Healthy pregnant women and those with threatened preterm labor were recruited in the Department of Obstetrics, The First Affiliated Hospital of Nanchang University, Jiangxi Province, China. Ethical approval of the study was granted by the Ethics Committee of the First Affiliated Hospital of Nanchang University (2019-061). All participants have been informed of the purpose, background, process, risks, and benefits of the study, and signed the informed consent to participate in this study. The Department of Gynecology and Obstetrics, the First affiliated Hospital of Nanchang University was responsible for the study.

The inclusion criteria of patients: (1) The patients delivered between 28 weeks and 37 weeks of gestation, which were classified into preterm group. (2) The patients at 28 weeks and less than 37 weeks gestation suffered regular or irregular contractions accompanied by progressive dilation or shortening of the cervical canal, which were classified into Sym.preterm group in this study.

The exclusion criteria of patients: (1) Administration of any antibiotic or probiotic treatment one month before sample collection. (2) Diseases that may affect microbiome composition such as thyroid disorders, asthma, lipid metabolic disorders, inflammatory bowel disease, irritable bowel syndrome, and celiac disease. (3) Other obstetric conditions complicating pregnancy, such as gestational hypertension, gestational diabetes, twin or multiple pregnancies, placenta previa. (4) Other chronic diseases, such as chronic hypertension, chronic kidney disease. (5) Termination of pregnancy due to fetal or maternal factors. Fecal samples were collected from all enrolled subjects at the hospital, while samples from patients with threatened preterm labor were taken at symptom initiation. The fecal samples were collected at admission and the average sampling time of preterm group was 25.9 gestational weeks and 26.5 gestational weeks for the control group. All fecal samples were stored at −80°C until further processing.

Bacterial genomic DNA was extracted using a MinkaGene Stool DNA kit (Magigene, Guangdong, China) according to the manufacturer’s instructions. After extraction, the 16S rRNA V4 region was amplified by quantitative real-time PCR with the following barcoded primers (shown from 5′ to 3′): V4F, GTGYCAGCMGCCGCGGTAA and V4R, GGACTACNVGGGTWTCTAAT. Sun and colleagues (Sun et al., 2013) showed that the V4 and V5 regions of bacteria 16S rRNA concentrate the least intragenomic heterogeneity. Further works verified its efficiency by revision to the region primer and avoided its bias to particular taxa like Proteobacteria (Caporaso et al., 2011; Caporaso et al., 2012; Parada et al., 2016; Walters et al., 2016). Therefore, in this study, 16S rRNA V4 region was selected for sequencing. Briefly, amplifications were performed using a step cycling protocol consisting of 98°C for 30 s, 35 cycles at 98°C for 10 s, 54°C for 30 s, and 72°C for 45 s, ending with the final elongation at 72°C for 10 min. PCR products were purified using an AxyPrep PCR Cleanup Kit (Axygen, California, U.S.A.). Sequencing libraries were generated using TruSeq DNA PCR-Free Sample Preparation Kit (Illumina, USA) following the manufacturer’s recommendations, and index codes were added. The library quality was assessed on the Qubit@ 2.0 Fluorometer (Thermo Scientific) and Agilent Bioanalyzer 2100 system. The MiSeq platform (Illumina, 2 × 250 bp paired-end, CA, USA) was employed for the 16S rRNA sequencing.

QIIME2 was used for controlling the sequencing data quality and clustering the 16S rRNA gene reads into Amplicon Sequence Variants [ASVs, based on DADA2 pipeline (Bolyen et al., 2019)], taxonomic assignment (based on the Greengenes Database V.13_8), and performing alpha diversity (Observed ASVs, Shannon Index and Phylogenetic Diversity Whole Tree Index), beta diversity, and PCoA analyses. All samples were rarefied to 3,000 sequences, and three samples were filtered. Permutational multivariate analysis of variance (PERMANOVA) was performed to determine if the microbiota composition differed between groups and generate the explained variation based on distance metrics. Linear discriminant analysis effect size (LEfSe) was performed to identify the bacterial biomarkers between groups (Segata et al., 2011). Distance-based redundancy analysis and variance projection were employed to show the tendency of variables. Oral bacteria were annotated according to the Human Oral Microbiome Database (HOMD, www.homd.org) (Chen et al., 2010). Sequencing reads had been uploaded on European Nucleotide Archive (access number: PRJEB39133).

The significance of differences between the two groups was determined by the Wilcoxon rank-sum test. The Kruskal–Wallis test was used for multi-groups. P values less than or equal to 0.05 were considered significant. The Benjamini and Hochberg method was used to adjust the P-value for multiple hypotheses (Benjamini and Yekutieli, 2001). Correlation analyses were performed based on the Pearson’s product–moment correlation. Statistical analyses and data visualization were performed using R V.3.5.0 (under RStudio V.1.1.453), with the vegan, ggplot2, pheatmap and corrplot packages.

After rarefied to 3,000 sequences, a total of 19 women with threatened preterm labor and 22 healthy pregnant women were included in the present study. In the threatened preterm group, eight patients with preterm symptoms delivered (Sym.PTB) at term, and 11 patients delivered preterm. The preterm-delivered group was further divided into an early week preterm group (seven women, Early PTB) and a late-week preterm group (four women, Late-PTB), depending on whether the condition was detected before 34 weeks of gestation. The clinical parameters are summarized in Table 1 .

The fecal samples of these participants were collected for DNA extraction and 16S rRNA sequencing. By employing QIIME2 bioinformatic software, we showed that alpha-diversity indices, including observed ASVs, Shannon Index, and PD Whole Tree Index, were increased in the preterm group, indicating a higher richness and evenness in preterm patients than in healthy controls. We did not observe significant differences in alpha diversity within the preterm subgroups ( Figure 1A , Supplementary Figures S1A, B ).

Beta diversity was calculated using the Bray–Curtis distance metrics to measure the extent of similarity in fecal microbial communities. According to the principal co-ordinate analysis, the gut microbiome of the preterm group significantly differed from that of the healthy group (PERMANOVA, P = 0.004, R² = 0.051, Figure 1B ), indicating that the structure of the microbiota of preterm patients differed from that of healthy controls. Interestingly, we observed that the preterm subgroups showed a gradually drifting tendency away from the healthy groups with increasing severity (PERMANOVA, P = 0.019, R² = 0.107, Figures 1C, D ), which suggests there are distinct gut microbiome profiles among laboring women who experience term birth, late preterm birth, or early preterm birth.

Subsequently, we analyzed the phylum-level profiles for the gut microbiota between preterm patients and healthy controls, which were fairly similar. The dominant phyla of both healthy and preterm groups were Firmicutes, Proteobacteria, Actinobacteria, Bacteroidetes, Verrucomicrobia, and Fusobacteria ( Figure 1E ).

Taken together, these results demonstrated the presence of gut microbiota dysbiosis in preterm patients compared with healthy controls.

Given that the gut microbiota was significantly different between preterm patients and healthy controls, we next investigated whether the gut microbiota was associated with the clinical parameters of patients.

To measure the extent of clinical parameters associated with the gut microbiome between preterm patients and healthy controls, we employed permutational multivariate analysis of variance to calculate the explained variation in host parameters based on the Bray–Curtis distance metric. Among 28 parameters with less than 20% missing data, five were significantly associated with gut microbial variations between preterm patients and healthy controls, namely, gestational age (recorded in weeks), neonatal weight, and Apgar scores (at 1, 5, and 10 min) ( Supplementary Figure S2A ). For the preterm subgroups, eight parameters were significantly associated with the subgroup microbiome variation, namely, C-reactive protein (CRP), tocolytic therapy, BMI at delivery, maternal age, gestational age, and Apgar scores (at 1, 5, and 10 min) ( Supplementary Figure S2B ). CRP is an inflammatory marker, while other parameters correlate with maternal–fetus prenatal and postpartum status, indicating that gut microbiome profiles are distinguished by gestational age at delivery and other clinical characteristics, many of which are expected to vary with gestational age at delivery.

We used distance-based redundancy analysis (db-RDA) to show the relationship of continuous parameters across samples by using R package vegan based on the Bray–Curtis distance metric. The projection of parameters demonstrated that gestational age (in weeks) and Apgar scores extended along axis-1, which distinguished between preterm patients and healthy controls ( Figures 2A–C ). While in the projection of preterm subgroup samples, CRP, BMI at delivery and maternal age were almost in the opposite direction to gestational age, neonatal weight, and Apgar scores, reflecting disease severity ( Figures 2D–J ). In the subgroup analysis, we included 13 preterm patients without missing value of all the included host parameters to avoid the projection change. Thus, the limited sample size could only provide us suggested indications, and further study on larger population is required. Altogether, these results again indicated that the gut microbiome profiles of preterm patients may be linked with other host parameters known to correlate with gestational age at birth.

To further identify the unique bacterial biomarkers between preterm and healthy groups, we performed LEfSe analysis based on the genus level and identified fifteen genera showing significant differences. Preterm patients exhibited a significant increase in the relative abundance of the genus Fusobacterium, Streptococcus, Neisseria, Haemophilus, Lautropia, Porphyromonas, Clostridium, Prevotella, Rothia, Oscillospira, Granuliccatella, Actinomyces, and Bilophila. On the contrary, genus Coprococcus and Gemmiger were depleted in preterm patients compared to those in healthy controls ( Figure 3A ).

Of note, several genera enriched in preterm groups were commonly considered oral pathogens, for example, Porphyromonas, Streptococcus, and Fusobacterium. As oral symptoms are common during pregnancy, we further examined the changes in the common resident oral bacteria in feces. We clustered and annotated bacteria to resident oral microbes according to the HOMD database. The relative abundance of common oral bacteria was markedly higher in the preterm group than in the healthy group (P < 0.001), as in the three subgroups ( Figure 3B ). We clustered the identified oral bacteria to examine their distribution among all individuals. Based on the heatmap, there was one cluster with thirteen genera enriched in preterm patients. Of these, nine were identified by LEfSe, and three were reported potential pathogens, indicating the possible higher intensity of bacterial oral–gut-translocation in preterm patients ( Figure 3C ). Among those enriched bacteria, Fusobacterium, Porphyromonas, and Streptococcus were common oral opportunistic pathogens.

Collectively, these results demonstrated that the gut microbiota of preterm patients differed from that of healthy individuals. Furthermore, a high proportion of genera enriched in preterm patients coincided with oral opportunistic pathogens.

Given that the gut microbiome profile was associated with the status of preterm patients and potential oral–gut translocated bacteria were enriched in preterm patients, we next investigated the association between potential oral–gut translocated bacteria and host parameters. We used Pearson’s product–moment correlation to evaluate the link between the relative abundance of oral-to-gut bacteria and host parameters. Results showed that oral-to-gut bacteria were negatively correlated with gestational age (in weeks), neonatal weight and Apgar scores ( Figure 4 ). Altogether, the oral bacteria were correlated with clinical parameters reflecting the maternal and fetal status.