Separation and characterization of cardiolipin molecular species by reverse-phase ion pair high-performance liquid chromatography-mass spectrometry.

An improved high-performance liquid chromatography-mass spectrometry method for the separation and characterization of cardiolipin molecular species is presented. Reverse-phase ion pair chromatography with acidified triethylamine resulted in increased chromatographic retention and resolution when compared with chromatography without acidified triethylamine. Using a hybrid triple quadrupole linear ion trap mass spectrometer to generate MS/MS spectra revealed three regions within each spectrum that could be used to deduce the structure of the cardiolipin molecular species: the diacylglycerol phosphate region, the monoacylglycerol phosphate region, and the fatty acid region. Cardiolipin standards of known composition were analyzed and exhibited expected chromatographic and mass spectral results. Two minor components in commercial bovine heart cardiolipin, (with the same molecular weight but different chromatographic retention times), were shown to differ by fatty acid composition: (C18:2)2(C18:1)2 versus (C18:2)3(C18:0)1. These compounds were then analyzed by HPLC-MS3 to examine specific diac ylglycerol phosphate generated fatty acid fragmentation. Also, two commercial sources of bovine heart cardiolipin were shown to have minor differences in cardiolipin species content. Cardiolipin isolated from rat liver, mouse heart, and dog heart mitochondria were then characterized and the relative distributions of the major cardiolipin species were determined.

. Separation and characterization of commercial standards of tetra-myristoyl cardiolipin (C14:0) 4 , tetra-oleoyl cardiolipin (C18:1) 4 , and bovine cardiolipin [containing tetra-linoleoyl cardiolipin (C18:2) 4 ] using the HPLC-MS/MS system. A-C show TIC and spectra of these standards injected alone. The inset spectra labeled 'MS' show the MS spectra of the [M Ϫ ] species envelope. The inset spectra labeled 'MS/MS' are the MS/MS spectra generated by selection of the fi rst mass of the cluster, which is then subjected to collision induced dissociation. D is of a mixture of the three standards.
To improve the chromatography of cardiolipin molecular species, we added acidifi ed triethylamine to the chromatographic eluents. This positively charged eluent component dynamically forms ion-pairs with the negatively charged phosphates in the cardiolipin molecule, resulting in increased chromatographic retention and greatly improved chromatographic resolution. To improve the mass spectral information generated from the fragmentation of cardiolipin molecular species, we used a linear ion trap that had no inherent low mass cut-off, and therefore the MS/MS spectrum of cardiolipin contained the fatty acid fragment. These changes have resulted in a substantial enhancement to our system, which we now report below.

Biological samples
All animal protocols used were approved by the Animal Care and Use Committee of Case Western Reserve University. C57BL/6 mice, 6 months of age, were obtained from the National Institute of Aging. Mice were euthanized by cervical dislocation and their hearts were rapidly excised. Three mouse hearts were combined for each experiment and their cardiac mitochondria were isolated using the procedure of Palmer ( 13 ), except that trypsin was used as the protease ( 14 ). Cardiac mitochondria exist in two functionally distinct populations within the myocyte. Subsarcolemmal mitochondria (SSM) are located underneath the plasma membrane and interfi brillar mitochondria are situated among the myofi brils. For the purposes of this report, SSM mitochondria were used. Male Sprague-Dawley rats (250-300 g) were obtained from Charles River Laboratories (Portage, MI). Animals were euthanized by decapitation and their livers were rapidly excised. The mitochondria were isolated by differential centrifugation, yielding a mitochondrial protein concentration of 80-100 mg/ml ( 15 ). Further purifi cation of rat liver mitochondria was achieved using self-forming Percoll gradient centrifugation ( 16 ). Dog heart mitochondria were obtained from a previously reported study ( 17 ).
Isolated mitochondria (1.5 mg, diluted to a fi nal volume of 0.25 ml using 50 mM KCl) were extracted using the method of Folch et al. ( 18 ). Samples were then subjected to silica gel chromatography and the lipid classes were serially eluted in six fractions using solvents of increasing polarity ( 19 ): Fraction 1 contained cholesterol esters and was eluted with isooctane:ethylacetate (95:1), fraction 2 contained triglycerides and was eluted with isooctane:ethylacetate (20:1), fraction 3 contained cholesterol and diglycerides and was eluted with isooctane:ethylacetate (75:25), fraction 4 contained free fatty acids and was eluted with isooctane:ethylacetate:acetic acid (75:25:2), fraction 5 contained monoglycerides and was eluted with isooctane:ethylacetate:acetic acid (75:25:2), and fraction 6 contained the combined phospholipids and was eluted with methanol. This fi nal fraction, containing cardiolipin, was evaporated to dryness while the other factions were stored for other experiments. Cardiolipin was isolated from fraction 6 by normal phase HPLC as described ( 10 ). 3 The HPLC-MS/MS system used an HP1100 series quaternary pump with an on-line degasser, autosampler, and column heater (Agilent Technologies, Wilmington, DE). The column used was a Symmetry® C 18 5µm, 150 × 3.9 mm analytical column (Waters Corporation, Milford MA). The column heater was operated at 35°C. Two eluents were used: eluent A contained 450 ml acetonitrile, 50 ml water, 2.5 ml triethylamine, and 2.5 ml glacial acetic acid, whereas eluent B contained 450 ml 2-propanol, 50 ml water, 2.5 ml triethylamine, and 2.5 ml glacial acetic acid. The gradient was formulated as follows: 50%B for 5 min, 50%B to 80% B over 10 min, 80% B to 100% B over 15 min, and hold 100% B for 10 min. The fl ow rate was 400 µl/min. The mass spectrometer used was a 3200 Q TRAP® hybrid triple quadrupole / linear ion trap mass spectrometer, with a Turbo V™ ion source (Applied Biosystems/ MDS SCIEX, Concord, Ontario, Canada). The source was operated in the Turbo IonSpray mode with instrument parameters including curtain gas: 10.00, ion spray voltage: -4500.00, temperature: 500°C, nebulizer gas: 60.00, and heater gas: 50.00. The instrument's linear ion trap mode was employed with an information dependent acquisition method, which included a survey scan in the enhanced MS (EMS) mode followed by an enhanced product ion scan on the largest ion in the EMS scan. The EMS scan was performed at a rate of 250 Da/s with a spectral range of either 1400-1600 Da (for biological samples) or 1200-1600 Da (for the standards), whereas the enhanced product ion scan was performed at a rate of 1000 Da/s with a spectral range of 100-900 Da. HPLC-MS 3 was performed using the same source parameters as HPLC-MS/MS. HPLC-MS 3 data collection used a three-experiment method, with each experiment having 1451.70 Da as the fi rst precursor and the second precursors were 695.20, 697.20, or 699.20 Da. The MS 3 spectrum was collected at a scan rate of 250 Da/s with a spectral range of 278-285 Da. A six port valve was employed to divert the fi rst 3 min of HPLC eluent to waste to prevent unretained compounds from fouling the ion source.

HPLC-MS/MS and HPLC-MS
Aliquots (100 nmoles) of cardiolipin standards were evaporated to dryness with nitrogen and reconstituted in 200 µl 50/50 eluent A/eluent B. These standards were both individually inected (5 µl) into the HPLC and mixed in equal proportions and injected (5 µl). For biological samples, aliquots (100 µl) from the normal phase HPLC cardiolipin containing fraction were withdrawn and evaporated to dryness with nitrogen, and reconstiuted in 100 µl 50/50 eluent A/eluent B. These were injected (60 µl) into the HPLC-MS/ MS system. This represents a 20-fold decrease in the amount material used for analysis from our previous procedure ( 10 ). Figure 1 shows the separation and characterization of commercial standards of tetra-myristoyl cardiolipin (C14:0) 4 , tetra-oleoyl cardiolipin (C18:1) 4 , and bovine cardiolipin (containing tetra-linoleoyl cardiolipin (C18:2) 4 ) using the HPLC-MS/MS system. Figure 1A-C show total ion chromatograms (TIC) of each of these standards injected alone. Figure 1D is of a mixture of the three standards and shows the separation of these cardiolipin molecular species. The inset spectra labeled 'MS' show the MS spectra of the [M Ϫ ] species envelopes of 1240, 1456, and 1448 m/z for tetra-myristoyl, tetra-oleoyl, and tetralinoleoyl cardiolipin, respectively. These spectra show the resolution of the 13 C species using the linear ion trap with data collection set at the highest resolution (250 Da/s). The inset spectra labeled 'MS/MS' are the MS/MS spectra generated by selection of the fi rst mass of the cluster, which is then subjected to collision induced dissociation. Three distinctive areas of the resulting spectra are characteristic of each cardiolipin species: the diacylglycerol phosphate region ('D'), the monoacylglycerol phosphate region ('M'), and the fatty acid region ('F'). In the case of tetralinoleoyl cardiolipin ( Fig. 1C ), the diacylglycerol phosphate mass is 695 m/z , the monoacylglycerol phosphate mass is 415 m/z , and the fatty acid mass is 279 m/z . Structures consistent with these fragments are shown in Fig. 2 . Figure 3 shows an expanded view of the bovine heart cardiolipin species shown in Fig. 1C , and focuses in on the cardiolipin molecular species other than (C18:2) 4 cardiolipin. In  Figure 3B and C are two cardiolipin species with the same mass (1452 Da, MS spectra) but these two chromatographic peaks are identifi ed as di-linoleoyl-di-oleoyl cardiolipin [(C18:2) 2 (C18:1) 2 ] and tri-linoleoyl-mono-stearoyl cardiolipin [(C18:2) 3 (C18:0) 1 ], respectively, based on the MS/MS 'D', MS/MS 'M', and MS/MS 'F' spectra. In addition, a species identifi ed as (C18:2) 3 (C18:3) 1 cardiolipin was observed with a mass of 1446 Da, the diacylglycerol phosphate spectrum containing fragments at 693 and 695 m/z , the monoacylglycerol phosphate spectrum containing fragments at 413 and 415 m/z , and the fatty acid spectrum containing fragments at 277 and 279 m/z (see Table 1 ).

RESULTS AND DISCUSSION
To determine the relative distribution of the different cardiolipin species, we used the Analyst® software package included with the 3200 Q TRAP to formulate XIC (extracted ion chromatograms) of each of these cardiolipin molecular species based on chromatographic retention time and the MS spectral response data. We then determine the peak areas of these different cardiolipin species. Calculation of the relative amounts was performed by determining the total peak area (sum of all the peak areas determined within a chromatogram), dividing the individual peak areas by the total peak area, and multiplying by 100%. The response of all the cardiolipin molecular species is assumed to be nearly identical because the phosphate species common to all cardiolipin species is largely responsible for the MS response ( 8 ). Bovine heart cardiolipin purchased from Avanti was then compared with bovine heart cardiolipin purchased from Sigma and their compositions were slightly different, as shown in Table 1 . Although (C18:2) 4 is the major cardiolipin observed from both sources, the material from Sigma contains three times more (C18:2) 3 (C18:3) 1 cardiolipin than the Avanti material. Additionally, the Sigma material has seven times less of the (C18:2) 3 (C18:0) 1 cardiolipin than the Avanti material.
Following these experiments, analysis of multiple control samples from various animal protocols was performed. Figure 4 shows TICs generated from control rat liver mitochondria, mouse heart mitochondria, and dog heart mitochondria. Figure 5A shows a chromatogram and spectra of cardiolipin species isolated from Sprague-Dawley rat liver mitochondria. There is an additional peak appearing as a shoulder on (C18:2) 4 cardiolipin, and the spectra displayed are from this additional species. The molecular weight of this species is 1422 Da and pattern of ions is consistent with (C18:2) 3 (C16:1) 1 Figure  5B shows heart mitochondria (SSM) from the C57BL/6 mice. There are three additional peaks appearing slightly before or slightly after (C18:2) 4 cardiolipin. The molecular weight of the fi rst species is 1543 Da and the pattern of ions consistent is with (C18:2) 2 (C22:6) 2 : 695 and 743 m/z -'D, 415 and 463 m/z -'M', and 279 and 327 m/z -'F', for a mass difference of 48 m/z for each of these pairs of ions ( Fig. 5B ). The molecular weight of the second species is 1496 Da and the pattern of ions consistent is with (C18:2) 3 (C22:6) 1 : 695 and 743 m/z -'D', 415 and 463 m/z -'M', and 279 and 327 m/z -'F'. The molecular weight of the third species is 1498 Da and the pattern of ions is con-sity ratios using HPLC-MS 3 . In this experiment, 1452 m/z was fi rst isolated and fragmented, then the diacylglycerol fragments (695, 697, and 699 m/z ) were isolated and fragmented. We then compared fatty acid ion intensity ratios generated using HPLC-MS 3 with those observed using HPLC-MS/MS. Figure 3B shows that the diacylglycerol fragment for (C18:2) 2 (C18:1) 2 is 697 m/z and fatty acids fragments are 279 m/z (C18:2) and 281 m/z (C18:1). As shown in Fig. 6 , when using HPLC-MS 3 , (C18:2) 2 (C18:1) 2 displays the presence of 279 m/z and 281 fatty acids when fragmenting 697 m/z . This is consistent with the expectation that this diacylglycerol fragment would contain two C18:2 and two C18:1 fatty acids. Also, because the MS/MS fragmentation that forms the diacylglycerol fragments   Table 2 .
In addition to our procedure, other approaches for the analysis of cardiolipin molecular species include shotgun lipidomics and LC-MS methods. In all three approaches, biological samples are subjected to a lipid extraction. Shotgun lipidomics then operates directly on this extract and relies exclusively on the mass spectrometer for selectivity ( 8 ). LC-MS methods ( 9,20 ) often include normal phase chromatography, in which different classes of lipids are separated from each other. However, in normal phase chromatography, the individual molecular species of lipids are not chromatographically resolved; therefore, cardiolipin molecular species are differentiated by mass spectrometry only. With our method, we perform a Folch extraction, selectively isolate phospholipids by small column silica gel chromatography, selectively isolate cardiolipin by normal phase HPLC, and then separate the cardiolipin molecular species by reverse-phase ion pair HPLC. Therefore, the selectivity of our procedure is due to two-dimensional chromatography (normal phase followed by reverse-phase ion pair) in addition to mass spectrometry. Because we chromatographically separate cardiolipin constitutional isomers prior to mass spectrometry, we are able to unambiguously establish the presence of different cardiolipin species with the same mass [e.g., (C18:2) 2 (C18:1) 2 and (C18:2) 3 (C18:0) 1 ]. This is not possible using methods that do not chromatographically resolve cardiolipin molecular species (i.e., either shotgun lipidomics or LC-MS), because reliance on mass selectivity will be ineffective for differentiating isomeric or isobaric species. HPLC with fl uorescence detection of derivatized cardiolipin does chromatographically resolve cardiolipin molecular species ( 6 ). However, our method with ion pair chromatography displays good resolution of cardiolipin molecular species without the need for derivatization and is much more selective because of mass spectrometric detection. Some methods suggest that the MS/MS ion intensity ratios of the fatty acids can be used to determine the fatty acid composition of cardiolipin species because intensity ratios appear to refl ect the amount ratios of the fatty acids ( 8 ). From the data shown above, that is not always true in the case of our method, and this suggests that employing ion intensity ratios for determining fatty acid composition should be used with caution.
In conclusion, we describe an improved HPLC-MS/MS method for the separation and characterization of cardiolipin molecular species. This was accomplished with reverse-phase ion pair chromatography and a hybrid triple quadrupole linear ion trap mass spectrometer. We were able to chromatographically separate cardiolipin isomers: e.g., (C18:2) 2 (C18:1) 2 cardiolipin and (C18:2) 3 (C18:0) 1 cardiolipin. The MS/MS spectra generated by the linear ion trap, which has no inherent low mass cut-off, contained fragments ions derived from the diacylglycerol phosphate, monoacylglycerol phosphate, and fatty acid components of cardiolipin, allowing for characterization of cardiolipin molecular species found in commercial standards and biological samples. The relative distribution of the major cardiolipin species observed in both commercial standards and isolated mitochondria were determined, illustrating compositional differences in cardiolipin from different sources.