Article

Cite This: J. Proteome Res. 2017, 16, 4208-4216 pubs.acs.org/jpr

Urinary Metabolic Phenotyping of Women with Lower Urinary Tract

Symptoms
Rhiannon Bray,*,† Stefano Cacciatore,‡ Beatriz Jiménez,§,# Rufus Cartwright,† Alex Digesu,†
Ruwan Fernando,† Elaine Holmes,⊥,§ Jeremy K. Nicholson,⊥,§ Phillip R. Bennett,‡,¶
David A. MacIntyre,‡ and Vik Khullar*,†
†
Department of Urogynaecology, St. Mary’s Hospital, Imperial College Healthcare NHS Trust, London W2 1NY, U.K.
‡
Imperial College Parturition Research Group, Division of the Institute of Reproductive and Developmental Biology, §Division of
Computational Systems Medicine, Department of Surgery and Cancer, and ⊥Centre for Digestive and Gut Health, Imperial College
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

London, London SW7 2AZ, U.K.

Downloaded via UNIV NACIONAL AUTONOMA MEXICO on May 5, 2021 at 18:35:21 (UTC).

¶
Queen Charlotte’s Hospital, Imperial College Healthcare NHS Trust, London W12 0HS, U.K.
#
Imperial Clinical Phenotyping Centre, QEQM Building, Imperial College London, Saint Mary’s Hospital, London W21NY, U.K.
*
S Supporting Information

ABSTRACT: Lower urinary tract symptoms (LUTS), including

urinary incontinence, urgency and nocturia, affect approximately half
of women worldwide. Current diagnostic methods for LUTS are
invasive and costly, while available treatments are limited by side
effects leading to poor patient compliance. In this study, we aimed to
identify urine metabolic signatures associated with LUTS using proton
nuclear magnetic resonance (1H NMR) spectroscopy. A total of 214
urine samples were collected from women attending tertiary
urogynecology clinics (cases; n = 176) and healthy control women
attending general gynecology clinics (n = 36). Despite high variation in
the urine metabolome across the cohort, associations between urine
metabolic profiles and BMI, parity, overactive bladder syndrome,
frequency, straining, and bladder storage were identified using
KODAMA (knowledge discovery by accuracy maximization). Four distinct urinary metabotypes were identified, one of which
was associated with increased urinary frequency and low BMI. Urine from these patients was characterized by increased levels of
isoleucine and decreased levels of hippurate. Our study suggests that metabolic profiling of urine samples from LUTS patients
offers the potential to identify differences in underlying etiology, which may permit stratification of patient populations and the
design of more personalized treatment strategies.
KEYWORDS: metabolic profiling, LUTS, overactive bladder, NMR, KODAMA, metabolomics

■ INTRODUCTION
Lower urinary tract symptoms (LUTS) can be grouped into
storage (daytime frequency, nocturia, urgency), voiding
(hesitancy, intermittency, slow stream), and incontinence
symptoms (including urgency incontinence and stress incon-
tinence).1 These symptoms are highly prevalent across all ages,
with more than 50% of women worldwide experiencing one or
more.2 The most prevalent LUTS syndrome is overactive
bladder (OAB), which affects one in six women and
encompasses bothersome urinary urgency, frequency, nocturia,
and typically urinary incontinence.1,3 OAB is associated with a
significant economic burden, with directs costs estimated to be
US $11 billion in the United States alone per year.4 The
condition also has far-reaching effects on physical and mental
health, daily activities, and quality of life.5
The underlying mechanisms of many LUTS are poorly
characterized, particularly for women. Current competing
hypotheses variously locate the primary pathological defects
in the nervous supply to the bladder, the bladder or pelvic floor
musculature, or molecular changes relating the bladder
urothelium.6−8 Objective diagnostic criteria are lacking for the
majority of LUTS with symptomatic diagnoses mainly based on
standard definitions (Table S1).1 There is concern that current
diagnostic algorithms, especially those for OAB, fail to account
for the underlying pathogenesis of the symptoms or
incorporate any objective assessment of symptoms.9−11 Func-
tional testing of the lower urinary tract with urodynamic testing
has generally been used to confirm symptomatic diagnoses with
physiological findings.12 Although these urodynamic tests are
considered to be the most definitive available, they are invasive,
expensive, time-consuming, unreliable, and have very limited
prognostic significance.13,14 The most recent guidance
Received: August 10, 2017
Published: September 22, 2017

© 2017 American Chemical Society 4208 DOI: 10.1021/acs.jproteome.7b00568

J. Proteome Res. 2017, 16, 4208−4216
Journal of Proteome Research
recommends that urodynamic testing should not be used in
women prior to the use of conservative treatments.15,16
Treatment strategies for LUTS are often hindered by side
effects and poor patient compliance.17−20 Current diagnostic
tools are unable to provide evidence as to which women will
either benefit from or tolerate different treatments. The
identification of novel, noninvasive biomarkers would provide
substantial clinical benefit in helping to direct and tailor
treatments. While a large number of putative urinary
biomarkers have been suggested for OAB syndrome, they
lack specificity, reliability, and prognostic power and as result
have not been adopted in clinical practice.7,21−23
Urine metabolic profiling involves the quantitative measure-
ment and identification of metabolites and small molecules that
represent end products of normal and pathological cellular
processes.24 For many pathological conditions, metabolites can
be used as biomarkers of disease and patterns of metabolic
perturbations exploited to achieve better understanding of the
biochemical changes associated with pathophysiology. High
field proton nuclear magnetic resonance (1H NMR) spectros-
copy is widely used for metabolic profiling as it permits rapid
quantitative measurement of metabolites within a biological
sample with minimal preparation, in a highly reproducible and
nondestructive manner that enables recovery of the sample.25 It
has been extensively used to examine the biochemistry of
disease onset and for diagnostic applications in fields such as
cancer,26−29 vascular disease,30 and diseases of the nervous
system.31
In this study, we performed 1H NMR-based metabolic
profiling on urine samples to determine if perturbations in the
urinary metabolome can be associated with LUTS and OAB
syndrome and thus provide unique metabolic signatures useful
for diagnostic and stratification purposes.
■
METHODS
Patient Cohort and Classification
This study was conducted following NHS Health Research
Authority National Research Ethics Service (NRES) Commit-
tee Approval (REC 12/LO/0394). Women attending specialist
urogynecology and general gynecology clinics were recruited
following informed consent. Patients provided a midstream
urine sample under nonfasting conditions and completed the
validated International Consultation on Incontinence−Female
LUTS Questionnaire.32,33 This questionnaire has 13 questions
encompassing voiding, storage, and incontinence symptoms
that qualitatively assess the severity of individual LUTS and
provides separate summary scores for storage, voiding, and
incontinence symptoms. Scoring is generally quantified
according to the following; 0 (never), 1 (occasionally), 2
(sometimes), 3 (most of the time), and 4 (all of the time).
Frequency scoring is quantified using the following ranges; 0
(1−6 times/day), 1 (7−8), 2 (9−10), 3 (11−12), and 4 (≥13).
Control participants were considered to be those reporting
nocturia less than once (ICIQ-FLUTS score = 0), urgency,
urgency incontinence, stress incontinence, or bladder pain
“never” or “occasionally” (≤1) and frequency less than eight
times (≤1). Patients with OAB were identified as those
reporting a minimum composite of the following symptoms;
urgency ≥ occasional (≥1), frequency ≥7−8 times (≥1), and
nocturia ≥ once (≥1) (Table S1). Exclusion criteria for the
study were women: (i) aged less than 18 years or over 70, (ii)
unable to consent, (iii) with a known malignancy, (iv) with
4209
Article
urinary tract infection, and (v) receiving anticholinergic
medication during the study. Ethnicity was self-reported as
White (European ancestry), Asian (Pakistani, Indian, Banglade-
shi, or Sri Lankan ancestry), Arabic, Black (African or African-
Caribbean ancestry), or Mixed. Parity and menopausal status as
well as height and weight were self-reported with BMI
subsequently calculated.
Sample Preparation
All urine samples provided were placed immediately on ice,
aliquoted, and stored at −80 °C within 1 h of collection.
Preparation and analysis of samples was performed using
standardized and optimized protocols as previously described.34
Briefly, urine was defrosted on ice and 540 μL added to 60 μL
of buffer (1.5 M KH2PO4/D2O, 2 mM NaN3 and 0.1% 3-
(trimethl-silyl)propionic acid-d4) (TSP). The mixture was then
centrifuged at 13 000g, and 550 μL of the resulting supernatant
was transferred into 5 mm diameter NMR tubes. Samples were
immediately loaded onto a refrigerated SampleJet robot
(Bruker Corporation, Germany) and kept at 4 °C until
measurement. Pooled quality control (QC) samples were
generated from a composite of all samples and used to assess
potential batch effects.
NMR Experiments and Spectral Processing
Standard monodimensional NMR experiments were run in
automation at 300 K in a Bruker Advance III 600 spectrometer
working at 14.1 T equipped with a BBI probe (Bruker Biospin,
Germany). Each experiment consisted of 32 Free Induction
Decays (FID) in 65 536 points using a 20 ppm window
centered at 4.75 ppm. The relaxation delay was set at 4 s, and a
water presaturation pulse was applied during this period: an
additional presaturations pulse was added using the first
increment of the NOE pulse sequence to further cancel the
water signal. Processing of the spectra (i.e., Fourier trans-
formation, phasing, and baseline correction) was done in
automation using TopSpin (Version 3.2; Bruker Biospin,
Germany).
Spectra were first calibrated in automation to the TSP signal
(0.00 pm). However, as the chemical shift of TSP is susceptible
to changes in the sample matrix and macromolecule interaction,
a second calibration was performed using the alanine CH3
proton peak at 1.48 ppm. The regions corresponding to water
(4.50−4.90 ppm) and urea (5.50−6.10 ppm) were removed
from the subsequent analysis. Each spectrum in the range
between 0.2 and 10.0 ppm was segmented into 0.02-ppm
chemical shift bins, and the corresponding spectral areas under
the curve were integrated using AMIX software (Version
3.9.14; Bruker BioSpin, Germany) giving a total of 465
variables. To account for the potential dilution effect of urine,
each processed spectrum was subsequently normalized using
Probabilistic Quotient Normalization (PQN).35
Identification of signals associated with LUTS was under-
taken using the SBASE database in Amix (v3.9.11; Bruker
BioSpin, Germany) or available assignments in the literature.36
Metabolite identification was also undertaken using statistical
total correlation spectroscopy (STOCSY),37 which was
performed in Matlab (Mathworks v.2014a) using in-house
scripts as well as 2D NMR experiments (1H,1H−COSY, 1H,1H-
TOCSY, and 1H, 13C-HSQC) run on a Bruker 800 MHz
spectrometer working at 18.8 T. Relative metabolite concen-
trations were calculated by integrating the reference signals in
the raw spectra before normalizing to the PQN coefficients
DOI: 10.1021/acs.jproteome.7b00568
J. Proteome Res. 2017, 16, 4208−4216
Journal of Proteome Research
Table 1. Clinical Demographics for Controls and Casesa
participants, n
ethnicity, n
Arabic
Asian
Black
Mixed
White
age (year), median [CI95%]
BMI (m/kg2), median [CI95%]
parity (number), median [CI95%]
pre-/postmenopausal, n
a
Median with 95% confidence intervals (CI95%) are shown.
Figure 1. Unsupervised and supervised multivariate of 1H NMR urine metabolic profiles collected from controls (gray, n = 36) and LUTS cases
(blue, n = 178). (A) No clear clustering of cases or controls was observed when data were analyzed using unsupervised principal components analysis
(PCA). Total variance of 13.6% was explained by the first component and 5.9% by the second component. (B) Supervised partial least-squares
(PLS) modeling of the data was then performed; however, robust discrimination between LUTS cases and controls was not obtained (R2Y = 0.1
[0.1−0.1], Q2Y = 0.06 [−0.07−0.05], p = 0.09).
applied to the processed spectra. Integration of metabolite
signals was performed using an in-house R script.
Statistical Analysis
Fisher’s exact test was used to assess differences between
categorical variables in demographic data (e.g., ethnicity)
between control and case patients. Wilcoxon rank-sum test
was used to assess differences between continuous variables in
demographic data (e.g., age, BMI, parity, and pre/post-
menopausal status). Prior to multivariate analysis, data were
mean-centered and scaled to unit variance. 38 Principal
component analysis (PCA) was performed on the processed
spectra, in order, to identify anomalous samples and to obtain
an overview of the structure of the data.
Regression of metabolic profiles against questionnaire
responses and clinical features was performed using partial
least-squares (PLS) analysis. To assess the predictive ability of
the PLS regression model, a 10-fold cross-validation was
conducted as previously described.27 This involved iteratively
removing 10% of samples prior to any step of the statistical
analysis (including PLS component selection, mean-centering,
and univariate scaling) and back-predicting them into the
model obtained from the remainder of the data. Parameter
selection (i.e., best number of components for PLS) was carried
out by means of an inner 10-fold cross-validation on the
remaining 90% of the data. The overall procedure was repeated
10 times. The goodness of fit parameter (R2) and the predictive
Article
control case p-value
36 176
0.6668
0 3
1 11
1 12
1 2
32 137
35.5 [22.6−57.3] 43.0 [24.4−67.0] 0.0014
21.6 [18.0−42.0] 24.1 [17.9−40.5] 0.0010
0 [0−3] 1 [0−4] <0.0001
19/3 60/22 0.2702
ability parameter (Q2) were calculated using standard
definitions.39 To estimate the performance of the PLS
regression model, statistical significance (p-value) was further
assessed for the Q2 value through permutation testing.40,41
The KODAMA algorithm was used to facilitate identification
of patterns representing underlying metabolic phenotypes on
all samples in the data set, including controls.42,43 The K-test
was then performed to reveal any associations between the
KODAMA scores and the clinical and questionnaire data. This
test uses the R2 to assess the proportion of the variance in the
dependent variable (KODAMA scores) that is predicted from
the independent variable (e.g., clinical parameter) and can thus
be used as a measure of the goodness of fit.44
Partition around medoids (PAM) clustering45 was applied to
the KODAMA scores with the silhouette algorithm46 used to
validate the subsequent results. This algorithm provides a
succinct graphical representation of how well each object lies
within its cluster with the silhouette median value being used to
evaluate the optimal number of clusters with the number of
possible clusters varying from 2 to 10.46 The distributions of
scores for each clinical parameter (questionnaire responses and
demographic data) were then compared across clusters
identified by the PAM analysis using Kruskal−Wallis test.
Spearman’s test was used to calculate the correlation
coefficients (rho) between the metabolite concentrations and
the clinical features. Hierarchical clustering (Ward’s linkage)
was used to order the heatmap showing the different metabolite
4210 DOI: 10.1021/acs.jproteome.7b00568
J. Proteome Res. 2017, 16, 4208−4216
Journal of Proteome Research Article

Table 2. PLS Regression Models for Clinical Features and Questionnaire Data Urine Metabolic Profilesa
PLS model
R2Y [95%CI] Q2Y [95%CI] p-value FDR
Clinical Features
age 0.163 [0.163−0.163] 0.075 [0.048−0.087] <0.001 0.003
BMI 0.065 [0.065−0.12] −0.123 [−0.265−0.053] 0.956 0.965
parity 0.09 [0.059−0.09] −0.047 [−0.09−0.014] 0.438 0.611
menopausal status 0.201 [0.126−0.201] −0.042 [−0.116−0.018] 0.273 0.468
Questionnaire
control versus cases 0.022 [0.022−0.022] −0.014 [−0.034−0.003] 0.174 0.468
nocturia 0.08 [0.069−0.08] −0.007 [−0.058−0.011] 0.061 0.438
urgency 0.029 [0.029−0.029] −0.012 [−0.072−0.001] 0.161 0.468
bladder pain 0.047 [0.047−0.073] −0.039 [−0.054−0.019] 0.458 0.611
frequency 0.046 [0.046−0.076] 0 [−0.022−0.014] 0.077 0.438
hesitancy 0.089 [0.054−0.089] −0.121 [−0.19−0.084] 0.959 0.965
straining 0.075 [0.065−0.075] −0.016 [−0.046−0.019] 0.215 0.468
intermittancy 0.041 [0.041−0.041] −0.04 [−0.07−0.01] 0.396 0.595
urgency incontinence 0.051 [0.051−0.07] −0.026 [−0.063−0.004] 0.269 0.468
frequency of incontinence 0.052 [0.052−0.066] −0.028 [−0.056−0.009] 0.268 0.468
stress incontinence 0.045 [0.045−0.06] −0.072 [−0.119−0.045] 0.766 0.940
insensible incontinence 0.043 [0.043−0.065] −0.082 [−0.133−0.054] 0.866 0.965
nocturnal enuresis 0.087 [0.074−0.1] −0.123 [−0.195−0.075] 0.965 0.965
storage score 0.074 [0.063−0.074] −0.014 [−0.03−0.018] 0.182 0.468
voiding score 0.041 [0.041−0.053] −0.066 [−0.083−0.048] 0.783 0.940
incontinence score 0.051 [0.051−0.076] −0.042 [−0.073−0.015] 0.396 0.595
OAB 0.081 [0.048−0.081] 0 [−0.032−0.015] 0.091 0.438
a
Robustness of PLS models was assessed using goodness of fit (R2) and predictive ability parameters (Q2). Significance of model prediction was
examining using permutation testing with the reported p-value presented before and after false discovery rate correction (FDR). Age was identified as
having a significant impact upon the metabolic profile of the samples assessed in this study (p < 0.001; q = 0.003).

concentrations across the clinical and questionnaire features.
Metabolite over-representation analysis was used to investigate
relationships between identified metabolites and pathways.36,47
A p-value <0.05 was considered to be significant. To account
for multiple testing, a false discovery rate (FDR) of <20% was
applied to reduce identification of false positives.48 All
calculations were made using R41 scripts developed in-house.
predictive model could only be identified for age (p < 0.001,
FDR = 0.003) (Table 2, Figure S1).
To identify potential underlying urinary metabotypes present
in the LUTS cohort, including controls, the data set was then
analyzed using the KODAMA algorithm (Figure 3A). The
relationship between the relative position of each urine
metabolic profile on the resulting KODAMA plot and clinical

■
RESULTS
Urine metabolic profiles were acquired for a total of 214
women who provided a single sample, of which 38 met the
criteria for control participants and 176 were identified as
LUTS cases. As expected, patients with LUTS were significantly
older and had a higher BMI and parity compared to controls
(Table 1). No difference in the distribution of menopausal
status or ethnicity was detected between LUTS and control
groups.
Variance in the urine metabolic profiles of samples in the
patient and control cohorts was first examined using PCA
(Figure 1A). The first component accounted for 13.6% of the
total variance in the data set with a further 5.9% explained by
the second component. No clear separation between cases and
controls was observed. Supervised PLS analysis was then
performed to identify variance in the metabolic profiles
associated with LUTS; however, the resulting model did not
demonstrate any clear, robust discrimination between LUTS
cases and controls (R2Y = 0.1 [0.1−0.1], Q2Y = 0.06 [−0.07−
0.05], p = 0.09; Figure 1B). We next applied PLS regression to
clinical and questionnaire data in an attempt to identify
whether metabolic profiles could be associated with specific
symptoms and patient characteristics (Table 2). A significant
4211
and questionnaire data was then examined using the K-test.
This involved calculating the coefficient of determination (R2)
to assess proportion of the variance in the dependent variable
(i.e., the relative position of a sample in the KODAMA plot)
that is predictable from the independent variable (i.e., the given
clinical parameter). A permutation test was then performed by
randomly sampling the value of the labels to estimate the
significance of the observed association. In total, six clinical
parameters (BMI, parity, frequency, straining, storage score,
and OAB status) were shown to be significantly correlated with
the distribution of metabolite profiles within the KODAMA
plot (Table 3). To further investigate these relationships, we
quantified 22 metabolites and correlated their relative
concentrations to values within each of these six clinical
parameters. BMI was found to be negatively associated with
formate (rho = −0.19, p-value = 8.05 × 10−3; FDR = 4.43 ×
−10−2) and positively associated with tyrosine (rh0 = 0.27, p-
value = 1.00 × 10−3; FDR = 1.10 × 10−2), phosphocreatine
(rho = 0.25, p-value = 4.75 × 10−4; FDR = 1.05 × 10−2), 2-
ketoisovalerate (rho = 0.23, p-value = 7.31 × 10−3; FDR = 4.43
× 10−2), and valine (rho = 0.21, p-value = 1.77 × 10−2; FDR =
7.79 × 10−2) (Figure 2, Table S2). Metabolite over-
representation analysis of these metabolites identified enrich-
ment of protein biosynthesis (p-value = 4.84 × 10−3, FDR =
3.87 × 10−1) and branched chain amino acid (valine, leucine,
DOI: 10.1021/acs.jproteome.7b00568
J. Proteome Res. 2017, 16, 4208−4216
Journal of Proteome Research Article

Table 3. KODAMA (K-Test) Modelsa whereas frequency (rho = −0.23, p-value = 8.81 × 10−4; FDR =
1.94 × 10−2) and OAB (rho = −0.22, p = 1.38 × 10−3; FDR =
K-test
3.03 × 10−2) were both negatively associated with creatinine
p-value FDR levels. An inverse relationship between straining and trimethyl-
Clinical Features amine-N-oxide levels was observed (rho = −0.19, p = 5.96 ×
age 0.735 0.802 10−3; FDR = 1.31 × 10−1). No significant associations between
BMI 0.004 0.032 quantified metabolites and storage score were identified.
parity 0.028 0.139 PAM clustering was then performed on the KODAMA
menopausal status 0.335 0.702 scores to identify any distinct urine metabotypes within the
Questionnaire LUTS cohort. Four individual clusters or metabotypes were
control versus cases 0.651 0.781 identified using the silhouette median value (Figure S2) as
nocturia 0.351 0.702 presented in Figure 3A. BMI (p-value = 2.61 × 10−2) and
urgency 0.689 0.787 frequency (p = 4.84 × 10−2) were the main clinical parameters
bladder pain 0.175 0.467 associated with metabotype groupings. Specifically, an inverse
frequency 0.000 0.000 relationship between BMI and frequency distribution was
hesitancy 0.584 0.781 observed across the four metabotypes identified (Figure 3B and
straining 0.029 0.139 Figure S3). Metabotype 1 contained the highest proportion of
intermittency 0.806 0.806 patients with a BMI greater than or equal to 25 (51.3%) and
urgency incontinence 0.569 0.781 the lowest proportion of patients experiencing the need to
frequency of incontinence 0.543 0.781 strain when passing urine (18.2%). It also contained the lowest
stress incontinence 0.260 0.624 proportion of OAB patients (25%) and those reporting
insensible incontinence 0.616 0.781 increased urinary frequency of more than 6 times per day
nocturnal enuresis 0.800 0.806 (34.1%). This group was characterized by increased urine levels
storage score 0.046 0.174 of hippurate, phenylacetylglutamine, glucose, threonine, crea-
voiding score 0.501 0.781 tinine, phosphocreatinine, trimethylamine-N-oxide, alanine, 2-
incontinence score 0.618 0.781 ketoisovalerate, and decreased levels of creatine, acetate, lactate,
OAB 0.004 0.032 and isoleucine (Table S3).
a
P-value and FDR are shown. Results are divided into clinical features In contrast, metabotype 4 contained the greatest proportion
and questionnaire data. Significance of model prediction was of patients with increased urine frequency (65.7%) and the
examining using K-test with the reported p-value presented before lowest proportion of patients with BMI ≥ 25 (18.8%). Urine
and after FDR correction. Significant values are shown in bold. BMI, metabolic profiles of this group characteristically contained
parity, frequency, straining, storage score, and OAB were found to higher levels of 2-hydroxyisobutyrate and isoleucine and lower
have a significant impact upon the metabolic profile of the samples concentrations of hippurate, phenylacetylglutamine, tyrosine,
assessed in this study.
and threonine compared to other metabotypes.
Metabotype 2 was characterized by higher levels of tyrosine,
glucose, threonine, creatinine, phosphocreatinine, glycine,
alanine, 2-hydroxyisobutyrate, and 3-hydroxyisovalerate and
lower levels of trigonelline, whereas metabotype 3 contained
higher levels of trigonelline, acetate and lower concentrations of
1-methylnicotinamide, tyrosine, glucose, thronine, creatinine,
phosphocreatinine, trimethylamine-N-oxide, alanine, 2-hydrox-
yisobutyrate, 2-ketoisovalerate, methylsuccinate, and valine
levels. No differences in the distribution of parity or storage
scores were observed across any of the metabotypes.

Figure 2. Heat map of correlations identified between quantified
urinary metabolites and clinical and LUTS questionnaire data.
Metabolites positively associated with features are shown in blue
scale, and those negatively associated are shown in yellow. To assist
visualization, metabolites were ordered according to hierarchical
clustering (shown on left of figure) with significantly altered
metabolites highlighted by red boxes with rho values presented.
and isoleucine) degradation pathways (p = 1.71 × 10−2, FDR =
6.84 × 10−1). Parity was negatively associated with acetate
levels (rho = −0.19, p = 6.84 × 10−3; FDR = 1.51 × 10−1),
4212
■ DISCUSSION
Normal function of the lower urinary tract involves tightly
coordinated signaling between the central and peripheral
nervous system and anatomic components of the lower urinary
tract. Malfunction or malcoordination of any of these pathways
could potential lead to LUTS, which encompass a variety of
symptoms that are phenotypic manifestations of multiple
pathophysiologies. Consistent with the concept that urine
metabolic profiles represent underlying biochemistry of disease
processes, we observed high variance across the urinary
metabolomes of the LUTS patient cohort. However, we failed
to identify any robust metabolic signatures of LUTS compared
to controls using standard supervised modeling of the
metabolic profiles. Taken together with the variation in specific
patient symptoms, this provides further evidence that individual
LUTS is likely to be derived from multiple causal etiologies. For
example, increased urgency has been linked to infection or
inflammation of the urothelium,23 increased excitability and
DOI: 10.1021/acs.jproteome.7b00568
J. Proteome Res. 2017, 16, 4208−4216
Journal of Proteome Research
Figure 3. Identification of LUTS metabotypes and their association with clinical and questionnaire features. (A) PAM clustering of KODAMA
output of LUTS urinary metabolic profiles identified 4 distinct clusters or metabotypes. These were characterized by differences in urinary metabolite
concentrations described in Supplementary Table 3. (B) Analysis of the distributions of clinical features represented in each metabotype revealed
BMI and Frequency to significantly differ between clusters (BMI p-value = 0.0261, frequency p-value = 0.0434, Kruskal−Wallis test).
extracellular coupling of detrusor myocytes,49 and bladder
ischemia.11
We next analyzed the data set using KODAMA, an algorithm
specifically designed for feature extraction from noisy, high-
dimensional data. This approach permitted the identification of
relationships between perturbations of metabolic profiles and
specific clinical characteristics (BMI and parity) and reported
symptoms (frequency, straining, storage score and OAB). In
agreement with a recent large-scale, multicohort study
examining the impact of adiposity on the urinary metab-
olome,50 we observed a negative correlation between BMI and
urinary formate, which is often derived from gut microbial
cometabolism.51 We also observed positive associations
between BMI, tyrosine, and valine, which is consistent with
previous blood metabolic profiling studies.52−54 In addition to
this, we also observed a positive association between BMI and
2-ketoisovalerate, a precursor in valine synthesis. Increased
urine concentrations of branched chain amino acids (BCAAs)
such as valine have been associated with dysregulated BCAA
metabolism that contributes to insulin resistance and glucose
intolerance in obese patients.53,55,56
Our results also identified a relationship between increased
frequency and OAB with lower creatinine levels. Previous
studies have reported lower urinary levels of creatinine in both
human and animal models of type 2 diabetes mellitus (T2DM)
secondary to reduced glomerular filtration rate associated with
neuropathy.57,58 Increased urinary frequency and OAB are
associated with diabetes;59 therefore, these results may indicate
a neuropathic pathway underlying increased frequency and
OAB in some LUTS patients. However, urine levels of
creatinine can also be influenced by other factors including
diet, muscle mass, and BMI.50,60 BMI is normally positively
associated with creatinine excretion.50 In contrast to this, we
observed decreased creatinine in LUTS patients with raised
BMI suggesting that this is uncoupled to our observations.
Analysis of the KODAMA results using PAM clustering
identified four metabotypes, or urinary signatures, to which all
patients in the cohort could be allocated. The biggest difference
in metabolic profiles across these metabotypes was between
metabotypes 1 and 4, which could subsequently be linked to
4213
Article
BMI and frequency. Although metabotype 1 was enriched for
patients with raised BMI and decreased frequency, the opposite
was observed for patients clustered within metabotype 4. This
was surprising considering raised BMI is often associated with
increased prevalence of OAB.61 Urine metabolic profiling may
therefore permit the identification of a subcohort of LUTS
patients that is not associated with BMI. Stratification of
patients using unsupervised, objective assessment of urinary
metabolic profiles may permit targeted treatment strategies that
improve response and efficacy and reduce side effect profiles.
Consistent with this, certain anticholonergic treatements for
OAB have demonstrated a trend for decreasing efficacy in
patients with raised BMI.62
Clustering of patient samples to metabotypes is driven by
differences in relative concentrations of urinary metabolites. As
expected, since metabotype 1 correlated with BMI, many of the
urine metabolites characteristic of metabotype 1 could be
associated with BMI including higher concentrations of
creatine, alanine and glucose.50 Concentrations of hippurate
and isoleucine were inversely correlated between metabotypes
1 and 4. Urine levels of the gut microbial cometabolite
hippurate have been reported in numerous studies as being
inversely correlated to BMI.50,58,63,64 In contrast, we observed a
direct association between hippurate and BMI, which were both
lower in women reporting high frequency (metabotype 4),
whereas hippurate and BMI were relatively higher in those
patients clustered in metabotype 1. This indicates that altered
urine hippurate levels in LUTS patients are not related to BMI
and may instead reflect a previously undescribed relationship
between gut microbiota, diet and LUTS.
Similar to hippurate, we found that women reporting
increased frequency with a low BMI had increased levels of
isoleucine, which again is in contrast to previous studies
reporting a positive association between high BMI and urine
isoleucine levels.50,52,53 Serum metabolic profiling has shown a
strong association between isoleucine and insulin resistance in
lean individuals.65 It has been long established that insulin
resistance is associated with systemic inflammation,66−68 and
therefore, these findings may indicate altered inflammatory
status in this group of patients.
DOI: 10.1021/acs.jproteome.7b00568
J. Proteome Res. 2017, 16, 4208−4216
Journal of Proteome Research
It is important to recognize that our study has a number of
limitations. Urine collection was performed without fasting at
the time of patient visit. While this sampling approach is highly
practical for busy clinical settings and may facilitate the
identification of metabolite biomarkers that are both robust and
resistant to nondisease related augmentation, it can lead to the
introduction of variation associated with diet and food
intake69,70 or diurnal patterns.71−73 This may have in turn
masked more subtle LUTS-associated differences in the urine
metabolic profile. It is also possible that demographic
differences could have impacted on our results. For example
age and BMI are known to independently influence urinary
metabolic profiles,50,60 but are also correlated with LUTS.74−79
Regardless, we were still able to identify numerous urinary
metabolite changes that were associated with LUTS, partic-
ularly frequency, independent of BMI. Future metabolic
profiling studies of LUTS may benefit from detailed assessment
of stratified demographic cohorts (e.g., treatment responder
and nonresponder phenotypes indicative of underlying
etiology) and longitudinal analyses of patients receiving
interventions.
In summary, our study indicates that the multietiological
pathogenesis of individual LUTS is reflected in a highly variable
urinary metabolome. However, urine samples from LUTS
patients can be clustered into metabotypes that are enriched for
patients with specific LUTS (e.g., frequency). Urine metabolites
associated with LUTS can also be identified that are
independent of potential clinical confounders such as BMI.
Urine metabolic profiling and subsequence stratification of
LUTS patients may permit targeted and more personalized
pharmacotherapy leading to improved patient outcome.
■
ASSOCIATED CONTENT
*
S Supporting Information
The Supporting Information is available free of charge on the
ACS Publications website at DOI: 10.1021/acs.jproteo-
me.7b00568.
LUTS definition of individual LUTS as set out by IUGA
and ICS; correlation between metabolite concentrations
and clinical features; differences in metabolites between
clusters; PCA and PLS score plot for all samples
according to age; silhouette median value used to
identify the optimal number (four) of urine clusters or
metabotypes identified by PAM-analysis of KODAMA
output; KODAMA clinical data (PDF)
■
AUTHOR INFORMATION
Corresponding Authors
*E-mail: rhiannon.bray1@nhs.net.
*E-mail: vik.khullar@imperial.ac.uk.
ORCID
Rhiannon Bray: 0000-0002-1346-3069
Elaine Holmes: 0000-0002-0556-8389
Notes
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS
This study was supported by the Imperial College Academic
Health Sciences Centre, the National Institute for Health
Research Biomedical Research Centre, Imperial College
4214
Article
Healthcare National Health Service Trust, Imperial College
London, and the United Kingdom Medical Research Council
(Grant No. G1100377) D.A.M. was supported by a Career
Development Award from the Medical Research Council
(Grant No. MR/L009226/1). The Clinical Phenome Centre
(CPC) was also supported by the National Institute for Health
Research (NIHR) Biomedical Research Centre based at
Imperial College Healthcare NHS Trust and Imperial College
London. The views expressed are those of the authors and not
necessarily those of the NHS, the NIHR or the Department of
Health.
■
REFERENCES
(1) Haylen, B. T.; de Ridder, D.; Freeman, R. M.; Swift, S. E.;
Berghmans, B.; Lee, J.; Monga, A.; Petri, E.; Rizk, D. E.; Sand, P. K.;
Schaer, G. N. An International Urogynecological Association (IUGA)/
International Continence Society (ICS) joint report on the
terminology for female pelvic floor dysfunction. Int. Urogynecol J.
2010, 21 (1), 5−26.
(2) Coyne, K. S.; Wein, A. J.; Tubaro, A.; Sexton, C. C.; Thompson,
C. L.; Kopp, Z. S.; Aiyer, L. P. The burden of lower urinary tract
symptoms: evaluating the effect of LUTS on health-related quality of
life, anxiety and depression: EpiLUTS. BJU Int. 2009, 103 (Suppl 3),
4−11.
(3) Stewart, W. F.; Van Rooyen, J. B.; Cundiff, G. W.; Abrams, P.;
Herzog, A. R.; Corey, R.; Hunt, T. L.; Wein, A. J. Prevalence and
burden of overactive bladder in the United States. World J. Urol. 2003,
20 (6), 327−36.
(4) Hu, T. W.; Wagner, T. H.; Bentkover, J. D.; LeBlanc, K.;
Piancentini, A.; Stewart, W. F.; Corey, R.; Zhou, S. Z.; Hunt, T. L.
Estimated economic costs of overactive bladder in the United States.
Urology 2003, 61 (6), 1123−8.
(5) Khullar, V.; Chapple, C.; Gabriel, Z.; Dooley, J. A. The effects of
antimuscarinics on health-related quality of life in overactive bladder: a
systematic review and meta-analysis. Urology 2006, 68 (2), 38−48.
(6) Roosen, A.; Chapple, C. R.; Dmochowski, R. R.; Fowler, C. J.;
Gratzke, C.; Roehrborn, C. G.; Stief, C. G.; Andersson, K. E. A refocus
on the bladder as the originator of storage lower urinary tract
symptoms: a systematic review of the latest literature. Eur. Urol. 2009,
56 (5), 810−9.
(7) Liu, H. T.; Tyagi, P.; Chancellor, M. B.; Kuo, H. C. Urinary nerve
growth factor but not prostaglandin E2 increases in patients with
interstitial cystitis/bladder pain syndrome and detrusor overactivity.
BJU Int. 2010, 106 (11), 1681−5.
(8) Liu, H. T.; Chen, C. Y.; Kuo, H. C. Urinary nerve growth factor
levels in overactive bladder syndrome and lower urinary tract
disorders. J. Formosan Med. Assoc. 2010, 109 (12), 862−78.
(9) Blaivas, J. G. Overactive bladder and the definition of urgency.
Neurourol. Urodyn. 2007, 26 (6), 757−8.
(10) Madersbacher, H. Overactive bladder: a clinical entity or a
marketing hype? Eur. Urol. 2005, 47 (3), 273−6.
(11) Homma, Y. Lower urinary tract symptomatology: Its definition
and confusion. Int. J. Urol. 2008, 15 (1), 35−43.
(12) Cartwright, R.; Afshan, I.; Derpapas, A.; Vijaya, G.; Khullar, V.
Novel biomarkers for overactive bladder. Nat. Rev. Urol. 2011, 8 (3),
139−45.
(13) Nitti, V. W.; Rovner, E. S.; Bavendam, T. Response to
fesoterodine in patients with an overactive bladder and urgency urinary
incontinence is independent of the urodynamic finding of detrusor
overactivity. BJU Int. 2010, 105 (9), 1268−75.
(14) South, M. M.; Romero, A. A.; Jamison, M. G.; Webster, G. D.;
Amundsen, C. L. Detrusor overactivity does not predict outcome of
sacral neuromodulation test stimulation. Int. Urogynecol J. Pelvic Floor
Dysfunct 2007, 18 (12), 1395−8.
(15) Smith, A.; Bevan, D.; Douglas, H. R.; James, D. Management of
urinary incontinence in women: summary of updated NICE guidance.
BMJ. 2013, 347, f5170.
DOI: 10.1021/acs.jproteome.7b00568
J. Proteome Res. 2017, 16, 4208−4216
Journal of Proteome Research
(16) Lucas, M. G.; Bosch, R. J.; Burkhard, F. C.; Cruz, F.; Madden, T.
B.; Nambiar, A. K.; Neisius, A.; de Ridder, D. J.; Tubaro, A.; Turner,
W. H.; Pickard, R. S.; Urology, E. A. o. EAU guidelines on assessment
and nonsurgical management of urinary incontinence. Eur. Urol. 2012,
62 (6), 1130−42.
(17) Wagg, A.; Compion, G.; Fahey, A.; Siddiqui, E. Persistence with
prescribed antimuscarinic therapy for overactive bladder: a UK
experience. BJU Int. 2012, 110 (11), 1767−74.
(18) Sears, C. L.; Lewis, C.; Noel, K.; Albright, T. S.; Fischer, J. R.
Overactive bladder medication adherence when medication is free to
patients. J. Urol. 2010, 183 (3), 1077−81.
(19) Glazener, C. M.; Herbison, G. P.; MacArthur, C.; Grant, A.;
Wilson, P. D. Randomised controlled trial of conservative management
of postnatal urinary and faecal incontinence: six year follow up. BMJ.
2005, 330 (7487), 337.
(20) Dmochowski, R. R. Duloxetine: a summary of published clinical
experience. Rev. Urol. 2004, 6 (Suppl3), S56−63.
(21) Kuo, H. C.; Liu, H. T.; Chancellor, M. B. Urinary nerve growth
factor is a better biomarker than detrusor wall thickness for the
assessment of overactive bladder with incontinence. Neurourol. Urodyn.
2009, 29 (3), 482−7.
(22) Wang, L. W.; Han, X. M.; Chen, C. H.; Ma, Y.; Hai, B. Urinary
brain-derived neurotrophic factor: a potential biomarker for objective
diagnosis of overactive bladder. Int. Urol. Nephrol. 2014, 46 (2), 341−
7.
(23) Chuang, Y. C.; Tyagi, V.; Liu, R. T.; Chancellor, M. B.; Tyagi, P.
Urine and Serum C-Reactive Protein Levels as Potential Biomarkers of
Lower Urinary Tract Symptoms. Urol. Sci. 2010, 21 (3), 132−6.
(24) Nicholson, J. K.; Wilson, I. D. Opinion: understanding ’global’
systems biology: metabonomics and the continuum of metabolism.
Nat. Rev. Drug Discovery 2003, 2 (8), 668−76.
(25) Nicholson, J. K.; Lindon, J. C.; Holmes, E. ’Metabonomics’:
understanding the metabolic responses of living systems to
pathophysiological stimuli via multivariate statistical analysis of
biological NMR spectroscopic data. Xenobiotica 1999, 29 (11),
1181−9.
(26) Cacciatore, S.; Zadra, G.; Bango, C.; Penney, K. L.; Tyekucheva,
S.; Yanes, O.; Loda, M. Metabolic Profiling in Formalin-Fixed and
Paraffin-Embedded Prostate Cancer Tissues. Mol. Cancer Res. 2017, 15
(4), 439−447.
(27) Bertini, I.; Cacciatore, S.; Jensen, B. V.; Schou, J. V.; Johansen, J.
S.; Kruhøffer, M.; Luchinat, C.; Nielsen, D. L.; Turano, P.
Metabolomic NMR fingerprinting to identify and predict survival of
patients with metastatic colorectal cancer. Cancer Res. 2012, 72 (1),
356−64.
(28) MacIntyre, D. A.; Jimenez, B.; Lewintre, E. J.; Martin, C. R.;
Schafer, H.; Ballesteros, C. G.; Mayans, J. R.; Spraul, M.; Garcia-
Conde, J.; Pineda-Lucena, A. Serum metabolome analysis by 1H-NMR
reveals differences between chronic lymphocytic leukaemia molecular
subgroups. Leukemia 2010, 24 (4), 788−97.
(29) Spratlin, J. L.; Serkova, N. J.; Eckhardt, S. G. Clinical
applications of metabolomics in oncology: a review. Clin. Cancer Res.
2009, 15 (2), 431−40.
(30) Brindle, J. T.; Antti, H.; Holmes, E.; Tranter, G.; Nicholson, J.
K.; Bethell, H. W.; Clarke, S.; Schofield, P. M.; McKilligin, E.;
Mosedale, D. E.; Grainger, D. J. Rapid and noninvasive diagnosis of
the presence and severity of coronary heart disease using 1H-NMR-
based metabonomics. Nat. Med. 2002, 8 (12), 1439−44.
(31) Kaddurah-Daouk, R.; Krishnan, K. R. Metabolomics: a global
biochemical approach to the study of central nervous system diseases.
Neuropsychopharmacology 2009, 34 (1), 173−86.
(32) Brookes, S. T.; Donovan, J. L.; Wright, M.; Jackson, S.; Abrams,
P. A scored form of the Bristol Female Lower Urinary Tract
Symptoms questionnaire: data from a randomized controlled trial of
surgery for women with stress incontinence. Am. J. Obstet. Gynecol.
2004, 191 (1), 73−82.
(33) Jackson, S.; Donovan, J.; Brookes, S.; Eckford, S.; Swithinbank,
L.; Abrams, P. The Bristol Female Lower Urinary Tract Symptoms
4215
Article
questionnaire: development and psychometric testing. BJU Int. 1996,
77 (6), 805−12.
(34) Dona, A. C.; Jiménez, B.; Schäfer, H.; Humpfer, E.; Spraul, M.;
Lewis, M. R.; Pearce, J. T.; Holmes, E.; Lindon, J. C.; Nicholson, J. K.
Precision high-throughput proton NMR spectroscopy of human urine,
serum, and plasma for large-scale metabolic phenotyping. Anal. Chem.
2014, 86 (19), 9887−94.
(35) Dieterle, F.; Ross, A.; Schlotterbeck, G.; Senn, H. Probabilistic
quotient normalization as robust method to account for dilution of
complex biological mixtures. Application in 1H NMR metabonomics.
Anal. Chem. 2006, 78 (13), 4281−90.
(36) Wishart, D. S.; Jewison, T.; Guo, A. C.; Wilson, M.; Knox, C.;
Liu, Y.; Djoumbou, Y.; Mandal, R.; Aziat, F.; Dong, E.; Bouatra, S.;
Sinelnikov, I.; Arndt, D.; Xia, J.; Liu, P.; Yallou, F.; Bjorndahl, T.;
Perez-Pineiro, R.; Eisner, R.; Allen, F.; Neveu, V.; Greiner, R.; Scalbert,
A. HMDB 3.0–The Human Metabolome Database in 2013. Nucleic
Acids Res. 2013, 41 (D1), D801−7.
(37) Cloarec, O.; Dumas, M. E.; Craig, A.; Barton, R. H.; Trygg, J.;
Hudson, J.; Blancher, C.; Gauguier, D.; Lindon, J. C.; Holmes, E.;
Nicholson, J. Statistical total correlation spectroscopy: an exploratory
approach for latent biomarker identification from metabolic 1H NMR
data sets. Anal. Chem. 2005, 77 (5), 1282−9.
(38) van den Berg, R. A.; Hoefsloot, H. C.; Westerhuis, J. A.; Smilde,
A. K.; van der Werf, M. J. Centering, scaling, and transformations:
improving the biological information content of metabolomics data.
BMC Genomics 2006, 7, 142.
(39) Eriksson, L.; Jaworska, J.; Worth, A. P.; Cronin, M. T.;
McDowell, R. M.; Gramatica, P. Methods for reliability and
uncertainty assessment and for applicability evaluations of classi-
fication- and regression-based QSARs. Environ. Health Perspect 2003,
111 (10), 1361−75.
(40) Mukherjee, S.; Tamayo, P.; Rogers, S.; Rifkin, R.; Engle, A.;
Campbell, C.; Golub, T. R.; Mesirov, J. P. Estimating dataset size
requirements for classifying DNA microarray data. J. Comput. Biol.
2003, 10 (2), 119−42.
(41) Szymańska, E.; Saccenti, E.; Smilde, A. K.; Westerhuis, J. A.
Double-check: validation of diagnostic statistics for PLS-DA models in
metabolomics studies. Metabolomics 2012, 8 (S1), 3−16.
(42) Cacciatore, S.; Luchinat, C.; Tenori, L. Knowledge discovery by
accuracy maximization. Proc. Natl. Acad. Sci. U. S. A. 2014, 111 (14),
5117−22.
(43) Cacciatore, S.; Tenori, L.; Luchinat, C.; Bennett, P. R.;
MacIntyre, D. A. KODAMA: an R package for knowledge discovery
and data mining. Bioinformatics 2016, btw705.
(44) Cameron, A. C.; Windmeijer, F. A. G. An R-squared measure of
goodness of fit for some common nonlinear regression models. Journal
of Econometrics 1997, 77 (2), 329−342.
(45) Reynolds, A. P.; Richards, G.; de la Iglesia, B.; Rayward-Smith,
V. J. Clustering Rules: A Comparison of Partitioning and Hierarchical
Clustering Algorithms. Journal of Mathematical Modelling and
Algorithms 2006, 5 (4), 475−504.
(46) Rousseeuw, P. J. Silhouettes: a graphical aid to the interpretation
and validation of cluster analysis. Journal of computational and applied
mathematics 1987, 20, 53−65.
(47) Xia, J.; Sinelnikov, I.; Han, B.; Wishart, D. MetaboAnalyst 3.0-
making metabolomics more meaningful. Nucleic Acids Res. 2015, 43
(W1), W251−W257.
(48) Storey, J. D. A direct approach to false discovery rates. Journal of
the Royal Statistical Society: Series B (Statistical Methodology) 2002, 64
(3), 479−498.
(49) Brading, A. F. A myogenic basis for the overactive bladder.
Urology 1997, 50 (6), 57−67.
(50) Elliott, P.; Posma, J. M.; Chan, Q.; Garcia-Perez, I.;
Wijeyesekera, A.; Bictash, M.; Ebbels, T. M.; Ueshima, H.; Zhao, L.;
van Horn, L.; Daviglus, M.; Stamler, J.; Holmes, E.; Nicholson, J. K.
Urinary metabolic signatures of human adiposity. Sci. Transl. Med.
2015, 7 (285), 285ra62.
(51) Marchesi, J. R.; Holmes, E.; Khan, F.; Kochhar, S.; Scanlan, P.;
Shanahan, F.; Wilson, I. D.; Wang, Y. Rapid and noninvasive
DOI: 10.1021/acs.jproteome.7b00568
J. Proteome Res. 2017, 16, 4208−4216
Journal of Proteome Research
metabonomic characterization of inflammatory bowel disease. J.
Proteome Res. 2007, 6 (2), 546−51.
(52) Cheng, S.; Rhee, E. P.; Larson, M. G.; Lewis, G. D.; McCabe, E.
L.; Shen, D.; Palma, M. J.; Roberts, L. D.; Dejam, A.; Souza, A. L.;
Deik, A. A.; Magnusson, M.; Fox, C. S.; O’Donnell, C. J.; Vasan, R. S.;
Melander, O.; Clish, C. B.; Gerszten, R. E.; Wang, T. J. Metabolite
profiling identifies pathways associated with metabolic risk in humans.
Circulation 2012, 125 (18), 2222−31.
(53) Newgard, C. B.; An, J.; Bain, J. R.; Muehlbauer, M. J.; Stevens,
R. D.; Lien, L. F.; Haqq, A. M.; Shah, S. H.; Arlotto, M.; Slentz, C. A.;
Rochon, J.; Gallup, D.; Ilkayeva, O.; Wenner, B. R.; Yancy, W. S., Jr.;
Eisenson, H.; Musante, G.; Surwit, R. S.; Millington, D. S.; Butler, M.
D.; Svetkey, L. P. A branched-chain amino acid-related metabolic
signature that differentiates obese and lean humans and contributes to
insulin resistance. Cell Metab. 2009, 9 (4), 311−26.
(54) Xie, G.; Ma, X.; Zhao, A.; Wang, C.; Zhang, Y.; Nieman, D.;
Nicholson, J. K.; Jia, W.; Bao, Y.; Jia, W. The metabolite profiles of the
obese population are gender-dependent. J. Proteome Res. 2014, 13 (9),
4062−73.
(55) Tremblay, F.; Brule, S.; Um, S. H.; Li, Y.; Masuda, K.; Roden,
M.; Sun, X. J.; Krebs, M.; Polakiewicz, R. D.; Thomas, G.; Marette, A.
Identification of IRS-1 Ser-1101 as a target of S6K1 in nutrient- and
obesity-induced insulin resistance. Proc. Natl. Acad. Sci. U. S. A. 2007,
104 (35), 14056−14061.
(56) Um, S. H.; D’Alessio, D.; Thomas, G. Nutrient overload, insulin
resistance, and ribosomal protein S6 kinase 1, S6K1. Cell Metab. 2006,
3 (6), 393−402.
(57) Turner, N. C.; Morgan, P. J.; Haynes, A. C.; Vidgeon-Hart, M.;
Toseland, N.; Clapham, J. C. Elevated renal endothelin-I clearance and
mRNA levels associated with albuminuria and nephropathy in non-
insulin-dependent diabetes mellitus: studies in obese fa/fa Zucker rats.
Clin. Sci. 1997, 93 (6), 565−71.
(58) Salek, R. M.; Maguire, M. L.; Bentley, E.; Rubtsov, D. V.;
Hough, T.; Cheeseman, M.; Nunez, D.; Sweatman, B. C.; Haselden, J.
N.; Cox, R. D.; Connor, S. C.; Griffin, J. L. A metabolomic comparison
of urinary changes in type 2 diabetes in mouse, rat, and human. Physiol.
Genomics 2007, 29 (2), 99−108.
(59) Golbidi, S.; Laher, I. Bladder dysfunction in diabetes mellitus.
Front. Pharmacol. 2010, 1, 136.
(60) Swann, J. R.; Spagou, K.; Lewis, M.; Nicholson, J. K.; Glei, D. A.;
Seeman, T. E.; Coe, C. L.; Goldman, N.; Ryff, C. D.; Weinstein, M.;
Holmes, E. Microbial-mammalian cometabolites dominate the age-
associated urinary metabolic phenotype in Taiwanese and American
populations. J. Proteome Res. 2013, 12 (7), 3166−80.
(61) Hunskaar, S. A systematic review of overweight and obesity as
risk factors and targets for clinical intervention for urinary
incontinence in women. Neurourol. Urodyn. 2008, 27 (8), 749−57.
(62) Cardozo, L.; Herschorn, S.; Snijder, R.; Siddiqui, E.; Chapple, C.
R. Does BMI, gender or age affect efficacy/tolerability of solifenacin in
the management of overactive bladder? Int. Urogynecol J. 2017, 28 (3),
477−488.
(63) Calvani, R.; Miccheli, A.; Capuani, G.; Tomassini Miccheli, A.;
Puccetti, C.; Delfini, M.; Iaconelli, A.; Nanni, G.; Mingrone, G. Gut
microbiome-derived metabolites characterize a peculiar obese urinary
metabotype. Int. J. Obes. 2010, 34 (6), 1095−8.
(64) Lees, H. J.; Swann, J. R.; Wilson, I. D.; Nicholson, J. K.; Holmes,
E. Hippurate: the natural history of a mammalian-microbial
cometabolite. J. Proteome Res. 2013, 12 (4), 1527−46.
(65) Ho, J. E.; Larson, M. G.; Ghorbani, A.; Cheng, S.; Chen, M. H.;
Keyes, M.; Rhee, E. P.; Clish, C. B.; Vasan, R. S.; Gerszten, R. E.;
Wang, T. J. Metabolomic Profiles of Body Mass Index in the
Framingham Heart Study Reveal Distinct Cardiometabolic Pheno-
types. PLoS One 2016, 11 (2), e0148361.
(66) Lin, S. Y.; Sheu, W. H. An emerging link between insulin
resistance and inflammation. J. Chin. Med. Assoc. 2006, 69 (6), 245−7.
(67) Shoelson, S. E.; Lee, J.; Goldfine, A. B. Inflammation and insulin
resistance. J. Clin. Invest. 2006, 116 (7), 1793−801.
4216
Article
(68) Ndumele, C. E.; Pradhan, A. D.; Ridker, P. M. Interrelationships
between inflammation, C-reactive protein, and insulin resistance. J.
CardioMetab. Syndr. 2006, 1 (3), 190−6.
(69) Rubio-Aliaga, I.; de Roos, B.; Duthie, S. J.; Crosley, L. K.; Mayer,
C.; Horgan, G.; Colquhoun, I. J.; Le Gall, G.; Huber, F.; Kremer, W.;
Rychlik, M.; Wopereis, S.; van Ommen, B.; Schmidt, G.; Heim, C.;
Bouwman, F. G.; Mariman, E. C.; Mulholland, F.; Johnson, I. T.;
Polley, A. C.; Elliott, R. M.; Daniel, H. Metabolomics of prolonged
fasting in humans reveals new catabolic markers. Metabolomics 2011, 7
(3), 375−387.
(70) Yin, P. Y.; Lehmann, R.; Xu, G. W. Effects of pre-analytical
processes on blood samples used in metabolomics studies. Anal.
Bioanal. Chem. 2015, 407 (17), 4879−4892.
(71) Maher, A. D.; Zirah, S. F.; Holmes, E.; Nicholson, J. K.
Experimental and analytical variation in human urine in 1H NMR
spectroscopy-based metabolic phenotyping studies. Anal. Chem. 2007,
79 (14), 5204−11.
(72) Saude, E. J.; Adamko, D.; Rowe, B. H.; Marrie, T.; Sykes, B. D.
Variation of metabolites in normal human urine. Metabolomics 2007, 3
(4), 439−451.
(73) Slupsky, C. M.; Rankin, K. N.; Wagner, J.; Fu, H.; Chang, D.;
Weljie, A. M.; Saude, E. J.; Lix, B.; Adamko, D. J.; Shah, S.; Greiner, R.;
Sykes, B. D.; Marrie, T. J. Investigations of the effects of gender,
diurnal variation, and age in human urinary metabolomic profiles.
Anal. Chem. 2007, 79 (18), 6995−7004.
(74) Milsom, I. The prevalence of urinary incontinence. Acta Obstet.
Gynecol. Scand. 2000, 79 (12), 1056−9.
(75) Milsom, I.; Stewart, W.; Thuroff, J. The prevalence of overactive
bladder. Am. J. Manag Care 2000, 6 (11 Suppl), S565−73.
(76) Irwin, D. E.; Milsom, I.; Hunskaar, S.; Reilly, K.; Kopp, Z.;
Herschorn, S.; Coyne, K.; Kelleher, C.; Hampel, C.; Artibani, W.;
Abrams, P. Population-based survey of urinary incontinence, over-
active bladder, and other lower urinary tract symptoms in five
countries: results of the EPIC study. Eur. Urol. 2006, 50 (6), 1306−14.
(77) Rortveit, G.; Hannestad, Y. S.; Daltveit, A. K.; Hunskaar, S. Age-
and type-dependent effects of parity on urinary incontinence: the
Norwegian EPINCONT study. Obstet. Gynecol. 2001, 98 (6), 1004−
10.
(78) Khullar, V.; Sexton, C. C.; Thompson, C. L.; Milsom, I.; Bitoun,
C. E.; Coyne, K. S. The relationship between BMI and urinary
incontinence subgroups: results from EpiLUTS. Neurourol. Urodyn.
2014, 33 (4), 392−9.
(79) Bunn, F.; Kirby, M.; Pinkney, E.; Cardozo, L.; Chapple, C.;
Chester, K.; Cruz, F.; Haab, F.; Kelleher, C.; Milsom, I.; Sievart, K. D.;
Tubaro, A.; Wagg, A. Is there a link between overactive bladder and
the metabolic syndrome in women? A systematic review of
observational studies. Int. J. Clin Pract 2015, 69 (2), 199−217.
DOI: 10.1021/acs.jproteome.7b00568
J. Proteome Res. 2017, 16, 4208−4216