Glia-derived noncanonical fatty acid binding protein modulates brain lipid storage and clearance

Abstract

Glia-derived secretory factors are essential for brain development, physiology, and homeostasis, with their dysfunction linked to a variety of neurological disorders. Through genetic and biochemical approaches, we identified odorant binding protein 44a (Obp44a), a noncanonical α-helical fatty acid binding protein (FABP) highly expressed in Drosophila central nervous system glia. Obp44a binds long-chain fatty acids and shuttles between glia and neurons, acting as a secretory lipid chaperone and scavenger to support lipid storage, efflux, and redox homeostasis. Notably, Obp44a is recruited to apoptotic cells and injured axons, especially when glial engulfment is impaired, demonstrating its role in lipid waste management and clearance of cellular debris during development and in pathological states. Our findings highlight FABPs’ importance in regulating brain lipid dynamics and neuronal response to stress and injury. By visualizing FABP function in vivo, this study provides insights into how defective lipid regulation may contribute to neuronal stress and disease progression.

Glial lipid chaperone Obp44a maintains brain health by regulating lipid transport, waste clearance, and stress responses.

INTRODUCTION

The central nervous system (CNS) is a lipid-rich organ with a complex and specialized lipid composition, essential for proper function and structural integrity (1–3). Dysregulation of brain lipid homeostasis is closely associated with neurological disorders and neurodegenerative diseases, highlighting the importance of maintaining a stable lipid environment for optimal brain health (4–6). In the CNS, this regulation is achieved through multiple interconnected processes largely orchestrated by neuron-glia interactions: lipid uptake across the blood-brain barrier, intercellular lipid transport, local synthesis and metabolism, and the recycling and efflux of lipids from the brain (7–9). Notably, astrocytes play key roles in these processes, serving as primary sources of lipid binding proteins and transporters that shuttle essential lipids and metabolites between neuron and glia (10–15), thus contributing to the establishment and maintenance of the brain’s unique lipid environment (10, 16). While classical apolipoproteins, such as apolipoprotein E, have been extensively studied (17), a diverse set of astrocyte-derived secretory factors involved in lipid metabolism remains largely unexplored. Furthermore, recent lipidomic and metabolomic studies have revealed regional and cell type–specific lipid profiles in the CNS, underscoring the need to investigate distinct lipid transfer proteins that support CNS lipid dynamics across developmental and pathological contexts (18–21).

Among the lipid transfer proteins, fatty acid binding proteins (FABPs) play an essential role in brain lipid trafficking and metabolism, facilitating the distribution, oxidation, metabolism, and storage of fatty acids while supporting multiple signaling pathways (22). FABPs control fatty acid movement to various intracellular compartments and promote the uptake of fatty acids from the extracellular environment into intracellular organelles, such as the mitochondria, endoplasmic reticulum, and peroxisome, where they are used for energy production and membrane synthesis. In mammals, multiple FABPs have been identified, each with a specific pattern of tissue distribution (23). Notably, brain-specific FABP7 is highly enriched in astrocytes (24, 25) and has been implicated in neurodevelopmental and neuropsychiatric disorders, including Down syndrome and schizophrenia (26, 27).

To investigate the mechanism of lipid homeostasis in the brain and the in vivo activities of glia-derived lipid binding proteins, we leveraged the genetic and imaging tools of Drosophila and identified a noncanonical FABP, odorant binding protein 44a (Obp44a). Unlike classical odorant binding proteins, which interact with odorants and pheromones in chemosensory systems (28–30), Obp44a binds long-chain fatty acids and is among the most abundantly expressed secretory proteins in Drosophila astrocytes, revealing a previously uncharacterized glia-derived lipid chaperone. Our findings further indicate that Obp44a dynamically traffics within the brain and releases into the hemolymph, supporting lipid efflux during both development and stress responses. The secretory activity and antioxidant properties of Obp44a mirror known functions of mammalian lipid binding proteins, suggesting an evolutionarily conserved strategy for lipid clearance and neuroprotection. By characterizing OBP44a’s activity in vivo, our study provides insight into lipid regulation in the Drosophila brain and identifies a key pathway for managing lipid dynamics in neural development and injury.

RESULTS

Obp44a is a highly abundant secretory protein produced by the Drosophila CNS glia

Glia-derived secreted proteins serve key functions in neuron-glia interactions but have not been fully evaluated in the Drosophila system (11, 31–33). To analyze the astrocyte secretome, we performed in silico analysis using multiple published RNA sequencing (RNA-seq) datasets. They are generated either by single-cell RNA sequencing (scRNA-seq) or bulk RNA-seq of fluorescence-activated cell sorting–sorted astrocytes at three different developmental stages (34–36). Among the highly enriched transcripts in astrocytes, we specifically searched for secreted proteins identified through Drosophila extracellular domain database, FlyXCDB (37), which contains 1709 secreted proteins uncovered by robust computational predictions and manual curations.

The scRNA-seq dataset collected from the first instar larval (L1) brain contains 44 astrocytes, which are characterized by high enrichment of alrm transcripts (34). Transcripts of 258 secreted proteins were detected in these astrocytes. Among the most highly enriched transcripts, Obp44a is the most abundantly transcribed gene in L1 astrocytes (Fig. 1, A and B, and data file S1). Subsequent investigations, using astrocyte-specific bulk RNA-seq datasets collected from the third instar larval (L3) and adult stages (36), further reveal transcripts of 176 and 29 secreted proteins (data files S2 and S3), respectively. Notably, Obp44a maintains high expression during brain development (Fig. 1C and fig. S1, A to C). Among the 13 secreted molecules that express across all developmental stages in astrocytes, Obp44a also exhibits the highest expression level (Fig. 1C, fig. S1C, and data file S4).

Fig. 1. Obp44a is a highly abundant secretory protein produced by CNS glia in Drosophila.

(A) Heatmap showing the top 30 most highly expressed and enriched secreted proteins in L1 astrocytes. Obp44a ranks as the highest expressed secreted protein. Each column represents an individual astrocyte. (B) Expression and enrichment analysis of the top 30 secreted proteins in L1 astrocytes. (C) Comparative analysis of 13 secreted proteins expressed across L1, L3, and adult stages reveals that Obp44a consistently maintains high expression. (D) Left: Heatmap of the top 30 Obp family members in antenna, larval (L), and adult (A) brains. F, female; M, male. Unlike classical Obps, Obp44a is highly expressed in the brain rather than the antenna. Right: Tissue distribution heatmap shows Obp44a is prominently expressed in the brain and testis. RPKM, reads per kilobase per million mapped reads. (E) Single-cell atlas of the L1 larval brain highlights Obp44a expression in astrocytes, cortex glia, and ensheathing glia. UMAP-1, Uniform Manifold Approximation and Projection 1. (F) Top: Obp44a-Gal4 was generated using a 2.9-kb enhancer upstream of the transcriptional start site. Bottom: Obp44a-Gal4 drives CD8::GFP (membrane label) or redStinger (nuclear label) expression in a subset of glia in L3 larval brains. (G) Antibody staining shows that Obp44a protein is enriched in L3 larval neuropil. (H) An Obp44a::GFP knock-in line reveals in vivo localization of Obp44a in L3 larval brains. (Ha) Single-section images of astrocytes labeled with alrm > CD2::mCherry show low levels of Obp44a in astrocyte somas, consistent with its efficient secretion. (I) Left: Obp44a::GFP shows widespread and diffused distribution in the adult brain. Right: Single-section images reveal Obp44a’s localization in adult optic lobe chiasm (OLC) glia. (J) Blocking secretion via Obp44a-Gal4-driven Nsf2 RNAi knockdown causes Obp44a::GFP to accumulate in glial somas, confirming its glial production and secretion. Representative maximum intensity projections or single optic sections from confocal images are shown. L1, first instar larvae; L3, third instar larvae. Scale bars are as indicated.

Additional transcriptome analysis indicates that although Obp44a belongs to the odorant binding protein family, which facilitates the transport of hydrophobic odorant molecules, the expression of Obp44a in the olfactory sensory organ antennae is nearly undetectable (Fig. 1D) (28). Instead, Obp44a displays high expression levels in larval and adult brains, as well as in the male fly testis (Fig. 1D), shown by the ModEncode tissue-specific RNA-seq datasets (38). To further investigate specific brain cell types responsible for Obp44a production, we analyzed two published Drosophila brain scRNA-seq datasets (34, 35). Our analyses reveal highly enriched expression of Obp44a in astrocytes, cortex glia, and ensheathing glia in L1 (Fig. 1E and fig. S2A) and L3 brains (fig. S2B). These findings highlight the abundant expression of Obp44a in the CNS glia, suggesting specialized roles beyond the traditional functions of odorant binding proteins (28–30).

To validate the findings of transcriptome analysis, we studied the distribution and dynamics of Obp44a protein in the fly brain. Obp44a enhancer Gal4-driven fluorescent markers targeting cell surface [mCD8::GFP (mCD8-tagged green fluorescent protein)] or nuclei (redStinger) revealed a broad distribution of Obp44a in the larval brain (Fig. 1F). Furthermore, the enhancer specifically labels a subset of CNS glia, as indicated by costaining with an antibody against repo, a marker that labels glial nuclei (fig. S3A). Immunohistochemical studies using a custom-made Obp44a antibody revealed a diffused distribution of the protein, which infiltrates the entire neuropil region in the larval brain (Fig. 1G). With high-resolution images, we observed puncta labeled by the antibody, indicative of vesicles and secretory granules containing Obp44a.

Using CRISPR-Cas9–mediated genome editing, we generated an Obp44a::GFP knock-in line to endogenously tag Obp44a and visualize its in vivo localization (fig. S4). The Obp44a::GFP signal matches the antibody staining result closely, displaying a broad and diffused distribution within the larval brain neuropil (Fig. 1H). Notably, although produced in astrocytes, Obp44a protein is low in astrocyte soma and high in the surrounding neuropil region (Fig. 1Ha), suggesting its effective release as a secretory protein. Similarly, in the adult brain, the Obp44a::GFP signal is found throughout the brain (Fig. 1I, left), with visibly higher levels in regions with densely packed neuronal processes, such as the boundaries of antennal lobe (AL), antennal mechanosensory and motor center (AMMC), and subesophageal zone (SEZ). Furthermore, thin optic sections in the optic lobe region show the cytoplasmic distribution of Obp44a in the optic lobe chiasm (OLC) glia (Fig. 1I, right).

To verify the glial secretion of Obp44a, we used Obp44a-Gal4–driven Nsf2 RNA interference (RNAi) to specifically inhibit secretion from Obp44a expressing glia (39, 40) and examined the distribution of Obp44a::GFP. Blocking glia secretion resulted in altered Obp44a::GFP neuropil distribution, confining the protein within glia somata (Fig. 1J), thereby supporting the result that it is actively released from the glia. Furthermore, we conducted Nsf2 knockdown experiments in wild-type flies and detected the endogenous Obp44a protein using antibody staining. This revealed similar accumulation of Obp44a proteins in glia somata (fig. S3B), thus corroborating our observations of the knock-in lines and suggesting that the GFP tag does not notably alter the secretory properties of the Obp44a protein.

Obp44a is a FABP

While homologs of Obp44a are exclusively insect odorant binding proteins (30), the distribution of OBP44a suggests nonsensory roles. As one of the most abundant proteins produced by the CNS glia, what function does Obp44a serve? To answer this question, we first performed a protein structure homology search and identified Aedes aegypti (yellow fever mosquito) Obp22 (AeObp22) as the closest structural homolog of Obp44a. Despite a mere 42% identity between their amino acid sequences, the predicted three-dimensional (3D) structure of Obp44a closely resembles that of AeObp22, featuring the six α helices and a hydrophobic pocket region critical for ligand binding (Fig. 2A). Previous x-ray crystallographic and nuclear magnetic resonance (NMR) studies have indicated the role of AeObp22 as a FABP (41, 42). Upon interacting with arachidonic acid (C20:4), AeObp22 undergoes conformational changes that lead to the formation of an additional α helix, α7, at the C terminus of the protein. Similar conformational change upon ligand binding has also been observed in Obp44a (43). The strong structural homology displayed by AeObp22 prompted us to test the possible interactions between lipids and Obp44a.

Fig. 2. Obp44a is a FABP regulating lipid storage.

(A) Protein structure prediction and homology analysis reveal structural similarities between D. melanogaster Obp44a (DmObp44a) and A. aegypti Obp22 (AeObp22). Top: Protein sequence alignment excluding signal peptides highlights critical amino acids (black triangles) that form the hydrophobic pocket in DmObp44a, corresponding to the residues identified in AeObp22. Bottom: AlphaFold2-predicted 3D structures show structural similarities between DmObp44a (green) and AeObp22 (gray), both featuring six α helices and a hydrophobic pocket that accommodates fatty acid ligands, such as C20:4 arachidonic acid (magenta) shown in the model. (B) Fatty acid binding–induced conformational changes in Obp44a are detected by NMR. The Trp102 side-chain H_ε1 proton (red arrow) is sensitive to ligand interactions. Distinctive shifts from the apo form are detected in the presence of palmitic acid (C16:0), stearic acid (C18:0), eicosenoic acid (C20:1), and 9-HODE (C18:2) but not in the case of docosanoic acid (C22:0). (C) Native PAGE binding assay demonstrates the interaction between Obp44a protein and C16 fatty acid. Left: Coomassie-stained gel shows purified Obp44a protein. Right: BODIPY FL-labeled C16 fatty acid (FA-C16) alone remains in the well without migration. Upon addition of 3.5 μM Obp44a to 10.5 μM FA-C16, a migrating complex is detected, and the colocalization (white star) indicates binding between Obp44a and FA-C16. (D and E) Obp44a mutants exhibit reduced numbers of lipid droplets in the L3 larval neuropil (D) and in the OLC region of adult brains (E), detected by Nile red staining (magenta). Astrocytes are labeled by alrm > CD8GFP (green) as landmarks. Statistical significance is assessed by unpaired t test with Welch’s correction. **P < 0.01 and ***P < 0.001. Error bars represent mean ± SEM; n = 10 in (D) and n = 14 and 19 in (E). WT, wild type.

To evaluate the lipid binding ability of Obp44a, we obtained recombinant Obp44a protein and performed NMR analysis, where conformational changes induced by ligand-protein interactions can be detected through the amide proton resonance repositions (44). For Obp44a, when a putative ligand interacts with its hydrophobic pocket, the bound state is characterized by the chemical shift of H_ε1 NMR signal of Trp102 (W102) (Fig. 2B). In the presence of fatty acids, including palmitic acid (C16:0) and stearic acid (C18:0), eicosenoic acid (C20:1), as well as the oxidized fatty acid, 9-hydroxyoctadecadienoic acid (9-HODE; C18:2), the H_ε1 Trp102 (W102) signal consistently displays an upfield shift [i.e., toward lower parts per million (ppm) values] from 10 ppm, corroborating a binding event (Fig. 2B). However, the introduction of docosanoic acid (C22:0) elicits no such response, with the amide proton spectrum remaining the same as the one in the unbound apo state (Fig. 2B). These results are consistent with previous findings in AeObp22, indicating that the hydrophobic pocket of Obp44a can accommodate a fatty acid ligand, in both native and oxidized forms, with a saturated acyl chain length limited to 20 carbons.

In addition, we conducted native gel shifting experiments using a BODIPY-labeled fluorescent C16 fatty acid (Fig. 2C). The incubation of C16 fatty acid and Obp44a resulted in the strong shifting of the fluorescent signal, providing additional evidence for direct interactions between Obp44a and fatty acids. It is worth noting that the major fatty acids present in the Drosophila brain range from C14 to C20 (45). The capability of Obp44a to bind these fatty acid species is consistent with a potential role in regulating trafficking and metabolism within the fly brain.

Obp44a regulates lipid storage and homeostasis in the Drosophila brain

After establishing Obp44a as a FABP, we sought to evaluate its function in fatty acid trafficking and storage in vivo. We generated a null mutant of Obp44a using CRISPR-Cas9–mediated gene editing (fig. S5) and examined the number of lipid droplets in the mutant fly brain. The fly lipid droplets, primarily consisting of triacylglycerols (TAGs) and sterol esters, serve as a major source of fatty acid storage and protect against oxidative damage during larval development (46–48). Using the Nile red staining (49), we detected a 47% decrease in the numbers of lipid droplets in Obp44a mutant larval brains (Fig. 2D). Similarly, in adult brains, quantification of lipid droplets performed in the OLC glia revealed a 30% of reduction of lipid droplet numbers in Obp44a mutants (Fig. 2E). Together, data obtained from larval and adult stages consistently support the function of Obp44a in maintaining the number of lipid droplets in the fly CNS.

FABPs facilitate intra- and intercellular lipid trafficking in both vertebrate and invertebrate systems (46, 50). However, their impact on overall metabolism and brain lipid homeostasis remains unclear. Since Obp44a is abundantly expressed in the fly CNS, we took this opportunity to address this question experimentally and investigated metabolomic changes associated with the Obp44a deficiency. Using hydrophilic interaction liquid chromatography mass spectrometry (HILIC-MS) analysis (51) of L3 brains (Fig. 3, A and B), we detected alterations in ~300 metabolites in the Obp44a mutant samples. Among these, 139 metabolites exhibited increased levels, while 158 showed reduced levels compared to the wild-type controls (data file S5). These changed metabolites span a broad range of molecular categories (Fig. 3C and fig. S6), indicating widespread metabolic perturbations in the absence of Obp44a.

Fig. 3. Obp44a maintains lipid homeostasis in the fly brain.

(A) Schematic diagram illustrating the sample preparation procedure for metabolomic profiling of L3 larval brains using HILIC-MS. Schematic image is created in BioRender. Yin, J. (2025) https://BioRender.com/sc2v2lu. (B) Principal components (PC) analysis of the L3 larval brain metabolome across all biological and replicates (n = 9) for wild type, Obp44a mutants, and blank controls. (C) Volcano plot highlighting changes in total detected metabolite between wild-type and Obp44a^−/− L3 larval brains. Red and blue dots represent significantly increased or decreased metabolites, meeting the cutoff criteria of P < 0.05 and log₂ fold change > 0.25. (D) Heatmap displays metabolites with significant changes in Obp44a mutant brains, including phosphatidylethanolamines (PE), diacylglycerol (DAG), phosphatidylinositol (PI), monoacylglycerol (MG), carnitines, and oxidated fatty acid 13-HODE. (E) Altered levels of all detected TAG and DAG species, most of which are reduced in Obp44a mutant brains. (F) Consistent reductions in multiple carnitine species were detected in Obp44a mutants. Red asterisks indicate P < 0.05. The color-coded bar (bottom) illustrates the average intensity of each metabolite within this category.

Notably, among the top 25 metabolites displaying a significant change (P < 0.05) across biological replicates, several groups of molecules stood out, including up-regulated 13-hydroxyoctadecadienoic acid (13-HODE) and phosphatidylethanolamine, alongside down-regulated phosphatidylinositols, diacylglycerols (DAGs), and carnitines (Fig. 3D). Furthermore, a notable reduction in glycogen levels is also detected, suggesting a shift in the brain energy supply within the Obp44a mutant. In particular, 13-HODE, a major lipoxygenation product synthesized from linoleic acid (52, 53), emerges as one of the most significantly elevated metabolites in the mutant (Fig. 3, C and D), indicating an increased level of oxidized lipids associated with the loss of Obp44a.

In addition, Obp44a mutants exhibit a significant down-regulation of four DAGs (Fig. 3D), which serve as crucial phospholipid precursors and components of cell membranes (54), as well as intermediates for TAG, the primary form for fatty acid storage and a major component of lipid droplets (55, 56). Upon examining all TAGs and DAGs detected in the metabolomics data, we observed a systematic reduction in Obp44a mutants, with only a few exceptions (Fig. 3E). In combination with the reduced lipid droplet phenotype that we observed in Obp44a mutants (Fig. 2, D and E), the metabolomics data suggest a reduced lipid storage in the fly brain caused by Obp44a deficiency.

Another notable group showing significant and consistent changes in Obp44a mutants is carnitines (Fig. 3D). Of the 18 carnitines detected in the metabolomic dataset, 16 exhibited reduction and two remained unchanged (Fig. 3F). Carnitines serve as cofactors for the transport of long-chain fatty acids, i.e., more than 14 carbons, into the mitochondria for subsequent β-oxidation and energy production (57). The depletion of carnitines in Obp44a mutants suggests a deficit in intracellular long-chain fatty acid transport and an elevated level of fatty acid β-oxidation.

Obp44a deficiency leads to morphological, physiological, and behavioral defects

Given the abundance and broad distribution of Obp44a in the fly brain, as well as its impact on the brain lipid homeostasis, we went on to determine the physiological consequence of its depletion at the cellular and behavioral levels. While Obp44a mutants are viable and have no gross morphological deficits, a close examination of glia in the adult OLC revealed notable anatomical disorganization (Fig. 4A). The OLC glia in wild-type flies displays evenly sized soma, minimal vacuolation, and a neatly aligned layer of nuclei, labeled by anti-repo immunostaining (Fig. 4A, top). In contrast, in 28 Obp44a mutants we examined, 26 showed altered morphology of the OLC glia, characterized by irregular soma shapes, the presence of large vacuoles (Fig. 4A, bottom, arrows) and disorganized nuclei (Fig. 4A, bottom, stars).

Fig. 4. Obp44a deficiency leads to morphological and physiological defects.

(A) Obp44a mutants show disrupted astrocyte morphology (alrm > CD8GFP, in green) and nuclear arrangement (anti-repo, in gray) in the adult OLC (n = 28). The presence of large vacuoles and disorganized nuclei in mutants are marked by orange arrows and stars. (B) CRISPR-Cas9–mediated tissue-specific mutagenesis reduces Obp44a levels in glia. Left: Schematic of the mutagenesis. The image is created in BioRender. Yin, J. (2025) https://BioRender.com/sc2v2lu. Right: Western blot showing reduced Obp44a level in L3 larval brains following the knockdown. (C and D) Glia-specific Obp44a mutagenesis reduces light-elicited calcium responses in L3 larval LNvs, suggesting a role of Obp44a in supporting neuronal physiological response. (C) Left: Schematic diagram illustrating the setup for calcium imaging experiments. Light pulses detected by the Bolwig’s organ (BO) elicit calcium responses in LNvs, which are recorded at the axonal terminal (green dashed circle). Right: Representative images of Pdf > GCaMP7f recordings at 0 and 1 s after stimulation for control and glia-specific Obp44a knockdown brains. (D) Left: Traces showing average GCaMP responses. The shaded area represents SEM, and the dashed line indicates the 100-ms light pulse. Right: Quantification of peak amplitude of the change in GCaMP signals (ΔF/F), following light stimulations. Statistical analysis was performed using one-way analysis of variance (ANOVA) followed by Tukey post hoc test. **P < 0.01 and ***P < 0.001. Error bars represent mean ± SEM; n = 12, 23, and 23.

To probe whether Obp44a depletion influences neuronal function, we evaluated the physiological responses of larval ventral lateral neurons (LNvs), where the light activation of presynaptic photoreceptors leads to calcium influx in the postsynaptic LNvs that can be reported by the calcium indicator GCaMP7f (58–60). Using a combination of repo enhancer–driven Cas9 expression (repo > Cas9) and a guide RNA (gRNA) transgene specifically targeting Obp44a, we achieved glia-specific knockdown of Obp44a and confirmed its efficiency by Western blots (Fig. 4B). In L3, Obp44a knockdown significantly reduced light-evoked physiological responses in LNvs, with calcium transient amplitudes reduced by more than 50% compared to both control groups (Fig. 4, C and D), suggesting that Obp44a is essential for normal neuronal physiological function.

Next, to assess the impact of Obp44a deficiency at the organismal level, we conducted behavioral studies in adult flies. Notably, Obp44a mutants tend to stay at the bottom of the vials, in contrast to wild-type controls that typically climb to the top of the vials. Quantifications confirmed this observation and demonstrated that Obp44a mutants have reduced ability to climb and perform negative geotaxis (Fig. 5A and fig. S7, A and B). Furthermore, to gain additional information regarding their locomotion, circadian rhythm, and sleep behavior, we tested Obp44a mutants using the Drosophila activity monitor (DAM). Consistent with findings from the climbing assay, Obp44a knockout flies showed significantly reduced locomotor activity during their active periods (Fig. 5, B and C). In addition, although their circadian rhythm appears normal (Fig. 5B), mutant flies exhibited fragmented sleep (Fig. 5D). While their total sleep time is similar to controls, Obp44a mutants have increased number of sleep episodes and shorter sleep episode duration, suggesting a function of Obp44a in maintaining normal locomotor activities and facilitating sleep consolidation. Recent studies have shown that FABP7-deficient mice display disrupted sleep wake cycles (61, 62). The similar phenotypes suggest a functional homology between Obp44a and mammalian FABP7.

Fig. 5. Obp44a is required for normal locomotion, sleep behavior, and resistance to oxidative stress in adult flies.

(A) Obp44a mutants exhibit significantly reduced climbing ability, indicated by a reduction in climbing index (top) and climbing height (bottom). Statistical significance was assessed using an unpaired t test with Welch’s correction for the climbing index and Welch’s ANOVA test with Dunnett’s multiple posttests for the climbing curve, **P < 0.01 and ***P < 0.001. Error bars represent mean ± SEM; n = 12 and 13 groups, 20 flies per group. (B) Obp44a mutants show altered locomotion patterns under light:dark (LD) and constant dark (DD) conditions. Representative actograms of average activity of 3- to 5-day-old adult flies are shown. (C) Activity curve and quantification demonstrate reduced activity levels in Obp44a mutants. ZT, Zeitgeber Time; CT, Circadian Time. (D) Quantifications show that, compared to wild-type controls, Obp44a mutants exhibit increased number of sleep episodes and decreased episode durations. Statistical significance was assessed using an unpaired t test with Welch’s correction, **P < 0.01 and ***P < 0.001; ns, not significant. Error bars represent mean ± SEM; n = 88, 94. (E) Glutathione redox (GSH/GSSG) biosensor (mito-roGFP2-Grx1) analysis reveals elevated redox potential in Obp44a mutant brains (10-day-old adults). Left: Representative raw and ratiometric images of optic lobes expressing mito-roGFP2-Grx1. Right: Quantification of 405/488 ratiometric fluorescent changes in the optic lobe region indicates increased redox potential in mutant brains. Statistical significance was assessed by unpaired t test with Welch’s correction, ***P < 0.001. Error bars represent mean ± SEM; n = 24, 30. (F) Survival curve of wild-type and Obp44a^−/− adult flies subjected to 5% H₂O₂ treatment. Obp44a^−/− flies exhibit significantly reduced survival probability after 24-hour treatment. Statistical significance was assessed by Welch’s ANOVA test with Dunnett’s multiple posttests, ***P < 0.001. Error bars represent mean ± SEM; n = 9 groups, 16 flies per group.

Our metabolomic analysis indicates that Obp44a deficiency leads to diminished lipid storage, altered lipid transport for metabolism within the mitochondria, and elevated lipid oxidation (Fig. 3). To investigate the involvement of Obp44a in regulating the oxidative state of the brain, we performed experiments on 10-day-old adult flies to evaluate the redox potential using an in vivo redox biosensor mito-Grx1-roGFP under the control of a tubulin enhancer (63). We observed a significantly higher fluorescence ratio change (405/488) in Obp44a mutants, indicating an elevated glutathione redox potential in mitochondria, which are the major source of oxidants in cells (Fig. 5E). Furthermore, to assess sensitivity to oxidative stress, we exposed adult flies to H₂O₂-induced oxidative stress via feeding, a treatment that typically causes lethality in wild-type flies within 4 days. Obp44a mutants exhibited increased sensitivity to H₂O₂, with significantly reduced median survival times compared to controls (Fig. 5F). This heightened susceptibility suggests that elevated oxidative stress in Obp44a mutants compromises their ability to withstand oxidative challenges, consistent with findings from our metabolomic analysis. Together, these results support a role for Obp44a in maintaining redox homeostasis and protecting the Drosophila brain from oxidative stress.

Obp44a traffics between neuron and glia and is released into hemolymph

We identified Obp44a as a highly abundant secretory FABP produced by CNS glia. To directly assess its function in neuron-glia lipid shuttling, we used the Obp44a::GFP knock-in line to visualize the trafficking of the Obp44a protein in both brain tissues and dissociated cells.

As shown previously, endogenously tagged Obp44a::GFP is diffusedly distributed throughout the neuropil region (Fig. 1H). This distribution pattern can be recapitulated by expressing GFP tagged Obp44a in either neuron or glia in the larval brain using cell type–specific enhancers (Fig. 6A). Notably, Obp44a::GFP produced by neurons is found in cytoplasm of glia cells in the larval optic lobe (Fig. 6A, right). Experiments using acutely dissociated brain cells further confirmed this observation. When Obp44a::GFP expression was restricted in astrocytes via the alrm-Gal4 enhancer, GFP signals were detected in neurons (Fig. 6B). Among the 74 neurons analyzed, 84% exhibited GFP signals, indicating substantial glia-to-neuron transfer. Conversely, when Obp44a::GFP expression was restricted to neurons using the elav-Gal4 driver, GFP signals were detected in 100% of the 91 observed glial cells, demonstrating robust neuron-to-glia trafficking (Fig. 6B). These findings consistently demonstrate that Obp44a can be effectively secreted and taken up by both neuron and glia.

Fig. 6. Obp44a traffics between neuron and glia and is secreted into the hemolymph.

(A) Expression of Obp44a::GFP transgene in astrocytes or neurons recapitulates its localization observed in the Obp44a::GFP knock-in line, demonstrating efficient secretion and dynamic trafficking across cell types. Right: Higher magnification, single optical section of the larval optic lobe shows that neuronally expressed Obp44a::GFP (elav-Gal4 > UAS-Obp44a::GFP) is found in the soma of repo-positive glia, demonstrating intercellular transfer from neurons to glia. (B) Obp44a traffics between glia and neuron. Left: Schematic of the experimental setup using acutely dissociated cells from L3 larval brains. Schematic image is created in BioRender. Yin, J. (2025) https://BioRender.com/sc2v2lu. Right: Obp44a::GFP produced in astrocytes (alrm-Gal4 > UAS-Obp44a::GFP) is observed in elav-positive neurons, while neuronal expressed Obp44a::GFP (elav-Gal4 > UAS-Obp44a::GFP) is detected in repo-positive glia, indicating Obp44a’s ability to traffic between glia and neuron. n = 74 (glia to neuron), n = 91 (neuron to glia). (C) Time-lapse live imaging of L3 larval brain explants reveals Obp44a::GFP mobilization from the neuropil toward surface glia. All 25 explants imaged exhibited similar phenotypes. (D) Obp44a::GFP is secreted from the brain into surrounding medium. Western blot of L3 larval brain explants incubated in physiological saline shows Obp44a::GFP in both the supernatant (secreted fraction) and brain lysate. Quantification from four biological replicates is shown in the right. (E) Obp44a is present in circulating hemolymph. Representative Western blot of L3 larval hemolymph reveals both monomeric and dimeric forms of Obp44a (arrow) in wild type larvae, which are absent in Obp44a mutants. Asterisk indicates a nonspecific band. DAPI, 4′,6-diamidino-2-phenylindole.

Next, we examined Obp44a mobilization in the brain and its interaction with fatty acid cargos in vivo. Using larval brain explants, we performed live imaging experiments to monitor endogenously tagged Obp44a::GFP alongside fluorescently labeled C12 fatty acid (C12-red). After 24 hours of feeding, C12-red was readily detected in the brain tissue, primarily within lipid droplets that were surrounded by Obp44a::GFP. Upon dissection and incubation of L3 brain explants in physiological saline, Obp44a::GFP signals rapidly diminished from the neuropil and migrated outward toward the surface glia. This redistribution was consistently observed across all samples (n = 25) (Fig. 6C and fig. S8A). In nearly half of the samples (12 of 25 brains), Obp44a::GFP was also detected in the surrounding medium, released in association with vesicle-like structures of variable sizes (fig. S8B).

Notably, C12-red–labeled lipid droplets remained closely associated with Obp44a::GFP and appeared to migrate in parallel toward the brain surface (fig. S8A), suggesting that Obp44a may associate with mobilized lipid droplets during fatty acid efflux. Although this approach does not allow direct visualization of Obp44a binding and transporting individual fatty acid molecules in vivo, the observed redistribution and secretion of Obp44a::GFP, along with its biochemically demonstrated fatty acid binding activity (Fig. 2), support a role for Obp44a in facilitating the mobilization and efflux of fatty acids from the brain.

To directly assess secreted Obp44a protein, we incubated Obp44a::GFP knock-in larval brains in physiological saline for 30 to 60 min, followed by Western blot analysis of the supernatant. This revealed high levels of Obp44a::GFP released into the medium (Fig. 6D). We also performed Western blot analysis on hemolymph collected from larvae, which detected Obp44a in both dimeric and monomeric forms in wild-type animals but not in knockout mutants (Fig. 6E). This result is consistent with previously published hemolymph proteomic data identifying Obp44a by mass spectrometry (64). Last, we observed strong Obp44a::GFP signals in nephrocytes, “kidney-like” cells on the larval epidermis that filter hemolymph, further confirming the presence Obp44a in the circulatory system (fig. S4).

Obp44a is a lipid chaperone and scavenger during development and injury response

During larval and pupal development, the Drosophila CNS undergoes two waves of neurogenesis, accompanied by significant brain remodeling, mass synaptic pruning, and programmed cell death associated with metamorphosis. These processes result in significant amounts of cellular debris, particularly lipids, which need to be efficiently cleared to prevent neuroinflammation. Given that Obp44a exhibits high affinity for fatty acids and is abundantly expressed in the larval brain, we hypothesized that Obp44a might play a role in lipid clearance during this developmental transition.

CNS glia are the primary phagocytes responsible for clearing neuronal debris during both development and neuroinflammation in Drosophila (65, 66). Astrocytes and cortex glia, in particular, rely on the phagocytic receptor Draper (drpr), a homolog of mammalian MEGF10/Jedi-1 (67), to engulf neuronal debris (68). As reported previously, drpr loss-of-function mutants exhibit clusters of round-shaped apoptotic cells (66, 69), particularly in the optic lobes, indicating that impaired glial engulfment results in the persistence of apoptotic cells. Notably, anti-Obp44a immunostaining revealed significant accumulation of Obp44a within these cell clusters in both larval and adult drpr mutant brains (Fig. 7A). This observation suggests that Obp44a is recruited to apoptotic cells and may play a role in cellular debris clearance.

Fig. 7. Obp44a functions as a lipid chaperone and scavenger during development and injury response.

(A) Endogenous Obp44a accumulates in apoptotic-like cells when phagocytosis is impaired in larval or adult drpr mutant brains. (B) Obp44a is recruited to Caspase-3–positive apoptotic like cells in both L3 larval and adult brains upon glial drpr knockdown. Right: Higher magnification, single-section images from adult optic lobes. Magenta arrowheads indicate representative cells colabeled with Obp44a::GFP and caspase-3. Nuclei are marked with DAPI (blue). (C) Obp44a responds to injury-induced lipid stress following antennal axotomy. Top left: Schematic of unilateral axotomy in adult flies. Right and bottom: Representative confocal images of Obp44a::GFP-labeled brains at 6, 24, and 48 hours postinjury. At 6 hours, Obp44a::GFP distribution is comparable on both sides of the brain. By 24 hours, Obp44a::GFP accumulates on the injured side, particularly in the AL and AMMC. At 48 hours, Obp44a::GFP signal intensifies and spreads to adjacent regions. (D) Schematic of Obp44a’s function as a lipid chaperone and scavenger. Produced by cortex and neuropil glia, Obp44a is secreted into the neuropil, traffics through multiple brain cell types, including neurons and surface glia, and ultimately enters the hemolymph. By interacting with both native and oxidized fatty acids, Obp44a regulates lipid trafficking and metabolism, facilitating the clearance and efflux of oxidized lipids during development and neuroinflammation and contributing to brain homeostasis. FFA, free fatty acid; FA, fatty acid. Schematic images are created in BioRender. Yin, J. (2025) https://BioRender.com/sc2v2lu.

To further investigate the Obp44a’s role in lipid clearance, we examined the localization of endogenously tagged Obp44a::GFP in wild-type flies and flies with glial-specific drpr knockdown. In controls, Obp44a::GFP was broadly distributed throughout the brain. In contrast, drpr knockdown led to prominent accumulation of Obp44a::GFP in apoptotic cells marked by cleaved caspase-3, observed in both larval and adult brains and particularly enriched in the optic lobe region (Fig. 7B). High-magnification images of the adult optic lobe revealed that all Obp44a-positive cells were also caspase-3 positive, confirming the recruitment of Obp44a to apoptotic cells, likely to bind fatty acids and support lipid clearance.

Previous studies have shown that defective glial engulfment due to drpr dysfunction results in notable accumulation of mushroom body axonal debris in the adult brain, detectable by anti-FasII staining (70). To test whether Obp44a is recruited to these persistent debris, we examined Obp44a::GFP knock-in flies in a drpr knockdown background and observed Obp44a::GFP accumulation on FasII-positive debris, consistent with its localization to sites of impaired clearance (fig. S9A). While Obp44a mutants alone did not exhibit debris accumulation, Obp44a; drpr double mutants showed a significant increase in FasII staining intensity compared to drpr single mutants (fig. S9, B and C). These results indicate that while Obp44a is not required to initiate debris clearance, it facilitates the process, and its loss further compromises clearance efficiency when drpr-mediated engulfment is impaired.

Collectively, these results demonstrate that Obp44a is actively recruited to apoptotic cells and axonal debris poised for engulfment, supporting its role in facilitating lipid clearance during brain development. Notably, FABPs have not previously been implicated in glial-mediated lipid clearance. Our results provide evidence that Obp44a, and potentially other FABPs, acts as a lipid scavenger in this essential developmental process.

Axonal injury in Drosophila robustly activates glial engulfment and inflammatory responses, providing an ideal context to further evaluate Obp44a’s scavenging function (68). To precisely assess Obp44a’s response, we surgically removed antennal segments 2 and 3 on one side of adult flies to introduce antennal nerve damages (68, 71), using the contralateral side as an internal control. At 6 hours postinjury, Obp44a::GFP distribution remained largely diffused, showing no difference from the uninjured side (Fig. 7C). By 24 hours, Obp44a::GFP accumulated in AL and AMMC regions on the injured side, and by 48 hours, strong signals extended across the injured antennal nerve, AL, AMMC, and the surrounding glia (Fig. 7C and fig. S10). This progression aligns with the time course of Wallerian axon degeneration and clearance following similar injuries (68, 71). Modest Obp44a::GFP accumulation was also detected in the AL on uninjured side at 24 and 48 hours, likely reflecting degeneration of contralateral olfactory receptor neuron projections. Quantification of Obp44a::GFP intensity in the AL/AMMC at 48 hours postinjury confirmed a significant increase on the injured side compared to controls (fig. S10), supporting the notion that Obp44a is actively recruited to sites of injury and inflammation, likely contributing to lipid clearance and debris processing during the glial injury response.

Together, our studies support a model for the dynamic trafficking of Obp44a among brain cell types and its role as a chaperone and scavenger in lipid transport and clearance (Fig. 7D). Initially secreted by cortex and neuropil glia, Obp44a enters neurons and neuropil regions, where it acts as a chaperone for fatty acids and supports essential functions such as membrane biogenesis, energy production, and lipid storage. Under oxidative stress and neural injury, Obp44a binds to free and oxidized fatty acids, transporting them out of the brain via surface glia and the hemolymph (Fig. 7D), preventing toxic lipid by-product accumulation. Through interactions with phagocytic pathways, such as drpr-mediated engulfment, Obp44a facilitates efficient clearance and recycling of cellular debris, supporting brain development and recovery from neuronal injury, reducing neuroinflammation and promoting overall brain health.

DISCUSSION

Within the CNS, fatty acids are essential for structural, signaling, and metabolic functions (72, 73), and their dysregulation is linked to a range of physiological and pathological conditions (74, 75). As primary carriers that transport hydrophobic fatty acids, FABPs have been extensively studied in vitro, where their 3D structures, ligand binding affinities, and interactions with phospholipid containing vesicles and membranes are well-characterized (23). Furthermore, their specific tissue distributions and knockout phenotypes underscore their critical roles in brain function (76–78). However, in vivo assessments of FABPs’ effects on brain metabolism and their interactions with lipid cargos under physiological and pathological conditions remain limited. Here, we identify Drosophila Obp44a as a prominent, glia-derived, secretory FABP, addressing this gap by exploring its role in brain lipid dynamics and homeostasis.

Obp44a stands out as a highly abundant FABP in the CNS with a significant role in lipid metabolism. Its deficiency leads to widespread changes in the larval brain metabolome, including reduced TAG and DAG levels, indicating decreased lipid storage. In addition, the depletion of carnitines and alterations of lipid species such as DAG and phospholipids suggest impacts on fatty acid β-oxidation, membrane composition, and signal transduction. Notably, carnitines are essential cofactors for transporting fatty acids across mitochondrial membranes for β-oxidation pathways to produce energy (57, 79). Reduced carnitine levels in Obp44a-deficient brains may thus limit mitochondrial fatty acid import, impairing energy production, leading to the accumulation of free fatty acids, potentially contributing to oxidative stress and cellular dysfunction. These metabolic shifts likely underlie the impaired calcium response and behavior deficits observed in Obp44a mutants, emphasizing the importance of FABPs in maintaining normal physiology and stress response in the CNS.

A key finding of this study is Obp44a’s recruitment to apoptotic cells and injured axons, revealing a previously uncharacterized role for FABPs as CNS lipid scavengers. The release of fatty acids, such as docosahexaenoic acid and arachidonic acid, at sites of phagocytosis and injury poses risks for inflammation and lipotoxicity if not efficiently cleared (80, 81). Obp44a appears to function as a lipid scavenger at these sites by binding oxidized fatty acids and other lipid by-products, promoting their clearance through drpr-mediated engulfment. This mechanism prevents the buildup of potentially toxic lipid residues, thereby protecting neighboring neural and glial cells from oxidative damage. Furthermore, Obp44a may help recruit specific fatty acids needed for axonal repair, paralleling the role of mammalian FABPs in injury recovery (81–83). Collectively, these findings highlight a dual role of Obp44a in lipid transport and waste management, likely representing an evolutionarily conserved lipid management strategy that could extend to mammalian FABPs.

Odorant binding proteins are widespread across the animal kingdom and are notable for their ability to bind a diverse range of hydrophobic ligands, including odorants, small organic molecules, amino acids, and fatty acids (28–30). As a noncanonical CNS-specific FABP, Obp44a represents a great example of naturally occurring protein engineering during evolution, revealing molecular adaptation in protein-lipid interactions. The lipid storage deficits and physiological abnormalities observed in Obp44a mutants closely resemble those in FABP7 knockout mice (61, 62, 84), suggesting a functional similarity between Drosophila Obp44a and mammalian FABP7 despite their structural differences. Known mammalian FABPs, along with the Drosophila homolog, dFABP (fabp) (85), feature a conserved β barrel structure with an internal fatty acid binding pocket and a N-terminal “cap” domain (23, 86). In contrast, Obp44a shares structural features with other insect odorant binding proteins, with an internal hydrophobic cavity formed by α helices and a flexible C-terminal cap (28). These distinctions highlight structural diversity among lipid-binding proteins, suggesting potential convergent mechanisms in the evolution of protein-lipid interactions and presenting exciting opportunities to discover previously unknown lipid binding proteins through systematic biochemical and structural studies.

When comparing the expression level, tissue distribution, and molecular features of Obp44a to the canonical fabp (46) and the three fatty acid transfer proteins (Fatps) in the Drosophila genome (dFabp1, 2, and 3) (9, 80, 81), we found that Obp44a is the primary FABP enriched in both developing and adult CNS (fig. S11). In contrast, fabp is found in the CNS and multiple peripheral tissues, where it supports intracellular fatty acid transport and lipid droplet formation (46). The Fatps, on the other hand, show low expression overall and typically act as transmembrane proteins, facilitating the uptake of long-chain fatty acids from extracellular fluid (50, 87, 88). In addition to its abundance, Obp44a has an N-terminal signal peptide that enables its secretion and trafficking into the hemolymph, allowing it to function as both an intracellular and extracellular lipid chaperone. The Obp44a’s role as the primary FABP in the rapidly developing larval brain offers a clear adaptive advantage, supporting the high demands for lipid management generated by neurogenesis, axonal remodeling, and synaptic pruning.

However, the specific mechanisms driving Obp44a’s mobilization and recruitment to apoptotic cells or injury sites remain unknown. A major barrier to addressing this is the lack of reliable lipid tracers or markers that enable live imaging of lipid-protein interactions in vivo. Such tools are crucial for directly monitoring how Obp44a binds, traffics, and releases lipid cargo under both physiological and pathological conditions. Given our evidence that Obp44a is secreted from glia and released into the hemolymph, potentially via vesicle-mediated mechanisms, clarifying the underlying cell biological pathways, including endosomal sorting and extracellular vesicle formation, will be essential for understanding its function. Elucidating these processes could enhance Obp44a’s utility as a model for studying lipid homeostasis in the nervous system and as a platform for lipid delivery in disease contexts.

By integrating these findings, our study not only supports the vital role of glia-derived FABPs in CNS function but also lays the groundwork for exploring similar lipid-scavenging and recycling mechanisms in mammalian systems. Given the link between lipid dysregulation and neurological disorders, our research on Obp44a provides a compelling foundation for investigating FABPs in both invertebrate and vertebrate models.

MATERIALS AND METHODS

Fly culture and stocks

Fly stocks were maintained in the standard cornmeal-based fly food in a 25°C incubator with humidity control. Larvae and adults were cultured in the light:dark (LD) condition with a 12-hour light:12-hour dark light schedule. Unless otherwise noted, all larvae were collected between ZT1 and ZT3 (ZT: zeitgeber time in a 12:12-hour light dark cycle; lights on at ZT0, lights off at ZT12). The following Drosophila melanogaster stocks were used for experiments: Obp44a-gRNA, UAS-Obp44a::GFP, Obp44a::GFP CRISPR knock-in, and Obp44a CRISPR knockout are generated in this study. Obp44a-Gal4 [GMR90C03; Bloomington Drosophila Stock Center (BDSC) 47122]; UAS-mCD8::GFP (BDSC 5137); UAS-RedStinger (BDSC 8547); Canton-S (wild type for this study; BDSC 64349); UAS-Nsf2-RNAi (BDSC 27685); alrm-Gal4 (gifted by M. Freeman); repo-Gal4 (BDSC 7415); elav-Gal4 (BDSC 458); Pdf-Gal4 (BDSC 6899); UAS-GCaMP7f (BDSC 86320); tub-mito-roGFP2-Grx1 (BDSC 67669); Obp44a-gRNA (Weizmann Institute of Science); nos-Cas9 (gifted by F. Diao and B. White); drpr RNAi (VDC v27086); and repo-Cas9 and drpr-gRNA and drprΔ5 mutant (gifted by C. Han and M. Freeman). Unless otherwise specified, all experiments are performed either in the L3 larvae (96 to 120 hours after egg laying) or 3- to 5-day-old adult flies.

RNA-seq data collection and analysis

The scRNA-seq data at the L1 stages (Fig. 1, A, B, and E, and fig. S2A) were from a previous publication and can be accessed via the following link (https://cells.ucsc.edu/?ds=dros-brain) or the National Center for Biotechnology Information (NCBI) database (GSE134722) (34). Similarly, the L3 brain scRNA-seq data (fig. S2B) were obtained from https://github.com/aertslab/SCopeLoomR or the NCBI database (GSE157202) as detailed in the published work (35). The L3 (Fig. 1C and fig. S1, A to C) and adult astrocyte RNA-seq data (Fig. 1C and fig. S1, B and C) were retrieved from the published report (36). RNA-seq data of the Drosophila antenna were derived from the published study (89). The L3 CNS data (Fig. 1D) were from ModEncode (ID-4658) (http://data.modencode.org/cgi-bin/findFiles.cgi?download=4658) and analyzed according to our previous study (90). RNA-seq data corresponding to adult female and male Drosophila brains (Fig. 1D) were retrieved from NCBI under the accession code GSE153165, as reported (91). RNA-seq data (fig. S11) analyzed for tissue-specific expression levels of Obp44a, fabp, and Fatps and the reads per kilobase per million mapped reads (RPKM) values were extracted from FlyBase (FBlc0003498), originating from FlyAtlas2 (92).

Seurat data processing and secretome analysis

We used the Seurat version 4.0 pipeline (93) to process scRNA-seq datasets from both L1 and L3 Drosophila brains. The L1 brain dataset (Fig. 1, A to C and E, and fig. S2A) underwent a previously published processing method (34), with filters to retain cells exhibiting unique feature counts ranging between 200 and 4500, while also restricting mitochondrial gene content to less than 20%. This filtration yielded 4349 cells with 12,942 detected genes.

The L3 dataset (fig. S2B) had undergone an initial filtration step, incorporating several quality control criteria according to the publication (35). This rendered a dataset containing 5056 cells and 9853 detected genes. After the construction of Seurat objects, standard preprocessing steps were executed, including log normalization, using a scale factor of 10,000, to standardize gene expression across individual cells by considering the total gene expression within each dataset. A subsequent linear transformation was applied. To determine highly variable genes, we implemented the FindVariable Features function with default parameters, following the guidelines provided by the R package developer.

To determine the dimensionality of the dataset, we explored several strategies, including the Elbow-Plot and JackStraw-Plot tests, in conjunction with an evaluation of principal component analysis. Ultimately, we retained 31 and 35 dimensions for the L1 and L3 datasets, respectively, using these dimensions for cell cluster identification through a graph-based approach. The resolution parameters of 3 and 2.5 were used when using Uniform Manifold Approximation and Projection (UMAP) for visualization in these two datasets.

To analyze the Drosophila secretome dataset, we used the FlyXCDB online database (http://prodata.swmed.edu/FlyXCDB), which contains 1709 secreted proteins (37). In the L1 astrocyte cell cluster (34), 258 secreted proteins were identified (data file S1). Similarly, in the L3 astrocyte–specific dataset (36), 176 secreted proteins were identified (data file S2). Last, in the adult astrocyte–specific dataset (36), 29 secreted proteins were detected, based on the original published threshold of log₂ fold change ≥ 0.5 (data file S3). These findings are visualized through heatmaps and dot plots in Fig. 1 (A to C) and fig. S1.

Generation of Obp44a::GFP knock-in line and the Obp44a mutant

The Obp44a::GFP knock-in line was generated through the integration of a donor plasmid (pTEGM) (gifted by F. Diao and B. White) (94) and a gRNA expression plasmid (gifted by C. Han) (95). The Obp44a::GFP cassette was introduced into the donor plasmid, plus two arm sequences for recombination, including a 560-bp Obp44a 5′ untranslated region (5′UTR) sequence, the FRT flanked Obp44a::EGFP coding sequence, and a 620-bp Obp44a 3′UTR sequence. The two PAM (protospacer adjacent motif) sites used for Cas9 cleavage were rendered nonfunctional within the synthesized sequence. These sequences were introduced for recombination with the endogenous Obp44a genomic sequence, thereby replacing the native Obp44a locus.

To generate the gRNA expression vector, we used a modified gRNA cloning vector based on pAC-attB-CaSpeR4 (96). This vector featured a U6:3 promoter and two gRNA expression cassettes arranged in tandem (tRNA + gRNA + gRNA core elongation factor) (95). The specific targeting sequences for Obp44a within the gRNA expression vector were “ccgagcgagcattcagtcctca” and “ccgctcaggctctgcaatcctac.” Both the donor and gRNA plasmids were prepared at a concentration of 0.5 to 1 μg/μl for coinjection into the nos-Cas9 (attP2) fly line (by Genetivision). The resulting knock-in line was validated through DNA sequencing and Western blots.

The generation of the Obp44a mutant involved the use of the nos-Cas9 (attP2) fly line (gifted by F. Diao and B. White) and a published Obp44a-gRNA line (97). The mutant was produced through combining nos-Cas9 and Obp44a-gRNA through multiple rounds of crossing and selections of the progenies using genomic polymerase chain reaction. The final knockout line with a 5-bp deletion in the coding region was verified via genomic DNA sequencing. In homozygous mutants, Western blot analysis also confirmed the absence of Obp44a protein.

3D protein structure homolog search and structure modeling

To gain insights into the Obp44a protein structure and explore potential homologous proteins, we first used the I-TASSER server to search for the structure homologs of Obp44a protein: https://zhanggroup.org/I-TASSER/, which identified AeObp22. The protein alignment was performed using Clustal Omega (https://ebi.ac.uk/jdispatcher/msa/clustalo). The 3D structure comparison (Fig. 2A) was prepared with ESPript3.0 (https://espript.ibcp.fr/ESPript/cgi-bin/ESPript.cgi) (98). The 3D protein structures for AeObp22 and DmObp44a were obtained from the AlphaFold2 database (https://alphafold.ebi.ac.uk/) (99, 100).

Behavioral analysis

All behavioral assessments were performed on adult flies selected within a 24-hour window posteclosion and reared on standard fly medium until they reached an age ranging between 3 and 5 days.

Climbing assays

Twenty flies from each experimental group were introduced into a climbing vial marked with two target lines positioned at heights of 9.5 and 13.5 cm above the vial base. The vials were gently tapped at the initiation of the assay to ensure that all flies were situated at the bottom. Climbing assessments were carried out to determine the climbing index and the percentage of flies surpassing the designated target heights (Fig. 5A and fig. S7). Flies were allowed to move freely for a duration of 10 s, during which images of the climbing vials were captured to calculate the proportion of flies reaching each target height. Each experimental group underwent this procedure six times, with the final percentage representing the average of these six trials. Climbing index was computed as follows: (% of flies between 9.5 and 13.5 cm lines × 0.5) + (% of flies above the 13.5 cm line × 1). Climbing time curves were generated by allowing flies to move freely for 1 min, capturing images of the climbing vials every 10 s, and calculating the proportion of flies above the 13.5 cm line at each time point.

DAM recording for sleep and locomotion analysis

We followed the published protocol (101, 102) and the product manual provided by the manufacturer (Trikinetics Inc.). Briefly, individual male flies were introduced into each monitor tube, containing fly food on one side and a cotton wick on the other, and placed into DAM2 monitors. The monitors were subsequently positioned within a 25°C incubator with humidity control set to 60% relative humidity and left undisturbed for the duration of the recording. The recording schedule consisted of 3 LD days followed by 6 constant darkness days. Fly activity was recorded at 1-min intervals using DAMSystem311X data acquisition software. Data preprocessing was conducted using the DAMFileScan113X program, followed by analysis of locomotion and sleep parameters using two MATLAB programs: the Sleep and Circadian Analysis MATLAB Program (SCAMP) (103) and the SleepMat software program (104). Only data from flies that remained alive throughout the entire 9-day recording period were included in the analysis.

H₂O₂ survival probability

Monitor tubes and flies were prepared in a manner similar to the procedure for DAM recordings described above, with slight modifications. Male flies were loaded into monitor tubes containing fly food with 5% H₂O₂ on one side and a cotton wick soaked in a 5% H₂O₂ solution on the other side. Recording commenced immediately after all monitors were set up, capturing fly activity at 1-min intervals over 4 days using DAMSystem311X software. The first recorded minute for analysis was the first minute of the subsequent hour after monitor setup (e.g., if monitor setup occurred at 5:40 p.m., the initial minute for analysis would be 6:01 p.m.). Raw data were preprocessed using the DAMFileScan113X program, and locomotion activity was analyzed using SCAMP. Survival probability was determined by checking activity levels every 6 hours, with flies considered deceased if no activity was detected within that time frame. Sixteen flies per group and multiple groups were examined for each genotype. Survival probability was calculated as the mean percentage of living flies within each group at each time point.

Lipid droplet staining

Larval or adult brains were dissected and fixed in 4% of paraformaldehyde (PFA)/phosphate-buffered saline (PBS) at room temperature for 30 min and washed in PBST (PBS, 0.3% Triton X-100) three times for 20 min each time. Fixed brains were incubated in the Nile red solution [Sigma-Aldrich, 19123; diluted 1:100 in PBST from a stock solution (100 mg/ml) in acetone] at 4°C overnight and then washed in PBST for 30 min each time. Larval brains were mounted on glass slides with the antifade mounting solution and imaged with a Zeiss LSM700 upright confocal microscope.

Immunohistochemistry

For whole-mount brain immunohistochemistry, more than 10 samples of each genotype and conditions were analyzed, and experiments were replicated two to five times. Larval and adult brains were dissected and fixed in 4% PFA/PBS at room temperature for 30 to 40 min, followed by washing in PBST and incubating in the primary antibody overnight at 4°C. On the next day, brains were washed with PBST and incubated in the secondary antibody at room temperature for 1 to 3 hours before final washes in PBST and mounting on the slide with the antifade mounting solution.

For immunohistochemistry on acutely dissociated brain cells, third instar larval brains were dissected and transferred to clean dish containing cold Dulbecco’s Phosphate Buffered Saline (DPBS) and were cut into smaller pieces by needles. After proteinase treatment [collagenase/dispase (1 mg/ml) and liberase I (0.1 Wünsch U/ml) for 40 min at 25°C], media neutralization, and centrifugation, cells were resuspended in 50 μl of Schneider’s insect medium and transferred onto chambered cell culture slides (VWR International, LLC, 53106-306). The chamber slides were pretreated with concanavalin A (0.25 mg/ml; Sigma-Aldrich). After 10 min at room temperature, extra solution was removed from the slides. Adhered cells were then fixed in 4% PFA in PBS for 10 min, then washed with PBST three times, and incubated in the primary antibody overnight at 4°C. On the next day, slides were washed with PBST and incubated in the secondary antibody at room temperature for 1 hour before final washes in PBST and mount with the antifade mounting solution after removal of the chamber top.

Primary antibodies used were rabbit anti-Obp44a (polyclonal antibody generated against peptide -CFLGKNLPLVQAAVQKN-, YenZym Antibodies, 1:500), mouse anti-repo (DSHB 8D12, 1:10), rat anti-elav (DSHB 7E8A10, 1:20), and rabbit anti–cleaved caspase-3 (Asp¹⁷⁵) (Cell Signaling, #9661, 1:100). Secondary antibodies (1:500 dilution) used were donkey anti-rabbit rhodamine (Jackson ImmunoResearch Labs, 711-295-152), goat anti-mouse Alexa Fluor 647 (Invitrogen, A-32728), and donkey anti-mouse CY3 (JacksonImmuno Research Labs, 715165150). Images are taken either with Zeiss LSM700 and LSM780 confocal microscopes in the laboratory or a Zeiss LSM800 confocal microscope at National Institute of Neurological Disorders and Stroke NINDS Neurosciences Light Imaging Facility.

Calcium imaging

Late L3 expressing Pdf-Gal4 driving UAS-GCaMP7f were used for calcium imaging experiments as described (58). Imaging was performed on a Zeiss LSM 780 confocal microscope equipped with a Coherent Vision II multiphoton laser. Larval brain explants were dissected in the external saline solution [120 mM NaCl, 4 mM MgCl₂, 3 mM KCl, 10 mM NaHCO₃, 10 mM glucose, 10 mM sucrose, 5 mM N-tris(hydroxymethyl)methyl-2-aminoethanesulfonic acid (TES), 10 mM Hepes, and 2 mM Ca²⁺ (pH 7.2)] and maintained in a chamber between the slide and cover glass and imaged with a 40× water objective using 920-nm excitation for GCaMP signals. GCaMP7f signals were collected at 100 ms per frame for 2000 frames during each recording session. Light stimulations of 100-ms duration were delivered using a 561-nm confocal laser controlled by the photo bleaching program in the Zen software. The laser power was set at 10%. GCaMP7f signals at the axonal terminal region of LNvs were recorded and analyzed. Average GCaMP7f signals of 20 frames before light stimulation was taken as F0, and ΔF (F − F0)/F was calculated for each time point. The average value of ΔF/F for individual brain samples were used to generate the average traces of calcium transients. The shaded area represents the standard error of the mean. The sample number n represents the number of individual animals.

Protein expression and purification

The expression and purification of ¹⁵N-labeled Obp44a in the apo state were carried out following the recently published protocols (44). The concentration was determined by the absorbance at 280 nm measured using a NanoDrop microvolume spectrophotometer (Thermo Fisher Scientific) and the predicted extinction coefficient of the protein. Of crucial importance is the use of the SHuffle Escherichia coli expression host to ensure proper disulfide pairings and fold of the protein, as well as the final HPLC purification step to remove any fatty acids bound to the protein during the expression.

NMR spectroscopy

The ¹⁵N-labeled Obp44a NMR samples (250 μl) were prepared by dissolving the protein in a lyophilized form in a 20 mM potassium phosphate buffer (pH 6.6), 0.5 mM EDTA, and 10% D₂O. The pH of each sample was checked and adjusted by the addition of 0.1 N NaOH. The protein concentration used was between 70 and 400 μM in a Shigemi tube (Shigemi Co. Ltd). All fatty acids were dissolved in either dimethyl sulfoxide (DMSO; C16:0, C18:0, and C18:2) or ethanol (C22:0) to create stock solutions with concentrations between 5.5 and 25 mM depending on their solubility. The final ratio of protein to fatty acid was 1:1.2 in all NMR samples.

The 1D ¹⁵N-edited proton spectra (105) were acquired at 298 K on a Bruker 600 MHz spectrometer equipped with a 5-mm triple resonance probe with triaxial gradient using a minimum of 256 scans, 2000 time domain points, 8-kHz spectral width, and 1-s recycle delay. The number of scans for each sample was adjusted relative to their concentrations to result in spectra of comparable signal to noise. Spectra were processed and plotted using the Bruker TOPSPIN program (Bruker NMR, Billerica, MA).

Gel shifting assay

Fatty acid–C16 (Thermo Fisher Scientific, D3821) was dissolve in DMSO to make 2.1 mM stock solution. A total of 0.5 μl of 2.1 mM FA-C16 was added to 100 μl of PBS to make a final concentration of 10.5 μM. Purified OBP44a protein was added to make final concentrations of 0, 1.77, and 3.54 μM respectively. After incubating at room temperature for 40 min in the dark, 2× tris-glycine native sample buffer (Thermo Fisher Scientific, LC2673) was added to each sample. Samples were then loaded on to two 10 to 20% tris-flycine gels (Thermo Fisher Scientific, XP10205BOX) in one mini gel tank and subjected to gel electrophoresis with native running buffer (Thermo Fisher Scientific, LC2672) in the dark. After electrophoresis, one gel was stained with 0.01% Coomassie Blue R250 and then scanned at a 700-nm channel using an Odyssey infrared scanner (LI-COR). Another gel was directly imaged at a Cy2 channel using a Typhoon laser scanner (GE Bioscience).

Mass spectrometry and metabolomic analysis

Liquid chromatography–tandem mass spectrometry

Metabolite extraction from Drosophila larval brains was executed following a modified cold methanol method as previously described (106). For each biological replicate, a pool of 100 brains from third instar larvae of either wild type (Canton-S) or Obp44a^−/− was dissected and combined in PBS buffer. Three independent biological replicates and three technical replicates were prepared for each genotype. To extract metabolites, the brain samples were homogenized in 200 μl of cold methanol/water solution (80/20, v/v) and subjected to gentle sonication using a Bioruptor instrument (30 s on, 30 s off, 10 cycles) at 4°C. Subsequently, the lysates were centrifuged at 10,000g for 10 min at 4°C, and the resulting supernatants were collected for liquid chromatography–tandem mass spectrometry (LC-MS/MS) analysis. LC-MS/MS experiments were performed using a combination of Zwitterionic hydrophilic interaction liquid chromatography (ZIC-HILIC) with an acetonitrile/water/7 mM ammonium acetate solvent system, coupled with high-resolution mass spectrometry. The analysis was conducted using a maXis-II-ETD UHR-ESI-Qq-TOF mass spectrometer (Bruker Daltonics) equipped with a Dionex Ultimate-3000 liquid chromatography system. The ZIC-HILIC column (2.1 mm by 150 mm) operated under mildly acidic pH conditions enabled efficient separation of the target metabolites. Global metabolic profiling was achieved by conducting three technical replicates for each of the three biological replicates per genotype, ensuring robust and comprehensive metabolite analysis.

LC-MS/MS metabolomic analysis

Metabolomic analysis of samples was performed using MetaboScape-2023 and MetaboAnalyst software, following established protocols (107). Compound identification based on the Bruker MetaboBase Personal 3.0, MoNA, MSDIAL, METLIN, and HMDB97 metabolomic libraries resulted in the annotation of 1371 of the total 3140 features detected. To evaluate data quality and variation, principal components analysis was conducted on the 1371 annotated features. A 3D plot was generated using Omicshare tools (https://omicshare.com/tools). Accurate mass measurements, with an accuracy of less than 5 ppm, and MS/MS spectra were both used for robust metabolite and lipid identification. In total, ~300 metabolites were identified to be present in both wild-type and mutant samples. Statistical significance between the three biological replicates of wild-type and mutant animals was assessed using the Student’s t test. Metabolites with a P value of less than 0.05 were considered significant. The data were visualized using various tools. A volcano plot was generated to illustrate the differential expression of metabolites using TBtools (108). Heatmaps of changed metabolites were also constructed using TBtools.

Protein sample collection and Western blot

To examine the Obp44a level in mutant and wild type (fig. S5C), five L3 larval brains were dissected and homogenized in 25 μl of lysis buffer [Neuronal Protein Extraction Reagent (N-Per) lysis buffer (Thermo Fisher, 87792) containing 0.1 mM dithiothreitol (DTT) and 1:100 diluted proteinase inhibitor cocktail (Sigma-Aldrich, P8340)].

To test the secretion of Obp44a (Fig. 6D), five Obp44a::GFP knock-in larval brains were dissected and introduced into 25 μl of external saline solution [120 mM NaCl, 4 mM MgCl₂, 3 mM KCl, 10 mM NaHCO₃, 10 mM glucose, 10 mM sucrose, 5 mM TES, 10 mM Hepes, and 2 mM Ca²⁺ (pH 7.2)] and incubated at room temperature for 0, 30, and 60 min. The solution was recovered, and the remaining brains were homogenized in 25 μl of lysis buffer. Samples were then incubated with protein sample buffer containing 10 mM DTT and heated to 95°C for 5 min before loading onto SDS–polyacrylamide gel electrophoresis (SDS-PAGE) for further analysis.

To examine Obp44a protein in larval hemolymph (Fig. 6E), collections from L3 larvae were performed following a method described in an online video resource (https://youtube.com/watch?v=im78OIBKlPA), with 20 larvae used to obtain hemolymph for each genotype. To avoid protein aggregation, collected hemolymph samples were incubated with protein sample buffer containing 10 mM DTT at room temperature for 30 min, followed by SDS-PAGE gel analysis.

After SDS-PAGE, the proteins were transferred onto a polyvinylidene difluoride membrane and then incubated with a 5% nonfat milk in TBST for 1 hour. Following the blocking, the membrane was incubated overnight at 4°C with primary antibodies, including rabbit anti-Obp44a (1:5000) or rabbit anti-GFP (Abcam, ab6556, 1:2000) or rabbit anti–α-tubulin (Abcam, ab15246, 1:2000). After three washes with TBST, the membranes were incubated with horseradish peroxidase–conjugated secondary antibodies (1:10,000) and incubated for 1 hour at room temperature. Membrane exposure was performed using a chemiluminescence detection kit (Bio-Rad, 1705062).

In vivo imaging of Obp44a trafficking with C12-red fatty acid feeding

Obp44a::GFP knock-in larvae were transferred to food containing C12-red fluorescent fatty acids (1 mg/ml; BODIPY 558/568 C12, Thermo Fisher Scientific, D3835) at early L3 stage (72 hours after egg laying). After feeding for 24 hours, larval brains were dissected and incubated in external saline solution [120 mM NaCl, 4 mM MgCl₂, 3 mM KCl, 10 mM NaHCO₃, 10 mM glucose, 10 mM sucrose, 5 mM TES, 10 mM Hepes, and 2 mM Ca²⁺ (pH 7.2)]. The explants were maintained in a chamber between the slide and cover glass and imaged with a 40× objective using a Zeiss 800 confocal microscope or a Nikon A1R confocal microscope for 30 to 90 min.

Single-sided axotomy of olfactory receptor neuron

To induce axon injury in adult flies, axotomy is performed as described with modifications (68, 71). Surgical ablation was carried out to remove antennal segments 2 and 3 on one side of the adult fly head, which resulted in the complete removal of olfactory receptor neuron cell bodies, Johnston’s organ, and full transection of the antennal nerve on the ablated side. Following the ablation, the injured flies were cultured at 25°C for recovery for 6, 24, and 48 hours, respectively. The dissected brains were fixed in 4% PFA for 30 min at room temperature, washed, and subsequently processed for mounting to examine the effects of axotomy on Obp44a::GFP localization.

Quantification and statistical analysis

Sample size determination, randomization, and blinding

Sample sizes were limited by the availability of animals with desired genotypes and are consistent with previous studies. Specific numbers for individual experiments are listed in the figure legends. Sample collection was randomized within each genotype and condition. Group allocation could not be blinded because of experimental requirements, but data quantification was performed blindly by coauthors. No data were excluded. Key findings were reproduced using independent genetic reagents and replicated by different laboratory members.

For lipid droplet density quantification, images were processed using Imaris 3D image visualization software. The Spots module of the Imaris was used to detect the number of lipid droplets within a 3D volume of 88.7 μm–by–88.7 μm frame with a thickness of 17 μm in the larval brain (Fig. 2D). The total number was exported from Imaris into Excel. The density was calculated by lipid droplet numbers divided by the volume of the quantified region. For adult brain (Fig. 2E), the lipid droplet number was manually calculated within the optic lobe OLC region with the thickness covered OLC. The total numbers in OLC were used for statistical analysis.

In vivo glutathione redox (GSH/GSSG) biosensor (mito-roGFP2-Grx1) redox level (Fig. 5E) was calculated by the mean intensity of 405-nm channel divided by the mean intensity of 488-nm channel in the whole optic lobe region with a frame of 15.8 μm by 15.8 μm (63). For each brain optic lobe sample, the top, median, and bottom three sections were used for quantification. A ratio image was created by dividing the 405-nm image by the 488-nm image pixel by pixel and displayed in false colors using the lookup table “Fire” in ImageJ.

For 3D reconstruction and quantification of mushroom body axonal debris, confocal images of the α-lobe were processed by Imaris 3D image visualization software. The Surface module of the Imaris was used to detect anti–Fas II signals and reconstruct a surface encompassing the entire debris region. Total volume and total intensity values were exported from Imaris and used to calculate mean intensity.

For quantification of Obp44a::GFP intensity in response to the antennal axotomy–induced injury, optic sections containing the AL and AMMC, where the highest GFP intensity was observed, were selected for each sample. Using the Histo function in Zen Elite software (Zeiss), the AL and AMMC regions (outlined in fig. S10) were manually delineated and used for analysis. Mean GFP intensity was measured on both the left and right sides of the brain, and the intensity ratio between the two sides was calculated and compared between control and injured groups.

Graphing and statistics analysis of the quantifications were performed using GraphPad Prism (9.4.1). For statistical analyses between two groups of samples, a two-tailed unpaired Student’s t test (unpaired t test with Welch’s correction) was performed; for experiments with more than two groups, one-way analysis of variance (ANOVA) followed by multiple comparisons, Tukey post hoc test, or Welch’s ANOVA test with Dunnett’s multiple posttests was performed. The exact value of sample number n, the statistical tests used, the confidence intervals, and the precision measures for individual experiments are included in the figure legends. Most quantitative data are presented as bar plot overlaid with dot plot; bar plot shows the mean (height of bar) and SEM (error bars); dot plot displays individual data points. n represents groups with 20 flies per group in Fig. 5A and fig. S7. n represents the fly numbers in Fig. 5 (C and D). n represents groups with 16 flies per group in Fig. 5F. n represents the number of larvae brains in Figs. 2D and 4D. n represents the number of adult brains in Fig. 2E and figs. S9C and S10. n represents the number of quantified sections in Fig. 5E (three sections per brain) and fig. S10. Statistical significances were assigned as: *P < 0.05, **P < 0.01, and ***P < 0.001. Schematic images are created in BioRender. Yin, J. (2025) https://BioRender.com/sc2v2lu.

Supplementary Materials

The PDF file includes:

Figs. S1 to S11

Legends for data files S1 to S5

Other Supplementary Material for this manuscript includes the following:

Data files S1 to S5