Seipin is the best-studied ER–LD junction protein. It is encoded by BSCL2 in humans and FLD1/SEI1 in yeast (Magre et al., 2001; Szymanski et al., 2007; Fei et al., 2008). Although the exact mode of action of seipin is not yet clear, recent studies have provided some mechanistic insights into the function of seipin. Seipin is an integral membrane protein of the ER that plays an essential role in LD formation and normal adipogenesis. Seipin contains short N- and C-terminal domains oriented toward the cytoplasm, two transmembrane domains, and a highly conserved large ER luminal domain (Szymanski et al., 2007; Cartwright and Goodman, 2012). The N-terminal domain of seipin is crucial for the initiation of LD formation in yeast, since deletion of this domain results in a seipin null phenotype (Cartwright et al., 2015). In humans, mutations in seipin result in lipodystrophy syndromes, a group of disorders that are characterized by a selective loss of adipose tissue (Magre et al., 2001), and seipinopathy, neurological disorders affecting motor neurons (Windpassinger et al., 2004; Cartwright and Goodman, 2012). Seipin localizes to ER–LD junctions in both yeast and mammalian cells (Szymanski et al., 2007; Fei et al., 2008; Salo et al., 2016; Wang et al., 2016). Absence of seipin from these sites results in growth-arrested LDs, manifesting in numerous tiny LDs, and a few supersized LDs, many of which are not functionally connected with the ER membrane, resulting in an altered LD surface proteome (Grippa et al., 2015; Salo et al., 2016; Wang et al., 2016).

In yeast, seipin forms a complex with Ldb16, a seipin partner protein with no identified mammalian homolog (Wang et al., 2014; Grippa et al., 2015). Ldb16 is an ER-localized integral membrane protein that interacts with seipin through its transmembrane helices (Wang et al., 2014). Deletion of either partner protein results in size heterogeneity of LDs. Expression of human seipin rescues this LD phenotype, suggesting that human seipin functionally complements the yeast seipin–Ldb16 complex (Wang et al., 2014).

Fld1 and Nem1 together define discrete ER subdomains from where LDs are being formed (Choudhary et al., 2020). These seipin/Nem1 subdomains become coenriched with DAG, as revealed by colocalization with a fluorescent ER-DAG sensor, and recruit additional LD biogenesis factors, such as FITM, Pex30, and the TAG-synthases, Dga1, and Lro1 to promote NL synthesis and formation of nascent LDs (Choudhary et al., 2020) (Figures 2A,B). Nascent LDs then undergo further growth and expansion via three distinct mechanisms: Synthesis of NLs by LD-localized acyltransferases (Sorger and Daum, 2002; Wilfling et al., 2013), flow of NLs from the ER through ER–LD junctions (Jacquier et al., 2011; Kassan et al., 2013; Wilfling et al., 2013), and transfer of NLs between adjacent LDs through a ripening process, i.e., transfer of TAG from smaller to larger LDs (Thiam et al., 2013; Salo et al., 2019). Mature LDs remain associated with seipin/Nem1 puncta in the ER (Figure 2B). Absence of any of these factors impedes droplet biogenesis and can result in ectopic TAG synthesis throughout the ER membrane and the formation of aberrant LDs (Choudhary et al., 2020) (Figure 2C).

Seipin preferentially enriches at ER tubules rather than ER sheets, owing to the higher curvature of the tubular ER domains, and thereby controls LD nucleation from these highly curved domains (Santinho et al., 2020). Such an enrichment of LD biogenesis factors at ER subdomains might modulate the local lipid composition, particularly that of DAG/TAG to promote TAG synthesis and/or control its nucleation by preventing the diffusion of NLs into the bulk of the ER membrane. Consistent with such a lipid gatekeeper function at ER–LD junctions, loss of seipin results in flow of NLs from tiny LDs via the ER membrane into large LDs causing the pronounced heterogeneity in LD size (Salo et al., 2019). In addition, seipin has been shown to regulate PA levels at ER–LD junctions (Han et al., 2015; Wolinski et al., 2015; Pagac et al., 2016) and to act as a scaffold for lipid-synthesizing enzymes (Talukder et al., 2015). ER–LD junctions have been analyzed by EM in various cell types, and they appear to be conserved (Jacquier et al., 2011; Wilfling et al., 2013; Ohsaki et al., 2016; Salo et al., 2016; Wang et al., 2016; Romanauska and Kohler, 2018). Interestingly, besides cytoplasmic LDs, cells can also generate LDs in the nucleoplasm. Formation of these nucleoplasmic LDs (nLDs) depends on TAG producing enzymes that are localized to the inner nuclear membrane (INM) (Romanauska and Kohler, 2018). More recently, seipin has been shown to restrain the formation of nLDs, and seipin-deficient yeast mutants show dramatic increase in nLD formation (Cartwright et al., 2015; Romanauska and Kohler, 2018). Although seipin appears to be absent from the INM, deficiency of seipin results in elevated levels of PA in the INM and an increase in lipin-1 expression, which is likely to promote the formation of nLDs (Soltysik et al., 2021).

On the other hand, hepatocytes generate abundant nLDs that are derived from apolipoprotein B (ApoB)-free luminal LDs, a process that is dependent on microsomal triglyceride transfer protein (MTP) (Soltysik et al., 2019). Under conditions of ER stress, ApoB levels are reduced due to co- and posttranslational degradation (Ginsberg and Fisher, 2009), whereas MTP levels are maintained (Qiu et al., 2005). This lack of ApoB-dependent NL secretion promotes the accumulation of ApoB-free luminal LDs, which grow inside the ER lumen and eventually disintegrate the INM leading to nLD formation (Soltysik et al., 2019). Thereby, nLD formation in hepatocytes may contribute to non-alcoholic fatty liver disease and its associated complications (Seebacher et al., 2020).

Seipin assembles into a ring-shaped oligomeric structure as revealed by recent cryo-EM studies of the luminal domain of human (undecamers) and fly (dodecamers) Seipin (Sui et al., 2018; Yan et al., 2018). Each monomer contains a hydrophobic helix that is apposed toward the ER bilayer, and a ß-sandwich domain with a structural similarity to lipid-binding proteins, such as Niemann-Pick type C2 (NPC2), a cholesterol-binding protein (Sui et al., 2018). This ER luminal ß-sandwich domain has affinity toward anionic phospholipids such as PA (Yan et al., 2018). Based on these findings, seipin was proposed to scan the ER membrane for small TAG lenses, where it could get anchored and facilitate further growth of LDs by transferring newly synthesized TAG from the ER membrane into the growing LDs (Sui et al., 2018). Furthermore, the seipin complex at the base of ER–LD junctions might create a selectivity barrier and thereby uncouples the lipid composition of these ER subdomains from the bulk of the ER. Based on molecular dynamic simulations, seipin has recently been proposed to directly interact with, and thereby enrich, TAG and DAG within its ring-shaped structure (Prasanna et al., 2020; Zoni et al., 2020). Hence, seipin might function as a lipid transporter and/or gatekeeper to locally regulate lipid levels at ER–LD junctions.

The lipid droplet organization (Ldo) proteins, Ldo16, and its splice variant Ldo45 were recently identified as genetic interactors of seipin (Eisenberg-Bord et al., 2018; Teixeira et al., 2018). Ldo16 and Ldo45 play important roles in regulating LD biogenesis at a spatially confined ER domain adjacent to the nuclear–vacuolar junction (NVJ), a large contact site between the nuclear ER and the vacuole, during nutritional stress (Eisenberg-Bord et al., 2018). Ldo45 promotes TAG accumulation and hence proliferation of LDs, whereas Ldo16 is necessary for efficient lipophagy, LD consumption by autophagy (Teixeira et al., 2018). Surprisingly, overexpression of Ldo45 leads to clustering of LDs, mimicking the seipin deletion phenotype, suggesting an antagonistic role of Seipin and Ldo45 in regulating LD formation and maturation (Eisenberg-Bord et al., 2018). Ldo45 facilitates the targeting of Pdr16, a Sec14-like phosphatidylinositol transfer protein, to LDs at NVJ (Eisenberg-Bord et al., 2018) (Figure 3A). This family of proteins has previously been shown to localize at organelle contact sites (Selitrennik and Lev, 2016). Surprisingly, Pdr16 is mislocalized in cells lacking Ldo45; however, double deletion of Ldo45 together with Seipin, rescues the targeting of Pdr16 to LDs (Eisenberg-Bord et al., 2018). However, the precise mechanism of Pdr16 targeting to a subpopulation of LDs, and its role in LD metabolism, are not yet completely understood.

Promethin/lipid droplet assembly factor 1 (LDAF1) is a remote homolog of Ldo45 and a seipin partner protein identified in humans and widely conserved across species (Eisenberg-Bord et al., 2018; Chung et al., 2019). In mammalian cells, seipin collaborates with promethin/LDAF1 to define discrete ER sites of LD biogenesis (Chung et al., 2019). Promethin/LDAF1 localizes to LDs upon fatty acid supplementation, and copurifies with seipin, suggesting that these two proteins form a complex (Castro et al., 2019). In an independent study, promethin/LDAF1 was identified as an interactor with the conserved hydrophobic helix of seipin (Chung et al., 2019). During oleate-induced LD biogenesis, seipin remains localized at ER–LD junctions, whereas LDAF1 loses its association with seipin foci in the ER, and moves onto the periphery of expanding LDs. These findings suggest that TAG accumulation results in disruption of the interaction between promethin/LDAF1 and seipin and that the protein can move from a bilayer onto a monolayer membrane (Chung et al., 2019). In cells devoid of LDs, most of the LDAF1 puncta colocalize with seipin foci, while free LDAF1 and seipin foci also coexist, but de novo LDs begin to form at sites that contain both LDAF1 and seipin, suggesting that both proteins play a key role in defining ER sites of LD biogenesis (Chung et al., 2019). In agreement with this, relocalization of LDAF1 to ER–plasma membrane contacts results in localization of seipin at these ectopic sites from where LDs can then form de novo (Chung et al., 2019). Protease protection assays indicate that both N- and C-termini of LDAF1 are oriented toward the cytoplasm, suggesting a hairpin type of topology with a central hydrophobic domain that is sufficient for both colocalization with seipin and association with the LD monolayer (Chung et al., 2019). Analyses of copurifying lipids indicate that the promethin/LDAF1–seipin complex, but not seipin alone, binds TAG. Interestingly, absence of seipin also results in loss of promethin/LDAF1, suggesting that previous studies analyzing seipin deficiency may, in fact, represent conditions where both partner proteins, seipin and LDAF1, were missing (Chung et al., 2019).

FITM proteins constitute an evolutionarily conserved family of polytopic ER-localized transmembrane proteins (Kadereit et al., 2008). Mammals express two FITM isoforms: FITM1 is mainly expressed in skeletal muscles, while FITM2 is expressed in most tissues. FITM2 plays a major role in the differentiation of preadipocytes into mature adipocytes. It is functionally regulated by peroxisome proliferator-activated receptor gamma (PPARγ), and controls the accumulation of TAG within LDs (Kadereit et al., 2008). While overexpression of FITM2 in 3T3-L1 adipocytes and in mouse liver in vivo results in accumulation of TAG-rich LDs, its deficiency leads to a decrease in both size and number of LDs (Kadereit et al., 2008). Purified FITM2 binds both DAG and TAG in vitro, but the protein itself has no TAG biosynthetic activity (Gross et al., 2011). Based on the observation that FITM proteins become enriched at ER–LD junctions upon LD formation, it has been suggested that they promote the condensation/sequestration of TAG in the ER bilayer and/or packaging of TAG into LDs (Kadereit et al., 2008; Gross et al., 2011).

The knockout of FITM2 in adipose tissue in mice results in lipodystrophy and insulin resistance (Miranda et al., 2014). In humans, a homozygous nonsense mutation in FITM2 was identified in a family with Siddiqi syndrome, characterized by deafness–dystonia syndrome, delayed development, and regression of motor skills, dystonia, ichthyosis-like features, signs of sensory neuropathy, and low body mass index (Zazo Seco et al., 2017). Interestingly, RNAi-mediated knockdown of the sole FITM2 in Drosophila melanogaster recapitulated many of the characteristic features of the human phenotype, suggesting a link between the mutation in FITM2 and the phenotypic syndrome (Zazo Seco et al., 2017). The causal relation between the defect in TAG formation or storage and these neurological syndromes in organisms lacking FITM function, however, remains to be characterized.

FITM2 is more widely conserved than FITM1, since Drosophila, Caenorhabditis elegans, and Saccharomyces cerevisiae have only orthologs of FITM2. Budding yeast even has two FITM2 homologs, Yft2 and Scs3. Yeast FITM2 mutants have defects in ER membrane homeostasis, and Scs3 mutant cells are auxotrophic for inositol (Hosaka et al., 1994; Moir et al., 2012). Lack of FITM2 proteins in yeast results in the formation of LDs that emerge toward the ER lumen instead of the cytoplasm, suggesting that these proteins are required to promote proper budding of nascent LDs from the ER (Choudhary et al., 2015). This function of FITM2 is conserved as depletion of FITM2 in human 3T3-L1 fibroblasts, and lack of the sole FITM2 in C. elegans results is a similar phenotype (Choudhary et al., 2015, 2016). Remarkably, deletion of FITM2 in C. elegans and mice is lethal, indicating that FITM2 fulfills a hitherto uncharacterized but essential function in lipid metabolism (Choudhary et al., 2015, 2016; Goh et al., 2015).

Sequence alignment suggests that FITM2 proteins could harbor a lipid phosphatase/phosphotransferase activity on the ER luminal side, and mutations of residues that are conserved among lipid phosphatases affect FITM2 function in vivo (Hayes et al., 2017). Consistent with this role, FITM2 becomes transiently enriched at LD biogenesis sites, where it may play an important role in regulating DAG levels, a critical prerequisite for proper LD formation (Adeyo et al., 2011; Choudhary et al., 2018). More recently, FITM2 was shown to harbor acyl-CoA diphosphatase activity and to hydrolyze fatty acyl-CoA to yield acyl 4′-phosphopantetheine (Becuwe et al., 2020). This enzymatic activity of FITM2 is important in maintaining proper structure of the ER membrane as lack of FITM2 results in the formation of ER whorls, and impacts both ER and lipid homeostasis, but it is presently not clear how exactly the lack of this enzymatic activity manifests in the many different phenotypes associated with mutations in FITM (Becuwe et al., 2020).

Pex30, a membrane-shaping protein has recently been shown to be important for proper LD formation. Pex30 contains an N-terminal reticulon homology domain (RHD), a hairpin-type membrane domain similar to that present in the reticulon family of ER-shaping proteins, and a C-terminal Dysferlin (DysF) domain (Joshi et al., 2016). Similar to reticulons, Pex30 tubulates the ER membrane. However, unlike the abundantly expressed reticulons, which show uniform ER distribution, Pex30 is a low abundant protein that localizes to discrete ER subdomains from which nascent LDs and pre-peroxisomal vesicles originate (Joshi et al., 2016, 2018). Pex30 remains associated with mature LDs at ER–LD junctions and does not localize to ER exit sites at which COPII-coated vesicles form (Joshi et al., 2016). Deletion of Pex30, on the other hand, results in a significant delay in the rate of LD biogenesis, proliferation of ER membranes, and defects in peroxisome biogenesis, suggesting a shared common mechanism in the biogenesis of these two organelles (Joshi et al., 2018; Wang et al., 2018; Santinho et al., 2020). Dysferlin containing proteins in higher eukaryotes play an important role in the repair of damaged membranes due to mechanical stress as occurs in skeletal muscle cells, giving rise to muscular dystrophies, suggesting that Pex30 may harbor similar membrane repair properties (Glover and Brown, 2007). These properties are possibly important for seipin function and probably in the repair of ER membrane at sites where LDs and peroxisomes form.

The human homolog of Pex30, multiple C2 domain containing transmembrane protein (MCTP), also contains an RHD (Joshi et al., 2018). While mammals express two MCTP isoforms, worms and flies have only one MCTP (Shin et al., 2005). Similar to Pex30, MCTP2 is a low abundant protein that localizes to ER subdomains at which nascent LDs are born and associates with peroxisomes in a dynamic fashion (Joshi et al., 2018). The curvature-inducing property of Pex30/MCTP2 might be important to deform the ER bilayer at LD biogenesis sites to accommodate DAG and/or TAG. Moreover, Pex30 colocalizes with other LD biogenesis factors, such as seipin and Nem1, at DAG-enriched ER subdomains (Joshi et al., 2018). Interestingly, in cells lacking seipin, Pex30 is mislocalized to a single punctum (Joshi et al., 2018; Wang et al., 2018). It is possible that Pex30 directly interacts with seipin, and absence of seipin from ER subdomains is likely to destabilize Pex30 localization. Remarkably, simultaneous deletion of both Seipin and Pex30 results in strong synergistic growth defect, impaired LD formation, and elevated DAG levels, which antagonizes LD emergence, resulting in TAG accumulation throughout the ER membrane (Joshi et al., 2018; Wang et al., 2018). These growth and LD emergence defects in double mutant cells lacking both Pex30 and seipin can be rescued by manipulating the lipid composition of the ER by deletion of the cholinephosphate cytidylyltransferase, Pct1, the rate-limiting enzyme of phosphatidylcholine biosynthesis through the Kennedy pathway, thereby reducing phosphatidylcholine, phosphatidylinositol, and DAG levels but increasing TAG levels (Wang et al., 2018).

Many proteins that dually target both the ER and LDs contain hydrophobic domains that adopt a V-shaped helical hairpin topology. How these hairpin-containing proteins are targeted to the LD surface, however, is not well understood. However, one such protein, UBX8, is localized to the ER with the help of two peroxisomal proteins. UBX8 is first captured by the soluble cytoplasmic protein Pex19, and subsequently inserted into the ER membrane at subdomains containing Pex3, from where it then translocates to the LD monolayer (Schrul and Kopito, 2016). Thus, Pex30-marked ER subdomains could mediate inter-organelle trafficking of proteins/lipids both between ER–LD and ER–LD–peroxisome contacts.

Mdm1 in yeast and its human ortholog Snx14 have recently been uncovered as molecular tethers at ER–LD junctions, providing novel insights into how these unique junctions are established and function in regulating LD biogenesis in response to metabolic demands (Datta et al., 2019; Hariri et al., 2019). Under starvation, LDs accumulate at spatially confined ER domains adjacent to the NVJ (Hariri et al., 2018). Mdm1 is a key NVJ component that is required for accumulation of LDs at NVJ during starvation and functions as a triorganellar molecular tether that connects the ER with LDs and the vacuole, thereby facilitating LD degradation (Hariri et al., 2018). Mdm1 is anchored in the ER via its N-terminal integral membrane domain, which is necessary and sufficient for targeting Mdm1 to LD biogenesis sites, and it targets the vacuolar membrane via a phosphatidylinositol 3-phosphate-binding Phox (PX) domain (Hariri et al., 2019). A third domain within Mdm1, termed PX-associated (PXA) domain, is sufficient for LD targeting; hence, Mdm1 can physically link the three compartments, the ER, LDs, and the vacuolar membrane (Hariri et al., 2019) (Figure 3A).

Mutations in the human ortholog of Mdm1, Snx14, lead to spinocerebellar ataxia autosomal recessive 20 (SCAR20) disease, which is characterized by intellectual disability, cerebellar atrophy, ataxia, and speech disorders (Thomas et al., 2014). Like Mdm1, Snx14 is an ER protein that relocalizes to ER–LD junctions upon oleate addition (Datta et al., 2019). Snx14 is anchored in the ER via its amino-terminal integral membrane domain and contains an amphipathic helix in the C-terminus, termed the C-Nexin domain, which is necessary for binding LDs in trans (Datta et al., 2019). Consistent with an ER–LD tethering function, overexpression of Snx14 dramatically increases ER–LD junctions, whereas a lack of Snx14 reduces these junctions (Datta et al., 2019). Though Mdm1 and Snx14 have the same domain organization, the PX domain in Snx14 lacks the residues required for phosphoinositide binding (Mas et al., 2014), consistent with the fact that NVJ contacts are absent in mammalian cells. Remarkably, both Mdm1 and Snx14 were found to cooperate with fatty acyl-CoA synthetases to locally increase concentrations of activated fatty acids for the formation of TAG at ER–LD junctions. Thus, apart from acting as tethers between organelles, these proteins could also provide a scaffold for the assembly of enzymes to channel the flux of metabolites toward energy storage. Absence of Mdm1 sensitizes cells to fatty acid-induced lipotoxicity, suggesting a key role in lipid homeostasis, whereas Snx14 mutants display gross LD morphology defects, implying that Snx14 is required for the formation of functional LDs. It will be interesting to determine whether Mdm1 and Snx14 interact with other LD biogenesis factors at ER–LD junctions to coordinate lipogenesis with LD expansion.

Rab18 belongs to the small GTPases family, which regulates membrane trafficking. Rab18 shows diverse intracellular localization and is found both at the ER and on LDs. Rab18 functions as a common mediator of lipogenesis and lipolysis in adipocytes (Pulido et al., 2011), and has been implicated in mediating the formation of ER–LD junctions (Ozeki et al., 2005). In agreement with a tethering function, overexpression of Rab18 greatly enhances ER–LD junctions possibly by decreasing Perilipin 2 levels on LDs (Ozeki et al., 2005). Rab18 interacts with the ER-associated NAG-RINT1-ZW10 (NRZ) tethering complex and their associated SNARES, resulting in the recruitment of ER to LDs and the formation of direct ER–LD junctions thereby promoting lipogenesis and LD growth (Xu et al., 2018). Even though the Rab18–NRZ complex functions to tether LDs to the ER, this function of Rab18 is not essential for maintenance of LD growth in all cell types (Jayson et al., 2018). Another partner protein of Rab18 is the double zinc finger FYVE domain-containing protein 1 (DFCP1), lack of which leads to decreased ER–LD junctions, whereas overexpression of DFCP1 results in increased junction formation (Li et al., 2019). Physiologically, mutations in Rab18 manifests in Warburg Micro syndrome, a collection of neurodevelopmental disorder characterized by optical atrophy, microcephaly, intellectual disability, and hypogenitalism (Bem et al., 2011; Dejgaard and Presley, 2019).

Upon activation of lipolysis, fatty acids that are released from LDs are oxidized by either mitochondria or peroxisomes to generate precursors for mitochondrial oxidative phosphorylation and energy production (Poirier et al., 2006). LDs thus form extensive contacts with these organelles to facilitate metabolite channeling of free fatty acids (Bohnert, 2020). In tissues that have a high-energy demand such as heart, skeletal muscle, brown adipocytes, and hepatocytes, pronounced contacts between LDs and mitochondria are observed (Wang et al., 2013; Boutant et al., 2017). These contact areas expand dramatically during starvation (Nguyen et al., 2017) or during endurance training of athletes (Shaw et al., 2008).

The factors that tether LDs and mitochondria are not yet fully characterized. Recently, tethering between the LD scaffolding protein Perilipin 1 (PLIN1) and Mitofusin 2 (MFN2), a mitochondrial outer membrane protein, has been observed in brown adipocytes (Boutant et al., 2017) (Figure 3B). Consistent with a tethering function during active lipolysis in brown adipocytes, PLIN1 co-precipitates with MFN2, and cells lacking MFN2 show reduced LD–mitochondrial contacts, impaired respiratory capacity, and reduced response to adrenergic stimulation (Boutant et al., 2017). Adipose-specific MFN2 knockout mice are protected from high-fat diet-induced insulin resistance and hepatic steatosis (Seebacher et al., 2020), suggesting that LD–mitochondrial interactions affect whole-body energy homeostasis (Boutant et al., 2017). Another perilipin family member, PLIN5, also localizes at LD–mitochondrial contacts (Gemmink et al., 2018) and is upregulated in oxidative tissues (Yamaguchi et al., 2006; Vigelso et al., 2016). However, its interacting mitochondrial partner protein remains yet to be identified.

Recently, a novel tether at LD–mitochondrial contacts containing Mitoguardin 2 (MIGA2) has been identified in white adipocytes (Freyre et al., 2019). Consistent with a tethering function of MIGA2, its overexpression in adipocytes greatly enhances LD–mitochondrial as well as ER–mitochondrial contact formation (Freyre et al., 2019). Overexpression of MIGA2 in cells that display very little LD–mitochondrial contacts, such as in COS7 cells, induces formation of extensive LD–mitochondrial contacts, suggesting that MIGA2 is sufficient to promote tethering of LD and mitochondria (Freyre et al., 2019). MIGA2 contains two N-terminal transmembrane domains that anchor the protein to the outer mitochondrial membrane, a middle domain, and a C-terminal amphipathic domain, which interacts with the LD surface (Freyre et al., 2019) (Figure 3C). Surprisingly, a truncated version of MIGA2 lacking both transmembrane domains, is unable to localize to LDs, despite the presence of the amphipathic domain (Freyre et al., 2019). Remarkably, the middle domain of MIGA2 contains a conserved peptide sequence called “two phenylalanines in an acidic tract” (FFAT) (Murphy and Levine, 2016), which binds VAMP-associated proteins (VAPs), a conserved family of integral ER membrane proteins, which mediate contact site formation between the ER and many different organelles, including the plasma membrane, peroxisomes, and the Golgi apparatus (Lev et al., 2008). MIGA2 thus facilitates tethering between three organelles, LDs, mitochondria, and the ER (Freyre et al., 2019). Importantly, lack of MIGA2 in preadipocytes results in impaired TAG synthesis, defects in LD growth, and blocks their differentiation into mature adipocytes, suggesting a crucial role of MIGA2 in coordinating and de novo lipogenesis in mitochondria to TAG production in the ER (Freyre et al., 2019). How exactly this coordination is achieved is presently not well understood, but mitochondria that are bound to LDs have unique bioenergetics, composition, and dynamics to support LD expansion (Benador et al., 2018).

As with the mitochondria, extensive contacts between LDs and peroxisomes have frequently been observed; however, little is known about the tethering components that establish these contacts (Novikoff et al., 1980; Binns et al., 2006; Thazar-Poulot et al., 2015). In yeast cells cultivated in the presence of oleic acid, which not only promotes rapid expansion of LDs but also induction of peroxisomes, extensive contacts between peroxisomes and LDs were observed (Binns et al., 2006). LDs isolated from such cells contained enzymes required for beta-oxidation of fatty acids, suggesting metabolic coupling between NL lipolysis in LDs and peroxisomal degradation of fatty acids. Consistent with such a metabolic coupling, free fatty acids accumulate on LDs in mutants defective in fatty acid beta-oxidation (Binns et al., 2006).

In mammalian cells, a tether complex comprised of the M1 spastin protein, which localizes to the LD surface, and the peroxisomal fatty acid transporter ATP-binding cassette subfamily D member 1 (ABCD1) (Figure 3C) has recently been uncovered (Chang et al., 2019). M1 spastin contains an N-terminal hydrophobic hairpin domain, which anchors it to the LD monolayer, a microtubule interacting and trafficking (MIT) domain, that selectively binds two ESCRT-III (endosomal sorting complex required for transport III) proteins, and a C-terminal AAA-ATPase domain (Chang et al., 2019). Localization of ESCRT-III at LD–peroxisomal contacts is crucial for fatty acid transfer, suggesting that ESCRT-III proteins promote fatty acid mobilization possibly by deformation of the LD membrane (Chang et al., 2019). Consistent with a role in tethering, overexpression of M1 spastin results in extensive LD–peroxisomal contacts in a microtubule-independent manner (Chang et al., 2019). Remarkably, M1 Spastin lacking the ATPase domain is still able to localize to LDs; however, it is unable to induce LD–peroxisomal contact formation (Chang et al., 2019). Spastin has been implicated in membrane remodeling and microtubule dynamics (Lumb et al., 2012). Mutations in spastin result in autosomal-dominant hereditary spastic paraplegia disease. Similarly, mutations in the peroxisomal fatty acid transporter protein, ABCD1, are associated with X-linked adrenoleukodystrophy and also manifest neurological symptoms such as spasticity (Morita and Imanaka, 2012). ABCD1 co-precipitates with Spastin or with its PXI domain alone, suggesting that these two proteins form a complex (Chang et al., 2019). Consistent with this, lack of ABCD1 abolishes LD-peroxisomal contacts, which cannot be rescued by overexpression of M1 Spastin, implying that the two partner proteins must be present for functional LD–peroxisomal contact formation (Chang et al., 2019). In agreement with this, overexpression of Spastin, enhances transfer of fatty acids via LD-peroxisomal contacts as revealed by pulse-chase experiments with a fluorescent fatty acid analog (Chang et al., 2019).

Taken together, contacts between LDs, on the one hand, and either mitochondria or peroxisomes, on the other hand, provide a novel means to channel the flux of metabolic intermediates and thereby integrate spatially segregated biochemical pathways. At the same time, these organellar contacts may play an essential role in controlling the progression of differentiation (Freyre et al., 2019). Alteration in LD contact formation due to mutations in molecular tethers are associated with different hereditary neurodevelopmental, neurodegenerative, and metabolic diseases (Herker et al., 2021). However, the exact molecular mechanism of the role of these LD contact site proteins in the etiology of these diseases is not well understood. Moreover, various intracellular pathogens, including bacteria, viruses, and parasites, are known to hijack the host LD machinery and induce dramatic rearrangement of LD contacts with cellular organelles for their own benefit and promote novel contact formation between LDs and the pathogen’s own replication organelle (Herker et al., 2021).