Engrafted parenchymal brain macrophages differ from host microglia in transcriptome, epigenome and response to challenge

Microglia are yolk sac-derived macrophages residing in the parenchyma of brain and spinal cord, where they interact with neurons and other glial cells by constantly probing their surroundings with dynamic extensions. Following different conditioning paradigms and bone marrow (BM) / hematopoietic stem cell (HSC) transplantation, graft-derived cells seed the brain and persistently contribute to the parenchymal brain macrophage compartment. Here we establish that these cells acquire over time microglia characteristics, including ramified morphology, longevity, radio-resistance and clonal expansion. However, even following prolonged CNS residence, transcriptomes and epigenomes of engrafted HSC-derived macrophages remain distinct from yolk sac-derived host microglia. Furthermore, BM graft-derived cells display discrete responses to peripheral endotoxin challenge, as compared to host microglia. Also in human HSC transplant recipients, engrafted cells remain distinct from host microglia, extending our finding to clinical settings. Collectively, our data emphasize the molecular and functional heterogeneity of parenchymal brain macrophages and highlight potential clinical implications for patients treated by HSC gene therapy.


Introduction
Macrophages were shown in the mouse to arise from three distinct developmental pathways that differentially contribute to the respective tissue compartments in the embryo and adult. Like other embryonic tissue macrophages, microglia first develop from primitive macrophage progenitors that originate in the mouse around E7.25 in the yolk sac (YS), are thought to be independent of the transcription factor (TF) Myb and infiltrate the brain without monocytic intermediate [1][2][3] . YS macrophage-derived microglia persist throughout adulthood. Most other tissue macrophages are however shortly after replaced by fetal monocytes that derive from myb-dependent multipotent erythro-myeloid progenitors (EMP) that also arise in the YS, but are currently thought to be consumed before birth. Starting from E10.5, definitive hematopoiesis commences with the generation of hematopoietic stem cells (HSC) in the aorto-gonado-mesonephros (AGM) region.
HSC first locate to the fetal liver but eventually seed the bone marrow (BM) to maintain adult lymphoid and myeloid hematopoiesis. Most EMP-derived tissue macrophage compartments persevere throughout adulthood without significant input from HSC-derived cells. In barrier tissues, such as the gut and skin, but also other selected organs, such as the heart, HSC-derived cells can however progressively replace embryonic macrophages involving a blood monocyte intermediate 4 .
Differential contributions of the three developmental pathways to specific tissue macrophage compartments seem determined by the availability of limited niches at the time of precursor appearance 5 . In support of this notion, following experimentally induced niche liberation by genetic deficiencies, such as a Csf1r mutation, irradiation or macrophage ablation, tissue macrophage compartments can be seeded by progenitors other than the original ones [6][7][8][9] .
Tissue macrophages display distinct transcriptomes and epigenomes 10,11 , that are gradually acquired during their development 12,13 . Establishment of molecular macrophage identities depends on the exposure to tissue specific environmental factors 4,14 . Accordingly, characteristic tissue macrophage signatures, including gene expression and epigenetic marks, are rapidly lost upon ex vivo culture, as best established for microglia 11,15 .
Microglia have been recognized as critical players in central nervous system (CNS) development and homeostasis 16 . Specifically, microglia contribute to synaptic remodeling, neurogenesis and the routine clearance of debris and dead cells [17][18][19][20][21] . Microglia furthermore act as immune sensors and take part in the CNS immune defense 22 . Deficiencies affecting intrinsic microglia fitness can result in neuropsychiatric or neurologic disorders 23 . Therapeutic approaches to these 'microgliopathies' could include microglia replacement by wild type (WT) cells. Moreover, microglia replacement by BM-derived cells has also been proposed as treatment for metabolic disorders, such as adrenoleukodystrophy (ALD) and Hurler syndrome, as well as neuroinflammatory diseases (e.g. amyotrophic lateral sclerosis, Alzheimer's) in order to slow down disease progression or improve clinical symptoms 24 . HSC gene therapy was shown to arrest the neuro-inflammatory demyelinating process in a gene therapy approach to treat metachromatic leukodystrophy (MLD) albeit with delay 25 . Of note, replacement of YS-derived microglia by HSCderived cells is also a byproduct of therapeutic stem cell transplantations that are routinely used to treat monogenic immune disorders, such as Wiskott-Aldrich syndrome (WAS) and IL10 receptor deficiencies 26,27 . To what extend HSC-derived cells can replace the host microglia (especially after conditioning) and if these restore functions by cross-correction remains unclear.
Understanding how engrafted cells perform in the host, in particular following challenge, is therefore of considerable clinical importance, not only in HSC transplantation but also in HSC gene therapy approaches of disorders with a neurological phenotype.
Here we report a comparative analysis of YS-derived microglia and HSC graft-derived parenchymal brain macrophages. Using RNAseq and ATACseq of host and engrafted macrophages isolated from mouse BM chimeras, we show that these cells acquire microglia characteristics such as longevity, morphology and gene expression features, but still remain significantly distinct with respect to transcriptomes and epigenomes. Furthermore, host and graft cells display discrete responses to challenge by peripheral endotoxin exposure. Finally, and extending our finding to clinical settings, we confirm that in human HSC transplant patients, grafted cells also remain distinct from host microglia. Collectively, these data establish that engrafted macrophages differ from host microglia even after prolonged residence in the brain parenchyma and could have considerable clinical implications for patients treated by HSC gene therapy.

BM cells efficiently engraft recipient brains conditioned by irradiation or myeloablation and display clonal expansion
To further characterize the engraftment process, including clonal dynamics of the HSC-derived cells, we used two complementary approaches, comprising (1) transplantation of BM isolated from Cx3cr1 CreER :R26 Confetti ('Microfetti') mice 36 (Fig. 2 A) and (2) transplantation of lineage negative BM cells marked by a genetic barcode prior to transplantation 37 (Suppl. Fig. 2a).
Cx3cr1 CreER :R26 Confetti BM recipients were treated at two or ten weeks post engraftment with tamoxifen (TAM) to induce stochastic expression of one of the four fluorescent reporter proteins encoded by the Confetti construct 36 (Fig. 2 A). Thirty weeks later, distinct brain regions of the chimeras, including the olfactory bulb, cortex, hippocampus and cerebellum, were analyzed for the presence of graft-derived labeled cells and clonal clusters ( Fig. 2 B-F). Engraftment was seen in all brain regions analyzed ( Fig. 2 G), although the cerebellum displayed higher frequencies of HSCderived cells, as reported earlier 31 . Integration of engrafted cells into the endogenous microglial network was reflected by their Iba-1 expression, similar morphology and intercellular distances as compared to host microglia ( Fig. 2 C-F, Suppl. Fig. 3). Absence of Confetti + monocytes over 10 weeks after BM transplantation excluded sustained labeling of peripheral monocytes (data not shown), indicating that the Confetti-labeled macrophages observed in the 10-week TAM treatment group were derived from the initial engraftment. Single cells or clones (defined as ≥ two same-color cells with 100 µm nearest neighbor proximity) of Confetti-labeled cells were observed at similar frequency regardless of the timing of TAM treatment in the olfactory bulb, cortex and hippocampus ( Fig. 2 G-I), suggesting comparable proliferation capacities after engraftment. The higher number of cerebellar engrafted macrophages and clones in the 10-week TAM treatment group (Fig. 2 J) might be attributable to niche-dependent differences in kinetics of clonal expansion 36 . Clusters of same color Confetti-labeled grafts in close proximity provide evidence of local proliferation, since slow-renewing cortical microglia in 'Microfetti' brains do not form clones over 36 weeks 36 .
For the genetic barcoding procedure, lineage negative BM cells isolated from male donors (CD45.1) were transduced with lentiviral vectors harboring advanced genetic barcodes (BC32) and a BFP reporter gene 37,38 . Transduced HSC/HPCs were transferred into 8 week old female recipient mice (CD45.2) that were irradiated (950 cGy) or conditioned with the alkylsulfonate busulfan (BU) (125 mg) used for myeloablation in pediatric and adult patients 39 , as well as preclinical mouse models 40,41 . Six months after transplantation, the right hemisphere of the mice was used for immunohistochemistry (eBFP, Iba-1; Fig. 2K). From the left hemisphere, macrophages were sorted 42 and analyzed for presence and complexity of barcodes using bioinformatics 38 . Both conditioning protocols resulted in efficient engraftment of the periphery and the microglia compartment (Suppl. Fig 2B,C), as also reflected in the numbers of clones displaying unique barcodes (Fig. 2L, M, Suppl. Fig. 2D). Interestingly, while engrafted cells isolated from BM, blood and spleen, and also CD45 hi CNS cells representing hematopoietic nonmicroglial cell, displayed mainly shared clones, the majority of clones identified among brain macrophages of busulfan-conditioned animals did not have counterparts in other tissues and were private (Fig. 2N). Corroborating our above results and that of others 33,43 , this suggests that a major fraction of the grafted cells originates from transduced precursors that seed the host CNS early after engraftment and is maintained by local proliferation independent from ongoing hematopoiesis.
Taken together, these results establish that efficiently engrafted donor cells adopt the phenotype and distribution of resident microglia within this cellular network and expand locally with kinetics specific to brain regions and conditioning paradigms.
BM-derived parenchymal brain macrophages exhibit steady state transcriptomes distinct from host microglia BM graft-derived parenchymal brain macrophages acquire characteristics such as ramified morphology, longevity and radio-resistance and can hence be considered engrafted microglia-like cells that could potentially be employed for therapeutic purposes. Recent studies have highlighted the impact of the tissue environment on macrophage identities, including epigenomes and expression profiles 10,11 . To test whether graft-derived microglia acquire in the CNS such a global molecular imprint, we transplanted lethally irradiated 6 week old WT mice with congenic WT BM harboring CD45.1 alleles. Nine months post transplantation, chimeras were sacrificed and brain macrophages were isolated for transcriptome and epigenome analysis by RNAseq 44 and ATACseq 45 , respectively. At this time point, half of the CNS macrophages of the chimeras were of graft origin (Fig. 3A, Suppl. Fig. 4A).
Global RNAseq analysis of parenchymal host and graft brain macrophages isolated from individual BM chimeras revealed that engrafted cells and host microglia showed significant transcriptome overlap, clearly distinguishing them from monocytes that served as reference for a HSC-derived cell population 46 (Fig. 3B, C). Of the total 11,614 detected transcripts, 10,572 (91%) displayed a lower than 2 fold-difference between the engrafted cells and host microglia. On the other hand, 979 transcripts were differentially expressed between these two populations (absolute value of log2-fold change>1, p-value<0.05) (Suppl. Fig. 5A). Expression of 469 genes was restricted to host microglia, while 510 genes were uniquely expressed by the engrafted cells (Suppl. Fig. 5A, B). Engrafted macrophages, but not host microglia displayed for instance mRNA encoding CCR2, Lysozyme, CD38, CD74, Mrc1, ApoE and Ms4a7 (Fig. 3E, Suppl. Fig. 5B).
loop -basic helix the such as ) s (TF  transcription factors  genes included  Differentially expressed   RxRg and the  etinoic acid receptor  r  Klf12, the  TF  like zinc finger  -rueppel  K  Hes1, the  TF  helix TGFβ-associated signal transducer Smad3, that were preferentially transcribed in engrafted cells. estrogen receptor displayed increased expression of the acrophages Conversely, engrafted m , as compared to host Runx3 and the macrophage survival factor Nr4a1 TF domain -unt r Esr1, the cells (Fig. 3D). Of note, the host microglia in this case were irradiated and differences observed between the two populations could hence relate to radiation damage; transcriptomes of engrafted cells however also differed significantly from age-matched non-irradiated microglia (Suppl. Fig.   5C).
Supporting the notion of their microglia-like identity, engrafted cells expressed similar levels of the DNA-RNA binding protein TDP-43 encoded by the Tardbp gene recently implied as regulator of microglial phagocytosis 47 , the Two-Pore Domain Potassium Channel THIK-1, encoded by the Kcnk13 gene and shown to be critical for microglial ramification, surveillance, and IL-1b release 48 and the TF Mef2c, reported to restrain microglia responses 49 (Fig. 3F) Likewise, the graft also displayed expression of 'microglia signature' genes that have been proposed to distinguish these cells from other tissue macrophages and acute monocyte infiltrates associated with inflammation ), which is critical to br Tgf ( receptor β TGF ) and Fcrls ( like molecule -Fc receptor , including 50,51 establish microglia identity 50 (Fig. 3G). Other proposed 'microglia signature genes', such as P2ry12, Tmem119, SiglecH and HexB displayed either significantly reduced expression in the grafted cells or were exclusively expressed by host microglia, like the ones encoding the the and ) Cd34 ( CD34 phoglycoprotein protein phos the , ) a5 2 Slc ( Sodium/glucose cotransporter 1 transcriptional repressors Sall1 (Sall1) and Sall3 (Sall3) (Fig. 3H, Suppl. Fig. 5D). Of note, lack of some microglia markers had been reported before for cells retrieved from acutely engrafted brains 8,50,50,52,53 .
The expression signature of the engrafted macrophages showed a considerable overlap with the transcriptome of perivascular macrophages [54][55][56] , including present and absent transcripts, 55 (GSEA) Gene Set Enrichment Analysis respectively. , Sall1 and Slc2a5 , Msn4a7 , ApoE such as revealed that engrafted macrophages displayed an activation signature as compared to host microglia (Suppl. Fig. 6A). Finally and corroborating our data, the list of genes we report differentially expressed by engrafted and host cells also displayed a considerable overlap with results recently reported by two other groups 57,58 (Suppl. Fig. 6B).
Sall1, a member of the Spalt ('Spalt-like' (Sall)) family of evolutionarily conserved transcriptional regulators critical for organogenesis, acts as repressor by recruitment of the Nucleosome Remodeling and Deacetylase Corepressor Complex (NurD). Binding motifs of Sall1 and hence its direct genomic targets remain undefined precluding a direct assessment of the impact of the lack of the repressor on the expression signature of the grafted macrophages.
Interestingly though, comparison of the recently reported list of genes differentially expressed by WT and Sall1-deficient microglia 53 and that of host and graft brain macrophages revealed in this study, showed significant overlap (Fig. 3H). This included expression of genes otherwise restricted to macrophages residing in non-CNS tissues, such as Msr1, encoding a scavenger receptor, and grafted , 53 deficient microglia -like Sall1 Furthermore, ).  (Fig. 3I, Suppl. Fig. 6B). This suggests that a major fraction of the differential expression of host microglia and engrafted cells could be explained by the specific absence of the transcriptional repressor Sall1 from the former cells. Overall, these findings establish that engrafted macrophages that persist in the brain and acquire microglia characteristics such as morphology and radio-resistance, also show significant transcriptome overlap with host microglia, but remain a molecularly distinct population.

Engrafted CNS macrophages and host microglia exhibit distinct epigenomes
To further define engrafted cells and host microglia we performed an epigenome analysis using ATACseq that identifies open chromatin regions by virtue of their accessibility for 'tagmentation' by transposases 45 . Correlated ATACseq replicates (Suppl Fig 7A) performed on graft and host microglia isolated from BM chimeras detected 58,947 total accessible regions (corresponding to 16,156 genes). Corroborating the observed differential gene expression (Fig. 3), host microglia but not engrafted cells, displayed ATAC signals in the Sall1 and Klf2 loci as indicated in Integrative Genomics Viewer (IGV) tracks (Fig. 4A). ATAC peaks in other genomic locations, such as ApoE and Ms4a7, were restricted to genomes of engrafted cells, in line with mRNA detection in these cells, but not host microglia (Fig. 4B). As ATACseq does not discriminate between bound transcriptional activators and repressors, some differentially expressed loci did not show epigenetic differences. This included for instance the MHC II locus (H2-ab1), which displayed similar ATACseq peaks in host microglia and the graft that lack and display H2-ab1 transcripts, respectively (Fig. 4C). These loci might be transcriptionally silenced, but activated upon cell stimulation. Similar 'poised' states, that might be revealed following challenge, can be assumed for gene loci, that displayed differential epigenomes and ATAC peaks, but were transcriptionally active in neither the host nor the engrafted cells. Finally, rare genes, such as the Dbi locus showed equal expression, but differential ATAC profiles which might suggest that their transcription is driven by distinct TFs (Fig. 4C). Global quantification of differential ATAC peaks between the two brain macrophage populations revealed that 6 % of the total accessible regions (or 8.7 % of the associated genes) were distinct. Specifically, 1,506 peaks (corresponding to 941 genes) displayed a >4 fold significant (p-value<0.01) enrichment in host microglia and 2,176 peaks (corresponding to 1,465 genes) were increased in BM graft-derived cells (Fig. 4D).
To identify potential TFs responsible for the differential transcriptomes and epigenomes of engrafted and host microglia, we applied a TF Binding Analysis (TBA) machine Of the more intermediately ranked motifs, 41 motifs were more important in either host or graft cells as measured by the log likelihood ratio (LL) when comparing graft and host microglia (LL>=10e-2, blue points Fig. 4E). Ten motifs were preferentially detected in host microglia (Fig.   4E), including motifs for TFs such as IRF8, which is a critical regulator of microglia identity 59 , as well less characterized motifs such as the C2H2 Zinc Finger motif. The latter motif could potentially be recognized by Sall1, which contains tandem C2H2 zinc fingers. Motifs preferentially detected in grafted cells included CEBPa, RFX, and Jun motifs (Fig. 4E). Motifs that were preferentially detected in either graft or host cells exhibited even greater differences in comparison to other myeloid cell populations, consistent with environmental cues directing chromatin remodeling of grafted cells (Fig. 4F). Collectively, these observations substantiate the conclusion that while grafted cells adopt epigenetic characteristics similar to microglia, they remain distinct from host cells even after prolonged CNS residence.

HSC-derived macrophages display distinct responses compared to host microglia following peripheral endotoxin challenge
Given the significant transcriptome and epigenome differences between the host and BM-graft derived macrophages that persists for extended periods of time post-transplantation, we next examined whether the two populations are functionally distinct. To that end, chimeras were challenged nine months post transplantation by a peripheral injection of the bacterial endotoxin lipopolysaccharide (LPS), an established paradigm for inflammation associated with robust microglia responses to systemic cytokine secretion. Host and engrafted cells were isolated from the brains of the chimeras 12 hours post LPS challenge, global RNA and ATAC sequencing were performed, and results were compared to the samples obtained from non-challenged mice presented earlier (Fig. 3, 4). Principle component analysis (PCA) revealed a high degree of similarity within each group, but segregation of the host and graft samples (Fig. 5A). Host and grafted cells responded with altered expression of 745 shared genes. 940 genes were changed in grafted cells only, and 602 genes were changed in host microglia upon LPS challenge ( Figure 5B).
Examples for these three categories are shown in Figure 5C, D and Suppl. Fig. 8A Sall1 and Upk1b ( Figure 5D). Ingenuity analysis of transcriptomes of engrafted and host brain macrophages isolated from chimeras with and without peripheral LPS challenge revealed potential distinct upstream regulators acting on these populations, as well as a differential representation of activated functional pathways (Supp. Fig. 8C). Engrafted macrophages displayed for instance activation of pathways controlled by IL1b and Ifng and suppressed by IL-10. ATACseq analysis revealed differential epigenome alterations between engrafted and host cells in response to the LPS challenge. Specifically, correlated ATACseq replicates (Suppl. Fig   9A) performed on engrafted and host cells isolated from BM chimeras after LPS challenge detected 46,485 total accessible regions (corresponding to 15,390 genes). Global quantification of differential ATAC peaks between the two brain macrophage populations revealed a total of 552 peaks (corresponding to 391 genes) that displayed a >4 fold significant (p-value<0.01) enrichment in host microglia and 841 peaks (corresponding to 618 genes) that were increased to the same extent in BM graft-derived cells ( Figure 6A).
Overall, differences between host and graft cells were less pronounced after LPS challenge ( Figure 4D, 6A). Analysis of motifs with TBA models trained on intergenic peaks present after LPS challenge identified a core group of highly significant motifs common to graft and host cells in both no treatment and LPS conditions that may be important for maintaining microglia identify (p<10e-20, Suppl. Fig. 9B). Under LPS conditions, we observed that AP-1 family motifs (Jun-related, Fosrelated) and NF-kappaB (Rel, Nfkb1) motifs are more significant, which is consistent with the role AP-1 and NF-kappaB factors play in mediating the macrophage inflammatory response (Fig. 6B).
Interestingly, host and graft cells preferred different variants of the NF-kappaB and AP-1 motifs (motif logos, Fig. 6B). The NF-kappaB motif most highly enriched in host cells was more GC rich than that in graft cells, whereas the AP-1 motif most enriched in graft cells corresponded to an N(1) spaced motif (TGANTCA) in contrast to the N(2) motif (TGANNTCA) that was most highly enriched in host cells. Of the 27 motifs that were differentially detected in challenged host and graft cells, several motifs were also differentially detected in untreated cells, including motifs for IRF8, Zinc Finger factors, NK related factors, and Tal related factors (Fig. 4F, 6B).
Appearance of ATACseq peaks was correlated with differential gene expression between the two macrophage populations ( Figure 6C). Transcripts encoding the scavenger receptor Marco were absent from host microglia, but specifically induced in these cells but not engrafted cells following the LPS challenge ( Figure 6C). Likewise, the Marco locus (92kb) in microglia of unchallenged animals displayed 5 ATACseq peaks that were all restricted to the host cells (I-V; Figure 6C). LPS challenge resulted in loss of one peak (IV) and the induction of 3 additional peaks (VI-VIII), again restricted to host microglia ( Figure 6C). An induced peak located 53,411 bp downstream of the TSS displayed a host-specific Nfkb1 motif, whereas a second peak located 19,210 bp upstream of the Marco TSS displayed a host-specific AP-1 (Fos-related) motif ( Figure   6C). Engrafted macrophages on the other hand, displayed as compared to host microglia prominent induction of Secreted Phosphoprotein 1 (SPP1) / Osteopontin ( Figure 6D), a factor that got recent attention as part of a microglia signature that could be associated with certain CNS pathologies 60 . In accordance with the expression results, Spp1 loci (74kb) of engrafted macrophages displayed following LPS challenge 3 ATACseq peaks (IV-VI), that were significantly enhanced over microglia ( Figure 6D). An induced peak located 54,402 bp upstream of the TSS displayed a TBX5 motif, whereas a second peak located 38,600 bp upstream of the Spp1 TSS displayed an AP-1 (Jun-related) motif ( Figure 6D). Collectively these data establish that engrafted microglia respond differently from host microglia to a challenge and are hence functionally distinct.

HSC-derived macrophages in murine and human brain parenchyma differ from host microglia
Engrafted brain macrophages differ from host microglia by their gene expression (Fig. 3B). To confirm this finding for protein expression we performed a histological analysis of brains of the BM chimeras generated by TBI and BU conditioning (Suppl. Figure 2A). Engrafted cells and host microglia were identified by Iba-1 staining. Graft-derived cells were defined according to eBFP expression conveyed by the lentiviral construct ( Figure 2K). In concordance with the transcriptome data ( Figure 3G), analysis for Tmem119 and P2ry12 expression revealed absence of these markers from the graft (Fig. 7 A, B), while eBFP + cells displayed ApoE and MHC Class II (Suppl.

Fig. 10).
To finally extrapolate our finding to a human setting we analyzed post mortem brains of patients that had underwent HSC transplantation. Specifically, we took advantage of gendermismatched grafts that allowed identification of the transplant by virtue of its Y-chromosomes through chromogenic in situ hybridization (CISH). Ramified Iba1 + microglia-like cells harboring the Y chromosome could be readily identified juxtaposed to Y chromosome-negative host microglia in cortex, cerebellum and hippocampus sections of the subject brains ( Fig. 7 C). Expression of the purinergic P2Y 12 receptor has been proposed earlier to serve as marker for human microglia 52,61 .
have thus could and activation by nucleotides compromises microglial 2 1 P2Y absence of Moreover, functional implications 62 . As observed in the murine chimeras, also human P2Y 12 receptor expression was found to be restricted to host cells, but absent from Y chromosome-positive brain macrophages cells (Fig. 7 D). Collectively these data establish that in the CNS of patients that underwent HSC-transplantation graft-derived cells remained functionally distinct from host microglia and strengthen the conclusion that HSC-derived engrafted cells differ from host microglia.

Discussion
Here we established that HSC-derived brain macrophages that persistently seed the CNS of recipient organisms following irradiation or myelo-ablation remain distinct from host microglia with respect to their transcriptomes, epigenomes and response to challenge. The exact origin of the HSC-derived engrafted cells in the chimeric organisms remains to be defined. In a classic study, Ajami and colleagues established that non-parenchymal brain macrophages that can persistently seed the host brain originate from non-monocytic BM-resident myeloid progenitors characterized by the absence of CX 3 CR1 expression 28 . Likewise, studies by Biffi and co-workers suggested that a transient wave of early hematopoietic progenitors infiltrates the host CNS during transplantation and following local proliferation, establish the graft 33 . This notion is supported by the results of our 'Microfetti' and barcoding approaches that establish that engrafted macrophages undergo clonal proliferation and thereby likely progressively outcompete irradiation or busulfan-damaged host microglia. Moreover, the conclusion that engrafted cells arose from cells that do not contribute to long-term hematopoiesis in the chimeras is also in line with the prominent detection of private clones in this population, which are not shared with the other hematopoietic compartments.
Future studies could aim to identify cells that upon engraftment will give rise to closer mimics of host microglia, including for instance expression of Sall1. This could include cells linked with the unique developmental YS origin of microglia 57 , or otherwise manipulated cells such as microglia-like cells derived from induced-pluripotent-stem-cell (iPS)-derived primitive macrophages 63 . Furthermore, in the context of gene therapy, viral vectors could be used to express transgenes to engineer the engrafted cells to boost engraftment and modulate their function. Of note however, under certain pathological conditions the distinct engrafted HSC-derived macrophages we report might also be advantageous as compared to host microglia. HSC-derived cells could for instance be superior to YS-derived microglia in the handling of the debris burden associated with senescence 64 or amyloid plaques that arise during Alzheimer's disease 65 . Indeed, the latter hypothesis was proposed although the exact underlying mechanism remains unclear 66 ; elucidation of such scenarios should profit from the molecular definition of the engrafted cells and host microglia provided in this study. However, in vivo functions of microglia remain poorly understood and future dedicated experimentation will be required to compare the performance of engrafted macrophages and host microglia in different disease models during aging and specific challenges.
Recent studies revealed a signature of disease associated microglia, termed 'DAM' 55 or 'MGnD' 67 , that is induced by various brain-intrinsic changes in the absence of massive peripheral infiltrates, though not following peripheral LPS challenge 60,68 , and can occur on acute to chronic time scales 60 . Of note, engrafted brain macrophages displayed robust constitutive expression of some of theses DAM genes, such as ApoE and Axl. Likewise, as opposed to microglia 60 , grafted HSC-derived brain macrophages responded to the LPS challenge with the induction of DAM/ MGnD hallmarks, such as the CD44 ligand Spp1/Osteopontin. The latter might have implications when considering brain macrophage contributions to CNS pathologies. Results obtained from fate mapping models currently suggest that at least in the brain of unchallenged C57BL6 mice kept in specific-pathogen-free (SPF) facilities, parenchymal macrophages are exclusively comprised of YS-derived derived microglia. Further experimentation will however be required to assess how absolute this exclusion of HSC-derived macrophages is, in particular following challenges.
Moreover, it remains currently unclear to what extend HSC-derived macrophages might be able to seed the human brain, f.i. during extended aging, and could hence impact brain pathologies.

Tissue macrophages such as Kupffer cells (KC) and alveolar macrophages (AM)
have been reported to be faithfully replaced by HSC-derived cells in irradiation chimeras and other le these studies . Whi

Mice
For generation of BM chimeras, wild type C57BL/6 J mice (Harlan) were used as recipients, Cx3cr1 GFP or CX 3 CR1 Cre :R26-RFP fl/fl mice 34,70 were used as donors. Recipient mice were lethally irradiated with a single dose of 950 cGy using an XRAD 320 machine (Precision X-Ray (PXI)) and All experimenters were blinded during data acquisition and analysis.

Microglia isolation
To isolate microglia and HSC-derived parenchymal CNS macrophages, BM chimeras were perfused using ice-cold phosphate buffered saline (PBS) and brains were harvested. Brains were dissected, homogenized by pipetting and incubated for 20 min at 37°C in a 1 ml HBSS solution containing 2% BSA, 1 mg/ml Collagenase D (Sigma) and 1 mg/ml DNase1 (Sigma). The homogenate was then filtered through a 100 µm mesh and centrifuged at 2200 RPM, at 4°C, for 5 min. For the enrichment of microglia and BM-derived cells, the pellet was re-suspended with a 40% percoll solution (Sigma) and centrifuged at 2200 RPM, room temperature for 15 min. The cell pellet was next subjected to antibody labeling and flow-cytometry analysis.

RNA-seq
RNA-seq of populations was performed as described previously 10 . In brief, 10 3 -10 5 cells from each population were sorted into 50 µL of lysis/binding buffer (Life Technologies) and stored at 80 C. mRNA was captured with Dynabeads oligo(dT) (Life Technologies) according to manufacturer's guidelines. We used a derivation of MARS-seq 44

ATAC-seq
20,000-50,000 cells were used for ATAC-seq 45  Tagmentated DNA was isolated using 2X volumes of SPRI beads and eluted in 21 µl. For library amplification, two sequential PCRs (9 cycles, followed by an additional 6 cycles) were performed in order to enrich small tagmentated DNA fragments. We used the indexing primers as described by were called using HOMER 71 (using parameters compatible with IDR analysis: -L 0 -C 0 -fdr 0.9).
The Irreproducible Discovery Rate (IDR) was computed for each peak using the Homer peak score for each replicate experiment (https://github.com/nboley/idr); peaks with IDR > 0.05 were filtered away. Intergenic peaks annotated by Homer were used to train TBA models for each cell type using default parameters (https://github.com/jenhantao/tba). The significance of each motif in each cell type was assigned by comparing the predictive performance of the trained TBA model and a perturbed model that cannot recognize one motif using the chi-squared test.

Image acquisition and analysis
Brain images of Microfetti and lineage negative BM chimeras were acquired on the Keyence BZ-9000 inverted fluorescence microscope using a 20X / 0.75 NA objective lens. Images were processed and analyzed using ImageJ (NIH). Intercellular distances were determined by

Combined immunohistochemistry and in situ hybridization
After the death of the two patients, brains were transferred into 4% paraformaldehyde (PFA) within 48 h and fixed for at least 3 weeks. After fixation, representative tissues from several brain regions including the temporal cortex, hippocampus, and cerebellum were dissected and embedded in paraffin. Routine neuropathological examination revealed minimal signs of neurodegeneration in

Barcode analyses
DNA was extracted from peripheral blood, bone marrow, spleen and sorted microglia and CD45 high cells were used for barcode amplification, multiplexing, as well as the bioinformatic processing 39 .
Unique barcodes with more than 100 reads per sample were taken into further analyses. Venn diagrams were produced by an in-house R-script using the "VennDiagram" package.

Digital droplet PCR
To determine the chimerism in the sorted microglia, digital droplet PCR was performed. In a duplex reaction, a Y-chromosome-specific fragment (and a control amplicon (located in the erythropoietin receptor) were simultaneously amplified and analysed using the QX200 system (BioRad).

Statistical analysis
Mean data are shown. Mann-Whitney test, D'Agostino-Pearson test, two-tailed unpaired t-test, and Welch's t-test were performed in GraphPad Prism7. Statistical significance was taken at P < 0.05.

Fig. 2 BM cells efficiently engraft recipient brains conditioned by irradiation or myeloablation and display clonal expansion
A) Fate mapping scheme for donor 'Microfetti' BM cells in lethally irradiated WT hosts. A single dose of TAM was applied at 2 or 10 weeks after BM reconstitution. BM chimeras were analyzed at 30 weeks after BM transplantation. B) Representative sagittal brain section indicating the fields of view (rectangles) analyzed for IBA-1 (magenta) expressing graft microglia in the olfactory bulb (ob), cortex (cx), hippocampus (hc) and cerebellum (cb). DAPI (blue). Scale bar, 1 mm.            CD45 int CNS macrophages spleen CD45 hi cells blood

Supplementary Fig. 2 Clonal analyses on TBI and Busulfan treated animals.
A) Experimental set-up: Lineage-negative cells from 8-weeks old male donors were transduced with lentiviral vectors carrying eBFP as a fluorescent reporter as well as BC32-eBFP barcode system prior to transplantation into 8-weeks old female recipients conditioned either with total body irradiation (TBI) or chemotherapy (Busulfan). Peripheral blood samples were taken every 4-6 weeks post transplantation to monitor engraftment. 6 months post transplantation, the animals were sacrificed and peripheral blood, BM, spleen and the brain were taken for clonal analyses.
B) Flow cytometry on peripheral blood samples to measure the chimerism (CD45.1 on graft cells).
C) Chimerism of sorted microglia cells as detected by Y-chromosome-specific digital droplet PCR.
D) Barcode analyses on DNA extracted from peripheral blood, bone marrow, spleen and sorted CD45 int brain macrophages and CD45 high CNS cells. Numbers of barcodes for the samples are shown in the bar plots and shared barcodes of the respective sample are displayed in the Venn diagrams.

Supplementary Fig. 4 Flow cytometry analysis of chimeras used for isolation of cells for transcriptome epigenome analysis
A) Representative flow cytometry analysis of brain of [CD45.1 > CD45.2] chimera, 9 months after engraftment. Note presence of graft derived-parenchymal macrophages (CD45.1 int ) and perivascular macrophages (CD45.1 hi ), but absence of the latter from the CD45.1 + host compartment. Dot blot on the right indicates respective population in the sort gate (Fig. 3a) B) Representative flow cytometry analysis of brain of [CD45.1 > CD45.2] chimera, 9 months after engraftment and 12 hours after intra-peritoneal LPS treatment. Dot blot on the right indicates respective population in the sort gate (Fig. 3a).  Supplementary Fig. 4 Supplementary Fig. 5