Analytica Chimica Acta 903 (2016) 100e109

Contents lists available at ScienceDirect

Analytica Chimica Acta journal homepage: www.elsevier.com/locate/aca

Dansylation isotope labeling liquid chromatography mass spectrometry for parallel profiling of human urinary and fecal submetabolomes Xiaoling Su a, Nan Wang a, b, Deying Chen a, Yunong Li b, Yingfeng Lu a, Tao Huan b, Wei Xu a, Liang Li a, b, *, Lanjuan Li a, ** a

State Key Laboratory and Collaborative Innovation Center for Diagnosis and Treatment of Infectious Diseases, The First Affiliated Hospital, College of Medicine, Zhejiang University, Hangzhou 310003, China Department of Chemistry, University of Alberta, Edmonton, Alberta T6G 2G2, Canada

b

h i g h l i g h t s

g r a p h i c a l a b s t r a c t

 A method of parallel analysis of urine and fecal metabolomes is developed.  Dansylation isotope labeling LC-MS is used for amine/phenol submetabolome quantification.  Detection of 3089 metabolites in urine and 3012 metabolites in feces for a total of 5372 metabolites.  Parallel profiling of urine and feces increases the metabolomic coverage.

a r t i c l e i n f o

a b s t r a c t

Article history: Received 9 October 2015 Received in revised form 19 November 2015 Accepted 21 November 2015 Available online 25 November 2015

Human urine and feces can be non-invasively collected for metabolomics-based disease biomarker discovery research. Because urinary and fecal metabolomes are thought to be different, analysis of both biospecimens may generate a more comprehensive metabolomic profile that can be better related to the health state of an individual. Herein we describe a method of using differential chemical isotope labeling (CIL) liquid chromatography mass spectrometry (LC-MS) for parallel metabolomic profiling of urine and feces. Dansylation labeling was used to quantify the amine/phenol submetabolome changes among different samples based on 12C-labeling of individual samples and 13C-labeling of a pooled urine or pooled feces and subsequent analysis of the 13C-/12C-labeled mixture by LC-MS. The pooled urine and pooled feces are further differentially labeled, mixed and then analyzed by LC-MS in order to relate the metabolite concentrations of the common metabolites found in both biospecimens. This method offers a means of direct comparison of urinary and fecal submetabolomes. We evaluated the analytical performance and demonstrated the utility of this method in the analysis of urine and feces collected daily from three healthy individuals for 7 days. On average, 2534 ± 113 (n ¼ 126) peak pairs or metabolites could be detected from a urine sample, while 2507 ± 77 (n ¼ 63) peak pairs were detected from a fecal sample. In total, 5372 unique peak pairs were detected from all the samples combined; 3089 and 3012 pairs were found in urine and feces, respectively. These results reveal that the urine and fecal metabolomes are very different, thereby justifying the consideration of using both biospecimens to increase the probability of

Keywords: Mass spectrometry Liquid chromatography Isotope labeling Urine Feces Metabolomics

* Corresponding author. Department of Chemistry, University of Alberta, Edmonton, Alberta T6G 2G2, Canada. ** Corresponding author. The First Affiliated Hospital, Zhejiang University, Hangzhou 310003, China E-mail addresses: [email protected] (L. Li), [email protected] (L.J. Li). http://dx.doi.org/10.1016/j.aca.2015.11.027 0003-2670/© 2015 Elsevier B.V. All rights reserved.

X. Su et al. / Analytica Chimica Acta 903 (2016) 100e109

101

finding specific biomarkers of diseases. Furthermore, the CIL LC-MS method described can be used to perform parallel quantitative analysis of urine and feces, resulting in more complete coverage of the human metabolome. © 2015 Elsevier B.V. All rights reserved.

1. Introduction

2. Experimental

Urine is widely used for biomarker discovery research, as it contains metabolic signatures of many biochemical pathways associated with a health state. Profiling urine metabolome can generate a rich source of information about the global physiological status of an organism [1e3] and various pathological conditions such as kidney-related diseases [4e6], diabetes [7,8] and cancer [9e11]. Feces are also gaining more attention as a biospecimen for metabolomics since fecal samples contain not only endogenous human metabolites, but also gut microbiota metabolites and residues or metabolites of digested materials [12e15]. Thus, human feces are an excellent source to interrelate the microbiome and human metabolomes [16]. Recent studies have suggested that the gut microbiome performs numerous important biochemical functions for the host, and disorders of the microbiome are associated with many diverse human disease processes [17e20]. Because both urine and feces can be readily collected noninvasively and their metabolomes are likely different, it would make sense to profile both biospecimen simultaneously in order to provide more comprehensive information on the human metabolome. NMR has been used for parallel profiling of urine and feces in animal models such as mouse and rat [21e26] and in human studies [27]. However, NMR detects less than 100 metabolites in total. GC-MS and LC-MS are more sensitive, but only a few studies of metabolomic profiling of urine and feces in parallel have been reported. Cauchi et al. reported the use of GC-MS to profile blood, breath, feces and urine samples of individuals with gastrointestinal diseases and healthy controls, and found only the metabolic data from the fecal samples could discriminate between Crohn's disease and healthy controls [28]. LC-MS was applied to analyze mouse urine and fecal metabolomes in a study of how the intake of prebiotic fibres affects intestinal microbiota in a diet-induced obesity animal model [29] and in a study involving a humanized mouse model [13]. To our knowledge, there is no report of quantitative comparison of the human urinary and fecal metabolomes with a relatively large metabolome coverage (i.e., thousands of metabolites). In this work, we report a method based on the highperformance chemical isotope labeling (CIL) LC-MS platform for parallel analysis of urine and fecal samples. We have recently shown that in a dansyl labeled human urine sample, more than 1600 unique peak pairs or putative metabolites can be detected in a 25-min LC-MS run [30], while in a dansyl labeled human fecal sample, more than 1700 peak pairs can be found [12]. In this work, we demonstrate that differential isotope labeling of pooled urine and pooled feces followed by LC-MS analysis can be used to relate the concentrations of common metabolites found in the two biospecimen, allowing both qualitative and quantitative comparisons of the two metabolomes. We illustrate the metabolomes of urine and feces are quite different, justifying the parallel profiling to increase the overall human metabolome coverage. Finally, we apply this method to compare the human urine and fecal metabolomes of healthy individuals in a pilot study which could be expanded in the future for discovering disease biomarkers based on parallel profiling of urine and feces.

2.1. Sample collection and processing All samples were collected in compliance with prevailing human research ethics guidelines. The study protocol was approved by the Medical Ethics Committee of the 1st Affiliated Hospital of Zhejiang University, Hangzhou, China and the Ethics Approval Board of the University of Alberta, Edmonton, Canada. In this study, urine and fecal samples were collected daily from 3 female healthy volunteers (marked as W, J and S) in 7 separate days. These individuals were chosen randomly. Urine samples were collected twice in a day. One was the 2nd-void morning urine after overnight fasting (12 h), denoted as 2nd-void urine. The other one was urine collected at the same time when the fecal sample was collected which is denoted as urine@feces. The fecal sample was stored immediately after collection in a 20  C freezer while urine sample was processed as follows. Within 1 h of urine collection, the urine sample was centrifuged at 4000 rpm for 10 min, after which the supernatant was filtered twice through 0.22 mM filter (Agela Technologies, China). The filtered urine was aliquoted into 1.5 mL vials and stored at 20  C. For fecal metabolite extraction, fecal samples were subjected to sequential solvent extraction by water and acetonitrile (ACN) as descried before [12]. Briefly, 600 mL water was added into each sample for the first extraction, followed by using 600 mL ACN for the 2nd extraction. The supernatants from the two extractions were combined and dried with a SpeedVac and then stored at 20  C for further use. 2.2. Derivatization, normalization and mixing The synthesis of 13C2-dansyl chloride has been reported [31] and the dansyl labeling reagents are available from the University of Alberta (MCID.chem.ualberta.ca). Dansylation labeling of fecal and urine samples was done as reported [12,32]. The individual sample was labeled separately using 12C2-dansyl chloride and quantified by LC-UV based on absorption at 338 nm [33]. A pooled fecal sample (i.e., fecal-pool) was prepared by mixing the same amount of aliquot from each of the 63 extracts (i.e., triplicate extractions of 21 samples) and a pooled urine sample (i.e., urine-pool) was prepared from 126 extracts. A portion of the fecal-pool or urine-pool was taken and labeled by 13C-dansylation. An aliquot of the 13C-labeled pool was mixed with a 12C-labeled individual sample in 1:1 molar ratio to produce a mixture for LC-MS analysis. For the comparison of fecal and urine metabolomes, a small portion of the urine-pool or fecal pool was separately labeled using 12 C-dansylation and the total concentration of the labeled metabolites was determined by LC-UV. Another pooled sample (i.e., urine-fecal-pool) was prepared by mixing the same amount of the urine-pool and fecal-pool, and labeled using 13C-dansylation. According to the UV quantification results, an appropriate volume of 13 C-labeled urine-fecal-pool was combined with 12C-labeled urinepool or fecal-pool in 1:1 molar ratio to generate a pool-to-pool mixture for LC-MS for analysis.

102

X. Su et al. / Analytica Chimica Acta 903 (2016) 100e109

2.3. LC-MS An Agilent 1290 series binary UPLC system with an Waters ACQUITY UPLC BEH C18 column (2.1 mm  10 cm, 1.7 mm particle size, 130 Å pore size) connected to an Agilent electrospray ionization (ESI) time-of-flight mass spectrometer (Model 6230, Agilent, Palo Alto, CA, USA) was used for LC-MS analysis. For the TOF instrument, the ion source conditions were: nitrogen nebulizer gas: 1.38 Bar, dry gas flow: 5 L/min, dry temperature: 325  C, capillary voltage: 4000 V, end plate offset: 120 V, mass range: m/z up to 1700, and spectra rate: 1 Hz. The resolving power of the instrument was typically about 11,000 (FWHM) at m/z 622. All MS spectra were obtained in the positive ion mode. For LC-MS, LC solvent A was 0.1% (v/v) formic acid in water, and solvent B was 0.1% (v/v) formic acid in ACN. The gradient elution profile was as follows: t ¼ 0 min, 15% B; t ¼ 2 min, 15% B; t ¼ 15 min, 45% B; t ¼ 20 min, 65% B; t ¼ 26 min, 98% B; t ¼ 29 min, 98% B; t ¼ 29.1 min, 15% B. The flow rate was 250 mL/min. The sample injection volume varied, depending on the applications.

(A)

Individual Urine or Fecal Samples

Raw samples 1. Aliquoting equal amounts of the raw samples 2. 13C-Dansyl labeling

13C-pool

Normalized 12C-samples Mixing the same amount

12C-sample/ 13C-pool

2.4. Data processing The MS data were internally mass-calibrated and then processed using a peak-pair picking software, IsoMS [34]. The level 1 peak pairs, along with their peak intensity ratios, were aligned from multiple runs by retention time within 20 s and accurate mass within 10 ppm using IsoMS-Align to generate the initial metabolite-intensity table. The Zero-fill program was used to find the missing ratios in the table from the raw LC-MS peak list and then fill in these values to produce the final table [35]. Iso-Quant was finally applied to calculate the individual peak ratio based on chromatographic peak areas of the 12C- and 13C-labeled peaks with a peak pair in the table [36]. The ratio values were used for statistical analysis. 2.5. Statistical analysis and database search For statistical analysis of the urine or fecal samples, only the common peak-pairs shown up in at least 50% of the samples were retained for analysis. Principle component analysis (PCA) and orthogonal partial least square discriminant analysis (OPLS-DA) were performed using SIMCA-Pþ 12.0 (Umetrics, Umeå, Sweden). The data were mean-centered and pareto-scaled (unit variance) prior to analysis. Positive metabolite identification was performed based on mass and retention time match to the dansyl standard library containing 273 unique amines/phenols using DnsID [37]. Putative identification was done based on accurate mass match to the metabolites in the human metabolome database (HMDB) (8021 known human endogenous metabolites) and the Evidence-based Metabolome Library (EML) (375,809 predicted human metabolites with one reaction) using MyCompoundID [38]. The mass accuracy tolerance window was set at 10 ppm for database search. 3. Results and discussion 3.1. Overall workflow Fig. 1 shows the workflow for parallel fecal and urine metabolome profiling using CIL LC-MS. Urine was directly analyzed while feces were subjected to solvent extraction. The 12C-dansyl labeled individual samples were separately injected into LC-UV to measure the total concentration of labeled metabolites in each sample for sample amount normalization. Based on the total concentration, an appropriate volume of an individual unlabeled sample was taken to

1. Metabolite extraction (for feces) 2. 12C-Dansyl labeling 3. LC-UV for measuring total conc. of labeled metabolites 4. Taking aliquot with equal amount

(B)

1. LC-MS 2. Data processing & statistical analysis

Urine and Fecal Metabolome Comparison

12C-urine-pool

12C-urine-pool/ 13C-urine-fecal-pool

13C-urine-fecal-pool

Mixing 12C C-fecal-pool feca

12C-fecal-pool/ 13C-urine-fecal-pool

1. LC-MS 2. Data analysis Urine and fecal metabolome comparison

Fig. 1. Workflow of the differential chemical isotope labeling LC-MS method for parallel fecal and urine metabolomic profiling: (A) procedure for analyzing individual samples within a group (urine or feces) and (B) procedure for comparing the metabolomes of urine and fecal samples.

mix with an equal amount of other unlabeled samples to generate a pooled fecal sample (i.e., fecal-pool from 63 samples) or a pooled urine sample (i.e., urine-pool from 126 samples). The pooled sample was then labeled by 13C-dansylation which served as a reference or internal standard for the 12C-labeled samples. An equal amount of the 12C-labeled individual sample and the 13C-labeled pooled sample was mixed. A quality control (QC) sample was prepared by mixing an equal amount of the 12C-labeled and 13Clabeled pooled samples. The mixtures of 13C-pool and 12C-sample were analyzed by LC-TOF-MS. After peak pair extraction and peak ratio calculation, a metabolite-intensity table is produced in which the peak ratio values (12C2-peak vs. 13C2-peak) for a given metabolite peak pair in all individual samples reflect the relative concentration differences of the metabolite in these samples. The quantitative metabolome tables generated from all the urine samples or all the feces were used for data comparison and statistical analysis. To compare the two quantitative metabolome datasets, as Fig. 1B shows, aliquots of the urine-pool and the fecal-pool were separately labeled using 12C-dansylation, followed by LC-UV measurement of the total concentration of labeled metabolites in these two pooled samples. A new pooled sample (urine-fecal-pool) was prepared by mixing the same amount of the fecal-pool and the urine-pool. The urine-fecal-pool was labeled with 13C-dansylation to serve as a reference for 12C-labeled fecal-pool or urine-pool. The above experiment was repeated three times (i.e., experimental triplicate) and each mixture of 13C-labeled urine-fecal-pool and

X. Su et al. / Analytica Chimica Acta 903 (2016) 100e109

W-2nd J-2nd S-2nd W-F J-F S-F QC

(A) 40 20 0

t[2]

12 C-labeled urine-pool or fecal-pool was injected into LC-MS three times (i.e., injection triplicate). Thus a total of 18 LC-MS runs were carried out. The average intensity ratio of an individual metabolite peak pair i from 9 runs of the urine-pool/urine-fecal-pool mixtures and 9 runs of the fecal-pool/urine-fecal-pool mixtures was calculated to give Ui(ave) or Fi(ave), respectively. These values can be used to bridge the fecal and urine metabolomic datasets (see below).

103

-20 -40

3.2. Urine metabolome

-60 -40

(B)

-20

0

20

t[1]

40

W J S

30 20

t[2]

10 0 -10 -20 -30 -50 -40 -30 -20 -10 0 10 20 30 40 50

t[1]

(C) 30

W-2nd J-2nd S-2nd W-F J-F S-F

20

t[2]

10 0 -10 -20 -30

(D)

-50 -40 -30 -20 -10 0 10 20 30 40 50 t[1] 40

1 2

30 20 to[1]

10 0 -10 -20 -30 -40 -25 -20 -15 -10

-5

0

5

10

15

20

25

t[1]

(E) 40 t[2]

In total, 126 12C-/13C-mixtures were produced from triplicate experiments of 42 urine samples by using the workflow shown in Fig. 1A. These mixtures were individually analyzed by LC-MS and 12 injections of the QC sample spaced evenly among the 126 sample injections were also done. Fig. 2A shows the PCA plot of the 138 runs. The QC data are clustered together, indicating excellent analytical reproducibility in LC-MS data acquisition. Some separations of the three individuals are already visible. Fig. 2B shows the OPLS-DA plot of the urine dataset according to three individuals. These three individuals are clearly separated (R2X ¼ 0.369, R2Y ¼ 0.952 and Q2 ¼ 0.944). Since each person collected two types of urine samples per day (i.e., 2nd-void and urine@feces), the 126 metabolomic data were further classified into six groups and analyzed by OPLS-DA. The result is shown in Fig. 2C. The person-toperson separation of the metabolomic data is still significant (R2X ¼ 0.579, R2Y ¼ 0.938 and Q2 ¼ 0.880). However, there are some separations between the two types of urine samples in each person, although these separations are much smaller than the person-to-person separations. Fig. 2D shows the OPLS-DA plot of the 126 metabolomic data grouped according to two sample types. The separation of the two groups is very clear (R2X ¼ 0.459, R2Y ¼ 0.980 and Q2 ¼ 0.938). The day-to-day separation mainly caused by diet effect for an individual is also visible. For example, Fig. 2E shows the OPLS-DA plot of seven separated clusters for seven different days of urine samples from individual S (R2X ¼ 0.649, R2Y ¼ 0.938 and Q2 ¼ 0.809). In this case, the separation of the two types of urine samples within a day is not clear. The above results indicate that separation between individuals is much greater than the type of urine collected within a day or dayto-day variations within an individual. Although the sample size (three individuals) was small, we collected the samples over 7 days without diet control, which should be reflective of the current practice of urine sample collection for most biomarker discovery studies where diet is not controlled. The results shown in Fig. 2 demonstrate that our method of profiling the amine/phenol submetabolome using dansylation CIL LC-MS can reveal even minor differences among different comparative samples. Supplemental Table T1 lists the 632 significant metabolites that provide the separation of three individuals based on their urine metabolome profiles; these significant metabolites were determined from the OPLS-DA score plot shown in Fig. 2B with a VIP score of greater than 1.2. From the combined results of 126 runs, we detected a total of 3089 unique peak pairs or metabolites. An average of 2534 ± 113 peak pairs was found per run (see Supplemental Table T2 for the list of the peak pair numbers detected for each run). We have examined the distributions of the peak pair numbers among these samples. Fig. 3A shows the distribution of the numbers of peak pairs detected from three individuals. Out of 3089 peak pairs found in total in two types of urine, 651 pairs are detected commonly. If we consider only one type of urine (i.e., 2nd-void in Fig. 3B and urine@feces in Fig. 3C), more common peak pairs are detected (938 common pairs in 2nd-void urine and 753 common pairs in urine@feces). If we only compare the common peak pairs found from the 2nd-void urine (938 pairs) and urine@feces (753 pairs), a

S-1 S-2 S-3 S-4 S-5 S-6 S-7

30 20 10 0 -10 -20 -30 -40 -50 -40 -30 -20 -10 0 10 20 30 40 50

t[1]

Fig. 2. (A) PCA plot of all urine data obtained from the 138 LC-MS runs. The data points of individuals are color-coded to show the data distribution. OPLS-DA score plots of (B) three individuals, (C) six groups, (D) two types of urine (1 ¼ 2nd-void urine and 2 ¼ urine@feces) and (E) 7 daily groups from S.

total of 1040 peak pairs are detected (see Fig. 3D). Among them, 651 (62.6%) pairs are commonly detected and most of the pairs (86.5%) from urine@feces are found in the 2nd-void urine. These results indicate that despite metabolome composition variations among different individuals and among samples collected from different

104

X. Su et al. / Analytica Chimica Acta 903 (2016) 100e109

DnsID to search the dansyl standards library, we positively identified 74 metabolites based on retention time and accurate mass matches (this list is shown in Supplemental Table T3e1). There were 648 and 2045 metabolites putatively identified based on accurate mass match against the HMDB database and the EML library, respectively (the lists are shown in Supplemental Tables T3e2 and T3e3, respectively). Thus, 2767 out of the 3089 peak pairs (90%) could be matched. 3.3. Fecal metabolome For the fecal metabolome profiling work, 21 samples collected from three individuals in 7 days on a daily basis were analyzed in experimental triplicate. Seven QC injections were performed. The PCA score plot of the 70-run dataset is shown in Fig. 4A. Although the data have some overlaps among the three individuals, some separation is visible. Fig. 4B shows the OPLS-DA plot of the data. The individuals are clearly separated (R2X ¼ 0.389, R2Y ¼ 0.989 and Q2 ¼ 0.972). The day-to-day separation likely caused by diet effect is visible for individuals, as it is shown in the OPLS-DA plot in Fig. 4C using individual S as an example. The triplicate experimental data are clustered together, indicating good experimental reproducibility. The results shown in Fig. 4 indicate that the variations among individuals are greater than day-to-day variations within an individual. We have determined the significant metabolites that

(A)

QC W J S

40

t[2]

20 0 -20 -40 -40 -30 -20 -10

(B)

0

10 20 30 40 50

t[1] W J S

30 20

t[2]

10 0 -10 -20 -30 -40 -30 -20 -10

(C)

0

10

20

30

40

t[1]

40

t[2]

20

Fig. 3. Venn diagrams of peak pair numbers detected from urine of three individuals: (A) all common peak pairs from three individuals, (B) common peak pairs from the 2nd-void urine, (C) common peak pairs from urine@feces, and (D) common peak pairs from two types of urine.

days and times, there are still a greater number of common metabolites found in all the samples. Metabolite identification for the 3089 peak pairs detected from urine was carried out using two levels of database search. Using

0

S-1 S-2 S-3 S-4 S-5 S-6 S-7

-20 -40 -60 -50 -40 -30 -20 -10 0 10 20 30 40 50 60 t[1] Fig. 4. (A) PCA plot of all feces data obtained from the 70 LC-MS runs. The data points of individuals are color-coded to show the data distribution. OPLS-DA score plots of (B) three individuals and (C) 7 daily groups from S.

X. Su et al. / Analytica Chimica Acta 903 (2016) 100e109

separate the three individuals based on OPLS-DA analysis and the list with VIP score of greater than 1.2 is shown in Supplemental Table T4. From the 63 LC-MS runs of the fecal samples, we detected a total of 3012 unique peak pairs with an average of 2507 pairs per run. Supplemental Table T5 lists the number of peak pairs detected in each run. Fig. 5 shows the distributions of the peak pair numbers detected according to the individuals. Out of 3012 peak pairs found in total, 1150, 1246 and 1214 pairs are detected commonly from individual W, J and S, respectively. 704 pairs are detected commonly from all three individuals. Despite the complexity of fecal samples, our results indicate that many common metabolites can be detected from individual samples, as in the case of urine metabolome profiling. We have also identified some of the peak pairs detected in fecal samples and the identification results are shown in Supplemental Tables T3e4 to T3e6. Among the 3012 peak pairs, 64 could be positively identified using the dansyl standards library. 420 and 1739 were putatively identified using the accurate mass search against the HMDB database and the EML library with one reaction, respectively. Thus, a total of 2223 pairs out of the 3012 peak pairs (74%) were matched to the metabolites in the databases. Although the number of peak pairs or metabolites detected in feces is similar to that of urine, the percentage of matched metabolites is lower than the urine (90%). Since only the human metabolite databases were searched, feces appear to contain a lot more non-endogenous human metabolites, compared to urine. This finding is not surprising as one would expect that feces contain microbiome metabolites and residual small molecules from diets. The fact that we could see the compositional differences between urine and feces is re-assuring on the quality of our metabolomic data; the large metabolomic coverage by dansylation LC-MS offers the possibility of detecting the differences.

105

untargeted methods such as NMR and conventional LC-MS which detect all groups of compounds, but with lower coverage in each group. First of all, we compare the peak pair numbers detected in daily samples of 2nd-void urine, urine@feces and feces from three

2nd-void urine

urine@feces

(A)

1239

198 66

572

181 63

1568 feces

2nd-void urine

urine@feces

(B)

1406

181

3.4. Qualitative comparison of urine and fecal datasets

90

The human urine and fecal metabolomes are thought to be different, but the extent of the difference has not been examined thoroughly. As discussed above, our method affords the detection of 3089 peak pairs or metabolites from urine and 3012 pairs from feces. Thus we can compare the amine/phenol submetabolomes of urine and feces in a more comprehensive manner than other

589

258 91

1394 feces

2nd-void urine

W

177 142

127 704

urine@feces

(C)

222

J

193

175 S Fig. 5. Venn diagram of common peak pairs detected from fecal samples of three individuals.

1271

225

83

583

173

72

1435 feces Fig. 6. Venn diagrams of peak pair numbers from daily samples of three individual: (A) W, (B) J and (C) S.

106

X. Su et al. / Analytica Chimica Acta 903 (2016) 100e109

individuals. Fig. 6AeC shows the distributions of the peak pair numbers found in Day 1 from individual W, J and S, respectively. Similar distributions were found for Days 2e7 (data not shown). The two types of urine samples collected in a day have many common peak pairs. For example, in Fig. 6A, out of a total of 2319 pairs found in the 2nd-void urine and urine@feces, 1811 pairs (79%) are in common from individual W. In contrast, there are many more peak pairs uniquely detected in urine when compared to the pairs detected in feces. For example, for individual W, 1568 pairs are unique in feces, compared to 1437 and 1420 pairs are unique in the 2nd-void and urine@feces, respectively. These comparison results clearly show that there is a great difference in the metabolome compositions of urine and feces. Next we compare the common peak pairs found in 7 daily samples in urine and feces. The number distributions are shown in Fig. 7AeC for the three individuals. As in the case of daily sample comparison shown in Fig. 6, there are many more unique peak pairs found in urine or feces than the common ones, further suggesting that the metabolite compositions of the urine and feces are quite different. Finally, we compare the combined common peak pairs found from all three individuals in the 2nd-void urine, urine@feces and feces. The number distribution is shown in Fig. 8A. A total of 1498 pairs are detected and, among them, only 155 (10%) pairs are commonly detected. If we combine the two urine data into one and compare it to the pairs found in feces (see Fig. 8B), the 155 common pairs found in urine and feces represent 13% of the 1165 pairs detected in total. If we only compare the 2nd-void urine and feces (Fig. 8C), a total of 1408 pairs are found in the two samples and only 199 (14%) pairs are in common. Similarly, comparing the urine@feces with feces (Fig. 8D), a total of 1255 pairs are detected and only 167 (13%) pairs are in common. These results indicate that only about 13e14% of the metabolites found in urine and fecal samples collected daily from three individuals over a period of 7 days are in common. Moreover, urine collected at the same time as feces collection does not share more common metabolites than the 2ndvoid. It is clear that fecal metabolome is very different from urine metabolome. Even when the common pairs are used to generate the OPLS-DA plots, there are clear separations among the different urine and fecal samples in the plots (see Supplemental Fig. S1).

2nd-void urine

(A)

658

321 70

250

142 14

773 feces urine@feces

2nd-void urine

(B)

801

385 84

255

138 16

850 feces

2nd-void

urine

urine@feces

(C)

950

205

3.5. Quantitative metabolome comparison As it is shown in Fig. 8B, when we combine all the metabolome data, there are 155 peak pairs commonly detected in two types of urine and feces. We focused on these common metabolites to determine their relative concentration differences in urine and feces. In order to relate the two biospecimens quantitatively, we used the workflow shown in Fig. 1B to prepare the 13C-labeled urine-fecal-pool, 12C-labeled urine-pool and 12C-labeled fecal-pool in triplicate (i.e., experimental triplicate). Each mixture of 12Clabeled urine-pool or fecal-pool and the 13C-labeled urine-fecalpool in equal amount was analyzed three times by LC-MS (i.e., injection triplicate). We then searched the metabolite-intensity table generated from the 18 LC-MS runs to find the 155 peak pairs of interest. Of these, 140 pairs were found and the remaining 15 pairs were missing. The missing ones were likely due to the concentration changes of these metabolites in the urine-fecal-pool to a level that was not detected in the mixture. Out of the 140 common pairs found in the 18 runs, we calculated the average ratio of an individual metabolite in urine [Ui(ave)] or feces [Fi(ave)] from the 9 LC-MS runs of the mixtures of urine-pool or fecal-pool and urine-fecalpool. Supplemental Table T6 shows the results for the 140 pairs. Fig. 9A shows the plot of the number of peak pairs as a function of log2[Ui(ave)/Fi(ave)]. The average log2[Ui(ave)/Fi(ave)] value is 0.443.

urine@feces

34

288

123 27

824 feces Fig. 7. Venn diagrams of peak pair numbers from 7 daily samples of three individual: (A) W, (B) J and (C) S.

The ratio ranges from 0.0116 (i.e., 86-fold more concentrated in feces than in urine) to 44.3 (i.e., 44-fold more concentrated in urine than in feces) with an overall dynamic range of 3784-fold. Among the 140 common pairs, 39 pairs could be positively identified using the dansyl standards library. Table 1 lists the average relative concentrations of these metabolites in the urinepool, Ui(ave), and the fecal-pool, Fi(ave), referenced to the same urine-fecal-pool. For example, for the top row in Table 1, glutamine, the average relative concentration of this metabolite referenced to the urine-fecal-pool is 1.35 ± 0.01 (n ¼ 9) in the urine-pool and

X. Su et al. / Analytica Chimica Acta 903 (2016) 100e109

Number of peak pairs

60

107

(A)

50 40 30 20 10

0

Peak ratio referenced to urine-pool

-6 -5 -4 -3 -2 -1 1 2 3 log2[Ui(ave)/Fi(ave)] 1024

4

5

6

(B)

256

F

64 16 4 1

U

U

F

U

1/4 1/16 1/64 1/256

F

1,4-Diaminobutane 5-Hydroxylysine L-Histidine Fig. 9. (A) Number of peak pairs as a function of their corresponding peak ratio values for the 140 metabolites commonly found in the urine-pool and the fecal-pool. (B) Distribution of peak pair ratios of three metabolites referenced to the urine-pool in all the individual urine and fecal samples.

Fig. 8. Venn diagrams of peak pair numbers from (A) all pairs of three individuals, (B) pairs from two types of urine and feces, (C) pairs between 2nd-void urine and feces, and (D) pairs between urine@feces and feces.

0.373 ± 0.005 (n ¼ 9) in the fecal-pool, which indicates that glutamine is 3.62-fold more concentrated in urine than in feces. Thus, the Ui(ave)/Fi(ave) ratio in Table 1 can be used to directly compare the concentration difference of a given metabolite commonly found in urine and feces. We can extend this pool-to-pool comparison to compare the

relative concentration of a metabolite in all the individual samples of the two biospecimen. This is done by normalizing the relative concentration values or peak ratio values of a metabolite in individual samples referenced to urine-pool or fecal-pool to one type of specimen (e.g., urine-pool only). We apply Fi/[Ui(ave)/Fi(ave)] (Fi is the peak ratio in an individual fecal sample) to all individual fecal samples to calculate the peak ratios that are normalized to the pooled urine, instead of the pooled feces. In this way, we can directly examine and compare the ratio distributions of both urine and feces in a combined metabolomic dataset. As an example, Fig. 9B shows the individual ratio distributions of three metabolites normalized to urine-pool. In the case of 1,4-diaminobutane, the relative concentrations of this metabolite in individual urine samples (in blue) referenced to the urine-pool are clustered around 1. In comparison, as Table 1 indicates, the concentration of this metabolite in the urine-pool is much lower than that in the fecal-pool with Ui(ave)/Fi(ave) ¼ 0.0116. Thus this metabolite's relative concentrations in all individual fecal samples referenced to the urine-pool (in red in Fig. 9B) are distributed and centered around 45. There is also a wider concentration distribution for this metabolite in feces than in urine. For 5-hydroxylysine, the relative concentrations of urine samples referenced to the urine-pool are distributed around 1. According to Table 1, Ui(ave)/Fi(ave) is 1.26, indicating that the concentration of 5-hydroxylysine in the urine-pool is only 1.26-fold higher than that in the fecal-pool. Thus this metabolite in individual fecal samples have the relative concentrations referenced to the urine-pool distributed around 1 and the distributions for the urine

108

X. Su et al. / Analytica Chimica Acta 903 (2016) 100e109

Table 1 Average relative concentrations of identified common metabolites in the urine-pool, Ui(ave), and the fecal-pool, Fi(ave), referenced to the urine-fecal-pool. HMDB No.

Metabolite Name

Ui(ave)±std (n ¼ 9)

Fi(ave) ±std (n ¼ 9)

Ui(ave)/Fi(ave)

log2[Ui(ave)/Fi(ave)]

HMDB00641 HMDB00187 HMDB00167 HMDB00149 HMDB00123 HMDB00446 HMDB00056 HMDB00161 HMDB00112 HMDB01906 HMDB00292 HMDB00452 HMDB00650 HMDB00157_3 HMDB00162 HMDB00883 HMDB00696 HMDB00300 HMDB00929 HMDB00159 HMDB00172 HMDB11177 HMDB00557 HMDB00687 HMDB00301 HMDB04987 HMDB00450 HMDB00669 HMDB00440 HMDB02390 HMDB00182 HMDB00177 HMDB01414 HMDB06050 HMDB00158 HMDB01856 HMDB00152_2 HMDB00130 HMDB00306

L-Glutamine

1.35 ± 0.01 1.05 ± 0.01 0.658 ± 0.005 1.62 ± 0.01 1.42 ± 0.01 1.27 ± 0.03 0.387 ± 0.005 0.387 ± 0.005 0.30 ± 0.07 1.38 ± 0.02 0.528 ± 0.007 1.38 ± 0.02 1.38 ± 0.02 0.36 ± 0.01 0.04a 0.117 ± 0.005 0.11a 0.417 ± 0.005 0.84 ± 0.01 0.287 ± 0.005 0.08a 0.464 ± 0.007 0.08a 0.08a 0.306 ± 0.005 0.074 ± 0.009 0.92 ± 0.02 1.24 ± 0.01 1.24 ± 0.01 1.24 ± 0.01 0.383 ± 0.005 1.61 ± 0.01 0.02a 0.403 ± 0.005 0.403 ± 0.005 0.67 ± 0.09 0.67 ± 0.09 0.179 ± 0.008 1.01 ± 0.01

0.373 ± 0.005 0.73 ± 0.01 1.04 ± 0.02 0.193 ± 0.005 0.363 ± 0.005 0.570 ± 0.009 1.32 ± 0.02 1.32 ± 0.02 1.40 ± 0.02 0.28 ± 0.02 1.23 ± 0.02 0.28 ± 0.02 0.28 ± 0.02 1.21 ± 0.04 1.62 ± 0.02 1.54 ± 0.03 1.44 ± 0.03 1.33 ± 0.02 0.79 ± 0.02 1.36 ± 0.03 1.58 ± 0.04 1.24 ± 0.02 1.58 ± 0.04 1.58 ± 0.04 1.08 ± 0.01 1.68 ± 0.02 0.73 ± 0.02 0.497 ± 0.009 0.497 ± 0.009 0.497 ± 0.009 1.35 ± 0.03 0.04a 1.73 ± 0.02 1.35 ± 0.03 1.35 ± 0.03 1.56 ± 0.02 1.56 ± 0.02 1.53 ± 0.03 0.89 ± 0.02

3.62 1.44 0.633 8.39 3.91 2.23 0.293 0.293 0.214 4.93 0.429 4.93 4.93 0.298 0.0247 0.0760 0.0764 0.311 1.06 0.211 0.0506 0.374 0.0506 0.0506 0.283 0.0440 1.26 2.50 2.50 2.50 0.284 40.3 0.0116 0.299 0.299 0.430 0.430 0.117 1.13

1.856 0.526 0.660 3.069 1.967 1.157 1.771 1.771 2.224 2.302 1.221 2.302 2.302 1.747 5.339 3.718 3.710 1.685 0.0841 2.245 4.305 1.419 4.305 4.305 1.821 4.506 0.333 1.322 1.322 1.322 1.816 5.333 6.430 1.742 1.742 1.218 1.218 3.095 0.176

a

L-Serine L-Threonine Ethanolamine Glycine N-Alpha-acetyllysine Beta-Alanine L-Alanine Gamma-Aminobutyric acid 2-Aminoisobutyric acid Xanthine L-Alpha-aminobutyric acid D-Alpha-aminobutyric acid Hypoxanthine - Isomer L-Proline L-Valine L-Methionine Uracil L-Tryptophan L-Phenylalanine L-Isoleucine L-phenylalanyl-L-proline L-Alloisoleucine L-leucine Urocanic acid Alpha-Aspartyl-lysine 5-Hydroxylysine Ortho-Hydroxyphenylacetic acid 3-Hydroxyphenylacetic acid 3-Cresotinic acid L-Lysine L-Histidine 1,4-diaminobutane o-Tyrosine L-Tyrosine Protocatechuic acid Gentisic acid -multi-tags Homogentisic acid Tyramine

This ratio was near the limit of relative quantification and thus the same value was retained for all the samples (std was zero).

and feces are similar as shown in Fig. 9B; there are a few outliers with very low concentrations in feces. For histidine, according to Table 1, Ui(ave)/Fi(ave) is 40.3 and thus the concentration of this metabolite in urine is much higher than in feces. Histidine concentrations in urine samples referenced to the urine-pool cluster around 1, while its concentrations in fecal samples referenced to the urine-pool distribute around 1/40 in Fig. 9. The concentration distribution ranges are similar for the two types of samples. The results shown in Fig. 9B illustrate that our method of determining Ui(ave)/Fi(ave) can be used to bridge the urine and fecal metabolomic datasets quantitatively, allowing examining the relative concentration differences of individual metabolites found in both biospecimens. This information may be very useful in untargeted metabolomics for biomarker discovery research where not only the concentration changes of a potential metabolite biomarker in either urine or feces, but also the concentration ratios of the metabolite in urine vs. feces, can be used to correlate with a phenotype change such as disease development or progression. We note that although only relative concentration differences are shown in this work, absolute concentrations of individual metabolites could be determined, if needed, by spiking known concentrations of labeled standards into the pooled samples.

profiling of urine and fecal metabolomes and offer both qualitative and quantitative comparisons of the two biospecimens. This method increased the overall human metabolome coverage; 5372 unique peak pairs were detected from all the samples combined and, among them, 3089 and 3012 pairs were found in urine and feces, respectively. We have shown that the metabolomes of urine and feces are quite different, and our method of determining Ui(ave)/ Fi(ave) can be used to relate the two biospecimens quantitatively. Acknowledgments This collaborative research was in part supported by a visiting professorship to L. Li by Zhejiang University K. P. Chao's Hi-Tech Foundation for Scholars and Scientists. Additional support was provided by the Natural Sciences and Engineering Research Council of Canada, Canadian Institutes of Health Research, Canada Research Chairs, Genome Canada and Alberta Innovates (to L. Li), and the Independent Foundation of the State Key Laboratory for Diagnosis and Treatment of Infectious Diseases and National Science and Technology Major Project of China (2013ZX10004101-007) and Zhejiang Public Welfare Project, China (2014C37040) (to L.J. Li). Appendix. Supplementary data

4. Conclusions We have developed a method to perform parallel quantitative

Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.aca.2015.11.027.

X. Su et al. / Analytica Chimica Acta 903 (2016) 100e109

References [1] E.A. Thevenot, A. Roux, Y. Xu, E. Ezan, C. Junot, Analysis of the human adult urinary metabolome variations with age, body mass index, and gender by implementing a comprehensive workflow for univariate and OPLS statistical analyses, J. Proteome Res. 14 (2015) 3322e3335. [2] C.C. Jacob, G. Dervilly-Pinel, G. Biancotto, F. Monteau, B. Le Bizec, Global urine fingerprinting by LC-ESI(þ)-HRMS for better characterization of metabolic pathway disruption upon anabolic practices in bovine, Metabolomics 11 (2015) 184e197. [3] O.A.H. Jones, S. Murfitt, C. Svendsen, A. Turk, H. Turk, D.J. Spurgeon, L.A. Walker, R.F. Shore, S.M. Long, J.L. Griffin, Comparisons of metabolic and physiological changes in rats following short term oral dosing with pesticides commonly found in food, Food Chem. Toxicol. 59 (2013) 438e445. [4] M.H. Hanna, P.D. Brophy, Metabolomics in pediatric nephrology: emerging concepts, Pediatr. Nephrol. 30 (2015) 881e887. [5] C.C. Huang, B.S. Lou, F.L. Hsu, C.C. Hou, Use of urinary metabolomics to evaluate the effect of hyperuricemia on the kidney, Food Chem. Toxicol. 74 (2015) 35e44. [6] K. Sharma, Obesity, oxidative stress, and fibrosis in chronic kidney disease, Kidney Int. Suppl. 4 (2014) 113e117. [7] M.J. Pena, H.J.L. Heerspink, M.E. Hellemons, T. Friedrich, G. Dallmann, M. Lajer, S.J.L. Bakker, R.T. Gansevoort, P. Rossing, D. de Zeeuw, S.S. Roscioni, Urine and plasma metabolites predict the development of diabetic nephropathy in individuals with Type 2 diabetes mellitus, Diabet. Med. 31 (2014) 1138e1147. [8] N.A. Yousri, D.O. Mook-Kanamori, M.M.E. Selim, A.H. Takiddin, H. Al-Homsi, K.A.S. Al-Mahmoud, E.D. Karoly, J. Krumsiek, K.T. Do, U. Neumaier, M.J. MookKanamori, J. Rowe, O.M. Chidiac, C. McKeon, W.A. Al Muftah, S.A. Kader, G. Kastenmuller, K. Suhre, A systems view of type 2 diabetes-associated metabolic perturbations in saliva, blood and urine at different timescales of glycaemic control, Diabetologia 58 (2015) 1855e1867. [9] D.B. Liesenfeld, N. Habermann, R. Toth, R.W. Owen, E. Frei, J. Bohm, P. SchrotzKing, K.D. Klika, C.M. Ulrich, Changes in urinary metabolic profiles of colorectal cancer patients enrolled in a prospective cohort study (ColoCare), Metabolomics 11 (2015) 998e1012. [10] H.I. Wettersten, A.A. Hakimi, D. Morin, C. Bianchi, M. Johnstone, D.R. Donohoe, J.F. Trott, O. Abu Aboud, S. Stirdivant, B. Neri, R. Wolfert, B. Stewart, R. Perego, J.J. Hsieh, R.H. Weiss, Grade-dependent metabolic reprogramming in kidney cancer revealed by combined proteomics and metabolomics analysis, Cancer Res. 75 (2015) 2541e2552. [11] J. Peng, Y.T. Chen, C.L. Chen, L. Li, Development of a universal metabolomestandard method for long-term LC-MS metabolome profiling and its application for bladder cancer urine-metabolite-biomarker discovery, Anal. Chem. 86 (2014) 6540e6547. [12] W. Xu, D.Y. Chen, N. Wang, T. Zhang, R.K. Zhou, T. Huan, Y.F. Lu, X.L. Su, Q. Xie, L. Li, L.J. Li, Development of high-performance chemical isotope labeling LCMS for profiling the human fecal metabolome, Anal. Chem. 87 (2015) 829e836. [13] A. Marcobal, P.C. Kashyap, T.A. Nelson, P.A. Aronov, M.S. Donia, A. Spormann, M.A. Fischbach, J.L. Sonnenburg, A metabolomic view of how the human gut microbiota impacts the host metabolome using humanized and gnotobiotic mice, ISME J. 7 (2013) 1933e1943. [14] M.S. Donia, M.A. Fischbach, Small molecules from the human microbiota, Science 349 (2015) 395. [15] D.M. Jacobs, E. Gaudier, J. van Duynhoven, E.E. Vaughan, Non-digestible food ingredients, colonic microbiota and the impact on gut health and immunity: a role for metabolomics, Curr. Drug Metab. 10 (2009) 41e54. [16] J. Kuczynski, C.L. Lauber, W.A. Walters, L.W. Parfrey, J.C. Clemente, D. Gevers, R. Knight, Experimental and analytical tools for studying the human microbiome, Nat. Rev. Genet. 13 (2012) 47e58. [17] J.M. Kinross, A.W. Darzi, J.K. Nicholson, Gut microbiome-host interactions in health and disease, Genome Med. 3 (2011) 14. [18] O. Goulet, Potential role of the intestinal microbiota in programming health and disease, Nutr. Rev. 73 (2015) 32e40. [19] L. Biedermann, G. Rogler, The intestinal microbiota: its role in health and disease, Eur. J. Pediatr. 174 (2015) 151e167. [20] J.K. Nicholson, E. Holmes, J. Kinross, R. Burcelin, G. Gibson, W. Jia, S. Pettersson,

109

Host-gut microbiota metabolic interactions, Science 336 (2012) 1262e1267. [21] L.E. Romick-Rosendale, A.M. Goodpaster, P.J. Hanwright, N.B. Patel, E.T. Wheeler, D.L. Chona, M.A. Kennedy, NMR-based metabonomics analysis of mouse urine and fecal extracts following oral treatment with the broadspectrum antibiotic enrofloxacin (Baytril), Magn. Reson. Chem. 47 (2009) S36eS46. [22] Z.X. Miao, M.X. Jin, X. Liu, W. Guo, X.J. Jin, H.Y. Liu, Y.H. Wang, The application of HPLC and microprobe NMR spectroscopy in the identification of metabolites in complex biological matrices, Anal. Bioanal. Chem. 407 (2015) 3405e3416. [23] C.C. Yde, M.R. Clausen, D.B. Ditlev, H. Lillefosse, L. Madsen, K. Kristiansen, B. Liaset, H.C. Bertram, Multi-block PCA and multi-compartmental study of the metabolic responses to intake of hydrolysed versus intact casein in C57BL/6J mice by NMR-based metabolomics, Metabolomics 10 (2014) 938e949. [24] Y.X. Yang, L.L. Wang, S.M. Wang, R.B. Huang, L.Y. Zheng, S.W. Liang, L. Zhang, J.F. Xu, An integrated metabonomic approach to studying metabolic profiles in rat models with insulin resistance induced by high fructose, Mol. Biosyst. 10 (2014) 1803e1811. [25] R. Calvani, E. Brasili, G. Pratico, G. Capuani, A. Tomassini, F. Marini, F. Sciubba, A. Finamore, M. Roselli, E. Marzetti, A. Miccheli, Fecal and urinary NMR-based metabolomics unveil an aging signature in mice, Exp. Gerontol. 49 (2014) 5e11. [26] I.K.S. Yap, J.V. Li, J. Saric, F.P. Martin, H. Davies, Y.L. Wang, I.D. Wilson, J.K. Nicholson, J. Utzinger, J.R. Marchesi, E. Holmes, Metabonomic and microbiological analysis of the dynamic effect of vancomycin-induced gut microbiota modification in the mouse, J. Proteome Res. 7 (2008) 3718e3728. [27] T. Misawa, Y. Date, J. Kikuchi, Human metabolic, mineral, and microbiota fluctuations across daily nutritional intake visualized by a data-driven approach, J. Proteome Res. 14 (2015) 1526e1534. [28] M. Cauchi, D.P. Fowler, C. Walton, C. Turner, W.J. Jia, R.N. Whitehead, L. Griffiths, C. Dawson, H. Bai, R.H. Waring, D.B. Ramsden, J.O. Hunter, J.A. Cole, C. Bessant, Application of gas chromatography mass spectrometry (GC-MS) in conjunction with multivariate classification for the diagnosis of gastrointestinal diseases, Metabolomics 10 (2014) 1113e1120. [29] F. Respondek, P. Gerard, M. Bossis, L. Boschat, A. Bruneau, S. Rabot, A. Wagner, J.C. Martin, Short-chain fructo-oligosaccharides modulate intestinal microbiota and metabolic parameters of humanized gnotobiotic diet induced obesity mice, PLoS One 8 (2013) 12. [30] R. Zhou, L. Li, Effects of sample injection amount and time-of-flight mass spectrometric detection dynamic range on metabolome analysis by highperformance chemical isotope labeling LC-MS, J. Proteomics 118 (2015) 130e139. [31] K. Guo, L. Li, Differential 12C-/13C-isotope dansylation labeling and fast liquid chromatography/mass spectrometry for absolute and relative quantification of the metabolome, Anal. Chem. 81 (2009) 3919e3932. [32] C.L. Tseng, L. Li, High-performance isotope-labeling liquid chromatography mass spectrometry for investigating the effect of drinking Goji tea on urine metabolome profiling, Sci. CHINA Chem. 57 (2014) 678e685. [33] Y. Wu, L. Li, Determination of total concentration of chemically labeled metabolites as a means of metabolome sample normalization and sample loading optimization in mass spectrometry-based metabolomics, Anal. Chem. 84 (2012) 10723e10731. [34] R. Zhou, C.L. Tseng, T. Huan, L. Li, IsoMS: automated processing of LC-MS data generated by a chemical isotope labeling metabolomics platform, Anal. Chem. 86 (2014) 4675e4679. [35] T. Huan, L. Li, Counting missing values in a metabolite-intensity data set for measuring the analytical performance of a metabolomics platform, Anal. Chem. 87 (2015) 1306e1313. [36] T. Huan, L. Li, Quantitative metabolome analysis based on chromatographic peak reconstruction in chemical isotope labeling liquid chromatography mass spectrometry, Anal. Chem. 87 (2015) 7011e7016. [37] T. Huan, Y. Wu, C. Tang, G. Lin, L. Li, DnsID in MyCompoundID for rapid identification of dansylated Amine- and Phenol-containing metabolites in LCMS-based metabolomics, Anal. Chem. 87 (2015) 9838e9845. [38] L. Li, R. Li, J. Zhou, A. Zuniga, A.E. Stanislaus, Y. Wu, T. Huan, J. Zheng, Y. Shi, D.S. Wishart, G. Lin, MyCompoundID: using an evidence-based metabolome library for metabolite identification, Anal. Chem. 85 (2013) 3401e3408.

Dansylation isotope labeling liquid chromatography mass spectrometry for parallel profiling of human urinary and fecal submetabolomes.

Human urine and feces can be non-invasively collected for metabolomics-based disease biomarker discovery research. Because urinary and fecal metabolom...
564B Sizes 0 Downloads 12 Views