The study participants were women from Dalian, a northern coastal city in China. Participants were selected from women who underwent routine prenatal examination at the Dalian Women’s and Children’s Medical Center between 2017 and 2019. The mothers were enrolled at the hospital admission for delivery and provided written informed consent for participation. The study protocol was approved by the Dalian Women’s and Children’s Medical Center (approval number 20160021). All operations and evaluations were performed by the same doctor. In total, 19 women had conceived after IVF after a long protocol of conventional solutions of fresh-cycle IVF and conventional luteal support after 1 week; 6 of these delivered before term, whereas 13 delivered at term. Six pregnant women had visited the obstetrics and gynecology clinic for a routine pregnancy test during the same period; three of these women delivered before term, whereas three delivered at term.

The inclusion criteria for this study were as follows: (1) No symptoms of vaginitis, single live births, and primipara; (2) abstinence from sex for 2 weeks, no antibiotic use, no medication history related to the vulva and vagina for 1 month prior to the study, and no history of systemic use of hormone drugs and immunosuppressants; (3) no history of pregnancy complications such as vaginitis, gestational hypertension disease, gestational diabetes mellitus, or pregnancy with abnormal thyroid function; and (4) no history of smoking, drinking, or drug consumption.

Vaginal swab specimens were collected throughout the prenatal check-ups of the healthy participants in all three trimesters of pregnancy. The swabs were obtained at routine prenatal visits, and by rotating a sterilized swab five times along the vaginal lumen in a circular motion. Speculum was not used. No occurrences of premature rupture of membranes was observed, and no participant was administered intravenous antibiotics during delivery. The participants were divided into those who conceived after IVF and those who conceived naturally (controls), and according to the mode of delivery, as full term (≥37 weeks) and preterm (<37 weeks). The collected swabs were placed into sterile tubes by trained research staff and stored at −80°C for DNA extraction.

The vaginal swabs were collected at least in triplicate at three time points: first trimester (F, 10–13 + 6 weeks), second trimester (S, 20–27 + 6 weeks), and third trimester (T, 28–33 + 6 weeks) of pregnancy. The presence of specific vaginal bacterial communities in any single gestational period in participants who delivered before term (P, six participants in IVF pregnancy group, three participants in natural pregnancy group) was compared with that in participants who delivered at term (T, 13 participants in IVF pregnancy group, 3 participants in natural pregnancy group).

The standard long IVF regimen was used. A gonadotropin-releasing hormone (GnRH) agonist regimen was the main protocol used in this study, and all patients underwent embryo transfer in fresh cycles. Controlled ovarian hyper-stimulation, oocyte retrieval, and embryo transfer were carried out. Participants were treated with downregulation from the mid-luteal phase of the previous cycle. When the pituitary reached desensitization, recombinant FSH was begun at 150–225 IU/day. Human chorionic gonadotropin (hCG) was given (4000–10,000 IU) once two or more follicles had reached a size of 18 mm. Oocytes were extracted 34–36 h after the hCG trigger, and this was followed by intracytoplasmic sperm injection (ICSI). Participants were injected progesterone (lot NO. 1220507 Shanghai General Pharma, China) at 40 mg/day 48 h after oocytes fertilization. The progesterone supplementation was continued until 10 weeks of gestation after pregnancy was achieved.

Genomic DNA was extracted using cetyltrimethylammonium bromide; the concentration and purity of DNA were detected by 1% agarose gel electrophoresis. The appropriate amount of sample was placed into the tube and diluted with sterile water to 1 ng/μL. High-throughput sequencing technology was used to sequence the V3–V4 region of the 16S rRNA gene, which was amplified used a specific primer with the barcode 314F-806R (V3V4 primers: 314F-5′ CCTAYGGGRBGCASCAG 806R5′ GGACTACNNGGGTATCTAAT) Phusion^® High-Fidelity PCR Master Mix (New England Biolabs) was used for PCR amplification according to the selection of the sequencing region.

PCR products were mixed with the same volume of 1× loading buffer and detected on 2% agarose gel electrophoresis. Sample strips of 400–450 bp were chosen for further experiments. The PCR products were purified using Qiagen Gel Extraction Kit (Qiagen, Hilden, Germany).

TruSeq^® DNA PCR-Free Sample Preparation Kit (Illumina, San Diego, CA, USA) was used to generate sequencing libraries, and their quality was analyzed using the Qubit@ 2.0 Fluorometer (Life Technologies, Thermo Fisher Scientific, Waltham, MA, USA) and Agilent Bioanalyzer 2100 system. The library was sequenced on Illumina HiSeq2500 platform, and finally, 250-bp paired-end reads were generated.

The reads were merged by FLASH to obtain raw tags (V1.2.7) (25), and quality filtering was performed to obtain high-quality clean tags (26) by QIIME (V1.7.0) (27). The tags were compared with the Gold database, and finally, the effective tags were obtained using UCHIME algorithm (28, 29). Uparse software (Uparse v7.0.1001) was used to analyze the sequences (30), which were assigned to the same operational taxonomic units (OTUs) once the similarities were more than 97%. The Silva Database (31) was used based on the Ribosomal Databases Project classifier (Version 2.2) algorithm to annotate taxonomic information (32). MUSCLE software (Version 3.8.31) was used to align multiple sequences, to further analyze the phylogenetic relationships of OTUs, and to determine the dominant species in the different groups (33). Based on further information about the abundance of normalized OTUs, we subsequently analyzed their alpha/beta diversity.

Alpha diversity was exhibited by six indices, namely Observed-species, Chao1, Shannon, Simpson, ACE, and Good’s Coverage, which were calculated with QIIME (Version 1.7.0) and displayed using R software (Version 2.15.3). Beta diversity was calculated by QIIME software (Version 1.7.0) and R software (Version 2.15.3). Species analysis with significant differences between groups was performed using Student’s t-test and mapping between groups was done using R software (Version 2.15.3).

We Uploaded the raw 16S rRNA gene sequencing data to the National Center for Biotechnology Information (accession no. PRJNA728871).

The details of the participants are stated in Table 1 . The average age of participants who conceived via IVF(A)was 31.42 ± 10.23 years, and that of those who conceived naturally (B) was 31.50 ± 4.52 years (P > 0.05). The mean BMI of the participants in Group A was 26.57 ± 2.03 kg/m² and that of participants in Group B was 27.48 ± 0.95 kg/m² (P > 0.05); The mean fetal weight in Group A was 3157.89 ± 628.55 g and that in Group B was 2991.67 ± 452.12 g (P > 0.05).

The details of the specimens collected from the participants who conceived via IVF (Group A, 19 participants) and those who conceived naturally (Group B, 6 participants) are stated in Table 2 . The participants were further divided into women who delivered before term (P) and those who delivered at term (T).

16S rRNA of each sample was analyzed using high-throughput sequencing technology. The rarefaction curves tended to approach saturation in all samples except for sample BFT (women who conceived naturally (B), first trimester (F), and delivered at term (T))( Figure 1 ); Good’s Coverage revealed that 99%–100% of the species were detected in all samples, suggesting that the data were suitable for further analyses. The details are shown in Table 3 .

Among women who conceived via IVF (A) or naturally (B), the number of species and diversities in the bacterial community in those who delivered at term (T) were significantly higher than those in women who delivered before term (P) throughout pregnancy (F, first trimester; S, second trimester; T, third trimester), regardless of the pregnancy pattern. The numbers of the observed species were as follows: AFT > AFP, AST > ASP, ATT > ATP, BFT > BFP, BST > BSP, and BTT > BTP ( Figure 2 ), indicating that preterm delivery can be predicted by the diversity of vaginal bacterial communities, regardless of whether the pregnancy is natural or via IVF. In addition, the bacterial communities could be more easily detected in early gestation than in the next two trimesters; the number of bacteria sharply declined during late gestation, and AFT > AST > ATT, AFP > ASP > ATP, BFT > BST > BTT, BFP > BSP > BTP ( Figure 2 ).

Among women who conceived naturally as well as in those who conceived after IVF, and in those who delivered at term (T) and before term (P), the vaginal bacterial communities in the first trimester were higher than those in the other two trimesters. This indicated that the first trimester was a critical period during pregnancy, and the diversity of vaginal bacterial communities decreased with increasing gestational age in all samples ( Figure 2 ).

The results indicated that conception after IVF could be associated with significantly decreased number of VMB in the first trimester, and that reduced microbial diversity in the vagina could be associated with an increased risk of preterm birth.

At the genus level, Lactobacillus was dominant in all the samples (>93%), which was consistent with previous reports (34, 35). However, we found some differences among these samples ( Figure 3 ) with regard to bacterial diversity. In the group of participants who conceived after IVF, the vaginal bacterial diversity in those who delivered before term was lower in the first and second trimesters, compared with participants who delivered at term; however, there were no differences in the bacterial diversity in the third trimester between the groups. In the group of participants who conceived naturally, the vaginal bacterial diversity in those who delivered before term was lower in the first trimester, compared with participants who delivered at term; however, there were no differences in vaginal bacterial diversity in the second and third trimesters between the groups.

In all women who conceived naturally, regardless of whether they delivered at term or before term, bacterial diversity was obvious in the first trimester. Thus, detection of vaginal bacteria might be helpful to monitor the risk of preterm birth in pregnant women. We found that among the top 30 genera, Lactobacillus was dominant in all samples, Alloscardovia was found only in participants who conceived after IVF, no Bacillus was found in the third trimester in all samples, Vibrio appeared only in AFT, Sulfurospirillum and Propionispira were detected only in BFT, Sporosarcina was detected only in AFT and AST, Lactococcus was found only in BFT and BFP, and Oceanobacillus was found only in BFP. In women who conceive after IVF, Vibrio and Sporosarcina might act as negative biomarkers for the risk of preterm delivery. As per our research, in women who conceive naturally, Oceanobacillus can act as a positive biomarker, and Sulfurospirillum and Propionispira can act as negative biomarkers for the risk of preterm delivery.

We further analyzed the differences of vaginal bacterial diversities among the different groups: between participants who conceived after IVF (A) and those who conceived naturally (B), and between women who delivered before term (P) and those who delivered at term (T) ( Figure 4 ). We observed that group BT had the highest number of observed species, and although groups BP and AT had a similar number of species, group AT had higher diversity than group BP did, and group AP had the lowest number of observed species and diversity among the four groups, indicating that bacterial diversity might play an important role in a healthy pregnancy. Firmicutes was dominant (98.501%) in all the samples, followed by Proteobacteria (0.133%), Actinobacteria (0.052%), and Bacteroidetes (0.050%). Actinobacteria appeared only in groups AT and AP (mostly in group AT), and Sulfurospirillum was found only in group BT. Vibrio was found only in group AT ( Figure 5 ). The diversities and numbers of microorganisms were important for further outcome of pregnancy, especially for patients who conceived after IVF. Hence, standard protocols should be established to support a shift of vaginal microbiota during IVF therapy.

Student’s t-test was performed to determine the significant difference between the bacterial species in all the groups (P < 0.05), and the results showed significant differences between groups AP and AT. Clostridia were a significant differential species and were dominant in group AT; between groups BP and BT, unidentified Actinobacteria were a significant differential species and were dominant in group BT; between groups AT and BT, Lactobacillus equicursoris was a significant differential species and was dominant in group AT ( Figure 6 ). Thus, Clostridia and Lactobacillus equicursoris can act as positive biomarkers for preterm birth in patients who conceive after IVF.

Between groups P and T, 35 bacterial species were found only in group P ( Figure 7 ), and the annotation results indicated that they were novel species and may act as biomarkers for the risk of preterm birth.