Characterization and relative quantification of phospholipids based on methylation and stable isotopic labeling[S]

Phospholipids (PLs), one of the lipid categories, are not only the primary building blocks of cellular membranes, but also can be split to produce products that function as second messengers in signal transduction and play a pivotal role in numerous cellular processes, including cell growth, survival, and motility. Here, we present an integrated novel method that combines a fast and robust TMS-diazomethane-based phosphate derivatization and isotopic labeling strategy, which enables simultaneous profiling and relative quantification of PLs from biological samples. Our results showed that phosphate methylation allows fast and sensitive identification of the six major PL classes, including their lysophospholipid counterparts, under positive ionization mode. The isotopic labeling of endogenous PLs was achieved by deuterated diazomethane, which was generated through acid-catalyzed hydrogen/deuterium (H/D) exchange and methanolysis of TMS-diazomethane during the process of phosphate derivatization. The measured H/D ratios of unlabeled and labeled PLs, which were mixed in known proportions, indicated that the isotopic labeling strategy is capable of providing relative quantitation with adequate accuracy, reproducibility, and a coefficient of variation of 9.1%, on average. This novel method offers unique advantages over existing approaches and presents a powerful tool for research of PL metabolism and signaling.

quantitation reduces complexity and increases precision, and more than two samples can be analyzed simultaneously. However, they are limited to the relative quantitation of aminophospholipids. Previously, we have reported a strategy based on phosphate methylation and isotopic labeling for the relative quantitation of phosphoinositides from two different samples ( 13 ). This method involves isotopic labeling of endogenous phosphoinositides with deuterated diazomethane (CD 2 N 2 ) for quantitation of phosphoinositides. CD 2 N 2 was generated in situ through acid-catalyzed hydrogen/deuterium (H/D) exchange and methanolysis of TMS-diazomethane. However, whether such a strategy could be adapted directly for the relative quantitation of other major PLs and LPLs remains unclear, due to the high complexity of PLs and LPLs in the biological sample .
In this study, we present an integrated method that combines a fast and robust phosphate derivatization and isotopic labeling strategy using TMS-diazomethane with shotgun lipidomics technique that enables simultaneous profi ling and relative quantifi cation of PLs.

Cell lines and animals
The prostate cancer cell line, PC3, was obtained from ATCC. The PC3 cells were cultured in DMEM medium with 10% (v/v) FBS and 1% (v/v) penicillin/streptomycin and maintained at 37°C in a 5% CO 2 incubator . Wild-type C57BL/6J mice (male, 12 weeks of age) were used in the study.

Lipid extraction
The lipid extraction was performed using 2 ml polypropylene safe-lock tubes. The lipid extraction was performed as described previously with some modifi cations ( 14 ). Briefl y, cells were collected after digestion with trypsin and subjected to cell counting assays using the Neubauer plate. Prior to lipid extraction, ‫ف‬ 5 × 10 5 cells were centrifuged (13,000 g , 5 min, 4°C) and the pellets were washed two times with 1 ml ice-cold PBS. The supernatants were then discarded and the pellet was spiked with 0. 4  phosphatidic acid (PA) tend to form negatively charged ions and are detected in negative mode . The PE, PS, PG, and PI can also be detected in positive mode by forming as ammonium adducts, but with less ionization effi ciency . Previously, Clark et al. ( 7 ) showed that TMS-diazomethane could be used to methylate the phosphate groups of phosphoinositides, thus increasing the detection sensitivity of phosphoinositides. In their discussion, they also mentioned the potential of this method for the detection of the other PL species. However, in practice, whether phosphate methylation by TMS-diazomethane is feasible for the detection of these PL classes, as well as their lysophospholipid (LPL) counterparts, remains unclear. For instance, whether the TMS-diazomethane is suitable for the methylation of all these PLs or LPLs, and whether the derivatization effi ciency is affected by the chemical properties of individual PL or LPL species, such as FA chain length and double bond numbers, are still unknown. Moreover, identifi cation of PLs and LPLs using ESI-MS/ MS after derivatization also needs to be elaborated. Recently, a strategy based on in-solution concomitant quaternization of glycerophospholipid amino groups and methylation of phosphate groups via reaction with diazomethane was reported to be able to enhance the detection sensitivity of some PL subclasses ( 8 ). However, the strategy just covers the four subclasses of the PLs, and the preproduction of diazomethane with specifi c apparatus is required.
In addition, relative quantifi cation of PL species between different samples will be helpful to understand the dynamic change and functions of PLs in biological systems with different conditions. The quantitative analysis of PLs is usually achieved by the inclusion of multiple internal standards (ISDs) and run separately. An alternative strategy for quantifi cation of PLs would be to incorporate a given heavy nuclei (e.g., 2 H, 13 C, or 15 N) into endogenous PLs by metabolic or chemical labeling. Each population of PLs is reacted with a reagent that differs only in its isotopic composition, thereby creating "heavy" and "light" versions of derivatized PLs, which are distinguishable by MS. Using such a strategy, run-to-run, or even scan-to-scan, variance in lipid ion abundance can be avoided because different samples are ionized simultaneously after separate derivatization and mixing. Moreover, the relative abundance of PLs from different biological samples can be directly determined and compared from the relative MS signal intensities. To date, there are several approaches that have been developed for the relative quantitation of PLs using isobaric mass stable isotope labeling derivatization reagents. For example, Berry et al. ( 11 ) have reported the isobaric tags for relative and absolute quantitation ( 9, 10 ) and isotope-labeled 4-(dimethylamino)benzoic acid N-hydroxysuccinimide ester-based derivatization labeling of primary amino groups of PE species for their quantitative analysis in biological mixtures . Recently, Nie et al. ( 12 ) also demonstrated isobaric mass stable isotope-labeled S,S ′ -dimethylthiobutanoylhydroxysucccinimide ester reagents for characterization and multiplexed quantifi cation of aminophospholipids in biological samples. The advantages of these strategies are that the use of MS2 for of individual species in the labeled and unlabeled lipid extracts, respectively, while I labeled and I unlabeled represent the height of the peak of the labeled and unlabeled PL species after 13 C correction, respectively. The results are presented as mean ± SD .
Due to the overlapping nature of the pseudomolecular ion peak of the species of interest with the M+2 isotope peak from another more unsaturated species that has a 2 Da lower mass (e.g., the 13 C isotope effect of PC 16:0/18:2 on PC 16:0/18:1), a 13 C correction was required for quantifi cation of the PLs. Moreover, due to a mass difference of 2 Da between the 1 H-and Dlabeled isotopic pairs for PL species, as well as their LPL counterparts, there would be an overlap between the unlabeled ion peak of the species of interest with the M+2 isotope peak and the labeled one (e.g., the 13 C isotope effect of unlabeled PE 16:0/18:0 on labeled PE 16:0/18:0). Therefore, a two-step 13 C isotope correction was carried out for quantifi cation of the PLs. Briefl y, for the quantifi cation of the unlabeled species, the isobaric overlap can be corrected by multiplying the monoisotopic peak intensity by a factor as follows ( 15 )

Safety considerations
The TMS-diazomethane is considered a less toxic alternative to diazomethane, but it should still be handled with care, as inhalation of TMS-diazomethane may cause lung damage or central nervous system depression. Therefore, TMS-diazomethane was handled with the appropriate safety procedures in place, i.e., inside a fume hood, with adequate personal safety equipment. Excess TMS-diazomethane is to be neutralized using acetic acid. Care must be taken if handled in large volumes, as it is highly volatile and produces nitrogen gas.

Phosphate methylation enables simultaneous detection of PLs and LPLs under positive ionization mode
To test whether phosphate methylation by TMSdiazomethane was feasible for the detection of the six major PL classes, as well as their LPL counterparts, a standard mixture containing 1.2 nmol each of 17:0/14:1-PC, -PE, -PS, -PA, -PG, and -PI, as well as 17:1-LPC, -LPE, -LPS, -LPA, -LPG, and -LPI, was subjected to derivatization by TMS-diazomethane as described in the Experimental section (see scheme in supplementary Fig. 1 ) . Our results show that phosphate methylation enables detection of the six major PL classes, as well as their LPL counterparts, under positive ionization mode ( Fig. 1 ). However, the PA, PG, and PI classes preferred to be ionized by forming as ammonium adducts , whereas their LPL counterparts tended to be detected in the form of H + adducts, though they also can be adducted with ammonium ions. adding 0.25 ml of deionized H 2 O. After 10 min of incubation at room temperature, the sample was centrifuged at 1,000 g for 10 min. The upper (organic) phase was collected, and the lower phase was re-extracted with the upper phase of a solvent mixture (1.55 ml), whose composition was equivalent to the expected composition of the upper phase (obtained by mixing MTBE-MeOH-H 2 O (10:3:2.5, v/v/v) and collecting the upper phase. Combined organic phases were dried in a vacuum centrifuge.
For the liver sample, a powder sample (approximately 25 mg) from an individual liver sample of a mouse was weighed and further homogenized in 0.5 ml of ice-cold diluted (0.1×) PBS with a Branson Sonifi er S-450 digital ultrasonic cell disruptor/homogenizer. A protein assay on each individual homogenate was performed with a Pierce BCA protein assay kit (Pierce, Rockford, IL).Then, a similar procedure was employed for the lipid extraction from liver samples ( ‫ف‬ 0.5 mg protein).

Derivatization and stable isotope labeling of lipids
A washing step was performed before derivatization. Briefl y, the extracted lipids were resuspended in 682.5 l MTBE/MeOH (MeOD, in the case of isotopic labeling)/2 N HCl (DCl, in the case of isotopic labeling) (500:150:32.5, v/v/v). Phase separation was induced by adding 250 l of 0.1 N HCl (DCl, in the case of isotopic labeling). This step was completed within 5 min. The upper organic phase was immediately transferred to a clean tube and rewashed with 500 l of the lower phase from a two-phase solvent system (MTBE/MeOH or MeOD/0.01 N HCl or DCl, 100:30:25, v/v). After a brief vortex and centrifugation (6,500 g , 2 min at 4°C), the upper phase (lipid extracts) was transferred into another clean tube for derivatization.
Fifty microliters of 2 M TMS-diazomethane in hexane was added into the lipid extract ( ‫ف‬ 0.5 ml) and incubated for 20 min at room temperature. This reaction was terminated by adding 3 l glacial acetic acid, with a visible color change. In the case of labeling, the sample solution was washed twice with 500 l of the lower phase of the two-phase solvent system (MTBE/MeOH/ H 2 O, 100:30:25, v/v). After centrifugation (1,500 g , 3 min), the upper phase was collected and dried in a SpeedVac and redissolved in 80 l chloroform/MeOH (1:1, v/v) with 5 mM ammonium acetate before mass spectrometric analysis.

Mass spectrometric analysis
Mass spectrometric analyses were performed on the TSQ Vantages instrument (Thermo Fisher Scientifi c, Bremen), equipped with a robotic nanofl ow ion source, Triversa Nanomate® (Advion Biosciences), using chips with 5.5 m spraying nozzles. The ion source was controlled using the ChipSoft TM software (Advion Biosciences), with the following settings: ionization voltage was set to 1.25 kV, gas backpressure to 0.5 psi in the positive ion mode, 1 l air gap before chip, and prewetting one time was enabled and vent headspace was disabled. The ion transfer capillary temperature was set to 190°C and s-lens RF amplitude was set to 217V . All MS or MS/MS analyses were operated under Xcalibur software. The Q1 mass range was set to m/z 400-1,200. Precursor ion scan (PIS) or neutral loss scan (NLS) analysis was carried out at a CE of 40 eV, using a dwelling time of 500 ms at a step size of 0.2 Da at unit resolution of the Q1 quadrupole. A 2 min period of signal averaging from 0.5 s per scan was employed for each PIS or NLS spectrum.

Quantitative monitoring of changes in lipid profi le
For relative quantitation of PLs, the peak intensities of unlabeled lipid molecular species were extracted and ratiometrically compared with the corresponding labeled ones as C labeled / C unlabeled = I labeled / I unlabeled , where C labeled and C unlabeled represent the contents characterization of the individual derivatized SMs. Instead, the PIS of m/z 198 was employed for the profi ling of SMs. Therefore, caution should be required in the calculation of PC abundance, as the isotopes of SMs would interfere with it. However, the signal intensity of the PC species is much larger than that of the SM species after derivatization (supplementary Fig. 3C). Therefore, such interference would be negligible, in practice, or can be corrected by 13 C isotopic correction, as described previously.

Characterization of PL and LPL species by ESI-MS/MS
After derivatization, the PL and LPL species were subjected to ESI-MS/MS analysis in the product-ion mode. Importantly, the PA, PG, and PI species can be detected readily with high signal response after derivatization, which facilitates the detection of these PLs under positive mode .
To examine the overall derivatization effi ciency, we prepared a PL and LPL mixture containing the six major PL classes and their LPL counterparts from liver tissue . This mixture was divided into two parts and one of the two parts was subjected to derivatization. After derivatization, the derivatized solution and the one that was not derivatized were analyzed by ESI-MS/MS, respectively. The derivatization effi ciency was found to be good enough (more than 93%) for the PC class and signifi cant enhancement in the signal intensity of the PC species was also achieved by derivatization (supplementary Fig. 2). Similar results for other PL classes were obtained from the MS analysis (data not shown). Importantly, the mole percent of the different PC species calculated from the mass spectra obtained by PIS at m/z 184 or 198, respectively, is virtually equal within experimental errors after correction for differential 13 C distribution. These results confi rmed that the derivatization effi ciency was effective enough and not affected by the FA chain length or saturation. Moreover, it also suggested that the ionization effi ciencies of individual derivatized molecular species of PLs or LPLs are similar within each class. In addition, our results also reveal that most of the SMs (approximately 80%) can be successfully derivatized by our protocol (supplementary Fig. 3). The derivatized SMs tend to be ionized as proton adducts ([M+H] + ) rather than lithium adducts, even in the presence of a high concentration of LiOH in the analyzing solution. Thus, the NLS of 217(213+CH 2 ), corresponding to methylated phosphocholine plus methyl aldehyde from lithiated SMs, is unsuitable to be used for the pairs of PE, PG, PS, PA, and PI were 2, 2, 4, 4, and 2 Da, respectively. No unlabeled PLs were detected in the isotopically labeled samples, demonstrating that the process of generating CD 2 N 2 and phosphate methylation of PLs was achieved close to completion. These values matched the theoretical values calculated from the number of available phosphate or carboxylic acid groups in each PL class . For example, the PE species has only one phosphate group, whereas the PS species has one phosphate group and one carboxylic acid group. Therefore, PE and PS can be labeled with one and two -CHD 2 groups, respectively, which leads to a mass difference between the 1 H-and the D-labeled isotopic pairs of PtdEtn and PtdSer with 2 and 4 Da, respectively. Similarly, rapid and effi cient phosphate methylation and isotopic labeling of LPLs were achieved and the mass differences between the 1 H-and the D-labeled isotopic pairs were matched to the theoretical values ( Fig. 3 ).
However, similar to our previous study, we found that the isotopic cluster of D-methylated PLs and LPLs shows the presence of one less abundant peak that is 1 mass unit smaller than the monoisotopic peak for the D-methylated PLs and LPLs. This is partly due to a small percentage of 1 H atoms in the D-marked MeOD reagent used in our experiments, as well as due to the effects of back-exchange during the generation of D-diazomethane. To minimize such effects, a prederivatization wash, as described in the Experimental section, was employed, which could reduce the concentration of hydrogen atoms in the environment. However, this could lead to the production of a small proportion of -CD 3 group in the resulting methyl ester due to the H/D exchange of acidic protons from the phosphate group with deuterium atoms in the environment . Previous studies had shown that two of the protons in the resulting methyl ester are derived from the diazomethane and the other one is the "donated" acidic proton from the carboxylic acid (phosphate, in our case ). Therefore, the fi nal concentration of the DCl in the prederivatization wash solution was controlled to be less than 0.1 N and the process was completed within 5 min. Under such optimal conditions, minimal peaks that were 1 mass unit smaller or larger than the monoisotopic peak for the D-methylated PL species and their LPL counterparts were found ( Fig. 3C ). The concentration of DCl in derivative reagents and the reaction time were also investigated to optimize the derivatization of PL species and their LPL counterparts. The PL and LPL derivatives from the various reaction conditions showed no signifi cant differences from the ESI-MS or MS/MS spectra (data not shown).
Taken together, our results show that the use of TMSdiazomethane, with MeOD/DCl as a D reservoir, is feasible for isotopic labeling of PL and LPL species, which could then be employed for relative quantitation of PL and LPL species. In addition, our results also indicate that isotopic labeling enables simultaneous fractionation of PL species from two different samples to reduce the sample complexity, as the isotopic pairs were coeluted from the column. To test whether NLS or PIS was feasible for the identification and quantitation of PLs and LPLs from biological samples, we prepared a cellular PL mixture extracted from a PC3 prostatic cancer cell line spiked with ISDs of PLs and LPLs. This mixture was derivatized, and the derivatized solution was analyzed by the NLS-or PIS-based ESI-MS/MS mentioned above. Our results showed that phosphate methylation enables rapid detection of the six major PLs ( Fig. 2 ) and their LPL counterparts (supplementary Fig.  6) under positive ionization mode. Importantly, the ISDs for the PLs were readily detected under our conditions. In addition, as mentioned above, the derivatization effi ciency and ionization effi ciency were not affected by the FA chain length or saturation of individual molecular species. Therefore, the different derivatized PL and LPL species can be quantifi ed by PIS-or NLS-based ESI-MS/MS with one corresponding ISD after correction for differential 13 C isotope distribution and under a diluted lipid concentration to avoid lipid aggregation, as previously described ( 15 ).

Isotopic labeling of PLs through phosphate methylation by CD 2 N 2 generated in situ
We tested and optimized the strategy based on phosphate methylation and isotopic labeling for the relative quantitation of PLs and their LPL counterparts. It is reasonable to assume that when CD 2 N 2 is utilized instead of a light reagent, a mass increase should be detected by the mass analyzer. Such an increase should be directly proportional to the amount of available phosphate or carboxylic acid groups in the PL species (supplementary Fig. 7). It means that the larger the number of phosphate or carboxylic acid groups in a PL species, the larger the mass difference between the 1 H-and the D-labeled isotopic pair that can be identifi ed . To test our hypothesis, a PL and LPL standard mixture, containing 17:0/14:1-PC, -PE, -PS, -PA, -PG, and -PI, as well as 17:1-LPC, -LPE, -LPA, -LPG, and -LPI, was subjected to our stable isotopic labeling analysis. A rapid and effi cient phosphate methylation and isotopic labeling of PE, PG, PS, PA, or PI was achieved ( Fig. 3 ). The mass difference between the 1 H-and the D-labeled isotopic standards mixture was divided into two aliquots and separately labeled with either 1 H-or D-labeled diazomethane, as described in the Experimental section . The two product solutions were subsequently mixed to simulate relative quantifi cation . This mixture was subjected to [DG+H-H 2 O] + -specifi c PIS-based ESI-MS/MS analysis. The peak intensities of unlabeled PL or LPL standards were extracted from the [DG+H-H 2 O] + -specifi c PIS spectrum and compared ratiometrically to the corresponding labeled ones for the relative quantitation of labeled and unlabeled lipid species. The signal intensity of the isotopic pair of the Fig. 8A). However, the PLs were readily detected after prefractionation by a hydrophilic interaction liquid chromatographybased chromatography , which separates the PLs from the TGs in the lipid droplet sample (supplementary Fig. 8B). Moreover, the isotopic pair of PCs was coeluted from the column and the profi les of unlabeled or labeled PLs were not affected by the prefractionation.

Relative quantitation of PL and LPL species by ESI-MS/MS
To characterize the accuracy of relative quantitation by phosphate methylation with CD 2 N 2 , the PL and LPL two different conditions. To further characterize the analytical properties and the accuracy of relative quantitation of PL species from real samples by phosphate methylation with CD 2 N 2 , the prostatic cancer cell cultures were divided into several aliquots before PL extraction and separated into the following proportions of cell volume: 3:1, 1:1, and 1:3 and separate labeling with either 1 H-or D-labeled diazomethane. The mixture of 1 H-or D-labeled PLs at different ratios was subjected to NLS-or PIS-based ESI-MS/MS analysis and the distribution of the H/D ratios was examined . Our results showed that the observed H/D ratios were consistent with the theoretical ratios of PC, PE, PS, PG, PI, and PA present in the mixture. Example mass spectra are shown for PE and PS populations, one with a ratio at 1:1 ( Fig. 5 ) and the other with a ratio at 1:3 (supplementary PL and LPL standards was found to be in good agreement with the theoretically expected ratios ( Fig. 4 ). The ratios of peak intensities were calculated for each pair of heavyand light-labeled PL or LPL standards after 13 C correction from four runs and the average ratios are displayed in Table 1 . The average experimental ratios for the PL and LPL standards were found to be in good agreement with the theoretically expected ratios, with an average coefficient of variation of 9.1%.
The aim of our experiments was to develop a rapid procedure to measure differences and compare the relative amounts of specifi c PL species between two populations of PLs extracted from two different samples utilizing stable isotopic labeling. In practice, these two samples could be obtained, for example, from cells, tissues, or body fl uid at the Experimental section, or by combining our method with the appropriate chromatographic separation methods. Second, for quantifi cation of PS species, caution is required because parts of the free hydroxyls in the head group of PG species are methylated during the phosphate methylation process (defi ned here as over-methylated PG), which would lead to an overlap between the D-labeled PS species and the unlabeled, but over-methylated, PGs that have the same fatty acyl composition. For instance, there would be an overlap between the D-labeled PS 31:1 (752 Da, 748+2D) and the unlabeled, but over-methylated, 31:1 PG (752 Da, 738+CH 2 ) in the DAG + -specifi c PIS spectrum ( Fig. 3 ). However, such overlap can be dissolved by addition of LiOH instead of NH 4 Ac, as the PG species prefer to adduct with Li + rather than NH 4 + in the presence of LiOH. Third, the plasmalogens would be hydrolyzed upon methylation due to the presence of HCl in our experimental conditions, as the vinyl ether motif within these compounds is acid labile ( supplementary Fig. 10 ) .   Fig. 9). These spectra illustrate that the difference in signal intensity of the PE or PS species between the heavy and light forms was found to be in good agreement with the theoretically expected ratios.

Potential pitfalls in quantitation of PLs
Although high accuracy for quantitation of PLs is possible under optimal conditions, there are several situations where identifi cation and quantitation of PLs could be compromised. First, unlike the isobaric mass tag labeling strategies mentioned above, in which MS 2 space was used to reduce complexity and increase precision, our labeling strategy is not a "mass-balanced" method and can be analyzed only in MS 1 . Therefore, caution should be used with 13 C correction for quantifi cation of PLs and LPLs, as there is only a 2 Da difference between the 1 H-and the D-labeled isotopic pairs for PC, PE, PG, and PI species, as well as their LPL counterparts . However, such potential quantitation error can be resolved by careful correction, as described in  Therefore, caution is required for profi ling or quantitative applications of this protocol. However, a further optimization of our protocol, such as fi nding an alternative acid for HCl or decreasing the concentration of HCl, may resolve such a problem.

CONCLUSIONS
The results presented here demonstrate the feasibility of rapid profiling and relative quantitation of PLs and LPLs from biological samples on the basis of phosphate methylation and isotopic labeling by chemical derivatization with diazomethane. Our novel assay is not only rapid, simple, and sensitive, but also offers several important advantages for PL measurement. First, phosphate methylation allows fast and sensitive identifi cation of the six major PL classes, including their LPL counterparts, under positive ionization mode from biological samples, which excludes the need to switch between positive and negative mode. Importantly, the PA, PG, and PI species can be detected readily with high signal response after derivatization, which facilitates the detection of these PLs under positive mode . Second, isotopic labeling-based quantitation of PLs also minimizes matrix effects or run-to-run differences that could potentially affect results. Moreover, isotopic labeling enables lipid fractionation to reduce sample complexity, as the isotopic pairs were coeluted from the column. Therefore, our methods would present a powerful tool for research of PL metabolism and signaling.