Article pubs.acs.org/ac

Integrated Quantification and Identification of Aldehydes and Ketones in Biological Samples David Siegel,† Anne C. Meinema,‡ Hjalmar Permentier,§ Gérard Hopfgartner,⊥ and Rainer Bischoff*,† †

Department of Pharmacy, Analytical Biochemistry Group, University of Groningen, Antonius-Deusinglaan 1, Building Code XB20, level 6, Groningen, 9713 AV, The Netherlands ‡ Biomolecular Sciences and Biotechnology Institute, Molecular Systems Biology Group, University of Groningen, Groningen, The Netherlands § Department of Pharmacy, Mass Spectrometry Core Facility, University of Groningen, Groningen, The Netherlands ⊥ School of Pharmaceutical Sciences, Life Sciences Mass Spectrometry Group, University of Geneva, Geneva, Switzerland S Supporting Information *

ABSTRACT: The identification of unknown compounds remains to be a bottleneck of mass spectrometry (MS)-based metabolomics screening experiments. Here, we present a novel approach which facilitates the identification and quantification of analytes containing aldehyde and ketone groups in biological samples by adding chemical information to MS data. Our strategy is based on rapid autosampler-in-needle-derivatization with p-toluenesulfonylhydrazine (TSH). The resulting TSH-hydrazones are separated by ultrahighperformance liquid chromatography (UHPLC) and detected by electrospray ionization-quadrupole-time-of-flight (ESI-QqTOF) mass spectrometry using a SWATH (Sequential Window Acquisition of all Theoretical Fragment-Ion Spectra) data-independent high-resolution mass spectrometry (HR-MS) approach. Derivatization makes small, poorly ionizable or retained analytes amenable to reversed phase chromatography and electrospray ionization in both polarities. Negatively charged TSH-hydrazone ions furthermore show a simple and predictable fragmentation pattern upon collision induced dissociation, which enables the chemo-selective screening for unknown aldehydes and ketones via a signature fragment ion (m/z 155.0172). By means of SWATH, targeted and nontargeted application scenarios of the suggested derivatization route are enabled in the frame of a single UHPLC-ESI-QqTOF-HR-MS workflow. The method’s ability to simultaneously quantify and identify molecules containing aldehyde and ketone groups is demonstrated using 61 target analytes from various compound classes and a 13C labeled yeast matrix. The identification of unknowns in biological samples is detailed using the example of indole-3-acetaldehyde.

T

carbonyl-metabolites are too small and/or polar for efficient reversed phase chromatography (e.g., acetone, propionaldehyde, glycolaldehyde, glyceraldehyde, dihydroxyacetone, pyruvate, glyoxylic acid, and carbohydrates). Various approaches to resolving these issues, including different ionization principles (e.g., atmospheric pressure chemical ionization, atmospheric pressure photoionization),4 chromatographic techniques (e.g., porous graphitic carbon chromatography,5 hydrophilic interaction liquid chromatography,6 ion-pairing chromatography,7 or gas-chromatography8), and derivatization protocols, each with particular advantages and disadvantages, have been suggested.9 While their comprehensive discussion is beyond the scope of this introduction, it may be emphasized that derivatization enables performance enhancements both at the chromatography and mass spectrometry

he identification of unknown compounds remains to be a bottleneck of mass spectrometry (MS)-based metabolomics screening experiments.1 Analysts face experimental, analytical and computational challenges which are often tightly linked to physicochemical analyte properties. Certain compound classes are, for example, inherently difficult to analyze by liquidchromatography-electrospray ionization-mass spectrometry (LC-ESI-MS) and may thus escape detection. These compounds commonly have one or more of the following properties: (i) They are too hydrophilic to be sufficiently retained on common reversed phase LC columns. (ii) They show a poor ionization yield in ESI due to the absence of ionizable functional groups. (iii) They do not yield sufficiently abundant and/or selective fragment ions after collision-induced dissociation (CID). A wide range of aldehyde- or ketone-containing molecules belong to the group of nonideal LC-ESI-MS analytes.2,3 The nonconjugated carbonyl group ionizes poorly compared to other common structural motifs (cf. carboxylic acids, phosphates, amines, or phenolic hydroxyl-groups) and several important © 2014 American Chemical Society

Received: March 2, 2014 Accepted: April 18, 2014 Published: April 18, 2014 5089

dx.doi.org/10.1021/ac500810r | Anal. Chem. 2014, 86, 5089−5100

Analytical Chemistry

Article

fonylhydrazine (TSH), all sample preparation solvents, and chemicals as well as all standard compounds were purchased from Sigma-Aldrich (Zwijndrecht, The Netherlands) at the highest available purity. 13C6-glucose was purchased from Cambridge Isotope Laboratories (Tewksbury, MA, USA). Nanopure water was obtained from a Sartorius arium 611 V water purification system at a resistivity of 18.2 MΩ cm (Sartorius, Göttingen, Germany). Apparatus. The UHPLC-ESI-QqTOF system consisted of a Dionex Ultimate 3000 RS UHPLC system (Dionex, Germering, Germany) with the sample compartment permanently cooled to 4 °C, a Waters Acquity UPLC HSS T3 column with precolumn (dimensions: 150 mm × 2.1 mm; particle size: 1.8 μm; Waters, Milford, MA, USA), and an AB SCIEX TripleTOF 5600 system with DuoSpray ESI Source (AB SCIEX, Brugg, Switzerland) controlled by the Analyst 1.5.1 TF software. Preparation of Yeast Cell Sample Extracts (12C and 13 C). Preparation was done according to ref 20 with slight modifications. Yeast Saccharomyces cerevisiae, strain YSBN6 (MATa; genotype: FY3 ho::HphMX4 derived from the S288C parental strain),21 was grown in the chemical defined medium Yeast Nitrogen Base (YNB, Formedium, Norfolk, UK), supplemented with 2% glucose (12C6 or 13C6 glucose content: 20 g/L, respectively). For the 13C sample extracts, culture volumes of 12 × 85 mL were incubated in 1 L flasks at 30 °C with shaking at 250 rpm. All samples were grown to OD600 = 3.5 ± 0.2, and 0.213 g of glucose was added 10 min prior to harvesting. For harvesting, 80 mL of culture was mixed into 320 mL of methanol, precooled to −80 °C. The methanol− culture mix was kept below −40 °C in a cold ethanol/ethylene glycol bath (1:1 mix) cooled with dry ice.22 For the 12C6 samples, cultures of 10 mL in 100 mL flasks at 30 °C with shaking at 250 rpm were kept in the steady-state phase of exponential growth for more than 24 h. Samples were harvested at an OD600 between 0.8−1; an equivalent of 1.5 mL of OD600 = 1 (∼3 × 107 cells) from the culture was taken and mixed into 10 mL of methanol precooled to −80 °C. The samples were kept cold in dry ice. Cells were separated by centrifugation at −9 °C and 2500g for 15 min. Extraction was done as previously described20 with slight modifications. Briefly, samples were extracted twice with 2 mL of 60% (v) ethanol for 2 min at 78 °C. After each extraction step, the mixture was centrifuged for 1 min at 13,000g by means of an Eppendorf MiniSpin centrifuge (Eppendorf, Hamburg, Germany) and the supernatant was collected. The pooled supernatants were evaporated in an Eppendorf Concentrator 5301 (Eppendorf, Hamburg, Germany). The residue was resuspended in 1 mL of nanopure water, and the solution was ultrasonicated briefly. It was then centrifuged, transferred to a vial, and subjected to analysis. For the generation of a yeast extract to be used as global 13C-internal standard (13C-IS), the procedure was scaled to a culture volume of 500 mL. For 13C-samples, resuspension was done in extraction solvent (60% (v) ethanol). The resulting solutions were aliquoted and stored at −80 °C. Standard Solutions and Calibration Curves. Analytes were divided into groups of 10. Compounds with identical masses were assigned to different groups, so that no group contained two analytes with masses closer than 2 Da. Solid and liquid standards were weighed and dissolved/diluted in nanopure water (exception: solutions containing steroidal hormones, benzophenone, and zearalenone were prepared in 7:3 H2O/MeCN) to yield stock solutions of c = 1 mM (per analyte, respectively). The stock solutions were aliquoted and stored at −80 °C until analysis. For the construction of

stage without requiring the implementation of techniques limiting the method’s versatility, such as specialized ionization sources or ion-pairing agents like tributylamine. We and others have shown previously that phenylhydrazine derivatives in general and 2,4-dinitrophenylhydrazine (DNPH) in particular are well-suited derivatization reagents for aldehydes and ketones in biological samples with the potential to significantly enhance LC-MS/MS performance for these analytes.3,10−14 In the case of the triketone mycotoxin and food contaminant tenuazonic acid, for example, derivatization with DNPH yielded a 50-fold improvement in sensitivity and enabled both ESI+ and ESIoperation.12,15 In fact, excellent ionization performance using positive and negative electrospray polarities is a key feature of phenylhydrazones, allowing for flexibility in the design of multianalyte methods and extending the covered analyte spectrum. Building upon the above considerations, we describe here the development of a rapid, automated p-toluenesulfonylhydrazine(TSH)-based derivatization method for the global derivatization of aldehydes and ketones. This approach facilitates the comprehensive quantification of known compounds as well as the identification of unknowns in biological samples based on the detection of signature fragment ions originating from the derivatization reagent substructure. To simultaneously record MS data for both measurement scenarios, we employed ultrahigh performance-liquid chromatography-electrospray ionization-quadrupole-time-of-flight (UHPLC-ESI-QqTOF) mass spectrometry in conjunction with data independent SWATH (Sequential Window Acquisition of all Theoretical Fragment-Ion Spectra) at high mass resolution. SWATH recently became practically feasible with the introduction of TripleTOF 5600 instruments. In SWATH, precursor ions are selected with a Q1 mass window width in the range of 10−100 u and cofragmented before TOF analysis. While initially applied in proteomics,16,17 recent work suggests that this approach is equally suited for small molecule identification and quantification, with a performance superior to other QqTOF acquisition modes.18,19 While, e.g., for urine samples spiked with metabolite libraries, information dependent acquisition (IDA) approaches generally performed better with regards to the quality of MS2 spectra recorded, analyte coverage was only between 71% and 82%. SWATH achieved 100% analyte coverage and generated a spectral quality superior to MSALL, a third QqTOF acquisition concept based on comprehensive ion transmission by the Q1.19 Notably, the data-independent acquisition principle allows one to extract both targeted and nontargeted data from SWATH files.16 The combination of TSH derivatization and SWATH is hence bound to provide a maximum in analyte coverage for aldehyde and ketone containing molecules. While chemo-selective derivatization ensures sufficient chromatographic retention and ionization yield, at the same time providing structural information, SWATH enables the comprehensive detection of known and unknown compounds by means of their precursor and fragment ions. Below, we will illustrate the performance of the proposed derivatization ESI-QqTOF-SWATH strategy with respect to simultaneous identification and quantification and outline why we believe that SWATH and TSH derivatization are indeed a “perfect match”. Our concept may be adapted to a range of other derivatization strategies and compound classes.

■

EXPERIMENTAL METHODS Chemicals. Acetonitrile (MeCN, HPLC SupraGradient grade) and ethanol (analytical reagent grade) were purchased from BIOSOLVE (Valkenswaard, Netherlands). p-Toluenesul5090

dx.doi.org/10.1021/ac500810r | Anal. Chem. 2014, 86, 5089−5100

Analytical Chemistry

Article

calibration curves, aliquots were thawed and mixed to obtain a single 61-analyte stock solution (concentration level = 33.33 μM based on the analytes’ free acid forms, wherever applicable). This global stock solution was diluted in 1:1 steps to construct a 15-point calibration curve. Prior to analysis, 6 μL of standard solution was premixed with 4 μL of an extract of fully 13C-labeled yeast cells (equivalent to 7.5 × 107 cells per mL) in 60% (v) ethanol or 4 μL of 60% (v) ethanol in the autosampler needle. The concentration levels in the mixed solutions hence ranged from 20 μM to 1.25 nM, and the yeast cell equivalent was 3 × 105 cells per injection. Automated Derivatization. If not explicitly stated otherwise, all sample/standard solutions were automatically derivatized by the Dionex WPS3000 autosampler before injection. Two reagent solutions were stored in HPLC vials in the cooled autosampler compartment (T = 4 °C). Solution A contained TSH dissolved in MeCN at 333 mM; solution B was 1 M HCl. By means of a scripted autosampler program (cf. Supporting Information), 1.85 μL of solution A, 0.5 μL of solution B, and 10 μL of sample solution were consecutively drawn into the autosampler needle and mixed for 8 min by repeated drawing and ejection of a 6 μL air bubble. After completion of the mixing routine, the entire 12.35 μL was injected (100 μL full loop injection). The solvent ratio in the injected solution corresponded to the initial gradient composition (H2O/MeCN 85:15 v/v). The concentration of TSH in the reaction mix was 50 mM, and the concentration of HCl was 0.04 M. After each drawing step, the needle was washed externally in the autosampler’s wash port with 200 μL of MeCN. UHPLC Method. The UHPLC eluents were A, 10 mM ammonium acetate in H2O, and B, MeCN. The flow rate was 0.35 mL/min, and the column temperature was 50 °C. After derivatization and injection, the initial eluent composition (85:15 A/B) was kept constant for 5 min. From 5 to 25 min, it was linearly changed to 0:100 A/B. From 0 to 1.5 min and from 21 min to the end of the UHPLC run, the MS ion source was bypassed by means of a switching valve. At 25 min, the run stopped and the next sample was derivatized. During derivatization, the column was reconditioned with the starting eluent composition at 0.35 mL/min. MS Method. The ion-source parameters were as follows (ESI+/ESI-): curtain gas (N2), 25/25 au; gas 1, 50/50 au; gas 2, 50/50 au; spray voltage, 5000/4500 V; temperature, 475/ 475 °C; declustering potential, 55/55 V. Two MS methods were created for each ESI polarity. The settings are detailed in Table 1. Calibration of the MS instrument via infusion of a homemade calibrant (ESI-: purine, ibuprofen, sulfamethazine, benzobromanone, hexakis(2,2-difluoroethoxy)phosphazene, and hexakis(2,2,3,3-tetrafluoropropoxy)phosphazine in MeCN/H2O 95:5; ESI+: purine, caffeine, clomipramine, verapamil, reserpine, and hexakis(2,2,3,3-tetrafluoropropoxy)phosphazine in MeCN/H2O 95:5) was done automatically after each of the two runs using the Analyst 1.5.1 TF autocalibration function. Integration, Lower and Upper Concentration Limits, Matrix Effects. Integration was done using the AB SCIEX MultiQuant 2.6 package (AB SCIEX, Brugg, Switzerland). Precursor or fragment ion chromatograms were extracted from the corresponding SWATH experiment total ion currents (TICs) using a mass window of 0.03 Da and subjected to Gaussian smoothing (width = 1). The limit of detection (LOD) was defined as the lowest concentration with an S/N ratio higher than 3. To determine the linear dynamic range, the

Table 1. Overview of the Employed QqTOF-SWATH Methoda MS experiment 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

Q1 window range/m/z

Q1 window width/m/z

Injection 1 TOF Scan 50−1000 200−240 40 200−240 40 240−280 40 240−280 40 280−320 40 280−320 40 320−360 40 320−360 40 360−400 40 360−400 40 400−440 40 400−440 40 440−480 40 440−480 40 Injection 2 TOF Scan 50−1000 480−520 40 480−520 40 520−560 40 520−560 40 560−600 40 560−600 40 600−700 100 600−700 100 700−800 100 700−800 100 800−900 100 800−900 100 900−1000 100 900−1000 100

accumulation time/ms

collision energy/V

60 60 60 60 60 60 60 60 60 60 60 60 60 60 60

5 5 30 5 30 5 30 5 30 5 30 5 30 5 30

60 60 60 60 60 60 60 60 60 60 60 60 60 60 60

5 5 30 5 30 5 30 5 30 5 30 5 30 5 30

a

Two injections are done for each sample and ESI polarity. The MS cycle time was 0.99 s for both methods. The TOF scan range was m/z 50−1000 for all experiments.

coefficient of determination (R2) was computed for the curve from c = 20 μM to c = LOD. If R2 was below 0.99, the highest concentration data point was removed. This procedure was repeated until R2 was ≥0.99 or the number of data points fell below 5. Linearity and LOD were determined in the presence of a 13C-labeled yeast cell extract. For analytes that were fully 13 C-labeled in the yeast extract at a concentration of at least 3 × LOD (Table 2), IS correction via evaluation of the 12C/13C peak ratio was applied prior to linearity evaluation. Matrix effects (MEs) were calculated according to ME = sM × s−1 × 100, with sM and s being the slopes of the calibration curves obtained in the presence and absence of the yeast matrix, respectively. Absolute ME uncertainties (uME) were calculated from the relative standard errors of the slopes (uS and uSM) by error propagation according to uME = ME × (uS2 + uSM2)0.5. Identification and Limit of Identification. SWATH datafiles used for identification of aldehydes and ketones in 13Clabeled yeast were mass-recalibrated postacquisition with the AB SCIEX PeakView 1.2 LC-MS Peak Statistics package (AB SCIEX, Brugg, Switzerland) using the calculated exact masses of three features known to occur in the sample (derivatized 2-ketoglutarate, ketovaline, and phenylpyruvic acid). The masses of the signature fragments evaluated were m/z 155.0172 (ESI-) and 5091

dx.doi.org/10.1021/ac500810r | Anal. Chem. 2014, 86, 5089−5100

5092

cyclohexanone 7967 dihydroxyace670 tone erythrose and 94 176 threose (sum) erythrulose 5460032 glucoronolac92 283 tone glyceraldehyde 751 glycolaldehyde 756 glyoxal 7860 glyoxylic acid 760 hydroxyacetone 599 hydroxypyruvic 964 acid indole-pyruvic 803 acid ketoisoleucine 47 ketoleucine 70 ketovaline 49 malondialdehyde 10 964 mercaptopyruvic 98 acid methylglyoxal 880

Carbon Metabolites 2,3-butanedione 650 2-carboxybenzal8406 dehyde 2-ketobutyric 58 acid 2-ketoglutaric 51 acid 4-hydroxy-2-ke- 440 853 toglutaric acid 4-hydroxyphe979 nylpyruvic acid acetaldehyde 177 acetoin 179 acetone 180 benzoylformic 11 915 acid betaine aldehyde 249

1 1 2 1 1 1 1 1 1 1 2 1 2

90.0317 60.0211 58.0055 74.0004 74.0368 104.0110

203.0582

130.0630 130.0630 116.0473 72.0211 119.9881

72.0211

1

102.0919

1 1

1 1 1 1

44.0262 88.0524 58.0419 150.0317

120.0423 176.0321

1

180.0423

1

1

162.0164

120.0423

1

146.0215

1 1

1

102.0317

98.0732 90.0317

2 1

86.0368 150.0317

9.7

9.4

7.0 9.6 5.5

7.2

11.5 11.3 10.7 15.2

12.3 13.4

6.8 8.1 4.0 13.7 9.0

10.1 11.8

5.3 6.6 11.5 2.3 8.3 2.4

5.0 5.2 12.5 12.8

4.2

14.6 6.4

2.4

9.9 10.5 9.9 10.8 11.5 8.0 11.2

5.3

1.6

1.7

3.3

15.2 9.8

409.100

299.106 299.106 285.090 409.100 289.031

372.101

259.075 229.064 395.084 243.043 243.080 273.054

289.085 345.075

289.085

267.116 259.075

270.127

213.069 257.095 227.085 319.075

349.085

331.059

315.065

271.075

423.116 319.075

1 1

[M + H]+ [M + H]+

+ + + +

H]+ H]+ H]+ H]+

1 1 2 1 2 1 2 1 1

[M + H]+ +

[M + H]+ [M + H]+ 91.054 [M + H]+ 91.054 [M + H]+ 1 2 2 2 1 1 1

130.065 [M + H]+ [M + H]+ [M + H]+ 155.016 [M + H]+ [M + H]+

[M + H] 157.032

1 1

1

2 1 1 2

111.092 [M + H]+

60.081

[M [M [M [M

2

2

[M + H]+

139.021

1 1

#tR

[M + H]+ [M + H]+

name (neutral PubChem monoisotopic groups tR1 tR2 [M + H]+/ quantifier/ Da Da form) CID mass/Da derivatized /min /min

prederivatization

X

X X

X

X

X

X

0.63

0.04 0.08 0.04 0.16 ND

0.63

0.31 1.25 0.63 1.25 0.63 0.63

0.63 0.31

0.63

0.01 0.31

0.04

BW 0.08 1.25 0.31

0.31

0.16

0.63

0.63

0.16 0.04

20

20 20 20 10 ND

20

20 20 20 LIN 20 20

20 20

40

10 20

10

BW 10 10 20

20

20

20

20

10 10

C IS lower upper applied limit/μM limit/μM

13

ESI+

313.050 329.045

0.996 98 ± 4 0.985 80 ± 5

BW 99 ± 3 91 ± 7 104 ± 7

370.087 297.091 297.091 283.076 407.085 287.017 407.085

4 8 8 5

0.996 97 ± 4 ± ± ± ±

0.991 102 ± 8

0.993 0.997 0.989 0.995 ND

105 110 101 100 ND

257.060 227.050 393.070 241.029 241.065 271.039

107 ± 5 98 ± 6 99 ± 6 LIN 100 ± 5 103 ± 5 0.994 0.997 0.992 LIN 0.996 0.996

287.071 343.061

0.991 103 ± 8 0.996 99 ± 4

1 2 1 2 1 2 1 2 2 2 2 1 2

[M − H]− 155.017 [M − H]− [M − H]− [M − H]− [M − H]− [M − H]− [M − H]− [M − H]− [M − H]− 91.055 [M − H]− [M − H]−

[M − H] [M − H]−

1 2

1 −

[M − H]−

1 1 2

[M − H]− [M − H]− 91.055 117.045

287.071

2

[M − H]−

0.993 98 ± 5

1

[M − H]−

1 1

1

[M − H]−

0.996 103 ± 3 0.992 100 ± 7

211.055 255.081 225.070 317.060

2

1 1

63.963

265.075 [M − H]−

#tR

not not observed observed 265.102 63.963 257.060 [M − H]−

0.991 105 ± 6

BW 0.995 0.996 0.992

347.071

269.060

0.982 107 ± 11

0.996 103

421.101 317.060

[M − H]−/ quantifier/ Da Da

0.995 103 ± 6 0.996 103 ± 3

R2

matrix effect/%

postderivatization with p-toluenesulfonylhydrazine

Table 2. Overview of Targeted Analytes and Quantitative Performance in ESI+/ESI−a

X

X X X

X

X

2 X

X

X

0.01

0.01 0.01 0.01 0.08 ND

0.04

0.16 2.5 0.08 0.63 0.63 0.01

0.04 ND

0.08

0.31 0.31

−

0.08 2.5 0.01

0.04

0.01

0.01

0.16

0.01 0.01

20

10 10 20 20 ND

0.63

20 20 20 20 20 10

20 ND

10

5 5

−

BW 20 10 5

1.25

2.5

20

20

10 10

C IS lower upper applied limit/μM limit/μM

13

ESI−

−

BW 106 ± 4 100 ± 16 100 ± 2

106 102 100 112 100 109

± ± ± ± ± ±

110 ± 9 108 ± 5 107 ± 5 95 ± 4 ND

5 4 5 10 7 5

0.993 104 ± 4

0.993 0.999 0.998 0.993 ND

0.993 94 ± 6

0.992 0.997 0.991 0.99 0.991 0.993

0.995 96 ± 3 ND ND

0.995 102 ± 7

0.999 104 ± 12 0.992 94 ± 6

−

BW 0.991 − 0.998

0.992 96 ± 7

0.997 57 ± 5

0.992 69 ± 6

0.995 109 ± 4

0.996 102 ± 4 0.991 101 ± 2

R2

0.04

0.01 0.01 0.02 0.04 −

1.25

0.31 2.5 0.04 0.63 5 1.25

0.04 −

0.08

0.31 0.62

−

− 0.31 2.5 0.01

0.31

0.04

0.31

0.16

0.01 0.31

identificamatrix tion limit/ effect/% μM

Analytical Chemistry Article

dx.doi.org/10.1021/ac500810r | Anal. Chem. 2014, 86, 5089−5100

Carbon Metabolites (cont.) phenylpyruvic 997 164.0473 acid propanal 527 58.0419 pyruvic acid 1060 88.0160 Flavoring Agents (Naturally Occurring) 1-phenyl-1,211 363 148.0524 propanedione 2-hydroxybenzal6998 122.0368 dehyde 3-hydroxybenzal101 122.0368 dehyde 4-hydroxybenzal122 122.0368 dehyde benzaldehyde 240 106.0419 benzoin 8400 212.0837 benzophenone 3102 182.0732 cinnamaldehyde 637 511 132.0575 jasmone 1549018 164.1201 raspberry ketone 21 648 164.0837 thymoquinone 10 281 164.0837 tiglic aldehyde 5321950 84.0575 vanillin 1183 152.0473 Mycotoxins tenuazonic acid 580 202 197.1052 zearalenone 5281576 318.1467 Steroidal Hormones (cont.) aldosterone 5839 360.1937 androstenedione 6128 286.1933 androsterone 5879 290.2246 corticosterone 5753 346.2144 cortisone 222 786 360.1937 cortisone-acetate 5745 402.2042 dihydrotestoster- 10 635 290.2246 one estrone 5870 270.1620 hydrocortisone 5754 362.2093 hydrocortisone3641 404.2199 acetate pregnenolone 8955 316.2402 pregnenolone105 074 396.1970 sulfate

5093

12.0 12.7 2.7 7.6 15.0 16.7 13.4 12.9 12.6 15.3 16.1 17.1 18.5 16.2 19.2 13.7 14.1 16.5 15.4 12.8 10.7 17.0

1 1 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1

17.1 14.7 15.6 16.6 17.3 19.1 14.4

1 1 1 1 1

17.1 15.5 17.4 18.5

13.9 18.4 19.4 15.8 14.6 17.1 17.7

1 1 1 1 1 1 1

15.1 18.7

9.5 11.9

1

485.283 565.240

439.205 531.252 573.263

529.237 455.236 459.268 515.257 529.237 571.247 459.268

366.148 487.190

275.085 381.127 351.116 301.101 333.163 333.127 333.127 253.101 321.090

291.080

291.080

291.080

485.131

227.085 257.059

333.090

1 1

[M + H]+ 313.157

1 1 2 1 1

157.065 [M + H]+ [M + H]+ [M + H]+ [M + H]+

1 2 1 2 2 2 2

1 1 1 1 1 1 1 1 1

[M + H]+ 207.092 195.092 145.076 [M + H]+ 107.049 145.076 [M + H]+ [M + H]+

374.220 300.220 [M + H]+ 497.247 [M + H]+ [M + H]+ 303.243

1

[M + H]+

1

[M + H]+ 1

2

[M + H]+

135.055

2 1

1

#tR

71.060 [M + H]+

[M + H]+

name (neutral PubChem monoisotopic groups tR1 tR2 [M + H]+/ quantifier/ Da Da form) CID mass/Da derivatized /min /min

prederivatization

Table 2. continued

X

X

X

0.16 0.02

0.16 0.005 0.005

0.08 0.005 0.31 0.16 0.005 0.02 0.08

0.04 0.08

0.31 0.04 2.5 0.04 0.04 0.01 0.16 0.31 0.04

0.04

0.04

0.63

0.31

0.31 0.31

0.08

5 10

10 2.5 2.5

2.5 10 10 10 10 5 10

10 1.25

20 20 20 10 10 10 5 20 10

20

20

20

20

5 20

20

C IS lower upper applied limit/μM limit/μM

13

ESI+

225.070 255.044 483.117 289.065 289.065 289.065 273.070 379.112 349.102 299.086 331.149 331.112 331.112 251.086 319.076 364.134 485.175

0.995 100 ± 6 0.993 99 ± 4 0.989 105 ± 7 0.994 100 ± 5 0.997 99 ± 3 0.992 100 ± 5 95 ± 3 99 ± 6 106 ± 7 97 ± 4 102 ± 5 104 ± 2 102 ± 10 94 ± 5 99 ± 3

0.997 100 ± 3 0.989 100 ± 9

527.222 453.222 457.253 513.243 527.222 569.233 457.253 437.190 529.238 571.248 483.269 563.225

103 ± 8 99 ± 3 93 ± 9 106 ± 7 105 ± 3 102 ± 5 106 ± 5 0.989 82 ± 14 0.994 99 ± 4 0.994 104 ± 4 0.994 99 ± 5 0.994 96 ± 3

0.990 0.996 0.982 0.990 0.997 0.991 0.991

0.997 0.991 0.995 0.995 0.991 0.998 0.993 0.992 0.996

331.076

2 2 1 2 1 2 2 1 1 2 1 1

[M − H]− [M − H]− [M − H]− 155.017 [M − H]− [M − H]− [M − H]− [M − H]− [M − H]− [M − H]− [M − H]− 96.959

1 1

[M − H]− 301.145

1

[M − H]−

1 2 1 1 1 1 1 1 1

1

[M − H]−

[M − H]− 155.017 155.017 [M − H]− [M − H]− [M − H]− [M − H]− [M − H]− [M − H]−

2

[M − H]−

1

2 1

[M − H]− [M − H]−

63.963

2

#tR

[M − H]−

[M − H]−/ quantifier/ Da Da

0.994 96 ± 4

R2

matrix effect/%

postderivatization with p-toluenesulfonylhydrazine

X

X

0.31 0.001 25

1.25 0.0025 0.0025

0.001 25 0.001 25 0.04 0.0025 0.001 25 0.001 25 0.005

0.0025 0.02

0.02 2.5 0.005 1.25 0.01 0.16 2.5 0.01

0.005

0.0025

0.01

0.125

2.5 0.31

0.0025

10 10

10 1.25 1.25

0.63 1.25 2.5 2.5 0.63 0.63 10

10 1.25

LIN 20 20 10 20 5 2.5 LIN 2.5

10

10

1.25

20

20 20

20

C IS lower upper applied limit/μM limit/μM

13

ESI−

94 ± 6

LIN 85 ± 2 106 ± 7 103 ± 5 88 ± 8 100 ± 3 65 ± 4 LIN 97 ± 4

82 ± 3 104 ± 6 106 ± 14 101 ± 3 91 ± 3 68 ± 4 105 ± 8

0.997 105 ± 7 0.996 95 ± 3

0.989 97 ± 14 0.999 78 ± 4 0.999 81 ± 5

0.998 0.995 0.991 0.996 0.998 0.999 0.991

0.991 101 ± 10 0.992 96 ± 9

LIN 0.995 0.995 0.991 0.99 0.994 0.995 LIN 0.995

0.996 103 ± 3

0.996 102 ± 3

0.991 94 ± 5

0.99

0.992 109 ± 14 0.992 93 ± 11

0.999 99 ± 5

R2

0.08 0.01

1.25 0.01 0.01

0.63 0.01 0.31 0.005 0.01 0.01 0.02

0.04 0.04

0.31 0.02 2.5 0.02 0.61 0.02 1.25 2.5 0.02

0.01

0.005

0.01

0.25

2.5 0.63

0.01

identificamatrix tion limit/ effect/% μM

Analytical Chemistry Article

dx.doi.org/10.1021/ac500810r | Anal. Chem. 2014, 86, 5089−5100

0.54 0.994 97 ± 12 10.61 0.36 0.993 100 ± 5 14.25 0.31 average

#tR

2 1 [M + H]+ [M + H]+ 483.268 457.252 19.6 20.4 17.4 18.1 1 1 314.2246 288.2089 Steroidal Hormones (cont.) progesterone 5994 testosterone 6013

name (neutral PubChem monoisotopic groups tR1 tR2 [M + H]+/ quantifier/ Da Da form) CID mass/Da derivatized /min /min

All masses/m/z ratios shown are based on calculated exact masses. The retention times tR1 and tR2 correspond to E/Z isomer peaks observable for some analytes. The peak used for integration is indicated in the #tR columns. Acronyms: LIN = linear calibration curves could not be established, BV = calibration curves could not be established due to background value issues, and ND = not detectable.

2 1 [M − H]− [M − H]− 0.990 97 ± 8 0.992 103 ± 5 5 2.5 0.16 0.02

R2 C IS lower upper applied limit/μM limit/μM

13

ESI+

481.253 455.237

#tR [M − H]−/ quantifier/ Da Da matrix effect/%

postderivatization with p-toluenesulfonylhydrazine prederivatization

Table 2. continued

m/z 155.0161 (ESI+), corresponding to the sum formula C7H7O2S in both cases (the mass difference equals the mass of two electrons). The masses were extracted from the SWATH datafiles with a mass window width of 0.01 Da (the narrower mass window is enabled by mass-recalibration). The absence of signature fragment peaks in the underivatized sample was confirmed prior to screening by injections omitting the derivatization routine. Limits of identification were calculated for 61 target analytes as the lowest concentration at which (i) the signature fragment ion UHPLC peak could be observed with a S/N ratio of at least 3 and (ii) the mass of the [M − H]− parent ion could be observed at the same retention time with an intensity of at least 100 counts and a mass error below 5 ppm. Limits of identification were determined in the absence of the 13C-yeast matrix since the individual contributions of 12C/13C analytes to a given signature fragment UHPLC peak cannot be distinguished on the basis of the recorded SWATH data. The limit of identification can be lower than the LOD if the latter is, for selectivity reasons, obtained by evaluation of a quantifier ion which differs from the signature fragment.

■

RESULTS AND DISCUSSION Derivatization Chemistry and Automation. Among the phenylhydrazine derivatives, 2,4-dinitrophenylhydrazine (DNPH), has been used most extensively for the derivatization of aldehydes and ketones.23 However, particularly when combined with LC-MS analysis, DNPH suffers from several disadvantages such as limited solubility in aqueous solvents, explosiveness, and a tendency to deposit in the MS ion-source. In a screening of other commercially available reagents, we identified TSH as a suitable alternative to DNPH. TSH has a reactivity similar to DNPH and improved solubility characteristics (data not shown). The key advantage of TSH, however, is its volatility, which is due to its decomposition to gaseous species at elevated temperatures (decomposition temperature ∼150 °C24). The formation of hydrazones from hydrazines and carbonyls generally proceeds via a two-step mechanism implying the possibility of E/Z isomer formation.10,14,25 Analytes with more than one reactive carbonyl group can be derivatized multiple times. Carbonyl groups conjugated to heteroatoms, as in carboxylic acids and their amides or esters, are, however, not reactive under mild conditions and can thus be discriminated from aldehydes and ketones (cf. Figure 1: the derivatization of pyruvic acid proceeds exclusively via the keto-carbonyl). Figure 2 gives an overview of the derivatization procedure. The volatility of TSH allows for the injection of excess reagent with the sample solution. Due to the direct injection approach, sample solution and derivatization reagent expenditure are minimal.

Figure 1. Acid-catalyzed hydrazone formation with TSH for the example pyruvic acid.

a

1.25 0.04 0.989 114 ± 14 0.996 101 ± 12 10 1.25 1.25 0.0025

C IS lower upper applied limit/μM limit/μM

Article

13

ESI−

R2

identificamatrix tion limit/ effect/% μM

Analytical Chemistry

5094

dx.doi.org/10.1021/ac500810r | Anal. Chem. 2014, 86, 5089−5100

Analytical Chemistry

Article

Figure 2. Schematic representation of the automated derivatization routine as conducted by a Dionex WPS-3000 autosampler. Solution A: 333 mM TSH in MeCN. Solution B: 1 M HCl. All solutions were stored in HPLC vials in the autosampler compartment (T = 4 °C).

Figure 3. Tentative interpretation of the ESI-/ESI+-QqTOF CID spectra (CE 40 V) of the testosterone-TSH-hydrazone derivative.

In general, the fragmentation data allow for the following conclusions: (i) All studied carbonyls form hydrazones with TSH. (ii) ESI+-CID spectra are generally more complex than ESI- spectra. In ESI+, fragments are heterogeneous and frequently involve part of the analyte substructure (cf. Supporting Information). (iii) In ESI-, fragmentation spectra are simple and dominated by the m/z 155.0172 fragment attributable to a toluene-sulfinate ion emerging exclusively from the derivatization reagent substructure. All 61 studied hydrazones show this fragment at both low (15 V) and high (40 V) CEs. At low CE, it is the most intense fragment for

The method is directly compatible with aqueous samples, e.g., those typically encountered in metabolomics or food analysis workflows. Properties of TSH-Hydrazones in ESI-MS/MS. The ESI+ and ESI- response and CID fragmentation spectra for 61 derivatized target analytes compiled from different compound classes (Table 2) were studied at collision energies (CEs) of 15 and 40 V. A comprehensive overview of the fragmentation data is available in the Supporting Information. As a case example, the fragmentation behavior of derivatized testosterone is interpreted in Figure 3. 5095

dx.doi.org/10.1021/ac500810r | Anal. Chem. 2014, 86, 5089−5100

Analytical Chemistry

Article

maximum cycle time of 1 s, providing at least 15 data points per peak, was found to be acceptable. Table 1 summarizes the final SWATH method setup. Two injections per sample and ESI polarity were used to cover the precursor ion mass range from m/z 200 to 1000. Above m/z 600, we raised the Q1 mass window width from 40 to 100 u due to the lower number of detected features. Assuming [M − H]− and [M + H]+ ionization, the monoisotopic mass of the underivatized, un-ionized aldehyde or ketone (m0) was calculated from the mass of the respective precursor ion observed after single derivatization (mDS) or double derivatization (mDD) according to

56 out of 61 analytes (exceptions: 2-ketoglutarate, 4-hydroxy-2ketoglutarate, tenuazonic acid, pregnenolone sulfate, 2,3butanedione). For 41 analytes, it is the only ion observed apart from the residual precursor ion. Due to its relative abundance and presence in all ESI-CID spectra, the m/z 155.0172 fragment is suited as a signature fragment for derivatized aldehydes and ketones. Mechanistically, its prevalence may be explained by the high delocalization potential and electron affinity of the toluene-sulfinate group, facilitating the formation of a negatively charged ion. In ESI+, a m/z 155.0161 fragment, corresponding to the positively charged toluene-sulfinate equivalent, is observed for some analytes. However, occurrence and abundance in ESI+ were insufficient for application as a signature fragment, i.e., the fragment did occur for merely 48% of analytes at CE 15 V and for 18% at CE 40 V. SWATH. In SWATH, several precursor ions are cofragmented and appear together with their fragments in the SWATH spectra. Precursor and fragment ions thus have to be related postacquisition. This requires sufficient abundance of both ion types and thus a careful choice of the employed CEs. The ideal CE yields (i) the highest possible abundance of the signature fragment across a wide range of analytes and (ii) the highest possible abundance of the [M − H]− ions required to obtain accurate mass data for the unknown aldehyde/ketone. To ensure that both requirements were met, we employed an approach based on two QqTOF experiments per given mass window. The precursor ions were detected in a QqTOF experiment with virtually no fragmentation (CE 5 V, corresponding to the minimum CE required for ion focusing) while signature fragments were detected in a second MS experiment done at CE 30 V. This value resulted from the optimization experiment detailed in Figure 4.

m0 = mDS − 167.02846 Da and m0 = mDD − 335.06419 Da (ESI‐)

or m0 = mDS − 169.04301 Da and m0 = mDD − 337.07874 Da (ESI + )

With respect to the underivatized, un-ionized aldehyde or ketone, the method thus covers the mass range from 33 to 833 Da (ESI-, single derivatization) or 31 to 831 Da (ESI+, single derivatization). The smallest conceivable analyte is acetaldehyde (mmonoisotopic = 44.0262 Da). Formaldehyde is gaseous at room temperature and was hence not considered. The coupling of derivatization and SWATH adds a layer of defined chemical information to the accurate mass measurements provided by the TOF mass analyzer. The SWATH data obtained with our method contain (i) accurate precursor ion masses, (ii) chemical information derivable from the occurrence of the m/z 155.0172 signature fragment, and (iii) accurate masses of other precursor ion fragments. Taken together, this allows one (a) to screen for functional groups in molecules, (b) to obtain the molecule’s accurate mass/sum formula, and (c) to obtain fragment ion information with high-mass accuracy facilitating structure elucidation. Using such data, aldehydes and ketones can, for example, quickly be distinguished from carboxylic acids, esters, ethers, and alcohols of the same sum formula.2,10,13,26 Identification Exemplified. To identify molecules containing aldehyde and ketone groups in biological samples with the highest possible confidence, an injection of (i) the underivatized biological sample, (ii) a derivatized water blank, and (iii) the derivatized biological sample was performed. ESI- is the ionization polarity of choice due to signature fragment abundance (see above). Injection (i) was used to confirm the absence of signature fragments in the underivatized sample and injection (ii) to eliminate signature fragment signals that are due to derivatization artifacts. This example is based on 13C-labeled yeast, in which a total of 66 signature fragment UHPLC peaks were detectable across all SWATH windows (Figure 4). Of these 66 features, 21 could be attributed to targeted metabolites (see Quantification section) and at least 6 to 12C-solvent impurities (butanones, pentanones, hexanones) originating from ethanol used in yeast sample preparation. Figure 5, Panel A, shows the m/z 320−360 SWATH windows for injections (ii) and (iii), from which the signature fragment trace was extracted. In the resulting chromatograms, peaks that are exclusively observed for injection (iii) can be attributed to derivatized aldehydes and ketones. Peaks seen in both chromatograms are derivatization artifacts (e.g., the peak seen at retention time (tR) = 13.8 min is dimerized TSH). In Panel A, all major features but one (tR = 15.0 min) could be explained by previously determined analytes. The unknown feature was selected for further investigation. To obtain the accurate mass of the unknown precursor ion, the summed mass spectra for the 14.95−15.05 min tR range were extracted from the TICs of the low CE experiment

Figure 4. Effect of collision energy on the number of signature fragment peaks identifiable in yeast. Signature fragment chromatograms were extracted from SWATH datafiles using m/z 155.0172 ± 0.005 (ESI-). The ESI+ equivalent m/z 155.0161 ± 0.005 (ESI+) is included for comparison. For each collision energy, the number of signature fragment peaks observed for the underivatized sample (zero in all cases) and in a derivatized water blank were subtracted from the number of peaks observed for the derivatized sample.

The maximum number of MS experiments in one MS scanning cycle depends on the width of the chromatographic peaks and the speed of the QqTOF instrument. Considering peak properties of the UHPLC method as outlined below, a 5096

dx.doi.org/10.1021/ac500810r | Anal. Chem. 2014, 86, 5089−5100

Analytical Chemistry

Article

Figure 5. Representation of the workflow leading to the identification of indole-3-acetaldehyde in a 13C-labeled yeast sample using SWATH data. The software used was AB SCIEX PeakView 1.2. The sample (pink trace) and a blank (derivatized water) are overlaid. Panel A: Screening for signature fragments in ESI- by extraction of the signature fragment chromatographic trace from the high CE TIC. Panel B: Search for potential precursor ion masses in the low CE TIC at the retention time of the unknown analyte identified in panel A. Panel C: Sum formula generation with the PeakView Formula Finder on the basis of the accurate mass identified in panel B (m/z 336.1307). Subtraction of derivatization reagent substructure sum formula leads to the tentative analyte sum formula C10H9NO. Panel D: Results (structures and identifiers) of a KEGG LIGAND query (query date: 12.01.2014) using C10H9NO. The nonhighlighted structures were excluded due to the absence of carbonyl group (C06345 and C14789) or due to their reported absence in yeast (C10663). The identity of the highlighted structure (C00637, indole-3-acetaldehyde) was confirmed by comparison to the standard substance (cf. Supporting Information). 5097

dx.doi.org/10.1021/ac500810r | Anal. Chem. 2014, 86, 5089−5100

Analytical Chemistry

Article

were thus quantified via their precursor ions (Table 2). Notably, on QqQ instruments operating in SRM/MRM mode, the signature fragment may be used as quantifier, as specific precursor ions are selected with Q1. Averaged over all quantified analytes, LODs of 0.31 μM (ESI+ only), 0.36 μM (ESI- only), and 0.19 μM (ESI+ or ESI- wherever better) were determined (Table 2). For individual analytes, such as androstenedione, cortisone, or pregnenolone-sulfate, LODs were as low as 1.25 nM (Table 2). Few analytes performed poorly in one of the ionization polarities (e.g., betaine aldehyde has a permanent positive charge rendering its analysis in ESI- inefficient). RSDs for technical replicates (six subsequent injections inclusive yeast matrix at c = 3 × LOD) were in the range of 2−8%, e.g., for acetone: 4.3%/2.7%/2.6%/3.3%;for benzaldehyde 5.9%/6.5%/4.1%/5.1% (ESI+/ESI+ with IS correction/ESI-/ESI- with IS correction). Linearity was good in almost all cases (R2 averaged over all analytes was 0.993 for ESI+ and 0.994 for ESI-). For tiglic aldehyde and benzaldehyde, linear calibration curves could only be established in ESI+ while glyoxylic acid calibration was linear only in ESI-. The average MEs were 100 (±5) % and 97 (±12) % for ESI+ and ESI- (see Table 2 for individual values). The MEs determined comprise the effects of the matrix during derivatization, chromatography, and ionization/ detection. Taken together, the data on linearity and MEs thus show that the derivatization routine is largely unaffected by the yeast matrix and that MEs occur exclusively at the mass spectrometry stage. If, for some analytes, issues with linearity or MEs occurred, they were resolved by changing ionization polarity and are thus not due to derivatization yield. Various types of biological samples may be analyzed with our method. Liquid matrices, such as bodily fluids or beverages, require little sample preparation and are thus particularly wellsuited targets. In the Supporting Information, the applicability of our method to other matrices (urine and beverages), including sample chromatograms, is illustrated and an exemplary comparison of detection sensitivity to other dedicated, MS-based methods is given. Limitations. A key feature of the method design is that data on all precursor and fragment ions in the mass range of m/z 200 to 1000 are recorded, including nonderivatized analytes. Contrary to MRM methods, analytes that were unknown or untargeted at the time of analysis can hence be quantified months or years after sample acquisition by mining the SWATH datafiles. There are, however, limitations to the number of quantifiable analytes, which are mainly due to insufficient chromatographic separation. Analytes which (i) have the same exact mass and (ii) do not produce a unique fragment ion in CID require chromatographic separation for discrimination. Chromatographic limitations were observed in the case of carbohydrates in particular. While for the naturally occurring tetroses (the C4-sugars erythrulose, erythrose, and threose) a separation of derivatized erythrulose from derivatized erythrose and threose was possible, this was not the case for pentoses and hexoses (data not shown). Future work will thus be directed at methods for the improved separation of carbohydrate TSH-hydrazones. Commonly, the comprehensive separation of carbohydrates requires the increased separation power provided by gas chromatography in conjunction with other derivatization approaches.8 In this context, it may furthermore be mentioned that the LC method employed is optimized for the separation of tosylhydrazones. The detection of nonderivatized analytes, which might be of interest in the frame of the profiling of complete metabolic pathways, may thus require additional chromatographic approaches.

(Figure 5, Panel B). The most intense mass peak for injection (iii) (m/z 336.1307) is not seen for injection (ii) and was tentatively considered to represent the mass of the unknown metabolite’s [M − H]− ion. Using PeakView 1.2, a restricted search for suitable sum formulas was conducted (Figure 5, Panel C). The major restriction criteria were: (a) the sum formula must contain the subformula 12C7H7N2O2S originating from the derivatization reagent substructure, (b) the sum formula must not contain more than the seven 12C atoms introduced by the derivatization reagent (all other carbon atoms are assumed to be 13C, denoted by the custom symbol Q in PeakView), and (c) the calculated mass must not deviate more than 5 ppm from the measured accurate mass. On this basis, the sum formula search yielded a single hit (Q10C7H17N3O2S, m/z 336.1304, mass error: 0.8 ppm). If criteria (a) and (b) were ignored, the number of hits was 6. Subtraction of C7H8N2OS yielded the tentative analyte formula Q10H9NO. At this stage, we deduced from the features of the underlying reaction chemistry that, in the corresponding structure, the oxygen atom must be bound to a carbon atom by means of a double bond and the nitrogen atom is not bound to the same carbon atom. A search of the KEGG (Kyoto Encyclopedia of Genes and Genomes) LIGAND database27 yielded four hits with sum formula C10H9NO (Figure 5, Panel D). Two structures were excluded due to the lack of a carbonyl group. A third compound C10663, echinopsine, is a phytoalkaloid not expected in yeast and was thus also excluded. The remaining compound, C00637, indole-3-acetaldehyde, is a known microbial metabolite. Its identity was unambiguously confirmed by comparison to the standard substance (cf. Supporting Information). This example demonstrates the power of the SWATH-based identification approach in conjunction with the chemical information obtained from derivatization. In summary, it was possible to exclude all but one structure from the result set of a metabolite database query. In cases where multiple sum formulas are derived from the accurate mass of a [M − H]− precursor ion, accurate mass data from ESI+ and accurate masses of other fragments observed in the ESI+/ESI- high CE experiments may be used to arrive at the correct molecular structure. To determine the sensitivity of the identification approach, limits of identification were calculated for a range of target analytes (cf. Experimental Methods). The average limit of identification for the analytes of Table 2 was 0.54 μM and thus close to the average LOD of 0.36 μM, underscoring the capacity of our method to both quantify and identify metabolites. Quantification. A key feature of SWATH is the possibility of quantifying known compounds using the workflow applied for identification. We evaluated the corresponding analytical figures of merit for 61 target analytes compiled from different compound classes (Table 2). 15-point calibration curves (1:1 dilution), covering a dynamic range from 20 μM to 1.25 nM, were constructed, both in the presence and absence of a fully 13 C-labeled yeast matrix representative of a typical metabolomics sample. For those analytes occurring as 13C-labeled metabolites in the yeast extract, the application of IS correction was feasible if quantifier ions incorporating 13C were evaluated (Table 2). Using SWATH data, target analytes may be quantified via their precursor or fragment ions. However, since multiple compounds are cofragmented in SWATH, generally occurring fragments, such as the signature fragment (m/z 155.0172) used for identification, are unsuited as quantifiers unless complete chromatographic separation is achieved. In ESI-, most analytes 5098

dx.doi.org/10.1021/ac500810r | Anal. Chem. 2014, 86, 5089−5100

Analytical Chemistry

■

A feature inherent to the derivatization of carbonyls with phenylhydrazine derivatives is the formation of E/Z-isomers. For a range of analytes, distinct E/Z-isomer peaks were observable in the UHPLC chromatograms (Table 2 and Supporting Information 4). They were, however, baseline-separated in almost all cases (exception: erythrulose) and did not obstruct quantification. Aside from the mentioned exception, only one isomer peak was integrated for quantification (Table 2). Analyte Specific Remarks. Of the analytes listed in Table 2, mercaptopyruvic acid deserves special mention. Due to its thiolgroup, this analyte is prone to rapid oxidative dimerization, which effectively decreases its levels in solution. A preinjection reduction step, e.g., with dithiothreithol, can resolve this issue; it is, however, presently not included in the method.9,28 Furthermore, as the limited reaction of TSH with the solvent MeCN yields the same hydrazone as the reaction of TSH with acetaldehyde, significant background values for acetaldehyde were observed. While avoiding MeCN in solutions and eluents resolved this issue, we did not choose this option in view of the chromatographic benefits of MeCN. Despite the mentioned limitations, the proposed method is capable of discriminating all 61 compounds listed in Table 2, including a range of “difficult” isomer pairs, e.g., glyceraldehyde and dihydroxyacetone, ketoleucine and ketoisoleucine, acetone and propionaldehyde, or malondialdehyde and methylglyoxal.

■

CONCLUSIONS

■

ASSOCIATED CONTENT

Article

AUTHOR INFORMATION

Corresponding Author

*E-mail: r.p.h.bischoff@rug.nl. Funding

This work was partially funded by a Centre for Systems Biology Grant (SBC-EMA) of The Netherlands Organization for Scientific Research (NWO). David Siegel was partly funded by a travel grant of the Swiss National Science Foundation, Berne (IZK0Z3_150518). Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS David Siegel and Anne Meinema would like to acknowledge Silke Vedelaar (University of Groningen, Molecular System Biology Group) for supporting the preparation of yeast samples. Gérard Hopfgartner would like to thank Yves LeBlanc for fruitful input on the TripleTOF 5600 instrument.

■

REFERENCES

(1) Gerlich, M.; Neumann, S. J. Mass Spectrom. 2013, 48, 291−298. Gu, H.; Gowda, G. A. N.; Neto, F. C.; Opp, M. R.; Raftery, D. Anal. Chem. 2013, 85, 10771−10779. Nakabayashi, R.; Saito, K. Anal. Bioanal. Chem. 2013, 405, 5005−5011. Peironcely, J. E.; Rojas-Cherto, M.; Tas, A.; Vreeken, R.; Reijmers, T.; Coulier, L.; Hankemeier, T. Anal. Chem. 2013, 85, 3576−3583. (2) Eggink, M.; Wijtmans, M.; Ekkebus, R.; Lingeman, H.; de Esch, I. J. P.; Kool, J.; Niessen, W. M. A.; Irth, H. Anal. Chem. 2008, 80, 9042− 9051. (3) Iwasaki, Y.; Nakano, Y.; Mochizuki, K.; Nomoto, M.; Takahashi, Y.; Ito, R.; Saito, K.; Nakazawa, H. J. Chromatogr., B 2011, 879, 1159− 1165. Santa, T. Biomed. Chromatogr. 2011, 25, 1−10. (4) Nordstrom, A.; Want, E.; Northen, T.; Lehtioe, J.; Siuzdak, G. Anal. Chem. 2008, 80, 421−429. Werner, E.; Heilier, J.-F.; Ducruix, C.; Ezan, E.; Junot, C.; Tabet, J.-C. J. Chromatogr., B 2008, 871, 143−163. (5) Hanai, T. J. Chromatogr., A 2003, 989, 183−196. West, C.; Elfakir, C.; Lafosse, M. J. Chromatogr., A 2010, 1217, 3201−3216. (6) Cubbon, S.; Bradbury, T.; Wilson, J.; Thomas-Oates, J. Anal. Chem. 2007, 79, 8911−8918. Guo, Y.; Gaiki, S. J. Chromatogr., A 2011, 1218, 5920−5938. Idborg, H.; Zamani, L.; Edlund, P. O.; SchuppeKoistinen, I.; Jacobsson, S. P. J. Chromatogr., B 2005, 828, 9−13. Lu, W.; Bennett, B. D.; Rabinowitz, J. D. J. Chromatogr., B 2008, 871, 236−242. (7) Horvath, C.; Melander, W.; Molnar, I.; Molnar, P. Anal. Chem. 1977, 49, 2295−2305. (8) Koek, M. M.; Jellema, R. H.; van der Greef, J.; Tas, A. C.; Hankemeier, T. Metabolomics 2011, 7, 307−328. Lenz, E. M.; Wilson, I. D. J. Proteome Res. 2007, 6, 443−458. Mashego, M. R.; Rumbold, K.; De Mey, M.; Vandamme, E.; Soetaert, W.; Heijnen, J. J. Biotechnol. Lett. 2007, 29, 1−16. Pasikanti, K. K.; Ho, P. C.; Chan, E. C. Y. J. Chromatogr., B 2008, 871, 202−211. (9) Siegel, D.; Permentier, H.; Reijngoud, D.-J.; Bischoff, R. J. Chromatogr., B 2014, DOI: 10.1016/j.jchromb.2013.11.022. (10) Richardson, S. D.; Caughran, T. V.; Poiger, T.; Guo, Y. B.; Crumley, F. G. Ozone: Sci. Eng. 2000, 22, 653−675. (11) Zwiener, C.; Glauner, T.; Frimmel, F. H. Anal. Bioanal. Chem. 2002, 372, 615−621. Gao, S. M.; Zhang, Z. P.; Karnes, H. T. J. Chromatogr., B 2005, 825, 98−110. Siegel, D.; Merkel, S.; Koch, M.; Nehls, I. Food Chem. 2010, 120, 902−906. Lattova, E.; Perreault, H. Mass Spectrom. Rev. 2013, 32, 366−385. (12) Siegel, D.; Rasenko, T.; Koch, M.; Nehls, I. J. Chromatogr., A 2009, 1216, 4582−4588. (13) Zaikin, V. G.; Halket, J. M. Eur. J. Mass Spectrom. 2006, 12, 79− 115. (14) Vogel, M.; Büldt, A.; Karst, U. Fresenius’ J. Anal. Chem. 2000, 366, 781−791.

We present a novel strategy toward the global detection of analytes containing aldehyde and ketone groups in biological samples, which is based on the combination of derivatization and SWATH detection. TSH derivatization allows for efficient reversed phase chromatography of very small and/or polar analytes which usually require specialized chromatographic techniques. At the mass spectrometry stage, TSH-hydrazones ionize efficiently both in ESI+ and ESI-, providing sensitivities comparable to other derivatization-based, dedicated methods. Derivatization furthermore provides the option of chemoselective screening by introducing a CID pathway leading to an abundant signature fragment ion. By means of a QqTOF-SWATH, the quantitative and structural information generated by derivatization is comprehensively recorded in a single workflow at high mass resolution. The thus obtained SWATH datafiles, corresponding to a sample’s aldehyde/ketone fingerprint, allow, e.g., for the postacquisition quantification of analytes that were unknown or untargeted at the time of analysis. Our method is suited to seamlessly complement pre-existing LC-MS methods as injections may be performed directly out of LC vials containing aqueous or aqueous/organic solutions. In view of its analytical benefits and relative simplicity, the presented approach is a valuable asset in the comprehensive study of aldehyde and ketone analytes. By enhancing MS screening with a layer of chemical information, a clear view on these compound classes is obtained and corresponding metabolomics screening experiments are largely facilitated. Looking ahead, our concept may also be of value in the study of protein carbonylation due to disease or oxidative stress, for instance.29

S Supporting Information *

Autosampler script, two tables, and two figures as described in the text. This material is available free of charge via the Internet at http://pubs.acs.org. 5099

dx.doi.org/10.1021/ac500810r | Anal. Chem. 2014, 86, 5089−5100

Analytical Chemistry

Article

(15) Asam, S.; Lichtenegger, M.; Muzik, K.; Liu, Y.; Frank, O.; Hofmann, T.; Rychlik, M. J. Chromatogr., A 2013, 1289, 27−36. (16) Gillet, L. C.; Navarro, P.; Tate, S.; Roest, H.; Selevsek, N.; Reiter, L.; Bonner, R.; Aebersold, R. Mol. Cell Proteomics 2012, 11; DOI: 10.1074/mcp.O111.016717. (17) Liu, Y.; Huettenhain, R.; Surinova, S.; Gillet, L. C. J.; Mouritsen, J.; Brunner, R.; Navarro, P.; Aebersold, R. Proteomics 2013, 13, 1247− 1256. (18) Hopfgartner, G.; Tonoli, D.; Varesio, E. Anal. Bioanal. Chem. 2012, 402, 2587−2596. Tonoli, D.; Varesio, E.; Hopfgartner, G. Chimia 2012, 66, 218−222. (19) Zhu, X.; Chen, Y.; Subramanian, R. Anal. Chem. 2013, 86, 1202−1209. (20) Buescher, J. M.; Moco, S.; Sauer, U.; Zamboni, N. Anal. Chem. 2010, 82, 4403−4412. (21) Canelas, A. B.; Harrison, N.; Fazio, A.; Zhang, J.; Pitkanen, J.-P.; van den Brink, J.; Bakker, B. M.; Bogner, L.; Bouwman, J.; Castrillo, J. I.; Cankorur, A.; Chumnanpuen, P.; Daran-Lapujade, P.; Dikicioglu, D.; van Eunen, K.; Ewald, J. C.; Heijnen, J. J.; Kirdar, B.; Mattila, I.; Mensonides, F. I. C.; Niebel, A.; Penttila, M.; Pronk, J. T.; Reuss, M.; Salusjarvi, L.; Sauer, U.; Sherman, D.; Siemann-Herzberg, M.; Westerhoff, H.; de Winde, J.; Petranovic, D.; Oliver, S. G.; Workman, C. T.; Zamboni, N.; Nielsen, J. Nat. Commun. 2010, 1; DOI: 10.1038/ncomms1150. Mortimer, R. K.; Johnston, J. R. Genetics 1986, 113, 35−43. (22) Lee, D. W.; Jensen, C. M. J. Chem. Educ. 2000, 77, 629−629. (23) Siegel, D.; Merkel, S.; Koch, M.; Emmerling, F.; Nehls, I. Acta Crystallogr. 2009, E65, o988−989. (24) Jeong, J.; Yang, J.; Ha, C.-S.; Cho, W.-J.; Chung, I. Polym. Bull. 2012, 68, 1227−1238. (25) Katsuki, H.; Tanegashima, C.; Tokushige, M.; Tanaka, S. Bull. Chem. Soc. Jpn. 1972, 45, 813−817. (26) Johnson, D. W. Rapid Commun. Mass Spectrom. 2007, 21, 2926− 2932. (27) Kanehisa, M.; Goto, S.; Sato, Y.; Furumichi, M.; Tanabe, M. Nucleic Acids Res. 2012, 40, D109−D114. (28) Siegel, D. Analyst 2012, 137, 5457−5482. (29) Dalle-Donne, I.; Rossi, R.; Giustarini, D.; Milzani, A.; Colombo, R. Clin. Chim. Acta 2003, 329, 23−38. Grimsrud, P. A.; Xie, H.; Griffin, T. J.; Bernlohr, D. A. J. Biol. Chem. 2008, 283, 21837−21841. Levine, R. L. Free Radical Biol. Med. 2002, 32, 790−796. Madian, A. G.; Regnier, F. E. J. Proteome Res. 2010, 9, 3766−3780.

5100

dx.doi.org/10.1021/ac500810r | Anal. Chem. 2014, 86, 5089−5100