Microglia-astrocyte crosstalk regulates synapse remodeling via Wnt signaling

Summary

Astrocytes and microglia are emerging key regulators of activity-dependent synapse remodeling that engulf and remove synapses in response to changes in neural activity. Yet, the degree to which these cells communicate to coordinate this process remains an open question. Here, we use whisker removal in postnatal mice to induce activity-dependent synapse removal in the barrel cortex. We show that astrocytes do not engulf synapses in this paradigm. Instead, astrocytes reduce contact with synapses prior to microglia-mediated synapse engulfment. We further show that the reduced astrocyte-synapse contact is dependent on the release of Wnts from microglia downstream of neuron-to-microglia fractalkine ligand-receptor (CX3CL1-CX3CR1) signaling. These results demonstrate an activity-dependent mechanism by which microglia instruct astrocyte-synapse interactions, providing a permissive environment for microglia to remove synapses. We further show that this mechanism is critical to remodel synapses in a changing sensory environment, and this signaling is upregulated in several disease contexts.

In Brief

In the postnatal mouse cortex, microglia use Wnts to direct astrocytes to reduce their contact with synapses prior to synapse engulfment by microglia. This mechanism allows microglia to instruct astrocyte-synapse interactions to facilitate synapse removal and remodeling.

Graphical Abstract

Introduction

Astrocytes and microglia are two types of non-neuronal, central nervous system (CNS)-resident cell types known to regulate synapse structure and function.^1,2 Astrocytes are resident CNS glial cells that have an extensive network of fine processes that juxtapose synapses where they play critical roles in regulating neurotransmission, ion homeostasis, and synaptic plasticity.³ Astrocytes can also promote synapse formation, stabilization, and maturation or directly engulf and remove synapses.⁴ Intriguingly, microglia are resident CNS macrophages that also regulate neurotransmission, synaptic plasticity, synapse formation/stabilization, and synapse removal.² Despite many overlapping synaptic functions, whether and how these two cell types communicate with each other to coordinate their interactions with neurons and synapses remain open questions. This type of glia-glia communication may be particularly important during synapse remodeling where neurons, astrocytes, and microglia must collectively determine which synapses to remove or spare.

Synapse pruning is a developmental process in which a subset of synapses within a circuit is removed as the circuit matures, which involves both microglia and astrocytes.⁵ For example, removing whiskers in postnatal mice, which reduces neuronal activity in the corresponding barrel cortex, elicits the removal of previously formed layer IV thalamocortical (TC) synapses by microglia through neuronal fractalkine (CX3CL1)-microglial fractalkine receptor (CX3CR1) signaling.⁶ In the developing retinogeniculate circuit, microglia also remove synapses from the less active neurons within the circuit.^7–9 However, this process does not involve CX3CL1-CX3CR1 signaling but rather complement-dependent induction of synapse engulfment by microglia. Astrocytes can similarly engulf retinogeniculate synapses from less active neurons through MEGF10 and MERTK receptors.⁸ It has further been shown that astrocytes can engulf and remodel synapses in the adult striatum,¹⁰ hippocampus,¹¹ and cerebellum.¹² These data beg further understanding of whether and how astrocytes and microglia communicate to coordinate the remodeling of synapses in response to changes in activity. Along these lines, IL-33 from astrocytes has been shown to stimulate microglial synaptic pruning in the cortex, thalamus, and spinal cord.^13–15 Still, it remains unknown if microglia can signal to astrocytes to regulate synapse remodeling and whether this crosstalk is dependent on sensory-experience-driven changes in neural activity.

Here, we find that whisker lesioning in postnatal mice, which we previously showed induces microglia to engulf and remove thalamocortical synapses in the barrel cortex,⁶ does not induce astrocyte engulfment of synapses. Instead, astrocyte processes decrease their physical association with synapses as microglia begin to engulf and remove synapses following whisker removal. Using cell type-specific RNA sequencing and molecular genetic approaches, we further identify a Wnt-dependent mechanism by which microglia communicate with astrocytes to stimulate a reduction in synapse contacts and orchestrate subsequent synapse removal. Together, these data highlight an activity-dependent mechanism by which microglia and astrocytes coordinate synaptic remodeling with implications for multiple disease conditions.

Results

Astrocytes decrease their association with thalamocortical synapses following whisker lesioning

The barrel cortex is a well-characterized circuit in the rodent brain, which is important for sensory processing. Synaptic inputs from the thalamus into layer IV of the barrel cortex are spatially arranged into a grid-like pattern where each cluster of synapses (i.e. a “barrel”) corresponds to an individual whisker on the snout. It is well described that the removal of whiskers during a critical window in postnatal mouse development (postnatal day 1–3; P1-P3) prevents the formation of this grid-like pattern of TC synapses.^16–18 In our initial studies, we showed that after this critical window, there is a subsequent stage of developmental plasticity in which a subset of initially-formed TC synapses is removed upon whisker removal.⁶ Specifically, we showed that removing whiskers from P4 mice by trimming or cauterizing elicits the removal of previously-formed TC synapses by microglia without disrupting the overall grid-like pattern.⁶ The process of whisker removal-induced synapse loss is gradual, with observable decreases in synapses beginning ~48 hours after whisker lesioning (P6) and more substantial decreases over the course of 6 days (P4-P10).⁶ By 6 days (P10), whisker removal in postnatal mice induces a 50–75% reduction in the density of vesicular glutamate transporter 2 (VGluT2)-positive TC presynaptic inputs in the barrel cortex and a corresponding decrease in functional synapses that is maintained into adulthood.⁶

As astrocytes have been shown to engulf synapses in other brain regions,^5,8,10–12 we first compared the relative contribution of microglia and astrocytes to the engulfment of TC inputs in this paradigm. All whiskers on one side of the P4 mouse snout were removed by cauterization such that each animal had its own internal control (intact whiskers on the other side of the snout) (Figure 1A). In agreement with our previous work,⁶ within 6 days of P4 whisker lesioning, VGluT2⁺ TC presynaptic terminals within layer IV of the barrel cortex were reduced in males and females (Figure 1B–C; Figure S1A–C) with a concomitant increase in engulfed VGluT2⁺ material within microglial lysosomes within 5 days of whisker lesioning (Figure 1D–F; Figure S1D–F). In contrast, there was no significant increase in engulfed VGluT2⁺ material within astrocytes (Figure 1G–J). As astrocytes are known to have an extensive network of fine processes that closely associate with neuronal synapses, forming “tri-partite synapses,”¹⁹ we next assessed whether astrocyte process proximity to synapses – as opposed to synapse engulfment – could instead be impacted by whisker lesioning. Given the density of astrocyte processes and the resolution limitations of light microscopy, we used transgenic mice in which astrocytes were labeled with membrane-bound GFP (mGFP) (Aldh1l1^CreER/+; Rosa26^mTmG/+)^20,21 combined with expansion microscopy to achieve increased resolution (Figure S2A–E). On average, we achieved a 4-fold expansion of the tissue (Figure S2F), thereby providing a 4-fold increase in resolution. We first assessed astrocyte-synapse interactions at 4 days after whisker lesioning (lesioning at P4 and assessment at P8) in expanded tissue from layer IV of the barrel cortex (Figure 2A–B). Compared to the control hemisphere, the average distance between VGluT2⁺ TC presynaptic terminals and their nearest astrocyte process (nearest neighbor distance, NND) was increased in the deprived hemisphere (Figure 2C–G, S2G). Additionally, the percentage of VGluT2⁺ presynaptic terminals contacted by astrocytes was decreased (Figure 2F, H), the percentage of each presynaptic terminal’s surface area that was contacted by an astrocyte process was decreased (Figure S2H–J), and the total density of astrocyte processes in was decreased in the deprived hemisphere (Figure S3A–D) in both males and females (Figure S2K–Q and S3G–I). Importantly, all this occurred without changes in total astrocyte numbers in this region (Figure S3E–F). The reduction in astrocyte-synapse interaction in the barrel cortex following whisker lesioning was further confirmed using electron microscopy (Figure 2I–L).

Figure 1. Microglia, but not astrocytes, engulf and remove synapses after whisker lesioning.

(A) Experimental design for B-C. Layer IV barrel cortex, 6 days post-lesioning.

(B) Representative images of anti-VGluT2⁺ synapses in control vs. deprived hemispheres (hems). Scale bars 10 μm.

(C) VGluT2⁺ synapse density in control vs. deprived hems (n=5 mice. *p<0.05).

(D) Experimental design for E-F. Layer IV barrel cortex, 5 days post-lesioning.

(E) Representative max intensity projections of anti-VGluT2⁺ synapses, anti-CD68⁺ lysosomes, and anti-GFP⁺ microglia in control vs. deprived hems. Scale bars 5 μm. Inset: 3D render of anti-VGluT2⁺ synaptic material (arrow) within anti-CD68⁺ lysosomes inside anti-GFP⁺ microglia. Inset scale bar 2 μm.

(F) VGluT2⁺ synaptic material within microglial lysosomes in control vs. deprived hems (n=4 mice. *p<0.05).

(G) Experimental design for H-J. Layer IV barrel cortex, 5 days post-lesioning.

(H-I) Representative max intensity projections (H, I top) and 3D renders (I bottom) of anti-VGluT2⁺ synapses, anti-LAMP2⁺ lysosomes, and anti-GFP⁺ astrocytes in control vs. deprived hems. Boxed region in H shown on left in I. 3D renders only show synaptic material within astrocyte lysosomes, examples indicated (arrows). Scale bar 5 μm.

(J) VGluT2⁺ synaptic material within astrocyte lysosomes in control vs. deprived hems (n=6 mice).

Data represent mean ± SEM. Circles (female) and squares (male) represent individual animals.

Statistics: Ratio paired t test (C, F, J)

See also Figure S1.

Figure 2. Astrocytes decrease their association with thalamocortical synapses following whisker lesioning.

(A) Top: experimental design for B-H. Layer IV barrel cortex, 4 days post-lesioning. Bottom: diagram of astrocyte-synapse nearest neighbor distance (NND) and astrocyte-synapse contact.

(B-C) Representative images of anti-VGluT2⁺ synapses and anti-GFP⁺ astrocyte processes (B) before vs. after gel-expansion for expansion microscopy, (C) in control vs. deprived hemispheres (hems) post-expansion highlighting synapse-astrocyte contacts (asterisks) and synapse-astrocyte NNDs (dashed lines). Scale bars (B) 20 μm, (C) 10 μm (not adjusted by expansion index).

(D-E) Representative 3D renders of control vs. deprived hems showing (D) anti-VGluT2⁺ synapses and anti-GFP⁺ astrocyte processes, (E) anti-VGluT2⁺ synapses pseudo-colored by synapse-astrocyte NND (corrected for expansion index). Astrocyte-contacted synapses (NND=0) are transparent. Scale bars 2 μm (corrected for expansion index).

(F) Cumulative proportion plot of synapses’ synapse-astrocyte NND in control vs. deprived hems. Data pooled from n=5 mice.

(G-H) Mean synapse-astrocyte NND (G) and percentage of synapses contacted by astrocytes (H), measured by expansion microscopy in control vs. deprived hems (n=5 mice. *p<0.05).

(I) Experimental design for J-L. Layer IV barrel cortex, 4 days post-lesioning.

(J) Representative electron micrographs of control vs. deprived hems. Pseudo-coloring highlights example synapses and astrocyte processes. Scale bars 1 μm.

(K-L) Mean synapse-astrocyte NND (K) and percentage of synapses contacted by astrocytes (L), measured by electron microscopy in control vs. deprived hems (n=4 mice. *p<0.05).

Bar graphs represent mean ± SEM. Circles (female) and squares (male) represent individual animals.

Statistics: Ratio paired t test (G-H, K-L)

See also Figures S2 and S3.

Our previous work showed that synapse removal by microglia happens gradually over the course of 6 days post-whisker lesioning with a daily rate of synapse loss of ~12% per day.⁶ Interestingly, we observed a ~10% reduction in astrocyte-synapse contact at P8 (Fig. 2F, H, L), comparable to the daily rate of synapse loss. The percentage of synapses contacted by astrocytes and the total density of astrocyte processes was also decreased 6 days after whisker lesioning (Figure S3J–M), a time point with even larger decreases in synapse density.⁶ These data suggest that reduced astrocyte-synapse association after whisker removal is a progressive process over the course of 6 days post-whisker lesioning while microglia are engulfing and removing synapses.

Wnt receptor signaling is elicited in layer IV astrocytes following whisker removal

To identify the molecular basis for astrocytes’ reduced proximity to synapses after whisker lesioning, we performed translating ribosome affinity purification followed by RNA sequencing (TRAP-seq). Astrocyte ribosomes were labeled using transgenic Aldh1l1^EGFP-L10a/+ mice²² (Figure 3A), which underwent unilateral whisker lesioning at P4. 24 hours after whisker removal, the deprived and non-deprived barrel cortices were micro-dissected, and GFP-tagged ribosomes were immunoprecipitated with a GFP antibody followed by RNA sequencing. We first ensured that cell type-specific genes for astrocytes, and not other brain cell types, were enriched in the ribosome-bound RNA compared to the unbound fraction (Figure S4A–B and Table S1). We then compared the deprived and non-deprived barrel cortices and identified 147 upregulated genes and 39 downregulated genes in astrocyte ribosome-bound RNA (Figure 3B and Table S1). Strikingly, Gene Set Enrichment Analysis identified genes related to transcriptional regulation and the cytoskeleton (Figure S4C and Table S2), consistent with the induction of astrocyte morphological changes after whisker lesioning. To identify potential signaling pathways upstream of these astrocyte transcriptional changes, we performed Ingenuity Pathway Analysis (IPA) of higher-order root regulators (Figure 3C and Table S3). IPA identified many regulators related to the canonical Wnt signaling pathway including β-catenin (CTNNB1) and its co-activated transcription factors T-cell factor/lymphoid enhancer factor (TCF/LEF).

Figure 3. Wnt receptor signaling is elicited in layer IV astrocytes following whisker removal.

(A) Experimental design for B-C. Barrel cortex, 24 hours post-lesioning.

(B) Volcano plot of differentially expressed genes (DEGs) in astrocytes in control vs. deprived hemispheres (hem). n=5 mice.

(C) Plot of Ingenuity Pathway Analysis (IPA) root regulators of DEGs in control vs. deprived hem astrocytes. x-axis position indicates activation z-score. Dot size indicates −log₁₀ p-value. Wnt signaling pathway molecules: bold and blue.

(D) Diagram of canonical Wnt siganling pathway.

(E) Experimental design for F-I. Layer IV barrel cortex, 24 hours post-lesioning.

(F) Representative images of control vs. deprived hems of TCF/Lef:H2B-GFP mice showing anti-SOX9⁺ astrocyte nuclei, DAPI⁺ nuclei, and anti-GFP⁺ nuclei. Right-most panels depict anti-GFP signal in astrocytes (SOX9⁺; examples outlined in insets) and non-astrocytes (DAPI⁺ SOX9-negative). Scale bars and inset scale bars 10 μm.

(G) Average TCF/Lef:H2B-GFP mean fluorescence intensity (MFI) in control vs. deprived hem astrocytes (n=3 mice. ***p<0.001).

(H) Representative images of anti-β-catenin, anti-SOX9⁺ astrocyte nuclei, and DAPI⁺ nuclei in control vs. deprived hems. Two right-most columns show anti-β-catenin in astrocyte nuclei (SOX9⁺; examples outlined in insets) and non-astrocytes (DAPI⁺ SOX9-negative). Scale bars and inset scale bars 10 μm.

(I) Average nuclear β-catenin MFI in astrocytes and non-astrocytes in control vs. deprived hems (n=6 mice. *p<0.05).

Bar graphs represent mean ± SEM. Circles (female) and squares (male) represent individual animals.

Statistics: Ratio paired t test (G), Repeated measures 2-way ANOVA with Holm-Sidak post-hoc test (I).

See also Figure S4, Tables S1–S3.

To validate increased Wnt receptor signaling in astrocytes within layer IV of the barrel cortex following whisker removal, we assessed two molecules in the canonical Wnt signaling pathway downstream of Wnt receptor activation that were identified by IPA analysis: β-catenin and TCF/LEF. Upon binding of the Wnt ligand to its receptors on the plasma membrane, β-catenin translocates to the nucleus and induces TCF/LEF-mediated transcriptional regulation of Wnt-responsive genes²³ (Figure 3D). To test whether whisker lesioning induces TCF/LEF activity in astrocytes, we used a TCF/Lef:H2B-GFP reporter mouse line in which a multimerized TCF/LEF response element drives expression of nuclear GFP (Figure 3E).²⁴ Using immunofluorescence confocal microscopy, we confirmed canonical Wnt signaling induction in layer IV astrocytes (identified by SOX9⁺ nuclei) in the deprived hemisphere at 24 hours after P4 whisker lesioning, indicated by significantly increased TCF/Lef:H2B-GFP reporter intensity (Figure 3F–G). Interestingly, we also observed TCF/Lef:H2B-GFP signal in non-astrocytes. However, this signal may have been due to the use of an artificial multimerized TCF reporter line.²⁵ We, therefore, also performed immunostaining for β-catenin. β-catenin can be observed in the cytosol and at the membrane in all conditions, but to focus on β-catenin downstream of Wnt receptor activation, we analyzed levels of β-catenin in nuclei, similar to previous studies²⁶ (Figure 3E). In agreement with our results from TCF/Lef:H2B-GFP reporter mice, β-catenin was increased in the nuclei of astrocytes in the deprived barrel cortex compared to the non-deprived cortex (Figure 3H–I). Importantly, this induction of canonical Wnt signaling in astrocytes following whisker lesioning was astrocyte-specific as indicated by increased nuclear β-catenin in astrocytic SOX9⁺ nuclei, but no increase in nuclear β-catenin in non-astrocyte, SOX9-negative nuclei in the deprived barrel cortex (Figure 3H–I). Together, these data demonstrate activation of Wnt receptor signaling specifically in astrocytes within layer IV barrel cortex following whisker removal in postnatal mice.

Activation of the Wnt signaling pathway in astrocytes is sufficient to induce a reduction in astrocyte-synapse interactions

To further investigate how astrocyte Wnt receptor signaling affects astrocyte-synapse interactions, we assessed mice in which Wnt signaling pathway activity is increased in astrocytes in the absence of whisker lesioning. To activate the Wnt signaling pathway specifically in astrocytes, we used astrocyte-specific conditional knockout (cKO) mice for adenomatous polyposis coli (APC) (Apc^F/F; Aldh1l1^CreER/+; Rosa26^mTmg/+). APC is a critical component of the β-catenin destruction complex and a negative regulator of Wnt signaling²³ (Figure 4A). In this Apc^Flox line, Cre-mediated recombination removes exon 14 of Apc, disrupting APC’s armadillo repeat region. This leads to stabilization of β-catenin, allowing β-catenin to translocate to the nucleus and constitutively activate the canonical Wnt signaling pathway even in the absence of upstream Wnt receptor activation²⁷ (Figure 4B). We injected astrocyte-specific APC cKO mice and littermate controls (Apc^+/+; Aldh1l1^CreER/+; Rosa26^mTmg/+) with tamoxifen from P1-P4 and observed increased nuclear β-catenin in astrocytes at P12, indicating increased Wnt signaling pathway activity at baseline (Figure 4C–E). We then performed ExM in layer IV of the barrel cortex at P12 in mice without any whisker manipulations. We first examined the total density of astrocyte processes and found no difference between APC cKO mice and their controls (Figure 4F–G). Next, we measured astrocyte-synapse interactions. While there was still the same percentage of synapses contacted by astrocytes (defined as NND=0 measured by ExM) (Figure 4H–I), the average separation between VGluT2⁺ TC synapses and the nearest astrocyte process (including synapses with NND=0 and NND>0) was increased in APC cKO mice versus controls (4H, J), comparable to the increase in NND observed in P8 mice after whisker lesioning (Figure 2G, S2G). Notably, synaptic activity is not decreased in this context as whiskers are still intact. This could explain why astrocyte-contacted synapses are less affected than uncontacted synapses. Together, these results suggest that activation of the Wnt signaling pathway in astrocytes is sufficient to induce reductions in astrocyte-synapse interaction in vivo, even in the absence of whisker lesioning.

Figure 4. Activation of Wnt signaling in astrocytes induces increased separation between synapses and astrocyte processes.

(A) Diagram of Wnt signaling pathway and β-catenin destruction complex.

(B) Diagram of Apc ^Flox allele.

(C) Experimental design for D-J. Layer IV barrel cortex, P12.

(D) Representative images of anti-β-catenin, anti-SOX9⁺ astrocytes, and DAPI⁺ nuclei in Aldh1l1^CreER/+; Apc^+/+; Rosa26^mTmG/+ (control) mice vs. Aldh1l1^CreER/+; Apc^Flox/Flox; Rosa26^mTmG/+ (APC cKO) mice. Right-most panels show anti-β-catenin in astrocyte nuclei (SOX9⁺; examples outlined in insets). Scale bars and inset scale bars 10 μm.

(E) Average nuclear β-catenin MFI in astrocytes in control vs. APC cKO mice (n=5 control, 3 APC cKO mice. *p<0.05).

(F) Representative 3D renders of anti-GFP⁺ astrocyte processes in control vs. APC cKO mice. Scale bars 2 μm (corrected for expansion index).

(G) Total density of astrocyte processes in control vs. APC cKO mice (n=6 control, 5 APC cKO mice).

(H) Representative 3D renders of anti-VGluT2⁺ synapses in control vs. APC cKO mice, pseudo-colored by synapse-astrocyte nearest neighbor distance (NND; corrected for expansion index). Astrocyte-contacted synapses (NND=0) are transparent. Insets: 3D renders of anti-VGluT2⁺ synapses and anti-GFP⁺ astrocyte processes. Scale bars and inset scale bars 2 μm (corrected for expansion index).

(I-J) Percentage of synapses contacted by astrocytes (I) and mean synapse-astrocyte NND (J) in control vs. APC cKO mice (n=6 control, 5 APC cKO mice. *p<0.05).

(K) Diagram of ligand-receptor analysis.

(L) Heatmap of predicted ligand activities for Wnts in public astrocyte TRAP-Seq datasets. Superscripts: reference numbers.

Bar graphs represent mean ± SEM. Circles (female) and squares (male) represent individual animals.

Statistics: Student’s t test (E, G, I-J).

See also Table S4.

If activation of Wnt receptor signaling in astrocytes is sufficient to induce reduced astrocyte-synapse interaction, this would suggest that Wnt signaling could similarly be used in other contexts to facilitate synapse removal by microglia via reductions in astrocyte-synapse contact. To compare levels of astrocyte Wnt receptor signaling after whisker removal to other conditions in which morphological changes in astrocytes, increases in microglial synapse engulfment, and decreases in synapses have all been documented (day vs. night, aging, Alzheimer’s disease, etc.), we performed a NicheNet ligand-receptor analysis of our astrocyte TRAP-Seq dataset and other published astrocyte TRAP-Seq datasets^28–32, comparing each dataset to its respective control (Figure 4K). Similar to whisker lesioning, Wnt ligands ranked among the ligands with the highest predicted activity after seizures,³⁰ during aging,²⁹ during sleep vs. wake,²⁸ and in an amyloid-β model and a tau model of Alzheimer’s disease³¹ (Figure 4L and Table S4). These data suggest that activation of astrocyte Wnt receptor signaling could be a general mechanism involved in regulating astrocyte-synapse contact and synapse removal by microglia.

Whisker lesioning induces microglia-astrocyte crosstalk via Wnts

We next sought to identify the upstream mediator of Wnt receptor signaling activation in astrocytes after whisker lesioning. There are multiple different levels of regulation of Wnt receptor signaling including expression of Wnt ligands and Wnt receptors, Wnt ligand protein release, and interactions with extracellular molecules that impact Wnt ligand distribution and activity²³. Based on our previous work in which microglia were required for the engulfment and removal of synapses after whisker lesioning,⁶ we were particularly intrigued by the possibility that Wnt receptor signaling in astrocytes could be influenced by changes in microglia elicited by whisker lesioning.

To examine how microglia might contribute to Wnt receptor activation in astrocytes, we examined transcriptional changes in microglia in the deprived and control hemispheres of the somatosensory cortex at 24 hours after whisker lesioning. Similar to astrocytes, we performed TRAP-seq in microglia using neonate Cx3cr1^{CreER/+ (Litt)}; Eef1a1^{LSL-EGFP-L10a/+} mice³³ with 24 hours of whisker deprivation (Figure 5A) and confirmed that cell-type specific genes for microglia, but not other cell types, were highly enriched in the ribosome-bound RNA versus the unbound fraction (Figure S5A–B and Table S5). We then used this dataset along with the astrocyte-specific TRAP-seq dataset (Figure 3A–B) to perform NicheNet³⁴ ligand-receptor analysis between microglia and astrocytes (Figure 5B and Table S6). For this analysis, we took all ligand-receptor pairs expressed in microglia and astrocytes and probed which microglia ligands have the strongest regulatory association with the genes that are transcriptionally increased or decreased in astrocytes following whisker lesioning. Strikingly, several microglia-derived Wnt ligands were identified as the top putative ligands involved in microglia-astrocyte crosstalk after whisker lesioning (Figure 5C and Table S6). Among the Wnts expressed in microglia, WNT7A, WNT2B, and WNT4 had the highest predicted activity.

Figure 5. Whisker lesioning induces microglia-astrocyte crosstalk via Wnts.

(A) Experimental design for microglia TRAP-Seq data. Barrel cortex, 24 hours post-lesioning.

(B) Diagram of ligand-receptor analysis.

(C) Predicted ligand activity vs. enrichment in microglia for top 20 ligands. Wnts bolded and blue.

(D) Experimental design for E-F. 24 hours post-lesioning

(E) Representative MERFISH sample highlighting barrel cortex layers.

(F) Representative image of DAPI⁺ nuclei and anti-VGluT2⁺ synapses overlaid with MERFISH-detected transcripts for microglia marker genes. Layer IV outlined. Scale bar 250 μm. Inset highlights DAPI⁺ microglia expressing Wnt7a (arrow). Inset scale bar 5 μm.

(G-H) Dot plots of gene expression in microglia in barrel cortex layers for (G) individual Wnt ligands and (H) the aggregate of all Wnt ligands. Data pooled from 2 males, 2 females.

(I) Experimental design for J-M.

(J) Representative images of GFP⁺ primary astrocytes with no treatment vs. exposure to WNT4, WNT5A, or WNT7A. Scale bars 50 μm.

(K-M) Plots of astrocyte morphological complexity after no treatment vs. exposure to WNT4, WNT5A, or WNT7A (n=3 independent culture experiments. **p<0.01; ****p<0.0001; ns: p>0.05).

Data in K-M represent the locally estimated scatterplot smoothing (LOESS) line of best fit (line) and 95% confidence interval (outline).

Statistics: Mixed-model ANOVA with Tukey’s post-hoc test (K-M)

See also Figure S5 and Tables S5–S7.

We next performed spatial transcriptomics by multiplexed error-robust fluorescence in situ hybridization (MERFISH) to examine the expression of Wnt ligands genes in microglia in different layers of the barrel cortex at 24 hours after whisker lesioning (Figure 5D–F, S5C–G, and Table S7). Importantly, MERFISH allowed us to specifically analyze which genes are expressed in layer IV where TC inputs are removed and astrocyte Wnt receptor signaling is activated in response to whisker lesioning. Within layer IV of the barrel cortex, the Wnt ligands with the highest expression in microglia were Wnt2, Wnt2b, Wnt4, Wnt5a, Wnt5b, and Wnt8a. We also computed the aggregate expression of all Wnt ligand genes and found that microglia in layer IV had higher expression of Wnt ligand genes than microglia in other layers of the barrel cortex independent of whisker lesioning (Figure 5H). These data suggest that layer IV microglia may be primed to engage in Wnt release, but that whisker lesioning-dependent Wnt signaling is likely operating at the post-transcriptional level.

We then assessed the capacity of these Wnts found in microglia to induce astrocyte morphological changes in vitro. To do so, we used a culture system that was established to assess astrocyte morphological changes.³⁵ In these cultures, EGFP-labeled astrocytes are cultured on a layer of methanol-fixed neurons, which induces astrocytes to form a complex, ramified morphology (Figure 5I) but prevents the neurons from actively signaling to astrocytes.³⁵ We exposed astrocyte cultures for 72 hours to either WNT4, WNT5A, or WNT7A, which had high predicted ligand activity by NicheNet (Figure 5C) and/or high expression in microglia in layer IV of the barrel cortex (Figure 5G). WNT4 did not induce changes in astrocyte morphology (Figure 5J–K and S5H–I), but WNT5A and WNT7A both induced significant, concentration-dependent reductions in astrocyte morphological complexity (Figure 5J, L–M and S5H–I). Together, these data suggest that multiple Wnts are expressed in layer IV cortical microglia and, although not induced by whisker lesioning at the transcriptional level, the proteins are capable of inducing changes in astrocyte process morphology.

Microglial Wnt release is necessary in vivo to induce decreased astrocyte-synapse interactions and synapse removal following whisker lesioning

To test the in vivo significance of microglia-astrocyte Wnt signaling as the factor driving changes in astrocyte morphology and synapse remodeling and the possibility that regulation of this signaling from microglia is post-transcriptional, we tested mice with microglia cell type-specific deficiency in Wnt release. To genetically block Wnt release from microglia, we leveraged mice harboring floxed alleles of wntless (Wls),³⁶ which is required for trafficking and secretion of Wnt proteins from cells (Figure 6A).^37–39 This approach allowed us to test the combined impact of blocking all Wnt ligands expressed by microglia, including those that induced reductions in astrocyte morphological complexity in vitro. Given the narrow time window for our experiments, gene deletion approaches using the inducible form of CreER in microglia⁴⁰ were not compatible with our experiments. Therefore, we chose to specifically ablate Wls in microglia using a BAC transgenic Cx3cr1^Cre line^41,42 in which Cx3cr1^Cre was randomly inserted into the genome such that endogenous Cx3cr1 expression is not impacted. We refer to these mice as Wls^Flox/Flox; Cx3cr1^Cre (WLS cKO) mice. We note that, in addition to microglia, Cx3cr1 is also expressed by other brain border macrophages and peripheral immune cells at this age. However, our analyses are all performed in the brain parenchyma within the barrel cortex. In this region and context, microglia are the only Cx3cr1 expressing cells. Deletion of Wls specifically in microglia at this age was confirmed by in situ hybridization (Figure S6A–B). WLS cKO mice were then intracerebroventricularly injected at P1 with PHP.eB::GfaABC1D-lck-smV5–4×6T adeno-associated virus (AAV)⁴³ to label astrocytes with membrane bound smV5 (Figure S6C–E). In WLS cKO mice, there was no longer a change in the total density of astrocyte processes (Figure S6F–H) or astrocyte association with synapses (Figure 6B–E) in layer IV of the barrel cortex after whisker lesioning. Most importantly, this attenuation of the reduction in astrocyte-synapse interactions was accompanied by a block in microglial engulfment of VGluT2⁺ synaptic material at 5 days post-lesioning (Figure 6F–H) and an attenuation of the loss of VGluT2⁺ TC presynaptic terminals in layer IV in the deprived hemisphere at 6 days post-lesioning in WLS cKO mice (Figure 6I–K). The attenuated synapse loss was phenocopied in postnatal mice treated daily from P3-P9 with XAV939,⁴⁴ a brain-permeable pharmacological inhibitor of canonical Wnt signaling (Figure S6I–L). Taken together, these data show that microglial Wnt release is necessary to induce astrocyte process retraction, microglia-mediated synapse engulfment, and synapse loss following sensory deprivation.

Figure 6. Microglial Wnt release is necessary to induce decreased astrocyte-synapse interactions and synapse removal following whisker lesioning.

(A) Diagram of microglia-astrocyte Wnt signaling.

(B) Experimental design for C-E. Layer IV barrel cortex, 4 days post-lesioning.

(C) Representative 3D renders of control vs. deprived hemispheres (hems) in Cx3cr1^Cre; Wls^+/+ (control) vs. Cx3cr1^Cre; Wls^Flox/Flox (WLS cKO) mice showing anti-VGluT2⁺ synapses pseudo-colored by synapse-astrocyte nearest neighbor distance (NND; corrected for expansion index). Astrocyte-contacted synapses (NND=0) are transparent. Insets: 3D renders of anti-VGluT2⁺ synapses and anti-GFP⁺ astrocyte processes. Scale bars and inset scale bars 2 μm (corrected for expansion index).

(D-E) Mean synapse-astrocyte NND (D) and percentage of synapses contacted by astrocytes (E) in control vs. deprived hems of control and WLS cKO mice (n=4 control, 4 WLS cKO mice. *p<0.05).

(F) Experimental design for G-H. Layer IV barrel cortex, 5 days post-lesioning.

(G) Representative max intensity projections (left) and 3D renders (right) of anti-VGluT2⁺ synapses, anti-CD68⁺ lysosomes, and anti-IBA1⁺ microglia in control vs. deprived hems of control and WLS cKO mice. 3D renders only show synaptic material within microglia lysosomes, examples indicated by arrows. Scale bars 5 μm.

(H) VGluT2⁺ synaptic material within microglial lysosomes in control vs. deprived hems of control and WLS cKO mice (n=4 control, 5 WLS cKO mice. *p<0.05).

(I) Experimental design for J-K. Layer IV barrel cortex, 6 days post-lesioning.

(J) Representative images of anti-VGluT2⁺ synapses in control vs. deprived hems of control and WLS cKO mice. Scale bars 10 μm.

(K) VGluT2⁺ synapse density in control vs. deprived hems of control and WLS cKO mice (n=6 control, 6 WLS cKO mice. ***p<0.001).

Data represent mean ± SEM. Circles (female) and squares (male) represent individual animals.

Statistics: Repeated measures 2-way ANOVA with Holm-Sidak post-hoc test (D-E, H, K)

See also Figure S6.

Microglial fractalkine receptor (CX3CR1) signaling acts upstream to induce Wnt receptor signaling in astrocytes for synapse remodeling

What is upstream to induce Wnt ligand-receptor signaling between microglia and astrocytes? In our previous work, we showed that microglial synapse engulfment and TC input elimination after whisker lesioning in neonate mice depend on signaling between the microglia-expressed CX3CR1 receptor and its ligand CX3CL1,⁶ which is restricted to neurons in the neonate barrel cortex and proteolytically cleaved by ADAM10 in response to whisker lesioning. When CX3CL1-CX3CR1 signaling was blocked, the synapses that remained after whisker lesioning were functional, and the effect was sustained. However, CX3CR1 is a G-protein coupled chemokine receptor, not an engulfment receptor, and we observed no effect on microglial numbers in the barrel cortex. We thus reasoned that CX3CL1-CX3CR1 activation must be facilitating synapse remodeling in some other manner. This inspired us to test whether CX3CL1-CX3CR1 signaling could be upstream of Wnt release from microglia, induction of Wnt signaling in astrocytes, and subsequent reductions in astrocyte-synapse interactions.

First, we used microglia-specific TRAP-seq to compare microglial transcriptional responses 24 hours after whisker lesioning in Cx3cr1^−/− mice versus Cx3cr1^+/− mice (Figure 7A). Notably, CX3CR1 signaling remains intact in Cx3cr1^+/− mice and whisker lesioning-induced synapse loss and synapse engulfment by microglia is comparable to wild-type mice.⁶ Comparing microglia isolated from the deprived vs. control hemispheres of Cx3cr1^+/− mice, we found 471 upregulated genes and 156 downregulated genes (Figure 7B and Table S5). However, in Cx3cr1^−/− mice, there were only 98 upregulated genes and 45 downregulated genes in the deprived hemisphere (Figure 7C and Table S5). Among these, only 5 genes were consistently upregulated in the deprived hemispheres of both Cx3cr1^+/− and Cx3cr1^−/− mice, and 0 genes were downregulated in both genotypes. These data suggest that the majority of gene changes in microglia within the first 24 hours after whisker lesioning are downstream of CX3CR1 activation.

Figure 7: CX3CR1-CX3CL1 signaling is upstream of microglia-astrocyte crosstalk via Wnts.

(A) Experimental design for B-C. Barrel cortex, 24 hours post-lesioning.

(B-C) Volcano plots of differentially expressed genes in microglia in the control vs. deprived hemispheres (hems) of (B) n=4 Cx3cr1^+/− mice and (C) n=5 Cx3cr1^−/− mice.

(D) (Top) experimental design for E-F. Layer IV barrel cortex, 24 hours post-lesioning. (Bottom) diagram of Wnt signaling pathway.

(E) Representative images of control vs. deprived hemispheres (hems) of TCF/Lef:H2B-GFP; Cx3cl1^−/− mice showing anti-SOX9⁺ astrocyte nuclei, DAPI⁺ nuclei, and anti-GFP⁺ nuclei. Right-most panels depict anti-GFP signal in astrocytes (SOX9⁺; examples outlined in insets). Scale bars and inset scale bars 10 μm.

(F) Average TCF/Lef:H2B-GFP mean fluorescence intensity (MFI) in control vs. deprived hem astrocytes (n=4 mice. ns: p>0.05).

(G) Experimental design for H-I. Layer IV barrel cortex, 24 hours post-lesioning.

(H) Representative images of anti-β-catenin, anti-SOX9⁺ astrocyte nuclei, and DAPI⁺ nuclei in control vs. deprived hems of Cx3cr1^−/− mice. Right-most column shows anti-β-catenin in astrocyte nuclei (SOX9⁺; examples outlined in insets). Scale bars and inset scale bars 10 μm.

(I) Average nuclear anti-β-catenin MFI in astrocytes in control vs. deprived hems of Cx3cr1^−/− mice (n=5 mice. ns: p>0.05).

(J) Experimental design for K-M. Layer IV barrel cortex, 4 days post-lesioning.

(K) Representative 3D renders of control vs. deprived hems in Aldh1l1^CreER/+; Rosa26^mTmG/+; Cx3cl1^−/− mice showing anti-VGluT2⁺ synapses pseudo-colored by synapse-astrocyte nearest neighbor distance (NND; corrected for expansion index). Astrocyte-contacted synapses (NND=0) are transparent. Insets: 3D renders of anti-VGluT2⁺ synapses and anti-GFP⁺ astrocyte processes. Scale bars and inset scale bars 2 μm (corrected for expansion index).

(L-M) Mean synapse-astrocyte NND (L) and percentage of synapses contacted by astrocytes (M) in control vs. deprived hems of Aldh1l1^CreER/+; Rosa26^mTmG/+; Cx3cl1^−/− mice (n=4 mice. ns: p>0.05).

Bar graphs represent mean ± SEM. Circles (female) and squares (male) represent individual animals.

Statistics: Ratio paired t test (F, I, L-M).

See also Figure S7.

We then tested whether CX3CR1-CX3CL1 signaling is upstream of Wnt receptor activation in layer IV barrel cortex astrocytes using mice deficient in CX3CL1-CX3CR1 signaling. First, using our MERFISH spatial transcriptomics dataset (Figure 5D–H and S5C–G), we examined gene expression of Wls, Wnt ligands, Wnt receptors, and core Wnt signaling components in Cx3cl1^−/− and wild-type mice (Figure S7A–D). Including cells from the whole brain section, we found no changes in gene expression in microglia or astrocytes, suggesting that CX3CL1 ablation does not affect baseline expression levels of Wnt signaling components. Then, to examine whether CX3CL1 ablation affects activation of Wnt receptor signaling in astrocytes after whisker lesioning, we crossed the TCF/Lef:H2B-GFP Wnt receptor signaling reporter to Cx3cl1^−/− mice and found that the increased reporter activity in astrocytes observed in wild-type mice at 24 hours after whisker lesioning was abolished in Cx3cl1^−/− mice (Figure 7D–F). We performed similar experiments in mice deficient in the receptor, Cx3cr1^−/− mice. Because Cx3cr1^−/− mice (also referred to as Cx3cr1^EGFP/EGFP) express GFP that interferes with the TCF/Lef:H2B-GFP, we instead measured levels of nuclear astrocytic β-catenin in these mice. Similar to Cx3cl1^−/− mice, the induction of nuclear β-catenin in astrocytes observed in wild-type mice following whisker lesioning was blocked in Cx3cr1^−/− mice (Figure 7G–I). Finally, we tested the impact of CX3CL1-CX3CR1 signaling on astrocyte-synapse interactions. Consistent with CX3CL1-CX3CR1 signaling in microglia being upstream of astrocyte Wnt receptor induction and reduced synapse interactions following whisker lesioning, we found that the total density of astrocyte processes is no longer decreased (Figure S7E–G) and synapse association with astrocyte processes is no longer decreased in response to P4 whisker lesioning in Cx3cl1^−/− mice (Figure 7J–M and S7H).

Taken together, these experiments demonstrate that the induction of astrocytic Wnt receptor signaling and decreased astrocyte-synapse association after whisker lesioning is downstream of CX3CR1 activation in microglia by CX3CL1, which all subsequently leads to increased removal of synapses by microglia and reduced synapse density. Because changes in Wnt-related molecule transcription were not observed in Cx3cl1^−/− microglia at baseline or wild-type microglia after whisker lesioning, this is likely through modulation of Wnt ligand translation and/or release from microglia to then activate Wnt receptor signaling in astrocytes.

Discussion

It is increasingly appreciated that both microglia and astrocytes can carry out synapse pruning in the developing brain.⁵ In some cases, this is within the same brain region.⁸ This begs the question of whether these two cell types are communicating to carry out this critical function for sculpting and remodeling neural circuits. Here, we show that microglia communicate directly with astrocytes through Wnt release to stimulate astrocytes to redistribute their processes and subsequently permit microglia-mediated synapse removal. In addition, numerous studies have shown a role for CX3CL1-CX3CR1 signaling in microglia-mediated regulation of synapses,^45–48 but mechanisms downstream of this ligand-receptor signaling have remained elusive. Our study provides insight into how CX3CL1-CX3CR1 signaling is regulating microglia-mediated synapse remodeling. In this context, CX3CL1-CX3CR1 activation does not induce classical chemokine signaling but instead engages activity-dependent microglia-to-astrocyte communication via Wnts prior to microglia-mediated synapse engulfment. Wnt release from microglia engages astrocytes to then redistribute their processes further away from synapses, which is subsequently permissive for microglia to engulf and remove synapses.

Previously, it has been shown that both microglia and astrocytes can similarly contribute to synapse pruning by engulfing and remove synapses. However, it has been less clear if and how these two cell types coordinate synapse removal. In the mouse cortex, microglia and astrocytes engulf separate domains of the cell during apoptotic cell death with microglia engulfing the cell body while the astrocytes engulf the processes.⁴⁹ Also, in the mouse spinal cord and thalamus, astrocyte secretion of IL-33 stimulates microglia to upregulate phagocytic pathways to engulf and remove synapses.^13–15 These studies indicate intercellular crosstalk occurs between astrocytes and microglia to coordinate synapse removal. Also, in disease contexts, several examples of microglia-astrocyte crosstalk have been identified.⁵⁰ For example, C1q, IL-1α, and TNF produced by microglia can induce C3 expression in astrocytes,⁵¹ which can tag synapses for removal in models of neuroinflammatory neurodegeneration.⁵² Nevertheless, during physiological settings, it remains unclear if microglia communicate with astrocytes to regulate synapse remodeling and whether this crosstalk is activity-dependent. Here, we show that CX3CR1-dependent Wnt release from microglia is an activity-dependent mechanism by which microglia signal to astrocytes to facilitate synapse removal by microglia.

In the current study, we identified that Wnt release from microglia induced Wnt receptor signaling in astrocytes to induce astrocyte process remodeling away from synapses. Wnt signaling is a known regulator of cell morphology and the actin cytoskeleton, including in astrocytes.²⁶ Our current study showed in vitro that activation of the Wnt signaling pathway in astrocytes was sufficient to induce changes in astrocyte morphology and in vivo this led to increased separation between astrocytes and synapses. We identified multiple Wnts expressed by microglia within layer IV of the barrel cortex that can induce reductions in astrocyte morphology. How CX3CR1 signaling regulates this release remains an open question. One possibility is through Ca²⁺⁻dependent Wnt release from microglia downstream of CX3CR1 activation. CX3CR1 is a GPCR coupled to Gi and Gz proteins, which can induce release of Ca²⁺ from internal stores.⁵³ It is possible that Wnt release from microglia is Ca²⁺ dependent. Also, it is noted that in addition to microglia, CX3CR1 is also expressed by brain border macrophages and peripheral immune cells. In disease contexts, peripheral-derived macrophages can communicate with astrocytes.⁵⁴ However, in the context of layer IV of the developing barrel cortex where the blood brain barrier is intact, microglia are the primary source of CX3CR1. Thus, our results are most consistent with a microglia-dependent mechanism.

Our observation that astrocytes decrease interaction with synapses prior to their removal by microglia suggests that reduced astrocyte-synapse interaction is permissive for microglia-mediated synapse removal. There are multiple potential mechanisms by which reduced astrocyte-synapse interaction could facilitate microglia-mediated synapse removal, which remain open questions. One possibility is that astrocytes physically shield synapses from removal. Microglia are known to remove specific subsets of synapses by recognizing “eat-me” signals such as C1q and/or phosphatidyl serine that “tag” synapses for removal,⁵ requiring direct contact between microglia and synapses for their removal. Reductions in astrocyte-synapse contact could allow for increased contact between microglia and synapses, allowing greater access to these “eat me” signals. Another possibility is that reduced astrocyte-synapse interaction could weaken synapses prior to removal by microglia. Astrocytes are known to perform important metabolic, trophic, and signaling functions at synapses,¹ which are likely altered as astrocyte-synapse contact is reduced. As a result, synapses without astrocyte contact could have long term depression, reduced synapse stability, or impaired synapse maturation, leading to their weakening and eventual removal by microglia. Indeed, one study genetically reduced astrocyte morphological complexity via CRISPR/Cas9 depletion of ezrin in astrocytes.⁵⁵ This manipulation in adult mice led to increased glutamate spillover and increased NMDA receptor activation in the hippocampus. Importantly, these two alternatives are not mutually exclusive. Reduced astrocyte-synapse interaction could weaken synapses, leading to increased surface-tagging of synapses with “eat me” signals while also allowing microglia greater access to contact and remove synapses.

The mechanisms of microglia-astrocyte Wnt signaling and the retraction of astrocyte processes from synapses in the regulation of microglia-mediated synapse remodeling are demonstrated here in the neonate barrel cortex, but there is an intriguing possibility that this could be a broader mechanism by which microglia and astrocytes coordinate to remove synapses in health and disease. Due to its accessibility and the wealth of knowledge on its development and function, the barrel cortex is an ideal circuit to work out cellular/molecular mechanisms of synapse remodeling, which can then be tested in other contexts.⁵ By comparing our results to other transcriptomic studies in astrocytes, we found evidence that Wnt receptor signaling is activated in astrocytes in other contexts such as Alzheimer’s disease and epilepsy in which synapse loss and astrocyte morphological changes are known to occur. These data match a recent comparison of reactive astrocytes showing that β-catenin, the main effector of the Wnt signaling pathway, is among the core transcriptional regulators of astrocyte reactivity shared across divergent conditions.⁵⁶ Also, among the defining features of reactive astrocytes are morphological changes,⁵⁷ which may affect their association with synapses. Further evidence of the broad applicability of our findings is that WLS, the gene required for Wnt secretion, is among the genes most significantly reduced in microglia in human patients with Autism spectrum disorder⁵⁸ and that the Wnt signaling pathway is a target for neuropsychiatric drugs including lithium, methylphenidate (Ritalin), fluoxetine, and clozapine.^59,60 Evidence from mouse models also indicates that Wnt ligand expression in microglia can be affected by early life inflammatory insults, another key risk factor for neuropsychiatric disorders.⁶¹ Thus, microglia-astrocyte Wnt signaling may play a role in these disorders. Finally, under physiological conditions such as during sleep or learning and memory tasks, electron microscopy studies have shown that astrocytes change their association with synapses concomitant with changes in synapse numbers^28,55,62,63 and during in vivo time lapse imaging experiments, astrocyte-synapse coverage is predictive of synapse stability.⁶⁴ It is, thus, possible that Wnt-dependent microglia-astrocyte interactions also regulate synapse remodeling in these contexts, which are important future directions of our work.

In summary, our findings provide deep mechanistic insight into how microglia and astrocytes coordinate to remove synapses via Wnt signaling downstream of CX3CL1-CX3CR1 signaling in microglia. We further provide data that supports that astrocyte process rearrangement away from synapses is a key step in permitting microglia to engulf and remove synapses. Our in silico analysis supports that this Wnt-dependent mechanism could occur in a broad range of physiological and disease conditions where changes in astrocyte morphology and synapse elimination have been documented — an important and disease-relevant future direction of this work.

Limitations of the Study

Because our study relies on static, high and super resolution imaging, the temporal dynamics of when astrocytes retract and when microglia engulf synapses is still an open question. Three color, super-resolution live imaging would be necessary, but is not currently within our lab’s capabilities. It also remains unclear how CX3CR1 signaling regulates Wnt-mediated microglia-astrocyte crosstalk and which specific microglial Wnt is mediating this crosstalk in vivo. We suspect that regulation of Wnts in microglia via CX3CR1 is post-translational given the lack of transcriptional changes in Wnts in microglia following whisker removal in the barrel cortex. But, measuring the release of different Wnts at the protein level in vivo, unfortunately, lacks feasibility. A final caveat of our work is that the ‘eat me’ signal that drives microglia to engulf a synapse after astrocyte processes remodel remains to be determined. We ruled out complement in our previous work.⁶ Future proteomic work will be necessary to better define synapses in the barrel cortex destined for removal following astrocyte process remodeling—an important future direction of this work.

Resource Availability

Lead contact

Requests for further information and resources should be directed to and will be fulfilled by the lead contact, Dorothy Schafer (Dorothy.Schafer@umassmed.edu).

Materials availability

This study did not generate new unique reagents.

Data and code availability

TRAP-seq data and MERFISH processed data have been deposited at GEO as GSE252628, GSE298314 and are publicly available as of the date of publication. Processed MERFISH .vzg files for browsing on MERSCOPE visualizer software (Vizgen) and the MERFISH raw output files are available upon request. All original code is available at https://github.com/SchaferLabUMassChan/Faust_2025 and archived at Zenodo at https://doi.org/10.5281/zenodo.16764343. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

STAR Methods

Experimental Model and Study Participant Details

Wild-type C57Bl/6J (Stock #000664), Rosa26^mTmG (Stock# 007676, B6.129(Cg)-Gt(ROSA)^{26Sortm4(ACTB-tdTomato,-EGFP)Luo}/J),²⁰ Cx3cr1^EGFP (Stock# 005582, B6.129P2(Cg)-Cx3cr1^tm1Litt/J),⁶⁵ and TCF/Lef:H2B-GFP (Stock# 013752, STOCK Tg(TCF/Lef1-HIST1H2BB/EGFP)61Hadj/J)²⁴ mice were obtained from Jackson Laboratories. Aldh1l1^CreER mice (Stock# 031008, B6N.FVB-Tg(Aldh1l1-cre/ERT2)1Khakh/J)²¹ were obtained from Jackson Laboratories and backcrossed over 12 generations on a C57Bl/6J (Jackson Laboratories) background. Apc^Flox mice containing the Apc^tm1Tno allele²⁷ were generated by obtaining Apc^Flox/Flox Kras^LSL-G12D/+ Cdx2^+/+ mice (Stock #035169, B6.Cg-Kras^tm4Tyj Apc^tm1Tno Tg(CDX2-cre/ERT2)752Erf/MaraJ)⁶⁶ from Jackson Laboratories and crossing with C57Bl/6J mice to generate Apc^Flox/+; Kras^+/+; Cdx2^+/+ mice. Cx3cl1^−/− (Cx3cl1^tm1Lira) mice⁶⁷ were provided by Dr. Sergio Lira (Ichan School of Medicine, Mount Sinai). Wls^Flox (129S-Wls^tm1.1Lan/J) mice³⁶ were provided by Dr. David Rowitch (University of California, San Francisco). Aldh1l1^EGFP (STOCK Tg(Aldh1l1-EGFP)OFC789Gsat/Mmucd) mice⁶⁸ on a C57Bl6 background were provided by Dr. Ben Barres (Stanford University).⁵¹ Cx3cr1^Cre (STOCK Tg(Cx3cr1-cre)MW126Gsat/Mmcd) mice⁴¹ on a C57Bl/6N background were provided by Dr. Staci Bilbo (Harvard University)⁴² and subsequently backcrossed over 12 generations on a C57Bl/6J (Jackson Laboratories) background. Cx3cr1^{CreER/+ (Litt)} (B6.129P2(Cg)-Cx3cr1^{tm2.1(cre/ERT2)Litt}/WganJ) mice⁶⁹ were obtained from (Jackson Laboratories). Eef1a1^{LSL.eGFPL10a/+} (B6.Cg-Eef1a1^tm1Rck/J) mice⁷⁰ were made by Ana Domingos (Instituto Gulbenkian de Ciência, PT) and Dr. Jeff Friedman (Rockefeller University). Aldh1l1^EGFP-L10a (B6;FVB-Tg(Aldh1l1-EGFP/Rpl10a)JD130Htz/J) mice²² were originally made by Dr. Nathaniel Heintz (Rockefeller University) and provided by Dr. Staci Bilbo (Duke University). All mice were maintained on a C57Bl/6J background. Sprague-Dawley Rats (Catalog # SD-001, Crl:CD(SD)) were obtained from Charles River Laboratories. Animals were maintained on a 12 hour light/dark cycle with food and water provided ad libitum. All animals were healthy and were not immune compromised. Experimental animals were not involved in previous procedures or studies. Littermates of the same sex were randomly assigned to experimental groups. Both males and females were used for all experiments. The sex of each individual data point is indicated in the figures by squares (male) or circles (female). Experiments were performed in mice aged 1–12 days old or rats 1 day old. The specific ages used for each experiment are indicated in the figures. All animal experiments were performed in accordance with Animal Care and Use Committees (IACUC) and under NIH guidelines for proper animal use and welfare.

Method Details

Whisker lesioning

Whisker lesioning was performed as described previously.⁶ Briefly, whiskers of anesthetized P4 mice were unilaterally lesioned by cauterization of the right whisker pad with a high temperature cautery (Bovie Medical Corporation; Clearwater, FL). Care was taken to ensure that all whiskers on that side were lesioned.

Immunohistochemistry

Mice were anesthetized and transcardially perfused with 0.1 M phosphate buffer (PB), followed by 4% paraformaldehyde (PFA, Electron Microscopy Sciences; Hatfield, Pennsylvania) in 0.1 M PB. For analyses in layer IV of the barrel cortex, cortical flatmounts were prepared as described previously in order to visualize the entire barrel field in a single section.⁶ Briefly, the cortex from the right and left hemisphere was dissected away from the midbrain, placed in open-face custom molds with a 250 μm-thick chamber 3D-printed using ABS-M30 material (Stratasys Inc.; Eden Praire, Minnesota), submerged in 4% PFA solution, covered with a glass slide to gently flatten, and post-fixed overnight at 4°C. For analyses in coronal sections, whole brain tissue was post-fixed overnight at 4°C in 4% PFA solution. The following day, tissue was washed three times in 0.1 M PB before storing in 30% sucrose in 0.1 M PB. 40 μm thick sections were cut using a microtome or a vibratome. They were subsequently blocked for 1 hour in 0.1 M PB containing 10% normal goat serum (NGS; G9023 Sigma-Aldrich; St. Louis, MO) and 0.3% Triton X-100 (X100 Sigma-Aldrich) (PBTGS). Blocked sections were then incubated with primary antibodies in PBTGS overnight at room temperature, washed 3 times with 0.1 M PB, incubated for 2–3 hours with secondary antibodies in PBTGS at room temperature, washed 3 times with 0.1 M PB, and mounted on glass slides using Fluoroshield mounting media containing DAPI (F6057 Sigma-Aldrich). For immunofluorescent analysis of TCF/Lef:H2B-GFP tissue, sections were stained with Hoechst 33342 dye (H3570 Thermo Fisher Scientific; Waltham, MA) and mounted using CitiFluor mounting medium (CFM-3 Electron Microscopy Sciences).

For immunofluorescent analysis of β-catenin, sucrose-incubated whole brain tissue was cryo-embedded in a 2:1 mixture of 30% sucrose and Tissue-Tek OCT (Sakura Finetek; Torrance, California), frozen on dry-ice and stored at −80°C. 20 μm coronal sections containing whisker barrels were mounted on slides, washed 3 times in phosphate buffered saline (PBS) containing 0.2% Triton-X (PBST), and incubated in Liberate Antibody Binding Solution (24310 Polysciences, Inc; Warrington, PA) overnight to quench endogenous Cx3cr1^EGFP signal. Sections were then washed 3 times with PBST, placed in 10 mM sodium citrate pH 6.0 with 0.05% Tween warmed to 95°C for 40 min for antigen retrieval. Following antigen retrieval, sections were washed once with PBST, blocked for 1 hour in 5% normal donkey serum (NDS) with 0.3 M glycine in PBS and incubated overnight at 4°C with primary antibodies in 1% NDS in PBS. After 3 washes in PBST, sections were then incubated with secondary antibodies raised in donkey in 1% NDS in PBS for 1 hour at room temperature, washed 3 times with PBST, and mounted with Fluoroshield mounting media containing DAPI (F6057 Sigma-Aldrich).

Primary antibodies used: guinea pig anti-VGluT2 (1:2000 for synapse area; 1:1000 for synapse engulfment; 135 404 Synaptic Systems; Göttingen, Germany), chicken anti-GFP (1:1000; ab13970 Abcam; Cambridge, United Kingdom), rat anti-CD68 (1:200; MCA1957 Bio-Rad; Hercules, CA), rat anti-LAMP2 (1:200; ab13524 Abcam), rabbit anti-SOX9 (1:500; ab185966 Abcam), mouse anti-β-catenin Alexa Fluor 647 (1:100; sc-7963 AF647 Santa Cruz Biotechnology; Dallas, TX), chicken anti-IBA1 (1:500; 234 009 Synapatic Systems), rat anti-V5 (1:100; Ab00136–6.1 Absolute Antibody; Boston, MA). Secondary antibodies were used at 1:1000 dilution: goat anti-chicken IgY Alexa Fluor 488 (A-11039 Thermo Fisher Scientific), goat anti-guinea pig IgG Alexa Fluor 594 (A-11076 Thermo Fisher Scientific), goat anti-rat IgG Alexa Fluor 647 (A-21247 Thermo Fisher Scientific), goat anti-rabbit IgG Alexa Fluor 647 (A-21245 Thermo Fisher Scientific), goat anti-rabbit IgG Alexa Fluor 488 (A-11034 Thermo Fisher Scientific), goat anti-rat IgG Alexa Fluor 488 (A-11006 Thermo Fisher Scientific), donkey anti-guinea pig IgG Alexa Fluor 488 (706–545-148 Jackson ImmunoResearch Labs; West Grove, PA), donkey anti-rabbit IgG Alexa Fluor 594 (A-21207 Thermo Fisher Scientific).

Synapse density analysis

Analysis of synapse density was performed as previously described.⁶ Briefly, fluorescent immunolabeled 40 μm thick brain sections were imaged at 63x on a Zeiss LSM700 scanning confocal microscope equipped with 405nm, 488nm, 555nm, and 639nm lasers and ZEN acquisition software (Zeiss; Oberkochen, Germany). For each animal, single plane images centered within a barrel in layer IV of the barrel cortex were acquired from 4 fields of view (FOV) per hemisphere. Images were analyzed in FIJI software by experimenters blind to the identity of the samples.⁷¹ After background subtraction (rolling ball radius 10 pixels) and despeckling, the VGluT2⁺ signal was manually thresholded and the Analyze Particles function (particle size 0.2 pixels to infinity; ImageJ plugin, NIH) was used to measure the total area of VGluT2⁺ presynaptic terminals. Data for each hemisphere per animal were averaged across the 4 FOV.

Engulfment analysis

Analysis of engulfment of VGluT2⁺ material was performed as previously described.^6,9,72 Briefly, fluorescent immunolabeled 40 μm thick brain sections were imaged at 63x on a Zeiss Observer Spinning Disk Confocal microscope equipped with diode lasers (405nm, 488nm, 594nm, 647nm) and ZEN acquisition software (Zeiss). For each animal, z-stack images (0.22 μm step size) centered within a barrel in layer IV of the barrel cortex were acquired from 6 to 8 FOV per hemisphere for microglia and 4 FOV for astrocytes. Images were background subtracted (rolling ball radius 50 pixels for microglia and astrocytes, 10 pixels for CD68 and LAMP2, 10 pixels for VGluT2) using FIJI software, then processed in Imaris (Bitplane; Zurich, Switzerland) as previously described.^6,9,72 Data analysis was performed by experimenters blind to the identity of the samples. For engulfment analysis in microglia, data for each hemisphere per animal was averaged across 15–20 individual cells per animal. For engulfment analysis in astrocytes, since it was difficult to determine the boundaries of individual astrocytes, surface renderings were generated for all astrocytes within the barrel region of each z-stack image. Each FOV contained ~6–10 astrocytes within the barrel region for a total of ~24–40 astrocytes per hemisphere per animal. Engulfment analysis was restricted to anti-VGluT2⁺ material contained within CD68⁺ lysosomes for microglia and LAMP2⁺ lysosomes for astrocytes. Data for each hemisphere per animal were averaged across the 4 FOV.

Tamoxifen injections

For expansion microscopy experiments (see Expansion microscopy), Aldh1l1^CreER; Rosa26^mTmG/+ mice, Aldh1l1^CreER; Rosa26^mTmG/+; Cx3cl1^−/− mice, and Aldh1l1^CreER; Rosa26^mTmG/+; Apc^Flox/Flox mice were injected once daily intraperitoneally (i.p.) with 50 μg tamoxifen (T5648 Sigma-Aldrich) dissolved in corn oil at 1 mg/mL at ages P1, P2, P3, and P4. For TRAP-Seq experiments (see TRAP-Seq), Cx3cr1^{CreER (Litt)}; Eef1a1^{LSL.eGFPL10a/+} mice were orally administered 50 μg tamoxifen (10 mg/mL in corn oil) once daily at ages P1, P2, and P3.

Expansion microscopy

For expansion microscopy, 40 μm tangential brain sections were prepared as described above (see Immunohistochemistry). Sections were incubated overnight in Liberate Antibody Binding Solution (24310 Polysciences, Inc) for antigen retrieval and to eliminate endogenous mTomato fluorescence in Rosa26^mTmG mice, then washed three times in 0.1 M PB. Sections were then blocked for 4 hours in PBTGS or 0.1 M PB with 10% NDS and 0.3% Triton-X (PBTDS) and incubated 3 days overnight at 4°C with primary antibody diluted in PBTGS or PBTDS. For samples from mice aged P10 or older, we used PBTGS containing 1% Triton-X. Following primary antibody incubation, sections were washed three times in 0.1 M PB, incubated overnight at room temperature with secondary antibody diluted in PBTGS or PBTDS, and washed 3 times in 0.1 M PB.

Primary antibodies used: guinea pig anti-VGluT2 (1:1500; 135 404 Synaptic Systems), chicken anti-GFP (1:1000; ab13970 Abcam), rat anti-V5 (1:100; Ab00136–6.1 Absolute Antibody), rabbit anti-ALDH1L1 (1:500; 85828 Cell Signaling Technology; Danvers, MA). Secondary antibodies were used at 1:300 dilution: goat anti-chicken IgY Alexa Fluor 488 (A-11039 Thermo Fisher Scientific), goat anti-guinea pig IgG Alexa Fluor 594 (A-11076 Thermo Fisher Scientific), goat anti-rat IgG Alexa Fluor 594 (A-11007 Thermo Fisher Scientific), goat anti-guinea pig IgG Alexa Fluor 488 (A-11073 Thermo Fisher Scientific), goat anti-rabbit IgG, Atto 647N (40839 Sigma-Aldrich), donkey anti-guinea pig IgG Alexa Fluor 488 (706–545-148 Jackson ImmunoResearch), donkey anti-rat IgG Alexa Fluor 594 (A-21209 Thermo Fisher Scientific).

Prior to gel embedding and expansion, pre-expansion images of the barrel cortex region of immunolabeled brain sections were acquired at 10x and 20x on a Zeiss Observer Spinning Disk Confocal microscope equipped with diode lasers (405nm, 488nm, 594nm, 647nm) and ZEN acquisition software (Zeiss). Sections were then embedded in polyacrylamide gels and expanded in de-ionized H₂O as previously described.^73,74 Briefly, brain sections were incubated overnight at room temperature in PBS containing 0.01% acryloyl-X SE (A20770 Thermo Fisher Scientific), washed twice with PBS for 15 minutes, and incubated on ice for 30 minutes in gelling solution [0.2% tetramethylethylenediamine (T9281 Sigma-Aldrich), 0.2% ammonium persulfate (A3678 Sigma-Aldrich), 0.01% 4-hydroxy-TEMPO (176141 Sigma-Aldrich), 8.11% sodium acrylate (QC-1489 Combi Blocks; San Diego, CA), 2.35% acrylamide (A9099 Sigma-Aldrich), 0.14% N,N′-methylenebisacrylamide (M7279 Sigma-Aldrich), 10.987% NaCl (S271 Sigma-Aldrich) in PBS]. Gelation chambers were then constructed by transferring brain sections onto glass slides (one section per slide), flanking the section with two stacks of two #1.5 glass coverslips (~170 μm thick per coverslip), covering the sections with gelling solution, and topping with a glass slide wrapped in parafilm. Completed gelation chambers were incubated for 2 hours at 37°C to polymerize the gels. Polymerized gels were then removed from the gelation chambers, trimmed down with a razor blade, and digested overnight at room temperature with 8 units/mL Proteinase K (P8107S New England BioLabs; Ipswich, MA) in buffer containing 0.5% Triton X-100 (X100 Sigma-Aldrich), 2 mM ethylenediaminetetraacetic acid (EDTA), 50 mM Tris, and 0.8 M NaCl. Digested gels were either expanded for imaging or washed 3 times for 20 minutes with PBS and stored at 4°C for up to a week.

Gels were expanded for imaging by washing 3 times for 20 minutes with de-ionized H₂O, including a 20 minute incubation in Hoechst 33342 dye (H3570 Thermo Fisher Scientific) during the middle wash. Gels were further trimmed down to the region of interest using a razor blade and transferred to a 6-well plate with a #1.5 cover glass bottom (P061.5HN Cellvis; Burlington, Ontario) pre-coated with 0.1% poly-L-lysine (P8920 Sigma Aldrich) for imaging. Images were acquired using a Zeiss Observer Spinning Disk Confocal microscope equipped with diode lasers (405nm, 488nm, 594nm, 647nm), inverted objective lenses, and ZEN acquisition software (Zeiss) and a Leica Microsystems STELLARIS 8 STED microscope with a 405 nm diode laser and a white light laser with HyD detectors, inverted objective lenses, and LASX software (Leica Microsystems). Post-expansion images of each gel were acquired at 10x in a FOV contained within the FOV of the pre-expansion images of the same gel. The expansion index for each gel was calculated using FIJI software by comparing the distances between landmarks identified in both the pre-expansion and post-expansion image for at least 3 sets of measurements per gel. Examples of VGluT2⁺ and GFP⁺ landmarks are shown in Figure S2E. The expansion index was calculated for each gel by averaging across these measurements for each gel. For data acquisition, z-stack images centered within a barrel in layer IV of the barrel cortex were acquired from 4–6 FOV with water-immersion objectives (40x/1.1 on Zeiss Observer or 63x/1.2 NA on Leica STELLARIS 8). Care was taken to avoid unlabeled astrocytes.

For data analysis, z-stack images were background subtracted using FIJI software, then normalized by layer in Imaris (Bitplane; Zurich, Switzerland). Astrocytes labeled with membrane-bound GFP in Aldh1l1^CreER/+; Rosa26^mTmG/+ tissue were 3D rendered using the Imaris surface tool with auto thresholding, 0.25 μm smoothing, voxel size > 50, and sphericity < 0.73. Astrocytes labeled with anti-V5 were 3D rendered using the Imaris surface tool using LabKit, a FIJI plugin for machine learning pixel classification.⁷⁵ For each animal, the same LabKit classification settings were used for all z-stack images for both control and deprived hemispheres. The total density of astrocyte processes was determined by dividing the volume of the astrocyte surface rendering by the total volume of the z-stack image. VGluT2⁺ synapses were 3D rendered using the Imaris surface tool with manual thresholding with 4 μm local background subtraction, 0.25 μm smoothing, and voxel size > 50. Synapse-astrocyte nearest neighbor distances (NND) were calculated by Imaris using VGluT2⁺ surfaces less than 60 μm³. VGluT2 surfaces with a NND of 0 were considered contacted by astrocytes and VGluT2 surfaces with NND less than 0 were excluded from analysis. Synapse ensheathment by astrocytes (the percent of synapse surface area contacted) was calculated for each VGluT2⁺ surface less than 60 μm³ using the Imaris Surface-Surface Contact Area XTension (https://github.com/Ironhorse1618/Python3.7-Imaris-XTensions/blob/master/XT_MJG_Surface_Surface_ContactArea2.py). For each z-stack image, we calculated the total density of astrocyte processes, the mean synapse-astrocyte NND, the percentage of synapses contacted by the astrocytes, and the mean synapse ensheathment by astrocytes. Data for each hemisphere per animal were averaged across the 4–6 FOV.

Electron microscopy

For electron microscopy, Rosa26^mTmG/+ mice were anesthetized on ice and transcardially perfused at a rate of 6–8 mL/min with 37°C 0.1 M PB to remove blood followed by 60–80 mL of 37°C fixative solution containing 2.5% glutaraldehyde (Electron Microscopy Sciences) and 2% PFA (Sigma) in cacodylate buffer (0.1 M sodium cacodylate (Electron Microscopy Sciences), 0.04% CaCl₂ (Sigma), pH 7.4). Perfused mice were held head-down for 15 minutes, then brains were dissected out and post-fixed in the same fixative at 4°C overnight. After washing the samples with ice-cold cacodylate buffer, brains were sectioned to 200 μm thickness using a vibratome. Coronal brain sections containing layer IV of the barrel cortex region were identified by endogenous Rosa26^mTmG fluorescence using a Zeiss Observer microscope equipped with Zen acquisition software (Zeiss). Layer IV of the left and right barrel cortex regions was dissected out and stored in fixative solution on ice before processing and embedding.

Samples were then post-fixed in 1% osmium tetroxide aqueous solution (Electron Microscopy Sciences) with 2.5% potassium ferrocyanide (Millipore Sigma) at room temperature for 1 hour. Sections were rinsed in water and then maleate buffer (Millipore Sigma) (pH=6.0) and stained in 0.05 M maleate buffer containing 1% uranyl acetate (Electron Microscopy Sciences) at 4°C overnight. Sections were dehydrated in series of washes from 5% to 100% ethanol, then two changes of propylene oxide before being infiltrated overnight at room temperature with 1:1 Embed 812 resin: propylene oxide. Finally, sections were embedded in 100% Embed 812 resin between two sheets of Aclar Embedding film (Ted Pella, Inc.) and polymerized at 60°C for 48 h. Once polymerized, the embedded brain sections were mounted on blank sectioning blocks with a single drop of embedding resin and allowed to polymerize overnight before being trimmed and serial sectioned (60 nm thick). The sections were placed on 200 mesh grids, then stained with 1% uranyl acetate and aqueous lead citrate before being examined using a Tecnai 12 Spirit TEM at 120 kV accelerating voltage. 11500x images were collected using a Gatan Rio9 CCD camera and Gatan DM4 Digital Micrograph software. For each sample, images from 30–50 FOV were collected from the central portion of the section, the portion corresponding to layer IV of the barrel cortex. To avoid double-counting synapses that might appear in multiple consecutive sections, we only collected images from every 3rd section, for a total of 2–3 sections per sample.

Data analysis was performed using FIJI software by experimenters blind to the identity of the samples. In each image, synapses were identified by the presence of an electron-dense synaptic cleft and astrocytes were identified by their irregular shape, electron-sparse cytoplasm, and glycogen granules. For each synapse, it was determined whether there was direct contact with an astrocyte process and the distance was measured from the synaptic cleft to the nearest astrocyte process. We identified a total of 50–70 synapses per sample and averaged the data for each hemisphere per animal.

Astrocyte cell body density analysis

Analysis of astrocyte cell body density was performed using single plane 20x epifluorescence images of fluorescent immunolabeled 40 μm thick brain sections containing the barrel cortex region acquired with a Zeiss Observer microscope equipped with Zen acquisition software (Zeiss). The total number of SOX9⁺ astrocyte nuclei was counted in 4 VGluT2⁺ barrels per hemisphere using FIJI software and divided by the total area of the 4 barrels to determine the astrocyte cell body density.

Translating Ribosome Affinity Purification (TRAP) sequencing

Mice were euthanized with CO₂, and brain regions of interest were dissected and flash-frozen for TRAP. Ribosome-associated mRNA from microglia was isolated from each region as previously described,⁷⁶ where each sample corresponds to a single mouse. Briefly, the brain tissue was thawed in an ice-cold Wheaton 33 low extractable borosilicate glass homogenizer containing 1 mL cell-lysis buffer (20 mM HEPES-KOH, pH 7.3, 150 mM KCl, 10 mM MgCl2, and 1% NP-40, 0.5 mM DTT, 100 μg/ml cycloheximide, and 10 μl/ml rRNasin (Promega; Madison, WI) and Superasin (Thermo Fisher Scientific). After manually homogenizing the samples with 3–5 gentle strokes using the PTFE homogenizer (grinding chamber clearance 0.1 to 0.15mm), samples. Samples were then homogenized in a motor-driven overhead stirrer at 900 r.p.m. with 12 full strokes. The samples were transferred to chilled Eppendorf tubes, and a post-nuclear supernatant was prepared by centrifugation at 4°C, 10 minutes, 2,000 × g. To the supernatant, NP-40 (final concentration = 1%) and DHPC (final concentration = 30 mM) were added. Post-mitochondrial supernatant was prepared by centrifugation at 4°C, 10 minutes, 16,000 × g. 200 uL Streptavidin MyOne T1 Dynabeads (Thermo Fisher Scientific), conjugated to 1 μg/μl biotinylated Protein L (Pierce Biotechnology; Waltham, MA) and 50 μg each of anti-eGFP antibodies Htz-GFP-19F7 and Htz-GFP-19C8, bioreactor supernatant (Memorial-Sloan Kettering Monoclonal Antibody Facility) were added to each supernatant. After overnight incubation at 4°C with gentle end-over-end rotation, the unbound fraction was collected using a magnetic stand. The polysome-bound beads were washed with high-salt buffer (20 mM HEPES-KOH, pH 7.3, 350 mM KCl, 10 mM MgCl2, 1% NP-40, 0.5 mM DTT, and 100 μg/ml cycloheximide). RNA clean-up from the washed polysome-bound beads and 5% of the unbound fractions was performed using RNeasy Mini Kit (Qiagen; Venlo, Netherlands) following the manufacturer’s instructions. RNA integrity was assayed using an RNA Pico chip on a Bioanalyzer 2100 (Agilent; Santa Clara, CA), and only samples with RIN > 9 were considered for subsequent analysis. Double-stranded cDNA was generated from 1–5 ng of RNA using the Nugen Ovation V2 kit (NuGEN; San Carlos, CA) following the manufacturer’s instructions. Libraries for sequencing were prepared using Nextera XT kit (Illumina; San Diego, CA) following the manufacturer’s instructions. The quality of the libraries was assessed by 2200 TapeStation (Agilent). Multiplexed libraries were directly loaded on NovaSeq (Ilumina) with single-read sequencing for 75 cycles. Raw sequencing data were processed by using Illumina bcl2fastq2 Conversion Software v2.17.

Bioinformatic analysis of RNA sequencing datasets

Raw sequencing reads were first quality checked and trimmed using Trim Galore (https://www.bioinformatics.babraham.ac.uk/projects/trim_galore/ v0.6.4; a wrapper program implementing Cutadapt v2.9 https://journal.embnet.org/index.php/embnetjournal/article/view/200 and FastQC v0.11.9 https://www.bioinformatics.babraham.ac.uk/projects/fastqc/) and mapped to the mouse genome (mm10) using the HISAT2 package (v2.2.0).⁷⁷ Reads were counted using featureCounts (v2.0.0) against the Ensembl v99 annotation.⁷⁸ The raw counts were processed through a variance stabilizing transformation (VST) procedure using the DESeq2 package⁷⁹ to obtain transformed values more suitable than the raw read counts for certain data mining tasks. All pairwise comparisons were performed on the count data of entire gene transcripts using the DESeq2 package (v1.36.0).⁷⁹

To determine TRAP-enriched genes, gene expression was compared between bound and unbound fractions. We used statistical cutoffs for astrocytes (p-value < 0.05, |fold change| > 2, basemean > 5) and microglia (p-value < 0.05, |fold change| > 2, basemean >5) to determine input-enriched genes for each cell type for downstream analysis. For analysis of differentially expressed genes (DEGs) between deprived and non-deprived samples, we restricted analysis to the input-enriched genes and used statistical cutoffs p-value < 0.05 and |fold change| > 1.2 for both astrocytes and microglia. Analysis of root regulators was performed on astrocyte DEGs with the use of QIAGEN Ingenuity Pathway Analysis (QIAGEN Inc., https://digitalinsights.qiagen.com/IPA).⁸⁰ Gene Set Enrichment Analysis was performed on astrocyte TRAP-enriched genes for gene sets corresponding to gene ontology (GO) terms using GSEA software (v4.3.3)^81,82 and the M5 Gene Ontology Molecular Signatures Database (m5.go.v2024.1.Mm.symbols.gmt).^83,84 The GSEA Preranked module was used for the analysis. Pathways with less than 3 genes or more than 2000 genes were excluded. Pathways with FDR < 0.1 were identified as significantly over-represented sets and loaded into Cytoscape for visualization. The Enrichment Map plug-in⁸⁵ was used to identify gene sets with overlapping genes. Manual labels were given to 4 groups of gene sets with similar functions and high numbers of overlapping genes.

Ligand-receptor analysis was performed in RStudio using NicheNet v2.0.2 (https://github.com/saeyslab/nichenetr).³⁴ NicheNet analysis of microglia-astrocyte ligand-receptor interactions was performed using astrocyte DEGs as the “target” gene list, astrocyte input-enriched genes as the “receiver-expressed” gene list and “background” gene list, and microglia input-enriched genes as the “sender-expressed” gene list. To compare the results of this study to publicly available astrocyte TRAP-Seq datasets, NicheNet was performed for each dataset using astrocyte DEGs as the “target” gene list, astrocyte expressed genes as the “receiver-expressed” gene list and “background” gene list, and all ligands in the NicheNet database to reflect an unknown sender. Putative ligands were ranked based on the area under the precision-recall curve (AUPR). The ligands with the highest predicted activity in each dataset were given the rank of 1.

Analysis of Wnt signaling by immunofluorescence

Analysis of TCF/Lef:H2B-GFP reporter activity and nuclear β-catenin were performed using single plane 63x confocal images of immunolabeled brain sections containing the barrel region using a Zeiss Observer Spinning Disk Confocal microscope equipped with diode lasers (405nm, 488nm, 594nm, 647nm) and ZEN acquisition software (Zeiss). Data analysis was performed by experimenters blind to the identity of the samples. For each animal, images centered within a barrel in layer IV of the barrel cortex were acquired from 6 to 10 FOV per hemisphere. Using FIJI software, images were background subtracted (rolling ball radius 50 pixels) and manually thresholded using the SOX9 channel to identify astrocyte nuclei or using DAPI to identify all nuclei. The Analyze Particles function (particle size 200 pixels to infinity; ImageJ plugin, NIH) was used to measure the area and mean fluorescence intensity (MFI) in the nucleus (SOX9⁺ and/or DAPI⁺ area) of each cell. For each animal, we quantified the average MFI for all cells in each hemisphere with area > 15 μm².

Preparation of primary astrocyte-fixed neuron cultures

Primary rat cortical astrocytes and cortical neurons were prepared from P1 Sprague-Dawley Rats obtained from Charles River Laboratories. Cortices were dissected after rapid decapitation of rat pups, then chopped into ~1 mm³ pieces. Cortical pieces from 3 animals were added to 1 vial of Papain (LK003178 Worthington Biochemical; Lakewood, NJ) supplemented with DNAse (LS002007 Worthington Biochemical) for 45 minutes at 32°C. Cortical pieces were then triturated in sequential ovomucoid solutions (LS003086 Worthington Biochemical) and pelleted. At this point, the samples were split for astrocyte and neuron cultures.

For neuron cultures, cell pellets were resuspended in panning buffer (DPBS (Gibco; 14287 Thermo Fisher Scientific) supplemented with bovine serum albumin (A8806 Sigma-Aldrich) and insulin (I1882 Sigma-Aldrich). The cells were filtered through a 20 μm mesh filter (03–20/14 Elko Filtering Co.) then incubated on two sequential negative panning dishes coated with Bandeiraea simplicifolia Lectin I (L-1100–5 Vector Laboratories; Youngstown, OH), followed by goat anti-mouse IgG + IgM (115–005-044 Jackson ImmunoResearch) and goat anti-rat IgG + IgM (112–005-044 Jackson Immunoresearch). The cells were then moved to positive panning dishes coated with anti-rat L1 (ASCS4, Developmental Studies Hybridoma Bank, Univ. Iowa) to bind the cortical neurons. After 45 minutes, the plate was washed with DPBS 5 times and then cells were washed off the dish by forceful pipetting with a P1000. The neurons were pelleted at 200 × g for 11 minutes and resuspended in serum-free neuron growth media (NGM; Neurobasal (21103049 Thermo Fisher Scientific), B27 supplement (12587010 Thermo Fisher Scientific), 2 mM L-glutamine (25–030-081 Thermo Fisher Scientific), 100 U/mL Pen/Strep (15140122 Thermo Fisher Scientific), 1 mM sodium pyruvate (11360070 Thermo Fisher Scientific), 4.2 μg/mL forskolin (F6886 Sigma-Aldrich), 50 ng/mL BDNF (450–02 Pepro-tech; Cranbury, NJ), and 10 ng/mL CNTF (450–13 PeptroTech)). 100,000 neurons were plated on 12 mm glass coverslips coated with poly-D-lysine (P6407 Sigma) and 2 μg/mL laminin (3400–010-02 R&D Systems; Minneapolis, MN). Neurons were cultured in a 37°C incubator in 10% CO₂. After 2 days in vitro (DIV), half of the neuronal media per well was replaced with neurobasal plus growth media (Neurobasal Plus (A3582901 Thermo Fisher Scientific), B27 Plus (A3653401 Thermo Fisher Scientific), 100 U/mL Pen/Strep, 1 mM sodium pyruvate, 4.2 μg/mL forskolin, 50 ng/mL, BDNF, and 10 ng/mL CNTF) supplemented with 4 μM AraC (C1768 Sigma-Aldrich). On DIV 3, the media was removed completely and replaced with neurobasal plus growth media. Every 3 days, half of the media was replaced with fresh neurobasal plus growth media. On DIV 10, neurons were fixed with 500 μl of ice-cold methanol for 3 minutes. The methanol was carefully removed and the plate was left in the fume hood to dry completely. Fresh neurobasal plus growth media was added and equilibrated in the incubator before astrocytes were added.

For astrocyte cultures, cell pellets were resuspended in astrocyte growth media (DMEM (GIBCO; 11960 Thermo Fisher Scientific), 10% fetal bovine serum (F2442 Sigma-Aldrich), 10 μM hydrocortisone (H0888 Sigma-Aldrich), 100 U/mL Pen/Strep, 2 mM L-glutamine, 5 μg/mL insulin, 1 mM sodium pyruvate, 5 μg/mL N-acetyl-L-cysteine (A8199 Sigma-Aldrich)). 15 to 20 million cells were seeded on 75mm² flasks coated with poly-D-lysine and incubated at 37°C in 10% CO₂. On DIV 3, non-astrocytic cells were removed via vigorous shaking of the flasks until only a monolayer of astrocytes remained. On DIV 5, astrocytes were treated with AraC to reduce fibroblast proliferation. On DIV 7, astrocytes were trypsinized (0.05% Trypsin-EDTA (25300054 Thermo Fisher Scientific)) and plated into 6 well plates at a density of 400,000 cells per well. On DIV 10, astrocytes were trypsinized again, resuspended in serum-free neurobasal plus growth media (see above) and plated on top of methanol-fixed neurons (see above) at a density of 15,000–20,000 astrocytes per well.

Morphological analysis of primary astrocytes

For analysis of astrocyte morphology, astrocytes were transfected with 2 μg per well of pMAX-GFP (Lonza; Basel, Switzerland) via Lipofectamine LTX with PLUS Reagent (15338100 Thermo Fisher Scientific) on DIV 8, prior to plating onto methanol-fixed neurons. On DIV10, 6 hours after plating onto methanol-fixed neurons, astrocytes were treated with Wnt recombinant proteins at either 50 or 100 ng/mL. Wnt proteins were obtained from Creative Biomart (Wnt4: WNT4–517HFL, Wnt5a: Wnt5a-2054M, and Wnt7a: WNT7A-42H), reconstituted in PBS and then stored in single-use aliquots at −20°C. 72 hours later, cells were fixed with warmed 4% PFA at room temperature for 7 minutes. After fixing, the cells were blocked in 5% Normal Goat Serum (005–000-121 Jackson ImmunoResearch) and 0.02% Triton X-100 (28314 Thermo Fisher Scientific) for 30 minutes at room temperature and then incubated with a primary antibody against GFP (Aves GFP-1020, dilution 1:1000) diluted in blocking solution without Triton X-100 for 1 night at 4°C. Alexa-fluor 488 conjugated Goat anti-Chicken IgY (A11039 Thermo Fisher Scientific) was diluted 1:500 in blocking solution and incubated for 2 hours at room temperature. Coverslips were mounted with Vectashield with DAPI (H-1200–10 Vector Laboratories) on Superfrost plus slides (48311–703 VWR; Radnor, PA) and sealed with nail polish.

Coverslips were imaged on a Keyence BZ-X800 fluorescent microscope. Images were acquired with a 40X dry objective, at standard resolution, blind to experimenter. Astrocyte morphological analysis was performed with the Sholl Analysis function in ImageJ (NIH). Sholl analysis data were analyzed using custom R code (https://github.com/Eroglu-Lab/In-Vitro-Sholl).⁸⁶ Three independent culture experiments were performed to achieve an n of 3.

Multiplexed error-robust fluorescent in situ hybridization (MERFISH)

For MERFISH experiments, mice were anesthetized on ice and sacrificed, then brains were immediately extracted and embedded in Tissue-Tek OCT (Sakura Finetek), immersed in 2-methylbutane that had been chilled to −80° C with dry ice, and stored at −80°C. Samples were prepared according to the Vizgen Fresh and Fixed Frozen Tissue Sample Preparation guide (Revision E9160002 Vizgen; Cambridge, MA). Throughout the protocol, all surfaces and tools were cleaned with RNAse-Zap (R2020, Sigma Aldrich) and solutions were prepared using RNAse-free distilled water (10977023, Thermo Scientific). 10 μm thick coronal slices containing the barrel cortex were prepared using a cryostat. Sections containing the barrel cortex were adhered to functionalized coverslips (10500001, Vizgen) pre-coated with fluorescent beads. Each coverslip contained 1 section from a Cx3cl1^−/− mouse and 1 section from a wild-type littermate control. Tissue on coverslips was fixed for 15 minutes at room temperature in 4% paraformaldehyde in PBS, followed by three washes with PBS. Tissue was then photobleached for 3 hours to remove autofluorescence using the MERSCOPE Photobleacher (1010003, Vizgen) and permeabilized in 70% ethanol overnight at 4°C. The tissue was then washed with PBS, and incubated with blocking solution containing 1:10 murine RNAse inhibitor (M0314L New England Biolabs) in blocking buffer (20300100, Vizgen) for 1 hour. This was followed by 1 hour of incubation with the primary antibody, purified rabbit anti-VGluT2 (ab227906, Abcam) diluted 1:100 in blocking solution. The tissue was then washed three times with PBS and incubated for 1 hour with an oligo-conjugated secondary anti-rabbit antibody (20300102, Vizgen), 1:100 in blocking solution containing 1:10 murine RNAse inhibitor (M0314L New England Biolabs). The sample was washed three times with PBS and fixed for 15 minutes at room temperature in 4% paraformaldehyde in PBS, followed by three washes with PBS. Next, the tissue was washed with 2X saline sodium citrate (SSC) and then formamide wash buffer (30% formamide in 2X saline sodium citrate, or SSC) for 30 minutes at 37°C. Afterwards, a custom MERFISH library mix (Table S7) was added and allowed to hybridize for 48 hours at 37°C in a humidified chamber. Sample was then washed and incubated at 47°C with formamide wash buffer twice for 30 minutes each, and once with 2X SSC for 2 minutes at room temperature. The tissue was then embedded in a polyacrylamide gel (4% 19:1 acrylamide/bis-acrylamide, 17% 5M NaCL, 17% 1M Tris pH 8) followed by incubation with tissue clearing solution (2X SSC, 2% SDS, 0.5% v/v Triton X-100, and proteinase K 1:100) overnight at 37°C. To prepare for imaging on the MERSCOPE instrument, a 140-gene imaging kit (10400004 Vizgen) and the included reagents were thawed and inverted 10 times to mix before adding subsequent reagents. Following the MERSCOPE Instrument User Guide (Revision G 91600001 Vizgen), the tissue was washed with DAPI and PolyT staining reagent (2030021 Vizgen) for 15 minutes, formamide wash buffer for 10 minutes, and 2X SSC for 2 minutes, all at room temperature. After washing, the slide was assembled into the flow chamber and placed into the MERSCOPE instrument for imaging. MERFISH imaging was performed as previously described⁸⁷ with parameter files provided by Vizgen and protein intensity set to medium. Briefly, a low-resolution mosaic image was first acquired (405 nm channel) with a low magnification objective (10x). The imaging area was then manually selected from this mosaic, with both brains included in the same area. After switching to a high-magnification objective, the MERSCOPE completed a fully-automated MERFISH experiment.

Following image acquisition, the resulting data were decoded using Vizgen’s analyzing pipeline incorporated in the MERSCOPE. Cell boundaries were segmented with Vizgen’s Postprocessing Tool (Vizgen) using the CellPose algorithm based on the DAPI signal. Segmentation was performed on Z=3 z-plane and cell borders were propagated to z-planes above and below. The volume, X position, and Y position of these cell boundaries, as well as the probe counts within each cell boundary, were output for further analysis. Filtering, quality control, and cell type annotation were performed in RStudio using the Seurat package⁸⁸ and custom scripts as previously described.⁸⁹ Cells with transcript counts ≤ 10 or volume < 100 μm³ were filtered out. Then, to account for global differences in transcript counts between coverslips, we normalized data to the total transcripts/cell for each coverslip. Cells with ≤ 10 unique genes were further excluded. X, Y coordinates outlining the wild-type and Cx3cl1^−/− brain sections on each coverslip, the deprived (left) and control (right) hemispheres of the barrel cortex on each brain section, and the cortical layers of each barrel cortex hemisphere were then acquired by manual tracing in the MERSCOPE visualizer software (Vizgen) using Cx3cl1 transcripts, anti-VGluT2 immunolabeling, DAPI signal, Cux2 transcripts, and Rorb transcripts as landmarks. Cells contained within these outlines were identified in RStudio using the point.in.polygon function from the sp package and annotated by their genotype, barrel cortex region, and barrel cortex cortical layer. Data from all samples were then merged into a single Seurat object for clustering and cell type annotation. Data were normalized by dividing gene counts/μm³ for each cell by total count/μm³ for that cell, multiplied by 10,000 and log-transformed. Data were then scaled and principal components were calculated on all 140 measured genes. JackStraw was used to determine the number of significant principal components (p < 0.05), which were then used to perform clustering analysis using Louvain algorithm and calculate UMAP embedding. Clusters were manually annotated based on the spatial distribution of the cells in the tissue and the expression cell type-specific marker genes. A few clusters composed of cells surrounding technical artifacts outside the cortex regions on one of the coverslips were removed from the analysis. In addition, because of imperfections in cell boundary segmentation, a small fraction of cells expressed cell type markers for multiple cell types. Clusters composed of these “hybrid” cells were also removed from the analysis. Cell clusters were recalculated after removal of artifacts and “hybrids” and then given their final cell type annotation. Analysis plots were generated using Seurat 5.0.3 and scCustomize 3.0.1⁹⁰ R packages and R 4.2.2.

In situ hybridization

Mice were transcardially perfused with PBS and 4% PFA, then brains were extracted, post-fixed in 4% PFA, and cryo-embedded in a 2:1 mixture of OCT : 30% sucrose as described above (see Immunohistochemistry). 10 μm thick coronal sections were cut on a cryostat and mounted on glass slides. In situ hybridization was performed using the RNAscope^® Multiplex Fluorescent Reagent Kit v2 with TSA Vivid Dyes (323270 Advanced Cell Diagnostics; Newark, CA) following manufacturer’s instructions using probes Mm-Wls-C2 (405011-C2 Advanced Cell Diagnostics) and Mm-P2ry12-C3 (317601-C3 Advanced Cell Diagnostics). On each slide, one section was hybridized with the RNAscope 3-plex Negative Control Probe (320871 Advanced Cell Diagnostics). Briefly, slides were washed with PBS for 5 minutes, baked for 30 minutes at 60°C, post-fixed in 4% PFA for 15 minutes at 4°C, and dehydrated in ethanol. Slides were then incubated in RNAscope hydrogen peroxide for 10 minutes at room temperature, washed in deionized H₂O, steamed in 99°C RNAscope Target Retrieval solution for 15 minutes, washed in deionized H₂O and 100% ethanol, and dried at 60°C for 5 minutes. Individual brain sections were outlined on slides using a hydrophobic pen and dried overnight at room temperature. Slides were then treated with RNAscope Protease III for 30 minutes at 40°C, washed with deionized H₂O, incubated with probes for 2 hours at 40°C, and washed in RNAscope wash buffer. Sections were then hybridized with RNAscope Multiplex FL v2 AMP1, AMP2, and AMP3, labeled with TSA Vivid 570 (1:1500) for C2 and TSA Vivid 650 (1:1500) for C3, and mounted with Fluoroshield mounting media containing DAPI (F6057 Sigma-Aldrich).

Single plane 40x confocal images were acquired using a Zeiss Observer Spinning Disk Confocal microscope equipped with diode lasers (405nm, 488nm, 594nm, 647nm) and ZEN acquisition software (Zeiss). Images were taken of 6–10 FOV per animal from sections labeled with Wls and P2ry12 RNAscope probes and 2 FOV from sections labeled with negative control probes. Data analysis was performed by experimenters blind to the identity of the samples. Using FIJI software, images were background subtracted for each animal such that < 8 dots per FOV were visible in the negative control images. Images were then manually thresholded using P2ry12 signal to identify microglia. The total P2ry12⁺ area and Wls⁺ P2ry12⁺ area were recorded for each image, and the percentage of P2ry12⁺ area containing Wls⁺ puncta was calculated for each animal.

XAV939 injections

For inhibition of canonical Wnt signaling, mice were injected once daily i.p. with 7.5 mg/kg XAV939 (X3004 Sigma-Aldrich) dissolved in 10% DMSO at ages P3, P4, P5, P6, P7, P8, and P9. Vehicle-injected mice were injected with 10% DMSO.

Intracerebroventricular (icv) injections

To label astrocytes with smV5, P0-P1 mice were anesthetized and injected with 2 μl injections in each hemisphere of PHP.EB::GfaABC1D-lck-smV5–4×6T adeno-associated virus (AAV) at a titer of 6.37 × 10¹³ viral genomes/mL using a hamilton syringe with a 33 gauge needle. Pups were placed on heat pads for recovery.

Quantification and Statistical Analysis

Statistical analysis of TRAP-Seq data were performed using Ingenuity Pathway Analysis software (QIAGEN), GSEA software (v4.3.3), and in R using the DESeq2 (v1.36.0) package. Sholl analysis data were analyzed using a Linear Mixed Model ANOVA with Tukey’s post-hoc, in R, using custom code (https://github.com/SchaferLabUMassChan/Faust_2025). All other statistical analyses were performed using GraphPad Prism 10.5.0. The exact number of samples (n), what n represents, the statistical tests used, and p values for each experiment are noted in the figure legends. For pairwise comparisons between control and deprived hemispheres, we used the ratio paired t test. For comparisons between two unpaired groups, we used Student’s t-test. For 2-way repeated measures comparisons, we used 2-way repeated measures ANOVA followed by Holm-Sidak post-hoc test.

Supplementary Material

1

Table S1. Astrocyte TRAP-Seq: Differential Gene Expression Analysis. Related to Figure 3.

2

Table S2. Astrocyte TRAP-Seq: Gene Set Enrichment Analysis. Related to Figure 3.

3

Table S3. Astrocyte TRAP-Seq: Ingenuity Pathway Analysis. Related to Figure 3.

4

Table S4. NicheNet Analysis of Public Astrocyte TRAP-Seq Datasets. Related to Figure 4.

5

Table S5. Microglia TRAP-Seq: Differential Gene Expression Analysis. Related to Figure 5.

6

Table S6. NicheNet Microglia-Astrocyte Ligand-Receptor Analysis. Related to Figure 5.

7

Table S7. MERFISH probe library. Related to Figure 5.

8

Figure S1. No evidence of sex differences in synapse loss or synapse engulfment after whisker lesioning Related to Figure 1

(A) Experimental design for B-C. Layer IV barrel cortex, 6 days post-lesioning.

(B-C) VGluT2⁺ synapse density (B) and the log₂ fold-change in VGluT2⁺ synapse density (C) in control vs. deprived hemispheres (hems) of male and female mice. Data pooled from control mice in Figures 1A–C, 6I–K, and S6I–L (n=9 male, 7 female mice. ***p<0.001, ****p<0.0001, ns: p>0.05).

(D) Experimental design for E-F. Layer IV barrel cortex, 5 days post-lesioning.

(E-F) VGluT2⁺ synaptic material within microglial lysosomes (E) and the log₂ fold-change in VGluT2⁺ synaptic material within microglial lysosomes (F) in control vs. deprived hems of male and female mice. Data pooled from control mice in Figures 1D–F and 6F–H (n=4 male, 4 female mice. *p<0.05, ns: p>0.05).

Data represent mean ± SEM. Circles (female) and squares (male) represent individual animals. Statistics: Repeated measures 2-way ANOVA with Holm-Sidak post-hoc test (B, E), Student’s t-test (C, F).

9

Figure S2. Expansion microscopy reveals reductions in astrocyte-synapse interactions after whisker lesioning in both males and females Related to Figure 2

(A) Experimental design for B-J. Layer IV barrel cortex, 4 days post-lesioning.

(B) Representative images of endogenous mTomato signal in Rosa26^mTmG brain sections untreated vs. treated with liberate antibody binding solution for 1 hour or 22 hours. Scale bars 20 μm.

(C) Representative image of barrel cortex of tamoxifen-injected Aldh1l1^CreER/+_; Rosa26^mTmG/+ mice showing anti-GFP⁺ astrocytes and anti-VGluT2⁺ synapses. Scale bar 20 μm.

(D) Diagram of expansion microscopy protocol.

(E) Representative images showing anti-GFP⁺ landmarks (crosses) and anti-VGluT2⁺ landmarks (asterisks) within boxed region of Aldh1l1^CreER/+_; Rosa26^mTmG/+ barrel cortex tissue pre-expansion vs. post-expansion. Scale bars 100 μm (not adjusted by expansion index).

(F) Gel expansion factor in control vs. deprived hemispheres (hems) (n=5 mice).

(G) Mean synapse-astrocyte nearest neighbor distance (NND), measured by expansion microscopy in control vs. deprived hems (n=5 mice. *p<0.05).

(H) Diagram of astrocyte ensheathment of synapses.

(I) Cumulative proportion plot of synapses’ ensheathment by astrocytes in control vs. deprived hems. Data pooled from n=5 mice.

(J) Mean synaptic ensheathment by astrocytes, measured by expansion microscopy in control vs. deprived hems, 4 days post-lesioning (n=5 mice. *p<0.05).

(K) Experimental design for L-Q. Layer IV barrel cortex, 4 days post-lesioning.

(L-Q) Mean synapse-astrocyte NND (L), log₂ fold-change in synapse-astrocyte NND (M), percentage of synapses contacted by astrocytes (N), log₂ fold-change in percentage of synapses contacted by astrocytes (O), mean synaptic ensheathment by astrocytes (P), and log₂ fold-change in synaptic ensheathment by astrocytes (Q) in control vs. deprived hems of male and female mice. Data pooled from control mice in Figures 2A–H, S2A–J, and 6A–E (n=5 male, 3 female mice. *p<0.05, ns: p>0.05).

Data in F-G, J, L-Q represent mean ± SEM. Circles (female) and squares (male) represent individual animals.

Statistics: Ratio paired t test (F-G, J), Repeated measures 2-way ANOVA with Holm-Sidak post-hoc test (L, N, P), Student’s t-test (M, O, Q).

10

Figure S3. The total density of astrocyte processes is reduced after whisker lesioning Related to Figure 2

(A) Experimental design for C-F. Layer IV barrel cortex, 4 days post-lesioning.

(B) Diagram of the “total density of astrocyte processes” measured using membrane-bound GFP (mGFP) labeling.

(C) Representative images of anti-GFP⁺ astrocyte processes in control vs. deprived hemispheres (hems). Insets: representative 3D renders of anti-GFP⁺ astrocyte processes. Scale bars 5 μm (corrected for expansion index) and inset scale bars 2 μm (corrected for expansion index).

(D) Total density of astrocyte processes in control vs. deprived hems (n=5 mice. *p<0.05).

(E) Representative images of anti-SOX9⁺ astrocytes and anti-VGluT2⁺ synapses in control vs. deprived hems. Scale bars and inset scale bars 50 μm.

(F) Density of astrocytes in control vs. deprived hems (n=4 mice. ns: p>0.05).

(G) Experimental design for H-I. Layer IV barrel cortex, 4 days post-lesioning.

(H-I) Total density (H) and log₂ fold-change in total density (I) of astrocyte processes in control vs. deprived hems of male and female mice. Data pooled from control mice in Figures 2A–H and 6A–E (n=5 male, 3 female mice. *p<0.05, ns: p>0.05).

(J) Experimental design for K-M. Layer IV barrel cortex, 6 days post-lesioning.

(K) Representative 3D renders of anti-VGluT2⁺ synapses in control vs. deprived hems, pseudo-colored by synapse-astrocyte nearest neighbor distance (NND; corrected for expansion index). Astrocyte-contacted synapses (NND=0) are transparent. Insets: 3D renders of anti-GFP⁺ astrocyte processes. Scale bars and inset scale bars 2 μm (corrected for expansion index).

(L-M) Total density of astrocyte processes (L) and percentage of synapses contacted by astrocytes (M) in control vs. deprived hems at 4 days post-lesioning (data from Figures S3D and 2H) and 6 days post-lesioning (n=5 P8, 4 P10 mice. *p<0.05)

Data represent mean ± SEM. Circles (female) and squares (male) represent individual animals. Statistics: Ratio paired t test (D, F), Repeated measures 2-way ANOVA with Holm-Sidak post-hoc test (H, L-M)

11

Figure S4. Whisker lesioning induces transcriptional changes in astrocytes in cytoskeletal genes and transcriptional regulation genes Related to Figure 3

(A) Experimental design for B-C. Barrel cortex, 24 hours post-lesioning.

(B) Heatmap of astrocyte TRAP-Seq data. Dashed line indicates “TRAP-enriched” cutoff (p<0.05, |fold change|>1.2, expression>5).

(C) EnrichmentMap summary of gene sets enriched in control vs. deprived hemisphere (hem) astrocytes, 24 hours post-lesioning. Manual labels (bold) given to groups of gene sets with similar functions and overlapping genes (grey lines).

12

Figure S5. Analysis of Wnt ligands expressed by microglia Related to Figure 5

(A) Experimental design for B. Barrel cortex, 24 hours post-lesioning.

(B) Heatmap of microglia TRAP-Seq data. Dashed line indicates “TRAP-enriched” cutoff (p<0.05, |fold change|>1.2, expression>5).

(C) Experimental design for D-G. 24 hours post-lesioning.

(D) Uniform manifold approximation and projection (UMAP) of MERFISH-identified cells, colored by cell type. Abbreviations: OPC: oligodendrocyte precursor cell; OL: oligodendrocyte; EpC: ependymal cell; ChP: choroid plexus; BAM: border-associated macrophage.

(E) Dot plot of cell-type specific marker gene expression in cell types from D.

(F) MERFISH coverslips (1 WT, 1 KO on each), color-coded by cell types from D.

(G) UMAP of MERFISH-identified cells, colored by coverslip.

(H) Experimental design for I.

(I) Representative images of GFP⁺ primary astrocytes exposed to 50 ng/mL of WNT4, WNT5A, or WNT7A. Scale bars 50 μm.

13

Figure S6. Inhibition of microglia-astrocyte Wnt signaling prevents synapse loss and reductions in the total density of astrocyte processes after whisker lesioning Related to Figure 6

(A) (Left) experimental design for A-B. Barrel cortex, P5. (Right) representative images of Wls and P2ry12 (microglia marker; outlined) in situ hybridization and DAPI⁺ nuclei in Cx3cr1^Cre; Wls^+/+ (control) vs. Cx3cr1^Cre; Wls^Flox/Flox (WLS cKO) mice, P5. Scale bars 10 μm.

(B) Percentage of P2ry12⁺ area overlapping with Wls⁺ signal in control vs. WLS cKO mice, P5 (n=3 control, 3 WLS cKO mice. **p<0.01).

(C) Experimental design for D-E. P8.

(D-E) Representative images showing anti-VGluT2 synapses and anti-V5 immunolabeling of PHP.EB::GfaABC1D-lck-smV5–4^×6T adeno-associated virus. Boxed region in (D) shown in (E). Scale bars (D) 500 μm and (E) 100 μm.

(F) Experimental design for G-H. Layer IV barrel cortex, 4 days post-lesioning.

(G) Representative images of anti-V5⁺ astrocyte processes in control vs. deprived hemispheres (hem) of control and WLS cKO mice. Scale bars 5 μm (corrected for expansion index).

(H) Total density of astrocyte processes in control vs. deprived hems of control and WLS cKO mice (n=4 control, 4 WLS cKO mice. **p<0.01).

(I) Diagram of Wnt signaling pathway inhibitor XAV939.

(J) Experimental design for K-L. Layer IV barrel cortex, 6 days post-lesioning.

(K) Representative images of anti-VGluT2⁺ synapses in control vs. deprived hems of vehicle-injected and XAV939-injected mice. Scale bars 10 μm.

(L) VGluT2⁺ synapse density in control vs. deprived hems of vehicle-injected and XAV939-injected mice (n=5 vehicle, 8 XAV939 mice. ****p<0.0001).

Data represent mean ± SEM. Circles (female) and squares (male) represent individual animals.

Statistics: Student’s t test (B), Repeated measures 2-way ANOVA with Holm-Sidak post-hoc test (H, L)

14

Figure S7. Expression of Wnt signaling genes is not affected in microglia or astrocytes in CX3CL1 knockout mice Related to Figure 7

(A) Experimental design for B-D. 24 hours post-lesioning.

(B) Uniform manifold approximation and projection of MERFISH-identified cells, colored by genotype.

(C-D) Dot plots in wild-type (WT) vs. Cx3cl1^−/− (KO) mice for (C) gene expression in microglia of Wnt ligands and Wnt release machinery, (D) gene expression in astrocytes of Wnt receptors and Wnt signaling machinery. Data pooled from 2 males, 2 females.

(E) Experimental design for F-H. Layer IV barrel cortex, 4 days post-lesioning.

(F) Representative images of anti-GFP⁺ astrocyte processes in control vs. deprived hemispheres of Aldh1l1^CreER/+_; Rosa26^mTmG/+; Cx3cl1^−/− mice. Scale bars 5 μm (corrected for expansion index).

(G) Total density of astrocyte processes in control vs. deprived hems of Aldh1l1^CreER/+_; Rosa26^mTmG/+; Cx3cl1^−/− mice (n=4 mice. ns: p>0.05).

(H) Mean synaptic ensheathment by astrocytes in control vs. deprived hems of Aldh1l1^CreER/+_; Rosa26^mTmG/+; Cx3cl1^−/− mice (n=4 mice. ns: p>0.05).

Bar graphs represent mean ± SEM. Circles (female) and squares (male) represent individual animals.

Statistics: Ratio paired t test (G-H)

Key resources table.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Guinea pig anti-VGluT2	Synaptic Systems	Cat# 135 404, RRID:AB_887884
Chicken anti-GFP	Abcam	Cat# ab13970, RRID:AB_300798
Rat anti-CD68	Bio-Rad	Cat# MCA1957, RRID:AB_322219
Rat anti-LAMP2	Abcam	Cat# ab13524, RRID:AB_2134736
Rabbit anti-SOX9	Abcam	Cat# ab185966, RRID:AB_2728660
Mouse anti-β-catenin, Alexa Fluor 647	Santa Cruz Biotechnology	Cat# sc-7963 AF647, RRID:AB_626807
Chicken anti-IBA1	Synaptic Systems	Cat# 234 009, RRID:AB_2891282
Rat anti-V5	Absolute Antibody	Cat# Ab00136-6.1, RRID:AB_3065207
Rabbit anti-ALDH1L1	Cell Signaling Technology	Cat# 85828, RRID:AB_3065208
Mouse anti-GFP (19F7)	Heintz Lab; Rockefeller University	Cat# Htz-GFP-19F7, RRID:AB_2716736
Mouse anti-GFP (19C8)	Heintz Lab; Rockefeller University	Cat# Htz-GFP-19C8, RRID:AB_2716737
Goat anti-chicken IgY (H+L), Alexa Fluor 488	Thermo Fisher Scientific	Cat# A-11039, RRID:AB_2534096
Goat anti-Guinea Pig IgG (H+L) Highly Cross-Adsorbed, Alexa Fluor 594	Thermo Fisher Scientific	Cat# A-11076, RRID:AB_2534120
Goat anti-Rat IgG (H+L) Cross-Adsorbed, Alexa Fluor 647	Thermo Fisher Scientific	Cat# A-21247, RRID:AB_141778
Goat anti-rabbit IgG (H+L) highly cross-adsorbed, Alexa Fluor 647	Thermo Fisher Scientific	Cat# A-21245, RRID:AB_2535813
Goat anti-Rabbit IgG (H+L) Highly Cross-Adsorbed, Alexa Fluor 488	Thermo Fisher Scientific	Cat# A-11034, RRID:AB_2576217
Goat anti-rat IgG (H+L) Cross-Adsorbed, Alexa Fluor 488	Thermo Fisher Scientific	Cat# A-11006, RRID:AB_2534074
Goat anti-rat IgG (H+L) highly cross-adsorbed, Alexa Fluor 594	Thermo Fisher Scientific	Cat# A-11007, RRID:AB_10561522
Goat anti-guinea pig IgG (H+L) highly cross-adsorbed, Alexa Fluor 488	Thermo Fisher Scientific	Cat# A-11073, RRID:AB_2534117
Donkey anti-guinea pig IgG (H+L) AffiniPure, Alexa Fluor 488	Jackson ImmunoResearch Labs	Cat# 706-545-148, RRID:AB_2340472
Donkey anti-rabbit IgG (H+L) highly cross-adsorbed, Alexa Fluor 594	Thermo Fisher Scientific	Cat# A-21207, RRID:AB_141637
Goat anti-Rabbit IgG, Atto 647N	Sigma Aldrich	Cat# 40839, RRID:AB_1137669
Donkey anti-rat IgG (H+L) highly cross-adsorbed, Alexa Fluor 594	Thermo Fisher Scientific	Cat# A-21209, RRID:AB_2535795
Goat anti-mouse IgG + IgM	Jackson ImmunoResearch Labs	Cat# 115-005-044, RRID:AB_2338451
Goat anti-rat IgG + IgM	Jackson ImmunoResearch Labs	Cat# 112-005-044, RRID:AB_2338094
Mouse anti-rat L1	Developmental Studies Hybridoma Bank, Univ. Iowa	Cat# ascs4, RRID:AB_528349
Chicken anti-GFP	Aves Labs	Cat# GFP-1020, RRID:AB_10000240
Rabbit anit-VGLUT2, carrier free	Abcam	Cat# ab227906, RRID: AB_3697231
Oligo-conjugated secondary anti-rabbit antibody	Vizgen	Cat# 20300102
Bacterial and virus strains
PHP.EB::GfaABC1 D-lck-smV5-4x6T	Dr. S. Thomas Carmichael, University of California, Los Angeles (Gleichman et al.43)	N/A
Chemicals, peptides, and recombinant proteins
Tamoxifen	Sigma Aldrich	Cat# C8267
Papain	Worthington Biochemical	Cat# LK003178
DNAse	Worthington Biochemical	Cat# LS002007
Ovomucoid	Worthington Biochemical	Cat# LS003086
Poly-D-lysine	Sigma Aldrich	Cat# P6407
DPBS	Gibco	Cat# 14287
Bovine serum albumin	Sigma Aldrich	Cat# A8806
Insulin	Sigma Aldrich	Cat# I1882
Bandeiraea simplicifolia Lectin I	Vector Laboratories	Cat# L-1100-5
Neurobasal	Thermo Fisher Scientific	Cat# 21103049
B27 supplement	Thermo Fisher Scientific	Cat# 12587010
L-glutamine	Thermo Fisher Scientific	Cat# 25-030-081
Pen/Strep	Thermo Fisher Scientific	Cat# 15140122
Sodium Pyruvate	Thermo Fisher Scientific	Cat# 11360070
Forskolin	Sigma Aldrich	Cat# F6886
BDNF	Pepro-tech	Cat# 450-02
CNTF	Pepro-tech	Cat# 450-13
Laminin	R&D Systems	Cat# 3400-010-02
Neurobasal Plus	Thermo Fisher Scientific	Cat# A3582901
B27 Plus	Thermo Fisher Scientific	Cat# A3653401
AraC	Sigma Aldrich	Cat# C1768
DMEM	Gibco	Cat# 11960
Fetal bovine serum	Sigma Aldrich	Cat# F2442
Hydrocortisone	Sigma Aldrich	Cat# H0888
N-acetyl-L-cysteine	Sigma Aldrich	Cat# A8199
Trypsin-EDTA	Thermo Fisher Scientific	Cat# 25300054
Lipofectamine LTX with PLUS reagent	Thermo Fisher Scientific	Cat# 15338100
XAV939	Sigma Aldrich	Cat# X3004
Murine RNAse inhibitor	New England Biolabs	Cat# M0314L
Recombinant Full Length Human WNT4 Protein, N-Fc-Flag-tagged	Creative Biomart	Cat# WNT4-517HFL
Active Recombinant Mouse Wnt5a	Creative Biomart	Cat# Wnt5a-2054M
Active Recombinant Human WNT7A Protein	Creative Biomart	Cat# WNT7A-42H
MERFISH blocking buffer	Vizgen	Cat# 20300100
MERFISH DAPI and PolyT staining reagent	Vizgen	Cat# 2030021
Critical commercial assays
RNAscope® Multiplex Fluorescent Reagent Kit v2 with TSA Vivid Dyes	Advanced Cell Diagnostics	Cat# 323270
MERSCOPE 140-gene imaging kit	Vizgen	Cat# 10400004
Deposited data
Raw and analyzed TRAP-Seq data	This paper	GEO: GSE252628
Processed MERFISH data	This paper	GEO: GSE298314
Astrocyte TRAP-Seq: day vs. night; sleep vs. wake	Bellesi et al.28	GEO: GSE69079
Astrocyte TRAP-Seq: aging (2 years vs. 10 weeks)	Clarke et al.29	BioProject: PRJNA417856
Astrocyte TRAP-Seq: seizure (pentylenetetrazol vs. saline)	Sapkota et al.30	GEO: GSE147830
Astrocyte TRAP-Seq: APP/PS1 and MAPT-P301S mice	Jiwaji et al.31	Array Express: E-MTAB-10985
Astrocyte TRAP-Seq: SOD1G37R mice	Sun et al.32	GEO: GSE74724
Experimental models: Organisms/strains
Mouse: C57B/6J	The Jackson Laboratory	RRID:IMSR_JAX:00 0664
Mouse: B6.129(Cg)-Gt(ROSA)26Sortm4(ACTB-tdTomato-EGFP)Luo/J	The Jackson Laboratory	RRID:IMSR_JAX:00 7676
Mouse: B6.129P2(Cg)-Cx3cr1tm1Litt/J	The Jackson Laboratory	RRID:IMSR_JAX:00 5582
Mouse: STOCK Tg(TCF/Lef1-HIST1H2BB/EGFP)61 Hadj/J	The Jackson Laboratory	RRID:IMSR_JAX:01 3752
Mouse: B6N.FVB-Tg(Aldh1l1-cre/ERT2)1Khakh/J	The Jackson Laboratory	RRID:IMSR_JAX:03 1008
Mouse: B6.Cg-Krastm4Tyj Apctm1Tno Tg(CDX2-cre/ERT2)752Erf/MaraJ)	The Jackson Laboratory	RRID:IMSR_JAX:03 5169
Mouse: Cx3cl1tm1Lira	Dr. Sergio Lira, Icahn School of Medicine, Mount Sinai (Cook et al.67)	N/A
Mouse: 129S-Wlstm1.1Lan/J	Dr. David Rowitch, University of California, San Francisco	RRID:IMSR_JAX:01 2888
Mouse: STOCK Tg(Aldh1l1-EGFP)OFC789Gsat/Mmucd	Dr. Ben Barres, Stanford University	RRID:MMRRC_0110 15-UCD
Mouse: STOCK Tg(Cx3cr1-cre)MW126Gsat/Mmcd	Dr. Staci Bilbo, Harvard University	RRID:MMRRC_0363 95-UCD
Mouse: B6.129P2(Cg)-Cx3cr1tm2.1(cre/ERT2)Litt/WganJ	Dr. Dan Littman, New York University	RRID:IMSR_JAX:02 1160
Mouse: B6.Cg-Eef1a1tm1Rck/J	Dr. Jeff Friedman, Rockefeller University and Dr. Ana Domingos, Instituto Gulbenkian de Ciência	RRID:IMSR_JAX:03 0305
Mouse: B6;FVB-Tg(Aldh1l1 -EGFP/Rpl10a)JD130Htz/J	Dr. Nathaniel Heintz, Rockefeller University	RRID:IMSR_JAX:03 0247
Rat: Crl:CD(SD)	Charles River Laboratories	RRID:RGD_734476
Oligonucleotides
RNAscope Probe Mm-Wls-C2	Advanced Cell Diagnostics	Cat# 405011-C2
RNAscope Probe Mm-P2ry12-C3	Advanced Cell Diagnostics	Cat# 317601-C3
RNAscope 3-plex Negative Control Probe	Advanced Cell Diagnostics	Cat# 320871
Custom 140-plex MERFISH library (See Table S7)	This Paper	N/A
Recombinant DNA
pMAX-GFP	Lonza	N/A
Software and algorithms
Prism 10.5	GraphPad	https://www.graphpad.com/
FIJI	Schindelin et al.71	http://fiji.sc
Ingenuity Pathway Analysis	Qiagen	RRID:SCR_008653
Gene Set Enrichment Analysis 4.3.3	Broad Institute	RRID:SCR_003199
ZEN black/blue	Zeiss	RRID:SCR_013672
Imaris 10.2	Oxford Instruments	RRID:SCR_007370
Other
In vitro Sholl Analysis	Tan et al.86	https://github.com/Eroglu-Lab/In-Vitro-Sholl
MERFISH and TRAP-Seq Analysis	This paper	https://github.com/SchaferLabUMassChan/Faust_2025

Highlights.

• Astrocyte processes reduce contact with synapses before synapse removal by microglia

• Microglia-derived Wnts drive astrocytes to reduce their association with synapses

• Wnt signaling between microglia and astrocytes is critical for synapse remodeling

• Microglia-astrocyte Wnt crosstalk is activated downstream of CX3CL1/CX3CR1 signaling