Lipid droplet motility and organelle contacts

Abstract

Lipid droplets (LDs) are fat storage organelles integral to energy homeostasis and a wide range of cellular processes. LDs physically and functionally interact with many partner organelles, including the ER, mitochondria, lysosomes, and peroxisomes. Recent findings suggest that the dynamics of LD inter-organelle contacts is in part controlled by LD intracellular motility. LDs can be transported directly by motor proteins along either actin filaments or microtubules, via Kinesin-1, Cytoplasmic Dynein, and type V Myosins. LDs can also be propelled indirectly, by hitchhiking on other organelles, cytoplasmic flows, and potentially actin polymerization. Although the anchors that attach motors to LDs remain elusive, other regulators of LD motility have been identified, ranging from modification of the tracks to motor co-factors to members of the perilipin family of LD proteins. Manipulating these regulatory pathways provides a tool to probe whether altered motility affects organelle contacts and has revealed that LD motility can promote interactions with numerous partners, with profound consequences for metabolism. LD motility can cause dramatic redistribution of LDs between a clustered and a dispersed state, resulting in altered organelle contacts and LD turnover. We propose that LD motility can thus promote switches in the metabolic state of a cell. Finally, LD motility is also important for LD allocation during cell division. In a number of animal embryos, uneven allocation results in a large difference in LD content in distinct daughter cells, suggesting cell-type specific LD needs.

Advances in microscopy, genome editing, and computational biology have dramatically increased our understanding of how cells organize themselves. A major paradigm shift has been the recognition that cellular organelles frequently interact with each other via specific contact sites. These contacts are dynamic and highly regulated, and they are key for understanding how organelle functions are integrated into the cell as a whole. As the reviews in this special issue demonstrate, this new perspective is transforming our understanding of lipid droplets (LDs), the organelles specialized for intracellular fat storage.

LDs play crucial roles in energy homeostasis, diverse aspects of lipid metabolism, and the handling of specific proteins (Walther and Farese, 2012; Hashemi and Goodman, 2015; Welte, 2015b; Kimmel and Sztalryd, 2016; Welte and Gould, 2017). They sequester harmful lipids and proteins and provide fatty acids for energy production, membrane lipids, and signaling molecules. LDs have thus critical functions in diverse cells, including in adipose tissue, liver and skin, many reproductive tissues, as well as the immune and nervous systems (Saka and Valdivia, 2012; Walther and Farese, 2012; Krahmer et al., 2013; Arrese et al., 2014; Gluchowski et al., 2017; Pennetta and Welte, 2018). After originating from the ER, LDs physically and functionally interact with many cellular organelles, including other LDs, the ER, mitochondria, peroxisomes, and lysosomes; eventually, they are broken down by cytoplasmic lipases and via autophagy.

In many cells, LDs show extensive intracellular motility (Welte, 2009; Valm et al., 2017). Such motility has the potential to bring LDs into the vicinity of other organelles and thus is expected to contribute to the formation, dynamics, and dissolution of inter-organellar contacts. Such roles for motility are certainly well established for other intracellular organelles: for example, in mitochondria, a complex of Milton and Miro1/2 mediates the recruitment of both Dynein and Kinesin for transport along microtubules as well as of Myosin 19 for transport along actin filaments (Oeding et al., 2018). This motility plays a crucial role in bringing different mitochondria together for fusion (Cagalinec et al., 2013), an adaptive stress response that improves mitochondrial metabolism and survival. Lysosomes are highly motile as well (reviewed in Pu et al., 2016), displaying Kinesin-1, Dynein-Dynactin, and Myosin IIa driven transport; such active transport is, for example, crucial for lysosomal fusion with autophagosomes during autophagy (Nakamura and Yoshimori, 2017). Below, we review what is known about LD motility and its contribution to inter-organelle contacts, with a particular emphasis on recent papers. For a summary of the older literature on LD motility, we refer the reader to three comprehensive reviews (Welte, 2009, 2015a; Welte and Gould, 2017).

LD motility is widespread

Like other intracellular structures, LDs can display a variety of states of motion: free diffusion, confined Brownian motion (e.g., around a tether point), and directed motion along linear tracks (Maucort et al., 2014; Norregaard et al., 2017). A given LD can repeatedly switch between such states, though the exact pattern varies dramatically between different types of cells. For example, in early Drosophila embryos, essentially all LDs constantly move back-and-forth along linear paths, with only occasional brief pauses and no obvious diffusive behavior (Gross et al., 2000, 2002). In contrast, in the hepatocyte cell line HuH-7, LDs typically oscillate in a confined area, but some switch to rapid unidirectional movement in a transient fashion (Targett-Adams et al., 2003). Because linear motion of LDs is by far the best understood, both molecularly and functionally, the following discussion will focus on this type of motility.

Directed LD motion has been observed in a wide variety of systems: from yeasts and fungal hyphae, to insect and vertebrate embryos, to a plethora of cultured mammalian cells (for reviews, see Welte, 2009, 2015a; Welte and Gould, 2017). Recently, advances in imaging technologies have uncovered many additional instances of LD motility; these discoveries expand both the experimental approaches to monitor LD motion and the systems in which its mechanisms and functions can be studied. Particularly exciting are label-free methods which have been used to follow LDs in real time: methods based on Raman scattering for plant seedlings, mouse oocytes, and insect embryos (Dou et al., 2012; Bradley et al., 2016; Waschatko et al., 2016); confocal reflection microscopy for Drosophila ovaries (Gaspar and Szabad, 2009); and near-infrared spectroscopy for fish embryos (Ishigaki et al., 2016). In addition, new fluorescent dyes have been developed for in-vivo imaging of LDs (e.g., Collot et al., 2018; Zheng et al., 2019). These methods are well suited for live imaging, making them complementary with older, fixed-imaging techniques. Concordantly, advances in computational analysis allow for monitoring many organelles at the same time through demixing of multi-spectral imaging. They reveal, e.g., with which types of organelles LDs make contact (Valm et al., 2017). Finally, intra-vital imaging has, for the first time, been able to monitor the motion of individual LDs in mammary gland cells during lactation (Masedunskas et al., 2017).

Molecular mechanisms of LD motility

To evaluate to what extent LD motility contributes to inter-organellar contacts, it is essential to experimentally modulate this motility. Thus, a good understanding of its molecular basis is crucial.

Directed motion of LDs is typically brought about by cytoskeletal motors, such as Kinesins, Dyneins, and Myosins (Welte, 2009; Welte and Gould, 2017). In the simplest case, LDs are directly pulled by motors, either unidirectionally along actin filaments or microtubules or – if opposing motors work together – back-and-forth along microtubules, displaying so-called bidirectional transport (Fig. 1 A, B) (Welte, 2004). Motors can also bring about LD motion indirectly, such as in the filamentous fungus Ustilago where LDs “hitchhike” on motor-driven early endosomes (Fig. 1C) (Guimaraes et al., 2015). In a second indirect mechanism, cytoskeletal motors can power bulk motion of the entire cytoplasm that then drags LDs along – together with many other organelles (Fig. 1D). Two such instances occur during Drosophila oogenesis: nurse cells, the fifteen sister cells of the oocyte, undergo actin and Myosin-II dependent contraction, resulting in LDs, as well as other cytoplasmic content, being squeezed through cytoplasmic bridges into the oocyte (Wheatley et al., 1995). At the same time, the oocyte undergoes Kinesin-1/microtubule-driven cytoplasmic streaming that mixes the incoming nurse cell contents with the existing ooplasm (Palacios and St Johnston, 2002).

Figure 1:

Mechanisms of LD motility.

(A, B) LDs are transported as direct cargo by molecular motors along cytoskeletal filaments: transport along actin filaments is driven by plus-end directed Myosins (A); transport along microtubules can be unidirectional towards either the plus-or minus end (powered by Kinesin-1 or Cytoplasmic Dynein, respectively) or bidirectional, using both types of motors (B). (C) LDs hitchhike on other organelles that are transported by motors along microtubules. (D) Molecular motors promote bulk flow of the cytoplasm that drags many organelles, including LDs, along. (E) Polymerization of actin filaments may push LDs through the cytoplasm. (F) Myosin II may move LDs indirectly, by sliding actin filaments with attached LDs against each other. (G). LD moving by Brownian motion until it is captured by a tether on a target organelle.

Some cases of LD motility may not be due to molecular motors, but nevertheless depend on the cytoskeleton. For example, in U2OS cells, localized polymerization of actin is necessary to drive clustered LDs away from each other (Pfisterer et al., 2017). Conceivably, LDs are pushed directly by growing actin filaments (Fig. 1E); alternatively – or in addition – actin filaments polymerized from LDs might be pushed against each other by Myosin II, a motor that slides filaments (Fig. 1F) (Pfisterer et al., 2017). Actin-based mechanisms also promote the dissociation of LDs from peroxisomes in seedlings of Arabidopsis (Cui et al., 2016).

Unequivocal proof that actin-and microtubule-based mechanisms are responsible for LD motion exists only for a moderate number of systems. However, the presence of either actin or tubulin in numerous LD proteomes is consistent with the idea that these cytoskeletal elements drive LD motility in many cells. In addition to the cases summarized previously (Welte, 2009, 2015a; Welte and Gould, 2017), there are recent examples ranging from plants and fungi to mammalian systems (Brocard et al., 2017; Yu et al., 2017; Zhi et al., 2017; Speziali et al., 2018, Pfisterer, 2017 #394). These data have to be interpreted cautiously as actin and tubulin in LD proteomes might simply represent biochemical contamination, given that they are abundant cytoplasmic proteins. However, actin, tubulin, and a motor subunit (KIF16B) were also detected in LD proteomes using much more stringent proximity-labelling strategies (Bersuker et al., 2018). As such stringent techniques become more wide-spread, the true extent of LD-cytoskeletal associations will become clearer. Of course, the mere presence of cytoskeletal elements does not prove by itself that they are responsible for LD motion; after all, LDs can act as storage sites for diverse proteins (Welte and Gould, 2017). Thus, functional tests are required, such as those detailed in the next section.

Finally, simple Brownian motion might lead to an inhomogeneous LD distribution in the cell if diffusing LDs get captured at specific sites, e.g., when they form contacts with partner organelles (Fig. 1G). For example, newly formed LDs in Drosophila cultured cells will change their spatial distribution over the course of several hours and eventually congregate into clusters (Guo et al., 2008). These clusters seem to result from slow aggregation of randomly moving LDs, not concerted transport to a specific region. This pattern is consistent with the idea that diffusing LDs happen to collide and then remain attached, potentially through a tethering mechanism (Fig. 1G).

Strategies for modulating LD motility: targeting tracks and motors

Given that most LD motility depends on actin filaments or microtubules, there is a simple general strategy to probe whether LD motility contributes to organelle contacts: alter the actin or microtubule tracks, pharmacologically or genetically, and determine effects on LD contacts. For example, when COS-7 cells are treated with Nocodazole to disassemble microtubules, the contact landscape of LDs changes dramatically: fewer LDs are engaged in contacts with mitochondria or peroxisomes, ternary contacts between LDs, peroxisomes, and the Golgi decrease, and simultaneous contacts between LDs, ER, and lysosomes increase (Valm et al., 2017). These observations implicate microtubule-driven motility of LDs and/or those other organelles in controlling inter-organellar contacts. In a hepatocyte cell line, microtubules are stabilized in response to Septin-9 upregulation, which then promotes LD relocalization to the perinuclear region (Akil et al., 2016), with the potential of novel inter-organelle contacts. The actin-disrupting drug Latrunculin prevents the migration of LDs towards the animal pole in zebrafish embryos (Dutta and Kumar Sinha, 2015) as well as proper allocation of LDs into the spores of fission yeast (Yang et al., 2017). In the fungus Metarhizium acridum deletion of the beta tubulin gene impairs LD trafficking (Zhang et al., 2017). Finally, post-translational modification of microtubule tracks can have a profound impact on LD distribution and organelle contacts: in Vero cells, nutrient deprivation results in the assembly of a network of microtubules enriched in detyrosinated tubulin; transport along these microtubules promotes LD dispersion which in turn increases LD-mitochondrial contacts (Herms et al., 2015). In contrast, in hepatic cells, microtubule acetylation impairs LD motility (Groebner et al., 2019); it was proposed that Cytoplasmic Dynein binds more strongly to these modified tracks, thus stalling LD motion. Intriguingly, cells with increased microtubule acetylation displayed reduced LD-LD interactions, suggesting that LD motion promotes contacts between LDs.

A downside of abolishing LD motility via altering cytoskeletal tracks is that many other cytoskeleton-dependent processes may also be disrupted, including the trafficking, biogenesis, and turnover of various organelles. Such pleiotropic effects make it challenging to confidently attribute changes in LD inter-organellar contacts to altered LD motion. A more targeted approach is to specifically abolish the motors powering LD transport. In those cases where the microtubule motors responsible for LD motion have been identified, they are almost invariably the plus-end motor Kinesin-1 and the minus-end motor Cytoplasmic Dynein (Welte, 2015a; Welte and Gould, 2017). They can be found on their own or in combination. Thus, inhibiting these two motors is typically the first choice for manipulating LD motility. Kinesin-1 even powers the indirect transport of LDs by cytoplasmic flow in Drosophila oocytes (Serbus 2005; Palacios 2002). The only exception we are aware of is in Ustilago, where LDs are dragged along by early endosomes propelled via Kinesin-3/Cytoplasmic Dynein (Guimaraes et al., 2015). Intriguingly, the Kinesin-3 family member KIF16B was recovered as a high-confidence LD protein in two mammalian cell lines (Bersuker et al., 2018), though whether it promotes motility in those cases has not yet been tested. Actin-based transport of LDs involves motors from three different myosin families: in budding yeast, LDs are transported from the mother to the daughter cell via a type V Myosin (Knoblach and Rachubinski, 2015), a motor that carries cargoes along its tracks analogous to Kinesin-1 (Fig. 1A). Declustering of LDs in U2OS cells (Pfisterer et al., 2017) and transport of LDs to the cortex of zebrafish embryos (Dutta and Kumar Sinha, 2015) depends on a different Myosin, non-muscle Myosin II. This myosin does not directly transport cargo, but rather slides actin filaments against each other. Thus, in these instances, LDs might be pushed by moving actin filaments (Fig. 1F). Finally, the single-headed Myosin I is also implicated in LD motility; its pharmacological inhibition results in local accumulation of LDs at the cleavage furrow in zebrafish embryos (Gupta et al., 2017). However, the underlying mechanism is not yet clear.

The identification of LD motors has led to an unexpected side-benefit for motor studies, namely new opportunities to study the activity of these motors in vivo (reviewed in Welte, 2015a; Welte and Gould, 2017). Although in principle any organelle could be used for such investigations, LDs are particularly suited because of their unique biophysical properties. On the one hand, the size and shape of LDs make them ideal for motion tracking, as the position of the LD center can be pinpointed within a few nanometers even by standard light microscopy (Gross et al., 2000). On the other hand, LDs can be manipulated in vivo with optical tweezers, allowing the forces driving LD motion to be determined at the single LD level, including their variation over time (Shubeita et al., 2008). Most recently, this combination of properties was used to uncover that Cytoplasmic Dynein has an intrinsic ability to upregulate force production if it encounters an opposing force (Reddy et al., 2016; Chapman et al., 2019).

Strategies for modulating LD motility: focus on LD-specific interventions

Kinesin-1, Cytoplasmic Dynein, and Myosins are not restricted to LDs, but typically perform multiple independent functions in cells, including the trafficking of other organelles. Thus, to manipulate just LD motility, it is necessary to target LD-specific co-factors of the motor machinery.

A powerful approach would be to specifically abolish motor recruitment to LDs. Although it has been known for almost two decades that specific motors move LDs (Gross et al., 2000), their mode of attachment to LDs remains unclear. This is in contrast to many other motor cargoes, where motor adapters are well characterized (Akhmanova and Hammer, 2010). Physical interactions between microtubule motors and various LD associated proteins have indeed been uncovered, such as with the perilipin Plin3 (Gu et al., 2019), the nesprin Klar (Gaspar et al., 2014), and the GTPase Arf1 (Rai et al., 2017). However, with the data currently available, it is hard to determine whether those interactions recruit the motors to the LDs or if they control the activity of motors recruited by other means. For example, in hepatic cells, a Dynein subunit co-immunoprecipitates with PLIN3 and partially colocalizes with LDs by immunostaining, but it is not yet known if that colocalization is PLIN3-dependent (Gu et al., 2019). In Drosophila embryos lacking Klar, LDs still move bidirectionally, with much reduced forces and velocities (Welte et al., 1998); thus Klar is at best one of several anchors that recruit motors to LDs (for alternative possibilities, see Welte, 2015a). Arf1 levels on LDs, Kinesin-1 levels on LDs, and LD motility all change together in response to metabolic signals (Rai et al., 2017). However, it remains possible that a common upstream signal promotes independent recruitment of Kinesin-1 and Arf1 to LDs, and Arf1 then modulates the Kinesin-1 activity on LDs. In addition, since Arf1 can affect the movement of multiple organelles (Kaczmarek et al., 2017), it cannot be a specific LD motor anchor. A yet unexplored possibility is that motors might be anchored to lipids in the phospholipid monolayer of the LDs, rather than to LD proteins. A precedent exists in the Golgi and endosome localization of Kinesin-1, which is recruited via a protein complex that recognizes specific membrane lipids (Mahajan et al., 2019).

Even though motor anchors on LDs remain elusive, the motor co-factors already identified allow modulation of LD motility. Inactivation of Arf1 via Brefeldin-A in hepatic cells disrupts the intracellular distribution of LDs similar to Kinesin-1 knockdown (Rai et al., 2017). In the absence of the LD-specific isoform of Klar, most LDs in Drosophila embryos are mislocalized to the yolk cell (Guo et al., 2005), which presumably results in globally altered LD-organelle contacts. Similarly, lack of the Kinesin-1 cofactor Halo causes global redistribution of LDs in these embryos (Arora et al., 2016), but apparently not of other organelles (Gross et al., 2003). And several perilipins are known to localize to LDs and alter their motility and intracellular distribution, such as ectopically expressed PLIN1 in fibroblasts (Marcinkiewicz et al., 2006; Orlicky et al., 2013), PLIN2 in Huh-7 and NIH3T3 cells (Boulant et al., 2008; Cruz et al., 2019), and LSD-2/PLIN2 in Drosophila embryos (Welte et al., 2005). For PLIN1 and LSD-2, the phosphorylation state of the protein correlates with distinct intracellular distributions (Welte et al., 2005; Marcinkiewicz et al., 2006); this is also the case for PLIN6 in fish, a protein that localizes to carotenoid droplets, proposed to be homologous to LDs (Granneman et al., 2017). For ectopically expressed PLIN1, it was shown that the relevant kinase is Protein Kinase A and that phosphorylation of specific sites is responsible for modulating LDs’ intracellular distribution (Marcinkiewicz et al., 2006). More broadly, various experimental treatments have been reported to alter LD motility in cultured cells, such as exposure to certain herbal extracts (Haselgrubler et al., 2019) or modulation of the extracellular pH (Nardi et al., 2019a).

Reciprocal relationship between LD motility and organelle contacts

Just as LD motion can promote or break interactions with other organelles, so too can organelle contacts positively and negatively modulate LD motility. As discussed above, LDs in Ustilago tightly interact with early endosomes, which allows them to utilize the endosomal motor machinery for transport (Guimaraes et al., 2015). In turn, organelle tethers are implicated in impeding LD motion. A prime example are members of the VPS13 protein family that tether mitochondria and the ER to each other (Kumar et al., 2018; Yeshaw et al., 2019).They are also present on LDs (Bersuker et al., 2018; Kumar et al., 2018; Yeshaw et al., 2019) and may help establish contacts between the ER and LDs. Live imaging of cells expressing a VPS13A GFP fusion revealed that transient active motion of LDs correlates with loss of VPS13A from LDs; when VPS13A was regained, LD motion ceased. This correlation suggests that tethering of LDs to the ER restricts their motion (Yeshaw et al., 2019). Consistent with this model, overexpression of VPS13A reduces LD motility (Yeshaw et al., 2019). A related phenomenon may occur during cell division in budding yeast. Here, some LDs remain in contact with the nuclear envelope in the mother cell, while the rest are transported along actin filaments to the daughter (Knoblach and Rachubinski, 2015). Nuclear-envelope anchoring may prevent some LDs from productively engaging with Myosin motors, thus retaining them in the mother. In an analogous manner, mitochondria bound to LDs tend to display much reduced motility (Benador et al., 2019); it will be interesting to determine whether mitochondria-LD contacts in turn impair LD motion.

Reciprocal relationship between LDs and the cytoskeleton

Not only do cytoskeletal elements control LD motility, but LDs can modulate actin and microtubule organization. In prostate cancer cells, LDs regulate a non-centrosomal microtubule organizing center which is important for proliferation (Nardi et al., 2018; Nardi et al., 2019b). LDs in U2OS cells carry the actin nucleator Formin-like 1 (FMNL1) and can nucleate actin filaments in vitro. This activity is likely important in vivo, as FMLN1 promotes the recruitment of non-muscle Myosin to LDs, which in turn helps LDs to dissociate from each other (Pfisterer et al., 2017). This declustering role may be widespread as LD proteomes from numerous systems include regulators of actin polymerization. Furthermore, the LD-associated protein ORP2 physically interacts with actin regulators and affects the formation of actin filaments in a nutrient dependent manner (Olkkonen et al., 2019), though the detailed mechanisms remain to be worked out. In a potentially analogous case, the LD-associated protein Utrophin forms an atypical cytoskeletal filament when complexed with beta-2Syntrophin. These filamentous structures appear to be independent from other cytoskeletal filaments and regulate LD size in adipocytes (Krautbauer et al., 2019).

Functional consequences of LD motility

Contacts between LDs and other organelles play crucial roles for inter-organelle metabolite exchange. For example, fatty acids from triglyceride breakdown can be shuttled to mitochondria for beta-oxidation (Rambold 2015). Reciprocally, mitochondria bound to LDs can provide the substrates necessary for the synthesis of neutral lipids as demonstrated in mammalian brown adipose tissue (Benador et al., 2018). As LD motility alters the spatial position of LDs relative to the rest of the cell and, thus, presumably to other organelles, it is expected to have a profound effect on establishing and breaking LD-organelle contacts and therefore, on the flux of lipids through the cell.

Indeed, such functional effects of LD motion have already been demonstrated in several cases. LD motility along microtubules can control the extent of contacts between LDs and mitochondria, and thus, how well fatty acids from LDs are used as fuel (Herms et al., 2015). In the liver, activation of Kinesin-1 on LDs promotes lipoprotein secretion; it has been hypothesized that active motion of LDs increases their contacts with the ER, leading to the transfer of lipids needed for the assembly of lipoproteins in the ER lumen (Rai et al., 2017).

In additional cases, a role of LD motility for lipid transfer between organelles has been inferred, though not yet conclusively demonstrated. A number of intracellular pathogens, both bacteria and protozoa, deploy a specialized strategy for taking advantage of LD-stored nutrients, while remaining protected from attack inside phagosome-derived organelles, called parasitophorous vacuoles. Infection somehow promotes the translocation of LDs to the parasitophorous vacuole, establishment of intimate LD-vacuole contacts, and even uptake of the LDs into the vacuole (Cocchiaro et al., 2008; Toledo et al., 2016). Whether active LD motion is required for this process has not yet been established but seems likely. In mammary gland cells, LDs originate all over the cell and move along linear paths towards the apical surface, in an intermittent manner (Masedunskas et al., 2017). Once apical, they are released by an unconventional secretion mechanism, so that the LDs that end up in milk are wrapped by remnants of the plasma membrane. This pattern of LD motility is presumably functionally important: LDs grow in part by LD-LD fusion. It seems likely that fusion is enhanced by confining LDs to linear paths. In addition, active transport to the apical side is expected to promote contact with the plasma membrane for secretion. Once the mechanism driving LD motility in these cells is better described, these conjectures become testable.

Modulation of lipid exchange by LD motility may also be important in cancer. LDs often accumulate in cancer cells and profoundly affect their metabolism (Koizume and Miyagi, 2016). Intriguingly, in a number of cancer types, the velocity with which LDs move positively correlates with cancer severity (Nardi et al., 2019a). The mechanistic basis for this correlation remains to be worked out but might involve exchange of lipids between LDs and other organelles.

LDs do not only store lipids, but also specific proteins (Welte, 2007; Welte and Gould, 2017). Thus, LD motility might also promote delivery of proteins to their correct cellular destination. The strongest case for such a role comes from the life cycle of the Hepatitis C virus. Upon Hepatitis C infection, newly synthesized viral core protein accumulates on LDs and promotes LD relocalization to the perinuclear region (Boulant et al., 2008). This motion is hypothesized to promote contacts between sites of viral RNA replication and the core-loaded LDs to promote assembly of mature viral particles. In early Drosophila embryos, LDs store large amounts of specific histones (Cermelli et al., 2006); these histones promote proper chromatin assembly (Li et al., 2012; Johnson et al., 2018) and can help fight bacterial infections (Anand et al., 2012). LDs in these embryos are highly motile, which might enhance LD contacts with bacteria and thus killing activity as well as affecting how quickly histones reach the nucleus.

There may even be a role for LD motility in cytokinesis. In both mammalian cells (Atilla-Gokcumen et al., 2014) and budding yeast (Yang et al., 2016), interfering with triglyceride synthesis causes cytokinesis defects, possibly because toxic diglycerides accumulate (Yang et al., 2016). Intriguingly, during mammalian cytokinesis, neutral lipids – and thus presumably LDs – accumulate at the cytokinetic furrow and persist in the midbody, the structure that links the two daughter cells (Atilla-Gokcumen et al., 2014). Although these LDs might be generated on site, it is conceivable that they are transported from elsewhere in the cell to act as a local sink for newly produced toxic lipids. Alternatively, LDs might accumulate there via capture by actin filaments/microtubules after diffusion, as midbodies are particularly rich in cytoskeletal elements.

LD motility and global cellular LD distribution

A common pattern associated with LD motility is that it seems to switch LDs between two distinct distributions within the cell: clustered and dispersed (Fig. 2). The dramatic spatial rearrangement of LDs between these two states has the potential to radically alter the landscape of inter-organellar LD contacts (Fig. 2, insets). In at least one case, such a remodeling of the LD “interactome” has been directly demonstrated and linked to profound changes in the physiology of the cell (Herms et al., 2015). Namely, in starved Vero cells, LD dispersal promotes contacts with mitochondria, transfer of fatty acids from LDs to mitochondria, and thus both mitochondrial respiration and LD consumption (Herms et al., 2015).

Figure 2:

Alternative LD distributions may control organelle contacts and cellular metabolism.

The main cartoon depicts the two distributions LDs adopt in cells; insets highlight consequences for organelle contacts. The black arrows indicate the internal and external factors that can lead to switching between the two LD distributions. In the clustered distribution, LDs are stored away from interacting organelles and are thus preserved for later use. In the dispersed distribution, LDs have higher chance of contacting a partner organelle, allowing for LD utilization. For simplicity, the whole cell view displays a reduced number of cellular structures. The blown-up panels emphasize those structures implicated in affecting LD function through contacts: LDs (yellow), lysosomes (magenta), unidentified “other” vesicle (grey), mitochondria (oblong, two-tone grey), microtubule (purple filaments) emanating from centrosome (purple), nuclei (large blue circle), ER (red lines with a darkened interior lumen), and cytoplasm (light blue).

Switching between a dispersed and a clustered state has been observed under diverse circumstances. For example, it can occur in response to the cell’s nutritional status (Herms et al., 2015; Nguyen et al., 2017) or due to cell cycle progression (Cruz et al., 2019). Similar LD repositioning also happens developmentally, such as in canine oocytes (Ariu et al., 2016) as well as in newly fertilized mouse embryos (Bradley et al., 2016); in the latter case, nutritional changes in the culture medium alter LD distribution further. Early Drosophila embryos also undergo precisely timed LD repositioning, with LDs first spreading broadly through the peripheral cytoplasm, then accumulating near the central yolk, and finally returning to a dispersed configuration (Welte, 2015b). LD redistribution from dispersed to clustered can also result from infection with diverse intracellular pathogens (Boulant et al., 2008; Cocchiaro et al., 2008; Toledo et al., 2016). Finally, ectopic expression of Plin1 in fibroblasts or HEK293 cells causes perinuclear clustering of LDs; subsequent activation of Protein Kinase A disperses the LDs (Marcinkiewicz et al., 2006; Orlicky et al., 2013).

We propose that toggling between LD clustering and dispersal may allow cells to achieve distinct metabolic states, by quickly and globally altering the contacts LDs engage in. Details will depend on the particular cell type, especially the distribution of potential partner organelles. For example, lysosomes degrade LDs through multiple types of autophagy and show a polarized cellular distribution, often with a perinuclear pool that actively engages in autophagy (Cabukusta and Neefjes, 2018). In cells with that configuration, co-accumulation of LDs around the nucleus would then promote LD turnover via autophagy. Indeed, interfering with the perinuclear accumulation of lysosomes (via interfering with their Dynein-dependent motility) can impair LD turnover (Tapia et al., 2019).

A possible correlate of this idea of a motility-driven metabolic switch are recent observations on the intracellular positioning of LDs in human muscle. Muscle contains two distinct types of mitochondria, intermyofibular and subsarcolemmal, each with a corresponding pool of LDs. In highly trained skiers, LDs positioned in the intermyofibrillar region, but not LDs positioned in the subsarcolemmal region, were consumed during 1 hour of exhaustive exercise (Koh et al., 2017). Another study obtained similar results, showing that trained individuals store the bulk of their LDs in the intermyofibrillar region while individuals with type 2 diabetes had a higher portion in the subsarcolemmal region. Upon exercise training, the type 2 diabetic patients shifted to an ‘athlete like’ intermyofibrillar distribution of their LDs (Daemen et al., 2018). In obese human males, exercise increases intracellular interactions between LDs and mitochondria (Shepherd et al., 2017). As of yet, it is unclear whether the observed differential LD positioning in human muscle is due to regulated LD motility or other mechanisms, but the similarities to intracellular positioning in cultured cells are intriguing and warrant further studies.

The mechanisms that bring about the switch between clustered and dispersed states has been elucidated in two cases. In Drosophila embryos, LD repositioning is mediated by precise temporal regulation of the synthesis and destruction of the Kinesin co-factor Halo, which upregulates motion away from the nucleus and thus causes LD clustering near the yolk (Arora et al., 2016). In Vero cells, nutritional status is sensed via AMP Kinase; in response to kinase activation, the amount of detyrosinated tubulin increases and the detyrosinated microtubules arrange into a non-centrosomal array (Herms et al., 2015). LDs preferentially interact with these detyrosinated microtubules and disperse on them.

LD motility controls LD allocation during cell division

In addition to the two LD distribution states discussed in the previous section, there is another case where LD distribution is globally controlled, namely during cell division. As detailed below, there are numerous cases in which the LDs of the mother cell are unequally distributed to the daughters. This is particularly dramatic in many animal embryos, where different progeny derived from the same fertilized egg can end up with vastly different LD inheritance. Presumably, this differential inheritance amongst the daughters is determined by the different lipid needs and by different potential LD interactomes in their respective lineages. For example, some cells might rely heavily on triglycerides for energy metabolism and thus would favor interactions with mitochondria, while others predominantly require fatty acids substrates for lineage-specific signaling molecule production. Embryonic sister cells with differing LD inheritances have yet to be studied, but we propose that they potentially display related differences in LD interactomes and LD utilization. In the following, we summarize known instances of such uneven allocation and document what is known about the underlying LD motility events.

The behavior of LDs during cell division has not been extensively studied, but early indications are that LDs are actively allocated between daughter cells. For example, in NIH 3T3 cells, LDs associate tightly with spindle poles during mitosis, which presumably promotes even allocation amongst the daughters (Cruz et al., 2019). In the budding yeast S. cerevisiae, cytokinesis is asymmetric with the larger daughter (‘mother’ cell) retaining the majority of the cytoplasmic contents and the smaller daughter (bud) receiving a lesser share. LDs, in contrast, are actively transported by Myosin along actin filaments into the forming bud (Wolinski et al., 2011; Knoblach and Rachubinski, 2015). This LD-bud deposition method is similar to the mechanism that deposits mitochondria into the bud (Vevea et al., 2014). These similarities suggest that LDs and mitochondria are especially valuable, warranting an active transport mechanism into the nascent bud, unlike the randomly allocated bulk of cytoplasmic contents.

Interestingly, the allocation of LDs between different cells is not always balanced. In the fission yeast S. pombe, growth media low in nutrients induces spore formation, a type of specialized division in which a diploid precursor cell (ascus) generates four stress-resistant haploid spores. The spores are loaded with LDs from the ascus cytoplasm, using actomyosin machinery and providing the spore with energy until nutrient conditions improve (Yang et al., 2017). As a result of this allocation, the remaining ascus cytoplasm is depleted of LDs, while the spores are enriched for them. An analogous situation occurs during Drosophila oogenesis where sixteen sister cells share a common cytoplasm as a result of incomplete cytokinesis. One of the sisters, the oocyte, undergoes meiosis and eventually becomes haploid; the remaining fifteen nurse cells generate tens of thousands of LDs that are transported to the oocyte in an actomyosin dependent manner and provide a major energy source for the future embryo. It would be interesting to know if other types of asymmetric cell divisions (for example, those of stem cells) also are characterized by uneven allocation of LDs to daughter cells.

During animal embryogenesis, LD allocation from a fertilized egg to an ever-growing number of embryonic cells can be particularly dramatic, often involving large-scale LD redistribution (see Welte, 2009). In the eggs of the Japanese rice fish (Medaka), LDs are initially everywhere. As development proceeds, they accumulate via microtubule-dependent motion at the vegetal pole. In Drosophila oocytes, LDs are homogenously distributed, but by 90 minutes into embryogenesis, they are enriched near the plasma membrane and depleted from the center of the embryo (Welte, 2015a). In mouse oocytes, LDs are inhomogeneously distributed into aggregates at meiosis II, but after fertilization disperse all over the zygote (Bradley et al., 2016).

Many of these LD rearrangements result in uneven allocation of LDs between the cells of the developing embryo. In Medaka, LDs are depleted at the animal pole, i.e., the precursor of the embryo proper (blastodisc), and instead the yolk sac contains the majority of LDs. A similar pattern was observed in zebrafish, where the yolk region stains much more heavily with the LD dye Nile Red than the blastodisc; however, LDs are continually being transported into the blastodisc using actomyosin machinery (Dutta and Kumar Sinha, 2015; Gupta et al., 2017). In Xenopus oocytes, triglycerides (and thus presumably LDs) are also highly enriched at the vegetal versus the animal pole (Lee et al., 2006; Papan et al., 2007; Shrestha et al., 2014); as a result, at gastrula stages, LDs are highly enriched in the endoderm and depleted from the ectoderm (Sehy et al., 2001; Papan et al., 2007). In Drosophila, at the cellular blastoderm stage, the epithelial cells inherit the majority of LDs, while the central yolk cell contains relatively few (Welte et al., 1998). At the moment, the organismal consequences of such uneven allocation across the embryo remain unknown. However, it would not be surprising if it profoundly affects embryo physiology and development.

Conclusion

Since the topic of intracellular LD motility was first reviewed a decade ago (Welte, 2009), many more cases of such motility have been discovered, driven by the widespread use of live imaging and improvements in imaging technology. In a few cases, the basic mechanisms underlying motility are fairly well understood, providing a framework for probing its biological role. As the examples covered in this review show, there is already good evidence that LD motility contributes to establishing and breaking the contacts LDs make with other organelles. Most likely, this is just the tip of the iceberg, as the relationship between LD motility and contacts has so far been examined only in a limited number of cell types and for a few types of organelles. New ways to manipulate LD motility will provide ever sharper tools to distinguish the effect of LD motility from other cellular processes. Further, the ongoing search for specific regulators of that motility, including LD-specific motor anchors, will help to improve such tools and is thus important beyond simply understanding the mechanisms of motility. Eventually, the insights from studies of single cells should help make it possible to attack the challenging problem of why animal embryos invest considerable energy to rearrange their LD content and allocate it differentially between cells.