C H A P T E R
9 Liquid Chromatographic Methods Combined with Mass Spectrometry in Metabolomics Georgios A. Theodoridis1, Helen G. Gika2, Robert Plumb3, Ian D. Wilson4 1
Department of Chemistry, Aristotle University, Thessaloniki, Greece, Department of Chemical Engineering, Aristotle University, Thessaloniki, Greece, 3 Waters Corporation, Milford, MA, USA, 4 Biomolecular Medicine, Department of Surgery and Cancer, Faculty of Medicine, Imperial College, London, UK
2
O U T L I N E Introduction Chromatographic Methods for Metabolite Profiling Reversed-Phase LC Separations Hydrophilic Interaction Liquid Chromatography (HILIC) Other Approaches to the Profiling of Polar and Ionic Metabolites Miniaturized LC Systems
Proteomic and Metabolomic Approaches to Biomarker Discovery http://dx.doi.org/10.1016/B978-0-12-394446-7.00009-1
146 147 147 151 153 154
Detection
155
Quality Control, Data Analysis, and Biomarker Detection
156
Metabolite Identification and Biomarker Validation
158
Conclusions
159
References
159
145
Copyright Ó 2013 Elsevier Inc. All rights reserved.
146
9. LIQUID CHROMATOGRAPHIC METHODS COMBINED WITH MASS SPECTROMETRY IN METABOLOMICS
INTRODUCTION The application of liquid chromatography (LC) for metabolomic/metabonomic studies, in which untargeted analytical methods are used in an attempt to obtain metabolic profiles of samples of biofluids or tissues that are as comprehensive as possible in order to advance biological understanding and discover novel biomarkers of various conditions, forms an increasingly important area of research.1,2 Although in principle a wide range of analytical methodologies could be used for such work, in practice the major techniques employed on a routine basis are gas chromatography (GC), nuclear magnetic resonance (NMR, especially 1 H NMR) spectroscopy, capillary electrophoresis (CE), and increasingly liquid chromatography (LC). Indeed, the rise of LC-based methodsdin particular those using LC-MSdhas been very rapid, given that published applications only began to appear early in this century (see, e.g., Plumb, Stumpf, et al. and Plumb, Granger, et al.3,4). However, LC-based methods have rapidly expanded in their application in this type of holistic, or global, form of metabolic profiling because of their ready compatibility with aqueous biological samples, whichdlike urine or blood plasmadcan usually be analyzed with minimal sample preparation compared to GC, for example. The use of LC-MS also potentially affords high sensitivity (though this is analyte-dependent) combined with spectrometric information that greatly aids analyte identification. As a result of these favorable analytical properties, there has been an enormous expansion in the use of LC-MS-based techniques.5e7 Initially, the LC systems employed were based on conventional high-performance liquid chromatography (HPLC) with separations performed on 3 to 5 mm particles. More recently, however, there has been a general move towards the use of separations on sub-2 mm particles and employing higher pressures than conventional HPLC in ultra-highpressure LC (UHPLC, or ultra performance,
UPLC) techniques to take advantage of the improved chromatographic resolution provided by these more advanced LC techniques.8,9 A more limited number of applications has also been published for capillary and miniaturized LC systems. Obviously, the operation of the hyphenated technique of LC-MS, despite the enormous technical advances made in the last two decades, is not without difficulty, and factors such as sample-dependent ion suppression/ enhancement represent an ever-present problem when the composition of the sample is both unknown and variable. Such considerations mean that although LC-MS can be used to rapidly generate large volumes of data in the pursuit of metabolic profiling, the value of that data depends very much on the ability of the analyst to ensure its quality. Clearly, there is always some level of conflict between the desire for the most comprehensive metabolite profile technically possible and the urgent need for high sample throughput. Indeed, in order to maximize throughput, many researchers advocate eliminating the separation step completely and infuse the samples directly into the MS. However, our view is that except for very simple matrices, the disadvantages of such a “separation-free” approach with its attendant problems based on matrix interferencesdin particular, ion suppression or ion enhancement and the difficulty of distinguishing between isobaric species and structural isomersd outweigh any possible increase in throughput. In particular, the requirement to obtain the most comprehensive metabolite profiles requires a chromatographic separation. The need to reduce the length of the chromatographic run to a minimum has, however, been one of the driving forces behind the adoption of UHPLC-MS and the displacement of the lower resolution HPLC-MS methods. Thus, the peaks obtained via UHPLC are usually only 2 to 5 seconds wide at the base, affording significantly greater peak capacity than conventional HPLC methods. However, if the need for high throughput is paramount, then it is quite possible to use UPLC run times of
CHROMATOGRAPHIC METHODS FOR METABOLITE PROFILING
approximately 2 min (see, e.g., Plumb et al.9) and analyze 20 to 30 urine samples/h while perhaps achieving the same level of metabolome coverage as a 10 to 15 min HPLC analysis. Currently, however, chromatographic run times (for both HPLC and UHPLC) are typically in the range of 10 to 20 min. This range allows for reasonable sample throughput and good chromatographic resolution and detailed protocols of this type for biofluids such as serum/plasma and urine and tissue extracts have been described.12e16 When, on the other hand, a more detailed investigation of one or a few samples is the aim, then resolution, rather than speed of analysis, is the main requirement, and the use of very high resolution systems such as multidimensional or capillary LC may represent the best strategy. Irrespective of the desired analytical outcome, there are a limited number of options with respect to the mode of chromatography and chromatographic phases that are available for use in metabolomics/metabonomics. Currently, the most widely employed type of separation is conventional gradient reversedphase (RP) chromatography. Such separations need to be carefully optimized for the metabolites likely to be present in a given matrix; for example, a method optimized for urine would be much less suitable for another matrix such as plasma, as the latter is lipid rich and urine is composed of much more “polar” metabolites. Indeed, one of the limitations of RPLC is that the analysis of polar/ionic metabolites represents a considerable challenge because these are often poorly retained, eluting at, or near, the solvent front. RPLC is therefore most suitable for the profiling of “medium” to nonpolar metabolites. The profiling of these unretained metabolites demands a different strategy, and the use of separations based on either hydrophilic interaction (HILIC), ion-pair (IPLC), ion-exchange (IEC), or aqueous normal phase chromatography (ANPC). In this chapter, the practice of LC-MS for metabonomic/metabolomic studies is
147
considered and the various modes and options for chromatography described.
CHROMATOGRAPHIC METHODS FOR METABOLITE PROFILING Reversed-Phase LC Separations HPLC-based separations in global metabolic profiling studies employ 3 to 5 mm particle size stationary phases packed into 2.1 to 4.6 mm i.d. columns 5 to 15 cm in length. Typical RP gradients for HPLC that we have found useful for the analysis of samples such as urine used a 2.1 mm i.d. 10 cm column packed with, e.g., 3.5 mm, C18-bonded SymmetryÒ packing held at 40 C and with a solvent delivery flow rate of 600 ml/min.17 For UHPLC (UPLC)-based metabolite profiling, in which sub 2 mm stationary phases are employed, the columns used vary from 5 to 15 cm in length and are of aproximately 2 mm i.d.9,14e17 Apart from the higher operating pressures used in UPLC compared to HPLC, the mobile phases, gradient elution profiles, and flow rates used are similar. RPLC is currently the most widely used separation mode for metabolomic/metabonomic profiling, stemming partly from the fact that it is well suited to the analysis of aqueous samples (e.g., urine, bile, and proteinprecipitated plasma/serum). Gradient elution is usually over 10 to 30 min depending on the type of analysis with 2 to 5 min at the end of the runs for column washing and reequilibration. Thus, for urine a typical method would use 0.1% aqueous formic acid and acetonitrile (also containing 0.1% formic acid) for chromatography with a gradient beginning at 100% aqueous formic acid (held for 0.5 min) rising in a linear gradient to 20% acetonitrile over 0.5 to 4 minutes and then continuing to 95% acetonitrile at 8 min. The solvent composition would then be held at 95% acetonitrile for a further minute to wash strongly retained contaminants from the column, then returning to 100% aqueous formic
148
9. LIQUID CHROMATOGRAPHIC METHODS COMBINED WITH MASS SPECTROMETRY IN METABOLOMICS
acid at 9.1, re-equilibrating for 0.9 min ready for the next sample. For plasma extracts, protocols for the RPLC-MS analysis of both serum/plasma and urine have been reported.10,11 For samples such as urine, minimal sample preparation is required (often only centrifugation and dilution11,14) and a typical UPLC separation is shown in Figure 1. With samples, such as plasma or serum9,10,15,16 or tissues18e20, it is essential to remove proteins prior to LC analysis in order to protect the column. It is also the case, as indicated previously, that relative contributions of polar and nonpolar metabolites (e.g., lipids) differ considerably between urine- and blood-derived or tissue samples. As a result, the gradient used for the analysis of such samples often begins at a higher proportion of the organic solvent than employed for urine. Lipids are also more strongly retained on RPLC columns and, as a result, their elution (and
column cleaning after the analysis) requires more eluotropic solvent for elution and wash steps. Generally, methanol (or other alcohols) appears to be a better solvent for eluting lipidic contaminants compared to, for example, acetonitrile and a typical (UPLC) analysis of such a sample would use separation on a 1.7 mm, 2.1 100 mm, Acuity BEH column 50 C eluted with 0.1% aqueous formic acid (solvent A) and methanol (also containing 0.1% formic acid, solvent B) at 0.4 mL/min. The gradient program employed a series of linear steps, beginning at 95% aqueous formic acid for 0.5 min., and changing to 60% A at 2.5 min, then to 30% A at 4.5 min and finally rising to 100% B at 10 min. This solvent composition was held for 2 min before the column was returned to 95% aqueous formic acid, where it was held for 2.5 min prior to the next sample. The very high efficiency of UHPLC often means that chromatographic
FIGURE 1 UHPLC-TOFMS (+ESI) of rodent urine on an Acquity BEH C18 2.1 100 mm 1.7 mM column (maintained at 40 C) using a solvent gradient of 0.1% aqueous formic acid versus acetonitrile (containing 0.1% formic acid). The upper trace shows the metabolites recovered from a “blood spot” paper and the lower trace is of the unextracted urine. Details can be found in reference 29.
CHROMATOGRAPHIC METHODS FOR METABOLITE PROFILING
peaks are only 3 to 5 seconds wide, requiring similar speed of response from the detector and thus a rapid duty cycle from the mass spectrometer. Although retention in RPLC is generally controlled via changes in mobile phase composition, temperature can also be used as a variable and, with the recent resurgence of interest in high temperature (HT) LC, there have also been some limited demonstrations of the potential usefulness of operating at higher temperatures for profiling samples containing mainly polar metabolites, such as urine. Thus, operating at high temperature decreases solvent viscosity, meaning that higher flow rates can be used if desired, and also changes solvent properties such that the need for organic modifiers can be reduced or even eliminated. In one example of the application of HT-UPLC to metabolite profiling, the temperature of the UPLC column was maintained at 90oC21 while a normal RP-solvent gradient was used to elute the urinary components. In another example, the organic solvent was dispensed with and a temperature gradient was performed to elute urinary-excreted metabolites.22 Lipids represent an important class of analytes in their own right, which has led to the development of the subfield of lipidomics. Gradient RPLC has formed the basis of a number of methods23e25 with gradient systems based on the use of acetonitrile-aqueous ammonium formate (10 mM) 2:3 v/v versus acetonitrile-isopropanol 1:9 v/v (containing ammonium formate (10 mM)),24e25 or 10 mM ammonium acetate versus acetonitrile-isopropanol 5:2 v/v (containing ammonium formate (10 mM).26 Elution of lipids was accomplished at 0.4 ml min and 55oC on an AcquityÔ HSS T3 column (2.1 mm i.d. 100 mm) via linear gradients rising first from 100% of the aqueous solvent to 40% of the organic mobile phase over 3 min, then to 100% organic solvent over the next 10 min, with this composition held for 2 min before returning to the starting conditions for 3 min prior to injection of the next
149
sample.24 An example of this type of separation is shown in Figure 2. Although specimens such as urine generally present few problems for metabolite profiling by LC-based methods, with sample preparation often limited to centrifugation and dilution, protein-rich samples such as plasma/serum and others are more problematic. Typically, the proteins in such samples are removed by the simple expedient of mixing the sample with an excess (three volumes) of a solvent such as methanol or acetonitrile, causing the proteins to precipitate followed by their removal by centrifugation. Solid-phase extraction (SPE) can also be applied27 to remove proteins from plasma and an online method for protein removal,28 turbulent flow chromatography (TFC), has been the subject of a short investigation for plasma metabolome profiling. Plasma samples are first passed though a large particle-containing column (25e50 mm) in which small molecules are retained and the proteins are eluted to waste as a result of “turbulent flow” caused by the combination of high flow rates and large chromatographic particles. Plasma can be directly injected onto the LC system without compromising chromatographic performance. The methodology can be thought of as either online extraction or a crude 2D-LC method, as the next step is to elute the retained analytes from the TFC column (using a reversed-phase solvent gradient) on to a conventional HPLC column to obtain the metabolic profile. The preliminary investigation was promising28 but revealed interesting differences between profiles found by TFC versus those of methanol-precipitated plasma showing, for example, reduced amounts of phospholipids (ca. tenfold reduction) for the former. Finally, for samples such as blood (but also other biofluids such as urine or bile), techniques such as dried blood spots, in which the sample is applied to a paper matrix and then subsequently extracted with solvent, can be used to remove protein. This type of blood spot technique has been briefly explored for metabonomic
150
9. LIQUID CHROMATOGRAPHIC METHODS COMBINED WITH MASS SPECTROMETRY IN METABOLOMICS
RPLC – MZ 160 13W6
1: TOF MS ES+ 160.604 1.79e3
5.36
100
100 80 60 40
9.01
20 0
%
-20 -40
2.30
-60 -80
2.43
-100
7.19
2.69
-100
3.87
0.09 0.73
100
0
4.99 5.75
SIMCA-P 11 - 16/11/2007 9:04:23 ìì
4.79
0
Time 2.00
4.00
6.00
8.00
10.00
12.00
HILIC-MZ-160 13W6
1: TOF MS ES+ 160.604 1.79e3
5.36
100
100
9.01
%
0
2.30 -100
2.43 7.19
2.69 3.87
0.09 0.73
4.99 5.75
-200
4.79
-100
0
100
200
SIMCA-P 11 - 16/11/
0
Time 2.00
4.00
6.00
8.00
10.00
12.00
FIGURE 2 Base peak chromatogram of human plasma extract. Retention time windows for the lipid classes are seen in the UPLC-TOF MS chromatogram. Analysis was performed on a 1.8 mm particle 100 2.1 mm i.d. Waters Acquity HSS T3 column at 55 C using gradient of acetonitrile:water (40:60, v/v) 10 mM ammonium acetate (eluent A) versus acetonitrile: isopropanol (10:90, v/v) 10 mM ammonium acetate (eluent B). Elution started with 60% A; for 10 min the gradient was ramped in a linear fashion to 100% B, where it was held for 2 min. The flow rate was 0.4 mL/min and the injection volume was 10 mL. MS detection was performed on SYNAPT HDMS, in electrospray in both positive (top pane) and negative (bottom pane) ionization. (Reprinted from reference 24, with permission from the American Chemical Society.)
CHROMATOGRAPHIC METHODS FOR METABOLITE PROFILING
studies29,30 and an example of the UPLC profile of urine obtained following this type of dried blood spotting compared to direct analysis of the sample is shown in Figure 1. All of these techniques render samples suitable for RPLC but each method gives slightly different metabolite profiles.
Hydrophilic Interaction Liquid Chromatography (HILIC) As indicated earlier, polar/ionizable, metabolites (sugars, amino acids, organic acids, etc.) pose a problem for RPLC methods due to poor retention. For the analysis of such analytes, the current standard approach in metabolomics is to use HILIC. Such separations employ solvents containing a very high proportion of organic modifier, and indeed the water content should be limited to no more than 50%; as a result, MS sensitivity in HILIC is often better than for RP methods because ionization efficiency is improved due to efficient generation of spray conditions. The value of HILIC for untargeted metabolite profiling has now been demonstrated in a wide range of applications31e34 and, when combined with RPLC on the same samples, probably provides the most comprehensive metabolome coverage currently available by LC using “routine” methodology. HILIC can either be used directly on samples such as urine or samples can be extracted using, for example, SPE first to separate polar from nonpolar metabolites. This approach was employed for the analysis of rat urine samples31 in which the unretained portion of the sample following SPE was analyzed via HILIC/electrospray ionization (ESI)-MS (and the retained portion by RP-LC/ ESI-MS). The polar metabolites obtained in this way were profiled using HPLC on a ZIC-HILIC column (100 2.1 mm, 3.5 mm). Although still not as widely employed as RPLC, the use of HILIC has continued to increase with examples of both HPLC and UHPLC-based metabolomic/metabonomic studies (reviewed in Spagou
151
et al.33). Examples using the format HPLC include separations performed with ZIC-HILIC (100 4.6 mm, 3.5 mm)34 or the Aphera NH2 polymer column (150 2 mm, 5 mm)35 and UHPLC applications of HILIC include examples based on the Acquity BEH HILIC (2.1 50 mm, 1.7 mm) material for profiling32,36 in, for example, animal disease model metabotyping,32 toxicological investigations, or for cancer biomarker discovery.37 In the latter application, the profiling of metabolites present in the urine of rats exposed to galactosamine (a model hepatotoxin) was performed by UPLC on an Acquity BEH HILIC column (2.1 100 mm, 1.7 mm) at a flow rate of 0.4 mL/min and a column temperature of 40 C. A solvent gradient separation was used with 0.1% (v/v) formic acid and 10 mM ammonium acetate in acetonitrile-water 95:5 v/v as the initial organic and 0.1% aqueous formic acid and 10 mM ammonium acetate in acetonitrile-water 50:50 as aqueous solvent. The initial solvent composition employed was 99.0% organic and 1% aqueous solvents for 1.0 min, followed by a linear gradient of the aqueous solvent over the next 11.0 min. On completion of the gradient, the solvent composition was returned to the starting conditions (0.1 min) and allowed to re-equilibrate for 4 min before injection of the next sample. Although HILIC has been used for polar metabolites as described earlier, it has also been applied as a complementary method for polar lipid separations on a Diol column to complement an RPLC separation in a 2D separation. Here the eluent from the HILIC column was collected manually, desalted, and concentrated before reanalysis by RPLC.37 The isocratic separation used a 5 mM NucleosilÔ 100-5OH packing contained in a 10 mm i.d. 250 mm column, with a mobile phase consisting of a mixture of cyclohexane and isopropanol-water-acetic acid28% aqueous ammonia 86:13:1:0.12 (v/v) in a ratio of 1:9 v/v at a flow rate of 1 ml min.38 It is clear that RP and HILIC, if used alone, will not provide comprehensive metabolome
152
9. LIQUID CHROMATOGRAPHIC METHODS COMBINED WITH MASS SPECTROMETRY IN METABOLOMICS
PC, SM, PE & DG 5.03 Positive ESI ion mode LPC & LPE
2.25 5.60 5.47
4.95 TG & ChoE
3.66
4.78 5.93
1.46 1.71 0.59 1.11
8.82 9.00 6.06 9.22 9.31 9.47
8.60
6.18
4.64
1.96
Time 1.00
2.00
3.00
4.00
5.00
6.00
7.00
8.00
9.00
10.00
PG, PI, PS, PE 5.09
FA Negative ESI ion mode
2.24 5.03 5.47
5.61
2.12 1.80 1.66
2.86
1.46
4.21
3.57
4.95 4.78
4.57
6.05
6.44
8.37
6.93 Time
1.00
2.00
3.00
4.00
5.00
6.00
7.00
8.00
9.00
10.00
FIGURE 3 Profiles from RP (upper) and HILIC (lower) -ToF-MS analysis of urine, obtained from Zucker lean and fat rats, for the ion m/z160 showing the better retention of these polar substances on the HILIC phase . The PCA plots to the right of the chromatograms show a clear separation (based on the whole metabolite profiles) between the two groups regardless of which method of profiling was used (but based on different markers). Red ¼ lean; black ¼ fat.
CHROMATOGRAPHIC METHODS FOR METABOLITE PROFILING
coverage and the best results are obtained if both types of separation are used. Although they can be used separately, it is also technically feasible to combine them into a single 2D system and indeed such a combined HILIC and RP separation.39 The resulting separation method was successfully used to provide metabolic profiles of urine samples obtained from lung cancer patients. Here polar compounds were separated on a TSK gel amide-80 column (2.0 50 mm, 3 mm) and the less polar ones using a HypersilÔ ODS2, C18 column (4.6 150 mm, 5 mm) and RPLC, and was found to be reproducible and robust. An example of the use of UPLC for both HILIC and RPLC analysis of urine from Zucker rats32 is shown in Figure 3.
Other Approaches to the Profiling of Polar and Ionic Metabolites Although HILIC has been widely used for polar compounds as discussed previously, it does not yet provide a complete answer to the analysis of polar, ionic compounds and it is arguable that techniques such ion exchange chromatography (IEC) should provide another approach to the separation of such metabolites present in samples. Preliminary studies using IEC for profiling rat urine samples from (fa/fa) and (e/e) Zucker rats, representing obese and lean phenotypes, were promising, demonstrating separation between the two classes on multivariate statistical analysis by principal component analysis (PCA). In this study, samples were analyzed on an IonPacÔ AS18 2 250 mm column. Separations were performed using gradient elution with the starting conditions being 5 mM NaOH for 1 min, changing with an exponential curve to 100 mM NaOH at 12 min at a flow rate of 300 ml/min. These conditions were maintained for a further 3 min before returning to the starting conditions for the next 8 min (to re-equilibrate the column) prior to the next injection. Another alternative to HILIC is ion-pair liquid chromatography (IPLC), but it has to be
153
recognized that this methodology has long-term consequences for MS-based applications as a result of the contamination of the instrument by the ion pair reagent. This contamination can prove extremely resistant to cleaning, and it may therefore be necessary to dedicate a mass spectrometer solely to this type of analysis. However, where the profiling of polar acidic metabolites is required, the disadvantages of IPLC may be compensated for by the utility of the method. Though many examples of the use of the IPLC technique involve targeted analysis, an example of this type of method, with detection via an OrbitrapÔ mass spectrometer that was used for the sensitive and specific detection of both unknown and known metabolites (5 ng/ml for 80 metabolite standards) used of RP-IPLC on a SynergiÔ Hydro-RP 2.5 mm C18 column (100 mm 2 mm i.d.) employing a flow rate of 200 mL/min.40 In a study on a strain of yeast engineered such that a gene of unknown function (YKL215C) was knocked out, the accumulation of oxoproline was observed, allowing the gene to be identified as oxoprolinase. The IP reagent was tributylamine (10 mM) (and 15 mM acetic acid) with a water-methanol gradient. The starting conditions for the gradient were 97:3 water/ methanol for 2.5 min, increasing the methanol content to 20% methanol at 5 min, then, from 7.5 min, increasing to 55% methanol at 13 min, then, at 15.5 min, rising to 95% methanol at 18.5 min. This solvent composition was maintained for a further 0.5 min before returning to the starting conditions. Tributylamine was also used for the analysis of cell extracts of Methylobacterium extorquens with a nanoscale LC-MS system41 and a hexylamine-based IP system42 has been used to analyze extracts of Lactobacillus plantarum, E. coli, and B. subtilis by LC-MS. IPLC, despite the manifest disadvantage of the need to essentially dedicate an instrument to the technique, provides a robust and reliable method for polar acidic metabolites and may well become firmly established in this role in the future.
154
9. LIQUID CHROMATOGRAPHIC METHODS COMBINED WITH MASS SPECTROMETRY IN METABOLOMICS
Last, a further separation tool with potential for the separation of highly polar metabolites is aqueous normal phase chromatography (ANPC).43 This technique was used to profile metabolites in human urine and plant extracts on a Cogent Diamond HydrideÔ column (100 2.1 mm, 4 mm) via gradient elution using 15.9 mM ammonium formate or 13.0 mM ammonium acetate and acetonitrile-water (containing 15.9 mM ammonium formate or 13.0 mM ammonium acetate) 90:10 v/v at 0.4 mL/min.
Miniaturized LC Systems In addition to the various different modes of chromatography that can be used for sample analysis, there is also a range of technical options that can be useful for metabolome profiling. So, although not yet widely used for routine analysis, there are obvious benefits to increasing both sensitivity and resolution, as well as providing minimal sample consumption, to be derived from the use of miniaturized LC column formats. Examples have been shown using micro bore columns, with internal diameters from 0.5 to 1.0 mm, and capillary columns of varying lengths in a number of metabolic profiling studies.44e46 For example, C18-bonded silica monolithic columns (0.2 mm i.d., 30e90 cm) were used to profile extracts of Arabidopsis thaliana43 with MS detection. In another example, an
FIGURE 4 A gradient RPLC-MS separation of rat urine on a capillary scale microfluidic device.
impressive, and high-resolution, separation using a 50 mm i.d. fused silica capillary, 200 cm long and filled with 3 mm porous C18-bonded particles was applied to the separation of metabolites present in cell lysates of the microorganism Shewanella oneidensis.46 A reversed-phase gradient was used, at a pressure of 20 Kpsi, resulting in the detection of more than 5,000 metabolites over some 2,000 min. Although long capillaries provide numerous benefits, there are also some advantages to be obtained from the use of shorter ones, so, for example, metabolic profiling of the urine of Zucker rats (a model for Type II diabetes) was performed on a 320 mm i.d. (10 cm) column packed with a 3.5 mm C18-bonded stationary phase.45 This analysis provided a more comprehensive profile than conventional HPLC-MS on a column containing the same packing material, and of the same length, with about twice as many ions observed via the capillary method (presumably as a result of reduced ion suppression). More recent developments have led to capillary scale microfluidic systems such as the Integrated Ceramic Microfluidic DevicesÔ.47 In this device, a 10 cm long, 300 mm i.d. separations channel, containing the 1.7 mm Acquity BEH stationary phase, was employed for the separation. An example of the type of separation that can be achieved using this type of platform is shown in Figure 4. Here rat urine (1 ml) was
DETECTION
profiled using a typical RP-gradient from 5 min to 95% acetonitrile versus 0.1% aqueous formic acid over 10 min at a flow rate of 12 ml/min. Clearly, whatever type of separation and mode of chromatography is employed, if metabolic phenotypes are to be compared, the generation of repeatable profiles from samples is of central importance and therefore retention time and peak shape must be stable over the whole length of an analytical run (which may well exceed 24 hrs on a routine basis). Although techniques such as UPLC can demonstrate impressive repeatability, researchers should be aware that, depending upon the sample matrix, the age of the column, and its previous exposure to samples, it can take a number of injections before the system is equilibrated, or conditioned, and retention times stabilize.10e16 In terms of the chromatographic process, we assume that what is happening during this “conditioning phase” is that matrix components are interacting with the column packing (or some other part of the system, e.g., tubing or frits) in such a way that active sites and silanols are masked. Conditioning is matrix dependent, with samples such as urine usually requiring fewer conditioning injections than serum or plasma. Once conditioned, we have found retention to be stable, with the buildup of contamination in the ion source of the mass spectrometer causing the bulk of the analytical variability thereafter. It is also our experience that in addition to conditioning, column lifetime is matrix dependent. Thus, analysis of urine generally provides longer column lifetimes (several thousand samples) than, for example, serum or plasma (up to 1,000 samples) before replacement is required.
DETECTION Clearly, a range of LC detectors are available that could, in principle, be used to interrogate the separation and produce metabolite profiles. These include sensitive and inexpensive but
155
relatively low information content detectors such as UV/vis (DAD) spectrophotometers and electrochemical detectors up to MS and NMR spectrometers of much greater expense but potentially of high information content. However, as a result of its combination of sensitivity and structure determination/identification potential, MS probably currently provides the best choice for LC-metabolomic/metabonomic profiling. MS is not without limitations, however, and it should always be remembered that ionization efficiency is compound specific and may differ widely even for structurally related compounds. Currently, ESI, combined with time-of-flight mass spectrometry (TOF-MS), is the method of choice for LC-MS-based metabolomic/metabonomic profiling, using both positive and negative modes to maximize the coverage of the metabolome. Other ionization techniques, such as atmospheric pressure chemical ionization (APCI), are available and could potentially have advantages for nonpolar metabolites, but APCI has not yet been widely applied in metabolome profiling. Along with the use of TOF instruments, metabolome profiles have also been determined using linear ion traps, as well as hybrid spectrometers such as IT-TOF48 or the QTRAPÔ13,48 although, where the highest mass resolution is required, Fourier transform ion cyclotron MS (FT-ICR) or Orbitrap MS instruments can be used. The high-resolution spectrometers provide excellent mass accuracy, enabling more accurate determination of atomic composition and so on to aid in metabolite characterization and identification. One area of difficulty that those performing this type of profiling for biomarker research should be aware of, though, is that different types of mass spectrometer do not necessarily give the same responses. Thus, although the dangers of trying to compare the results of metabolite using different types of chromatography are obvious, the problems of comparing and correlating LC-MS data derived from different MS instrumentsdeven when the same separation is useddare less obvious. That
156
9. LIQUID CHROMATOGRAPHIC METHODS COMBINED WITH MASS SPECTROMETRY IN METABOLOMICS
such differences can be problematic is illustrated by the use of QTOF and QTRAP mass spectrometers to simultaneously profile the same eluent (split 50:50 as it emerged from the column). The outcome of the study was that the data from both instruments for the urinary metabolite profiles obtained from control and test animals were readily separated by PCA, the “markers” were spectrometer dependent.49
QUALITY CONTROL, DATA ANALYSIS, AND BIOMARKER DETECTION Analytical variability resulting from changes in system performance can pose a major threat to the successful performance of metabolic phenotyping studies, and in LC-MS data such variability can result from changes in chromatography due to column degradation (e.g., gradual or catastrophic changes in peak shape and retention time) or changes in spectrometer performance (e.g., changes in mass accuracy or signal intensity). Such changes need to be monitored and corrective action taken to avoid severe problems in any subsequent analysis of the data. Our approach to minimizing such problems revolves around a standard “quality control” or QC sample.10e16 These QCs can be prepared by pooling aliquots of the samples being analyzed, thereby providing a representative sample, or, where this is not practicable, by using a bulk sample of the matrix (e.g., plasma or serum obtained from a blood bank or commercial supplier). These QCs are then interspersed every 5 or 10 samples throughout the run. As each of these samples is identical in composition, by monitoring the variability observed in them when they are analyzed, the quality of the analysis can be assessed. The same QC samples can also be used to condition the LC-MS system prior to the start of the run, as discussed earlier. The metabolite profiling exercise for each of the samples in the run generates a large amount of
LC-MS data on signal intensity, analyte mass (over the range w70 to 1,200 amu) and retention time. The raw data is usually in the form of a series of full scan mass spectra comprising the spectral data for the metabolites (including adducts, isotopic peaks, and systematic noise, etc.) that have been acquired over successive time points (each of w2e20 ms). To process the enormous amount of data residing in these files and extract useful information, specialized software is needed. This demand has led to the development of many proprietary, freeware, and in-house programs; these programs allow instrument noise to be removed, baselines to be corrected, centering, normalization, peak picking, peak integration and alignment, de-isotoping, adduct removal, and other tasks to be performed to reveal the metabolite peaks that were present in the samples. This work allows the construction of a peak table that lists the samples, the ions for the metabolites, and their intensities with the 3D retention time/mass/intensity information compressed into two dimensions by combining the mass and retention time data into a single “feature.” Data analysis can be undertaken using either the proprietary software generally available from the manufacturer or freely available open source programs such as MZmine,50 MetAlign,3,51 and XCMS,52 which can be used to analyze data once it has been converted into a suitable format such as netCDF or mzXML. An advantage to the operator of open source software packages is that they can be customized for the individual needs of the study (e.g., a requirement to accommodate particularly broad or narrow peaks, etc.). The initial means of examining this type of metabolic phenotyping data is most commonly to use multivariate statistical analysis (e.g., PCA) to highlight differences between samples from test and control groups; however, in order to have confidence that the detected features are genuine indicators of the condition being investigated, additional univariate statistical analysis and manual examination of the raw data represent good practice. Data analysis
QUALITY CONTROL, DATA ANALYSIS, AND BIOMARKER DETECTION
FIGURE 5 A typical workflow for LC-MS-based metabolomics.
157
158
9. LIQUID CHROMATOGRAPHIC METHODS COMBINED WITH MASS SPECTROMETRY IN METABOLOMICS
still remains one of the most time-consuming steps of any metabolomic/metabonomic study, but care taken with this aspect of the work will be repaid in terms of genuine possibilities of discovering metabolic associations with the condition under study (MS-based data analysis for metabolomics is reviewed in the literature52).
METABOLITE IDENTIFICATION AND BIOMARKER VALIDATION Once potential biomarkers have been highlighted by the data analysis, they must then be characterized and unambiguously identified. The sheer variety of molecular structures that may be encountered means that such work can be quite time- and resource-consuming and continues to represent a significant bottleneck in LC-MS-based biomarker discovery. The work is still hampered by a lack of readily transferable spectral libraries for LC-MS (in contrast to, for example, GC-MS) analysis, and retention time data is by definition stationary and mobile phaseespecific and also not readily transferred between systems/studies/laboratories. Currently, unless the potential biomarker is easily identified on the basis of in-house databases, the strategy for identification would involve the pragmatic use of a combination of using any accurate mass data available to determine a probable elemental composition and the examination of any molecular fragmentation data to inform searches made on the available databases such as METLIN,54 lipidmaps,55 the Human Metabolome Database (HMDB)56,57 and Chemspider,58 etc. Such searches can narrow the field of likely candidates to a more manageable number, and this number can be further reduced by using clues such as the chromatographic retention data to attempt to discriminate between polar and nonpolar candidates. Biochemical “plausibility” may also be brought into play to further narrow the field, at which point further targeted MS experiments
may be used to narrow the field even more. Comparison with an authentic standard then represents the final confirmation, and arguably identification is only provisional until that point. If standards are not available, then isolation and further spectroscopic analysis such as NMR spectroscopy or chemical synthesis may be necessary for unequivocal identification. Identification of the putative biomarker represents only part of the process of biomarker discovery, and this initial metabolic phenotyping work is sufficient only to generate the hypothesis that certain molecules are indicative of a particular condition. Proof of the hypothesis depends on the development of properly validated and specific methods to allow re-analysis of the incurred samples to demonstrate that the changes are real and not artifactual and then prospective studies to fully validate their value and convert the analytes from potential to actual biomarkers. For example, following metabonomic/metabolomic investigations both 5-oxoproline and ophthalmic acid have been proposed as biomarkers of glutathione depletion following exposure of animals and cells to compounds such as acetaminophen (paracetamol). The development of specific analytical methods for these compounds enabled their quantitative measurement59,60 and their role as informative “downstream” biomarkers of reactive metabolite-mediated toxicity to be understood via a systems biology approach.61 Similarly, in investigations on human urine undertaken as part of a large epidemiological study (INTERnational collaborative study of MAcronutrients, micronutrients, and blood Pressure, or INTERMAP), a number of urinary metabolites were seen that appeared to be associated with blood pressure. These observations were then followed by the development of quantitative UPLC-MS methodology for the quantification of urinary phenylacetylglutamine, 4-cresyl sulphate, and hippurate and provide reference ranges for the 24-hr urinary excretion of these metabolites in free-living individuals.62
REFERENCES
In Figure 5, a typical workflow for metabolic profiling and data analysis and other work of the type that would be used in metabolomic/ metabonomic investigations is shown.
5.
6.
CONCLUSIONS 7.
As currently practiced, metabolomic and metabonomic studies are increasingly being performed using LC-MS-based techniques. Obtaining the most comprehensive metabolite profile does, however, require the use of more than one separation mode (e.g., RPLC and HILIC), and more than one mode of ionization (i.e., both positive and negative ESI, perhaps with positive and negative APCI as well). Although it is clear that the methodology is still rapidly developing, robust protocols are appearing for various types of sample and continuing rapid advances in instrumentation will further improve efficiency. The difficulties that remain center mainly on the identification of unknowns detected as potential biomarkers during the course of metabolic profiling experiments. However, great strides are being made in the development of metabolite databases and other resources that will greatly help reduce the complexity of the task in the future.
8.
9.
10.
11.
12
13.
References 1.
Nicholson JK, Connelly J, Lindon JC, et al. Metabonomics: a platform for studying drug toxicity and gene function. Nat Rev Drug Discov 2002;1:153e61. 2. Nicholson JK, Lindon JC. Systems biology: metabonomics. Nature 2008;455:1054e6. 3. Plumb RS, Stumpf CL, Gorenstein MV, et al. Metabonomics: the use of electrospray mass spectrometry coupled to reversed-phase liquid chromatography shows potential for the screening of rat urine in drug development. Rapid Commun Mass Spectrom 2002; 2002(16):1991e6. 4. Plumb RS, Granger J, Stumpf C, et al. Metabonomic analysis of mouse urine by liquid-chromatographytime of flight mass spectrometry (LC-TOFMS): detection of strain, diurnal variation and gender differences. Analyst 2003;128:819e23.
14.
15.
16
17
159 Theodoridis G, Gika HG, Wilson ID. LC-MS-based methodology for global metabolite profiling in metabonomics/metabolomics. Trend Anal Chem 2008; 27:251e60. Theodoridis GA, Gika H, Wilson ID. Mass spectrometry-based holistic analytical approaches for metabolite profiling in systems biology studies. Mass Spectrom Rev 2011;30. 884-806. Wu Z, Huang Z, Lehmann R, et al. The application of chromatography-mass spectrometry: methods to metabonomics. Chromatographia 2009;69:S23e32. Wilson ID, Nicholson JK, Castro-Perez J, et al. High resolution “ultra performance” liquid chromatography coupled to oa-TOF mass spectrometry as a tool for differential metabolic pathway profiling in functional genomic studies. J Proteome Res 2005;4:591e8. Plumb RS, Granger JH, Stumpf CL, et al. A rapid screening approach to metabonomics using UPLC and oa-TOF mass spectrometry: application to age, gender and diurnal variation in normal/Zucker obese rats and black, white and nude mice. Analyst 2005;130: 844e9. Dunn WB, Broadhurst D, Begley P, et al. Procedures for large-scale metabolic profiling of serum and plasma using gas chromatography and liquid chromatography coupled to mass spectrometry. Nat Protocols 2011;6: 1060e83. Want EJ, Wilson ID, Gika H, et al. Global metabolic profiling procedures for urine using UPLC-MS. Nat Protocols 2010;5:1005e18. Want EJ, Masson P, Michopoulos F, et al. Global metabolic profiling procedures for animal and human tissues via UPLC-MS. Nature Protocols 2013;8:17e32. Gika HG, Theodoridis GA, Wingate JE, et al. Withinday reproducibility of an HPLC-MS-based method for metabonomic analysis: application to human urine. J Proteome Res 2007;6:3291e3303. Gika HG, Macpherson E, Theodoridis GA, et al. Evaluation of the repeatability of ultra-performance liquid chromatography-TOF-MS for global metabolic profiling of human urine samples. J Chromatogr B 2009;871: 299e305. Michopoulos F, Lai L, Gika H, et al. UPLC-MS-based analysis of human plasma for metabonomics using solvent precipitation or solid phase extraction. J Proteome Res 2009;8:2114e21. Zelena E, Dunn WB, Broadhurst D, et al. Development of a robust and repeatable UPLCMS method for the long-term metabolomic study of human serum. Anal Chem 2009;81:1357e64. Wilson ID, Plumb R, Granger J, et al. HPLC-MS-based methods for the study of metabonomics. J Chrom B 2005; 817:67e76.
160
9. LIQUID CHROMATOGRAPHIC METHODS COMBINED WITH MASS SPECTROMETRY IN METABOLOMICS
18. Masson P, Alves AC, Ebbels TMD, et al. Optimization and evaluation of metabolite extraction protocols for untargeted metabolic profiling of liver samples by UPLC-MS. Anal Chem 2010;82:7779e86. 19. Masson P, Spagou K, Nicholson JK, et al. Technical and biological variation in UPLC-MS-based untargeted metabolic profiling of liver extracts: application in an experimental toxicity study on galactosamine. Anal Chem 2011;83:382e90. 20. Loftus N, Barnes A, Ashton S, et al. Metabonomic investigation of liver profiles of non-polar metabolites obtained from alcohol-dosed rats and mice using high mass accuracy MSn analysis. J Proteome Res 2011;10:705e13. 21. Plumb RS, Rainville P, Smith BW, et al. Generation of ultrahigh peak capacity LC separations via elevated temperatures and high linear mobile-phase velocities. Anal Chem 2006;78:7278e83. 22. Gika HG, Theodoridis G, Extance J, et al. High temperature-ultra performance liquid chromatographymass spectrometry for the metabonomic analysis of Zucker rat urine. J Chromatogr B 2008;871:279e87. 23. Rainville PD, Stumpf CL, Shockcor JP, et al. Novel application of reversed-phase UPLC-oa TOF-MS for lipid analysis in complex biological mixtures: a new tool for lipidomics. J Proteome Res 2007;6:552e8. 24. Castro-Perez JM, Kamphorst J, DeGroot J, et al. Comprehensive LC-MSE lipidomic analysis using a shotgun approach and its application to biomarker detection and identification in osteoarthritis patients. J Proteome Res 2010;9:2377e89. 25. Bird SS, Marur VR, Sniatynski MJ, et al. Lipidomics profiling by high-resolution LC-MS and high energy collisional dissociation fragmentation: Focus on characterisation of mitochondrial cardiolipins and monolysocardiolipins. Anal Chem 2011;83:940e9. 26. Fauland A, Kofeler H, Trotzmuller M, et al. A comprehensive method for lipid profiling by liquid chromatography-ion cyclotron resonance mass spectrometry. J Lipid Res 2011;11:2314e22. 27. Michopoulos F, Lai L, Gika H, et al. UPLC-MS-based analysis of human plasma for metabonomics using solvent precipitation or solid phase extraction. J Proteome Res 2009;8:2114e21. 28. Michopoulos F, Edge AM, Theodoridis G, et al. Application of turbulent flow chromatography to the metabonomics analysis of human plasma: comparison with protein precipitation. J Sep Sci 2010;33:1472e9. 29. Michopoulos F, Theodoridis G, Smith CJ, et al. Metabolite profiles from dried biofluids spots for metabonomics studies using UPLC combined with oaToF-MS. J Proteome Res 2010;9:3328e34. 30. Michopoulos F, Theodoridis G, Smith CJ. Metabolite profiles from dried blood spots for metabonomic
31.
32.
33.
34
35.
36.
37.
38.
39.
40.
41.
42.
43.
studies using UPLC combined with oaToF-MS: effects of different papers and sample storage stability. Bioanalysis 2011;3:2757e67. Idborg H, Zamani L, Edlund PO, et al. Metabolic fingerprinting of rat urine by LC/MS: Part 1. Analysis by hydrophilic interaction liquid chromatographyelectrospray ionization mass spectrometry. J Chromatogr B 2005;828:9e13. Gika HG, Theodoridis GA, Wilson ID. Hydrophilic interaction and reversed-phase ultra-performance liquid chromatography TOF-MS for metabonomic analysis of Zucker rat urine. J Sep Sci 2008;31:1598e608. Spagou K, Tsoukali H, Raikos N, et al. Hydrophilic interaction chromatography coupled to MS for metabonomic/metabolomic studies. J Sep Sci 2010;33: 716e27. Cubbon S, Bradbury T, Wilson J, et al. Hydrophilic interaction chromatography for mass spectrometric metabonomic studies of urine. Anal Chem 2007;79: 8911e8. Kim K, Aronov P, Zakharkin SO, et al. Urine metabolomics analysis for kidney cancer detection and biomarker discovery. Molecular & Cellular Proteomics 2009;8:558e70. Spagou K, Wilson ID, Masson P, et al. HILIC-UPLC-MS for exploratory urinary metabolic profiling in toxicological studies. Anal Chem 2011;83:382e90. Cai X, Dong J, Zou L, et al. Metabonomic study of lung cancer and the effects of radiotherapy on lung cancer patients: analysis of highly polar metabolites by ultraperformance HILIC coupled with Q-TOF MS. Chromatographia 2011;74:391e8. Fauland F, Kofeler H, Trotzmuller M, et al. A comprehensive method for lipid profiling by liquid chromatography-ion cyclotron resonance mass spectrometry. J Lipid Res 2011;52:2314e22. Yang Q, Shi X, Wang Y, Wang W, et al. Urinary metabonomic study of lung cancer by a fully automatic hyphenated hydrophilic interaction/RPLC-MS system. J Sep Sci 2010;33:1495e503. Lu W, Clasquin MF, Melamud E, et al. Metabolomic analysis via reversed-phase ion-pairing liquid chromatography coupled to a stand alone Orbitrap mass spectrometer. Anal Chem 2010;82:3212e21. Kiefer P, Delmotte N, Vorholt JA. Nanoscale ion-pair reversed-phase HPLC-MS for sensitive metabolome analysis. Anal Chem 2010;83:850e5. Coulier L, Bas R, Jespersen S, et al. Simultaneous quantitative analysis of metabolites using ion-pair liquid chromatography-electrospray ionisation mass spectrometry. Anal Chem 2006;78:6573e82. Matyska MT, Pesek JJ, Duley J, et al. Aqueous normal phase retention of nucleotides on silica hydride-based
REFERENCES
44.
45.
46.
47.
48.
49.
50.
51.
columns: method development strategies for analytes relevant in clinical analysis. J Sep Sci 2010;33:930e8. Tolstikov VT, Lommen A, Nakanishi K, et al. Monolithic silica-based capillary reversed-phase liquid chromatography/electrospray mass spectrometry for plant metabolomics. Anal Chem 2003;75:6737e40. Granger J, Plumb R, Castro-Perez J, et al. Metabonomic studies comparing capillary and conventional HPLC-oaTOF MS for the analysis of urine from Zucker obese rats. Chromatographia 2005;61:375e80. Shen Y, Zhang Y, Moore RJ, et al. Automated 20 kpsi RPLC-MS and MS/MS with chromatographic peak capacities of 10001500 and capabilities in proteomics and metabolomics. Anal Chem 2005;77:3090e100. Rainville PD, Smith NW, Wilson ID, et al. Addressing the challenge of limited sample volumes in in vitro studies with capillary-scale icrofluidic LC-MS/MS. Bioanalysis 2011;3:873e82. Loftus N, Miseki K, Iida J, et al. Profiling and biomarker identification in plasma from different Zucker rat strains via high mass accuracy multistage mass spectrometric analysis using liquid chromatography/mass spectrometry with a quadrupole ion trap-time of flight mass spectrometer. Rapid Comm Mass Spectrom 2008;22: 2547e54. Gika HG, Theodoridis GA, Earll M, et al. Does the mass spectrometer define the marker? A comparison of global metabolite profiling data generated simultaneously via UPLC-MS on two different mass spectrometers. Anal Chem 2010;82:8226e34. Katajamaa M, Miettinen J, Oresic M. MZmine: toolbox for processing and visualization of mass spectrometry based molecular profile data. Bioinformatics 2006;22:634e6. Lommen A. MetAlign: Interface-driven, versatile metabolomics tool for hyphenated full-scan mass
52.
53.
54. 55. 56.
57. 58. 59.
60.
61.
62.
161 spectrometry data reprocessing. Anal Chem 2011;81: 3079e86. Smith CA, Want EJ, O’Maille G, et al. XCMS: processing mass spectrometry data for metabolite profiling using nonlinear peak alignment, matching, and identification. Anal Chem 2006;78:779e87. Katajamaa M, Oresic M. Data processing for mass spectrometry-based metabolomics. J Chromatogr A 2007; 1158:318e28. METLIN (http://metlin.scripps.edu). Lipidmaps (http://www.lipidmaps.org). Wishart DS, Knox C, Guo AC, et al. HMDB: a knowledgebase for the human metabolome. Nucleic Acids Research 2009;37:D603e10. Human Metabolome Data Base (HMDB) (http://www. hmdb.ca). Chemspider (http://www.chemspider.com). Geenen S, Michopoulos F, Kenna JG, et al. HPLCMS/MS methods for the quantitative analysis of ophthalmic acid in rodent plasma and hepatic cell line culture medium. J Pharm Biomed Anal 2011;54: 1128e35. Geenen S, Guallar-Hoyas C, Michopoulos F, et al. HPLC-MS/MS methods for the quantitative analysis of 5-oxoproline (pyroglutamate) in rat plasma and hepatic cell line culture medium. J Pharm Biomed Anal 2011;56: 655e63. Geenen S, du Preez FB, Reed M. A mathematical modelling approach to assessing the reliability of biomarkers of glutathione metabolism. European J Pharm Sci 2011;46:233e43. Wijeyesekera A, Clarke PA, Bictash M, et al. Quantitative UPLC-MS/MS analysis of the gut microbial cometabolites phenylacetylglutamine, 4-cresyl sulphate and hippurate in human urine: INTERMAP study. Anal Methods 2012;4:65e72.