Article pubs.acs.org/jpr
Discovery and Validation of Plasma Biomarkers for Major Depressive Disorder Classification Based on Liquid Chromatography−Mass Spectrometry Xinyu Liu,†,∥ Peng Zheng,‡,§,∥ Xinjie Zhao,† Yuqing Zhang,‡,§ Chunxiu Hu,† Jia Li,† Jieyu Zhao,† Jingjing Zhou,‡,§ Peng Xie,*,‡,§ and Guowang Xu*,† †
Key Laboratory of Separation Science for Analytical Chemistry, Dalian Institute of Chemical Physics, Chinese Academy of Sciences, 457 Zhongshan Road, Dalian 116023, China ‡ Department of Neurology, The First Affiliated Hospital of Chongqing Medical University, 1 Youyi Road, Yuzhong District, Chongqing 400016, China § Institute of Neuroscience, Chongqing Medical University, 1 Yixueyuan Road, Yuzhong District, Chongqing, 400016, China S Supporting Information *
ABSTRACT: Major depressive disorder (MDD) is a debilitating mental disease with a pronounced impact on the quality of life of many people; however, it is still difficult to diagnose MDD accurately. In this study, a nontargeted metabolomics approach based on ultra-high-performance liquid chromatography equipped with quadrupole time-offlight mass spectrometry (UPLC-Q-TOF/MS) was used to find the differential metabolites in plasma samples from patients with MDD and healthy controls. Furthermore, a validation analysis focusing on the differential metabolites was performed in another batch of samples using a targeted approach based on the dynamic multiple reactions monitoring method. Levels of acyl carnitines, ether lipids, and tryptophan pronouncedly decreased, whereas LPCs, LPEs, and PEs markedly increased in MDD subjects as compared with the healthy controls. Disturbed pathways, mainly located in acyl carnitine metabolism, lipid metabolism, and tryptophan metabolism, were clearly brought to light in MDD subjects. The binary logistic regression result showed that carnitine C10:1, PE-O 36:5, LPE 18:1 sn-2, and tryptophan can be used as a combinational biomarker to distinguish not only moderate but also severe MDD from healthy control with good sensitivity and specificity. Our findings, on one hand, provide critical insight into the pathological mechanism of MDD and, on the other hand, supply a combinational biomarker to aid the diagnosis of MDD in clinical usage.
1. INTRODUCTION Major depressive disorder (MDD) is a serious debilitating mental disease and leads to a high rate of suicide and economic burden. It has greatly affected the quality of life of 350 million diagnosed patients worldwide.1 Biological research findings have demonstrated that MDD is closely linked to the hyperactivity of the hypothalamic−pituitary−adrenal axis,2 hippocampal atrophy,3,4 and glial reduction in the subgenual prefrontal cortex;5 however, the diagnosis of MDD is still difficult due to the high error rate.6 Several methods have been used in clinical diagnosis for depression, such as a structured or semistructured interview (SDI) and the dexamethasone suppression test (DST).7 Although these methods can be used in routine clinical practice, there is still a “gap” to achieve an accurate diagnosis. “Omics” techniques including genomics,8 proteomics,9,10 and metabolomics11 have been involved in the research of discovering biomarkers for MDD diagnosis. Metabolomics is preferred for the study of small molecular compounds, which have a close relationship with the physiological and pathological © XXXX American Chemical Society
processes in an organism due to its high-throughput quantification of metabolites. Nowadays, it has become a powerful tool in discovering biomarkers and key pathways involved in many diseases.12−14 Metabolomics technologies based on NMR,15−18 liquid chromatography−mass spectrometry (LC−MS),17,18 and gas chromatography−mass spectrometry (GC−MS)19−22 have been used to investigate metabolic changes in patients with depressive disorder. The outcome of these studies revealed the significant changes of glucose, low-density lipoprotein (LDL), very low-density lipoprotein (VLDL), lysophosphatidylcholine (LPCs), branched-chain amino acids, and organic acids in patients with the depressive disorder. Moreover, malonate, formate, N-methylnicotinamide, m-hydroxyphenylacetate, and alanine were defined as the potential urinary biomarkers for MDD.11 Currently, LC−MS has been a pillar in Received: February 14, 2015
A
DOI: 10.1021/acs.jproteome.5b00144 J. Proteome Res. XXXX, XXX, XXX−XXX
Article
Journal of Proteome Research
belonged to M-MDD and 19 subjects belonged to S-MDD. The antidepressive drugs taken by MDD subjects included fluoxetine, escitalopram, mirtazapine, venlafaxine, deanxit, and citalopram. The clinical information is presented in Table 1.
metabolomics due to its advantages of high sensitivity, high resolution, and wide coverage of metabolites.23−25 In this study, a nontargeted metabolomics approach based on UPLC-Q-TOF/MS technology was first applied to discover the candidate biomarker for MDD identification and disclose the key pathways in MDD; then, a targeted metabolomics approach focusing on the candidate metabolites was executed for further validation (Figure 1). Our aim is to identify the disturbed metabolites involved in MDD and provide potential biomarkers for MDD classification.
2.2. Sample Preparation for the Discovery and the Validation Sets
To remove the protein, we added 400 μL of acetonitrile containing 10 internal standards to 100 μL of plasma. After 2 min of vortexing, the mixture was centrifuged at 13 000 rpm for 10 min at 4 °C. The supernatant was divided into two aliquots (200 μL for each) and then frozen to dry for subsequent analysis. The residues were dissolved in 50 μL of 80% methanol. The information on the 10 internal standards included in the extraction reagent were as follows: tryptophan-d5 (4.25 μg/ mL), carnitine C2:0-d3 (0.16 μg/mL), FFA16:0-d3 (2.5 μg/mL), FFA 18:0-d3 (2.5 μg/mL), chenodeoxycholic acid-d4 (1.49 μg/ mL), cholic acid-d4 (1.8 μg/mL), LPC 19:0 (0.75 μg/mL), SM (d18:1/12:0) (0.13 μg/mL), PE 34:0 (0.32 μg/mL), and PC 38:0 (0.63 μg/mL). Notably, quality-control (QC) samples were prepared by mixing equal aliquots of plasma from each real sample and were pretreated in the same manner as the real samples. One QC was inserted regularly after running every 10 real samples. 2.3. Metabolic Profiling Data Acquisition
Metabolic profiling of plasma samples in a discovery set based on nontargeted analysis was collected by a Waters ACQUITY ultraperformance liquid chromatography system (UPLC) (Waters, Milford, MA) coupled to an AB SCIEX Triple TOF 5600 System (AB SCIEX, Framingham, MA). An ACQUITYTM C8 BEH column (1.7 μm, 2.1 × 100 mm) and a T3 HSS column (1.8 μm, 2.1 × 100 mm) were employed in positive and negative modes, respectively, and the column temperature was kept at 55 °C. The mobile phases of (A) 0.1% formic acid in water and (B) 0.1% formic acid in acetonitrile were used in positive ionization mode, while (C) 6.5 mM NH4HCO3 in water and (D) 6.5 mM NH4HCO3 in 95% methanol/water were used in negative ionization mode. In positive mode, the gradient elution initially started from 5% B, after 0.5 min linearly increased to 100% B in 23.5 min and maintained for 4 min, and then returned to the initial ratio for 2 min of equilibrium. In negative mode, the gradient initially started from 5% B, after 0.5 min linearly increased to 100% B in 17.5 min and maintained for 5.5 min, and then returned to the initial ratio for about 2 min of equilibrium. Data acquisition was performed in full scan mode with m/z 100− 1000 in positive mode and m/z 80−1000 in negative mode. Parameters of mass spectrometry were as follows: ion spray voltage, 5500 V (+) and 4500 V (−); curtain gas of 35 PSI, declustering potential, 100 V (+) and −100 V (−); collision energy, 10 V (+) and −10 V (−); and interface heater temperature of 550 °C. Dynamic multiple reactions monitoring (MRM) data were acquired for validation set by Shimadzu LC (30AD)−MS (TQ 8050) (Shimadzu, Kyoto, Japan). The columns were the same as the nontargeted analysis. Shorter gradients were applied in both positive mode and negative mode to increase throughput. The same mobile phases were used in positive mode, while the lower concentration of NH4HCO3 (5 mM) was used in negative mode for matching the shorter gradient. For positive mode, the gradient initially started from 10% B, after 1 min linearly increased to 40% B in 4 min and then increased to 100% B in 12 min and maintained for 5 min, and then returned to the initial ratio for 3 min of equilibrium. In negative mode, the gradient
Figure 1. Flowchart of the study strategy in this work.
2. MATERIALS AND METHODS 2.1. Samples Collection and Clinical Information
The plasma samples of MDD subjects and healthy controls, including discovery set and validation set, were sampled, respectively, in the psychiatric center and medical examination center of the First Affiliated Hospital of Chongqing Medical University (Chongqing, China). The study was approved by the ethics committee of Chongqing Medical University. Written informed consent was obtained from each participant. All of the MDD patients were diagnosed according to the fourth Diagnostic and Statistical Manual of Mental Disorders criteria for MDD (DSM-IV).26 The Hamilton depression scale (HAMD) value of the subjects in the MDD group was higher than 18. ̈ MDD (DNIn the discovery set, 60 first-episode drug-naive MDD) patients (male/female: 30/30) and 59 (male/female: 30/ 29) healthy control (HC) subjects were included. In the MDD group, the HAMD value of MDD subjects ranged from 18 to 37, 33 cases of subjects with HAMD value ranged from 18 to 24 were divided in moderate MDD group (M-MDD), and the other 27 cases of subjects with HAMD value greater than 24 were assigned in the severe MDD group (S-MDD).27 In the validation set, the other 75 MDD patients and 52 HC subjects were assigned. The HAMD value of MDD subjects ranged from 18 to 58, 45 cases belonged to M-MDD, and 30 cases were classified in S-MDD. Moreover, among the 75 MDD patients, 35 subjects (male/female: 20/15) were first-episode DN-MDD, 24 of them were in M-MDD group, and 11 of them were in S-MDD group. The other 40 cases were drug-treatment (DT) MDD subjects (male/female: 20/20), and 21 subjects B
DOI: 10.1021/acs.jproteome.5b00144 J. Proteome Res. XXXX, XXX, XXX−XXX
Article
HAMD, Hamilton depression scale; BMI, body mass index; CHOL, cholesterol; TG, triglyceride; LDLC2, low density lipoprotein cholesterol; APOBT, apolipoprotein B; APOAT, apolipoprotein A; HDLC3, high density lipoprotein cholesterol. b*, p < 0.05; **, p < 0.01; ***, p < 0.001 compared with control. ###, p < 0.001, S-MDD compared with M-MDD.
21.58 ± 2.70 4.51 ± 0.62 1.01 ± 0.38 2.41 ± 0.58 0.76 ± 0.19 1.50 ± 0.24 1.52 ± 0.37 22.25 ± 2.11 4.69 ± 0.75 1.22 ± 0.69 2.61 ± 0.72 0.84 ± 0.21 1.45 ± 0.25 1.42 ± 0.41
initially started from 100% A, after 1 min linearly increased to 40% B in 2 min and then increased to 100% B in 9 min and maintained for 4 min, and then returned to the initial ratio for about 4 min of equilibrium. The main parameters of mass spectrometry were the same as those in positive and negative modes: heating gas flow, 10 L/min; dry gas flow, 10 L/min; nebulizing gas flow, 3 L/min; DL temperature, 250 °C; heat block temperature, 400 °C; and interface heater temperature; 300 °C. 2.4. Data Analysis
the raw data from the nontargeted analysis acquired by UPLC-QTOF/MS were imported to MarkerView workstation (AB SCIEX, USA) to obtain the matched and aligned peak table. Isotope-labeled internal standards were used to correct the areas of ion masses in the peak table. First, all ion mass features in QC samples were corrected by each internal standard; then, a suitable internal standard was defined according to the minimum RSD % of each ion mass feature. After ion mass features were corrected by the selected internal standard, multivariate and univariate analyses were conducted. Lab Solutions Main (Shimadzu, Kyoto, Japan) was used to extract the area of metabolites in the targeted analysis; then, the peak area of each metabolite was corrected by the same internal standard as the nontargeted analysis. SIMCA-P software (version 11.0; Umetrics) was applied to fulfill the orthogonal partial least-squares (OPLS) analysis of the data from both positive and negative modes, and unit variance (UV) scaling was utilized before multivariate analysis. Mann− Whitney U test with FDR limit equal to 0.05 was used, and the metabolites with adjusted p value
