Lipidome-wide C flux analysis: a novel tool to estimate the turnover of lipids in organisms and cultures

Lipid metabolism plays an important role in the regulation of cellular homeostasis. However, because it is difficult to measure the actual rates of synthesis and degradation of individual lipid species, lipid compositions are often used as a surrogate to evaluate lipid metabolism even though they provide only static snapshots of the lipodome. Here, we designed a simple method to determine the turnover rate of phospholipid and acylglycerol species based on the incorporation of 13C6-glucose combined with LC-MS/MS. We labeled adult Drosophila melanogaster with 13C6-glucose that incorporates into the entire lipidome, derived kinetic parameters from mass spectra, and studied effects of deletion of CG6718, the fly homolog of the calcium-independent phospholipase A2β, on lipid metabolism. Although 13C6-glucose gave rise to a complex pattern of 13C incorporation, we were able to identify discrete isotopomers in which 13C atoms were confined to the glycerol group. With these isotopomers, we calculated turnover rate constants, half-life times, and fluxes of the glycerol backbone of multiple lipid species. To perform these calculations, we estimated the fraction of labeled molecules in glycerol-3-phosphate, the lipid precursor, by mass isotopomer distribution analysis of the spectra of phosphatidylglycerol. When we applied this method to D. melanogaster, we found a range of lipid half-lives from 2 to 200 days, demonstrated tissue-specific fluxes of individual lipid species, and identified a novel function of CG6718 in triacylglycerol metabolism. This method provides fluxomics-type data with significant potential to improve the understanding of complex lipid regulation in a variety of research models.


INTRODUCTION
Measurements of lipid compositions provide static snapshots of the lipidome but do not capture dynamic features, such as the rates of synthesis, inter-conversion, and degradation. In order to obtain such information, stable isotope experiments are necessary (1)(2)(3). A variety of different precursors have been applied, including 2 H 3 -methionine, 2 H 13 -choline (4), 15 N-serine, 15 N-choline (5), 13 C 5 -glutamine (6), labeled glucose (7,8), labeled glycerol (9), labeled fatty acids (9)(10)(11), and 2 H 2 O (12). Most of these precursors target specific lipid classes, but some may gain access to the entire lipidome because their metabolites label a spectrum of building blocks used for lipid synthesis.
However, the latter can make data analysis very complex. Lipids are heterogeneous molecules, in which different moieties turn over independently of each other. For instance in phospholipids, fatty acids have a faster turnover than the glycerol backbone and that again is different from the turnover of the head group.
If a single isotope-labeled precursor is incorporated into multiple moieties, the isotopic label will be exposed to more than one turnover mechanism. Thus, it is not surprising that most previous studies have taken a reductionist approach by either targeting specific types of lipids, such as glycosphingolipids (13), or a specific moiety, such as phosphorylcholine (4).
Despite of that, first attempts have been made to establish a "fluxomics"-type analysis of lipid metabolism. For instance, non-targeted isotopomer filtering and matching has been applied to identify 692 new isotopomers in primary muscle cultures labeled with 13 C 16 -palmitic acid (11). Furthermore, the advent of new ultra-high resolution mass spectrometers has enabled the separation of 15 N-isotopomers from 13 C-isotopomers, which allowed the simultaneous measurement of the abundance and the labeling of many nitrogen-containing lipids (5).
4 flies are ideally suited for this type of work because (i) they do not change their body mass after reaching maturity, which eliminates growth as a source of 13 C incorporation, (ii) they are small, which limits the cost for stable isotope reagents, and (iii) they can be genetically modified, which makes them excellent tools to study the effect of various genes on lipid metabolism.

Drosophila strains and labeling with stable isotope precursors
The fly strain, y [1]  -VIA), is inactivated in these flies (14). Adult male flies of either the wild-type or the mutant were chosen for all labeling experiments. The animals were kept in vials that contained 0.2 mL of the labeling solution placed on a piece of filter paper (0.5 x 0.5 inch) as the only nutrient. The vials were placed in wet chambers at 25 C for up to 2 weeks. The labeling solutions contained 1M 13 C 6 -glucose, 1M 2 H 5 -glycerol, or 1M 2 H 9choline plus 1M unlabeled glucose in water. They were replenished at least every 4 days. Five flies were collected for each time point. In some experiments, heads, thoraces, and abdomens were dissected under the microscope. In that case, 15 flies were collected per time point. Whole flies or their body parts were stored at -80 C before analysis.

Lipidomics data acquisition by LC-MS/MS
Fly samples were homogenized in water and lipids were extracted with methanol and chloroform (15).
Internal standards were added immediately after the initial application of chloroform and methanol. They consisted of 100μl of 4-fold diluted SPLASH LipidoMIX and 50 μl of 4-fold diluted CL standard mix by guest, on March 5, 2020 www.jlr.org Downloaded from (both from Avanti Polar Lipids). The lipid extracts were dried under nitrogen and re-dissolved in 100 μl chloroform/methanol (1:1) except for the fly head samples, which were re-dissolved in 25 μl chloroform/methanol (1:1). Lipid extracts were analyzed by a QExactive HF-X mass spectrometer (ThermoScientific) coupled to an Agilent 1100 high-performance liquid chromatograph (Agilent Technologies) equipped with a Restek Ultra C18 reversed-phase column (particle size: 3μm, dimensions: 100 x 2.1 mm) operated at room temparature, using similar conditions as described by others (16). An aliquot of 7 μl of the extract was injected and chromatographed at a flow rate of 0.25 ml/min. Solvent A contained 600 ml acetonitrile, 399 ml water, 1 ml formic acid, and 0.631 g ammonium formate. Solvent B contained 900 ml 2-propanol, 99 ml acetonitrile, 1 ml formic acid, and 0.631 g ammonium formate. The chromatographic run time was 30 minutes, changing the proportion of solvent B in a non-linear gradient from 30 to 35% (0-2 minutes), from 35 to 67% (2-5 minutes), from 67 to 83% (5-8 minutes), from 83 to 91% (8-11 minutes), from 91 to 95% (11-14 minutes), from 95 to 97% (14-17 minutes), from 97 to 98% (17-20 minutes), from 98 to 100% (20-25 minutes), and from 100 to 30% (25-26 minutes). For the remainder of the run time the proportion of solvent B stayed at 30% (26-30 minutes). The mass spectrometer was operated alternatively in negative or in positive ion mode. Triacylglycerol data were extracted from the positive ion spectra and all other data were extracted from the negative ion spectra.
The spray voltage was set to 4 kV and the capillary temperature was set to 350 C. MS1 scans were acquired at a resolution of 120,000, an AGC target of 10 6 , a maximal injection time of 65 ms, and a scan range of 300-2000 m/z. MS2 scans were acquired at a resolution of 30,000, an AGC target of 3x10 6  First, mass spectral data were extracted and processed for lipid identification, lipid quantitation, and isotopomer analysis. All analyzed species and their shorthand notations are listed in Table 1. Next, kinetic parameters were estimated, including the labeled fraction of tissue glycerol-3-phosphate (p), fractional syntheses of lipid species (q), as well as their turnover rate constants (k), half-life times (t h ), and flux rates (j).

1.3.
Isotopomer analysis. In order to read the signal intensities of isotopomers, spectra were imported into Xcalibur 4.0. The intensities were averaged over a 0.1-min interval centered at the peak retention time of the species to be analyzed. For most lipids (1 glycerol group), we read the intensities of the isotopomers with 0 and 3 13 C atoms. For PG species (2 glycerol groups), we read the intensities of the isotopomers with 0, 3, and 6 13 C atoms. For CL species (3 glycerol groups), we read the intensities of the isotopomers with 0, 3, 6, and 9 13 C atoms. according to binomial distribution theory (17). Since PG contains 2 glycerol groups, its isotopomer pattern at time t equals: In equation (1), m 0 , m 1 , m 2 are the normalized intensities of the isotopomers with 0, 1, or 2 labeled glycerol groups (m 0 +m 1 +m 2 =1), p is the labeled fraction of glycerol-3-phosphate, and q is the fractional synthesis of PG. First, p and q values were calculated by equation (1). Second, the p values were substituted into equation (2) in order to calculate fractional syntheses (q) of lipids with one glycerol group: In equation (2), m 0 and m 1 are the normalized intensities of the isotopomers with 0 and 1 labeled glycerol group (m 0 +m 1 =1). Finally, p values obtained from equation (1) were substituted into equation (3) in order to calculate fractional syntheses (q) of CL species: In equation (3), m 0 , m 1 , m 2 , m 3 are the normalized intensities of the isotopomers with 0, 1, 2, or 3 labeled glycerol groups (m 0 +m 1 +m 2 +m 3 =1).

2.2.
Estimation of k, t h , and j. Turnover rate constants (k) were estimated from serial q values by non-linear regression to equation (4): Half-life times (t h ) were calculated from the turnover rate constants: Synthetic fluxes (j) were calculated by multiplying the turnover rate constant of a lipid species with its concentration (c) as determined in step 1.2.

Measurement of the isotopic labeling of water-soluble metabolites
In order to measure the 13 C abundance of water-soluble lipid precursors, samples from the same cohort analyzed by lipidomics, were processed for metabolomics. Flies (5 animals per time point) were homogenized in 100 μl methanol/water (8+2) and kept at -80 C overnight. Homogenates were spun for 5 minutes at 15,000 g in Eppendorf tubes. The supernatants were analyzed on a QExactive HF-X mass spectrometer directly coupled to a Vanquish UHPLC system (ThermoFisher Scientific, Waltham, MA, USA). A 5-μL aliquot of the supernatant was injected onto a 5 μm C18 Acclaim 120 column (4.6 x 100 mm, ThermoScientific). Metabolites were eluted with a 5-50% methanol gradient in water containing 0.1% formic acid and 0.2g/l ammonium acetate over 50 minutes at a flow rate of 1ml/min. The mass spectrometer was operated in positive ion mode and data-dependent mode with survey scans acquired at a resolution of 120,000 over a scan range of 100-1000 m/z. Up to five of the most abundant precursors from the survey scan were selected with an isolation window of 1.7Th and fragmented by higher-energy collisional dissociation with normalized collision energy of 30.

Overview of the methodology
We incubated adult Drosophila melanogaster with 13 C 6 -glucose in order to observe the incorporation of 13 C atoms into various lipid species. Flies were separated into 3 body parts, including head, thorax, and abdomen, which roughly represent the nervous system, the indirect flight muscles, and the digestive/reproductive tract, respectively. Lipids were extracted from these tissues or from whole flies and analyzed by LC-MS/MS. The mass spectral data were processed in 3 steps. First, lipids were identified by the mass-to-charge (m/z) ratios of intact molecules and their daughter ions, using the commercial software LipidSearch (ThermoScientific, version 4.1 SP1). Second, lipids were quantified by comparing their MS1 intensity to that of structurally related internal standards. Third, for each lipid species, the distribution of intensities among its isotopomers was determined in MS1 scans acquired within ±0.05 min of its peak retention time. The isotopomer patterns were transformed into normalized vectors M, which can take different forms depending on the number of glycerol groups in the lipid molecule (see Methods).
The evolution of M from the unlabeled state (M 0 ) to progressively labeled states (M t ) formed the data set from which the fraction of labeled glycerol-3-phosphate (labeling of the lipid precursor) and the fractional syntheses of lipids (newly formed molecules/total molecules) were calculated. Those data in turn were used to calculate turnover rate constants, half-life times, and fluxes (Fig. 1).

Intramolecular distribution of 13 C atoms
Since 13 C 6 -glucose is metabolized by the glycolytic pathway and the Krebs cycle, 13 C atoms gain access to the entire metabolic network and ultimately to all lipid moieties, including fatty acids, glycerol groups, and head groups. As a result, multiple new isotopologues have been observed in cell cultures exposed to 13 C 6 -glucose (3,7,8). However, among the many isotopologues, molecules with 3 13 C atoms are particularly abundant (7,8). We confirmed this observation in fruit flies fed with 13 C 6 -glucose. For example we found 13 C 3 molecules to be most abundant among the isotopologues of PC 16:1/16:1 (Fig.   2a). Collision-induced dissociation of 13 C 3 -PC 16:1/16:1 produced the expected fragments LPC16:1, the fatty acid FA16:1, phosphoryl-choline, and glycerol phosphate. 13 C atoms were only present in fragment ions that carried glycerol, such as LPC 16:1 and glycerol-phosphate (Fig. 2b). These data demonstrate that the 13 C atoms of 13 C 3 -PC 16:1/16:1 were confined to the glycerol moiety, a finding we confirmed for other phospholipids, among them PG.
In the PG spectra, we found that 13 C 6 -PG isotopologues were most abundant, consistent with the presence of 2 glycerol groups per molecule (Fig. 3a). Fragmentation of 13 C 6 -PG produced monoisotopic acyl ions and 13 C 3 -labeled LPA ions, corroborating the positional specificity of 13 C atoms in the glycerol groups ( Fig. 3b). In conclusion, we demonstrated the preferential labeling of glycerol moieties and identified specific isotopomers in which all 13 C atoms were confined to the glycerol group. Our data are consistent with previous reports showing glycerol-specific 13 C 9 -isotopomers of CL, a lipid with 3 glycerol groups (8,18). Thus, 13 C 3 -isotopomers (or 13 C 6 -isotopomers in the case of PG or 13 C 9 -isotopomers in the case of CL) can be used to monitor selectively the incorporation of 13 C atoms into the glycerol backbone, which provides an opportunity to measure the rate of de novo synthesis of phospholipid and acylglycerol species.

Estimation of precursor labeling
by guest, on March 5, 2020 www.jlr.org

Downloaded from
In order to determine the fractional synthesis of lipids, it is necessary to know the abundance of 13 C in the intracellular precursor molecule, which in our case is glycerol-3-phosphate. The labeling of glycerol-3phosphate can be determined by LC-MS/MS of the Drosophila metabolome. However, this requires the analysis of each sample on a metabolomics platform, in addition to the lipidomics analysis. Alternatively, it is possible to calculate the labeling of glycerol-3-phosphate from lipidomics data by mass isotopomer distribution analysis (17). Lipids with more than one glycerol group, such as PG, bis(monoacylglycerol)phosphate, and CL, are amenable to this technique, which estimates the glycerol-3-phosphate labeling from the isotopomer pattern of lipids harboring multiple glycerol units. CL is the most abundant of these lipids, but its slow turnover prevents any substantial accumulation of 13 C atoms and consequently makes it a poor choice for mass isotopomer distribution analysis (18). Instead, we performed this analysis with PG, measuring the signal intensities of the glycerol-specific isotopomers 13 C 0 -PG, 13 C 3 -PG, and 13 C 6 -PG.
We verified that the 3 signals had the same retention times (Fig. 4a) and a reproducible and highly accurate mass difference between each other (Fig. 4b). The high accuracy of the mass difference between isotopomers (<1 ppm) was important because it allowed the unequivocal identification of isotopomers in spectra crowded with different 13 C-labeled species (Fig. 4c).
As expected, the incorporation of 13 C atoms shifted the distribution of abundancies gradually from 13 C 0 to 13 C 6 isotopomers (Fig. 5a). From that distribution, we calculated the fraction of labeled glycerol-3phosphate (p) and the fractional synthesis of PG (q) by solving equation 1 for p and q. In order to test the accuracy of this calculation, we measured the fraction of labeled glycerol-3-phosphate by LC-MS/MS.
The measured labeling of glycerol-3-phosphate was slightly higher than the calculated labeling but the difference was small and it vanished after reaching saturation at about 2 days of incubation (Fig. 5b). As expected, the fractional synthesis of PG increased from 3-9 days (Fig. 5c). Together these data demonstrate that the labeling of tissue glycerol-3-phosphate can be estimated with sufficient accuracy from the PG spectra, which makes the method solely reliant on lipidomics, circumventing the need for any additional measurement.

Turnover rates of lipid molecular species
Next we determined the fractional syntheses of various lipid species. These values were calculated by equations 2 and 3, using the various isotopomer distributions and the glycerol-3-phosphate labeling data obtained from PG analysis. The rate by which the fractional synthesis increased during the incubation varied tremendously between different lipid classes (Fig. 6a). From the time evolution of the fractional syntheses, we calculated turnover rate constants, which showed a strong head-group specificity (Fig. 6b).
In contrast, we found less turnover variations within a given lipid class except for PE, where large differences between species were observed. For instance PE 16:0/16:1 turned over 4 times faster than PE 16:1/18:1 and 10 times faster than the alkyl-acyl species PE 18:0e/18:2 and PE18:0e/18:3 (Fig. 6c). Halflife times calculated from the turnover rates of Drosophila lipids ranged from 2 to 200 days. PC, PI, and PG species had the fastest turnover whereas CL and alkyl-acyl-PE species had the slowest (Fig. 6d). The half-lives of CL and ether-PE exceeded the average life span of the flies, which was about 60 days, underscoring the remarkable stability of these two phospholipids. Furthermore, our data showed a strong dependence of the TG turnover rate on the length of the acyl chains (Fig. 6e).
In order to assess the accuracy of our method, we compared the results to turnover measurements obtained with 2 other precursors, including 2 H 5 -glycerol and 2 H 9 -choline. When we analyzed the glycerol turnover of PC by measuring the relative intensities of 2 H 5 -isotopomers in 2 H 5 -glycerol-labeled flies, we found that the turnover rate constants were remarkably similar to those obtained with 13 C 6 -glucose (Fig.   7). In contrast, a similar approach with 2 H 9 -choline yielded much higher turnover rates, which is consistent with established pathways for head group recycling of PC (19). In summary, these data support the validity of our method. They demonstrate distinct turnover kinetics of Drosophila lipids with half-life times stretching over 2 orders of magnitude and identify CL and ether-PE's as the most stable phospholipids.

Tissue-specific lipid fluxes in wild-type
To determine whether lipid dynamics varies between different tissues, we divided fruit flies into body parts containing the nervous system (head), indirect flight muscles (thorax), and the digestive and reproductive systems (abdomen). Comparison between these tissues identified both similarities and differences. For instance, the half-life of PE 18:1/18:2 was longer in indirect flight muscles (>20 days) than in the head or the abdomen (<10 days) whereas the half-lives of PI 18:1/18:2, PC 18:1/18:2, and dMePE 18:1/18:2 were similar (<10 days) (Fig. 8a). In all tissues, the half-lives of PC species were much shorter than the half-lives of PE species (Fig. 8b). Equal half-lives were observed of TG in different tissues (Fig. 8c). However, when we determined flux rates, we found the turnover of most TG species to be lower in the nervous system than in the muscular or the digestive/reproductive system (Fig. 8d).
Furthermore, the TG fluxes were dependent on the number of acyl carbons, peaking at a carbon number of 44. Tissue-specific flux rates were also found for various PE and PC species, with each species having its own pattern of specificity (Fig. 8e, f).
Finally, we applied our method to a Drosophila mutant in order to demonstrate its power to identify specific changes in lipid metabolism in response to the deletion of a lipid-metabolizing enzyme. We chose -VIA), an important phospholipase involved in membrane remodeling and signal transduction (20). Deletion of iPLA2 had no effect on the half-lives of either PC or PE (Fig. 8b) but significantly prolonged the half-life of TG, suggesting that CG6718 is involved either directly or indirectly in TG lipolysis of Drosophila (Fig. 8c). Thus, our data reveal tissue-specific fluxes of individual lipid species

Conclusions
In summary, we have developed a method to measure the dynamics of phospholipids and acyl glycerol lipids. It yields precursor labeling, fractional syntheses, turnover rate constants, half-life times, and flux rates from a series of mass spectra collected at different time points during the incorporation of 13 C 6glucose. In the steady state, the fluxes represent the rates of de novo synthesis and degradation, which must be distinguished from the recycling rates of acyl chains and head groups. Thus, our method measures specifically the turnover of the glycerol backbone, the most stable part of lipid molecules that does not undergo recycling. It has the distinct advantage that it enables comparison among glycerol containing lipids, a very large section of the lipidome, whereas other methods are limited to lipids with specific fatty acids (9)(10)(11) or with specific head groups (4,5).
We applied our method to Drosophila melanogaster, but it is readily applicable to cell cultures and to other organisms, including mouse models. Interestingly, we observed a wide range of turnover rates among the lipid species of Drosophila and showed that the rates are tissue-specific. CL and ether-PE were the lipids with the longest half-life. Previously, we have already demonstrated the extraordinary stability of CL and its dependence on tafazzin and respiratory enzymes (18,21). Our data also show that lipid flux analysis in gene knockout models can reveal novel functions of lipid metabolizing enzymes. As a case in point, we identified the unexpected involvement of the fly homologue of the calcium-independent metabolism. We therefore believe that the present method is a very useful tool to study lipid metabolism.