Dynamic metabonomic and microbiological response of rats to lincomycin exposure: an integrated microbiology and metabonomics analysis

Abstract

Humans are associated with a consortium of gut microbiome and individual variations in the microbes influence host metabolism. To probe the relationship between microbiome and the host metabolic changes, we analyzed the metabonomic and microbiological response of rats exposed to lincomycin (LM) by an integrated approach combining 16S rRNA gene sequencing and ¹H NMR-based metabolomics profiling. LM exposure resulted in decreased levels of hippurate, short chain fatty acids (SCFAs) and primary bile acids and increased levels of choline and oligosaccharides. Levels of Barnesiella and Prevotella decreased sharply, whereas level of Clostridium cluster XIVa increased slightly. In addition, strong correlations were observed between metabolites and the levels of Barnesiella, Prevotella and Escherichia coli. Meanwhile, some metabolites, such as N-methylnicotinate and trigonelline, showed association with osmotic homeostasis and nucleic acid synthesis. These results suggest that LM exposure lead to significantly suppressed fermentation, gut microbial modification of bile acids and influenced liver and kidney homeostasis. This applicable method reveals the effects regarding metabonomic and microbiological responses of LM, and the combination of microbiology and metabonomics as a powerful approach offers a non-invasive means to elucidate the progression of drugs and diet. The correlation between the host and gut microbiota identifies potential biomarkers and provides substantial insight into the bacterial function, which in turn could provide a rationale for development of potential microbiota-based early prevention and therapeutic interventions.

---

Introduction

The human intestine contains a diversity of gut microbiota, approximately 10 times more than human cells.¹ The gut microbiota is involved in numerous significant biochemical functions, such as metabolic processing, immune system regulation, digestion and epithelial homeostasis.^1,2 For example, the gut microbiome contributes to modifying the chemical structure of absorbed bile acids through processes including deconjugation, dehydrogenation and dehydroxylation, leading to a disturbance of the host immune system, lipid and glucose metabolism.^1,3 Accumulating evidence indicates that metabolic perturbations associated with changes in the gut microbiome play important roles in the development of a variety of diseases such as obesity,⁴ diabetes,⁵ cardiovascular diseases,⁶ allergies asthma⁷ and inflammatory bowel disease.⁸ For example, gut microbiome-generated butyrate from nondigestive fibers has been strongly associated with IBD patients.^9,10 Given the covariation between key members of the gut microbiota community and the metabolic phenotype, it is of particular interest to probe the microbial–mammalian metabolic axis. In this aspect, antibiotics provide excellent tools to examine the relationship between the gut microbiota and metabolic changes. Commonly-used antibiotics include gentamicin and/or cetriaxone,⁹ imipenem/cilastatin sodium,¹¹ streptomycin and penicillin G¹² and vancomycin.¹³ These previous studies are helpful in providing valuable data on the interaction between host and gut microbiota. However, the gut microbiome's effect on host metabolism extends well beyond local effects in the gut to diverse remote organ systems; a single biological sample cannot provide a complementary view of the human metabonome.^14,15 Shi et al.¹⁶ revealed that the use of multiple biological matrices allowed us to systematically analyze the metabonomic responses of gallic acid exposure and substantially enhance the level of metabolome coverage. Therefore, exploring a better non-invasive research for understanding the interaction between host and gut microbiome in both urine and faeces is imperative. The faeces is composed a large quantity of gut bacteria and their metabolic products, and the urine excretes numbers of metabolites associated with the host metabolism.

Metabonomics, an array of analytical techniques including high-resolution ¹H nuclear magnetic resonance (¹H NMR) spectroscopy and chromatography-mass spectrometry (MS), coupled with multivariate statistical analysis is considered as a well-established tool to comprehensively characterize the metabolic phenotype of mammalian hosts that can be related with the microbial community in the intestinal tract.¹¹ For instance, Nicholson et al.^17–19 suggested that a ¹H NMR-based metabolomics approach is a useful platform for untargeted metabolic profiling, and they demonstrated that the metabolic variations in gastrointestinal compartments, mammalian tissues from kidney and liver, and biofluids such as urine and blood are directly correlated to the activities of various microorganisms that inhabit in the gut. In addition, polymerase chain reaction (PCR) and denaturing gradient gel electrophoresis (DGGE) of the V3 region of the 16S ribosomal ribonucleic acid (16S rRNA) sequences, commonly employed for description of the substantial diversity of the gut microbiome, is a powerful technique for identifying and quantifying the complex microbial communities.²⁰ Metabonomics coupled with microbiology analysis has been extensively used to investigate the correlation between the host and gut microbes.^1,21

In the present study, the dynamic metabonomic and microbiological response of LM exposure at the systematic level was performed by an integrated approach combining 16S rRNA gene sequencing and ¹H NMR spectroscopy-based metabolomics profiling. The aim of the study was to assess the time-dependent effects of LM exposure on the gut microbiome and its metabolic profiles in urine and faeces. In addition, we investigated the correlation between gut microbiota and metabolites in faeces. Exploring this information is particularly important for exploring the systematic biochemical progression of LM and revealing the symbiotic relationships between the host and gut microbes, which in turn will provide a foundation for development of microbiota-based personal and public health care solutions.

Materials and methods

Animal experiment and sample collection

The animal experiment was reviewed and approved by the Institutional Animal Care and Use Committee of Sun Yat-sen University, and conformed to the National Institute of Health guidelines on the ethical use of animals. A total of 14 male Sprague-Dawley (SD) rats (180–200 g), obtained from the Laboratory Animal Center of Sun Yat-sen University, were placed in individual metabolic cages and were allowed to acclimatize for one week in an SPF animal facility (temperature, 24 °C; humidity, 50–70%; light–dark cycle, 12–12 h) with free access to water and commercial rodent food at the Laboratory Animal Center of Sun Yat-sen University. After one week of acclimatization, rats were randomly divided into two groups of seven rats, namely control group (CTR) and lincomycin group (LIN). While LM in ultrapure water (Milli-Q system, Millipore, Billerica, MA) was administered by oral gavages at the dosage of 1.5 g kg⁻¹ of body weight from days 1 to 14. The CTR received the same volume of water. After antibiotic exposure, all rats were housed in a controlled environment for another seven days as a recovery period. The body weight of each rat was assessed daily just before drug administration. Daily 8 h (8:00 a.m. to 16:00 p.m.) urine and faeces (8:00 a.m.) samples were collected on day 1 preintervention and 5, 14 and 21 days postintervention for metabonomics and microbiology analysis. The whole experimental workflow is shown in Fig. S1 and the chemicals used in this study are summarized in the ESI text.‡

Sample preparation and NMR spectroscopy

Urine and faecal samples were extracted as previously optimized with slight modifications (detailed in ESI text‡).^22,23 The proton spectra of urine and faeces were collected on a Bruker AVIII 600 MHz spectrometer at a 600.13 MHz. The acquisition parameters were essentially referred in previous study.²⁴ A one-dimensional (1D) pulse sequence was used [recycle delay (RD)-G₁-90°-t₁-90°-t_m-G₂-90°-acquisition]. The water signal was suppressed by irradiation during RD of 2 s, and mixing time (t_m) of 80 ms. For each sample, the 90° pulse length was adjusted to approximately 10 μs and a total of 64 transients were accumulated into 32k data points over a spectral width of 20 ppm. For NMR signal assignment purpose, a range of two-dimensional (2D) NMR experiments (¹H–¹H correlation spectroscopy (COSY), total correlation spectroscopy (TOCSY), ¹H–¹³C heteronuclear single quantum correlation spectroscopy (HSQC) and ¹H–¹³C heteronuclear multiple bond correlation spectroscopy (HMBC))were carried out on the selected samples as previously reported.^25,26

NMR data processing

All ¹H NMR spectra of urine and faecal extracts were manually phased and baseline-corrected using TOPSPIN and referenced to the TSP resonance at δ 0.00. The urinary spectral region δ 0.50–9.50 and the faecal spectral region δ 0.50–8.50 were bucked with an equal spectra width of 0.004 ppm widths by use of AMIX software (V2.1, Bruker Biospin, Germany). To obtain only the endogenous metabolite changes as a result of LM exposure, signals from LM were carefully discarded together with regions containing urea and H₂O signals. In urine spectra, these regions included δ 4.76–4.86 (for H₂O) and δ 5.50–6.00 (for urea). For faecal spectra, the discarded regions included δ 0.87–0.92, δ 1.15–1.19, δ 1.25–1.55, δ 2.14–2.45, δ 2.73–2.9, δ 4.38–4.43, δ 5.36–5.39 (for LM), and δ 4.76–4.86 (for H₂O). Subsequently, each integral region was normalized to the total sum of all integral regions for each spectrum before pattern recognition analysis.

Microbiological analysis of faecal samples

Total bacterial DNA extractions were isolated from 180–220 mg faecal pellets using a TIANamp Stool DNA Kit according to the manufacturers' instructions. All resultant DNA samples were stored at −20 °C for further analysis. PCR amplification was performed using universal primers 357f_GC and 518r, which target the variable V3 region of the 16S rRNA gene for the predominant bacteria and were reported previously²⁷ for DGGE analysis. DGGE with 8% (w/v) polyacrylamide gels containing a linear gradient from 38% to 58% (100% denaturant corresponds to 7 M urea and 40% formamide) were used. Electrophoresis was conducted under constant voltage of 70 V for 12 h at 60 °C in 1× Tris–acetate–EDTA (TAE) buffer. Gels were stained with silver nitrate and then visualized using a digital camera (Canon, Japan). DGGE image analysis and data output were performed by Quantity One, and the intensity and position of bands in each lane were read into a spectrum of 200 variables. The data matrix from DGGE was normalized to the sum of the intensity prior to pattern recognition analysis. Bands with Q² > 0.755 and |r| > 0.755 were selected and subjected to further identification by sequencing as previously described (detailed in ESI text‡).^9,21,49 Additional PCR was performed using specific primers (Clostridium leptum subgroup or the Bacteroides spp.) based on the extracted DNA and the PCR products that were electrophoresed in DGGE gels (Tables S1–S3‡).

Data processing and statistical analysis

SPSS software (V17.0, SPSS, Chicago, USA) and SIMCA-P⁺ 12.0 (Umetrics, Sweden) were used for all analyses. Data of body weight gain and the ratio of body weight (weight gain to weight at week 0), presented as the means ± standard deviation and were processed using a two-tailed unpaired t-test with the critical p value set as 0.05. For the spectral data from NMR and DGGE, normalization to the total sum of the residual spectrum was carried out prior to pattern recognition analyses followed by scaling of the data. Initially, PCA was applied to reveal an overview intrinsic similarity/dissimilarity within each data set as a non-supervised model. To screen metabolites and DGGE bands with striking changes contributing to the separations between CTR and LIN, OPLS-DA models were then carried out using the NMR- or DGGE-data as the X-matrix and class information as the Y-matrix with Pareto scaling. The score plots showed the group clusters. The loading plots, provided potential biomarkers, were generated using a MATLAB (Version 8.0, The MathWorks Inc., USA) script developed in-house following back-scaling transformation. The color-coded correlation coefficients indicate the variable contributions to class separation, those contributing the most to the prediction of the response are shown in red, whereas those with slight or no association with the response are shown in blue.²⁸ The quality of the models was assessed by model parameters, such as R² and Q², which represent the quality of the fit and predictability of the model respectively.²⁹ For further ensuring the validity of all the models, the response permutation testing (RPT) was performed.³⁰ The correlation between NMR faecal spectra and DGGE gel data was modeled by using OPLS regression and Pearson's correlation coefficient as previous studies reported.^1,21 In the current study, a cutoff value |r| > 0.755 was chosen as discrimination significance based on the test for significance of the Pearson's product-moment correlation coefficient.⁹ The resultant metabolites and sequences were searched against Human Metabolome Database (HMDB), http://www.hmdb.ca/, (accessed October 2014), Kyoto Encyclopedia of Genes and Genomes (KEGG), http://www.genome.jp/kegg/, (accessed October 2014) and Ribosomal Database Project (RDP), http://www.rdp.cme.msu.edu, (accessed October 2014) followed by metabolic pathway or function analysis.

Results

Influence of LM on weight

During the 14 day administration, no animals died, but experimental rats displayed diarrhoea throughout the 14 day antibiotic treatment period. Diarrhoea stopped at the end of the experiment. The representative change in tendency of body weight during the experimental period is presented in Fig. S2A.‡ Significantly higher levels of body weights were observed in the LIN than in CTR after 11 to 14 days of treatment. The ratio of body weight (weight gain to weight at week 0) of LIN was approximately 5% higher than CTR. In contrast, there was no significant difference on days 5 and 21 between CTR and LIN (Fig. S2B‡). Our results were similar to those of agricultural industry study in which LM treatment is used extensively as a growth promoter.³¹

Metabolites assignment with ¹H NMR spectroscopy

Typical ¹H NMR spectra of urine and faecal extracts obtained from CTR and LIN on day 14 were shown in Fig. 1. The endogenous metabolites involved in the spectra were assigned based on the literature^9,32 and were further confirmed by 2D NMR spectra. A total of 96 metabolites were assigned (71 urine-derived and 55 faeces-derived metabolites) in Table S4.‡ The spectra were mainly comprised of choline, amino acid derivatives, bile acids, SCFAs and indole derivatives, etc. These spectra exhibited obvious systematic differences between the groups. The levels of faecal oligosaccharides increased substantially, whereas the levels of hippurate and SCAFs decreased significantly. In order to obtain more detailed information about the metabolic alterations from LIN exposure, multivariate data analysis including PCA and OPLS-DA were performed.

Fig. 1 Typical 600 MHz ¹H NMR spectra of urine from lincomycin-treated rats (A), untreated rats (B) and faecal extracts from lincomycin-treated rats (C), untreated rats (D) at day 14 post treatment. (E) The expansion of δ 0.65–0.76, (F) the expansion of δ 4.50–4.68, (G) the expansion of δ 5.15–5.34. Metabolite keys are shown in Table S4.‡

Dynamic metabolic profiles in urinary and faecal samples after LM treatment

PCA analysis was performed to generate an overview of the metabolic disturbance resulting from LM exposure in urine and faeces over the experimental period (Fig. S3‡). The trajectory plots that manifested LIN underwent a marked shift from day 1 to day 5 and continued to drift away from day 5 to day 14, and then moved closer to CTR at the end time point (Fig. S3A and S3B‡). In Fig. S3C and S3D,‡ those points of each animal on day 5 were located far from the pre-dose points, and those points representing day 21 were still different but moved closer to the control cluster. The metabolic profiles of rats dosed with LM displayed the dynamic changes during 21 days of the experiment along the first principal component, and showed aged-related physiological changes along the second principal component in both groups. The analysis showed that metabolic profiles might reflect metabonomic perturbations at different time points.

LM-induced urinary and faecal metabolites changes

After the crude screening by PCA, OPLS-DA was performed to maximize the discrimination of experimental groups and to focus on metabolic variations for the different time points in both urinary and faecal extract spectra. R²X and Q², represented the quality of fit and predictability of the sublevel mathematical models, were summarized in Tables S5 and S6.‡ The clear classifications between CTR and LIN at different time points were shown in the OPLS-DA score plots (Fig. 2 and 3), whereas the correlation coefficient loading plots illustrated significantly altered metabolites induced by LM.

Fig. 2 OPLS-DA score plots (left panel) and the correlation coefficient loading plots (right panel) of urine metabolites derived from CTR (black squares) and LIN (red circles) at each time point. Correlation coefficients of the metabolites labeled in the figures are shown in Table S5.‡ CTR: control group; LIN, lincomycin group; NMN, N-methylnicotinamide.

Fig. 3 OPLS-DA score plots (left panel) and the correlation coefficient loading plots (right panel) of faecal metabolites derived from CTR (black squares) and LIN (red circles) at each time point. Correlation coefficients of the metabolites labeled in the figures are shown in Table S6.‡ CTR: control group; LIN, lincomycin group; TCA, taurocholic acid.

In Fig. 2, the coefficient loading plot showed similar feature perturbances such as increased levels of allantoin, creatine and decreased levels of hippurate, butyrate and 3-methyl-2-oxovalerate through the whole experiment in urine. Of particular note, decreased levels of α-mannose, β-glucose, 2-oxoglutarate and succinate, as well as increased levels of taurine, hypoxanthine, fumarate, dimethylamine and alanine were observed on day 14. Higher levels of lysine, trigonelline, N-methylnicotinamide (NMN), lactate and creatinine, and decreased levels of α-ketoisovalerate and propionate were observed during the 14 day treatment period, but they all drifted back to the normal level after a seven day recovery period. Interestingly, a number of key metabolites, such SCFAs in urine were down-regulated to a less extent than those in faeces.

In Fig. 3, similar metabolic perturbations such as higher levels of oligosaccharides, taurine, choline, threonine, uridine, β-glucose, and lower levels of SCFAs and taurocholic acid (TCA), isoleucine, α-ketoisocaproate, aspartate, formate, adenine, uracil, phenylalanine, tyrosine and urocanate were observed in each time point. Decreased levels of α-ketoisovalerate, creatine and β-galactose, and increased level of glycine were observed during the 14 day treatment. Although the metabolic variations were similar in different time points, they were significantly different from those in CTR. Additionally, the discriminating metabolites of day 14 illustrated a greater extent than those on days 5 and 21. On day 21, fewer changes were found compared with CTR in terms of the number of altered metabolites and the degree of changes. The concentrations of creatine, choline, glycine and α-ketoisovalerate drifted back to the normal levels at the end of the experiment.

The dominant metabolites associated with rats treated LM in coefficient loading plots were summarized in Tables S5 and S6.‡ Based on the statistical analysis of the normalized integrals of metabolites, Tables 1 and 2 summarized the discriminating metabolites screened out in Fig. 2 and 3, accounting for the differentiation between LIN and CTR at the three time points. The results of the permutation tests are shown in Fig. S4 and S5.‡

Table 1 Statistical analysis results of the main metabolite change in urine at days 5, 14 and 21^a

Metabolites	Chemical shift	Variations
		Day 5	Day 14	Day 21
a *: indicates significant changes compared with control *p < 0.05, **p < 0.01. LIN: lincomycin group; CTR: control group.
Fumarate	6.53(s)	—	↑*	—
Taurine	3.275(t), 3.437(t)	—	↑**	—
Dimethylamine	2.729(s)	—	↑**	—
Lactate	1.33(d), 4.12(q)	↑**	↑**	—
Hippurate	3.975(d), 7.55(t), 7.64(t), 7.835(t)	↓**	↓**	↓**
Butyrate	0.90(t), 1.56(m), 2.15(t)	↓*	↓**	↓**
3-Methyl-2-oxovalerate	1.10(d), 0.88(t)	↓*	↓**	↓**
α-Ketoisovalerate	1.13(d), 3.02(m)	↓**	↓**	—
Propionate	1.06(t), 2.19(q)	↓*	↓**	—
Acetate	1.934(s)	↓*	—	—
α-Mannose	5.195(s)	—	↓*	—
β-Glucose	4.657(d), 3.235(d), 3.733(dd)	—	↓**	—
2-Oxoglutarate	2.454(t), 3.02(t)	—	↓*	—
Trigonelline	4.44(s), 8.09(m), 8.85(m), 9.125(s)	↑**	↑**	—
Allantoin	5.39(s)	↑**	↑**	↑**
Creatine	3.045(s), 3.93(s)	↑**	↑**	↑*
Creatinine	3.043(s), 4.05(s)	↑*	↑*	—
Lysine	3.03(t), 3.76(t)	↑*	↑*	—
Glycine	3.57(s)	↑**	—	—
N-Methyl nicotinamide	4.48(s), 8.90(d), 8.97(d), 9.29(s)	↑**	↑*	—
Alanine	1.492(d), 3.78(q)	—	↑**	—
Hypoxanthine	8.22(s), 8.20(s)	—	↑*	—

---

Table 2 Statistical analysis results of the main metabolite change in faece at days 5, 14 and 21^a

Metabolites	Chemical shift	Variations
		Day 5	Day 14	Day 21
a *: indicates significant changes compared with control *p < 0.05, **p < 0.01. LIN: lincomycin group; CTR: control group.
β-Glucose	4.657(d), 3.235(d), 3.733(dd)	↑**	↑**	↑**
Uridine	3.81(d), 3.92(d), 4.14(q), 4.24(t), 4.36(t), 5.90(d), 5.91(d), 7.87(d)	↑**	↑**	↑**
Threonine	1.32(d), 3.58(d)	↑**	↑**	↑**
Taurine	3.275(t), 3.437(t)	↑**	↑**	↑**
Choline	4.075(t), 3.515(t), 3.20(s)	↑**	↑*	↑**
Glycine	3.57(s)	↑**	↑**	—
Formate	8.46(s)	↓**	↓**	↓**
Adenine	8.19(s), 8.21(s)	↓**	↓**	↓**
Uracil	5.81(d), 7.54(d)	↓**	↓**	↓**
Urocanate	6.40(d), 7.31(d), 7.43(s), 7.89(s)	↓**	↓**	↓**
Tyrosine	6.91(d), 7.20(d)	↓**	↓**	↓**
β-Galactose	3.48(dd), 3.65(dd), 3.93(m), 3.71(m)	↓**	↓**	—
Creatine	4.59(d), 3.74(m)	↓**	↓**	—
Aspartate	3.045(s), 3.93(s)	↓**	↓**	↓*
Taurocholic acid (TCA)	0.67(s)	↓**	↓**	↓**
α-Ketoisocaproate	0.92(d), 2.06(m), 2.61(d)	↓**	↓**	↓**
Acetate	1.934(s)	↓**	↓**	↓**
Propionate	1.06(t), 2.19(q)	↓**	↓**	↓*
Butyrate	0.90(t), 1.56(m), 2.15(t)	↓*	↓**	↓*
Isoleucine	0.94(t), 1.01(d), 1.25(m), 1.48(m)	↓**	↓**	↓**
α-Ketoisovalerate	1.13(d), 3.02(m)	↓**	↓**	—
Phenylalanine	3.13(dd), 3.29(dd), 3.98(dd), 7.33(m), 7.38(m), 7.43(m)	↓**	↓**	↓**
Xanthine	7.91(s)	—	↓**	↓**

---

Microbial profiles of faecal samples using DGGE analysis

High and low levels of metabolic variations from NMR analysis were found on day 14. To probe the disturbances to gut microbiota, the microbial structure on day 14 was represented by DGGE patterns of the 16S rRNA gene V3 region for predominant bacteria and of regions for two specific groups, Bacteroides spp. and Clostridium leptum subgroups. Each DGGE gel was loaded with samples from CTR and LIN to avoid integral variations for directly comparing two groups. A total number of 41 bands were detected from the 16S rRNA gene V3 regions (Fig. 4). The bacterial diversity of LIN was found to reduce compared with the CTR.

Fig. 4 OPLS-DA score plots (left panel), PCR-DGGE profiles (middle panel) and correlation coefficient loading plots (right panel) derived from 16S rRNA gene V3 region, indicating the discrimination between CTR (black) and LIN (red) on day 14. The bands of information can be found in Table 3. CTR: control group; LIN: lincomycin group.

Predictions for the difference between two groups were shown in OPLS-DA score plots and corresponding coefficient loading plots in Fig. 4. Twelve differential bands were identified by cloning and sequencing of 16S rRNA gene V3 regions (Table 3). The sequences are deposited in the GenBank with access numbers of KP 666050–666061. They mainly belonged to three phyla: four species from Bacteroidetes, five from Proteobacteria and three from Firmicutes. Levels of Barnesiella and Prevotella, which belongs to Bacteroidetes, significantly decreased, whereas levels of Proteobacteria and Firmicutes markedly increased. Although Bacteroides spp. showed no difference in the total bacterial abundance, there exhibited small but no significance disturbance in a slice of subspecies (Fig. S6A‡). Additionally, Clostridium leptum represented a significant increase in bacterial diversity compared with CTR (Fig. S6B‡).

Table 3 Closest relatives of 16S rDNA V3 regions sequences derived from DGGE bands

DGGE band	Seq. length (bp)	Domain	Phylum	Class	Order	Family	Genus	Species	S_ab score	Accession numbers
Band 1	194	Bacteria	Proteobacteria	Gammaproteobacteria	Enterobacteriales	Enterobacteriaceae	Escherichia	Uncultured bacterium	1.000	KP 666050
Band 2	169	Bacteria	Firmicutes	Clostridia	Clostridiales	Lachnospiraceae	Clostridium XIVa	Uncultured bacterium	0.853	KP 666051
Band 3	172	Bacteria	Firmicutes	Clostridia	Clostridiales	Ruminococcaceae	Flavonifractor	Uncultured bacterium	0.915	KP 666052
Band 4	169	Bacteria	Firmicutes	Clostridia	Clostridiales	Lachnospiraceae	Lachnoanaerobaculum	Uncultured bacterium	0.913	KP 666053
Band 5	195	Bacteria	Proteobacteria	Gammaproteobacteria	Pseudomonadales	Moraxellaceae	Acinetobacter	Acinetobacter baumannii	1.000	KP 666054
Band 6	189	Bacteria	Bacteroidetes	Bacteroidia	Bacteroidales	Porphyromonadaceae	Prevotella	Uncultured bacterium	0.967	KP 666055
Band 7	194	Bacteria	Proteobacteria	Gammaproteobacteria	Enterobacteriales	Enterobacteriaceae	Escherichia	Escherichia coli	1.000	KP 666056
Band 8	194	Bacteria	Proteobacteria	Gammaproteobacteria	Enterobacteriales	Enterobacteriaceae	Escherichia	Escherichia coli	0.960	KP 666057
Band 9	194	Bacteria	Proteobacteria	Gammaproteobacteria	Enterobacteriales	Enterobacteriaceae	Escherichia	Uncultured bacterium	1.000	KP 666058
Band 10	189	Bacteria	Bacteroidetes	Bacteroidia	Bacteroidales	Porphyromonadaceae	Barnesiella	Uncultured bacterium	0.978	KP 666059
Band 11	190	Bacteria	Bacteroidetes	Bacteroidia	Bacteroidales	Porphyromonadaceae	Barnesiella	Uncultured bacterium	0.962	KP 666060
Band 12	189	Bacteria	Bacteroidetes	Bacteroidia	Bacteroidales	Porphyromonadaceae	Barnesiella	Uncultured bacterium	0.961	KP 666061

---

Integration of ¹H NMR spectral data and DGGE microbial data

To reveal the potential correlations between gut microbiome and metabolites, NMR and DGGE data of faeces were correlated by using Orthogonal Projection to Latent Structure (OPLS) regression and Pearson's correlation coefficient. The resulting association, shown in Fig. 5 and summarized in Table S7,‡ displayed both positive and negative correlations between metabolites and the identified bacterial groups. Many of the identified compounds were well-known gut microbial co-metabolites, e.g., SCFAs, TCA and glycine. Bacteria belonging to the same genus shared similar correlative metabolites. The genus of E. coli (Band 1, 7, 8, 9) exhibited positive correlations with choline, tyrosine, TCA and uridine and reverse correlations with α-ketoisovalerate, glycine, propionate and acetate. Within the identified three Barnesiella (Band 10, 11, 12) genera, there were positive correlations with the levels of SCFAs and TCA, and negative correlations with urocanate and isoleucine. The genus Lachnoanaerobaculum (Band 4) was positively correlated with α-ketoisovalerate and negatively correlated with creatine. The genus Acinetobacter (Band 5) displayed a positive correlation with acetate and a negative correlation with adenine and phenylalanine. Clostridium XIVa (Band 2) showed a positive link to TCA, asparagine and choline and a reverse correlation with α-ketoisovalerate, adenine and glycine. However, the genus Prevotella (Band 6) only showed a strong negative correlation with α-ketoisovalerate. Furthermore, Flavonifractor (Band 3) was also significantly associated with creatine and glycine in a negative manner. In summary, LM exposure induced a significant taxonomic perturbation in the gut microbiome, which in turn substantially altered the metabolomic profiles.

Fig. 5 DGGE bands and associations with specific faecal metabolites are shown for bacterial species with the direction of correlation indicated by red (positive) or blue (negative) lines. TCA: taurocholic acid. The correlation results were summarized in Table S7.‡

Discussion

A number of different antibiotics have thus far been applied to investigate the effect on microbial perturbation. LM is a clinically useful antibiotic against Gram-positive bacteria, especially used in patients allergic to the penicillins.^33,34 Additionally, on the basis of its poor systemic bioavailability, LM was selected at the dosage of 1.5 g kg⁻¹ of body weight in our study to rapidly and completely remove the microbes from the gut.³⁵ We combined 16S rRNA sequencing and metabolomics to characterize the time-dependent microbiological and endogenous metabolic response to LM in urine and faeces. The data clearly indicated that LM exposure disturbed either the gut bacteria at the abundance level or the urinary/faecal metabolomic profiles. Correlation analysis in the faeces identified some gut bacteria were highly correlated with altered metabolites. These results shown that a dynamic dual metabonomics analysis combined with DGGE analysis provides a better non-invasive strategy for identifying meaningful biomarkers and functional bacteria compared with any single biological sample in isolation. This study revealed that LM directly impacted the abundance level and diversity of gut microbiota, which in turn could affect energy metabolism, bile acid enterohepatic recycling and SCFA fermentation through the modulation of bacterial metabolic products.

The strajectories displayed time-dependence of urinary and faecal metabolic profiles of LM exposure across the 14 day treatment period and the seven day recovery in Fig. S3.‡ In Fig. 2 and 3, compared with the metabolic variations of days 5 and 21, greater numbers of altered metabolites and a higher degree of changes on day 14 indicated that the impact of LM on the rats altered continuously with time. LM perturbation altered faecal metabolites to a greater extent than urinary metabolites because faecal metabolic profile is a direct reflection of changes in microbial composition. The key altered metabolites were widely distributed across the mammalian–microbial metabolic system involved in gut microbiome metabolism, energy metabolism and nucleic acids synthesis as shown in Fig. 6.

Fig. 6 Metabolic pathways altered by LM exposure. ↑, up-regulated; ↓, down-regulated; red colour, faeces; blue colour, urine. Abbreviations: TCA, taurocholic acid; TβMCA, tauro-β-muricholic acid; DCA, deoxycholic acid; CA, cholic acid; NMN, N-methylnicotinamide.

LM disturbs carbohydrates metabolism

In the current investigation, the energy harvesting capacity of gut microbiota exerts a strong influence on host metabolism. This is mainly reflected by the increased level of oligosaccharides and the decreased levels of monosaccharides (β-glucose, β-galactose) and fermentation products (SCFAs) in LIN shown in Fig. 2 and 3. This result indicates that the metabolism of oligosaccharides-producing monosaccharides and SCFAs was disturbed by LM treatment. Nondigestible carbohydrates are degraded by gut microbiota because they manifest variety of glycoside hydrolase.³⁶ Antibiotic treatment resulted in the accumulation of oligosaccharides,⁹ which is supported by our investigation. More specifically, SCFAs were discovered to positively correlate with the abundance of Barnesiella, Prevotella, which suggest that these bacteria play important roles in metabolism of SCFAs. A recent study revealed that genus Barnesiella and Prevotella belong to the Porphyromonadaceae family, and that Porphyromonadaceae can produce a significant amount of SCFAs from glucose.³⁷ LM treatment has certainly restrained the growth of these two species bacteria, resulting in the accumulation of oligosaccharides and the decrease of SCFAs. Another example to support LM exposure's perturbation of energy metabolism stems from the observation that ketoisovalerate and isoleucine were significantly increased in faecal extractions of LM-treated rats (Fig. 6). Isovalerate and 2-methyl-butyrate are produced by the oxidation of leucine and isoleucine via the degradation from proteins, peptides, and amino acids.^38,39 Previous studies have shown that the increased relative abundance of Firmicutes and Bacteroidetes in the gut is associated with obesity both in mice and humans.^40,41 Interestingly, we also found that body weight gain was positively related to the ratio of Clostridium leptum and Bacteroides spp., which may serve as a novel method for gaining insight into how microbial components contribute to mammal growth.

LM disturbs aromatic amino acids metabolism

Aromatic amino acids were significantly altered after LM treatment. They were highly correlated with perturbed gut bacterial families. Hippurate, as a biomarker for assessing the balance of the gut microbial community,¹³ decreased significantly in the current study. Decreased level of hippurate positively correlated to decrease in the abundance of Clostridia discovered in CD.^21,42 Interestingly, three Firmicutes families, classified into the same order of Clostridiales, significantly decreased in abundance after LM exposure in Fig. 4. This result supported our notion that the metabolic biomarkers were highly correlated with perturbed gut bacterial families. Some proteins and other aromatic amino acids such as phenylalanine, tyrosine and tryptophan also can be catabolized by the gut microbiota.^38,43 Notably, we discovered that tyrosine was negatively correlated with the abundance of the E. coli in Fig. 5, which indicated that E. coli may involve in the catabolism of tyrosine. The elevation of E. coli's relative abundance in faeces was associated with excessive weight gain in adolescents, and pregnant women.^44,45 Therefore we tentatively hypothesize that the increased E. coli act as one factor to interpretate the body weight gain in the current study.

LM disturbs bile acids metabolism

After LM treatment, the depleted level of TCA and the increased level of urinary β-alanine were observed in LIN. These results suggest that metabolism of bile acid enterohepatic recycling was hampered by LM treatment. It was known that bile acids are synthesized from cholesterol in the liver and further undergo extensive enterohepatic recycling and gut microbial processes, including deconjugation, dehydrogenation and dehydroxylation.¹ The primary bile acids can be transformed to secondary bile acids by intestinal bacteria via 7α dehydroxylation, which has been identified from cell extracts from Bacteroides, while 7α-dehydroxylase was not discovered in Barnesiella and Prevotella.^46,47 These results supported our investigation that taurocholic acid (TCA) was positively correlated with Barnesiella abundance in Fig. 5. The underlying mechanisms of the positive correlation between TCA and Escherichia, Clostridium XIVa abundance remain elusive; however, such associations could be achieved by other mechanisms besides changing the diversity and abundance of bacteria present in the gut, such as the regulation of gene and protein expression in bacteria.⁴⁸ Moreover, bile acid metabolism can affect lipid and glucose metabolism through the modulation of FXR and TGR5 signalling.¹⁸ It seems that increased body weight may stem from the disturbance of bile acids. Urinary β-alanine, a competitive inhibitor of taurine, is untaken in the kidney and liver because it shares the same transporter.⁴⁹ In our results, it showed a higher level of excretion in the LIN (Fig. 2), which illustrated the different requirements of taurine in the metabolism of bile acids. Moreover, the role of gut bacterial in the regulation of bile acids has raised the question of whether bile acids modulated by the gut microbiome are associated with regulation of the host immune system. Previous studies demonstrated that LM could modulate the immune function.⁵⁰ Clearly, further research is needed to shed light on the role of the gut microbiome and its associated metabolites, including bile acids, in LM-induced impaired immune responses and the body weight gain.

LM disturbs liver and kidney functions

Liver and kidney are the main organs for the detoxification of xenobiotic drugs; hence the homeostasis of the liver and kidney may be disturbed by LM exposure, which is happens to be demonstrated in this investigation. Among the metabolic biomarkers identified in faeces, we observed fluctuation of the concentration of nucleotides and nucleoside in rats exposed to LM. Since the liver is the primary organ for nucleotide synthesis, our observation suggests that LM exposure disturbed nucleotide synthesis and nucleic acid synthesis (Fig. 6). Moreover, we observed elevated levels of N-methylnicotinate (NMN) and trigonelline in the urine (Fig. 6), which is suggestive of antioxidation activity by LM. This is because trigonelline and NMN are the methylated metabolites of niacin (vitamin B3) and nicotinamide, respectively, which can be generated during the conversion from S-adenosylmethionine to S-adenosylhomocysteine in the process of biosynthesis of cysteine, an essential amino acid of glutathione synthesis.⁵¹ This is another important line of evidence supporting the notion that LM exposure causes a perturbation of liver function. Furthermore, the observation of choine positively correlated to the abundance of Proteobacteria (Fig. 5) indicated that LM may be associated with liver during choline depletion. Spencer et al.⁵² revealed that a low quantity of Proteobacteria in human faecal microbiota was associated with hepatic status. Elevated levels of a range of organic osmolytes, such as taurine and choline, observed in this investigation, imply that LM exposure may lead to disturbances in the kidney osmolarity.⁵³ Additionally, we observed depletions in the levels of amino acids, such as isoleucine and tyrosine in LIN, which suggests LM exposure could also result in the inhibition of protein synthesis. This view is particularly supported by the depleted level of tyrosine because tyrosine is known to be used only in protein synthesis.⁵⁴

Future studies are needed to address the mechanistic basis of their association. LM exposure may result in shifted metabolic profiles by affecting the physiology of the gut microbiota without changing the species and abundance. Therefore, an altered metabolome may not entirely result from shifts in the spectrum of microbes demonstrated by 16S rRNA gene sequencing. Such metabolic alterations could be achieved via other mechanisms besides changing the composition and abundance of bacteria that inhabit the gut, such as the regulation of gene and protein expression in bacteria. A recent study revealed that xenobiotics significantly changed the physiology and gene expression in gut microbiota.⁴⁸ Therefore, metatranscriptomics and metaproteomics profiling are warranted in the future to elucidate the role of the gut bacteria.

Conclusion

We used 16S rRNA gene sequencing and metabolomics profiling to study the impact of LM exposure on the gut microbiome and its metabolic profiles. Our results showed LM not only perturbed the liver and kidney functions but also modulates the host systematic metabolism through disturbance of the gut bacteria, which were manifested in decreased levels of urinary hippurate, short chain fatty acids (SCFAs), primary bile acids and increased levels of choline and oligosaccharides. In addition, strong correlations between metabolites and the levels of Bacteroides, Barnesiella, Prevotella, Flavonifractor, Lachnoanaerobaculum, Acinetobacter and Escherichia coli were observed to guide further research. Taken together, these data demonstrated substantial insight into the effects regarding microbiology and metabolomics response to LM treatment. Advances in technology in both metabolic phenotyping and microbial profiling methods have improved our ability to derive correlations between microbial and metabolic phenotypes, which are of particular important for optimizing therapeutic strategies and early interventions.

Conflict of interest

The authors declare no competing financial interests.

Acknowledgements

We acknowledge financial support from the National Natural Science Foundation of China (No. 81173564, No. 81274028, No. 81473319 and No. 81473540) and Guangdong Provincial Key Laboratory of New Drug Design and Evaluation (2011A060901014).

References

1. K. Lu, R. P. Abo, K. A. Schlieper, M. E. Graffam, S. Levine, J. S. Wishnok, J. A. Swenberg, S. R. Tannenbaum and J. G. Fox, Environ. Health Perspect., 2014, 122, 284–291.
2. V. B. Young, R. A. Britton and T. M. Schmidt, Interdiscip. Perspect. Infect. Dis., 2008, 2008, 296873.
3. C. Thomas, A. Gioiello, L. Noriega, A. Strehle, J. Oury, G. Rizzo, A. Macchiarulo, H. Yamamoto, C. Mataki, M. Pruzanski, R. Pellicciari, J. Auwerx and K. Schoonjans, Cell Metab., 2009, 10, 167–177.
4. P. J. Turnbaugh, R. E. Ley, M. A. Mahowald, V. Magrini, E. R. Mardis and J. I. Gordon, Nature, 2006, 444, 1027–1031.
5. L. Zhao, Nat. Rev. Microbiol., 2013, 11, 639–647.
6. Z. Wang, E. Klipfell, B. J. Bennett, R. Koeth, B. S. Levison, B. Dugar, A. E. Feldstein, E. B. Britt, X. Fu, Y. M. Chung, Y. Wu, P. Schauer, J. D. Smith, H. Allayee, W. H. Tang, J. A. DiDonato, A. J. Lusis and S. L. Hazen, Nature, 2011, 472, 57–63.
7. S. L. Russell, M. J. Gold, M. Hartmann, B. P. Willing, L. Thorson, M. Wlodarska, N. Gill, M. R. Blanchet, W. W. Mohn, K. M. McNagny and B. B. Finlay, EMBO Rep., 2012, 13, 440–447.
8. N. A. Nagalingam and S. V. Lynch, Inflammatory Bowel Dis., 2012, 18, 968–984.
9. Y. Zhao, J. Wu, J. V. Li, N. Y. Zhou, H. Tang and Y. Wang, J. Proteome Res., 2013, 12, 2987–2999.
10. J. R. Marchesi, E. Holmes, F. Khan, S. Kochhar, P. Scanlan, F. Shanahan, I. D. Wilson and Y. Wang, J. Proteome Res., 2007, 6, 546–551.
11. X. Zheng, G. Xie, A. Zhao, L. Zhao, C. Yao, N. H. Chiu, Z. Zhou, Y. Bao, W. Jia, J. K. Nicholson and W. Jia, J. Proteome Res., 2011, 10, 5512–5522.
12. J. R. Swann, K. M. Tuohy, P. Lindfors, D. T. Brown, G. R. Gibson, I. D. Wilson, J. Sidaway, J. K. Nicholson and E. Holmes, J. Proteome Res., 2011, 10, 3590–3603.
13. I. K. S. Yap, J. V. Li, J. Saric, F. P. Martin, H. Davies, Y. Wang, I. D. Wilson, J. K. Nicholson, J. Utzinger, J. R. Marchesi and E. Holmes, J. Proteome Res., 2008, 7, 3718–3728.
14. S. P. Claus, S. L. Ellero, B. Berger, L. Krause, A. Bruttin, J. Molina, A. Paris, E. J. Want, I. de Waziers, O. Cloarec, S. E. Richards, Y. Wang, M. E. Dumas, A. Ross, S. Rezzi, S. Kochhar, P. van Bladeren, J. C. Lindon, E. Holmes and J. K. Nicholson, mBio, 2011, 2, e00271.
15. R. D. Heijtz, S. Wang, F. Anuar, Y. Qian, B. Bjorkholm, A. Samuelsson, M. L. Hibberd, H. Forssberg and S. Pettersson, Proc. Natl. Acad. Sci. U. S. A., 2011, 108, 3047–3052.
16. J. T. Bjerrum, O. H. Nielsen, F. Hao, H. Tang, J. K. Nicholson, Y. Wang and J. Olsen, J. Proteome Res., 2010, 9, 954–962.
17. S. P. Claus, T. M. Tsang, Y. Wang, O. Cloarec, E. Skordi, F. P. Martin, S. Rezzi, A. Ross, S. Kochhar, E. Holmes and J. K. Nicholson, Mol. Syst. Biol., 2008, 4, 219–233.
18. F. P. J. Martin, M. E. Dumas, Y. Wang, C. Legido-Quigley, I. K. Yap, H. Tang, S. Zirah, G. M. Murphy, O. Cloarec, J. C. Lindon, N. Sprenger, L. B. Fay, S. Kochhar, P. van Bladeren, E. Holmes and J. K. Nicholson, Mol. Syst. Biol., 2007, 3, 112–127.
19. F. P. J. Martin, Y. Wang, I. K. S. Yap, N. Sprenger, J. C. Lindon, S. Rezzi, S. Kochhar, E. Holmes and J. K. Nicholson, J. Proteome Res., 2009, 8, 3464–3474.
20. S. O. Noor, K. Ridgway, L. Scovell, E. K. Kemsley, E. K. Lund, C. Jamieson, I. T. Johnson and A. Narbad, BMC Gastroenterol., 2010, 10, 134.
21. M. Li, B. Wang, M. Zhang, M. Rantalainen, S. Wang, H. Zhou, Y. Zhang, J. Shen, X. Pang, M. Zhang, H. Wei, Y. Chen, H. Lu, J. Zuo, M. Su, Y. Qiu, W. Jia, C. Xiao, L. M. Smith, S. Yang, E. Holmes, H. Tang, G. Zhao, J. K. Nicholson, L. Li and L. Zhao, Proc. Natl. Acad. Sci. U. S. A., 2008, 105, 2117–2122.
22. C. Xiao, F. Hao, X. Qin, Y. Wang and H. Tang, Analyst, 2009, 134, 916–925.
23. J. Wu, Y. An, J. Yao, Y. Wang and H. Tang, Analyst, 2010, 135, 1023–1030.
24. F. Dong, L. Zhang, F. Hao, H. Tang and Y. Wang, J. Proteome Res., 2013, 12, 2958–2966.
25. O. Beckonert, H. C. Keun, T. M. Ebbels, J. Bundy, E. Holmes, J. C. Lindon and J. K. Nicholson, Nat. Protoc., 2007, 2, 2692–2703.
26. J. F. Wu, E. Holmes, J. Xue, S. H. Xiao, B. H. Singer, H. R. Tang, J. Utzinger and Y. L. Wang, Int. J. Parasitol., 2010, 40, 695–703.
27. H. Wei, L. Dong, T. Wang, M. Zhang, W. Hua, C. Zhang, X. Pang, M. Chen, M. Su, Y. Qiu, M. Zhou, S. Yang, Z. Chen, M. Rantalainen, J. K. Nicholson, W. Jia, D. Wu and L. Zhao, FEMS Microbiol. Ecol., 2010, 73, 577–586.
28. G. Liu, G. Yang, T. Fang, Y. Cai, C. Wu, J. Wang, Z. Huang and X. Chen, RSC Adv., 2014, 4, 23749–23758.
29. L. Zhang, L. Wang, Y. Hu, Z. Liu, Y. Tian, X. Wu, Y. Zhao, H. Tang, C. Chen and Y. Wang, Biomaterials, 2013, 34, 7117–7126.
30. Y. Yang, Z. Han, Y. Wang, L. Wang, S. Pan, S. Liang and S. Wang, RSC Adv., 2015, 5, 36732–36741.
31. Y. Wang, R. Wang, J. Wang, H. Yang, J. Song, Y. Wang and A. Deng, J. Sci. Food Agric., 2010, 90, 2083–2089.
32. X. Shi, C. Xiao, Y. Wang and H. Tang, J. Proteome Res., 2013, 12, 991–1006.
33. J. Spizek and T. Rezanka, Appl. Microbiol. Biotechnol., 2004, 64, 455–464.
34. I. B. R. Duncan and B. Jeans, Can. Med. Assoc. J., 1965, 93, 685–691.
35. G. A. Albarellos, L. Montoya, G. A. A. Denamiel, S. M. Passini and M. F. Landoni, J. S. Afr. Vet. Assoc., 2013, 84, 968–972.
36. S. R. Gill, M. Pop, R. T. Deboy, P. B. Eckburg, P. J. Turnbaugh, B. S. Samuel, J. I. Gordon, D. A. Relman, C. M. Fraser-Liggett and K. E. Nelson, Science, 2006, 312, 1355–1359.
37. K. M. Wylie, R. M. Truty, T. J. Sharpton, K. A. Mihindukulasuriya, Y. Zhou, H. Gao, E. Sodergren, G. M. Weinstock and K. S. Pollard, PLoS One, 2012, 7, e35294.
38. E. A. Smith and G. T. Macfarlane, Anaerobe, 1997, 3, 327–337.
39. G. L. Gall, S. O. Noor, K. Ridgway, L. Scovell, C. Jamieson, I. T. Johnson, I. J. Colquhoun, E. K. Kemsley and A. Narbad, J. Proteome Res., 2011, 10, 4208–4218.
40. P. J. Turnbaugh, M. Hamady, T. Yatsunenko, B. L. Cantarel, A. Duncan, R. E. Ley, M. L. Sogin, W. J. Jones, B. A. Roe, J. P. Affourtit, M. Egholm, B. Henrissat, A. C. Heath, R. Knight and J. I. Gordon, Nature, 2009, 457, 480–484.
41. F. H. Karlsson, D. W. Ussery, J. Nielsen and I. Nookaew, Microb. Ecol., 2011, 61, 473–485.
42. H. Sokol, C. Lay, P. Seksik and G. W. Tannock, Inflammatory Bowel Dis., 2008, 14, 858–867.
43. E. Diaz, A. Ferrandez, M. A. Prieto and J. L. Garcia, Microbiol. Mol. Biol. Rev., 2001, 65, 523–569.
44. A. Santacruz, M. C. Collado, L. Garcia-Valdes, M. T. Segura, J. A. Martin-Lagos, T. Anjos, M. Marti-Romero, R. M. Lopez, J. Florido, C. Campoy and Y. Sanz, Br. J. Nutr., 2010, 104, 83–92.
45. A. Santacruz, A. Marcos, J. Warnberg, A. Marti, M. Martin-Matillas, C. Campoy, L. A. Moreno, O. Veiga, C. Redondo-Figuero, J. M. Garagorri, C. Azcona, M. Delgado, M. Garcia-Fuentes, M. C. Collado, Y. Sanz and E. S. Group, Obesity, 2009, 17, 1906–1915.
46. R. W. Owen, R. F. Bilton and M. E. Tenneson, Biochem. Soc. Trans., 1977, 5, 1711–1713.
47. K. Kwawamoto, I. Horibe and K. Uchida, J. Biochem., 1989, 106, 1049–1053.
48. C. F. Maurice, H. J. Haiser and P. J. Turnbaugh, Cell, 2013, 152, 39–50.
49. M. E. Bollard, H. C. Keun, O. Beckonert, T. M. Ebbels, H. Antti, A. W. Nicholls, J. P. Shockcor, G. H. Cantor, G. Stevens, J. C. Lindon, E. Holmes and J. K. Nicholson, Toxicol. Appl. Pharmacol., 2005, 204, 135–151.
50. D. M. Radomska-Lesniewska, E. Skopinska-Rozewska and J. Malejczyk, Cent. Eur. J. Immunol., 2010, 35, 217–222.
51. X. J. Zhao, F. Hao, C. Huang, M. Rantalainen, H. Lei, H. Tang and Y. Wang, J. Proteome Res., 2012, 11, 4712–4721.
52. M. D. Spencer, T. J. Hamp, R. W. Reid, L. M. Fischer, S. H. Zeisel and A. A. Fodor, Gastroenterology, 2011, 140, 976–986.
53. W. Neuhofer and F. X. Beck, Physiology, 2006, 21, 171–180.
54. M. M. Hallemeesch, P. B. Soeters and N. E. Deutz, Am. J. Physiol.: Renal, Fluid Electrolyte Physiol., 2002, 282, F316–F323.

---

Footnotes

† Accession codes: The GenBank submission accession numbers are KP 666050–666061.

‡ Electronic supplementary information (ESI) available. See DOI: 10.1039/c5ra10626e

§ These authors contributed equally to this work.

---