TrACh 2014analytical Metabolomics-Based Approaches To Pancreatic Cancer
TrACh 2014analytical Metabolomics-Based Approaches To Pancreatic Cancer
TrACh 2014analytical Metabolomics-Based Approaches To Pancreatic Cancer
a r t i c l e i n f o a b s t r a c t
Keywords: We provide a comprehensive review of the analytical technologies applied in pancreatic cancer (PC)
Analytical method research, investigating tumor-associated metabolites identified as possible new biomarkers for screening
Biomarker
the population at risk and for potential early tumor diagnosis. We briefly introduce PC tumorigenesis, risk
Extraction
factors and some of the benefits of studying changes to the metabolome associated with PC. We focus on
Mass spectrometry
Metabolic profile the different matrices used in the studies and sample preparation for the analytical methods applied to PC
Metabolomics specimens. We then delineate the advantages and the disadvantages of each analytical platform, pointing
Pancreatic cancer out the importance of extraction techniques in metabolomics approaches employed in PC research. We
Pathogenesis cover a number of case studies, and then review identified metabolites associated with PC. In the final sec-
Physiopathology tion, we discuss possible approaches to understanding the link between metabolic profile and the onset of
Sample preparation PC, which may provide insights into the physiopathological mechanisms involved in PC pathogenesis.
Ó 2014 Elsevier Ltd. All rights reserved.
Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
2. Metabolomics and cancer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
3. Metabolomics approaches in biomarker discovery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
4. Metabolomics-based studies in PC research. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
4.1. Biological fluids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
4.2. Analytical platforms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
4.3. NMR-based studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
4.4. MS-based studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
5. Sample preparation in approaches using metabolomics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
6. Metabolomics data export and analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
7. Commercial software and library sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
8. Analysis and summary of the results obtained in PC studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
9. Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
10. Conclusions and perspectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
Appendix A. Supplementary data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
Abbreviations: CDA, Cytidine deaminase; CP, Chronic pancreatitis; CT, Computed 1. Introduction
tomography; CTPS, Cytidine 50 -triphosphate synthetase; dCK, Deoxycytidine
kinase; dFdC, 20 ,20 -difluorodeoxycytidine or gemcitabine; dFdU, 20 ,20 -difluorode-
oxyuridine; hENT, Human equilibrative nucleoside transporter; MRI, Magnetic
Pancreatic cancer (PC) is one of the leading causes of cancer-
resonance imaging; MRP, Multidrug resistance-associated protein; PC, Pancreatic related deaths worldwide, reflected by an incidence of 277,668
cancer; PCA, Principal-component analysis; PDCA, Pancreatic ductal adenocarci- new cases and almost the same mortality rate (266,029 cases)
noma; PLS-DA, Partial least squares discriminant analysis; TOF, Time-of-flight; per year [1]. The incidence of PC is constantly on the rise, especially
UHPLC, Ultra-high-performance liquid chromatography.
⇑ Corresponding author. Tel.: +39 0461 615136; Fax: +39 0461 615200. in the USA among the black male population (15/100,000) [2]
E-mail address: [email protected] (I.M. Di Gangi).
(Fig. 1) and increases with age, according to GLOBOCAN data [3]
http://dx.doi.org/10.1016/j.trac.2013.12.006
0165-9936/Ó 2014 Elsevier Ltd. All rights reserved.
I.M. Di Gangi et al. / Trends in Analytical Chemistry 55 (2014) 94–116 95
Fig. 1. Data for estimated 5-year prevalence for both sexes (Panel A) and the incidence of pancreatic tumor for males (Panel B) of all ages around the world. (Adapted from
GLOBOCAN 2008, IARC [3]).
96 I.M. Di Gangi et al. / Trends in Analytical Chemistry 55 (2014) 94–116
Fig. 2. Curves of estimated incidence (A) and mortality (B) for PC in males and females in southern Europe. Data are expressed as age-standardized rates per 100,000.
(Adapted from GLOBOCAN 2008, IARC [3]).
(Fig. S1). PC is also the tenth most frequent cancer in Europe, diabetes, overweight and Helicobacter pylori have long been known
accounting for some 2.6% of cancer in both sexes, even though [4–8]. Furthermore, over five decades, large independent studies
the age-specific rates of incidence and mortality in men are have examined ABO blood groups and the risk of PC in Western,
one-and-a-half times higher than in women (Fig. 2). Asian and other populations, suggesting that blood-type A might
A number of risk factors for developing PC have been estab- be associated with a higher risk of developing pancreatic tumor
lished: a family history of the disease, smoking, alcohol, age and [9] and observing a lower frequency of blood-type O in patients
I.M. Di Gangi et al. / Trends in Analytical Chemistry 55 (2014) 94–116 97
with exocrine PC, compared to patients with other forms of cancer to chemotherapy and recurrence following surgical resection in pa-
[10]. In another study, Pelzer et al. [11] found that the majority of tients with PC, but has demonstrated modest effectiveness in the
patients suffering from PC had blood-group A, and they confirmed screening of symptomatic individuals on an outpatient basis, with
that blood-group 0 was less frequent in patients with PC. However, a median sensitivity of 79% (range 70–90%) and median specificity
to date, the mechanisms by which ABO blood type, or closely- of 82% (range 68–91%). However, it has been shown to be ineffec-
linked genetic variants, may influence the risk of PC have not been tive for the mass screening of asymptomatic subjects [22]
clearly defined. (Table S1).
Rapid tumor progression and early metastasis make it a partic- Gold et al. [23] developed a monoclonal antibody (Mab PAM4),
ularly lethal solid tumor, with a 5-year survival rate of less than 5%. which is highly specific for a MUC1 produced by pancreatic adeno-
Consequently, most treatment options are based on palliative care carcinoma. This antigen is identified in over 90% of pancreatic car-
[12]. cinoma and its precursor lesions, but is not detected in the normal
Because early PC often does not cause symptoms and later pancreas [23]. The authors demonstrated that the sensitivity and
symptoms are usually non-specific and varied, less than 20% of the specificity of the immunoassay for PC were 77% and 95%,
patients can have their cancer resected with a curative intent, be- respectively, suggesting that PAM4 is potentially useful for the
cause more than 80% of patients have a locally advanced or meta- early detection of PC [23] (Table S1).
static tumor that is unresectable at the time of diagnosis [13–15]. Unfortunately, none of these markers has achieved sufficient
The clinical symptoms of PC are usually unremarkable until the levels of sensitivity and specificity to be recommended for screen-
cancer has progressed to an advanced stage. The use of imaging ing asymptomatic patients in the general population, so the search
studies (e.g., abdominal CT or MRI), in the context of strong clinical for new biomarkers that would allow early detection of disease is
suspicion of pancreatic ductal adenocarcinoma (PDAC), is inade- still an open issue.
quate for diagnosing PC at an early stage, as these techniques do This review has the aims of:
not reliably detect pancreatic tumors <1–2 cm in size, so, at the
moment, there is no effective method for early detection [16]. (1) examining the various analytical technologies that have
20 ,20 -Difluorodeoxycytidine (gemcitabine, dFdC) is currently been applied in PC research;
considered the preferred host chemotherapeutic drug for meta- (2) investigating tumor-associated metabolites identified as
static and advanced pancreatic tumors, but it is far from being possible new biomarkers for screening of the population at
the gold standard, since the overall patient-survival rate is unac- risk; and,
ceptably low and new therapeutic approaches are urgently needed (3) studying the potential role of metabolomics in the field of
[17,18]. One of the reasons for the poor response to dFdC is thought early tumor diagnosis and personalized therapy.
to be acquired resistance [19], which occurs with four possible
processes, as suggested by Ohmine et al. [20], on observing the 2. Metabolomics and cancer
changes in protein expression in RPK9 cells. These authors sur-
mised that the acquisition of dFdC resistance was through: Metabolomics has developed rapidly in the past few years and it
represents a useful tool for the identification of new biomarkers,
(1) attenuation of dFdC phosphorylation via suppression of dCK, which may help to understand tumorigenesis mechanisms in
a kinase protein that has a key role in phosphorylating gem- cancer research and to develop ‘‘cancer models’’ [24]. This technol-
citabine in order to activate its function; ogy, when used as a translational research tool, can provide a link
(2) facilitation of dFdC inactivation to dFdU by induction of between the laboratory and the clinic, particularly because
CDA; metabolic and molecular imaging technologies, such as positron
(3) facilitation of dCTP production by induction of ribonucleo- emission tomography and magnetic resonance spectroscopic imag-
tide reductase and CTPS; and, ing, enable non-invasive discrimination of metabolic markers
(4) attenuation of dFdC uptake or facilitation of metabolite ‘‘in vivo’’.
excretion by induction of MRP expression or by decreasing Metabolomics is one of the pillars of systems biology, providing
expression of hENT1 [20]. a quantitative snapshot of low-molecular-weight compounds (typ-
ically 1000 Da) in biofluids and other complex matrices [25,26].
Screening tests for the general population need to be accurate, Thus pancreatic tumor can represent a rich source for the discovery
safe, and convenient, as well as capable of high throughput. Many of biomarkers with appreciable specificity and sensitivity. In this
serologic markers have been found to be potential biomarkers dur- context, the use of metabolomics-based approaches providing ac-
ing screening for PC detection, belonging to different classes of cess to biochemical-phenotype information can offer an opportu-
compounds, such as: nity for non-invasive screening of early tumor-associated
perturbations in the cellular metabolism [27].
(1) tumor markers: CA 19-9, CA-242, CA-50, CEA, CEACAM1, It also has the potential to identify new targets for diagnosis and
MIC1, MUC1, alpha-fetoprotein, DU-PAN-2 mAb, alpha4GnT; intervention at various stages of malignancy, which could support
(2) apoptosis markers: NF-kB, hTert and CK-19; early, personalized therapy [28]. A critical step in defining the
(3) cytokines, such as IL-8; correct personalized anticancer therapy is identification of altered
(4) adhesion molecules: ICAM-1, MMP-2, MMP-9; and, genes and pathways in the patient’s tumor and elucidation of their
(5) growth and angiogenesis factors, such as EGFR, IGFBP-1. particular oncogenic role. The success of molecular studies in iden-
tifying potential molecular targets for therapeutic intervention is
These and others serum biomarkers are summarized in exemplified by developments in the treatment of chronic myelog-
Table S1, which presents an overview of significant evaluated mol- enous leukemia (CML) [29] and the development of clinical appli-
ecules, potentially modulated in patients diagnosed with PC. cations from biomarker discovery using metabolomics to
CA19-9 is the most extensively studied of these tumor markers, therapeutic targets in breast cancer [30], ovarian cancer [31] and
but it has poor specificity for PC, being high in the case of many renal cell carcinoma [32].
cancers of the upper gastrointestinal tract, ovarian cancer, hepato- A growing number of metabolomics-based studies have been
cellular cancer and benign conditions of the hepatobiliary system carried out, with more than 700 published articles on cancer re-
[21]. CA19-9 is considered the standard for monitoring response search (Fig. S2A). Furthermore, in the past few years, we have
98
I.M. Di Gangi et al. / Trends in Analytical Chemistry 55 (2014) 94–116
Fig. 3. Overall flowchart of strategy for metabolomics-based investigations directed at biomarker discovery in PC (PC). The process of biomarker discovery using metabolomics begins with a well-designed experimental plan,
which most importantly must include appropriate controls for the disease concerned. A prediction model needs to be created, using independent validation datasets with a minimum of three classes: healthy control, disease group
and an additional related-disease control group (i.e. biological samples from PC versus non-PC controls versus CP patients). The analytical methods and statistical validation employed in biomarker discovery are of critical
importance. The final step is to identify and to validate new therapeutic targets, starting from the possible pathways involved in the onset and the progression of PC, which should ultimately be tested both ‘‘in vitro’’ and ‘‘in vivo’’ to
establish the utility of the biomarkers discovered in a routine clinical setting.
I.M. Di Gangi et al. / Trends in Analytical Chemistry 55 (2014) 94–116 99
observed increasing interest in the application of the metabolo- greatest potential to provide diagnostic information on the general
mics approach to PC (Fig. S2B), which represents a challenge in health of the organism. Metabolites in blood or urine reflect the
terms of survival rate and disease prognosis. This research is begin- state of the disease and the host’s response to the disease. When
ning to provide new information and search for metabolic stored at 40°C, both blood and urine samples can be kept for ex-
biomarkers using NMR, GC–MS, LC–MS, in the hope that it may tended periods without alteration for subsequent analysis [37].
allow early diagnosis and/or screening for PC in selected popula- Samples can be procured by relatively non-invasive means at mul-
tions [24]. tiple time points, permitting temporal-based analysis. It is conceiv-
able that more immediate changes in metabolites may be seen in
blood during acute physiological events. Moreover, metabolite pro-
3. Metabolomics approaches in biomarker discovery
files from serum have demonstrated less circadian variation, and
decreased inter-subject and intra-subject variability compared to
Currently, many different approaches are used for metabolo-
urine [27].
mics investigations (e.g., targeted and untargeted metabolomics
However, preparation and processing of serum or plasma is
are complementary strategies that are commonly combined in
much more complex, and even small changes in technique alter
cancer-biomarker discovery) [33]. Using the targeted method, a
the recovery of certain metabolites [38,39]. Furthermore, lipid
particular metabolite or a small group of related metabolites can
and protein present in serum specimens may induce degraded
be measured, typically focusing on one or more related pathways
spectral resolution of serum metabolites, an interference absent
of interest [34]. In other cases, an untargeted or global approach
in urine samples [40].
may be taken, in which as many metabolites as possible are mea-
The metabolomics profile of serum/plasma, urine and other
sured simultaneously and compared in biological samples without
biological fluids, such as bile, saliva, cell culture or tissue extract,
bias [35]. In contrast to targeted metabolomics results, untargeted
has been investigated for diagnostic purposes for a variety of
metabolomics datasets are exceedingly complex, but they give ac-
malignancies, including PC [41–57]. From these data, candidate
cess to semi-quantitative measurement of unknown compounds,
biomarkers, such as glutamate [50], 3-hydroxybutyrate
which is essential for generating new hypotheses [35].
[41,43,44], lysine [48,50], phenylalanine [41,44], arachidonic acid
Metabolomics also includes other types of experiments that can
[48], glucuronic acid [53], tauroursodeoxycholic acid [48] and
be applied to the field of cancer research, including metabolic pro-
deoxycholylglycine [48], have been proposed, due to the significant
filing and metabolic fingerprinting.
changes between PC patients and healthy controls.
Metabolic fingerprinting usually analyses all resectable analytes
The key pathways that have been demonstrated to behave dif-
in a given sample, with subsequent classification of samples and
ferently in pancreatic tumor and normal controls include glycolysis
identification of differentially-expressed metabolites, which define
and the pentose phosphate, nucleotide and protein biosynthesis, li-
the sample classes. Metabolic profiling instead focuses on the anal-
pid and phospholipid turnover, TCA cycle and redox stress
ysis of a specific group of metabolites or a class of compounds,
[47,54,55]. The phosphocholine metabolism has also been sug-
whose data can be integrated with pathway maps to enhance bio-
gested to be implicated in PC [54], supporting the recent finding
logical understanding.
of a new biomarker, PC-594, as an opportunity for identifying sub-
Dettmer et al. [33] provided a clear outline of these two comple-
jects with PC or at high risk of developing PC [55].
mentary methods, showing that their combination is a potential
All the observations of these metabolomics-based studies, per-
tool for metabolomics investigations.
formed in around 60% of cases on plasma or serum [41–45,48,49]
However, the number and the chemical properties of the
and in 15% of cases on urine [46,47], support the feasibility of using
metabolites to be studied are defining attributes of any metabolo-
serum/plasma or urine specimens for clinical research on the met-
mics experiment and determine the experimental design with
abolomics state of PC patients, and for biomarker discovery. The
respect to sample preparation and choice of instrumentation.
use of other biological fluids, such as saliva [56], bile [53] and pan-
Fig. 3 presents an overall flowchart of the strategy for metabolo-
creatic tissue [57], was also investigated (25% of studies) for this
mics-based investigations directed at biomarker discovery in PC.
purpose.
The process of biomarker discovery using metabolomics begins
Saliva is an important biological fluid that provides various
with a well-designed experimental plan, which includes three dif-
functions and is made up of 99% water, with minerals, mucus, elec-
ferent steps, as proposed by Wettersten and Weiss [32]:
trolytes, nucleic acids and proteins, such as enzymes, enzyme
inhibitors, growth factors, cytokines, immunoglobulins and other
Phase 1: discovery;
glycoproteins [58]. It is a filtrate of blood, reflecting the physiolog-
Phase 2: pre-validation and quantification; and,
ical condition of the body, so it could be used to evaluate clinical
Phase 3: validation and application.
status and predict systemic diseases [58]. Compared to blood, sal-
iva offers distinct advantages for diagnostic or research purposes
For any biomarker discovered through metabolomics analysis
(e.g., its collection is cost effective, safe, easy and non-invasive).
to be used in routine clinical practice, sample collection and an
Indeed, many of the characteristics of biological fluids, such as
analytical methodology need to be available, allowing accurate
blood and urine, are applicable to saliva, including circadian varia-
quantification of the metabolite and offering high throughput,
tion and the presence of diverse diagnostic analytes [58].
prompt turnaround and low costs [32]. As in all biomarker studies,
A comprehensive metabolite analysis of saliva samples was
the likelihood of finding a true marker for the general population is
conducted by Sugimoto et al. [56] using capillary electrophoresis
exceedingly low. It is more likely that biofluid markers will be most
time-of-flight mass spectrometry (CE-TOF-MS). The authors identi-
useful for assessing the population at risk.
fied 48 principal metabolites, mostly amino acids, as biomarker
candidates using salivary diagnosis for the detection of PC [56].
4. Metabolomics-based studies in PC research Bile is among the other matrices that can be used to study and
to discover molecular markers. Bile composition is very complex
4.1. Biological fluids and varies according to the nutritional state of the individual.
The main components include bile acids, phospholipids (mainly
The vast majority of clinical research has been carried out on ur- phosphatidylcholine), cholesterol, bilirubin (mostly in its
ine and serum samples [36], perhaps because these fluids have the conjugated form) and inorganic salts (potassium, sodium and
100 I.M. Di Gangi et al. / Trends in Analytical Chemistry 55 (2014) 94–116
bicarbonate), and very small amounts of copper and other metals. amino acids, fatty acids, organic acids, sugars and enzymes and
Recent methodological developments have enabled identification transporter proteins.
of a number of bile metabolites that have links with the biliary Furthermore, the development of high-resolution and ultra-
tract and PCs [53,59,60]. Investigations of human bile are also con- high-resolution analyzers (TOF-MS and FT-MS, respectively) has
sidered to help the biomarker discovery process in vitro and pro- yielded accurate mass measurements, whereas ion-trap analyzers,
vide avenues for translational research in detecting and following among other things, are used due to their capacity to perform two-
dynamic variations of biomarkers in clinical settings using stage, three-stage or multi-stage MS/MS (i.e. MSn) experiments to
non-invasive approaches, such as in-vivo magnetic resonance spec- obtain the additional structural information needed for metabolite
troscopy [61]. The findings of Bezabeh et al. [53], using 1H NMR identification.
spectroscopy in human-bile samples of PC patients, clearly indi- Most MS-based studies have enhanced analysis for global
cated that there is an alteration in bile composition in the course profiling of the PC ‘‘metabolome’’, in a non-targeted metabolomics
of pancreatic tumor, suggesting that elevated levels of GlcU ob- approach, by using the combination of UHPLC or GC with Q-TOF-
served could be of diagnostic value in the detection of PC [53]. MS [57,65,67], flow injection with Fourier-transform ion-cyclotron
resonance MS (FI-FT-ICR-MS) [55], or LC with LTQ-Orbitrap [48] to
generate comprehensive metabolomics profiles containing thou-
4.2. Analytical platforms sands of accurate mass measurements from PC patients and dis-
ease-free subjects.
Metabolomics experiments in a biological sample can employ Other analytical platforms showing promise for PC metabolo-
one or more complementary analytical platforms to obtain exten- mics research include CE-MS [56] and pressurized capillary elec-
sive coverage, because of the large number of molecules, with sub- trochromatography (pCEC), which is a hybrid technique between
stantial diversity of the metabolites in terms of chemical structure LC and CE, coupled with a UV detector [49]. However, CE based
and concentration. techniques are still at an early stage as compared to techniques
In PC research, metabolomics-based studies directed at bio- based on GC and LC.
marker discovery have been performed by employing nuclear mag-
netic resonance (NMR) [43,44,46,47,50,53] or MS [20,41,42,45,
48,62], the latter being coupled with a separation technique, such 4.3. NMR-based studies
as gas chromatography (GC) [41,45,48,62], liquid chromatography
(LC) [20,48,62] or CE [49,56] and with different ion sources and Recently NMR-based metabolomics was applied to toxicity,
mass analyzers, depending on the type of experiment. drug efficacy and disease diagnosis both in vitro and in clinical tri-
Each of these analytical techniques has advantages and disad- als [25,59,68–71], providing evidence of differing spectral regions
vantages, most notably differences in sensitivity, reproducibility, in cancer patients and healthy controls [72]. This approach led to
accuracy and equipment costs, as well summarized elsewhere discrimination of patients with PC from those with chronic pancre-
[63,64]. atitis (CP) and healthy controls using the metabolic profiles of rat
NMR techniques have been applied to targeted and untargeted pancreatic tissue, demonstrating the relationship with disease pro-
studies of PC, using different kinds of experiments: gression in an animal model [70]. There is an established associa-
tion between carcinoma of the pancreas and the sporadic or
(1) One-dimensional 1H-NMR spectroscopy [43,46] with NOESY hereditary forms of CP, and most studies are designed using PC
experiments [47,50]; and CP as two different classes of patients. Results from a prospec-
(2) Two-dimensional 2D-NMR spectroscopy [44,53], in particu- tive case–control study conducted by Ge et al. [73] seem to suggest
lar, 1H-1H-DQF-COSY and TOCSY experiments for analysis of that CP may be closely related to PC, although the answer to the
D-glucuronic acid in bile samples [53]; question of a cause-and-effect relationship is unclear. In this
(3) 31P-NMR spectroscopy [53]; context, Zhang and colleagues [50] found different metabolomics
(4) Two-dimensional J-resolved 1H and H, 13C-HSQC-NMR spec- patterns in the plasma samples of PC and CP patients using a
troscopy [47]; and, NMR metabolomics-based approach. The PC group showed ele-
(5) High-resolution magic angle spinning (HRMAS) 1H-NMR vated concentrations of N-acetyl glucosamine (NAG), dimethyl-
spectroscopy applied to tissue samples in PC rats [52]. amine (DMA), very low-density lipoprotein (VLDL) and acetone
and reduced levels of 3-hydroxybutyrate, lactate, high-density
Of the ionization techniques, atmospheric pressure ionization lipoprotein (HDL), low-density lipoprotein (LDL), citrate, alanine,
(API), especially electrospray ionization (ESI), is a popular choice glutamate, glutamine, histidine, isoleucine, lysine and valine,
as the ion source in MS-based applications in PC research compared to CP patients [50]. These results suggest that PC may
[20,48,65]. When combined with LC, it shows good sensitivity develop in different ways and lead to different metabolic changes
and a high dynamic range and versatility, but also provides soft in PC and CP patients and healthy individuals [50].
ionization conditions, giving access to the molecular mass of intact A study of targeted profiling, using the 1H-NMR spectra of urine
molecules from complex mixtures. samples from patients with benign pancreatic conditions and
Only a few targeted studies have been conducted using MS PDAC, showed a clear distinction between controls and PDAC with
without prior separation chromatography, by direct injection anal- early-stage or locally-advanced disease [46]. The authors observed
ysis [42,55], which is a high-throughput approach allowing a unique metabolite-expression patterns for PDAC, with strong pre-
sample to be processed within a few minutes, but with major lim- dictive accuracy for cancer, which could reflect changes occurring
itations in terms of ion suppression and loss of sensitivity. Typi- at both tumor-microenvironment level and the global metabolism
cally, labelled standards are used in this approach to correct [46]. Recently, Napoli et al. [47] identified a discrete urinary meta-
matrix effects [66], but standards are not available for all com- bolomics signature for PDAC, distinguishing patients from healthy
pounds and they are sometimes very expensive. Targeted controls and discriminating samples based on tumor localization.
approaches, using a single-quadrupole or triple-quadrupole mass The possibility of following disease evolution and deducing tu-
analyzer, coupled with GC [41,45,62] or LC [20,62], have instead mor-site mapping was achieved with data extracted from NMR
been used to quantify particular classes of compounds, presumed spectra using multivariate statistical analysis. However, a large
to be altered in PC in comparison to the control group, such as proportion of the patients were at stage IV and gender effects were
I.M. Di Gangi et al. / Trends in Analytical Chemistry 55 (2014) 94–116 101
not examined, as the profiles were developed solely from male and supporting a possible link between lipid metabolism and can-
patients [47]. cer development.
An NMR-based analytical method was applied by Bathe et al. Other possible pathways involved in PC were explored in the
[44] to establish a serum-metabolomics profile capable of discrim- past few years {e.g., MALDI-TOF-MS was used to study the fucosy-
inating PC patients from patients with benign pancreatic biliary lation pattern of glycans in MS and MS/MS modes for PC (stages IA,
disease. The authors were able to distinguish metabolites that IIA and B, III, and IV) versus CP and normal controls, surmising the
are strongly related to disease alone (elevated glutamate, acetone, use of haptoglobin-fucosylation changes as a marker for disease in
and 3-hydroxybutyrate) and those related to disease and age (ele- pancreatic tumors [75]}. Finally, a UHPLC-TOF-MS/MS method was
vated glucose, phenylalanine, formate, and mannose) [44]. These applied to perform small-molecule-metabolite profiling of
results were congruent with serum studies limited in sample size matched normal and PC tissue, identifying a sub-set of known
(17 PC) [43], which related the differences in metabolite profiles metabolites that were found to be significantly de-regulated in
in PC patients and healthy subjects to physiological and patholog- pancreas tumor tissue [57].
ical variations. Nishiumi et al. [45] instead used a GC-MS-based metabolomics
Using 1H-NMR, Nishijima et al. [60] discovered significantly- approach to analyze serum samples from 21 PC patients and 9
elevated lactate levels in the serum and the bile of patients with healthy volunteers. PC patients, at early (I) and advanced (III, IVA
malignant hepatobiliary disease. In 2009, Bezabeh et al. [53] ob- and IVB) stages of the disease, were shown to have significant
served that PC patients had very high levels of D-glucuronic acid differences in the metabolome compared to the control group,
(GlcU) in bile samples, whereas this was absent or negligible in showing a variation in the profile of 18 out of 60 identified metab-
control and CP patients. The exact mechanism inducing elevated olites [45]. The authors later confirmed these data, creating a ser-
levels of GlcU in bile remains unclear. Nevertheless, they suggest um metabolomics-based diagnostic model for detecting
that this metabolite could be a very promising additional marker resectable PC patients [41].
in these patient categories for evaluation of definitive diagnosis Although some published metabolome-profiling studies of pan-
[53]. creatic carcinoma are heterogeneous regarding the MS techniques
Finally, another study reported successful discrimination of used and the metabolites studied, they all share canonical vari-
malignant from benign biliary disease using bile samples, pointing ance-based evaluation strategies with two classes (PC and healthy
out the better performance of the metabolomics approach in the subjects) or three classes (PC, CP and healthy subjects) (Table S2)
diagnosis of biliary tract cancer, in comparison with conventional for comparison and make use of multivariate statistical analysis
diagnostic tools, including serum-tumor markers (CA19-9, CEA) of the data.
and bile cytology [59]. However, invasive methods are required To date, only a few applications of a CE-MS based approach in
for the collection of bile and it is generally inconvenient to sample. PC have been published. Using CE-TOF-MS, it was recently demon-
An altered blood-lipid profile was observed in plasma extracts strated that the salivary metabolomics profile of a small number of
from a large number of PC patients (n = 82) by Beger and col- patients with PC was significantly different from normal controls
leagues [74]. They surmised that this could be related to altered [56]. pCEC coupled with UV detection was instead used by Zhang
pancreatic functionality and altered levels of pancreatic enzymes, and co-workers [49] to study the plasma samples of 17 PC patients.
such as lipase, phospholipase A and cholesterol esterase, which Metabolomics profiling, analyzed using OPLS-DA multivariate
help the body digest fat [74]. analysis, showed different metabolite phenotypes for the PC group
and healthy controls [49]. These results suggest that pCEC can be
considered a promising technique, which, when coupled with
4.4. MS-based studies MS, could serve as a more powerful platform for disease-biomarker
discovery.
There are different applications of metabolomics-based chro-
matography-MS methods for PC. A crucial recent study, conducted
by Bi and colleagues [65], focused on the optimization of quench-
ing, harvesting and extraction of metabolites from human pancre- 5. Sample preparation in approaches using metabolomics
atic cell line Panc-1, in order to obtain the maximum analytical
recovery of all the compounds considered in cell lysate. The most The quality of sample preparation is an essential factor for the
efficient, reproducible method for extraction of Panc-1 cells was success of all analytical procedures. When working with biological
by using 75% of a 9:1 MeOH/chloroform solution. Separation matrices, such as serum, plasma, whole blood, tissue homogenates,
efficiency was also evaluated by comparing HILIC and RPLC separa- saliva, urine, cell pellets or cell media, analyte extraction usually
tion, using a UHPLC-QTOF-MS system [65]. HILIC provided better becomes very sensitive because the analytical procedure may be
separation of metabolites than RPLC, thus enhancing Panc-1 cell- affected by interference from matrix components [76]. In the met-
metabolome coverage and making it possible to achieve the best abolomics approach, all small molecules are the targets and only
reproducibility for the analytical procedure proposed [65]. salts and macromolecules, such as proteins or larger peptides,
A targeted approach was applied by Leichtle et al. [42] using can be considered as matrices. Sample preparation for the analysis
FIA-MS/MS analysis of 26 amino acids in the sera of PC patients, of biological samples can include liquid-liquid extraction (LLE), so-
healthy controls and pancreatitis patients [42]. This obtained data lid-phase extraction (SPE), supercritical fluid extraction, acceler-
consistent with a previous publication, in which the authors ated solvent extraction, protein precipitation and membrane
revealed 26 significantly different metabolites in the plasma met- methods, such as dialysis or ultracentrifugation [33]. In the case
abolomics profiles of 5 PC patients in comparison with 5 mixed of low-abundance metabolites, the extraction procedure can
pancreatitis/healthy controls, by combining GC-TOF-MS, LC-ESI- include a pre-concentration step to achieve the detection limits
MS, and LC-LTQ-Orbitrap studies [48]. Using HILIC-LC-ESI-MS anal- for the analytical technique applied [76].
ysis, naturally-occurring isomers of polar phosphatidylcholine PCs Since the polarity of biological metabolites varies widely, opti-
(34:2) were also detected in the same plasma samples [48]. Orešič mization of extraction and analytical methodology is extremely
and co-workers [67] had already performed ‘‘lipidomics studies’’ in important to recover and to detect as many metabolites as possi-
breast-cancer research, making use of a GC-TOF-MS metabolomics ble. Moreover, sample preparation should be highly reproducible,
approach, demonstrating up-regulation of many membrane lipids effective, and fast to allow high-throughput studies [77].
102 I.M. Di Gangi et al. / Trends in Analytical Chemistry 55 (2014) 94–116
Nevertheless, it should be noted that any kind of sample prep- The sample-preparation procedures used in PC metabolomics
aration would cause analyte losses. In particular, extraction proce- studies are summarized in Table 1.
dures, such as SPE or LLE, when applied to aqueous samples, result
in poor recovery of very polar compounds [33].
In the case of an NMR metabolomics-based approach for clinical 6. Metabolomics data export and analysis
diagnostics, the advantage is that only minimal sample preparation
is necessary and it can be applied to different matrices [78]. How- Metabolomics is not limited to individual biomarkers, but rep-
ever, some factors may affect the NMR spectra (i.e. chemical-shift resents a new approach to diagnostics, in which large overall data-
signal perturbation or inconsistency of the peak position) in a sets can be employed to develop new models. Both metabolic
way that subsequently decreases the efficacy of multivariate statis- fingerprinting and profiling can be used in the search for new
tical analysis of the data. Careful sample preparation can help to biomarkers.
limit peak shifts (e.g., by attempting to control the pH of biological The metabolic fingerprinting approach usually allows detection
samples). In PC, many studies performed using NMR for serum or of all metabolites in biological samples, with subsequent classifica-
plasma [43,44] and urine [46,47] treat the specimens simply by tion into different groups and identification of differentially-ex-
adding NaOH or HCl [46] or phosphate-buffer solution [47] to min- pressed analytes among them. To perform this complex task,
imize variations in NMR chemical shifts of metabolites arising from data-analysis tools, metabolite libraries and databases are
differences in pH. required.
NMR analysis was done after derivatization with phenyl The first step in data export is to convert analytical data from
hydrazine in order to obtain target analysis for detecting levels of metabolomics-based experiments into a standard, uniform format
D-glucuronic acid in PC patients [53]. [34], which includes data pre-processing in order to reduce noise
Analysis of volatile compounds, which can be found in urine, and background, alignment of the spectra from NMR or MS and
plasma or exhaled air, as an important component of the metabo- automated picking and annotation of MS or total-ion/extracted-
lome, is usually performed using GC-MS. ion chromatographic peaks of metabolites [81]. The data may
Conventional sample preparation techniques (e.g., LLE or SPE) eventually be pre-treated (e.g., centering, and scaling) [82], and
are often infeasible, because the analytes may be incompletely ex- finally processed using multivariate statistical analysis, which
tracted and losses will occur if the extract is concentrated. More- facilitates data visualization through data-dimensionality reduc-
over, the solvent can interfere with the analytes during GC tion, and highlights relevant biological information [83].
separation, due to incomplete separation of the solvent peak from However, for metabolomics-data analysis, it is necessary to
the analyte peaks. In order to analyze organic acids, for example, extract information about all compounds, including mass and
Wahl et al. [79] used headspace sampling with a gas-tight syringe, retention time in the case of chromatography, compound identifi-
combined with cold trapping in a temperature-controlled cold- ers and intensity (i.e. chromatographic height or area) as quantita-
injection system and GC-MS for the analysis of volatile metabolites tive representations of concentration. The process is inherently
in urine samples. difficult for the processing of an untargeted dataset, due to the
Non-polar lipid classes have also been analyzed in PC patient greater complexity of this approach, in particular as regards data
biofluids using GC separation and MS, with a derivatization step alignment, data normalization and annotation [33].
(e.g., methylation, silylation or oximation) necessary to expand fur- Finally, raw data from each experiment are stored in sample
ther the range of lipids accessible with this technique [41,45,67]. files as spectra acquired at given time points or scans. There are
However, GC–MS analysis is limited to thermally-stable com- currently several public and proprietary software tools available
pounds with sufficient vapor pressure for volatilization during to carry out data export and subsequent statistical analysis.
injection. Multivariate statistical techniques are the most useful tools for
LC–MS experiments are often affected by deviations in reten- investigating this array of information efficiently, by reducing the
tion time from sample to sample, as a consequence of column deg- dimensionality of complex datasets. They generally fall into two
radation or sample carry-over, small fluctuations in room categories of analysis – supervised and unsupervised – which
temperature or mobile-phase pH, as well as other variations. Be- should be selected according to the aim of the study [33].
fore injection, serum or plasma samples should usually be treated If the aim is sample classification and prior information about
with organic solvent (acetonitrile or methanol) for precipitation the sample identity is unknown, unsupervised methods are used
protein [42,74]. Furthermore, the supernatants are sometimes [e.g., hierarchical clustering analysis (HCA), principal-component
derivatized to carry out direct-injection analysis of the metabolites analysis (PCA), or independent component analysis (ICA)] [35].
concerned {e.g., amino acids [42]}, or extracted with chloro- However, if sample identity is known, and the aim of the study
form:methanol mixtures for the analysis of hydrophobic com- is to discover characteristic biomarkers (e.g., in order to search for
pounds {e.g., phospholipids [74]}. biomarkers of PC by comparing samples from healthy and diseased
Urine samples are usually diluted, because they contain a high subjects), supervised methods [e.g., generally discriminant analysis
concentration of salt, which can cause ion suppression and adduct (DA)], using prior information about sample class may be suitable.
formation in the electrospray process and may also cause rapid In this case, particular care must be taken to ensure an appropri-
deterioration of instrumental performance, due to contamination ately large number of observations, in order to reduce the possibil-
by non-volatile residues. ity of generating false positives using supervised methods [63].
For the treatment of in-vitro cell or human-tissue samples, the In any case, data mining of metabolomics datasets is generally a
procedure usually requires washing and quenching with ice-cold complex procedure, initially requiring unsupervised methods to
methanol and the use of trypsin or cell scrapers in order to remove investigate the underlying data structure, unbiased by knowledge
adherent cells from the growth surface [20,65]. A series of freeze- of experimental design [84]. The most popular unsupervised
thaw cycles for metabolite extraction needs to be performed in the data-analysis tool is PCA. Subsequent analysis of the metabolome
solution collected and, following high-speed centrifugation, can assist in understanding the link between factors (e.g., genetic
analysis can be carried out in the supernatant [80]. A similar pro- mutation, natural variation, environmental conditions and the
tocol can be applied to the preparation of tissue samples, where overall characteristics of the subjects being considered).
freeze-thaw cycles are used in the process of tissue The application of these technologies in cancer research has re-
homogenization. vealed system-wide alterations of unexpected metabolic pathways
I.M. Di Gangi et al. / Trends in Analytical Chemistry 55 (2014) 94–116 103
Table 1
Metabolite-extraction techniques employed in metabolomics studies of pancreatic cancer
related to phenotypic perturbations [72]. Wen et al. [59] applied A previous approach to classification of premalignant PC was
NMR-based metabolomics to bile collected from individuals with performed by Ge and Wong [85] using three feature-selection
biliary tract cancer or benign biliary tract disease, including pat- schemes (Student t test, Wilcoxon rank sum test and genetic algo-
tern recognition and multivariate data analysis to note significant rithm) to compensate for systematic differences due to sample
differences between individuals with cancer and benign diseases, loadings and instrument errors and to reduce the dimensionality
compared to other conventional approaches. of the proteomic dataset to a manageable number [85].
104 I.M. Di Gangi et al. / Trends in Analytical Chemistry 55 (2014) 94–116
Table 2
Metabolites found differentially expressed in PC metabolomics studies included in this comprehensive review. The table shows chemical formula, molecular weight, PubChem ID,
platform(s) and matrix used for the identification of each compound
Metabolite Chemical Molecular Pubchem ID Matrix NMR GC-MS LC- CE- Fold inductiona
formula weight MS MS
Amino acids and
Peptides
Alanine (Ala) C3H7NO2 89.0932 CID 5950 Plasma [50] ;
Serum [44] [41,45] [42] 1.23 [41]
Saliva [56] 3.67
Urine [47] M
Arginine (Arg) C6H14N4O2 174.2010 CID 6322 Serum [44] [42] " [44]
Asparagine (Asn) C4H8N2O3 132.1179 CID 6267 Serum [44] [41,45] ; (x0.87) [41]; " (x1.35)
[45]; ; [44]
Aspartic acid (Asp) C4H7NO4 133.1027 CID 424 Serum [44] [41,45] [42] 0.98 [41]
C4H7NO4 133.1027 CID 424 Saliva [56] 4.1
Beta-Alanine (b-Ala) C3H7NO2 89.0932 CID 239 Saliva [56] 3.04
3-Methylhistidine (MeHis) C7H11N3O2 169.1811 CID 64969 Serum [42]
Urine [47] M
4-Hydroxyproline C5H9NO3 131.1299 CID 5810 Serum [41,45] 1.17 [41]
5-Oxoproline (Pyroglutamic acid) C5H7NO3 129.1140 CID 7405 Serum [44] [45]
Alpha-Aminobutyric acid C4H9NO2 103.1198 CID 80283 Saliva [56] 4.01
Urine [46] ""
Carnosin (Carn) C9H14N4O3 226.2325 CID 439224 Serum [42]
Citrulline (Cit) C6H13N3O3 175.1857 CID 9750 Saliva [56] 3.1
Serum [42]
Creatine C4H9N3O2 131.1332 CID 586 Plasma [50]
Serum [44] ;
Cysteine + Cystine CID 5862 + CID Serum [41] 1.02
67678
Gamma-Aminobutyric acid (GABA) C4H9NO2 103.1198 CID 119 Saliva [56] 3.31
Glutamic acid (Glu) C5H9NO4 147.1293 CID 33032 Plasma [50] ;
Serum [44] [41,45] [42] "" [44]; 1.33 [41]
Saliva [56]
Glutamine (Gln) C5H10N2O3 146.1445 CID 5961 Serum [44] [41,45] [42] ; (x0.86) [41]; ; [44]
C5H10N2O3 146.1445 CID 5961 Plasma [50] [48] ; (x1.20) [48]; ; [50]
C5H10N2O3 146.1445 CID 5961 Saliva [56] 4.96
Glutathione C10H17N3O6S 307.3235 CID 124886 Tissue [57] ; (x0.6)
Glycine (Gly) C2H5NO2 75.0666 CID 750 Serum [44] [41,45] [42] 0.98 [41]; ; (x0.76) [45]
Urine [47] ;
Saliva [56] 3.1
Glycylglycine (Gly-Gly) C4H8N2O3 132.1179 CID 11163 Serum [41] 0.76
Histidine (His) C6H9N3O2 155.1546 CID 6274 Plasma [50] ;
Serum [44] [41] [42] ; (x0.7) [41]
Saliva [56] 2.02
Hippuric acid C9H9NO3 179.1727 CID 464 Urine [47] ;
Hydroxyproline (trans) C5H9NO3 131.1299 CID 5810 Serum [42]
Isoleucine (Ile) C6H13NO2 131.1729 CID 6306 Serum [43,44] [45] ;; [34]
Plasma [50] ;
Leucine (Leu) C6H13NO2 131.1729 CID 6106 Serum [43,44] [45] " [34]
Urine [47] "
Plasma [50] ;
Leucine + Isoleucine (Leu+Ile) CID 6106 + CID Saliva [56] 7.71
6305
Serum [42]
Lysine (Lys) C6H14N2O2 146.1876 CID 5962 Plasma [50] [48] " (x1.03) [48]; ; [50]
Serum [44] [41] [42] ; (x0.78) [41]; ; [44]
Saliva [56] 3.97
Methionine (Met) C5H11NO2S 149.2113 CID 6137 Serum [44] [41,45] [42] ; (x0.86) [41]
N-Acetyltyrosine C11H13NO4 223.2252 CID 89216 Serum [45] " (x1.57)
N-Acetylglutamic acid (NAG) C7H11NO5 189.1659 CID 70914 Plasma [50] "
N-Methylalanine C4H9NO2 103.1198 CID 5288725 Plasma [48] "" (x2.81)
Ornithine (Orn) C5H12N2O2 132.1609 CID 6262 Serum [44] [41] [42] 1.11 [41]
Saliva [56] 1.97
Phenylalanine (Phe) C9H11NO2 165.1891 CID 6140 Serum [44] [41,45] [42] 0.92 [41]; " [44]
Plasma [50] [48] ; (x1.15) [48]
Saliva [56] 3.54
Pro-Gly-Pro or Pro-Pro-Gly Saliva [56] 4.27
Proline (Pro) C5H9NO2 115.1305 CID 8988 Serum [44] [41,45] [42] 0.88 [41]; ; [44]
Saliva [56] 3.99
Sarcosine (Sarc) C3H7NO2 89.0932 CID 1088 Serum [42]
Serine (Ser) C3H7NO3 105.0926 CID 5951 Serum [44] [45] [42]
CID 5951 Saliva [56] 4.34
I.M. Di Gangi et al. / Trends in Analytical Chemistry 55 (2014) 94–116 105
Table 2 (continued)
Metabolite Chemical Molecular Pubchem ID Matrix NMR GC-MS LC- CE- Fold inductiona
formula weight MS MS
Amino acids and Taurine (Tau) C2H7NO3S 125.1469 CID 1123 Serum [42]
Peptides
Urine [46] "
Saliva [56] 0.376
Tissue [57] " (x1.6)
Threonine (Thr) C4H9NO3 119.1192 CID 6288 Serum [44] [41,45] [42] ; (x0.86) [41]; ; [44];
Urine [46] "
Saliva [56] 4.75
Tryptophan (Trp) C11H12N2O2 204.2252 CID 6305 Serum [44] [45] [42] ; [44]
Urine [46] "
Saliva [56] 6.47
Tyrosine (Tyr) C9H11NO3 181.1885 CID 6057 Serum [44] [41,45] [42] ; (x0.84) [41];
Plasma [50]
Saliva [56] 2.9
Valine (Val) C5H11NO2 117.1463 CID 6287 Serum [44] [41,45] [42] ; (x0.81) [41]
Plasma [50] ;
Urine [47] M
Saliva [56] 5.92
Lipids
Cardiolipin C81H158O17P2 1466.0585 CID 449005 Plasma [56] M
Glycerophosphocholine C8H20NO6P 257.2213 CID 657272 Saliva [56] 0
Glyceryl of lipids Plasma [50] "
LDL, -CH3-(CH2)n Plasma [50] ;
Lipid, -CH2-C=O Plasma [50]
Lipid, -CH2-CH=CH Plasma [50]
LysoPC(16:0) (Palmitoyl C24H50NO7P 495.6301 CID 86554 Plasma [48] " (x1.33)
lysophosphatidyl choline)
LysoPC(18:2) (1-18:2- C26H50NO7P 519.6515 CID 11988421 Plasma [48] " (x1.59)
lysophosphatidylcholine)
Posphatidylcholine PC (34:2) C42H80NO8P 758.0603 CID 6021688 Plasma [48] " (x1.32)
Posphatidylethanolamine, PE(12:0/ C31H62NO8P 607.7996 CID 9546763 Plasma [48] " (x1.81)
14:0)
Posphatidylinositol PI (18:0,18:2) Plasma [56] ;
m/z 861
Posphatidylinositol PI (18:0,20:2) Plasma [56] ;
m/z 887
Posphatidylinositol PI (18:0,20:4) Plasma [56] ;
m/z 885
Posphatidylserine C13H24NO10P 385.3041 CID 44147587 Plasma [56] M
Triglycerides Serum [43] ""
VLDL, -CH2-CH2-C=O Plasma [50] "
VLDL, -CH3-(CH2)n Plasma [50] "
VLDL, -CH3-(CH2)n Plasma [50] "
Fatty acids
2-Aminobutyrate (Butyric acid) C4H9NO2 103.1198 CID 517460 Serum [44]
2-Phenylacetamide C8H9NO 135.1632 CID 7680 Urine [47] "
2-Hydroxybutyric acid C4H8O3 104.1045 CID 11266 Serum [44] [41,45] 0.97 [41]
2-Hydroxyglutaric acid C5H8O5 148.1140 CID 43 Serum [45]
2-Hydroxyisobutyric acid C4H8O3 104.1045 CID 11671 Serum [44]
Urine [46,47] " [46]; M [47]
2-Hydroxyisovaleric acid C5H10O3 118.1311 CID 99823 Serum [44] [45]
3-Hydroxybutyric acid C4H8O3 104.1045 CID 441 Serum [43,44] [41,45] ;; [43]; " [44]; 1.7 [41];
Plasma [50] ;
3-Hydroxyisobutyric acid C4H8O3 104.1045 CID 87 Serum [44] [45] " [44]
3-Hydroxyisovaleric acid C5H10O3 118.1311 CID 69362 Serum [43,44] ;; [43]
Urine [47] ;
3-Hydroxyphenylacetic acid C8H8O3 152.1473 CID 12122 Serum [45]
3-Hydroxypropionic acid C3H6O3 90.0779 CID 68152 Serum [45]
3-Methyl-2-oxovaleric acid C6H10O3 130.1418 CID 47 Serum [44] "
4-Hydroxyphenylacetic acid C8H8O3 152.1473 CID 127 Serum [45]
Urine [46,47] " [46]; M [47]
7-Hydroxyoctanoic acid C8H16O3 160.2108 CID 167627 Serum [45] " (x1.38)
Acetic acid C2H4O2 60.0520 CID 176 Serum [44]
Arachidonic acid C20H32O2 304.4669 CID 444899 Plasma [48] " (x1.49)
Decanoic acid C10H20O2 172.2646 CID 2969 Serum [45] ; (x0.76)
Isobutyric acid (2-Methylpropionic C4H8O2 88.1051 CID 6590 Serum [44]
acid)
Urine [47] M
Isovaleric acid (3-Methylbutyric C5H10O2 102.1317 CID 10430 Serum [44]
acid)
Lauric acid (n-Dodecanoic acid) C12H24O2 200.3178 CID 3893 Serum [41,45] 0.79 [41]; ; (x0.76) [45]
Margaric acid (n-Heptadecanoic C17H34O2 270.4507 CID 10465 Serum [45] ; (x0.80)
acid)
Table 2 (continued)
Metabolite Chemical Molecular Pubchem ID Matrix NMR GC-MS LC- CE- Fold inductiona
formula weight MS MS
Fatty acids Myristic acid (n-Tetradecanoic C14H28O2 228.3709 CID 11005 Serum [45] ; (x0.77)
acid)
N-Caprylic acid (n-Octanoic acid) C8H16O2 144.2114 CID 379 Serum [41,45] ; (x0.77) [41]; ; (x0.73)
[45]
Nonanoic acid (C9) C9H18O2 158.2380 CID 8158 Serum [41] ; (x0.53)
Palmitic acid (Hexadecanoic acid) C16H32O2 256.4241 CID 985 Serum [45] ; (x0.79)
Palmitoleic acid (cis-9- C16H30O2 254.4082 CID 445638 Serum [45]
Hexadecenoic acid)
Stearic acid (n-Octadecanoic acid) C18H36O2 284.4772 CID 5281 Serum [45] ; (x0.74)
Suberic acid C8H14O4 174.1944 CID 10457 Serum [44]
Sterols
Cholesterol C27H46O 386.6535 CID 5997 Plasma [48] " (x1.85)
HDL, cholesterol Plasma [50] ;
Bile acids
Bile salts Urine [47] M
Glycocholic Acid (Cholylglycine) C26H43NO6 465.6227 CID 23617285 Plasma [48] "" (x2.61)
Glycodeoxycholic Acid C26H43NO5 449.6233 CID 22833539 Plasma [48] " (x1.31)
(Deoxycholylglycine)
Taurocholic acid C26H45NO7S 515.7030 CID 6675 Plasma [48] " (x1.75)
Tauroursodeoxycholic acid C26H45NO6S 499.7036 CID 3034759 Plasma [48] "" (x2.01)
(TUDCA)
Thiodiglycolic acid (TDGA) C4H6O4S 150.1530 CID 31277 Serum [45] "" (x6.27)
Other organic acids
2-Oxoisocaproate (Alpha- C6H10O3 130.1418 CID 70 Serum [44]
ketoisocaproic acid)
Abscisic acid C15H20O4 264.3169 CID 5280896 Serum [42]
Acetoacetic acid C4H6O3 102.0886 CID 96 Urine [47] "
Aconitic acid (Cis) C6H6O6 174.1082 CID 643757 Serum [45] " (x1.54)
Urine [46] "
Aconitic acid (Trans) C6H6O6 174.1082 CID 444212 Urine [46] "
Adipic acid C6H10O 146.1412 CID 196 Serum [44]
Citraconic acid C5H6O4 130.0987 CID 643798 Serum [45]
Citric acid C6H8O7 192.1235 CID 311 Plasma [50] ;
Serum [44] [45]
Urine [47] ;
Citric acid + Isocitric acid CID 311 + CID Serum [41] 1.1
5318532
Formic acid CH2O2 46.0254 CID 284 Serum [44] "
Plasma [50]
Urine [47] M
Fumaric acid C4H4O4 116.0722 CID 444972 Serum [45]
Glucuronic acid (GlcU) C6H10O7 194.1394 CID 444791 Serum [41] 4.57
Bile [53] ""
Glucuronolactone C6H8O6 176.1241 CID 92283 Serum [45]
Glutaconic acid C5H6O4 130.0987 CID 5280498 Serum [45]
Glyceric acid C3H6O4 106.0773 CID 752 Serum [45] ; (x0.53)
Glycolic acid C2H4O3 76.0514 CID 757 Serum [41,45] 0.97 [41]
Glyoxylic acid C2H2O3 74.0355 CID 760 Serum [45]
Homogentisic acid C8H8O4 168.1467 CID 780 Serum [45] " (x1.26)
Hydrocinnamic acid (3- C9H10O2 150.1745 CID 107 Plasma [48] ; (x1.38)
phenylpropionic acid)
Isocitric acid C6H8O7 192.1235 CID 1198 Serum [45]
Lactic acid (2-hydroxypropionic C3H6O3 90.0779 CID 612 Serum [43,44] [41,45] 1.14 [41]; ; [43]; "
acid) (x1.48) [45]
Plasma [50] ;
Urine [47] M
Malic acid C4H6O5 134.0874 CID 525 Tissue [57] ; (x0.3)
Malonic acid C3H4O4 104.0615 CID 867 Serum [44] [45]
Methylsuccinic acid C5H8O4 132.1146 CID 10349 Serum [44]
Oxalic acid C2H2O4 90.0349 CID 971 Serum [45]
Pyruvic acid C3H4O3 88.0621 CID 1060 Serum [44] [45]
Pyruvic acid + Oxalacetic acid CID 1060 + CID Serum [41] 1.13
970
Succinic acid C4H6O4 118.0880 CID 1110 Serum [44]
Tissue [57] ; (x0.3)
Tartaric acid C4H6O6 150.0868 CID 875 Serum [45]
Carbohydrates
Sugars
1,5-Anhydro-D-glucitol C6H12O5 164.1565 CID 64960 Serum [41] ; (x0.55)
Arabinose C5H10O5 150.1299 CID 66308 Serum [41] 1.75
Arabitol C5H12O5 152.1458 CID 94154 Serum [41] 0.96
Erythritol (Meso-erythritol) C4H10O4 122.1198 CID 222285 Plasma [48] " (x1.53)
CID 222285 Serum [41] 2.48
I.M. Di Gangi et al. / Trends in Analytical Chemistry 55 (2014) 94–116 107
Table 2 (continued)
Metabolite Chemical Molecular Pubchem ID Matrix NMR GC-MS LC- CE- Fold inductiona
formula weight MS MS
Carbohydrates Fucose C6H12O5 164.1565 CID 19466 Urine [46] ""
Glucose C6H12O6 180.1559 CID 5793 Urine [46,47] " [46,47]
Serum [44] [41] 1.05 [41]; " [44]
Glucose-a C6H12O6 180.1559 CID 79025 Plasma [50]
Glucose-b C6H12O6 180.1559 CID 64689 Plasma [50]
Inositol C6H12O6 180.1559 CID 892 Serum [41] 1.13
Mannose C6H12O6 180.1559 CID 18950 Serum [44] [41] 0.97 [41]; " [44]
Ribulose C5H10O5 150.1299 CID 151261 Serum [41] 1.39
Xylitol C5H12O5 152.1458 CID 6912 Serum [41] 0.75
Xylose C5H10O5 150.1299 CID 644160 Urine [46] "
Nucleotides and nucleosides
50 -AMP (Adenosine C10H14N5O7P 347.2212 CID 6083 Tissue [57] ; (x0.5)
50 -monophosphate)
50 -UMP (Uridine C9H13N2O9P 324.1813 CID 6030 Tissue [57] ; (x0.4)
50 -monophosphate)
Adenine C5H5N5 135.1267 CID 190 Urine [47] M
Hypoxanthine C5H4N4O 136.1115 CID 790 Serum [44]
Urine [46] ""
Saliva [56] 3.35
Inosine (Hypoxanthosine) C10H12N4O5 268.2261 CID 6021 Plasma [48] ; (x1.40)
Nicotinamide adenine dinucleotide C21H27N7O14P2 663.4251 CID 5288979 Tissue [57] ; (x0.08)
(NAD)
UDP-N-acetyl-d-glucosamine C17H27N3O17P2 607.3537 CID 445675 Tissue [57] ; (x0.3)
(UDP-GlcNAc)
Uridine C9H12N2O6 244.2014 CID 6029 Tissue [57] ; (x0.1)
Chemicals and Drugs
Category
Alcohols
Ethanol C2H6O 46.0684 CID 702 Serum [44] ;
Glycerol C3H8O3 92.0938 CID 753 Serum [44] [41,45] 0.86 [41]; ; [44]
Methanol CH4O 32.0419 CID 887 Urine [46] ;
Serum [44] ;
Propylene Glycol C3H8O2 76.0944 CID 1030 Serum [44]
Acetylated compounds
Acetylated compounds Urine [47] "
O-Acetylcarnitine C9H17NO4 203.2356 CID 18230 Serum [44]
Urine [46] "
Amines
Betaine C5H11NO2 117.1463 CID 247 Serum [44]
Saliva [56] 0.822
Carnitine C7H15NO3 161.1989 CID 10917 Serum [44]
Saliva [56] 1.88
Choline C5H14NO+ 104.1708 CID 305 Serum [44]
Urine [46] "
Saliva [56] 2.63
Cadaverine C5H14N2 102.1781 CID 273 Saliva [56] 6.15
Dimethylamine (DMA) C2H7N 45.0837 CID 674 Serum [44] "
Urine [46,47] " [46]; M [47]
Plasma [50] "
Ethanolamine C2H7NO 61.0831 CID 700 Serum [41] 0.91
Saliva [56] 4.35
O-Phosphoethanolamine C2H8NO4P 141.0630 CID 1015 Serum [41] 0.77
Putrescine C4H12N2 88.1515 CID 1045 Saliva [56] 3.98
Trimethylamine C3H9N 59.1103 CID 1146 Saliva [56] 4.13
Trimethylamine-N-oxide (TMAO) C3H9NO 75.1097 CID 1145 Urine [46] ""
Serum [43] ;
Urine [47] M
Tryptamine C10H12N2 160.2157 CID 1150 Plasma [48] ; (x1.07)
Heterocyclic
Compounds
1-Methylnicotinamide C7H9N2O+ 137.1592 CID 457 Urine [46,47] "" [46]; M [47]
4-Pyridoxic acid C8H9NO4 183.1614 CID 6723 Urine [46] "
Serum [44] ;
Creatinine C4H7N3O 113.1179 CID 588 Serum [43,44] [41] 0.82 [41]; " [43]
Urine [47] ;
Indolelactic acid C11H11NO3 205.2099 CID 92904 Serum [45]
Methylimidazoleacetic acid C6H8N2O2 140.1399 CID 75810 Saliva [45] [56] 4.64
Pipecolic acid (PiPA) C6H11NO2 129.1570 CID 849 Serum [42]
Saliva [56] 1.11
Piperidine C5H11N 85.1475 CID 8082 Saliva [56] 3.73
Pyridine (1-Piperideine) C5H9N 83.1317 CID 68161 Saliva [56] 0.665
Pyrroline hydroxycarboxylic acid C5H7NO3 129.1140 CID 1059 Saliva [56] 2.28
Table 2 (continued)
Metabolite Chemical Molecular Pubchem ID Matrix NMR GC-MS LC- CE- Fold inductiona
formula weight MS MS
Heterocyclic Trigonelline C7H7NO2 137.1360 CID 5570 Urine [46,47] ; [46,47]
Compounds Urea CH4N2O 60.0553 CID 1176 Serum [44] [45] ; [44]; ; (x0.80) [45]
Uric acid C5H4N4O3 168.1103 CID 1175 Serum [45] 0.85 [41]; ; (x0.61) [45]
Inorganic chemicals
Phosphate (Triphosphoric acid) H5O10P3 257.9550 CID 983 Serum [41] 0.93
Phosphoric acid H3O4P 97.9952 CID 1004 Serum [45]
[45]
Neurotransmitter Agents
Burimamide C9H16N4S 212.3151 CID 3032915 Saliva [56] 2.56
Organic chemicals
4-Cresol (para-Cresol) C7H8O 108.1378 CID 2879 Serum [45]
Acetone C3H6O 58.0791 CID 180 Plasma [50] [45] "
Urine [46] ""
Serum [44] "
Pyrogallol (Pyrogallic acid) C6H6O3 126.1100 CID 1057 Serum [41] 0.71
Unknowns
compounds
C17H26N4O5 Saliva [56] 2.69
C18H32N6O6 Saliva [56] 6.7
C2H6N2 Saliva [56] 4.42
C30H55N27O3S Saliva [56] 5.56
C30H62N19O2S3 Saliva [56] 6.65
C32H48O13 Saliva [56] 2.03
C3H7NO2 Saliva [56] 3
C4H12N5 Saliva [56] 6.98
C4H5N2O11P Saliva [56] 0
C4H7N Saliva [56] 1.14
C4H9N Saliva [56] 2.7
C4H9NO2 Saliva [56] 5.16
C5H11NO2 (10.05 min.) Saliva [56] 2.1
C5H11NO2 (13.42 min.) Saliva [56] 2.17
C5H14N5 Saliva [56] 5.35
C6H6N2O2 Saliva [56] 2.89
C7H12N2O3 Saliva [56] 2.48
C7H8O3S Saliva [56] 0.592
C8H9N Saliva [56] 0.817
Unknown 132.0773 m/z Tissue [57] 4.4
Unknown 187.1201 m/z Plasma [48] ; (x1.11)
Unknown 188.1283 m/z Tissue [57] 2.0
Unknown 236.9944 m/z Tissue [57] 4.3
Unknown 265.0721 m/z Plasma [48] " (x1.17)
Unknown 287.2436 m/z Tissue [57] 87.9
Unknown 332.0745 m/z Plasma [48] " (x1.37)
Unknown 366.1397 m/z Tissue [57] 14.1
Unknown 414.1563 m/z Plasma [48] " (x1.69)
Unknown 459.2029 m/z Tissue [57] 3.52
Unknown 523.9791 m/z Tissue [57] 73.1
Unknown 633.1954 m/z Plasma [48] " (x1.80)
Unknown 753.1246 m/z Plasma [48] " (x1.27)
The citation number is listed in the column under the platform used to detect each metabolite.
Some metabolites were identified by more platforms in different matrices and each author reported different value of fold induction.
a
Fold induction values are reported as calculated from the PC to controls ratio. Data were not reported when not available in the study.
A recent study, conducted using an intrinsic three-class bioin- allow the user to analyze metabolomics data and to link them with
formatics approach, allowed comparison of the metabolomics pro- other ‘omics’ data [86].
files of patients with pancreatitis and pancreatic carcinoma, General information about the physical and chemical properties
evaluating their selectivity and computing multi-marker panels of metabolites can be obtained by searching general chemical
combined with the conventional tumor marker CA 19-9 to differ- databases (e.g., PubChem, ChemSpider or CAS), which can also be
entiate PC from various benign lesions [42]. useful for identifying the chemical structure of unknowns by start-
ing with elemental composition, obtained from accurate mass
7. Commercial software and library sources measurements. The development of specific metabolomics MS dat-
abases is in progress (e.g., the METLIN database catalogues metab-
One of the major challenges of informatics is to provide tools that olites, MS/MS spectra and the LC–MS profiles of human plasma and
link metabolomics data with other types of high-throughput molec- urine samples) [87].
ular data (e.g., transcriptomics, genomics and proteomics) and incor- Biochemical databases can later be used to identify unknown
porate prior knowledge of pathways and molecular interactions. metabolites or to determine the biological function of identified
While many tools have been developed for the analysis of gene- metabolites. Information about biochemical pathways and the
expression and proteomics data, to date, there are few tools that metabolites involved is available in the KEGG [88], BioCyc [89]
I.M. Di Gangi et al. / Trends in Analytical Chemistry 55 (2014) 94–116 109
and SMPD [90] databases, which provide data visualization in the kinds of experiments, mainly with an untargeted approach, as
form of individual pathway charts or an overall view of metabolic mentioned above.
pathways. GC–MS was responsible for identifying 49.8% (108) of all metab-
More specific lipidomics databases exist, such as LipidMaps olites, a higher percentage than LC–MS, which identified 31.8%
[91,92], SphinGOMAP [93] and Lipid Bank [94], containing struc- (69), whereas a single CE-MS study, using CE-TOF-MS [56] identi-
tural and nomenclatural information as well as standard analytical fied 56 (25.8%) of the metabolites in saliva samples, mainly amino
protocols. acids, listed in Table 2.
In 2010, Gao et al. [95] developed Metscape software as a plug- Amino acids are also the most common metabolites identified
in for Cytoscape, to visualize and to interpret metabolomics data in in plasma, serum or urine of PC subjects. Tyrosine, glutamate and
the context of human metabolic networks [96]. The new version, histidine were detected in serum samples using both GC–MS and
MetScape 2 [97], provides functions for creating pathways and net- LC–MS and using NMR in serum and plasma.
work-level views and analyzing several types of experimental data. The results were quite different for glutamate concentration in
It allows users to: serum (significantly increased) [44] and in plasma (decreased)
[50], whereas data for Tyrosine and Histidine were in agreement
upload lists of metabolites and genes with experimental across different authors. High levels of glutamate, acetone and
measurements; 3-hydroxybutyrate were observed in the sera of cancer patients
identify related genes, metabolites, reactions, enzymes and [44], while increased glucose, phenylalanine, formate and mannose
pathways; were found to correlate with both disease and age. Differences
build and analyze gene and metabolite networks; between the serum metabolic profiles of PC patients and healthy
visualize changes in experimental data over time or experimen- controls have also been reported [43], although the investigation
tal conditions; and, finally, was limited to PCA.
to identify and to visualize enriched pathways from expression- Ornithine, serine and proline were found to be significantly dif-
profiling data. ferent in PC patients, compared to controls in sera, and valine, gly-
cine, alanine, threonine and tryptophan showed a trend towards
Metscape 2 uses an internal relational database that integrates significance in urine, plasma or serum analyzed using NMR,
data from the KEGG and EHMN (Edinburgh Human Metabolic Net- GC–MS or LC–MS.
work) [98] databases. Metscape 2 has been applied to obtain a bet- All these amino acids were also identified by Sugimoto et al.
ter understanding of the underlying metabolic processes [56] in saliva samples using the CE-MS method of analysis
associated with PC [97]. However, many of the molecules detected (Table 2).
are currently not included in databases and metabolite reposito- Highly-significant cancer-specific increases in threonine, tryp-
ries, showing the extent to which our picture of cellular metabo- tophan and 4-hydroxyphenylacetate in urine were noted by Davis
lism in relation to PC is incomplete [99,100]. et al. [46], who surmised that this pattern reflected muscle-protein
breakdown, during which all the constituent amino acids entered
oxidative pathways. Effectively, the highest incidences of cachexia
8. Analysis and summary of the results obtained in PC studies occurred among PC patients with solid-epithelial malignancies, re-
flected in a weight loss of more than 5% at diagnosis. Cachexia re-
Different classes of compounds (e.g., amino acids, organic acids, sults in severe metabolic disturbances in energy and protein
sugars, and fatty acids) were analyzed and identified in PC as metabolism and takes place within a complex metabolic syndrome
metabolites with a relevant significance in the development and characterized by anorexia and high rates of fat and skeletal muscle
the progression of the disease [41–57]. degradation [101]. However, some authors instead observed
All the metabolites identified in PC patients using the metabolo- decreasing levels of threonine [41,44], tryptophan [44] and other
mics-based approaches described in this review are listed in amino acids (valine, n-caprylic acid, nonanoic acid, methionine,
Table 2. asparagine, glutamine, lysine, histidine and tyrosine) as well as
NMR spectroscopy and GC–MS have provided identification for long-chain fatty acids in the serum samples of PC [55,62] and CP
most of these metabolites, which were verified using reference patients [50], relating these significant decreases to the pathogen-
standards or metabolomics databases (KEGG, PubChem or HMDB). esis of pancreatic disease. These findings were supported by other
It is evident from Table 2 that known metabolites can be research groups, which found similar decreases in plasma samples
grouped into four principal classes: from both PC and CP using HPLC analysis [102].
It is well known that the uptake and the catabolism of amino
amino acids; acids and fatty acids are improved to support rapid cell prolifera-
fatty acids (fatty acyls, glycerophospholipids, glycerolipids and tion in cancer tissues, and these decreases may be explained as a
sterols); result of enhanced usage in tumors [103]. Patients with pancreatic
carbohydrates; and, disease are also troubled by malnutrition because of pancreatic
bile acids. endocrine and exocrine insufficiency [104], so there is a possibility
that decreases in their serum-metabolite levels may also reflect
Some metabolite classes are better detected across platforms malnutrition.
than others (e.g., many amino acids are usually detected by analy- Schrader et al. [102] suggested mainly inflammatory effects –
sis with NMR or MS coupled with GC or LC). apart from malnutrition – and pointed at the inverse relationship
However, some metabolites (acetone, creatinine, dimethyl- between circulating amino-acid concentrations and the degree of
amine, glucose, glutamine, glycine, isoleucine, lactate, leucine, ly- inflammation present (e.g., in hemodialysis patients) [102].
sine, methanol, phenylalanine, taurine, threonine, tryptophan, Whether increased tumor-associated proteolytic activity [105]
tyrosine, urea and valine) identified by two or more studies contributes to not only the generation of specific peptide-decay
showed similar or contrasting results in terms of increased or de- profiles, but also the specificity of the amino-acid profiles is still
creased levels in the different matrices. unknown.
NMR techniques have contributed towards discovering about The untargeted metabolomics study on urine, performed by Na-
73.7% (135) of the total number of 217 metabolites, using different poli and co-workers [47], confirmed the association of the metab-
110
I.M. Di Gangi et al. / Trends in Analytical Chemistry 55 (2014) 94–116
Fig. 4. The interaction between metabolic pathways, drug resistance and downstream signaling pathways. K-Ras and the p53 tumor-suppressor gene are two of the main genes mutating in PC, linked to impaired metabolic
pathways. In K-Ras⁄ mutated cells, high levels of glucose are subjected to fermentation in anaerobic conditions, leading to increased production of lactic acid. Moreover, high levels of glucose concentrations and/or of arginine,
inhibit the complex formation of hCHOP-C/EBPa, down-regulating hENT1 expression and consequently inhibiting the transport of chemotherapeutic compounds. K-Ras⁄ modulates glutamine concentrations, which are involved in
cell survival, whilst, in p53⁄ cells, cell survival and cell cycle are regulated by serine concentration. For further details, please refer to the text.
I.M. Di Gangi et al. / Trends in Analytical Chemistry 55 (2014) 94–116 111
olites produced in ketogenesis (e.g., acetoacetate) with the disease correlated positively with cancer. Furthermore, it has been sug-
[47], as well as a decreased level of citrate, which has often been gested that the growth-promoting effect of lipids on PC cells occurs
correlated with a down-regulated TCA cycle as a consequence of at multiple levels:
the so-called ‘‘Warburg effect’’ [106].
A reduction in the serum level of 1,5-anhydro-D-glucitol was transduction of signals induced by hormones;
observed in PC (Table 2) by Kobayashi et al. [41], using a GC–MS a structural role in cell membranes; and,
metabolomics-based approach. energy supply [115].
1,5-Anhydro-D-glucitol is usually a serum biomarker for short-
term glycemic control and a decreased serum level for this metab- Chronic inflammation is heavily mediated by lipid-signal-trans-
olite shows the presence of hyperglycemia and glycosuria in the duction pathways, and correlation between chronic inflammation
last few days [103]. These results suggest that glucose tolerance and the development of cancer has been recognized in recent years
was impaired in these patients because of pancreatic insufficiency, [116].
reflecting systemic responses to the metabolic changes induced by In a different study, MS metabolic profiling was applied to the
pancreatic disease [41]. plasma samples of PC patients, proposing as candidate biomarkers
High cancer-specific levels of hypoxanthine, choline, trimethyl- fatty-acid arachidonic acid, lipids lysoPC(18:2), PC(34:2), and
amine-N-oxide (TMAO), O-acetylcarnitine and acetone were PE(26:0), and bile acids tauroursodeoxycholic acid, taurocholic
observed in the serum or plasma and urine of PC patients using acid, deoxycholylglycine and cholylglycine. Increased levels of ara-
NMR analysis, and the results of the different studies were consis- chidonic acid in the plasma of PC patients may be a result of sys-
tent, with the exception of TMAO, which showed different trends temic equilibrium of the compounds involved in inflammation
[43,46]. These changes, perhaps indicative of metabolic distur- [48].
bances associated with tumor metabolism, may represent an en- In general, a high concentration of omega-6 fatty acids (e.g., ara-
hanced capacity for DNA synthesis and energy production chidonic acid) promotes inflammation, increasing the production
(hypoxanthine), cell-membrane formation (choline, trimethyla- of pro-inflammatory enzymes and cytokines, including COX-2,
mine-N-oxide), as well as increased rates of lipolysis and fatty acid tumor necrosis factor (TNF-a), and interleukin 1beta (IL-1b),
metabolism (O-acetylcarnitine, acetone) seen in rapidly proliferat- whereas omega-3 fatty acids (e.g., eicosapentaenoic acid and doco-
ing tumor cells [107–110]. sahexaenoic acid) have anti-inflammatory properties [117]. Fur-
Down-regulation of citric-acid-cycle intermediates succinate thermore, TNF-a is a potent activator of nuclear transcription
and malate may have an overall impact on the energy metabolism factor NF-kB, which controls the expression of numerous genes in-
of the cell, while lower levels of uridine, 50 -uridine monophosphate volved in inflammation and immune response processes, including
(50 -UMP) and 50 -adenosine monophosphate (50 -AMP) could reflect proliferation, invasion and adhesion, angiogenesis and apoptosis
rapid turnover of these nucleotides in tumor tissue [57]. The [118]. Consistent with this, TNF-a, IL-1 and NF-kB were over-ex-
authors also observed down-regulation of the powerful antioxi- pressed and activated in PC patients [119,120].
dant glutathione in pancreatic tumor tissue [57]. A similar de- Significant increased levels of dimethylamine (DMA), formate
crease in the serum levels of glutathione has been observed in and acetone were observed in urine, plasma and serum [46,47],
breast carcinoma with a simultaneous increase in lipid peroxida- and significant increased levels of NAG and VLDL were reported
tion in plasma, which becomes more pronounced with ageing of in plasma using NMR analysis [50]. This pattern was integrated
the patients [111]. with significant decreased levels of HDL, LDL and citrate, confirmed
Moreover, increased levels of the amino-acid taurine in the pan- by NMR and GC-MS in the plasma, serum or urine of PC tumors
creatic tumor tissue [57], serum and urine of PC patients [46], com- [41,50] (Table 2).
pared to the normal controls, were found, as shown in Table 2. The overall significance of these results may be explained by
Taurine is a major constituent of bile and has been reported to considering the role of possible immune-system impairment in
be significantly higher in urinary bladder cancer [111]. pancreatic disease, manifested through the increase in acetylated
Metabolomic profiling of the plasma of PC patients by Urayama glycoproteins levels (high levels of NAG in PC cases), involved in
et al. [48] showed a rise in taurocholic acid and tauroursodeoxy- white blood-cell recognition, especially in mammals [121,122].
cholic acid, which are also constituents of bile. The study of Beza- The high VLDL level in the plasma of the PC group, associated with
beh and colleagues [53] focused on the analysis of D-glucuronic low LDL and HDL concentrations, was related to a well-known in-
acid, which was very high in bile samples obtained from PC creased risk of coronary artery disease [123]. DMA, an animal
patients, whereas it was absent or negligible in control and CP pa- carcinogen, could instead be related to the possibility of increased
tients. The presence of high levels of GlcU, observed in pancreatic damage to the genome or disruption of cellular metabolic pro-
tumor, may have diagnostic implications, but, at the moment, the cesses [124].
reason remains unclear. D-Glucuronic acid is synthesized in the li- Data on lactate levels are instead partly inconsistent for differ-
ver from UDP-glucose mediated by UDP-glucose dehydrogenase, ent authors. An increase in the cancer serum levels of lactate was
and is involved in a number of key detoxification pathways by reported in PDAC, using GC-MS [45], confirmed in pancreatic
removing a variety of non-polar drugs, environmental toxins and tumor tissue [52] and in many human cancers [125–127], whereas
carcinogens from the body [112]. It is also responsible for the analysis of plasma serum and urine using 1H-NMR showed a con-
removal of bilirubin, a major end-product of heme catabolism. In trasting trend [43,50]. Lactate is the product of anaerobic glycolysis
the liver, D-glucuronic acid is conjugated to bilirubin, forming bili- and increases in hypoxia, ischemia and poorly vascularized cancer.
rubin diglucuronide, and is excreted into the bile [113]. A possible These inconsistencies may be caused by metabolic patterns spe-
explanation can be found in a significant increase of b-glucuroni- cific to species and tissue and partly by the small sample sizes in
dase activity observed in PC tissue and CP patients, compared to the studies carried out.
normal subjects [114]. This may be released into the bile along Many authors have observed significantly lower levels of 3-
with pancreatic secretions and could bring about deconjugation hydroxybutyrate {except Bathe and colleagues [44]], isoleucine
of the bilirubin diglucuronide present in bile, liberating free GlcU. and creatine in serum and plasma using NMR or GC–MS analysis
The serum/plasma and tissue of PC patients has also been [44,45,50], and 3-hydroxyisovalerate in serum and urine, whereas
shown to have an altered lipid profile based on NMR [52,74] and discordant trends in the level of leucine have been observed in ur-
FI-FT-ICR-MS analysis [55], highlighting lipid compounds ine, plasma or serum from PC patients [43,44] (Table 2). Isoleucine
112 I.M. Di Gangi et al. / Trends in Analytical Chemistry 55 (2014) 94–116
and leucine are essential amino acids, and changes in the serum of tion in turn triggers hCHOP-C/EBPa complex formation, reducing
PC patients could be attributed to the impaired metabolism of hENT1 expression [142]}.
these amino acids in PC. The 3-hydroxybutyric acid level is instead hENT1 is a nucleoside transporter acting as a drug transporter,
higher in ketosis; indeed, in humans, beta-hydroxybutyrate is and it mediates the cellular uptake of nucleotides (including gem-
synthesized in the liver from acetyl-CoA and can be used as an citabine, a nucleoside analogue) in order to convert drugs into their
energy source by the brain when blood glucose is low [128]. bioactive form. It has already been demonstrated in several studies
3-Hydroxyisovalerate derives from isovaleryl-CoA, a catabolic that hENT1 down-regulation is responsible for the poor response
intermediate of leucine. Interestingly, a recent study suggested to gemcitabine therapy [143–145].
that modulation of the expression of leucine zipper tumor suppres- Decreased hENT1 levels are also due to a higher glucose concen-
sor 2 has direct effects on the proliferation of various cancer cells tration [142], which was found to be altered in PC patients’ urine
[129]. Thus, a significant decrease in the 3-hydroxyisovalerate and serum by Napoli et al. [47], Davis et al. [46] and Bathe et al.
level in the serum may suggest the development of PC. [44] (Table 2). Higher glucose concentrations have the same inhib-
Compared to controls, PC patients also showed significant in- itory effect on the complex formation of hCHOP-C/EBPa, which in
creased concentrations of serum creatinine, combined with de- turn down-regulates hENT1 expression, consequently inhibiting
creased excretion in urine [43,47] (Table 2). An increase in the chemotherapeutic effect of gemcitabine. These data suggest
creatinine within the first 48 h is strongly associated with the that not only a gene therapy could be effective in improving the
development of pancreatic necrosis [130], which, when analyzed therapeutic response to gemcitabine, as already reported [146],
histologically, can be a simple, accurate and reproducible predictor but also an amino-acid-modified diet or a diet based on caloric/glu-
of post-operative outcome in PDAC patients [131]. cose restriction could help to overcome chemotherapy resistance
by enhancing the effects of gemcitabine due to hENT1 modulation.
9. Discussion Research suggests that there are major health benefits associated
with dietary restriction.
Cancer is commonly considered a genetic disease that is driven Fig. 4 illustrates the interaction between the impaired meta-
by mutations of oncogenes and tumor-suppressor genes. However, bolic pathways mentioned in PC, chemotherapy drug resistance
one of the major underlying purposes of these genetic and gene- and downstream signaling pathways.
expression changes is to create a metabolic phenotype for cancer Recent studies in rodent and in-vitro models uncovered a poten-
cells that is essential for tumor-cell growth and survival tial link between short-term starvation and the improved efficacy of
[132,133]. The metabolic phenotype of cancer includes alterations chemotherapy, which has already been demonstrated for some
in glycolysis, amino-acid metabolism, nucleotide metabolism and types of cancer [147,148]. An interesting study performed by Safdie
glycerophospholipid metabolism. Indeed, over and beyond aerobic et al. [149] outlined the link between fasting and increased temozol-
glycolysis, also called the ‘‘Warburg effect’’ [134,135], cancer cells omide chemotherapeutic effect in a mouse model of glioma. In order
exhibit increased protein and nucleotide synthesis [136,137], in- to define a potential predictor for the efficacy of chemotherapy in
creased fatty-acid synthesis and changes in fatty-acid metabolism patients affected by different types of cancer, a new technology,
[132,138,139]. In particular, during malignant transformation in miRNA, allowing for comprehensive and rapid mRNA-expression
cancer, Hilvo and co-workers observed drastic changes in the lipid profiling, has been developed and applied in recent years [150,151].
metabolism of cells, so they suggested that these alterations are a Advances in our understanding of functional genomics in the
prominent feature of solid tumors [140]. field of cancer could be best addressed by supplementary studies
Approximately 75% of all pancreatic carcinomas occur within including measurement of mRNA, proteins and low-molecular-
the head or the neck of the pancreas, 15–20% are located in the weight metabolites over time and in varied conditions.
body of the pancreas and 5–10% in the tail [141], leading to dys- For this reason, connecting metabolomics studies to the differ-
function of this organ. As the pancreas is a dual-function gland, ent stages of the disease would help scientists to uncover new
being made up of two types of tissue (exocrine and endocrine tis- pathways and/or new therapeutic approaches.
sue) with two different activities (secreting the digestive enzymes
and hormones, respectively), pancreatic dysfunction due to carci-
nogenetic processes results in an imbalance of metabolic profiles 10. Conclusions and perspectives
(reviewed in Table 2). Metabolomic studies have the potential both
to detect a class of altered metabolites deriving from the status of Recently, MS, NMR and multivariate statistical techniques were
the disease, which can predict PC at an early stage for early diagno- incorporated into a multidisciplinary approach to profile changes
sis, and to predict the follow-up of patients during chemotherapy. in small molecules associated with the onset and the progression
It is likely that no single biomarker can selectively characterize PC, of human diseases. The aim of this research is to identify the pro-
and all metabolites with significant changes could be candidate files of metabolites that are uniquely correlated with a specific hu-
biomarkers for new PC diagnosis based on specific patterns of mul- man disease, such as cancer, in order to diagnose accurately and to
tiple metabolites. treat the disease.
An algorithm or Web-based risk calculator, which takes into ac- Efforts to discover and to distinguish specific metabolic abnor-
count different and combined altered parameters/metabolites, will malities in cancer cells continue to involve scientists around the
probably be a better tool for metabolomics studies, enabling clini- world and changes in energy metabolism are considered an ‘‘emerg-
cians to improve understanding of the state of the disease. More- ing hallmark’’ of cancer [152]. Indeed, although PC is characterized
over, identification of the small metabolites involved in PC by several oncogenes and tumor-suppressor genetic aberrations
prediction/development may open the way for economical, com- (e.g., K-Ras and p53-protein mutation), which lead to uncontrolled
mercially-available kits, with the aim of carrying out extensive proliferation and are responsible for enhanced cancer cell survival,
screening of patients at risk. Ultimately, different metabolite con- by modulating downstream signaling pathways (MAPK and PI3K-
centrations resulting from metabolomics studies may also explain mTORpathways) [153], these genetic aberrations are also closely in-
the possible pathways involved in the onset and the progression of volved in modulating the main metabolic pathways (glycolysis,
PC, suggesting specific new targeted therapies {e.g., the altered amino acid metabolism and different biosynthetic pathways).
concentration of arginine found by Bathe [44] (Table 2) represents Oncogenic K-Ras mutation plays a key role in tumor initiation in
a critical factor, by which increasing eNOS levels and NO produc- PDAC [153] and has also been shown to promote glycolysis
I.M. Di Gangi et al. / Trends in Analytical Chemistry 55 (2014) 94–116 113
[34] E. Dudley, M. Yousef, Y. Wang, W.J. Griffiths, Targeted metabolomics and hepatobiliary diseases: its utility in detection of malignancies, Jpn. J. Clin.
mass spectrometry, Adv. Protein Chem. Struct. Biol. 80 (2010) 45–83. Oncol. 27 (1997) 13–17.
[35] A. Krastanov, Metabolomics – The State of Art, Biotechnol. Biotechnol. Equip. [61] G.A.N. Gowda, Human bile as a rich source of biomarkers for
24 (2010) 1537–1543. hepatopancreatobiliary cancers, Biomark. Med. 4 (2010) 299–314.
[36] J.C. Lindon, E. Holmes, J.K. Nicholson, So what’s the deal with metabonomics?, [62] G. Zhang, P. He, H. Tan, A. Budhu, J. Gaedcke, B.M. Ghadimi, et al., Integration
Anal Chem. 75 (2003) 384A–391A. of metabolomics and transcriptomics revealed a fatty acid network exerting
[37] O. Beckonert, H.C. Keun, T.M.D. Ebbels, J. Bundy, E. Holmes, J.C. Lindon, et al., growth inhibitory effects in human PC, Clin. Cancer Res. Off. J. Am. Assoc.
Metabolic profiling, metabolomics and metabonomic procedures for NMR Cancer Res. 19 (2013) 4983–4993.
spectroscopy of urine, plasma, serum and tissue extracts, Nat. Protoc. 2 [63] D. Nagrath, C. Caneba, T. Karedath, N. Bellance, Metabolomics for
(2007) 2692–2703. mitochondrial and cancer studies, Biochim. Biophys. Acta. 2011 (1807)
[38] A.L. Castle, O. Fiehn, R. Kaddurah-Daouk, J.C. Lindon, Metabolomics Standards 650–663.
Workshop and the development of international standards for reporting [64] E. Werner, J.-F. Heilier, C. Ducruix, E. Ezan, C. Junot, J.-C. Tabet, Mass
metabolomics experimental results, Brief. Bioinform. 7 (2006) 159–165. spectrometry for the identification of the discriminating signals from
[39] E.M. Lenz, J. Bright, I.D. Wilson, S.R. Morgan, A.F.P. Nash, A 1H NMR-based metabolomics: current status and future trends, J. Chromatogr. B. 871
metabonomic study of urine and plasma samples obtained from healthy (2008) 143–163.
human subjects, J. Pharm. Biomed. Anal. 33 (2003) 1103–1115. [65] H. Bi, K.W. Krausz, S.K. Manna, F. Li, C.H. Johnson, F.J. Gonzalez, Optimization
[40] N.V. Reo, NMR-based metabolomics, Drug Chem. Toxicol. 25 (2002) 375–382. of harvesting, extraction, and analytical protocols for UPLC-ESI-MS-based
[41] T. Kobayashi, S. Nishiumi, A. Ikeda, T. Yoshie, A. Sakai, A. Matsubara, et al., A metabolomics analysis of adherent mammalian cancer cells, Anal. Bioanal.
novel serum metabolomics-based diagnostic approach to PC, Cancer Chem. 405 (2013) 5279–5289.
Epidemiol. Biomark. Prev. Publ. Am. Assoc. Cancer Res. Cosponsored Am. [66] B.K. Matuszewski, M.L. Constanzer, C.M. Chavez-Eng, Strategies for the
Soc. Prev Oncol. 22 (2013) 571–579. assessment of matrix effect in quantitative bioanalytical methods based on
[42] A.B. Leichtle, U. Ceglarek, P. Weinert, C.T. Nakas, J.-M. Nuoffer, J. Kase, et al., HPLCMS/MS, Anal. Chem. 75 (2003) 3019–3030.
Pancreatic carcinoma, pancreatitis, and healthy controls: metabolite models [67] J. Budczies, C. Denkert, B.M. Müller, S.F. Brockmöller, F. Klauschen, B. Györffy,
in a three-class diagnostic dilemma, Metabolomics Off. J. Metabolomic Soc. 9 et al., Remodeling of central metabolism in invasive breast cancer compared
(2013) 677–687. to normal breast tissue – a GC-TOFMS based metabolomics study, BMC
[43] D. OuYang, J. Xu, H. Huang, Z. Chen, Metabolomic profiling of serum from Genomics 13 (2012) 334.
human PC patients using 1H NMR spectroscopy and principal component [68] J.C. Lindon, J.K. Nicholson, E. Holmes, H. Antti, M.E. Bollard, H. Keun, et al.,
analysis, Appl. Biochem. Biotechnol. 165 (2011) 148–154. Contemporary issues in toxicology the role of metabonomics in toxicology and its
[44] O.F. Bathe, R. Shaykhutdinov, K. Kopciuk, A.M. Weljie, A. McKay, F.R. evaluation by the COMET project, Toxicol. Appl. Pharmacol. 187 (2003) 137–146.
Sutherland, et al., Feasibility of identifying PC based on serum [69] K.A. Vermeersch, M.P. Styczynski, Applications of metabolomics in cancer
metabolomics, Cancer Epidemiol. Biomark. Prev. Publ. Am. Assoc. Cancer research, J. Carcinog. 12 (2013) 9.
Res. Cosponsored Am. Soc. Prev Oncol. 20 (2011) 140–147. [70] J.T. Brindle, H. Antti, E. Holmes, G. Tranter, J.K. Nicholson, H.W.L. Bethell, et al.,
[45] S. Nishiumi, M. Shinohara, A. Ikeda, T. Yoshie, N. Hatano, S. Kakuyama, S. Rapid and noninvasive diagnosis of the presence and severity of coronary
Mizuno, T. Sanuki, H. Kutsumi, E. Fukusaki, T. Azuma, T. Takenawa, M. heart disease using 1H NMR-based metabonomics, Nat. Med. 8 (2002) 1439–
Yoshida, Serum metabolomics as a novel diagnostic approach for PC, 1444.
Metabolomics 6 (2010) 518–528. [71] K. Odunsi, R.M. Wollman, C.B. Ambrosone, A. Hutson, S.E. McCann, J.
[46] V.W. Davis, D.E. Schiller, D. Eurich, O.F. Bathe, M.B. Sawyer, Pancreatic ductal Tammela, et al., Detection of epithelial ovarian cancer using 1H NMR-based
adenocarcinoma is associated with a distinct urinary metabolomic signature, metabonomics, Int. J. Cancer J. Int. Cancer. 113 (2005) 782–788.
Ann. Surg. Oncol. (2012). [72] J.C. Lindon, E. Holmes, J.K. Nicholson, Metabonomics and its role in drug
[47] C. Napoli, N. Sperandio, R.T. Lawlor, A. Scarpa, H. Molinari, M. Assfalg, Urine development and disease diagnosis, Expert Rev. Mol. Diagn. 4 (2004) 189–199.
metabolic signature of pancreatic ductal adenocarcinoma by (1)h nuclear [73] U.K. Bhanot, P. Möller, Mechanisms of parenchymal injury and signaling
magnetic resonance: identification, mapping, and evolution, J. Proteome Res. pathways in ectatic ducts of chronic pancreatitis: implications for pancreatic
11 (2012) 1274–1283. carcinogenesis, Lab. Investig. J. Tech. Methods Pathol. 89 (2009) 489–497.
[48] S. Urayama, W. Zou, K. Brooks, V. Tolstikov, Comprehensive mass [74] Richard.D. Beger, Laura.K. Schnackenberg, Ricky.D. Holland, Donghui. Li,
spectrometry based metabolic profiling of blood plasma reveals potent Yvonne. Dragan, Metabonomic models of human PC using 1D proton NMR
discriminatory classifiers of PC, Rapid Commun. Mass Spectrom. RCM 24 spectra of lipids in plasma, Metabolomics. 2 (2006) 125–134.
(2010) 613–620. [75] T. Ikegami, H. Kobayashi, H. Kimura, V. Tolstikov, O. Fiehn, N. Tanaka, High-
[49] H. Zhang, Y. Wang, X. Gu, J. Zhou, C. Yan, Metabolomic profiling of human performance liquid chromatography for metabolomics: high-efficiency
plasma in PC using pressurized capillary electrochromatography, separations utilizing mono- lithic silica columns., in: Metabolomics Front.
Electrophoresis 32 (2011) 340–347. Syst. Biol., Springer-Verlag, Tokyo, 2005: pp. 107–126.
[50] L. Zhang, H. Jin, X. Guo, Z. Yang, L. Zhao, S. Tang, et al., Distinguishing PC from [76] B.L. Proctor, Recent developments in mass spectrometry for the analysis of
chronic pancreatitis and healthy individuals by (1)H nuclear magnetic complex mixtures, Annu. Rev. Pharmacol. Toxicol. 23 (1983) 171–192.
resonance-based metabonomic profiles, Clin. Biochem. 45 (2012) 1064–1069. [77] K. Dettmer, N. Nürnberger, H. Kaspar, M.A. Gruber, M.F. Almstetter, P.J.
[51] N.J. Zyromski, A. Mathur, G.A.N. Gowda, C. Murphy, D.A. Swartz-Basile, T.E. Oefner, Metabolite extraction from adherently growing mammalian cells for
Wade, et al., Nuclear magnetic resonance spectroscopy-based metabolomics metabolomics studies: optimization of harvesting and extraction protocols,
of the fatty pancreas: implicating fat in pancreatic pathology, Pancreatol. Off. Anal. Bioanal. Chem. 399 (2011) 1127–1139.
J. Int. Assoc. Pancreatol. IAP Al 9 (2009) 410–419. [78] A. Zhang, H. Sun, S. Qiu, X. Wang, NMR-based metabolomics coupled with
[52] F. Fang, X. He, H. Deng, Q. Chen, J. Lu, M. Spraul, et al., Discrimination of pattern recognition methods in biomarker discovery and disease diagnosis,
metabolic profiles of PC from chronic pancreatitis by high-resolution magic Magn. Reson. Chem. MRC 51 (2013) 549–556.
angle spinning 1H nuclear magnetic resonance and principal components [79] H.G. Wahl, A. Hoffmann, D. Luft, H.M. Liebich, Analysis of volatile organic
analysis, Cancer Sci. 98 (2007) 1678–1682. compounds in human urine by headspace gas chromatography-mass
[53] T. Bezabeh, O.B. Ijare, N. Albiin, U. Arnelo, B. Lindberg, I.C.P. Smith, Detection spectrometry with a multipurpose sampler, J. Chromatogr. A. 847 (1999)
and quantification of D-glucuronic acid in human bile using 1H NMR 117–125.
spectroscopy: relevance to the diagnosis of PC, Magma N Y N 22 (2009) [80] C.A. Sellick, R. Hansen, G.M. Stephens, R. Goodacre, A.J. Dickson, Metabolite
267–275. extraction from suspension-cultured mammalian cells for global metabolite
[54] D.S. Longnecker, N. Chandar, D.G. Sheahan, J.E. Janosky, B. Lombardi, profiling, Nat. Protoc. 6 (2011) 1241–1249.
Preneoplastic and neoplastic lesions in the pancreas of rats fed choline- [81] M. Katajamaa, M. Oresic, Data processing for mass spectrometry-based
devoid or choline-supplemented diets, Toxicol. Pathol. 19 (1991) 59–65. metabolomics, J. Chromatogr. A. 1158 (2007) 318–328.
[55] S.A. Ritchie, H. Akita, I. Takemasa, H. Eguchi, E. Pastural, H. Nagano, et al., [82] S. Bijlsma, I. Bobeldijk, E.R. Verheij, R. Ramaker, S. Kochhar, I.A. Macdonald,
Metabolic system alterations in PC patient serum: potential for early et al., Large-scale human metabolomics studies: a strategy for data (pre-)
detection, BMC Cancer 13 (2013) 416. processing and validation, Anal. Chem. 78 (2006) 567–574.
[56] M. Sugimoto, D.T. Wong, A. Hirayama, T. Soga, M. Tomita, Capillary [83] J. Trygg, E. Holmes, T. Lundstedt, Chemometrics in metabonomics, J. Proteome
electrophoresis mass spectrometry-based saliva metabolomics identified Res. 6 (2007) 469–479.
oral, breast and PC-specific profiles, Metabolomics Off. J. Metabolomic Soc. [84] N. Takahashi, Three-dimensional mapping of N-linked oligosaccharides using
6 (2010) 78–95. anion-exchange, hydrophobic and hydrophilic interaction modes of high-
[57] P. Kaur, K. Sheikh, A. Kirilyuk, K. Kirilyuk, R. Singh, H.W. Ressom, A.K. Cheema, performance liquid chromatography, J. Chromatogr. A. 720 (1996) 217–225.
Metabolomic profiling for biomarker discovery in PC, Intern. J. Mass Spec. 310 [85] G. Ge, G.W. Wong, Classification of premalignant PC mass-spectrometry data
(2012) 44–51. using decision tree ensembles, BMC Bioinformatics. 9 (2008) 275.
[58] P.D.V. de Almeida, A.M.T. Grégio, M.A.N. Machado, A.A.S. de Lima, L.R. [86] A.B. Leichtle, J.-F. Dufour, G.M. Fiedler, Potentials and pitfalls of clinical
Azevedo, Saliva composition and functions: a comprehensive review, J. peptidomics and metabolomics, Swiss Med. Wkly. 143 (2013) w13801.
Contemp. Dent. Pract. 9 (2008) 72–80. [87] C.A. Smith, G. O’Maille, E.J. Want, C. Qin, S.A. Trauger, T.R. Brandon, et al.,
[59] H. Wen, S.S. Yoo, J. Kang, H.G. Kim, J.-S. Park, S. Jeong, et al., A new NMR-based METLIN: a metabolite mass spectral database, Ther. Drug Monit. 27 (2005)
metabolomics approach for the diagnosis of biliary tract cancer, J. Hepatol. 52 747–751.
(2010) 228–233. [88] M. Kanehisa, S. Goto, M. Hattori, K.F. Aoki-Kinoshita, M. Itoh, S. Kawashima,
[60] T. Nishijima, M. Nishina, K. Fujiwara, Measurement of lactate levels in serum et al., From genomics to chemical genomics: new developments in KEGG,
and bile using proton nuclear magnetic resonance in patients with Nucleic Acids Res. 34 (2006) D354–357.
I.M. Di Gangi et al. / Trends in Analytical Chemistry 55 (2014) 94–116 115
[89] P. Romero, J. Wagg, M.L. Green, D. Kaiser, M. Krummenacker, P.D. Karp, [119] P.W. Szlosarek, F.R. Balkwill, Tumour necrosis factor alpha: a potential target
Computational prediction of human metabolic pathways from the complete for the therapy of solid tumors, Lancet Oncol. 4 (2003) 565–573.
human genome, Genome Biol. 6 (2005) R2. [120] P. Ariapart, S. Bergstedt-Lindqvist, V. van Harmelen, J. Permert, F. Wang, I.
[90] A. Frolkis, C. Knox, E. Lim, T. Jewison, V. Law, D.D. Hau, et al., SMPDB: the Lundkvist, Resection of PC normalizes the preoperative increase of tumor
small molecule pathway database, Nucleic Acids Res. 38 (2010) D480–487. necrosis factor alpha gene expression, Pancreatol. Off. J. Int. Assoc. Pancreatol.
[91] B.H. Junker, C. Klukas, F. Schreiber, VANTED: a system for advanced data IAP Al 2 (2002) 491–494.
analysis and visualization in the context of biological networks, BMC [121] K.S. Pedersen, W.R. Bamlet, A.L. Oberg, M. de Andrade, M.E. Matsumoto, H.
Bioinform. 7 (2006) 109. Tang, et al., Leukocyte DNA methylation signature differentiates PC patients
[92] R.W. Byrnes, D. Cotter, A. Maer, J. Li, D. Nadeau, S. Subramaniam, An editor for from healthy controls, PloS One 6 (2011) e18223.
pathway drawing and data visualization in the Biopathways Workbench, [122] M.J. Baine, M. Menning, L.M. Smith, K. Mallya, S. Kaur, S. Rachagani, et al.,
BMC Syst. Biol. 3 (2009) 99. Differential gene expression analysis of peripheral blood mononuclear cells
[93] A.H. Merril Jr., SphinGOMAP – a web-based biosynthetic pathway map of reveals novel test for early detection of PC, Cancer Biomark. Sect. Dis. Markers
sphingolipids and glycosphingolipids, Glycobiology 15 (2005) 15G. 11 (2011) 1–14.
[94] E. Yasugi, K. Watanabe, LIPIDBANK for Web, the newly developed lipid [123] J. Ren, S.M. Grundy, J. Liu, W. Wang, M. Wang, J. Sun, et al., Long-term
database, Tanpakushitsu Kakusan Koso 47 (2002) 837–841. coronary heart disease risk associated with very-low-density lipoprotein
[95] P. Shannon, A. Markiel, O. Ozier, N.S. Baliga, J.T. Wang, D. Ramage, et al., cholesterol in Chinese: the results of a 15-Year Chinese Multi-Provincial
Cytoscape: a software environment for integrated models of biomolecular Cohort Study (CMCS), Atherosclerosis 211 (2010) 327–332.
interaction networks, Genome Res. 13 (2003) 2498–2504. [124] A.Q. Zhang, S.C. Mitchell, R.L. Smith, Dimethylamine formation in the rat from
[96] J. Gao, V.G. Tarcea, A. Karnovsky, B.R. Mirel, T.E. Weymouth, C.W. Beecher, various related amine precursors, Food Chem. Toxicol. Int. J. Publ. Br. Ind.
et al., Metscape: a Cytoscape plug-in for visualizing and interpreting Biol. Res. Assoc. 36 (1998) 923–927.
metabolomics data in the context of human metabolic networks, [125] D.M. Brizel, T. Schroeder, R.L. Scher, S. Walenta, R.W. Clough, M.W. Dewhirst,
Bioinforma. Oxf. Engl. 26 (2010) 971–973. et al., Elevated tumor lactate concentrations predict for an increased risk of
[97] A. Karnovsky, T. Weymouth, T. Hull, V.G. Tarcea, G. Scardoni, C. Laudanna, metastases in head-and-neck cancer, Int. J. Radiat. Oncol. Biol. Phys. 51
et al., Metscape 2 bioinformatics tool for the analysis and visualization of (2001) 349–353.
metabolomics and gene expression data, Bioinforma. Oxf. Engl. 28 (2012) [126] G. Schwickert, S. Walenta, K. Sundfør, E.K. Rofstad, W. Mueller-Klieser,
373–380. Correlation of high lactate levels in human cervical cancer with incidence of
[98] H. Ma, A. Sorokin, A. Mazein, A. Selkov, E. Selkov, O. Demin, et al., The metastasis, Cancer Res. 55 (1995) 4757–4759.
Edinburgh human metabolic network reconstruction and its functional [127] S. Walenta, T.-V. Chau, T. Schroeder, H.-A. Lehr, L.A. Kunz-Schughart, A.
analysis, Mol. Syst. Biol. 3 (2007) 135. Fuerst, et al., Metabolic classification of human rectal adenocarcinomas: a
[99] M. Baker, Metabolomics: from small molecules to big ideas, Nat. Meth. 8 novel guideline for clinical oncologists?, J Cancer Res. Clin. Oncol. 129 (2003)
(2011) 117–121. 321–326.
[100] T. Kind, M. Scholz, O. Fiehn, How large is the metabolome? A critical analysis [128] E.T. Baran, N. Ozer, V. Hasirci, Poly(hydroxybutyrate-co-hydroxyvalerate)
of data exchange practices in chemistry, PloS One 4 (2009) e5440. nanocapsules as enzyme carriers for cancer therapy: an in vitro study, J.
[101] K.C. Fearon, M.D. Barber, J.S. Falconer, D.C. McMillan, J.A. Ross, T. Preston, PC Microencapsul. 19 (2002) 363–376.
as a model: inflammatory mediators, acute-phase response, and cancer [129] J.M. Kim, J.S. Song, H.H. Cho, K.K. Shin, Y.C. Bae, B.J. Lee, et al., Effect of the
cachexia, World J. Surg. 23 (1999) 584–588. modulation of leucine zipper tumor suppressor 2 expression on proliferation
[102] H. Schrader, B.A. Menge, O. Belyaev, W. Uhl, W.E. Schmidt, J.J. Meier, Amino of various cancer cells functions as a tumor suppressor, Mol. Cell. Biochem.
acid malnutrition in patients with chronic pancreatitis and pancreatic 346 (2011) 125–136.
carcinoma, Pancreas. 38 (2009) 416–421. [130] V. Muddana, D.C. Whitcomb, A. Khalid, A. Slivka, G.I. Papachristou, Elevated
[103] T. Yamanouchi, Y. Akanuma, Serum 1,5-anhydroglucitol (1,5 AG): new serum creatinine as a marker of pancreatic necrosis in acute pancreatitis, Am.
clinical marker for glycemic control, Diabetes Res. Clin. Pract. 24 (Suppl) J. Gastroenterol. 104 (2009) 164–170.
(1994) S261–268. [131] N. Hiraoka, Y. Ino, S. Sekine, H. Tsuda, K. Shimada, T. Kosuge, et al., Tumour
[104] J.E. Domínguez-Muñoz, Pancreatic exocrine insufficiency: diagnosis and necrosis is a postoperative prognostic marker for PC patients with a high
treatment, J. Gastroenterol. Hepatol. 26 (Suppl 2) (2011) 12–16. interobserver reproducibility in histological evaluation, Br. J. Cancer. 103
[105] P. Findeisen, M. Neumaier, Functional protease profiling for diagnosis of (2010) 1057–1065.
malignant disease, Proteomics Clin. Appl. 6 (2012) 60–78. [132] K. Glunde, C. Jie, Z.M. Bhujwalla, Molecular causes of the aberrant choline
[106] K. Garber, Energy boost: the Warburg effect returns in a new theory of cancer, phospholipid metabolism in breast cancer, Cancer Res. 64 (2004) 4270–4276.
J. Natl. Cancer Inst. 96 (2004) 1805–1806. [133] R.A. Gatenby, R.J. Gillies, Why do cancers have high aerobic glycolysis?, Nat
[107] H.J. Atherton, N.J. Bailey, W. Zhang, J. Taylor, H. Major, J. Shockcor, et al., A Rev. Cancer. 4 (2004) 891–899.
combined 1H NMR spectroscopy- and mass spectrometry-based [134] D.A. Tennant, R.V. Durán, H. Boulahbel, E. Gottlieb, Metabolic transformation
metabolomics study of the PPAR-alpha null mutant mouse defines in cancer, Carcinogenesis 30 (2009) 1269–1280.
profound systemic changes in metabolism linked to the metabolic [135] M.G. Vander Heiden, L.C. Cantley, C.B. Thompson, Understanding the
syndrome, Physiol. Genomics. 27 (2006) 178–186. Warburg effect: the metabolic requirements of cell proliferation, Science
[108] E.C.Y. Chan, P.K. Koh, M. Mal, P.Y. Cheah, K.W. Eu, A. Backshall, et al., 324 (2009) 1029–1033.
Metabolic profiling of human colorectal cancer using high-resolution magic [136] L. Rahman, D. Voeller, M. Rahman, S. Lipkowitz, C. Allegra, J.C. Barrett, et al.,
angle spinning nuclear magnetic resonance (HR-MAS NMR) spectroscopy and Thymidylate synthase as an oncogene: a novel role for an essential DNA
gas chromatography mass spectrometry (GC/MS), J. Proteome Res. 8 (2009) synthesis enzyme, Cancer Cell 5 (2004) 341–351.
352–361. [137] J.L. Griffin, J.P. Shockcor, Metabolic profiles of cancer cells, Nat. Rev. Cancer 4
[109] C. Denkert, J. Budczies, W. Weichert, G. Wohlgemuth, M. Scholz, T. Kind, et al., (2004) 551–561.
Metabolite profiling of human colon carcinoma–deregulation of TCA cycle [138] F.P. Kuhajda, Fatty acid synthase and cancer: new application of an old
and amino acid turnover, Mol. Cancer. 7 (2008) 72. pathway, Cancer Res. 66 (2006) 5977–5980.
[110] J. Wang, C. Ma, Z. Liao, B. Tian, J.-P. Lu, Study on chronic pancreatitis and PC [139] J.A. Menendez, R. Lupu, Fatty acid synthase and the lipogenic phenotype in
using MRS and pancreatic juice samples, World J. Gastroenterol. WJG. 17 cancer pathogenesis, Nat. Rev. Cancer 7 (2007) 763–777.
(2011) 2126–2130. [140] M. Hilvo, M. Orěsîc, Regulation of lipid metabolism in breast cancer provides
[111] S. Srivastava, R. Roy, S. Singh, P. Kumar, D. Dalela, S.N. Sankhwar, et al., diagnostic and therapeutic opportunities, Clin. Lipidol. 7 (2012) 177–188.
Taurine – a possible fingerprint biomarker in non-muscle invasive bladder [141] M.W. Saif, Pancreatic neoplasm in 2011, An update, JOP J. Pancreas 12 (2011)
cancer: A pilot study by 1H NMR spectroscopy, Cancer Biomark. Sect. Dis. 316–321.
Markers 6 (2010) 11–20. [142] A. Pandolfi, N. Di Pietro, High glucose, nitric oxide, and adenosine: a vicious
[112] D.L. Nelson, M.M. Cox, Carbohydrate biosynthesis., in: Lehninger Princ. circle in chronic hyperglycaemia?, Cardiovasc Res. 86 (2010) 9–11.
Biochem., 3rd ed., Worth, New York, 2000: pp. 743–744. [143] J.J. Farrell, H. Elsaleh, M. Garcia, R. Lai, A. Ammar, W.F. Regine, et al., Human
[113] E.R. Gordon, C.A. Goresky, T.H. Chang, A.S. Perlin, The isolation and equilibrative nucleoside transporter 1 levels predict response to gemcitabine
characterization of bilirubin diglucuronide, the major bilirubin conjugate in in patients with PC, Gastroenterology 136 (2009) 187–195.
dog and human bile, Biochem. J. 155 (1976) 477–486. [144] I. Borbath, L. Verbrugghe, R. Lai, J.F. Gigot, Y. Humblet, H. Piessevaux, et al.,
[114] B. Sperker, U. Werner, T.E. Mürdter, C. Tekkaya, P. Fritz, R. Wacke, et al., Human equilibrative nucleoside transporter 1 (hENT1) expression is a
Expression and function of beta-glucuronidase in PC: potential role in drug potential predictive tool for response to gemcitabine in patients with
targeting, Naunyn. Schmiedebergs Arch. Pharmacol. 362 (2000) 110–115. advanced cholangiocarcinoma, Eur. J. Cancer Oxf. Engl. 1990 (48) (2012)
[115] P. Clerc, N. Bensaadi, P. Pradel, A. Estival, F. Clemente, N. Vaysse, Lipid- 990–996.
dependent proliferation of PC cell lines, Cancer Res. 51 (1991) 3633–3638. [145] D. Santini, G. Schiavon, B. Vincenzi, C.E. Cass, E. Vasile, A.D. Manazza, et al.,
[116] C.A. Dinarello, The paradox of pro-inflammatory cytokines in cancer, Cancer Human equilibrative nucleoside transporter 1 (hENT1) levels predict
Metastasis Rev. 25 (2006) 307–313. response to gemcitabine in patients with biliary tract cancer (BTC), Curr.
[117] M. Azrad, C. Turgeon, W. Demark-Wahnefried, Current evidence linking Cancer Drug Targets. 11 (2011) 123–129.
polyunsaturated fatty acids with cancer risk and progression, Front. Oncol. 3 [146] S. Pérez-Torras, J. García-Manteiga, E. Mercadé, F.J. Casado, N. Carbó, M.
(2013) 224. Pastor-Anglada, et al., Adenoviral-mediated overexpression of human
[118] H. Algül, G. Adler, R.M. Schmid, NF-kappaB/Rel transcriptional pathway: equilibrative nucleoside transporter 1 (hENT1) enhances gemcitabine
implications in PC, Int. J. Gastrointest. Cancer 31 (2002) 71–78. response in human PC, Biochem. Pharmacol. 76 (2008) 322–329.
116 I.M. Di Gangi et al. / Trends in Analytical Chemistry 55 (2014) 94–116
[147] C. Lee, L. Raffaghello, S. Brandhorst, F.M. Safdie, G. Bianchi, A. Martin- [153] S. Gysin, M. Salt, A. Young, F. McCormick, Therapeutic strategies for targeting
Montalvo, et al., Fasting cycles retard growth of tumors and sensitize a range ras proteins, Genes Cancer 2 (2011) 359–372.
of cancer cell types to chemotherapy, Sci. Transl. Med. 4 (2012) 124ra27. [154] E. Racker, R.J. Resnick, R. Feldman, Glycolysis and methylaminoisobutyrate
[148] L. Raffaghello, C. Lee, F.M. Safdie, M. Wei, F. Madia, G. Bianchi, et al., uptake in rat-1 cells transfected with ras or myc oncogenes, Proc. Natl. Acad.
Starvation-dependent differential stress resistance protects normal but not Sci. USA 82 (1985) 3535–3538.
cancer cells against high-dose chemotherapy, Proc. Natl. Acad. Sci. USA 105 [155] J. Yun, C. Rago, I. Cheong, R. Pagliarini, P. Angenendt, H. Rajagopalan, et al.,
(2008) 8215–8220. Glucose deprivation contributes to the development of KRAS pathway
[149] F. Safdie, S. Brandhorst, M. Wei, W. Wang, C. Lee, S. Hwang, et al., Fasting mutations in tumor cells, Science 325 (2009) 1555–1559.
enhances the response of glioma to chemo- and radiotherapy, PloS One 7 [156] Sophie Vasseur, Richard Tomasini, Roselyne Tournaire, Juan.L. Iovanna,
(2012) e44603. Cancers, Hypoxia Induc. Tumor Metab. Switch Contrib. Pancreat Cancer
[150] S. Marchini, R. Fruscio, L. Clivio, L. Beltrame, L. Porcu, I.F. Nerini, et al., Aggress 2 (2010) 2138–2152.
Resistance to platinum-based chemotherapy is associated with epithelial to [157] J.P. Morton, P. Timpson, S.A. Karim, R.A. Ridgway, D. Athineos, B. Doyle, et al.,
mesenchymal transition in epithelial ovarian cancer, Eur. J. Cancer Oxf. Engl. Mutant p53 drives metastasis and overcomes growth arrest/senescence in
1990 (49) (2013) 520–530. PC, Proc. Natl. Acad. Sci. USA 107 (2010) 246–251.
[151] H.-L. Zhang, L.-F. Yang, Y. Zhu, X.-D. Yao, S.-L. Zhang, B. Dai, et al., Serum [158] C. Muñoz-Pinedo, N. El Mjiyad, J.-E. Ricci, Cancer metabolism: current
miRNA-21: elevated levels in patients with metastatic hormone-refractory perspectives and future directions, Cell Death Dis. 3 (2012) e248.
prostate cancer and potential predictive factor for the efficacy of docetaxel- [159] O.D.K. Maddocks, C.R. Berkers, S.M. Mason, L. Zheng, K. Blyth, E. Gottlieb,
based chemotherapy, The Prostate 71 (2011) 326–331. et al., Serine starvation induces stress and p53-dependent metabolic
[152] D. Hanahan, R.A. Weinberg, Hallmarks of cancer: the next generation, Cell remodelling in cancer cells, Nature. advance online publication (2012).
144 (2011) 646–674.