Urinary Amine Metabolomics Characterization with Custom 12-plex Isobaric DiLeu Labeling

Abstract

Lower urinary tract symptoms (LUTS) are common in aging males. Disease etiology is largely unknown, but likely includes inflammation and age-related changes in steroid hormones. Diagnosis is currently based on subjective symptom scores, and mainstay treatments can be ineffective and bothersome. Biomarker discovery efforts could facilitate objective diagnostic criteria for personalized medicine and new potential druggable pathways. To identify urine metabolite markers specific to hormone-induced bladder outlet obstruction, we applied our custom synthesized multiplex isobaric tags to monitor the development of bladder outlet obstruction across time in an experimental mouse model of LUTS. Mouse urine samples were collected before treatment and after 2, 4, and 8 weeks of steroid hormone treatment and subsequently analyzed by nanoflow ultrahigh-performance liquid chromatography coupled to tandem mass spectrometry. Accurate and high throughput quantification of amine-containing metabolites was achieved by twelve-plex DiLeu isobaric labeling. Metandem, a novel online software tool for large-scale isobaric labeling-based metabolomics, was used for identification and relative quantification of labeled metabolites. A total of 59 amine-containing metabolites were identified and quantified, 9 of which were changed significantly by the hormone treatment. Metabolic pathway analysis showed that three metabolic pathways were potentially disrupted. Among them, the arginine and proline metabolism pathway was significantly dysregulated both in this model and in a prior analysis of LUTS patient samples. Proline and citrulline were significantly changed in both samples and serve as attractive candidate biomarkers. 12-plex DiLeu isobaric labeling with Metandem data processing presents an accessible and efficient workflow for amine-containing metabolome study in biological specimens.

Brief Synopsis of the Graphic:

The urine sample is collected from the LUTS mouse model. Extracted metabolites from distinct urine samples are differentially labeled with isobaric DiLeu tags and each metabolite is shown as a single peak at the MS¹ level. The subsequent MS² analysis enables relative quantification based on reporter ion intensities. Metandem software is used for data analysis. Finally, the quantification results can be used for biomarker discovery of LUTS.

Graphical Abstract

INTRODUCTION

Lower urinary tract symptoms (LUTS) is a major public health problem in the aging population. Symptoms include frequency, urgency, nocturia, weak urinary stream, straining to void, and a sense of incomplete emptying (1–3). Recent studies have shown that the etiology of LUTS is multi-factorial (4, 5). In men, prostate enlargement, prostatic inflammation, age-related changes in detrusor function and steroid hormones, and heightened sensitivity to bladder filling may all contribute (4–8). The complex and variegated composition of the LUTS patient population makes it challenging to tease out and validate contributions of different mechanisms. Currently no objective biomarkers exist to inform treatment strategies, so patients are generally treated with drugs to decrease prostate size (5α-reductase inhibitors) or relax smooth muscle (alpha blockers) (9, 10). These treatments are not completely effective, durable in response, nor are they without complication (11). Patients with LUTS that persists after these mainline treatments may undergo invasive transurethral resection of the prostate (12). Overall, better biomarkers and treatment options need to be found. Here, we attempt to characterize the contribution of hormone changes to the urine metabolome under controlled, experimental conditions in a mouse model of bladder outlet obstruction (BOO). Understanding the urine signature of hormone-induced BOO will inform both future efforts to study druggable pathways and validate LUTS biomarkers for patient stratification and personalized treatments.

Our hormone-induced BOO mouse model relies on slow-release, subcutaneous implants of testosterone and 17β-estradiol in adult male C57/BL6 mice to generate marked increases in urinary frequency, bladder volume, bladder mass, and prostate mass as well as decreased urinary volume/void before the 8 week timepoint (13). Additionally, this model, if treated longer (4 months), develops prostatic intraepithelial neoplasia lesions, making this model useful for the study of carcinogenesis in the prostate (14). To characterize the urine metabolome of hormone-induced BOO, we collected samples across the development of this phenotype: before treatment and after 2, 4, and 8 weeks of treatment.

Tandem mass spectrometry (MS/MS)-based relative quantification by isobaric labeling is a useful technique for comparative quantitative metabolomics in biological systems (15–17). N,N-dimethyl leucine (DiLeu) labels are isobaric. This means that analyte precursor m/z between channels is indistinguishable at low MS resolution, but distinct reporter ions are apparent in the low m/z region upon MS² fragmentation. The intensities of these reporter ions in MS² spectra reflect the labeled metabolites’ abundance in each sample and, thus, can be compared for relative quantification. Compared to label-free metabolomics, multiplexed isobaric labeling of metabolites greatly reduces run-to-run variation, enhances ionization efficiency, improves chromatographic separation of polar metabolites, and decreases instrument time demand (by 12-fold when using 12-plex DiLeu labeling) (18). DiLeu utilizes a triazine ester to label primary and secondary amines, making this derivatization scheme applicable to many metabolites. Additionally, DiLeu labeling enables polar metabolites to be separated and detected on nanoUPLC systems, improving chromatographic resolution and detection sensitivity over typical standard flow separation techniques (19). Twelve-plex DiLeu isobaric labels, which are designed and synthesized in our lab (20, 21), were employed here for the relative quantification of amine-containing metabolites in urine samples via nanoflow ultrahigh-performance liquid chromatography coupled to tandem mass spectrometry (nanoUPLC-MS/MS).

Isobaric labeling has been widely adopted for quantitative proteomics and peptidomics, but it has only recently been applied to metabolomics analyses (22–24). Thus, data processing workflows for isobaric labeling in proteomics and peptidomics have matured more rapidly than for metabolomics. To date, few metabolomics software tools can process stable isotope label-based metabolomics data, particularly when using reporter ions produced by MS² for quantification. Therefore, our lab developed a novel online software tool for isobaric labeling-based metabolomics, called Metandem, which integrates metabolite quantification, identification, and statistical analysis in the same software package and is freely available at http://metandem.com/web/ (25). Metandem is also the first omics data analysis software to provide straightforward, online parameter optimization functionality for customization to a particular dataset (25). Here, we employed the Metandem software tool to analyze the 12-plex DiLeu-labeled urinary metabolites at multiple time points after hormone treatment. Metabolite identification, quantification, and statistical analysis were achieved in less than 15 min using our Metandem software tool. It is expected that these powerful tools may identify clinically useful biomarkers of hormone-induced bladder outlet obstruction and new targets for drug treatment.

MATERIALS AND METHODS

Hormone-induced urinary obstruction mouse model

All animal procedures were approved by the University of Wisconsin-Madison Animal Care and Use Committee. Adult male C57BL/6 mice were used for this study (n = 3) (Charles River, Wilmington, MA). Urine was collected from each animal via metabolic cage (26) for a two-hour period. After the pre-treatment urine collection, mice were treated with subcutaneous, slow-release implants containing compressed testosterone (T, 25 mg) and 17β-estradiol (E2, 2.5 mg + 22.5 mg cholesterol binder), as described previously (13). Urine samples from the same mice were then collected at 2, 4, and 8 weeks post-treatment, as above. All samples were stored at −80 °C until further processing.

Mouse urine sample preparation

Urine samples were thawed on ice and mixed well by vortexing. Raw urine (220 μL) was centrifuged at 10,000 g for 10 min to remove particulates and cellular debris. Centrifugal filters (3 kDa, Millipore Amicon Ultra, Burlington, MA) were pre-rinsed 3 times with 500 μL of Milli-Q water at 14,000 g for 20 min. Clarified supernatant (200 μL) of each urine sample was added to the filter unit and centrifuged at 14,000 g for 30 min, followed by 2 rinses with Milli-Q water (200 μL; same centrifugation speed and time) to obtain the metabolite fraction (~600 μL total). Osmolality was determined via a freezing-point depression osmometer (Osmometer Model 3250, Advanced Instruments, Norwood, MA) (27) and all samples were normalized to 50 mOsmoles/kg H₂O. Osmolite concentration can represent the total urinary metabolite content, and normalized to osmolality has been proven to be a better method for metabolite normalization prior to instrumental analysis (28). Then all the aliquoted urine samples were lyophilized and stored at −80 °C until labeling.

DiLeu synthesis and labeling procedure

Twelve-plex isobaric DiLeu reagents were synthesized and used for labeling reaction as previously described (29). Briefly, the DiLeu 12-plex reagents were synthesized, aliquoted in inactivated form at 4 °C in a dry box and activated prior to labeling. Lyophilized urine metabolite samples were re-dissolved in 0.5 M triethylammonium bicarbonate solution prior to derivatization of primary and secondary amines by excess activated DiLeu reagent (Fig. 1D). The organic: aqueous ratio was maintained at ~70% via anhydrous dimethylformamide. Reactions were allowed to proceed for 2 hrs at room temperature with vigorous vortexing. Labeling reactions were quenched with 0.25% hydroxylamine (v/v), and labeled samples were combined in equal ratios (v/v) to form pooled 12-plex samples. Excess DiLeu reagents were removed from pooled 12-plex samples via SCX Ziptips (OMIX-SCX, Agilent, Santa Clara, CA) as previously described (19) before lyophilization and storage at −80 °C until analysis.

Figure 1. Details of 12-plex DiLeu labeling.

A: Precursor ion of DiLeu-labeled citrulline; B: An MS² spectrum of the 12-plex DiLeu-labeled citrulline acquired in the Orbitrap at 60 K resolving power. Low m/z region showing distinct DiLeu reporter channels (bottom) and after zooming in, twelve distinct reporter ion peaks are present (bottom); C: Citrulline changing trends from different time points of three biological replicates. D: Sample labeling map showing the 12-plex DiLeu tags and time point (randomized)

LC-ESI-MS analysis

Twelve-plex pooled samples were reconstituted in 3% acetonitrile, 0.1% formic acid (v/v) in water before injection. UPLC-MS/MS analysis was conducted using a Thermo Dionex UltiMate™ 3000 nanoLC system coupled to a Thermo Q Exactive™ HF Orbitrap MS. The analytical column was self-made with an integrated emitter tip and dimensions of 75 μm inner diameter × 15 cm length, packed with 1.7 μm, 150 Å, BEH C₁₈ material (Waters, Milford, MA). Mobile phase A was 0.1% formic acid in water, and mobile phase B was 0.1% formic acid in acetonitrile (Optima Solvents, Thermo, Waltham, MA). The flow rate was 0.3 μL/min, and the 70-min gradient was as follows: 0–16 min, 3% solvent B; 16–20 min, 3–25% B; 20–30 min, 25–45% B; 30–50 min, 45–70% B; 50–56 min, 70–95% B; 56–60 min 95% B; 60–60.5 min, 95–3% B; 60.5–70 min, 3% B. Positive ionization mode was used for the MS analysis. Full MS scans were acquired from m/z 180 to 1000 at a resolution of 60 K, automatic gain control (AGC) at 1 × 10⁶, and maximum injection time of 50 ms. The top 20 precursors were selected for higher-energy C-trap dissociation tandem mass spectrometry (HCD MS²) analysis with an isolation window of 1 m/z, normalized collision energy (NCE) of 30, resolving power of 60k, AGC target of 1 × 10⁵, maximum injection time of 30 ms, and a lower mass limit of 110 m/z.

Data analysis

Raw data files were acquired using Thermo Scientific Xcalibur™ software and converted into .txt format via the COMPASS software suite (30). Metandem was then used to batch-process three technical replicates of each 12-plex sample for metabolite quantification. Accurate mass of reporter ions was obtained by averaging the mass across several MS/MS spectra. Then, the mass tolerance for metabolite identification was 20 ppm. DiLeu label purity was predetermined, and correction for each channel was performed as previously described (19). Average precursor mass shift due to labeling was 145.1280 Da. Data analysis parameters, such as reporter ion mass tolerance, batch mass tolerance, and retention time tolerance were optimized using the parameter optimization graphs in the Metandem software. Output files with reporter ion information were merged and median-normalized. Molecular weights of detected compounds were calculated based on the charge and mass shift caused by labeling, and then searched against the Human Metabolome Database (HMDB). The unreasonable matches were excluded from the identification results and only the metabolites with primary or secondary amine were kept. Paired t-test was performed comparing metabolite abundance of 2/4/8 weeks to that of 0 week. Any of the three comparisons showing fold change > 1.5 and p-value < 0.05 were considered as significantly changed metabolites. MetaboAnalyte 4.0 software (31) and Kyoto Encyclopedia of Genes and Genomes (KEGG) database were used for metabolic pathway analysis. Pathways were considered dysregulated when more than two metabolites were identified in the pathway and at least one was significantly changed due to the treatment. Candidate biomarkers and dysregulated pathways were compared with prior patient analyses and other previous reports.

RESULTS AND DISCUSSION

Efficacy of twelve-plex DiLeu isobaric labeling for metabolomics

Twelve-plex DiLeu isobaric labeling allowed multiplexed metabolite identification and quantification in mouse urine samples while also reducing instrumentation time demand, decreasing run-to-run variation, and also improving quantification accuracy. The same labeled metabolite from 12 urine samples showed a single peak in the MS¹ spectrum with a mass shift of 145.1280 Da (Fig. 1A). For this peak, twelve distinct reporter ion peaks are present in the MS² low m/z region (Fig. 1B). The intensity of each reporter ion in MS² spectra reflects the labeled metabolites’ abundance in each sample and, thus, can be compared for relative quantification (Fig. 1C). The absence of unlabeled metabolite ions in the MS¹ (mass difference of 145.1280 compared to unlabeled counterparts) suggested highly efficient labeling of the metabolites by the DiLeu reagent.

Metandem parameter optimization

Metandem is a newly developed custom software platform for large-scale stable isotope labeling-based metabolite identification and quantification. It is also the first omics data analysis software that contains functionality to perform online parameter optimization for customization to a dataset. Results of automated parameter optimization were as follows: optimal reporter ion mass tolerance, 0.5 mDa (Fig. 2A); optimal batch processing mass tolerance, 5 ppm (Fig. 2B); and optimal batch processing retention time tolerance, 0.5 min (Fig. 2C). Metandem also provides the histogram for retention time (Fig. S1A) and detected precursor (Fig. S1B) distribution results.

Figure 2. Metandem parameter optimization results:

A: Optimizing reporter ion mass tolerance (0.5 mDa); B: Optimizing batch processing retention time tolerance (5 ppm); C: Optimizing batch processing retention time tolerance (0.5 min).

Mouse urine metabolite identification and quantification

Three technical replicates of each 12-plex injection were merged in Metandem. A total of 312 features were identified as putative metabolites after accurate mass matching against the HMDB. After excluding unreasonable matches from the database, 59 were primary or secondary amine-containing metabolites were identified (Table S1). Thirty-seven of these 59 were documented in the urine metabolome database (32), and the twenty-two additional amine-containing metabolites were, to our knowledge, first reported here. After comparing metabolite abundance between timepoints, 9 metabolites were identified as statistically significant biomarker candidates (Paired t-test, fold change > 1.5 and p-value < 0.05. The smallest p-value out of the three was show in Table 1). Among them, eight metabolites were generally increased at all timepoints, while only one metabolite was generally decreased at all timepoints (Fig. 3). For the increased expression patterns, leucine and 5-aminopentanamide have the highest concentration at 2 weeks; N-acetylputrescine at 4 weeks; and proline, citrulline, D-alanyl-D-alanine, O-phosphohomoserine, and 2-Aminobenzoic acid peaked at 8 weeks.

Table 1.

Nine significantly changed urine metabolites after hormone treatment.

Compound	Molecular Weight	p-value	Expression Pattern	ppm	tr (min)	HMDB_ID
Proline	115.0623	0.006	Increase	9.2	13.6	HMDB00162
5-Aminopentanamide	116.0938	0.029	Increase	10.2	14.4	HMDB12176
N-Acetylputrescine	130.1108	0.042	Increase	1.2	15.2	HMDB02064
Citrulline	175.0936	0.005	Increase	11.7	15.5	HMDB00904
D-Alanyl-D-alanine	160.0827	0.005	Increase	13.1	16.5	HMDB03459
O-Phosphohomoserine	199.0278	0.001	Increase	16.1	16.7	HMDB03484
Purine	120.0450	0.001	Decrease	11.3	8.6	HMDB01366
2-Aminobenzoic acid	137.0464	0.023	Increase	9.7	19.0	HMDB01123
Leucine	131.0942	0.006	Increase	3.0	19.0	HMDB00687

Figure 3.

Quantification trends of 9 significantly changed urine metabolites (n = 3; Paired t-test, fold change > 1.5 and p < 0.05).

Metabolic pathway analysis

Metabolic pathway analysis is based on the association between identified metabolites and their related biological processes (33). Herein, all identified metabolites were input into the MetaboAnalyte 4.0 software for metabolic pathway analysis. Three potentially perturbed metabolic pathways were identified: (1) the arginine and proline metabolism pathway; (2) the aminoacyl-tRNA biosynthesis pathway; and (3) the tryptophan metabolism pathway (Table 2. p-value is from Fisher’s exact test). Although tryptophan metabolism pathway has a p-value slightly higher than 0.05, it was also shown here as a candidate disrupted metabolic pathway.

Table 2.

Potentially disrupted metabolic pathways via MetaboAnalyte 4.0 and KEGG pathway analysis, metabolites highlighted with red bold font are significantly changed metabolites (p-value is from Fisher’s exact test).

Metabolic Pathway	KEGG ID	Matched Metabolites	p-value
Arginine and proline metabolism	Map00330	citrulline, N-acetylputrescine, proline, glutamic acid, creatine, GABA, 4-aminobutyraldehyde	4.90E-05
Aminoacyl-tRNA biosynthesis	Map00970	leucine, proline, cysteine, glycine, alanine, glutamic acid	4.88E-03
Tryptophan metabolism	Map00380	2-aminobenzoic acid, 3-hydroxyanthranilic acid, 5-hydroxyindoleacetic acid	6.81E-02

Comparison with LUTS patient samples and other previous reports

Prior analyses have demonstrated some of the metabolomic and immunohistochemical features of LUTS in patient samples, which represent processes like fibrosis and inflammation (27). Both LUTS patients and this hormone-induced mouse model of BOO show perturbation of the arginine and proline metabolic pathway in the urine metabolome. In particular, the candidate biomarkers citrulline and proline were significantly changed in both LUTS and this mouse model (Fig. 4). For proline, it was increased in both LUTS patients and the hormone-treated mice, represents a strong candidate biomarker for hormone-induced BOO. Urinary proline was also increased in a mouse model of hepatic injury and fibrosis, indicating fibrosis is a potential relevant pathway in the present BOO model (34). Next, citrulline was decreased in LUTS patients but increased in this mouse model. Citrulline is poorly understood in the context of prostate diseases. Citrulline is a non-essential amino acid, a precursor to arginine, and displays antioxidant properties (35). Increased urine citrulline in these mice could be explained by inflammation-induced expression of nitric oxide synthase 2, the enzyme responsible for conversion of arginine to citrulline (36), possibly in response to oxidative stress generated by catechol estrogen metabolites (37).

Figure 4.

Arginine and proline metabolism pathway is potentially disrupted (MetaboAnalyte, KEGG; Fisher’s exact test, p < 0.07).

Other processes were significantly changed in the present study but not observed in LUTS patients. For example, the aminoacyl-tRNA biosynthesis pathway was disrupted in the urine metabolome of this mouse model but not in LUTS patients. This pathway is pivotal in determining how the genetic code is interpreted as amino acids (38). More specifically, leucine and proline were significantly changed in this metabolic pathway. The tryptophan metabolism pathway, represented by 2-aminobenzoic acid, was also disrupted in this mouse model. Overall, these pathway-level results were similar to a prior report of urine metabolomics in liver injury and fibrosis (39). Interestingly, leucine may play an important role in prostatic proliferation. Leucine is an essential branched-chain amino acid that signals through the mTOR pathway (40). This signaling is pro-proliferative in prostate cancer cells, and as such decreasing leucine transport into tumors is an attractive therapeutic target (41). It is possible the strong initial increases in leucine in this model contribute to the increased prostate mass (benign hyperplasia) or even the development of prostatic intraepithelial neoplasia observed later in this model (14).

CONCLUSIONS

Twelve-plex DiLeu isobaric labeling is an attractive high-throughput strategy for identification and quantification of amine-containing metabolites, and Metandem is a useful tool for large-scale stable isotope labeling-based metabolomics data analysis. Paired together, these tools offer a powerful and accessible method for relative quantification of amine-containing metabolites in disease biomarker research. In this study of urinary amine metabolomics of a hormone-induced LUTS mouse model, we have identified and quantified 59 amine metabolites, and 22 of them were identified in urine for the first time. LUTS patients and this mouse model shared common pathways that are dysregulated compared to control groups, for instance, the arginine and proline metabolism pathway. Proline presents an especially attractive candidate biomarker for hormone-induced BOO, as it was significantly increased in both human LUTS and this mouse model. Future experiments will test the hypothesis that this hormone treatment results in fibrosis of the lower urinary tract, ultimately leading to the pronounced BOO phenotype.

Supplementary Material

Supplementary Information

Figure S1. Additional statistical graphs from Metandem

Table S1. List of 59 identified metabolites of interest