Lipid droplets are intracellular organelles that play central roles in lipid metabolism. In many cells, lipid droplets undergo active motion, typically along microtubules. This motion has been proposed to aid growth and breakdown of droplets, to allow net transfer of nutrients from sites of synthesis to sites of need and to deliver proteins and lipophilic signals. This review summarizes the current understanding of where, why and how lipid droplets move.
Lipid droplets are the prime sites where cells store triacylglycerols, sterol esters and other neutral lipids. Their structure is unique: a core of neutral lipids is surrounded by a monolayer of amphipathic lipids, with proteins either embedded in this monolayer or attached to its surface. Fat storage may not be the only biological role of lipid droplets: they have been proposed to actively participate in lipid metabolism and in signalling, to control intracellular lipid trafficking and to transiently capture proteins from other cellular compartments [1–3].
Lipid droplets are surprisingly complex and highly dynamic. Proteomic studies suggest that droplets contain dozens (if not hundreds) of different proteins , and RNAi (RNA interference) screens identified hundreds of genes involved in droplet biology [5,6]. Droplets can vary in size, intracellular distribution, lipid composition and protein content, sometimes even within a single cell. Droplets can grow and shrink, exchange proteins with their surroundings and physically and functionally interact with other organelles.
A particularly dramatic example of the dynamic nature of lipid droplets is that they can move around the cytoplasm. Sometimes droplets perform a seemingly random dance; sometimes many droplets move in a directed, highly co-ordinated manner. Droplet motion has been observed in cells as diverse as fungal symbionts of plants, insect oocytes, fish eggs and human liver cells (Figure 1) and has been proposed to play important roles in droplet biology, from biogenesis to communication with other organelles.
Examples of lipid-droplet motion in mammals (
A– F), fish ( G), insects ( H, I) and fungi ( J)
Droplet motion: the basics
Live imaging reveals that in many cells, lipid droplets move in a directed manner. These movements are typically fast (hundreds of nanometres per second) and sustained over seconds to minutes. The patterns of droplet displacement suggest that there two distinct driving forces: in some cases, the entire cytoplasm flows and droplets are just one of many cellular contents that get dragged along. Such ‘cytoplasmic streaming’, powered by rearrangements of the cytoskeleton, can be actin- or microtubule-based. In other cases, it is specifically the lipid droplets that move; in this ‘droplet motility’, droplets are pulled along cytoskeletal filaments by molecular motors.
In many systems, droplet motility is dominated by microtubule-based transport. For example, pharmacological disruption of microtubules typically stops droplet motion, whereas disruption of actin filaments has little effect [7–11]. Also, cytoplasmic dynein, a minus-end-directed microtubule motor, is often physically associated with droplets [12–14] and required for droplet motion [12,14–17]. Yet, the presence of actin and myosin I in some droplet proteomes [18,19] hints at ancillary roles for actin-based transport.
In some cells, the distribution of the entire cellular droplet population can undergo dramatic shifts, over timescales from tens of minutes to hours (Figures 1C, 1F, 1G, 1H and 1I). It is tempting to speculate that these changes are due to lipid droplets moving from one location in the cell to another. It is also conceivable that these droplets do not move as units, but are rather broken down into their components and regenerated at their destination. Distinguishing between these possibilities requires in vivo imaging at high enough temporal and spatial resolutions to follow the fate of individual droplets. In some cases, these global changes in droplet distribution clearly arise from the motility of individual droplets (e.g. [7,20]). In others, the mechanisms responsible for redistribution remain to be resolved.
Where do droplets move?
Cultured mammalian cells
Droplet motion has been reported for mammalian cell lines of diverse origin, including cells derived from hepatocytes, fibroblasts, kidney epithelia and the adrenal cortex [8–11,15,19,21]. In these cells, most droplets just oscillate, possibly stuck to the ER (endoplasmic reticulum); a small subset displays rapid, bidirectional motion along microtubules (Figure 1A). Individual droplets can switch between these two motility states .
Several intracellular pathogens induce droplet motion (Figures 1B and 1C). The bacterium Chlamydia trachomatis lives inside membrane-bound ‘parasitophorous vacuoles’ and induces the translocation of host lipid droplets into the vacuole . When hepatitis C virus infects cells, newly synthesized viral core protein accumulates on lipid droplets and causes them to aggregate near the microtubule-organizing centre, in a dynein-dependent redistribution .
The fat globules in mammalian milk originate from intracellular lipid droplets of mammary-gland epithelial cells. These droplets arise throughout the cells and are secreted apically (Figure 1D), in a unique process that involves enveloping them with plasma membrane remnants . Apical–basally oriented microtubules might provide highways for rapid droplet transit, and proteomic studies indeed identified dynein on mammary-gland droplets . Since tests with microtubule-depolymerizing drugs have yielded contradictory results , live imaging is needed to resolve whether droplet motility mediates droplet secretion.
The eggs of many animals are full of lipid droplets, which provide an important energy source for embryogenesis. As these eggs mature, they frequently undergo dramatic rearrangements. This ‘ooplasmic segregation’ often includes massive redistribution of lipid droplets, for example in fish, insects and annelids [20,24,25]. In the eggs of the fish Medaka, lipid droplets are initially evenly distributed and then accumulate at the vegetal pole (Figure 1G), via directed motility along microtubules [7,24].
Drosophila oocytes and embryos
During Drosophila oogenesis, lipid droplets and most other oocyte contents arise in nurse cells and are transferred through cytoplasmic bridges to the oocyte (Figure 1H). Transfer of nurse-cell components involves both actin-based cytoplasmic streaming and, for specific cargoes, active transport along microtubules. Timing of droplet transfer and inhibitor studies  suggest that most droplets travel passively by streaming. As nurse-cell components arrive in the oocyte, they are thoroughly mixed via large-scale, microtubule-motor-driven cytoplasmic streaming. Although most oocyte lipid droplets are carried by the bulk flow of streaming, a subset moves actively along microtubules, in both directions .
In early Drosophila embryos, essentially all droplets move bidirectionally along radially organized microtubules (Figure 1I, right). At specific times in embryogenesis, the balance between inward and outward motion shifts, resulting in dramatic global redistribution of the droplets : from an initial peripheral position, they first accumulate around the centre of the embryo and then spread back into the periphery (Figure 1I, left). Embryos may be able to compensate for unusual droplet distributions, as embryos with defects in inward accumulation [28,29] or outward spreading [14,20] develop apparently normally.
Arbuscular mycorrhizal fungi are symbionts that live in and around plant roots. Like many other fungi, they grow as thread-like filaments (hyphae), often connected in an elaborate branching network. Nutrients received from the plant host are stored as triacylglycerol in lipid droplets. These droplets undergo dramatic, bidirectional movements along the hyphae . Some droplets are carried by bulk cytoplasmic streaming; others move persistently in the opposite direction (Figure 1J).
Many pathogenic fungi use specialized structures, called appressoria, to break through the tough surfaces of their hosts. Lipid droplets appear to be crucial for appressorium function [31,32]. In the rice blast fungus Magnaporthe grisea  and the insect pathogen Metarhizium anisopliae , lipid droplets originate in fungal spores and redistribute to the incipient appressorium. The underlying mechanism is unknown, although several kinases are implicated .
How common is droplet motion?
Droplet motion has been detected in species from fungi to mammals. It is therefore reasonable to surmise that this phenomenon may be both ancient and much more widespread than published examples would suggest, especially since detecting droplet motion faces several experimental challenges. It requires live imaging at high acquisition rates; in some cell types, only a small subset of droplets moves at any given time ; and droplet motility can be modulated by extracellular stimuli (such a modulation was initially reported for unidentified intracellular vesicles ; these vesicles were later concluded to be lipid droplets ). I expect that future systematic studies will uncover many new instances of droplet motion.
Why do droplets move?
Bulk droplet movements probably promote nutrient delivery from one location to another. In mycorrhizal fungi, droplet motion may distribute the energy received from the plant throughout the hyphal network (Figure 1J). Milk secretion moves triacylglycerol from locations inside the cells into the extracellular space (Figure 1D), and in Drosophila, cytoplasmic streaming delivers droplets from their site of synthesis in nurse cells to the oocyte and thus ultimately to the embryo (Figure 1H). Finally, droplet uptake into Chlamydiabearing vacuoles provides the parasite with lipids essential for growth and survival (Figure 1B).
Biogenesis and breakdown of lipid droplets
Nascent lipid droplets can increase in size by fusing with each other, a dynein- and microtubule-dependent process [12,34] (Figure 1E). Droplet motion and growth often occur simultaneously; for example, in adipocytes, lipid droplets arise in peripheral lamellipodia and grow as they move towards the cell centre ; it remains to be explored whether growth and motion are functionally linked.
Conversely, when lipolysis is chronically stimulated in adipocytes, their large perinuclear lipid droplets fragment into tiny droplets that disperse throughout the cell  (Figure 1F). Fragmentation increases overall droplet surface area and thus presumably promotes access by lipases; dispersion may facilitate delivery of lipids throughout the cell. An interesting possibility is that microtubule motors help with both fragmentation and dispersion. An alternative proposal envisions breakdown of droplets in one location and resynthesis at another; in the latter case, lipids from shrinking droplets might diffuse through the expansive ER network to growing ones elsewhere in the cell .
Exchange with other cellular compartments
Lipid droplets display extensive interactions with other organelles, including the ER, mitochondria, endosomes and peroxisomes [3,37]. These contacts can be quite transient (e.g. ‘kiss-and-run’ interactions with phagosomes ) and may allow exchange of lipids [3,4]. Droplet motility may thus shuttle lipids between different membranous compartments.
Proteins that cannot travel efficiently through the aqueous intracellular environment might also ride piggyback on lipid droplets . A case in point is the newly synthesized core protein of hepatitis C; droplet motion towards the nucleus has been proposed to bring core-loaded droplets into close proximity with perinuclear sites of viral RNA replication, thus promoting assembly of new virus .
How do droplets move?
The motors powering droplet motion
Most examples of droplet motility represent motion along microtubules, powered by molecular motors. Such motors are directional, moving either towards the plus or the minus ends of the microtubules.
In all cases where a minus-end droplet motor has been identified, it is cytoplasmic dynein. This motor is physically associated with lipid droplets in mammary glands, fibroblasts and Drosophila embryos [12–14]. Functional inactivation of dynein blocks droplet motion and droplet fusion in fibroblasts [15,34], disrupts droplet motility in Drosophila embryos [14,16] and prevents hepatitis C-driven droplet clustering in hepatocytes .
How dynein attaches to lipid droplets is unknown. In fibroblasts, recruitment of dynein to droplets is regulated via ERK2 (extracellular-signal-regulated kinase 2)-dependent phosphorylation , although no binding partners have been identified. For other types of cargo, BicD (Bicaudal D) has been proposed to act as dynein tether. However, BicD is not required for dynein recruitment to Drosophila embryonic droplets; rather, it regulates the activity of both dynein and kinesin-1 .
In hepatocytes as well as in Drosophila embryos, lipid droplets are associated with the plus-end motor kinesin-1 [19,41]. In the embryos, this motor indeed powers droplet motion; when it is impaired either genetically or by antibody inhibition, droplet motion is disrupted . Force measurements indicate that multiple copies of kinesin-1 can be active simultaneously on a single droplet .
Like other cargoes that move back and forth along microtubules, bidirectionally moving droplets must employ plus- and minus-end-directed motors in quick succession . Vigorous droplet motion implies that the opposing motors do not compete with each other in a tug of war, but rather that their activities are co-ordinated .
The mechanism of co-ordination remains hotly debated . Proposals range from ‘co-ordinators’ that turn one type of motor off while the opposite-polarity motor is active  to the possibility that co-ordinated movement arises out of motor competition . Proper co-ordination requires the dynein cofactor dynactin [15,16] and, in Drosophila, the novel protein Klar [20,45], as disruption of these molecules severely impairs droplet motion in both directions.
PAT [perilipin/ADRP (adipose differentiation-related protein)/TIP47 (47 kDa tail-interacting protein)] proteins
PAT proteins form an ancient family of droplet-localized proteins with important roles in lipid metabolism; for example, they regulate the access of lipases to lipid droplets in mammals, insects and possibly fungi [1,46]. Intriguingly, three family members, ADRP, perilipin and LSD-2 (lipid storage droplet 2), have been shown to also control the intracellular distribution of droplets.
The Drosophila protein LSD-2 is required for correct developmental regulation of droplet motion in embryos . During embryogenesis, LSD-2 undergoes changes in its phosphorylation state that correlate with the direction of net droplet transport. LSD-2 binds to the proposed co-ordinator Klar  and might thus transmit developmental signals to motors and co-ordinators. One of these signals is probably the transiently expressed Halo protein: it modulates LSD-2's phosphorylation state  and promotes net plus-end transport .
In mammalian cultured cells, droplets are typically dispersed throughout. When ADRP is knocked down in human hepatoma cells  or when perilipin is ectopically expressed in mouse fibroblasts , lipid droplets cluster near the cell centre. Perilipin-induced droplet aggregation is reversed on sustained activation of PKA (protein kinase A) . Perilipin is a substrate of PKA, and a point mutation in one of PKA's target sites in perilipin prevents droplet dispersion .
It is intriguing that these PAT proteins can regulate droplet motion as well as lipid metabolism . In fibroblasts ectopically expressing perilipin, PKA-dependent phosphorylation of perilipin even induces lipolysis and droplet dispersion simultaneously in the same cells . Yet, these two processes can be uncoupled as they depend on distinct PKA target sites in perilipin . Thus dispersion and lipolysis are not necessarily linked, but can be regulated in a co-ordinated manner.
Caveolins are membrane proteins that shuttle between the plasma membrane, the Golgi and lipid droplets. When expressed in mammalian cultured cells, certain dominant-negative caveolin mutants (e.g. Cav3DGV) localize to lipid droplets constitutively, block microtubule-based droplet motility and prevent global droplet redistributions [8,21]. These mutants also interfere with the metabolic turnover of the droplets, revealing yet another link between droplet motion and lipid homoeostasis.
One target of Cav3DGV is the small G-protein Rab18. Rab18 localizes to lipid droplets in wild-type cells and is displaced from droplets by Cav3DGV expression . In turn, high levels of Rab18 cause loss of ADRP from the droplet surface . Rab18 is an attractive candidate for a regulator of droplet motion since many Rab proteins control trafficking and motility of specific vesicles.
Although it is clear that lipid droplets can display highly regulated motion in many different types of cells, the underlying mechanisms have so far been explored only in a few systems. Nevertheless, it is already apparent that certain components play similar roles in droplet transport in species as distinct as flies and mammals. It will be fascinating to uncover how widely conserved this droplet transport machinery is and how a shared machinery is modified to generate cell-type-specific droplet behaviour. A deeper understanding of the mechanisms driving droplet motion will also make it possible to critically test why droplets move, by examining the biological consequences of altered droplet motility.
The Dynamic Cell: Joint Biochemical Society and British Society for Cell Biology Focused Meeting held at Appleton Tower, University of Edinburgh, U.K., 1–4 April 2009. Organized and Edited by Ian Dransfield (Edinburgh, U.K.), Margarete Heck (Edinburgh, U.K.), Kairbaan Hodivala-Dilke (Cancer Research UK, London, U.K.), Robert Insall (Beatson Institute for Cancer Research, Glasgow, U.K.), Andrew McAinsh (Marie Curie Research Institute, Oxted, U.K.) and Barbara Reaves (Bath, U.K.).
I am grateful to Dawn Brasaemle, Sven-Olof Olofsson and Steven Gross for stimulating discussions, and Yi Guo, Susan Tran and Arno Müller for comments on this paper.
Research on lipid-droplet motion in my laboratory is supported by the National Institutes of Health [grant number GM64687].