Journal ofi lai g

proteomeresearch

A^c s •• 0015^000

pubs.acs.org/jpr

1Fecal Metabolomics of Healthy Breast-Fed versus Formula-Fed2Infants before and during in Vitro Batch Culture Fermentation3JoMay Chow/ Matthew R. Panasevich/ Danny Alexander/ Brittany M. Vester Boler/4Mariana C. Rossoni Serao/ Trevor A. Faber/ Laura L. Bauer/ and George C. Fahey*'̂

s ^Abbott Nutrition, Columbus, Ohio 43219, United States6^Department of Animal Sciences, University ofIllinois, Urbana, Illinois 61801, United States7^Metabolon, Inc., Durham, North Carolina 27713, United States

ABSTRACT: Nontargeted metabolomics analyses were used (l) to comparefecal metabolite profiles of healthy breast-fed (BF) and formula-fed (FF)infants before and during in vitro fermentation in batch culture and (2) toevaluate fecal metabolomics in infant diet. Samples from healthy BF (n = 4) orFF (n = 4) infants were individually incubated at 37°C in anaerobic mediacontaining 1% (wt/vol) galactooligosaccharides, 6'-$ialyllactose, 2'-fucosyllactose,lacto-N-neotetraose, inulin, and gum arable for up to 6 h,and supernatants wereanalyzed using GC/MS and LC/MS/MS to assess changes in variouscompoimds. Comparison ofover 250 metabolites prior to incubation showed Hthat BF samples contained higher relative concentrations (P < 0.05) of H14 compounds including but not limited to human milk oligosaccharides and :[ Iother metabolites presumably transferred through breast feeding (linoelaidate,myo-inositol) (P < 0.05). Conversely, feces from FF infants contained 41identified metabolites at higher levels (P < 0.05). Our data are consistent with the notion that carbon-limited cultures catabolizeprotein and amino acids to obtain energy, whereas the provision offermentable carbohydrate creates anabolic conditions relying onanuno adds for bacterial growth. Results also suggest that fecal metabolomics can be a useful tool for studying interactions amongdiet, microbes, and host.

KEYWORDS: infant, feces, metabolomics, metabonomics, oligosaccharides, fermentation

26 • INTRODUCTION

27 The human gut microbiota play a crudal role in shaping infant28 health and development. Abnormal patterns of gut microbiota29 have been clinically linked to late-onset sepsis' and necrotizing30 enterocolitis,^ two important causes ofmorbidity and mortality31 in preterm infants. A growing body of evidence also implicates32 the intestinal microbiota in the development of colic,^ allergic33 disease,"* and obesity later in life^ as well as the formation of34 a key metabolite involved in melamine-induced kidney stone35 formation. Indeed, thegutmicrobiota modulate brain develop-36 ment and subsequent adult behavior in animal models.^37 The composition of the infant gutmicrobiome isprofoundly38 affected by diet. In general, studies using molecular techniques39 indicate that a majority of breast- and formula-fed infants harbor40 significant numbers of bifidobacteria.® Feces from exclusively41 breast-fed infants, however, are dominated by bifidobacteria,42 whereas those from formula-fed infants also contain other

43 bacterial species, including E. coli, Clostridium difficile, bacter-44 oides, and lactobadlli.®'̂45 Despite the abundance of studies regarding the effect of diet46 on infant fecal microbiota andhealth,relatively little attention47 has been paid to the influence of diet on the infant fecal meta-48 bolome.""'̂ The metabolomics approach, which can sometimes49 involve the analysis and identification of thousands of small

ACS PubliCStiOnS exxxx American Chemical society

metabolites, offers the potential to gain insight into the complex sointeractions among gut microbes, diet, and host. In turn, these sidata may be used to gain a clearer understanding of how diet 52affects both the gut microbiota and the host and to identify S3potential areas for future research. Overall, the main objectivesof S4the current studywere to use nontargeted metabolomicsanalysis ssof fecal samples to (l) compare metabolite profiles of healthy 56breast-fed to healthy formula-fed infants before and during S7in vitro fermentation in batch culture and (2) to evaluate the 58usefulness of metabolomics in assessing the infant diet. S9

• MATERIALS AND METHODS 60

Donors 6i

The infant fecal samples usedwerecollected for use in aprevious 62study.'"* Briefly, eight infants were enrolled for participation 63in this study between March and June of 2010 from the 64Champaign-Urbana area, and infants consumed their normal diet 6Sofexclusive breast milk(« =4) or exclusive infantformula (« =4) 66for at least 2 months immediately prior to fecal collection. 67Formula-fed infants were fed one of three lactose-based 68

formulas: Similac Advance (Abbott Laboratories Columbus, 69

Received: January 3, 2014

dx.doi.0rg/lO.lO2i/prSOOOllwlJ. Proteome Rei. XXXX XXX, XXX-XXX

Journal of Proteome Research

70 OH; « = l), Member's Mark Infant formula (Sam's Club^71 Bentonville, AR; « = 1), or Enfamil Premium (Mead Johnson,72 Glenview, IL; «=2). These formulas contained as much as 4g/L73 galactooligosaccharides (COS), but they did not contain74 maltodextrins, probiotics, or added oligosaccharides otherthan75 COS. Other inclusion/exclusion criteria included the following:76 the infant was full term at birth with a gestational age of38 to77 42wk; the infant was atorabove the fifth percentile for weight at78 birth; the infant had no maternal medi^ history ofdiabetes,79 tuberculosis, or perinatal infection with proven adverse effects on80 the fetus; were vaginal births; were at least 2 mo ofage at study81 entry but not older than 4 mo of age; had no known cardiac,82 respiratory, gastrointestinal, or other systemic disease such as83 urinarytract infection orotitis media; had nohistory ofblood group84 incompatibilityserious enough toresult inhematological problems;85 and were notreceiving any medications (except for supplemental86 vitamins) and have never received antibiotics. The e^cperimental87 protocol was approved by the University of Illinois Institutional88 Review Board, and all legally acceptable representatives signed an89 informed consent priorto initiation ofthe e:q>eriment

90 Substrates

91 Substrates used included galactooligosaccharides (GOS) 9592 (GOS; Inalco Pharmaceuticals, Italy), a-(2—6')-W-acetylneur-93 aminyl-lactose sodium salt (6'-si^yllactose) (6'SL; Inalco94 Pharmaceuticals, Italy); 2'- cr-L-fucopyranosyl-D-lactose (2'-95 fucos)dlactose) (2'FL; Inalco Pharmaceuticals, Italy); lacto-N-96 neotetraose (LNnT; Boehringer Mannheim, Germany); Orafti97 HP inulin (HP; BENEO-Orafti, Belgium); and gum arable (GA;98 Fisher Scientific, Pittsburgh, PA).

99 Fecal Collection

100 On the day of the in vitro eiqieriments, fecal samples were101 collected in diapers within 15 min of defecation. Diapers were102 double-sealed inplastic bags andtransported to thelaboratory in103 a cooler containing tepidwater. Fecal samples were diluted 1:10104 (wt/vol) in anaerobic diluent, homogenized for 15 s in aWaring105 blender, filtered throughfourlayers ofcheesecloth, andsealed in106 serum bottles under CO2. Serum bottles containing the inocula107 were stored at 37 °C until use.

108 In Vitro Fermentation and Sample Collection

109 Each substrate (80 mg) was weighed intriplicate for each infant at110 each sampling time into16mLBalch tubes. Analiquot (7.2mL)111 of media (Table l) was aseptically transferred into the Balch112 tubes, capped with but}d rubber stoppers, and sealed with113 aluminum caps. Tubes containing HP and GA were stored with114 media at 4 ®C for approximately 12h to enable hydration ofthe115 substrates before initiating fermentation. These tubes were placed116 in a 37 ®C waterbath approximately 30 min before inoculation.117 Due to the limited supply of substrates and unpredictability118 associated with obtaining fecal samples from in^ts, tubes119 containing GOS, 6'SL, 2'FL, and LNnT were hydrated upon120 obtaining fecal samples. Fecal samples fromthe individual infants121 were also incubated in media witiiout added carbohydrate, and122 these cultures are referred to as "blank" tubes. Supematant123 samples were collected at 0, 3, and 6 h offermentation and were124 immediately frozen at —80 °C imtil further analysis.

125 Metabolomic Profiling

126 Frozen samples were shipped under dry ice to a commercial127 laboratory (Metabolon, Durham, NC) for metabolite analysis.128 Procedures for metabolic profiling have been described129 previously for the three platforms used in combination for the130 analysis, including GC/MS'̂ and two LC/MS systems, '̂' one

Table 1.CompositionofMicrobiological MediumUsedin thein Vitro Experiment

component concentration in medium

mL/L

solution A" 330.0

solution b'' 330.0

trace mineral solution*^ 10.0

water-soluble vitamin solution'' 20.0

folate/biotin solution' 5.0

riboflavin solution/ 5.0

hemin solution^ 2.5

resazurin'' 1.0

distilled H2O 296.1

g/LNaiCOj 4.0

cysteine HCI-H2O 0.5

trypticase 0.5

yeast extract 0.5

"Composition (g/L): NaCl, 5.4; KH2PO4, 2.7; CaCb-HzO, 0.16;MgCb-dHzO, 0.12; MnCl2-4H20, 0.06; C0CI2-6H2O, 0.06;(NH4)2S04, 5.4. ''Composition (g/L): K2HPO4, 2.7. '̂ Composition(mg/L): ethylenediaminetetraacetic add (disodium salt),500; FeS04-7HiO, 200; ZnS04-7H20, 10; MnCl2-4H20, 3; H3PO4, 30; C0CI2-6H2O, 20; CUCI2-2H2O, I; NiCl2-6H20, 2; Na2Mo04-2H20, 3.''composition (mg/L): tiiiamin-HCI, 100; d-pantothenic add, 100;niacin, 100;pyridoxine, 100;p-aminobenzoic add, 5; vitamin3,2,0.25.'̂ Composition (mg/L): foiic add, 10; d-biotin, 2; NH4HCO3, 100.-^Composition: riboflavin, 10 mg/mL in 5 mmoI/L of Hepes.^Composition: hemin, 500 mg/mL in 10 mmoI/L of NaOH.''Composition: resazurin, 1g/L in distilled H2O.

optimized for positive ionization and one optimizedfor negative 131ionization. Peak extraction, data curation, and QC procedures 132were performed using proprietary sofhvare.'̂ '̂ ® At the time 133of analysis, samples were thawed and prepared according to 134the described standard methanol extraction protocol, which is 13sdesigned to remove proteins, dislodge small molecules boundto i36protein or physically trapped in the precipitatedprotein matrb^ 137recovera wide range of chemically diversemetabolites,and split i38into aliquotsfor analysis on the three platforms. 139

Data Normalization i4o

Data were collected over multiple platform run days and were 141adjustedbyscaling to the medianvaluesfor eachgroup-balanced 142run-day block for each individual compound. This minimizes 143any interdayinstrumentgain or driftbut does not interferewith 144intraday sample variability. Datawere not otherwise adjusted or 145normalized. 146

Statistical Analysis 147

Repeatedmeasures ANOVAwasperformed to leverage the data i48fromthe multiple timepoints withinthe study.There were three 149tests embedded within the repeated measures ANOVA: (l) 150DIET Main, (2) TIME Main, and (3) Interaction (DIET X isiMAIN). Essentially, the "DIET" Main effect tested whether 152the means of the three groups were differentfrom the BLANK 153when averaged across all time points. The "TIME" Main effect 154examines whether the means at each time point were different isswhenaveraged across the groups. Finally, the "Interaction" asks is6whether the time profiles are nonparallel between the groups 157(nonparallel profiles signify a difference during the time-course 158between thegroups). The calculations also tookintoaccount the 159repeated measure inherent in sampling at each time point the 160cultures from individual infants. 161

d)tdot.o:g/10.1021/prS<B011w 12 Proteome Res.XXXX, XXX XXX-XXX

Journal of Proteome Research

Higher In BF Blank at Time = 0 h

(BF/FF)°

Iacto4il-fucopentaose2-fueesytlactoMliRMlatdato (tr 18:2rS)1,2-pRiiwiiawlfucmemyOijnosJtollaurochotonata sultetol-palmKoylglyvaroptiosphochoUnoBuantna

HMv «*{44tytfro)Qfph6fly()(ftctfttom^m tthanalSRilna

UlidiMlaelaM

B phmphatsisovtlafita

S-amlRovil«istoganuna-tocophcfol

piMfiytecttit*b*tMttOSttral

VllSnto(Ipha-toCOphBtl

pdntltoyltthanotemkleinnttoll-phMplttto(IIP)

4-lqrdroxyphiiflylac«tit*dooxycftfrotlM

taunts (12:0)3-4alwilractnnins

>sts{n3DPA;ZdocoMpantsnanosts (nS DPA; 22:5n3)Riyrtatsts (14:0)

urobOlnostnttnotansto (alpha or s<nima; (10:303 or 6)] am

pyifcloxats ampalmlM(16dn

pngnan^Iol dtawfktscleats (ia:1n9)

/•kstoIKhociwtatslUfimaStSfOl ma

capcsla(10:0)anotaats(18:2n6)

13-intttqrfanytittlcsciacapiytets (8:0)

2-(sap<opytnistata3-phsnytpfoplonats (tiydwcmnamats)

trtettbaOytetsgamffla-muitehotate

Nsoocsnts(24KIIpstaigonats (9:0) —

niotlnatsilbonudsoaMe •••Naeatylglulaaisla

2-hydroxyptlmltats •asstiyptamlns ^

caproats (8:0)Zamlnobutynts ^

pantothanate ^pantadacanoata (15a) ••

Higher in FF Blank at Time = 0 h

(FF/BF)®

40X 30X 20X 10X 1 10X 20X 30X 40X

Magnitude of Higher Concentration

Figure 1. Ratios, as indicated, ofbreast-fed (BP) and formula-fed (FF) group means for compounds that are significantly difierent between fecalsupematants at time= 0 in the absence ofadded carbohydrate substrates (blank), (a) BF/FF= breast-fed to formula-fed andFF/BF= formula-fed tobreast-fed. (b) Indicates compound that has not been officially "plexed" (based on astandard), but we are confident in its identity.

Table 2.Metabolic Pathway Classification and Fold-Change Values forCompounds Significantly Higher inBlank BF Samples atTime = 0 h

super-pathway sub-pathway biochemical name BF/FF"

amino add phenyialanine and tiyrosinemetabolism 3-(4-hydroxyphenyl)lactate 2.56

carbohydrate fructose, mannose, galactose, starch, andsucrose metabolism fiicose 25.00

2-fucosyllactose >30.00

lacto-N-fricopentaose >30.00

glycoiysis, gluconeogenesis, pyruvate metabolism lactate 1.96

ketone bodies 1,2-propanediol 25.00

energy oxidative phosphorylation phosphate 1.25

lipid long-chain frtty add linoelaidate (tr 18:2k6) >30.00

bile add metabolism taurocholenate sulhte'' 9.09

glycerolipid metabolism ethanolamine 2.33

inositol metabolism myo-insoitol 14.29

lysolipid 1-palmitoylglycerophosphocholine 6.25

nucleotide purine metabolism, guanine containing guanine 3.45

pyrimidine metabolism, uracil containing uiidine 2.13

BF/FF —breast-fed to formula-fed. ''Indicates compound that has not been officially "plexed" (based on astandard), but we are confident in itsidentity.

162 Missing values (if any) wereassumed to be below the level of163 detection for a particular biochemical with the instrumentation164 used and were imputed with the observed minimum for that165 particular biochemical. Welch's Two-Sample t tests were used166 to analyze fold differences between FF and BF groups. Forall167 analyses, missing values were imputed with the observed168 minimum for that particular compound (imputed values were169 added after block normalization). The statistical analyses were170 performed on natural log-transformed data to reduce the effect171 ofany potential outliers in the data. Welch's Two-Sample t test

comparisons were made between the means of eacdi biochemical musing statistical software: Array Studio (Omicsoft, Inc.) or "R" 173fi*om the Free Software Foundation, Inc. Significance was 174determined by P < 0.05. 17S

• RESULTS 176

Comparison of Initial Conditions 177

Figure 1 displays fold differences of identified metabolites that I78were presentathigher levels (P < 0.05) in eitherthe BForinthe 179

dx.doi.OFg/10.1021/pr50001 Iwl / Pmteome Ites. XXXX, XXX, XXX-XXX

Journal of Proteome Research

Table 3. MetaboUc Pathway Classification and Fold-Change Values for Compounds Significantly Higher in Blank FF Samples atTime = 0 h

super-pathway

amino add

carbohydrate

energy

tipid

cofactors and vitamins

identity.

sub-pathway biochemical name FF/BF

glutamatemetabolism N-acetylglutamate 2.08

phenylalanine and tyrosinemetabolism phenylacetate 10.39

4-hydroxyphenylacetate 6.91

3-phen)dpropionate (hydrocinnamate) 2.52

tryptophan metabolism tryptamine 1.85

ureacycle, arginine and prolinemetabolism S-aminovalerate 27.89

butanoate metabolism 2-aminobutyrate 1.78

glycolysis, gluconeogenesis, pyruvate metabolism Z-isopropylmalate 2.53

Krebs cycle tricarballylate 2.48

essentialhttty add linoleate (I8:2n6) 2.87

linolenate [alpha or gamma; (I8:3n3 or 6)] 4.23

docosapentaenoate (n3 DPA; 22:5n3) 5.50

short-chain fatty add valerate 8.52

isovalerate 46.99

medium-chainfattyadd caproate (6:0) 1.79

caprylate (8:0) 2.68

pelargonate (9:0) 2.19

caprate (10:0) 2.87

laurate (12:0) 6.49

long-chainfatty add myristate (14:0) 4.94

pentadecanoate (1S:0) 1.18

palmitate (16:0) 3.75

oleate(I8;ln9) 3.34

lignocerate (24:0) 2.20

fettyadd, monohydroxy 2-hydroxypalmitate 1.89

&tty add, branched 13-methylmyristic add 2.76

endocannabinoid palmitoylethanolamide 8.00

camitine metabolism deoxycamitine 6.82

3-dehydrocamitine'' 5.51

bile add metabolism 7-ketolithocholate 3.27

gamma-muricholate 2.26

inositol metabolism inositol 1-phosphate (UP) 7.34

sterol/steroid beta-sitosterol 9.02

stigmasterol 2.97

pregnen-doil disuliate'' 3.35

hemoglobin and poiphyrin metabolism urobilinogen 4.63

nicotinate and nicotinamide metabolism nicotinate ribonudeoside'' 2.16

pantothenate and CoA metabolism pantothenate 1.61

tocopherol metabolism alpha-tocopherol 8.05

gamma-tocopherol 18.36

vitamin B6 metabolism pyiidozxate 4.19

180 FF groups. Thesedifierences were calculated by comparing the181 levels of the various metabolites in the blank tubes at time = 0.

182 Since thecontents oftheblanktubes were identical except forthe183 fecal samples from the individual infants, anydifferences in the184 levels ofthevarious metabolites between thegroups presumably185 represent differences in the relative levels of fecal metabolites186 between the twodietgroups. Samples from BFinfants contained187 significantly greater concentrations of 14compounds compared188 to FF infants. As expected, supematants from BF infants had189 greater levels of the human milk oligosaccharides (HMO),190 2'-fucosyllactose and lacto-N-fiicopentaose, and the HMO191 precursor, fiicose, than those from FF infants. These two

192 oligosaccharides were not detected in FFsamples, but based on193 detection limits, these values must have been at least 2S-fold194 higher instoolsamples fromBFinfants (Table 2). Furthermore,

time = 0 samples from BF infrmts were higher in a singleamino 19sadd super-pathway-related compound, [3-(4-hydroxyphenyI) 196lactate], a metabolite of tyrosine. Samples from the BFinfwts 197also were higher in five lipid super-pathway-related compounds 198(Table 2) andwere particularly elevated in the C18:2, n-6 trans 199fatty add, linoelaidate (>30-fold) and in the cyclic alcohol 200structural membrane component, myo-inositol (14.29-fold). 201

In contrast, supematants from FF infrnts contained a total of 20241 identified metabolites thatwerepresentat significantly higher 203concentrations compared to those from BF infants (Figure l). 204More specifically, samples from FF infants contained seven 20scompounds related to the amino add super-pathway, a total of 206two compoimds related to carbohydrate and/or energy super- 207pathways, 26 compounds related to the lipid super-pathway, 208and seven compounds related to cofactor and/or vitamin 209

dx.doi.0rg/IO.IO2i/prSOOOIIwlJ. ProKomeRes. XXXX, XXXXXX-XXX

Journal of Proteome Research

210 super-pathways thatwere present at greater concentrations than211 those from BF infants (Table 3).The amino acid super-pathway-212 related compounds that were particularly elevated in FF infants213 included the amino acid catabolites: phenyllactate (l0.39-fold),214 4-hydroxyphenylacetate (6.91-fold), and 5-aminovalerate215 (27.89-fold). Likewise, the lipid super-pathway compounds216 that were noticeably greater in FF infant samples included the217 following: fatty acids, valerate (8.52-fold) and isovalerate (46.99-218 fold); the endocannabinoid, palmitoyl ethanolamide (8.00-fold);219 thesterol beta-sitosterol (9.02-fold); and both a- (8.05-fold) and220 y-tocopherol (l8.36-fold).221 The relative levels of the essential fatty add, linoleate (18:2,222 n6), and itsomega conversion product, linoelaidate (trans, trans-223 9,12-octadecadienoic add), in samples from BF and FF infants,224 are presented in Figure 2.Linoleate was markedly higher in FF

Relative

Concentration i-S

••1

rillinoleate linoelaidate

Fatty Acid

Figure 2. Relative levels oflinoleate anditsomega conversion product,linoelaidate, in BFversus FF infant samples.

225 samples, butwas also relatively abundant in the BF supernatants.226 However, the figure also indicates markedly higher levels of trans,227 trans form in the BF infant compared to those from the FF228 infants, which had almost none.

229 Infant Variability

230 Somecompounds exhibited highindividual infant variation, often231 being orJy detected from one infant, orbeing many fold higher in232 one individual compared to the population. For instance, one of233 theBF infwtshad considerably higher levels oftheomega-3 fatty234 acid, docosahexanoate (DHA), compared to theothers, aswell as235 uniquely detectable levels of creatine, acetylcamitine, and urate236 (datanot shown). Widevariation wasalso notedin theamountof237 2'FLdetected in feces from BF andFF infrnts (data not shown).238 Other examples of individual infent variation, including differ-239 ences in several amino acidcatabolites and longchain fatty adds,240 wereobserved in both the BF and FF groups.

Addition of Exogenous Carbohydrates 241

A heat map displaying tlie compounds whose fermentation 242patterns were altered by the presence of exogenously added 243carbohydrate in one or both of the diet groups (BF or FF) is 244shown in Figure 3. Fermentation patterns for GOS, 2'FL, and 245LNnT were highly similar to each other, with corresponding 246increases (red) and decreases (green) in various amino acids, 247carbohydrate, and lipid-related metabolites over the 6 h 248fermentation. By comparison, the profiles for 6'SL and HP 249showed similarity to one another but differed substantially from 250those of GOS, 2'FL, and LNnT. Lastly, the profile for GAwas 251essentially identical to the blank (no carbohydrate added 252control) and did not resemble those of the other carbohydrate 253treatments. In addition, Figure 3 indicates that some metabolites 254that responded differentially to oligo addition between the BF 25sand FF groups (flagged as blue cells in the far right column), 256which included various metabolites related to amino acid 257

metabolism, the urea cycle, pyruvate metabolism, and specific 258SCFA (e.g., valerate and isovalerate). 259

As an example of a differential response to oligos, Figure 4 260displays in greater detail changes in the relative concentrations of 261the fattyacid, valerate, in FFversus BFsamples duringthe course 262of fermentation. Valerate, a product of proline degradation, 263accumulated rapidly during the 6 h fermentation in the blank, 2646'SL, HP, and GAtreatments, but the increasewasonlyobserved 265in samples collected from FF infants. Relative levels of valerate 266

changed very little for BF infants, regardless of carbohydrate 267substrate. Similar resultswere observedfor isovalerate, a product 268ofvaline fermentation (data not shown). 269

The relative concentrations ofselected biogenic amines during 270the course of fermentation arepresentedin Figure 5.With a few 271exceptions, the fold differences between FF and BF (FF/BF) 272infonts changed very little upon addition of carbohydrate. 273However, the addition of either HP or GA resulted in signifi- 274cantly greater concentrations of putrescine after 6 h of fermenta- 275tion from FF compared to BF infant inoculum (P < 0.05). 276Likewise, supplementation with LNnT led to higher relative 277concentrations of agmatine in FF compared to BF infants at 6 h 278(P < 0.05). 279

• DISCUSSION 280

Fecal Metabolites: Breast-Fed versus Formula-Fed 281

A major objective of this study was to compare fecal metabolite 282profiles of FF and BF infants. To accomplish this objective, 283we identified compounds present in differing amounts in the 284time = 0 h blank FF and BF samples, because these compounds 285would presumably represent those residing in the fecal inocula. 286Inoculum from FF infants contained elevated levels of many 287amino acid catabolites such as phenyllactate from the degrada- 288tion of phenylalanine, 4-hydroxyphenylacetate from the break- 289down of tyrosine, and 5-amino valerate from the metabolism 290of lysine. Samples from this group also contained higher levels 291of isovalerate and valerate, which are indicative of protein 292fermentation. 293

The predominance of protein fermentation, as previously 294noted by Heavy etal.,'̂ may be due toanumber ofcontributing 295factors. To compensate for differences in the amino acid 296composition between bovine and human milk, infant formulas 297have been formulated to contain greater concentrations of 298protein thanhuman milk,^° soagreater quantity ofprotein could 299potentially escape digestion and absorption in the small intestine. 300Moreover, samples from FF infants presumably contain a more 301

d*doi.org/]0.1021/prS00011wl2PfOfeofne fles. XXXX XXX, XXX-XXX

Journal of Proteome Research

PATHWAY SUB-PATHWAYAntno acid Oycine. serlne and dveonine metabolam

Atanineand aspertale metacoism

GUarrate ir^OdoHm

Hsidine tnettboiam

Lysine melabodam

FYcnylalanine and tyrosine matatuistn

Tryptophan rreiaboiam

Valne. laucina, and Isdeuctne metatnism

CyslBina. mathionina, SAM. laurinematabolsm

Ureacycle: arglNna-. proine-, meiabolam

Pdyantne nwabelwn

PapMa Dipepdda

CarboHydrate Airinosugar matabdsni

Nuclaotde

Cctactonand

vtarrin*

Fnjctose, rrannose. galactose. stareh, endsucrose malabofsm

Qycolysis. gluconeoganarts. pyruveteme(abcisrn

Kdecttde sugars, panloea iretaboism

STiorl-chainfady acid

Madunvcrain fatly acid

Lcng-chain fady acid

Fatty acid, dicarborylaleFatty acidmetabcdsmGtycerolpid meabolsmKeienebodte

_LyscipidPjine maetcAsnv adefttna cortaining

Ucodnale and rScodnarrMe malabotsRi

PanWhenale and CoA iretaboism

IXanobiotics IChenical

[Sugar, sugarsuteWma. siareh

BIOCHEMCAL NAME

beta-hydfoaypynntfa

O-acetytKmosarine

deB-alanne

cada>etine

2-anYnopenanoale

phenylaiaale (PIA)

4-<iydroawnenrt)ytuvB<e

. N«afrlYro^

indolelaelale

andranibleN-sce(ylryptophan

ttdeudne

ieuclnaN-acefrKaine

_ ataha-hydfcayiaoiialei" cysMlne

N-acatybysteine

eyalloB_ 2-hydroxybutyfate (Al

emidiineS-amnoselereie

_ MacatytaniidiinepUraacine

N-aeetylpiitreacino

_ aomanna

_ B^cybroineglueoMfrfnaNaeetyipiueoaanYne

N^eaNlBAicoaanina I

afYtYongaa'NacatylnauaTii'iate

lactose

J'-fucoeylactDse

fruciDse-6-pho8pnade

1,3-dihvdroxyacetona

pyruale

laewe_ acatolacBle

sraWolttireilol

_ aedcnectdo86-7-phospha>e

- ''•fc"'*'caproele (6:0)

caprylalet&O)

patsroenaie lao)

eacraie 110:0)

lBurBlB(t2d))

nvftsieta M4:0)

patrilaledSd))pa>TtiBlaBte(16:1n7)iTBrgar8le(17:0)

cts-veccanaie(18:1n7)

_ lnoetadal8(18:2n6)_Z-hydragrgUtgate_ isowlerate_elhanotairtne

1,2-prcpane<tol

1-pa»TilDylgVeeroptesphoe»«anolanine

. BdenosineS'-npnoplioaphalB(AMP)

nicgtinanlde adenifte dinuctea*de (NAD'nicottnateaeatylCoA

Z.S-dinyAoxyisotaierale

BLANK I GOS I S'SL I ZfL ' HP OA OtfT

Figure 3.Compounds altered by the addition ofexogenously added carbohydrate (relative tothe blank) regardless ofdiet groups. Red cells indicate anincrease inthecompound; green cells indicate adecrease inthecompound; light red cells indicate aninitial increase at3hwith asubsequent decrease at6 h; blue cells indicate that the compound changed differentially depending on treatment group, (a) Compounds flagged as blue cells respondeddifferentially between the breast-fed and formula-fed groups, (b)Indicates compound that has not been officially "plexed" (based onastandard), but weare confident in its identity.

302 diverse fecal microbiome than those from BF infants with a

503 greaterproportion of bacteroides and dostridia than BFinfents.®Because bacteroides and dostridia possess proteolytic and/or 304

amino-add-fermenting capabilities, samples from FF infants 305

dx.doi.0rg/lO.lO2i/pfSOOOI Iwl/. Proteome Res.XXXX, XXX, XXX-XXX

Journal of Proteome Research

o -ti> 2

•i §o

O

-1.MEDIA : 2BLANK saGos 4.6SL

5-

0

5

0

5

0 . 2—7.1NUUN ^ acuM s

5

0 : ^ . /•4=36=6 --A A . ' A.' -A : ; A. X

•1 0123456 -1 0123456 -1 0123456 -1 0123456

Fermentation Time (h)

Figure 4. Relative concentrations of valerate over 6 h of fermentation.Each triangle represents theaverage offour infants in each diet group.Blue triangles = BF, orange triangles = FF.

306 would be expected to contain higher levels of protein307 fermentation products.308 Very little isknown about thefermentation ofdietary proteins309 in the infant gut and its potential effects on human health.310 However, recently published studies using a formula-fed piglet311 model suggest thatexcessive protein intake during theneonatal312 period could potentially lead to a variety of physiological con-313 sequences such as compromised intestinal barrier function,314 disturbed regulation of intestinal permeability by acetylcholine315 and vasoactive intestinal peptide, ^ early implantation of ileal316 microbiota,^^ and altered colonic immune cell development.^^317 Indeed, the piglet data^ '̂̂ ^ suggest that neonatal feeding of318 high protein diets might also result in metabolic consequences319 laterin life by altering sensitivity of the colonic mucosa to pro-320 inflammatory insults at maturity. Nonetheless, given the limited321 number of animal studies, the scarcity of mechanistic data, and322 thequalitative natureofourmetabolomics data, it ispremature to323 extrapolate the results to human infants.

324 Carbohydrate Supplementation

325 The second objective of the current study was to determine326 the effect of carbohydrate supplementation on fermentation327 profiles of the fecal inocula. As expected, the 0 h blankcultures328 (no carbohydrate added), as well as those treated with GA,329 mainly produced various fatty adds suchascaproate, isovalerate,330 andvalerate as a result of protein fermentation. Bycontrast, the

addition of fermentable carbohydrates, such as COS, 2'FL, and 331LNnT, to the cultures reduced the levelsof variousamino acids, 332decreased the accumulation of amino add-related metabolites, 333and increased the levels of metabolites related to energy 334generation. These observations are consistent vdth the idea 335that cultures lacking a source of fermentable carbohydrate will 336ferment amino acids to obtain energy, whereas the provision 337ofcarbon would obviate the need for metabolism ofamino adds 338

for energy and create anabolic conditions that would require 339utilization of amino acids for growth. 340

A more detailed examination of these data revealed that 341

cultures from FF infants responded differently to carbon 342limitation than those of BF infants. Although both BF and FF 343samples generated amino add super-pathway-related compounds 344in response to carbohydratesupplementation,onlythose firom FF 345in&nts also fermented protein under carbon-limited conditions 346(e.g., 0 h blank or GA). Continuous degradation of amino 347nitrogen by samples from FF infants during carbon-limited 348conditions was likely due to the presence of bacteria that are 349able to exdushrely utilize amino adds and/or peptides such as 350dostridia, Shieella, enterococd, bacteroides, Escherichia coli, or 351staphylococci. 352

Interestingly, fecal microbiota from allfour FF infantsquickly 353degraded2'FL in vitrowithinthe first3 h ofincubation, but those 354fromBFinfantsvariedconsiderably in their ability to fermentthis 355oligosacdiaride (data not shown). The rapid disappearance of 3562'FL in culturesfromFF infantsisconsistentwithpreviouswork 357showing that the introductionof foodsother than breast milkto 358in^ts improvesthe ability of fecal bacteriato ferment complex 359carbohydrate^ induding HMO.^ More specifically, the rapid 360degradation of2'FL might possibly be attributed to the presence 361of a significant population of bacteroides in FF infants, some 362strains of which have been documented to readily utilize HMO 363as a sole source ofcarbohydrate.^^ In contrast, the variation in 364the ability of microbiota fi'om BF infants to catabolize 2'FL 365might have been caused by either the presence or absence 366ofbifidobacteria species that are capable ofutilizing HMO^® or 367differences in the levels of bifidobacteriaspecies that are capable 368of utilizing HMO in the gut. Although we did not determine 369whether the mothers of the infantswho partidpated in the study 370were secretors ornonsecretors, '̂ itisalso possible thatvariability 371in the fiicosylated oligosaccharide content of breast milkcould 372influence the abilityof the infant colonic microbiota to at least 373degrade thisparticularHMO.Additional e^eriments areneeded 374to verify the difference in the ability of fecal microbiota from 375BFand FF to degrade various HMO and to determine whether 376

Fold of Change (FF / BF)*

COMPOUND Fonnula-Fed vs. Breast-4^ed

Blank COS 6'SL 2TL LNnT HP GA

Oh 3h 1 Oh 3h 6h Oh 3h 6h Oh 3h 6h Oh 3h 6h Oh 3h 6h Oh 3h 6h

gamma-aminobutyrate (GABA)

0.77 0.47 C*3'1 0.54 0.63 1.15 0.38 0.58 1.02 0.30 0.50 0.67 0.69 0.89 0.61 0.58 0.51 1.19 1.53 0.51 0.96

cadaverfne 2.87 3.11 0 1.71 1.80 6.04 2.70 2.73 0.94 0.94 3.85 0.96 3.09 1.88 0.81 6.19 1.62 1.15 1.99 0.13 2.32

tyiamine 0.71 1.14 0.88 0.73 0.68 1.23 0.51 1.10 0.47 0.62 2.43 1.02 0.91 1.74 1.31 0.46 1.29 0.94 0.76 0.98 1.50

tryptamine 1.85 1.27 1.22 1.23 1.23 1.22 1.28 1.50 1.30 1.32 1.37 1.23 1.32 1.24 0.98 1.11 1.22 1.64 1.10 1.35 1.29

putresdne 2.99 1.84 0.44 2.00 0.19 3.53 0.92 8.39 1.50 0.34 7.98 2.94 1.42 3.10 2.57 0.85 1.26 1.28 1.45 V3 3S

agmatlne 4.15 0.38 0.08 0.98 0.38 0.40 0.71 0.43 0.09 0.80 2.02 1.10 0.93 0.68 3.5f 1.37 0.78 [Tm 1.42 0.14 0.19

spemtidlne 2.00 1.29 0.89 1.38 1.09 0.83 1.43 1.32 1.03 1.34 0.70 0.68 1.42 0.78 0.67 1.18 1.32 0.92 1.52 1.20 1.36

"Shaded ceUs indicate P < 0.05 (red indicates that the mean values are significantly higher for that comparison; green values significantly lower).Blue-bolded text indicates O.OS <p < 0.10. Noncolored text and cells indicate that mean values are notsignificandy different for thatcomparison.Figure 5. Fold changes (FF/BF) ofbiogenic amines in FFvs BF infants during 6 h fermentation.

clx.doi.Org/10.1021/pr500011w I/. Proteome Res.XXXX, XXX, XXX-XXX

Journal of Proteome Research

377 secretor status influences the ability ofthe microbiota todegrade378 fiicosyiated oligosaccharides.

379 Biogenic Amines

380 Emerging research has linked the intestinal microbiome to both381 brain development and behavior^ via the brain—gut—enteric382 microbiota axis.^° This relationship is mediated in part through383 signaling molecules generated bythe gut bacteria^ such as the384 biogenic amine/neurotransmitter gamma-amino butyric acid.^^385 On the basis of previous work showing elevated levels of the386 biogenic amine, tyramine, infeces from healthy infants fed either387 cow's milk or cow milk formula compared to those who were388 breast fed,^^ we originally hypothesized that (a) stool samples389 from the FF infants woiJd contain higher relative levels of the390 various amines than thosefrom BF infants and (b) the addition391 of fermentable carbohydrate would suppress accumulation of392 amines. However, our limited data suggest that this was not393 necessarily thecase, at least with the biogenic amines thatwere394 identified in the current study. To illustrate, at time = 0 h,outof395 seven biogenic amines identified, only asingle amine, tryptamine,396 was higher in FF than in BF infants (Table 3). Furthermore,397 provision of fermentable carbohydrate actually resulted in398 increases rather than decreases in several biogenic amines (e.g.,399 GABA, cadaverine, and agmatine) regardless ofdiet (Figure 3).400 Lastly, the excess production of amines by samples from FF401 infimts depended on the specific amine analyzed and on the402 type ofcarbohydrate that was added to thecultures (Figure 5).403 Although we did not anticipate that supplementation with404 fermentable carbohydrate would increase generation ofamines,405 our data are consistent with the observation that the ability to406 produce amines iswidespread among human intestinal bacteria. '̂̂407 Other Compounds of Dietary Origin

408 Direct comparison ofthe time = 0 h blank samples also showed409 that BF and FF infants contained elevated levels of several410 metabolites that presumably originated from breast milk and411 infant formula, respectively. To illustrate, samples from FF412 infants contained at least 8-fold more a- and y-tocopherol than413 those of BF infants. Because tocopherols are considered an414 essential nutrient and cannot be synthesized endogenously, the415 presence of elevated levels of tocopherols in the stool samples416 ofFF suggest two possible scenarios: (a) tocopherols are being417 provided in the diet at levels in excess of infrnt need or (b)418 tocopherols provided in infrnt formula are not completely419 absorbed byin the small intestine. If (a) were true, one might420 consider lowering the amount of tocopherols that are421 incorporated into infrnt formula. On the other hand, if (b)422 were true, further research would beneeded to identify methods423 for improving tocopherol absorption by infants.

424 • CONCLUSIONS

425 In this study, we used metabolomics analyses to compare the426 fermentation profiles generated byfecal inocula from BF andFF427 infents. Comparison ofthe samples at time = 0 h revealed signs428 of carbon limitation and predominant protein fermentation in429 samples from FF infants versus the presence of HMO and less430 carbon restriction inthe BF group. Furthermore, thecomparison431 revealed differences in relative levels of some compounds that432 were most likely acquired through either breast milk (e.g.,433 linoelaidate) or infant formula (e.g., tocopherols, soy-based434 compounds). Supplementation with fermentable carbohydrates435 led to the accumulation of compounds indicative of energy436 generation. Cultures from BF infants that were not supple-437 mented with carbohydrate (blank) did not accumulate

compounds indicative of protein fermentation, but those from 438FF infants continued to accumulate compounds involved in 439protein catabolism. Although our results suggest that supple- 440mentation with fermentablecarbohydrateshad little effecton the 441accumulation of biogenic amines, further investigations are 442needed to first determine thebiological significance ofamino add 443catabolites, induding biogenic amines, on infant gut health and, 444more broadly, on overall infant health. In light ofthese findings, 445fecal metabolomics appears to be a useful tool for assessing 446the quality of infant diets regardless of whether the source of 447nourishment comes from breast feeding or formula feeding. 448

• AUTHOR INFORMATION 449

Corresponding Author 450

*E-mail: gcfahey(®illinois.edu. Fax: 217-333-7861. Tel.: 217- 451333-2361. 452

Notes 453

The authors declare the following competing financial 454interest(s): J.M.C. is an employee of Abbott Nutrition. M.R.P. 455is a University of Illinois graduate student and co-op student 456employee of Abbott Nutrition. 457

• ACKNOWLEDGMENTS 458

This research was supported by finandal backing from Abbott 459Nutrition. 460

ABBREVIATIONS USED 461

COS, galactooligosaccharide; GA, gum arabic; 2'FL, 2'- 462fucosyllactose; 6'SL, 6'-sialyllactose; LNn, lacto-N-neotetraose; 463HP, inulin; SCFA, short-chain fatty adds; BF, breast-fed; FF, 464formula-fed; HMO, human milkoligosaccharides 465

• REFERENCES 466

(1) Mai, V.j Torrazza, R. M.; Ukhanova, M.; Wang, X.; Sun, Y.; Li, N.; 467Shuster, J.; Sharma, R,; Hudak, M. L.; Neu, J. Distortions in 468development of intestinal microbiota assodated withlate onset sepsis 469in preterm infants.PLoSOne2013, 8, e52876. 470(2) Mai, v.; Young, C.M.; Ukhanova, M.; Wang, X.; Sun, Y.; Casella, 471

G.; Theriaque, D.; Li, N.; Sharma, R.; Hudak, M.; Neu, J. Fecal 472microbiota in prematureinfants priorto necrotizing enterocolitis. PLoS 4730«c 2011, 6, e20647. 474(3)deWeerth, C.; Fuentes, S.; deVos, W.M.Crying ininfants: Onthe 475

possible role of the intestinal microbiota in the development of colic. 476Gut Microbes 2013, 4, 416—421. 477(4) Hufifhagle, G. B. The microbiota and allergies/asthma. PLoS 478

Pathog. 2010, 6, el000549. 479(5) Kalliom^, M.; CoUado, M. C.; Salminen, S.; Isolauri, E. Early 480

differences in fecal microbiota composition in children may predict 481overweight. Am.J. Clin. Nutr. 2008, 87, 534—538. 482(6) Zheng, X.; Zhao, A; Xie, G.; Chi, Y.; Zhao, L; Li, H.; Wang, C.; 483

Bao, Y.; Jia, W.; Luther, M.; Su, M.; Nicholson,J. K.;Jia, W. Melamine- 484inducedrenaltoxidtyismediated bythegutmicrobiota. Set. Transl. Med. 4852013,5, 172ra22. 486

(7) Diaz Heijtz, R.; Wang, S.; Anuar, F.; Q|an, Y.; Bjbrkholm, B.; 487Samuelsson, A; Hibberd, M. L.; Forssberg, H.; Pettersson,S. Nonnal 488gut microbiota modulates brain development and behavior.Proc. NatL 489Acad.Sci. U.SA. 2011, 7, 3047-3052. 490(8) Harmsen, H. J. M.; Wildeboer-Veloo, A C. M.; Raangs, G. C.; 491

Wagendorp, A A; Klijn, N.; Bindels, J. G.; Welling, G. W. Analysis of 492intestinal flora development in breast-fed and formula-fed infimts by 493using molecular identification and detection methods. J. Pediatr. 494GastroenteroL Nutr. 2000, 30, 61—67. 495(9) Fenders, J.; Thijs, C.; Vink, C.j Stelma, F. F.; Snijders, B.; 496

Kummeling, L; van den Brandt, P. A; Stobberingh, E. E. Factors 497

dx.dowrg/10.1021/pr50001 Iw IJ. Proteome Res.XXXX, XXX, XXX-XXX

Journal of Proteome Research

498 influencing the composition ofthe intestinal microbiota in early infancy.499 Pediatrics 2006, 118, 511-521.500 (10) Le Huerou-Luron, I.; Blat, S.; Boudry, G. Breast- v. formula-501 feeding; Impacts on the digestive tract and immediate and long-term502 health effects. Nutr. Res. Rev. 2010,23, 23-36.503 (11) Martin, F. J.; Sprenger, N.; Montoliu, I.; Rezzi, S.; Kochhar, S.;504 Nicholson,). K. Dietary modulation ofgut functional ecology studied by505 fecal metabonoraics. ]. Proteome Res. 2010, 9,5284-5295.505 (12) Poroyko, V.) Morowitz, M.; Bell, T.; Ulanov, A.; Wang, M.j507 Donovan, S.j Bao, N.j Gu, S.; Hong, L.; Alverdy,J.C.; Bergelson,J.; Liu,508 D. C. Diet creates metabolic niches in the "immature gut" that shape509 microbial communities. Nutr. Hasp. 2011,26, 1283-1295,510 (13) Francavilla, R,; Calasso, M.; Calace, L.j Siragusa, S.; Ndagijimana,511 M.; Vernocchi, P.; Brunetti, L.; Mando, G.; Tedeschi, G.; Guerzoni, E.;512 Indrio, F.; Laghi, L.; Miniello, V. L.; Gobbetti, M.; De Angelis, M. Effect513 oflactose on gut microbiota andmetabolome ofinfants withcow'smilk514 allergy. Pediatr. Allergy Immunol. 2012, 23,420-427.515 (14) Vester Boler, B. M.; Rossoni Serao, M. C.; Faber, T.A.j Bauer, L.516 L.; Chow, J. M.; Murphy, M. R.; Fahey, G.C., Jr. Invitro fermentation517 characteristics of select nondigestible oligosaccharides by infant fecal518 inocula. /. Agric. Food Chem. 2013, 61, 2109-2119.519 (15) Lawton, K. A.; Berger, A.; Mitchell, M.; Milgram, K. E.; Evans, A.520 M.; Gui, L.; Hanson, R, W.; Kalhan, S. C.; Ryals, J. A.; Milbum, M. V.521 Analysis of theadult human plasma metabolome. Future Med. 2008, 9,522 3 83-397.

523 (16) Evans, A. M.; DeHaven, C. D.; Barrett, T.; Mitchell, M.; Milgram,524 E. Integrated, nontargeted ultrahigh performance liquid chromatog-525 raphy/electrospray ionization tandem mass spectrometry platform for526 the identification and relative quantification of the small-molecule527 complement of biological systems. Anal. Chem. 2009, 81,6656-6667.528 (17) DeHaven, C. D.; Evans, A. M.; Dai, H.; Lawton, K. A529 Organization of GC/MS and LC/MS metabolomicsdata into chemical530 libraries. ]. Chemiinjormatics 2010,2, 9.531 (18) DeHaven, C. D.; Evans, A. M.; Dai, H.; Lavrton, K. A. Software532 Techniques for Enabling High-Throughput Analysis of Metabolomic533 Datasets, Metabolomics. Roessner, U., Ed.j InTech: Rijeka, Croatia,534 2012; pp 167—192. http://www.intechopen.com/books/535 metabolomics/software-techniques-for-enabling-high-throughput-536 analysis-on-metaboloraic-datasets.537 (19) Heavey, P. M.; Savage, S. H.; Parrett, A;Cecchini, C.; Edwards, C.538 A; Rowland, I. R. Protein-degradation products and bacterial enzyme539 activities infaeces ofbreast-fed and formula-fed infants. Br.J. Nutr. 2003,540 89, 509-515.

541 (20) Center for Food Safety and Applied Nutrition, Food and Drug542 Administration, Department of Health and Human Services. Assess-543 ment of nutrient requirements for infant formulas. ). Nutr. 1998, J2g,544 2110S-2I29S.

545 (21) Boudry, G.; Morise, A; Seve, B.; Le Huerou-Luron, I. Effect of546 milk formula protein content onintestinal barrier function inaporcine547 model of LBW neonates. Pediatr. Res. 2011, 69, 4—9.548 (22) Chatelais, L.;Jamin, A; Gras-Le Guen, C.; Lailes,J.; Le Huerou-549 Luron,I.; Boudry, G.The level of protein in milkformula modifies ileal550 sensitivity to LPS later in life in a piglet model. PLoS One 2011, 6,551 eI9594.

552 (23) Boudry, G.; Jamin, A;Chatelais, L; Gras-LeGuen, C.; Michel, C.;553 Hueron-Luron, I. Dietary protein excess during neonatal life alters554 colonic microbiota and mucosal response to inflammatory mediators555 later in life in female pigs. J. Nutr. 2013, 143, 1225-1232.556 (24) Richardson, AJ.; McKain, N.; Wallace, R.J.Ammonia production557 by human faecal bacteria, and the enumeration, isolation and558 characterization ofbacteria capable ofgrowth on peptides and amino559 acids. BMC Microbiol. 2013,13, 6.560 (25) Edwards, C. A; Parrett, A M. Intestinal flora during the first561 months of life: Newperspectives. Br. J. Nutr. 2002,88,S11-S18.562 (26) Abrecht, S.; Stools, H. A; van den Heuvel, E. G. H. M.; Voragen,563 A G.J.;Gruppen, H.Occurrence ofoligosaccharides infeces ofbreast-564 fed babies in their first six months oflife and the corresponding breast565 milk. Carbohydr. Res. 2011, 346, 2540-2550.

(27)Marcobal, A; Barboza, M.; Froehlich,J.W.; Block, D.E.; German, 566J. B.; Lebrilla, C. B.; Mills, D. A. Consumption of human milk 567oligosaccharides by gut-related microbes. /. Agric. FoodChem. 2010, 58, 5685334-5340. S69

(28)Ward, R. E.; Ninonuevo, M.; Mills, D.A; Lebrilla, C.B.; German, 570J. B. In vitro fermentability of human milk oligosaccharides by several 57istrains of bifidobacteria. Mol. Nutr. Food Res. 2007, 51, 1398—1405. 572

(29) Kunz, C.; Rudloff, S.; Baier, W.; Klein, N.; Strobel, S. 573Oligosaccharides in human milk; Structural, functional, and metabolic 574aspects. Annu. Rev. Nutr. 2000, 20, 699—722. 575

(30) Rhee, S. H.; Pothoulakis, C.; Mayer, E. A Principles and clinical 576implications of the brain—gut—enteric microbiota axis. Nat. Rev. 577Gastroenterol. Hepatol. 2009, 6, 306-314. 578(31) Matsumoto, M.; Kibe, R.; Ooga, T.; Aba, Y.; Sawaki, E.; Koga, Y.; 579

Benno, Y. Cerebral low-molecular metabolites influencedby intestinal 580microbiota: A pilot study. Front. Syst. Neurosci. 2013, 7, 1—19. 581(32) Barrett, E.; Ross, R P.; O'Toole, P. W.; Fitzgerald, G. F.; Stanton, 582

C. y-aminobutyric acid production by culturable bacteria from the 583human intestine./.Appl. Microbiol. 2012,113, 411—417. 584(33)Murray, K. E.j Adams, R. F.; Earl, J.W.j Shaw, K.J.Studies of the 585

free frecal amines of infants vrith gastroenteritis and ofhealthy infonts. 586Gut 1986, 27, 1173-1180. 58?

(34) Macfarlane, S.; Macfarlane, G. T. Proteolysis and amino acid 588fermentation. In Human ColoHic Bucferifl; Role in Nutrition, Physiology 589and Pathology, Gibson, G. R.,Macfarlane, G. T., Eds.; CRC Press: Boca 590Raton, FL, 1995; pp 75—100. 591

dx.doi.or9/lO.lo2i/prSOOOl lwl7. Proteome fles. XXXX, XXX. XXX-XXX