Q-TOFMS-based plasma metabolomics of polycystic ovary syndrome patients with and without insulin resistance

Journal of Pharmaceutical and Biomedical Analysis 121 (2016) 141–150

Contents lists available at ScienceDirect

Journal of Pharmaceutical and Biomedical Analysis journal homepage: www.elsevier.com/locate/jpba

UHPLC/Q-TOFMS-based plasma metabolomics of polycystic ovary syndrome patients with and without insulin resistance Ya-Xiao Chen a,1 , Xiao-Jing Zhang b,1 , Jia Huang a , Shu-Jun Zhou b , Fang Liu b , Lin-Lin Jiang a , Meiwan Chen b , Jian-Bo Wan b,∗ , Dong-Zi Yang a,∗∗ a b

Department of Obstetrics and Gynecology, Sun Yat-Sen Memorial Hospital, Guangzhou, China State Key Laboratory of Quality Research in Chinese Medicine, Institute of Chinese Medical Sciences, University of Macau, Macao, China

a r t i c l e

i n f o

Article history: Received 6 July 2015 Received in revised form 29 December 2015 Accepted 12 January 2016 Available online 15 January 2016 Keywords: Polycystic ovary syndrome Insulin resistance Metabolomics UHPLC/Q-TOFMS

a b s t r a c t Polycystic ovary syndrome (PCOS), characterized with menstrual irregularities, hyperandrogenism and ovulatory abnormalities, is usually companied with insulin resistance (IR) and accounts for one of the most prevalent reproductive dysfunction of premenopausal women. Despite accumulating investigations, diagnostic standards of this pathological condition remain obscure. The aim of present study is to characterize the plasma metabolic characteristics of PCOS patients with and without IR, and subsequently identify the potential biomarkers for the diagnosis of PCOS and its IR complication. A total of 59 plasma samples from eligible healthy controls (CON, n = 19), PCOS patients without IR (non-IR PCOS, n = 19) and PCOS patients with IR (IR PCOS, n = 21) were profiled by an ultra high-performance liquid chromatography–quadrupole time-of-flight mass spectrometry (UHPLC/Q-TOFMS) followed by multivariate statistical analysis. Compared to the healthy controls, significant decrease in the levels of phosphocholines (PCs) and lyso PC (18:2), and increase in trilauric glyceride level were observed in the plasma of IR PCOS. Meanwhile, the significant increase in the levels of saturated fatty acids (palmitic acid and stearic acid) and decanoylcarnitine, and decrease in PC (36:2) and PS (36:0) were found in non-IR PCOS patients. Trilauric glyceride and decanoylcarnitine were identified as the potential biomarkers with the highest sensitivity and specificity for the diagnosis of PCOS patients with and without IR, respectively. Furthermore, based on these alterations of metabolites, MetPA network pathway analysis suggested a profound involvement of the abnormalities of glycerophospholipid, glycerolipid and fatty acid metabolisms in the pathogenesis of PCOS and IR complications. Collectively, LC–MS-based metabolomics provides a promising strategy for complementary diagnosis of PCOS and its IR complication and offers a new insight to understand their pathogenesis mechanisms. © 2016 Elsevier B.V. All rights reserved.

1. Introduction Polycystic ovary syndrome (PCOS) is a common endocrine and metabolic disease that affects approximately 5–10% of the reproductive-age women, leading to the increased risk of anovulatory infertility and recurrent pregnancy loss [1,2]. Insulin resistance (IR) is the most common complication of PCOS, occurring in at least 50% of PCOS women, independent of obesity and age [3]. Insulin resistance has been recognized to be the major cause of

∗ Corresponding author at: Institute of Chinese Medical Sciences, University of Macau, Taipa, Macao, China. Fax: +853 2884 1358. ∗∗ Corresponding author. Fax: +86 20 8133 2853. E-mail addresses: [email protected], [email protected] (W. Jian-Bo), [email protected] (D.-Z. Yang). 1 The authors contributed equally to this work. http://dx.doi.org/10.1016/j.jpba.2016.01.025 0731-7085/© 2016 Elsevier B.V. All rights reserved.

hyperandrogenenia and chronic anovulation and plays a critical role in pathogenesis and development of PCOS [4]. PCOS patients with IR (IR PCOS) are more susceptible to develop infertility [5], type II diabetes mellitus [6] and cardiovascular diseases [7] than those without IR (non-IR PCOS). The polygenic trait and complex environmental factors of PCOS cause its diverse phenotypes and the heterogeneous diagnostic criteria [8,9]. According to the crucial clinical characteristics of PCOS, including hyperandrogenism, chronic anovulation and polycystic ovary (PCO), three international diagnostic criteria for PCOS have been proposed at 1999 [10], 2003 [5,11] and 2006 [12], respectively. However, the definition and diagnosis of PCOS are still controversial [13,14]. Moreover, patients with PCOS can be assigned into two phenotypes based on the presence or absence of insulin resistance that diagnosed by oral glucose tolerance test (OGTT), homeostasis model assessment (HOMA) or insulin–glucose clamp studies [15]. These diagnostic approaches

142

Y.-X. Chen et al. / Journal of Pharmaceutical and Biomedical Analysis 121 (2016) 141–150

Fig. 1. PLS-DA score plot (A) and permutation test (B) based on UHPLC/Q-TOFMS analysis of plasma samples from IR PCOS (n = 21), non-IR PCOS (n = 19) and healthy controls (n = 19).

for IR are invasive, time consuming and uncomfortable for patients. Therefore, the effective molecular biomarker(s) in plasma are critically important for clinical diagnosis and therapy of PCOS and its IR complication. Metabolomics, the youngest member in the “omics” family, is an emerging platform that allows to understanding global endogenous metabolites in the biological systems and its dynamic changes in response to endogenous and exogenous factors [16,17]. It has been extensively applied in the study of disease pathogenesis and discovery of diagnostic biomarkers [18,19]. In the recent years, the metabolomics studies have also been conducted to identify possible PCOS biomarkers and to explain the latent mechanisms responsible for the pathogenesis and progression of PCOS by using nuclear magnetic resonance (NMR) [20,21], gas chromatography–mass spectrometry (GC–MS) [22,23] and liquid chromatography–mass spectrometer (LC–MS) techniques [24]. These studies mainly focused on metabolic characteristics of different PCOS phenotypes and obesity. Recently, the serum metabolic profiles of PCOS with and without IR were also conducted using

ultra high-performance liquid chromatography–quadrupole timeof-flight mass spectrometry (UHPLC/Q-TOFMS) [25]. Although both serum and plasma are derived from circulating blood and metabolites in these two biological fluids are similar, they are not entirely and accurately equivalent. Significant differences were found in both types and levels of metabolites in serum and plasma, such as organic acids, amino acids, amines, carbohydrates, fatty acids, and lipids [26–28]. The aim of present study is to characterize plasma metabolic profiles of PCOS patients with and without IR complication using UHPLC/Q-TOFMS-based metabolomics approach, and subsequently to identify the potential biomarkers for the complementary prognosis of PCOS and its IR complication. 2. Materials and methods 2.1. Subjects All of the PCOS patients and health subjects were recruited from Department of Obstetrics and Gynecology, Sun Yat-Sen Memorial

Y.-X. Chen et al. / Journal of Pharmaceutical and Biomedical Analysis 121 (2016) 141–150

143

Fig. 2. OPLS-DA score plots of pairwise groups, (A) healthy controls vs IR PCOS, (B) healthy controls vs non-IR PCOS, (C) IR PCOS vs non-IR PCOS, (D) healthy controls vs total PCOS (including non-IR PCOS and IR PCOS).

Hospital, Guangzhou, China, from January 2013 to October 2013. The informed consent letters have been obtained from all the participants prior to inclusion in this study. This study was approved by the Medical Ethical Committee of Sun Yat-Sen Memorial Hospital. According to the Rotterdam criteria [5,11], PCOS diagnosis was made in patients that present at least two out of the three following clinical traits: (a) oligo- and/or anovulation; (b) clinical and/or biochemical hyperandrogenism, e.g., acne, hirsutism and androgenic alopecia; and (c) PCO (presence of more than 12 follicles in each ovary with the diameter of 2–9 mm, and/or increase in ovary size more than 10 cm3 ). Patients with IR were diagnosed by oral glucose tolerance test (OGTT) and met the following criterias, (a) plasma fasting glucose concentration <6.1 mmol/L; (b) 2 h after glucose load between 7.8 and 11.1 mmol/L in OGTT assay; (c) fasting insulin level >12 mU/L; and (d) 2-h insulin level in OGTT assay >80 mU/L. The clinical subjects were excluded with any of the following condition, including (a) age <20 or >40 years; (b) current pregnant, delivery or miscarriage within the preceding 3 months; (c) congenital adrenal hyperplasia, androgen-secreting tumors, and other diseases with hyperandrogenism; (d) cardiovascular diseases; and (e) androgenic drug or sex steroid therapy. The control subjects were recruited from health women who visited the hospital for routine checkup with matching ages, regular menstrual cycles, normal androgen levels, no PCO and no IR. According to the above mentioned inclusion/exclusion criteria, a total of 40 PCOS patients,

i.e., 21 patients with IR (IR PCOS) and 19 patients without IR (nonIR PCOS), and 19 health participants were included in the present study. 2.2. Sample collection and clinical biochemical assay The overnight fasting blood samples (3 mL) were collected from PCOS and control subjects into anticoagulant tubes. Alcohol, greasy food and excess exercise were avoided at least 24 h prior to sample collection. Plasma were obtained through centrifuging at 3000 × g for 10 min at 4 ◦ C and stored at −80 ◦ C until biochemical assay or UHPLC/Q-TOFMS analysis. All of the clinical and biochemical characteristics of indicated groups were examined at the Core Laboratory at Sun Yat-Sen Memorial Hospital, Guangzhou, China. 2.3. Sample preparation For UHPLC/Q-TOFMS analysis, an aliquot of 100 ␮L of plasma sample was mixed with 400 ␮L of acetonitrile followed by centrifugation at 11,200 × g for 10 min at 4 ◦ C for deproteinization. The supernatant (400 ␮L) was separated and then evaporated to dryness under N2 gas. The residue was reconstituted in 100 ␮L of 80% acetonitrile aqueous solution before UHPLC/Q-TOFMS analysis. A pooled “quality control” (QC) sample was prepared by mixing equal volumes (10 ␮L) from each of the 59 plasma samples, and an

144

Y.-X. Chen et al. / Journal of Pharmaceutical and Biomedical Analysis 121 (2016) 141–150

Fig. 3. Heat-map (A) and intensity comparison (B) of differential metabolites visualized the variations of metabolite profiles of IR PCOS, non-IR PCOS and healthy controls. *p < 0.05, **p < 0.01.

aliquot of 100 ␮L QC sample was prepared in parallel with that of test samples for the method validation. 2.4. UHPLC/Q-TOFMS analysis The samples were analyzed by Waters ACQUITYTM UHPLC system coupled with Xevo G2S Q-TOFMS system (Waters Corp., Milford, MA, USA) which equipped with electrospray ionization (ESI) source. The separation of metabolites was achieved on an ACQUITY UHPLC HSS T3 column (100 mm × 2.1 mm i.d., 1.7 ␮m) using a binary gradient elution system consisted of acetonitrile containing 0.1% formic acid (A) and 0.1% aqueous formic acid solution containing 1% acetonitrile (B) under the following gradient program: linear gradient from 1% A to 99% A (0–6 min) followed by linear gradient back to 1% A in an additional 2 min (6–8 min). 10 ␮L aliquot of each sample was subjected to the column that was maintained at 45 ◦ C. The flow rate was 0.45 mL/min. The LC eluent

was introduced into the MS system and individually detected in ESI positive (ESI+) and ESI negative (ESI−) ion models. The optimized condition was as follows: source temperature at 120 ◦ C, desolvation gas flow of 800 L/h at 450 ◦ C, cone gas flow at 20 L/h, capillary voltage of 3.0 kV (ESI+) or 2.0 kV (ESI−), sampling cone voltage of 30 V (ESI+) or 45 V (ESI−), extraction cone voltage of 4.0 V. MS data were acquired in full scan mode from m/z 50 to m/z 1200 at acquisition rate of 0.2 s/scan, the calibrant solution, leucine-enkephalin (200 ng/mL), was continuously introduced to MS system at a flow-rate of 50 ␮L/min via the LockSprayTM interface to ensure the accuracy and reproducibility of TOFMS. 2.5. Data processing The acquired raw UHPLC/Q-TOFMS was originally processed using Progenesis QI (Waters Corp., Milford, MA, USA), deriving normalized label-free quantitation results of metabolites peaks of each

Y.-X. Chen et al. / Journal of Pharmaceutical and Biomedical Analysis 121 (2016) 141–150

145

Fig. 4. Receiver operating curves (ROC) of potential biomarkers contributing to the separations of (A) IR PCOS and healthy controls, (B) non-IR PCOS and healthy controls.

sample. The normalization across samples was based on total ion intensity of each chromatogram. The resulting data matrix consisting of sample code, RT (retention time)—m/z pair, and the peak area was introduced into SIMCA-P software (version 13.0, Umetrics, Umeå, Sweden) for multivariate pattern recognition analysis. After dataset pre-treatment using mean-centered and Pareto (Par)scaled mathematical methods, principal component analysis (PCA), partial least-squares-discriminant analysis (PLS-DA) and orthogonal partial least squares-discriminant analysis (OPLS-DA) were performed. All models were further evaluated using variance analysis of sevenfold cross-validation predictive residual to avoid model overfitting. The permutation testing was also used to evaluate the constructed PLS-DA model. The parameters of R2 (cum) and Q2 (cum) were used to evaluate fitness and predictive ability of the constructed model, respectively. 2.6. Statistics The clinical characteristics of all the subjects and the relative contents of metabolites in plasma samples were presented as mean ± standard deviation (SD). A one-way analysis of variance (ANOVA) followed by Tukey’s post hoc test or student t-test was performed to examine the statistic difference among three groups or between two groups, respectively, using SPSS version 19.0 software (SPSS, Inc., Chicago, IL, USA). A value of p < 0.05 was considered significant. 3. Results and discussion 3.1. Baseline characteristics of PCOS patients and healthy controls The clinical and biochemical characteristics of PCOS patients with and without IR as well as healthy controls were shown in Table 1. Age, systolic blood pressure (BP), diastolic BP, triglyceride (TG), cholesterol (CHOL), low density lipoprotein (LDL), high density lipoprotein (HDL), folliclestimulating hormone (FSH)/luteinizing hormone (LH), and fasting glucose were not significantly different in three groups. There were no significant differences in body mass index (BMI), WHR, fasting insulin, OGTT 2-h insulin, and OGTT 2-h glucose between controls and non-IR PCOS, but these characteristics were significantly increased in the IR PCOS patients than the other groups. Compared to controls, PCOS

patients (IR and non-IR patients) showed the elevated testosterone (T), luteinizing hor-mone (LH), free testosterone (FT), androstendione (A2) and ovarian follicles, while those indicators have no statistically differences between non-IR PCOS and IR PCOS groups. 3.2. UHPLC–Q-TOF/MS analysis of plasma samples In order to guarantee the accuracy and credibility of raw data, LC–MS method validation was performed prior to the measurement of plasma samples. 10 ␮L of each plasma sample was mixed to generate a pooled QC sample that provided a representative “mean” sample containing all analytes. Pooled QC sample was analyzed three times at the beginning of the analytical run to ensure system equilibration and then once or twice every ten samples to provide robust quality assurance for each metabolic feature detected. Ten representative peaks in the chromatograms of QC sample detected in both positive and negative ion mode were selected for method validation. Overall, the retention times and molecular weight of the ten selected peaks in all the analyzed UHPLC chromatographs were precisely the same, while the relative standard deviations (RSD, %) of peak area were measured as less than 11.6% for ESI+ and 15.2% for ESI− mode, respectively. This result suggests that the developed LC–MS method was robust with good precision, stability and repeatability for a metabolomics study. Under the optimal conditions, the metabolomes of plasma samples from IR PCOS, non-IR PCOS, and healthy controls were profiled by the validated UHPLC/Q-TOFMS method in positive and negative ion modes. Based on their chromatograms, a three-dimensional dataset encompassing the retention time, m/z value, and normalized peak area of metabolites was generated. A total of 3275 peak indices (RT-m/z pairs) in positive mode and 546 peak indices in negative mode were observed across samples. Thus, the dataset derived from positive ion mode was selected for further multivariate statistical analysis because it appeared more sensitive for LC–MS analysis of plasma metabolites. 3.3. Pattern recognition of healthy control, non-IR PCOS and IR PCOS The resulting dataset was analyzed to examine the clustering of each group using multivariate statistical analysis. Unsupervised PCA provided unsatisfactory separation of three groups (figure not shown). In order to maximize the separation among groups of

146

Y.-X. Chen et al. / Journal of Pharmaceutical and Biomedical Analysis 121 (2016) 141–150

Fig. 5. Summary of ingenuity pathway analysis with MetPA (A), including (B) glycerophospholipid metabolism, (C) glycerolipid metabolism and (D) fatty acid metabolism.

observations, the supervised PLS-DA, a regression extension of PCA that utilized class information, was also carried out to examine the separation of three groups. After mean-centering and Pareto scaling, the score plot and permutation test of PLS-DA model were generated according to the metabolite profiles detected in positive ion mode (Fig. 1). All observations were located in the Hotelling T2 (0.95) ellipse, visual inspection of PLS-DA score plot exhibited tight clusters of samples from each group and clearly discrimination among IR PCOS, non-IR PCOS and healthy controls. Moreover, compared to IR PCOS, the cluster of non-IR PCOS tended to be closer to healthy controls (Fig. 1), which indicated that the disturbance of metabolite profile by IR is greater than that by PCOS condition. Several parameters, including R2 and Q2 , are commonly used to evaluate the quality and reliability of model. Herein, R2 X, R2 Y and Q2 of the constructed PLS-DA model were 0.749, 0.903 and 0.514, respectively, suggesting good fitness and prediction.

To study the respective impact of PCOS and insulin resistance on metabolic changes in details, the metabolic differences between pairwise groups were characterized using OPLS-DA models (Fig. 2). The samples from controls and IR PCOS patients (R2 Y = 0.998, Q2 = 0.411, Fig. 2A), controls and non-IR PCOS patients (R2 Y = 0.982, Q2 = 0.810, Fig. 2B), and non-IR PCOS and IR PCOS patients (R2 Y = 0.921, Q2 = 0.765, Fig. 2C) were unambiguously distinguished according to their differences in globally metabolic profiles. Additionally, in order to determine whether metabolites profiles sufficiently distinguished between controls and total PCOS patients, irrespective of IR complication, the data of IR PCOS and non-IR PCOS groups were combined and defined as one class, and then subjected to same OPLS-DA model along with those of healthy controls. Very clear differentiation between total PCOS patients and healthy controls was observed in scores plots, with values of R2 Y and Q2 of 0.905 and 0.413, respectively (Fig. 2D). These data

Y.-X. Chen et al. / Journal of Pharmaceutical and Biomedical Analysis 121 (2016) 141–150

147

Table 1 Clinical and biochemical characteristics of healthy controls, non-IR PCOS and IR PCOS groups. Control (n = 19) Age (y) BMI (kg/m2 ) WHR Systolic BP (mmHg) Diastolic BP (mmHg) T (mmol/L) FT (mmol/L) A2 (mmol/L) LH (U/L) LH/FSH Fasting insulin (mU/L) Fasting glucose (mmol/L) 2-h insulin OGTT (mU/L) 2-h glucose OGTT (mmol/L) CHOL(mmol/L) TG (mmol/L) HDL (mmol/L) LDL (mmol/L) Ovarian follicles (n)

27.7 19.7 0.77 121 77.2 1.77 4.13 3.15 8.63 1.23 9.55 5.22 61.2 6.93 4.12 0.74 1.77 2.55 6.4

± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

4.3 4.1a 0.44a 21 17.8 0.33a 1.55a 2.09a 2.13a 0.56 1.89a 1.32 13.5a 0.85a 1.76 0.48 0.53 1.24 3.6a

Non-IR PCOS (n = 19) 30.3 19.5 0.77 118 69.7 3.15 5.96 4.96 12.5 1.38 9.25 5.13 56.7 7.15 4.56 0.76 1.72 2.93 12.5

± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

4.1 3.4a 0.41a 12 9.5 0.41b 1.44b 2.36b 2.8b 0.72 2.34a 0.96 12.1a 0.96a 1.78 0.34 0.48 1.18 2.4b

IR-PCOS (n = 21) 28.5 22.7 0.85 123 75.3 3.76 6.71 4.77 11.3 1.58 15.4 5.24 127 8.23 5.47 1.49 1.27 3.15 14.7

± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

3.5 5.3b 0.39b 13 10.5 0.57b 2.24b 2.12b 2.3b 0.44 3.76b 1.11 22b 1.08b 2.12 0.73 0.45 1.23 3.5b

BMI, body mass index; WHR, waist-to-hip ratio; BP, blood pressure; T, testosterone; FT, free testosterone; A2, androstenedione; LH, luteinizing hormone; FSH, follicle stimulating hormone; OGTT, oral glucose tolerance test; TG, triglyceride; CHOL, cholesterol; HDL, high density lipoprotein; LDL, low density lipoprotein. Data represents mean ± SD. Averages followed by different letters are significantly different at p < 0.05.

indicated that there are substantial differences in the composition and/or content of metabolites between the compared groups. 3.4. Metabolic alterations correlated with PCOS condition and IR complication In order to reveal endogenous metabolites contributing most to pairwise group separations, the S-plot was constructed following the OPLS-DA model, in which the X-axis and Y-axis represent variable contribution and variable confidence, respectively. The further the ion departs from zero of X-axis and Y-axis, the more the ion contributes to the difference with higher confidence level (figure not shown). The variable importance in the projection (VIP) values was also implemented to search potential biomarkers. Only variables with VIP values higher than 2.0 were highlighted to be important for discrimination, and the variables were further filtered by Student t-test to determine whether potential biomarker were statistically significant between two groups. The metabolites with a significant difference (p < 0.05) were kept and considered as the potential biomarkers. TOFMS and MS/MS data provide accurate molecular mass of the parent ion and highlight fragment ion, which allows to deduce their exact molecular formulas and structural characteristics. By combination of searching biochemical databases, e.g., PubChem (http:// ncbi.nim.nih.gov/), ChemSpider database (www.chemspider.com), and MassBank (http://www.massbank.jp/), the endogenous mammalian metabolites with matched molecular formula and structure information were rigorously assigned as potential biomarkers. The results were summarized in Table 2, and the corresponding contents in non-IR PCOS, IR PCOS and control groups were shown in Fig. 3. From OPLS-DA/S-plot between IR PCOS and healthy controls, 11 ions from plasma samples in the positive ion mode were deemed discriminatory (VIP >2.0 and p < 0.05), and 7 potential biomarkers were tentatively identified as lysophosphatidylcholine (LysoPC) (18:2), phosphatidylcholine (PC) (34:2), PC (36:2), PC (36:4), PC (38:3), PC (38:4), and trilauric glyceride. Intriguingly, except for the increased trilauric glyceride level in the plasma of IR PCOS patients, plasma contents of the 5 PCs and Lyso PC (18:2) were significantly decreased when compared to those of healthy controls (Table 2). The metabolic variation between IR PCOS and controls resulted from the integrative effects of PCOS condition and IR condition,

as evidenced by the targeted metabolites identified in the comparisons of non-IR PCOS vs control group, and these two subgroups of PCOS patients. For instance, the content of plasma PC (36:2) in non-IR PCOS patients was lower than that of healthy controls, and higher than that of IR PCOS, thus, leading to the dramatically decreased concentration of PC (36:2) in IR PCOS plasma when compared to control group. Although similar variation tendency of several metabolites was found in IR PCOS and non-IR PCOS patients, diverse metabolite variations have also been induced by PCOS and IR, and thus the individual PCOS subgroups showed different representative metabolite profiles. Interestingly, we observed that the plasma content of palmitic acid was significantly higher in non-IR PCOS when compared to control group, whereas the IR complication decreased its content, which might explained the comparable levels of palmitic acid in IR PCOS and control groups. Our findings suggested that IR condition was major factor contributed to the separation of IR PCOS and control groups. The lower levels of lysoPC and PCs, except PC (40:5), were detected in plasma samples from IR PCOS patients, compared to non-IR PCOS (Table 2), which was consistent with the previous studies [23,24]. An accumulating number of studies have documented that PC, the major phospholipid component of eukaryotic membranes, could improve insulin sensitivity and contribute to both proliferative growth and programmed cell death [29]. PC is also the biosynthetic precursor of lysoPC [30]. A numerous of studies have shown that lyso PC play an critical role in glucose metabolism, lyso PC activates adipocyte glucose uptake and lowers blood glucose levels in murine models of diabetes [31], and the decreased plasma level of lyso PCs was found in Type 2 diabetes [32]. Low lyso PC was also considered as marker molecules of insulin resistance [32]. When combined the two subgroups of PCOS patients and compared their metabolite profiles with those of control subjects, a total of 11 endogenous small molecules have been highlighted for their contributions to the discrimination of healthy controls and total PCOS, irrespective of IR complication. Among them, nine metabolites were identified and were also shown in Table 2. Most of these metabolites were also same as the biomarkers identified from IR PCOS vs control and IR PCOS vs non-IR PCOS, but none from non-IR PCOS vs control, which demonstrated that IR was associated with large metabolic changes in total PCOS patients. Our findings also

148

Y.-X. Chen et al. / Journal of Pharmaceutical and Biomedical Analysis 121 (2016) 141–150

Table 2 Summary of differential biomarkers of pairwise groups in ESI positive mode. tR (min) Mean measured mass (Da) Theoretical exact mass (Da) Error (ppm) Molecular formula p-value Trend Identity IR PCOS vs control

4.60 5.11 5.11 5.11 5.18 5.22 5.26

520.3404 758.5700 782.5703 786.5992 702.5600 834.6000 810.6011

520.3403 758.5700 782.5700 786.6013 702.5649 834.5989 810.6013

0.2 0.0 0.4 −2.7 −6.9 1.3 −0.2

C26 H50 NO7 P C42 H80 NO8 P C44 H80 NO8 P C44 H84 NO8 P C39 H74 O6 C46 H86 NO8 P C46 H84 NO8 P

0.020 <10e-4 0.006 <10e-4 <10e-4 <10e-4 <10e-4

↓ ↓ ↓ ↓ ↑ ↓ ↓

LysoPC(18:2) PC(34:2) PC(36:4) PC(36:2) Trilauric glyceride PC(38:3) PC(38:4)

Non-IR PCOS vs control

3.49 3.70 4.09 4.96 5.11 5.71

316.2479 274.2741 302.3051 820.6019 786.5992 256.2629

316.2488 274.2746 302.3059 820.6068 786.6013 256.2640

−2.9 −2.0 −2.7 −6.0 −2.7 −4.6

C17 H33 NO4 C16 H32 O2 C18 H36 O2 C44 H86 NO10 P C44 H84 NO8 P C16 H33 NO

0.012 0.021 0.024 0.003 0.003 0.026

↑ ↑ ↑ ↓ ↓ ↓

Decanoylcarnitine Palmitic acid Stearic acid PS(36:0) PC(36:2) Palmitic amide

IR PCOS vs non-IR PCOS 3.70 4.60 5.11 5.11 5.11 5.11 5.22 5.26 5.49 5.51

274.2741 520.3404 758.5700 782.5703 786.5992 794.6033 834.6000 810.6011 868.6461 927.1894

274.2746 520.3403 758.5700 782.5700 786.6013 794.6064 834.5989 810.6013 868.6431 927.1890

−2.0 0.2 0.0 0.4 −2.7 −3.8 1.3 −0.2 3.4 0.4

C16 H32 O2 C26 H50 NO7 P C42 H80 NO8 P C44 H80 NO8 P C44 H84 NO8 P C46 H84 NO7 P C46 H86 NO8 P C46 H84 NO8 P C48 H86 NO8 P C27 H44 N7 O17 P3 S

0.001 <10e-4 <10e-4 <10e-4 0.014 <10e-4 <10e-4 0.003 0.001 0.002

↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓ ↑ ↑

Palmitic acid LysoPC(18:2) PC(34:2) PC(36:4) PC(36:2) PE(38:4) PC(38:3) PC(38:4) PC(40:5) trans-2-Hexenoyl-CoA

5.11 5.11 5.11 5.11 5.11 5.14 5.18 5.22 5.26

730.5375 758.5700 782.5703 786.5992 794.6033 551.5021 702.5600 834.6000 810.6011

730.5387 758.5700 782.5700 786.6013 794.6064 551.5039 702.5649 834.5989 810.6013

−1.6 0.0 0.4 −2.7 −3.8 −3.4 −6.9 1.3 −0.2

C40 H76 NO8 P C42 H80 NO8 P C44 H80 NO8 P C44 H84 NO8 P C46 H84 NO7 P C35 H66 O4 C39 H74 O6 C46 H86 NO8 P C46 H84 NO8 P

0.032 0.043 0.003 0.001 0.015 0.048 <10e-4 0.027 <10e-4

↓ ↓ ↓ ↓ ↓ ↓ ↑ ↑ ↑

PC(32:2) PC(34:2) PC(36:4) PC(36:2) PE(38:4) DG(P-14:0/18:1) Trilauric glyceride PC(38:3) PC(38:4)

Total PCOS vs control

PC, Phosphatidylcholine; LysoPC, lysophosphatidylcholine; PS, Phosphatidylserine; PE, phosphatidylethanolamine; DG, diacylglycerol

suggested that the disturbance of metabolites by IR was greater than that by the condition of PCOS. PC (38:3) and PC (38:4) were present at a higher level in total PCOS patients, but were much lower in IR PCOS samples when compared to control subjects, the most plausible explanation was that PCOS and IR conditions have the inconsistent impacts on the metabolisms of PC (38:3) and PC (38:4). Thus, these metabolites are not sufficient to be biomarkers of PCOS patients. To further evaluate the sensitivity and specificity of the identified biomarkers of PCOS with or without IR complication, receiver operating curve (ROC) of targeted metabolites was generated and the corresponding area under curve (AUC) was calculated. The AUC value near to 1.0 has been considered with high diagnostic efficiency [33]. As shown in Fig. 4A, the potential biomarker contributing to the discrimination between IR PCOS patients and healthy controls was identified as trilauric glyceride with a high AUC of 0.937, while decanoylcarnitine (AUC = 0.715) was highlighted for distinguishing the non-IR PCOS from controls (Fig. 4B). To our best knowledge, the direct relationship between decanoylcarnitine and PCOS as well as the correlation of trilauric glyceride with IR PCOS have not been reported previously. However, it has been indicated that the content of a derivative of decanoylcarnitine, 6-keto-decanoylcarnitine, in the urine samples of PCOS patients is dramatically altered compared with healthy controls [34]. Our present study and the previous study consistently suggested a promising application of decanoylcarnitine and/or its related metabolites as potential diagnostic markers for PCOS without IR complication, while trilauric glyceride might represent a promising diagnostic biomarker for the pathological condition of IR PCOS.

3.5. Potential signaling pathways involved in PCOS with or without IR The underlying signaling pathways and molecular networks influenced by PCOS with or without IR complication were explored and visualized by MetPA, a web application for metabolomics analysis [35]. Potential biomarkers contributing to the separation of pairwise groups were imported into MetPA. The “Homo sapiens” library was selected as database, while hypergeometric test and relative-betweeness centrality were performed for over representation analysis and pathway topology analysis, respectively. In an intuitive manner, the graphic output containing 3 levels of view, i.e., metabolome view, pathway view and compound view, was obtained. As shown in Fig. 5A, the node with high impact value suggested potential targeted pathways, and glycerophospholipid metabolism, glycerolipid metabolism and fatty acid metabolism pathways were highlighted with the impact values of 0.14, 0.06 and 0.03, respectively. The positions of identified metabolite biomarkers were labeled in dark red in the individual targeted signaling pathway views. In Fig. 5B and C, the key positions of 4 metabolites, including PC (38:4), LysoPC (18:2) (C04230), PC (34:2) (C00416) and trilauric glyceride, that contribute to the separation of IR PCOS and control groups, indicated that the disturbance of glycerophopholipid and glycerolipid metabolisms are largely involved in the process of PCOS with IR complication. Consistently, the involvement of glycerophopholipid and glycerolipid metabolic dysfunction in the development of PCOS and the close association of this pathway with the susceptibility of insulin resistance were also reported in previous studies [34,36]. Additionally, palmitic acid and decanoylcarnitine, the biomarkers for distinguishing non-IR PCOS from control group, were presented in fatty acid metabolism

Y.-X. Chen et al. / Journal of Pharmaceutical and Biomedical Analysis 121 (2016) 141–150

view, which suggested that fatty acid metabolism was greatly disturbed when non-IR PCOS occurred (Fig. 5D). Actually, our previous study revealed that the abnormalities of lipid metabolism have been implicated in the pathogenesis of PCOS, as indicated by plasma phospholipid fatty acid profiles of PCOS patients was greatly altered when compared to the healthy controls [37]. 4. Conclusion In our present study, the metabolic alterations in PCOS patients with or without IR were characterized using LC–MSbased metabolomics approach. Clear metabolic differences were observed among IR PCOS, non-IR PCOS and healthy controls, and the disturbance of metabolite profile by IR complication was greater than that by PCOS condition. These variations involved significant perturbations in glycerophospholipid, glycerolipid and fatty acid metabolisms. Trilauric glyceride and decanoylcarnitine were identified as the potential biomarkers with the highest sensitivity and specificity for the diagnosis of PCOS patients with and without IR, respectively. Due to the limited size of samples used in the present study, it remains to be further investigated by using validation sample set and large-scale metabolomics study. Our findings also suggest that LC–MS-based metabolomics offers a promising strategy for the complementary diagnosis of PCOS disease and its IR complication. Counterbalance of the corresponding molecular events might contribute to the alleviation and treatment of PCOS and IR complication. Conflict of interest None of the authors have conflicts to disclose. Acknowledgements This work was supported by the grants from the Research Committee of the University of Macau (MYRG123-ICMS12 and MYRG111-ICMS13 to JB Wan) and from Macao Science and Technology Development Fund (010/2013/A1 to JB Wan). References [1] A. Dunaif, K.R. Segal, D.R. Shelley, G. Green, A. Dobrjansky, T. Licholai, Evidence for distinctive and intrinsic defects in insulin action in polycystic ovary syndrome, Diabetes 41 (1992) 1257–1266. [2] S. Franks, Polycystic ovary syndrome, N. Engl. J. Med. 333 (1995) 853–861. [3] F. Gonzalez, Inflammation in polycystic ovary syndrome: underpinning of insulin resistance and ovarian dysfunction, Steroids 77 (2012) 300–305. [4] G.A. Burghen, J.R. Givens, A.E. Kitabchi, Correlation of hyperandrogenism with hyperinsulinism in polycystic ovarian disease, J. Clin. Endocrinol. Metab. 50 (1980) 113–116. [5] M.E. Lujan, D.R. Chizen, R.A. Pierson, Diagnostic criteria for polycystic ovary syndrome: pitfalls and controversies, J. Obstet. Gynaecol. Can. 30 (2008) 671–679. [6] E. Diamanti-Kandarakis, A. Dunaif, Insulin resistance and the polycystic ovary syndrome revisited: an update on mechanisms and implications, Endocr. Rev. 33 (2012) 981–1030. [7] L.J. Shaw, C.N.B. Merz, R. Azziz, F.Z. Stanczyk, G. Sopko, G.D. Braunstein, S.F. Kelsey, K.E. Kip, R.M. Cooper-DeHoff, B.D. Johnson, V. Vaccarino, S.E. Reis, V. Bittner, T.K. Hodgson, W. Rogers, C.J. Pepine, Postmenopausal women with a history of irregular menses and elevated androgen measurements at high risk for worsening cardiovascular event-free survival: results from the National Institutes of Health—National Heart, Lung, and Blood Institute sponsored women’s ischemia syndrome evaluation, J. Clin. Endocrinol. Metab. 93 (2008) 1276–1284. [8] M. Kahsar-Miller, R. Azziz, The development of the polycystic ovary syndrome: family history as a risk factor, Trends Endocrinol. Metab. 9 (1998) 55–58. [9] C. Christakou, E. Diamanti-Kandarakis, Polycystic ovary syndrome—phenotypes and diagnosis, Scand. J. Clin. Lab. Invest. Suppl. 244 (2014) 18––22, discussion 21. [10] R. Azziz, Diagnostic criteria for polycystic ovary syndrome: a reappraisal, Fertil. Steril. 83 (2005) 1343–1346.

149

[11] E.A.-S.P.C.W.G. Rotterdam, Revised 2003 consensus on diagnostic criteria and long-term health risks related to polycystic ovary syndrome, Fertil. Steril. 81 (2004) 19–25. [12] R. Azziz, E. Carmina, D. Dewailly, E. Diamanti-Kandarakis, H.F. Escobar-Morreale, W. Futterweit, O.E. Janssen, R.S. Legro, R.J. Norman, A.E. Taylor, S.F. Witchel, Androgen Excess Society, Positions statement: criteria for defining polycystic ovary syndrome as a predominantly hyperandrogenic syndrome: an Androgen Excess Society guideline, J. Clin. Endocrinol. Metab. 91 (2006) 4237–4245. [13] B.A. Swiglo, M.H. Murad, H.J. Schunemann, R. Kunz, R.A. Vigersky, G.H. Guyatt, V.M. Montori, A case for clarity, consistency, and helpfulness: state-of-the-art clinical practice guidelines in endocrinology using the grading of recommendations, assessment, development, and evaluation system, J. Clin. Endocrinol. Metab. 93 (2008) 666–673. [14] J. Rycroft-Malone, Formal consensus: the development of a national clinical guideline, Qual. Health Care 10 (2001) 238–244. [15] D.R. Matthews, J.P. Hosker, A.S. Rudenski, B.A. Naylor, D.F. Treacher, R.C. Turner, Homeostasis model assessment: insulin resistance and beta-cell function from fasting plasma glucose and insulin concentrations in man, Diabetologia 28 (1985) 412–419. [16] D. Ryan, K. Robards, Metabolomics: the greatest omics of them all? Anal. Chem. 78 (2006) 7954–7958. [17] R. Kaddurah-Daouk, B.S. Kristal, R.M. Weinshilboum, Metabolomics: a global biochemical approach to drug response and disease, Annu. Rev. Pharmacol. Toxicol. 48 (2008) 653–683. [18] A.Z. Buzatto, A.C. de Sousa, S.F. Guedes, Z. Cieslarova, A.V. Simionato, Metabolomic investigation of human diseases biomarkers by CE and LC coupled to MS, Electrophoresis 35 (2014) 1285–1307. [19] O. Beckonert, H.C. Keun, T.M. Ebbels, J. Bundy, E. Holmes, J.C. Lindon, J.K. Nicholson, Metabolic profiling metabolomic and metabonomic procedures for NMR spectroscopy of urine, plasma, serum and tissue extracts, Nat. Protoc. 2 (2007) 2692–2703. [20] W. Atiomo, C.A. Daykin, Metabolomic biomarkers in women with polycystic ovary syndrome: a pilot study, Mol. Hum. Reprod. 18 (2012) 546–553. [21] L. Sun, W. Hu, Q. Liu, Q. Hao, B. Sun, Q. Zhang, S. Mao, J. Qiao, X. Yan, Metabonomics reveals plasma metabolic changes and inflammatory marker in polycystic ovary syndrome patients, J. Proteome Res. 11 (2012) 2937–2946. [22] H.F. Escobar-Morreale, S. Samino, M. Insenser, M. Vinaixa, M. Luque-Ramirez, M.A. Lasuncion, X. Correig, Metabolic heterogeneity in polycystic ovary syndrome is determined by obesity: plasma metabolomic approach using GC–MS, Clin. Chem. 58 (2012) 999–1009. [23] Y. Zhao, L. Fu, R. Li, L.N. Wang, Y. Yang, N.N. Liu, C.M. Zhang, Y. Wang, P. Liu, B.B. Tu, X. Zhang, J. Qiao, Metabolic profiles characterizing different phenotypes of polycystic ovary syndrome: plasma metabolomics analysis, BMC Med. 10 (2012) 153. [24] X. Zhao, F. Xu, B. Qi, S. Hao, Y. Li, Y. Li, L. Zou, C. Lu, G. Xu, L. Hou, Serum metabolomics study of polycystic ovary syndrome based on liquid chromatography–mass spectrometry, J. Proteome Res. 13 (2014) 1101–1111. [25] F. Dong, D. Deng, H. Chen, W. Cheng, Q. Li, R. Luo, S. Ding, Serum metabolomics study of polycystic ovary syndrome based on UPLC–QTOF–MS coupled with a pattern recognition approach, Anal. Bioanal. Chem. 407 (2015) 4683–4695. [26] W.B. Dunn, D. Broadhurst, P. Begley, E. Zelena, S. Francis-McIntyre, N. Anderson, M. Brown, J.D. Knowles, A. Halsall, J.N. Haselden, A.W. Nicholls, I.D. Wilson, D.B. Kell, R. Goodacre, C. Human Serum Metabolome, Procedures for large-scale metabolic profiling of serum and plasma using gas chromatography and liquid chromatography coupled to mass spectrometry, Nat. Protoc. 6 (2011) 1060–1083. [27] L. Liu, J. Aa, G. Wang, B. Yan, Y. Zhang, X. Wang, C. Zhao, B. Cao, J. Shi, M. Li, T. Zheng, Y. Zheng, G. Hao, F. Zhou, J. Sun, Z. Wu, Differences in metabolite profile between blood plasma and serum, Anal. Biochem. 406 (2010) 105–112. [28] Z. Yu, G. Kastenmuller, Y. He, P. Belcredi, G. Moller, C. Prehn, J. Mendes, S. Wahl, W. Roemisch-Margl, U. Ceglarek, A. Polonikov, N. Dahmen, H. Prokisch, L. Xie, Y. Li, H.E. Wichmann, A. Peters, F. Kronenberg, K. Suhre, J. Adamski, T. Illig, R. Wang-Sattler, Differences between human plasma and serum metabolite profiles, PLoS One 6 (2011) e21230. [29] N.D. Ridgway, The role of phosphatidylcholine and choline metabolites to cell proliferation and survival, Crit. Rev. Biochem. Mol. Biol. 48 (2013) 20–38. [30] T. Matsumoto, T. Kobayashi, K. Kamata, Role of lysophosphatidylcholine (LPC) in atherosclerosis, Curr. Med. Chem. 14 (2007) 3209–3220. [31] K. Yea, J. Kim, J.H. Yoon, T. Kwon, J.H. Kim, B.D. Lee, H.J. Lee, S.J. Lee, J.I. Kim, T.G. Lee, M.C. Baek, H.S. Park, K.S. Park, M. Ohba, P.G. Suh, S.H. Ryu, Lysophosphatidylcholine activates adipocyte glucose uptake and lowers blood glucose levels in murine models of diabetes, J. Biol. Chem. 284 (2009) 33833–33840. [32] M.N. Barber, S. Risis, C. Yang, P.J. Meikle, M. Staples, M.A. Febbraio, C.R. Bruce, Plasma lysophosphatidylcholine levels are reduced in obesity and type 2 diabetes, PLoS One 7 (2012) e41456. [33] R.G. Moore, A.K. Brown, M.C. Miller, S. Skates, W.J. Allard, T. Verch, M. Steinhoff, G. Messerlian, P. DiSilvestro, C.O. Granai, R.C. Bast Jr., The use of multiple novel tumor biomarkers for the detection of ovarian carcinoma in patients with a pelvic mass, Gynecol. Oncol. 108 (2008) 402–408. [34] T. Zhou, M. Wang, H. Cheng, C. Cui, S. Su, P. Xu, M. Xue, UPLC-HRMS based metabolomics reveals the sphingolipids with long fatty chains and olefinic bonds up-regulated in metabolic pathway for hypoxia preconditioning, Chem. Biol. Interact. 242 (2015) 145–152.

150

Y.-X. Chen et al. / Journal of Pharmaceutical and Biomedical Analysis 121 (2016) 141–150

[35] J. Xia, D.S. Wishart, MetPA: a web-based metabolomics tool for pathway analysis and visualization, Bioinformatics 26 (2010) 2342–2344. [36] R.S. Legro, J.M. McAllister, Heirarchical clustering and beyond in PCOS endometrium: brave new world, J. Clin. Endocrinol. Metab. 94 (2009) 1084–1085.

[37] X.J. Zhang, L.L. Huang, H. Su, Y.X. Chen, J. Huang, C. He, P. Li, D.Z. Yang, J.B. Wan, Characterizing plasma phospholipid fatty acid profiles of polycystic ovary syndrome patients with and without insulin resistance using GC–MS and chemometrics approach, J. Pharm. Biomed. Anal. 95 (2014) 85–92.