Excitatory Neuron-Derived Interleukin-34 Supports Cortical Developmental Microglia Function

Summary

Neuron-microglia interactions dictate the development of neuronal circuits in the brain. However, the factors that regulate these processes across development are largely unknown. Here, we found that IL-34, a neuron-derived cytokine, was upregulated in early development and maintained neuroprotective, mature microglia in the anterior cingulate cortex (ACC) of mice. IL-34 was upregulated in the second week of postnatal life and was expressed primarily in excitatory neurons. Excitatory-neuron specific deletion of IL-34 reduced microglia numbers and microglial TMEM119 expression and increased aberrant microglial phagocytosis of excitatory thalamocortical synapses in the ACC. Acute, low dose blocking of IL-34 at postnatal day 15 similarly decreased microglial TMEM119 and aberrantly increased microglial phagocytosis of synapses. Viral overexpression of IL-34 induced TMEM119 expression and prevented appropriate microglial phagocytosis of synapses. These findings establish IL-34 as a key regulator of neuron-microglia crosstalk in postnatal brain development, controlling both microglial maturation and synapse engulfment.

eTOC Blurb

While neurons and microglia (the brain’s resident immune cells) are known to talk to one another, the signals that regulate their complex, important interactions are not comprehensively defined. Devlin et al. identify a novel role for the neuron-derived cytokine interleukin-34 (IL34) in instructing the development and functional maturation of microglia.

Graphical Abstract:

Introduction

Microglia serve diverse roles in normal brain development including functions in vasculogenesis, neurogenesis, myelination, tissue repair, and synaptic remodeling¹. Microglia are tissue-resident macrophages and progress through different functional stages in development² dependent upon precise molecular cues they receive from other cells in their neural environment. These cues are still being defined and likely depend on the distinct brain region and/or specific circuits within a given region. In many brain regions, it is not until the second postnatal week of life in mice that microglia adopt a transcriptional and functional profile consistent with adult microglia^2,3; however, little is known about the precise timing of this transition, or the brain-specific signals that contribute to it⁴.

As macrophages, microglia depend on signaling through the colony-stimulating factor 1 receptor (CSF1r) for their proliferation, differentiation, and survival^5–7. For this reason, CSF1r is often targeted for microglial depletion experiments⁸, and blockage or genetic deletion of this receptor potently reduces microglia number in the brain at every stage of development⁹. Often overlooked is that there are two known ligands that signal through this receptor: CSF-1, which is expressed in all tissues, and the more recently identified interleukin-34 (IL-34), whose expression is limited to keratinocytes in the skin and neurons in the forebrain^10,11. Further, there is a spatio-temporal pattern of microglial dependence on, and presence of, CSF-1 and IL-34. In the embryonic and early postnatal mouse brain, CSF-1 is predominantly expressed in all brain regions, and the loss of CSF-1 alone is sufficient to deplete virtually all microglia. This is in contrast to the adult brain, where IL-34 mRNA and protein is highly expressed in the forebrain; thus roughly 70–90% of gray matter microglia in the forebrain depend on IL-34 signaling for survival, while microglia in the cerebellum and brainstem still depend on CSF-1^7,10–13. It is notable that these signals come from different cellular sources, as CSF-1is primarily expressed by microglia early in life² and later by oligodendrocytes and astrocytes, while IL-34 is expressed by neurons¹³. Thus, given these differences in timing, brain region, and cell source, it is highly unlikely these two ligands play a redundant role for microglial survival and differentiation, however no work to date has empirically tested this in vivo. There does exist some preliminary evidence that CSF-1and IL-34 drive distinct transcriptional programs in microglia¹⁴, but it is still unknown (1) when forebrain microglia become primarily dependent on IL-34, (2) how neurons, and which neurons, express IL-34, and (3) how IL-34 could differentially influence microglia cell state and function, in vivo, in cortical development.

Here, we demonstrate that IL-34 is developmentally upregulated in the second week of postnatal life in mice in multiple forebrain regions. We show that IL-34 expression in the cortex is regulated by neuronal subtype, with elevated mRNA expression in glutamatergic (VGlut1+) neurons compared to GABAergic (Gad2+) neurons. Constitutive genetic deletion of IL-34 impacted microglia numbers and developmental cell state as demonstrated by decreased TMEM119 protein and increased lysosomal content at P15. Excitatory neuron-specific IL-34 deletion in VGlut2^Cre;IL-34^fl/fl mice impacted microglia similarly to global deletion and caused increased phagocytosis of VGlut2+ synaptic material in the ACC. Acute, low-dose antibody blocking of IL-34 at P15 prevented drastic microglia loss but decreased the amount of homeostatic (TMEM^Hi/CD68^Lo) microglia, increased phagocytic microglia (TMEM^Lo/CD68^Hi), and increased aberrant microglial phagocytosis of synapses. Finally, viral overexpression of IL-34 in neurons at a developmentally inappropriate timepoint (P1-P8) prematurely increased microglial TMEM119 and decreased microglial engulfment of synapses.

Results

IL-34 expression is primarily regulated by development and neuronal subtype.

We examined the time course of IL-34 and CSF-1 expression in the developing mouse brain. Previous literature demonstrated that microglia in the embryonic brain depend solely on CSF-1 signaling for survival, whereas microglia in the adult brain depend on either IL-34 or CSF-1, depending on brain region^7,10–12. Despite this, little was known about when IL-34 becomes the primary signal for microglia in regions that express IL-34 in the adult brain (e.g. cortex and striatum)¹³. We collected brains from wild-type, C57Bl6/j male and female mice at 6 ages (postnatal day 7, 14, 20, 30, 38, 55) spanning early life, adolescence, and early adulthood in mice¹⁵. We measured IL-34 mRNA in three IL-34-dependent brain regions (Anterior Cingulate Cortex (ACC), Nucleus Accumbens (NAc), Amygdala (AMY)) and from the cerebellum (CBM), which is CSF-1-dependent (Figure 1A). We found no developmentally regulated expression of CSF-1 mRNA in any region, but a developmental increase in IL-34 mRNA, which began between postnatal day 7 (P7) and P14 in all regions except the cerebellum (Figure 1B). We collected mice at P8 and P15 and used ELISA to measure total IL-34 protein in the ACC and CBM and similarly found an increase in protein between P8 and P15 in the ACC, but no increase in the cerebellum (Figure 1C). We focus on ACC in all following experiments, as previous work in our lab has detailed a developmental period of microglia-neuron interactions in this brain region¹⁶. Together, IL-34 expression increases in IL-34-dependent brain regions in the second postnatal week.

Figure 1. IL-34 expression is primarily regulated by development and neuronal subtype.

(A) Experimental timeline for WT brain collection across development and brain regions punched for qPCR analysis.

(B) Quantification of IL-34 and CSF-1 mRNA levels in all four brain regions across six postnatal developmental timepoints. (n = 8 male and 8 female C57Bl6/j wild-type mice/age, P55 cerebellum excluded due to poor RNA quality, one-way ANOVA, data normalized to P7).

(C) Quantification of IL-34 protein using ELISA from tissue punches of the ACC and CBM. (n = 5 male and 5 female C57Bl6/j wild-type mice/age, two-way ANOVA age × brain region, Sidak’s post hoc test, main effect of brain region in legend).

(D) Experimental timeline for early postnatal viral injections and DREADDs experiments.

(E) Representative image of RNA-FISH stain including VGlut1 to distinguish excitatory neurons from inhibitory neurons.

(F) Quantification of IL-34 mRNA puncta in excitatory (VGlut1+) and inhibitory (Gad2+) neurons in control mice. (n = 4 mice/sex, data shown are animal averages of the mRNA expression of all neurons of that type, unpaired t-test).

Data in (B), (C) and (F) are from one independent cohort consisting of four or more litters. See also Figure S1.

To test whether neuronal activity or subtype influences IL-34 expression, we artificially increased neuronal activity using a designer receptor exclusively activated by designer drug (DREADDs) approach in the first postnatal week (Figure 1D). We injected the designer drug deschloroclozapine (DCZ) at postnatal day 8, 9, and 10. This approach reliably activated neurons in Gq-expressing mice compared to GFP controls as measured by increased Fos mRNA (Figures S1A–S1D). Using RNAScope, we stained and quantified IL-34 mRNA and found no difference between GFP control and Gq mice, suggesting that heightened neuronal activity did not increase overall IL-34 mRNA in the circuit (Figures S1E–F). We then quantified IL-34 mRNA at single-cell resolution using probes for Fos and Gad2 to distinguish recently active (Fos+, i.e. greater than 6 Fos puncta per cell adjusted for cell size) from inactive (Fos−), GABAergic (Gad2+) or glutamatergic (Gad2−) neurons (Figures 1G). We found higher expression of IL-34 in Fos+ (active) neurons in Gq, but not GFP, mice. (Figures 1H). Notably, we observed twice as much IL-34 in Gad2− (presumably excitatory) neurons versus Gad2+ inhibitory neurons. Thus, to follow-up, we used VGlut1 and Gad2 probes to precisely quantify IL-34 in glutamatergic (VGlut1+) and GABAergic neurons and confirmed that IL-34 is expressed two-fold higher in excitatory neurons (Figures 1E–F). We next tested whether activating only excitatory neurons impacted IL-34 expression using VGlut1-Cre mice and a cre-dependent DREADD virus (Figures 1I). We again saw no change in overall IL-34 (Figures S1J–K), but an increase in IL-34 in Fos+ neurons compared to Fos− neurons (Figures 1L). These data suggest there is a homeostatic setpoint for IL-34 that is unchanged by activation, while at the level of individual neurons, activity predisposes them to increase IL-34. It is worth noting that we did not measure protein or neuronal release, which may be impacted by activation and occlude our findings within cell somas.

Constitutive IL-34 deletion decreases adult microglial number and TMEM119 protein.

Previous reports show that adult mice lacking IL-34 (IL-34^LacZ/LacZ) have fewer microglia in the cortex, striatum, and hippocampus, but not the cerebellum or brainstem^10,11. We confirmed these findings in IL-34^LacZ/LacZ adult male and female mice, and observed a moderate, yet significant, reduction in Iba1+ cell number in IL-34^LacZ/+ heterozygous mice (Figures 2A–B). Microglial numbers in the cortex and striatum of IL-34^LacZ/LacZ mice were equivalent to cerebellum and brainstem numbers, suggesting that, when IL-34 is lost, a CSF-1-dependent population of microglia is maintained in these regions¹³. We found a reduction in TMEM119, a microglia-specific marker that is developmentally regulated and present in mature, homeostatic microglia^3,17 in the remaining microglia of IL-34^LacZ/LacZ mice, suggesting that these forebrain, CSF-1-dependent microglia are immature and non-homeostatic (Figures 2C). This diminished TMEM119 in IL-34^LacZ/LacZ microglia was similar to the low expression seen in CSF-1-maintained brainstem and cerebellar microglia (Figures 2D). In line with the hypothesis that there is developmental consequence of IL-34 loss on forebrain microglia, we re-analyzed bulk RNA-Sequencing on isolated microglia from WT and IL-34^LacZ/LacZ mice¹⁴ and found that IL-34^LacZ/LacZ microglia have elevated neonatal marker gene expression (Itgax, Apoe, Clec7a) and comparably low expression of adult markers (P2yr12, Tmem119, Olfml3)^2,3,18 (Figures 2E). Notably, IL-34^LacZ/LacZ microglia upregulate CSF-1, which is highly expressed in immature microglia^2,19, and suggests that, in the absence of the primary signal (IL-34), CSF-1 upregulation by microglia is a potential compensatory survival mechanism. These gene expression changes were not observed in CSF-1-deletion mice¹⁴. Additionally, many genes associated with microglia phagocytosis were dysregulated in forebrain IL-34^LacZ/LacZ microglia (C3ar1, Itgam, Axl)^20–22(Figures 2E). These data suggest that when IL-34 signaling is lost, remaining CSF-1-dependent microglia in the forebrain resemble cerebellar microglia and have potentially disrupted phagocytic capabilities. This supports the idea that there are distinct IL-34- and CSF-1-maintained subpopulations of microglia that are transcriptionally and functionally distinct.

Global IL-34 deletion impacts microglia state in the second postnatal week.

Our observation that IL-34 increases in the second postnatal week coincides with many known developmental microglial processes, including proliferation¹⁶, upregulation of TMEM119³, increased ramification²³, establishment of region-specific heterogeneity²⁴, and closure of a period of synaptic pruning in the ACC¹⁶. To test whether IL-34 functionally controls these processes, we generated and collected IL-34^LacZ/+ (heterozygous control) and IL-34^LacZ/LacZ (IL-34 deletion) mice at P8 and P15 (Figures 3A). We confirmed the developmental increase in IL-34 by staining for LacZ in IL-34^LacZ/+ mice and saw robust expression at P15 (Figure 2A). We next stained the ACC with Iba1 and TMEM119 and found that IL-34^LacZ/LacZ mice have a reduction in microglia at P8 and P15 and a reduction in TMEM119 relative to Iba1 at P15 (Figure 2B–E), suggesting that IL-34 is necessary for the developmental upregulation of TMEM119 in cortical microglia. We also observed a marked reduction in expression of another homeostatic protein (P2YR12) at both P8 and P15 (Figures 3B). To assess whether IL-34^LacZ/LacZ microglia have dysregulated phagocytic function we measured the volume of CD68 (lysosomal marker) within microglia (Iba1) and found that P15, but not P8, microglia in IL-34^LacZ/LacZ mice have elevated lysosomal content (Figure 2F–G). Cell volume was unchanged (Figures 3C). We also found a decrease in process complexity in IL-34^LacZ/LacZ microglia at P15 (Figure 2H–J). These results suggest that IL-34’s role in the brain extends beyond survival and has developmentally relevant functions for microglial proliferation, maturation, phagocytosis, and ramification.

Figure 2. Constitutive IL-34 deletion impacts microglia in the second postnatal week.

(A) Representative image of LacZ in the ACC of IL-34^LacZ/+ mice at postnatal day 8 (P8) and P15.

(B-C) Representative images of Iba1 and quantification of microglia number in the ACC of P8 and P15 IL-34^LacZ/LacZ and IL-34^LacZ/+ mice. (n = 5–8 mice/sex/age/genotype, two-way ANOVA age × genotype, Sidak’s post-hoc test, main effect of genotype and interaction in legend). Scale = 100μm.

(D) Method for masking the Iba1 channel to quantify TMEM119 fluorescence.

(E) Representative images and quantification of TMEM119 in the ACC of P8 and P15 IL-34^LacZ/LacZ and IL-34^LacZ/+ control mice (n = 5–8 mice/sex/age/genotype, two-way ANOVA age × genotype, Sidak’s post-hoc test, main effect of genotype and interaction in legend).

(F) Representative IMARIS 3D reconstructions of microglia (Iba1) and lysosomes (CD68) from ACC of P8 and P15 IL-34^LacZ/LacZ and IL-34^LacZ/+ control mice. Scale = 5μm for P8 microglia, 15μm for P15 microglia.

(G) Quantification of lysosomal content (volume of CD68 / total microglia volume * 100, n = 3 mice/sex/age/genotype, 4–6 cells analyzed per mouse, individual microglia represented by gray circles, animal averages represented by black dots, two-way ANOVA age × genotype, Sidak’s post-hoc test, main effect of genotype and interaction in legend).

(H-J) Representative images of individual microglia (Iba1) and quantification of ramification using sholl analysis. (n = 5–8 mice/sex/age/genotype, 6 cells analyzed per mouse, two-way ANOVA distance from soma × genotype, main effect of genotype in legend).

Data in (C), (E), (G), and (J) are all from one independent cohort consisting of eight litters. See also Figure S2–S4.

In the brainstem, a CSF-1-dependent region, we found no effect of IL-34 deletion on microglia number, TMEM119 protein, or P2YR12 protein (Figures 3D–F). We also found no change in neuron, astrocyte, or oligodendrocyte cell numbers (Figures 3H–J) in ACC, and bulk RNASequencing of whole forebrain from IL-34^LacZ/LacZ and IL-34^LacZ/+ control mice demonstrated downregulation of microglia genes (e.g. P2yr12, Csf1r, Tmem119) but no changes in non-microglial genes (Figures 3K). These data confirm that genetic deletion of IL-34 impacts microglia in the forebrain and not the brainstem, and does not affect the number or transcriptional profile of neurons, astrocytes, or oligodendrocytes^5,10.

Global IL-34 deletion imparts an anxiolytic phenotype at P15 and adulthood.

To test whether the functional changes to microglia in IL-34^LacZ/LacZ mice influenced behavioral outcomes (Figures 4A), we first tested neonatal ultrasonic vocalizations (USVs), which have been shown to be developmentally regulated in the second week of postnatal life and impacted by early-life microglial perturbations^16,25. We found no difference in call number at P8 but a decrease at P15 in IL-34^LacZ/LacZ mice (Figures S4B–C). IL-34^LacZ/LacZ mice also had increased weight at P8 and P15 (Figures S4F–G) and spent an increased percentage of time in the open arms of the elevated plus maze (EPM) in adulthood (Figures S4D–E), suggestive of an anxiolytic phenotype. There were no differences in locomotor activity or sociability (Figures 4H–I), but IL-34^LacZ/LacZ males made more errors during Barnes Maze reversal (Figures 4J–K), suggesting they have a moderate deficit in cognitive flexibility. Taken together, IL-34 deletion induces an anxiolytic phenotype at P15 (reduced vocalizations) that persists to adulthood (increased open arm time in the EPM). These findings make sense given the ACC’s role in regulating anxiety²⁶, and suggest that the microglia changes in IL-34^LacZ/LacZ mice impact behavior early in life and in adulthood.

Excitatory neuron-specific IL-34 deletion decreases microglia number and increases phagocytosis of synapses

To functionally link excitatory neuron-derived IL-34 to the microglia changes we observed in the global deletion mice, we generated VGlut2^Cre;IL-34^fl/fl mice and collected their brains at P15. This VGlut2^Cre genetic strategy predominantly affects excitatory thalamic neurons, thus testing whether IL-34 released from thalamic projections, i.e. synaptic terminals, into the ACC was sufficient to influence cortical microglia maturation and function. Comparable to the global IL-34^LacZ/LacZ mice, we found a reduction in microglia number and TMEM119 protein in VGlut2^Cre;IL-34^fl/fl mice (Figure 3A–C).

Figure 3. Excitatory neuron-specific deletion of IL-34 reduces microglia number, increases phagocytosis of synaptic material, and reduces overall synapse numbers in the ACC at P15.

(A) Representative images of Iba1 and TMEM119 in the ACC of P15 VGlut2^Cre IL-34^+/+, IL-34^fl/+, and IL-34^fl/fl mice. Scale = 100μm.

(B-C) Quantification of microglia number and TMEM119 in the three groups. (n = 2–4 mice/sex/genotype, one-way ANOVA, Sidak’s post-hoc test, main effect of genotype in legend).

(D) Previous findings detailing microglia-neuron interactions in the ACC during the second postnatal week. Conceptual diagram constructed based on data from Figure 4 in Block et al., 2022.

(E) Quantification of total lysosomal content in the ACC of P15 VGlut2^Cre IL-34^+/+, IL-34^fl/+, and IL-34^fl/fl mice. (n = 2–4 mice/sex/genotype, 4–6 cells analyzed per mouse, individual microglia represented by gray circles, animal averages represented by black dots, nested one-way ANOVA, Sidak’s post-hoc test, main effect of genotype in legend).

(F) Representative IMARIS reconstructions and quantification of microglia synaptic engulfment from VGlut2Cre IL-34+/+, IL-34fl/+, and IL-34fl/fl mice. (n = 2–4 mice/sex/genotype, 4–6 cells analyzed per mouse, individual microglia represented by gray circles, animal averages represented by black dots, nested one-way ANOVA, Sidak’s post-hoc test, main effect of genotype in legend).

(G) Thalamocortical synapse numbers are quantified by VGlut2 and PSD95 overlap.

(I-J) Representative images and quantification of Vglut2+/PSD95+ overlap. (n = 2–4 mice/sex/genotype, 3 sections imaged per animal, images represented by gray circles, animal averages represented by black dots, nested one-way ANOVA, Sidak’s post-hoc test, main effect of genotype in legend). Scale = 5μm.

Data in (B), (C), (E), (G), and (J) are all from one independent cohort consisting of four litters.

It is well established that microglia engulf aberrant or unnecessary synapses in neuronal development. This process requires precise regulation to ensure appropriate synapses are eaten, while necessary synapses are maintained²⁷. While some signals have been identified that act as “eat me” tags on synapses destined for destruction^22,28–31, very few “don’t eat me” signals have been discovered that inhibit synaptic engulfment, protecting structurally mature, functioning circuits³². Recent work from our lab has shown that microglia in the ACC at P8 exhibit elevated phagocytic capacity, actively engulf VGlut2+ thalamocortical presynaptic material, and stop by P10¹⁶ (Figure 3D). Because this timing coincides with the IL-34 increase (P8-P15, Figure 1), and because IL-34 deletion induces phagocytic, immature microglia (Figure 2F–G), we hypothesized that IL-34 may act as a “brake” on microglial synaptic pruning. To quantify engulfment of thalamocortical synapses (which are the neurons affected in our VGlut2^Cre;IL-34^fl/fl deletion strategy), we triple-stained for Iba1, CD68, and VGlut2 and 3D reconstructed individual microglia in IMARIS. We found elevated lysosomal content in VGlut2^Cre;IL-34^fl/fl mice (Figure 3E), and increased engulfment of excitatory presynaptic material in IL-34^fl/fl mice compared to IL-34^fl/+ and IL-34^+/+, confirming our hypothesis that excitatory neuron-derived IL-34 prevents microglial engulfment of synapses (Figures 3F–G and S4L). To check if microglial over-engulfment led to less excitatory synapses in the ACC, we stained for presynaptic VGlut2 and postsynaptic PSD95 and counted overlap of the two signals (Figure 3H). We found the expected decrease in IL-34^fl/fl mice (Figure 3I–J), providing strong initial evidence that excitatory neuron-derived IL-34 acts as a “don’t eat me” signal for thalamocortical synapses in the ACC.

Acute IL-34 inhibition at P15 mimics constitutive genetic loss of IL-34 and increases phagocytic microglia.

One limitation of our findings thus far was that there was a significant reduction in the number of microglia in IL-34 deletion mice (Figures 2B–C and 3C). Thus, it is reasonable to assume that the remaining microglia increase their phagocytic activity to compensate for the fact that there are less of microglia overall. To control for this and any off-target developmental effects of genetic deletion, we injected function blocking antibodies directly into the brain intracerebroventricularly (ICV) to block IL-34 (anti-IL-34), CSF-1 (anti-CSF-1), or control IgG¹² in a temporally precise manner at P15 (Figure 4A). We titrated antibody dose (1mg/kg in our study vs. 100mg/kg in original publication¹²) to reduce microglial cell death. We collected these mice 48 hours after surgery and found only a slight, non-significant decrease in microglia number in the ACC (Figure 4B–C), while there was still a significant decrease in TMEM119 in anti-IL-34-treated microglia (Figure 4D). The anti-IL-34 microglia were also less ramified (Figures 4E–F).

Figure 4. Acute, low-dose IL-34 inhibition at P15 mimics constitutive genetic loss of IL-34 without significant cell loss.

(A) Schematic of blocking antibody injection surgeries and tissue collection.

(B) Representative images of Iba1 and TMEM119 in the ACC of mice administered either control (IgG) antibody, anti-IL-34, or anti-CSF-1. Scale = 100μm.

(C-D) Quantification of microglia numbers and TMEM119 in the three groups. (n = 3–6 mice/sex/antibody, one-way ANOVA, Sidak’s post-hoc test, main effect of antibody shown in legend).

(E) Representative images of individual microglia (Iba1) and quantification of cell ramification from control, anti-IL-34, and anti-CSF-1 mice. (n = 5–8 mice/sex/treatment, 6 cells analyzed per mouse, two-way ANOVA distance from soma × genotype, main effect of genotype in legend).

Data from (C), (D), and (F) are from one independent cohort consisting of six litters. See also Figure S7.

We next tested whether decreased TMEM119 in anti-IL-34 microglia correlated with increased lysosomal activity. We noticed significant heterogeneity where some microglia were abnormally high in CD68 (suggestive of heightened phagocytic activity) and low in TMEM119 (Figure 5A). We quantified this by measuring expression of the two markers on a per cell basis and binning them into three categories: “Lo”, “Mid”, and “Hi” based on fluorescent value histograms (Figures S5A–B). Anti-IL-34 mice had fewer CD68^LoTMEM119^Hi microglia, an expression profile typical of homeostatic cortical microglia at P15, and more CD68^HiTMEM119^Lo phagocytic, non-homeostatic microglia (Figures 5B–C). There was also a decrease in TMEM119^Hi cells but no change in CD68^Hi cells (Figures S5C–D). We then sequenced isolated microglia and whole forebrain from control and anti-IL-34 mice (Figure 5D); anti-IL-34 treated microglia transcriptomes clustered separately from control microglia via principal component analysis (Figure 5E) and there were over 400 differentially expressed genes (DEGs) between these two groups. Several mature, homeostatic marker genes (e.g. P2yr12, Tmem119, Cd164) were upregulated in control microglia while anti-IL-34 microglia adopted an interferon-responsive inflammatory profile (Cxcl9, Ifit3b, Ccl5, Mx1). We observed moderate enrichment for “disease-associated” microglia genes in a-IL-34 microglia³³ (Figures S5E–F), but the gene signature was most similar to type-I-interferon-responsive microglia that shape cortical development in the first week of postnatal life³⁴ (Figures 5F and S5G). We calculated the “microglia developmental index” of the samples using our previously published methods¹⁹ and found that anti-IL-34 treated microglia had reduced transcriptional maturity (Figure 5G). Comparing whole forebrain differences revealed a similar DEG pattern to the isolated microglia, confirming that IL-34-blocking induces changes specific to microglia, and not astrocytes, oligodendrocytes, or neurons, all of which have been shown to express alternative receptors for IL-34, although these receptors have a much lower binding affinity to IL-34 compared to CSF1r^35–38 (Figure 5H). Anti-IL-34 whole forebrain samples also had enrichment for genes related to microglial response to Lipopolysaccharide (LPS), which make up an “Inflammation Index”¹⁹ (Figure 5I). This implies reduced IL-34 contributes to brain-wide inflammation. Together, preventing IL-34 signaling acutely at P15 phenocopies the effects of genetic IL-34 deletion, inducing phagocytic, immature microglia. We acknowledge the limitations that come with injecting blocking antibodies, which causes tissue damage and microglia activation. However, we feel confident in these data because they corroborate findings from the non-invasive, genetic IL-34 deletion experiments, and we observed no tissue damage in or around our analysis region as the injections were 2mm posterior to the ACC (Figures 6A).

Figure 5. Acute IL-34 inhibition at P15 decreases mature, homeostatic microglia and increases phagocytic microglia.

(A) Representative images of CD68^LoTMEM119^Hi and CD68^HiTMEM119^Lo microglia.

(B-C) Quantification of homeostatic and phagocytic microglia in the ACC of the three groups. (n = 3–6 mice/sex/antibody, data shown are a percentage of 12 cells measured per animal across 3 images, one-way ANOVA, Sidak’s post-hoc test, main effect of antibody in legend).

(D) Schematic of bulk RNASequencing of isolated microglia and whole forebrain.

(E) PCA plot of isolated microglia transcriptomes from a-gp120 control and a-IL-34 treated mice.

(F and H) Volcano plot showing differentially expressed genes between a-gp120 control and a-IL-34 treated microglia or whole forebrain (n=2–3 mice/sex/treatment, genes shown as significant passed a threshold of padj < 0.001 and LogFC > 1).

(G and I) Quantification of microglia developmental index and LPS-induced “inflammatory” index from whole transcriptome data of isolated microglia or whole forebrain using methods described in Hanamsagar et al., 2017. (n=2–3 mice/sex/treatment, unpaired t-test).

All data are from one independent cohort consisting of four or more litters. See also Figure S5 and S7.

IL-34 inhibition causes aberrant eating of VGlut2+ thalamocortical synapses.

We next tested whether IL-34-blocked microglia had elevated engulfment of VGlut2+ thalamocortical synapses. Consistent with the excitatory neuron-specific IL-34 deletion mice (Figure 3), we found increased lysosomal content in anti-IL-34 microglia and a two-fold increase in synaptic material contained within microglial lysosomes (Figures 6A–C and S6B). There was a decrease in synaptic engulfment in anti-CSF-1 mice, implying that blocking CSF-1 has the opposite effect on microglia. We also found that anti-IL-34 and anti-CSF-1 had opposite effects on cell volume (anti-IL-34 decreased volume while anti-CSF-1 increased cell volume) (Figures 6C). There was a reduction in VGlut2+/PSD95 synapses in anti-IL-34 treated mice, concordant with the synapse over-engulfment in these mice (Figures 6D–E). Anti-CSF-1 mice also had reduced synapse numbers, although this was presumably via a non-microglial mechanism since we observed decreased synaptic engulfment in these mice. The synapse decrease in anti-IL-34 mice was driven by a reduction in VGlut2+ presynaptic puncta (Figure 6F). In sum, IL-34 and CSF-1 signaling have differential effects on microglia function at P15, and blocking IL-34 increases thalamocortical synaptic engulfment at an inappropriate developmental window, leading to synapse loss

Figure 6. IL-34 Inhibition Causes Aberrant Eating of VGlut2+ Synapses.

(A-C) Quantification and representative images of lysosomal content and VGlut2 synaptic engulfment in control, anti-IL-34, and anti-CSF-1 treated mice. (n = 3–5 mice/sex/antibody, 4–6 cells analyzed per mouse, individual microglia represented by gray circles, animal averages represented by black dots, nested one-way ANOVA, Sidak’s post-hoc test, main effect of antibody in legend).

(D-F) Representative images of VGlut2 and PSD95 and quantification of synapses in control, anti-IL-34, or anti-CSF-1 brains. (n = 3–5 litters/sex/antibody, 3 sections imaged per animal, 1–2 animals per litter, images represented by gray circles, litter averages represented by black dots, nested one-way ANOVA, Sidak’s post-hoc test, main effect of antibody in legend).

(G) Schematic showing where hippocampal analysis was performed.

(H-I) Quantification and IMARIS representative images of engulfed VGlut2 synaptic material in control, anti-IL-34, and anti-CSF-1 treated mice. (n = 1–3 mice/sex/antibody, 4–6 cells analyzed per mouse, individual microglia represented by gray circles, animal averages represented by black dots, nested one-way ANOVA, Sidak’s post-hoc test, main effect of antibody in legend).

Data in (A), (B), (E), (F), and (H) are all from one independent cohort consisting of six litters. See also Figure S6 and S7.

We next tested whether these effects were generalizable to other IL-34-dependent brain regions by performing similar analyses in hippocampal CA1 in our blocking antibody-treated mice (Figure 6G). Consistent with our findings in cortex, we saw no effect of anti-IL-34 on microglia number in the hippocampus, but we did see a decrease in TMEM119 protein (Figures S6D–E). IMARIS reconstructions of hippocampal microglia revealed a significant increase in VGlut2 synaptic engulfment in anti-IL-34 microglia, similar to cortical microglia (Figure 6H–I), despite no change in microglial volume or lysosomal content (Figures S6F–G). These data suggest that IL-34 acts as a broad “don’t eat me” signal in both the ACC and hippocampus..

Acute IL-34 inhibition does not cause mass microglia death and repopulation.

To validate that the phagocytic, immature microglia in anti-IL-34 mice were not newly born microglia repopulating after microglial cell death, we generated P2yr12^CreER/+;tdtomato^fl/+ mice to tag microglia before surgery. We injected 4-hydroxytamoxifen (4OHT) at P14, followed by ICV control or anti-IL-34 antibodies at P15 and collected the mice 48 hours later (Figures 7A). We stained for Iba1 and RFP to determine what percentage of microglia after anti-IL-34 were present before surgery and found >90% of Iba1+ cells were also RFP+ in both control and anti-IL-34 mice (Figures S7B–C), confirming that the phagocytic microglia after anti-IL-34 administration are the same cells from before the antibody injection that have developmentally regressed.

We next sought to confirm that the microglia were not phagocytic just because they were eating dying microglia. We did not expect this to be the case, as there is evidence that astrocytes, not microglia, clear microglial debris following low-dose CSF1r-inhibition³⁹. We injected control or anti-IL-34 at P15 and collected mice 12- or 24-hours post-surgery. We observed an increase in microglia number in control mice 12 hours following surgery when compared to anti-IL-34 mice or even control mice at 24- or 48-hours post-surgery (Figures 7D, dotted line). We believe that this is microglia proliferation occurring acutely after surgery caused by either physical stress or anesthesia (isoflurane) which is known to cause inflammatory responses in aged microglia⁴⁰. We did not see this increase in microglia in anti-IL-34 mice and only observed a slight reduction in microglia from baseline at 24 hours post-surgery. There was an increase in phagocytic cups (which are rare in P17 mouse cortex), in anti-IL-34 mice at 12 hours post-surgery (Figures S7E–F). This corresponded to an increase in phagocytic, CD68^Hi/TMEM^Lo microglia (Figures S7H and S7J), but no change in homeostatic, CD68^Lo/TMEM^Hi microglia (Figures S7G).We believe this is due to a floor effect, as few control-treated microglia met the criteria for “homeostatic” this shortly after surgery. Finally, we observed a significant interaction between post-surgery time and antibody treatment in glial fibrillary acidic protein (GFAP) fluorescence demonstrating increased GFAP in anti-IL-34 mice 24 hours post-surgery, which is temporally aligns with the slight decrease in microglia and suggests to us that astrocytes are eating any dying microglia (Figures 7I) in support of previous literature³⁹, although we did not rigorously confirm this. In sum, our ICV injection causes microglial proliferation in control mice 12 hours post-surgery, an effect that is blunted by anti-IL-34. Anti-IL-34 increases phagocytic microglia 12 hours post-surgery, and any microglial death 24 hours post-surgery coincides with reactive astrocytes. We believe that these data confirm that the anti-IL-34 microglia phenotypes are not a consequence of microglial engulfment of dying microglia immediately after surgery.

IL-34 overexpression at P1 increases TMEM119 and decreases synaptic engulfment at P8.

Finally, we tested the sufficiency of IL-34 signaling to control microglia maturation and synapse pruning. We took advantage of the fact that IL-34 is lowly expressed at P8 (Figures 1 and 2) and virally overexpressed IL-34 in neurons in the PFC of P1 mice (Figure 7A). This increased IL-34 protein at P8 in the ACC (Figure 7A) and increased microglia number and TMEM119 protein (Figures 7C–E). This is notable, as TMEM119 is not expressed in cortical microglia until the second postnatal week in mice³, suggesting that overexpressing IL-34 before it is endogenously produced is, at least partially sufficient to accelerate the maturation of microglia. Next, we used IMARIS to measure lysosomal content and VGlut2 synaptic engulfment. IL-34 overexpression reduced lysosomal content and thalamocortical synaptic engulfment (Figures 7F–H and S7K), resulting in an increase in Vglut2+/PSD95 synapses (Figures 7I–J). Finally, we observed a significant increase in microglial ramification in AAV-IL-34 mice (Figure 7K).

Figure 7. IL-34 Viral Overexpression at P1 increases TMEM119 expression and reduces microglial engulfment of synapses.

(A) Schematic of viral injections and tissue collection.

(B) IL-34 ELISA data measuring IL-34 protein in control and AAV-IL-34 mice. (n = 4–5 mice/group, unpaired t-test).

(C-E) Representative images of Iba1 and TMEM119 and quantification of microglia numbers and TMEM119 in the ACC of mice injected with either control GFP or AAV-IL-34. (n = 3–5 mice/sex/virus, unpaired t test). Scale = 100μm.

(F) Quantification of microglia lysosomal content and VGlut2 synaptic engulfment and IMARIS representative images in control GFP and AAV-IL-34 mice. (n = 3–5 mice/sex/virus, 4–6 cells analyzed per mouse, individual microglia represented by gray circles, animal averages represented by black dots, nested t-test).

(I) Representative images and quantification of Vglut2+/PSD95+ overlap in control GFP or AAV-IL-34 brains. (n = 3–4 litters/sex/antibody, 3 sections imaged per animal, 1–2 animals per litter, images represented by gray circles, litter averages represented by black dots, nested t-test).

(K) Quantification of microglial ramification in control GFP and AAV-IL-34 mice. (n = 3–5 mice/sex/virus, 4–6 cells analyzed per mouse, two-way ANOVA distance from soma × virus, p values shown are main effect of virus).

Data in (B) is from one independent experiment consisting of three litters, and data in (D), (E), (F), (G), (J), and (K) are pooled from two independent experiments consisting of six total litters.

Discussion

It is increasingly recognized that microglia exhibit a vast array of functional states throughout the developing brain^18,41–43. We provide evidence that IL-34, a neuron-derived cytokine that signals through the CSF1r on microglia, functionally matures cortical microglia in a discrete window of neurodevelopment. IL-34 is upregulated in the second week of postnatal life and IL-34 mRNA is higher in active, glutamatergic neurons in the ACC at P10. This developmental increase in IL-34 expression in the ACC corresponds increased microglia numbers, ramification, TMEM119 expression, and a decrease in phagocytic capacity. Global genetic deletion and neuron-specific deletion of IL-34 prevents these microglial developmental processes from progressing and causes overeating of excitatory synapses in the ACC. Acute blocking of IL-34 at P15 increases immature, phagocytic microglia, decreases ramification and TMEM119 expression, and re-activates microglia to eat thalamocortical synapses during an inappropriate developmental window (P15). In turn, virally overexpressing IL-34 in neurons early in development (P8) is sufficient to accelerate microglial functional maturity and inhibit synaptic pruning. These results establish IL-34 as a “brake” for microglial phagocytosis in the context of synaptic pruning, as well as unveiling its greater role in fine-tuning microglial identity and functional state in development.

We found that IL-34 expression is higher in Fos+, glutamatergic neurons. Despite these exciting initial findings, follow up work is needed to define its release dynamics from neurons. We were limited with our RNAFISH experiments to assessing mRNA only within the nucleus, but there is transcriptomic data suggesting that IL-34 mRNA is present at the synapse and may be locally translated there⁴⁴. This is relevant since our finding that VGlut2-specific deletion of IL-34, which results in the loss of IL-34 in excitatory thalamic projection neurons, was sufficient to reduce microglia numbers and increase engulfment of synapses in the cortex, suggesting the potential for thalamic neurons to influence cortical microglia via synaptic IL-34 release. Additionally, we found that blocking IL-34 specifically increased microglia engulfment of presynaptic (VGlut2+) terminals in the ACC. This does not, however, inform what role it may play in engulfing post-synaptic material, which microglia have been shown to do⁴⁵, or other neuronal components such as spines or cell bodies in other brain regions. As a secreted protein, it is unlikely that IL-34 is tagging individual synapses for engulfment like complement²². Thus, based on our data, we believe that IL-34 has a broad effect on local microglia once the circuit has stabilized. The discovery that IL-34 gene expression is lower in GABAergic neurons compared to glutamatergic neurons at postnatal day 10 was not surprising to us. While the timing of putative inhibitory synaptic pruning in the ACC is not well described, GABAergic pruning generally occurs later in development than excitatory synaptic pruning⁴⁶, in part due to the delayed maturation of GABA circuits as the GABA switch happens in the second postnatal week⁴⁷. It stands to reason that GABAergic neurons may upregulate IL-34 expression directly following their circuit maturation, and this should be addressed in future work.

While both IL-34 and CSF-1 bind to the same receptor on microglia, our findings demonstrate that they have distinct effects on microglial functional differentiation, with separate roles in development. Prior to our work, Kana et al., demonstrated that microglia in culture treated with IL-34 vs. CSF-1 have different transcriptional responses, but no data existed that tested the role of these two molecules on microglia function in an intact, developing brain^14,48. IL-34 binding to CSF1r in macrophages leads to increased phosphorylation of downstream targets in the MAPK pathway, while CSF-1-binding preferentially phosphorylates members of the STAT and ribosomal S6K pathways⁴⁹, which may provide one explanation for their differential effects on cell function. Additionally, IL-34 polarizes macrophages to a less “activated” state, by increasing IL-10 and decreasing phagocytosis in vitro⁵⁰, and microglia in gray matter are sensitive to IL-34 blocking or ablation, while microglia in white matter, cerebellum, and brainstem are sensitive to CSF-1 blocking or ablation^12,13. This is consistent with the finding that white matter microglia exhibit different functional and transcriptomic properties compared to gray matter microglia^51,52.Our experiments demonstrate that IL-34 signaling is critical not only for gray matter microglial survival and proliferation at these early stages of life, but also for dictating the proper functional development of these cells.

Microglia number and TMEM119 in the cortex of IL-34^LacZ/LacZ mice were indistinguishable from cerebellar microglia, suggesting CSF-1 maintains a subset of microglia in the cortex when the primary signal (IL-34) is lost^5,11. The question remains, however, whether two distinct populations of microglia are maintained solely by IL-34 or CSF-1 in the intact brain. This is unlikely given that CSF-1 is still lowly expressed in the forebrain of wild type mice¹³. Thus, we hypothesize that IL-34/CSF-1 functional regulation exists along a spectrum – where the balance of total IL-34 vs. CSF-1 present in the system can tip cells back and forth between various states. This suggests that IL-34/CSF-1 signaling can differentially modulate microglial function across various developmental and disease states and warrants future studies that interrogate the potential role for IL-34 signaling in diseases where microglia play a causal role.

A stop-gain mutation in the IL-34 gene may confer risk for Alzheimer’s disease (AD) in humans^53,54. Further, IL-34 is decreased (while CSF-1 is increased) in Alzheimer’s disease and in animal models^55,56, and IL-34 infusion is sufficient to rescue associative learning deficits in APP/PS1 mice⁵⁷. In addition to the potential implications for AD, an abnormal balance in IL-34 and CSF-1 signaling has been demonstrated in several neurodegenerative diseases, including Huntington’s, multiple sclerosis, amyotrophic lateral sclerosis, and chronic neuropathic pain^58–62. The basic biological mechanisms uncovered from our studies provide a foundation for future experiments to interrogate the functional role of IL-34 signaling in these conditions and position it as a potential therapeutic target in a range of degenerative conditions.

We observed an unexpected decrease in synapses in the anti-CSF-1 mice, despite a decrease in microglial engulfment of pre-synaptic structures compared to control-treated mice (Figure 6). This suggests that CSF-1 may be important in other aspects of synaptic development. It is possible that our anti-CSF-1 manipulation is directly impacting neurons, since there is evidence that neurons can express the CSF-1 receptor and that signaling through this receptor is protective³⁵, however we did not rigorously test this hypothesis. There are other reports suggesting CSF-1 is produced by neurons, but only in the context of stress⁶³, injury⁵⁸, or ethanol withdrawal⁶⁴, and that this upregulation has negative consequences for neurons and synaptic transmission. These findings are in line with our observations that blocking CSF-1 decreased microglial engulfment of synapses and suggest that CSF-1 may be primarily involved in maintaining immature phagocytic microglia. More work is needed to tease out the precise effects of CSF-1 and IL-34 signaling in different cell types in different contexts. Our data also demonstrate a need for greater attention to and interpretation of results utilizing CSF1r inhibitors. It sets a precedent for modeling different microglial functional states in vitro using different growth factors (CSF-1 vs. IL-34) and demands additional scrutiny of these models for replicating an intact brain environment (with homeostatic levels of these cytokines across contexts)⁶⁵.

To summarize, our results unveil a mechanism by which neurons communicate with microglia to control their function in postnatal development. In addition to promoting survival, IL-34 signaling to microglia acts as an important cue that matures microglia and prevents over-eating of thalamocortical synapses. These data shed light on the functional relevance of IL-34, making it a unique and interesting target for future studies of microglia and neurons.

Limitations of the study

While our blocking antibody approach complemented our genetic deletion data and prevented significant microglia cell loss, there were some limitations in that the manipulation lacked cell-type specificity - it blocked binding of all cellular sources of IL-34 and was not specific to microglia. Additionally, perhaps the largest limitation of our study was that we have yet to identify the mechanism by which IL-34 signaling programs microglia function. While we expect that the CSF-1 receptor is involved, we were unable to empirically test this because of the receptor’s necessary role in microglial survival. Further, questions remain regarding how and why IL-34-deficient microglia overeat excitatory synapses. We hypothesize that it is by programing microglia to change their “sensitivity” to other, more specific signals including: C3/C1q, Sirpa/CD47, CX3CR1/CX3CL1, Adam10, or Trem2/externalized phosphatidylserine^{22,27,29,31,32,66} however we did not experimentally confirm this. Finally, we did not overexpress CSF-1, so we are unable to confirm that CSF-1 overexpression in a cell type where CSF-1 is not normally expressed (neurons) would have a different effect.

Resource Availability

Lead Contact

Requests for further information and resources should be directed to and will be fulfilled by the lead contact, Dr. Staci Bilbo (staci.bilbo@duke.edu).

Materials Availability

This study did not generate new unique reagents.

Data and code availability

Bulk RNASequencing data generated is deposited in GEO and is publicly available from the date of publication. Accession number is listed in the Key resources table. All original code for analysis is available on Github (repositories and DOIs provided in STAR methods and Key resources table). Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Key resources table.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Guinea pig anti-VGlut2 1:2000	Synaptic Systems	Cat#135404; RRID: AB_887884
Chicken anti-IBA1 1:1000	Synaptic Systems	Cat# 234 009, RRID:AB_2891282
Rat anti-CD68 1:1000	Biolegend	Cat# 137001 (also 137002), RRID:AB_2044003
Chicken anti-Beta Galactosidase (LacZ) 1:1000	Aves Labs	Cat# BGL1010, RRID:AB_2313508
Guinea pig anti-TMEM119 1:1000	Synaptic Systems	Cat# 400 004, RRID:AB_2744645
Rabbit anti-PSD95 1:250	Thermo Fisher Scientific	Cat# 51-6900, RRID:AB_2533914
Rabbit anti-c-Fos 1:1000	Millipore	Cat# ABE457, RRID:AB_2631318
Rabbit anti-P2y12 1:2000	Anaspec	Cat#AS-55043A: RRID: AB_2298886
Rabbit anti-Sox9 1:2000	Millipore	Cat# AB5535, RRID:AB_2239761
Mouse anti-Olig2 1:2000	Millipore	Cat#MABN50; RRID: AB_10807410
Guinea pig anti-Neun	Synaptic Systems	Cat# 266 004, RRID:AB_2619988
Rabbit anti-RFP	Rockland	Cat# 600-401-379, RRID:AB_2209751
Rabbit anti-GFAP	Abcam	Cat# ab7260 RRID: AB_305808
Anti-gp120 ragweed control blocking antibody	Easley-Neal et al., 2019	Genentech (Gift)
Anti-IL34 blocking antibody	Easley-Neal et al., 2019	Genentech (Gift)
Anti-CSF1 blocking antibody	Easley-Neal et al., 2019	Genentech (Gift)
Bacterial and virus strains
AAV8-hSyn1-hM3D(Gq)-mCherry	Bryan Roth (Unpublished)	Addgene Cat#50474
AAV8-hSyn1-EGFP	Bryan Roth (Unpublished)	Addgene Cat#50475
AAV5-hSyn1-eGFP-T2A-mIL34-FLAG-WPRE	Anne Schaefer (Unpublished)	N/A
AAV5-hSyn1-eGFP-WPRE	Anne Schaefer (Unpublished)	N/A
AAV8-hSyn-DIO-mCherry	Bryan Roth (Unpublished)	Addgene Cat#50459
AAV8-hSyn-DIO-hM3D(Gq)-mCherry	Bryan Roth (Unpublished)	Addgene Cat#44361
Biological samples
Chemicals, peptides, and recombinant proteins
TRIzol Reagent	Thermo Fisher Scientific	Cat#15596026
DAPI	Thermo Fisher Scientific	Cat#D1306
Fluoromount-G	Thermo Fisher Scientific	Cat#00-4959-52
Deschloroclozapine (DCZ)	Tocris	Cat#7193
Vectashield PLUS Antifade Mounting Medium w/ DAPI	Vector Laboratories	Cat# H-2000-10
4-hydroxytamoxifen (4OHT)	Sigma-Aldrich	Cat#H6278
Cell Lysis Buffer 2	RND Systems	895347
Chloroform	Sigma-Aldrich	Cat#C2432
Normal Goat Serum	VWR	Cat#102038-610
Triton X-100 Surfact-Amps	Thermo Fisher Scientific	Cat#28314
Tris-buffered Saline (TBS)	Thermo Fisher Scientific	Cat#28358
2-Propanol	Sigma-Aldrich	Cat#I9516
Qiagen QuantiTect Reverse Transcription kit	Qiagen	205311
Critical commercial assays
IL34 ELISA	RND Systems	Cat#M3400
RNAScope	ACD Bio	RNAScope Multiplex Fluorescent V1 Assay
HCR RNAFISH	Molecular Instruments	HCR RNA-FISH Bundle with custom probes for Gad2, Fos, and IL34
Bradford Protein Assay	BioRad	Cat#5000111
Deposited data
Bulk RNASequencing Data	This Paper	GEO: GSE290856
Bulk RNASequencing Data	Kana et al.14	GEO: GSE133362
Experimental models: Cell lines
Experimental models: Organisms/strains
WT Mice (C57BL/6J)	Jackson Labs	Cat#000664
IL34 KO Mice	Dr. Daniel Saban (Gift)	Greter et al., 2012
P2Y12CreER mice	INIA Consortium (Gift), Jackson Labs	Cat#034727
Ai14 Mice	Jackson Labs	Cat#007914
Slc17a7-IRES2-Cre (VGlut1-Cre) Mice	Jackson Labs	Cat#037512
Vglut2-ires-cre knock-in (C57BL/6J) Mice	Jackson Labs	Cat#028863
Oligonucleotides
qPCR IL34: Forward	Harvard Primer Bank ID:18921437a1	TTGCTGTAAACAAAGCCCCAT
qPCR IL34: Reverse	Harvard Primer Bank ID:18921437a1	CCGAGACAAAGGGTACACATTT
qPCR CSF1: Forward	Harvard Primer Bank ID:192801a1	ATGAGCAGGAGTATTGCCAAGG
qPCR CSF1: Reverse	Harvard Primer Bank ID:192801a1	TCCATTCCCAATCATGTGGCTA
qPCR 18s: Forward	This paper	GAATAATGGAATAGGACCGC
qPCR 18s: Reverse	This paper	CTTTCGCTCTGGTCCGTCTT
Recombinant DNA
Software and algorithms
Ethovision	Noldus Technology	RRID:SCR_000441; https://www.noldus.com/ethovision-xt
Prism 9.0.0	GraphPad Software, Inc.	RRID:SCR_002798; https://www.graphpad.com
MUPET v2.1	Van Segbroeck et al., 2017	https://github.com/mvansegbroeck/mupet
FV31S-SW	Olympus Corporation	https://www.olympus-lifescience.com/en/support/downloads/
Imaris 9.5.1	Oxford Instruments	RRID:SCR_007370; https://imaris.oxinst.com/
Python 2019.3.3	Python Programming Language	RRID:SCR_008394; http://www.python.org/
Ilastik	Berg et al., 2019	RRID:SCR_015246; http://ilastik.org/
Avisoft-Ultrasound Gate recording software	Avisoft Bioacoustics	RRID:SCR_014436; http://www.avisoft.com/downloads/
MatlabR_2022a	Mathworks	RRID:SCR_001622; https://www.mathworks.com/
ImageJ	NIH	RRID:SCR_003070; https://imagej.nih.gov/ij/
Microglia image analysis pipeline custom script	This paper	DOI: 10.5281/zenodo.15528736
Sholl analysis python custom script	This paper	DOI: 10.5281/zenodo.15528742
Microglia tmem119/cd68 heterogeneity custom scripts	This paper	DOI:10.5281/zenodo.15528748
IL34 blocking sequencing analysis custom scripts	This paper	DOI: 10.5281/zenodo.15528752
RNA in situ single cell quantification custom scripts	This paper	DOI: 10.5281/zenodo.15528731
Other

STAR Methods

Experimental models and study participant details

Mice.

All procedures relating to animal care and treatment conformed to Duke Institutional Animal Care and Use Committee and National Institutes of Health guidelines (protocol no. A107–19-05 and renewal no. A062–22-03). Animals were group housed in a standard 12:12-hour light–dark cycle. In all experiments, both male and female mice were included at sufficient power (n ≥ 3 biological replicates). In cases where there was not a main effect of sex, male and female mice were combined for subsequent analysis. The age of the mice for every given experiment is specified in the figure and text. The C57BL/6J mouse line was obtained from Jackson Laboratory, stock no. 000664. The IL-34^LacZ/LacZ mice were a gift from Dr. Daniel Saban at Duke University and were originally generated by Dr. Marco Colonna¹⁰. VGlut1-Cre male mice (B6.Cg-Slc17a7^{tm1.1(cre)Hze/J}) were ordered from Jackson Laboratory, stock no. 037512 and bred in house with C57BL/6J females to produce VGlut1-Cre offspring for cre-dependent chemogenetics experiments. VGlut2-Cre mice (B6J.129S6(FVB)-Slc17a6^tm2(cre)Lowl/MwarJ) were ordered from Jackson Laboratory, stock no. 028863, and bred in house with IL-34^fl/fl mice (generous gift from Dr. Marco Colonna to Dr. Anne Schaefer^10,13) to produce VGlut2-Cre;IL-34^fl/fl offspring for excitatory neuron-specific IL-34 deletion experiments. Ai14 tdtomato^fl/fl mice were ordered as breeder females from Jackson Laboratory, stock no. 007914 and B6(129S6)-P2ry12^{em1(icre/ERT2)Tda}/J (P2yr12Cre^ER/+) mice were recently restored from cryopreservation by the INIA (Integrative Neuroscience Initiative on Alcoholism) consortium and were gifted to S.D.B. These mice are now available from Jackson Laboratory, stock no. 034727.

Method details

qPCR.

For the qPCR time course, wild-type mice were sacrificed and saline perfused at the six described ages (P7, P14, P21, P30, P38, P55) and their brains flash-frozen in 2-methylbutane in dry ice and stored at −80C. Tissue punches were taken from the anterior cingulate cortex, nucleus accumbens, amygdala, and cerebellum and stored in 500μL TRIzol (ThermoFisher, 15596026). RNA was extracted from the tissue using TRIzol based chloroform extraction, followed by isopropanol precipitation. Resulting RNA concentration was measured using a NanoDrop spectrophotometer (ND-1000) and either 200 ng or 1000 ng (depending on starting concentration) of RNA were reverse transcribed into cDNA using a Qiagen QuantiTect Reverse Transcription kit (205311). qPCR was performed on an Eppendorf Realplex ep Mastercycler using Sybr/Rox amplification with a QuantiFAST PCR kit (204056). Primer sequences are listed in Supplementary Table 1. All samples were loaded in triplicate and all plates were run on the same day to ensure comparability across plates. Fold change was calculated using the 2^−ΔΔCT method with 18S as an endogenous control for sample normalization.

IL-34 ELISA.

For all ELISA tissue collection, mice were anaesthetized with CO₂ and transcardially perfused with ice-cold saline prior to brain collection. For tissue punches, brains were flash-frozen in dry ice and embedded in O.C.T. (Sakura Finetek) and stored at −80C. Tissue was sectioned on the Leica CM50 cryostat until brain region of interest was visible, then brain region was punched using a 1mm RapidCore tissue punch (Ted Pella) and frozen at −80C. Prior to the tissue-punch ELISA, tissue was homogenized in 300μL 1X PBS using the Dremel Tissue-Tearor (BioSpec, 985370–04). Then, 300μL Cell Lysis Buffer 2 was added to the samples (R&D Biosystems, 895347) and they were placed on an orbital shaker at 200rpm for 30 minutes before spinning down in a 4°C centrifuge at 13,200 rpm for 20 minutes. Samples were stored at −20C overnight or −80C long term. All samples were run according to the manufacturer’s instructions for the kit (R&D Biosystems, M3400). All protein was normalized to total protein in the sample using a standard bradford assay kit (BioRad, DC^™ Protein Assay Kit, 5000111), which was assessed the same day as the ELISA. The equation used to calculate pg/100mg protein is: (IL-34 concentration (pg/mL) *100) / (total protein (mg/mL) *1000).

Microglia Staining.

For all immunohistochemistry experiments, mice were perfused with ice-cold saline followed by 4% paraformaldehyde (PFA). Tissue was then post-fixed in 4% PFA for 24 hours prior to cryoprotection in 30% sucrose with 0.1% sodium azide. Brains were flash-frozen in 2-methylbutane in dry ice and stored at −80C until sectioned. Sections were collected at 40μM and kept free-floating at −20C in cryoprotectant until stained. Free-floating sections were washed 3X in 1X PBS prior to blocking in a 10% normal goat serum (NGS) solution with 0.3% Triton-X in 1X PBS for 1 hour. Sections were incubated in primary antibodies overnight at 4°C. The primary antibodies used were Chicken anti-Iba1 (Synaptic Systems, 234 009, 1:1000), Guinea Pig anti-TMEM119 (Synaptic Systems, 400 004, 1:1000), Rat anti-CD68 (Biolegend, 137002, 1:1000). The following day, sections were washed 3X in 1X PBS and incubated in secondary antibodies for 2 hours at room temperature prior to mounting on gelatin-subbed slides and cover-slipping with Vectashield plus DAPI (Vector Labs, H-2000–10).

Microglia Imaging and Analysis.

Images of microglial Iba1, TMEM119, and CD68 staining were acquired on a Zeiss AxioImager M2 at 20X magnification. 19 Z-Stacks at a 1μM step size were acquired and maximum intensity projections were used for all analyses. For Microglia Cell Counting, single channel, maximum intensity, Iba1 images were counted using the supervised machine learning tool Ilastik⁶⁷ with the Pixel + Object Segmentation pipeline. All automatic counts were validated by blinded hand counts. Object segmentation outputs were then combined and analyzed using custom python scripts. For Microglial TMEM119 expression, first, an Iba1 mask was generated using Ilastik’s pixel segmentation pipeline on single channel Iba1 images. This mask was used to control for any mean gray value differences that may be a result of decreased cell number or process complexity. Following mask generation, TMEM119 mean gray value within the Iba1 mask was calculated using a custom Fiji script. All details and code for these analyses can be found at github.com/bendevlin18/mgla_img_analysis_pipeline, DOI: 10.5281/zenodo.15528736. Finally, for Microglia 2D Sholl Analysis, using the Iba1 stain, individual microglia were selected for analysis from each image by an individual blind to sex, genotype, and age. Those individual microglia were put through the Ilastik pixel segmentation pipeline to generate binary segmentations of the images. These binary segmentations were processed using a custom python script based on the sholl analysis fiji plugin that skeletonizes the image and plots concentric rings from the soma of each cell, measuring the intersections of those rings with cell processes. All code for 2D sholl analysis is also available on GitHub. https://github.com/bendevlin18/sholl-analysis-python, DOI: 10.5281/zenodo.15528742.

Microglia Phagocytic Capacity Imaging and Analysis

For microglia phagocytic capacity, microglia (Iba1) and CD68 staining was imaged using a Leica SP8 upright confocal. Individual microglia were imaged at a 63X magnification with a Z-stack step size of 0.33 μM. These images were then imported into IMARIS (Oxford Instruments, v9.0) and surfaces rendered of the Iba1 and CD68 channels to generate 3D reconstructions of the microglia and their lysosomes. The phagocytic capacity calculation was as follows: (total CD68 volume within microglia / total microglial volume (Iba1)) *100.

Microglia Heterogeneity Quantification.

To quantify microglial heterogeneity, representative microglia were selected by a blind experimenter from every image using the Iba1 channel. Each cell was selected using the freehand selection tool in Fiji (ImageJ) and TMEM119 and CD68 mean gray value were measured within each selection. 4 cells were measured per 20X images, and 3 images were analyzed per animal. TMEM119 and CD68 measurements were then normalized to Iba1 expression and histograms representing the distribution of expression values were plotted to determine value cutoffs for high, low, and mid expressing cells of each marker (supp. Fig. 7b–c). Individual cells were assigned high, low, or mid based on these cutoffs, and percentages of cells (out of 12 total) falling into each category were calculated per animal. The entire python script used for this analysis is available on GitHub. https://github.com/bendevlin18/microglia_tmem_cd68_heterogeneity, DOI:10.5281/zenodo.15528748.

RNA Sequencing

Mice were saline-perfused, and their brains dissected and split at the midline using a razor blade. Whole forebrain was collected, and microglia were isolated from one half using a CD11b antibody-based procedure, according to published methods⁶⁸. RNA was extracted from whole forebrain and CD11b+ isolated microglia using the TRIzol based chloroform extraction and samples were transferred to MedGenome (Foster City, California) for library preparation and sequencing. Raw fastQ files were aligned to Mus Musculus GRCm38 mm9 using STAR (v2.7.5c) and featurecounts (v1.6.3) on the Duke Compute Cluster with custom bash scripts. Genes were filtered and only included in analysis if they were present at 10 counts in at least 4 samples. Differentially expressed genes were calculated using DESeq2⁶⁹ in R 4.4.0. All analysis scripts are available on GitHub. https://github.com/bendevlin18/IL34blocking_seq_analysis_2024, DOI: 10.5281/zenodo.15528752.

Neuron, Astrocyte, Oligodendrocyte Stain and Quantification

Free-floating sections were washed 3X in 1X PBS prior to blocking in a 10% normal goat serum (NGS) solution with 0.3% Triton-X in 1X PBS for 1 hour. Sections were incubated in primary antibodies overnight at 4°C. The primary antibodies used were Rabbit anti-Sox9 (Millipore, AB5535, 1:1000), Guinea Pig anti-Neun (Synaptic Systems, 266 004, 1:1000), Mouse anti-Olig2 (Millipore, MABN50, 1:250). The following day, sections were washed 3X in 1X PBS and incubated in secondary antibodies (1:200) for 2 hours at room temperature prior to mounting on gelatin-subbed slides and cover-slipping with Vectashield plus DAPI (Vector Labs, H-2000–10). To quantify overall numbers of neurons, astrocytes, and oligodendrocytes in the cortex and brainstem, images were acquired on a Zeiss AxioImager M2 at 20X magnification. 19 Z-Stacks at a 1μM step size were taken and maximum intensity projections were used for all analyses. Single channel, maximum intensity, images of each of the three stains were counted using Ilastik with the Pixel + Object Segmentation pipeline. All automatic counts were validated by blinded hand counts. Object segmentation outputs were then combined and analyzed using custom python scripts.

Synaptic Staining.

Samples were collected and processed as described above. Free-floating sections were rinsed in 1X TBST (0.2% Triton X-100) 3 times for 10 minutes prior to block. Sections were then incubated in 5% NGS in 1X TBST for 1 hour at room temperature. Sections were then incubated in primary antibody diluted in 5% NGS in 1X TBST overnight at 4°C. The primary antibodies used were Guinea Pig anti-VGlut2 (Synaptic Systems, 135 404, 1:2000), Rabbit anti-PSD95 (ThermoFisher, 51–6900, 1:350) for overall synapse counts, or Guinea Pig anti-VGlut2 (Synaptic Systems, 135 404, 1:2000), Rat anti-CD68 (Biolegend, 137002, 1:1000), and Chicken anti-Iba1 (Synaptic Systems, 234 009, 1:500). Following primary incubation, sections were washed 3X for 10 minutes in TBS and then incubated in secondary antibody diluted in 5% NGS in 1X TBST for 2 hours at room temperature. Sections were mounted on gelatin-subbed slides and cover-slipped with Fluoromount G (ThermoFisher, 00–4959-52).

Synaptic Imaging and Analysis.

Synapse images were acquired on an Olympus FV3000 inverted confocal. Images were taken at 60X magnification with a 1.6X optical zoom. A 4×4 tile scan with 4, 0.33 μM step Z-stacks were obtained. Synapse images were analyzed using SynBot⁷⁰. Synapses were identified by the colocalization of pre- (Vglut2) and post- (PSD95) synaptic puncta.

Synaptic Engulfment Imaging and Analysis.

Synaptic Engulfment images were acquired on an Olympus FV3000 inverted confocal. Images were taken at 60X magnification with a 2.3X optical zoom. A 2×2 tile scan with 100–110, 0.33 μM step Z-stacks were obtained. Whole, stitched image stacks were imported into IMARIS (Oxford Instruments, v9.0) and microglial surfaces were rendered using the Iba1 channel. Individual microglia were selected and reconstructed (up to 3/image) and used to mask the CD68 channel. A surface was created from each Iba1-masked CD68 channel and the CD68 surface was used to mask the VGlut2 channel. Surfaces were created of the CD68-masked VGlut2 signal and volumes of all surfaces were exported for analysis. The phagocytic capacity was calculated as described above. The volume of synaptic material engulfed was normalized to Iba1 cell volume.

Behavior.

Ultrasonic Vocalizations.

We performed ultrasonic vocalization testing as previously described²⁵. Briefly, pups were removed from their cage and individually placed in a cotton-lined cup within a sound-attenuating chamber under an Avisoft Condenser ultrasound microphone (Avisoft Bioacoustics CM16/CMPA). suspended four inches above the contained pup. The dam and any remaining littermates were removed from the testing room in the home cage. After 3 min of recording, pups were weighed and sexed before being returned to their home cage. After the dam attended to each returned pup, the home cage was returned to the colony room. Analysis of USV .wav files was done using the MATLAB (MatlabR_2022a) machine-learning program MUPET to quantify the number of calls, the energy of the calls, and all syllable information. Forty-unit syllable repertoires were generated for each dataset and manually inspected before USV similarity was calculated and heatmaps were generated.

Open Field.

Adult mice (P50–55) were placed in a 50 cm × 50 cm square enclosure with 38-cm-high walls and allowed to explore freely for 10 minutes. Behaviors (distance moved and velocity) were recorded and analyzed using EthoVision video tracking software (Noldus). Center avoidance was assessed by comparing time spent at the periphery of the chamber to time spent in the center (middle third of chamber).

Social Preference.

Adult mice (P60–65) were assessed for social preference using a three-chambered preference test. On test day, subject and stimulus animals were habituated to the testing room for a minimum of for 1 h before testing. Each mouse was then tested for the preference to investigate a novel object (rubber duck) versus a novel social stimulus (age and sex-matched stimulus animal). Mice were placed in the middle chamber with the rubber duck confined in a clear plexiglass cylindrical cage on one side of the test and the novel conspecific confined in an identical cage on the other side for 10 minutes. The test was recorded with a Logitech webcam, and all behavior was manually quantified using Solomon Coder by an observer blind to sex and treatment. The social preference score was calculated as: (time spent investigating social stimulus / total time spent investigating either stimulus) * 100.

Elevated Plus Maze.

Male and Female adult mice (P70–75) were tested individually in the elevated plus maze (each arm 10cm wide by 52 cm long) to measure anxiety-like behavior. On the day of testing, mice were habituated to the testing room for at least 1 hour prior to start. Males and females were tested on separate days. One by one, each mouse was placed in the center of the elevated plus maze and their movement throughout the maze was tracked for 5 minutes. Behaviors were recorded and analyzed using Ethovision (Noldus), and the endpoints measured are as follows: Distance Moved, Velocity, Time in Open Arms, Time in Closed Arms, Time in Center, # of Transitions from Open to Closed Arms. The “Percent in Open Arms” metric was calculated by (time in open arms / time in closed arms) *100.

Barnes Maze.

Male adult mice (P80–90) were tested for memory/cognitive flexibility on Barnes maze over a 5-day period. In the first 3 days, each mouse was placed in the center of a large (45cm, radius) circular platform and given 4 minutes to find and enter the target hole (1 of 20 holes with an escape hatch). This was repeated a total of 3 times for each mouse on each of the first 3 days (Training Phase), and the amount of time taken to find the target hole drastically decreased across the 3 days, indicating that the mice had learned the location of the target hole. On the fourth day, the target hole was moved to another quadrant of the maze, and the mice were tested for 3 trials on day 4 and 5 to measure their reversal learning potential. Ethovision (Noldus) was used to track their performance in this task, and the endpoints analyzed were Time to Escape, Total Nose poke Errors, Time Spent in Original Target Quadrant (reversal only).

P1 Viral Injections

Postnatal day 1 mouse pups were placed in a 15mL conical tube and anesthetized in an ice bath for 10 minutes. Once pups were unresponsive via toe pinch, they were removed from the conical tube and placed on an ice pack for the duration of the injections. Mice were injected with a glass pulled needle pre-filled with virus connected to a microinjector (KD Scientific, 788100). 200nL was injected over a 30 second period into each hemisphere, and prefrontal cortex was targeted by injecting 2mm posterior to the inferior cerebral vein and 1mm lateral to the sagittal sinus. For the DREADDs experiments, we injected one of the following viruses: AAV8-hSyn-hM3D(Gq)-mCherry (Addgene, 50474-AAV8), AAV8-hSyn-hM4D(Gi)-mCherry (Addgene, 50475-AAV8), AAV8-hSyn-DIO-mCherry (Addgene, 50459), AAV8-hSyn-DIO-hM3D(Gq)-mCherry (Addgene, 44361). For the IL-34 Overexpression experiments, we injected one of the following custom-made viruses: AAV5-hSYN1-eGFP-WPRE (control) or AAV5-hSYN-eGFP-T2A-mIL-34-FLAG-WPRE (IL-34 overexpressing).

P15 Blocking Antibody Injections

Mice were anesthetized using 4% isoflurane pretreated with Ketophen anesthetic (5mg/kg, S.C.), and fitted into a Stoelting stereotaxic apparatus for the duration of the surgery. The fur on the head was shaved and, following betadine and ethanol preparation of the skin, a single incision was made. The surface of the skull was cleaned and Bregma was identified. All blocking antibodies were obtained from Genentech¹², stored at 4°C and diluted to a working stock of 10mg/mL in sterile saline on the day of surgery. A 10uL 1701RN Hamilton syringe (Thomas Scientific, 1207K95) with attached pulled glass needle was loaded with 4μL of antibody and lowered into each ventricle (−.7AP, 1ML, 2.4DV to bregma). 2μL were diffused over 2 minutes into each ventricle.

Microglia P2RY12^CreER experiments

Mice were injected IP with a single dose of 75mg/kg 4OHT or saline dissolved in corn oil at P14 and underwent blocking antibody surgery the following day at P15. Surgeries were performed in the same manner as all other blocking antibody experiments and mice were collected 2 days after surgery at P17.

DREADD DCZ Injections

DCZ (Tocris, 7193) was dissolved in 1% DMSO in sterile saline at a working dilution of 0.05 mg/mL. For the neonatal (P8–10) DREADD DCZ injections, mice were weighed daily prior to injection and injected with a 1mg/kg dose of DCZ at the same time each day (11a.m.). The mice were collected 2 hours following DCZ administration on the third day of injections (P10). For the adult DREADD DCZ injections, mice were weighed and injected 15 minutes apart to allow for collection 90 minutes post DCZ administration.

RNA In Situ Staining, Imaging, and Analysis

RNAScope

Brains for RNAScope were collected 2 hours following the final DCZ injection at P10. The mice were deeply anesthetized using Avertin (tribromoethanol) and transcardially perfused with ice-cold saline and their brains embedded in O.C.T. (Sakura Finetek) and flash frozen in dry ice. The brains were sectioned at 20uM thickness and thaw mounted onto Superfrost Plus slides (Fisher Scientific, 12–550-15). Slides were stored at −80C until stained. We performed RNAScope (ACDBio) in accordance with the manufacturer’s protocol. Briefly, tissue was fixed for 15 minutes in 4% PFA followed by gradual dehydration for 5 minutes each in 50%, 70% and 100% EtOH solutions. The samples were then treated with protease for 30 minutes, briefly washed, and then incubated with primary probes for 2 hours at 37°C. The primary probes used were C1-Gad2, C2-VGlut1, and C3-IL-34. Following hybridization, the tissue went through a series of 4 amplification steps per kit instructions. After the fourth amplification, the tissue was washed, and coverslipped with Vectashield plus DAPI (Vector Labs, H-2000–10). Slides were stained in batches of 4 and imaged within 72 hours of the staining on a Leica SP8 upright confocal. 19 step Z-Stack images at a thickness of 0.33μM of the ACC were taken with the 40X objective. 3 sections per animal were imaged and included in the analysis.

HCR RNAFish

Brains for HCR RNAFish (Molecular Instruments) were collected and sectioned and pretreated (fixation and dehydration) in the same manner as RNAScope. Following fixation and dehydration, the tissue was placed in primary probe solution overnight at 37°C. The primary probes used were B1-Gad2, B2-Fos, and B3-IL-34. The next day, tissue was washed at 37°C in increasing concentrations of 5X SSCT buffer and then incubated with hairpins directed against the primary probes overnight at 37°C. On the final day, the tissue was washed in 100% 5X SSCT and cover slipped using Fluoromount G (ThermoFisher, 00–4959-52). Imaging of these samples was performed using an Olympus Fluoview 3000 inverted confocal with the 30X objective on the same settings as the RNAScope imaging.

Analysis

Images from RNAScope and RNAFish pipelines were analyzed in IMARIS (Oxford Instruments, v9.0) using the “Imaris for Cell Biologists” plugin to generate cells from the DAPI stain and to count each RNAScope puncta as a separate “vesicle” type within that cell. The volume of the DAPI stain was included in the output as well as the number of each vesicle type. Every cell in the image was included in the analysis and cells were determined to be “glutamatergic” or “GABAergic” based on a threshold adjusted number of Vglut1 puncta vs. Gad2 puncta. Additionally, cells were determined to be Fos+ or Fos− based on an average of 6 puncta adjusted for cell size. IL-34 puncta number was normalized to total DAPI volume and averaged for all cells of a given type within a given animal. The code used to process the data is publicly available at https://github.com/bendevlin18/RNA_in_situ_single_cell_quant, DOI: 10.5281/zenodo.15528731.

Quantification and Statistical analysis

Statistical tests were performed using GraphPad Prism 9. Raw data as well as a description of the tests and results (including multiple comparison corrections and post-hoc analyses) are provided. In all experiments, both male and female mice were included at sufficient power (n ≥ 3 biological replicates) and statistics were first run including sex as a variable (e.g. three-way ANOVA with sex × age × genotype). In all cases where there was not a main effect of sex, male and female mice were then combined for subsequent statistical testing. Unless otherwise noted, data are mean ± S.E.M. and all p-values that fall below p<0.05 are denoted with an asterisk, while p-values between 0.05 and 0.10 denoted with a #.

Highlights.

• Interleukin-34 (IL34) increases in the second week of postnatal life

• IL34 deficiency reduces homeostatic microglia and increases phagocytic microglia

• Blocking IL34 at postnatal day 15 re-activates synaptic pruning

• Overexpressing IL34 accelerates microglial maturity and inhibits synaptic pruning

Inclusion and Diversity

We support inclusive, diverse, and equitable conduct of research.