Senescent glia link mitochondrial dysfunction and lipid accumulation


Fly stocks and maintenance

Flies were raised at 25 °C and 60% relative humidity on standard cornmeal fly food under a 12–12 h light–dark cycle. All experiments were performed with male flies to minimize biological variance, as male and female flies age at notably different rates. Flies were transferred to fresh food vials every 48 h, housed in cohorts of 20 and randomly assigned to experimental conditions. For geneSwitch (inducible GAL4-UAS)41 experiments, food was prepared with either 100 μl of RU-486 (4 mg ml−1 in 100% EtOH; Sigma-Aldrich, M8046-1G) or vehicle (100% EtOH), pipetted onto food vials and allowed to dry for 24 h. See Supplementary Table 1 in the Supplementary Information for genotypes and stock information.

For neuronally expressed UAS-RNAi experiments, flies were collected onto RU-486 food at adult eclosion (0 days of age) and reared at 29 °C.

For AP1 blockade experiments, flies were collected onto vehicle or RU-486 food at adult eclosion (0 days of age). Animals were maintained on RU-486 continuously (7 days per week) or intermittently (3 or 1 day per week) by flipping flies to vehicle food. All experiments and sample collection for 1 day per week RU-486 treated flies were performed with animals on vehicle food, specifically at 6 days after last RU-486 feeding to ensure geneSwitch termination41. Controls were selected on the basis of assay. For BODIPY experiments, genotype-matched vehicle-fed animals were used as controls, as GFP interferes with BODIPY detection.

Behavioural assays

To measure survival, the number of dead and/or censored flies was recorded every 2 days after flipping flies to fresh food; flies were housed in vials of 20 each, with a minimum of 100 flies per genotype and experiment, repeated a minimum of two times. To measure climbing, flies were single-housed in empty vials and allowed to acclimate for 30 min. Climbing was measured by tapping flies to the bottom of the vial then recording height climbed after 30 s over three trials with a 5 min testing interval. Averaged climbing height was determined in Fiji. Data are expressed as a percentage of the maximum vial height (8 cm). Heat shock assessment was performed as described in ref. 61. In brief, flies were transferred to clear plastic 13 ml vials, and each vial contained 15 flies. Vials plus flies were transferred to a water bath for 1 h at 38.5 °C for stress. The flies were then transferred to fresh food and allowed to recover overnight at 25 °C then the percentage of flies alive versus dead were recorded per vial. Oxidative stress (H2O2 feeding): adult flies were single-housed and loaded in the Drosophila Activity Monitoring system on either 5% sucrose-agar or 1% H2O2 sucrose-agar. Activity was recorded for 10 days then analysed in the R environment using the rethomics package62 to determine time of animal death and generate survival curves.

FACS-based isolation of AP1+ glia, AP1neg glia and neurons for bulk RNA-seq or lipidomic analysis

All work was performed in RNAse-free conditions. To create a cell suspension for FACS-based sorting, adult fly brains (n = 20 brains per replicate for RNA-seq; n = 40 brains per replicate for lipidomic analysis) were rapidly dissected in Schneider’s medium with 45 μM actinomycin D and stored on ice until dissections were complete. Brains were then washed in cold phosphate-buffered saline (PBS) (3×). A single cell suspension was achieved by enzymatic and physical dissociation as follows: whole brains were incubated in dissociation buffer (300 μl of activated papain, Worthington PAP2 LK003178 and 4.1 μl liberase, Roche 5401119001) at 25 °C at 1,000 rpm on a shaker for a total of 20 min. During incubation, at 5 and 10 min, tissue was gently homogenized by pipetting. At 15 min, the entire homogenate was passed through a 25G 5/8 needle (7×). At 20 min, enzymatic activity was halted by the addition of cold Schneider’s medium. Cells were then strained (35 μM filter), pelleted (800g, 7 min) and resuspended in cold Schneider’s medium with actinomycin D and 2.5 μl of RNAse inhibitor (Takara Recombinant RNase Inhibitor, catalogue no. 2313A). Cells were resuspended in 250 μl, counterstained with 5 μM 4,6-diamidino-2-phenylindole (DAPI) and 50 nM syto60 (nuclear marker; ThermoFisher, catalogue no. S11342) and sorted by the Penn Cytomics and Cell Sorting Facility using a BD FACS Aria II SORP (100 μM nozzle; purity). Dead cells were excluded through DAPI uptake. Doublets were excluded through FSC-H by FSC-W and SSC-H by SSC-W parameters. Nucleated cells were included by syto60. Glia were identified by GFP, and neurons were GFP negative. AP1 activity was identified by dsRed. The gating strategy is shown in Extended Data Fig. 1c.

For bulk RNA-seq, 500 neurons, 500 AP1+ glia and 500 AP1neg glia were collected per replicate, with four replicates per cell type. For lipidomic profiling, 100,000 neurons, 100,000 AP1neg glia and 35,000 AP1+ glia were collected per replicate, with 5–6 replicates per cell type. Cells were immediately frozen at −80 °C until further processing. Total processing of tissue and cell isolation took roughly 3 h. Data from the sort were analysed using FlowJo v.10.8.1. To generate cell and DNA content plots, cell populations with different N were overlaid using absolute cell counts normalized to mode (to the peak height at mode of the distribution).

For immunostained FACS-isolated cells, following dissociation and resuspension as above, cells were fixed in 4% paraformaldehyde for 15 min at room temperature. Cells were washed then resuspended in 5% normal goat serum (NGS) for 5 min on ice. Cells were next incubated in primary antibody (1:20 mouse anti-γH2Av, DSHB UNC93-5.2.1) for 30 min at room temperature, washed and incubated in secondary antibody (1:200 goat antimouse AlexaFluor 647, ThermoFisher Scientific, catalogue no. A-21235) for 30 min at room temperature. Gating parameters were as above. Cells were immediately analysed using a BD FACS Aria II SORP as above.

Bulk RNA-seq and analysis

For sorted cells, RNA isolation, library preparation (SMART-Seq v.4) and RNA-seq (Illumina 2 × 150 40 million paired-end reads per sample; 20 million each direction) were performed by Admera Health. For whole brains, roughly 10–12 brains were dissected per condition. Total RNA was extracted using the Zymo RNA clean & concentrator−5 kit (Zymo, R1013), using their RNA clean-up from the aqueous phase after Trizol/chloroform extraction protocol plus on-column DNaseI treatment. RNA amount was measured by nanodrop, and integrity was validated by an Agilent 2100 Bioanalyzer using an RNA nano chip. The RNA-seq libraries (TruSeq stranded with Poly-A selection) and sequencing (Illumina NovaSeq S4 with 40 million paired-end reads; 2 × 150 bp) were performed by Admera Health. Four biological replicates were generated for each sample type, experimental timepoint, condition and genotype.

Demultiplexed reads passing the quality control filter (Q > 30) were obtained from BaseSpace then merged across sequencing lanes for each sample, with roughly 20 million reads total per sample. Paired-end reads were aligned to the fly genome using HISAT2 (v.2.1.0)63. The HISAT2 index was built from FlyBase’s Drosophila melanogaster reference genome r6.17. Alignment sorted BAM files (samtools v.15) for each sample were merged across sequencing runs (picard)64. Reads that uniquely aligned to exonic regions were counted with HTSeq (v.0.9.1)65 with the union setting to produce a count matrix for differential expression analysis using the DESeq2 (ref. 66) package in the R environment. The design model formula was ‘~group’ if there were two or more key variables involved (that is, genotype and age) or design model formula was the single key variable (that is, genotype). Pairwise comparisons were made between samples (that is, ‘contrast=c(’group’)’), with an alpha cut-off of 0.05 with lfcShrink() applied. Gene ontology and Reactome pathway enrichment were performed with tools at, using all expressed genes as background (n roughly 15,694). Refer to indicated Supplementary Data for differentially expressed genes between samples and/or groups across experiments.

Unbiased lipidomics for multiple reaction monitoring profiling and analysis

For lipidomic profiling cells were FACS-isolated as above. Brains were rapidly dissected in PBS, pelleted by centrifugation and excess PBS was removed for freezing at −80 °C until further processing (n = 8 brains per replicate; 5–6 replicates per genotype and/or age). Lipid extracts from FACS-sorted cells and whole-brain samples were prepared using a slightly modified Bligh & Dyer extraction procedure67. In brief, the frozen samples were thawed for 10 min at room temperature, and 200 μl of ultrapure water was added to promote lysis, followed by 450 μl of methanol and 250 μl of high-performance liquid chromatography-grade chloroform. Samples were vortexed for 10 s, resulting in a one-phase solution, and incubated at 4 °C for 15 min. Next, 250 μl of ultrapure water and 250 μl of chloroform were added, creating a biphasic solution. The samples were centrifuged at 14,000g for 10 min, which resulted in three phases in the tubes. The bottom organic phase containing the lipids was transferred to new tubes, then evaporated in a vacuum concentrator leaving behind the dried lipid extracts.

Multiple reaction monitoring profiling of the extracted lipids was performed as described previously68. The dried lipid extracts were dissolved in 100 μl of methanol:chloroform (3:1 v/v) to make lipid stock solutions. The lipids were further diluted threefold in the injection solvent 7:3 methanol:acetonitrile with 10 mM ammonium formate immediately before analysis. The injection solvent alone without any lipids was used as the ‘blank’ sample.

Mass spectrometry data were acquired for 3 min by flow injection (that is, no chromatographic separation). Briefly, 8 μl of diluted lipid extract stock solution delivered to the jet stream technology ion source (AJS) source of an Agilent 6495C Triple Quadrupole mass spectrometer. Multiple reaction monitoring methods were organized into 11 methods on the basis of the ten main lipid classes based on the LipidMaps database; see Extended Data Fig. 9d for total n of lipids screened and Supplementary Data 17 for individual species. TAGs were divided into two separate methods on the basis of fatty acid neutral loss residues.

Statistical analysis was performed using the EdgeR package69. EdgeR uses a generalized linear model to identify differentially expressed lipids. The generalized linear model is based on the negative binomial distribution that incorporates the blank with a dispersion term using the common dispersion method70. This allows it to model the technical and biological variability. This method was previously described in detail in ref. 68. Significant lipids were chosen on the basis of a false discovery rate value <0.1 (ref. 71).

Whole-mount brain immunofluorescence

A standard protocol was used for fixation and staining. In brief, adult fly brains were dissected in cold PBS and fixed in 4% paraformaldehyde (v/v) for 50 min at room temperature. Brains were washed and permeabilized in PBS-0.1% Triton-X (PBST; 3×, 10 min). Samples were blocked in PBST-5% NGS at room temperature for 1 h, then incubated for 24–48 h at 4 °C with 1° antibody (1:25 mouse anti-repo, DSHB 8D12; 1:20 rat anti-elav, DSHB 7E8A10). Brains were washed in PBST then incubated with fluorescently conjugated 2° antibody for 1 h at room temperature. For AP1 activity (all genotypes containing TRE-dsRed) and tdTomato, endogenous fluorophore luminescence was measured without additional antibody staining. Brains were counterstained with Hoechst (0.10 mg ml−1 in PBS) for 15 min, cleared in mounting media (20 mM Tris pH 8.0, 0.5% N-propyl gallate, 80% glycerol, PBS), mounted in mounting media and cover slipped. Brains were imaged by confocal microscopy (Leica DM 6000 CS) with identical laser power and gain settings across experiments. Images were acquired throughout the full brain at 2 μM steps at 1,024 × 1,024 resolution by ×20 (dry) or ×63 (oil) objectives.

For BODIPY, brains were dissected and fixed as above then incubated for 24–48 h at room temperature in 1:250 dilution of 10 mg ml−1 BODIPY 493/503 (Invitrogen D3922) prepared in NGS. Brains were washed in PBST, counterstained with Hoechst and prepared for imaging as above.

For DHE, fly brains were rapidly dissected in cold Schneider’s medium, incubated in 60 μM DHE (ThermoFisher, catalogue no. D11347) for 7 min at room temperature shaking. Brains were washed in Schneider’s medium (2×, 5 min) then PBS (1×, 5 min), mounted in mounting media and imaged immediately (excitation 488 nm, emission 515–656 nm).

Fiji v.2.0 was used for analysing all images. For TRE-dsRed quantification, dsRed was measured in Fiji as raw integrated density in scaled images of the z stacked brain. For BODIPY 493/503 quantification, background was first subtracted from scaled images of the z stacked brains. Automatic thresholding was applied and Analyze Particles (Analyze>Analyze Particles) was used to determine the number, average size and total area of BODIPY+ droplets.

Whole-mount brain immunohistochemistry for SA-β-Gal activity

A protocol was adapted72 for staining in fixed dissected whole Drosophila brains. In brief, adult fly brains were dissected in cold PBS and fixed in 2% paraformaldehyde and 0.2% glutaraldehyde (v/v) for 30 min at room temperature. Brains were washed in PBS (3×, 5 min) then incubated in 150 μl of X-Gal staining solution (40 mM citric acid phosphate buffer, 5 mM potassium hexanocyanoferrate(II) trihydrate, 5 mM potassium hexanocyanoferrate(III), 150 mM NaCl, 2 mM MgCl2-6H2O, 2.44 mM x-Gal) at 37 °C in the dark shaking (300 rpm) for a predetermined time on the basis of genotype (roughly 12–24 h). Brains were washed in PBS (3×, 5 min) and cleared in mounting media as above overnight. Brains were imaged on APO16 microscope and staining was quantified in Fiji v.2.0 by converting to a red, green and blue stack and measuring area and median value in the red channel only. Inverted density was calculated by subtracting median grey value from 255 and normalized to controls processed in parallel to account for variability across experiments.

Western immunoblot

Fly brains were rapidly dissected in cold PBS (n = 8 brains per biological replicate), then homogenized in 5 μl of sample buffer per brain (1× Laemmli Buffer (Bio-Rad, catalogue no. 161-0737), 1× cOmplete mini EDTA-free protease inhibitor cocktail, 1 mM phenylmethylsulfonyl fluoride (Sigma, catalogue no. P7626), 50 μl β-mercaptoethanol (BME): Sigma, catalogue no. M6250). Samples were denatured (98 °C for 3 min) before loading onto 4–12% Bis-Tris gel. Volume equivalent of one brain per sample was run in 1× MES buffer, transferred to 0.45 μM nitrocellulose membrane overnight by electrophoresis. Membranes were stained by Ponceau S to confirm transfer. Membranes were blocked in 3% bovine serum albumin in 1× Tris-buffered saline, 0.1% Tween 20 detergent, incubated in primary antibody overnight at 4 °C (1:200 mouse anti-γH2Av, DSHB UNC93-5.2.1; 1:2,000 mouse anti-tubulin, DSHB AA4.3). Blots were incubated with 1:5,000 dilution of species-appropriate HRP-conjugated secondary antibody for 1 h at room temperature, then detected by ECL (Cytiva (formerly GE Healthcare Life Sciences), catalogue no. RPN2232) using an Amersham Imager 600. Quantification was performed in Fiji by region of interest. Sample protein was normalized to the loading control alpha tubulin. Mammalian cells were lysed in modified RIPA buffer and blotted using standard techniques as previously described73 using an antibody to JUN (1:1,000; Cell Signaling Technology, catalogue no. 9165).

Cell proliferation by EdU labelling

Flies were maintained on 0.2 mM EdU food from eclosion through dissection. EdU staining was performed according to the manufacturer’s protocol (Click-iT EdU Imaging Kit; ThermoFisher, catalogue no. mp10338). In brief, brains were dissected and fixed as above for immunohistochemistry. Following permeabilization, brains were incubated in Click-iT reaction mixture overnight at 4 °C. Brains were washed, counterstained with Hoechst, cleared, mounted and imaged as above.

Mitochondrial function assay

The ratio of ATP/cytotoxicity was determined using the Promega Mitochondrial ToxGlo Assay (G8001). The manufacturer’s instructions were followed. Assay lysates consisted of individual dissected fly brains (n = 3–4 brains per genotype or condition) across a minimum of three experiments. Samples were always normalized to parallel processed controls.


A protocol was adapted for measuring mtDNA in adult fly heads34. For head collection, whole flies were anaesthetized by CO2, frozen by submersion in liquid nitrogen, then vortexed to separate heads from bodies. Total cellular DNA was extracted by homogenizing five heads (per replicate) in 30 µl working solution (10 mM Tris-HCl, pH 8.0, 1 mM EDTA, 0.1% Triton-X-100 and 10 μg ml−1 protease K). Samples were then incubated at 37 °C for 60 min, and followed by inactivation of protease K at 95 °C for 10 min. Head cuticles were pelleted by centrifuging samples at 12,000g for 10 min at room temperature. Supernatants were transferred into a new tube, before measuring DNA concentration by Nanodrop. mtDNA was quantified using nuclear DNA (GAPDH) as control in real-time quantitative PCR (qPCR). Primer sequences: mtDNA-F1, GAATTAGGACATCCTGGAGC and mtDNA-R1, GCACTAATCAATTTCCAAATCC; GAPDH-F1, GACGAAATCAAGGCTAAGGTCG and GAPDH-R1, AATGGGTGTCGCTGAAGAAGTC.

Real-time qPCR

Total RNA was isolated from fly brains or heads (n = 8–20 per replicate) by RNeasy Mini Kit (Qiagen, catalogue no. 74104), with on-column removal of genomic DNA (Qiagen, catalogue no. 79254). Complementary DNA (cDNA) was prepared from total RNA (Applied Biosystems, catalogue no. 4368814) then quantified by Qubit ssDNA Assay (Invitrogen, catalogue no. Q10212). Real-time qPCR reactions were set up using Fast SYBR Green reagents (ThermoFisher, catalogue no. 4385612) in 384-well plates with 20 ng of cDNA per reaction and analysed on a ViiA 7 Real-Time PCR System (Applied Biosystems). Relative expression was determined using the ∆∆CT method. For each sample, mean CT values were determined from 2–3 technical replicates. ∆CT was determined relative to the housekeeping gene, β-tubulin. ∆∆CT was then calculated as fold change relative to the control group. Real-time qPCR primers were sourced from FlyPrimerBank74 or previous publications, BLASTd against the Drosophila genome for specificity and optimized by serial dilution curve and melt curve analysis. See Supplementary Table 2 for primer sequences.

Mammalian cell culture

IMR90 primary human fibroblasts (ATCC CCL-186) were grown at 37 °C, 3.5% O2, 5% CO2, in Dulbecco’s modified Eagle’s medium (Gibco, catalogue no. 10313-121) with 10% FBS (Corning, catalogue no. 35-011-CV), 1% penicillin–streptomycin (Gibco, catalogue no. 15140-122) and 2 mM glutamine (Gibco, catalogue no. 25030-081). Cultures were checked routinely for mycoplasma contamination. Irradiation senescence was induced by 20 Gray of X-ray irradiation of 20–30% confluent cells. Cells were split after returning to confluence 3 days after irradiation. On days 4 and 7 after irradiation, cells were transfected with a pool of four small-interfering RNA against JUN (Dharmacon siGENOME) or non-targeting control (siNTC #3, Dharmacon siGENOME) to a final concentration of 100 nM with 0.8% Dharmafect reagent following the manufacturer’s protocol. Medium was changed 18–20 h after each transfection. Medium from days 8–10 after irradiation, or from normal proliferating IMR90 cells cultured in parallel, was collected, centrifuged at 500g for 3 min to remove whole cells and large debris, then added to 20–30% confluent proliferating IMR90 cells plated on 96-well imaging plates (Perkin Elmer, catalogue no. 6055302) for 48 h. Cells were fixed in 10% neutral buffered formalin (Epredia, catalogue no. 9400-1) and stained with 500 µg ml−1 DAPI and 5 µg ml−1 BODIPY 493/503 (Cayman, catalogue no. 25892) in PBS. Automated imaging of cells was done on a Nikon Ti2 microscope and images were analysed in NIS Elements.

Statistical analysis

Statistical analysis and data visualization were performed in the R Environment using RStudio with base R and packages as indicated including with tidyverse (dplyr, ggplot2), ggrepel, cowplot, ggsurvplot. No statistical method was used to predetermine sample sizes; standard sample sizes for Drosophila, pooled across two or three independent experiments, were used. See the Source Data files for data and statistical reporting corresponding to each main and Extended Data figure.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Source link