A rapid screening approach to metabonomics using UPLC and oa-TOFmass spectrometry: application to age, gender and diurnal variation innormal/Zucker obese rats and black, white and nude mice

Robert S. Plumb,*a Jennifer H. Granger,a Chris L. Stumpf,a Kelly A. Johnson,a Brian W. Smith,a

Scott Gaulitz,a Ian D. Wilsonb and Jose Castro-Perezc

Received 2nd February 2005, Accepted 5th April 2005

First published as an Advance Article on the web 27th April 2005

DOI: 10.1039/b501767j

The use of Ultra Performance Liquid Chromatography (UPLC), with a rapid 1.5 minute reversed-

phase gradient separation on a 1.7 mm reversed-phase packing material to provide rapid ‘‘high

throughput’’ support for metabonomic screening is demonstrated. The peak capacity and the

number of marker ions detected using these fast UPLC separations and oa-TOF MS was found to

be similar to that generated by conventional HPLC-MS methods with a 10 minute separation.

The UPLC-MS methodology was applied to the analysis of urine samples from rodents, including

normal and Zucker obese rats and three strains of mice (of both sexes), and was found to provide

rapid discrimination between age, strain, gender and diurnal variation.

Introduction

One of the major scientific challenges of the 21st century will

be determining the relationship between the human genome

and the risks of developing major diseases such as cancer,

diabetes, arthritis etc. Understanding the relationship between

the genome, the proteome and the expressed endogenous

metabolites (metabolic phenotyping or metabotyping1) requires

the development of ever more powerful analytical chemistry

techniques and data interpretation tools. To address metabolic

variation information rich analytical procedures have been

combined with multivariate statistical techniques, such as

principal components analysis (PCA), resulting in the develop-

ment of metabonomics.2

Initially metabonomics data were largely generated using

proton NMR spectroscopy2–5 However, the application of

high performance liquid chromatography combined with mass

spectrometry (LC/MS) to metabonomics studies is now

becoming more widely employed.6,7 From these exploratory

studies it is clear that LC/MS and LC/MS/MS, which is the

current analytical tool of choice for proteomics, quantitative

bioanalysis and metabolite identification,8–12 will also become

increasingly important for metabolite profiling. In metabo-

nomics there is a tension between the need for both com-

prehensive sample analysis and high sample throughput.

Whilst rapid analysis times have been achieved in quantitative

drug bioanalysis, where only a small number of analytes are

determined, this speed is obtained via lengthy optimization,

appropriate sample preparation and minimal chromatographic

resolution. The high selectivity of the mass spectrometer

operating in multiple reaction monitoring (MRM) mode is

relied upon to improve the specificity of such rapid assays. The

lack of chromatographic resolution of these rapid LC/MS/MS

analysis systems makes this approach unacceptable for

metabonomic analysis where both the maximum possible

chromatographic resolution of a complex mixture and a full

mass range are required.

An obvious method for achieving these aims is to improve

chromatographic resolution, which can be done by employing

small particle size chromatographic stationary phases down to

and including sub-2 mm particles coupled to very high operat-

ing pressures.13–17 The use of these small particles was pioneered

by Jorgenson, principally in the analysis of peptides13–15 where

he used capillary scale separations with sub-2 mm particles to

affect ultrahigh resolution chromatography. However, the use

of columns of larger format (and correspondingly higher

sample loadings) results in the generation of high back

pressures requiring the use of specialized solvent delivery

equipment capable of operating at pressures well above those

of conventional HPLC to ensure reasonable flow rates.

This work has led to the development and commercializa-

tion of ‘‘Ultra Performance Liquid Chromatography’’ (UPLC)

which employs sub-2 mm stationary phase particles and high

linear mobile phase velocities to generate high resolution

chromatograms with very short analysis times. An example of

its utility was recently provided by Perez et al at a microbore

scale (2.1 and 1 mm ID) for the analysis of Midazolam

metabolites in rat bile;18 here the extra resolution and

sensitivity of UPLC resulted in the facile separation of two

conjugated metabolites in just 6 min, previously unresolved by

HPLC using a 30 min analysis time.

Here the use of UPLC, coupled to oa-TOF MS for the

metabonomic analysis of rodent urine samples, using a 1.5 min

separation combined with pattern recognition techniques, is

described. The methodology was used for the investigation

of the urinary metabolic fingerprints of male (fa/fa) Zucker

rats and male Alderely Park (AP) (Wistar-derived) rats as well

as male and female black, white and nude mice.

Experimental

Chemicals

Fisher Optima acetonitrile (HPLC grade) was purchased from

Fisher Scientific (Hampton, NH, USA), ammonium formate

PAPER www.rsc.org/analyst | The Analyst

844 | Analyst, 2005, 130, 844–849 This journal is � The Royal Society of Chemistry 2005

and formic acid (spectroscopic grade) were purchased

from Sigma/Aldrich (MO, USA). Distilled water was purified

‘‘in-house’’ using a MilliQ system Millipore (MA, USA).

Leucine-enkephalin was obtained from Sigma-Aldrich

(MO, USA).

Animal studies

Urine samples were collected from male Wistar-derived (AP)

and obese (fa/fa) Zucker rats (BABU, Alderley Park) aged 6

and 20 weeks. Samples were stored at 220 uC. The animals

were housed in polypropylene cages and allowed free access to

water and food.

Samples were collected from black (C57BL19J) white

(Alpk:ApfCD) and nude male and female mice (n 5 10) at

different times of the day (am and pm) by minimal

manipulation of the lower abdomen as described elsewhere.8,9

Samples were stored at 220 uC prior to analysis. Prior to the

study animals were housed in polycarbonate solid bottom

cages, according to strain. Animals were allowed free access

to water and food (pelleted irradiated rat and mouse diet 1)

(Special Diet Services, Witham, Essex, UK), from weaning

until the end of the experiment. Animals were maintained at

room temperature with artificial 12 h dark/12 h light cycles.

Prior to analysis the samples were centrifuged at 13 000 rpm

and a 100 mL aliquot of the supernatant was diluted 1:4 with

distilled water and transferred to a total recovery auto-sampler

vial for analysis by UPLC-MS.

Chromatography

Chromatographic separations were performed on a 50 62.1 mm ACQUITY2 1.7 mm C18 column (Waters Corp,

Milford, USA) using an ACQUITY2 Ultra Performance

Liquid Chromatography system (Waters). The column was

maintained at 40 uC and eluted with a bilinear gradient of

0–20% A (0–30 s) and then 20–95% (30–60 s) B where A 5 0.1%

formic acid and B 5 acetonitrile 0.1% formic acid. The

gradient duration was 1 min at a flow rate of 1 mL min21. The

column eluent was directly split such that y150 mL min21 was

directed to the mass spectrometer. A 10 mL injection of each

sample was made onto the column.

Mass spectrometry

Mass spectrometry was performed on a Micromass LCT

Premier2 (Waters MS Technologies, Manchester, UK) ortho-

gonal acceleration time of flight mass spectrometer operating

in positive ion mode. The nebulization gas was set to 550 L h21

at a temperature of 200 uC the cone gas set to 0 L h21 and the

source temperature set to 110 uC. The capillary voltage and

the cone voltage were set to 3200 and 100 V, respectively. The

LCT-Premier2 was operated in W optics mode with 12 000

resolution. The data acquisition rate was set to 0.1 s, with a

0.05 s interscan delay using dynamic range enhancement

(DRE); data were collected for 1.5 min. All analyses were

acquired using the lock spray to ensure accuracy and

reproducibility; leucine-enkephalin was used as the lock mass

(m/z 556.2771) at a concentration of 50 fmol mL21 and flow

rate 30 mL min21. Data were collected in centroid mode from

100–950 m/z with a lockspray frequency of 11 s, and data

averaging over 10 scans.

Chemometric analysis

UPLC-MS data were divided into subsets for analyses to assess

strain and gender differences, as well as diurnal variation. The

raw data were analysed by the Micromass MarkerLynx

applications manager Version 1.0 (Waters, UK); this applica-

tion manager integrates peaks in the LC/MS data by using

ApexTrack2 peak detection. The LC/MS data are peak-

detected and noise-reduced in both the LC and MS domains

such that only true analytical peaks are further processed by

the software (e.g., noise spikes are rejected). A list of the

intensities of the peaks detected is then generated for the first

sample, using the retention time (RT) and m/z data pairs as the

identifier for each peak. An arbitrary number is then assigned

to each of these RT–m/z pairs in order of elution, (1, 2, 3, 4,….

etc). This process is repeated for each run; once this is

completed the data from each LC/MS analysis in the batch

are then sorted such that the correct peak intensity data for

each RT–m/z pair is aligned in the final data table. The ion

intensities for each peak detected are then normalized, within

each sample, to the sum of the peak intensities in that sample.

The resulting normalized peak intensities are then multiplied

by 10 000. The resulting three-dimensional data, peak number

(RT–m/z pair), sample name, and ion intensity were analysed

by principal components analysis (PCA). The resulting data

was exported to the Simca-P+ Software package (Umetrics,

Umea, Sweden) for subsequent processing by Partial Least

Square Discriminant Analysis PLS-DA.

Results and discussion

We have previously demonstrated the ability of urinalysis by

HPLC-MS to discriminate between strains and gender, and to

reveal differences due to diurnal variation, using the mouse as

a model system.6 Those mouse studies however, were based on

a 10 min analysis time and, at best, provided a ‘‘medium

throughput’’ means of metabonomic profiling. For the present

study we have investigated urine samples from two rat and

three mouse strains (see later). The rat urine samples were

obtained from either male Zucker obese (fa/fa or +/+) or male

Wistar-derived (AP) animals. The Zucker obese rat is impor-

tant as it represents a widely used rodent model of obesity and

insulin-resistance of the type that can lead to the development

of type II diabetes.19,20 The data presented in Fig. 1 show

the positive ion total ion chromatograms (TIC) of the urine

from a typical male Zucker +/+ animal obtained on samples

from 6- and 20-week old rats (Fig. 1(b) and (d)) together with

the equivalent samples from a normal male animal (Fig. 1(a)

and (c)). Visual examination of the TICs for week 6 shows

significantly more peaks in the AP (normal) animals sample

compared to that for the Zucker +/+ animal, with major

differences in ion intensities occurring between 0.6 and 0.8 min.

The week 20 samples also showed a marked difference between

the Zucker +/+ animals and the normal AP animals. The

Zucker +/+ animals show a prominent peak at 0.8 which was

detected at a lower level in the normal animals. Conversely the

control animal sample obtained for week 20 showed a very

This journal is � The Royal Society of Chemistry 2005 Analyst, 2005, 130, 844–849 | 845

large peak at 0.5 min which was absent in the equivalent

Zucker animal sample. The average peak width for com-

ponents detected in these urine samples was determined to be

1 s giving a peak capacity of 60, which is comparable to

previously reported HPLC/MS metabonomics results obtained

using a 10 min analysis time21 and a conventional chromato-

graphic column. In order to process the data it is essential to

have a good linear range for detection and highly reproducible

chromatography. To ensure a wide linear range data was

collected using the LCT-Premier time of flight mass spectro-

meter in dynamic range enhancement mode (DRE) giving a

linear range of four orders of magnitude. The reproducibility

of the chromatography was determined by evaluating the

retention time variation of the m/z 5 180 (hippuric acid), 206

(xanthurate) and 165 (phenylaniline) ions. The average reten-

tion time variation was determined to be less that 0.2% RSD.

Fig. 1 UPLC/MS total ion current (TIC) resulting from the injection

of AP and Zucker rat urine from weeks 6 and 20 onto a 2.1 6 50 mm

1.7 mm ACQUITY C18 column. The column was eluted over 1 min

with a 0–95% aqueous formic acid-acetonitrile gradient at 1 mL min21.

The column effluent was split such that 150 mL min21 entered the

MS. The MS was operated in positive ion mode over the mass range

50–850 m/z.

Fig. 2 The scores (a) and loadings plot (b) from the PLS-DA analysis

of the LC/MS data generated from the AP and Zucker rats on weeks 6

and 20.

Fig. 3 The extracted ion chromatograms of m/z 5 255 and 332 for the Zucker rats on weeks 6 and 20

846 | Analyst, 2005, 130, 844–849 This journal is � The Royal Society of Chemistry 2005

The UPLC/MS data obtained for these rat urine samples

was subjected to peak detection and integration via the

MarkerLynx program. Partial least squares-discriminant

analysis (PLS-DA) was then performed on an exported peak

list from MarkerLynx. The resulting scores plot from PLS-DA

of the UPLC/MS data is displayed in Fig. 2(a). In this figure

we can clearly see that the four groups of samples are clearly

separated. It is also apparent that there is more variance in the

Zucker obese animals at 20 weeks than observed at 6 weeks.

The number of marker ions detected from the 1.5 min UPLC

analysis was in the region of 3900. This compares favourably

with the value of 1000 ions detected in urine samples for the

Fig. 4 The scores (a) and loadings plot (b) from the PCA analysis of the LC/MS data generated from the male black mice showing the AM-PM

variation. The scores (c) and loadings plot (d) from the PCA analysis of the LC/MS data generated from the black, white and nude mice, (e) shows

the PCA analysis of the LC/MS data from the white female and male data (AM & PM).

This journal is � The Royal Society of Chemistry 2005 Analyst, 2005, 130, 844–849 | 847

previously reported study by Haselden and Plumb employing a

10 min HPLC separation.21

The data displayed in Fig. 2(b) is the resulting loadings

plot from the PLS-DA analysis. The ions responsible for

the clustering are clearly visible in this figure; with the ions

m/z 5 196, 328, 255. 332 and 525 contributing strongly to the

position of the of the 20-week old Zucker animals, whereas

the ions m/z 5 457, 285 and 255 contributed more strongly to

the position of the week 6 Zucker animals. The position of the

6-week old AP rats was determined by the ion m/z 5 255. A

visual comparison between the extracted ion chromatograms

for m/z 5 255 and 332 for the Zucker animal on weeks 6 and

20 shows a clear difference in the intensities of the peaks

as shown in Fig. 3. From the data we can see that the con-

centration of the m/z 5 255 was reduced in the Zucker week 20

urines compared to that of the control animals for both week 6

and 20 samples and the Zucker rat sample for week 6. The

opposite is observed for the m/z 5 332 ion where the con-

centration was increased in week 20 Zucker urines compared

to the other samples.

Biochemical differences due to sampling time, gender and

strain all offer the potential for the introduction of confound-

ing factors with the potential for the misinterpretation of the

data and false conclusions. It is essential to account for these

effects and eliminate them prior to drawing any biological

inferences from the data. Some of this variability can be seen

in the data obtained from examining samples from mice as

reported previously using HPLC-MS with a 10 min run time.21

The data shown in Fig. 4(a) and (c) illustrate the diurnal

variation in the positive ion UPLC/MS data obtained on the

urine samples obtained from the black (C57BL19J) male mice

using the analytical conditions described above for the rat

samples. This figure clearly illustrates the ability of UPLC/MS

and chemometric techniques to differentiate between urine

samples collected in the morning and those collected in the

evening. In the scores and loading plot we can easily observe

two separated sample clusters. The ions contributing most to

the variance seen in the data appear to be at m/z 5 259, 297,

206 (Fig. 4(c)).

The data in Fig. 4(d) and (e) illustrate the variation observed

between the black C57BL19J, white Alpk:ApfCD and nude

mice, the best clustering was obtained using PC2 vs. PC3.

In this data set we can see a clear resolution of the normals

from the nude strain. The ions contributing most to the

clustering are given in the loading plot and were determined

to be, m/z 5 255, 131, 345, 267 and 206. In this complex

data there are three major variables, strain, sampling time

(AM-PM) and gender. Here it is possible to visualize the strain

variation, overlaid with other biological differences caused by

gender and diurnal effects which are masked to an extent by

the more dominant strain differences. To help visualize the

gender variation the mouse-derived data was further sub-

divided in to strains and processed separately. The results

are displayed in Fig. 4(e), and illustrate the ability of the

methodology to identify the gender variation in the white

mouse data. Here we can see that the male and female animal

samples are clearly separated in to two distinct groups using

PC2 vs. PC3. The ion m/z 5 206 was also found to contribute

significantly to this clustering. This ion has been previously

determined to be 4,8-dihydroxyquinoline-2-carboxylic acid,

part of the tryptophan catabolism pathway.9

In this instance prior knowledge enabled us to rapidly

identify one of the important marker molecules present in the

urine of these animals. In situations where identity is not so

easily determined the accurate mass derived from TOF/MS

can be used to derive elemental compositions which can be

used to narrow the field, enabling a significant number of

potential candidates to be eliminated, whilst MS/MS experi-

ments can be used to narrow the field further (hopefully to a

point where a comparison with an authentic standard can

be contemplated). Where an unknown compound remains

refractory to this approach mass directed fraction collection

can be used to provide sufficient material for further spectro-

scopic characterisation by NMR spectroscopy etc.

Conclusion

The data generated in this study clearly shows that UPLC can

be used with a rapid 1.5 min separation to support high-

throughput metabonomics screening activities. The peak

capacity and the number of peaks detected using these fast

UPLC gradients and oa-TOF MS was similar to, or better

than, that generated with a 10 min HPLC separation. This

rapid LC/MS analysis approach to metabonomics allows an

overall view of any variation or biochemical changes in the

sample to be rapidly acquired. Once the whole sample set has

been screened for any gross metabolic difference in the data

set, a smaller subset of the samples can then be correctly

selected and subjected to more comprehensive analysis to

determine the entire biomarker set.

Robert S. Plumb,*a Jennifer H. Granger,a Chris L. Stumpf,a

Kelly A. Johnson,a Brian W. Smith,a Scott Gaulitz,a Ian D. Wilsonb andJose Castro-Perezc

aWaters Corporation, 34 Maple Street, Milford, MA 01757, USAbAstraZeneca, Dept. of Drug metabolism and Pharmacokinetics,Mereside, Alderley Park, Macclesfield, Cheshire, UK SK10 4TGcWaters Corporation, Floats Rd, Wythenshawe, Manchester, UKM23 9LZ. E-mail: Rob_Plumb@Waters.com

References

1 J. K. Nicholson, J. Connelly, J. C. Lindon and E. Holmes, Nat.Rev. Drug Disc., 2002, 1, 153–161.

2 E. Holmes and H. Antti, Analyst, 2002, 127, 1549–1557.3 J. K. Nicholson, J. C. Lindon and E. Holmes, Xenobiotica, 1999,

29, 1181–1189.4 D. G. Robertson, M. D. Reilly, J. C. Lindon, E. Holmes and

J. K. Nicholson, Compr. Toxicol., 2002, 585–610.5 E. M. Lenz, J. Bright, I. D. Wilson, S. r. Morgan and A. F. Nash,

J. Pharm. Biomed. Anal., 2004, 33, 1103–1115.6 R. S. Plumb, J. H. Granger, C. L. Stumpf, I. D. Wilson, J. A. Evans

and E. M. Lenz, Analyst, 2003, 128, 819–823.7 H. Idborg-Bjorkman, P. O. Edlund, O. M. Kvalheim, I. Schuppe-

Koistinen and S. P. Jacobsson, Anal. Chem., 2003, 75, 4784–4792.8 G. J. Dear, R. S. Plumb, B. C. Sweatman, P. S. Parry, A. D. Robert,

J. C. Lindon, J. K. Nicholson and I. M. Ismail, J Chromatogr., B,2000, 748, 295–309.

9 E. M. Lenz, J. Bright, R. Knight, I. D. Wilson and H. Major,Analyst., 2004, 129, 535–541.

10 C. E. Hop, P. R. Tiller and L. Romanyshyn, Rapid Commun. MassSpectrom., 2002, 16, 212–219.

11 J. Ayrton, R. A. Clare, G. J. Dear, D. N. Mallett and R. S. Plumb,Rapid Commun. Mass Spectrom., 1999, 13, 1657–1662.

848 | Analyst, 2005, 130, 844–849 This journal is � The Royal Society of Chemistry 2005

12 R. S. Plumb, G. J. Dear, D. N. Mallett and J. Ayrton, RapidCommun. Mass Spectrom., 2001, 15, 986.

13 J. E. MacNair, K. C. Lewis and J. W. Jorgenson, Anal. Chem.,1997, 69, 983–989.

14 K. Lan and J. W. Jorgenson, Anal. Chem., 1999, 71, 709–714.15 J. S. Mellors and J. W. Jorgenson, Anal. Chem., 2004, 76, 5441–5450.16 J. A. Lippert, B. Xin, N. Wu and M. L. Lee, J. Microcolumn Sep.,

1997, 11, 631–643.17 Y. Xiang, B. Yan, C. V. McNeff, P. W. Carr and M. L. Lee,

J. Chromatogr., A, 2003, 1002, 71–78.

18 R. S. Plumb, J. Castro-Perez, J. H. Granger, I. Beattie, K. Joncourand A. Wright, Rapid Commun. Mass Spectrom., 2004, 18,2331–2337.

19 L. M. Zucker and H. N. Antoniades, Endocrinology, 1972,1320–1330.

20 H. King, R. E. Aubert and W. H. Herman, Diabetes Care, 1998,1414–1431.

21 J. N. Haselden, R. S. Plumb, C. L. Stumpf, J. H. Granger,J. Castro-Perez and G. J. Dear, Rapid Commun. Mass Spectrom.,2003, 17, 2632–2638.

This journal is � The Royal Society of Chemistry 2005 Analyst, 2005, 130, 844–849 | 849