Astrocytes engage unique molecular programs to engulf pruned neuronal debris from distinct subsets of neurons

Abstract

Precise neural circuit assembly is achieved by initial overproduction of neurons and synapses, followed by refinement through elimination of exuberant neurons and synapses. Glial cells are the primary cells responsible for clearing neuronal debris, but the cellular and molecular basis of glial pruning is poorly defined. Here we show that Drosophila larval astrocytes transform into phagocytes through activation of a cell-autonomous, steroid-dependent program at the initiation of metamorphosis and are the primary phagocytic cell type in the pupal neuropil. We examined the developmental elimination of two neuron populations-mushroom body (MB) γ neurons and vCrz⁺ neurons (expressing Corazonin [Crz] neuropeptide in the ventral nerve cord [VNC])-where only neurites are pruned or entire cells are eliminated, respectively. We found that MB γ axons are engulfed by astrocytes using the Draper and Crk/Mbc/dCed-12 signaling pathways in a partially redundant manner. In contrast, while elimination of vCrz⁺ cell bodies requires Draper, elimination of vCrz⁺ neurites is mediated by Crk/Mbc/dCed-12 but not Draper. Intriguingly, we also found that elimination of Draper delayed vCrz⁺ neurite degeneration, suggesting that glia promote neurite destruction through engulfment signaling. This study identifies a novel role for astrocytes in the clearance of synaptic and neuronal debris and for Crk/Mbc/dCed-12 as a new glial pathway mediating pruning and reveals, unexpectedly, that the engulfment signaling pathways engaged by glia depend on whether neuronal debris was generated through cell death or local pruning.

Keywords: Drosophila; apoptosis; astrocyte; phagocytosis; pruning; synapse elimination.

Figure 1.

Astrocytes transform into phagocytes at the initiation of neural circuit remodeling. (A) Astrocytes in controls were labeled with GFP (alrm-Gal4, UAS-mCD8∷GFP; green) in A, C, and D, the neuropil was labeled with HRP-Cy3 (red), and glial nuclei were labeled with anti-Repo (blue). Time points are as indicated (APF). Thoracic and brain regions are shown. (Inset) High-magnification view of the boxed region. Single z-sections are shown for A, C, and D. Bars: 20 μm; insets, 10 μm. (B) TEM image from a cross-section of the abdominal VNC of a control animal. An astrocyte (outlined in yellow) at the edge of the neuropil, exhibiting cytoplasmic vacuoles (magenta). Genotype used was as follows: alrm-Gal4, UAS-mCD8∷GFP/+. Bar, 2 μm. (C) Draper was labeled in red. Time points are as indicated (APF). (Insets) High-magnification view of the boxed regions. Genotype used was as follows: alrm-Gal4, UAS-mCD8∷GFP. Bars: 20 μm; insets, 10 μm. (D) LysoTracker was labeled in red. Note that all LysoTracker⁺ puncta in the neuropil were surrounded by GFP⁺ membranes of astrocytes. (Inset) High-magnification view of vacuole filled with LysoTracker. Genotype used was as follows: alrm-Gal4, UAS-mCD8∷GFP. Bars: 20 μm; inset, 5 μm.

Figure 2.

EcR signaling cell-autonomously regulates transformation of astrocytes into phagocytes. (A) Astrocytes in controls were labeled with GFP (alrm-Gal4, UAS-mCD8∷GFP/+; green) in A–E, and EcR-B1 (blue) and glial nuclei were labeled with anti-Repo (red). (Insets) High-magnification view of an astrocyte cell body. (A–E) All confocal images are single z-confocal slices. Bars: 20 μm; inset, 10 μm. (B) The neuropil was labeled with HRP-Cy3 (red) and, in B and D, glial nuclei were labeled with anti-Repo (blue). Time points are as indicated (APF). Note the lack of transformation of astrocytes at 6 h APF. (B,D,E) Genotypes used were as follows: control (alrm-Gal4,UAS-mCD8∷GFP/+) and alrm>EcR^DN (alrm-Gal4,UAS-mCD8∷GFP/UAS-EcR^DN). Bars, 20 μm. (C) Astrocyte MARCM clones at L3 and 6 h in controls (alrm-Gal4,UAS-mCD8∷GFP, repoflp^6-2;FRT^2A/FRT^2A,tub-Gal80) or astrocytes expressing EcR^DN (alrm-Gal4,UAS-mCD8∷GFP, repoflp^6-2/UAS-EcR^DN;FRT^2A/FRT^2A,tub-Gal80) at 6 h APF. Bars, 20 μm. (D) Draper was labeled in red. Time points are as indicated (APF). Note the strong knockdown of draper in astrocytes in alrm>EcR^DN animals compared with controls at 6 h APF. Bars, 20 μm. (E) LysoTracker was labeled in red. Time points are as indicated (APF). LysoTracker⁺ puncta in astrocytes were blocked at 6 h APF in alrm>EcR^DN animals. Bars, 20 μm.

Figure 3.

Astrocytes engulf and clear synaptic material from the neuropil. (A) In A and D, confocal images are single z-confocal slices. Astrocytes were labeled with GFP (alrm-Gal4, UAS-mCD8∷GFP/+; green), glial nuclei were labeled with anti-Repo (blue), and, in A and D, active zones were labeled with nc82 (antibody for Brp; red). Time points are as follows: L3 (third instar larva) and APF (hours APF). (Top inset) High-magnification view of the boxed region. (Middle and bottom insets) Other examples of vacuoles. Bars: 20 μm; insets, 1 μm. (B) The graph shows fluorescence intensity per pixel in each channel for the inset in A of the astrocyte membrane and nc82⁺ puncta along a line drawn through a vacuole. (C) TEM image from a cross-section of the abdominal VNC of an alrm>EcR^DN animal. An astrocyte (outlined in yellow) at the edge of the neuropil. Time point is as indicated (APF). Genotype used was as follows: alrm>EcR^DN (alrm-Gal4,UAS-mCD8∷GFP/UAS-EcR^DN). Bars, 2 μm. (D) Time points are as indicated (APF). Genotypes used were as follows: control (alrm-Gal4/+) and alrm>EcR^DN (alrm-Gal4/UAS-EcR^DN). Bars, 20 μm. (E,F) Quantification of pixel intensity of nc82⁺ puncta at L3 and 18 h APF in SOG, thorax, and abdominal regions of the VNC. (E) Genotypes used were as follows: EcR^DN/+ (UAS-EcR^DN/+) and alrm>EcR^DN (alrm-Gal4/UAS-EcR^DN). Error bars represent ±SEM. N-values are as follows: alrm-Gal4/+, N ≥ 20; UAS-EcR^DN/+, N ≥ 16; and alrm-Gal4/UAS-EcR^DN, N ≥ 20 (N denotes number of measurements). (*) P < 0.05; (**) P < 0.01; (****) P < 0.0001. (F) Genotypes used were as follows: shi^ts/+ (UAS-shi^ts/+) and alrm>shi^ts (alrm-Gal4/UAS-shi^ts). Flies were raised at 18°C and then shifted to the restrictive temperature (30°C) at 0 h APF for 18 h. Error bars represent ±SEM; N ≥ 12; (****) P < 0.0001.

Figure 4.

Draper and Crk/Mbc/dCed-12 pathways are required for the formation of astrocytic vacuoles. (A) Astrocytes were labeled with GFP (alrm-Gal4, UAS-mCD8∷GFP/+; green). Images zoomed at the abdominal tip of the VNC at 6 h APF. (Insets) One representative vacuole. Images are single z-confocal slices. Bars: 20 μm; insets, 2 μm. Genotypes used were as follows: control (alrm-Gal4,UAS-mCD8∷GFP/+), drpr^Δ5/+ (alrm-Gal4,UAS-mCD8∷GFP/+; drpr^Δ5/+), drpr^Δ5 (alrm-Gal4,UAS-mCD8∷GFP/+; drpr^Δ5), alrm>dCed-12^RNAi (alrm-Gal4, UAS-mCD8∷GFP/UAS-dCed-12^RNAi), and alrm>dCed-12^RNAi+drpr^Δ5 (alrm-Gal4, UAS-mCD8∷GFP/UAS-dCed-12^RNAi/+; drpr^Δ5). (B) Quantification of number of vacuoles of figures in A. Error bars represent ±SEM; number of brains used was five or more; two hemisegments were measured in each brain. (***) P < 0.001; (****) P < 0.0001. (C) Quantification of a cross-section area of vacuoles of the figures in A. Error bars represent ±SEM; number of brains used was five or more; two hemisegments were measured in each brain. (*) P < 0.05; (**) P < 0.01; (****) P < 0.0001. (D) Quantification of pixel intensity of nc82⁺ puncta in SOG, thorax, and abdominal regions of the VNC at L3 and 18 h APF from Supplemental Figure 11. Genotypes used were as follows: alrm>dCed-12^RNAi (alrm-Gal4/UAS-dCed-12^RNAi) and alrm>dCed-12^RNAi+drpr^Δ5 (alrm-Gal4/UAS-dCed-12^RNAi; drpr^Δ5). Error bars represent ±SEM; N-values are as follows: alrm-Gal4/+, N ≥ 18; alrm>dCed-12^RNAi, N ≥ 20; alrm-Gal4/+; drpr^Δ5, N ≥ 16; and alrm>dCed-12^RNAi+drpr^Δ5, N ≥ 18. (*) P < 0.05; (**) P < 0.01; (***) P < 0.001; (****) P < 0.0001.

Figure 5.

EcR, Draper, and Crk/Mbc/dCed-12 function in astrocytes to promote MB γ neuron clearance. (A) Astrocytes were labeled with GFP (alrm-Gal4, UAS-mCD8∷GFP/+; green) in A and C, glial nuclei were labeled with anti-Repo (blue), and MB lobes were labeled with anti-FasII (red) in A, C, and D. The adult-specific dorsal lobe is outlined (white line). Time point is as indicated (APF). Genotypes used were as follows: wild type (wt; alrm-Gal4,UAS-mCD8∷GFP/+) and alrm>EcR^DN (alrm-Gal4,UAS-mCD8∷GFP/UAS-EcR^DN). (A–D) All confocal images are z-projections. Bars, 10 μm. (B) Quantification of MB γ neuron pruning phenotype from A using the categories shown in Supplemental Figure 14 (see the Materials and Methods). Wild type, N = 24; alrm>EcR^DN, N = 20 hemisegments quantified. (C) Image of magnified dorsal MB lobe. Note the FasII⁺ debris inside astrocytic vacuoles. Time point is as indicated (APF). Genotype used was as follows: alrm-Gal4, UAS-mCD8∷GFP/+. Bars, 10 μm. (D) MB debris was scored at the time points indicated (APF). Arrows indicate extra MB γ debris. Bars, 10 μm. Genotypes used were as follows: Wild type (wt; alrm-Gal4/+), alrm>dCed-12^RNAi (alrm-Gal4/UAS-dCed-12^RNAi), drpr^Δ5 (alrm-Gal4/+; drpr^Δ5), and alrm>dCed-12^RNAi+drpr^Δ5 (alrm-Gal4/UAS-dCed-12^RNAi; drpr^Δ5). (E–G) Quantification of the MB γ neuron pruning phenotype from D. N-values are as follows: wild type, N = 36; alrm>dCed-12^RNAi, N = 22; drpr^Δ5, N = 32; and alrm>dCed-12^RNAi+drpr^Δ5, N = 23 hemisegments quantified at 18 h APF. N-values are as follows: wild type, N = 58; alrm>dCed-12^RNAi, N = 30; drpr^Δ5, N = 26; and alrm>dCed-12^RNAi + drpr^Δ5, N = 22 hemisegments quantified at 48 h APF. N-values are as follows: wild type, N = 38; alrm>dCed-12^RNAi, N = 20; drpr^Δ5, N = 18; and alrm>dCed-12^RNAi+drpr^Δ5, N = 32 hemisegments quantified at newly eclosed adults.

Figure 6.

EcR, Draper, and Crk/Mbc/dCed-12 function in astrocytes to promote clearance of vCrz⁺ neuronal debris. (A) vCrz⁺ neurons were labeled with anti-Crz (green) in A and B. Time points used were as follows: L3 (third instar larva). Genotypes used were as follows: control (alrm-Gal4, UAS-mCD8∷GFP/+) and alrm>EcR^DN (alrm-Gal4, UAS-mCD8∷GFP/UAS-EcR^DN). Confocal images in A and B are z-projection images. Bars, 20 μm. (B) Time points are as indicated in A. Bars, 20 μm. Genotypes used were as follows: control (alrm-Gal4/+), alrm>drpr^RNAi (alrm-Gal4/drpr^RNAi), drpr^Δ5 (alrm-Gal4/+; drpr^Δ5), alrm>dCed-12^RNAi (alrm-Gal4/dCed-12^RNAi), and alrm>dCed-12^RNAi+drpr^Δ5 (alrm-Gal4/dCed-12^RNAi; drpr^Δ5). (C) Quantification of pixel intensity of vCrz⁺ debris from B. (C,D) N-values are as follows: alrm-Gal4/+, N = 32; alrm-Gal4/drpr^RNAi, N = 11; alrm-Gal4/+; drpr^Δ5, N = 11; alrm-Gal4/dCed-12^RNAi, N = 14; and alrm-Gal4/dCed-12^RNAi; drpr^Δ5, N = 12 brains quantified. Error bars represent ±SEM. (****) P < 0.0001. (D) Quantification of the percentage of remaining vCrz⁺ cell bodies from B. Error bars represent ±SEM. (****) P < 0.0001.

Figure 7.

Loss of Draper delays neurite degeneration and clearance of vCrz⁺ neuronal cell bodies. (A) vCrz⁺ neurons were labeled with GFP (Crz-Gal4, UAS-mCD8∷GFP; green). Time points are as indicated (APF). Bars, 20 μm. Genotypes used were as follows: control (Crz-Gal4, UAS-mCD8∷GFP; +), drpr^Δ5/+ (Crz-Gal4, UAS-mCD8∷GFP; drpr^Δ5/+), and drpr^Δ5 (Crz-Gal4, UAS-mCD8∷GFP; drpr^Δ5). (B–D) Quantification of the percentage of intact lateral (B), medial (C), and horizontal (D) axons of vCrz⁺ neurons per hemisegment from A. (B–E) Control, N = 30; drpr^Δ5/+, N = 30; and drpr^Δ5, N = 32 hemisegments quantified. Error bars represent ±SEM. (*) P < 0.05; (***) P < 0.001; (****) P < 0.0001. (E) Quantification of the percentage of cell bodies of vCrz⁺ neurons per hemisegment from A.