A Novel Urinary Metabolite Signature for Diagnosing Major

Article pubs.acs.org/jpr

A Novel Urinary Metabolite Signature for Diagnosing Major Depressive Disorder Peng Zheng,†,‡,#,§ Jian-jun Chen,†,‡,#,§ Ting Huang,†,‡,#,§ Ming-ju Wang,†,‡,# Ying Wang,†,‡,# Mei-xue Dong,†,‡,# Yuan-jun Huang,†,‡,# Lin-ke Zhou,†,‡,# and Peng Xie*,†,‡,# †

Department of Neurology, The First Affiliated Hospital of Chongqing Medical University, Chongqing, China Chongqing Key Laboratory of Neurobiology, Chongqing, China # Institute of Neuroscience, Chongqing Medical University, Chongqing, China ‡

S Supporting Information *

ABSTRACT: Major depressive disorder (MDD) is a prevalent and debilitating mental disorder. Yet, there are no objective biomarkers available to support diagnostic laboratory testing for this disease. Here, gas chromatography−mass spectrometry was applied to urine metabolic profiling of 126 MDD and 134 control subjects. Orthogonal partial least-squares discriminant analysis (OPLS-DA) was used to identify the differential metabolites in MDD subjects relative to healthy controls. The OPLS-DA analysis of data from training samples (82 first-episode, drug-naı̈ve MDD subjects and 82 well-matched healthy controls) showed that the depressed group was significantly distinguishable from the control group. Totally, 23 differential urinary metabolites responsible for the discrimination between the two groups were identified. Postanalysis, 6 of the 23 metabolites (sorbitol, uric acid, azelaic acid, quinolinic acid, hippuric acid, and tyrosine) were defined as candidate diagnostic biomarkers for MDD. Receiver operating characteristic analysis of combined levels of these six biomarkers yielded an area under the receiver operating characteristic curve (AUC) of 0.905 in distinguishing training samples; this simplified metabolite signature classified blinded test samples (44 MDD subjects and 52 healthy controls) with an AUC of 0.837. Furthermore, a composite panel by the addition of previously identified urine biomarker (N-methylnicotinamide) to this biomarker panel achieved a more satisfactory accuracy, yielding an AUC of 0.909 in the training samples and 0.917 in the test samples. Taken together, these results suggest this composite urinary metabolite signature should facilitate development of a urine-based diagnostic test for MDD. KEYWORDS: major depressive disorder, major depression, MDD, biomarker, diagnosis, diagnostic, metabonomic, gas chromatography−mass spectrometry, GC−MS, nuclear magnetic resonance, NMR

■

INTRODUCTION Major depressive disorder (MDD) is a common and lifethreatening mental disorder, which affects up to 15% of the general population and is a vital cause of disability worldwide.1 Currently, MDD diagnosis solely relies on symptomatic clusters and other clinical features, as there are no objective laboratorybased tests available. The current diagnostic method results in high rates of underdiagnosis, delayed diagnosis, and misdiagnosis on account of the high heterogeneity of clinical symptoms.2 Thus, the identification of objective biomarkers for MDD would be of considerable clinical value in facilitating the development of laboratory-based diagnostic tests for MDD, © 2013 American Chemical Society

while also improving our understanding of the underlying pathophysiological mechanisms of the disease. Metabonomics, which focuses on the quantitative measurement of endogenous metabolites in biosamples such as plasma and urine,3 has been increasingly used to capture diseasespecific metabolic signatures as possible biomarkers.4 Metabolic profiling and metabolite biomarkers have been successfully used in the discrimination or diagnosis of various neuropsychiatric disorders such as stroke, Parkinson’s disease, bipolar disorder, Received: September 13, 2013 Published: November 14, 2013 5904

dx.doi.org/10.1021/pr400939q | J. Proteome Res. 2013, 12, 5904−5911

Journal of Proteome Research

Article

Table 1. Demographic and Clinical Details of Recruited Subjectsa training set sample size sex (M/F) age (year)c BMIc HDRS scoresc

test set

HC

MDD

Pb

HC

MDD

Pb

82 53/29 34.2 ± 10.1 20.9 ± 2.6 0.2 ± 0.6

82 46/36 32.2 ± 10.3 21.6 ± 2.7 22.4 ± 4.6

0.26 0.20 0.15 0.00

52 27/25 28.8 ± 9.8 21.3 ± 2.4 0.3 ± 0.7

44 17/27 34.1 ± 9.1 22.1 ± 3.1 25.7 ± 3.9

0.19 0.01 0.14 0.00

a

Abbreviations: HC, healthy controls; MDD, major depressive disorder; Y/N, yes/no; M/F, male/female; BMI, body mass index; HDRS, Hamilton Depression Rating Scale. bTwo-tailed Student’s t-test for continuous variables (age, BMI, and HDRS scores); Chi-square analysis for categorical variables (sex). cValues expressed as means ± SDs.

autism, and schizophrenia.5−11 Previously, this method has also been used to characterize the metabolic changes in depressive animal models,12−16 which provides valuable clues in uncovering the molecular mechanisms of depression. In this group’s recent nuclear magnetic resonance (NMR) spectroscopy-based metabolomic studies on MDD, a panel of metabolite biomarkers capable of discriminating MDD subjects from healthy controls with high accuracy was identified in both the plasma and urine.17,18 The two studies exemplified the ability of metabolomics to identify diagnostic biomarkers for MDD from clinical samples. Given that a single analytical technology cannot provide complete coverage of the human metabonome due to the diverse physicochemical properties and wide concentration ranges, 19 it is essential to apply complementary metabolomic platforms to identify novel biomarkers for MDD. This complementation has been proven to be especially valuable for psychiatric disorders such as schizophrenia.7 Here, a gas chromatography−mass spectrometry (GC−MS)based metabonomic platform was therefore used to profile metabolites in urine samples from 126 MDD subjects and 134 healthy controls. The first objective of this study was to identify the differential urinary metabolites in MDD subjects relative to healthy controls using highly homogeneous training samples (82 first-episode, drug-naı̈ve MDD subjects and 82 wellmatched healthy controls). The second objective was to optimize a simplified metabolite signature for MDD and independently validate its diagnostic performance in diverse test samples including medicated samples (44 MDD subjects and 52 healthy controls). Lastly, a combined analysis of the biomarkers identified by NMR and GC−MS technologies was performed to find a robust diagnostic biomarker panel for MDD.

■

history of neurological, DSM-IV Axis I/II, or systemic medical illness. In total, 126 MDD subjects and 134 healthy controls were recruited into this study. The same samples obtained from these subjects had been used to identify and validate urinary biomarkers for MDD in a previous NMR-based metabolomic study. Briefly, the 126 MDD subjects and 134 healthy controls were divided into a training set and a test set. The majority of these MDD subjects (n = 95) were first-episode and drugnaı̈ve, and the remaining MDD subjects (n = 31) were being treated with various antidepressants. The training set, composed of 82 first-episode, drug-naı̈ve MDD subjects and 82 healthy controls, was used to identify urinary diagnostic markers for MDD; the remaining subjects were used to construct the test set to independently validate the diagnostic generalizability of these biomarkers. The detailed demographic and clinical data of the participants are presented in Table 1. All MDD subjects scored significantly higher on the HDRS than all healthy controls in both the training and test sets. In both sets, the MDD group and healthy control group did not significantly differ in gender and BMI. As for age, MDD subjects and healthy controls were matched in the training set, but not in the test set. Prior to sample collection, written informed consents were obtained from all subjects. The protocols of this study were reviewed and approved by the Ethical Committee of Chongqing Medical University. Sample Preparation and GC−MS Acquisition

After overnight fasting, morning urine samples were collected with a sterile cup and transferred into a sterile tube. All urine samples were then centrifuged at 1500g for 10 min. The resulting supernatant was divided into equal aliquots and stored at −80 °C until later analysis. Prior to GC−MS analysis, a 15 μL aliquot of urine was vortexed after adding 10 μL of internal standard solution (Lleucine-13C6, 0.02 mg/mL). Then, 15 μL of urease was added into this mixed solution. The urea was degraded for 60 min at 37 °C. The mixture was extracted with 240 μL of ice-cold methanol and then 80 μL of ice-cold methanol. After vortexing for 30 s, the mixture was centrifuged at 14000 rpm for 5 min at 4 °C. The 224 μL supernatant was transferred to a glass vial and vacuum-dried at room temperature. The dried metabolic extract was derivatized first with 30 μL of methoxyamine (20 mg/mL) for 1.5 h at 37 °C. Subsequently, 30 μL of N,Obis(trimethylsilyl) trifluoroacetamide (BSTFA) with 1% trimethylchlorosilane (TCMS) was added to the mixture and heated for 1 h at 70 °C to form trimethylsilyl (TMS) derivatives. After derivatization and cooling to room temperature, 1.0 μL of this derivative was injected into the GC−MS system for analysis.

MATERIALS AND METHODS

Participants

All MDD subjects were recruited from the psychiatric center of the First Affiliated Hospital at Chongqing Medical University. MDD diagnosis relied on a Structured Psychiatric Interview using DSM-IV-TR criteria.20 Depression severity was assessed by the 17-item version of the observer-rated Hamilton Depression Rating Scale (HDRS).21 Only depressed subjects with HDRS scores of greater than 17 were enrolled. Exclusion criteria for MDD subjects included any pre-existing physical or other mental disorders and/or illicit drug use. Healthy control subjects were recruited from the medical examination center of First Affiliated Hospital at Chongqing Medical University. Healthy controls were required to display no previous lifetime 5905



Article

Figure 1. The OPLS-DA model. (A) The scores plots of the OPLS-DA model showing an obvious separation between 82 first-episode, drug-naı̈ve MDD subjects (green diamonds) and 82 healthy controls (red triangles). (B) A 300-iteration permutation test showing that the values of permuted R2 and Q2 (bottom left) were significantly lower than the corresponding original R2 and Q2 values (top right), suggesting the constructed OPLS-DA model was valid and not-overfitted.

was performed.18 If the values of Q2 and R2 resulting from the original model were higher than the corresponding values from the permutation test, the model was deemed valid.25

GC−MS Analysis

GC−MS analysis was carried out according to this group’s previously published work with minor modifications.14 Briefly, each 1 μL of the derivative solution was injected into an Agilent 7980 GC system (Agilent Technologies Inc., USA). An HP-5 MS fused silica capillary column (30 m × 0.25 mm × 0.25 μm, Agilent, USA) was used for the separation with helium carrier gas at a flow rate of 1 mL/min. The injector temperature was set at 280 °C. The column temperature was initially kept at 80 °C for 2 min and then increased from 80 to 320 °C at 10 °C/ min, where it was held for 6 min. The column effluent was introduced into the ion source of an Agilent 5975 mass selective detector (Agilent Technologies). The MS quadrupole temperature was set at 150 °C, and the ion source temperature was set at 230 °C. Data acquisition was performed in the full scan mode from m/z 50 to 550.

Identification of Urinary Biomarkers for MDD

As a diagnosis based on quantification of a small number of metabolites would be more economically feasible and convenient in clinical practice,7 a stepwise optimization algorithm based on Bayesian Information Criterion (BIC) was employed to optimize the metabolite biomarker combination.26 This procedure could identify the simplified diagnostic metabolite signature and construct a discriminative model: P(Y=1)=1/(1+e-y); y=(−360.85)x1+(−1.58)x2+(−0.99)x3+(340.14)x4+(1.59)x5+(5.22)x6−0.57. Applying this model, the probability of illness in each sample could be calculated. In addition, to find a robust diagnostic biomarker panel for MDD, the metabolite biomarkers identified by NMR technology was individually added into the logistic regression model constructed with the urinary diagnostic biomarkers identified by GC−MS technology. Using SPSS 13.0 software, the resulting receiver-operating characteristic (ROC) curve and AUC values can be obtained from the disease probability values.27

Metabonomic Data Analysis

GC−MS metabolite profiles were processed after conversion into a NetCdf file format using TagFinder.22 This processing enabled deconvolution, alignment, and data reduction to produce a list of mass and retention time pairs with corresponding intensities for all detected peaks from each data file in the data set. The resulting three-dimensional data setincluding peak index (RT-m/z pair), sample names (observations), and normalized peak area percentageswas imported into SIMCA-P 12.0 (Umetrics, Umeå, Sweden). A supervised multivariate approach, termed orthogonal partial least-squares discriminant analysis (OPLS-DA), was performed on the unit variance-scaled spectral data to visualize discrimination between MDD subjects and healthy controls.23,24 By analysis of OPLS-DA loadings, the differential metabolites responsible for the discrimination between the two groups were identified (variable importance plot (VIP) > 1, p < 0.05). The levels of these metabolites were normalized to creatinine. The quality of the OPLS-DA models was described by three parameters (R2X, R2Y, and Q2Y), which were calculated by the default leave-one-out (LOO) procedure. R2X and R2Y were used to quantify the goodness-of-fit; Q2Y was employed to assess model predictability.25 To rule out nonrandomness of separation between groups, a 300-iteration permutation test

Statistical Analysis

Comparisons of demographic characteristics between groups were performed on SPSS 13.0 using the parametric Student’s ttest, the nonparametric Mann−Whitney U test, or the Chisquare test where appropriate. The concentration of differential metabolites was calculated relative to that of the creatinine. The student’s t-test was applied to detect statistical differences between depressed and control groups. A p-value of less than 0.05 was considered to be statistically significant.

■

RESULTS

Identification of Differential Urinary Metabolites in MDD

The first stage of this study sought to identify differential metabolites in first-episode, drug-naı̈ve depressed patients relative to healthy controls. A high homogeneity of MDD subjects was chosen for this analysis to ensure that the identified biomarkers were related to MDD pathophysiology rather than drugs and/or chronic impairment. OPLS-DA 5906



Article

analysis was performed to explore the metabolic differences between the two groups, and representative GC−MS total ion chromatograms of urine samples from a MDD subject and a healthy control are shown in Supplemental Figure 1, Supporting Information. The resulting OPLS-DA scores plot clearly distinguished MDD subjects from healthy controls (R2X = 0.27, R2Y = 0.788, and Q2 = 0.557; Figure 1A). Permutation tests demonstrated the OPLS-DA model was valid or not overfitted (Figure 1B). To systematically identify the metabolites responsible for differentiating between the MDD and healthy control groups, the corresponding OPLS-DA loading plots were analyzed. This analysis resulted in the identification of 23 differential urinary metabolites responsible for discriminating between MDD subjects and healthy controls (Table 1). The urine samples of MDD subjects exhibited higher levels of azelaic acid, sucrose, sorbitol, fructose, glucose, alanine, uric acid, and hydroxylamine in combination with lower levels of phenylalanine, cysteine, tyrosine, aminoisobutyric acid, Nacetyl-D-glucosamine, α-hydroxyisobutyric acid, 2,3-dihydroxybutanoic acid, 2,4-dihydroxypyrimidine, quinolinic acid, aminoethanol, hippuric acid, 1-methylinosine, indoxyl sulfate, hypoxanthine, and pseudouridine.

the simplified metabolite signature was capable of distinguishing 82 MDD subjects from 82 healthy controls with an AUC of 0.905 (95% confidence interval: 0.862−0.949) (Figure 4A). As an AUC of 1.0 represents a strongly discriminating test while an AUC of 0.5 represents a weakly discriminating test, this finding indicates that this simplified urinary metabolite signature is a “good” classifier of MDD subjects and healthy controls. Validation of the Simplified MDD Biomarker Panel

In the second stage of this study, a simplified MDD metabolite signature was identified and preliminarily validated as an effective classifier of MDD subjects and healthy controls. Given the majority of MDD patients in clinical practice are medicated, an entirely independent sample cohort including medicated MDD subjects was required to validate this metabolite signature before proceeding to a larger-scale clinical trial. Thus, as a final step, the simplified biomarker panel was used to classify blinded diverse samples from an independent test cohort of 44 MDD subjects and 52 healthy controls. ROC analysis yielded an AUC of 0.837 (95% confidence interval: 0.755−0.920; Figure 4B) in discriminating MDD subjects from healthy controls. Quantification of Diagnostic Performance of the Composite Urinary Metabolite Signature

Identification of a Simplified MDD Metabolite Signature

In the first stage of this study, 23 candidate urinary biomarkers of MDD were identified. However, diagnosis based on quantification of so many metabolites would not be convenient and economical in clinical practice. Thus, to identify a simplified urinary metabolite signature that would be more practical in diagnosing MDD, the 23 identified differential metabolites were used as candidates in the next phase of the study. A stepwise optimization algorithm based on Bayesian information criterion (BIC) was used to optimize a urinary metabolite signature for MDD. This analysis showed that six metabolitesazelaic acid, sorbitol, uric acid, quinolinic acid, hippuric acid, and tyrosinecan be attributed to the most significant deviations between the MDD and healthy control groups, indicating that these six metabolites should yield the highest predictive power for future diagnostic applications (Figure 2). The relative concentrations of these six urinary metabolite biomarkers for depression are presented in Figure 3. To quantitatively assess the diagnostic performance of this simplified metabolite signature, the area under the ROC curve (AUC) was calculated. The ROC analysis demonstrated that

The metabolite biomarkers identified by NMR technology alanine, formate, m-hydroxyphenylacetate, malonate, N-methylnicotinamidewere individually added into the logistic regression model.18 The AUCs of each logistic regression model are calculated and presented in Supplemental Table 1, Supporting Information. The majority of the newly generated panels had increased AUCs over the original panel. Of note, a composite panel constructed by the addition of previously identified urine biomarker (N-methylnicotinamide) to this biomarker panel achieved a more satisfactory accuracy, yielding an AUC of 0.909 (95% confidence interval: 0.867−0.951, Figure 5A) in the training samples and 0.917 (95% confidence interval: 0.862−0.973, Figure 5B) in the test samples.

■

DISCUSSION MDD is a prevalent and debilitating mental disorder. Currently, the lack of disease biomarkers to support objective laboratorybased testing (e.g., a urine test) constitutes a bottleneck in the clinical diagnosis of MDD. Here, a metabolite signature consisting of six urinary metabolite biomarkersazelaic acid, sorbitol, uric acid, quinolinic acid, hippuric acid and tyrosine was identified as an effective diagnostic tool, 0.905 in the training set and 0.837 in the test set. Furthermore, a composite panel constructed by the addition of previously identified urine biomarker (N-methylnicotinamide) to this biomarker panel achieved a more satisfactory accuracy, yielding an AUC of 0.909 the training samples and 0.917 in the test samples. These findings suggest that this metabolite signature should aid in the development of objective laboratory-based diagnostic tools for MDD. To explore the underlying molecular functions of these urinary metabolite biomarkers, metabolic pathway analysis was performed. The seven metabolites were found to be primarily involved in (i) glucose metabolism, (ii) the kynurenine pathway (quinolinic acid), (iii) the tyrosine-phenylalanine pathway (tyrosine and hippuric acid), (iv) tryptophan−nicotinic acid metabolism (N-methylnicotinamide), and (iv) oxidative stress (uric acid and azelaic acid). It is not surprising that these metabolite biomarkers are involved in disturbances across

Figure 2. Construction of various logistical regression models. Different combinations of urine metabolites were applied to construct various logistical regression models. Bayesian information criterion (BIC) of each model are presented. The model constructed with six select urine metabolites (sorbitol, uric acid, azelaic acid, quinolinic acid, hippuric acid, and tyrosine) showed the highest predictive ability. 5907



Article

Figure 3. The relative concentrations of these six urinary metabolite biomarkers for depression.

Figure 4. Quantification of the diagnostic performance of the six urinary metabolite biomarkers. Combined levels of the six metabolite biomarkers (i.e., sorbitol, uric acid, azelaic acid, quinolinic acid, hippuric acid, and tyrosine) yielded (A) an area under the receiver operating characteristic curve (AUC) of 0.905 in the training set and (B) an AUC of 0.837 in the test set.

Figure 5. Quantification of the diagnostic performance of the composite urinary metabolite biomarker panel. The composite urinary biomarker panel composed through the addition of N-methylnicotinamide to the original urinary metabolite panel (i.e., sorbitol, uric acid, azelaic acid, quinolinic acid, hippuric acid, tyrosine, and N-methylnicotinamide) discriminated MDD subjects from healthy controls with (A) an area under the receiver operating characteristic curve (AUC) of 0.909 in the training set and (B) an AUC of 0.917 in the test set.

5908



Article

several alternative molecular pathways. This phenomenon has also been observed in previous investigations that have successfully identified a diverse set of protein and genetic biomarkers for MDD.28,29 The disturbed metabolic pathways are discussed in detail below with the exception of “tryptophan−nicotinic acid metabolism,” which has been thoroughly discussed in a previous study.18

Table 2. Key Urinary Metabolites Attributed to Discriminating MDD Subjects and Healthy Controls

Disturbed Glucose Metabolism

Here, a significantly higher level of sorbitol was observed in MDD subjects relative to healthy controls, suggesting the glucose flux through the sorbitol pathway is upregulated in MDD subjects. Interestingly, increased TCA cycle flux has also been observed in a previous NMR study.18 Increased flux through these two metabolic pathway may aggravate the deficiency of peripheral glucose supply observed in MDD patients in a previous plasma metabolic study.17 These consistent set of findings suggests that MDD is associated with disturbances in peripheral glucose metabolism. Disturbed Kynurenine Metabolic Pathway

Quinolinic acid (QUIN), a product of the kynurenine pathway, is an excitotoxic neuromodulator. Here, urinary secretion of QUIN was significantly decreased in MDD subjects relative to healthy controls, suggesting that perturbed kynurenine pathway metabolism is implicated in MDD. In agreement with this presumption, previous studies have consistently reported that this metabolic pathway plays an important role in MDD pathogenesis.30 Given CNS kynurenine is mainly derived from the bloodstream, the increased CNS kynurenine levels previously observed in MDD patients may result in decreased circulating kynurenine levels, which may account for the decreased urinary secretion of QUIN observed here.

P-valueb

fold changec

metabolites

RT (min)_m/z

VIPa

azelaic acid sucrose sorbitol fructose glucose β-alanine uric acid hydroxylamine phenylalanine cysteine tyrosine β-aminoisobutyric acid N-acetyl-Dglucosamine α-hydroxyisobutyric acid 2,3-dihydroxybutanoic acid 2,4dihydroxypyrimidine quinolinic acid aminoethanol hippuric acid 1-methylinosine indoxyl sulfate hypoxanthine pseudouridine

14.92_317 22.76_361 16.59_319 15.83_147 16.45_319 10.67_174 18.15_441 6.38_133 13.14_218 12.32_220 16.49_218 11.1_248

2.07 1.27 1.08 1.83 1.44 1.08 2.22 1.18 1.05 1.11 2.02 1.05

2.07 1.74 6.80 4.07 1.18 1.25 8.80 1.34 5.41 3.40 3.78 6.26

10−10 10−04 10−04 10−08 10−03 10−02 10−03 10−03 10−03 10−03 10−09 10−04

−2.06 −1.46 −0.99 −0.80 −0.45 −0.53 −0.21 −0.45 0.18 0.19 0.73 1.05

18.01_205

1.13

6.43 × 10−04

0.21

5.65_131

1.44

3.93 × 10−03

0.22

9.64_117

1.57

4.68 × 10−04

0.25

9.51_241

1.56

4.07 × 10−06

0.37

14.29_296 8.53_174 15.43_206 25.03_259 14.08_277 15.1_265 20.21_357

1.88 2.04 1.58 1.80 1.55 1.98 1.53

1.04 3.45 5.14 2.01 2.01 9.78 6.59

10−10 10−10 10−04 10−09 10−06 10−11 10−13

0.42 0.43 0.52 0.44 0.57 1.25 0.58

× × × × × × × × × × × ×

× × × × × × ×

a

Variable importance in the projection (VIP) was obtained from OPLS-DA with a threshold of 1.0. bP-values were derived from the Student’s t-test. cPositive fold change values indicate relatively lower concentrations in MDD subjects relative to healthy controls. Negative fold change values indicate relatively higher concentrations in MDD subjects relative to healthy controls.

Disturbed Tyrosine-Phenylalanine Pathway

Hippuric acid is a metabolite of phenylalanine. Here, significantly decreased levels of hippuric acid were observed in MDD subjects compared to healthy controls. Interestingly, significantly decreased levels of tyrosine and phenylalanine were also found in MDD subjects (Table 2), although phenylalanine was not included in the diagnostic biomarker panel. The consistent pattern of decreased levels of metabolites in the tyrosine−phenylalanine pathway suggests that this pathway may be inhibited in MDD. Consistent with these findings, a previous urine metabolomic study in a depressive animal model has also demonstrated that depressive behavior is associated with significantly reduced levels of metabolites in the tyrosine− phenylalanine pathway.31

differentiating MDD subjects from healthy controls, further studies that recruit other psychiatric disorders are required to assess the specificity of these urinary biomarkers; (ii) except for matching common demographic factors such as age and gender in recruiting subjects, other possible confounding factors such as lifestyle, smoking, and alcohol consumption should be considered in further studies; (iii) ideally, multiple clinical sites should be used for the recruitment of patients and controls; and (iv) the current results do not address whether changes in these urinary metabolites are a cause or a consequence of MDD, which should be investigated in further studies.

Metabolic Evidence for Increased Oxidative Stress

Uric acid is the final product of purinergic catabolism and plays an important role in antioxidant defense.32 In response to increased oxidative stress, purine is preferentially metabolized to uric acid. Here, uric acid levels were significantly increased in MDD subjects relative to healthy controls, suggesting increased oxidative stress in MDD. In addition, azelaic acid levels were significantly increased in MDD subjects relative to healthy controls. Azelaic (nonanedioic) acid has been shown to inhibit reactive oxygen species (ROS) generation through NADPH oxidase inhibition.33 Through azelaic acid’s inhibition of ROS generation, this finding is consistent with the increased oxidative stress status in MDD. Our findings have to be cautiously interpreted due to the following limitations: (i) as we only assessed the diagnostic performance of the urinary metabolite biomarker panel in

■

CONCLUSIONS In conclusion, using a GC−MS based metabonomic approach, a panel of urinary metabolite biomarkers, composed of azelaic acid, sorbitol, uric acid, quinolinic acid, hippuric acid, tyrosine, and N-methylnicotinamide, was capable of discriminating MDD subjects from healthy controls with high accuracy in a training set. The diagnostic performance of this metabolite signature was further validated in a test set. These results suggest this composite urinary metabolite signature should facilitate development of a urine-based diagnostic test for MDD. 5909



■

Article

(11) He, Y.; Yu, Z.; Giegling, I.; Xie, L.; Hartmann, A.; Prehn, C.; Adamski, J.; Kahn, R.; Li, Y.; Illig, T. Schizophrenia shows a unique metabolomics signature in plasma. Transl.Psychiatry 2012, 2 (8), e149. (12) Zhang, F.; Jia, Z.; Gao, P.; Kong, H.; Li, X.; Lu, X.; Wu, Y.; Xu, G. Metabonomics study of urine and plasma in depression and excess fatigue rats by ultra fast liquid chromatography coupled with ion traptime of flight mass spectrometry. Mol Biosyst. 2010, 6 (5), 852−61. (13) Li, Z. Y.; Zheng, X. Y.; Gao, X. X.; Zhou, Y. Z.; Sun, H. F.; Zhang, L. Z.; Guo, X. Q.; Du, G. H.; Qin, X. M. Study of plasma metabolic profiling and biomarkers of chronic unpredictable mild stress rats based on gas chromatography/mass spectrometry. Rapid Commun. Mass Spectrom. 2010, 24 (24), 3539−46. (14) Shao, W.-H.; Fan, S.-H.; Lei, Y.; Yao, G.-E.; Chen, J.-J.; Zhou, J.; Xu, H.-B.; Liu, H.-P.; Wu, B.; Zheng, P. Metabolomic identification of molecular changes associated with stress resilience in the chronic mild stress rat model of depression. Metabolomics 2013, 9 (2), 433−443. (15) Zheng, S.; Zhang, S.; Yu, M.; Tang, J.; Lu, X.; Wang, F.; Yang, J.; Li, F. An 1 H NMR and UPLC−MS-based plasma metabonomic study to investigate the biochemical changes in chronic unpredictable mild stress model of depression. Metabolomics 2011, 7 (3), 413−423. (16) Zheng, S.; Yu, M.; Lu, X.; Huo, T.; Ge, L.; Yang, J.; Wu, C.; Li, F. Urinary metabonomic study on biochemical changes in chronic unpredictable mild stress model of depression. Clin. Chim. Acta 2010, 411 (3), 204−209. (17) Zheng, P.; Gao, H. C.; Li, Q.; Shao, W. H.; Zhang, M. L.; Cheng, K.; Yang, D. Y.; Fan, S. H.; Chen, L.; Fang, L. Plasma metabonomics as a novel diagnostic approach for major depressive disorder. J. Proteome Res. 2012, 11 (3), 1741−1748. (18) Zheng, P.; Wang, Y.; Chen, L.; Yang, D.; Meng, H.; Zhou, D.; Zhong, J.; Lei, Y.; Melgiri, N.; Xie, P. Identification and validation of urinary metabolite biomarkers for major depressive disorder. Mol. Cell. Proteomics 2013, 12 (1), 207−214. (19) Williams, R.; Lenz, E.; Wilson, A.; Granger, J.; Wilson, I.; Major, H.; Stumpf, C.; Plumb, R. A multi-analytical platform approach to the metabonomic analysis of plasma from normal and Zucker (fa/fa) obese rats. Mol. BioSyst. 2006, 2 (3−4), 174−183. (20) Association, A. P. Diagnostic and Statistical Manual of Mental Disorders: DSM-IV-TR; American Psychiatric Publishing: Arlington, VA, 2000. (21) Williams, J. B. A structured interview guide for the Hamilton Depression Rating Scale. Arch. Gen. Psychiatry 1988, 45 (8), 742. (22) Luedemann, A.; Strassburg, K.; Erban, A.; Kopka, J. TagFinder for the quantitative analysis of gas chromatographymass spectrometry (GC-MS)-based metabolite profiling experiments. Bioinformatics 2008, 24 (5), 732−737. (23) Trygg, J.; Wold, S. Orthogonal projections to latent structures (O-PLS). J. Chemom. 2002, 16 (3), 119−128. (24) Bylesj, M.; Rantalainen, M.; Cloarec, O.; Nicholson, J. K.; Holmes, E.; Trygg, J. OPLS discriminant analysis: combining the strengths of PLS-DA and SIMCA classification. J. Chemom. 2006, 20 (8−10), 341−351. (25) Mahadevan, S.; Shah, S. L.; Marrie, T. J.; Slupsky, C. M. Analysis of metabolomic data using support vector machines. Anal. Chem. 2008, 80 (19), 7562−7570. (26) Oikonomopoulou, K.; Li, L.; Zheng, Y.; Simon, I.; Wolfert, R.; Valik, D.; Nekulova, M.; Simickova, M.; Frgala, T.; Diamandis, E. Prediction of ovarian cancer prognosis and response to chemotherapy by a serum-based multiparametric biomarker panel. Br. J. Cancer 2008, 99 (7), 1103−1113. (27) Bradley, A. P. The use of the area under the ROC curve in the evaluation of machine learning algorithms. Pattern Recognit. 1997, 30 (7), 1145−1159. (28) Papakostas, G.; Shelton, R.; Kinrys, G.; Henry, M.; Bakow, B.; Lipkin, S.; Pi, B.; Thurmond, L.; Bilello, J. Assessment of a multi-assay, serum-based biological diagnostic test for major depressive disorder: a Pilot and Replication Study. Mol. Psychiatry 2013, 18, 332−339. (29) Spijker, S.; Van Zanten, J. S.; De Jong, S.; Penninx, B. W.; van Dyck, R.; Zitman, F. G.; Smit, J. H.; Ylstra, B.; Smit, A. B.; Hoogendijk,

ASSOCIATED CONTENT

S Supporting Information *

Figure S1, representative GC−MS total ion chromatograms of urine samples. Table S1, the AUCs of each logistic regression model. This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*Address: Department of Neurology The First Affiliated Hospital at Chongqing Medical University 1 Youyi Road, Yuzhong District, Chongqing, P.R.C. 400016. Tel: +86-2368485490. Fax: +86-23-68485111. E-mail: [email protected]. cn. Author Contributions § Peng Zheng, Jian-jun Chen, and Ting Huang contributed equally to this work.

Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS We thank Dr. N. D. Melgiri for editing and proofreading the manuscript. This work was supported by the National Basic Research Program of China (973 Program) (Grant No. 2009CB918300) and the National Natural Science Foundation of China (Grant No. 30900456).

■

REFERENCES

(1) Kupfer, D. J.; Frank, E.; Phillips, M. L. Major depressive disorder: new clinical, neurobiological, and treatment perspectives. Lancet 2012, 379 (9820), 1045−1055. (2) Mitchell, A. J.; Vaze, A.; Rao, S. Clinical diagnosis of depression in primary care: a meta-analysis. Lancet 2009, 374 (9690), 609−619. (3) Nicholson, J. K.; Lindon, J. C.; Holmes, E. “Metabonomics”: understanding the metabolic responses of living systems to pathophysiological stimuli via multivariate statistical analysis of biological NMR spectroscopic data. Xenobiotica 1999, 29 (11), 1181−1189. (4) Rhee, E. P.; Gerszten, R. E. Metabolomics and cardiovascular biomarker discovery. Clin. Chem. 2012, 58 (1), 139−147. (5) Jung, J. Y.; Lee, H. S.; Kang, D. G.; Kim, N. S.; Cha, M. H.; Bang, O. S.; Ryu do, H.; Hwang, G. S. 1H-NMR-Based Metabolomics Study of Cerebral Infarction. Stroke 2011, 42 (5), 1282−1288. (6) Yap, I. K. S.; Angley, M.; Veselkov, K. A.; Holmes, E.; Lindon, J. C.; Nicholson, J. K. Urinary metabolic phenotyping differentiates children with autism from their unaffected siblings and age-matched controls. J. Proteome Res. 2010, 9 (6), 2996−3004. (7) Yang, J.; Chen, T.; Sun, L.; Zhao, Z.; Qi, X.; Zhou, K.; Cao, Y.; Wang, X.; Qiu, Y.; Su, M. Potential metabolite markers of schizophrenia. Mol. Psychiatry 2013, 18, 67−78. (8) Bogdanov, M.; Matson, W. R.; Wang, L.; Matson, T.; SaundersPullman, R.; Bressman, S. S.; Beal, M. F. Metabolomic profiling to develop blood biomarkers for Parkinson’s disease. Brain 2008, 131 (2), 389−396. (9) Zheng, P.; Wei, Y.-D.; Yao, G.-E.; Ren, G.-P.; Guo, J.; Zhou, C.-J.; Zhong, J.-J.; Cao, D.; Zhou, L.-K.; Xie, P. Novel urinary biomarkers for diagnosing bipolar disorder. Metabolomics 2013, 800−808. (10) Lan, M. J.; McLoughlin, G. A.; Griffin, J. L.; Tsang, T. M.; Huang, J. T.; Yuan, P.; Manji, H.; Holmes, E.; Bahn, S. Metabonomic analysis identifies molecular changes associated with the pathophysiology and drug treatment of bipolar disorder. Mol. Psychiatry 2009, 14 (3), 269−79. 5910



Article

W. J. Stimulated gene expression profiles as a blood marker of major depressive disorder. Biol. Psychiatry 2010, 68 (2), 179−186. (30) Raison, C. L.; Dantzer, R.; Kelley, K. W.; Lawson, M. A.; Woolwine, B. J.; Vogt, G.; Spivey, J. R.; Saito, K.; Miller, A. H. CSF concentrations of brain tryptophan and kynurenines during immune stimulation with IFN-α: relationship to CNS immune responses and depression. Mol. Psychiatry 2009, 15 (4), 393−403. (31) Zheng, S.; Yu, M.; Lu, X.; Huo, T.; Ge, L.; Yang, J.; Wu, C.; Li, F. Urinary metabonomic study on biochemical changes in chronic unpredictable mild stress model of depression. Clin. Chim. Acta 2010, 411 (3−4), 204−9. (32) Yao, J. K.; Dougherty, G. G., Jr; Reddy, R. D.; Keshavan, M. S.; Montrose, D. M.; Matson, W. R.; McEvoy, J.; Kaddurah-Daouk, R. Homeostatic imbalance of purine catabolism in first-episode neuroleptic-naive patients with schizophrenia. PLoS One 2010, 5 (3), e9508. (33) Jones, D. A. Rosacea, Reactive Oxygen Species, and Azelaic Acid. J. Clin. Aesthetic Dermatol. 2009, 2 (1), 26.

5911


A Novel Urinary Metabolite Signature for Diagnosing Major

Recommend Documents