Metabolic profiles characterizing different phenotypes of polycystic ovary syndrome: plasma metabolomics analysis

This randomized study population consisted of 217 PCOS patients and 48 women of similar age as controls, who visited the Division of Reproductive Center, Peking University Third Hospital, from March 2010 to March 2011. According to 2003 Rotterdam criteria, the diagnostic traits of PCOS are the presence of two or more of: oligo-ovulation and/or anovulation, clinical and/or biochemical signs of hyperandrogenism, and polycystic ovaries after exclusion of other etiologies (congenital adrenal hyperplasia, androgen-secreting tumors, Cushing's syndrome, 21-hydroxylase-deficient non-classic adrenal hyperplasia, androgenic/anabolic drug use or abuse, thyroid dysfunction, hyperprolactinemia, type 2 diabetes mellitus and cardiovascular disease). Women who had received any hormonal treatment or insulin-lowering agent during the last 3 months were excluded from the study. The control subjects were selected from women attending the clinic on account of male azoospermia. All controls had regular menstrual cycles and normal androgen levels. This study was approved by the Ethics Committee of Peking University Third Hospital. Informed consent was obtained from all participants prior to inclusion in this study.

Venous blood (3 ml) was collected into a heparin sodium tube and the plasma was collected by centrifugation. Each plasma sample was split into two aliquots and run in parallel using two different analytical platforms, as described previously [16, 17].

For the ¹H NMR measurement, an aliquot of 300 μl of plasma was mixed with 250 μl D₂O and 50 μl 3-trimethylsilyl-²H₄-propionic acid sodium salt (TSP) in D₂O (1 mg/ml) in a 5 mm NMR tube. The D₂O provided a field-frequency lock solvent for the NMR spectrometer and the TSP served as an internal reference of chemical shift. ¹H NMR spectra of the plasma samples were acquired on a Varian INOVA 600 MHz NMR spectrometer at 27 °C by using Carr-Purcell-Meiboom-Gill (CPMG) spin-echo pulse sequence with a total spin-spin relaxation delay (2nτ) of 320 ms. The free induction decays (FIDs) were collected into 32K data points with a spectral width of 8,000 Hz and 64 scans. The FIDs were zero-filled to double size and multiplied by an exponential line-broadening factor of 0.5 Hz prior to Fourier transformation (FT). In addition, diffusion-edited experiments were also carried out with bipolar pulse pair-longitudinal eddy current delay (BPP-LED) pulse sequence. The gradient amplitude was set at 35.0 G/cm, with a diffusion delay of 100 ms. A total of 128 transients and 16K data points were collected with a spectral width of 8,000 Hz. A line-broadening factor of 1 Hz was applied to FIDs before Fourier transformation. All plasma ¹H NMR spectra were manually phased and baseline-corrected using VNMR 6.1C software (Varian, Inc.). For CPMG spectra, each spectrum over the range of δ 0.4 to 4.4 was data reduced into integrated regions of equal width (0.01 ppm). For BPP-LED data, each spectrum over the range of δ 0.1 to 6.0 was segmented into regions of equal width (0.01 ppm). The regions containing the resonance from residual water (δ 4.6 to 5.1) were excluded. The integral values of each spectrum were normalized to a constant sum of all integrals in a spectrum in order to reduce any significant concentration differences between samples. Identification of metabolites in spectra was accomplished based on information in the literature and the Chenomx NMR Suite 5.0 (Chenomx, Calgary, Canada).

For the GC/TOF-MS measurement, plasma samples (100 μl) were thawed before the immediate addition of 500 μl of methanol (100%) to stop enzymatic activity. The samples were vortexed thoroughly, and 20 μl of a ribitol stock solution (0.2 mg/ml) was added as an internal reference. The mixture was placed on a shaker at 70°C for 15 minutes and centrifuged at 10,000 g for 10 minutes. The supernatant was mixed with 500 μl of pure water and 250 μl of chloroform, and was centrifuged at 4,000 rpm for 15 minutes. The upper (polar) phase was separated and then evaporated to dryness under a stream of N₂ gas in a thermostatically controlled water bath (60°C). Methoxyamine hydrochloride (20 μl, 20 mg/ml pyridine) was then added to the dried fraction of the polar phase. Following continuous shaking at 30°C for 90 minutes, 40 μl of N-methyl-N-trimethylsilyltrifluoroacetamide (MSTFA) was added, and the tube was incubated at 37°C for 30 minutes, then kept at room temperature for 120 minutes. Solutions (0.3 μl) were injected at a split ratio of 25:1 into a GC/TOF-MS system consisting of an HP 6890 gas chromatograph and a time-of-flight mass spectrometer (Waters Co., Milford, MA, USA) on a 30 m DB-5 column (250 μm inner diameter., 0.25 μm film; Agilent Technologies, Palo Alto, CA, USA). The injection temperature was 230°C, the interface temperature was set to 290°C, and the ion source temperature was adjusted to 220°C, with electron energy of 70 eV. Helium was set to a column flow rate of 1 ml/min. After a 5-minute solvent delay time at 70°C, the oven temperature was increased to 310°C in increments of 5°C/min, followed by a 1-minute isocratic cool down to 70°C and an additional 5-minute delay. MassLynx software (Waters Co.) was used to acquire the chromatographs. NIST02 libraries with electron impact (EI) spectra were searched rigorously for all the peaks detected with the total ion current (TIC), to identify the metabolites. Compounds were also identified by comparison of their mass spectra and retention times with those of commercially available reference compounds. The integral values of each spectrum were normalized to the integral of ribitol stock in order to reduce any significant concentration differences between samples.

Multivariate pattern recognition analysis was carried out by using SIMCA-P plus software (V. 10.0). Principal component analysis (PCA) was performed for data from different groups to detect the distributions and separations among those groups. Prior to PCA, all data variables were mean centered and preprocessed using orthogonal signal correction (OSC) to remove variations from non-correlated factors such as the instability of the spectrometer, inconstancy in sample preparation, and variability of some metabolites depending on the subject.

Of 217 PCOS patients according to the Rotterdam guidelines, they were 72 patients (33.2%) presented with the classic PCOS phenotype (HA+AO+PCO); 74 patients (34.1%) with chronic anovulation and polycystic ovaries but no clinical or biochemical hyperandrogenism (AO+PCO); 33 patients (15.2%) with hyperandrogenism and chronic anovulation but normal ovaries (HA+AO); 38 patients (17.5%) with hyperandrogenism and polycystic ovaries but ovulatory cycles (HA+PCO). Baseline characteristics for patients with different phenotypes of PCOS and controls were described in Table 1. A broad spectrum of metabolic and biochemical changes in PCOS subgroups compared with the control group were indicated, including the obviously enhanced triglyceride and low-density lipoprotein (LDL) level and reduced HDL level, a marked augment in serum concentrations of androgens such as androstenedione, increased ratio of luteinizing hormone (LH) to follicle-stimulating hormone (FSH). Overall, these changes in PCOS patients were clinically associated with the phenotypes and complemented the metabolomic analysis.

The plasma samples from the PCOS and control groups were analyzed using ¹H NMR spectroscopy. OSC-PCA was performed to evaluate the population structure of each PCOS subgroup comparing to the control group (Figure 1). The metabolites responsible for the differences among the various groups were identified and summarized in Table 2. The results showed that all of the four PCOS phenotypes had higher levels of very low-density lipoprotein (VLDL), LDL, fatty acids, unsaturated fatty acids and an unidentified sugar (δ 3.74 ppm), but had lower levels of phosphatidylcholine and lysyl-albumin as compared with the control group, respectively, which indicated the aberrance of lipid metabolism in PCOS. Specially, elevated level of lactate and reduced level of glucose in PCOS phenotypes except the feature (HA+AO) showed the carbohydrate metabolic disorder in the patients with polycystic ovaries.

However, a total of 45 plasma metabolites were identified as endogenous metabolites by GC/TOF-MS analysis, including amino acids, fatty acids, sugars, and organic acids. Multivariate statistical analysis was performed, and the OSC-partial least squares (PLS) scores plot illustrated that each PCOS phenotype exhibited obvious differences with the control group (Figure 2). Moreover, there were diverse significant differences were noted in the plasma levels of each PCOS phenotypes compared with the control group. The productions of lactate and long-chain fatty acids (for example, linoleic acid, palmic acid and stearic acid) were induced, whereas the production of glucose was inhibited in PCOS groups, which was consistent with ¹H NMR data (Table 3).

Amino acids play important roles both as basic substrates and as regulators in many metabolic pathways. Interestingly, marked change in the plasma amino acid pattern was detected by GC/TOF-MS analysis in each PCOS phenotype comparing to the control, suggesting the abnormality of amino acids catabolism and biosynthesis (Table 3). We observed that the concentrations of valine and tryotophan were generally elevated in PCOS groups. By contrast, the levels of glycine and proline in four phenotypes of PCOS were decreased significantly compared with the corresponding control group levels. Additionally, higher levels of alanine, serine, threonine, phenylalanine, ornithine and tyrosine were observed in anovulatory PCOS patients. Surprisingly, plasma serine and threonine levels were inhibited in women with PCOS with normal ovulation as compared with controls. Moreover, the concentrations of total endogenous amino acids and gluconeogenic amino acids were markedly increased in the classic and nonhyperandronetic PCOS phenotypes (A and B groups) compared to the control samples, indicating the induced proteolysis and inhibited gluconeogenesis in these women with PCOS. In particular, although both BACC and AAA concentrations were augmented in PCOS samples, the ratio of BCAA to AAA was dramatically reduced in anovulatory PCOS phenotypes as compared with the control group, respectively.

Approximately 40% patients with PCOS in our study were obese (BMI ≥25) and 30% were suffered from insulin resistance (Homeostasis Model Assessment of Insulin resistance (HOMA-IR) ≥2.69) [18]. To evaluate the potential interferences of obesity and insulin resistance with the metabolic abnormalities in PCOS, we performed the correlation analysis to test the associations between specific metabolites and PCOS, BMI and IR. Taken together, we observed the significantly positive association of linoleic acid, stearic acid, alanine, serine and tryptophan concentrations with the occurrence of PCOS disease, and the obviously negative relations of glucose, proline and isoleucine levels with PCOS controlling for age, BMI and IR (Additional file 1). However, plasma concentrations of valine, glycine, serine and threonine were closely correlated with IR and obesity (Additional file 2). Additionally, significant positive association of lactate and leucine concentrations with IR was observed independently of obesity. Extraordinarily, significant decreases of serine and threonine levels were observed in both obese women with PCOS and patients with IR comparing to the nonobese PCOS subjects and patients with normal insulin sensitivity, respectively. In contrast, serine and threonine concentrations in PCOS samples were notably enhanced as compared with the normal controls independently of obesity or IR, indicating opposite impacts of PCOS and its two common features on the metabolism of serine and threonine (Additional files 3 and 4).

With the aid of the Kyoto Encyclopedia of Genes and Genomes (KEGG) database, metabolic pathways associated with PCOS were summarized in Figure 3 based on the changes of intermediates concentrations detected in this study. The affected metabolic pathways in PCOS patients included the tricarboxylic acid (TCA), glycolysis, ketogenesis, lipolysis, proteolysis and urea cycles. Elevated rates of peripheral glucose uptake and diminished hepatic lactate conversion to glucose suggested the enhanced glycolysis in muscle and inhibited gluconeogenesis in liver during occurrence of PCOS. Additionally, the PCOS patients were accompanied by marked impairment of TCA cycle. This was reflected in the reduced citrate level and elevated levels of plasma threonine, valine, phenylalanine and tyrosine, inducing a drop in succinyl-CoA and fumarate. In addition, the women with PCOS advanced quickly to a state of lipolysis and protein catabolism, evidenced by increased levels of plasma fatty acids and amino acids. An elevated level of ornithine and decreased level of arginine implied the imbalance of urea cycle (a reduced level of citrate and arginine has been reported [19]).