Variation in Antibiotic-Induced Microbial ... - ACS Publications

Variation in Antibiotic-Induced Microbial Recolonization Impacts on the Host Metabolic Phenotypes of Rats. Jonathan R. Swann*†‡ ... Department of ...
0 downloads 7 Views 1MB Size
ARTICLE pubs.acs.org/jpr

Variation in Antibiotic-Induced Microbial Recolonization Impacts on the Host Metabolic Phenotypes of Rats Jonathan R. Swann,*,†,‡ Kieran M. Tuohy,‡,§ Peter Lindfors,‡ Duncan T. Brown,‡ Glenn R. Gibson,‡ Ian D. Wilson,|| James Sidaway,^ Jeremy K. Nicholson,† and Elaine Holmes*,† †

)

Biomolecular Medicine, Department of Surgery and Cancer, Faculty of Medicine, Imperial College London, South Kensington, London SW7 2AZ, United Kingdom ‡ Department of Food and Nutritional Sciences, School of Chemistry, Food and Pharmacy, University of Reading, Whiteknights, Reading, RG6 6AP, United Kingdom Department of Drug Metabolism and Pharmacokinetics IM, AstraZeneca UK Ltd, Alderley Park, Macclesfield, Cheshire SK10 4TG, United Kingdom ^ Global Safety Assessment, AstraZeneca UK Ltd, Alderley Park, Macclesfield, Cheshire, SK10 4TG, United Kingdom § IASMA Research and Innovation Centre, Fondazione Edmund Mach, Food Quality and Nutrition Area, Via E. Mach, 1, 38010 S. Michele all'Adige, Italy

bS Supporting Information ABSTRACT:

The interaction between the gut microbiota and their mammalian host is known to have far-reaching consequences with respect to metabolism and health. We investigated the effects of eight days of oral antibiotic exposure (penicillin and streptomycin sulfate) on gut microbial composition and host metabolic phenotype in male Han-Wistar rats (n = 6) compared to matched controls. Early recolonization was assessed in a third group exposed to antibiotics for four days followed by four days recovery (n = 6). Fluorescence in situ hybridization analysis of the intestinal contents collected at eight days showed a significant reduction in all bacterial groups measured (control, 1010.7 cells/g feces; antibiotic-treated, 108.4). Bacterial suppression reduced the excretion of mammalianmicrobial urinary cometabolites including hippurate, phenylpropionic acid, phenylacetylglycine and indoxyl-sulfate whereas taurine, glycine, citrate, 2-oxoglutarate, and fumarate excretion was elevated. While total bacterial counts remained notably lower in the recolonized animals (109.1 cells/g faeces) compared to the controls, two cage-dependent subgroups emerged with Lactobacillus/ Enterococcus probe counts dominant in one subgroup. This dichotomous profile manifested in the metabolic phenotypes with subgroup differences in tricarboxylic acid cycle metabolites and indoxyl-sulfate excretion. Fecal short chain fatty acids were diminished in all treated animals. Antibiotic treatment induced a profound effect on the microbiome structure, which was reflected in the metabotype. Moreover, the recolonization process was sensitive to the microenvironment, which may impact on understanding downstream consequences of antibiotic consumption in human populations. KEYWORDS: antibiotics, gut microbiota, recolonizing, bacteroidetes, firmicutes, metabolic profiling

’ INTRODUCTION The gut microbiota are now recognized as a significant factor in determining host health and with respect to conferring an r 2011 American Chemical Society

Received: March 16, 2011 Published: May 17, 2011 3590

dx.doi.org/10.1021/pr200243t | J. Proteome Res. 2011, 10, 3590–3603

Journal of Proteome Research extended metabolic capacity on the host. Substantial interindividual variability occurs within the human gut microbiota and numerous host factors exert selective pressure upon the microbiome including host genome,1,2 diet,3 age,4 and drug therapy.5,6 In turn, microbial variation is known to influence host metabolic phenotype7 and the metabolic activity of the gut microbiota has been found to variously potentiate or ameliorate the toxicity of xenobiotics810 and to impact upon several diseases. The association of certain microbial compositions with host insulin resistance11 and obesity12 has led to a renewal in interest in the impact of the gut microbes on host health. Although host and microbial metabolism are known to be inextricably linked, little work has been conducted on elucidating specific molecular associations between microbes and their mammalian host. Exploration of direct mammalianmicrobial interactions promises to achieve a better understanding of the underpinning role of the gut microbiota in mammalian health. The contribution of the microbiota toward interanimal variation in disease etiology, drug metabolism and toxicity can be studied using models of reduced or simplified microbiota. Germ-free (GF) animals, with or without subsequent implantation of a reduced microbiota, have been widely used to investigate the relationship between host and microbial metabolism. These animals represent “knockout” models with lifelong gut microbial absence allowing covariation between phenotypic features or metabolic function and microbial composition to be studied. However, these “knockout” animals lack morphological and physiological characteristics associated with supporting an intestinal microbiota. GF animals have reduced body weight due to reduced calorific harvest from the diet, underdeveloped gut structure and absorptive capacity, restricted maturation of the immune system, lower metabolic rate, smaller lungs, livers, and hearts and therefore reduced cardiac output. Arguably, this does not produce a meaningful comparison with “conventional” animals and therefore a more appropriate model may be the antibiotic-treated (AB) animal, since it is more reflective of the morphology and physiology of a conventional (CV) animal while possessing a reduced microbiota. Suppression of the microbiota through the oral administration of antibiotics was first described by Van der Waaij and Sturm.13 Although the exact microbial composition of AB animals cannot be controlled, different antibiotics can be used to selectively target different groups of microbiota. Historically it has been difficult to evaluate the extent of microbial perturbation following antibiotic treatment because the majority (5090%) of the microbiota could not be cultured. In this study fluorescence in situ hybridization (FISH) analysis, a culture-independent, quantitative molecular-based technique, has been applied to assess the microbiota, allowing enumeration of several key bacterial groups. We combine this with high resolution 1H nuclear magnetic resonance (NMR) spectroscopic based profiling of urine and fecal water to provide an alternative surrogate means of ascertaining the influence of the gut microbiota on host metabolism. High resolution spectroscopy, in particular 1H NMR spectroscopy has been used to metabolically phenotype both GF14,15 and AB16 models of modulated microbiota and to provide a window into gut microbial composition and activity, where aromatic compounds such as hydroxyphenylpropionic acid (HPPA) and hippurate, produced either directly by the microbiota or through the concerted trans-genomic metabolism of the microbiome and host genome, can be profiled over time. Antibiotics such as vancomycin,16 enrofloxacin17 and antibiotic cocktails18 have been

ARTICLE

studied using this approach and were associated with modulation of multiple urinary and fecal metabolites such as amines and microbial metabolites of aromatic amino acids, which normalized after approximately three weeks (Supplementary Table S1, Supporting Information). A variety of antibiotics, including bacitracin, kanamycin, neomycin, streptomycin and penicillin, have been utilized either alone or in combination to effectively decontaminate the digestive tract. In the present study penicillin and streptomycin sulfate were selected for attenuation of the gut microbiota as they have complementary bacteriocidal qualities. Penicillin G, a β-lactam antibiotic, is most effective against Gram-positive bacteria and is incompletely absorbed by the GIT (bioavailability about 1530% on an empty stomach).19 Streptomycin is effective against Gramnegative bacteria and is poorly absorbed from the GIT and therefore should not enter systemic circulation preventing interference with host biochemistry. Here, we have used a parallel microbiological and metabonomic approach to characterize and integrate the microbial and metabolic profile of rats following both temporal microbiotal perturbation and also recolonization via the administration of antibiotics.

’ EXPERIMENTAL SECTION Study Design

Male, Wistar derived AlpkHsdRccHan:WIST rats (n = 18) were obtained from the Rodent Breeding Unit at AstraZeneca, Alderley Park. Animals were approximately 10 weeks old and housed 3 per cage. Food and water were available ad libitum and all animals were subjected to 12-h light, 12-h dark artificial light cycle. Animals were acclimatized to these housing conditions prior to the start of the dosing regime. The rats were separated into 3 groups of 6 and antibiotic treatment was carried out following the protocol of Guo et al.20 The antibiotic solution, containing 4 mg/mL Streptomycin and 2 mg/mL Penicillin G, was provided in the animals’ drinking water. The drinking water of the control (CN) group contained no antibiotics. The bacterially suppressed (SP) group was given free access to the antibiotic-water for 8 days. Animals in the recolonizing (RC) group were given free access to the antibiotic-water for the first 4 days and were then switched to antibiotic-free water for the remaining 4 days of the study. All animals were euthanased by halothane inhalation on the 8 day of the study. Sample Collection

Urine sample collection was carried out overnight for 16 h on days 3 4 and days 78 using metabolism cages. Fecal samples were taken on day 8 immediately prior to the animals being removed from the metabolism cages. Intestinal contents for FISH analysis were taken at necropsy. Enumeration of Faecal Microbial Populations by Fluorescence in situ Hybridization (FISH)

The enumeration of fecal microbial populations was carried out using fluorescently labeled 16S rRNA targeted oligonucleotide probes and fluorescence in situ hybridization (FISH). Oligonucleotide probes (MWG-Biotech, Milton Keynes, UK) targeting the Atopobium group (50 -GGT CGG TCT CTC AAC CC),21 Bacteroides spp. (50 -CCA ATG TGG GGG ACC TT),22 Bifidobacterium spp. (50 -CATCCGGCATTACCACCC),23 Eubacterium rectale group (50 -GCTTCTTAGTCARGTACCG),24 Clostridium histolyticum group (50 -TTATGCGGTATTAATCYCCTTT) and the lactobacilli/enterococci (50 -GGTATTA3591

dx.doi.org/10.1021/pr200243t |J. Proteome Res. 2011, 10, 3590–3603

Journal of Proteome Research

ARTICLE

Figure 1. Microbial profiling analysis of the intestinal contents collected at day eight from all study animals. (A) Total and target bacterial populations in the intestinal contents at day eight. Values are mean þ standard error of the mean and expressed as Log10 cells/g faeces as determined by FISH. (B) PCA scores plot constructed from the FISH data comparing control, recolonizing and suppressed animals. Circled are the two subgroups that emerge in the recolonizing group (RCR and RCβ) (R2 = 0.97). (C) Microbiotal composition at the phylum level. (D) Impact of antibiotic-induced suppression and subsequent recolonization on the composition of the gut microbiota. Values represent the total probe count for the individual bacterial probes for control, recolonizing subgroups RCR and RCβ and suppressed animals.

GCAYCTGTTTCCA) were used for enumeration of dominant members of the gut microbiota. The probes were labeled at the 50 end with the fluorescent dye Cy3. For the enumeration of total cells, samples were stained with the nucleic acid stain 40 ,60 diamino-2-phenylindole (DAPI). Slides were enumerated using

a Nikon microscope with an EPI-fluorescence attachment (Nikon U.K., Kingston-upon-Thames, U.K.). Principal components analysis (PCA) was applied to the normalized FISH data to identify trends within the bacterial populations relating to antibioticinduced bacterial suppression and subsequent recolonization. 3592

dx.doi.org/10.1021/pr200243t |J. Proteome Res. 2011, 10, 3590–3603

Journal of Proteome Research 1

H Nuclear Magnetic Resonance (NMR) Spectroscopy

Urine samples were prepared by mixing 400 μL of urine with 200 μL of phosphate buffer at pH 7.4 containing 10% D2O and 1 mM 3-trimethylsilyl-1-[2,2,3,3-2H4] propionate (TSP) and then transferring the mixture to 5 mm outer diameter NMR tube. For the fecal samples, 200300 mg (one pellet) of stool was mixed with 1 mL of phosphate buffer at pH 7.4 containing 10% D2O, 3 mM sodium azide, and 1 mM TSP. The samples were sonicated for 30 min and then centrifuged at 14 000 g for 10 min to remove particulates before being transferred to a 5 mm outer diameter NMR tube. 1 H NMR spectra were acquired for each sample using a Bruker Avance 600 spectrometer operating at 600.13 MHz. A standard one-dimensional NMR spectrum was acquired for each sample with water peak suppression using a standard pulse sequence (recycle delay (RD)-90-t1-90-tm-90-acquire free induction decay (FID)). The RD was set at 2 s and the mixing time (tm) 100 ms. The 90 pulse length was approximately 12 μs and t1 was set to 3 μs. For each sample, 8 dummy scans were followed by 64 scans and collected in 64 K data points using a spectral width of 20 ppm and an acquisition time per scan of 2.73 s. Data Preprocessing and Analysis 1

H NMR spectra were manually corrected for phase and baseline distortions and then referenced to the TSP resonance at δ 0.0. Spectra were digitized using an in-house MATLAB (version R2009b, The Mathworks, Inc.; Natwick, MA) script. To prevent baseline effects that arise from imperfect water saturation the region containing the water resonance was excised. An in-house peak alignment algorithm25 was then performed on each spectrum in MATLAB to adjust for shifts in peak position due to small pH differences between samples and then each spectrum was normalized using a probabilistic quotient approach.26 For both the urine and fecal NMR spectra, principal components analysis (PCA) using unit variance scaling was applied in SIMCA (Umetrics, Umea). Orthogonal projection to latent structure discriminant analysis (OPLS-DA) models were constructed using unit variance scaling to aid the interpretation of the model and distinguish the metabolites that differed between the groups. Here, 1H NMR spectroscopic data were used as the descriptor matrix and class information (CN, SP, or RC) as the response variable. The contribution of each variable (metabolite) to sample classification was visualized by back-scaling transformation, generating a correlation coefficient plot. These coefficient plots are colored according to the significance of correlation to “class” (e.g., CN, SP, RC), with red indicating high significance and blue indicating low significance. The direction and degree of the signals relate to covariation of the metabolites with the classes in the model. For all models, one orthogonal component was used to remove systematic variation unrelated to class. Predictive ^ parameter. performance was assessed using the Q2Y Since a decrease in the fecal short chain fatty acids was characteristic of the antibiotic-treated group, propionate, butyrate and acetate were quantified. An OPLS model was constructed to elucidate correlations between these individual fecal metabolites and components of the urinary metabolic profile. As diminished fecal short chain fatty acids were strongly characteristic of antibiotic administration, the relationship with the urinary metabolite profile was tested independently in the control (n = 5) and treatment (n = 6) groups, in order to discriminate specific relationships between fecal short chain fatty acids and urinary metabolites from correlations arising as a

ARTICLE

global response to antibiotic treatment. For propionate, the integral from the triplet at δ 1.04 in the fecal spectrum, arising from methyl group on the propionate molecule, formed the response variable (Y matrix) and was regressed against the whole urinary NMR spectrum (X matrix). The integrals from the singlet at δ 1.92 and triplet at δ 0.90 corresponding to the methyl groups from acetate and butyrate respectively were correlated with the urinary metabolic profiles.

’ RESULTS Clinical Observations and Histopathology

There were no adverse clinical observations recorded throughout the study. Water consumption remained constant in the CN group, however there was a pronounced increase in water intake in the SP and RC groups on days 23 and 57 compared to the CN animals. FISH Analysis of the Bacterial Profiles of Intestinal Contents Collected at Day Eight Following Antibiotic Dosing

Following eight days of antibiotic administration the average total number of bacterial cells (determined by the nucleic acid stain 40 ,60 -diamino-2-phenylindole (DAPI)) contained within the intestinal contents was reduced from 10.7 log10 cells/g in CN animals to 8.4 log10 cells/g SP animals. Those animals treated with antibiotics for four days and then allowed to recolonize for four days (RC) still supported substantially less bacteria (9.1 log10 cells/g of faeces) than the CN animals (Figure 1A). Individual probe counts were all significantly lower (p < 0.001) in the SP group compared to the CN group with the greatest change witnessed in counts for members of the Gram negative Bacteriodetes phylum covered by the Mib724 probe (CN = 10.35 log10 cells/g; SP = 7.58 log10 cells/g). Similar observations were noted in the RC animals except for the Lab158 counts (members of Gram positive Firmicutes phylum) where large intragroup variation occurred. Principal components analysis (PCA) performed on the FISH count data identified trends in the bacterial profiles between study animals. From the PCA scores plot (Figure 1B) it can be seen that CN animals were clearly discriminated from those groups treated with antibiotics and this was explained by higher counts of all probes in the CN group. The SP group clustered tightly, indicating greater homogeneity within the SP bacterial profiles than in the CN or RC groups, and was positioned marginally further away from CN animals than the RC animals. In the RC animals, two subgroups emerged in the bacterial signatures (denoted RCR and RCβ). This subgrouping appeared to correspond to the cage, with the three animals of the RCR subgroup being housed together and those of RCβ housed together. RCR had significantly higher Lab158 counts (Lab158 = 9.3 log10cells/g) than RCβ (Lab158 = 7.8 log10 cells/g, p = 0.003, determined by a two-tailed, unpaired Student’s t test). The PCA scores plot indicated that the bacterial signature of RCβ bore greater similarity to those animals treated with antibiotics for eight days than RCR suggesting recolonization was more extensive in the RCR subgroup. This is corroborated by the significantly higher total average bacterial counts of RCR (9.5 log10 cells/g feces) compared to RCβ (8.7 log10 cells/g feces, p < 0.001) animals. Gut microbial composition can be seen for each animal in Figure 1D where the relative abundance of each individual probe count can be observed. Bacteria enumerated by the 3593

dx.doi.org/10.1021/pr200243t |J. Proteome Res. 2011, 10, 3590–3603

Journal of Proteome Research

ARTICLE

Figure 2. Principal components analysis (PCA) of the urinary metabolic profiles of all study animals at day eight. PCA scores plot comparing the urinary metabolic profiles at day eight from (A) control, recolonizing and suppressed animals (R2 = 0.50); (B) control and suppressed (R2 = 0.45); (C) control and recolonizing (R2 = 0.47); (D) recolonizing and suppressed (R2 = 0.44); and (E) the two subgroups of the recolonizing animals (RCR and RCβ). 3594

dx.doi.org/10.1021/pr200243t |J. Proteome Res. 2011, 10, 3590–3603

Journal of Proteome Research

ARTICLE

Table 1. Urinary Metabolite Differences Identified at Day Eight between Control (CN), Recolonizing (RC) and Suppressed (SP) Animals by OPLS-DAa metabolite

^ = 0.91b SP vs CN Q2Y

^ = 0.87b RC vs CN Q2Y

^ = 0.51b SP vs RCQ2Y

Microbially Derived 3-HPPAc

0.64

0.67

4-HPPA Hippuratec

0.86 0.80

0.64 0.80

 0.59 

Indoxyl sulfate

0.94

0.72

0.66

PAGc

0.81

0.82

þ0.71

Energy Metabolism 2-oxoglutarate

þ0.53

þ0.48



Citrate

þ0.54

þ0.53



Fumaratec

þ0.47

Glucose

þ0.65

þ0.67



D-3-hydroxybutyrate

þ0.47

þ0.47







Amino Acids Alanine

þ0.49

þ0.75



Taurine

þ0.75

þ0.62



Betaine

þ0.74

þ0.49

Choline

þ0.75

þ0.50



DMG

þ0.57

þ0.64



Glycinec

þ0.64

þ0.67



Creatinec

þ0.80

þ0.62



Creatininec

þ0.75

þ0.74



Formate

þ0.57





Acetatec

0.55





Lactate

þ0.48

N-acetyl-glycoproteinc

þ0.84

Choline Metabolism þ0.55

Other





þ0.73



Allantoin

þ0.82

þ0.62



TMAOc Phosphocholine

0.74 þ0.84

0.77 þ0.64

 

a Values are the correlation coefficient derived from the OPLS-DA model. b þ, relatively higher in the treatment group (SP or RC); , relatively lower in the treatment group with respect to the control group. c Observed after administration of vancomycin, enrofloxacin or an antibiotic cocktail in mice and rats.1618 3-HPPA, 3-hydroxyphenylpropionic acid; 4-HPPA, 4-hydroxyphenylpropionic acid; DMG, dimethylglycine; PAG, phenylacetylglycine; TMAO, trimethylamine-N-oxide.

Bac303 and Mib724 probes dominated the CN profile. Both of these probes measure bacteria belonging to the Bacteroidetes phylum. Probes specific to genera in the Firmicutes (Chis150, Erec482, Lab 158) and Actinobacteria (Bif164, Ato291) phyla were lower in abundance with both sets of probes measuring approximately 5-fold lower than those of the Bacteroidetes phylum. Microbiotal composition was modulated in the SP group with a marked reduction in the proportion of Mib724 in relation to the total count and a notable rise in the proportion of Erec482, Ato291 and Bac303. Bacteria measured by the Bif164 probe were also noted to decrease in abundance. Collectively, probes measuring genera belonging to the Bacteroidetes phylum were lower in the overall profile of SP animals compared to CN animals although this was driven by the pronounced reduction in Mib724 probe counts. In contrast, the proportion of probe counts enumerating genera of the Firmicutes and Actinobacteria phyla were observed to

increase following eight days of antibiotic treatment, measuring approximately 2-fold lower than the Bacteroidetes. The two subgroups of the RC rats emerged with markedly different bacterial profiles. RCR was dominated by the Firmicutes phylum with an apparent bloom in the Grampositive lactobacilli/enterococci (Lab158 probe). Diversity was narrow in these animals with Bif164 and Chis150 counts being low. RCβ was more similar in composition to the SP animals, with Bacteroidetes constituting the majority of the bacterial profile. As witnessed in the SP animals, the Bacteroidetes: Firmicutes ratio was altered in the RCβ group with the abundance of the measured Firmicutes increased and Bacteroidetes decreased compared to the CN. Lab158 and Erec482 counts were higher proportionally in RCβ compared to the CN group. RCβ also contained a higher ratio of Actinobacteria with Ato291 and Bif164 constituting a greater amount of the overall profile compared to CN animals. 3595

dx.doi.org/10.1021/pr200243t |J. Proteome Res. 2011, 10, 3590–3603

Journal of Proteome Research

ARTICLE

Figure 3. Orthogonal projection to latent structures-discriminant analysis (OPLS-DA) of urine collected from all study animals at day four. (A) Predictive scores plot from the OPLS-DA model comparing control animals (black) with animals treated with antibiotics (blue) at day four. (B) Coefficient plot extracted from the OPLS-DA model constructed from the urinary 1H NMR profiles at day four discriminating between the control and antibiotic-treated animals (recolonizing and suppressed). 3-HPPA, 3-hydroxyphenyl propionic acid; 4-HPPA, 4-hydroxyphenylpropionic acid; DMG, dimethylglycine; PAG, phenylacetylglycine; TMAO, trimethylamine-N-oxide.

Effect of Antibiotic Dosing upon Urine 1H NMR Spectroscopic Profiles (Day Eight)

At day eight, the urinary metabolic profiles of SP and RC animals were noticeably different in composition to CN animals as determined by PCA (Figure 2A). These differences were more pronounced in the SP animals. Pair-wise comparisons of each group showed clear differentiation in the urinary metabolic profiles between all groups (Figure 2BD). OPLS-DA coefficients plots were generated from the models of these pairwise comparisons to aid interpretation, indicating the variables (1H NMR chemical shifts) responsible for the separation witnessed in the scores plots (Table 1). As expected, the ^= model calculated from comparing CN and RC animals (Q2Y 0.87) was weaker than that comparing CN and SP animals ^ = 0.91). SP animals were found to excrete higher amounts (Q2Y of glucose, the ketone body D-3 hydroxybutyrate, and the TCA cycle intermediates 2-oxoglutarate, citrate and fumarate than CN animals. The amino acids, taurine, glycine, and alanine, in addition to allantoin, creatine, creatinine, choline, phosphocholine, betaine, dimethylglycine, lactate, formate and N-acetyl-glycoprotein fragments were also excreted in greater amounts in the SP animals. Bacterial suppression reduced the amount of microbially derived metabolites excreted including 3-hydroxyphenylpropionic acid (3-HPPA), 4-HPPA, hippurate, indoxyl sulfate and phenylacetylglycine (PAG); trimethylamine-N-oxide (TMAO) and acetate excretion was also lower in the SP animals at day eight. The urinary metabolic profiles of RC animals differed to those of CN animals (Figure 2C and Table 1) with higher concentrations of alanine, allantoin, creatine, creatinine, N-acetyl-glycoprotein and taurine observed. Metabolites involved in choline metabolism (choline, phosphocholine, betaine, dimethylglycine and glycine) were elevated in the urine of RC animals at day eight as were glucose, citrate, 2-oxoglutarate and D-3 hydroxybutyrate. After four

days of recolonization urinary concentrations of 3-HPPA, 4-HPPA, hippurate, PAG, indoxyl sulfate and TMAO remained lower than in controls. The subgroups observed in bacterial profiles also emerged in the metabolic profiles of the urine at day eight (Figure 2E). The RCR profile was closer in composition to the CN animals than RCβ, which correlated to the higher bacterial load in the RCR animals. The loadings plot from the PCA model revealed RCR animals excreted more 4-HPPA, PAG, indoxyl sulfate, acetate and less 2-oxoglutarate, D-3 hydroxybutyrate, alanine, citrate, creatinine, glycine, lactate, N-acetyl-glycoprotein, and N-methylnicotinamide when compared to the RCβ group. Differences were observed between the RC and SP urine at day eight although the separation between the two groups was less pronounced than comparisons to CN animals (Figure 2D). SP animals excreted greater amounts of betaine and surprisingly PAG. RC animals excreted higher amounts of 4-HPPA and indoxyl sulfate. The urine of subgroup RCβ was more similar in composition to the SP urine than RCR. 1

H NMR Spectra of the Urine Collected from All Study Animals (Day Four)

At day four, both SP and RC animals had received antibiotics in their drinking water for the same period of time and were therefore at this point maintained under identical experimental conditions. An OPLS-DA model was built where the CN animals constituted one class (n = 6) and the SP and RC animals were combined to form the second class (n = 12). The cross-validated ^ = 0.8) predictive scores plot extracted from the OPLS-DA (Q2Y model (Figure 3B) show the urinary metabolic profiles of control and treated animals were clearly discriminated at day four and that all treated animals shared a similar metabolic profile. The corresponding OPLS-DA coefficients plot is shown in Figure 3A, explaining the metabolites responsible for the separation 3596

dx.doi.org/10.1021/pr200243t |J. Proteome Res. 2011, 10, 3590–3603

Journal of Proteome Research

ARTICLE

Table 2. Metabolites Significantly Different in the Feces Following Penicillin/Streptomycin Administration Determined by OPLS-DAa metabolite

^ = 0.97)b RC vs CN (Q2Y ^ = 0.95)b SP vs CN (Q2Y Antibiotic-related metabolites

Penicillamine-related

þ0.91

þ0.81

Short chain fatty acids (SCFA) Acetatec

0.91

0.90

n-butyratec

0.71

0.71

Propionatec

0.92

0.92

Amino acids Alaninec

0.86

0.89

Taurine

þ0.84

þ0.75

Tryptophan

þ0.92

þ0.92

Asparagine

þ0.88

þ0.73

Methioninec

0.86

0.93

Oligosaccharidesc

þ0.99

þ0.99

Cholinec Succinate

þ0.87 0.83

þ0.84 0.85

Lactatec

0.90

0.92

Other

a

Values are the correlation coefficient derived from the OPLSDA model. b þ, relatively higher in the treatment group (SP or RC); , relatively lower in the treatment group with respect to the control group. c Observed after administration of vancomycin, enrofloxacin or an antibiotic cocktail in mice and rats.16,17

witnessed in Figure 3B. Antibiotic dosing was found to increase relative concentrations of the TCA cycle intermediates 2-oxoglutarate and citrate and, in addition, amounts of creatinine, taurine, and N-acetyl-glycoprotein fragments were also increased. Decreased concentrations of 3-HPPA, 4-HPPA, PAG, hippurate, trimethylamine-N-oxide, and indoxyl sulfate (Table 2) were observed in the AB group. Impact of Antibiotic Administration on the Metabolic Profiles of Fecal Extracts (Day Eight)

At day eight, the fecal metabolite profiles of SP and RC animals were markedly different from the CN group. This was confirmed by the construction of OPLS-DA models comparing ^ = 0.98) and CN with RC animals CN with SP animals (Q2Y ^ = 0.98) (Figure 4, metabolic differences listed in Table 2). (Q2Y Interestingly, unlike the urinary profiles, SP and RC fecal samples were found to be similar in metabolic composition, despite the four days of recolonization in the RC group (as illustrated by the poor discriminatory model strength between the two classes, ^ = 0.06). SP and RC animals were observed to excrete Q2 Y higher amounts of taurine, tryptophan, asparagine and choline, and noticeably higher amounts of oligosaccharides than the CN animals. The microbially synthesized short chain fatty acids (SCFA), namely acetate, butyrate and propionate were found to be reduced in the SP and RC rats along with alanine, lactate, methionine and succinate levels. Penicillamine, a major metabolite of penicillin, was measured in the feces of both SP and RC animals, with higher amounts observed in the SP animals. The composition of fecal SCFA was modulated by the antibiotic treatment in favor of butyrate and propionate (Supplementary Table S2, Supporting Information). The CN animals excreted

acetate/butyrate (1:0.21) and acetate/propionate (1:0.13) while the SP animals excreted acetate/butyrate (1:0.39) and acetate/ propionate (1:0.35). Similar alterations were observed in the RC animals (acetate/butyrate (1:0.35); acetate/propionate (1:035)). Correlation of Fecal SCFAs to Urinary Metabolites

To elucidate the relationship between microbial activity in the gastrointestinal tract and host endogenous metabolism, fecal levels of the SCFA, propionate, butyrate and acetate, were correlated to the urinary metabolic profiles of CN animals using OPLS (Supplementary Table S3, Supporting Information). Metabolic connections were observed between the SCFA and urinary components. All fecal SCFA positively correlated with urinary formate, creatine, and taurine and negatively correlated with 2-oxoglutarate, citrate, malonate, malate, N-acetyl-glycoproteins and inosine. The strongest model was obtained when fecal butyrate was regressed ^ = 0.90) with the model being against the urine profile (Q2Y additionally driven by positive correlations with acetate and negative ^= correlations with creatinine and glycine. Fecal propionate (Q2Y 0.76) correlated strongly with metabolites implicated in energy metabolism including negative correlations with fumarate. Hippurate, methylamine, phosphocholine and lactate were also negatively correlated with propionate while 4-HPPA was positively associated ^ = 0.77) correlated positively with (Figure 5). Fecal acetate (Q2Y urinary acetate and negatively with fumarate, PAG, methylamine, phosphocholine, creatinine, glycine and lactate.

’ DISCUSSION Penicillin and Streptomycin Shift Microbiotal Balance in Favor of Firmicutes, Modulate Energy Metabolism

After eight days of penicillin and streptomycin administration, the bacterial counts of all of the probes analyzed in the intestinal contents of the dosed rats were significantly reduced. Compositional makeup of the microbiota was modulated by antibiotic therapy, whereby the relative proportions of bacteria enumerated by the Erec482, Ato291, and Bac303 probes were increased in the overall profile and the proportion of those selected by the Mib724 probe were noticeably decreased. At the phylum level, a reduction in the relative proportion of the Bacteroidetes was observed following eight days of antibiotic treatment with a subsequent rise in the Firmicutes and Actinobacteria. Such changes are noteworthy due to increasing literature reports implicating such shifts in the etiology of obesity. Work by Ley et al.12,27 reported that the proportion of Bacteroidetes was lower in obese individuals in both mice and humans, while the proportion of Firmicutes was higher. Bacteroidetes abundance was independent of food intake but increased with weight loss on two types of low calorie diets. Enhanced capacity for energy harvest by the obese microbiota through the production of SCFA was proposed as the mechanism promoting weight gain.28 Schwiertz et al. confirmed the association between SCFA metabolism and obesity but reported a decrease in the Firmicutes to Bacteroidetes ratio in overweight and obese adult humans.29 A further study found weight loss diets in obese subjects to significantly modulate gut microbial fingerprints but found no connection between Bacteroidetes and Firmicutes in human obesity.30 These findings suggest that the connection between the gut microbiota and obesity may be more intricate in nature than a simple phylum-level shift would account for but nevertheless 3597

dx.doi.org/10.1021/pr200243t |J. Proteome Res. 2011, 10, 3590–3603

Journal of Proteome Research

ARTICLE

Figure 4. Orthogonal projection to latent structures-discriminant analysis (OPLS-DA) of feces collected from all study animals at day eight. OPLS-DA coefficient and predictive scores plots from the 1H NMR profiles of fecal water at day eight discriminating between control and suppressed (A and B: ^ = 0.98) and between control and recolonizing rats (C and D: Q2Y ^ = 0.98). Q2 Y

implicates energy harvest from the diet as a potential underlying mechanism. Antibiotic consumption in the 20th century has been proposed as a factor in increased human obesity31 with the gut microbiota suggested as a mechanism. Since the 1940s, antibiotics have been recognized to exert growth promoting effects by raising feedconversion efficiency32 and were utilized in livestock farming to improve weight gain until 2006, when subtherapeutic antibiotic use in agriculture was banned in the EU. Since antibiotic supplements need to be taken orally to be effective in weight gain, community modulation of the gut microbiota must underlie these growth alterations. Moreover, weight gain does not occur in

germ-free animals after oral administration of antibiotics.33 Modulation of the Firmicutes/Bacteroidetes balance observed in the current study supports the notion that antibiotics may elicit these growth effects through disruption of energy extraction metabolism due to either antibiotic resistance in the Firmicutes or antibiotic sensitivity in the Bacteroidetes. Specifically, the Mib724 population of Bacteroidetes appeared to be the most susceptible of the probes measured to the antibiotics. Consistent with altered intestinal ecology influencing energy extraction efficiency, the microbial fermentation products, SCFA and lactate, were notably reduced in the fecal water following antibiotic administration while the SCFA-precursors, 3598

dx.doi.org/10.1021/pr200243t |J. Proteome Res. 2011, 10, 3590–3603

Journal of Proteome Research

ARTICLE

Figure 5. Metabolic correlations between fecal propionate and urinary metabolites determined by OPLS. Coefficient plot of the urinary metabolites ^ = 0.84). from the control animals that correlated with fecal propionate signal (Q2Y

oligosaccharides, were elevated. In alignment with bacterial compositional changes the relative proportions of butyrate and propionate to acetate were increased in the antibiotic-treated animals (Supplementary Table S2, Supporting Information). Modulations were also observed in host endogenous energy metabolism with increased urinary excretion of TCA cycle intermediates (citrate, 2-oxoglutarate and fumarate), glucose, the ketone body D-3 hydroxybutyrate, and metabolites indirectly related to energy metabolism alanine, and lactate. The close relationship between microbial SCFA and host energy metabolism was demonstrated in the control animals where the fecal abundance of SCFA, in particular propionate, were observed to negatively correlate to several intermediates of the TCA cycle in the urine (citrate, 2-oxoglutarate, fumarate, malate, malonate). Taken together, these findings suggest that increased production of SCFA by the gut microbiota, and therefore increased bioavailability along the GIT, has a negative impact on the TCA cycle, thus reflecting a shift in energy strategy dependent upon the availability of these microbial products. While modulation of energy metabolism may be attributable to either reduction in bacterial activity or compositional shifts, these findings reinforce the connection between the microbiota and host energy metabolism. Endogenous and Combinatorial Metabolic Modulation Following Microbial Depletion

Covariation in the urinary metabolic signature and the bacterial profile demonstrate bacterial modulation had a pronounced effect upon the cometabolic phenotype of the host. Bacterial abundance, and therefore bacterial metabolic activity, was reduced by the penicillin/streptomycin regime. Additionally, variation witnessed in the microbial composition was statistically associated with variation in the microbiome with potential to influence different metabolic end points. A high degree of metabolic redundancy is encoded within the microbiome where common functional categories of genes and metabolic pathways are provided by several distinct bacterial species. Collectively, these contributions yield a “core microbiome” with a combined core metabolic capacity, and which share similarity in function across unrelated healthy individuals.34 In this study, antibiotic-induced

microbial disruption was sufficient to perturb the host metabolic profile suggesting the core microbiome of these animals had been disrupted. Metabolites derived from concerted mammalian-microbial cometabolism were reduced in the urine of SP animals following gut bacterial reduction. These compounds included 3-HPPA, 4-HPPA, PAG, hippurate and indoxyl-sulfate. Hippurate is a well understood mammalian-microbial cometabolite originating from bacterial action upon plant phenols to produce benzoate which becomes conjugated with glycine in the liver to produce hippurate. The aromatic species, 3-HPPA and 4-HPPA arise through the bacterial hydrolysis of dietary derived chlorogenic acid and further bacterial reduction of caffeic acid and a subset of microbiota have trytophanase activity which metabolizes tryptophan to indole and subsequently undergoes hepatic biotransformation to produce indoxyl-sulfate. Anaerobic bacterial metabolism of phenylalanine and phenols leads to the formation of phenylacetate, which can be conjugated with glycine in the liver to form PAG. Conversely, the loss of bacterial activity in the intestinal tract leads to an increase in metabolites typically degraded by such catabolism, improving their bioavailability, increasing the amount reaching circulation and consequently amount excreted. Increased excretion of taurine in the urine and feces of SP animals most likely relates to this reduced microbial action upon dietary taurine. High concerntrations of urinary taurine is characteristic of germ-free and acclimatizing germ-free rodents.14,15 Further evidence of microbially modulated taurine metabolism is the increase in tauro conjugated bile acids in various tissues (heart, liver, kidney) in germ-free rats.35 A similar trend toward an increase in the tauro-conjugated bile acid pool was noted following antibiotic treatment. Increased urinary excretion of creatine and creatinine may also arise from quelling of bacterial metabolism, with bacteria isolated in human feces possessing the metabolic capacity to degrade creatine and creatinine.36,37 Endogenous choline metabolism was increased following suppression of bacterial activity, with choline, betaine, DMG and glycine increased in the urine of SP animals. Choline can be metabolized to glycine through a purely mammalian pathway or it can be converted in the gut via bacterial metabolism to trimethylamine (TMA) and subsequently oxidized in the liver to 3599

dx.doi.org/10.1021/pr200243t |J. Proteome Res. 2011, 10, 3590–3603

Journal of Proteome Research form TMAO.38 Increased urinary excretion of these metabolites and decreased TMAO excretion, in addition to increased choline in the feces, reflects an increased flux toward glycine synthesis in the host after microbially mediated methylamine formation was removed. Although choline is required to synthesize phosphocholine and raised methylamine production has been implicated in nonalcoholic fatty liver disease, the elevated excretion of betaine, dimethylglycine and glycine observed in the current study indicates that choline bioavailability exceeds the necessary threshold for phosphocholine synthesis and is instead metabolized to glycine. Raised amounts of glycoproteins in the SP urine may indicate an inflammatory immune response to disrupted microbial ecology, where N-acetyl-glycoproteins are late biomarkers of inflammation and the acute-phase response.39 A fine balance exists between the mucosal immune system and the gut microbiota where a state of “controlled inflammation” is maintained. However, fluctuations within this enteric community have the potential to provoke an inflammatory response to protect the host from translocating bacteria. There is also the possibility of alterations in barrier function. Indeed, certain intestinal bacteria, such as the bifidobacteria, have been shown to directly regulate expression of tight junction proteins on mucosal epithelial cells, thus regulating intestinal permeability.40,41 Stimulation of this group of intestinal bacteria using specific dietary fiber supplements, for example, the prebiotic oligofructose, has been shown to reduce chronic systemic inflammation associated with metabolic disease and weight gain.40,42 Another contributing factor to inflammation may be the reduction in SCFA, with research demonstrating that SCFA have a role in the prevention of mucosal inflammation.43 This is corroborated by the negative correlation observed between fecal SCFA and urinary N-acetyl-glycoproteins in the control animals. Antibiotic-derived metabolites were not detected in the urine but the major metabolite of penicillin (a penicillamine structure δ 1.46 (singlet), 1.57 (singlet), 3.63 (singlet)) was observed in the fecal water. Microbiotal profiles were highly consistent within each study group for both the CN and SP animals, and across the two cages that constituted each study group, which was also reflected in their metabolic signatures. Short-term recovery from antibiotic administration was evaluated to investigate the impact of early GIT recolonization processes on the microbial and metabolic profiles and interactions. Early Bacterial Recolonization Is Dynamic and Influenced by Housing

Intrinsic population dynamics in the early phases of recolonization influence bacterial establishment and composition. This may include: the presence of symbiotic bacteria and the emergence of cooperative food webs and promotional cross-feeding (mutualistic relationship) between species, competition for resources and mucosal binding sites, or the presence of inhibitory strains that would otherwise restrict proliferation via interference mechanisms such as the production of toxic metabolites and antimicrobial compounds. The presence of existing bacteria resistant to the antibiotic effects would also influence the reestablishment of the microbiota. In addition, those enteric bacteria with a greater ability to adapt to or alter the local environment may have a competitive advantage over less fit organisms. Spatially, whether a species inhabits the mucosa or luminal environment may also contribute to growth advantage. Collectively, these factors ensure a dynamic recolonization process.

ARTICLE

Following four days of bacterial recolonization, the abundance of the microbiota remained significantly lower in the RC animals than the CN animals. Abundance was higher than SP animals, although the RC animals received only four days of antibiotictreatment in the first instance compared to eight days in the SP group. By observing the RC group, information regarding the recolonizing properties of the specific bacterial groups can be obtained. The different subsets emerging in the microbiota of RC animals postcessation of the antibiotic treatment demonstrate the influence of the microenvironment, in this case animal caging, upon recolonization. This is of interest considering host genotype, age, diet, and water were the same, and that animals were housed in the same room yet cage assignment alone was sufficient to profoundly alter early recolonization. Housing variation represents one form of stressful experience that influences recolonization following antibiotic therapy. A variety of stressful challenges have been demonstrated to influence both the microbiome and metabotype of rats.4446 Elucidating the relationship between environmental stressors, including physiological and psychological, and bacterial re-establishment postantibiotic treatment may improve our understanding of the response to antibiotic therapy. By characterizing such interactions, we gain a more reflective picture of real life environmental variation across human populations and may explain variations witnessed in the effectiveness of antibiotics across human populations. In the current study, what is also striking is the homogeneity of the microbial profiles within cage-sharing animals for all study groups, possibly a consequence of coprophagy, illustrating the influence of cohabitants upon the microbial profile. The variability in recolonization response observed in both the microbial and metabolic profiles has the potential to impact on the metabolism of drugs and other xenobiotics. The gut microbiota are recognized as an important variable in defining the metabolic fate of some xenobiotics in the host; thus, subtle variations in tightly controlled animal studies can modulate microbial composition with potential to modulate metabolic outcomes of drugs.47,48 In general, there was an increased proportion of Firmicutes in the recolonizing profiles suggesting that bacteria belonging to this phylum may be better early colonizers than those of the phylum Bacteroidetes, or at least recover more quickly following cessation of antibiotic ingestion. The high proportion of Lab158 probe counts in the RCR subgroup may relate to a particular bacterial strain or species, not present in RCβ, with a growth advantage in the early stage of recolonization. Alternatively the phylotype of these dominant organisms may encode resistance to the antibiotic regime and would therefore not endure the same degree of attenuation following antibiotic-treatment as those comprising the Lab158 probe count in the RCβ animals. Bacteria belonging to this group, including the lactobacilli and enterococci, have been shown to transfer antibiotic resistance genes readily within the gut, and certain strains are resistant to penicillin.4951 Given the similarity in urinary metabolic profiles between all RC animals at day four yet clear distinction between the two subgroups at day eight, the concept of competitive recolonization appears more feasible. The urinary metabolic profiles show some of the findings observed in the SP animals persisted in the RC animals but to a lesser extent indicating that four days of recolonization was insufficient to restore the microbiome to its predose functional state. These deviations from CN animals included elevated concentrations of taurine, DMG, creatine, creatinine, allantoin, 3600

dx.doi.org/10.1021/pr200243t |J. Proteome Res. 2011, 10, 3590–3603

Journal of Proteome Research and reduced TMAO and the cometabolites, 3-HPPA, 4-HPPA, PAG, hippurate, and indoxyl sulfate. While the urinary metabolic profiles of RC animals differed in composition to the SP animals at day eight, the fecal profiles were comparable, highlighting the greater sensitivity of the urinary profile to differences in microbial composition and abundance. Metabolic differences between the RCR and RCβ subgroups aligned with the observed variance in the microbial profiles. However, given the higher bacterial load in RCR, it is difficult to elucidate if the covariation in the metabotype and the gut microbiota was related to differences in the gut microbial composition (microbiomic variation) or gut microbial abundance (variation in metabolic capacity). The higher bacterial load of RCR was mirrored by greater excretion of the cometabolites 4-HPPA, PAG, and indoxyl-sulfate. Interestingly, RCR had a higher percentage of Firmicutes and excreted a higher amount of the SCFA acetate in the urine, while the animals of subgroup RCβ excreted greater amounts of metabolites implicated in energy metabolism (2-oxoglutarate, citrate, N-methyl nicotinamide, and D-3 hydroxybutyrate). The RCβ group bore greater metabolic resemblance to SP animals than RCR, as did the microbial profile, further emphasizing the close correlation between microbiome and host metabotype. The variation observed in the microbial and metabolic profiles of the two subgroups may reflect either, different stages in a dynamic recolonization process that would eventually converge upon a common stable climax community (temporal variation) or relate to different recolonization trajectories that stabilize upon compositionally different microbiota. In the latter, the biotransformation potential encoded in the microbiome may vary or in keeping with the concept of a core microbiome, while compositional variation may occur, functional homogeneity would be achieved. Consistent with these observations, Dethlefsen and Relman52 demonstrated large interindividual variability in the recovery of gut microbial composition of humans receiving ciprofloxacin and noted that in all cases the postdose composition was distinct from initial state. However this was based on a low study size (n = 3). While it has been shown in previous studies that a period of three weeks is required for the gut microbiome to stabilize,15,16 in this study we were concerned primarily with establishing the suitability of this model for studying the effects of antibiotic treatment as an alternative to the use of germ-free animals. We were interested in the initial phase of the recovery process, rather than the long-term consequences of the treatment, which will be studied in more depth in future work now that the model has been validated for short-term studies. Antibiotic intake has been associated with a number of human diseases including Clostridium difficile associated diarrhea or colitis,53 susceptibility to gastroenteritis and more chronic disorders including autism.54 Niche construction theory, where organisms alter their local environment and that of others, to modulate the selection pressures exerted upon themselves and their descendants, extend the effects of historical contingencies, such as antibiotic insult.3 It is therefore plausible that antibiotic dosing may be reorganizing the gut microbiota permanently and in an uncontrollable manner and that the effects of this realignment would persist until further perturbations take place.

’ CONCLUSIONS Antibiotic treatment exerted a profound effect upon the gut microbiota of rats and subsequently the cometabolic phenotype

ARTICLE

of the host. Such modulation was not exclusive to mammalian microbial cometabolic pathways but also appeared to impact endogenous host metabolism. These findings suggest the core microbiome was altered by the administration of antibiotics and is of interest given the multidimensional recolonization that can occur once antibiotic treatment ceases and reflects both local host and microbiota specific parameters. Restoration of the microbiome and it is influence on host metabolism appears to have been largely “uncontrolled” and in the present study, we are unable to ascertain whether the microbiome would return to the predose state and thus recover to the predose metabolic phenotype. However, the dramatic differences in both the metabolic phenotype and the microbiome witnessed in this study raise interesting questions regarding the subsequent metabolic state of the host after antibiotic intervention.

’ ASSOCIATED CONTENT

bS

Supporting Information Supplemental Tables S1S3. This material is available free of charge via the Internet at http://pubs.acs.org.

’ AUTHOR INFORMATION Corresponding Author

*Dr Jonathan Swann, Department of Food & Nutritional Sciences, The School of Chemistry, Food and Pharmacy, The University of Reading, Whiteknights, Reading, RG6 6AP, U.K. Tel: 0118 378 6649. Fax: 0118 931 0080. E-mail: [email protected]. Professor Elaine Holmes, Biomolecular Medicine, Department of Surgery and Cancer, Faculty of Medicine, Imperial College London, South Kensington, London SW7 2AZ, U.K. Tel: 020 7594 3220. Fax: 020 7594 3226. E-mail: elaine.holmes@imperial. ac.uk.

’ ACKNOWLEDGMENT We thank Olivier Cloarec from Korrigan Sciences Ltd. for data analysis advice. ’ REFERENCES (1) Zoetendal, E. G.; Akkermans, A. D.; Akkermans-van Vliet, W. M.; de Visser, J. A.; De Vos, W. M. The host genotype affects the bacterial community in the human gastrointestinal tract. Microb. Ecol. Health Dis. 2001, 13 (3), 129–134. (2) Van de Merwe, J. P.; Stegeman, J. H.; Hazenberg, M. P. The resident faecal flora is determined by genetic characteristics of the host. Implications for Crohn’s disease? Antonie Van Leeuwenhoek 1983, 49 (2), 119–24. (3) Dethlefsen, L.; Eckburg, P. B.; Bik, E. M.; Relman, D. A. Assembly of the human intestinal microbiota. Trends Ecol. Evol. 2006, 21 (9), 517–23. (4) Hopkins, M. J.; Sharp, R.; Macfarlane, G. T. Variation in human intestinal microbiota with age. Dig. Liver Dis. 2002, 34 (Suppl 2), S12–8. (5) Finegold, S. M. Interaction of antimicrobial therapy and intestinal flora. Am. J. Clin. Nutr. 1970, 23 (11), 1466–71. (6) Nyhlen, A.; Ljungberg, B.; Nilsson-Ehle, I.; Nord, C. E. Impact of combinations of antineoplastic drugs on intestinal microflora in 9 patients with leukaemia. Scand. J. Infect. Dis. 2002, 34 (1), 17–21. (7) Holmes, E.; Nicholson, J. Variation in gut microbiota strongly influences individual rodent phenotypes. Toxicol. Sci. 2005, 87 (1), 1–2. (8) Kuroda, K.; Yoshida, K.; Yoshimura, M.; Endo, Y.; Wanibuchi, H.; Fukushima, S.; Endo, G. Microbial metabolite of dimethylarsinic acid 3601

dx.doi.org/10.1021/pr200243t |J. Proteome Res. 2011, 10, 3590–3603

Journal of Proteome Research is highly toxic and genotoxic. Toxicol. Appl. Pharmacol. 2004, 198 (3), 345–53. (9) Swann, J.; Wang, Y.; Abecia, L.; Costabile, A.; Tuohy, K.; Gibson, G.; Roberts, D.; Sidaway, J.; Jones, H.; Wilson, I. D.; Nicholson, J.; Holmes, E. Gut microbiome modulates the toxicity of hydrazine: a metabonomic study. Mol. Biosyst. 2009, 5 (4), 351–5. (10) Upreti, R. K.; Shrivastava, R.; Chaturvedi, U. C. Gut microflora & toxic metals: chromium as a model. Indian J. Med. Res. 2004, 119 (2), 49–59. (11) Dumas, M. E.; Barton, R. H.; Toye, A.; Cloarec, O.; Blancher, C.; Rothwell, A.; Fearnside, J.; Tatoud, R.; Blanc, V.; Lindon, J. C.; Mitchell, S. C.; Holmes, E.; McCarthy, M. I.; Scott, J.; Gauguier, D.; Nicholson, J. K. Metabolic profiling reveals a contribution of gut microbiota to fatty liver phenotype in insulin-resistant mice. Proc. Natl. Acad. Sci. U.S.A. 2006, 103 (33), 12511–6. (12) Ley, R. E.; Backhed, F.; Turnbaugh, P.; Lozupone, C. A.; Knight, R. D.; Gordon, J. I. Obesity alters gut microbial ecology. Proc. Natl. Acad. Sci. U.S.A. 2005, 102 (31), 11070–5. (13) van der Waaij, D.; Sturm, C. A. Antibiotic decontimination of the digestive tract of mice. Technical procedures. Lab. Anim. Care 1968, 18 (1), 1–10. (14) Claus, S. P.; Tsang, T. M.; Wang, Y.; Cloarec, O.; Skordi, E.; Martin, F. P.; Rezzi, S.; Ross, A.; Kochhar, S.; Holmes, E.; Nicholson, J. K. Systemic multicompartmental effects of the gut microbiome on mouse metabolic phenotypes. Mol. Syst. Biol. 2008, 4, 219. (15) Nicholls, A. W.; Mortishire-Smith, R. J.; Nicholson, J. K. NMR spectroscopic-based metabonomic studies of urinary metabolite variation in acclimatizing germ-free rats. Chem. Res. Toxicol. 2003, 16 (11), 1395–404. (16) Yap, I. K.; Li, J. V.; Saric, J.; Martin, F. P.; Davies, H.; Wang, Y.; Wilson, I. D.; Nicholson, J. K.; Utzinger, J.; Marchesi, J. R.; Holmes, E. Metabonomic and microbiological analysis of the dynamic effect of vancomycin-induced gut microbiota modification in the mouse. J. Proteome Res. 2008, 7 (9), 3718–28. (17) Romick-Rosendale, L. E.; Goodpaster, A. M.; Hanwright, P. J.; Patel, N. B.; Wheeler, E. T.; Chona, D. L.; Kennedy, M. A. NMR-based metabonomics analysis of mouse urine and fecal extracts following oral treatment with the broad-spectrum antibiotic enrofloxacin (Baytril). Magn. Reson. Chem. 2009, 47 (Suppl 1), S36–46. (18) Williams, R. E.; Eyton-Jones, H. W.; Farnworth, M. J.; Gallagher, R.; Provan, W. M. Effect of intestinal microflora on the urinary metabolic profile of rats: a (1)H-nuclear magnetic resonance spectroscopy study. Xenobiotica 2002, 32 (9), 783–94. (19) Rammelkamp, C. H.; Keefer, C. S. The Absorption, Excretion, and Distribution of Penicillin. J. Clin. Invest. 1943, 22 (3), 425–37. (20) Guo, W.; Ding, J.; Huang, Q.; Jerrells, T.; Deitch, E. A. Alterations in intestinal bacterial flora modulate the systemic cytokine response to hemorrhagic shock. Am. J. Physiol. 1995, 269 (6 Pt 1), G827–32. (21) Harmsen, H. J.; Wildeboer-Veloo, A. C.; Grijpstra, J.; Knol, J.; Degener, J. E.; Welling, G. W. Development of 16S rRNA-based probes for the Coriobacterium group and the Atopobium cluster and their application for enumeration of Coriobacteriaceae in human feces from volunteers of different age groups. Appl. Environ. Microbiol. 2000, 66 (10), 4523–7. (22) Manz, W.; Amann, R.; Ludwig, W.; Vancanneyt, M.; Schleifer, K. H. Application of a suite of 16S rRNA-specific oligonucleotide probes designed to investigate bacteria of the phylum cytophaga-flavobacterbacteroides in the natural environment. Microbiology 1996, 142 (Pt 5), 1097–106. (23) Langendijk, P. S.; Schut, F.; Jansen, G. J.; Raangs, G. C.; Kamphuis, G. R.; Wilkinson, M. H.; Welling, G. W. Quantitative fluorescence in situ hybridization of Bifidobacterium spp. with genusspecific 16S rRNA-targeted probes and its application in fecal samples. Appl. Environ. Microbiol. 1995, 61 (8), 3069–75. (24) Franks, A. H.; Harmsen, H. J.; Raangs, G. C.; Jansen, G. J.; Schut, F.; Welling, G. W. Variations of bacterial populations in human feces measured by fluorescent in situ hybridization with group-specific

ARTICLE

16S rRNA-targeted oligonucleotide probes. Appl. Environ. Microbiol. 1998, 64 (9), 3336–45. (25) Veselkov, K. A.; Lindon, J. C.; Ebbels, T. M.; Crockford, D.; Volynkin, V. V.; Holmes, E.; Davies, D. B.; Nicholson, J. K. Recursive segment-wise peak alignment of biological (1)h NMR spectra for improved metabolic biomarker recovery. Anal. Chem. 2009, 81 (1), 56–66. (26) Dieterle, F.; Ross, A.; Schlotterbeck, G.; Senn, H. Probabilistic quotient normalization as robust method to account for dilution of complex biological mixtures. Application in 1H NMR metabonomics. Anal. Chem. 2006, 78 (13), 4281–90. (27) Ley, R. E.; Turnbaugh, P. J.; Klein, S.; Gordon, J. I. Microbial ecology: human gut microbes associated with obesity. Nature 2006, 444 (7122), 1022–3. (28) Turnbaugh, P. J.; Ley, R. E.; Mahowald, M. A.; Magrini, V.; Mardis, E. R.; Gordon, J. I. An obesity-associated gut microbiome with increased capacity for energy harvest. Nature 2006, 444 (7122), 1027–31. (29) Schwiertz, A.; Taras, D.; Schafer, K.; Beijer, S.; Bos, N. A.; Donus, C.; Hardt, P. D. Microbiota and SCFA in lean and overweight healthy subjects. Obesity (Silver Spring) 2010, 18 (1), 190–5. (30) Duncan, S. H.; Lobley, G. E.; Holtrop, G.; Ince, J.; Johnstone, A. M.; Louis, P.; Flint, H. J. Human colonic microbiota associated with diet, obesity and weight loss. Int. J. Obes. (London) 2008, 32 (11), 1720–4. (31) Ternak, G. Antibiotics may act as growth/obesity promoters in humans as an inadvertent result of antibiotic pollution? Med. Hypotheses 2005, 64 (1), 14–6. (32) Feighner, S. D.; Dashkevicz, M. P. Subtherapeutic levels of antibiotics in poultry feeds and their effects on weight gain, feed efficiency, and bacterial cholyltaurine hydrolase activity. Appl. Environ. Microbiol. 1987, 53 (2), 331–6. (33) Forbes, M.; Park, J. T. Growth of germ-free and conventional chicks: effect of diet, dietary penicillin and bacterial environment. J. Nutr. 1959, 67 (1), 69–84. (34) Turnbaugh, P. J.; Ley, R. E.; Hamady, M.; Fraser-Liggett, C. M.; Knight, R.; Gordon, J. I. The human microbiome project. Nature 2007, 449 (7164), 804–10. (35) Swann, J. R.; Want, E. J.; Geier, F. M.; Spagou, K.; Wilson, I. D.; Sidaway, J. E.; Nicholson, J. K.; Holmes, E. Microbes and Health Sackler Colloquium: Systemic gut microbial modulation of bile acid metabolism in host tissue compartments. Proc. Natl. Acad. Sci. U.S.A. 2010, 108, 4523–4530. (36) Dunn, S. R.; Gabuzda, G. M.; Superdock, K. R.; Kolecki, R. S.; Schaedler, R. W.; Simenhoff, M. L. Induction of creatininase activity in chronic renal failure: timing of creatinine degradation and effect of antibiotics. Am. J. Kidney Dis. 1997, 29 (1), 72–7. (37) Wyss, M.; Kaddurah-Daouk, R. Creatine and creatinine metabolism. Physiol. Rev. 2000, 80 (3), 1107–213. (38) al-Waiz, M.; Mikov, M.; Mitchell, S. C.; Smith, R. L. The exogenous origin of trimethylamine in the mouse. Metabolism 1992, 41 (2), 135–6. (39) Bell, J. D.; Brown, J. C.; Nicholson, J. K.; Sadler, P. J. Assignment of resonances for ’acute-phase’ glycoproteins in high resolution proton NMR spectra of human blood plasma. FEBS Lett. 1987, 215 (2), 311–5. (40) Cani, P. D.; Possemiers, S.; Van de Wiele, T.; Guiot, Y.; Everard, A.; Rottier, O.; Geurts, L.; Naslain, D.; Neyrinck, A.; Lambert, D. M.; Muccioli, G. G.; Delzenne, N. M. Changes in gut microbiota control inflammation in obese mice through a mechanism involving GLP2-driven improvement of gut permeability. Gut 2009, 58 (8), 1091–103. (41) Ewaschuk, J. B.; Diaz, H.; Meddings, L.; Diederichs, B.; Dmytrash, A.; Backer, J.; Looijer-van Langen, M.; Madsen, K. L. Secreted bioactive factors from Bifidobacterium infantis enhance epithelial cell barrier function. Am. J. Physiol. Gastrointest. Liver Physiol. 2008, 295 (5), G1025–34. (42) Cani, P. D.; Neyrinck, A. M.; Fava, F.; Knauf, C.; Burcelin, R. G.; Tuohy, K. M.; Gibson, G. R.; Delzenne, N. M. Selective increases 3602

dx.doi.org/10.1021/pr200243t |J. Proteome Res. 2011, 10, 3590–3603

Journal of Proteome Research

ARTICLE

of bifidobacteria in gut microflora improve high-fat-diet-induced diabetes in mice through a mechanism associated with endotoxaemia. Diabetologia 2007, 50 (11), 2374–83. (43) Venkatraman, A.; Ramakrishna, B. S.; Shaji, R. V.; Kumar, N. S.; Pulimood, A.; Patra, S. Amelioration of dextran sulfate colitis by butyrate: role of heat shock protein 70 and NF-kappaB. Am. J. Physiol. Gastrointest. Liver Physiol. 2003, 285 (1), G177–84. (44) O’Mahony, S. M.; Marchesi, J. R.; Scully, P.; Codling, C.; Ceolho, A. M.; Quigley, E. M.; Cryan, J. F.; Dinan, T. G. Early life stress alters behavior, immunity, and microbiota in rats: implications for irritable bowel syndrome and psychiatric illnesses. Biol. Psychiatry 2009, 65 (3), 263–7. (45) Teague, C. R.; Dhabhar, F. S.; Barton, R. H.; Beckwith-Hall, B.; Powell, J.; Cobain, M.; Singer, B.; McEwen, B. S.; Lindon, J. C.; Nicholson, J. K.; Holmes, E. Metabonomic studies on the physiological effects of acute and chronic psychological stress in Sprague-Dawley rats. J. Proteome Res. 2007, 6 (6), 2080–93. (46) Wang, Y.; Holmes, E.; Tang, H.; Lindon, J. C.; Sprenger, N.; Turini, M. E.; Bergonzelli, G.; Fay, L. B.; Kochhar, S.; Nicholson, J. K. Experimental metabonomic model of dietary variation and stress interactions. J. Proteome Res. 2006, 5 (7), 1535–42. (47) Clayton, T. A.; Baker, D.; Lindon, J. C.; Everett, J. R.; Nicholson, J. K. Pharmacometabonomic identification of a significant host-microbiome metabolic interaction affecting human drug metabolism. Proc. Natl. Acad. Sci. U.S.A. 2009, 106 (34), 14728–33. (48) Wilson, I. D.; Nicholson, J. K. The role of gut microbiota in drug response. Curr. Pharm. Des. 2009, 15 (13), 1519–23. (49) Feld, L.; Schjorring, S.; Hammer, K.; Licht, T. R.; Danielsen, M.; Krogfelt, K.; Wilcks, A. Selective pressure affects transfer and establishment of a Lactobacillus plantarum resistance plasmid in the gastrointestinal environment. J. Antimicrob. Chemother. 2008, 61 (4), 845–52. (50) Klare, I.; Konstabel, C.; Werner, G.; Huys, G.; Vankerckhoven, V.; Kahlmeter, G.; Hildebrandt, B.; Muller-Bertling, S.; Witte, W.; Goossens, H. Antimicrobial susceptibilities of Lactobacillus, Pediococcus and Lactococcus human isolates and cultures intended for probiotic or nutritional use. J. Antimicrob. Chemother. 2007, 59 (5), 900–12. (51) Tuohy, K.; Davies, M.; Rumsby, P.; Rumney, C.; Adams, M. R.; Rowland, I. R. Monitoring transfer of recombinant and nonrecombinant plasmids between Lactococcus lactis strains and members of the human gastrointestinal microbiota in vivo--impact of donor cell number and diet. J. Appl. Microbiol. 2002, 93 (6), 954–64. (52) Dethlefsen, L.; Relman, D. A. Microbes and Health Sackler Colloquium: Incomplete recovery and individualized responses of the human distal gut microbiota to repeated antibiotic perturbation. Proc. Natl. Acad. Sci. U.S.A. 2010, 108, 4554–4561. (53) George, W. L.; Sutter, V. L.; Finegold, S. M. Antimicrobial agent-induced diarrhea--a bacterial disease. J. Infect. Dis. 1977, 136 (6), 822–8. (54) Adams, J. B.; Romdalvik, J.; Ramanujam, V. M.; Legator, M. S. Mercury, lead, and zinc in baby teeth of children with autism versus controls. J. Toxicol. Environ. Health A 2007, 70 (12), 1046–51.

3603

dx.doi.org/10.1021/pr200243t |J. Proteome Res. 2011, 10, 3590–3603