LC/ESI-MS/MS detection of FAs by charge reversal derivatization with more than four orders of magnitude improvement in sensitivity.

Quantitative analysis of fatty acids (FAs) is an important area of analytical biochemistry. Ultra high sensitivity FA analysis usually is done with gas chromatography of pentafluorobenzyl esters coupled to an electron-capture detector. With the popularity of electrospray ionization (ESI) mass spectrometers coupled to liquid chromatography, it would be convenient to develop a method for ultra high sensitivity FA detection using this equipment. Although FAs can be analyzed by ESI in negative ion mode, this method is not very sensitive. In this study, we demonstrate a new method of FA analysis based on conversion of the carboxylic acid to an amide bearing a permanent positive charge, N-(4-aminomethylphenyl)pyridinium (AMPP) combined with analysis on a reverse-phase liquid chromatography column coupled to an ESI mass spectrometer operating in positive ion mode. This leads to an ∼60,000-fold increase in sensitivity compared with the same method carried out with underivatized FAs. The new method is about 10-fold more sensitive than the existing method of gas chromatography/electron-capture mass spectrometry of FA pentafluorobenzyl esters. Furthermore, significant fragmentation of the precursor ions in the nontag portion improves analytical specificity. We show that a large number of FA molecular species can be analyzed with this method in complex biological samples such as mouse serum.

due to the fact that analytes that form isobaric precursor ions and coelute during LC will not be distinguished in the mass spectrometer if they give rise to the same detected fragment ion, essentially eliminating any advantage a MS/ MS experiment has over a MS experiment. This loss of specifi city represents a signifi cant limitation when analyzing complex biological samples.
We recently reported a straightforward LC-ESI-MS/MS derivatization procedure for the targeted lipidomic analysis of eicosanoids via stable isotope dilution ( 32 ). The carboxyl group is derivatized with a newly developed reagent, N -(4aminomethylphenyl)pyridinium (AMPP), that results in a permanent positive charge (charge reversal). This derivatization results in a 10-to 20-fold improvement in detection sensitivity by LC-ESI-MS/MS ( 32 ). Our methodology employed a simple solid-phase extraction procedure of eicosanoids from a variety of biological matrices followed by a mild quantitative derivatization step with AMPP. The resulting derivatives can be directly submitted to LC-ESI-MS/MS and display robust fragmentations in their analyte segments making them attractive candidates for high-sensitivity/specifi city SRM experiments. Here we utilize a similar approach, with the exception of an alternative extraction method, to monitor the free FA profi les in complex biological samples. We developed and validated a stable isotope dilution LC-ESI-MS/MS method that is able to detect essentially all saturated and unsaturated FAs in a single chromatographic run. Sensitivity improvement over LC-ESI-MS/MS of underivatized FAs in negative ion mode is ‫ف‬ 60,000-fold.

Preparation of FA-free glassware and reagents
Low abundant FAs such as AA are usually not present as a contaminant in glassware and reagents; however, abundant FAs such as oleic, palmitic, and stearic acids are present as common contaminants. It has not been possible to remove these contaminants to a level below the FA detection limit for the method described in this paper. The procedure described here reduces abundant FA contamination to a level usually below the amounts to be detected in the sample of interest.

Preparation of FA stock solutions
The following FA standards from Cayman Chemicals were used (d 14 -palmitoleic acid, d 14 -␣ -linolenic acid, d 4 -linoleic acid, d 5 -eicosapentaenoic acid, d 8 -AA, d 17 -oleic acid, d 6 -dihomo-␥ -linolenic acid, d 5 -DHA, stearidonic acid, and AA ( -3). d 31 -Palmitic acid complex biological mixtures. Although useful, these methods are still limited by dynamic range limitations and compound volatility considerations ( 11 ). Although electroncapture detection of pentafl uorobenzyl esters of FAs provides exceptionally high sensitivity, there are many laboratories worldwide that now have access to electrospray ionization (ESI) machines rather than electroncapture instruments. However, a major obstacle to the ESI technique is that FAs undergo less than ideal fragmentation behavior in negative ion mode via collision-induced dissociation (CID). Under low-energy (<100 eV) CID conditions typical to most commercial instruments, fragmentation of the featureless backbone of a saturated FA is minimal. Furthermore, the most prominent fragments originate from the loss of CO 2 ( Ϫ 44 Da) and elimination of water ( Ϫ 18 Da) from the carboxylic acid group, neither of which are specifi c enough for reliable quantitation in complex matrices. Unsaturated FAs do undergo, to some extent, fragmentations that are specifi c to their structure. However, the abundances of these fragmentations are relatively weak and result in SRM measurements of poor sensitivity. Another major limitation of this approach is related to a FA's relative ionization effi ciency and the manner in which the ions are analyzed. For compounds that contain free carboxylates, such as FAs, ionization is best achieved in negative ion mode under basic pH conditions where the carboxylate is ionized (12)(13)(14)(15). Unfortunately, optimal LC resolution is facilitated by acidic pH conditions, to keep the carboxyl group protonated where ionization of the carboxylate is suppressed. Post-column addition of base could potentially alleviate this problem at the expense of the method's simplicity and sensitivity.
One group recently reported a LC-ESI-MS/MS method of FA analysis in plasma using post-column infusion of a barium ion solution ( 16 ). The formation of positively charged adduct ions promotes diagnostic fragmentation reactions of unsaturated FA species with enhanced SRM detection sensitivity. Other cation reagents, including alkaline earth metals and copper ions, also proved suitable for enhanced sensitivity for FA analysis in the SRM mode ( 17,18 ). An alternative strategy for enhanced sensitivity is to improve the ionization effi ciency of FAs via specifi c derivatization with reagents that introduce either readily chargeable or fi xed charge groups such as tertiary or quaternary amines, respectively. Many derivatives of this nature have been reported including: pyrolidides (19)(20)(21), picolinyl esters ( 22,23 ), dimethyloxazolines ( 24-28 ), benzofurazans ( 29 ), pyridiniums ( 30 ), and cholines ( 31 ). The advantages of these derivatives include improved MS sensitivity and reproducible chromatography profi les. A major limitation of these methodologies is the relatively harsh conditions usually required for derivatization, which can result in unwanted oxidation, isomerization, or degradation of some FAs. This limitation could potentially be addressed by the development of robust derivatization procedures that require milder conditions. Another major limitation is the tendency of these derivatives to fragment via CID in immediate proximity to the chargeable/cationic site. Fragmentation in the derivatization tag is undesirable transferred to a 12 × 75 mm glass culture tube. To each culture tube, 50 l of absolute ethanol containing 1 ng of each internal standard was added. The sample was adjusted to 125 l by adding purifi ed water (Milli-Q, Millipore Corp.). Aliquots of 250 l of methanol (Fisher Optima grade, catalog #A456-4) and 12.5 l of 1 N HCl were added to each sample. A bi-phasic solution was formed via addition of 750 l of isooctane. This solution was vortexed for 60 s, and the phases were separated by centrifugation at 3,000 rpm for 60 s. The upper isooctane phase was removed via an oven-baked glass Pasteur pipette and transferred to an ovenbaked Waters Total Recovery vial. The remaining aqueous phase was extracted once more with an additional 750 l of isooctane. The combined isooctane phases were evaporated to dryness under a stream of fi ltered N 2 and derivatized with AMPP as described below.
Derivatization with AMPP. AMPP was synthesized in-house as described previously ( 32 ). Subsequent to our lead publication, the AMPP reagent was made commercially available by Cayman Chemical Company (catalog #710000) under the product name AMP+ Mass Spectrometry Kit.
To the residue in the oven-baked Waters Total Recovery autosampler vial was added 10 l of ice-cold acetonitrile/DMF (4:1, v/v). Ten microliters of ice-cold 1 M EDCI in distilled Milli-Q water (freshly prepared daily) was added. The vial was briefl y and d 35 -stearic acid were from Sigma-Aldrich. GLC-463 standard (Nu-Check Prep, Inc.), containing 52 distinct FA molecular species, was used for the rest of the calibration standards. Stock solutions of FAs were prepared at concentration of 25-100 pg/ l in absolute ethanol and stored at Ϫ 80°C under Ar in 1.5 ml amber vials (Agilent, catalog #5182-0716) with polytetrafl uoroethylene/ silicone septum screw caps (Agilent, catalog #5185-5838). Serial dilutions of the stock solutions were made in absolute ethanol for standard curve and extraction recovery analyses. Internal standards were diluted to a working stock of 100 pg/ l in absolute ethanol.

Preparation of samples and derivatization with AMPP
Standard curves. Each sample contained 1 ng of each internal standard and various amounts of nondeuterated FAs (added from serial dilutions of the accurate concentration stock solution made from milligram amounts of FA as described above) transferred to a glass auto-sampler vial (Waters Total Recovery screw cap vial, catalog #186002805). Solvent was removed with a stream of nitrogen, and the residue was derivatized with AMPP as described below.

Extraction of FAs from mouse serum. Analysis of endogenous
FAs in serum was carried out with commercial mouse serum (Atlantic Biologicals, catalog #S18110). A 10 l aliquot of serum was

LC/ESI-MS/MS analysis
Studies were carried out on a Waters Xevo TQ triple quadrupole mass spectrometer interfaced to an Acquity UPLC. The MassLynx 4.1 software package was used for data collection and ana lysis. Chromatography was carried out with a C18 reversephase column (Waters Acquity UPLC BEH Shield RP18, 2.1 × 100 mm, 1.7 m, catalog #186002854). Solvent A was 100% water (Fisher Optima grade, catalog #L-13780)/0.1% formic acid (Fisher mixed on a vortex mixer and placed on ice while other samples were processed as above. To each vial was added 20 l of 5 mM HOAt/15 mM AMPP in distilled acetonitrile (stored at Ϫ 20°C and warmed to 65°C immediately prior to use). The vials were mixed briefl y on a vortex mixer, capped with a split-septum screw cap (Agilent, catalog #5185-5824), and placed in a 60°C incubator for 30 min. Samples were analyzed on the same day and kept in the auto-sampler rack at 10°C while queued for injection.  protonation resulting from the addition of weak organic acid such as acetic or formic acid, which is necessary for optimal LC on reverse-phase columns. As shown in Fig. 1 , AMPP amides of FAs give rise to spectral signature ions at m / z 169 and 183 due to CID of the AMPP tag. Additionally, abundant high molecular weight fragments are also generated, for example, m / z 239 for fragmentation between C3 and C4 in most FA species ( Fig. 1 ). AMPP amides of oleic acid and its deuterated analog show an abundant product ion at m / z 295 due to cleavage between C7 and C8, thus leaving a relatively stable allylic radical. This ion is not present in the spectrum of the AMPP amide of petroselenic acid ( Fig. 1 ). Likewise, vaccenic AMPP amide shows a major product ion at m / z 323, due to cleavage of the C9-C10 bond to generate an allylic radical. This species is not present in the other 18:1 spectra. These high molecular weight product ions provide for high analytical specifi city, which may be important for analysis of FAs in complex biological samples. If a product ion resulting Optima grade, catalog #A117-50), and solvent B was acetonitrile (Fisher Optima grade, catalog #L-14338)/0.1% formic acid. The solvent program was (linear gradients): 0-0.5 min, 90% A; 0.

RESULTS AND DISCUSSION
As noted in the introduction, conversion of the carboxyl group of lipids, such as eicosanoids and FAs, to the AMPP amide results in an analyte with a permanent positive charge, which can be analyzed by LC-ESI-MS/MS in positive ion mode. This is more sensitive than negative ion mode detection of the underivatized caboxylate anion because ionization of the latter is greatly suppressed by the To gauge the increase in FA detection sensitivity, we analyzed various amounts of d 8 -20:4 AMPP amide in positive ion mode and various amounts of d 8 -20:4 free acid in negative ion mode. For the latter we monitored a major high mass product ion due to the loss of CO 2 . We also tuned the instrument to optimize the cone voltage and collision energy for this transition in negative ion mode. Results are shown in Fig. 2 . Injection of 0.78 pg of d 8 -20:4 AMPP amide gives rise to a peak area of 22,800 in positive ion mode versus 17,100 for 50 ng of d 8 -20:4 free acid in negative ion mode. Thus, the increase in sensitivity for the AMPP derivatization method is 64,000-fold.
Next we analyzed the FAs present in mouse serum, and the results are summarized in Table 2 . Intra-assay coeffi cients of variation based on fi ve injections of the same sample were typically less than 4%. Inter-assay coeffi cients of variations based on injections of six independent extractions of the same serum were typically less than 6%. Thus, the method is highly reproducible. For these studies we used the m/z 239 production ion. As noted above, this is present in all of the FAs, but its use is adequate in the case of mouse serum. Additional analytical specifi city can be obtained by monitoring analytespecifi c precursor ions, such as those noted above for the 18:1 species.
It should be mentioned that accurate quantifi cation of the absolute amount of any particular FA species requires a chemically identical isotopic substituted internal standard. Only in this way can one account for differences in ionization from cleavage of the AMPP tag is used for MS/MS, it would not be possible to distinguish isobaric AMPP-labeled species that coelute during LC. High molecular weight product ions were observed for all FAs analyzed. Precursor and product m / z values for all FAs are given in Table 1 , and product ion mass spectra are shown in Fig. 1  Isobaric species (i.e., cis / trans isomers or double bond positional isomers) were addressed via LC retention times. This was true for all species with the exception of the 18:1 isomers, which were not completely resolved. Although not applicable to the current study, alternative SRM transitions for the isobaric species could also be used as a method to resolve these species as each isomer has a distinct fragmentation pattern.
The limits of quantifi cation for the AMPP amide method were all on the order of 50-100 femtograms on-column, as determined by standard curve analysis . We used accurate concentration FA stock solutions made from milligram amounts of FAs and carried out serial dilution to obtain low concentration stock solutions. AMPP derivatization and pre-MS/MS sample clean-up were carried out on fully diluted FA solutions, so the limits-of-quantifi cation we report include any losses due to AMPP derivatization and pre-MS/MS sample clean-up. The limit-of-quantifi cation of FA by gas chromatography/electron-capture mass spectrometry of pentafl uorobenzyl esters is reported to be about 10 femtomoles (3,000 femtograms) ( 33 ). Thus, our method is about 10-fold more sensitive than this previous method of FA analysis. A trace amount of 12:1 fatty acid was seen but variability was high due to its low level. %CV, percent coeffi cient of variation . effi ciencies in the mass spectrometer source and in differences in precursor-to-product ion generation for the different FA molecular species. Also, if a deuterated FA is used as in internal standard, deuterium should not be present at a site that leads to an isotope effect on the amount of product ion generated. Another option is to use a limited number of heavy atom substituted FA internal standards and to determine the relative MS/MS signal intensities of each FA molecular species by using standard curves for the appropriate species. This is not as accurate as an internal standard for absolute quantifi cation. Background contamination of solvents was particularly bad for the saturated series of chains 12-18 carbons in length as well as for the monounsaturated 18 carbon series. Baking of glassware and trituration of reagents improved background levels signifi cantly.
Large amounts of derivatization reagents relative to FAs are used to ensure quantitative conversion to AMPP amides. All reagents and their products elute in the void volume of the LC run and do not enter the ESI-MS/MS source because a diversion value is used to direct LC output to waste during the initial part of the run. Thus, the method does not lead to excessive loading of the ESI-MS/ MS source.
In summary, we have developed a new FA quantitative analysis using readily available LC/ESI-MS/MS equipment that provides a sensitivity close to that of the most sensitive FA method so far developed (gas chromatography of pentafl uorobenzyl esters with electron-capture detection). Although LC does not provide the resolving power of capillary gas chromatography, the use of unique MS/MS channels is usually suffi cient to resolve isobaric species that coelute during LC. The new method should fi nd widespread use given the relatively large number of ESI-MS/MS instruments available in modern analytical laboratories.