Dual Parallel Mass Spectrometers for Analysis of … Glycerophospholipid and Plasmalogen Molecular Species Wm. Craig Byrdwell* FQS, NCAUR, ARS, USDA 1815 N. University Street, Peoria,

Dual Parallel Mass Spectrometers for Analysis ofSphingolipid, Glycerophospholipid and PlasmalogenMolecular Species

Wm. Craig Byrdwell*FQS, NCAUR, ARS, USDA 1815 N. University Street, Peoria, Illinois 61604, USA

Analysis of phospholipids was performed using a liquid chromatographic separation with two massspectrometers in parallel providing electrospray ionization (ESI) and atmospheric pressure chemicalionization (APCI) data simultaneously from a triple quadrupole instrument and a single quadrupoleinstrument, respectively. The output from UV-Vis and evaporative light scattering detectors were alsoacquired by the two mass spectrometers, respectively, for four detectors overall. This arrangement was usedto identify and calculate area percents for molecular species of dihydrosphingomyelin (DHS) andsphingomyelin (SPM) in commercially available bovine brain SPM, in human plasma extract and in porcinelens extract. Molecular species of phosphatidylethanolamine and its plasmalogen, and phosphatidylcholineand its plasmalogen were identified and semi-quantitative analysis performed. Commercially availablebovine brain SPM was found to contain 11.5% DHS and 88.5% SPM. The only DHS molecular speciesidentified in human plasma was 16:0-DHS, at or below 1% of the sphingolipid content. Porcine lensmembranes were found to contain 14.4% DHS and 85.6% SPM. Other findings reported here include: (1)phospholipids were found to undergo dimerization in the electrospray source, giving masses representingcombinations of species present. (2) Triacylglycerols gave usable mass spectra under electrospray ionizationconditions, as well as under APCI-MS conditions. (3) Triacylglycerols gave ammonium adducts as basepeaks in their APCI mass spectra, which reduced fragmentation and increased the proportions of molecularions. (4) Mass spectra were obtained for phospholipids which underwent both protonation and sodiumadduct formation in different chromatographic runs. # 1998 John Wiley & Sons, Ltd. This paper wasprepared under the auspices of the US Government and it is therefore not subject to copyright in the US.

Received 12 December 1997; Revised 7 January 1998; Accepted 10 January 1998Rapid Commun. Mass Spectrom.12, 256–272 (1998)

Dihydrosphingomyelins (DHSs), which differ from sphin-gomyelins (SPMs) only by the absence of 4,5-unsaturationin the sphingoid backbone, are difficult to analyze becausethey usually occur chromatographically overlapped andindistinguishable from sphingomyelins. In mixtures, identi-fication of some DHS species has been previously reportedusing mass spectrometry (MS),1 but only saturated mole-cular species of DHS could be identified. Without priorchromatographic separation, only DHS species with nounsaturation in the acyl chains have unique masses, whichare two mass units higher than sphingomyelins. Reversed-phase (RP) high performance liquid chromatography(HPLC) has been used successfully for separation ofmolecular species of SPMs (and coeluting DHSs).2 How-ever, differentiation of DHSs from SPMs required that thefractions be collected, and then different analyses wereperformed after derivatization to identify (i) the fatty acylchain, and (ii) the sphingosine base. This method was veryeffective at elucidating the identities of the species, but waslabor-intensive and time consuming. A similar RP-HPLCseparation, but which utilized chemical ionization (CI) MSinstead of chemical analysis of the sphingolipids, identifiedmany SPM and DHS species, but the data still resulted insome ambiguities between SPM and DHS species.3 Andagain, the method utilized fraction collection followed byanalysis.

As technology has advanced, online liquid chromato-graphy/mass spectrometric analysis has been employed.Thermospray mass spectrometry following the RP-HPLCseparation of sphingomyelin did accomplish the identifica-tion of three minor DHS components in bovine brain SPM.4

This represented a significant advance in the analysis ofsphingolipids, but, since this was a RP-HPLC separation,prior phospholipid class separation was still required beforemolecular species could be separated and identified. A morerecent study5 also demonstrated the application of RP-HPLC/ thermospray MS to sphingomyelins, but in thisreport no data were given for dihydrosphingomyelins. Fastatom bombardment (mostly of unseparated mixtures),plasmaspray and other ionization methods have also beenapplied to phospholipids. Many of these applications havebeen recently reviewed,6,7 and are not discussed here sincethey shed no light on the differentiation of DHS from SPM.

One of the most common interface/ionization sources forLC/MS analysis of phospholipids, including sphingomye-lins, has become electrospray ionization (ESI).1,8–10 Be-cause most common phospholipids are naturally charged,they are particularly amenable to ESI-MS. However, moststudies employing ESI-MS have either focused on unsepa-rated mixtures, or have not included sphingomyelin in thesamples. Thus, overall in the literature, there is a paucity ofdata which demonstrate the separation and identification ofmolecular species of DHSs differentiated from SPMs. Werecently reported11 the application of an HPLC separationusing an amine column, adapted from a previous method,12

*Correspondence to: W. C. Byrdwell, FQS, NCAUR, ARS, USDA,1815 N. University Street, Peoria, IL 61604, USA.

CCC 0951–4198/98/050256–17 $17.50 # 1998 John Wiley & Sons, Ltd. This paper was prepared under the auspices of the US Government and it is therefore not subject to copyright in the US

RAPID COMMUNICATIONS IN MASS SPECTROMETRY, VOL.12, 256–272 (1998)

which allowed the partial separationof DHS speciesfromSPM.We thendemonstrated13 that this partial resolution ofthemixture of DHS speciesfrom SPMspecieswasentirelysufficient to allow differentiation of all DHS and SPMspecies by mass using either ESI-MS or atmosphericpressure chemical lonization (APCI)-MS. A recent studyprecededourreportof theuseof APCI-MS for phospholipidanalysisby presenting datafor dioleoylphosphatidylcholinestandard,8 but SPM wasnot resolved from other phospho-lipids, and DHS was not identified, if present. In ourinvestigation, we found that the reason for the partialseparation of DHSfrom SPMwasthatthelongchain(20:0–26:0)DHS species,which elutedfirst, wereseparatedfromthe short chain (14:0–18:0) DHS species, and the DHSspecies eluted before SPM species (which were alsoseparated into long and short chain species). The shortchain DHS speciesremained chromatographically over-lapped with the long chain SPM species, but thesewereeasily differentiated by mass.Using this separation, thecolumn allowedseparation by class,sothatseveraltypesofphospholipids wererepresented,andwithin eachclass,longchainspecieselutedbeforeshortchainspecies. TheESI-MSandAPCI-MS datawereseento becomplementaryto eachother, with ESI-MS providing only protonatedmolecules,while APCI-MS produced diagnostic fragments aswell assmallamountsof [M�H]� ions. In orderto accomplishbothtypes of ionization, separaterunswere performed13 andtheionization sourceswere changedbetween runs. All datawereobtained using a single quadrupole instrument.

The extension and improvement of the previouslydemonstrated methodology13 are reported here. In thecurrent study, two massspectrometersutilizing the sameeluentstreamin parallel areemployed. APCI-MSdatawereobtained usinga single quadrupole instrument, while ESI-MS data were obtainedusing a triple-quadrupole instru-ment. Thetriple quadrupoleinstrumentwasoperated in bothfull scanandMS/MS modes.Collision-induceddissociationof parent ions was performedon the ESI-MS instrumentusing an automatedprocedurewhich selectedthe mostabundant parent ions in the preceding scansfor use asprecursorsfor daughterion formation. Additional datawereobtained from UV-vis and evaporative light scatteringdetectors. Threetypes of tissuesphingolipids were exam-ined: humanplasma,bovine brain and porcine eye lens.Changes in the chromatographic performance led to someunexpectedresults.

EXPERIMENTA L

Materials

All solventsexcept iso-propanol (IPA) and water were ofHPLC quality and were purchased from Sigma-AldrichChemical (Milw aukee,WI, USA), or EM Science(CherryHill, NJ, USA) andwereusedwithout further purification.Water wasobtained from purification of housede-ionizedwaterusinga Mill iporepurificationsystem.A.C.S. reagentgrade isopropanol (IPA) (Fisher Chemical, Fairlawn, NJ,USA) wasuseduntil anomaliesbegan,atwhich time HPLCgrade IPA was used. Phospholipid standardswere fromAvanti Polar Lipids (Alabaster,AL, USA) and were usedwithout furtherpurification. Two setsof standard solutionsweremade: onebasedon approximateplasmaphospholipidproportions, referredto asthe Plasma Model solution, andone basedon approximate human lens phospholipid pro-

portions,referredto astheLensModelsolution.ThePlasmaModel solution contained19.9%bovine brain SPM, 5.0%eggyolk phosphatidylethanolamine(PE),and75.1%bovinebrain phosphatidylcholine (PC) in methanol. The LensModel solution contained69.4%bovine brain SPM, 5.3%bovine brain PC and 25.3% PE, madeup from the com-bination of bovine brain and egg PE, in methanol. Initi alsolutions were diluted 1:3 in methanol, so the injectedconcentrations were 8.44mg/mL and 8.36mg/mL of theLens Model andPlasmaModel, respectively.

Human plasma was extracted using the procedure ofFolch et al.14 Whole plasmalipid extractwasusedwithoutprior separationof lipid classes. Porcine globes wereobtained from a local slaughterhouse and the lenseswereexcisedwithin 24 hourspost-mortemandkeptunder liquidN2 until extraction. Five lenseswerehomogenizedusingablade homogenizer and were extractedaccording to themethodof Folchet al.14

Liq uid chromatography

HPLC was performedusing an LDC-4100-MS quaternarypump with membranedegasser (Thermo SeparationProd-ucts, Schaumburg, IL, USA) andan HP 1050autosampler(Hewlett-Packard, Wilmington, DE, USA). The LC wasoperated underthe control of the singlequadrupole instru-ment. The triple quadrupole instrument was started via a24V relayattachedto theautosampler to produce a contactclosure.Theautosamplerstartsignalstarteddataacquisitionon both mass spectrometers.The column wasan Adsorbo-sphere NH2, 25cm� 4.6mm, 5 mm particle size (AlltechAssociates,Deerfield, IL, USA). A gradientprogram wasused which was composed of: (a) 33% ammoniumhydroxide solution/isopropanol(40:60), (b) hexane/isopro-panol (40:60), (c) methanol/isopropanol (40:60), and (d)water/isopropanol(40:60).Thegradientwasasfollows: 5%A throughout; initial 68%B, 12% C and15%D heldfor 10minutes;rampto 52%B, 16%C,27%Dat 15 minutes,helduntil 25minutes;rampto 95%Cat45minutes,helduntil 60minutes;recycled to initial conditions at 75 minutes.Thisresulted in acompositionof 0.66%NH4OH throughout.Theeffluent flow rate was 0.85mL/min throughout.10mL ofeachsamplewasinjected.

The column eluent split was accomplished by use ofdifferent lengths and diameters of polyether ether ketone(PEEK) andcapillary tubingandtee-junctionsasfollows: aValco Tee (Valco Instruments, Houston, TX, USA) wasattachedto the outlet of the column. To the teesideof thejunction was attached a length (1.5 ft.) of 0.005 in. i.d.PEEK tubing which went directly to the UV-vis detector,theoutlet from theUV-visdetector wasconnectedvia 0.005in. i.d. PEEK tubing to the inlet of the evaporative lightscatteringdetector(ELSD). Thestraight-throughsideof thefirst Valco Tee was attachedto a second Valco Tee via ashortpieceof stainlesssteel0.005in. i.d. tubing,from eachof thetwo otheroutletsof thesecond tee,anequallength (3ft.) of 0.1mm (=0.0039 in.) i.d. deactivated fused silicacapillary tubing was attachedvia an adaptingferrule; onecapillary wasattachedto theAPCI-MS inlet, theotherwasattachedto the ESI-MS inlet.

The UV-vis detectorwas a Spectroflow 757 (AppliedBiosystems, Foster City, CA, USA) operated at a wave-length of 206nm. The ELSD was a model ELSD MKIII(Varex,Burtonsville, MD, USA). The drift tubewassetto140°C, and the nebulizer gas (nitrogen) was set to 3.0

# 1998JohnWiley & Sons,Ltd. Rapid Communicationsin MassSpectrometry, Vol. 12, 256–272(1998)

DUAL PARALLEL MS FOR ANALYSIS OF PHOSPHOLIPIDS 257

standardliters per minute and a pressureof 26.6 psi. TheUV-vis spectrophotometer was interfaced to the triplequadrupoleinstrumentfor dataacquisition,while theELSDwasinterfacedto the single quadrupoleinstrument.

MassSpectrometry

A FinniganMAT SSQ-710C(FinniganMAT, SanJose,CA,USA) was used for acquisition of APCI-MS data. AFinniganMAT TSQ 700 wasusedfor acquisition of ESI-MS data. Al l conditions used for both APCI and ESIionization sources were the same as recently described,13

except the ESI spray voltage was set to 5.5kV. Foracquisitionof daughter ion datausing ESI-MS,anautomaticacquisition program was written using the InstrumentControl Language (ICL) which is part of the FinniganMAT software. A brief description of the acquisitionprogram is given. During automatic acquisition, thecollision-induced dissociation (CID) gas was off duringfull scanmode, and turned on during daughter analysis.With the CID gasoff in full scanmode, signal level wasincreasedby a factorof �100.TheCID gaspressurevalvewassetto 2.0 mtorr, andrequiredseveral seconds to reachthisvalue,duringwhichMS/MSacquisitionbegan.Thefullscanparameterswereasfollows: scanrangem/z200to 950in 1.0 s; number of parentscansbeforethresholdtest= 9;parent offset=ÿ5.5; collision offset=ÿ10.0; daughteroffset=ÿ10.0; parent msmscorrection factor= 70. At thestart of each run, the MS/MS threshold value wasautomatically set to two thresholdincrements (1 200 000counts)abovetheinitial baseline.If thetotal signal detectedfor massesabovem/z600 (signal cutoff value) passedthethresholdlevel, then MS/MS acquisition began.The CIDgasvalveopened, andtheprogramdeterminedthemassesofthe threemostabundant parentsfrom the precedingscans.The parentmasseswere sorted high to low (since longerchains elutedfirst) andthe programobtained daughterionspectrafor thesemassesin sequence.The parametersusedfor daughterion analysiswereasfollows: scanrangem/z50to ‘parentmass� 25’ in 1.0 s;numberof daughterscansforeach mass= 5; parent offset=ÿ5.5; collision off-set=ÿ35.0; daughter offset=ÿ42.3; parentmsmscorrec-

tion factor= 5.All parameterswerefully adjustableduringarun, most by usingsoft buttons.

RESULTS AND DISCUSSION

Standards

An ELSD chromatogram and APCI-MS reconstructedionchromatogram for the Lens Model solution are shown inFig. 1(a) and 1(b). AUV-Vis chromatogramand ESI-MS/MS reconstructed ion chromatogram for the Lens ModelSolution are shown in Fig. 1(c) and 1(d). Figure 1(d)demonstratesthecyclingbetweenfull scananddaughterionmodesperformed by the automatic acquisition program.Similarly, Fig. 2(a)and2(b) represent ELSD andAPCI-MSchromatogramsobtainedfor thephospholipid mixturemadein theapproximateproportionsof human plasma, while Fig.2(c) and2(d) representthecorrespondingUV-Vi s andESI-MS chromatogramsfor the Plasma Model solution.Figure2(d) again demonstratestheautomatic cycling between fullscan and daughter ion modes performed by the triplequadrupole instrument.Thefour detectorsshoweddifferentsensitivitiesfor the various classesof phospholipids andexhibited different susceptibilities to backgroundchangescausedby the gradientrun, discussedbelow.Although Fig1(d) and 2(d) exhibit a high background level in thereconstructed ion chromatograms(RIC) obtained by ESI-MS, this arose primarily from column bleedat m/z425,sotheuppermassregion showeda muchbettersignalto noiseratio, demonstratedin Fig. 3. Although the thrust of thepresentreportis thedemonstrationof theutility of thedualmass spectrometer system for phospholipid molecularspeciesidentification, especiallyof dihydrosphingomyelins,extensivedetail is given for all phospholipids studied.

APCI massspectra of the three commercially availablephospholipids in the model mixtures are shownin Fig. 4.The first phospholipid which elutedfrom the column wasPE (PE plus PE plasmalogen).Figure 1 showsthat thesespecieseluted in two unresolved peaks. Figure 4(A) and4(B) representaveragemassspectra across the first andsecondunresolved peaks, respectively. Both spectrashowthat these speciesproduced mostly protonatedmolecules,

Figure 1. Chromatogramsfrom four detectorsof phospholipidstandardsin approximateproportionsto humanlensmembrane.(a)Evaporativelightscatteringdetector(ELSD) chromatogram,(b) reconstructed,or total, ion chromatogram(RIC) obtainedby APCI-MS, (c) UV-Vis detectorchromatogram(set to 206 nm) and(d) ESI-MS/MSreconstructedion chromatogram.ESI-MS/MSobtainedin automaticparent/daughtercyclingmode.Phospholipidmixtureof 25.3%egg/bovinebrainphosphatidylethanolamine5.3%bovinebrainphosphatidylcholineand69.4%bovinebrainsphingomyelin.All Chromatogramsobtainedsimultaneouslyusingaminecolumnwith gradientasgiven in experimental.

Rapid Communicationsin MassSpectrometry, Vol. 12, 256–272(1998) # 1998JohnWiley & Sons,Ltd.

258 DUAL PARALLEL MS FOR ANALYSIS OF PHOSPHOLIPIDS

with little fragmentation.Theprotonatedmoleculesallowedthe identification of the speciesin the first peak as PEplasmalogens,while the second peak contained primarilyPE, with some PE plasmalogen. The PE plasmalogenspeciesexhibited less fragmentationthan the PE species.Thelack of fragmentationof bothof thesephosphoethanol-aminesis in contrastto the APCI-MS spectra previouslyshown13 andto spectrashown belowfor PC.Theonly majordifferencein thestartingconditionsbetween this study andthat previously reported13 was that the NH4OH was notpremixed with the solventsin this study,but waskept asaseparate solvent channel andincorporated into the effluentin the mixing valve. Whether sucha small differencecanaccount for the unusual phenomenonobserved during thisstudy will require further investigation. Despite thedifferencesin the abundances,the fragmentswhich the PEdid produce were primarily diacylglycerol fragmentsidentical to thosereportedearlier for PCandshownbelow,and are the sameas those produced by triacylglycerols.15

The diacylglycerol fragmentsconfirmedthe identification

of the molecular species, while the differenceof m/z141between the fragmentsand the corresponding protonatedmoleculesconfirmed the presenceof the phosphoethanol-amineheadgroup.The massesgiven for the loss of headgroupfragments representthefully protonatedheadgroups.For PE molecules, this means oneproton at the phosphateoxygen,andone protonon thenitrogen to form the ionizedamine.For PCmolecules,themassesincludeoneprotononthe phosphate oxygen.When headgroup fragmentswereobserved in daughter ion spectra, they included anadditional hydrogen on the phosphate oxygen whichcleavedfrom theglycerolbackbone.Extractedion chroma-tograms(EICs)takenfrom theRIC in Fig.1(b)areshowninFig.5.Figure5(a)revealedthat,usingthiscolumn,24:n and22:n PE plasmalogen species eluted in the first peakbetween 5 and6 1/2 minutes,20:n specieselutedbetweenthe two peaks, and 18:n and 16:n specieseluted in thesecond peak.All PEspecieselutedin thesecond peakin thissample.Theareapercentageof themolecular speciesin eggPE (from the Human PlasmaModel solution), calculatedfrom the areas under peaks in EICs of the massesofprotonatedmoleculesobtained by APCI-MS, is given inTable 1. The EICs in Fig. 5(a) depict the seven mostabundant molecularspeciesof PE plasmalogen and PE inthe LensModel solution. The elution of thesespecieswasaffectedby therelative proportionsof thetwo classes,sotheappearance of the overall peak in PE from egg wasnarrower, and there was less difference in the elution ofPE plasmalogenversusPE.All PE plasmalogen molecularmasseswere calculatedbasedon a 16:1 ether-linked fattychain, the most common fatty chain for plasmalogens.Daughter ion spectra obtained from ESI-MS/MS didindicate the possible presenceof an 18:1 plasmalogenbackbone,but furtherdiscussion is beyondthescopeof thisiniti al report.

The PE plasmalogen/PE peak is seento exhibit verydifferentresponsesfrom thefour detectorsused.Figure1(b)illustrates that this peak was larger than expected in theAPCI-MS chromatogrambasedon its known concentrationin the Lens Model solution, and the quantitative analysisfrom APCI-MS datarevealedthatthesumof theareasof themolecularspeciesaccounted for 58.0%of thetotal areaforall phospholipids, though it was known to be present at25.3% by weight. In the Plasma Model solution, which

Figure 2. Chromatogramsfrom four detectorsof phospholipidstandardsin approximateproportionsto humanplasma:(a)ELSDchromatogram,(b)RIC obtainedby APCI-MS, (c) UV-Vis detectorchromatogramand (d) ESI-MS/MS reconstructedion chromatogram.ESI-MS/MS obtainedinautomaticparent/daughtercycling mode.Phospholipidmixture of 5.0%eggPE,75.1%bovinebrain PCand19.9%bovinebrain SPM.

Figure 3. FilteredESI-MS/MSextractedion chromatograms(EICs).All chromatogramsfiltered to displayonly scansobtainedin full scanmode. (a) Mass region from 700 to 850Da which containedmostphospholipidprotonatedmolecularion masses,(b) lower massregionfrom 700to 775Da which includedmassesof short-chain(14:0–18:0)protonatedmolecularions.(c) Uppermassregionfrom 775to 850Dawhich includedlong-chain(20:0–26:0)protonatedmolecularions.



contained5.0%PEby weight, theareapercentageanalysisby APCI-MS gavea composition of 11.7%. Thus, in bothsamples,thePEgaveapercentcompositionwhichwasnear2.3 times higher than the known composition.Within thisclass, however, the quantitation appeared to be morereliable. The LensModel solution, which containedmostlybovine brain PE, showed 48.7% plasmalogen and 51.3%PE,which is in agreementwith thenearly50%plasmalogenindicatedby thesupplier. TheeggPEin thePlasmaModelsolution gaveareapercent compositions of 12.7%plasma-logen and 87.3% PE. The Avanti Polar Lipids catalogreported that,usingtheir HPLC method,they foundthePEto be> 99% pure PE. This demonstratesthe difficulty inseparationof molecular speciesof plasmalogensapartfromthe parentclass.The UV-Vi s detectorcan be seenin Fig.

1(c) to give more responsefrom polyunsaturatedplasmalo-gensat thefront of thepeakthannormalPEat theback(seethe elution demonstratedfor this sample in Fig. 5(a). Thegreatdisparity betweenthediacyl phospholipids versusthesphingolipids, as well as the background drift, madequantitative analysis using the UV-Vi s detector fruitless.The ELSD detector proved to be better at yieldingpercentagecomposition data. The chromatogram in Fig.1(a) showed 15.8% PE,andthat in Fig. 2(a) showed 4.8%PE. However, useof the mass spectrometer’s datasystemfor integrationof ELSDchromatograms,althoughexcellentfor masschromatogram integration, produced inadequateresolution in the areaunits of ELSD chromatograms.

Basedonthespectraobtained for thePEplasmalogenandPE species, we can now conclusively identify the uni-

Figure 4. MassspectraobtainedusingAPCI-MS of phospholipidstandards.Bovine brain/eggPE mixture shownin Fig. 1: (a) front peakelutedbetween5.25to 5.5min, (b) secondPEpeakelutedat 6 to 6.5min (seealsoFig. 5(a)).BovinebrainPCandPCplasshownin Fig. 2: (c) front of PCpeakelutedfrom 17 to 17.5min, (d) sectionof PCpeakelutedfrom 18.5to 19min (seealsoFig. 5(b)).BovinebrainSPMshownin Fig. 1: (e) firstsphingolipidpeak(SL1) elutednear21min. Peakcomposedof long chaindihydrosphingomyelin(DHS,or D). (f) Secondsphingolipidpeak(SL2)elutedfrom 22 to 22.75min. Peakcomposedof short-chainDHS and long-chainSPM. (g) Third sphingolipidpeak(SL3) elutedfrom 22.75to23.75min. Peakcomposedof short-chainSPM(seealsoFig. 5(c)).



dentified phospholipid which wasreported earlier13 to havestructural features similar to phosphatidyl ethanolamineplasmalogenasdefinitely being composedof PEplasmalo-gen,aswell assome PE.The massspectrum shownearliermaynow be interpretedto showthat thepeakarosemostlyfrom PE plasmalogen having 20:1 (m/z589.6,704.6)and18:0 (m/z 563, 730.6) acyl chains. These identificationswerenot given conclusively in the previousreportbecauseearlier results16 from a commercially available PEstandardhad exhibited contradictory chromatographic behavior.However, in the first report,the eluentdid not contain anyNH4OH which wasaddedlater aselectrolyte for the ESI-MS. Suchadifferencecanchangetheionizationstateof thephospholipid, and causea changein the chromatographicbehavior on theaminecolumn.Thispossibility is supportedby the similarity between the peak broadening for PEplasmalogenseenin the first report16 to that shownin Fig.5(a).

Bovinebrainphosphatidyl cholineelutedasabroadpeak,with long-chain specieselutedbeforeshort-chain species, asshown in Fig.3 and5(b). MostPCplasmalogenselutedonlyslightly beforethe similar normal PC, but the short-chainspecieselutedbefore the long-chainspecies, andtheserantogether to form one broad peak,which appeared homo-geneous. The ESI-MS dataclearly showedthe presenceofPC plasmalogenalong with PC.While easily identified byESI-MS, quantitation was performed using APCI-MSfragments. Since ESI-MS is more sensitive than APCI-MS, more speciescould be qualitatively identified usingESI than were quantified using APCI, such as 24:nplasmalogen speciesand 40:n PC specieswhich wereobserved in EICs obtained by ESI-MS, but not quantified.Using APCI-MS data to give areapercentresults, it wasfound that bovine brain PC in the PlasmaModel solutiongave a total areapercentof 29.0% PC plasmalogen and

71.0% for diacyl PC. Figure 4(c) and 4(d) show thatabundant protonatedmolecules were apparentin the massspectra of PC plasmalogen and PC. As with the PEplasmalogen, thePCplasmalogendid notexhibitsignificantfragmentation, but ratherformedprotonatedmolecules.Thediacyl PC, conversely, formed abundantdiacylglycerolfragment ions. As with PE, these fragmentswere usedtoconfirm the identities of the acyl chains andthe differenceof m/z183 between these fragmentsandthe correspondingprotonated molecules confirmed the presence of thephosphocholineheadgroup(asdid thedaughterion spectraobtained by ESI-MS/MS). The percent composition of themolecularspeciesof bovinebrainPC,determinedby APCI-MS, is given in Table 1. Area percent calculationswereperformedusingtheprotonatedmoleculesfor PCplasmalo-gen,while the areas underthe peaksfor the diacylglycerolfragmentsand the protonatedmolecules were summedtodeterminethetotal areasfor PC,similar to themethodusedfor triacylglycerolquantification.Thephosphocholines,asaclass, showed less responsethan expectedbasedon thepercentby weight. ThePlasmaModel wascalculatedfromAPCI-MS datato contain1.3%,while theLensModel wascalculatedto contain62.5%. Thesesolutionswereknown tocontain 5.3% and 75.1% by weight, respectively. Thecompositions calculated from the ELSD chromatogramswere5.3%and76.2%, respectively,in goodagreementwithexpected results.It wasfortuitousthat, in thecases of bothPEandPCclasses,theplasmalogenmolecular massesfell inbetween the parent class molecular masses. This allowedfacile differentiationof the species. Only the most highlyunsaturated plasmalogen species (6 or more sites ofunsaturation) had massesthe sameas thoseof saturatedparent species. The highly unsaturated PE plasmalogenspecies, if they occurredat all, were chromatographicallyseparatedfrom their isobaric parents.However, in the case

Figure 5. Extractedion chromatogramsobtainedunderAPCI-MSconditions.(a) Phosphatidylethanolaminepeakseenin Fig. 1(b).Long-chainPEplasmalogen(plas)elutedfirst, followed by shorter-chainPEplasandall PEspecies.(b) BovinebrainPCandPCplaspeakseenin Fig. 2(b).Longchainspecieselutedbeforeshortchainspecies.(c) BovinebrainSPMpeakseenin Fig. 1(b). Long-chainDHS elutedasthe first peak,short-chainDHS andlong-chainSPMelutedasthesecondpeak,andshort-chainSPMelutedasthe third peak.



of PCplasmalogen,only partialchromatographic separationwas achieved and overlap occurredbetween thesehighlyunsaturatedplasmalogenswith their parentcompounds, iftheywerebothpresent. In thecasesof overlap(e.g.22:6PCplas= 36:0 PC), the ESI-MS chromatogramwas used tomake the assignment (areaassignedto 36:0 PC).For mostspeciesthis wasnot a concernandanalysiswasstraightfor-ward. Another potentially ambiguous overlap of masseswhich required careful interpretation was that, for PCspecies, therewasasmallpercentageof fragment formationwhich occurred from the replacement of a methyl groupfrom the choline headgroup with a hydrogen atom, for amassdifferenceof 14Da.Thispresentedthepossibility thata saturatedPC plasmalogen could be confusedwith a PCparent with one more site of unsaturation. In thesecases,althoughthespecieswereisobaric,thechromatography wassufficient to partially differentiate the contribution to aplasmalogenpeakfrom the parent, as shown in Fig. 5(b).However, because resolution of PC plasmalogen speciesfrom PCspecieswasnot complete,it is possible that somearea attributed to PC plasmalogen speciesarose fromdemethylated PC species. Similarly, a protonated PCmolecule in which all three choline methyl groupshavebeenreplaced by hydrogenatomsin theAPCI sourcewould

have the same massas a plasmalogen having two fewercarbons(e.g.(38:1PC- 42)= 20:0PCplas,m/z774.7). Thepossibility exists that the area attributed to 20:0 PCplasmalogen could arise from the tri-demethylatedparentPC. However, if this were the case all PC protonatedmoleculeswould give this fragmentwithout regard to theidentitiesof the fatty acyl chains,so the PC plasmalogenspecieswould be in a constantproportionto the PC parentions. This was not observed in the data.For instance,theratio of the 20:0 PC plasmalogen percentage to thepercentageof 38:1 PC is 7.2, while the ratio of 20:4 PCplas to 38:5 PC is 1.6. Most importantly, the ESI-MS/MSdaughterion spectraof 20:0 PC plasmalogen (m/z774.7)and18:0 PCplasmalogen (m/z746.6)exhibitedbasepeaksat m/z184,clearly indicating the intact (not demethylated)cholineheadgroup.

In the chromatogramsin Figs 1 and 2, the region overwhich SPM eluted (from 20 –25 minutes) showed threepeaks,asmallleadingpeakandtwo largerpeaks. As wasthecasewith human lens membranes, the first peak in thisregion represented the presenceof long-chain dihydro-sphingomyelins. The second peak represented long-chainsphingomyelins and short-chain dihydrosphingomyelins,andthe third peakrepresentedshort-chainsphingomyelins.

Table 1. Phospholipid molecular speciesarea percent for commercial phospholipid standards

PEfrom Egg PCfrom BovineBrain SPMfrom BovineBrain(HeadGroup�2H)� = 142amu (HeadGroup�H)� = 184amu (HeadGroup�H)� = 184amu

PEplas [M�2H]� % PE [M�2H]� % PCplas [M�H]� % PC [M�H]� % DHS [M�H]� % SPM [M�H]� %

24:0 788.7 0.4 40:0 804.7 0.2 24:0 830.7 0.5 40:0 846.7 0.4 26:0 845.8 0.0 26:0 843.8 0.224:1 786.7 1.9 40:1 802.7 0.1 24:1 828.7 0.3 40:1 844.7 0.2 26:1 843.8 1.1 26:1 841.7 0.924:2 784.7 2.3 40:2 800.6 0.0 24:2 826.7 0.3 40:2 842.7 0.2 26:2 841.7 0.0 26:2 839.7 0.124:3 782.7 3.7 40:3 796.6 0.1 24:3 824.7 0.4 40:3 840.7 0.3 26:3 839.7 0.0 26:3 837.7 0.024:4 780.7 5.0 40:4 796.6 0.7 24:4 822.7 0.4 40:4 838.7 0.6 26:5 835.7 1.0 26:4 835.7 0.024:5 778.7 6.0 40:5 794.6 2.5 24:5 820.7 0.6 40:5 836.6 0.7 24:0 817.7 10.5 26:5 833.7 0.122:0 760.6 0.6 40:6 792.6 1.4 22:0 802.7 0.3 40:6 834.6 1.9 24:1 815.7 16.6 26:6 831.7 0.122:1 758.6 2.0 40:7 790.6 0.3 22:1 800.7 0.4 40:7 832.6 1.0 24:2 813.7 1.0 24:0 815.7 5.722:2 756.6 1.8 38:0 776.6 0.5 22:2 798.7 0.5 38:0 818.7 0.4 24:3 811.7 0.3 24:1 813.7 14.522:3 754.6 3.7 38:1 774.6 0.2 22:3 796.7 2.0 38:1 816.7 0.6 24:4 809.7 0.2 24:2 811.7 0.822:4 752.6 6.6 38:2 772.6 0.3 22:4 794.6 1.8 38:2 814.7 0.9 24:5 807.7 1.6 24:3 809.7 0.122:5 750.6 7.5 38:3 770.6 3.7 22:5 792.6 3.7 38:3 812.6 1.0 22:0 789.7 14.7 24:4 807.7 0.120:0 732.6 7.5 38:4 768.6 19.9 20:0 774.7 4.1 38:4 810.6 3.9 22:1 787.7 3.6 24:5 805.7 0.120:1 730.6 15.5 38:5 766.6 6.1 20:1 772.6 2.2 38:5 808.6 2.4 22:2 785.7 0.0 22:0 787.7 5.420:2 728.6 11.8 38:6 764.5 3.0 20:2 770.6 2.0 38:6 806.6 2.9 22:5 779.7 1.2 22:1 785.7 1.920:3 726.6 2.4 38:7 762.5 0.2 20:3 768.6 7.5 38:7 804.6 0.6 20:0 761.7 9.9 22:2 783.7 0.020:4 724.6 1.3 36:0 748.6 1.5 20:4 766.6 3.8 36:0 790.7 2.4 20:1 759.7 0.6 22:3 781.7 0.018:0 704.6 4.8 36:1 746.6 11.8 20:5 764.6 5.1 36:1 788.6 16.3 20:5 751.6 3.5 22:4 779.7 0.018:1 702.6 10.8 36:2 744.6 14.3 18:0 746.6 19.7 36:2 786.6 6.9 18:0 733.6 30.9 20:0 759.7 11.218:2 700.6 3.9 36:3 742.6 3.0 18:1 744.6 6.5 36:3 784.6 1.7 18:1 731.6 1.4 20:1 757.6 0.618:3 698.5 0.0 36:4 740.5 6.2 18:2 742.6 1.6 36:4 782.6 3.6 18:2 729.6 0.4 20:2 755.6 0.216:0 676.6 0.4 36:5 738.5 0.2 18:3 740.6 5.0 36:5 780.6 0.6 18:3 727.6 0.4 20:3 753.6 0.016:1 674.5 0.0 36:6 736.5 0.2 16:0 718.6 27.2 36:6 778.6 0.3 16:0 705.6 1.3 20:4 751.6 0.0

36:7 734.5 0.0 16:1 716.6 1.8 36:7 776.6 0.6 16:1 703.6 0.0 20:5 749.6 0.5Sum 100.0 34:0 720.6 1.3 14:0 690.6 2.2 34:0 762.6 5.1 18:0 731.6 52.0

34:1 718.6 13.9 34:1 760.6 31.8 Sum 100. 18:1 729.6 2.434:2 716.5 7.4 Sum 99.9 34:2 758.6 4.3 18:2 727.6 0.834:3 714.5 0.3 34:3 756.6 0.3 18:3 725.6 0.332:0 692.5 0.0 32:0 734.6 3.7 16:0 703.6 2.032:1 690.5 0.2 32:1 732.6 3.4 16:1 701.6 0.032:2 688.5 0.2 32:2 730.6 0.4 14:0 675.6 0.132:3 686.5 0.0 30:0 706.6 0.2

Sum 100.1Sum 99.7 Sum 99.6

Percentof phosphoethanolamines Percentof phosphocholines Percentof sphingolipidsPEplas= 12.7PE= 87.3 PCplas= 29.0PC= 71.0 DHS= 11.5SPM= 88.5



The presenceof thesespeciesis demonstrated in Fig. 3which representsthefiltering of theESI-MS chromatogramin Fig. 2(d) to showonly scansobtained in full scanmode.The massregion from m/z 700 to 850 (Fig. 3(a)), whichcontainedall protonatedmolecules for PC,DHS andSPMspecies, was subdivided into regionsfrom m/z700 to 775(Fig. 3(b)), which contained massesof short-chain species,andm/z775 to 850 (Fig. 3(c)), which containedmassesoflong-chain species. Between 20 and 25 minutes in thechromatogramsin Fig. 3 two peaksin theshort-chain massregion (Fig. 3(b)) and two peaksin the long-chain massregion(Fig. 3(c)) mayclearlybeseen.Thefirst peakin thisregion in eachchromatogram representsDHS, while thesecond peakrepresentsSPM.Figure3 also demonstratestheelution of long-chain andshortchain PCandPCplasmalo-genbetween 15and20minutes.Figure5(c)showsEICsforthemost abundantDHSandSPMspecies. Sincethesamplecontained about 8 times as much SPM as DHS, thecontributions from isotopepeaksof SPM may be seeninchromatogramsof completelysaturatedDHSspecies. Also,unsaturatedDHS specieswere isobaric with SPM specieshavingonelesssiteof unsaturation in theamide-linked acylchain (e.g., 24:1-DHS= 24:0-SPM). However, the DHSspecies were completely chromatographically separatedfrom their isobaric SPM counterparts so no ambiguityarose. The integrated areas underextracted ion chromato-grams indicated that bovine brain sphingomyelin wascomposedof 11.5%DHS and88.5%SPM. The molecularspeciescompositions of these sphingolipids is given inTable1. Themajor speciesin DHS andSPMweresimilar,with 18:0, 24:1, 22:0 and 20:0 species predominant.However, thereweresubstantial differences in the relativeproportionsof speciessuchas22:0and18:0.Because DHShaspreviously beendifficult to identify, the importance ofsuchcompositionaldifferencesbetween thesesphingolipidshasnot beenaddressedat all. It hasbeenfound that DHSleads to a more ordered membranestructure,17 but theimportanceof this finding, and the impact that differentmolecular specieshaveon membraneorder, is still to bedetermined.It is interesting to note thateggyolk SPMwasfoundto contain 8.2%DHS, but this wascomposedonly ofshort-chainspecies, havingnoneof thelonger-chainspeciesshown to bepresent in bovine brain SPM.

Massspectra of the first, secondand third sphingolipidpeaksare shown in Fig. 4(e)–(g) respectively. The massspectrapresented hereare similar to the spectrafor thesecompoundswhich werepreviously reported,13 but therearedifferences.First is thepresencein thesespectraof adductsformed from the addition of 106Da to the diacylglycerolfragments, with another adductat 18Da more,or 124 Da.As mentionedabove,the primary differencein the LC/MSmethodappliedpreviously versusthat usedfor this studywasthattheNH4OH wasnotpre-mixedwith thesolventsinthis chromatographic system. Whether this differenceproduced the undesirable side effect of adduct formationwill have to be further explored using APCI-MS/MS.Anotherdifferencein thespectrawasthelossof 14Dafromthe protonatedmolecules, assumedto be replacementof amethyl moiety from the cholineheadgroupby a hydrogenatom,asmentionedabovefor PC.Thismechanismappearedto occurto a largerextentwith sphingolipidsthanwith PC.Thelast differenceobservedin thesespectrawasafragmentarising from thelossof 18Da from thesphingoid fragment.Thisoccurredfrom lossof thehydroxy groupin theform ofdehydration.Lossof thehydroxy groupby dehydrationhas

been reportedas the primary fragmentation pathway forhydroxy-containing triacylglycerols during APCI-MS.18

However,the primary fragments in thesemassspectrastilldefinitively identifiedtheDHSandSPMmolecular species,andwereusedfor semi-quantitative analysis.

The chromatographicseparation using theaminecolumnis thekeydifferencewhichallowedusto report thepresenceof numerousintactDHS speciesin bovinebrainandin eggyolk. Sinceall specieselutedoverashorttimeperiod,it waseasier to identify speciespresentat low levels. Forexample,26:5-DHS,presentas1.0%of DHS, represented67.3ng,or80.6 pmol,basedon the10mL of sample injected.We wereunable, however, to confirm thepresenceof a sphingolipidcontaining a 20:1 sphingoid backbonewhich has beenreported in bovine brain. Jungalwala et al. reported2 theidentification of a20:1sphingolipidasthebiphenylcarbonylderivativeof onefractionfrom areversed-phaseseparation.The 20:1 sphingolipid was reportedto occur only with an18:0fatty acid.This would havea massisobaric with 20:0-SPM, sothesphingoidfragmentwould appearat m/z576.6.Since 18:0 and 18:1 sphingolipids are separated soeffectively using the amine column, if the 20:1 werepresent, it would beexpectedto bedifferentiable. Theonlypeakapparent in the masschromatogram for m/z576.6inFig. 5(C) is that for 18:0-SPM. On theotherhand,we wereable to identify more polyunsaturated DHS and SPMspeciesthanhavebeenreported, especially speciescontain-ing 5 sitesof unsaturation. Thesegive peaksin their EICswhich have longer retention times than their saturatedhomologs. The chromatographic data provide valuableconfirmation of the mass data to support the presenceofthese species.

Most daughter ion spectraof phosphocholine-containingphospholipids(PC, PC plas, DHS and SPM) obtained bypositive ion ESI-MS/MS showed only a basepeakat m/z184 arising from the intact fully protonated choline headgroup,andsomefragmentsof theheadgroup,similar to theheadgroupregionin thedaughterspectrumshown for Lyso-PC, below. In some spectra, small amounts of usefulfragmentation wereobserved, but mostoften the daughterspectra were simple and showed only phosphocholinefragments. Plasmalogens produced more useful spectrathan the diacylglycerol compounds. Plasmalogens moreoften gave fragments useful for identification of the acylchains along with the head group fragments. Usefulfragmentation wasonly achievedastheCID gasapproached2.0 mtorr. Before the collision cell pressure rose to thislevel, only headgroupspectrawereobserved.Sphingolipidsproduced virtually only headgroup fragments in daughterion spectra.Giventhelargenumberof experimentspossibleusing dualparallel massspectrometers, useof negative ionMS/MS modefor identification of RCOOÿ fragmentsis anobviousextensionof the demonstratedmethodology.

Human plasmaextract

Figure 6 shows ELSD and APCI-MS chromatograms ofhuman plasmawhole extract. Because of the magnitudeofthe peakelutedbetween 3 and4 min, chromatogramswithexpandedvertical scalearealso given.Figure6(e)and6(f)show the UV-Vis and ESI-MS chromatogramsobtainedduring the same run. Figure 7(e) shows the ESI-MSchromatogramfor the massrange between m/z 700 and850, the rangecontaining the protonatedmoleculesof PC,PCplas,DHSandSPMspecies. Thechromatogramsin Fig.



7(a)–(d)representschromatogramswhichhavebeenfilteredto show only data obtained in full scan mode. For thepurposes of this study, the most important observationarising from thesedatais thatin thetimebetween 20and25minutes,when DHS andSPMeluted,only onepeakis seenin the low massrange(Fig. 7(b)), associated with short-chain DHS andSPMspecies, andonly onepeakis seeninthehigher massrange(Fig. 7(c)), associatedwith long-chainDHS and SPM species. This indicates that humanplasmacontains virtually no dihydrosphingomyelin. Examination

of EICs for DHS speciesindicatedthat a small amount of16:0-DHScould be identified, and this was at a very lowlevel. Semi-quantitative analysis, presented in Table 2,showstheareapercentagesof themolecular speciesof SPMandindicatesthat16:0-DHS waspresentat a level of about1% of total sphingolipid. Small amountsof other DHSspeciesmaybepresentwhich arebelowthedetectionlimitin the whole extractsample.Nevertheless, becauseof theclearseparationof the isobaric DHS andSPMspecies, thiscolumn allowed unambiguousconfirmation of the generallack of DHS in plasma, just as it indicatedits presenceinothersamples.Previously, failure to identify DHS did notmeanit wasnot present,ashasbeenshownusingtheaminecolumn for analysis of numerous speciesof DHS whichwerenot previously identified. Usingthis method,chroma-tographicand massdataboth support the lack of DHS inplasma.

As in brainPC,Fig. 7(b) and(c) indicate thepresenceoflong andshortchain PC plasmalogenandPC species. Theareapercentcompositionof thesespeciesis given in Table2. In Fig. 7(a), another peakmay be seento elutenear29minutes,havinglower massesthantheotherphospholipidsdiscussed thusfar. Daughter ion spectraobtainedacrossthispeak allowed its identification. ESI massspectraof theparentcompoundand an averageddaughterion spectrumare shown in Fig. 8. These data clearly identify thisphospholipid as lyso-phosphatidylcholine. The full scanspectrumprovidedprotonatedmoleculesfrom m/z� 500to600, while the daughterspectrashoweda peakat m/z184(Fig. 8(b)) which identified it as a phosphocholine, and apeakatm/z339whicharosefrom lossof thephosphocholinehead group, and indicated the acyl-linked fatty chain.Becauseof the structureof lyso-compounds, this fragmenthad the same massas [RCOO�58]� ions observed fordiacylphospholipids.

The most abundant lipids in human plasma werecholesterol, cholesterol-related compounds (cholesterolesters,etc.) and triacylglycerols (TAGs). Thesecoelutedin the large peak at short retention times in Fig. 6.Cholesterol elutedfirst, followedby cholesterolesters,with

Figure 6. Chromatogramsof humanplasmawholeextractby four detectors.(a) ExpandedELSD chromatogram,(b) raw ELSD chromatogram,(c)expandedRIC by APCI-MS, (d) raw RIC by APCI-MS, (e) UV-Vis chromatogram,(f) RIC by ESI-MS/MS,ESI-MS/MSobtainedin automaticparent/daughtercycling mode.

Figure 7. Extractedion chromatogramsobtainedby ESI-MS/MSandfiltered to displayonly scansobtainedin full scan(parent)mode.(a)Mass region from 470 to 600Da which containedthe protonatedmolecularions of lyso-phospholipids,aswell as fragmentsof diacylandacyl-alkenylphospholipids.(b) Massregionfrom 700 to 775Dawhich containstheprotonatedmolecularion shortacyl chainspecies.(c) Massregionfrom 775to 850Dawhichcontainsthelongacylchainmolecularspecies.(d) Massregionfrom 700to 850Dawhichcontainsthe protonatedmolecularion massesfor shortandlong-chaindiacyl,acyl-alkenyl,andsphingolipidspecies.(e)Unfiltered,rawESI-MS/MSchromatogramshowingthe massregionfrom 700 to 850Da.



TAGs coeluted over the first two-thirds of the large peak.APCI massspectrafor thesecompoundsareshownin Fig.9.Figure 9(b) showedunexpectedresultsobtainedfor TAGs.Instead of the protonated molecules and diacylglycerolfragments normally observed for TAGs using APCI–MS,15,19,20 ammonium adducts were observed, instead.The ammonium adductsgave much more molecular ionintensity thanis normallyobtained for TAGs with few sitesof unsaturation. The formation of ammonium adductsofTAGs is not without precedent. Recently, Laakso andManninen reported21 that [M�18]� adducts could beformedwhen ammoniumhydroxide in methanol wasused

asa reagentby running thesheathgasthroughthissolution.For the work presentedhere, as for the previous work,13

ammonium hydroxidewaschosen simply asan electrolyteto produce a stablecurrentunderESI-MS conditions, andbecauseit allowedtheuseof a UV-Vi s detector, in contrastwith acetic acid.As thespectrum in Fig. 9(b) indicates,thesystem producedammonium adductsexclusively, with noprotonated molecule formation, and greatly reducedamounts of diacylglycerol fragments. The ammoniumadducts actedto produce muchlargermolecular ion adductpeaksthanhavepreviously beenreported usingAPCI-MS.Usually, under APCI-MS conditions, triacylglycerols gave

Table 2. Phospholipid molecular speciesarea percent of two human plasmaphospholipids

PC SPM(HeadGroup�H)� = 184amu (HeadGroup�H)� = 184amu

PCplas [M�H]� % PC [M�H]� % DHS [M�H]� % SPM [M�H]� %

22:0 802.7 0.0 38:0 818.7 0.0 16:0 705.6 100.0 24:0 815.7 6.322:1 800.7 0.0 38:1 816.7 0.1 24:1 813.7 16.022:2 798.7 0.8 38:2 814.7 0.8 24:2 811.7 8.322:3 796.7 1.5 38:3 812.6 3.5 24:3 809.7 0.522:4 794.6 1.0 38:4 810.6 5.3 24:4 807.7 0.022:5 792.6 1.3 38:5 808.6 3.1 24:5 805.7 0.322:6 790.6 0.7 38:6 806.6 2.4 24:6 803.7 0.420:0 774.7 0.9 38:7 804.6 0.5 22:0 787.7 7.820:1 772.6 4.3 36:0 790.7 0.0 22:1 785.7 6.320:2 770.6 5.3 36:1 788.6 3.8 22:2 783.7 1.420:3 768.6 6.9 36:2 786.6 14.9 22:3 781.7 0.620:4 766.6 3.2 36:3 784.6 11.1 20:0 759.7 4.120:5 764.6 3.3 36:4 782.6 8.0 20:1 757.6 2.018:0 746.6 6.1 36:5 780.6 1.1 20:2 755.6 0.318:1 744.6 16.9 36:6 778.6 0.2 20:5 749.6 0.218:2 742.6 11.0 36:7 776.6 0.0 20:6 747.6 1.018:3 740.6 8.8 34:0 762.6 1.2 18:0 731.6 4.816:0 718.6 11.1 34:1 760.6 16.2 18:1 729.6 3.116:1 716.6 16.7 34:2 758.6 22.9 18:2 727.6 0.514:0 690.6 0.3 34:3 756.6 2.5 16:0 703.6 28.8

32:0 734.6 0.0 16:1 701.6 4.4Sum 100.1 32:1 732.6 1.5 16:2 699.6 0.5

32:2 730.6 0.9 14:0 675.6 2.230:0 706.6 0.0

Sum 99.8Sum 100.0

Percentof phosphocholines Percentof sphingolipidsPCplas= 32.4PC= 67.6 DHS= 1.1 SPM= 98.9

Figure 8.Massspectraof lysophosphatidylcholineobtainedby ESI-MSandESI-MS/MS.(a)Averageof ESI-MSmassspectraacrosspeakat26 to 26.5min in Fig. 7 obtainedin full scan(parent)mode.(b) Daughterion massspectrumobtainedby ESI-MS/MSof ion at522.4Da. Collision induceddissociation(CID) usingargoncollision gasusedfor ESI-MS/MS.



very low abundancesof protonatedmolecules for specieswith little unsaturation, while highly unsaturated species

gavemostly protonatedmolecules.This behavior hasbeenreportedextensively.15,19–21It wasanunexpectedancillarybenefit of this study to observe the enhancement ofmolecular ion adduct abundancesbrought about by theammonium hydroxide incorporated into the LC system.AlthoughtheTAGsarenot resolved into molecular species,they can be identified by overall acyl chain length anddegreeof unsaturation. Quantitative analysisof the TAGmolecular speciesin the sampleusedfor this study waspreviously performed(Ref. 22, andmanuscript in prepara-tion) using the RP-HPLC/APCI-MS methodology pre-viously established.20,21 The complete TAG compositionthus obtained was converted to the format of (carbonnumber:sites of unsaturation) for comparisonwith quanti-tative analysis performed using the ammonium adductsshown in Fig. 9(b). The results using both methodsareshownin Table 3. While the specificity allowed by a fullRP-HPLC/APCI-MS run is invaluable for identificationofindividual molecular species of TAGs, the ability tocharacterizeall TAGs by acyl carbonnumber in the samerun asphospholipid molecular speciescould be usefulasatotal lipid screening procedure.

In addition to the lipids discussedabove,phosphatidylethanolamines were also observed in the human plasmasample.Though presentat comparatively low levels, themajor individual molecular speciescould be identified.Asseen in Fig. 9(d), these specieswere similar to thoseobservedin eggPEshown in Fig.4(a)and4(b)(aswell asinbovinebrain,not shown).

Figure 9.APCI-MSmassspectraof lipid constituentsof humanplasmawholeextract.(a)Massspectrumof first partof peakelutedat 2.5 to 4 min in Fig. 6; peakcontainsmostly cholesterol.(b) Massspectrumshowingammoniumadductsof triacylglycerols(TAGs)andfragmentsformedfrom lossof fatty acylchains[M - RCOO]� from TAGs.(c) Massspectrumof thesecondhalf of thepeakelutedbetween2.5and4 min; containscholesterol-relatedspecies.(d) Massspectrumshowingmolecularionsof phosphatidylethanolaminespecies.

Table 3. Human plasmatriacylglycerol compositioncomparison

TAGAPCI-MS

[M � NH4]�

from APCI-MS completeTAG composition

54:0 0.0 0.354:1 0.1 0.554:2 1.8 2.654:3 8.7 8.654:4 9.7 10.354:5 3.7 5.754:6 0.7 2.152:0 0.2 0.452:1 1.6 3.652:2 22.4 20.552:3 23.7 22.052:4 7.4 9.052:5 0.5 1.450:0 0.2 0.750:1 5.7 3.950:2 8.1 4.950:3 3.1 2.250:4 0.4 0.448:0 0.4 0.248:1 1.1 0.548:2 0.8 0.4

Sum 100.3 100.2



Other observations

Two other observations arosefrom this work. First, it wasfound that when the scan mass range was extendedsufficiently, peaksarising from dimerization of phospholi-pids, presumably in the ESI source,wereobserved.Figure10 showsdimers formed from long and short chain PCspecies, long andshortchainSPMspeciesandLPC speciesfrom thehumanplasmaextract. In thesespectra, almostallcombinationsof thespeciespresentareobserved.Thedimer

massesrepresentedeitherthesumof a protonatedandnon-protonatedphospholipid or both protonatedphospholipids.Thelabelsonthefiguresgivethemassesfor bothprotonatedphospholipids.For this experiment, theESI-MSinstrumentwasrun in full scanmodeonly. In somecases these dimerswere presentat levels ashigh as5 or 6 percentof the basepeakabundance.While theanalytical utility of thesedimersis marginal, their occurrenceunderESI conditions providesan indication of a low energy of formation, which mighthave biological implications. Formation of similar ion

Figure 10.ESI-MSmassspectrashowingformationof dimersof phospholipidmolecularspeciesof humanplasmain ESI source.(a)Massspectrumof rangefrom 1400to 1700Da,showingpeaksarisingfrom thecombinationof PCphospholipidselutedbetween15and17minutesin Fig.7. (b) Massspectrumshowingpeaksarisingfrom thecombinationof PCphospholipidselutedbetween18and20minutesin Fig.7. (c) Massspectrumof secondSPMpeakshowing combinationof sphingolipids. (d) Massspectrumof thirdSPM peakshowingcombinationof shorter-chainsphingomyelins.(e) Massspectrumof rangefrom 900 to 1200Da showingcombinationof lyso-PCspecieselutedfrom 28 to 29.5min in Fig. 7. For massesof PCandSPMspecies,seeTables.For lyso-PCusethesemasses:16:0-LPC:496.3Da;18:0-LPC:524.4Da;18:1-LPC:522.4Da;18:2-LPC:520.3Da;18:3-LPC:518.3Da;20:0-LPC: 552.4Da; 20:4-LPC:544.3Da.



clusters from several phospholipid standardswas recentlyreported using liquid secondaryion mass spectrometry(LSIMS).23

Another observation which arose waslessbenign. Afterallowing the column to sit unused for a week, thecharacteristics of the chromatographic system changeddramatically. When further analyseswere performed, thechromatographic retention times were dramatically ex-tended. Examinationof the resultant spectraindicatedthatsodium adductswere beingelutedinstead of theprotonatedmolecules to which we had become accustomed. Achromatogramof the humanplasmasampleis shown inFig. 11 along with an ESI massspectrum of PC speciesobtained across the top of the peak which now elutedbetween 34and42min (comparedto 16to 20min in Fig.6).The elution of SPM in this run was not completeafter60min. Figure 11(b) showsexclusively massesassociatedwith sodiumadducts,[M�23]�, of thePCspeciesinstead of

protonatedmolecules, [M�1]�. APCI massspectrasimi-larly showedsodiatedmoleculesfor phospholipids, but stillprovided the normal fragmentsresulting from loss of thehead group. While trying to resolve the problematicchromatographic behavior, a new column was installedandall solventswerechanged. Attemptsto regenerate theold columnusing aceticacid,phosphoric acidor ammoniumhydroxide solutions had little effect or resulted in anadmixture of sodiated and protonated molecules, withcontinued poor chromatography. This result was ratherunexpected since this chromatographic system had pre-viously beenapplied to different studiesover a period ofyears.On theotherhand,manyotherauthorshavereportedsodium adductsof phospholipids exclusively when usingother chromatographic systems.Af ter severalattempts tosolvetheproblem, it wasfoundthattheproportioningvalveon thepumphadmalfunctioned, andthatChannelD on thepump,which hadbeenusedto deliver theH2O/IPA portion

Figure 11.ESI-MSchromatogramsandmassspectraof humanplasmaextractsodiatedadducts.(a) ESI-MSchromatogramshowingtheRIC andshortandlongchainmassregions.(b) Massspectrumshowingsodiatedmolecularionsfor PCspecieselutedbetween36and40 minutes.(c) Massspectrumshowingsodiatedtriacylglycerolmolecularionsandnormaldiacylglycerolfragments.



of the eluent,was not properly delivering flow. Thus, theabsenceof asufficientamountof aqueouscomponentled tothechromatographicchange. Fromthis changein behavior,it appearedthattheaminecolumnhaspreviously actedasanion exchangecolumnto removesodiumfrom phospholipidpreparations,thusyielding only protonatedmolecules.Themalfunction of the HPLC instrument thus provided anunexpectedandusefulinsight into theretentionmechanismof the column.

Concomitant with the change in the appearance ofphospholipid spectrawas a changein the appearanceoftriacylglycerol spectraobtained by theESI-MS instrument.One would not expect large neutral TAGs to give goodresponseunderESI-MS conditions, althoughESI/MS haspreviously beenusedfor analysis of TAGs.24–26However,when thechromatographic systemchanged its behavior, weobtained spectrafor TAGs which showedboth sodiated

moleculesanddiacylglycerol fragments.Figure11(c)showsa massspectrum for the TAG speciesin humanplasmaobtainedusing ESI.Thisspectrum maybecompared to thatin Fig. 9(b),which showedtheammonium adductsof TAGobtained using APCI. The differenceof 5 Da between thequasi-molecular ionsin Fig. 11(c)versus9(b) representsthedifferencebetweensodium adducts (�23) versusammo-nium adducts (�18). During this samerun, while sodiumadductsappearedin ESI-MSspectra,theAPCI massspectraof TAGs continued to show ammonium adducts, andappeared the sameasthe spectrum in Fig. 9(b).

Porcine lensextract

Although the sodium adductformation wasnot eliminateduntil the proportioning valve was replaced,washing thecolumnwith ammoniumhydroxidesolution did at leastgive

Figure 12.ESI-MSchromatogramsandmassspectraof porcinelensextractsodiatedadducts.(a)ESI-MSchromatogramshowingtheRIC andshortandlong chainphospholipidmassregions.(b) Massspectrumof thesecondof threesphingolipidpeaks.PeakcontainedshortchainDHSandlongchainSPMspecies,aswell asresidualPC,(c) Massspectrumof thethird of threesphingolipidpeaks.PeakcontainedshortchainSPMspecies.



betterpeakshapesandsphingomyelin could be eluted.AnHPLC/ESI-MS chromatogram of porcine lens extract isshown in Fig. 12. The chromatogram and resultantmassspectraallowed the identification of both DHS and SPMspecies. Examplespectraof the sodiated sphingolipids areshown in Fig. 12(b)and(c). Thoughsodiumadducts of themolecules wereobserved(both by ESI-MS andAPCI-MS,though only small abundancesin APCI spectra), thesodiumion was associated with the headgroup,so the sphingoidbackbone fragments formed during APCI were the exactsame as those previously observed.Therefore, quantifica-tion was performed using the same massesas before thechangein the chromatographic behavior. The molecularspeciesaregiven in Table4. Extractedion chromatogramsof thesphingoidbackbonefragmentsobtained underAPCI-

MS conditions areshown in Fig. 13(a), while EICs for thesodiated molecules simultaneously obtained under ESIconditions, in parallel, are shown in Fig. 13(b). Allmolecular species identified were confirmed by thepresenceof simultaneouspeaksin the APCI-MS andESI-MS extractedion chromatograms.Porcine Lens sphingo-lipids werecalculatedto be composedof 14.4%DHS and85.6% SPM. The most abundant molecular specieswere16:0,18:0,22:1and24:1,althoughtherelativepercentagesof thesediffered between the DHS andSPM.

Conclusions

The dual massspectrometer system allowed data to beobtainedunder different ionization conditions using onechromatographicrun. This providedthebenefit of virtuallyidentical retention times between the different systems.Comparedto the methodpreviously used13 for obtainingthis same type of data,the amountof time, aswell as theamountof solventandsampleused,werecut by half. Moreimportantly, the amountof information obtained per runwas increased dramatically. For componentsfor which nostandards hadbeenrun (e.g.lyso-PCin humanplasma) theacquisitionof daughter ion spectraobtainedby ESI-MS/MSwas decisive in identification of the compounds.The twoionization methodsdemonstratedhere produced comple-mentarydatafor severaltypes of molecules.PC,DHS andSPMproducedonly intact ionizedmoleculesunderESI-MSconditions, while they produced diagnostically usefulfragmentsunderAPCI-MS conditions.Having both typesof datawas especially useful for the sphingolipids whichproducedvery low abundancesof protonatedmolecules inAPCI-MS spectra.The useof EICs of sodiatedmoleculesfrom ESI data, alongwith EICsof fragmentsin APCI data,prevented misinterpretation of potentially ambiguousmassesin porcine lensextract.

We have chosen only the most simple conditions todemonstrate the application of the dual massspectrometersystem.The number of permutations is large when oneconsiders that both positive andnegative ion analysismaybe performed on either instrument,APCI andESI may beperformedon eithermachine(the sourcesare interchange-able),andthetriple quadrupolemachinemayberun in full

Table 4. Porcine lenssphingolipid molecular speciesarea

[HeadGroup�H]� = 184amuDHS [M�H]� % SPM [M�H]� %

24:0 817.7 1.0 26:0 843.8 0.024:1 815.7 6.8 26:1 841.7 0.624:2 813.7 2.6 26:2 839.7 0.524:3 811.7 0.0 26:3 837.7 0.222:0 789.7 2.3 26:4 835.7 0.122:1 787.7 7.8 24:0 815.7 3.422:2 785.7 1.1 24:1 813.7 15.120:0 761.7 4.2 24:2 811.7 3.520:1 759.7 4.2 24:3 809.7 0.220:2 751.6 0.0 22:0 787.7 4.318:0 733.6 12.3 22:1 785.7 13.418:1 731.6 2.0 22:2 783.7 0.718:2 729.6 1.2 20:0 759.7 3.716:0 705.6 41.6 20:1 757.6 3.316:1 703.6 2.9 20:2 755.6 0.314:0 677.6 9.0 18:0 731.6 10.014:1 675.6 0.9 18:1 729.6 1.9

16:0 703.6 30.6Sum 99.9 16:1 701.6 3.4

14:0 701.6 4.9Sum 100.1

Percentof sphingolipidsDHS= 14.4 SPM= 85.6

Figure 13.Extractedion chromatogramsof porcinelenssphingolipidsby APCI-MSandESI-MScorrespondingto themostabundantspeciesgivenin Table4. (a)APCI-MS EICsof sphingolipidbackbone([M-headgroup]�) fragmentsfor sphingolipidselutedbetween26and33minutes.(b) ESI-MS EICsof themassesof sodiatedmolecularionsof themostabundantspeciesgiven in Table4.



scanmodeexclusively or MS/MSmaybeperformed.Giventhelargenumberof possibleexperiments, theacquisition ofdata from two machines simultaneously is an importanttime-savingarrangement.Of course, dedicatingtwo com-pletemass spectrometersystemsto analysis is not feasiblein most laboratories.The cost of such instrumentation isoftenprohibitive.Ontheother hand,theadventof benchtopAPCI and ESI ion trap technology makes use of dualinstruments more feasible than ever before. Given theextensive amount of data provided, future studies ofphospholipids in our laboratorywill certainly employ thedual mass spectrometric systemasa matter of course.

Theuseof two other detectorshadsignificant advantages,aswell. The quantitationof the relative proportions of thedifferentclassesof moleculeswasbestperformedusing theELSD, becausewe observedsubstantial differences in theresponses of different phospholipid classesunder MSconditions. However, within each phospholipid class,identification andquantification werebestdoneusingmasschromatograms.SinceweusedESI-MSfor MS/MSanalysisin this study, APCI-MS alone was used for quantitativeanalysis. Using ESI-MS in full scanmodedoes,however,allow quantitative analysis to be performedin addition toAPCI, aspreviously shown.13 TheUV-Vi sdetectorshowedsevere limi tations in its selective responseto differentphospholipid classes, as well as its susceptibility toimpurities in the solvent system. If available, a secondELSD detectoror a flameionizationdetectorfor LC shouldbe substituted for the UV-Vis detector. This would alsoallow more flexibility in the choiceof electrolyte for ESI.

The use of both mass spectrometers allowed theidentification of numerousdihydrosphingomyelin molecu-lar speciesin bovine brain, and to determine that humanplasmacontainsvirtually noDHS. Useof bothsystemswasespecially valuable for the porcinelens extract,with bothmachines operating in full scan mode. As the DHSmolecular species are identified in an increasingnumberof tissue types, distinct tendenciesfor several commonspeciesstartto becomeevident.However, therearedistinctcompositionaldifferencesbetween tissuetypes.Therelativeproportion of DHS to SPM from porcine eye lensmembraneswas much more similar to bovine brain thanto human eye lens membranes. But the identity of themolecular specieswassomewherein between thecomposi-tionsof bovinebrainandhuman lens(high amountsof 16:0like the humanlens, but the levelsof 18:0 weremore likebrainextract).Therolesof theindividual molecular speciesneed to be elucidated and, given the effect of DHS onmembranefluidity , its biological significance needsto befurther assessed.The natureof sphingomyelins from othersourcesshould bere-examinedin light of thenewanalyticalcapabilities which haveevolved.

Acknowledgment

The work of Lynne Copeson the extractionof the humanplasmasampleis gratefully acknowledged.

REFERENCES

1. J.L. Kerwin, A. R. TuiningaandL. H. Ericsson,‘IdentificationofMolecular Speciesof Glycerophospholipidsand SphingomyelinUsing ElectrosprayMassSpectrometry’,J. Lipid Res.35, 1102–1114(1994).

2. F. B. Jungalwala,V. H. Hayssen, J. M. Pasquini and R. H.

McCluer, ‘Separationof MolecularSpeciesof SphingomyelinbyReversed-PhaseHigh PerformanceLiquid Chromatography’,J.Lipid Res.20, 579–587(1979).

3. J. I. Tengand L. L. Smith, ‘Improved High-PerformanceLiquidChromatographyof Sphingomyelin’, J. Chromatogr.322, 240–245(1985).

4. H.-Y. Kim and N. Salem, ‘Application of ThermosprayHigh-PerformanceLiquid Chromatography/MassSpectrometryfor theDeterminationof Phospholipidsand RelatedCompounds’,Anal.Chem.59, 722–726(1987).

5. A. Valeur,N. U. Olsson,P. Kaufmann,S.Wada,C.-G.Kroon,G.Westerdahland G. Odham,‘Quantification and ComparisonofSome Natural Sphingomyelinsby On-line High-PerformanceLiquid Chromatography/Discharge-AssistedThermosprayMassSpectrometry’,Biol. MassSpectrom.23, 313–319(1994).

6. A. Kuksis and J. J. Myher, ‘Application of Tandem MassSpectrometryfor the Analysisof Long-ChainCarboxylicAcids’,J. Chromatogr.B 671, 35–70(1995).

7. N. U. Olsson and N. Salem, ‘Molecular SpeciesAnalysis ofPhospholipids’,J. Chromatogr.B 692, 245–256(1997).

8. A. A. Karlsson,P.Michelson,A. LarsenandG. Odham,‘Normal-Phase Liquid ChromatographyClass Separationand SpeciesDetermination of Phospholipids Utilizing Electrospray MassSpectrometry/TandemMass Spectrometry’, Rapid Commun.Mass.Spectrom.10, 775–780(1996).

9. X. Han andR. W. Gross,‘StructuralDeterminationof PicomoleAmounts of PhospholipidsVia ElectrosprayIonization TandemMassSpectrometry’,J. Am. Soc.MassSpectrom.6, 1202–1210(1995).

10. H.-Y. Kim, T.-C. L. Wang and Y.-C. Ma, ‘Liquid Chromato-graphy/MassSpectrometryof PhospholipidsUsing ElectrosprayIonization’, Anal. Chem.66, 3977–3982(1994).

11. W. C. Byrdwell, D. Borchman,R. A. Porter,K. G. Taylor andM.C. Yappert, ‘Separationand Characterizationof the UnknownPhospholipidin Human Lens Membranes’,Invest. Ophthalmol.Vis. Sci.35, 4333–4343(1994).

12. V. L. Hanson,J. Y. Park,T. W. OsbornandR. M. Kiral, ‘High-PerformanceLiquid ChromatographicAnalysis of Egg YolkPhospholipids’,J. Chromatogr.205, 393–400(1981).

13. W. C. Byrdwell andD. Borchman,‘Liquid Chromatography/MassSpectrometricCharacterizationof Sphingomyelinand Dihydro-sphingomyelinof HumanLensMembranes’,OphthalmicRes.29,191–206(1997).

14. J.Folch,M. LeesandG. H. SloaneStanley,‘A SimpleMethodforthe Isolation and Purification of Total Lipides from AnimalTissue’,J. Biol. Chem.226, 497–509(1957).

15. W. E. Neff and W. C. Byrdwell, ‘SoybeanOil TriacylglycerolAnalysisby Reversed-PhasesHigh PerformanceLiquid Chroma-tographyCoupledwith AtmosphericPressureChemicalIonizationMassSpectrometry’,J. Am.Oil Chem.Soc.72, 1185–1191(1995).

16. W. C. Byrdwell, D. Borchman,R. A. Porter,K. G. Taylor andM.C. Yappert, ‘Separationand Characterizationof the UnknownPhospholipidin Human Lens Membranes’,Invest. Ophthalmol.Vis. Sci.35, 4333–4343(1994).

17. D. Borchman,W. C. Byrdwell and M. C. Yappert, ‘Thermo-dynamic PhaseTransition Parametersof HumanLens Dihydro-sphingomyelin’,OphthalmicRes.28(suppl1), 81–85(1996).

18. W. C. Byrdwell andW. E. Neff, ‘Analysisof Hydroxy-ContainingSeedOils UsingAtmosphericPressureChemicalIonizationMassSpectrometry’,J. Liq. Chromatogr.Rel. Technol., in press.

19. W. C. Byrdwell, E. A. Emken, W. E. Neff and R. O. Adlof,‘Quantitative Analysis of Triglycerides Using AtmosphericPressureChemical Ionization Mass Spectrometry’,Lipids 31,919–935(1996).

20. W. C. Byrdwell and W. E. Neff, ‘Qualitative and QuantitativeAnalysisof TriacylglycerolsUsingAtmosphericPressureChemi-cal Ionization Mass Spectrometry’, in New TechniquesandApplications in Lipid Analysis, R. E. McDonald and M. M.Mossoba(Eds),pp. 45–80,AOCSPress,Champaign,IL (1997).

21. P. Laakso and P. Manninen, ‘Identification of Milk FatTriacylglycerols by Capillary SupercriticalFluid Chromatogra-phy-AtmosphericPressureChemical Ionization Mass Spectro-metry’, Lipids 32, 1285–1295(1997).

22. W. C. Byrdwell and E. A. Emken, ‘Liquid Chromatography/AtmosphericPressureChemical Ionization Mass SpectrometricCharacterizationof HumanChylomicronTriacylglycerols’, 88thAOCSAnnualMeetingAbstracts,p.58,AOCSPress,Champaign,IL (1997).

23. Y. Wu, J.WangandS.Sui, ‘Characterizationof Phospholipidsby



ElectronImpact,FieldDesorptionandLiquid SecondaryIon MassSpectrometry’,J. MassSpectrom.32, 616–625(1997).

24. K. L. Duffin, J. D. Henion and J. J. Shieh, ‘ElectrosprayandTandem Mass SpectrometricCharacterizationof AcylglycerolMixturesThat Are Dissolvedin NonpolarSolvents’,Anal. Chem.63, 1781–1788(1991).

25. J.J.Myher,A. KuksisandP.W. Park,‘StereospecificAnalysisofDocosahexaenoicAcid-Rich Triacylglycerols by Chiral-Phase

HPLC With Online ElectrosprayMass Spectrometry’,in NewTechniquesand Applicationsin Lipid Analysis, R. E. McDonaldandM. M. Mossoba(Eds),pp.100–120,AOCSPress,Champaign,IL (1997)

26. A. Kuksis and J. J. Myher, ‘Application of Tandem MassSpectrometryfor the Analysisof Long-ChainCarboxylicAcids’,J. Chromatogr.671, 35–70(1995).



Dual Parallel Mass Spectrometers for Analysis of … Glycerophospholipid and Plasmalogen Molecular Species Wm. Craig Byrdwell* FQS, NCAUR, ARS, USDA 1815 N. University Street, Peoria,

Documents