Pre- and post-menopausal women

This study was approved by the Ethics Committee of Harbin Medical University. One hundred and fifteen women aged 39 to 66 years from the Daoli district in Harbin, China were recruited with informed written consents. None of the participants had any metabolic diseases, liver/renal diseases or inflammatory diseases when entering the current study. Menstrual history questionnaires were used to obtain information on menopausal status. Fifty-two women aged 39 to 52 years reported regular menstruation without hormonal agents and were deemed as pre-menopausal, and 63 women aged 46 to 66 years reported amenorrhea for more than 12 months and were defined as post-menopausal. After fasting for 12h, 5 mL of whole blood was collected by venipuncture from each participant. The fresh blood samples were centrifuged at 1,000×g for 10 min and the supernatant was extracted and frozen at –80°C.

Sample pretreatment

The plasma samples were thawed in a 4°C refrigerator for 40 min. A 200 μL aliquot of each plasma sample was mixed with acetonitrile (3×200 μL) and vortexed for 1 min. After mixing, the solutions were centrifuged at 12,000 ×g for 15 min at 4°C. An aliquot (300 μL) of each supernatant was transferred into a clear vial and dried in a vacuum rotary dryer, and the residue was dissolved in 100 μL of acetonitrile/water (1:3, v/v). The solutions were vortexed for 5 min, and then centrifuged at 12,000 ×g for 15 min at 4°C. The supernatants were then placed into sample vials for analysis. Before UPLC/MS analysis, all the samples were randomized. The pooled quality control (QC) samples were prepared by mixing equal amounts of plasma samples from 5 pre-menopausal subjects and 5 post-menopausal subjects. One QC sample was run every 10 samples.

Chromatography

A 10 μL aliquot of the pretreated sample was injected into an ACQUITY UPLC BEH C18 column (2.1 × 100mm, 1.7μm; Waters Corporation, Milford, MA) and analyzed using an UPLC system (Waters Corporation). The UPLC mobile phase consisted of acetonitrile containing 0.1% formic acid (solvent A) and water containing 0.1% formic acid (solvent B). The flow rate was set at 0.3 mL/min. A linear mobile phase gradient was used as follows: 1% A, held for 0.5 min; 0.5–4.0 min, increased from 1 to 15% A; 4.0–4.5 min, increased from 15 to 55% A; 4.5–11.5 min, increased from 55 to 90% A; 11.5–12.0 min, increased from 90 to 99% A; and held at 99% A from 12.0–15.0 min. After the analytical run, the mobile phase was returned to 1% A in 0.1 min and equilibrated at 1% A for 1 min.

Mass spectrometry

Mass spectrometry was performed on an Agilent 6520-QTOF (Agilent Technologies) operating at both positive-ion (ESI+) and negative-ion (ESI–) electrospray ionization (ESI) modes. The capillary voltage was 4.0 kV in ESI+ mode and 3.5 kV in ESI- mode. The desolvation temperature was 330°C, the gas flow rate was set at 10 L/min, and nitrogen was used as the drying gas. Centroid data were collected in the full scan mode from 50 to 1000 m/z.

Data preprocessing and annotation

The raw UPLC-QTOF/MS data were first converted into mzData-format files by MassHunter Qualitative Analysis Software (Agilent Technologies). Then these files were imported to the XCMS package in the R platform for preprocessing, which included peak detection, peak matching and retention time alignment.[13] The following XCMS parameters were utilized: xcmsSet (method = "centWave", peakwidth = c(5, 20)); group (bw = 10); retcor (method = "obiwarp"). The other parameter settings were as default. The CAMERA package in R was further used for annotation of isotope peaks, adducts and fragments.[14]

Statistical analysis

Principal component analysis (PCA) and partial least-squares discriminant analysis (PLS-DA) were performed on SIMCA-P 11.5 software. Welch’s t-test, heat map and false discovery rate (FDR) were performed using R. PCA was first used to detect grouping trends and outliers.[15] PLS-DA was then performed to understand global metabolic changes between pre-menopausal women and post-menopausal women.[15] To avoid overfitting, permutation tests with 100 iterations were performed to validate the supervised model.[16] Variable importance in the projection (VIP) for each metabolite was calculated based on the established PLS-DA model. In addition, the Welch’s t test was performed to determine the significance of each metabolite, and the corresponding FDR value was estimated for correcting multiple comparisons.[17] To visualize the metabolite changes by menopause, the standardized metabolite level in 115 women was plotted in a heat map.

A metabolic view on menopause

Previous metabolomic studies on menopause have observed evident relationships between menopause and metabolic changes.[8, 21] Auro et al. utilized a NMR-based platform to investigate the metabolism of menopause at the population level and found that menopause status was associated with total, monounsaturated, omega-7 and -9 fatty acids and several amino acids.[8] However, this study only targeted 135 molecules, of which, 74 were macromolecular lipoproteins. Yamatani et al. also reported that levels of fatty acid metabolites were significantly higher in the visceral fat of post-menopausal women than in premenopausal women.[21] In addition to fatty acid metabolism, our study look at lipid profiles in more details and at hormone metabolism and amino acid metabolism at a small-molecular level, adding to the current understanding of menopause-induced metabolic changes.

The most remarkable changes in post-menopausal women were characterized by significantly increased lipid-related metabolites, including LPCs, LPEs, acylcarnitines, and fatty acids. Post-menopausal women are known to be highly susceptible to these diseases. Studies have shown that LPC acts synergistically with endothelin-1 in inducing vascular smooth muscle cell proliferation.[22] Accumulation of fatty acid intermediates in the cardiomyocyte cytosol could result in lipotoxicity and increase cardiovascular risk in menopausal women.[23] In addition, studies have demonstrated that products of lipid and lipoprotein oxidation may contribute to reduced bone mass and bone fragility.[24, 25] An increasing body of biological and epidemiological evidence has also provided support for a link between CVDs and osteoporosis.[26] From this point of view, lipid metabolism might be a one of the common mechanisms underlying the pathogenesis of both CVD and osteoporosis. After all, the significant changes observed in lipid-related metabolites in postmenopausal women certainly warrant further investigation with respect to chronic diseases such as CVDs and osteoporosis.

Significantly decreased concentrations of pregnanediol-3-glucuronide, androsterone sulfate and DHEAS were detected in post-menopausal women. These compounds are all hormone metabolites or precursors,[19] and the changes observed in them indicated a significant drop of hormones after menopause. Hormonal changes could lead to a cascade of physiological and pathological changes, which could account for many phenomena and diseases associated with menopause (e.g. osteoporosis).[27] Studies have also shown that DHEAS levels are inversely associated with mortality of coronary artery diseases,[28] although the underlying mechanisms have yet to be determined. In addition, two amino acid metabolites, p-hydroxyphenylacetic acid and dihydrolipoic acid, were found down-regulated in post-menopausal women. In recent years, amino acid changes have been linked to increased risk of diabetes and CVDs.[29] It has been suggested that dihydrolipoic acid has antioxidant activity against microsomal lipid peroxidation, which contributes to atherosclerosis.[19] The exact roles of these metabolic changes in menopause-related diseases deserve our attention and await further investigation.

Menopause as a potential confounder in metabolomics

Over the last decade, the number of research articles published in the field of metabolomics has increased dramatically.[30] Compared with genomics or proteomics, metabolomics can reflect changes in biological phenotypes more directly and immediately.[5] However, metabolic changes can be attributed to a variety of factors, including both biological sources of interest and uncontrolled confounding sources. Our study demonstrated that menopause could be an important contributor to the metabolic profile of body fluids, which makes menopause as a possible confounder in metabolomic studies. Many metabolomic studies have used designs that do not account for the effects of menopause.[10, 31–33] Remarkably, some of these studies have identified biomarkers that overlapped with menopause-related biomarkers identified in the present study.[10, 31, 32] In this case, it is difficult to distinguish whether these biomarkers resulted from biological sources of interest or uncontrolled menopause status, and these biomarkers require verification. In future studies, menopause should be considered and balanced in metabolomics designs. Alternatively, menopause should be at least adjusted for in statistical analysis to avoid any false discoveries. This would increase the reliability and repeatability of metabolomic studies.

In this study, total cholesterol (TC) and low density lipoprotein (LDL) were imbalanced between pre-menopausal women and post-menopausal women. However, all the biomarkers remained statistically significant after adjusting for TC and LDL in logistic regression models. In addition, the widely-used acetonitril extraction would restrict lipid profiling to more hydrophilic lipids, which prevented a more comprehensive understanding of lipid metabolism of menopause. Lipidomics would be needed to validate and complement current findings.