Kinetic and J-resolved statistical total correlation NMR spectroscopy approaches to structural information recovery in complex reacting mixtures: Application to acyl glucuronide intramolecular transacylation reactions

Article abstract of DOI:10.1021/ac702614t

We demonstrate here a new variant on a statistical spectroscopic method for recovering structural information on unstable intermediates formed in reaction mixtures. We exemplify this approach with respect to the internal acyl migration reactions of 1-β-O-acyl glucuronides (AGs), which rearrange at neutral or slightly alkaline pH on a minute to hour time scale to yield a series of positional glucuronide ring isomers and α/β anomers from the 1-β (starting material), i.e. 2-β, 2-α, 1-α, 3-β, 3-α, and 4-β, 4-α isomers together with the aglycon and α- and β-glucuronic acid hydrolysis products. Multiple sequential 800 MHz cryoprobe ¹H NMR spectra (1D and 2D J-resolved, JRES) were collected on a 5.1 mM solution of a synthetic model drug glucuronide, 1-β-O-acyl (S)-α-methyl phenylacetyl glucuronide (MPG) in 0.1 M sodium phosphate buffer in D₂O at pD 7.4 over 18 h to monitor the reaction which leads to the formation of the eight positional isomers and hydrolysis products. As the reaction proceeds and new isomers form, the NMR signal intensities vary accordingly allowing the application of a novel kinetic variant on statistical total correlation spectroscopy (K-STOCSY) method to recover the connectivities between proton signals on the same reacting molecule based on their intensity covariance through time. We performed K-STOCSY analysis on both the standard ID NMR spectra and the skyline projected singlets of the ¹H- ¹H JRES NMR spectra through time, i.e. the K-JRES-STOCSY experimental variant, which increases the effective spectral dispersion and is ideally suited for the analysis of heavily overlapped spin systems. High statistical correlations were observed between mutarotated α-and β-anomers of individual positional isomers, as well as directly acyl migrated products and anticorrelation observed between signals from compounds that were being depleted as others increased, e.g. between the 1-β and 2-α/2-β isomers. This statistical kinetic approach enabled the recovery of structural connectivity information on all isomers allowing unequivocal resonance assignment, and this approach to spectroscopic information recovery has wider potential uses in the study of reactions that occur on the second-to-minute time scale in conditions where multiple sequential NMR spectra can be collected. JRES-STOCSY is also of potential use as a method for recovering spectroscopic information in highly overlapped NMR signals and spin systems in other types of complex mixture analysis.

Full text of DOI:10.1021/ac702614t

Anal. Chem. 2008, 80, 4886–4895
Kinetic and J-Resolved Statistical Total
Correlation NMR Spectroscopy Approaches to
Structural Information Recovery in Complex
Reacting Mixtures: Application to Acyl Glucuronide
Intramolecular Transacylation Reactions
Caroline H. Johnson,† Toby J. Athersuch,† Ian D. Wilson,‡ Lisa Iddon,§ Xiaoli Meng,§
Andrew V. Stachulski,§ John C. Lindon,† and Jeremy K. Nicholson*,†
Department of Biomolecular Medicine, Division of Surgery, Oncology, Reproductive Biology and Anaesthetics (SORA),
Faculty of Medicine, Sir Alexander Fleming Building, Imperial College London, South Kensington, London SW7 2AZ,
U.K., Department of Drug Metabolism and Pharmacokinetics, AstraZeneca, Macclesfield, Cheshire SK10 4TG, U.K.,
and Department of Chemistry, The Robert Robinson Laboratories, University of Liverpool, Liverpool L69 7ZD, U.K.
We demonstrate here a new variant on a statistical
spectroscopic method for recovering structural informa-
tion on unstable intermediates formed in reaction mix-
tures. We exemplify this approach with respect to the
internal acyl migration reactions of 1-ꢀ-O-acyl glucu-
ronides (AGs), which rearrange at neutral or slightly
alkaline pH on a minute to hour time scale to yield a series
of positional glucuronide ring isomers and r/ꢀ anomers
from the 1-ꢀ (starting material), i.e. 2-ꢀ, 2-r, 1-r, 3-ꢀ,
3-r, and 4-ꢀ, 4-r isomers together with the aglycon and
r- and ꢀ-glucuronic acid hydrolysis products. Multiple
sequential 800 MHz cryoprobe 1H NMR spectra (1D and
2D J-resolved, JRES) were collected on a 5.1 mM solution
of a synthetic model drug glucuronide, 1-ꢀ-O-acyl (S)-r-
methyl phenylacetyl glucuronide (MPG) in 0.1 M sodium
phosphate buffer in D2O at pD 7.4 over 18 h to monitor
the reaction which leads to the formation of the eight
positional isomers and hydrolysis products. As the reac-
tion proceeds and new isomers form, the NMR signal
intensities vary accordingly allowing the application of a
novel kinetic variant on statistical total correlation spec-
troscopy (K-STOCSY) method to recover the connectivities
between proton signals on the same reacting molecule
based on their intensity covariance through time. We
performed K-STOCSY analysis on both the standard 1D
NMR spectra and the skyline projected singlets of the
1H-1H JRES NMR spectra through time, i.e. the K-JRES-
STOCSY experimental variant, which increases the effec-
tive spectral dispersion and is ideally suited for the
analysis of heavily overlapped spin systems. High statisti-
cal correlations were observed between mutarotated r-
and ꢀ-anomers of individual positional isomers, as well
as directly acyl migrated products and anticorrelation
observed between signals from compounds that were
being depleted as others increased, e.g. between the 1-ꢀ
and 2-r/2-ꢀ isomers. This statistical kinetic approach
enabled the recovery of structural connectivity information
on all isomers allowing unequivocal resonance assign-
ment, and this approach to spectroscopic information
recovery has wider potential uses in the study of reactions
that occur on the second-to-minute time scale in condi-
tions where multiple sequential NMR spectra can be
collected. JRES-STOCSY is also of potential use as a
method for recovering spectroscopic information in highly
overlapped NMR signals and spin systems in other types
of complex mixture analysis.
Two-dimensional (2D) nuclear magnetic resonance (NMR)
spectroscopic methods are invaluable molecular structural as-
signment tools, and can readily be used to identify nuclei from
within the same scalar J-coupled spin-system or from nuclei close
in space via dipolar interactions. For less sensitive nuclei or for
dilute samples, long total acquisition times can be required for
some 2D experiments, for example a 1H-1H 2D total correlation
spectrum (TOCSY) data set can take several hours to acquire (at
high resolution in F1) on biological samples. In dynamic reactions,
real-time NMR observation of the degradation process and
structural identification of the products can be nontrivial as the
changing speciation results in low sensitivity. One such example
of a spontaneous reacting metabolic system is provided by the
ester glucuronides of drugs.
Many drugs are metabolized to acyl glucuronides (AGs), and
these have been studied extensively, because of their possible
reactivity toward proteins and consequent immunotoxicity.1–4 The
1-ꢀ-O AG is formed initially as a result of conjugation via UDP-
glucuronosyltransferase that transfers a glucuronic acid moiety
* To whom correspondence should be addressed. Tel: +44(0)2075943195.
Fax: +44(0)2075944336. E-mail: j.nicholson@imperial.ac.uk.
† Imperial College London.
(1) Spahn-Langguth, H.; Benet, L. Z. Drug Metab. Rev. 1992, 24, 5–48
(2) Mulder, G. J. Annu. Rev. Pharmacol. Toxicol. 1992, 32, 25–49
(3) Faed, E. M. Drug Metab. Rev. 1984, 15, 1213–1249
(4) Fenselau, C. In Conjugation-deconjugation reactions in drug metabolism and
toxicity; Kauffman, F. C., Ed.; Spring-Verlag: Berlin, 1994; pp 367-389.
.
.
.
‡ AstraZeneca.
§ University of Liverpool.
.
4886 Analytical Chemistry, Vol. 80, No. 13, July 1, 2008
10.1021/ac702614t CCC: $40.75 2008 American Chemical Society
Published on Web 05/10/2008

Scheme 1. Rearrangement Reactions of Acyl Glucuronidesa
a All isomers can hydrolyze, but the 1ꢀ isomer has the highest rate and contributes most to the aglycon pool.
to a carboxylate group of the drug. This type of metabolic phase
II conjugation is common in nonsteroidal anti-inflammatory drugs
(NSAIDs) where many commonly used drugs contain a carboxy-
late group.5 The resulting 1-ꢀ-O AG can then undergo a cascade
of spontaneous chemical reactions in aqueous solution, resulting
in acyl migration (through ortho-ester intermediate structures),
mutarotation (via ring-opened aldehyde forms) and hydrolysis,
to produce seven further positional isomers: 2-ꢀ, 2-R, 1-R, 3-ꢀ, 3-R,
4-ꢀ, 4-R the aglycon and R- and ꢀ-glucuronic acid as shown in
Scheme 1. The rates at which these reactions proceed are a
function of the aglycon structure6,7 (steric and electronic features),
the pH8 and ionic strength of the solution.1 The positional isomers
can also bind irreversibly to endogenous proteins, and the
formation of protein adducts by acyl migration can potentially
result in immunopathological reactions.1,3 Several NSAIDs (tol-
metin, zomepirac, ibufenac) have been withdrawn from the
market, due to different types of adverse reactions.9,10 Since these
withdrawn drugs all have fast AG degradation rates (0.26 h-1
,
0.45 h-1, 1.1 h-1 8,11,12
for the 1-ꢀ-O-AG in aqueous solutions at
)
pH 7.4, and also exhibit protein binding, tentative links have been
made between AG toxicity and the AG degradation rate. However,
no clear connection has yet been made between simple AG
degradation rates and the incidence of reported toxicity,13 and it
is likely that a complex interplay between AG transacylation rates
and pharmacokinetic processes is involved.
(5) Gibson, G. G.; Skett, P. Introduction to drug metabolism, 3rd ed.; Nelson
Thornes: Cheltenham, U.K, 2001.
(6) Nicholls, A. W.; Akira, K.; Lindon, J. C.; Farrant, R. D.; Wilson, I. D.; Harding,
(9) Samuel, S. A. N. Engl. J. Med. 1981, 304, 978
(10) Herxheimer, A. Drug Ther. Bull. 1968, 6, 48
(11) Castillo, M.; Smith, P. C. Drug Metab. Dispos. 1995, 23, 566–572
(12) Hyneck, M. L.; Smith, P. C.; Munafo, A.; McDonagh, A. F.; Benet, L. Z.
Clin. Pharmacol. Ther. 1988, 44, 107–114
(13) Stachulski, A. V.; Harding, J. R.; Lindon, J. C.; Maggs, J. L.; Park, B. K.;
Wilson, I. D. J. Med. Chem. 2006, 49, 6931–6945
.
.
J.; Killick, D. A.; Nicholson, J. K. Chem. Res. Toxicol. 1996, 9, 1414–1424
(7) Vanderhoeven, S. J.; Troke, J.; Tranter, G. E.; Wilson, I. D.; Nicholson,
J. K.; Lindon, J. C. Xenobiotica 2004, 34, 889–900
(8) Hasegawa, J.; Smith, P. C.; Benet, L. Z. Drug Metab. Dispos. 1982, 10,
469–473
.
.
.
.
.
.
Analytical Chemistry, Vol. 80, No. 13, July 1, 2008 4887

Figure 1. NMR spectra for the (S)-R-methyl phenylacetic AG. (a) 1H NMR spectrum. (b) A 1H JRES skyline projection on to the F2 axis. (c)
A
1H-1H 2D JRES NMR spectrum.
1H NMR spectroscopy has been applied successfully to study
heavily overlapped even at high observation frequencies,27 such
that it is not possible to assign many of the peaks to the positional
isomers in partially equilibrated samples using one-dimensional
(1D) NMR methods. Kinetic studies have also been performed
using 19F NMR spectroscopy for fluorinated drugs,6 but this gives
little direct structural information, without some form of correlated
1H NMR data. 13C NMR spectroscopy has also been applied to
study glucuronide transacylation kinetics, but at only 1% 13C
natural abundance there are insensitivity problems leading to long
spectral acquisition times (usually many hours) even in 1D
experiments and consequently few time-points can be analyzed.28,29
An alternative approach has been to apply liquid chromatography
(LC)-NMR to chromatographically resolve individual isomers,
which can then be both structurally assigned and held in the NMR
probe for individual kinetic experiments;24,26,30 however, full
chromatographic resolution is not always possible with such
structurally similar compounds depending on the aglycon structure.
the degradation of 1-ꢀ-O AGs 1,14–26 but resolution and identifica-
tion of the isomers can be difficult in partly equilibrated or
equilibrium mixtures. With the exception of the anomeric proton
signals, the glucuronide ring protons are highly coupled and
(14) Bischer, A.; Zia-Amirhosseini, P.; Iwaki, M.; McDonagh, A. F.; Benet, L. Z.
J Pharmacokinet. Biopharm. 1995, 23, 379–395
(15) Ruelius, H. W.; Young, E. M.; Kirkman, S. K.; Schillings, R. T.; Sisenwine,
S. F.; Janssen, F. W. Biochem. Pharmacol. 1985, 34, 451–452
(16) Mortensen, R. W.; Sidelmann, U. G.; Tjornelund, J.; Hansen, S. H. Chirality
2002, 14, 305–312
(17) Volland, C.; Sun, H.; Dammeyer, J.; Benet, L. Z. Drug Metab. Dispos. 1991,
19, 1080–1086
(18) Benet, L. Z.; Spahn-Langguth, H.; Iwakawa, S.; Volland, C.; Mizuma, T.;
Mayer, S.; Mutschler, E.; Lin, E. T. Life Sci. 1993, 53, PL141-146
(19) Spahn, H.; Iwakawa, S.; Lin, E. T.; Benet, L. Z. Pharm. Res. 1989, 6, 125–
132
(20) Spahn, H. J. Chromatogr. 1988, 430, 368–375
(21) Smith, P. C.; Liu, J. H. Xenobiotica 1993, 23, 337–348
.
.
.
.
.
.
.
.
(22) Gavaghan, C. Ph.D. Thesis, Birkbeck College, University of London, 1997.
(23) Vanderhoeven, S. J.; Lindon, J. C.; Troke, J.; Nicholson, J. K.; Wilson, I. D.
(27) Farrant, R. D.; Spraul, M.; Wilson, I. D.; Nicholls, A. W.; Nicholson, J. K.;
J. Pharm. Biomed. Anal. 2006, 41, 1002–1006
(24) Sidelmann, U. G.; Hansen, S. H.; Gavaghan, C.; Carless, H. A. J.; Lindon,
.
Lindon, J. C. J. Pharm. Biomed. Anal. 1995, 13, 971–977
(28) Akira, K.; Taira, T.; Hasegawa, H.; Sakuma, C.; Shinohara, Y. Drug Metab.
Dispos. 1998, 26, 457–464
(29) Akira, K.; Taira, T.; Shinohara, Y. J. Pharmacol. Toxicol. Methods 1997,
37, 237–243
.
J. C.; Farrant, R. D.; Wilson, I. D.; Nicholson, J. K. Anal. Chem. 1996, 68,
.
2564–2572
(25) Johnson, C. H.; Wilson, I. D.; Harding, J. R.; Stachulski, A. V.; Iddon, L.;
Nicholson, J. K.; Lindon, J. C. Anal. Chem. 2007, 79, 8720–8727
(26) Corcoran, O.; Mortensen, R. W.; Hansen, S. H.; Troke, J.; Nicholson, J. K.
Chem. Res. Toxicol. 2001, 14, 1363–1370
.
.
.
(30) Mortensen, R. W.; Corcoran, O.; Cornett, C.; Sidelmann, U. G.; Lindon,
J. C.; Nicholson, J. K.; Hansen, S. H. Drug Metab. Dispos. 2001, 29, 375–
380.
.
4888 Analytical Chemistry, Vol. 80, No. 13, July 1, 2008

Figure 2. (a) A series of 1D 1H spectra at specific time-points for the glucuronide ring proton resonances. (b) A series of 2D 1H J-RES skyline
projections at the same specific time-points for the glucuronide ring proton resonances.
Developments in cryoprobe NMR spectroscopy have permitted
much greater real-time sensitivity for most NMR experiments,31
and the use of higher magnetic fields gives greater frequency
dispersion to aid structure elucidation in complex mixtures.32 The
2D JRES NMR experiment is one of the few 2D approaches that
can be performed rapidly (∼10 min) with high sensitivity as it
requires only a relatively small number of F1 increments to enable
effective dispersion of the J-couplings and patterns.33 Tilting and
symmetrizing the JRES spectrum after the second Fourier
transform allows projection of the spectrum on to the F2 axis
giving what is effectively a proton-decoupled proton spectrum
(singlets only), which is ideal for studying heavily overlapped spin-
systems and gives resolution enhancement through 2D signal
processing (use of sine-bell apodization function) and also virtual
dispersion improvement.
We have shown previously the value of the JRES approach
for studies on small molecule studies in complex biofluids, the
1H NMR spectra of which contain multiple overlapped signals from
many metabolites.1,34–38 However, we have been seeking a more
(31) Lindon, J. C.; Holmes, E.; Nicholson, J. K. FEBS J. 2007, 274, 1140–1151
(32) Keun, H. C.; Beckonert, O.; Griffin, J. L.; Richter, C.; Moskau, D.; Lindon,
J. C.; Nicholson, J. K. Anal. Chem. 2002, 74, 4588–4593
(33) Aue, W. P.; Karhan, J.; Ernst, R. R. J. Chem. Phys. 1976, 64, 4226–4227
.
.
(34) Tomlins, A. M.; Foxall, P. J. D.; Lynch, M. J.; Parkinson, J.; Everett, J. R.;
.
Nicholson, J. K. Biochim. Biophys. Acta 1998, 1379, 367–380.
Analytical Chemistry, Vol. 80, No. 13, July 1, 2008 4889

Figure 3. A scheme showing (a) two overlapped doublets, each with J ) 7 Hz, separated by a chemical shift of 1 Hz; (b) the JRES projection
of the two doublets; (c) two doublets showing the inner peaks resolved by 1 Hz, i.e. a chemical shift difference of 8 Hz, requiring an 8-fold
increase in NMR observation frequency; (d) two triplets each with J ) 7 Hz, separated by a chemical shift of 1 Hz; (e) the JRES projection of
the two triplets; and (f) the two triplets showing the inner peaks resolved by 1 Hz, i.e. a chemical shift of 15 Hz, requiring a 15-fold increase in
NMR observation frequency.
general solution to the problem of overlapped signals in reaction
mixtures and have turned to statistical spectroscopic recovery
methods, including STOCSY, for which we have found multiple
applications in other areas.39–45 The STOCSY approach to infor-
mation recovery takes advantage of the colinearity of the intensity
variables in a set of NMR spectra, so that correlations from
resonances in the same molecule can be identified. The method
calculates statistical correlation matrices for individual spectral
data points corresponding to peaks of interest, and plots the
correlation coefficients onto a pseudo NMR spectrum with a color-
coding for visualization purposes. STOCSY has the advantage over
traditional 2D NMR correlation methods such as TOCSY46,47 as
positive correlations can be seen between resonances with no spin-
coupling, so all peaks from the same molecule can, in principle,
be identified assuming that there is not extensive overlap of peaks
with those of other compounds. This overlap problem is acute in
samples containing multiple saccharides, polyols or glucuronides
as may often be seen in biofluids. Analysis of lower or negative
(but still statistically significant) correlations has the potential to
allow molecules in the same chemical or metabolic pathways to
be determined. This method has recently been used to character-
ize structural pathway connectivities of drug metabolites,43 and
analyze complex spectral data from liquid chromato-
graphy-NMR spectroscopy data.44 It has also been extended to
visualize diffusion-edited NMR data,41 to enhance biomarker
detection by applying heteronuclear 1H-31P STOCSY to MAS
NMR spectra of liver and gut samples,39,42 and 1H-19F STOCSY
has been utilized for fluorinated drug metabolism studies.45
The main aims of the present study were to investigate the
efficiency of the STOCSY approach for assignment of NMR spectral
peaks in a chemically reacting system, and the development of JRES-
STOCSY to improve information recovery from heavily overlapped
signals and spin-systems. We have exemplified the approach using
the degradation of a model drug AG, (S)-R-methyl phenylacetyl AG
(MPG), which is a stuctural analogue of (S)-ibuprofen AG, the
statistical analysis of spectra collected over the time-course of the
acyl migration reactions being designed to extract connectivity
information through correlation of signals from the different isomers
as they are formed or degraded.
(35) Foxall, P. J. D.; Parkinson, J. A.; Sadler, I. H.; Lindon, J. C.; Nicholson,
J. K. J. Pharm. Biomed. Anal. 1993, 11, 21–31
(36) Nicholson, J. K.; Foxall, P. J.; Spraul, M.; Farrant, R. D.; Lindon, J. C. Anal.
.
Chem. 1995, 67, 793–811.
(37) Lynch, M. J.; Masters, J.; Pryor, J. P.; Lindon, J. C.; Spraul, M.; Foxall,
P. J. D.; Nicholson, J. K. J. Pharm. Biomed. Anal. 1994, 12, 5–19.
(38) Sweatman, B. C.; Farrant, R. D.; Holmes, E.; Ghauri, F. Y.; Nicholson, J. K.;
Lindon, J. C. J. Pharm. Biomed. Anal. 1993, 11, 651–664.
(39) Wang, Y.; Cloarec, O.; Tang, H.; Lindon, J. C.; Holmes, E.; Kochhar, S.;
Nicholson, J. K. Anal. Chem. 2008, 80, 1058–1066
(40) Cloarec, O.; Dumas, M. E.; Craig, A.; Barton, R. H.; Trygg, J.; Hudson, J.;
.
Blancher, C.; Gauguier, D.; Lindon, J. C.; Holmes, E.; Nicholson, J. K. Anal.
Chem. 2005, 77, 1282–1289
(41) Smith, L. M.; Maher, A. D.; Cloarec, O.; Rantalainen, M.; Tang, H.; Elliott,
.
P.; Stamler, J.; Lindon, J. C.; Holmes, E.; Nicholson, J. K. Anal. Chem. 2007,
79, 5682–5689
.
(42) Coen, M.; Hong, Y. S.; Cloarec, O.; Rhode, C. M.; Reily, M. D.; Robertson,
D. G.; Holmes, E.; Lindon, J. C.; Nicholson, J. K. Anal. Chem. 2007, 79,
8956–8966
.
(43) Holmes, E.; Loo, R. L.; Cloarec, O.; Coen, M.; Tang, H. R.; Maibaum, E.;
Bruce, S.; Chan, Q.; Elliott, P.; Stamler, J.; Wilson, I. D.; Lindon, J. C.;
EXPERIMENTAL SECTION
Nicholson, J. K. Anal. Chem. 2007, 79, 2629–2640
(44) Cloarec, O.; Campbell, A.; Tseng, L. H.; Braumann, U.; Spraul, M.; Scarfe,
G.; Weaver, R.; Nicholson, J. K. Anal. Chem. 2007, 79, 3304–3311
.
Materials. 3-[2,2,3,3-2H4]Trimethylsilyl propionate sodium salt
(TSP), sodium dihydrogen phosphate, and disodium hydrogen
phosphate were purchased from Sigma-Aldrich Company, Ltd.
.
(45) Keun, H. C.; Athersuch, T. J.; Beckonert, O.; Wang, Y.; Saric, J.; Shockcor,
J. P.; Lindon, J. C.; Wilson, I. D.; Holmes, E.; Nicholson, J. K. Anal. Chem.
2008, 80, 1073–1079
(46) Bruschweiler, R. J. Chem. Phys. 2004, 121, 409–414
.
.
(47) Bruschweiler, R.; Zhang, F. J. Chem. Phys. 2004, 120, 5253–5260.
4890 Analytical Chemistry, Vol. 80, No. 13, July 1, 2008

Figure 4. (a) Localized pseudo 3D correlation matrix of 1H 1D NMR spectra showing correlations between all isomers. (b) Localized pseudo
3D matrix of 1H 1D JRES projections showing correlations between all isomers. 1ꢀH1 etc. denotes the H1′ proton resonance of the 1-ꢀ-isomer.
(Gillingham, Dorset, U.K.). HPLC-NMR grade deuterium oxide
(2H2O) was obtained from Goss Scientific Instruments (Essex,
U.K.).
Chemical Synthesis of 1-ꢀ-O-Acyl (S)-r-methyl Pheny-
lacetyl Glucuronide (MPG). A model drug glucuronide [1-ꢀ-O-
acyl (S)-R-methyl phenylacetyl glucuronide] was chemically
synthesized using a variation of the selective acylation method
as recently documented.13,48,49 The method involves alkylation of
(48) Perrie, J. A.; Harding, J. R.; Holt, D. W.; Johnston, A.; Meath, P.; Stachulski,
A. V. Org. Lett. 2005, 7, 2591–2594
(49) Bowkett, E. R.; Harding, J. R.; Maggs, J. L.; Park, B. K.; Perrie, J. A.;
Stachulski, A. V. Tetrahedron 2007, 63, 7596–7605
.
.
Analytical Chemistry, Vol. 80, No. 13, July 1, 2008 4891

Figure 5. STOCSY plot calculated from the covariance to the data point corresponding to the maximum of the (S)-R-methyl phenylacetyl AG
H1′ proton resonance at δ 5.58 (0-4.5 h) (*minor AG contaminant).
benzyl glucuronate49 with racemic R-methyl phenylacetic acid
following activation of the carboxylic acid using Carpino’s uronium
reagent O-(7-azabenzotriazol-1-yl)-N,N,N′,N′-tetramethyluronium
hexafluorophosphate (HATU) and 1,4-diazabicyclo[2.2.2]octane
(DABCO). The diastereoisomeric benzyl esters were separated
by chromatography, and then the (S)-benzyl ester was deprotected
by catalytic dehydrogenation to afford the title compound as the
highly pure 1-ꢀ isomer.
17.65 h. An exponential apodization function corresponding to a
line-broadening factor of 0.3 Hz was applied to the 1D spectral
FIDs. The JRES data were zero-filled to 64 K (F2) and 128 (F1)
points prior to Fourier transformation, and a sine-bell apodization
function was applied followed by conventional tilting and
symmetrization.
Spectral Processing and STOCSY Calculations. JRES
skyline projections on to the F2 axis were calculated for each
experiment and treated as 1D NMR spectra for STOCSY analysis
purposes. These data were imported into MATLAB 7.4.0 using
in-house software (MetaSpectra: O. Cloarec, and T. Ebbels
Imperial College London), and scaled to a constant value for the
TSP peak intensity at δ 0.0. The region containing the small water
residual peak between δ 4.56 and 4.68 was removed to eliminate
the variation in water suppression. For calculation of 2D STOCSY
correlation matrices, each spectrum was imported into 4800 data
points with a computer point resolution of 1.6 Hz prior to data
analysis.
Correlation coefficients were calculated for data points corre-
sponding to peaks in the JRES projection selected for each of the
eight positional isomers. A pseudo-2D NMR matrix was plotted
showing the covariance between all the variables in the spectrum.
Pseudo-1D NMR plots were made by back-projecting the correla-
tion coefficients between a selected variable and all the other
variables of the spectrum on to the final NMR spectrum acquired
after equilibration for 18 h. The plots were color-coded to represent
the degree of correlation between the selected variable and all
the other variables in the spectrum.
1H NMR Spectroscopy. All NMR experiments were carried
out at 310 K on a Bruker Avance 800 spectrometer equipped with
a 5 mm z-gradient triple-resonance inverse (TXI) 1H/13C/15
N
cryoprobe at 800.32 MHz. The 1H 1D NMR spectra were recorded
using a standard water peak presaturation pulse sequence36 (to
suppress the residual H2O signal), and 16 scans were acquired
with a spectral width of 16025.5 Hz and a relaxation delay (RD)
of 2 s, with a total acquisition time of 1.55 min per spectrum. The
1H-1H 2D JRES spectra were acquired using the standard pulse
sequence RD-90°-t1/2-180°-t1/2-ACQ (t2), where t1 is an
increment delay and t2 is the acquisition time. Four scans per
increment for 20 increments were acquired into 16 K data points,
with a spectral width of 7684.4 Hz in the F2 domain (from t2) and
25 Hz in the F1 dimension (from t1) and a RD of 2 s. Each
spectrum was acquired in 8.73 min. Once the acquisition param-
eters had been optimized using a standard sample, 550 µL of 100
mM sodium phosphate buffer in 2H2O (pD 7.4) was added quickly
to 1.5 mg of MPG. The sample was transferred to a 5 mm NMR
tube containing 50 µL of TSP (0.5 mg/mL in 2H2O), and an
automated program was set up to acquire 103 loops of the 1D
and 2D experiments, which resulted in a total experiment time of
4892 Analytical Chemistry, Vol. 80, No. 13, July 1, 2008

Figure 6. STOCSY plot calculated from the covariance to the data point corresponding to the maximum of the (S)-R-methyl phenylacetic acid
AG 2R H1′ proton resonance at δ 5.40 (4.5-9 h).
the proton resonances in the anomeric region and for the
glucuronide ring protons at ∼δ 3.5. However, it is clear that the
JRES data is of great value for assigning resonances and multiplic-
ity in these regions.
The full advantage of acquiring the JRES NMR spectra can be
seen in Figure 2, which compares a series of 1H 1D NMR spectra
with the 1H 1D JRES skyline projections at the same time-points
for the δ 3.2-6.2 region of the glucuronide ring protons. The
complexity of the 1H 1D spectra in (a) illustrates the difficulty in
monitoring a dynamic system when there are a large number of
overlapped spin-coupled multiplet resonances. The skyline projec-
tions seen in (b) allow the peaks to be more resolved, and thus
the change in peak heights can be seen more clearly. The JRES
experiment is based on the spin-echo pulse sequence, and peak
intensities are dependent on the proton T2 values as well as
Figure 7. A statistically reconstructed 1D NMR spectrum showing
concentration. In addition, these data are often heavily weighted,
8 overlaid coefficient plots for the median spectra of each isomer.
particularly in the J-coupling (F1 axis), using functions such as
the shifted sine-bell, or Lorentzian-Gaussian transformation, of
which the pseudoecho is a special case. Thus both of these effects
can, in principle, cause errors in peak intensities that might affect
any subsequent correlation analysis. However, in the reacting
glucuronide system as studied herein, where all of the species
involved are either isomers or closely related structurally, proton
T2 values likely to be similar.
STOCSY experiments are always made less effective by peak
overlap as independent signals lower the covariance. The degree of
1H NMR spectral dispersion seen in Figure 2 could normally only
be achieved with unattainably high NMR magnetic field strengths
Correlation coefficients >0.9 are shown only, allowing each isomer
to be isolated. The color scheme is chosen to define each isomer
separately.
RESULTS AND DISCUSSION
1
1D and 2D NMR Spectroscopy. A H 1D NMR spectrum
of MPG is shown in Figure 1(a). Figures 1(b) and 1(c) show the
1H-1H JRES NMR spectrum in the form of a 1D skyline projection
and 2D contour plot respectively. The contour plot displays the
chemical shifts and coupling constants in two orthogonal fre-
quency domains, with the skyline projections displaying these data
as a 1D spectrum of singlets. Even though these spectra show
resonances from only one compound, there is some overlap in
Analytical Chemistry, Vol. 80, No. 13, July 1, 2008 4893

as illustrated in Figure 3. Figure 3(a) shows a system of two
overlapped doublets separated by a chemical shift of 1 Hz, each with
coupling constants of J ) 7 Hz. If this result had been achieved at a
frequency of 800 MHz, in order to separate them into two completely
resolved doublets by a spacing of 1 Hz, as shown in Figure 3(c), i.e.
chemical shift difference of 8 Hz, the NMR experiment would need
to be carried out on a magnet with 8 × Bo field strength (ca. 150 T
operating at 6.4 GHz) which is not possible with current technology.
For comparison, the JRES projection is shown in Figure 3(b), and
because of the removal of J-coupling splittings, the 1 Hz separation
is easily seen. A similar argument applies to partially overlapped
triplets, where the hypothetical increase in observation frequency
to cause complete separation of the two multiplets with a chemical
shift difference of 1 Hz, by 1 Hz, would be a factor of 15, as shown
in Figures 3(d) to (f). Therefore, the effect of projecting 1D skyline
NMR spectra from 2D JRES data produces an apparent or virtual
spectral dispersion that is not achievable using conventional NMR
spectrometers, and this is most useful in STOCSY analysis of skyline
intensity covariance to demonstrate intramolecular connectivity.
STOCSY Analysis of NMR Data. The STOCSY analysis was
carried out across the full set of 103 time course 1D NMR
experiments to show correlations between spectral peak intensities
in the reacting system. Figure 4(a) shows a correlation matrix of
the 1H 1D spectra and can be interpreted as a pseudo-3D NMR
spectrum (the axes hold information about the coupling constants,
chemical shift and time domain). Each peak from the NMR spectra
can be seen on the diagonal line, and correlations between peak
intensities appear as diagonal peaks similar to a standard TOCSY
seen that the highest spectral correlations (r >0.9) with the peak
of interest (δ ) 5.58, H1′ 1-ꢀ driver) all arise from the 1-ꢀ isomer,
and clearly differentiate the peaks of this isomer from those of
the other isomers. (Nomenclature: 2-ꢀ H1′ refers to the proton at
the C-1 position on the glucuronide ring of the 2-ꢀ positional
isomer etc.) Correlations are not seen between the 1-ꢀ and 1-R
anomers as formation of the latter is not by mutarotation but by
positional isomerization, with the 1-R isomer being formed through
a back-reaction from the 2-R isomer.
Targeted 1D K-JRES-STOCSY analyses were carried out to
identify peaks from the other isomers within the spectral sets
where the peaks were at their maximum intensity for that isomer,
in order to maximize variation for distinguishing each isomer. A
K-JRES-STOCSY plot for experiments acquired in the 4.5-9 h
time-period, showing correlation to the 2-R H1′ proton, is shown
in Figure 6. A strong correlation (r >0.9) is seen with the other
protons in the 2-R isomer, as well as those from the 2-ꢀ isomer.
As seen in Scheme 1, the 2-R isomer can undergo mutarotation
to the 2-ꢀ isomer, hydrolysis to the aglycon, and also acyl
migration to form the 3-R and 1-R isomers, and apart from
hydrolysis all are reversible reactions. Correlations to the 3-R and
1-R isomers have high coefficients of r ) 0.88 and r ) 0.72, while
anticorrelations are seen to the 1-ꢀ isomer r ) -0.92.
A K-JRES-STOCSY analysis carried out on the 13.5-18 h time-
period using the same driver peak gives different correlation
coefficients from those calculated in the 0-4.5 h analysis. The 2-R
isomer now appears anticorrelated to the 3-R and 1-R isomers, r )
-0.45 and r ) -0.12 correspondingly, while correlation to the 1-ꢀ
isomer is higher than previously observed at r ) 0.19. This can be
explained by the rate of formation of the 2-R isomer, which increases
rapidly from 0 to 10 h and then slows from 10 h onward, and the
2-R/2-ꢀ isomer concentrations then decline. At 10 h the 1-ꢀ isomer
concentration was still decreasing, and the 3-R and 1-R isomers were
increasing, generating anticorrelations to the 1-ꢀ signal intensities.
Correlations can also be investigated over the time-course of
the entire 18 h experiment. For each of the 8 isomers, an edited
median spectrum was generated such that only points in the
spectrum which corresponded to a threshold correlation coef-
ficient r > 0.9 were displayed, and the other points were set to 0.
An overlay of these 8 edited spectra is shown in Figure 7,
illustrating how the use of a correlation coefficient threshold can
help differentiate specific isomer resonances.
The acquisition of JRES spectra decreases the spectral com-
plexity and facilitates the generation of each STOCSY correlation
plot by increasing the spectral dispersion. Therefore the consider-
able peak overlap is eliminated, and furthermore, good signal-to-
noise ratios are obtained in a relatively short period of time. The
JRES-STOCSY method also allows structural information about
protons from the same molecule that are not spin-spin coupled
to be detected, and allows connectivities to be established in the
dynamic pathway. This technique could also be used effectively
to help assign metabolite identities in complex samples such as
biofluids, where there is considerable overlap of multiple com-
pound peaks throughout the NMR spectral regions. The JRES
technique has already been shown to be valuable for analysis of
complex mixtures of endogenous metabolites,34–38 and so the
combination JRES-STOCSY technique will be particularly useful
for metabonomic/metabolomic applications.
1
1H NMR spectrum. A corresponding correlation matrix of the H
1D JRES projections can be seen in (b) illustrating the decrease of
spectral complexity, and the increase in the spectral dispersion when
compared to (a). High correlations are seen for peaks arising from
the same positional isomer and also between mutarotating isomers
since this reaction is fast50 compared to the degradation rate.28
Due to the sequential nature of the isomerization kinetics, not
all peaks vary strongly in intensity over the whole time course,
and so a “time slice” variant was applied to the kinetic STOCSY
1
approach and H 1D K-JRES-STOCSY analyses were carried out
over 4 time intervals, 0-4.5, 4.5-9, 9-13.5 and 13.5-18 h. By
analyzing each interval separately, it was possible to identify the
time-period which best discriminated the individual isomers and
taking into consideration only the highest correlation coefficient,
this permitted the identification of the NMR peaks arising from
each individual isomer. Within each interval a “driver” peak (the
peak from which a data point is chosen at its apex to initiate the
statistical analysis) was selected for the isomer of interest, and
correlations to this peak were calculated using spectra acquired
in the time-course interval and back-projected onto the final
spectrum at the end of the experiment where all isomers are
present. A 1D K-JRES-STOCSY plot calculated using the 1D JRES
projection NMR spectra acquired in the 0-4.5 h time-period is
shown in Figure 5. Correlation coefficients were calculated for
correlations between a data point belonging to the H1′ proton of
the 1-ꢀ isomer and the rest of the data points in the matrix over
the 4.5 h, and the coefficients for each data-point were back-
projected onto the final (18 h) NMR spectrum. It can be clearly
(50) Sidelmann, U. G.; Gavaghan, C.; Carless, H. A. J.; Farrant, R. D.; Lindon,
J. C.; Wilson, I. D.; Nicholson, J. K. Anal. Chem. 1995, 67, 3401–3404
.
4894 Analytical Chemistry, Vol. 80, No. 13, July 1, 2008

1
By obtaining multiple NMR spectra, in this case H 1D skyline
ACKNOWLEDGMENT
This work was supported by the BBSRC and AstraZeneca
through the award of a CASE studentship to Caroline H. Johnson.
projections of the JRES data, it is possible to capture the reaction
processes, and with the addition of STOCSY analysis, the pathway
of these processes can be defined, and mixtures of isomers can be
assigned. This type of analysis is particularly useful when dealing
with reacting systems, operating at the minute time-scale which is
conducive to rapid sequential measurements on highly sensitive
cryoprobes. For more rapid chemical reactions, simple one-shot 1H
NMR spectra obtainable in as little as one second could be used.
The application of the K-STOCSY approach here has been observed
with transacylation reactions of the 1-ꢀ-O-AGs, and correlations
between isomers have been shown to differ as the experimental time
increases, but this technique could also be applied to other dynamic
or enzyme-catalyzed metabolic systems.
NOTE ADDED AFTER ASAP PUBLICATION
This paper was posted on 5/10/08. An error in the labeling in
Scheme 1 was noted and corrected. The paper was reposted on
5/29/08.
Received for review December 24, 2007. Accepted March
21, 2008.
AC702614T
Analytical Chemistry, Vol. 80, No. 13, July 1, 2008 4895