Research article Received: 12 February 2014

Revised: 8 May 2014

Accepted: 12 May 2014

Published online in Wiley Online Library

(wileyonlinelibrary.com) DOI 10.1002/jms.3397

Determination of the enrichment of isotopically labelled molecules by mass spectrometry Ana González-Antuña, Pablo Rodríguez-González and J. Ignacio García Alonso* A general method for the determination of the enrichment of isotopically labelled molecules by mass spectrometry (MS) is described. In contrast to other published procedures, the method described here takes into account and corrects for measurement errors such as the contribution at M  1 due to loss of hydrogen or lack of spectral resolution and provides an uncertainty value for the determined enrichment. The general procedure requires the following steps: (1) evaluation of linearity in the mass spectrometer by injecting the natural abundance compound at different concentration levels, (2) determination of the purity of the mass cluster using the natural abundance analogue, (3) calculation of the theoretical isotope composition of the labelled compound using different tentative isotope enrichments, (4) calculation of ‘convoluted’ isotope distributions for the labelled compound taking into account the purity of the mass cluster determined with the natural abundance analogue and (5) comparison of the isotope distributions measured for the labelled compound with those calculated for different isotope enrichments using linear regression. The method was applied to a series of commercially available 13C- and 2H-labelled compounds and to a suite of singly 13C-labelled β2-agonist prepared in-house both by gas chromatography (GC)–MS, GC–tandem MS (MS/MS) and liquid chromatography–MS/MS with satisfactory results. It was observed that the main uncertainty source for the isotope enrichment was the uncertainty in the purity of the measured cluster as determined with the natural abundance compound. Copyright © 2014 John Wiley & Sons, Ltd. Additional supporting information may be found in the online version of this article at the publisher’s web site. Keywords: isotope enrichment; labelled compounds; GC–MS; LC–MS; isotope distribution

Introduction

J. Mass Spectrom. 2014, 49, 681–691

* Correspondence to: J. Ignacio García Alonso, Department of Physical and Analytical Chemistry, University of Oviedo, Julián Clavería 8, 33006, Oviedo, Spain. E-mail: [email protected] Department of Physical and Analytical Chemistry, University of Oviedo, Julián Clavería 8, 33006, Oviedo, Spain

Copyright © 2014 John Wiley & Sons, Ltd.

681

Molecules labelled with enriched isotopes are employed in scientific studies for many purposes: in biological studies to unravel metabolic pathways and the size of metabolic pools,[1] in chemical metrology to perform isotope dilution calculations,[2] in environmental studies to follow the transformation and diffusion of pollutants[3] and, since many years, in organic synthesis to understand reaction mechanisms[4] to cite only a few relevant areas. The knowledge of the isotope composition of the labelled molecule employed and, particularly, the isotope enrichment of the labelling atoms (13C, 2H, 15N, 18O, etc) is required for many calculation procedures employed in these studies. For example, in isotope dilution mass spectrometry (IDMS), the isotope composition of the labelled molecule is employed in the isotope pattern deconvolution mode in combination with singly labelled compounds.[2] However, it is also well known that, when the labelled molecules are employed as internal standards for IDMS, the accurate knowledge of their isotope composition is not required.[2] That is why, traditionally, manufacturers of labelled molecules provide a nominal enrichment for the purchased molecules and, in most cases, without explanation on how this enrichment was calculated. Furthermore, the nominal enrichment for commercially available labelled molecules has no uncertainty value associated to it. In light of modern trends in chemical metrology and the recent advances in MS instrumentation this situation can no longer be considered acceptable. So far, there are very few procedures described in the literature for the measurement of isotope enrichments and no compounds

certified for their absolute isotope composition which could be employed as reference materials. Additionally, very few studies were published about the suitability of different mass spectrometers for isotope distribution measurements (spectral accuracy). For example, the accurate measurement of the isotope composition of labelled compounds by MS depends on the linearity of the mass spectrometer. The non-linear signal response of Orbitrap mass spectrometers was recently reported for the determination of arsenobetaine by IDMS.[5] The reasons for the non-linearity in the instrumental response were not studied, but potential sources of errors included: (1) space charge effects, (2) overfilling of the ion trap, (3) collision of ions with background gas and (4) Fourier transform data processing. Similar results were described by other authors when using the Orbitrap for molecular formula assignment,[6] metabolomic database building[7] or stable isotope probing.[8] Erve et al.[6] determined the isotope distribution observed for ten commercially available natural products with masses between 639 and 1664 u by continuous direct infusion of standard solutions both in positive and negative detection modes. They observed that ‘spectral accuracy’, the relative error in the (M + 1)/(M) ratio, degraded when increasing resolution from 7500 to 100 000 at m/z 400 and that less abundant isotopomers were

A. González-Antuña, P. Rodríguez-González and J. I. García Alonso

682

under represented in comparison with their theoretical abundance. In a similar study Xu et al.[7] analysed 137 solutions of commercial compounds from 75 to 810 u by flow injection analysis (FIA). They observed that the errors on the abundance ratios for (M + 1)/(M) at the top of the FIA peak were a function of the total counts measured. When the counts for M were below ca. 106–107 counts, relative errors in the (M + 1)/(M) ratio higher than -20% were obtained. For more classical analysers, such as the quadrupole and the time of flight, the reasons for spectral inaccuracy have been better studied. First and most important, the mass clusters obtained for different molecules in the ion source may not be pure. Different fragmentation mechanisms may occur simultaneously in the ion source, and fragment ions containing different numbers of hydrogen atoms can be obtained which overlap in the mass spectrum, particularly for electron ionisation (EI).[9,10] Second, the resolution of the mass spectrometer employed may not be adequate, and tailing of a given mass peak M at masses M  1 and M + 1 could be observed even in a properly tuned and calibrated instrument. Quadrupoles may show a nonsymmetrical tailing at adjacent masses.[11] Another effect which could be observed is the change in the measured isotope distributions with the amount of compound injected. This observed change in the isotope distribution could be due to different causes either at the ion source, the analyser or the detector. For example, Fagerquist et al.[12] observed different isotope distributions for fatty acid methyl esters at different concentration levels, and this effect was ascribed to proton transfer in the gas phase between a fragment ion and a neutral molecule which disturbed the isotope distribution of the molecular ion at high concentration levels. On the other hand, the effect of the selected mass analyser and detector on the measured isotope composition of molecules has not been discussed in depth in the literature. For example, Böcker et al.[13] observed that the experimental abundances for the M + 1 ion for 86 compounds in an orthogonal accelerator TOF instrument were always biased low with respect to the M ion. The relative errors were usually lower than 20% but significant when accurate isotope distributions needed to be measured. This behaviour was ascribed to two possible effects: (1) a non-linearity effect in the detector at low counts which produced proportionally less counts for low abundance masses in comparison with those with high abundance or (2) a high detector threshold value which reduced the area counts for low abundance peaks proportionally more than for high abundance peaks. Concentration effects on isotope distributions by gas chromatography (GC)–MS were discussed also by Dauner and Sauer[14] using a MD800 quadrupole analyser. These authors observed that, when working at very high concentrations, the calculation of the isotope distribution during the whole chromatographic peak profile showed a dip for the monoisotopic mass M at the maximum of the chromatographic peak compatible with a non-linearity or saturation effects of the MS detector. Similar non-linearity results were obtained by Antoniewicz et al.[15] for the study of amino acid isotope enrichments. These authors employed an old HP 5971 quadrupole mass spectrometer and indicated that the newer Agilent 5975B did not show any concentration effect. So, there is a clear need to develop methods to evaluate spectral accuracy of mass spectrometers before they can be applied to the study of isotopically labelled compounds. Calculation methodologies usually compare the experimental isotope distributions obtained by MS with theoretically derived spectra using correlation analysis.[16] The ‘right’ isotope enrichment

wileyonlinelibrary.com/journal/jms

is obtained from the best fit (maximum correlation coefficient or minimum sum of squared residuals) between the calculated and experimental abundances. A similar procedure was described by Gonzalez-Gago et al.[11] for PBDEs labelled with 81Br and by Castillo-Tirado et al.[17] for diclofenac-d4. In this paper, we expand on our previous publications[11,17] and present a general procedure which can be applied both to the evaluation of mass spectrometers for accurate isotope distribution measurements (spectral accuracy) and to determine the isotope enrichment of labelled molecules with its associated uncertainty values. We have applied this procedure to a number of labelled compounds obtained commercially and to a suite of β2-agonist singly labelled with 13C. Additionally, we compare different mass spectrometers under different experimental conditions and discuss their suitability for accurate isotope distribution measurements.

Experimental Instrumentation GC–MS measurements in the selective ion monitoring mode (SIM) were performed with a gas chromatograph model 6890N (Agilent Technologies, Wilmington, DE, USA) coupled to a mass spectrometric detector model 5975B (Agilent Technologies) and with a gas chromatograph model 7890A (Agilent Technologies) coupled to a triple quadrupole mass spectrometric detector model 7000 (Agilent Technologies). Both instruments used an HP-5MS capillary column (cross-linked 5% phenyl-methyl siloxane, 30 m × 0.25 mm i.d., 0.25 μm coating), and the EI was performed at 70 eV. The column temperature was initially held at 50 °C for 1 min, and then a temperature ramp of 15 °C/min was applied until 300 °C. This temperature was sustained for 5 min. Helium was used as carrier gas at flow rate of 2 ml/min. The temperature of the injector, transfer line, ion source and quadrupole was set at 270, 280, 230 and 150 °C, respectively. A sample volume of 2 μl was injected automatically in pulsed splitless mode (1-min purge time and 60 psi of pressure pulse). Qualitative analysis was performed from m/z 40 to 600 in the full-scan mode. The measurements of isotope abundances were performed in the SIM mode using the following ion molecular fragments: M-43 clenhexyl, M-167 clencyclohexerol, M-30 clenpenterol and M-15 (loss of a methyl group) for the rest of β2-agonists. Two ion molecular fragments were selected for octylphenol (M-71 and M-113) and for n-nonylphenol (M-42 or M-155). All measurements were carried out using a 10-ms dwell time per mass. Five consecutive masses were usually employed in the SIM mode. Daily optimisation of the instrumental conditions was performed using the ‘autotune’ option of the software supplied with the instrument. For this purpose perfluorotributylamine (PFTBA) was used as the tuning compound. The same compound was employed also to check the contributions at M  1 and M + 1 from tailing of the peak at mass M. A centrifuge Centromix (JP Selecta, Barcelona, Spain) was used for the preparation of the derivatisation reagent employed in the analyses of β2-agonists. The triple quadrupole liquid chromatography–tandem MS (LC–MS/MS) system employed consisted in a high-performance liquid chromatography (HPLC) system model 1290 (Agilent Technologies) with a reverse-phase zorbax-C18 analytical column (2.1 mm × 50 mm and 1.8-μm particle size, Agilent Technologies). Mobile phases A and B were water and acetonitrile both with 0.1% formic acid. The chromatographic method for the β2-agonists consists of an initial mobile phase composition (2% B) constant for

Copyright © 2014 John Wiley & Sons, Ltd.

J. Mass Spectrom. 2014, 49, 681–691

Enrichment of isotopically labelled molecules 2 min, a linear gradient to 25% B in 9 min and a final gradient until 90% B in 10 min. These final conditions were kept for 2 min. The chromatographic method employed for alkylphenols was 20% B for 1 min and then up to 100% B in 11 min keeping these conditions for 2 min. In all cases, the flow rate used was 0.3 ml min1. The HPLC system was connected to a triple quadrupole mass spectrometer Agilent 6460 equipped with an electrospray interface operating in positive and negative ion modes for the β2-agonist and alkylphenols, respectively, using the following operation parameters: capillary voltage, 3500 V; nebuliser pressure, 45 psi; drying gas flow rate, 11.0 L min1; and gas temperature, 400 °C. The measurements of isotope abundances were performed with five masses using SIM acquisition mode with the molecular ion MH+ or M  H for agonists and alkylphenols, respectively. For clenproperol, the molecular ion showed a very low intensity due to the loss of H2O. For this compound the (M  18)H+ ion was employed in the isotope enrichment calculations. An analytical balance model AB204-S (Mettler Toledo, Zurich, Switzerland) was used for the gravimetric preparation of all solutions.

enrichment) was obtained from Cambridge Isotope Laboratories Inc. (Andover, MA, USA) in pure solid form. Stock solutions of each analyte, either natural or labelled, were prepared in methanol and stored at 4 °C. These solutions were employed to prepare daily gravimetrically diluted working standard solutions in methanol. All solvents methanol, toluene, hexane and acetonitrile were purchased from Fisher Scientific (Madrid, Spain). The organic solvents used were HPLC grade or LC–MS/MS grade. Ultra-pure water was obtained from a Milli-Q system (Millipore Co., Bedford, USA). Formic acid, sodium chloride, diethylamine, chloro (chloromethyl)dimethylsilane and acetic acid anhydride were also supplied by Sigma-Aldrich. Procedures Derivatisation reactions for gas chromatography

Alkyl-phenols were derivatised with acetic acid anhydride by the procedure described previously,[19] while β2-agonists were derivatised with chloro(chloromethyl)dimethylsilane following the procedure described elsewhere.[20] The molecular formulas of the derivatised β2-agonists are included in Fig. 1.

Reagents and materials Natural abundance clenproperol, clenbuterol, salbutamol, brombuterol, terbutaline, ractopamine, tert-octylphenol and n-nonylphenol were purchased from Sigma-Aldrich (St. Louis, MO, USA). Clenpenterol and clencyclohexerol were purchased from WITEGA Laboratorien (Berlin-Adlershof GmbH, Berlin). Natural abundance clenhexyl was obtained from ISC-Science (Oviedo, Spain). The structure of all β2-agonists tested is shown in Fig. 1. All 13C1-labelled β2-agonists except clenbuterol were provided in pure solid form by ISC-Science. Clenbuterol 13C1 was synthesised in our laboratory as described elsewhere.[18] All compounds have the carbon-13 label in the benzylic position. Deuterated clenbuterol (2H9) was obtained from Sigma-Aldrich (St. Louis, MO, USA) with a nominal enrichment of >97.0%. Labelled alkylphenols such as 4-tert-octylphenol (ring 13 C6, 93.5% nominal enrichment) and 2H2-tert-octylphenol (97.7% nominal enrichment) were purchased in acetone solution from Sigma-Aldrich, while 4-n-nonylphenol (ring 13C6, 99% nominal

Measurement of experimental isotope distributions

All experimental isotope distributions, Aiexp, were calculated as the ratio of each peak area obtained for each mass divided by the sum of all peak areas measured for the given compound either after LC or GC separation in the SIM mode. For each mass cluster, the monoisotopic mass M was measured together with M  1 (to check for loss of hydrogen and/or tailing in the mass spectrometer) and M + 1, M + 2, M + 3, etc. Isotope distributions were expressed as fractional abundances where ∑ Ai = 1. Replicate injections (n ≥ 5) were performed to calculate the experimental uncertainties of the isotope distributions. In some particular situations (e.g. alkylphenols measured by GC–MS) also the M  2 mass was measured as a significant contribution was detected at some clusters. The initial evaluation of ‘spectral accuracy’ was performed based on the peak area ratios for M + 1 and M + 2 versus M, respectively.

683

Figure 1. Structure of the β2-agonists studied.

J. Mass Spectrom. 2014, 49, 681–691

Copyright © 2014 John Wiley & Sons, Ltd.

wileyonlinelibrary.com/journal/jms

A. González-Antuña, P. Rodríguez-González and J. I. García Alonso 2

Calculation of theoretical isotope distributions

We have employed the method of Kubinyi[21] based on a matrix multiplication algorithm[2] for the polynomial expansion of the binomial probability equations. This method was implemented as a Visual Basic macro for Excel[22,23] and can calculate the isotope distribution for natural abundance molecules containing any combination of H, B, C, N, O, F, Si, P, S, Cl, Br, Se and I atoms. The Excel spreadsheet for low resolution isotope distribution calculations (IDC.xls) is included as a supplementary material with this manuscript. This spreadsheet calculates also the uncertainties of the isotope distributions because of natural variability. These uncertainty values are not employed in this manuscript because of their very small effect on the results obtained. Calculation of cluster purity

The contribution of the loss of hydrogen (M  H) or the tailing of the peaks at the low mass side (M  1) in the experimental spectrum can be calculated by multiple linear regression. Both contributions are calculated simultaneously as the isotopic composition of the M  H ion is almost the same as that of the M ion but shifted one nominal mass unit.[10,11,19] The differences between the distribution of M  1 or M  H were so small for all compounds that we decided to employ the distribution of M  1 in all calculations. In brief, the experimental isotope distribution measured for the natural abundance compound, Aiexp, at n given masses, including mass M  1, was compared with the theoretically calculated isotope distributions for the given cluster (Aitheo) using the matrix equation: 2

3

2 0 6 M 7 6 M 6 Aexp 7 6 A theo 7 6 6 Mþ1 7 6 6 Aexp 7 6 AMþ1 6 theo 7 6 6 … 7¼6 7 6 … 6 6 Mþn2 7 6 Mþn2 7 6 6A 5 4 Atheo 4 exp AMþn1 AMþn1 theo exp AM1 exp

AM theo

3

2

e1

3

7 6 2 7 6 e 7 AMþ1 theo 7 7 " 7 # 6 Mþ2 7 6 e3 7 x Atheo 7 M 7 6 þ6 7 7 (1) 6 … 7 x M1ðHÞ … 7 7 7 6 7 6 n1 7 AMþn1 5 4e 5 theo AMþn en theo

The least squares solution to Eqn 1, the values of xM and xM  1(H), are the fractions of the experimental spectrum corresponding to the cluster of mass M and mass M  1 (and/or M  H), respectively. For some compounds, the calculation of cluster purity was performed at different concentration levels to rule out non-linearity effects. Additionally, the residuals of the multiple linear regression, ei, were studied for normality. An example spreadsheet for the calculation of the cluster purity of brombuterol by LC–MS/MS (Brombuterol LC.xls) is given as supplementary information. Evaluation of spectral accuracy

684

We employed two methods for the evaluation of the spectral accuracy of the different mass spectrometers employed. In all cases, the natural abundance compounds were employed as reference standards. In the first method, the ratio of peak areas for (M + 1)/M and (M + 2)/M was employed, and the results were compared with the expected ratios based on the natural isotope distributions. In the second method, we computed theoretical ‘convoluted’ distributions taking into account the observed cluster purity and tailing at the low mass side of the spectrum. For that purpose, we employed the equation:

wileyonlinelibrary.com/journal/jms

AM1 conv

3

2

3

0

7 6 AM 6 AM 6 conv 7 6 theo 7 6 6 6 … 7 ¼ x M 6 AMþ1 7 6 6 theo 6 Mþn2 7 6 4 Aconv 5 4 … AMþn1 conv

Mþn1 Atheo

2

AM theo

3

7 6 AMþ1 7 7 6 theo 7 7 7 6 7 þ x M1ðHÞ 6 … 7 7 7 6 7 6 Mþn1 7 5 4 Atheo 5

(2)

AMþn theo

where Aiconv is the convoluted isotope distribution for the natural abundance compound, including mass M  1 and n consecutive masses, taking into account the observed purity of the mass cluster, the fragmentation factors xM and xM  1(H) and the theoretically calculated isotope distributions for the cluster of mass M (Aitheo). Then, the ratio of peak areas for (M + 1)/M and (M + 2)/M was compared with the expected ratios based on the convoluted distributions observed for each compound. In this way, we can correct for cluster purity and lack of spectral resolution and evaluate other possible sources of spectral inaccuracy such as mass bias. Calculation of isotope enrichments

A Visual Basic macro for Excel was prepared to calculate 100 different isotope distributions for a labelled molecule between two, user selectable, low and high isotope enrichment values for the labelling isotope. After calculation, the purity of the mass cluster, measured with the natural abundance compound, was taken into account to compute 100 theoretical ‘convoluted’ distributions at the different enrichment values selected using Eqn 2. In this case, Aiconv are the convoluted isotope distributions for a labelled compound, while Aitheo are the theoretically calculated isotope distributions for the cluster of mass M in the labelled compound at a given isotope enrichment. Finally, the experimentally measured isotope distributions for the labelled compound were compared with the 100 convoluted distributions by linear regression analysis.[11,17,19,24] Then, the ‘right’ isotope enrichment was selected as to that giving the minimum in the square sum of residuals (SSR) between the calculated and observed distributions. The Excel spreadsheet employed for the calculation of the isotope enrichments for isotopically labelled compounds (IEC.xls) is included also as a supplementary material with this manuscript.

Results and discussion Suitability of different mass spectrometers for isotope distribution measurements In this paper, we compare three different instruments: a quadrupole GC–MS, a triple quadrupole GC–MS/MS and a triple quadrupole LC–MS/MS. All measurements were performed in SIM mode, and five to ten consecutive masses, introduced with one decimal figure, were measured depending on the compound. For all compounds the LC–MS/MS instrument was run both under standard resolution (0.7 u full width at half maximum, FWHM) and under higher resolution (0.5 u FWHM) to compare the results obtained. Unfortunately, not all compounds could be measured in all instruments and instrumental configurations, but, at least, every compound was measured in two instruments with a different chromatographic separation (GC or LC), ion source (EI or ESI) and mass spectrometer configuration. For gas chromatography, a chemical derivatisation was performed, while LC allowed the direct injection of the compounds.

Copyright © 2014 John Wiley & Sons, Ltd.

J. Mass Spectrom. 2014, 49, 681–691

Enrichment of isotopically labelled molecules For all instruments, it was observed that the quality of the results depended strongly on the magnitude of the peak areas measured. At low concentration levels or for mass fragments giving low intensities, the results showed high uncertainties. It was decided to work in the higher part of the linear range of the instrument and perform always a mass calibration before any measurement. Spectral accuracy of the different quadrupole mass spectrometers The capability of mass spectrometers to measure accurate isotope ratios in organic molecules is termed ‘spectral accuracy’ in the modern literature.[6,7] The typical test is to measure the ratio M + 1/M for a series of compounds of natural abundance and compare these results with theoretically calculated ratios.[6,7,13] In our case, most of the compounds evaluated contained Cl or Br so the M + 2/M ratio could also be employed to evaluate spectral accuracy with reasonable counting statistics. So, all natural abundance β2-agonists were injected in the three instruments in quintuplicate and the ratio of peak areas for masses M + 2 and M + 1 with respect to M calculated. The data obtained with the three instruments are summarised in Fig. 2 as a function of the monoisotopic mass M considered for each compound. We have included data for M + 1/M (clear symbols) and M + 2/M (grey symbols). As it can be observed, most of the results produce errors lower than ±10% for M + 1/M and lower than ±5% for M + 2/M using the three instruments, and these errors are independent of the monoisotopic mass considered. As can be observed, better results were obtained for the M + 2/M ratios in comparison with the M + 1/M ratios. This could be due to the fact that most of the compounds tested contained chlorine or bromine for which the M + 2 peak was higher than the M + 1 peak. Additionally, large relative errors were obtained for the M + 1/M ratios for some of the compounds both by GC–MS and LC–MS/MS. There could be three different factors which explain these poor results. First, the mass cluster measured may not be pure (e.g. coexistence of M+ and (M  H)+ in the mass spectrum); second, the resolution of the mass spectrometer may not be adequate (e.g. tailing of the peaks at the low mass side of the spectrum) and, third, the transmission in the mass spectrometer

Relative error (M+1)/M and (M+2)/M (%)

25

20

could be a function of the mass of the ion considered (mass bias) as it is usual for inorganic mass spectrometers.[2] So, we decided to employ a more powerful calculation procedure to study the sources of error in spectral accuracy. This alternative calculation procedure allows us to compute the contribution of the tailing at the low mass side of the spectrum or the formation of M  H+ ions in the EI source. This procedure is based on multiple linear regression as described in Eqn 1. The whole isotope distribution measured for each compound, Aiexp, at n given masses, including mass M  1, was compared with the theoretically calculated isotope distributions for the given cluster (Aitheo) as shown in Eqn 1. In this way, the contribution at M  1 (and/or M  H) could be calculated for each compound, and instrumental configuration was employed. Once the contribution of cluster purity and mass resolution are corrected the contribution of mass bias to spectral accuracy can be calculated. The results of this alternative procedure are given below for the different instruments. GC–MS instrument All compounds, except ractopamine which could not be derivatised under the conditions employed,[20] were measured in this instrument. The fragmentation factors for M  1(H), and M  2H when required, obtained for all compounds tested are shown in Table 1. For each compound, we have included information on the number of consecutive masses measured in SIM mode and the nominal mass of the monoisotopic mass M at the cluster measured. For the alkylphenols, two main clusters were measured for comparison of their fragmentation fractions. The measured isotope distribution of all β2-agonists was consistent with a mixture of M and M  H (or M  1 due to tailing at the low mass side of the spectrum) in the measured clusters. To check the tailing at the low mass side of the spectrum, PFTBA, which does not contain any hydrogen, was employed. Under the working conditions used, the M  1 contribution due to tailing was between 0.3 and 0.7% of the M peak depending on the mass range selected.[11] For the alkylphenols, the presence of an M  2H cluster was detected and had to be included in the multiple linear regression calculations. The combined standard uncertainties for the fragmentation fractions, u(x), were calculated taking into account the standard deviation from the individual fragmentation fractions obtained for each injection, sx, and the standard uncertainties of each fragmentation factor obtained from the multiple linear regression calculations, si. The equation employed to calculate the combined standard uncertainties was:[25] sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ∑ðsi Þ2 uðx Þ ¼ ðsx Þ2 þ n

15

10

5

0

-5 200

250

300

350

400

450

500

550

600

Monoisotopic mass

J. Mass Spectrom. 2014, 49, 681–691

where n is the number of injections employed in the calculation (from 5 to 9 depending on the compound). As it can be observed, fragmentation factors for M  1(H) between 0.57 and 4.35% were obtained depending on the compound and the selected cluster. The correlation coefficients (r2) were, in all cases, better than 0.999 indicating good correlation between the experimental and theoretical abundances using Eqn 1. In some cases, the SSR was relatively high (of the order of 104), and that resulted in higher uncertainties for the fragmentation factors. This was the case particularly for tertoctylphenol at the 177 cluster, n-nonylphenol at the 107 cluster and clenhexyl at the 329 cluster. In general, the GC–MS provided

Copyright © 2014 John Wiley & Sons, Ltd.

wileyonlinelibrary.com/journal/jms

685

Figure 2. Spectral accuracy for the β2-agonists studied expressed as relative error (%) in the M + 1/M and M + 2/M ratios. Clear symbols: M + 1/M ratio; grey symbols: M + 2/M ratio. (○) GC–MS; (Δ) GC–MS/MS; (□) LC–MS/MS.

(3)

A. González-Antuña, P. Rodríguez-González and J. I. García Alonso Table 1. Fragmentation factors xM  1 and xM  2 (in %) obtained by GC–MS, GC–MS/MS operating in the SIM mode, LC–MS/MS operating in the SIM mode at standard resolution (0.7 FWHM) and LC–MS/MS operating in the SIM mode at a higher resolution (0.5 FWHM) for the different compounds tested. Uncertainties are expressed as combined standard uncertainties Instrument

Compound (number of masses measured)

GC–MS

tert-Octylphenol (6) tert-Octylphenol (5) n-Nonylphenol (5) n-Nonylphenol (5) Clenbuterol (10) Brombuterol (8) Clenproperol (9) Clenpenterol (8) Clenhexyl (10) Clencyclohexerol (9) Salbutamol (9) Terbutaline (10) Clenbuterol (10) Brombuterol (8) Clenproperol (8) Clenpenterol (8) Clenhexyl (9) Clencyclohexerol (9) Salbutamol (9) Terbutaline (10) tert-Octylphenol (4) n-Nonylphenol (4) Clenbuterol (6) Brombuterol (7) Clenproperol (6) Clenproperol (6) Clenpenterol (6) Clenhexyl (6) Clencyclohexerol (6) Salbutamol (5) Terbutaline (5) Ractopamine (5) tert-Octylphenol (4) n-Nonylphenol (4) Clenbuterol (6) Brombuterol (7) Clenproperol (6) Clenproperol (6) Clenpenterol (6) Clenhexyl (6) Clencyclohexerol (6) Salbutamol (5) Terbutaline (5) Ractopamine (5)

GC–MS/MS (SIM mode)

LC–MS/MS (SIM mode) 0.7 FWHM

LC–MS/MS (SIM mode) 0.5 FWHM

results well in agreement with the theoretical distributions when the contribution of M  H (and/or M  1) was taken into account. GC–MS/MS instrument

686

Only the β2-agonists, except ractopamine for the reasons given above, were measured in this instrument. The results obtained for the fragmentation factors are also given in Table 1 including their combined uncertainties. In all cases, the SSR in Eqn 1 was in the 105–106 range which indicated a very good fit with

wileyonlinelibrary.com/journal/jms

Nominal mass of the cluster

xM  1 (%)

xM  2 (%)

135 177 107 220 331 419 317 331 329 329 506 492 331 419 317 331 329 329 506 492 205 219 277 365 263 + 245 (M  18)H 291 303 319 240 226 302 205 219 277 365 263 + 245 (M  18)H 291 303 319 240 226 302

4.35 ± 0.18 0.75 ± 0.90 0.78 ± 0.96 0.63 ± 0.18 1.54 ± 0.45 1.82 ± 0.16 1.32 ± 0.75 0.85 ± 0.26 3.3 ± 1.2 0.57 ± 0.38 0.79 ± 0.43 2.60 ± 0.51 0.29 ± 0.48 0.08 ± 0.34 0.22 ± 0.48 0.48 ± 0.25 0.95 ± 0.44 0.14 ± 0.20 0.01 ± 0.50 0.34 ± 0.14 0.16 ± 0.06 0.28 ± 0.10 0.56 ± 0.55 0.53 ± 0.39 0.85 ± 2.0 0.67 ± 0.57 0.52 ± 0.54 0.59 ± 0.52 0.45 ± 0.50 0.26 ± 0.08 0.38 ± 0.14 0.52 ± 0.34 0.18 ± 0.23 0.20 ± 0.42 0.10 ± 1.4 0.25 ± 0.66 0.19 ± 0.93 0.46 ± 0.49 0.13 ± 0.43 0.29 ± 0.79 0.16 ± 0.54 0.23 ± 0.18 0.90 ± 0.11 0.26 ± 0.65

1.00 ± 0.17 0.93 ± 0.89 0.53 ± 0.95 0.33 ± 0.18 – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – –

the theoretical abundances. The fragmentation factors obtained for M  1(H) were lower than those obtained with the single quad instrument. We employed PFTBA to check the tailing at the low mass side of the spectrum under the experimental conditions employed, and the results are shown in Table 2. As can be observed, the contribution at M  1 due to tailing was negligible (97.0% for each compound, respectively, without any uncertainty value. Figure 4 shows the plot of the SSR obtained

wileyonlinelibrary.com/journal/jms

0.01

Square sum of residuals

ratios are plotted against the theoretically convoluted ratios in Fig. 3A. Grey symbols indicate M + 2/M ratios, while clear symbols refer to M + 1/M ratios for all β2-agonists studied. As can be observed, there is a good agreement between the experimental and theoretical ratios for a wide range of ratio values (from 0.014 to 2.00) using all three instruments. Figure 3B shows the same results as in Fig. 2 but after correction for cluster purity and tailing. As can be observed, the relative errors for the (M + 1)/M and (M + 2)/M ratios are now well below ±5% on average and do not depend on the mass of the considered ion. So, we can rule out the existence of mass bias effects at the level of confidence given by the experimental variability. In summary, we can conclude that the different quadrupole mass spectrometers tested provide isotope compositions in agreement with those theoretically calculated when the contribution at M  1 and/or M  H was taken into account. So, we propose an alternative procedure to evaluate the spectral accuracy of mass spectrometers taking into account the purity of the mass cluster and the resolution of the mass spectrometer by measuring the whole isotopic profile of the compounds. The application of Eqn 1 allows the estimation of these parameters and provides an alternative mode for the evaluation of spectral accuracy: the SSR of the multiple linear regression. Residual values of the order of 105 to 106 were routinely obtained for the three quadrupole mass spectrometers evaluated which indicated a very good agreement with the theoretical abundances. This is a clear advantage of the instruments tested over other mass spectrometers described in the recent literature[5–7,13] where non-linear response seems to be the limiting factor to achieve good spectral accuracy.

0.001

0.0001

0.00001

0.000001 0.98

0.985

0.99

0.995

1

Isotope enrichment 13

Figure 4. Determination of the isotope enrichment for nonylphenol C6 13 2 (white points), tert-octylphenol C6 (black points) and clenbuterol H9 (grey points).

for one typical injection of three of these compounds when changing the isotope enrichment from 98 to 100% in the software. As can be observed, clear minima are obtained for all compounds with SSR values lower than 104 at the optimum enrichment value. Similar plots were obtained for all injections of each compound with the different instruments employed. The different isotope enrichments measured for these compounds both by GC–MS and LC–MS/MS are shown in Table 3. As can be observed, similar enrichment values were obtained by both MS techniques which provide high confidence in the results. For nonylphenol 13C6 and clenbuterol 2H9, the results were in agreement with those given by the manufacturers. However, for the two octylphenols, there was a clear discrepancy between the measured and the indicated values. In order to understand these differences between the indicated and experimental abundances, we compared the experimental abundances observed with those calculated at the isotope enrichment provided by the manufacturers and at that obtained using our method. Then, we compared the results with the mass spectrum shown in the Certificate of Analysis provided by the manufacturer. The results for tert-octylphenol 13C6 are shown in Fig. 5. As can be observed, the experimental isotope pattern compares very well with that calculated at 99.3% enrichment when taking into account the lack of purity of the mass cluster. On the other hand, the isotope distribution calculated for a 93.5% abundance is completely different from the experimental abundance even when the lack of purity of the mass cluster is not taken into account in the calculations. The inset in Fig. 5 shows a snapshot of the mass spectrum provided by the manufacturer in the Certificate of Analysis. No numerical values can be extracted from the mass spectrum given, but the isotope distribution observed is very similar to that measured here and is completely different from the theoretical distribution assuming a 93.5% enrichment. We could not obtain any information from the manufacturer on how the isotope enrichments were calculated. The isotope enrichment of β2-agonists All labelled β2-agonists were singly 13C labelled in the carbon adjacent to the aromatic ring (see structure in Fig. 1). In order to determine their final isotope enrichment after synthesis, both

Copyright © 2014 John Wiley & Sons, Ltd.

J. Mass Spectrom. 2014, 49, 681–691

Enrichment of isotopically labelled molecules Table 3. Isotope enrichment (at%) determined for commercially available compounds using alternative MS techniques (in parentheses standard deviation from n = 5 injections). In the case of nonylphenol, only the clusters at mass 220 were employed in the calculations. The last column shows the average isotope enrichment (at%) and their propagated standard uncertainties (in parentheses) as determined from the different MS techniques Compound tert-Octylphenol

13

C6

2

tert-Octylphenol H2 13

n-Nonylphenol C6 2 Clenbuterol H9 a

GC–MS

LC–MS/MS standard resolution

LC–MS/MS high resolution

Indicated value

Average isotope enrichment

99.37 (0.03)a 99.25 (0.02)b 99.02 (0.06)a 98.89 (0.06)b 99.24 (0.05) 98.55 (0.01)

99.38 (0.04)

99.36 (0.02)

93.5

99.3 (0.1)

98.95 (0.02)

98.85 (0.02)

97.7

98.9 (0.1)

99.11 (0.03) 98.52 (0.01)

98.78 (0.11) 98.47 (0.01)

99 >97.0

99.0 (0.3) 98.5 (0.1)

Cluster at nominal mass 135. Cluster at nominal mass 177.

b

GC–MS, GC–MS/MS and LC–MS/MS measurements were performed as indicated in the procedures. The final results are collected in Table 4. For the singly 13C-labelled compounds, the isotope enrichments obtained were higher than 98% except for terbutaline where only ca. 90% enrichment was achieved. In most cases, the results from different techniques agree for the same compounds. However, there were some disagreements (e.g. clenproperol by GC and LC), and then a more detailed study of the uncertainty sources of the different techniques was carried out.

1 Exp 13C6

0.9

Theo 13C6 99.3% M-H corr 141

Theo 13C6 93.5% M-H corr

Fractional abundance

0.8

Theo 13C6 93.5% no M-H corr

0.7 0.6 0.5 0.4 0.3 120

0.2

130

140

The uncertainties of the measured isotope enrichments

150 m/z

0.1 0 138

139

140

141

142

143

Nominal mass

Figure 5. Comparison of fractional abundances obtained for tert13 octylphenol C6 by GC–MS at different isotope enrichments with and without cluster purity correction. The error bars in the experimental abundances correspond to the standard deviation of n = 9 injections. The inset shows part of the mass spectrum of the compound provided by the manufacturer in the certificate of analysis.

From Tables 3 and 4, we can observe that the standard deviations of the results for five independent injections were, in general, very low for all compounds and instruments tested. For example, the isotope enrichments calculated for clenproperol by LC–MS/MS have standard deviations of 0.02–0.03% which make the results obtained by LC–MS/MS statistically different from those obtained by GC–MS and GC–MS/MS. However, in the calculation of the isotope enrichment of these molecules, another important source of uncertainty is the fragmentation factors measured for the natural abundance compounds (Table 1). We have looked into the error propagation of the fragmentation factors to the computed isotope

Table 4. Carbon isotope enrichment (at%) for the nine β2-agonists using alternative MS techniques (in parentheses standard deviation from n = 5 injections). The last column shows the average isotope enrichment (at%) and their propagated standard uncertainties (in parentheses) as determined from the different MS techniques Compound 13

Clenbuterol C1 13 Brombuterol C1 13 Clenproperol C1 13 Clenpenterol C1 13 Clenhexyl C1 13 Clencyclohexerol C1 13 Salbutamol C1 13 Terbutaline C1 13 Ractopamine C1

GC–MS

GC–MS/MS

LC–MS/MS standard resolution

LC–MS/MS high resolution

Average isotope enrichment (at%)

98.78 (0.06) 99.06 (0.07) 98.64 (0.28) 99.08 (0.13) a,c 100 (0.00) 99.30 (0.11) 99.18 (0.12) 91.08 (0.59) –b

97.98 (0.07) 98.35 (0.06) 98.79 (0.17) 99.16 (0.19) 99.85 (0.03) 98.72 (0.19) 99.72 (0.01) 89.41 (0.83) –b

98.57 (0.41) 99.09 (0.09) 99.24d (0.02) 99.40 (0.08) 99.37 (0.06) 98.63 (0.22) 99.43 (0.09) 91.84 (0.02) 99.38 (0.04)

98.56a (0.05) 99.08 (0.04) 99.30d (0.03) 99.33c (0.04) 99.06a (0.05) 98.45 (0.10) 99.22 (0.14) 91.70 (0.09) 99.05 (0.11)

98.5 (0.9) 98.9 (0.6) 99.0 (0.7) 99.2 (0.4) 99.4 (0.7) 98.8 (0.6) 99.4 (0.4) 91.0 (1.3) 99.2 (0.4)

4

Square sum of residuals for the natural abundance compound >10 . Not measured. 4 c Square sum of residuals for the labelled compound >10 . + d Determined on the (M  18)H fragment. a

b

Copyright © 2014 John Wiley & Sons, Ltd.

689

J. Mass Spectrom. 2014, 49, 681–691

wileyonlinelibrary.com/journal/jms

A. González-Antuña, P. Rodríguez-González and J. I. García Alonso enrichments and observed that the uncertainties in the M  1(H) factors, uM  1(H), propagated directly into the uncertainties of the isotope enrichments, uie, with the equation: uie ð%Þ≈

uM1ðHÞ ð%Þ m

(4)

where m is the number of enriched isotopes in the molecule. So, the propagation of the uncertainty in the fragmentation factors is lower for multiply labelled compounds than for singly labelled. If we take into account the uncertainties given in Table 1 and propagate those for the different measurements given in Tables 3 and 4 using Eqns 3 and 4, we obtain the final results shown in the last column of both tables. Now, the final uncertainties cover the small differences observed between instruments. For clenproperol, the isotope enrichment is now given as 99.0% with 0.7% combined uncertainty which includes all the results obtained both by GC and LC. As can be observed in Table 4, except for terbutaline, all compounds show enrichments in excess of 98% with uncertainties lower than 0.9%. Better uncertainty values were obtained for the multiply labelled compounds shown in Table 3 as m > 1. So, it is important to determine the uncertainty in the fragmentation factors to calculate the uncertainties in the isotope enrichments, particularly for singly labelled compounds. These final uncertainty values can be employed for uncertainty propagation when these compounds are used for IDMS in combination with isotope pattern deconvolution.[2,19,20]

Conclusions We have developed and applied a procedure for the evaluation of spectral accuracy in different mass spectrometers and applied it for the calculation of the isotope enrichment of isotopically labelled compounds. The method measures the whole isotopic cluster and takes into account the purity of the mass cluster and/or the tailing of the mass peak at the low mass side of the spectrum. When these contributions are not taken into account, the calculated isotope enrichment values could be biased, particularly for singly labelled compounds. The method provides also the means to calculate the uncertainty of the measured isotope enrichment. We have observed that quadrupole-based GC–MS, GC–MS/MS and LC–MS/MS instruments are well suited for isotope enrichment measurements when the contribution at M  1(H) is taken into account. The method was applied to the determination of the isotope enrichment of β2-agonists singly labelled with 13C and alkylphenols with satisfactory results in terms of high isotope enrichment of the labelled compounds and low propagated isotope enrichment uncertainties. Acknowledgements

690

The authors are grateful for financial support from the Spanish Ministry of Economy and Competitiveness through Projects Ref. CTQ2009-12814 and Ref. CTQ2012-36711. The EU is acknowledged for the provision of FEDER funds for the purchase of the GC–MS/MS and LC–MS/MS instruments. Ana González-Antuña acknowledges her doctoral grant Ref. BP 09056 to FICYT (Asturias). Pablo Rodríguez-González acknowledges his research contract RYC-2010-06644 to the Spanish Ministry of Economy and Competitiveness through the Ramón y Cajal Program.

wileyonlinelibrary.com/journal/jms

References [1] M. C. Hellerstein, R. A. Neese. Mass isotopomer distribution analysis at eight years: theoretical, analytic and experimental considerations. Am. J. Physiol. 1999, 282, E1146. [2] J. I. García Alonso, P. Rodríguez-González. Isotope dilution mass spectrometry. Royal Society of Chemistry: Cambridge (UK), 2013. [3] P. Rodriguez-Gonzalez, S. Bouchet, M. Monperrus, E. Tessier, D. Amouroux. In situ experiments for element species-specific environmental reactivity of tin and mercury compounds using isotopic tracers and multiple linear regression. Environ. Sci. Pollut. Res. 2013, 20, 1269 [4] J. G. Burr. Tracer Applications for the Study of Organic Reactions. Interscience: New York, 1957. [5] L. Ouerdane, J. Meija, S. Bakirdere, L. Yang, Z. Mester. Nonlinear signal response in electrospray mass spectrometry: implications for quantitation of arsenobetaine using stable isotope labeling by liquid chromatography and electrospray Orbitrap mass spectrometry. Anal. Chem. 2012, 84, 3958. [6] J. C. L. Erve, M. Gu, Y. Wang, W. DeMaio, R. E. Talaat. Spectral accuracy of molecular ions in an LTQ/Orbitrap mass spectrometer and implications for elemental composition determination. J. Am. Soc. Mass Spectrom. 2009, 20, 2058. [7] Y. Xu, J.-F. Heilier, G. Madalinski, E. Genin, E. Ezan, J.-C. Tabet, C. Junot. Evaluation of accurate mass and relative isotopic abundance measurements in the LTQ-Orbitrap mass spectrometer for further metabolomics database building. Anal. Chem. 2010, 82, 5490. [8] M. Taubert, S. Baumann, M. von Bergen, J. Seifert. Exploring the 13 limits of robust detection of incorporation of C by mass spectrometry in protein-based stable isotope probing (protein-SIP). Anal. Bioanal. Chem. 2011, 401, 1975. [9] M. C. B. Moraes, J. G. A. Brito-Neto, V. F. Juliano, C. L. do Lago. Eliminating interferences of M-nH ions in isotope patterns from low-resolution mass spectra. Int. J. Mass Spectrom. 1998, 178, 129. [10] J. Meija, J. A. Caruso. Deconvolution of Isobaric Interferences in Mass Spectra. Am. Soc. Mass Spectrom. 2004, 15, 654. [11] A. González-Gago, J. M. Marchante-Gayón, M. Ferrero, J. I. Garcia 81 Alonso. Synthesis of Br-labeled polybrominated diphenyl ethers and their characterization using GC(EI)MS and GC(ICP)MS. Anal. Chem. 2010, 82, 2879. [12] C. K. Fagerquist, R. A. Neese, M. K. Hellerstein. Molecular ion fragmentation and its effects on mass isotopomer abundances of fatty acid methyl esters ionized by electron impact. J. Am. Soc. Mass Spectrom. 1999, 10, 430. [13] S. Böcker, M. C. Letzel, Z. Lipták, A. Pervukhin. SIRIUS: decomposing isotope patterns for metabolite identification. Bioinformatics 2009, 25, 218. [14] M. Dauner, U. Sauer. GC-MS analysis of amino acids rapidly provides rich information for isotopomer balancing. Biotechnol. Prog. 2000, 16, 642. [15] M. R. Antoniewicz, J. K. Kelleher, G. Stephanopoulos. Accurate assessment of amino acid mass isotopomer distributions for metabolic flux analysis. Anal. Chem. 2007, 79, 7554. [16] M. J. MacCoss, C. C. Wu, D. E. Matthews, J. R. Yates. Measurement of the isotope enrichment of stable isotope-labeled proteins using high-resolution mass spectra of peptides. Anal. Chem. 2005, 77, 7646. [17] A. Castillo, E. Gracia-Lor, A. F. Roig-Navarro, P. Rodríguez González, J. I. Garcia Alonso, F. Hernandez Hernandez. Isotope pattern deconvolution-tandem mass spectrometry for the determination and confirmation of diclofenac in wastewaters. Anal. Chim. Acta 2013, 765, 77–85. [18] A. González-Antuña, I. Lavandera, P. Rodríguez-González, J. Rodríguez, 13 J. I. García Alonso. Straightforward route to obtain C1-labeled clenbuterol. Tetrahedron 2011, 67(31), 5577. [19] A. González-Antuña, P. Rodríguez-González, G. Centineo, J. I. García Alonso. Evaluation of minimal 13C-labelling for stable isotope dilution in organic analysis. Analyst 2010, 135, 953. [20] A. González-Antuña, P. Rodríguez-González, I. Lavandera, G. Centineo, V. Gotor, J. I. García Alonso. Development of a routine method for the simultaneous confirmation and determination of clenbuterol in urine by minimal labeling isotope pattern deconvolution and GC-EI-MS. Anal. Bioanal. Chem. 2012, 402, 1879.

Copyright © 2014 John Wiley & Sons, Ltd.

J. Mass Spectrom. 2014, 49, 681–691

Enrichment of isotopically labelled molecules [21] H. Kubinyi. Calculation of isotope distributions in mass spectrometry. A trivial solution for a non-trivial problem. Anal. Chim. Acta 1991, 247, 107. [22] J. I. García Alonso, P. Rodríguez-González, A. González-Gago, A. González-Antuña. Determination of the uncertainties in the theoretical mass isotopomer distribution of molecules. Anal. Chim. Acta 2010, 664, 68. [23] J. I. García Alonso, P. Rodríguez-González. Response to “Comments on the uncertainties in isotope patterns of molecules” by J. Meija and Z. Mester. Anal. Chim. Acta 2011, 694, 177. [24] N. Fabregat-Cabello, A. Castillo, J. V. Sancho, F. V. González, A. F. Roig-Navarro. Fast methodology for the reliable determination of

nonylphenol in water samples by minimal labeling isotope dilution mass spectrometry. J. Chrom. A. 2013, 1301, 19. [25] Quantifying Uncertainty in Analytical Measurements, EURACHEM/ nd CITAC Guide, 2 edition, 2000.

Supporting information Additional supporting information may be found in the online version of this article at the publisher’s web site.

691

J. Mass Spectrom. 2014, 49, 681–691

Copyright © 2014 John Wiley & Sons, Ltd.

wileyonlinelibrary.com/journal/jms

Determination of the enrichment of isotopically labelled molecules by mass spectrometry.

A general method for the determination of the enrichment of isotopically labelled molecules by mass spectrometry (MS) is described. In contrast to oth...
583KB Sizes 0 Downloads 3 Views