Challenges Associated With the Design and Deployment of Food Intake Urine Biomarker Technology for Assessment of Habitual Diet in Free-Living Individuals and Populations-A Perspective

Abstract

Improvement of diet at the population level is a cornerstone of national and international strategies for reducing chronic disease burden. A critical challenge in generating robust data on habitual dietary intake is accurate exposure assessment. Self-reporting instruments (e.g., food frequency questionnaires, dietary recall) are subject to reporting bias and serving size perceptions, while weighed dietary assessments are unfeasible in large-scale studies. However, secondary metabolites derived from individual foods/food groups and present in urine provide an opportunity to develop potential biomarkers of food intake (BFIs). Habitual dietary intake assessment in population surveys using biomarkers presents several challenges, including the need to develop affordable biofluid collection methods, acceptable to participants that allow collection of informative samples. Monitoring diet comprehensively using biomarkers requires analytical methods to quantify the structurally diverse mixture of target biomarkers, at a range of concentrations within urine. The present article provides a perspective on the challenges associated with the development of urine biomarker technology for monitoring diet exposure in free-living individuals with a view to its future deployment in "real world" situations. An observational study (n = 95), as part of a national survey on eating habits, provided an opportunity to explore biomarker measurement in a free-living population. In a second food intervention study (n = 15), individuals consumed a wide range of foods as a series of menus designed specifically to achieve exposure reflecting a diversity of foods commonly consumed in the UK, emulating normal eating patterns. First Morning Void urines were shown to be suitable samples for biomarker measurement. Triple quadrupole mass spectrometry, coupled with liquid chromatography, was used to assess simultaneously the behavior of a panel of 54 potential BFIs. This panel of chemically diverse biomarkers, reporting intake of a wide range of commonly-consumed foods, can be extended successfully as new biomarker leads are discovered. Towards validation, we demonstrate excellent discrimination of eating patterns and quantitative relationships between biomarker concentrations in urine and the intake of several foods. In conclusion, we believe that the integration of information from BFI technology and dietary self-reporting tools will expedite research on the complex interactions between dietary choices and health.

Keywords: biomarker of food intake (BFI); dietary intake; habitual diet; metabolomics; urinary biomarkers.

Figure 1

Workflow for biomarker panel development. Where: MAIN, Metabolomics at Aberystwyth, Imperial and Newcastle; RP, reverse phase; HILIC, Hydrophilic Interaction Liquid Chromatography; LoD, limit of detection; LoQ, limit of quantification.

Figure 2

Screening biomarkers to detect those with concentrations in spot urine that reflect levels found in 24 h urine from Study 1. (A) Boxplot of total creatinine content and refractive index (RI) in First Morning Void (FMV) and 24 h urine samples. (B) Scatter plot showing the linear association between creatinine concentration and RI in FMV and 24 h urine samples. (C) Scatter plots showing the linear association of selected biomarker concentrations between FMV and 24 h urine samples. Where: Biomarker 7,7-Methyl xanthine; Biomarker 8, Acesulphame-K; Biomarker 12, Calystegine A3; Biomarker 19, D,L-Sulphoraphane-N-acetyl-L-cysteine; Biomarker 23, DHPPA [3-(3,5-Dihydroxyphenyl)-1-propanoic acid]; Biomarker 42, Proline betaine; Biomarker 50, Tartarate; Biomarker 53, Trimethylamine-N-oxide (Full list of number codes for biomarkers is in Supporting Data 4).

Figure 3

Comparison of stability of example biomarkers in First Morning Void (FMV) urine collected in vacuum tubes stored at 4°C for 2 weeks (Vacuum transfer method) or Universal tubes (Standard Universal collection) stored at −80°C. Where: Biomarker 1, 1-Methyl histidine; Biomarker 3, 3-Methyl histidine; Biomarker 15, Carnosine; Biomarker 17, Creatinine; Biomarker 18, D,L-Sulphoraphane L-cysteine; Biomarker 23, DHPPA [3-(3,5-Dihydroxyphenyl)-1-propanoic acid]; Biomarker 31, Ferulic acid-4-O-sulphate; Biomarker 37, L-Anserine; Biomarker 50, Tartarate; Biomarker 53, Trimethylamine-N-oxide (Full list of number codes for biomarkers is in Supporting Data 4).

Figure 4

In silico overview of the chemical diversity of candidate biomarkers. (A) Chemical class and Superclass classifications of 54 biomarkers using ClassyFire. (B) Multi-dimensional scaling of Tanimoto Distances between MACCS (Molecular ACCess System) fingerprints of biomarkers. (C). Visualisation of biomarkers in chemical space. Where: logP, partition coefficients [Full list of number codes for biomarkers in panels (B,C) is in Supporting Data 4].

Figure 5

Relationship between relative standard deviation (RSD), biomarker concentration in calibration standard mixtures, and median concentration in First Morning Void (FMV) urines from Study 1. RSD data of biomarkers used to monitor exposure to six example foods/food groups when measured at nine concentration levels over a 6-month period. The concentration range within which the median concentration of the same biomarkers was measured in FMV urine taken from 95 free-living participants are highlighted by dotted boxes.

Figure 6

Biomarker panel to characterise eating behaviour on individual Menu Days in Study 2. (A) Principal Components Analysis (PCA) scores plot of biomarker panel measurements in First Morning Void (FMV) urine of 15 individuals across three Menu Days in Study 2. (B) PCA variable loadings plot showing the variance contributions of biomarkers on each Menu Day (Full list of number codes for biomarkers in panel B is in Supporting Data 4). (C) Boxplots illustrating the concentration in FMV urine of top ranked biomarkers discriminating Menu Days following Random Forest classification. Text box provide details of meals consumed on each Menu Day. Where: Biomarker 3,3-Methyl histidine; Biomarker 7,7-Methyl xanthine; Biomarker 19, D,L-Sulphoraphane-N-acetyl-L-cysteine; Biomarker 22, DHBA-3-O-sulphate; Biomarker 50, Tartarate; Biomarker 53, Trimethylamine-N-oxide.

Figure 7

Demonstration that biomarker panel can be extended without any loss of classification power. Multi-dimensional scaling (MDS) of Random Forest proximity extracted from a classification model of 15 individuals consuming three unique menus in Study 2. (A) MDS using a panel of 38 biomarkers of food intake (BFIs) which was extended to 54 BFIs and (B) 2 years later.