Antioxidant Role for Lipid Droplets in a Stem Cell Niche of Drosophila

Abstract

Stem cells reside in specialized microenvironments known as niches. During Drosophila development, glial cells provide a niche that sustains the proliferation of neural stem cells (neuroblasts) during starvation. We now find that the glial cell niche also preserves neuroblast proliferation under conditions of hypoxia and oxidative stress. Lipid droplets that form in niche glia during oxidative stress limit the levels of reactive oxygen species (ROS) and inhibit the oxidation of polyunsaturated fatty acids (PUFAs). These droplets protect glia and also neuroblasts from peroxidation chain reactions that can damage many types of macromolecules. The underlying antioxidant mechanism involves diverting PUFAs, including diet-derived linoleic acid, away from membranes to the core of lipid droplets, where they are less vulnerable to peroxidation. This study reveals an antioxidant role for lipid droplets that could be relevant in many different biological contexts.

Graphical abstract

Figure 1

Neuroblast Proliferation Is Spared during Environmental Hypoxia (A) Cartoon of Drosophila larva, showing the developing wing discs and CNS. Most proliferating cells correspond to neuroblasts (NBs) and ganglion mother cells (GMCs) within the CNS and to epithelial progenitors within the wing disc. (B) Larval development timeline (hours after larval hatching), depicting the three larval instars and the hypoxia (2.5% oxygen) regimen used. (C) In vitro EdU assay shows that cell proliferation in neuroblast lineages (CNS) is more resistant to hypoxia than that of epithelial progenitors (wing disc). After 22 hr of normoxic or hypoxic NR, tissues were dissected and incubated in vitro with EdU for 1 hr. (D) In vivo EdU assay indicates that cell proliferation in neuroblast lineages (CNS) is more resistant to hypoxia than that of epithelial progenitors (wing disc). The CNS and wing disc are shown at 73 hr (after 3 hr of normoxic NR with EdU) and also at 92 hr (after a further 19 hr of normoxic or hypoxic NR in the presence of EdU). CNS insets in (D) show high-magnification views of EdU⁺ cells, large neuroblasts (dotted circle) are already EdU⁺ at 73 hr and contribute to the final EdU volumes. Histograms in (C) and (D) depict the average volume of EdU⁺ cells incorporated per wing disc or per thoracic CNS. Scale bars, 50 μm, inset scale bars, 10 μm. In this and subsequent figures, histograms show the mean, scatterplots show individual data points, and error bars are 1 SD. The key shows that histogram bars are colored according to t tests (p < 0.05): significant decrease (red), significant increase (green), or no significant change (gray) from the control (black). See also Figure S1.

Figure 2

Hypoxia Induces Lipid Droplets in the CNS (A) Oil red O-stained neutral lipids in the CNS of a late-third instar larva (control genotype: w¹¹¹⁸) raised on a standard diet. (B) LipidTOX stained neutral lipids in the CNS of a late-third instar larva raised on a standard diet. Neutral lipids (red) are prominent in the central brain and ventral nerve cord but do not overlap with neuroblast lineages (green: nab-Gal4 > CD8::GFP). Scale bar, 50 μm. (C) Neutral lipids (LipidTOX, red) in the thoracic ganglion of the late third instar CNS accumulate in clusters of lipid droplets labeled with Lsd-2::YFP (green). Scale bars, 1 μm. (D) Lipid droplets (LipidTOX) in the CNS of a mid-third instar (70 hr) larva fed on a standard diet increase during NR normoxia and become more numerous during NR hypoxia. The lipid droplet (LD) content of the wing disc does not change significantly during hypoxia. It is expressed as percentage of volume within the thoracic region of interest of the ventral nerve cord (white boxes, upper row) or the entire wing disc (white outlines, lower row). Scale bars, 50 μm. (E) Lipid droplets increase ∼25-fold in normoxia and ∼80-fold in hypoxia in the CNS but do not significantly change in the wing disc. Note that fold change values in this figure are normalized to the 70 hr start points of Figure 2D. See also Figure S2.

Figure 3

Lipid Droplets Accumulate in Cortex and Subperineurial Glia (A) Transverse sections (dorsal up) of the thoracic CNS from late-third instar larvae raised on standard diet. Lipid droplets (LipidTOX) are observed nearby, but not within neuroblast lineages (labeled with nab > CD8::GFP). They localize to glial cells (repo > CD8::GFP) including the cortex glia (nrt^GMR37H03> CD8::GFP) and the subperineurial glia (moody > CD8::GFP). Scale bars, 10 μm. (B) Lipid droplets (LipidTOX) localize to glia (repo > Lsd-2::GFP) but not to neuroblast lineages (elav^C155> Lsd2::GFP). Strong Lsd2::GFP (GFP) signal localizes to the surface of glial lipid droplets but low-intensity signal in glia and neuroblast lineages may indicate localization to other organelles such as the ER (Fauny et al., 2005). Scale bars, 10 μm. (C and D) Glial-specific knockdown of Lsd-2 (repo > Lsd-2 RNAi) inhibits lipid droplet accumulation in late-third instar larvae raised on standard diet (control: repo > w¹¹¹⁸). Scale bars, 50 μm. (E) Diagram showing the distribution of lipid droplets (red) within cortex and subperineurial glia (green) surrounding a neuroblast lineage (blue), comprising a single neuroblast, a ganglion mother cell, and multiple neuronal progeny.

Figure 4

Neuroblast Proliferation during Hypoxia Requires Glial TAG Synthesis (A) TAG synthetic pathway showing the enzymes (bold, see text for details) knocked down by cell-type-specific RNAi in this study. Dietary fatty acids that are saturated (SFA), monounsaturated (MUFA), or polyunsaturated (PUFA) can be incorporated into the pathway at multiple steps, but, for clarity, single entry points are depicted. (B) Multi-isotope imaging mass spectrometry (MIMS) of glial lipid droplets induced by hypoxia in larvae raised on diets containing 1-¹³C-glucose, 1-¹³C acetate, or ¹³C-U-linoleate (PUFA). Hue saturation intensity (HSI) images show that high ¹³C/¹²C ratios are selectively detected in CNS lipid droplets (arrowhead shows one example). Scale bars, 5 μm. (C) Quantification of hypoxia-induced lipid droplets (fold change at endpoint versus repo > w¹¹¹⁸ control) with glial-specific gene knockdowns (repo>RNAi) of TAG synthetic genes. (D and E) Quantification of neuroblast proliferation (in vivo EdU assays), during normoxia or hypoxia with neuroblast lineage (nab-Gal4) or glial (repo-Gal4) -specific DGAT1 knockdown (> DGAT1 RNAi, control: > w¹¹¹⁸). Pooled data are represented as mean ± SD. See also Figure S3.

Figure 5

ROS Induce CNS Lipid Droplets (A) Hypoxia significantly increases ROS (measured by DHE oxidation) in many cells of the larval CNS including neuroblasts (red arrow marks one of the neuroblasts). Scale bars, 10 μm. (B) CNS lipid droplets increase (fold change relative to NR endpoint) in response to 22 hr of in vivo exposure to the following pro-oxidants: chronic hypoxia (2.5% oxygen), H₂O₂ (0.5% v/v), tBH (100 mM), ethanol (10% v/v), intermittent hypoxia (44 cycles of 10 min normoxia: 10 min anoxia) or FeCl₂ (100 mM). Note that 22 hr of osmotic stress (1.25 M NaCl) does not significantly increase lipid droplets. (C) Dietary supplementation with the antioxidant N-acetyl cysteine (100 mM) inhibits lipid droplet induction by NR or tBH (100 mM). (D) Glial-specific overexpression of Catalase (repo>Cat) or Superoxide Dismutase 2 (repo>SOD2) inhibits lipid droplet induction by hypoxia (2.5% oxygen) or tBH (100 mM) (control genotype: repo > w¹¹¹⁸). See also Figure S4.

Figure 6

PUFAs Redistribute from Phospholipids to TAGs during Oxidative Stress (A) Lipid droplets following NR, hypoxia (2.5% oxygen), or tBH (100 mM) are more abundant on a diet supplemented with 20 mM linoleate (C18:2) than with 20 mM stearate (C18:0). (B) Neutral loss scans indicate that palmitate (C16:0), stearate (C18:0), oleate (C18:1), and linoleate (C18:2) are the major fatty acids present in CNS TAGs following NR. Dietary 20 mM linoleate (C18:2) leads to a relative enrichment of this PUFA in the TAG pool. X values are m/z, and Y values are relative signal intensities, scaled to the largest C16:0 peak. (C) Positive ion scans indicate the major PC mass envelopes (gray boxes) in the CNSs of NR larvae. Dietary 20 mM linoleate (C18:2) leads to a relative decrease in PC 16:0/16:0 and a relative enrichment of PC species containing C18:2. X values are m/z and Y values are relative signal intensities, scaled to the PC 16:0/18:2 peak. (D) Total FA composition (by GC-MS) of larvae raised on 20 mM linoleate diet. Neither hypoxia (n = 3) nor 100 mM tBH (n = 3) lead to major changes, compared to NR (n = 3), in the relative abundance of the major fatty acids (top to bottom: C18:2, C18:1, C18:0, C16:1, C16:0, and C14:0). (E) Relative distribution of major fatty acids (C16:0, C18:1, and C18:2) in the TAG, PC, and PE pools of CNSs from larvae on 20 mM linoleate diet. Hypoxia (n = 3) and 100 mM tBH (n = 4) both lead to a strong decrease, compared to NR (n = 3), in the proportion of all three major fatty acids in PCs and PEs, with a correspondingly strong increase in their proportions in TAG. (F) Positive ion scans of PC 18:x/18:x from the CNSs of larvae on a 20 mM linoleate diet. The ratio of 18:2/18:2 to 18:0/18:2 is decreased by hypoxia (n = 3, mean = 4.03, SD = 0.43, p = 0.008) and tBH (n = 3, m = 4.84, SD = 0.3, p = 0.09) relative to the NR control (n = 3, m = 5.7, SD = 0.62). For each treatment group, the mean spectrum is normalized to the 18:0/18:2 peak (m/z 786) of NR, and the maximum/minimum peak heights at m/z 782 (brackets) indicate variation between biological replicates. (G) CNS lipid peroxidation following 100 mM tBH treatment of larvae raised on a 20 mM linoleate diet. Co-incubation with the PUFA peroxidation sensor (C11-BODIPY 581/591) and LipidTOX 633 shows that the ratio of oxidized: non-oxidized sensor is lower in CNS lipid droplets (white arrow) than in cell membranes. Scale bars, 10 μm. See also Figure S5.

Figure 7

Lipid Droplets Protect Dividing Neuroblasts from PUFA Peroxidation (A) Lipid peroxidation following low-dose (10 mM) tBH treatment for 20 hr of larvae raised on a 20 mM linoleate (C18:2) diet. Co-incubation with the PUFA peroxidation sensor (C11-BODIPY 581/591) and LipidTOX 633 shows an increase in the ratio of oxidized: non-oxidized sensor in neuroblast lineages (dotted outlines) and also in glia following the knockdown of lipid droplets (repo > Lsd-2 RNAi, control genotype: repo > w¹¹¹⁸). With repo > w¹¹¹⁸, there is a low peroxidation ratio in LipidTOX⁺ droplets (arrow), but, with repo > Lsd-2 RNAi, there is a high peroxidation ratio in rare abnormal LipidTOX⁺ puncta (arrow). The same two genotypes are used in (B)–(E). Scale bars in this and subsequent panels, 10 μm. (B) Glial lipid droplet knockdown increases 4-HNE protein adducts in many cells of the CNS, including neuroblasts (marked with anti-Miranda), neurons (marked with anti-Neurotactin), and glia (marked with anti-Repo). (C) Quantification of 4-HNE protein adducts in neuroblasts. Glial lipid droplet knockdown in the presence of 10 mM tBH (+ tBH), but not in NR controls (– tBH), leads to a significant increase in 4-HNE protein adducts. This is further increased when larvae are raised on a 20 mM linoleate diet. Mann-Whitney test: gray p > 0.1, green p < 0.001 relative to the appropriate genotype control. (D) ROS increase following glial lipid droplet knockdown. Oxidized DHE levels increase significantly throughout the CNS (histogram) with elevated levels in glia (blue arrows) and in neuroblasts (white arrows). (E) Neuroblast proliferation is significantly inhibited by 10 mM tBH if glial lipid droplets are knocked down. Inhibition is stronger when larvae are raised on a 20 mM linoleate diet and can be significantly rescued by dietary supplementation with 40 μg/ml AD4. In the control genotype, tBH and 20 mM dietary linoleate do not significantly decrease neuroblast proliferation. Pooled data are represented as mean ± SD. (F) PUFA protection model for the antioxidant role of lipid droplets. The neural stem cell niche (glia) and the neural stem cell (neuroblast) are depicted in the presence of oxidative stress, in a CNS that is wild-type (left) or RNAi knockdown for a glial lipid droplet gene (right). Oxidative stress stimulates the biogenesis of glial lipid droplets that protect vulnerable PUFAs in the TAG core from ROS-induced peroxidation. Peroxidation of omega-6 PUFAs produces 4-HNE, which, in turn, can generate more ROS, leading to PUFA chain reactions that damage membrane lipids and other macromolecules such as proteins. Following oxidative stress, the proportion of total PUFAs and other fatty acids increases in the TAG pool, relative to the phosphatidylcholine pool. This may correspond to bulk redistribution of lipids from membranes to lipid droplets, which would decrease the amount of membranes and thus minimize the fraction of PUFAs that are susceptible to ROS-induced peroxidation. The mechanism by which ROS and/or 4-HNE inhibit stem cell proliferation and the molecule(s) communicating between the niche and the stem cell are not yet clear (dotted arrows). See Discussion for details. See also Figure S6.

Figure S1

Hypoxia Inhibits Larval Food Intake, Related to Figure 1 Top panels show petri dishes containing colored food (yeast paste containing bromophenol blue) and their corresponding lids. Petri dishes contained mid-third instar larvae subject to normoxia (20.9% oxygen, left panels) or hypoxia (2.5% oxygen, right panels) for 2.5 hr. Bottom panels show high magnification views of larvae exposed to normoxia (left panels) or hypoxia (right panels) at the end of the experiments. Under normoxia, larvae remain in the food and ingested food is clearly visible in the gut. In contrast, hypoxia induces larvae to crawl out of the food and to ingest very little.

Figure S2

CNS Lipid Droplets Accumulate and the Neuroblast-Tracheal Distance Increases during Normal Development, Related to Figure 2 (A) LipidTOX staining of the CNS at early (48 hr), mid (72 hr) or late (96 hr) third instar showing a progressive lipid droplet accumulation during development. Larvae were raised on standard diet in ambient normoxia. Scale bar: 50 μm. (B and C) The CNS tracheal network and neuroblasts are shown at early (B: 48 hr) and late (C: 96 hr) third instar. Tracheal-specific GFP expression (btl > CD8::GFP, green) and the neuroblast marker Miranda (red) are shown. The main tracheal network is in the neuropil, whereas neuroblasts are in the cortex. Scale bars: 50 μm. (D) The tracheal-neuroblast distance, measured from confocal z-plane reconstructions of the ventral nerve cord (white boxes in B and C). Between early and late third instar, a significant ∼1.5 fold increase is observed in the average distance (yellow arrows) between ventral neuroblasts and the tracheal network, although there is no significant change in lateral regions (blue arrows).

Figure S3

Metabolic Labeling of Glial Lipid Droplets Using MIMS, Related to Figure 4 (A) Stitched composite transverse cross section of the thoracic region of an unlabeled CNS (dorsal up) visualized with multi-isotope imaging mass spectrometry (MIMS) at mass 26 (¹²C¹⁴N-) to reveal morphology. Scale bar: 50 μm. (B and C) Cross section of the optic lobe of an unlabeled CNS. Images show a single superficial glial cell containing multiple lipid droplets, recognized by characteristic signals of low nitrogen (B, ¹²C¹⁴N- channel) and high carbon (C, ¹²C channel). (D and E) Quantitations of Hue Saturation Intensity (HSI) images of sectioned lipid droplet profiles. Graphs show ¹³C/¹²C ratio (fold change above background) versus distance (μm) for individual droplet profiles incorporating dietary ¹³C-glucose (D) or ¹³C-acetate (E). MIMS analysis of many lipid droplets gives diameters of 1-2 μm, in agreement with confocal analysis, but apparently smaller droplets likely correspond to ‘glancing’ cross sections that include only the periphery of a droplet. (F) Corresponding ¹³C/¹²C ratio HSI and ¹²C¹⁴N images of ¹³C-acetate labeled lipid droplets. Scale bars: 1 μm. (G) An example of the ¹³C/¹²C ratio (fold change above background) versus distance (μm) profile for an individual lipid droplet with characteristic low nitrogen content (¹²C¹⁴N) and a diameter of ∼1.5 μm. (H) GC-MS FAME analysis of major CNS fatty acids incorporating ¹³C following metabolic labeling with dietary U-¹³C-linoleic acid. After NR, hypoxia or tBH (100 mM) stress, ∼80% of ¹³C labeled fatty acids correspond to linoleate (C18:2, n = 3 for each condition, error bars show s.d.). Thus, in the corresponding MIMS analysis (Figure 4B), the majority of ¹³C label remains as linoleate not other fatty acids.

Figure S4

Hypoxic Induction of Lipid Droplets Does Not Require Neuroblast Proliferation or HIF Signaling, Related to Figure 5 (A) Pros::YFP misexpression in neuroblast lineages inhibits proliferation. In vivo EdU incorporation during normoxia or hypoxia is decreased in CNSs from NB > Pros::YFP (elav^GMR71C07, tubG80^ts> Pros::YFP) but not NB > control (elav^GMR71C07, tubG80^ts> ) larvae. Embryos were raised at 18°C and first instar larvae were then transferred to 29°C to alleviate Gal80 repression. Scale bars: 10 μm. (B) Blocking neuroblast proliferation does not inhibit the hypoxic induction of lipid droplets. Genotypes as in A. Scale bars: 10 μm. (C) Expression of a UAS-ODD::GFP reporter for stabilization of the oxygen-dependent degradation domain (ODD) of Hypoxia Inducible Factor 1 (HIF1). A 2.5 hr period of hypoxia (2.5% oxygen) significantly increases ODD::GFP intensity in the tracheal system (btl > ODD::GFP) but very little increase is observed in glia (repo > ODD::GFP) and no significant change is observed in neuroblast lineages (elav^GMR71C07> ODD::GFP). Scale bars: Tracheal system, 20 μm; Glia and Neuroblast lineages, 50 μm. (D) Sima (HIF1) is not required for hypoxia induced CNS lipid droplets. Histogram shows comparable lipid droplet increases during hypoxia for heterozygous (sima⁰⁷⁶⁰⁷/TM6, Tb) and homozygous (sima⁰⁷⁶⁰⁷/sima⁰⁷⁶⁰⁷) larvae. Scale bars: 10 μm. (E) Distribution of ROS in the CNS during normal development. ROS (oxidized DHE) were detected throughout the CNS, but at higher levels in neuroblasts (arrow), of a late third instar raised on standard diet in ambient normoxia. Neuroblast lineages were marked with GFP (ase > CD8::GFP). Scale bars: 10 μm.

Figure S5

Iron Stress Selectively Depletes Linoleate from PCs, Related to Figure 6 (A) In vitro induction of neutral lipid puncta after 24 hr culture of the CNS in sugar-free saline. Note that the in vitro LipidTOX puncta are smaller than in vivo CNS lipid droplets. Scale bars: 10 μm. (B) In vitro induction of lipid droplets is inhibited by glial-specific knockdown of PLD, Lipin, DGAT1 or Lsd-2 (repo>RNAi, control: repo > w¹¹¹⁸). (C) In vivo induction of CNS lipid droplets by 100 mM FeCl₂ is inhibited by glial-specific knockdown of PLD, Lipin and DGAT1 (repo>RNAi, control: repo > w¹¹¹⁸). (D) Tandem mass spectrometry of CNS phosphatidylcholines (PCs) from larvae subject to FeCl₂ stress. Three major PC mass envelopes are shown at low (left) and high (right) resolution. High-resolution images compare the relative distributions of species containing different numbers of fatty acyl carbon-carbon double bonds under control (NR) and 100 mM FeCl₂ stress. Two independent spectra for each condition are shown and the signal intensity (y axis) is normalized for each mass envelope to the species with 1 fatty acyl double bond. FeCl₂ stress strongly decreases the relative abundance of PCs with at least one C18 chain (34:x and 36:x mass envelopes) and at least two C-C double bonds (red arrows). (E) Overlaid neutral loss scans from tandem mass spectrometry indicate that, in both control and FeCl₂ conditions, CNS TAGs contain fatty acids that are saturated (18:0, stearic acid), monounsaturated (18:1, oleic acid) and polyunsaturated (18:2, linoleic acid). Y values correspond to relative signal intensities scaled to the largest C18:1 peak. (F) Tandem mass spectrometry shows that TAG composition does not change substantially between NR controls (n = 3), hypoxia (2.5% oxygen, n = 3) or tBH (100 mM, n = 4). Larvae were raised on the linoleate (C18:2) rich diet, and the relative abundance of the following TAG species was calculated: 16:1/16:1/16:0 (m/z 820), 16:1/16:0/16:0 (822), 16:0/16:0/16:0 (824), 18:2/16:1/16:0 (846), 18:2/16:0/16:0 (848), 18:3/18:2/16:0 and 18:2/18:2/16:1 (870), 18:2/18:2/16:0 (872), 18:1/18:2/16:0 (874), 18:1/18:1/16:0 and 18:2/18:0/16:0 (876), 18:1/18:0/16:0 (878), 18:2/18:2/18:2 (896), 18:2/18:2/18:1 (898), 18:2/18:1/18:1 and 18:2/18:2/18:0 (900). The sn1 to sn3 order of fatty acyl chains was not analyzed.

Figure S6

Exogenous 4-HNE Increases ROS and Inhibits Neuroblast Proliferation, Related to Figure 7 (A) In vitro incubation with 4-HNE (100 μM for 1 hr) generates 4-HNE protein adducts in cells throughout the CNS, including neuroblasts (marked with Miranda, red). Scale bars: 10 μm. (B) In vitro incubation with 4-HNE (100 μM for 30 min) increases ROS (DHE oxidation) in the CNS, particularly in neuroblasts (arrows). Scale bars: 10 μm. (C) In vitro incubation with 4-HNE (100 μM for 2 hr) decreases neuroblast proliferation (EdU incorporation into neuroblast lineages). Main panels show confocal projections, insets show single confocal sections at a higher magnification. Scale bars: 10 μm. (D) Quantification of 4-HNE inhibition of neuroblast proliferation and rescue by the antioxidant AD4 (40 μg/ml). Conditions as in C. (E) In vivo neuroblast proliferation is inhibited by intermittent hypoxia, weakly in a control genotype (repo > w¹¹¹⁸) and more strongly when glial lipid droplets are knocked down (repo > Lsd-2 RNAi). Intermittent hypoxia corresponds to 44 cycles of 5 min anoxia, 25 min normoxia.