Screening and identification of endometrial proteins as novel potential biomarkers for repeated implantation failure

Sample collection

This study was approved by the Ethics Committee of the Faculty of Medicine (Hangzhou Women’s Hospital, China) (Approval Number: 2016001-10). Written informed consent was obtained from all subjects before endometrial collection.

The workflow of our study is shown in Fig. 1. Failure of three or more cycles in which reasonably high-quality embryos were transferred was defined as RIF (Margalioth et al., 2006). Grade I and grade II embryos were determined to be high-quality embryos (Gardner et al., 1998). The blastocyst quality was determined according to the definitions by Gardner et al. (1998). Blastocysts were considered high-quality if they had a grade 3 or 4 blastocoel, a grade A or B inner cell mass, and a grade A or B trophectoderm on days 5 or 6. We collected data from fifty-two RIF cases and 135 pregnancy cases undergoing IVF-ET treatment at our hospital between September 2016 to December 2019. Data including age, follow-up outcomes, and clinical examination findings were collected. The endometrium was collected at WOI by hysteroscopy biopsy at LH+7. The endometrium was then washed with saline immediately and frozen in liquid nitrogen and stored until protein extraction.

Endometrium protein extraction

We randomly selected 2 pregnancy cases and 3 RIF cases for endometrial protein extraction. Protein extraction was performed using a lysis buffer after grinding the sample (100 mg) to powder in liquid nitrogen. Phenylmethanesulfonyl fluoride (PMSF) was added to a final concentration of 1mM and ethylene diamine tetraacetic acid (EDTA) was added to a final concentration of 2mM, held for 5 min, and then dithiothreitol (DTT) was added to a final concentration of 10mM. The sample was placed in an ice bath ultrasound for 5 min and the lysate was centrifuged at 15,000 g for 20 min. The supernatant was precipitated 5 times with 1 mL acetone and incubated at 20 °C for 2 h and then centrifuged at 15,000 g at 4 °C for 20 min. The precipitate was washed in chilled acetone, incubated at 20 °C for 30 min, and centrifuged again at 15,000 g and 4 °C for 20 min. The washing was repeated twice, then air dried and the precipitate was re-dissolved with the lysis buffer. The precipitate was centrifuged again at 15,000 g and 4 °C for 20 min after an ice bath ultrasound for 5 min. The supernatant was determined using the Bradford Protein Assay Kit to detect the protein concentration.

Protein digestion and iTRAQ-2D LC-MS/MS

A solution of 1:50 trypsin (Promega, USA)-to-protein mass was prepared. Then, a total of 100 μg of protein from each group was digested with prepared solution at 37 °C for 16 h. The peptides were reconstituted in 0.2M TEAB and processed according to the manufacturer’s protocol for 8-plex iTRAQ reagent (AB SCIEX, Framingham, MA, USA). Three biological replicates of the RIF group were labeled with 113, 114, and 115 isobaric tags, respectively. The peptides with two biological replicates from the pregnant group were labeled with 116 and 117 isobaric tags, respectively.

A high pH reversed-phase chromatography column (Phenomenex, Gemini-NX 3u C18110A, 150*2.00 mm) was used for the first-dimensional fractionation procedure. We collected 16 fractions in all, and then dried them for next LC-MS analysis. The fractions were re-suspended in 2% acetonitrile containing 0.1% formic acid, and then loaded into a C18 trap column (Acclaim PepMap 75 μm ×150 mm, C18, 3 μm, 100A). Then, online chromatography separation was performed on the nanoLC system (Dionex Ultimate 3000 RSLCnano) (Fan, Wang & Wang, 2019). The trapping and desalting procedures were carried out at a flow rate of 3 μL/min for 5 min with 100% solvent A (0.1% formic acid, 2% acet- onitrile and 98% water). The peptides were eluted using a 65 min gradient of buffer A (0.1% formic acid) to buffer B (80% ACN containing 0.1% formic acid) at 300 nL/min on an analytical column (Acclaim PepMap 75 μm ×15 cm C18-CL, 3 μm 100 Å, Thermo160321). Q Exactive system (Thermo Scientific) fitted with a Nanospray ion source was used to acquire tandem MS data. Specific steps are as follows: data were acquired using an ion spray voltage of 2.2 kV. MS spectra across the scan range of 350–1,800 m/z with a 70,000 resolution using maximum injection time (60 ms) per spectrum. Twenty of the most intense precursors per MS cycle were selected for fragmentation and were detected with 100 ms maximum injection time. Tandem mass spectra were recorded at a 17,500 resolution with the rolling collision energy turned on and iTRAQ reagent collision energy adjustment turned on. The lock mass option was enabled for more accurate measurements. Dynamic exclusion was set for 10 s.

Finally, the acquired MS/MS data were analyzed using IPeak and IQuant software as former researchers (Wen et al., 2015; Wen et al., 2014; Fan, Wang & Wang, 2019). Only proteins identified at global FDR ≤1% with ≥1 peptide were considered for further downstream analysis. A differentially expressed protein was determine only if it is identified and quantified with at least one significant peptide with the P <0.05 and fold change >1.5.

Bioinformatics analysis

Principal components analysis (PCA) was performed to confirm the sample repeatability. Functional annotation was performed using the Gene Ontology (GO) database (http://www.geneontology.org) and included the cellular component, molecular function, and biological process. The differentially expressed protein–protein network was analyzed by STRING software (http://www.string-db.org/). The Kyoto Encyclopedia of Genes and Genomes (KEGG) database (http://www.genome.jp/kegg/or http://www.kegg.jp/) was used to predict the main metabolic pathways (Kanehisa et al., 2007). We obtained the significantly enriched GO/pathway items by hypergeometric test. The EggNOG database (http://eggnogdb.embl.de) was used for pairwise orthology predictions, functional annotation, and classification (Huerta-Cepas et al., 2015).

Western-blot analysis

We lysed endometrial tissues with 200 μ L of RIPA lysate (P0013B, Beyotime, Shanghai, China) plus 1 mM PMSF at 4  °C for 30 min, and then harvested the supernatant with centrifugation at 11,000 g for 10 min. The harvested protein concentrations were measured with a BCA quantitative kit (P0009, Beyotime). Samples were subjected to polyacrylamide gel electrophoresis, and were transferred onto a PVDF membrane (IPVH00010, Millipore, Massachusetts, USA). The membrane was blocked by 5% skimmed milk powder solution at room temperature for 2 h and was incubated with primary antibodies, including antithrombin III (rabbit monoclonal, ab126598, abcam, Cambridge, United Kingdom), cortisol binding globulin (rabbit monoclonal, ab110648, abcam), fetuin-A (alpha-2-HS-glycoprotein, rabbit monoclonal, ab137125, abcam), GAPDH (mouse monoclonal, 60004-1-Ig, proteintech, Beijing, China), and alpha tubulin (rabbit polyclonal, 11224-1-AP, proteintech) at 4 °C overnight. Secondary antibodies as goat anti-mouse IgG-HRP (BK0023, BEST, Xian, China) and goat anti-rabbit IgG-HRP (BK0027, BEST) were then incubated with membrane at room temperature for 1.5 h. The blots were visualized using the ECL Plus Luminous Kit (S17851, Yeasen, Shanghai, China). At last, the results were measured with Image J software.

Statistical analysis

Parametric data were tested using the chi-square test for the composition ratios and t-tests for means of two groups. Nonparametric analysis was carried out using the Mann–Whitney U-test. Parametric data were presented as mean ± SD while nonparametric data were presented as median ± IQR, and P < 0.05 was considered to be statistically significant by the SPSS software, version 16.0 (SPSS, Chicago, IL). Our clinical data was able to identify significant differences in 81.45% of RIF cases and controls at a statistical support level of α = 0.05 with a d = 0.5 applying a one tail model calculated by Gpower 3.0.5.

Clinical data analysis

We recruited 52 RIF patients and 135 pregnant patients undergoing IVF-ET treatment between September 2016 to December 2019. Data, including age, follow-up outcomes, and clinical examination findings were collated into databases. After correcting for age, since age may affect the pregnancy rate, a total of 40 subjects with RIF under 40 years old were paired with 80 pregnant subjects by age, body mass index, and treatment time and then analyzed. There were no significant differences between RIF patients and pregnant controls in general and clinical data (P > 0.05, Table 1).

Endometrial proteomics results

We identified a total of 6,102 proteins through iTRAQ-2D LC-MS/MS from 113,384 spectra and 31,024 peptides, respectively (Table S1). Among the 6,102 identified proteins, 5,840 had GO annotations (95.71% of all proteins); 5,504 had KEGG annotations (90.20% of all proteins); and 6,097 had EggNOG annotations (99.92% of all proteins).

We performed quality control on the quantitative results with volcano maps and the distributions of coefficient of variation. We selected the fold change for 6,102 proteins and found 285 proteins with fold change in RIF cases/pregnant controls >1.5 or <0.67. And we only used proteins with P <0.05. Further screening revealed 82 differentially expressed proteins in RIF patients compared with the pregnant controls, including 55 up-regulated proteins (>1.50-fold, P < 0.05) and 27 down-regulated proteins (<0.67-fold, P < 0.05) (Table 2). The hierarchical clustering provided a visualized mode to display the clustering patterns of the differentially expressed proteins between the groups (Fig. 2).

Bioinformatics analysis results

Gene Ontology analysis of differentially expressed proteins revealed that most of the proteins were involved in the response to stimulus (42 proteins), extracellular region (35 proteins), and structural molecule activity (10 proteins) (Fig. 3A). The most significantly enriched GOs were negative regulation of hydrolase activity, blood microparticle, and enzyme inhibitor activity through hypergeometric testing (Fig. 3B). Seven proteins (SPB6, APOA1, GMIP, THBG, CBG, ANT3, and FETUA) were identified in the hydrolase activity term, seven proteins (VTDB, IGHG4, APOA1, A1AG2, FETUA, ANT3, and A1AG1) were identified in the blood microparticle term, and another seven proteins (ANT3, FETUA, CBG, ASPN, SPB6, THBG, and APOA1) were identified in the enzyme inhibitor activity term. Among these proteins, ANT3 and FETUA were identified in prior studies (Hannan et al., 2010; DeSouza et al., 2005) which played an important molecular function in endopeptidase inhibitor activity by GO analysis (Fig. 3C). String analysis found interactions between these proteins (Fig. 3D) which assembled in two bunches: ribosomal proteins and blood homeostasis proteins.

In addition, the KEGG pathway mapping revealed the immune system (seven proteins), transport and catabolism (five proteins), and translation (five proteins) pathways (Fig. 4A). Enriched KEGG pathway analysis showed the ribosome and primary immunodeficiency pathways as significant with P <0.05 (Fig. 4B). Finally, we used the EggNOG database to determine that the differential proteins associated with RIF are mostly clustered in classifications including posttranslational modification, protein turnover, chaperones (22 proteins), translation, ribosomal structure and biogenesis (eight proteins), and carbohydrate transport and metabolism (six proteins) (Fig. 4C).

Western-blot results

We verified endometrial antihrombin-III (ANT3, P01008), corticosteroid-binding globulin (CBG, P08185), and fetuin-A (FETUA, P02765) levels using the Western-blot. We found significantly higher levels of CBG and fetuin-A in RIF patients (Fig. 5). A significant difference in CBG and fetuin-A was found in RIF patients using grayscale detection by Image J (1.39 fold, P = 0.003; 1.47 fold, P = 0.002; respectively).