Astrocyte-derived MFG-E8 facilitates microglial synapse elimination in Alzheimer's disease mouse models

Abstract

Region-specific synapse loss is an early pathological hallmark in Alzheimer's disease (AD). Emerging data in mice and humans highlight microglia, the brain-resident macrophages, as cellular mediators of synapse loss; however, the upstream modulators of microglia-synapse engulfment remain elusive. Here, we report a distinct subset of astrocytes, which are glial cells essential for maintaining synapse homeostasis, appearing in a region-specific manner with age and amyloidosis at onset of synapse loss. These astrocytes are distinguished by their peri-synaptic processes which are 'bulbous' in morphology, contain accumulated p62-immunoreactive bodies, and have reduced territorial domains, resulting in a decrease of astrocyte-synapse coverage. Using integrated in vitro and in vivo approaches, we show that astrocytes upregulate and secrete phagocytic modulator, milk fat globule-EGF factor 8 (MFG-E8), which is sufficient and necessary for promoting microglia-synapse engulfment in their local milieu. Finally, we show that knocking down Mfge8 specifically from astrocytes using a viral CRISPR-saCas9 system prevents microglia-synapse engulfment and ameliorates synapse loss in two independent amyloidosis mouse models of AD. Altogether, our findings highlight astrocyte-microglia crosstalk in determining synapse fate in amyloid models and nominate astrocytic MFGE8 as a potential target to ameliorate synapse loss during the earliest stages of AD.

Fig 1.. Identification of a unique p62-accumulated bulbous astrocyte subset in the NL-F KI hippocampus.

A) Schematic of NL-F KI mouse model and viral tdTomato labeling strategy to visualize complex hippocampal astrocyte processes. Immunostaining for GFAP (green) post intra-hippocampal AAV2/5-GfaABC₁D-tdTomato (tdTomato) (magenta) injection. Scale bar = 10 μm. B) Representative images of tdTomato-labeled astrocytes (magenta) in the 6-mo WT and NL-F KI hippocampus. Scale bar = 10 μm. Images are modified on ImageJ as shown by white ROI. C) Quantification of the tdTomato labeled astrocyte area (μm²) using ImageJ. Transparent points = individual astrocytes (16–58 astrocytes per mouse; 104 WT and 183 NL-F KI astrocytes sampled in total); full points = mouse average of ROIs (n=4–6 male mice per genotype). Unpaired two-tailed student’s t-test on mouse average. D) Representative images of tdTomato-labeled (magenta) astrocytes in the 6-mo hippocampus showing a bushy and bulbous astrocyte. Scale bar = 20 μm. Insets show representative zoom-ins of processes. Scale bar = 2 μm. E) Quantification of the tdTomato-labeled astrocytic volume (μm³) using Imaris 3D surface rendering. Transparent points = individual astrocytes (5–7 astrocytes per condition; 35 NL-F KI bushy and 31 NL-F KI bulbous astrocytes sampled in total); full points = mouse average of ROIs (n=6 male mice). Paired two-tailed student’s t-test on mouse average. F) Representative image of p62 (green) and S100β (magenta) immunoreactivity in the hippocampus. Scale bar = 10 μm. Images are modified on ImageJ as shown by white ROI. Inset shows representative Imaris 3D reconstruction of p62 within the S100β⁺ astrocytic surface. G) Quantification of the % colocalization between p62 and S100β⁺ bulbous astrocytes in the hippocampus using ImageJ. Transparent points = individual astrocytes (3 astrocytes per mouse; 15 astrocytes sampled in total per condition), full points = mouse average of ROIs (n=5 male mice). Ratio paired two-tailed student’s t-test on mouse average. H) Representative image of p62 (green) immunoreactivity in the 6-mo NL-F KI hippocampus. Scale bar = 400 μm. Insets highlight some regions with bulbous astrocytes. I) Representative images of p62 (green) and S100β (magenta) immunoreactivity in the 2-, 6- and 20-mo hippocampus. Scale bar = 10 μm. Insets show representative zoom-ins of p62-immunostaining. J) Quantification of the number of p62⁺ S100β⁺ bulbous astrocytes per mm² in the hippocampus in 6-mo (left) and 20-mo (right). Each point = 1 mouse (n=5–6 male mice). Two-tailed Mann-Whitney (6-mo) or unpaired two-tailed student’s t-test (20-mo) on mouse average. Dashed line represents graph split. K) Representative images of p62 (green) and S100β (magenta) immunoreactivity in the hippocampus, motor cortex (M1) and thalamus. Scale bar = 10 μm. Insets show representative zoom-ins of p62-immunostaining. L) Quantification of the number of p62⁺ S100β⁺ bulbous astrocytes per mm² in hippocampus, motor cortex (M1) and thalamus. Each point = 1 mouse (n=5 male mice). Kruskal-Wallis test (p<0.001) followed by the Dunn’s multiple comparisons test on mouse average. All data shown as mean ± SEM. p-values shown ns P>0.05; *P<0.05; **P<0.01; ***P<0.001; ****P<0.0001.

Fig 2.. Loss of astrocyte-synapse interactions and increased loss of excitatory post-synaptic marker Homer1 near hippocampal NL-F KI bulbous astrocytes.

A) Schematic of astrocytic-synapse interactions. B-F) Representative images and quantification of astrocytic EAAT1-, excitatory post-synaptic Homer1- and pre-synaptic Synaptophysin-immunoreactive puncta in the 6-mo hippocampal CA1 SR using super-resolution Airyscan confocal microscopy. B) Quantification of the number of colocalized Homer1- and Synaptophysin-immunoreactive spots shown as density per μm³ using Imaris. Transparent points = individual ROIs (3 ROIs per mouse; 15 WT and 21 NL-F KI ROIs sampled in total); full points = mouse average of ROIs (n=5–7 male mice per genotype). Unpaired two-tailed student’s t-test on mouse average. C) Quantification of the number of EAAT1-immunoreactive spots shown as density per μm³ using Imaris. Transparent points = individual ROIs (3 ROIs per mouse; 15 WT and 21 NL-F KI ROIs sampled in total); full points = mouse average of ROIs (n=5–7 male mice per genotype). Unpaired two-tailed student’s t-test on mouse average. D) Quantification of the number of colocalized EAAT1, Homer1 and Synaptophysin spots divided by the number of colocalized Homer1 and Synaptophysin spots shown as %. Transparent points = individual ROIs (3 ROIs per mouse; 15 WT and 21 NL-F KI ROIs sampled in total); full points = mouse average of ROIs (n=5–7 male mice per genotype). Unpaired two-tailed student’s t-test on mouse average. E) Quantification of the number of colocalized EAAT1, Homer1 and Synaptophysin spots divided by the number of EAAT1 spots using Imaris shown as %. Transparent points = individual ROIs (3 ROIs per mouse; 15 WT and 21 NL-F KI ROIs sampled in total); full points = mouse average of ROIs (n=5–7 male mice per genotype). Unpaired two-tailed student’s t-test on mouse average. F) Representative image of B-E for EAAT1 (blue), Homer1 (green) and Synaptophysin (red). Scale bar = 1 μm. Insets show regions of triple colocalization. G) Representative images of Ezrin (green) immunoreactivity in the 6-mo NL-F KI hippocampus near bushy or bulbous S100β⁺ astrocytes (magenta). Scale bar = 10 μm. Images are modified on ImageJ as shown by white ROI. H) Quantification of the total Ezrin spot volume on S100β⁺ astrocytes/S100β volume using Imaris 3D surface rendering and surface-spot colocalization. Transparent points = individual astrocytes (5–8 astrocytes per condition; 36 NL-F KI bushy and 30 NL-F KI bulbous astrocytes sampled in total); full points = mouse average of ROIs (n=5 male mice). Paired two-tailed student’s t-test on mouse average. I) Representative images for excitatory post-synaptic Homer1-immunoreactive puncta (green) in the 6-mo NL-F KI hippocampus near bushy or bulbous S100β⁺ astrocytes (magenta) using super-resolution Airyscan confocal microscopy. Scale bar = 1 μm. J) Quantification of the number of Homer1-immunoreactive puncta shown as density per μm³ using Imaris. Transparent points = individual ROIs (3 ROIs per condition; 18 ROIs sampled in total per condition), full points = mouse average of ROIs (n=6 male mice). Wilcoxon matched pairs signed rank test on mouse average. All data shown as mean ± SEM. p-values shown ns P>0.05; *P<0.05; **P<0.01; ***P<0.001; ****P<0.0001.

Fig 3.. Increased microglia-Homer1 engulfment near hippocampal NL-F KI bulbous astrocytes.

A) Representative 3D rendered images of excitatory post-synaptic Homer1 (white), Lamp1 lysosomes (red) and GFAP (green) in the 6-mo hippocampal CA1 SR. Scale bar = 10 μm. Inset shows representative zoom of Homer1 inside Lamp1⁺ astrocytic lysosomes. B) Quantification of astrocytic Homer1 engulfment using Imaris 3D surface rendering shown as: Homer1 volume in Lamp1⁺ lysosomes in GFAP⁺ astrocytic surface/GFAP volume x100. Transparent points = individual astrocytes (5–10 astrocytes per mouse; 40 WT and 39 NL-F KI astrocytes sampled in total); full points = mouse average of ROIs (n=5 male mice per genotype). Unpaired two-tailed student’s t-test with Welch’s correction on mouse average. C) Representative 3D rendered images of excitatory post-synaptic Homer1 (white), Lamp1 lysosomes (green) and tdTomato (magenta) in the 6-mo hippocampal CA1 SR. Scale bar = 10 μm. Inset shows representative zoom of Homer1 inside Lamp1⁺ astrocytic lysosomes. D) Quantification of astrocytic Homer1 engulfment using Imaris 3D surface rendering shown as: Homer1 volume in Lamp1⁺ lysosomes in tdTomato⁺ astrocytic surface/tdTomato volume x100. Transparent points = individual astrocytes (4–8 astrocytes per condition; 26 WT, 35 NL-F KI bushy and 31 NL-F KI bulbous astrocytes sampled in total), full points = mouse average of ROIs (n=4–6 male mice). Unpaired (WT vs NL-F KI) or paired (NL-F KI bushy vs NL-F KI bulbous) two-tailed student’s t-test on mouse average. Dashed line represents graph split. E) Representative 3D rendered images of excitatory post-synaptic Homer1 (white), CD68 lysosomes (red) and P2Y12 (green) in the 6-mo hippocampal CA1 SR. Scale bar = 10 μm. Inset shows representative zoom of Homer1 inside CD68⁺ microglial lysosomes. F) Quantification of microglial Homer1 engulfment using Imaris 3D surface rendering shown as: Homer1 volume in CD68⁺ lysosomes in P2Y12⁺ microglial surface/P2Y12 volume x100. Transparent points = individual microglia (5–6 microglia per mouse; 36 WT and 28 NL-F KI microglia sampled in total); full points = mouse average of ROIs (n=5–6 male mice per genotype). Two-tailed Mann Whitney test on mouse average. G) Representative 3D rendered images of excitatory post-synaptic Homer1 (white), CD68 lysosomes (red) and P2Y12 (green) in the 6-mo hippocampal CA1 SR near bushy or bulbous tdTomato⁺ (magenta) astrocytes. Scale bar = 10 μm. Inset shows representative zoom of Homer1 inside CD68⁺ microglial lysosomes. H) Quantification of microglial Homer1 engulfment using Imaris 3D surface rendering shown as: Homer1 volume in CD68⁺ lysosomes in P2Y12⁺ microglial surface/P2Y12 volume x100. Transparent point = individual microglia (5–7 microglia per condition; 32 and 28 microglia sampled in total near (< 5 μm away) bushy and bulbous NL-F KI astrocytes respectively), full points = mouse average of ROIs (n=6 male mice). Paired two-tailed student’s t-test on mouse average. I) Quantification of the number of Iba1⁺ microglia near bushy or bulbous NL-F KI astrocytes. Transparent points = individual astrocyte regions (4–10 regions per condition; 30 NL-F KI bushy and 33 NL-F KI bulbous astrocyte regions sampled in total); full points = mouse average of ROIs (n=5 male mice). Wilcoxon matched-pairs signed rank test followed by a Bonferroni-Dunn correction on mouse average. J) Representative 3D rendered images of MFG-E8 (green) inside S100β⁺ (magenta) bushy or bulbous astrocytic processes in the 6-mo NL-F KI hippocampus. Scale bar = 5 μm. K) Quantification of the volume of MFG-E8 within bushy or bulbous S100β⁺ astrocytic processes per μm³ using Imaris 3D surface rendering. Transparent points = individual astrocytes (3–5 astrocytes per condition; 27 NL-F KI bushy and 30 NL-F KI bulbous astrocytes sampled in total); full points = mouse average of ROIs (n=5 male mice). Paired two-tailed student’s t-test on mouse average. All data shown as mean ± SEM. p-values shown ns P>0.05; *P<0.05; **P<0.01; ***P<0.001; ****P<0.0001.

Fig 4.. Increased astrocytic MFG-E8 signaling at the onset of synapse loss in the NL-F KI hippocampus.

A) Representative image of RNAScope probing for Mfge8 (green), Aldh1l1⁺ astrocytes (magenta) and Cx3cr1⁺ myeloid cells (yellow) in the hippocampus. Scale bar = 25 μm. Inset shows exemplary ROI. B) Quantification of the number of Mfge8 mRNA spots on Aldh1l1⁺ astrocytes, Cx3cr1⁺ myeloid cells or Aldh1l1⁻Cx3cr1⁻ cells in the hippocampus using Imaris. Data shown as a percentage of the total number of Mfge8 spots in a large area of interest. Each point = 1 mouse (n=5 female mice). One-way ANOVA (F (2, 12) = 378.9, p<0.0001) followed by Bonferroni’s multiple comparisons post-hoc test on mouse average. C) RT-qPCR probing for Mfge8 expression in FACS sorted ACSA2⁺CD45⁻ astrocytes and CX3CR1⁺CD45⁺CD11b⁺ microglia. Gene expression is normalized to the geomean of 3 house-keeping genes (Actb, Gapdh and Rpl32) by the Delta CT method. Each point = 1 mouse (n=3–4 female mice). Unpaired two-tailed student’s t-test with a Welch’s correction on mouse average. D) Representative images of MFG-E8 (magenta) and GFAP (green) immunoreactivity in the post-mortem human frontal cortex. Scale bar = 10 μm. E) RNAScope probing for Mfge8 (green) on Slc1a2⁺ (magenta) astrocytes in the 6-mo WT and NL-F KI hippocampus. Scale bar = 15 μm. F) Quantification of the total number of Mfge8 spots in a large region of interest using Imaris. Each point = 1 mouse (n=4–5 mixed male and female mice per genotype). Unpaired two-tailed student’s t-test on mouse average. G) Quantification of the number of Slc1a2⁺ astrocytes in a large region of interest using Imaris. Each point = 1 mouse (n=4–5 mixed male and female mice per genotype). Unpaired two-tailed student’s t-test on mouse average. H) Quantification of the number of Mfge8 mRNA spots in the hippocampus shown as percent distribution on Slc1a2⁺ astrocytic nuclei. Each point = mouse average (n=4–5 mixed male and female mice per genotype). Two-way ANOVA (Interaction: F (2, 21) = 6.188, p<0.01) followed by Bonferroni’s multiple comparisons post-hoc test on Y=log(Y) transformed mouse average. I) Representative images of MFG-E8 (white) immunoreactivity in the 6-mo hippocampal CA1 SR in WT and NL-F KI mice. Scale bar = 30 μm. J) Quantification of MFG-E8 immunoreactivity in large region using ImageJ. Each point = mouse average (n=7 male mice per genotype). Unpaired two-tailed student’s t-test on mouse average. K) Western blot probing for MFG-E8 (55–75 kDa) with respect to GAPDH loading control (39 kDa) in 6-mo WT and NL-F KI hippocampal homogenates. 1 lane is 1 mouse. L) Quantification of MFG-E8/GAPDH using ImageJ densitometry analysis. Each point = mouse average (n=5 male mice per genotype). Unpaired two-tailed student’s t-test on mouse average. M) Representative images of RNAScope followed by immunostaining probing for either Itga_v (cyan) and Itgb₅ (magenta) integrin mRNA subunits (left) or Itga_v (cyan) and Itgb₃ (magenta) integrin mRNA subunits (right) with microglial P2Y12 (green) in the hippocampus. Scale bar = 20 (left) and 15 (right) μm. Inset highlight ROIs. N) Flow cytometry probing for CD61 integrin on CD45⁺CD11b⁺CX3CR1⁺ hippocampal microglia from 6-mo WT and NL-F KI mice. O) Quantification of the levels of CD61 on CD45⁺CD11b⁺CX3CR1⁺ microglia. Each point = mouse average (n=5 female mice per genotype). Unpaired two-tailed student’s t-test followed by a Welch’s correction on mouse average. All data shown as mean ± SEM. p-values shown ns P>0.05; *P<0.05; **P<0.01; ***P<0.001; ****P<0.0001.

Fig 5.. Viral CRISPR-saCas9 deletion of astrocytic MFG-E8 prevents microglia-Homer1 engulfment and excitatory loss in the NL-F KI hippocampus.

A) Schematic of CRISPR-saCas9 paradigm to knock-down Mfge8 from hippocampal astrocytes with spatio-temporal control. Immunostaining for gRNA-mCherry (magenta) and Cas9-HA (green) constructs post-CRISPR-saCas9 injection. Scale bar = 5 nm. Representative image of MFG-E8 (white) immunoreactivity post-CRISPR-saCas9 injection. Scale bar = 1 nm. Inset highlights ROIs. mCherry injection zone drawn with white ROI on ImageJ. B) Representative images of gRNA-mCherry (red) and Cas9-HA (green) constructs with either GFAP (blue, left), NeuN (blue, middle) or Iba1 (blue, right) post-CRISPR-saCas9 injection. Scale bar = 30 μm. C) Quantification of the colocalized mCherry-labeled signal with GFAP⁺ astrocytes, NeuN⁺ neurons or Iba1⁺ microglia in the hippocampus using ImageJ shown as percentage of the total mCherry-labeled signal. Each point = 1 mouse (n=6 mixed male and female mice). Brown-Forsythe (643.8 (2.000, 6.920, p<0.0001) and Welch ANOVA (339.9 (2.000, 8.829), p<0.0001) with Dunnett’s T3 multiple comparisons test. D) Quantification of the colocalized GFAP⁺ astrocytes with mCherry-labelled signal in the hippocampus using ImageJ shown as percentage of the total GFAP⁺ astrocytes. 8–10 male mice sampled. E) Representative images of MFG-E8 (white) immunoreactivity in the 6-mo WT and NL-F KI SR hippocampus post control or test CRISPR-saCas9 injection. Scale bar = 20 μm. F) Quantification of MFG-E8 immunoreactivity using ImageJ. Data shown as normalized to contra control CRISPR injected hemisphere. Transparent points = individual ROIs (3 ROIs per condition; 24 ROIs sampled in total per condition), full points = mouse average of ROIs (n=8 male mice per condition). Two-way ANOVA (Interaction: F (1, 28) = 218.6, p<0.0001) followed by Bonferroni’s multiple comparisons post-hoc test on Y=log(Y) transformed mouse average. G) Representative 3D rendered images of excitatory post-synaptic Homer1 (white), CD68 lysosomes (red) and P2Y12 (green) in the 6-mo WT and NL-F KI CA1 SR hippocampus post control or test CRISPR-saCas9 injection. Scale bar = 10 μm. Inset shows representative zoom of Homer1 inside CD68⁺ microglial lysosomes. H) Quantification of microglial Homer1 engulfment using Imaris 3D surface rendering shown as: Homer1 volume in CD68⁺ lysosomes in P2Y12⁺ microglial surface/P2Y12 volume x100. Transparent points = individual microglia (5–6 microglia per mouse; 24 WT control, 24 NL-F KI control, 24 WT test and 34 NL-F KI test injected microglia were sampled in total), full points = mouse average of ROIs (n=4–6 male mice per condition). Two-way ANOVA (Interaction: F (1, 14) = 4.929, p<0.05) followed by Bonferroni’s multiple comparisons post-hoc test on Y=log(Y) transformed mouse average. I) Representative images of excitatory post-synaptic Homer1 (green) and pre-synaptic Bassoon (magenta) immunostaining in the 6-mo WT and NL-F KI CA1 SR hippocampus post control or test CRISPR-saCas9 injection using super-resolution Airyscan confocal microscopy. Scale bar = 1 μm. Insets show regions of colocalization. J) Quantification of the number of colocalized Homer1 and Bassoon spots shown as density per μm³ using Imaris. Transparent points = individual ROIs (3 ROIs per mouse; 12 WT control, 12 NL-F KI control, 12 WT test and 15 NL-F KI test injected ROIs were sampled in total), full points = mouse average of ROIs (n=4–5 male mice per condition). Two-way ANOVA (Interaction: F (1, 13) = 11.02, p<0.01) followed by Bonferroni’s multiple comparisons post-hoc test on Y=log(Y) transformed mouse average. All data shown as mean ± SEM. p-values shown ns P>0.05; *P<0.05; **P<0.01; ***P<0.001; ****P<0.0001.