scispace - formally typeset

Journal ArticleDOI

1H NMR global metabolic phenotyping of acute pancreatitis in the emergency unit.

08 Oct 2014-Journal of Proteome Research (American Chemical Society)-Vol. 13, Iss: 12, pp 5362-5375

TL;DR: It is demonstrated that combinatorial biomarkers have a strong diagnostic and prognostic potential in AP with relevance to clinical decision making in the emergency unit.

AbstractWe have investigated the urinary and plasma metabolic phenotype of acute pancreatitis (AP) patients presenting to the emergency room at a single center London teaching hospital with acute abdominal pain using 1H NMR spectroscopy and multivariate modeling. Patients were allocated to either the AP (n = 15) or non-AP patients group (all other causes of abdominal pain, n = 21) on the basis of the national guidelines. Patients were assessed for three clinical outcomes: (1) diagnosis of AP, (2) etiology of AP caused by alcohol consumption and cholelithiasis, and (3) AP severity based on the Glasgow score. Samples from AP patients were characterized by high levels of urinary ketone bodies, glucose, plasma choline and lipid, and relatively low levels of urinary hippurate, creatine and plasma-branched chain amino acids. AP could be reliably identified with a high degree of sensitivity and specificity (OPLS-DA model R2 = 0.76 and Q2Y = 0.59) using panel of discriminatory biomarkers consisting of guanine, hippurate ...

Summary (3 min read)

Introduction

  • Acute pancreatitis (AP) is an inflammatory condition associated with a progressive systemic inflammatory response (SIRS) and, in severe cases, autonecrosis of pancreatic tissue, organ failure, and death, also known as KEYWORDS.
  • This lacks sensitivity and specificity, and it is subject to variation in its diagnostic threshold over time after the initial pancreatic insult.
  • 6−10 Other generic physiological scoring systems are used in critical care such as the Acute Physiology and Chronic Health Evaluation (APACHE II) score,11 a generic physiological measurement based on 12 parameters, that is designed to measure the severity of disease for adult patients 24 hrs following admission to intensive care units.
  • 14−16 The determination of metabolite changes that describe a biological phenotype based on either 1H nuclear magnetic resonance (NMR) spectroscopy or mass spectrometry (MS) has been widely applied for global profiling to define diagnostic or prognostic biofluid profiles for physiological or pathological states.
  • A key feature of AP is abdominal pain, a common and often nonspecific presenting complaint for a large array of other surgical pathologies that require urgent treatment.

Study Design

  • This was an observational control study of consecutive patients presenting with acute abdominal pain to a single center “Accident and Emergency” unit at St. Mary’s hospital, London, U.K.
  • Patients were excluded if they had been discharged from hospital within 72 h with the same pain, if they were unconscious at presentation, if they were pregnant, if they were under the age of 18, or if they had undergone surgery within the previous 6 weeks or received a blood transfusion.
  • Patients with congenital pancreatic malformations or those with cystic fibrosis were also excluded.
  • Patients were investigated and managed as per the current British Society of Gastroenterology guidelines.
  • Samples were collected from each patient for the first 5 days of their admission.

Sample Collection and Preparation

  • If catheterized, a “clean catch” sample was taken.
  • Urine samples were kept on ice after collection prior to being stored frozen at −80 °C.
  • The mixture was transferred into a 5 mm outer diameter NMR tube.

D2O.

  • 1H NMR spectra of the plasma samples were acquired employing two 1-D NMR experiments.
  • All method characteristics and details are detailed elsewhere.
  • (OPLS-DA)27 was also carried out to optimize recovery of potential biomarkers, that is, compounds with high correlation and covariance with the class (e.g., disease severity or etiology).
  • Modeling was conducted in SIMCA-P v.12.1 using unit variance scaling.
  • Metabolite identification was based on chemical shifts published in literature and statistical total correlation spectroscopy of peaks29 (e.g., of urinary 3-hydroxyisovalerate is shown in Supplemental Figure 1, SI).

Clinical Data

  • Data were analyzed by unpaired two-tailed Student’s t test.
  • The etiologies of pain in the non-AP patients group were heterogeneous, reflecting the typical distribution of the causes of acute abdominal pain presenting to a typical accident and emergency department32 (Table 1).
  • One of the cholelithiasis AP cohorts had necrotic pancreatitis.
  • One patient with cholelithiasis AP and one patient with alcoholic AP diagnosis had a myocardial infarction during their admission, both of whom were treated with a standard acute coronary syndrome protocol.
  • The APACHE II score11 varied significantly between the groups, indicating the increased general severity of illness in the AP group compared with the non-AP group.

Characterizing the Urinary Phenotype of AP

  • The urinary 1H NMR spectra from this clinical study demonstrated substantial interindividual variation.
  • Figure 1, shows urinary 1H NMR spectra from an alcoholic AP patient (A) and a non-AP patient (B).
  • Unsurprisingly, large concentrations of acetate, ethanol, acetone, and ethyl glucuronide were also observed in this patient, reflecting the underlying etiology .
  • NMR spectral profiles of acetaminophen and ibuprofen and their urinary metabolites have been published previously.
  • The drug-related origin of the mannitol was supported by a Pearson correlation test35 using all compounds in the profile and a selected driver peak for the mannitol signal (δ 3.805 ppm, (5-CH)), with an established cutoff of p < 0.05.

Plasma NMR Spectra

  • There were few obvious disease-specific metabolites that varied systematically between raw spectra; however, clear quantitative changes in the spectra were observed, particularly in concentrations of VLDL, LDL, and nonesterified fatty acids.
  • Likewise, a comparison of CPMG raw spectra from different patients, an AP and non-AP patient diagnosed with diverticulitis, is presented in Figure 4 in the SI, where differences of VLDL were observed.

Multivariate Modeling of Patients with Severe Abdominal Pain

  • Three main analyses were performed based on clinical criteria: (1) diagnosis of AP (AP vs non-AP), (2) etiology (alcohol consumption vs cholelithiasis), and (3) AP severity, based on a validated modified Glasgow score comparison of mild (Glasgow score 1) versus moderate to severe (Glasgow score 2 and 3).
  • PLS-DA plots for each clinical scenario in urine and plasma (CPMGand 1-D-spectral) data showed positive predictive models for all with the exception of prediction of etiology in the plasma (1-D-plasma and CPMG) data set.
  • Discriminant metabolites found in plasma and urine for the three clinical scenarios are summarized in Tables 2−4 and discussed below.
  • Because hippurate is a wellestablished gut microbial cometabolite derived from glycine conjugation of benzoic acid,37,38 a subgroup analysis of antibiotic-treated patients (n = 10; 5non-AP/5AP) and nonantibiotic-treated (n = 21; 12non-AP/9AP) patients was performed to determine if antibiotic ingestion in the AP diagnostic model was responsible for the difference in urinary excretion of hippurate between the two groups .
  • Moreover, the cross-validated OPLS-DA was also more robust .

Diagnosis of AP

  • Multivariate analysis of both urinary and plasma 1H NMR spectra was able to accurately stratify patients presenting with acute abdominal pain in a clinical setting into those with AP and those with other heterogeneous causes of abdominal pain ).
  • Urine provided a stronger predictor of AP diagnosis than plasma .
  • Thus, it is very likely that the striking rise in plasma VLDLs seen here relates to the generalized severity of the systemic response rather than AP-specific changes.

Etiology of AP Caused by Alcohol Consumption and Cholelithiasis

  • The two causes of AP in this series were alcohol (9/15) and cholelithiasis, which is representative of the epidemiology of the condition.
  • Metabolites such as acetone, acetoacetate, and 3- hydroxyisovalerate (Table 3) were predictive for alcoholinduced AP.
  • It is well established that alcohol consumption influences the regulation of key pathways such as gluconeogenesis.
  • Ketone bodies are products of the oxidative pathway of alcohol metabolism53 and are thus concordant with this observation discriminating the alcohol-induced AP patients, although because there were no dietary records for the patients, the possibility that anorexia, associated with severe abdominal pain, may have been systematically different between the groups and therefore a contributor to the ketone body profile.

AP Severity Based on the Glasgow Severity Score

  • The raw NMR spectra from severely unwell AP patients were highly variable in structure compared with control patients .
  • These data provide a strong metabolic phenotype of the severity of illness and a clear metabolic description of AP disease severity.
  • AP patients with moderate to severe AP disease (Glasgow 2−3) had higher levels of glucose and reduced plasma relative concentrations of alanine, valine, choline, the acetyl signal from α1-acid dx.doi.org/10.1021/pr500161w | J. Proteome Res. XXXX, XXX, XXX−XXXI glycoprotein (NAC1), and urinary creatinine (Table 4).
  • As previously described, this combined cohort of metabolic changes is suggestive of insulin resistance54 and probable failure of pancreatic function, with the NAC1 most likely reflecting generalized inflammation.

Differentiation of Metabolic Phenotypes within the Non-AP Group

  • It was possible to distinguish between subclasses of the non-AP group using the metabolic profiles, which is of importance to the clinical setting.
  • High levels of indoleacetate have typically been related to gastrointestinal cancer and hepato-biliary tract cancer, and high concentrations have been observed in patients with cirrhosis, diabetes, and cholelithiasis occasionally.
  • Nevertheless, the authors have demonstrated a convincing metabolic phenotype for the diagnosis, etiology, and severity of AP.
  • Scores plot of the first versus second component of the principal component analysis using the most discriminant metabolites from urine and CPMG.
  • This material is available free of charge via the Internet at http://pubs.acs.org.

Did you find this useful? Give us your feedback

...read more

Content maybe subject to copyright    Report

1
H NMR Global Metabolic Phenotyping of Acute Pancreatitis in the
Emergency Unit
Alma Villasen
or,
,
James M. Kinross,*
,§,
Jia V. Li,
Nicholas Penney,
§
Richard H. Barton,
Jeremy K. Nicholson,
Ara Darzi,
§
Coral Barbas,
and Elaine Holmes*
,
Section of Biomolecular Medicine, Division of Computational and Systems Medicine, Department of Surgery and Cancer, Faculty of
Medicine, Imperial College London, Sir Alexander Fleming Building, Exhibition Road, South Kensington, London SW7 2AZ, United
Kingdom
Centre for Metabolomics and Bioanalysis (CEMBIO), Faculty of Pharmacy, Universidad San Pablo CEU, Campus Monteprincipe,
Boadilla del Monte, 28668 Madrid, Spain
§
Section of Biosurgery & Surgical Technology, Division of Surgery, Department of Surgery and Cancer, Faculty of Medicine, Imperial
College London, QEQM Building, St. Marys Hospital, London W2 1NY, United Kingdom
*
S
Supporting Information
ABSTRACT: We have investigated the urinary and plasma
metabolic phenotype of acute pancreatitis (AP) patients presenting
to the emergency room at a single center London teaching hospital
with acute abdominal pain using
1
H NMR spectroscopy and
multivariate modeling. Patients were allocated to either the AP (n =
15) or non-AP patients group (all other causes of abdominal pain,
n = 21) on the basis of the national guidelines. Patients were
assessed for three clinical outcomes: (1) diagnosis of AP, (2)
etiology of AP caused by alcohol consumption and cholelithiasis,
and (3) AP severity based on the Glasgow score. Samples from AP
patients were characterized by high levels of urinary ketone bodies,
glucose, plasma choline and lipid, and relatively low levels of
urinary hippurate, creatine and plasma-branched chain amino acids.
AP could be reliably identied with a high degree of sensitivity and specicity (OPLS-DA model R
2
= 0.76 and Q
2
Y = 0.59) using
panel of discriminatory biomarkers consisting of guanine, hippurate and creatine (urine), and valine, alanine and lipoproteins
(plasma). Metabolic phenotyping was also able to distinguish between cholelithiasis and colonic inammation among the
heterogeneous non-AP group. This work has demonstrated that combinatorial biomarkers have a strong diagnostic and
prognostic potential in AP with relevance to clinical decision making in the emergency unit.
KEYWORDS: acute pancreatitis, abdominal pain, metabonomics, NMR, patient strati cation
INTRODUCTION
Acute pancreatitis (AP) is an inammatory condition associated
with a progressive systemic inammatory response (SIRS) and,
in severe cases, autonecrosis of pancreatic tissue, organ failure,
and death.
1,2
In Europe, the mortality rate for this condition
remains at 10%, and in severe cases associated with multiple
organ failure it can be as high as 30%. AP aects between 5 and
80/100 000 people globally, and the mainstay of therapy in the
acute phase is the rapid initiation of supportive treatments, the
treatment of reversible underlying causes, and the allocation of
intensive therapies, where needed. In the U.K., an AP diagnosis
is conventionally made on patient history, clinical examination,
and the measurement of serum amylase. However, this lacks
sensitivity and specicity, and it is subject to variation in its
diagnostic threshold over time after the initial pancreatic
insult.
3
Although other biomarkers are available (e.g., serum
pancreatic lipase or urinary amylase), they do not provide
prognostic value and they are expensive. Imaging modalities
such as ultrasound (USS) and dynamic computed tomography
(CT) scanning may also be used to conrm the diagnosis of
AP;
4
however, USS is nonspecic and a CT is only used to
assess severity at 72 h or to rule out other diagnoses.
Prognostication is essential in AP for early organ support and
minimization of mortality. Unfortunately, clinical risk scores
such as the Glasgow score,
5
are only marginally more accurate
than clinical intuition.
610
Other generic physiological scoring
systems are used in critical care such as the Acute Physiology
and Chronic Health Evaluation (APACHE II) score,
11
a generic
physiological measurement based on 12 parameters, that is
designed to measure the severity of disease for adult patients 24
hrs following admission to intensive care units. However, this is
not disease-specic for AP. Ultimately, diagnosing the etiology
of the underlying condition remains the only methodology for
Received: February 19, 2014
Article
pubs.acs.org/jpr
© XXXX American Chemical Society A dx.doi.org/10.1021/pr500161w | J. Proteome Res. XXXX, XXX, XXXXXX

preventing disease progression and further episodes of AP.
Although alcohol consumption and cholelithiasis are respon-
sible in 6080% of cases of AP,
12,13
the causes are multiple, and
10% of cases are classed as idiopathic, severely limiting
denitive therapeutic options in this cohort of patients.
Metabolic proling coupled to computational modeling has
emerged as a tool for describing metabolic systems and is
capable of characterizing dierent time points in diseases.
1416
The determination of metabolite changes that describe a
biological phenotype based on either
1
H nuclear magnetic
resonance (NMR) spectroscopy or mass spectrometry (MS)
has been widely applied for global proling to dene diagnostic
or prognostic biouid proles for physiological or pathological
states.
14,1719
However, to demonstrate that a metabonomic
strategy for characterizing pathology has genuine translational
capacity, it must be robust enough to cope with the complex
clinical heterogeneity encountered in real-world clinical
environments such as the emergency room, which dier greatly
from standardized and tightly controlled experimental environ-
ments. To our knowledge, this approach has yet to be
prospectively applied to patients with AP presenting in the
acute setting, although a small NMR-based comparison has
been made between AP patients from an inpatient clinic ( n =5)
with healthy controls from an outpatient (n = 5).
20
A key
feature of AP is abdominal pain, a common and often
nonspecic presenting complaint for a large array of other
surgical pathologies that require urgent treatment. However,
40% of all cases of acute abdominal pain are labeled, as being
nonspecic. The aim of the current study was therefore to
determine the potential of a metabonomic approach in the
diagnosis and prognostic staging of AP and to ascertain the
clinical utility of this approach in the analysis of acute
abdominal pain in the emergency setting.
MATERIALS AND METHODS
Study Design
This was an observational control study of consecutive patients
presenting with acute abdominal pain to a single center
Accident and Emergency unit at St. Marys hospital, London,
U.K. All work was approved the by the ethical committee in St.
Marys Hospital, London (Rec 05/Q0403/201). The inclusion
criteria were designed to be pragmatic. Therefore, patients were
included if they presented with acute severe abdominal pain of
<72 h duration or if they demonstrated signs of peritonitis
when examined by the emergency physician on their rst
examination. Patients were excluded if they had been
discharged from hospital within 72 h with the same pain, if
they were unconscious at presentation, if they were pregnant, if
they were under the age of 18, or if they had undergone surgery
within the previous 6 weeks or received a blood transfusion.
Patients with congenital pancreatic malformations or those with
cystic brosis were also excluded. Patients with AP were
identied by their medical history and examination and an
elevated amylase that was greater than three times the standard
range (090 units/L). A pancreatic lipase assay was not
routinely available for use within the hospital. Patients were
investigated and managed as per the current British Society of
Gastroenterology guidelines.
21
Therefore, all patients under-
went an abdominal ultrasound (USS) within 24 hrs and a CT
of the abdomen at 72 hrs if no clinical improvement was
obtained or sooner if clinically indicated. Patients underwent
clinical severity scoring using the modied Glasgow criteria
5
at
24 and 48 hrs. All patients were given a diagnosis for their
etiology, and underwent an endoscopic retrograde cholangio-
pancreatography (ERCP) on the same admission if evidence of
obstructive biliary cholelithiasis was demonstrated. Patients
with a previous history of pancreatitis (i.e., > 2 episodes) were
subcategorized to have either recurrent or acute chronic
pancreatitis (CP). CP has a dierent mechanism and clinical
presentation,
22
and these patients were therefore excluded from
the nal analysis. The clinical diagnosis on discharge from
hospital was used for patient phenotyping, as determined by the
attending consultant physician. The aim was to achieve a
practical representation of how clinical, biochemical, and
imaging data can be used to reach the diagnosis during the
admission phase and to determine if a metabonomic method
was able to augment this approach and to dierentiate patients
or subsets of patients according to presence, severity, or cause
of AP in an early stage. The 30 day follow-up data were also
recorded to conrm the nal diagnosis. The non-AP patients
group was investigated according to the discretion of the
attending surgeon. However, their diagnosis was conrmed on
the basis of the clinical presentation, basic serum, and urine
biochemistry and on imaging by USS or CT or at the time of
surgery with histopathological analysis. All patients had clinical
observation data (pulse, temperature, and blood pressure) and
routine clinical biochemical data collected on admission and
then at 24 h intervals until the patient underwent surgery or
was discharged. Samples were collected from each patient for
the rst 5 days of their admission. Samples for metabonomic
analysis were collected as soon as possible after attendance of
the patient in the emergency room or within 24 hrs of
admission to hospital. Patients not recruited prior to this cuto
were excluded.
Sample Collection and Preparation
Urine Samples. Sampling was performed precatheterization
wherever possible. If catheterized, a clean catch sample was
taken. Urine samples were kept on ice after collection prior to
being stored frozen at 80 °C. Urine samples were thawed on
ice, vortexed, and centrifuged; they were prepared by adding
200 μL of phosphate buer pH 7.4 containing 20% of D
2
Oto
400 μL of urine. The mixture was transferred into a 5 mm outer
diameter NMR tube. All
1
H NMR spectra were acquired using
a Bruker DRX600 spectrometer (Rheinstetten, Germany) with
a 5 mm TXI probe operating at 600.13 MHz. The eld
frequency was locked on to D
2
O solvent. Primary acquisitions
were made using a standard 1-D pulse program [recycle delay
(RD)-90°-t
1
-90°-t
m
-90°-acquire free induction decay (FID)].
The 90° pulse length was adjusted to 12 μs. A total of 64
scans were recorded into 32 K data points with a spectral width
of 20 ppm at 300 K. An exponential function was applied to the
FID prior to the Fourier transformation, which resulted in a
line broadening of 0.3 Hz.
Plasma Samples. Plasma samples were collected in 5 mL
aliquots and in heparinized lithium tubes, and they were kept
on ice before being centrifuged. The supernatant obtained was
frozen at 20 °C until the day of analysis. 200 μL of plasma
was added to 400 μL of 0.9% saline solution containing 10% of
D
2
O.
1
H NMR spectra of the plasma samples were acquired
employing two 1-D NMR experiments. These were a standard
1-D pulse sequence (as described for urine) and a Carr
PurcellMeiboomGill (CPMG) pulse sequence [RD-90°-(τ-
180°-τ)
n
-acquire FID] giving a total spinspin relaxation delay
(2
n
τ) of 80 ms, determined by the number of spin echoes and τ
Journal of Proteome Research Article
dx.doi.org/10.1021/pr500161w | J. Proteome Res. XXXX, XXX, XXXXXXB

= 400 μs. Typically, in the standard 1-D and CPMG
experiments, 256 and 128 transients were collected into 32 k
data points, respectively, at 300 K. All method characteristics
and details are detailed elsewhere.
23
Data Processing of NMR Spectra. All urine and plasma
NMR spectra were automatically phased, baseline-corrected,
and referenced to either sodium 3-(trimethylsilyl) propionate-
2,2,3,3-d
4
(TSP; δ 0.0) or α-glucose (δ 5.23) for urine and
plasma, respectively, using in-house scripts. The spectra were
imported to Matlab and the region containing water and urea
resonances (δ 4.7 to 6.2) and TSP (δ 0.1 to 0.1 for urinary
spectra only) were removed. Peak alignment
24
and spectral
normalization using a probabilistic quotient algorithm
25
were
performed in the full-resolution spectra using an in-house
Matlab script (The MathWorks, Natick, MA).
Multivariate Analysis. Principal component analysis
(PCA) was applied to NMR spectral data sets to visualize
outliers and data trends.
26
Partial least-squares discriminant
analysis (PLS-DA) was performed to optimize the classication
of samples and aid identication of candidate biomarkers
associated with clinical phenotype. Orthogonal-PLS-DA
(OPLS-DA)
27
was also carried out to optimize recovery of
potential biomarkers, that is, compounds with high correlation
and covariance with the class (e.g., disease severity or etiology).
Modeling was conducted in SIMCA-P v.12.1 using unit
variance scaling. Pseudoloadings plots were created with a
color code projected onto the spectrum
28
to indicate the
correlation of the metabolites discriminating between the
classes in each comparison, that is, representing AP and other
abdominal pain etiologies. Red indicates high correlation and
dark blue denotes no correlation with sample class. The
direction and magnitude of the signals relate to the covariance
of the metabolites with the class in the model. The quality of
the models was assessed by the cumulative R
2
and Q
2
,
indicating the goodness of t and the predictive power of each
model using SIMCA software (Umetrics SIMCA-P+12.0.1
software). Metabolite identi cation was based on chemical
shifts published in literature and statistical total correlation
spectroscopy (STOCSY) of peaks
29
(e.g., STOCSY of urinary
3-hydroxyisovalerate is shown in Supplemental Figure 1, SI).
HMDB
30
and COLMAR
31
online databases were also used to
conrm the chemical shift values for all metabolites. Finally,
where unknown metabolites were identied or there was
uncertainty in the metabolite identity, 2D NMR experiments,
for example,
1
H
1
H total correlation spectroscopy (TOCSY),
1
H
13
C heteronuclear multiple quantum coherence (HMQC),
and heteronuclear single quantum coherence (HSQC), were
performed for complete chemical structure elucidation. Para-
metric clinical data were analyzed by unpaired two-tailed
Students t test using SPSS v.19.0.
Exclusion of Outliers and Sample Processing. A total of
46 patients were enrolled into the study (AP = 22, non-AP =
24). From this initial recruitment, ve patients had chronic
pancreatitis (CP) and were excluded; an example of urinary CP
spectrum is shown in Supplemental Figure 2, SI. Additionally,
strong outliers detected either by extremely unusual NMR
spectra dominated by a few signals or by signicant broadened
signals due to, for example, excess proteinuria were removed.
Finally, after removing the patients unable to provide samples,
the numbers of patients for urine analysis were AP = 13 and
non-AP = 18, while those for plasma were AP = 15 and non-AP
= 21.
RESULTS
Clinical Data
Patients were aged 18 to 77 years old, with a median age of 42.5
years (SD = 16.8). The remaining demographic data are
Table 1. Summary of Clinical Data
non-AP (n = 23) pancreatitis (n = 15) p value
age (median, range) 42 (1877) 50 (2270) 0.980
sex F:M 12:11 3:12 0.001
cholecystitis 1
biliary colic 1
ascending cholangitis 1
appendicitis 5
diverticulitis 3
colitis 2
adhesions 1
unknown 2
UTI 3
renal colic 1
cecal volvulus 1
gastritis 1
mumps orchitis 1
gall stone pancreatitis 6
EtOH pancreatitis 9
amylase (mean, SD) 62.1 (73.7) 874.6 (841.8) 1.3 × 10
06
symptom duration (mean, SD) 33.4 (24.1) 41.9 (23.8) 0.860
Glasgow 24 h (mean, SD) 0 2.1 (1.8)
Glasgow 48 h 0 1.1 (1.3)
APACHE 24 h (mean, SD) 3.3 (3.2) 9.6 (6.3) 0.014
APACHE 48 h 1.9 (3.0) 5.8 (3.9) 0.140
length of stay, days (median, range) 3 (014) 7 (2132) 0.020
a
Data presented is the average (range or SD). Data were analyzed by unpaired two-tailed Students t test.
Journal of Proteome Research Article
dx.doi.org/10.1021/pr500161w | J. Proteome Res. XXXX, XXX, XXXXXXC

provided in Table 1. The etiologies of pain in the non-AP
patient s group were heterogeneous, reecting the typical
distribution of the causes of acute abdominal pain presenting
to a typical accident and emergency department
32
(Table 1).
Two patients in this cohort were not given a denitive
diagnosis for their abdominal pain, and three patients had a
primary biliary pathology but with a normal blood amylase level
at presentation. The etiologies of the AP group were due to
either cholelithiasis (6/15) or alcohol (9/15). One of the
cholelithiasis AP cohorts had necrotic pancreatitis. The mean
Glasgow score (a disease-specic clinical risk score of severity
ranging between 0 and 3, where 3 represents a severe episode)
was 2.07 at 24 h and 1.07 at 48 h, indicating that this cohort of
AP suered moderately severe episodes. However, one patient
with cholelithiasis AP and one patient with alcoholic AP
diagnosis had a myocardial infarction during their admission,
both of whom were treated with a standard acute coronary
syndrome protocol. The APACHE II score
11
varied signi-
cantly between the groups, indicating the increased general
severity of illness in the AP group compared with the non-AP
group. This was also reected in the hospital length of stay data,
which diered signicantly (average for non-AP group = 3 and
AP = 7; p < 0.02) between diagnostic groups (Table 1).
Characterizing the Urinary Phenotype of AP
The urinary
1
H NMR spectra from this clinical study
demonstrated substantial interindividual variation. Figure 1,
shows urinary
1
H NMR spectra from an alcoholic AP patient
(A) and a non-AP patient (B). In general, spectra from AP
patients were characterized by high concentrations of urinary
glucose and ketone bodies. Unsurprisingly, large concentrations
of acetate, ethanol, acetone, and ethyl glucuronide were also
observed in this patient, reecting the underlying etiology
(Figure 1A). Ethanol was not consistently observed in the six
patients with alcohol-induced pancreatitis due to the varying
volume of alcohol consumed between patients and time since
ingestion (range 24 h to 1 week). Furthermore, it was apparent
from visual inspection across both groups (AP and non-AP) of
urinary spectra that metabolites from common analgesic
nonsteroidal anti-inammatory drugs (such as ibuprofen)
were observed (data not shown), as might be expected in a
cohort of patients presenting with abdominal pain. Signals from
the drug acetaminophen metabolites were present in both
groups, but they varied between patients; higher concentrations
of acetaminophen metabolites can be observed in the
cholecystitis sample in Figure 1B. NMR spectral proles of
acetaminophen and ibuprofen and their urinary metabolites
have been published previously.
33,34
Urine specimens from
several patients contained high concentrations of mannitol [δ
3.65(dd); δ 3.73(m); δ 3.77(d); δ 3.84(dd)] used as an
excipient in the acetaminophen preparation and which was
therefore correlated with acetaminophen and its related urinary
metabolites [acetaminophen sulfate (AS) = δ 2.18(s), δ
7.31(d), δ 7.46(d); acetaminophen glucuronide (AG) = δ
2.17(s), δ 3.62(m), δ 5.10(d), δ 7.13(d), δ 7.36(d);
acetaminophen (A) = δ 2.16(s), δ 6.91(d), δ 7.25(d); N-
acetylcysteine conjugate (NAC) = δ 1.86(s)]. The drug-related
origin of the mannitol was supported by a Pearson correlation
test
35
using all compounds in the prole and a selected driver
peak for the mannitol signal (δ 3.805 ppm, (5-CH)), with an
established cuto of p < 0.05. The resulting graph showed
mannitol signals and acetaminophen (A, AG, AS, and NAC)
was positively and highly correlated, showing correlations
values above 0.897 (from NAC 6.98(d)) with a p value of 0.039
(Supplemental Figure 3, SI). This relationship was conrmed
by
1
H NMR analysis of the intravenous acetaminophen
preparation, a common analgesic given in the acute setting
and by correlation wit h the patient drug charts that
demonstrated the drug was given in the emergency room.
Mannitol is a commonly used excipient in acetaminophen
formulations.
36
The regions containing mannitol (from δ 3.65
to δ 3.97) and acetaminophen metabolites (from δ 1.85 to δ
1.88; δ 2.14 to δ 2.2; δ 3.59 to δ 3.65; δ 7.11 to δ 7.18; and δ
7.29 to δ 7.49) were therefore removed from the raw data for
the remaining analysis to minimize any confounding eects
unrelated to the disease etiology or severity.
Plasma NMR Spectra
There were few obvious disease-specic metabolites that varied
systematically between raw spectra; however, clear quantitative
changes in the spectra were observed, particularly in
concentrations of VLDL, LDL, and nonesteried fatty acids.
A comparison of 1-D NMR spectra from an AP and a non-AP
patient with appendicitis is shown in Figure 2, illustrating the
gross dierences in the lipoproteins. Likewise, a comparison of
CPMG raw spectra from dierent patients, an AP and non-AP
patient diagnosed with diverticulitis, is presented in Figure 4 in
the SI, where dierences of VLDL were observed.
Figure 1. Typical 600 MHz
1
H NMR spectra of urine obtained from a
patient with acute pancreatitis and alcoholic background (A) and a
non-AP patient with cholecystitis (B). Sections from the NMR spectra
(raw data) show the metabolite dierences describing the physical
condition. Key: 1, hippurate; 2, acetaminophen sulfate (AS); 3,
acetaminophen glucuronide (AG); 4, acetaminophen (A); 5, N-
acetylcysteine conjugate of acetaminophen (NAC); 6, creatinine; 7,
guanidinoacetate; 8, glycine; 9, trimethylamine-N-oxide (TMAO); 10,
dimethylamine; 11, citrate; 12,
D-3-hydroxybutyrate; 13, ethanol; 14,
acetone; 15, alanine; 16, succinate; 17, 3-hydroxyisovalerate; 18,
malonic acid; 19, creatine.
Journal of Proteome Research Article
dx.doi.org/10.1021/pr500161w | J. Proteome Res. XXXX, XXX, XXXXXXD

Multivariate Modeling of Patients with Severe Abdominal
Pain
Three main analyses were performed based on clinical criteria:
(1) diagnosis of AP (AP vs non-AP), (2) etiology (alcohol
consumption vs cholelithiasis), and (3) AP severity, based on a
validated modied Glasgow score comparison of mild
(Glasgow score 1) versus moderate to severe (Glasgow score
2 and 3). It was not possible to build predictive models for
organ failure or mortality based on the small numbers of
patients. PCA, PLS-DA, and OPLS-DA models were therefore
created to determine the capacity of the metabolic proles to
classify the sample set according to the above outcomes and to
determine how well the model could predict the class of
samples according to the principal diagnosis. PCA models did
not show clear separation for any of the clinical scenarios in this
study due to the extreme heterogeneity of the data. PLS-DA
plots for each clinical scenario in urine and plasma (CPMG-
and 1-D-spectral) data (Figure 3) showed positive predictive
models for all with the exception of prediction of etiology in the
plasma (1-D-plasma and CPMG) data set. Discriminatory
metabolites were found in the OPLS-DA pseudoloading plots
for each of the three sets of models for AP prediction, severity,
and etiology in urine and plasma (Figure 4) and diered in
composition between models assessing dierent outcomes.
Discriminant metabolites found in plasma and urine for the
three clinical scenarios are summarized in Tables 24 and
discussed below.
1. Diagnosis. Consistent changes in concentrations of
urinary and plasma metabolites were found between the AP and
the non-AP gro up. Plasma acetone,
D-3-hydroxybutyrate,
acetoacetate, glucose, lipid signals (CH
3
CH
2
,CH
2
CH
CH) fraction, and plasma choline concentrations were
increased in AP patients. Low urinary levels of hippurate,
creatine, and plasma valine and alanine were also characteristic
of the AP group (Table 2). Because hippurate is a well-
established gut microbial cometabolite derived from glycine
conjugation of benzoic acid,
37,38
a subgroup analysis of
antibiotic-treated patients (n = 10; 5non-AP/5AP) and non-
antibiotic-treated (n = 21; 12non-AP/9AP) patients was
performed to determine if antibiotic ingestion in the AP
diagnostic model was responsible for the dierence in urinary
excretion of hippurate between the two groups (Figure 5 and
Supplemental Table 1, SI). The PLS-DA analysis demonstrated
class separation according to antiobiotic usage (R
2
= 0.85, Q
2
=
0.47) based on modulation of a combination of metabolites
including decreased urinary citrate, methylamine, and crea-
tinine, suggesting that antibiotic use did have a systemic
metabolic impact in this group of patients. Some of the
dierentiating signals derived directly from antibiotics (e.g.,
metronidazole
39
). Concentrations of 3-hydroxyisovaleric acid
previously reported to be a gut microbial cometabolite
40
were
decreased in the antibiotic group. Ho wever, hippurate
concentrations did not vary signicantly (driver peak = δ
1
H
7.85 (6-CH); correlation (r)=0.29; p = 0.11) between
groups, suggesting that the AP pathophysiology has a greater
eect on this mammalian-host cometabolite than antibiotic use.
In addition to signals from analgesics and NSAID metabolites
(Figure 5, SI), the multivariate urinary model of antibiotic
usage (Figure 5 and Supplemental Table 1, SI) shows unknown
compounds whose excretion was increased in the antibiotic
group, which are most likely to be xenometabolites. Antibiotic
usage was associated with decreased urinary concentrations of
3-hydroxyisovalerate, methylamine, citrate, and creatinine.
To compare the robustness of the metabonomic models in
urine and plasma, we created receiver operating characteristic
(ROC) curves (Figure 6A,B). The urine
1
H NMR model
(AUC = 0.91) produced a stronger diagnostic model than
plasma (AUC = 0.86). The urinary multivariate model had a
high sensitivity with true-positive rates (TPRs) of 0.5 to 0.6,
low false-positive rates (FPR) of <0.2, and a classication rate
of 80%. Additionally, 10 metabolites (5 from each biouid)
were combined to generate a denitive diagnostic ROC curve
(Figure 6C). This demonstrated a stronger diagnostic model
for AP (AUC = 0.96) when compared with each biouid
separately. Moreover, to emphasize the quality of the
dierential biomarker for the diagnosis, a PCA analysis was
performed based on these 10 discriminant metabolites. The
scores plot (Supplemental Figure 6A, SI) showed a clear
separation of disease groups. The permutation test (Figure 6B,
SI) for these metabolites indicated that the model was robust.
Moreover, the cross-validated OPLS-DA was also more robust
(Figure 6D, R
2
= 0.76, Q
2
= 0.59). Serum amylase data had a
sensitivity of 100% and a specicity of 95%, which although
comparable to the metabonomic data is likely to be of little
signicance, as it was used as a diagnostic criterion and it had
no prognostic value.
2. Etiology . Multivariate models of alcohol versus
cholelithiasis AP had a good classication capacity (R
2
Y from
51 to 90%) but variable predictive capacity; Q
2
ranged from 0
to 24% (Figure 3(ii)). The models generated from the plasma
(CPMG and 1-D) spectra carried a negative predictive value in
the rst component, indi cating that there was no clear
diagnostic signature for etiology based on the low-molecular-
weight components in plasma. However, OPLS-DA from
urinary
1
H NMR spectra generated a stronger model (R
2
Y =
0.76, Q
2
= 0.44), conrming that urinary ketone body excretion
Figure 2. 600 MHz
1
H NMR spectra (δ 0.0 to 4.7) of plasma from
(A) pancreatitis patient and (B) non-AP patient diagnosed with
appendicitis. Abbreviations: LDL1 and VLDL1 refer to the terminal
CH
3
groups of fatty acids in low-density and very-low-density
lipoprote ins, respectively. HDL refer s to the C18 signal from
cholesterol in high-density lipoprotein. NAC1 and NAC2 refer to
composite acetyl signals from α1-acid glycoprotein.
Journal of Proteome Research Article
dx.doi.org/10.1021/pr500161w | J. Proteome Res. XXXX, XXX, XXXXXXE

Figures (11)
Citations
More filters

Journal ArticleDOI
Abstract: The gut microbiota is composed of a huge number of different bacteria, which produce a large amount of compounds playing a key role in microbe selection and in the construction of a metabolic signaling network. The microbial activity is affected by environmental stimuli leading to the generation of a wide number of compounds, which influence the host metabolome and human health. Indeed, metabolic profiles related to the gut microbiota can offer deep insights on the impact of lifestyle and dietary factors on chronic and acute diseases. Metagenomics, metaproteomics and metabolomics are some of the meta-omics approaches to study the modulation of the gut microbiota. Metabolomic research applied to biofluids allows to: define the metabolic profile; identify and quantify classes and compounds of interest; characterize small molecules produced by intestinal microbes; and define the biochemical pathways of metabolites. Mass spectrometry and nuclear magnetic resonance spectroscopy are the principal technologies applied to metabolomics in terms of coverage, sensitivity and quantification. Moreover, the use of biostatistics and mathematical approaches coupled with metabolomics play a key role in the extraction of biologically meaningful information from wide datasets. Metabolomic studies in gut microbiota-related research have increased, focusing on the generation of novel biomarkers, which could lead to the development of mechanistic hypotheses potentially applicable to the development of nutritional and personalized therapies.

210 citations


Cites methods from "1H NMR global metabolic phenotyping..."

  • ...This has also been reported by Bro et al. (2015), who used plasma to determine breast cancer biomarkers, or by Villaseñor et al. (2014), who described the global metabolic phenotyping of acute pancreatitis, and by Dumas et al. (2016), who used this technique to study metabolic syndrome and fatty…...

    [...]


Journal ArticleDOI
TL;DR: The findings suggest that modeling against a continuous ELF-derived score of fibrosis provides a more robust assessment of the metabolic changes associated with fibrosis than modeling against the categorical METAVIR score.
Abstract: Metabolic Phenotyping for Enhanced Mechanistic Stratification of Chronic Hepatitis C-Induced Liver Fibrosis

29 citations


Journal ArticleDOI
TL;DR: It is demonstrated that siderophore biosynthesis coordinately modulated the differential metabolomes and thus may indicate novel targets for virulence-based diagnosis, therapeutics, and drug development related to urinary tract infections.
Abstract: Urinary tract infections impose substantial health burdens on women worldwide. Urinary tract infections often incur a high risk of recurrence and antibiotic resistance, and uropathogenic E. coli accounts for approximately 80% of clinically acquired cases. The diagnosis of, treatment of, and drug development for urinary tract infections remain substantial challenges due to the complex pathogenesis of this condition. The clinically isolated UPEC 83972 strain was found to produce four siderophores: yersiniabactin, aerobactin, salmochelin, and enterobactin. The biosyntheses of some of these siderophores implies that the virulence of UPEC is mediated via the targeting of primary metabolism. However, the differential modulatory roles of siderophore biosyntheses on the differential metabolomes of UPEC and non-UPEC strains remain incompletely understood. In the present study, we sought to investigate how the differential metabolomes can be used to distinguish UPEC from non-UPEC strains and to determine the associ...

22 citations


Journal ArticleDOI
TL;DR: This review critically discusses recent findings of urine (non-invasive) and blood serum and plasma (minimally invasive) metabolomics, focusing on key role of metabolites and their use as potential biomarkers for diagnostic purposes.
Abstract: The metabolome is affected by individual physiologic and pathophysiologic states as well as the environment. Metabolomics has become crucial approach for clinical studies, providing a better understanding of disease mechanisms. The expansion of analytical methods aiming at performing detailed analysis of biofluids has led to the characterization of many disease biomarkers. NMR provides faster and more comprehensive assessment of the biological samples in human models. NMR-based profiling studies aimed at identifying biomarkers for specific diseases has significantly increased over the last few years. These have attempted to correlate human pathophysiology with alterations in the metabolic profile of common biofluids such as urine, plasma and serum. In this context, NMR-based untargeted metabolomics has become an important adjunct for the identification of biomarkers in disease research, not only for early diagnosis purposes, but also for therapy prediction, precise prognosis or monitoring of disease progression. This review critically discusses recent findings of urine (non-invasive) and blood serum and plasma (minimally invasive) metabolomics, focusing on key role of metabolites and their use as potential biomarkers for diagnostic purposes.

20 citations


Journal ArticleDOI
TL;DR: It is suggested that GC-MS based serum metabolomics method can be used in the clinical diagnosis of AP by profiling potential biomarkers.
Abstract: Acute pancreatitis (AP) is defined as an acute inflammation of pancreas that may cause damage to other tissues and organs depending upon the severity of symptoms. The diagnosis of AP is usually made by detection of raised circulating pancreatic enzyme levels, but there are occasional false positive and false negative diagnoses and such tests are often normal in delayed presentations. More accurate biomarkers would help in such situations. In this study, the global metabolites' changes of AP patients (APP) were profiled by using gas chromatography-mass spectrometry (GC-MS). Multivariate pattern recognition techniques were used to establish the classification models to distinguish APP from healthy participants (HP). Some significant metabolites including 3-hydroxybutyric acid, phosphoric acid, glycerol, citric acid, d-galactose, d-mannose, d-glucose, hexadecanoic acid and serotonin were selected as potential biomarkers for helping clinical diagnosis of AP. Furthermore, the metabolite changes in APP with severe and mild symptoms were also analyzed. Based on the selected biomarkers, some relevant pathways were also identified. Our results suggested that GC-MS based serum metabolomics method can be used in the clinical diagnosis of AP by profiling potential biomarkers.

17 citations


Cites background from "1H NMR global metabolic phenotyping..."

  • ...Their results suggested that combinatorial biomarkers consisting of guanine , 74 hippurate and creatine (urine), and valine, alanine a d lipoproteins (plasma) have a strong diagnostic 75 and prognostic potential in APP [18]....

    [...]


References
More filters

Journal ArticleDOI
TL;DR: The Human Metabolome Database is designed to address the broad needs of biochemists, clinical chemists, physicians, medical geneticists, nutritionists and members of the metabolomics community.
Abstract: The Human Metabolome Database (HMDB) is currently the most complete and comprehensive curated collection of human metabolite and human metabolism data in the world. It contains records for more than 2180 endogenous metabolites with information gathered from thousands of books, journal articles and electronic databases. In addition to its comprehensive literature-derived data, the HMDB also contains an extensive collection of experimental metabolite concentration data compiled from hundreds of mass spectra (MS) and Nuclear Magnetic resonance (NMR) metabolomic analyses performed on urine, blood and cerebrospinal fluid samples. This is further supplemented with thousands of NMR and MS spectra collected on purified, reference metabolites. Each metabolite entry in the HMDB contains an average of 90 separate data fields including a comprehensive compound description, names and synonyms, structural information, physico-chemical data, reference NMR and MS spectra, biofluid concentrations, disease associations, pathway information, enzyme data, gene sequence data, SNP and mutation data as well as extensive links to images, references and other public databases. Extensive searching, relational querying and data browsing tools are also provided. The HMDB is designed to address the broad needs of biochemists, clinical chemists, physicians, medical geneticists, nutritionists and members of the metabolomics community. The HMDB is available at: www.hmdb.ca

2,316 citations


Journal ArticleDOI
TL;DR: Metabonomics is a systems approach for studying in vivo metabolic profiles, which promises to provide information on drug toxicity, disease processes and gene function at several stages in the discovery-and-development process.
Abstract: The later that a molecule or molecular class is lost from the drug development pipeline, the higher the financial cost. Minimizing attrition is therefore one of the most important aims of a pharmaceutical discovery programme. Novel technologies that increase the probability of making the right choice early save resources, and promote safety, efficacy and profitability. Metabonomics is a systems approach for studying in vivo metabolic profiles, which promises to provide information on drug toxicity, disease processes and gene function at several stages in the discovery-and-development process.

1,754 citations


Journal ArticleDOI
TL;DR: The main NMR spectroscopic applications in modern metabolic research are summarized, and detailed protocols for biofluid and tissue sample collection and preparation are provided, including the extraction of polar and lipophilic metabolites from tissues.
Abstract: Metabolic profiling, metabolomic and metabonomic studies mainly involve the multicomponent analysis of biological fluids, tissue and cell extracts using NMR spectroscopy and/or mass spectrometry (MS). We summarize the main NMR spectroscopic applications in modern metabolic research, and provide detailed protocols for biofluid (urine, serum/plasma) and tissue sample collection and preparation, including the extraction of polar and lipophilic metabolites from tissues. 1H NMR spectroscopic techniques such as standard 1D spectroscopy, relaxation-edited, diffusion-edited and 2D J-resolved pulse sequences are widely used at the analysis stage to monitor different groups of metabolites and are described here. They are often followed by more detailed statistical analysis or additional 2D NMR analysis for biomarker discovery. The standard acquisition time per sample is 4-5 min for a simple 1D spectrum, and both preparation and analysis can be automated to allow application to high-throughput screening for clinical diagnostic and toxicological studies, as well as molecular phenotyping and functional genomics.

1,611 citations


Journal ArticleDOI
TL;DR: The probabilistic quotient normalization is introduced in this work, which is based on the calculation of a most probable dilution factor by looking at the distribution of the quotients of the amplitudes of a test spectrum by those of a reference spectrum.
Abstract: For the analysis of the spectra of complex biofluids, preprocessing methods play a crucial role in rendering the subsequent data analyses more robust and accurate. Normalization is a preprocessing method, which accounts for different dilutions of samples by scaling the spectra to the same virtual overall concentration. In the field of 1H NMR metabonomics integral normalization, which scales spectra to the same total integral, is the de facto standard. In this work, it is shown that integral normalization is a suboptimal method for normalizing spectra from metabonomic studies. Especially strong metabonomic changes, evident as massive amounts of single metabolites in samples, significantly hamper the integral normalization resulting in incorrectly scaled spectra. The probabilistic quotient normalization is introduced in this work. This method is based on the calculation of a most probable dilution factor by looking at the distribution of the quotients of the amplitudes of a test spectrum by those of a refer...

1,217 citations


Journal ArticleDOI
TL;DR: APR-induced alterations initially protect the host from the harmful effects of bacteria, viruses, and parasites, however, if prolonged, these changes in the structure and function of lipoproteins will contribute to atherogenesis.
Abstract: Infection and inflammation induce the acute-phase response (APR), leading to multiple alterations in lipid and lipoprotein metabolism. Plasma triglyceride levels increase from increased VLDL secretion as a result of adipose tissue lipolysis, increased de novo hepatic fatty acid synthesis, and suppression of fatty acid oxidation. With more severe infection, VLDL clearance decreases secondary to decreased lipoprotein lipase and apolipoprotein E in VLDL. In rodents, hypercholesterolemia occurs attributable to increased hepatic cholesterol synthesis and decreased LDL clearance, conversion of cholesterol to bile acids, and secretion of cholesterol into the bile. Marked alterations in proteins important in HDL metabolism lead to decreased reverse cholesterol transport and increased cholesterol delivery to immune cells. Oxidation of LDL and VLDL increases, whereas HDL becomes a proinflammatory molecule. Lipoproteins become enriched in ceramide, glucosylceramide, and sphingomyelin, enhancing uptake by macrophages. Thus, many of the changes in lipoproteins are proatherogenic. The molecular mechanisms underlying the decrease in many of the proteins during the APR involve coordinated decreases in several nuclear hormone receptors, including peroxisome proliferator-activated receptor, liver X receptor, farnesoid X receptor, and retinoid X receptor. APR-induced alterations initially protect the host from the harmful effects of bacteria, viruses, and parasites. However, if prolonged, these changes in the structure and function of lipoproteins will contribute to atherogenesis.

1,186 citations


Frequently Asked Questions (1)
Q1. What are the contributions mentioned in the paper "H nmr global metabolic phenotyping of acute pancreatitis in the emergency unit" ?

The authors have investigated the urinary and plasma metabolic phenotype of acute pancreatitis ( AP ) patients presenting to the emergency room at a single center London teaching hospital with acute abdominal pain using H NMR spectroscopy and multivariate modeling. This work has demonstrated that combinatorial biomarkers have a strong diagnostic and prognostic potential in AP with relevance to clinical decision making in the emergency unit.