Deregulated Tyrosine–Phenylalanine Metabolism in Pulmonary

Feb 19, 2015 - Using these variables, a receiver operating characteristics curve was plotted and the area under the curve was calculated to be 0.85 (9...
0 downloads 0 Views 3MB Size
Article pubs.acs.org/jpr

Deregulated Tyrosine−Phenylalanine Metabolism in Pulmonary Tuberculosis Patients Mrinal Kumar Das,† Subasa Chandra Bishwal,† Aleena Das,†,∥ Deepti Dabral,†,⊥ Vinod Kumar Badireddy,† Bhaswati Pandit,‡ George M. Varghese,§ and Ranjan Kumar Nanda*,† †

Immunology Group, International Centre for Genetic Engineering and Biotechnology (ICGEB), Aruna Asaf Ali Road, New Delhi 110067, India ‡ National Institute of Biomedical Genomics, Kalyani, West Bengal 741251, India § Department of Medicine, Christian Medical College, Vellore, Tamil Nadu 632004, India S Supporting Information *

ABSTRACT: Metabolic profiling of biofluids from tuberculosis (TB) patients would help us in understanding the disease pathophysiology and may also be useful for the development of novel diagnostics and host-directed therapy. In this pilot study we have compared the urine metabolic profiles of two groups of subjects having similar TB symptoms and categorized as active TB (ATB, n = 21) and non-TB (NTB, n = 21) based on GeneXpert test results. Silylation, gas chromatography mass spectrometry, and standard chemometric methods were employed to identify the important molecules and deregulated metabolic pathways. Eleven active TB patients were followed up on longitudinally for comparative urine metabolic profiling with healthy controls (n = 11). A set of 42 features qualified to have a variable importance parameter score of > 1.5 of a partial least-squares discriminate analysis model and fold change of > 1.5 at p value < 0.05 between ATB and NTB. Using these variables, a receiver operating characteristics curve was plotted and the area under the curve was calculated to be 0.85 (95% CI: 0.72−0.96). Several of these variables that represent norepinephrine, gentisic acid, 4-hydroxybenzoic acid, hydroquinone, and 4-hydroxyhippuric acid are part of the tyrosine− phenylalanine metabolic pathway. In the longitudinal study we observed a treatment-dependent trend in the urine metabolome of follow-up samples, and subjects declared as clinically cured showed similar metabolic profile as those of asymptomatic healthy subjects. The deregulated tyrosine−phenylalanine axis reveals a potential target for diagnostics and intervention in TB. KEYWORDS: urine analysis, tuberculosis, silyalation, metabolomics, GC−MS, tyrosine−phenylalanine metabolism



INTRODUCTION Tuberculosis (TB), caused by Mycobacterium tuberculosis (Mtb), continues to be a major public health threat. In 2013 alone, 9.0 million new TB cases were identified and 1.5 million lives were lost worldwide, the majority of them from resourcelimited settings in Asia and Africa.1 The loss of disabilityadjusted life years (DALYs) due to the disease and lengthy treatment impedes socioeconomic development. It has been suggested that novel, more effective diagnostics and therapeutics in the form of host-directed therapy and protective vaccines are the need of the hour to reach the goal of TB eradication by 2050.2,3 Up to one-third of world’s population is thought to have latent TB infection, and ∼200 million may develop active disease in their lifetime.4 Multiple factors contribute to the reactivation of latent infection to active disease including immune status, metabolic phenotype of the host, and pathogenicity of the strain. During the transition of latent to active disease, metabolic homeostasis in the host gets disrupted.5 Exploring this shift in balance at molecular levels © 2015 American Chemical Society

in biofluids of patients will improve understanding in disease biology and may also help in developing translatable solutions in terms of novel diagnostics, host-directed therapy, and vaccines.6,7 Recent developments in metabolomics, the metabolic component of functional genomics, provide a practical approach for elucidating altered host metabolic phenotypes under different disease conditions.8 Few biofluidbased metabolomics studies have been carried out to elucidate altered host metabolism in TB, and most of these focused on sputum, serum, and plasma.9−15 Notably, however, urine can provide insight into the system-wide changes that are seen in response to the physiological challenges during disease onset and therapeutic intervention.16 Thus, a comparative urine metabolomics study may be valuable to explore novel metabolic pathways and may supplement existing information on the cytokine profiling, transcriptomics and proteomics from other biofluids to better understand TB pathophysiology.14,17,18 Received: January 9, 2015 Published: February 19, 2015 1947

DOI: 10.1021/acs.jproteome.5b00016 J. Proteome Res. 2015, 14, 1947−1956

Article

Journal of Proteome Research

Table 1. Epidemiological Details of Study Subjects Used in This Comparative Urine Metabolome Analysis for Tuberculosis study types

I. case and control

II. follow up after therapeutic interventions and healthy

subject details

ATB

NTB

0M

2M

4M

6M

healthy

no. of study subjects (E/S/N) mean age (range) in years gender (male in %) BMIb smoking habit (y/n/na) alcohol use (y/n/na) cough (y/n/na) expectoration (y/n/na) chest pain (y/n/na) fever (y/n/na) AFB (+ve/−ve/na) abnormal chest X-ray (y/n/na) cavity (y/n/na) GeneXpert (+ve/−ve)

21 (11/10/−) 38 (22−60) 100 16.43 ± 5.57b 11/10/− 9/12/− 21/−/− 21/−/− 12/9/− 14/7/− 21/−/− 14/−/7 11/2/8 21/−

21 (8/13/−) 46 (17−69) 100 22.19 ± 4.08b 11/10/− 4/8/9 21/−/− 21/−/− 3/18/− 12/1/8 −/11/10 2/2/17 2/−/19 −/21

11 (11/−/−) 35 (23−49) 100 17.83 ± 2.80 7/4/− 4/7/− 11/−/− 11/−/− 9/2/− 9/2/− 11/−/− 10/−/1 9/1/1 11/−

11 (11/−/−) 35 (23−49) 100

11 (11/−/−) 35 (23−49) 100

9 (9/−/−) 36 (23−49) 100

7/4/− 4/7/− 11/−/− 11/−/− −/−/11 −/−/11 1/10/− −/−/11 −/−/11

7/4/− 4/7/− 11/−/− 11/−/− −/−11 −/−11 −/11/− −/−11 −/−11

6/3/− 2/7/− 7/−/2 7/−/2 −/−/9 −/−/9 −/9/− −/−/9 −/−/9

11 (−/−/11) 34 (20−51) 100 25.12 ± 4.08 1/7/3 4/3/4 −/11/− −/11/− −/11/− −/11/−

a

Study subjects were recruited from E, Eastern; S, Southern; and N, Northern parts of India. bBody mass index (BMI in kg/m2) of active TB patients (ATB) was significantly lower than nontuberculosis (NTB) at p < 0.05. y, yes; n; no; na; information not available; AFB; acid-fast bacilli staining test.

Figure 1. Schematic presentation of methods used for subject classification and chemical derivatization of biological material. (a) Subjects with similar clinical symptoms of tuberculosis were classified as active TB (ATB) and non-TB (NTB) based on nucleic acid amplification test (GeneXpert). (b) Steps used for MOX and MSTFA derivatization of urine samples adopting methodology reported by Chan et al.,23 with minor modification. RT, room temperature (22 °C); MOX, methoxamine-HCl; MSTFA, N-methyl-N-(trimethylsilyl)trifluoroacetamide; TMCS, trimethylchlorosilane; GC−MS, gas chromatography and mass spectrometry.



Metabolite profiling of urine is carried out using multiple techniques including liquid−gas chromatography and mass spectroscopy: LC−MS/GC−MS and nuclear magnetic resonance (NMR) to identify deregulated molecular features and perturbed metabolic pathways. Each of these methods has its own advantages and limitations, and no method alone can capture the complete metabolomic picture.19 In this study, we used GC−MS-based metabolomic tools to explore the TBspecific biological activities by comparing urine metabolites of active TB and non-TB subjects sharing similar clinical symptoms.

EXPERIMENTAL PROCEDURES

Subject Recruitment

Study subjects for this pilot-scale urine metabolomics study were part of an ongoing project to validate putative urine volatile markers of TB previously identified by this group.20 Because this was a pilot scale study, power calculation was not carried out and a small sample set of 21 active TB (ATB) and similar number of age-matched, non-TB (NTB) male subjects showing similar symptoms were taken for this study. Epidemiological details of all study subjects are presented in Table 1, and metadata are available in Table S1 in the Supporting Information. All ATB and NTB subjects were 1948

DOI: 10.1021/acs.jproteome.5b00016 J. Proteome Res. 2015, 14, 1947−1956

Article

Journal of Proteome Research

0.25 μm; Restek USA) in a GC−MS (7890A GC, 5975 MSD from Agilent Technologies, USA) for separation. Helium was used as a carrier gas at a constant flow rate of 1 mL/min. The front inlet temperature was fixed at 250 °C during injection; temperature gradients of 50 to 150 °C (ramp of 10 °C/min) and 150 to 310 °C (ramp of 7 °C/min) with a hold time of 3 min between two ramps and after reaching final temperature were used. Electron ionization (EI) mode was fixed at −70 eV to scan ions of 35 to 600 m/z range. Maximum scan speed was 5 Hz with a 6 min solvent delay. The ion-source temperature and quadrupole temperatures were fixed at 230 and 150 °C, respectively. Sample introduction to data acquisition parameters (both GC separation and mass spectrometry) were controlled through ChemStation software (Agilent Technologies, USA), and the run time was 38.43 min per sample. Data Preprocessing. Before analysis, the sample codes were opened by a team member not participating in sample processing and GC−MS data acquisition. Raw GC−MS data files in .d format from ChemStation were converted to .elu and .fin using AMDIS 32 (National Institute of Standards and Technology, USA). Preprocessing of these files was carried out using Mass Professional Profiler (MPP, ver B.12.01 from Agilent Technologies, USA) for peak picking and alignment. Parameters used for peak alignment were retention time tolerance, 0.5 min; match factor, 0.6; and delta m/z, 0.2. Aligned features with peak area information were exported as .csv, and total area normalization was carried out manually using Excel (Microsoft, USA). Statistical Analysis to Select Important Deregulated Molecules. Uni- and multivariate analyses of the total area normalized metadata were carried out using MetaboAnalyst 2.0 to select important deregulated molecules.24 Missing values were imputed with half of the minimum value of study population, and the data matrix was normalized with the peak area of the feature representing the internal spike in standard Lysine D4, following which generalized log transformation and universal scaling method was employed to obtain a near Gaussian distribution of the variables to carry out multivariate analysis such as principal component analysis (PCA) and partial least-squares discriminate analysis (PLS-DA). Hierarchical clustering using Ward linkage and Euclidean cluster method, univariate analysis in terms of fold change (FC: ATB/NTB), and t test (Wilcoxon-Mann−Whitney test at p < 0.05) were carried out using MetaboAnalyst. Features qualifying criteria of FC > 1.5, t test p value < 0.05, and variable important parameter (VIP) score > 1.5 of a PLS-DA model were selected as important deregulated molecular features for TB. Selecting these features, area under the receiver operating characteristic curve (AUC of ROC) was calculated using ROCCET.25 Spearman correlation values (r2, p value, and false discovery rate (FDR) adjusted p value) among these important molecules within ATB group were calculated using MetaboAnalyst. Identification of Important Features. NetCDF files of three randomly selected QC samples, from the study, were exported from ChemStation and processed using ChromaTOF (4.50.8.0; Leco USA). Each feature is represented by a composite spectrum generated by Mass Profiler Professional (MPP, Agilent) software. The MPP-generated composite spectra were compared manually with output of ChromaTOF for fragmentation pattern matching. A similarity index (SI) of 700 was set in ChromaTOF for library annotation (NIST 11, total 243 893 spectra). Library annotations were further validated by matching retention indices (RIs) of those reported

recruited from Christian Medical Center, Vellore (CMC) and the National Institute of Biomedical Genomics, Kalyani (NIBMG), located in Southern and Eastern India, respectively. Sputum samples from study subjects were used for GeneXpert test (Cephid U.S.A.) following manufacturer’s instructions. GeneXpert positive cases were grouped as ATB and the negative ones as NTB (Figure 1a). Exclusion criteria for the subject recruitment were persons of < 15 years of age, taking immunosuppressive drugs, HIV +ve cases (based on ELISA test), Rifampicin drug resistance cases (as per GeneXpert results), and already under anti-TB medication. At least two follow-up samples at 2 months interval, until completion of therapy (6 months), were available for 11 out of 21 ATB subjects and included in the longitudinal part of the study. All of these follow-up subjects were receiving the same anti-TB drug regimen recommended under directly observed treatment short-course (DOTs) of revised national TB control program (RNTCP) in India. The drug regimen includes an intensive phase (2H3R3Z3E3) of 2 months and a continuation phase (4H3R3) of 4 months with thrice weekly doses of Isoniazide (H: 600 mg), Rifampicin (R: 450 mg), Pyrazinamide (Z: 1500 mg), and Ethambutol (E: 1200 mg). Subjects without a previous history of TB and not on any kind of medication in last 1 week were recruited from the International Centre for Genetic Engineering and Biotechnology (ICGEB), New Delhi to serve as asymptomatic healthy controls only for the longitudinal study. Sample Collection. Random, midstream urine samples (40 mL) were collected in 50 mL falcon tubes from the recruited subjects and kept in ice. Coded urine samples were centrifuged at 5000g for 10 min at 4 °C to collect supernatant for further processing. Protease inhibitors (33 μL sodium azide; 100 mM), 500 μL of phenylmethanesulfonyl fluoride (PMSF; 2%), and 1 μL of leupeptin (100 mM) per 50 mL of urine were added prior to storing them at −80 °C, and these steps were completed within 4 h of sample collection. Sample transportation and long-term storage until further analysis were carried out at −80 °C. Quality Control and Randomization. Following recommended protocols, an equal volume of all urine samples was pooled to prepare a quality control (QC).21,22 For the longitudinal study, urine samples from the active TB patients on follow-up (0, 2, 4, and 6 months on treatment) and asymptomatic healthy subjects were pooled, and a separate QC was prepared. Coded samples were randomized using a webbased tool (www.randomizer.org) to process those in batches, followed by GC−MS data acquisition within 24 h. Sample Processing. Five or six test urine samples with one QC in a batch were brought to room temperature by thawing in ice, avoiding more than two freeze−thaw cycles, before undertaking TMS (trimethylsilyl) derivatization following the method described by Chan et al. with minor modification (Figure 1b).23 Urine samples were processed in a −ve pressure laboratory meant for TB clinical sample handling at ICGEB following BSL-III level biosafety guidelines. To identify the feature of internal standard, that is, lysine D4 (Cambridge Isotope, USA), we spiked a healthy urine sample (male, 33 years) with an equal amount of lysine (Sigma, USA) and lysine D4 (5 μL of 10 mg/mL) before further processing. GC−MS Data Acquisition. Using an automatic liquid sampler (7683B ALS, Agilent, USA), 1 μL of derivatized urine sample was loaded using split less mode to an RTx-5 column (5% diphenyl, 95% dimethylpolysiloxane; 30 m × 0.25 mm × 1949

DOI: 10.1021/acs.jproteome.5b00016 J. Proteome Res. 2015, 14, 1947−1956

Article

Journal of Proteome Research

Figure 2. Urine metabolome isolated from active (ATB) and non-TB (NTB) tuberculosis patients show variation within and between study groups. (a) z-score plot of urine molecular features of ATB (red dots) and NTB (green dots) subjects showing variation of individual features within group. Each point represents its deviation from the mean (‘0’ at Y axis) among the observations. Variations of all 422 features are plotted on the X axis and variation within 21 observations, belonging to either study groups, are presented on the Y axis. (b) Score plot generated from partial least-squares discriminate analysis (PLS-DA) using all 422 variables from ATB (n = 21) and NTB (n = 21) to identify important features based on variable importance score (VIP > 1.5). (c) Heatmap with hierarchically clustered 42 important features selected based on both uni- and multivariate analyses shows individual variations in their peak intensities. The name of the important features is expanded for better view. (d) Area under the receiver operating characteristic curve (AUC of ROC) calculated to find accuracy of the model using a web-based tool ROCCET by selecting 42 important features.

the ICGEB, NIBMG, and CMC. Informed consent was obtained from all of the study subjects.

in databases (NIST and Golm Database). A few of the identities were confirmed by running commercial standards available using same separation and chemometric method. Pathway Analysis. Pathway analysis was carried out in MetaboAnalyst 2.0 to identify deregulated molecular pathways using identified molecules as input.24 An additional manual update from KEGG database and literature search was used to build pathways with maximum hits. Statistical Analysis Used in Longitudinal Study. A PLSDA model was built using molecular features resulting from the ATB subgroup followed up on for longitudinal study and asymptomatic healthy subjects. Data transformation and normalization were carried out following similar procedures previously discussed.



RESULTS

Performance of GC−MS Method and QC Samples

Molecular features of the internal standard lysine D4 and its normal isotopic form-lysine as observed in a test urine sample are presented (Figure S1 in the Supporting Information). Peak alignment of all raw data files generated in the study (total: n = 53; ATB: n = 21, NTB: n = 21, QC: n = 11) resulted in a metadata with 773 features (> 50% in any class). PCA showed that all 11 QC samples (run in 11 days) were tightly clustered (Figure S2 in the Supporting Information). Chemometrics To Identify Deregulated Molecular Features Associated with Tuberculosis

Ethical Statement

Features extracted from the ATB and NTB groups, present in > 50% in any class, result in a 42 × 422 matrix. From the z-score plot we observed that more variables of ATB are skewed

Study protocols used in this report were approved by the Institutional Review Board and human ethical committees of 1950

DOI: 10.1021/acs.jproteome.5b00016 J. Proteome Res. 2015, 14, 1947−1956

Article

Journal of Proteome Research

Table 2. List of Important Features with Fold Change (FC: ATB/NTB) at Statistical Significance (Wilcoxon−Mann−Whitney Test, p value) between the Groups (ATB/NTB) Used in the Case-Control Study, Variable Important Parameter (VIP) Score from Partial Least-Square Discriminate Analysis (PLS-DA), Tentative Identification, and Retention Index (RI) Calculated and Reported in Database Using RTX-5 equiv Column (NIST and Golm Database)a selected features

FC (ATB/NTB)

p value

VIP score

identification (tentative/confirmed*)b

RI (calculated)

RI (reported)

[email protected] [email protected] 75 [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] 11 [email protected] [email protected]

3.26 3.69 5.60 2.66 8.29 1.68 4.22 2.46 5.41 1.67 2.44 1.53 5.81 1.65 1.89 2.50 1.98 1.99 3.27 2.32 1.53 2.09 4.72 1.63 10.00 2.93 1.91 6.02 2.39 2.52 4.05 1.71 3.94 4.11 4.08 2.65 4.97 2.94 4.03 2.39 1.92 1.95

0.0008 0.046 0.0001 0.005 0.024 0.038 0.0002 0.038 0.005 0.02 0.012 0.01 0.013 0.014 0.025 0.004 0.01 0.013 0.025 0.007 0.049 0.024 0.033 0.025 0.024 0.049 0.049 0.020 0.010 0.003 0.006 0.007 0.043 0.015 0.0001 0.000 0.035 0.015 0.007 0.016 0.004 0.009

1.93 1.74 2.50 1.55 1.73 1.67 2.22 1.70 2.47 1.69 1.97 1.73 1.85 1.66 1.63 1.68 2.08 1.85 1.69 1.54 1.55 1.51 1.62 1.54 1.96 1.55 1.51 1.61 1.74 2.24 1.70 1.80 1.50 1.67 3.19 1.90 1.78 2.16 1.78 1.54 2.50 1.84

2,4-dihydroxybutyric acid NI NI lactic acid* GABA NI NI norepinephrine* NI 4-hydroxyhippuric acid NI NI NI NI NI malic acid* NI hydroquinone* NI 4-hydroxybenzoic acid* NI NI NI 4-hydroxyhippuric acid NI glucose* dihydroxybiphenyl NI NI gentisic acid 13-a-ethyl-3-oximinogon-4-ene-17-one NI NI NI phosphate phosphate phosphate phosphate glyceric acid NI NI NI

1419

1403

1077

1067

2202

2149

2236

2233

1463

1479

1405

1410

1634

1633

1634

1633

1931 1878

1932 NA

1832 1988

1892 NA

1237 1240 1240 1241 1332

1261 1261 1261 1261 1341

a NI: not identified; NA: not available in database; GABA: gamma-aminobutyric acid. bMolecules with asterisk (*) are confirmed from the in-house database generated from running standard molecules of major pathways.

A subset of 18 features was tentatively identified by NIST library (Figure S3 in the Supporting Information) and RI matching. Further confirmation was carried for six of these molecules using in-house library generated by running standards from commercial suppliers and matching their retention time and fragmentation pattern (RIs). Peak areas of each molecule in ATB and NTB groups are presented in Figure S4 in the Supporting Information.

toward positive scale than NTB (Figure 2a). On the basis of univariate (FC: ATB/NTB > 1.5 with a significant difference at p = 0.05 based on Wilcoxon-Mann−Whitney test) and multivariate analyses (variable important parameter: VIP > 1.5 of PLS-DA), 42 features were selected as important features (Figure 2b and Table 2). Differences in their relative abundances are presented in a heatmap (Figure 2c). The calculated area under the receiver operating characteristic curve (AUC of ROC) using these important features was 0.85 (95% confidence interval: CI: 0.72−0.96) with an accuracy of 0.79 in 100 cross validation test (Figure 2d).

Elucidation of Altered Biological Pathways from Deregulated Metabolites in Urine of Tuberculosis Patients

The pathway analysis revealed alteration of 18 metabolic activities with tyrosine metabolism having the maximum 1951

DOI: 10.1021/acs.jproteome.5b00016 J. Proteome Res. 2015, 14, 1947−1956

Article

Journal of Proteome Research

Figure 3. Pathway analysis of a subset of identified important molecules showed involvement of tyrosine−phenylalanine metabolism. (a) Norepinephrin, gentisic acid, hydroquinone, 4-hydroxybenzoic acid (4-HBA), and 4-hydroxyhippuric acid (4-HHA) are contributed from tyrosine and phenylalanine metabolism. Hydroxyquinone is produced from gentisic acid and 4-HBA. Metabolic conversion catalyzed by known enzyme expressed from host genome is shown in black arrows, and steps catalyzed by enzymes of gut microbe origin are presented in blue arrows. Spearman Rank correlation coefficients between identified metabolites are shown with #. (b) Relative abundances of five identified urine metabolites involved in the tyrosine−phenylalanine pathway are presented in box and whisker plot (*, **, ***: p value < 0.05, < 0.01, and < 0.005, respectively, of Wilcoxon-Mann−Whitney test).

number of hits (three) (Table S1 in the Supporting Information). With supporting information from the KEGG database and literature search, we could link five molecules, viz. norepinephrine, gentisic acid, 4-hydroxybenzoic acid (4-HBA), hydroquinone, and 4-hydroxyhippuric acid (4-HHA), to tyrosine−phenylalanine metabolism (Figure 3a). The relative abundances of these molecules in ATB and NTB are presented in Figure 3b. We identified a higher abundance of 4-HBA and 4-HHA in ATB than NTB with a Spearman rank correlation value of 0.74 (p value = 0.0001, FDR: 0.003) between these two molecules. Decarboxylation of 4-HBA gives rise to hydroquinone, and we observed a Spearman rank correlation of 0.55 (p value = 0.01, FDR: 0.14) between them. Most of the steps in tyrosine metabolism are catalyzed by enzymes that are produced by the host, whereas several steps involved in phenylalanine metabolism, as obtained from KEGG, literature,

and the NCBI database analyses, have been reported to be catalyzed by enzymes of gut microbe origin (Figure 3a). Effect of Natural Variations on the Identified Important Molecules

Natural variations such as age, body mass index (BMI), geographical diversity, and smoking habit did not influence the separation of the two study groups using these selected 42 features (Figure S5a−d in the Supporting Information). Antituberculosis Therapeutic Intervention Leads to Alteration in the Urine Molecular Profiles of Tuberculosis Patients

PLS-DA score plot using all extracted features (294) from the longitudinal study showed a specific trend from 0 to 6 months of receiving treatment and healthy controls cluster away from freshly diagnosed TB patients (Figure 4a,b). However, we observed that the urine profile of subjects completing 1952

DOI: 10.1021/acs.jproteome.5b00016 J. Proteome Res. 2015, 14, 1947−1956

Article

Journal of Proteome Research

Figure 4. Follow-up urine samples of fresh TB patients undertaking therapeutic interventions and healthy control subjects showed altered metabolic phenotypes. (a) 11 TB active subjects were followed up on for 6 months with 2 month intervals, and healthy subjects (n = 11) were recruited for comparison with clinically cured subjects completing therapeutic interventions. (b) Score plot obtained from a partial least-squares discriminate analysis (PLS-DA) including subjects receiving TB therapy from a fresh case (0 M) and at different intervals of 2, 4 and 6 months (0 M, 2 M, and 6 M) of intervention and healthy (n = 11) subjects shows a trend. The three components on three axes explained a total variance of 31.1%. The fresh TB subjects’ metabolic phenotypes are significantly different from healthy subjects and clinically cured subjects. However, a certain degree of similarity between healthy and clinically cured subjects was also observed.

Extracted molecular features showed a greater number of variables in ATB samples were skewed to the positive scale, indicating greater metabolic heterogeneity within the freshly detected TB cases than in non-TB. Elucidating the possible role of the identified deregulated molecules, we observed altered metabolism of essential amino acid phenylalanine and semiessential amino acid tyrosine in TB patients. Previous studies on TB patients and Mtb-infected animal models showed change in the abundance of phenylalanine and tyrosine levels in serum samples.9,12−14 However, we observed insignificant variation in abundances of tyrosine (FC: ATB/NTB = 2.2 at p value: 0.33) and phenylalanine levels (FC: ATB/NTB = 1.2 and p value: 0.09) in urine of TB patients. Interestingly, some downstream products of metabolism of these two amino acids showed higher abundance in ATB when compared with NTB. Higher abundance of 4-HHA and 4-HBA, that is, breakdown products of phenyl alanine, was also reported in the urine of autistic subjects.28 Glycine conjugation in liver is an important mechanism to eliminate unwanted carboxylic acids such as salicylic and benzoic acids. While 4-HHA could be formed in liver or kidney by glycine conjugation of 4-HBA,29 enzymes that catalyze these steps from phenylalanine to 4-HBA are yet to be reported in human. However, gut microbes catalyze these specific reactions and may contribute to host biofluids.30 Clayton et al. hypothesized that deregulated gut microbial activity may have contributed to higher 4-HBA and 4-HHA levels in urine of autistic patient.31 In a recent report, altered gut microbe diversity was demonstrated in Mtb-infected mice models.32 Our findings of increased downstream waste products of phenylalanine metabolism indicate a possible role of altered gut microbiota in TB patients; however, that needs to be established in a separate focused study including information on dietary, body mass, and environmental factors that influence gut microbial−mammalian cometabolic pathways. Additionally, the level of chronic lung pathology information on TB patients

therapeutic interventions and declared as clinically cured were closer with healthy subjects (Figure 4b).



DISCUSSION

In this pilot study we carried out a nontargeted comparative metabolic profiling to identify deregulated molecules in urine of diagnostically important study groups sharing similar clinical symptoms, that is, ATB and NTB. To minimize role of confounders like dietary habits and environmental factors, we analyzed a similar number of subjects in ATB and NTB from each clinical site. In the longitudinal study, we followed TB patients undertaking therapeutic interventions to explore if any treatment-dependent metabolic phenotypes exist. Such studies may demonstrate the importance of metabolic profiling to elucidate altered host metabolic activities in TB. Recent publications have demonstrated the importance of metabolic profiling to understand TB pathogenesis in animal models and human.9−15 However, our study includes ATB and NTB groups having similar symptoms unlike ATB and healthy compared in previous studies.9,10,14,15 This was carried out with an intention to find a TB-specific metabolic phenotype other than those that could be contributed by inflammation and stress response. This could be considered as the first report of ATB-NTB metabolic comparison study using GC−MS following silylation. Close clustering of QC samples showed negligible variation related to methodology adopted for this study. In the urine metabolite profiling study, creatinine normalization is routinely used; however, few reports have shown varied concentration of creatinine and its precursor creatine levels in serum of Mtbinfected animal models and TB patients.9,12−14 Therefore, we followed an alternate method of total area normalization to adjust for the dilution effect that we expect from random sample collection and diurnal variation. Female urine metabolic profiles change significantly during the menstrual/reproductive cycles of the subject.26,27 Therefore, only male subjects were included in this study to limit these confounding factors. 1953

DOI: 10.1021/acs.jproteome.5b00016 J. Proteome Res. 2015, 14, 1947−1956

Journal of Proteome Research

Article



CONCLUSIONS Comparative urine metabolome analysis using silylation and GC−MS demonstrated that differences do exist between clinically confirmed active TB patients and non-TB subjects. It appears that the tyrosine−phenylalanine metabolic axis toward waste production is shifted in TB. Because these metabolic activities are mediated by enzymes expressed by host as well as gut microbe genome, our findings support a possible role of altered gut microbiota in TB pathogenesis. We have also demonstrated a time-dependent trend that exists in the metabolic phenotype of TB subjects undertaking therapy. The outcome of this study opens up new avenues to target the tyrosine-phenylalanine axis by modulating microbiome or host metabolism as an improved strategy for better physiological recovery of TB patients in future.

may also aid in understanding whether metabolic changes are due to disease or associated pathologies including cachexia or both. The lack of individual dietary details and information on weight lost due to disease onset of all study subjects are limitations of this study. Excretion of higher levels of norepinephrine in biofluid of TB patients, as observed in our study, corroborates previous reports.33,34 In a previous study, higher concentration of normetanephrine, a derivative of norepinephrine, in sputum of TB patients was also reported.10 Catecholamines, like norepinephrine and normetanephirne, are found to be released by both innate and adaptive immune cells, in contrast with classifying them only as neurotransmitters and hormones. Their role in immunity and inflammation has been evidently reported in many studies.35,36 Interestingly Alaniz et al. have shown that dopamine β-hydroxylase knockout mice that do not produce epinephrine and norpeineprhine are more susceptible to TB.37 Although previous reports demonstrated low phosphate concentration in serum of TB patients,38 we observed higher phosphate levels in urine of TB patients. Reabsorption of phosphate in kidney is regulated by parathyroid hormone, fibroblast growth factor 23, and 1,25-dihydroxyvitamin D.39 Because vitamin D deficiency plays a key role in TB pathogenesis, it may contribute to lowering renal reabsorption of phosphate, leading to higher urine phosphate excretion.40 A subgroup of important molecules, gamma-aminobutyric acid (GABA), glucose, and lactic acid, corroborates previous reports on their association with TB pathogenesis.10,13−15 duPreez and Loots demonstrated increased GABA levels in sputum of TB patients and predicted it as a host response molecule that fuels the citric acid cycle by supplying succinic acid.10 GABA metabolizes to succinic acid through succinic acid semialdehyde by catalytic action of succinic acid semialdehyde dehydrogenase. Interestingly, succinic acid semialdehyde dehydrogenase deficiency caused elevated urinary excretion of 2,4-dihydroxybutyric acid, another important molecule identified in this study.41 In the current study, we attempted to generate proof-of-concept data of the effect of therapeutic intervention on metabolites of urine of the curable ATB group. We observed a treatment-dependent systematic alteration of these metabolites, and these changes could possibly be contributed by molecules from therapy-induced altered host metabolism and products of metabolized antibiotics. Although the sample size in this study is low (n = 11), our findings support previous reports of Mahapatra et al.11 and, more importantly, the urine profile of TB patients completing therapy was found to be closer to healthy subjects with limited similarity and thus provides an important insight and impetus to carry out a more focused study with a large sample size. Because 6 month follow-up samples were collected within 48 h of taking the last dose of medication, it would be interesting to follow up on them beyond the treatment completion period to nullify the contribution of TB drug metabolites in the urine of clinically cured subjects while comparing them with asymptomatic healthy controls. Immunomolecular profiling studies also demonstrated differences in the levels of interleukine-2 receptor (IL-2R), activation of T cells, cytokine profile, and sICAM-1 between healthy and TB patients completing medications.42−44 Similar studies from sites with higher drug-resistant Mtb incidents and in an effort to find differences that may exist between drug-resistant and -susceptible cases may yield interesting insights and are part of an ongoing effort in our laboratory.



ASSOCIATED CONTENT

S Supporting Information *

Table S1: Epidemiological metadata of subjects used in the study. Table S2: List of identified deregulated pathways in tuberculosis patients as resulted from pathway analysis using 14 important molecules. Figure S1: Total ion chromatogram of internal standard spiked healthy urine sample. Figure S2: Validation of repeatability of methods used for derivatization of urine samples and data acquisition using GC−MS by running QC samples in every batch. Figure S3: Tentative identification of molecules by matching composite spectra of extracted from MPP and those matching similar RT and EI-mass spectra in ChromaTOF and library database. Figure S4: Variation in abundance of each identified important molecular feature between active (ATB) and non-TB (NTB) subjects is observed. Figure S5: Influence of natural variations such as age, body mass index (BMI), geographical locations, and smoking habit on the identified important molecular features are found to be insignificant. This material is available free of charge via the Internet at http://pubs.acs.org.



AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Tel: 00-91-11-26741358. Present Addresses ∥

A.D.: Malaria Group, International Centre for Genetic Engineering and Biotechnology (ICGEB), Aruna Asaf Ali Marg, New Delhi, India. ⊥ D.D.: Centre for Cellular and Molecular Platforms, National Centre for Biological Sciences and Tata Institute of Fundamental Research, Bellary Road, Bangalore 560065, India. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS Financial support from the Bill and Melinda Gates Foundation, Grand Challenges Canada and Department of Biotechnology, and Government of India to Ranjan Kumar Nanda are acknowledged. Critical suggestion of Dr Suresh Nair, Staff Research Scientist, International Centre for Genetic Engineering and Biotechnology, New Delhi on the manuscript is acknowledged.



REFERENCES

(1) WHO. Global Tuberculosis Report 2014; WHO: Geneva, 2014.

1954

DOI: 10.1021/acs.jproteome.5b00016 J. Proteome Res. 2015, 14, 1947−1956

Article

Journal of Proteome Research (2) Kaufmann, S. H.; Lange, C.; Rao, M.; Balaji, K. N.; Lotze, M.; Schito, M.; Zumla, A. I.; Maeurer, M. Progress in tuberculosis vaccine development and host-directed therapies–a state of the art review. Lancet Respir. Med. 2014, 2 (4), 301−320. (3) Smith, R. Eradication of tuberculosis by 2050 impossible without new vaccine. BMJ 2009, 338, b1291. (4) Herrera, V.; Perry, S.; Parsonnet, J.; Banaei, N. Clinical application and limitations of interferon-gamma release assays for the diagnosis of latent tuberculosis infection. Clin. Infect. Dis. 2011, 52 (8), 1031−1037. (5) M?ller, M.; Hoal, E. G. Current findings, challenges and novel approaches in human genetic susceptibility to tuberculosis. Tuberculosis 2010, 90 (2), 71−83. (6) Zumla, A. I.; Gillespie, S. H.; Hoelscher, M.; Philips, P. P.; Cole, S. T.; Abubakar, I.; McHugh, T. D.; Schito, M.; Maeurer, M.; Nunn, A. J. New antituberculosis drugs, regimens, and adjunct therapies: needs, advances, and future prospects. Lancet Infect. Dis. 2014, 14 (4), 327− 340. (7) Blankley, S.; Berry, M. P.; Graham, C. M.; Bloom, C. I.; Lipman, M.; O’Garra, A. The application of transcriptional blood signatures to enhance our understanding of the host response to infection: the example of tuberculosis. Philos. Trans. R. Soc., B 2014, 369 (1645), 20130427. (8) Yanes, O.; Clark, J.; Wong, D. M.; Patti, G. J.; Sanchez-Ruiz, A.; Benton, H. P.; Trauger, S. A.; Desponts, C.; Ding, S.; Siuzdak, G. Metabolic oxidation regulates embryonic stem cell differentiation. Nat. Chem. Biol. 2010, 6 (6), 411−417. (9) Che, N.; Cheng, J.; Li, H.; Zhang, Z.; Zhang, X.; Ding, Z.; Dong, F.; Li, C. Decreased serum 5-oxoproline in TB patients is associated with pathological damage of the lung. Clin. Chim. Acta 2013, 423, 5−9. (10) du Preez, I.; Loots, D. T. New sputum metabolite markers implicating adaptations of the host to Mycobacterium tuberculosis, and vice versa. Tuberculosis 2013, 93 (3), 330−337. (11) Mahapatra, S.; Hess, A. M.; Johnson, J. L.; Eisenach, K. D.; DeGroote, M. A.; Gitta, P.; Joloba, M. L.; Kaplan, G.; Walzl, G.; Boom, W. H.; Belisle, J. T. A metabolic biosignature of early response to antituberculosis treatment. BMC Infect. Dis. 2014, 14, 53. (12) Shin, J. H.; Yang, J. Y.; Jeon, B. Y.; Yoon, Y. J.; Cho, S. N.; Kang, Y. H.; Ryu do, H.; Hwang, G. S. (1)H NMR-based metabolomic profiling in mice infected with Mycobacterium tuberculosis. J. Proteome Res. 2011, 10 (5), 2238−2247. (13) Somashekar, B. S.; Amin, A. G.; Rithner, C. D.; Troudt, J.; Basaraba, R.; Izzo, A.; Crick, D. C.; Chatterjee, D. Metabolic profiling of lung granuloma in Mycobacterium tuberculosis infected guinea pigs: ex vivo 1H magic angle spinning NMR studies. J. Proteome Res. 2011, 10 (9), 4186−4195. (14) Weiner, J., 3rd; Parida, S. K.; Maertzdorf, J.; Black, G. F.; Repsilber, D.; Telaar, A.; Mohney, R. P.; Arndt-Sullivan, C.; Ganoza, C. A.; Fae, K. C.; Walzl, G.; Kaufmann, S. H. Biomarkers of inflammation, immunosuppression and stress with active disease are revealed by metabolomic profiling of tuberculosis patients. PLoS One 2012, 7 (7), e40221. (15) Zhou, A.; Ni, J.; Xu, Z.; Wang, Y.; Lu, S.; Sha, W.; Karakousis, P. C.; Yao, Y. F. Application of (1)h NMR spectroscopy-based metabolomics to sera of tuberculosis patients. J. Proteome Res. 2013, 12 (10), 4642−4649. (16) Want, E. J.; Wilson, I. D.; Gika, H.; Theodoridis, G.; Plumb, R. S.; Shockcor, J.; Holmes, E.; Nicholson, J. K. Global metabolic profiling procedures for urine using UPLC-MS. Nat. Protoc. 2010, 5 (6), 1005−1018. (17) Berry, M. P.; Graham, C. M.; McNab, F. W.; Xu, Z.; Bloch, S. A.; Oni, T.; Wilkinson, K. A.; Banchereau, R.; Skinner, J.; Wilkinson, R. J.; Quinn, C.; Blankenship, D.; Dhawan, R.; Cush, J. J.; Mejias, A.; Ramilo, O.; Kon, O. M.; Pascual, V.; Banchereau, J.; Chaussabel, D.; O’Garra, A. An interferon-inducible neutrophil-driven blood transcriptional signature in human tuberculosis. Nature 2010, 466 (7309), 973−977. (18) De Groote, M. A.; Nahid, P.; Jarlsberg, L.; Johnson, J. L.; Weiner, M.; Muzanyi, G.; Janjic, N.; Sterling, D. G.; Ochsner, U. A.

Elucidating novel serum biomarkers associated with pulmonary tuberculosis treatment. PLoS One 2013, 8 (4), e61002. (19) Bouatra, S.; Aziat, F.; Mandal, R.; Guo, A. C.; Wilson, M. R.; Knox, C.; Bjorndahl, T. C.; Krishnamurthy, R.; Saleem, F.; Liu, P.; Dame, Z. T.; Poelzer, J.; Huynh, J.; Yallou, F. S.; Psychogios, N.; Dong, E.; Bogumil, R.; Roehring, C.; Wishart, D. S. The human urine metabolome. PLoS One 2013, 8 (9), e73076. (20) Banday, K. M.; Pasikanti, K. K.; Chan, E. C.; Singla, R.; Rao, K. V.; Chauhan, V. S.; Nanda, R. K. Use of urine volatile organic compounds to discriminate tuberculosis patients from healthy subjects. Anal. Chem. 2011, 83 (14), 5526−5534. (21) Sangster, T.; Major, H.; Plumb, R.; Wilson, A. J.; Wilson, I. D. A pragmatic and readily implemented quality control strategy for HPLCMS and GC-MS-based metabonomic analysis. Analyst 2006, 131 (10), 1075−1078. (22) Dunn, W. B.; Broadhurst, D.; Begley, P.; Zelena, E.; FrancisMcIntyre, S.; Anderson, N.; Brown, M.; Knowles, J. D.; Halsall, A.; Haselden, J. N.; Nicholls, A. W.; Wilson, I. D.; Kell, D. B.; Goodacre, R. Procedures for large-scale metabolic profiling of serum and plasma using gas chromatography and liquid chromatography coupled to mass spectrometry. Nat. Protoc. 2011, 6 (7), 1060−1083. (23) Chan, E. C.; Pasikanti, K. K.; Nicholson, J. K. Global urinary metabolic profiling procedures using gas chromatography-mass spectrometry. Nat. Protoc. 2011, 6 (10), 1483−1499. (24) Xia, J.; Wishart, D. S. Web-based inference of biological patterns, functions and pathways from metabolomic data using MetaboAnalyst. Nat. Protoc. 2011, 6 (6), 743−760. (25) Xia, J.; Broadhurst, D. I.; Wilson, M.; Wishart, D. S. Translational biomarker discovery in clinical metabolomics: an introductory tutorial. Metabolomics 2013, 9 (2), 280−299. (26) Wallace, M.; Hashim, Y. Z.; Wingfield, M.; Culliton, M.; McAuliffe, F.; Gibney, M. J.; Brennan, L. Effects of menstrual cycle phase on metabolomic profiles in premenopausal women. Hum. Reprod. 2010, 25 (4), 949−956. (27) Xu, X.; Duncan, A. M.; Merz-Demlow, B. E.; Phipps, W. R.; Kurzer, M. S. Menstrual cycle effects on urinary estrogen metabolites. J. Clin. Endocrinol. Metab. 1999, 84 (11), 3914−3918. (28) Lis, A. W.; McLaughlin, I.; Mpclaughlin, R. K.; Lis, E. W.; Stubbs, E. G. Profiles of ultraviolet-absorbing components of urine from autistic children, as obtained by high-resolution ion-exchange chromatography. Clin. Chem. 1976, 22 (9), 1528−1532. (29) Wikoff, W. R.; Anfora, A. T.; Liu, J.; Schultz, P. G.; Lesley, S. A.; Peters, E. C.; Siuzdak, G. Metabolomics analysis reveals large effects of gut microflora on mammalian blood metabolites. Proc. Natl. Acad. Sci. U. S. A. 2009, 106 (10), 3698−3703. (30) Nicholson, J. K.; Holmes, E.; Kinross, J.; Burcelin, R.; Gibson, G.; Jia, W.; Pettersson, S. Host-gut microbiota metabolic interactions. Science 2012, 336 (6086), 1262−1267. (31) Clayton, T. A. Metabolic differences underlying two distinct rat urinary phenotypes, a suggested role for gut microbial metabolism of phenylalanine and a possible connection to autism. FEBS Lett. 2012, 586 (7), 956−961. (32) Winglee, K.; Eloe-Fadrosh, E.; Gupta, S.; Guo, H.; Fraser, C.; Bishai, W. Aerosol Mycobacterium tuberculosis infection causes rapid loss of diversity in gut microbiota. PLoS One 2014, 9 (5), e97048. (33) Vaccarezza, J. R.; Ruiz, D. C. The excretion of a urinary metabolite of epinephrine and norepinephrine in tuberculosis. Am. Rev. Respir. Dis. 1969, 100 (3), 398−400. (34) Hafeiz, A. A.; Issa, H. A.; el-Kammah, B.; Abdel-Hafez, M. A.; Abdou, M. A.; Abdel-Khalek, M.; Ramadan, S. M. Plasma catecholamines in pulmonary tuberculosis. Kekkaku 1992, 67 (10), 647−652. (35) Flierl, M. A.; Rittirsch, D.; Nadeau, B. A.; Chen, A. J.; Sarma, J. V.; Zetoune, F. S.; McGuire, S. R.; List, R. P.; Day, D. E.; Hoesel, L. M.; Gao, H.; Van Rooijen, N.; Huber-Lang, M. S.; Neubig, R. R.; Ward, P. A. Phagocyte-derived catecholamines enhance acute inflammatory injury. Nature 2007, 449 (7163), 721−725. (36) Bergquist, J.; Tarkowski, A.; Ekman, R.; Ewing, A. Discovery of endogenous catecholamines in lymphocytes and evidence for 1955

DOI: 10.1021/acs.jproteome.5b00016 J. Proteome Res. 2015, 14, 1947−1956

Article

Journal of Proteome Research catecholamine regulation of lymphocyte function via an autocrine loop. Proc. Natl. Acad. Sci. U. S. A. 1994, 91 (26), 12912−12916. (37) Alaniz, R. C.; Thomas, S. A.; Perez-Melgosa, M.; Mueller, K.; Farr, A. G.; Palmiter, R. D.; Wilson, C. B. Dopamine beta-hydroxylase deficiency impairs cellular immunity. Proc. Natl. Acad. Sci. U. S. A. 1999, 96 (5), 2274−2278. (38) Rohini, K.; Bhat, S.; Srikumar, P. S.; Mahesh Kumar, A. Assessment of serum calcium and phosphorus in pulmonary tuberculosis patients before, during and after chemotherapy. Indian J. Clin. Biochem. 2014, 29 (3), 377−381. (39) Fukumoto, S. Phosphate metabolism and vitamin D. Bonekey Rep. 2014, 3, 497. (40) Gibney, K. B.; MacGregor, L.; Leder, K.; Torresi, J.; Marshall, C.; Ebeling, P. R.; Biggs, B. A. Vitamin D deficiency is associated with tuberculosis and latent tuberculosis infection in immigrants from subSaharan Africa. Clin. Infect. Dis. 2008, 46 (3), 443−446. (41) Brown, G. K.; Cromby, C. H.; Manning, N. J.; Pollitt, R. J. Urinary organic acids in succinic semialdehyde dehydrogenase deficiency: evidence of alpha-oxidation of 4-hydroxybutyric acid, interaction of succinic semialdehyde with pyruvate dehydrogenase and possible secondary inhibition of mitochondrial beta-oxidation. J. Inherited Metab. Dis. 1987, 10 (4), 367−375. (42) Chan, C. H.; Lai, C. K.; Leung, J. C.; Ho, A. S.; Lai, K. N. Elevated interleukin-2 receptor level in patients with active pulmonary tuberculosis and the changes following anti-tuberculosis chemotherapy. Eur. Respir. J. 1995, 8 (1), 70−73. (43) da Silva, M. V.; Figueiredo, A. A.; Machado, J. R.; Castellano, L. C.; Alexandre, P. B.; Oliveira, R. F.; Faria, G. E.; Pereira, S. A.; Rodrigues, D. B.; Rodrigues, V., Jr. T Cell Activation and Proinflammatory Cytokine Production in Clinically Cured Tuberculosis Are Time-Dependent and Accompanied by Upregulation of IL10. PLoS One 2013, 8 (6), e65492. (44) Demir, T.; Yalcinoz, C.; Keskinel, I.; Demiroz, F.; Yildirim, N. sICAM-1 as a serum marker in the diagnosis and follow-up of treatment of pulmonary tuberculosis. Int. J. Tuberc. Lung Dis. 2002, 6 (2), 155−159.

1956

DOI: 10.1021/acs.jproteome.5b00016 J. Proteome Res. 2015, 14, 1947−1956