Lipid droplets (LDs) are organelles that compartmentalize nonbilayer-forming lipids in the aqueous cytoplasm of cells. They are ubiquitous in most organisms, including in animals, protists, plants and microorganisms. In eukaryotes, LDs are believed to be derived by a budding and scission process from the surface of the endoplasmic reticulum, and this occurs concomitantly with the accumulation of neutral lipids, most often triacylglycerols and steryl esters. Overall, the mechanisms underlying LD biogenesis are difficult to generalize, in part because of the involvement of different sets of both evolutionarily conserved and organism-specific LD-packaging proteins. Here, we briefly compare and contrast these proteins and the allied processes responsible for LD biogenesis in cells of animals, yeasts and plants.
LD biogenesis in animals, yeasts and plants: common versus specific proteins and processes
A general model for the formation of a lipid droplet (LD) is difficult because of the variation in cell-type-, tissue-, organ- and organism-specific proteins, as well as differences in physiological and/or energetic demands on cells that produce LDs. Information in animal and yeast systems has accumulated in recent years on the identity of the specific proteins involved in LD biogenesis [1–5]. Several key proteins and corresponding genes have been identified as important for LD biogenesis, and several of these have come from mutant screens in model organisms or investigations into the molecular basis of human neutral lipid storage diseases, which are collectively referred to as lipodystrophies .
Although not an exhaustive list, LD biogenesis proteins in animals are considered to include perilipins (PLINs), triacylglycerol (TAG) biosynthetic enzymes [including glycerol-3-phosphate acyltransferase (GPAT), lysophosphatidic acid acyltransferase (LPAAT), phosphatidic acid phosphatase or lipin and diacylglycerol acyltransferase (DGAT)], fat-inducing transmembrane proteins (FIT1 and FIT2), SEIPIN and fat-specific storage protein 27 (FSP27) . Some evidence suggests that additional proteins involved in membrane dynamics, such as coatomer protein 1, SNAREs, Rabs and atlastin, may participate also in LD biogenesis; however, their precise role(s) remains unclear , particularly in plants where these proteins are only beginning to be considered in LD biogenesis . For the most part, LD ‘biogenesis’ proteins or their homologs collectively act at the endoplasmic reticulum (ER) membrane and/or LD surface to modulate the formation, growth and dynamics of LDs. Briefly and as depicted in Figure 1, TAGs are considered to be synthesized and accumulate within the ER membrane bilayer and aggregate into ‘lens’ domains that, with the assistance of FIT2, emerge from the ER surface toward the cytoplasm as a TAG-filled bulge [9,10]. Subsequent growth of the LD occurs while it either remains connected to the ER by a membrane neck or after release from the ER into the cytoplasm followed by LD fusion or reattachment to the ER. Regardless of which process, the nascent LD is coated with a monolayer of phospholipids and proteins, thought to be facilitated in part by SEIPIN and the binding of PLIN proteins to the LD surface.
Model for LD Biogenesis in eukaryotes.
Overall, the co-ordinated synthesis of TAG in specific subdomains of the ER is considered to make the compartmentalization of neutral lipids, a more efficient process. However, the mechanisms that regulate the formation of these LD subdomains in the ER remain elusive, although emerging evidence suggests that the same ER subdomains are involved in both LD and peroxisome formation  and that protein–protein and/or protein–lipid interactions likely play important roles in the LD assembly process [3,12–16]. For instance, new structural information for human and Drosophila SEIPIN proteins, obtained by cryo-electron microscopy (EM), suggests that SEIPIN forms an oligomeric, toroid-shaped complex in the ER membrane through extensive interactions of its lumenal domains [17,18]. These ER lumenal domains in animal SEIPIN have homology to lipid-transfer proteins and include hydrophobic helices that might allow SEIPIN to associate with the nascent TAG ‘lens’ in the ER membrane through interaction with membrane-packing defects. The N-terminus of Drosophila SEIPIN is orientated toward the cytoplasm and is predicted to interact directly with the growing LD, holding it in proximity to the ER as it fills with neutral lipid , but this remains to be determined experimentally. In yeast, the N-terminus of SEIPIN is required for directional budding of LDs into the cytoplasm , along with a cooperating protein, Ldb16p, that interacts with SEIPIN to support proper LD formation [20,21]. Recent evidence also suggests that, in some cell types, TAG and phosphatidylcholine (PC) synthesis enzymes remain associated with the LD surface for continued growth, and that FSP27 proteins mediate, among other things, fusion and growth of these LDs after release from the ER [1,5,7,15,22]. Hence, the emerging, generalized model for LD biogenesis requires the co-ordinated actions of several proteins that together synthesize, package and cover this hydrophobic compartment as it emerges from specific subdomains in the ER.
The information to date in plant systems about LD biogenesis is more fragmented, and emerging models [23,24] have some similarities and some differences from the details described above for animal and yeast systems (Figure 1). Most historical experimental evidence on LD formation in plants comes from studies of developing oilseeds, where oleosin proteins, which are not present in animal and yeast cells, are co-translationally inserted into the ER  and thought to migrate toward sites of TAG accumulation that form between the leaflets of the ER bilayer (Figure 1). Thereafter, the hairpin loop conformation of oleosin may support the formation of a ‘lens’ or bulge in the ER membrane, although the process responsible for the subsequent growth and release of the LD from the ER surface is largely unknown . For instance, no obvious homologs of the PLIN proteins are found in plants, so oleosins serve an analogous function in forming the LD coat in developing seeds. However, oleosins are not expressed in many plant cell types that produce and store neutral lipids, including vegetative organs, such as leaves and stems, and most oleaginous fruits [27,28]. Instead, it seems that oleosins, rather than serving a general role in LD formation in plant cells, have evolved specialized functions in seeds (and pollen where they are also highly expressed) to prevent dehydration/rehydration-induced fusion of LDs [23,29–31]. Recently, additional proteins have been identified that target to LDs in various plant cell types [e.g. the plant-specific LD-associated proteins (LDAPs) and LDAP-interacting protein (LDIP)] (Figure 1), such that the inventory of LD proteins in plants is increasing to include those with possible novel roles in LD formation, stability, dynamics and/or turnover [24,32–38].
While oleosins represent one conspicuous difference between animal/yeast systems and plants, there are other differences as well. Obvious homologs for FIT2 and FSP27 also are missing in plants , and while this suggests that there are as-of-yet undiscovered proteins that have adopted these functions, these differences also raise the question of how much conservation exists for LD biogenesis between plant and animal/yeast systems. Still, a common subcellular site for TAG synthesis (i.e. ER) and the capacity for packaging neutral lipids into the aqueous cytoplasmic environment are features shared by all eukaryotes, and this suggests that common, evolutionarily conserved mechanisms underlie the biophysics and biogenesis of this organelle. Indeed, expression of mouse FIT2 in plant leaves promoted LD biogenesis , an organ/tissue in plants where LDs are normally relatively scarce. Additional support for a conserved biogenesis protein complex comes from evidence that plant homologs of SEIPIN modulate LD formation and size in plant cells in a manner that appears to be similar to the mechanisms described for animals and yeast . However, unlike yeast, insects and humans that contain only a single SEIPIN gene, three SEIPIN genes were identified in Arabidopsis and other angiosperm species , indicating that this fundamental component of LD biogenesis has been elaborated in plants. Thus, there are aspects of neutral lipid storage in plants that share some consistencies with those found in animal and yeast systems, but we are far from a unifying theme for eukaryotic organisms that describes the detailed requisite proteins and mechanisms for lipid storage, which likely involves both conserved and cell-type- or organism-specific components.
Lipid synthesis enzymes
The Kennedy pathway leading to the synthesis and accumulation of TAGs is inextricably linked to the biogenesis of cytoplasmic LDs, and the steps for the sequential acyl-CoA-dependent acylation of glycerol have been known for years [41,42]. For instance, the acyltransferases (GPAT, LPAAT, DGAT) catalyzing these reactions reside in the ER in most eukaryotic cells, and presumably occupy the same subcellular location as the sterol acyltransferases, and the more recently described acyl-CoA-independent DAG acyltransferases (phospholipid:DGAT, PDAT in plants; Lro1 in yeast) [43,44]. Consequently, the biogenesis of LDs in eukaryotic cells (Figure 1) begins necessarily with the synthesis of these nonbilayer lipids — TAGs and/or steryl esters (StEs) . Most of the work with LD biogenesis has focused on TAG-containing LDs, but similar biophysical processes are likely to operate in the initial stages of StEs-containing LDs as well [10,45]. It is even possible that TAGs and StEs are co-mixed in LDs in some cell types. On the other hand, distinct LD types with different surface or internal compositions have been observed to coexist in cultured mammalian cells , fission yeast  and plants .
Neutral lipids account for ∼1–4% of biological membranes, but above these levels, they coalesce to form neutral lipid aggregates that appear as ‘lens’-like structures between the leaflets of ER membranes . In synthetic membrane systems, modulation of phospholipids on one leaflet or other that reduce surface tension leads to the directional ‘budding’ of the lens-like structure from the membrane bilayer, resulting in a nascent LD covered with a phospholipid monolayer that is contiguous with the originating ER membrane leaflet . While this process can occur spontaneously in synthetic systems when surface tension is reduced, the process is apparently carefully orchestrated in biological systems by a variety of proteins, including SEIPIN, FIT2 and PLINs. Notably, in the absence of either SEIPIN or FIT2 in mammals and yeast, LD budding can occur on either side of the ER membrane , suggesting that one role for these proteins is to determine the polarity of bud emergence from the ER into the cytoplasmic space.
Once a nascent LD has emerged from the ER, there is a coordination between neutral lipid and phospholipid synthesis for proper growth and expansion of the droplet, and not surprisingly, the ratios of phospholipid classes and/or molecular species have been found to influence the size of LDs released into the cytoplasm [9,51,52]. Enzymes of phospholipid synthesis are located in the ER to support this coordination, but some of the enzymes associated with PC biosynthesis also have been localized to the LD surface in some cell types, especially under conditions that promote LD proliferation . It is of interest to note that some models for yeast LD biogenesis suggest that LDs remain attached to the ER and are not released into the cytoplasm (Figure 1), and undergo expansion or retraction based on the metabolic needs of the cell [53,54].
FIT proteins (FIT1 and FIT2)
FIT proteins were discovered ∼10 years ago, and while they have clearly been shown to be involved in LD formation and storage lipid deposition in a variety of eukaryotic cell types, their roles in these processes are somewhat enigmatic. Mammals contain two FIT isoforms, FIT1 and FIT2, and both were shown to localize to the ER and promote LD formation when overexpressed in cultured cells [55–57]. FIT2, however, was more effective in this process, and the necessity of this isoform for neutral lipid accumulation in adipose tissue [58,59] suggests that this protein is a fundamental component of storage lipid accumulation in eukaryotes. However, biochemical analyses indicated that purified FIT2 proteins bind to TAG, but have no inherent TAG biosynthetic activity . Given that FIT proteins become enriched in ER–LD regions upon lipid-storing conditions, it was proposed that they help sequester TAG in the ER bilayer at the early stages of LD formation [50,59] (Figure 1).
Homologous genes encoding FIT2-like proteins have been identified in yeast and Caenorhabditis elegans and were also shown to be important for LD biogenesis in these organisms. Notably, in the absence of FIT2, LDs remained embedded in ER membranes, revealing a critical role for FIT proteins in LD emergence . Clues to the underlying mechanism have been suggested from studies showing that disruption of FIT2 proteins increases ER content of DAG, a lipid species that disfavors LD budding . Additional mutations that decreased DAG content in FIT2-disrupted cells complemented the LD phenotype, supporting a role for FIT2 in modulating DAG metabolism .
Other studies, however, have shown that FIT2 is important for phospholipid metabolism and maintenance of general ER morphology . FIT2 proteins share sequence similarity with lipid phosphate phosphatases (LPPs) , including conserved amino acids known to be important for catalytic activity [60,61]. Purified FIT2 proteins were also active toward phosphatidic acid (PA) and lysophosphatidic acid (LPA) substrates, and disruption of FIT2 increased cellular PA content, suggesting that PA is a possible substrate for FIT2 in vivo . The importance of LPP activity for FIT2 function was further supported by studies showing that PA content and LD biogenesis were restored in FIT2-disrupted cells by expression of native FIT2 proteins, but not by modified versions of FIT2 proteins lacking LPP activity. Disruption of FIT2 also reduced the synthesis of other phospholipids, such as PC and phosphatidylethanolamine, and resulted in general defects in ER morphology. Thus, the failure of LDs to emerge from the ER in FIT2-disrupted cells might be due to generalized defects in ER membranes resulting from aberrant phospholipid metabolism.
It was also proposed, however, that FIT2 might be important for maintaining the balance of phospholipids in the more immediate vicinity of LD formation by facilitating the transfer of phospholipids from the inner to the outer leaflet of the ER membranes, thereby increasing phospholipid availability for growth of the LD monolayer membrane . Topology mapping studies place the active site of FIT2 proteins on the lumenal side of ER membranes , and activity of FIT2 toward PA would produce DAG, which could readily move or ‘flip’ to the outer leaflet membrane. Conversion of DAG back to phospholipid would trap the glycerolipid species on the outer leaflet of the membrane, thereby making it available for LD growth. While this model remains to be experimentally tested, there are many studies showing that phospholipid content, composition and metabolism are essential features of LD biogenesis in eukaryotic cells [62–64].
While obvious homologs of FIT2 are not found in plants, the essential nature of FIT2 in animals and yeasts suggests that other proteins in plants might provide an analogous function. There are indeed other ER-bound LPPs (or lipins) in plants, and disruption of these proteins leads to severe changes in ER membrane organization [65–67]. Plants also have phospholipid flippases, albeit not extensively characterized, that are likely involved in the exchange of phospholipids between inner and outer leaflets of membranes, including at the ER . However, whether these proteins play a role in LD biogenesis has not been considered, except for the role of phosphatases in supplying DAG for TAG biosynthesis. Perhaps the new information about FIT2 and its reported phosphatase activity will give insights to the initiation of LD biogenesis in plant systems. Interestingly, expression of mouse FIT2 in plants is able to promote a dramatic proliferation of cytoplasmic LDs, even in leaves where LDs do not normally accumulate . This increase in LD formation in leaves was observed in both transient and stable expression systems and even resulted in elevated oil content in seeds of several lines of FIT2-overexpressing plants. Like in other cell types, mouse FIT2 also localized to the ER in plant cells and was especially concentrated at domains considered to be involved in LD formation . This suggests that FIT2 interact with some endogenous protein(s) (or biophysically conserved property of LD formation) in plants to initiate LD biogenesis, or at least that the activity of FIT2 in plants is sufficient to drive ectopic LD formation. Furthermore, both TAG and StE content were increased in leaves expressing mouse FIT2, suggesting a general initiation of LD formation rather than LDs with TAG only . Thus, it may be that, in the ER in plants, FIT2 promotes a localized environment that favors the general assembly of LDs, perhaps through its phosphatase activity, but this remains to be investigated. In any case, the ectopic, heterologous expression of mouse FIT2 is sufficient to induce the formation of LDs in both insect and plant cells (Figure 2), confirming an important role in promoting LD biogenesis in a variety of eukaryotic cell types.
Heterologous expression of mouse FIT2 is sufficient to promote the cross-kingdom formation of LDs in evolutionarily-diverse cell types.
The protein designated SEIPIN is important for normal LD formation at ER–LD contact sites (Figure 1). This protein is the product of the human Berardinelli-Seip Congenital Lipodystrophy 2 (BSCL2) gene and mutations result in the most severe form of congenital lipodystrophy . The yeast homolog of human SEIPIN, Fld1p, was originally identified in a mutant screen for aberrant LDs [70,71] and like its human counterpart, the yeast SEIPIN protein is required for normal LD biogenesis. Absence of Fld1p results in the production of small-sized LDs that tend to aggregate and become tangled in ER membranes and also fuse together to form supersized LDs . A different genetic screen identified a similar supersized LD phenotype when the yeast LDB16 gene was knocked out, and subsequent studies revealed that Fld1p and Ldb16p physically interact with each other to form a functional complex required for proper LD assembly . In the absence of either Fld1p or Ldb16, PA was also shown to aggregate in the vicinity of the small LDs entangled in the ER . Based on these and other data, it was proposed that PA, which is a cone-shaped lipid, has an affinity for the extreme curvature present at the ER–LD membrane junction, thereby becoming non-specifically trapped at this location in fld1/ldb16-disrupted cells. It also appeared that this pool of PA was inaccessible to PA-metabolizing enzymes and that no specific enzymatic pathway was responsible for the PA detected in these regions . These PA ‘foci' were suppressed only when Fld1p and Lbd16p were co-expressed together, suggesting that one role for the SEIPIN complex in eukaryotes is to prevent the accumulation of PA, serving as a vital signaling and structural metabolite, at the ER–LD junction.
Interestingly, the single human SEIPIN (BSCL2) gene can functionally complement the LD biogenesis and PA accumulation phenotypes present in fld1/ldb16-disrupted yeast cells , suggesting that the activities of Fld1p and Ldb16p are shared by the mammalian SEIPIN protein. There is no detectable sequence homology, however, between Ldb16p and any region of the mammalian SEIPIN protein sequence, suggesting functional, convergent evolution. Indeed, this appears to be an underlying theme for certain proteins and LD-related activities in yeast, mammals and plants, where proteins of similar functions appear to lack any obvious sequence conservation across phyla.
Unlike mammals and yeast, where single SEIPIN genes have been identified, angiosperm plant species contain multiple SEIPIN homologs, all of which appear to play a role in LD biogenesis [40,72]. In fact, differences in LD morphology were observed when any of the three different SEIPIN isoforms were overexpressed in plants, and the N-terminus of these proteins was shown to play a critical role in the determination of LD size and number . Furthermore, singular expression of plant SEIPIN genes in yeast fld1 mutant yeast cells only partially complemented the LD biogenesis phenotype , suggesting that other factors or proteins from plants, or perhaps a proper stoichiometry of all three SEIPIN proteins, might be required for the production of normal sizes and numbers of LDs. The loss-of-function of two or all three of the SEIPIN isoforms in Arabidopsis severely compromised normal LD biogenesis in plants, both in terms of normal LD morphology and subcellular location of LD deposition .
In general, the SEIPIN protein has been implicated in multiple aspects of LD biogenesis and lipid metabolism in eukaryotes, including proper initiation of LD formation at the ER, modulation of glycerophospholipid metabolism (particularly for PA) and maintenance of the ER–LD junction [2,16,20,63]. How SEIPIN performs these various functions remains an open question, but recent structural studies, as well as increased knowledge of SEIPIN interacting proteins, are beginning to provide a more comprehensive picture of how SEIPIN works. For instance, topology mapping and membrane association studies revealed that the SEIPIN protein has two TMDs, with the N- and C-termini orientated toward the cytoplasm, and a large, highly conserved domain present on the lumenal side of the ER. Motif-based sequence analysis of the three SEIPINs from Arabidopsis (AtSEIPIN1–3) and their homologs from human and Drosophila, identified three motifs that are common among the different SEIPINs (Figure 3). These three motifs span the conserved ER lumenal domain of SEIPINs. In contrast, the five homologs differ mainly in their N-terminal regions, particularly AtSEIPIN2 and AtSEIPIN3, which have significantly longer N-terminal regions relative to the other three SEIPINs (Figure 3). Analysis of the native conformation of yeast SEIPIN revealed that it assembled into high-ordered complexes where the subunit compositions were predicted from Stokes radii and sedimentation coefficients to be homo-oligomers comprised of nine subunits . More recent cryo-EM analysis of the human and Drosophila SEIPIN complexes revealed ring-like homo-oligomers of 11  (Figure 4A) or 12 SEIPIN subunits  (Figure 4B), respectively, consistent with the toroid shape of the negatively stained SEIPIN complexes isolated from yeast . The only well-resolved region in the cryo-EM density map was that of the ER lumenal domain, which showed extensive interactions between the SEIPIN monomers to help facilitate the organization of the SEIPIN oligomer. Mutations to key residues involved in protein–protein contacts disrupted oligomer formation, and the inability of these modified proteins to restore LD biogenesis in SEIPIN-disrupted cells confirmed that oligomerization is essential for SEIPIN function in vivo.
Common motifs of SEIPIN proteins from different species.
Structural models of plant SEIPINs compared with human and Drosophila SEIPIN.
The lumenal domain of each monomer consists of eight β-strands (β1–β8) and three or four α-helices (α1–α3 in human; α1–α4 in Drosophila) (Figure 4F,G,K,L). The eight β-strands form a β-sandwich fold that resembles lipid-binding domains in other proteins, notably the Niemann–Pick type C2 (NPC2) protein that binds sterols, and the C2 domain of protein kinase C (PKC-C2) that binds phosphatidylserine. The lipid-binding domain of human SEIPIN was expressed and purified and used in protein–lipid overlays, which revealed selective binding to PA with 16:0–18:1, but not 16:0–16:0 fatty acyl groups, suggesting that the acyl composition might be an important factor in PA recognition . Usage of full-length SEIPIN protein in the protein–lipid overlays showed binding to both PA and PI-3P , but these latter results were interpreted by the authors with caution given there is little evidence that connects PI-3P to LD biogenesis, and PI-3P is typically not associated with ER membranes. Nonetheless, the binding of the SEIPIN lumenal domain to PA is tantalizing in that it suggests a potential mechanism by which SEIPIN might suppress the accumulation of PA at the ER–LD junction. SEIPIN is also known to interact with enzymes that modulate PA metabolisms, such as AGPAT and lipin , and thus the SEIPIN lipid-binding domains might help preserve the metabolic accessibility of PA, especially in the vicinity of the ER–LD junction. Determining a cryo-EM 3D structure of SEIPIN with bound phospholipids will be helpful in determining and confirming the specific lipid species that binds to SEIPIN in vivo, and what conformational changes might take place upon lipid binding.
The hydrophobic helices present in the lumenal domains of SEIPIN are positioned towards the surface of the ER inner leaflet membrane, and comparison of SEIPIN sequences across distantly related species revealed strong conservation in the position of large, hydrophobic amino acids . Notably, large hydrophobic residues are often found on amphipathic helices that bind to the packing defects present on the LD monolayer, thereby serving as targeting signals for various LDAPs. Expression of fusion proteins consisting of mCherry and the SEIPIN hydrophobic helices resulted in targeting of the protein to LDs, confirming that these regions of SEIPIN have the capacity to recognize the LD surface. Taken together with the relative positioning of these helices on the lumenal face of the ER, Sui at al.  hypothesized that these helices help the SEIPIN complex recognize the packing defects on the lipid ‘lens' as they begin to bulge in the ER membrane. Similar helices present in the N-terminal cytoplasmic region might recognize the lens on the outer surface of the ER.
Homology modeling of the ER lumenal domains of the three plant (Arabidopsis) SEIPINs matches well to the main structural features of Drosophila and human SEIPIN, including the β-sandwich lipid-binding motif and membrane-interacting helices (Figure 4). Quaternary structure prediction for the three Arabidopsis SEIPINs showed that AtSEIPIN1 and AtSEIPIN2 are capable of assembling from 12 subunits, while AtSEIPIN3 is predicted to form an 11-mer (Figure 4C-E). In the human and Drosophila SEIPIN structures, it was shown that in addition to the membrane-interacting helices, other regions of the protein also contributed to oligomerization, including parts of the loop regions between β1 and β2 (named L2) and between β7 and β8 (named L9) [17,18] (Figure 4F,G). In comparison with human and Drosophila SEIPINs, the Arabidopsis SEIPIN1 homology model has a relatively shorter L2 (Figure 4H), while both AtSEIPIN2 and AtSEIPIN3 have longer L9 regions (Figure 4I,J). Analysis of the physicochemical properties of the membrane-anchored helices in human SEIPIN (α2 and α3), Drosophila SEIPIN (α3 and α4) and the predicted models of Arabidopsis SEIPINs demonstrates conserved amphipathic features of these helices, where the distribution of hydrophobic and polar residues between the opposite faces of these α-helices is well suited for membrane binding (Figure 4P–T). The sequence analysis of the identified common motifs (Figure 3B) indicates highly conserved amino acids in the different regions of the ER lumenal domain. For instance, positions 53–70 within motif 1 represent the amino acid sequence of the membrane-anchored helices. This region includes several absolutely conserved amino acids across all SEIPINs, especially serine and threonine at positions 53 and 61, respectively (Figures 3B and 4P–T). Other conserved residues include those at positions 15–17 of motif 1 (i.e. proline, glutamic acid and serine, respectively), methionine (or valine) at position 25 of motif 1, the aromatic amino acid (tyrosine/phenylalaine) at position 51 of motif 1, glutamic acid (or glutamine) at position 75 of motif 1, tyrosine at position 3 of motif 2 and the aromatic amino acid (tyrosine/phenylalanine) at position 24 of motif 3 (Figure 3B). All of these conserved amino acid residues are reportedly involved in the oligomerization of SEIPIN subunits in human and Drosophila SEIPINs [17,18]. It remains to be determined how the different plant SEIPINs organize, and whether a hetero-oligomeric complex of the three SEIPIN proteins is formed in vivo, with a specific stoichiometric ratio, to produce LDs in plants. However, these conserved residues at specific places in the predicted structures suggest important homologous structural features that likely participate in the organization of all SEIPINs.
There is evidence that LD formation from the ER bilayer may be mediated by changes in membrane surface tension and hydrophobic interactions within the bilayer . Furthermore, the surface phospholipid composition has been postulated to modulate LD size [9,51]. How SEIPIN and other associated proteins co-ordinate these biophysical factors to initiate and control the rate of directional cytoplasmic LD biogenesis remains to be determined. For instance, does the SEIPIN complex recognize subtle curvature sites where PA resides at nascent TAG aggregation sites to mark the position of TAG synthesis and LD emergence? Are there additional interacting proteins that participate with SEIPIN in this process, such as the peroxisome biogenetic factor peroxin 30 ? Also, is FIT2 a gatekeeper that interacts with SEIPIN proteins to initiate LD formation? It is likely that the answers to these questions in the near future will reveal more mechanistic details about how this hydrophobic LD compartment is formed and stabilized in the cytoplasm.
LDAPs and LDIP: plant-specific LD biogenesis factors?
Apart from oleosins, several other LD proteins have been identified in proteomics studies of purified plant LDs, and the protein compositions of this compartment appear to differ depending upon the source. One particularly interesting protein family initially identified in the proteome of avocado LDs, are the LDAPs, which represent a plant-specific family of LD surface proteins (Figure 1), but have no obvious homologs in yeasts or animals . LDAPs share homology, however, with the small rubber particle protein family that localize to rubber particles, which are similar in structural respects to LDs, except they contain a nonpolar, polyisoprenoid core instead of a core of TAGs or StEs [26,77].
LDAPs specifically localize to LDs, and because they exhibit minimal hydrophobic stretches (Figure 5A), they are unlikely to be embedded into the LD hydrophobic matrix, but instead are most likely associated with the LD surface (Figure 5B). Furthermore, disruption of any portion of the LDAP sequence diminishes their association with LDs [78,79]. Because LDAPs are associated with various types of cytoplasmic LDs, including rubber particles, these proteins may represent plant-specific proteins evolved to stabilize neutral lipids in the cytoplasm . Although beyond the scope of this review, LDAPs are not associated with the LDs present inside plastids known as plastoglobules, which compartmentalize a variety of pigments, pigment esters, TAGs and other neutral lipids . Arabidopsis has three LDAP genes that exhibit complex expression patterns during growth and development and in response to plant stress , and the roles of these proteins (also termed SRPs for stress-related proteins) in LD accumulation have been examined [82,83]. Importantly, the LDAP proteins were discovered to interact directly with another plant-specific LD protein that has been designated LDIP (for LDAP-interacting protein). This interaction was initially revealed through a yeast 2-hybrid screen and confirmed through additional studies [24,84]. LDIP contains several stretches of hydrophobic amino acid sequence (Figure 5A), and so it may anchor itself to the LD and facilitate the binding of LDAP to the LD surface (Figure 5B). Furthermore, modulating the expression of either LDIP or LDAPs alters the numbers and/or size of LDs in plant cells [24,79,82]. However, considerable gaps remain in the understanding of how these plant-specific proteins, or even the better-known oleosin proteins, interact and function with the more broadly conserved SEIPIN proteins in the organization and formation of LDs in plant cells.
Structural model for plant LDAP and LDIP proteins.
fat-specific storage protein 27
lysophosphatidic acid acyltransferase
Support for research in K.D.C.’s, J.M.D.s’ and R.T.M.’s laboratories on LDs was provided by the U.S. Department of Energy, BER Division (DE-SC0000797) and more recently, BES-Physical Biosciences program (DE-SC0016536). Funding support was also provided by the Natural Sciences and Engineering Research Council of Canada (to R.T.M).
We are grateful to Dr David Silver (Duke-NUS Medical School) for permission to show previously published micrographs in Figure 2a, Ian Smith (University of Guelph) for assistance with generating Figures 1 and 5 and Dr Charlene Case for assistance with preparation of the manuscript. We apologize to those researchers whose work was not cited because of space limitations. Mention of trade names or commercial products in this article is solely for the purpose of providing specific information and does not imply recommendation or endorsement by the U. S. Department of Agriculture. USDA is an equal opportunity provider and employer.
The Authors declare that there are no competing interests associated with the manuscript.