The isoprenoid squalene is an important precursor for the biosynthesis of sterols. The cellular storage of squalene and its impact on membrane properties have been the subject of recent investigations. In a screen for abnormal lipid droplet morphology and distribution in the yeast Saccharomyces cerevisiae, we found significant lipid droplet clustering (arbitrarily defined as an aggregation of six or more lipid droplets) in a number of mutants (e.g. erg1) that are defective in sterol biosynthesis. Interestingly, these mutants are also characterized by accumulation of large amounts of squalene. Reducing the level of squalene in these mutants restored normal lipid droplet distribution. Moreover, inhibition of squalene monooxygenase in two mammalian cell lines (CHO-K1 and 3T3-L1) by terbinafine also resulted in lipid droplet clustering. These results indicate that the level of squalene may affect the growth and distribution of lipid droplets.
Lipid droplets (LDs) are energy storage organelles that comprise a neutral lipid core of triacylglycerols (TAG) and sterol esters (SE), surrounded by a monolayer of phospholipids with proteins embedded [1-5]. Formation of LDs is thought to occur in the endoplasmic reticulum (ER) [6, 7]. The most favored model of LD biosynthesis [6-9] proposes that newly synthesized neutral lipids accumulate between two leaflets of the endoplasmic reticulum (ER) membrane before budding into the cytosol [10, 11]. Varying in size and composition, LDs are found in nearly all eukaryotic cells [12-15], and provide a store of energy and of bioactive lipids, such as fatty acids and sterols . For instance, although mature adipocytes usually contain one or a few huge TAG-rich LDs (25-300 μm in diameter) , other mammalian cell types contain smaller neutral LDs with different TAG : SE ratios. Plant seeds usually store abundant TAG-rich LDs, and some seed tissues may contain as much as 76% lipid w/w . Depending on the growth phase, cells of the budding yeast Saccharomyces cerevisiae contain up to a dozen lipid particles/droplets (~ 0.4 μm in diameter) per cell [1, 18], which are made up of equal amounts of TAG and SE [1, 19, 20].
LDs may grow in size in response to increased lipid synthesis and/or uptake . Giant or supersized LDs represent the most efficient form of lipid storage in terms of surface to volume ratio. Small LDs, on the other hand, provide more surface area for LD-associated proteins, such as lipases. Giant LDs may also affect the cell structure and cytoskeleton in a negative way, given their sheer volume. Therefore, the growth and final size of LDs have important implications in cell biology and function. Recent studies have suggested that LDs may grow through fusion, and the clustering of LDs may be a prerequisite for LD growth and fusion . Despite rapid progress in LD research, the fundamental mechanisms that govern the formation, size and distribution of cellular LDs are still unclear. Little is known about how the size of LDs is determined and what factors regulate the localization/distribution of mature LDs .
In an attempt to isolate novel regulators of the cellular dynamics of LDs, we have performed screens of yeast mutants using Nile red to stain LDs, in order to uncover mutations that affect cellular LD dynamics, in particular LD size and distribution. We have previously found that Fld1p (a human seipin homolog) and phospholipids, especially phosphatidic acid, play important roles in determining LD size [18, 22-25]. We report here a screen examining the role of essential genes in LD dynamics, in which we have uncovered a unique ‘LD clustering’ phenotype in cells lacking ERG1, which encodes squalene epoxidase, an important enzyme in ergosterol biosynthesis (Fig. 1). We have extended this finding by showing that accumulation of squalene in yeast through other means also leads to LD clustering. Finally, we have shown that increased squalene is also associated with LD clustering in mammalian cells.
ERG1 suppression results in LD clustering
To identify gene disruptions that affect the cellular dynamics of LDs in S. cerevisiae, Nile red, a vital dye specific for intracellular LDs , was used to visually examine the morphology of LDs in the Hughes collection of yeast essential gene mutants . Under the conditions used, wild-type cells usually produce three to eight LDs of ~ 0.4 μm diameter that are loosely distributed in a cell (Fig. 2A). Occasionally, two or three LDs make contact with each other, but clustering of LDs (arbitrarily defined as an aggregation of at least six LDs in one cluster) is rarely seen. We screened for mutants that harbor LDs with striking morphological changes, and found a mutant with supersized LDs, details of which were reported recently . We also identified a single mutant (erg1) with extensive clustering of LDs. In this mutant, LD clustering/aggregation was observed under fluorescent microscopy (Fig. 2B) and electron microscopy (Fig. 2C). LDs of erg1 cells were very irregular in terms of localization. Up to 80% of the total population of erg1 cells contained clusters of LDs of normal size (0.3–0.5 μm diameter). The remaining ~ 20% of the erg1 cells showed a similar LD phenotype to the wild-type strain. Under the same growth conditions, wild-type cells contained 6.5 ± 3.0 LDs per cell (mean ± SD; n =200), and 95% of the cells displayed small, almost spherical LDs (< 0.3 μm diameter) that were equally distributed in the cell.
To confirm a specific role for Erg1p in LD morphology, complementation was performed using a vector expressing GFP-tagged ERG1. Expression of ERG1–GFP in erg1 cells restored the normal morphology of LDs (Fig. 2D). Localization of ERG1–GFP in both LDs and the ER (Fig. 2E) confirmed the previously reported localization of Erg1p .
The ERG1 gene encodes squalene epoxidase (monooxygenase), which converts squalene to 2,3-oxidosqualene during sterol biosynthesis  (Fig. 1). Squalene is known to accumulate in the erg1 mutant strain, and its level may be as much as four times higher than that of the wild-type strain. The level of squalene, TAG and SE was examined under our experimental conditions, and, as expected, a fourfold increase in squalene was found in the erg1 mutant, but no significant change in TAG was detected (Fig. 2F). Almost no sterol esters were detected in the erg1 mutant, indicating a severe defect in ergosterol synthesis due to Erg1p deficiency. The increased level of squalene, but not TAG or SE, in the erg1 mutant implied that squalene may be responsible for LD clustering.
Squalene accumulation by other genetic or pharmacological means is associated with LD clustering
To further establish a link between squalene and LD clustering, we used additional genetic and pharmacological means to manipulate cellular squalene levels, followed by examination of LD dynamics. We first tested the inhibition of squalene epoxidase in wild-type cells by adding 10 nm terbinafine , which led to ~ 70% of the cell population showing LD clustering, whereas only 7.5% of control cells had a similar phenotype (Fig. 3A,B). The squalene level of treated cells was increased by ~ 350% compared to untreated cells. The TAG content of treated cells was also increased to the same extent as squalene, but the SE content decreased by ~ 30% compared to untreated cells (Fig. 3C).
During sterol biosynthesis, squalene epoxide is converted to lanosterol, the first sterol in the pathway, catalyzed by the oxidosqualene cyclase (lanosterol synthase) Erg7p. Lanosterol is then further converted in several steps to ergosterol, the major membrane sterol in yeast . Disruption of ERG7 causes an increase in cellular squalene in S. cerevisiae . As expected, clustering of LDs (Fig. 4A,B) was observed in ~ 60% of the cell population of the erg7 strain in the presence of doxycycline. Under our growth conditions, the level of squalene in the erg7 mutant was 1.5-fold higher than in the wild-type (Fig. 4C). In contrast, levels of TAG and SE in the erg7 strain were 65% and 0%, respectively, of those of the wild-type under the same growth conditions (Fig. 4C).
Another strategy to increase cellular squalene in yeast is disruption of HEM1 . hem1∆ strains have been commonly used as a model for anaerobic growth , and heme is essential for the activity of Erg11p. Squalene accumulates in hem1 mutant cells when heme is deficient, and such cells are unable to produce sterol . hem1∆ cells showed a defect in LD morphology in which ~ 90% of the cells had a clustering phenotype (Fig. 5A,B). The squalene level in this mutant was increased ~ 3.5-fold compared to the wild-type (Fig. 5C).
Squalene synthase (Erg9p) converts farnesyl pyrophosphate to squalene. Over-expressing ERG9 may cause squalene accumulation and LD clustering. Indeed, ~ 50% of yeast cells over-expressing ERG9 showed LD clustering (Fig. 6A,B), and the level of squalene in the ERG9 over-expressing strain was 1.5-fold higher than in the wild-type (Fig. 6C). The level of TAG and SE also increased (Fig. 6C).
Accumulation of upstream intermediates of ergosterol biosynthesis does not cause LD clusters
Squalene accumulation may lead to secondary accumulation of upstream intermediates of ergosterol biosynthesis (Fig. 1). It is possible that some upstream intermediates other than squalene may induce LD clustering. To assess this possibility, we examined the LD morphology and lipid content of cells with compromised function of Erg9p (squalene synthase), Erg8p (phosphomevalonate kinase) or Erg12p (mevalonate kinase) (Fig. 1). Erg9p is inhibited by zaragozic acid (10 μg/ml) . The majority (94%) of cells treated with zaragozic acid showed a similar LD phenotype to the wild-type (Fig. 7A). The squalene level in treated cells was decreased by ~ 20% compared to control cells (Fig. 7E). Two other upstream conditional mutants (erg8 and erg12) from the Hughes collection were treated with doxycycline. Less than 8% of erg8 cells (Fig. 7B) showed the LD clustering phenotype, and their squalene level was 15% lower than that of the wild-type (Fig. 7E). Approximately 11% of erg12 cells treated with doxycycline showed LD clustering (Fig. 7C), and the squalene level in the mutant strain was ~ 23% lower than that in the wild-type (Fig. 7E). The levels of squalene, TAG and SE from wild-type and various mutants were compared and correlated with LD clustering (Fig. 8A–D). Based on statistical analysis (Fig. 8E), there is a strong (r =0.75) and significant (P <0.005) positive correlation between the percentage of cells showing the LD clustering phenotype and the squalene content, but a non-significant correlation between the percentage of cells showing the LD clustering phenotype and the SE (r =0.08) or TAG (r =0.11) content (Fig. 8F,G). Together, these data indicate that accumulation of squalene, but not other intermediates, appears to account for LD clustering.
Squalene accumulation in mammalian cells is also associated with LD clustering
Squalene accumulation may also cause LD clustering in mammalian cells. To study this, terbinafine was used to inhibit squalene epoxidase. Two mammalian cell lines (CHO-K1 and 3T3-L1) were treated with: (a) no terbinafine or oleic acid (control), (b) 10 nm terbinafine but no oleic acid, (c) oleic acid (0.4 mm) only, and (d) terbinafine (10 nm) and oleic acid. Microscopy revealed that the clustering of LDs was increased in CHO-K1 (Fig. 9A,d) and 3T3-L1 (Fig. 9B,d) cells by up to 48% and 53%, respectively, under terbinafine and oleic acid treatment.
Lipid droplets  are dynamic cellular organelles whose number, size and distribution respond quickly to environmental changes [4, 35]. However, the factors and mechanisms that govern LD size, number and localization remain to be elucidated [10, 22]. Recent genome-wide studies have identified some genes that affect LD dynamics, but additional factors may also exist [22, 36]. In this study, we provide strong evidence supporting a previously unrecognized role of squalene in LD clustering in both yeast and mammalian cells.
Clustering/aggregation of lipid droplets has been observed under various conditions and in various cell types [18, 22, 36, 37], but has never been clearly defined to our knowledge. We arbitrarily defined clustering as an aggregation of at least six LDs. Based on this criterion, we found almost no LD clustering in wild-type yeast cells but extensive LD clustering upon suppression of ERG1, which encodes squalene epoxidase. As squalene is a highly hydrophobic neutral lipid that is found in LDs and between membrane leaflets , we reasoned that squalene accumulation may be the cause of LD clustering. Indeed, in other mutants and under conditions in which squalene is known to accumulate, LDs invariably aggregate. In addition, squalene also appeared to cause LD aggregation in two mammalian cell lines.
Importantly, the levels of major LD neutral lipids, TAG and SE, did not increase significantly under most conditions used (Fig. 8B,C). It is interesting to note that SE was almost absent in some mutants because of defective sterol biosynthesis (Fig. 8B). Moreover, blocking enzymes upstream of Erg1p did not result in LD clustering. Statistical analyses indicated that there is a strong positive correlation between the percentage of cells showing clustering and the squalene level. Together, these data support a role for squalene, but not TAG, SE or sterol intermediates upstream of squalene, in triggering LD aggregation under the conditions used.
Why do LDs aggregate? This is probably a normal response to the increased load of neutral lipids. It may also be a prelude to LD fusion, as LD clustering is probably a required step for fusion of LDs. In this regard, it is interesting to note that, in yeast cells lacking Fld1p (a human seipin homolog), supersized LDs as well as clusters of multiple small LDs were easily identifiable . These observations suggest that, at least in the fld1∆ mutant, small LDs may aggregate before they fuse, giving rise to supersized LDs. However, the molecular machinery that promotes LD fusion remains to be identified in wild-type yeast cells and other cell types under normal growth conditions.
Precisely how excess squalene causes clustering of LDs is presently unclear, but may relate to its ability to regulate biophysical membrane properties . Elegant work by Spanova et al. has shown that squalene may be accommodated in yeast lipid particles and organelle membranes without deleterious effects . Moreover, they have shown that accumulation of squalene alone (i.e. in the absence of TAG and SE) is not sufficient to initiate the formation of lipid droplets/particles . This indicates that squalene behaves differently from TAG/SE in the initial ‘budding’ events of LD formation. However, our results suggest that squalene promotes LD clustering, just as TAG and SE. Therefore, it is possible that accumulation of other neutral lipids may be able to promote LD clustering. It is important to note that we observed prominent LD clustering but, surprisingly, no supersized LDs in the mutants in this study. It appears that a significant increase in the level of neutral lipids such as squalene, TAG or SE may be sufficient to trigger LD aggregation, but not sufficient for LD fusion under the conditions used. Therefore, although LD clustering may be a prerequisite for LD fusion, LD coalescence may require additional protein factors and/or changes to the phospholipid monolayer [39, 40].
Squalene is found in health stores due to its anti-cancer and antioxidant effects. It is also commonly used as an adjuvant in vaccines. Squalene may accumulate upon increased dietary intake or injection. Squalene accumulation may also occur in a normal human physiological context, based on recent work on the mammalian protein equivalent of ERG1, squalene monooxygenase . This work showed that cholesterol accelerates the proteasomal degradation of squalene monooxygenase, leading to accumulation of squalene, and suggests that this enzyme serves as a crucial control point in cholesterol synthesis beyond 3-hydroxy-3-methyl-glutaryl (HMG) CoA reductase . Although the corresponding yeast protein lacks the mammalian regulatory N-terminal domain, it is noteworthy that squalene monooxygenase has been isolated from lipid droplets in both yeast  and mammalian cells . Importantly, the enzyme (at least in the yeast system ) was found to be inactive in LDs, which is expected to result in squalene accumulation. Together, our results provide important insights into the impact of squalene on cellular lipid storage in the form of LDs.
In summary, we have characterized LD clustering for the first time, and identified squalene as an important regulator of LD aggregation by screening of yeast essential gene mutants. Additional protein and lipid factors that regulate LD distribution may also exist and await future exploration. As we are just beginning to appreciate the cellular dynamics of LDs, it is critical to identify all genetic factors that regulate LD size, movement and distribution. The results presented in this study should facilitate future endeavors to understand LD dynamics.
Both wild-type R1158 (URA3::CMV-tTA MATa his3-1 leu2-0 met15-0) and single deletion mutants (Table 1) were obtained from the yeast Tet promoters Hughes (yTHC) collection . Each strain was verified by its growth phenotype on yeast extract/peptone/dextrose (YPD) medium and minimal synthetic defined medium  containing 10 μg·mL−1 doxycycline. Without doxycycline in the medium, the Tet promoter is fully activated. Addition of doxycycline allows down-regulation of the promoter in a titratable manner until the gene of interest is no longer expressed at detectable levels.
Table 1. Genes for which mutants of the yeast Tet promoters Hughes (yTHC) collection were used
Nile red, oleic acid, TAG and SE standards, squalene, Brij58, doxycycline, terbinafine and zaragozic acid were purchased from Sigma-Aldrich (St Louis, MO, USA). Antibodies were purchased from Invitrogen (Carlsbad, CA, USA).
Yeast cells were grown with rotary shaking at 30 °C in liquid YPD medium (1% yeast extract, 2% w/v bactopeptone, 2% w/v dextrose). Alternatively, yeast cells were grown in minimal synthetic defined medium (0.67% Difco nitrogen base without ammonium sulfate, 0.5% ammonium sulfate and 2% dextrose, with all amino acids supplemented). Plasmid-carrying strains were grown on synthetic medium with appropriate selection as described previously .
Inactivation of the HEM1 gene by insertion of a hem1::hphNT1 disruption cassette was performed as described previously . Plasmid pYM16 (euroscarf, European Saccharomyces Cerevisiae Archive for Functional Analysis) was used as the template for PCR to create the disruption cassette, using the primers listed in Table 2. The disruption cassette was transformed into the wild-type strain by standard procedures to integrate the cassette into the yeast genome via homologous recombination. The disruption of HEM1 in the mutant was confirmed by colony PCR. The hem1∆ mutation was confirmed by a spot test in which the hem1∆ strain and wild-type control were grown on minimal synthetic defined medium  lacking histidine and supplemented with 20 μg·mL−1 ergosterol and 0.06% Tween-80.
Construction of a vector expressing GFP-tagged Erg1p
A 2.3 kb fragment containing the native promoter and the entire coding sequence of ERG1 was amplified by PCR using the primers listed in Table 2. The amplified fragment was then sub-cloned into HindIII- and XbaI-cleaved YCplac111-GFP plasmid , which has the GFP-coding sequence inserted between the BamHI and EcoRI restriction sites. The confirmed construct was transformed into the tetO7-erg1 strain. Expression of ERG1-GFP complemented the tetO7-erg1 phenotype, as confirmed by growing the transformed strain in SD medium without leucine.
Over-expression of ERG9
A 1.3 kb fragment carrying the entire coding sequence of ERG9 and lacking the upstream promoter was amplified by PCR using the primers listed in Table 2. The amplified fragment was then sub-cloned into BamHI- and XhoI-cleaved pESC-His-c-myc tag plasmid (Agilent Technologies, Santa Clara, CA, USA). The confirmed construct was transformed into the wild-type strain by standard procedures. Successful transformation in the wild-type was confirmed by standard western blot analysis in which Myc antibody was used as primary antibody against the c-Myc tag and a donkey anti-mouse secondary antibody (Jackson ImmunoResearch Laboratories, West Grove, PA, USA) was used.
Fluorescence imaging was performed on a DM5500B microscope (Leica, Deerfield, IL, USA). Samples were viewed using a 100 × NA 1.30 oil-immersion objective lens (Leica). Images were taken using a DFC480 digital camera (Leica) and LAS AF (Leica Application Suite AutoFocus) version 2.0.0 software (Leica).
Nile red (Sigma-Aldrich) is a specific vital dye for intracellular LDs. A 450–490 nm band-pass excitation filter, a 510 nm dichromatic mirror, and a 515 nm long-pass emission filter (filter cube I3; Leica) were used to view LDs.
The GFP signal was visualized using a 470/40 nm band-pass excitation filter, a 500 nm dichromatic mirror, and a 525/50 nm band-pass emission filter (filter cube GFP; Leica).
Transmission electron microscopy
Cells were grown in YPD and SD media with doxycycline until stationary phase, harvested, fixed with 2.5% w/v glutaraldehyde, and post-fixed with 2% w/v osmium tetroxide. The samples were subsequently dehydrated in a graded ethanol series and embedded in Spurr's resin (Sigma-Aldrich, St. Louis, MO, USA). Ultra-thin sections (80 nm) were stained with uranyl acetate and lead citrate, and examined under a JEM-1230 electron microscope (JEOL, JOEL Australia, Frenchs Forest, NSW, Australia).
Thin-layer chromatographic quantification of neutral lipids and squalene
Lipid was extracted from yeast cells as previously described, with minor modifications . In brief, cells were grown in the required medium until the required growth phase (quantified by A at 600 nm), harvested and washed once with ice-cold NaCl/Pi. The dried cell pellets were resuspended in 100 μL lyticase (Sigma-Aldrich) (1700 U/ml in 10% glycerol), then incubated at 37 °C for 1 h, at −70 °C for 1 h and at 37 °C for 15 min. Lipids were extracted with hexane and air-dried. Quantification of neutral lipids was performed as described previously . Samples and TAG, SE and squalene standards were dissolved in 100 μL chloroform/methanol (2 : 1 v/v) and spotted on silica gel 60 F 254 plates (Merck, Kilsyth, Victoria Australia) using a TLC auto-spotter (Analtech Inc. Newark, DE, USA). Chromatograms were developed in hexane/diethyl ether/acetic acid (85:15:1). Dried plates were derivatized by heating, and scanned using a Canon scanner (North Ryde, NSW, Australia). For each assay, at least three independent samples were tested, and mean and SD values were calculated.
Mammalian cell culture
Chinese hamster ovary cells K1 (CHO-K1) from the American Type Culture Collection were grown as a monolayer in a humidified incubator at 37 °C in a 5% CO2 atmosphere. Normal medium included 10% newborn calf serum (NCS) in Ham's Nutrient Mixture F-12 (DF12) medium (Life Technologies, Mulgrave, Victoria, Australia). Medium containing oleic acid was prepared by mixing 11.6 ml 5% BSA, 3.5 ml oleate (20 mm) and 160 ml of normal medium (DF12 + NSC).
Normal medium for the 3T3-L1 adipocyte cell line (from the American Type Culture Collection) included 10% fetal bovine serum and Dulbecco's modified Eagle's medium. Medium containing oleic acid was the same as described above.
On the first day, cell lines were set up in triplicate at a density of 4 × 105 cells per 60 mm dish in normal medium. On the second day, cells were transferred into a new Petri dish with a cover glass. Four treatments were performed, including: (a) normal medium without oleic acid and terbinafine, (b) medium with oleic acid, (c) normal medium with terbinafine (10 nm), and (iv) medium with terbinafine (10 nm) and oleic acid.
Microscopy of mammalian cells
Cells were rinsed once with 1 × Dulbecco's modified Eagle's medium/NaCl/Pi and fixed with 3.7% v/v formaldehyde/NaCl/Pi for 10 min at room temperature. Samples were washed twice with 1 x NaCl/Pi, and incubated with LipidTOX (Invitrogen, Mulgrave, Victoria, Australia)/NaCl/Pi (1:1000 v/v) in the dark at room temperature for 1 h. Cells were washed once with 1 x NaCl/Pi and fixed to glass slides using mounting buffer (90% glycerol, 1 m Tris pH 6.8). Images were obtained using a Leica DM5500B microscope. GFP was excited at 488 nm from an argon laser. Images were edited using LCS Lite confocal software (Leica).
This work is jointly supported by a PhD scholarship from the Vietnamese Government and the University of New South Wales, Australia, and by grants from the Australian Research Council (DP120101543 and DP0984902). H.Y. is a Future Fellow of the Australian Research Council.