Marrow from control and Histiocytosis patients. Statistics: p-value was calculated using a mixed-effects linear regression model (see methods). (B) Venn diagrams represent the repartition per cell type of single-nucleotide variations (SNVs) identified in NeuN⁺, PU.1⁺, DN and matching blood in samples from 8 histiocytosis patients (NeuN, n=62; DN, n=69; PU.1, n=71; Blood/BM, n=12) and 35 control individuals (NeuN, n=107; DN, n=108; PU.1, n=107; Blood/BM, n=22). (C) Mean Variant allelic frequency (VAF, %, HemePACT) for BRAF c.1799T>A (V600E) (red), TET2 and DNMT3A variants (blue) in brain PU.1+nuclei and blood/bone marrow nd: not detected (*mean depth 5600x, see Figure S6), ^$: ddPCR only. For patient #1 and #2 myeloid (HLA-DR+, Lin-) and lymphoid (Lin+) cells were flow-sorted. (D) Left, mutual exclusivity analysis of mutations found by single-cell genotyping (Tapestri) of PU.1+ cells isolated from Pons and Cerebellum from patient #6. The number represents the probability that two mutations are mutually exclusive in single cells by random chance. The smaller the probability, the more likely they are in different cell populations. Right, plot depicts probable origin of the BRAFV600E and CH clones in patients #6. Numbers show the range of allelic frequency in different brain samples for each mutation. (E) Top, mutual exclusivity analysis of mutations found by single-cell genotyping (Tapestri) of PU.1+ cells isolated from Pons from patient #8. The number represents the probability that two mutations are mutually exclusive in single cells by random chance. The smaller the probability, the more likely they are in different cell populations. Bottom, plot depicts probable origin of the BRAFV600E and CH clones in patients #8. Numbers show the range of allelic frequency in different brain samples for each mutation. (F) Left, mutual exclusivity analysis of mutations found by single-cell genotyping (Tapestri) of PU.1+ cells isolated from cerebellum from patient #3. The number represents the probability that two mutations are mutually exclusive in single cells by random chance. The smaller the probability, the more likely they are in different cell populations. Right, plot depicts probable origin of the BRAFV600E and CH clones in patients #3. Numbers show the range of allelic frequency in different brain samples for each mutation. (G) Plot depicts detection of BRAFV600E clones in patients #1, #2 and #5 based on targeted-sequencing data and ddPCR. Numbers show the range of allelic frequency in different brain samples for each mutation. (H) Plot depicts detection of the BRAFV600E clones and CH clones in patients #4 and #7 based on targeted-sequencing data and ddPCR. Numbers show the range of allelic frequency in different brain samples for each mutation. nd: not detected Ω Absent from cerebellum; ^& Present in frontal cortex; ^@ Absent from midbrain

Nevertheless, the BRAF^V600E variant was detected in the in the bone marrow and blood from two ECD patient (#8 and #3) and clonal hematopoiesis (CH) ^36,37 carrying TET2 or DNMT3A variants was identified in the blood/BM of 5 patients, also detected in the brain of 3 of them (#8, #3 and #6) (Figure 1B and Figure 4C), and analysis, of the SNV burden for the most frequent CHIP mutations (TET2, DNMT3A and ASXL1) indicated a higher mutation burden in the patients’ blood and PU.1+ nuclei (Figure S6A). These results are consistent with our previous report of a high frequency of CHIP in the blood of ECD patients ¹⁴, although there was no significant association found between the presence of CHIP and neurodegeneration in this epidemiological study. We therefore investigated the lineage relationships between BRAF and CH clones in the brain and blood/bone marrow in our series of patients.

Single nuclei genotyping (Tapestri) of brain PU.1+ nuclei from patient #6, which presented with clonal hematopoiesis (TET2 and DNMTA3 variants at AF~50%) with a JAK2 variant subclone at 40% AF, as well as a brain specific BRAF^V600E clones (Figure 1B, Figure 4C, Table S3), showed that the BRAF^V600E nuclei present in the patient’s brain (at AF 3.5%) also carried the TET2 and DNMT3A variants (Figure 4D). It is of note that the PU.1⁺ TET2/ DNMT3A/JAK2 variant was absent from the cerebellum by bulk targeted sequencing (Table S3). Therefore, it is possible to conclude that a bone-marrow derived brain PU.1⁺ TET2/DNMT3A/BRAF subclone was associated to patient #6 brain lesions (Figure 4D and Table S3).

Interestingly, single nuclei genotyping (Tapestri) of brain PU.1+ nuclei from patients #8 and #3, who also presented with TET2 and BRAF clones detectable in the brain and blood or bone marrow (Figure 1B, Figure 4C and Table S3), showed that the TET2 and BRAF variants were mutually exclusive at the single nuclei level (Figure 4E), indicating that the patients carried independent hematopoietic clones, both able to colonize the brain. Of note, the TET2 clone was also detected in the unaffected frontal cortex of patient 8 (Table S3), indicating that the BRAF clone was a better match with the brain lesions than the TET2 clone.

In contrast to the above, the 2 pediatric-onset LCH patients (patients #1 and #2), did not present with detectable CH, and the BRAF variant was not detected in their white blood cells, even after separation into myeloid and lymphoid cells by flow-sorting (Figure S6A) and analysis by ddPCR at a depth >8000x (Figure 4C F, figure S6B, Table S3). Patient #1 and #2 were diagnosed as infants presenting with skin lesions, which were sampled at ~6 month old, i.e ~2 decades before brain PU.1 cells were analysed. We found that the skin sample form of patient #1, carried the BRAF^V600E as well as a silent intronic mutation (MSH6_c.260+47C>T), which was also identified in brain PU.1 cells, but was absent from patient blood lymphoid or myeloid cells by ddPCR (Figure S6C). We also found that the skin sample from patient #2 carried the BRAF^V600E, but not the additional pathogenic ARAF and KRAS mutations present in brain PU.1 cells (Figure 1 and Figure S6C), while all 3 mutations were absent from patient blood lymphoid or myeloid cells. Thus, although it is not possible to formally exclude that bone marrow clones spontaneously disappeared from the blood of patients #1 and #2, despite these 2 patients receiving minimal chemotherapy, late in the course of the disease, and never presented with sign of a myeloproliferative disease (see Extended Data Patients), it appears more likely that the BRAF mutation may have occurred in the resident macrophage lineage, giving rise to a multifocal disease in the absence of bone marrow clone as observed in the corresponding mouse model²⁶.

Similarly, neither CH clones nor the BRAF variant were detected in the bone marrow of patient #5, at a sequencing depth of ~5000x (Figure 4C, G, Figure S6B and Table S3). This patient presented with multifocal histiocytosis in the absence of sign of a bone marrow myeloproliferative disease and did not receive chemotherapy, suggesting again that either that a BRAF^V600E bone marrow clone had spontaneously disappeared, or that the BRAF mutation occurred in the resident macrophage lineage.

Patient #7 presented with multiple cancers, and clonal hematopoiesis (TET2 mutation at AF 35%), however, the TET2 variant was undetectable in the brain, while the BRAF variant was undetectable in the bone marrow at a depth of ~5000x (Figure 4C and Table S3). A similar pattern was observed in patient #4 (Figure 4C and Table S3). These data suggested that the BRAF mutation may have occurred either in the resident macrophage lineage (Figure 4G), or that a small BRAF^V600E bone marrow clone (as observed in patient #3 for example) had spontaneously disappeared in these patients.

These results therefore identified 3 possible different natural histories of the microglial BRAF^V600E clones in histiocytosis patients. In one ECD patient, the brain BRAF clone was a subclone of a known myeloproliferative disease. In 2 other ECD patients the BRAF and TET2 mutations were identified as independent clones. Finally, in the 2 pediatric-onset LCH patients and 3 late onset LCH/ECD patients, the PU.1+ brain BRAF^V600E clones may originate from the resident macrophage lineage, although we cannot eliminate the possibility that they originated from transient bone marrow BRAF^V600E clones. It is also of note that in this small series, clinical symptoms and the topography of the disease appear independent of the putative origin of the BRAF mutant clones.

Conserved BRAF^V600E -driven inflammatory microglia in patients and mice

PCA and pathway analysis of differentially expressed genes in RNAseq from FACS-sorted microglia from the cortex and brainstem of experimental and control mice confirmed that, in addition to the proliferative signature, microglia from the brainstem of mutant mice also presented with strong inflammatory signatures (Figure 3D, Figure S4 and Table S5), reminiscent of the transcriptional profile of the patients’ hindbrain (see Figure 2D). Accordingly, analysis of common DEG between patients and controls hindbrain and between microglia from the brainstem of experimental and control mice showed a core phagocytic and inflammatory BRAF-associated microglia (BAM) signature characterized by IL1b, NADPH oxidase (Cybb), complement, and phagocytic receptors (Figure 5A, Figure S7A and Table S6). A single nucleus (sn)-RNAseq analysis of dissected cortex and brainstem from experimental and control mice (Figure 5B and Figure S7B-E), confirmed the presence of activated microglia in the brainstem of mutant mice, expressing cathepsins, lysozyme, and APOE (Figure S7F-I and Table S7). Overall, these results strongly support the hypothesis that neuro-histiocytosis is a clonal neuro-inflammatory microglial disease associated with the BRAF^V600E mutation.

Figure 5. Inflammatory signatures and neuron loss in mouse models of histiocytosis.

(A) Top, pathway analysis using g:profiler webtool of common differential expressed genes between 2 month old Cx3cr1^CreERt2 Braf^LSL-V600E mice microglia in Figure 3 and human whole brain samples in Figure 2. Bottom, hierarchical clustering of common genes in end stage mouse microglia. Expression values are Z score transformed. Samples were clustered using average linkage and cluster similarity was determined using the Euclidean distance. (B) Single-nuclei RNAseq (snRNAseq) analysis of dissected cortex and brainstem from Braf^VE/WT Cx3cr1^CreER mice pulsed with OH-TAM at E9.5 and analyzed at 6 month of age (end stage, VE) (n=2) and littermate controls (Ctrl, n=2). After QC and data processing nuclei were clustered and annotated using Uniform Manifold Approximation and Projection for Dimension Reduction (UMAP) by cell-type. (C) Bar plot showing the relative frequency of neurons, microglia and stromal cells by brain region and condition. Statistics: p-values were calculated with two-sided proportion test for each cell type. (D) UMAP of all nuclei color-coded by brain region (left) or by condition (right). (E) Analysis of Neurons. Top, UMAP of neuronal nuclei. color-coded by brain region (left) or by condition (right). Bottom, schematic of brainstem depicting the localization of neuronal clusters reduced in VE samples and NeuN staining (iDISCO) of the pons tegmental nucleus (TRN) from mutants and littermate control. Plot shows the quantification of NeuN staining by immunofluorescence in BRAFV600E and control mice. Each dot represents the mean of three fields per mouse. Statistics: p-values were calculated with Student t test. (F) Analysis of Astrocytes. Left, Dot plot showing the expression level (color scale) and the percent of cells expressing (dot size) the most significantly upregulated DEG (log2FC >= 0.5 & FDR <= 0.05) between brainstem astrocytes, A3 (VE) vs A1,2 (control). Barplot represents number of cells per cluster. Right, pathway analysis of DEG in astrocytes using enrichR of upregulated (red) and downregulated (blue) genes in cluster A3 (FDR <0.05, log2FC >= or <= 0.5/−0.5) in comparison to clusters 1,2. Plot shows the quantification of the % of pSTAT3 expressing cells by immunofluorescence, among IBA1+, GFAP+, and IBA1+ GFAP+ cells in the brainstem of Cx3cr1^CreERT2; Braf ^LSL-V600E and control mice. (G) Violin plots showing expression scores for previously defined disease-associated astrocytes signatures, across astrocytes clusters. Statistics: p-values between A3 and A0, A1 and A2 were calculated with one-way ANOVA, using Dunnett’s multiple comparisons test. (H) Analysis of Oligodendrocytes. Left, Dot plot showing the expression level (color scale) and the percent of cells expressing (dot size) the most significantly upregulated genes (log2FC >= 0.5 & FDR <= 0.05) between brainstem oligodendrocytes O2,4,5 (VE) vs O 0,1 (control). Right, pathway analysis of DEG in oligodendrocytes using enrichR of upregulated (red) and downregulated (blue) genes in cluster O2,4,5 (FDR <0.05, log2FC >= or <= 0.5/−0.5) in comparison to clusters 0,1.

BRAF^V600E microglia causes massive loss of grey nuclei glutamatergic and GABAergic neurons in the brainstem.

(Sn)-RNAseq confirmed that the mutant microglial compartment was expanded at the expense of the neuronal compartment, selectively in the mouse brainstem (Figure 5C), reminiscent of changes observed in patients. In addition, nuclei corresponding to all brain cell types from mutant and control mice formed common clusters in the cortex, but distinct clusters in the brainstem, suggesting that all cell types are affected in the brainstem of mutant mice (Figure 5D). Analysis of neuronal clusters suggested preferential reduction in numbers of glutamatergic excitatory neurons (HB GLU6,7,8,9, clusters N12 and N0), and GABAergic inhibitory neurons (HBINH5,7,8, cluster N13) from the pons and medulla oblongata grey nuclei, in mutant compared with WT mice (Figure 5E and Figure S8). Immunofluorescence analysis of NeuN+ neurons in the pons tegmental reticular nucleus (TRN) confirmed the loss of ~60% neurons in the TRN of mutant mice (Figure 5E). These data confirmed the observation in patients (see Figure 2) and allow to conclude that BRAF^V600E microglia causes neuronal death in the hindbrain, and specifically the loss of activating and inhibitory neurons within the pons grey nuclei.

Neurotoxic astrocyte response and reduced oligodendrocyte metabolism.

(Sn)-RNAseq analysis also characterized activation of astrocytes from in the brainstem of mutant mice (Cluster Astro_3, Figure S9 and Table S8). This cluster presented with high GFAP expression (Figure 5F), as observed in patients, associated with complement activation, oxidative stress, and JAK-STAT signaling (Figure 5F-G and Table S8). Activation of JAK-STAT signaling is reported to promote neuroinflammation in neurodegenerative diseases ³⁸, and was confirmed by immunofluorescence (Figure 5F). Of note, pathways associated with glutamatergic and GABAergic synaptic processes were downregulated in nuclei from the Cluster Astro_3 (Figure 5F), likely in in relation to the decrease in corresponding neurons.

Finally, analysis of oligodendrocytes indicated a global downregulation of metabolism in the brainstem from mutant mice (Oligo clusters O2, O4, O5, Figure 5H and Table S9), together with high expression of C4b and α1-antichymotrypsin/Serpina3n (Figure 5H and Figure S9), resembling a reactive oligodendrocyte signature previously described as disease associated 1 (DA1) oligodendrocytes in PS2APP and TauPS2APP mice³⁹ (Figure S9). As observed in astrocytes, pathways associated with glutamatergic and GABAergic synaptic processes were downregulated (Figure 5H).

Altogether, these results strongly suggest that mutant microglia promote an astrocytic neurotoxic response which may contribute to neuronal death and predominates in the brainstem grey nuclei and cerebellum in patients and mice. Of note the microglia-driven neurotoxic astrocyte response, described in several human neurodegenerative diseases, and characterized by activation of the JAK-STAT pathway may represent a novel therapeutic target.

CSF1R inhibition during preclinical disease depletes mutant microglia, limits neuronal death and improves symptoms and survival.

Progressive accumulation of microglial BRAF^V600E clones cause neuronal damage and clinical symptoms. Previous studies have shown that BRAF or MEK inhibitors are an effective treatment of LCH and ECD and may delay the onset of neuro-histiocytosis in a mouse model ²⁶. However, their effect is only suspensive and long-term treatment carry significant risks of toxicity. We reasoned that if BRAF^V600E mutant microglia remained sensitive to depletion by a CSF1R inhibitor, which depletes wild-type microglia in vivo ⁴⁰, this treatment may represent an alternative or complement the use of MAP Kinase inhibitors ⁴¹ and may prevent neuronal loss and paralysis. We found that treatment with the CSF1R inhibitor PLX5622 ⁴⁰, initiated at 3 months, right before the onset of clinical symptoms, treatments delayed the onset of neurological symptoms by several months (Figure 6A), increased survival time (Figure 6B), and limited neuronal loss in the TRN (Figure 6C) at least as efficiently as the BRAF Inhibitor PLX4720 ^26,42. We did not observe a clear synergy or additive effects of a treatment with both inhibitors in comparison to each inhibitor separately (Figure 6A-C). The CSF1R inhibitor PLX5622 decreased microgliosis more efficiently than PLX4720 in 6 months old mice as expected (Figure 6D), and decreased the microglial and astrocyte activation markers SOC3, IL1b, and C1q slightly better than PLX4720 (Figure 6E). These results suggest that CSF1R inhibitors may represent an alternative to MAP-Kinase inhibitors to limit or prevent neurodegeneration associated with BRAF^V600E microglial clones during the preclinical phase of incipient neurodegeneration. Further studies would be needed to investigate possible additive effects of the combination treatment in terms of survival and disease score.

Figure 6. Early microglia depletion with CSF1R inhibitors limits neuronal loss and improves symptoms and survival.

(A) Disease score progression and (B) survival curves for mice in A. Ticks indicate animal death/experimental endpoint. Statistics: Mantel-Cox test. p values are for comparison with control diet. Hazard ratio (logrank): for (a) [CSF1R inh/Ctrl Diet (HR 0.38; 95% CI, 0.19–0.76)], [Braf inh/Ctrl Diet (HR 0.30; 95% CI, 0.15–0.63)], [Both inh/Ctrl Diet (HR 0.20; 95% CI, 0.09 to 0.45)], for (b) [CSF1R inh/Ctrl Diet (HR 0.28; 95% CI, 0.06 to 1.24)], [Braf inh/Ctrl Diet (HR 0.18; 95% CI, 0.03 to 0.92)], [Both inh/Ctrl Diet (HR 0.16; 95% CI, 0.03 to 0.81. (C) Quantification of NeuN staining by immunofluorescence in the pons tegmental reticular nuclei (TRN) from control and Csf1r^MerCreMer; Braf^LSL-V600E mice treated with control diet, BRAF inhibitor (PLX4720) CSF1R inhibitor (PLX5622) or the combination. Each dot represents the mean of three fields per mouse. Statistics: p-values were calculated with ANOVA. (D) Percentage of IBA1+ area in brain from Csf1r^MerCreMer; Braf^LSL-V600E mice and littermate controls pulsed with OH-TAM at E8.5 and treated from 3 months of age with food formulated with CSF1R inhibitor (PLX-5622) n=6, Braf-V600E inhibitor (PLX-4720) n=6, both n=6, or control diet n=5. Statistics: p-values were calculated with one-way ANOVA, using Dunnett’s multiple comparisons test. (E) qPCR analysis in brain tissue from mice treated with BRAF inhibitor (PLX4720) CSF1R inhibitor (PLX5622) or the combination. Statistics: p-values were calculated with one-way ANOVA, using Dunnett’s multiple comparisons test

Discussion

In this study, we aimed to describe the mechanisms of Histiocytosis-associated neurodegenerative disease (neuro-histio), a late complication of two related clonal myeloproliferative diseases of macrophages, LCH and ECD. Previous studies based on biopsies have shown that the BRAF mutation was detectable in the patient brains ^22,26. Our results show that the cells carrying the BRAF mutation are microglia (or microglia-like cells), which predominantly accumulate in the brainstem, cerebellum, and hippocampus due to a local proliferative advantage of mutant microglia, where they are associated with the loss of grey nuclei and Purkinje neurons. These data identify mutant microglia /microglia-like cells as the likely drivers of the neurodegenerative disease, account for the clinical symptoms of neuro-histiocytosis, and suggest that the lack of reported efficacy of chemotherapy or MAPK inhibitors on established neurodegenerative disease is due to irreversible neuronal loss. Our studies also identify potential molecular targets, as transcriptomics studies identify a BRAF^V600E microglia inflammatory and neurotoxic profile ^43–45, dominated by IL1b, CYBB, complement and phagocytosis, without a clear senescence signature^46–50, and a neurotoxic astrocyte response reminiscent of other human neurodegenerative diseases, characterized by activation of the JAK-STAT pathway, TNF signaling, oxidative stress and complement production.

Importantly, our results also showed that in 4 patients who died without clinical symptoms of neurodegeneration within 2 to 7 years of the initial diagnosis of histiocytosis (ECD), BRAF mutant microglia clones were also pervasive in the patients’ brains, with the same preferential hindbrain distribution as in neuro-histio patients who died within 8 to 25 years of the initial diagnosis of histiocytosis and were associated with neuroinflammation. However, the mutant microglia burden was lower in patients without clinical symptoms, and the presence of clinical symptoms was correlated with a larger size of mutant microglia clones. These data show that histiocytosis patients can present with incipient neurodegeneration, possibly years before the onset of clinical symptoms. This suggest the hypothesis that microglia clones in these patients might be targeted by early therapeutic interventions to prevent or limit the development of a potentially lethal neurodegenerative disease. In this regard, we provide here a proof of principle that BRAF^V600E microglia remain CSF1-dependent, and that microglia depletion with a CSF1R inhibitor⁵¹ started during the incipient phase delay neurodegeneration in a mouse model. Future studies should evaluate the role of brain imaging and biomarkers of neuronal death to select at risk patients.

We acknowledge that the relatively small of size of this series of patients is a limitation of our studies given that LCH and ECD are rare orphan diseases. We studied here 6 patients with adult-onset ECD, two of them presenting with neuro-histio and 4 with incipient neurodegeneration, but only 2 patients with pediatric-onset LCH, both diagnosed with neuro-histio. We were not able to obtain brains from LCH patients without neuro-histio, because the mortality rate in young LCH patient is (fortunately) low, families rarely authorize autopsy, and follow up of the patients into adulthood is limited. Therefore, we do not know if the incipient neurodegeneration we identified in ECD patients also applies to the pediatric LCH population.

Our study also provides insight into the cellular origin of the mutant cells in the brain and the relationship between clonal hematopoiesis (CHIP) and neurodegeneration. A recent cohort study showed that the diagnosis of CHIP was more frequent in control cohorts than in patients with Alzheimer Disease (AD), suggesting that mutant activated macrophage or microglia clones may protect against neurodegeneration ⁵². In contrast to AD, ECD is associated with a higher prevalence of CHIP ¹⁴, as well as with hematologic and solid malignancies. However, there was no statistically significant association between the presence of CHIP and the diagnosis of neuro-histio ¹⁴. Consistently with these data, 5 of our 8 patients presented with CHIP, 2 of them with neuro-histio. Using deep-sequencing and single nuclei genotyping, we found that the BRAF mutant clone was a subclone of a CHIP clone in the brain of one patient, suggesting that CHIP clones can cause neurodegeneration. However, we also found that independent BRAF mutant and CHIP clones were present in 2 other patients, which does not support a synergy between CHIP and BRAF mutations at the cellular level. Finally, although we cannot eliminate the possibility of transient bone marrow clones, our results open the possibility that the BRAF mutant clone may originate from the yolk sac resident macrophage lineage^23,24,26 in the two LCH patients and several ECD patients. Altogether our data shows that in one patient the CHIP clone carrying the BRAF mutation is responsible for neurodegeneration, but this is not the case in the other patients, although we cannot exclude that CHIP may be either aggravating or protective from neurodegeneration in ECD. Our results also suggest that, as observed in mouse models, a multifocal disease with neurodegeneration can be observed when the BRAF mutation is carried by brain microglia or microglia-like cells, irrespectively of their putative lineage of origin ^21,26.

Notwithstanding the caveats listed above, we propose that our studies identify a human neurodegenerative disease characterized by neuronal death associated with clonal proliferation of inflammatory microglia (CPIM) of either bone marrow or possibly of yolk sac origin, a concept that might apply to other patients with NDD of unknown mechanisms³². The identification of patients with incipient neurodegeneration, together with the effect of microglia depletion in a mouse model also suggest that at risk patients may benefit from preventative treatments.

Resource availability.

Lead contact:

requests for further information and resources should be directed to and will be fulfilled by Frederic Geissmann geissmaf@mskcc.org. Materials availability: requests for materials should be directed to and will be fulfilled by Frederic Geissmann geissmaf@mskcc.org.

Data and code availability:

• DNA sequencing data processed for selection of somatic variants are available for all patients and samples in Supplementary table 3. Patient DNA raw sequencing datasets and additional information required to reanalyze the reported data are available from the lead contact upon request.

• Human whole brain RNAseq, GEO: GSE273766

• Mouse microglia RNAseq GEO: GSE273797

• Mouse brain single-nuclei-RNAseq: GEO GSE288761. Analyzed data are also accessible on shinyapp, https://weillcornellmed.shinyapps.io/Histiocytosis

  • All cell types https://weillcornellmed.shinyapps.io/Histiocytosis_6/
  • Neurons https://weillcornellmed.shinyapps.io/Neurons_6/
  • Astrocytes https://weillcornellmed.shinyapps.io/Astrocytes_6/
  • Oligodendrocytes https://weillcornellmed.shinyapps.io/Oligodendrocytes_6/

STAR Methods

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Human samples

The study was conducted according to the Declaration of Helsinki, and human tissues were obtained with patient-informed consent and used under approval by the Institutional Review Boards from Memorial Sloan Kettering Cancer Center (IRB protocols #X19–027). Samples from histiocytosis patients were collected under GENE HISTIO study (approved by CNIL and CPP Ile-de France) from Pitié-Salpêtrière Hospital and Hospital Trousseau and from Memorial Sloan Kettering Cancer Center (IRB #14–201). Control frozen brain samples and matching-blood samples were provided by the Netherlands Brain Bank (NBB), the Human Brain Collection Core (HBCC, NIH), the Hospital Sant Joan de Déu, Pitié-Salpêtrière Hospital and the MSKCC Rapid Autopsy Program (IRB #15–021). Samples obtained were clinically and neuropathologically classified by the collaborating institutions as Histiocytoses, and unaffected controls. Patients age, gender and disease status and cause of death is depicted in table S1.

Patients extended description:

Patient #1 was diagnosed with LCH in his first year of life for a persistent skin rash and lytic bone lesions of the skull and shoulders on X-rays. Biopsy of the scalp was positive for the BRAF^V600E mutation. He was treated with chemotherapy (oral corticosteroids and vinblastine) with good response. The patient was considered in remission until age 9 when he presented with diabetes insipidus. PET scan showed hypometabolism in the cerebellum and the thalamic region⁵³ (see Figure S2B). At this time no other location of the disease was observed. He was diagnosed 4 years later, at age 13, with cognitive difficulties, disinhibited behavior and a cerebellar syndrome, and was placed in a special need education programme. MRI at that time shows moderate damage to the pons, dentate nuclei, and the cerebellum (see Figure S2B). Insidious progression of cerebellar and cognitive symptoms and new onset motor deficits led him to be wheelchair-bound by the age of 17. Repeated examination of his blood and bone marrow did not identify the BRAF^V600E mutation or hematological involvement. At age 18, the patient was briefly treated with the BRAF inhibitor at a dose of 17 mg/kg p.o daily for a week, then at 9 mg/kg for another week before being discontinued for fever hypotonia and pain. The patient did not receive further chemotherapy and died aged 21.

Neuropathological and molecular examination of the patient brain indicated focal, non-systematized, areas of microglial activation, astrogliosis, and axonal spheroids in the white matter and neuronal loss in grey matter. The lesions predominate in cerebellum, where Purkinje cells are replaced by hyperplasic Bergmann glia, the dentate nucleus, the striatum, the hippocampus (CA4/3) and subiculum, the amygdala and the brainstem, while the frontal and temporal cortex is comparable to control. Histological lesions corresponded to the detection of BRAF^V600E in matching brain samples (see Figure 2) at 5% to 20% AF indicating that 10% up to 40% of PU.1⁺ cells are mutated in these areas. The non-coding SNV in the MSH6 gene was also detected at similar allelic frequency. A re-analysis of the initial skin biopsy in the present study also identified the mutation in the MSH6 gene at AF similar to the BRAF^V600E mutation (see Figure S6C). Retrospective analysis by deep sequencing and ddPCR of facs-purified blood cells sampled before treatment with the BRAF inhibitor did not identify BRAF or MSH6 mutations.

Patient #2 was also diagnosed with LCH in his first year of life on a biopsy of persistent scalp rash and bone lesions, in zygomatic bone and vertebra, and later developed diabetes insipidus, but no hematological involvement. She was treated as patient 1 and remained with no evidence of active disease during her childhood, until the age of 16, when she developed progressively clumsy gait that evolved to cerebellar syndrome with spasticity, and cognitive impairment. No other lesions or hematological involvement were found at this time and the patient was diagnosed with possible neurodegeneration related to LCH. She received chemotherapy followed by IVIG, without benefit. MRI at age 26 showed cerebellar lesions similar to patient #1(see Figure S2B). Following cerebellar biopsy the patient was started on treatment with a MEK inhibitor.

Neuropathological and molecular examination:

A core cerebellar biopsy for diagnostic purpose at age 26 demonstrated regions of Purkinje cell and granule cell loss, microcalcifications and gliosis of the cerebellar white matter see Figure 2). Molecular analysis indicated the presence of BRAF^V600E ARAF and KRAS mutations in PU.1+ nuclei at an allelic frequency of 2.5%, indicating 5% mutant cells. Analysis by deep sequencing and ddPCR of facs-purified blood cells sampled at the time of the cerebellar biopsy did not identify BRAF, ARAF, or KRAS mutations. Analysis by deep sequencing and ddPCR of the original skin biopsy showed the BRAF^V600E mutation, but the ARAF and KRAS mutations were not detectable (see Figure S6C).

Patient #3 was born in 1941 and diagnosed with LCH aged 52 in 1993 on a scalp eruption and received local treatment. The patient developed diabetes insipidus 7 years later at age 59, and later xanthelasmas, bone, cardiac and lung lesions positive for the BRAF^V600E mutation, and diagnosed with ECD. He was treated with oral corticosteroids in 2000 and vinblastine in 2003, and interferon alpha in 2006. He did not receive further treatment and later developed progressive static and kinetic cerebellar syndrome, documented at age 77 in 2017. The MRI at the time showed superior cerebellar peduncles abnormalities (Figure S2B). The patient then developed abnormal behaviour, and episodes of confusion and falls and died age 79 of brain hemorrhage.

Neuropathological and molecular examination:

Retrospective analysis of the patient blood sampled in 2017 identified a TET2 mutation at AF ~1.5% and a BRAF V600E mutation at AF ~0.7% (see Figure 4). Ischemic and hemorrhagic lesions made the neuropathological analysis difficult, but focal, non-systematized areas of microglial activation, astrogliosis, and axonal spheroids and neuronal loss are identified in the cerebellum (dentate nuclei) (see Figure 2). Molecular analysis indicates the presence of large PU.1⁺ BRAF^V600E clones (AF: 10 to 20%) in the cerebellum, brainstem and hippocampus, and smaller clones (AF: ~1% in the frontal and temporal cortex.

Patient #4 was diagnosed with ECD at age 59, with skin and multifocal bone disease, retroperitoneal fibrosis, and lung and cardiac involvement. The BRAF^V600E mutation was identified in a skin biopsy. No hematological involvement or BRAF mutations was found in bone marrow aspirates. He was treated initially with IFNa from 2011, subsequently replaced with the BRAF inhibitor vemurafenib in 2016. Brain MRI performed in 2015 and 2016 showed hypersignals in the dentate nucleus (see Figure S2B), but the patient neurological examinations in 2015, 2016, 2017, concluded to the lack of neurological symptoms including cerebellar signs or cognitive complain. The patient was diagnosed in 2018 with stage IV pancreatic mucinous adenocarcinoma at age 66 and died the same year.

Neuropathological and molecular analysis.

Retrospective analysis of the patient bone marrow sampled in 2017 identified a DNMT3 mutation at AF ~1.5% but no BRAF mutation. Neuropathological analysis revealed severe focal, non-systematized, areas of microglial activation, astrogliosis, axonal spheroids and neuronal loss in the pons, and milder lesions in hippocampus and cerebellum (see Figure 2), which correspond to the detection of a large PU.1⁺ BRAF^V600E clones in the pons (AF ~10%), a smaller clone in the hippocampus (AF ~1%) and no detectable clone in the cerebellum. The DNMT3 mutation was not detected in the brain.

Patient #5 was diagnosed with LCH and ECD at age 75 in 2018 with severe sclerosing cholangitis with a BRAF^V600E mutation identified on liver biopsy, a skin eruption, and bone, lung and cardiac involvement, but no hematological disease or neurological disease although MRI performed at diagnosis showed bilateral dentate nuclei T2 hypersignal (Figure S2B). The patient did not receive chemotherapy because of her hepatic insufficiency and died of hepatic insufficiency at age 78.

Neuropathological and molecular analysis.

Retrospective analysis of the bone marrow at diagnosis did not detect the BRAF^V600E mutation or CHIP (Figure 4, Figure S6).. Neuropathological analysis revealed severe histological lesions in the dentate nuclei, hippocampus (junction CA1/2), pons and medulla oblongata (see Figure 2). Molecular analysis showed corresponding large (AF~10%) PU.1⁺ BRAF^V600E clones in the hippocampus, midbrain, pons and medulla oblongata, and smaller (1%) PU.1⁺ BRAF^V600E clones in the cerebellum and temporal lobe (Figure 2).

Patient #6 was diagnosed with ECD at age 59, with skin and pulmonary involvement and hypermetabolic bone lesions. Skin biopsy showed a BRAF^V600E mutation. The patient was also diagnosed with clonal hematopoiesis (with DNMT3A, JAK2, and TET2 mutations at ~50% AF), but the BRAF mutation was not detected in the bone marrow (see also Figure 4, Figure S6). MRI showed small and multiple dentate nuclei T1 hyperintensities, but neurological examination was normal, without sign of motor, cerebellar, or cognitive involvement. The patient was treated for 4 months with the MEK inhibitor cobimetinib. At the age of 60 he developed a T-cell anaplastic lymphoma, responsible for his death age 61.

Neuropathological and molecular analysis found focal, non-systematized, areas of microglial activation, and astrogliosis in the pons and cerebellum (see Figure 2). Molecular analysis indicated the presence of 2 small PU.1⁺ BRAF^V600E clones in the cerebellum and pons. The bone marrow clone carrying DNMT3A/ JAK2/ TET2 mutations at AF ~50% was also detected in the brain, at AF 2% to 9% (average 7%). Retrospective analysis of the bone marrow at diagnosis confirmed the TET2, DNMT3 and JAK 2 mutations and the absence of BRAF mutation.

Patient #7 was diagnosed with ECD at age 85 with retroperitoneal fibrosis which biopsy was positive for the BRAF^V600E mutation, and hypermetabolic long bone lesions. The patient was treated for kidney adenocarcinoma, prostatic adenocarcinoma, and ischemic cardiopathy. Bone marrow analysis identified CHIP with a TET2+ clone in 34% of BM cells, but the BRAF mutation was absent. Brain MRI was normal (see Figure S2B).The patient developed lung metastasis of the kidney adenocarcinoma, acute left hemispheric subdural hematoma after a fall age 88 and died of denutrition and nosocomial infection at age 89.

Neuropathological and molecular analysis:

Retrospective analysis of the bone marrow at diagnosis confirmed a TET2 mutation (AF ~35%) and absence of the BRAF mutation (see Figure 4, Figure S6). Neuropathological analysis found focal, non-systematized, areas of microglial activation, and astrogliosis in the pons and cerebellum. Molecular analysis indicated the presence of a PU.1⁺ BRAF^V600E clone in the pons (see Figure 2). The TET2 mutation was not detected in the brain (see Figure 4).

Patient #8 had a previous history of bilateral metastatic breast cancer at age 48, treated bilateral mastectomy, irradiation and chemotherapy. She was diagnosed with ECD at age 58 on skin biopsy, with diabetes insipidus and skin, bone, heart and retroperitoneal lesions. Bone marrow analysis at diagnosis identified a TET2 mutation (AF 41%, a BRAF mutation (AF 5%) and a NRAS mutation (AF 1%). MRI documented brain lesions predominant in the cerebellum (see Figure S2B). The patient was treated with the BRAF inhibitor vemurafenib, but a static and kinetic cerebellar syndrome and cognitive disturbances with dysexecutive syndrome was diagnosed at the age of 66. The patient was treated again for metastatic breast cancer at age 68, and subsequently died of sepsis.

Neuropathological and molecular analysis:

Retrospective analysis of the bone marrow at diagnosis confirmed the presence of a TET2 mutation at AF 31% and BRAF at AF 10%. Neuropathological analysis found focal, non-systematized, areas of microglial activation, and astrogliosis in all brain regions examined except the frontal cortex (see Figure 2). Molecular analysis indicated the presence of PU.1⁺ BRAF^V600E clones in the same regions. The TET2 mutation was also identified in the brain, including the frontal cortex, but single nuclei genotyping indicated that the 2 clones were distinct (see Figure 4).

Mice

Mice were housed at a maximum of five animals per cage with a 12-h light/dark cycle (lights on from 0700 to 1900 h) at constant temperature (23°C) with ad libitum access to food and water. Mice were maintained at Memorial Sloan Kettering Cancer Center (MSKCC) Zuckerman Research Center animal facility under specific-pathogen-free conditions. All experiments were performed according to the guidelines set by the Institutional Animal Care and Use Committee as well as the National Institutes of Health Guide for the Care and Use of Laboratory Animals, Institutional Review Board (IACUC 15–04-006 and 13–04-003) from MSKCC. Csf1r^MeriCreMer mice ^24,30 (FVB/NJ) were kindly provided by Dr Jeffrey Pollard, Cx3cr1^CreER mice ⁵⁴ (C57BL/6J, Jackson Stock No. 021160) (kindly provided by Dr Dan Littman), Rosa26^LSL-YFP mice ⁵⁵, and Braf ^LSL-V600E mice ²⁹ (C57BL/6J) were kindly provided by C. Pritchard (Leicester, UK) were previously described, including genotyping protocols. Braf ^WTcre⁻, Braf ^V600Ecre⁻ and Braf ^WTcre⁺ littermates were considered Braf WT for representation of data. For targeting of BRAF(V600(E) mutation to microglia progenitors, we genetically targeted EMPs by pulse-labelling Csf1r^MeriCreMer;Braf ^LSL-V600E;Rosa26^LSL-YFP E8.5 embryos as previously described ^23,24 or Cx3cr1^CreER;Braf ^LSL-V600E E9.5 embryos ( with a single injection of 37.5 mg per kg (body weight) of 4-hydroxytamoxifen (4-OHT, Sigma-Aldrich) into pregnant females. To target of BRAF(V600E) mutation after birth to CX3CR1 positive cells Cx3cr1^CreER;Braf ^LSL-V600E mice were pulsed at 1 month of age with 37.5 mg per kg (body weight) of Tamoxifen (Sigma-Aldrich) intraperitoneally for 5 consecutive days. EMPs appear in the mouse yolk sac at embryonic day E8.5 and express the Csf1 receptor (Csf1r) and one day later (E9.5) express the chemokine receptor CX3CR1^23-25. They colonize the fetal liver from E9.5 and give rise to macrophage precursors that distribute in embryonic tissues and differentiate into tissue-specific macrophage subsets, such as microglia in the central nervous system ²⁴. Embryonic development was estimated considering the day of vaginal plug formation as 0.5 days post-coitum (dpc). A short treatment with 4-OHT leads to transient nuclear translocation of the estrogen receptor-Cre-recombinase fusion protein (MeriCreMer or CreER) in cells expressing the Csf1r^MeriCreMer transgene or Cx3cr1^CreER Knock-in allele and deletion of a floxed stop cassette (LSL) in the Braf^LSL-V600E and Rosa26^LSL-YFP alleles. 4-OHT was supplemented with 18.75 mg per kg (body weight) progesterone (Sigma-Aldrich) to counteract the mixed oestrogen agonist effects of tamoxifen, which can result in fetal abortions. We did not observed differences associated with sex in disease development ²⁶, and female mice were used.

METHOD DETAILS

Nuclei isolation from frozen brain samples.

All samples were handled and processed under Air Clean PCR Workstation. ~250–400 mg of frozen brain tissue were homogenized with a sterile Dounce tissue grinder using a sterile non-ionic surfactant-based buffer to isolate cell nuclei (250 mM Sucrose, 25 mM KCL, 5 mM MgCl2, 10 mM Tris buffer pH 8.0, 0.1% (v/v) Triton X-100, 3 μM DAPI, Nuclease Free Water). Homogenate was filtered in a 40-μm cell strainer and centrifuged 800g 8 min 4°C. To clean-up the homogenate, we performed a iodixanol density gradient centrifugation as follow: pellet was gently mixed 1:1 with iodixanol medium at 50% (50% Iodixanol, 250 mM Sucrose, 150 mM KCL, 30 mM MgCl2, 60 mM Tris buffer pH 8.0, Nuclease Free Water) and homogenization buffer. This solution layered to a new tube containing equal volume of iodixanol medium at 29% and centrifuged 13.500g for 20 min at 4°C. Nuclei pellet was gently resuspended in 200 μl of FACS buffer (0.5% BSA, 2mM EDTA) and incubated on ice for 10 min. After centrifugation 800g 5 min 4°C, sample was incubated with anti-NeuN (neuronal marker, 1:500, Anti-NeuN-PE, clone A60 Milli-Mark^™) for 40 min. After centrifugation 800g 5 min 4°C, sample was washed with 1X Permeabilization buffer (Foxp3 / Transcription Factor Staining Buffer Set, eBioscience^™) and centrifuged 1300g 5 min, without breaks to improve nuclei recovery. Staining with anti-Pu.1 antibody in 1X Permeabilization buffer (microglia marker 1:50, PU.1-AlexaFluor 647, 9G7 Cell Signaling) was performed for 40 min. After a wash with FACS buffer sample were ready for sorting. Nuclei are FACS-sorted in a BD FACS Aria with a 100-μm nozzle and a sheath pressure 20 psi, operating at ~1000 events per second Pellet is gently resuspended in 200 μl of FACS buffer (0.5% BSA, 2mM EDTA) and incubated on ice for 10 min. After centrifugation 800g 5 min 4°C, sample was incubated with anti-NeuN (neuronal marker, 1:500, Anti-NeuN-PE, Milli-Mark^™) for 40 min. After centrifugation 800g 5 min 4°C, sample is washed with 1X Permeabilization buffer (Foxp3 / Transcription Factor Staining Buffer Set, eBioscience^™) and centrifuged 1300g 5 min, without breaks. Staining with anti-Pu.1 antibody in 1X Permeabilization buffer (microglia marker 1:50, PU.1-AlexaFluor 647, 9G7 Cell Signaling) was performed for 40 min. After a wash with FACS buffer sample were ready for sorting. Nuclei are FACS-sorted in a BD FACS Aria with a 100-μm nozzle and a sheath pressure 20 psi, operating at ~1000 events per second. For purity analysis, snRNAseq was performed on sorted PU.1+ nuclei from one control brain, with a resulting purity of >93% microglia ³². For DNA sequencing, nuclei were sorted into 1.5 ml certified RNAse, DNAse DNA, ATP and Endotoxin free tubes containing 100 μl of sterile PBS. For each population we sorted >10⁵ nuclei. Nuclei pellets were centrifuged 20 min at 6000g and processed immediately for gDNA extraction with QIAamp DNA Micro Kit (Qiagen) following manufacture instructions. DNA from whole-blood or bone marrow samples was extracted using the same protocol. Flow cytometry data was collected using DiVa 8.0.1 Software. Subsequent analysis was performed with FlowJo_10.6.2.

Library preparation and sequencing.

DNA samples were submitted to the Integrated Genomics Operation (IGO) at MSKCC for quality and quantity analysis, library preparation and sequencing. After PicoGreen quantification, ~200ng of genomic DNA were used for library construction using the KAPA Hyper Prep Kit (Kapa Biosystems KK8504) with 8 cycles of PCR. After sample barcoding, 2.5ng-1μg of each library were pooled and captured by hybridization with baits specific to either the HemePACT (Integrated Mutation Profiling of Actionable Cancer Targets related to Hematological Malignancies) assay, designed to capture all protein-coding exons and select introns of 576 (2.88Mb) commonly implicated oncogenes, tumor suppressor genes and or HemeBrainPACT (716 genes, 3.44 Mb) an expanded panel that included additional custom targets related to neurological diseases (see Supplementary Table 2).Capture pools were sequenced on the HiSeq 4000, using the HiSeq 3000/4000 SBS Kit (Illumina) for PE100 reads. Samples were sequenced to a mean depth of coverage of 1106x (Control samples: 1088Xx, Histiocytosis samples 1178x).

Mutation data analysis.

The data processing pipeline for detecting variants in Illumina HiSeq data is as follows. First the FASTQ files are processed to remove any adapter sequences at the end of the reads using cutadapt (v1.6). The files are then mapped using the BWA mapper (bwa mem v0.7.12 ). After mapping the SAM files are sorted and read group tags are added using the PICARD tools. After sorting in coordinate order the BAM’s are processed with PICARD MarkDuplicates. The marked BAM files are then processed using the GATK toolkit (v 3.2) according the best practices for tumor normal pairs. They are first realigned using ABRA (v 0.92) and then the base quality values are recalibrated with the BaseQRecalibrator. Somatic variants are then called in the processed BAMs using MuTect (v1.1.7) for SNV and ShearwaterML ^56,57.

MuTect1: to identify somatic variants and eliminate germline variants, we run the pipeline as follow: PU.1, DN and Blood samples against matching-NeuN samples, and NeuN samples against matching-PU.1. In addition, we ran all samples against a Frozen pool of 10 random genomes. We selected Single Nucleotide Variations (SNVs) mutations [Missense, Nonsense, Splice Site, Splice Regions] that were supported by at least 4 or more mutant reads and with coverage of 50x or more. Fill-out file for each project (~27 samples per sequencing pool), allowed to exclude mutations with high background noise. This resulted in 612 mutations (missense, nonsense, splice_site, splice_region).

ShearwaterML, was used to look for low allelic frequency somatic mutations as it has been shown to efficiently call mutations present in a small fraction of cells with true positives being ~90%. Briefly, the basis of this algorithm is that is uses a collection of deep-sequenced samples to learn for each site a base-specific error model, by fitting a beta-binomial distribution to each site combining the error rates across all normal samples both the mean error rate at the site and the variation across samples, and comparing the observed mutation rate in the sample of interest against this background model using a likelihood-ratio test. For detailed description of this algorithm please refer to^56,57. In our data set, for each cell type (NeuN, DN, PU.1) we used as “normal” a combination of the other cell types (from Histiocytosis as well as control samples), i.e PU.1 vs NeuN+DN, DN vs NeuN+PU.1, NEUN vs PU.1+DN, Blood vs NeuN+DN. Since all samples were processed and sequence the same way, we expect the background error to be even across samples. More than 400 samples were used as background leading to an average background coverage >400.000x. Resulting variants for each cell type were filtered out as germline if they were present in more than 20% of all reads across samples. Additionally, mutations with coverage of less than 50x and more than 35% variant allelic frequency (VAF) were removed from downstream analysis. p-values were corrected for multiple testing using Benjamini & Hochberg’s False Discovery Rate (FDR) ⁵⁸ and a q-value of cutoff of 0.01 was used to call somatic mutations. Mutations were required to have a least one supporting read in each strand. Somatic mutations within 10bp of an indel were filtered out as they typically reflect mapping errors. We selected Single Nucleotide Variations (SNVs) mutations [Intronic, Intergenic, Missense, Nonsense, Splice Site, Splice Regions] that were supported by at least 4 or more mutant reads and annotated them using VEP. Finally, we excluded mutations with a MAF (minor allelic frequency) cutoff of 0.01 using the gnomeAD database. This resulted in 424 SNVs.

We compared the final mutant calls of MuTect and ShearwaterML and found that >20% of the events (137 mutations) called by MuTect1 were also called by ShearwaterML. The aggregation of MuTect and ShearwaterML resulted in a total of 899 unique variants, with a mean coverage at the mutant site of 659x (10% percentile: 313x, 90% percentile: 1095x) and a mean of 20 mutant reads (10% percentile: 4, 90% percentile: 42), and 87% of mutated supported by at least 5 mutant reads (mean variant allelic frequency of 3%).

Validation of mutations.

We performed validation of 7.56% of mutations (82/899) by droplet-digital-PCR (ddPCR) on pre-amplified DNA or on libraries (when DNA not available) and confirmed 81/82 of mutations tested (>98%). Most assays were performed in all cell types isolated from the same brain region and matching blood/bone marrow). The mean depth of ddPCR was ~4000x. For BRAFV600E ddPCR we used BRAF_V600E Bio-Rad validated assay (Unique Assay ID: dHsaMDV2010027). Other assays specific for the detection of mutations were designed and ordered through Bio-Rad. For newly designed assays, cycling conditions were tested to ensure optimal annealing/extension temperature as well as optimal separation of positive from empty droplets. All reactions were performed on a QX200 ddPCR system (Bio-Rad catalog # 1864001). When possible, each sample was evaluated in technical duplicates or quartets. Reactions contained 10ng gDNA, primers and probes, and digital PCR Supermix for probes (no dUTP). Reactions were partitioned into a median of ~31,000 droplets per well using the QX200 droplet generator. Emulsified PCRs were run on a 96-well thermal cycler using cycling conditions identified during the optimization step (95°C 10’; 40–50 cycles of 94°C 30’ and 52–56°C 1’; 98°C 10’; 4°C hold). Plates were read and analyzed with the QuantaSoft sotware to assess the number of droplets positive for mutant DNA, wild-type DNA, both, or neither.

Quantification of mutational load and statistics.

We defined mutational load or mutational burden as the number of synonymous and non-synonymous somatic single-nucleotide-variations (SNV) per megabase of genome examined ⁵⁹. To quantify mutational load we took into consideration the panel used for sequencing each sample: HemePACT (2.88 Mb) or the extended panel HemeBrain-PACT (3.44 Mb) (see Supplementary Table 1-3). Therefore, the number of mutations was normalized by the number of Mb sequenced for that specific sample. In the cases where we calculated mutational load per patient, we averaged the mutational load of each sample from that patient for a given cell type [(i.e if for one patient, 2 PU.1 samples were sequenced, one from hippocampus and one from cortex (with BRAIN-PACT) then the mutational load for PU.1 for that patient is the mean of the mutational load of the 2 PU.1 samples analyzed). Statistical significance was analyzed with GraphPad Prism (v9) and R (3.6.3). Non-parametric tests were used when data did not follow a normal distribution (Normality test: D’Agostino-Pearson and Shapiro-Wilk test). In all the statistical tests, significance was considered at P < 0.05. For Venn Diagram plots, we used ⁶⁰.

Single-nuclei library preparation and sequencing (Tapestri).

PU.1+ nuclei from Pons and Cerebellum from patient #6, Cerebellum from patient #3 and Pons from patient #8 were isolated as detailed above and processed as in ⁶¹. Briefly, nuclei were suspended in Mission Bio cell buffer at a maximum concentration of 4000 nuclei/μl, encapsulated in Tapestri microfluidics cartridge, lysed and barcoded. Barcoded samples were then put through targeted PCR amplification with the Myeloid panel (MB03–0036, Mission Bio) which included the mutations to be tested. PCR products were removed from individual droplets, purified with Ampure XP beads and used as templates for PCR to incorporate Illumina i5/i7 indices. PCR products were purified again, quantified with an Agilent Bioanlyzer for quality control, and sequenced on an Illumina NovaSeq. The minimum total read depth was determined by same formula as used in ⁶¹. A total of 2769 nuclei were analyzed from patient #6, 4919 from patient #3 and 710 from patient #8.

Demultiplexing and alignment.

FASTQ files for single-nuclei DNA libraries were first processed through Mission Bio’s Tapestri pipeline v3 part1 with default parameters for demultiplexing and alignment that resulted in two outputs:

1. Single-barcode BAM files

2. Barcode-by-amplicon read-count matrix $X1$. This matrix is meant for single-cell copy number analysis, which was not relevant for this study.

Briefly, the Mission Bio pipeline v3 part1 proceeds in the following steps:

1. Adaptor sequences are trimmed from the reads; barcode sequences are extracted; the reads are aligned to the hg19 genome (UCSC)

2. The reads are assigned to individual cell barcodes while filtering out the low-quality cell barcodes.

3. For each barcode, the number of forward reads aligned to each amplicon is used to form matrix $X1$.

A more detailed documentation of the Mission Bio pipeline is available at: https://support.missionbio.com/hc/en-us/categories/360002512933-Tapestri-Pipeline and ⁶².

Single-nuclei library preparation and sequencing (Tapestri).

PU.1+ nuclei were isolated as detailed above and processed as in ⁶¹. Briefly, nuclei were suspended in Mission Bio cell buffer at a maximum concentration of 4000 nuclei/μl, encapsulated in Tapestri microfluidics cartridge, lysed and barcoded. Barcoded samples were then put through targeted PCR amplification with the Myeloid panel (MB03–0036, Mission Bio) which included the mutations to be tested. PCR products were removed from individual droplets, purified with Ampure XP beads and used as templates for PCR to incorporate Illumina i5/i7 indices. PCR products were purified again, quantified with an Agilent Bioanlyzer for quality control, and sequenced on an Illumina NovaSeq. The minimum total read depth was determined by same formula as used in ⁶¹.

Demultiplexing and alignment.

FASTQ files for single-nuclei DNA libraries were first processed through Mission Bio’s Tapestri pipeline v3 part1 with default parameters for demultiplexing and alignment that resulted in two outputs:

1. Single-barcode BAM files

2. Barcode-by-amplicon read-count matrix $X1$. This matrix is meant for single-cell copy number analysis, which was not relevant for this study.

Briefly, the Mission Bio pipeline v3 part1 proceeds in the following steps:

1. Adaptor sequences are trimmed from the reads; barcode sequences are extracted; the reads are aligned to the hg19 genome (UCSC).

2. The reads are assigned to individual cell barcodes while filtering out the low-quality cell barcodes.

3. For each barcode, the number of forward reads aligned to each amplicon is used to form matrix $X1$.

A more detailed documentation of the Mission Bio pipeline is available at: https://support.missionbio.com/hc/en-us/categories/360002512933-Tapestri-Pipeline and ⁶².

Single-cell variant calling and joint genotyping.

We used a custom variant calling pipeline [Zhang et al., bioRxiv 2024]. Briefly:

1. Single-cell level variant calling for discovery: Mutect2 (GATK 4.2.5.0) and custom filters are used for de novo mutation calling and filtering at single-cell level. All cells’ variant lists are merged into a consensus list.

2. Sample-level joint genotyping: bcftools (v1.11) is used for genotyping across all cells in the same sample using the consensus list from above. This results in a second matrix $X2$, which is variant-by-cell.

3. Patient-level joint genotyping: If multiple samples of the same patient are surveyed, variant lists from the samples are merged into a consensus list for genotyping across all cells of that patient.

Analysis of clonal relationship between a pair of mutations.

Given two SNVs’ distribution of reads across single cells, we developed a statistical measure of their clonal relationship – whether they are colocalizing in the same set of cells or mutually exclusive. Suppose we have a probability matrix $P\in {[0,1]}^{n\times m}$ for $n$ cells and $m$ mutations, where $p_{i,j}$ is the probability that mutation $j$ is present in cell $i$. We will describe how the probability is modeled in the next section. For any two mutations $j$ and $j^{\prime }$, let $Y$ be the random variable that denote the number of observations of mutual exclusivity across the $n$ cells, i.e. the number of observations of either (0,1) or (1,0) gametes in the mutation matrix. Assuming that the mutation states of cells are independent, $Y$ is a Poisson binomial variable with parameters:

$$p=\left\{p_{i,j}\left(1-p_{i,j^{\prime }}\right)+\left(1-p_{i,j}\right)p_{i,j^{\prime }}:\forall i\in [n]\right\}.$$

We define a null distribution that conserves the marginal expectation of the number of occurrences of each mutation across the $n$ cells.

Specifically, we define probability matrix $P^{\prime }$ such that:

$$p_{i,j}^{\prime }=\frac{{\sum }_{i=1}^n{p}_{i,j}}{n}.$$

Let $X$ be the random variable that denote the number of observations of mutual exclusivity for mutations $j$ and $j^{\prime }$ across the $n$ cells under the null distribution.

$X$ is a binomial variable with $n$ trials and probability of success given by:

$$p=\frac{{\sum }_{i=1}^n{p}_{i,j}}{n}\left(1-\frac{{\sum }_{i=1}^n{p}_{i,j^{\prime }}}{n}\right)+\left(1-\frac{{\sum }_{i=1}^n{p}_{i,j}}{n}\right)\frac{{\sum }_{i=1}^n{p}_{i,j^{\prime }}}{n}$$

The p-value for rejecting the null hypothesis is given by $Pr(X\ge Y)$. We formally define hypothesis testing problem as follows:

Given a probability matrix $P$ and two mutations $j$ and $j^{\prime }$, find the p-value $\mathrm{Pr}(X\ge Y)$, where $Y$ and $X$ is the number of occurrences of gametes (0,1) and (1,0) for mutation $j$ and $j^{\prime }$ under the alternate and the null hypothesis, respectively.

We compute the p-value $\mathrm{Pr}(Y\le X)$ by marginalizing $X$ as follows,

$$\begin{array}{l}\mathrm{Pr}(Y\le X)=\sum _{k=1}^{n}P(Y\le k)\mathrm{Pr}(X=k)\\ =\sum _{k=1}^n\left(P{(Y\le k)}_k^n\right)p^{k_{{(1-p)}^{n-k}}},\end{array}$$

where $\mathrm{Pr}(Y\le X)$ is the cumulative probability of the Poisson Binomial variable. If we want to test for occurrence of different gametes, we simply modify the indicators in the model: i.e. if we testing for colocalization, we would set the gametes to be (0,0) and (1,1).

Read count model.

We use the mutation genotype model used by ConDoR to model the probability of a mutation being present in a cell ⁶³. We assume that measurement of mutation across cells is independent. As such, let $s_{i,l}$ denote the state of mutation $j$ in copy number cluster $l$; let $q_{i,j}$ and $r_{i,j}$ denote the variant reads and total reads of mutation $j$ in cell $i$, respectively. We use the beta-binomial model for the read counts $r_{i,j}$, of each mutation in each cell. When the mutation is present, i.e. $s_{i,j}=1$, we model the read counts as:

$$q_{i,j}\mid s_{i,\sigma (i)}=1~\mathrm{Beta}-\mathrm{Binom}\left(r_{i,j},\alpha ,\text{β}\right),$$

where we set $\text{α}=\text{β}=1$ for this study. When the mutation is absent, i.e. ${\text{s}}_{\text{i},\text{j}}=0$, there may still be variant reads, i.e. $r_{i,j}>0$, because of sequencing errors. Let $f_p$ be the sequencing error rate to produce a false positive variant read and $d$ be the dispersion parameter. We model the read counts for $s_{i,j}=0$ case as follows,

$$q_{i,j}\mid s_{i,\sigma (i)}=0~\mathrm{Beta}-\mathrm{Binom}\left(r_{i,j},f_{p}d,\left(1-f_p\right)d\right).$$

We set $f_p=0.001$ and $d=15$.

Neuro-histological analysis of human samples.

Neuropathological analysis of human brains was performed by 3 of us (DS, IP, MR). Immunohistochemistry of tissue of patients and controls was carried out on 3–4-μm thick paraffin sections, fixed with 4% formaldehyde. Hematoxylin and eosin and immunohistochemical analysis with rabbit anti-IBA1 (1:500, 019–19741, Wako), GFAP (1:500, 6F2, Dako) was performed on paraffin sections, in Ventana XT platform.

RNA extraction from human brain samples, library preparation and sequencing.

RNA was extracted from frozen tissue using the Qiagen all-prep DNA/RNA mini kit (80204) according to the manufacturer’s instructions. RNAs were used for ribogreen quantification and quality control on Agilent BioAnalyzer. Subsequently, 500 ng of total RNA was used for polyA selection and Truseq library preparation according to the instructions provided by Illumina (TruSeq RNA Sample Prep Kit v.2), with 8 cycles of PCR. Samples were barcoded and run on a Hiseq 4000 in a 125 bp–125 bp paired-end run, using the TruSeq SBS Kit v.3 (Illumina). An average of 75 million paired reads was generated per sample.

Analysis of human and mice bulk RNAseq.

FastQ files of 2×100bp paired-end reads were quality checked using FastQC (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/, 2012). Samples were with high quality reads (Phred score >= 30) were aligned to the Human reference genome (GRCh378/hg38) using STAR aligner (version 2.7) ⁶⁴. Gene quantification was performed using feature counts from the Subread package in R ⁶⁵. Aligned reads were visualized using IGV (version 2.8.9). Gene expression levels were normalized and log2 transformed using the Trimmed Mean of M-values (TMM) method. Differential expression analysis was performed using the edgeR package in R ⁶⁶. Significantly differentially expressed genes were selected with controlled False Positive Rate (B&H method) at 5% (FDR <= 0.05). Upregulated genes were selected at a minimum log₂ fold change of 2 and downregulated genes at a minimum log₂ fold change of −2. Heatmaps were drawn on the normalized expression matrix using the pheatmap package in R. Euclidean distance similarity metric and complete linkage algorithm were used for hierarchical clustering. Gene ontology (GO) and pathway (WikiPathways, KEGG, Reactome) analysis using g:profiler webtool was performed on the list of upregulated genes (log₂ >= 2 and FDR <=0.05) ordered by importance ⁶⁷. GO terms and pathways were selected based on an FDR <= 0.05.

Flow cytometry of mouse microglia, cell sorting and RNA isolation for bulk-RNAseq.

Animals where lethally anesthetized by an intravenous injection (IV) of 10 μl/g body weight of ketamine, xylazine, acepromazine and followed by a transcardial perfusion with 20 ml ice cold PBS (gibco 14190–144). Brain was dissected and regions of one hemisphere were further separated into the Cortex, Striatum, Hippocampus, Midbrain/Interbrain, Brainstem (Pons/Medulla), Cerebellum and cervical spinal cord and placed on ice in PBS. Brain areas were individually Dounce homogenized in 6ml ice cold FACS buffer (5% Bovine Serum Albumin Proteins, 2 mM Ethylenediaminetetraacetic Acid (EDTA) in PBS (Gibco 14190–144), sterile filtered) 10–20 times with the loose and then the tight pestles. Homogenate was strained through a 100 μm cell strainer (Falcon 352360) into a 15 ml falcon and Douncer was washed out with 2 ml FACS buffer and strained as well into the 15 ml falcon. The single cell suspension was centrifuged at 300 g for 5 min at 4 °C. Resulting cell pellet was resuspended in ice cold PBS-buffered 40% Percoll^™ (GE Healthcare 17–0891-01 and centrifuged for 30 min at 500 g at 4 °C with full acceleration and braking. Supernatant was discarded and cell pellet was washed once with 10 ml cold FACs buffer centrifuged again for 5 min at 300 g at 4 °C. An aliquot of the cell pellet was used for RNA isolation with Qiagen all-prep DNA/RNA mini kit (80204) according to the manufacturer’s instructions. Remaining cell pellet was resuspended in 50 μl antibody staining mix and transferred to a 96 well plate to incubate for 30 min on ice. Antibody staining mix contained FACs buffer with F4/80 (APC) 1:100, CD45 (APC-Cy7) 1:100, CD115 (PE) 1:100, CD11b (PE-Cy7) 1:200, and CD16/32 blocking 1:100 diluted. Samples were spun down at 300 g for 5 min at 4 °C and then resuspended in 200 μl FACs buffer with 0.1 ug/mL DAPI and kept on ice. Stained cells from separated brain areas were individually sorted for CD11b⁺, CD45^int microglia with the BD FACSAria II in purity mode using a 100 μm nozzle into individual Eppendorfs filled with 750 μl FACS buffer.

DNA extraction from mouse microglia.

DNA was isolated from sorted microglia by resuspending cell pellet in 10 μl or 20 μl QuickExtract (20 μl for samples with more than 50 000 cells) and pulse-vortexed for 15 seconds. Samples were incubated for 6 min at 65 °C, followed by vortexing for 15 s. Reaction was halted by incubating samples for 2 min at 98 °C. All samples were diluted with 20 μl water and stored at −20 °C until submission to the Integrated Genomics Operation facility of MSKCC for digital PCR.

Digital PCR for Braf V600E allelic frequency in mice.

Recombination frequency of the Braf^LSL-V600E allele was assessed in pooled microglia samples from each brain region using a multiplex digital PCR (ddPCR) assay detecting the novel junction formed after Cre-mediated recombination (Braf-Rec) and the targeted unrecombined allele (Braf-minigene). The endogenous untargeted allele was not detected by the assay. Recombination frequency was calculated as (counts Braf-Rec)/(counts Braf-Rec + counts Braf-minigene). Because experimental mice have one germline copy of Braf^LSL-V600E, mutant microglia frequency was expressed as Allelic Frequency = Recombination frequency/2. To confirm equal efficiency for the two assays, a gBLOCK(IDT) containing the Braf-Rec and Braf-minigene amplicons separated by a KpnI restriction enzyme site was used as a 1:1 positive control. The gBLOCK was digested with KpnI and purified using the Qiagen Reaction Cleanup Kit prior to use. Genomic DNA from WT and Braf^LSL-V600E Cre⁻ mice were run as negative controls.

Mouse tissue preparation for histology.

Dissected brain hemisphere and whole vertebra was fixated for 24 h in 4% Paraformaldehyde in PBS at 4 °C. Tissues were washed for 10 min 3 times in PBS at room temperature. After the fixation, the spinal cord was dissected out of the vertebra. Tissue was stored in 70 % Ethanol until they were embedded in Paraffin. Paraffin blocks were sectioned with a 5 μm thickness. Brains were sectioned from the midline starting in a serial manner and 6 slices were collected representing a block of 600 to 700 μm thickness. The spinal cord was split in 3 – 6 parts and embedded to receive cross sections.

Immunohistochemistry analysis in mouse tissues.

The immunohistochemistry detection of IBA1 and GFAP were performed at Molecular Cytology Core Facility of Memorial Sloan Kettering Cancer Center using Discovery XT processor (Ventana Medical Systems). For IBA1: The tissue sections were blocked for 30 minutes in 10% normal goat serum.2% BSA in PBS. The primary antibody incubation (rabbit polyclonal IBA1 antibody, Wako, cat#019–19741) was used at 0.2 ug/ml. The incubation with the primary antibody was done for 5 h, followed by 60 min incubation with biotinylated goat anti-rabbit IgG (Vector labs, cat#:PK6101) in 1:200 dilution (6.5 ugr/mL). Blocker D, Streptavidin-HRP and DAB detection kit (Ventana Medical Systems) were used according to the manufacturer instructions. For GFAP: The tissue sections were blocked for 30 minutes in 10% normal goat serum.2% BSA in PBS.The primary antibody incubation (rabbit polyclonal GFAP antibody, Dako, cat#Z0334) was used at 1 ug/ml. The incubation with the primary antibody was done for 5 h, followed by 32 min incubation with biotinylated goat anti-rabbit IgG (Vector labs, cat#:PK6101),1:200 (6.5 ugr/mL). Blocker D, Streptavidin-HRP and DAB detection kit (Ventana Medical Systems) were used according to the manufacturer instructions. IBA1 and GFAP positive area was determined based on IHC staining. Sagittal brain images were segmented manually in QuPath into Cortex, Striatum, Hippocampus, Midbrain (including basal ganglia), Pons/Medulla, Spinal cord and the Cerebellum with additional separation of the white matter and the grey matter in the Cerebellum. Segmented areas and the unsegmented image were individually exported to ImageJ with a 1:4 compression. Segmented brain areas were individually analyzed in ImageJ by subtracting the background (Substract Background) with rolling ball radius of 50 pixels. Color Deconvolution 1.7 plugin was used to split the DAB channel from hematoxylin by using the H DAB vector. The DAB channel was further used and transferred into a binary image by selecting the Threshold value given by the Default or Otsu algorithm run on the unsegmented sagittal brain image for GFAP or IBA1, respectively. The DAB positive area was determined by running Analyze Particles taking all particles in the rage of 2 – Infinity Pixel into account.

Immunofluorescence in mouse tissue.

Immunofluorescence of mouse tissue was carried out on 3–4-μm thick paraffin sections, fixed with PFA with anti-pSTAT3-Tyr705 (1:100 D3A7, XP^® Rabbit mAb, Cell Signaling), anti-IBA1 (1:200, AIF-1/IBA1 Antibody Goat Polyclonal antibody, Novus Biologicals) and anti-GFAP (1:200, ab4674, Chicken polyclonal, Abcam), NEUN (1:200, MAB377, Millipore). Secondary Alexa647, Alexa555 and Alexa488 (Invitrogen) were added 1:200. mages were taken with a Zeiss Laboratory.A1, BondIII (Leica-Microsystems), BZ-9000 BIOREVO microscope (Keyence) and analyzed using the BZ-II Analyzer (Keyence) or with a LSM880 Zeiss microscope with 40×/1.4 (oil), performing a tile scan and z stack on the whole tissue at a 512 × 512 or 1,024 × 1,024 pixel resolution and manually analyzed using Imaris (Bitplane) software.

Isolation of mouse brain nuclei for sn-RNAseq.

All samples were handled and processed under Air Clean PCR Workstation. Frozen brain tissue was dissociated using a Dounce homogenizer with 1X lysis buffer (Nuclei PURE Lysis Buffer, Sigma L9286). Nuclei suspension is filtered using a 35 um Cell Strainer and centrifuged at 600g for 5 min at 4 °C. Pellet was resuspended in wash buffer [1X SCC (Invitrogen, cat no AM9770) 20 mM DTT, 1% BSA and RNAse inhibitor (Ambion, cat no AM2682)] and centrifuged at 600g for 5 min at 4 °C. Nuclei were stained with DAPI at 1 mg/1 mL (Invitrogen, cat no D1306 and DAPI+ nuclei were FACS-sorted with a BD FACS Aria with a 100-μm nozzle and a sheath pressure 20 psi. Sorted nuclei were centrifuged in 5mL tubes in a swinging bucket centrifuge at 600g for 5min. RIN was determined using BioAnalyzer.

Single-cell barcoding, library preparation and sequencing.

The single-nuclei RNA-Seq of FACS-sorted nuclei suspensions was performed on Chromium instrument (10X genomics) following the user guide manual for 3′ v3.1. In brief, FACS-sorted cells were washed once with PBS containing 1% bovine serum albumin (BSA) and resuspended in PBS containing 1% BSA to a final concentration of 700–1,300 cells per μl. The viability of cells was above 80%, as confirmed with 0.2% (w/v) Trypan Blue staining (Countess II). Cells were captured in droplets. Following reverse transcription and cell barcoding in droplets, emulsions were broken and cDNA purified using Dynabeads MyOne SILANE followed by PCR amplification per manual instruction. Samples were multiplexed together on one lane of 10X Chromium (using Hash Tag Oligonucleotides - HTO) following previously published protocol ⁶⁸. Final libraries were sequenced on Illumina NovaSeq S4 platform (R1 – 28 cycles, i7 – 8 cycles, R2 – 90 cycles). The cell-gene count matrix was constructed using the Sequence Quality Control (SEQC) package ⁶⁹. Viable cells were identified on the basis of library size and complexity, whereas cells with >20% of transcripts derived from mitochondria were excluded from further analysis.

Sn-RNA-seq data preprocessing.

All downstream analyses were performed using AnnData v.0.7.4 and the scanpy package v.1.5.1. A matrix of 22,276 genes x 15,067 cells was used for quality control and further analysis. Cells with more than 20% of transcripts derived from mitochondria were excluded. Genes detected in fewer than ten cells or genes with low expression levels were also excluded. Mitochondrial and ribosomal genes were removed using a curated list. To remove cellular doublets, we used the DoubletDetection package with P value threshold = 1 × 10−7, voter thresh = 0.08 followed by manual inspection of the co-occurrence of contradictory markers. A matrix of 18,730 genes x 12,603 cells was retained for downstream analysis. The count matrix was normalized using the scanpy.pp.normalize total function log2-transformed using an increment of 0.1.

Clustering analysis and cell type annotations.

Principal component analysis was computed on the top 2,000 highly variable genes. Clustering was performed using PhenoGraph ⁷⁰ and the knee point (eigenvalues smaller radius of curvature) was used to select the optimal number of principal components. For the PhenoGraph k parameter, values between k = 10 and k = 55 in steps of 5 were tested. The adjusted Rand index was used to determine the consistency of clustering across different values for k (10 to 50). The optimal k = 30 for all cell compartments was chosen from the window where the adjusted Rand index was consistently high. This approach identified N= 24 clusters. Clusters were displayed using a Uniform Manifold Approximation and Projection (UMAP). Clusters of interest were identified based on marker genes using a manually curated panel. For microglia, (CX3CR1, C1QA, PTPRC), oligodendrocytes, (MBP, MOBP), OPCs, (NEU4, PDGFRA), Astrocytes, (AQP4, GJA1), Neurons, (SYT1, SNAP25, GRN1), Vascular/Fibroblasts, (DCN, VTN), T cells, (CD3D, CD52). Each cell type of interest was re-clustered as described above and analyzed separately.

Annotation of neuronal clusters.

The identities of neuron clusters from brainstem snRNA-seq were annotated using Mousebrain.org (http://mousebrain.org/adolescent/genesearch.html). The top 10 unique genes from each cluster were searched as a group within the adolescent brain atlas and best match was identified based on magnitude of expression, number of genes expressed, and anatomical annotation to the hindbrain. Neuron identities were further confirmed by examining anatomical localization of top gene expression via Spatial RNAseq data of adult brains from Mousebrain.org (http://mousebrain.org/adult/spatial.html) and in situ hybridization data from Allen Brain Atlas (https://mouse.brain-map.org/search/index).

Differential gene expression.

Differential expression analysis between the PhenoGraph-generated clusters was performed using MAST ⁷¹ with default parameters. Differentially expressed genes were considered statistically significant if they had an FDR <= 0.05 and the absolute value of the log2(fold change) was greater or equal to 0.5.

Gene set enrichment.

Lists of qualified differentially expressed genes between clusters within different cell types were used for pathway analysis using the enrichr module from the gseapy package. Gene sets and pathways included, GO_Biological_Process_2018, MSigDB_Hallmark_2020, KEGG_2019_Mouse and Reactome_2016. Results with FDR<0.05 were reported as significantly enriched pathways. Gene signatures of reactive astrocytes were obtained from ^{72 73}. Scoring of gene signatures across clusters was done by applying the sc.tl.score_genes function from scapy on the clusters of interest. Briefly, the score is the average log-transformed expression of a set of genes subtracted with the average log-transformed expression of a reference set of genes.

Mouse drug treatments.

For BRAF and/or CSF1R inhibition, mice were placed on an ad libitum PLX4720 (BRAF inhibitor), PLX5622 (CSF1R inhibitor), or combined diet at three months of age. PLX4720 chow (417 mg/kg) and PLX5622 chow (1,200 mg/kg) were provided by Plexxikon Inc)^40,42. Braf WT and Braf VE female littermates were assigned randomly into the control or treated group. Scoring of mice was performed blinded weekly by assessing hindlimb reflexes and other behavioral phenotypes such as axial rolling, as described previously ²⁶. The investigators were not blinded to allocation during experiments and outcome assessment.

Quantitative PCR (qPCR) in mouse brains

RNA was isolated from cell pellets from dissected brain regions using RNA using the Qiagen all-prep DNA/RNA mini kit (80204) according to the manufacturer’s instructions. RNAs were used for ribogreen quantification and quality control on Agilent BioAnalyzer. cDNA was prepared using SuperScript^™ IV First-Strand Synthesis System (FisherScientific) following manufacturers instructions. Resulting cDNA was used for q-PCR using Taqman Gene Expression Assays using TaqMan Fast Advanced Master Mix (Applied Biosystems). Assays used were: IL1B (Mm00434228_m1); C1QC (Mm00776126_m1); SOCS3 (Mm00545913_s1) and GAPDH (Mm99999915_g1). The Comparative Ct Method (ΔΔCT Method) was used to calculate determine the relative quantity of the target genes.

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistical analysis.

Statistical analyses were performed with GraphPad Prism (v8) and R (3.6.3) and significance was determined at p value < 0.05. Tests used are detailed in corresponding figure legends. Non-parametric tests were used when data did not follow a normal distribution (Normality test: D’Agostino-Pearson and Shapiro-Wilk test).

Mixed-effects linear regression model.

Age, sex, sequencing depth, anatomical locations of the sample, and neurological status were incorporated as fixed effects, while donor as random effects. For Figure 1C, the analysis was performed separately for all cell types. The presence of histiocytosis in PU1 cells was the only factor significantly (P < 0.0167 = 0.05/3) associated with VAF. There was a positive and significant association between VAF and histiocytosis in PU1 (P=0.002 by a Wald test). The conditional R² coefficient was 0.454. For Figure 2F, neurological status and age were significant predictors of SNV burden (P = 0.007 and 0.041, respectively, by a Wald test). Sex, sequence depth, and anatomical locations were not associated with SNV burden. The conditional R² coefficient was 0.478. For Figure 4A, the analysis was performed separately for the four cell types. Bonferroni’s correction makes P value < 0.0125 (0.05/4) an appropriate threshold for statistical significance. The presence of histiocytosis in PU.1 cells was the only factor significantly associated with SNV burden. All other factors in all other cell types did not pass the P < 0.0125 threshold. Histiocytosis in PU.1 was positively associated with SNV burden (P < 0.001 by a Wald test). The conditional R² coefficient was 0.407.

Supplementary Material

2

Table S1: List of control and histiocytosis patients analyzed, related to Figure 1

3

Table S2: List of genes in the HemePACT panel, related to Figure 1

4

Table S3: List of somatic mutations identified by HemePACT in brain nuclei from patients and controls, with ddPCR validation, related to Figure 1

5

Table S4: RNA-seq analysis of human whole brain samples. Differentially Regulated Genes and Pathway Analysis, related to Figure 2

6

Table S5: Bulk RNA-seq analysis of sorted microglia from mouse cortex and brainstem at 2 months and end stage of the disease, and differentially Regulated Genes and pathway analysis, related to Figure 3.

7

Table S6: Conserved gene signature between human brain cells and mouse microglia. Common genes and corresponding pathway analysis, related to Figure 3.

8

Table S7: sn-RNAseq. Differential Expression Analysis between Microglia Cluster 4 and Cluster 0,1,3, related to Figure 5

9

Table S8: sn-RNAseq. Differential Expression Analysis and Pathway Analysis between Astrocytes Cluster 3 and Cluster 2 and 1, related to Figure 5.

10

Table S9: sn-RNAseq: Differential Expression Analysis and Pathway Analysis between Oligodendrocytes Cluster 2, 4 and 5 and Cluster 0 and 1, related to Figure 5.

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
NeuN-PE (clone A60)	Millipore	Cat# FCMAB317PE, RRID:AB_11212465
PU.1-AlexaFluor 647 (clone 9G7)	Cell Signaling	Cat# 2240, RRID:AB_2186911
IBA1	Wako	Cat# 019-19741, RRID:AB_839504
GFAP (clone 6F2)	Dako	Cat# MON3002-5, RRID:AB_2247498
F4/80 (APC)	Biolegend	Cat# 123116, RRID:AB_893481
CD45 (APC-Cy7)	Biolegend	Cat# 103116, RRID:AB_312981
CD115 (PE)	eBioscience	Cat# 12-1152-82, RRID:AB_465808
CD11b (PE-Cy7)	BD Biosciences	Cat# 552850, RRID:AB_394491
goat anti-rabbit IgG	Vector labs	Cat# PK-6101, RRID:AB_2336820
pSTAT3-Tyr705 (clone D3A7)	Cell Signaling	Cat# 9145, RRID:AB_2491009
IBA1	Novus Biologicals	Cat# NB100-1028, RRID:AB_3148646)
GFAP	Abcam	Cat# ab4674, RRID:AB_304558
NEUN	Millipore	Cat# MAB377, RRID:AB_2298772
Biological samples
Human postmortem brain tissue	Pitié-Salpêtrière Hospital and Hospital Trousseau and from Memorial Sloan Kettering Cancer Center. Netherlands Brain Bank (NBB), the Human Brain Collection Core (HBCC, NIH), the Hospital Sant Joan de Déu, Pitié-Salpêtrière Hospital and the MSKCC Rapid Autopsy Program. See Table S1	N/A
Chemicals, peptides, and recombinant proteins
4-hydroxytamoxifen	Sigma-Aldrich	Cat# H7904-25MG
Tamoxifen	Sigma-Aldrich	Cat# T5648-5G
Sucrose	Sigma-Aldrich	Cat# S0389-500G
KCL	Fisher Scientific	Cat# AM9640G
MgCl2	Fisher Scientific	Cat# AM9530G
Tris buffer	Fisher Scientific	Cat# AM9855G
Triton X-100	Sigma-Aldrich	Cat# T8787-100ML
OptiPrep	Sigma-Aldrich	Cat# D1556-250ML
Goat serum	Sigma-Aldrich	Cat# G9023-10ML
Nuclei PURE Lysis Buffer	Sigma	Cat# L9286
SCC	Invitrogen	Cat# AM9770
DTT	Fisher Scientific	Cat# R0861
RNAse inhibitor	Ambion	Cat# AM2682
DAPI	Invitrogen	Cat# D1306
SuperScript™ IV First-Strand Synthesis System	Fisher Scientific	Cat# 18-091-050
PLX5622	Plexxikon	N/A
PLX4720	Plexxikon	N/A
Percoll™	GE Healthcare	Cat# 17-0891-01
PBS	Gibco	Cat# 14190-144
Critical commercial assays
Foxp3 / Transcription Factor Staining Buffer Set	eBioscience	Cat# 00-5523-00
QIAamp DNA Micro Kit	QIAGEN	Cat# 56304
All-prep DNA/RNA mini kit	QIAGEN	Cat# 80234
QIAquick PCR & Gel Cleanup Kit	QIAGEN	Cat# 28506
DAB detection Kit	Ventana Medical Systems	Cat# 760-700
KAPA HyperPrep Kits	Kapa Biosystems	Cat# 07962347001
Quant-iT PicoGreen dsDNA Assay Kit	Invitrogen	Cat#P7589
HiSeq 3000/4000 SBS Kit	Illumina	N/A
digital PCR supermix for probes	Bio-Rad	Cat# 186-3028
Myeloid panel	Mission Bio	Cat# MB03-0036
Ampure XP beads	Beckman Coulter	Cat# A63881
Quant-it™ RiboGreen RNA Assay Kit	Thermo Fisher	Cat# R11490
TruSeq RNA Library Prep Kit v2	Illumina	N/A
TruSeq SBS Kit v3 - HS	Illumina	N/A
Dynabeads™ MyOne™ Silane	Invitrogen	Cat# 37002D
Chromium Next GEM Single Cell 3’ GEM, Library & Gel Bead Kit v3.1, 16 rxns	10X Genomics	Cat# PN-1000121
Chromium Next GEM Chip G Single Cell Kit	10X Genomics	Cat# PN-1000120
HemeBrainPACT	This paper	N/A
Deposited data
Human whole brain bulk RNAseq	This paper	GSE273766
Mouse-microglia bulk RNAseq	This paper	GSE273797
Mouse brain snRNAseq	This paper	GSE286627
Experimental models: Organisms/strains
Mouse
Csf1r MeriCreMer	Dr Jeffrey Pollard	Cat# 019098; RRID:IMSR_JAX:01 9098 ref#30
Cx3cr1 CreER	Jackson	Cat# 021160; RRID: IMSR_JAX:021160
Rosa26 LSL-YFP	Srinivas, S. et al	CAT#:006148 RRID: IMSR_JAX:006148
Braf LSL-V600E	Mercer, K et al.	N/A ref#29
C57BL/6J	Jackson Laboratory	RRID:IMSR_JAX:00 0664
FVB/NJ	Jackson Laboratory	RRID:IMSR_JAX:00 1800
Oligonucleotides
TaqMan IL1B	Applied Biosystems	Cat# Mm00434228_m1
TaqMan C1QC	Applied Biosystems	Cat# Mm00776126_m1
TaqMan SOCS3	Applied Biosystems	Cat# Mm00545913_s1
TaqMan GAPDH	Applied Biosystems	Cat# Mm99999915_g1
BRAF_V600E Bio-Rad validated assay	Bio-Rad	Cat# dHsaMDV2010027
Software and algorithms
BD FACS DiVa 8.0.1 Software	BD Biosciences	https://www.bdbiosciences.com
FlowJo_10.6.2	BD Biosciences	https://www.flowjo.com/
Variant detection pipeline		https://github.com/sosocc/Variant-PostProcess and https://github.com/soccin/Variant-PostProcess
ShearwaterML	Martincorena, I. et al refs# 56, 57	N/A
QuantaSoft Software	BioRad	Cat#186-4011
Mission Bio’s Tapestri pipeline v3 part1	Mission Bio	https://support.missionbio.com/hc/en-us/categories/360002512933-Tapestri-Pipeline
Single-cell variant calling and joint genotyping	Zhang et al., bioRxiv 2024	N/A
FastQC	N/A	(https://www.bioinformatics.babraham.ac.uk/projects/fastqc/
STAR aligner (version 2.7)	Dobin, A. et al ref# 64	http://code.google.com/p/rna-star/.
Subread package in R	Liao, Y. et al ref# 65	http://bioinf.wehi.edu.au/Rsubread/
IGV (version 2.8.9)	N/A	https://igv.org/doc/desktop/
edgeR package in R	McCarthy, D. J et al. ref# 66	http://bioconductor.org )
pheatmap package in R	N/A	https://cran.r-project.org/web/packages/pheatmap/index.html
g:profiler	N/A	https://biit.cs.ut.ee/gprofiler/gost
QuPath	N/A	https://qupath.github.io
ImageJ	N/A	https://imagej.net/ij/
Imaris 10.2	N/A	https://imaris.oxinst.com/
AnnData v.0.7.4	N/A	https://anndata.readthedocs.io/
scanpy package v.1.5.1	N/A	https://scanpy.readthedocs.io/en/stable/index.html
PhenoGraph	Levine, J. H ref# 70	N/A
Mousebrain.org	N/A	http://mousebrain.org/adolescent/genesearch.html)
Allen Brain Atlas	N/A	https://mouse.brain-map.org/search/index
MAST	Finak, G. et al ref# 71	N/A
enrichr	N/A	https://gseapy.readthedocs.io
Other

Highlights.

• Microglia clones predominate in rhombencephalon grey nuclei via local clonal expansion

• Microglia clones are inflammatory and cause astrocytosis and progressive neuronal loss

• Patients with the largest clones present with clinical neurodegeneration

• The long subclinical phase represents a therapeutic window before irreversible damage