An LC-MS/MS method to quantify acylcarnitine species including isomeric and odd-numbered forms in plasma and tissues.

Acylcarnitines are intermediates of fatty acid and amino acid oxidation found in tissues and body fluids. They are important diagnostic markers for inherited diseases of peroxisomal and mitochondrial oxidation processes and were recently described as biomarkers of complex diseases like the metabolic syndrome. Quantification of acylcarnitine species can become challenging because various species occur as isomers and/or have very low concentrations. Here we describe a new LC-MS/MS method for quantification of 56 acylcarnitine species with acyl-chain lengths from C2 to C18. Our method includes amino acid-derived positional isomers, like methacrylyl-carnitine (2-M-C3:1-CN) and crotonyl-carnitine (C4:1-CN), and odd-numbered carbon species, like pentadecanoyl-carnitine (C15:0-CN) and heptadecanoyl-carnitine (C17:0-CN), occurring at very low concentrations in plasma and tissues. Method validation in plasma and liver samples showed high sensitivity and excellent accuracy and precision. In an application to samples from streptozotocin-treated diabetic mice, we identified significantly increased concentrations of acylcarnitines derived from branched-chain amino acid degradation and of odd-numbered straight-chain species, recently proposed as potential biomarkers for the metabolic syndrome. In conclusion, the LC-MS/MS method presented here allows robust quantification of isomeric acylcarnitine species and extends the palette of acylcarnitines with diagnostic potential derived from fatty acid and amino acid metabolism.


LC-MS/MS
The analysis was performed on a triple quadrupole QTRAP5500 LC-MS/MS system (AB Sciex, Framingham, MA), equipped with a 1200 series binary pump, a degasser, and a column oven (Agilent, Santa Clara, CA) connected to a HTC pal autosampler (CTC Analytics, Zwingen, Switzerland). A Turbo V ion spray source operating in positive ESI mode was used for ionization. The source settings were: ion spray voltage, 5,500 V; heater temperature, 600°C; source gas 1, 50 psi; source gas 2, 50 psi; curtain gas, 40 psi; and collision gas (CAD), medium.
Chromatographic separation was achieved on a Zorbax Eclipse XDB-C18 column (length 150 mm, internal diameter 3.0 mm, particle size 3.5 m; Agilent) at a column temperature of 50°C. Eluent A consisted of 0.1% formic acid, 2.5 mM ammonium acetate, and 0.005% HFBA in water. Eluent B consisted of 0.1% formic acid, 2.5 mM ammonium acetate, and 0.005% HFBA in acetonitrile. Gradient elution was performed with the following program: 100% A (0.5 ml/min) for 0.5 min, a linear decrease to 65% A (0.5 ml/min) for 2.5 min, hold for 3 min, a linear decrease to 40% A (0.5 ml/min) for 3.7 min, a linear decrease to 5% A (0.5 ml/min) for 1 min. Elution was then carried out at 100% B (1 ml/min) for 0.5 min, hold for 7.3 min (1.5 ml/min). Re-equilibration was fi nally performed at 100% A (0.5 m/min) for 4 min. The complete running time of the program was 22 min. Analytes were measured in scheduled multiple reaction monitoring (MRM) (total scan time, 0.5 s; scan time window, 0.5 min). Quadrupoles were working at unit resolution.

Calibration and quantifi cation
Calibration was achieved by spiking plasma and liver samples with different quantities of acylcarnitine standards. A ten-point calibration was performed by addition of increasing amounts of each standard and IS, as described in the sample preparation section. Calibration curves were constructed and fi tted by linear regression without weighting. Data analysis was done using Analyst 1.5 ® software (AB Sciex).

Animal experiment and sample collection
C57BL6/N mice (n = 10), at an age of 12 weeks, were injected intraperitoneally with a single dose of STZ (180 mg/kg body weight dissolved in 0.1 M citric acid buffer) and mice (n = 10) injected intraperitoneally with citric acid buffer served as controls. All mice were fed a standard chow diet (ssniff V1534-0 R/M-H) and were not fasted prior to sampling of blood and organs. Body elusive. Additionally, the abundance of many species in biological samples is often too low for robust detection and quantifi cation.
Here we describe a robust and highly sensitive LC-MS/ MS method for comprehensive quantifi cation of an array of acylcarnitine species with a special focus on amino acidderived intermediates, which often occur at very low concentrations and in isomeric forms. Because alterations in plasma odd-numbered fatty acids were recently shown to associate with impaired glucose tolerance and diabetes ( 13 ), we also focused on quantifi cation of the corresponding odd-numbered acylcarnitines. The method was validated and applied to plasma and liver samples from streptozotocin (STZ)-induced insulin-defi cient mice, which serve as a model of type I diabetes.

Synthesis of reference substances
O-acylated acylcarnitines were synthesized from carnitine chloride and free fatty acids using a modifi ed protocol from Ziegler et al. ( 14 ). Briefl y, the free fatty acid was incubated with thionyl chloride (15 mg free fatty acid per microliter thionyl chloride) under continuous shaking for 4 h at 70°C to generate fatty acyl chlorides. Carnitine hydrochloride dissolved in trichloroacetic acid was added to the acyl chloride and incubated at 45°C for 18 h. After cooling, the product was precipitated and washed three times in cold diethyl ether. Sodium-(s)-␤ -hydroxyisobutyric acid was fi rst treated with hydrochloric acid to produce the free fatty acid, followed by evaporation of water at 90°C for 30 min, before incubation with thionyl chloride. For (S)-and (R)-␤ -hydroxybutyric spectra were acquired. Butylation of acylcarnitines (especially of dicarboxylic species) increased the ionization effi ciency (data not shown). Representative fragmentation patterns for valeryl-carnitine (C5:0-CN; monocarboxylic) and 3-methylglutaryl-carnitine (C5:0-M-DC-CN; dicarboxylic) are shown in Fig. 1 . Fragmentation patterns of acylcarnitines were reported to have a prominent fragment ion at m/z 85 in common ( 15 ). This fragment ion displayed the highest intensity in our measurements and was thus used for quantifi cation. It is proposed to be formed due to loss of the butyl group(s) ( m/z 56), the acyl chain [ m/z 102 and m/z 146 for valeryl-carnitine (C5:0-CN) and 3-methylglutaryl-carnitine (C5:0-M-DC-CN), respectively], and the trimethylamine fragment ( m/z 59), as initially described by Chace et al. ( 15 ) in 1997. A compilation of all acylcarnitine species covered by our method, including the corresponding MS parameters, is shown in Table 1 .

Chromatographic separation and identifi cation of isomeric acylcarnitine species
Because all acylcarnitines form a similar product ion ( m/z 85) and many compounds are isobaric or occur as positional isomers, appropriate chromatographic separation is essential. Chromatographic separation was established using a C18-reversed phase HPLC column (length, 15 cm; internal diameter, 3.0 mm; particle size, 3.5

m).
We favored an HPLC column over an ultra-performance LC column, because larger sample volumes of complex matrices were injected . Peaks were well shaped, sharp ( ‫ف‬ 0.1 min; Fig. 2 ), and well separated. A coelution of analyte and IS was achieved, which is important to compensate for matrix effects and varying ionization effi ciencies during gradient elution ( 16 ). The addition of HFBA to the eluents as an ion-pairing reagent ( 17 ) improved peak weight and blood glucose levels were measured daily using tailvein puncture. After 5 days, mice were anesthetized with isofl urane. Blood was collected into EDTA-coated tubes and was centrifuged for 10 min at 1,200 g and 4°C. Plasma was separated and snap-frozen in liquid nitrogen. Mice were euthanized and liver tissue was collected and immediately snap-frozen in liquid nitrogen. All samples were stored at Ϫ 80°C until the day of measurement. The research was conducted in conformity with the Public Health Service policy on humane care and use of laboratory animals . All experiments were performed according to the German guidelines for animal care and were approved by the state of Bavaria (Regierung von Oberbayern) ethics committee (reference number 55.2.1-54-2532-22-11).

Statistical analysis
Levels of signifi cance between metabolite concentrations in STZ-treated and control mice were assessed using two-sided independent Student's t -tests.

Acylcarnitine fragmentation
Mass spectrometric analysis of acylcarnitine species was performed using ESI in positive mode and product ion  Figure 2A shows a typical chromatogram (total ion count) derived from a liver sample containing acylcarnitines ranging from 0 to 18 carbon atoms. Retention times increased with the number of carbons and decreased with the number of double bonds . Because >50 mass transitions were required to quantify all metabolites, data were acquired in the scheduled MRM mode to yield more and shorter scan cycles per peak, to facilitate peak integration and data analysis. corresponding IS, analyte-IS pairs were selected according to the closest retention times and the best accuracies for these analyte-IS allocations [e.g., nonanoyl-carnitine (C9:0-CN) to d 3 -octanoyl-carnitine (d 3 -C8:0-CN) ]. For all analytes and matrices, calibration curves were linear and correlation coefficients (r 2 ) were greater than 0.99 ( Table 2 ). Limits of detection (LODs) were determined as the signal-to-noise ratio of three, and limits of quantitation (LOQs) were calculated as the triple fold of the LOD and were sufficient enough for quantification of acylcarnitine species in plasma and liver samples ( Table 2 ).

Calibration and quantifi cation
For quantifi cation of metabolite concentrations and to compensate for variation in sample preparation and ionization effi ciency, an acylcarnitine standard mixture containing 13 deuterium-labeled acylcarnitine species was added to the samples ( Table 2 ). Calibration lines were generated by addition of different concentrations of acylcarnitine species to a pool of liver and plasma samples. The ratio between analyte and IS was used for quantifi cation ("stable isotope dilution"). For those analytes without We found that concentration ranges of acylcarnitine species in plasma and liver were in good agreement with levels reported previously ( 21 ) Whereas almost all acylcarnitine species could be found in liver samples, the concentrations in plasma were generally far lower and a number of species (especially monounsaturated species like crotonyl-and 3-methyl-crotonyl-carnitine) could not be detected.

DISCUSSION
Acylcarnitines have recently gained considerable interest as markers for impairments in fatty acid and amino samples ( Table 4 ) . With a few exceptions, intra-and interday coeffi cients of variation (CVs) for liver and plasma samples were below 10% for high-abundant acylcarnitines (>0.01 pmol/mg tissue; >0.01 pmol/ l plasma) and between 10 and 20% for low-abundant acylcarnitines. Recoveries were calculated as the difference between area ratios of prespiked samples (addition of standards before extraction) and postspiked samples (addition of standards to the methanol extract). Recoveries between 59 and 99% were found (supplementary Table 1). Matrix effects were determined by comparing the peak areas of acylcarnitine species spiked in matrix samples (three plasma and three liver samples) and corresponding amounts dissolved in methanol at two levels (supplementary Tables 2, 3). The area ratios of spiked acylcarnitine species in plasma samples ranged between 110 and 120% compared with nonmatrix samples. Area ratios of spiked acylcarnitine species in liver were between 85 and 122% compared with nonmatrix samples, with a few exceptions for shortchain species.
To assess sample stability in plasma, samples were stored at room temperature for 1 h before extraction and peak areas were compared with samples that were directly extracted. Acylcarnitines in plasma remained stable over 1 h at room temperature. For liver tissue, thawing of samples in three freeze-thaw cycles strongly affected the abundance of various analytes, but postpreparative storage of plasma and liver samples at Ϫ 20°C showed a high stability of most analytes. See supplementary Table 4 for a summary on plasma and liver sample stabilities.

Method application
The developed method was applied to plasma and liver samples from STZ-treated insulin-defi cient and healthy control mice. STZ causes pancreatic ␤ -cell destruction resulting in insulin-defi cient type I diabetes ( 20 ). STZtreated animals showed very low circulating insulin levels accompanied by a dramatic weight loss and a severe increase in blood glucose levels 5 days after treatment (see supplementary Table 5). ion at m/z 85 for all analytes, which was described to be a specifi c product ion of acylcarnitine fragmentation ( 15 ). LC with a C18-reversed phase HPLC column allowed the chromatographic separation of isomeric acylcarnitine species, and by spiking with reference compounds, we could subsequently assign the individual isomers to the respective peaks. Proper chromatographic separation is mandatory because positional isomers have similar mass transitions [e.g., crotonyl-carnitine (C4:1-CN) and methacrylyl-carnitine (2-M-C3:1-CN), m/z 286 → m/z 85]. In addition, quantifi cation of acylcarnitine concentrations in a direct infusion approach is challenged by potential interfering matrix components which lead to an overestimation of analyte concentrations ( 9 ). Although a separation of underivatized acylcarnitine isomers has been described ( 11 ), we applied a butylation to achieve an enhanced sensitivity, particularly for dicarboxylic acylcarnitines, as described previously ( 24 ). To further improve chromatographic separation and to enhance peak shapes, the ion pairing agent, HFBA, was added to the eluents in very low concentrations (0.005%). HFBA was favored over the stronger trifl uoroacetic acid, which is more commonly used ( 9 ), because trifl uoroacetic acid causes very strong ion suppression ( 25 ). We could achieve baseline separation of all acylcarnitine species including positional isomers, and all peaks were well shaped. The total running time of 22 min is in the range of other reported methods (9)(10)(11).
Calibration was performed by stable isotope dilution. A coelution of standards and available ISs was achieved, which is important for compensation of matrix effects and differences in ionization effi ciency ( 16 ). LODs and LOQs were sufficient to determine analytes in plasma and/or liver (supplementary Table 6). In addition, we observed that matrix effects of liver tissue are comparable to other tissue types such as muscle, heart, or kidney tissue, and thus suggest that the method can be applied to other tissues to determine analyte concentrations in these matrices (data not shown). Importantly, extraction of increased tissue quantities (>40 mg) does not improve method performance, because this leads to increased noise and reduced signal-to-noise ratios. In general, for comparison of analyte concentrations between samples, it is recommended to extract similar quantities of tissue to reduce variation resulting from matrix effects.
To evaluate LC-MS/MS method performance, we validated our method in plasma and liver samples as described previously ( 26 ). Precisions showed CVs <15% for highabundant metabolites and CVs <20% for low-abundant compounds. Accuracies displayed CVs of ±10% with a few exceptions. This is a major improvement over previously published methods, for which up to 30% variation was reported in plasma ( 9 ), or for which no measures were reported at all ( 11 ). We observed that plasma samples were stable when stored for 1 h at room temperature prior to preparation . Furthermore, most acylcarnitine species in plasma and liver showed a high postpreparative stability.
acid oxidation associated with the metabolic syndrome ( 5-7 ). Moreover, the metabolites were proposed to have predictive quality for disease initiation and progression ( 5 ). Various acylcarnitines occur as positional isomers and many species often appear at very low concentrations. Even though several LC-MS/MS methods have been described for quantifi cation of isomeric acylcarnitine species ( 9-12 ), a comprehensive method that covers low-abundant species and includes odd-numbered acylcarnitines is lacking.
Here we describe a sensitive and robust method for quantifi cation of individual isomeric acylcarnitine species, including low-abundant amino acid-derived forms (see Fig. 4 ). In addition, our method includes acylcarnitines with odd-numbered acyl-chain lengths, which gained increased attention recently as metabolite markers in disease states ( 5,22,23 ).
We applied ESI in positive mode for ionization. Collision-induced dissociation produced an intense product The displayed accuracy is the average of the assayed concentration (corrected by endogenous levels of plasma and liver samples) in percent of the actual spiked concentration. Concentrations in liver are given per milligram of liver wet weight. w.w., wet weight.
We found that, particularly, the branched-chain acylcarnitines, as intermediates of branched-chain amino acid degradation, were increased in plasma and liver, and among them, most strongly, propionyl-carnitine (C3:0-CN), 2-methy lbutyrylcarnitine We applied our method to plasma and liver samples of STZ-treated insulin-defi cient mice. STZ is selectively taken up into the pancreas and causes ␤ -cell toxicity via alkylation, resulting in ␤ -cell necrosis ( 20 ). STZ-treated mice showed very low circulating insulin levels paralleled by a strong increase in blood glucose and a dramatic decrease in body weight (see supplementary Table 5). while alterations in succinyl-carnitine (C4-DC-CN) concentrations remained insignifi cant. Finally, we found elevated levels of several odd-numbered acylcarnitines in diabetic animals, most signifi cantly the short-chain species, propionyl-carnitine (C3-CN), valeryl-carnitine (C5-CN), and heptanoyl-carnitine (C7-CN) in liver. Interestingly, oddnumbered fatty acids were recently proposed as potential markers of the metabolic syndrome ( 22 ). The origin and cause for these alterations in disease states is not known.
In summary, we have developed a sensitive and robust method which allows quantifi cation of a large number of acylcarnitine species, including isomeric forms derived from a variety of metabolic pathways in amino acid and fatty acid oxidation. Compared with existing methods, we expanded the palette of analytes with novel acylcarnitine species, including low-abundant odd-numbered forms. Our LC-MS/MS method was validated in plasma and liver samples from mice and was applied to samples from STZtreated type I diabetic mice. These mice display increased concentrations of BCAA-derived and odd-numbered acylcarnitine species, which were recently proposed as potential biomarkers for the metabolic syndrome.
The authors thank Mena Eidens for excellent participation in conduction of mouse experiments and sample collection.
we also detected elevated levels of the 5-carbon straightchain valeryl-carnitine (C5:0-CN), for which the metabolic origin is obscure. Increases in plasma levels of 3-carbonchain and 5-carbon-chain acylcarnitines derived from breakdown of branched-chain amino acids, as determined by a direct infusion MS/MS approach, were recently reported as important predictors for type 2 diabetes ( 5 ). Furthermore, while concentrations of 3-hydroxybutyryl-carnitine (C4-OH-CN) were strongly increased (by ‫ف‬ 40%), in diabetic mice, probably as a result of increased ketone body production ( 27 ), levels of 3-hydroxyisobutyryl-carnitine (2-M-C3-OH-CN), an intermediate in valine breakdown, displayed a strong decrease (by ‫ف‬ 50%) in both plasma and liver. For a correct interpretation of changes in these different pathways, it is thus necessary to include LC separation prior to MS for individual quantifi cation of these metabolites. Similarly, while using direct infusion MS/MS, Mihalik et al. ( 6 ) reported C4-dicarboxyl-carnitine as a marker for glucolipotoxicity in type 2 diabetes, but did not discriminate between succinyl-carnitine (C4-DC-CN) and methylmalonyl-carnitine (C3-M-DC-CN), which are interconverted via methylmalonyl-CoA mutase. With our method, we found that only levels of methylmalonyl-carnitine (C3-M-DC-CN) were significantly increased in diabetic mice,