To investigate whether vital staining by Nile Red or BODIPY could label both LROs and lipid droplets, we grew wild-type and glo-4(ok623) animals on OP50 E. coli diet supplemented with Nile Red or BODIPY. glo-4 encodes a putative guanine nucleotide exchange factor (GEF) for the GLO-1 Rab GTPase. glo-1 and glo-4 mutants lacked LROs 7 . Consistent with a previous report 7 , there was no detectable Nile Red-labelled structure in glo-4 mutants (Figure 1A and 1B). Using the same imaging parameters, no autofluorescence was detected in either wild-type or glo-4 animals that were not vital-stained with Nile Red (Figure 1C and 1D). These results support the idea that Nile Red labels LROs only. In contrast, we observed in gut epithelial cells of wild-type animals two populations of BODIPY-labelled structures, one with high fluorescence intensity and the other with lower fluorescence intensity (Figure 1E). In glo-4 mutants, the low-intensity population remained (Figure 1F), while the high-intensity population (Figure 1E) and LRO-specific autofluorescence 7 were lost (Figure 1G and 1H). These results suggest that the high-intensity BODIPY structures are LROs while the low-intensity BODIPY structures are putative fat storage structures.

To investigate whether the high-intensity BODIPY structures are LROs, we grew wild-type animals bearing a transgenic marker for LROs (hjIs9 [glo-1::gfp]) on OP50 diet supplemented with a red BODIPY fatty acid analog, BODIPY558/568-C12. Indeed, GLO-1::GFP encircled high-intensity red BODIPY structures (Additional File 1). This is consistent with a previous finding that the high-intensity green BODIPY signal co-localizes with lysoTracker, a marker for LROs 8 . However, red BODIPY also labelled GLO-1::GFP-negative structures, albeit with lower fluorescence intensity (Additional File 1). In peroxisomal β-oxidation mutants, enlarged lipid droplets which are GLO-1::GFP-negative, can be labelled by BODIPY 21 . Thus, the GLO-1::GFP-negative structures with low BODIPY intensity in wild-type animals may also be genuine fat storage structures, i.e., putative lipid droplets.

To confirm that bona fide neutral fat storage structures in live wild-type animals can be labelled by BODIPY, we adopted and modified previous differential centrifugation approaches 22 23 to partially isolate lipid droplets from dhs-28(hj8) and wild-type animals that were vital-labelled by BODIPY. This approach exploits the fact that lipid droplets have a lower density than water and most other intracellular organelles. Indeed, enlarged and small lipid droplets isolated from dhs-28 mutants were shown to be labelled by green BODIPY (Figure 2A-2C). Also, small putative lipid droplets isolated from wild-type animals were shown to be labelled by green BODIPY, albeit with low intensity (Figure 2D-2F). Similarly, enlarged lipid droplets isolated from dhs-28 were labelled by red BODIPY(Figure 2G-2I), and small putative lipid droplets isolated from wild-type animals were labelled by red BODIPY with low intensity (Figure 2J-2L). Furthermore, we prepared lipid droplet fraction and cytoplasmic fraction from dhs-28 mutant animals bearing the glo-1::gfp transgene (dhs-28;glo-1::gfp). Western blot experiments showed that the lipid droplet fraction had no detectable GLO-1::GFP protein, which was found predominantly in the cytoplasm fraction (Additional File 2). This result demonstrated that our biochemical approach could yield lipid droplet fractions devoid of LROs.

A previous report 24 and our unpublished observations (see below) demonstrated that Nile Red could vital-label enlarged lipid droplets and other small spherical structures in peroxisomal β-oxidation mutants. We have shown that although the loss of glo-4 function leads to a loss of LROs, it did not prevent the formation of enlarged lipid droplets in mutants with defective peroxisomes 21 . Interestingly, in daf-22; glo-4 double mutants, enlarged lipid droplets and small putative lipid droplets could be vital-labelled by Nile Red (Figure 1I) as well as by BODIPY (Figure 1J). It further supports the notion that, in contrast to lipid droplets in wild-type animals, lipid droplets in peroxisomal β-oxidation mutants can be vital-labelled by Nile Red.

Here, we sought to compare fluorescence properties of Nile Red-labelled LROs in wild-type animals and Nile Red-labelled lipid droplets and LROs in peroxisomal β-oxidation mutants. We scanned Nile Red emission spectra in wild-type and dhs-28 animals. In wild-type animals, all Nile Red-positive structures, i.e., LROs, displayed the same emission spectrum that peaked at 641 nm (Figure 3A, 3B and 3G). We classified these structures as "Spec-1". In contrast, in dhs-28 mutants, there were structures with a different spectrum that peaked at 631 nm and shouldered at 590 nm and 610 nm in addition to Spec-1s (Figure 3C, 3D and 3G). We classified these structures as "Spec-2". In wild-type and dhs-28 mutant animals, the Spec-1s were mostly smaller than 3 μm in diameter and encircled by the LRO marker GLO-1::GFP (Figure 3E and 3F). In dhs-28 mutant animals, the Spec-2s were rather heterogeneous in size, ranging from below 1 μm to 8 μm in diameter (Figure 3D and 3F). Importantly, all enlarged lipid droplets were Spec-2s (Figure 3D and 3F). Consistent with the notion that glo-4 encodes a guanine nucleotide exchange factor for GLO-1, GLO-1::GFP marking was lost in glo-4 mutants (Figure 3H and 3I). In daf-22;glo-4 double mutants, all Spec-1s were lost, while Spec-2s, i.e., enlarged lipid droplets and small putative lipid droplets, remained (Figure 3J).

We next sought to obtain vital Nile Red-stained subcellular structures using our differential centrifugation procedure. To this end, we grew dhs-28 mutants on 10-cm NGM/OP50 plates that were overlaid with Nile Red. We then prepared lipid droplet fractions from ~12,000 dhs-28 animals harvested from 10-cm NGM/OP50/Nile Red plates, and pelleted cytoplasmic membranous organelles that included LROs. As shown in Additional File 3, isolated membrane pellets displayed Spec-1 spectrum similar to Spec-1 in intact dhs-28 animals; and isolated lipid droplets displayed Spec-2 spectrum similar to Spec-2 in intact dhs-28 animals. For reasons currently unknown, we noted that dhs-28 animals grown and stained on 10-cm NGM/OP50/Nile Red plates displayed more prominent 590 nm and 610 nm peaks of Spec-2s than their counterparts that were grown and stained on 6-cm NGM/ OP50/Nile Red plates (compare Additional File 3D with Figure 3G).

We also probed whether red BODIPY had different fluorescence emission spectra when labelling LROs vs. lipid droplets. In dhs-28 mutant animals bearing GLO-1::GFP marker, both GFP-encircled LROs and non-GFP-encircled large lipid droplets displayed the same emission spectrum peaked at 577 nm and shouldered at 620-631 nm (Additional File 4). Thus, BODIPY labelling did not distinguish lipid droplets from LROs as Nile Red labelling did.

We interpret these results as indicating that in wild-type C. elegans, lysosome related organelles (LROs) are vital-labelled by both BODIPY and Nile Red with high intensity. On the other hand, lipid droplets, the fat storage structures, can be vital-labelled by BODIPY with low intensity but not by Nile Red. Under special conditions such as peroxisomal β-oxidation dysfunction, lipid droplets can be labelled by Nile Red and be distinguished from LROs by a distinct fluorescence emission spectrum. Since in any situation, both BODIPY and Nile Red stain LROs strongly, fluorescence intensities of the two vital lipophilic dyes should be used with great caution as surrogates of intracellular fat levels.

Recently, it was found that the intensities of post-fix Nile Red or Oil-Red-O staining correlated more closely with quantitative TAG measurements by GC-MS than the intensities of vital Nile Red or BODIPY staining 8 20 . However, the underlying principle of the post-fix staining as a better proxy of fat level is not clear. The post-fix approach involves permeating cell membrane and intracellular membrane with isopropanol and freeze-fracture. This treatment presumably allows direct penetration of the two dyes.

We hypothesized that the invasive manner in which Nile Red and Oil-Red-O were delivered allowed targeting of these dyes to lipid droplets that bypassed LROs. To test this hypothesis, we stained fixed wild-type and glo-4 animals with Nile Red. Consistent with a previous report 20 , post-fix Nile Red staining labelled spherical structures in both wild-type and glo-4 animals (Figure 4A and 4C), which were not otherwise vital-labelled by Nile Red in glo-4 mutants (Figure 1B). More interestingly, spectral analyses of emission spectra of Nile Red in wild-type and glo-4 animals revealed that all labelled structures displayed the same Spec-2 property as that of lipid droplets in peroxisome mutants (Figure 4B, 4D and 4E). Moreover, post-fix Nile Red staining did not label any Spec-1 structures, i.e., LROs (Figure 4B and 4D), which were otherwise the only structures labelled by Nile Red in live wild-type animals (Figure 3B).

The above imaging and biochemical isolation experiments indicate that fat storage structures in wild-type and glo-4 animals possess the same properties as enlarged lipid droplets in mutants defective in peroxisomal function. Next, we sought to confirm that they also shared the same properties at the ultrastructural level, similar to mammalian lipid droplets that are characterized as electron-translucent structures with a phospholipid monolayer membrane 3 .

Close inspection of electron-translucent structures in wild-type animals revealed that they were indeed similar to enlarged lipid droplets in daf-22 mutant animals 21 . These structures, with diameters normally smaller than 3 μm, possessed a phospholipid monolayer membrane (Figure 5A-5C), the hallmark of lipid droplets. In glo-4 mutants, lipid droplets were not eliminated (Figure 5D-5F). In daf-22;glo-4 double mutants, neither small nor large lipid droplets were eliminated (Figure 5G-5H). These results demonstrate that lipid droplets are universal fat storage structures in C. elegans and that lipid droplets are fundamentally distinct from LROs in their structural properties and their sensitivity to different genetic regulatory mechanisms.

The wild-type strain was N2 Bristol. The following mutant and transgenic strains were used: daf-22(ok693), glo-4(ok623), dhs-28(hj8), daf-22(ok693);glo-4(ok623), Is[ges-1p::glo-1::gfp](hjIs9), dhs-28(hj8);Is[ges-1p::glo-1::gfp](hjIs9), glo-4(ok623);Is[ges-1p::glo-1::gfp](hjIs9).

Bacteria culture and seeding of 6-cm and 10-cm NGM plates were the same as previously described 21 .

BODIPY and Nile Red vital staining in live animals was essentially the same as previously described 21 . 100 μL 5 μM green BODIPY (#D-3823, Invitrogen) or red BODIPY (#D-3835, Invitrogen) in 1X phosphate buffer saline (PBS, pH 7.2) was added onto the ~2 cm-in-diameter OP50 bacteria lawn of a 6-cm NGM plate. To vital-label animals in large quantity, 1 mL 12.5 μM green or red BODIPY was added onto the 10 cm-in-diameter OP50 bacteria lawn of a 10-cm NGM plate. The plates were immediately dried in laminar flow hood and were used to grow and stain animals. For Nile Red staining, 100 μL 5 μg/mL Nile Red (#N-1142, Invitrogen) in PBS was added onto the ~2 cm-in-diameter OP50 bacteria lawn of a 6-cm NGM plate. 1 mL 12.5 μg/mL Nile Red in PBS was added onto the 10 cm-in-diameter OP50 bacteria lawn of a 10-cm NGM plate. The Nile Red plates were immediately dried in laminar flow hood and were equilibrated at room temperature for 24 hr before use.

Channel mode confocal images of BODIPY, Nile Red, and GFP fluorescence in late stage L4 animals were acquired essentially the same as previously described 21 . To compare BODIPY and Nile Red fluorescence of wild-type, glo-4, and daf-22;glo-4 animals, we used imaging parameters that detected fluorescence of wild-type animals within the linear range. In order to carry out emission spectrum analyses of vital Nile Red fluorescence, lambda mode confocal imaging and image processing by linear un-mixing were conducted 34 . A Zeiss LSM 510 inverted confocal microscope equipped with a META detector was used. Images were acquired with a C-Apochromat 40X/1.20 W Korr UV-VIS-IR M27 objective. Key parameters were: excitation laser DPSS 561 nm, 15 mW, 15%; emission spectrum 577-738 nm at 10 nm per channel; pinhole, 90 μm; optical zoom, 2X; line scan, mean of 4; gain and offset were tailored to individual images so that no pixel was saturated in each channel. Lambda mode images were acquired for non-stained wild-type and dhs-28 animals as negative controls. No autofluorescence was detected. Two types of spectra were identified for Nile Red-stained structures in dhs-28 intestinal cells in lambda mode images. The spectrum with peak shifted towards a shorter wavelength was classified as "Spec-2", while the other, "Spec-1". In wild-type animals, only Spec-1 was identified. The two spectral profiles of dhs-28 animals were used to linearly un-mix lambda mode images of wild-type, dhs-28, dhs-28;Is[ges-1p::glo-1::gfp], and daf-22;glo-4 animals with the same parameter settings. Spec-1s were pseudo-colored red. Spec-2s, green. No apparent co-localization of red and green structures and minimal un-assigned pixels indicated reliable spectral classification and un-mixing. Lambda imaging of red BODIPY fluorescence in dhs-28;Is[ges-1p::glo-1::gfp] animals and in wild-type animals were essentially the same as that of Nile Red fluorescence except that the emission spectrum spanned 566-738 nm at 10 nm per channel. Only one emission spectrum was identified in Red BODIPY-stained dhs-28 and wild-type animals. For each imaging experiment, 10-30 animals were imaged.

Late L4 stage wild-type and glo-4 animals were fixed and stained essentially as previously described 20 . In brief, late L4 animals were fixed in 1% formaldehyde/PBS for 10 minutes. Then samples were immediately frozen on dry ice/ethanol and thawed on running tap water. Samples were frozen and thawed again, 3 times in total. Samples were washed 3 times with 1X PBS, dehydrated in 60% isopropanol for 2 minutes, stained with 0.5 mL 1 μg/mL Nile Red in 60% isopropanol for 30 minutes. Stained samples were then washed 3 times with 1X PBS, re-hydrated in 1X PBS, and mounted in 1X PBS onto agarose-padded slides for fluorescence imaging. Channel mode imaging and lambda imaging were conducted the same as described in previous section.

To isolate BODIPY-labelled lipid droplets, 4,000 freshly hatched wild-type or dhs-28 animals were loaded onto one 10-cm NGM/OP50 plate supplemented with green or red BODIPY. Animals were allowed to grow to 1-day-adult stage, i.e., 24 hr past L4 stage. Animals were harvested and washed 4 times with 1X PBS/0.001% Triton X-100. Worm pellet was re-suspended in 1X PBS added with protease inhibitors on ice. Worm samples were then homogenized by 20 strokes using a dounce homogenizer with a tight-fit pestle. Homogenates were transferred in 10 mL 1X PBS/protease inhibitors into a falcon tube. Homogenates were centrifuged for 10 minutes at 1000 g at 4°C. 1 mL top layer was aspirated and transferred into a new tube and washed by adding 9 mL 1X PBS/protease inhibitors. The tube was centrifuged in the same way. The 1 mL top layer was collected and washed with 9 mL 1X PBS/protease inhibitors, and centrifuged again. The final 1 mL top layer was collected for further analysis. To image isolated lipid droplets, the final top layer fraction was mounted onto a cytometer and imaged on a Zeiss Observer.Z1 inverted microscope with a 40X/NA1.20 W objective. Epi-fluorescence imaging parameters were set so that the BODIPY fluorescence of lipid droplets isolated from dhs-28 animals was detected within the linear range. Lipid droplets from wild-type animals were imaged with the same parameters.

To isolate Nile Red-labelled lipid droplets from dhs-28 animals, 4,000 freshly hatched L1s were loaded onto each of three 10-cm NGM/OP50/Nile Red plates. The 12,000 animals were allowed to grow to late stage L4 and were then harvested. Lipid droplets were isolated in the same way as for BODIPY-labelled animals. The post-nuclear cytoplasmic fraction generated during the first centrifugation step was further spun at 100,000 g for 30 min at 4°C. The membrane pellet was re-suspended in 1 mL 1X PBS/protease inhibitors. Lipid droplet fraction and cytoplasmic membrane pellet fraction were mounted onto a cytometer for confocal imaging. Lambda mode confocal imaging was conducted for the two fractions in the same way as described in "Confocal Microscopy and Spectral Analysis" section. Experiments were repeated three times. More than 30 lipid droplets and 30 membrane pellets were imaged.

To prepare lipid droplet fraction and cytoplasm fraction for western blot experiment, 4000 freshly hatched dhs-28;Is[ges-1p::glo-1::gfp] L1s were loaded onto a 10-cm NGM/OP50 plate. Animals were harvested at late L4 stage. One plate of animals was treated as one sample. The sample was subjected to lipid droplet isolation with the same procedure, except that the extraction volume was scaled down from 10 mL to 1 mL and the lipid droplet fraction volume was scaled down from 1 mL to 0.1 mL. The 0.1 mL final lipid droplet fraction and the 0.9 mL post-nuclear cytoplasm fraction generated during the first centrifugation step were concentrated and extracted for protein in an equal final volume of 60 μL. 7.5 μL of each protein sample was loaded onto a SDS-PAGE gel and was subjected to western blot using a custom made rabbit anti-GFP antibody (Y2769). Wild-type lipid droplet fraction and cytoplasm fraction were prepared in the same way and were loaded alongside as a negative control. Experiments were conducted two times. Three samples each time.

Electron microscopy was conducted the same as previously described 21 .