Determination of thiol metabolites in human urine by stable isotope labeling in combination with pseudo-targeted mass spectrometry analysis

Precursor ion scan and multiple reaction monitoring scan (MRM) are two typical scan modes in mass spectrometry analysis. Here, we developed a strategy by combining stable isotope labeling (IL) with liquid chromatography-mass spectrometry (LC-MS) under double precursor ion scan (DPI) and MRM for analysis of thiols in 5 types of human cancer urine. Firstly, the IL-LC-DPI-MS method was applied for non-targeted profiling of thiols from cancer samples. Compared to traditional full scan mode, the DPI method significantly improved identification selectivity and accuracy. 103 thiol candidates were discovered in all cancers and 6 thiols were identified by their standards. It is worth noting that pantetheine, for the first time, was identified in human urine. Secondly, the IL-LC-MRM-MS method was developed for relative quantification of thiols in cancers compared to healthy controls. All the MRM transitions of light and heavy labeled thiols were acquired from urines by using DPI method. Compared to DPI method, the sensitivity of MRM improved by 2.1–11.3 folds. In addition, the concentration of homocysteine, γ-glutamylcysteine and pantetheine enhanced more than two folds in cancer patients compared to healthy controls. Taken together, the method demonstrated to be a promising strategy for identification and comprehensive quantification of thiols in human urines.


Results
In our proposed strategy, the IL combined with LC-DPI-MS and LC-MRM-MS analysis was performed for the qualitative and relative quantitative analysis, respectively. The schematic diagram of the principle of this method was shown in Fig. 1.
Optimization of TCEP and BQB conditions. The effect of the TCEP and BQB contents on reduction efficiencies of disulfide bonds and chemical labeling were investigated. Equal volumes of pooled urine samples from five cancer types and healthy control were mixed as the investigated samples. 5 compounds at m/z of 261, 377, 379, 401, 429 and retention times of 4.7, 16.5, 21.4, 27.9, and 33.6 min were extracted, and their peak areas were calculated to evaluate the TCEP and BQB effects.
Determination of thiols is complicated due to their occurrence in multiple forms, since their free sulfhydryl group is prone to oxidation 9 . TCEP has frequently been used as reducing agent and is considered as a suitable choice for low molecular weight disulfides 1 . The reducing effect of TCEP was investigated in the range of 10-500 nmol and the content of BQB was fixed at 100 nmol. As shown in Figure S1A, the peak areas of five thiol derivatives increased with the increase in TCEP concentration from 10 to 100 nmol. Whereas, the peak area dropped with further increase in TCEP concentration (> 100 nmol) for the four analytes (m/z 377, 379, 401, and 429), and 200 nmol for the fifth (m/z 261), indicating that the efficiencies of chemical labeling may be suppressed by the excess of TCEP. Consequently, for identification of most of thiols in urine, 100 nmol of TCEP was used in the following experiments.
The amount of labeling reagent ranging from 5 to 200 nmol was further optimized. As shown in Figure S1B, the peak areas of thiol derivatives increased with the increase of BQB content from 5 to 20 nmol and finally reached a plateau when BQB contents exceeded 20 nmol. For more reliable quantitation of thiols in urine, 50 nmol of BQB was used in the following experiments.

Qualitative analysis of thiols in human urine by IL-LC-DPI-MS.
After optimization, we qualitatively analyzed the presence of thiols in human urine by IL-LC-DPI-MS method and pooled samples (n = 10) of each cancer type were prepared to minimize the variation between individuals. Figure 2A shows the total ion Scientific RepoRts | 6:21433 | DOI: 10.1038/srep21433 chromatograms of urine in nasopharyngeal cancer analyzed by IL-LC-DPI-MS. The two chromatograms derived from the BQB and BQB-d 7 labeled urine samples displayed almost identical peak patterns. The other four cancer samples (i.e., esophagus cancer, gastric cancer, lymph cancer and lung cancer) analyzed by LC-DPI-MS were shown in Figure S2. Extracted peak-pair data from the two ion chromatograms according to a mass difference of 7 Da (i.e., M BQB-d7 labeled − M BQB labeled = 7 Da) and only peak pairs with the same retention time and intensity were assigned to be the thiol candidates. Taking compound 7 and 8 as the examples (Fig. 2B), two peak pairs at retention times of 14.4 and 16.3 min were observed between the extracted ion chromatograms at m/z 333 and 340 from BQB and BQB-d 7 labeled samples, respectively. Same peak intensities and retention times of those two peaks in two labeled chromatograms suggest that these two compounds were all thiol candidates. The structures of all the  identified thiols were further elucidated by product-ion scan (MS/MS) and high resolution mass spectrometry (QTOF-MS) analysis.
In our previous research, a phenomenon was observed that if the compound containing n sulfhydryl groups (i.e., n = 2-6), all of the sulfhydryl groups could be labeled with BQB, and highest intensity peaks of these derivatives with n charge states (i.e., [M + n × BQB] n+ ) were observed among several of precursor ions with different charge states 22 . Similar peak pattern was observed in case of the BQB-d 7 labeled compounds with m/z of ([M + n × BQB-d 7 ] n+ ). The mass shift of BQB and BQB-d 7 labeled derivatives was always 7 Da no matter the compounds contain one or more sulfhydryl groups. To distinguish the numbers of sulfhydryl groups in thiol candidates, the charge number of the precursor ions from derivatives should be examined.
The charge numbers of thiol were further examined by QTOF-MS analysis and total 103 ion pairs were detected in all the 5 types of cancer urine samples (Table 1). These results were consistent with the results obtained from healthy urine in previous report 25 . Most of the detected thiols contain single charge, indicating that most of thiols only have one sulfhydryl group. The other 19 compounds could not be assigned the charge number by the QTOF data analysis software, which may be attributed to the low abundance of those thiols and the ion suppression effect of matrix interference.
Among the 103 identified thiols, 5 (No. 1-5) have been recognized as Cys, HCys, Nac, γ -GluCys, and GSH by the standards, and 12 (No. [6][7][8][9][10][11][12][13][14][15][16][17] were given the possible structures by the MS/MS and QTOF-MS information from our previous work 25 . In current research, it was found through standard comparison that the compound no. 6 named as cysteamine in previous reports, was not actually cysteamine(data not shown). Also, we found that the prospective molecular formulas (C 11 H 22 N 2 O 4 S) of compound 82 was identical to pantetheine, which have been found in Arabidopsis thaliana extracts 30 . Through comparison of retention times ( Figure S3) and MS/MS data ( Figure S4) with standards, compound 82 was identified to be pantetheine. The proposed structures of product ion derived from MS/MS spectra of pantetheine are also shown in the Figure S5. It is worth noting that the existence of pantetheine in urine is first reported by our developed method.

Development of IL-LC-MRM-MS method.
We further investigated the content changes of the 103 thiols in 5 types of cancer urine compared to healthy control by the IL-LC-DPI-MS method. The peak area ratios of each identified thiol from the cancer urines relative to healthy control were calculated by the forward and reverse labeling tests. However, results showed that only 64 (62%) of the 103 thiols could be calculated with the RSDs lower than 30% (data not shown). Among the other 38% of thiols not being quantified, which have lower signals or the signals were originally near the limit of quantifications (LOQs), were largely affected by the instrument signal fluctuations. So, the calculated RSDs were higher than 30% or their signals became lower than LOQs.
LC-MS operated in MRM mode has been widespread for targeted metabolite quantification 31 . The precursor ion and corresponding product ion of metabolite is monitored simultaneously in MRM mode, which improves the detection specificity and sensitivity by reducing the matrix interferences signals. In current study, we proposed an IL-LC-MRM-MS method to investigate the content changes of thiol between cancers and healthy controls. In this method, the product ion was fixed at m/z 218.  Figure 3A shows the extracted ion chromatograms of nasopharyngeal cancer urine analyzed by IL-LC-MRM-MS. In contrast to the DPI methods, all the MRM information was generated in single spectrum. Similar with DPI method, the peak intensities of m/z 333 and 340 from BQB and BQB-d 7 labeled thiols at retention times of 14.4 and 16.3 min in the MRM mode were also same ( Fig. 3B). We also compared the S/N of DPI and MRM methods. Result shows that the S/N significantly improved by 2.1-11.3 folds in MRM compared to DPI method. For example, the S/N of compounds 75 and 80 in the DPI analysis ( Figure S6A To evaluate the accuracy of the relative quantification obtained by our developed method, BQB and BQB-d 7 labeled pooled urines were mixed at different volume ratios (1:10, 1:5, 1:2, 1:1, 2:1, 5:1, and 10:1) and the samples were analyzed by LC-MRM-MS in triplicate measurements. The peak area ratios of BQB/BQB-d 7 labeled samples were calculated from sixty peak pairs with high intensities. The obtained average isotopic ratios were 0.1, 0.2, 0.5, 1.0, 2.1, 5.3, and 10.3 for the 1:10, 1:5, 1:2, 1:1, 2:1, 5:1, and 10:1 mixtures, respectively, with relative standard deviations (RSDs) being less than 9.0%. The correlation coefficient (R 2 ) was 0.9998 and the slopes of linear regressions were approximately 1.00 (1.0291), which shows that the peak area ratios highly matched with the concentration ratios of the different isotope labeled analytes ( Figure S7).

Relative quantitative analysis of thiols in human urine by IL-LC-MRM-MS.
We further investigated the content changes of the 103 thiols in 5 types of cancer urine samples (nasopharyngeal cancer, esophagus cancer, gastric cancer, lymph cancer, and lung cancer) compared to the healthy controls by our developed IL-LC-MRM-MS method. Before analysis, the creatinine in each pooled samples were quantified according to previously reported method 16,[32][33][34] (Table S1). The creatinine is a standard manner to normalize the concentrations of urine sample since the excretion of creatinine is rather constant over a longer time interval. The results of the measured peak areas ratios (cancer/healthy) are shown in Table S2. It is worth noting that, compared to the LC-DPI-MS method, the number of thiols being accurately quantified changed from 64 (62%) to 99 (96%) by the LC-MRM-MS method.
The unpaired student's t-test was performed to examine the statistical significance of fold changes from six independently biological experiments (three from forward labeling and the other three from reverse labeling). The result of statistical test including p-values and 95% confidence interval estimates are shown in Table S3. The fold change of increased or decreased more than 2.0 and the p-values less than 0.01 were considered as a statistically significant difference. Figure 4 shows a volcano plot, where the -log 10 (p value) was plotted against its corresponding log 2 (fold change of cancer/healthy control). The blue plots represent the significantly decreased and the red plots represent the significantly increased thiols compared to healthy controls.  Table 2, amongst the decreased thiols, compound 32 decreased more than 2.0 folds in all types of cancers, which may be employed as potential indicator for the screening of cancers. Similarly, compound 77 decreased more than 2.0 folds in 3 types of cancers (nasopharyngeal cancer, lymph cancer, and lung cancer), whereas, it increased more than 2.0 folds in the gastric cancer (Table 3). Compounds 6, 12, 33, and 82 (pantetheine) were also found more than 2.0 folds decrease in nasopharyngeal and lymph cancer, lymph and lung cancer, nasopharyngeal and gastric cancer, and esophagus and lung cancer, respectively. Interestingly, compounds 12 also increased more than 2.0 folds in the esophagus cancer (Table 3).

As shown in
For the increased thiols, as shown in the Table 3, compound 40 was found more than 2.0 folds increase in 3 types of cancers (nasopharyngeal cancer, esophagus cancer and gastric cancer). Compound 41 increased 2.0 foldsin nasopharyngeal and esophagus cancers. Both compounds 18 and 56 increased 2.0 folds in esophagus and gastric cancers, but compounds 18 decreased more than 2.0 folds in lung cancer ( Table 2). All 5 types of cancers, including nasopharyngeal cancer, esophagus cancer, gastric cancer, and lung cancer, have their own characteristic thiols with increased level, except lymph cancer. It is worth noting that the two known thiols of HCys and γ -GluCys increased 2.0 folds in nasopharyngeal cancer and gastric cancer, respectively. The result of HCys was consistent to the previous reports on plasma sample analysis from breast, colorectal, and cervical cancer 2-4,6 . However, there was no study reported so far about the content changes of thiol in urine of cancer. Thus, our study presents the first report for the increased level of HCys and γ -GluCys in urine of nasopharyngeal cancer and gastric cancer, respectively.
For further elucidation of significantly increased and decreased thiols in cancer (Tables 2 and 3), 3 compounds (pantetheine, HCys and γ -GluCys) were successfully identified by comparing to the standards; and 2 compounds (compound 11 and 12) were given the possible structures by MS/MS and QTOF-MS information. However, most of the compounds could not be identified. We could not found their structures in the HMDB and METLIN database by the prospective formulas or molecular weight information, suggesting that they may not been previously found in biological samples. To give reference in the following study, we could found prospective structures in the ChemSpider database, which is a free chemical structure database providing fast structure search access to over 35 million structures from hundreds of data sources.
In order to reach a reliable result, we took all types of 50 cancer urines as "cancer samples", also collected 30 healthy urines as "healthy control", then, the pooled samples of cancer and healthy control were labeled with BQB and BQB-d 7 to form the forward and reverse labeling samples to access the content changes of thiol in mixed cancers compared to healthy control (Table S4). As shown in Table S4, compound 32 was decreased in the mixed cancers, the average peak area ratios (cancer/healthy control) was about 0.6, similar to the results from individual types of cancer (Table S2, from 0.3 to 0.6). Some other thiols, which were significantly increased or decreased in their corresponding cancers compared to healthy controls, have not been found changes through mixing all types of cancer as one sample pool, i.e., compound 4 (γ -GluCys) in gastric cancer and compound 102 in lung cancer. So, future research should be focused on the individual types of cancer and more numbers of sample would be analyzed to verify our findings.

Discussion
Biothiol imbalances in biological samples are associated with different kinds of disease, such as cardiovascular disease, neurodegenerative disease, cancer, kidney dysfunction, and diabetes mellitus 1 . Based on these biochemical  NO.  Here, we developed a novel method for the comprehensive analysis of thiols in 5 types of cancer urine. In this method, the IL-DPI-LC-MS was firstly applied for non-targeted profiling of thiols in 5 cancer urines. The DPI method can significantly improve the identification accuracy by generating two individual ion chromatograms corresponding to BQB and BQB-d 7 labeled urines. Using this strategy, 103 thiol candidates were discovered in all the cancer urines and 6 thiol candidates in urines were confirmed as Cys, HCys, Nac, γ -GluCys, GSH and pantetheine by standards. In these identified thiol compounds, pantetheine has been firstly discovered in human urine, which extends the diversity of the thiol metabolites present in human urine. The pantetheine is considered as an intermediate in the production of coenzyme A in mammalian liver by preparation and purification of enzymes in vitro 35 . However, through comparative genomics, pantetheine was not found in coenzyme A biosynthesis pathway in the body 36 . However, it is unclear whether the pantetheine detected in urine is the metabolic product of 4'-phosphopantetheine, dephosphocoenzyme A, or coenzyme A, all of which have pantetheine structure moiety and are considered as cofactors in coenzyme A biosynthesis by the body, or an intermediate in other biological pathway. Thus, it is essential to investigate this compound in human cell extracts to elucidate its existence and biological pathway in our future work.

Commom compounds Individual compounds
Then, the IL-LC-MRM-MS method was firstly developed and applied to compare content changes in 5 types of cancer and healthy controls. Compared to DPI method, the quantification sensitivity of MRM method improved by 2.1-11.3 folds. Then, the number of compounds, which could be accurately quantified, changed from 64 (62%) to 99 (96%). We found that different content changes of thiols are associated with different types of cancers. Every cancer has their own characteristic thiols which significantly increased or decreased compared to healthy controls. The phenomenon may be due to the heterogeneity of different cancers. The HCys and γ -GluCys were firstly reported more than 2.0 folds increase in the urine of nasopharyngeal cancer and gastric cancer, respectively, then, the two thiols could be considered as potential biomarker for the nasopharyngeal and gastric cancers. The pantetheine were found more than 2.0 folds decrease in both esophagus and lung cancer urines. In addition, compounds 32 decreased more than 2.0 folds in urines of all the examined types of cancers, which may be employed as potential indicator for the screening of cancers. However, most of the compounds that showed significant changes could not be identified. So, further study should focus on the identification of thiols and provide an insight into the better use of urinary thiols as biomarkers for cancers. Taken together, the IL-LC-DPIS-MS method combined with IL-LC-MRM-MS method demonstrated to be a promising strategy for the identification and quantification of compounds with identical groups in metabolomics study. Urine sample collection and preparation. The 5 types of cancer (nasopharyngeal cancer, esophagus cancer, gastric cancer, lymph cancer, and lung cancer) and healthy control were collected from Hubei Cancer Hospital, China. The 10 samples of first morning urine of every types of cancer and healthy controls were collected (5 males and 5 females; 60 ± 5 years old). All the patients were diagnosed with cancer for the first time and had not been given any treatment at the time point of urine samples collection. Healthy controls were selected based on medical history and physical examination. Written informed consent was obtained from the study subjects, and an approval was granted by the Hubei Cancer Hospital Ethics Committee and met the declaration of Helsinki. All the experiments were performed in accordance with Hubei Cancer Hospital Ethics Committee's guidelines and regulations.

Reagents
The urine samples were pretreated according to previously described method 16 . Briefly, 200 μ L of each urine sample was added to a prepared screw-cap vial (1.5 ml) containing 18 μ L of EDTA (10 mmol/L) and 2 μ L of formic acid. Six pooled samples of five cancers and healthy control were prepared by taking equal volume of their   Principle of the strategy. Firstly, each type of cancer urine was subjected to IL-LC-DPI-MS method for the non-targeted profiling of thiols. Equal volume of sample was labeled with BQB and BQB-d 7 , respectively. Then the light and heavy labeled samples were mixed and analyzed by LC-DPI-MS. The LC-DPI-MS method generated two individual ion chromatograms corresponding to the precursor ion of BQB and BQB-d 7 labeled thiols, respectively. Peak-pair data were extracted from the two ion chromatograms according to a characteristic mass difference and only peak pairs with the same retention time and intensity were assigned as thiol candidates. Secondly, the targeted relative quantification of the thiols between the cancer and healthy control was investigated by the IL The DPI method consists of two PI (m/z 218 and 225) in the mass range of m/z 200-600. DPI was carried out under positive ion mode. IsoSpray voltage was set at 5.2 kV and vaporizer temperature was set at 550 °C. The mass spectrometer was operated with gas settings of 40 psi for nebulizer gas, 30 psi for curtain gas, and 60 psi for collision gas. Scan time per cycle was 2.0 s with a pause of 5.0 ms for each scan. Resolution of Q1 and Q3 was set to "low" and "unit", respectively. Declustering potential, entrance potential, cell entrance potential, collision energy and cell exit potential were set at 45 V, 7 V, 15 V, 38 V and 3 V, respectively.
For structural identification (MS/MS analysis), IDA (Information Dependent Acquisition) mode was performed under positive ion mode. The criteria were set as that EPI was triggered when signals of the pre-selected compounds by PI exceeding 1000 counts/s at their retention times. The mass tolerance was set to 250 mDa, and retention time tolerance was set to 60 s.