Comparative Nontargeted Profiling of Metabolic Changes in Tissues and Biofluids in High-Fat Diet-Fed Ossabaw Pig

Comparative Nontargeted Profiling of Metabolic Changes in Tissuesand Biofluids in High-Fat Diet-Fed Ossabaw PigKati Hanhineva,*,† Thaer Barri,‡ Marjukka Kolehmainen,† Jenna Pekkinen,† Jussi Pihlajamaki,†

Arto Vesterbacka,† Gloria Solano-Aguilar,§ Hannu Mykkanen,† Lars Ove Dragsted,‡ Joseph F. Urban, Jr.,§

and Kaisa Poutanen†,∥

†Institute of Public Health and Clinical Nutrition, University of Eastern Finland, P.O. Box 1627, FIN-70211 Kuopio, Finland‡Department of Nutrition, Exercise and Sport, University of Copenhagen, Rolighedsvej 30, DK-1958 Frederiksberg-C, Denmark§U.S. Department of Agriculture, Agricultural Research Service, Beltsville Human Nutrition Research Center, Diet, Genomics, andImmunology Laboratory, Beltsville, Maryland 20705, United States∥VTT Technical Research Centre of Finland, P.O. Box 1000, Tietotie 2, FI-02044 VTT, Finland

*S Supporting Information

ABSTRACT: Typical clinical biomarker analyses on urine and plasma samplesfrom human dietary interventions do not provide adequate information aboutdiet-induced metabolic changes taking place in tissues. The aim of this studywas to show how a large-scale nontargeted metabolomic approach can be usedto reveal metabolite groups for generating new hypotheses of obesity-relatedmetabolic disturbances produced in an animal model. A large spectrum ofmetabolites in the semipolar region, including small water-soluble moleculeslike betaine and dihydroxyindole, and a wide range of bile acids as well asvarious lipid species were detected. The high-fat diet influenced metabolichomeostasis of Ossabaw pigs, especially the lipid metabolome, throughout allthe analyzed sample types, including plasma, urine, bile, liver, pancreas, braincortex, intestinal jejunum and proximal colon. However, even dramaticmetabolic changes in tissues were not necessarily observed in plasma and urine.Metabolite profiling involving multiple sample types was shown to be a feasible method for the examination of a wide spectrumof metabolic species extending from small water-soluble metabolites to an array of bile acids and lipids, thus pointing to thepathways of metabolism affected by the dietary treatment.

KEYWORDS: metabolite profiling, metabolomics, high-fat diet, nutrition, pig

■ INTRODUCTIONHuman dietary interventions have the inherent restriction inlimiting analysis mostly to easily accessible biofluid samples.Although analysis of urine and/or plasma may be linked toclinical phenotypes and complications (e.g., for biomarkerdiscovery), they are not adequate to assess the metabolic effectsinduced by dietary change in various tissues and organs. Morestudies at the tissue level could reveal effects of dietary changeon metabolic homeostasis and thereafter the function of vitalorgans that ultimately reflect whole body health status. This canbe achieved by using animal models.Large-scale metabolomic approaches offer a wider analyte

spectrum than available with typical clinical biomarkers.Improvements in mass spectrometry technologies within thepast decade have provided an extremely accurate, sensitive, andhigh-throughput approach to explore the content of smallmetabolites (<1000 Da) in virtually any biological material.This capacity can be utilized in nontargeted metabolite profilingthat examines a wide range of chemical species, in contrast totraditional, hypothesis-driven, targeted metabolite analyses.1,2

The hypothesis-free profiling is particularly valuable in assessing

alterations in metabolism under various dietary and diseaseperturbations by its potential to provide robust wide-scalemetabolite information.Obesity is a nutrition-related disorder that has reached

epidemic proportions in many populations around the worldand is a risk factor for the metabolic abnormalities related toseveral chronic diseases.3 The risk is mainly mediated throughthe metabolic syndrome,4 which is featured by central obesity,dyslipidemia, hypertension, and insulin resistance. Theresponse of liver to the energy excess is one of the keydeterminants of obesity-related metabolic risks, since theabnormalities are to a large extent explained by liver fataccumulation.5−7 Still, the molecular mechanisms by whichdifferent nutrients may be involved in the pathogenesis ofobesity are poorly understood.8 Over the past few years therehas been a steep increase in metabolomic studies that examine,for example, insulin resistance and type 2 diabetes mellitus(T2DM), and their potential to provide further insight into the

Received: March 22, 2013Published: June 27, 2013

Article

pubs.acs.org/jpr

© 2013 American Chemical Society 3980 dx.doi.org/10.1021/pr400257d | J. Proteome Res. 2013, 12, 3980−3992

pathophysiological mechanisms related to obesity-associateddisorders is high.9−12 However, many studies have beenrestricted to metabolite profiles of blood13−17 or urine.18,19

Only a few studies investigating tissue biopsies, such as adiposetissue20 and liver21,22 have been reported. In order to widen thescope of plausible analytics and range of sample types, animalmodels like the C57BL/6J mouse model of diet-inducedobesity23 and diet-induced diabetic Wistar rats24,25 have beenused. Most recently, swine has been used as an alternativeanimal model26 because of the many similarities to humans interms of diet and metabolism.27 In particular, the Ossabawminiature pig breed has been used to model human obesity,metabolic syndrome, and cardiovascular disease.28−30 Thisbreed has been adapted to store large amounts of fat forsurvival because of a feral life cycle, and it has a higherpropensity to obesity and obesity-related diseases under high-caloric diet compared to other pig models such as the Yukatan,Gottingen, and domestic swine.31−33

The aim of this work was to examine the potential of aparallel multiorgan and biofluid LC−MS-based nontargetedmetabolite profiling to investigate diet-induced metabolic

alterations in an experimental animal model. We analyzedbiofluids and tissues from Ossabaw pigs fed either a basal orhigh-fat (HF) diet. By data-driven clustering, we showed thatthe metabolic effects of the HF diet leading to obesity wereobserved in all the biofluids examined: plasma, urine, and bile.More importantly, we observed tissue-specific metabolitepattern changes in liver, pancreas, brain cortex, jejunum, andproximal colon, not all similarly reflected in plasma or urine.We thus propose that parallel profiling of multiple biofluids andtissues in an animal model is a useful approach when studyingdiet-induced metabolic effects.

■ RESULTS

Female Ossabaw pigs of approximately 19 weeks of age wererandomized by weight into two groups of five pigs each (bodyweights 10.2 ± 0.4 and 10.0 ± 0.3 kg, respectively) that werefed either a basal or HF diet (Table 1). A whole body DXAscan taken at week 28 showed no significant difference in totalfat in pigs fed the HF diet (7.1 ± 1.5 kg) compared to the pigsfed the basal diet (6.4 + 2.1 kg), but at week 35 there was a

Table 1. Consumption, Nutritional Content, and Composition of the Test Diets

phase I phase II

est nutr content (% dry matter) basal high fat basal high fat

avg energy intake (kcal/day) 1020−2041 1135−2269 1996−3706 2450−4550avg feed consumed (kg/day) 0.35−0.7 0.35−0.7 0.7−1.3 0.7−1.3protein (%) 16.47 16.43 15.92 15.49energy from protein (kcal/kg) 658 657 637 619carbs (%) 48.01 44.12 44.83 42.78energy from carbs (kcal/kg) 1920 1764 1793 1710from starch (%) 45.06 38.87 41.83 35.32from fructose (%) 0 2.5 0 5from other (%) 2.95 2.75 3 2.46fat (%) 3.75 9.12 4.68 13energy from fat (kcal/kg) 338 821 421 1170fiber (%) 2.6 2.43 4.71 2.26ash (%) 6.1 6.11 6 6.09added amino acids (%) 3.19 3.22 2.86 2.81vitamin and trace minerals 6.97 6.77 8.6 6.56tot dry matter (%) 87.09 88.2 87.6 88.9tot water (%) 12.9 11.8 12.4 11

phase I phase II

ingred (% tot vol) basal high fat basal high fat

corn 72.5 62.5 62.5 56.78soy (47% protein) 21.5 23.2 16.5 22.5oat bran 0 0 15 0soybean oil 0.8 5.1 1.12 5.1corn oil 0 0.7 0 4.5fructose 0 2.5 0 5cholesterol 0 0.7 0 1dicalcium phosphate dihydrate 1.9 2 1.5 1.8calcium carbonate 0.65 0.6 0.9 0.7tryptophan 0.05 0.05 0.03 0.03methionine 0.14 0.16 0.1 0.1L-lysine 0.47 0.45 0.37 0.28L-threonine 0.1 0.1 0.09 0.07sodium chloride 0.4 0.4 0.4 0.6coline chloride 0 0.05 0 0.05selenium premix 0.04 0.04 0.04 0.04swine mineral 0.8 0.8 0.8 0.8swine vitamin 0.65 0.65 0.65 0.65

Journal of Proteome Research Article

dx.doi.org/10.1021/pr400257d | J. Proteome Res. 2013, 12, 3980−39923981

significant (p < 0.03) increase in total fat in the HF groupcompared to pigs fed the basal diet (Figure 1A), as well as a

significant difference in body weight at week 37 (Figure 1B).Serum total, HDL and LDL cholesterol (p < 0.001) andglucose (p < 0.01) levels in pigs fed the HF diet at week 35differed significantly from those in pigs fed the basal diet (Table2). Serum insulin, triglyceride, LDH, ALT, AST, and GGTlevels between the two groups were not significantly different(Table 2).

The nontargeted LC−MS metabolite profiling of tissues andbiofluids from the two dietary groups of pigs resulted in a dataset containing ca. 9000 features. First, the data set was dividedinto subsets of each biological sample type, containing both dietgroups, and the data was ranked based on the intensity of thefeatures within each sample type. Hierarchical clustering wasperformed on the relative peak area values of the 2000 mostintense features observed in each sample type as exemplified forplasma, liver, and pancreas (Figure 2). The clustering reveals adistinct diet-induced pattern in metabolite content in eachsample type and suggests that the clusters are more clearlyseparated in case of liver and pancreas than plasma.

Next, the metabolite markers in each sample type wereranked based on univariate comparison (Student’s t-test). Themost significantly altered metabolites across the entire sampleset were selected by retaining the 2000 most differentialmarkers based on the p-value and subjected to partial leastsquares discriminant analysis (PLS-DA) in order to visualizethe clustering pattern of all the different sample types fromboth of the diet groups. The PLS-DA plots separately depictclusters of tissues (liver, proximal colon, jejunum, pancreas, andbrain cortex), plasma and bile, and urine (Figure 3).Furthermore, the clusters are separated based on the twodietary treatments for all sample types, as can be seen in theinsets in Figure 3. In plasma samples, there is HF-dependentprojection toward the left in the chart. In the case of bile, thereis also clear separation between the HF and basal diet groups,but the projection of the metabolite change is in a differentdirection, most likely resulting from a different metabolitecomposition. Notably, within the different tissues, the clearestseparation between HF and basal groups is in the liver samples.The data set of 2000 metabolite markers was further reduced

by selecting only markers having a remarkable change betweenthe two diets in at least one sample type (p < 0.05; fold change>50%), and with a peak intensity value >10. This data set(containing 141 markers) represented those metabolite signalsthat showed the maximum difference between the two diets atleast in one of the examined sample types, and that had a highenough signal in the MS for structural elucidation by MS/MSanalysis. The list of assigned metabolites is available asSupporting Information (Table S1). The marker group wasclustered according to the occurrence of the metabolite signalin the different samples. The quality threshold (QT) clusteringresulted in seven clusters holding markers with similar features.The clusters are shown as heat map plots in Figure 4.The largest cluster (Cluster 1) in the analysis is formed of

mainly bile acids and phosphatidylcholines (PC) that are highlyabundant in the bile samples, and scattered among othertissues. This cluster includes two highly abundant PCs whichare present also in the plasma sample, the PC(38:4) andPC(36:4) (Figure 5). These two lipids have high intensity inthe plasma, bile, and liver samples and show lower signals inpigs fed the HF diet. These PCs are also detected in smalleramounts in the other tissues with HF diet-induced lowerintensity in the pancreas, jejunum, and proximal colon buthigher levels in the brain cortex from pigs fed a HF diet. Inaddition, other phosphatidylcholines such as PC(36:5) andPC(38:6) were detected at various levels in different sampletypes. For example, PC(36:5) is lower and PC(38.6) higher inliver from pigs fed a HF diet.The most abundant bile acid detected in the analysis is

glycocholic acid that is represented in Cluster 1 with severaldifferent ions including the pseudomolecular ion, dimer, andfragment ions after several losses of water. All of the ions arehigher in the bile and jejunum samples from pigs fed a HF diet(Figure 5). Other minor metabolites classified as bile acidswithout further identification were detected based on theelution region, elemental composition, and fragmentationpattern and are distributed specifically among the differentorgans with either higher or lower signals in tissues from pigsfed a HF diet (Figure 5).Cluster 2 contains the most abundant lysophosphatidylcho-

lines (LPC) found in all the sample types except urine, and arehigher in plasma, bile, and liver from the HF-fed pigs.LPC(18:1) and LPC(18:2) are higher and LPC(16:1) and

Figure 1. (A) Total body fat (kg) determined by DXA scan at week35. (B) Body weight (kg) at week 37. Mean weight (n = 5/group) ±SEM are plotted.

Table 2. Biochemical Profile in Serum from 24-h-FastedPigsa

basal high fat p-value

body wt (kg) 35.6 ± 0.6 42.2 ± 1.1* <0.0001plasma cholesterol(mg/dL)

92.20 ± 7.4 201.4 ± 22.6* <0.0001

HDL cholesterol(mg/dL)

44.2 ± 3.8 70.0 ± 2.4 * <0.0001

LDL cholesterol(mg/dL)

35.60 ± 3.35 130.60 ± 21.53* <0.0001

plasma triglycerides(mg/dL)

38.0 ± 3.2 37.2 ± 3.6 0.8372

LDH (IU/L) 424.0 ± 11.89 563.60 ± 134.4 0.2175ALT (IU/L) 49.8 ± 1.59 43.80 ± 2.49 0.405AST (IU/L) 36.0 ± 1.51 52.0 ± 10.15 0.2265GGT (IU/L) 37.40 ± 1.77 46.8 ± 5.47 0.1147glucose (mg/dL) 93.60 ± 3.50 111.20 ± 3.13 0.0104insulin (μg/L) 0.72 ± 0.19 0.80 ± 0.23 0.8442aMean values ± SE for biochemical parameters. P-values after one wayANOVA comparison of means between treatment groups.

Journal of Proteome Research Article

dx.doi.org/10.1021/pr400257d | J. Proteome Res. 2013, 12, 3980−39923982

LPC(16:0) are lower in pancreas from pigs fed a HF diet(Figure 5). Cluster 2 has also the marker for betaine that wasfound in all the sample types except for bile, and gave the mostintense signal in the pancreatic tissue. Betaine is lower in all thesamples from pigs fed a HF diet (except brain cortex). InCluster 2 there are also various other lipid signals most likelyfor carnitines, phosphatidylethanolamines, phosphatidylserines,and ceramides that were not subjected to detailed identification.Cluster 3 contains metabolites found particularly in the

proximal colon. These include bile acids and lysophospholipids.For example, LPC(P-18:0) is lower and LPC(20:2) is higher inproximal colon tissue from pigs fed a HF diet. However, bothare higher in liver of pigs fed a HF diet. The proximal colonsamples show several minor bile acid signals which weredetected in the bile in only trace levels such as a metabolitetentatively identified as hydroxy oxo-cholanoic acid ordihydroxycholenoic acid (Figure 5). Notably, glycocholic acidis the most abundant metabolite in bile, but it was not detectedin proximal colon.Cluster 4 contains mainly urinary metabolites. Dihydrox-

yindole and salicyluric acid are lower in urine samples from pigsfed a HF diet. Although hippuric acid is also lower in urinesamples from pigs fed a HF diet (Figure 6), it is notrepresented in the cluster map because it did not pass thefiltering criteria for significant metabolites (p < 0.05; foldchange >50%) in any sample type (values for hippuric acidwere p-value 0.007, fold change 0.67).Cluster 5 contains phosphatidylcholines including PC(38:5)

and PC(36:3) as well as lysophosphatidylcholine LPC(20:3)that is higher in pigs fed a HF diet, especially in the liver,plasma, and bile. Cluster 6 contains phosphatidylcholines andother lipids that have high values in plasma. Some, likePC(42:7), increased, while PC(40:5) decreased, and others likePC(36:2) remained relatively constant in samples from pigs feda HF diet. Cluster 7 (Figure 4) contains a group of unknownmetabolites found in some tissues, especially in the brain cortex.This cluster also contains LPE lipids primarily in tissue frombrain cortex.A further examination of the most abundant bile acids

revealed an interesting pattern especially in the bile samples.

The most striking difference was observed between the tauro-and glycine conjugated metabolites, since the two identifiedtauro -conjugated bile acids (tauro-ursodeoxycholic acid andtauro-muricholic acid) were decreased in the bile of the pigs fedthe HF diet, while the two most abundant glycine conjugatedspecies (glycocholic acid and trihydroxyoxocholanyl-glycine)were increased in the HF diet fed pigs in bile, jejunum, and liversamples (Figure 7). Notably, in the plasma and tissue samplesmainly the secondary bile acids were found (glycodeoxycholicacid and glycoursodeoxycholic acid) (Figure 7). In the braincortex, we did not detect any bile acids.

■ DISCUSSION

The objective of the present work was to examine whether aparallel multiorgan and biofluid metabolite profiling approachwould be feasible in an experimental animal model toinvestigate diet-induced metabolic alterations at the organand tissue levels. The Ossabaw pigs fed a HF diet were used tomodel these comparisons. The HF diet induced significantweight gain, hypercholesterolemia and modest increase infasting glucose level indicating the effect of HF diet on theenergy metabolism, which has been observed also in otherstudies with the same pig model.31 In the metabolite profilingassay, diet-dependent alterations were detected in themetabolite profile of all the examined organs and biofluidsextending through the various metabolite classes detected. Thenontargeted approach captured a large spectrum of metabolitesin the semipolar region, including small water-solublemolecules like betaine and dihydroxyindole, and a wide rangeof bile acids as well as various lipid species. A parallel profilingof metabolite clusters based on their intensity throughout allthe examined sample types provided inter-related patterns ofthe metabolites in each of the tissue types examined. Evendramatic metabolic changes in tissues were not necessarilyobserved in the biofluids like plasma and urine, which highlightsthe importance of inclusion of tissue samples when addressingthe question of dietary impact on metabolism.As expected, the urinary metabolome was very distinct from

all of the other sample types, whereas the metabolite

Figure 2. Heat map charts based on hierarchical clustering for the 2000 most intense metabolite markers in samples from plasma, liver, and pancreasfrom Ossabaw pigs on a basal or high-fat diet.

Journal of Proteome Research Article

dx.doi.org/10.1021/pr400257d | J. Proteome Res. 2013, 12, 3980−39923983

compositions in various tissues shared more similar profilesthan in plasma or bile samples. Clustering based on PLS-DAanalysis indicated a metabolic shift in all sample types.However, the trajectory was different for plasma and bile dueto different metabolite composition. The pancreas, jejunum,and brain cortex from the HF-fed pigs projected upward in thechart similar to bile, whereas the proximal colon samples had aleftward trend like plasma; although the separation of thegroups was not as clear for proximal colon as for the other

sample types. Interestingly, the liver groups shifted up andleftward and could thus be interpreted as metabolic alterationssimilar for both plasma and bile. Importantly, the mostresponsive tissue in the Ossabaw pig model was liver, whichis also one of the key tissues in human metabolic syndrome.5−7

In addition to the general metabolite shift that wasdemonstrated throughout all the examined sample types, wecould identify several interesting metabolite patterns in this firstgeneral profiling assay, which will be briefly discussed here. A

Figure 3. PLS-DA analysis of the most differential metabolite markers throughout all the sample types from pigs fed either a basal (B) or high-fat(HF) diet. (Insets) Close-ups from the three separate clusters formed by tissues: (A) pancreas, liver, jejunum, proximal colon, and cortex; (B)plasma and bile; (C) urine.

Journal of Proteome Research Article

dx.doi.org/10.1021/pr400257d | J. Proteome Res. 2013, 12, 3980−39923984

more focused analysis of specific organs and metabolite groupsin terms of diet-induced obesity will be the subject of asubsequent works.Hippuric acid was decreased in the urine samples of the HF

diet fed pigs. Hippuric acid is generally regarded as a biomarkerof the dietary phenolic content as it is the final end product ofmicrobial degradation of phenolic compounds to benzoic acidand is then conjugated in liver with glycine.34,35 The potentialdietary sources for hippuric acid in our study varied betweenthe two groups at the latter phase of the study, since thecontent of corn was higher in the basal diet (62.5% in basal vs56.8% in HF), and there was added oat bran to the basal diet toadjust for the added fat containing ingredients in the HF diet.On the other hand, the HF diet was higher in soy and soybeanand corn oil, which also are potential sources of phenoliccompounds. However, in the data set of the most discriminateions between the two dietary groups (141 markers), we did notobserve any phenolic derived metabolites in the urine wherethey would typically accumulate after ingesting a phenolic richdiet.36,37 The reduced urinary hippuric acid level could be dueto HF diet-induced modulation in colonic microbiota, as hasbeen suggested also in other studies. Decrease in the hippuricacid levels in the urinary samples has been reported in animaltrials38 and human studies39 in the context of obesity relatedresearch. Similar observation has been made in a study ofCrohn’s disease where decreased hippurate levels were causedby altered gut microbial metabolism.40 Gut microbialmetabolism is strongly linked with the obesity phenotype,41,42

and thus the difference in the hippuric acid level observed inthe Ossabaw pigs most likely implicates modulation of thecolonic microbial composition or its function by the HF diet.One of the most interesting metabolite findings was the

decrease in betaine (trimethylglycine) levels in plasma,pancreas, jejunum, and proximal colon in the HF diet-fedpigs. Betaine is naturally present in a wide variety of tissues andcells, and it is gained either from dietary sources or producedendogenously from choline by oxidation. In tissues, betaine hasa dual role functioning as an osmolyte and as a methyl groupdonor in the formation of methionine and S-adenosylmethio-nine.43 Several studies with different animal models of obesityshow decreased betaine levels in liver, plasma or serum.44−46 Inhumans, low circulating betaine levels are associated withobesity related metabolic diseases.43 In fact, betaine supple-mentation attenuated development of the HF diet-induced fattyliver and associated insulin resistance in a mouse model,46 aswell as induced the accumulation of carnitines in muscle andliver of mice fed a HF diet (Pekkinen et al., accepted forpublication in “Molecular Nutrition & Food Research”),elucidating the importance of the balance in betaine levels interms of energy homeostasis. In our study, the betaine levelsdecreased in the pancreas, intestinal jejunum, and proximalcolon in pigs fed a HF diet, indicating a previously undetectedmetabolic regulation of betaine in different intestinal andendocrine tissues. The higher betaine levels in the basal-diet-fedgroup might also reflect the oat bran added to adjust thevolume of the basal diet in phase II, as grains are an important

Figure 4. Most significantly affected metabolite markers detected in tissue and biofluid samples. The chart shows markers having p < 0.05, foldchange >50%, and peak area intensity >10 clustered according to their intensity profiles.

Journal of Proteome Research Article

dx.doi.org/10.1021/pr400257d | J. Proteome Res. 2013, 12, 3980−39923985

dietary source of betaine. The amount of oat bran added to thebasal diet in phase II of the study provided approximately 30mg of betaine daily (150 g of bran containing 20 mg/100 gbetaine according to USDA Food composition database:http://fnic.nal.usda.gov/food-composition/usda-nutrient-data-laboratory). The total betaine levels in the pig diet were notdetermined, but the daily dietary intake of betaine in humans isin the range 0.5−2 g.47 It is notable that the HF diet containedadded choline chloride for enhanced cholesterol digestibility,and this could have been as precursor for betaine synthesis invivo, yet the betaine levels were still lower in the HF group.Therefore, it can be postulated that the difference in betainelevels observed in the profiling reflected the altered (energy)metabolism due to the HF diet. In general, the role of betainein the development of obesity and related metabolic diseasescertainly needs more attention.The plasma metabolite profile was characterized by various

lipid species showing obvious changes between the two dietarygroups. In addition to plasma, the lipid profile varied in the bileand in different tissues with some sharing of metabolite clusterswith distinct patterns of expression that were clearly affected bythe diet. There have been various publications wherenontargeted metabolite profiling revealed alterations in the

lipid profile due to obesity or obesity-related symptoms both inanimal models and human studies mainly in plasma samples. Acharacteristic of these studies is extensive variation and evendiscrepancy in the results when individual lipid metaboliteshave been examined. For example, lysophosphatidylcholineLPC(18:1) decreased in plasma from subjects with impairedglucose tolerance,19 healthy obese men,48 and in HF diet-induced obese mice,45 but increased in obese twins49 and obeseZucker (fa/fa) rats.50 Likewise, in our study, pigs fed a HF diethad increased LPC(18:1) in all the samples analyzed exceptbrain cortex (and urine). A very similar comparative literaturereview can be done on many of the PC or LPC metabolitesdetected in our study. The heterogeneity of published datalikely resulted from differences in study design, the animalmodel, and characteristics of free living human studypopulations, as well as intensive interindividual variation. Inaddition, differences in analytical methods (MS, NMR),extraction protocols, and identification of relevant molecules,computing, analysis, and reporting of results often confoundcomparisons between studies. In our study, the changesobserved in the lipid patterns between and within each tissueand biofluid were at least diet-related but complex, suggestingthat each of the examined tissues has specific regulation of the

Figure 5. Metabolite differences (peak area intensities) in Clusters 1−3. Blue bars represent samples from pigs fed the basal diet and the red barsrepresent samples from pigs fed the HF diet. Mean peak area values (n = 5/group) ± SEM are plotted.

Journal of Proteome Research Article

dx.doi.org/10.1021/pr400257d | J. Proteome Res. 2013, 12, 3980−39923986

lipid pattern that is affected by the dietary modulation, andwould necessitate a more detailed lipid-focused analysis. Thefact that phospholipids often come out as markers withoutconsistent patterns between studies necessitates analysis withexactly the same diet but varying, for example, in study design,animal species, or analytical procedures to assess whether thesemetabolites can be regarded as robust biomarkers. With currentmetabolomic techniques, detailed lipidomic profiling can beused to gain a deeper insight into the proportional regulation ofvarious lipid species in addition to PCs and LPCs such asglycerophospholipids and sphingolipids.51−53

Another abundant metabolite group encountered in ouranalysis was the bile acids. Bile acids are a diverse class ofcholesterol derived metabolites in the enterohepatic circulationthat are synthesized in liver and metabolized by the colonicmicrobiota.54 They play a central role in the absorption ofdietary lipids, and more recently have been discussed aspotential signaling molecules. Recent studies on germ-freerodent models clearly show that microbiota has a central role in

controlling the bile acid synthesis and absorption evidenced byaltered composition of bile acids throughout the enterohepaticsystem and peripheral tissues in absence of microbiota.55,56

Dietary habits further impact this interplay, as there is a linkbetween the HF diet and microbial composition, and recentinvestigations suggest that the modulation of the microbiotacould be at least partly mediated by the increased excretion ofbile acids caused by HF diet, which puts strong selectivepressure on the gut microbiota.57 Interestingly, serum bile acidlevels increased after weight loss following Roux-en-Y gastricbypass surgery, indicating that bile acid metabolism maycontribute to improved glucose and lipid metabolism after thesurgery.58,59 In our study, the bile acid profiles were clearlyaltered in plasma, bile, and all tissue samples, especially jejunumand proximal colon. Most importantly, there was a clear HFdiet-related trend in the conjugation pattern of the mostabundant bile acids, as the taurine conjugated metabolitesdecreased (bile samples) whereas glycine conjugates increased(bile, liver, and jejunum samples). Studies with other animal

Figure 6. Metabolite differences (peak area intensities) in Clusters 4−6. Blue bars represent samples from pigs fed the basal diet and the red barsrepresent samples from pigs fed the HF diet. Mean peak area values (n = 5/group) ± SEM are plotted.

Figure 7. Accumulation of (A) taurine-conjugated and (B) glycine-conjugated bile acids in the various sample types. Mean peak area values (n = 5/group) ± SEM are plotted. Abbreviations are pancr, pancreas; jejun, jejunum; prox c, proximal colon.

Journal of Proteome Research Article

dx.doi.org/10.1021/pr400257d | J. Proteome Res. 2013, 12, 3980−39923987

models have shown a similar trend in plasma and liver samplesof rats on HF diet.60 Interestingly, similar shift in the bile acidconjugation pattern was observed also in rats subjected tochronic ethanol consumption.61 Analysis of mice fed a diet highin saturated (milk) fat showed a remarkable increase in thetaurine conjugated bile acids in bile samples.62 This is contraryto our findings and could be due to a different animal model or(most likely) because the fat in our study was mostly based onvegetable oils. Interestingly, in our study the most intensive bileacid signals in the various tissue samples all represent secondarybile acids that have undergone dehydroxylation by the colonicmicrobiota. These preliminary results obtained by the non-targeted profiling clearly show that the Ossabaw pig has apronounced diet induced response in the bile acid compositionthat could represent a model closer to human metabolism thanrodents in the context of investigating diet−bile acid−microbiota homeostasis and function. This approach wouldwork especially well with a targeted metabolite analysis of thebile acid pool in the various biofluids and intestinal andperipheral tissues.The metabolite information presented here was a fraction of

the total metabolite content measured because the limits set (p< 0.05; fold change >50%) for inclusion of metabolites that hada pronounced change in at least one sample type representedonly ca. 2% of the total number of signals detected. Thus, onlythe most relevant changes were reported to demonstrateclustered effects of a HF diet on a variety of tissues andbiofluids. However, smaller changes in metabolites of unknownbiological relevance were widely observed, and analysis of thesemetabolites could provide further insight into HF diet-inducedchanges in metabolism. Notably, in the tissue samples we couldsee numerous metabolite alterations which were not detected inthe fasting plasma, thus underlining the importance of animalstudies which enable the inclusion of tissue samples. Ouranalysis pointed out the metabolite classes that were mostaffected by the HF diet, for example, bile acids and betaine andrelated metabolites, and would benefit from focused examina-tion on multiple tissues and biofluids (including muscle andadipose tissue that were not included in this study) to provide acomprehensive view of energy metabolism and dysfunction tobetter understand the relationship between metabolic disordersand pathophysiological events related to obesity.

■ CONCLUSIONThis study illustrates the impact of a high-fat diet onmetabolome both at the tissue and biofluid level, and suggeststhat the multisample nontargeted metabolite profiling approachmay be useful in analysis of the molecular mechanisms ofdietary impact on whole body homeostasis. The methodologyused is feasible for concomitant examination of a vast pool ofmetabolites from different tissues and biofluids taken fromexperimental animals fed different diets.

■ MATERIALS AND METHODS

Maintenance of Animals and Experimentation

Ossabaw miniature female swine (sus scrofa) were obtainedfrom the Indiana University Ossabaw Production Unit (IndianaUniversity School of Medicine, Indianapolis, IN). According tolocal Animal Care and Use Committee (ACUC) recommen-dations, all pigs in the breeding unit were serologically testedand confirmed negative for Porcine Reproductive andRespiratory Syndrome virus, swine Influenza virus serotypes

H1N1 and H3N2, Actinobacillus pleuropneumoniae, andCircovirus. Fecal samples were confirmed absent of foodborne pathogens, Salmonella spp., pathogenic Escherichia colispp. and Campylobacter spp. After weaning at four weeks of age,pigs were vaccinated against Swine Influenza H1N1+ H3N2(FluSure XP, Pfizer Inc., New York, NY), ActinobacillusPleuropneumoniae-Bordetella Bronchiseptica-Erysipelothrix Rhu-siopathiae-Haemophilus Parasuis-Pasteurella multocida Bacterin(Parapleuro ShieldP + BE, Novartis Animal Health, US, Inc.)and Circovirus (CircoFLEX, Ingelvac USA). At five weeks ofage, pigs were transported in a temperature controlled vehicle(Transportech LLC, Brockton, MA) in kennels with access towater and delivered overnight to the United States Departmentof Agriculture, Beltsville Area Agricultural Research Center,Beltsville, Maryland according to standardized procedures ofquarantine and under the approval and supervision of BeltsvilleArea Animal Care and Use Committee. After arrival inBeltsville, pigs were housed in individual pens with free accessto water and fed a preweighed amount of pig grower diet basedon ground corn and soybean supplemented with recommendedlevels of amino acids, vitamins and minerals. The diet contained16.5% of protein, 48% from carbohydrates and 3.75% from fat(Table 1).At 19 weeks of age, pigs were randomized by weight and split

into two treatment groups fed either a basal (B) or high-fat(HF) diet. Pigs in the B group (n = 5) continued with initialgrower diet, while pigs in the HF group (n = 5) switched to adiet containing similar level of protein (16.4%), lesscarbohydrate (44.0%) and more fat (9.1%) (Table 1). Duringphase I (week 19−26) pigs from both dietary groups had anequal biweekly stepwise increase in the amount of feed given toeach pig each day to adjust for nutrient requirements duringgrowth. Starting at week 27 until week 37 (Phase II), pigs in theB group received a diet with 15.9% protein, 44.8% carbohydrateand 4.7% fat with a gradual increase in the amount of feed toprovide 3706 kcal/day through 37 weeks of age. Oat bran wasadded (15% of total volume) to the basal diet in phase II tobring the volume of the diet to the same level as the HF dietcontaining extra of fat from added corn oil (Table 1). At week27, pigs in the HF-diet-fed group were converted to a dietcontaining 15.5% protein, 42.8% carbohydrates and 13% fatwith a gradual increase in the amount of feed to provide 4550kcal/day through 37 weeks of age. The diets for HF-diet-fedgroups also contained 0.7% cholesterol, 2.5% fructose, and 0.7%of corn oil in phase I (week 19−26) and 1.0% cholesterol, 5%fructose, and 4.5% corn oil in phase II (week 27−week 37). Inaddition, 0.05% choline chloride was added to the HF diets inphase I and II (Table 1). Food rations were individuallyweighed, monitored and recorded daily. Pigs in both groupsconsumed all feed given each day.Body weight was monitored throughout the experiment and

a dual-energy X-ray absorptiometry (DXA) measurement(Lunar Prodigy, Madison WI) was taken at week 28 and 35of age to provide a noninvasive indicator of total body fat. Priorto the DXA scan, pigs were anesthetized with 500 mg ketamine(Ketaset, Fort Dodge Animal Health, Iowa), 80 mg tiletamine(Telazol, Fort Dodge Animal Health, Iowa), 80 mg zolazepam(Telazol), and 333 mg xylazine (Xyla-Ject, Phoenix Pharma-ceutical, St Joseph, MO) per 100 kg body weight. At the time ofsacrifice, 24-h-fasted pigs were euthanized using EuthasolR(Virbac AH, Inc., Fort Worth, TX) followed by thoracotomy.Tissue samples were dissected aseptically from the left lobe ofthe liver, anterior surface of the pancreas, proximal colon at

Journal of Proteome Research Article

dx.doi.org/10.1021/pr400257d | J. Proteome Res. 2013, 12, 3980−39923988

∼30 cm from the ileal/cecal junction, jejunum at the midgut,frontal lobe of the brain cortex and were placed in cryotubes forimmediate freezing in liquid nitrogen. Biofluids (urine and bile)were aseptically collected with a 18 gauge needle attached to asyringe and were frozen immediately. Blood samples werecollected aseptically the day before necropsy from the rightjugular fossa of 24-h-fasted pigs using a 18-gauge needle and aBD vacutainer (Franklin Lakes, NJ) for serum and a Kendallvacutainer with 7.5 mg EDTA (Tyco Healthcare, Mansfield,MA) for plasma that were separated after centrifugation andstored at −80 °C. All samples were transported with dry ice(−78.5 °C) to Denmark for metabolomics analysis and storedat −80 °C.

Biochemical Parameters

Serum triglycerides (TG), total cholesterol (CHOL), low-density lipoprotein (LDL) cholesterol, high-density lipoprotein(HDL) cholesterol, alanine aminotransferase (ALT), aspartateaminotransferase (AST), gamma-glutamyltransferase (GGT),lactate dehydrogenase (LDH), glucose and insulin levels withcorresponding standards were measured using a VetAceClinical chemistry analyzer (Alfa Wasserman DiagnosticTechnologies, West Caldwell, NJ).

Sample Preparation for Metabolite Profiling

Plasma and urine samples were prepared in 96-well Sirroccoplate as described previously.63 Briefly, samples weresandwiched between two volumes (each 90 μL) of extractionsolvent (1:1, acetonitrile/methanol). The extract was thencollected and the precipitate was washed twice with two moresolvent volumes. The pooled extract was evaporated to drynessand the residue was dissolved in 100 μL of MeOH/H2O (1:1).The frozen tissue samples stored in cryotubes (liver,

pancreas, brain cortex, intestinal jejunum, and proximalcolon) were ground to fine powder with mortar and pestle inliquid nitrogen. Aliquots of the frozen tissue were weighed toprechilled eppendorf tubes (∼100 mg aliquots) and kept frozenin liquid nitrogen until extraction. The extraction wasperformed in a two-step procedure as described by Masson etal., 201064 with some modifications. First, MeOH:H2O (1:1)was added in a ratio of 3 μL of solvent/mg frozen tissue (∼100mg of tissue in each sample) to extract the polar and semipolarmetabolites. The sample was sonicated at room temperature for15 min, vortexed, centrifuged, and the supernatant wascollected. Next, dichloromethane/MeOH (1:1) was added tothe residue in the same ratio as in the first step to extractlipophilic metabolites. After sonication, vortexing and centrifug-ing, the supernatant was collected to a separate tube, and thetwo extracts were evaporated to dryness in a speedvaccentrifuge. Both samples were redissolved in 200 μL ofMeOH/H2O (1:1), and 50 μL aliquots from the two extractionsteps were combined and centrifuged prior to the LC−MSanalysis.

UPLC−qTOF-MS Analysis

An ultraperformance liquid chromatography (UPLC) systemcoupled to quadruple time-of-flight (Premier QTOF) massspectrometer (Waters Corporation, Manchester, UK) was usedfor sample analysis. The analysis was performed as describedpreviously for “Chromatography Method II”.63 In short, thechromatographic column used was HSS T3 C18 with mobilephase consisting of 0.1% formic acid in water (A) and 0.1%formic acid in “70% acetonitrile/30% methanol” (B). Positiveion acquisition mode in electrospray ionization was imple-

mented at a probe voltage of 3.2 kV. The selected m/z rangewas from 50 to 1000 Da. Mass spectral scan time and interscandelay were, respectively, 0.08 and 0.02 s. Leucine enkephalinwas infused intermittently every 10 s and used for accurateonline mass calibration. To get more structural information, alow-to-high collision energy ramp (MSE mode) was imple-mented for selected samples. The MSE collision energy wasramped between 10 and 35 V during each individual scan of0.08 s with an interscan delay of 0.02 s. In some samples, MS/MS fragmentation was performed for further confirmation ofmarker identity. The mass resolution for precursor and productions scans was 8300.

Data Processing, Chemometrics and Statistical Analysis

The LC−MS data were extracted using the MarkerLynxsoftware (Waters). Peaks were collected by automaticallycalculating the peak width and baseline noise. The intensitythreshold for the collected peaks was set to 20, with masstolerance of 0.05 Da in 0.05 min retention time window.Isotopic peaks were excluded. The data set was evaluated forthe presence of zero values to distinguish the true absent valuesfrom the erroneous zeros caused by the software.65,66 Thecomplete data set was first observed by ranking metabolitesbased on their abundance in each of the different sample types,and hierarchical clustering was performed. The heat maps weremade with R 2.7.2 using heatmap.2-function from the gplots-package. The data for the heat maps were hierarchicallyclustered both by markers (rows) and samples (columns).Next, the data was observed by ranking the whole data setacross all the sample types based on the p-values from Student’stwo-tailed t-test to each sample type (Microsoft Excel). The2000 most differential markers throughout the entire data setwere subjected to the partial least-squares discriminant analysis(PLS-DA) using the software SIMCA-P+ 12. All the data waslogarithmically transformed using ten as base and pareto-scaled.The PLS-DA models were validated using SIMCA-P+’s internal7-fold cross-validation. Lastly, the data set was reduced tocontain only such metabolites that had the p-value <0.05, foldchange >50%, and intensity value for the metabolite marker>10. The quality threshold (QT) cluster analysis wasperformed for this data set by the open-source software Multiexperiment Viewer (http://www.tm4.org/http://www.tm4.org/).

Metabolite Identification

The metabolite signals showing altered diet-dependentexpression profiles were identified by retention time and massspectra using our in-house database of authentic standards. Theelemental composition of the molecular markers was calculatedin the MassLynx software, and compared against commondatabases such as the Human Metabolome Database (http://www.hmdb.ca), Metlin (http://metlin.scripps.edu/), SciFinderScholar (SciFinder ScholarTM 2007), and ChemSpider(http://www.chemspider.com/). The MS/MS fragmentationspectra of the examined compounds were compared withcandidate molecules found in databases, and verified withcommercial standard compounds when available. The detailedlist of assigned metabolic features is available as SupportingInformation.

Journal of Proteome Research Article

dx.doi.org/10.1021/pr400257d | J. Proteome Res. 2013, 12, 3980−39923989

■ ASSOCIATED CONTENT*S Supporting Information

Assigned metabolite list. This material is available free of chargevia the Internet at http://pubs.acs.org.

■ AUTHOR INFORMATIONCorresponding Author

*E-mail: [email protected]. Tel: +358-40-3552364. Fax:+358-17 162 131.

Notes

The authors declare no competing financial interest.

■ ACKNOWLEDGMENTSThe funding from the Nordforsk Nordic Centre of Excellenceprojects “HELGA − whole grains and health” and “SYSDIET −Systems biology in controlled dietary interventions and cohortstudies” is gratefully acknowledged, as well as funding fromAcademy of Finland. A portion of the study was funded byUSDA/ARS Project Plan 1235-51530-053-00D. Mention oftrade names or commercial products in this publication is solelyfor the purpose of providing specific information and does notimply recommendation or endorsement by the USDA; theUSDA is an equal opportunity provider and employer.

■ REFERENCES(1) Patti, G. J.; Yanes, O.; Siuzdak, G. Innovation: Metabolomics: theapogee of the omics trilogy. Nat. Rev. Mol. Cell Biol. 2012, 4, 263−269.(2) Manach, C.; Hubert, J.; Llorach, R.; Scalbert, A. The complexlinks between dietary phytochemicals and human health deciphered bymetabolomics. Mol. Nutr. Food Res. 2009, 10, 1303−1315.(3) Despres, J. P.; Lemieux, I. Abdominal obesity and metabolicsyndrome. Nature 2006, 7121, 881−887.(4) Alberti, K. G.; Eckel, R. H.; Grundy, S. M.; Zimmet, P. Z.;Cleeman, J. I.; Donato, K. A.; Fruchart, J. C.; James, W. P.; Loria, C.M.; Smith, S. C., Jr Harmonizing the metabolic syndrome: a jointinterim statement of the International Diabetes Federation Task Forceon Epidemiology and Prevention; National Heart, Lung, and BloodInstitute; American Heart Association; World Heart Federation;International Atherosclerosis Society; and International Association forthe Study of Obesity. International Diabetes Federation Task Force onEpidemiology and Prevention; Hational Heart, Lung, and BloodInstitute; American Heart Association; World Heart Federation;International Atherosclerosis Society; International Association for theStudy of Obesity. Circulation 2009, 16, 1640−1645.(5) Browning, J. D.; Horton, J. D. Molecular mediators of hepaticsteatosis and liver injury. J. Clin. Invest. 2004, 2, 147−152.(6) Kotronen, A.; Yki-Jarvinen, H. Fatty liver: a novel component ofthe metabolic syndrome. Arterioscler. Thromb. Vasc. Biol. 2008, 1, 27−38.(7) Adiels, M.; Taskinen, M. R.; Packard, C.; Caslake, M. J.; Soro-Paavonen, A.; Westerbacka, J.; Vehkavaara, S.; Hakkinen, A.; Olofsson,S. O.; Yki-Jarvinen, H.; Boren, J. Overproduction of large VLDLparticles is driven by increased liver fat content in man. Diabetologia2006, 4, 755−765.(8) Kolb, H.; Mandrup-Poulsen, T. The global diabetes epidemic as aconsequence of lifestyle-induced low-grade inflammation. Diabetologia2010, 1, 10−20.(9) Wang, T. J.; Larson, M. G.; Vasan, R. S.; Cheng, S.; Rhee, E. P.;McCabe, E.; Lewis, G. D.; Fox, C. S.; Jacques, P. F.; Fernandez, C.;O’Donnell, C. J.; Carr, S. A.; Mootha, V. K.; Florez, J. C.; Souza, A.;Melander, O.; Clish, C. B.; Gerszten, R. E. Nat. Med. 2011, 4, 448−453.(10) Xie, B.; Waters, M. J.; Schirra, H. J. Metabolite profiles and therisk of developing diabetes. J. Biomed. Biotechnol. 2012, 805683.

(11) Bain, J. R.; Stevens, R. D.; Wenner, B. R.; Ilkayeva, O.; Muoio,D. M.; Newgard, C. B. Metabolomics applied to diabetes research:moving from information to knowledge. Diabetes 2009, 11, 2429−2443.(12) Griffin, J. L.; Nicholls, A. W. Metabolomics as a functionalgenomic tool for understanding lipid dysfunction in diabetes, obesityand related disorders. Pharmacogenomics 2006, 7, 1095−1107.(13) Gall, W. E.; Beebe, K.; Lawton, K. A.; Adam, K. P.; Mitchell, M.W.; Nakhle, P. J.; Ryals, J. A.; Milburn, M. V.; Nannipieri, M.;Camastra, S.; Natali, A.; Ferrannini, E. RISC Study Group. Alpha-Hydroxybutyrate is an early biomarker of insulin resistance andglucose intolerance in a nondiabetic population. PLoS One 2010, 5,e10883.(14) Adams, S. H.; Hoppel, C. L.; Lok, K. H.; Zhao, L.; Wong, S. W.;Minkler, P. E.; Hwang, D. H.; Newman, J. W.; Garvey, W. T. Plasmaacylcarnitine profiles suggest incomplete long-chain fatty acid beta-oxidation and altered tricarboxylic acid cycle activity in type 2 diabeticAfrican-American women. J. Nutr. 2009, 6, 1073−1081.(15) Lanza, I. R.; Zhang, S.; Ward, L. E.; Karakelides, H.; Raftery, D.;Nair, K. S. Quantitative metabolomics by H-NMR and LC-MS/MSconfirms altered metabolic pathways in diabetes. PLoS One 2010, 5,e10538.(16) Fiehn, O.; Garvey, W. T.; Newman, J. W.; Lok, K. H.; Hoppel,C. L.; Adams, S. H. Plasma metabolomic profiles reflective of glucosehomeostasis in non-diabetic and type 2 diabetic obese African-American women. PLoS One 2010, 12, e15234.(17) Wang, C.; Kong, H.; Guan, Y.; Yang, J.; Gu, J.; Yang, S.; Xu, G.Plasma phospholipid metabolic profiling and biomarkers of type 2diabetes mellitus based on high-performance liquid chromatography/electrospray mass spectrometry and multivariate statistical analysis.Anal. Chem. 2005, 13, 4108−4116.(18) Wang, C.; Feng, R.; Sun, D.; Li, Y.; Bi, X.; Sun, C. Metabolicprofiling of urine in young obese men using ultra performance liquidchromatography and Q-TOF mass spectrometry (UPLC/Q-TOFMS). J. Chromatogr., B 2011, 27, 2871−2876.(19) Zhao, X.; Fritsche, J.; Wang, J.; Chen, J.; Rittig, K.; Schmitt-Kopplin, P.; Fritsche, A.; Haring, H. U.; Schleicher, E. D.; Xu, G.;Lehmann, R. Metabonomic fingerprints of fasting plasma and spoturine reveal human pre-diabetic metabolic traits. Metabolomics 2010, 3,362−374.(20) Pietilainen, K. H.; Rog, T.; Seppanen-Laakso, T.; Virtue, S.;Gopalacharyulu, P.; Tang, J.; Rodriguez-Cuenca, S.; Maciejewski, A.;Naukkarinen, J.; Ruskeepaa, A. L.; Niemela, P. S.; Yetukuri, L.; Tan, C.Y.; Velagapudi, V.; Castillo, S.; Nygren, H.; Hyotylainen, T.; Rissanen,A.; Kaprio, J.; Yki-Jarvinen, H.; Vattulainen, I.; Vidal-Puig, A.; Oresic,M. Association of lipidome remodeling in the adipocyte membranewith acquired obesity in humans. PLoS Biol. 2011, 6, e1000623.(21) Puri, P.; Baillie, R. A.; Wiest, M. M.; Mirshahi, F.; Choudhury, J.;Cheung, O.; Sargeant, C.; Contos, M. J.; Sanyal, A. J. A lipidomicanalysis of nonalcoholic fatty liver disease. Hepatology 2007, 4, 1081−1090.(22) Garcia-Canaveras, J. C.; Donato, M. T.; Castell, J. V.; Lahoz, A.A comprehensive untargeted metabonomic analysis of human steatoticliver tissue by RP and HILIC chromatography coupled to massspectrometry reveals important metabolic alterations. J. Proteome Res.2011, 10, 4825−4834.(23) Kim, H. J.; Kim, J. H.; Noh, S.; Hur, H. J.; Sung, M. J.; Hwang, J.T.; Park, J. H.; Yang, H. J.; Kim, M. S.; Kwon, D. Y.; Yoon, S. H.Metabolomic analysis of livers and serum from high-fat diet inducedobese mice. J. Proteome Res. 2011, 2, 722−731.(24) Sampey, B. P.; Freemerman, A. J.; Zhang, J.; Kuan, P. F.;Galanko, J. A.; O’Connell, T. M.; Ilkayeva, O. R.; Muehlbauer, M. J.;Stevens, R. D.; Newgard, C. B.; Brauer, H. A.; Troester, M. A.;Makowski, L. Metabolomic profiling reveals mitochondrial-derivedlipid biomarkers that drive obesity-associated inflammation. PLoS One2012, 6, e38812.(25) Fardet, A.; Llorach, R.; Martin, J. F.; Besson, C.; Lyan, B.; Pujos-Guillot, E.; Scalbert, A. A liquid chromatography-quadrupole time-of-flight (LC-QTOF)-based metabolomic approach reveals new meta-

Journal of Proteome Research Article

dx.doi.org/10.1021/pr400257d | J. Proteome Res. 2013, 12, 3980−39923990

bolic effects of catechin in rats fed high-fat diets. J. Proteome Res. 2008,6, 2388−2398.(26) He, Q.; Ren, P.; Kong, X.; Wu, Y.; Wu, G.; Li, P.; Hao, F.; Tang,H.; Blachier, F.; Yin, Y. Comparison of serum metabolite compositionsbetween obese and lean growing pigs using an NMR-basedmetabonomic approach. J. Nutr. Biochem. 2012, 2, 133−139.(27) Baker, D. H. Animal models in nutrition research. J. Nutr. 2008,2, 391−396.(28) Sturek, M.; Alloosh, M.; Wenzel, J.; Byrd, J. P.; Edwards, J. M.;Lloyd, P. G.; Tune, J. D.; March, K. L.; Miller, M. A.; Mokelke, E. A.;Brisbin, I. L. Ossabaw Island Miniature Swine: CardiometabolicSyndrome Assessment. In Swine in the Laboratory: Surgery, Anesthesia,Imaging, and Experimental Techniques; Swindle, M. M., Ed.; CRCPress: Boca Raton, FL, 2007; pp 397−402.(29) Spurlock, M. E.; Gabler, N. K. The development of porcinemodels of obesity and the metabolic syndrome. J. Nutr. 2008, 2, 397−402.(30) Spence, L. A.; Weaver, C. M. Calcium intake, vascularcalcification, and vascular disease. Nutr. Rev. 2013, 1, 15−22.(31) Lee, L.; Alloosh, M.; Saxena, R.; Van Alstine, W.; Watkins, B. A.;Klaunig, J. E.; Sturek, M.; Chalasani, N. Nutritional model ofsteatohepatitis and metabolic syndrome in the Ossabaw miniatureswine. Hepatology 2009, 1, 56−67.(32) Bell, L. N.; Lee, L.; Saxena, R.; Bemis, K. G.; Wang, M.;Theodorakis, J. L.; Vuppalanchi, R.; Alloosh, M.; Sturek, M.;Chalasani, N. Serum proteomic analysis of diet-induced steatohepatitisand metabolic syndrome in the Ossabaw miniature swine. Am. J.Physiol. Gastrointest. Liver Physiol. 2010, 5, G746−754.(33) Rødgaard, T.; Stagsted, J.; Christoffersen, B. Ø.; Cirera, S.;Moesgaard, S. G.; Sturek, M.; Alloosh, M.; Heegaard, P. M. H.Orosomucoid expression profiles in liver, adipose tissues and serum oflean and obese domestic pigs, Gottingen minipigs and Ossabawminipigs. Vet. Immunol. Immunopathol. 2013, 3−4, 325−330.(34) Krupp, D.; Doberstein, N.; Shi, L.; Remer, T. Hippuric acid in24-h urine collections is a potential biomarker for fruit and vegetableconsumption in healthy children and adolescents. J. Nutr. 2012, 7,1314−1320.(35) Lees, H. J.; Swann, J. R.; Wilson, I. D.; Nicholson, J. K.; Holmes,E. Hippurate: The Natural History of a Mammalian-MicrobialCometabolite. J. Proteome Res. 2013, 12 (4), 1527−1546.(36) Aura, A. Microbial metabolism of dietary phenolic compoundsin the colon. Phytochem. Rev. 2008, 3, 407−429.(37) Spencer, J. P.; Abd El Mohsen, M. M.; Minihane, A. M.;Mathers, J. C. Biomarkers of the intake of dietary polyphenols:strengths, limitations and application in nutrition research. Br. J. Nutr.2008, 1, 12−22.(38) Shearer, J.; Duggan, G.; Weljie, A.; Hittel, D. S.; Wasserman, D.H.; Vogel, H. J. Metabolomic profiling of dietary-induced insulinresistance in the high fat-fed C57BL/6J mouse. Diabetes Obes. Metab.2008, 10, 950−958.(39) Calvani, R.; Miccheli, A.; Capuani, G.; Tomassini Miccheli, A.;Puccetti, C.; Delfini, M.; Iaconelli, A.; Nanni, G.; Mingrone, G. Gutmicrobiome-derived metabolites characterize a peculiar obese urinarymetabotype. Int. J. Obes. (London) 2010, 6, 1095−1098.(40) Williams, H. R.; Cox, I. J.; Walker, D. G.; Cobbold, J. F.; Taylor-Robinson, S. D.; Marshall, S. E.; Orchard, T. R. Differences in gutmicrobial metabolism are responsible for reduced hippurate synthesisin Crohn’s disease. BMC Gastroenterol. 2010, 108.(41) Backhed, F.; Manchester, J. K.; Semenkovich, C. F.; Gordon, J. I.Mechanisms underlying the resistance to diet-induced obesity in germ-free mice. Proc. Natl. Acad. Sci. U.S.A. 2007, 3, 979−984.(42) Turnbaugh, P. J.; Backhed, F.; Fulton, L.; Gordon, J. I. Diet-induced obesity is linked to marked but reversible alterations in themouse distal gut microbiome. Cell Host Microbe 2008, 4, 213−223.(43) Lever, M.; Slow, S. The clinical significance of betaine, anosmolyte with a key role in methyl group metabolism. Clin. Biochem.2010, 9, 732−744.

(44) Serkova, N. J.; Jackman, M.; Brown, J. L.; Liu, T.; Hirose, R.;Roberts, J. P.; Maher, J. J.; Niemann, C. U. Metabolic profiling of liversand blood from obese Zucker rats. J. Hepatol. 2006, 5, 956−962.(45) Kim, H. J.; Kim, J. H.; Noh, S.; Hur, H. J.; Sung, M. J.; Hwang, J.T.; Park, J. H.; Yang, H. J.; Kim, M. S.; Kwon, D. Y.; Yoon, S. H.Metabolomic analysis of livers and serum from high-fat diet inducedobese mice. J. Proteome Res. 2011, 2, 722−731.(46) Kathirvel, E.; Morgan, K.; Nandgiri, G.; Sandoval, B. C.; Caudill,M. A.; Bottiglieri, T.; French, S. W.; Morgan, T. R. Betaine improvesnonalcoholic fatty liver and associated hepatic insulin resistance: apotential mechanism for hepatoprotection by betaine. Am. J. Physiol.Gastrointest. Liver Physiol. 2010, 5, G1068−1077.(47) Olthof, M. R.; Verhoef, P. Effects of betaine intake on plasmahomocysteine concentrations and consequences for health. Curr. DrugMetab. 2005, 1, 15−22.(48) Kim, J. Y.; Park, J. Y.; Kim, O. Y.; Ham, B. M.; Kim, H. J.; Kwon,D. Y.; Jang, Y.; Lee, J. H. Metabolic profiling of plasma in overweight/obese and lean men using ultra performance liquid chromatographyand Q-TOF mass spectrometry (UPLC-Q-TOF MS). J. Proteome Res.2010, 9, 4368−4375.(49) Pietilainen, K. H.; Sysi-Aho, M.; Rissanen, A.; Seppanen-Laakso,T.; Yki-Jarvinen, H.; Kaprio, J.; Oresic, M. Acquired obesity isassociated with changes in the serum lipidomic profile independent ofgenetic effects–a monozygotic twin study. PLoS One 2007, 2, e218.(50) Loftus, N.; Miseki, K.; Iida, J.; Gika, H. G.; Theodoridis, G.;Wilson, I. D. Profiling and biomarker identification in plasma fromdifferent Zucker rat strains via high mass accuracy multistage massspectrometric analysis using liquid chromatography/mass spectrome-try with a quadrupole ion trap-time of flight mass spectrometer. RapidCommun. Mass Spectrom. 2008, 16, 2547−2554.(51) Meikle, P. J.; Christopher, M. J. Lipidomics is providing newinsight into the metabolic syndrome and its sequelae. Curr. Opin.Lipidol. 2011, 3, 210−215.(52) Quehenberger, O.; Armando, A. M.; Brown, A. H.; Milne, S. B.;Myers, D. S.; Merrill, A. H.; Bandyopadhyay, S.; Jones, K. N.; Kelly, S.;Shaner, R. L.; Sullards, C. M.; Wang, E.; Murphy, R. C.; Barkley, R.M.; Leiker, T. J.; Raetz, C. R.; Guan, Z.; Laird, G. M.; Six, D. A.;Russell, D. W.; McDonald, J. G.; Subramaniam, S.; Fahy, E.; Dennis, E.A. Lipidomics reveals a remarkable diversity of lipids in human plasma.J. Lipid Res. 2010, 11, 3299−3305.(53) Nygren, H.; Seppanen-Laakso, T.; Castillo, S.; Hyotylainen, T.;Oresic, M. Liquid chromatography-mass spectrometry (LC-MS)-basedlipidomics for studies of body fluids and tissues. Methods Mol. Biol.2011, 247−257.(54) Midtvedt, T. Microbial bile acid transformation. Am. J. Clin.Nutr. 1974, 11, 1341−1347.(55) Swann, J. R.; Want, E. J.; Geier, F. M.; Spagou, K.; Wilson, I. D.;Sidaway, J. E.; Nicholson, J. K.; Holmes, E. Systemic gut microbialmodulation of bile acid metabolism in host tissue compartments. Proc.Natl. Acad. Sci. 2011, Supplement 1, 4523−4530.(56) Sayin, S. I.; Wahlstrom, A.; Felin, J.; Jantti, S.; Marschall, H. U.;Bamberg, K.; Angelin, B.; Hyotylainen, T.; Oresic, M.; Backhed, F. Gutmicrobiota regulates bile acid metabolism by reducing the levels oftauro-beta-muricholic acid, a naturally occurring FXR antagonist. Cell.Metab. 2013, 2, 225−235.(57) Yokota, A.; Fukiya, S.; Islam, K. B. M. S.; Ooka, T.; Ogura, Y.;Hayashi, T.; Hagio, M.; Ishizuka, S. Is bile acid a determinant of thegut microbiota on a high-fat diet? Gut Microbes 2012, 5, 455−459.(58) Patti, M. E.; Houten, S. M.; Bianco, A. C.; Bernier, R.; Larsen, P.R.; Holst, J. J.; Badman, M. K.; Maratos-Flier, E.; Mun, E. C.;Pihlajamaki, J.; Auwerx, J.; Goldfine, A. B. Serum bile acids are higherin humans with prior gastric bypass: potential contribution toimproved glucose and lipid metabolism. Obesity (Silver Spring) 2009,9, 1671−1677.(59) Simonen, M.; Dali-Youcef, N.; Kaminska, D.; Venesmaa, S.;Kakela, P.; Paakkonen, M.; Hallikainen, M.; Kolehmainen, M.;Uusitupa, M.; Moilanen, L.; Laakso, M.; Gylling, H.; Patti, M. E.;Auwerx, J.; Pihlajamaki, J. Conjugated bile acids associate with altered

Journal of Proteome Research Article

dx.doi.org/10.1021/pr400257d | J. Proteome Res. 2013, 12, 3980−39923991

rates of glucose and lipid oxidation after Roux-en-Y gastric bypass.Obes. Surg. 2012, 9, 1473−1480.(60) Suzuki, Y.; Kaneko, R.; Nomura, M.; Naito, H.; Kitamori, K.;Nakajima, T.; Ogawa, T.; Hattori, H.; Seno, H.; Ishii, A. Simple andrapid quantitation of 21 bile acids in rat serum and liver by UPLC-MS-MS: effect of high fat diet on glycine conjugates of rat bile acids.Nagoya J. Med. Sci. 2013, 1−2, 57−71.(61) Xie, G.; Zhong, W.; Li, H.; Li, Q.; Qiu, Y.; Zheng, X.; Chen, H.;Zhao, X.; Zhang, S.; Zhou, Z.; Zeisel, S. H.; Jia, W. Alteration of bileacid metabolism in the rat induced by chronic ethanol consumption.FASEB J. 2013, DOI: 10.1096/fj.13-231860.(62) Devkota, S.; Wang, Y.; Musch, M. W.; Leone, V.; Fehlner-Peach,H.; Nadimpalli, A.; Antonopoulos, D. A.; Jabri, B.; Chang, E. B.Dietary-fat-induced taurocholic acid promotes pathobiont expansionand colitis in Il10−/− mice. Nature 2012, 7405, 104−108.(63) Barri, T.; Holmer-Jensen, J.; Hermansen, K.; Dragsted, L. O.Metabolic fingerprinting of high-fat plasma samples processed bycentrifugation- and filtration-based protein precipitation delineatessignificant differences in metabolite information coverage. Anal. Chim.Acta 2012, 47−57.(64) Masson, P.; Alves, A. C.; Ebbels, T. M.; Nicholson, J. K.; Want,E. J. Optimization and evaluation of metabolite extraction protocolsfor untargeted metabolic profiling of liver samples by UPLC-MS. Anal.Chem. 2010, 18, 7779−7786.(65) Malitsky, S.; Blum, E.; Less, H.; Venger, I.; Elbaz, M.; Morin, S.;Eshed, Y.; Aharoni, A. The transcript and metabolite networks affectedby the two clades of Arabidopsis glucosinolate biosynthesis regulators.Plant Physiol. 2008, 4, 2021−2049.(66) Hanhineva, K.; Aura, A. M.; Rogachev, I.; Matero, S.; Skov, T.;Aharoni, A.; Poutanen, K.; Mykkanen, H. In vitro microbioticfermentation causes an extensive metabolite turnover of rye branphytochemicals. PLoS One 2012, 6, e39322.

Journal of Proteome Research Article

dx.doi.org/10.1021/pr400257d | J. Proteome Res. 2013, 12, 3980−39923992