An overlooked subset of Cx3cr1wt/wt microglia in the Cx3cr1CreER-Eyfp/wt mouse has a repopulation advantage over Cx3cr1CreER-Eyfp/wt microglia following microglial depletion

Background Fluorescent reporter labeling and promoter-driven Cre-recombinant technologies have facilitated cellular investigations of physiological and pathological processes, including the widespread use of the Cx3cr1CreER-Eyfp/wt mouse strain for studies of microglia. Methods Immunohistochemistry, Flow Cytometry, RNA sequencing and whole-genome sequencing were used to identify the subpopulation of microglia in Cx3cr1CreER-Eyfp/wt mouse brains. Genetically mediated microglia depletion using Cx3cr1CreER-Eyfp/wtRosa26DTA/wt mice and CSF1 receptor inhibitor PLX3397 were used to deplete microglia. Primary microglia proliferation and migration assay were used for in vitro studies. Results We unexpectedly identified a subpopulation of microglia devoid of genetic modification, exhibiting higher Cx3cr1 and CX3CR1 expression than Cx3cr1CreER-Eyfp/wtCre+Eyfp+ microglia in Cx3cr1CreER-Eyfp/wt mouse brains, thus termed Cx3cr1highCre−Eyfp− microglia. This subpopulation constituted less than 1% of all microglia under homeostatic conditions, but after Cre-driven DTA-mediated microglial depletion, Cx3cr1highCre−Eyfp− microglia escaped depletion and proliferated extensively, eventually occupying one-third of the total microglial pool. We further demonstrated that the Cx3cr1highCre−Eyfp− microglia had lost their genetic heterozygosity and become homozygous for wild-type Cx3cr1. Therefore, Cx3cr1highCre−Eyfp− microglia are Cx3cr1wt/wtCre−Eyfp−. Finally, we demonstrated that CX3CL1–CX3CR1 signaling regulates microglial repopulation both in vivo and in vitro. Conclusions Our results raise a cautionary note regarding the use of Cx3cr1CreER-Eyfp/wt mouse strains, particularly when interpreting the results of fate mapping, and microglial depletion and repopulation studies. Supplementary Information The online version contains supplementary material available at 10.1186/s12974-022-02381-6.


Background
Microglia are derived from the yolk sac during early embryonic development and represent approximately 10% of the healthy adult brain's total cell population [1,2]. They play critical roles in maintaining brain development and function [3,4]. In their homeostatic state, they display a highly ramified morphology, efficiently surveying the central nervous system (CNS) microenvironment, recognizing and clearing cell debris [5]. Microglia play pivotal roles in diverse CNS diseases including neurodevelopmental disorders, neurodegenerative disorders, and high-grade glioma [6][7][8][9][10][11][12][13]. Targeting microglia has thus emerged as an attractive strategy to modulate neuroinflammation in the context of various CNS diseases [14,15].
In this study, we report the existence of an unexpected, small population of Cx3cr1 high Cre − Eyfp − microglia in Cx3cr1 CreER-Eyfp/wt mice. After genetically mediated microglial depletion using Cx3cr1 CreER-Eyfp/wt Rosa26 DTA/wt mice, Cx3cr1 high Cre − Eyfp − microglia escape depletion, display a repopulation advantage and eventually constitute one-third of the total repopulated microglial pool. We further determined that the Cx3cr1 high Cre − Eyfp − microglia are Cx3cr1 wt/wt Cre − Eyfp − . Finally, we demonstrate the vital role of CX3CL1-CX3CR1 signaling in regulating microglial repopulation post-depletion. Not being aware of this population may result in misinterpretation of the results generated since these cells escape detection (not carrying the Eyfp or Gfp) and cannot be modified (lacking Cre expression) as expected.

PLX3397 chow treatment
PLX3397 (HY-16749, MedChemExpress) was formulated with standard chow at 290 mg/kg by SAFE Nutrition Service, France. Four-week-old mice were treated with PLX3397 chow for 3 weeks and sacrificed at different timepoints afterwards.
Iba-1 + microglia were counted using an Axio Imager M2 microscope with Apotome attachment (Carl Zeiss). Cells were counted exhaustively on day 1, day 3, day 7 and day 10 after the final Tam injection. While a fractionator was used to estimate the positive cells on day 42 after the final Tam injection, the counting frame width and height are 150 μm, sample grid X and Y are 300 μm. Thus, all the data has a 2nd estimated coefficient of errors (Schmitz-Hof ) less than 0.1 [32]. Then, relative microglia number ratios were calculated with reference to the control group.
All the fluorescence images were captured using an LSM700 laser scanning confocal microscope (Axioobserver Z1; CarlZeiss microscopy, Germany) and analyzed using ZEN software (the black edition; Zeiss).

Morphological analysis
Confocal microscopy was used to acquire images at 3 µm intervals, with a 40× objective lens used for Cx3cr1 depletion analysis ( Fig. 2g and Additional file 1: Fig. S4G) and a 20× lens used for Cx3cr1 CreER-Eyfp/wt physiological analysis (Additional file 1: Fig. S5A) (Plan-Apochromat lens, 20×/0.8 and 40×/1.3 Oil objective, Carl Zeiss). Microglial morphology was analyzed with the skeleton analysis method using FIJI open-source image analysis software (Image J 1.51s, NIH) [33,34]. Briefly, the maximum intensity projections of z-series stacks were created. The Iba-1 + channel and the CX3CR1-EYFP channel were enhanced in brightness to visualize all processes, followed by de-speckling to eliminate single-pixel backgrounds. Then, each channel image was processed into a binary image, and a topological skeleton image was created (Additional file 1: Fig. S2A). The Analyze Skeleton plugin feature was applied to the whole image frame, and the process length and the number of endpoints were normalized by the cell number per frame. Three morphological parameters were used: (1) total length of the processes, (2) number of endpoints, and (3) microglial process area. The process length and the number of endpoints of CX3CR1 +/+ EYFP − Iba-1 + cells were calculated by subtracting the CX3CR1-EYFP channel data from the Iba-1 channel data. Each cell's process area was represented as the convex hull area by connecting the process ends using the polygon tool (Additional file 1: Fig. S2B, C) [35]. The selection criteria for the process area analysis were relatively isolated from the processes of the surrounding cells, and whole processes were not truncated and are within the image.

Microglial isolation
Single-cell suspensions were prepared using a Neural Tissue Dissociation Kit (T, Miltenyi Biotec), and myelin was removed using 38% Percoll. After passing through a 40 μm cell strainer, cells were incubated with the CD11b magnetic bead for 20 min. CD11b + cells were isolated and collected using a magnetic field and MS column (Miltenyi Biotec, Germany).

Flow cytometry
After microglia isolation, cells were transferred into 96-well V-bottom plates and incubated at 4 °C for 20 min with the following antibodies: yellow dead cell stain kit (L34959, Thermo Fisher Scientific), Near-IR dead cell stain kit (L34975, Thermo Fisher Scientific), CD11b

Cells
were sorted into CD11b + CD45 + Ly6C − CX3CR1 + EYFP + and CD11b + CD45 + Ly6C − CX3CR1 + EYFP − populations using a BD Influx Cell Sorter. Briefly, 100 cells were collected directly into lysis buffer, and the library was built by the smart-seq method. Gene expression data were analyzed using Qlucore Omics Explorer 3.4. For twogroup and multi-group comparisons of candidate genes, expression was considered as significantly different if p < 0.05 using a heteroscedastic two-tailed Student's t test or an F test, respectively. For transcriptome-wide analyses, the Benjamini-Hochberg method was used to correct multiple tests.

Whole-genome sequencing
Whole-genome sequencing was performed essentially as in DNTR-seq [36] with minor modifications. After PCR amplification, the reactions were cleaned-up by SPRIbeads (at 0.9× volume) for size selection and were pooled together according in equal parts according to their DNA concentration (qubit dsDNA quantification assays, ThermoFisher). The pooled library was then cleanedup one more time by SPRI-beads (at 0.9× volume) and sequenced on an Illumina NextSeq 550 using the high output 75 bp kit. Genomic reads were trimmed for adaptor sequences using TrimGalore and mapped to human genome reference build hg38 with the non-genomic parts of the Cx3cr1-CreERT2 targeting vector added as a separate template. Duplicates were removed using Picard Mark-Duplicate, along with read pairs with MAPQ < 20. Reads mapping to the part of the vector which is inserted into the genome (position 1173-6936), and to the region surrounding the insertion site (chr9:120,040,000-120,080,000) were extracted from the bam files. Physical coverage (eg. the number of times a base is spanned by paired read mapping positions) was visualized using a custom R script.

Primary microglia proliferation and migration assay
Primary microglia were cultured as follows: cerebrum tissues were collected from postnatal days 3 mice and used for generating mixed glia cultures in T75 flasks. After 2 weeks of primary culture (DMEM F-12 with 20 ng/ml M-CSF), the microglia were isolated with CD11b magnetic bead sorting (MACS, Miltenyi Biotec) according to the manufacturer's protocol. For proliferation assay, isolated CD11b positive cells were seeded at a density of 0.2 × 10 5 on the coverslips (83.1840.002, Sarstedt) in 24-well plate for 2 days with DMEM F-12 medium containing 10% FCS and 20 ng/ml M-CSF, followed by replacement with serum-free medium (DMEM F-12) and serum-free medium (DMEM F-12) with 100 ng/ml CX3CL1 (472-FF-025/CF, R&D systems). 24 h later the cells were fixed by 4% paraformaldehyde, then incubated overnight in primary antibody solution (Goat anti-Iba-1, 1:500, ab5076, Abcam; Rat anti-Ki67, 1:500, 14-5698-82, eBioscience). After washing, the coverslips were incubated for 2 h at room temperature with a secondary antibody (Donkey anti-goat, 1:1000, Alexa Fluor 488, 1942238, invitrogen; Donkey anti-Rat, 1:1000, Alexa Fluor 555, ab150154, Abcam). The rate of ki67 positive cells in Iba-1 positive cells were calculated as the proliferating rate of microglia. For migration assay, isolated CD11b positive cells were seeded at a density of 0.5 × 10 5 in an Incucyte ImageLock 96-well plate (4379, Sartorius) for overnight with the addition of 10% FCS and 20 ng/ml M-CSF. On the following day, a standardized scratched wound was induced in each well simultaneously using an Incucyte WoundMaker kit (Cat. No. 4493, Sartorius), followed by replacement with serum-free, M-CSF free DMEM/F-12 medium containing indicated concentration of CX3CL1 (472-FF-025/CF, R&D systems). The plate was then placed in an Incucyte Zoom System, and live images were taken every 2 h. The images were analyzed according to the manual of Incucyte Zoom.
To confirm that Cx3cr1 high Cre − Eyfp − microglia were not an anomaly of one specific mouse strain, we investigated the microglial populations in Cx3cr1 GFP/wt mice (JAX, 005582) by flow cytometry (gating strategy depicted in Additional file 1: Fig. S3a) and immunohistochemistry. The results revealed that Iba-1 + Tmem119 + GFP − cells with microglial morphology could also be detected in the Cx3cr1 GFP/wt adult mouse brain, constituting 0.59% ± 0.27% of total microglia (Fig. 1i, j).

42) after three consecutive
We next addressed if these newly repopulated Cx3cr1 high Cre − Eyfp − microglia were derived from the periphery. We and others have previously demonstrated that the empty microglial niche can be repopulated within weeks through resident microglia proliferation and concomitant infiltration of monocytes [38,39]. Two novel specific surface markers (Tmem119 and P2ry12) were used to identify CNS-resident microglia [40,41]. Our results demonstrated that all newly repopulated Cx3cr1 high Cre − Eyfp − microglia expressed both Tmem119 (Fig. 2h) and P2ry12 (Fig. 3a). Furthermore, the vast majority of Cx3cr1 high Cre − Eyfp − microglia had low expression of F4/80 (Fig. 2g).
Given the lack of Cre and Eyfp expression on the mRNA and protein levels, we reasoned that the CreER-Eyfp knockin-allele was either epigenetically/ genetically silenced or deleted. We thus investigated Cx3cr1 high Cre − Eyfp − microglia at the DNA level. Genotyping of DNA extracted from Cx3cr1 high Cre − Eyfp − and Cx3cr1 CreER-Eyfp/wt Cre + Eyfp + microglia revealed that Cx3cr1 high Cre − Eyfp − microglia apparently lacked the CreERT2-Eyfp insert (Fig. 4a). This analysis was performed using the primers and protocols Jax lab recommended, which only target part of the CreERT2-Eyfp knockin-allele (Additional file 1: Fig. S6a). We, therefore, further designed primers and optimized protocols to target longer sequences, including the whole Cx3cr1 exon 2 containing the inserted CreERT2-Eyfp fusion gene (Additional file 1: Fig. S6a). The mutant band was not detected in Cx3cr1 high Cre − Eyfp − microglia (Fig. 4b), again indicating that they lack the CreERT2-Eyfp insert. Whole-genome sequencing of the DNA extracted from Cx3cr1 high Cre − Eyfp − and Cx3cr1 CreER-Eyfp/wt Cre + Eyfp + microglia further confirmed the complete absence of the CreERT2-Eyfp knockin-allele in Cx3cr1 high Cre − Eyfp − microglia. Interestingly, the coverage of the integrated Cx3cr1 coding exon in Cx3cr1 high Cre − Eyfp − microglia is twice that of the coverage in the Cx3cr1 CreER-Eyfp/wt Cre + Eyfp + microglia (Fig. 4c- 1: Fig. S6d). We then hypothesized that homologous recombination during microglial mitosis leads to loss of the CreERT2-Eyfp insert. To address this hypothesis, we tested whether breeding Cx3cr1 CreER-Eyfp/CreER-Eyfp Rosa26 DTA/wt mice would result in the loss of the CreERT2-Eyfp insert, given the fact that there is no wild type allele to recombine with. Our results revealed that the EYFP − F4/80 low microglia, here denoted as Cx3cr1 wt/ wt Cre − Eyfp − microglia, were not detected in Cx3cr1 CreER-Eyfp/ CreER-Eyfp Rosa26 DTA/wt mice after genetically microglial depletion and repopulation (Additional file 1: Fig. S6e). These results indicate that homologous recombination is a possible mechanism leading to loss of heterozygosity (LOH) of Cx3cr1 wt/wt Cre − Eyfp − microglia in Cx3cr1 CreER-Eyfp/wt mice.

Discussion
The Cx3cr1 CreER-Eyfp mouse line constitutes a central tool for studying microglia biology [30]. In the present study, we unexpectedly identified the presence of Cx3cr1 wt/wt Cre − Eyfp − microglia lacking CreERT2-Eyfp locus in the Cx3cr1 CreER-Eyfp/wt mouse brain, which can be repopulated after microglial depletion following Tam injections in Cx3cr1 CreER-Eyfp/wt Rosa26 DTA/wt mice.
One previous study reported that all repopulated microglia were derived from CX3CR1 + cells, as measured by fate mapping in Cx3cr1 CreER::Ai14 mice. All CX3CR1 + microglia were tdTomato + , and all Iba-1 + cells were tdTomato + , leading to the interpretation that all the repopulated Iba-1 + cells arose from the surviving CX3CR1 + cells following microglial depletion [43]. However, in both our Cx3cr1 CreER-Eyfp/wt and Cx3cr1 GFP/wt transgenic mice, the presence of Cre − EYFP − and GFP − microglia was recorded. Proliferating GFP − microglia in Cx3cr1 GFP/+ mice have also been reported following microglia depletion using the CSF1 receptor inhibitor PLX3397 [18]. In that study, the authors showed that proliferating GFP − cells could become Iba-1 + microglia; however, whether these Iba-1 + GFP − microglia were also CX3CR1 + was not addressed [18]. Our results indicated that these cells could be de facto microglia expressing all thus far established canonical markers and that they existed before microglial depletion and repopulation. Another study reported that most newly repopulated microglia were tdTomato + using the CSF1 receptor inhibitor PLX5622 in Cx3cr1 CreER/+:tdTomato mice, and the Iba1 + tdTomato − cells were interpreted as being infiltrating peripheral monocytes [26]. However, based on our study, we consider that the Iba1 + tdTomato − cells could be derived from the local microglial cell pool, which does not carry the transgene Cx3cr1 CreER . Furthermore, another study demonstrated that EYFP − microglia escaped Cre-mediated recombination and could repopulate the CNS following microglial depletion in Rosa26 −STOP−Eyfp Cx3cr1 CreER :iDTR mice [39]. Phagocytosis and activation of microglia is, at least partly, dependent on the CX3CR1/CX3CRL axis [44][45][46][47][48][49]. Hence, the EYFP-microglia, expressing twice the amount of CX3CR1, might have a different phagocytosis capacity. In light of these findings, interpretation of results from Cx3cr1 CreER/wt and Cx3cr1 GFP/wt transgenic mice should be made with caution.
We consider two possible explanations for the presence of Cx3cr1 wt/wt Cre − Eyfp − microglia in the Cx3cr1 CreER-Eyfp/wt mouse brain: (i) maternal-derived macrophages or microglia, or (ii) LOH of microglia during mitosis. The possibility of maternal-derived microglia or macrophages was excluded by our results, as the Cx3cr1 wt/wt Cre − Eyfp − microglia inherited the paternal DTA gene. Alternatively, LOH due to homologous recombination could explain the existence of Cx3cr1 wt/wt Cre − Eyfp − microglia in the Cx3cr1 CreER-Eyfp/wt mouse brain. LOH is a phenomenon whereby the cells only possess the genetic information from one of the parental chromosomes, as previously described in the cancer cells [50] and mammalian cells in vivo and in vitro [51][52][53]. Our results showing a similar percentage of Cx3cr1 wt/wt Cre − Eyfp − microglia in each of the heterozygous Cx3cr1 CreER-Eyfp/wt mouse brains and the absence of Cx3cr1 wt/wt Cre − Eyfp − microglia in the homozygous Cx3cr1 CreER-Eyfp/CreER-Eyfp mouse brain support the theory of LOH through homologous recombination during mitosis. The mechanisms of microglial LOH in Cx3cr1 CreER-Eyfp/wt and Cx3cr1 GFP/wt mouse strains and whether the phenomena of LOH also occurs in other analogous genetic modified heterozygous strains need to be further investigated.
Microglial depletion and repopulation studies have expanded our knowledge of microglia in physiological and pathological states, exerting favorable effects in different preclinical disease models [54,55]. Moreover, new microglia rapidly repopulate the brain parenchyma following microglial depletion, although the origin of these newly repopulated microglia has been debated. Elmore and colleagues reported that CX3CR1 + GFP − cells were potential microglial precursor cells during the microglial repopulation period. However, this viewpoint has been challenged, and newly repopulated microglia are proposed to only arise from the surviving microglia [56]. This is further supported using fate mapping approaches showing that The new forming microglia only temporary expression nestin, and no microglia were derived from None microglia nestin + cells [43]. Our RNA sequencing data revealed no increased expression of precursor of stem cell genes in Cx3cr1 wt/wt Cre − Eyfp − microglia. Moreover, microglial single-cell data from wild type mice revealed no high Cx3cr1-expressing microglia that clustered separately from other microglia. Rather, the cells expressing higher levels of Cx3cr1 were evenly distributed (Additional file 1: Fig. S3C, D). Taken together, these data support that Cx3cr1 wt/wt Cre − Eyfp − microglia are not precursors or stem cells. How these surviving microglia could survive depletion and whether they were unaffected by the depletion period remain unclear. Microglia cannot actually be fully depleted using currently available depletion models [54] and the existence of a microglial subset that may be resistant to depletion has been previously proposed [57]. Cx3cr1 wt/wt Cre − Eyfp − microglia could be one such subset, at least in the Cx3cr1-Cre derived depletion setting. Our results indicate that both unaffected Cx3cr1 wt/wt Cre − Eyfp − microglia and surviving Cx3cr1 CreER-Eyfp/wt Cre + Eyfp + microglia repopulate the brain competitively, with contribution from peripherally derived macrophages. Moreover, it is widely known that individual microglia occupy non-overlapping spatial territories [58,59]. Herein, we report a novel finding that non-overlapping niches exist between repopulated Cx3cr1 wt/wt Cre − Eyfp − and Cx3cr1 CreER-Eyfp/wt Cre + Eyfp + microglia; however, related signal pathways need to be further investigated.
Tissue macrophages with a self-renewing capacity are seeded during embryonic development, and some macrophages can be replaced or renewed postnatally by peripheral monocytes [60][61][62][63][64][65]. In the brain, the resident microglial pool is self-renewing without contribution from peripheral monocytes during the whole life span [1]. However, in disease states, such as CNS injury and neurodegenerative diseases, monocytes can infiltrate the brain and become microglia-like cells, albeit with different functionalities [66]. Here we used a mouse model in which peripheral monocytes entered the brain after microglial depletion [38]. Our results demonstrate that the main transcriptomic difference is noted between the F4/80 high and F4/80 low groups, but not between the EYFP − and EYFP + groups after competitive microglial repopulation, which indicates that the infiltrating monocytes, imprinted by the CNS microenvironment, are different from the resident repopulated microglia, and confirming our previous publication [38].
Microglia replacement therapy has been proposed for CNS diseases linked to microglial dysfunctions or gene mutations. Replacing microglia by genetically modified or engineered cells may hold promise for distinct CNS diseases, yet the factors regulating competitive engraftment of different populations including microglia, monocytes and engineered cells are poorly understood [67,68]. CX3CR1 regulates microglia colonization and distribution in the brain [69], and Cx3cr1 gene-deleted mice exhibited lower microglial density in the developing brain [70]. The CX3CL1-CX3CR1 axis also regulates microglial repopulation following microglial depletion in the mouse retina [26], but whether this axis regulates competitive microglial repopulation in the brain has not previously been addressed. Our results indicate that repopulating microglia originate from three different predominant sources, resident microglia, including both Cx3cr1 wt/wt Cre − Eyfp − and Cx3cr1 CreER-Eyfp/wt microglia, competing with infiltrating peripheral-derived microglia-like cells. Our data demonstrate that Cx3cr1 wt/wt Cre − Eyfp − microglia have a competitive advantage over Cx3cr1 CreER-Eyfp/wt microglia. Furthermore, resident microglia lacking Cx3cr1 (from Cx3cr1 CreER-Eyfp/CreER-Eyfp mice) were unable to compete with the peripheral-derived microglia-like cells following microglial depletion. This indicates that the resident microglia repopulation, but not peripheral derived microglia-like cell repopulation relies on CX3CR1. The proliferation rate of microglia in vitro was decreased in Cx3cr1 gene-depleted primary microglia, and this was not affected when the cells were challenged with CX3CL1 protein. This likely explains the repopulation advantage of Cx3cr1 wt/wt Cre − Eyfp − over Cx3cr1 CreER-Eyfp/wt microglia post-depletion, at least partly. Cx3cr1 deficiency decreased the microglial migration rate, consistent with previous studies reporting that Cx3cr1 deficiency impaired microglia migration in vivo [71,72] and in vitro [73]. However, microglial migration was not impaired in Cx3cr1 CreER-Eyfp/wt microglia. CX3CL1 increased migration rates of Cx3cr1 wt/wt Cre − Eyfp − microglia but not of Cx3cr1 CreER-Eyfp/wt and Cx3cr1 CreER-Eyfp/CreER-Eyfp microglia, which indicates that CX3CL1-CX3CR1 regulates microglial migration. The migration rate is higher in Cx3cr1 wt/wt Cre − Eyfp − than Cx3cr1 wt/wt Cre − Eyfp − microglia after adding CX3CL1, which further explains the repopulation advantage of Cx3cr1 wt/wt Cre − Eyfp − over Cx3cr-1 CreER-Eyfp/wt microglia post-depletion. Taken together, we conclude that the CX3CL1-CX3CR1 axis is important for the resident microglial repopulation and for competition with peripheral monocyte-derived microglia-like cells. Thus, limiting residential microglia repopulation by inhibiting CX3CL1-CX3CR1 signaling improves the microglial replacement efficiency by peripheral derived monocytes.

Conclusions
A small portion (less than 1%) of Cx3cr1 wt/wt Cre − Eyfp − microglia do not carry genetic labels in the widely used Cx3cr1 GFP/wt or Cx3cr1 CreER-Eyfp/wt mouse models. Not being aware of this population may lead to significant data misinterpretation since these cells may escape detection (not carrying the Eyfp or Gfp) and cannot be modified (lacking Cre expression) as expected. We further demonstrate the important role of the CX3CL1-CX3CR1 axis in regulation of microglial repopulation post-depletion. These findings raise an important cautionary note, not only when using the strains mentioned above but also for other strains that might display similar phenomena.
Additional file 1: Figure S1. Additional data related to Fig. 1. A Representative 8 sections of Hoechst staining of Cx3cr1 CreER-Eyfp/wt mouse brain sequential sagittal slices, the numbers representing the location and number of EYFP − microglia in that location, each color represents one mouse, n = 6. Figure S2. Morphology analysis. A The process to prepare topological skeleton from the original photomicrographs. The full-size maximum intensity projection images (top) were processed into binary images (middle) and then skeletonized (bottom) following the ImageJ plugin protocol. Cropped images (right) from the original full-size images (left) were shown to improve the visualization. Scale bar, 50 μm. B The Analyze Skeleton plugin was applied, and skeletonized endpoints were tagged purple, the slab is orange, and the junction is pink. The tagged data are summarized as total length (sum of endpoints, slab, and junction) and the total number of endpoints. C The process area is represented as the convex hull area by connecting the process ends using the ImageJ polygon tool. Figure S3. Additional data related to Fig. 1. A Gating strategies of flow cytometry analyses. B EYFP − microglia ratio of total microglia at 3 weeks, 4 weeks, 6 weeks, 9 weeks and 15 weeks old mice, n = 4, 4, 6, 4, respectively, mean ± s.d. No significant difference by one-way ANOVA. C UMAP plots from single-cell sequencing, each dot represents one cell, the color represents the expression levels of Cx3cr1. D Histogram plot of Cx3cr1. The x-axis represents Cx3cr1 expression read counts; the y-axis represents cell numbers. Figure S4. Additional data related to Fig. 2. A Graph showing relative Iba-1 + microglia at 1 day, 3 days and 7 days after the final Tam injection, respectively; n = 3-4, mean ± s.d. *p < 0.05, **p < 0.01, ***p < 0.001 by Student's two-tailed unpaired t test. B Representative images of Iba-1 microglia staining of Cx3cr1 CreER-Eyfp/wt mice and Cx3cr1 CreER-Eyfp/wt Rosa26 DTA/wt mice at 1 day after the final Tam injection in cortex and cerebellum. Scale bar, 200 μm. C Representative 3 sequential sagittal sections (25 µm/section, with 36 sections interval) after Hoechst staining of Cx3cr1 CreER-Eyfp/wt Rosa26 DTA/wt mice brains. The numbers represent the location and number of Cx3cr1 high Cre − Eyfp − microglia at 1 day and 3 days after the final Tam injection, each color represents one mouse, n = 3. D Quantitative data showing the total number of microglia in the 3 sections in Ctrl group, 1 day and 3 days after the final Tam injection, n = 4, 3, 3, mean ± s.d. Figure S5. Additional data related to Fig. 4. A-C Bar graphs showing RPKM of Cre, EYFP, and Cx3cr1, respectively, n = 5, mean ± s.d. ***p < 0.001 by Student's two-tailed unpaired t test. D The expression of CX3CR1 in both Cx3cr1 CreER-Eyfp/wt Cre + Eyfp + and Cx3cr1 high Cre − Eyfp − microglia in Cx3cr1 CreER-Eyfp/wt Rosa26 DTA/wt mice at day 42 after the final Tam injection. E Bar graph showing the mean fluorescence intensity (MFI) of CX3CR1, n = 4, mean ± s.d. ***p < 0.001 by Student's two-tailed unpaired t test. Figure S6. Additional data related to Fig. 4. A Schematic diagram showed how the primers were designed. The red color indicates the common primers used in all the PCR reactions; the green and blue color indicate the WT primer and Mutant primer used in PCR reaction for Fig. 4f; the orange color indicates the long PCR product targeting primer used in PCR reaction for Fig. 4g