Urinary Metabolic Phenotyping of Women with Lower Urinary Tract

Subscriber access provided by University of Sussex Library

Article

Urinary metabolic phenotyping of women with lower urinary tract symptoms Rhiannon Bray, Stefano Cacciatore, Beatriz Jimenez, Rufus Cartwright, Alex Digesu, Ruwan Fernando, Elaine Holmes, Jeremy Kirk Nicholson, Phillip R. Bennett, David A. MacIntyre, and Vik Khullar J. Proteome Res., Just Accepted Manuscript • DOI: 10.1021/acs.jproteome.7b00568 • Publication Date (Web): 22 Sep 2017 Downloaded from http://pubs.acs.org on September 23, 2017

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Journal of Proteome Research is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 29

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Journal of Proteome Research

Urinary metabolic phenotyping of women with lower urinary tract symptoms Rhiannon Bray1*, Stefano Cacciatore2, Beatriz Jiménez3, Rufus Cartwright1, Alex Digesu1, Ruwan Fernando1, Elaine Holmes4,3, Jeremy K. Nicholson4,3, Phillip R. Bennett2, 5, David A. MacIntyre2 and Vik Khullar1* 1: Department of Urogynaecology, St. Mary’s Hospital, Imperial College Healthcare NHS Trust, London 2: Imperial College Parturition Research Group, Division of the Institute of Reproductive and Developmental Biology, Imperial College London, London 3: Division of Computational Systems Medicine, Department of Surgery and Cancer, Imperial College London, London 4: Centre for Digestive and Gut Health, Imperial College London, London 5: Queen Charlotte’s Hospital, Imperial College Healthcare NHS Trust, London *

Corresponding authors: R.B.: E-mail: [email protected]; V.K.: [email protected] Department of Urogynaecology, St Mary's Hospital, Praed St, London W2 1NY, UK.

ACS Paragon Plus Environment

0


Page 2 of 29

Abstract: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

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 urogynaecology 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 aetiology, which may permit stratification of patient populations and the design of more personalized treatment strategies.

Keywords: Up to 10 max Metabolic profiling, LUTS, Overactive Bladder, NMR, KODAMA, metabolomics


1

Page 3 of 29


INTRODUCTION 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

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 incontinence). (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 1 in 6 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 characterised, 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) Functional 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 recommends that urodynamic testing should not be used in women prior to the use of conservative treatments. (15, 16)


2


Page 4 of 29

Treatment strategies for LUTS are often hindered by side effects and poor patient 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

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, non-invasive 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 measurement 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) spectroscopy 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 non-destructive 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 (2629), 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.


3

Page 5 of 29


METHODS 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Patient cohort and classification This study was conducted following NHS Health Research Authority National Research Ethics Service (NRES) Committee Approval (REC 12/LO/0394). Women attending specialist urogynaecology and general gynecology clinics were recruited following informed consent. Patients provided a mid-stream urine sample under non-fasting 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 (ICIQFLUTS 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 urinary tract infection, and v) receiving anticholinergic medication during the study. Ethnicity was self-reported as White (European ancestry), Asian (Pakistani, Indian, Bangladeshi or Sri Lankan ancestry), Arabic, Black (African or African-Caribbean ancestry) or Mixed. Parity, menopausal status as well as height and weight were self-reported with BMI subsequently calculated.


4


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 6 of 29

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 standardised and optimised 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 13000 g 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 centred at 4.75 ppm. The relaxation delay was set at 4 s and a water pre-saturation 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 transformation, 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 (0.00pm). 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 ACS Paragon Plus Environment

5

Page 7 of 29


removed from the subsequent analysis. Each spectrum in the range between 0.2 and 10.0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

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 undertaken 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 inhouse 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 concentrations were calculated by integrating the reference signals in the raw spectra before normalizing to the PQN coefficients 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.


6


Page 8 of 29

Regression of metabolic profiles against questionnaire responses and clinical features was 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

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 backpredicting 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 10fold 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 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 dataset, 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) clustering (45) was applied to the KODAMA scores with the silhouette algorithm (46) 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 distribution 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 ACS Paragon Plus Environment

7

Page 9 of 29


used to calculate the correlation coefficients (rho) between the metabolite concentrations 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

and the clinical features. Hierarchical clustering (Ward’s linkage) was used to order the heatmap showing the different metabolite concentrations across the clinical and questionnaire features. Metabolite overrepresentation 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

Urinary Metabolic Phenotyping of Women with Lower Urinary Tract

Recommend Documents