Tau is required for glial lipid droplet formation and resistance to neuronal oxidative stress

Abstract

The accumulation of reactive oxygen species (ROS) is a common feature of tauopathies, defined by Tau accumulations in neurons and glia. High ROS in neurons causes lipid production and the export of toxic peroxidated lipids (LPOs). Glia uptake these LPOs and incorporate them into lipid droplets (LDs) for storage and catabolism. We found that overexpressing Tau in glia disrupts LDs in flies and rat neuron-astrocyte co-cultures, sensitizing the glia to toxic, neuronal LPOs. Using a new fly tau loss-of-function allele and RNA-mediated interference, we found that endogenous Tau is required for glial LD formation and protection against neuronal LPOs. Similarly, endogenous Tau is required in rat astrocytes and human oligodendrocyte-like cells for LD formation and the breakdown of LPOs. Behaviorally, flies lacking glial Tau have decreased lifespans and motor defects that are rescuable by administering the antioxidant N-acetylcysteine amide. Overall, this work provides insights into the important role that Tau has in glia to mitigate ROS in the brain.

Extended Data Fig. 1 |. Glial lipid droplet pathway and characterization of lipid droplets in the fly retina.

(a) Elevating ROS levels in neurons causes an SREBP-dependent production of peroxidated lipids (LPOs). These lipids are exported by ABC-lipid transporters, transferred to glia via apolipoproteins (for example APOE), and endocytosed by the glia. Within the glia, the LPO are integrated into lipid droplets (LDs) via the endoplasmic reticulum (ER). LDs promote lipid β-oxidation by the mitochondria, producing energy. Thus, glial LDs remove the toxic, LPO and protect both cells from ROS-induced damage. A growing list of established and predicted AD risk genes (red text) fit into this pathway, with most genes being required for proper glial LD formation and protection against neuronal ROS^,–. Created with Biorender.com. (b) The fly retina is composed of ~800 highly structured ommatidia. Diagram: a single ommatidium with seven photoreceptor neurons (blue) surrounded by pigment glia (green). Each photoreceptor contains a lipid-rich rhabdomere (hashed circle) which can be visualized with lipid stains. (c) Elevating ROS in photoreceptors causes LDs to accumulate in surrounding glia. LDs are visualized using neutral lipid dyes, Nile red and BODIPY 493/503. (d) TEM image of the fly retina, including 4 ommatidia. Each ommatidium contains seven visible photoreceptors with the rhabdomeres (R) clustered at the center and the cell body (CB) extending distally. Photoreceptors are surrounded by retinal glia (green). (e) The genetic manipulation used to investigate regulators of glial LD formation in the fly retina. UAS-transgenes are expressed either in mature photoreceptors, using Rh1-GAL4, or in retinal glia, using 54C-GAL4. Rh:ND42 RNAi is used to induce ROS only in the photoreceptors, causing glial LDs. (f) LDs can be induced in photoreceptors either by causing neuronal ROS (Rh:ND42 RNAi) or by overexpressing the activated SREBP transcription factor (nSyb-GAL4 > UAS-SREBP). LDs produced by Rh:ND42 RNAi can be prevented by feeding flies the antioxidant NACA, consistent with. LDs formed by neuronal SREBP upregulation in non-ROS conditions are not impacted by NACA. Lipid stain = Nile red. Data presented as mean ± SD. A datapoint represents the average LD number per ommatidium per animal (n). n = 7–10 (detailed in Supporting Data 1). Statistics: 1-way ANOVA with Tukey’s multiple comparisons test. P-values = not significant (n.s.) >0.05, **=0.004, ***=0.0002, ****≤0.0001. (g) LDs within retinas co-stain for both Nile red and BODIPY 493/503, validating that these puncta are LDs. (f-g) Retinas from 2d flies were fixed, dissected, and stained with neutral lipid dyes. Arrow = glial LD.

Extended Data Fig. 2 |. Additional data for hTau overexpression in flies.

(a-b) UAS transgenes were expressed in photoreceptors using Rh1-GAL4. Fixed retinas were stained with Nile red or BODIPY 493/503 to visualize LDs (white puncta, arrow) within retinal glia. (a) Few LDs were seen with UAS-Control (LacZ), UAS-Tau or UAS-Tau^R406W expression up to 6d. (b) Rh:ND42 RNAi was used to induce neuronal ROS and robust LD formation within glia. No change in LD number was observed when UAS-Tau or UAS-Tau^R406W were also expressed using Rh1-GAL4. Arrow = glial LD. A datapoint represents the average LD number per ommatidium per animal (n). n = 3–6 (detailed in Supporting Data 1). Data presented as mean ± SD. Statistics: 1-way ANOVA with Tukey’s multiple comparisons test. P-values = not significant, >0.05. (c) UAS-Control or UAS-hTau were expressed in retinal glia using 54C-GAL4 and TEM was performed. Glial cells = green. Shown are representative images for the quantification provided in Fig. 1c. (d) UAS-Ctrl (LacZ) or UAS-hTau were expressed for 10 days in adult brain glia using repoGS, starting at 2-days old (d). Both normal (total Tau) and hyperphosphorylated hTau (AT8, AT100) protein is produced in fly heads, assayed using western immunoblots (WBs). Shown are 3 replicate samples for UAS-hTau collected in parallel.

Extended Data Fig. 3 |. Assessment of existing tau loss-of-function alleles.

(a) Schematic of the genomic region deleted in tau^MR22 (green line) flies and the impacted genes. (b) Schematic of fly tau gene and isoform G. Green line = the deleted genomic region in tauΔex6–14 (more commonly known as “dTau KO”) flies. Red shows exons known to encode microtubule binding domains (MTBDs) that are deleted. A large exon (orange) that is exclusive to isoform tau-G is intact. (c) PCR was used to confirm the deleted genomic region in tauΔex6–14 homozygous flies. (d) qPCR was performed on control flies, tauΔex6–14 homozygous flies, tauΔex6–14 heterozygous flies, or transheterozygous tauΔex6–14/Df flies to measure tau-G transcript levels. Df = deficiency allele. tau-G is significantly upregulated in response to the tauΔex6–14 allele. Data were consistent using two independent deficiency alleles, Df-1 and Df-2. All flies are in a w[1118] background. Total RNA was extracted from 2d fly heads. A datapoint represents the relative expression per technical replicate (n). n = 4–8. Data presented as mean ± SD. Statistics: 1-way ANOVA with Tukey’s multiple comparisons test. P-values = not significant (n.s.) >0.05, *=0.0117, **≤0.01, ****≤0.0001. See Supporting Data 1 for details on n and P-values. (e) In silico analyses were performed on the protein sequence encoded by tau-G to define MTBDs. Unhighlighted text is sequence unique to tau-G and not deleted in tauΔex6–14 flies. Pink highlighted text is sequence deleted in tauΔex6–14 flies that is also found in tau-G. Red and green text denotes specific MTBD identified by MAPanalyzer. Italicized text denotes sequence identified as common to MT associated proteins using InterPro. Note the overlap in predictions at two MTBD within the sequence unique to tau-G (red, italicized). These predictions correctly identify established MTBD (green, italicized) within the sequence deleted in tauΔex6–14. Overall, tau-G is likely upregulated to compensate for the loss of other tau transcripts in tauΔex6–14 flies.

Extended Data Fig. 4 |. RNA analysis of tau-CRIMIC flies.

(a) Schematic of the CRIMIC cassette inserted between exons 10 and 11 of tau. This disrupts all transcripts carrying established (red) and predicted (orange) microtubule binding domains. Primer sets (blue) were designed to measure the levels of the different tau transcripts by qPCR. (b-d) qPCR data for flies including yw control, heterozygous tau-CRIMIC/+, and trans-heterozygous tau-CR/Df mutant flies carrying one copy of tau-CRIMIC and one copy of a deficiency (Df) allele. tau transcript levels were consistently decreased using two independent deficiency alleles, Df-1 and Df-2. A datapoint represents the relative expression per technical replicate (n). n = 4–10. Data presented as mean ± SD. Statistics: 1-way ANOVA with Tukey’s multiple comparisons test. P-values = not significant (n.s.) >0.05, *≤0.05, **≤0.01, ***≤0.001, ****≤0.0001. See Supporting Data 1 for details on n and P-values.

Extended Data Fig. 5 |. Additional data for Drosophila tau RNAi studies.

(a) UAS-Control (Ctrl) RNAi or UAS-tau RNAi was expressed ubiquitously using da-GAL4. Protein extracted from 2d heads was analyzed by WB for dTau levels. A datapoint represents the relative band density per technical replicate (n). n = 6. (b) UAS-Control (LacZ), UAS-Ctrl RNAi or UAS-tau RNAi were expressed in adult flies starting at 2d using the pan-neuronal, drug inducible driver, elavGS. n = 234–325 per genotype. (c) UAS-Ctrl RNAi or UAS-tau RNAi were expressed through development and adulthood using the pan-neuronal driver, elav-GAL4, at 29 °C. Control RNAi n = 161, tau RNAi n = 125. As the GAL4/UAS system is temperature-dependent, this produces higher RNAi expression than standard 25 °C rearing temperatures (see Fig. 3e). (b-c) Survival was measured using Kaplan-Meier plots. P-value ≤ 0.0001 seen with UAS-tau RNAi versus UAS- Ctrl RNAi in b and c, with UAS-tau RNAi versus UAS-Control in c, and UAS- Ctrl RNAi versus UAS-Control in c. UAS-Ctrl RNAi and UAS-tau RNAi are in a w⁺ background. UAS-Control is in a w* background. (d) UAS-Control (GFP) RNAi-2 or UAS-tau RNAi-2 were expressed in retinal glia using 54C-GAL4. Rh:ND42 RNAi was used to induce neuronal ROS and glial LDs. LDs were visualized in fly retinas from 2d animals using BODIPY 493/503 (white puncta, arrows). A datapoint represents the average LD number per ommatidium per animal (n). n = 6. (e) UAS-Control RNAi or UAS-tau RNAi was expressed in retinal glia using 54C-GAL4 and TEM was performed to visualize morphological changes in the fly retina. Glial cells = green. Shown are representative images for the quantification shown in Fig. 5d. (f) UAS-Control RNAi or UAS-tau RNAi were expressed in glia of 1–2d adult flies for 10d. Fixed brains were stained for LDs using BODIPY 493/503 (white puncta, arrows). Shown are representative images for the quantification shown in Fig. 5f. (g) The total levels of triglycerides (TG) were measured in whole heads from either 10d or 26d flies expressing UAS-RNAi in adult glia using repoGS. A datapoint represents the relative TG levels per technical replicate (n). n = 8. Statistics: unpaired 2-sided student t-test (a, d), Log-rank test (b, c), 1-way ANOVA with Tukey’s multiple comparisons test (g). P-values = not significant (n.s.) >0.05, *≤0.05, **≤0.01, ***≤0.001, ****≤0.0001. Data presented as mean ± SEM. See Supporting Data 1 for details on technical replicates, n, and P-values.

Extended Data Fig. 6 |. Downregulating dTau in adult glia causes increased daytime sleep.

UAS-Control RNAi or UAS-tau RNAi were expressed in glia of adult animals starting at 2d using the drug-inducible glial driver, repoGS. Starting at 5d, DAM assays were used to measure fly activity over the next 3 days. (a) Average activity profiles in Control (black, n = 32) or tau RNAi (red, n = 31) expressing flies. (b-d) tau RNAi expressing flies have more recorded total and daytime sleep. A datapoint represents the mean measurement per n animal over the 3 days of analysis. Statistics: unpaired 2-sided student t-test. Data presented as mean ± SEM. P-values = not significant (n.s.) >0.05, ****≤0.0001. Experiments were performed in a w⁺ background.

Extended Data Fig. 7 |. Astrocyte monocultures require Tau for proper LD formation during stress.

(a) HBSS was used to induce stress in rat astrocytes grown in the absence (monoculture, n = 6) or presence (co-culture, n = 5) of neurons. LDs were then stained in these astrocytes using BODIPY 493/503. A datapoint represents the average LD number or size from 10 glia per experimental replicate (n). (b) WB showing GFP (control) or GFP-hTau expression in astrocyte monocultures prior to co-culturing with neurons. (c-e) LipidTox was used to assess LD number in astrocyte monocultures. CM = complete glia media (control). Glia were treated with HBSS to induce oxidative stress. A datapoint represents the average LD number from 5–10 glia per experimental replicate (n). n = 3–4. (c) Astrocytes were transduced to overexpress GFP (control) or GFP-hTau. (d-e) Astrocytes were transduced to express RNAi targeting Mapt to a control RNAi. GFP marks cells that were successfully transduced. Statistics: unpaired 2-sided student t-test (a), 1-way Anova with Tukey’s multiple comparisons test (b and c). P-values = not significant (n.s.) >0.05, *≤0.05, **≤0.01, ***≤0.001, ****≤0.0001. Data presented as mean ± SEM. See Supporting Data 1 for details on n and P-values. All images are maximum intensity projections.

Extended Data Fig. 8 |. Overexpression of futsch in glia disrupts LD formation similar to hTau overexpression.

Rh:ND42 RNAi was used to induce neuronal ROS and glial LD. LDs were visualized in fly retinas of 2d animals using BODIPY 493/503 (white puncta, arrows). UAS-Control (LacZ), UAS-futsch, or UAS-ensconsin (ens) were overexpressed in glia using 54C-GAL4. A datapoint represents the average LD number per ommatidium per animal (n). n = 7–8. Statistics: 1-way ANOVA with Tukey’s multiple comparisons test. P-values = not significant (n.s.) >0.05, *≤0.05, **≤0.01, ***≤0.001, ****≤0.0001. Data presented as mean ± SEM. See Supporting Data 1 for details on n and P-values.

Extended Data Fig. 9 |. Characterization of 1xUAS-hTau flies.

(a) Schematic of a standard UAS construct for hTau that carries five upstream activation sequences (5xUAS). (b) Schematic of our new 1xUAS construct for hTau that carries one upstream activation sequence (1xUAS). (c) WBs comparing expression of total Tau protein in flies carrying a standard (5x)UAS-GFP-hTau transgene versus flies carrying a 1xUAS-GFP-hTau transgene. For accurate comparisons, transgenes were expressed for 4d using the drug-inducible, ubiquitous driver da-GAL4^Geneswitch (daGS) starting in 2d animals. Protein was extracted from whole fly heads. (d) Viability analysis for flies expressing UAS-hTau versus flies expressing 1xUAS-hTau using the tau-CRIMIC allele. (e) WBs showing the successful production of hTau protein when 1xUAS-hTau was expressed using the tau-CRIMIC allele. Protein was extracted from 2d fly heads. Hyperphosphorylated Tau is not detected in these animals using the AT8 antibody (no data to show). Note that expression of hTau using 1xUAS-hTau and the tau-CRIMIC allele can result in secondary bands with this well-established anti-Tau antibody due to the low expression of the protein. In contrast, this same antibody does not produce secondary bands when hTau is expressed at higher levels using 5xUAS-hTau transgenes or a stronger GAL4 driver, as per c and Extended Data Fig. 3b. A datapoint represents the relative band density per technical replicate (n). n = 4. Data presented as mean ± SD. Statistics: 1-way ANOVA w/Tukey’s multiple comparison’s test. P-values = not significant (n.s.) >0.05. (f) Expression of 1xCtrl or 1xhTau using the tau-CRIMIC allele does not induce LDs. Quantification seen in Fig. 7c.

Extended Data Fig. 10 |. Additional data for BODIPY-C12/ER tracker studies and expression data on PLIN2 versus MAPT in aged human brains.

(a) Human oligodendrocyte-like MO3.13 cells were transiently transfected with siRNA targeting MAPT or scrambled control. Cells were fixed and stained for total hTau levels (red) or DAPI (blue). Shown are representative images for the quantification of hTau levels provided in Fig. 8b. Note that the level of hTau downregulation varies between cells. This will make quantifications presented in Fig. 8b,c and Extended Data Fig. 10c,d underestimates as cells with intact Tau will be included. (b) Zoomed out images for those seen in Fig. 8c. Dashed square = the representative image used. (c) Total levels of BODIPY-C12 were measured in transfected cells. MAPT siRNA results in an overall increase in BODIPY-C12 levels, indicating defects in the breakdown of externally sourced lipids. (d) Total levels of ER tracker signal were measured in transfected cells, showing ER enlargement with MAPT siRNA-2. (c-d) A datapoint represents the relative fluorescence per cell in an image (n). n = 5. Statistics: 1-way ANOVA with Dunnett’s multiple comparisons test. Data presented as mean ± SD. P-values = not significant (n.s.) >0.05, *≤0.05, **≤0.01, ***≤0.001, ****≤0.0001. (e-g) Publicly available RNA-seq data from human postmortem brains was analyzed for expression levels of MAPT (encodes Tau) and the LD marker, PLIN2. Pearson r correlations were performed to define a correlation in expression between these two genes. Two analyses were performed: (1) includes all individuals independent of clinical status; (2) includes only individuals with a clinical diagnosis of AD (shown in red). (e) RNA-seq data in the MAYO cohort was obtained from temporal cortex (TCX) post-mortem tissue. This cohort includes individuals with AD, progressive supranuclear palsy, and cognitively unimpaired individuals. (f) RNA-seq data in the ROSMAP cohort was obtained from dorsolateral prefrontal cortex (DLPFC) post-mortem tissue. ROSMAP is a prospective study recruiting individuals from the community setting without known dementia who are followed longitudinally and consent for brain donation and autopsy at the time of death. See Supporting Data 1 for details on n and statistics.

Fig. 1 |. Overexpression of hTau in glia disrupts LDs and glial cell morphology in response to neuronal ROS.

UAS-Control (LacZ) or UAS-hTau transgenes were expressed in retinal glia using 54C-GAL4. When applicable, Rh:ND42 RNAi was used to induce neuronal ROS and glial LDs. a, LDs were visualized in fly retinas using BODIPY 493/503 (white puncta, arrowheads). A datapoint represents the average LD number per ommatidium per animal (n); n = 6–12. b, Representative TEM images of 2 d or 40 d fly retinas expressing Control or hTau within retinal glia and in the presence of Rh:ND42 RNAi. Green, glial tissue; red, vacuoles. See Extended Data Fig. 2c for images of Control or hTau expression in the absence of Rh:ND42 RNAi. c,d, Quantification of TEM images showing the total area occupied by glial tissue (c) or vacuoles (d) where a datapoint represents the total cumulative area surrounding photoreceptor R1 in a single ommatidium (n); n = 18–27. Ommatidia were from three animals (2 d) or 4 animals (40 d). e–g, The drug (RU486)-inducible, glial driver repoGS was used to express UAS-Control or UAS-hTau in the glia of adults starting at 2 d for 10 days. LDs were visualized in the central brain using BODIPY 493/503 (white puncta, arrowheads) (e). A datapoint represents the total number of LDs per brain (n; outlined); n = 7 (f). LPO levels were measured in 10 d or 21 d fly heads. A datapoint represents the relative LPO levels per technical replicate (n); n = 9 (g). Statistics: one-way ANOVA with Tukey’s multiple comparisons test in a, c and g; unpaired two-sided Student’s t-tests in d and f. Data presented as mean ± s.e.m.; n.s., not significant (P > 0.05); *P < 0.05; **P ≤ 0.01; ***P ≤ 0.001; ****P ≤ 0.0001. See Supporting Data 1 for details on n and P values. Ctrl, Control.

Fig. 2 |. tau loss disrupts neuronal ROS-induced glial LDs and causes degenerative phenotypes.

a, Schematic of a CRIMIC cassette inserted in the fly tau gene. Features of the CRIMIC cassette include FRT-flanking sites for cassette removal in the presence of FLP, a splice acceptor (SA) sequence to truncate tau transcripts and a T2A-GAL4 sequence to produce GAL4 under control of tau’s regulatory elements. b, Western immunoblots on fly heads show a severe loss of dTau in trans-heterozygous tau-CRIMIC (CR)/deficiency (Df) mutant flies using an antibody that targets microtubule-binding domains within dTau. Loss of dTau can be rescued by using a genomic rescue construct (GR). A datapoint represents the relative normalized band density per technical replicate (n); n = 4–12. c, Glial LD formation is severely disrupted in heterozygous tau-CRIMIC animals. This can be rescued by reintroducing tau using GR. Neuronal ROS was induced using Rh:marf RNAi. Glial LDs were visualized in fly retinas using BODIPY 493/503 (white puncta, arrows). A datapoint represents the average LD number per ommatidium per animal (n); n = 10–12. d, tau-CR/Df flies have climbing defects that are rescuable using the GR in a negative geotaxis assay. A datapoint represents the mean number of seconds (up to 60 s) required by an individual fly (n) to reach a 6.5 cm line; n = 21. e, tau-CR/Df flies have reduced survival that is rescuable using the GR based on Kaplan–Meier plots. n = 199–226. Statistics: one-way ANOVA with Tukey’s multiple comparisons test in b, c and d; log-rank test in e. n.s., not significant (P > 0.05); *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ****P ≤ 0.0001. Data presented as mean ± s.e.m. See Supporting Data 1 for details on n and P values. All experiments were performed in a yw background.

Fig. 3 |. tau is expressed in neurons and glia.

a,b, tau-CRIMIC (T2A-GAL4) was used to express UAS-mCherry.NLS; 2–3 d adult brains were fixed and counterstained for Elav (neurons) (a) or Repo (glia) (b). The overlapping signal (white) between mCherry.NLS (green) and Elav or Repo (purple) indicates neurons or glia that express tau. Zoomed images are focused on regions rich in neurons or glia. c,e, UAS-Control (Luc) RNAi or UAS-tau RNAi were expressed in neurons starting in embryos and through adulthood using elav-GAL4. d, UAS-Control RNAi or UAS-tau RNAi were expressed in neurons of adult animals starting at 2 d using drug-inducible neuronal driver, nSybGS. Negative geotaxis was used to measure climbing abilities shown in c and d; a datapoint represents the mean number of seconds required (up to 60 s) by an individual fly (n) to reach a 6.5 cm line; n = 20–21. Data presented as mean ± s.e.m. Survival defects were measured using Kaplan–Meier plots (e); n = 250–253. Statistics: one-way ANOVA with Tukey’s multiple comparisons test in c and d; log-rank test in e. Not significant (P > 0.05); **P ≤ 0.01; ****P ≤ 0.0001. In c–e, experiments were performed in a w⁺ background. See Supporting Data 1 for details on n and P values. Ctrl, Control.

Fig. 4 |. tau loss in glia contributes to climbing and lifespan defects.

a–c, repoFLP was introduced into tau-CR/Df flies to excise the CRIMIC cassette only in glia; tau remained mutated in all cells except glia. Experiments were performed in a yw background. Schematic of repoFLP/FRT manipulation in tau-CR/Df flies describing the cell-type-specific manipulation within each genotype (underlined text); created with BioRender.com (Methods) (a). d,e, UAS-Control RNAi or UAS-tau RNAi were expressed in glia of adult animals starting at 2 d using repoGS. Experiments were performed in a w⁺ background. Negative geotaxis assays were used to measure climbing abilities of flies in b and d; a datapoint represents the mean number of seconds required (up to 60 s) by an individual fly (n) to reach a 6.5 cm line; n = 20–41. Data presented as mean ± s.e.m. Statistics: one-way ANOVAs with Tukey’s multiple comparisons test. Survival defects were measured using Kaplan–Meier plots in c and e; n = 197–250. Statistics: log-rank test. n.s., not significant (P > 0.05); *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001. See Supporting Data 1 for details on n and P values. Ctrl, Control.

Fig. 5 |. Glial dTau loss disrupts LD formation and glial cell morphology in response to neuronal ROS.

a–e, Rh:ND42 RNAi was used to induce neuronal ROS and glial LDs. UAS-Control RNAi or UAS-tau RNAi were expressed in retinal glia using 54C-GAL4. LDs were visualized in fly retinas of 2 d animals using BODIPY 493/503 (white puncta, arrowheads) (a). A datapoint represents the average LD number per ommatidium per animal (n); n = 6–11 (b). c, Representative TEM images of 2 d or 40 d fly retinas expressing Control or tau RNAi within retinal glia and in the presence of Rh:ND42 RNAi. Green, glial tissue; red, vacuoles. See Extended Data Fig. 5e for images of Control or tau RNAi expression in the absence of Rh:ND42 RNAi. d,e, Quantification of TEM images showing the total area occupied by glial tissue (d) or vacuoles (e) where a datapoint represents the total cumulative area surrounding photoreceptor R1 in a single ommatidium (n); n = 18–30. Ommatidia were from three animals per condition. f,g,i, repoGS was used to express UAS-Control RNAi or UAS-tau RNAi in brain glia starting in 2 d adult animals. LDs were quantified in fixed brains after 10 d of RNAi expression. Representative images are shown in Extended Data Fig. 5f. A datapoint represents the total number of LDs per brain (n; outlined); n = 8–9 (f). LPO levels were measured in heads from either 10 d or 26 d flies. A datapoint represents the relative LPO levels per technical replicate (n); n = 9 (g). h, If the climbing defects caused by tau loss in brain glia were dependent on its role in glial LD formation, we predicted that turning off the pathway by feeding the flies the antioxidant NACA would result in no climbing defects in flies lacking glial tau. i, Negative geotaxis assays were used to measure climbing abilities in RNAi-expressing flies. NACA was fed to flies to reduce ROS. A datapoint represents the mean number of seconds required (up to 60 s) by an individual fly (n) to reach a 6.5 cm line; n = 39–42. Statistics: unpaired two-sided Student’s t-test in b, e and f; one-way ANOVA with Tukey’s multiple comparisons test in d, g and i. n.s., not significant (P > 0.05); *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ****P ≤ 0.0001. Data presented as mean ± s.e.m. See Supporting Data 1 for details on n and P values. Experiments were performed in a w⁺ background. Ctrl, Control.

Fig. 6 |. Tau-associated disruptions in glial LD formation is conserved in mammalian astrocytes.

a, Schematic of the neuron–astrocyte lipid transfer assay. First, primary rat hippocampal neurons or astrocytes are cultured separately. The neurons uptake Red-C12-labeled fatty acids. Tau levels are upregulated or downregulated (with RNAi) in astrocytes using lentiviral transductions. Upon co-culturing, the neurons and astrocytes are separated by wax, and ROS is induced using HBSS, causing Red-C12-labeled lipids from neurons to accumulate in LDs of astrocytes. b,c,f,g, After HBSS-induced stress, the number of LDs containing Red-C12 within astrocytes was quantified. A datapoint represents the average Red-C12/LD number from ten cells per experimental replicate (n). GFP or GFP-hTau overexpressing astrocytes were co-cultured with Red-C12-loaded neurons; n = 4 (b,c). Astrocytes expressing a Control RNAi or Mapt RNAi were co-cultured with Red-C12-loaded neurons; n = 3 (f,g). d,h, LPO levels were measured in astrocyte monocultures after 4 h treatment with neural-condition media from HBSS-stressed neurons (NCM-HBSS). When applicable, NACA was added at the same time as NCM-HBSS. A datapoint represents the relative LPO levels per technical replicate (n); n = 12 (d), n = 8 (h). e, Western immunoblots were used to measure rTau levels in astrocytes (monoculture) expressing RNAi before co-culturing with neurons. Three independent RNAi targeting Mapt were compared to Control (scrambled) RNAi. A datapoint represents the relative band density per experimental replicate (n); n = 4. Statistics: unpaired two-sided Student’s t-test in c; one-way ANOVA with Dunnett’s multiple comparisons test in e and g; one-way ANOVA with Tukey’s multiple comparisons test in d and h. n.s., not significant (P > 0.05); *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ****P ≤ 0.0001. Data presented as mean ± s.e.m. in c, d, g and h or as mean ± s.d. in e. See Supporting Data 1 for details on n and P values. Ctrl, Control.

Fig. 7 |. Tau’s function as a microtubule-binding protein impacts LD formation.

a–c, Rh:ND42 RNAi was used to induce neuronal ROS and glial LD formation. LDs were visualized in fly retinas of 2 d animals using BODIPY 493/503 (white puncta, arrows). UAS-hTau-2 lines were overexpressed in glia using 54C-GAL4. Phosphoresistant mutant hTau-2[S262A] has a strong affinity for MT; phosphomimetic mutants hTau-2[S262E] and hTau-2[S262D] are predicted to have reduced affinity for MT (a). Flies heterozygous for LOF futsch[N94] or ensΔC alleles have reduced glial LD number (b). Glial LD number is disrupted in heterozygous tau-CRIMIC animals (c). This can be rescued with ectopic expression of hTau using a 1xUAS-hTau (1xhTau) transgene to avoid toxicity associated with overexpressing hTau. 1xhTau transgenes carrying disease-associated mutations in hTau that disrupt MT–Tau interactions and confer AD risk are unable to rescue defects in LD formation like wild-type hTau protein. A datapoint represents the average LD number per ommatidium per animal (n); n = 5–17. Statistics: one-way ANOVA with Tukey’s multiple comparisons test. n.s., not significant (P > 0.05); *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ****P ≤ 0.0001. Data presented as mean ± s.e.m. See Supporting Data 1 for details on n and P values.

Fig. 8 |. Tau loss in oligodendrocyte-like cells disrupts LD budding from the ER.

a, Schematic of BODIPY-C12/ER tracker pulse–chase assay. First, human oligodendrocyte-like MO3.13 cells were transfected with Control or MAPT-targeting siRNA. Cells were then allowed to take up the fluorescent fatty acid, BODIPY-C12. The ER was subsequently stained with ER tracker, and stress was induced using HBSS. b, Quantification of hTau levels in transfected M03.13 cells stained for Tau protein using immunofluorescence. A datapoint represents the relative total fluorescence taken from a single image (n) captured from an experimental replicate; n = 4–7. c, In the merged images, green puncta (white arrowheads) are mature LDs that have successfully budded from the ER. Yellow puncta (yellow arrowheads) are BODIPY-rich puncta and regions localized at the ER that are future LDs. These would need to bud from the ER to become a mature LD. A datapoint represents the total number of puncta per cell (n); n = 10. Statistics for b and c, one-way ANOVA with Dunnett’s multiple comparisons test. Data presented as mean ± s.d. **P ≤ 0.01; ***P ≤ 0.001; ****P ≤ 0.0001. Full images are shown in Extended Data Fig. 10b. See Supporting Data 1 for details on n and P values. d, Endogenous Tau has an important role in glial LD formation, potentially by maintaining microtubule stability, and is required for the proper budding of LDs from the ER. Glia lacking Tau become enlarged in response to neuronal ROS. These enlarged glia degenerate with age as toxic LPOs accumulate due to defective LD formation and resulting defects in lipid catabolism. In tauopathies, the aberrant accumulation of Tau (modeled using overexpression) within glia also disrupts glial LD formation and contributes to glial cell degeneration in response to neuronal ROS. e, A proposed model for glial LD disruptions that occur in response to Tau loss in disease progression. Endogenous Tau loss in glia destabilizes MTs and disrupts LD formation. Data support that in disease progression, Tau loss can occur as a result of the prion-like sequestration of endogenous Tau by pTau. Overexpression of Tau overstabilizes MT in models. Our data show that Tau overexpression in glia disrupts glial LDs similar to Tau loss, suggesting that too little or too much Tau will similarly disrupt glial LD formation in disease. Ctrl, Control.