Mitochondrial Metabolomics Using High-Resolution Fourier-Transform Mass Spectrometry

Summary

High-resolution Fourier-transform mass spectrometry (FTMS) provides important advantages in studies of metabolism because more than half of common intermediary metabolites can be measured in 10 min with minimal pre-detector separation and without ion-dissociation. This allows unprecedented opportunity to study complex metabolic systems, such as mitochondria. Analysis of mouse liver mitochondria using FTMS with liquid chromatography shows that sex and genotypic differences in mitochondrial metabolism can be readily distinguished. Additionally, differences in mitochondrial function are readily measured, and many of the mitochondria-related metabolites are also measurable in plasma. Thus, application of high-resolution mass spectrometry provides an approach for integrated studies of complex metabolic processes of mitochondrial function and dysfunction in disease.

1. Introduction

Mitochondria have a central role in metabolism, with about 10% of the proteins encoded by the nuclear genome being targeted to mitochondria. While much of mitochondrial metabolomics has been focused on the citric acid cycle and energy metabolism, mitochondria contain complete complements of intermediates to support mitochondrial replication (1), transcription (2), translation (3, 4), and post-translational modifications (5, 6). Proteolytic systems, such as the ATP-dependent protease La (LON), are also present (7), so that peptides and modified amino acids are produced. Mitochondria contain enzymes for steps of purine and pyrimidine biosynthesis (8). Mitochondria also contain enzyme systems for carbohydrate, amino acid and fatty acid oxidation and anabolic systems to use these precursors for biosynthesis and elimination of nitrogen through ureagenesis (9, 10, 11). Key catabolic processes include β-oxidation of fatty acids, requiring import of fatty acids as acyl carnitines, and oxidation of branch-chain amino acids through branched chain fatty acid intermediates (12, 13). The anabolic systems include pathways for terpene and terpenoid products, such as cholesterol and coenzyme Q, and squalene precursors for vitamin D (14, 15). Mitochondria are involved in key steps of porphyrin and heme biosynthesis (16, 17) and are important sites of activation of vitamin D and biosynthesis of steroids, including sex hormones, mineralocorticoids and glucocorticoids(18). Mitochondria contain detoxification enzymes, such as cytochromes P-450 (Cyp) and glutathione S-transferases (19, 20), and the rhodanese system for sulfur (H₂S and thiocyanate) metabolism (21).

Although targeted analyses are available for most of these pathways, studies of integrated mitochondrial metabolism become costly and complex when large numbers of targeted analyses are required. High-resolution metabolomics of isolated mitochondria provides an alternative approach, capturing intermediates from 136 out of 154 mammalian metabolic pathways (Fig 1) (22).

Figure 1.

Metabolites extracted from isolated liver mitochondria map to 136 of the 154 metabolic pathways found in the KEGG human metabolic pathways database. The black dots (•) represent matched metabolites [Reproduced by permission of Plos One (22)]

High-resolution mass spectrometry represents an under-utilized analytic approach for study of metabolism in biologic systems. Comisaro and Marshall (23) described Fourier-transform (FT) Ion Cyclotron Resonance Mass Spectrometry in 1974, and noted advantages in terms of mass resolution, mass accuracy and sensitivity compared to other mass detectors. While triple quadrupole instruments are frequently described as superior in terms of sensitivity for specific targeted analyses, the capacity of FT instruments to measure greater than 20,000 metabolites in plasma (24) shows that high-resolution instruments are superior in terms of detection of large numbers of low abundance metabolites in complex biologic matrices. Multiple high-resolution mass spectrometers are commercially available and their characteristics were compared by Marshall and Hendrickson (25).

High-resolution mass spectrometry of mouse liver mitochondria measures metabolites of many of the pathways described based on matches to metabolic pathways obtained using pathway-mapping tools. For instance, each of the following pathways by Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis (http://www.genome.jp/kegg/) showed matches to >10 ions in a single analysis: arachidonate, amino acid, purine, pyrimidine, porphyrin, steroid and tyrosine metabolism (22). Matches to intermediates were also present for many individual pathways (Table 1). Although verification of identities by retention time and tandem mass spectrometry (MS/MS) are available for only subsets of these metabolites, such as amino acids, these data show that high-resolution metabolomics offers means to broadly evaluate metabolism of mitochondria using a single assay that simultaneously measures a large number of chemicals present in the mitochondrial metabolome.

Table 1.

Representative matches of mitochondrial features to metabolomics databases (MMCD and Metlin).

Chemical Name	m/z	Chemical Formula	Adduct
Leucine/Isoleucine	132.1015	C6H13NO2	[M+H]
Aspartate	134.0443	C4H7NO4	[M+H]
Glutamine	147.0762	C5H10N2O3	[M+H]
Glutamate	148.0595	C5H9NO4	[M+H]
Methionine	150.058	C5H11NO2S	[M+H]
Phenylalanine	166.0859	C9H11NO2	[M+H]
Tyrosine	182.0807	C9H11NO3	[M+H]
Tryptophan	205.0966	C11H12N2O2	[M+H]
Acetyllysine	189.1238	C8H16N2O3	[M+H]
N,N-Dimethyltryptamine	211.1192	C12H16N	[M+Na]
Cystathionine sulfoxide	280.0944	C7H14N2O5S	[M+ACN+H]
S-Methylglutathione	322.105	C11H19N3O6S	[M+H]
Cysteineglutathione disulfide	427.0915	C13H22N4O8S2	[M+H]
Inosine	269.0864	C10H12N4O5	[M+H]
Methylcytidine	258.1085	C10H15N3O5	[M+H]
Adenosine; Deoxyguanosine	268.1034	C10H13N5O4	[M+H]
Methylguanosine	299.1252	C11H16N5O5	[M+H]
Methyladenosine	323.1431	C11H15N5O4	[M+ACN+H]
CMP	324.0575	C9H14N3O8P	[M+H]
Succinoadenosine	384.1125	C14H17N5O8	[M+H]
ADP; Deoxyguanosine diphosphate (dGDP)	428.034	C10H15N5O10P2	[M+H]
Adenosine 5-O-(3-thiodiphosphate)	465.9958	C10H15N5O9P2S	[M+Na]
3-Dehydrocarnitine	182.0803	C7H13NO3	[M+Na]
L-Carnitine	203.1395	C7H15NO3	[M+ACN+H]
Lipoamide	206.0713	C8H15NOS2	[M+H]
Topaquinone	212.0579	C9H9NO5	[M+H]
Dihydrobiopterin; 6-lactoyltetrahydrobiopterin; 4a- carbinolamine tetrahydrobiopterin	240.109	C9H13N5O3	[M+H]
Pantothenic acid	261.1448	C9H17NO5	[M+ACN+H]
Thiamine	266.1211	C12H17N4OS	[M+H]
7,8-dihydro-L-Biopterin	281.1376	C9H13N5O3	[M+ACN+H]
Dihydropteroic acid	315.1213	C14H14N6O3	[M+H]
Dihydropteroic acid	356.1461	C14H14N6O3	[M+ACN+H]
Ubiquinone	341.1718	C19H26O4	[M+Na]
Biocytin	373.189	C16H28N4O4S	[M+H]
10-Formyldihydrofolate	472.1548	C20H21N7O7	[M+H]
Tetrahydrofolate	446.1773	C19H23N7O6	[M+H]
Estradiol	295.1661	C18H24O2	[M+Na]
Testosterone sulfate	391.153	C19H28O5S	[M+Na]
7a-Hydroxy-3-oxo-4-cholenoic acid	389.2671	C24H36O4	[M+H]
Coprocholic acid	473.3241	C27H46O5	[M+Na]
Triterpenoid	575.3022	C30H48O7S	[M+Na]
Polyprenol	245.1866	C15H26O	[M+Na]
Trihydroxy-dimethyl-tetradehydrocholecalciferol	457.3278	C29H44O4	[M+H]
Tetrahydroxy vitamin D3	490.3494	C27H44O5	[M+ACN+H]
3β,5α,9α,14α,22E,24R) Tetrahydroxy vitamin D3	502.3491	C28H44O5	[M+ACN+H]

The analytic framework for high-resolution mitochondrial metabolomics is similar to other mass spectral methods in that extraction, chromatography and other parameters can be optimized to bias the chemicals detected. The method described here is a top-down, untargeted, analytic approach (Fig 2) designed to study identified and unidentified chemicals. In practice, this is a hybrid approach because ongoing identification of mass to charge ratio (m/z) features (termed m/z feature, defined in terms of accurate mass m/z and retention time on the column) with characteristic retention times under defined conditions supports targeted analysis. High-resolution mass spectrometry is superior to other types of mass spectrometry for a top-down metabolomics approach because the mass resolution and mass accuracy of the instrument effectively detects and separates chemicals based upon accurate mass (Fig 3). With the soft ionization techniques available for high-resolution mass spectrometers interfaced to liquid chromatographic (LC) systems, the analysis can be completed without ion dissociation. Signal intensity for accurate mass to charge ratio in complex biological matrices can be used directly with stable isotope dilution for quantification with relatively short LC run times and no prefractionation prior to sample analysis (26). With advanced data extraction algorithms such as adaptive processing (apLCMS) (27) and xMSanalyzer (24), >20,000 ions can be reproducibly detected in biological extracts. In this top-down workflow, bioinformatics and biostatistical methods are then used to select metabolites or groups of metabolites for subsequent mapping to pathways or chemical database searches. This approach places identification by co-elution with standards and ion dissociation (MSⁿ) in a distal rather than proximal position in the workflow. Thus, application of high-resolution mass spectral analyses and metabolic pathway analysis using bioinformatics provides global metabolic information about mitochondria within a single analytic platform.

Figure 2.

Comparison of top-down and bottom-up workflow for metabolomics analysis. The terminology is derived from systems engineering where a top-down view entails dissection of the total system successively into more detailed parts, while a bottom up view starts with the component parts and examines how those are put together to create the whole. The high-resolution mass spectrometry-based metabolomics described here starts with measurement of all ions that are detectable and uses a workflow to successively gain insight into groups of ions that have similar biologic behavior and then selectively test to confirm identities by co-elution with standards and ion dissociation (MSⁿ) studies. This approach differs from commonly used (bottom-up) targeted metabolomics analyses in which analytic methods are developed and validated for specific metabolites prior to application. The reproducibility of LC and high-resolution mass spectrometers are sufficient to allow a hybrid approach in which known targets can be measured along with untargeted analysis of unidentified ions.

Figure 3.

Comparison of mass spectrometry (MS)-based metabolic profiling approaches for chemicals that have similar but not identical masses. (A) Analysis with gas chromatography (GC) or liquid chromatography (LC) with a single low-resolution mass detector requires separation of chemicals prior to detection. (B) Analysis with a tandem mass spectrometer using either GC or LC often does not require complete separation of metabolites prior to detection because ion dissociation and detection of product ions support identification. With this approach, prior knowledge of the ion-dissociation characteristics is generally required, mostly limiting the approach to known chemicals. (C) LC coupled to high-resolution mass spectrometry supports high-throughput analysis because chemicals are resolved by mass and have less demand for chromatographic separation. This is very useful for complex biologic systems where many unidentified chemicals are present. High-resolution instruments include Fourier-transform ion cyclotron resonance, Orbitrap (Thermo), and newer time-of-flight instruments [Reproduced by permission of Annual Reviews, (50)].

The present article describes methods for use of high-resolution mass spectrometry to measure identified and unidentified metabolites in mammalian mitochondria. The methods are generally applicable to all biological fluids and tissue extracts. In application to mitochondria, results show that much of the mammalian mitochondrial metabolome is uncharacterized. Analysis using available online databases suggests that some of these may be related to transporters of unknown function and/or to environmental agents. Comparison of mitochondrial metabolites to plasma metabolites suggest that associations may be present that could allow plasma metabolites to be developed as biomarkers of mitochondrial function in vivo.

2. Materials

Analysis of the mitochondrial metabolome by high-resolution mass spectrometry is divided into 5 major processes, mitochondrial isolation, sample preparation for metabolomics, analysis by mass spectrometry, data extraction/processing, and biostatistical/bioinformatic analyses. As indicated above, high-resolution mass spectrometers allow a different workflow to be utilized in analytic chemistry, cell physiology and medical research compared to low-resolution instruments (Fig 2). In a metabolomics analytical approach, the mass spectrometer primarily functions as a detector, i.e., an instrument used in conjunction with an analyte separation platform such as gas chromatography, liquid chromatography or capillary electrophoresis, to detect and measure a chemical based upon its characteristic m/z. In the method described here, high-performance metabolic profiling is interfaced with liquid chromatography (LC) that utilizes both anion exchange (AE) and reverse phase chromatography by parallel analytical columns via a column switching technique that is coupled to Fourier Transform ion cyclotron resonance mass spectrometry (LC-FTMS) or linear trap quadrupole (LTQ)-Velos Orbitrap mass spectrometry. Both instruments are high resolution Fourier-transform based instruments, which vary only in the motion of the ions within the mass analyzer electromagnetic field (28). Following acquisition of the mass spectrometric data, feature extraction is completed by utilizing an advanced set of adaptive processing algorithms, apLCMS (27, 29) with enhancements using xMSanalyzer (24) that result in more than double the number of features extracted using traditional data processing techniques for noise reduction and feature extraction. With the acquisition and workflow protocol described above, extracted m/z features are provided as feature tables consisting of descriptive information on thousands of ions detected reproducibly within samples.

For the analytical approach described above to work effectively, rigorous standard operating and quality control procedures are required. These include column pre-conditioning to the sample matrix prior to analysis with the sample matrix, triplicate analysis in a continuous series without interrupted solvent flow, quality control reference standards, and addition of a known concentration of representative, ¹³C- and/or ¹⁵N-labeled internal standards with an exact mass m/z and known retention time. Under these conditions, the data tables contain the accurate mass m/z, retention time (with minimum and maximum values), ion intensity, and ion intensity coefficient of variation for each detected ion on a per sample basis. The m/z features are then analyzed with biostatistics and bioinformatics methods to select subsets of ions for pathway mapping and post-hoc confirmation with retention time matching of known reference standards and MSⁿ methods (Fig 2).

2.1 Materials for Mitochondria Isolation

1. Isolation medium

2. Dounce homogenizer and 15 mL chilled centrifuge tubes

3. Homogenization buffer [220 mM mannitol, 70 mM sucrose, 2 mM Hepes (pH 7.4), 1 mM EGTA]

4. Incubation medium (220 mM mannitol, 2 mM Hepes, pH 7.0)

2.2 Metabolomic Sample Preparation

1. Acetonitrile (ACN, HPLC grade), formic acid (FA, Sigma-Aldrich), water (HPLC grade).

2. Internal standard consisting stable isotopic chemicals: [¹³C₆]-D-glucose, [¹⁵N]-indole, [2-¹⁵N]-L-lysine dihydrochloride, [¹³C₅]-L-glutamic acid, [¹³C₇]-benzoic acid, [3,4-¹³C₂]-cholesterol, [¹⁵N]-L-tyrosine, [trimethyl-¹³C₃]-caffeine, [¹⁵N₂]-uracil, [3,3-¹³C₂]-cystine, [1,2-¹³C₂]-palmitic acid, [¹⁵N,¹³C₅]-L-methionine, [¹⁵N]-choline chloride, and 2’-deoxyguanosine-¹⁵N₂,¹³C₁₀-5’-monophosphate (Cambridge Isotope Laboratories, Inc Andover, PA). Note that this mixture is arbitrary, selected to provide a diverse mixture of chemical properties and minimize cost from available stable isotopic chemicals with more than one ¹³C or ¹³C with ¹⁵N, selected to avoid overlap with natural abundance ¹³C.

2.3 Components for LC/FT-MS Analysis

1. HPLC grade water with 2% formic acid (LC/MS grade, Solvent A)

2. ACN (Solvent B)

3. HPLC grade water (Solvent C)

4. ACN with 2% formic acid (Solvent D)

5. Anion exchange (AE) column (Hamilton PRP-X110S, 100 × 2.1 mm, 7 μm)

6. Reverse phase C18 column (Higgins Analytical Targa C18, 100 × 2.1 mm, 5 μm)

2.4 Data Extraction and Processing

1. apLCMS: http://www.sph.emory.edu/apLCMS (27).

2. xMSanalyzer: http://userwww.service.emory.edu/~kuppal2/xMSanalyzer/ (24). The program depends on XCMS (Bioconductor) or apLCMS (www.sph.emory.edu/apLCMS), XML (CRAN), R2HTML (CRAN), snow (CRAN) and limma (Bioconductor).

2.5 Data Selection and Annotation (Biostatistics and Bioinformatics Tools)

1. Limma (Bioconductor)

2. Batch annotation utility (xMSanalyzer)

   This utility in xMSanalyzer uses KEGG, which links to other databases, and Metlin

2.6 Metabolite Verification

1. Coelution of authentic standard and MSⁿ requires appropriate authentic standards

2. Deconvolution MS/MS

   DeconMSn is used for deconvolution of MS/MS spectra (30)

3. Methods

3.1 Isolation of Mitochondria

A number of different extraction and separation techniques have been developed to optimize mitochondrial isolation from other cellular components, including perfusion of organs to remove blood, use of chelating agents other than EGTA, homogenization by alternative means to improve quality or yield, variations in centrifugation to optimize purity and improve separation from other organelles and variations in solutions to optimize respiratory and metabolic characteristics (31, 32, 33). Metabolite contents of mitochondria can vary significantly based on the isolation technique. This can be especially important because high-resolution mass spectrometric methods are very sensitive and can detect features from reagents and also contaminants from glassware and plastics that are introduced during mitochondria and sample preparation phases. Therefore, it is imperative that a consistent protocol is followed to assure repeatability, and that an adequate description of the methods of isolation is provided to enable comparison across multiple studies. For mitochondrial preparation, it is also necessary to consider a medium that is compatible with mass spectrometry analysis. Media such as Percoll, used to separate organelles, may contaminate the instrument and should be avoided for this application. Consequently, the present method does not separate mitochondria from contaminating peroxisomes and lysosomes; these are relatively minor components but their exact contribution is presently undefined.

1. Remove livers from mice, wash with ice-cold isolation medium twice, and place in 5 mL ice-cold-isolation medium [220 mM mannitol, 70 mM sucrose, 2 mM Hepes (pH 7.4), 1 mM EGTA].

2. Cut liver (all liver lobes) into small pieces and homogenize (15 strokes with Dounce homogenizer) in 5 mL volume of isolation medium on ice.

3. Place homogenate into 15 mL chilled centrifuge tubes and separate large cell debris by centrifuging at 600 × g for 5 min at 4°C (bench top centrifuge).

4. Transfer the supernatants (containing mitochondria) by aliquoting 750 μL each into multiple 1.5 ml microcentrifuge tubes (pre-chilled on ice) and centrifuge at 11,000 × g for 10 min at 4 °C.

5. Discard supernatant using aspiration/vacuum and resuspend the pellet in 500 μL isolation medium. Note that a loosely pelleted portion contains some plasma membrane and “light” mitochondria that do not have well coupled respiration. This portion is normally removed and discarded, keeping only the more densely pelleted “heavy” mitochondria, which have better respiratory characteristics.

6. Centrifuge at 600 × g for 5 min at 4 °C.

7. Transfer supernatant to new microcentrifuge tube.

8. Separate by centrifugation at 8000 × g for 10 min at 4 °C.

9. The isolated mitochondria will then be in a pellet form. Note that this preparation is not completely pure as it also contains peroxisomes and lysosomes, which can be separated by alternate methods (34).

10. Re-suspend the pellet in 400 μL incubation medium (220 mM mannitol, 2 mM Hepes, pH 7.0) to recover the mitochondria and for metabolomics analysis by mass spectrometry.

11. A simple assay such as measurement of the activation of mitochondrial permeability transition (MPT) should be conducted to make sure that the mitochondria are intact and functional (36). This assay is conducted by first monitoring the baseline absorbance at 540 nm. The MPT is then induced by adding CaCl₂ to a final concentration of 100 μM. If the mitochondria are intact, induction of MPT results in a progressive decrease in absorbance at 540 nm within minutes.

12. Perform protein assay (e.g., Bio-Rad protein assay) to normalize amounts of mitochondrial metabolites for metabolomics study.

3.2 Sample Preparation for Metabolomic Analysis

In optimization of detection of total number of ions (m/z features) by the chromatographic procedures as follows, we found little difference between the performance with acetonitrile and methanol for extraction. The procedure as described uses acetonitrile, with additions of deionized water to provide 2:1 ACN: water ratio. For other purposes, such as lipidomics, alternative extraction procedures should be used (35, 36).

1. A stock solution of the following extraction solution should be made (per sample): Add 2.5 μL of an internal standard mixture consisting 14 stable isotopic chemicals (see 2.2 under Materials) that cover a broad range of chemical properties represented in small molecules to 100 μL acetonitrile.

2. Add 100 μL of the above ACN solution including internal standards to 50 μL of isolated mitochondrial samples corresponding to 250 μg of mitochondrial protein.

3. Centrifuge sample at 14,000 × g for 10 min at 4° C to remove protein. Load supernatants onto an autosampler maintained at 4° C until injection.

3.3 LC/FT-MS Analysis

Optimization with ACN and a formic acid gradient for the mobile phase demonstrated that ESI in the positive ion mode produced more m/z features than negative ion mode. Correlation analyses for ions detected using C18 and AE indicated about 30% redundancy in detection, i.e., the dual chromatography method enhanced metabolic coverage by >50% compared to a single column approach. Because the dual chromatography approach washes and re-equilibrates the column during the analysis on the opposite column, the time frame for analysis is the same on the dual chromatography method as the single column approach. Direct comparison of results for known chemicals detected on both columns showed good agreement (22).

1. Inject 10 μL for each run of the LC/FT-MS analysis. Analyses are performed alternating between AE and C18 LC columns by use of a switching valve. This allows one column to undergo a wash cycle while separation is performed on the other column. The wash cycle includes a 7.5 min isocratic flow with 2% (v/v) formic acid in ACN at 0.5 mL/min. For the remainder of the 10 min, the column is re-equilibrated to initial conditions before the next sample injection.

2. AE analysis; analyte separation is performed by AE (Hamilton PRP-X110S, 2.1×10 cm) with a mobile phase formic acid gradient at 0.35 mL/min. A short, end-capped C18 pre-column (Higgins Analytical, Targa guard) is included for desalting and improved separation. The gradient for the mobile phase is provided in Table 2.

3. C18 analysis; analyte separation is performed with a C18 column (Higgins Analytical, Targa, 2.1×10 cm) and an acetonitrile gradient. See Table 3 for mobile phase conditions.

4. Analysis by a Thermo LTQ-FTICR mass spectrometer (Thermo Fisher, San Diego, CA); The mass spectrometer is operated in full scan mode over a mass range of m/z 85 to 850. Optimized operating conditions (37) are as follows; spray voltage, 4.5 kV; sheath gas, 40 (arbitrary units); capillary temperature, 275 °C; capillary voltage, 60 V; tube lens, 150 V; ion transfer optics, optimized automatically; maximum injection time, 500 msec; the maximum number of ions collected for each scan, 3 × 10⁶. The number of ions per scan exceeds that recommended by the manufacturer and represents an operational condition that improved detection of m/z features with minimal deterioration of mass resolution and accuracy. A wide range scan mode is used for the FT-ICR with mass resolution of 50,000 (37, 38). Analysis with apLCMS (27) resulted in 2872 m/z features and C18 resolved 2754 m/z features (22). Because mitochondrial isolation uses biologically derived materials that could contribute to mitochondrial signal, extracts of preparation media should also be examined to serve as analytic blanks. Importantly, mitochondrial incubation media contained about half of the m/z features found in mitochondria, indicating a need for caution related to possible artifacts of isolation. Coefficients of variation range from below 10% for many high abundance metabolites to >30% for low abundance metabolites. Thus, the number of analytic replicates must be determined based upon research needs. We have now adopted a default protocol with analysis in triplicate to improve detection of low abundance metabolites.

5. Analysis using an LTQ Orbitrap mass spectrometer; Although methods were originally developed on the LTQ-FT, due to the greater sensitivity of the LTQ Velos Orbitrap (Thermo Fisher, San Diego, CA), a successful analytical method was also developed for this mass analyzer using the dual chromatography approach. Optimized operating conditions for the LTQ Orbitrap are as follows; HESI probe with S-lens combination for ESI; MS¹ mode scanning m/z range of 85-2000; resolution, 60,000; maximum number of ions collected, 5.00 × 10⁵. The maximum injection time, 5 μL/s; capillary temperature, 275 °C; source heater; 45 °C; voltage, 4.6 kV; sheath gas, 45; auxillary gas flow, 5; sweep gas flow, 0. Each sample is run in triplicate with dual chromatography (AE and C18) and 10 μL injection volume on each column. Analysis with apLCMS (24) resulted in 13,034 m/z features using AE. Analysis on C18 resolved 14,317 m/z features. The differences in data recovery between instruments represents progress in recognizing the value of triplicate analyses to detect low abundance features on the LTQ-Velos Orbitrap (duplicates were used for analysis on the LTQ-FT), improved data extraction by xMSanalyzer (24) compared to apLCMS (27) alone, and decreased background due to use of the LTQ-Velos Orbitrap mass spectrometer strictly for metabolomics investigations while the LTQ-FT is a common use instrument in the Chemistry Department. Low abundance ions detected by the LTQ-Velos Orbitrap with xMSanalyzer but below the detection limit of the LTQ-FT with apLCMS were examined manually on the LTQ-FT to determine whether differences were due to instrumentation or data extraction. Results indicated that differences were largely due to data extraction and noise reduction, i.e., the results obtained on the LTQ-Velos Orbitrap were recapitulated on the LTQ-FT when the same samples were analyzed and specific, selected targets were examined.

Table 2.

Mobile phase gradient for anion exchange (AE) analysis of mitochondrial metabolites

Time (min)	Solvent A (2% Formic Acid in H2O), %	Solvent B (CAN), %	Solvent C (H2O), %	Solvent D (2% Formic Acid in ACN), %	Flow rate (μ,L/min)
0	5	50	45	0	350
2	5	50	45	0	350
7	50	50	0	0	350
10	50	50	0	0	350

Table 3.

Mobile phase gradient for C18 analysis of mitochondrial metabolites.

Time (min)	Solvent A (2% Formic Acid in H2O), %	Solvent B (ACN), %	Solvent C (H2O), %	Solvent D (2% Formic Acid in ACN), %	Flow rate (μ,L/min)
0	5	35	60	0	350
2	5	35	60	0	350
7	5	95	0	0	400
10	5	95	0	0	400

3.4 Data Collection and Processing

1. Data are collected continuously over the 10-min chromatographic separation and stored as .raw files. These files are converted using XCalibur file converter software (Thermo Fisher, San Diego, CA) to .cdf files for further data processing. The .raw files are time and date stamped and electronically archived as the original data for use in subsequent reanalysis as necessary.

2. An adaptive processing software package (apLCMS, http://www.sph.emory.edu/apLCMS) designed for LC-FTMS data is used for peak extraction and quantification of ion intensities (27). This software provides m/z feature tables containing m/z values, retention time, and integrated ion intensity for each m/z feature, obtained through 5 major processing steps: 1) noise filter, 2) peak identification, 3) retention time correction, 4) m/z peak alignment across multiple spectra, and 5) re-analysis to capture peaks originally missed because of weak signal relative to the signal to noise filter. This software is used on a daily basis for quality control to evaluate changes during the daily analysis sequence on a run-by-run basis: total ion intensity, internal standard ion intensity, internal standard accurate mass m/z, internal standard retention time.

3. After all sample sets are analyzed (up to 25 sets of 20), data are re-extracted as a complete dataset with xMSanalyzer (24). xMSanalyzer is an R package available at http://userwww.service.emory.edu/~kuppal2/xMSanalyzer/ that improves the performance of existing peak detection methods such as apLCMS and XCMS by allowing detection of low abundance metabolites and enhancing the quantitative reproducibility of features. In addition to this, xMSanalyzer includes utilities that allow quality assessment of both features and samples, metabolite overlap analysis between two to three datasets, and batch annotation of metabolites to known compounds and pathways in databases such as KEGG, HMDB, Metlin, LipidMAPS, CAS, and ChEBI (See Note 1). The xMSanalyzer utilities are classified into four main modules: 1) feature detection module to increase the number of quantitatively reproducible features by processing samples at two or more parameter settings, merging the resulting data, and selecting data based upon feature consistency, 2) sample quality module to support quality control analysis, 3) feature overlap module to detect overlap among multiple datasets or software packages and visualize the extent of overlaps, and 4) batch annotation module to facilitate annotation of metabolites (24). The re-extraction of all samples together is necessary to assure consistent alignment of m/z and retention time for subsequent analyses.

3.5 Biostatistics and Bioinformatics Data Analysis of High-Resolution Metabolomics

1. In the top-down metabolomics workflow (Fig 2), extracted data for ions are analyzed and discussed as m/z features without knowledge of chemical identity (See Note 2). The first step in the data analysis process is to perform feature and sample filtering based on coefficient of variation (CV) and Pearson correlation between technical replicates, respectively. Only the features that have a median CV less than 50% and the samples with Pearson correlation greater than 0.7 are used for further analysis. The technical replicates are averaged following the quality assessment, and only the features with at least 70% signal in either one of the test conditions (control, test) are retained. A log₂ transformation followed by quantile normalization is performed to reduce the effect of technical errors on downstream statistical analysis and biological interpretation. The preprocessed data is then used as input for Limma (39, 40) to perform hypothesis testing to identify differentially expressed metabolites between the experimental conditions. The p-values are adjusted for multiple comparisons using Benjamini Hochberg false discovery rate (FDR) procedure. A Manhattan plot, where the x-axis is the m/z value of the metabolite and the y-axis corresponds to the negative logarithm of the p-value for each m/z, can be used to represent the metabolome-wide association study (MWAS). An MWAS of sex-dependent differences in high-resolution metabolomics of liver mitochondria is provided in Fig 4. The significant features can then be annotated by searching metabolic databases such as KEGG and Metlin for putative matches to known chemicals using a mass to charge threshold (e.g., ±10 ppm). The selection of threshold is somewhat arbitrary due to the operational conditions of analysis and the specific ions analyzed; direct comparisons of 5 and 10 ppm show little overall difference, but some known chemicals are matched at 10 ppm that are not matched with 5 ppm threshold even though specifications and calibration indicate that <5 ppm should be obtained. The identities of the features matching known compounds are validated using MSⁿ.

2. The significance value for FDR, designated as a q-value to distinguish it from the raw p-value, should be selected according to practical issues and interpreted accordingly. A stringent q-value (smaller raw p value) decreases the risk of false discovery but can exclude otherwise important metabolites if the metabolite is of relatively low abundance and has high analytic noise. On the other hand, a liberal q value has a greater likelihood of including spurious metabolites that are only associated by chance. We use a more liberal q value at this initial level of analysis to enhance sensitivity to detect pathways that differ; subsequent statistical tests are used to determine likelihood that pathways are detected based upon multiple matches to the same pathway.

   Comparison of mitochondria from male and female mice using a liberal q value of 0.1 resulted in 246 significant features from AE and 211 significant features from C18 chromatography (Fig 4) (22). These included matches to amino acids, dipeptides, tripeptides, nucleotides and a number of other known and unknown metabolites. Of the 246 features obtained using AE chromatography, 197 features had greater ion intensity in males and 49 had greater intensity in females. Ions with greater ion intensity in males included matches to leucine/isoleucine, glutamate, and methionine. Ions with greater ion intensity in females matched adenosine, amino octadecanoic acid and sphinganine.

3. Confirmation of chemical identity of ions is obtained by MSⁿ by reanalysis of the samples, specifically to verify that the product ions in the MS/MS are correct for the tentatively matched metabolite. In the example shown, the changes in amino acids are consistent with previous findings of sex differences in mitochondrial use of amino acids (41). Furthermore, for the examples shown, the metabolites were detected on both C18 and AE chromatography, and these showed good agreement in response characteristics (Fig 5). In a practical sense, once chemicals have been identified in this analytical framework, one can maintain stringent standard operating procedures and quality control and eliminate the need to perform MS/MS on every metabolite in every sample.

4. Deconvolution of MS/MS spectra; High abundance ions that elute separately from other ions with the same nominal mass can be studied directly by MSⁿ, allowing confirmation of identities relative to authentic standards (26, 38, 42). However, for low abundance ions, direct analysis is often not possible because multiple ions with nearly identical m/z are within the ion selection window in the ion trap (Fig 6A). Deconvolution procedures are used to obtain predicted MS/MS for these ions (Fig 6B). In this procedure, analysis as a function of elution time allows selection of product ions that are correlated with precursor ions. To improve reliability in determination of these predicted spectra, replicate analyses of individual deconvolution spectra are averaged. A two-dimensional (2D) deconvolution procedure further improves reliability and also provides predicted spectra for low abundance ions that are poorly resolved by elution time (Fig 6C). In this 2D-deconvolution procedure, 5 different samples are analyzed in triplicate; product ions correlated with the co-eluting precursor ions are selected and correlated with precursor ion abundance across the series of samples to provide the predicted spectra for each precursor.

5. The statistical analysis by FDR tests individual metabolites for significant differences, while pathway effects may occur in a group-wise manner in which an entire pathway is changed, but individual metabolites do not show significant differences. To evaluate this type of pathway effect, partial least squares-discriminant analysis with orthogonal signal correction (OPLS-DA) provides an advantage over the univariate approach with FDR. Application of OPLS-DA with a statistical test (PCLS; Principal Component Loading Statistics) to identify the top 5% of metabolites that contribute to 95% separation of metabolic profiles from thioredoxin-2 transgenic mice and wild type littermates showed pathway effects even when no individual metabolite was significant by FDR. In this case, the transgenic mouse mitochondria differed from wild type littermates mitochondria in contents of a group of metabolites related to 1-carbon metabolism, including methylated nucleotides, folate and choline (22). Application of OPLS-DA with PCLS thereby provides a complementary approach to detect pathway effects.

Figure 4.

Metabolome-Wide Association Study (MWAS) of sex differences in the mouse liver mitochondrial metabolome. Significant differences in high-resolution metabolomics studies are readily visualized by a plot of the −log p as a function of m/z obtained by perform a false discovery rate (FDR) analysis. This type of plot is widely used in genome-side association studies (GWAS) and is termed a Manhattan plot. In the example shown, 211 significantly different features from C18 chromatography were obtained for a comparison between male and female mice (22) using a liberal q value of 0.1 (−log p = 1.47, indicated by arrow head). These m/z include accurate mass matches to amino acids, dipeptides, tripeptides, nucleotides and a number of other identified and unidentified metabolites.

Figure 5.

Comparison of metabolites measured by dual chromatography establishes reliability of the high-resolution method. About 30% of the ions detected using C18 chromatography are also detected using anion exchange chromatography. Some of the mitochondrial metabolites that were found to discriminate male from female mitochondria using false discovery rate analysis (q = 0.1) are shown. Metabolites with greater ion intensity in male mitochondria include (A) glutamate and (B) methionine while metabolites with greater ion intensity in female mitochondria included (C) adenosine. Data were analyzed using one-way ANOVA and Tukey’s post hoc test (* p<0.05, ** p<0.01, *** p<0.001) (22).

Figure 6.

Use of deconvolution MS/MS methods to obtain predicted MS/MS spectra of low abundance ions. The mass accuracy, mass resolution and sensitivity of high-resolution mass spectrometry allow measurement of large numbers of unidentified ions in mitochondria and other biologic extracts. Using statistical tests such as FDR, many of these are found to be significantly associated with biologic characteristics of interest. Software such as DeconMSn (30) supports generation of predicted MS/MS spectra for such low-abundance ions. In (A), a high-resolution MS spectrum illustrates low-abundance ions within the ion-selection window of the ion trap. (B) The deconvolution software selects product ions that correlate with the precursor ion designated by an arrow in Panel A, allowing generation of a predicted MS/MS spectrum. (C) Because the ions are present at low abundance within a complex mixture, the process needs to be repeated for multiple samples to confirm results and average data to obtain a reliable predicted MS/MS spectrum. This 2-dimensional deconvolution MS/MS approach also has an advantage that predicted spectra are obtained for co-eluting ions not resolved by 1-dimensional deconvolution with respect to elution time.

3.6 Functional Interactions of the Mitochondrial Proteome

1. MitoCarta, a mitochondrial proteome database contains over 1000 mitochondrial proteins, including about 150 enzymes for which substrates and products are known. This provides a basis to use metabolomics to develop integrated models of mitochondrial function. In an analysis of plasma amino acid profiles during an oral glucose tolerance test, patterns of plasma amino acids transported by the glutamate-aspartate transporter (SLC25A13) were correlated (43), indicating that a) mitochondrial transport is an important determinant of the mitochondrial metabolome and b) that groups of plasma metabolites may be useful to monitor mitochondrial function in vivo. The richness of metabolic information provided by high-resolution metabolomics will allow such functional links between mitochondrial function and accessible fluids to be established and used for medical diagnostics and disease management. Thus, the functional interactions of mitochondria can be addressed in a global manner by comparing mitochondrial metabolomics to other biological compartments.

2. Comparison of the mitochondrial metabolome to the plasma metabolome provides information about plasma metabolites that could serve as biomarkers of mitochondrial function in vivo (See Note 3). By eliminating metabolites from the >150 metabolites predicted from the MitoCarta list that are also substrates and products of cytoplasmic enzymes, an enriched list of 40 potentially selective mitochondrial indicators is obtained which could potentially be used as plasma indicators of mitochondrial function. Of the 40 potential mitochondrial metabolites selected that could be present in plasma, 30 of the metabolites matched m/z detected in human plasma from a cohort of 99 samples (Table 4). This high-resolution metabolomics data indicates that matches are present in human plasma for many expected adducts, therefore suggesting that with appropriate model development and testing, plasma metabolomics may be useful to monitor mitochondrial function as a component of routine healthcare and act as in vivo biomarkers of mitochondrial dysfunction.

   The MitoCarta list is thought to include most, but not all, mitochondrial enzymes. A comparison of features found in mitochondria with those detected in human plasma using the getVenn function of xMSanalyzer (24) is shown in Fig 7. The results show that 1425 (17.5%) among 8130 ions were common between plasma and mitochondrial extracts and matched within 10 ppm (Fig 7A). Despite this, the projected list of metabolites from about 150 mitochondrial enzymes is very small relative to the 6705 unique ions detected in isolated mouse liver mitochondria. Consistent with published data for the LTQ-FT data showing only 745 ions matching to KEGG human metabolic pathways (Fig 1), analysis of the 6705 ions in liver mitochondria analyzed using the LTQ-Velos Orbitrap system showed only a small percentage that matched known metabolites in the mammalian metabolic databases (Fig 7B). These high-resolution data suggest that most of the mitochondrial metabolome is uncharacterized. Thus, in addition to the mitochondria-related metabolites in Table 4, hundreds of other mitochondrial metabolites could potentially be developed as markers of mitochondrial function. A database search for the 1425 ions present in both mitochondria and plasma showed that half of these match metabolites in KEGG (Fig 7C), indicating that the 745 ions mapped in Fig 1 are extensively represented within the plasma metabolome.

3. Among the ions detected in the mitochondria isolated from mouse liver tissue using the described procedure, approximately 140 of the m/z features were potential matches to environmentally relevant chemicals. These included fungicides, herbicides, insecticides, a pesticide synergist, plant growth regulators, plasticizers, food preservatives, a flame retardant and an antiscald agent (Table 5). Among these classes, a large proportion of the chemicals are commonly used as herbicides and fungicides in agriculture and food production. For the triazine class of herbicides, metabolic products of the herbicides were observed. For instance, ions detected in the isolated mitochondria were matches for propazine, simazine and cyanazine. Three different atrazine-glutathione adducts and a cyanazine-glutathione adduct were also observed. Glutathione conjugated adducts provide one of the key detoxification mechanism for protecting mitochondria from oxidative stress (44), and the presence of these adducts could indicate a mitochondrial defense response for exposure to environmental agents. Consequently, the data also indicate that environmental agents are commonly present in mitochondria, even though the mice were maintained in a controlled research environment. Possible sources include water, food and bedding, which contain a number of cultivated vegetative products including corn, alfalfa and soybeans. Specific studies will be needed to verify identity of these matches to environmental agents.

4. The composition of the high-resolution mitochondrial metabolome depends upon the activity of transporters in the mitochondria, which allow uptake and efflux of metabolites. The detection of many ions matching environmental agents and related detoxification products raises important questions concerning the role of mitochondrial transport systems in controlling the mitochondrial metabolome. Mitochondria contain both ABC (ATP-Binding Cassette) and SLC (Solute Carrier) transporters. The former are primary-active membrane proteins that use the energy of ATP hydrolysis to transport a variety of molecules across the membranes, while the latter include passive transporters, ion-coupled transporters and exchangers. The ATP-hydrolysis based efflux is well known for ABC proteins present in the plasma membrane of human cells; the expression of these transporters in the mitochondrial membrane is less characterized. Based on homology to mouse MitoCarta genes, the human MitoCarta list supports the mitochondrial localization of 10 ABC transporters of the 48 known human ABC proteins (ABCA9, ABCB2, ABCB6, ABCB8, ABCB10, ABCC12, ABCD1, ABCD2, ABCD3, ABCF2), most without known substrates or functions. ABCB6 was shown to bind heme and porphyrins and to perform their uptake into the mitochondria (45). On the other hand, ABCD1, ABCD2 and ABCD3 were shown to be expressed in peroxisome (46, 47) and ABCB2 is involved in the pumping of degraded cytosolic peptides across the endoplasmic reticulum (48), but little is known about their presence in mitochondria. Therefore, the evidence of the actual expression of these proteins in human mitochondrial membrane is needed to further understand their possible involvement in metabolites transport.

   Among the 52 SLC gene families, SLC25 represents the mitochondrial carrier family, a large family of 53 nuclear-encoded transporters present in the mitochondrial inner membrane. Only 47 mitochondrial SLC transporters are predicted in the human MitoCarta database, but this may be incomplete due to the mouse homology based methodology. Thirty-one SLC transporters have identified functions in transport of nucleotides, citric acid cycle intermediates, amino acid transport, fatty acid transport and ion transport (49). Specificities of many of these are not well established, and sixteen putative transporters (Table 6) do not have known functions. The ability to obtain detailed information about mitochondrial metabolites using high-resolution metabolomics provides opportunities to address the important gaps in knowledge concerning these functions.

Table 4.

Selected putative mitochondrial metabolites matching ions present in plasma; Delta = theoretical m/z - detected m/z; Mass error = delta / (theoretical m/z) × 10⁶

Chemical Name	Chemical Formula	Monoisotopic Mass	Detected m/z	Adduct	Mass Error (ppm)	Comment
3-Methyl-2- oxovaleric acid	C6H10O3	130.0630	153.0540	[M+Na]+	11.775	Mitochondrial metabolite of Ile
Acetoacetate	C4H6O3	102.0317	103.0396	[M+H]+	5.278	Used with β- hydroxybutyrate to measure mitochondrial NAD/NADH redox potential
a-Ketoisocaproate	C6H10O3	130.0630	131.0690	[M+H]+	8.628	Mitochondrial metabolite of Leu
a-Ketoisovalerate	C5H8O3	116.0473	139.0381	[M+Na]+	11.196	Mitochondrial metabolite of Val
Betaine aldehyde	C5H11NO	101.0841	102.0907	[M+H]+	5.341
β-Hydroxybutyrate	C4H8O3	104.0473	127.0383	[M+Na]+	14.145	Used with acetoacetate to measure mitochondrial NAD/NADH redox potential
Carnitine	C7H15NO3	161.1052	162.1118	[M+H]+	3.515	Mitochondrial fatty acid transport
Choline	C5H13NO	103.0997	142.0634	[M+K]+	4.388	Choline oxidase is mitochondrial, produces betaine semialdehyde (to dimethylglycine)
Creatine	C4H9N3O2	131.0695	132.0780	[M+H]+	8.174
Creatine phosphate	C4H10N3O5P	211.0358	234.0246	[M+Na]+	1.752
Creatinine	C4H7N3O	113.0589	114.0655	[M+H]+	4.761
Dimethylglycine	C4H9NO2	103.0633	104.0702	[M+H]+	3.193	Mitochondrial intermediate choline to glycine
Fumarylacetoacetate	C8H8O6	200.0321	201.0418	[M+H]+	10.877	Tyr metabolism
γ-Butyrobetaine	C7H15NO2	145.1103	146.1169	[M+H]+	4.091	Precursor to carnitine
Glutamic γ- semialdehyde	C5H9NO3	131.0582	132.0649	[M+H]+	3.718
Glutaric semialdehyde	C5H8O3	116.0473	139.0381	[M+H]+	10.909	Lys degradation
Kynurenine	C10H12N2O3	208.0848	209.0955	[M+H]+	14.842	Trp degradation
Lauroylcarnitine	C19H37NO4	343.2722	366.2629	[M+Na]+	4.021	Mitochondrial fatty acid metabolism
Myristoylcarnitine	C21H41NO4	371.3036	372.3115	[M+H]+	1.688	Mitochondrial fatty acid metabolism
N-Acetylglutamate	C7H11NO5	189.0637	228.0253	[M+K]+	7.406	Regulates urea cycle
Oleylcarnitine	C25H47NO4	425.3505	426.3562	[M+H]+	3.553	Mitochondrial fatty acid metabolism
Palmitoylcarnitine	C23H45NO4	399.3349	400.3408	[M+H]+	3.194	Mitochondrial fatty acid metabolism
Pantothenate	C9H17NO5	219.1107	242.1008	[M+Na]+	3.584	Coenzyme A precursor
Protoporphyrinogen	C34H40N4O4	568.3050	569.3066	[M+H]+	9.563	Mitochondrial heme intermediate
Pyrroline-5- carboxylate	C5H7NO2	113.0477	114.0544	[M+H]+	4.356	Pro and ornithine degradation to Glu; required to connect urea cycle and citric acid cycle
Saccharopine	C11H20N2O6	276.1321	277.1387	[M+H]+	2.332	Lys degradation
Sarcosine	C3H7NO2	89.0477	90.0544	[M+H]+	5.367	Demethylated to glycine in mitochondria
Stearoylcarnitine	C25H49NO4	427.3662	428.3724	[M+H]+	2.294	Mitochondrial fatty acid metabolism
Succinic acid semialdehyde	C4H6O3	102.0317	103.0396	[M+H]+	5.278	GABA metabolism
δ-Aminolevulinate	C5H9NO3	131.0582	132.0649	[M+H]+	3.439	Mitochondrial heme precursor

Figure 7.

A subset of m/z features is present in mitochondria that match features in plasma. (A) Common m/z features were obtained using the getVenn function of xMSanalyzer by comparing 8130 m/z features in mitochondria [85-850 m/z range, extracted from 14,317 m/z features (m/z range; 85-2000)] and 7027 m/z features (85-850 m/z range) identified in plasma. (B). Of the 6705 features present only in mitochondria, 11% matched known metabolites in KEGG. C. Of the 1425 features common to mitochondria and plasma, 50% matched known metabolites in KEGG. Thus, the data indicate that mitochondria contain a very large content of unidentified chemicals and also that many plasma metabolites are likely to be useful to indirectly evaluate mitochondrial function.

Table 5.

High resolution matches of ions (m/z) measured in liver mitochondria to environmental chemicals.

Environmental Chemical	Chemical Formula	Classification	Detected m/z [M+H]+	Delta	Mass Error (ppm)
Propamocarb hydrochloride	C9H21ClN2O2	Fungicide	225.1370	−0.0005	−2.174
Cyproconazole	C15H18ClN3O	Fungicide	292.1225	−0.0014	−4.752
Triadimefon	C14H16ClN3O2	Fungicide	294.0998	0.0006	1.947
Tridemorph	C19H39NO	Fungicide	298.3084	0.0020	6.758
Benzimidazole	C16H14F3N3O2S	Fungicide	119.0606	−0.0002	−1.595
Piperonyl butoxide	C19H30O5	Pesticide synergist	339.2183	−0.0017	−5.121
Clopyralid	C6H3Cl2NO2	Herbicide	191.9608	0.0006	3.049
Acetochlor	C14H20ClNO2	Herbicide	270.1258	−0.0002	−0.841
Butachlor	C17H26ClNO2	Herbicide	312.1725	0.0000	0.066
Propazine	C9H16ClN5	Herbicide	230.1146	0.0021	9.078
Atrazine GSHa Adduct Ib	C18H30N8O6S	Herbicide	487.2083	−0.0001	−0.291
Atrazine GSH Adduct IIb	C18H29N8O6SCl	Herbicide	521.1655	0.0037	7.041
Atrazine GSH Adduct IIIb	C18H29N8O7SCl	Herbicide	537.1657	−0.0016	−2.930
Cyanazine GSH Adductb	C19H29N9O6S	Herbicide	512.2088	−0.0054	−10.477
Simazine	C7H12ClN5	Herbicide	202.0836	0.0018	8.835
Cyanazine	C9H13ClN6	Herbicide	241.0931	0.0032	13.066
Diethyltoluamide (DEET)	C12H17NO	Insect repellant	192.1371	0.0012	6.375
Acetamiprid	C10H11ClN4	Insecticide	223.0737	0.0008	3.464
Bifenazate	C17H20N2O3	Insecticide	301.1533	0.0014	4.697
Cypermethrin	C22H19Cl2NO3	Insecticide	416.0792	0.0023	5.535
Indoxacarb	C22H17ClF3N3O7	Insecticide	528.0787	−0.0007	−1.380
Pirimicarb	C11H18N4O2	Insecticide	239.1527	−0.0024	−9.946
Ethephon	C2H6ClO3P	Plant growth regulator	144.9811	0.0004	3.046
Forchlorfenuron	C12H10ClN3O	Plant growth regulator	248.0574	0.0011	4.469
Daminozide	C6H12N2O3	Plant growth regulator	161.0909	0.0012	7.487
Dimethyl phthalate	C10H10O4	Plasticizer	195.0639	0.0013	6.690
Monobutyl phthalate	C12H14O4	Plasticizer	223.0950	0.0015	6.706
Benzylbutyl phthalate	C19H20O4	Plasticizer	313.1411	0.0023	7.216
Di-n-heptyl phthalate	C22H34O4	Plasticizer	363.2507	0.0023	6.467
Diisodecyl phthalate	C28H46O4	Plasticizer	447.3435	0.0034	7.641
Phenoxyethanol	C8H10O2	Preservative	139.0744	0.0010	6.930
3-Iodo-2- propynylbutylcarbamate	C8H12INO2	Preservative	282.0013	−0.0027	−9.702
Triphenyl phosphate	C18H15O4P	Flame retardant	327.0757	0.0024	7.249
Diphenylamine	C12H11N	Anti-scald agent	170.0952	0.0012	7.158

^aGSH=Glutathione;

^bas reported in LeBlanc et al. (51) Delta = theoretical m/z -detected m/z; Mass error = delta / (theoretical m/z) × 10⁶

Table 6.

Putative mitochondrial transporters which do not have known functions

SYM	SYNONYMS	DESCRIPTION
SLC25A14	BMCP1|MGC149543|UCP5	solute carrier family 25 (mitochondrial carrier, brain), member 14
SLC25A16	D10S105E|GDA|GDC|HGT.1| MGC39851|ML7|hML7	solute carrier family 25 (mitochondrial carrier; Graves disease autoantigen), member 16
SLC25A27	FLJ33552|UCP4	solute carrier family 25, member 27
SLC25A30	KMCP1	solute carrier family 25, member 30
SLC25A34	DKFZp781A10161|RP11- 169K16.2	solute carrier family 25, member 34
SLC25A35	FLJ40217|MGC120446|MGC1 20448	solute carrier family 25, member 35
SLC25A39	CGI-69|FLJ22407	solute carrier family 25, member 39
SLC25A40	MCFP	solute carrier family 25, member 40
SLC25A43	-	solute carrier family 25, member 43
SLC25A44	FLJ90431|KIAA0446|RP11- 54H19.3	solute carrier family 25, member 44
SLC25A46	-	solute carrier family 25, member 46
C14orf68	HDMCP|HMFN1655	chromosome 14 open reading frame 68
LOC153328	-	similar to CG4995 gene product
MTCH1	CGI-64|MGC131998|PSAP	mitochondrial carrier homolog 1 (C. elegans)
MTCH2	HSPC032	mitochondrial carrier homolog 2 (C. elegans)
MCART1	CG7943|MGC14836	mitochondrial carrier triple repeat 1