metabolomics
Koek, Maud Marijtje
Citation
Koek, M. M. (2009, November 10). Gas chromatography mass spectrometry: key technology in metabolomics. Retrieved from https://hdl.handle.net/1887/14328
Version: Corrected Publisher’s Version
License: Licence agreement concerning inclusion of doctoral thesis in the Institutional Repository of the University of Leiden
Downloaded from: https://hdl.handle.net/1887/14328
4
Towards quantitative metabolomics with silylation-based gas chromatography- mass spectrometry:
key issues and recommendations
ABSTRACT
Silylation in combination with gas chromatography-mass spectrometry has become one of the standard methods in metabolomics analysis. Silylation significantly increases the application range of gas chromatography in metabolomics analysis, however, precise analysis of all metabolites that can be silylized is very challenging. In this paper, the differences in method performance for different types of metabolites were investigated and the strategies to improve the repeatability of the analysis of derivatized metabolites were explored and demonstrated. Finally, some recommendations for silylation-based metabolomics are provided.
74
INTRODUCTION
The aim in metabolomics research is to measure all (or at least as many as possible) metabolites in cells, body fluids or tissue, in order to gain new insights in the functioning of biological systems.
Gas chromatography coupled to mass spectrometry (GC-MS) is a highly suitable technique for comprehensive non-target analysis as it combines a very high separation power and repeatable retention times with sensitive and selective mass detection. In addition, the possibility of using an electron ionization interface in GC-MS allows for the indiscriminate detection of almost all compounds amendable to GC. However, many metabolites contain polar functional groups that are thermally labile at temperatures required for their separation or are not volatile at all. In addition, the peak shapes or recoveries of polar compounds can be unsatisfactory, due to adsorption on the surface of the analytical column. Therefore, a derivatization prior to analysis is required to measure metabolomics samples adequately with GC-MS.
In previous papers a one-dimensional GC-MS method1 and a comprehensive two- dimensional GC×GC-MS method2 suitable for the analysis of a broad range of small polar metabolites were described. These methods use using a derivatization with an oximation reagent, followed by silylation. Several other GC-MS based methods for metabolomics, also using a comprehensive silylation approach prior to analysis, have been reported.3-10
Silylation reagents are the most versatile and universally applicable reagents to render polar metabolites GC-amenable. They are therefore considered very suitable for comprehensive non-target analysis. Nearly all polar functional groups which present problems in gas chromatographic analysis, such as hydroxyl, amine, amide, phosphate and thiol groups, can be converted to their alkylsilyl derivatives. Additionally, the introduction of an alkylsilyl group increases the MS response, due to the additional mass of the derivatized metabolite. However, the reactivity of different functional groups towards silylation and the stability of the derivatized metabolites can vary. In general the ease of silylation of the parent compound and the stability of the derivatized compound decreases in the following order: - hydroxyl (alcohol; primary > secondary
> tertiary) > hydroxyl (phenol) > carboxylic acid > amine (primary > secondary) >
amide.15
The ease of silylation or reactivity of metabolites is influenced by the thermodynamics and kinetics of the reaction, i.e. the increase or decrease of the free Gibbs energy (∆G
= enthalpy (∆H) – entropy (S)· temperature (T); thermodynamics) and the activation energy of the reaction (kinetics) determine the feasibility of a transformation. The reactivity of a silyl group towards oxygen is higher than with nitrogen due to several factors, i.e. (1) oxygen has two unshared electron pairs while nitrogen has only one, (2) the oxygen electrons are sterically more available and (3) the greater possibility of
(p→d)π back-bonding, i.e. delocalization of unshared electrons in the empty d-orbitals of Si in Si-O compared to Si-N lowers the energy of the transition state more (lower activation energy).15 In addition, the bond strength increases with increasing (p→d)π back-bonding. Therefore, the enthalpy of the reaction of Si with O is generally more exothermic than the reaction with N. Furthermore, the reactivity decreases in the order alcohol > phenol > carboxylic acid as the unshared electron pairs are increasingly used in conjugative effect within the molecule, decreasing the extent of (p→d)π back- bonding 15. The order of stability of the derivatized metabolites mentioned above can also be explained by the thermodynamics and kinetics of the silylation reaction.
Derivatized metabolites with the largest differences in enthalpy upon derivatization (endothermic reaction back to parent metabolite) and highest bond strengths (higher activation energy for reaction back to parent compound), i.e. silylated alcohols, are the most stable.
Early on in the development of the GC-MS method for metabolomics, the performance for metabolites originating from different compound classes was assessed and metabolites were classified into three groups based on their analytical performance (Chapter 3).1 Performance class-1 metabolites are metabolites containing hydroxylic and carboxylic functional groups, such as sugars, fatty acids and organic acids. The analytical performance for these metabolites is generally very satisfactory.
Performance class-2 type metabolites, metabolites containing amine or phosphoric functional groups, can also be measured with satisfactory derivatization efficiencies, repeatability and intermediate precision. However the analysis of these metabolites is more critical compared to class-1 metabolites, when the method is not carried out under
‘optimal’ conditions. Metabolites with amide, thiol or sulfonic functional groups, so- called performance class-3 compounds, are more difficult to derivatize and analyze.
Additionally these derivatized metabolites are less stable and therefore more prone to degrade or adsorb on the surface of the liner and/or analytical column.
An important effect observed in GC(×GC)-MS analysis is the matrix-enhancement effect, i.e. for some metabolites the MS response when sample matrix is present is higher compared to the MS response in academic standards, i.e. samples without any matrix.1 This effect has been described earlier in pesticide analysis11-14; most probably certain compounds from the matrix block active sites in the analytical system, causing the increased response for metabolites that are prone to adsorb or degrade on the active sites on the surface of the liner and/or the analytical column. As a result the method performance for some metabolites is significantly better in matrix than in academic standards. The matrix-enhancement effect is not a constant factor, but is influenced by the composition of the matrix and depends on the ‘activity’ state of the GC system. In contrast, some compounds can lower the recovery of metabolites when these are present at high concentrations in the matrix, such as certain components from growth
76
media or buffer components in microbial metabolomics or anti-coagulants in blood plasma.
The level of difficulty of analysis can differ significantly between matrices. Based on our experience, serum and plasma are considered class 1 matrices, i.e. relatively easy to analyze with generally high intermediate precision. Tissues, plants and urine (class 2) are more difficult matrices, especially due to larger differences between the compositions of samples. Microbial matrices (class 3) are the most difficult matrices due to the extremely large differences between samples and sometimes very high concentrations of compounds from the extraction/quenching buffer or residuals of the fermentation medium. Therefore, the precise analysis of these matrices is more critical1. Obviously, this is a general classification and for every matrix specific challenges and problems can occur that complicate the analysis. For example the use of EDTA as anticoagulant in plasma can have a significant adverse effect on the performance of the GC-MS method.
During the last years, large numbers of metabolomic samples of many different sample matrices were analyzed in our laboratory. In this chapter, the differences in method performance were investigated for different metabolites from different performance classes. Furthermore, the influence of different matrices and the inertness of the analytical system on the performance for the different performance classes of metabolites were systematically investigated. For this purpose, an evaluation mixture was used, composed of metabolites from the three defined classes (Table 1). In several experiments, critical factors in OS-GC-MS analysis are illustrated and ways to improve the precision of the analysis are presented.
Table 1 Metabolites and their performance class used for evaluation of the performance of silylation based GC-MS analysis
Class 1 (green) Class 2 (orange) Class 3 (red)
Cholic acid-d4 Leucine-d3 Nicotinamide
Citric acid Phenylalanine-d5 Glutamic acid-d3
Fructose Glucose-6-phosphate
Glucose-d7 Ribose-5-phosphate
For an explanation of the classes, cf. Introduction.
EXPERIMENTAL
Chemicals and materials
Pyridine (Baker analyzed) was purchased from Mallinkrodt Baker (Deventer, The Netherlands) and pyridine hydrochloride (analytical grade), was purchased from Sigma-Aldrich (Zwijndrecht, The Netherlands). A solution of 56 mg/mL ethoxyamine hydrochloride (> 99%, Acros Organics, Geel, Belgium) in pyridine was used for oximation, N-methyl-N-trimethylsilyl trifluoroacetamide (MSTFA, Alltech, Breda, The Netherlands) was used for silylation and trimethylchlorosilane (analytical grade, Sigma-Aldrich) was used in some experiments as a catalyst of the silylation reaction.15 All chemicals used to study matrix enhancement effects were purchased from Sigma- Aldrich.
OV1701 deactivated liners were purchased from BGB Analytik (Boeckten, Switzerland), Siltek liners were purchased from Restek (Bellefonte, PA, USA) and Gerstel deactivated liners were purchased from Gerstel (Mülheim an der Ruhr, Germany). Teflon autosampler vials (Savilex, Minotonka, Minnesota, USA) and deactivated borosilicate glass autosampler vials (Chromacol, Herts, UK) were purchased from Alltech.
Standards
Leucine-d3, glutamic acid-d3, phenylalanine-d5, glucose-d7 and cholic acid-d4 were purchased from Cambridge Isotope Laboratories (Andover, Massachusetts, USA). All other standards and compounds were purchased from Sigma-Aldrich Chemie. US Defined Fetal Bovine Serum was purchased from Hyclone (Logan, Utah, USA).
Evaluation mix
An evaluation mix composed of ten metabolites (100 ng/µL per metabolite, Table 1) with varying physio-chemical properties, i.e. citric acid, cholic acid-d4, leucine-d3, glutamic acid-d3, phenylalanine-d5, fructose, glucose-d7, ribose-5-phosphate, glucose- 6-phosphate and nicotinamide, was prepared in methanol/water 1:4 v/v to evaluate the reactivity, stability and repeatability of analysis for different compound classes. For all experiments the amount of evaluation mixture used was adjusted to obtain a final concentration of 10 ng/µL per metabolite in the derivatized samples. An internal standard (IS) solution containing dicyclohexylphthalate in pyridine was spiked to all samples prior to analysis to correct for small variations in injection volume and MS response. The final concentration of dicyclohexylphthalate in the injection solution was 10 ng/µL. If a sample did not contain any matrix (i.e. no serum, microbial extract, etc.), the sample was referred to as an academic standard.
78
Blank fetal bovine-serum matrix
Methanol was added to the fetal bovine serum to precipitate the proteins; a volume of four times the sample volume was used. Then the mixture was centrifuged for 10 min at 10,000 rpm. The supernatant was transferred to a glass container and stored at -20°C until analysis, but no longer than one month after preparation.
Unless stated otherwise, 500 µL of supernatant (equivalent of 100µl serum) was used to prepare samples with fetal bovine serum matrix.
Derivatization of samples
After addition of the evaluation mix and eventually blank fetal bovine serum matrix, samples were dried under a nitrogen flow. Unless stated otherwise, samples were dried and derivatized in Teflon autosampler vials as follows. 35 µL ethoxyamine reagent was added and the samples were oximated for 90 min at 40°C on a tube roller in a stove.
Subsequently, 10 µL of internal standard (dicyclohexylphthalate) solution and 100 µL MSTFA were added and the samples were silylated for 45 min at 40°C on a tube roller in a stove.
GC-MS analysis
Unless stated otherwise, samples were analyzed with the following method. The derivatized extracts were analyzed with an Agilent 6890 gas chromatograph coupled with an Agilent 5973 mass selective detector. 1-µL aliquots of the extracts were injected into a HP5-MS capillary column (30 m x 250 µm I.D., 0.25 µm film thickness;
Agilent technologies, Santa Clara, CA, USA) using PTV-injection (Gerstel CIS4) in the splitless mode. The temperature of the PTV was 70°C during injection and 0.6 min after injection the temperature was raised to 300°C at a rate of 2°C per second and held at 300°C for 20 min. The initial GC oven temperature was 70°C; 5 min after injection the GC oven temperature was increased by 15°C/min to 320°C and held for 5 min at 320°C. Helium was used as a carrier gas and the pressure was programmed such that the helium flow was kept constant at a flow rate of 1.7 mL per min. Detection was achieved using MS in electron ionization mode and full-scan monitoring mode (m/z 15–800). The temperature of the ion source was set at 250°C and that of the quadrupole at 200°C.
Reactivity and stability of metabolites from different compound classes
Silylation reaction curves
The influence of the amount of MSTFA on the derivatization efficiency of metabolites was studied by derivatizing 50 µL of the evaluation mix, dried under nitrogen flow, with increasing amounts of MSTFA. The amount of oximation reagent was kept constant and the total volume of sample was also kept constant by adding varying amounts of pyridine (Table 2). In this way curves of the MS response versus the MSTFA amount, so-called silylation reaction curves, could be constructed, using
different experimental conditions, including different derivatization times and temperatures, addition or no addition of trimethylchlorosilane (TMCS; a possible catalyst for silylation15) and with or without addition of 500 µL fetal bovine-matrix (Table 3). In addition, the repeatability of derivatization and the stability of the derivatized metabolites (stored at -20°C) at different MSTFA concentrations was investigated.
Stability of derivatized metabolites
Curves 1, 4, 5, 6 and 7 were re-analyzed after storage for 28 days in a freezer at -20°C.
In addition, the stability of the derivatized metabolites in fetal bovine serum matrix at - 20°C using the standard derivatization protocol was evaluated by comparing the normalized MS response of spiked metabolites in fetal bovine serum samples on day 1, i.e. the day of derivatization of the samples, and after storage for 84 days at -20°C (cf.
Supplemental data, Table S1).
Table 2 Preparation of curves of MS response versus MSTFA amount a) Mix standard
(µL)
Ethoxyamine reagent (µL)
MSTFA (µL) Pyridine (µL) Total volume (µL)
1 50 50 25 375 450
2 50 50 40 360 450
3 50 50 50 350 450
4 50 50 75 325 450
5 50 50 75 325 450
6 50 50 100 300 450
7 50 50 150 250 450
8 50 50 250 150 450
9 50 50 400 0 450
a) A total of 9 samples with different MSTFA amounts per curve were prepared.
Table 3 Experimental conditions of different curves (MS response vs MSTFA amount) Curve Silylation time
(min)
Silylation temperature (°C)
1% TMCS added?
Fetal bovine serum added?
1 40 45 No No
2 90 45 No No
3 45 100 No No
4 90 100 No No
5 40 45 Yes No
6 65 70 No No
7 65 70 No No
8 40 45 No Yes
9 90 100 Yes Yes
80
Inertness of analytical system
Liner deactivation
Long sequences of academic standards (no matrix added) of the evaluation mix alternated with spiked (evaluation mix) fetal bovine serum or microbial samples were analyzed using differently deactivated injection liners (i.e. deactivated liners (Gerstel), siltek liners (Restek), quartz liners deactivated with high temperature silylation, hexamethyldisilazane deactivated liners and OV1701 deactivated liners (BGB)) to assess the performance for varying metabolites. For the test with fetal bovine serum, the sequence was as follows: first ten academic standards followed by three spiked serum samples, one academic standard, three serum samples, one academic standard, etc.. For the test with Propionibacterium freudenreichii extract, the analysis of 2 standards was alternated with 5 microbial extracts, etc.. The microbial extracts were prepared as reported earlier1, 0.027 g (dry weight) of microbial cell content was present in the final extract (1 mL) after derivatization.
Glass versus teflon autosampler vials
Furthermore, the performance of the method was tested using deactivated glass or teflon autosampler vials. For this purpose the evaluation mix (additionally containing glycerol-3-phosphate) was derivatized in quintuplicate in teflon and glass autosampler vials. In addition, the mix was spiked to 500 µL blank fetal bovine serum in glass and teflon autosampler vials and derivatized in quintuplicate.
Matrix-enhancement effect
Experiment I: Comparison of protective abilities of amides, amines and sugars with varying volatilities
Several compounds from different compound classes and with different volatilities were tested, including amides (octafluorohexanediamide and already silylized bis- (trimethylsilyl)-urea), sugars (deoxyribose, glucose and melibiose) and amines (1,12- diaminododecane). Six vials with 100 µL of the mix standard were dried, and 2.5 mg of each compound were added to a separate vial and derivatized further according to the method described above; bis(trimethylsilyl)-urea is a silylation reagent and was added only after oximation to prevent untimely silylation.
Experiment II: Comparison of academic standards with fetal bovine serum matrix The response of the test metabolites from the evaluation mix with added bis- (trimethylsilyl)-urea (2.5 mg/100 µL), 1,12-diaminododecane (2.5 mg/100 µL) or a combination of bis-(trimethylsilyl)-urea (2.5 mg/100 µL) and 1,12-diaminododecane (2.5 mg/100 µL) was compared to the response of the metabolites spiked to fetal bovine serum matrix. Furthermore, a combination of bis-(trimethylsilyl)-urea and 1,12- diaminododecane (2.5 mg per 100 µL for both compounds) was added to fetal bovine-
serum matrix spiked with the standard mix to study if the matrix enhancement of serum would increase.
Experiment III: ‘Worst case’ scenario
Six samples were prepared containing the mix standard. To four of the dried samples 1.7 mg of ammoniumsulfate was added. To two of the samples with the ammoniumsulfate, bis-(trimethylsilyl)-urea (2.5 mg per 100 µL) was added. All samples were derivatized with the method described above.
82
RESULTS
The influence of three critical factors, i.e. inertness of the analytical system, the compound class of the metabolite and the sample matrix, on the analytical performance of a range of different metabolites (Table 1) was studied. Although these factors are interrelated and the final method performance for a specific metabolite is influenced by all factors, these critical factors are discussed separately for the clarity of the representation of the results.
Table 4 Relative MS responses a) and RSDs of metabolites derivatized in glass and Teflon autosampler vials
Average relative response (n=5) a) RSDs (n=5)
Compound Name Academic standard Serum matrix Academic standard Serum matrix Teflon Glass Teflon Glass Teflon Glass Teflon Glass Class 1
Citric acid 100 70 n.d. b) n.d. b) 11 8 3 7
Fructose 100 65 n.d. b) n.d. b) 3 19 6 8
Glucose-d7 100 45 n.d. c) n.d. c) 3 32 4 9
Cholic acid-d4 83 67 100 96 4 8 9 7
Class 2
Leucine-d10 91 68 100 105 2 8 7 12
Phenylalanine-d5 77 42 100 92 3 22 7 8
Glycerol-3- phosphate
91 64 100 91 5 11 7 12
Ribose-5-phosphate 65 42 100 103 8 9 3 15
Glucose-6-phosphate 57 23 100 94 11 23 4 19
Class 3
Nicotinamide 59 38 100 82 7 5 6 10
a) First the average normalized response was calculated by correction of the peak areas with that of the IS (dicyclohexylphthalate); next the averaged normalized response per metabolite was divided by that obtained in the serum matrix and using Teflon vials and multiplied with 100; only for citric acid, fructose and glucose-d7 the averaged normalized response was divided by that obtained in the academic standard using Teflon vials.
b) n.d.: not determined; these compounds were also present in fetal bovine-serum matrix, therefore, the relative response was not calculated.
c) n.d.: not determined; due to the presence of a large amount of glucose in the serum matrix the relative response of glucose-d7 could not be determined accurately.
Inertness of the analytical system
Liner deactivation
Early on in the development of the GC method it was observed that the repeatability of PTV injection was improved significantly when performing an extra deactivation of the deactivated borosilicate injection liner. Several differently deactivated liners (cf.
Experimental) were assessed, and the performance for metabolites from all three performance classes (see Table 1) was evaluated. The performance of the OV1701 deactivated liner was satisfactory for our application and was chosen based on the
general performance and repeatable results obtained over a longer period of time for different metabolites and varying sample matrices (data not shown). The method performance varied significantly between different matrices, for example microbial or plasma/serum extracts. For microbial samples the liner had to be exchanged every 20 – 30 samples to maintain acceptable results for critical metabolites, such as glutamic acid, nicotinamide and cholic acid.1 However, the performance for the same metabolites in serum samples was satisfactory even after 54 injections of serum and 28 injections of academic standards (data not shown); for serum and plasma samples the injection liner is usually changed every 40 – 45 injections.
Inert and repeatable liner deactivation and timely exchange of the injection liner can considerably improve the precision of GC-MS methods. It should be noted that the amount of matrix injected into the GC-MS has an important influence when to exchange the liner. Besides the OV1701-deactivated liner, most certainly other deactivated liners can also be suitable for analyzing metabolomics samples with a GC- MS method. However, based on our experience we used the OV1701-deactivated liner.
Glas versus teflon autosampler vials
The relative response and the repeatability of derivatization in glass and Teflon autosampler vials were compared for metabolites from different compound classes (Table 4). The metabolites were derivatized directly in an autosampler vial and injected without transfer of the sample according to the standard procedure (cf. Experimental).
The relative responses of metabolites derivatized in Teflon autosampler vials were significantly higher than the relative responses obtained in glass autosampler vials (for the normalization procedure see Table 4). In addition, the relative standard deviations (RSDs) were better in Teflon vials. For academic standards the difference in performance was more pronounced. However, also in serum matrix the performance using Teflon vials was better.
The low recovery and high RSDs of glucose and fructose in glass vials was caused by the formation of artefacts. Besides the two normal peaks (cis- and transethoxime form of acyclic glucose and fructose) two extra peaks were formed for both glucose and fructose when the derivatization was performed in glass autosampler vials. Especially for nicotinamide and the sugar-phosphates a matrix-enhancement effect was observed, i.e. the relative response was higher in the serum sample compared to the academic solutions. This is probably caused by decreased adsorption and/or degradation on the surface of the liner and analytical column in the presence of matrix components. This effect is investigated in the section Matrix-enhancement effect below and further discussed in Discussion.
In conclusion, the use of Teflon autosampler vials was preferred over glass autosampler vials. It should be noted that the authors are aware that a wide range of
84
deactivated autosampler vials are available, and that there might be glass autosampler vials performing better than the ones investigated here.
Reactivity and stability of metabolites from different compound classes
Silylation reaction curves
The reactivity of metabolites from different compound classes was examined by constructing curves (so-called silylation reaction curves) of the MS response versus the MSTFA amount (as described in Table 2) at different experimental conditions (Table 3). For all metabolites tested (see Table 1) no significant differences in the response of the metabolites at the maximal amount of MSTFA or the shape of the curve were observed for the same metabolite under the different experimental conditions of Table 3; as an example, the curves for phenylalanine-d5 are shown in S-Figure 1.
Consequently, a silylation time of 45 min and temperature of 40°C were sufficient to reach a thermodynamic equilibrium between the silylation reagent and the investigated metabolites. Furthermore, the addition of TMCS as possible catalyst had no effect on the recovery of the different metabolites. Actually, the MSTFA amount was the critical factor determining the maximum response and completeness of the reaction as discussed below. The shape of the equilibrium curves for the different metabolites, however, was varying with compound class, i.e. the equilibrium constant (slope of the curves) of the silylation reaction differed per metabolite class (Figure 1). Class-3 metabolites, i.e. glutamic acid-d3 and nicotinamide, and the amino acids (class 2), i.e.
phenylalanine-d5 and leucine-d3, needed the largest excess of MSTFA to reach their maximum response. For the sugar-phosphates (class 2) and citric acid (class 1) an intermediate amount of MSTFA was needed, and the sugars (class 1) reached their maximum response at a relatively low amount of MSTFA (40 µL). Additionally the variation of the MS response for the metabolites measured under the different experimental conditions tested (Table 3) varied with compound class (Figure 1). For class-1 metabolites, i.e. fructose, typically small variations (RSD < 6%) of MS response were obtained over the entire range of MSTFA concentrations (40 – 400 µL).
For class-2 metabolites, i.e. phenylalanine-d5, the variation at low levels of MSTFA (<
150 µL) was larger, however at higher MSTFA concentrations variations of the MS response were small (RSD < 7%). For class-3 metabolites, i.e. glutamic acid-d3, the variation in MS response over the entire range of MSTFA concentrations was relatively large. However at MSTFA concentrations greater than 150 µL, RSDs of 11%
or better were obtained.
In summary, at an amount larger than 150 µL of MSTFA, equivalent to approximately a molar excess of 15 times (including the derivatization of the oximation reagent ethoxyamine), the recovery and repeatability of MS response were satisfactory for all classes of metabolites. We use 100 µL in the standard derivatization procedure, corresponding with an amount of 340 µL in the silylation reaction curves.
Figure 1 Silylation reaction curve and variation of MS response of metabolites in academic standard from different compound classes; class 1, fructose (A); class 2, phenylalanine-d5 (B); class 3, glutamic acid-d3 (C).
The recovery of the metabolites was also significantly influenced by the inertness of the analytical system. The curves of MS response versus MSTFA amount for different metabolites (academic standards) were measured on the same day on two different analytical systems, (i) with an ‘active’ column, i.e. using an old column with active sites resulting in tailing, low recoveries and high RSDs for derivatized metabolites, and (ii) with an ‘inert’ column, i.e. using a relatively new column with few active sites (Figure 2). For class-3 metabolites, such as glutamic acid, the peak areas on the ‘active’
column were much lower than on the ‘inert’ column over the entire range of MSTFA concentration. For class-1 metabolites, i.e. citric acid, this effect was also present, but only at low concentrations of MSTFA. The peak areas for class-2 metabolites were also influenced by the ‘active’ column; however, to a lesser extent than class-3 metabolites (data not shown). The decrease in peak areas was probably caused by an increase in adsorption and degradation due to active sites present on the surface of the
‘active’ column.
Cholic acid-d5 was classified in class 1 on the basis of the functional groups present in the metabolite. Although the shape of the curve was similar to other class-1 compounds, such as citric acid, the variation in MS response resembled the variations obtained for class-2 metabolites and at low levels of MSTFA (150 µL) the variations even resembled the variation in response found for class-3 metabolites. Additionally, cholic acid-d4 was more prone to degrade on an ‘active’ column than other class-1 metabolites. Possibly, some functional groups in cholic acid are more difficult to silylize due to steric hindrance, causing the larger variation in MS response at lower MSTFA concentrations. Probably the enthalpy of formation (∆H) of silylated cholic acid is relatively small, explaining the limited stability on the active column compared to other class-1 compounds.
0 500000 1000000 1500000 2000000 2500000 3000000
0 200 400 600
MSTFA (ul) 0
500000 1000000 1500000 2000000 2500000 3000000
0 200 400 600
MSTFA (ul) 0
500000 1000000 1500000 2000000 2500000 3000000
0 200 400 600
MSTFA (ul)
MS response
A B C
86
Figure 2 Curves of MS response versus MSTFA amount on ‘good column’ (
○
) and ‘bad column’ (■
) for class 1: citric acid (A) and class 3: glutamic acid-d3 (B).In general, the observed shapes of the silylation reaction curves of all metabolites in academic standards and in fetal bovine serum samples were comparable. For example, the maximum responses for the different metabolites were reached at the same amounts of MSTFA. Adsorption and/or degradation effects were similar in fetal bovine serum and academic standards, i.e. class-2 and especially class-3 metabolites were more susceptible towards degradation in the analytical system than class-1 metabolites, although the effects were less pronounced in serum matrix than in academic standards (data not shown).
Stability of derivatized metabolites at -20°C
The short-term stability of the different derivatized metabolites in fetal bovine-serum matrix at -20°C was investigated by re-injecting curves 1, 4, 5, 6 and 7 after 28 days.
The curves of the different metabolites were comparable with the curves on day 1 (data not shown), only the stability was lower at low levels of MSTFA (< 100 µL) resulting in lower relative responses for most metabolites. The long-term stability of the derivatized metabolites in fetal bovine-serum matrix at -20°C using the standard derivatization protocol (Cf. Experimental) was tested by comparing the normalized MS response of spiked metabolites in fetal bovine-serum samples on day 1 and day 84 (S- Table 1). The responses of all metabolites on day 84 were between 90 and 130 % compared to the response on day 1.
Matrix enhancement effect
Due to differences in stability class-2 and class-3 metabolites are more prone to adsorb or degrade on the surface of the liner or on the analytical column, resulting in lower recoveries of these metabolites. In some matrices, e.g. plasma or serum, these effects are much less pronounced than in academic standards, probably due to the shielding of active sites by components from the matrix. Several experiments were carried out, to examine this so-called matrix enhancement effect and to investigate which matrix
0 500000 1000000 1500000 2000000 2500000 3000000
0 100 200 300 400 500
MSTFA (µl)
MS response
0 500000 1000000 1500000 2000000 2500000 3000000 3500000
0 100 200 300 400 500
MSTFA (µl)
MS response
B
A
compounds (or compound classes) are responsible for the decrease in adsorption and degradation effects in metabolomics analysis.
In a first experiment several silylated compounds, e.g. trimethylsilylimidazole, 4- bromo-N,N-bis(trimethylsilyl)-aniline, N-benzyl-trimethylsilylamine and bis- (trimethylsilyl)-urea, were screened on their ability to protect the test metabolites.
These compounds were chosen as they were expected to be good trimethylsilyl (TMS) donors, able to prevent degradation of other metabolites by donating a TMS group to a degraded metabolite, or to derivatize impurities in the system. From these compounds bis-(trimethylsilyl)-urea gave the most promising results (highest recovery and lowest RSDs of the test metabolites, data not shown). Urea is also one of the most abundant compounds present in extracts of plasma and serum samples and could be one of the compounds explaining the relatively good performance of the method in class-1 matrices. At a concentration of 2.5 mg bis-(trimethylsilyl)-urea per 100 µL the highest recoveries and lowest RSDs were obtained in academic standards (S-Table 2), this was approximately a 50-fold higher molar amount than the total amount of all test compounds. The matrix enhancement effect was significant, especially for the class-2 and class-3 metabolites and cholic acid. The results for class-1 metabolites were unaffected; these metabolites were already maximally recovered in the academic standard without matrix.
Table 5 Relative responses and repeatability (RSD, n=5) of the test metabolites using different protective compounds added to an academic mixture a) (cf. Experimental – Matrix enhancement effect – experiment 1)
Protective compound 3one Urea Glucose Diaminododecane
Compound Name Relative
response RSD
(%)
Relative response
RSD (%)
Relative response
RSD (%)
Relative response
RSD (%) Class 1
Citric acid 100 5 106 6 95 5 100 4
Fructose 100 4 105 3 n.d. b) n.d. 108 2
Glucose-d7 100 6 105 3 n.d. b) n.d. 108 1
Cholic acid-d4 100 19 118 9 125 3 138 2
Class 2
Leucine-d3 100 3 110 4 107 3 110 2
Phenylalanine-d3 100 18 119 6 143 4 130 3
Ribose-5-phosphate 100 10 115 2 105 2 147 7
Glucose-6-phosphate 100 22 113 4 130 3 112 32
Class 3
Nicotinamide 100 8 120 7 92 20 79 24
Glutamic acid-d3 100 24 128 6 144 6 134 8
a) Relative response was calculated by comparing the average peak areas of the metabolites with the average peak areas in the academic standard after correction for the internal standard.
b) n.d.: not determined. Due to the high concentration of glucose these metabolites could not be quantified.
88
Subsequently, experiments were performed with silylized compounds from different compound classes with different volatilities to compare their protective abilities and the influence of their elution temperature. The studied compounds included the amides octafluorohexanediamide and bis-(trimethylsilyl)-urea, the sugars deoxyribose, glucose and melibiose and the amine 1,12-diaminododecane (cf. Experimental – Matrix enhancement effect – Experiment 1). The results from bis-(trimethylsilyl)-urea (class 3), glucose (class 1) and 1,12-diaminododecane (class 2) are shown in Table 5 using a normal, inert GC-MS system. All results on recoveries and RSDs are shown in the supplement (S-Table 3, S-Table 4). All ‘protective’ compounds enhanced the response for some metabolites compared to the academic standard without an additive. The addition of bis-(trimethylsilyl)-urea and octafluorohexanediamine provided the best overall results for all tested compounds with the highest recoveries and the lowest RSDs of the peak areas. In general, the protective abilities increased with the instability of the protective compound. Amides (bis-(trimethylsilyl)-urea, octafluorohexaneamide) and amines (diaminododecane) that have the weakest bond with TMS were the best protectors. This effect is even more pronounced when the standards with and without the protective compounds are injected on an analytical system with an ‘active’
analytical column, i.e. an old column with active sites resulting in tailing of peaks, low recoveries and high RSDs for critical derivatized metabolites (see S-Tables 5 and 6).
For this situation the better performance upon addition of the amides and diaminododecane compared to the other compounds is evident. No significant influence of the elution temperature of the protective compounds on the improvement of the performance of the metabolites with GC-MS could be detected. In conclusion, addition of bis-(trimethylsilyl)-urea and 1,12-diaminododecane appeared to be most promising. In addition, nicotinamide appeared to be the most critical metabolite from the evaluation mixture and is a good marker for the early detection of a decrease in system performance.
Next, the responses of the test metabolites from the evaluation mix spiked to fetal bovine-serum matrix were compared with the responses of the metabolites in academic standards containing bis-(trimethylsilyl)-urea, 1,12-diaminododecane or a combination of bis-(trimethylsilyl)-urea and 1,12-diaminododecane. Additionally bis- (trimethylsilyl)-urea and 1,12-diaminododecane were added to fetal bovine-serum matrix to investigate whether the response for the spiked metabolites could be increased compared to normal serum matrix (for a summary, see Table 6; for all results see S-Tables 7 and 8). The responses of the test metabolites in the academic standard with the combination of bis-(trimethylsilyl)-urea and 1,12-diaminododecane were comparable with the results in serum matrix, except for nicotinamide and glutamic acid, but for these the addition of the protective compounds still improved the relative response compared to no addition. In other words, a class-1 matrix could be mimicked to a large extent by the addition of bis-(trimethylsilyl)-urea and 1,12-diaminododecane
to an academic standard. The responses for the test metabolites in fetal bovine-serum matrix were not enhanced further by adding additional protective compounds, indicating that the metabolites were already at their maximum response in fetal bovine- serum (S-Table 7). As discussed earlier (cf. Introduction) serum is considered to be a class-1 sample matrix, i.e. typically the performance of the GC-MS method in serum is very satisfactory and most metabolites can be measured with high presision in this matrix. Most probably the protective abilities of class 1 sample matrices are sufficient to obtain maximum recovery for most metabolites and therefore the addition of more
‘protective’ compounds is unnecessary.
Table 6 Relative response of the test metabolites in academic standards and fetal bovine serum matrix with different protective compounds compared to the response in serum without added compounds a)
Matrix Serum 3o Urea and diaminododecane
Compound Name Relative
response a) RSD (%) Relative
response a) RSD (%) Relative
response a) RSD (%) Class 1
Citric acid b) n.d. 2 100 2 100 5
Fructose b) n.d. 2 100 1 103 2
Glucose-d7 c) n.d. 1 100 1 105 1
Cholic acid-d4 100 2 73 9 107 3
Class 2
Leucine-d3 100 3 70 10 91 9
Phenylalanine-d5 100 5 74 13 98 5
Ribose-5-phosphate 100 1 76 7 107 2
Glucose-6-phosphate 100 2 70 8 98 4
Class 3
Nicotinamide 100 10 49 35 77 1
Glutamic acid-d3 100 2 40 13 60 1
a) First the average normalized response was calculated by correction of the peak areas with that of the IS (dicyclohexylphthalate); next the average normalized response per metabolite was divided by that obtained in the serum matrix and multiplied with 100; only for citric acid, fructose and glucose-d7 the normalized response was divided by that obtained in the academic standard.
b) n.d.: not determined. These metabolites are present in the fetal bovine serum matrix. Therefore the relative response compared to the response in the academic standard (no matrix) was not determined.
c) n.d.: not determined. Due to the high concentration of glucose in the fetal bovine-serum matrix, glucose-d7 could not be determined accurately.
The results shown above show that the method performance can be improved by the addition of protective compounds. This hypothesis was tested for a ‘worst-case’ matrix.
For this purpose an artificial class 3 matrix was prepared and the effect of the addition of bis-(trimethylsilyl)-urea to this matrix on the recovery of the metabolites from the evaluation mix was assessed (cf. Experimental – Matrix-enhancement effect – Experiment 3). In our experience the presence of large amounts of sulfate in extracts
90
has significant adverse effects on the recovery and RSDs of all metabolites in microbial samples. Bis-(trimethylsilyl)-urea was added to the evaluation mix with a large amount of ammoniumsulfate and the responses for the test metabolites were compared with the response in an academic standard (Table 7). It was found that the addition of bis- (trimethylsilyl)-urea significantly improved the recovery of most metabolites, compared to the recovery in samples containing ammoniumsulfate without bis- (trimethylsilyl)-urea as protective reagent (Table 7). Compounds eluting close after sulfate, such as leucine-d3, glutamic acid-d3, were still influenced despite the addition of urea.
Table 7 Influence of sulfate on the relative response of metabolites in academic standards a) (cf.
Experimental – Matrix enhancement effect – Experiment 3)
Matrix 3o Sulfate Sulfate and urea
Compound Name b) (compound class)
Leucine-d3 (2) 100 0 0
Glutamic acid-d3 (3) 100 0 0
Phenylalanine-d5 (2) 100 0 64
Citric acid (1) 100 84 104
Fructose (1) 100 99 102
Glucose-d7 (1) 100 98 103
Ribose-5-phosphate (2) 100 51 124
Glucose-6-phosphate (2) 100 3 83
Cholic acid-d4 (1) 100 5 94
a) Relative response was calculated by comparing the average peak areas of the metabolites with the average peak areas in the academic standard after correction for the internal standard
(dicyclohexylphthalate).
b) Compounds are sorted according to their elution order.
Discussion
In this chapter, the differences in method performance for metabolites were investigated for different performance classes (cf. Introduction, Table 1). Besides compound class (reactivity, stability), the method performance was largely dependent on the sample matrix (e.g., plasma, serum, tissue, microbial) and the inertness of the analytical system.
The reactivity towards silylation and stability of the derivatized compounds from the different performance classes were investigated by the construction of silylation reaction curves. The results were consistent with the theory on silylation reactivity and stability of different functional groups (cf. Introduction). In general, the excess of MSTFA needed to reach the maximum peak areas for the metabolites and the RSDs of the peak areas was the lowest for class 1, higher for class 2 and highest for class 3.
Additionally, the stability of the derivatized metabolites decreased in the order: class 1
> class 2 > class 3. It should be noted that the least-reactive and least-stable functional group present in the metabolite determined their classification to the different performance classes.
Due to the differences in stability, class-2 and, especially, class-3 metabolites were much more influenced by the existence of active sites in the analytical system and the inertness of materials used during sample work-up than class-1 metabolites.
In the matrix-enhancement effect, i.e. the increased response for metabolites compared to academic standards when sample matrix is present, all three factors, i.e. compound class, sample-matrix class and inertness of the analytical system, played an important role. In general, only the recoveries and RSDs of class-2 and class-3 metabolites were significantly improved in the presence of matrix or by the addition of protective compounds, due to a decrease in adsorption and degradation in the analytical system.
Class-1 metabolites are less prone to degradation and therefore the performance of these metabolites is already optimal without matrix. However, not all sample matrices improve the performance for critical metabolites. Whereas class-1 matrices (cf.
Introduction), such as plasma, possess excellent protective abilities and significantly improve the method performance for critical metabolites, class-3 matrices can even decrease the performance for critical metabolites compared to academic standards. In our attempts to mimic the matrix enhancement effect of class-1 matrices, it was observed that especially class-3 type metabolites were good protectors, when added in large quantities to academic standards. As class-3 metabolites are prone to degrade and/or react on active sites in the analytical column and are in fact good TMS donors, these metabolites were capable of preventing adsorption and degradation of other metabolites. By adding a combination of 1,12-diaminododecane (class 2) and bis(trimethylsilyl)-urea (class 3) (~ 50-fault higher molar amount compared to the amount of test metabolites) to academic standards, the responses for the majority of the evaluated metabolites in the standard were comparable with the response in a serum
92
sample (class-1 matrix) (Table 6). Also the responses for class-2 metabolites could be significantly increased in an artificial class-3 matrix (academic standard containing ammoniumsulfate) (Table 7). In Table 8 the performances for the different metabolite classes in different matrices are summarized. The grey areas indicate the matrix and metabolite combinations, for which the performance could possibly be improved by the addition of extra protective compounds (urea and/or 1,12-diaminododecane).
Table 8 Performance of GC-MS for different matrix and metabolite classes a)
Compound Matrix b)
Class 1 Class 2 Class 3
Class 1 + + + + +
Class 2 + + + + –
Class 3 + + + – –
a) + +: good, +: reasonable, + -: critical, -: very critical; the grey areas indicate the matrix and metabolite combinations that possibly benefit from the addition of extra protective compounds.
b) Matrix classes as described in the introduction, i.e. class 1: plasma, serum class 2: tissues, plants, urine, class 3: microbial samples
In summary, GC-MS based metabolomics analysis using silylation is not as straightforward as it may seem. Silylation can be applied to convert many functional groups that pose a problem in GC-MS analysis, extending the application range of the method to a large fraction of the small polar metabolites present in biological samples.
However, the precision of GC-MS data for all these metabolites has to be carefully monitored, because different functional groups have different reaction behavior towards silylation and different stabilities. In addition, the stability of derivatized metabolites can vary with the sample matrix.
To obtain reliable GC-MS data for metabolomics analysis, a good protocol should involve an elaborate optimization and validation prior to the implementation of a method (Chapter 2). An extensive set of metabolites, from all compound classes and with different functional groups, should be chosen to assess the performance of the GC-MS method. Especially compounds from class 2 (amino acids) and class 3 should be well represented as these compounds are less reactive towards silylation and most likely to adsorb and/or degrade during sample work-up and analysis. Additionally, the method should be validated for every new matrix to assess the performance for all metabolite classes in that specific sample matrix.
While it is true that method performance is varying with compound class and matrix, still some general rules can be formulated to obtain the best possible results for metabolomics analysis.
(1) The inertness of the analytical system and all materials used during sample workup is of utmost importance (autosampler vials, deactivated injection liners, clean analytical column).
(2) All possibilities of moist or other disturbances entering the sample should be avoided after derivatization. For example, vials should be recapped after analysis, when extract must be kept over longer periods of time, they should be stored at - 20°C to prevent breakdown of metabolites and transferring to another autosampler vial or dilution of the sample prior to analysis should be avoided, as this can result in irreproducible data.
(3) A sufficient amount of silylation reagent should be used during derivatization, an excess of over 15 times was needed to maintain the stability of metabolites during analysis and storage in the freezer (-20°C).
(4) Representative quality-control standards (preferably isotopically labeled metabolites) of all performance classes of interest should be added prior to extraction and prior to derivatization to control the method performance for every sample. For example, standards can be added to check the completeness of the oximation (e.g. glucose-d7) and silylation reaction (e.g. phenylalanine-d5) or the inertness of the analytical system (e.g. nicotinamide, glutamic acid-d3, cholic acid- d4).
(5) Furthermore, a set of representative internal standards (preferably isotopically labeled metabolites) should be added prior to extraction, to eventually correct for disturbances during sample workup or analysis. At least one exogenous standard should be added to correct for small variations in injection volume and MS response.
(6) The use of extra ‘protective’ compounds, such as bis-(trimethylsilyl)-urea and/or 1,12-diaminododecane, can assist in the improvement of the method performance for critical metabolites in critical matrices (Table 8).
94
CONCLUSIONS
In this chapter, critical factors for the performance of a metabolomics method based on a silylation prior to GC-MS analysis was investigated. The most-critical factors were the compound class of the metabolite, the sample matrix and the inertness of the analytical system. The differences in performance between different compound classes could be explained by theoretical considerations of the thermodynamic and kinetic properties of the silylation reaction. Based on an extensive study, some recommendations could be given to improve the repeatability of analysis, especially for critical metabolites. It could be demonstrated that the addition of protective compounds significantly improved the recovery and RSDs for many evaluated metabolites in critical matrices. Actually, by following the guidelines presented in this paper, many sample series of class-1, class-2 and class-3 matrices were successfully analyzed and relevant biological information could be obtained.
Reference List
1. Koek, M. M.; Muilwijk, B.; vanderWerf, M. J.; Hankemeier, T. Anal. Chem.
2006, 78, 1272-81.
2. Koek, M. M.; Muilwijk, B.; van Stee, L. L. P.; Hankemeier, T. J. Chromatogr. A 2008, 1186, 420-29.
3. Fiehn, O.; Kopka, J.; Trethewey, R. N.; Willmitzer, L. Anal. Chem. 2000, 72, 3573-80.
4. Jonsson, N.; Gullberg, J.; Nordstrom, A.; Kusano, M.; Kowalczyk, M.;
Sjostrom, M.; Moritz, T. Anal. Chem. 2004, 76, 1738-45.
5. Pierce, K. M.; Hope, J. L.; Hoggard, J. C.; Synovec, R. E. Talanta 2006, 70, 797-804.
6. Roessner, U.; Wagner, C.; Kopka, J.; Trethewey, R. N.; Willmitzer, L. Plant J.
2000, 23, 131-42.
7. Shellie, R. A.; Fiehn, O.; Welthagen, W.; Zimmermann, R.; Zrostlikovβ, J.;
Spranger, J.; Ristow, M. J. Chromatogr. A 2005, 1086, 83-90.
8. Strelkov, S.; Elstermann, M. v.; Schomburg, D. Biol. Chem. 2004, 385, 853-61.
9. Villas-Bôas,J.M; Jesper Højer-Pedersen, J.; Kesson, M.; Smedsgaard, J.;
Nielsen, J. Yeast 2005, 22, 1155-69.
10. O'Hagan, S.; Dunn, W. B.; Knowles, J. D.; Broadhurst, D.; Williams, R.;
Ashworth, J. J.; Cameron, M.; Kell, D. B. Anal. Chem. 2007, 79, 464-76.
11. Anastassiades, M.; Mastovska, K.; Lehotay, S. J. J. Chromatogr. A 2003, 1015, 163-84.
12. Cajka, T.; Mastovska, K.; Lehotay, S. J.; Hajslova, J. J. Sep. Sci. 2005, 28, 1048-60.
13. Erney, D. R.; Gillespie, A. M.; Gilvydis, D. M.; Poole, C. F. J. Chromatogr. A 1993, 638, 57-63.
14. Mastovska, K.; Lehotay, S. J.; Anastassiades, M. Anal.Chem. 2005, 77, 8129-37.
15. Pierce, A. E. Silylation of organic compounds, Pierce Chemical Co.: 1968.
96
SUPPLEMENTAL
S-Figure 1 Silylation reaction curves of phenylalanine-d5 under different experimental conditions (for a description of the reaction conditions for the different curves see Table 3, p.83).
S-Table 1 Stability of derivatized metabolites in fetal bovine serum after 84 days a) Relative response (%)
Compound name day 1 day 84
Class 1
Citric acid 100 131
Fructose 100 89
Cholic acid-d4 100 108
Class 2
Leucine-d3 100 102
Phenylalanine-d5 100 109
Ribose-5-phosphate 100 110
Glucose-6-phosphate 100 112
Class 3
Glutamic acid-d3 100 103
a) Response of the metabolites was normalized on dicyclohexylphthalate (IS) and subsequently the response on day 1 was set on 100%; Glucose-d7 could not be determined accurately due to the large amounts of endogenous glucose in the serum sample.
S-Table 2 Relative responses a) of the test metabolites using different amounts of bis(trimethylsilyl)urea on an ‘active’ b) analytical column
Compound name Relative response (%) Relative standard deviations (%) Added amount urea no 1 mg /
100 µL
2.5 mg /
100 µL no 1 mg /
100 µL
2.5 mg / 100 µL Class 1
Citric acid 100 99 104 6 8 3
Fructose 100 93 95 6 9 4
Glucose-d7 100 94 95 6 9 4
Cholic acid-d4 100 807 1000 267 19 15
Class 2
Leucine-d3 100 105 122 15 12 4
Phenylalanine-d5 100 178 237 33 14 8
Ribose-5-phosphate 100 166 241 26 16 11
Glucose-6-phosphate 100 207 326 38 16 12
Class 3
Glutamic acid-d3 100 188 263 29 13 9
a) Relative response was calculated by comparing the average peak areas of the metabolites with the average peak areas in an academic standard (no additions) after correction for the internal standard (dicyclohexylphthalate).
b) Active column: an old column with active sites resulting in tailing of peaks, low recoveries and high RSDs for critical derivatized metabolites.
