From as far back as Alois Alzheimer’s 1907 description of glial “adipose saccules”, numerous correlations have been made between LD accumulation in the brain and neurodegenerative diseases such as amyotrophic lateral sclerosis (ALS), Huntington’s disease, Parkinson’s disease and Alzheimer’s disease. The links between lipid metabolism, LDs and these neurodegenerative diseases have been discussed in detail in a number of recent reviews (Hamilton and Fernandes, 2018; Pennetta and Welte, 2018; Farmer et al., 2020; Haidar et al., 2021; Tadepalle and Rugarli, 2021; Teixeira et al., 2021). For some hereditary neurodegenerative conditions, causal links have been made to mutations in genes encoding proteins regulating LD biogenesis or turnover. For example, in the case of ALS caused by mutations in human Vesicle-Associated Membrane Protein (VAMP)-associated protein B (hVapB), combined human and Drosophila analyses implicate defective LD biogenesis as a contributory factor (Sanhueza et al., 2015; Pennetta and Welte, 2018; Farmer et al., 2020). Dominant mutations in an ER protein that regulates LD assembly, Seipin, can lead to seipinopathies including some forms of motor neuron disease (Windpassinger et al., 2004; Ito and Suzuki, 2009; Guo et al., 2013; Tadepalle and Rugarli, 2021). Related to this, the ER shaping factor Receptor Expression-Enhancing Protein 1 (REEP1) is required for forming appropriate numbers of LDs in the mouse brain and dominant mutations in this protein are associated with human hereditary spastic paraplegia (Züchner et al., 2006; Renvoisé et al., 2016). Furthermore, loss-of-function mutations in Sorting nexin 14 (Snx14), an ER-LD tethering protein, are associated with a form of spinocerebellar ataxia called SCAR20 (Datta et al., 2019; Datta et al., 2020). In the case of LD lipolysis, recessive mutations in a brain TAG lipase (DDHD Domain-Containing 2 (DDHD2)), which hydrolyzes LD core lipids, underlie a form of complex hereditary spastic paraplegia (Schuurs-Hoeijmakers et al., 2012; Inloes et al., 2018). Additionally, mutations in Huntingtin (Htt), a scaffold protein connecting the selective autophagy receptor p62 to LD cargo disrupt LD macroautophagy (lipophagy) and lead to Huntington’s disease (Rui et al., 2015). For many other neurodegenerative diseases, it is clear that lipid metabolism is perturbed but direct links between specific LD components and pathologies have not yet been made.

Many different stresses are known to induce LDs in the mammalian nervous system. In Schwann cells of the peripheral nervous system (PNS), infection by Mycobacterium leprae leads to myelin breakdown and Peroxisome Proliferator-Activated Receptor gamma (PPARγ)-dependent induction of LDs (Mattos et al., 2011; Díaz Acosta et al., 2018; Mietto et al., 2020). In the adult central nervous system (CNS), microglia accumulate LDs in response to innate inflammation, dietary high fat or low glucose, neurodegeneration, neuronal excitotoxicity or injury (Tamosaityte et al., 2016; Churchward et al., 2018; Chali et al., 2019; Ogrodnik et al., 2019; Raas et al., 2019; Marschallinger et al., 2020; Claes et al., 2021; Gouna et al., 2021; Zhuang et al., 2022). Microglia also accumulate LDs during ageing in the mouse and human brain, and this is associated with defective phagocytosis and a proinflammatory cell state (Marschallinger et al., 2020). Glial-like ependymal cells of the vertebrate CNS can also accumulate LDs in response to injury or a high fat diet (Enos et al., 2019; Maya-Monteiro et al., 2021). Moreover, astrocytes display increased LDs in vivo in response to a high fat diet and ex vivo/in vitro upon many different stresses including nutrient deprivation, hypoxia, excess fatty acids, γ-secretase inhibition, adrenergic receptor stimulation and neuronal excitotoxicity (Kwon et al., 2017; Ioannou et al., 2019; Ogrodnik et al., 2019; Gutierrez et al., 2020; Smolič et al., 2021). In the case of a high-fat diet, astrocytes of the hypothalamus that accumulate LDs also express proinflammatory cytokines and may therefore contribute to obesity-induced hypothalamic inflammation (Kwon et al., 2017).

In the invertebrate genetic model organism Drosophila, stress-induced LDs have been well characterized in both the developing and adult nervous systems (Bailey et al., 2015; Liu et al., 2015). In the developing CNS, LDs form predominantly in cortex and subperineurial glia, which constitute the niche for multipotent self-renewing neural stem cells called neuroblasts, and they increase following exposure to hypoxia or oxidant chemicals (Bailey et al., 2015; Kis et al., 2015; Dong et al., 2021). At adult stages, LDs are also present in glia of the CNS and increase during hypoxia (Smolič et al., 2021). In the adult retina, part of the PNS, several different genetic models of neurodegeneration lead to an increase in LDs in glial-like retinal pigment cells (RPCs) (Liu et al., 2015; Liu et al., 2017; Cabirol-Pol et al., 2018; Van Den Brink et al., 2018; Yeshaw et al., 2019; Girard et al., 2020; Muliyil et al., 2020). Sparse LDs have also been reported in Drosophila CNS and photoreceptor neurons and, in the latter, it is known that they increase in abundance in several neurodegeneration models (Van Den Brink et al., 2018; Wat et al., 2020; Girard et al., 2021). In both the mammalian and Drosophila nervous systems, a growing body of evidence indicates that a common feature of many of the stresses and pathologies that induce glial LDs is redox imbalance, which is associated with high levels of reactive oxygen species (ROS) (Bailey et al., 2015; Liu et al., 2015; Liu et al., 2017; Ioannou et al., 2019; Cheng et al., 2020; Muliyil et al., 2020).

LDs perform a myriad of context-dependent cellular functions beyond energy homeostasis, including the storage of vitamin and signalling lipid precursors, the suppression of ER stress and lipotoxicity, as well as the maturation, storage, turnover and quality control of proteins [reviewed in (Welte and Gould, 2017; Roberts and Olzmann, 2020)]. In glia, the accumulation of abundant LDs tends to correlate with the presence of stress and disease (Section 2.1). In principle, therefore, LDs could either be a driver or a mitigator (albeit not a 100% efficient one) of cellular dysfunction. Consistent with this, beneficial and harmful roles have been ascribed to stress-induced glial LDs, depending upon biological context. In addition to contextual differences, it is challenging to assign specific functions to LDs as few, if any, genetic or pharmacological manipulations are completely specific for this organelle and interpreting phenotypes is not always straightforward. Nevertheless, some of the more specific perturbations of glial LDs have targeted the enzymes catalyzing the biosynthesis and lipolysis of their neutral lipid cargos—CEs and TAGs.

In the developing mammalian brain, cholesterol is abundant and the majority of it is synthesized in oligodendrocytes and utilized in myelination (Dietschy, 2009). In the adult brain, however, most cholesterol is synthesized in astrocytes and it can be transferred to neurons in order to maintain axonal integrity (Dietschy, 2009; Mou et al., 2020; Staurenghi et al., 2021). In mouse models of Alzheimer’s disease, brain CEs—as well as TAGs—are elevated and LDs accumulate in forebrain ependymal cells of the neural stem cell niche (Chan et al., 2012; Yang et al., 2014; Hamilton et al., 2015). LDs in microglia are also implicated as a recent study showed that a genetic risk factor for Alzheimer’s, the apolipoprotein E4 (ApoE4) allele, increases their abundance and also alters microglial properties such as phagocytosis (Machlovi et al., 2022). In several Alzheimer’s models, the enzyme synthesizing the CE cargo of LDs—ACAT also known as sterol O-acyltransferase (SOAT)—has been blocked using genetic or pharmacological methods. In the context of human mutant amyloid precursor protein (APP) and triple-transgenic mouse models, ACAT inhibition is beneficial as it substantially reduces APP processing and the production of extracellular amyloid plaques (Hutter-Paier et al., 2004; Shibuya et al., 2014). Brain CEs can also be lowered indirectly by converting cholesterol to 24 (S)-hydroxycholesterol, which can then be secreted from cells and eliminated via the blood brain barrier (Moutinho et al., 2016). Consistent with this, a chemical activator of cholesterol 24-hydroxylase (CYP46A1) increased 24-hydroxycholesterol secretion from APP mutant iPSC-derived neurons (but not astrocytes) lowering CEs and increasing proteosomal degradation of phosphorylated Tau, a hallmark of Alzheimer’s Disease (van der Kant et al., 2019). Although it is not yet clear how lowering CEs decreases APP processing and phospho-Tau degradation, altered trafficking and autophagy in microglia and neurons rather than in ependymal cells are likely to be relevant (Puglielli et al., 2001; Shibuya et al., 2014; Shibuya et al., 2015; van der Kant et al., 2019). In summary, the ACAT and CYP46A1 manipulations suggest that biosynthesis of the CE cargo of LDs can be harmful, contributing to the pathogenesis of Alzheimer’s disease.

Several recent Drosophila and mammalian studies have blocked glial LD accumulation by targeting TAG metabolism, using DGAT1 inhibition or ATGL overexpression (Bailey et al., 2015; Liu et al., 2015; Van Den Brink et al., 2018; Nakajima et al., 2019; Muliyil et al., 2020; Smolič et al., 2021). Comparisons between these studies provide some useful insights into the roles of glial LDs. The two Drosophila studies using DGAT1 (Mdy) knockdown both reported that this method of blocking glial LDs leads to non-cell autonomous cellular dysfunction: in one context late-onset adult photoreceptor degeneration and, in the other, underproliferation of neural stem cells and increased ROS during hypoxia (Bailey et al., 2015; Van Den Brink et al., 2018). Similarly, in cultured mammalian astrocytes, DGAT1 and/or DGAT2 inhibitors were used to block LDs. This resulted in a concomitant decrease in astrocyte cell number, suggesting that LD biosynthesis is important for glial proliferation and/or cell survival (Nakajima et al., 2019; Smolič et al., 2021). The outcomes of all four Drosophila and mammalian studies that inhibited DGAT1/2 are therefore consistent in showing that glial biosynthesis of TAGs can be beneficial in diverse contexts.

Three adult Drosophila retinal studies have utilized ATGL (Bmm) overexpression to boost TAG lipolysis and thus delete LDs. One found that this manipulation increases age-dependent photoreceptor degeneration (Van Den Brink et al., 2018). In contrast, the two other studies observed that ATGL overexpression substantially rescues photoreceptor degeneration in retinal cells mutant for a metalloprotease [A Disintegrin And Metalloproteinase Domain-Containing Protein 17 (ADAM17)/Tumor Necrosis Factor (TNF)-Alpha Converting Enzyme (TACE)] or for mitochondrial components [Mitofusin, Nicotinamide adenine dinucleotide (NADH) Dehydrogenase (ubiquinone) 42 kDa subunit (ND-42) or Methionyl-transfer Ribonucleic Acid (tRNA) synthetase, mitochondrial (MetRS-m)], leading the authors to conclude that, in these contexts, glial LDs promote neurodegeneration (Liu et al., 2015; Muliyil et al., 2020). However, a reinterpretation of this conclusion was recently suggested by systematic side-by-side comparisons of DGAT1 knockdown and ATGL overexpression, not in glia but in Drosophila renal cells (Lubojemska et al., 2021). This study showed that, although both manipulations efficiently block LD accumulation, the former is harmful whereas the latter is beneficial for cell function. Based on these and other findings, it was argued that overexpression of the lipid-droplet resident enzyme ATGL equates to a gain, not a loss, of an LD subfunction, enhancing the ability of the LD to stimulate TAG lipolysis (Lubojemska et al., 2021). This may also be the case in glia, such that DGAT1 and ATGL work in the same not opposite “directions” to promote a beneficial flux of fatty acids through the TAG compartment. More generally, the comparisons of DGAT1 and ATGL manipulations in different glial contexts illustrate that assigning an overall protective or a harmful role to LDs can be confusing and, at best, is an oversimplification. Instead, it may be useful to adopt a more nuanced approach, parsing the individual subfunctions of LDs using specific manipulations that avoid targeting more general aspects of lipid metabolism such as fatty acid synthesis or uptake.

A growing body of evidence is now shedding light on the mechanisms by which LDs in glia, and in other cell types, function to protect against lipotoxicity and redox imbalance (high ROS) during metabolic stress and disease (Ioannou et al., 2019; Bailey et al., 2015; Cheng et al., 2020; Lubojemska et al., 2021; Bensaad et al., 2014; Nguyen et al., 2017; Ackerman et al., 2018; Islam et al., 2019; Taïb et al., 2019; Liu et al., 2021). These protective LD roles appear to be intimately linked with mitochondria via at least four non-mutually exclusive mechanisms, whose relative importance is likely to be context dependent (Figure 1B). First, LDs in glia can provide an efficient conduit for delivering lipids via lipolysis or lipophagy to mitochondria for β-oxidation, in order to prevent fatty acid accumulation to toxic levels and/or to generate adenosine triphosphate (ATP) and ketone bodies (Rambold et al., 2015; Schulz et al., 2015; Ioannou et al., 2019; Wu et al., 2020). Second, this lipid trafficking route may help to minimize the potentially toxic effects of oxidized polyunsaturated fatty acids (PUFAs) via the LD core acting to protect against lipid peroxidation (Bailey et al., 2015; Li et al., 2018) or to sequester already peroxidated lipids (Liu et al., 2015; Ioannou et al., 2019). Third, in the context of mouse embryonic fibroblasts and glioma cells, LDs have been shown to act as a lipid buffer that is not required to deliver fatty acids to mitochondria but to sequester them, thus preventing acylcarnitine accumulation and lipotoxic dysregulation of mitochondria (Nguyen et al., 2017; Cheng et al., 2020). Fourth, LDs are also known, at least in non-neural contexts, to generate lipid signals that promote mitochondrial biogenesis and function (Haemmerle et al., 2011; Najt et al., 2020; Lubojemska et al., 2021). Hence, ATGL lipolysis at the surface of LDs can release fatty acids that activate the nuclear receptor PPARα, a partner of PPAR_γ Coactivator-1α (PGC1α), either directly or via the Sirtuin 1 deacetylase (Haemmerle et al., 2011; Najt et al., 2020; Lubojemska et al., 2021).

Distinct from a role in mitochondrial regulation, glial LDs can also regulate the activity of the intercellular signalling protein Hedgehog (Hh). In cortex glia of the developing Drosophila CNS, a proportion of the total Hh protein colocalizes with markers of the LD surface (Dong et al., 2021). Glial knockdown of a Drosophila Perilipin called Lipid storage droplet-2 (Lsd2), a negative regulator of ATGL, prevented glial overexpressed Hh from mediating an anti‐proliferative effect on neighbouring neural stem cells (Dong et al., 2021). It is therefore possible that the association of Hh with LDs modulates its secretion and/or activity.

Glia are known to secrete many types of lipids including cholesterol, fatty acids, phospholipids and phosphoglycolipids [reviewed in (Ralhan et al., 2021; Lee et al., 2021; Lane-Donovan et al., 2014; Mahley, 2016)]. These lipids are bound to extracellular proteins such as ApoE and can be taken up by neurons via low‐density lipoprotein (LDL) receptors or fatty acid transporters. They are known to be essential for the maintenance of multiple aspects of neuronal function, including membrane homeostasis, neurite outgrowth and intracellular signalling. Importantly, lipids can also be transferred in the reverse direction, from neurons to glia. A series of elegant papers (Ioannou et al., 2019; Liu et al., 2015; Liu et al., 2017; Van Den Brink et al., 2018; Muliyil et al., 2020; Moulton et al., 2021) used Drosophila and mammalian models of redox imbalance and neurodegeneration to demonstrate that metabolically stressed neurons deliver potentially toxic fatty acids to glia (Figure 2) [reviewed in detail in (Ralhan et al., 2021)]. In the context of Drosophila photoreceptor neurons, genetic knockdowns of mitochondrial components such as ND-42 generate redox imbalance, which activates c‐Jun‐N‐terminal Kinase (JNK) and Sterol Regulatory Element Binding Protein (SREBP) and stimulates the synthesis of fatty acids (Liu et al., 2015; Liu et al., 2017; Van Den Brink et al., 2018; Moulton et al., 2021). These fatty acids then become peroxidated in the presence of high ROS, exported from neurons via ATP-binding cassette transporter A (ABCA) transporters and transferred via an apolipoprotein D (ApoD) orthologue, Glaz, to glial-like RPCs. In RPCs, lipidated Glaz is thought to be taken up by an ApoD/E receptor, Lipoprotein Receptor-related Protein 1 (Lrp1), and fatty acids then trafficked in a clathrin-dependent manner via fatty acid transport protein (FatP) to be esterified via DGAT1 into TAGs stored in LDs (Liu et al., 2017; Van Den Brink et al., 2018; Moulton et al., 2021). Similarly, in a mammalian neuron-astrocyte coculture model, excitotoxicity was used to induce redox imbalance and increase autophagy, leading to neuronal production of excess free fatty acids (Ioannou et al., 2019). These fatty acids are then transferred to astrocytes via an ApoE and clathrin-dependent mechanism, where they are likely trafficked via the brain-specific fatty acid-binding protein, Fabp7, into LDs (Liu et al., 2017; Ioannou et al., 2019; Islam et al., 2019; Moulton et al., 2021; Qi et al., 2021). Together, the mammalian and Drosophila studies show that glia take up fatty acids form neurons and this can protect them from lipotoxicity. In this context, glial LDs can play a neuroprotective role, sequestering potentially toxic or peroxidated lipids (Liu et al., 2017; Ioannou et al., 2019; Moulton et al., 2021).

Neural stem and progenitor cells (NSPCs), like other stem cells, are regulated by many different aspects of lipid metabolism [reviewed in (Hamilton and Fernandes, 2018; Knobloch and Jessberger, 2017; Harkins et al., 2021; Madsen et al., 2021)]. Although NSPCs have been extensively characterized, there are very few reports of LDs in these cells in vivo in physiological wildtype conditions. However, in the embryonic mouse brain, conditional knockout in NSPCs of squalene synthase, an enzyme of cholesterol biosynthesis, results in LD accumulation and this correlates with the apoptosis of newborn neuronal progeny (Saito et al., 2009). These mutant embryonic NSPCs also upregulate vascular endothelial growth factor, although it is not clear if this process is linked to LDs (Saito et al., 2009). In the NSPC niches of the adult mammalian brain, LDs have mostly been described in niche cells such as ependymal cells not in the progenitors themselves [reviewed in (Ralhan et al., 2021; Hamilton and Fernandes, 2018)]. Recently, however, a study of the adult mouse brain found that NSPCs in the subventricular zone (SVZ) and dentate gyrus (DG) niches express the LD marker gene perilipin 2 (plin2) and, when cultured in vitro, they accumulate abundant Plin2⁺ LDs (Ramosaj et al., 2021). In cultured SVZ NSPCs, Plin2⁺ LDs are smaller in size during the proliferative than the quiescent (non-dividing) state. Furthermore, Plin2⁺ LD content per NSPC varies and correlates positively with oxygen consumption and extracellular acidification rates as well as with proliferative ability (Ramosaj et al., 2021). NSPCs with more abundant LDs also tend to have higher ROS levels, although not an increase in lipid peroxidation. This suggests that LDs in NSPCs could safeguard PUFAs, as they are reported to do in glia of the developing Drosophila CNS (Bailey et al., 2015; Ramosaj et al., 2021). Interestingly, genetic or pharmacological knockdown of ATGL in cultured NSPCs increased LDs and led to a decrease in proliferation (Ramosaj et al., 2021). Conditional knockdown of fatty acid synthase (Fasn) in SVZ or DG NSPCs has the opposite effect on LDs, decreasing them, yet it also impairs proliferation (Knobloch et al., 2013; Ramosaj et al., 2021). Conversely, a Fasn gain-of-function mutation associated with a human cognitive disorder leads to an accumulation of TAGs and ER stress, again impairing the proliferation of DG NSPCs (Bowers et al., 2020). It is therefore tempting to speculate that fatty acid flux through the TAG compartment promotes NSPC proliferation. The beneficial role of TAG lipolysis in NSPCs could therefore be related to that observed in Drosophila glia (Section 2.3).

The effector mechanisms by which ATGL activity influences NSPC properties remain to be identified. One possibility is that TAG lipolysis provides an efficient route for delivering fatty acids to mitochondria for β-oxidation in order to fuel oxidative phosphorylation (Section 2.2; Figure 1B). For cultured SVZ NSPCs, however, the validity of this explanation is not yet clear, in part because there is no consensus on the contribution that fatty acid β-oxidation makes to overall oxygen consumption rate (Stoll et al., 2015; Ramosaj et al., 2021). Nevertheless, fatty acid import into mitochondria does play an important role in SVZ and DG NSPCs as strong pharmacological inhibition of a key enzyme in this pathway, carnitine palmitoyltransferase 1a (CPT1a), decreases their proliferation (Stoll et al., 2015; Knobloch et al., 2017). In addition, pharmacological and genetic approaches indicate that maintenance of the quiescent state of DG NSPCs in vitro and in vivo also requires CPT1a and, by inference mitochondrial β-oxidation (Knobloch et al., 2017). A connection between LDs and β-oxidation may also be important for NSPCs in the embryonic neocortex (Xie et al., 2016). In this context, LD lipolysis, carnitine biosynthesis and CPT1a are all required to maintain the pool size of Paired Box 6 (Pax6)⁺, T-box transcription factor Eomes/Tbr2⁺ neural stem cells. Carnitine biosynthesis and CPT1a were also shown to function in the balance between self-renewing and differentiative divisions and to maintain the mitochondrial redox balance of embryonic neural stem cells (Xie et al., 2016). Together, the available data suggest that, under physiological conditions, fatty acid flux through the LD compartment of NSPCs acts to promote mitochondrial β-oxidation, in turn regulating multiple stem cell properties including the cell division mode and proliferative state.

Neurons in non-pathological and unstressed conditions tend to contain few if any LDs in vivo. An important question is why this is the case, given that neurons (and most other cell types) can form LDs in vitro when cultured under appropriate conditions. At least part of the explanation lies in the greater propensity of glia, ependymal cells and microglia to take up and process extracellular brain lipids (Sections 2.2 and 2.3). This in vivo “lipid sink” role has been mimicked in transwell co-cultures, where ectopic LDs in hippocampal neurons from ApoE3 or ApoE4 humanized mouse models of Alzheimer’s disease are cleared by astrocytes via ApoE-dependent extracellular lipid transport (Qi et al., 2021). Cell-intrinsic metabolic processes also make an important contribution towards preventing LDs from accumulating in neurons. For example, neurons express at least two different TAG lipases—ATGL and DDHD2—and their loss-of-function or chemical inhibition can lead to ectopic LDs in mammalian, Drosophila and C. elegans neurons (Inloes et al., 2014; Inloes et al., 2018; Yang et al., 2020a; Wat et al., 2020). In C. elegans, it has also been shown that mutations in an Abhydrolase Domain-Containing Protein 5 (ABHD5)/Comparative Gene Identification-58 (CGI-58) orthologue, a known co-activator of ATGL, or overexpression of DGAT1/2 orthologues leads to ectopic LDs in neurons (Yang et al., 2020a). Similarly, some perilipins protect LDs from lipolysis, and overexpression of either the Lsd-1 or Lsd-2 perilipins in Drosophila photoreceptors results in a large increase in the usually sparse LDs in these neurons (Girard et al., 2021). Collectively, these findings provide evidence that neurons do not usually accumulate LDs in vivo, because they actively turnover TAGs, favouring lipolysis over biosynthesis Figure 2. This raises the important general question of how neurons and other cell types regulate their rates of neutral lipid synthesis and lipolysis. In the case of TAGs in hepatocytes, an ER and LD-associated protein called hypoxia inducible lipid droplet associated (HILPDA) may contribute towards coordinating synthesis and lipolysis rates as it both stimulates DGAT1 and inhibits ATGL (de la Rosa Rodriguez et al., 2021). It is not yet clear whether or not HILPDA functions in similar way in neurons but it is known to be expressed and strongly hypoxia-inducible in human primary astrocytes (Allen et al., 2020). In neurons, neutral lipid turnover is likely to be beneficial for their function, at least during unstressed homeostatic conditions (Inloes et al., 2014; Inloes et al., 2018).

Neurons accumulate ectopic LDs in several neurodegenerative diseases and during ageing (Shimabukuro et al., 2016; Farmer et al., 2020; Conte et al., 2021). In sporadic and familial forms of Parkinson’s disease, α-synuclein in neurons aggregates into inclusion bodies known as Lewy bodies [reviewed in (Stok and Ashkenazi, 2020)]. It is linked to multiple aspects of lipid metabolism in complex ways, associating with LDs and directly binding phospholipids and unsaturated fatty acids [reviewed in (Teixeira et al., 2021; Roberts and Olzmann, 2020)]. The toxicity of α-synuclein in human induced pluripotent stem cell (iPSC) neurons likely involves unsaturated fatty acids as it is ameliorated by inhibitors of stearoyl-CoA desaturase (Vincent et al., 2018; Fanning et al., 2019). In the yeast S. cerevisiae, α-synuclein inhibits growth and this is rescued by inhibition of Pah1, a Lipin phosphatidate phosphatase, suggesting that diacylglycerol synthesis is harmful, contributing to toxicity (Soste et al., 2019). A closer functional link to TAG metabolism is suggested by a Drosophila model of Parkinson’s disease where human α-synuclein is expressed in photoreceptor neurons (Girard et al., 2021). Coexpression of the perilipin Lsd2 induces LDs in photoreceptors, which recruit α-synuclein to their surface and increases the proportion of protease-resistant α-synuclein, a characteristic associated with α-synuclein aggregation and neurodegeneration (Cremades et al., 2012; Suzuki et al., 2015; Girard et al., 2021). In a related α-synuclein photoreceptor model, co-expression of ATGL with α-synuclein decreased the protease-resistant fraction (Girard et al., 2021). Together, these findings suggest that TAG lipolysis in Drosophila photoreceptor neurons inhibits the formation of a toxic form of α-synuclein, thus playing a protective role. TAG lipolysis is also beneficial during the recovery of PNS neurons from optic nerve injury in mice, although it remains unclear if LDs accumulate (Yang et al., 2020b). Nevertheless, in this context, neuronal regeneration requires ATGL and DDHD2 but is inhibited by DGAT1/2 (Yang et al., 2020b). Given that phospholipid synthesis enzymes can also facilitate regeneration, it may be that redirecting fatty acids away from TAGs into membrane lipids is beneficial for neuronal regrowth (Yang et al., 2020b; Roy and Tedeschi, 2021). TAG lipolysis in neurons is not, however, universally beneficial. In a C. elegans genetic model of excitotoxicity, LDs accumulate in neurons during their degeneration but this is rescued by inactivation of ATGL or ABHD5/CGI-58 and worsened by C20 PUFA incorporation into phospholipids (Yang et al., 2020a). Hence, in this excitotoxic model, TAG lipolysis in neurons exacerbates their degeneration, perhaps because it redirects PUFA into membranes where they are vulnerable to peroxidation. ATGL-dependent lipolysis also appears to be detrimental in cultured mammalian motor neurons expressing Seipin N88S, a mutant associated with a dominant spastic paraplegia that localizes to LDs and induces ER stress (Holtta-Vuori et al., 2013). In summary, there appears to be no universal truth about whether TAG lipolysis in neurons protects or harms from stress or disease—it all depends upon biological context.

LDs are a common feature of the developing and adult nervous systems of vertebrates and invertebrates. They have been observed in essentially all major cell types of the nervous system, albeit to differing degrees and in some cases only in stress or disease contexts. Under physiological conditions in vivo, LDs primarily accumulate in glia, ependymal cells and microglia. However, even in the absence of detectable LDs, cells such as neurons are still actively turning over neutral lipids. LDs in the nervous system become more numerous as a hallmark of several neurodegenerative diseases and also as a consequence of stresses involving redox imbalance and lipotoxicity. LDs in the nervous system during health and disease participate in multiple complex functions, which are dependent upon the cell type that they accumulate in. It is often difficult to make conclusions about whether LDs overall are beneficial or harmful. A more useful approach is to assign functions to individual biochemical reactions that are directly linked to LDs, such as TAG synthesis or lipolysis.

Looking forwards, a central challenge is to understand how lipid metabolic networks become wired differently in glia, neural stem cells and neurons, and how this influences adaptation to stress and disease. An important part of addressing this issue will be to determine how stresses change the LD proteome differently in each cell type of the nervous system. Rapid advances in technologies such as spatially-resolved and single-cell transcriptomics and proteomics are likely to help greatly in this quest (Armand et al., 2021; Goto-Silva and Junqueira, 2021; Maniatis et al., 2021). In addition, comparisons between the LD proteomes of glia, ependymal cells, microglia and neurons may also be facilitated by genetically encoded strategies that provide cell-type specific proximity labelling using Ascorbate Peroxidase 2 (APEX2) or related enzymes (Bersuker et al., 2018). A surface analysis technology from the physical sciences, mass spectrometry imaging, is an approach that promises to offer lipidomics and perhaps even proteomics with single-cell resolution in brain tissue sections (Gilmore et al., 2019; Taylor et al., 2021). Several different mass spectrometry imaging (MSI) platforms have already been used to spatially resolve amino acid, protein, lipid and LD metabolism, where they are beginning to deliver exciting new insights into regional and cell-type specific features of brain metabolism (Steinhauser et al., 2012; Bailey et al., 2015; Narendra et al., 2020; Newell et al., 2020; Wang et al., 2022).