Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2015 Jun 1.
Published in final edited form as: Anal Biochem. 2014 Mar 18;454:23–32. doi: 10.1016/j.ab.2014.01.020

Identification of Metabolites from LC-EC Profiling: GC-MS and Re-Fractionation Provide Essential Information Orthogonal to LCMS/microNMR

Rose M Gathungu 1,2, Susan S Bird 1, Diane P Sheldon 1, Roger Kautz 2, Paul Vouros 2, Wayne R Matson 3, Bruce S Kristal 1
PMCID: PMC4034759  NIHMSID: NIHMS578303  PMID: 24657819

Abstract

HPLC-coulometric electrode-array detection (LC-EC) is a sensitive, quantitative and robust metabolomics profiling tool that complements the commonly used MS and NMR-based approaches. However, LC-EC provides little structural information. We recently demonstrated a workflow for the structural characterization of metabolites detected by LC-EC profiling, combined with LC-ESI-MS and microNMR. This methodology is now extended to include: (i) GC-EI-MS analysis to fill structural gaps left by LC-ESI-MS and NMR, and (ii) secondary fractionation of LC-collected fractions containing multiple co-eluting analytes. GC-EI-MS spectra have more informative fragment ions that are reproducible for database searches. Secondary fractionation provides enhanced metabolite characterization by reducing spectral overlap in NMR and ion-suppression in LC-ESI-MS. The need for these additional methods in the analysis of the broad chemical classes and concentration ranges found in plasma is illustrated with discussion of four specific examples, including: (i) characterization of compounds for which one or more of the detectors is insensitive (e.g., positional isomers in LC-MS, the direct detection of carboxylic groups and sulfonic groups in 1H NMR, or non-volatile species in GC-MS).; (ii) detection of labile compounds, (iii) resolution of closely eluting and/or co-eluting compounds and, (iv) the capability to harness structural similarities common in many biologically-related, LC-EC detectable compounds.

Keywords: Metabolite characterization, Secondary Fractionation, LC-EC, LC-MS, NMR, GC-MS

Introduction

The objective of our liquid chromatography - Coulometric array detection (LC-EC) metabolomic profiling studies is to identify redox active metabolites that vary relative to a biological state e.g., diet, disease [1-8]. LC-EC is sensitive (with limits of detection at the femtomole level), precise, and robust -- all strengths for metabolic profiling studies [6, 9-12]. Because LC-EC identifies metabolites only by retention time and oxidation potentials, structural annotation of metabolites of interest requires follow-up with additional analytical approaches, such as mass spectrometry (MS) and nuclear magnetic resonance (NMR) [13-15]. Developing an approach for structural annotation of LC-EC-detected metabolites requires consideration of: (i) the physico-chemical diversity (e.g., polarity, boiling point, etc) of the metabolome (thus no single analytical technique provides complete structural characterization of all species) [16]; (ii) many LC-EC detectable metabolites are labile (easily oxidized and/or reduced), (iii) metabolites vary over a wide concentration range (down to picomolar), thus requiring subsequent analytical platforms to be sensitive in addition to structurally informative, and (iv) the use of buffer ingredients such as pentane sulfonic acid (PSA) in LC-EC to maximize sensitivity and for reproducibility, but they can interfere with MS and NMR analysis [1, 10, 12, 15].

We recently developed a platform that enabled characterization of LC-EC detectable metabolites by combining the strengths of LC-MS, NMR, and liquid chromatography-ultraviolet detection (LC-UV) for metabolite fractionation and pre-concentration. This platform had been designed to structurally annotate metabolites that had been identified as biologically significant using LC-EC profiling and statistical analysis [13]. The combined platform enabled the mapping of metabolites between LC-EC, LC-MS and NMR and succeeded in structurally characterizing high-concentration metabolites that are readily detectable by LC-MS and NMR, but it falters when: (i) addressing trace-level metabolites; (ii) the metabolite is not readily detectable by LC-MS and/or NMR, or; (iii) the LC-UV fractionation produces a mixture of metabolites. These cases suggested that a complementary analytical method and possibly a secondary fractionation step would be necessary for a more complete characterization of LC-EC identified species in the plasma metabolome.

LC-MS analysis of trace-level unknowns is limited to ionizable compounds; LC-MS also requires authentic standards to unambiguously identify analytes, but such standards are not always available [17-19]. LC-MS uses electrospray ionization-mass spectrometry (ESI-MS), a soft-ionization technique that produces limited structural information due to limited fragmentation and, is often plagued with ion suppression, especially between closely eluting and co-eluting compounds. Although the use of different column chemistries in LC (e.g. reversed-phase vs. HILIC) can often resolve co-eluting metabolites, structural information of analytes is still limited by ESI ionization [20, 21]. Ion suppression can lead to non-detection of a metabolite or to a reduction of an ion's abundance, thus inhibiting the acquisition of structurally informative MS/MS data [19]. Moreover, LC-ESI-MS can result in non-specific metabolite fragmentation from the loss of common functional groups (e.g., –CO2, -NH2, and H2O), and in varied fragmentation patterns of the same molecule, depending on the instrument and/or the mobile phase composition used for analysis. In comparison, gas chromatography-mass spectrometry (GC-MS) uses electron ionization (EI) or chemical ionization (CI) to analyze volatile metabolites. The more commonly used EI-MS, unlike ESI-MS, does not suffer from ion suppression. EI-MS data complements ESI-MS data because EI-MS is a hard ionization technique that generates extensive and predictable fragmentation for structural characterization of metabolites of interest. EI mass spectra are well-defined, are consistent across instruments and laboratories, and are searchable against well-annotated databases [22-24]. GC-EI-MS retention indices and fragmentation patterns can also be leveraged to selectively recognize metabolites that share common structural features, thus facilitating characterization of unknowns [19, 22-24]. A weakness of EI-MS is its inability to reproducibly detect the intact molecule, which can be compensated by parallel use of LC-ESI-MS [24].

Mixtures of metabolites complicate both LC-MS and NMR analysis because signals from high concentration metabolites can mask signals from low concentration metabolites (e.g., by ion suppression, spectral overlap).[21, 25, 26] In our previous work, a single reversed-phase LC fractionation method produced samples that were reduced in complexity relative to the initial biofluid (sera/plasma), but most fractions still contained multiple compounds. Additional sample purification prior to LC-MS and NMR analysis should enable improved detection/characterization of closely/co-eluting metabolites following initial fractionation.

In line with the above considerations, the current study expands our initial platform to include GC-MS and secondary fractionation to deal with the challenge of the complexity of the plasma samples, specific characterization challenges of certain compounds, and the high dynamic range of the metabolites following profiling. In addition, GC-MS was added to our workflow to characterize EC-detectable metabolites with common structural features. Due to the predictability of EI-MS spectra, metabolites with common structural features share diagnostic fragment ions that can be used facilitate their characterization. While the specific individual analytical techniques have been used previously in metabolomics profiling and metabolite identification, the utility of GC-MS to characterize EC-detectable metabolites was not apparent in advance, as the obvious transition of metabolite identification from LC-EC is typically LC-MS. Indeed, to our knowledge, the work presented in this manuscript, represents the first time GC-MS was successfully used to structurally-identify EC-detected peaks of interest. Similarly, for the characterization of closely eluting and co-eluting metabolites, secondary fractionation using a time-slicing approach of an LC-chromatogram into smaller time-points allowed for the collection of purer metabolites without having to resort to changing the LC column chemistry, which would have made it difficult to correlate the characterized metabolites across the different platform. Furthermore, since all analytical techniques (i.e. LC-EC, LC-MS, GC-MS and NMR) were used offline from each other, fractionation served as a link to ease the co-ordination of data across the different detectors.

Materials and Methods

Chemicals

LC-MS grade acetonitrile (ACN), methanol (MeOH), and isopropyl alcohol (IPA), triethylamine, ammonium acetate, acetic acid, and N,O-bis(trimethylsilyl)trifluoroacetamide with 1% trimethylchlorosilane (BSTFA/TMCS) were purchased from FisherScientific (Pittsburgh, PA). All deuterated solvents were purchased from Cambridge Isotopes (Andover, MA). All standards were purchased from Sigma-Aldrich (St. Louis, MO). Human pooled plasma (from both males and females) was purchased from Interstate Blood Bank, INC (Memphis, TN)

Sample Preparation

Metabolite extraction from plasma and the initial fractionation protocols were used as previously described [13]. In brief: Metabolites were extracted with cold acidified acetonitrile from 2.5 mL of human plasma and the supernatant dried under vacuum. For LC fractionation, the metabolites were reconstituted in 200 μl of the starting mobile phase. Ninety-six fractions were then collected and dried down under vacuum for subsequent MS and NMR analysis. More details in the Supplemental Information.

Fractions that were found to contain multiple metabolites based on LC-MS, GC-MS and 1H NMR analysis were re-fractionated and re-analyzed. The conditions (i.e., gradient, columns, mobile phase) used for re-fractionation were identical to those used in the initial LC-UV fractionation, but the eluent was time-sliced into 100 μl fractions (vs. 1 ml) in the region of the closely eluting and co-eluting metabolites. The collected fractions were then dried under vacuum for subsequent analysis.

Analytical Protocols for Metabolite Identification

The LC-MS, LC-EC and 1H NMR experiments and instrument conditions for metabolite identification were used as previously described [13]. Briefly, fractions for LC - high resolution MS and LC-EC analysis reconstituted in either 100 μl (LC-MS) or 50 μl (LC-EC) of mobile phase A. LC separations was performed using gradient elution under reversed phase conditions (Details in the Supplemental Information). The injection volume for LC-MS was 80 μl while that for LC-EC was 50 μl. For 1H NMR analysis, each fraction was reconstituted in 10 μl deuterated DMSO (DMSO-d6), with 8 μl loaded into a microcoil NMR probe.

GC-MS analysis

After evaporation of the LC solvent, each fraction was reacted with 40 μl of BSTFA/TMCS at 70°C (1 hour). Samples were then stored at -80 °C awaiting GC-MS analysis. Each derivatized fraction was analyzed on a HP6890 series GC coupled to a single quadrupole HP5973 mass selective detector and on a HP-5MS GC capillary column (30m×0.25mm×0.25μm, 5% phenyl-siloxane) with helium as a carrier gas at a flow rate of 1 mL/min. The injector was set to 250°C; 5 μl injections of each sample were made in the split mode (1:10). Separation used a temperature gradient: the initial oven temperature (130°C) was held for five minutes, ramped up at 10°C/minute to a final temperature of 300°C where it was held for ten minutes. For MS analysis, the source temperature was set to 230°C, while the quadrupole temperature was set to 150°C. After a two minute solvent delay, full-scan EI spectra (electron energy =70 eV) were acquired across the mass range of m/z 50-800. Spectra were background subtracted and searched against the NIST database (NIST08.L).

Data Analysis for Metabolite Identification

The initial structural annotation of each metabolite was based on database searches of each unique exact mass (both positive and negative ions) against the METLIN [27] and HMDB [28] databases using a mass tolerance of 5 ppm.

Results and Discussion

Our long-range goal is the structural characterization of biologically-relevant, electrochemically-active metabolites following LC-EC profiling and statistical analysis [3, 4, 7, 8]. The present study extends our previous structural identification platform [13], by utilizing the synergistic advantages of multiple analytically diverse platforms (i.e., LC-EC, LC-MS, 1H-NMR and GC-MS) (Figure 1). The four examples presented below highlight solutions to the different challenges encountered in metabolite characterization, including: (i) metabolites with structural features that are only detectable in certain detectors, therefore requiring the combination of results from all detectors for their full characterization, (ii) metabolites that cannot be isolated as individual compounds with a single LC-fractionation step, and therefore require secondary re-fractionation to purify them, (ii) low concentration metabolites that are detected in the LC-EC and MS platforms, and (iv) metabolites with similar structural features that are selectively identifiable in a particular analytical platform, facilitating their structural annotation.

Figure 1.

Figure 1

Open in a new tab

General flow-chart of the strategy for the structural characterization of LC-EC-detected plasma metabolites.

Prior to structural characterization, metabolites were concentrated and extracted from plasma, then separated and fractionated (Figure 1). The fractionation step which was necessary to concentrate metabolites, allowed us to work within the limits of detection of the different detectors while reducing the complexity of the plasma pool. For example, MS is about 10x less sensitive than LC-EC while NMR is about 100x less sensitive than LC-EC. Furthermore, fractionation prior to analysis with each detector, served to ensure that the metabolite identified during LC-EC profiling was the same one identified in subsequent analysis (i.e. LC-MS, NMR and GC-MS). In order to obtain a sufficient metabolite concentration, it required the use of large volumes of a commercially available human plasma pool that was determined to contain all the metabolites of interest. We note that the need for structural identification can arise in two very distinct situations during profiling studies. For peaks of interest that are consistently present in a study, e.g., endogenous metabolites, we can work with pooled samples and create fractions that have this peak enriched and isolated, and we then use aliquots of this fraction on the different structurally informative platforms. For peaks of interest that are not consistently present, we can create pools from plasma samples that contain the peak of interest.

All LC isolated fractions were first analyzed by LC-with high resolution MS. Initial structural annotation of the metabolite(s) in each fraction was based on a database search of each unique m/z ion, and enabled the provisional assignment of one or more molecular formulae to each analyte. Because database searches often yield several possible matches, database filtration for structural assignment of metabolites was based on a comparison of the top hits with HCD fragmentation results, GC-EI-MS, and 1H NMR data. In all the cases discussed below, database hits led to initial hypotheses as to the identity of the molecules of interest. Had the database search not yielded any logical metabolite hits, the seven rules developed by Kind and Fiehn would have been used to calculate probable elemental formulae of metabolites from exact mass data [29], followed by the use MS/MS, EI-MS fragment patterns and 1H NMR data for detailed structural information.

Because of the high sample mass requirements of NMR, samples for NMR were isolated by pooling the collected fractions from three separate injections. Although this increased the LC-fractionation time 3-fold, this made the analytes 30-fold more concentrated than those used for LC-MS thus reducing the NMR analysis time. After NMR analysis, the analytes were recovered and analyzed by GC-MS. Those metabolites that could not be isolated in sufficient quantity for NMR analysis were directly isolated for GC-MS analysis. Before GC-MS analysis, each fraction was first derivatized by silylation to increase its volatility and stability. All available data was then combined and unequivocal structural identification of the identified metabolites was confirmed based on comparisons of analytical results with those of authentic standards. To confirm that each structurally-identified metabolite was the one detected and identified during LC-EC profiling, each isolated fraction and authentic standard was re-analyzed by LC-EC using the LC-EC profiling conditions previously described [1, 2, 7, 11] (details in the supplemental section). This step was significant because it enabled the correlation of the metabolites detected during LC-EC profiling with the structural data that had been determined offline using MS and/or NMR.

Metabolite Characterization by combining LC-MS, NMR and GC-MS

We now show the structural characterization of a relatively high concentration EC-detected metabolite from fraction D12 as an example of the need for complementary analytical detectors. This metabolite was detected across the four detectors (LC-EC, LC-MS, GC-MS and 1H-NMR), facilitating its complete and unambiguous characterization. The negative ion MS spectrum with HCD fragmentation of fraction D12 exhibited an ion at m/z 174.0546 [M-H]- (Figure 2A). A database search of that exact mass yielded two possibilities (Figure 2A, right), indole-3-acetic-acid (I-3AA) and 5-hydroxyindole acetaldehyde (5-HIA), as having the closest exact mass values. The MS/MS results indicated that metabolite D12 contained a carboxylic acid group, based on the loss of CO2 (m/z 130.0646), thus eliminating 5-HIA. The identification of metabolite D12 as I-3AA was further supported by results from GC-MS and 1H NMR analysis. The aromatic region of the 1H-NMR spectrum (Figure 2B) of fraction D12 exhibited two ortho-coupled doublets at 7.33 ppm (J=7.8 Hz) and 7.49 ppm (J = 7.6 Hz), two triplets at 7.06 ppm (J=7.8 Hz) and 6.97 ppm (J= 7.0 Hz), a singlet at 7.2 ppm, and a highly de-shielded proton at 10.83 ppm attributed to the –NH proton on the pyrrole. These 1H NMR results were consistent with a mono-substituted indole compound with the substituent on the pyrrole moiety. These results were further supported by GC-EI-MS analysis of the fraction that had been recovered after NMR analysis. The EI spectrum of the TMS-derivative of the metabolite exhibited an ion at m/z 319 as the peak of highest mass (Figure 2C). This m/z value was consistent with the addition of two trimethylsilyl (TMS) groups to the metabolite when compared to the exact mass LC-MS measurements. The EI-MS spectrum also exhibited a fragment ion at m/z 202 [M-117]+ as the base peak, this arises from α-cleavage at the β carbon of the trimethylsilylated carbonyl ester.

Figure 2.

Figure 2

Open in a new tab

A: The high resolution MS spectrum in the negative ion mode of fraction D12 B: the aromatic region (6.5-11ppm) of the 1H NMR analysis of fraction D12. C: The GC-MS spectrum of fraction D12.

The data above shows the need to incorporate additional analytical techniques to the MS and NMR platform described in the previous manuscript. In this example, the identification of I-3AA required data from LC-MS, NMR and GC-MS that together provided structural information that could not be gained from one technique alone. LC-MS provided an exact mass that suggested the metabolite's molecular formula, and the MS/MS data indicated the presence of a carboxylic acid functional group. Moreover, a 1H NMR spectrum indicated the substitution and exact atomic position of hydrogens on the indole moiety (derived from the splitting pattern of the aromatic protons, and their chemical shift) -- information unobtainable from LC-MS. The addition of GC-MS results determined the number of acidic hydrogens (through TMS derivatization) and also overlapped with the LC-MS/MS information and established the presence of the carboxylic acid moiety. These are functionalities that are not detectable by 1H-NMR; and thus structural cross-validation was provided. The unambiguous structural assignment of metabolite D12 as 1-3AA was further established by the comparison of all analytical results with those of an authentic standard of 1-3AA.

After the structural characterization of each metabolite, the most important step is ensuring the assigned structure correlates to the LC-EC metabolite detected during profiling. After the metabolite in fraction D12 was identified as I-3AA, fraction D12 and an authentic standard of I-3AA were subjected to LC-EC analysis (Figure 3A). LC-EC analysis of authentic I-3AA exhibits the same retention time and oxidation potential as the metabolite identified from fraction D12. Furthermore, addition of authentic 1-3AA standard into fraction D12 resulted in an increase in the peak intensity of the putative I-3AA (Figure 3B). These results confirm that I-3AA was the metabolite in fraction D12.

Figure 3.

Figure 3

Open in a new tab

A: The re-analysis of fraction D12 using LC-EC to confirm that the structurally characterized metabolite is indeed the marker of interest identified during profiling. B: Co-injection of a mixture of fraction D12 and I-3AA shows co-elution confirming that metabolite D12 is I-3AA. The compounds labeled (a) and (b) are degradation products of 1-3AA.

In our second of four examples, we examine the concern that EC-detectable compounds are usually labile and can be degraded via redox processes. Metabolite degradation can lead to the assignment of degradation products as the molecules of interest or to underestimation of metabolite levels during quantitation. Fraction D12 presented a case where the LC-EC metabolite degraded. LC-EC analysis of fraction D12 showed the presence of three compounds, one of which corresponded to I-3AA (Figure 3). Two other peaks labeled (a) and (b) were degradation products of I-3AA (Figure 3B). The identification of these as degradation products was based on a stability test of an authentic standard of I-3AA, which showed that, with time, I-3AA breaks down to form these two compounds. In the current study, we did not attempt follow-up studies to determine the structural identity of these compounds, which were not observed upon LC-MS or NMR analysis (likely due to their being below the limit of detection of either method). While identification of these degradation products is not the focus of this study, given the likely structural homology to the parent compound (I-3AA), we would expect that, if a sufficient concentration of the degradation products can be obtained, e.g. by inducing degradation of a high concentration of the authentic standard, the degradation products could be structurally characterized using MS and NMR as was done with the other metabolites of interest in this study.

Resolution of closely-eluting and co-eluting metabolites

Our third example examines the challenge posed by multiple co-eluting and/or closely eluting metabolites. This scenario requires secondary re-fractionation to obtain pure (or at least purer) metabolites to acquire full structural assignment of each individual compound. During the initial LC-UV fractionation step of the plasma metabolite extract, three distinct UV peaks all eluting within the same one-minute time frame were observed in the fraction D05. The LC-MS results of this fraction (D05) also showed three unique molecular masses at m/z 195.0815 [M+H]+, 212.0019 [M-H]-, and 137.0233 [M-H]-. Based on these three exact molecular masses and their HCD fragmentations, the three compounds were putatively identified as (i) caffeine, (ii) indoxyl sulfate (ISA) (by fragment ions at m/z 132.0444 and m/z 79.9560) and (iii) a di-substituted hydroxyl benzoic acid (fragment ion at m/z 93.0332 consistent with a loss of CO2). The MS and MS/MS results of the metabolite with an exact mass of m/z 137.0233 yielded three possible positions of the hydroxyl group on the benzyl ring: 2 (ortho), 3 (meta) or, 4 (para) hydroxybenzoic acid. These positional isomers are indistinguishable by MS and MS/MS, but can be resolved by 1H NMR. To confirm these putative identifications, fraction D05 was also subsequently analyzed by 1H NMR and GC-MS.

The LC-UV and LC-MS results showed three compounds, but the 1H NMR results indicated that fraction D05 also contained a fourth compound based on comparison of the NMR spectrum of D05 and the spectra of individual authentic standards that had tentatively been assigned from the LC-MS results. The 1H NMR spectrum of D05 also indicated that the hydroxyl benzoic acid metabolite (m/z 137.0233) was 1,2 hydroxybenzoic acid (o-hydroxybenzoic acid). Comparison of the 1H NMR spectrum of fraction D05 with authentic standards of caffeine, o-hydroxybenzoic acid (salicylic acid), and indoxyl-sulfate (ISA) confirmed the presence of these three compounds in fraction D05. As previously reported, however, only five of the six protons from ISA were observed by 1H NMR, as the sixth proton overlaps with resonances of co-eluting metabolite [13]. Furthermore, resonances in addition to those expected from the caffeine, o-hydroxybenzoic acid, and ISA identified by LC-MS were also observed by 1H NMR of Fraction D05. The unassigned protons in the aromatic region of 1H NMR spectrum of fraction D05, indicated that this fourth metabolite contained a 1,2 (ortho) di-substituted benzyl ring, based on the splitting patterns (two ortho-coupled doublets and two singlets).

The presence of four metabolites in fraction D05 was further supported by the GC-MS analysis. GC-MS analysis of TMS-derivatized fraction D05, showed four well-resolved metabolites. The EI-MS spectra of the four metabolites are shown in Figure 4A-D. The EI-MS spectrum of the first compound (RT 9.01 minutes) exhibited an ion at m/z 267 as the peak of highest mass, consistent with a TMS derivative of a hydroxybenzoic acid after the loss of a methyl radical. The EI mass spectrum of the second compound (Figure 4B), which eluted at 11.79 minutes, exhibited an ion at m/z 277 as the base peak, consistent with ISA. The ion at 277 is derived from the migration of the TMS group from the sulfur to the oxygen on the indole ring, followed by a loss of SO3H. In the gas phase, TMS groups are prone to migration towards electron-rich groups like oxygen [30]. The GC-MS results of this second metabolite support the previous identification of ISA using LC-MS and 1H NMR [13]. The EI-MS spectrum of the third compound (RT 13.37) exhibited an ion at m/z 194 as the peak of highest mass, consistent with its identification as caffeine (Figure 4C).

Figure 4.

Figure 4

Open in a new tab

GC-EI-MS spectra of the four metabolites in fraction, identified as salicylic acid (A), indoxyl sulfate (B), caffeine (C) and 2-hydroxyhippuric acid (D).

The fourth metabolite, which was not observed in the LC-MS analysis, was detected at 15.69 minutes in the GC-MS. Its EI-MS spectrum exhibited an ion at m/z 339 as the peak of highest mass, as well as fragment ions at m/z 324, m/z 206 and m/z 193 (Figure 4D). A NIST database search of this EI-MS spectrum suggested that the likely candidate was a hydroxyhippuric acid, modified with two TMS groups with the hydroxyl moiety located either at the 2(ortho), 3(meta), or 4(para) positions. Hydroxyhippuric acid contains three hydrogens that can be derivatized by TMS. Partial TMS-derivatization of molecules with multiple acidic hydrogens is common in GC-MS analysis.[19, 24] The 1H NMR results indicated the most likely structure was 2-hydroxyhippuric acid (2-OH-HA) and GC-MS analysis of an authentic standard of 2-OH-HA yielded an identical retention time and EI spectrum. GC-MS results of the 2-OHHA standard also showed the completely TMS derivatized 2-OH-HA (all three acidic hydrogens derivatized) in addition to the partial derivative. These two 2-OH-HA-derivatives were observed at different GC retention times at 14.59 and 15.69 minutes (results not shown) with the completely derivatized 2-OH-HA eluting earlier (RT 14.59 minutes). The fourth metabolite was therefore identified as 2-hydroxyhippuric acid.

The results presented above confirmed that the remaining two of the four metabolites present in fraction D05 co-eluted; the next task was therefore to identify the two co-eluting metabolites. This could be accomplished by re-fractionation of fraction D05 time-sliced into shorter intervals i.e., instead of one-minute fractions (1 mL/ fraction), each minute was time-sliced into 100 μl/ fraction. To ensure that the retention order was maintained, the same LC conditions (i.e., mobile phase solvents and columns) used in the first fractionation step were maintained. While this specific example focuses on metabolites that elute within the same one-minute time frame, it is also possible to have metabolites spread across two fractions, whereby the leading and the trailing edge of a peak are collected in two different fractions. Each of the collected fractions was then analyzed by GC-MS.

The LC-UV chromatogram of re-fractionated D05 shows only three UV peaks (Figure 5A). Each isolated fraction was analyzed by GC-MS, which established that the two fractions collected from the first UV peak (annotated D05_1 in Figure 5) each contained caffeine. The second UV peak (annotated D05_2) was ISA. The third UV peak, which was split across two separate 100 μL fractions, contained SA in the first fraction (annotated D05_3-1) and a mixture of SA and 2-OH-HA in the second fraction (annotated D05_3-2, Figure 5B). The two GC peaks correspond to salicylic acid (RT 9.01 minutes) and the fully TMS-derivatized version of 2-OH-HA (RT 14.59 minutes). Further confirmation of the co-elution of these two metabolites was established by co-injection of authentic standards of these two compounds on the LC-UV, which yielded a single UV peak (results not shown). The complexity of fraction D05 necessitated its re-fractionation, which then enabled the structural resolution of the metabolites that co-eluted under LC-UV and LC-MS conditions using an orthogonal platform (GC-MS).

Figure 5.

Figure 5

Open in a new tab

A: LC re-fractionation of fraction D05. The tic marks on the UV trace indicate the time-based fractions collected for GC-MS analysis. B: GC chromatogram of the tail of peak 3, D05_3-2 (Left), and the EI-MS spectrum of metabolite eluting at 14.59 minutes (Right).

Following the structural characterization of the metabolites in fraction D05, the fraction and the respective metabolite standards was also analyzed by LC-EC using the profiling conditions to establish electrochemical activity. Caffeine ionizes well under LC-MS conditions and is also detected as a high concentration metabolite upon 1H NMR analysis (based on peak intensity). On the LC-EC, however, caffeine (at the same concentration in UV and NMR) produces a very small EC peak, indicating that it is only slightly electrochemically active (it likely requires a higher voltage). This highlights the differences in detection mechanisms and the dependence of responses of the different detectors to functional groups of the individual metabolites, thus exemplifying the need for LC-EC re-confirmation after structural characterization. It is also interesting to note that under the original LC-EC profiling separation conditions, SA and 2-OH-HA are well-resolved chromatographically (Supplemental Figure 1).

Characterization of Structurally Related Species: Indole Metabolites

Our fourth and final example for discussion is a common occurrence during metabolomic studies whereby multiple metabolites of interest have structural similarities. Metabolites that arise from the same metabolic pathway tend to share structural features, (e.g., the same core structure), which may be useful to identify unknown metabolites based on similarities in their analytical data. For example, indole metabolites, including the previously characterized I-3AA (fraction D12), all have a beta carbon at the 3-position of the indole ring and a carboxylic acid moiety. These structural similarities were observed in the form of common fragment ions upon GC-MS analysis. These “signature” ions were used to assist in the characterization of the unknowns by structurally linking them.

The GC-EI-MS spectra shown in Figure 6 (A and B) and Figure 2C, share a common, thus potentially diagnostic, fragment ion at m/z 202, even though the three compounds had distinct molecular ions of the trimethylsilylated species of each metabolite, observed at m/z 319, 333 and 421 for fractions D12, E09 and D06 respectively. Distinct molecular ions were found in LC-MS analysis i.e. 174.0546 (D12), 188.0715 (E09), 204.0662 (D06), with all the ions detected in the negative mode. The fragment ion at m/z 202 observed in all three EI-MS spectra (Figures 5 and 2C), arises from an α-cleavage at the β carbon of the trimethylsilylated carbonyl ester. Although these three metabolites all have a different number of carbons, they all share a beta carbon at the 3-position on the indole moiety, and a carboxylic acid functionality. Even with a longer carbon chain on the substituted indole compounds (e.g., Indole-3-Propionic acid (I-3PA) has a longer chain than I-3AA) the fragment ion at m/z 202 is expected to be the base peak, because it can be stabilized by resonance from the indole ring and can be used as a diagnostic ion marker. These GC-MS patterns should be usable to detect and characterize other indole species substituted with alkyl carbonyl functionality.

Figure 6.

Figure 6

Open in a new tab

EI-MS spectrum of fractions D06 (A) and E09 (B). The fragment ion at m/z 202 is a diagnostic ion that is observed in indoxyl compounds substituted with an alkylated carboxylic acid moiety.

Summary and Conclusion

In summary, the results presented show how expanding our LC-MS and NMR platform to include GC-MS and secondary fractionation addresses several fundamental challenges encountered in the structural elucidation of metabolites of interest following metabolite profiling. These include: (i) the presence of co-eluting metabolites, which is resolved using secondary fractionation to collect purer metabolites, thus enhancing their detection and, in turn increasing the probability of characterizing unknowns and; (ii) the chemical diversity of metabolites of interest, which was addressed by the parallel use of multiple analytical techniques. Specific to the work presented, the use of GC-MS, LC-MS and NMR following LC-EC profiling enables the structural identification of previously unidentified compounds shown by statistical analysis of LC-EC profiles to be biologically meaningful.

The parallel use of a combination of analytical technologies (LC-EC, LC-ESI MS, GC-EI-MS and NMR) each with their own advantages, provides structural data that, when taken together, can unambiguously characterize unknown species. MS is critical in identifying the presence of functional groups like carboxylic acids or sulfonic groups that are not detected by NMR. NMR allows for positional isomers to be distinguished. The addition of GC-EI-MS, although limited to the analysis of volatile and/or semi-volatile compounds, produces well-defined, highly-reproducible and structurally-diagnostic fragment ions which can facilitate the characterization of metabolites with similar structural features as demonstrated in the identification of the indole metabolites. These features enable the detection of related compounds with structural similarities across a chromatogram and provide a quick and sensitive way to survey volatile metabolites from an LC chromatogram. Although the method presented was specifically developed for the structural characterization of plasma metabolites following LC-EC metabolite profiling, it is also applicable for the comprehensive analysis of other complex biological matrices e.g. tissue, cells, urine, in which the structural identification of unknown metabolites of interest requires multiple analytical techniques. This method of combining multiple analytical technologies, each with its own strengths, is also generally applicable in any large scale metabolomics profiling study, whereby structural characterization of metabolites identified as biomarkers is needed. In most cases, a single analytical technique is usually insufficient for unambiguous identification of a biomarker metabolite. While the use of multiple analytical techniques all used offline from each other reduces analytical throughput (and thus reducing utility for routine metabolomics analysis) their combined use increases the probability of structurally identifying biologically relevant metabolites, to address emerging challenges as metabolomics research shifts from initial broad discovery surveys to specific etiological hypotheses.

Supplementary Material

01

Acknowledgements

The studies reported were funded by U01-ES16048 (BSK, PI), a part of the NIH Genes and Environment Initiative (GEI), R01-HL109239 (BSK, PI), R01-AG28996 (BSK, PI), 1P01CA168530 (Le Marchand - U. Hawaii; Kristal, Project leader, project 3), Brigham and Women's Hospital Department of Neurosurgery (BSK), The Lamas Family Hydrocephalus Research Initiative (BSK), and R01CA69390 (PV, PI). The authors also thank Thermo Fisher for the loan of an Exactive Bench top orbitrap for demonstration testing and for financial support to attend scientific meetings.

BSK is the inventor on general metabolomics-related IP that has been licensed to Metabolon via Weill Medical College of Cornell University and for which he receives royalty payments via Weill Medical College of Cornell University. He also consults for and has a small equity interest in the company. Metabolon offers biochemical profiling services and is developing molecular diagnostic assays detecting and monitoring disease. Metabolon has no rights or proprietary access to the research results presented and/or new IP generated under these grants/studies. BSK's interests were reviewed by the Brigham and Women's Hospital and Partners Healthcare in accordance with their institutional policy. Accordingly, upon review, the institution determined that BSK's financial interest in Metabolon does not create a significant financial conflict of interest (FCOI) with this research. The addition of this statement where appropriate was explicitly requested and approved by BWH.

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Current Address: Counterpoint Health Solutions, Inc, 135 South Rd. Bedford, MA 01730

References

  • 1.Kristal BS, Vigneau-Callahan K, Matson WR. Simultaneous analysis of multiple redox-active metabolites from biological matrices. Methods Mol Biol. 2002;186:185–94. doi: 10.1385/1-59259-173-6:185. [DOI] [PubMed] [Google Scholar]
  • 2.Kristal BS, Vigneau-Callahan KE, Matson WR. Simultaneous analysis of the majority of low-molecular-weight, redox-active compounds from mitochondria. Anal Biochem. 1998;263(1):18–25. doi: 10.1006/abio.1998.2831. [DOI] [PubMed] [Google Scholar]
  • 3.Paolucci U, et al. Development of biomarkers based on diet-dependent metabolic serotypes: characteristics of component-based models of metabolic serotypes. OMICS. 2004;8(3):221–38. doi: 10.1089/omi.2004.8.221. [DOI] [PubMed] [Google Scholar]
  • 4.Shi H, et al. Development of biomarkers based on diet-dependent metabolic serotypes: practical issues in development of expert system-based classification models in metabolomic studies. OMICS. 2004;8(3):197–208. doi: 10.1089/omi.2004.8.197. [DOI] [PubMed] [Google Scholar]
  • 5.Shi H, et al. Characterization of diet-dependent metabolic serotypes: proof of principle in female and male rats. J Nutr. 2002;132(5):1031–8. doi: 10.1093/jn/132.5.1031. [DOI] [PubMed] [Google Scholar]
  • 6.Shurubor Y, et al. Relative contribution of specific sources of systematic errors and analytical imprecision to metabolite analysis by HPLC–ECD. Metabolomics. 2005;1(2):159–168. [Google Scholar]
  • 7.Vigneau-Callahan KE, et al. Characterization of diet-dependent metabolic serotypes: analytical and biological variability issues in rats. J Nutr. 2001;131(3):924S–932S. doi: 10.1093/jn/131.3.924S. [DOI] [PubMed] [Google Scholar]
  • 8.Shi H, et al. Characterization of diet-dependent metabolic serotypes: primary validation of male and female serotypes in independent cohorts of rats. J Nutr. 2002;132(5):1039–46. doi: 10.1093/jn/132.5.1039. [DOI] [PubMed] [Google Scholar]
  • 9.Shurubor Y, et al. Analytical precision, biological variation, and mathematical normalization in high data density metabolomics. Metabolomics. 2005;1(1):75–85. [Google Scholar]
  • 10.Matson WR, et al. n-Electrode three-dimensional liquid chromatography with electrochemical detection for determination of neurotransmitters. Clin Chem. 1984;30(9):1477–88. [PubMed] [Google Scholar]
  • 11.Shurubor YI, et al. Biological variability dominates and influences analytical variance in HPLC-ECD studies of the human plasma metabolome. BMC Clin Pathol. 2007;7:9. doi: 10.1186/1472-6890-7-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Kristal BS, et al. High-performance liquid chromatography separations coupled with coulometric electrode array detectors: a unique approach to metabolomics. Methods Mol Biol. 2007;358:159–74. doi: 10.1007/978-1-59745-244-1_10. [DOI] [PubMed] [Google Scholar]
  • 13.Bird SS, et al. Structural Characterization of Plasma Metabolites Detected via LC Electrochemical Coulometric Array Using LC-UV Fractionation, MS, and NMR. Anal Chem. 2012;84(22):9889–98. doi: 10.1021/ac302278u. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Schiavo S, et al. Metabolite identification using a nanoelectrospray LC-EC-array-MS integrated system. Anal Chem. 2008;80(15):5912–23. doi: 10.1021/ac800507y. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Ebbel EN, et al. Identification of phenylbutyrate-generated metabolites in Huntington disease patients using parallel liquid chromatography/electrochemical array/mass spectrometry and off-line tandem mass spectrometry. Analytical Biochemistry. 2010;399(2):152–161. doi: 10.1016/j.ab.2010.01.010. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Moco S, et al. Metabolomics technologies and metabolite identification. TrAC Trends in Analytical Chemistry. 2007;26(9):855–866. [Google Scholar]
  • 17.Chen J, et al. Practical Approach for the Identification and Isomer Elucidation of Biomarkers Detected in a Metabonomic Study for the Discovery of Individuals at Risk for Diabetes by Integrating the Chromatographic and Mass Spectrometric Information. Analytical Chemistry. 2008;80(4):1280–1289. doi: 10.1021/ac702089h. [DOI] [PubMed] [Google Scholar]
  • 18.Evans AM, et al. Integrated, Nontargeted Ultrahigh Performance Liquid Chromatography/Electrospray Ionization Tandem Mass Spectrometry Platform for the Identification and Relative Quantification of the Small-Molecule Complement of Biological Systems. Analytical Chemistry. 2009;81(16):6656–6667. doi: 10.1021/ac901536h. [DOI] [PubMed] [Google Scholar]
  • 19.Kind T, Fiehn O. Advances in structure elucidation of small molecules using mass spectrometry. Bioanal Rev. 2010;2(1-4):23–60. doi: 10.1007/s12566-010-0015-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Annesley TM. Ion suppression in mass spectrometry. Clin Chem. 2003;49(7):1041–4. doi: 10.1373/49.7.1041. [DOI] [PubMed] [Google Scholar]
  • 21.Trufelli H, et al. An overview of matrix effects in liquid chromatography-mass spectrometry. Mass Spectrom Rev. 2011;30(3):491–509. doi: 10.1002/mas.20298. [DOI] [PubMed] [Google Scholar]
  • 22.Fiehn O. Extending the breadth of metabolite profiling by gas chromatography coupled to mass spectrometry. Trends Analyt Chem. 2008;27(3):261–269. doi: 10.1016/j.trac.2008.01.007. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Fiehn O, et al. Identification of Uncommon Plant Metabolites Based on Calculation of Elemental Compositions Using Gas Chromatography and Quadrupole Mass Spectrometry. Analytical Chemistry. 2000;72(15):3573–3580. doi: 10.1021/ac991142i. [DOI] [PubMed] [Google Scholar]
  • 24.Halket JM, et al. Chemical derivatization and mass spectral libraries in metabolic profiling by GC/MS and LC/MS/MS. Journal of Experimental Botany. 2005;56(410):219–243. doi: 10.1093/jxb/eri069. [DOI] [PubMed] [Google Scholar]
  • 25.Exarchou V, et al. LC-NMR coupling technology: recent advancements and applications in natural products analysis. Magn Reson Chem. 2005;43(9):681–7. doi: 10.1002/mrc.1632. [DOI] [PubMed] [Google Scholar]
  • 26.van der Hooft JJJ, et al. Structural Elucidation and Quantification of Phenolic Conjugates Present in Human Urine after Tea Intake. Analytical Chemistry. 2012;84(16):7263–7271. doi: 10.1021/ac3017339. [DOI] [PubMed] [Google Scholar]
  • 27.Smith CA, et al. METLIN: a metabolite mass spectral database. Ther Drug Monit. 2005;27(6):747–51. doi: 10.1097/01.ftd.0000179845.53213.39. [DOI] [PubMed] [Google Scholar]
  • 28.Wishart DS, et al. HMDB: a knowledgebase for the human metabolome. Nucleic Acids Res. 2009;37(Database issue):D603–10. doi: 10.1093/nar/gkn810. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Kind T, Fiehn O. Seven Golden Rules for heuristic filtering of molecular formulas obtained by accurate mass spectrometry. BMC Bioinformatics. 2007;8:105. doi: 10.1186/1471-2105-8-105. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Annan M, LeQuesne P, Vouros P. Trimethylsilyl group migration in the mass spectra of trimethylsilyl ethers of cholesterol oxidation products. product ion characterization by linked-scan tandem mass spectrometry. Journal of The American Society for Mass Spectrometry. 1993;4(4):327–335. doi: 10.1016/1044-0305(93)85055-3. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

01
NIHMS578303-supplement-01.doc (108KB, doc)

RESOURCES