Lipid-droplet-accumulating microglia represent a dysfunctional and proinflammatory state in the aging brain

Abstract

Microglia become progressively activated and seemingly dysfunctional with age, and genetic studies have linked these cells to the pathogenesis of a growing number of neurodegenerative diseases. Here we report a striking buildup of lipid droplets in microglia with aging in mouse and human brains. These cells, which we call 'lipid-droplet-accumulating microglia' (LDAM), are defective in phagocytosis, produce high levels of reactive oxygen species and secrete proinflammatory cytokines. RNA-sequencing analysis of LDAM revealed a transcriptional profile driven by innate inflammation that is distinct from previously reported microglial states. An unbiased CRISPR-Cas9 screen identified genetic modifiers of lipid droplet formation; surprisingly, variants of several of these genes, including progranulin (GRN), are causes of autosomal-dominant forms of human neurodegenerative diseases. We therefore propose that LDAM contribute to age-related and genetic forms of neurodegeneration.

Extended Data Fig. 1

Lipid droplet accumulating microglia are abundant in the hippocampus but rare in other brain regions of aged mice. a–d, Representative confocal images of the cortex (a), thalamus (b), corpus callosum (c) and hippocampal dentate gyrus (d) from 20-month old male mice stained for BODIPY⁺ (lipid droplets) and Iba1⁺ (microglia). Scale bar: 20 μm. Arrows point towards BODIPY⁺ lipid droplets. e, Quantification of BODIPY+/Iba1+ cells. n = 4 mice per group. One-way ANOVA followed by Tukey’s post hoc test. Error bars represent mean ± SD. ***P< 0.001.

Extended Data Fig. 2

LDAM have a unique transcriptional signature that minimally overlaps with published gene expression profiles of microglia in aging and neurodegeneration. a,b, IPA pathway analysis of genes that are significantly upregulated (a) or downregulated (b) in LD-hi microglia in aging. Analysis based on top 100 down- and up-regulated genes (Fisher’s exact test, Benjamini-Hochberg FDR). c-g, Expression plots comparing RNA-Seq data of LDAM (see Fig. 2) with published RNA-Seq data of microglia in aging (c), AD (d), ALS (e), disease-associated microglia (DAM) (f) and neurodegenerative microglia (MGnD) (g). Data are expressed as signed fdr, i.e the product of log2 FC and log10 fdr. h, Paired dot plot showing FPKM values of LD-lo and LD-hi microglia for ApoE (paired Student’s t-test; P= 0.423). Dotted lines connect LD-lo and LD-hi microglia sorted from the same samples. i, Heatmap showing expression changes of LDAM genes (genes differentially expressed in LD-hi microglia in aging) in LD-hi microglia from GRN^−/− mice, from LPS treated mice, and in microglia clusters revealed by Li et al. (2019) and Hammond et al. (2019)^,. Sample size in a,b,h: n = 3 samples per group. Each sample is a pool of microglia from the hippocampi of 3 mice. LD, lipid droplet.

Extended Data Fig. 3

LPS treatment induces lipid droplet formation in microglia and in BV2 cells. a,b, 3-month-old male mice were given intraperitoneal (i.p.) injections of LPS (1 mg/kg BW) for four days. Representative confocal images of BODIPY⁺ and Tmem119⁺ in the hippocampus (a) and of BODIPY and Iba1 staining in the cortex, corpus callosum, and thalamus (b). c-e, Lipidome profiling of lipid droplets from LPS-treated BV2 cells, primary microglia, and liver tissue. c, Pie charts showing that the lipid composition of lipid droplets from young and aged microglia is highly similar, but differs between young and aged liver tissue. d,e, Distribution of MAG chain lengths (d) and TAG saturation levels (e) of lipid droplets isolated from LPS-treated BV2 cells and from microglia and liver tissue from aged mice. young = 5-month-old male mice, old= 20-month-old male mice; n = 4 mice per group. Data in a-b were replicated in at least two independent experiments. Error bars represent mean ± s.e.m. Scale bars, 20 μm.

Extended Data Fig. 4

Aged plasma induces lipid droplet formation in BV2 cells. a, Representative micrographs of BODIPY⁺ staining and of phagocytosis of pHrodo red Zymosan in BV2 cells treated with 5% plasma from young (3-months) and aged (18-months) mice for 12 hours. Scale bars, 5 μm. b, Quantification of BODIPY⁺ staining in BV2 cells treated with young and aged plasma. c,d, Quantification of Zymosan uptake in BV2 cells treated with young and aged plasma (c), and in aged plasma treated BODIPY-low and BODIPY-high cells (d). Statistical tests: two-sided Student’s t-test. Error bars represent mean ± SD. *P< 0.05, ***P< 0.001.

Extended Data Fig. 5

Lipid droplet containing microglia in the cortex, corpus callosum, and thalamus of GRN^−/− mice. a-c, Representative confocal images of BODIPY⁺ (lipid droplets) and Iba1⁺ (microglia) in the cortex (a), corpus callosum (b), (c) and thalamus from 9-month-old male GRN^−/− mice. BODIPY⁺/Iba1⁺ cells were frequently found in the thalamus and were detected to a lesser extent in cortex and corpus callosum. Data were replicated in at least three independent experiments.

Extended Data Fig. 6

Expression changes of LDAM genes in lipid droplet-rich microglia from normal aging, GRN^−/− and LPS-treated mice. a, Heatmap showing expression changes of LDAM genes (genes differentially expressed in LD-hi microglia in aging; 692 genes) in LD-hi microglia from GRN^−/− mice and from LPS treated mice.

Extended Data Fig. 7

LDAM show signs of metabolic alterations. a, Paired dot plot showing FPKM values of LD-lo and LD-hi microglia for ACLY (data obtained from RNA-Seq analysis, see Fig. 2). Dotted lines connect LD-lo and LD-hi microglia sorted from the same samples. P=b, NAD colorimetric assay showing the NAD⁺/NADH ratio of primary hippocampal microglia from 3-month old mice (young) and of LD-lo and LD-hi primary microglia from 20-month old male mice. Experiments were performed two times in technical triplicates. n=3 mice per group per experiment. Statistical tests: paired two-sided Student’s t-test (a) one-way ANOVA (b) followed by Tukey’s post hoc test. Horizontal lines in the box plots indicate medians, box limits indicate first and third quantiles, and vertical whisker lines indicate minimum and maximum values. *P< 0.05, ***P< 0.001.

Fig. 1. Microglia in the aged brain accumulate lipid droplets.

a, Electron microscopy of microglia from young and aged mice. b, Hippocampus from aged mice stained for BODIPY⁺ (lipid droplets) and TMEM119⁺ (microglia). Right panel shows 3D reconstruction of BODIPY⁺/TMEM119⁺ microglia. Arrows indicate lipid droplets. c-e, Quantification of BODIPY⁺ lipid droplet numbers (c), percent BODIPY⁺/TMEM119⁺ cells (d), and average BODIPY⁺ lipid droplet size (e) in the hippocampus (dentate gyrus). n = 6 mice per group. f, Representative image of Plin3⁺ (lipid droplets) TMEM119⁺ microglia in aged mice. g, Confocal images of Plin2⁺ (lipid droplets) and Iba1⁺ (microglia) in the human hippocampus of a 22-year-old and 67-year-old individual. Arrows indicate Plin2⁺Iba1⁺ cells. h,i, Representative images (h) and quantification (i; P=0.01) of CARS⁺ signal (2845 cm⁻¹) in TMEM119⁺ microglia in the hippocampus of young and aged mice. n = 5 mice per group. j,k Experimental schematic for lipidomics analysis of lipid droplets isolated from whole hippocampus and from FACS-sorted microglia from 20-month old mice (j), and pie charts showing the composition of lipid droplets (k); n = 4 mice per group. Statistical tests: two-sided Student’s t-test. Error bars represent mean ± SD. **P< 0.01, ***P< 0.001. Data in f-g were replicated in at least two independent experiments. Scale bars, 1 μm (a), 20 μm (b,g,h), 10 μm (f). n, nucleus; LD, lipid droplet; Ly, Lysosome; TAG, triacylglycerol; DAG, diacylglycerol; MAG, monoacylglycerol; CE, cholesteryl ester. Young= 3-month-old male mice; Aged= 20 month-old male mice

Fig. 2. RNA-Seq of lipid droplet-low and lipid droplet-high microglia from aged mice reveals transcriptional changes linked to phagocytosis and ROS production.

a, Flow sorting scheme for isolation of BODIPY^lo (=LD-lo) and BODIPY^hi (=LD-hi) CD11b⁺CD45^lo cells from the hippocampus of 18-month old male mice. n = 3 samples per group. Each sample is a pool of microglia from the hippocampi of 3 mice. b, Representative images of microglia after brain homogenization and marker staining, before (upper panel) and after (lower panel) FACS sorting. Scale bars, 5 μm. c, Volcano plot showing differentially expressed genes in LD-hi versus LD-lo microglia. Dashed line represents q-value < 0.05 cutoff (two-sided Student’s t-test, Benjamini-Hochberg FDR). Genes involved in phagosome maturation (red) and ROS production (purple) are highlighted. d, Heatmap showing the top 50 differentially expressed genes (q<0.05, ranked by p-value; R DeSeq2 Package, pairwise comparisons, Benjamini-Hochberg FDR). e, Top canonical pathways identified by IPA that are differentially regulated between LD-hi and LD-lo microglia. Analysis based on top 200 genes ranked by p-value (Fisher’s exact test, Benjamini-Hochberg FDR). f, IPA upstream regulator analysis of top 200 differentially expressed genes between LD-lo and LD-hi microglia (Fisher’s exact test, Benjamini-Hochberg FDR). g, Overlap between genes changing in microglia in aging and neurodegeneration (Aging, AD, ALS, DAM, MGnD), and genes upregulated (yellow) or downregulated (blue) in LD-hi microglia. Percent overlap denotes the fraction of genes in each gene list that are up- or down-regulated in LD-hi microglia. LD, lipid droplet.

Fig. 3. LPS treatment induces lipid droplet formation in microglia.

a-d, BV2 cells were treated with PBS or LPS (5 μg/ml) for 18 hours and co-treated with Triacsin C (1μM) or saline. Representative micrographs of BODIPY⁺ staining in BV2 cells (a) and quantification of BODIPY⁺ cells (b). c,d, Representative flow cytometry histogram (c) and quantification (d) of BODIPY fluorescence in BV2 cells. e-h, Lipidome profiling of lipid droplets from LPS-treated BV2 cells, and from primary microglia and liver tissue from aged (20-month-old) mice. Overall composition of lipid droplets (e), percentage of neutral lipids (f), and distribution of TAG species (g,h). i,j, BODIPY⁺ and Iba1⁺ in the hippocampus of 3-month old male mice given intraperitoneal (i.p.) injections of PBS or LPS (1 mg/kg BW) for four days. Representative confocal images (i) and quantification (j; P=0.008) of BODIPY⁺/Iba1⁺ microglia. n = 6 mice per group. k-m, RNA-Sequencing of BODIPY^lo (=LD-lo) and BODIPY^hi (=LD-hi) CD11b⁺CD45^lo microglia from the hippocampus of 3-month old LPS-treated mice. n = 3 biologically independent samples per group. Each sample is a pool of microglia from the hippocampi of two mice. k, Volcano plot showing differentially expressed genes in LD-hi versus LD-lo microglia. Dashed line represents q-value < 0.05 cutoff (two-sided Student’s t-test, Benjamini-Hochberg FDR). LDAM genes are highlighted in green. l, EnrichR pathway analysis of genes differentially regulated between LD-hi and LD-lo microglia (Fisher’s exact test, Benjamini-Hochberg FDR). m, Scatterplot showing gene expression intensities (mean normalized counts) of LD-hi microglia in LPS-treated mice compared to LDAM in aging. Genes differentially expressed in LDAM are highlighted in red. Experiments on BV2 cells were performed three times in technical triplicates. Statistical tests: two-sided Student’s t-test (j) and one-way ANOVA followed by Tukey’s post hoc test (b,d). Error bars represent mean ± SD (j). Horizontal lines in the box plots indicate medians, box limits indicate first and third quantiles, and vertical whisker lines indicate minimum and maximum values (b,d). *P< 0.05, **P< 0.01, ***P< 0.001. Scale bars, 5 μm (a), 20 μm (i). CE, cholesteryl ester; DAG, diacylglycerol; LD, lipid droplet; MAG, monoacylglycerol; MFI, mean fluorescent intensity; TAG, triacylglycerol; TrC Triacsin C.

Fig. 4. LDAM and lipid droplets in BV2 cells are associated with impaired phagocytosis.

a, Pathway map of genes related to phagosome maturation that are differentially expressed in LDAM (see Fig. 2). b, Confocal images showing BODIPY⁺ (lipid droplets), CD68⁺ (endosomes/ lysosomes), and Iba1⁺ in the hippocampus from 20-month old mice. c, Percentage of BODIPY^- and BODIPY⁺ Iba1⁺ microglia with high levels of CD68 (CD68^hi). n = 5 mice per group; P=0.002. d, 3D reconstruction of Iba1⁺ microglia showing BODIPY⁺ lipid droplets closely surrounded by CD68⁺ vesicles. e, Electron microscopy image showing lysosomal accumulation in LDAM from a 20-month old mouse. f,g, Confocal images (f) and quantification (g) of BODIPY⁺ and Zymosan⁺ in BV2 cells treated with LPS (5 ug/ml) for 18 hours. h,i, Phagocytosis of pHrodo red Zymosan in BV2 cells treated with PBS or LPS and co-treated with Triacsin C (1μM) or saline. Representative images of (h) and time lapse imaging and quantification (i) of Zymosan uptake in BV2 cells. j,k, 250 μm organotypic brain slices from 12-month old mice were incubated for 4 hours with pHrodo red Zymosan particles. Representative confocal images of the hippocampus (j) and pie chart showing the percentages of Zymosan-containing BODIPY^- and BODIPY⁺ Iba1⁺ cells (k). P-value for Zymosan⁺BODIPY^- vs Zymosan⁺BODIPY⁺ Iba1⁺ cells = 0.0012. n = 3 mice per group. l-n, Myelin debris labelled with Alexa Fluor 555 was stereotactically injected into the hippocampus of 20-month old male mice, and phagocytosis was analyzed 48 hours after injection. m, Representative images of AF555-labelled myelin (left panel) and of Iba1⁺ cells with and without lipid droplets (BODIPY⁺) and A555⁺ myelin. n, Quantification of myelin uptake in BODIPY⁺/Iba1⁺ and BODIPY^- Iba1⁺ cells. n = 6 mice per group; P=0.038. Experiments on BV2 cells were performed three times in technical triplicates. Statistical tests: two-sided Student’s t-test (c,g,n), two-way ANOVA followed by Tukey’s post hoc test (i,k). Error bars represent mean ± SD (c,g,n) and mean ± SEM (i). *P< 0.05, **P< 0.01, ***P< 0.001. Data in d-e were replicated in at least two independent experiments. Scale bars, 20 μm (b,j,m), 0.5 μm (e), 5 μm (f), 200 μm (h). LD, lipid droplet; Ly, Lysosomes; n, nucleus; TrC, Triacsin C.

Fig. 5. LDAM and lipid droplet-rich BV2 cells show increased ROS production and LDAM secrete elevated levels of inflammatory cytokines.

a, Pathway map of genes related to ROS production that are differentially expressed in LDAM (see Fig. 2). b, Experimental schematic of ROS analysis in primary microglia from young and aged mice. c, Representative flow cytometry histogram and quantification of CellROX fluorescence in primary microglia from young and aged mice; P=0.0013. d, Gating scheme for BODIPY^lo (LD-lo) and BODIPY^hi (LD-hi) microglia from aged mice. e, Histogram and quantification of CellROX fluorescence in LD-lo and LD-hi microglia from aged mice; P=0.0025. f-h, CellROX fluorescence in BV2 cells treated with PBS or LPS (5 ug/ml) for 18 hours, co-treated with Triacsin C (1μM) or saline. f, Representative images of CellROX⁺ signal in BV2 cells. g, Confocal images showing BODIPY⁺ and CellROX⁺ in LPS treated BV2 cells. h, Flow cytometry histogram and quantification of CellROX fluorescence in BV2 cells. i,j Acutely isolated LD-lo and LD-hi primary microglia from aged mice were treated with LPS (100 ng/ml) for 8 hours, and cytokine concentrations in the medium were measured using multiplex array. N=6 biologically independent samples, pooled from two independent experiments (3 mice per experiment); each sample corresponds to microglia isolated from the hippocampus of one mouse. Heatmap showing changes in cytokine secretion under baseline conditions and after LPS treatment (i) and individual dot plots of selected cytokines (j). Experiments were performed three (BV2 cells) or two (primary cells) times in technical triplicates. Primary cells were isolated from three mice per group per experiment. Statistical tests: two-sided Student’s t-test (c,e), one-way ANOVA (h,j) followed by Tukey’s post hoc test. Horizontal lines in the box plots indicate medians, box limits indicate first and third quantiles, and vertical whisker lines indicate minimum and maximum values (c,e,h). Error bars represent mean ± SD (j) *P< 0.05, **P< 0.01, ***P< 0.001. Data in f-g were replicated in at least two independent experiments. Scale bars, 20 μm (f), 5 μm (g). LD, lipid droplet; MFI, mean fluorescent intensity; TrC, Triacsin C. Young= 3-month-old male mice; Aged= 20 month-old male mice.

Fig. 6. CRISPR-Cas9 screen identifies genetic regulators of lipid droplet formation.

a, Structure of BODIPY 493/503 and of iodo-BODIPY. b,c, Calcein⁺ signal in BV2 cells that were treated with 5 μg/ml LPS for 18 hours, stained with iBP, and exposed to photoirradiation for 3 hours. Representative micrographs (b) and quantification (c) of Calcein⁺ (live cells) and iBP⁺ (lipid-droplet containing cells) in control (non-irradiated) and irradiated BV2 cells; P=0.00046. d, Experimental schematic for pooled CRISPR-Cas9 screen to identify regulators of lipid droplet formation in LPS-treated BV2 cells. e, Volcano plot showing hits for genetic regulators of lipid droplet formation from the CRISPR-Cas9 knockout screen. Dashed line represents P-value < 0.05 cutoff. Two-sided Student’s t-test. Positive effect size represents genes targeted by sgRNAs that were enriched in lipid droplet- negative cells; negative effect size represents genes targeted by sgRNAs that were under-represented in lipid droplet- negative cells. Genes previously associated with neurodegeneration are highlighted in blue and yellow. Screens were performed as technical duplicates. f-i, Single CRISPR-Cas9 knockout BV2 cell lines of selected screen hits (SNX17^−/−, GRN^−/−, SLC33A1^−/−, VPS35^−/−). Cas9-expressing BV2 cells were used as control (CTR). Representative micrographs of BODIPY⁺ staining in PBS or LPS treated cells (f) and quantification of the percentage of BODIPY⁺ cells (g). Quantification of CellROX MFI in PBS or LPS treated cells (h). Time lapse imaging and quantification of Zymosan uptake in cells treated with PBS or LPS (i). Experiments on BV2 cells were performed three times in technical triplicates. Statistical tests: two-sided Student’s t-test (c), two-way ANOVA (g,h,i) followed by Tukey’s post hoc test. Error bars represent mean ± SD (c,g,h) and mean ± SEM (i). Horizontal lines in the box plots indicate medians, box limits indicate first and third quantiles, and vertical whisker lines indicate minimum and maximum values. *P<0.05, **P< 0.01, ***P< 0.001. Scale bars, 20 μm (b,f). LD, lipid droplet; iBP, iodo-BODIPY; MFI, mean fluorescent intensity.

Fig. 7. GRN^−/− mice possess high numbers of lipid droplet-rich microglia which functionally and partially transcriptionally resemble LDAM.

a-c, BODIPY⁺ and Iba1⁺ expression in the hippocampus of 9-month old male WT mice and age- and sex-matched GRN^−/− mice. Representative confocal images (a), and quantification of BODIPY⁺/Iba1⁺ microglia (b, P=0.0064) and of lipid droplet numbers per Iba1⁺ microglia (c). n = 5 mice per group. d,e, 250 μm organotypic brain slices from 9-month old GRN^−/− mice were incubated for 4 hours with pHrodo red Zymosan particles. Representative confocal images of the hippocampus (d) and quantification of Zymosan uptake in BODIPY^- and BODIPY⁺ Iba1⁺ cells (e). n = 3 mice per group. f, Quantification of CellROX fluorescence in primary LD-lo and LD-hi microglia from 9-month old GRN^−/− mice and age-matched wild type controls. g, Acutely isolated LD-lo and LD-hi primary microglia from 10-month old GRN^−/− mice were treated with LPS (100 ng/ml) or PBS for 8 hours. Heatmap shows changes in cytokine secretion under baseline conditions (PBS) and after LPS treatment. h-k, RNA-Sequencing of BODIPY^lo (=LD-lo) and BODIPY^hi (=LD-hi) CD11b⁺CD45^lo microglia from the hippocampus of 10-month old GRN^−/− mice. n = 3 biologically independent samples per group. Each sample is a pool of microglia from the hippocampi of two mice. h, Volcano plot showing differentially expressed genes in LD-hi versus LD-lo microglia. Dashed line represents q-value < 0.05 cutoff (two-sided Student’s t-test, Benjamini-Hochberg FDR). LDAM genes are highlighted in green. i, EnrichR pathway analysis of genes differentially regulated between LD-hi and LD-lo microglia (Fisher’s exact test, Benjamini-Hochberg FDR). j, Scatterplot showing gene expression intensities (mean normalized counts) of LD-hi microglia in GRN^−/− mice compared to LDAM in aging. Genes differentially expressed in LDAM are highlighted in red. k, Scatterplot showing expression changes of LDAM DE genes in LD-hi microglia from GRN^−/− mice and in LDAM in aging. Spearman correlation’s coefficient was used for correlation analysis. l, Overlap of genes differentially expressed between LD-hi and LD-lo microglia in aging (LDAM), in GRN^−/− mice, and in LPS-treated young mice, and list of genes that are shared between all groups. m, Box plots showing the scaled expression of up- and downregulated DE LDAM genes (692 genes) in aging, GRN^−/− mice, and LPS-treated young mice. n,o, Heatmap showing expression changes (n) and EnrichR pathway analysis (o) of the shared genes (red) and the 20 most significant genes shared between LDAM and LD-hi microglia in GRN^−/− mice (ranked by P-value; Fisher’s exact test, Benjamini-Hochberg FDR). sample size: Aging (j-o): n = 3 samples per group. Each sample is a pool of microglia from the hippocampi of 3 mice. GRN^−/− mice (h-o): n = 3 samples per group. Each sample is a pool of microglia from the hippocampi of 2 mice. LPS (m-o): n = 4 samples per group. Each sample is a pool of microglia from the hippocampi of 2 mice. Statistical tests: two-sided Student’s t-test (b,c,e) and one-way ANOVA (f) followed by Tukey’s post hoc test. Error bars represent mean ± SD. *P<0.05, **P< 0.01, ***P< 0.001. Experiments on primary cells (f,g) were performed two times in technical triplicates. Scale bars, 20 μm (a), 10 μm (d). DE genes, differentially expressed genes; LD, lipid droplet; MFI, mean fluorescent intensity.