1. Immunology and Inflammation
  2. Neuroscience
Download icon

Subventricular zone/white matter microglia reconstitute the empty adult microglial niche in a dynamic wave

  1. Lindsay A Hohsfield
  2. Allison R Najafi
  3. Yasamine Ghorbanian
  4. Neelakshi Soni
  5. Joshua Crapser
  6. Dario X Figueroa Velez
  7. Shan Jiang
  8. Sarah E Royer
  9. Sung Jin Kim
  10. Caden M Henningfield
  11. Aileen Anderson
  12. Sunil P Gandhi
  13. Ali Mortazavi
  14. Matthew A Inlay
  15. Kim N Green  Is a corresponding author
  1. Department of Neurobiology and Behavior, United States
  2. Institute for Memory Impairments and Neurological Disorders, United States
  3. Sue and Bill Gross Stem Cell Research Center, United States
  4. Department of Molecular Biology and Biochemistry, United States
  5. Department of Developmental and Cell Biology, United States
  6. Department of Anatomy and Neurobiology, United States
  7. Department of Physical Medicine & Rehabilitation, University of California, Irvine, United States
Research Article
  • Cited 0
  • Views 1,156
  • Annotations
Cite this article as: eLife 2021;10:e66738 doi: 10.7554/eLife.66738

Abstract

Microglia, the brain’s resident myeloid cells, play central roles in brain defense, homeostasis, and disease. Using a prolonged colony-stimulating factor 1 receptor inhibitor (CSF1Ri) approach, we report an unprecedented level of microglial depletion and establish a model system that achieves an empty microglial niche in the adult brain. We identify a myeloid cell that migrates from the subventricular zone and associated white matter areas. Following CSF1Ri, these amoeboid cells migrate radially and tangentially in a dynamic wave filling the brain in a distinct pattern, to replace the microglial-depleted brain. These repopulating cells are enriched in disease-associated microglia genes and exhibit similar phenotypic and transcriptional profiles to white-matter-associated microglia. Our findings shed light on the overlapping and distinct functional complexity and diversity of myeloid cells of the CNS and provide new insight into repopulating microglia function and dynamics in the mouse brain.

Introduction

Microglia represent the largest population of immune cells in the brain, constituting 5–10% of brain cells in the adult central nervous system (CNS). As resident tissue macrophages, microglia are responsible for immune defense and resolution, tissue maintenance, neuronal support, and synaptic integrity (Salter and Stevens, 2017; Tay et al., 2017; Wolf et al., 2017). Their central role in the CNS makes microglia attractive drug targets for neurological disorders/injuries. However, developing effective therapies that manipulate microglia requires further understanding of microglial origins, diversity, homeostasis, and dynamics.

Microglia arise from yolk sac-derived erythromyeloid progenitors and colonize the brain as embryonic microglia during early stages of development (i.e. E8.5 – E9.5) (Ginhoux et al., 2010; Kierdorf et al., 2013). These immature myeloid cells, displaying amoeboid morphology and high proliferative potential, enter the brain via the meninges and ventricles in mice (Lelli et al., 2013; Ueno et al., 2013; Swinnen et al., 2013; Xavier et al., 2015), as well as, via the leptomeninges, choroid plexus, and ventricular zone in humans (Tay et al., 2017; Verney et al., 2010; Monier et al., 2007; Ginhoux et al., 2013). Microglial colonization occurs first in the white matter (WM) (e.g. internal capsule, external capsule, and cerebral peduncle) and continues to the sub- and then cortical plate as cells proliferate and migrate in a radial and tangential manner (Verney et al., 2010). In adulthood (P28 and onward), microglia become fully mature, exhibiting ramified morphology and expressing canonical microglial signature genes: P2ry12, Tmem119, Siglech, Cx3cr1, Olfml3, Fcrls, and Sall1 (Dubbelaar et al., 2018; Baufeld et al., 2018).

Recent single-cell RNA sequencing studies have identified transcriptionally distinct microglial gene signatures associated with disease (e.g. disease-associated microglia [DAM], microglial neurodegenerative [MGnD] phenotype [Keren-Shaul et al., 2017; Krasemann et al., 2017; Mathys et al., 2017; Masuda et al., 2019]), injury (e.g. injury-responsive microglia [IRM] Hammond et al., 2019), and brain region-specific areas/developmental stages (e.g. proliferative-region-associated microglia [PAM], axon tract-associated microglia [ATM] Masuda et al., 2019; Hammond et al., 2019; Li et al., 2019; Matcovitch-Natan et al., 2016), including the recent discovery of white matter-associated microglia (WAM) (Safaiyan et al., 2021). WAMs have been identified as a population of microglia in WM tracts from the corpus callosum that share parts of the DAM gene signature, including genes involved in phagocytosis, and increase with aging and disease (Safaiyan et al., 2021). In line with this, a distinct subset of microglia (PAMs/ATMs) has been described in the axonal tracts of the corpus callosum during development, sharing not only a similar location to WAMs, but morphological (i.e. amoeboid) and phagocytosis-associated gene profile (Masuda et al., 2019; Hammond et al., 2019; Li et al., 2019). Despite this, homeostatic microglia appear less heterogeneous during adulthood (Hammond et al., 2019; Li et al., 2019). These findings shed light on microglial diversity and state changes during health and disease; however, it remains unclear whether adult homeostatic microglia exist as one population, with the ability to change from one transcriptional/functional state to another, or whether they exist as heterogeneous subpopulations with distinct propensities.

Microglial homeostasis and dynamics are maintained by many signaling factors, including transforming growth factor-beta, Il-34, and colony-stimulating factor 1 (CSF1) (Butovsky et al., 2014; Elmore et al., 2014). Recent studies exploring the homeostatic kinetics of the microglia in the adult brain have revealed that these cells are long-lived (Tay et al., 2017; Réu et al., 2017) and self-renew, even after acute 80–95% depletion and subsequent repopulation (Elmore et al., 2014; Zhan et al., 2019; Bruttger et al., 2015). However, no approach to date has been able to deplete all microglia (Waisman et al., 2015; Han et al., 2017), and the rapid proliferation of surviving microglial cells would obscure the detection of other myeloid cells that contribute to the CNS environment and/or repopulation in the adult brain. While conventional depletion paradigms have shown that microglia have a remarkable capacity to repopulate from the presence of few surviving cells, we sought to investigate the consequences of eliminating these few remaining cells on microglial population dynamics.

To address this, we have optimized a colony-stimulating factor 1 receptor (CSF1R) inhibitor approach that involves sustained inhibitor administration, building on our prior work that microglia are dependent on this signaling for their survival (Elmore et al., 2014). This approach results in a delayed repopulation of myeloid cells that reconstitute the brain in a sequential manner previously unseen in the adult brain. We show that repopulating cells emerge from the subventricular zone (SVZ)/WM areas and traffic throughout the brain parenchyma via WM tracts before spreading out radially and tangentially through the rest of the brain in a dynamic wave of proliferating cells. Following full brain reconstitution, these repopulating cells remain phenotypically, transcriptionally, and functionally distinct from endogenous microglia, demonstrating unique gene expression profiles that are enriched for DAM genes and unique phenotypic properties similar to WAMs. Together, these data highlight the utility of CSF1R inhibitors in identifying and studying myeloid cell homeostasis and dynamics.

Results

Sustained high dose of CSF1R inhibitor unmasks a distinct form of myeloid cell CNS repopulation

In previous studies, we have shown that 7 day treatment of the brain penetrant CSF1R/KIT/FLT3 inhibitor PLX3397 (Pexidartinib; 600 ppm in chow) eliminates ~90–98% of microglia in the CNS (Elmore et al., 2014; Najafi et al., 2018). During depletion, surviving microglia are seen scattered throughout the brain (Figure 1A–B,C) and subsequent withdrawal of the inhibitor results in rapid and spatially homogenous microglial repopulation within 3 days, with cells exceeding control numbers by 7 days (Figure 1C,E). Recent studies show that repopulation is dependent on the local proliferation and clonal expansion of surviving microglia (Elmore et al., 2014; Zhan et al., 2019; Bruttger et al., 2015; Huang et al., 2018; Zhan et al., 2020; Mendes et al., 2021); thus, we refer to this type of repopulation as global microglial (GLOBAL) repopulation.

Sustained high dose of CSF1R inhibitor unmasks a distinct form of myeloid cell CNS repopulation.

(A–B) Experimental paradigm and schematic depicting dose and duration of PLX3397 (600 ppm) treatment and subsequent inhibitor withdrawal allowing for global microglial (GLOBAL) and white matter microglia (WM) repopulation. For GLOBAL repopulation: 2-month-old wild-type (WT) mice were treated with 600 ppm of PLX3397 for 7 days, achieving ~90–98% brain-wide microglial depletion, with remaining microglia visibility dispersed throughout the brain parenchyma, and then placed on control diet for 7 days (7d recovery) allowing for microglial repopulation. At 7 day recovery, repopulating microglia reconstitute the brain from areas in which previously remaining microglia were deposited. For WM repopulation: 2-month-old WT mice were treated with 600 ppm of PLX3397 for 14 days, achieving 99.98% brain-wide microglial depletion, and then placed on control diet for 7 days allowing for microglial repopulation. At 7 day recovery, repopulating myeloid cells reconstitute the brain in specific neuroanatomical niches (e.g. ventricle, subventricular zone, white matter tracts, caudoputamen). (C–D) Representative immunofluorescence whole brain images of myeloid cells (IBA1, green) at each time point of treatment and recovery during GLOBAL (C) and WM (D) repopulation, with white dots superimposed over microglia. Due to the differential kinetics of these two forms of repopulation, repopulation (i.e. recovery) was evaluated at different timepoints, including initial stages with few cells, mid-repopulation, and full brain reconstitution. For GLOBAL repopulation: 2-month-old WT mice treated with control, 7 days of PLX3397 (7d PLX3397), 7 days of PLX3397 followed by 1 day on control diet (1d Recovery), 7 day of PLX3397 followed by 2 days on control diet (2d Recovery), 7 days of PLX3397 followed by 3 days on control diet (3d Recovery), and 7 days of PLX3397 followed by 7 days on control diet (7d Recovery). For WM repopulation: 2-month-old WT mice treated with control, 14 days of PLX3397 (14d PLX3397), 14 days of PLX3397 followed by 3 days on control diet (3d Recovery), 14 days of PLX3397 followed by 7 days on control diet (7d Recovery), 14 days of PLX3397 followed by 14 days on control diet (14d Recovery), and 14 days of PLX3397 followed by 28 days on control diet (28d Recovery). (C1–C4, D1–D4) Inserts of higher resolution confocal images of IBA1+ cells during repopulation. White dotted lines and yellow arrows highlight ‘wave’ edge and direction. (E–G) Quantification of IBA1+ cells per field of view (FOV) at each time point in cortical and white matter regions, respectively during GLOBAL (E) and WM (F–G) repopulation. (H) Pharmacokinetics analysis of PLX3397 levels in plasma and brain of mice treated with 7 day and 14 day PLX3397 (600 ppm). (I) Representative 63x immunofluorescence images of myeloid cells (IBA1, white) display morphological alterations. (J–L) Quantification of IBA1+ cell morphology: cell body area in the white matter tract (J), average process/filament length (K), and process complexity/branching (L) in the piriform cortex. Level 1–3+ indicates level of branching from the cell body. Data are represented as mean ± SEM (n=3–4). *p < 0.05, ** p < 0.01, *** p < 0.001; significance symbols represent comparisons between groups: (E) control *, 0d #, 1d ⋄, 2d Δ, 3d Φ; (F–G) control *, 0d #, 3d ⋄, 7d Δ. CC, corpus callosum; CTX, cortex; HC, hippocampus; LV, lateral ventricle.

Figure 1—source data 1

Sustained high dose of CSF1R inhibitor unmasks a distinct form of myeloid cell CNS repopulation.

https://cdn.elifesciences.org/articles/66738/elife-66738-fig1-data1-v2.xlsx

Here, we set out to examine CNS myeloid cell repopulation dynamics in the absence of remaining microglia in the brain. To accomplish this, we utilized a sustained high dose of PLX3397 (600 ppm) for 14 days. This treatment results in pharmacologically unprecedented microglial depletion, in which we observe no IBA1+ cells across whole brain sections (Figure 1A–B,D). Although PLX5622 is a CSF1R inhibitor (CSF1Ri) that is more active against CSF1R compared to other related kinases, studies in our lab have shown that high dose PLX3397 vs. high dose PLX5622 achieves higher CNS exposure and microglial depletion efficiency. Pharmacokinetic analysis of microglia-depleted brains at both 7d and 14d treatment of PLX3397 shows that PLX3397 levels remain the same in the CNS despite longer drug exposure (Figure 1H).

To explore the differential repopulation dynamics between these microglial elimination paradigms, we treated mice with PLX3397 for 14 days and then withdrew the inhibitor, allowing the CNS to recover for 3, 7, 14, and 28 days (Figure 1D,F–G), and compared it to GLOBAL repopulation (Figure 1C,E). At 3 day recovery following 14 day PLX3397 treatment, no IBA1+ cells are detectable in most brain sections. By 7 day recovery, IBA1+ cells appear, but are exclusively located near the lateral ventricle and in WM tracts lining the ventricles (Figure 1D2). By 14 day recovery, IBA1+ cells have spread throughout most of the CNS; however, some areas of the cortex (e.g. the piriform cortex) remain unoccupied (Figure 1D3). These areas display a distinct ‘wave’ of cells in adjacent unoccupied cortical areas (Figure 1D3). At 28 day repopulation, all brain regions are populated with IBA1+ cells, but absolute cell numbers remain 50% lower compared to microglia in control animals, as seen in somatosensory cortices (Figure 1D4, F). We subsequently refer to this form of repopulation as white matter microglia (WM) repopulation, due to its distinct characteristics from GLOBAL repopulation.

We have previously shown that within 14–21 days of GLOBAL repopulation, repopulating microglia not only attain similar densities to resident microglia, but also display similar morphologies, cell surface marker expression, gene expression profiles, and inflammatory responses to LPS (Elmore et al., 2014; Elmore et al., 2015; Elmore et al., 2018). In contrast, WM repopulating cells display larger cell bodies (Figure 1I–J) after 7 day recovery (which normalize by 14 day recovery), as well as reduced process/filament length (Figure 1I,K) and reduced process/dendrite branching and complexity (Figure 1I,L) compared to homeostatic microglia even after 28 days recovery.

WM repopulation elicits a dynamic wave of repopulating proliferative myeloid cells

In GLOBAL repopulation, microglia repopulate the brain parenchyma in a homogeneous fashion, with repopulating cells displaying no preference for specific locations (Figure 1C). In contrast, WM repopulating cells first appear in precise ventricular and WM locations (Figure 1D2). To expand upon this initial observation, we sought to define the anatomical niches of this distinct form of repopulation and built a spatial heat map at three brain positions (Bregma 2.58 mm, 1.10 mm, −2.06 mm) along the rostral-caudal axis in brains of mice at 7 day recovery (Figure 2A). Throughout this axis, IBA1+ cells initially repopulate the brain within the caudoputamen, particularly in areas near the lateral ventricle and associated WM tracts. At rostral regions of the brain, cells are found within the rostral migratory stream (RMS), a projecting axonal tract from the SVZ to the olfactory bulb. In more caudal brain regions, repopulating cells are seen near the SVZ, caudoputamen, and corpus callosum. Subsequent analysis of the entire brain confirms the presence of these early repopulating cells in areas near the SVZ/ventricular zones, WM tracts, and caudoputamen.

Figure 2 with 1 supplement see all
White matter microglia (WM) repopulation elicits a dynamic wave of repopulating proliferative myeloid cells.

(A) Spatial heat map of WM repopulation at 7 day recovery in three brain positions (Bregma 2.58 mm, 1.10 mm, −2.06 mm) depicting % of brains with cells present in specific brain regions (n=5–19). (B–C) Immunofluorescence whole brain coronal (B) and sagittal (C) section images of IBA1+ cells (green) show a sequential time course of the ‘wave’ (see higher resolution image of wavefront in B1 insert) of myeloid cells filling the brain between 7 and 14 day recovery during WM repopulation. White dotted lines and yellow arrows indicate the edge and direction of the ‘wave’, highlighting the radial migratory patterns of WM repopulating cells. Yellow dashed arrows indicate the direction of the tangential migratory pattern of WM repopulating cells, mostly utilizing WM tracts. The straight white dashed line in C shows the Bregma position at which coronal sections were taken for B. (E–F) Representative immunofluorescence whole brain images of myeloid cells (IBA1, green) and proliferating cells (Ki67, red) at 7 (A) and 14 (B) day recovery. Insets show higher resolution of IBA1 and Ki67 colocalization in initially repopulating cells (A1) and in cells outside and inside the wavefront (B2-4). Data are represented as mean ± SEM (n=3–8).

Figure 2—source data 1

White matter microglia (WM) repopulation elicits a dynamic wave of repopulating proliferative myeloid cells.

https://cdn.elifesciences.org/articles/66738/elife-66738-fig2-data1-v2.xlsx

Repopulating cells migrate in a radial and tangential fashion, initially filling the WM and caudoputamen before spreading out through the cortex in a dynamic wave between 7 and 14 day recovery (Figure 2B–C). Analysis of the migratory wave during this recovery time from Bregma 0.445 to Bregma −0.08 shows repopulating cell deposition appears to occur in stages, with symmetrical distribution and expansion (Figure 2—figure supplement 1A). At Stage I, cells are visible near the lateral ventricle at the intersection of the corpus callosum, caudoputmen, and SVZ and appear to migrate in a radial migration pattern from the ventricular zone into and filling the caudoputamen in an inferior direction. At this stage, cells also appear in WM tracts, specifically in the corpus callosum. At Stage II, cells begin to migrate in a tangential migration pattern utilizing WM tracts to migrate into WM areas or areas near WM tracts. At Stage III, repopulating cells continue to migrate in both a radial and tangential migratory pattern filling ~80–95% of the striatum, including the caudoputamen, lateral septal complex, and pallidum. It is also apparent that repopulating cells are restricted or unable to migrate past the WM tract or corpus callosum between the cerebral nuclei and cerebral cortex (Figure 2—figure supplement 1B). At Stage IV, the cells have broken through this WM tract barrier and migrate out in a radial migration pattern moving from the subcortical to the cortical plate (Figure 2B–D; Figure 2—figure supplement 1B). During Stages III-IV, patches of cortical microglial expansion are occasionally seen; however, the majority of proliferating/Ki67+ cells that contribute to WM repopulation are found in the wave rather than in cortical clusters of expanding microglia.

In contrast to the proliferative profile of previously described GLOBAL-repopulating microglia, in which remaining microglia proliferate throughout the brain to give rise to newly repopulating cells, proliferating WM repopulating myeloid cells are initially found near ventricles and WM tracts (Figure 2E and E1, Figure 2F–G). As the cells spread and migrate through the brain, proliferation remains localized within the cell ‘wavefront’ (Figure 2F, F and F4). Once out of the front (i.e. in the wake of the wave) IBA1+ cells appear to stop proliferating (i.e. Ki67-; Figure 2F3). This wave of proliferating cells is most apparent at Stage IV, led by a wavefront with an average width of ~100–150 µm of amoeboid and proliferating myeloid cells.

Extensive CSF1R inhibition unveils the presence of CSF1Ri-resistant myeloid cell in the subventricular zone/white matter areas

Having demonstrated that 14 day PLX3397 (600 ppm) treatment and subsequent withdrawal results in reconstitution of the adult brain with a phenotypically distinct myeloid cell with unique tempo-spatial migratory patterns, we next sought to determine the source of these cells. To that end, we first confirmed the extent of microglial depletion with multiple myeloid markers, including microglial-specific P2RY12 and TMEM119 (Figure 3—figure supplement 1A–B), as well as in Cx3cr1CreERT2 mice, to explore whether surviving cells were present but just downregulating myeloid markers. In these mice, YFP is permanently expressed in microglia following tamoxifen-inducible lineage tracing, illustrating that depletion is not due to a downregulation in microglial markers, but a loss of cells (Figure 3—figure supplement 1C). While in previous analyses, we observed no IBA1+ (including Cd11b+, P2RY12+, and TMEM119+) cells throughout the brains following 14 day PLX3397 treatment, we next conducted an examination throughout the entire brain along the rostral-caudal axis (i.e. every 6th section). With this extensive analysis, we observe a very small number of surviving IBA1+ cells in treated brains (~15 in 14 day treated PLX3397 brains vs. ~132,000 in control brains = 99.98% depletion; Figure 3—figure supplement 1D). Despite this, we describe the highest reported loss of microglial cells in the adult brain. Notably, these few cells (0.02% of cells) are seen exclusively in ventricular (i.e. SVZ) and adjacent WM areas (Figure 3A–D). These cells display a lack of canonical microglial markers, including P2RY12 and TMEM119, as well as distinct morphological profiles (Figure 3E–G). Examination of brains depleted for 3.5 month PLX3397 (600 ppm) revealed no further surviving cells in the WM/WVZ, suggesting that these cells eventually succumb to CSF1Ri (Figure 3—figure supplement 1F–H). These findings indicate that surviving SVZ/WM microglia may possess a different sensitivity to CSF1R inhibitors, possibly relying on other growth factors for survival, or that CSF1Ri kinetics in WM may be different to grey matter areas due to lipid abundance or inhibitor solubility in lipids. Reports show that the population of microglia residing in the adult SVZ and adjacent RMS display a distinct morphological profile with an amoeboid cell body and fewer/shorter branched processes and exhibit an activated phenotype, similar to WM repopulating cells (Ribeiro Xavier et al., 2015; Goings et al., 2006; Böttcher et al., 2019). Prior descriptions of myeloid cells found in the adult SVZ have found lower expression levels of the microglial-specific marker P2RY12 (Ribeiro Xavier et al., 2015). Here, we find repopulating cells are initially negative for both microglial-specific P2RY12 and TMEM119 surface markers, however, express these markers by 28 day recovery (Figure 3H–I, Figure 3—figure supplement 1E).

Figure 3 with 1 supplement see all
Extensive CSF1R inhibition unveils the presence of CSF1Ri-resistant myeloid cells in the subventricular zone and white matter tracts.

Two-month-old WT mice treated with vehicle or PLX3397 (600 ppm in chow) for 14 days to evaluate extent of microglial depletion. (A) Brain section schematic of SVZ and WM areas where CSF1Ri-resistant myeloid cells are present following 14 day PLX3397 (600ppm in chow) treatment. (B) Representative whole brain slice image of CSF1Ri-resistant IBA1+ cells (white arrows) near the SVZ/WM areas (white box) surrounding the lateral ventricle. (C–D) Representative tile scan (C) and high-resolution confocal (D) images of control and 14 day PLX3397 mice showing the deposition of surviving SVZ/WM IBA1+ (green) cells in WM areas (white dotted lines). (E–G) Representative tile scan (E–F) and confocal immunofluorescence image of IBA1+ cells co-stained with microglial-specific markers TMEM119 (E) and P2RY12 (F–G) in control and 14 day PLX3397 mice, showing that CSF1Ri-resistant cells are TMEM119- and P2RY12-. Higher resolution images illustrate atypical morphological profile of CSF1Ri-resistant cells. (H) Representative 20x images of myeloid cells (Cd11b, red) and P2RY12 (green). (I) Quantification of % colocalization of CD11b+ and P2RY12+ cells as seen in (H). (J) Control, 14 day PLX3397, 7 day recovery, and 28 day recovery mouse hemispheres were collected and analyzed for bulk-tissue gene expression changes using Nanostring Immune Profile. (K–L) Volcano plots displaying the fold change of genes (log2 scale) and their significance (y axis, -log10 scale) between 14 day PLX3397 depleted vs. 7 day recovery mice (K) and control vs 28 day recovery (L). Data are represented as mean ± SEM (n=3–5). *p < 0.05, ** p < 0.01, *** p < 0.001; significance symbols represent comparisons between groups: control *, 0d #, 3d ⋄, 7d Δ, 14d Φ. CC, corpus callosum; CP, caudoputamen; CTX, cortex; HC, hippocampus; LV, lateral ventricle; SVZ, subventricular zone; WM, white matter.

Figure 3—source data 1

Extensive CSF1R inhibition unveils the presence of CSF1Ri-resistant myeloid cells in the subventricular zone and white matter tracts.

https://cdn.elifesciences.org/articles/66738/elife-66738-fig3-data1-v2.xlsx

To gain insight into the transcriptional profile of these WM repopulating cells, we measured mRNA transcript levels from control, 14 day PLX3397-treated, 7 day recovery, and 28 day recovery whole brain hemispheres, using a Nanostring Immune Profiling panel (~700 immunology-related genes) (Figure 3J). Comparing 14 day PLX3397-treated brains (i.e. microglial-depleted) to 7 day recovery brains allowed us to explore the gene expression profile of the initial repopulating cells. This comparison revealed that the most upregulated genes in the 7 day recovery brain are involved in myeloid cell activation/priming, pathogen sensing, and monocyte-macrophage signaling (e.g. Mrc1, C3ar1, Ccl12, Clec7a, Ccr2/Ccl2, Cybb, and Ccl9), rather than homeostatic microglia signature genes (Figure 3K). In comparing 28 day recovery brains to controls, increased expression was detected in several genes associated with myeloid cell signaling (Ccl8, Cmah, Ly9, Lyz2, Tlr8, C4b), in particular, major histocompatibility complex I (H2-D1) and II (H2-Aa, H2-Ab1, Cd74) components and microglial priming (Clec7a, Cybb) (Figure 3L).

To explore whether WM repopulating microglia maintain resistance to CSF1Ri following their migration from the SVZ, we treated 7 and 14 day recovery mice for 7 days with PLX3397 (600 ppm) (Figure 3—figure supplement 1I). At both recovery times, the majority of cells were eliminated (Figure 3—figure supplement 1K) showing that WM repopulated cells are susceptible to CSF1Ri treatment and require CSF1R signaling for their survival. Together, these data provide evidence for the existence of a very small population of myeloid cells located in the SVZ and adjacent WM tracts that can uniquely survive 14 day PLX3397 high dose CSF1Ri treatment.

WM repopulation occurs due to an unprecedented level of microglial depletion

To conclusively determine whether this unique form of repopulation occurs as a result of the unprecedented level of microglial depletion vs. a 14 day requisite CSF1Ri drug treatment, we utilized H2K-BCL2 mice, a transgenic mouse that overexpresses BCL2 in all hematopoietic cells (Domen et al., 1998). Similar to reports in Vav-Bcl2 mice (Askew et al., 2017), these mice display elevated microglial densities (Figure 4A). In H2K-BCL2 mice, 14 day treatment with PLX3397 (600 ppm) leads to a 61–95% elimination of microglia (Figure 4B–D). In line with this, a previous study has shown that overexpression of BCL2 in myeloid cells affords some resistance to tissue macrophage loss in Osteopetrotic (op/op) mice, a mouse lacking functional CSF1 (Lagasse and Weissman, 1997).

WM repopulation occurs due to an unprecedented level of microglial depletion.

(A) Representative 20x immunofluorescence images of IBA1+ (green) cells in WT and H2K-BCL2 mice. Quantification of IBA1+ cells per region of interest (ROI) in subiculum (SUB) and cortex (CTX). (B–D) H2K-BCL2 mice were treated with PLX3397 for 14 days (600 ppm in chow), the drug was withdrawn, and then mice were provided with 7 days to recover, allowing for repopulation. (B) Representative whole brain images of IBA1+ (green) cells in control, 14 day PLX3397, and 7 day recovery mice, with white dots superimposed over microglia, showing incomplete microglial depletion leads to GLOBAL repopulation. (C) Quantification of IBA1+ cells per whole brain slice, as seen in (B). (D) Higher resolution images of IBA1+ (green) and PU.1+ (red), a myeloid cell marker, cells. Data are represented as mean ± SEM (n=4–6). # p < 0.1, *p < 0.05, *** p < 0.001.

Figure 4—source data 1

WM repopulation occurs due to an unprecedented level of microglial depletion.

https://cdn.elifesciences.org/articles/66738/elife-66738-fig4-data1-v2.xlsx

BCL2 is a major regulator of apoptosis, and enhanced Bcl2 expression promotes the survival of cells of myeloid cells (Ogilvy et al., 1999), thus we postulate that BCL2 protection from cell death occurs through reduction in apoptosis. We and others have shown that CSF1Ri-induced microglial cell death is caspase-dependent (Elmore et al., 2014; Hagan et al., 2020); however, other mechanisms outside of apoptosis and necroptosis (Bohlen et al., 2017) could also play a contributing role (e.g. protease/autophagy). Importantly, subsequent withdrawal of CSF1Ri elicits GLOBAL repopulation, rather than WM repopulation (Figure 4B–D), indicating that WM repopulation occurs due to the unprecedented level of microglia depletion rather than drug treatment.

WM repopulating myeloid cells derive from an existing Cx3cr1+ cell source originating from the SVZ/WM area

Since repopulating cells first appear in the SVZ and the SVZ is a notable neurogenic/proliferative niche in the brain, we next stained sections containing the SVZ in control, 14 day PLX3397, 5 day and 7 day recovery groups for known precursor cell markers, as well as other cell lineage markers. Between 3 and 5 day recovery, repopulating cells transiently express NESTIN (92% of IBA1 cells were NESTIN+ at 5 day recovery; Figure 5A, Figure 5—figure supplement 1F), MASH1 (81% of IBA1 cells were MASH1+ at 5 day recovery; Figure 5B, Figure 5—figure supplement 1G), and TIE2 (50% of IBA1 cells were TIE2+ at 5 day recovery; Figure 5—figure supplement 1A,H), but are negative for GFAP (0% of IBA1 cells were GFAP+ at any timepoint during recovery; Figure 5—figure supplement 1B), DCX (0% of IBA1 cells were DCX+ during recovery; Figure 5—figure supplement 1C), OLIG2 (0% of IBA1 cells were OLIG2+ during recovery; Figure 5—figure supplement 1D), and SOX2 (0% of IBA1 cells were SOX2+ during recovery; Figure 5—figure supplement 1E) at all timepoints. Consequently, we performed lineage tracing using tamoxifen-inducible Cre-recombinases under control of the Nestin (NestinCreERT2) and Ascl (Ascl1CreERT2, note: Ascl1 encodes for MASH1) promoters, along with the myeloid cell-specific line (Cx3cr1CreERT2). Cre lines were crossed with YFP reporters for visualization of induced expression (Figure 5—figure supplement 2A). Tamoxifen was given immediately following PLX3397 treatment to track the lineage of repopulation cells, except for Cx3cr1CreERT2 mice - which were given a 21 day washout period (i.e. tamoxifen was administered 21 days prior to PLX3397 treatment) to restrict labeling to resident vs short-lived peripheral myeloid cells (Goldmann et al., 2013). These studies revealed that repopulating cells do not originate from Ascl1+ (0% of IBA1 cells were YFP+ during recovery; Figure 5C) or Nestin+ (0% of IBA1 cells were YFP+ during recovery; Figure 5D) cell sources, despite their transient expression of these markers. Consistent with previous reports (Fonseca et al., 2017; Zhao et al., 2019), we found that the Cre-recombinase from the Cx3cr1CreERT2 line is leaky in the absence of tamoxifen (8% of microglia express YFP in control brains and 12% of microglia express YFP during at 7 day recovery; Figure 5—figure supplement 2C–D). Despite this, we show that 94% of microglia in control brains expressed YFP (Figure 5—figure supplement 2C,B), 100% of surviving microglia expressed YFP following 14 days of PLX3397 (Figure 5—figure supplement 2C,B), and 95% of repopulating microglial cells expressed YFP at 7 day recovery (Figure 5—figure supplement 2B), thereby demonstrating that the majority of repopulating cells derive from a Cx3cr1+ cell source (Figure 5E). These repopulating cells exhibit a similar wave-like migration pattern, appearing first near ventricular areas and lastly in cortical regions (Figure 5—figure supplement 2G).

Figure 5 with 2 supplements see all
WM repopulating myeloid cells derive from an existing Cx3cr1+ cell source originating from the SVZ/WM area.

(A–H) Two-month-old WT mice were treated with PLX3397 (600 ppm) for 14 days, then allowed to recover without PLX3397 for 3, 5, and 7 days. (A–B) Representative 63x immunofluorescence images of proliferating (Ki67+, blue) myeloid cells (IBA1+, green) staining for positive for common cell lineage/precursor cell markers: NESTIN (red, A) and MASH1 (red, B) in the SVZ of control, 14 day PLX3397, 3 day recovery, 5 day recovery, and 7 day recovery mice. (C–E) CreER-directed lineage-specific labeling. In these mouse lines, tamoxifen-inducible Cre-recombinase is expressed under control of the promoter of interest. When activated by tamoxifen, the CreER fusion protein translocates to the nucleus allowing transient recombination to occur and, when crossed to a YFP reporter, visualization of induced expression via eYFP. (C–D) Representative 20x images of IBA1+ (red) and associated promoter-driven lineage-derived (YFP, green) cells in control and 7 day recovery Ascl1CreERT2/YFP (B) and NestinCreERT2/YFP (C) mice. (E) Representative 20x images of IBA1+ (red) and Cx3cr1+ lineage derived (YFP, green) cells in control and 7 day recovery mice. (F–H) Representative coronal (F) and sagittal (G–H) brain images of IBA1+ (green) and Ki67+ (red, B–C) cells near SVZ/RMS regions between 5 and 14 days recovery. Inserts provide higher resolution images of cells near the SVZ/RMS proliferative site of repopulation, illustrating the spread of repopulating cells via WM/axonal tracts (i.e. RMS) between the CP and OB and between other WM regions (IC and CeP). CP, caudoputamen; CC, corpus callosum; CTX, cortex; LV, lateral ventricle; RMS, rostral migratory stream; SVZ, subventricular zone; OB, olfactory bulb; IC, internal capsule; CeP, cerebral peduncle.

Figure 5—source data 1

WM repopulating myeloid cells derive from an existing Cx3cr1+ cell source originating from the SVZ/WM area.

https://cdn.elifesciences.org/articles/66738/elife-66738-fig5-data1-v2.xlsx

The role of the SVZ/WM area in myeloid cell proliferation and migration signaling during early WM repopulation

In addition to the site of surviving CSF1Ri-resistant microglia, we next explored what role the SVZ/WM area plays in cell proliferation and the spatio-temporal expansion of repopulating cells during WM repopulation. Immunohistochemical analyses for IBA1 and Ki67 show that repopulating cells initially appear between 3 and 7 days recovery, with IBA1+/Ki67+ cells apparent within the SVZ and the adjacent caudoputamen by 7 days recovery (Figure 5F). Further evaluation of sagittal sections from mice at 5 day recovery confirms that repopulating cells first populate the parenchyma in the SVZ, specifically from a Ki67+-dense region located along the alpha arm of the SVZ (αSVZ) and posterior RMS (pRMS) (Figure 5G). These cells subsequently accumulate in the WM areas adjacent to the SVZ (i.e. the corpus callosum), caudoputamen, RMS, olfactory bulb, internal capsule, and cerebral peduncle before eventually filling the parenchymal grey matter (Figure 5H). The internal capsule connects the cerebral peduncle, caudoputamen, and RMS, while the RMS connects the SVZ to the olfactory bulb, providing anatomical pathways by which repopulating cells travel to specific brain locations. Of note, the aforementioned niches are the precise locations in which microglia initially colonize the developing brain (Ueno et al., 2013; Verney et al., 2010; Ginhoux et al., 2013). During the 14th – 17th week of gestation in the developing human brain, microglia are found near or within: the optic tract, the WM junction between the thalamus and internal capsule, and the junction between the internal capsule and the cerebral peduncle (Menassa and Gomez-Nicola, 2018). Furthermore, a recent study has shown that a second population of amoeboid CX3CR1-expressing microglia emerge from the ventricular zone at embryonic day 20 (E20) and infiltrate the corpus callosum during post-natal day 3–7 (P3-P7) (Nemes-Baran et al., 2020), corresponding to a similar postnatal timepoint and location as PAMs/ATMs identified via scRNA-seq (Hammond et al., 2019; Li et al., 2019).

To examine the transcriptional changes occurring in the SVZ during this early stage of WM repopulation, we micro-dissected the SVZ from control, 14 day PLX3397, and 5 day recovery mice, and performed bulk tissue gene expression analyses via RNA-seq (Figure 6A). Gene expression data can be explored at http://rnaseq.mind.uci.edu/green/alt_repop_svz/gene_search.php. In comparing control vs. 5 day recovery, 227 DEGs were identified (FDR < 0.05) between the two groups (Figure 6B), with the majority being downregulated microglia-enriched/related genes, reflecting the reduced pool of myeloid cells in the CNS during the early stages of repopulation. Upregulated non-myeloid enriched DEGs in depleted vs. 5 day recovery mice (Figure 6B) consisted of genes implicated in cell cycle regulation (Pak3, Swi5, Psmd11, Stat3), DNA transcription/recombination/repair/expression (Alyref2, Swi5, Zfp612, Zfp51, Thumpd1, Prmt5, Taok3, Psmd11, Tox, Stat3), cell adhesion/migration/proliferation (Pak3, Anxa1, Cadm1) and development (Gfap, Rab14, Zfp612). Gene ontology (GO) analysis of DEGs between control and 5d recovery SVZ tissue identified the following top four enriched pathways: myeloid cell differentiation, leukocyte immunity, leukocyte activation, leukocyte chemotaxis and phagocytosis (Figure 6C). Focusing on myeloid genes, P2ry12, Siglech, Trem2, Cd33 and Cx3cr1 were least enriched during initial repopulation, whereas Ccl12Cd52, Lyz2, Itgb2, and Cd84 were highly enriched (Figure 6D). To explore the biological relevance of these findings and the effect on early repopulation dynamics due to a loss in one of these important genes/signals, we administered an antibody against CCL12, the most highly upregulated gene during early WM repopulation (Figure 6E). Here, we show that neutralization of CCL12 results in a significant reduction in repopulating cell numbers at 7 day recovery, but not total distance of cell spreading (Figure 6F–H), indicating that this chemokine may play an important role in early repopulating cell proliferation or survival. Together, these data highlight the role of the SVZ and signaling during early WM repopulation.

The role of the SVZ/WM area in myeloid cell proliferation and migration signaling during early WM repopulation.

(A–D) Transcriptional analysis of SVZ tissue during early WM repopulation. (A) Bulk tissue RNA-seq analysis was performed on micro-dissected SVZ tissue from control, 14 day PLX3397, and 5 day recovery brains (n=5). (B) Heatmap of DEGs between 14 day PLX3397 (Elim) and 5 day recovery SVZ tissue. Gene expression data can be explored at http://rnaseq.mind.uci.edu/green/alt_repop_svz/gene_search.php. (C) Gene ontology chord plot of DEGs between control and 5 day SVZ tissue. (D) Plot highlighting expression (% of control) changes in myeloid-associated genes in depleted (14d PLX3397) and repopulated (5d recovery) SVZ tissue. (E–H) Neutralization of CCL12 during WM repopulation. (E) Experiment schematic of CCL12 neutralization study: 2-month-old WT mice were treated for 14 days with PLX3397 (600 ppm) and then placed on control diet for 7 days allowing for WM repopulation. Four i.p. injections were administered of anti-CCL12 antibody or goat IgG (Isotype control) at 1 day recovery, 3 day recovery, 5 day recovery, and 6 day recovery. (F) Representative whole brain images of IBA1+ cell (green) deposition following treatment. (G–H) Quantification of number of total IBA1+ cells and total distance traveled by IBA1+ wavefront in (F). Total distance was calculated by measuring the length from the ventricular edge of SVZ to the leading edge of the IBA1+ cell wavefront.

Figure 6—source data 1

The role of the SVZ/WM area in myeloid cell proliferation and migration signaling during early WM repopulation.

https://cdn.elifesciences.org/articles/66738/elife-66738-fig6-data1-v2.xlsx

WM repopulating cells do not derive from the periphery

Our data show that extensive microglial depletion results in repopulation of the adult brain from myeloid cells that originate from SVZ/WM areas and utilize WM ‘highways’ to spread throughout the brain before filling the cortex in a distinctive wave-like pattern. As WM repopulating cells maintain distinct phenotypes from microglia, even after extended periods of time in the brain, we concluded that either (1) surviving SVZ/WM myeloid cells represent a distinct myeloid cell type with the capacity to spread and fill the empty microglial brain niche, or (2) extensive microglial depletion stimulates an influx of peripheral cells, which enter near the ventricles and then spread throughout the brain maintaining distinct profiles to their microglial counterparts. Indeed, previous studies have shown that under certain conditions (e.g. in an empty microglial niche) induced by microglial depletion, peripheral myeloid cells can infiltrate and serve as the source for microglial repopulation in the brain parenchyma (Varvel et al., 2012; Lund et al., 2018; Cronk et al., 2018; Paschalis et al., 2018). Thus, we reasoned that repopulation could be occurring from peripheral sources and undertook several complementary experimental approaches to explore this.

The choroid plexus is a major route of cellular entry into the CNS (Ge et al., 2017), thus we explored the contribution of this site to WM repopulation. Here, we observe that choroid plexus myeloid cells do not repopulate until 14 days recovery, despite the appearance of myeloid cells in the adjacent brain parenchyma (Figure 7A–B). Utilizing Cx3cr1GFP/+/Ccr2RFP/+ mice, in which CCR2+ cells (mainly monocytes) express RFP, we show that WM repopulating cells are CCR2/RFP-, or not a result of the infiltration of CCR2+ monocytes (Figure 7—figure supplement 1A–C). A recent study has posited that CSF1R inhibition suppresses CCR2+ monocyte progenitor cells and CX3CR1+ BM-derived macrophages (among other BM populations) and that these populations do not recover after cessation of CSF1R inhibition (Lei et al., 2020). In this study, we evaluated the effects of 14 day PLX3397 600 ppm treatment on peripheral myeloid cells, including CCR2+ and CX3CR1+ myeloid cells, and BM myeloid precursors (Figure 7—figure supplement 1D–F). Here, we observe an expansion, not suppression, of HSCs, common myeloid progenitors (CMPs), and granulocyte/monocyte progenitors (GMPs) following CSF1R inhibition and/or recovery (Figure 7—figure supplement 1F). However, these changes do not result in significant changes to blood or spleen myeloid cell populations (Figure 7—figure supplement 1D–E). CCR2/CCL2 signaling is implicated in many neuropathologies with peripheral cell CNS infiltration (Chu et al., 2014), however, we show that CX3CR1 and CCR2 KO in Cx3cr1GFP/GFP/Ccr2RFP/RFP mice elicits no alterations in WM repopulation (Figure 7C–D) indicating that this form of repopulation is not dependent on these signaling axes. We further explored the CCR2/CCL2 signaling axis using Ccl2-/- mice. Interestingly, we found that a lack of Ccl2 conveyed resistance to CSF1Ri-induced microglial death, with 14 day PLX3397 treatment only eliminating ~90–95% of microglia (Figure 7—figure supplement 1G–H). As a result, GLOBAL repopulation was observed in these mice rather than WM repopulation. In combination with findings from H2K-BCL2 mice, these data confirm that >99% microglial elimination is a requirement for WM repopulation, which may not be achieved by 14 day PLX3397 treatment under certain conditions.

Figure 7 with 1 supplement see all
WM repopulating cells do not derive from the periphery.

(A) Representative immunofluorescence 10x (A) and 63x (B) images of IBA1+ cell deposition within the choroid plexus (labeled with Collagen IV) in control, 14 day PLX3397, 7 day recovery, and 28 day recovery mice. (B) Quantification of IBA1+ cell deposition in the choroid plexus and parenchymal space. (C, D) Sustained microglial depletion and WM repopulation in Cx3cr1GFP/GFP/Ccr2RFP/RFP (i.e., Cx3cr1 and Ccr2 KO). Representative immunofluorescence images (C) and quantification of the number (D) of CX3CR1-GFP+ (green) and CCR2-RFP+ (red) cells in control, 14 day PLX3397, and 7 day recovery mice (n=3–5). (E) Experimental paradigm: Schematic depicting generation of BM GFP+ chimeras, achieved by head-shielded (HS) irradiation and transplantation of donor GFP+ BM cells. After 3 months, mice were treated for 14 days with PLX3397 and then allowed to recover for 7 or 14 days on control diet. (F) FACS gating strategy to determine % chimerism in BM GFP+ chimeras achieved by head-shielded irradiation. (G–J) Representative whole brain images of GFP+ (green) and IBA1+ (red) cell deposition in HS irradiated control (G), 14 day PLX3397 (H), and mice following 7 (I) and 14 day recovery (J). (J1) Higher resolution images of repopulating cell wavefront seen in HS chimeras during WM repopulation. (K) Quantification of number of GFP+ and IBA1+ cells treated HS irradiated mice. % above bar graph indicates %GFP+IBA1+/IBA1+ cells. Data are represented as mean ± SEM (n=3–7). *p < 0.05, ** p < 0.01, *** p < 0.001. CTX, cortex; CP, caudate putamen; LV, lateral ventricle; ChP, choroid plexus; WM, white matter.

Figure 7—source data 1

WM repopulating cells do not derive from the periphery.

https://cdn.elifesciences.org/articles/66738/elife-66738-fig7-data1-v2.xlsx

Next, we utilized bone marrow (BM) chimeric mice to further determine whether repopulating cells originate from a peripheral source (i.e. non-CCR2+ monocytes or other BM-derived cells). Two-month-old wild-type mice underwent head-shielded (HS) irradiation, followed by retro-orbital administration of GFP+ donor BM cells and 12 weeks of recovery for immune system reconstitution (Figure 7E–F). Previous studies have shown that CNS infiltration can occur from the BM upon exposure to head irradiation and consequent BBB permeability (Eglitis and Mezey, 1997; Priller et al., 2001; Mildner et al., 2007). With HS irradiation, however, no GFP+ cells were visible in the parenchyma of control chimeric mice (Figure 7G,K), confirming that under normal conditions circulating peripheral cells do not enter the brain. GFP+ cells were visible in the choroid plexus of control HS chimera (Figure 7G1), consistent with their partial turnover by circulating BM-derived cells (Kierdorf et al., 2019). Fourteen-day PLX3397 eliminated all myeloid cells in HS-irradiated brains (Figure 7H,K). By 7 day recovery, WM repopulation was apparent in HS-irradiated chimeric mice, however, IBA1+ cells were GFP- (Figure 7I–K), thus ruling out peripheral BM-derived cells as the source of this form of microglial repopulation. At 14 day recovery, the wave of repopulating myeloid cells was visible in the cortex, as seen in WT mice (Figure 7J and J1). It should be noted that since parenchymal repopulating IBA1+ cells were GFP- in HS chimeric mice, these data provide strong evidence that BM-derived cells, including choroid plexus macrophages - which are GFP+ in HS chimeric control mice - do not contribute to WM repopulation. Together, these data indicate that surviving SVZ/WM myeloid cells serve as the major source of this unique form of CNS myeloid cell repopulation.

Repopulating myeloid cells are transcriptionally distinct and mount a differential response to inflammatory stimulus compared to homeostatic microglia

We next sought to gain insight into the transcriptional profile of WM repopulating cells once in residence in the brain. In addition, we also investigated whether phenotypic and transcriptional alterations translated into functional consequences and thus explored their response to immune challenge, via LPS administration (Figure 8A). Here, we performed RNA sequencing (RNAseq) on FACS-sorted CD11b+CD45int control and 28 day recovery cells at 6 and 24 hr following LPS-induced immune challenge (Figure 8A, Figure 8—figure supplement 1A). For controls, cells were also collected from mice that did not receive LPS, referred to as 0 hr post LPS. Flow cytometry analysis of CD11b+CD45int cells collected at 28 day recovery shows that these repopulating cells exhibit higher CD45 and CD11b intensities compared to control microglia at baseline. Control microglia show increased CD45 and CD11b in response to LPS, while repopulated cells do not change (Figure 8—figure supplement 1B–C).

Figure 8 with 1 supplement see all
Repopulating myeloid cells are transcriptionally distinct and mount a differential response to inflammatory stimulus compared to homeostatic microglia.

(A) Two-month-old WT control or 28 day recovery mice were given intraperitoneal injections of either PBS or LPS (0.33 mg/kg) and then collected at 6 or 24 hr post injection. Controls, which were mice that did not receive LPS, are referred to as 0 hr post LPS. Myeloid cells were extracted from whole brain hemispheres, isolated using FACS gating for CD11b+CD45int and processed for RNA-seq. (B) Volcano plots displaying the fold change of genes (log2 scale) and their significance (y axis, -log10 scale) between control vs. 28 day recovery mice. (C) Gene ontology chord plot of DEGs between control and 28 day recovery myeloid cells. (D) Heatmap showing expression of genes enriched in DAM, HSC/BM-derived cells, canonical microglia, BAM, PAM, and WAM signatures in control and 28 day repopulating myeloid cells. (E–F) Representative immunofluorescence 20x images of IBA1+ (red) and AXL+ (green, I) or CLEC7A+ (green, J) cells shown in areas with high repopulating cell deposition in control, 7, 14, and 28 day recovery mice. (G) Principal component analysis plot of extracted control and 28 day recovery cells, across time (0 hr, 6 hr, 24 hr) and treatment (+/- LPS), depicting the separation of groups into six clusters. (H) Heatmap of selected time-series cluster analysis of control and 28 day recovery cells. Provided number indicates number of genes per cluster. (I) Time-series cluster analysis of control vs. 28 day recovery myeloid cell response (during WM repopulation) to LPS challenge following 14 day PLX3397 (600 ppm in chow; from H). Clusters showing distinct responses to LPS between control and 28 day WM repopulated cells, across time, were plotted as eigengene values, along with the top represented genes within each cluster. Data are represented as mean ± SEM (n=3–5). *p < 0.05, ** p < 0.01, *** p < 0.001. CP, caudoputamen; CC, corpus callosum; CTX, cortex; LV, lateral ventricle; PirCTX, piriform cortex. Gene expression data can be explored at http://rnaseq.mind.uci.edu/green/alt_repop_lps/gene_search.php.

Figure 8—source data 1

Repopulating myeloid cells are transcriptionally distinct and mount a differential response to inflammatory stimulus compared to homeostatic microglia.

https://cdn.elifesciences.org/articles/66738/elife-66738-fig8-data1-v2.xlsx

RNA was extracted from FACS-sorted CD11b+CD45int cells and RNA-seq analysis was performed to establish a high-resolution transcriptome profile of these cells in the absence and presence of LPS. Gene expression data can be explored at http://rnaseq.mind.uci.edu/green/alt_repop_lps/gene_search.php. Unlike GLOBAL repopulation, in which control and repopulating cells display few transcriptional differences (25 DEGs in GLOBAL vs. control Figure 8—figure supplement 1F–G), we identified 69 DEGs in WM repopulated myeloid cells compared to control microglia in the absence of LPS (logFC > 1, FDR > 0.05; Figure 8B). GO analysis of DEGs between control and 28 day repopulated cells identified the following top five enriched pathways: regulation of cellular component size, negative regulator of chemotaxis, ameboidal-type cell migration, negative regulation of response to external stimulus, and wound healing (Figure 8C). To visualize differences in these cells vs. other myeloid cell subsets, we compared their gene expression profile to previously established myeloid cell signatures, including homeostatic microglia (Butovsky and Weiner, 2018; Bennett et al., 2016), HSC/BM-derived myeloid cells (Weiskopf et al., 2016), border-associated myeloid cells (BAMs) (Mrdjen et al., 2018)/CNS-associated macrophages (CAMs) (Jordão et al., 2019), DAMs/MGnD/ARMs (Keren-Shaul et al., 2017; Krasemann et al., 2017; Mathys et al., 2017; Sala Frigerio et al., 2019), PAMs/ATMs (Masuda et al., 2019; Hammond et al., 2019; Li et al., 2019), and WAMs (Safaiyan et al., 2021). Notably, we observe robust enrichment of DAM and WAM-associated genes in repopulating cells, including Clec7a, Axl, Apoe, Cst7, Ctsd, and Ly9 (Figure 8D). AXL and CLEC7A, recently identified as WAM markers (Safaiyan et al., 2021), immunostaining is apparent in repopulating myeloid cells, particularly in the early stages of repopulation, and in cells located at the wavefront, whereas undetected in microglia from control brains (Figure 8E–F). In addition, we also observe a reduction in expression of Sall1, a transcription factor unique to microglia (Buttgereit et al., 2016), in repopulating compared to homeostatic microglia.

After confirming distinct transcriptome differences between control and 28 day recovery isolated microglia at 0 hr LPS (i.e. in the absence of LPS), we next evaluated their gene expression profiles at 6 hr and 24 hr following LPS challenge. Principal component analysis demonstrated that biological replicates were highly correlated, with samples clustering into six distinct groups (Figure 8G). To identify global patterns in gene expression changes over time between our experimental groups, we employed K-means clustering (Conesa et al., 2006), revealing nine distinct clusters of genes (Figure 8H, Figure 8—figure supplement 1H). Cluster 3 contains genes that are significantly different between control and repopulating cells across all time points, including genes implicated in clearance (Apoe, Atg2a, Axl, Wdr45b), cell growth/differentiation (Cd34, Cdk9), stress (Ier2, Ppp1r15a), inflammation (Ccrl2, H2-K2, Scimp, Tlr3), and senescence (Ubn1, Wdr45b). Clusters 2 (e.g. C4b, Ccl2, Ccl3, Ccl4, Ctsl, Ctss, Il1a, Nlrp3, Pcna, and Tnf), 5, 7 (e.g. Ccl12, Cd63, Csf1, Cxcl13, Il1b, Myc, Plek, and Nfkbib), and 8 shows differential responses to LPS at the 24 hr timepoint between control and repopulating cells (Figure 8I). GO term analysis of Cluster 3 revealed top enriched pathways: phagopore assembly site membrane, extrinsic component of membrane, and endosome. Cluster 8 GO term analysis revealed the following enriched pathways: negative regulation of glucocorticoid secretion, negative regulation of interleukin-1 mediated signaling pathway, and connective tissue replacement involved in inflammatory response wound healing. Together, these findings provide evidence that WM repopulating cells represent a myeloid cell population that are transcriptionally and functionally distinct from adult homeostatic microglia with similarities to the recently identified subset of WAMs, cells absent in the adult (4 months) mouse brain, but comprise 20% of microglia in the aged (24 months) mouse brain (Safaiyan et al., 2021). Further studies are needed to evaluate whether these WM repopulating cells exist in the naïve brain as WAMs or whether CSF1Ri treatment induces this phenotype. Further exploration is also warranted on the contributions and possible cross talk with astrocytes and other cell types during repopulation.

Repopulating myeloid cells elicit few functional differences in behavior, injury, and neuronal-associated structures

Given the transcriptional and functional distinction between WM and homeostatic microglia, we next sought to determine the physiological and functional consequences of filling the adult brain with these cells. Previous studies have shown that replacement of endogenous microglia with GLOBAL repopulating microglia results in no detectable changes in cognitive or behavioral function (Elmore et al., 2018; Rice et al., 2017). Here, we performed a battery of cognitive and behavior tasks in 28 day recovery mice (Figure 9A–H) and observe that these mice exhibit reduced locomotion (Figure 9A–B) and reduced ability to discriminate a novel place change compared to controls (Figure 9H). However, we observe no other significant behavior or cognitive disruptions. Overall, these animals do not appear to exhibit overt cognitive, behavioral, or health changes.

Figure 9 with 1 supplement see all
Repopulating myeloid cells elicit few functional differences in behavior, injury, and neuronal-associated structures.

(A–H) Two-month-old WT mice treated with vehicle or PLX3397 (600 ppm in chow) for 14 days, followed by 28 days of recovery, filling the brain with WM repopulated myeloid cells. Mice underwent behavioral assessment by Open Field, Y-Maze, and Elevated Plus Maze, Social Interaction Test, and Novel Object/Place recognition. In Open Field, distance traveled (A) and average speed (B) were reduced in 28 day recovery groups, but time spent in each zone (C) was unchanged. No changes in performance were seen in the Y-maze, as measured by the number of alternations (D). No changes in anxiety behaviors were seen in the elevated plus maze, as measured by time spent in the open or closed arms (E). No changes in social preference were seen in the social interaction test (F). No changes in novel object recognition memory were seen (G), but 28 day recovery mice had a significant impairment in novel place recognition memory (H). Data are represented as mean ± SEM (n=10). *p < 0.05. (I–L) Two-month-old Cx3cr1-GFP+ mice treated with vehicle or PLX3397 (600 ppm in chow) for 14 days, followed by 35 days of recovery, filling the brain with WM repopulated myeloid cells. (I) Analysis of motility and focal laser response in WM repopulating myeloid cells. Representative images of Cx3cr1-GFP+ myeloid cell response to laser ablation, over time, obtained via two-photon imaging in control and 35 day recovery mice. (J) Quantification of the average normalized GFP+ intensity measured within a 50 µm radius of the site of damage over time. (K) Representative image of Cx3cr1-GFP+ myeloid cell at process motility from 0 min (red) to 2 min (green) time-period. (L) Quantification of process motility (i.e. extension of process µm per min) measuring the difference in visibly moving processes over 2 min in control and 28 day recovery myeloid cells. Data are represented as mean ± SEM (n=4). (M–N) Two-month-old WT mice treated with vehicle or PLX3397 (600 ppm in chow) for 14 days, followed by 28 days of recovery, filling the brain with WM repopulated myeloid cells. (M) Representative immunofluorescence 20x images of Parvalbumin+ (PV, green), WFA (a marker for perineuronal nets, red), and IBA1+ (blue) cells shown in the cortex of control, 14 day PLX3397, and 28 day recovery mice. (N) Quantification of % area of WFA per field of view (FOV) in the cortex. Data are represented as mean ± SEM (n=4–6). ** p < 0.01.

Figure 9—source data 1

Repopulating myeloid cells elicit few functional differences in behavior, injury, and neuronal-associated structures.

https://cdn.elifesciences.org/articles/66738/elife-66738-fig9-data1-v2.xlsx

Based on transcriptional changes in myeloid signaling and priming, we next explored how these repopulating cells would respond to injury. Live two-photon imaging of cortical myeloid cells was performed in control and 35 day recovery Cx3cr1GFP/+/Ccr2RFP/+ mice (Figure 9I–L). Cx3cr1-GFP+ repopulating myeloid cells react to a focal laser injury with similar rates of migration as homeostatic microglia (Figure 9J, Video 1 and Video 2). Furthermore, extension and retraction of processes (i.e. motility) were similar between control microglia and repopulating cells, although repopulating cells displayed fewer and thicker processes as previously demonstrated (Figure 9K–L).

Video 1
Cx3cr1-GFP+ myeloid cell response in control mice after laser ablation.

Representative video of Cx3cr1-GFP+ myeloid cell response to laser ablation (1–5 s long) captured over a 62 min time period in control mice obtained via two-photon imaging. Each frame captures 30 s of elapsed time.

Video 2
Cx3cr1-GFP+ myeloid cell response in 35 day recovery mice after laser ablation.

Representative video of Cx3cr1-GFP+ myeloid cell response to laser ablation (1–5 s long) captured over a 62-min time period in 35 day recovery mice obtained via two-photon imaging. Each frame captures 30 s of elapsed time.

Our recent studies have identified a novel contribution of microglia in modulating perineuronal nets (PNNs) in the adult brain (Crapser et al., 2020b; Crapser et al., 2020a; Liu et al., 2021), specialized extracellular matrix assemblies that enwrap neurons and proximal dendrites to regulate synaptic plasticity (Pizzorusso et al., 2002), protect against neurotoxins (Cabungcal et al., 2013), and enhance signal propagation, among other functions (reviewed in Crapser et al., 2021). In control brains, PNNs (as detected by Wisteria floribunda agglutinin (WFA) staining) are preferentially found on parvalbumin (PV)-expressing cells. Fourteen-day treatment of high-dose PLX3397 results in significant elevation in PNN staining, corroborating previous findings that microglia play a critical homeostatic role in modulating these structures (Crapser et al., 2020b). Following CSF1Ri withdrawal and subsequent 28 day recovery, PNN staining returns to normal levels (Figure 9M–N), indicating that WM repopulating cells share similar PNN-regulating capacities as homeostatic microglia.

Accumulating evidence indicates that WM and developing microglia contribute to myelinogenesis/oligodendrocyte progenitor maintenance (Marsters et al., 2020; Hagemeyer et al., 2017). In early postnatal development, a population of amoeboid microglia migrating from the ventricular zone into the corpus callosum phagocytose oligodendrocyte precursor cells (OPCs) prior to myelination (Nemes-Baran et al., 2020), which share similar timepoint and location-specific distributional overlaps with PAMs/ATMs (Hammond et al., 2019; Li et al., 2019). Thus, we next explored the effects of filling the brain with repopulating cells on WM and WM-associated cells. However, unlike during development in which these cells participate in oligo/myelinogenesis, in 28 day recovery mice we observe no changes in Olig2, a marker for cells of oligodendrocyte lineage (Figure 9—figure supplement 1A,D), or PDGFRα, a marker for oligodendrocyte progenitor cells (Figure 9—figure supplement 1B,E). Given the similarity of WM cells to WAMs, which have been shown to engulf MBP+ particles/myelin debris, we next stained for MBP, a marker for myelin (Figure 9—figure supplement 1C,F–G). Although not statistically significant there appears to be a slight decrease in MBP staining, indicating that these cells could be involved in MBP clearance. Together, these data provide evidence that although WM repopulating cells maintain an altered phenotypic profile, they can still fill the empty microglial niche with few functional consequences.

Lasting phenotypic and transcriptional profiles of WM repopulating cells after three mo recovery

To determine whether the phenotypic and transcriptional differences in WM repopulating microglia are a result of a slower restoration of the microglial phenotype (i.e. temporary) or sustained after longer recovery time (i.e. reflective of a distinct cell subtype), we treated 2-month-old wild-type mice for 14 days with PLX3397 and then allowed for 3 months of recovery (Figure 10A–B). Morphological analysis of IBA1+ cells shows that WM repopulating cells maintain reduced cell density (Figure 10C), and amoeboid-like cell characteristics, including larger cell body size (Figure 10D), reduced number of processes (Figure 10E), and reduced process length (Figure 10F). Bulk tissue RNAseq analysis (Figure 10H) highlights that transcriptional differences remain in the brain three mo following CSF1Ri withdrawal (Figure 10I). Similar to earlier WM repopulation timepoints, DEG analysis highlights a downregulation in homeostatic-associated microglial gene P2ry12 and upregulation in DAM/WAM-associated genes Clec7a (Figure 10I–J). GO term analysis shows that upregulated DEGs are involved in neuron projection development, axon guidance, synaptic transmission, and synaptic potential, providing evidence that WM repopulating cells could play an important role in axonal homeostasis. Together, these data show that WM repopulating microglia represent a distinct population of myeloid cells rather than a slower restoration of the microglial phenotype.

Lasting phenotypic and transcriptional profiles of WM repopulating cells after 3 month recovery.

Two-month-old WT mice treated with vehicle or PLX3397 (600 ppm in chow) for 14 days. Chow was then withdrawn, and animals were allowed to recover on control diet for three months, assessing the long-term effects of filling the brain with WM repopulated myeloid cells. (A-B) Representative whole brain slice and 20x immunofluorescence confocal images of IBA1+ cells in control and 3-month recovery mice. (C-G) Quantification of IBA1+ cells and morphology: IBA1+ cells per field of view (C), cell body area (D), total number of IBA1+ processes (E), average process/filament length (F), and average process/filament diameter (G). (H-L) Transcriptional analysis of 3-month recovery WM repopulation brain tissue. (A) Bulk tissue RNA-seq analysis was performed on whole brain hemispheres from control and 3-month recovery brains. (I) Volcano plot displaying the fold change of DEGs (log2 scale) and their significance (y axis, -log10 scale) between control vs. 3 month recovery. (J) Quantification of relative expression (FPKM, fragments per kilobase of transcript per million) for microglial-associated genes in control and 3 month recovery brain tissue. (K-L) Gene ontology analysis of upregulated (K) and downregulated (L) DEGs between control and 3 month recovery mice. Data are represented as mean ± SEM (n=3-4). * p < 0.05, ** p < 0.01. DG, dentate gyrus; SS CTX, somatosensory cortex; Pir CTX, piriform cortex.

Figure 10—source data 1

Lasting phenotypic and transcriptional profiles of WM repopulating cells after 3-month recovery.

https://cdn.elifesciences.org/articles/66738/elife-66738-fig10-data1-v2.xlsx

Discussion

As central players in CNS homeostasis, defense, and disease, intense focus has recently been placed on microglia and our understanding of their cell origins, function and dynamics. For decades the identity and ontogeny of microglial precursors remained controversial; the scientific community debated whether microglia derive from embryonic progenitors or blood-derived monocytes (Rio-Hortega, 1939; Ginhoux and Prinz, 2015). It is now well-established that microglia arise from yolk sac-derived erythromyeloid progenitors (Ginhoux et al., 2010; Kierdorf et al., 2013).

We previously reported that adult microglia are dependent on CSF1R signaling for their survival and identified several CSF1R inhibitors that eliminate microglia for extended periods of time without peripheral cell infiltration (Elmore et al., 2014). Following CSF1Ri-dependent microglial depletion, we and others have shown that subsequent withdrawal of CSF1R inhibitors from the microglial-depleted brain results in rapid microglial repopulation derived from surviving microglia (Elmore et al., 2014; Zhan et al., 2019; Najafi et al., 2018; Elmore et al., 2018; Rice et al., 2017; O'Neil et al., 2018). The resultant tissue is reconstituted within 14–21 days in a homogenous, tile-like fashion with the replacing microglia fully resembling the original tissue (Elmore et al., 2014; Zhan et al., 2019; Elmore et al., 2015; Elmore et al., 2018). Due to the rapid proliferation of these surviving microglia, exploration into the contribution of specific myeloid cell subtypes to the adult CNS has proven difficult, and so we set out to develop a paradigm without notable surviving microglia.

Previous attempts to achieve an empty microglial niche have fallen short, reporting ~95% or less microglial depletion efficiency (Lund et al., 2018). Here, we utilize sustained high-dose CSF1Ri administration (specifically 14 days of PLX3397 600 ppm) and show we can obtain 99.98% microglial depletion. In doing so, we identify a CNS myeloid cell subset that repopulates the brain parenchyma from SVZ/WM areas, without contributions from the periphery. We describe this form of repopulation as WM repopulation due to its unprecedented level of depletion efficiency and distinct characteristics from global microglial (GLOBAL) repopulation. Unlike GLOBAL repopulation paradigms, in which surviving microglia proliferate in clusters to give rise to new microglia (Bruttger et al., 2015) or more uniformly throughout the brain (Elmore et al., 2014; Zhan et al., 2019), WM repopulation dynamics involve specific spatiotemporal patterns and a dynamic migratory wave of proliferating cells. In WM repopulation, SVZ/WM microglia give rise to the majority of repopulating cells, including cortical microglia; however, this is not a unique property of WM microglia. Under these conditions, sufficient depletion of local microglia favors renewal from more distal cells that proliferate and migrate to fill the niche over local repopulation. Here, and in previous studies we have shown that incomplete (≤99%) microglial elimination leads to repopulation (i.e. GLOBAL repopulation) from all surviving microglia, including cortical microglia (Elmore et al., 2014), suggesting that there is a specific threshold of surviving microglia necessary for local repopulation. In addition, previous studies report a major contribution of peripheral BM-derived myeloid cells to the repopulating cell population following an empty microglial niche or the persistent loss of microglia (in which microglia cannot repopulate the niche) (Lund et al., 2018; Cronk et al., 2018). Since these models rely on tamoxifen administration, it could be possible that BM-derived myeloid cell engraftment in the CNS results from an experimental or technical caveat related to toxin administration rather than the presence of an empty microglial niche. In line with this, a recent study has reported that tamoxifen expands macrophage populations and should be reconsidered as a neutral agonist in myeloid cell lineage studies (Rojo et al., 2018). Recent scrutiny has also been placed on the use of CSF1R inhibitors, implicating long-term changes in BM-derived macrophages (Lei et al., 2020). In this study, we show that high dose and sustained CSF1Ri treatment can result in alterations to monocyte precursor populations; however, these changes do not translate in significant changes to peripheral monocyte populations, which we again show do not contribute to CNS myeloid cell repopulation in the absence of toxin, irradiation or injury.

WM repopulating cells initially appear in specific neuroanatomical niches (first in the SVZ – the site of where surviving CSF1Ri-resistant SVZ/WM myeloid cells reside) and spread throughout the brain in a distinct pattern: via WM tracts to the caudoputamen, optical tract, internal capsule, cerebral peduncle, and finally to cortical areas. The caudoputamen is closely associated with the lateral ventricle, SVZ and WM tracts in which myeloid cells initially appear and migrate, and we believe this spatial association plays a large role in why certain parenchymal areas see more appreciable repopulating cell deposition. Notably, this distribution of initially repopulating cells in ventricular regions and subsequent migration pattern exhibit strong anatomical parallels to microglial colonization and distribution of the embryonic and postnatal brain, in which microglia enter the brain via ventricular routes and remain restricted in WM zones before migrating out to the rest of the brain, with the cortical plate being one of the last areas of colonization (Tay et al., 2017; Ueno et al., 2013; Verney et al., 2010; Ginhoux et al., 2013; Xavier et al., 2014; Lopez-Atalaya et al., 2018). Del Rio-Hortega observed the accumulation of microglia along ventricles and white matter areas in the developing brain within the first postnatal week (Rio-Hortega, 1939; Rio-Hortega, 1932). In 1939, Kershman used the term ‘fountain of microglia’ referring to ‘hot spots’ of activated microglia during human embryogenesis. This observation has been confirmed by several studies over the decades, and recent publications have identified a distinct microglia population that appears postnatally in myelinated regions (Hagemeyer et al., 2017; Benmamar-Badel et al., 2020; Wlodarczyk et al., 2017). A recent study reported that the fountain of microglia (amoeboid microglia migrating from the ventricular zone) phagocytoses OPCs in the corpus callosum during development (Nemes-Baran et al., 2020). It should be noted that we do not believe WM repopulating cells represent a microglial precursor. Most importantly, our findings highlight important anatomical structures that facilitate microglial/myeloid cell migration in an empty microglial niche, which play an essential role in development, but also appear intact in the adult brain.

Utilizing a microglial depletion and repopulation paradigm that successfully achieves an empty microglial niche, we identified a unique subset of myeloid cells in the SVZ/WM area that appears to serve as the source for WM repopulating cells. Notably, our novel paradigm results in the replacement of microglia with Cx3cr1+ myeloid cells originating from the SVZ and associated WM areas, allowing us to study the biology of these cells, and how they adapt to extrinsic environmental cues from grey matter, rather than the WM areas they are normally restricted.

SVZ/WM myeloid cells are initially resistant to CSF1R inhibition, which we believe owes to unique properties of the SVZ environment, as once WM repopulating cells (deriving from SVZ/WM myeloid cells) take up residence in other parenchymal areas, they once again are susceptible to CSF1Ri treatment. Although the signals responsible for shaping macrophage/microglia identity are still being discovered, studies have revealed that the local niche or microenvironment can play an active role in establishing macrophage identity (Lavin et al., 2014; Gosselin et al., 2014). We postulate that a combination of local factors, such as transcription factors, availability of CSF1, and epigenetic regulation (via tissue-specific enhancers) (Blériot et al., 2020), contribute to the uniqueness of SVZ/WM microglia and their ability to escape initial PLX3397 treatment. Studies indicate that WAMs/DAMs upregulate CSF1 and TREM2 (Keren-Shaul et al., 2017; Safaiyan et al., 2021), signaling factors that promote microglial survival (Elmore et al., 2014; Wang et al., 2015; Zheng et al., 2017; Easley-Neal et al., 2019), which may help explain enhanced SVZ/WM microglia survival. It remains unclear whether PLX3397 concentration levels are lower in specific brain regions (i.e. bioavailability of the compound is lower in WM compared to grey matter areas possibly due to differential lipid abundance and compound solubility). However, we believe this is unlikely given the extent of microglial elimination in most WM areas and the close proximity of surviving SVZ/WM microglia to the ventricles (and exposure to higher peripheral doses of the compound).

Previous studies have shown that Csf1- and Csf1r-deficient mice exhibit severe macrophage population deficits, but that not all macrophages are depleted (Rojo et al., 2019; Dai et al., 2002; Cecchini et al., 1994), possibly through contributions from GM-CSF, IL-3 or other important myeloid cell growth factors (Pixley and Stanley, 2004). It was also shown that some macrophages persist in the absence of CSF1R in Csf1r-deficient zebrafish (Kuil et al., 2020). Since 3.5 months treatment with PLX3397 resulted in the eventual loss of SVZ/WM microglia, we hypothesize that this population of SVZ/WM microglia is partially resistant to CSF1R and requires a different threshold for cell death. These cells may rely on other factors for cell survival. A recent study identified a MAC2+ CSF1Ri-resistant microglial population following 14 days of PLX5622 (1200 ppm) treatment, postulating that alterations in of TREM2-TYROBP could explain CSF1Ri resistance (Zhan et al., 2020). This study highlights the existence of a progenitor-like microglial cells that is resistant to CSF1Ri; however, it is important to note that the authors report 88% depletion of IBA1+ cells with similar repopulation dynamics to what we term GLOBAL repopulation. Thus, we do not believe the identified MAC2+ population significantly contributes to WM repopulation.

In addition, several studies have highlighted the existence of microglial cells residing in brain-specific regions with distinct identities and properties (Masuda et al., 2019; Li et al., 2019; Böttcher et al., 2019). WM repopulating cells maintain altered phenotypes and responses to LPS, even after extended periods of time, suggesting that these SVZ/WM-derived cells are highly distinct from other microglia. In line with identifying distinct properties of myeloid cells in the SVZ/WM, previous studies have reported the unique heterogeneity of this microglial subset compared to subsets in other brain regions (Stratoulias et al., 2019Tan et al., 2020). Similar to the morphological and molecular findings in WM repopulating cells, SVZ microglia are less branched (or less ramified), express surface markers that are commonly associated with an alternatively active phenotype (i.e. expressing high levels of anti-inflammatory cytokines IL-4 and IL-10), and lower expression levels of P2RY12 (Ribeiro Xavier et al., 2015). Interestingly, the existence of Cx3cr1+ IBA1- cells in the SVZ and RMS has been reported (Ribeiro Xavier et al., 2015). In humans, microglia in the SVZ exhibit a more activated phenotype that was distinct from all other brain regions, and show higher expression of CD45, CD64, CD68, CX3CR1, EMR1, HLA-DR, indicative of microglial activation, and proliferation markers like Ki67 compared to other subsets (Böttcher et al., 2019). Moreover, WM microglia have been reported to display unique properties during postnatal stages, including an amoeboid morphology and enriched expression of genes related to microglial priming, phagocytosis, and migration (Staszewski and Hagemeyer, 2019). Thus, it appears that WM repopulating myeloid cells fit many of the previously reported characteristics of SVZ/WM microglia.

Once established in the brain, the gene expression profile of the WM repopulating cell shows enhanced enrichment for DAM genes (e.g., Clec7a, Axl, Ly9, Apoe, Itgax). A recent single-cell RNAseq analysis has revealed a WM microglia-specific cluster and specifically a population of white matter-associated microglia (WAMs) that are dependent on TREM2 signaling (Safaiyan et al., 2021). The similarities between WAMs and WM repopulating cells are striking: WAMs are characterized by the downregulation in homeostatic microglial genes (P2ry12, Hexb) and upregulation in DAM-associated genes (Apoe, B2m, Lyz2, and Clec7a), cathepsins, and major histocompatibility complex (MHC) class II-related genes (H2-D1, H2-K1), which are also observed in WM repopulating cells. Droplet-based scRNAseq analysis revealed that these cells are located in the corpus callosum, a WM region with close proximity to the SVZ. WAM signature genes also include Lpl and Itgax, which are also elevated in 28 day WM repopulating cells. Furthermore, WM repopulating cells and WAMs express markers associated phagocytic activity, including CLEC7A and AXL (Safaiyan et al., 2021). An early postnatal phagocytic subset of microglia located in the WM (termed PAMs) has been identified, which also shares a gene signature with DAMs (Li et al., 2019). It appears that WM repopulating cells closely resemble WAMs, however, PAMs/ATMs may represent the developmental equivalent of this cell population. These data reinforce the discovery and existence of a distinct population of microglia in the WM; however further study is needed to examine whether surviving SVZ/WM microglia exist in the naïve brain as WAMs or whether CSF1Ri treatment induces this WAM-like phenotype.

WM is comprised primarily of myelinated axons that connect neurons in different regions of the brain. While WM abnormalities have been detected in a variety of neurological disorders, several microglial-related diseases (adult-onset leukoencephalopathy with axonal spheroids and pigmented glia; ALSP, Nasu-Hakola disease), caused by mutations in microglial-associated genes (e.g. Csf1r, Tyrobp, Trem2), have implicated microglial dysfunction/activation in extensive WM loss and altered cognitive function (Sirkis et al., 2021). Recent evidence shows that WAMs engulf MBP+ particles during disease and facilitate myelin debris clearance (Safaiyan et al., 2021), thus developing ways to manipulate or enhance WM microglial function could prove therapeutically beneficial. Furthermore, since WAM cell numbers increase dramatically with age, identifying their source, properties, and mechanisms of emergence in the brain could prove vital to our understanding of aging and age-related disorders.

In conclusion, this study unveils the presence of a myeloid cell subtype originating from the SVZ and associated WM areas with increased CSF1Ri resistance that yields a dynamic wave of repopulating cells to reconstitute the microglial-depleted brain. These cells exhibit distinct properties compared to homeostatic microglia, sharing similar phenotypic and transcriptional profiles to DAMs and WAMs. Together, these results not only highlight the complexity and diversity of myeloid cells in the adult brain, but establish a model system that provides new insight on myeloid cell homeostasis and dynamics in the brain.

Materials and methods

Key resources table
Reagent type
(species) or
resource
DesignationSource or
reference
IdentifiersAdditional
information
Strain, strain background (M. musculus)B6.129P2(C)-Cx3cr1tm2.1(cre/ERT2)JungJackson LaboratoryIMSR Cat# JAX:020940, RRID:IMSR_JAX:020940
Strain, strain background (M. musculus)STOCK Ascl1tm1.1(Cre/ERT2)Jejo/JJackson LaboratoryIMSR Cat# JAX:012882, RRID:IMSR_JAX:012882
Strain, strain background (M. musculus)C57BL/6-Tg (Nes-cre/ERT2)KEisc/JJackson LaboratoryIMSR Cat# JAX:016261, RRID:IMSR_JAX:016261
Strain, strain background (M. musculus)B6.129X1-Gt(ROSA)26Sortm1(EYFP)Cos/JJackson LaboratoryIMSR Cat# JAX:006148, RRID:IMSR_JAX:006148
Strain, strain background (M. musculus)B6.129(Cg)-Cx3cr1tm1LittCcr2tm2.1Ifc/JernJJackson LaboratoryIMSR Cat# JAX:032127, RRID:IMSR_JAX:032127
Strain, strain background (M. musculus)B6.129S4-Ccl2tm1Rol/JJackson LaboratoryIMSR Cat# JAX:004434, RRID:IMSR_JAX:004434
Strain, strain background (M. musculus)C57BL/6-Tg(CAG-EGFP)131Osb/LeySopJJackson LaboratoryIMSR Cat# JAX:006567, RRID:IMSR_JAX:006567
Strain, strain background (M. musculus)C57BL/6JJackson LaboratoryIMSR Cat# JAX:000664, RRID:IMSR_JAX:000664
AntibodyAnti-Iba1 (Rabbit polyclonal)FUJIFILM Wako ShibayagiCat# 019–19741, RRID:AB_839504IF(1:1000)
AntibodyAnti-Iba1 (Goat polyclonal)AbcamCat# ab5076, RRID:AB_2224402IF(1:1000)
AntibodyAnti-Cd11b (Rat monoclonal)Bio-radCat# MCA711, RRID:AB_321292IF(1:50)
AntibodyAnti-P2RY12 (Rabbit polyclonal)Sigma-AldrichCat# HPA014518, RRID:AB_2669027IF(1:200)
AntibodyAnti-TMEM119 (Rabbit monoclonal)AbcamCat# ab209064, RRID:AB_2800343IF(1:200)
AntibodyAnti-AXL (Goat polyclonal)R and D SystemsCat# AF854, RRID:AB_355663IF(1:100)
AntibodyAnti-Dectin-1 (CLEC7A) (Rat monoclonal)InvivoGenCat# mabg-mdect, RRID:AB_2753143IF(1:30)
AntibodyAnti-Ki67 (Rabbit monoclonal)AbcamCat# ab16667, RRID:AB_302459IF(1:200)
AntibodyAnti-Ccl12 (Goat polyclonal)R and D SystemsCat# AF428, RRID:AB_2070875100 ug/0.5 ml sterile HBSS
AntibodyAnti-Nestin (Mouse monoclonal)AbcamCat# ab6142, RRID:AB_305313IF(1:200)
AntibodyAnti-MASH1 (Mouse monoclonal)BD BiosciencesCat# 556604, RRID:AB_396479IF(1:200)
AntibodyAnti-TIE2 (Mouse monoclonal)AbcamCat# ab24859, RRID:AB_2255983IF(1:100)
AntibodyAnti-GFAP (Chicken polyclonal)AbcamCat# ab4674, RRID:AB_304558IF(1:3000)
AntibodyAnti-doublecortin (DCX) (Goat polyclonal)Santa Cruz BiotechnologyCat# sc-8066, RRID:AB_2088494IF(1:200)
AntibodyAnti-Olig2 (Rabbit monoclonal)AbcamCat# ab109186, RRID:AB_10861310IF(1:200)
AntibodyAnti-SOX2 (Goat polyclonal)R and D SystemsCat# AF2018, RRID:AB_355110IF(1:200)
AntibodyAnti-GFP (Rabbit polyclonal)AbcamCat# ab6556, RRID:AB_305564IF(1:200)
AntibodyAnti-GFP (Chicken polyclonal)AbcamCat# ab13970, RRID:AB_300798IF(1:200)
AntibodyAnti-PU.1 (Rabbit monoclonal)Cell Signaling TechnologyCat# 2258, RRID:AB_2186909IF(1:200)
AntibodyAnti-Collagen IV (Rabbit polyclonal)AbcamCat# ab6586, RRID:AB_305584IF(1:200)
AntibodyAnti-Parvalbumin (Mouse monoclonal)MilliporeCat# MAB1572, RRID:AB_2174013IF(1:200)
AntibodyAnti-Wisteria floribunda lectin (WFA) (Biotinylated)Vector LaboratoriesCat# B-1355, RRID:AB_2336874IF(1:1000)
AntibodyAnti-PDGF Receptor alpha (PDGFRα) (Rabbit polyclonal)AbcamCat# ab124392, RRID:AB_10978090IF(1:200)
AntibodyAnti-Myelin Basic Protein (MBP) (Rat monoclonal)MilliporeCat# MAB386, RRID:AB_94975IF(1:200)
AntibodyAnti-NeuN (Mouse monoclonal)MilliporeCat# MAB377, RRID:AB_2298772IF(1:1000)
Software, algorithmLeica Application Suite X (LASX)LeicaRRID:SCR_013673
Software, algorithmImarisBitplaneRRID:SCR_007370
Software, algorithmImage JImage JRRID:SCR_003070
Software, algorithmFijiFijiRRID:SCR_002285
Software, algorithmWeighted Gene Co-expression Network AnalysisSoftware R packageRRID:SCR_003302
Software, algorithmEdgeRBioconductor software packageRRID:SCR_012802
Software, algorithmDESeqBioconductor software packageRRID:SCR_000154
Software, algorithmLIMMABioconductor software packageRRID:SCR_010943
Software, algorithmGlimmaBioconductor software packageRRID:SCR_017389
Software, algorithmggplot2Software R packageRRID:SCR_014601
Software, algorithmEnhancedVolcanoBioconductor software packageRRID:SCR_018931
Software, algorithmclusterProfilerBioconductor software packageRRID:SCR_016884
Software, algorithmmaSigProBioconductor software packageRRID:SCR_001349
Software, algorithmEnrichrEnrichrRRID:SCR_001575
Software, algorithmEthovision XTNoldusRRID:SCR_000441
Software, algorithmGraphPad PrismGraphPadRRID:SCR_002798

Compounds

PLX3397 was provided by Plexxikon Inc (Berkeley, CA) and formulated in AIN-76A standard chow by Research Diets Inc at the doses indicated in the text. PLX3397 was provided in chow at 600 ppm.

Mice

All mice were obtained from The Jackson Laboratory (Bar Harbor, ME) unless otherwise indicated. Cx3cr1CreERT2 (020940), Ascl1CreERT2 (012882), and NestinCreERT2 (016261) mice were bred to R26-YFP (006148) reporter mice. Cx3cr1GFP/GFPCcr2RFP/RFP (032127) mice were bred to C57BL/6 to obtain Cx3cr1GFP/+Ccr2RFP/+ mice. H2K-BCL-2 transgenic mice were gifted from Irving Weissman. Ccl2 KO (004434) mice were obtained from The Jackson Laboratory. For transplant studies, bone marrow cells were isolated from CAG-EGFP mice (006567). All other mice were male C57BL/6 (000664) mice. Animals were housed with open access to food and water under 12 hr/12 hr light-dark cycles. All mice were aged to 1.5 months unless otherwise indicated.

Animal treatments

Request a detailed protocol

All rodent experiments were performed in accordance with animal protocols approved (AUP-17–179) by the Institutional Animal Care and Use Committee at the University of California, Irvine (UCI). Microglial depletion: Mice were administered ad libitum with PLX3397 at a dosage of 600 ppm (to eliminate microglia) or vehicle (control) for 14 days. To stimulate repopulation, PLX3397 was withdrawn and replaced with vehicle. LPS treatment: Lipopolysaccharide (LPS; Escherichia coli 0111:B4; L4130, Sigma-Aldrich, St. Louis, MO) was dissolved in phosphate-buffered saline (PBS) and administered intraperitoneally (IP) at a dose of 0.33 mg/kg animal body weight either 6 or 24 hr prior to sacrifice. BrdU labeling: Bromodeoxyuridine (BrdU; 000103, Thermo Fisher Scientific, Waltham, MA) was administered at a dose of 1 ml/100 g body weight (per manufacturer’s instructions) for four consecutive days. Mice were sacrificed 24 hr following last BrdU injection. Tamoxifen treatment: Tamoxifen (10540-29-1, Sigma-Aldrich) was suspended in corn oil for 60 min at 50°C. To obtain efficient conversion of loxP alleles a dose of 5 mg tamoxifen/ 25 g animal body weight was delivered orally over five consecutive days. Animals were injected with tamoxifen immediately following PLX3397 inhibition to track the lineage of the repopulating cells (for all lines except Cx3crCreERT2, in which TAM was administered 21 days prior to PLX3397 treatment). In vivo neutralization of CCL12: 100 ug of polyclonal goat anti-CCL12/MCP-5 (AF428, R and D Systems, Minneapolis, MN) or goat IgG in 0.5 ml of sterile HBSS was administered per mouse via i.p. injection at day 1, day 3, day 5, and day 6 (25 ug x 4) recovery (i.e., days post PLX3397 withdrawal). Bone marrow transplant: C57BL/6 mice were anesthetized with isoflurane and then irradiated with 1000 cGy (head-shielded) and reconstituted via retroorbital injection with 2 x 106 whole BM cells from CAG-EGFP mice. Blood was measured 4, 8, and 12 weeks post transplantation to track granulocyte chimerism. At 12 weeks post transplantation, the mice were euthanized and BM was harvested and analyzed by flow cytometry for HSC chimerism. This established an average percent chimerism of >40% in HS irradiated mice. Tissue collection: Following treatments, adult mice were sacrificed via carbon dioxide inhalation and perfused transcardially with 1X PBS. For mouse pups (below the age of postnatal day P9), animals were fully sedated using ice and then decapitated. Brains were extracted and dissected down the midline, with one half flash-frozen for subsequent RNA and protein analyses, and the other half drop-fixed in 4% paraformaldehyde. Fixed brains were cryopreserved in PBS + 0.05% sodium azide + 30% sucrose, frozen, and sectioned at 40 μm on a Leica SM2000 R sliding microtome for subsequent immunohistochemical analyses. Subventricular zone microdissection and isolation: Extracted brains were immersed in ice-cold HBSS (14025092, Thermo Fisher Scientific) and cut in half along the sagittal axis. Following removal of the septum, a thin layer of the rostral and lateral walls of the lateral ventricles were extracted from each hemisphere with a microsurgical stab knife (52–1501, Unique Technologies, Mohnton, PA) and immediately frozen in RNA isolation buffer solution.

Histology and confocal microscopy

Request a detailed protocol

Fluorescent immunolabeling followed a standard indirect technique as described previously (Elmore et al., 2014). Brain sections were stained with antibodies against: IBA1 (1:1000; 019–19741, Wako Chemicals, Richmond, VA; and ab5076, Abcam, Cambridge, UK), CD11b (1:50, MCA711, Bio-Rad Laboratories, Hercules, CA), P2RY12 (1:200, HPA014518, Sigma-Aldrich), TMEM119 (1:200, ab209064, Abcam), AXL (1:100, AF854, R and D Systems), Dectin-1 (also known as CLEC7A; 1:30, mabg-mdect, Invivogen, San Diego, CA), Ki67 (1:200, ab16667, Abcam), NESTIN (1:200, ab6142, Abcam), MASH1 (1:200, 556604, BD Biosciences, San Jose, CA), TIE2 (1:100, ab24859, Abcam), GFAP (1:3000, ab4674, Abcam), DCX (1:200, sc-8066, Santa Cruz, Dallas, TX), OLIG2 (1:200, ab109186, Abcam), SOX2 (1:200, AF2018, R and D Systems), YFP/GFP (1:200, ab6556, Abcam), GFP (1:200, ab13970, Abcam), PU.1 (1:200, 2258, Cell Signaling Technology, Danvers, MA), Collagen IV (1:200, ab6586, Abcam), Parvalbumin (1:200, MAB1572, Millipore, Burlington, MA), WFA (1:1000, B-1355, Vector Laboratories, Burlingame, CA), PDGFRα (1:200, ab124392, Abcam), MBP (1:200, MAB386, Millipore), and NeuN (1:1000, MAB377, Millipore). For DAPI staining, mounted brain sections were cover-slipped using Fluoromount-G with DAPI (00-4959-52, Invitrogen, Carlsbad, CA). High resolution fluorescent images were obtained using a Leica TCS SPE-II confocal microscope (Leica Microsystems, Wetzlar, Germany) and LAS-X v 3.3.0 software. Images in the cortex were taken in the somatosensory cortex unless otherwise indicated. For confocal imaging, one field of view (FOV) per brain region was captured per mouse unless otherwise indicated. To capture brightfield images and whole brain stitches, automated slide scanning was performed using a Zeiss AxioScan.Z1 equipped with a Colibri camera (Zeiss, Oberkochen, Germany) and Zen AxioScan 2.3 software. Microglial morphology was determined using the filaments module in Bitplane Imaris 7.5 (Bitplane, Zurich, Switzerland), as described previously (Elmore et al., 2015). Cell quantities were determined using the spots module in Imaris. Percent coverage measurements were determined in Image J (NIH, Bethesda, MD).

Cranial window implantation

Request a detailed protocol

Mice were anesthetized with isoflurane (Patterson Veterinary, Greeley, CO) in O2 (2% for induction, 1–1.5% for maintenance). To provide perioperative analgesia, minimize inflammation, and prevent cerebral edema, Carprofen (10 mg/kg, s.c., Zoetis, Parsippany-Troy Hills, NJ) and Dexamethasone (4.8 mg/kg, s.c. Phoenix Pharmaceuticals, St. Joseph, MO) were administered immediately following induction. Ringer’s lactate solution (0.2 mL/20 g/hr, s.c, Hospira, Lake Forest, IL) was given throughout the surgery to replace fluid loss. Sterile eye ointment (Rugby Laboratories, Hempstead, NY) was applied to prevent corneal drying. Surgical tools were sterilized using a hot glass bead sterilizer (Germinator 500, CellPoint Scientific, Gaithersburg, MD). Following hair removal, Povidone-iodine (Phoenix) and Lidocaine Hydrochloride Jelly (2%, Akorn, Lake Forest, IL) was used to disinfect and numb the scalp, respectively. The scalp and underlying connective tissue were removed to expose the parietal and interparietal bone. Lidocaine Hydrochloride injectable (2%, Phoenix) was used for muscle analgesia and the right temporal muscle detached from the superior temporal line. The skull was dried using ethanol (70% in DI water) and a thin layer of Vetbond Tissue Adhesive (3M, Saint Paul, MN) applied to the exposed surface. Custom-printed ABS headplates were attached using Contemporary Ortho-Jet liquid and powder (Lang Dental, Wheeling, IL) at an angle parallel to the skull. A small craniotomy (3 mm diameter) was performed over the right hemisphere 2.5 mm anterior and 3 mm lateral lambda. Hemostatic gelfoam sponges (Pfizer, New York, NY) pre-soaked in sterile saline (CareFusion AirLife Modudose, CareFusion/BD, San Diego, CA) were used to absorb dural bleeding. Surgery was terminated if dural tears or intracerebral bleeding was observed. A 4 mm glass coverslip (World Precision Instruments, Sarasota, FL) was placed over the exposed brain and its edges attached to the skull first with a thin layer of Vetbond and second with dental acrylic. Following surgery, mice recovered in their home cage over a warm heating pad until normal behavior resumed (~15–30 min). Postoperative care consisted of daily Carprofen injections (10 mg/kg, s.c.) for 1 week.

Two-photon imaging

Request a detailed protocol

Fluorescence was gathered with a resonant two-photon microscope (Neurolabware, Los Angeles, CA) with 900 nm excitation light (Mai Tai HP, Spectra-Physics, Santa Clara, CA). A 20x water immersion lens (1.0 NA, Olympus, Tokyo, Japan) was used with magnification 4. Emissions were filtered using a 510/84 nm and 607/70 nm BrightLine bandpass filter (Semrock, Rochester, NY). Image sequences were gathered using Scanbox acquisition software (Scanbox, Los Angeles, CA) at a depth of 200–260 µm below the meninges. An electrically tunable lens was used to image 20 planes (326x325µm, 3 µm z step), each sampled at 0.5 Hz. Laser damage consisted of line scanning at magnification 25 for 1–10 s at 800 nm.

Quantification of homeostatic motility and response to laser ablation

Request a detailed protocol

All image stacks were processed and analyzed using the image processing package FIJI, a distribution of NIH Image J software (Schindelin et al., 2012). Stacks were temporally binned by taking the sum for each pixel over 30 frames (1 min.). Homeostatic motility was quantified manually by measuring the difference of visibly moving processes over 2 min. For each mouse, 10 microglia were chosen at random and the first five extension or retraction observed were recorded for a total of 50 observations per mouse. Microglia respond to laser damage by extending their processes toward the site of injury. We took advantage of increasing fluorescence at the site of damage from infiltrating GFP-positive processes to quantify microglia response to laser ablation. The average GFP intensity within a circle (r = 50 µm) centered at site of damage at any timepoint (tx) was normalized to the intensity in that area at 1 min. post ablation (t1). To determine differences in GFP intensity within groups over time and between groups at any timepoint, we used a repeated measures two-way ANOVA corrected for multiple comparison (Geisser-greenhouse correction).

PK analysis

Request a detailed protocol

PLX3397 concentration in plasma and cerebellum were analyzed for pharmacokinetic (PK) data by Integrated Analytical Solutions, (Inc).

FACS analysis

Request a detailed protocol

Myeloid cells were extracted from whole hemispheres, isolated into single-cell suspensions and identified using fluorescence-activated cell sorting (FACS) gating for CD11b+CD45int as previously described (Elmore et al., 2014). Cells were stained with the following surface antibodies purchased from Biolegend (San Deigo, CA) at 1:200 unless otherwise indicated: CD34-eFlour660 (1:50, 50-0341-80, eBioscience, San Diego, CA), Sca-1-AF700 (1:100, 108141), CD16/32-PE (101307), Ter119-PE/Cy5 (116209), ckit/CD117-PE/Cy7 (25-1171-81, eBioscience), CD150/SLAM-PerCP-eFlour710 (46-1502-82, eBioscience), CD11b-APC (101212), CD11b-PE (101208), Gr1-AF700 (108422), CD45-AF700 (103128), CD45-APC/Cy7 (103116), NK1.1-PE (108707), CD3-PE/Cy7 (100220), CD19-Per-Cyanine5.5 (45-0193-82, eBioscience), CD11c-APC/Cy7 (117323), Ly6C-PE (1:400, 128007), Ly6G-5.5 (127615). For HSCs, CMPs, and GMPs, all cells were gated on live (PI-), Ter119- cells and then identified with the following gating strategy: HSCs: FcyR-, ckit+ Sca+ CD34-, SLAM+, CMPs: FcyR-, ckit+, Sca-, CD35+, and GMPs: FcyR+, ckit+, Sca-, CD34+. Samples were acquired with the BD LSRII or BD Fortessa X20, and sorted with the BD FACS Aria II.

Nanostring analysis

Request a detailed protocol

RNA was extracted and purified from frozen half brains using an RNA Plus Universal Mini Kit (Cat. #73404, Qiagen). For nCounter analysis, total RNA was diluted to 20 ng/μl and probed using a mouse nCounter PanCancer Immune Profiling Panel (Nanostring Technologies, Seattle, WA, USA) profiling ~700 immunology-related mouse genes. Counts for target genes were normalized to the best fitting house-keeping genes as determined by nSolver software. The WGCNA package was used to evaluate the quality of reads, as well, as identify and remove appropriate outliers, based on standard deviation within normalized expression values. PCA plots were generated using plot3D. Negative binomial linear regression analysis was performed using EdgeR, DESeq, and Limma packages to generate FDR and log fold change values. Top significant genes are displayed as a volcano plot constructed using GLimma, ggplot2, and EnhancedVolcano.

RNA-sequencing and analysis

Total RNAs were extracted by using RNeasy Mini Kit (Qiagen, Hilden, Germany). RNA integrity number (RIN) was measured by Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA) and samples with RIN >= 7.0 were kept for cDNA synthesis. cDNA synthesis and amplification were performed followed by Smart-seq2 (Picelli et al., 2014) standard protocol. Libraries were constructed by using the Nextera DNA Sample Preparation Kit (Illumina, San Diego, CA). Libraries were base-pair selected based on Agilent 2100 Bioanalyzer profiles and normalized as determined by KAPA Library Quantification Kit (Illumina). The libraries were sequenced using paired-end 43 bp mode on Illumina NextSeq500 platform with around 10 million reads per sample. Read alignment and expression quantification: Pair-end RNA-seq reads were aligned using STAR v.2.5.1b with the options (--outFilterMismatchNmax 10 --outFilterMismatchNoverReadLmax 1 --outFilterMultimapNmax 10) (Dobin et al., 2013). Rsubread was used to generate feature counts (Liao et al., 2019). Gene expression was measured using Limma, edgeR, and org.Mm.eg.db packages with expression values normalized as RPKM (Robinson et al., 2010; McCarthy et al., 2012; Carlson, 2018; Ritchie et al., 2015).

Differential expression analysis

Request a detailed protocol

Libraries with uniquely mapping percentages higher than 80% were considered to be of good quality and kept for downstream analysis. Protein coding and long non-coding RNA genes, with expression RPKM > = one in at least three samples, were collected for subsequent analysis. Differential expression analysis was performed by using Limma, edgeR, and org.Mm.eg.db (Robinson et al., 2010; McCarthy et al., 2012; Carlson, 2018; Ritchie et al., 2015). Differentially expressed genes (DEGs) were selected by using false discovery rate (FDR)<0.05. Top significant genes are displayed as a volcano plot constructed using GLimma, ggplot2, and EnhancedVolcano (FDR < 0.05, LogFC >1) (Blighe, 2019). PCA plots were generated using plot3D (Soetaert, 2017).

Gene ontology and pathway analysis

Request a detailed protocol

DEGs were analyzed for Gene ontology (GO) enrichment by clusterProfiler using a hypergeometric test with corrected p-values < 0.05 (Yu et al., 2012). These results were then plotted with GOplot. maSigPro package was used to identify genes that show different gene expression profiles over time (Conesa et al., 2006). Heatmaps were generated by mapping RPKM values to genes identified in maSigPro and then constructed using gplots. Normalized (min max normalization for each individual gene) log2-transformed expression values are displayed as a heatmap with hierarchical clustering utilizing gplots. maSigPro-selected gene clusters were identified and enriched for GO using clusterProfiler. GOplot was used to correlate genes and important pathways. DEGs of 3 mo recovery mice were analyzed for Gene ontology (GO) enrichment by enrichR with corrected p-values < 0.05.

Behavioral and cognitive analysis

Mouse behavior, motor function, and cognition was evaluated using the following tasks: elevated plus maze, open field, novel place/novel object, sociability test, and spontaneous alternation Y-maze in the order listed, and as previously described unless otherwise indicated (Elmore et al., 2014; Elmore et al., 2015; Spangenberg et al., 2019Spangenberg et al., 2016). Testing was conducted at 28 days recovery (i.e. after CSF1Ri removal and microglial repopulation).

Sociability test

Request a detailed protocol

The Crawley’s or Three-Chamber Sociability test assesses general sociability, or time spent with another rodent. In brief, animals were placed in a Three-chamber Sociability Test box (19 cm x 45 cm) with two dividing walls made of clear Plexiglas allowing free access to each chamber. During habituation, the subject mouse is placed in the middle chamber for 5 min for adaption. During testing (24 hr after habituation), a stranger mouse (inside a wire containment cup) is placed in one of the side chambers, and the subject mouse was placed in the center chamber and allowed to access and explore all three chambers for 10 min. The placement of the stranger mouse in the left and right chambers is systemically altered between trials. The duration of time spent in each chamber, velocity, and distance traveled was measured. Spontaneous Alternation Y-Maze: For this task, mice were placed in a Y-maze (35.2 cm arm length x 5 cm width x 20 cm sidewall height). Each animal was allowed to freely explore the arena for 8 min. Distinct intra-maze visual cues were positioned at the end of each arm for spatial orientation. Spontaneous alternation, which measures the willingness of an animal to explore new environments, was measured by the number of triads, or entry of all three arms in a consecutive sequence (i.e. ABC and not BAB). Unless otherwise indicated, behavioral readouts for all tasks were calculated from video using the EthoVision XT 14 tracking system (Noldus, Leesburg, VA).

Data analysis and statistics

Request a detailed protocol

Statistical analysis was performed with Prism Graph Pad (v.8.1.1, GraphPad Software, San Diego, CA). To compare two groups, the unpaired Student’s t-test was used. To compare multiple groups, a one-way ANOVA with Tukey’s posthoc test was performed. For all analyses, statistical significance was accepted at p < 0.05. All bar graphs are represented as means +/- SEM and significance expressed as follows: *p < 0.05, **p< 0.01, ***p < 0.001. n is given as the number of mice within each group.

Data availability

Sequencing data have been deposited in GEO under accession code GSE166092, and can be explored in an interactive fashion at http://rnaseq.mind.uci.edu/green/. All other data generated or analysed during this study are included in the manuscript and support files.

The following data sets were generated
    1. Green KN
    2. Hohsfield LA
    3. Soni N
    (2021) NCBI Gene Expression Omnibus
    ID GSE166092. Subventricular zone/white matter microglia reconstitute the empty adult microglial niche in a dynamic wave.

References

    1. Rio-Hortega P
    (1932)
    Cytology and cellular pathology of the nervous system
    Canadian Medical Association Journal 27:576.

Decision letter

  1. Jaime Grutzendler
    Reviewing Editor; Yale University, United States
  2. Carla V Rothlin
    Senior Editor; Yale School of Medicine, United States
  3. Jaime Grutzendler
    Reviewer; Yale University, United States

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

This is a comprehensive cellular and molecular investigation of the repopulation dynamics of microglia following extensive elimination through administration of the CSF1r antagonist PLX3397. The investigators describe an intriguing capacity for a small subpopulation of white matter and subventricular zone surviving microglia to repopulate the entire brain. They provide extensive characterization of the "wave" pattern of migration and proliferation during repopulation and provide important information about the transcriptional profile of these cells. This study has important implications for our understanding of microglia development and repopulaiton dynamics as well as implications for therapeutics with CSF1r antagonists.

Decision letter after peer review:

Thank you for submitting your article "Subventricular zone/white matter microglia reconstitute the empty adult microglial niche in a dynamic wave" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, including Jaime Grutzendler as the Reviewing Editor and Reviewer #1, and the evaluation has been overseen by Carla Rothlin as the Senior Editor.

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

These are the main issues to address but please look at the individual reviews for further details that should also be considered:

1) Since this paper is entirely dependent on the idea that there is a subpopulation of microglia cells originating from SVZ, that repopulates the cortex after PLX3397 treatment, it will be critical to show better quality high- resolution imaging through confocal tiling spanning large areas of the cortex and white matter. Stain and show several markers of microglia to more convincingly demonstrate that there are no surviving microglia in the cortex that would explain local repopulation. Related to this, In lines 164 for Figure 2C, the authors claim that microglial markers are lost rather than downregulated. But the images have faint (green) signals that seem to be more indicative of downregulation rather than loss. Please explain this.

2) On the basis of the appearance of microglia at different locations after 14 days of PLX3397 treatment, the authors claim that the repopulating cells elicit a dynamic wave migrating radially and tangentially from the SVZ/ ventricular zone, white matter tract, and caudoputamen. To establish this observation more convincingly and to rule out the possibility of sparse local repopulation specifically in the cortex, as some cells can be seen in figure 3 B (stage III) in the cortex, the authors need to present a more detailed temporal and spatial resolution of the "Wavefront". More detailed imaging is required using high-resolution confocal tiling capturing different time points between stage III and IV to show the gradual emergence of the wavefront from the white matter tract and invading the cortex. A similar extensive analysis of the wavefront in the lineage tracing using cx3cr1-creER should also be performed.

3) The authors make several quantitative pronouncements without adequate quantitative data that is needed. Eg., while co-expression of certain proteins and myeloid cells (or the lack thereof) are mentioned (lines 247-249), quantitative assessments are required i.e. what percent of IBA1+ cells are NESTIN+ or MASH+ etc? Similarly, the authors make claims of ~25% or ~100% labeling in (lines 260-261) but precise numbers should be provided. It is clear to this reviewer that at least for Figure 4E, there are red (IBA1+ cells) that are not CX3CR1+. Moreover, for the various "stages" in Figure 3B-D, there is no indication of time (days) following PLX withdrawal which would help substantiate temporal claims. Data in Figure 6A-B are just images and would be helped by some quantification of the number of cells. Similarly, data in Figure 6G-K can be presented as percent of GFP cells amongst IBA1 cells as discussed in lines 361-367 and 371. In lines 394-395 while the number of DEG genes is specified for END repopulated cells compared to control microglia (65), the number is not stated for PND repopulated cells. The authors should provide the exact number rather than stating it as a "few" genes.

4) Do the authors have any sense of what is unique about the surviving cells in the SVZ, why do they escape PLX3397 treatment? If they are simply cx3cr1 positive myeloid cells, the transcriptome data provided only answers how these cells differ once they start repopulation the brain, however it does not answer how they were unique prior to PLX3397 treatment that made them survive. Additional data characterizing this would be important to further determine the relevance of this finding. This issue was also raised by Reviwer#3 (1a).

5) The issue of pharmacokinetic analysis of PLX3397 is important. One possibility is that the cells in the white matter rather than being more resistant to PLX3397 receive a relatively lower concentration of the drug than in the cortex due to either reduced perfusion of these areas or alternatively since PLX3397 is lipophilic, it could be incorporated in the myelin sheets thereby reducing its effective concentration. Ideally, the methods used to measure PLX3397 brain concentration should be applied separately to cortical and white matter tissues. What happens if the concentration of PLX3397 used is higher? Is the white matter wavefront still formed?

Perhaps complete elimination of myeloid cells would be achieved with a longer PLLX treatment paradigm. The authors should consider an up to 2-3 month treatment protocol. In addition, the mention of "sustained" in line 196 should be replaced with the specific duration i.e. 14 days to remove any potential confusion that the authors could mean longer that 14 days.

6) After 28 days of recovery period the authors observed that the repopulating cells after PLX3397 treatment were transcriptionally different from the control population. It would be important to know whether these transcriptional differences in the empty niche repopulating microglia are temporary or are sustained even after a longer recovery time. It is possible that these cells show a *delayed* transition to more control microglial-like features which may be detected at longer than 28 days. Looking at up to 3 months after reconstitution would help to clarify whether the differences between reconstituted cells and control microglia are indeed not due to a slower restoration in the microglial phenotype.

7) In the AD mouse model, the authors administered PLX3397 and analyzed the effect on amyloid plaque deposition after 3 months of recovery (Figure 9). On the basis of this experiment, the authors claim that the repopulating cells could reduce the amyloid load. However, the significance of this experiment is unclear and does not fit with the central theme of the paper. There is a complete lack of controls addressing amyloid synthesis, processing, secretion, and clearance before and after PLX 3397 treatment. Therefore, the data provided is superficial and not sufficient to demonstrate such a causal relationship. This part should be completely revamped or eliminated from the study.

Also, it is difficult to draw appropriate conclusions from the 5XFAD studies provided in Figure 9E-P because they lack the 14d PLX treatment results. This is needed to determine whether some of the features are unchanged between control and repopulated 5XFAD results or else the 14d depletion is could have itself changed things and then the repopulation would restore the change. Therefore, the 14d PLX results is critically needed data.

8) The relative resistance of the microglia in the SVZ and adjacent WM track could very well be due to the local cues/environment, which alter the signaling and lead to relative resistance. While interesting from the experimental perspective, it might not be physiologically relevant for how endogenous microglia repopulate the brain.

9) The comparison of control microglia vs. 28-day repopulated microglia in response to LPS is not particularly informative because of the artificial nature of the timing and treatment. It could also be due to the cross talk with astrocytes and other cell types, which are also altered under the 2-week 3397 treatment.

(10) The H2k-BCL2 experiment (Figure S1B-C ) is potentially interesting. Why put it in supplementary? It would be best to move to main figures and add proper quantification. The H2K-BCL2 studies are a bit confusing. First, the statement that these mice provide "overexpression of BCL2 in myeloid lineages affords some resistant to CSF1Ri-induced cell death" is not substantiated by published (or new) data. In addition, no evidence is provided that microglia indeed show over expression of BCL2. Third, the mechanism of BCL2 protection from cell death is not mentioned (presumably through reduction in apoptosis?) Fourth, in the initial study by this group (Elmore et al., 2014), they suggested that CSF1Ri-induced cell death using PLX3397 occurs by apoptosis stating "To confirm that microglia undergo cell death with CSF1R inhibition, we found that many microglia stained for active caspase-3 in the same animals, a classic marker of apoptosing cells (Figure S2J- S2L)" on page 381. Since blocking apoptosis with BCL2 overexpression only increases survival from 0.2-5-10%, do the authors want to expand on whether cell death primarily occurs by apoptosis as they suggested in a previous study?

Reviewer #1:

In this article, Hohsfield et al. developed a method to achieve complete elimination of microglia in the adult brain using a sustained high dose of PLX3397 for 14 days. On the basis of immunofluorescence staining of microglia and different transgenic lines, the authors show that the repopulating myeloid cells are cx3cr1 positive and originate from the subventricular zone and white matter tract. The extensive CSF1R inhibition unveils the presence of inhibitor-resistant myeloid cells in the subventricular zone/white matter areas. The authors show that the repopulating cells were CCR2-RFP negative, eliminating the possibility of peripheral immune cell infiltration. Furthermore, the repopulating microglia are transcriptionally distinct from the control microglia and were enriched with disease-associated microglia genes, like AXL, ApoE, Clec7a, and cst7. To understand the functional impact of these empty niche repopulating microglia on mice, the authors performed various behavioral tests. Also, in AD-like mice, the authors show that administration of PLX3397 for 14days could reduce the amyloid load in the cortex and amygdala.

Overall, this study is comprehensive and depicts a potentially interesting microglial repopulation dynamic under extensive microglial depletion. The potential relevance of this information is that these method for microglia ablation is widely used and that the repopulation mechanism could have relevance for understanding mechanisms of microglia development. If PLX3397 or similar drugs are ever approved , it is important to know how they affect the brain. However, the major claims of this paper, the dynamic wave of microglia repopulating the whole brain and complete elimination after a sustained high dose of PLX3397, need more extensive and rigorous analysis. Following are the major concerns we have with this paper.

1. Since this paper is entirely dependent on the idea that there is a subpopulation of microglia cells originating from SVZ, that repopulates the cortex after PLX3397 treatment, it will be critical to show better quality high- resolution imaging through confocal tiling spanning large areas of the cortex and white matter. Stain and show several markers of microglia to more convincingly demonstrate that there are no surviving microglia in the cortex that would explain local repopulation.

2. On the basis of the appearance of microglia at different locations after 14 days of PLX3397 treatment, the authors claim that the repopulating cells elicit a dynamic wave migrating radially and tangentially from the SVZ/ ventricular zone, white matter tract, and caudoputamen. To more convincingly establish this observation and to rule out the possibility of sparse local repopulation specifically in the cortex, as some cells can be seen in figure 3 B (stage III) in the cortex, the authors need to present a more detailed temporal and spatial resolution of the "Wavefront". More detailed imaging is required using high-resolution confocal tiling capturing different time points between stage III and IV to show the gradual emergence of the wavefront from the white matter tract and invading the cortex. A similar extensive analysis of the wavefront in the lineage tracing using cx3cr1-creER should also be performed.

3. The issue of pharmacokinetic analysis of PLX3397 is important. One possibility is that the cells in the white matter rather than being more resistant to PLX3397 receive a relatively lower concentration of the drug than in the cortex due to either reduced perfusion of these areas or alternatively since PLX3397 is lipophilic, it could be incorporated in the myelin sheets thereby reducing its effective concentration. Ideally, the methods used to measure PLX3397 brain concentration should be applied separately to cortical and white matter tissues. What happens if the concentration of PLX3397 used is higher? Is the white matter wavefront still formed?

4. Do the authors have any sense of what is unique about the surviving cells in the SVZ, why do they escape PLX3397 treatment? If they are simply cx3cr1 positive myeloid cells, the transcriptome data provided only answers how these cells differ once they start repopulation the brain, however it does not answer how they were unique prior to PLX3397 treatment that made them survive. Additional data characterizing this would be important to further determine the relevance of this finding.

5. After 28 days of recovery period the authors observed that the repopulating cells after PLX3397 treatment were transcriptionally different from the control population. It would be interesting to know whether these transcriptional differences in the empty niche repopulating microglia are temporary or are sustained even after a longer recovery time.

6. In the AD mouse model, the authors administered PLX3397 and analyzed the effect on amyloid plaque deposition after 3 months of recovery (Figure 9). On the basis of this experiment, the authors claim that the repopulating cells could reduce the amyloid load. However, the significance of this experiment is unclear and does not fit with the central theme of the paper. There is a complete lack of controls addressing amyloid synthesis, processing, secretion, and clearance before and after PLX 3397 treatment. Therefore, the data provided is superficial and not sufficient to demonstrate such a causal relationship.

7. The H2k-BCL2 experiment (Figure S1B-C ) is potentially interesting. Why put it in supplementary? It would be best to move to main figures and add proper quantification.

Reviewer #2:

The manuscript by Hohsfield et al. used high dose of CSF1R/FLT/Kit inhibitor to diminish brain microglia to less than 0.1%, which mostly located in the SVZ and adjacent white matter, called END. They then described their molecular characteristics and the repopulation routes. There are interesting observations with broad description of the characteristics and the resistant population and the population routes. The study also confirmed some of the previous findings, such as the brain-origin of the repopulation. They also showed even after 28 days, the repopulated microglia from END exhibit differential responses to LPS.

The major concern of the study is that the findings failed to lead to deeper understanding of microglial homeostasis regulation in health and in disease in its current form.

1) The relative resistance of the microglia in the SVZ and adjacent WM track could very well be due to the local cues/environment, which alter the signaling and lead to relative resistance. While interesting from the experimental perspective, it might not be physiologically relevant for how endogenous microglia repopulate the brain.

2) Some of the conclusions are confirmation of previous studies, such as the origin of END population.

3) The comparison of control microglia vs. 28-day repopulated microglia in response to LPS is not particularly informative because of the artificial nature of the timing and treatment. It could also be due to the cross talk with astrocytes and other cell types, which are also altered under the 2-week 3397 treatment.

Reviewer #3:

In this study by Hohsfield et al., the authors set out to develop a complete elimination of microglia in the brain followed by its reconstitution. Using a high dose (600ppm) PLX3397 14 day treatment, they show the most effective depletion strategy for IBA1+-brain cells at 99.98%. This is indeed the most extensive microglial depletion strategy to date. They refer to this as a microglial "empty niche" environment and then proceed to determine the reconstitution mechanisms. Interestingly, unlike previous paradigms (referred to as a partially depleted niche), where microglia are either reconstituted by proliferation of residual microglia or infiltration of peripheral myeloid cells, in this empty niche scenery, reconstitution is slower and only partial by a month. Furthermore, reconstitution follows defined routes similar to that previously reported for developmental colonization of myeloid cells that become microglia. However, even after a month of populating the brain, these cells remain phenotypic and functionally distinct from endogenous microglia. remarkable, this paradigm when applied to AD its shown to reduce amyloid plaques burden. These are interesting findings, many of which are novel. However, some of the conclusions are derived from insufficiently substantiated data that lack some appropriate controls and enough quantification and so can be improved by addressing these concerns for an otherwise interesting subject matter.

This is manuscript is an attempt to show a novel CSF1Ri-resistent myeloid cell-type that originates in the subventricular zone/white matter and reconstitutes an empty microglial niche following near total CSF1R-dependent microglial elimination. The authors claim that reconstituting myeloid cells populate the empty microglia niche in the brain through migratory routes akin to that employed by endogenous microglia during development. These myeloid cells show a distinct profile from homeostatic microglia and maintain a distinct profile following extended residence in the brain. Finally, in a model of AD, reconstituting the brain with this approach and these cells resulted in a lower AD load that suggests this could be a promising approach to plaque burden in AD. Despite these extremely exciting findings, the manuscript in its current form suffers from significant concerns that need to be addressed before it can be suitable for publication at eLife.

1. Overstated or insufficiently substantiated claims: several details in the manuscript, while interesting overstate the evidence provided by the data or are not substantiated by the current data.

a. Figure 3 seeks to provide spatio-temporal evidence for the "dynamic wave" of repopulation. however, while these findings are interesting, they seem to be based of mainly spatial (without temporal) data taken "between 7 and 14" days of PLX withdrawal. While the impression is given that the authors assess the myeloid cell migration at different timepoints between these time periods, the authors seem to make these conclusions and claims solely from static images at (rather than between) 7 and 14 days. To substantiate 'temporal" claims, the authors need to provide analysis at different time points between 7 and 14 days which they don't seem to have done.

b. The authors make several quantitative pronouncements without adequate quantitative data that is needed. Eg., while co-expression of certain proteins and myeloid cells (or the lack thereof) are mentioned (lines 247-249), quantitative assessments are required i.e. what percent of IBA1+ cells are NESTIN+ or MASH+ etc? Similarly, the authors make claims of ~25% or ~100% labeling in (lines 260-261) but precise numbers should be provided. It is clear to this reviewer that at least for Figure 4E, there are red (IBA1+ cells) that are not CX3CR1+. Moreover, for the various "stages" in Figure 3B-D, there is no indication of time (days) following PLX withdrawal which would help substantiate temporal claims. Data in Figure 6A-B are just images and would be helped by some quantification of the number of cells. Similarly, data in Figure 6G-K can be presented as percent of GFP cells amongst IBA1 cells as discussed in lines 361-367 and 371. In lines 394-395 while the number of DEG genes is specified for END repopulated cells compared to control microglia (65), the number is not stated for PND repopulated cells. The authors should provide the exact number rather than stating it as a "few" genes.

c. A major claim that permeates the manuscript is that the residual myeloid cells are CSF1Ri-resistant. However, this is based upon a 14-day treatment regimen. Perhaps complete elimination of myeloid cells would be achieved with a longer PLLX treatment paradigm. The authors should consider an up to 2-3 month treatment protocol. In addition, the mention of "sustained" in line 196 should be replaced with the specific duration i.e. 14 days to remove any potential confusion that the authors could mean longer that 14 days.

d. Similarly, the density phenotypes (Figure 1F) and the morphology phenotypes (Figure 1 J-L) were only assessed up to 28 days following PLX withdrawal. It is possible that these cells show a *delayed* transition to more control microglial-like features which may be detected at longer than 28 days. This reviewer recommends looking at up to 3 months after reconstitution (which the authors did with the 5XFAD data in Figure 9). This would help to clarify whether the differences between reconstituted cells and control microglia are indeed not due to a slower restoration in the microglial phenotype.

e. In lines 164 for Figure 2C, the authors claim that microglial markers are lost rather than downregulated. But the images have faint (green) signals that seem to be more indicative of downregulation rather than loss. Please explain this.

f. Lines 150-151 suggest that repopulating cells display larger cell bodies with no indication of a recovery to normal levels despite the fact that the corresponding Figure shows this in Figure 1J.

g. For data in Figure 3E, the cells are referred to as "proliferating END repopulating myeloid cells" (line 233) but there is no clear evidence of colocalization of Ki67 and IBA1.

h. The authors state that the myeloid cells they identify are CSF1Ri-resistant after 14 days of PLX3397 (600ppm) treatment. But to ensure that they are indeed truly resistant, a longer (and "sustained"-line 1960) exposure of perhaps up to 2 months would be needed.

i. In lines 308 – 311, the authors claim that since CCL12 antibodies reduced populating cell numbers but not the total distance of spreading. However, it is not clear what "the total distance of cell spreading" is. In the images, the territory occupied by the green cells (myeloid cells) seems to be smaller in antibody treated tissues. Moreover, although the authors claim that this combined observation leads to the conclusion that CCL12 may play a role in the proliferation of repopulating cells, this is not the only possibility. For example, the antibody could affect the survival of cells rather than the proliferation. The authors could detect proliferation using markers in control and CCL12 antibody treatment.

j. In line 340-341, the authors claim the CMPs are expanded by flow but they do not provide the gating strategy here. This will help to compare to the cells assessed in the other study.

k. Finally, it is difficult to draw appropriate conclusions from the 5XFAD studies provided in Figure 9E-P because they lack the 14d PLX treatment results. This is needed to determine whether some of the features are unchanged between control and repopulated 5XFAD results or else the 14d depletion is could have itself changed things and then the repopulation would restore the change. Therefore, the 14d PLX results is critically needed data.

2. Internally inconsistent findings: Some aspects of the manuscript provide some inconsistent results that need to be addressed:

a. In Figure 1E and 1F, the regions are the same (ssCTX) but the values for thee control conditions are wildly different (~80 cells per FOV in 1E v. ~190 cells per FOV in 1F). This inconsistency is glaring and needs to be explained and/or rectified.

b. In Figure 1F-G, 14d recovery from PLX treatment leads to an INCREASE of ~50% (from ~190 – ~90 cells) in the ssCTX or ~20% (from ~320 – ~250 cells). However, in Figure 6F, 14d recovery from PLX leads to a 100% INCREASE in cell numbers (from ~3000 – ~6000 cells). These results are inconsistent and should be explained or reconciled.

3. Poor data presentation and Figures:

a. In Figure 1, the flow of data is supposed to be a comparison between PND and END repopulation. However, for most of the results, only END data is presented e.g. 1G; 1I-L.

b. To be considerate of color-blind individuals, the manuscript will be helped by color coding images in green and magenta rather than green and red.

c. Many of the Figures provide full brain images that do not allow sufficient assessment of the claims given based on the images provided. This reviewer thinks the points will be clearer if large brain images are saved for supplemental figures and more detailed images are left in the manuscript proper. E.g. the 14d PLX condition in Figure 1D seems to have some green puncta that may be considered cells but is better determined by more focused images as in D1-4. These kinds of concerns are present in many of the Figures including Figure 2A-C (rather than the whole brain images, individual P2YR12/ IBA1 single frames in addition to the merged frames would be helpful), Figure 6G-J and Figure 9C (see 9D where the cells and amyloid can be more clearly seen).

d. For some figure panels, there isn't always consistency in what is being shown.

e. It would be helpful if the protein stains are consistent. E.g. IBA is red in Figure 4C-E but green in Figure 4F-H. Similarly, Ki67 is magenta in Figure 4F but red in Figure 4G-H. It would be helpful to maintain the same color scheme (especially for a marker like IBA that is used throughout the manuscript) as best as possible.

f. It is not clear why the supplemental Figures are presented as e.g. for Figure S1 "Figure 1-Supplement 1". This supplemental Figure is related to Main Figure 2 not main Figure 1. This Figure is also labeled as Tmem119 in panel A but the legend say "P2RY12". The pattern in the figure is also inconsistent because the control, 14 and 28 days are in the cortex but the 7d is in the WM. This is given without explanation. If necessary, both regions can be shown but the inconsistency is confusing.

g. For Figure 2H, it seems like the expression of IBA1 is very low in the original image that is brightened in the insert. Why is this?

4. Outstanding questions:

a. Although the study highlights the existence of the CSF1Ri-resident population, it is not clear whether this population exists in the normal brain (which would imply a "progenitor" pool) or is induced to take on this phenotype following PLX treatment. The study does not seem interested in this important question. This could potentially be determined by staining for some of the upregulated DEG genes (proteins) in the SVZ/WM of the untreated brain. If the number following 14d PLX (~15 in line 170) is much more reduced than that in the naïve brain, then it becomes difficult to argue for a "resistant" population since the population would have been reduced from whatever the number in the naive brain to 15.

b. The H2K-BCL2 studies are interesting but a bit confusing. First, the statement that these mice provide "overexpression of BCL2 in myeloid lineages affords some resistant to CSF1Ri-induced cell death" is not substantiated by published (or new) data. In addition, no evidence is provided that microglia indeed show over expression of BCL2. Third, the mechanism of BCL2 protection from cell death is not mentioned (presumably through reduction in apoptosis?) Fourth, in the initial study by this group (Elmore et al., 2014), they suggested that CSF1Ri-induced cell death using PLX3397 occurs by apoptosis stating "To confirm that microglia undergo cell death with CSF1R inhibition, we found that many microglia stained for active caspase-3 in the same animals, a classic marker of apoptosing cells (Figure S2J- S2L)" on page 381. Since blocking apoptosis with BCL2 overexpression only increases survival from 0.2-5-10%, do the authors want to expand on whether cell death primarily occurs by apoptosis as they suggested in a previous study?

c. The language used in the discussion on lines 547-549 is strong. The authors seem to be confusing different models. Their findings do not "refute" those other findings since they used different approaches and unlike the current findings, it is likely that an empty microglial niches did not occur in those studies.

d. Finally, it is not clear that the identity of the remaining IBA1+ cells in this study is different from.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Subventricular zone/white matter microglia reconstitute the empty adult microglial niche in a dynamic wave" for further consideration by eLife. Your revised article has been evaluated by Carla Rothlin (Senior Editor) and a Reviewing Editor.

The manuscript has been improved but there are some remaining issues that need to be addressed, as outlined below:

1) The paper has significantly improved. However, there are still concerns about the scope of the claims about the surviving microglia being the so-called WAMs. The fact that this small number of cells in WM survives could be due to many factors including lower drug bioavailability in WM. Please tone down this claim and explain possible alternative reasons for the survival of these cells post-drug treatment.

2) While in this particular case it appears that WM microglia are responsible for most of the cortical repopulation, this is a unique situation and may not reflect a specific property of WM microglia. Several papers including recent ones in eLife have clearly shown repopulation from surviving cortical microglia. Please include these in the discussion and tone down the claims of the uniqueness of WM microglia in their ability to repopulate the brain ( PMID: 33054973, PMID: 34250902)

3) Please go through the critiques from the 3 reviewers and address them as much as possible. No need for further experiments, just data clarifications, and additions to the Discussion section.

Reviewer #1:

In the revised manuscript Hohsfield et al. have answered many of the concerns raised. Specifically:

1. In the earlier version of this manuscript the images showing the "wavefront" of repopulating microglia in the cortex was not very convincing. In the revised manuscript, as per our suggestion, the authors have added high-resolution confocal tiling of large areas showing the wavefront at different stages of the recovery. This data does show more convincingly that dividing microglia emerge from areas newa corpus callosum and appear to advance towards cortex. However, this data also seems to show patchy cortical areas with proliferating microglia that to not appear to emerge from the Wavefront. This suggest that other sources of regenerating microglia exist in the courtext (likely small number of surviving microglia)

2. To address our concerns about the uniqueness of microglia in the white matter tract and their resistance to PLX3397 treatment, the authors postulate based on recent literature that these microglia are resistant due to enhanced TREM2 and CSF1 expression. Additionally, the authors added bulk tissue transcriptome data of 14 days PLX-treated and after 3-5 days recovery time. The authors show that the initial repopulating microglia have genes upregulated in the pathways related to myeloid cell activation/priming, pathogen sensing, and monocyte-macrophage signaling.

3. We had concerns that whether after longer recovery time post PLX treatment, the repopulated microglia would still be transcriptionally different or not. To address this the authors have added transcriptome data after 3 months of recovery and suggest that they are transcriptionally different from the control.

4. We asked whether a longer duration of PLX3397 treatment will eliminate microglia from SVZ and white matter tract or not. The authors in the revised manuscript report that a longer duration of PLX3397 (3 months) eliminates microglia from the white matter region. This suggests that the small population of surviving microglia in WM has a different threshold for cell death, or the bioavailability of the compound is somehow lower in WM compared to cortex. Ideally the authors should discuss this in greater detail in their final version as it is difficult to say with the current data the precise reason for the relative sparing of some WM microglia

5. The authors took our suggestion to completely remove the Alzheimer's disease model data as the data was not conclusive and to add the H2K-BCL2 with more characterization in the main figures.

Overall, as it is the paper is comprehensive and contains interesting data related to the repopulation dynamics of microglia post extensive elimination using PLX3397 for 14 days. The repopulating microglia are transcriptionally different than the control homeostatic microglia and they maintain this difference up to 3 months of recovery. Since the origin of this most (although not likely all) repopulating microglia is from white matter, the authors claim to uncover new insights into white matter microglia function and dynamics. However, we are not very confident about the authors claim of the uniqueness of the white matter microglia. As the authors showed that a longer duration of PLX3397 treatment (3 months) could completely eliminate microglia from white matter, suggesting that they are susceptible to PLX3397 too. Also, the perfusion of this drug in different regions of the brain is not understood. The transcriptional differences observed in the repopulating microglia could be due to the enhanced load of phagocytosis due to extensive PLX3397 treatment. A recent article published in eLife by Mendes et al. (eLife 2021;10:e61173) using in vivo 2 photon imaging showed that the microglia post depletion in cortex repopulate rapidly and locally rather than from migrating microglia and attain morphological maturity within days. Thus, we suggest authors tone down their conclusion about uncovering the uniqueness of white matter microglia. I would suggest changes in the title and abstract which at the moment to us appear to strongly suggest that WM are very unique and the pathway for migration towards cortex could have implications during development, etc. These are interesting hypotheses but are not necessarily well supported by the data

Reviewer #2:

The manuscript addressed many of the concerns, and is much improved in the quality of the presentation. A comparison with the WAM from the recent study is interesting. A. highly relevant eLife paper published last year reported single cell RNA-seq analyses of a relative resistant population ( eLife. 2020 Oct 15;9:e51796. doi: 10.7554/eLife.51796). How do these two resistant populations compare in terms of transcriptomic signatures?

Reviewer #3:

This seems to be a much-improved manuscript but this reviewer cannot fully assess the revisions because it lacks appropriate notification of the changes in the text to allow visualization / assessments of the amendments. For example, for Reviewer 3's comment 1f, the authors state that they corrected the reference to the recovery to normal levels of the cell soma. But this reviewer does not see this mention in the text. The same is true for the response to comment 1g and 1i. Moreover, page numbers referenced in the Essential Revisions do not match up with the text. Also, in response to Reviewer 3 question 1d, Figure 9 is mentioned rather than Figure 10. Also, for 2b, "Figure 6F" is no longer that and the new figure should be mentioned in the response. In response to question 4b, the authors reference Essential Revisions 9 but the response is Essential Revisions 10. This reviewer considers this a set of very unfortunate failures that excessively prolonged this review that has made reviewing the revision extremely difficult.

At a minimum for the next revision, the authors should (i) easily identify the parts of the text that have been revised in a different-colored font and (ii) accurately cite the line numbers in the text where the said revisions are located in the response letter. This should be done for ALL Reviewer 3's comments (both the individualized and the Essential revisions) especially Essential comment 10, points 2-4.

For this reviewer, the following questions have been addressed and DO NOT need to be revisited in a revised resubmission: Questions 1a, b, d, j, k, 2a, 3a, d, f, g, 4c.

For Figure 1c, answers are provided in Essential Revisions 5. But these are not quite satisfactory. The question was whether "residual myeloid cells are CSF1Ri-resistant" with a 14-day treatment regimen. The authors state that "[w]e would like to point out that in many cases we do achieve 100% microglial elimination in whole brain sections following 14d PLX3397 treatment". This actually implies that these cells are not "resistant". In addition, the additional data provided for 3.5 months indeed shows that all microglia are eliminated but the important question of whether the wavefront still forms in this context is not answered as the authors do not show a recovery data after 3.5 months.

For 3b, please work with the journal to generate color blind-sensitive figures.

For 3c, while this reviewer does not think this answer is satisfactory, I will defer to the Editor and other reviewers. If they agree with this response, then I will agree to it too.

For 3e, it is a simple enough request to change the color scheme to facilitate ease of understanding by readers. Please, do this.

For question 4a, the location of the amended tests should be identified. Moreover, the authors do not really answer the question. The question was whether the "resistant" pool (~15 cells) of cells exist in the naïve brain. From the response, it seems like the authors think these cells are equivalent too recently published WAMs. This WAMs are more numerous that these resistant cells in the naïve brain. Therefore, can the authors differentiate between naïve cells becoming WAMs with PLX treatment and naïve WAMs?

https://doi.org/10.7554/eLife.66738.sa1

Author response

Essential revisions:

These are the main issues to address but please look at the individual reviews for further details that should also be considered:

1) Since this paper is entirely dependent on the idea that there is a subpopulation of microglia cells originating from SVZ, that repopulates the cortex after PLX3397 treatment, it will be critical to show better quality high- resolution imaging through confocal tiling spanning large areas of the cortex and white matter. Stain and show several markers of microglia to more convincingly demonstrate that there are no surviving microglia in the cortex that would explain local repopulation. Related to this, In lines 164 for Figure 2C, the authors claim that microglial markers are lost rather than downregulated. But the images have faint (green) signals that seem to be more indicative of downregulation rather than loss. Please explain this.

We thank the reviewer for this comment. To address this concern, we have put together a new figure (Figure 3) that focuses on this issue, providing high-resolution confocal images of large regions of the cortex and white matter stained for several microglial markers (see Figure 3B-G, Figure 5—figure supplement 4F). We have also re-generated all our figures in Adobe Illustrator, which provide much improved quality images to explore. For Figure 2C, the faint green signals (Cx3cr1-CreERT2) are unspecific or vascular staining that arises when no staining is present in the brain. This is more apparent when the green channel is boosted to show that the brain is indeed in focus, however, we have adjusted the green channel back to its original levels. We apologize for the confusion it caused. Furthermore, if the cells were still present in these brains (with downregulated staining) then we would also observe faint red (Iba1+ staining), which we do not, and repopulation would occur in clusters radiating out from surviving cells, which we do not observe. Additionally, downregulation of signal would not explain the disappearance of cells from our Cx3cr1-CreERT2 linage tracing experiments, or the vast literature that has confirmed that microglia are killed with CSF1Ri (or genetic ablation of CS1/IL34, or CSF1R).

2) On the basis of the appearance of microglia at different locations after 14 days of PLX3397 treatment, the authors claim that the repopulating cells elicit a dynamic wave migrating radially and tangentially from the SVZ/ ventricular zone, white matter tract, and caudoputamen. To establish this observation more convincingly and to rule out the possibility of sparse local repopulation specifically in the cortex, as some cells can be seen in figure 3 B (stage III) in the cortex, the authors need to present a more detailed temporal and spatial resolution of the "Wavefront". More detailed imaging is required using high-resolution confocal tiling capturing different time points between stage III and IV to show the gradual emergence of the wavefront from the white matter tract and invading the cortex. A similar extensive analysis of the wavefront in the lineage tracing using cx3cr1-creER should also be performed.

To address this concern, we have added a new supplemental figure (Figure 2—figure supplement 1) that provides high-resolution whole brain (Figure 2—figure supplement 1A) and confocal images (Figure 2—figure supplement 1B) of various timepoints of the “wavefront.” Figure 5—figure supplement 4 provides two whole brain images of the wavefront in Cx3cr1-CreERT2 mice, with a high-resolution image of the “wavefront.”

3) The authors make several quantitative pronouncements without adequate quantitative data that is needed. Eg., while co-expression of certain proteins and myeloid cells (or the lack thereof) are mentioned (lines 247-249), quantitative assessments are required i.e. what percent of IBA1+ cells are NESTIN+ or MASH+ etc? Similarly, the authors make claims of ~25% or ~100% labeling in (lines 260-261) but precise numbers should be provided. It is clear to this reviewer that at least for Figure 4E, there are red (IBA1+ cells) that are not CX3CR1+. Moreover, for the various "stages" in Figure 3B-D, there is no indication of time (days) following PLX withdrawal which would help substantiate temporal claims. Data in Figure 6A-B are just images and would be helped by some quantification of the number of cells. Similarly, data in Figure 6G-K can be presented as percent of GFP cells amongst IBA1 cells as discussed in lines 361-367 and 371. In lines 394-395 while the number of DEG genes is specified for END repopulated cells compared to control microglia (65), the number is not stated for PND repopulated cells. The authors should provide the exact number rather than stating it as a "few" genes.

The reviewer makes a valid point and we have added quantifications to support our claims accordingly. For lines 247-249 (now lines 280-287), we have added quantifications of the percentage of IBA1+ cells positive for precursor and other cell lineage markers (see Figure 5—figure supplement 3F-H). For lines 260-261 (now lines 297-306), we have quantified the percentage of IBA1+ cells that express YFP (see Figure 5—figure supplement 4B). For Figure 3B-D (now Figure 2), we have added approximate time/days following PLX withdrawal, but it should be noted that this timing is dependent on extent of depletion and can vary among animals. For Figure 6A-B (now Figure 7), we have added a quantification of the number of IBA1+ cells within the choroid plexus and outside of it (in the parenchymal areas) (see Figure 7B). For Figure 6G-K (now Figure 7), we have added quantification of the percentage of GFP+ IBA1+ cells above the graph (see Figure 7K). For lines 394-395, we have added the number of DEGs for GLOBAL repopulated cells (now line 445).

4) Do the authors have any sense of what is unique about the surviving cells in the SVZ, why do they escape PLX3397 treatment? If they are simply cx3cr1 positive myeloid cells, the transcriptome data provided only answers how these cells differ once they start repopulation the brain, however it does not answer how they were unique prior to PLX3397 treatment that made them survive. Additional data characterizing this would be important to further determine the relevance of this finding. This issue was also raised by Reviwer#3 (1a).

We postulate local cues from the white matter environment and ventricular zone allow these myeloid cells to escape PLX3397 treatment. This is demonstrated by the fact that once they leave these areas, they become susceptible to PLX3397 treatment, as demonstrated in Figure S2I-J, similar to non-SVZ/WM microglia. Although the signals responsible for shaping macrophage/microglia identity are still being discovered, studies have revealed that the local niche or microenvironment can play an active role in establishing macrophage identity (Gosselin et al., 2014 Cell; Lavin et al., 2014 Cell). We postulate that a combination of transcription factors, availability of CSF1, and epigenetic regulation (via tissue-specific enhancers) contribute to the uniqueness of WM/SVZ microglia and their ability to escape PLX3397 treatment. Several studies have shown that the parenchymal microglial population is heterogeneous highlighting the existence of cells residing in brain-specific regions with distinct identities (Bottcher et al., 2019 Nat Neurosci, Li et al., 2019 Neuron, Masuda et al., 2019 Nature). In fact, a recent single-cell RNAseq analysis has revealed a white matter microglia-specific cluster and specifically a population of white matter-associated microglia (WAMs) that are dependent on TREM2 signaling. The similarities between these WAMs and WM repopulating cells are striking, including a shared disease-associated microglia (DAM) gene signature, along with an upregulation in genes and markers associated phagocytic activity, including CLEC7A and AXL (Safaiyan et al., 2020 Neuron). These recent data indicate that our WM repopulating cells are likely WAMs. Notably, DAMs are also resistant to CSF1Ri (including PLX3397; Spangenberg et al., 2016 Brain). We postulate that surviving SVZ/WM microglia could exhibit enhanced survival due to an upregulation of TREM2 and CSF1 (upregulated in DAMs/WAMs), which can also act as survival signals (Zheng et al., 2018 Front Aging Neurosci). We have added text regarding this issue to the discussion (see lines 634-646). To address the reviewer’s concern about further characterization, we have performed transcriptional characterization of the tissue following 14d PLX3397 treatment (prior to repopulation) and following 5d/7d recovery (see Figure 3J-L and Figure 6A-D), and characterization of WM repopulating cells at 28d recovery (see Figure 8B-D). We have also added a new figure (see Figure 3) that characterizes the expression of multiple microglial markers in surviving WM/SVZ microglia. Unfortunately, due to the extremely low amount of cells that are present at 14d, it is not feasible at this time to perform transcriptional analyses of these cells. It should be noted that we do not believe these cells change their identity during repopulation.

5) The issue of pharmacokinetic analysis of PLX3397 is important. One possibility is that the cells in the white matter rather than being more resistant to PLX3397 receive a relatively lower concentration of the drug than in the cortex due to either reduced perfusion of these areas or alternatively since PLX3397 is lipophilic, it could be incorporated in the myelin sheets thereby reducing its effective concentration. Ideally, the methods used to measure PLX3397 brain concentration should be applied separately to cortical and white matter tissues. What happens if the concentration of PLX3397 used is higher? Is the white matter wavefront still formed?

Perhaps complete elimination of myeloid cells would be achieved with a longer PLLX treatment paradigm. The authors should consider an up to 2-3 month treatment protocol. In addition, the mention of "sustained" in line 196 should be replaced with the specific duration i.e. 14 days to remove any potential confusion that the authors could mean longer that 14 days.

We thank the reviewers for this comment. We would like to point out that most white matter microglia are eliminated by 14d PLX3397. This is apparent throughout the corpus callosum. It is only specifically in areas near white matter/ventricular zone (near the ventricles) that CSF1R-resistant surviving cells are found. Thus, it doesn’t appear that the cells in the white matter receive a different concentration, in fact with their close proximity to the ventricles, one would assume the opposite was true. We have added a discussion of this issue in the discussion (see lines 641-644). We would also like to point out that in many cases we do achieve 100% microglial elimination in whole brain sections following 14d PLX3397 treatment. In these cases, the wavefront still forms. As requested, we have also added data in which animals were treated with PLX3397 (600 ppm) for 3.5 mo (see Figure 3—figure supplement 2F-H). We have replaced “sustained” in line 196 with 14d PLX3397 (see line 253-4).

6) After 28 days of recovery period the authors observed that the repopulating cells after PLX3397 treatment were transcriptionally different from the control population. It would be important to know whether these transcriptional differences in the empty niche repopulating microglia are temporary or are sustained even after a longer recovery time. It is possible that these cells show a *delayed* transition to more control microglial-like features which may be detected at longer than 28 days. Looking at up to 3 months after reconstitution would help to clarify whether the differences between reconstituted cells and control microglia are indeed not due to a slower restoration in the microglial phenotype.

We agree with the author that it would be important to analyze repopulating microglia at a later timepoint and have isolated brain tissue at 90d recovery and performed RNA sequencing analysis. The new data from 90d (~3 mo) recovery mice can be found in Figure 10. These data provide evidence that WM repopulating cells remain phenotypically/morphologically and transcriptionally distinct from homeostatic and GLOBAL repopulating cells rather than regain a homeostatic microglial phenotype.

7) In the AD mouse model, the authors administered PLX3397 and analyzed the effect on amyloid plaque deposition after 3 months of recovery (Figure 9). On the basis of this experiment, the authors claim that the repopulating cells could reduce the amyloid load. However, the significance of this experiment is unclear and does not fit with the central theme of the paper. There is a complete lack of controls addressing amyloid synthesis, processing, secretion, and clearance before and after PLX 3397 treatment. Therefore, the data provided is superficial and not sufficient to demonstrate such a causal relationship. This part should be completely revamped or eliminated from the study.

Also, it is difficult to draw appropriate conclusions from the 5XFAD studies provided in Figure 9E-P because they lack the 14d PLX treatment results. This is needed to determine whether some of the features are unchanged between control and repopulated 5XFAD results or else the 14d depletion is could have itself changed things and then the repopulation would restore the change. Therefore, the 14d PLX results is critically needed data.

We agree with the reviewer. Given the recent findings by Safaiyan et al. (2020), which highlight the existence of WAMs that represent a potentially protective response to aging and disease (including increased phagocytic potential), and the striking similarities between WAMs and our WM repopulating cells, the significance of this experiment and our study has changed. We have chosen to remove these data and will provide a more thorough analysis of these data and findings in a subsequent manuscript.

8) The relative resistance of the microglia in the SVZ and adjacent WM track could very well be due to the local cues/environment, which alter the signaling and lead to relative resistance. While interesting from the experimental perspective, it might not be physiologically relevant for how endogenous microglia repopulate the brain.

It should be noted that the appearance of repopulating cells is similar in phenotypical profiles and anatomical locations to microglia in the developing brain (Ueno et al., 2013 Nat Neurosci). We also have follow-up data highlighting how infiltrating myeloid cells use similar routes to spread throughout the adult brain. In addition, a recent study has highlighted a population of amoeboid microglia that migrate from the ventricular zone into the corpus callosum and engulf OPCs during early postnatal development (Nemes-Baran et al., 2020 Cell Rep). Given the recent findings by Safaiyan et al. (2020), which highlight the existence of WAMs, and the striking similarities between WAMs and our WM repopulating cells, the significance of this experiment has changed. Notably, WAMs are present at low numbers in the adult brain but increase dramatically with age, indicating their relevance to brain aging and age-related diseases. Understanding the sources, properties, and emergence of these cells in the brain is therefore of great importance, which our study sheds light on. Thus, we argue that these findings may have high physiological relevance and novel future application for understanding disease pathogenesis. We would also like to point out that the study of any microglial depletion model (e.g. Elmore et al. 2014, Neuron; Buttger et al. 2016, Immunity) is non-physiological, however, depletion methods provide us with a tool to better understand the biological role and function of myeloid cells as well as their population dynamics in the CNS. All microglial depletion models have caveats, despite this, the field still has gained much insight on microglial biology, and this is especially important for a cell that has been implicated in many neurological disorders. Given the disparate findings regarding microglial heterogeneity, in what appears to be dependent on the method of isolation and RNAseq technique utilized, microglial depletion models could serve an important role in helping decipher these differences.

9) The comparison of control microglia vs. 28-day repopulated microglia in response to LPS is not particularly informative because of the artificial nature of the timing and treatment. It could also be due to the cross talk with astrocytes and other cell types, which are also altered under the 2-week 3397 treatment.

We agree with the reviewer that it is important to consider cross talk with astrocytes and other cell types. Unfortunately, this study did not focus on reporting changes in astrocytes in this paradigm. It is important to note though that during LPS administration, astrocytes are also present, along with repopulating cells. We have added bulk tissue RNAseq analysis of 28d recovery so readers can observe other cellular contributions (see Figure 3J-L). Several studies using microglial depletion and repopulation models have relied on LPS challenge to evaluate the functional immune response differences between endogenous and repopulating cells (O’Neil et al. 2018 Acta Neuropathol. Commun; Lund et al. 2018 Nat Comm; Elmore et al. 2018 Aging Cell). Previously, we have shown that GLOBAL repopulating (at 21d recovery or repopulation) microglia respond in a similar fashion to LPS challenge as control microglia (Elmore et al. 2015 PLoS One), whereas in this study we show that WM repopulating cells exhibit functionally alterations compared to control microglia. To address this comment, we have tempered our conclusions from this study and added text about the possible caveats of this experiment to the discussion, including timing/treatment and astrocytes/other cell type contributions (see lines 486-87).

(10) The H2k-BCL2 experiment (Figure S1B-C ) is potentially interesting. Why put it in supplementary? It would be best to move to main figures and add proper quantification. The H2K-BCL2 studies are a bit confusing. First, the statement that these mice provide "overexpression of BCL2 in myeloid lineages affords some resistant to CSF1Ri-induced cell death" is not substantiated by published (or new) data. In addition, no evidence is provided that microglia indeed show over expression of BCL2. Third, the mechanism of BCL2 protection from cell death is not mentioned (presumably through reduction in apoptosis?) Fourth, in the initial study by this group (Elmore et al., 2014), they suggested that CSF1Ri-induced cell death using PLX3397 occurs by apoptosis stating "To confirm that microglia undergo cell death with CSF1R inhibition, we found that many microglia stained for active caspase-3 in the same animals, a classic marker of apoptosing cells (Figure S2J- S2L)" on page 381. Since blocking apoptosis with BCL2 overexpression only increases survival from 0.2-5-10%, do the authors want to expand on whether cell death primarily occurs by apoptosis as they suggested in a previous study?

We agree with the reviewer and have now added the H2K-BCL2 experiment to the main figures as well as proper quantification (see Figure 4A-D). To address confusion regarding these studies, we have also done the following: (1) added microglial quantification of control (see Figure 4A) and 14d PLX3397 treated H2K-BCL2 mice (see Figure 4B-D) – which addresses the reviewer concern about the enhanced survival of microglia in these mice and the resistance of these mice to CSF1Ri, (2) updated citations and statements about BCL expression in transgenic mice, (3) added text to postulate the mechanism of BCL2 protection from cell death, and (4) added text to expand on how CSF1Ri results in cell death (see Lines 259-269).

Reviewer #1:

All comments are addressed in Essential Revisions above.Reviewer #2:

The manuscript by Hohsfield et al. used high dose of CSF1R/FLT/Kit inhibitor to diminish brain microglia to less than 0.1%, which mostly located in the SVZ and adjacent white matter, called END. They then described their molecular characteristics and the repopulation routes. There are interesting observations with broad description of the characteristics and the resistant population and the population routes. The study also confirmed some of the previous findings, such as the brain-origin of the repopulation. They also showed even after 28 days, the repopulated microglia from END exhibit differential responses to LPS.

The major concern of the study is that the findings failed to lead to deeper understanding of microglial homeostasis regulation in health and in disease in its current form.

(1) The relative resistance of the microglia in the SVZ and adjacent WM track could very well be due to the local cues/environment, which alter the signaling and lead to relative resistance. While interesting from the experimental perspective, it might not be physiologically relevant for how endogenous microglia repopulate the brain.

See Essential Revisions #8.

(2) Some of the conclusions are confirmation of previous studies, such as the origin of END population.

To date, no microglial depletion study has reported the existence of a CSF1Ri resistant microglial cell located in the SVZ/WM tracts. Previous microglial depletion models have shown that repopulation of the microglial niche arises from surviving microglia (microglia we term GLOBAL), which we demonstrate in this manuscript are transcriptionally and functionally distinct despite arising from a Cx3cr1+ cell source. Given the recent discovery of WAMs and the striking similarities to WM microglia, we believe the origin of WM microglia does not share the same origin as surviving microglia in gray matter areas (which we will follow up on in a future publication).

(3) The comparison of control microglia vs. 28-day repopulated microglia in response to LPS is not particularly informative because of the artificial nature of the timing and treatment. It could also be due to the cross talk with astrocytes and other cell types, which are also altered under the 2-week 3397 treatment.

See Essential Revisions #9.

Reviewer #3:

This is manuscript is an attempt to show a novel CSF1Ri-resistent myeloid cell-type that originates in the subventricular zone/white matter and reconstitutes an empty microglial niche following near total CSF1R-dependent microglial elimination. The authors claim that reconstituting myeloid cells populate the empty microglia niche in the brain through migratory routes akin to that employed by endogenous microglia during development. These myeloid cells show a distinct profile from homeostatic microglia and maintain a distinct profile following extended residence in the brain. Finally, in a model of AD, reconstituting the brain with this approach and these cells resulted in a lower AD load that suggests this could be a promising approach to plaque burden in AD. Despite these extremely exciting findings, the manuscript in its current form suffers from significant concerns that need to be addressed before it can be suitable for publication at eLife.

1. Overstated or insufficiently substantiated claims: several details in the manuscript, while interesting overstate the evidence provided by the data or are not substantiated by the current data.

a. Figure 3 seeks to provide spatio-temporal evidence for the "dynamic wave" of repopulation. however, while these findings are interesting, they seem to be based of mainly spatial (without temporal) data taken "between 7 and 14" days of PLX withdrawal. While the impression is given that the authors assess the myeloid cell migration at different timepoints between these time periods, the authors seem to make these conclusions and claims solely from static images at (rather than between) 7 and 14 days. To substantiate 'temporal" claims, the authors need to provide analysis at different time points between 7 and 14 days which they don't seem to have done.

See Essential Revisions #2.

b. The authors make several quantitative pronouncements without adequate quantitative data that is needed. Eg., while co-expression of certain proteins and myeloid cells (or the lack thereof) are mentioned (lines 247-249), quantitative assessments are required i.e. what percent of IBA1+ cells are NESTIN+ or MASH+ etc? Similarly, the authors make claims of ~25% or ~100% labeling in (lines 260-261) but precise numbers should be provided. It is clear to this reviewer that at least for Figure 4E, there are red (IBA1+ cells) that are not CX3CR1+. Moreover, for the various "stages" in Figure 3B-D, there is no indication of time (days) following PLX withdrawal which would help substantiate temporal claims. Data in Figure 6A-B are just images and would be helped by some quantification of the number of cells. Similarly, data in Figure 6G-K can be presented as percent of GFP cells amongst IBA1 cells as discussed in lines 361-367 and 371. In lines 394-395 while the number of DEG genes is specified for END repopulated cells compared to control microglia (65), the number is not stated for PND repopulated cells. The authors should provide the exact number rather than stating it as a "few" genes.

See Essential Revisions #3.

c. A major claim that permeates the manuscript is that the residual myeloid cells are CSF1Ri-resistant. However, this is based upon a 14-day treatment regimen. Perhaps complete elimination of myeloid cells would be achieved with a longer PLLX treatment paradigm. The authors should consider an up to 2-3 month treatment protocol. In addition, the mention of "sustained" in line 196 should be replaced with the specific duration i.e. 14 days to remove any potential confusion that the authors could mean longer that 14 days.

See Essential Revisions #5.

d. Similarly, the density phenotypes (Figure 1F) and the morphology phenotypes (Figure 1 J-L) were only assessed up to 28 days following PLX withdrawal. It is possible that these cells show a *delayed* transition to more control microglial-like features which may be detected at longer than 28 days. This reviewer recommends looking at up to 3 months after reconstitution (which the authors did with the 5XFAD data in Figure 9). This would help to clarify whether the differences between reconstituted cells and control microglia are indeed not due to a slower restoration in the microglial phenotype.

We would like to point out to the reviewer that we did evaluate microglial density and morphology phenotypes in Figure 9 and did observe differences between reconstituted cells and control microglia in wildtype mice at 3 mo recovery. However, to make these observations more apparent, we have removed the 5xFAD data and instead focused our data on the cellular (and transcriptional) alterations in wildtype animals after 3 months of reconstitution (see Figure 10).

e. In lines 164 for Figure 2C, the authors claim that microglial markers are lost rather than downregulated. But the images have faint (green) signals that seem to be more indicative of downregulation rather than loss. Please explain this.

See Essential Revisions #1.

f. Lines 150-151 suggest that repopulating cells display larger cell bodies with no indication of a recovery to normal levels despite the fact that the corresponding Figure shows this in Figure 1J.

We thank the reviewer for pointing this out and have corrected the text accordingly.

g. For data in Figure 3E, the cells are referred to as "proliferating END repopulating myeloid cells" (line 233) but there is no clear evidence of colocalization of Ki67 and IBA1.

We thank the reviewer for pointing this out and have added a reference to this colocalization in the text.

h. The authors state that the myeloid cells they identify are CSF1Ri-resistant after 14 days of PLX3397 (600ppm) treatment. But to ensure that they are indeed truly resistant, a longer (and "sustained"-line 1960) exposure of perhaps up to 2 months would be needed.

See Essential Revisions #5.

i. In lines 308 – 311, the authors claim that since CCL12 antibodies reduced populating cell numbers but not the total distance of spreading. However, it is not clear what "the total distance of cell spreading" is. In the images, the territory occupied by the green cells (myeloid cells) seems to be smaller in antibody treated tissues. Moreover, although the authors claim that this combined observation leads to the conclusion that CCL12 may play a role in the proliferation of repopulating cells, this is not the only possibility. For example, the antibody could affect the survival of cells rather than the proliferation. The authors could detect proliferation using markers in control and CCL12 antibody treatment.

We have clarified “total distance” in the Figure Legend and added this additional interpretation to the manuscript.

j. In line 340-341, the authors claim the CMPs are expanded by flow but they do not provide the gating strategy here. This will help to compare to the cells assessed in the other study.

We have included our written gating strategy in the Methods section for HSCs, CMPs and GMPs. However, we would be happy to provide a visual of our gating strategy to the Reviewer.

k. Finally, it is difficult to draw appropriate conclusions from the 5XFAD studies provided in Figure 9E-P because they lack the 14d PLX treatment results. This is needed to determine whether some of the features are unchanged between control and repopulated 5XFAD results or else the 14d depletion is could have itself changed things and then the repopulation would restore the change. Therefore, the 14d PLX results is critically needed data.

We have removed this data.

2. Internally inconsistent findings: Some aspects of the manuscript provide some inconsistent results that need to be addressed

a. In Figure 1E and 1F, the regions are the same (ssCTX) but the values for thee control conditions are wildly different (~80 cells per FOV in 1E v. ~190 cells per FOV in 1F). This inconsistency is glaring and needs to be explained and/or rectified.

We apologize for the inconsistency and have re-quantified these samples.

b. In Figure 1F-G, 14d recovery from PLX treatment leads to an INCREASE of ~50% (from ~190 – ~90 cells) in the ssCTX or ~20% (from ~320 – ~250 cells). However, in Figure 6F, 14d recovery from PLX leads to a 100% INCREASE in cell numbers (from ~3000 – ~6000 cells). These results are inconsistent and should be explained or reconciled.

We apologize for the reviewer if this presents a conflicting inconsistency. However, we would like to point out that Figure 1F-G only focuses on a small area of the ssCTX at 14d recovery, whereas Figure 6F involves the quantification of the entire brain. We observe in many cases, especially during a stage in which the wave is still making its way across the brain (or the cortex during that 14d recovery), some variability; some parts of the cortex will be fully repopulated and some will be packed with cells in a “wave.” We hope this helps explaining why there would be such a dramatic increase in the number of cells when looking at the entire brain vs. one small brain region.

3. Poor data presentation and Figures:

a. In Figure 1, the flow of data is supposed to be a comparison between PND and END repopulation. However, for most of the results, only END data is presented e.g. 1G; 1I-L.

We have performed extensive analysis of GLOBAL repopulation over the years (Elmore et al. 2014; Elmore et al. 2015; Rice et al. 2017; Elmore et al. 2018; Najafi et al. 2018), and use this manuscript instead to characterize and focus on WM repopulation, which is why GLOBAL characterization is limited in the manuscript.

b. To be considerate of color-blind individuals, the manuscript will be helped by color coding images in green and magenta rather than green and red.

We agree with the reviewer about the importance of inclusion and consideration for color-blind individuals, unfortunately changing our color scheme at this time will be difficult, but we will utilize green and magenta for future manuscripts. We are also willing to work with the journal after acceptance to address this issue and generate more color-blind friendly images.

c. Many of the Figures provide full brain images that do not allow sufficient assessment of the claims given based on the images provided. This reviewer thinks the points will be clearer if large brain images are saved for supplemental figures and more detailed images are left in the manuscript proper. E.g. the 14d PLX condition in Figure 1D seems to have some green puncta that may be considered cells but is better determined by more focused images as in D1-4. These kinds of concerns are present in many of the Figures including Figure 2A-C (rather than the whole brain images, individual P2YR12/ IBA1 single frames in addition to the merged frames would be helpful), Figure 6G-J and Figure 9C (see 9D where the cells and amyloid can be more clearly seen).

We thank the reviewer for this suggestion. We agree that more detailed images for the reader to explore would be important, but also argue that given the nature and motility of repopulation, we believe whole brain images do provide a greater lens for exploration. To address the issue of better assessment of claims, we have imported our Figures into Adobe Illustrator, which now allow for better resolution of whole brain images and individual cells.

d. For some figure panels, there isn't always consistency in what is being shown.

If the reviewer provided a specific example, we would be happy to correct the figure panels.

e. It would be helpful if the protein stains are consistent. E.g. IBA is red in Figure 4C-E but green in Figure 4F-H. Similarly, Ki67 is magenta in Figure 4F but red in Figure 4G-H. It would be helpful to maintain the same color scheme (especially for a marker like IBA that is used throughout the manuscript) as best as possible.

We apologize for the inconsistency in our color schemes and did make every effort to try to keep IBA1 in the same channel. Unfortunately, changing the color schemes at this time for all Figures would not be feasible, but again we will be aware of this issue for future manuscripts.

f. It is not clear why the supplemental Figures are presented as e.g. for Figure S1 "Figure 1-Supplement 1". This supplemental Figure is related to Main Figure 2 not main Figure 1. This Figure is also labeled as Tmem119 in panel A but the legend says "P2RY12". The pattern in the figure is also inconsistent because the control, 14 and 28 days are in the cortex but the 7d is in the WM. This is given without explanation. If necessary, both regions can be shown but the inconsistency is confusing.

We have corrected these labeling inconsistencies. As for using WM for 7d recovery, this is because microglia are only found near the SVZ and WM areas at this timepoint. This pattern was also presented in Figure 1.

g. For Figure 2H, it seems like the expression of IBA1 is very low in the original image that is brightened in the insert. Why is this?

The Insert for Figure 2H is a 63X high resolution image of that microglial cell. The image was not brightened.

4. Outstanding questions:

a. Although the study highlights the existence of the CSF1Ri-resident population, it is not clear whether this population exists in the normal brain (which would imply a "progenitor" pool) or is induced to take on this phenotype following PLX treatment. The study does not seem interested in this important question. This could potentially be determined by staining for some of the upregulated DEG genes (proteins) in the SVZ/WM of the untreated brain. If the number following 14d PLX (~15 in line 170) is much more reduced than that in the naïve brain, then it becomes difficult to argue for a "resistant" population since the population would have been reduced from whatever the number in the naive brain to 15.

We are interested in determining whether this population exists in the normal brain. We indeed explored whether our findings indicated a progenitor pool and went about staining for several precursor cell markers and lineage traced for those in Figure 5. A recent single-cell RNAseq analysis has revealed a white matter microglia-specific cluster and specifically a population of white matter-associated microglia (WAMs) that are dependent on TREM2 signaling. The similarities between these WAMs and WM repopulating cells are striking, including a shared disease-associated microglia (DAM) gene signature, along with an upregulation in genes and markers associated phagocytic activity, including CLEC7A and AXL (Safaiyan et al., 2020 Neuron). These recent data indicate that our WM repopulating cells are WAMs, reinforcing the existence of a distinct population of microglia in the WM that does indeed exist in the normal brain. We have updated our manuscript accordingly.

b. The H2K-BCL2 studies are interesting but a bit confusing. First, the statement that these mice provide "overexpression of BCL2 in myeloid lineages affords some resistant to CSF1Ri-induced cell death" is not substantiated by published (or new) data. In addition, no evidence is provided that microglia indeed show over expression of BCL2. Third, the mechanism of BCL2 protection from cell death is not mentioned (presumably through reduction in apoptosis?) Fourth, in the initial study by this group (Elmore et al., 2014), they suggested that CSF1Ri-induced cell death using PLX3397 occurs by apoptosis stating "To confirm that microglia undergo cell death with CSF1R inhibition, we found that many microglia stained for active caspase-3 in the same animals, a classic marker of apoptosing cells (Figure S2J- S2L)" on page 381. Since blocking apoptosis with BCL2 overexpression only increases survival from 0.2-5-10%, do the authors want to expand on whether cell death primarily occurs by apoptosis as they suggested in a previous study?

See Essential Revisions #9.

c. The language used in the discussion on lines 547-549 is strong. The authors seem to be confusing different models. Their findings do not "refute" those other findings since they used different approaches and unlike the current findings, it is likely that an empty microglial niches did not occur in those studies.

We have tempered our language in the discussion accordingly.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

(1) The paper has significantly improved. However, there are still concerns about the scope of the claims about the surviving microglia being the so-called WAMs. The fact that this small number of cells in WM survives could be due to many factors including lower drug bioavailability in WM. Please tone down this claim and explain possible alternative reasons for the survival of these cells post-drug treatment.

We thank the editors for their comment. We have toned down the claim that the surviving microglia are WAMs (see lines 47, 49, 110, 486 (deletion), 488-9, 709-11, 720, 723 (deletion), 727-8). We have also added to our section in the discussion explaining the alternative reasons for microglial survival in WM areas post-drug treatment (see previously added lines 645-57, newly added lines 654-5, 657-8, 659-72) as well as to the Results section (see lines 226-9).

(2) While in this particular case it appears that WM microglia are responsible for most of the cortical repopulation, this is a unique situation and may not reflect a specific property of WM microglia. Several papers including recent ones in eLife have clearly shown repopulation from surviving cortical microglia. Please include these in the discussion and tone down the claims of the uniqueness of WM microglia in their ability to repopulate the brain ( PMID: 33054973, PMID: 34250902)

We agree with the editors that repopulating cortical areas is not a specific or unique property of WM microglia. In 2014, we identified the unique ability of microglia (including cortical microglia) to repopulate the brain following CSF1R inhibition (Elmore et al., 2014 Neuron). Thus, it was never our intention to make this claim. Under repopulation conditions in which an empty microglial niche is not achieved (< 98% IBA1+ cell loss), so in the papers that are cited above in which 88% (Zhan et al. 2020) and 75% (Mendes et al. 2021) microglial loss were reported, cortical repopulation can be derived from cortical microglia. Under conditions in which < 98% microglial loss is achieved, we term this form of repopulation “GLOBAL repopulation,” and would like to clarify that repopulating cells derive from the nearest surviving microglia in the brain region where the repopulating cells appear. For this reason, we state on lines 122-3, “repopulation is dependent on the local proliferation and clonal expansion of surviving microglia.” However, under WM repopulation (> 98% IBA1+ cell loss), there are no surviving cells in a particular brain region to give rise to cells, thus cortical repopulation (even hippocampal or thalamic repopulation) does not occur due to surviving microglia in that area. We apologize for the confusion and have included a clarification of this in the discussion (see lines 590-97). We have also toned down our claims about the uniqueness of WM microglia to repopulate the brain (see lines 43 (deletion), 44 (deletion), 45 (deletion), 725-6 (deletion)). We have also added these suggested citations (see lines 123, 668).

(3) Please go through the critiques from the 3 reviewers and address them as much as possible. No need for further experiments, just data clarifications, and additions to the Discussion section.

We have gone through the critiques from the 3 reviewers and have addressed them as much as possible (see below).

Reviewer #1:

In the revised manuscript Hohsfield et al. have answered many of the concerns raised. Specifically:

1. In the earlier version of this manuscript the images showing the "wavefront" of repopulating microglia in the cortex was not very convincing. In the revised manuscript, as per our suggestion, the authors have added high-resolution confocal tiling of large areas showing the wavefront at different stages of the recovery. This data does show more convincingly that dividing microglia emerge from areas newa corpus callosum and appear to advance towards cortex. However, this data also seems to show patchy cortical areas with proliferating microglia that to not appear to emerge from the Wavefront. This suggest that other sources of regenerating microglia exist in the courtext (likely small number of surviving microglia)

We have noted this observation in the Results section (see lines 190-2).

2. To address our concerns about the uniqueness of microglia in the white matter tract and their resistance to PLX3397 treatment, the authors postulate based on recent literature that these microglia are resistant due to enhanced TREM2 and CSF1 expression. Additionally, the authors added bulk tissue transcriptome data of 14 days PLX-treated and after 3-5 days recovery time. The authors show that the initial repopulating microglia have genes upregulated in the pathways related to myeloid cell activation/priming, pathogen sensing, and monocyte-macrophage signaling.

We are glad to have addressed the reviewer’s concerns.

3. We had concerns that whether after longer recovery time post PLX treatment, the repopulated microglia would still be transcriptionally different or not. To address this the authors have added transcriptome data after 3 months of recovery and suggest that they are transcriptionally different from the control.

We are glad to have addressed the reviewer’s concerns.

4. We asked whether a longer duration of PLX3397 treatment will eliminate microglia from SVZ and white matter tract or not. The authors in the revised manuscript report that a longer duration of PLX3397 (3 months) eliminates microglia from the white matter region. This suggests that the small population of surviving microglia in WM has a different threshold for cell death, or the bioavailability of the compound is somehow lower in WM compared to cortex. Ideally the authors should;d discuss this in greater detail in their final version as it is difficult to say with the current data the precise reason for the relative sparing of some WM microglia

We have added to our section in the discussion explaining the alternative reasons for microglial survival in WM areas post-drug treatment (see previously added lines 645-57, newly added lines 658-72) as well as to the Results section (see lines 226-9).

5. The authors took our suggestion to completely remove the Alzheimer's disease model data as the data was not conclusive and to add the H2K-BCL2 with more characterization in the main figures.

We are glad to have addressed the reviewer’s concerns.

Overall, as it is the paper is comprehensive and contains interesting data related to the repopulation dynamics of microglia post extensive elimination using PLX3397 for 14 days. The repopulating microglia are transcriptionally different than the control homeostatic microglia and they maintain this difference up to 3 months of recovery. Since the origin of this most (although not likely all) repopulating microglia is from white matter, the authors claim to uncover new insights into white matter microglia function and dynamics. However, we are not very confident about the authors claim of the uniqueness of the white matter microglia. As the authors showed that a longer duration of PLX3397 treatment (3 months) could completely eliminate microglia from white matter, suggesting that they are susceptible to PLX3397 too. Also, the perfusion of this drug in different regions of the brain is not understood. The transcriptional differences observed in the repopulating microglia could be due to the enhanced load of phagocytosis due to extensive PLX3397 treatment. A recent article published in eLife by Mendes et al. (eLife 2021;10:e61173) using in vivo 2 photon imaging showed that the microglia post depletion in cortex repopulate rapidly and locally rather than from migrating microglia and attain morphological maturity within days. Thus, we suggest authors tone down their conclusion about uncovering the uniqueness of white matter microglia. I would suggest changes in the title and abstract which at the moment to us appear to strongly suggest that WM are very unique and the pathway for migration towards cortex could have implications during development, etc. These are interesting hypotheses but are not necessarily well supported by the data

We thank the reviewer for their comment. We have toned down our claims about the uniqueness of WM microglia to repopulate the brain (see lines 43 (deletion), 44 (deletion), 45 (deletion), 725-6 (deletion)). However, we would like to highlight that we provide several examples of the uniqueness of these cells to homeostatic microglia in our manuscript, aside from their initial resistance to 14d PLX3397 (600 ppm) treatment, including their migratory patterns during repopulation, morphology, expression of cell surface markers, and transcriptional profile. We also provide experiments that describe their ability to fill the microglial niche without repercussions on other cell types or behavior, indicating the nuanced complexity of distinct microglial populations and their functions. We would also like to point out that we are not alone in our claim that white matter microglia are unique, other studies have demonstrated this as well (Safaiyan et al., 2021, van der Poel, 2019, Hagemeyer et al., 2017). It should also be noted that the report by Mendes et al. (2021) uses 7d PLX5622 treatment, which results in partial depletion – 75% of microglia are depleted. Under this paradigm, GLOBAL repopulation would occur, which we state in the manuscript derives from surviving microglia and exhibits distinct properties (such as the appearance of clusters of surviving microglia that proliferate to give rise to repopulating cells) compared to WM repopulation (with the majority of cells deriving from a wave of proliferating cells). We postulate that there is a certain threshold of microglial depletion (what we define as an empty microglial niche of > 98% depletion) that must be reached in order for WM repopulation to occur. We have made changes to the abstract, but would argue that the title succinctly and appropriately describes our findings in this manuscript. We would like to know if there is room for discussion with the editors on other appropriate titles.

Reviewer #2:

The manuscript addressed many of the concerns, and is much improved in the quality of the presentation. A comparison with the WAM from the recent study is interesting. A. highly relevant eLife paper published last year reported single cell RNA-seq analyses of a a relative resistant population ( ELife. 2020 Oct 15;9:e51796. doi: 10.7554/eLife.51796). How do these two resistant populations compare in terms of transcriptomic signatures?

The reviewer highlights an interesting paper that we also found extremely relevant for our study. We have added this paper to our Discussion section (see lines 666-72). The authors identify a MAC2+ progenitor-like microglial cell that is resistant to CSF1Ri. It is important to note that the authors do not achieve an empty microglial niche, reporting a loss in 88% of IBA1+ cells, meaning this type of repopulation (GLOBAL repopulation) is different than WM repopulation we describe in the current manuscript. Although we have not done comprehensive transcriptional analysis between these cells, the authors report an upregulation in Lyz2 and MHC-associated genes, and a downregulation in Tmem119, Mafb, Cx3cr1, and Csf1r. Our repopulating cells also upregulate Lyz2 and MHC-associated genes, and downregulate many canonical microglial signature genes, however, we would like to point out that DAMs and WAMs also display a similar transcriptional profile. Thus, we cannot draw many conclusions from this data. We have also stained our WM repopulating cells for MAC2 and although some of the surviving cells in the WM are MAC2+, the repopulating cells that give rise to the microglial population are MAC2-. We are in the midst of preparing a follow-up manuscript that will detail and clarify these findings.

Reviewer #3:

This seems to be a much-improved manuscript but this reviewer cannot fully assess the revisions because it lacks appropriate notification of the changes in the text to allow visualization / assessments of the amendments. For example, for Reviewer 3's comment 1f, the authors state that they corrected the reference to the recovery to normal levels of the cell soma. But this reviewer does not see this mention in the text. The same is true for the response to comment 1g and 1i. Moreover, page numbers referenced in the Essential Revisions do not match up with the text. Also, in response to Reviewer 3 question 1d, Figure 9 is mentioned rather than Figure 10. Also, for 2b, "Figure 6F" is no longer that and the new figure should be mentioned in the response. In response to question 4b, the authors reference Essential Revisions 9 but the response is Essential Revisions 10. This reviewer considers this a set of very unfortunate failures that excessively prolonged this review that has made reviewing the revision extremely difficult.

We apologize to the reviewer for their frustrating reviewing experience. It was not our intention to make this revision difficult. We provided a manuscript that identified revised text changes in red in our last revision. We are unclear as to why the reviewer did not receive this version. We have also noticed that the PDF version of the article file does not have accurate line numbers. We are unsure as to why this occurs, but would recommend that the reviewer use the word file for accurate line numbers. We have added line numbers to this response so the reviewer can better find our changes. Please note that previous text changes made in the last round of revisions are colored in orange and the recently made text changes for this round of revision are in blue (both chosen based on colorblind friendly recommendations).

https://doi.org/10.7554/eLife.66738.sa2

Article and author information

Author details

  1. Lindsay A Hohsfield

    1. Department of Neurobiology and Behavior, Irvine, United States
    2. Institute for Memory Impairments and Neurological Disorders, Irvine, United States
    Contribution
    Conceptualization, Data curation, Formal analysis, Funding acquisition, Validation, Investigation, Visualization, Methodology, Writing - original draft, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-6018-656X
  2. Allison R Najafi

    1. Department of Neurobiology and Behavior, Irvine, United States
    2. Institute for Memory Impairments and Neurological Disorders, Irvine, United States
    Contribution
    Data curation, Formal analysis, Investigation, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
  3. Yasamine Ghorbanian

    1. Sue and Bill Gross Stem Cell Research Center, Irvine, United States
    2. Department of Molecular Biology and Biochemistry, Irvine, United States
    Contribution
    Data curation, Formal analysis, Investigation, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
  4. Neelakshi Soni

    1. Department of Neurobiology and Behavior, Irvine, United States
    2. Institute for Memory Impairments and Neurological Disorders, Irvine, United States
    Contribution
    Formal analysis, Investigation, Methodology
    Competing interests
    No competing interests declared
  5. Joshua Crapser

    1. Department of Neurobiology and Behavior, Irvine, United States
    2. Institute for Memory Impairments and Neurological Disorders, Irvine, United States
    Contribution
    Conceptualization, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
  6. Dario X Figueroa Velez

    Department of Neurobiology and Behavior, Irvine, United States
    Contribution
    Data curation, Investigation, Visualization, Methodology
    Competing interests
    No competing interests declared
  7. Shan Jiang

    Department of Developmental and Cell Biology, Irvine, United States
    Contribution
    Formal analysis, Validation, Visualization, Methodology
    Competing interests
    No competing interests declared
  8. Sarah E Royer

    1. Department of Neurobiology and Behavior, Irvine, United States
    2. Sue and Bill Gross Stem Cell Research Center, Irvine, United States
    3. Department of Anatomy and Neurobiology, Irvine, United States
    Contribution
    Investigation, Methodology
    Competing interests
    No competing interests declared
  9. Sung Jin Kim

    1. Department of Neurobiology and Behavior, Irvine, United States
    2. Institute for Memory Impairments and Neurological Disorders, Irvine, United States
    Contribution
    Data curation, Investigation, Methodology
    Competing interests
    No competing interests declared
  10. Caden M Henningfield

    1. Department of Neurobiology and Behavior, Irvine, United States
    2. Institute for Memory Impairments and Neurological Disorders, Irvine, United States
    Contribution
    Data curation
    Competing interests
    No competing interests declared
  11. Aileen Anderson

    1. Department of Neurobiology and Behavior, Irvine, United States
    2. Institute for Memory Impairments and Neurological Disorders, Irvine, United States
    3. Sue and Bill Gross Stem Cell Research Center, Irvine, United States
    4. Department of Anatomy and Neurobiology, Irvine, United States
    5. Department of Physical Medicine & Rehabilitation, University of California, Irvine, Irvine, United States
    Contribution
    Supervision, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-8203-8891
  12. Sunil P Gandhi

    Department of Neurobiology and Behavior, Irvine, United States
    Contribution
    Resources, Supervision
    Competing interests
    No competing interests declared
  13. Ali Mortazavi

    Department of Developmental and Cell Biology, Irvine, United States
    Contribution
    Resources, Software, Supervision
    Competing interests
    No competing interests declared
  14. Matthew A Inlay

    1. Department of Neurobiology and Behavior, Irvine, United States
    2. Sue and Bill Gross Stem Cell Research Center, Irvine, United States
    3. Department of Molecular Biology and Biochemistry, Irvine, United States
    Contribution
    Resources, Supervision, Writing - review and editing
    Competing interests
    No competing interests declared
  15. Kim N Green

    1. Department of Neurobiology and Behavior, Irvine, United States
    2. Institute for Memory Impairments and Neurological Disorders, Irvine, United States
    Contribution
    Conceptualization, Resources, Software, Formal analysis, Supervision, Funding acquisition, Project administration, Writing - review and editing
    For correspondence
    kngreen@uci.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-6049-6744

Funding

National Institute of Neurological Disorders and Stroke (R01NS083801)

  • Kim N Green

National Institute on Aging (RF1AG056768)

  • Kim N Green

National Institute on Aging (P50AG016573)

  • Kim N Green

National Institute of Neurological Disorders and Stroke (F31NS108611)

  • Joshua Crapser

National Institute of Neurological Disorders and Stroke (T32NS082174)

  • Yasamine Ghorbanian

Alzheimer's Association (AARF-16-442762)

  • Lindsay A Hohsfield

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

 We thank Edna Hingco and Ayer Darling Jue for their excellent technical assistance, and Brian L West and Andrey Rymar at Plexxikon, Inc for providing and formulating CSF1Ri chow and pharmacokinetics analysis. We are grateful to Claudia I Czimczik for proofreading the manuscript. This work was supported by the National Institutes of Health (NIH) under awards: R01NS083801 (NINDS), R01AG056768 (NIA), and P50AG016573 (NIA) to KNG; F31NS108611 (NINDS) to JDC and T32NS082174 (NINDS) to YG. LAH was supported by the Alzheimer’s Association Research Fellowship (AARF-16–442762).

Ethics

Animal experimentation: All rodent experiments were performed in accordance with animal protocols approved (AUP-17-179) by the Institutional Animal Care and Use Committee at the University of California, Irvine (UCI).

Senior Editor

  1. Carla V Rothlin, Yale School of Medicine, United States

Reviewing Editor

  1. Jaime Grutzendler, Yale University, United States

Reviewer

  1. Jaime Grutzendler, Yale University, United States

Publication history

  1. Received: January 20, 2021
  2. Preprint posted: February 18, 2021 (view preprint)
  3. Accepted: August 22, 2021
  4. Accepted Manuscript published: August 23, 2021 (version 1)
  5. Version of Record published: September 8, 2021 (version 2)

Copyright

© 2021, Hohsfield et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,156
    Page views
  • 174
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Evolutionary Biology
    2. Immunology and Inflammation
    Srijan Seal et al.
    Review Article

    Researchers worldwide are repeatedly warning us against future zoonotic diseases resulting from humankind’s insurgence into natural ecosystems. The same zoonotic pathogens that cause severe infections in a human host frequently fail to produce any disease outcome in their natural hosts. What precise features of the immune system enable natural reservoirs to carry these pathogens so efficiently? To understand these effects, we highlight the importance of tracing the evolutionary basis of pathogen tolerance in reservoir hosts, while drawing implications from their diverse physiological and life-history traits, and ecological contexts of host-pathogen interactions. Long-term co-evolution might allow reservoir hosts to modulate immunity and evolve tolerance to zoonotic pathogens, increasing their circulation and infectious period. Such processes can also create a genetically diverse pathogen pool by allowing more mutations and genetic exchanges between circulating strains, thereby harboring rare alive-on-arrival variants with extended infectivity to new hosts (i.e., spillover). Finally, we end by underscoring the indispensability of a large multidisciplinary empirical framework to explore the proposed link between evolved tolerance, pathogen prevalence, and spillover in the wild.

    1. Immunology and Inflammation
    2. Microbiology and Infectious Disease
    Hao Gu et al.
    Research Article

    Vaccination strategies for rapid protection against multidrug-resistant bacterial infection are very important, especially for hospitalized patients who have high risk of exposure to these bacteria. However, few such vaccination strategies exist due to a shortage of knowledge supporting their rapid effect. Here, we demonstrated that a single intranasal immunization of inactivated whole cell of Acinetobacter baumannii elicits rapid protection against broad A. baumannii-infected pneumonia via training of innate immune response in Rag1-/- mice. Immunization-trained alveolar macrophages (AMs) showed enhanced TNF-α production upon restimulation. Adoptive transfer of immunization-trained AMs into naive mice mediated rapid protection against infection. Elevated TLR4 expression on vaccination-trained AMs contributed to rapid protection. Moreover, immunization-induced rapid protection was also seen in Pseudomonas aeruginosa and Klebsiella pneumoniae pneumonia models, but not in Staphylococcus aureus and Streptococcus pneumoniae model. Our data reveal that a single intranasal immunization induces rapid and efficient protection against certain Gram-negative bacterial pneumonia via training AMs response, which highlights the importance and the possibility of harnessing trained immunity of AMs to design rapid-effecting vaccine.