Lipid droplets are the universal cellular organelles for the transient or long-term storage of lipids. The number, size and composition of lipid droplets vary greatly within cells in a homogenous population as well as in different cell types. The variability of intracellular lipid-storage organelles reflects the diversification of lipid droplet composition and function. Lipid droplet diversification results, for example, in two cellular lipid droplet populations that are prone to diminish and grow, respectively. The aberrant accumulation or depletion of lipids are hallmarks or causes of various human pathologies. Thus, a better understanding of the origins of lipid droplet diversification is not only a fascinating cell biology question but also potentially serves to improve comprehension of pathologies that entail the accumulation of lipids. This Commentary covers the lipid droplet life cycle and highlights the early steps during lipid droplet biogenesis, which we propose to be the potential driving forces of lipid droplet diversification.
The storage of lipids is a universal feature of cells and organisms, which evolved as a mechanism that allowed survival by buffering energy fluctuations. Within cells, lipids are stored in specialized organelles called lipid droplets. All lipid droplets share the same structure – a hydrophobic oil core of the storage lipids, which mainly comprise triacylglycerols (TAG) and sterol esters, is shielded by a phospholipid monolayer that contains specific proteins (Ohsaki et al., 2014). The lipid droplet life cycle (Fig. 1) includes lipid nucleation at the endoplasmic reticulum (ER) membrane, a growth phase that results in the potential budding and detachment of the nascent lipid droplet, followed by persistence of the lipid droplet in the cytosol before it potentially degenerates owing to the remobilization of the stored lipids by cytosolic lipases (Lass et al., 2011) or autophagy, and/or lipophagy (Cingolani and Czaja, 2016).
Throughout these steps, lipid droplets exhibit diversity. Within the past few years, several lines of evidence have demonstrated the existence of diverse multifunctional lipid droplet populations based on their distinct protein or lipid compositions (for examples, see Beller et al., 2006; Hsieh et al., 2012; Martin et al., 2005; Spandl et al., 2011; Wilfling et al., 2013). How and why lipid droplets diversify is not only an intriguing cell biology question given that lipid droplet diversity undoubtedly also has an important effect on cell and organism physiology. For example, a subset of cellular lipid droplets has been found to be prone to growth (Wilfling et al., 2013) and is thus likely to remobilize to a lesser extent. Thus, identifying lipid droplet diversification mechanisms might provide a better understanding of the pathophysiological conditions that are marked by altered lipid storage dynamics, which result, for example, in either excessive loss of lipids (Arner and Langin, 2014; Das et al., 2011; Patni and Garg, 2015; Robbins and Savage, 2015) or an increase in their storage (Byrne, 2013; Gaggini et al., 2015; Montani et al., 2004; Shulman, 2014; Sztalryd and Kimmel, 2014).
Lipid droplets are the only organelles that are surrounded by a phospholipid monolayer. As such, cytosolic lipid droplets can be considered an intracellular emulsion (Thiam et al., 2013b) whereby the hydrophobic lipid phase is dispersed in the hydrophilic cytoplasm in the form of droplets protected by a phospholipid monolayer. The monolayer that encloses lipid droplets thus has unusual properties compared to bilayer membranes, and the monolayer controls the selective binding of proteins to the lipid droplet surface (Thiam et al., 2013b). This key feature of lipid droplets has become an important point of focus in the lipid droplet field as the proteins that are bound to lipid droplets define lipid droplet identity, regulation and function. Still, how proteins selectively bind to lipid droplets remains poorly understood. Knowing the basics of emulsion physics (Thiam and Forêt, 2016; Thiam et al., 2013b) offers a unique angle to study and answer this question. Emulsion science has provided for some time a good understanding of the thermodynamic stability of droplet dispersion that is stabilized by surfactants, such as phospholipids or proteins. The use of approaches that employ emulsion science can thus be expected to have a tremendous impact on lipid droplet biology, similar to the impact of liposomal reconstitutions on membrane biology.
This Commentary summarizes the currently known examples of lipid droplet diversification and links lipid droplet diversity to the lipid droplet life cycle. Based on the available data, we discuss the potential regulatory mechanisms of protein binding to lipid droplets that allow the biogenesis of multifunctional lipid droplets (Pol et al., 2014; Welte, 2015).
Are all lipid droplets the same?
The answer is simply no. Everyone who has ever looked at the lipid droplets of different cell lines or tissues quickly realizes that lipid droplets vary within and between cell types in terms of their spatial organization, number and size (Fig. 2) (for example, see Gocze and Freeman, 1994; Herms et al., 2013; Szymanski et al., 2007). Drosophila S2 cells, for example, carry many small lipid droplets of equal size (Krahmer et al., 2011) (Fig. 2A). By contrast, most other cultured cells, such as monkey COS-7 kidney or murine AML12 hepatocytes (Tschapalda et al., 2016) (Fig. 2B) harbor an array of numerous and differently sized lipid droplets. The same is true for most organs in vivo, as is for example the case for the cells from the Drosophila fat storage organ, the fat body (Beller et al., 2010b). The mammalian white adipocyte cells are probably the most extreme with regard to the way fat is stored. In these cells, a singular lipid droplet fills up the entire cytoplasm (Cushman, 1970), and even pushes the nucleus into the periphery (Fig. 2C). In addition to differences in lipid droplet size and number, the intracellular positioning of lipid droplets can also vary significantly. Cells can exhibit either dispersed (Fig. 2A), clustered (Fig. 2D) or polarized (Fig. 2E) pools of lipid droplets (Beller et al., 2010a) depending, for instance, on the cell type and metabolic state.
In addition to their appearance, the protein and lipid composition of lipid droplets can also vary between different cell types. In terms of lipids, the core of most lipid droplets is enriched with TAG and some sterol ester (Bartz et al., 2007). However, the specialized function of certain cell types is reflected by the lipid composition of their lipid droplets. Examples include the enrichment of stellate cell lipid droplets with retinyl esters (D'Ambrosio et al., 2011) and of foamy monocyte lipid droplets with the uncommon fatty acid cis-7-hexadecanoic acid (Guijas et al., 2016), as well as the accumulation of polyunsaturated fatty acids – such as linoleic acid in larval (Bailey et al., 2015) and peroxidated lipids – in glia cells of adult Drosophila (Liu et al., 2015). In terms of the lipid-droplet-associated proteins, a core set of about 50 proteins has been found to decorate lipid droplets in various cellular systems (reviewed in Hodges and Wu, 2010). However, a diverse and large number of additional proteins has been identified across many studies (a selection of studies is given in Beilstein et al., 2013; Beller et al., 2006; Cermelli et al., 2006; Chen et al., 2016; Currie et al., 2014; Dahlhoff et al., 2015; Ivashov et al., 2013; Khan et al., 2015; Khor et al., 2014; Krahmer et al., 2013; Li et al., 2016; Na et al., 2013; Schmidt et al., 2013; Su et al., 2014; Zhang et al., 2011; Zhu et al., 2015), suggesting a significant amount of variability in terms of the lipid droplet protein composition between lipid droplets isolated from different cell types.
Even within a single cell, some of the lipid droplet proteins might only associate with a subset of lipid droplets. Indeed, several examples of differential protein targeting have been revealed by fluorescence microscopy imaging of the localization of distinct proteins (Fig. 2F). For example, the annotated short chain dehydrogenase/reductase CG2254 of Drosophila targets lipid droplet subsets in Drosophila tissue culture cells and larval fat bodies (Beller et al., 2006). The triacylglycerol synthesis enzymes diacylglycerol acyltransferase 2 (DGAT2) and glycerol-phosphate acyltransferase 4 (GPAT4) (Fig. 2G) (Wilfling et al., 2013) are also found to localize to a subset of lipid droplets, which keep growing under feeding conditions and overgrow the non-decorated lipid droplets (Wilfling et al., 2013). In contrast, localization of the small GTPase Rab18 has been shown to mark lipolytic lipid droplets (Fig. 2G) (Martin et al., 2005; Pulido et al., 2011). Furthermore, different members of the founding lipid-droplet-associated protein family, the so-called perilipins (Greenberg et al., 1991; Kimmel et al., 2010), associate with distinct sets of lipid droplets. For instance, as lipid droplets increase in size during mammalian adipocyte differentiation (Wolins et al., 2006), the perilipins PLIN3 and PLIN4, PLIN2 and PLIN1 sequentially bind to them (Wolins et al., 2005). The different mammalian perilipin proteins have been suggested to have an important role in either promoting the biogenesis of lipid droplets (for example, PLIN3; Bulankina et al., 2009) or in the stabilization and regulated remobilization of lipid droplets (for example, PLIN1; Kimmel and Sztalryd, 2016). Thus, the localization of specific perilipins provides a clue as to whether a given lipid droplet is in a transient (e.g. growing) or persistent state. The potential usefulness of perilipin localization as a diagnostic marker has been recently demonstrated in hepatosteatosis and hepatic cancerogenesis. For those diseases, the localization of PLIN1 to large lipid droplets, which are probably more difficult to remobilize, has been suggested to serve as a prognostic marker for the progression of steatosis to steatohepatitis, liver cirrhosis and cancer (Straub et al., 2008, 2010).
Differences in the lipid composition of individual lipid droplets within cells have also been described (Fig. 2H). For example, coherent anti-stokes Raman scattering (CARS) microscopy has revealed that lipid droplets within human adipose-derived stem cells and microalgae have different lipid constituents (Cavonius et al., 2015; Di Napoli et al., 2016). Furthermore, in McArh7777 rat liver cells, lipid droplets are either enriched in TAG or cholesterol esters (Hsieh et al., 2012).
The observed variation and potential discrepancies among proteomics and lipidomics studies of lipid droplets that have been isolated from various cellular systems might thus arise from any experimental and/or technical differences, or could reflect a biologically meaningful lipid droplet diversification between different cell types or tissues. To distinguish between these possibilities, lipid droplets should be purified from different cell types or tissues and processed and analyzed under identical technical, metabolic and physiological conditions. To investigate lipid droplet diversification within singular cells, the protein and lipid composition of purified lipid droplet populations should be analyzed. The ability to sort lipid droplets to high purity by using fluorescent markers and a modified fluorescence-activated cell sorting (FACS) methodology has been very elegantly demonstrated (Hsieh et al., 2012). Although this methodology has not yet been adapted extensively, it provides a powerful means to catalog lipid droplet diversification at the level of the protein and lipid composition.
Why are lipid droplets diverse?
The biological meaning of lipid droplet diversification is still unclear. In fact, there is unlikely to be a single generally valid explanation, as lipid droplets are multitasking organelles, which in addition to their energy storage function, fulfill different roles depending on the cell type and its physiological state (Pol et al., 2014; Welte, 2015). In terms of the variability of lipid droplet number in cells of a given population, it has been proposed that the lipid droplets of some cells act as ‘sponges’ by absorbing excess fatty acids that have been esterified into TAGs (Herms et al., 2013). Such a protective mechanism potentially also explains the accumulation of lipid droplets in glial cells during oxidative stress (Bailey et al., 2015; Liu et al., 2015). A protection from lipotoxicity could also explain the existence of small lipid droplets, which are prone to disappearing, in cells that also include growing lipid droplets (Wilfling et al., 2013). Such small transient lipid droplets could be part of a cellular ‘overflow’ mechanism that becomes important under conditions of elevated levels of free fatty acids, for example during the postprandial phase or in pathological scenarios.
Differences in the size and positioning of intracellular lipid droplets could help to achieve particular lipid droplet functions. The unilocular appearance of the large lipid droplet in mammalian white adipocytes probably relates to the primary function of the droplet in long-term lipid storage. Here, the reduced surface-to-volume ratio presumably helps to stabilize and protect the lipid store from improper remobilization. Upon stimulation of lipolysis, however, a number of small lipid droplets arise through the esterification of fatty acids that are liberated by the lipolytic activity. These lipid droplets could form to increase the total surface area that is accessible, for instance, to lipases to further enhance lipolytic activity (Ariotti et al., 2012) or to auxiliary proteins that were absent on the large lipid droplet, as well as possibly to mitochondria to optimize the channeling of liberated fatty acids into the β-oxidation pathway. Such a lipolysis-induced spatial repositioning of lipid droplets to mitochondria by transport along microtubules has indeed been demonstrated in Vero and COS-7 cells (Herms et al., 2015). Alternatively, the small lipid droplets could be simply formed to prevent lipotoxicity due to the sudden increase in an excess of free fatty acids. Both interpretations of the appearance of small lipid droplets during lipolysis are still under debate (Hashimoto et al., 2012; Paar et al., 2012).
When do lipid droplets diversify?
Although several pathways have been identified that are linked to the question of ‘how and when do lipid droplets diversify?’, the understanding of the underlying mechanisms is still lagging behind. Lipid droplet diversification can potentially arise at different steps of the lipid droplet life cycle (Fig. 1) – during lipid droplet emergence, in the steady state or over the course of remobilization. In the following, we present the different stages of the lipid droplet life cycle in detail and discuss possible steps that are crucial for lipid droplet diversification.
Lipid droplet emergence
Step 1 – lipid synthesis and nucleation
Lipid droplet biogenesis starts with the synthesis of neutral lipids by enzymes that are localized at the ER. The newly formed neutral lipids – for instance TAGs, which are the most prevalent – are transferred into the intermembrane space of the ER. When the neutral lipid concentration reaches a critical level, it becomes thermodynamically favorable that the molecules coalesce to nucleate to a lipid lens. Indeed, recent electron microscopy data clearly show the presence of such lipid lenses of around 50 nm in size (Choudhary et al., 2015). However, it is unknown whether the formation of these lenses occurs stochastically and/or on controlled sites of the ER. In fact, both mechanisms are possible and could potentially result in the biogenesis of different lipid droplets.
The nucleation of lipid lenses can be described by a phase de-mixing phenomenon, which is controlled by energetic constraints in response to the increased accumulation of TAGs in the intermembrane space of the ER (Fig. 3). Lipid nucleation thus requires the need to overcome an energy barrier. Several factors contribute to the energy barrier – the energetic costs of interactions between homo- and heterotypic molecules (i.e. interactions among TAG molecules versus interactions between TAGs and, for example, phospholipids), the thermal energy, the propensity entropy to disperse TAG molecules in the bilayer and the concentration of TAGs (Thiam and Forêt, 2016). This energy barrier might be lower in some ER sites, and these would therefore constitute the preferential sites for lipid droplet formation (Thiam and Forêt, 2016). If the energy barrier is not met, the rapid and efficient packaging of lipids into droplets will fail and, subsequently, the TAG content within the ER bilayer will increase, ultimately resulting in the formation of relatively big lipid droplets.
Defined nucleation sites with a lower energy barrier will constitute the main sites of lipid droplet formation as long as the TAG concentration in the bilayer remains below a critical concentration. This concentration is affected by the rate of TAG synthesis, the abundance and/or activity of the DGAT enzymes (Harris et al., 2011), which catalyze the final step of TAG synthesis, or the distribution of phospholipids in the membrane. At high TAG concentrations, aberrant or alternative nucleation sites could arise owing to thermal fluctuations; this could occur, for instance, in oleate-loaded cells or under pathological conditions that are marked by highly elevated levels of free fatty acids.
A nucleated lipid lens can enlarge (Fig. 3) by either acquiring more TAG generated through ongoing lipogenesis, through lateral fusion of lipid lenses that are still in the ER bilayer (Thiam and Forêt, 2016) or by a ripening process, which involves the transfer of lipids between lenses and/or nascent lipid droplets (Thiam and Forêt, 2016; Thiam et al., 2013b). This growth of the lipid lenses will result in a budding – i.e. the accumulated lipids become spherical – into a nascent lipid droplet of about 100–300 nm in size (Fig. 3). During all these steps, the droplet shares its surrounding phospholipid monolayer with the ER. Intriguingly, however, the phospholipid monolayer composition of purified lipid droplets is different from the phospholipid composition of the ER membrane (Tauchi-Sato et al., 2002; Zanghellini et al., 2010). Currently, it is unknown whether only specific phospholipids are recruited to the surface of nucleated and/or nascent lipid droplets, whether different lipid droplets are nucleated in different phospholipid environments of the ER and whether lipid droplet biogenesis occurs at sites where the ER membrane composition is altered. The concomitant occurrence of these mechanisms could be a source of lipid droplet diversity.
Nascent lipid droplets will keep growing to become mature lipid droplets – i.e. lipid droplets that reach steady state. Whether lipid droplets remain in contact with the ER at this stage, or ultimately pinch off either spontaneously or through the action of specific proteins, is not yet clear. This might also vary between different cell types and the physiological and metabolic characteristics of the given cell. Because more and more ER-derived proteins are known to target lipid droplets, it is crucial to address this question to better understand whether the identity or function of a given lipid droplet, which is tied to the composition of its proteome, is fixed or can be dynamically adjusted through the exchange of proteins between the ER and lipid droplets.
Step 2 – regulation of the early steps of lipid droplet biogenesis
The energy barrier that controls lipid droplet nucleation is very likely to be lowered at regions of the membrane where the two phospholipid monolayers of the bilayer do not adhere or are partially ‘unzipped’. Common examples of such situations are membrane bends (Fig. 4A). Here, the increased exposure of the hydrophobic interior induces stress on the membrane (Thiam and Forêt, 2016). The accumulation of TAG molecules in these defect regions will be favorable as it relieves the overall membrane stress. Nucleation sites could thus coincide with membrane bends (Fig. 4A). One way to induce similar defects is the insertion of proteins into the membrane that generate membrane curvature or hydrophobic mismatches (McMahon and Boucrot, 2015) – i.e. the total length of the hydrophobic acyl chains of the phospholipids is smaller than the bilayer thickness (Fig. 4B,C). As DGATs are the final enzymes in TAG synthesis, they are good candidates to generate lipid droplet nucleation sites through membrane bending. Most organisms, in fact, encode two DGAT isoforms (Harris et al., 2011). The isoforms possess different structural features and only show partially redundant functions (Liu et al., 2012). DGAT1 is an integral membrane protein of the ER (McFie et al., 2010), whereas membrane association of DGAT2 is likely to be caused by a hairpin structure that mediates only a partial insertion into the membrane (McFie et al., 2014; Stone et al., 2006). These different modes of membrane localization could both result in hydrophobic mismatches and membrane curvature, which can give rise to local defects that are favorable for TAG accumulation (Fig. 4B,C).
The structural differences among the DGAT proteins, as well as any potential functional differences that are not related to TAG synthesis, could not only affect lipid droplet formation but could also be the reason for the occurrence of lipid droplet subtypes (Wilfling et al., 2013). For instance, the ability of DGAT1 to laterally diffuse within the membrane is likely to be much smaller than that of DGAT2 owing to its integral membrane insertion (Fig. 4B,C). Therefore, it is possible that DGAT1-mediated TAG synthesis is locally constrained. In contrast, the spatial flexibility of DGAT2 would result in a much more disperse distribution of TAG within the ER intermembrane space. Along those lines, primary and secondary lipid droplet nucleation sites might exist. Primary lipid droplet nucleation sites would provide controlled lipid droplet biogenesis and are marked by the presence of the lipogenic machinery and additional proteins, which potentially facilitate the regulated biogenesis of lipid droplets and the controlled association of proteins. By contrast, secondary sites are sites of spontaneous events along the ER membrane, marked by a low nucleation energy barrier, which facilitates the appearance of lipid lenses in an unregulated manner (Thiam and Forêt, 2016). Owing to stochastic protein distribution, the lipid droplets that arise at such secondary random sites are likely to comprise a different set of proteins as compared to those lipid droplets that originate at defined primary sites. This stochastic component might also result in variations in the phospholipid or cargo lipid composition, which ultimately gives rise to different sets of lipid droplets.
Three other ER-localized protein families profoundly impact lipid droplet formation. The seipin protein (also known as FLD1 and SEI1 in yeast, and BSCL2 in mammals) appears to regulate ER organization and lipid droplet distribution, and additionally impacts lipogenesis that is coupled to lipid droplet formation (Bi et al., 2014; Cui et al., 2011; Fei et al., 2008, 2011; Wang et al., 2016; Wolinski et al., 2011). Individuals that lack functional seipin are affected by a severe form of lipodystrophy called Berardinelli–Seip congenital lipodystrophy (Magré et al., 2001). Further, the ER-localized acyl-CoA synthetase 3 (ACSL3) is rapidly recruited to sites of lipid droplet biogenesis, where it is likely to be important for stabilizing lipid droplet nucleation and lipid storage (Kassan et al., 2013). The growth of the lipid lens additionally depends on the action of the fat-storage-induced transmembrane (FIT) proteins, which are needed for the budding of lipid droplets from the ER membrane (Choudhary et al., 2015). Accordingly, a loss of FIT function in yeast and higher eukaryotes causes lipid droplets to remain in the ER membrane (Choudhary et al., 2015). Loss of function of the single FIT-encoding gene in Caenorhabditis elegans is lethal, suggesting that lipid droplet formation is a vital process in this organism (Choudhary et al., 2015).
The exact mechanisms by which these proteins affect lipid droplet formation sites remain to be elucidated. Nevertheless, a local lowering of the nucleation barrier and stabilization of the nucleated droplets at these sites, as recently proposed for seipin (Wang et al., 2016), could involve a remodeling of the ER membrane or a change of the lipid distribution – e.g. through a direct interaction with TAG or phospholipids.
Step 3 – protein association with nascent lipid droplets
Many proteins localize to lipid droplets through the ER (for examples, see Kory et al., 2016; Thiel et al., 2013; Zehmer et al., 2009), and it is thus likely that nascent lipid droplets recruit cytosolic or ER proteins during their emergence (Fig. 5). The site of lipid droplet nucleation, either at primary (predetermined) or secondary (random) ER sites as outlined above, can be also a determining factor of the protein coat composition of lipid droplets through multiple non-exclusive mechanisms.
With regard to lipid droplet nucleation at secondary sites, protein association is predominantly regulated by coincidental contact between a putative lipid droplet protein and the nascent droplet, as well as the overall biophysical properties. For example, association of a protein with the lipid droplet could be favored through a hydrophobic interaction with the lipid droplet oil core or the curvature of the forming lipid droplet (Fig. 5A). So far, there is only little information regarding the importance of the composition of the lipid droplet oil core for protein binding. In this context, the selective targeting of PLIN4 to lipid droplets containing cholesterol esters and of PLIN5 to those that have TAGs, could provide support for this notion (Hsieh et al., 2012). At the level of nascent lipid droplets, a number of proteins could potentially have access to a lipid droplet, but proteins with lower affinities could be displaced by competition or crowding effects (Kory et al., 2015) (Fig. 5B). This could result in the sequential formation of lipid droplets that comprise distinct pools of proteins. During a lipogenic burst with excessive lipid droplet formation on secondary sites, such a mechanism would, however, be less relevant for affinity-based lipid droplet diversification given that a hypothetically sufficient lipid droplet surface area is present to host most available proteins.
Primary ER lipid droplet nucleation sites are potentially enriched in multiple proteins, such as DGAT1, seipin, FITs and/or ACSL3 (see above). Additionally, yet to be identified, curvature-inducing proteins could locally further minimize the energy barrier that is necessary to stabilize lipid droplet nucleation. In that manner, they could control the propensity of lipid droplet formation and affect the recruitment of additional proteins to nascent droplets. This could be achieved through the formation of a protein diffusion barrier that simply shields the lipid droplet nucleation site, or by generating membrane properties that are unfavorable for diffusion (Fig. 5C). Such a mechanism would allow the binding of cytosolic proteins to nascent lipid droplets. Indeed, several proteins target lipid droplets without having any prior membrane association, for example PLIN1 (Brasaemle et al., 1997). Finally, ER proteins present at lipid droplet biogenesis sites could transiently interact with transmembrane proteins that either promote their retention in the ER during lipid droplet formation or promote lipid droplet targeting (Fig. 5D). This mechanism, for example, is proposed to mediate the localization of the hepatitis C virus core protein to lipid droplets through interaction with the ER-localized DGAT1 enzyme (Herker et al., 2010).
Protein-induced lipid lens formation can also result in secondary lipid droplet nucleation sites. For example, damaged and/or unfolded proteins in the ER can expose hydrophobic amino acid residues to the ER intermembrane space, which results in a partial unzipping of the membrane and in ER membrane stress. This then would favor the accumulation of neutral lipid molecules and, subsequently, the formation of lipid droplets. Furthermore, these damaged proteins could be incorporated into the forming lipid droplets and subsequently degraded through microlipophagy, for example, thus representing an ER protein quality control pathway (Vevea et al., 2015).
Steady-state lipid droplets – differential protein targeting to existing lipid droplets
Several examples of the differential protein targeting to larger and cytosolic droplets exist. Among these, the above-described localization of distinct perilipin proteins to lipid droplets of increasing size during mammalian adipocyte maturation (Wolins et al., 2006) stands out. However, the basis of the binding variability of the different perilipins is currently unknown. Although PLIN3 and PLIN2 have been proposed to localize to nascent lipid droplets in the ER, the specific membrane association mechanism for PLIN4 and PLIN1 is unknown. The exchange of the ER-localized perilipins PLIN3 and PLIN2 with PLIN1 could be linked to the translation of PLIN1 on free ribosomes in combination with its degradation in the absence of lipid droplets (Servetnick et al., 1995; Xu et al., 2006). Growth of the lipid droplets – with only limited protein replenishment occurring at the same time – could result in the availability of lipid droplet surface area, which facilitates a co- or post-translational targeting of PLIN1; this, in turn, might displace already localized PLIN2 and PLIN3 owing to the higher affinity of PLIN1 for the lipid droplet phospholipid monolayer. As an alternative, the observed sequential recruitment of proteins could be based on an enzyme-mediated remodeling of the phospholipid monolayer, which would result in altered biophysical properties that are potentially recognized by the given protein. Of course, such differences in biophysical properties could also already arise during lipid droplet nucleation owing to the particular properties of the primary or secondary nucleation sites (see above).
With regard to a secondary remodeling of the lipid droplet coat, only one mechanism has so far been described in detail. A number of independent studies have revealed a role of the coat protein complex I (COPI) machinery in the regulation of the amount of stored lipids (Beller et al., 2008; Guo et al., 2008; Soni et al., 2009). COPI is best known for its function in the retrograde trafficking of proteins and lipids from the Golgi back to the ER (Beck et al., 2009) and in the anterograde trafficking of small cargo between Golgi stacks (Pellett et al., 2013). Mechanistically, COPI activity has been shown to control the targeting of key lipid storage regulators, such as the adipose triglyceride lipase (ATGL; also known as PNPLA2) (Beller et al., 2008; Soni et al., 2009) or the lipogenic enzymes GPAT4 and DGAT2 (Wilfling et al., 2014), to the surface of lipid droplets. This is achieved through a modification of the lipid droplet surface composition by COPI-mediated budding of nano-lipid droplets from the surface of lipid droplets. In vitro experiments with isolated artificial lipid droplets have demonstrated that COPI alone is sufficient to bud nano-lipid droplets of 60 nm in diameter (Thiam et al., 2013a). The budding of nano-lipid droplets alters the phospholipid composition of the lipid droplet, which in turn results in an increased surface tension (Thiam et al., 2013a) that subsequently promotes the formation of bridges between cytosolic lipid droplets and the ER. Proteins and lipids can travel between the two compartments through these bridges allowing, for instance, the secondary recruitment of lipogenic enzymes to existing mature lipid droplets (Wilfling et al., 2014). Intriguingly, COPI activity is restricted to a subset of lipid droplets and, accordingly, lipogenic enzymes also only reach a subset of lipid droplets. How COPI activity is targeted to selected lipid droplets is currently unknown.
Remobilizing lipid droplets – protein targeting to lipolytic lipid droplet subsets
The small GTPase Rab18 specifically targets a subset of lipid droplets following stimulation of lipolysis (Martin et al., 2005; Pulido et al., 2011). Its localization to lipid droplets also enhances basal lipolysis rate in murine adipocytes (Pulido et al., 2011). Interestingly, the insulin-mediated activation of lipogenesis also triggers localization of Rab18 to lipid droplets in murine 3T3-L1 cells, and overexpression of Rab18 further enhances the basal lipogenesis rate (Pulido et al., 2011). Thus, Rab18 has a role in lipogenesis and lipolysis (Pulido et al., 2011). How Rab18 is targeted to subsets of lipid droplets is currently unknown. Insulin-mediated localization, however, has been found to depend on the phosphorylation of Rab18 by phosphatidylinositol-3-kinase (PI3K) (Pulido et al., 2011), suggesting post-translational modifications play an important role.
As mentioned above, targeting of the lipase ATGL to lipid droplets depends on COPI (Beller et al., 2008; Soni et al., 2009), whose activity appears to be restricted to certain subsets of lipid droplets. At the same time, COPI also mediates the targeting of lipogenic enzymes, such as GPAT4 and DGAT2, to subsets of lipid droplets (Wilfling et al., 2014). Thus, like Rab18, the COPI machinery appears to have paradoxical activities in both lipogenic and lipolytic processes. However, the proposed COPI action on lipid droplets involves multiple steps (budding of nano-lipid droplets, alteration of surface tension, formation of monolayer–bilayer bridges between the lipid droplets and the ER), which raises the possibility that other – yet to be identified – factors could mediate its switch between lipogenesis- and lipolysis-promoting effects. The same might be true for Rab18 function.
So when do lipid droplets diversify? Although lipid droplet diversification is clearly tied to the progression of the lipid droplet life cycle – in particular to the growth of lipid droplets and their secondary editing by factors such as COPI – we believe there is convincing evidence that the foundation for lipid droplet diversification, as well as the fate of lipid droplets, has already been determined during lipid droplet biogenesis. Thus, important questions to answer next are whether indeed ‘primary’ and ‘secondary’ lipid droplet biogenesis sites exist and, if so, whether they employ dedicated protein machineries that control the emergence of a lipid droplet and the associated proteins, as well as how lipid droplets are targeted by secondary editing machineries such as COPI.
Discussion and perspectives
Lipid droplet plasticity has a big impact on lipid physiology during health and disease. The emerging appreciation that not all lipid droplets are equal thus promises new insights into the regulation of lipid storage. Solving the important questions of how, when and why lipid droplets within a cell and between cells diversify, will result in a much better understanding of the regulation of the lipid droplet life cycle.
Lipid droplet diversification within a cell is probably the least characterized. So far, COPI is the only factor known to affect the localization of proteins to subsets of lipid droplets. However, COPI a priori only functions at already formed lipid droplets. Thus, a paramount question is whether the COPI machinery recognizes and selects certain lipid droplets to act upon stochastically or by the sensing of a particular signal. If such a specification signal exists, the next question is when and how it is imprinted. In this Commentary, we highlight the early steps of lipid droplet biogenesis at the ER as a possible time point for lipid droplet diversification and specification. The observation that the size of nascent lipid droplets can vary in response to the amount of available free fatty acids (Kassan et al., 2013) and the fact that lipid droplets can differ very early after biogenesis in terms of their protein or lipid composition support the notion that lipid droplets diversify during their emergence. As outlined above, a number of mechanisms can potentially result in the differential targeting of proteins to newly formed lipid droplets (Fig. 5), which would result in different lipid droplet protein coats. These varying protein coats would not only result in a functional diversification of lipid droplets, but such differences could also be recognized by secondary lipid droplet editing machineries such as COPI.
Another question is how stable are certain lipid droplet subsets? Although individual droplets interact with each other to exchange lipids (Gong et al., 2011), a translocation of proteins from one lipid droplet to another has not been described to the best of our knowledge. Furthermore, our own photoconversion experiments on several lipid-droplet-associated proteins do not support protein exchange between neighboring droplets (M.B., unpublished results). Still, protein translocation between droplets, as well as the heterotypic or homotypic fusion of lipid droplets (Boström et al., 2007), could affect lipid droplet diversification and needs to be studied in greater detail.
Although, the past few years have produced an ever increasing number of examples for intracellular lipid droplet diversification, we are still in the early days of understanding how lipid droplet diversification is achieved and what the functional implications of lipid droplet subsets are. Yet, the knowledge of how to enhance the prevalence of growing lipid droplets or the amount of lipid droplets that are prone to remobilization provides an entry point to identify suitable targets to enhance the lipid content of cells for biofuel production or nutritional purposes, as well as to tackle diseases associated with lipid storage.
We thank Carole Sztalryd and the members of our groups for constructive discussions during the writing of the manuscript.
A.R.T. is supported by grants ATIP-Avenir (from Institut National de la Santé et de la Recherche Médicale) and Programme Emergence de la Ville de Paris. M.B. is supported by grants from Bundesministerium für Bildung und Forschung (031 A 306) and Deutsche Forschungsgemeinschaft (BE 4597 / 2-1).
The authors declare no competing or financial interests.