Quantitative structural characterization of phosphatidylinositol phosphates from biological samples[S]

The aspects of cellular metabolism controlled by phosphatidylinositol phosphates (PtdInsPs) have been broadly expanded, and these phospholipids have drawn tremendous attention as pleiotropic signaling molecules. PtdInsPs analysis using LC/MS/MS has remained challenging due to the strong hydrophilicity of these lipids. Multiple reaction monitoring (MRM) or a neutral loss scan has been performed to quantitatively measure PtdInsPs after chemical derivatization on the phosphate groups of inositol moieties. Only predefined PtdInsPs can be measured in MRM mode, and fatty acyl compositions of sn-1 and sn-2 positions of PtdInsPs cannot be obtained from a neutral loss scan. In our present study, we developed a simple LC/MS/MS method for structural identification of sn-1 and sn-2 fatty acids of PtdInsPs and their relative quantitation. Precursor ion scans of sn-1 monoacylglycerols (MAGs) of PtdInsPs provided structural information about the lipids, and ammonium adduction enhanced signal intensities of PtdInsPs. The relative amount of observed PtdInsPs in biological samples could be compared using chromatographic peak areas from the neutral loss scans. Using precursor ion scans of sn-1 MAG and neutral loss scans of headgroups, major PtdInsPs in cells and tissues were successfully identified with structural information of sn-1 and sn-2 fatty acids, and their relative amounts in different samples were compared.

columns under strong alkaline conditions. In addition, a normal-phase column has been tried to separate highly polar PtdInsPs (28,31).
TMS-diazomethane has been used to methylate carboxylic acids or phosphates (14,32,33). Wasslen et al. (33) showed that trimethylsilylation enhanced the ionization efficiencies and sensitivities of glycerophospholipids and provided clearer fragmentation patterns than unmodified glycerophospholipids. Polar lipids, including glycerophospholipids, showed increased signal enhancement due to better ionization efficiencies after methylation. In addition, methylation of glycerophospholipids has resulted in better chromatographic peak shapes and resolution, compared with those of nonmethylated molecular species (34). Methylation of the phosphate groups of the inositol moiety could enhance an RPLC separation and ionization efficiencies of PtdInsPs, as well as prevent nonspecific binding of PtdInsPs within LC/ MS systems. Thus, methylation of phosphate groups should be especially beneficial for the analysis of highly phosphorylated PtdInsP 2 and PtdInsP 3 (35), which typically exist in very low levels in biological samples. PtdIn-sPs from cells have been measured using a neutral loss scan upon headgroup fragmentation in the positive ion mode after derivatization with TMS-diazomethane (14). Clark et al. (35) successfully quantified PtdInsP 3 along with other PtdInsPs from stimulated cells and liver tissues using LC/MS/MS after TMS-diazomethane derivatization. They performed MS 2 analysis for individual PtdInsPs to get the information of the sn-1 and sn-2 fatty acyl compositions of the observed PtdInsPs.
Methylation of PtdInsPs successfully enables detection of PtdInsPs in biological extracts using LC/MS/MS but was at the limits of detection for quantitation of PtdInsP 3 . We have determined whether detection limits can be improved by using ammonium adduction. This would be especially crucial to profile PtdInsPs using neutral loss scans to search for more PtdInsPs in biological extracts because the sensitivity in scan mode is lower than that in nonscanning technique such as multiple reaction monitoring. In addition, it would be advantageous for precursor ion scans of sn-1 monoacylglycerol (MAG) in positive ion mode to get fatty acid compositions at sn-1 and sn-2 positions of Pt-dInsPs. In this study, the combined strategy of precursor ion scans of sn-1 MAG and neutral loss scans of headgroups was successfully applied to compare the relative amounts of PtdInsPs in biological samples. Hepatocytes and mice livers were used to explore the changes of PtdIn-sPs after insulin stimulation. PtdInsPs have been deacylated and analyzed by measuring 32 P-labeled inositides or 3 H-myo-inositol using HPLC (18)(19)(20). TLC has been conventionally used for the analysis of phospholipids and has been applied to analyze isomers of PtdInsPs (21,22). A TLC analysis is relatively nonspecific, and lipids extracted from a TLC plate have been further analyzed with GC or GC/MS (23). These methods using radiolabeling with 32 P or a TLC separation are very laborious and have limited applications. Mass spectrometric detection of phospholipids has been performed in a positive or negative ion mode. Collision-induced dissociation (CID) is the most commonly used tandem mass spectrometric strategy. CID in a positive ion mode provides structural information about the polar headgroups of phospholipids. Haag et al. (17) determined PtdInsPs from human T cells before and after stimulation in a positive ion mode, through diacylglycerol (DAG) ions resulting from neutral loss of the headgroup from PtdIn-sPs. On the other hand, the fatty acid composition attached at the sn-1 or sn-2 position of the phospholipids was obtained from negative ion mode CID. Several research groups have performed structural characterization of Pt-dInsPs using CID in negative ion mode, and determined fatty acyl compositions using the neutral loss of free fatty acid substituents or corresponding ketens (13,16,24,25). Thus, mass spectrometric analysis from both positive and negative ion modes should be required for complete structural elucidation of PtdInsPs. However, neutral loss scans of free fatty acid substituents may not be suitable for routine analysis of PtdInsPs due to the lack of sensitivity in negative ion mode, compared with those in positive ion mode. An ELISA has been used to analyze PtdInsP, PtdInsP 2 , or PtdInsP 3 . However, different types of ELISA kits are necessary for each class of PtdInsPs species, and the fatty acyl composition of PtdInsPs cannot be identified by this approach (26,27).

Materials and sample preparation
Recently, LC/MS or LC/MS/MS has been widely used for analysis of various phospholipids including PtdInsPs (28)(29)(30). LC/MS/MS is able to detect intact phospholipids without deacylation or chemical derivatization and provides valuable structural information for the identification of phospholipids. Phosphate groups on the inositol moiety render distinct chemical properties on PtdInsPs compared with other types of phospholipids. Observing positively charged PtdInsPs during a mass spectrometric analysis is difficult due to the acidic nature of phosphates. Thus, structural identifications using a tandem mass spectrometric analysis for PtdInsPs have been published using the deprotonated forms (13,16,25,30). Ion suppression during the electrospray ionization process could be more problematic for PtdInsPs because they are low abundance lipids in biological samples. MS combined with an LC separation module could be a suitable method for analysis of PtdInsPs from various biological samples to discriminate isomers and to overcome ion suppression. Reverse-phase liquid chromatography (RPLC) separation has generally resulted in poor performance due to the high polarity of the phosphate groups on PtdInsPs. Ogiso et al. (30) reported successful separation of deprotonated PtdInsPs using C8 reconstituted in 100 l methanol, and 20 l of 2 M TMS-diazomethane in hexane (Sigma-Aldrich) was added. The derivatization reaction was allowed to proceed for 30 min at room temperature, and 3 l glacial acetic acid was added to the reaction solution. Four hundred fifty microliters CHCl 3 /methanol/water (8:4:3, v/v/v) was added to the reaction solution, and the solution underwent vigorous vortexing. After centrifugation, the organic layer of the solution was collected and dried under vacuum. The dried extract was reconstituted in 30 l methanol for the LC/MS/ MS analysis.

Cell culture and insulin treatment
The mouse hepatocyte cell line AML12 cells were purchased from the ATCC (Manassas, VA). Cells were maintained in DMEM/ F12 (Welgene Inc., Gyeongsan-si, Republic of Korea) supplemented with 10% FBS, dexamethasone (100 nM) and 1× insulintransferrin-selenium-pyruvate supplement solution (Welgene Inc.). Cells were cultured in DMEM/F12 without any supplements overnight before insulin treatment. For experiments, the cells were incubated in DMEM/F12 with PBS or insulin (100 ng/ ml or 1,000 ng/ml) for 20 min. After the incubation, the cells were immediately collected and washed with PBS. The cells were lysed with cold methanol and vigorous vortexing, then kept at -80°C until processing. PtdInsPs from biological samples were extracted by using the Folch method with minor modifications (36). Five hundred microliters of CHCl 3 /methanol/12M HCl (80:40:1, v/v/v) and 100 l of internal standard solution (10 M each of 17:0/20:4 PtdIn-sPs in methanol) were added to approximately a million AML cells or 20-50 mg of mice livers. Samples were homogenized using a TissueLyzer (Qiagen Inc., Valencia, CA), and allowed to stand for 25 min at room temperature to ensure all PtdInsP dissolved in solvents. After centrifugation, the supernatant was collected for the extraction of PtdInsPs. Then, 170 l of CHCl 3 and 340 l of 0.01 M HCl were added to the supernatant, followed by vigorous vortexing. The bottom organic layer containing the lipids was collected and transferred to a new tube. Another 170 l of CHCl 3 was added to the remaining solution. After vigorous vortexing and centrifugation, the bottom organic layer was collected and added to the previously collected organic layer. The organic layer presumably containing PtdInsPs was dried under vacuum. The dried extract was kept at 20°C prior to chemical derivatization. Standard solutions of PtdInsPs were prepared using the same procedure as described for biological samples. The dried extract was Framingham, MA), and a reverse-phase column (Pursuit 5 C18 150 × 2.0 mm; Agilent Technologies, Santa Clara, CA) was used. MS was operated in the positive ion mode with turbo ion-spray voltage of 5,500 V, using 20 psi curtain gas, 50 psi nebulizer gas, and 50 psi drying gas at a temperature of 400°C. The LC separation for PtdInsPs used mobile phase A (5 mM ammonium formate-methanol-tetrahydrofuran; 500:200:300) and mobile phase B (5 mM ammonium formate-methanol-tetrahydrofuran; 100:200:700), and proceeded at 200 µl/min and 23°C. The separation gradient was as follows: 5% of B at 0 min, 5% of B for 10 min, 5% to 90% of B for 35 min, 90% of B for 0.1 min, and then 90% to 5% of B for 14.9 min. For comparison, mobile phase A (0.1% formic acid-methanol-tetrahydrofuran; 500:200:300) and mobile phase B (0.1% formic acid-methanol-tetrahydrofuran; 100:200:700) were also used with the same gradient. A neutral

Insulin-treated mice
Mice were housed at 21°C-23°C with 12 h light and 12 h dark cycles in the Specific Pathogen Free animal facility at the University of Ulsan with free access to water and standard rodent chow. All animal care and procedures were conducted according to the protocols and guidelines approved by the University of Ulsan Animal Care and Use Committee. After anesthesia, female C57BL/6 mice at 8 weeks of age were administered with PBS or insulin (1 U/kg body weight via the vena cava). Five minutes after insulin infusion, the liver was removed and frozen in liquid nitrogen and kept at -80°C until processing.

Ammonium adducts for PtdInsPs
PtdInsPs were permethylated via chemical derivatization with TMS-diazomethane. Incubation with TMS-diazomethane at room temperature for 30 min reproducibly formed permethylated PtdInsPs. Oxygen atoms within the phosphate hydroxyl groups were known to be methylated (35). All phosphate groups were permethylated after the reaction in our study. For example, 17:0/20:4 PtdInsP 2 contains five hydroxyl groups on phosphates, including a loss scan and a precursor ion scan, specific to the structural characteristics of PtdInsPs, were performed in the positive ion mode. Collision energies of 23-46 V were used for the neutral loss scans depending on neutral loss masses, and 53 V collision energy was used for the precursor ion scans. The extracted ion chromatogram (EIC) corresponding to the specific lipid was used for relative quantitation after normalization using the in-   The signal enhancement was the peak area ratio of ammonium ion adducted PtdInsPs to protonated PtdInsPs. Peak areas of ammonium ion adducted or protonated PtdInsPs were obtained when either ammonium formate or formic acid was included in LC mobile phases.
18:1/20:4 PtdInsPs. The 17:0/20:4 PtdInsPs, used as internal standards, were also found at another retention time. The number of phosphorylations of inositol phosphates did not affect the LC elution time among PtdInsPs with the same fatty acyl compositions. Fewer PtdInsPs species were observed when ammonium ion additive was not included in the LC mobile phases (supplemental Fig. S2). When the same PtdInsPs were observed, the signal enhancement due to ammonium ion adduction ranged between 2-and 10fold depending on PtdInsPs species.
The benefit of addition of ammonium ion in mobile phases for signal enhancement and reproducible analysis was demonstrated by several groups using small chemicals (37,38). We wanted to identify PtdInsPs from biological samples such as cells and mice livers, and also to obtain the relative quantitation of PtdInsPs. Reproducibility and signal enhancement of ammonium ion adducted PtdInsPs were evaluated from three independent experiments. Signal enhancement due to ammonium ion adduction was evaluated using peak areas of ammonium ion-adducted PtdInsPs to protonated PtdInsPs in Table 1. Generally, 2-to 6-fold signal enhancement was expected due to ammonium ion adduction. Reproducible ammonium ion adduct formation of PtdInsPs was evaluated, and the relative standard deviation for each type of PtdInsPs was shown in Table 2. Linearity of neutral loss scans of each PtdInsP was evaluated as shown in supplemental Fig. S3. Ammonium ion adduct formation of PtdInsPs was stable and reproducible, and neutral loss scans provided a linear relationship between signal responses and amounts of PtdInsPs. Precursor ion scans also enabled reproducible quantitation of Pt-dInsPs; however, the ion abundances of precursor ion scans of sn-1 MAG were lower than neutral loss scans of headgroups. Thus, we used neutral loss scans for relative quantitation of PtdInsPs from biological samples.

Structural identification of sn-1 and sn-2 fatty acids of PtdInsPs
Permethylated PtdInsPs provided DAGs as major fragment ions in CID spectra as shown in the previous section. Thus, neutral loss scans, corresponding to phosphoinositol headgroups, have been performed to profile PtdInsPs. Clark et al. (35) also used a similar strategy to profile PtdIn-sPs, and they used a neutral loss scan of 598 to profile Pt-dInsP 3 . In the present study, a neutral loss scan of 615 was used to profile potential PtdInsP 3 isomers due to ammonium phosphate attached to the glycerol backbone; 17:0/20:4 PtdInsP 2 + 5Met was mainly observed, although overmethylated PtdInsPs were found as well, but at lower abundances (Fig. 1).
Adduct formation such as [M + NH 4 ] + or [M + Na] + is commonly observed in mass spectrometric analyses. Although adduct formation often enhances ionization efficiencies of analytes, these phenomena may not be reproducible and may complicate the quantitative analysis and interpretation of MS/MS spectra. We rarely observed sodium or potassium ion adduct formation for permethylated PtdInsPs. On the contrary, signal intensities of ammonium adducts with permethylated PtdInsPs were generally comparable to those of protonated forms in the positive ion mode. Figure 1A shows the mass spectrum of permethylated PtdInsPs without ammonium ion additive in the LC mobile phase. . As shown in Fig. 2A Neutral loss scans were performed, and peak areas for specific PtdInsPs were used to evaluate the reproducibility of ammonium ion adduct formation. The peak area of each PtdInsPs species was normalized to that of the corresponding internal standard. RSD indicates the relative standard deviation of three experiments. acyl compositions and comparison of LC retention times from neutral scan spectra enabled us identify the unknown PtdInsPs with sn-1 fatty acid information. Biological samples were known to primarily contain 16:0, 18:1, and 18:0 fatty acids at the sn-1 position of phospholipids (40). Thus, we included precursor ion scans for 16:0, 18:1, and 18:0 fatty acids, as well as 17:0 for internal standards, and the precursor ion scan could then be modified depending on the fatty acid compositions of any PtdInsPs.
The neutral loss scans for insulin-treated AML12 cells could observe 18:1/20:4 PtdInsPs, 18:0/20:4 PtdInsPs, 18:0/20:3 PtdInsPs, 16:0/20:4 PtdInsPs, 16:0/20:3 PtdInsPs, and so forth. The EICs generated from the neutral loss scan of 291 for PtdIns are shown in Fig. 4. The sn-1 and sn-2 fatty acids of PtdIns were assigned with specific fatty acid compositions using the result from precursor ion scans shown in Fig.  5. PtdInsPs having fewer unsaturated fatty acids or longer chain fatty acids were likely to elute late as shown in Fig. 3 and Fig. 4. Precursor ion scan of 341 was performed to find PtdIn-sPs having stearic acid at the sn-1 position shown in Fig. 5, where mass spectra were generated for the shaded retention time areas. EIC for m/z 920.7 from neutral loss scan of 291 shows two peaks in Fig. 4. The peak eluted later (42.5-43.5 min) should come from 18:0/20:3 PtdIns, which is confirmed from Fig. 5B. However, the peak eluted earlier (41.5-42.5 min) might not be any PtdIns species or might be generated from the second isotope of 18:0/20:4 PtdIns, as assumed from Fig. 5A. For that retention time, 18:0/20/4 ion adduction. Likewise, neutral loss scans of 507, 399, and 291 were used to profile PtdInsP 2 , PtdInsP, and PtdIns species, respectively, from cells and mice livers.
A neutral loss scan only provides the number of phosphate groups in the inositol phosphates of PtdInsPs, but fatty acyl compositions at the sn-1 and sn-2 positions could not be identified. CID spectra for PtdInsPs should be obtained to confirm the complete structural characterization of the lipid species (35). Potential PtdInsPs in biological samples should be selected from a neutral loss scan result for complete structural elucidation of the PtdInsPs, and this process is often complicated and tedious. Then, MS 2 experiments for all potential PtdInsPs should be performed with frequent modification of the method depending on specific PtdInsPs observed in biological samples. Otherwise, the composition of fatty acyl chains in PtdInsPs could not be identified only with the neutral loss scan result. For example, m/z 1,134.6 from a neutral loss scan of 507 would be identified as 38:4 PtdInsP 2 , but its fatty acyl composition could be any of the following molecular species: 16:0/22:4 PtdInsP 2 , 18:0/20:4 PtdInsP 2 , 18:1/20:3 PtdInsP 2, or other possibilities. DAG-specific precursor ion scans could not elucidate the fatty acyl compositions at the sn-1 and sn-2 positions of PtdInsPs either (39).
In our present study, precursor ion scans of sn-1 MAG were performed to identify sn-1 fatty acid of PtdInsPs in positive ion mode. Mass differences of 108 Da among differentially phosphorylated PtdInsPs having the same fatty  Table S1A.
Changes of PtdInsPs in insulin-stimulated mice livers are shown in Fig. 7. Peak areas of all PtdInsPs observed are shown in supplemental Table S1B. Highly phosphorylated 18:0/20:4 PtdInsP 3 were increased the most for insulin stimulation, compared with the corresponding PtdInsP 3 from control livers. Generally, PtdInsP expression in livers PtdInsPs were the only PtdInsPs molecular species observable; thus the peak eluted at 41.5-42.5 min was assigned as 18:0/20:4 PtdIns. Likewise, shaded peaks for m/z 922.7 and m/z 924.8 were assigned as 18:0/20:2 PtdIns and 18:0/20:1 PtdIns, respectively. Precursor ion scans also enabled reproducible quantitation of PtdInsPs; however, the ion abundances of precursor ion scans of sn-1 MAG were lower than neutral loss scans of headgroups in positive ion mode. Thus, we used neutral loss scans for relative quantitation of PtdIn-sPs from biological samples. PtdInsP 3 's were observed only from neutral loss scan analysis, and their sn-1 fatty acyl compositions were assigned when fewer phosphorylated PtdInsP 2 and PtdInsP species with the same fatty acyl compositions were observed at the same retention time, as shown in Fig. 5A and supplemental Fig. S4.

Quantitative profiling of PtdInsPs
Insulin has been shown to stimulate Akt phosphorylation via activation of PI3K (41). Western blot analysis shows was increased after insulin treatment, but PtdIns expression was decreased. Interestingly, PtdInsP 2 levels in cells and livers were not increased when the PtdInsP 3 increment was obvious after insulin stimulation. The major fatty acyl composition of PtdInsPs in mice livers was 18:0/20:4 PtdIn-sPs, and the amount of 18:0/20:4 PtdInsPs constitutes 80%-90% of all observed PtdInsPs.

CONCLUSION
Our result shows that the combined strategy of the neutral loss scans of phosphoinositol headgroups and precursor ion scans of sn-1 MAG is able to profile quantitatively PtdInsPs with structural information of the sn-1 and sn-2 fatty acids, and stable ammonium ion adduction is especially beneficial to observe more PtdInsPs. Quantitative profiling of PtdInsPs in cells and tissues was successfully performed with structural information of the sn-1 and sn-2 fatty acid compositions. Pt-dInsP 3 's were observed only for some of PtdInsPs in the samples we used. Although major fatty acid compositions of PtdInsPs were 18:0/20:4 PtdInsPs for hepatocytes and livers, these fatty acid compositions might vary depending on the types of cells and tissues. Regioisomers of PtdInsPs, considering different phosphorylation sites on the inositol groups, could not be resolved with this method, and solving this problem remains part of our future work. Changes of PtdInsPs after 20 min incubation with 100 ng/ml of insulin. The 1,000 ng/ml of insulin-treated cells shows a similar result; data not shown. Peak area was the normalized peak area of each PtdInsPs to the corresponding internal standard from neutral loss scan analysis. Changes of PtdInsPs were calculated by dividing the average peak areas of PtdInsPs in insulin-stimulated cells by those in control cells. Error bars indicate the relative standard deviations of independently prepared samples (n = 3). Neutral loss scans of headgroups were used for quantitation, and EICs for specific PtdInsPs can be found in Fig. 4 and supplemental Fig. S4.   Fig. 7. Changes of PtdInsPs in insulin-stimulated mice livers. Changes of PtdInsPs were calculated by dividing the average peak areas of PtdInsPs in insulin-stimulated mice livers by those in control livers. Error bars indicate the relative standard deviations of independently prepared samples (n = 4).