Mass Spectrometry-based Metabolomics for the Discovery of

Feb 20, 2013 - (28) For the present study, we generated an in-house customized ..... Joshipura , K. J.; Ascherio , A.; Manson , J. E.; Stampfer , M. J...
0 downloads 0 Views 5MB Size
Article pubs.acs.org/jpr

Mass Spectrometry-based Metabolomics for the Discovery of Biomarkers of Fruit and Vegetable Intake: Citrus Fruit as a Case Study Estelle Pujos-Guillot,†,‡ Jane Hubert,†,‡ Jean-François Martin,†,‡ Bernard Lyan,†,‡ Mercedes Quintana,†,‡ Sylvain Claude,†,‡ Bruno Chabanas,†,‡ Joseph A. Rothwell,†,‡ Catherine Bennetau-Pelissero,§,∥ Augustin Scalbert,⊥ Blandine Comte,†,‡ Serge Hercberg,# Christine Morand,†,‡ Pilar Galan,# and Claudine Manach*,†,‡ †

INRA, UMR 1019, UNH, CRNH Auvergne, F-63000 Clermont-Ferrand, France Clermont University, Université d’Auvergne, Unité de Nutrition Humaine, BP 10448, F-63000 Clermont-Ferrand, France § Université Bordeaux, Physiopathologie de la plasticité neuronale, U862, F-33000 Bordeaux, France ∥ INSERM, U862, F-33000 Bordeaux, France ⊥ International Agency for Research on Cancer (IARC), Nutrition and Metabolism Section, Biomarkers Group, F-69372 Lyon, France # Research Unit on Nutritional Epidemiology, Université Paris 13, Sorbonne Paris cité, INSERM U557, INRA U1125, CNAM, F-93017 Bobigny, France ‡

S Supporting Information *

ABSTRACT: Elucidation of the relationships between genotype, diet, and health requires accurate dietary assessment. In intervention and epidemiological studies, dietary assessment usually relies on questionnaires, which are susceptible to recall bias. An alternative approach is to quantify biomarkers of intake in biofluids, but few such markers have been validated so far. Here we describe the use of metabolomics for the discovery of nutritional biomarkers, using citrus fruits as a case study. Three study designs were compared. Urinary metabolomes were profiled for volunteers that had (a) consumed an acute dose of orange or grapefruit juice, (b) consumed orange juice regularly for one month, and (c) reported high or low consumption of citrus products for a large cohort study. Some signals were found to reflect citrus consumption in all three studies. Proline betaine and flavanone glucuronides were identified as known biomarkers, but various other biomarkers were revealed. Further, many signals that increased after citrus intake in the acute study were not sensitive enough to discriminate high and low citrus consumers in the cohort study. We propose that urine profiling of cohort subjects stratified by consumption is an effective strategy for discovery of sensitive biomarkers of consumption for a wide range of foods. KEYWORDS: metabolomics, dietary assessment, biomarkers of intake, phytochemicals, fruits and vegetables, citrus fruits



INTRODUCTION The ability to read, in a plasma or urine sample, the dietary habits of individuals, their recent food intake, and how they deal physiologically with ingested energy, proteins, micronutrients, bioactives or contaminants, has become an achievable goal. Recent technological advances in mass spectrometry (MS)-based metabolomics have enabled the simultaneous measurement of hundreds of small molecules in biological samples, providing information-rich profiles.1 Human diets provide hundreds of phytochemicals, food additives and contaminants, and numerous studies have demonstrated that even low exposure to some of these compounds (polyphenols, endocrine disruptors, etc.) may have a significant impact, beneficial or detrimental, on human health. Therefore, the era © 2013 American Chemical Society

in which assessment of the intakes of 40 macro- and micronutrients was considered sufficient to describe dietary exposure is ending and a more comprehensive phenotyping is now required. Further, controlled intervention studies conducted in the last 20 years have often described interindividual variation in response to a nutritional challenge, due to differences in gene polymorphisms, epigenetic patterns, enterotypes, or lifestyle.2,3 These observations brought about the need for large-scale prospective studies on multiethnic cohorts with genotyping and extensive phenotyping of individuals based on biochemical, functional, imaging and omic measurements to Received: October 23, 2012 Published: February 20, 2013 1645

dx.doi.org/10.1021/pr300997c | J. Proteome Res. 2013, 12, 1645−1659

Journal of Proteome Research

Article

examination. The exclusion criteria were as follows: use of medication, use of antioxidant or vitamin supplements, smoking, high alcohol consumption, intestinal disorders, recent surgery, hyperlipemia, intense physical activity (>5 h/week), vegetarian or vegan diet. Volunteers received a single 600 mL dose of orange juice (OR group), grapefruit juice (GR group) or control drink (CO group) in 3 experimental periods separated by 6-day wash-out periods. The citrus juices were Andros brand commercial pasteurized pure juices. The control drink (PYC laboratories, Aix-en-Provence, France) was composed of water with 45 g/L saccharose, 26 g/L fructose, 23 g/L dextrose, 1.9 g/L K, 0.1 g/L Mg, 0.09 g/L Na, 9 g/L citrate, 0.7 g pectins, 380 mg/L vitamin C, and 370 mg/L vitamin B9, providing approximately the same energy supply (240 kcal for 600 mL) as the orange juice. Participants were instructed to consume polyphenol-free diets 24h before the experimental juice intake, and on the days of sample collection. They also received a list of all citrus derived-foods to be avoided (fruit, juice, jelly, ice-cream, etc.) for at least 3 days before each juice challenge. The composition of all polyphenol-free meals taken the day before the juice challenges was specified by the investigators. The volunteers collected the first void urine at home, and arrived at the experimental unit at 7.45 a.m. in a fasted state. At 8 a.m. they consumed one of the experimental drinks with a controlled breakfast composed of ham, white bread and butter. At 1.30 p.m. they consumed a standardized low-polyphenol meal (instant mashed potatoes, butter, chopped steak, white bread, cottage cheese, sugar). In the evening they ate a meal at home whose composition (pasta, butter, chicken, white bread, emmental cheese, yogurt and sugar) was specified by the investigators. The 24-h urines were collected in 5 fractions: (F1 = first void urine before juice intake, F2 = 0−6 h after juice intake, F3 = 6−12 h, F4 = any urine collected after 12 h and during the night, F5 = first void urine on day 2). Volumes were recorded and aliquots stored at −80 °C. The experimental protocol was approved by the Clermont-Ferrand Ethics Committee (N° AU688, ID RCB 2006-A00628−43) and performed according to the principles stated in the Declaration of Helsinki, as revised in 2000. Written informed consent was obtained from each participant. MTI Design. Twenty-four slightly overweight men (BMI 27.4 ± 0.3 kg/m2, aged 56 ± 1 y) participated in a controlled randomized crossover intervention trial. Each consumed 500 mL/day of orange juice (OJ) or 500 mL/day of control drink (CO) in addition to their usual diet for a duration of 4 weeks. Details of the study design have been published elsewhere.15 Orange juice from concentrate was provided by Florida department of Citrus (Lake Alfred, FL). The control drink was composed of 45 g/L carbohydrates including 50% sucrose, 25% glucose and 25% fructose. At the end of each intervention period, 24 h urines were collected. Of the 24 subjects enrolled in the study, twelve were chosen to participate in the present metabolomics study, and only the urine samples collected at the end of CO and OJ 4-weeks supplementation periods were analyzed. The study was approved by the French Human Ethics Committee of the South East VI, and is registered at clinicaltrials.gov as NCT00983086. CHT Design. Data and samples were taken from the SUpplémentation en VItamines Minéraux et AntioXidants (SU.VI.MAX) cohort, a large sample of middle-aged adults living in France. SU.VI.MAX (1994−2002) was a randomized double-blind, placebo-controlled, primary prevention trial that included a total of 12741 individuals to test the potential

better understand genotype−diet−health relationships. While considerable progress has been made in the genotyping and phenotyping of individuals, dietary assessment methods offer only limited accuracy and coverage.4,5 Subjects generally report their own food intakes via food diaries, 24-h dietary recalls and food frequency questionnaires (FFQs), whose limitations have long been recognized. First, recall errors and misreporting make questionnaires imprecise and inappropriate for some populations (e.g., the elderly, children and obese people). Second, subjects tend to change their dietary habits when they must record their food consumption. In addition FFQs do not assess portion sizes well and are unable to capture a large spectrum of food diversity. Indeed, much of the inconsistency in nutritional epidemiology findings may be due to inaccurate dietary assessment.6,7 This is especially true for fruit and vegetables (F&V), whose consumption is often overestimated by dietary questionnaires.8 A complementary approach to questionnaires is the use of biomarkers, which indicates physiological exposure to nutritional factors and indirectly reflects food intake. Biomarkers have been rather underused so far and very few have been successfully validated and applied in nutritional epidemiology.9 The classic strategy for biomarker identification is based on prior knowledge of a specific food constituent and of its metabolism and pharmacokinetics after food ingestion. However, this approach has been unable to supply the whole set of markers necessary to reflect the diversity of foods consumed by humans.9 Metabolomics offers a conceptual breakthrough for biomarker discovery. Each food, especially those of plant origin, typically contains hundreds of components that together form a specific chemical signature. All food constituents are absorbed and metabolized to a variety of phase I, phase II and microbial metabolites. Together, these constitute what has been dubbed the food metabolome.10 The metabolomic approach exploits the complexity of food composition and knowledge of the fate of food components in the human body to identify specific signatures that reflect the consumption of a food. Citrus fruits are among the most widely consumed fruits in the world and represent a rich source of micronutrients and bioactive compounds. Their consumption has been associated with a lower risk of acute coronary events and stroke.11 Plasma vitamin C and carotenoids have been used as markers of citrus intake, although they are not specific to citrus fruits and have even been used for assessing total F&V intake.12 Recently, proline betaine has been proposed as a robust biomarker of citrus intake.13,14 The aim of the present work was to use citrus fruits as a case study to investigate the ability of MS-based metabolomics to identify the most sensitive and specific biomarkers of intake. To determine the best strategy for biomarker discovery, metabolomic profiling was applied to three complementary studies with different designs. STI was a short-term intervention study on a small number of volunteers with a strictly controlled diet, MTI was a medium-term, intervention study on a larger group of volunteers consuming orange juice as part of their usual diet, and CHT was a cohort study.



EXPERIMENTAL SECTION

Design of Human Studies

STI Design. Four healthy male volunteers (aged 33 ± 7 y, BMI 22.8 ± 2.3 kg/m2) participated in the study after clinical 1646

dx.doi.org/10.1021/pr300997c | J. Proteome Res. 2013, 12, 1645−1659

Journal of Proteome Research

Article

Data Processing

efficacy of daily supplementation with antioxidant vitamins and minerals on the incidence of cancers and ischemic heart disease.16 Of these subjects, 6850 individuals then participated in the SU.VI.MAX 2 study (2007−2009), which investigated the effect of nutrition on the quality of aging. SU.VI.MAX and SU.VI.MAX 2 were conducted according to the guidelines of the Declaration of Helsinki and were approved by the Ethical Committee for Studies with Human Subjects of Paris-Cochin Hospital (CCPPRB No 706 and 2364, respectively) and the Comité National Informatique et Liberté (No 334641 and 907094, respectively). During the SU.VI.MAX study, subjects were asked to complete a 24 h record every 2 months so that all days of the week and all seasons were covered. An instruction manual for coding food portions, that included validated photographs of more than 250 foods, was issued to all participants. SU.VI.MAX2 male volunteers were stratified into quintiles according to their usual intake of citrus fruits (sum of orange, mandarin, clementine, lemon and grapefruit intake) using dietary questionnaire data collected between 1994 and 2009. Two groups of 40 subjects were randomly selected from the lowest and highest quintiles of citrus consumers. One spot urine sample per subject, collected in the fasted state between 2007 and 2008 and stored at −80 °C in the SU.VI.MAX biobank, was used for metabolomic profiling.

The preprocessing treatments were identical for the studies STI, MTI and CHT. The raw data were transformed to centroid mode and mass corrected before analysis with MarkerLynx Applications Manager v4.1. The liquid chromatography−MS data were peak detected and noise reduced for both the liquid chromatography and MS components. Each peak in the resulting 3-dimensional data set was represented by retention time, m/z and intensity in each sample. All ions that did not appear in more than 25% of the samples of at least one group were considered to be noise and removed. The study designs of STI, MTI and CHT differed most notably in the numbers of subjects and the extent of control of the diet, leading to significant variation in the amplitude of the effect of citrus intake on the metabolomic profiles, relative to other sources of variation. Furthermore, STI had a temporal dimension whereas MTI and CHT captured profiles as a snapshot. Statistical treatments were thus tailored to each study design, as described below. All Principal Component Ananlysis (PCA) and Partial Least Square-Discriminant Analysis (PLSDA) were performed using SIMCA-P+ software (version 12.0, Umetrics AB, Umea, Sweden). In PLS-DA, the ability to classify each individual in the correct consumption group was assessed by R2Y. All PLS models were built using the 7-fold crossvalidation method. The prediction power of the model was assessed by the Q2 parameter. To check that PLS components could not lead to a correct classification by chance, a permutation test was carried out (n = 100). For each test, samples are randomly assigned to each experimental group, a PLS model is carried out and R2Y and Q2 are computed. The result of the tests is displayed on a validation plot, which shows the correlation coefficient with the original non permuted sample, having a value 1 on the horizontal axis and R2Y and Q2 values on the vertical axis. Logically, permuted samples must lead to poor predictive models with low Q2 values and therefore the regression line of Q2 values must have a negative intercept. Variable Importance Projection (VIP) values were obtained as indicators of importance of each ion in the discrimination. STI Data Analysis. In STI, ion intensities were adjusted according to the volume of each urine fraction collected. Data were log transformed and a two-way mixed linear model on repeated measures (factors diet and time) was applied using the proc mixed function of SAS software (SAS v9.1, SAS Institute, Cary, NC). All ions with p-value