Neonatal microglia replacement in mice modulates seizure severity in adulthood

Abstract

Microglia replacement therapy, where endogenous brain macrophages are depleted and replaced by adoptively transferred surrogates, holds promise for treating pediatric neurologic diseases, but little is known about how early-life microglia replacement impacts the brain. We sought to investigate how early postnatal microglia depletion and adoptive macrophage transfer, essential components of microglia replacement, durably impact neural circuits in a mouse model. Using both pharmacologic and genetic models, postnatal microglia depletion worsened adult seizure severity, mortality, and neuropathology in a chemical seizure model. Replacement of endogenous microglia by adoptive transfer of monocytes rescued this effect, while transfer of authentic microglia from a donor mouse did not, and even worsened seizure phenotypes. RNA sequencing of transplanted microglia, monocyte-derived surrogates, and endogenous microglia revealed distinct state changes across groups in response to chemically induced seizures, demonstrating that both ontogeny and adoptive transfer significantly impact resident macrophage responses to the excitotoxic brain environment. In sum, we established models for neonatal microglia depletion and replacement, then applied them to identify durable impacts of depletion and reconstitution on the brain environment. We ultimately identified differential responses of macrophages to excitotoxic challenge based on their ontogeny, underscoring focus areas for ongoing preclinical development of microglia replacement therapies.

Graphical abstract

Replacing dysfunctional microglia by transplanting healthy surrogate cells holds promise for treating pediatric neurological diseases. Bennett and colleagues show that neonatal depletion of microglia worsens seizure outcomes in adulthood, while transplantation of monocytes, but not microglia, rescues this effect. Their findings inform future strategies for developing effective microglia replacement therapies.

Introduction

Microglia are long-lived brain-resident macrophages that arise embryonically. After depletion, microglia rapidly repopulate the brain parenchyma through self-renewal.¹ Hematopoietic stem cell transplantation (HSCT) depletes some microglia and leads to engraftment of donor-derived macrophages in their place.² This so called “microglia replacement,” thought to underlie the efficacy of HSCT for pediatric degenerative diseases such as metachromatic leukodystrophy, Krabbe disease, and X-linked adrenoleukodystrophy, includes depletion of dysfunctional endogenous microglia and adoptive transfer of therapeutic cellular surrogates that engraft the brain parenchyma and differentiate into competent brain-resident macrophages.^3,4,5,6,7 Although microglia replacement holds great therapeutic potential,^8,9,10,11 little is known about how its core components (microglia depletion, adoptive transfer, and donor cell chimerism) affect the long-term functioning of the central nervous system (CNS), particularly when performed during early development.

Because microglia play particularly important roles in circuit development,^{12,13,14,15,16,17,18,19,20} both the impact of early postnatal microglia depletion, as would be required to treat pediatric diseases, and the lasting consequences of long-term donor cell engraftment, are poorly understood but essential considerations for microglia cell therapies. Microglia also undergo developmentally regulated programming thought to be important in establishing their homeostatic setpoint, and therefore appropriate functioning.²¹ It is unknown how microglia depletion, repopulation, or transplantation may impact this developmental programming. Therefore, our goal was to model key steps required for early postnatal microglia replacement, to better understand their long-term impacts on neural circuit function. A promising translational approach to microglia depletion is small-molecule CSF1R inhibitors (CSF1Ri) that (1) when paired with HSCT, greatly enhance donor chimerism in the brain parenchyma,^9,22 and (2) when used with engineered, “inhibitor resistant” CSF1R variants, achieve high-efficiency microglia replacement without the need for traditional chemotherapy.²³ We made use of one such FDA-approved CSF1Ri, PLX3397, as an initial model of postnatal microglia depletion.

For adoptive cell transfer, a major additional uncertainty is appropriate selection of donor cells to generate microglia surrogates. Circulating blood cells, such as hematopoietic stem cells (HSCs) and monocytes, are accessible, readily engraft the brain parenchyma, and become “microglia-like,” here referred to as monocyte-derived microglia (mo-microglia). Mo-microglia, however, remain transcriptionally distinct from authentic yolk sac-derived microglia (embryonically derived microglia, em-microglia), which are highly specialized to the functional demands of the brain parenchyma,^3,10,22,24 including when transplanted into the brain.^3,22,24 To test how transplantation with true microglia versus monocytes affects the brain, we used a genetic model of microglia depletion paired with concurrent intracranial transplantation of donor cells, which leads to high-efficiency microglia replacement.²⁵

To study the long-term impacts of neonatal microglia replacement on neural circuit development, we measured seizure susceptibility as a broadly relevant neurological disease endophenotype, prevalent in many neurological diseases potentially treated by microglia replacement. Additionally, true microglia express distinctly high levels of P2Y12 that allow them to mitigate exuberant neuronal activity during seizures,^26,27 while mo-microglia are P2Y12 low, raising the possibility that monocyte-derived donor cells may respond differently to neuronal excitation. P2Y12 impacts on seizure are most extensively studied in chemical models such as kainic acid (KA) treatment,^26,27 a paradigm directly compatible with our microglia replacement models.

Based on this rationale, we report several discoveries relevant to preclinical development of microglia replacement therapies. First, early postnatal microglia depletion using the small-molecule CSF1Ri PLX3397 worsened seizure outcomes in adulthood, leading to increased neuropathology 5 days after seizure challenge. A genetic model of microglia depletion and repopulation, which also allows rapid postnatal microglia replacement by intracranial adoptive transfer, phenocopied these results. Interestingly, transplantation of monocytes, but not microglia, rescued both behavioral and neuropathologic seizure phenotypes associated with postnatal depletion. Bulk RNA sequencing of transplanted monocytes, transplanted microglia, and unmanipulated endogenous microglia after seizures revealed distinct ontogeny- and transplant-dependent state changes, suggesting that both the origin of the repopulating cell, as well as early-life microglia de- and repopulation itself, have long-lasting impacts on neural circuitry.

Results

Early postnatal PLX3397 treatment increases seizure susceptibility in adulthood

Prior studies used early depletion by CSF1R antagonism to identify important contributions of microglia to brain development^20,28 but it remains largely unknown whether transient early loss of microglia causes persistent phenotypes into adulthood. This is particularly important for the emerging preclinical development of microglia replacement therapies for pediatric diseases, which require early microglia depletion for donor macrophage engraftment.

For microglia depletion, we treated neonatal mice with CSF1Ri PLX3397 daily during the first postnatal week, after which we allowed endogenous microglia to fully repopulate the brain (Figure 1A). With daily administration of PLX3397 starting on postnatal day 1 (P1), we found by P3 a significant depletion of microglia, with near-complete ablation by P7 (Figure S1). Following PLX3397 cessation, microglia fully repopulated the brain (Figure S1), as predicted by prior studies of adult depletion.²⁹

Figure 1.

Early postnatal treatment with PLX3397 increases seizure susceptibility in adulthood

(A) Schematic of PLX3397 treatment paradigm. C57Bl/6 mice were given daily injections with PLX3397 from P1 to P7. Endogenous microglia were then allowed to repopulate 2–3 months before seizure induction. (B) Racine scores indicating seizure severity over time in male mice. No PLX, control C57Bl/6 mice; PLX Repop, C57Bl/6 mice with postnatal depletion and repopulation of microglia. No PLX n = 6, PLX Repop n = 9, data pooled from two independent experiments. (C) Seizure severity scores were determined as the total Racine score over the course of the experiment. No PLX n = 6, PLX Repop n = 9, unpaired Student’s t test. (D) Percent of male mice that died during status epilepticus in No PLX and PLX-repopulated groups. No PLX n = 6, PLX Repop n = 9. (E) Time to peak seizure severity in male mice. No PLX n = 6, PLX Repop n = 9, unpaired Student’s t test. (F) Representative traces showing fEPSP (local field potential) from the CA1 region of the hippocampus in No PLX (left, black) and PLX-repopulated (right, purple) male mice. Dotted line indicates the increased amplitude observed in PLX-repopulated mice. (G) Input-output curve showing increased slopes at higher stimulation in male PLX-repopulated animals. No PLX n = 8, PLX Repop n = 9, data combined from three independent cohorts, repeated measures ANOVA (with Bonferroni correction for multiple comparisons). All data represented as mean ± SEM. ns, not significant; ∗p < 0.05, ∗∗p < 0.01.

To model persistent impacts of CSF1Ri on neural circuit function following repopulation, we tested the behavioral response to induced seizures in a repeated low-dose KA model.³⁰ Two to 3 months after P1-P7 microglia depletion, adult mice received repeated, low doses of KA to induce seizures (Figure 1A). In this model of seizure induction, mice initially receive a moderate dose of KA (10 mg/kg), followed by progressively lower doses (5 mg/kg, then 2.5 mg/kg) every 20 min until a stage behavioral 5 seizure is observed or a maximum 6 KA doses are given. Unlike giving a single, large dose of KA, this model allows most animals to survive, giving us the opportunity to assess not only initial seizure behavior but also downstream neuropathological outcomes. Male mice given postnatal PLX3397 and then repopulated (“PLX-repopulated”) displayed more severe behavioral seizures in adulthood (Figures 1B and 1C), as measured by Racine score and overall seizure severity scores, while females (estrus stage not tested) were not affected (Figures S2A and S2B). PLX-repopulated males but not females also had increased mortality after reaching status epilepticus (Figures 1D and S2C). Latency to peak seizure score was not significantly different between normal development controls and PLX-repopulated mice (Figures 1E and S2D), suggesting that early postnatal PLX3397 exposure does not impact the time to peak behavioral seizures, but rather the overall time spent in convulsive seizure stages. One potential confounder of using a repeated low dose is that KA dosing may be variable between mice. However, we did not observe significant differences in cumulative KA dose between untreated and PLX-repopulated groups (Figure S2E).

To test network excitability as a reason for the observed seizure sensitivity differences, we prepared live brain slices for local field recording in hippocampal subregions of male PLX-repopulated and control mice in the absence of KA challenge. We found no difference in field excitatory postsynaptic potential (fEPSP) slopes in response to stimulation of either inner or outer molecular dentate gyrus afferent input (data not shown) but found significantly increased fEPSP slopes in area CA1 in brain slices from PLX-repopulated animals (Figures 1F and 1G), consistent with hyperexcitable network development, which may increase susceptibility to seizure activity. In sum, early postnatal treatment with PLX3397 causes enduringly increased excitatory network activity in area CA1 of the hippocampus and decreased seizure threshold in adulthood.

Neonatal treatment with PLX3397 increases gliosis and neuropathology following seizures in adulthood

After assessing behavioral seizure phenotypes, we recovered mice for 5 days in order to determine if seizure-induced neurodegeneration or gliosis was affected with postnatal PLX3397 treatment. We observed no obvious differences in morphology between control microglia and PLX-repopulated microglia at baseline, but after seizures PLX-repopulated microglia had a more abnormal morphology, including reduced ramification, and abundant rod-shaped and rounded microglia, suggestive of a persistently reactive state (Figure 2A). To further characterize microglial morphology before and after seizure, we assigned microglia to one of four morphology types: baseline/homeostatic with highly complex processes (type 0), those with shorter, thicker, less complex processes (type 1), those with very large soma and many short, “fluffy”-appearing processes (type 2), and those with a rounded shape with few to no processes (type 3) (Figure S3). Using this typing system, we observed no significant differences between control microglia and PLX-repopulated microglia before seizure induction, with type 0 being the most common morphology type (Figure 2B, top). After seizures, however, 52.3% of unmanipulated microglia had returned to a homeostatic morphology by 5 days post-seizure, compared with only 2% of PLX-repopulated microglia, while 98% displayed one of the three abnormal morphologies (Figure 2B, bottom). In line with abnormal microglia morphology, we found markedly increased astrogliosis (Figures 2C and 2D) and more severe neurodegeneration as visualized by Fluorojade C, a dye used to label degenerating neurons³¹ (Figures 2E and 2F). These findings demonstrate that the prolonged severe seizures associated with early postnatal pharmacologic microglia depletion and repopulation lead to persistently increased glial activation and neurodegeneration.

Figure 2.

Neonatal treatment with PLX3397 leads to increased gliosis and neuropathology following seizures in adulthood

(A) Representative IBA1 staining (magenta) showing morphology of untreated or PLX-repopulated microglia before seizures (baseline) or 5 days post-seizure (5dps) in the hippocampus (top row) and cortex (bottom row). Representative of n = 3 (No PLX-baseline), n = 4 (No PLX-5dps), n = 3 (PLX Repop-baseline), and n = 6 (PLX Repop-5dps), two independent experiments. Scale bar, 50 μm. (B) Microglia morphology typing before and after seizures. Microglia were assigned one of four morphology types: baseline/homeostatic with highly complex processes (type 0), short, thick, less complex processes (type 1), very large soma with many short, “fluffy”-appearing processes (type 2), and rounded shape with few to no processes (type 3). The percentage of cells of each type is shown before seizures (top) and after seizures (bottom) for No PLX and PLX Repop groups. No PLX (baseline) n = 2, PLX Repop (baseline) n = 3, No PLX (5dps) n = 4, PLX Repop (5dps) n = 5, data pooled from two independent experiments. (C) Representative astrocyte GFAP expression (white) in the cortex of mice with or without postnatal PLX3397 treatment before seizures (baseline) or 5dps. Scale bar, 50 μm. (D) Quantification of area covered by GFAP-expressing astrocytes. Ctx, cortex; Hipp, hippocampus; Stria, striatum; Thal, thalamus. No PLX-baseline n = 4, PLX Repop-baseline n = 4, No PLX-5dps n = 6, PLX Repop-5dps n = 9, data pooled from two independent experiments, one-way ANOVA with Tukey’s multiple comparisons. (E) Representative Fluorojade C staining (green) indicating neurodegeneration in the hippocampus, striatum, and thalamus of untreated mice (top) and PLX-repopulated mice (bottom) 5dps. Scale bar, 100 μm. (F) Quantification of Fluorojade C+ cells in the hippocampus, striatum, and thalamus. No PLX n = 5, PLX Repop n = 6, data pooled from two independent experiments, unpaired Student’s t test. All data represented as mean ± SEM. ns, not significant; ∗p < 0.05, ∗∗p < 0.01.

Neonatal transplantation of monocytes, but not microglia, rescues depletion-induced seizure phenotypes

Although FDA-approved PLX3397 efficiently depletes microglia, off-target effects on oligodendrocyte progenitor cells (OPCs) and the peripheral immune system^32,33 could contribute to the above effects. Having observed a durable seizure phenotype after PLX-mediated depletion and repopulation, we tested a second model of early postnatal microglia depletion, one compatible with subsequent microglia replacement by adoptive transfer, to also test how microglia replacement altered depletion-induced seizure phenotypes.

We crossed the inducible Cx3cr1-CreER mouse³⁴ to Csf1r fl/fl mice to generate Cx3cr1-CreER; Csf1r fl/fl mice, which allows for microglial depletion following tamoxifen administration.^22,24,25 We administered tamoxifen on postnatal days P1 and P2, followed by intracranial transplant of donor GFP-expressing monocytes (mo-microglia) or microglia (em-microglia) on P3.²⁵ Another cohort received a sham injection that allowed for repopulation with endogenous microglia (“tamoxifen-repopulated”), an orthogonal model to validate PLX-repopulation experiments. We then allowed the donor cells or endogenous microglia to repopulate the parenchyma (Figure 3A). At age 3 months, donor mo-microglia and em-microglia had engrafted approximately 50% of the brain, most consistently in the cortex and hippocampus (Figures 3B and S4A). Because donor cell engraftment levels could not be assessed prior to seizure experiments, we induced seizures in all transplanted animals, but in downstream analysis chose a priori to include only those animals that had at least 50% replacement of endogenous microglia by donor cells in the hippocampus, as this region is thought to initiate seizure activity in models of KA-induced seizures.^35,36,37 Donor mo-microglia generally engrafted at higher densities than transplanted or endogenous -microglia (data not shown), consistent with previous reports, including using the same transplant and engraftment approach.^25,38 As a readout of the state of transplanted and repopulated cells once engrafted in the brain, we stained for homeostatic parenchymal macrophage marker TMEM119. As previously reported,²⁴ both transplanted em- and mo-microglia expressed TMEM119 at baseline, as did tamoxifen-repopulated and unmanipulated microglia (Figure S4B). As a barometer of baseline CNS state after microglia replacement, we characterized reactive gliosis before seizure induction. Early after tamoxifen treatment and transplant (P16), we found elevated GFAP expression in all tamoxifen-treated groups, which largely normalized by age 3 months (Figure S4C). We also found similar numbers of OPCs in the corpus callosum of all groups at both P16 and age 3 months (Figure S4D). Finally, we quantified baseline inhibitory structural synapses in CA1 of the hippocampus, where KA-induced seizures are thought to initiate, and found no significant differences between groups at age 3 months (Figures S4E and S4F).

Figure 3.

Postnatal repopulation with em-microglia, but not mo-microglia, results in worsened seizure outcomes

(A) Schematic of microglia depletion and transplantation paradigm. Cx3cr1-CreER; Csf1r fl/fl mice were treated with tamoxifen on P1 and P2 to deplete endogenous microglia. On P3, donor GFP+ microglia (em-MG) or monocytes (mo-MG) were transplanted and allowed to engraft in the brain for 2–3 months. Some mice received a sham PBS injection that allowed endogenous microglia to repopulate the brain (Tam Repop). At age 2–3 months, kainic acid was used to induce seizures. (B) Representative dot renderings of engraftment levels for transplanted microglia (em-MG, top) and transplanted monocytes (mo-MG, bottom) at age 3 months. Representative of n = 9 (em-MG) and n = 13 (mo-MG) animals per group, three independent experiments (em-MG) and four independent experiments (mo-MG). (C) Racine scores indicating seizure severity over time in male mice. em-MG, transplanted microglia; mo-MG, transplanted monocytes; Endog-MG, unmanipulated endogenous microglia; Tam Repop = tamoxifen + sham injection. em-MG n = 19, mo-MG n = 16, Endog-MG n = 10, Tam Repop n = 12, data pooled from seven independent experiments. (D) Total seizure severity scores in male mice. Dotted line represents average seizure severity scores for PLX-repopulated male mice. em-MG n = 19, mo-MG n = 16, Endog-MG n = 10, Tam Repop n = 12, one-way ANOVA with Tukey’s multiple comparisons. (E) Percent of male mice in each group that died during status epilepticus. em-MG n = 19, mo-MG n = 16, Endog-MG n = 10, Tam Repop n = 12. (F) Time to peak seizure severity in male mice. em-MG n = 19, mo-MG n = 16, Endog-MG n = 10, Tam Repop n = 12, one-way ANOVA with Tukey’s multiple comparisons. All data represented as mean ± SEM. ns, not significant; ∗p < 0.05, ∗∗p < 0.01.

Following baseline assessment of engraftment in the Cx3cr1CreER; Csf1r fl/fl model of microglia replacement, we induced seizures as described for PLX3397 studies in Figure 1. We observed that both tamoxifen-repopulated mice and em-microglia-transplanted mice displayed more severe seizure behaviors compared with their unmanipulated counterparts, similar to the phenotype observed in PLX-repopulated mice (Figures 3C and 3D). Mice transplanted with mo-microglia, however, had milder seizures, similar to unmanipulated controls (Figures 3C and 3D). In line with increased behavioral seizure severity, we observed increased mortality in both tamoxifen-repopulated and transplanted em-microglia but not mo-microglia groups (Figure 3E). Latency to peak seizure severity, however, was not different between any groups, as seen in our PLX-repopulated animals (Figure 3F).

We had initially hypothesized that transplanted mo-microglia would be more likely than em-microglia to worsen seizure phenotypes, given their low expression of purinergic receptor P2Y12, which is important for microglial regulation of seizure activity in the brain.^26,27 Given the result that mo-microglia engraftment rescued microglia depletion and repopulation-induced seizure severity, we were concerned that the different methods used to isolate donor cells (primary microglia versus bone marrow-derived monocytes) might impact seizure responses. While donor populations for both cell types were isolated using the Miltenyi MACS (magnetic bead sorting) system, microglia isolation used positive selection CD11b beads, while bone marrow-derived monocytes used negative selection (see materials and methods). As such, magnetic beads bound to CD11b on the surface of microglia could potentially impact their state or function compared with negatively selected monocytes (Figure S5A). To address this, we tested the impact of bead exposure on transplanted monocytes and microglia. Following isolation using our standard negative selection method, we incubated unbound monocytes with the same CD11b-positive selection cocktail as used for microglia, and transplanted these bead-bound monocytes into the brain (Figure S5A). To spare transplanted microglia from bead exposure, we used FACS to isolate primary microglia, followed by transplantation without magnetic sorting (Figure S5A). In both cases, we found no differences in seizure responses compared with normal cell isolation methods: FACS-isolated transplanted em-microglia led to similar seizure outcomes as MACS-sorted em-microglia, and bead exposed mo-microglia similarly rescued behavioral seizure activity associated with depletion and endogenous microglia repopulation (Figures S5B and S5C), suggesting that isolation method is not responsible for the observed differences.

Although we did not control for estrus stage in females, we remained interested in whether a sex-specific effect would also be present in the genetic transplant model as it was in PLX-repopulated mice. Interestingly, both males and females with transplanted em-microglia displayed more severe seizures and increased mortality compared with controls (Figures S5D–S5F). Only males with transplanted mo-microglia, however, had significantly rescued seizure severity and mortality (Figures 3C–3E), while females were not similarly rescued (Figures S5G–S5I). Latency to peak seizure activity was not impacted by transplant status, once again suggesting that overall time in convulsive seizure stages accounted for differing seizure outcomes (Figure 3F). Like in PLX-repopulation experiments, we wanted to confirm whether different cumulative doses of KA may explain the different seizure phenotypes between groups. We found small differences between groups in their cumulative KA dose received over the course of the experiment, with untransplanted males and females usually receiving the maximum dose, but there were no significant differences in cumulative dose between em- and mo-microglia groups (Figure S5J). Finally, given the variability in engraftment levels observed in our transplant model, we tested for correlations between donor cell engraftment level and seizure severity without an engraftment cutoff. We found a moderate negative correlation between mo-microglia engraftment levels and seizure severity score (Figure S5K), suggesting that higher mo-microglia engraftment is more protective against seizure susceptibility in this model. Conversely, we found a positive correlation between em-microglia engraftment and seizure severity score (Figure S5L). Taken together, these results highlight a durable effect of early postnatal microglial depletion and repopulation on seizure severity, which is rescued in a sex-dependent manner by monocyte transplantation.

Monocyte engraftment rescues seizure-associated neuropathology elicited by early postnatal microglia depletion and repopulation

As in PLX-repopulation experiments, we allowed KA-challenged mice to recover for 5 days and assessed seizure-induced gliosis and neurodegeneration. We used the same morphology typing system as described for Figure 2 to characterize microglia morphologies in our transplant model. Because the morphology of mo-microglia remains distinct from that of true microglia both at baseline and following challenge, we typed mo-microglia as follows: baseline/homeostatic defined as mo-microglia with fewer than five primary processes (type 0), mo-microglia with five to seven primary processes with minimal secondary branching (type 1), and mo-microglia with greater than eight primary processes with extensive secondary branching (type 2) (Figure S3). At baseline, we found no significant differences in morphology between em-microglia, untransplanted microglia, and tamoxifen-repopulated microglia, with most microglia in all groups being type 0 (Figures 4A and 4B – top). Similarly, most mo-microglia at baseline were characterized as type 0. Five days after seizures, both transplanted and tamoxifen-repopulated em-microglia displayed extensive non-homeostatic morphology, with only 4.2% and 6% of microglia being type 0, respectively, compared with 70% of control microglia being type 0 (Figures 4A and 4B, bottom). As opposed to the significant morphology changes observed in em-microglia conditions, mo-microglia displayed reduced morphological changes following seizures, with 50% of mo-microglia displaying homeostatic morphology (Figures 4A and 4B, bottom). We also assessed astrogliosis, and observed increased GFAP expression in several brain regions in mice with transplanted em-microglia (Figures 4C and 4D). Because there were no significant differences in GFAP expression between groups prior to seizure induction, these results demonstrate that repopulated and transplanted em-microglia exuberantly amplify astrocyte activation after seizure, consistent with prior studies showing that reactive microglia can directly activate astrocytes.³⁹ Transplanted mo-microglia, however, rescued this effect when compared with transplanted em-microglia and were similar to unmanipulated controls (Figures 4C and 4D). Interestingly, unlike in mice with transplanted em-microglia, tamoxifen-repopulated em-microglia mice did not display increased astrogliosis (Figures 4C and 4D). Neurodegeneration, however, was increased in both transplanted em-microglia and tamoxifen-repopulated mice compared with unmanipulated and mo-microglia-transplanted mice (Figures 4E and 4F). Together, these results uncover ontogeny-dependent impacts of neonatal microglia replacement on post-ictal damage responses.

Figure 4.

Mice with transplanted em-microglia or tamoxifen-repopulated em-microglia have more neuropathology following seizures compared with mo-MG-repopulated mice

(A) Representative IBA1 staining (magenta) showing morphology of transplanted yolk sac-derived microglia (em-MG), transplanted monocyte-derived microglia (mo-MG), unmanipulated endogenous microglia (Endog-MG), and tamoxifen-repopulated microglia (Tam Repop) at baseline (top row) and 5 days post-seizure (5dps) (bottom row) in the hippocampus of male mice. Representative of n = 4 (em-MG-baseline), n = 4 (mo-MG-baseline), n = 4 (Endog-MG-baseline), n = 4 (Tam Repop-baseline), n = 6 (em-MG-5dps), n = 5 (mo-MG-5dps), n = 5 (Endog-MG-5dps), n = 5 (Tam Repop-5dps), three independent experiments (baseline) and four independent experiments (5dps). Scale bar, 50 μm. (B) Macrophage morphology typing before and after seizures. em-MG, Endog-MG, and Tam Repop microglia were assigned one of four morphology types as described in Figure 2. mo-MG were typed as follows: baseline/homeostatic defined as mo-microglia with fewer than five primary processes (type 0), five to seven primary processes with minimal secondary branching (type 1), and greater than eight primary processes with extensive secondary branching (type 2). The percentage of cells of each type is shown before seizures (top) and after seizures (bottom) for all groups. em-MG (baseline) n = 3, mo-MG (baseline) n = 3, Endog-MG (baseline) n = 2, Tam Repop (baseline) n = 3, em-MG (5dps) n = 7, mo-MG (5dps) n = 5, Endog-MG (5dps) n = 5, Tam Repop (5dps) n = 5, data pooled from animals from three independent experiments (baseline) and five independent experiments (5dps). (C) Representative astrocyte GFAP expression (white) in the hippocampus of male mice with em-MG, mo-MG, Endog-MG, and Tam Repop microglia 5dps. Scale bar, 150 μm. (D) Quantification of area covered by GFAP-expressing astrocytes in males. Hipp, hippocampus; Stria, striatum; Ctx, cortex; Thal, thalamus. em-MG n = 8, mo-MG n = 6, Endog-MG n = 6, Tam Repop n = 4, data pooled from four independent experiments, one-way ANOVA with Tukey’s multiple comparisons. (E) Representative Fluorojade C staining (green) indicating neurodegeneration in the hippocampus of mice 5dps. Scale bar, 150 μm. (F) Quantification of Fluorojade C+ cells in the hippocampus and striatum of male mice. em-MG n = 7, mo-MG n = 5, Endog-MG n = 5, Tam Repop n = 5, data pooled from two independent experiments, one-way ANOVA with Tukey’s multiple comparisons. All data represented as mean ± SEM. ns, not significant; ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

Ontogeny and transplantation modulate microglial responses to excitotoxic challenge

We were intrigued that repopulation with mo- but not em-microglia prevented seizure phenotypes associated with microglia depletion in males. To identify potential mechanisms, we performed bulk RNA sequencing on FACS purified mo-microglia, transplanted em-microglia, and endogenous em-microglia 5 days after seizure, along with saline controls (Figure S6A). We first verified sample purity by confirming lack of expression of non-microglia/macrophage genes (Figure S6B). Consistent with prior studies,^3,22,24 we found a large number of differentially expressed genes (DEGs) between transplanted em- and mo-microglia (1,266) at baseline, and a smaller number in transplanted versus endogenous em-microglia (462, Figure S6C). We confirmed expected expression differences in ontogeny-specific genes such as Sall1 and P2ry12 (em-microglia), and Apoe, Ms4a7, and Clec12a (mo-microglia; Figures S6D and S6E). We also observed immune response-related genes upregulated in transplanted em-microglia at baseline (Figure S6F), suggesting that transplantation itself has long-term impacts on microglial gene expression.

Having validated baseline differences between em- and mo-microglia, we assessed seizure-elicited responses within each cell type. While mo-microglia and endogenous em-microglia had similar numbers of DEGs 5 days post-seizure (92 and 97, respectively), transplanted em-microglia had 203 (Figure 5A), suggesting that transplantation impacts magnitude or duration of microglia state change following induced seizures.

Figure 5.

Ontogeny and transplantation modulate microglial responses to excitotoxic challenge

(A) Total number of upregulated genes (red) and downregulated genes (blue) in KA-challenged as compared with saline populations. Total number of DEGs for each comparison is indicated above each bar. em-MG, transplanted microglia; mo-MG, transplanted monocytes; Endog-MG, unmanipulated endogenous microglia (cutoffs used: adjusted p < 0.05, logFC > 1 or <−1, and average TPM > 100 in at least one of the comparison groups). (B) PCA plot showing em-MG, mo-MG, and Endog-MG 5 days post-seizure. (C) Total number of DEGs (solid bars) and DEGs exclusive to post-seizure populations (dotted bars) between cell types 5 days post-seizure (cutoffs used as above). (D) Selected seizure-exclusive GO terms enriched in mo-MG as compared with em-MG after seizures. (E) Volcano plot showing DEGs in mo-MG versus em-MG after seizures. Dotted lines indicate cutoffs at log fold change equal to 1 and adjusted p value less than 0.05. (F) Gene expression values for selected DEGs in post-seizure mo-MG and em-MG. (G) Selected seizure-exclusive GO terms enriched in em-MG as compared with Endog-MG after seizures. (H) Volcano plot showing DEGs in em-MG versus Endog-MG after seizures. Dotted lines indicate cutoffs at log fold change equal to 1 and adjusted p value less than 0.05. (I) Gene expression values for selected DEGs in post-seizure em-MG and Endog-MG. For all RNA sequencing analysis, n = 2 (Endog-MG-saline), n = 3 (em-MG-saline), n = 3 (mo-MG-saline), n = 3 (Endog-MG-seizure), n = 4 (em-MG-seizure), n = 6 (mo-MG-seizure), data pooled from two independent experiments.

We next compared transcriptional states between cell types after KA challenge, noting first that cell ontogeny rather than KA treatment drove sample clustering by principal component analysis (PCA) (Figure 5B). We found many DEGs between mo-microglia and both transplanted and endogenous em-microglia, and a smaller number between transplanted and endogenous em-microglia (Figure 5C). Among them, 377 (37.2%—mo-microglia versus transplanted em-microglia), 689 (54.5%—mo-microglia versus endogenous em-microglia) and 91 (84.3%—transplanted versus endogenous em-microglia) were exclusive to the seizure condition, demonstrating that both ontogeny and transplantation affect responses to excitotoxic challenge.

To generate more specific molecular hypotheses, we tested for Gene Ontology (GO) term enrichment among the genes upregulated in mo-microglia following seizure and noted enrichment for multiple inhibitory immune response and wound healing gene sets (Figure 5D). We also noted that, among the top 100 seizure-exclusive DEGs, KA-challenged mo-microglia strongly upregulated genes related to lipid metabolism (Ascl4, Osbpl3), synapse regulation (Lrrtm2, Nrxn1, Slc1a2), potassium buffering (Kcnj10), and anti-inflammatory responses (Il10, Oasl1) (Figures 5E and 5F). On the other hand, transplanted em-microglia did not show enrichment for these pathways, and upregulated some genes related to inflammation and immune cell activation (Tnfrsf8, Il1rl1, H2-Oa) (Figures 5E and 5F). These results point to avenues for future study of shared and distinct functional responses between mo- and em-microglia to seizure challenge.

Focusing next on the impact of transplantation on em-microglia responses to seizures, GO term enrichment analysis using upregulated DEGs in transplanted cells identified innate immune and interferon beta response gene sets (Figure 5G). Also among the top 100 seizure-exclusive DEGs in transplanted em-microglia were upregulated inflammatory including Irak3, Cfb, Ifi213, and Mmrn2 (Figures 5H and 5I). Although much work is needed to determine if and how gene expression differences relate to the impacts of transplanted mo- and em-microglia on seizure phenotypes, these findings support a growing body of literature that ontogeny greatly influences the responses of brain macrophages to challenge, and that these responses are additionally influenced by transplantation itself.

Discussion

Microglia replacement, the depletion of endogenous microglia and their substitution with adoptively transferred surrogates, is a promising modality for treatment of CNS diseases across the lifespan, but is of immediate interest for rapid onset pediatric diseases such as leukodystrophies. For several leukodystrophies, HSCT is the standard of care, thought to provide therapeutic benefit by facilitating some level of microglia replacement by peripheral, HSC-derived donor macrophages.⁷ HSCT, however, requires high-dose chemotherapy or irradiation for successful donor macrophage engraftment in the brain, which causes CNS damage, including permanent loss of neurogenesis and microglia senescence.² As such, there is great interest in developing safer, faster, more specific and efficient approaches to microglia replacement, including by making use of small-molecule CSF1R antagonists.²³ With this bright future for CNS innate immune therapies there is a critical need to understand how microglia depletion and adoptive transfer of microglial surrogates affect the brain.

While modeling the impact of early microglia replacement on circuit excitability, we noted that postnatal microglia depletion itself increased adult susceptibility to severe seizures, associated with increased hippocampal excitability as measured by electrophysiology in live brain slices. The exacerbated seizure phenotype was furthermore associated with increased microglia perturbation, astrogliosis, and neurodegeneration 5 days after seizures. The finding that postnatal microglia depletion and repopulation has lasting impacts on brain functioning is consistent with prior studies demonstrating that microglia play important roles in healthy brain development through both prenatal and perinatal stages.^17,20,40,41 Interestingly, microglia depletion and repopulation in adulthood does not impact seizure susceptibility,⁴² suggesting that a critical developmental window exists in which microglia are required for proper neural circuit development.

Having described the impact of postnatal microglia depletion and repopulation on adult seizure phenotypes, we next tested the impact of microglia replacement, either with intracranially transplanted microglia (em-microglia), or monocytes, circulating bone marrow-derived cells capable of engrafting the brain and becoming microglia-like (mo-microglia). Although mo-microglia become very similar to em-microglia, their gene expression profile overlaps with disease reactive states,^22,24 including notably low expression of P2Y12, a purinergic receptor implicated in seizure inhibition.^26,27 We were therefore surprised to find that transplanted monocytes and not transplanted microglia prevented severe seizures and damage responses associated with three other forms of microglia repopulation (PLX3397-mediated, and tamoxifen-mediated with or without em-microglia transplant).

These results have several important implications. We used PLX3397 for CSF1R inhibition because it is approved for use in humans,^43,44 although it has off-target effects on other brain and immune cells.^32,33 Cross-validation with a genetic depletion model supports the idea that early-life depletion and repopulation of microglia negatively impacts seizure susceptibility in adults, but it will be important for future preclinical studies to test these phenotypes with more specific CSF1R inhibitors that may show reduced impacts on neurodevelopment. Additionally, we found that the PLX3397-mediated increases in seizure severity, as well as their rescue by transplanted mo-microglia, only affected males. While we did not track estrus cycle in female mice, which may explain the lack of observed phenotype, in future studies it will be important to more fully interrogate what, if any, sex-dependent differences exist in neonatal microglia replacement, especially given the potential for human therapies.

Our studies also highlight the impacts of donor cell type on the effects of microglia replacement. Given the result that repopulation of the brain niche with mo-microglia spared the severe seizure phenotypes observed with depletion and repopulation of em-microglia, we used RNA sequencing as a first step to uncover seizure-elicited responses. These transcriptomic data revealed distinct state changes in each cell type. Transplanted em-microglia had more DEGs than mo- and endogenous microglia 5 days post-seizure, suggestive of an increased or prolonged response to seizures. Genes upregulated by transplanted em-microglia were also enriched for signatures of abnormal inflammatory and stress responses, suggesting they may be uniquely sensitive to transplantation. Meanwhile, post-seizure mo-microglia had gene signatures associated with inhibition of immune responses and wound healing that could relate to their protective effects. Although we did not discover the mechanisms underlying their differential impacts on seizure, these data demonstrate that em-microglia and mo-microglia respond differently to excitotoxic injury, in line with studies showing distinct responses to other insults.^3,45 These data provide important avenues for future studies to understand the molecular mechanisms governing ontogeny- and transplantation-dependent impacts of macrophages on neuronal network development and seizure phenotypes. Identifying these mechanisms will not only reveal important basic aspects of microglia function and injury response, but also point to ways to improve cell therapies. For example, our studies utilized fully differentiated postnatal microglia as a donor cell population which, as opposed to a non-terminally differentiated cell type such as HSC-derived monocytes, may be more sensitive to the process of cell isolation, which in turn may prime them toward exuberant reaction to subsequent insults. Our findings regarding the effects of donor cell type on seizure underscore the need for further study into what makes an optimal donor cell for microglia replacement. Most immediately, investigating whether transplantation of less-differentiated yolk sac progenitors, IPSC-derived microglia, or HSCs reduces adult seizure phenotypes could reveal better surrogate cells for use in microglia replacement. Additional important variables to consider in future studies include the age of donor cells. On one hand, using donor cells from an adult may support expedited transplantation by allowing for donor cell sourcing prior to birth, following in utero genetic testing. On the other hand, donor cells isolated from older subjects may come with risks associated with aging. It will be vital for future preclinical studies to consider each these factors in ultimately determining the optimal donor cell to replace diseased microglia.

With the growing potential for microglia replacement to treat a number of pediatric neurological diseases, it is critical to understand the mechanisms by which (1) early-life microglia depletion and (2) repopulation with surrogate cells, impact the brain. Our studies serve as an important first step in identifying distinct functional responses to excitotoxic injury, and provide insight into crucial avenues of future study to fully realize microglia replacement therapies.

Limitations of the study

Our study has several important limitations. First, we did not identify the mechanisms that explain why microglia depletion worsens seizure responses, whether it is due to depletion, repopulation, or both, nor why mo-microglia but not em-microglia rescue this effect. Second, although our study uses large in vivo cohorts with orthogonal validation, it is unlikely but still possible that indirect effects of our microglia depletion methods contribute to observed phenotypes. Third, we found no effect on seizure severity following microglia replacement in females. Because estrus stage influences seizure susceptibility and was not tracked, we cannot conclusively determine that microglia replacement does not affect neural circuit development in females, which requires further testing. It is also critical to state that many additional translational steps are required to bring microglia replacement to humans. Despite these limitations, this study provides critical insights into how early microglia replacement affects neural circuit function in adulthood.

Materials and methods

Mice

All animal studies were performed with prior approval from the Children’s Hospital of Pennsylvania Institutional Animal Care and Use Committee panel in accordance with institutional and national regulations. All animals used were housed in a non-barrier facility with 12-h light-dark cycles at 23°C ± 2°C, in ventilated cages, and were provided standard chow and water ad libitum. For PLX3397 experiments, WT C57Bl/6J mice (JAX 000664) were obtained from JAX and then bred in-house. Cx3cr1-CreER+/+; Csf1r fl/fl mice were generated by crossing Cx3cr1-CreER+/+ mice (JAX 021160) with Csf1r fl/fl mice (JAX 021212). All experimental animals used were Cx3cr1-CreER+/−; Csf1r fl/fl. Donor cells for monocyte and microglia transplants were obtained from C57BL/6-Tg(CAG-EGFP)131Osb/LeySopJ “Osb-GFP” mice (JAX 006567).

Isolation of donor monocytes

Isolation of bone marrow-derived monocytes for transplantation was performed as previously described.²⁴ Briefly, femurs and tibias were dissected from adult Osb-GFP (aged 6–12 weeks). Whole bone marrow was flushed out of the bones with cold 1× PBS using a 1 mL syringe, spun down at 300 × g at 4°C, and then red blood cells were lysed with ACK lysis buffer for 5 min at room temperature. Total bone marrow cells were then resuspended in MACS buffer (2% BSA, 1 mM EDTA in 1× PBS) for enrichment of bone marrow-derived monocytes. Monocytes were enriched using a monocyte isolation kit following manufacturer’s instructions (Miltenyi, catalog no. 130-100-629). Finally, monocytes were resuspended in sterile 1× PBS for transplantation at 1.5 × 10⁵ cells/μL. In some experiments, isolated monocytes were further incubated with the CD11b microglia isolation kit. For these experiments, monocytes were first negatively selected with the MACS monocyte isolation kit (Miltenyi, catalog no. 130-100-629) as described above, then further selected for using the MACS CD11b microglia isolation kit (Miltenyi, catalog no. 130-093-636), following manufacturer’s instructions, before final resuspension for transplant.

Isolation of donor microglia

Isolation of microglia was also performed as previously described.²⁴ Briefly, P7-P16 Osb-GFP pups were anesthetized and intracardially perfused with 15 mL cold 1× PBS. Brains were then harvested and placed in cold medium A buffer (15 mM HEPES, 0.5% glucose, 250 U/mL DNAse in 1× HBSS). Brains were then homogenized using a cold glass Dounce homogenizer (using 3–5 pulls) and homogenized tissue was spun down at 300 × g at 4°C. Pellets were resuspended with 30% Percoll solution in HBSS (Percoll Plus, GE Healthcare, catalog no. 17544501) and spun at 700 ×g, 4°C, for 20 min (0 brake) to remove myelin. Following this Percoll spin, the top myelin layer was discarded and cell pellets were washed twice in medium A buffer, followed by resuspension in MACs buffer. Microglia were then enriched using the MACS CD11b microglia isolation kit (Miltenyi, catalog no. 130-093-636) following the manufacturer’s instructions. Enriched microglia were then resuspended in sterile 1× PBS for immediate transplantation at 1.5 × 10⁵ cells/μL. For some experiments, FACS was used to isolate microglia instead of MACS. For these experiments, the above protocol was followed to obtain a single-cell suspension, then microglia were washed twice in cold FACS buffer (1× PBS, 0.05% BSA, 0.5 mM EDTA). Then, microglia were incubated with Fc Block for 10 min at room temperature. Following Fc Block, microglia were incubated for 30 min at 4°C with the following antibody cocktail: PE-Cy7-conjugated CD45 (Clone 30-F11, BioLegend, catalog no. 103114), PE-conjugated CD11b (Clone M1/70, BioLegend, catalog no. 101207), and a live-dead stain (Far Red, ThermoFisher, catalog no. L34973). Cells were then washed twice and resuspended with cold FACS buffer. Microglia were then sorted using a FACSJazz or FACSMelody flow sorter (live cell singlets, CD11b+, CD45-intermediate, GFP+).

Microglia depletion and donor cell transplant

For PLX3397 experiments, C57Bl/6 pups were injected subcutaneously each day with 50 mg/kg PLX3397 (Selleck Chem, catalog no. S7818) from P1 to P7 to deplete microglia. For macrophage transplantation experiments, 100 mg/kg tamoxifen (Sigma, catalog no. T5648-1G) was administered subcutaneously to Cx3cr1-CreER+/−; Csf1r fl/fl pups on P1 and P2 to deplete endogenous microglia. On P3, GFP+ donor cells (isolated as described above) were transplanted, or a sham PBS injection given, intracranially using a glass microcapillary pulled pipette. Approximately 1 μL containing 1.5 × 10⁵ cells was injected bilaterally into each hemisphere of the cortex. Pups were then allowed to recover and donor cells to engraft in the brain for 2–3 months before seizure experiments were performed. Due to variability in engraftment for transplanted animals, inclusion criteria in all downstream analyses was predetermined to be at least 50% engraftment of donor cells in the hippocampus.

Repeated low-dose KA-induced seizure model and Racine scoring

To induce seizures, we followed the “expedited KA” model as previously described.³⁰ Briefly, animals were placed singly in cages and allowed to acclimate for 10 min. For KA administration, animals were given an initial intraperitoneal dose of 10 mg/kg KA, then a 5 mg/kg dose, then repeated 2.5 mg/kg doses (each dose separated by 20 min), until a stage 5 seizure was observed or a maximum of 6 doses given. Animals were observed up to 300 min following initial KA dose, and scored according to a modified Racine scoring system⁴⁶ as follows every 5 min: stage 1, absence, immobility; stage 2, hunching with facial automatisms and/or tail lifting; stage 3, rearing on hind legs (no falling), with facial automatisms and/or forelimb clonus; stage 4, repeated rearing and falling with continuous facial automatisms and forelimb clonus; stage 5, generalized tonic-clonic convulsions with lateral recumbence or “wild running/jumping” or death. At the end of 300 min, surviving animals were returned to their home cage and monitored for 5 days.

Immunofluorescence and histology staining

Deeply anesthetized mice were intracardially perfused with first 20 mL cold 1× PBS, followed by 20 mL cold 4% PFA (EMS, 15714). Following dissection, brains were drop-fixed in 4% PFA overnight at 4°C. Brains were then cryoprotected by placing in 30% sucrose solution for 2–5 days at 4°C, then frozen on dry ice in OCT blocks. Brain slices (14–16 μm thick) were cryosectioned onto Superfrost Plus slides (Fisher, 1255015) and stored at −80°C until staining.

For immunofluorescence staining, slides were placed in a staining chamber and rehydrated for 2–5 min with 1× PBS. Sections were then blocked for 1 h at room temperature using 10% donkey serum containing 0.5% Triton X-100. Primary antibodies were diluted in 1% donkey serum containing 0.5% Triton, and incubated on slides in the staining chamber overnight at 4°C. Primary antibodies used: rabbit anti-IBA1(1:500, Wako, 019-19741), goat anti-IBA1 (1:250, Abcam, ab5076), rabbit anti-GFAP (1:500, Agilent, Z033429-2), goat anti-PDGFRa (1:400, R&D Systems, AF1062), and rabbit anti-TMEM119 (1:300, Abcam, ab209064). After overnight incubation, slides were washed three times in 1× PBS for 5 min each. Secondary antibodies were also diluted in 1% donkey serum containing 0.5% Triton X-100 and incubated on slides for 1.5 h at room temperature in covered staining chambers. Slides were then washed three times (5 min each) in 1× PBS. Finally, coverslips were mounted using DABCO mounting medium containing DAPI for a nuclear stain (EMS, 17989-60) and sealed with quick-dry nail polish.

For synapse staining, floating 40 μm thick brain sections were cut and placed in 3 mL 1× PBS in a 24-well plate. Sections were washed once with 1× PBS before blocking in 1 mL blocking solution (10% donkey serum containing 0.5% Triton X-100) for 1 h at room temperature. Following blocking, sections were incubated in primary antibody solution (diluted in 1% donkey serum containing 0.5% Triton X-100) for 20–24 h at room temperature in the dark. Primary antibodies used: rabbit anti-VGAT (1:1,000; Synaptic Systems, 131002) and mouse anti-neuroligin 2 (1:500; Synaptic Systems, 129511). Sections were then washed three times (5 min each) with 1× PBS and incubated with AlexaFluor-conjugated secondary antibodies for 90 min at room temperature. Tissue sections were then mounted on slides and coverslips added with DABCO mounting medium containing DAPI and sealed with quick-dry nail polish.

For Fluorojade C staining for degenerating neurons, the Fluoro-Jade C (FJC), RTD Ready-to-Dilute Staining Kit (VWR TR-100-FJ) was used according to the manufacturer’s instructions for non-paraffin-embedded sections. Non-aqueous DPX mounting medium (Sigma, 1005790507) was used to mount coverslips.

Image acquisition and analysis

Stained tissue sections were imaged using a BZ-X800 fluorescent microscope (Keyence) for epifluorescence or a Leica SP8 confocal microscope equipped with white light laser and diode 405 laser for confocal images.

For microglia/macrophage morphology typing, representative 20× images of the cortex and CA1 of the hippocampus were used. First, images were imported into FIJI and converted to a thresholded, binary image. Each cell was then assigned an ROI and corresponding number, and a random number generator was used to select 10 cells to score within each image. Microglia were assigned to one of four morphology types according to the following criteria: baseline/homeostatic with highly complex processes (type 0), those with shorter, thicker, less complex processes (type 1), those with very large soma and many short, “fluffy”-appearing processes (type 2), and those with a rounded shape with few to no processes (type 3). For mo-MG, one of three morphology types was assigned as follows: baseline/homeostatic was defined as mo-microglia with fewer than five primary processes (type 0), mo-microglia with five to seven primary processes with minimal secondary branching (type 1), and mo-microglia with greater than eight primary processes with extensive secondary branching (type 2). A total of two to seven biological replicates were used in each group, and the scorer was blinded to condition. Following scoring, the percentage of cells of each type were calculated for each group and condition.

For quantification of GFAP area, brain regions of interest were outlined using the polygon tool in FIJI. Following background subtraction and thresholding (held consistent across all samples within an experiment), GFAP percent area coverage was calculated. The final percent area covered value was calculated by averaging two to three brain sections per animal.

For quantification of NeuN+ cells in the cortex, two to three 20× images of NeuN staining were acquired per mouse. Following background subtraction (maintained the same for all samples), the Cell Counter function in FIJI was used to count NeuN+DAPI+ cells. This count was then normalized to area. Because of the high density of neurons in the hippocampus it was difficult to count individual cells, therefore, for quantification of NeuN+ cells in the hippocampus, we first outlined the hippocampus using the polygon tool in FIJI. Following background subtraction and thresholding (settings were maintained equal among all samples) we measured the percent area covered by NeuN in this region.

For quantification of PDGFRa+ OPCs, two to three 10× tile scans were acquired for each mouse. The corpus callosum was outlined in FIJI using the polygon tool, then the Cell Counter function was used to count PDGFRa+DAPI+ cells.

For synapse density quantification, two to three 15 μm z stacks (0.5 μm step size) were acquired using a Leica SP8 confocal microscope. Image acquisition settings were kept consistent for all images within a staining set. Using Imaris, the surfaces function was used to create a surface for each synaptic marker, then the total number of colocalized pre- and postsynaptic surfaces was measured as a readout of structural synapse density.

For Fluorojade C quantification, the brain region of interest was outlined using the polygon tool in FIJI, and the area of this region was recorded. Fluorojade C+ cells were then counted using the Cell Counter tool in FIJI.

RNA extraction

Isolation of endogenous or donor microglia/monocytes was carried out as described above in isolation of donor microglia with FACS. Briefly, once a single-cell suspension was obtained, cells were incubated with Fc Block for 10 min at room temperature, then stained for FACS using CD45 (PE-Cy7, Clone 30-F11, BioLegend, catalog no. 103114), CD11b (PE, Clone M1/70, BioLegend, catalog no. 101207), and a live-dead stain (Far Red, ThermoFisher, catalog no. L34973). Donor cells were then sorted using a FACSJazz or FACSMelody flow sorter (live cell singlets, CD11b+, CD45-intermediate, GFP+), directly into 500 μL TRIzol LS Reagent (Fisher, 10296028) to lyse cells. Following sorting, cells were homogenized fully by vortexing for 1 min, then incubated for an additional 5 min at room temperature. After incubation, 0.2 volumes of chloroform (Sigma, C2432-500ML) were added to each tube and thoroughly mixed for 15 s, followed by 3 min of incubation at room temperature. After incubation, samples were centrifuged at 12,000 × g for 15 min at 4°C, resulting in a lower red phenol-chloroform layer, an interphase layer, and an upper colorless aqueous layer. The upper aqueous layer containing RNA was then transferred to a fresh tube and 1 volume fresh 70% EtOH was added and thoroughly mixed by vortexing. Then, 700 μL of this solution was added to an RNEasy spin column, and further extraction and purification was done using the QIAGEN RNEasy Micro kit (QIAGEN, 74004) according to the manufacturer’s instructions. Extracted RNA was then stored at −80°C until further sequencing.

Library preparation and sequencing

Sequencing libraries were prepared using the NEBNext Single Cell/Low Input RNA Library Prep Kit for Illumina (NEB, E6420) according to the manufacturer’s instructions, with 8 μL of RNA input per sample. Quantity and quality of cDNA libraries was determined using an Agilent TapeStation using High Sensitivity D5000 ScreenTape (5067-5592), and High Sensitivity D5000 Reagents (5067-5593). Libraries were then sequenced by the Children’s Hospital of Philadelphia Center for Applied Genomics Core Facility using a NovaSeq 6000 system (Illumina) with an S1 or SP Reagent Kit v.1.5 (2 × 100 bp). Data were then de-multiplexed and shared with us through BaseSpace (Illumina) as raw FastQ files.

RNA sequencing analysis

FastQ files were processed using the nf-core/rnaseq pipeline (v.3.17.0), a standardized bioinformatics best-practice workflow implemented in Nextflow (v.24.10.1).^47,48 Quality control of raw sequencing data was performed using FastQC (v.0.11.9, available online at: https://www.bioinformatics.babraham.ac.uk/projects/fastqc/) followed by MultiQC (v.1.14) for summarization of quality metrics across all samples.⁴⁹ Adapter sequences and low-quality bases were trimmed using Trim Galore (v.0.6.7), a wrapper around Cutadapt (v.3.4), with default quality threshold parameters (Phred score < 20) and minimum read length of 20 bp after trimming.⁵⁰

The preprocessed reads were aligned to the Mus musculus reference genome (GRCm39 primary assembly) using STAR (v.2.7.10a) with default alignment parameters.⁵¹ The reference annotation used was Ensembl release 111 (Mus_musculus.GRCm39.111.gtf.gz). Aligned reads were quantified at the gene level using featureCounts (Subread package v.2.4.5), counting only reads that mapped to exonic regions.⁵² Post-alignment quality control was performed using RSeQC (v.4.0.0) to assess read distribution across genomic features, gene body coverage, and to detect potential sample anomalies.⁵³ Duplicate reads were marked using Picard MarkDuplicates (v.2.27.4, available at http://broadinstitute.github.io/picard/) but were retained for quantification. The mapping rates, duplication rates, and gene assignment rates were assessed using MultiQC to identify potential outliers and technical biases. Sample-to-sample correlation analysis was performed to evaluate biological and technical reproducibility between replicates.

Following alignment and quantification, gene-level counts were extracted from the featureCounts output and processed using DESeq2 (v.1.36.0) in R (v.4.2.2).⁵⁴ Prior to differential expression analysis, low-abundance genes were filtered by requiring a minimum TPM (transcripts per million) count of 10 in at least 2 samples. Normalized counts were obtained using the variance stabilizing transformation method.

Enriched GO terms were identified by input of DEGs (with cutoffs of p < 0.05, logFC > 1, and average TPM >100 in at least one comparison group) into the Panther enrichment analysis tool in the GO Consortium (available at https://geneontology.org/docs/go-enrichment-analysis/).

Slice electrophysiology

All slice electrophysiology experiments were performed on 2- to 3-month-old male animals, and done according to previously published protocols.⁵⁵ Briefly, animals were anesthetized with isoflurane, brains were dissected and removed quickly (>1 min), and immediately placed in ice-cold oxygenated (95% O₂/5% CO₂) sucrose-based artificial cerebrospinal fluid (aCSF) containing: 202 mM sucrose, 3 mM KCl, 2.5 mM NaH₂PO₄, 26 mM NaHCO₃, 10 mM glucose, 1 mM MgCl₂, and 2 mM CaCl₂. Coronal hippocampal slices (350 μm thick) were then cut using a vibratome (VT1200S, Leica Microsystems, Buffalo Grove, IL). Slices were then transferred to a holding chamber filled with 32°C–33°C oxygenated (95% O₂/5% CO₂) control aCSF containing: 130 mM NaCl, 3 mM KCl, 1.25 mM NaH₂PO₄, 26 mM NaHCO₃, 10 mM glucose, 1 mM MgCl₂, 2 mM CaCl₂. Slices were allowed to incubate for at least 60 min, then were returned to room temperature before recording experiments.

Electrodes for recording fEPSPs were fabricated using borosilicate glass (World Precision Instruments, Sarasota, FL, no. 1B150F-4), and pulled to a tip resistance of 2–4 MΩ when filled with aCSF. The stimulating electrode used was a non-concentric bipolar electrode (World Precision Instruments, no. ME12206). fEPSPs in area CA1 were recorded by placing the recording electrode in the stratum radiatum while stimulating the Schaffer collateral pathway. To record fEPSPs in the dentate gyrus, this hippocampal subregion was separated into two pathways. In the inner molecular layer, the stimulating electrode was placed in the medial perforant pathway with a recording electrode in the inner molecular layer of the dentate (bottom third of the molecular layer). On the other hand, recordings in the outer molecular layer were performed by stimulating the lateral perforant pathway and recording from the outer layer of the molecular layer (upper third). Electrical stimuli were 100 μs in duration, and field potential input-output relationships were recorded using the following parameters: 50–500 μA stimulation, 50 μA increments, 8 s inter-stimulus interval. Recordings were obtained using an Axoclamp 900 A amplifier and pClamp10 data acquisition software (Molecular Devices, Sunnyvale, CA) filtered at 2 kHz. fEPSP data were analyzed using pClamp11.2. All statistical tests for fEPSP data were conducted using two-way repeated measures ANOVA with Sidak’s multiple comparison test in order to test for treatment effect and stimulation intensity effect.

Experimental design and statistical analysis

GraphPad Prism v.9.0.2 was used for data presentation and statistical analysis. For all figures, mean ± standard error of the mean is shown. ns, not significant; p ≥ 0.05, ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. Details of specific statistical tests are described in individual figure legends.

Data availability

RNA sequencing datasets from these studies are available at the Gene Expression Omnibus repository (GEO: GSE306032). Other data are available upon reasonable request to corresponding author.

Acknowledgments

We would like to thank the Children’s Hospital of Philadelphia Center for Applied Genomics for assistance with RNA sequencing, and the Children’s Hospital of Philadelphia Flow Cytometry Core for assistance with FACS. This work was supported by funding from the Children’s Hospital of Philadelphia Training Program in Neurodevelopmental Disabilities (T32NS007413-26) (to C.A.O.), NIH DP5OD036159 (to M.L.B.), NIH R01-NS-120960 (to F.C.B.), The Klingenstein-Simons fellowship in neuroscience (to F.C.B.), NIH R01-NS-129737-01 (to F.C.B.), the Silicon Valley Community Foundation grant 2023-33184 (to F.C.B.), NIH 5R01-NS-120099 (to A.S.C.), and NIH 5R21-NS-128745 (to A.S.C.).

Author contributions

Conceptualization, C.A.O., M.L.B., and F.C.B.; development of methodology, C.A.O. and F.C.B.; implementation and analysis of experiments, C.A.O. and E.M.L.; interpretation of results, C.A.O. and F.C.B.; designing methodology and interpretation of electrophysiology experiments, C.A.O., S.A.S.D., and A.S.C.; RNA sequencing analysis and data visualization, S.S. and C.A.O.

Declaration of interests

F.C.B. and M.L.B. are co-inventors on a patent filed by The Board of Trustees of The Leland Stanford Junior University (application 16/566,675) related to methods of microglia replacement. F.C.B. also holds shares in NovoGlia Inc. M.L.B. holds shares in Alector Inc.

Supplemental information

Table S1. DEG lists for each cell type comparison and condition

In each tab, a single comparison is shown. For each, only DEGs meeting the cutoff of log2FC > 1 or <–1, p_adj 100 are shown. The first tab includes a key to abbreviations used for each comparison.

Table S2. List of Gene Ontology (GO) terms enriched in each cell type comparison both before and after seizure

Each tab contains GO terms for a single comparison, with headings found at the top of each list indicating the cell type and condition being compared.