Neuronal lipid droplets play a conserved and sex-biased role in maintaining whole-body energy homeostasis

Abstract

Lipids are essential for neuron development and physiology. Yet, the central hubs that coordinate lipid supply and demand in neurons remain unclear. Here, we combine invertebrate and vertebrate models to establish the presence and functional significance of neuronal lipid droplets (LD) in vivo. We find that LD are normally present in neurons in a non-uniform distribution across the brain, and demonstrate triglyceride metabolism enzymes and lipid droplet-associated proteins control neuronal LD formation through both canonical and recently-discovered pathways. Appropriate LD regulation in neurons has conserved and male-biased effects on whole-body energy homeostasis across flies and mice, specifically neurons that couple environmental cues with energy homeostasis. Mechanistically, LD-derived lipids support neuron function by providing phospholipids to sustain mitochondrial and endoplasmic reticulum homeostasis. Together, our work identifies a conserved role for LD as the organelle that coordinates lipid management in neurons, with implications for our understanding of mechanisms that preserve neuronal lipid homeostasis and function in health and disease.

Keywords: AgRP; C. elegans; Drosophila; Lipid droplet; adipokinetic hormone; adipose triglyceride lipase (ATGL); arcuate nucleus; energy homeostasis; mouse; neuron; sex difference; triglyceride.

Figure 1.. Lipid droplets are present in neurons across species under normal physiological conditions.

a, Illustration of Drosophila brain; mushroom bodies and optic lobes indicated. b,d, Z-projection of confocal images of Drosophila whole brain and c,e, Kenyon cell soma region in 5-day-old female (b,c) and male (d,e) in elav>GFP-LD(2.6) animals. Green punctae represent neuronal lipid droplets (LD). (b,d) Scale=100 μm; (c,e) scale=20 μm. f, Bodipy^493/503-stained LD in GT1–7 hypothalamic neurons. Scale=10 μm. g, Percentage of neurons with one or more LD in GT1–7 and N46 hypothalamic neurons in basal conditions. N=9033 GT1–7 and 28265 N46 cells from 6 independent experiments. h,i, Profile of fatty acids (FA) esterified in triglyceride (TG) in GT1–7 (N=3) and N46 neurons (N=4). C14:0= Myristic acid; C16:0= Palmitic acid; C16:1= Palmitoleic acid; C18:0= Stearic acid; C18:1= Oleic acid. Data are represented as mean ± SEM. j, LD stained with Bodipy^493/503 in GT1–7 or with LipidTox in NPY (k) and POMC (l) primary neurons, supplemented with oleate for 5h. Scales=10, 25, 25 μm. m, Percentage of neurons with one or more LD in GT1–7, NPY and POMC neurons treated with vehicle (BSA) or oleate. GT1–7, N=7735 BSA and 6193 oleate; NPY, N=58 BSA and 59 oleate; POMC, N=77 BSA and 105 oleate. n, Illustration of the arcuate nucleus (ARC) of the hypothalamus containing hunger-activated Neuropeptide Y (NPY)/Agouti-related peptide (AgRP) and hunger-inhibited Pro-opiomelanorcortin (POMC) neurons which regulate energy homeostasis. o-q, Transmission electron microscopy (TEM) of LD (red arrows) in mouse ARC neurons (yellow outline). Scale=10 μm (o), 1 μm (p-q). r, Percentage of neurons containing at least one LD in males (turquoise) (N=53 cells from 2 mice) and females (orange) (N=41 cells from 2 mice). See related data in Supplemental Figure 1.

Figure 2.. A network of genes regulates neuronal lipid droplets.

a, LD formation relies on enzymes including GPAT and AGPAT; LIPIN and DGAT/DIESL while LD hydrolysis is mediated by lipases including ATGL, HSL and MGL (Suppl Table 1). The recruitment and activity of lipases is regulated by ABHD5, G0S2, PLIN at the surface of LD and SEIPIN, a docking protein. b, Number of LD in GT1–7 neurons treated with vehicle (DMSO) or ATGListatin (24h), N=5. c, Total amount of FA esterified into TG in GT1–7 neurons treated with DMSO or ATGListatin (24h). d, Profile of FA esterified into TG in GT1–7 cells treated with control (DMSO, control data regraphed from Fig. 1g) or ATGListatin. C14:0, Myristic acid; C16:0, Palmitic acid; C16:1, Palmitoleic acid; C18:0, Stearic acid; C18:1, Oleic acid. N=3 independent experiments. e,f, Relative proportion of FA esterified into TG in GT1–7 neurons treated with DMSO or ATGListatin (24h). g, Number of LD in GT1–7 neurons incubated with oleate for 24h, or 24h after oleate withdrawal with or without ATGListatin, N=4–6. h, Number of LD in GT1–7 neurons (preloaded with oleate) treated with Forskolin (FSK) or FSK + ATGListatin (2.5h), N=5–7. i, Number of LD in GT1–7 neurons incubated with oleate and vehicle or DGAT1 inhibitor A-922500 (24h), N=6. j,l, Maximum Z-projections and k,m, quantification of neuronal LD (green punctae) in the Drosophila Kenyon cell soma region of elav>GFP-LD(3.4) in adult female (orange) and male (turquoise) flies with neuronal loss of dATGL (j,k) and dHSL (l,m). k,l, Scale=20 μm. n-s, Quantification of neuronal LD in the Kenyon cell soma region of adult elav>GFP-LD(3.4) females and males with neuron-specific loss of dSREBP, dAGPAT3, dDIESL dPLIN1, dPLIN2, dSEIPIN. (b,c,i) Student’s t-test, (h) Kruskall-Wallis, (e,f) multiple t-test, (g) two-way ANOVA with Sidak post-hoc test, (k,m-o,q-s) Two-way ANOVA with Tukey post-hoc test. (p) Mann-Whitney Test. ns indicates not significant; *p<0.05, **p<0.01; ****p<0.0001. Data are represented as mean ± SEM. See related data in Supplemental Figures 2 and 3.

Figure 3.. Neuronal lipid droplet regulation affects whole-body energy homeostasis in worms and flies.

a-l, Whole-body energy homeostasis in adult Drosophila males (turquoise) and females (orange) with pan-neuronal loss of genes that encode LD-associated proteins dHSL (a-c), dPLIN1 (d-f), dPLIN2 (g-i) and dDIESL (j-l). (a,d,g,j) Percent body fat in fed conditions. Mean percent body fat +/− SEM. (b,e,h,k) Magnitude of body fat loss from 0–12h post-fasting (early). (c,f,i,l) Magnitude of body fat loss from 12–24h post-fasting (late). Fat breakdown data expressed as the mean body fat loss over a given period post-fasting +/− coefficient of error. Two-way ANOVA: ns indicates not significant, *p<0.05, *** p<0.001, ****p<0.0001 from RNAi genotype interaction, # indicates control genotype interaction. m,n, Fluorescence microscopy of unc-119p::YFP neuronal expression (yellow) and Bodipy^558/568C12-stained LD (red) in neuronal RNAi sensitive worms fed with either empty vector (EV) or atgl-1 RNAi (ATGL- RNAi). Scale=100 μm. o, Quantification of Bodipy^558/568C12 fluorescence in the anterior gut of EV (n=28) vs. ATGL-RNAi (n=42) worms, N=3. p, Quantification of Oil Red O staining (ORO) in EV (n=26) vs. ATGL-RNAi (n=43) worms, N=3. q-u, ORO staining in EV and ATGL-RNAi fed or fasted worms n=92–103, N=6. Scale=100 μm. (o,p) Mann-Whitney, (u) Student’s t-test, data are represented as mean ± SEM. See related data in Supplemental Figure 3.

Figure 4.. Neuronal ATGL influences whole-body energy homeostasis in mammals.

a,d, Body weight, b,e, fat and lean mass, and c,f, cumulative food intake in 16-week-old male and female ARC^ATGLCRE and ARC^ATGLKO mice, N=24–28 males and 13 females. g-n, Energy expenditure (EE) and body temperature traces with corresponding quantifications. j,o, Fatty acid oxidation (FAOx), k,p, respiratory quotient (RQ), q-t, food intake, and satiety in ARC^ATGLCRE and ARC^ATGLKO males and females measured in metabolic cages during 24h at 21 °C or 24h at 4 °C. N=6–7 males and 8–9 females. Data are represented as mean ± SEM. (a-f) Student’s t-test, *p<0.05; (g-t) Two-way ANOVA: #p<0.05, ####p<0.0001, time interaction and *p<0.05, **p<0.01, genotype interaction, Sidak post-hoc. See related data in Supplemental Figure 4.

Figure 5.. ATGL function within AgRP neurons plays a conserved role in regulating whole-body energy homeostasis.

a,b, Body weight, b,e, cumulative food intake, c,f, fat and lean mass in 16- week-old male and female AgRP^ATGLCRE and AgRP^ATGLKO mice, N=14–21 males and 6–11 females. g-j, EE, k,m, RQ and l,n, FAOx in AgRP^ATGLCRE and AgRP^ATGLKO males and females in ad libitum (Fed) or fasted (16h) conditions (Fast). N=10–11 males and 6–11 females. o-r, Food intake and satiety in AgRP^ATGLCRE and AgRP^ATGLKO males and females over 24h at 21 °C or 24h at 4 °C. N=8–11 males and 6–10 females. Data are represented as mean ± SEM. (a-f) Student’s t-test, *p<0.05. (g-r) Two-way ANOVA: #p<0.05, ##p<0.01, ####p<0.0001, time interaction and *p<0.05, **p<0.01, genotype interaction, Sidak post-hoc. s, Whole-body fat breakdown 0–24h post-fasting in female (orange) and male (turquoise) flies with loss of dATGL in the adipokinetic hormone (Akh)-producing cells (APC). t, Fat breakdown 12–24h post-fasting in flies in which the APC were ablated via overexpression of proapoptotic gene reaper (rpr). u, Fat breakdown 0–12h post-fasting in Akh mutant flies (Akh^A). v, Fat breakdown 12–24h post-fasting in flies with APC-specific loss of dDIESL. w, Body fat in flies with APC-specific loss of dDGAT1. x, Fat breakdown 12–24h post-fasting in flies with APC-specific loss of dDGAT1. Body fat shown as mean +/− SEM. Fat breakdown data expressed as the mean percent body fat loss post-fasting +/− coefficient of error. Two-way ANOVA: ns indicates not significant, *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001, RNAi genotype interaction. See related data in Supplemental Figures 5 and 6.

Figure 6.. Profound and sex-specific lipid remodeling caused by neuronal loss of ATGL.

a, Differentially regulated lipid species in brains of Drosophila adult males and b, females with neuronal loss of dATGL. c, Differentially regulated cardiolipin species in Drosophila adult male brains with neuronal loss of dATGL; black boxes indicate significantly altered lipids and grey boxes indicate trends. d, Principal component analysis and e, volcano plot from LC-QTOF-based lipidomics in GT1–7 cells from the 188 annotated MS features obtained following MS data processing. In the volcano plot, the × axis represents the fold changes of MS signal intensities expressed as log2 for all the features in the ATGListatin group compared with control group. The y axis corresponds to the p values expressed as −log10. The color plots show the annotated lipid entities. N=5–6. f,g, Box plots of annotated unique lipids (expressed as log2) significantly discriminating 24h ATGListatin treatment from control group. Multiple t-test; N=5–6. PE= Diacylglycerophosphoethanolamines; PC= Diacylglycerophosphocholines; PEO= 1-alkyl,2-acylglycerophosphoethanolamines; CL= Cardiolipin; Cer= Ceramide; SM= Ceramide phosphocholines (sphingomyelins); GlcCer= Simple Glc series; S= Diacylglycerophosphoserines; CAR= Fatty-acyl-carnitines; PI= Diacylglycerophosphoinositols; PCO= 1-alkyl,2-acylglycerophosphocholines; LPCO= Monoalkylglycerophosphocholines; LPE= Monoacylglycerophosphoethanolamines; DG= Diacylglycerols; PC= Monoacylglycerophosphocholines; CE= Steryl esters; Chol derivates= Cholesterol and derivates; TG= Triacylglycerols. See related data in Supplemental Figures 7 and 8.

Figure 7.. Loss of ATGL is associated with mitochondrial defects and ER stress in hunger-activated neurons.

a, Mitochondrial number in Drosophila female (orange) and male (turquoise) adipokinetic hormone (Akh)-producing cells (APC) with APC-specific loss of dATGL. Mean +/− SEM; two-way ANOVA and Tukey post-hoc test. b, Electron microscopy of mitochondria in AgRP neurons identified by GFP immunostaining (*yellow). Scale=1 μm. c, Mitochondria number, d, length, and e, aspect ratio (length/width) in male AgRP^ATGLCRE (n= 2) vs AgRP^ATGLKO (n=2) mice. N= 26 CRE and 53 KO neurons. (d-e) Mean +/− SEM; student t-test. f, Relative metabolite levels in response to control and ATGListatin treatment (24h) in GT1–7 neurons. Significant species are annotated by an asterisk (increased red, decreased blue). Mean +/− SEM; multiple t-test; N=7–8. g, ER cisternae (*orange) in AgRP neurons. Scale=1 μm. h, Percentage of neurons with ER cisternae in AgRP^ATGLCRE and AgRP^ATGLKO males. Fisher’s exact test. i, Levels of a GFP-based ER stress reporter in Drosophila APC. GFP is produced only in contexts where Xbp1 is spliced in an IRE1-dependent manner. GFP levels in females (orange) and males (turquoise) with APC-specific loss of dATGL. Mean +/− SEM; two-way ANOVA, Tukey post-hoc test. j, Nascent protein synthesis in APC of male controls (grey) and males with APC-specific dATGL loss (turquoise). Mean +/− SEM; Student’s t-test. k, Akh protein levels in the APC of adult females (orange) and males (turquoise) with APC-specific loss of dATGL. Mean +/− SEM. Two-way ANOVA, Tukey post-hoc test. l, AgRP mRNA level in the ARC of AgRP^ATGLCRE vs AgRP^ATGLKO males. N=6–10. m,n, AgRP immunofluorescence in the ARC (m) and the PVN (n) from AgRP^ATGLCRE and AgRP^ATGLKO males and females (fold change from controls). Mean +/− SEM, (l) Student’s t-test, (m) one-way ANOVA, Sidak post-hoc test; N=4–5 males and 3–4 females. o-p, Spontaneous action potentials (sAP) of AgRP neurons in male and female AgRP^ATGLCRE vs AgRP^ATGLKO mice. N=13 vs 25 males; 11 vs 11 females. Kruskal-Wallis test * indicates p<0.05, ** indicates p<0.01, *** indicates p<0.001, **** indicates p<0.0001, ns: not significant. q, Graphical abstract. See related data in Supplemental Figure 9.