The study was conducted in full compliance with government policies and the Helsinki Declaration, and the procedures were approved by institutional ethics committee of Nanjing Medical University. The subjects were the volunteers sequentially recruited from affiliated hospitals of Nanjing Medical University (NJMU). Every participant was fully informed about the purpose of this research, and written consent was obtained from each participant. All participants were ethnically Han Chinese. A questionnaire was used to collect information including personal background, lifestyle factors, occupational and environmental exposures, genetic risk factors, sexual and reproduction status, medical history, and physical activity. Cases included in this study were UMI with normal semen quality in accordance with WHO guidelines and normal physical as well as endocrine status. They were husbands of couples who attended the affiliated hospitals of NJMU because of conception failure for at least 12 months. Their wives had undergone a series of clinical tests and their reproductive function was evaluated as normal. The men had undergone complete historical and physical examinations and measurement of hormones as well as routine semen analysis; their partners had undergone complete historical and physical examinations, including conventional gynecological examination, serum hormone level measurements, tubal assessment, ovarian function testing, hysteroscopy, β-ultrasonic examination, immunologic test, microbial inspection, and karyotype analysis. Fertile controls (FC) in the present study were fertile men recruited and sampled at the same period as the UMI cases. They were from the early pregnancy registry of the same hospitals who were in the third month following a successful pregnancy. They were healthy men with normal reproductive function, prospectively confirmed by them fathering healthy babies 6~8 months later. Eighty UMI cases and 80 matched FCs were randomly included in training set; 50 UMI cases and 50 matched FC were randomly included in the validation set.

Semen samples were obtained in private by masturbation into a sterile wide-mouth and metal-free glass container after an at least 3-day abstinence. Routine semen analyses were performed using a computer aided semen analyzer (CASA) (WLJY 9000, Weili New Century Science and Tech Dev., China), while referring to guidelines in the fourth edition World Health Organization (WHO) Laboratory Manual for the Examination of Human Semen. After collection, the semen was liquefied at 37°C for 30 min and then analyzed in accordance with WHO guidelines; variables examined included semen volume, concentration, number per ejaculum, motility, vitality, progression, and motion parameters. Strict quality control measures were enforced throughout the entire study. Semen samples were assessed twice, and the counting procedure was automatic and performed by two technicians according to WHO recommendations. CASA results were compared directly with manual assessments using an improved Neubauer hemocytometer over a wide range of sperm concentrations. For sperm motility analysis, we used prepared videotapes to ensure that all motile spermatozoa were being identified using the current and consistent setup parameters.

After CASA analysis, each semen sample was prepared by Percoll gradient methods for separating viable spermatozoa and seminal plasma in the laboratory. Then all samples were stored at -80°C prior to GC-MS metabolic analysis.

The metabolic analysis and data processing were performed basically according to previous report [17]. Briefly, a 25 μl mixture of chloroform, methanol, and water (2:5:2) was added into 25 μl of seminal plasma and the samples were vortexed for 1 min. The samples were placed at -20°C for 20 min to extract metabolites, followed by centrifugation at 12,000 rpm for 10 min, and the liquid layer was transferred into a new tube. The residue was extracted with 50 μl of methanol using the homogenizer for 10 min followed by centrifugation at 12,000 rpm for 10 min. The supernatant was then combined with the previous extraction. After vortexing, a volume of 100 μl mixture was transferred to a glass vial spiked with internal standards (10 μl heptadecanoic acid at 1 mg/ml and 4-chlorophenylalanine at 0.3 mg/ml). Then the samples were vacuum dried at room temperature. The residue was chemically derivatized with a two-step procedure and then analyzed with Pegasus HT system (Leco Corporation, St Joseph, USA) coupled with an Agilent 6890N gas chromatography. Briefly, a 1 μl derivate was injected in splitless mode at 270°C. Helium was used as carrier gas; the flow rate was 1.0 ml/min. The transfer interface and ion source were set to 270°C and 220°C respectively. The oven temperature was set as a program which started at 80°C for 2 min, and then ramped to 180°C at a rate of 10°C/min, to 230°C with a rate of 6°C/min, finally to 295°C with a rate of 40°C/min and held for 8 min. MS data were acquired with m/z range of 30–600 at an acquisition rate of 20 spectrum/second. Compound identification for GC-TOFMS was performed by comparing the mass fragments with NIST Standard mass spectral databases using ChromaTOF software with a similarity of more than 70% and verified by available reference compounds.

An aliquot of 5 μl of each seminal plasma sample was pooled as the quality control sample. The QC samples were stored at -80°C and prepared once for each batch, and were injected every 10 samples in the GC-MS sequence. Responses of internal standards were used for creating a QC chart, those samples where deviation of internal standards were over control limit of 3 times of relative standard deviation (RSD) were re-analyzed. RSD of compounds in each QC sample were calculated; compounds that possessed an RSD over 30% were obsoleted.

Fisher's exact test was used to analyze the differences of categorical variables such as drinking and smoking status between cases and controls. T-test was used to analyze for differences in continuous variables such as age and body mass index (BMI). Student-T tests were engaged to identify the differential seminal plasma metabolites between cases and controls, the data were log-transformed when appropriate. To obtain statistically robust results, the Bonferroni method was employed for correcting for multiple comparisons.

Those statistical methods above mentioned were all carried out within R programming language. Principal component analysis (PCA) and orthogonal partial least squares discriminant analysis (OPLS-DA) models were performed with SIMCA-P+ 13.5 (Umetrics, Umea, Sweden).

As shown in Table 1, cases and controls were well matched for age, BMI, drinking status, and smoking status. No significant difference was observed between the two groups (Table 1).

A total of 333 metabolites were obtained from metabolic profiling of seminal plasma. 153 metabolites were identified by library matching, and 96 of those were validated by local standards. After application of QC criteria, 136 metabolites were found with RSD < 0.3. Ninety compounds of those were identified, and 70 were confirmed by local standards. A PCA model was created with a learning set using the 136 metabolites. PCA score plots (Fig 1) showed the separation trend between UMI and FS groups.

Subsequently, OPLS-DA models were established and demonstrated satisfactory modeling and predictive abilities with 1 predictive component and 2 orthogonal components (R²Xcum = 0.333, R²Ycum = 0.833, Q2cum = 0.634) with metabolic profiling data (Fig 2A). As a comparison, quality parameters of seminal plasma, including semen volume, concentration, number per ejaculum, motility, vitality, progression, and motion parameters, were used for constructing another OPLS-DA model (R²Xcum = 0.360, R²Ycum = 0.423, Q2cum = 0.331). The score plot showed a poorer discriminate efficiency (a sensitivity of 82% and specificity of 92%) than that created with metabolomics data (a sensitivity of 98% and specificity of 95%) (Fig 2B).

Furthermore, an independent set of samples were profiled as the validation set which consisted of 50 UMI and 50 FC. Profiling data from the validation set were used to test the prediction ability of the established OPLS-DA model. In the Y prediction scores plot, 41 out of 50 UMI samples and 46 out of 50 control samples in validation groups were correctly assigned to proper class, using a cutoff value of 0.5 (Fig 3, the Y value was set at 1 for UMI, and 0 for controls in the learning group). This result showed the ability of the OPLA model to predict the unknown samples to the right groups with a sensitivity of 82% and specificity of 92%. As all these UMI were diagnosed as fertile male under current WHO seminal plasmatic parameter standards, the OPLS-DA model is of great significance in the diagnosis of UMI.

Significantly altered metabolites in seminal plasma with a VIP threshold (VIP > 1) in above mentioned OPLS-DA model, as well as the Mann-Whitney U test between UMI and FC, were selected and summarized in Table 2. Among these 44 differential metabolites, 37 were confirmed by reference standards as shown in Table 2. Variations of these differential metabolites were described by fold changes where UMI subjected to FC. P-values and fold changes in validation group were also listed, and coincided well with those in the training set.

Pearson correlation coefficient was calculated with metabolic profiling data and quality parameters of seminal plasmatic, including semen volume, concentration, number per ejaculum, motility, vitality, progression, and motion parameters. Significant correlated metabolites and quality parameters were listed in Table 3.

Metabolic pathway analysis were carried out with MetaboAnalyst [18]. The seminal plastic biomarkers of UMI were input as data source. Seven from 80 homo-sapiens (human) pathways were significantly changed as shown in Table 4.