The Ethical Committee of the Institutional Review Board of Saint Mother Obstetrics and Gynecology Clinic and Institute for ART, Fukuoka, Japan, approved this study in November 2017 (No.17-ST-03). The study was registered in the University Hospital Medical Information Network-Clinical Trial Registration, Japan, in March 2018 (UMIN-CTR000031731). All recruited women provided written informed consent.
We enrolled 211 consecutive women diagnosed with RIF after ≥3 ET cycles using Gardner scoring system [15] grade 3BB or higher blastocysts 5–6 days post-fertilization between October 2017 and June 2018. All patients’ uterine and endometrial structures were evaluated via transvaginal ultrasonography. Patients with uterine cavity ultrasounds that revealed possible causes of infertility received a diagnostic hysteroscopy to rule out intrauterine disorders (e.g., endometrial polyps, submucosal myomas, and intrauterine adhesions). Twenty-nine women were excluded after the hysteroscopy. We also excluded 30 women with other possible risk factors for reproductive failure, including 13 with thrombophilia (e.g., antiphospholipid syndrome), 15 with endocrinologic abnormalities or collagen disease, and 2 with parental chromosomal imbalances or translocations. Seven women who had received antibiotics within 1 month of sampling were excluded because antibiotics can affect microbiota communities.
Finally, 145 women were included. Forty did not provide vaginal specimens; thus, we obtained 145 endometrial samples and 105 vaginal aspirates. We also recruited 21 healthy women with etiologies of infertility because their husbands had azoospermia as the control group; these women had regular menstrual cycles without causes of infertility such as tubal factors, ovulation disorder, endometriosis, endocrinologic abnormalities, or immunological abnormalities. Figure 1 shows the participant selection methods.
Vaginal and endometrial sampling
Vaginal and endometrial samples were taken 5–7 days after ovulation or the beginning of the high-temperature phase in the basal body temperature. For women with irregular menstrual cycles, samples were taken during the hormone replacement cycle. From days 1–3 of the menstrual cycle, 2–8 mg of oral estradiol valerate (Progynova®, Bayer Health Care, Schering, Germany) was administered. From day 13, oral chlormadinone acetate (8 mg; Lutoral, Shionogi, Osaka, Japan) was administered for 13 days. Samples were obtained 5–7 days after initiating oral progesterone intake.
Vaginal discharge was collected using two sterilized swabs after placing a sterilized vaginal speculum. One swab was submitted for Nugent scoring [16], which indicates BV; the other was used to analyze the microbiota. The latter swab was immediately soaked in OMNIgene® VAGINAL (DNA Genotek Inc., Ottawa, Ontario, Canada). The vagina was then washed with physiological saline and wiped three times with dry tampons to remove vaginal secretions and cervical mucus. To minimize the risk of contaminating the endometrial samples in the vagina, a Medgyn pipette IV (Harada Corporation, Tokyo, Japan) was inserted into the uterine cavity, avoiding contact with the vaginal walls. We inserted the pipette approximately 5 mm from the bottom of the uterus and pulled it back toward the cervical canal under abdominal ultrasound guidance and absorbed the samples while turning the pipette slowly for 45–60 seconds. All uterine samples were placed in an in utero solution with the endometrial tissue. We stopped the absorption to prevent uptake of any cervical mucus left after washing when the pipette neared the cervix, then quickly removed the pipette from the uterus. The pipette tip was cut to 3 cm with sterilized surgical scissors to prevent contamination with cervical mucus. The endometrial samples were placed into the OMNIgene® VAGINAL without touching the pipette to the liquid.
Evaluation of the Nugent score
Vaginal samples were smeared on glass slides, fixed over a flame, and Gram stained. The stained slides were then examined at 1000× magnification to evaluate the Nugent scoring [16], an index for microscopically evaluating Lactobacillus, Gardnerella and Mobiluncus in vaginal samples. The scoring requires training but is a highly reproducible standard method, with scores ranging from 0: >30 lactobacilli or no Gardnerella-like bacteria in the visual field to 4: no lactobacilli or >30 Gardnerella-like bacterium in the visual field. Mobiluncus presence is an additional 2 points. Subjects with Nugent scores of 0–3, 4–6, and ≥7 points were categorized into the normal, intermediate, and BV groups, respectively.
DNA extraction and 16S rRNA sequencing
Varinos Inc., Shinagawa, Tokyo, Japan, extracted and sequenced the bacterial DNA. The vaginal and endometrial samples were treated with proteinase K and lysozyme solution per the manufacturer’s instructions. Genomic DNA was extracted using an Agencourt Genfind v2 Blood & Serum DNA Isolation Kit (Beckman Coulter, Inc., Miami, FL, USA) or MagNA Pure 24 (Roche Diagnostics, Grenzach-Wyhlen, Germany) Pathogen 200 hp 1.0 protocol. The dsDNA concentration was quantified fluorometrically with a Qubit dsDNA HS Assay Kit (Thermo Fisher Scientific Inc., Waltham, MA, USA). The V4 hypervariable region of the bacterial 16S rRNA gene was amplified from the specimen’s DNA using the modified primer pair, 515f (5'-TCG TCG GCA GCG TCA GAT GTG TAT AAG AGA CAG GTG YCA GCM GCC GCG GTA A-3') and 806rB (5'- GTC TCG TGG GCT CGG AGA TGT GTA TAA GAG ACA GGG ACT ACN VGG GTW TCT AAT-3'), with Illumina Nextera XT adapter overhang sequences (underlined; Illumina, Inc., San Diego, CA, USA) [17]. PCR was performed using 25 ng/µL DNA, 200 µmol/L of each deoxynucleotide triphosphate, 400 nmol/L of each primer, 2.5 U FastStart HiFi polymerase, 4% of 20 mg/mL bovine serum albumin (Sigma-Aldrich, St. Louis, MO, USA), 0.5 mol/L betaine (Sigma), and the appropriate buffer with MgCl2 supplied by the manufacturer (Roche Molecular Systems, Inc., Pleasanton, CA, USA). Thermal cycling consisted of initial denaturation at 94°C for 2 minutes, followed by 30 cycles of denaturation at 94°C for 20 seconds, annealing at 50°C for 30 seconds, extension at 72°C for 1 minute, and a final extension at 72°C for 5 minutes. The amplicon mixture was purified using Agencourt AMPure XP (Beckman Coulter, Inc.). Purified PCR samples were multiplexed using a dual-index approach with the Nextera XT Index kit v2 (Illumina, Inc.) per the Illumina 16S Metagenomic Sequencing Library preparation protocol. PCR indexing was performed using KAPA HiFi HotStart ReadyMix (Roche Sequencing Solutions Inc., Pleasanton, CA, USA) in a 50-µl reaction volume and subsequently purified using Agencourt AMPure XP beads. The final library was paired-end sequenced at 2× 200 bp using a MiSeq Reagent Kit v3 on the Illumina MiSeq platform. The ZymoBIOMICS Microbial Community Standard (Zymo Research, Irvine, CA, USA), containing a mixture of Pseudomonas, Escherichia, Salmonella, Lactobacillus, Enterococcus, Listeria, Bacillus, and two yeast species, was used as a positive control. UltraPure™ DNase/RNase-Free Distilled Water (Thermo Fisher Scientific, Inc.) was used as a blank control.
Sequenced reads were merged using EA-Utils fastq-join [18] to obtain a 291-bp median merged sequence length. Quality control of the merged sequence was performed using USEARCH v10.0.240 [19] to remove PhiX reads, truncate primer-binding sequences, and discard sequences of <100 bp and with a sequence quality <Q20. Quantitative Insights Into Microbial Ecology (QIIME) 1.9.1 was used with the default parameters for quality filtering, chimera checking, sequence clustering into operational taxonomic units (OTUs), and taxonomic assignment [20]. Sequences were clustered into OTUs using an open-reference OTU-picking strategy using the UCLUST method based on 97% sequence identity. Taxonomy was assigned to each OTU using RDP Classifier [21] with a 0.50 confidence threshold against the Greengenes database, version 13_8 [22].
Sequencing results and OTU analysis
Seven endometrial samples with insufficient sequence reads were excluded. Thus, 39,716,693 reads were obtained from 159 endometrial samples and 126 vaginal samples. Seven endometrial samples with <1,000 reads were excluded. The average read count per endometrial sample was 66,762 (range 192–422,265) and per vaginal sample was 185,311 (range 203–611,776). After quality filtering and OTU clustering, the average read counts were 11,742 (range 551–42,236) and 38,368 (range 36–46,155) for the endometrial and vaginal samples, respectively. Six endometrial samples and two vaginal samples with <1,000 OTU hit reads were excluded. Low-abundance taxa (0.01%) were filtered from the OTU tables (Figure 1). Bacterial taxa in a blank control were assumed to be contaminants from various reagents; therefore, blank-characteristic taxa were subtracted to reduce background noise as in previous studies [11, 23]. Fourteen bacterial taxa detected in a blank control and known to be reagent contaminants were excluded using QIIME: Acidovorax, Acinetobacter, Chryseobacterium, Citrobacter, Escherichia, Flavobacterium, Janthinobacterium, Leptothrix, Methylobacterium, Pseudomonas, Rhodococcus, Sphingomonas, Stenotrophomonas, and Yersinia (Table 1).
Statistical analysis
We calculated the Shannon diversity index and Chao1 richness, which became the index of the microbiota’s α-diversity, then conducted t-tests. We calculated the weighted UniFrac distance for analyzing the β-diversity of the microbiotas between the samples and conducted PERMANOVA tests. The tests were analyzed using QIIME 1.9.1. We performed Welch’s t-tests using R 3.4.3 (https://www.r-project.org/) to compare the bacterial abundances between groups. Hierarchical analysis was performed using R 3.4.3. Distances based on the squared Euclidean distance were calculated and clustered via Ward’s method.
Table 1
Microbial genera detected in blank controls.
| Manual protocol batch 1 (n = 10) | Manual protocol batch 2 (n = 14) |
Taxonomy | Mean abundance (%) | Detection rate (%) | Mean abundance (%) | Detection rate (%) |
Pseudomonas | 20.7 | 10/10 (100) | 18.9 | 14/14 (100) |
Escherichia | 41.3 | 10/10 (100) | 17.6 | 14/14 (100) |
Rhodococcus | 2.5 | 8/10 (80) | 0.1 | 1/14 (7.1) |
Janthinobacterium | 2.8 | 10/10 (100) | 7.4 | 13/14 (92.9) |
Sphingomonas | 4.4 | 10/10 (100) | 0.6 | 7/14 (50) |
Flavobacterium | 3.6 | 9/10 (90) | 0.8 | 12/14 (85.7) |
Methylobacterium | 1.8 | 1/10 (10) | 0.4 | 2/14 (14.3) |
Stenotrophomonas | 2.5 | 8/10 (80) | 0.6 | 9/14 (64.3) |
Acinetobacter | 1.9 | 3/10 (30) | 0.6 | 10/14 (71.4) |
Leptothrix | 0.0 | 0/10 (0) | 1.1 | 9/14 (64.3) |
Acidovorax | 0.0 | 0/10 (0) | 0.1 | 3/14 (21.4) |
Chryseobacterium | 0.0 | 0 /10 (0) | 1.1 | 12/14 (85.7) |
Citrobacter | 0.0 | 0 /10 (0) | 0.4 | 9/14 (64.3) |
Yersinia | 0.0 | 0 /10 (0) | 48.0 | 14/14 (100) |
The composition of the bacteria detected from the blank controls varied with each lot of reagents, and bacteria detected from multiple lots were used as background bacteria. |