1

The use of metabolomics for studying the effects of pre- and probiotics Katrine Overgaard Poulsen

Bachelor’s thesis in Agrobiology - Food Science Katrine Overgaard Poulsen 201407648

2

Bachelor’s thesis (15 ECTS) June 2017

Title:

The use of metabolomics for studying the effects of pre- and probiotics

Danish title:

Brugen af metabolomics til at undersøge effekterne af præ- og probiotika

Purpose of the thesis (copied from bachelor’s contract):

The worldwide obesity epidemic and changes in lifestyle to less physical activity have given rise to

growing numbers of people suffering from metabolic syndrome. Pre- and probiotics have been

found to have potential health benefits after intake, especially in relation to obesity, diabetes and

cardiovascular diseases. Based on literature, the potential health benefits associated with pre- and

probiotics for human health and the role of metabolomics to investigate the effect of pre- and

probiotics will be described.

Author:

Katrine Overgaard Poulsen, 201407648

Institution:

Aarhus University, Department of Agrobiology

Main supervisor:

Ulrik Kræmer Sundekilde*, Postdoc

Co-supervisor:

Jette F. Young*, Associate professor

* Aarhus University, Department of Food Science - Differentiated & Biofunctional Food

Total number of pages: 34.8 pp. (2400 counts per page, text only)

3

I Abstract

Pre- and probiotics have proven to modulate the colonic microbiota and activity, and to induce

health effects in the host regarding metabolic syndrome, gut diseases and improvements of immune

system function in vitro and in vivo. However, interactions with the host are complex and the

biochemical effects on host’s metabolism remain unclear. In order to elucidate the actions of pre-

and probiotics, the technique metabolomics is emerging. Thus, this thesis includes a pilot study

were two participants consumed one commercial probiotic drink á day for three weeks followed by

a wash-out period, and collected six urine samples individually. The metabolic profile of the urine

samples were investigated by 1H nuclear magnetic resonance (NMR) spectroscopy based

metabolomics. By the use of multivariate statistical analysis (principal components analysis)

changes in urinary metabolites of the samples were examined. The study proved great inter- and

intra-individual differences especially due to variations in creatinine excretion, but no clear relation

to probiotic intake was found. Literature on the matter demonstrates indications of relations

between pre- and/or probiotic supplementations and changes in the metabolic profile of a biological

source, albeit it is mostly based on animal studies. Thus, metabolomics offers potential in exploring

effects of pre- and probiotics on biochemical pathways and host’s metabolism, but there is a need of

greater human intervention studies.

II Sammendrag Præ- og probiotika har vist sig at modulere tarmens mikrobiota og aktivitet samt at inducere

sundhedseffekter i værten angående metabolisk syndrom, tarmsygdomme og forbedringer af

immunsystemets funktion in vitro og in vivo. Dog er interaktionerne med værten komplekse, og de

biokemiske effekter på værtens metabolisme er ukendte. For at belyse virkningen af præ- og

probiotika er teknikken metabolomics fremspirende. Denne opgave inkluderer derved et pilot

interventionsstudie med to deltagere, som indtog én kommerciel probiotisk drik om dagen i tre uger

efterfulgt af en wash-out periode og afgav seks individuelle urinprøver. Urinprøvernes metabolske

profil blev undersøgt ved brug af 1H nuclear magnetic resonance (NMR) spectroscopy baseret

metabolomics. Via multivariate statistisk analyse (principal komponent analyse) blev ændringer i

urinprøvernes metabolitter beset. Forsøget viste klare inter- og intra-individuelle forskelle, særligt

grundet variationer i creatinine udskillelse, men viste ingen klar relation til indtag af probiotika.

Litteratur på området demonstrerer indikationer på sammenhæng mellem indtag af præ- og/eller

probiotika og ændringer i den metabolske profil af en biologisk kilde, men er dog primært baseret

på dyrestudier. Der er således potentiale for, at metabolomics kan udlede de effekter, som præ- og

probiotika har på biokemiske pathways og værtens metabolisme, men der er et behov for større

interventionsstudier med mennesker.

4

Table of content I Abstract 3II Sammendrag 31 Introduction 5

a) The human microflora 5a.i) Microbial activity of the human gut flora and the end-products produced 5

b) The definition of pro- and prebiotics 7b. i) Introduction to potential beneficial effects of pre- and probiotics 8

c) Exploring the human metabolome 11c. i) Metabolomics 12Chemical analysis used in metabolomics 12Basic principles of nuclear magnetic resonance spectroscopy (NMR) in regards to metabolomics 13Chemometrics 17c. ii) Metabolic composition of urine 18

2 Materials and methods 22a) Pilot study and sample collection 22b) Sample preparation and data acquisition by 1H NMR 22c) Identification of metabolites and multivariate analysis 23

3 Results 25a) 1H NMR spectra and some identified metabolites 25b) Inter-individual differences reflected by PCA plots 26c) Intra-individual differences reflected by PCA plots 27

c.i) Intra-individual differences observed in subject U 27c.ii) Intra-individual differences observed in subject K 29

4 Discussion 32a) Characteristics of urine profiles 32b) Associations between observed discriminators and probiotic intake 34c) Methodological considerations 41

c.i) Some considerations regarding NMR vs. MS 41c.ii) Some considerations regarding study design 42

5 Conclusion and perspectives 436 Acknowledgements 437 Literature 44Appendix 54

5

1 Introduction Studies indicating that certain species of the colonic microbiota influence the host’s health have led

to the concept of probiotics and later prebiotics. Intake of pre- and probiotics is an attempt to

modify the composition of the gut flora, and has shown to modulate immune response,

improvements of the gut health and to exert positive effects on symptoms associated with metabolic

syndrome (Ebel et al., 2014). However, the exact mechanisms leading to modifications of the host’s

metabolism after intake remain elusive, but they are generally considered to be due to changes in

microbial activity and metabolites thereof.

Metabolomics is “a comprehensive analysis in which all the metabolites of a biological system are

identified and quantified” (Lindon & Nicholson, 2008) and offers a holistic analysis of biological

samples such as tissues and urine. Hence, by the use of metabolomics, we can identify and quantify

metabolites in a biological system, and by multivariate analysis explore changes induced by

consuming pre- and/or probiotics.

a) The human microflora The human microflora (the microbiome) consists of more than 400 different species and more than

1014 bacteria cells, who inhabit the gastro-intestinal system and exist in a symbiotic relationship

with the host (Eckburg et al., 2005; Scott et al., 2008).

The quantitatively greatest bacterial species are of the phyla Bacteroidetes and Firmicutes (Eckburg

et al., 2005), including the genera Prevotella, Eubacteria, Lactobacillus, Ruminococcus and

Clostridium as well as the genera Bifidobacterium of Actinobacteria (Power et al., 2013; Tremaroli

& Bäckhed, 2012). General categorisation of the human gut flora has been attempted and gave rise

to different enterotypes predominated by Bacteroides, Prevotella and Ruminococcus (Arumugam et

al., 2011). However, this categorisation is argued to be less discrete (Tremaroli & Bäckhed, 2012).

The composition of the gut flora is established early in life during the neonate period and is greatly

determined by birth mode of delivery (i.e. vaginal or cesarean) (Murgas et al., 2011; Ravel et al.,

2011), but can be remodelled by environmental factors such as lifestyle, diet and diseases. This

affect homeostasis and thus the composition of gut flora through complex interactions (Nicholson et

al., 2012). A distortion in the composition of the gut flora is referred to as dysbiosis and is a partial

underlying mechanism in the development of metabolic diseases, inflammatory bowel diseases

(IBDs) and colon cancer (Bringiotti et al., 2014; Turroni et al., 2014).

a.i) Microbial activity of the human gut flora and the end-products produced The majority of bacteria in the proximal colon are anaerobic saccharolytic and produce hydrolytic

enzymes (e.g. glycosidases, polysaccharidases) in order to degrade and ferment non-digestible

carbohydrates that enter the colon (Hamer et al., 2008; Macfarlane & Macfarlane, 2012).

6

A variety of intermediates and end-products including CO2, H2 and organic acids such as lactate,

succinate, acetate, propionate and butyrate are produced in the gastrointestinal system (Cummings

& Macfarlane, 1997). The latter three comprise the quantitatively most prominent short-chain fatty

acids produced (SCFAs) (Manas et al., n.d.).

Nitrogen constituents derived from protein typically pass through the proximal colon, since

carbohydrates are the preferred source of energy (Preter et al., 2011). Hence, the majority of

nitrogen compounds enter the distal colon for proteolytic fermentation, and are degraded to smaller

amounts of SCFAs, branched-chain fatty acids, ammonia, hydrogen sulphide, indole and volatile

amines dependent on substrate and energy availability (Cummings & Macfarlane, 1997; Hamer et

al., 2008). Methanogens and other bacteria growing on intermediate products are also present in the

colon (Gibson et al.,1993).

Derivatives not being assimilated by bacteria in the colon are absorbed by the colonocytes across

the apical membrane (Verbeke et al., 2015). Considering the SCFAs produced, butyrate is

predominantly used by the colonocytes for energy, and therefore not further transported across the

basolateral membrane. Acetate and propionate are transported to the blood stream and liver (Preter

et al., 2011). Resultantly, the human gut flora secures a pathway of energy salvage for the host

(Tremaroli & Bäckhed, 2012).

However, some compounds produced by the gut flora are potentially toxic. Here among are phenols

and indoles from dissimilatory metabolism of aromatic amino acids in the gut. These are detoxified

by glucoronide or sulphate conjugation in the liver and excreted in the urine. Ammonia from

deamination of amino acids is likewise excreted as urea (Cummings & Macfarlane, 1991; Smith &

Macfarlane, 1996). Furthermore, some bacteria exert enzyme activities such as azoreducatase,

nitroreductase, urease and β-glucorinase, which have been found to be associated with the

conversion of pro-carcinogenic compounds to carcinogenic compounds (Preter et al., 2008;

Ouwehand, et al., 2002).

The aforementioned divergence between the microbial activity in the proximal colon and the distal

colon is partly due to differences in the energy available and the substrate entering the colon. This

affect the distribution and composition of bacteria and the end-products produced (Cummings &

Macfarlane, 1991). A greater availability of non-digestible carbohydrates in the proximal colon

leads to greater production of SCFAs, which lowers pH in the colon (Cummings et al., 1987).

A reduction in pH changes the microbial activity from decreased proteolytic activity to increased

saccharolytic activity, thus preventing production of potentially toxic derivatives of amino acid

fermentation (Smith & Macfarlane, 1996). The effect of SCFAs as bioactive molecules in regards to human biology has been widely studied.

7

A review by Cummings and Macfarlane (1997) emphasizes how SCFAs (especially butyrate) affect

transport across the gut epithelium layer, the metabolism of the colonocytes, growth and

differentiation of cells, hepatic control of lipid and carbohydrates, and finally energy salvage among

other potential effects. Moreover, it is assumed that the symbiosis between microbiota and humans

influences the host’s immune system and disease development (Macfarlane & Macfarlane, 2012;

Martin et al., 2012). The identification that certain microorganisms have a positive impact on

human health has led to the concept of pro- and prebiotics.

b) The definition of pro- and prebiotics The human body is as before mentioned host for a variety of microorganisms, but not all exert a

probiotic effect. The definition of probiotics has been redefined since its introduction the 1900s

(Metchnikoff, 1908). Today, the most common definition has been established by Guarner &

Schaafsma (1998), who define probiotics as ”live microorganisms, which upon ingestion in certain

numbers, exert health effect on the host”. They must survive passing the human gastrointestinal

system in order to fulfil their potential beneficial effect in the host (FAO (Food and Agriculture

Organization) & WHO, 2006).

According to Shah (2007), 56 species of the genus Lactobacillus have been identified as being

probiotic, with Lactobacillus acidophilus being the most prevalent. Of the genus Bifidobacterium,

29 species are identified, here among are Bifidobacterium adolescentis, B. infantis and B. breve

(Fijan, 2014; Shah, 2007). Species of Escherichia Coli, Streptococcus, Saccharomyces,

Enterococcus, Leuconostoc and Pediococcus have also been identified to have probiotic effect

(Fijan, 2014; Ohland & Macnaughton, 2010). It should be emphasised that probiotic genera are

commonly present in the human gastrointestinal system despite lack of dietary supplements

(Fontana et al., 2013). Additionally, other authors stress that the probiotics do not have to colonize

the gastrointestinal tract in order to exert a beneficial effect. Non-colonizing strains thus exert

effects as they pass through the colon by remodelling the gut flora (Ohland & Macnaughton, 2010).

Probiotic species express β-galactosidase, fructofuranosidase and β-glucosidase activities among

others, thus ferment carbohydrates to particularly lactic acid and acetate, which lowers pH in the

colon and reduces proteolytic activity (Preter et al., 2008; Tabasco et al., 2014). Lactate and acetate

can subsequently be converted into butyrate by other microbial species in the gut (Rajkumar et al.,

2015). However, the expressions of the aforementioned enzyme activities are species and strain

specific (Tabasco et al., 2014).

Consumption of certain foodstuff can selectively influence the composition of bacteria in the colon

and induce the growth of these, thus leading to the concept of prebiotics (Gibson & Roberfroid,

1995).

8

The definition of prebiotics has been widely revisited since its introduction in 1995 (Gibson &

Roberfroid, 1995). The currently accepted definition of dietary prebiotics is ”a selectively fermented

ingredient that results in specific changes, both in the composition and/or activity in the

gastrointestinal microflora thus conferring benefits upon host health” (Gibson et al., 2010).

Prebiotics provide an attentive manipulation of the gut flora, and denote foodstuff aimed at inducing

the growth of especially strains of Lactobacillus spp. and Bifidobacterium spp. in the

gastrointestinal system by stimulating saccharolytic activity (Gibson & Roberfroid, 2004). They

must be resistant towards hydrolysis and absorption in the upper part of the gastrointestinal system

in order to exert beneficial effects. Intake of inulin-type fructans and galacto-oligosaccharides have

been demonstrated to improve growth of bifidobacterium and lactobacilli in multiple studies

(Kaplan & Hutkins, 2000; Roberfroid et al., 2010). A review by Gibson & Roberfroid (2004)

proposes that prebiotics are non-digestible carbohydrates such as frucotooligosaccharides (FOS),

galacto-oligosaccharides (GOS) and lactulose.

b. i) Introduction to potential beneficial effects of pre- and probiotics Today, various commercial probiotic products are available (Tripathi & Giri, 2014). In order to be

ascribed a health claim, probiotic supplementations must provide evidence of health benefits in

vitro and from animal and human studies in vivo (FAO (Food and Agriculture Organization) &

WHO, 2006). Several review articles gather some demonstrated health benefits of pre- and

probiotics. Here among are Preter et al. (2011), who distinguish between the effect of probiotic

intake on modification of the host’s immune reactivity, and interference with colonic metabolism

respectively. This is demonstrated by in vitro and in vivo studies testing the effects of selected

bacterial strains and validating how certain strains are able to adhere to the mucosa and colonize the

gastro-intestinal tract, resulting in a changed microbial activity towards increased saccharolytic

activity (Preter et al., 2011).

Shah (2007) has systematically reviewed the effect of colonization by probiotic strains in the colon

and assesses some health effects of lactobacilli and bifidobacteria. These effects include an induced

immune response, anti-mutagenic activity, improvements of bowel diseases and contribution to

enhanced epithelial cellular barrier growth (Ewaschuk et al., 2008; Shah, 2007).

Immunological and anti-mutagenic effects are inter alia provided by the ability of some strains to

increase IgA antibody levels and to reduce nitroreductase, azoreductase and β-glucuronidase

activity, the latter three enzymes involved in activation of mutagens (Goldin & Gorbach, 1984;

Shah, 2007).

Additionally, multiple reviews (Arora et al., 2013, Gibson & Roberfroid, 1995; Preter et al., 2011;

Shah, 2007) highlight potential beneficial effects of Bifidobacteria and Lactobacillus by reducing

9

blood level ammonia, lowering blood cholesterol, improvement of serum lipid profile and to inhibit

the growth of pathogens.

Based on in vitro studies, the cholesterol lowering effect is assumed to be either due to 1) inhibition

of dietary cholesterol absorption from the small intestine by binding of bile acids, 2) assimilation of

cholesterol by lactic acid bacteria or 3) excretion of bile acids after binding of them (Pereira &

Gibson, 2002).

Table 1 summarizes selected studies demonstrating health benefits after administration of pre-

and/or probiotics. Search criteria for table 1 were results of in vivo human studies.

The search objective was aimed at 3 categories: obesity and serum lipids, metabolism and disease

markers (markers being related to inflammatory bowel’s disease (IBD), lactose intolerance, immune

system) and alterations in gut flora composition owing to the statements of the aforementioned

reviews.

Table 1: Demonstrated health benefits from pre- and/or probiotic administration in human trials

Health marker Genus or prebiotic Species (strain) Duration and administration

Study design Reference

Obesity and cholesterol level

Reduced total cholesterol, reduced body-mass index, reduced LDL-lipoprotein

Bifidobacterium lactis (HN019) FM 80mL/day 45 d

R, PC n = 51 (metabolic syndrome symptoms)

Bernini et al. (2016)

Reduction in total plasma cholesterol level compared to placebo group

Lactobacillus Bifidobacterium

acidophilus lactis

FM 3x100 g/day 6 weeks

SB, PC n = 14 (moderate hypercholesterolemia)

Ataie-Jafari et al. (2009)

Reduction in total plasma cholesterol from baseline and placebo-group

Lactobacillus plantarum (CECT 7527, CECT 7528, CECT 7529)

1 capsule/day 12 weeks

R, DB, PC, parallel trial n = 60

Fuentes et al. (2012)

Reduced LDL cholesterol correlated with an increase in plasma de-conjugated bile acids Decreased plasma fibrinogen

Lactobacillus reuteri (NCIMB 30242)

Capsule 2/day, 9-week intervention

R, DB, PC n = 127 (hypercholesterolemic)

Jones et al. (2012)

Reduction on LDL, total cholesterol and HDL (probiotic or symbiotic)

Lactobacillus salivarius (UBL S22)

Capsule 1/day, 6 weeks

R, SB, PC pilot study n = 45 3 groups; P, w/L. salivarius, w/L. salivarius + FOS

Rajkumar et al. (2015)

Metabolism and disease markers

Reduce the mutagenicity of urine and faeces after ingestion of cooked meat

Lactobacillus acidophilus FM 2/day

Intervention study with multiple phases n = 11 (healthy)

Lidbeck et al. (1992)

Reduction in serum concentrations of CRP, IL-1β/IL-6, TNF-α (probiotic/synbiotic)

Lactobacillus salivarius (UBL S22)

Capsule 1/day, 6 wk

R, SB, OC pilot study n = 45 3 groups; P, w/L. salivarius, w/L. salivarius + FOS

Rajkumar et al. (2015)

10

Decreased TNF-α, IL-6 cytokines

Bifidobacterium lactis (HN019) FM 80mL/day 45 d

R, PC n = 51 (symptoms of metabolic syndrome)

Bernini et al. (2016)

Improvement of IBD symptoms Increased faecal β-galactosidase activity

Lactobacillus Bifidobacterium Streptococcus

acidophilus, casei, brueckii spp. Bulgaricus, plantarum longum (Y10), breve (Y8), infantis (Y1) salivarius, spp. thermophilus

3/day, 20 d VSL-3 (pharmaceutical probiotic) administered

Brigidi et al. (2001)

Decreased β-glucuronidase activity

Lactulose, oligofructose-inriched inulin

2/day, 4 wk R, C-O, PC study n = 53 (healthy)

Preter et al. (2008)

Decreased β-glucuronidase activity Increased β-glucosidase activity (B. brevis)

Lactobacillus Saccharomyces Bifidobacterium

casei (Shirota) boulardii brevis

FM 2/day, 4 wk

R, C-O, PC study n = 53 (healthy)

Preter et al. (2008)

Alterations in the intestinal microflora

Reduced faecal count of E. Coli (synbiotics) Increased faecal lactobacilli counts following (probiotics and synbiotics) Decreased faecal E. Coli counts (probiotics and synbiotics)

Lactobacillus FOS

salivarius (UBL S22)

Capsule 1/day, 6 wk

R, SB, PC, pilot study n = 45 3 groups; P, w/L. salivarius, w/L. salivarius + FOS n = 45

Rajkumar et al. (2015)

Increased counts of Bifidobacterium and Lactobacillus

XOS Rice porridge 1/day 6 wk

R, PC n = 20 (healthy)

Lin et al. (2016)

No significant effect on faecal concentration of total lactobacilli or bifidobacterium

Lactobacillus Lactococcus Streptococcus Bifidobacterium Propionobacterium

acidophilus, rhamnosus, plantarum, casei, helveticus, bulgarius lactis thermophiles bifidum shermani

2 capsules/day 1 wk

Short-term intervention study n = 11 (healthy) Commercial product Measurement of faecal composition of lactobactobacilli and Bifidobacterium using qPCR

Taverniti et al. (2014)

Abbreviations: C-O, cross-over; CRP, C-reactive protein; d, days; DB, double-blinded; FM, fermented milk; FOS, fructo-oligosaccharide; IL, interleukin; P, placebo; PC, placebo controlled; R, randomized; SB, single-blinded; TNF, tumor necrosis factor; w/, with; wk, weeks; XOS, xylo-oligosaccharide.

Indications of alterations in microbial activity and anti-mutagenic activity following pre- and/or

probiotic intake have been observed by Preter et al. (2008), who found decreased β-glucoronidase

activity in faeces samples from patients receiving Lactulose, Lactobacillus Casei (Shirota),

Bifidobacterium brevis or combined pre- and probiotics (synbiotic). The authors assigned this

decrease to have an impact on the development of potentially carcinogenic compounds.

Furthermore, Rajkumar et al. (2015) observed indications of an improved immune response due to a

reduction in pro-inflammatory cytokines in the plasma. The authors assigned the effect to

improvements of intestinal barrier function induced by Lactobacillus salivarius and/or FOS.

11

Despite demonstrations of improved health effects in humans following pre- and/or probiotic

supplementation, results of different studies are frequently contradictory.

For example, Ataie-Jafari et al. (2009) and Fuentes et al. (2012) found improved effects on plasma

cholesterol following probiotic supplementation.

However, a comprehensive systematic review of human studies on probiotics and lipid profile

(triglyceride, total cholesterol, LDL cholesterol, HDL cholesterol and blood pressure) found that out

of 22 studies with 16 different probiotic species only 8 species proved to have an effect, indicating

great inconsistency (Kallioma et al., 2010). Another thorough review by Preter et al. (2011) on the

effect of pre- and probiotics stresses how prebiotics generally prove to increase the production of

SCFAs in human, animal and in vitro studies. However, the effects of probiotics on microbial

activity in human studies are more inconsistent as compared to animal studies or in vitro studies.

Furthermore, Flaherty et al. (2015) reviewed the impact of supplementation on mucosal immunity

and found that the potential beneficial effector molecules produced by probiotics are highly species

and strain specific. Thus, probiotics exert species and strain specific changes on the host’s health,

and the effect of one strain cannot be extrapolated to other strains (Fijan, 2014; Shah, 2007).

To sum up: the human gastrointestinal tract is colonized by a wide range of microorganisms that

ferment non-digested foodstuff entering the colon to a variety of end-products. Some of these

microorganisms improve the host’s health and are said to be probiotic. The activities of these can be

selectively induced by prebiotics. The underlying biochemical mechanisms of pre- and probiotic

impacts on host’s metabolism are not clearly elucidated, but are generally assumed to be exerted

through modulation of microbial activity and microbial produced metabolites (Arora et al., 2013;

Preter et al., 2011; Turroni et al., 2014). These metabolites either exert beneficial effect in the colon

or in the circulatory system of the host, thus influencing other biochemical reactions through

complex interactions.

c) Exploring the human metabolome Through a symbiotic relationship with the host and subsequent degradation of foodstuff, the gut

flora influences the metabolome of an individual. The metabolome is a collection of metabolites

(<2000 Da*) of both extrinsic and intrinsic origin found in a biological system (Wishart et al.,

2013). Therefore, the metabolome of an individual is composed of endogenous compounds and

xenobiotics of extrinsic origin (Holmes et al., 2008). Metabolites can be perceived as the “end-point

of the bio-molecular gene-protein-metabolite cascade” (Claus & Swann, 2013). Hence, the

metabolites and the metabolome expose information regarding the biochemical processes in an

organism and/or a biological compartment (Alonso et al., 2015; Holmes et al., 2008).

Several factors exhibit great influence on the composition of metabolites in an organism. *The limit is not clearly defined, others refer to metabolites as being <1500Da, (Wishart, 2014).

12

Among others are gender, age, gut flora and diet contributors to variation (Claus & Swann, 2013;

Slupsky et al., 2007), making the composition highly complex. In order to analyse the human

metabolome by quantifying and identifying metabolites, the technique metabolomics is emerging

(Alonso et al., 2015; Wishart, 2014).

c. i) Metabolomics The terms metabolomics and metabonomics are frequently used interchangeably. However,

metabonomics is defined as ”the quantitative measurements of the time-related multiparametric

metabolic response of living systems to pathophysiological stimuli or genetic modification”

(Nicholson et al., 1999), whereas metabolomics can be defined as “a comprehensive analysis in

which all the metabolites of a biological system are identified and quantified” (Lindon &

Nicholson, 2008). The term metabolomics will be used in the current study.

Metabolomics offers analysis of endogenous and exogenous metabolites in different biological

samples by the use of multitargeted analysis (Bijlsma et al., 2006; Lindon, 2004) and have been

used in studies regarding the gut flora (Nicholls et al., 2003; Nicholson et al., 2005).

Metabolomics aims at exploring the full metabolome of an organism by either targeted and/or non-

targeted analysis. Targeted metabolomics can be described as being biased for selected metabolites.

Opposite, un-targeted metabolomics is an analysis of all detectable metabolites, where one aims at

profiling the entire metabolic profile of a sample without a prior hypothesis regarding the

composition (Alonso et al., 2005; Ryan et al., 2013).

The procedure of metabolomics progresses from chemical analysis to identification and

quantification of compounds to a final interpretation of the data in a biological context (Roessner &

Bowne, 2009), resulting in a biological integrated analysis of the organism (Holmes et al., 2008;

Slupsky et al., 2007). This follows the need of analytical software and databases in order to interpret

the results of the chemical analysis (Beckonert et al., 2007).

The widespread use of metabolomics (and metabonomics) in different scientific fields has given

rise to the Human Metabolome Database (HMDB: http://www.hmdb.ca/), which “brings together

quantitative, chemical, physical, clinical and biological data about all experimentally ‘detected’

and biologically ‘expected’ human metabolites” (Wishart et al., 2013).

Chemical analysis used in metabolomics The following two types of chemical analysis techniques are most frequently used in metabolomics:

Nuclear magnetic resonance spectroscopy (NMR) and mass spectroscopy (MS) (gas

chromatography-mass spectrometry (GC-MS) or liquid chromatography-mass spectrometry (LC-

MS)) (Slupsky et al., 2007; Wishart, 2014).

13

A description of NMR-spectroscopy is pursued in the following with emphasis on the exploitation

of physical and chemical properties in regards to metabolomics and to a lesser extend on the

physical and chemical principles behind NMR analysis. Finally, an introduction to interpretation

and exploration of spectroscopic data by multivariate analysis is given in the section Chemometrics.

Basic principles of nuclear magnetic resonance spectroscopy (NMR) in regards to metabolomics Data acquisition NMR is a quantitative, non-destructive technique and can be used for analysis of molecular

structures in a biological sample (Beckonert et al., 2007). NMR is able to identify and quantify

approximately 50-200 metabolites simultaneously with a detection limit of 5µM, and has been

utilised in the past decades for characterization of human derived biofluids (Wishart, 2014).

The molecules detected by NMR are polar molecules such as amino acids, sugars and organic acids

(Wishart, 2014). The technique is based on nuclei spin, energy absorption of specific nuclei

subjected to a magnetic field and re-emission (Bothwell & Griffin, 2011).

Nuclei of atoms are charged and have a certain spin quantum number, I, which differs between

different nuclei, though always in multiples of ±½ (Claridge, 1999). Spin properties of nuclei create

distinct magnetic moments when the nuclei are placed in an external magnetic field. This makes the

nuclei behave like a magnet and causes them to be oriented in a certain direction (Lane, 2012).

In a magnetic field, B, the individual nuclei precess around their own axis. They absorb photons of

the frequency, ν, determined by the gyromagnetic ratio, γ, of the nuclei and the strength of the

magnetic field (Pavia et al., 2001), thus:

𝜈 = 𝛾 ∗ 𝐵

By applying a mathematical approach, the orientations (magnetic moments) of nuclei are expressed

through a vector model by which the bulk of magnetic moments of nuclei is along the z-axis in a

Cartesian co-ordinate system, parallel to the magnetic field (Claridge, 1999).

Nuclei with spin take up 2I+1 possible orientations. Therefore, protons with spin ½ have 2 distinct

energy states: in the orientation parallel to the magnetic field (lower energy state, α) or opposite to

the magnetic field (higher energy state, β). Not all nuclei possess spin properties, and these

therefore have a quantum number of 0. This makes them undetectable in NMR analysis (Claridge,

1999). The isotopes 1H, 13C or 31P spin ±½ and are frequently used.

Individual nuclei have distinct magnetic moments and thus absorb radio waves at different

frequencies. NMR exploits spin properties and magnetic moments and induces transition of nuclei

to a higher energy state (Lane, 2012). The energy of photons must match the energy difference

between the energy states of the nuclei.

14

The energy of a photon is related to Planck’s constant, h, and the frequency, ν. Therefore the

energy, E, of a photon needed to cause transition is expressed by:

𝐸 = ℎ ∗ 𝛾 ∗ 𝐵 (Claridge, 1999).

Thus the frequency needed to excite the system is given by:

𝜈 = ∆𝐸/ℎ (Lane, 2012).

However, despite some nuclei having the same spin (such as 1H, 13C or 31P), they do not have the

same differences in energy state. Therefore, one can apply a certain energy pulse in order to

selectively induce energy transition in specific nuclei (Lane, 2012).

Resultantly, NMR spectroscopy induces a transition in energy states of nuclei with spin properties

and distinct magnetic moments by applying a defined energy pulse on a sample, causing the nuclei

to have resonance with the incoming wave (Pavia et al., 2001).

Nuclei are not equally distributed in the distinct energy states, but separated according to

Boltzmann’s distribution. Some will be in the greater energy state and others in the lower energy

state. The stronger the energy applied on nuclei, the greater the difference in population. Hence,

more nuclei will be in the low energy state for excitation (Pavia et al., 2001).

Nuclei in a complex sample are present in different molecules, and subject to different local

magnetic environments, which affect the resonance frequency by which they absorb. This leads to

individual chemical shift, ∂, of nuclei (Bothwell & Griffin, 2011). Therefore two magnetic sources

affect the nuclei; namely the applied magnetic field and the local magnetic environment.

Application of a 90° pulse causes the bulk magnetization of specific nuclei to rotate from the z-axis

to the x-y plane in a Cartesian co-ordinate system. When turning off the applied pulse, the nuclei

relax back to equilibrium and return the magnetization to the z-axis (Lane, 2012), subsequently

emitting radio waves with wavelengths dependent on the nature of the molecules and the chemical

environment. The signals emitted from the nuclei and the time it takes to restore to the z-plane are

characteristic for different nuclei in different molecules and measured in NMR. This time is

determined by the spin-lattice relaxation time, T1, and the spin-spin relaxation time, T2. T1 refers to

the time it takes for the longitudinal magnetization to recover, whereas T2 refers to the decay of

transverse magnetization (Hornak, 1997). This energy emission is collected as signals called free

induction decays (FIDs) by a detector (Bothwell & Griffin, 2011).

The electromagnetic energy emitted can be translated to a function of frequencies by a Fourier

transformation, which converts time-domain data into a spectrum or plot of intensity vs. frequency

(ppm) (Bothwell & Griffin, 2011) as depicted in figure 1.

15

Figure 1: Transformation of radio-waves emission into a 1D spectrum, adapted from Bothwell & Griffin (2011)

Data of NMR spectroscopy are reflected in one-dimensional (1D) spectra or two-dimensional (2D)

spectra with 1D being the most prevalent. A 1D spectrum has two axes. A y-axis represents the

intensity of absorption and a x-axis represents the chemical shifts (values in parts per million, ppm)

expressed by the difference in resonance between a reference substance and a given substance

(Blümich & Callaghan, 1995).

In order to characterize and quantify the peaks, one must consider the chemical shifts, the area of

the peaks and the pattern of the peaks regarding spin-coupling.

Spectral analysis - identification and quantification of metabolites Chemical shifts are essential for identification of molecules in a sample (Lane, 2012). The chemical

shifts reflect the nature of nuclei and the local magnetic environment they are situated in, which

causes the nuclei to absorb at different frequencies than they otherwise would in a single atom

(Chemistry, n.d.). The chemical shifts are relative to a reference set at 0 ppm.

Adjacent nuclei located in different magnetic environments influence each other’s magnetic field,

reflected by intramolecular interactions called scalar coupling.

Scalar coupling reflects bond order and strength between nuclei in a molecule, and is portrayed in a

NMR spectrum by splitting of the resonance signal.

This separation is determined by the coupling constant, J. A general rule is that a nucleus situated in

a molecule with mutually coupled nuclei less than three bonds away results in n+1 peaks (Hornak,

1997; Lane, 2012). A triplet therefore reflects a nucleus scalar coupled to two equivalent nuclei.

16

Through examination of the pattern of spectral peaks, one can obtain information concerning the

chemical structure of the molecules, and identify the metabolites detected by NMR (Alonso et al.,

2015). A table of chemical shifts in 1H NMR is depicted in appendix A1.

The area under a peak is proportional to the relative number of hydrogens given rise to the peak

(proton NMR). Therefore, one can integrate the peaks and quantify the concentration, provided the

presence of a reference and knowledge of chemical shifts (Lindon, 2004; Pavia et al., 2001;

Smolinska et al., 2012). The references for chemical shift that are most frequently used are 3-

trimethylsilylpropionic acid (TSP), 2,2-dimethyl-2-siapentane-5-sulfonate sodium salt (DSS), and

tetramethylsilane (TMS) for organic solvents (Beckonert et al., 2007) and are set at 0.00 ppm.

A basic pulse sequence of a 1D NMR experiment demands setup of relaxation delay, pulse width

and acquisition time of the time domain data. The latter can be zero-filled in order to maximize the

data points used for computing the spectrum (Pauli et al., 2005). Deuterated solvents such as

deuterium oxide (D2O) are added in order to lock on a deuterium signal.

The aforementioned HMDB offers great assistance in identification of metabolites in a spectrum.

However, owing to the complexity of a biological sample, peaks in a spectrum might overlap,

making it difficult to distinguish between compounds. To overcome this, one can run another

experiment with an increased concentration of one of the compounds, thereby spiking the sample

(Dona et al., 2016). By subsequent spectrum analysis, the peaks can be identified. Other methods

include running a heteronuclear 2D spectrum analysis. Heteronuclear 2D spectrum analysis

provides increased signal dispersion and divides overlapping peaks by exploring the scalar

interactions between a proton and another magnetic nucleus (Lane, 2012; Wishart, 2014).

Graphically, the analysis is displayed by a spectrum of two axes, with chemical shift values of one

of the nuclei represented on either of the axis and chemical shifts of the other nuclei on the other

axis.

Data pre-processing Data pre-processing is an intermediate step between the raw data obtained from the chemical

analysis and data interpretation. This follows the need of phase correction, baseline correction,

alignment, normalization, binning and scaling. Some authors distinguish between data pre-

processing and data pre-treatment (Goodacre et al., 2007), however in this thesis, the terms will be

considered in concert as according to Smolinska et al. (2012).

Phasing and baseline-correction are done manually or automatically by NMR spectrum software.

Phasing aims at correcting asymmetrical peaks following Fourier transformation. Base-line

correction aims at reducing bends across the base-lines (Goodacre et al., 2007).

In order to align the peaks in a spectrum, the signal of the internal reference is adjusted to 0 ppm.

17

Alignment by the method icoshift (interval correlated shifting) divides spectra into various

segments and aligns the spectra to the corresponding segments of a reference spectrum (Tomasi et

al., 2011).

Normalization of the urinary metabolite concentration is important, since the overall concentration

of metabolites in urine can vary to a great extent. The peak area of creatinine is often used as a

reference for normalization, since creatinine clearance in urine is considered constant (Smolinska et

al., 2012), and for compensation of potential variations in urine volume, which can affect the

concentrations of metabolites (Craig et al., 2006; Saude et al., 2007).

Other frequently applied methods include normalization according to urine volume, area of TSP

signal, the use of probalistic quotient normalization, the sum of the squared value of all variables

for a given sample (2-Norm) and sum of value of all variables for a given sample (1-Norm)

(Rasmussen et al., 2011).

Binning divides the spectrum into segments of either equal size (equidistant binning) or non-

equidistant, and therefore reduces the dimensionality of the spectrum (Ross et al., 2007).

Mean-centering and scaling are frequently applied in concert. Mean-centering is done by

subtracting the mean from each value, thereby differences between high-concentrated and low-

concentrated metabolites are adjusted to vary around zero (Craig et al., 2006; Smolinska et al.,

2012). Scaling is applied to reduce variations of data values (e.g. spectral intensities).

Scaling can be done by unit variance (UV) scaling, where one calculates the standard deviation of

each variable, sk, and subsequently multiply with a scaling weight 1/sk (Eriksson et al., 2006) or by

Pareto scaling, where each variable is divided by the square root of sk (Trygg et al., 2007).

Semiautomatic identification and/or quantification of metabolites in a sample can be obtained by

commercial software such as Chenomx NMR suite, which also refers to HMDB for peak

identification. This programme also allows for phase correction, baseline correction, reference

adjustment and pH determination (Wishart, 2014).

Chemometrics Data analysis aims at exploring the data and spectra following data pre-processing and detecting

significant trends and differences (Smolinska et al., 2012). Chemometrics is a mathematical

interpretation of chemical data. Both supervised and unsupervised analysis can be employed.

Principal component analysis (PCA) is an unsupervised, multivariate projection method, which

aims at portraying the total variance in a data set (Eriksson et al., 2006; Trygg, 2003).

It summarizes variation in a dataset into scores (T) and different loadings (P) of these scores

resulting in “principals components” (PCs). A dataset is summed in a matrix X with theoretically N

observations and K variables (Eriksson et al., 2006). The observations are plotted in a K-

dimensional space; therefore each observation is situated according to its measured value.

18

The first PC (PC1) represents the greatest variance explained, and is an approximation of data

according to least square sense (Eriksson et al., 2006). Each point in the dataset is projected onto

the PC1 line and has a co-ordinate named a score. The second PC (PC2) is orthogonal to PC1, and

represents the second greatest variance. These two PCs now define a plane with the observations

having different co-ordinates (scores) and provide insight in the maximum explained variation in

two dimensions. However, multiple PCs can be computed dependent on the size of the data set with

each PC explaining the observed variance in descending order. A general rule is that at least 4

observations are needed to compute one PC.

A graphical presentation of the scores is depicted from a scores plot, which presents the

observations by the use of the co-ordinate values resulting from the PCs and groups the

observations according to variance. Each variable influence the variance between observations to

different extends and therefore have individual vectors, which are applied on the PCs. These vectors

are termed loading vectors, and the resulting influence of the individual loadings on the PCs is

depicted in a loading plot (Eriksson et al., 2006). Thus, the latter represents the weight by which

different factors affect the observed variance.

In order to validate one’s model as plotted by PCA, one can confer with the Q2 value. This value

indicates the quality of the model in terms of consistency between predicted and original data. The

value has a theoretical maximum of 1, and generally one can accept a model of Q2 ≥ 0.40 (Worley

& Powers, 2013).

c. ii) Metabolic composition of urine The metabolic composition of urine reflects metabolic processes in both liver and kidneys (Lindon

et al., 2000; Wishart, 2014). The primarily role of the kidneys is to maintain homeostasis by

filtering body fluids and excreting toxins and other waste products.

Therefore, urine is apt for studying degradation or detoxification pathways (Ryan et al. 2013), and a

suitable matrix for investigating biomarkers, which can be defined as “measureable changes

associated with physiological and pathological processes” (Wu & Gao, 2015).

Urine sampling is non-invasive, and the complex biofluid has been largely investigated by several

chemists in order to characterize and identify the human urine metabolome (Bouatra et al., 2013;

Saude & Sykes, 2007).

Through review of preceding studies and literature, Bouatra et al. (2013) composed the Urine

Metabolome Database (UMDB: http://www.urinemetabolome.ca) as an integrated part of HMDB.

This database provides knowledge of metabolites found in human urine, their corresponding

concentrations and potential disease association (Bouatra et al., 2013). Urine has a final pH of

around 6.2, and the most abundant metabolites include urea, creatinine, hippurate and citrate,

inorganic ions and salts, water-soluble toxins and pigments (Bouatra et al., 2013; Rose et al., 2015).

19

Extended information on the composition of the final urine and concentration of metabolites can be

found in Bouatra et al. (2013).

A typical spectrum of urinary metabolites is visualised in figure 2 showing the complexity of urine

and illustrating some quantitatively great metabolites such as hippurate, urea, creatinine and citrate

measured by 1H NMR.

Figure 2: Spectrum of metabolites found in urine by the use of 1H NMR. TMAO, trimethylamine-N-oxide; TSP, 3-trimethylsilylpropionic acid (reference substance), adapted from Beckonert et al., 2007.

The metabolic composition of urine depends inter alia on diet, time of day and age, leading to intra-

and inter-individual differences with an average concentration range of ±50 % (Bouatra et al.,

2013). This makes quantification, identification and analysis of the urine in regards to elucidating

characteristics of phenotypes challenging (Bouatra et al., 2013; Lindon et al., 2000; Rose et al.,

2015). Table 2 assembles examples of sources to variations and their effect on the metabolic

composition of urine. Table 2: Sources of variation in the metabolic composition of human urine

Source of variation Main effects of human NMR studies (urine) References Gender W:

HL: 1,2,3,4citrate, 2,3glycine, 3hippurate, 3lactate, 4fumarate, 4creatine M: HL: 1,3taurine, 1,2,3,4creatinine/creatine, 3TMAO, 33-hydroxybutyrate, 4carnitine, 4acetylcarnitine, 4acetone

1 Kochhar et al. (2006) 2Wang et al. (2005) 3Psihogios et al. (2008) 4Slupsky et al. (2007)

Age Older individuals: LL: 1,3,4creatine/creatinine, 33-hydroxybutyrate, 43-hydroxyisovalerate, 4cis-aconitate, 4lactate, 4alanine, 4carnitine HL: 3hippurate, 3TMAO, 4trigonelline

1 Kochhar et al. (2006) 3Psihogios et al. (2008) 4Slupsky et al. (2007)

20

Abbreviations: HL, higher level; LL, lower level; M, men; NAG, N-acetyl glycoprotein; P, plasma; TMAO, trimethylamine-N-oxide; W, women; a reference and a study are related, if they have the same subscripts.

As depicted in table 2, dietary differences have contributed to variations in acetate, hippurate,

lactate, citrate and glycine excretions in urine (Holmes et al., 1994). Additionally, intake of protein

sources also contributes to variations in the composition of urinary metabolites, especially due to

changes in trimethylamine-N-oxide, creanitine and citrate excretions (Lenz et al., 2004; Rasmussen

et al., 2012; Stella et al., 2006).

Furthermore, gender differences have been reflected by greater citrate and glycine excretions in

females than males, whereas creatinine is excreted in higher concentrations in males as it correlates

with body weight and muscle mass (Kochhar et al., 2006; Psihogios et al., 2008; Wang et al., 2005).

Adding to the complexity, the gut flora of an individual also affects the metabolic profile of urine.

As previous stated, the gut flora produces several compounds when fermenting foodstuff in the

colon, leading to great impact on the metabolites circulating in the body and therefore also the

composition of metabolites in urine. The composition of gut flora is dependent on multiple factors

such as host genotype, diet, age, and can be linked to certain metabolic phenotypes and metabolic

profiles (Calvani et al., 2010; Li et al., 2008).

Prior knowledge of the metabolic composition of urine and the normal physiological differences

that are present between individuals is therefore of great importance for urine metabolomic

analysis, since one must differentiate between normal inter- and intra-individual variations and the

Diurnal Morning: HL: 4creatinine, 4dimethylamine

4Slupsky et al. (2007)

Diet 5Standardization proved less inter-individual variation 6Greatest proportional variation between people in citrate, hippurate, lactate, acetate, glycine HL: hippurate after 13fruit and vegetables intake and 14tea consumption 7,8,9,10Protein diet (fish/meat/protein): HL: 7,8,9,10TMAO, 9,10creatinine, 9carnitine, 9taurine, 9creatine, 91- and 3-methylhistidine, 10nitrogen excretion LL: 10citrate 12Vegetarian diet: HL: NAG, glycine, succinate, citrate, hippurate LL: taurine, formate, TMAO, phenylalanine, methylhistidine

5Walsh et al. (2006) � 6Holmes et al. (1994) 7Lenz et al. (2004) 8Dumas et al. (2006) 9Stella et al. (2006) 10Rasmussen et al. (2012) 12Xu et al. (2010) 13Krupp et al. (2012) 14Mulder et al. (2005)

Body mass index 8Obese, BMI >25: HL: 2-hydroxyisobutyrate LL: hippurate, xanthine, trigonelline

8Calvani et al. (2010)

Gut flora derived metabolites

4TMAO, 4,11dimethylamine 8,11hippurate, 8,112-hydroxyisobutyrate, 9phenolic, benzoyl, 11phenyl derivatives, 9derivatives of choline metabolism, 9indole derivatives, 10p-cresol, 113-aminoisobutyrate, 114-cresol sulphate, 11taurine, 11lactate

4Slupsky et al. (2007) 8Calvani et al. (2010) 9Nicholson et al. (2012) 10Preter et al. (2008) 11Li et al. (2008)

21

differences correlating to a physiological response of a disease state or intervention (Saude &

Sykes, 2007). In regards to elucidating isolated effects of probiotic supplementation on the

metabolic composition of urine, several factors such as diet, lifestyle, gender and age must be taken

into account.

22

2 Materials and methods a) Pilot study and sample collection

Two participants were enrolled in the study: subject K (female) and subject U (male).

Both participants were equipped with 22 Actimel Danone Strawberry yoghurts fermented with

Lactobacillus Casei Danone-culture (Lactobacillus paracasei subsp. paracasei ((DN-

114001/CNCM I-1518), see appendix A2 for ingredients and brand details).

The participants were asked to continue their regular diet and regular life routines though should not

ingest any other products with claimed probiotic content. No controls were used in the study, and

the subjects were not asked to keep track of their diet.

Urine sample number I was collected on day 0. The first yoghurt was subsequently ingested in the

morning of day 0 after sample number I had been collected.

One yoghurt á day was consumed for 21 days. Last yoghurt was ingested on day 21.

The yoghurts were consumed in the morning preferable at the same time of day each day.

The samples should represent the “morning pee”, thereby should be collected before breakfast in

the early morning. Complete sample homogeneity and integrity cannot be obtained, however were

attempted by collecting the urine approximately at the same time of day and by storing the samples

in a freezer around [-20;-18°C] immediately after collection.

The intervention period was followed by a wash-out period of 6 weeks, after which a final urine

sample was collected.

Figure 3 is an outline of the course of the intervention study including number of day and the

corresponding sample number. See appendix A3 for a detailed outline of the intervention.

Figure 3: Illustration of the course of intervention reflecting the time points by which samples were collected and the first and last yoghurt ingested. S, sample number in Roman numbers (appendix A3).

b) Sample preparation and data acquisition by 1H NMR

The urine samples were stored in a regular freezer (-18° C) until chemical analysis.

The 12 urine samples thawed in a water bath and were centrifuged at 12000g for 5 minutes at +4 °

C. The samples were kept cool in a refrigerator (+5 ° C) until pH measurements were executed.

SIII from each observation was used for pH measurements and buffer adjustments in order to verify

that the samples were in the range of pH [6.5-7.5].

23

After verification, a Master Mix was prepared by mixing 1380µL buffer and 1620µL D2O water

(Sigma-Aldrich, Copenhagen, Denmark) to a total volume of 3000µL. The buffer was composed of

NaH2PO4 (583 mM) (Sigma-Aldrich, Copenhagen, Denmark) sodium formate (1.162 mM) with a

final concentration of 0.25 mM DSS (Sigma-Aldrich, Copenhagen, Denmark) as internal chemical

shift reference (see appendix A4 for ingredients list).

A duplicate was prepared for each of the 12 samples resulting in 24 samples (see appendix A5 for

IDs). Therefore, 24 standard 5 mm NMR tubes were prepared by adding 130 µL Master Mix and

570 µL of a given sample to a corresponding tube. All tubes were gently mixed and allocated in a

random order for analysis.

Spectral measurements were obtained by the use of a Bruker Avance 600 NMR spectrometer

(Bruker Biospin, Rheinstetten, Germany) with a frequency of 600.13 MHz, and data acquisition

based on Beckonert et al. (2007). Samples were analysed at 298 K and individually shimmed before

each run. Samples in an aqueous matrix (e.g. urine and serum) need water suppression. This is done

by running pulse experiments such as a presaturation 1D NOESY (Nuclear Overhauser

Enhancement Spectroscopy) - experiment (Beckonert et al., 2007). Thus, one-dimensional spectra

were acquired by the use of 1D NOESY -presat with a 90°pulse sequence and automatic water

suppression during the relaxation delay (Bruker standard pulse: noesypr1). The relaxation delay was

set at 2 seconds and mixing time 0.1 s. A total of 64 scans for each sample were collected. The

spectral width was set at 20.03 ppm with 32K data points. Subsequent zero-filling resulted in

approximately 65K data points.

The 1H spectra were processed in Topspin 3.0 (Bruker Biospin) and adjusted to the signal of the

internal reference DSS at 0 ppm. The spectra were individually phased and baseline-corrected

before further processing.

Prior to Fourier transformation, the FIDs were multiplied by 0.30 Hz line-broadening and the water

region was eliminated from each spectra.

c) Identification of metabolites and multivariate analysis Chenomx NMR suite 7.0 was used to identify the metabolites in the spectra. Automatic fit was

applied (internal library: Chenomx NMR 600 MHz) in order to ease the identification.

The spectra were analysed in order to construct metabolic profiles of the urine samples. Afterwards,

peaks that were not identified or verified by the automatic approach were analysed and evaluated by

comparing the spectra with literature (Bouatra et al., 2013) and HMDB.

A targeted approach was used to examine for peaks of known microbial origin such as SCFAs.

Raw NMR data were pre-processed in Matlab. Alignment was done using Icoshift (Tomasi et al.,

2011) by setting the reference peak to 0.00 ppm and subsequently overlaying every spectrum on top

of each other.

24

The spectra were subsequently divided in intervals and aligned again in order to reduce variation in

chemical shift between the spectra. Prior to normalization, the water region ranging from 4.7-4.9

ppm was removed, and likewise were the regions below 0.05 ppm and above 10.0 ppm. The

resulting spectra therefore ranged from [0.05;10.0] ppm with the water region excluded.

Normalization was proceeded by 1-Norm and subsequent binning of spectra by binning sizes of 0.1

reduced the dimensionality of the spectra to consist of 927 data points per spectrum.

The resulting data were transferred to SIMCA 14.1 for multivariate analysis by Principal

component analysis (PCA) and were scaled by Pareto scaling. The computed models were selected

by conferring to the Q2 value, and data were then explored.

Mayor metabolites contributing to variation were quantified by normalization to creatinine.

25

3 Results a) 1H NMR spectra and some identified metabolites The main objective of this metabolomics study was to analyse changes in urinary metabolic profiles

following probiotic intake. By the use of automatic fit in Chenomx NMR suite 7.0, an average of

109 out of 338 metabolites were identified per spectrum. After analysis of the individual spectra

and fitting of the identified pre-fitted metabolites to each spectrum, a total of 36 different

metabolites were frequently observed (appendix A6). Additionally, 12 unknown compounds (UN)

were observed in the different spectra (appendix A6). 1H NMR spectra of sample V (K5 and U5) from each individual are portrayed in figure 4 and 5.

Figure 4: Representative spectra of subject K with some identified metabolites. Sample K5 is portrayed, collected 22 days after the start of

the intervention. Horizontal axis: chemical shift (ppm), vertical axis: intensity (arbitrary units).

Figure 5: Representative spectra of subject U with some identified metabolites. Sample U5 is portrayed, collected 22 days after start of the

intervention. Horizontal axis: chemical shift (ppm), vertical axis: intensity (arbitrary units).

26

Figure 4 and figure 5 illustrate great similarity regarding the overall urinary metabolic profile, but

the intensity of some identified metabolites differs to a great extend between the subjects. Hippurate

and urea are seemingly more intense in subject K than subject U, whereas peaks corresponding to

creatinine and citrate are greater in subject U (figure 4 and figure 5).

b) Inter-individual differences reflected by PCA plots In order to examine the variance in the data set, a principal component analysis of 1H NMR spectra

from every urine sample collected was performed (including replicates). The model (PCA1)

resulted in 5 principal components (PCs) with Q2 = 0.70. A scores plot with PC1 vs. PC2 is

depicted in figure 6.

Figure 6: PCA1 scores plot of the first and second principal component (PC1 and PC2 respectively), representing variance between individuals K (green) and U (blue). Ellipses encircle the two individuals.

Figure 7: PCA1 loading plot of variation across PC1 with identified peaks. Encircled areas comprise a given identification.

27

The greatest source of variance is reflected by PC1 (54.6 % of total variance) and separates the

samples according to individual with subject K located on the left side of the plot and subject U on

the right side (figure 6). Separation by PC2 proves less variation in samples of subject U, while

samples of subject K are dispersed to a greater extent (figure 6).

The peaks accountable for the variation across the PC1 line were identified in Chenomx and the

identifications are depicted in figure 7. Metabolites strongly contributing to variation are creatinine,

citrate, TMAO/betaine, trigonelline, urea, UN1 and hippurate. Metabolites elevated in subject U

were creatinine and citrate, whereas metabolites elevated in subject K were trigonelline, urea, UN1

and hippurate (figure 7).

The concentration of citrate, TMAO (subject U only), trigonelline and hippurate were subsequently

normalized to creatinine and quantified (appendix A7 and appendix A8).

PCA1 was explored in order to investigate any mutual correlation between the observations and

different time points. Therefore, scores plots of different combinations of PCs were analysed.

However, no noteworthy pattern was observed, since scores were always grouped according to

individuals (results not shown).

Another attempt was preceded to examine any relation to the different time points. Here, the mean

of sample 1 (collected at time point 0) from both individuals was subtracted from each of the

resultant duplicates. Using this matrix, another PCA model was made (PCA2), in order to elucidate

a generalization between the values and time point. Again, samples from the two subjects collected

at the same time point did not correlate. Thus, this model was rejected (results not shown).

c) Intra-individual differences reflected by PCA plots Due to the dominance of inter-individual differences, PCA models of intra-individual differences

were computed (figure 8 and figure 11). The observations of each subject were divided into two

groups: group “No” representing the samples collected before and after the intervention (i.e. U1 and

U6 respectively, K1 and K6 respectively) and group “Yes” representing the samples collected

during the course of the intervention.

c.i) Intra-individual differences observed in subject U PCA3 was computed to examine intra-individual differences in subject U, with 3 PCs and Q2 =

0.58.

Scores plots of different combinations of PCs were explored in order to elucidate a link between

members of group “No” or “Yes”. No clear grouping was found. The combination reflecting the

most noticeable grouping is depicted in figure 8. Some members of group “Yes” are located on the

positive side of both PC1 and PC3, and hence distinct from U1 and U6 (green and purple

respectively) (figure 8).

28

Figure 8: PCA3 scores plot (PC1 vs. PC3) of intra-individual differences observed in U, colour legend to the right reflects sample number

according to appendix A5. Scores are assigned according to NMR numbers in appendix A5. An ellipse encircles members of group “Yes”.

Figure 9: Corresponding loading plot of PCA3 (PC1 vs. PC3). Green: the metabolites accounting for less variance between group “No” and

“Yes”, Red: metabolites accounting for greater variance between group “No” and “Yes”. Loadings related are encircled.

The replicates of U2 (blue) and U3 (red) are not situated in couples as expected, thus hampering the

distinction between the groups (figure 8). However, the analysis was still preceded, and the

corresponding loading plot of PCA3 was computed (figure 9).

The loadings with the greatest influence on variance between “Yes” and “No” across PC1 and PC3

are coloured red on figure 9, and the corresponding peaks were identified in Chenomx and are

marked on figure 10.

29

Figure 10: 1H NMR spectrum of samples U1-U6. Areas in red reflect metabolites contributing to observed differences between groups “No” and “Yes”. Areas encircled by a black cover a given identification. Combined it is apparent that metabolites accounting for discrimination between the two groups

“No” and “Yes” in subject U are: TMAO/betaine, UN12/3-methylxanthine, N-

phenylacetylglycine/phenyl-derivatives, citrate, urea, dimethylamine, UN1, glycolate/1,7-

dimethylxanthine and a new unknown compound at 1.9 ppm (UN13) (figure 9 and figure 10).

Phenyl-derivatives, hippurate, TMAO/betaine and citrate are seemingly more abundant in samples

on the positive side of PC3 (figure 9), thus distinct from U1 situated on the negative side (figure 8).

Identification of the majority of highlighted metabolites between [3;4] ppm was neglected due to

overlapping peaks (figure 10). The presence of N-phenylacetylglycine in urine is controversial and

could be mistaken for phenylacetylglutamine (Wishart et al., 2013). It will therefore also be

regarded as a region of phenyls and aromatic derivatives in general terms.

A reduced excretion of trigonelline is found after normalization to creatinine (appendix A7), though

this is not reflected in PCA3 (figure 9 and figure 10). Additionally, the concentrations of hippurate,

citrate and TMAO/betaine are found to fluctuate during the intervention after normalization to

creatinine (appendix A7). Hence, no clear association between variations of the aforementioned

metabolites and the intervention is evident when combining both PCA3 and the normalized values.

The discrepancies could be due to differences in normalization methods, and the fact that PC1 and

PC3 only accounts for 36.6 % and 14 % of the observed variance in subject U respectively. Hence,

the variance analysed in figure 8-10 accounts for a minor part of the observed variance.

Furthermore, the excretion of creatinine additionally varies to a great extend (appendix A7) and

accounts for some variance observed in PCA3 (figure 10).

c.ii) Intra-individual differences observed in subject K The samples of subject K were also grouped into “No” and “Yes”. 3 PCs were computed with Q2 =

0.65.

30

The model was explored using scores plots of different PC combinations in order to detect any

linkage between group members of “No” or “Yes”. However, no clear association was evident. The

combination reflecting the most association between members of the groups is depicted in figure

11.

Figure 11: PCA4 scores plot (PC1 vs. PC2) of intra-individual differences observed in subject K. Colour legend to the right reflects sample number according to appendix A5. Scores are assigned according to NMR number in appendix A5. Members of group “No” are encircled by a red line. Members of group “Yes” are encircled by a black line.

Figure 12: Loading-line plot of variance (PC2) observed in subject K. Areas encircled cover a given identification.

The replicates on figure 11 are well coupled, and group “No” and group “Yes” are somewhat

separated across PC2 (accounting for 28.3 % of the total variance) with group “No” located on the

negative side. Thus, a loading-line plot of PCA4 was computed and is depicted in figure 12.

The chemical shifts accountable for some of the variation across the PC2 line were identified in

Chenomx, and are also denoted on figure 12.

Identified discriminators include: trigonelline, betaine, creatinine, hippurate, urea and UN1, with

the latter three being more abundant in K1 and K6 (figure 12).

31

However, both trigonelline and hippurate normalized to creatinine concentration resulted in highly

fluctuating concentrations, with the lowest quantified concentration of trigonelline measured in K6

(48.12 µM/mM creatinine) and the highest measured in K1 (146.62 µM/mM creatinine), and the

lowest concentration of hippurate in K6 (435 µM/mM creatinine) and the highest in K2 (1904.28

µM/mM creatinine) (appendix A8). As with subject U, the discrepancies can be reasoned with the

different normalization methods and additionally fluctuations in creatinine concentration (appendix

A8), as well as the fact that PC2 of PCA4 accounts for 28.3 % of the observed intra-individual

variance in subject K. Hence, the identified metabolites are not the only source of variation.

32

4 Discussion In the present pilot study, the effect of consuming probiotic yoghurts for three weeks was

investigated by the use of 1H NMR metabolomics in order to elucidate the potential in

metabolomics as a tool for characterizing effects of probiotics on the urinary metabolome. In the

following discussion section, an assessment of the overall results through a metabolomic approach

is presented. Furthermore, associations between the results and the probiotic treatment will be

examined by reviewing literature regarding metabolomic studies on the effect of pre- and/or

probiotics. Lastly, the discussion will bring about some methodological considerations regarding

the current study.

a) Characteristics of urine profiles A table of frequently identified peaks is found in appendix A6, here among are creatinine, urea,

hippurate and citrate as expected from section c.ii Urine metabolites.

As stressed in c.ii, different food constituents can affect the metabolic composition of urine. This

influence was apparent in the current study, more specifically concerning TMAO/betaine of choline

metabolism, dimethylamine of choline metabolism, tartrate, phenyl derivatives, dimethyl sulfone,

trigonelline and xanthine derivatives.

The identification of betaine and trimethylamine-N-oxide should be considered with caution.

Betaine and TMAO both have a singlet peak at around 3.26 ppm. The peaks overlap, which makes

it hard to distinguish between the two compounds. Both can be derived from dietary choline, though

differ in metabolic fate as visualised in figure 13. The path for betaine production is mainly

mammalian, whereas the path of TMAO formation is sym-xenobiotic, therefore both mammalian

and microbiological (Xie et al., 2013). Additionally, TMAO derives directly from dietary intake of

fish (Lenz et al., 2004).

Figure 13: Metabolism of choline, adapted from Martin et al. (2009). Red ellipses encircle betaine and TMAO.

33

In order to clarify the identification, one could either spike a sample with e.g. TMAO, or one could

acquire a heteronuclear [1H,13C] 2D NMR spectrum (see appendix A10 for 2D spectra).

Various xanthine derivatives were identified, including theophylline, 3-methylxanthine and 1,3

dimethylurate (results not shown, confer to appendix A6). Following pattern was observed: when

xanthine derivatives were detected by automatic fit, caffeine could also be detected, albeit with

some uncertainty. The individual metabolites were analysed by HMDB (Wishart et al., 2013) and

KEGG (Kanehisa et al., 2017), and it is hypothesised that the metabolites derive from the

metabolism of caffeine when present in the same sample.

Rothwell et al. (2014) also suggested a relation between coffee consumption and caffeine derived

metabolites, with the greatest discriminators of difference being 1-methylxanthine, 1,7-

dimethylerate, paraxanthine and theophylline among others, thus further supports the generalization

made based on the present results. In order to verify the hypothesis, one could make a correlation of

the metabolites, though this is beyond the scope of this bachelor thesis.

Dimethyl sulfone originates from intestinal metabolism and foodstuff containing sulfoxides (e.g.

onions) (Winning et al., 2009). These are metabolised in the liver and kidneys.

Tartrate is a food additive, but also found in wine. When ingested, tartrate is metabolized by certain

colonic bacteria species. A minor part is excreted in the urine (Wishart et al., 2013). Trigonelline is

both involved in regeneration of S-adenosylmethionine, found in coffee and a product of niacin

metabolism (Lang et al., 2011; Wishart et al., 2013). Furthermore, niacin itself derives from diet

constituents and to some extent from metabolism of gut flora (Wishart et al., 2013).

Additionally, as stressed in c.ii, the urinary metabolic profile is highly affected by the composition

of the gut flora. In the present study, identified metabolites related to microbiota activity were:

trimethylamine N-oxide, hippurate, tartrate, dimethylamine and 2-hydroxyisobutyrate. Thus greatly

associated with diet-derived metabolites.

The formation of hippurate occurs in 2 steps. The gut flora metabolizes dietary polyphenols and

other aromatics to benzoate. Subsequently, benzoate is conjugated with glycine in the liver or

kidneys, resulting in hippurate (Williams et al., 2010).

Hippurate is found to correlate with Clostridia spp. (Li et al., 2008), be reduced in patients with

irritable bowel diseases (Dawiskiba et al., 2014; Williams et al., 2010) and to be dependent on

dietary intake of polyphenols or benzoate (Krupp et al., 2012; Mulder et al., 2005; Walsh et al.,

2006). Additionally, hippurate has been consistently reported to be negatively correlated with

obesity in animal and human studies, and to vary between individuals (Holmes, 1994; Holmes et al.,

2011).

Thus, the production of hippurate might be related to a certain metabolome or diet, and lower levels

of hippurate excretion could be an indication of dysbiosis in the gut.

34

Combined, the urinary metabolic profiles in the present study are greatly influenced by diet and the

gut flora.

Inclusion of PCA1 allowed for clear separation between individuals, suggesting how these are

subject to different environmental factors. This inter-individual difference accounts for the greatest

variation (reflected by PC1, figure 6). By analysis of the loading-line plot (figure 7) discriminators

include creatinine, urea, hippurate, citrate, trigonelline and UN1. The sources of the majority of

these metabolites have already been considered. Urea is a product of protein catabolism and

produced after deamination of amino acids. Creatinine excretion correlates with body and muscle

mass and found to be associated with a high-protein diet (Rasmussen et al., 2012; Stella et al.,

2006).

A study by Walsh et al. (2006) found how standardization of diet reduced variation in hippurate and

creatinine in urine. The authors highlight the influence of non-nutrient food components as a source

of variation, thus stressing the impact of diet on the urinary metabolic profile. The observed inter-

individual differences are therefore ascribed to gender differences, differences in diet and to other

environmental and physiological differences.

To sum up: analysis of the urine samples by 1H NMR led to identification of some common

metabolites between individuals, albeit those found to be associated with gender differences and

diet predominantly attribute to great inter-individual differences. Hence, no metabolites indicate

clear changes in gut flora activity thus far.

Therefore, it became relevant to explore the intra-individual differences in terms of changes in

metabolite composition as preceded by PCA3 (subject U, figure 8) and PCA4 (subject K, figure

11). Identified metabolites accountable for some intra-individual differences in subject U were:

TMAO/betaine, citrate, phenyl derivatives, dimethylamine, hippurate, urea and creatinine, and in

subject K: betaine, trigonelline, hippurate, urea and creatinine.

The following section seeks to explore any associations between the aforementioned discriminators

and probiotic intervention. The most prominent identified gut-derived metabolites contributing to

variation reflected by PCA3 and PCA4 are considered.

These include: TMAO/betaine, citrate and phenyl derivatives (identified as N-phenylacetylglycine)

for subject U (figure 9 and figure 10) and hippurate and trigonelline for subject K (figure 12).

b) Associations between observed discriminators and probiotic intake A typical biomarker of changes in gut flora activity is the concentration of SCFAs measured. As

stated in section b.i, probiotic activity is mainly saccharolytic, thus they ferment non-digestible

carbohydrates and induce production of SCFAs. The SCFAs are absorbed across the gut epithelium

layer or excreted with faeces.

35

In the present study, a targeted approach was applied to all of the spectra before multivariate

analysis in order to assess whether probiotic intake changed the excretion of SCFAs in the urine and

hence, could indicate changes in microbial activity. However, no SCFAs were identified (appendix

A6).

Zheng et al. (2011) successfully detected butyrate in urine and faeces samples of rats by the use of

GC-MS and LC-MS. Additionally, Verbeke et al. (2010) successfully detected 13C-SCFAs in

human urine samples by the use of GC-MS after supplementation of radio-active labelled non-

starch polysaccharides. In another study, Boets et al. (2016) investigated the availability of colonic

derived SCFAs in healthy adults after supplementation of labelled 13C-SCFAs and detection by GC-

MS, and found that only 0.05% of supplemented SCFAs were found in urine.

The aforementioned studies and the results of the current study thus suggest at least 3 following

hypotheses: 1) 1H NMR might be too insensitive to identify and quantify SCFAs in human urine, 2)

SCFAs in urine are suitable biomarkers of changes in microbial activity, but the study design was

insufficient or the commercial product was not able to affect the composition of microbiota, 3)

SCFAs are not suitable biomarkers of probiotic supplementation when analysing human urine

samples by 1H NMR, and other metabolites in urine might be suitable biomarkers. The 2 former

hypotheses will be considered in Methodological considerations, whereas the latter will be

discussed in the following.

Studies using metabolomics in order to elucidate potential effects of pre- or probiotic treatment in

humans are not plentiful, therefore the following table includes animal studies on the effect of pre-

or probiotic supplementation. Table 3: Intervention studies using metabolomic techniques in order to elucidate the effect of pre- or probiotic supplementation

Prebiotic or probiotic used Study design Overall results Reference L. rhamnosus Galactooligosaccharides Combined (synbiotic)

1H NMR (HBM mice, urine, serum) Microbial Profilling (fecal and jejunal contents) With controls

Urine: Probiotic: HL: α-ketoisovalerate, TMA LL: butyrate, isovalerate Prebiotic: HL: carnitine, acylcarnitine, taurine LL: α-ketoisocaproate, butyrate, lysine, α-aminoadipate Synbiotic: Similar to prebiotic, though increase in creatine and α-ketoisovalerate

Martin et al. (2009)

L. paracasei or L. rhamnosus Changes in metabolite profile of multiple biological compartments (HBM mice, plasma, urine, liver, fecal) Supplemented with L. rhamnosus or L. paracsei

Urine: HL: tryptamine, indoleacetylglycine, phenylacetylglycine LL: creatine, citruline (L. rhamnosus), LL: 3-hydroxyisovalerate (L. paracasei), LL: α-ketoisocaproate, citrate

Martin et al. (2008a)

36

L. paracasei or L. rhamnosus with two types galactooligosaccharides (Pre1 and Pre2)

Co-administration of probiotics and prebiotics Integrative metabolic profiling (HBM mice, urine, fecal, plasma, liver)

Urine: HL: 1-methylnicotinamide, LL: unidentified lipids (L. paracasei) HL: creatine, taurine, LL: arginine, citruline (L. rhamnosus) Pre1: LL: phenylacetylglycine, N-acetyl- and O-acetylglycoproteins, tryptamine, HL: citrate Pre2: HL: TMA, glycerate, creatine, LL: α-ketoisovalerate, arginine, citrulline

Martin et al. (2008b)

Changes in metabolic profile and microbiome after consumption of prebiotics

1H NMR (human, faeces) n = 12 PC

Residues of PDX in faeces Positive correlation between acetate and Bacteriodes in faeces Negative correlation between bifidobacteria and succinate and acetate

Lamichhane et al. (2014)

Lactobacillus Casei Shirota in fermented milk Lactulose

Supplementation of lactose-[15N]ureide and [2H4]tyrosine to asses the effect of pro- prebiotics on intestinal flora GC-MS (human, urine) R, PC, C-O n = 29

Probiotic consumption: LL(%): p-[ring-2H4]cresol and 15N in 12-24h urine excretion Prebiotic: LL(%): p-[ring-2H4]cresol in 0-24h urine excretion, 15N in 6-24h urine excretion

Preter et al. (2004)

Lactobacillus casei Shirota Bifidobacterium breve Yakult OF-IN

Effect on colonic nitrogen-protein metabolism Short term and long term effect Combustion-isotope ratio MS (human, 15N in urine and faeces) GC-MS (human, p-cresol content in urine and faeces) lactose[15N, 15N]ureide biomarker R, PC n = 20

LL: p-cresol (urine) LL: 15N excretion in urine accompanied by HL: 15N excretion in faeces

Preter et al. (2007)

Bifidobacterium longum Lactobacillus acidophilus and prebiotic (FOS) (synbiotic)

1H NMR (human faeces) n = 16 1 month In vitro study

Less change in lactobacilli counts LL: Succinate, tyrosine, phenylalanine, alanine, glutamate, lysine, glycine, valine, isoleucine HL: propionate, butyrate, acetate, lactate Favour of amino acid assimilation rather than catabolism

Ndagijimana et al. (2009)

Lactobacillus salivarius PS2 1H NMR (women with mastitis, urine) n = 23 BL-C

HL: hippurate, TMAO, creatine

Vázquez-Fresno et al. (2014)

Lactobacillus acidophilus La5 Bifidobacterium lactis Bb12

1H NMR (mice, urine and faeces) Aged mice Adult mice

Adult mice: HL: N-methylnicotinamide, dimethylglycine, choline LL: sarcosine, phenylacetylglycine, nicotinate, α-ketoisocaproate, 2-oxo-4-methylvalerate

Brasili et al. (2013)

37

Aged mice: HL: dimethylglycine, butyrate, 3-hydroxyisovalerate, 2-oxo-4-methylvalerate, threonine LL: succinate, N-methylnicotinamide, citrate Response dependent on age

Dietary inulin-type fructans 1H NMR (Obese women, urine and plasma) qPCR analysis of 16s rDNA (faeces) DB, PC n = 30

Increased Bifidobacterium and Faecalibacterium prausnitzii No clear clustering of plasma and urine metabolic profiles between subjects Positive correlation between higher urinary hippurate and Collinsella

Dewulf et al. (2013)

Potato fibres Pea starch Lactobacillus acidophilus

Lactose-[15N2]ureide degration Isotope ratio mass spectrometry (human, 15N in blood, urine, faeces) n = 14

LL: renal excretion by supplementation of fibres (potato or pea) vs. control Prebiotic intake lowered colonic generation and renal excretion of 15NH3

Wutzke et al. (2010)

Lactobacillus paracasei F19 Lactobacillus acidophilus LA-5 Bifidobacterium lactis BB-12

GC-MS (IBD patients, serum) 8 w n = 61 Consumption of acidified milk with/without probiotics

Effect independent of probiotic supplementation HL: lactate, glutamine, proline, creatinine/creatine, aspartic acid LL: glucose

Pedersen et al. (2011)

Abbreviations: BL-C, baseline controlled; C-O, cross-over; DB, double-blinded; HBM, human baby microflora; HL, higher level; IBD, irritable

bowel disease; LL, lower level; OF-IN, oligofructose-enriched inulin; PC, placebo-controlled; Pre, prebiotic, R, randomized, w, weeks.

Table 3 demonstrates that changes in the concentration of some discriminators in the present study

have previously been linked to pre- or probiotic treatment, and that other metabolites than SCFAs

are identified following an intervention.

Preter et al. (2004) and Preter et al. (2007) investigated the use of labelled supplementation of

lactose-[15N]ureide and [2H4]tyrosine as biomarkers of changes in microbial activity towards

saccharolytic fermentation following supplementation of Lactobacillus Casei Shirota and Lactoluse

(Preter et al., 2004) and Lactobacillus casei Shirota, Bifidobacterium breve Yakult and OF-IN

(Preter et al., 2007).

The authors found reduced urinary excretion of 15N and p-cresol, thus supplying indications of other

biomarkers than production of SCFAs. These findings are further substantiated by Wutzke et al.

(2010) and indicated by Ross et al. (2013), who found that intake of a whole-grain diet (though not

assigned prebiotic) led to changes in protein metabolism in men and lowered the urinary urea

excretion. Additionally, Martin et al. (2008b) observed a decrease in phenylacetylglycine following

prebiotic supplementation and related the results to indications of decreased proteolytic activity.

Contradictory, Martin et al. (2008a) reported an increase in specific phenyl and indole derivatives

by supplementation of probiotics alone (table 3). The authors highlight how these derivatives imply

proteolytic fermentation in the gut and are linked to the metabolism of aromatic amino acids.

38

In the current study, the region of phenolic and aromatic derivatives was assigned to contribute to

variation observed in subject U (figure 10). These peaks were initially identified as N-

phenylacetylglycine, a glycine conjugate of phenylacetic acid metabolised through phenylalanine

metabolism (Wishart et al., 2013). The metabolism has been correlated to Bacteroides, Clostridia

and E. Coli in vitro (Smith & Macfarlane, 1996), thus demonstrating the compound’s association

with the composition of gut flora.However, the peaks could also reflect other aromatic and phenolic

derivatives, and therefore not only a product of protein metabolism. Lee et al. (2006) demonstrated

that certain phenolic derivatives of black tea suppress the growth of pathogenic bacteria found in

faeces in vitro. Albeit not completely comparable with the current study, it is interesting that an

increase in the concentration of phenolic derivatives has proven to be an indication of modulated

gut flora, thus the variation observed could indicate changes in microbial activity in agreement with

Martin et al. (2008a) and Lee et al. (2006). However, Martin et al. (2008a) also found reduced level

of 3-hydroxyisovalerate following supplementation of L. paracasei and reduced level of citrate.

This is contradictory with the current study. Citrate is identified as a contributor to variance (figure

10) with increasing concentration during the course of the intervention (appendix A9). Thus, the

observed changes in phenolic derivatives/N-phenylacetylglycine are more likely due to diet and not

probiotic supplementation.

Combined, these findings indicate that mainly prebiotic intake alters the metabolism of the gut flora

towards increased saccharolytic activity, perhaps owing to the nature of their structure. An increase

in energy available for saccharolytic fermentation leads to greater demand of nitrogen sources for

bacterial growth, and hence reduces the formation of potential toxic compounds and the excretion

of nitrogen in the urine (Cummings & Macfarlane, 1991; Scheppach et al., 2001).

Probiotic supplementation alone has not proven to affect saccharolytic activity as much (Martin et

al., 2008a; Wutzke et al., 2010).

In the present study, trigonelline was identified to contribute to intra-individual differences in

subject K (figure 12).

Trigonelline has been found to be a constituent in coffee (Ashihara et al., 2014; Lang et al., 2011).

A study found that 50 % of ingested trigonelline was excreted 0-8 h after coffee consumption (Lang

et al., 2010). In the present study, samples were taken in the early morning, 12 h past last coffee

consumption. It might also be expected that other biomarkers of coffee consumption and caffeine

metabolism were contributors to discrimination, if coffee was the only factor accountable for

differences. Therefore, it is relevant to explore other sources of urinary trigonelline excretion.

Trigonelline derives both from the metabolism of vitamin B3 (niacin), and from the conversion of

S-adenosylmethionine (SAM) to S-adenosylhomocysteine. The latter cycle is linked to the

metabolism of betaine (figure 13).

39

Brasili et al. (2013) found indications of modulations of betaine-homocysteine metabolism and

NAD metabolism after probiotic supplementation in mice of different age by 1H NMR analysis of

urine and faeces. The probiotic treatment induced urinary excretion of dimethylglycine in both adult

and aged mice. Additionally, the authors linked probiotic supplementation to modulation of other

pathways regulating homocysteine amounts, namely methylation of nicotinate to N-

methylnicotinate (trigonelline) and N-methylnicotinamide. Combined, the results indicated changes

in betaine-homocysteine and NAD metabolism. The authors also found increased concentration of

3-hydroxyisovalerate among others, though no such changes were identified in the current study,

and opposing the findings of Martin et al. (2008a).

Interestingly, by multicompartmental metabolic profiling, Martin et al. (2009) found modulation of

transmethylation pathways (betaine-homocysteine metabolism) after supplementation of pre- and

probiotics in HBM mice linked to the regeneration of SAM. They argued that probiotic

supplementation induced the conversion of homocysteine to methionine, “crucial for maintaining

methionine homeostasis, detoxifying homocysteine and producing S-adenosylmethionine” (Martin et

al., 2009). Additionally, other studies have linked reduced trigonelline excretion with depletion of

SAM and increased oxidative stress (Calvani et al., 2010; Sun et al., 2008). In the present study, it

is not possible to clearly relate the observed differences in trigonelline to modulations of betaine-

homocysteine metabolism, since there is no indication of induced production of other derivatives,

which might be expected considering the aforementioned studies, but it is interesting that

trigonelline - inter alia produced from demethylation of SAM to S-adenosylcysteine, is found to

contribute to variation (figure 12).

1-methylnicotinamide (another derivative of nicotinate metabolism) was identified in the current

study (appendix A6), but not assigned to contribute to observed variation in the present PCA model

(figure 12). The variation in trigonelline concentration could also be an indication of modulation of

NAD metabolism, and it can be speculated, whether probiotic supplementation results in

modulations of homocysteine and NAD metabolism as observed in the aforementioned studies, and

whether this modulation has an impact on health. However, after quantification of trigonelline, no

clear pattern of variation is observed in subject K (appendix A8).

Contradictory, subject U excreted less trigonelline during the course of the intervention (appendix

A7), but the differences were not identified in PCA3 (figure 10).

Considering the potential discriminator TMAO, Martin et al. (2008a) found increased level of TMA

and TMAO in the liver of mice supplemented with Lactobacillus rhamnosus, indicating modulation

in methylamine and choline metabolism. The authors referred to Allison & Macfarlane (1989) and

argued that supplementation of L. rhamnosus induced production of methylamines by Bacterioides

and Clostridium perfringens, thus modulating the gut flora.

40

L. paracasei was found to have different metabolic effect on choline and not related to increased

TMAO production. As stressed previously, TMAO originates from more than one source, and the

compound could be mistaken for betaine in the current study. Despite an observed variation (figure

10), evidence does not support a relation between the changes in TMAO/betaine concentration and

probiotic intake in the current study.

Citrate was also identified as a potential discriminator in subject U (figure 10). Citrate is an

intermediate in the TCA cycle and derives from dietary sources. Additionally, citrate is found to

positively correlate with milk intake (Zheng et al., 2015). After quantification of citrate no clear

association to the intervention became evident, since the quantitatively greatest concentration of

citrate was measured in the post-intervention sample (U6, 411.09 µM/mM creatinine, appendix

A7). Citrate was also found in increasing but diminishing amounts in subject K (appendix A8). The

reason for a seemingly elevated excretion of citrate could be due to an increased consumption of

carbohydrate or from the intake of fermented yoghurts, since no other TCA cycle intermediates

were identified as contributors of variation.

In the present study, the level of hippurate led to intra-individual differences in subject K (figure

12). The excretion ranged from 814.52 µM/mM creatinine to 1904.28 µM/mM creatinine, with a

lower level of hippurate during the course of the intervention (appendix A8). This is contradictory

with other studies, which have found increased level of hippurate after pre- or probiotic intake and

further associated this increase with an improvement of gut-health (Dewulf et al., 2013; Vázquez-

Fresno et al., 2014). The referred studies are not fully comparable, since one concerns obese women

and the other concerns women with mastitis, but it remains interesting that the level of a human-gut

co-metabolite is found in decreasing amounts during the current intervention.

This could indicate either changes in the composition and activities of the gut flora owing to the

aforementioned correlation to certain species, or simply changes due to diet as observed in other

studies (Krupp et al., 2012; Mulder et al., 2005; Walsh et al., 2006). Since the excretion of

hippurate varies to a great extend between K1 and K2 (representing day to day changes, appendix

A8), it remains more likely that the variation is due to diet, as likewise suggested in a study by

Walsh et al. (2006).

To sum up: the present pilot study did not allow for any clear elucidation of the effect of probiotics

by the use of 1H NMR based metabolomics. The metabolomic approach revealed great inter- and

intra-individual differences, especially concerning gut flora derived metabolites as hippurate,

TMAO and phenyls and diet related metabolites. A relation to intake of a probiotic drink cannot be

verified, since the results were not unanimously related to specific changes in urinary metabolic

profile and could not be fully substantiated by literature. Additionally, by normalizing the results of

the current study to creatinine excretion, it is not possible to elucidate clear associations.

41

It was therefore not possible to verify hypothesis 3, but it remains interesting that the use of labelled

compounds has proven to be a suitable biomarker of changes in microbial activity following mainly

prebiotic supplementation.

The most obvious metabolites contributing to differences were discussed. These differences only

account for 14 % and 28.3 % of intra-individual differences in subject U and K respectively,

indicating a relative minor impact on the total variance. It cannot be rejected that other metabolites

contributing to variation are linked to probiotic supplementation.

Literature indicates a relation to specific metabolites after pre- or probiotic ingestion, but there is a

great lack of human intervention studies. Based on table 3 it is additionally hard to decipher

consistency between the outcomes of studies using metabolomics, thus stressing a significance of

study design and the prebiotic or probiotic strain investigated.

Furthermore, the microbial activity in the colon is highly influenced by the nature of substrate

available as stressed in section a.i. As already discussed and indicated by literature, induction of

saccharolytic fermentation is dependent on appropriate carbohydrate sources e.g. in terms of

prebiotics. Thus, the intake of a synbiotic combination could lead to greater impact on microbial

activity.

c) Methodological considerations

c.i) Some considerations regarding NMR vs. MS Several review papers give insight in factors, which can be considered when comparing NMR

spectroscopy and MS, here among are Wishart (2014) and Lindon & Nicholson (2008).

Factors of importance in regards to the outcome and procedure of a metabolomic analysis include

sample size, preservation of sample integrity and sensitivity. Samples for NMR require relative

little preparation, however, there is a need of greater sample size (0.5µL) than GC-MS and LC-MS

(50µL and 10µL respectively) (Wishart, 2014). Thus, some biological samples are harder to analyse

by NMR (e.g. urine of mice) due to sample size. Though, NMR based metabolomic analysis is

widely used owing to a relatively easy data acquisition and the non-destructive nature, thereby

maintaining sample integrity.

Opposite, samples for GC-MS and LC-MS are most often chemically modified and derivatized to

gas phase and liquid phase respectively before analysis (Lindon & Nicholson, 2008).

Whereas NMR can detect 50-75 compounds in a human biofluid, GC-MS is assigned 50-100

compounds and LC-MS “hundreds to thousands of features” (Wishart, 2014). GC-MS and LC-MS

are therefore more sensitive techniques than NMR and therefore hypothesis 1 (NMR might be too

insensitive to detect SCFAs in urine) cannot be excluded.

42

However, an advance concerning NMR is the ability to detect molecular structures rather than ion

masses. In MS analysis, the output corresponds to ion masses, which might fit multiple compounds

and therefore be harder to identify and separate (Claus & Swann, 2013).

Other noteworthy factors relevant for both NMR and MS include differences in pH, temperature

and salt, since these can influence the chemical analysis (Wishart, 2008).

c.ii) Some considerations regarding study design Referring hypothesis 2, the present study design has multiple drawbacks, and only some aspects

will be elaborated. First: the pilot study includes only two participants, uses no controls and only

spans the course of 21 days. The design is therefore not statistically reasoned and the findings

cannot be generalized. Additionally, a lack of criteria for the participants in regards to factors such

as lifestyle and diet presents a challenge when wanting to elucidate biases. However, the producer

of the probiotic drink involved claims that the drink is beneficial “as part of a regular healthy diet”

(appendix A2), therefore screening and criteria assessment should prove to be negligible, since

potential health benefits are be mutual between healthy subjects. However, neither the current study

nor the producer has clearly defined term “healthy”, but this thesis has sought to elucidate different

health effects observed by other studies.

As indicated by the former discussion on the identified discriminators in the present study, it is

difficult to distinguish whether a change in metabolite composition is due to changes in gut flora

activity induced by intake of probiotics, or induced by differences in diet. Hence, the study would

benefit from a standardized diet or at least a dietary record in order to identify potential biases.

Additionally, factors as age, weight, gender, etc. must be taken into account, as it is evident that

these influence the metabolic profile of urine. A cross-over, placebo controlled study with a greater

number of participants and approval of the Ethical Comitee in accordance with the Declaration of

Helsinki would provide more statistical power and a validation for an intervention.

Multiple chemical techniques and multiple biological sources can be used in order to interpret the

outcome of an intervention, as demonstrated by studies in table 3. In the current study, this notion is

relevant in regards to hypothesis 2, since the use of one biological source and one chemical

technique is evidently insufficient in order to achieve the ideal goal of elucidating an effect of an

intervention on the entire metabolome (Xie et al., 2013).

43

5 Conclusion and perspectives The present pilot study demonstrates the use of 1H NMR urine metabolomics to investigate the

effect of probiotic intake on urinary metabolic profile. The study found variations in creatinine,

citrate, hippurate and trigonelline excretions during the course of the intervention.

However, the variations could not reasonably be related to intake of probiotics. Multiple

methodological factors must be improved in order to fully confirm these findings. Other

metabolomics studies prove significant but inconsistent results dependent on the supplement

consumed and study design. However, several of these studies have assigned certain metabolites as

biomarkers of changed microbial activity in the gut, and indicators of modulations in biochemical

pathways in the host following an intervention. Hence, metabolomics has proved promising

potential in improving our understanding of the biochemical mechanisms behind pre- and/or

probiotic activities. Nevertheless, there is a need of greater human intervention trials and a potential

in holistic integration of multiple biological compartments, when wanting to understand the

complete biochemical effects of pre- and probiotic action on host’s health.

6 Acknowledgements I will like to thank my supervisor Ulrik Sundekilde and co-supervisor Jette F. Young for taken the

time to help me during the making of this bachelor thesis. Special thanks to Ulrik Sundekilde for

given great guidance and for volunteering as a participant in the pilot-study. Thanks to Christian

Clement Yde for initially setting up the project.

44

7 Literature Allison, C., & Macfarlane, G. T. (1989). Influence of pH, nutrient availability, and growth-rate on amine production by

Bacteroides fragilis and Clostridium perfringens. Applied and Environmental Microbiology, 55(11), 2894–2898.

Alonso, A., Marsal, S., & Julià, A. (2015). Analytical methods in untargeted metabolomics: state of the art in 2015.

Frontiers in Bioengineering and Biotechnology, 3(March), 23. https://doi.org/10.3389/fbioe.2015.00023

Arora, T., Singh, S., & Sharma, R. K. (2013). Probiotics: Interaction with gut microbiome and antiobesity potential.

Nutrition, 29(4), 591–596. https://doi.org/10.1016/j.nut.2012.07.017

Arumugam, M., Raes, J., Pelletier, E., Le Paslier, D., Yamada, T., Mende, D. R., … Bork, P. (2011). Enterotypes of the

human gut microbiome. Nature, 473(7346), 174–80. https://doi.org/10.1038/nature09944

Ashihara, H., Ludwig, I. A., Katahira, R., Yokota, T., Fujimura, T., & Crozier, A. (2014). Trigonelline and related

nicotinic acid metabolites: occurrence, biosynthesis, taxonomic considerations, and their roles in planta and in

human health. Phytochemistry Reviews, 14(5), 765–798. https://doi.org/10.1007/s11101-014-9375-z

Ataie-Jafari, A., Larijani, B., Alavi Majd, H., & Tahbaz, F. (2009). Cholesterol-lowering effect of probiotic yogurt in

comparison with ordinary yogurt in mildly to moderately hypercholesterolemic subjects. Annals of Nutrition and

Metabolism, 54(1), 22–27. https://doi.org/10.1159/000203284

Beckonert, O., Keun, H. C., Ebbels, T. M. D., Bundy, J. G., Holmes, E., Lindon, J. C., & Nicholson, J. K. (2007).

Metabolic profiling, metabolomic and metabonomic procedures for NMR spectroscopy of urine, plasma, serum

and tissue extracts. Nature Protocols, 2, 2692–2703. https://doi.org/10.1038/nprot.2007.376

Bernini, L. J., Simao, A. N. C., Alfieri, D. F., Lozovoy, M. A. B., Mari, N. L., de Souza, C. H. B., … Costa, G. N.

(2016). Beneficial effects of Bifidobacterium lactis on lipid profile and cytokines in patients with metabolic

syndrome: A randomized trial. Effects of probiotics on metabolic syndrome. Nutrition, 32(6), 716–719.

https://doi.org/10.1016/j.nut.2015.11.001

Bijlsma, S., Bobeldijk, I., Verheij, E. R., Ramaker, R., Kochhar, S., Macdonald, I. A., … Smilde, A. K. (2006). Large-

Scale Human Metabolomics Studies : A Strategy for Data ( Pre- ) Processing and Validation Large-Scale Human

Metabolomics Studies : A Strategy for Data ( Pre- ) Processing and Validation, 78(2), 567–574.

https://doi.org/10.1021/ac051495j

Blümich, B., & Callaghan, P. T. (1995). Principles of Nuclear Magnetic Resonance Microscopy. Magnetic Resonance

Chemistry, 33, 322–322. https://doi.org/10.1002/mrc.1260330417

Boets, E., Gomand, S. V, Deroover, L., Preston, T., Vermeulen, K., De Preter, V., … Verbeke, K. A. (2016). Systemic

availability and metabolism of colonic-derived short-chain fatty acids in healthy subjects - a stable isotope study.

The Journal of Physiology, 0, 1–15. https://doi.org/10.1113/JP272613

Bothwell, J. H. F., & Griffin, J. L. (2011). An introduction to biological nuclear magnetic resonance spectroscopy.

Biological Reviews, 86(2), 493–510. https://doi.org/10.1111/j.1469-185X.2010.00157.x

Bouatra, S., Aziat, F., Mandal, R., Guo, A. C., Wilson, M. R., Knox, C., … Wishart, D. S. (2013). The Human Urine

Metabolome. PLoS ONE, 8(9). https://doi.org/10.1371/journal.pone.0073076

Brasili, E., Mengheri, E., Tomassini, A., Capuani, G., Roselli, M., Finamore, A., … Miccheli, A. (2013).

[i]Lactobacillus acidophilus[/i] La5 and [i]Bifidobacterium lactis[/i] Bb12 induce different age-related metabolic

profiles revealed by [up]1[/up]H-NMR spectroscopy in urine and feces of mice. The Journal of Nutrition,

143(10), 1549–57. https://doi.org/10.3945/jn.113.177105.with

Brigidi, P., Vitali, B., Swennen, E., Bazzocchi, G., & Matteuzzi, D. (2001). Effects of probiotic administration upon the

composition and enzymatic activity of human fecal microbiota in patients with irritable bowel syndrome or

functional diarrhea. Res.Microbiol., 152(0923–2508), 735–741.

45

Bringiotti, R., Ierardi, E., Lovero, R., Losurdo, G., Di Leo, A., & Principi, M. (2014). Intestinal microbiota: The

explosive mixture at the origin of inflammatory bowel disease? World Journal of Gastrointestinal

Pathophysiology, 5(4), 550–9. https://doi.org/10.4291/wjgp.v5.i4.550

Calvani, R., Miccheli, A., Capuani, G., Tomassini Miccheli, A., Puccetti, C., Delfini, M., … Mingrone, G. (2010). Gut

microbiome-derived metabolites characterize a peculiar obese urinary metabotype. Int J Obes (Lond), 34(6),

1095–1098. https://doi.org/ijo201044 [pii]\r10.1038/ijo.2010.44

Chemistry, R. S. of. (n.d.). Introduction to NMR spectroscopy. Retrieved April 9, 2017, from http://www.rsc.org/learn-

chemistry/wiki/Introduction_to_NMR_spectroscopy

Claridge, T. D. W. (1999). High-resolution NMR techniques in organic chemistry. Chapter 8: Correlations through

space: The nuclear Overhauser effect (2nd ed., Vol. 19). Retrieved from http://www.gbv.de/dms/ohb-

opac/30584704X.pdf

Claus, S. P., & Swann, J. R. (2013). Nutrimetabonomics: Applications for Nutritional Sciences, with Specific Reference

to Gut Microbial Interactions. Annu. Rev. Food Sci. Technol., 4(December 2012), 381–399.

https://doi.org/10.1146/annurev-food-030212-182612

Craig, A., Cloarec, O., Holmes, E., Nicholson, J. K., & Lindon, J. C. (2006). Scaling and normalization effects in NMR

spectroscopic metabonomic data sets. Analytical Chemistry, 78(7), 2262–2267. https://doi.org/10.1021/ac0519312

Cummings, J. H., & Macfarlane, G. T. (1991). The control and consequences of bacterial fermenation in the human

colon. Journal of Applied …, (Wiggins 1983), 443–459. Retrieved from

http://onlinelibrary.wiley.com/doi/10.1111/j.1365-2672.1991.tb02739.x/abstract

Cummings, J. H., & Macfarlane, G. T. (1997). Role of intestinal bacteria in nutrient metabolism. Clinical Nutrition, 16,

3–11. Retrieved from papers2://publication/uuid/3172BFDE-68FC-49A0-AB0A-A4D2EF55A3BE

Cummings, J. H., Pomare, E. W., Branch, H. W. J., Naylor, C. P. E., & Macfarlane, T. (1987). Short chain fatty acids in

human large intestine , portal , hepatic and venous blood. https://doi.org/10.1136/gut.28.10.1221

Dawiskiba, T., Deja, S., Mulak, A., Zabek, A., Jawień, E., Pawełka, D., … Młynarz, P. (2014). Serum and urine

metabolomic fngerprinting in diagnostics of inflammatory bowel diseases. World Journal of Gastroenterology,

20(1), 163–174. https://doi.org/10.3748/wjg.v20.i1.163

Dewulf, E. M., Cani, P. D., Claus, S. P., Fuentes, S., Puylaert, P. G., Neyrinck, A. M., … Delzenne, N. M. (2013).

Insight into the prebiotic concept: lessons from an exploratory, double blind intervention study with inulin-type

fructans in obese women. Gut, 62(8), 1112–1121. https://doi.org/10.1136/gutjnl-2012-303304

Dona, A. C., Kyriakides, M., Scott, F., Shephard, E. A., Varshavi, D., Veselkov, K., & Everett, J. R. (2016). A guide to

the identification of metabolites in NMR-based metabonomics/metabolomics experiments. Computational and

Structural Biotechnology Journal, 14, 135–153. https://doi.org/10.1016/j.csbj.2016.02.005

Dumas, M. E., Maibaum, E. C., Teague, C., Ueshima, H., Zhou, B., Lindon, J. C., … Holmes, E. (2006). Assessment of

analytical reproducibility of 1H NMR spectroscopy based metabonomics for large-scale epidemiological

research: The INTERMAP study. Analytical Chemistry, 78(7), 2199–2208. https://doi.org/10.1021/ac0517085

Ebel, B., Lemetais, G., Beney, L., Cachon, R., Langella, P., Gervais, P., … Langella, P. (2014). Impact of Probiotics on

Risk Factors for Cardiovascular Diseases . A Review Impact of Probiotics on Risk Factors for Cardiovascular

Diseases . A Review, 8398(January 2017). https://doi.org/10.1080/10408398.2011.579361

Eckburg, Paul B.; Bik, Elisabeth M.; Bernstein, Charles N.; Purdom, Elizabeth; Dethlefsen, Les; Sargent, Michael; Gill,

Steven R.; Nelson, Karen E.; Relman, D. A. . (2005). Diversity of the Human Intestinal Microbial Flora. Science,

308(5728), 1635–1638. https://doi.org/10.1126/science.1110591

Eriksson, L., Johansson, E., Kettaneh-Wold, N., Trygg, J., Wikström, C., & Wold, S. (2006). PCA. In Multi- and

46

Megavariate Data Analysis (pp. 43–70).

Ewaschuk, J. B., Diaz, H., Meddings, L., Diederichs, B., Dmytrash, A., Backer, J., … Madsen, K. L. (2008). Secreted

bioactive factors from, 8, 1025–1034. https://doi.org/10.1152/ajpgi.90227.2008.

FAO (Food and Agriculture Organization), & WHO. (2006). Probiotics in food FOOD AND NUTRITION. In

Probiotics in Food: Nutrition Paper 85.

Fijan, S. (2014). Microorganisms with claimed probiotic properties: An overview of recent literature. International

Journal of Environmental Research and Public Health, 11(5), 4745–4767.

https://doi.org/10.3390/ijerph110504745

Flaherty, S. O., Saulnier, D., Pot, B., Versalovic, J., Flaherty, S. O., Saulnier, D., … Versalovic, J. (2015). immunity ?

How can probiotics and prebiotics impact mucosal immunity ? Gut Microbes, 976(October), 293–300.

https://doi.org/10.4161/gmic.1.5.12924

Fontana, L., Bermudez-Brito, M., Plaza-Diaz, J., Munoz-Quezada, S., & Gil, A. (2013). Sources, isolation,

characterisation and evaluation of probiotics. Br J Nutr, 109 Suppl, S35-50. https://doi.org/S0007114512004011

[pii]\r10.1017/S0007114512004011

Fuentes, M. C., Lajo, T., Carrión, J. M., & Cuñé, J. (2012). Cholesterol-lowering efficacy of Lactobacillus plantarum

CECT 7527, 7528 and 7529 in hypercholesterolaemic adults. British Journal of Nutrition, 1–7.

https://doi.org/10.1017/S000711451200373X

Gibson, G. R., Macfarlane, S., & Macfarlane, G. T. (1993). Metabolic Interactions Involving Sulfate-Reducing and

Methanogenic Bacteria in the Human Large-Intestine. Fems Micriobiology Ecology, 12(2), 117–125.

Gibson, G. R., & Roberfroid, M. B. (1995). Dietary modulation of the human colonic microbiota: introducing the

concept of prebiotics. The Journal of Nutrition, 125(6), 1401–1412. https://doi.org/10.1079/NRR200479

Gibson, G. R., & Roberfroid, M. B. (2004). Dietary modulation of the human colonic microbiota: introducing the

concept of prebiotics. The Journal of Nutrition, 125(6), 1401–1412. https://doi.org/10.1079/NRR200479

Gibson, G. R., Scott, K. P., Rastall, R. A., Tuohy, K. M., Hotchkiss, A., Dubert-Ferrandon, A., … Buddington, R.

(2010). Dietary prebiotics: current status and new definition. Food Science and Technology Bulletin: Functional

Foods, 7(1), 1–19. https://doi.org/10.1616/1476-2137.15880.

Goldin, B. R., & Gorbach, S. L. (1984). Alterations of the intestinal microflora by diet, oral antibiotics, and

lactobacillus: Decreased production of free amines from aromatic nitro compounds, azo dyes, and glucuronides.

Journal of the National Cancer Institute, 73(3), 689–695. https://doi.org/10.1093/jnci/73.3.689

Goodacre, R., Broadhurst, D., Smilde, A. K., Kristal, B. S., Baker, J. D., Beger, R., … Wulfert, F. (2007). Proposed

minimum reporting standards for data analysis in metabolomics. Metabolomics, 3(3), 231–241.

https://doi.org/10.1007/s11306-007-0081-3

Guarner, F., & Schaafsma, G. J. (1998). Probiotics, 39(November 1995), 237–238.

Hamer, H. M., Jonkers, D., Venema, K., Vanhoutvin, S., Troost, F. J., & Brummer, R. J. (2008). Review article: The

role of butyrate on colonic function. Alimentary Pharmacology and Therapeutics, 27(2), 104–119.

https://doi.org/10.1111/j.1365-2036.2007.03562.x

Holmes, E., Foxall, P. J. D., Nicholson, J. K., Nield, G. H., Brown, S. M., Beddell, C. R., … Neidig, P. (1994).

Automatic Data Reduction and Pattern Recognition Methods for Analysis of H Nuclear Magnetic Resonance

Spetra of Human Urin from Normal and Pathological States. Analytical Biochemistry, (220), 284–296.

Holmes, E., Li, J. V, Athanasiou, T., Ashrafian, H., & Nicholson, J. K. (2011). Understanding the role of gut

microbiome – host metabolic signal disruption in health and disease. Trends in Microbiology, 19(7), 349–359.

https://doi.org/10.1016/j.tim.2011.05.006

47

Holmes, E., Wilson, I. D., & Nicholson, J. K. (2008). Metabolic Phenotyping in Health and Disease. Cell, 134(5), 714–

717. https://doi.org/10.1016/j.cell.2008.08.026

Hornak, J. P. (1997). Spin physics. Retrieved May 11, 2017, from https://www.cis.rit.edu/htbooks/nmr/inside.htm

Jones, M. L., Martoni, C. J., & Prakash, S. (2012). Cholesterol lowering and inhibition of sterol absorption by

Lactobacillus reuteri NCIMB 30242: a randomized controlled trial. European Journal of Clinical Nutrition,

66(11), 1234–1241. https://doi.org/10.1038/ejcn.2012.126

Kallioma, M., Antoine, J., Herz, U., Rijkers, G. T., Wells, J. M., & Mercenier, A. (2010). Guidance for Substantiating

the Evidence for Beneficial Effects of Probiotics : Prevention and Management of Allergic. The Journal of

Nutrition, 140, 713–721. https://doi.org/10.3945/jn.109.113761.713S

Kanehisa, M., Furumichi, M., Tanabe, M., Sato, Y., & Morishima, K. (2017). KEGG: New Perspectives on Genomes,

Pathways, Diseases and Drugs. Nucleic Acids Research, 45(Database Issue), 353–361.

https://doi.org/10.1093/nar/gkwl1092

Kaplan, H., & Hutkins, R. W. (2000). Fermentation of Fructooligosaccharides by Lactic Acid Bacteria and

Bifidobacteria Fermentation of Fructooligosaccharides by Lactic Acid Bacteria and Bifidobacteria †, 66(6), 6–9.

https://doi.org/10.1128/AEM.66.6.2682-2684.2000.Updated

Kochhar, S., Jacobs, D. M., Ramadan, Z., Berruex, F., Fuerholz, A., & Fay, L. B. (2006). Probing gender-specific

metabolism differences in humans by nuclear magnetic resonance-based metabonomics. Analytical Biochemistry,

352(2), 274–281. https://doi.org/10.1016/j.ab.2006.02.033

Krupp, D., Doberstein, N., Shi, L., & Remer, T. (2012). Hippuric Acid in 24-Hour Urine Collections Is a Potential

Biomarker for Fruit and Vegetable Consumption in Healthy Children and Adolescents. Journal of Nutrition,

142(7), 1314–1320. https://doi.org/10.3945/jn.112.159319

Lamichhane, S., Yde, C. C., Forssten, S., Ouwehand, A. C., Saarinen, M., Jensen, H. M., … Bertram, H. C. (2014).

Impact of dietary polydextrose fiber on the human gut metabolome. Journal of Agricultural and Food Chemistry,

62(40), 9944–9951. https://doi.org/10.1021/jf5031218

Lancashire, R. J. (2003). Introduction to NMR spectroscopy, proton assignment chart. Retrieved September 10, 2017,

from http://wwwchem.uwimona.edu.jm/spectra/nmrintro.html

Lane, A. N. (2012). Principles of NMR for Applications in Metabolomics. In Teresa Whei-Mei Fan et al (Ed.), The

Handbook of Metabolomics (pp. 127–197). https://doi.org/10.1007/978-1-61779-618-0

Lang, R., Wahl, A., Skurk, T., FiratYagar, E., Schmiech, L., Eggers, R., … Hofmann, T. (2010). Development of a

hydrophilic liquid interaction chromatography-high- performance liquid chromatography-tandem mass

spectrometry based stable isotope dilution analysis and pharmacokinetic studies on bioactive pyridines in human

plasma and urine after coffe. Analytical Chemistry, 82(4), 1486–1497. https://doi.org/10.1021/ac902616k

Lang, R., Wahl, A., Stark, T., & Hofmann, T. (2011). Urinary N-methylpyridinium and trigonelline as candidate dietary

biomarkers of coffee consumption. Molecular Nutrition and Food Research, 55(11), 1613–1623.

https://doi.org/10.1002/mnfr.201000656

Lee, H. C., Jenner, A. M., Low, C. S., & Lee, Y. K. (2006). Effect of tea phenolics and their aromatic fecal bacterial

metabolites on intestinal microbiota. Research in Microbiology, 157(9), 876–884.

https://doi.org/10.1016/j.resmic.2006.07.004

Lenz, E. M., Bright, J., Wilson, I. D., Hughes, A., Morrisson, J., Lindberg, H., & Lockton, A. (2004). Metabonomics,

dietary influences and cultural differences: A 1H NMR-based study of urine samples obtained from healthy

British and Swedish subjects. Journal of Pharmaceutical and Biomedical Analysis, 36(4), 841–849.

https://doi.org/10.1016/j.jpba.2004.08.002

48

Li, M., Wang, B., Zhang, M., Rantalainen, M., Wang, S., Zhou, H., … Zhao, L. (2008). Symbiotic gut microbes

modulate human metabolic phenotypes. Proceedings of the National Academy of Sciences, 105(6), 2117–2122.

https://doi.org/10.1073/pnas.0712038105

Lidbeck, A., Overvikf, E., Rafter, J., Nord, C. E., & Gustafssonf, J. (1992). Effect of Lactobacillus acidophilus

Supplements on Mutagen Excretion in Faeces and Urine in Humans. Microbial Ecology in Health and Disease,

5(March), 59–67. https://doi.org/10.3109/08910609209141305

Lin, S. H., Chou, L. M., Chien, Y. W., Chang, J. S., & Lin, C. I. (2016). Prebiotic Effects of Xylooligosaccharides on

the Improvement of Microbiota Balance in Human Subjects. Gastroenterology Research and Practice, 2016.

https://doi.org/10.1155/2016/5789232

Lindon, J. C. (2004). Metabonomics – Techniques and Applications. Buisness Briefing: Future Drug Discovery,

1(1999), 12–17.

Lindon, J. C., & Nicholson, J. K. (2008). Spectroscopic and statistical techniques for information recovery in

metabonomics and metabolomics. Annual Review of Analytical Chemistry (Palo Alto, Calif.), 1, 45–69.

https://doi.org/10.1146/annurev.anchem.1.031207.113026

Lindon, J. C., Nicholson, J. K., Holmes, E., & Everett, J. R. (2000). Metabonomics: metabolic processes studied by

NMR spectroscopy of biofluids. Concepts in Magnetic Resonance, 12(5), 289–320. https://doi.org/10.1002/1099-

0534(2000)12:5<289::AID-CMR3>3.0.CO;2-W

Macfarlane, G. T., & Macfarlane, S. (2012). Bacteria, Colonic Fermentation and Gastrointestinal Healt. Journal of

AOAC …, 95(1), 50–60. https://doi.org/10.5740/jaoacint.SGE

Manas, M; Martinez de Victoria, E; Gil, A; Yago, M; Mathers, J. (n.d.). Nutrition and Metabolism. In H. M. Gibney,

Michael J; Macdonald, Ian A; Roche (Ed.) (pp. 218–222).

Martin, F. P. J., Collino, S., Rezzi, S., & Kochhar, S. (2012). Metabolomic applications to decipher gut microbial

metabolic influence in health and disease. Frontiers in Physiology, 3 APR(April), 1–11.

https://doi.org/10.3389/fphys.2012.00113

Martin, F. P. J., Sprenger, N., Yap, I. K. S., Wang, Y., Bibiloni, R., Rochat, F., … Nicholson, J. K. (2009).

Panorganismal Gut Microbiome - Host Metabolic Crosstalk research articles. Journal of Proteome Research, 8,

2090–2105. https://doi.org/10.1021/pr801068x

Martin, F. P. J., Wang, Y., Sprenger, N., Yap, I. K., Lundstedt, T., Lek, P., … Nicholson, J. K. (2008a). Probiotic

modulation of symbiotic gut microbial-host metabolic interactions in a humanized microbiome mouse model. Mol

Syst Biol, 4(157), 157. https://doi.org/10.1038/msb4100190

Martin, F. P. J., Wang, Y., Sprenger, N., Yap, I. K. S., Rezzi, S., Ramadan, Z., … Nicholson, J. K. (2008b). Top-down

systems biology integration of conditional prebiotic modulated transgenomic interactions in a humanized

microbiome mouse model. Molecular Systems Biology, 4(205), 205. https://doi.org/10.1038/msb.2008.40

Metchnikoff, E. (1908). The Prolongation of Life. Optimistic Studies. (P. C. Mitchell, Ed.). G. P. Putnam’s Sons.

Mulder, T. P., Rietveld, A. G., & van Amelsvoort, J. M. (2005). Consumption of both black tea and green tea results in

an increase in the excretion of hippuric acid into urine...Proceedings of the 1st International Conference on

Polyphenols and Health held in Vichy, France, November 18-21, 2004. American Journal of Clinical Nutrition,

81(1), 256S–60S 1p. Retrieved from

http://search.ebscohost.com/login.aspx?direct=true&db=rzh&AN=106632009&site=ehost-live

Murgas Torrazza, R., & Neu, J. (2011). The developing intestinal microbiome and its relationship to health and disease

in the neonate. Journal of Perinatology : Official Journal of the California Perinatal Association, 31 Suppl 1(S1),

S29-34. https://doi.org/10.1038/jp.2010.172

49

Ndagijimana, M., Laghi, L., Vitali, B., Placucci, G., Brigidi, P., & Guerzoni, M. E. (2009). Effect of a synbiotic food

consumption on human gut metabolic profiles evaluated by 1H Nuclear Magnetic Resonance spectroscopy.

International Journal of Food Microbiology, 134(1–2), 147–153.

https://doi.org/10.1016/j.ijfoodmicro.2009.04.016

Nicholls, A. W., Mortishire-Smith, R. J., & Nicholson, J. K. (2003). NMR Spectroscopic-Based metabonomic Studies

of Urinary Metabolite Variation in Accliatizing Germ-Free Rats. Chemical Research in Toxiology, 16.

https://doi.org/10.1021/tx0340293

Nicholson, J. K., Holmes, E., Kinross, J., Burcelin, R., Gibson, G., Jia, W., & Pettersson, S. (2012). Metabolic

Interactions, 108(June), 1262–1268. https://doi.org/10.1126/science.1223813

Nicholson, J. K., Holmes, E., Lindon, J. C., & Wilson, I. D. (2004). The challenges of modeling mammalian

biocomplexity. Nature Biotechnology, 22(10), 1268–1274. https://doi.org/10.1038/nbt1015

Nicholson, J. K., Holmes, E., & Wilson, I. D. (2005). Gut microorganisms, mammalian metabolism and personalized

health care. Nature Reviews. Microbiology, 3(5), 431–438. https://doi.org/10.1038/nrmicro1152

Nicholson J K, Lindon J C & Holmes E, (1999) “Metabonomics: Understanding the Metabolic Responses of Living

Systems to Pathophysiological Stimuli via Multivariate Statistical Analysis of Biological NMR Spectroscopic

Data”, Xenobiotica, 29, pp. 1,181–1,189

Ohland, C. L., & Macnaughton, W. K. (2010). Probiotic bacteria and intestinal epithelial barrier function. American

Journal of Gastrointestinal Liver Physiology, 298(167), G807-819. https://doi.org/10.1152/ajpgi.00243.2009.

Ouwehand, A. C., Salminen, S., & Isolauri, E. (2002). Probiotics an overview of beneficial effects.pdf, 279–289.

Pauli, G. F., Jaki, B. U., & Lankin, D. C. (2005). Quantitative 1 H NMR: Development and Potential of a Method for

Natural Products Analysis §. Journal of Natural Products, 68(1), 133–149. https://doi.org/10.1021/np0497301

Pavia, D. L., Lampman, G. M., & Kriz, G. S. (2001). Nuclear Magnetic Resonance Spectroscopy Part one: Basic

concepts. In Introduction to Spectroscopy (3rd ed., pp. 102–167).

Pedersen, S. M. M., Nebel, C., Nielsen, N. C., Andersen, H. J., Olsson, J., Simrén, M., … Malmendal, A. (2011). A

GC-MS-based metabonomic investigation of blood serum from irritable bowel syndrome patients undergoing

intervention with acidified milk products. European Food Research and Technology, 233(6), 1013–1021.

https://doi.org/10.1007/s00217-011-1599-1

Pereira, D. I. a, & Gibson, G. R. (2002). Cholesterol Assimilation by Lactic Acid Bacteria and Bifidobacteria Isolated

from the Human Gut Cholesterol Assimilation by Lactic Acid Bacteria and Bifidobacteria Isolated from the

Human Gut, 68(9), 1–6. https://doi.org/10.1128/AEM.68.9.4689

Poulsen, K. O. (2017). Photograph of commercial yoghurt.

Power, S. E., O’Toole, P. W., Stanton, C., Ross, R. P., & Fitzgerald, G. F. (2013). Intestinal microbiota, diet and health.

British Journal of Nutrition, 111(3), 387–402. https://doi.org/10.1017/S0007114513002560

Preter, V. De, Geboes, K., Verbrugghe, K., De Vuyst, L., Vanhoutte, T., Huys, G., … Verbeke, K. (2004). The in vivo

use of the stable isotope-labelled biomarkers lactose-[15N]ureide and [2H4]tyrosine to assess the effects of pro-

and prebiotics on the intestinal flora of healthy human volunteers. The British Journal of Nutrition, 92(3), 439–

446. https://doi.org/10.1079/BJN20041228

Preter, V. De, Hamer, H. M., Windey, K., & Verbeke, K. (2011). The impact of pre- and/or probiotics on human

colonic metabolism: Does it affect human health? Molecular Nutrition and Food Research, 55(1), 46–57.

https://doi.org/10.1002/mnfr.201000451

Preter, V. De, Raemen, H., Cloetens, L., Houben, E., Rutgeerts, P., & Verbeke, K. (2008). Effect of dietary intervention

with different pre- and probiotics on intestinal bacterial enzyme activities. European Journal of Clinical

50

Nutrition, 62(2), 225–231. https://doi.org/10.1038/sj.ejcn.1602706

Preter, V. De, Vanhoutte, T., Huys, G., Swings, J., Vuyst, L. De, Rutgeerts, P., & Verbeke, K. (2007). Oligofructose-

Enriched Inulin on Colonic Nitrogen-Protein Metabolism in Healthy Humans, 358–368.

https://doi.org/10.1152/ajpgi.00052.2006.

Psihogios, N. G., Gazi, I. F., Elisaf, M. S., Seferiadis, K. I., & Bairaktari, E. T. (2008). Gender-related and age-related

urinalysis of healthy subjects by NMR-based metabonomics. NMR in Biomedicine, 21, 195–207.

https://doi.org/10.1002/nbm.1176

Rajkumar, H., Kumar, M., Das, N., Kumar, S. N., Challa, H. R., & Nagpal, R. (2015). Effect of Probiotic Lactobacillus

salivarius UBL S22 and Prebiotic Fructo-oligosaccharid on Serum Lipids, Inflammatory Markers, Insulin

Sensitivity, and Gut Bacteria in Healthy Young Volunteers: A Randomized Controlled Single-Blind Pilot Study.

Journal of Cardiovascular Pharmacology and Therapeutics, 20(3), 289–298.

https://doi.org/10.1177/1074248414555004

Rasmussen, L. G., Savorani, F., Larsen, T. M., Dragsted, L. O., Astrup, A., & Engelsen, S. B. (2011). Standardization

of factors that influence human urine metabolomics. Metabolomics, 7(1), 71–83. https://doi.org/10.1007/s11306-

010-0234-7

Rasmussen, L. G., Winning, H., Savorani, F., Toft, H., Larsen, T. M., Dragsted, L. O., … Engelsen, S. B. (2012).

Assessment of the effect of high or low protein diet on the human urine metabolome as measured by NMR.

Nutrients, 4(2), 112–131. https://doi.org/10.3390/nu4020112

Ravel, J., Gajer, P., Abdo, Z., Schneider, G. M., Koenig, S. S. K., McCulle, S. L., … Forney, L. J. (2011). Vaginal

microbiome of reproductive-age women. Proceedings of the National Academy of Sciences, 108(Supplement_1),

4680–4687. https://doi.org/10.1073/pnas.1002611107

Roberfroid, M. B., Gibson, G. R., Hoyles, L., McCartney, A. L., Rastall, R. A., Rowland, I., … Meheust, A. (2010).

Prebiotics concept and health. British Journal of Nutrition, 106(S2), 1–63.

https://doi.org/https://doi.org/10.1017/S0007114510003363

Roessner, U., & Bowne, J. (2009). What is metabolomics all about? BioTechniques, 46(5 SPEC. ISSUE), 363–365.

https://doi.org/10.2144/000113133

Rose, C., Parker, A., Jefferson, B., & Cartmell, E. (2015). The Characterization of Feces and Urine: A Review of the

Literature to Inform Advanced Treatment Technology. Critical Reviews in Environmental Science and

Technology, 45(17), 1827–1879. https://doi.org/10.1080/10643389.2014.1000761

Ross, A. B., Pere-Trepat, E., Montoliu, I., Martin, F.-P. J., Collino, S., Moco, S., … Kochhar, S. (2013). A Whole-

Grain-Rich Diet Reduces Urinary Excretion of Markers of Protein Catabolism and Gut Microbiota Metabolism in

Healthy Men after One Week. Journal of Nutrition, 143(6), 766–773. https://doi.org/10.3945/jn.112.172197

Ross, A., Schlotterbeck, G., Dieterla, F., & Senn, H. (2007). NMR Spectroscopy Techniques. In J. C. Lindon, J. K.

Nicholson, & E. Holmes (Eds.), The Handbook of Metabolonomics and Metabolomics (1st ed., pp. 55–113).

Elsevier B.V.

Rothwell, J. A., Fillâtre, Y., Martin, J. F., Lyan, B., Pujos-Guillot, E., Fezeu, L., … Manach, C. (2014). New

biomarkers of coffee consumption identified by the non-targeted metabolomic profiling of cohort study subjects.

PLoS ONE, 9(4). https://doi.org/10.1371/journal.pone.0093474

Ryan, E. P., Heuberger, A. L., Broeckling, C. D., Borresen, E. C., Tillotson, C., & Prenni, J. E. (2013). Advances in

Nutritional Metabolomics. Current Metabolomics, 1(2), 109–120. https://doi.org/10.2174/2213235X11301020001

Saude, E. J., & Sykes, B. D. (2007). Urine stability for metabolomic studies: Effects of preparation and storage.

Metabolomics, 3(1), 19–27. https://doi.org/10.1007/s11306-006-0042-2

51

Saude, E. J., Adamko, D., Rowe, B. H., Marrie, T., & Sykes, B. D. (2007). Variation of metabolites in normal human

urine. Metabolomics, 3(4), 439–451. https://doi.org/10.1007/s11306-007-0091-1

Scheppach, W., Luehrs, H., & Menzel, T. (2001). Beneficial health effects of low-digestible carbohydrate consumption.

The British Journal of Nutrition, 85 Suppl 1, S23–S30. https://doi.org/10.1079/BJN2000259

Scott, K. P., Duncan, S. H., & Flint, H. J. (2008). Dietary fibre and the gut microbiota. Nutrition Bulletin, 33(3), 201–

211. https://doi.org/10.1111/j.1467-3010.2008.00706.x

Shah, N. P. (2007). Functional cultures and health benefits. International Dairy Journal, 17, 1262–1277.

https://doi.org/10.1016/j.idairyj.2007.01.014

Slupsky, C. M., Rankin, K. N., Wagner, J., Fu, H., Chang, D., Weljie, A. M., … Marrie, T. J. (2007). Investigations of

the effects of gender, diurnal variation, and age in human urinary metabolomic profiles. Anal Chem, 79(18),

6995–7004. https://doi.org/10.1021/ac0708588

Smith, E. A., & Macfarlane, G. T. (1996). Enumeration of human colonic bacteria producing phenolic and indolic

compounds: effects of pH, carbohydrate availability and retention time on dissimilatory aromatic amino acid

metabolism. The Journal of Applied Bacteriology, 81(3), 288–302. https://doi.org/10.1111/j.1365-

2672.1996.tb04331.x

Smolinska, A., Blanchet, L., Buydens, L. M. C., & Wijmenga, S. S. (2012). NMR and pattern recognition methods in

metabolomics: From data acquisition to biomarker discovery: A review. Analytica Chimica Acta, 750, 82–97.

https://doi.org/10.1016/j.aca.2012.05.049

Stella, C., Beckwith-Hall, B., Cloarec, O., Holmes, E., Lindon, J., Powell, J., … Nicholson, J. (2006). Susceptibility of

human metabolic phenotypes to dietary modulation. J. Proteome Research, 5(10), 2780–2788.

Sun, J., Schnackenberg, L. K., Holland, R. D., Schmitt, T. C., Cantor, G. H., Dragan, Y. P., & Beger, R. D. (2008).

Metabonomics evaluation of urine from rats given acute and chronic doses of acetaminophen using NMR and

UPLC/MS. Journal of Chromatography B: Analytical Technologies in the Biomedical and Life Sciences, 871(2),

328–340. https://doi.org/10.1016/j.jchromb.2008.04.008

Tabasco, R., Fernández de Palencia, P., Fontecha, J., Peláez, C., & Requena, T. (2014). Competition mechanisms of

lactic acid bacteria and bifidobacteria: Fermentative metabolism and colonization. LWT - Food Science and

Technology, 55(2), 680–684. https://doi.org/10.1016/j.lwt.2013.10.004

Taverniti, V., Scabiosi, C., Arioli, S., Mora, D., & Guglielmetti, S. (2014). Short-term daily intake of 6 billion live

probiotic cells can be insufficient in healthy adults to modulate the intestinal bifidobacteria and lactobacilli.

Journal of Functional Foods, 6(1), 482–491. https://doi.org/10.1016/j.jff.2013.11.014

Tomasi, G., Savorani, F., & Engelsen, S. B. (2011). Icoshift: An effective tool for the alignment of chromatographic

data. Journal of Chromatography A, 1218(43), 7832–7840. https://doi.org/10.1016/j.chroma.2011.08.086

Tremaroli, V., & Bäckhed, F. (2012). Functional interactions between the gut microbiota and host metabolism. Nature,

489(7415), 242–249. https://doi.org/10.1038/nature11552

Tripathi, M. K., & Giri, S. K. (2014). Probiotic functional foods: Survival of probiotics during processing and storage.

Journal of Functional Foods, 9(1), 225–241. https://doi.org/10.1016/j.jff.2014.04.030

Trygg, J. (2003). Homepage of Chemometrics. Science, 1–3.

Trygg, J., Holmes, E., & Lundstedt, T. (2007). Chemometrics in Metabonomics. Journal of Proteome Research, 6,

469–479.

Turroni, F., Ventura, M., Buttó, L. F., Duranti, S., O’Toole, P. W., Motherway, M. O. C., & Van Sinderen, D. (2014).

Molecular dialogue between the human gut microbiota and the host: A Lactobacillus and Bifidobacterium

perspective. Cellular and Molecular Life Sciences, 71(2), 183–203. https://doi.org/10.1007/s00018-013-1318-0

52

Vázquez-Fresno, R., Llorach, R., Marinic, J., Tulipani, S., Garcia-Aloy, M., Espinosa-Martos, I., … Andres-Lacueva,

C. (2014). Urinary metabolomic fingerprinting after consumption of a probiotic strain in women with mastitis.

Pharmacological Research, 87, 160–165. https://doi.org/10.1016/j.phrs.2014.05.010

Verbeke, K., Ferchaud-Roucher, V., Preston, T., Small, a C., Henckaerts, L., Krempf, M., … Priebe, M. G. (2010).

Influence of the type of indigestible carbohydrate on plasma and urine short-chain fatty acid profiles in healthy

human volunteers. European Journal of Clinical Nutrition, 64(7), 678–684. https://doi.org/10.1038/ejcn.2010.92

Verbeke, K. A., Boobis, A. R., Chiodini, A., Edwards, C. A., Franck, A., Kleerebezem, M., … Tuohy, K. M. (2015).

Towards microbial fermentation metabolites as markers for health benefits of prebiotics. Nutr Res Rev, 28(1), 42–

66. https://doi.org/10.1017/S0954422415000037

Walsh, M. C., Brennan, L., Malthouse, J. P. G., Roche, H. M., & Gibney, M. J. (2006). Effect of acute dietary

standardization on the urinary, plasma, and salivary metabolomic profiles of healthy humans. American Journal

of Clinical Nutrition, 84(3), 531–539. https://doi.org/84/3/531 [pii]

Wang, Y., Tang, H., Nicholson, J. K., Hylands, P. J., Sampson, J., & Holmes, E. (2005). A metabonomic strategy for

the detection of the metabolic effects of chamomile (Matricaria recutita L.) ingestion. Journal of Agricultural and

Food Chemistry, 53(2), 191–196. https://doi.org/10.1021/jf0403282

Williams, H. R., Cox, I. J., Walker, D. G., Cobbold, J. F., Taylor-Robinson, S. D., Marshall, S. E., & Orchard, T. R.

(2010). Differences in gut microbial metabolism are responsible for reduced hippurate synthesis in Crohn’s

disease. BMC Gastroenterology, 10, 108. https://doi.org/10.1186/1471-230X-10-108

Winning, H., Roldán-Marín, E., Dragsted, L. O., Viereck, N., Poulsen, M., Sánchez-Moreno, C., … Engelsen, S. B.

(2009). An exploratory NMR nutri-metabonomic investigation reveals dimethyl sulfone as a dietary biomarker

for onion intake. The Analyst, 134(NOVEMBER 2009), 2344–2351. https://doi.org/10.1039/b918259d

Wishart, D. S. (2008). Quantitative metabolomics using NMR. TrAC - Trends in Analytical Chemistry, 27(3), 228–237.

https://doi.org/10.1016/j.trac.2007.12.001

Wishart, D. S. (2014). Exploring the Human Metabolome by Nuclear Magnetic Resonance Spectroscopy and Mass

Spectrometry. In N. W. Lutz, J. V Sweedler, & R. A. Wevers (Eds.), Methodologies for Metabolomics,

Experimental Strategies and Techniques (pp. 3–29).

Wishart, D. S., Jewison, T., Guo, A. C., Wilson, M., Knox, C., Liu, Y., … Scalbert, A. (2013). HMDB 3.0-The Human

Metabolome Database in 2013. Nucleic Acids Research, 41(D1), 801–807. https://doi.org/10.1093/nar/gks1065

Worley, B., & Powers, R. (2013). Multivariate Analysis in Metabolomics. Current Metabolomics, 1, 92–107.

https://doi.org/10.2174/2213235X11301010092

Wu, J., & Gao, Y. (2015). Physiological conditions can be reflected in human urine proteome and metabolome. Expert

Review of Proteomics, 12(6), 623–636. https://doi.org/10.1586/14789450.2015.1094380

Wutzke, K. D., Lotz, M., & Zipprich, C. (2010). The effect of pre- and probiotics on the colonic ammonia metabolism

in humans as measured by lactose-[(1)(5)N(2)]ureide. Eur J Clin Nutr, 64(10), 1215–1221.

https://doi.org/10.1038/ejcn.2010.120

Xie, G., Zhang, S., Zheng, X., & Jia, W. (2013). Metabolomics approaches for characterizing metabolic interactions

between host and its commensal microbes. Electrophoresis, 34(19), 2787–2798.

https://doi.org/10.1002/elps.201300017

Xu, J., Yang, S., Cai, S., Dong, J., Li, X., & Chen, Z. (2010). Identification of biochemical changes in lactovegetarian

urine using 1H NMR spectroscopy and pattern recognition. Analytical and Bioanalytical Chemistry, 396(4),

1451–1463. https://doi.org/10.1007/s00216-009-3338-z

Zheng, H., Yde, C. C., Clausen, M. R., Kristensen, M., Lorenzen, J., Astrup, A., & Bertram, H. C. (2015).

53

Metabolomics investigation to shed light on cheese as a possible piece in the French paradox puzzle. Journal of

Agricultural and Food Chemistry, 63(10), 2830–2839. https://doi.org/10.1021/jf505878a

54

Appendix

A1: Table 1.A: Chemical shifts of protons in different molecules, adapted from Lancashire (2003).

A2 :Ingredients and brand:

Photo 1.A: Actimel Danone strawberry yoghurt drink, ingredients and brand, (Poulsen, 2017)

Composition of yoghurt: Translated ingredients list from Danish to English based upon photo 1.A:

Ingredients: Pasteurised milk, sugar, strawberries 2.1 %, glucose, milkminerals, modified tapioca

starch, natural aroma, acidity regulator (sodium citrate), yoghurtculture, L. Casei Danone-culture

(Lactobacillus paracasei subsp. Paracasei (DN-114001/CNCM I-1518)), vitamins (B6, D)

55

Photo 2.A: Recommended daily intake of Actimel, Danone strawberry yoghurt drink (Poulsen, 2017) Translated recommendations from Danish to English based upon the highlighted section of photo

2.A:

“1 bottle a day as part of a balanced diet and healthy lifestyle”

Below:

“Actimel contains vitamin B6 and D, which help maintaining normal function of the immune

system”

A3: Set-up of pilot study Table 2.A: Set-up of pilot study Date Time point in number of

days Sample, S

17/2 0 I 18/2 1 II 19/2 2 20/2 3 21/2 4 22/2 5 23/2 6 24/2 7 25/2 8 III 26/2 9 27/2 10 28/2 11 1/3 12 2/3 13 3/3 14 4/3 15 IV 5/3 16 6/3 17

56

7/3 18 8/3 19 9/3 20 10/3 21 11/3 22 V 22/4 64 VI

A4: Ingredients of Master Mix with corresponding concentrations Table 3.A: Composition of stock buffer used for chemical analysis of urine samples

Ingredients 11.7x stock (100mL) Concentration Final concentration Na2HPO4

10.377 g 583 mM 50 mM

DSS

0.065 g 2.9 mM 0.25 mM

Sodium formate

0.0079 g 1.162 mM 100 µM

A5: Sample IDs Table 4.A: ID of samples for NMR analysis. K, Katrine; U, Ulrik

Sign on NMR tube

ID Corresponding sample, S

Sample number

I U1 I 1 II U2 II 2 III U3 III 3 IV U4 IV 4 V U5 V 5 VI U1 I 1 VII U2 II 2 VIII U3 III 3 IX U4 IV 4 X U5 V 5 U1 U6 VI 6 U2 U6 VI 6 1 K1 I 1 2 K2 II 2 3 K3 III 3 4 K4 IV 4 5 K5 V 5 K1 K6 VI 6 7 K1 I 1 8 K2 II 2 9 K3 III 3 10 K4 IV 4 11 K5 V 5 K2 K6 VI 6

57

A6: Identification of metabolites Table 5.A: Frequently identified metabolites from the 24 spectra

Identified metabolite Chemical shifts (multiplicity) Assignments 3-hydroxyisovalerate 2.36(s), 1.3(s)

CH2, CH3

UN1 2.33(s) - Citrate 2.5(d), 2.7(d) CH2 Dimethylamine 2.72 (s) CH3 Succinate 2.4 (s) CH3 Alanine 1.48(d), 3.8 (q) CH3, CH 2-hydroxyisobutyrate 1.35(s) CH3 UN2 2.28-2.25 - Methylamine 2.60(s) CH3 Creatinine 3.04(s), 4.1(s) CH3, CH2 Malonate 3.11(s) CH2 Dimethylsulfone 3.13(s) CH3 UN3 3.19(s) - Betaine 3.25(s), 3.9(s) CH3, CH2 Trimethylamine N-oxide (TMAO) 3.26(s) CH3 1,3 dimethylurate 3.31(s), 3.44(s) CH3, CH3 UN4 3.35(s) - 4-hydroxyphenylacetate 3.44(s), 6.9 (d), 7.2(d) CH2, ArH, ArH 3-methylxanthine 3.5(s), 8.0 (s) CH3, CH Glycine 3.56(s) CH2 Glycolate 3.95(s) CH2

Hippurate 3.96(d), 7.54(-), 7.63(tt), 7.82(dd), 8.5(s) CH2, ArH, ArH, ArH, NH

Tartrate 4.34(s) CH

Trigonelline 4.43(s), 8.07(t), 8.82(d), 8.84 (d), 9.0 (s)

CH3, ArH, ArH, ArH, ArH

Urea 5.8(s) NH2 UN5 7.19/s) - UN6 7.20(s) - UN7 7.48(s) - Hypoxanthine 8.17(s), 8.20(s) CH, CH UN8 8.19(s) or (d) - Formate 8.45(s) H-COOH

1-methylnicotinamide 4.5(s), 8.2 (t) 8.88 (d), 8.95(d) 9.26(s) CH3, ArH, ArH, ArH,ArH

N-phenylacetylglycine 8.0 (s), 7.4(t), 7.35(t), 7.35(d), 3.7(d), 3.7(s) NH, ArH, ArH, ArH, CH2, CH2

UN9 8.10 (s) - UN10 5.00-5.02 - UN11 1.22(d) - UN12 7.97 (-) -

58

A7: Quantification of selected metabolites (U), normalized to creatinine

35.09

24.4626.73 24.75

30.78

40.67

051015202530354045

U1 U2 U3 U4 U5 U6

Trigonelline,µM/mMcreatinine

Trigonelline

211.20253.13

352.94

264.98 244.50

321.43

050100150200250300350400

U1 U2 U3 U4 U5 U6

Hippurate,µM/mMcreatinine

Hippurate

72.29114.36

173.96

295.85

177.21

411.09

050100150200250300350400450

U1 U2 U3 U4 U5 U6

Citrate,µM/mMcreatinine

Citrate

59

Figures 1.A: Quantified concentration of selected metabolites normalized to creatinine (subject U)

A8: Quantification of selected metabolites (subject K), normalized to creatinine

10263.25

20616.45

10880.3112885.90

10977.559524.80

0

5000

10000

15000

20000

25000

U1 U2 U3 U4 U5 U6

Creatinineaverage(µM)

Creatinineaverage(microM)

2.91

65.53

27.35 27.01

0.68

23.96

0

10

20

30

40

50

60

70

U1 U2 U3 U4 U5 U6

TMAO,µM/mMcreatinine

TMAO

146.62 139.48

95.99

127.82 119.89

48.12

0

50

100

150

200

K1 K2 K3 K4 K5 K6

Trigonelline,µM/mMcreatinine

Trigonelline

60

Figures 2.A Quantified concentration of selected metabolites normalized to creatinine (subject K)

814.52

1904.28

451.83 357.81

1589.25

435.00

0

500

1000

1500

2000

K1 K2 K3 K4 K5 K6

Hippurate,µM/mMcreatinine

Hippurate

0.00 0.00

20.10 21.82 24.44

52.40

0

10

20

30

40

50

60

K1 K2 K3 K4 K5 K6

Citrate,µM/mMcreatinine

Citrate

2435.652132.65

4065.054700.50

2988.753116.25

0

1000

2000

3000

4000

5000

K1 K2 K3 K4 K5 K6

Creatinineaverage(µM)

Creatinineaverage(microM)

61

A9: Average concentration of selected metabolites Table 6.A: Average concentrations of selected metabolites (µM/mM creatinine) from each observation and corresponding

standard deviation (Std)

Sample

Trigonelline (µM/mM creatinine)

Hippurate (µM/mM creatinine)

Citrate (µM/mM creatinine)

Creatinine(mM)

TMAO (µM/mM creatinine)

U1 35.09 211.20 72.29 10263.25 2.91 U2 24.46 253.13 114.36 20616.45 65.53 U3 26.73 352.94 173.96 10880.31 27.35 U4 24.75 264.98 295.85 12885.90 27.01 U5 30.78 244.50 177.21 10977.55 0.68 U6 40.67 321.43 411.09 9524.80 23.96 Average 28.76 259.06 175.59 10928.93 25.48 Std(average) 6.45 52.55 125.10 4118.69 23.38 K1 146.62 814.52 0.00 2435.65 K2 139.48 1904.28 0.00 2132.65 K3 95.99 451.83 20.10 4065.05 K4 127.82 357.81 21.82 4700.50 K5 119.89 1589.25 24.44 2988.75 K6 48.12 435.00 52.40 3116.25 Average 123.85 633.17 20.96 3052.50 Std(average) 36.32 663.04 19.35 976.26

A10: Heteronuclear 1H 13C 2D spectra of trimethylamine N-oxide and betaine

Figure 3.A: Heteronuclear [1H ,13C] 2D spectrum of trimethylamine N-oxide, adapted from HMDB.com

62

Figure 4.A: Heteronuclear [1H ,13C] 2D spectrum of betaine, adapted from HMDB.com