Astrocyte morphogenesis requires self-recognition

Abstract

Self-recognition is a fundamental cellular process across evolution and forms the basis of neuronal self-avoidance^1-4. Clustered protocadherin (cPcdh) proteins, which comprise a large family of isoform-specific homophilic recognition molecules, have a pivotal role in the neuronal self-avoidance that is required for mammalian brain development^5-7. The probabilistic expression of different cPcdh isoforms confers unique identities on neurons and forms the basis for neuronal processes to discriminate between self and non-self^5,6,8. Whether this self-recognition mechanism also exists in astrocytes remains unknown. Here we report that γC3, a specific isoform in the Pcdhγ family, is enriched in human and mouse astrocytes. Using genetic manipulation, we demonstrate that γC3 acts autonomously to regulate astrocyte morphogenesis in the mouse visual cortex. To determine whether γC3 proteins act by promoting recognition between processes of the same astrocyte, we generated pairs of γC3 chimeric proteins that are capable of heterophilic binding to each other, but incapable of homophilic binding. Co-expression of complementary heterophilic binding isoform pairs in the same γC3-null astrocyte restored normal morphology. By contrast, chimeric γC3 proteins individually expressed in single γC3-null mutant astrocytes did not. These data establish that self-recognition mediated by γC3 contributes to astrocyte development in the mammalian brain.

Extended Data Fig.1:. Efficient labeling of cortical astrocytes using AAV.

a, Sparse labeling of astrocytes across all cortical layers (P14 mouse brain). AAV.PhP.eB expressing Lck-GFP, under the control of an astrocyte-specific promoter (GfaABC₁D), which were retro-orbitally injected into neonates at P1. b, Astrocytes were labeled from P8 to P21 in the visual cortex, and tissues were subsequently stained with GFP antibodies. Representative images were obtained from three mice and have been routinely observed in injections with similar results.

Extended Data Fig.2:. Identification of astrocyte γC3 gene expression in V1.

a, Low-magnification image illustrating γC3 expression in astrocytes across both upper (L2/3) and lower (L5/6) cortical layers. γC3 expression is also expressed in other cell types in these regions. Scale bar, 40 μm. Representative images are from experiments quantified in (b), which were repeated independently in three mice. b, Astrocyte cell areas were segmented based on the expression of the astrocyte-specific marker Slc1a3, detected using an RNAscope in situ hybridization probe. γC3 expression was detected using an isoform-specific RNA probe. Solid outlines indicate the cell boundaries of identified single astrocytes. γC3 RNA puncta per cell were quantified within Slc1a3-positive regions in layers 2/3 and 5/6 of V1 in WT mice at P21. N = 280 cells from three mice. Statistical comparisons were performed using the two-sided Wilcoxon rank-sum test, with nested analysis treating the animal as the unit of analysis. Error bars represent the standard error of the mean (s.e.m.). The probe specificity to γC3 is supported by two observations. First, each of the multiple probes (referred to as Z probes) in the probe set is assessed computationally for cross-reactivity with non-cognate sequences in the transcriptome. Second, signal amplication requires adjacent Z probes to bind simultaneously, which contributes additional stringency in signal detection. Scale bar, 10 μm.

Extended Data Fig.3:. Evaluating astrocyte morphology in the hippocampus CA1 using multiple morphological metrics.

a, To assess the morphological characteristics of astrocytes, we retro-orbitally injected WT or γC3 KO mice with AAVs expressing Lck-smV5, Lck-smMyc, and Lck-GFP under the control of the astrocyte-specific GfaABC1D promoter. The mice were harvested at P21 and subjected to immunostaining with anti-Myc, anti-V5, and anti-GFP antibodies. Roundness: Measures how closely the shape’s minor and major axes resemble a perfect circle. Circularity: Quantifies how similar the object’s area and perimeter are to a perfect circle. b, Representative images of single astrocytes, flattened in a confocal volume, obtained from the CA1 hippocampus of both WT and γC3 KO mice. Representative images were obtained from three mice. c, The results are summarized in plots representing various morphological parameters. Two-sided unpaired t-tests with Welch’s correction was used to compare WT and γC3KO groups. Apparent cell volume: WT, n=19 astrocytes from three mice; γC3 KO, n=19 astrocytes from three mice. Feret max, Feret min, aspect ratio, territory size, roundness, and circularity: WT, n=30 astrocytes from three mice; γC3 KO, n=24 astrocytes from three mice. Error bars, s.e.m. Scale bars, 10 μm. **p < 0.01; ***p < 0.001. Nested analysis was performed for all statistical comparisons to confirm the results, and details are provided in Supplementary Table 4.

Extended Data Fig.4:. Brain-wide multicolor labeling of astrocyte morphology.

a, To enhance the labeling of fine astrocytic processes, smFPs were targeted to the plasma membrane using the Lck domain and packaged into AAV.PhP.eb serotype, which was delivered via the retroorbital route into P1 mice. Brains were harvested at P21. b, Multicolor labeling of astrocytes in the brain. WT mice were injected with AAV expressing Lck-smV5, Lck-smMyc, and Lck-GFP driven by the astrocyte-specific GfaABC1D promoter. Stochastic multicolor labeling of astrocytes is observed throughout the brain, including the hippocampus, thalamus, and visual cortex. The right panel shows a 3D reconstruction of neighboring astrocytes in layer 6 of V1. Scale bars 1mm (hippocampus), 40μm (thalamus), 10μm (V1). c, Astrocyte volumes were computed through surface reconstruction. The voxels inside each surface that overlap with each other were calculated and highlighted in yellow, generating a new surface from the overlapping regions. d,e, Astrocyte tiling index was calculated by dividing the overlapping volume between adjacent astrocytes by the volumes of a single astrocyte. Both WT and γC3 KO astrocytes exhibited minimal overlap with adjacent astrocytes. Two-sided unpaired t-tests with Welch’s correction was used to compare WT and γC3KO groups. WT, n=44 astrocytes from three mice; γC3KO, n=21 astrocytes from three mice. Error bars, s.e.m. Scale bars 10 μm. ****p<0.0001. Nested analysis was performed for all statistical comparisons to confirm the results, and details are provided in Supplementary Table 4.

Extended Data Fig.5:. Astrocyte-specific Cre induction and gene recombination in Aldh1l1-Cre/ERT2 mice.

a, Schematic illustrating the crossing of Aldh1l1-Cre/ERT2 mice with ROSA26-LSL-Cas9-P2A-eGFP mice. Cre recombinase expression was induced by tamoxifen injection from P1 to P3, and V1 tissues were harvested at P21. b, Astrocytes were immunostained with an anti-Kir4.1 antibody, neurons with an anti-NeuN antibody, and GFP with an anti-GFP antibody. GFP co-localized with Kir4.1 staining, indicating astrocyte-specific Cre-mediated gene recombination, with no detectable GFP expression in neurons. Representative images were obtained from three mice. Scale bars: 10 μm.

Extended Data Fig.6:. Validation of astrocyte-specific Pcdhγ KO and rescue of γC3 in cortical astrocytes.

a, Diagrams of conditional Pcdhγ KO and Cre-inducible γC3 alleles. Each Pcdhγ protein is encoded by an mRNA comprising one of 22 variable exons (yellow) and the 3 constant “C” exons (blue). In the Pcdhg^fcon3 KO allele, loxP sites flank the final constant exon, which is fused with GFP at the carboxy-terminus. Cre recombination results in loss of GFP-tagged Pcdhγ proteins. In the ROSA26-CAG::lox-Stop-lox-γC3-mCherry Cre-inducible mice, Cre-mediated excision of the stop codon leads to the expression of γC3 with mCherry fused to the carboxy-terminus. In animals carrying both alleles and astrocyte-specific Cre, GFP is lost. As mCherry sequences are incorporated into the 3’-end of the γC3 mRNA, in the absence of Cre, the “Stop” cassette leads to transcription termination. Thus, mCherry containing transcripts are only seen upon excision of the “Stop” cassette. b, c, In situ detection of Pcdhγ expression in visual cortex from whole-mount expanded tissues by EASI-FISH. (b) Low-magnification view, (c) High-magnification view of single optical sections from whole-mount preparations of the visual cortex from the indicated genotypes (see Methods). Upper panel: Control, Pcdhγ/Pcdhγ (GFP+). Middle panel: Pcdhγ-KO, Aldh1l1-Cre/ERT2; Pcdhγ/Pcdhγ (GFP-). Lower panel: Pcdhγ-KO; γC3, Aldh1l1-Cre/ERT2; Pcdhγ/Pcdhγ; γC3 (GFP- and mCherry+). Astrocytes were labeled with Slc1a3 probes. d, Quantification of Cre-mediated recombination in astrocytes. RNA in situ hybridization confirmed efficient deletion of the Pcdhγ genes in conditional KO mice. In control mice, GFP-tagged RNA from Pcdhγ locus is expressed. In Pcdhγ-KO mice, the GFP-tagged RNA is removed from the Pcdhγ locus. In Pcdhγ-KO; γC3 mice, the GFP-tagged RNA is removed, and the γC3 RNA transcript expressed from the ROSA26 locus is tagged with mCherry sequence. Note control and Pcdhγ-KO do not contain the ROSA26-CAG::lox-Stop-lox-γC3-mCherry construct. Thus, the sparse mCherry puncta is non-specific hybridization. The Kruskal-Wallis test was used to compare the number of RNA puncta among the groups, followed by Dunn’s multiple comparison test for post-hoc pairwise comparisons. Specific p-values are provided in Supplementary Table 4. Control: n=61 astrocytes from two mice; Pcdhγ-KO, n=58 astrocytes from three mice; Pcdhγ-KO; γC3, n=58 astrocytes from three mice. Error bars, s.e.m. Scale bars, 10 μm. ****p < 0.0001.

Extended Data Fig.7:. Expression of AAVs expressing γC3FL and γC3 homophilic binding mutants in vivo.

AAV.PhP.eB was used to express Lck-smMyc, γC3 full-length (γC3FL), γC3-L87E, and γC3-L342E mutants under an astrocyte-specific GfaABC₁D promoter in γC3 KO mice. AAV.PhP.eB expressing γC3FL, γC3-L87E, and γC3-L342E, tagged with a C-terminal 3xV5, were retro-orbitally delivered into P1 mice. Astrocyte morphology was labeled by co-injecting AAV.GfaABC1D expressing Lck-smMyc and visualized with an anti-Myc antibody (green). Expression of γC3FL, γC3-L87E, and γC3-L342E was detected using an anti-V5 antibody (red). a, Expression of γC3FL. Representative images were obtained from five mice. b, Expression of γC3-L87E. Representative images were obtained from five mice. c, Expression of γC3-L342E. Representative images were obtained from five mice. Scale Bars 10 μm.

Extended Data Fig.8:. Design of heterophilic protocadherin chimera pairs that lost homophilic binding.

a, Contact between cell membranes (grey) of astrocyte sister branches (WT). γC3 molecules (blue) forming a trans-dimer (in curly brackets) are shown schematically with extracellular cadherin (EC) domains as ellipses. b,c, Homophilic-deficient cPcdh chimeras. d, A pair of chimeras (one from b and another from c) co-expressed in the same astrocyte. e, A pair of chimeras in D modified with mutations (asterisk, cyan) that enable heterophilic binding. f, Summary of AUC experiments of WT γC3, γC4, and γC5, parts of which were used for chimera design. Of note, AUC on γC5 was done in the context of the EC1-EC5 fragment, but only the EC1-EC4 trans-dimer is shown in the schematic. g, Summary of AUC experiments on the designed chimeras. See Methods for details on mutation (cyan) design. A sign with a circle and a line through it depicts inability to form dimers.

Extended Data Fig.9:. Trans-dimer models and their properties at the EC2-EC3 boundary.

a, Relative FoldX energies (in parenthesis in kcal/mol) of the EC1-EC4::EC4’-EC1’ trans-homodimers assuming all complexes form dimers identical in Cα backbone to γC4::γC4. The chimeras are color-coded based on the sequence composition: γC3 (blue), γC4 (green) or γC5 (red). All structures shown in ribbon representation with calcium atoms as green balls. b, Comparison of the amino acid properties at the EC2EC3 boundary of γC4 trans-dimer and γC3. Protein backbone is in ribbon representation. Residues are shown as sticks in the expanded view of the EC2EC3::EC3’EC2’ interface. Residues that differ in properties between γC4 and γC3 in a diamond-shaped area correspond to thicker sticks. Polar residues predicted to destabilize γC3 trans-dimer in the γC4-like orientation are underlined in cyan. c, Schematic representation of the expanded views shown in b for all WT and chimera proteins. H – hydrophobic boundary, PH – polar/hydrophobic non-complementary boundary.

Extended Data Fig.10:. Summary of astrocyte self-recognition and morphogenesis via γC3 and chimeric isoform binding.

Astrocyte morphology relies on self-recognition mediated by γC3. Binding between γC3 proteins is likely to activate intracellular signaling pathways which specify distinct morphological consequences. Chimeras which have lost homophilic binding (e.g. M1, M6, M3, and M8) do not promote normal morphogenesis on their own. By contrast, pairs of complementary chimeras (e.g. M1+M6 and M3+M8) which bind heterophilically promote normal morphogenesis when expressed in the same astrocyte. The precise mechanism by which γC3 regulates morphogenesis is unclear. Binding could activate repulsion^,. The initial repulsive response may direct process extension away from sister branches and this would indirectly promote process outgrowth. Alternatively, transient binding between γC3 on opposing processes may directly promote the assembly of signaling complexes which could promote process outgrowth^,(see Discussion).

Fig. 1:. γC3 is the predominant cPcdh isoform in mouse and human astrocytes.

a, The mouse cPcdh gene cluster contains exons encoding 58 extracellular and transmembrane domains. The α and γ proteins share common exons encoding a terminal segment of the cytoplasmic domain. Each β comprises a distinct C-terminal encoded segment. b, RNA-Seq transcriptional profiling of cPcdh genes from the mouse cerebral cortex in neurons, oligodendrocytes, and microglia at P7 (upper panels) and at three stages of development in astrocytes (lower panels) from the same region. Note the difference in the scale on the Y-axis between the upper and lower panels. These data are from Zhang et al. 2014 and Clarke et al. 2018. c, Expression of cPcdh genes in astrocytes from different regions of the mouse central nervous system from Endo et al., 2022 (upper panel). d. RNAseq data for human astrocytes immunopanned from fetal (18–18.5 weeks of gestation) and adult human brains from Zhang et al. 2016 (lower panel). Heatmaps show the log₂ FPKM values of the cPcdh cluster genes. The data for panel b and d are from https://brainrnaseq.org/. The data for panel c are from http://astrocyternaseq.com/.

Fig. 2:. Astrocyte morphology is disrupted in γC3 KO mice.

a, AAV vectors expressing Lck-GFP (myristoylated GFP localizing to the cytoplasmic face of the plasma membrane), controlled by an astrocyte-specific promoter were retro-orbitally injected into P1 neonates. Astrocyte morphology analyzed using multiple metrics (see Methods). b, Astrocyte morphology in WT and γC3KO mutants. Quantification of astrocyte apparent volumes (see Methods). Astrocytes in L6 showed reduced volumes from P8 to P21 in γC3KO mice. P8: WT (n=48 astrocytes from three mice) and γC3KO (n=75 astrocytes from three mice); P14: WT (n=29 astrocytes from three mice) and γC3KO (n=21 astrocytes from three mice); P21: WT (n=35 astrocytes from six mice) and γC3KO (n=21 astrocytes from six mice). Two-sided unpaired t-test with Welch’s correction was used for comparisons. c, Astrocytes in P21 WT and γC3KO animals. WT: L2/3, n=10; L4, n=23; and L5/6, n=35 astrocytes each from six mice. γC3KO: L2/3 n=56; L4, n=16; and L5/6 n=17 astrocytes each from six mice. Two-sided unpaired t-test with Welch’s correction was used for comparisons across cortical layers (L2/3, L4, and L5/6). d, Astrocyte morphology analyzed with multiple morphological metrics. Representative images of single astrocytes flattened in a confocal volume from both WT and γC3 KO in layer 5/6 V1 mice. The results are summarized in plots representing various morphological parameters. WT, n=24 astrocytes from six mice; γC3 KO, n=21 astrocytes from four mice. For neuropil infiltration volume (NIV, see methods): WT, n=32 cells from three mice; γC3 KO, n=42 cells from three mice. Error bars represent the standard error of the mean (s.e.m.). Two-sided unpaired t-test with Welch’s correction was used. Scale bars, 10 μm. ***p < 0.001, ****p < 0.0001. Nested analysis was performed for all statistical comparisons to confirm the results, and details are provided in Supplementary Table 4.

Fig. 3:. Replacement of the Pcdhγ cluster with a single isoform rescues astrocyte morphology.

a, Strategy to determine if a single isoform alone is sufficient for normal astrocyte morphology (see text). Mice were injected with two AAVs each expressing a different membrane marker (Lck-smV5 or Lck-smMyc) under the astrocyte-specific GfaABC1D promoter to label astrocyte morphology. Using two different labeled viruses enabled delineation of boundaries between adjacent astrocytes. b,c, Astrocytes lacking the entire Pcdhγ complex exhibit marked defects in morphology. The expression of γC3 only in astrocytes (see Methods) substantially rescues astrocyte morphological defects. See Methods for genetic scheme. Morphological differences were analyzed using one-way ANOVA with post-hoc Tukey’s tests for pairwise comparisons. Specific p-values are provided in Supplementary Table 4. Control: n=40 cells from three mice; Pcdhγ-KO, n=50 cells from three mice; Pcdhγ-KO; γC3, n=40 cells from three mice. Error bars, s.e.m. Scale bars, 10 μm. ****p < 0.0001. Nested analysis was performed to confirm the results, and details are provided in Supplementary Table 4.

Fig. 4:. γC3 homophilic recognition specificity is required for astrocyte morphology.

a, Schematic of proteins tested for rescue in WT and γC3 null mice. b, Summary of AUC experiments on the EC1-EC4 WT and mutant proteins. c, Structure-based design of mutations disrupting homophilic binding. Unsatisfied buried charges (red spheres) disrupt homophilic binding. d,e, Rescue experiments using AAV to drive different γC3 constructs under the control of the astrocyte-specific GfaABC₁D promoter in WT and γC3 KO mutants in layer 6 of V1. Morphological differences were analyzed using one-way ANOVA with post-hoc Tukey’s tests for pairwise comparisons. WT, n=32 cells from three animals; γC3 KO-control, n=45 cells from five animals; γC3 KO-γC3FL, n=14 cells from five animals; γC3 KO-γC3L87E, n=33 cells from five animals; and γC3 KO-γC3L342E, n=25 cells from five animals. Error bars, s.e.m. Scale bars, 10 μm. ***p < 0.001, ****p < 0.0001. Nested analysis was performed to confirm the results, and details are provided in Supplementary Table 4.

Fig. 5:. Complementary chimeras expressed in astrocytes only, rescue the morphology defect in γC3 null mutant astrocytes.

a, Schematic showing protein constructs used in rescue experiments (see text). b, K_Ds for EC1-EC4 fragments of chimeras shown in A as determined using AUC. c, Astrocytes transduced with virus-expressed chimeras under the control of the astrocyte-specific GfaABC₁D promoter were identified by staining using antibodies to the epitope tags (arrows). Weak signal to noise was often observed with anti-HA antibody. Astrocyte morphology was visualized using anti-Myc staining to visualize AAV.Lck-smMyc. d, Quantification of astrocyte apparent volumes in different genotypes as indicated. Morphological differences were analyzed using one-way ANOVA with post-hoc Tukey’s tests for pairwise comparisons. WT, n=32 cells in three mice; γC3 KO-control, n=45 cells in five mice; M1, n=26 cells in five mice; M6, n=25 cells in five mice; M1::M6, n=18 cells in five mice; M3, n=31 cells in five mice; M8, n=20 cells in five mice; M3::M8: n=16 cells in five mice. Error bars, s.e.m. e, The schematic shows that complementary pairs of chimeras (M1::M6 or M3::M8) enable heterophilic binding within the same astrocytes. Scale bars, 10 μm. ***p < 0.001, ****p < 0.0001. Nested analysis was performed to confirm the results, and details are provided in Supplementary Table 4.