Review

pubs.acs.org/ac

NMR Spectroscopy for Metabolomics and Metabolic Profiling

Cynthia K. Larive,*,† Gregory A. Barding, Jr.,‡ and Meredith M. Dinges†
†
Department of Chemistry, University of California−Riverside, Riverside, California 92521, United States
‡

■
Chemistry and Biochemistry Department, California State Polytechnic University, Pomona, California 91768, United States
CONTENTS
Applications of NMR Metabolomics 133
Biofluids 134
Downloaded via INST NAC DE TECNOLOGIA INDUSTRIAL on September 23, 2022 at 11:56:07 (UTC).

Cells and Tissues 134

See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

Plants 134
Natural Products 134
Environmental Interactions 135
Food and Beverages 135
Experimental Design 135
Sample Storage and Preparation 136
Extraction 136
Development of New Methods and Experiments: 1H
NMR 136
Relaxation Effects 136
New Pulse Sequences 137
Fast and Ultrafast NMR Experiments 137
Probe Developments 137
High-Resolution Magic Angle Spinning (HR-MAS) 138
Development of New Methods and Experiments: Figure 1. Growth in publications on NMR metabolomics/
Nuclei Other than 1H 138 metabonomics since 2000. These results were obtained from a topic
Isotope Labeling 138 search of all documents on the Web of Knowledge using the keywords
Metabolic Flux Analysis 139 “NMR and metabonomic*” and “NMR and metabolomics*”.
Hyperpolarization and DNP 140
Data Processing and Chemometrics 140 illustrates the rapid growth in publications on topics that
Spectral Processing 141 include the keywords NMR and metabolomics/metabonomics
Resonance Assignments and Databases 142 since the year 2000. Due to space limitations, this Review
Quantitation 142 covers only papers published between 2011 and the first half
Chemometrics 142 of 2014.
Perspective
Author Information
Corresponding Author
143
144
144
■ APPLICATIONS OF NMR METABOLOMICS
The omics revolution in systems biology has provided an
Notes 144
unprecedented level of understanding of how organisms
Biographies 144
respond to environmental factors such as diet, aging, and
Acknowledgments 144
disease. As the end products of transcription and translation,
References 144
small molecule metabolites can be considered as the output
of the biological system, in essence a molecular phenotype.

A pplications of NMR for metabolomics and metabolic

profiling continue to grow rapidly as does the refinement
of methods for the measurement, analysis, and interpretation of
complex data sets. Metabolomics (metabonomics) is a set of
global measurements performed on biological samples with
the goal of quantifying as many metabolites as possible and
evaluating changes in metabolite levels as a result of an applied
stress. Metabolic profiling experiments follow a more limited
set of metabolites often through specific pathways. NMR is also
well suited for metabolite fingerprinting, which involves the
comprehensive and simultaneous analysis of a wide variety of
compounds. For the purpose of this Review, we do not distin-
guish between metabolomics and metabonomics and have
elected to use the term metabolomics throughout. Figure 1
Because metabolites can impact gene transcription through
feedback loops, the flow of information from the genome
through the transcriptome and proteome to the metabolome is
bidirectional (Figure 2).1,2 Though a comprehensive review
of NMR metabolomics applications is beyond the scope of
this Review, the impact of this approach can be appreciated by
a survey of recently published review articles. The section
below provides an overview of recent reviews describing appli-
cations of NMR-based metabolomics. In addition, the review by
Special Issue: Fundamental and Applied Reviews in Analytical
Chemistry 2015
Published: November 6, 2014

© 2014 American Chemical Society 133 dx.doi.org/10.1021/ac504075g | Anal. Chem. 2015, 87, 133−146
Analytical Chemistry
Figure 2. Complex interactions of functional levels (genome, transcriptome, proteome, and metabolome) in biological systems. Bidirectional flows of
biological information are observed between the genome, transcriptome, proteome, and metabolome. The complex interaction of components from
all the functional levels and the environment produces the phenotype, the output of the system measured in systems-level metabolomics and systems
biology. Adapted from ref 1 with permission of The Royal Society of Chemistry, copyright 2011.
Bothwell and Griffin provides an excellent introduction to
biological NMR.3
Biofluids. Biofluids are an attractive sample matrix for
metabolomics studies because they can be obtained in a
noninvasive (e.g., saliva, urine) or minimally invasive manner
(e.g., blood plasma or serum, cerebrospinal fluid). NMR
metabolomics studies using biofluids offer potential for a
deeper understanding of disease pathogenesis and the
identification of metabolic biomarkers useful for disease
diagnosis or treatment monitoring.4−7 Because of the aberrant
metabolism of cancer cells involving enhanced glucose uptake
and glycolytic activity, a shift in the TCA cycle from oxidation
to lipogenesis, increased glutaminolysis, and nucleotide biosyn-
thesis, metabolic screening of biofluids has the potential to
contribute to early cancer screening, improved diagnostic
accuracy, and prediction of a patient’s response to treatment.8
For example, urine metabolic profiling shows excellent promise
as a screening tool for bladder cancer.9
For a metabolic biomarker to be clinically useful, its level
must clearly associate with disease risk or progression, should
be insensitive to variables like ethnicity, diet, and location, and
should not vary too much over the short-term within an
individual. To assess the sources of variation among individuals,
blood plasma and urine 1H NMR spectra were measured for
samples obtained from 154 healthy Caucasian, postmenopausal
female twins (identical and nonidentical).10 Thirty four of the
identical twins provided two samples over the space of several
months in this study. On the basis of these results, Nicholson
et al. estimate that stable variation derived from familial and
individual-environmental factors accounts for on average 60%
of the biological variation in the 1H NMR spectra of plasma and
47% of the variation in the urine spectra. The authors postulate
that clinically predictive metabolic variation potentially useful
for biomarker discovery lies within this stable component. In
an effort to catalog the metabolic components of urine, a
comprehensive study of the human urine metabolome using
NMR, GC/MS, ICPMS, and direct infusion- and LC-MS/MS
reported the identification of 445 and quantification of 378
unique urine metabolites or metabolic species, with 179
metabolites detected using NMR.11 The potential of 1H
NMR metabolic profiling to identify trace quantities of body
fluids (blood, urine, saliva, and semen,) singly or in binary
134
Review
mixtures, was also demonstrated and validated as a rapid and
nondestructive tool for forensic analysis.12
Cells and Tissues. In addition to the global metabolic view
provided by blood and urine, sampling of intact tissues can
provide a more specific, local response to disease. For example
blood serum, bronchoalveolar lavage fluid, and excised lung
tissue were used to identify metabolic biomarkers of sepsis in
rats.13 Another recent review documents the progress made in
the application of NMR and magnetic resonance spectroscopic
imaging (MRSI) to detection, diagnosis, and characterization of
human prostate cancer.14 NMR metabolomics was used to
investigate the effects of environmental stressors associated
with biofilm development by Staphlococcus epidermis, strongly
implicating a central role for the TCA cycle.15 NMR studies on
the effect, mode of action, delivery, and toxicity of therapeutic
drugs has been recently reviewed.16 NMR metabolomics also
has potential to provide a thorough assessment of the toxicity
and mechanisms of action of new nanomaterial-based
therapeutics.17
Plants. Plant metabolomics studies have been used to
address the effects of genotype, ecotype, and environmental
stressors such as drought and submergence18−20 and can
provide a metabolic phenotype in chemical genomics studies in
plants.21 The application of 1H NMR spectroscopy to plant
metabolomics is in many ways more complicated than biofluid
analysis and requires consideration of whether the whole plant,
a portion of the plant (i.e., leaves, roots, or fruit), or a single cell
type is most relevant to the question being explored.22 Most
plant metabolomics studies are carried out on plant tissue
extracts;18,22 however, the review by Serra et al. discusses the
use of solid state NMR for the characterization of the insoluble
protective biopolymers cutin and suberin.23 In addition to their
primary metabolome, plants also produce an immense number
of secondary metabolites (∼200 000) that allow them to
interact with beneficial or harmful organisms. Leiss et al.
provide an overview on the use of NMR metabolomics to
identify compounds important for host plant resistance to
western flower thrips (Frankliniella occidentalis).24
Natural Products. NMR metabolomics approaches are also
finding use in discovery-oriented natural products chemistry.25
Comparison of high-resolution 2D NMR spectra of unfrac-
tionated sample extracts can facilitate the detection and
dx.doi.org/10.1021/ac504075g | Anal. Chem. 2015, 87, 133−146
Analytical Chemistry Review

identification of novel compounds that might be lost during

chromatographic fractionation because of their low abundance,
chemical properties, or instability. Activity guided fractionation,
a common approach in natural products research, also fails
when several biogenic small molecules (BSMs) act synergisti-
cally. Through the statistical analysis of NMR spectra of
complex mixtures of BSMs, unique spectral features can be
identified and correlated to a phenotype or biological property
of interest.26
Environmental Interactions. Environmental metabolo-
mics addresses the effects of complex environmental stressors
including abiotic factors (e.g., exposure to xenobiotic com-
pounds and temperature shifts) and biotic factors (e.g.,
herbivory and competition) and can be used to elucidate
their biochemical modes of action.27 As NMR metabolomics
studies using samples derived from human patients can be a
powerful tool for the discovery of disease biomarkers, metabolic
shifts in terrestrial or aquatic organisms are potential early
bioindicators of environmental stress. To integrate the results
of genomic, transcriptomic, and metabolomics data, Ogata et al.
introduced ECOMICS, a public domain web-based toolkit that
allows for monitoring of biomass metabolism and facilitates
understanding of relationships between molecular and micro-
bial elements.28
Food and Beverages. Though applications of NMR
metabolomics are most prevalent in studies using biofluids or
samples derived from living organisms, this approach has also
been used to address a diverse range of applications involving
other types of complex samples. For example, NMR
metabolomics has been used to assess the quality of plant
herbal products and juices.29 For wines, NMR metabolomics
measurements can reflect the grape variety and age of the
vintage, as well as provide insights to the “terroir”, which can
encompass the effects of geographic region, climate, soil,
agricultural practices, and fermentation organisms, parameters
well-known to impact the taste of wine as well as its metabolic
profile.30 The application of NMR metabolomics using wine,
juice, and milk samples has been described for student
instruction in a variety of courses while also illustrating how
students can use 1H NMR metabolic profiling to follow the
utilization of glucose by yeast.31 Figure 3. Work flow for clinical metabolomics/metabolic NMR

■ EXPERIMENTAL DESIGN
A typical metabolomics workflow is summarized diagrammati-
cally in Figure 3.32 The study begins with consideration of
factors such as determining the number of subjects, establishing
appropriate controls, the types of samples that will be provided,
how frequently samples will be taken, and how the samples will
be stored. The collected samples undergo preparation for NMR
analysis, and the NMR spectra are recorded according to a
standardized protocol selected by the investigator. The NMR
data sets are then subjected to statistical analyses to identify
variance between the test groups and those NMR resonances
important for distinguishing the sample groupings. Once key
metabolites are identified, a univariate analysis can be
performed to establish trends in metabolite levels as a function
of a perturbation (i.e., dose, time) or to quantify differences
between control and treatment samples. After significance has
been established, metabolic pathways are examined to under-
stand the biochemical basis for the effect.
Proper experimental design is critical to obtaining meta-
bolomics data that addresses the experimental hypothesis and
can be translated to other laboratories or situations. Keun and
135
spectroscopy. Samples are collected in a uniform way to minimize
variability and are analyzed by an NMR profile to collect data on all
metabolites potentially present in the sample. Pattern recognition
approaches (PRA) include principal component analysis, partial least-
squares discriminant analysis, orthogonal projections to latent
structures, heat map, support vector machines, the random forests
method, and other modes aiming to highlight underlying trends and
visualization tools such as contribution. Trend and box plots are used
to further evaluate these. Receiver operating characteristic (ROC)
curves are generally considered the method of choice for evaluating the
performance of potential biomarkers. The markers are eventually
placed in a metabolic pathway to provide insight on the biochemical
phenomena. Reprinted from ref 32 with permission from John Wiley
& Sons Inc., copyright 2013.
Athersuch have reported typical experimental protocols for
NMR metabolomics analyses of biofluids, tissues, cell extracts,
and culture media.33 A comparison study of technical replicates
of human urine samples at two universities using 5 mm and
3 mm probes revealed the importance of parameter stand-
ardization in producing reliable and reproducible quantitative
NMR metabolomics data.34 An important finding of this study
is that previously unrecognized parameters, especially those
dx.doi.org/10.1021/ac504075g | Anal. Chem. 2015, 87, 133−146
Analytical Chemistry Review

related to solvent suppression, can have a dramatic effect on the

results. Though the metabolomics community has not yet
arrived at a set of standardized methods, standardization of
controlled vocabularies (or ontologies) and agreement on
reporting standards for the experimental methods and bio-
logical data facilitate data exchange.1 An important step toward
reducing interlaboratory variability, verifying method accuracy,
and validation of new measurement methods is the availability
of a standard reference material SRM 1950 “Metabolites
in Human Plasma” by the National Institute of Standards
and Technology (NIST).35 SRM 1950 is provided with
comprehensive metabolomics characterization using GC/MS,
LC-MS/MS, and NMR.
Sample Storage and Preparation. As in almost any
analysis, reproducible and effective methods of sample storage,
preservation, and preparation are critical to the outcome of
NMR metabolomics studies. Factors affecting the inter- and
intrasubject variability in human urine samples, including
storage conditions and diet, were investigated.36 Bacterial
metabolites were minimized by storing the samples under
refrigeration, and diet standardization attenuated the effects of
diet culture and cohabitation, though gender-specific differ-
ences remained. The effects of sample handling and storage
conditions on human plasma samples for NMR metabolomics
studies have also been investigated.37 Due to the high protein
content of blood plasma or serum samples, which obscures
the resonances of small molecule metabolites, it is typically
necessary to physically remove the proteins by precipitation or
ultrafiltration or to attenuate the intensity of protein resonances
in the 1H NMR spectrum based on their faster rates of
T2 relaxation. Gowda and Raftery compared these approaches
and found that precipitation of blood serum proteins using
methanol gave the best performance, producing significantly
higher levels of some metabolites than other approaches.38 An
optimized method for attenuating intersample differences in the
pH and dication concentrations in NMR spectra of rat, mouse,
and human urine samples has also been reported.39
Extraction. Extraction is a critical component of sample
preparation protocols for cell and tissue analysis,32 but it
can also be slow and labor-intensive. Matheus et al. report a Figure 4. 1H NMR spectra of the “wash”, “elute 1”, and “elute 2”
streamlined approach using sonication to disrupt cell mem- fractions compared with the reference (nonfractionated) profile from
branes or tissue structures during the extraction step.40 A study the same urine sample. The blue profile was reconstructed by adding
exploring extraction protocols using solvents with varying the “wash”, “elute 1”, and “elute 2” subprofiles. The aromatic and
polarities for NMR metabolomics of adherent mammalian aliphatic regions are shown in (a) and (b), respectively. Reprinted
cells demonstrated that a higher proportion of hydrophilic from ref 42 with kind permission from Springer Science and Business
metabolites was obtained with a solvent of methanol/ Media, copyright 2012.
chloroform/water.41 Compared with liquid−liquid extraction
methods, solid phase extraction (SPE) can be useful for strategy for structure elucidation and compound identification
in metabolomics studies.43−45

fractionating complex samples prior to measurement of NMR
spectra and can be more amenable to automation. SPE-NMR
was used to isolate 3 fractions from urine samples based on
differences in polarity and improved the selectively of the NMR
measurements. Figure 4 shows the 1H NMR spectra obtained
for urine samples applied to hydrophilic−lipophilic balance
(HLB) SPE cartridges and collected into three fractions which
were analyzed separately: wash (load + 3 wash volumes of
0.5 mL of water), elute 1 (3 elutions with 0.5 mL of 10%
methanol/90% water), and elute 2 (3 elutions of 0.5 mL of
methanol).42 Mass-guided SPE-trapping of selected compounds
for NMR measurements were used in conjunction with HPLC-
FTMS to elucidate the structures of 36 phenolic conjugates in
human urine following a single bolus intake of black or green
tea.43 Hyphenation of LC-MS-SPE-NMR has proven a useful
136
■
DEVELOPMENT OF NEW METHODS AND
EXPERIMENTS: 1H NMR
Because of its high sensitivity and natural abundance and its
nearly ubiquitous presence in organic metabolites, many
metabolomics studies rely on the measurement of 1H NMR
spectra. In this section, efforts to improve the reproducibility,
throughput, sensitivity, and selectivity of 1H NMR meta-
bolomics measurements are summarized.
Relaxation Effects. Most NMR metabolomics experiments,
and especially those performed on biofluids, depend on effec-
tive suppression of the water resonance. One of the most pop-
ular methods of solvent suppression is the 1D-NOESY pulse
sequence, because of its robustness and ease of implementation.
dx.doi.org/10.1021/ac504075g | Anal. Chem. 2015, 87, 133−146
Analytical Chemistry Review

Figure 5. Schematic diagram for MaxEnt. MaxEnt reconstruction begins with empirical data and a preliminary trial spectrum f (typically a blank
spectrum). Spectrum f is inverted (DFT−1) to create “mock” data (m) that is compared with the empirical data (d). An update to the trial spectrum
is computed by searching along the gradients of the entropy and the constraint (the agreement between the empirical and mock data).
The algorithm converges to the unique MaxEnt solution when the gradient of the objective function Q = S − λC is zero and the gradients of S and C
are antiparallel. Reprinted from ref 52. Copyright 2014 American Chemical Society.

Parameter optimization and the mechanism of 1D NOESY
for water suppression, which in part depends on differences in
the T1 relaxation properties of water and the metabolites of
interest, have been recently reviewed.46 In a different app-
lication of the NOESY experiment, Farooq et al. report the use
of supercooled water in 1.4 mm capillaries to measure high
quality 2D NOESY spectra of small molecule metabolites
at −15 °C.47 At this temperature, the solvent viscosity is 3−4
times higher than water at room temperature affecting the rate
of proton cross relaxation and producing strong NOESY
correlations that can be used in concert with 2D experiments
based on J-coupling for resonance assignment and metabolite
identification.
New Pulse Sequences. To overcome problems of res-
onance overlap in the 1H NMR spectra of complicated samples,
Martin-Pastor described a one-dimensional singlet-filtered
experiment that edits complex spectra to reveal only the
peaks of singlet resonances and weakly coupled signals.48
Experimental schemes that produce high quality 2D J-resolved
spectroscopy and spin−echo correlated spectroscopy (SECSY)
spectra in inhomogeneous magnetic fields have also been
reported and demonstrated to be useful for in vivo measure-
ments.49,50 Longitudinal relaxation enhancements using
selective excitation and refocusing pulses were demonstrated
to increase the sensitivity of in vivo detection of mouse brain
metabolites by magnetic resonance spectroscopy (MRS).51
Fast and Ultrafast NMR Experiments. Pulsed NMR
experiments typically rely on the discrete Fourier transform
(DFT) to convert digitized time domain free induction decays
(FIDs) to frequency domain data. Multidimensional NMR
spectra using DFT require uniform sampling intervals with
sufficiently small time increments in the indirect dimension
to satisfy the Nyquist condition and long evolution times for
high resolution. As a result, multidimensional NMR spectra often
require long acquisition times and resolution in the indirect
137
dimension is limited. One way to overcome this limitation is
through fast NMR experiments which use nonuniform sampling
with maximum entropy (MaxEnt) reconstruction, a non-Fourier
method of spectrum analysis, illustrated in Figure 5.52 Rai and
Sinha report the use of nonlinear sampling, forward maximum
entropy reconstruction, and the J-compensated heteronuclear
single quantum coherence (HSQC) pulse sequence to achieve a
22-fold reduction in the data collection time for body fluid
samples without any compromise in the quantitation of low
abundance metabolites.53
Though maximum entropy reduces the time required for
multidimensional NMR experiments by reducing the number
of data points acquired, ultrafast NMR experiments which
grew out of magnetic resonance imaging techniques, allow
the acquisition of 2D NMR spectra in a single scan.54 These
ultrafast 2D NMR experiments open the door to high
resolution and high throughput 2D NMR-based metabolomics
studies.55 An ultrafast heteronuclear 2D J-resolved spectros-
copy pulse sequence has been reported.56 As an extension of
this experiment, an ultrafast 3D UFJCOSY pulse sequence was
described that produces COSY-type correlations with hetero-
nuclear couplings displayed in the third dimension.57 This
experiment allows direct measurement of the level of isotopic
enrichments, which are critical for metabolic flux experiments.58
In addition to greatly reducing the acquisition time, these
ultrafast experiments are also widely applicable for quantita-
tive NMR measurements.59 Coupled with stable isotope
labeling and polarization methods that enhance sensitivity,
this breakthrough promises to revolutionize in vivo NMR
spectroscopy.
Probe Developments. The NMR probe contains the
hardware elements necessary to apply radio frequency pulses
to the sample and to detect the resultant signal prior to
amplification and digitization. Therefore, probe design and
optimization is an avenue frequently explored to improve
dx.doi.org/10.1021/ac504075g | Anal. Chem. 2015, 87, 133−146
Analytical Chemistry
detection sensitivity or to enable new types of measurements.
Renslow et al. describe a biofilm microreactor seated on a
custom built NMR probe that allows for simultaneous
electrochemical and NMR measurements at the microscale.60
Other advances in probe design include development of a
13
C-optimized probe with a 35 μL sample volume constructed
from a high temperature superconducting oxide deposited on a
sapphire substrate.61 To facilitate experiments involving hetero-
nuclear hyperpolarization at low fields, a single input, double-
tuned probe circuit has been developed that enables double
resonance NMR to be performed with single channel consoles
using PANORAMIC (Precession And Nutation for Observing
Rotations At Multiple Intervals about the Carrier) waveforms.62
Compared with traditional 5 mm NMR probes, microcoil
probes offer improved mass sensitivity. The application of
microcoil NMR for metabolomics of urine and serum samples,
including the effect of concentrating biofluid samples on
metabolite quantification, has been investigated using a probe
designed for flow injection analysis.63 A stripline NMR probe
has been integrated into a microfluidic flow system allowing
electrochemical generation of reactive products concentrated
via SPE before measurement of the NMR spectrum in the
150 nL active volume.64 Though microfluidic chip-based flow
probes are potentially attractive for high-throughput metab-
olomics measurements, obtaining high spectral resolution
remains a challenge because mismatch in magnetic suscepti-
bility of the sample and the chip material results in suscepti-
bility broadening. An innovative strategy to overcome
susceptibility broadening in chip-based NMR probes relies on
the simulation of field maps to optimize the shape of air filled
compensating structures that ensure a flat magnetic field
distribution inside the sample detection region.65
High-Resolution Magic Angle Spinning (HR-MAS).
HR-MAS experiments provide 1H NMR spectra of semisolid
materials such as cells, tissues, biopsy samples, organs, and
organisms. Spinning the sample about an axis at the magic
angle (54°44′) reduces line broadening by averaging magnetic
field gradients at compartment boundaries, the effects of
residual homonuclear dipolar interactions, and chemical shift
anisotropy.66 A complication in interpreting HR-MAS spectra
is that metabolite chemical shifts can vary slightly according to
the local microenvironment in tissues or cells (e.g., pH) ham-
pering accurate quantitation of metabolite concentrations. Lazariev
et al. describe an accurate method coupling quantum mechan-
ical simulations and quantitation algorithms (QM-QUEST)
that corrects mismatches producing corrected metabolite finger-
prints.67
An exciting advance in HR-MAS is the reduction in sample
size requirements using microcoil NMR probes. Wong et al.
report the use of magic angle coil spinning (MACS) in which a
62 μm diameter copper wire was wound around a quartz
capillary tube and soldered to a nonmagnetic capacitor
producing a detection volume of 690 nL.68 The capillary and
coil were fit inside a ceramic insert and placed in a standard
4 mm rotor. This setup allowed the measurement of 1H NMR
spectra for a 500 ng sample of muscle tissue and produced a
7-fold gain in sensitivity compared with a spectrum measured
for 16 mg of tissue packed into a 30 μL disposable insert,
though the resonances in the MACS spectra were broader than
desired for detailed metabolic phenotyping. Refinements of
the initial MACS design were reported that minimize line
broadening from temperature gradients and anisotropic
magnetic susceptibility to produce resonance line widths less
138
Review
than 1 Hz.69 This technology was used to obtain 1H NMR
spectra at 18.8 and 23.5 T for two strains of C. elegans
nematodes highlighting the potential of NMR microprobe
technologies for metabolic screening of small model organ-
isms.70 To minimize heating from eddy currents and reduce the
centrifugal force felt by the worms, a spinning frequency of
350 Hz was used.
■
DEVELOPMENT OF NEW METHODS AND
EXPERIMENTS: NUCLEI OTHER THAN 1H
A fundamental advantage of using NMR to probe biological
systems is the ability to measure a wide variety of spin 1/2
nuclei, including (but not limited to) the biologically relevant
1
H, 13C, 31P, and 15N in addition to the pharmaceutically
relevant 19F. Of these, 1H, 19F, and 31P are found at high
percentages of natural abundance. The lower natural abundance
of 13C and 15N can be useful in measuring enrichment due to
catabolism of labeled precursors as a function of specific bio-
logical processes or pathways. More specifically, NMR can
easily be used to monitor the positional enrichment of a
molecule for both 13C and 15N NMR to establish which
pathways are being activated or inactivated. The sensitivity
disadvantage of NMR compared to mass spectrometry-based
techniques is somewhat overcome by the inherent capability of
NMR to monitor isotopomer enrichment. The disadvantage of
mass spectrometry for measuring isotopomer distribution is
based on spectral interpretation; the investigator must deduce
from fragmentation patterns the location of enrichment,
reducing sensitivity and throughput. The fate of labeled
precursors (especially 13C labeled glycolytic intermediates)
has been tracked with NMR (or MRS) in vivo using hyper-
polarization and DNP. To better appreciate the versatility of
NMR for exploring biological processes, the application of
NMR for metabolic flux measurements (including isotope
tracing), isotope labeling, and DNP are discussed below.
Isotope Labeling. Although NMR has inherent advantages
for measuring in vivo isotopologue enrichment from 13C
and 15N labeled substrates, it may still be difficult to identify
and quantify enrichment. Many experiments permit the direct
detection of 13C to determine isotopomer distribution;71−74
however, these experiments can still suffer from a lack of
sensitivity for isotopomers that are only partially enriched.
Although the enrichment of a given isotopomer may be minor,
this change may still be important to understanding the
biological mechanisms of the system or pathway. Alternatively,
1D 1H experiments can be used by taking advantage of the
splitting in the 1H spectrum resulting from 13C enrichment,
though 1H spectra at natural abundance can be convoluted
even without the additional complication arising from the extra
signals due to 13C couplings. To reduce the spectral complexity
arising from 13C enrichment, Cahoreau et al. used 2D
heteronuclear J-resolved spectroscopy (2D-JRES), which
records the scalar couplings in the second dimension instead
of chemical shifts, decreasing experiment time compared with
other 2D experiments.75 Other 2D experiments (such as
1
H−13C HSQC) can also be used to monitor isotopomer
enrichment by taking advantage of the dispersion of a second
dimension and the enhanced detection sensitivity of the 1H
nucleus coupled to 13C.76 A recent publication by Fan and Lane
applied a variety of 2D and 2D-edited experiments on either
[U−13C]-glucose or [U−13C, 15N]-glutamine enriched cancer
cells.77 The 2D experiments described include 1H−1H TOCSY,
dx.doi.org/10.1021/ac504075g | Anal. Chem. 2015, 87, 133−146
Analytical Chemistry
Figure 6. Workflow implemented in IsoDesign to determine the optimal label input in a 13C-MFA experiment. Blue, orange, and green boxes detail
steps performed by the calculation module, the visualization module, and a spreadsheet program, respectively. Reprinted from ref 92 with permission
from John Wiley & Sons Inc., copyright 2014.
1
H−1H HCCH TOCSY, 1H−13C HSQC, 1H−15N HSQC
HACACO (correlating carbonyl carbons with neighboring
protons), 1H−13C HSQC-TOCSY, and 1H−31P HSQC
TOCSY, and the authors discuss how the isotopomer distribution
is determined from the spectra. Despite isotopic enrichment,
traditional 2D experiments can still be time-consuming and less
amenable to high-throughput isotope tracing than 1D experiments.
The advent of ultrafast 2D NMR experiments, which allow the
acquisition of a multidimensional NMR experiment in a single
scan, has greatly reduced experiment times.78 Giraudeau and co-
workers measured the 13C enrichment in a biological sample with
a traditional TOCSY experiment, which had a run time of greater
than 2 h, compared to an ultrafast TOCSY with an experiment
time of 3 s on a 400 MHz spectrometer equipped with a room
temperature probe.58 Continued development and applications of
ultrafast 2D NMR experiments will significantly increase the
throughput of isotope tracing by NMR.
This section has predominantly discussed the use of isotope
labeling for the use of in vivo isotope labeling for the analysis of
biological pathways. Another interesting approach is to use
isotope labeling ex vivo to improve NMR sensitivity for general
metabolomics experiments. By incorporating an NMR active
isotope onto a tag, specific chemical moieties, such carboxylic
acid groups or amines, can be derivatized significantly
increasing sensitivity and selectivity while decreasing spectral
convolution.79,80 For example, Tayyari and co-workers designed
a cholamine tag with a 15N label on the primary amine, allowing
derivatization of carboxylic acid functional groups.81 The
authors demonstrate the sensitivity of the tag using 1H−15N
HSQC experiments by collecting only 1−4 scans over 128
increments. The tag also included a tertiary amine which
provides a stable, permanent positive charge attached to all
derivatized molecules, enhancing mass spectrometry sensitivity
as well. Although not yet widely used, the use of derivatization
to enhance NMR sensitivity and selectivity and decrease NMR
resonance overlap while enhancing cross-platform utility has
strong potential for improving NMR metabolite profiling
experiments.
Metabolic Flux Analysis. Metabolic flux analysis (MFA)
is a popular and growing topic in analytical chemistry and
139
Review
molecular biology and has recently been extensively re-
viewed.82,83 The ultimate goal of MFA is to establish the
metabolic flux through a particular pathway using labeled
reagents and to compare the flux under different stresses and in
various tissue types including cancers,84−86 oysters,87 plants,88
and bacteria.89 These analyses are often limited to specific, well-
defined biochemical pathways. Flux experiments measure time-
based isotope enrichments and require that the activities of the
enzymes involved in the pathway are known. For well-studied
pathways, enzymatic activities can often be approximated;
otherwise, parameters must be estimated to understand how
the pathways may be affected. Regardless, computational
methods are necessary to model the pathways and estimate
the flux parameters. One such program reported by Hettling
and co-workers estimated TCA fluxes with noisy NMR data in
heart tissue using a computational model to estimate the flux
parameters.90 In addition to data analysis, the actual design of a
MFA study can be challenging, as one must consider what
precursors and isotopic composition to use and the desired
outcome (dependent on the biological question under
investigation). Work by Millard et al. presents a novel software
program for the design of 13C-MFA experiments, termed
IsoDesign.91 As illustrated in Figure 6, the program allows the
investigator to generate models using different label inputs,
simulate the isotopic data, calculate the precision expected on
each flux for each label input, and outputs these results in
a visually interpreted sensitivity landscape. Isodesign can be
used for comparison of isotopic data from MS (including
MS/MS) and NMR (both 1H and 13C). Tools such as those
provided by Hettling et al., Millard et al., and others are
increasingly in need as more biological questions are being
probed with substrates labeled specifically for the investigation
of selected pathways.90−92
A major advantage of NMR over other popular MFA
analytical platforms is that it is nondestructive which allows for
in vivo NMR spectroscopy under specific conditions. A
comprehensive review and guide for in vivo NMR by de
Graaf and co-workers has recently been published and discusses
topics such as substrate choice, metabolic model, and choice of
NMR experiment (direct vs indirect detection).93 Although this
dx.doi.org/10.1021/ac504075g | Anal. Chem. 2015, 87, 133−146
Analytical Chemistry
portion of the Review will briefly cover a few of the recently
published in vivo (or in situ) NMR experiments, the reader is
directed to the work by de Graaf and co-workers for a more in-
depth discussion of the many considerations for in vivo 13C
NMR. Solid state NMR (ssNMR) can also be used for
MFA/isotope tracing. Work by Norris et al. on tree seedlings
investigated both 13C and 15N allocation as a function of time,
allowing the researchers to monitor several different “pulses” of
tracer and the spatial dispersion of the tracer before/after
administration.94 Although the high spin rates can result in
detrimental effects on an in vivo system, it may provide a
qualitative evaluation of substrate localization. Another
approach that has been applied to study substrate localization,
while avoiding the physiological stress associated with high spin
rates of ssNMR, is to use a “metabolomics-on-a-chip” method
for MFA. Ouattara and co-workers recently reported the
application of a chip-based approach to explore the metabolic
network of a hepatoma cell line.95 The authors compared the
chip environment with that of a traditional Petri dish and were
able to determine that the cells in the chip showed evidence of
a higher oxygenation rate, indicating a more oxygen-rich
environment. Although not necessarily an in vivo analysis, the
biochip method does allow for MFA to model biochemical
pathways and to deduce intracellular metabolite distributions in
controlled environments.
Hyperpolarization and DNP. For much of its history,
NMR has been considered to be a low sensitivity technique
because small energy differences give rise to similar populations
of the nuclear spin states at ambient temperatures. Hyper-
polarization is a method of increasing polarization by altering
the populations away from their thermal equilibrium value,
enhancing sensitivity by several orders of magnitude. Because
the energy difference between two spin states is directly
proportional to the magnitude of the applied magnetic field, a
simple (but expensive) mechanism for increasing polarization is
to perform the measurements in a higher field magnet.
However, the magnetic field required to reach a polarization
of even 10% is outside the current technical limits. Another way
to increase polarization is by cooling to very low temperatures;
for example, at 0.1 K, the polarization of protons reaches 23%.96
Though theoretically possible, measurements of NMR spectra
of solid samples at such low temperatures would be quite
challenging and the processes driving polarization buildup at
low temperatures become very slow.
Hyperpolarization of various nuclei has proven to be a
powerful approach for enhancing the use of NMR for molecular
and cellular imaging.97 For noble gases like 129Xe and 3He,
hyperpolarization can be accomplished by optical pumping of
alkali metal atoms (typically Rb) which then transfer their
polarization to the noble gas via spin exchange, increasing the
NMR signal by 10 000-fold. 129Xe gas has been used in many in
vivo imaging studies, particularly those targeting the lungs.
Palaniappan et al. review the development of novel 129Xe
imaging agents composed of cryptophane cages coupled with a
biological recognition element like an antibody.98 Crypto-
phanes bind xenon with moderate affinity (∼4000 M−1), and
residence times in the 30−300 ms range make them well-suited
for in vivo MRI experiments. Another approach to hyper-
polarization involves reactions of para-hydrogen. Gaseous
hydrogen can exist in two nuclear spin isomers (ortho- and
para-). The concentration of the para-isomer can be increased
by cooling the gas and inducing the conversion using a cata-
lyst. This para-hydrogen induced polarization (PHIP) can be
140
Review
transferred to organic molecules by reacting hyperpolarized
hydrogen with precursor molecules in the presence of a
transition metal catalyst. The use of PHIP for imaging,
including the use of 13C or 15N labeled metabolite precursors,
has been reviewed by Glöggler et al.99
An exciting development in hyperpolarization experiments for
metabolic profiling is the rapidly expanding use of dynamic
nuclear polarization (DNP), which has been accelerated by the
introduction of commercial instruments that facilitate its
implementation. An excellent overview of DNP in liquids has
been recently authored by Günther.97 DNP takes advantage of
the comparably larger polarization of electron spins, which can
be nearly 100% at low temperatures. Saturation of the electron
spin transfers this polarization from the unpaired electrons to
nuclear spin via hyperfine and dipolar interactions. Though
DNP experiments can be executed in a variety of experimental
arrangements, the most useful configuration for metabolic
profiling experiments is dissolution DNP, in which polarization
is carried out at low temperatures (<1.5 K) and the polarized
liquid sample is generated by dissolution of the polarized solid
and transferred to an NMR magnet for acquisition of the
NMR spectrum.
13
C-labeled compounds are widely used in dissolution DNP
studies (in vitro and in vivo) because of their large chemical
shift range (∼200 ppm), low natural abundance (1.1%), which
gives rise to a weak background, and the possibility of long
T1 relaxation times depending on the structural environ-
ment.100 DNP has been shown to increase the sensitivity of 13C
detection by more than 10 000-fold enabling its use to probe
enzymatic processes in vivo.101 For example, DNP has been
used to follow the kinetics of pyruvate metabolism in cancer
cells85,102 and in vivo in a breast cancer mouse model103 and to
monitor the metabolism of U−13C2H7−glucose in perfused
human breast cancer cells.104 Schilling et al. demonstrated
the measurement of apparent diffusion coefficients using
hyperpolarized 13C lactate generated in vivo from [1-13C]-
pyruvate in tumor spheroids.105 Central carbon metabolism
has also been probed in Saccharomyces cerevisiae106,107 includ-
ing an investigation of the effects of sulfite as an inhibitor of
glycolysis.108
Transfer of hyperpolarization to protons through indirect
detection experiments using spatially selective coherence trans-
fers offers a promising tool for implementation with standard
MRI scanners.109 As illustrated in Figure 7, the creative design
and application of hyperpolarized NMR probes can enable real-
time biological assays of probe uptake and export, pH, redox
state, reactive oxygen species, ion concentrations, signaling
pathways, etc.110 Hyperpolarized 89Y complexes are potentially
attractive as NMR or MRI imaging probes because of the
unusually long T1 relaxation times (e.g., 10 min) typical of this
very low γ nucleus. Lumata et al. demonstrate optimized
DNP experiments leading to signal enhancements for 89Y over
60 000 times the thermal signal.111
■ DATA PROCESSING AND CHEMOMETRICS
A significant challenge in metabolomics is the amount of data
obtained from a single experiment. Even for NMR, which is
inherently less sensitive than mass spectrometry, hundreds
of peaks are often detected and only a subset of these is iden-
tifiable and quantifiable. For example, Bouatra et al. recently
published a report on the human urine metabolome identi-
fying some 179 peaks by NMR alone.11 Data processing
and chemometrics analyses convert NMR spectral data into
dx.doi.org/10.1021/ac504075g | Anal. Chem. 2015, 87, 133−146
Analytical Chemistry Review

Figure 7. Principle of biological assays using hyperpolarized NMR probes. Hyperpolarization is optimized ex situ, and the hyperpolarized probe or
label is added to a biomolecule, cell extracts, or living cells to conduct biological assays for detection inside an NMR spectrometer. Reprinted from
ref 110. Published under Creative Commons 4.0, copyright 2014.

Figure 8. Principal component analysis scores plots of (a) 16 metabolites fitted independently from round one; (b) same as for part (a) but with one
metabolite (threonine) and one outlying sample (QXB1) excluded; (c) 37 metabolites fitted on the basis of exemplar spectra in round two, with one
outlying sample (QXB1) excluded. Data were log transformed and mean centered. Reprinted from ref 112. Copyright 2011 American Chemical
Society.

information that can be used to address the particular biological
problem being interrogated.
Spectral Processing. Before the NMR data can be
interrogated for statistical changes, the raw data must be
processed, including baseline correction, phasing, apodization,
referencing, integration, and normalization. This can be
accomplished manually, but manual data processing is time-
consuming and can be less reliable than automated processing.
Because of the time associated with data processing, several
groups have been working on improving and automating
different aspects (or the entirety) of the process. Data
processing steps should be treated systematically to minimize
introduction of errors. Tredwell et al. conducted a variability
study in which 5 different analysts independently processed
the same data set.112 The five participants were asked to use
commercially available software (Chenomx NMR Suite) to fit
metabolite resonances in the same spectra of yeast extracts.
Comparison of the results obtained by each analyst gave an
overall coefficient of variation (CV) of 20%. In contrast, when
a single person performed the analysis of the same data
5 times, the CV was only 2.4%. The variance between analysts
was largely attributed to differences in fitting spectral regions
having extensive peak overlap. Figure 8a shows PCA score
plots representing the 16 metabolites identified and quantified
by all 5 participants. Person 2 appears to be an outlier in
all samples, but with the removal of threonine and one out-
lier sample (QXB1), Figure 8b shows better clustering.112
Figure 8c shows the results of a second round of data
processing in which spectra were first processed by
automation and then refit by each analyst, resulting in
identification of 37 metabolites. The availability of a single
141
pipeline for the preprocessing and chemometric analysis of
NMR data could help to reduce analyst-dependent variations.
Such a pipeline would ideally include Fourier transformation,
zero filling, apodization, phase correction, referencing, align-
ment, integration, and inclusion of various chemometric
approaches (such as PCA, PLS, etc.). Recently, Worley and
Powers presented a package, termed MVAPACK, for the
complete handling of NMR data, from preprocessing, through
alignment, integration, and chemometrics.113 Although not
necessarily entirely automated, the package does allow for
importing both Bruker and Agilent data through secondary
programs.
The individual aspects of the NMR preprocessing pipeline
have also been addressed. Improvements in addressing phase
correction,114 baseline correction,115 alignment,116−118 normal-
ization,119,120 integration/binning,121 deconvolution,122−124 and
quantitation125 have also been recently published. An interest-
ing approach reported by Romano et al. provides a program
PRICONA that uses the free induction decay (FID), rather
than the processed spectrum for PCA analysis, and allevi-
ates the need for any operator manipulation.126 Continued
improvement and incorporation of emerging techniques for
NMR data preprocessing is essential for improving the high-
throughput characteristics of NMR for metabolite profiling.
Figure 9 provides examples of resonance deconvolution using
an automated quantum mechanical total line shape model
(QMTLS) for analysis of the NMR spectrum of a serum
sample treated by ultrafiltration (UF).127 Figure 9b highlights a
deconvoluted region in which the acetate resonance is over-
lapped with the multiplets of arginine and lysine. The quality
of each fit was determined using relative-root-square-error
dx.doi.org/10.1021/ac504075g | Anal. Chem. 2015, 87, 133−146
Analytical Chemistry
Figure 9. Examples of deconvolution with doublets of valine, isoleucine,
and ketoleucine and triplets of leucine and isoleucine (A) and singlet of
acetate and complex multiplets of arginine and lysine (B) of a single UF
serum sample. Reprinted from ref 127 with kind permission from
Springer Science and Business Media, copyright 2014.
(RRMSE), and a good fit was characterized by a low RRMSE
value as well as a high R factor.
Resonance Assignments and Databases. One of the
major challenges of NMR metabolic profiling is the ability to
make resonance assignments from the obtained spectral data.
Often, 1H NMR spectra have a high degree of spectral overlap
making assignments challenging.128 Peak assignments in these
cases can be made with the help of 2D HSQC spectra.
The added dispersion in the 13C dimension improves the
resolution of peaks overlapped in the 1H spectrum and adds
13
C chemical shifts as an additional parameter to aid in
resonance assignments.128
While multidimensional NMR experiments can aid in the
process of assigning resonances, this strategy can be time-
consuming.129 When the identity of the metabolites in a sample
are known (or suspected), resonance assignments can be
facilitated using libraries or databases of 1H NMR spectra such
as HMDB and MMCD.130 These databases typically display
spectra as static images which do not allow for detailed
inspection of a particular region,131 lack the ability to perform
automated assignments, and can be time-consuming to
search.132 While databases commonly provide lists of chemi-
cal shifts for each metabolite, they may lack information about
coupling constants. Coupling constants provide information
about atom connectivity, stereochemistry, and molecular con-
formation making them useful for resonance assignments,
especially in complex spectra.131 Many investigators have begun
developing their own databases including MetaboID,129
Metabohunter,130 MetIDB,131 MeRy-B,133 and BML-NMR.134
142
Review
Quantitation. The popularity of quantitative NMR (qNMR)
has grown substantially over the past two decades. In
metabolomics and metabolic profiling experiments, qNMR can
provide absolute or relative quantification of multiple metabolites
within a sample without prior separation of components.135,136
While qNMR is a useful technique for metabolite quantification,
it is a demanding process with complications arising from water
suppression, T2-editing, protein interactions, relaxation differ-
ences, macromolecule background, and spectral overlap.137,138
To address these complications, quantification software
applications have been developed including MetaboQuant139
and IQMNMR.140 Several alternatives to quantitative 1D 1H
NMR spectra have been introduced, such as 2-D [1H−13C]
HSQC,141 2-D 1H INADEQUATE,142 and 1-D TOCSY.143
While 2-D [1H−13C] HSQC spectra offer the benefit of
reduced spectral overlap and improved resolution, metabo-
lite quantification can be compromised by resonance-specific
signal attenuation during coherence transfer delays.141 Hu
et al. demonstrated that, for accurate quantitation, a series of
HSQC spectra can be acquired with incremented repetition
times. Back extrapolation of the data yields a time-zero HSQC
spectrum in which metabolite resonance intensities are directly
proportional to concentrations.141 Because their acquisition
can be time-consuming, these 2D-experiments benefit from the
use of cryogenic probes, which increase measurement
sensitivity by reducing noise and thereby decreasing time
spent signal averaging or by the use of fast/ultrafast NMR
experiments.
Chemometrics. Characterization of the processed NMR
data is another important facet of the NMR metabolomics
pipeline. With hundreds of resonances possible in a biological
sample, making sense of the peak identities and intensity
changes can be challenging. Such analyses are often performed
using complicated algorithms and combining several chemo-
metric approaches to indicate the important resonances prior to
metabolite identification. Several different approaches were
evaluated for their ability to facilitate classification using five
biofluid 1H NMR data sets, including urine, plasma, milk, and
serum.144 The methods interrogated by Hochrein et al. include
Support Vector Machines, Elastic Net, Random Forests, Top
Scoring Pairs, and Nearest Shrunken Centroids. The authors
found that, unsurprisingly, many methods were dependent on
the data set being investigated. Only Support Vector Machines
and Random Forests combined with other statistical ap-
proaches for filtering gave the best performance for a majority
of the data sets.144 Similarly, Maraschin et al. investigated the
combination of several different methods to classify the
geographical origin of a native Brazilian plant and found that
a combination of several algorithms was optimal.145 Other
advances in chemometrics include the application of techniques
specifically for certain NMR experiments in which multidimen-
sional data sets (such as two-dimensional J-resolved spectros-
copy) are decomposed into matrices that describe the different
components of the 2D experiment (J-couplings, chemical shift,
and intensities).124 The resultant matrices can then be
subjected to traditional multivariate statistical analysis with-
out sacrificing the additional information provided by the
2D experiment.
Other methods for reducing the complexity of NMR data
include the statistical integration of NMR data from different
biofluids/tissues from the same specimen,146 removal of inter-
fering resonances from pharmaceuticals,147 or high concen-
trations of metabolites, such as sugars in diabetic patients that
dx.doi.org/10.1021/ac504075g | Anal. Chem. 2015, 87, 133−146
Analytical Chemistry Review

Figure 10. Comparison of STOCSY and STORM for the identification of N-methylnicotinic acid signals in the INTERMAP BMI nondiabetic U.S.
data set. Scatterplot of correlations determined by STOCSY (x-axis, n = 1880) and by STORM (y-axis, n = 112), using an N-methylnicotinic acid
(NMNA) peak at δ 8.84 as driver, is shown. For each axis, the kernel density estimation of the distribution of each class is shown in the same color
on the opposite side. NMNA is shown in orange, hippuric acid in green, an unknown metabolite in light blue, and all other variables in dark blue.
For each of the metabolites, each signal is shown with a different symbol. For clarity, each distribution has been normalized to the same total area.
It shows that the density estimate of NMNA overlaps less with others in STORM compared to STOCSY. Reprinted from ref 149. Copyright 2012
American Chemical Society.

might be obscuring other resonances.123 Wei et al. addresses
the problem of identifying resonances in a complex sample
using ratio analysis NMR spectroscopy (RANSY).128 Like
STOCSY (statistical total correlation spectroscopy),148 RANSY
is a statistical approach centered on identifying resonances
of the same molecule from a series of spectra, though it does
not rely on statistical correlations. Instead with RANSY, the
ratios of the detected peaks are used to determine molecular
connectivity. Recent variations of STOCSY have also been
reported including iterative STOCSY (I-STOCSY)132 and
STORM (subset optimization by reference matching).149
STORM provides subset optimization by reference matching
using a 3-step algorithm that includes subset selection,
STOCSY of the subset, and reference updating. One downfall
of STOCSY is difficulty in identifying low concentration metab-
olites, but Posma et al. posit that this is improved in STORM
through the addition of subset selection.149 Figure 10 provides
a comparison of STORM (y-axis) and STOCSY (x-axis) for
resonances in a large data set of urine samples from nondiabetic
patients.149 Note that the N-methylnicotinic acid (NMNA)
density estimate, shown in orange, has less overlap with other
components using STORM, making it possible to recover rare
signals obscured by resonance overlap. Another promising
technique for biomarker identification is visual interpretation of
z-score ratios (VIZR) developed by Barding et al.150 VIZR
identifies unique resonances using ratios of z-scores calculated
by comparing a spectrum obtained for a subject sample to
143
spectra measured for a collection of samples from a general
population. Additional approaches for identifying and quantify-
ing substituents in complex samples have also been published
or reviewed151−153 and are of continuing interest to the
bioanalytical community.
■ PERSPECTIVE
NMR-based metabolomics has rapidly become established as a
useful approach for the analysis of complex biosamples
contributing to biomarker discovery, disease diagnosis, and
treatment monitoring as well as providing insights into the
impact of environmental stressors on organisms and eco-
systems. In addition to untargeted metabolomics measure-
ments, isotopic labeling and metabolic flux experiments provide
unique insights into specific metabolic pathways. Innovations
related to automated data processing, resonance assignments,
and statistical data analysis continue to improve the reliability
and robustness of NMR metabolomics results. The introduc-
tion of fast and ultrafast multidimensional NMR experiments
promises to greatly improve the throughput of NMR
measurements. In addition, approaches that affect nuclear
polarization such as DNP dramatically enhance NMR
sensitivity. Rather than choosing between NMR and mass
spectrometry-based metabolomics, strategies that make use
of both platforms provide better coverage of the metabolome
and improved measurement reliability. Chemometric tools
that integrate the results of NMR, GC/MS, and LC-MS and
dx.doi.org/10.1021/ac504075g | Anal. Chem. 2015, 87, 133−146
Analytical Chemistry
results from other analytical techniques will benefit the adop-
tion of multiplatform metabolomics approaches.
■
AUTHOR INFORMATION
Corresponding Author
*E-mail: clarive@ucr.edu.
Notes
The authors declare no competing financial interest.
Biographies
Cynthia K. Larive is a Professor of Chemistry at the University of
California−Riverside (UCR) and currently serves as Divisional Dean
within the College of Natural and Agricultural Sciences. She received
her B.S. degree from South Dakota State University in 1980 and a
Ph.D. from UCR in 1992 under the direction of Dallas Rabenstein.
She was on the faculty of the University of Kansas from 1992 to 2004,
before returning to UCR in 2005. Larive is a fellow of the AAAS and
ACS and has received awards for teaching, research, and service. Her
research interests are in the application of NMR, GC/MS and LC-MS
for metabolomics, and metabolic profiling in a variety of biological
systems and in the structural characterization of the glycosaminogly-
cans heparin and heparan sulfate.
Gregory A. Barding, Jr. is an Assistant Professor of Chemistry at
California Polytechnic University, Pomona. He received his B.S.
degree from California State University−San Bernardino in 2008 and
his Ph.D. degree in 2013 from UCR where he was mentored jointly by
Cynthia Larive and Julia Bailey-Serres in the Department of Botany
and Plant Sciences. He was a postdoctoral scientist at the University of
Washington School of Medicine in the laboratory of professor Daniel
Raftery from 2013 to 2014. Dr. Barding’s research interests are focused
on applying various analytical techniques including GC, LC, NMR,
UV−vis, and mass spectrometry for general metabolite profiling
studies as well as stable isotope tracing in several biological systems.
Meredith M. Dinges is a Ph.D. candidate at UCR under the mentorship
of Cynthia Larive. She received her B.A. degree from Drury University
in 2012. Her research interests are focused on metabolomics and
metabolic profiling of fecal samples along the length of the colon to
understand the dynamic processes related to microbial metabolism
and transport of metabolites across the lumen and into the
bloodstream. She carries out her research in collaboration with
Professor Christian Lytle in the UCR School of Medicine.
■
ACKNOWLEDGMENTS
Support by the National Science Foundation (Grant No. IOS-
1121626) and the U.S. Department of Agriculture, National
Institute of Food and Agriculture - Agriculture and Food
Research Initiative (Grant No. 2011−04015) to C.K.L. is
gratefully acknowledged.
■
REFERENCES
(1) Dunn, W. B.; Broadhurst, D. I.; Atherton, H. J.; Goodacre, R.;
Griffin, J. L. Chem. Soc. Rev. 2011, 40, 387−426.
(2) Robinette, S. L.; Holmes, E.; Nicholson, J. K.; Dumas, M. E.
Genome Med. 2012, 4, 30.
(3) Bothwell, J. H. F.; Griffin, J. L. Biol. Rev. 2011, 86, 493−510.
(4) Gebregiworgis, T.; Powers, R. Comb. Chem. High Throughput
Screening 2012, 15, 595−610.
(5) Emwas, A.-H. M.; Salek, R. M.; Griffin, J. L.; Merzaban, J.
Metabolomics 2013, 9, 1048−1072.
(6) Mikkonen, J. J. W.; Herrala, M.; Soininen, P.; Lappalainen, R.;
Tjaderhane, L.; Seitsalo, H.; Niemela, R.; Salo, T.; Kullaa, A. M.;
Myllymaa, S. Metabolomics 2013, 3, 128.
144
Review
(7) Duarte, I. F.; Diaz, S. O.; Gil, A. M. J. Pharm. Biomed. Anal. 2014,
93, 17−26.
(8) Duarte, I. F.; Gil, A. M. Prog. Nucl. Magn. Reson. Spectrosc. 2012,
62, 51−74.
(9) Hyndman, M. E.; Mullins, J. K.; Bivalacqua, T. J. Urol. Oncol.:
Semin. Orig. Invest. 2011, 29, 558−561.
(10) Nicholson, G.; Rantalainen, M.; Maher, A. D.; Li, J. V.;
Malmodin, D.; Ahmadi, K. R.; Faber, J. H.; Hallgrimsdottir, I. B.;
Barrett, A.; Toft, H.; Krestyaninova, M.; Viksna, J.; Neogi, S. G.;
Dumas, M. E.; Sarkans, U.; Silverman, B. W.; Donnelly, P.; Nicholson,
J. K.; Allen, M.; Zondervan, K. T.; Lindon, J. C.; Spector, T. D.;
McCarthy, M. I.; Holmes, E.; Baunsgaard, D.; Holmes, C. C. Mol. Syst.
Biol. 2011, 7, 525.
(11) Bouatra, S.; Aziat, F.; Mandal, R.; Guo, A. C.; Wilson, M. R.;
Knox, C.; Bjorndahl, T. C.; Krishnamurthy, R.; Saleem, F.; Liu, P.;
Dame, Z. T.; Poelzer, J.; Huynh, J.; Yallou, F. S.; Psychogios, N.;
Dong, E.; Bogumil, R.; Roehring, C.; Wishart, D. S. PLoS One 2013, 8,
No. e73076.
(12) Scano, P.; Locci, E.; Noto, A.; Navarra, G.; Murgia, F.; Lussu,
M.; Barberini, L.; Atzori, L.; De Giorgio, F.; Rosa, M. F.; d’Aloja, E.
Magn. Reson. Chem. 2013, 51, 454−462.
(13) Izquierdo-García, J. L.; Nin, N.; Ruíz-Cabello, J.; Rojas, Y.; de
Paula, M.; López-Cuenca, S.; Morales, L.; Martínez-Caro, L.;
Fernández-Segoviano, P.; Esteban, A.; Lorente, J. A. Intensive Care
Med. 2011, 37, 2023−2032.
(14) DeFeo, E. M.; Wu, C. L.; McDougal, W. S.; Cheng, L. L. Nat.
Rev. Urol. 2011, 8, 301−311.
(15) Zhang, B.; Halouska, S.; Schiaffo, C. E.; Sadykov, M. R.;
Somerville, G. A.; Powers, R. J. Proteome Res. 2011, 10, 3743−3754.
(16) Garcia-Á lvarez, I.; Fernández-Mayoralas, A.; Garrido, L. Curr.
Top. Med. Chem. 2011, 11, 27−42.
(17) Duarte, I. F. J. Controlled Release 2011, 153, 34−39.
(18) Rolin, D.; Deborde, C.; Maucourt, M.; Cabasson, C.; Fauvelle,
F.; Jacob, D.; Canlet, C.; Moing, A. In Metabolomics Coming of Age with
Its Technological Diversity; Rolin, D., Ed; Adv. Bot. Res. 67; Academic
Press: London, UK, 2013; pp 1−66.
(19) Barding, G. A., Jr.; Fukao, T.; Beni, S.; Bailey-Serres, J.; Larive,
C. K. J. Proteome Res. 2012, 11, 337−347.
(20) Barding, G. A., Jr.; Fukao, T.; Beni, S.; Bailey-Serres, J.; Larive,
C. K. J. Proteome Res. 2013, 12, 898−909.
(21) Orr, D. J.; Barding, G. A.; Tolley, C. E.; Hicks, G. R.; Raikhel, N.
V.; Larive, C. K. In Plant Chemical Genomics: Methods and Protocols;
Hicks, G. R., Robert, S., Eds.; Methods Mol. Biol. 1056; Springer: New
York, USA, 2014; pp 225−239.
(22) Kim, H. K.; Choi, Y. H.; Verpoorte, R. Trends Biotechnol. 2011,
29, 267−275.
(23) Serra, O.; Chatterjee, S.; Huang, W. L.; Stark, R. E. Plant Sci.
2012, 195, 120−124.
(24) Leiss, K. A.; Choi, Y. H.; Verpoorte, R.; Klinkhamer, P. G. L.
Phytochem. Rev. 2011, 10, 205−216.
(25) Forseth, R. R.; Schroeder, F. C. Curr. Opin. Chem. Biol. 2011, 15,
38−47.
(26) Robinette, S. L.; Bruschweiler, R.; Schroeder, F. C.; Edison, A. S.
Acc. Chem. Res. 2012, 45, 288−297.
(27) Lankadurai, B. P.; Nagato, E. G.; Simpson, M. J. Environ. Rev.
2013, 21, 180−205.
(28) Ogata, Y.; Chikayama, E.; Morioka, Y.; Everroad, R. C.; Shino,
A.; Matsushima, A.; Haruna, H.; Moriya, S.; Toyoda, T.; Kikuchi, J.
PLoS One 2012, 7, No. e30263.
(29) Heyman, H. M.; Meyer, J. J. M. S. Afr. J. Bot. 2012, 82, 21−32.
(30) Hong, Y. S. Magn. Reson. Chem. 2011, 49, S13−S21.
(31) Abriata, L. A. Concepts Magn. Reson., Part A 2012, 40A, 171−
178.
(32) Zhang, A. H.; Sun, H.; Qiu, S.; Wang, X. J. Magn. Reson. Chem.
2013, 51, 549−556.
(33) Keun, H. C.; Athersuch, T. J. In Metabolic Profiling: Methods and
Protocols, Metz, T. O., Ed.; Methods Mol. Biol. 708; Springer: New
York, USA, 2011; pp 321−334.
dx.doi.org/10.1021/ac504075g | Anal. Chem. 2015, 87, 133−146
Analytical Chemistry
(34) Lacy, P.; McKay, R. T.; Finkel, M.; Karnovsky, A.; Woehler, S.;
Lewis, M. J.; Chang, D.; Stringer, K. A. PLoS One 2014, 9, No. e85732.
(35) Simon-Manso, Y.; Lowenthal, M. S.; Kilpatrick, L. E.; Sampson,
M. L.; Telu, K. H.; Rudnick, P. A.; Mallard, W. G.; Bearden, D. W.;
Schock, T. B.; Tchekhovskoi, D. V.; Blonder, N.; Yan, X. J.; Liang, Y.
X.; Zheng, Y. F.; Wallace, W. E.; Neta, P.; Phinney, K. W.; Remaley, A.
T.; Stein, S. E. Anal. Chem. 2013, 85, 11725−11731.
(36) Rasmussen, L. G.; Savorani, F.; Larsen, T. M.; Dragsted, L. O.;
Astrup, A.; Engelsen, S. B. Metabolomics 2011, 7, 71−83.
(37) Pinto, J.; Domingues, M. R. M.; Galhano, E.; Pita, C.; Almeida,
M. D.; Carreira, I. M.; Gil, A. M. Analyst 2014, 139, 1168−1177.
(38) Gowda, G. A. N.; Raftery, D. Anal. Chem. 2014, 86, 5433−5440.
(39) Jiang, L. M.; Huang, J.; Wang, Y. L.; Tang, H. R. Analyst 2012,
137, 4209−4219.
(40) Matheus, N.; Hansen, S.; Rozet, E.; Peixoto, P.; Maquoi, E.;
Lambert, V.; Noël, A.; Frédérich, M.; Mottet, D.; de Tullio, P.
Phytochem. Anal. 2014, 25, 342−349.
(41) Martineau, E.; Tea, I.; Loaec, G.; Giraudeau, P.; Akoka, S. Anal.
Bioanal. Chem. 2011, 401, 2133−2142.
(42) Jacobs, D. M.; Spiesser, L.; Garnier, M.; de Roo, N.; van
Dorsten, F.; Hollebrands, B.; van Velzen, E.; Draijer, R.; van
Duynhoven, J. Anal. Bioanal. Chem. 2012, 404, 2349−2361.
(43) van der Hooft, J. J. J.; de Vos, R. C. H.; Mihaleva, V.; Bino, R. J.;
Ridder, L.; de Roo, N.; Jacobs, D. M.; van Duynhoven, J. P. M.;
Vervoort, J. Anal. Chem. 2012, 84, 7263−7271.
(44) van der Hooft, J. J. J.; Mihaleva, V.; de Vos, R. C. H.; Bino, R. J.;
Vervoort, J. Magn. Reson. Chem. 2011, 49, S55−S60.
(45) van der Hooft, J. J. J.; de Vos, R. C. H.; Ridder, L.; Vervoort, J.;
Bino, R. J. Metabolomics 2013, 9, 1009−1018.
(46) McKay, R. T. Concepts Magn. Reson., Part A 2011, 38A, 197−
220.
(47) Farooq, H.; Soong, R.; Courtier-Murias, D.; Anklin, C.;
Simpson, A. Anal. Chem. 2012, 84, 6759−6766.
(48) Martin-Pastor, M. J. Agric. Food Chem. 2014, 62, 1190−1197.
(49) Huang, Y. Q.; Cal, S. H.; Zhang, Z. Y.; Chen, Z. Biophys. J. 2014,
106, 2061−2070.
(50) Huang, Y. Q.; Zhang, Z. Y.; Cai, S. H.; Chen, Z. PLoS One 2013,
9, No. e84032.
(51) Shemesh, N.; Dumez, J. N.; Frydman, L. Chem.Eur. J. 2013,
19, 13002−13008.
(52) Hoch, J. C.; Maciejewski, M. W.; Mobli, M.; Schuyler, A. D.;
Stern, A. S. Acc. Chem. Res. 2014, 47, 708−717.
(53) Rai, R. K.; Sinha, N. Anal. Chem. 2012, 84, 10005−10011.
(54) Pathan, M.; Akoka, S.; Tea, I.; Charrier, B.; Giraudeau, P.
Analyst 2011, 136, 3157−3163.
(55) Le Guennec, A.; Giraudeau, P.; Caldarelli, S. Anal. Chem. 2014,
86, 5946−5954.
(56) Pathan, M.; Akoka, S.; Giraudeau, P. J. Magn. Reson. 2012, 214,
335−339.
(57) Giraudeau, P.; Cahoreau, E.; Massou, S.; Pathan, M.; Portais, J.
C.; Akoka, S. ChemPhysChem 2012, 13, 3098−3101.
(58) Giraudeau, P.; Massou, S.; Robin, Y.; Cahoreau, E.; Portais, J.-
C.; Akoka, S. Anal. Chem. 2011, 83, 3112−3119.
(59) Giraudeau, P.; Akoka, S. In Metabolomics Coming of Age with Its
Technological Diversity; Rolin, D., Ed; Adv. Bot. Res. 67; Academic
Press: London, UK, 2013; pp 99−158.
(60) Renslow, R. S.; Babauta, J. T.; Majors, P. D.; Mehta, H. S.;
Ewing, R. J.; Ewing, T. W.; Mueller, K. T.; Beyenal, H. Water Sci.
Technol. 2014, 69, 966−973.
(61) Ramaswamy, V.; Hooker, J. W.; Withers, R. S.; Nast, R. E.; Brey,
W. W.; Edison, A. S. J. Magn. Reson. 2013, 235, 58−65.
(62) Tadanki, S.; Colon, R. D.; Moore, J.; Waddell, K. W. J. Magn.
Reson. 2012, 223, 64−67.
(63) Grimes, J. H.; O’Connell, T. M. J. Biomol. NMR 2011, 49, 297−
305.
(64) Falck, D.; Oosthoek-de Vries, A. J.; Kolkman, A.; Lingeman, H.;
Honing, M.; Wijmenga, S. S.; Kentgens, A. P. M.; Niessen, W. M. A.
Anal. Bioanal. Chem. 2013, 405, 6711−6720.
(65) Ryan, H.; Smith, A.; Utz, M. Lab Chip 2014, 14, 1678−1685.
145
Review
(66) Hu, J. Z. In Metabolic Profiling: Methods and Protocols; Metz, T.
O., Ed.; Methods Mol. Biol. 708; Springer: New York, USA, 2011; pp
335−364.
(67) Lazariev, A.; Allouche, A. R.; Aubert-Frecon, M.; Fauvelle, F.;
Piotto, M.; Elbayed, K.; Namer, I. J.; van Ormondt, D.; Graveron-
Demilly, D. Meas. Sci. Technol. 2011, 22, 114020.
(68) Wong, A.; Jiménez, B.; Li, X. N.; Holmes, E.; Nicholson, J. K.;
Lindon, J. C.; Sakellariou, D. Anal. Chem. 2012, 84, 3843−3848.
(69) Wong, A.; Li, X. N.; Sakellariou, D. Anal. Chem. 2013, 85,
2021−2026.
(70) Wong, A.; Li, X. N.; Molin, L.; Solari, F.; Elena-Herrmann, B.;
Sakellariou, D. Anal. Chem. 2014, 86, 6064−6070.
(71) El Hage, M.; Conjard-Duplany, A.; Baverel, G.; Martin, G.
Neurochem. Int. 2011, 58, 896−903.
(72) Gonzalez-Rodriguez, I.; Gaspar, P.; Sanchez, B.; Gueimonde,
M.; Margolles, A.; Neves, A. R. Appl. Environ. Microb. 2013, 79, 7628−
7638.
(73) Li, W.; Bian, F.; Chaudhuri, P.; Mao, X.; Brunengraber, H.; Yu,
X. NMR Biomed. 2011, 24, 176−187.
(74) Marin-Valencia, I.; Cho, S. K.; Rakheja, D.; Hatanpaa, K. J.;
Kapur, P.; Mashimo, T.; Jindal, A.; Vemireddy, V.; Good, L. B.;
Raisanen, J.; Sun, X.; Mickey, B.; Choi, C.; Takahashi, M.; Togao, O.;
Pascual, J. M.; DeBerardinis, R. J.; Maher, E. A.; Malloy, C. R.; Bachoo,
R. M. NMR Biomed. 2012, 25, 1177−1186.
(75) Cahoreau, E.; Peyriga, L.; Hubert, J.; Bringaud, F.; Massou, S.;
Portais, J.-C. Anal. Biochem. 2012, 427, 158−163.
(76) Gallinger, A.; Biet, T.; Pellerin, L.; Peters, T. Angew. Chem., Int.
Ed. 2011, 50, 11672−11674.
(77) Fan, T. W.-M.; Lane, A.N. J. Biomol. NMR 2011, 49, 267−280.
(78) Frydman, L.; Scherf, T.; Lupulescu, A. Proc. Natl. Acad. Sci.
U.S.A. 2002, 99, 15858−15862.
(79) Shanaiah, N.; Desilva, M. A.; Gowda, G. A. N.; Raftery, M. A.;
Hainline, B. E.; Raftery, D. Proc. Natl. Acad. Sci. U.S.A. 2007, 104,
11540−11544.
(80) Ye, T.; Mo, H.; Shanaiah, N.; Gowda, G. A. N.; Zhang, S.;
Raftery, D. Anal. Chem. 2009, 81, 4882−4888.
(81) Tayyari, F.; Gowda, G. A. N.; Gu, H. W.; Raftery, D. Anal.
Chem. 2013, 85, 8715−8721.
(82) Sims, J. K.; Manteiga, S.; Lee, K. Curr. Opin. Biotechnol. 2013,
24, 933−939.
(83) Fernie, A. R.; Morgan, J. A. Plant Cell Environ. 2013, 36, 1738−
1750.
(84) Yang, Y.; Lane, A. N.; Ricketts, C. J.; Sourbier, C.; Wei, M.-H.;
Shuch, B.; Pike, L.; Wu, M.; Rouault, T. A.; Boros, L. G.; Fan, T. W.
M.; Linehan, W. M. PLoS One 2013, 8, No. e72179.
(85) Yang, C.; Harrison, C.; Jin, E. S.; Chuang, D. T.; Sherry, A. D.;
Malloy, C. R.; Merritt, M. E.; DeBerardinis, R. J. J. Biol. Chem. 2014,
289, 6212−6224.
(86) Cheng, T.; Sudderth, J.; Yang, C.; Mullen, A. R.; Jin, E. S.;
Mates, J. M.; DeBerardinis, R. J. Proc. Natl. Acad. Sci. U.S.A. 2011, 108,
8674−8679.
(87) Tikunov, A. P.; Stoskopf, M. K.; Macdonald, J. M. Metabolites
2014, 4, 53−70.
(88) Gilbert, A.; Robins, R. J.; Remaud, G. S.; Tcherkez, G. G. B.
Proc. Natl. Acad. Sci. U.S.A. 2012, 109, 18204−18209.
(89) Rados, D.; Turner, D. L.; Fonseca, L. L.; Carvalho, A. L.;
Blombach, B.; Eikmanns, B. J.; Neves, A. R.; Santos, H. Appl. Environ.
Microbiol. 2014, 80, 3015−3024.
(90) Hettling, H.; Alders, D. J. C.; Heringa, J.; Binsl, T. W.;
Groeneveld, A. B. J.; van Beek, J. H. G. M. BMC Syst. Biol. 2013, 7, 82.
(91) Millard, P.; Sokol, S.; Letisse, F.; Portais, J.-C. Biotechnol. Bioeng.
2014, 111, 202−208.
(92) Moseley, H. N. B.; Lane, A. N.; Belshoff, A. C.; Higashi, R. M.;
Fan, T. W. M. BMC Biol. 2011, 9, 37.
(93) de Graaf, R. A.; Rothman, D. L.; Behar, K. L. NMR Biomed.
2011, 24, 958−972.
(94) Norris, C. E.; Quideau, S. A.; Landhausser, S. M.; Bernard, G.
M.; Wasylishen, R. E. Sci. Rep. 2012, 2, 719.
dx.doi.org/10.1021/ac504075g | Anal. Chem. 2015, 87, 133−146
Analytical Chemistry
(95) Ouattara, D. A.; Prot, J.-M.; Bunescu, A.; Dumas, M.-E.; Elena-
Herrmann, B.; Leclerc, E.; Brochot, C. Mol. BioSyst. 2012, 8, 1908−
1920.
(96) Günther, U. L. Top. Curr. Chem. 2013, 335, 23−69.
(97) Witte, C.; Schröder, L. NMR Biomed. 2013, 26, 788−802.
(98) Palaniappan, K. K.; Francis, M. B.; Pines, A.; Wemmer, D. E. Isr.
J. Chem. 2014, 54, 104−112.
(99) Gloggler, S.; Colell, J.; Appelt, S. J. Magn. Reson. 2013, 235,
130−142.
(100) Keshari, K. R.; Wilson, D. M. Chem. Soc. Rev. 2014, 43, 1627−
1659.
(101) Koptyug, I. V. Mendeleev Commun. 2013, 23, 299−312.
(102) Harrison, C.; Yang, C.; Jindal, A.; DeBerardinis, R. J.;
Hooshyar, M. A.; Merritt, M.; Sherry, A. D.; Malloy, C. R. NMR
Biomed. 2012, 25, 1286−1294.
(103) Li, L. Z.; Kadlececk, S.; Xu, H. N.; Daye, D.; Pullinger, B.;
Profka, H.; Chodosh, L.; Rizi, R. NMR Biomed. 2013, 26, 1308−1320.
(104) Harris, T.; Degani, H.; Frydman, L. NMR Biomed. 2013, 26,
1831−1843.
(105) Schilling, F.; Düwel, S.; Köllisch, U.; Durst, M.; Schulte, R. F.;
Glaser, S.; Haase, A.; Otto, A. M.; Menzel, M. I. NMR Biomed. 2013,
26, 557−568.
(106) Jensen, P. R.; Karlsson, M.; Lerche, M. H.; Meier, S. Chem.
Eur. J. 2013, 19, 13288−13293.
(107) Meier, S.; Karlsson, M.; Jensen, P. R.; Lerche, M. H.; Duus, J.
O. Mol. Biosyst. 2011, 7, 2834−2836.
(108) Meier, S.; Solodovnikova, N.; Jensen, P. R.; Wendland, J.
ChemBioChem 2012, 13, 2265−2269.
(109) Harris, T.; Giraudeau, P.; Frydman, L. Chem.Eur. J. 2011, 17,
697−703.
(110) Meier, S.; Jensen, P. R.; Karlsson, M.; Lerche, M. H. Sensors
2014, 14, 1576−1597.
(111) Lumata, L.; Jindal, A. K.; Merritt, M. E.; Malloy, C. R.; Sherry,
A. D.; Kovacs, Z. J. Am. Chem. Soc. 2011, 133, 8673−8680.
(112) Tredwell, G. D.; Behrends, V.; Geier, F. M.; Liebeke, M.;
Bundy, J. G. Anal. Chem. 2011, 83, 8683−8687.
(113) Worley, B.; Powers, R. ACS Chem. Biol. 2014, 9, 1138−44.
(114) Worley, B.; Powers, R. Chemom. Intell. Lab. Syst. 2014, 131, 1−
6.
(115) Wang, K. C.; Wang, S. Y.; Kuo, C. H.; Tseng, Y. F. J. Anal.
Chem. 2013, 85, 1231−1239.
(116) Liebeke, M.; Hao, J.; Ebbels, T. M. D.; Bundy, J. G. Anal.
Chem. 2013, 85, 4605−4612.
(117) MacKinnon, N.; Ge, W.; Khan, A. P.; Somashekar, B. S.;
Tripathi, P.; Siddiqui, J.; Wei, J. T.; Chinnaiyan, A. M.; Rajendiran, T.
M.; Ramamoorthy, A. Anal. Chem. 2012, 84, 5372−5379.
(118) Vu, T. N.; Valkenborg, D.; Smets, K.; Verwaest, K. A.;
Dommisse, R.; Lemiere, F.; Verschoren, A.; Goethals, B.; Laukens, K.
BMC Bioinformatics 2011, 12, 405.
(119) Wang, B.; Goodpaster, A. M.; Kennedy, M. A. Chemom. Intell.
Lab. Syst. 2013, 128, 9−16.
(120) Fages, A.; Pontoizeau, C.; Jobard, E.; Levy, P.; Bartosch, B.;
Elena-Herrmann, B. Anal. Bioanal. Chem. 2013, 405, 8819−8827.
(121) Sousa, S. A. A.; Magalhaes, A.; Ferreira, M. M. C. Chemom.
Intell. Lab. Syst. 2013, 122, 93−102.
(122) Hao, J.; Liebeke, M.; Astle, W.; De Iorio, M.; Bundy, J. G.;
Ebbels, T. M. D. Nat. Protoc. 2014, 9, 1416−1427.
(123) Ye, T.; Zheng, C.; Zhang, S. C.; Gowda, G. A. N.; Vitek, O.;
Raftery, D. Anal. Chem. 2012, 84, 994−1002.
(124) Yilmaz, A.; Nyberg, N. T.; Jaroszewski, J. W. Anal. Chem. 2011,
83, 8278−8285.
(125) Wu, H. D. I.; Fushing, H. Stat. Interface 2011, 4, 443−450.
(126) Romano, R.; Canonico, R.; Acernese, F.; Indovina, P. L.;
Barone, F. Proc. SPIE, Health Monit. Struct. Biol. Syst. 2013, 8695,
869538.
(127) Mihaleva, V. V.; Korhonen, S. P.; van Duynhoven, J.; Niemitz,
M.; Vervoort, J.; Jacobs, D. M. Anal. Bioanal. Chem. 2014, 406, 3091−
3102.
146
Review
(128) Wei, S. W.; Zhang, J.; Liu, L. Y.; Ye, T.; Gowda, G. A. N.;
Tayyari, F.; Raftery, D. Anal. Chem. 2011, 83, 7616−7623.
(129) MacKinnon, N.; Somashekar, B. S.; Tripathi, P.; Ge, W. C.;
Rajendiran, T. M.; Chinnaiyan, A. M.; Ramamoorthy, A. J. Magn.
Reson. 2013, 226, 93−99.
(130) Tulpan, D.; Leger, S.; Belliveau, L.; Culf, A.; Cuperlovic-Culf,
M. BMC Bioinf. 2011, 12, 400.
(131) Mihaleva, V. V.; te Beek, T. A. H.; van Zimmeren, F.; Moco, S.;
Laatikainen, R.; Niemitz, M.; Korhonen, S. P.; van Driel, M. A.;
Vervoort, J. Anal. Chem. 2013, 85, 8700−8707.
(132) Sands, C. J.; Coen, M.; Ebbels, T. M. D.; Holmes, E.; Lindon, J.
C.; Nicholson, J. K. Anal. Chem. 2011, 83, 2075−2082.
(133) Ferry-Dumazet, H.; Gil, L.; Deborde, C.; Moing, A.; Bernillon,
S.; Rolin, D.; Nikolski, M.; de Daruvar, A.; Jacob, D. BMC Plant Biol.
2011, 11, 104.
(134) Ludwig, C.; Easton, J. M.; Lodi, A.; Tiziani, S.; Manzoor, S. E.;
Southam, A. D.; Byrne, J. J.; Bishop, L. M.; He, S.; Arvanitis, T. N.;
Gunther, U. L.; Viant, M. R. Metabolomics 2012, 8, 8−18.
(135) Bharti, S. K.; Roy, R. TrAC, Trends Anal. Chem. 2012, 35, 5−
26.
(136) Barding, G. A., Jr.; Salditos, R.; Larive, C. K. Anal. Bioanal.
Chem. 2012, 404, 1165−1179.
(137) Sokolenko, S.; Blondeel, E. J. M.; Azlah, N.; George, B.;
Schulze, S.; Chang, D.; Aucoin, M. G. Anal. Chem. 2014, 86, 3330−
3337.
(138) Tiainen, M.; Soininen, P.; Laatikainen, R. J. Magn. Reson. 2014,
242, 67−78.
(139) Klein, M. S.; Oefner, P. J.; Gronwald, W. BioTechniques 2013,
54, 251−256.
(140) Song, X.; Zhang, B. L.; Liu, H. M.; Yu, B. Y.; Gao, X. M.; Kang,
L. Y. BMC Bioinformatics 2011, 12, 337.
(141) Hu, K. F.; Westler, W. M.; Markley, J. L. J. Am. Chem. Soc.
2011, 133, 1662−1665.
(142) Martineau, E.; Giraudeau, P.; Tea, I.; Akoka, S. J. Pharm.
Biomed. Anal. 2011, 54, 252−257.
(143) Sandusky, P.; Appiah-Amponsah, E.; Raftery, D. J. Biomol.
NMR 2011, 49, 281−290.
(144) Hochrein, J.; Klein, M. S.; Zacharias, H. U.; Li, J.; Wijffels, G.;
Schirra, H. J.; Spang, R.; Oefner, P. J.; Gronwald, W. J. Proteome Res.
2012, 11, 6242−6251.
(145) Maraschin, M.; Somensi-Zeggio, A.; Oliveira, S. K.; Kuhnen, S.;
Tomazzoli, M. M.; Zeri, A. C. M.; Carreira, R.; Rocha, M. Lect. Notes
Bioinf. 2012, 7632, 129−140.
(146) Maher, A. D.; Cysique, L. A.; Brew, B. J.; Rae, C. D. J. Proteome
Res. 2011, 10, 1737−1745.
(147) Maher, A. D.; Fonville, J. M.; Coen, M.; Lindon, J. C.; Rae, C.
D.; Nicholson, J. K. Anal. Chem. 2012, 84, 1083−1091.
(148) 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. Anal. Chem. 2005, 77, 1282−1289.
(149) Posma, J. M.; Garcia-Perez, I.; De Iorio, M.; Lindon, J. C.;
Elliott, P.; Holmes, E.; Ebbels, T. M. D.; Nicholson, J. K. Anal. Chem.
2012, 84, 10694−10701.
(150) Barding, G. A., Jr.; Orr, D. J.; Sathnur, S. M.; Larive, C. K. Anal.
Bioanal. Chem. 2013, 405, 8409−8417.
(151) Wang, B.; Shi, Z. Q.; Weber, G. F.; Kennedy, M. A. Anal.
Bioanal. Chem. 2013, 405, 8419−8429.
(152) Shao, X. G.; Miao, L. N.; Liu, Z. C.; Liu, H. L.; Cai, W. S.
Spectrosc. Lett. 2011, 44, 244−250.
(153) Smolinska, A.; Blanchet, L.; Buydens, L. M. C.; Wijmenga, S. S.
Anal. Chim. Acta 2012, 750, 82−97.
dx.doi.org/10.1021/ac504075g | Anal. Chem. 2015, 87, 133−146