Study participants

83 asymptomatic women with RIF were recruited in the Center for Reproductive Medicine at the Second Affiliated Hospital of Chongqing Medical University between October 2022 and December 2023 with 6 months of follow-up after the last embryo transfer. RIF was defined as the failure of clinical pregnancy after 4 good quality cleavage stage embryo transfers or 2 good quality blastocyst transfers, with at least three fresh or frozen IVF cycles. Clinical pregnancy was defined as an intrauterine pregnancy up to 12 weeks of gestation. Miscarriage was defined as the loss of a pregnancy before the completion of 12 weeks of gestation. The inclusion and exclusion criteria were modified from a previous study (Supplemental Table S1) [20].

The clinical characteristics of the study population are detailed in Table 1. The initial cohort of 83 patients with RIF was reduced to 80 for the final analysis, with 40 patients in each of the non-CE and CE groups, as illustrated in Fig. 1. There were no significant differences between the groups in terms of age, BMI, duration of infertility, history of previous conception, AMH level, basal FSH level, basal LH level, basal E2 level, basal P level, and basal PRL (P > 0.05).

Table 1 Clinical characteristics of the study populationFig. 1

Flow diagram of the study

All patients with RIF underwent vaginal smear examination, and those with a dominant Lactobacillus flora were selected for inclusion. A total of 83 RIF patients were enrolled. These patients underwent hysteroscopy and endometrial biopsy to confirm the presence of uterine structural normalities and endometritis. No uterine structural abnormalities were detected. Based on the examination results, patients were categorized into endometritis and non-endometritis groups. Endometritis group patients were treated with antibiotics and underwent repeat endometrial biopsy. Three patients with persistent endometritis were excluded. The remaining endometritis and non-endometritis group patients had their endometrial flushing fluid collected for 16 S rRNA sequencing analysis

RIF, recurrent implantation failure; CE, chronic endometritis

Examination of vaginal smear

During speculum examination, vaginal swabs were collected and smeared on glass-slides. Then, the slides were stained with Gram’s method and observed by oil lens (1000×). The Nugent scoring was performed according to a standard protocol [21]. Large gram-positive rods were considered as Lactobacillus.

Diagnosis and management of CE

At present, there is no universally accepted consensus on the diagnosis of CE. The diagnosis of CE in this study was based on a combination of hysteroscopy and immunohistochemistry, as previously described [22, 23]. For the clinical management of CE, particularly in the context of patients with RIF, it is essential to identify the underlying causes of implantation failure. This was achieved through hysteroscopy, which allowed for the examination of abnormal uterine shape or structure and the detection of chronic endometritis. Patients without structural abnormalities were included in the study. Following this initial assessment, endometrial biopsy was conducted to obtain a tissue sample for histopathological examination. This examination was then performed to confirm the presence of CE.

Endometrial biopsy sampling and immunohistochemistry for CD138

Endometrial biopsy was performed in the follicular phase as described previously [24]. Endometrial samples were fixed in neutral formalin and embedded in paraffin for immunohistochemistry. All biopsy blocks were serially sectioned at a thickness of 6 μm and incubated with rabbit anti-human monoclonal CD138 antibody (10593-1-AP, Proteintech, No.666, Gaoxin Avenue D3-3, Wuhan, China) and the secondary antibody used was a horseradish peroxidas-conjugated affinipure goat anti-rabbit IgG (PR30011, Proteintech) following the protocol. Immunorecative signals were visualized and photographed with a Nikon DS-Fi3 camera on a Nikon E200 microscope (Nikon, Shinagawa Intercity Tower C, 2-15-3, Konan, Minato-ku, Tokyo, Japan). The number of CD138+ cells more than 4/HPF+ was diagnosed as CE [23].

Treatment of CE

In case CE diagnosis, Doxycycline (Jiangsu Lianhua Pharmaceutical Co., LTD, No.21, Wenfeng Road, Yangzhou, Jiangsu, China) 100 mg twice a day for 14 days was employed. In the follicular phase of the cycle following the therapy, endometrial biopsy and were immunohistochemistry for CD138 repeated. Three patients were excluded for persistent CE in the following analysis.

Cervical mucus and endometrial fluid Collection

Cervical mucus and endometrial fluid were collected from the same patient seven days after the luteinizing hormone surge in natural cycles. The perineum was cleaned by cotton swabs soaked in iodophor solution (Shandong Lilkang Medical Technology Co. Ltd, No.1, Lierkang Road, Dezhou, Shandong, China) with the patient in classic lithotomy position. After inserting a vaginal speculum, the vaginal secretions were removed by cotton swabs soaked in saline solution. Cervical samples were obtained using nylon flocked swabs. After removal of cervical mucus, endometrial fluid were collected with a double-lumen embryo transfer catheter (T-1,731,511, Pacific Contrast Scientific Instruments Co. Ltd, No.1777,Dazheng road, Jinan, Shandong, China) as previously described [17]. Briefly, the outer sheath of the catheter was inserted into the endocervix avoiding contact with the vaginal wall. Subsequently, the inner catheter was inserted into the sheath and advanced into the uterine cavity. 1.0 mL of saline solution was injected and withdrawn to collect the endometrial flushing fluid. After the inner catheter was re-sheathed, both the sheath and catheter were withdrawn from the uterine cavity.

DNA extraction

DNA isolation was performed as described previously [25]. Endometrial fluid samples were pre-treated with lysozyme (9001-63-2, Sigma-Aldrich, PO Box 14,508, St. Louis, MO, USA), lysostaphin (9011-93-2, Sigma-Aldrich), and mutanolysin (55466-22-3, Sigma-Aldrich). Total DNA was further extracted using a QIAamp DNA Blood Mini kit (Qiagen, Strasse 1, Hilden, Germany) according to the manufacturer’s instructions. The genomic DNA was quantified by NanoDrop 2000 (Thermo Scientific, 168 3rd Ave, Waltham, MA, USA) and its integrity was assessed by agarose gel electrophoresis. Additionally, no-template controls (NTCs) were included in the DNA extraction process to confirm the absence of contaminants from the extraction kit.

Polymerase chain reaction and 16 S ribosomal RNA gene sequencing

The V3-V4 region of the 16S rRNA gene was amplified by PCR with the barcode-index primers 338F (5’-ACTCCTACGGGAGGCAGCAG-3’) and 806R (5’-GGACTACHVGGGTWTCTAAT-3’) using a TransStart FastPfu DNA polymerase (TransGen Biotech, No.1 Yongtaizhuang North Rd, Beijing, China) on a GeneAmp 9700 thermocycler (Applied Biosystems, 42 North Rd, Wakefield, RI, USA). PCR reactions were performed as following program: 3 min of denaturation at 95℃; 25 cycles of denaturation at 95℃ for 30 s, annealing at 55℃ for 30 s, and elongation at 72℃ for 45 s; and a single extension at 72℃ for 10 min. To confirm the absence of contaminants, we included NTCs in the PCR amplification. The absence of amplification in the NTCs indicated that the PCR reagents were free from contaminants. The PCR products were recovered from a 2% agarose gel by AxyPrep DNA Gel Extraction Kit (Axygen Biosciences, 33,170 Central Ave, Union City, CA, USA). The quality and concentration of the purified amplicons were assessed by a Quantus™ Fluorometer (Promega, 2800 Woods Hollow Rd, Madison, WI, USA). Purified amplicons were pooled in equimolar amounts. Then, the library was constructed using NEXTFLEX Rapid DNA-Seq Kit (PerkinElmer, 940 Winter St, Waltham, MA, USA) following the manufacturer’s instruction. Paired-end sequencing was performed using a NovaSeq system (Illumina, 5200 Illumina Way, San Diego, CA) with PE250 platform.

Sequencing data analysis

Sequencing reads were treated for quality control and length-filter using fastp (0.19.6) [26]. Taxonomic assignment was performed following QIME2 pipeline [27]. Paired-end reads were merged by Fast Length Adjustment of Short reads (FLASh) tool and de-noised using DADA2 [28, 29]. The amplicon sequence variants (ASVs) generated by DADA2 were used for taxonomic assignment by naive Bayes classifier based on aligning the representative sequences to the SILVA rRNA database (SSU 138 release) [30].

Linear discriminant analysis effect size was used with the parameters (α = 0.05 and LDA score 3.0) to identify any genera that differential in relative abundance between the two groups [31].