Competitive repopulation of an empty microglial niche yields functionally distinct subsets of microglia-like cells

Circulating monocytes can compete for virtually any tissue macrophage niche and become long-lived replacements that are phenotypically indistinguishable from their embryonic counterparts. As the factors regulating this process are incompletely understood, we studied niche competition in the brain by depleting microglia with >95% efficiency using Cx3cr1CreER/+R26DTA/+ mice and monitored long-term repopulation. Here we show that the microglial niche is repopulated within weeks by a combination of local proliferation of CX3CR1+F4/80lowClec12a– microglia and infiltration of CX3CR1+F4/80hiClec12a+ macrophages that arise directly from Ly6Chi monocytes. This colonization is independent of blood brain barrier breakdown, paralleled by vascular activation, and regulated by type I interferon. Ly6Chi monocytes upregulate microglia gene expression and adopt microglia DNA methylation signatures, but retain a distinct gene signature from proliferating microglia, displaying altered surface marker expression, phagocytic capacity and cytokine production. Our results demonstrate that monocytes are imprinted by the CNS microenvironment but remain transcriptionally, epigenetically and functionally distinct.

A ll organs in the body house resident macrophage populations with duties that are tailored to the function of that organ. For example, brain microglia promote neuronal wiring, splenic red pulp macrophages (RPM) eliminate dying erythrocytes in the spleen and alveolar macrophages recycle surfactant in the lung. It is now well-recognized that the majority of these tissue macrophages are derived from embryonic precursors that self-maintain throughout adulthood 1-4 rather than from continuous monocyte input, as was previously suggested 5 . During development of the embryo, tissues are seeded with successive waves of hematopoietic precursors from the yolk sac and fetal liver 1,6-8 , which following birth are displaced by BMderived monocytes in some organs [9][10][11][12][13][14] . Furthermore, upon infection or injury, circulating blood monocytes can give rise to long-lived replacements of macrophages in many organs including the heart 15 , liver 16 , lung 17 , and brain 18,19 . The final macrophage composition can thus be heterogeneous across organs, and this has complicated recent attempts to assign macrophage nomenclature based on their ontogeny 20 . Whether the ontogeny of resident macrophage populations ultimately dictates functionality is not well understood and may vary depending on the circumstances generating the macrophage 21 .
The diversity of tissue macrophage populations has been proposed to be the result of exposure to specialized microenvironments 8,[22][23][24][25] . Significant reprogramming of macrophage precursors transplanted into adult tissue microenvironments has been demonstrated 24,26,27 , arguing for an imprinting capacity of the macrophage niche as was recently proposed 28 . Peripheral myeloid cells that colonize the microglial niche adopt microglialike morphology 29,30 , display ATP-sensing capacity 29 , can promote repair following cranial irradiation 19 and require TGF-β signaling for functional integration into the central nervous system (CNS) 31 . However, to what degree monocytes become imprinted by the microglial niche and what factors regulate this imprinting remain unknown.
In this study, we develop a novel strategy to efficiently deplete microglia, and observe competitive repopulation giving rise to a permanent mosaic of myeloid cells derived both from proliferating microglia and from brain-engrafting Ly6C hi monocytes. We provide evidence of niche imprinting, as monocyte-derived macrophages adopt both microglia gene expression and epigenetic profiles. However, monocyte-derived macrophages display a unique gene signature, giving rise to a distinct surface marker phenotype and functional profile. Our results demonstrate that both niche imprinting and myeloid origin define macrophage identity within the CNS, and this may have implications for understanding macrophage biology, disease, and therapy.
We analyzed additional groups of mice at day 28, and as expected from previous reports, 29,30,35 we detected repopulation of the CX3CR1 + compartment. Flow cytometric analyses revealed two distinct populations within the CX3CR1 + gate that differentially expressed the F4/80 antigen (Fig. 1a), and this was confirmed by immunohistochemistry ( Supplementary Fig. 2B). The surface phenotype of the CX3CR1 + F4/80 low population overlapped with microglia from control mice, whereas the CX3CR1 + F4/80 hi population expressed higher levels of CD45 and lower levels of Siglec H (Fig. 1c). Siglec H has previously been suggested to be specifically expressed by microglia 36 . Importantly, the appearance of the F4/80 hi population was not transient, since they remained 12 weeks after TAM administration (Supplementary Fig. 2D) and retained their surface marker phenotype ( Supplementary Fig. 2E). In further support of repopulation by two distinct types of microglia we observed that the homogenous distribution of CX3CR1 + P2ry12 + microglia in control brains was replaced by day 28 with pockets of CX3CR1 + P2ry12 + and CX3CR1 + P2ry12microglia-like cells (Fig. 1d). The P2ry12 antibody has been proposed to only label microglia and not peripherally-derived myeloid cells 34 and we observed that F4/80 hi parenchymal macrophages were indeed P2ry12 -( Supplementary  Fig. 2C). qPCR analysis of whole brain further confirmed gradual increases in P2ry12, Cx3cr1 and Siglech expression from days 7 to 28 (Supplementary Fig. 2A).
Local proliferation and infiltration repopulate the niche. Based on these data we hypothesized that the F4/80 low cells represented CNS-resident microglia that had expanded following depletion and that F4/80 hi cells were peripherally derived. To more conclusively address the origin of CNS-repopulating F4/80 low and F4/80 hi microglial cells we irradiated Cx3cr1 CreER/+ R26 DTA/+ and control Cx3cr1 CreER/+ mice (CD45.2) and reconstituted them with congenic (CD45.1) BM. To limit the effects of irradiation on the ability of microglia to repopulate the niche 30,37 we protected the head from irradiation. After 8 weeks of reconstitution we administered TAM and analyzed the chimeras after a further 8 weeks. Head-protected CD45.1 → Cx3cr1 CreER/+ R26 DTA/+ chimeras displayed frequencies of F4/80 low and F4/80 hi cells that were similar to those in non-irradiated mice (Fig. 2a). Using this system we could thus demonstrate that F4/80 low cells were completely host (CD45.2)-derived and that F4/80 hi cells were completely donor (CD45.1)-derived (Fig. 2a). To control for irradiation-induced damage we used a chemotherapy regimen to achieve myeloablation, a strategy that has previously been reported to lead to almost complete donor chimerism without spontaneously inducing CNS myeloid engraftment 38,39 . We thus produced Cx3cr1 GFP/+ Ccr2 RFP/+ → Cx3cr1 CreER/+ R26 DTA/+ chimeras where donor and host CX3CR1 + cells could be differentiated by GFP and YFP expression, respectively (Supplementary Fig. 3). Using this experimental setup we found that after depletion of microglia, practically all F4/80 hi cells were GFP + (Fig. 2b), confirming their peripheral origin.
These results supported our hypothesis of a dual origin of CNS repopulating macrophages and further suggested the involvement of a combination of local microglial proliferation and infiltration of BM-derived precursors in order to repopulate the niche. This was supported by high frequencies of Ki67 + microglia at day 7 and their absence at day 28 (Fig. 2c). To conclusively address whether the F4/80 low population arose by microglial proliferation in response to depletion we administered EdU during the repopulation phase, which confirmed a burst of proliferation in F4/80 low microglia during days 0-14 (Fig. 2d).
BM-derived microglial replacements are Ly6C hi monocytes. While Ly6C hi monocytes can act as precursors of many peripheral macrophage populations 21 , BM-derived microglia have been proposed to arise from myeloid progenitors without passing through a monocyte intermediate 30,40 . Furthermore, it has been suggested that blood brain barrier (BBB) disruption is a prerequisite for brain colonization by peripheral myeloid cells 41 . Our experiments using head protection and chemotherapy-induced myeloablation demonstrated that brain irradiation was not required for niche competition by peripheral precursors. In addition, we did not detect disruption of the BBB as a consequence of microglial depletion ( Supplementary Fig. 4A). Instead, we recorded increased mRNA levels in the CNS of the chemokine receptor Ccr2 and of the monocyte chemoattractants Ccl2, Ccl3, Ccl4, and Ccl5 at day 7 ( Supplementary Fig. 4B), suggesting chemotactic recruitment of Ly6C hi monocytes. In addition, we detected elevated mRNA levels of several other cytokines and chemokines at day 7, but most were back to baseline by day 14 (Supplementary Fig. 4C). This was accompanied by astrocytosis, as evidenced by both increased Gfap mRNA expression ( Supplementary Fig. 4D) and GFAP immunostaining ( Supplementary Fig. 2E).
To address the importance of Ly6C hi monocyte release from the BM we produced WT:Ccr2 −/− → Cx3cr1 CreER/+ R26 DTA/+ mixed BM chimeras (Fig. 3a) that lack CCR2 -/monocytes in the circulation (ref. 42 and Fig. 3b). We observed that the CNS F4/ 80 hi compartment was exclusively repopulated by WT cells (Fig. 3a), demonstrating a crucial role of CCR2 in this process. Consistent with this observation, a time course experiment encompassing depletion/repopulation revealed a wave of Ly6C hi monocytes entering the brain that peaked at day 2 after TAM, prior to the establishment of the F4/80 hi macrophage pool (Fig. 3b). At the peak of monocyte infiltration into the CNS we observed vascular activation in the brain, as evidenced by    Kinetics of depletion and repopulation of microglia in Cx3cr1 CreER/+ R26 DTA/+ mice. a Dot plots detailing depletion and repopulation of microglia in Cx3cr1 CreER/+ R26 DTA/+ mice on the indicated days after TAM administration. Top panel is gated on live singlet cells. Bottom panel is gated on CD11b + CD45 + Ly6C -Ly6G -. b Quantification of CX3CR1 + F4/80 low and CX3CR1 + F4/80 hi populations by flow cytometry. n = 12, 5, 9 mice, mean ± s.e.m. ***p < 0.001 by one-way ANOVA with Dunnett's Multiple Comparison test. Data are pooled from three separate experiments. The experiment was repeated twice for each time point. c Surface expression by flow cytometry on repopulated F4/80 low and F4/80 hi cells at day 28 compared to naive microglia. Gated on CD11b + CD45 + Ly6C -Ly6G -. n = 4 mice/group. Lines represent mean values. The experiment was performed twice. d Quantification of P2ry12 and CX3CR1-YFP staining in the indicated brain regions. Representative images are from the cortex. n = 3-7 mice/group, mean ± s.e.m. ***p < 0.001, **p < 0.01, *p < 0.05 by one-way ANOVA with Dunnett's Multiple Comparison test. Scale bar top panel 200 µm, bottom three panels 50 µm. Hippo. Hippocampus. Thal. Thalamus increased ICAM-1 staining (Fig. 3c). This suggested that peripherally derived macrophages could enter the CNS through the vasculature. Furthermore, Iba-1 staining did not reveal obvious infiltrates in meninges or choroid plexus, arguing against these anatomical areas as major infiltration sites ( Supplementary  Fig. 5A).
CCR2 is not an exclusive marker for Ly6C hi monocytes and can also be expressed on myeloid precursors and stem cells. However, we did not detect increased numbers of c-kit + progenitor cells in either the CNS ( Supplementary Fig. 5B) or the blood ( Supplementary Fig. 5C), arguing against repopulation from BM-derived stem cell/progenitors 43    not detect changes in neutrophils or lymphocyte numbers in the CNS following depletion ( Supplementary Fig. 5D-E). More importantly however, adoptive transfer of purified Ly6C hi monocytes from Cx3cr1 GFP/+ Ccr2 RFP/+ mice into microgliadepleted Cx3cr1 CreER/+ R26 DTA/+ mice during the peak of monocyte entry into the brain (days 0, 1, and 2) resulted in specific reconstitution of >70% of the F4/80 hi compartment without contributing to the F4/80 low microglial pool (Fig. 3d). Collectively, our results suggested a sequence of events whereby the remaining microglia proliferate extensively in order to refill the niche following depletion. Monocytes simultaneously infiltrate the brain and give rise to F4/80 hi macrophages.
To address whether Ly6C hi monocytes enter in one wave or continuously infiltrate the brain, we administered EdU at different time points following depletion. CNS-retrieved Ly6C hi monocytes were consistently~80% EdU + . When EdU was given during days 0-14 after TAM, the F4/80 hi macrophages were alsõ 80% EdU + . However, when EdU is administered during days 14-28 or days 56-80 after TAM, EdU incorporation into the F4/ 80 hi pool was not different from controls ( Supplementary Fig. 5F). This indicates that F4/80 hi macrophages are generated from Ly6C hi monocytes entering the brain directly following depletion.
Taken together, our data imply that in Cx3cr1 CreER/+ R26 DTA/+ mice the availability of an empty niche allows circulating monocytes to infiltrate the brain and to give rise to long-lived microglia-like cells.
Engrafting macrophages adopt TGF-β driven transcription. Cx3cr1 CreER/+ R26 DTA/+ mice thus represent a useful tool to study the impact of the CNS microenvironment on peripherallyderived macrophages. To investigate whether monocyte-derived macrophages adopted microglia gene expression in the microglial niche, we depleted microglia and profiled the transcriptomes of sorted F4/80 low and F4/80 hi populations ( Fig. 4a) after 4 and 12 weeks. In parallel we sorted Ly6C hi monocytes, common myeloid progenitors (CMP) and granulocyte-macrophage progenitors (GMP) from the BM, as well as splenic RPM and intestinal macrophages from naive mice for comparison. Hierarchical clustering of the 1000 most variable genes in the dataset separated samples according to tissue origin ( Supplementary  Fig. 6). Principal component analysis (PCA) based on all genes similarly organized all CNS macrophages into one cluster ( Fig. 4b) including F4/80 low and F4/80 hi subsets as well as naive microglia and recently proliferated microglia sorted from Cx3cr1 CreER/+ R26 DTR/+ mice, which express the diphtheria toxin receptor (DTR) 30,32 . This indicated a closer relationship between F4/80 hi macrophages and microglia compared to their BM precursors or other tissue macrophages, which clustered separately. This was further substantiated by the fact that F4/80 hi macrophages expressed several genes that have previously been described to be microglia-specific 44,45 (Fig. 4c), including Fcrls, P2ry12, P2ry13, Siglech, Olfml3, and Tmem119. In addition, several transcription factors that are associated with microglia homeostasis 46 were detected in F4/80 hi macrophages, including Egr1, Mafb, Mef2a, and Jun, with the exception of Sall1 (ref. 37 ) (Fig. 4d). Several of these genes are induced by TGF-β signaling 34 and we have recently demonstrated that abrogation of TGF-β signaling in monocyte-derived macrophages results in fatal demyelinating disease 31 . Consistently, we observed high levels of Tgfbr1 expression in both F4/80 low and F4/80 hi populations, which were similar to levels in naive microglia (Fig. 4e).
These results indicate that monocytes adopted key microgliaspecific genes subsequent to CNS engraftment.

Repopulating cells differ transcriptionally and functionally.
Comparison of the transcriptomes of naive microglia with other populations resulted in 4967 differentially expressed genes (>2 fold both directions, adj. p < 0.05) with Ly6C hi monocytes, 5240 with RPMs and 4370 with intestinal macrophages, respectively. F4/80 low microglia 4 weeks after TAM administration differentially expressed 361 genes compared to naive microglia, and only 56 genes differed after 12 weeks (Fig. 5a). For F4/80 hi macrophages the corresponding numbers were 886 and 1173, respectively (Fig. 5a). This indicated that after repopulation the F4/80 low transcriptome is affected, but almost returns to baseline within 12 weeks. Conversely, F4/80 hi macrophages do not completely adopt the naive microglia transcriptome. This was further supported by visualization of the genes that were differentially expressed in F4/80 hi macrophages compared to naive microglia ( Fig. 5b).
To assess if the unique gene signature in monocyte-derived macrophages resulted in functional changes, we first confirmed increased surface protein expression of Clec12a, CD36, and CXCR4 on F4/80 hi macrophages compared to F4/80 low microglia using flow cytometry (Fig. 5e). We next examined phagocytic capacity, since F4/80 hi macrophages expressed higher levels of scavenger receptors. We observed increased uptake of both E. coli microparticles as well as fluorescentlylabelled myelin by F4/80 hi macrophages compared to in both naive and repopulated F4/80 low microglia (Fig. 5f). Finally, we assessed their inflammatory cytokine and chemokine production capacity. Ex vivo isolated F4/80 hi macrophages secreted similar or lower levels of TNF, CCL3, CCL4 or CCL5 without stimulation or following LPS stimulation compared to F4/80 low microglia sorted in parallel ( Supplementary Fig. 7). We did not detect production of IL-1β, IL-10, IL-12, IFN-γ or CCL2 under these conditions. Our results thus far demonstrated that monocyte-derived macrophages could adopt key components of the microglia transcriptome, but retained large transcriptional and functional differences, even after long-term integration into the CNS.
Monocyte-derived macrophages adopt microglia DNA methylome. Epigenetic reprogramming has been demonstrated in Ly6G-V450 CD115-APC   Fig. 8A). Detailed analysis of the CNS cluster revealed that F4/80 low microglia clustered closely with naive microglia, demonstrating that their DNA methylation profile is largely conserved during repopulation ( Supplementary  Fig. 8A, B). Consistently, we observed only 139 differentially methylated sites between naive microglia and F4/80 low microglia at 7 weeks (adj. p < 0.05). For F4/80 hi macrophages and Ly6C hi monocytes these numbers were 879 and 2081 differentially methylated sites, respectively. Analysis of the most variable sites in the dataset revealed that F4/80 hi macrophages adopted a methylation profile distinct from both BM progenitors and their F4/80 low counterparts as these three groups occupied the extremes of a multidimensional scaling (MDS) plot visualizing the differences in the dataset (Supplementary Fig. 8B). Contrasting the DNA methylation profiles of these groups using a stringent p-value (p < 0.001) revealed 1486 sites that were differentially methylated between all groups. Visualization of these CpG sites demonstrated that F4/80 hi macrophages adopted an intermediate methylation profile between Ly6C hi monocytes and F4/80 low microglia ( Supplementary Fig. 8C). These results are consistent with the interpretation that monocyte-derived macrophages partly adopted the DNA methylation profile of microglia following CNS niche colonization.
Engrafting macrophages adopt a conserved gene signature. To investigate whether our transcriptomic profiles were biased by our DT-mediated depletion system, we compared our F4/80 low and F4/80 hi signatures to recently published datasets comparing CNS-engrafting macrophages and microglia. Cronk et al. 52 recently demonstrated that peripheral macrophages engraft the brain as a consequence of impaired microglial self-renewal, using Cx3cr1 CreER/+ Csf1r fl/fl mice. The authors used two additional models to at least partly replace the microglial pool with peripheral macrophages: whole-body irradiation followed by 9 month recovery, or a combination of the CSF1R antagonist PLX5562 and whole-body irradiation. We determined that our F4/80 hi gene signature was highly expressed in peripherallyderived macrophages but not locally derived microglia in all three models, and that the opposite was true for our F4/80 low gene signature (Fig. 6a).
In another recent study Bennett et al. 53 used intracerebral transplantation (ICT) of microglia or peripheral immune cells into the empty microglial niche in Csf1r -/mice, demonstrating that only cells of yolk sac origin would adopt bona fide microglial gene expression. We determined that the F4/80 hi gene signature was highly expressed in microglia-like cells derived from intracerebrally transplanted BM or blood, but not in intracerebrally transplanted microglia (Fig. 6b). Conversely, the F4/80 low gene signature was highly expressed in intracerebrally transplanted microglia, but not in blood or BM-derived microglia-like cells (Fig. 6b).
Taken together these results are consistent with the idea that monocyte-derived macrophages adopt a unique and reliable gene signature as a consequence of encountering the CNS microenvironment. In further support of this conclusion, our F4/80 hi and F4/80 low gene signatures were conserved in engrafting macrophages and naive microglia reported in two additional studies, one by Bruttger et al. 30 (Fig. 6c) and one by ourselves 31 (Fig. 6d), both using irradiation combined with microglia depletion.
Next we performed gene set enrichment analysis on the transcriptomic profiles of F4/80 low and F4/80 hi macrophages and naive microglia and visualized their pairwise comparisons using BubbleGum 54 . We initially focused on studies comparing brainengrafting macrophages with microglia. Overall, the gene sets identified in brain-engrafting macrophages were highly enriched in F4/80 hi macrophages, and gene sets identified in CNS resident microglia were enriched in F4/80 low and naive microglia (Fig. 6e). This analysis used gene sets from a study comparing the transcriptomic profiles of peripherally-derived and ARTICLE microglia-derived tumor-associated macrophages in glioma 55 , as well as the signature genes in peripherally-derived macrophages (beMφ-50) and microglia (Mg-52) described by Cronk et al. 52 (Fig. 6e). In addition, F4/80 hi macrophages displayed enrichment of a common gene signature identified in microglia isolated from mouse models of AD, ALS, and MS, termed neurodegeneration-associated microglia (MGnD) 56 (Fig. 6e). Furthermore, genes differentially expressed in Sall1 deficient microglia 37 were also highly enriched in F4/80 hi macrophages (Fig. 6e), which was of interest since F4/80 hi macrophages did not express Sall1. Taken together our analysis demonstrates that the F4/80 low and F4/80 hi gene signatures are conserved across several experimental systems and are unlikely to be biased by our microglia depletion system.
Type I IFNs regulate macrophage phenotype and colonization.
To further probe potential regulators of peripheral macrophage engraftment, we performed gene set enrichment analysis on the transcriptional profile of F4/80 low and F4/80 hi macrophages. We tested for enrichment of the hallmark gene sets, which are 50 well-defined gene sets that convey specific biological states or processes 57 and used BubbleGum to visualize the results. This analysis demonstrated that responses to IFN-γ and to IFN-α were the most significantly enriched biological processes in F4/80 hi macrophages (Fig. 7a). This result was further supported by ingenuity pathway analysis performed on the genes differentially expressed in F4/80 hi macrophages (>2 fold change both directions compared to F4/80 low microglia, adj. p < 0.05), which predicted several molecules related to interferon (IFN) signaling (IRF7, IRF3, IFN-α) to most potently affect the F4/80 hi gene signature (Fig. 7b). Furthermore, F4/80 hi macrophages expressed higher levels of the receptors for type I IFNs, Ifnar1, and Ifnar2, as well as several IFN stimulated genes, including Oas1g and Oas2 and Ifi44 (Fig. 7c) 58 . We thus asked whether type I IFNs regulate colonization of the microglial niche by monocyte-derived macrophages. To test this experimentally we produced Ifnar1 −/− : WT → Cx3cr1 CreER/+ R26 DTA/+ competitive chimeras, which resulted in approximately 65/35% (Ifnar1 -/-:WT) chimerism in blood Ly6C hi monocytes. We then depleted microglia using TAM administration and analyzed chimerism in the F4/80 hi compartment after 3 weeks. We observed that Ifnar1 -/cells had a competitive advantage over WT cells in repopulating the F4/80 hi compartment, resulting in approximately 85/15% (Ifnar1 -/-:WT) chimerism (Fig. 7d). This indicates that type I IFNs have an inhibitory effect on microglial niche colonization.

Discussion
Microglia are derived from yolk-sac progenitors that arise early during embryogenesis and that colonize the primitive brain 1 . The formation of the BBB is believed to halt any further colonization of the brain by primitive macrophages during embryogenesis. It is now established that microglia self-maintain throughout adulthood with little or no input from peripheral precursors 1-3 , this occurring as a stochastic process of microglial proliferation 59,60 . During certain conditions, however, as demonstrated herein and by others 29,30,52 , myeloid cells can colonize the brain and give rise to long-lived microglia-like cells. Understanding how monocytes take over the microglia niche is not only of biological importance but also of therapeutic interest, since replacement of mutant microglia with wild-type BM-derived cells has been proposed for neurological disorders such as ALS 61,62 and Nasu-Hakola disease 63 . The current study was undertaken to characterize this process and its underlying mechanisms.
In this paper we report that after depletion of microglia using Cx3cr1 CreER/+ R26 DTA/+ mice, the microglial niche is simultaneously repopulated by a combination of surviving resident microglia and CNS-infiltrating monocytes. While previous reports have described microglial proliferation or peripheral myeloid cell engraftment to occur as a consequence of microglia depletion 29,30,35,52,[64][65][66] , our report is the first that demonstrates that these processes can occur simultaneously and cooperatively to repopulate an empty microglial niche, effectively showing that these processes are not mutually exclusive. Using head-shielded chimeras and adoptive transfer, we could demonstrate that brain irradiation or stem cell release into the circulation were not prerequisites for peripheral macrophage repopulation in our model. Furthermore, monocyte engraftment occurred without BBB breakdown. However, the absence of microglia caused a transient cytokine storm which likely led to the observed vascular activation, coinciding with Ly6C hi monocyte entry into the CNS. Future studies should address the importance of this inflammation in attracting monocytes into the CNS.
The fact that proliferating microglia and engrafting macrophages adopted distinct surface marker expressions (F4/80 low and F4/80 hi , respectively) could be utilized to sort these populations and perform transcriptional, epigenetic, and functional profiling. One obvious benefit of our experimental paradigm is that we could thus control for the effect of the CNS environment during depletion/repopulation, since F4/80 low and F4/80 hi cells were sorted from the same brains. Consequently, our first observation was that the global gene expression profile of CNS-engrafting macrophages was more similar to microglia than to other tissueresident macrophages or their Ly6C hi monocyte precursors. Consistent with this we observed upregulation of microgliaspecific genes including P2ry12, Fcrls, Siglech, and Tmem119 in monocyte-derived macrophages, as well as microglia-expressed transcription factors Mafb and Mef2a. In addition, we observed a similar trend for the DNA methylation profile. In concordance with a recent proposal, 28 we interpret this as imprinting of monocyte-derived macrophages by the microglial niche. Importantly, however, Sall1, a transcription factor regulating microglia identity 37 was not expressed in monocyte-derived macrophages, whose transcriptional profile displayed enrichment of genes expressed in Sall1 -/microglia. Enforced expression of Sall1 in CNS-engrafting macrophages would be an important experiment to address the requirement for Sall1 expression in coordinating the bona fide microglia gene expression program. Our F4/80 low and F4/80 hi gene signatures consistently overlapped with the published transcriptional profiles of microglia and engrafting macrophages across several models of microglia depletion, ruling out the possibility that they were biased by the nature of our microglia depletion model. Rather, the conserved gene signature observed in F4/80 low and F4/80 hi subsets instead indicates that these gene sets represent fixed ontogeny-dependent expression programs. We consistently observed a high similarity between our F4/80 hi gene signature and that described in microglia-like cells derived from intracerebral transplantation of cells of hematopoietic stem cell origin into an empty microglial niche 53 , including high expression of Ms4a7, Clec12a, and Apoe. This was in contrast to cells of yolk-sac origin 53 , which displayed higher expression of microglia-specific genes, including Sall1, Sall3, and Slc2a5.
Why do monocyte-derived macrophages not fully adopt the microglia gene expression program despite being situated in the same organ space? One possibility is that microglial education occurring during development 46,67 induces a transcriptional and epigenetic program not obtainable by monocyte-derived macrophages that enter the adult brain. Alternatively, monocytes may contain unique epigenetic programming from the BM. In concordance with this notion we observed that the DNA methylation pattern of monocyte-derived macrophages was frozen between BM and microglia states. Interestingly, Lavin et al. have demonstrated considerable variation in the ability of BM-derived transplants to recover tissue macrophage-specific chromatin states. Depending on the tissue, transplant-derived macrophages isolated from the liver, spleen, lung or peritoneum adopted 47-92% of tissue-specific enhancer regions 24 . Such variability could potentially explain why monocytes do not fully adopt microglia gene expression, which contrasts to monocytes repopulating Kupffer cell 68 or alveolar macrophage 26 niches following experimental depletion.
Finally, to identify potential regulators of monocyte engraftment and repopulation we performed gene set enrichment and ingenuity pathway analyses, which identified a robust IFN signature in monocyte-derived macrophages. Importantly, type I IFNs regulate the phenotype of both CNS resident and infiltrating myeloid cells. Exaggerated type I IFN signaling in microglia leads to their activation and engulfment of synapses, manifesting as    neurological symptoms in lupus-prone mice 69 . However, local production of type I IFNs in the CNS suppress autoimmune neuroinflammation by modulating CNS engrafting macrophages 58 . We used a competitive chimera approach to address the requirement for IFNAR signaling in monocyte-derived macrophages in repopulating the microglial niche, and demonstrated that type I IFNs impair this process. Interestingly, this contrasts with our recent observation that TGF-β signaling promotes microglia niche colonization 31 . Future studies should aim to identify the crucial microenvironmental factors shaping CNS engrafting monocytes and how they may be manipulated in order for monocytes to completely adopt microglia identity.

Methods
Mice. All mice were bred and maintained under specific pathogen-free conditions at Karolinska Institutet, in accordance with national animal care guidelines. All animal experiments were approved by the appropriate ethical review board (Stockholms djurförsöksetiska nämnd). Cx3cr1 CreER-EYFP , R26 DTR , R26 DTA , CD45.1 were originally obtained from The Jackson Laboratory. Cx3cr1 GFP/ + Ccr2 RFP/+ , and Ccr2 -/-(RFP/RFP) mice were a gift from Klas Blomgren at Karolinska Institutet. Ifnar1 -/were originally obtained from Ulrich Kalinke. Experiments were started when mice were 6-12 weeks.
In vivo treatments. Tamoxifen administration: Tamoxifen (TAM; Sigma) was suspended in corn oil at 75°C and 5 mg (200 µl) was unless otherwise specified administered subcutaneously (s.c.) on three consecutive days.
Diphtheria toxin administration: Microglia were depleted in Cx3cr1 CreER/+ R26 DTR/+ by two subcutaneous doses of 5 mg TAM separated by 48 h followed 3 weeks later by three daily doses of 25 ng/g diphtheria toxin, given intraperitoneally.
EdU administration: Mice were fed for 2 weeks with EdU in drinking water (0.2 mg/ml + 1% sucrose) from days 0-14 or 14-28 after TAM injections. The solution was protected from light at all times and changed once per week. EdU was detected by flow cytometry according to manufacturer's instructions (Invitrogen).
Evans Blue administration: To assess BBB integrity mice were injected i.p with 200 μl 2% Evan's Blue solution (Sigma). Twenty-four hours later mice were perfused and brains collected, weighed and homogenized in 1 ml 3.05 M trichloroacetic acid. The homogenate was centrifuged for 10 min at 10,000 × g and the supernatant mixed 1:3 with 95% ethanol. Samples were excited at 625 nm and emission read at 680 nm using a Glomax fluorescence plate reader.
Generation of BM chimeras. Mice were irradiated with 9.5 Gray using an X-RAD 320 irradiation source (0.95Gray/minute) with a 20 × 20 cm irradiation field. Head protection was accomplished by retaining mice under isofluorane anaesthesia and placing the head (from the neck up) outside the field of irradiation. Mice were monitored throughout the irradiation period to make sure the heads stayed outside the field (the irradiation source is equipped with a lamp to visualize the irradiation field). Mice were reconstituted on the same day with 2-5 × 10 6 BM cells by tail vein injection. Mice were considered reconstituted and used for experiments 6-8 weeks later. Head-protected mice generally resulted in 50-80% chimerism.
Chemotherapy. Myeloablation using chemotherapy was performed by administering 20 mg/kg Busulfan (Sigma) by i.p. injection on three consecutive days. Busulfan was solubilized in DMSO and diluted to 5 mg/ml in PBS. On the fourth day 12-14 × 10 6 BM cells were given by tail vein injection. This procedure resulted in 80-98% chimerism.
Preparation of single cell suspensions. Mice were sacrificed by injecting 100 µl pentobarbital (i.p.). When applicable, blood was collected from the right ventricle prior to perfusion. Mice were perfused with ice-cold PBS and organs dissected. CNS cells were prepared by enzymatic digestion using Collagenase D (1 mg/ml, Roche) and DNAse I (0.2 mg/ml, Roche) or Neural Tissue Dissociation Kit T (Miltenyi Biotec). Myelin was removed using 38% Percoll.
Microarray analysis. For RNA preparation cells were sorted into a solution of RNA later (Thermo Fisher Scientific). RNA was prepared using RNeasy Micro Kit (Qiagen). RNA quality and integrity was assessed using a Bioanalyzer 2100 (Agilent). All samples included had high quality (RIN = 7-10). cDNA preparation, hybridization and scanning were performed at the Array and Analysis Facility, Science for Life Laboratory at Uppsala Biomedical Center (BMC). 0.5-10 ng of total RNA from each sample was used to generate amplified and biotinylated sensestrand cDNA from the entire expressed genome according to the GeneChip WT Pico Reagent Kit User Manual (Affymetrix). GeneChip Mouse Gene 2.1 ST Arrays were hybridized for 16 h in a 45°C incubator, washed and stained and finally scanned at the GeneTitan Multi-Channel Instrument, according to the GeneTitan Instrument User Guide for Expression Arrays Plates (Affymetrix).
Gene Set Enrichment analysis was performed and visualized using BubbleGum using standard settings. Ingenuity pathway analysis (IPA, Qiagen) was used to analyze upstream regulators.
DNA methylation analysis. DNA was prepared using QIAamp DNA Micro Kit (Qiagen). DNA methylation analysis was performed using Infinium Methylatio-nEPIC BeadChip (Illumina) with ≥250 ng input DNA at the Bioinformatics and Expression Analysis (BEA) core facility, Karolinska Institutet. EPIC probes (50-mer oligonucleotides) were mapped to GRCm38/mm10 using Bismark (version 0.14.5) with default settings. A total of 19,420 probes were identified as unique hits (conserved between human and mouse) and used for downstream analysis. Data was analyzed using the ChAMP package, probes were filtered for mapping probes with 17,633 probes passing detection cutoff, and subsequently SWAN normalized. Differential probes were identified using Limma (p < 0.01) and clustering was preformed using PCA analysis.
qPCR. Hemibrains or coronal brain sections were homogenized in RLT-buffer (Qiagen) containing beta-mercaptoethanol (14.3 M, 10 µl/ml, Sigma) and RNA prepared using the RNeasy Mini Kit (Qiagen). cDNA was synthesized using the iScript cDNA Synthesis Kit (Bio-Rad). qPCR was performed with SYBR green reaction (Bio-Rad). All expression levels are reported relative to Hprt or Hprt and Gapdh. Primer sequences can be found in Supplementary Table 1.