Label-free quantitative analysis of lipid metabolism in living Caenorhabditis elegans.

The ubiquity of lipids in biological structures and functions suggests that lipid metabolisms are highly regulated. However, current invasive techniques for lipid studies prevent characterization of the dynamic interactions between various lipid metabolism pathways. Here, we describe a noninvasive approach to study lipid metabolisms using a multifunctional coherent anti-Stokes Raman scattering (CARS) microscope. Using living Caenorhabditis elegans as a model organism, we report label-free visualization of coexisting neutral and autofluorescent lipid species. We find that the relative expression level of neutral and autofluorescent lipid species can be used to assay the genotype-phenotype relationship of mutant C. elegans with deletions in the genes encoding lipid synthesis transcription factors, LDL receptors, transforming growth factor beta receptors, lipid desaturation enzymes, and antioxidant enzymes. Furthermore, by coupling CARS with fingerprint confocal Raman analysis, we analyze the unsaturation level of lipids in wild-type and mutant C. elegans. Our study shows that complex genotype-phenotype relationships between lipid storage, peroxidation, and desaturation can be rapidly and quantitatively analyzed in a single living C. elegans.


Biochemical assays
Experimental details of gas chromatography and mass spectrometry analysis of fatty acids composition have been previously described ( 10) . Cholesterol, phosphatidylcholine, and superoxide dismutase were performed using commercially available kits according to manufacturer's protocols (catalog numbers 10007640, 10009926, and 706002; Cayman Chemical, Ann Arbor, MI). To collect worms, 16 cm plates were washed with M9 media into Eppendorf tubes. Worms were further washed multiple times with M9 media to remove any residual bacteria. The expression level of phosphatidylcholine was used as a reference to adjust for the number of C. elegans analyzed.

RESULTS
By simultaneous CARS and TPEF imaging of living C. elegans , we discovered two distinctive lipid species ( Fig. 1 ; see supplementary Video I). One lipid species can be visualized by CARS only, which we identify as neutral lipid ( Fig. 1A ). Visualization of such neutral lipid species with CARS has been previously reported by Hellerer et al. ( 4) . The other lipid species can be visualized by both autofl uorescent signals identifi able with TPEF imaging and lipid signals identifi able with CARS imaging ( Fig. 1B, C ). Further confocal Raman spectral analyses of the neutral lipid droplets reveal strong chemical signatures typical of tria-

A multifunctional CARS microscope
A spectrometer with a 300 g/mm grating and a thermoelectrically cooled back-illuminated EMCCD (Newton 920-BRD; Andor Technology, Belfast, Ireland) is mounted to the side port of a laser scanning microscope (IX71/FV300; Olympus, Center Valley, PA) to allow TPEF imaging, CARS imaging, and spontaneous Raman spectral analysis on the same platform. Pump and Stokes lasers are tuned to 14,140 cm Ϫ 1 and 11,300 cm Ϫ 1 , respectively, to be in resonance with the CH 2 symmetric stretch vibration at 2,840 cm Ϫ 1 . The combined beams are focused into the sample through a ×60 water immersion microscope objective with a 1.2 numerical aperture. Forward-detected CARS signal is collected by an air condenser with a 0.55 numerical aperture, transmitted through a 600/65 nm band-pass fi lter, and detected by a photomultiplier tube (H7422-40; Hamamatsu, Japan). Simultaneously, backrefl ected TPEF signal is collected by the same illuminating objective, spectrally separated from the excitation source, transmitted through a 520/40 nm band-pass fi lter, and detected by a photomultiplier tube (H7422-40; Hamamatsu) mounted at the back port of the microscope. Following CARS and TPEF imaging, the Stokes beam is blocked and the pump laser-induced Raman scattering signal is directed toward the spectrometer to permit spectral analysis from 830 to 3,100 cm Ϫ 1 , which covers both the fi ngerprint and the CH-stretch vibration regions. Due to the utilization of a confocal pinhole with a diameter of 50 µm before the spectrometer, the Raman signals arise from a focal volume of 2.3 femtoliter at the center of the fi eld of view of a CARS image. Average acquisition time for a 512 × 512 pixel CARS image is 1.12 s, and a full-spectral Raman analysis is 4 s. The combined Stokes and pump laser power is kept constantly at 40 mW. For all Raman spectral measurements, pump laser power is reduced to 10 mW. Variability in Raman spectral measurements of neutral lipid droplets is discussed in supplementary Figs. II and III . Experimental details of Rama spectra acquisition and data analysis were previously described by Slipchenko et al. ( 9).

Imaging conditions and data analysis
All C. elegans were anesthetized in a droplet of 100 mM sodium azide and mounted on fresh 2% agarose slides prior to imaging. To evaluate the expression level of neutral and autofl uorescent lipid droplets, we defi ne a probed volume with xyz dimensions of 125 × 125 × 35 µm. We fi rst locate the midsection of an adult wildtype or mutant C. elegans and then perform simultaneous depth imaging with CARS and TPEF along the vertical (z) axis at 1 µm step size to obtain 36 frames. Background CARS and TPEF pixel intensity are subtracted from average pixel intensity of CARS and TPEF signals of the probed volume to obtain the expression level of neutral and autofl uorescent lipid droplets, respectively. Background CARS and TPEF pixel intensity are defi ned as the average pixel intensity of probed volumes devoid of neutral and autofl uorescent lipid droplets. The background CARS signal includes signals arising from the worm bodies. The expression levels of neutral and autofl uorescent lipid droplets are adjusted to 1 for the wild type and comparatively for mutant C. elegans . Because CARS signal is quadratically dependent on the concentration of CH 2 molecular vibration at p Ϫ S = 2,840 cm has also been reported by Hellerer et al. ( 4) . However, the effect of daf-4 deletion on the decrease of autofl uorescent lipid droplet level is reported here for the fi rst time. Finally, the fi fth phenotype is observed with rme-2 mutants, where a 1.3-fold increase in the expression of neutral lipid droplets is accompanied by a wild-type level expression of autofl uorescent lipid droplets ( Fig. 2A-C ). This phenotype suggests a possible role of RME-2 in neutral lipid droplet formation. However, this relationship has not been elucidated in the current literature. Taken together, our results show that the relative expression levels of neutral and autofl uorescent lipid species provide a reliable means to assay for phenotypes of C. elegans mutants.
Using biochemical assays, we fi nd that changes in the level of autofl uorescent lipid species in fat-5/fat-6 , sbp-1 , and sod-1 mutants are due to distinct mechanisms ( Fig.  2D ). By measuring the cholesterol level in total lipid extracts, we fi nd a 1.2-fold increase in fat-5/fat-6 mutants, a 4-fold reduction in sbp-1 mutants, and no change in sod-1 mutants. These observations suggest that ⌬ 9 desaturases and SREBP-1 are involved in cholesterol biosynthesis, whereas superoxide dismutase is not. Therefore, increases in autofl uorescent lipid species observed in fat-5/fat-6 and sbp-1 mutants are likely due to a different mechanism compared to sod-1 mutants ( Fig. 2D ).
By CARS imaging and spontaneous Raman analysis of single lipid droplets, we further evaluated the degree of lipid-chain unsaturation in wide-type and mutant C. elegans . Using C18 fatty acid methyl esters as standards, we show that the degree of lipid-chain unsaturation can be measured using three Raman-active bands, including 1280 cm Ϫ 1 , 1660 cm Ϫ 1 , and 3015 cm Ϫ 1 ( Fig. 3A-C ) ( 6,21) . Because the signal-to-noise ratio for C=C stretch is highest at 1660 cm Ϫ 1 band, we select this band to evaluate lipid chain unsaturation ( 6) . We observe that I 1660 /I 1445 is linearly correlated with lipid chain unsaturation ( Fig. 3C ). Using I 1660 /I 1445 as a reliable measure of ⌬ 9 desaturase enzymatic activity, we systematically evaluated lipid chain unsaturation of neutral lipid droplets. We observed signifi cant reduction in C=C stretch vibration signal in ⌬ 9 desaturase mutants compared to wild-type C. elegans ( Fig. 3D ). Quantitative analysis of lipid droplet I 1660 /I 1445 in six desaturase mutants reveals up to 2-fold reduction in lipid chain unsaturation in single and double ⌬ 9 desaturase mutants ( Fig. 3E ). Our Raman spectral analyses are further supported by GC-MS measurements of lipid chain unsaturation of total lipid extracts (see supplementary Fig. I ). We fi nd a dramatic decrease in the ratios of unsaturated oleic, linoleic, and eicosenoic fatty acids over saturated stearic acid in fat-5/fat-6 and fat-6/fat-7 mutants compared to the wild type ( Fig. 3F ). Complete analyses of lipid composition of ⌬ 9 desaturase mutants using GC-MS have been described previously ( 10) . However, unlike GC-MS, the combination of CARS imaging and confocal Raman spectral analysis, so called compound Raman microscopy ( 6) , enabled us to measure lipid chain desaturation noninvasively with single lipid droplet sensitivity. This capability should allow real-time dynamic studies of the activity of desaturases and other lipid metabolism enzymes in living C. elegans . cylglycerides ( Fig. 1D ). In contrast, the fl uorescence from the autofl uorescent lipid droplets dominates the Raman spectra ( Fig. 1D ). Several previous studies have also identifi ed such autofl uorescent particles and associated them with lipids, oxidative stress, and lifespan of C. elegans ( 11,12) .
To explore the potential of using the neutral and autofl uorescent lipid species as a readout of lipid metabolism, we evaluated their expression levels in wild-type and mutant C. elegans . All selected C. elegans mutants have been well characterized, with deletions in the genes encoding lipid metabolism proteins, including ⌬ 9 desaturases (palmitoyl-CoA desaturase fat-5 and stearoyl-CoA desaturases fat-6 and fat-7 ) ( 10) , sterol regulatory element binding protein ( sbp-1 ) ( 13,14) , copper/zinc superoxide dismutase ( sod-1 ) ( 15, 16) , type II transforming growth factor ␤ receptor ( daf-4 ) ( 4, 17) , and LDL receptor ( rme-2 ) 18 . Specifi cally, ⌬ 9 desaturases are lipogenic enzymes critical for the conversion of saturated fatty acids into monounsaturated fatty acids ( 19,20) . Sterol regulatory element binding protein 1 (SREBP-1) is a transcription factor that controls the expression of lipogenic enzymes ( 14) . Superoxide dismutase is an antioxidant enzyme that protects cells from reactive oxygen species ( 15) . Type II transforming growth factor ␤ receptor is a transmembrane serine/threonine kinase whose functions are implicated in many biological processes, including the insulin signaling pathway ( 17) . Receptor-mediated endocytosis 2 (RME-2) is an LDL receptor that mediates yolk endocytosis and fatty acid transport in oocytes ( 18) . Although the functions of these lipid metabolism proteins are well understood, their roles in the expression level of autofl uorescent lipid species in C. elegans have not been characterized.
Our analyses of neutral and autofl uorescent lipid droplet expression level in C. elegans mutants reveal fi ve distinctive phenotypes as compared to the wild type ( Fig. 2A ). The fi rst phenotype is observed in C. elegans mutants with double deletion of ⌬ 9 desaturases, fat-5/fat-6 , where a 1.4fold decrease in the expression level of neutral lipid droplets is accompanied by a 3-fold increase in the expression level of autofl uorescent lipid droplets ( Fig. 2A-C ). This phenotype is consistent with previous biochemical analyses where a decrease in fat storage and an increase in the expression of genes involved in fatty acid oxidation are observed in ⌬ 9 desaturase mutants ( 10 ). The second phenotype is observed in sbp-1 mutants, where there is a near complete suppression of both neutral and autofl uorescent lipid droplet expression ( Fig. 2A-C ). This phenotype is also supported by the established roles of SREBP-1 in cholesterol and fatty acids homeostasis ( 14) . The third phenotype is observed with sod-1 mutants, where the wild-type level of neutral lipid droplet expression is accompanied by a 2-fold increase in autofl uorescent lipid droplet expression ( Fig. 2A-C ). This phenotype suggests a direct role of antioxidant enzymes in regulating the level of autofl uorescent lipid droplets. The fourth phenotype is observed with daf-4 mutants, where a 1.4-fold increase in neutral lipid droplet is accompanied by a 2-fold reduction in autofl uorescent lipid droplet compared to the wild type ( Fig. 2A-C ). This increase in neutral lipid storage in daf-4 mutants cent lipid droplet formation. These observations suggest that SREBP-1, ⌬ 9 desaturases, and transforming growth factor ␤ receptor participate in shared pathways by both neutral and autofl uorescent lipid droplet formation. In contrast, deletion of antioxidant enzymes affects only autofl uorescent lipid droplet formation, and deletion of LDL receptor RME-2 affects only neutral lipid droplet formation. These observations suggest that antioxidant enzymes or LDL receptor RME-2 participates in specifi c autofl uorescent or neutral lipid droplet formation pathways, respectively. Thus, the relationship between the expression levels of neutral and autofl uorescent lipid species could potentially be used to identify the involvement of unknown proteins in lipid metabolism pathways.
In addition to visualization of neutral and autofl uorescent lipid species with CARS and TPEF signals, spontaneous DISCUSSION In this study, we report label-free visualization and quantitation of coexisting neutral and autofl uorescent lipid species in living C. elegans . We show that multimodal imaging allows rapid genotype-phenotype screening of lipid metabolism in C. elegans . Specifi cally, we fi nd that the expression of neutral and autofl uorescent lipid species are dynamically correlated to specifi c genes. Deletion of SREBP-1 transcription factor for lipid and cholesterol synthesis suppresses both neutral and autofl uorescent lipid droplet formation. Deletion of ⌬ 9 desaturases represses neutral lipid-droplet formation and promotes autofl uorescent lipid droplet formation. Conversely, deletion of transforming growth factor ␤ receptor represses autofl uorescent lipid droplet formation and promotes neutral autofl uores- Lipids play a ubiquitous role in human physiology. Membrane lipids actively regulate cell proliferation, apoptosis, migration, and senescence ( 25) . Lipid-mediated endocrine networks regulate systemic metabolic homeostasis ( 26) . Excessive lipid storage in obesity is associated with increased risk factors for diabetes, cardiovascular diseases, stroke, and cancer ( 27) . Given the signifi cance of lipids in biology, lipid metabolism should be thoroughly and systematically studied. The multifunctional CARS microscope described in this article, when combined with recent advances in genetics ( 28) and high-throughput screening ( 29) for C. elegans research, should enable functional studies of lipid metabolism enzymes, interaction of lipid metabolism networks, and discovery of new lipid metabolism pathways. Because CARS microscopy has been applied for in vivo imaging (30)(31)(32) , our lipid metabolism studies in living C. elegans should be extensible to both animals and humans. The versatility of the multifunctional CARS microscope would render it an indispensible tool to the study of lipids in diseases.
The authors thank Han-Wei Wang for help with experiments.
Raman microspectroscopy enables noninvasive quantitation of desaturation in single lipid droplets. In general, lipid storage, peroxidation, and desaturation are all critical to the health of animals. Indeed, lipid peroxidation is strongly linked to the lifespan of animals ( 22) . Loss of stearoyl-CoA desaturase-1 function has been shown to reduce body adiposity, increase insulin sensitivity, and resistance to diet-induced adiposity in mice ( 19) . However, loss of stearoyl-CoA desaturase-1 function is also associated with increased aorta atherosclerosis ( 23) and infl ammation ( 24) . Nonetheless, the impacts of lipid desaturation on lipid storage, peroxidation, or infl ammation are not clearly understood. Herein, we report that genetic deletions of ⌬ 9 desaturase genes in C. elegans are strongly associated with a reduction of lipid chain unsaturation ( Fig.  3E ) and neutral lipid storage ( Fig. 2B ) as well as a significant increase in autofl uorescent lipid species ( Fig. 2C ) and cholesterol synthesis ( Fig. 2D ). Given the strong conservation in lipid metabolism from C. elegans to humans ( 1) , it is conceivable that future in-depth investigation of lipid desaturation in C. elegans could bring new insights to the roles of desaturases in human health and diseases.