2.1.

Drosophila

$w^{1118}$ strain was provided by fly core facility in College of Medicine, National Taiwan University. FB-Gal4, UAS-Bmm, and UAS-Lsd-2 strains were kindly provided by Dr. Ronald Kühnlein.^12,16 To overexpress Bmm or Lsd-2 in the fat body, we crossed FB-Gal4 with UAS-Bmm or UAS-Lsd-2 to generate FB-Gal4:UAS-Bmm (FB-Bmm-overexpressing mutant) and FB-Gal4:UAS-Lsd-2 (FB-Lsd-2-overexpressing mutant), respectively. The $w^{1118}$ and FB-Gal4 strains were used as the controls in oenocyte and fat body experiments, respectively. All strains were raised at 22°C using a standard agar diet (1% autolyzed yeast, 5.8% cornmeal, 5% glucose, 0.6% agar).

2.2.

Oil Red O Staining

The procedure for staining Drosophila larval oenocytes and fat body is described as follows: larvae were carefully dissected in phosphate buffered saline (PBS) and fixed in 4% paraformaldehyde for 25 (oenocyte) or 10 min (fat body). Specimens were then rinsed with distilled water and incubated in Oil Red O stain for 20 min. (Saturated Oil Red O in $\text{isopropanol}/\text{water}=1:1\text{solution}$. Oil Red O was from Sigma-Aldrich, UK.) After staining, the specimens were rinsed with distilled water and ready for imaging by bright field microscopy incorporated with a CCD camera.

2.3.

Drosophila Larva Analysis

In the oenocyte experiment, the L3-larvae were chosen and placed on the moistened tissue paper for 0 (fed) or 16 h (starved). In the fat body experiment, the L2-larvae were chosen and placed on the moistened tissue paper for 0 (fed) or 48 h (starved). In the starvation assay, the L2-larvae were chosen and placed on the moistened tissue paper for 0 to 120 h. Their viability was determined by the pharynx movement every 12 h. Before in vivo observation, larvae were anesthetized by exposure to ether fume for 2 to 3 min. Temperature was kept at 22°C during observation.

2.4.

CARS/TPE-F Microscopy

In the setup, one Ti-sapphire laser (Mira-900P, Coherent, California) with 2.5-ps pulse width serves as pump/probe beam, and the other Ti-sapphire laser (Mira-900F) with 200-fs pulse width serves as Stokes beam in CARS, both are at 76 MHz repetition rate (Fig. 1). Here the 200-fs pulsed Ti-sapphire laser can provide better efficiency of two-photon excitation of autofluorescence and green fluorescent protein (GFP). Both lasers are pumped by a Nd:YVO4 laser (Verdi-10W, Coherent, Santa Clara, California). An additional synchronization system (Synchro-Lock, Coherent, Santa Clara, California) is applied to synchronize the two laser pulses. The laser beams are collinearly combined and directed into a laser scanning microscope (FV300 and IX-71, Olympus, Tokyo, Japan), and focused onto the sample by a 60X $\mathrm{N}.\mathrm{A}.=1.2$ water immersion objective (UPLSAPO 60XW, Olympus, Tokyo, Japan). To obtain CARS imaging of lipid C-H stretching mode at $\sim 2845{\mathrm{cm}}^{-1}$, the pump/probe beam is tuned at $\sim 710\mathrm{nm}$, and the Stocks beam is tuned at $\sim 890\mathrm{nm}$, with 30 and 15 mW at the sample, respectively. The forward CARS (F-CARS) signal (at $\sim 590\mathrm{nm}$) is collected by a condenser ($\mathrm{N}.\mathrm{A}.=0.55$), passing through a set of bandpass filters (FF01-630/92 and FF01-590/10, Semrock, Rochester, New York) and detected by a PMT (R7400U-02, Hamamatsu, Iwata-City, Japan). The TPE-F images are obtained simultaneously with CARS images. The fluorescence is collected by the same objective (epi-detection), passing through a short-pass dichroic mirror (685SPXR, Chroma, Bellows Falls, Vermont) and a bandpass filter (FF01-510/84, Semrock, New York), detected by a PMT (R3896, Hamamatsu, Japan). The scan speed is about 3 to 5 s for each CARS/TPE-F image. These images are shown as the overlay of CARS (pseudo-colored in yellowish orange) and TPE-F (pseudo-colored in green) images.

Fig. 1

Schematic Setup of CARS/TPE-F Microscope. F-ISO: Faraday isolator, BS: beam splitter, HWP: half-wave plate, P: polarizer, GS: galvo scanner, DM: dichroic mirror, BF: bandpass filter, C: condensor, Obj: objective.

2.5.

Image Analysis of Lipid Content in Oenocytes and Fat Body

For oenocytes, CARS/TPE-F images were focused on the plane, which showed clear contours of cell nuclei. LDs were identified from the thresholded CARS image and the CARS intensity of LDs was subtracted by the background, which is defined by nonlipid surrounding regions. The lipid content was determined by the square root of the background-subtracted CARS intensity,³³ and then multiplied by LDs area and divided by oenocytes area to give the normalized CARS intensity. On the other hand, for the fat body, we acquired images from several focal planes and estimated the lipid content by averaging the ratio of LDs area to the fat body area of these images. The area of the oenocyte and fat body was determined by the segmented TPE-F image. All image analyses were performed by Image-J. Statistical significance was calculated by using an unpaired two-tailed Student’s t test. Error bars represent standard deviations.

3.1.

In Vivo Imaging of Drosophila Larvae by CARS/TPE-F Microscopy

The larval oenocytes, clusters of pear-shaped cells, are found in the space between the dorsoventral muscles and the body wall.^45,46 Freshly dissected oenocytes often show greenish-yellow autofluorescence upon excitation with ultraviolet light, but it quenches rapidly.⁴⁵ Using CARS/TPE-F microscopy to observe the third-instar larva, we can recognize oenocytes by their cell morphology and autofluorescence [Fig. 2(b)]. The oenocytes are located at the basal surface of the lateral epidermis (at $\sim 5\mu \mathrm{m}$ depth from the cuticle), in the presence of repetitive clusters on the side of the larval body [Fig. 2(a) and 2(b)]. CARS was tuned at $\sim 2845{\mathrm{cm}}^{-1}$ corresponding with C-H bond stretching mode for lipid imaging. Figure 2(b) shows the CARS signal in oenocytes, indicating lipid accumulation. When we focus deeper at $\sim 20\mu \mathrm{m}$ depth from the cuticle, we can also find the fat body and the gut, as shown in Fig. 2(c). The fat body has thin lobes composed of fat cells with large LDs and is clearly resolved by its strong CARS signal [Fig. 2(c)]. Some part of the gut shows autofluorescence (perhaps from the food) and has small LDs on it [Fig. 2(c)]. The larval cuticle also has strong autofluorescence that gives us clear contour of the larval body in CARS/TPE-F images [Fig. 2(b) and 2(c)]. This demonstrates that CARS/TPE-F microscopy can be used to visualize lipid distribution and autofluorescent organs in living Drosophila larva, with good axial and spatial resolution at subcellular level. This approach enables us to observe and monitor lipid dynamics in a noninvasive way.

Fig. 2

The internal organs of a living Drosophila larva: (a) a schematic drawing of internal organs in a larva; (b) the CARS/TPE-F images of oenocytes in a larva, at $\text{depth}=5\mu \mathrm{m}$ from the cuticle; (c) the CARS/TPE-F images of the fat body and gut in a larva, at $\text{depth}=20\mu \mathrm{m}$ from the cuticle. Scale bar: 30 μm. (Yellowish orange: CARS, green: autofluorescence).

To visualize the lipid in larval oenocytes and the fat body, we compared conventional bright-field images of Oil Red O stained tissues [Fig. 3(a) and 3(c)] with CARS/TPE-F images of a living larva [Fig. 3(b) and 3(d)]. When the larva is starved, its oenocytes will accumulate lipid-like liver steatosis in mammals,⁷ shown as many Oil-Red-O-positive LDs in Fig. 3(a). In the in vivo CARS/TPE-F image [Fig. 3(b)], oenocytes are identified by their autofluorescence from the TPE-F signal (pseudo-colored in green) and accumulated LDs are also clearly resolved from the CARS signal (pseudo-colored in yellowish orange) [Fig. 3(b)]. Moreover, in vivo CARS/TPE-F image shows a better feature of each lipid droplet (LD) [Fig. 3(b)], while the image of Oil Red O stained tissue shows LDs as reddish fuzzy spots [Fig. 3(a)] due to possible aggregation caused by fixation and staining procedures.^47,48 In addition, from in vivo CARS/TPE-F image, we not only observe autofluorescence of oenocytes but also find some fluorescent globules in the cytoplasm (pseudo-colored in green) [Fig. 3(b)] which could be the protein granules,⁴⁹ providing us more information about its metabolic status. Likewise, compared with the image of Oil Red O stained tissue [Fig. 3(c)], in vivo CARS/TPE-F image shows better resolution of LDs and gives us the fine structure of the fat body [Fig. 3(d)]. Using fluorescent fusion proteins such as GFP can also achieve in vivo imaging of LDs in Drosophila fat body.²¹ However, the fluorescent protein is labeled on a specific protein (such as PERILIPIN1 or Lsd-1) that expresses on the surface of LDs, and therefore its fluorescence cannot truly reflect the lipid content in LDs. Unlike fluorescent protein labeling, the CARS signal can be specific to C-H bonds mainly from neutral lipid in LDs. In conclusion, in vivo imaging via CARS/TPE-F microscopy can preserve the most natural information including the feature of LDs, the lipid content, and the autofluorescence.

Fig. 3

Larval oenocytes and the fat body: (a) Oil Red O stained bright field and (b) in vivo CARS/TPE-F images of oenocytes. The lipid droplet and fluorescent globule are indicated by the white arrow and yellow arrow, respectively. Scale bar: 20 μm. (c) Oil Red O stained bright field and (d) in vivo CARS/TPE-F images of the fat body. Scale bar: 30 μm. (Yellowish orange: CARS, green: autofluorescence).

3.2.

Lipid Regulation in Larval Oenocytes

To investigate lipid regulation in larval oenocytes and fat body, each of two lipid regulatory proteins—Bmm or Lsd-2—was overexpressed in the fat body of Drosophila to observe their effects. The imbalance between these two proteins can alter lipid metabolism in Drosophila. Figure 4(a) through 4(f) are CARS/TPE-F images of oenocytes under fed and starved (16 h) conditions. In the control larva ($w^{1118}$), oenocytes normally do not accumulate much lipid under fed condition [Fig. 4(a)]. When it is starved, lipolysis (TAG hydrolysis) occurs in the fat body and the fatty acids are released from the fat body to oenocytes. As a result, more LDs accumulate in oenocytes ($N=13$ larvae) [Fig. 4(b)].⁷ However, in FB-Bmm-overexpressing mutant (FB-Gal4:UAS-Bmm), LDs accumulate in oenocytes even when it is fed [Fig. 4(c)]⁷ and accumulate more when starved ($N=8$ larvae) [Fig. 4(d)]. On the contrary, in FB-Lsd-2-overexpressing mutant (FB-Gal4:UAS-Lsd-2), LDs are less found in oenocytes both under fed and starved conditions ($N=12$ larvae) [Fig. 4(e) and 4(f)].⁷ The image analysis of lipid content in oenocytes is shown in Fig. 4(g). FB-Bmm-overexpressing mutant has $\sim \text{twofold}$ more lipid than wild-type under fed condition, while FB-Lsd-2-overexpressing mutant has $\sim \text{fourfold}$ less lipid than wild-type under starved condition [Fig. 4(g)]. Grönke et al. have found that Bmm would be upregulated upon starvation in normal flies.¹⁶ Thus, overexpression of Bmm in the fat body is likely to mimic the effect of starvation and lead to more LDs accumulation in oenocytes. On the other hand, Lsd-2 acts antagonistically to Bmm and hinders the access of Bmm to LDs,^9,16 thus can lower the lipolysis rate when overexpressed. Indeed, we found less accumulation of LDs in FB-Lsd-2-overexpressing mutant oenocytes even when starved. These results indicate that promotion and inhibition of TAG hydrolysis in the fat body can significantly perturb lipid metabolism in the larva under fed and starved condition.

Fig. 4

The CARS/TPE-F images of oenocytes under fed and starved conditions, in the control ($w^{1118}$) [(a) and (b)], Bmm overexpression under the control of FB-Gal4 (Bmm) [(c) and (d)], and Lsd-2 overexpression under the control of FB-Gal4 (Lsd-2) [(e) and (f)]. Scale bar: 20 μm. (Yellowish orange: CARS, green: autofluorescence). (g) Analysis of lipid content in oenocytes under fed and starved conditions. (Error bars represent standard deviations of oenocytes from 8 to 13 larvae. $***p<0.0001$.)

3.3.

Lipid Regulation in the Fat Body

We further investigated the effect of overexpression of Bmm and Lsd-2 on lipid regulation in larval fat body. Figure 5 shows CARS/TPE-F images of the fat body of FB-Gal4 [Fig. 5(a) and 5(b)] (as control), FB-Bmm-overexpressing mutant [Fig. 5(c) and 5(d)], and FB-Lsd-2-overexpressing mutant [Fig. 5(e) and 5(f)] under fed and starved (48 h) conditions. Because these mutants are generated by the FB-Gal4 driver which expresses GFP in fat cells,¹² both mutants show GFP fluorescence in their fat bodies. Therefore we can identify the LDs by CARS signal (pseudo-colored in yellowish orange) in the region of the fat body by TPE-F signal (pseudo-colored in green). Under fed condition, the average size of LDs [Fig. 5(g)] in the fat body of FB-Bmm-overexpressing mutant is much smaller ($\sim 4.5\mu \mathrm{m}$ in diameter) ($N=22$ LDs) [Fig. 5(c)] than that of the control larva ($\sim 6.1\mu \mathrm{m}$ in diameter) ($N=17$ LDs) [Fig. 5(a)], whereas the size of LDs in the fat body of FB-Lsd-2-overexpressing mutant is slightly larger ($\sim 6.4\mu \mathrm{m}$ in diameter) ($N=18$ LDs) [Fig. 5(e)]. Grönke et al. have found that overexpression of Bmm can reduce the size and number of LDs in fat cells of adult flies,¹⁶ and overexpression of Lsd-2 can increase total TAG content of adult flies.¹² Our results demonstrate that they can have different sizes of LDs at larval stage, especially in FB-Bmm-overexpressing mutant. The reduced size of LDs in FB-Bmm-overexpressing mutant larval fat body implies that promotion of lipolysis decreases the formation of large LDs. We further analyzed the lipid content in their fat bodies. Unlike oenocytes, which locate near the surface (at $\sim 5\mu \mathrm{m}$ depth), the fat body distributes more deeply (at $\sim 20\mu \mathrm{m}$ depth) in the larval body, and the intensity of its CARS signal decays with the depth. To avoid the problem it may cause in lipid quantification, we used the density of LDs in the fat body (the ratio of LDs area to the fat body area) instead of the CARS intensity to estimate the lipid content [Fig. 5(h)]. We found that FB-Bmm-overexpressing mutant has lower lipid content under fed condition, and the lipid content dramatically decreases after 48 h starvation ($N=10$ larvae) [Fig. 5(d)]. On the contrary, the lipid content in FB-Lsd-2-overexpressing mutant decreases less than in the control after starvation ($N=8$ larvae) [Fig. 5(f)]. It was reported that LDs can aggregate under starvation.⁵⁰ However, we did not observe discernible LDs aggregation in the fat body [Fig. 5(b) and 5(f)]. Our results demonstrate that overexpression of Bmm can cause smaller LDs and lower lipid content in the larval fat body and faster consumption of lipid under starvation, while overexpression of Lsd-2 results in slower consumption of lipid under starvation.

Fig. 5

The CARS/TPE-F images of the fat body under fed and starved conditions, in the control (FB-Gal4) [(a) and (b)], Bmm overexpression under the control of FB-Gal4 (Bmm) [(c) and (d)], and Lsd-2 overexpression under the control of FB-Gal4 (Lsd-2) [(e) and (f)]. Scale bar: 30 μm. (Yellowish orange: CARS, green: GFP) (g) Average LD size in the fat body. (Error bars represent standard deviations of 17 to 22 LDs for each mutant. $***p<0.0001$.) (h) The ratio of LDs area to the fat body area of each larva under fed and starved conditions. (Error bars represent standard deviations of fat bodies from $8\sim 11$ larvae. $*p<0.01$, $***p<0.0001$.)

To evaluate the correlation between the lipid metabolism and the viability under long-term food-deprived condition, we performed the starvation assay (0 to 120 h) on L2-larvae of $w^{1118}$, FB-Gal4, FB-Bmm-overexpressing mutant, and FB-Lsd-2-overexpressing mutant to measure their lifespans under starvation, as shown in Fig. 6. The mean lifespan of FB-Bmm-overexpressing mutant is $57.9\pm 2.9\mathrm{h}$ ($N=40$), while that of FB-Lsd-2-overexpressing mutant is $87.3\pm 3.4\mathrm{h}$ ($N=40$) and that of the control (FB-Gal4) is $68.7\pm 3.6\mathrm{h}$ ($N=40$) (Fig. 6). FB-Bmm-overexpressing mutant has a shorter lifespan whereas FB-Lsd-2-overexpressing mutant has a longer lifespan with respect to the control. In summary, Bmm and Lsd-2 regulate lipid in an opposite way, and overexpression of these proteins can result in different sizes of LDs and rate of lipid consumption in the larval fat body, which can strongly affect their viability under starvation.

Fig. 6

Life span of $w^{1118}$, FB-Gal4, FB-Bmm-overexpressing mutant (Bmm), and FB-Lsd-2-overexpressing mutant (Lsd-2) under starvation. (Error bars represent standard deviations from total 40 larvae in four independent experiments.)