Repopulated microglia induce expression of Cxcl13 with differential changes in Tau phosphorylation but do not impact amyloid pathology

All animal procedures were reviewed and approved by the University Committee on Animal Resources of the University of Rochester Medical Center and performed according to the Institutional Animal Care and Use Committee and guidelines from the National Institute of Health (NIH). Animals were housed in a 12-h light/12-h dark cycle with food ad libitum. 14-month-old male and female Tg(APPswe, PSEN1dE9)85Dbo mice (also known as APP/PS1) were obtained from an established colony (JAX stock no. 005864) maintained at the University of Rochester vivarium. Tg(APPSwe, tauP301L)1Lfa Psen1^tm1Mpm mice (also known as 3xTg) were initially obtained from Frank M. LaFerla and Salvatore Oddo by Howard Federoff and maintained at the University of Rochester as a homozygous line. The 3xTg mice express mutated human APP Swedish, MAPT P301L under the control of the Thy1.2 promoter and PS1 M146V under the Psen1 promoter. Age-matched non-transgenic (NTg) mice, bred continuously in a parallel colony to 3xTgs with a similar genetic background, were used as wild-type controls to 3xTg mice in RNA sequencing experiments. With age, the 3xTg mice develop Aβ plaque deposits and intraneuronal hyperphosphorylated Tau aggregates. Our studies used two cohorts of 21–22-month-old male 3xTg mice; females from these cohorts were not available as they had been used for unrelated studies.

Mice received a chow diet (AIN-76A-D1001i, Research Diets) containing 1200 mg/kg PLX5622 (Chemgood) ad libitum for 2 weeks to deplete microglia. Control chow with the same base formula without PLX5622 was given to the control group. After the 2-week treatment, mice were returned to the standard chow diet: AIN-76A (Research Diets) for the 3xTg Cohort 1 or 5053-Rodent Diet 20 (currently in use at the University of Rochester vivarium) to allow microglial repopulation for 1 month. All of these formulations were irradiated by the vendor.

Mice were injected 24 h before euthanasia with Methoxy-X04 (MeX04, i.p., 4 mg/kg, Tocris Biosciences), a brain-permeable Aβ fluorescent marker [41, 42]. On the day of euthanasia, animals were deeply anesthetized with a mixture of xylazine (i.p., 10 mg/kg) and ketamine (i.p., 100 mg/kg) and perfused intracardially with 0.15 M phosphate buffer (PB) containing 0.5% sodium nitrite and 2 IU heparin/ml. After perfusion, hemispheres were separated: one was either immediately submerged in fixative solution (4% paraformaldehyde (PFA), pH 7.2 in PB, 4 °C) to be used for immunofluorescence experiments or flash-frozen in cold isopentane for ELISA (both as described below), and the other was processed for flow cytometry as follows. The hippocampus from each half brain was dissected and homogenized in 3 ml FACS buffer (1× Phosphate Buffered Saline (PBS) + 0.5% BSA). Homogenates were filtered through a 70 µm cell strainer into a 15 ml tube containing 3 ml FACS buffer. The strainer was washed with an additional 3 ml of FACS buffer, and the cell suspensions were centrifuged at 400×g for 5 min at 4 °C. The supernatants were discarded, and the remaining pellets were resuspended in 40% Percoll (Cytiva) prepared with PBS, then centrifuged at 400g for 30 min with no braking. After removing the supernatants, the pellets were resuspended in 90 µl FACS buffer with 1:100 Fc block (2.4G2, 1:100, BioLegend) and transferred to a 96-well plate. After a 15 min incubation with Fc block at 4 °C, the following antibodies were added in a 10 µl master mix: CD11b-FITC (M1/70, Biolegend), CD45-APC/Cy7 (30F11, Biolegend), 7AAD (Invitrogen), P2Ry12-APC (S16007D, Biolegend) & TMEM119-PE (106-6, Abcam). The latter two cell surface molecules are considered homeostatic microglial markers [4]. The plate was then incubated for 30 min at 4 °C in the dark. The samples were washed once with FACS buffer and transferred to 5 ml tubes containing 7AAD such that its final dilution was 1:80. Appropriate fluorescent-minus-one (FMO) and single-stained bead controls (Ultracomp eBeads, Invitrogen) were prepared in tandem with samples. After excluding debris, doublets, and dead cells, CD45^lo/CD11b⁺ was used to gate for microglia on a FACSAria II (BD). MeX04⁺ and MeX04⁻ microglia were sorted. Samples from APP/PS1 mice were analyzed the same way, but with a LSR II flow cytometer (BD) without sorting. All events were recorded, and data were analyzed with FCS Express 7 (DeNovo Software).

Half-brains were fixed overnight in 4% PFA at 4 °C, dehydrated in 30% sucrose overnight, frozen in cold isopentane, and stored at − 80 °C until sectioning on a − 25 °C freezing stage microtome into 30 µm thick coronal slices stored in a cryoprotectant solution. For immunofluorescence, sections were washed extensively in PBS and blocked with 10% normal donkey or goat serum for 1 h at RT.

For amyloid pathology analysis, sections were immunolabeled for amyloid-beta (Aβ), microglia (Iba1 or P2RY12), a common microglial activation marker CD68, and a widely used marker for neuritic damage LAMP1 [43]. The following primary antibodies were used: biotin anti-Aβ (clone 6E10, 1:3000, BioLegend), rabbit anti-Iba1 (1:2000, Wako), rabbit anti-P2RY12 (1:2000, Anaspec), rat anti-CD68 (1:500, Bio-Rad) and rat anti-LAMP1 (1:2000, Abcam). Sections were incubated in primary antibodies for 48 h at 4 °C. The sections were washed and incubated in fluorescently labeled secondary antibodies/reagents (Alexa Fluor 488, Alexa Fluor 594 streptavidin conjugate and Alexa Fluor 647, Invitrogen; all at 1:1000) for 3 h at RT, then mounted and coverslipped (Prolong Gold, ThermoFisher Scientific).

For Tau pathology analysis, sections were incubated in biotinylated mouse anti-HT7 (1:1000, Invitrogen) and either rabbit anti-pT205 (1:1000, Invitrogen), rabbit anti-pS396 (1:1000, RayBiotech) or rabbit anti-pS409 (1:800, Invitrogen) for 2 h at RT then overnight at 4 °C. They were washed, incubated in fluorescently labeled secondary antibodies (Alexa Fluor 488, Invitrogen) and Alexa fluor 594 streptavidin conjugate (Invitrogen; all at 1:1000) for 3 h at RT, and then mounted and coverslipped.

For each animal, 3–4 coronal tissue sections that included the subiculum (S) and CA1 field of the hippocampus (CA1) were imaged with a Nikon A1R HD confocal microscope using a 10× (Plan Apo Lambda, NA: 0.40), 20× (Plan Apo VC, NA: 0.75) or 40× water-submersion (Apo LWD, NA: 1.15) objective lens as indicated in the figure legends. Imaging parameters were kept constant across all sections for each set of immunofluorescent labels. All image analysis was performed using ImageJ FIJI (NIH) with semi-automated custom macros. Experimenters were blinded to treatment.

Frozen hippocampi were homogenized in Tissue Protein Extraction Reagent (ThermoFisher Scientific) at a concentration of 50 mg/ml with 1× Halt Protease and Phosphatase Inhibitor Single-Use Cocktail (ThermoFisher Scientific), vortexed and sonicated. The homogenates were centrifuged for 100,000×g for 1 h. The supernatant was collected as the soluble fraction; whereas the pellet was incubated in guanidinium-HCl pH 6.0 for 4 h and centrifuged at 100,000×g for 1 h. This new supernatant was collected as the insoluble fraction. For Aβ40 ELISAs (ThermoFisher Scientific), the soluble fraction was diluted at 1:20, and the insoluble fraction was diluted at 1:3000. For Aβ42 ELISAs (ThermoFisher Scientific), the soluble fraction was diluted 1:2, and the insoluble was diluted 1:30. The soluble fraction diluted at 1:3 was used as input to the CXCL13 ELISA kit (R&D Systems). All dilutions were established empirically.

All statistical analyses were performed in Graphpad Prism v7.04. Comparisons between PLX and control-treated groups in male-only experiments were made using Student’s t-test. Comparisons of MeX04 and MFI of several homeostatic markers were made using two-way ANOVA with Bonferroni correction. All data points that represent individual animal averages are presented as mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001.

To explore the impact of microglial self-renewal on AD pathology, we utilized an established paradigm that partially depletes microglia using PLX5622 (PLX), a colony stimulating factor 1 receptor (CSF1R/c-kit/Flt3) inhibitor, in 3xTg and APP/PS1 mouse models of AD. After 2 weeks of PLX exposure, microglia numbers in the brains of aged 3xTg (22-month-old) and APP/PS1 (14-month-old) mice decreased by approximately 50% and 65%, respectively (Additional file 1: Fig. S1A–C, F). Consistent with previous findings [19], we observed greater depletion in plaque-devoid regions (50% in 3xTg and 70% in APP/PS1) versus 40% and 50% depletion of plaque-associated microglia in 3xTg and APP/PS1, respectively (Additional file 1: Fig. S1D, E, G, H). The impact of microglial repopulation was assessed 1 month after the discontinuation of PLX treatment (Figs. 1A, 2A). We found that self-renewal of microglia did not improve amyloid pathology, evidenced by the unaltered area fraction occupied by amyloid plaques, as well as the average size and number of plaques in either male 3xTg (Fig. 1B–E, G–I) or APP/PS1 mice of both sexes (Fig. 2B–E and Additional file 2: Fig. S2A–D).

Furthermore, when we examined dense-core plaques in 3xTg mice detected with MeX04, we observed no change with microglial repopulation (Fig. 1K–M). While ELISA measurements revealed a significant decrease of insoluble Aβ42, we did not detect any statistically significant changes in soluble forms of Aβ42 and Aβ40 or insoluble Aβ40 (Additional file 3: Fig. S3A–D). Since amyloid pathology only weakly correlates with neuronal loss and cognitive function [45], we sought to examine the impact of PLX treatment on the levels of dystrophic neurites in both APP/PS1 and 3xTg mice. In contrast to our expectations, we observed no improvement in levels of neuritic damage, indicated by similar levels of LAMP1 staining, a lysosomal membrane glycoprotein, under all conditions (Figs. 1F, J, 2F), except in the subiculum of female APP/PS1 mice, where repopulation increased LAMP1 levels (Additional file 2: Fig. S2E). Consistent with a lack of improvement in neuritic damage, microglial repopulation did not lead to changes in a battery of behavioral assays, including NOR, CFC, and Lashley III maze, in 3xTg mice (Additional file 4: Fig. S4A). Behavioral readouts for recognition, fear, spatial memory, and anxiety showed no difference in 21-month-old PLX-treated 3xTg mice when compared to control 3xTg mice (Additional file 4: Fig. S4B–F).

While microglial repopulation did not significantly attenuate plaque pathology or influence behavioral performance (Figs. 1 and 2; Additional file 2: Fig. S2, Additional file 3: Fig. S3, and Additional file 4: Fig. S4), we sought to determine whether repopulated microglia exhibited any changes in phagocytic potential or the levels of signature homeostatic and activation markers (Fig. 3). In order to assess microglial capacity to phagocytose plaques, we injected MeX04, a brain-permeable fluorescent probe for Aβ, 24 h prior to euthanasia and FACS-sorted hippocampal CD11b⁺CD45^int microglia into MeX04⁺ and MeX04⁻ fractions (Additional file 5: Fig. S5). The percentage of MeX04⁺ microglia, which represents the proportion of plaque-phagocytosing microglia, was not significantly different between PLX and control treatments, although there was a trend toward a higher fraction of microglia internalizing MeX04 with PLX treatment in the 3xTg model (3xTg: Fig. 3A, p = 0.14; APP/PS1: Additional file 6: Fig. S6A and Additional file 7: Fig. S7A). In agreement with these findings, microglia recruitment to plaque, quantified by the amount of Iba1 immunoreactivity in close proximity to plaque, also remained unaltered after repopulation (Fig. 3F, J; Additional file 6: Fig. S6E and Additional file 7: Fig. S7E). Similarly, we did not observe changes in total microglial volume coverage (Fig. 3G, K; Additional file 6: Fig. S6D and Additional file 7: Fig. S7D).

We next assessed the levels of CD68, P2RY12, and TMEM119 in repopulated microglia (Fig. 3B, C, H, I, L, M; Additional file 6: Fig. S6B, C, G and Additional file 7: Fig. S7B, C, F). Repopulated microglial levels of CD68, a well-known marker for reactive microglia, showed a trend toward decreasing in the subiculum of 3xTg mice (Fig. 3H, p = 0.06) and in the subiculum and CA1 in APP/PS1 mice, although the main effect of PLX treatment was only statistically significant in APP/PS1 female mice (Additional file 6: Fig. S6G and Additional file 7: Fig. S7F). Furthermore, microglial levels of CD68 were unchanged between the 3xTg groups in CA1, a region that does not display amyloid pathology in these mice (Fig. 3I). The levels of homeostatic markers P2RY12 and TMEM119 displayed more complex changes, reflecting the difference in pathologies between the two models. In the 3xTg model, the PLX treatment group displayed significantly lower levels of P2RY12 (Fig. 3B) but slightly higher levels of TMEM119 (Fig. 3C), albeit not significantly (p = 0.09). Due to the restricted amyloid pathology in the subiculum in 3xTg mice, the amount of MeX04⁺ microglia was too small to analyze separately and, therefore, was combined with MeX04⁻ microglia for analysis. Quantification of P2RY12 intensity in brain tissue revealed similarly decreased (although not significantly) levels of P2RY12 in the CA1 region of the hippocampus, a region with hyperphosphorylated Tau but lacking amyloid pathology (Fig. 3M, p = 0.15). This effect was not observed in the subiculum, which displays amyloid pathology (Fig. 3L). In the APP/PS1 model with only amyloid pathology, TMEM119 levels significantly increased with the repopulation of microglia in males but not in females (Additional file 6: Fig. S6B and Additional file 7: Fig. S7B). A similar pattern was observed with P2RY12 for males, which shows a trend (p = 0.15) toward higher levels with microglial repopulation (Additional file 6: Fig. S6C, p_F(treatment) < 0.05). As expected, MeX04⁺ microglia showed decreased levels of TMEM119 and P2RY12 compared to MeX04⁻ microglia in both sexes of APP/PS1 mice (Additional file 6: Fig. S6B, C and Additional file 7: Fig. S7B, C).

To examine the effects of repopulated microglia on Tau pathology, we quantified the levels of three different phosphorylated Tau epitopes, pT205, pS409 and pS396, in two cohorts of 3xTg mice of similar age (Fig. 4A). When normalized to total Tau levels, we found that repopulation of microglia led to a significant increase in pT205 signal ratio in the first cohort (Fig. 4B) and a slight but non-significant increase in the second cohort (Fig. 4F). Interestingly, levels of pS409 normalized to total Tau were significantly decreased in both of the cohorts (Fig. 4C, G). Similarly, levels of pS396 showed a strong trend for a decrease (p = 0.05) in the second cohort but remained unchanged in the first cohort (Fig. 4D, H). Importantly, there was no change in levels of total Tau in any of the cohorts (Fig. 4E, I).

To determine transcriptomic associates of microglial repopulation, we performed single-cell RNA sequencing (scRNAseq) of FACS-isolated CD45^int/+ cells from pooled hippocampi (Fig. 5A). After quality control, we identified 15,828 cells for further bioinformatic analysis. Cells identified as doublets or apoptotic were excluded as described in “Methods” section. Clustering analysis through the Seurat package revealed nine distinct clusters in our dataset, plotted on Uniform Manifold Approximation and Projection (UMAP) space, without any anchoring algorithms that may artificially coerce the position of individual data points (Fig. 5B). By using the single-cell Mouse Cell Atlas (scMCA package), we identified five major cell types: perivascular macrophages and microglia (PVMMicro), T cells, B cells, and neurons (Fig. 5B Inset). Microglial cells represented the overwhelming majority of identified cells and were further subdivided into six clusters (Fig. 5B–D).

In our dataset, most cells were identified as homeostatic microglia (Fig. 5C). These cells expressed markers such as P2ry12, Cxc3cr1, and Tmem119 (Fig. 5D). The second most abundant cell type observed in both groups resembled activated response microglia (ARMs; or disease-associated microglia, DAM). This well-established cluster was identified based on the expression of Tyrobp, Apoe, Lpl, Cst7, and other markers in accordance with previous research (Fig. 5D) [3, 4, 14, 46]. Similarly, we identified a cluster enriched in interferon-related genes such as Ifitm3 and Isg15 (Fig. 5D). We believe these cells might be related to interferon response microglia (IRMs) which have also been previously observed [14]. We also observed a small cluster of microglial cells with high expression levels of cell cycle and DNA replication genes, such as Mki67 (Fig. 5B, D), which could represent dividing microglia (“div mg”, also known as cycling/proliferating microglia or CPM). This small cluster has been identified in previous scRNAseq studies, too [14]. Together, these four clusters accounted for a similar proportion of total cells between control and repopulated microglia (Fig. 5C).

Furthermore, between 3 and 5% of CD45^int/+ cells in the hippocampus fell into a microglial cluster defined by downregulation of cytoskeletal genes such as Actb (translates to β-actin) and Tuba1b; but also by upregulation of the long non-coding RNA Malat1 (Fig. 5D). Gene Ontology analysis through clusterProfiler revealed that genesets pertinent to the regulation of actin, as well as mitochondrial function, were negatively enriched (Additional file 9: Fig. S9A, B). This cluster, which we refer to as “cyto-mg” for the purposes of this study, was found in similar proportions between control and PLX-repopulated (4.5% and 3.4%, respectively) groups but in much higher proportions in NTg mice (24.6%, Additional file 10: Fig. S10A–C).

Lastly, we identified a cluster (“PLX-microglia”) whose proportion changed between 3xTg control-treated and PLX-repopulated microglia (Fig. 5B). This “PLX-microglia” cluster made up 12.1% of CD45^int/+ cells in the PLX-repopulated group but only 2.3% of cells in control-treated group in 3xTg mice (Fig. 5C). The PLX-microglia cluster was very strongly and almost exclusively (|logFC threshold|> 1.9, Additional file 11: Table S1), characterized by upregulation of Cxcl13 (Fig. 5E). Detailed analysis of all microglial clusters revealed that PLX-repopulated microglia expressed more Cxcl13 in general (Fig. 5E and Additional file 11: Table S1). In contrast, Cxcl13 was detected in very few microglia isolated from control-treated NTg mice (Additional file 10: Fig. S10D). We confirmed increased CXCL13 levels in hippocampal lysates through ELISA (Fig. 5F), and we visualized Cxcl13 mRNA expression through RNAScope technology (Fig. 5G, H). In our observations, Cxcl13 staining colocalized with Iba1 immunopositivity, suggesting that Cxcl13 is of microglial origin (Fig. 5G). We found strong trends towards increased Cxcl13 expression in CA1 (p = 0.05) and subiculum (p = 0.06) in PLX-repopulated mice compared to the controls (Fig. 5I, J), but there was no meaningful difference between the groups in the cortex (p = 0.64, Fig. 5K).

While microglia have complex roles in AD pathology, several studies have shown the benefits of eliminating microglia in AD mouse models. Although prolonged PLX treatment did not alter plaque load or Aβ levels in either 5xFAD (4-week PLX treatment starting at 14 months) [19] or 3xTg mice (12-week PLX treatment starting at 15-months) [18], both studies found that chronic microglial elimination partially prevented cognitive dysfunction in these mice [18, 19], suggesting that microglia contribute to neuronal dysfunction possibly via release of inflammatory cytokines and chemokines in the chronic neuroinflammatory environment [19]. Thus, “rejuvenating” microglia by pharmacologically inducing their repopulation could replace these microglia that have been shaped by their prolonged residence in an inflammatory environment, resulting in new microglia that better perform homeostatic functions. However, our experiments show that microglial repopulation did not ameliorate neuritic damage or improve cognitive outcomes in old AD-like animals (Figs. 1, 2 and Additional file 2: Fig. S2, Additional file 4: Fig. S4). Furthermore, we observed an increase in neuritic damage with microglial repopulation in APP/PS1 female mice, which is consistent with Gratuze et al. which described increased BACE1 immunopositivity around plaques in their depletion/repopulation paradigm, suggestive of increased dystrophic neurites [33]. We hypothesize that one potential explanation for our results could be that the repopulated microglia after PLX-depletion are “primed” by the surrounding inflammatory CNS microenvironment in the aged brain rapidly after they are born [32]. In fact, other investigators have shown that repopulated microglia do not alter the response to immune challenges or modify their expression of inflammation related genes [30, 32], which is consistent with our observations of the microglial activation marker CD68.

Interestingly, a recent study demonstrated that PLX treatment caused a significant shift from compact to diffuse plaque morphology with increasing neuritic damage [34]. These findings support a pivotal role for microglia in limiting fibrillar plaque expansion by encapsulating Aβ to form a protective barrier, preventing toxic effects of filamentous Aβ on nearby neurons [43, 49]. Disruption to microglial clustering around plaques led to a shift in plaque structure from compact to more diffuse with complex fibrillar branching, resulting in more dystrophic neurites surrounding plaques [43]. However, when microglia were allowed to repopulate, amyloid pathology was comparable to non-depleted animals, suggesting that the repopulated microglia replaced the resident population, but did not offer further disease-modifying benefits [34]. This is consistent with our observation that microglial repopulation did not alter amyloid pathology or improve dystrophic neurites surrounding plaques (Figs. 1, 2 and Additional file 2: Fig. S2). Although we observed a decrease in insoluble Aβ42 measured by ELISA, the insoluble Aβ40 and the soluble fractions of both Aβ42 and Aβ40 remained unchanged (Additional file 3: Fig. S3). Thus, our data add to previous findings and suggest no clear benefit of repopulation to amyloid pathology or cognitive outcomes (Figs. 1, 2 and Additional file 2: Fig. S2, Additional file 3: Fig. S3, and Additional file 4: Fig. S4) in 3xTg or APP/PS1 models of AD at stages of advanced pathology. In fact, repopulation may worsen pathology in some cases, as in a recent study which depleted and repopulated microglia in a Tau seeding model of 5-month-old 5xFAD mice and found that amyloid pathology was exacerbated 3.5 months post-PLX treatment [33]. On the other hand, chronic microglia depletion that is started before pathology onset (i.e., prior to plaque formation) led to a decrease in plaque burden and reduction of dystrophic neurites [20, 21], suggesting that intervening before the environment becomes inflammatory may lead to better outcomes. Simply limiting microglial proliferation beginning at an earlier timepoint in pathogenesis also reduces pathology [50], consistent with recent findings on the critical role of microglia in amyloid plaque seeding [9]. However, we believe microglial depletion after pathology onset is more relevant to clinical settings. Our data add to the growing body of evidence that repopulating microglia in advanced amyloid pathology does not offer clear benefits.

Since microglia play a complex role in the development of Tau pathology, the impact of their depletion and/or repopulation is of substantial clinical importance. The outcomes from depletion-only paradigms have revealed conflicting results depending on models used and dosages of CSF1R-antagonists administered [18‐21, 51‐53]. However, this is the first study to our best knowledge that has characterized the impact of microglial repopulation post-depletion on a transgenic model of Tau pathology. This is worth investigating since the side effects of sustained CSF1R inhibition can lead to uncharacterized outcomes [54]. We found that PLX-based repopulation of microglia increased pT205 immunopositivity, but not that of pS396 in CA1 (Fig. 4). A study in 5xFAD mice, in which Tau was seeded after the depletion and repopulation phases, also found increased AT8⁺ Tau immunoreactivity [33] which recognizes epitopes pS202 and pT205. On the other hand, we found that pS409 epitope was consistently and significantly decreased in the PLX-repopulated groups, while overall Tau levels were unchanged (Fig. 4). These findings highlight the value of assessing Tau pathology through multiple approaches [55] and suggest a potential disease-modifying benefit of our repopulation paradigm.

Transcriptomic analysis of microglial repopulation in 3xTg mice identified previously reported clusters of microglia associated with AD but also noted two distinct, previously unreported clusters. The “Cyto-mg” cluster was similarly represented in PLX-repopulated and control 3xTg mice, but its prevalence was greatly reduced compared to the NTg control mice (Additional file 10: Fig. S10A, B). Gene ontology analysis of downregulated transcriptomic markers that define this cluster revealed significant enrichment of gene sets pertinent to cytoskeleton organization and mitochondrion organization (Additional file 9: Fig. S9). The most upregulated gene in this cluster was Malat1 (Fig. 5D, Additional file 11: Table S1), which has been shown to play roles in various neuroinflammatory insults—potentially through activation of the inflammasome [56]. However, we also acknowledge that downregulation of aforementioned gene ontology terms (Additional file 9: Fig. S9) may represent an artifactual outcome due to the lower number of detected RNA molecules or annotated genes (Additional file 8: Fig. S8). Further, it is unclear whether the transcriptomic signature of “Cyto-mg” cluster exists in microglia in other models of Tau pathology. Single-cell sequencing of microglia from Thy-Tau22 mice revealed a transcriptomic landscape characterized by increased proportions of ARM (or DAM) rather than a signature associated with cytoskeletal components or Malat1 [57]. However, in that study, the authors anchor their Tau datasets with APP/PS1 datasets using CCA as a dimensionality reduction method (as opposed to PCA) which might have masked rare populations that exist only in one group [58]. Nevertheless, questions pertinent to existence and/or function of “Cyto-mg” cluster in models of AD are beyond the scope of this paper.

The other previously unreported cluster that we observed here is strongly defined by upregulation of Cxcl13 (Fig. 5B, D). Although this “PLX-microglia” cluster was distinct from other types of activated microglia, we also observed Cxcl13 upregulation in Homeo and ARM/DAM clusters, which account for most of the microglia, but not in other CD45^int/+ cells (Fig. 5C, E and Additional file 11: Table S1). As expected, statistical testing on all cells identified as microglia revealed significant upregulation of Cxcl13 in PLX-treated groups (Additional file 11: Table S1). We confirmed this finding at the protein level as well (Fig. 5F). Interestingly, we observed strong trends towards upregulation of Cxcl13 transcripts with in situ hybridizations in CA1 and subiculum of PLX-repopulated mice, but this effect was not detected in the cortex (Fig. 5I–K). This suggests that Cxcl13 is preferentially expressed in regions that display AD pathology. Since CXCL13 plays an important role in homing and activation of CXCR5⁺ lymphocytes, such as B cells, follicular helper T cells, and Th17 cells [37], this transcriptomic signature might suggest an overall increased inflammatory microenvironment. This is further supported by decreased microglial levels of P2RY12 (Fig. 3B). CXCL13 expression has also been observed in models of multiple sclerosis, particularly in cerebrospinal fluid and meningeal tertiary lymphoid organs [35]. Whether this leads to increased lymphocyte infiltration or formation of tertiary lymphoid organs following microglial repopulation in AD mouse models warrants further research.

Lastly, we found similar proportions of DAM/ARM between control and repopulated groups in 3xTg mice. This finding contrasts another study [33] in which microglia did not re-establish DAM signature following a repopulation paradigm in the 5xFAD model. Although it is difficult to directly compare across studies due to the use of different sequencing technologies that provide different levels of resolution (microarray vs scRNAseq), we argue that the age of animals (~ 5 months vs. 23 months), as well as the presence of active Tau pathology in our model, may have created an environment that is more conducive to re-establishment of the DAM/ARM signature.