Quantitative analysis of sphingolipids for lipidomics using triple quadrupole and quadrupole linear ion trap mass spectrometers.

Sphingolipids are a highly diverse category of bioactive compounds. This article describes methods that have been validated for the extraction, liquid chromatographic (LC) separation, identification and quantitation of sphingolipids by electrospray ionization, tandem mass spectrometry (ESI-MS/MS) using triple quadrupole (QQQ, API 3000) and quadrupole-linear-ion trap (API 4000 QTrap, operating in QQQ mode) mass spectrometers. Advantages of the QTrap included: greater sensitivity, similar ionization efficiencies for sphingolipids with ceramide versus dihydroceramide backbones, and the ability to identify the ceramide backbone of sphingomyelins using a pseudo-MS3 protocol. Compounds that can be readily quantified using an internal standard cocktail developed by the LIPID MAPS Consortium are: sphingoid bases and sphingoid base 1-phosphates, more complex species such as ceramides, ceramide 1-phosphates, sphingomyelins, mono- and di-hexosylceramides, and these complex sphingolipids with dihydroceramide backbones. With minor modifications, glucosylceramides and galactosylceramides can be distinguished, and more complex species such as sulfatides can also be quantified, when the internal standards are available. JLR LC ESI-MS/MS can be utilized to quantify a large number of structural and signaling sphingolipids using commercially available internal standards. The application of these methods is illustrated with RAW264.7 cells, a mouse macrophage cell line. These methods should be useful for a wide range of focused (sphingo)lipidomic investigations.

Cell culture RAW264.7 cells, a macrophage-like cell line derived from tumors induced in male BALB/c mice by the Abelson murine leukemia virus ( 37 ), were obtained from the American Type Culture Collection (Manassas, VA) (cat# TIB-71; lot# 3002360), stored as frozen stocks to ensure the cells were never passaged more than 20 times, and cultured according to LIPID MAPS protocols (www.lipidmaps.org), as briefl y summarized here. The cells were grown in 60 mm plastic culture dishes in DMEM supplemented with 10% FBS, 4 mM L-glutamine, 4.5 g/L glucose, 1.5 g/L sodium bicarbonate, 100 U/ml penicillin, and 0.1 mg/ml streptomycin. RAW264.7 cells were cultured at 37°C, 95% relative humidity, and 5% CO 2 in a ThermoForma Stericult CO 2 incubator. Cells that were at 80% confl uence were rinsed with phosphate buffered saline (PBS), scraped from the dish, and seeded at 2.5 × 10 6 cells in 5 ml of media in 60 mm dishes, and analyzed after 24 h. The cells were quantifi ed by DNA assay using the Quant-iT DNA Assay Kit, Broad Range (Molecular Probes, cat #Q-33130).

Lipid extraction from RAW264.7 cells
The cells were washed twice with PBS, with the dishes tilted to aid in the removal of as much liquid as possible. Then the cells were scraped from the dish in the residual PBS (typically < 0.2 ml) 4 using Nalgene cell scrapers (Rochester, NY) and transferred into 13 × 100 mm borosilicate tubes with a Tefl on-lined cap (catalog #60827-453, VWR, West Chester, PA). After adding 0.5 ml of CH 3 OH and 0.25 ml of CHCl 3 , the internal standard cocktail (500 pmol of each species dissolved in a fi nal total volume of 10 l of ethanol) were added, and the contents were dispersed using a Branson 1510 ultra sonicator (Sigma) at room temperature for 30 s. This single phase mixture was incubated at 48°C overnight in a heating block, which affords optimal extraction of sphingolipids due to their high phase transition temperatures ( 3 ). After cooling, 75 µl of 1 M KOH in CH 3 OH was added and, after brief sonication, incubated in a shaking water bath for 2 h at 37°C to cleave potentially interfering glycerolipids. After cooling to room temperature, approximately 3 to 6 µl of glacial acetic labeled version of each analyte of interest, but it is impractical for these to be used for the large numbers of com pounds examined in a lipidomic study. Therefore, an alternative is to identify internal standards that are similar in structure and ionization and fragmentation characteristics for the categories of compounds under investigation. This study evaluates an internal standard cocktail (now commercially available) developed for the LIPID MAPS Consortium ( 34 ) that contains uncommon chain-length sphingoid bases (C17) for sphingosine (So), sphinganine (Sa) and the 1-phosphates (S1P and Sa1P) and C12:0 fatty acid analogs of Cer (and, as discussed in the text, the cocktail initially also had C25:0-Cer), ceramide 1-phosphate (Cer1P), SM and mono-and dihexosylCer (HexCer and diHexCer). Examples are also given for how the types of analytes can be expanded by supplementation with additional internal standards. Because it is common in mass spectrometry for laboratories to have different categories of instruments available, this article describes the optimization and validation of the sphingolipid internal standard cocktail by electrospray tandem mass spectrometry on two types of instruments, a triple quadrupole mass spectrometer and a quadrupole linear-ion trap mass spectrometer, both operating in the triple quadrupole modes. The analytes of interest were identifi ed and quantifi ed by LC ESI-MS/MS using multiple reaction monitoring (MRM), a technique where the eluate is repetitively scanned for selected precursor-product ion pairs to enhance the sensitivity and specifi city of the analysis. The method is applicable to relatively small samples (e.g., a Petri dish of cells), as exemplifi ed by analysis of RAW264.7 cells, a mouse macrophage cell line.
The elution times for these analytes ( Fig. 2 ) are discussed under "Results." In cases where there is signifi cant carryover of Cer1P on this LC column (i.e., over 1%, which occurs with reverse phase columns obtained from some suppliers as well as with some lots of the columns described in this article), Cer1P can be analyzed instead using a Supleco 2.1 (i.d.) × 50 mm Discovery C8 column (Sigma, St. Louis, MO), with the column heated to 60°C and a binary solvent system [based on reference ( 38 )] at a fl ow rate of 0.6 ml/min. Prior to the injection, the column is equilibrated for 2 min with a solvent mixture of 70% Mobile phase altA1 (CH 3 OH/ H 2 O/THF/HCOOH, 68.5/28.5/2/1, v/v/v, with 5 mM ammonium formate) and 30% Mobile phase altB1 (CH 3 OH/THF/ HCOOH, 97/2/1, v/v/v, with 5 mM ammonium formate), and after sample injection (30 µL), the altA1/altB1 ratio is maintained at 70/30 for 0.4 min, followed by a linear gradient to 100% altB1 over 1.9 min, which is held at 100% altB1 for 5.3 min, followed by a 0.5 min wash of the column with 70:30 altA1/altB1 before the next run.
Ceramides, sphingomyelins, monohexosylceramides (HexCer) and dihexosylceramides (LacCer). These compounds were analyzed using the organic-phase extract and normal-phase LC using a Supelco 2.1 (i.d.) × 50 mm LC-NH 2 column at a fl ow rate of 1.0 ml/min and a binary solvent system as shown in Fig. 2C . Prior to injection, the column was equilibrated for 1.0 min with 100% Mobile phase A2 (CH 3 CN/CH 3 OH/HCOOH, 97/2/1, v/v/v, with 5 mM ammonium formate), and after sample injection, Mobile phase A2 was continued for 3 min, followed by a 1.0-min linear gradient to 100% Mobile phase B2 (CH 3 OH/H 2 O/HCOOH, 89/6/5, v/v/v, with 50 mM triethylammonium acetate), which was held for 3.0 min, then restored to 100% A2 by a 1.0-min linear gradient, and maintained at 100% A2 for 1 min to reequilibrate the column. In addition to these analytes, Cer1P and sulfatides (ST) can also be analyzed, however, their recoveries are higher in the single-phase extract, which can be reconstituted in the solvent for normal phase chromatography (Mobile phase A2). The elution times for these analytes are discussed under "Results." Resolution of glucosylceramide (GlcCer) and galactosylceramide (Gal-Cer). If the biological sample contains both GlcCer and GalCer (the latter is less widely distributed, and only found in barely detectable amounts in RAW264.7 cells until stimulation with Kdo 2 -Lipid A, as described under "Results"), these can be resolved using a different normal phase column (Supelco 2.1 (i.d.) × 250 mm LC-Si) and an isocratic elution with Mobile phase A3 (CH 3 CN/CH 3 OH/H 3 CCOOH, 97/2/1, v/v/v, with 5 mM ammonium acetate) at 1.5 ml per min (this can be reduced to 0.75 ml/min if necessary for complete desolvation). After the column is preequilibrated for 1.0 min, the sample (dissolved in Mobile phase A3) is injected, and the column is isocratically eluted for 8 min. In most cases, the GlcCer and GalCer are separated by 0.5-1 min (as shown in Fig. 2D ), which should be confi rmed during the analysis by interspersing vials with these internal standards throughout the runs.
Other sphingolipids (sulfatides). Sulfatides were analyzed using the same normal phase chromatography as described for SM, Gl-cCer, etc. ( Fig. 2C ); however, prior to extraction, the samples were spiked with 500 pmol of C12-sulfatide (d18:1/C12-GalSulfate) (from Avanti Polar Lipids). The organic-phase extract can be used for a qualitative screen of whether or not sulfatides are present, but for quantitation, a separate run should be made using the single-phase extract because it has the higher recovery (over 50%). acid was added to bring the extract to neutral pH (checked with pH paper to ensure that the extract has been neutralized), 5 and a 0.4-ml aliquot was transferred to a new test tube to serve as the "single-phase extract" (which was centrifuged to remove the insoluble residue, the supernatant collected, the residue reextracted with 1 ml of methanol:CHCl 3 , 1:2, v:v , centrifuged, and the supernatants combined). To the remainder of the original extract was added 1 ml of CHCl 3 and 2 ml of H 2 O followed by gentle mixing then centrifugation using a table-top centrifuge, and the lower layer (the "organic-phase extract") was transferred to a new tube. The upper phase was extracted with an additional 1 ml of CHCl 3 , which was also added to the organic-phase extract.
The solvents were removed from the single-phase extract and the organic-phase extract using a Savant AES2000 Automatic Environmental Speed Vac. The dried residue was reconstituted in 0.3 ml of the appropriate mobile phase solvent for LC-MS/MS analysis (described herein and the summary in Fig. 1 ), sonicated for approximately 15 s, then transferred to a microcentrifuge tube and centrifuged at 14,000 to 16,000 g for several min (as needed) before transfer of the clear supernatant to the autoinjector vial for analysis. Fig. 1 are the types of LC columns and extraction conditions that were found to be optimal for analysis of different subcategories of sphingolipids that are often encountered in mammalian cells, such as the RAW264.7 cell line. This scheme illustrates, nonetheless, that the investigator has multiple options depending on the nature of the biological sample; for example, if a particular biological sample contains only GlcCer (as is the case for RAW264.7 cells under standard culture conditions), then a shorter amino-column can be used, but if GlcCer and Gal-Cer are both present, an additional run with a longer silica column will also be necessary to distinguish these isomers. Specifi c LC conditions and instrument parameters for specifi c analytes are described herein. For all methods, 0.03-0.05 ml were injected onto the column.

Summarized in
Sphingoid bases, sphingoid base 1-phosphates, and ceramide 1-phosphates. These compounds were analyzed using the single-phase extract because their recovery into the organic phase of a traditional lipid extraction can be variable. They were separated by reverse phase LC using a Supelco 2.1 (i.d.) × 50 mm Discovery C18 column (Sigma, St. Louis, MO) and a binary solvent system at a fl ow rate of 1.0 ml/min. If this fl ow rate does not afford complete desolvation (typically seen as a jagged elution profi le), the fl ow rate can be reduced and/or the ion source gas fl ow rate can be increased.
Prior to injection of the sample, the column was equilibrated for 0.4 min with a solvent mixture of 60% Mobile phase A1 (CH 3 OH/H 2 O/HCOOH, 58/41/1, v/v/v, with 5 mM ammonium formate) and 40% Mobile phase B1 (CH 3 OH/HCOOH, 99/1, v/v, with 5 mM ammonium formate), and after sample injection (typically 50 l), the A1/B1 ratio was maintained at 60/40 for 0.5 min, followed by a linear gradient to 100% B1 over 1.8 min, which was held at 100% B1 for 5.3 min, followed by a 0.5 min wash of the column with 60:40 A1/B1 before the next run. 5 This differs from previous reports from our laboratory where neutralization of the single phase extract was not necessary, however, we are now fi nding some degradation unless this is done. Whether this is due to contaminants in the test tubes, a shift in the temperature of the Speed Vac, etc., has not been ascertained.
for the product ion of choice. Mass spectrometer conditions can also be optimized by injecting the sample via a sample loop with the solvents present at elution time at the given fl ow rate to better mimic LC conditions. Once the settings were identifi ed (see Tables 1-4 ), the precursor/product ion pairs (based on the characteristic fragmentations shown in Fig. 1B ) were placed in MRM methods as described herein.

Survey of subspecies in biological samples for selection of the Multiple
Reaction Monitoring (MRM) parameters. As the analysis of sphingolipids in cell extracts was conducted in MRM mode where specifi c precursor-product pairs are monitored during the LC elution, it is necessary fi rst to scan a representative cell extract to determine which subspecies are present. This should be conducted over a wide enough range that sphingoid base and acyl-chain subspecies that are shorter (e.g., C16-sphingoid base and C14:0 fatty acyl) as well as longer (such as the C20-sphingoid base and C30fatty acyl-chain length ceramides, the latter being found in skin) would be detected. If tissues contain signifi cant amounts of subspecies that are outside the range described in this report, the investigator will need to determine the optimal parameters for the additional species using the appropriate standards.
To perform the subspecies survey, the extract was resuspended in 0.5 ml of CH 3 OH and infused into the 4000 QTrap (which was the more sensitive instrument) at 0.6 ml/h and a scan of precursors for m/z 184.4 was used to detect the subspecies of SM, and scans of the precursors for m/z 264.4 (d18:1 backbone) and 266.4 (d18:0 backbone) were performed over a wide range of collision energies (35-75 eV) to check for other subspecies of sphingolipids. This also detects subspecies with a 4-hydroxysphinganine (phytosphingosine, t18:0) backbone and/or an ␣ -hydroxy-fatty acid because these have a 16 amu higher precursor m/z and also

Mass spectrometric analysis of standards and biological samples
Two systems were used for these analyses: a Perkin Elmer Series 200 MicroPump system coupled to a PE Sciex API 3000 triple quadrupole (QQQ) mass spectrometer (Applied Biosystems, Foster City, CA), and a Shimadzu LC-10 AD VP binary pump system coupled to a Perkin Elmer Series 200 autoinjector coupled to a 4000 quadrupole linear-ion trap (QTrap) (Applied Biosystems, Foster City, CA) operating in a triple quadrupole mode. For both instruments, Q1 and Q3 were set to pass molecularly distinctive precursor and product ions (or a scan across multiple m/z in Q1 or Q3), using N 2 to collisionally induce dissociations in Q2 (which was offset from Q1 by 30-120 eV); the ion source temperature varied between 300 and 500°C depending on solvent composition and mass spectrometer.

Identifi cation of ionization and fragmentation parameters for standards and analytes.
As the LC conditions were established (described above), the optimal conditions for ionization and fragmentation were determined for each analyte, and standard curves were established. Each sphingolipid standard was dissolved at a concentration of 1-10 pmol/µL in the appropriate solvent mixture in which it elutes from the column and infused into the ion source at a rate of 0.6 ml/h to determine the optimal ionization conditions by varying the declustering potential (DP) and focusing potential (FP) for the API3000, and the DP and entrance potential (EP) for the 4000 QTrap. After the Q1 settings were determined, product ion spectra were collected across a range of collision energies (CE), structurally specifi c product ions were identifi ed, and collision energies and collision cell exit potentials (CXP) were manipulated to produce optimal signal , half of the sample is used as the "single-phase extract" for analysis of the shown categories of sphingolipids and the other half is further extracted to obtain a lower (organic) phase for the other categories. For more details, see "Materials and Methods." B: Depiction of the major fragmentations of different categories of sphingolipids described in this article. For the sphingoid bases and ceramides depicted in the left two structures, -OR = -OH, -phosphate or mono-or di-hexosyl groups at position 1, and R and R ′ are the alkyl sidechains for the sphingoid base and fatty acid, respectively, which can vary in length and, to some extent, unsaturation. tion time when MRM pairs were being monitored for 8 analytes in a given period. Both Q1 and Q3 were set to unit resolution.

Standard curves for quantitative analysis of sphingolipids by LC-MS/ MS.
Once the LC and MRM protocols were selected, standard curves were determined under these conditions. The concentration of the individual components of the Avanti internal standard cocktail was 25 µM and the other sphingolipid subspecies were dissolved in methanol to produce stocks at 0.5 mg/ml. These were serially diluted into the appropriate LC solvent immediately before analysis to provide 0.5-1000 pmol of each standard per 50 µL injection. Each was then analyzed by LC-MS/MS in each of the mass spectrometers to generate the standard curves such as those shown in the fi gures and supplementary fi gures of produce a m/z 264.4 fragment (that can later be distinguished from the d18:1 species by LC mobility); to check for other sphingoid base backbones, the product m/z is varied by +/ Ϫ 14 amu for likely homologs (e.g., Ϫ 14 for 17:1 and +28 for 20:1) and by Ϫ 2 amu for additional unsaturation (e.g., d18:2). The results from these surveys are used to construct the MRM protocol for analysis of the cell extracts by LC-MS/MS (see Tables 1 -4 ) in which Q1 and Q3 were set to pass the precursor and most abundant, structurally specifi c product ion for each sphingolipid subspecies (for example, m/z 538.  ) to integrate both internal standard and analyte allowed quantitation to obtain the areas under the curves, then the pmol of the analyte was calculated using the following formula: pmol of analyte of interest = K analyte × (A analyte / A IS ) × pmol of added internal standard where K analyte = correction factor for the analyte versus the internal standard; A analyte = area of the analyte; and A IS = area of the added internal standard. The K analyte factor adjusts for differences between the analyte and the internal standard with respect to ion yield per this article, then to calculate the linear regression lines and fi t in Tables 1-4 .

Quantitative analysis of sphingolipids in cell extracts by LC-MS/MS.
To quantify the amounts of the sphingolipid analytes of interest in cell extracts, the elution profi les for each MRM pair were examined. The areas under the peaks generated for both analytes and internal standards can then be integrated with the mass spectrometer software (i.e., Analyst 1.4.2 for both Applied Biosystems instruments). Using identical integration settings (number of  due from the single-phase extract and the aqueous layer from the organic-phase extract were reextracted three times using the above method, and each extract was analyzed by LC-MS/MS. To validate the extraction protocol, ‫ف‬ 3.0 × 10 6 RAW 264.7 cells (N = 6) were spiked with internal standards and extracted, with repeated reextraction of the residue at the sphingoid base step and with repeated aqueous reextracts of the organic layer at the complex sphingolipid step. Using the optimized protocol, the reproducibility of the method was determined by analyzing three separate series of cultured RAW264.7 cells (six dishes prepared separately on three different days) and the coeffi cient of variation (CV) was calculated as the SD divided by the mean × 100.

Quality control
For each LC analysis, vials with the internal standards alone were analyzed at the beginning, middle, and end of the run. In addition, blank samples (containing only the LC solvent) were analyzed at varying intervals throughout the run to assess possible carryover. If carryover or shifts in the LC retention times for any of the analytes or standards were noticed, the column was cleaned before resuming the run. As noted above, some columns had an unacceptable level of carryover of Cer1P, and when that was the case, that analyte was reanalyzed using the alternative method. unit amount for the selected MRM pair, which also includes any correction for differences in isotopic abundance ( ‫ف‬ 1.1% per carbon), which are insignifi cant for analytes with alkyl chain lengths similar to the internal standard but become more substantial for very-longchain species. If the chain length is outside the number that have been corrected thus (i.e., for which there is an internal standard that can be used to determine the correction factor empirically), one can adjust the difference in isotopic distribution from the nearest standard using the ratio of (M+H) + , (M+H+1) + , and (M+H+2) + for the number of carbons in the analyte versus the internal standard using Analyst 1.4.2 or an online tool such as http://www2.sisweb. com/mstools/isotope.htm.
No correction was necessary for differences in extraction recovery because the extraction conditions were chosen to achieve similar recoveries, as described herein. Following the LIPID MAPS convention, the quantities of the sphingolipids are expressed as pmol/ g of DNA. For comparison, it has been our experience that 1 × 10 6 RAW 264.7 cells contain approximately 3 g of DNA and approximately 0.25 mg of protein.

Optimization of sphingolipid recovery from cells in culture
To determine the number of extraction steps necessary for recovery of the sphingolipids from RAW264.7 cells, the solid resi- There were some interesting differences between the instruments. Most notably, in the 4000 QTrap, the single dehydration product [( m/z 284.3 ( Fig. 3A )] and a small fragment with m/z 60.1 (C 2 H 6 NO from cleavage of the C2-C3 bond) were abundant ions for sphinganine [and the single dehydration product was also abundant for sphingosine, m/z 282.3 ( Fig. 3B )]; whereas in the API 3000, sphingosine and sphinganine yielded primarily m/z 264.4 and 266.4 fragments ( 7,15,34 ). The C2-C3 bond cleavage product is useful for identifi cation of 1-deoxysphingoid bases (i.e., to establish that they are 1-deoxy-and not 3-deoxycompounds), which have been found in various biological systems ( 39 ), because they are cleaved to m/z 44.1 (data not shown).
There were also differences in sensitivity that are discussed later, however, the greater sensitivity of the 4000 QTrap in negative ionization mode allowed analysis of Cer1P in both positive (supplementary Fig. V-A ) and negative (supplementary Fig. V-B ) ionization mode. Sulfatides were also analyzed in negative ion mode, with loss of sulfate ( m/z 96.8) as the major product ion (supplementary Fig. V ).
These fragmentation characteristics were used to select the MRM pairs used to optimize the LC and mass spectrometer settings in Tables 1-4 .

Optimization of LC ESI-MS/MS conditions
Using an initial set of ionization parameters for the MRM pairs for the selected sphingolipids, the LC conditions shown in Fig. 2 were devised, then the ionization parameters were optimized to ensure that they were appropriate for the electrospray solvent composition of the LC eluate. Standard curves for each analyte category were then determined by LC-MS/MS with the results described herein.
Sphingoid bases and sphingoid base 1-phosphates. Using the LC conditions for analysis of these compounds ( Fig. 2 ), the ionization and CID conditions for the internal standards and analytes were optimized. The results are summarized in Tables 1 and 2 . The highest recoveries of these compounds were obtained in the single-phase extract fraction (described herein); therefore, it was consistently used for analysis of these compounds (although the lower phase organic solvent extract could also be used, with this caveat regarding the lower recovery).
Graphs of the integrated ion intensities from these MRM analyses versus amounts for the sphingoid bases and sphingoid base 1-phosphates (both the naturally occurring d18 and the internal standard d17) species are shown for the API 3000 and 4000 QTrap ( Fig. 4 ) . Using both instruments, all of the species showed a linear signal response of 0.5-1000 pmol, and the d17 and d18 species behaved similarly. The major difference in signal response was for saturated and unsaturated species (e.g., d18:1 versus d18:0), which confi rms the need for internal standards for each. The 4000 QTrap was approximately 3-4 orders of magnitude more sensitive for analysis of these compounds than the API 3000.

RESULTS
The goal of these experiments was to test the utility of a commercially available internal standard cocktail for the quantitation of cellular sphingolipids using LC-MS/MS and two types of instruments, an API 3000 triple quadrupole (QQQ) mass spectrometer and a 4000 quadrupole linear-ion trap (QTrap) mass spectrometer in triple quadrupole mode. To achieve this objective, many of the LC-MS/MS conditions from previous publications ( 15,34 ) were used; however, a number of minor modifi cations were identifi ed during these studies, such as the observation that carryover of Cer1P occurs on some reverse phase columns.
The sections below describe the order in which each category of internal standard and analyte was evaluated, starting with identifi cation of the fragmentation profi les for selection of the precursor-product pairs, optimization of the parameters for LC-MS/MS, and application of the methods to cells, with the additional tests for extraction effi ciencies.

Identifi cation of major ions and fragments for LIPID MAPS™ internal standards and related analytes
Syringe infusion was used to select precursor/product ion pairs for MRM based on the structure-specifi c ions that were of highest intensity under optimized conditions.. As noted previously ( 15,34 )  The 4000 QTrap was several orders of magnitude more sensitive than the API 3000 for these compounds. In addition, in the 4000 QTrap, the ion yields for DHCer were close ( ‫ف‬ 85%) to the corresponding Cer ( c.f. Fig. 5 , panels B and D) whereas the ion yields for Cer on the QQQ were 6-8 times higher than their respective dihydroCer (c.f. Fig. 5 , panels A and C), although this is somewhat diffi cult to discern due to the log scale. In the API 3000, there was also a slightly greater difference between the C12-Cer and C25-Cer (4%).
The similarity in signal response across all these chain lengths (C12:0-C25:0) and for both Cer and dihydroCer using the 4000 QTrap allows the C12:0 internal standard to be used for all of these species (with a small correction for the 15% difference noted above). The same is the case for the API 3000, but with a larger correction factor (6-to 8-fold for DHCer versus Cer, and 4% for C12-to C25Cer). The initial LIPID MAPS TM sphingolipid internal standard cocktail included C25-Cer to provide an empirical confi rmation of this chain-length difference within each run. However, during studies of the RAW264.7 cells, we noted a discrepancy in the amounts of C24:1-Cer (d18:1/C24:1) using the API 3000 versus the 4000 QTrap that was caused by in-source dehydration of C25-Cer. This was surprising because the precursor-product pair of the in-source dehydration product of C25-Cer ( m/z 646.9/264.4) does not correspond to the MRM pair for C24:1-Cer ( Tables 1-4 ), however, the M+2 13 C isotopologue of C25-Cer ( m/z 648.9/264.4) provides this match. This explanation was confi rmed by analysis of the internal standard alone and by reverse phase LC-MS/MS (data not shown). Even though this interference can be minimized by careful selection of the ionization and fragmentation parameters for the instrument (indeed, these must be optimized for maximal sensitivity for the cellular ceramides), if other investigators fi nd that this artifact is produced in their mass spectrometer, they should use an internal standard cocktail mix without C25-Cer. This mix is now available from Ceramides. These compounds were analyzed using the lower phase organic solvent extract (information about extraction recoveries are described herein). Graphs of the integrated ion intensities for these analytes and standards versus amounts (0.5-1000 pmol) are shown in Fig. 5 for Cer (d18:1) with N-acyl chain lengths from C12:0 to C25:0 (panels A and B) and for dihydroCer (d18:0) with C16 to C24 (panels C and D). Using these LC conditions ( Fig. 2 ) and optimized ionization parameters ( Tables 3 and 4 ), all of the subspecies displayed a linear signal response of 0.5-1000 pmol on both instruments, with little difference in cps due to the length of the fatty acyl-chain.   Fig. VIII) . All of the subspecies displayed a linear signal response of 0.5-1000 pmol on both instruments. There was little difference in cps due to the length of the fatty acyl-chain. Overall signal for GlcCer was approximately 2.5 orders of magnitude higher on the 4000 QTrap versus the API 3000, and little correction is needed for the dihydro backbone in the 4000 QTrap, whereas for the API 3000, an 8-fold correction factor is needed (supplementary Fig. VIII ).
Comparisons of standard GlcCer and GalCer with comparable backbones did not reveal any differences in ion yield (data not shown); therefore, the C12-GlcCer internal standard is adequate for analysis of total HexCer (i.e., where the stereochemistry of the carbohydrate is not specifi ed). This might not be the case, however, for other backbones, such as ones with ␣ -hydroxy-fatty acids. Many cells do not contain galacto-family sphingolipids, so it is not necessary to distinguish GlcCer and GalCer in many instances, although the ability to do so using the modifi ed LC conditions described under " Materials and Methods" (see Fig. 2D ) allows this analysis when needed. Since Gal-Cer may be present in trace amounts compared with GlcCer, the separation of these compounds by LC should have greater sensitivity than a shotgun approach that distinguishes them in mixtures by only differences in the peak intensity ratio of the product ions at m/z 179 and 89 ( 31 ).
Dihexosylceramides (lactosylceramides). These compounds were analyzed using the lower phase organic solvent extract (information about extraction recoveries are de-Avanti Polar Lipids as LIPID MAPS TM sphingolipid internal standard Mix II (Catalog number, LM-6005).
Sphingomyelins. These compounds were analyzed using the lower phase organic solvent extract (information about extraction recoveries are described herein). After identification of LC conditions for analysis of these compounds ( Fig. 2 ) and optimization of the ionization parameters ( Tables 3 and 4 ), SM (d18:1) and dihydroSM (d18:0) with N-acyl chain lengths of C12:0-C24:0 were analyzed (supplementary Fig. VII) . All of the subspecies displayed a linear signal response of 0.5-1000 pmol on both instruments, and there was little difference in cps due to the length of the fatty acyl-chain. The signal response for SM on the 4000 QTrap (supplementary. Fig. VII-B ) was approximately an order of magnitude higher than for the API 3000 (supplementary Fig. VII-A ). Because SM and dihydroSM are analyzed by loss of the phosphocholine headgroup rather than backbone cleavage, the cps for the saturated and unsaturated backbone was similar for both instruments.
Although the positive ion mode CID does not allow defi nitive determination of fatty acid and sphingoid base combinations of SM (i.e., they are only inferred once the biological material is found to have little or no sphingoid base chain length variants), we have noticed that one can utilize the trap functionalities of the 4000 QTrap in negative ion mode to produce cleavage products with backbone information ( Fig. 6 ). This is done by selecting the same precursor ion m/z for the fi rst and second product ion, then offsetting Q2 by 5-10 eV so that no fragmentation occurs in the collision cell. Precursors are then induced to fragment in the ion trap by application of an amplitude frequency, allowing observation of structurally indicative peaks. This pseudo-MS 3 enhances the backbone structure specifi c fragment ( m/z 449.4 in Fig. 6B ) versus the amount of this fragment in typical MS/MS mode (c.f. Fig. 6B versus Fig. 6A), and thus allows postrun analysis of the SM in a sample for structural verifi cation.
Since SM is much more abundant than any of the other sphingolipids, it was more prone to showing a loss of linearity in the ion intensity versus amount (data not shown). We did not ascertain if this is due to column overloading or ionization suppression; however, it was readily detected and rectifi ed by analyzing the lipid extracts in several different dilutions (e.g., full strength and as a 1/10 dilution of the LC mobile phase used for sample injection).
Monohexosylceramides. These compounds were analyzed using the lower phase organic solvent extract (information about extraction recoveries are described herein). After identifi cation of LC conditions 6 for analysis of these compounds ( Fig. 2 ) and optimization of the ionization parameters ( Tables 3 and 4 ), both GlcCer (d18:1) and dihy- Fig. 6. Comparison of sphingomyelin product ion by traditional negative ion mode MS/MS in the 4000 QTrap (A) and by "pseudo MS 3 " analysis (B). Panel A is the product ion scan for the major precursor ion, m/z 631.5 from the in-source demethylation of SM; panel B refl ects the results that are obtained by selecting the same precursor ion m/z for the fi rst and second product ion, then offsetting Q2 by 5-10 eV so that no fragmentation occurs in the collision cell. Precursors are then induced to fragment in the ion trap by application of an amplitude frequency, allowing observation of the peaks shown when scanned as a "pseudo MS 3 ." supplementation of the internal standard cocktail with the appropriate compounds, a commercially available C12sulfatide (C12-ST) was used to quantify the sulfatides in the RAW264.7 cells. The highest recoveries of these compounds were obtained in the single-phase extract fraction; therefore, it was consistently used for analysis of these compounds (although the lower phase organic solvent extract could also be used, with this caveat). The negatively charged sulfate product ion is highly abundant (supplementary Fig. V ). Comparison of the signal responses for sulfatides on the API 3000 and 4000 QTrap (supplementary Fig. XI) revealed that the latter was approximately one order of magnitude more sensitive than the former (c.f. supplementary Fig. XI, panels A and  B). Under optimized ionization conditions ( Tables 3 and  4 ), the signal response was almost identical for C12-, C16:0and C24:0-ST, and linear (0.5-1000 pmol) for both instruments. Similar comparisons with dihydro-ST and ␣ -hydroxy-fatty acid-containing ST were not possible due to lack of the necessary standards.

Analysis of sphingolipids in RAW 264.7 cells
RAW264.7 cells were used to test these methods and standards with cells in culture. First, the major molecular subspecies of the cells were determined (defi ned as variants in backbone composition that account for at least 1% of that category of sphingolipid), and the extraction and analysis effi ciencies of the endogenous analytes and internal standards were examined, then several independent cultures of the cells were analyzed and the results compared.
Identifi cation of sphingolipid subspecies in RAW264.7 cells. Before using MRM to quantify the sphingolipids in cells, the molecular subspecies were determined by precursor ion scans to ensure the protocol included all of the relevant MRM pairs. This was fi rst accomplished by syringe infusion of the sample dissolved in 1 ml of CH 3 ( Fig. 7B ). By these analyses, the major amide-linked fatty acids were C16:0, C18:0, C20:0, C22:0, C24:1, and C24:0, with very low amounts of C26:1 and C26:0 (and even lower amounts of C14:0 and C18:1, which were detectable but are not reported here because they were less than 1% of the total). The signals for C12 species were not detectable over background ( Fig. 7 ); therefore, this chain length could be used for internal standards.
Precursor ion scans for other backbone sphingoid bases (e.g., phytosphingosine and other chain length homologs) were also performed as described previously ( 15,34 ), 8 but scribed herein). After identifi cation of LC conditions for analysis of these compounds ( Fig. 2 ) and optimization of the ionization parameters ( Tables 3 and 4 ), both LacCer (d18:1) and dihydroLacCer (d18:0) with C16:0 and C24:0 acyl chains, as well as the C12-LacCer internal standard, were analyzed (supplementary Fig. IX) . All of the species displayed a linear signal response of 0.5-1000 pmol on both instruments. There was little difference in cps due to the length of the fatty acylchain. Therefore, the C12-internal standard can be used within this range of alkyl-chain lengths. Signal response for LacCer is the lowest of the sphingolipids observed in positive ion mode (supplementary Fig. IX ), with approximately an order of magnitude greater sensitivity in the 4000 QTrap than API 3000. When analyzed using the 4000 QTrap, the cps for dihydroLacCer (d18:0) was only 20% lower than for Lac-Cer (d18:1), but it was approximately 5-fold lower for the d18:0 versus d18:1 backbone species using the API 3000.
Ceramide 1-phosphates. The highest recoveries of these compounds were obtained in the single-phase extract (described herein), therefore, they were used for quantitative analysis. 7 The elution profi les for Cer-P on reverse phase and normal phase LC are shown in Fig. 2 . Reverse phase chromatography was selected because it was compatible with the high salt in the single-phase extract and positive ionization mode ESI-MS/MS gives the more structure-specifi c fragmentation (i.e., cleavage to the sphingoid base backbone ions versus loss of phosphate in negative ionization mode). Once the LC conditions were selected, ionization and CID conditions for the internal standards and analytes were optimized. The results are summarized in Tables 3 and 4 . Graphs of the integrated ion intensities from these MRM analyses versus amounts are shown in supplementary Fig. X for both Cer1P (d18:1) and dihydroCer1P (d18:0) with C16:0 and C24:0 acyl chains, compared with the C12-Cer1P internal standard. All species displayed a linear signal response of 0.5-1000 pmol on both instruments. There was little difference in cps due to the length of the fatty acyl-chain. Therefore, the C12-internal standard can be used with Cer1P within this range of alkylchain lengths. The 4000 QTrap is approximately an order of magnitude more sensitive than the API 3000 for analysis of Cer1P (supplementary Fig. X ).
Because we have sometimes encountered signifi cant (i.e., as high as 10%) carryover of Cer1P using some lots of the reverse-phase C18 chromatographic material, an LC method using another solid phase (C8) has been developed for those instances (see "Materials and Methods"). The LC conditions for complex sphingolipids ( Fig. 2C ) can also be used to analyze Cer1P, but the elution peaks are broad.
Analysis of additional sphingolipids by modifi cation of these protocols, as exemplifi ed by sulfatides. As an example of how it should be possible to analyze additional sphingolipids by 7 The lower phase organic solvent extract can also be used for screening or confi rmation since the low recovery (ca 10 to 20%) can be corrected using the internal standard, however, the most reliable results are obtained using the single-phase extract, where the recovery is over 90% as shown in Fig. 10). ysis of Cer, GlcCer, and SM using these methods were determined using three separate cultures of cells (with six individual dishes per culturing) and were 8 ± 4% for SM and somewhat higher (12 ± 5%) for the less abundant Cer and HexCer, except when the subspecies was below 1 pmol/ g DNA, which increased the CV as much as 2-fold) (supplementary Fig. XII). The CV for the sphingoid bases were 15-25% (also depending on abundance) (supplementary Fig. XIII), and for CerP, were as high as 50% (not shown), which might be due to the very low amounts of these metabolites or possibly dish-to-dish variation in the amounts, as this compound is thought to function mainly in cell signaling.
Amounts of sphingolipids in RAW264.7 cells. Fig. 10 summarizes all of the subspecies of sphingolipids that have been analyzed in RAW264.7 cells using the methods described in this article. The results have many interesting features, such as: a ) the relative amounts of the different sphingolipid subcategories, with SM being most abundant, followed by Cer and HexCer in about an order-of-magnitude lower amounts, then sphingosine and even smaller amounts of the other sphingoid bases and 1-phosphates, and Cer-1P; b ) the high proportion of C16-dihydroCer (d18:0/C16:0) versus C16-Cer (d18:1/C16:0) compared with the other chain length (DH)Cer in any of the sphingolipid subcategories (Cer, SM, and HexCer); c ) the noticeable differences in the (DH)Cer fatty acyl-chain lengths among the sphingolipid subcategories; d ) the higher amounts of sphingosine versus sphinganine, and relative to the 1-phosphates; e ) the presence of a substantial amount of an ␣ -hydroxy fatty acid (hC24:1) in the Cer backbone of sulfatides but none of the other lipid subcat-none were found to be present in signifi cant amounts (i.e., over 1%).
Other analytes can also be surveyed by such scans, such as sulfatides via precursor m/ z 96.9 scans. When applied to RAW264.7, no subspecies of sulfatide was signifi cantly above the background. However, when the RAW264.7 cells were activated with Kdo 2 -Lipid A for 24 h (following the LIPID MAPS protocol described on www.lipidmaps.org), the amounts were more substantial and included a Cer backbone with an ␣ -OH fatty acid, as described herein ( Fig. 10 ).
Analysis of the extraction effi ciencies. The use of an internal standard is based on the assumption that the behavior of the selected compound(s) will be similar not only for LC ESI-MS/MS (as has been shown to be the case for the compounds described herein) but also during the extraction procedures leading up to the analysis. Shown in Figs. 8 and 9 are the recoveries of the internal standards and endogenous sphingolipids in sequential reextractions of RAW264.7 cells as described in "Materials and Methods." To obtain the pmol shown, the integrated areas from the MRM analysis of each of the internal standards were compared with the areas for the internal standards analyzed directly (i.e., without extraction). It is evident that in essentially every case, over 90% of the internal standard spike (500 pmol) as well as each subspecies of analyte are recovered in the fi rst and second extracts (which are pooled in the standard extraction protocol), with very little being lost in subsequent extracts. Furthermore, there are no noteworthy differences in the behavior of the internal standards and the cellular sphingolipids, although it is impossible to exclude the possibility that some of the endogenous sphingolipids are trapped in aggregates that are inaccessible to the extraction solvents.
Estimation of the coeffi cient of variation for analysis of RAW264.7 cells. The coeffi cients of variation (CV) for anal- Fig. 8. Recovery of internal standards and cellular sphingoid bases and 1-phosphates at each cycle of extraction. Approximately 1 × 10 6 RAW264.7 cells (approximately 3 g DNA) were spiked with the internal standard cocktail (500 pmol each), extracted four times, and analyzed using LC-MS/MS on an 4000 QTrap mass spectrometer as described under "Materials and Methods." The amounts of the analytes in each extract were calculated using the MRM areas for the unknowns versus the areas for the internal standards injected directly (i.e., without extraction). The bars represent the mean ± SD for n = 6. be taken to ensure that none of the components are already present in the samples (for example, C17-sphingoid bases, which are found in some organisms-in which case one might obtain another standard, such as the C10-or C14-homolog, which is available from Matreya) ( 39 ) and that they are appropriate for any additional analytes beyond the ones validated in this study (for example, sphingolipids with "phyto"ceramide backbones). In addition, the amounts of the internal standard cocktail can be varied based on the analytes that are of greatest interest. In the case of this study, the internal standard cocktail was tailored to the most abundant analyte (SM); however, other choices can be made based on the experience of investigators with their particular application. 9 Since multiple reaction monitoring only provides information about the specifi c precursor-product pairs that have been selected, it is important to begin by identifying the molecular species that are present in each new biological material and to recheck the composition when the cells egories; and f ) the absence of detectable sulfatides in RAW264.7 cells until activation by Kdo 2 -Lipid A. It also warrants comment that reanalysis of the monohexosylCer to determine the amounts of GlcCer versus GalCer (using the normal phase conditions shown in Figs. 1 and 2 and described under "Materials and Methods") found that GlcCer accounted for approximately 90% (28.1 ± 1.9 pmol/ g DNA) versus 10% for GalCer (3.2 ± 0.6 pmol/ g DNA). The amounts of dihexosylCer were also low [(about half of HexCer (c.f. Fig. 9 )]. A full description (and interpretation) of the sphingolipid composition of these is outside the scope of this article.

DISCUSSION
This study has established methods for the quantitative analysis of all of the components of the sphingolipid metabolic pathway through backbone (Cer) and its initial metabolites (SM, GlcCer, GalCer, and Cer-P), including the sphingoid bases involved in sphingolipid turnover and cell signaling (sphingosine and S1P). A welcome fi nding was that a relatively small number of internal standards are needed to quantify these compounds using optimized LC ESI-MS/MS conditions, and as these are available commercially as a single cocktail, this should facilitate studies of this spectrum of metabolites. Of course, when using these internal standards for other applications, care should Fig. 9. Recovery of complex sphingolipids at each cycle of extraction. Approximately 1 × 10 6 RAW264.7 cells (approximately 3 g DNA) were spiked with the internal standard cocktail (500 pmol of each species), extracted four times, and analyzed using LC-MS/MS on a 4000 QTrap mass spectrometer as described under "Materials and Methods." The amounts of the analytes in each extract were calculated using the MRM areas for the unknowns versus the areas for the internal standards injected directly (i.e., without extraction). The bars represent the mean ± SD for n = 6. 9 We have examined cells that were spiked with 5, 50 or 500 pmol of the internal standards and the latter two resulted in the same analyte quantitations as are shown in Fig. 11; therefore, use of a lower amount of the internal standards (i.e., 50 pmol) might be preferable since this would place each internal standard within an order of magnitude of its respective analyte. One also has the option of formulating an internal standard cocktail with different amounts of the standards. MRM protocol, the potential for two or more compounds having overlapping precursor-product ion pairs becomes considerable. In addition, the overlap can occur with not just the nominal mass of the compounds but also with isotopologues (for example, the M +2 13 C isotopologue of d18:1 with d18:0) and ions that are produced by in-source degradation-and even the combination of both of these possibilities, as was found for C25-Cer in these studies. This argues strongly for use of a preMS separation method such as LC to reduce the likelihood of such artifacts as well as to confi rm the identity of the analytes.
LC has the additional advantages that it a ) reduces the complexity of the electrospray droplet, thereby decreasing the likelihood that it will be composed of compounds with greatly different gas phase basicity or acidity, which can suppress the ionization of the analyte(s) of interest ( 42 ); b ) separates salts in the biological extract from the compounds of interest; hence, it minimizes formation of multiple salt adducts and other sources of ionization suppression ( 43 ); c ) concentrates the analytes of interest in a small volume; and d ) with appropriate choice of LC conditions, separates isomers that might not be distinguished by MS or MS/MS alone ( 44,45 ).
In comparing the results with these mass spectrometers, several interesting differences were observed. For some of the compounds, the instruments produce different modes of fragmentation, such as the formation of m/z 60.1 as a signifi cant product from sphinganine in the 4000 QTrap. Signal response was also different for most of the com-are treated in ways that might alter the composition, as illustrated in these studies by the appearance of quantifiable amounts of sulfatides only after the RAW264.7 cells have been treated with Kdo 2 -Lipid A. It is fairly easy to survey the types of sphingoid base backbones by starting with precursor ion scans for m/z 264.4 and 266.4 as signatures for the sphingosine (d18:1) and sphinganine (d18:0), respectively ( 39 ). 10 The same principle is followed in scanning for sphingoid bases that differ in alkyl chain length (e.g., m/z 292.4 for eicosasphingosine, d20:1) and unsaturation (e.g., m/z 262.4 for sphingadienes that have been found in mammals and are more prevalent in plants) ( 40 ). "Phyto" (4-hydroxy-) sphingolipids often yield fragments with the same product m/z as sphingosines, but they can be distinguished by both their increased mass and retention time by LC ( 34 ). These precursor scans also provide information about the amide linked fatty acids, including special subcategories such as ␣ -hydroxy species ( 41 ). Once the sphingoid bases and N-acyl variants are known, the corresponding transitions can be incorporated into a new MRM protocol.
In optimizing the LC and mass spectrometer settings for analysis of large numbers of compounds using the selected Fig. 10. Quantitative analysis of sphingolipids and dihydrosphingolipids in RAW264.7 cells. The amounts of these lipids were measured by LC-MS/MS using the internal standard cocktail as described under "Materials and Methods." Shown are the means ± SE (n = 18) for three separate experiments with 6 dishes each. The upper insert shows the free sphingoid bases and 1-phosphates; the lower insert shows the amounts of sulfatides (ST) determined in an experiment where ST biosynthesis was induced by Kdo 2 -Lipid A, as described in the text. The N-acyl-chains of ST included hydroxy-24:1 (h24:1) and -24:0 (h24:0). 10 This is less true for SM since of the phosphocholine headgroup produces a fragment that does not distinguish the components of the lipid backbone, however, as described under "Results" the backbone composition can be determined using the 4000 QTrap in an atypical ion selection/fragmentation mode. pounds on the 4000 QTrap versus the API 3000, with the 4000 QTrap often being 3 to 4 orders of magnitude more sensitive than the API 3000. This increased sensitivity is most likely attributed to a combination of the orthogonal ion source and improved ion optics of the 4000 QTrap.
Although each new application of this methodology will require some degree of revalidation of these parameters, it is hoped that this provides a useful platform for lipidomics analysis of most, if not all, of the sphingolipids in the early steps of the metabolic pathway. The availability of an internal standard cocktail that is applicable to such a wide profi le of compounds will not only facilitate such studies but will help ensure that the results from different laboratories can be compared.