Intrinsic DNA damage repair deficiency results in progressive microglia loss and replacement

Abstract

The DNA excision repair protein Ercc1 is important for nucleotide excision, double strand DNA break, and interstrand DNA crosslink repair. In constitutive Ercc1‐knockout mice, microglia display increased phagocytosis, proliferation and an enhanced responsiveness to lipopolysaccharide (LPS)‐induced peripheral inflammation. However, the intrinsic effects of Ercc1‐deficiency on microglia are unclear. In this study, Ercc1 was specifically deleted from Cx3cr1‐expressing cells and changes in microglia morphology and immune responses at different times after deletion were determined. Microglia numbers were reduced with approximately 50% at 2–12 months after Ercc1 deletion. Larger and more ramified microglia were observed following Ercc1 deletion both in vivo and in organotypic hippocampal slice cultures. Ercc1‐deficient microglia were progressively lost, and during this period, microglia proliferation was transiently increased. Ercc1‐deficient microglia were gradually replaced by nondeficient microglia carrying a functional Ercc1 allele. In contrast to constitutive Ercc1‐deficient mice, microglia‐specific deletion of Ercc1 did not induce microglia activation or increase their responsiveness to a systemic LPS challenge. Gene expression analysis suggested that Ercc1 deletion in microglia induced a transient aging signature, which was different from a priming or disease‐associated microglia gene expression profile.

Main Points

• The DNA repair protein Ercc1 is essential for microglia

• Microglia‐specific deletion of Ercc1 transiently induces an aging phenotype

• Ercc1‐deficient microglia are gradually lost and replaced by non‐deficient microglia

1. INTRODUCTION

Organisms have developed intricate mechanisms to repair different forms of DNA damage, for example, base excision repair (BER), nucleotide excision repair (NER), mismatch repair (MMR), interstrand crosslink repair (ICR), and double‐strand break repair (DBR). Repair of DNA damage is important as persistent damage induces cell senescence or cell death (Rodier et al., 2009). Deficiencies in DNA repair pathways lead to various progeria syndromes both in human and mouse (Schumacher, Garinis, & Hoeijmakers, 2008). In many progeria syndromes, neurologic defects occur, which highlights the vital role of genome stability maintenance for the central nervous system (CNS; Heng, Eggen, Boddeke, & Kooistra, 2017; McKinnon, 2013).

Excision repair cross‐complementation group 1 (ERCC1) in complex with XPF is an essential nuclease in the NER, ICR and DBR pathways (Ahmad et al., 2008; Klein Douwel et al., 2014; Li et al., 2019). Besides their role in DNA damage repair, nucleotide excision repair factors like ERCC1 and XPF are recruited to active promoters to facilitate transcription (Le May et al., 2010). Mutations in ERCC1 and XPF cause cerebro‐oculo‐facioskeletal (COFS) syndrome, Cockayne syndrome (CS), and xeroderma pigmentosum (XP) in humans (Gregg, Robinson, & Niedernhofer, 2011; Kashiyama et al., 2013). Mice carrying a knockout (ko) and a hypomorphic (Δ) allele for Ercc1 (Ercc1 ^Δ/ko) display a range of progeroid changes, including reduced lifespan, loss of the body weight, and various aging‐related pathological changes in peripheral organs (Dolle et al., 2011). In addition, constitutive Ercc1‐knockout (Ercc1 ^Δ/ko) mice display premature CNS aging, such as motor abnormalities and cognitive decline, widespread astrogliosis, microgliosis and neuronal degeneration in the brain, and progressive motor neuron loss in the spinal cord (Borgesius et al., 2011; de Waard et al., 2010; Vegh et al., 2012). Microglia in Ercc1 ^Δ/ko mice exhibit a hypertrophic morphology with thickened primary processes and an increase in soma size (Raj et al., 2014). Functionally, microglia in Ercc1 ^Δ/ko mice show increased phagocytosis, proliferation and reactive oxygen species (ROS) production. Notably, microglia in Ercc1 ^Δ/ko mice are primed (Cunningham, 2013), indicated by an enhanced proinflammatory response to a systemic peripheral inflammatory challenge (intraperitoneal lipopolysaccharide [LPS] injection; Raj et al., 2014). Specific deletion of Ercc1 in forebrain neurons (Camk2 ^wt/cre :Ercc1 ^ko/loxP) showed that neuronal genotoxic stress was sufficient to induce microglia priming (Raj et al., 2014).

We previously reported that a microglia priming gene expression signature is shared between Ercc1 ^Δ/ko mice, Alzheimer's disease (AD), and amyotrophic lateral sclerosis (ALS) mouse models and naturally aged mice (Holtman et al., 2015). This profile is characterized by the upregulation of genes associated with immune, phagosome, and antigen presentation pathways and the downregulation of homeostatic microglia genes (Holtman et al., 2015). Similarly, microglia in CNS disease mouse models exhibit increased phagocytic and immune activity and are referred to as disease‐associated microglia (DAM) or microglia in neurodegenerative disease (MGnD; Butovsky & Weiner, 2018; Keren‐Shaul et al., 2017; Krasemann et al., 2017; Mathys et al., 2017). Where constitutive Ercc1 deletion induces microglia priming, the effect of microglia‐specific Ercc1‐deficiency is unclear. In this study, the Ercc1 gene was deleted from microglia and the effect on microglia density, morphology, survival, proliferation, and responsiveness to LPS‐induced inflammation were determined.

2. MATERIALS AND METHODS

2.1. Animals

Cx3cr1 ^wt/creERT2 (JAX stock #021160), Ercc1 ^wt/ko and Ercc1 ^loxP/loxP mouse lines were crossed to obtain the experimental lines Cx3cr1 ^wt/creERT2 :Ercc1 ^ko/loxP and Cx3cr1 ^wt/creERT2 :Ercc1 ^wt/loxP (Figure S1a). In brief, Cx3cr1 ^wt/creERT2:Ercc1 ^wt/ko mice were generated by crossing Ercc1 ^wt/ko and Cx3cr1 ^wt/creERT2 mice (both in a C57BL/6 background). Cx3cr1 ^{creERT2/creERT2}:Ercc1 ^wt/ko were crossed with Ercc1 ^loxP/loxP mice (FVB background), resulting in Cx3cr1 ^wt/creERT2 :Ercc1 ^ko/loxP and Cx3cr1 ^wt/creERT2 :Ercc1 ^wt/loxP mice (Figure S1a, hereafter referred to as Cx3cr1‐Ercc1 ^ko/loxP and Cx3cr1‐Ercc1 ^wt/loxP mice). The mice were group‐housed with 2–4 same‐sex littermates per cage under 12‐hour light/dark cycle conditions and ad libitum access to food and water. All experiments were performed in the Central Animal Facility (CDP) of the UMCG, with protocol (15360‐03‐002) approved by the Animal Care and Use Committee of the University of Groningen.

2.2. Genotyping

Genomic DNA was isolated from ear cuts for genotyping with MyTaq Extract‐PCR Kit (Bioline, BIO‐21127). Primer information is provided in Table S1. Cx3cr1‐cre was genotyped by PCR with cre primers (Table S1). Schematic representation of Ercc1 ^wt, Ercc1 ^ko, Ercc1 ^loxP, and recombined Ercc1 ^rec alleles are depicted in Figure S1b. The Ercc1 ^ko allele consisted of a neo cassette insertion interrupting Exon 7, aborting the essential carboxy‐terminal 74 amino acids of Ercc1 (Weeda et al., 1997). In the Ercc1 ^loxP allele, Exons 3–5 are flanked by loxP sites (Doig et al., 2006). After tamoxifen treatment, Exons 3–5 in the Ercc1 ^loxP allele will be deleted by homologous recombination, resulting in a recombined Ercc1 ^rec allele. In Ercc1 ^wt/ko and Cx3cr1‐Ercc1 ^wt/ko mouse line, genotyping of the Ercc1 ^wt and Ercc1 ^ko was done by duplex PCR using wt and neo primer pairs (Table S1) as described before (Ahmad et al., 2008). For Cx3cr1‐Ercc1 ^ko/loxP and Cx3cr1‐Ercc1 ^wt/loxP mouse lines, a duplex PCR was performed to distinguish Ercc1 ^wt, Ercc1 ^ko, and Ercc1 ^loxp alleles using loxp and neo primer pairs (Table S1).

After tamoxifen treatment, the Ercc1 ^loxP alleles in both Cx3cr1‐Ercc1 ^wt/loxP and Cx3cr1‐Ercc1 ^ko/loxP mice microglia are recombined (Ercc1 ^rec) (Figure S1c). This recombination results in full Ercc1 deletion in Cx3cr1‐Ercc1 ^ko/loxP mice microglia, but only partial deletion in Cx3cr1‐Ercc1 ^wt/loxP mice microglia, since one Ercc1 ^wt allele is still present (Figure S1c). To confirm the specific deletion of Ercc1 in microglia, all mice were genotyped by genomic PCR on sorted microglia (Figure S1d). Mice of 6–8 weeks of age received tamoxifen to induce Ercc1 gene recombination. At certain time points post tamoxifen treatment, microglia were FACS‐sorted and DNA was isolated from these microglia for genotyping. A duplex PCR was performed to distinguish Ercc1 ^wt, Ercc1 ^ko, and Ercc1 ^loxP allele using loxp and neo primer pairs (Table S1), and Ercc1 ^rec was genotyped using rec primer pair (Table S1). The Ercc1 ^wt (~1,000 bp), Ercc1 ^ko (~800 bp), and Ercc1 ^loxP (~500 bp) products were separated by electrophoresis on 1.2% agarose gels, and the Ercc1 ^rec allele generated a ~280 bp PCR product (Figure S1d).

2.3. Administration of tamoxifen

Mice of 6–8 weeks of age received 2 doses of 500 μl of tamoxifen (20 mg/ml, Sigma‐Aldrich, T5648) dissolved in corn oil (Sigma‐Aldrich, C8267) via oral gavage with a 48 hr interval as described previously (Parkhurst et al., 2013).

2.4. LPS treatment

Mice were given an intraperitoneal (i.p.) injection of 1 mg/kg LPS (Sigma‐Aldrich, Escherichia coli 011:B4, L4391) dissolved in Dulbecco's phosphate buffered saline (DPBS, Lonza, BE17512F). Control mice received a respective volume of DPBS. After 3 hr, animals were perfused with saline or PBS under deep anesthesia and the brains were collected.

2.5. Microglia isolation and flow cytometry

Microglia were isolated as described in our previous work (Gerrits, Heng, Boddeke, & Eggen, 2020). Mice were perfused with saline or PBS under deep anesthesia. Brains were placed in Hank's balanced salt solution (HBSS, Gibco, 14170‐088) with 0.6% glucose (Sigma‐Aldrich, G8769) and 15 mM HEPES (Lonza, BE17‐737E). All the following isolation procedures were performed on ice or at 4°C during centrifugation. Brains were mechanical dissociated using the Potter‐Elvehjem tissue homogenizer and centrifuged at 220g for 10 min. The pellets were resuspended in 25 ml 24% Percoll (GE Healthcare, 17–0891‐01) with a 3 ml PBS layer on top, followed by centrifugation for 20 min at 950g (accelerate 4 and brake 0) to remove myelin. The cell pellets were incubated with CD11b‐PE (eBioscience, 12–0112‐82), CD45‐PE/Cy7 (eBioscience, 25–0451‐82), and Ly‐6C‐APC (Biolegend, 128015) antibodies for 20–30 min on ice. Then the cells were washed once and filtered into FACS tubes. Microglia were FACS‐sorted as DAPI^negCD11b^high CD45^int Ly6c^neg events on sorter MoFlo‐Astrios or MoFlo‐XDP (Beckman Coulter).

Ki67 staining was performed according to the manufacturer's protocol. In brief, the cell pellets after the Percoll gradient were permeabilized by adding 1 ml cold 70% ethanol drop by drop while vortexing. After 1 h incubation at −20°C, the cells were twice washed with 1 ml PBS with 10% FBS and incubated with CD11b‐PE (eBioscience, 12‐0112‐82), CD45‐FITC (eBioscience, 11‐0451‐85), Ly‐6C‐APC/Cy7 (Biolegend, 128025), and Ki67‐Alexa Fluor® 647 antibody (Biolegend, 652407) for 30 min. Ki67⁺ microglia and Ki67⁻ microglia were collected.

2.6. Genomic DNA and RNA isolation from microglia

The AllPrep DNA/RNA Micro Kit (Qiagen, 80284) was used to extract genomic DNA and total RNA from sorted microglia.

2.7. Quantification of recombination efficiency in bulk microglia by quantitative real‐time PCR (qPCR)

To investigate recombination efficiency, DNA from microglia from Cx3cr1‐Ercc1 ^ko/loxP and Cx3cr1‐Ercc1 ^wt/loxP mice was analyzed by qPCR. Two primer sets were designed, based on the recombined Ercc1 ^rec allele: Ercc1‐rec1 and Ercc1‐rec2 (Table S1). Without recombination, the primer sets will not give products during the PCR due to the long span between the forward and reverse primer. In addition, a reference primer pair, Ref‐Il1, which amplifies a genomic fragment of the Il1 gene was included (Table S1). The PCR reaction mixture contained 5 μl DNA template from microglia samples, 5.5 μl iTaq™ Universal SYBR® Green Supermix (Bio‐Rad, 1725125), 0.3 μl ddH₂O and 0.2 μl 10 μM primer mix. Each sample was quantified with three technical replicates. The percentage of microglia with a recombinant Ercc1 ^rec allele was calculated by the following formula (Livak & Schmittgen, 2001).

$$\text{Percentage of microglia with}\mathit{\text{Ercc}}{1}^{\mathit{rec}}\text{allele}=\frac{\text{microglia}\left(\mathit{\text{Ercc}}{1}^{\mathit{rec}}\right)}{\text{microglia}\left(\mathit{\text{Ercc}}{1}^{\mathit{rec}}\right)+\text{microglia}\left(\mathit{\text{Ercc}}{1}^{\mathit{\text{loxP}}}\right)}=2^{-\left(\frac{\mathrm{Ct}\left(\mathit{\text{Ercc}}1-\mathit{rec}1\right)+\mathrm{Ct}\left(\mathit{\text{Ercc}}1-\mathit{rec}2\right)}{2}-\mathrm{Ct}\left(\mathit{Ref}-\mathit{Il}1\right)-1\right)}$$

2.8. Quantification of recombination efficiency by single‐cell qPCR (genomic DNA)

Individual microglia were FACS‐sorted in 384‐well PCR plates containing 5 μl ddH₂O in each well. The presence of the Ercc1 ^rec allele was determined using qPCR primer pair Ercc1‐rec1. As a positive control, individual microglia were analyzed using Ref‐Il1 primers. As negative controls, individual splenic macrophage (DAPI^negCD11b^highCD45^posLy6g^neg) were sorted in 384 well plates and analyzed. 5.5 μl of iTaq™ Universal SYBR® Green Supermix, 0.3 μl ddH₂O and 0.2 μl 10 μM primer mix were added to each well. Quantitative PCR reactions were performed using the QuantStudio 7 Real‐Time PCR system (Thermo Scientific). In the end, the number of PCR reactions resulting in specific DNA products from Ercc1 ^rec (with correct melting curves) were quantified. The percentage of microglia with recombinant Ercc1 ^rec allele was calculated by the following formula.

$$\text{Percentage of microglia with}\mathit{\text{Ercc}}{1}^{\mathit{rec}}=\frac{\text{number of}\mathrm{PCR}\text{reactions with products from}\mathit{\text{Ercc}}{1}^{\mathit{rec}}\text{allele}}{\text{total number of}\mathrm{PCR}\text{reactions}}$$

2.9. cDNA synthesis and qPCR

RNA isolated from microglia was mixed with 1 μl random primers (0.5 μg/μl, Invitrogen, 48190011) and ddH₂O to 10 μl. Samples were incubated at 65°C for 15 min and kept on ice. Thereafter, 8 U/μl M‐MuLV reverse transcriptase (Thermo Scientific, EP0442), 0.8 U/μl Ribolock RNase inhibitor (Thermo Scientific, EO0382), 0.5 mM dNTP‐mix (Thermo Scientific, R0192) and reverse transcriptase buffer were added and incubated on a thermal cycler at 42°C for 1 hr, at 70°C for 10 min and finally at 4°C. The resulting cDNA was used for qPCR reactions. The PCR reaction mixture contained 5 μl cDNA template from microglia samples, 5.5 μl iTaq™ Universal SYBR® Green Supermix, 0.3 μl ddH₂O and 0.2 μl 10 μM primer mix. Each sample was run with three technical replicates. Quantitative PCR reactions were performed using the ABI7900HT Fast Real‐Time PCR System (Thermo Scientific), LightCycler® 480 System (Roche) or QuantStudio 7 Real‐Time PCR system (Thermo Scientific). To determine relative expression levels, Hprt1 was used as the reference gene. Primer sequences are provided in Table S2.

2.10. QuantSeq 3′ mRNA‐Sequencing and bioinformatic analysis

RNA quantity and quality were analyzed on a Fragment Analyzer (Agilent), only RNA samples with a RIN value >6.5 were used. Sequencing libraries were prepared with the Quant Seq 3′ mRNA‐Seq Library Prep Kit FWD (Lexogen, 015.96). Quality control of the raw FASTQ files was performed with FASTQC. Bad quality bases were trimmed with FASTX_trimmer of the FASTX_toolkit (version 0.013). Sequences were aligned using default parameters on HiSAT2 version 2.1 to the M. musculus (GRCm38.85) reference template obtained from Ensembl. Quantification of the reads was performed with HTseq‐counts (version 0.6.1). Raw count matrices were loaded in R and processed with DESeq2. Genes were identified as differentially expressed with an FDR < 0.05 and fold change >1.5. Normalized values (counts per million) of the differentially expressed genes were used as heatmap input. Gene ontology (GO) term enrichment analysis was performed using Metascape (http://www.metascape.org/).

2.11. Immunohistochemistry and immunofluorescence

To collect brain tissue for immunostaining, mice were perfused with saline under deep anesthesia. Brains were fixed for 48 hr in 4% paraformaldehyde (PFA) at 4°C. After dehydration in 25% sucrose, the brain samples were embedded with O.C.T. compound (Sakura Finetek, 4583) and stored at −80°C.

For immunohistochemistry, 16 μm sections were prepared by cryo‐sectioning. After washing thrice with 1× PBS (identical for all subsequent washing steps), antigen retrieval was performed by pressure cooking in 10 mM sodium citrate, pH 6.0. The sections were washed and incubated in PBS with 1% hydrogen peroxide (H₂O₂) to block endogenous peroxidases. Again, the sections were washed and blocked for 30 min using 5% normal donkey serum (NDS; Jackson Immuno Research, 017‐000‐121) in PBS with 0.3% Triton X‐100 (PBS⁺). Afterward, the sections were incubated with the primary rabbit‐α‐ionized calcium‐binding adapter molecule 1 (Iba1) antibody (1:1,000; Wako, 01‐19741) overnight at 4°C. The following day, the slides were washed and incubated with the biotinylated secondary donkey‐α‐rabbit IgG antibody (1:400; Jackson Immuno Research, 711‐065‐152) for 1 hr. After washing, the sections were incubated with ABC solution (VECTASTAIN® ABC Kit, Vector Laboratories, PK‐6100) for 30 min. The sections were washed, stained using 0.04% 3,3′‐Diaminobenzidine (DAB) and 0.01% H₂O₂ for 8 min and subsequently dehydrated using a sequence of increasing ethanol concentrations. The slides were air dried for 30 min, mounted with coverslips using DePex (Serva) and stored at room temperature. All the slides were scanned with the NanoZoomer 2.0‐HT Digital Pathology system (Hamamatsu Photonics, K.K., Japan) at 40 times magnification.

For immunofluorescence, free‐floating brain sections were immunolabeled as described (Sierra et al., 2010). For organotypic hippocampal slice culture, slices were blocked for 1 hr with 5% normal donkey serum and thereafter incubated with a primary antibody against Iba1 (1:1,000; Wako, 019‐19741) overnight at 4°C. On the next day, after washing thrice with 1× PBS, Alexa Fluor 488 donkey anti‐rabbit (1:400; Invitrogen, A21206) secondary antibody was added. After 1 hr of secondary antibody incubation, sections were washed and incubated in Hoechst solution (1 μg/ml, Sigma‐Aldrich, 14530) for 5 min. After washing, the slides were mounted with Mowiol mounting medium on glass slides. Image acquisition was performed using a Leica SP8 confocal microscope system (TCS SP8, Leica Microsystems).

2.12. Microglia density and spatial distribution analysis

Microglia densities in the frontal cortex and in cornu ammonis (CA), and dentate gyrus (DG) were determined by counting all Iba1‐positive cells in a specified region of interest (ROI) of known dimensions using the cell counter plugin for the ImageJ software (http://rsb.info.nih.gov/ij/). To assess the spatial distribution of microglia in the frontal cortex, the nearest neighbor distances—that is, the average Euclidian distances between nearest cells—were determined using the NND plugin for ImageJ. 2–3 ROIs were selected per animal per group for the analysis. Three mice per group were used except the 22 months (n = 2).

2.13. Morphometric analysis of microglia

A pipeline was developed to analyze morphological changes in microglia (Van Weering et al., in prep.). Briefly, single‐cell images of iba1‐positive cells were first extracted from the whole slide scans, with at least 20 cells per region per animal. Prior to analysis, the single cell images were preprocessed to cell silhouette images by semi‐automated thresholding. Subsequently, the cell silhouettes were converted to cell skeleton images by repeated thinning and pruning of the branch areas. In the cell skeleton, branch endings (end nodes), branch crossings (junctions), and all branch points emanating from the cell soma (start nodes) were tagged to allow node quantification. Both cell silhouette‐ and cell skeleton images served as input for fully automated morphometric analysis. The outputs of the pipeline included Sholl analysis result and morphometric features per cell. A specified list of morphometric features, as well as a detailed description of the morphometrics pipeline is described elsewhere (van Weering et al., in prep.).

2.14. Clustering of microglia based on morphometric features

To identify groups of microglia of similar morphology, a nonsupervised clustering approach was applied (described in detail in van Weering et al., in prep.). In brief, after normalization and scaling of all morphometric features, a principal component analysis (PCA) was applied to reduce dimensionality and redundancy in the dataset. Subsequently, a hierarchical clustering (Ward's method) was performed on the top contributing principal components (PCs) with an eigenvalue >1 (here, PC1–4, Figure S2a), resulting in nine clusters of microglia with distinct morphological properties (Figure S2b). The morphometric properties of each cluster are depicted in Figure S2d.

2.15. Organotypic hippocampal slice culture

Organotypic hippocampal slice culture (OHSCs) were prepared as described previously (Stoppini et al., 1991) with minor modifications. In brief, brains were rapidly isolated from Cx3cr1‐Ercc1 ^ko/loxP and Cx3cr1‐Ercc1 ^wt/loxP mouse pups (p3) after decapitation. The hippocampi from both hemispheres were isolated in ice cold serum‐free HBSS supplemented with 0.5% glucose and 15 mM HEPES. Isolated hippocampi were cut into 375 μM thick slices using a tissue chopper (McIlwain) and were transferred to 0.4 μm culture plate inserts (Millipore, PICM03050). These culture plate inserts, containing 6 slice cultures each, were placed in 6‐well plates containing 1.2 ml of culture medium per well. Culture medium (pH 7.2) consisted of 50% minimum essential medium supplemented with 25% heat‐inactivated horse serum (Gibco, 16050‐122), 25% basal medium eagle, 2 mM glutamax and 0.65% glucose. The slice cultures were kept at 35°C in a humidified atmosphere (5% CO2). On the first day after preparation, OHSCs were treated with 1 nM 4‐hydroxy tamoxifen (Sigma‐Aldrich, T176) for 48 hr to induce Ercc1 deletion. OHSCs were kept for up to 3 months and the culture medium was refreshed every 2 days. After fixation with 4% PFA overnight at 4°C, OHSCs were processed for immunofluorescence staining.

2.16. Quantification and statistical analysis

Statistical significance was determined by either a two‐way ANOVA followed by Bonferroni correction or a two‐tailed Student's t‐test as indicated in the legends. For the morphometrics data, after hierarchical clustering, a Kruskal–Wallis test followed by a Wilcoxon rank sum test with Bonferroni correction was performed for comparison of morphometric features between microglia clusters. Statistical differences with p values lower than 0.05 were considered significant.

3. RESULTS

3.1. Microglia are progressively lost after Ercc1 deletion

Ercc1 is an essential endonuclease component in NER, ICR, and DBR, and microglia will accumulate DNA lesions after Ercc1 deletion. Cell cycle arrest, DNA repair, and apoptosis are the general responses to DNA damage (Norbury & Zhivotovsky, 2004). To obtain conditional Ercc1‐deficient mice, Ercc1 ^wt/ko and Cx3cr1 ^wt/creERT2 mice were crossed to generate Cx3cr1 ^wt/creERT2:Ercc1 ^wt/ko mice (Figure S1a). Then, Cx3cr1 ^{creERT2/creERT2}:Ercc1 ^wt/ko were crossed with Ercc1 ^loxP/loxP mice, resulting in Cx3cr1‐Ercc1 ^ko/loxP, and Cx3cr‐Ercc1 ^wt/loxP mice (Figure S1a). After tamoxifen‐induced nuclear translocation of CreER, the Ercc1 ^loxP alleles in both Cx3cr1‐Ercc1 ^wt/loxP and Cx3cr1‐Ercc1 ^ko/loxP mice microglia were recombined (Ercc1 ^rec) (Figure S1b,c). This recombination resulted in Ercc1 deficiency in Cx3cr1‐Ercc1 ^ko/loxP mice microglia, but not in Cx3cr1‐Ercc1 ^wt/loxP mice microglia, since one Ercc1 ^wt allele was still present (Figure S1b,c).

To determine the effect of Ercc1 deletion on microglia, first, the effect on microglia cell density was determined in the frontal cortex (Figure 1a). From 2 months after tamoxifen treatment onward, the density of microglia in the frontal cortex of Cx3cr1‐Ercc1 ^ko/loxP mice was significantly lower than in littermate controls, and this reduction persisted until 12 months after tamoxifen treatment (Figure 1b). A reduction in microglia density after Ercc1 deletion was also observed in other brain regions. In the DG and CA, a significant reduction was observed at 6 and 12 months after tamoxifen treatment (Figure 1b). At 22 months after tamoxifen treatment, microglia density in Cx3cr1‐Ercc1 ^ko/loxP mice was similar to control littermates in all brain regions investigated (Figure 1b). Together with the reduction in microglia density, a significant increase was observed in the nearest neighbor distance between microglia in the frontal cortex of Cx3cr1‐Ercc1 ^ko/loxP mice, suggesting that the observed microglia loss occurred throughout the brain and was not regional (Figure 1c). Similar to our histological data, the number of FACS‐sorted microglia from Cx3cr1‐Ercc1 ^ko/loxP mice was significantly lower than from littermate controls. This reduction in the number of isolated microglia persisted from 2 to 12 months after tamoxifen treatment (Figure 1d).

FIGURE 1.

Reduced microglia numbers in Cx3cr1‐Ercc1 ^ko/loxP mice after tamoxifen treatment. (a) Representative Iba1 staining of microglia in the frontal cortex of Cx3cr1‐Ercc1 ^wt/loxP and Cx3cr1‐Ercc1 ^ko/loxP mice at different time points after tamoxifen treatment. (b) Box plots depict the number of Iba1‐positive cells/mm² in three different brain regions. Two to three sections per animal were used for density analysis. n = 3 mice per group except the 22 months (n = 2). The box boundaries represent first and third quartiles and the center lines indicate the median. Whiskers extend from box boundaries to the minimum and maximum values, respectively. (c) Nearest neighbor distance analysis across groups at different time points after tamoxifen treatment. Two to three sections per animal were used for the analysis. n = 3 mice per group except the 22 months (n = 2). (d) Boxplot depicts the number of microglia (DAPI^negCD11b^highCD45^intLy‐6C^neg) sorted from the entire mouse brain. For Cx3cr1‐Ercc1 ^wt/loxP mice: n = 7 (1 m), 10 (2 m), 6 (3 m), 4 (5 m), 17 (6 m), 6 (9 m), 4 (10 m), 12 (12 m); Cx3cr1‐Ercc1 ^ko/loxP mice, n = 10 (1 m), 13 (2 m), 6 (3 m), 4 (5 m), 15 (6 m), 5 (9 m), 4 (10 m), 13 (12 m). A two‐way ANOVA followed by a Bonferroni correction for multiple comparisons was performed to assess significance. *, p < 0.05; **, p < 0.01; ***, p < 0.001

In addition, we generated OHSCs from Cx3cr1‐Ercc1 ^ko/loxP pups and deleted Ercc1 by ex vivo 4‐hydroxy‐tamoxifen treatment. Similar to our in vivo findings, 3 months after tamoxifen treatment, the number of microglia in Cx3cr1‐Ercc1 ^ko/loxP OHSCs was reduced by approximately 50% compared to control OHSCs, in the DG, CA1, and CA3 regions (Figure S3a,b).

3.2. Ercc1‐deficient microglia are progressively replaced

After tamoxifen treatment, the Ercc1 ^loxP allele recombines into an Ercc1 ^rec allele in both Cx3cr1‐Ercc1 ^ko/loxP and Cx3cr1‐Ercc1 ^wt/loxP mice (Figure S1b,c). To determine the recombination efficiency, genomic DNA was isolated from FACS‐isolated microglia and analyzed by qPCR for the Ercc1 ^rec allele. In Cx3cr1‐Ercc1 ^wt/loxP mice microglia, the percentage of microglia with the Ercc1 ^rec allele was high (80–100%) and comparable at all investigated time points after tamoxifen treatment (Figure 2a). The percentage of microglia with an Ercc1 ^rec allele in Cx3cr1‐Ercc1 ^ko/loxP mice initially was same as in littermate controls, but progressively declined over time, to approximately 5% at 12 months after tamoxifen treatment (Figure 2a). Ercc1 recombination efficiency was further determined by single‐cell PCR. The percentages of Ercc1 ^rec microglia progressively decreased in Cx3cr1‐Ercc1 ^ko/loxP mice, and very few microglia with an Ercc1 ^rec allele were detected at 22 months after tamoxifen treatment (Figure 2b), corroborating our previous observations.

FIGURE 2.

Ercc1‐deficient microglia are gradually replaced by Ercc1 ^ko/loxP microglia. (a) The percentage of microglia carrying an excised Ercc1 ^loxP allele (Ercc1 ^rec) at a range of time points after tamoxifen treatment are depicted. The percentage of microglia with an Ercc1 ^rec allele was determined by qPCR using Ercc1‐rec1 and Ercc1‐rec2 primers and normalized to an unaffected genomic locus (the Il1b gene). Each dot represents an individual animal. A two‐way ANOVA followed by a Bonferroni correction for multiple comparisons was performed to assess significance. ***, p < 0.001. (b) The percentage of microglia carrying an Ercc1 ^rec allele at a range of time points after tamoxifen treatment was determined by single cell genomic qPCR. Individual microglia were FACS‐isolated and PCR amplified using Ercc1‐rec1 primers specific for the Ercc1 ^rec allele. The percentage of wells with a PCR product of the Ercc1 ^rec allele is shown in red, wells without a PCR product are indicated in light grey. As a positive control, individual microglia (MG) were PCR amplified using primers for the Il1b gene, indicated in dark grey. As a negative control, individual splenic macrophages (Mφ) were analyzed with Ercc1‐rec1 primers. n = 1–3 animals per group. (c) A cartoon illustrating the progressive loss of microglia cells and the gradual replacement of Ercc1‐deficient (Ercc1 ^ko/rec) microglia by nondeficient (Ercc1 ^ko/loxP) microglia in Cx3cr1‐Ercc1 ^ko/loxP mice after tamoxifen treatment

These data indicate that after tamoxifen treatment, Ercc1‐deficient microglia (Ercc1 ^ko/rec) were gradually lost in Cx3cr1‐Ercc1 ^ko/loxP mice and were replaced by Ercc1 ^ko/loxP microglia (Figure 2c). The Ercc1 ^ko/loxP microglia are likely cells that escaped tamoxifen induced Ercc1 deletion, and still carried a functional Ercc1 ^loxp allele. At 12 months after tamoxifen treatment, Ercc1‐deficient microglia were almost completely replaced by Ercc1 ^ko/loxP microglia, but microglia numbers were still reduced by approximately 40–50% (Figures 1 and 2). At 22 months after tamoxifen treatment, Ercc1‐deficient microglia were fully replaced and no differences in microglia densities were observed between control and Cx3cr1‐Ercc1 ^ko/loxP mice (Figures 1 and 2).

3.3. Altered microglia morphology in Cx3cr1‐Ercc1^{ko /loxP} mice

An evident change in microglia morphology was observed in the frontal cortex of Cx3cr1‐Ercc1 ^ko/loxP mice from 2 to 12 months after tamoxifen treatment when compared to littermate controls (Figure 3a). Some of the microglia became enlarged, with increased soma sizes and branch lengths (Figure 3a). This altered microglia morphology was also observed in other brain regions, including cortex, hippocampus, cerebellum and olfactory bulb (data not shown). Strikingly, at 22 months after tamoxifen treatment, all microglia in the Cx3cr1‐Ercc1 ^ko/loxP mice displayed a morphology that was comparable to littermate controls (Figure 3a). In OHSCs, a similar change in microglia morphology was observed in Cx3cr1‐Ercc1 ^ko/loxP mice (Figure S3c). Next, morphological differences in microglia were quantified across groups at different time points after tamoxifen treatment. The generated 23 morphometric features of each microglia cell are provided in Table S3.

FIGURE 3.

Altered microglia morphology in Cx3cr1‐Ercc1 ^ko/loxP mice. (a) Representative images of Iba1‐stained microglia in the cortex of Cx3cr1‐Ercc1 ^wt/loxP and Cx3cr1‐Ercc1 ^ko/loxP mice at different time points after tamoxifen treatment. (b) Hierarchical clustering on principal components resulted in nine cell clusters (I–IX). (c) Representative cells for each microglia cluster. (d) Sholl analysis for microglia clusters I–IX, revealing distinct differences in cell size and branching complexity between clusters. Dots and vertical lines represent means and +/− standard deviations respectively. (e) Distribution analysis of microglia clusters across genotypes at different time points after tamoxifen; wt/loxP: Cx3cr1‐Ercc1 ^wt/loxP , ko/loxP: Cx3cr1‐Ercc1 ^ko/loxP

To identify subsets of microglia with a similar morphology, we performed hierarchical clustering on principal components. First, a PCA was applied to the morphometric feature dataset. The first four PCs with an eigenvalue >1 were retained for hierarchical clustering (Figure S2a,b), resulting in 9 microglia clusters with distinct morphological properties (Figure 3b). Cell silhouettes representative for each cluster are depicted in Figure 3c. Notably, microglia in cluster I and II were almost exclusively derived from Cx3cr1‐Ercc1 ^ko/loxP mice, indicating these microglia clusters are Ercc1‐deficiency related (Figure 3b). Cluster I and II microglia were characterized by a relatively large soma area, high total branch length values, a large number of end nodes and relatively low cell solidity values (Figures 3c and S2). These findings were corroborated by Sholl analysis with cluster I and II microglia being the largest cells with most extensive ramification patterns compared to other clusters (Figure 3d). Comparisons of all morphometric features between microglia clusters can be found in Table S3. Next, we analyzed the relative distribution of the microglia clusters over the different mouse groups (genotype and time after tamoxifen treatment). In control mice, the relative proportion of cluster II and cluster I microglia remained low or even absent at all timepoints in control animals (Figure 3e). In Cx3cr1‐Ercc1 ^ko/loxP mice, between 2 and 12 months after tamoxifen treatment, cluster I and II microglia accounted for 45–65% of the total population in the cortex (Figure 3e). At 22 months after tamoxifen treatment, the microglia cluster distribution in Cx3cr1‐Ercc1 ^ko/loxP and Cx3cr1‐Ercc1 ^wt/loxP mice was comparable and cluster I and II microglia were almost absent (Figure 3e).

To summarize, upon Ercc1 deletion, CNS microglia numbers were reduced by approximately 50%, which was accompanied by the emergence of a microglia subpopulation (Cluster I and II) with relatively large and hyper‐ramified cells. This reduction in cell number and changes in microglia morphology persisted until 12 months after tamoxifen treatment. Gradually, microglia numbers and morphology returned to control levels at 22 months after tamoxifen treatment.

3.4. Increased proliferation compensates for microglia loss after Ercc1 deletion

Under homeostatic conditions, the proliferation rate of microglia is relatively low and the population is maintained by balanced proliferation and apoptosis (Askew et al., 2017). After genetic or pharmacological depletion of microglia, the remaining microglia rapidly expand and repopulate the CNS (Bruttger et al., 2015; Elmore et al., 2014; Rubino et al., 2018). As Ercc1‐deficient microglia were gradually replaced, we determined if microglia proliferation was increased after Ercc1 deletion. The expression levels of Ki67, a gene expressed by proliferating microglia (McDonough et al., 2020), was determined in microglia at early (1 and 2 months) and late (12 months) time points after tamoxifen treatment. Microglia from Cx3cr1‐Ercc1 ^ko/loxP mice expressed significantly higher Ki67 levels at 1–2 months after tamoxifen, indicating increased microglia proliferation (Figure 4a). Dividing microglia were observed in the Cx3cr1‐Ercc1 ^ko/loxP mouse hippocampus 1.5 months after tamoxifen treatment (Figure 4b, indicated by white arrows). At 12 months after tamoxifen, when almost all Ercc1‐deficient microglia were replaced, Ki67 expression levels had returned to levels observed in control microglia (Figure 4a). Next, we determined if Ki67 expressing, proliferating microglia were cells carrying an Ercc1 ^loxP allele or if Ercc1‐deficient microglia also proliferated. Microglia were isolated 2–3 months after tamoxifen treatment when the majority of the cells are still Ercc1‐deficient (see Figure 2). These microglia were separated in Ki67⁺ and Ki67⁻ subpopulations by FACS (Figure 4c). The percentage of Ki67⁺ cells was more than two‐fold higher in microglia from Cx3cr1‐Ercc1 ^ko/loxP mice compared to Cx3cr1‐Ercc1 ^wt/loxP controls (Figure 4c,d). Genotyping of Ki67⁺ microglia population revealed that the percentage of microglia with an Ercc1 ^rec allele from both Cx3cr1‐Ercc1 ^ko/loxP and Cx3cr1‐Ercc1 ^wt/loxP mice was comparable, indicating that Ercc1‐deficient microglia (Ercc1 ^ko/rec) also proliferated (Figure 4e).

FIGURE 4.

Increased microglia proliferation in Cx3cr1‐Ercc1 ^ko/loxP mice after tamoxifen treatment. (a) Ki67 gene expression was determined by qPCR and normalized to Hprt1 expression levels. Each dot represents one animal, n = 3–5 mice per group. A two‐way ANOVA followed by a Bonferroni correction was performed. ***, p < 0.001. (b) Example of a mitotic microglia in a Cx3cr1‐Ercc1 ^ko/loxP mouse, 1.5 months after tamoxifen treatment. (c) FACS plots showing the isolation of Ki67⁺ and Ki67⁻ microglia from Cx3cr1‐Ercc1 ^wt/loxP and Cx3cr1‐Ercc1 ^ko/loxP mice at 2–3 months after tamoxifen treatment. Percentage of Ki67⁺ and Ki67⁻ microglia from Cx3cr1‐Ercc1 ^wt/loxP and Cx3cr1‐Ercc1 ^ko/loxP mice was indicated in the plot. (d) Quantification of the percentage of Ki67⁺ and Ki67⁻ microglia in Cx3cr1‐Ercc1 ^wt/loxP and Cx3cr1‐Ercc1 ^ko/loxP mice at 2–3 months after tamoxifen treatment, n = 4 mice per group. An unpaired two‐tailed t‐test was performed. ***, p < 0.001. (e) The percentage of Ki67⁺ and Ki67⁻ microglia with Ercc1 ^rec allele was determined by genomic qPCR, n = 3–6 mice per group. A two‐way ANOVA followed by a Bonferroni correction for multiple comparisons was performed to assess significance

In summary, tamoxifen treatment resulted in Ercc1 excision, leading to a progressive loss of Ercc1‐deficient microglia in Cx3cr1‐Ercc1 ^ko/loxP mice. In parallel, the remaining microglia, including Ercc1‐deficient microglia, displayed increased expression of proliferation marker Ki67. At 12 months after tamoxifen treatment, when nearly all Ercc1‐deficient microglia (Ercc1 ^ko/rec) were replaced by Ercc1 ^ko/loxP microglia, Ki67 expression returned to control levels.

3.5. Ercc1‐deficient microglia are not immune activated or primed

In Ercc1 ^Δ/ko mouse microglia, the expression of genes like Axl, Lgals3 (Mac2), Apoe, and Itgax (Cd11c) is upregulated (Holtman et al., 2015; Figure 5a). The expression of these genes was also determined in Ercc1‐deficient microglia from Cx3cr1‐Ercc1 ^ko/loxP mice at different time points after tamoxifen treatment. The expression of Axl was significantly higher in Cx3cr1‐Ercc1 ^ko/loxP microglia only at 12 months compared to Cx3cr1‐Ercc1 ^wt/loxP microglia, but the level of induction was much lower than in Ercc1 ^Δ/ko mice (Figure 5a,b). Lgals3, Apoe, and Itgax expression was not significantly induced in microglia from Cx3cr1‐Ercc1 ^ko/loxP mice, suggesting microglia were not primed (Figure 5b).

FIGURE 5.

Ercc1‐deficient microglia are not immune activated or primed. (a) The expression of Axl, Lgals3, Apoe, and Itgax was determined by qPCR in microglia from Ercc1 ^Δ/ko mice and normalized to Hprt1 expression levels. An unpaired two‐tailed Student's t‐test was performed for statistical analysis. ***, p < 0.001. (b) Axl, Lgals3, Apoe and Itgax gene expression was determined by qPCR in microglia from Cx3cr1‐Ercc1 ^wt/loxP and Cx3cr1‐Ercc1 ^ko/loxP mice at different time points after tamoxifen treatment and normalized to Hprt1 expression levels. A two‐way ANOVA followed by a Bonferroni correction for multiple comparisons was performed for statistical analysis. **, p < 0.01. Gene expression levels of Ccl2 and Tnf were determined by qPCR in microglia from Ercc1 ^Δ/ko and control mice 3 hr after an i.p. PBS (c) or 1 mg/kg LPS (d) and normalized to Hprt1 expression levels. Each dot represents a mouse. An unpaired two‐tailed Student's t‐test was performed for statistical analysis, ***, p < 0.001. Gene expression levels of Ccl2 and Tnf were determined by qPCR in microglia from Cx3cr1‐Ercc1 ^wt/loxP and Cx3cr1‐Ercc1 ^ko/loxP mice 3 hr after an i.p. PBS (e) or 1 mg/kg LPS (f) and normalized to Hprt1 expression levels. Each dot represents a mouse. A two‐way ANOVA followed a Bonferroni correction was performed. **, p < 0.01, ***, p < 0.001

In agreement with our previous findings, basal expression levels of Ccl2 and Tnf were slightly higher in Ercc1 ^Δ/ko mice (Figure 5c) and Ercc1 ^Δ/ko mice microglia showed an increased responsiveness to LPS compared to controls in terms of Ccl2 and Tnf expression (Raj et al., 2014; Figure 5d). For Cx3cr1‐Ercc1 ^ko/loxP microglia, basal expression levels of Ccl2 and Tnf were similar to Cx3cr1‐Ercc1 ^wt/loxP microglia, again suggesting Cx3cr1‐Ercc1 ^ko/loxP microglia were not primed (Figure 5e). In response to LPS, Cx3cr1‐Ercc1 ^ko/loxP microglia only showed a (modest) enhanced Ccl2 expression at 2 and 6 months after tamoxifen treatment (Figure 5f). For Tnf expression, no increase in LPS responsiveness was observed in Cx3cr1‐Ercc1 ^ko/loxP microglia. (Figure 5f). In summary, microglia‐specific deletion of Ercc1 did not induce an immune activated or primed phenotype in microglia.

3.6. Gene expression profiling of microglia after Ercc1 deletion

To delineate the effect of Ercc1 deletion on microglia, we compared the gene expression profiles between Cx3cr1‐Ercc1 ^ko/loxP and Cx3cr1‐Ercc1 ^wt/loxP microglia, before and at different times after tamoxifen treatment (Figures 6a and S4).

FIGURE 6.

Gene expression profile of microglia after Ercc1 deletion. (a) Outline of microglia sampling. Microglia were isolated from Cx3cr1‐Ercc1 ^wt/loxP and Cx3cr1‐Ercc1 ^ko/loxP mice before and 5 days to 12 months after tamoxifen treatment. The gene profiles were generated by 3′ QuantSeq. Cx3cr1‐Ercc1 ^ko/loxP mice microglia without tamoxifen treatment served as nontamoxifen controls. (b) Heatmap of 1,670 differentially expressed genes through pairwise comparisons as described in Figure S4 (FDR < 0.05 and fold change >1.5). These genes clustered into 7 groups. (c) GO analysis of different clusters. The genes in Clusters 2 and 3 were combined for GO analysis. The top 20 GO terms are shown. (d) The number of differentially up‐ and down‐regulated genes between the indicated comparisons and associated GO terms are depicted (FDR < 0.05 and fold change >1.5). (e) Venn diagram showing unique and overlapping genes and associated GO terms between 458 upregulated genes (ko 2m tam vs. ‐tam microglia) and priming signature of aged and disease‐associated microglia (Holtman et al., 2015). The full lists of genes and enriched GO terms above mentioned are available in Table S4. ctrl, Cx3cr1‐Ercc1 ^wt/loxP mice after tamoxifen treatment; ko, Cx3cr1‐Ercc1 ^ko/loxP mice after tamoxifen treatment; ‐tam, Cx3cr1‐Ercc1 ^ko/loxP mice without tamoxifen treatment

Genes in Cluster 1 were more enriched in Cx3cr1‐Ercc1 ^wt/loxP microglia at 12 months than 5 days after tamoxifen treatment (Figure 6b), suggesting changes in expression of these genes are age‐related. Cluster 1 genes were associated with GO terms such as brain development, neuronal system, synapse, and morphogenesis (Figure 6c). Some cluster 1 genes, that showed increased expression in microglia isolated from mice 12 months after tamoxifen (from both genotypes), were also transiently increased in Cx3cr1‐Ercc1 ^ko/loxP mice microglia at 1 and 2 months after tamoxifen treatment, when most microglia were still Ercc1‐deficient (Figures 6b and 2). However, with ongoing microglia replacement, around 6 months after tamoxifen treatment. The expression level of these genes was reduced, suggesting Ercc1 deficiency caused a transient aging phenotype in microglia 1–2 months after tamoxifen treatment (Figure 6b). When microglia were fully replaced in Cx3cr1‐Ercc1 ^ko/loxP mice at 12 months after tamoxifen treatment (Figure 2), the transcriptional profiles of Cx3cr1‐Ercc1 ^ko/loxP and Cx3cr1‐Ercc1 ^wt/loxP mice microglia were again very similar (Figure 6b).

To get more insight into the transient microglia aging phenotype, several homeostatic microglia signature genes, inflammatory genes expressed by potentially senescent human microglia (Geirsdottir et al., 2019), and DAM genes were investigated (Geirsdottir et al., 2019; Keren‐Shaul et al., 2017). Expression levels of homeostatic microglia genes, such as Sall1, Cx3cr1, and Csf1r were unaffected by Ercc1 deletion (Figure S5a). However, at 2 and 6 months after tamoxifen treatment, microglia isolated from Cx3cr1‐Ercc1 ^ko/loxP mice displayed a transient aging phenotype with higher expression levels of some senescence‐associated genes (Tnf, Ccl3 and Ccl4) and some of the DAM genes (Apoe, Cst7, and Axl) (Figure S5b,c). But this microglia aging phenotype was no longer detected at 12 months after tamoxifen treatment in Cx3cr1‐Ercc1 ^ko/loxP mice (Figure S5b,c).

The expression of the majority of the genes in Clusters 2 and 3 was increased at 1 and 2 months after Ercc1 deletion, and progressively returned to control levels from 6 to 12 months (Figure 6b). Some genes in these clusters were associated with mitosis/cell cycle, corroborating the observed increased microglia proliferation at 1 and 2 months after tamoxifen treatment (Figures 6c and 4). In addition, some GO terms of Clusters 2 and 3 genes were related with apoptosis processes such as p53 signaling, release of cytochrome c from mitochondria and intrinsic apoptotic signaling pathway in response to DNA damage (Figure 6c). This supports the excessive loss of Ercc1‐deficient microglia between 1 and 6 months after tamoxifen treatment (Figure 1).

Cluster 7 contained genes that were downregulated in Ercc1‐deficient microglia at 1, 2‐, and 6‐months after tamoxifen treatment. GO enrichment indicated that these genes were involved in regulation of protein folding, extracellular exosome assembly, and lysosomal transport (Figure 6c).

To determine if tamoxifen treatment affected gene expression, the transcriptomes of microglia from Cx3cr1‐Ercc1 ^ko/loxP mice without tamoxifen treatment and Cx3cr1‐Ercc1 ^wt/loxP mice 5 days after tamoxifen treatment were compared. We only identified 28 differentially expressed genes which were involved in p53 signaling and cell cycle (Figure 6d), indicating a limited influence of tamoxifen treatment on microglia gene expression. No differentially expressed genes were detected between Cx3cr1‐Ercc1 ^wt/loxP and Cx3cr1‐Ercc1 ^ko/loxP mice microglia 5 days after tamoxifen treatment, indicating that 5 days of Ercc1 deficiency was not sufficient to affect microglia gene expression (Figure 6b).

The number of differentially expressed genes, both up‐ and down‐regulated, in microglia at different times after tamoxifen treatment, and associated GO terms are depicted in Figure 6d. The gene lists and enriched GO terms are provided in Table S4.

Microglia‐specific deletion of Ercc1 resulted in a gene expression signature distinct from the priming signature of aged and disease‐associated microglia we previously reported (Holtman et al., 2015). Only 19 of the 458 genes upregulated at 2 months after tamoxifen treatment in Cx3cr1‐Ercc1 ^ko/loxP microglia overlapped with this priming gene expression module consisting of 295 upregulated genes, suggesting microglia were not primed by intrinsic DNA damage repair deficiency (Figure 6e).

4. DISCUSSION

Here, we analyzed the effect of Ercc1‐deficiency on microglia. Deletion of Ercc1 resulted in progressive microglia loss and to compensate for this cell loss, microglia proliferation was transiently increased. Interestingly, the remaining Ercc1‐deficient and nondeficient microglia both displayed increased proliferation activity. Gradually, likely due to ongoing loss of Ercc1‐deficient microglia, the CNS gradually repopulated with nondeficient (Ercc1 ^ko/loxp) microglia. Unlike constitutive Ercc1‐knockout mice, intrinsic Ercc1 deletion did not induce microglia activation or priming. Microglia‐specific deletion of Ercc1 transiently induced an aging‐associated gene expression profile, which was different from the gene expression signature of aged and CNS disease‐associated microglia (Holtman et al., 2015).

4.1. Phenotypes of microglia in conditional Cx3cr1‐Ercc1^{ko /loxP} mice compared to constitutive Ercc1^{Δ /ko} mice

In constitutive Ercc1 ^Δ/ko mice, Ercc1‐deficiency was already present in the zygote, and in all cell types. However, in Cx3cr1‐Ercc1 ^ko/loxP mice, Ercc1‐deficiency in Cx3cr1‐expressing cells was induced by tamoxifen treatment in young adult mice (6–8 weeks of age). In constitutive Ercc1 ^Δ/ko mice, microglia density was increased but in Cx3cr1‐Ercc1 ^ko/loxP mice, microglia were lost after Ercc1 deletion. Constitutive Ercc1 deletion resulted in a microglia priming gene expression signature (Holtman et al., 2015). In a previous study, we showed that Camk2 ^creER‐driven Ercc1 deletion in forebrain neurons also resulted in a microglia phenotype reminiscent of what was observed in Ercc1 ^Δ/ko mice (Raj et al., 2014). In contrast, although microglia from Cx3cr1‐Ercc1 ^ko/loxP mice transiently displayed an aging gene expression profile at 2 months after tamoxifen treatment, no clear gene expression signature of priming was detected. Together, these results suggest that microglia priming can be triggered by neuronal genotoxic stress (Raj et al., 2014), but not by microglia‐intrinsic genotoxic stress after Ercc1 deletion. Nonetheless, Ercc1 is an essential protein for microglia, as microglia deficient for Ercc1 are progressively replaced by microglia with a functional Ercc1 allele.

4.2. Turnover of microglia in Cx3cr1‐Ercc1^{ko /loxP} mice after tamoxifen treatment

Using genetic labeling and long‐term in vivo imaging, Füger et al. reported a median lifetime of mouse neocortical microglia of 15 months (Füger et al., 2017). In a similar study, the turnover time of mouse microglia was estimated to be 41, 15, and 8 months in the cortex, hippocampus, and olfactory bulb, respectively (Tay et al., 2017). Askew et al. showed that 0.69% of the microglia population is proliferating with an estimated turnover time of 96 days (Askew et al., 2017). Despite these differences in reported turnover times, microglia are relatively long‐lived cells in the CNS. Strikingly, after experimental depletion of microglia by either using mice expressing a CD11b‐TK transgene (Gowing, Vallières, & Julien, 2006), CSF1R inhibitors (Rice et al., 2017), or the Cx3cr1 ^creER :iDTR system, remaining microglia repopulated the CNS very fast (Bruttger et al., 2015; Elmore et al., 2014; Rubino et al., 2018; Varvel et al., 2012). The disadvantage of these approaches is the very fast depletion and repopulation, resulting in (transient) astrogliosis. And the repopulated microglia in the Cx3cr1 ^creER :iDTR system have an altered, interferon regulatory factor 7‐driven activation phenotype (Rubino et al., 2018; Waisman, Ginhoux, Greter, & Bruttger, 2015).

In the Cx3cr1‐Ercc1 ^ko/loxP mice used in this study, Ercc1‐deficient microglia were gradually lost, which was associated with a transient increase in proliferation. Similarly, Varol et al. reported a progressive replacement of Dicer‐deficient microglia in Cx3cr1 ^creER :Dicer ^ko/loxP mice after tamoxifen treatment. Importantly, they further showed that Dicer‐deficiency could induce DNA damage in newborn microglia (Varol et al., 2017). Most likely, Ercc1 deficiency resulted in microglia apoptosis, reflected by the increased expression of apoptosis‐related genes in Cx3cr1‐Ercc1 ^ko/loxP microglia at 1 and 2 months after tamoxifen treatment (Figure 6bc, Cluster 2 and 3). Importantly, this gradual replacement of microglia did not result in microgliosis.

After pharmacologic depletion by CSF1R inhibitors, the repopulated microglia are derived from the remaining microglia population without contribution from peripheral myeloid cells (Elmore et al., 2014; Huang et al., 2018; Zhan, Sohn, Zhou, Li, & Gan, 2019). For genetic microglia ablation using the Cx3cr1 ^creER :iDTR system, Bruttger et al. also showed that microglia exclusively renew from the remaining cells (Bruttger et al., 2015). However, Lund et al showed that the repopulated microglia originated from remaining CX3CR1⁺F4/80^loClec12a⁻ microglia and CX3CR1⁺F4/80^hiClec12a⁺ microglia‐like macrophages originated from Ly6C^hi monocytes (Lund et al., 2018). These monocyte‐derived macrophages can acquire some key features of microglia but still are transcriptionally and functionally distinct from CNS resident microglia even 12 weeks after depletion (Lund et al., 2018). In our study, at 12 months after tamoxifen treatment, the repopulated microglia are transcriptionally similar to microglia from control mice, suggesting that the repopulated microglia in Cx3cr1‐Ercc1 ^ko/loxP mice are most likely derived from microglia that escaped Cre recombination.

4.3. Morphology of microglia after Ercc1 deletion

Microglia are pleomorphic and can adapt to different environments. Ramified microglia are relatively quiescent and surveil the parenchyma, while microglia with a more amoeboid phenotype are more migratory, phagocytic and immune activated (Kettenmann, Hanisch, Noda, & Verkhratsky, 2011). Microglia morphometrics revealed significant alterations in microglia morphology in Cx3cr1‐Ercc1 ^ko/loxP mice after tamoxifen treatment. Further analysis revealed this was mainly due to changes in a subset of microglia. Interestingly, this subset of microglia consisted of both Ercc1‐deficient and nondeficient microglia, since this subset was detected at both 2 months (when the majority of microglia is Ercc1 ^ko/rec; Figure 2) and 12 months (when the majority of microglia is Ercc1 ^ko/loxP; Figure 2) after tamoxifen treatment. This suggests that, in compensation for the cell loss, both Ercc1‐deficient (Ercc1 ^ko/rec) and nondeficient (Ercc1 ^ko/loxP) microglia became larger in size and more extensively ramified, likely in order to surveil a larger parenchymal area. However, potential functional differences between these enlarged cells and homeostatic microglia are still unresolved.

4.4. Spontaneous recombinase activity of CreER

In this study, we used Cx3cr1 ^creER mice to delete Ercc1 in microglia. Microglia deficient for Ercc1 were gradually lost and replaced by Ercc1 ^ko/loxP microglia. The transcriptional changes we determined at different time points after tamoxifen treatment were generated using all microglia isolated from mouse brain. As a consequence, the effects of Ercc1‐deficiency on gene expression might be diluted by Ercc1 ^ko/loxP microglia that progressively populate the brain after tamoxifen treatment. In order to separate Ercc1‐deficient Ercc1 ^ko/rec microglia and nondeficient Ercc1 ^ko/loxP microglia, we generated a Cx3cr1 ^creERT2 :Ercc1 ^ko/loxP :R26 ^{CAG‐tdTomato} mouse line. In these mice, Ercc1 ^ko/rec microglia would also express the tdTomato reporter, where Ercc1 ^ko/loxP microglia do not. This assumes that the excision of the floxed Ercc1 and floxed stop cassette upstream of the tdTomato reporter gene are equally efficient. However, we found 70% of the microglia already expressed the tdTomato reporter prior to tamoxifen treatment (data not shown). This “leaky” Cre activity of the Cx3cr1 ^creER transgene was confirmed by other studies (Chappell‐Maor et al., 2020; Fonseca et al., 2017). In addition, for our Cx3cr1 ^creERT2 :Ercc1 ^ko/loxP :R26 ^{CAG‐tdTomato} reporter mouse line, we observed that the spontaneous activity of Cre only affected the floxed stop cassette upstream of tdTomato, but not the floxed sequence of our target gene Ercc1. This is most likely due to the size of the floxed fragment of our target gene Ercc1, which is around 2.5 kb (Doig et al., 2006), which is longer than the 0.8 kb floxed stop cassette in the tdTomato reporter (Madisen et al., 2010). The different susceptibility of constructs in response to basal CreERT2 activity has also been shown in other studies (Alvarez‐Aznar et al., 2020; Van Hove et al., 2020). This tamoxifen‐independent Cre activity made it impossible to distinguish between Ercc1‐deficient and nondeficient microglia in our Cx3cr1 ^creERT2 :Ercc1 ^ko/loxP :R26 ^{CAG‐tdTomato} mice.

In summary, our data indicate that Ercc1 is an essential protein for microglia and its deletion leads to cell death. As a consequence, microglia are gradually replaced by nondeficient Ercc1 ^ko/loxP microglia carrying a functional Ercc1 ^loxP allele. The replacement of Ercc1‐deficient microglia by nondeficient microglia is not accompanied by extensive immune activation or gliosis in the CNS. These data furthermore indicate that the functionality and gene expression changes observed in constitutive Ercc1 ^Δ/ko mice are not microglia intrinsic but likely caused by an aging environment, in agreement with earlier postulations (Raj et al., 2014).

Supporting information

FIGURE S1 Experimental strategy of microglia‐specific Ercc1 deletion. (a) Breeding strategy to generate microglia‐specific conditional Ercc1 knockout mice on a mixed genetic background (50% C57BL/6J, 50% FVB). (b) Cartoon depicting four different Ercc1 allelic variants. For the Ercc1 ^ko allele, a neo cassette was inserted into exon 7, the Ercc1 ^loxP allele contains two loxP sites in Introns 2 and 5. Excision of the Ercc1 ^loxP allele by Cre recombinase results in the Ercc1 ^rec allele. (c) After tamoxifen treatment, both the Ercc1 ^loxP alleles in Cx3cr1‐Ercc1 ^wt/loxP and Cx3cr1‐Ercc1 ^ko/loxP mice will become the Ercc1 ^rec allele. However, only microglia in Cx3cr1‐Ercc1 ^ko/loxP mice will become Ercc1‐deficient, since in Cx3cr1‐Ercc1 ^wt/loxP mice microglia, the Ercc1 ^wt allele is still present. (d) To confirm Ercc1 deletion in microglia, all mice were genotyped by genomic PCR on sorted microglia. A representative example is depicted. Mice of 6–8 weeks of age received 2 doses of 500 μl of tamoxifen (20 mg/ml) to induce Ercc1 gene recombination by oral gavage. Four months post tamoxifen treatment, DNA isolated from sorted microglia was used for genotyping. The Ercc1 ^rec allele generated a 280 bp PCR product (left). The Ercc1 ^wt, Ercc1 ^loxP, and Ercc1 ^ko alleles (as indicated by the arrows) were detected in microglia sorted from Cx3cr1‐Ercc1 ^wt/loxP and Cx3cr1‐Ercc1 ^ko/loxP mice. All the PCR products were run on the same agarose gel but not in adjacent lanes.

FIGURE S2 Clustering of microglia with similar morphometric features through hierarchical clustering on principal components. To reduce the dimensionality of the dataset, a PCA was performed. (a) A table depicts the eigenvalues, the variance retained and the cumulative variance for the first 10 PCs. PCs with an eigenvalue >1 (here, PC1–PC4) accounted for 85% of the accumulative variance and were retained for hierarchical clustering. (b) Hierarchical clustering on the first four PCs resulted nine distinct morphological clusters (I–IX). (c) the heatmap represents the normalized values for all morphometric features for microglia clusters I–IX. (d) Comparison of representative morphometric features (i.e., soma area, total branch length, the number of end nodes and cell solidity) between microglia clusters I–IX. Violin plots represent the kernel densities for each cluster. Upper, middle and lower hinges of the boxplots represent the first, second (median) and third quantiles, respectively. Whiskers extend to the minimum and maximum value still within the 1.5× the interquartile range of the first and third quantiles, respectively. Dots represent outliers.

FIGURE S3 morphometric features of microglia clusters Changes in microglia number and morphology in OHSC after microglia‐specific Ercc1 deletion. (a) Iba1 immunofluorescence staining of cultured organotypic hippocampal slice across groups. Hippocampal slices were prepared from Cx3cr1‐Ercc1 ^ko/loxP and Cx3cr1‐Ercc1 ^wt/loxP mouse pups (p3). On the first day after preparation, OHSCs were treated with 1 nM 4‐hydroxy tamoxifen for 48 hr to induce Ercc1 deletion ex vivo, after which slices were further cultured for 3 months. (b) comparison of microglia densities in the three hippocampal regions DG, CA1, and CA3 across groups (n = 3 pups per group). Box boundaries represent first and third quartiles and the center lines indicate the medians. Whiskers extend from box boundaries to the minimum and maximum values respectively. Significance was assessed using an unpaired two‐tailed t‐test. **. p < 0.01; ***, p < 0.001. (c) Morphological changes of microglia after Ercc1 deletion ex vivo.

Figure S4 Lists of genes and enriched GO terms Comparisons and respective cut‐offs used in the gene expression analysis. For gene expression analysis, 24 RNA samples from eight conditions were collected for 3′ QuantSeq. FDR < 0.05 and FC > 1.5 were used as the cutoff for differentially expression genes (DEGs). DEGs from 13 comparisons were pooled to create a list for clustering analysis. Abbreviation: ‐tam, Cx3cr1‐Ercc1 ^ko/loxP mice without tamoxifen treatment; Ctrl, Cx3cr1‐Ercc1 ^wt/loxP mice after tamoxifen treatment; KO, Cx3cr1‐Ercc1 ^ko/loxP mice after tamoxifen treatment; d, day; m, month; FC, fold change; FDR, false discovery rate.

Figure S5 Transient aging phenotype in microglia after Ercc1 deletion. Gene expression values (CPM) of selected microglia signature genes (a), senescence‐associated secretory phenotype genes (b), and DAM genes (c) at different time points before and after tamoxifen treatment. All groups are compared to the 5d‐WT group. Significance was assessed using an unpaired two‐tailed t‐test. *, p < 0.05; **, p < 0.01; ***, p < 0.001. Abbreviation: ‐tam, Cx3cr1‐Ercc1 ^ko/loxP mice without tamoxifen treatment; WT, Cx3cr1‐Ercc1 ^wt/loxP mice after tamoxifen treatment; KO, Cx3cr1‐Ercc1 ^ko/loxP mice after tamoxifen treatment.

Table S1 information of PCR primers for genotyping.

Table S2 Information of primers for qPCR (mRNA).

Table S3 Morphometric features of microglia clusters.

Table S4 Lists of genes and enriched GO terms.

Zhang X, Heng Y, Kooistra SM, et al. Intrinsic DNA damage repair deficiency results in progressive microglia loss and replacement. Glia. 2021;69:729–745. 10.1002/glia.23925

DATA AVAILABILITY STATEMENT

The data that supports the findings of this study are available in the supplementary material of this article