This intervention study invited pregnant women who received antenatal care at the Women and Children’s Hospital, School of Medicine, Xiamen University, from June 1^st, 2017 to August 15^th, 2017. The inclusion criteria were briefly listed as follows: 1) gestational age between 37 and 42 weeks; 2) pregnant women who had regular prenatal visits and for whom all clinical data could be obtained; and 3) newborns who were full-term deliveries. Pregnant women with the following complications were excluded: 1) infectious diseases caused by bacteria, viruses, or parasites; 2) inflammatory diseases (e.g., ankylosing spondylitis); 3) metabolic diseases such as diabetes mellitus; 4) pregnancy-associated illnesses including preeclampsia and gestational hypertension; 5) reproductive system disorders (e.g., an ovarian cyst); 6) abnormal pregnancy state (e.g., preterm birth); 7) genetic diseases such as thalassemia; and 8) tumors, including pituitary adenomas and uterine fibroids. The specific analytical protocol of this study is shown in Figure S1 . The basic information of the pregnant were collected through the last prenatal health checkup and lifestyle questionnaires.

The protocol of our present research was approved by the Medical Ethics Committee of the Women and Children’s Hospital, School of Medicine, Xiamen University (KY-2018-020). All procedures were conducted in accordance with the Declaration of Helsinki. All patients were required to provide written informed consent prior to participation. No adverse events were reported for any of the newborns in this study.

Pregnant women scheduled to have a CS surgery were administrated with prophylactic antibiotics 15 to 60 minutes before skin incision and divided into the two groups according to their willingness to have their newborns swabbed with the vaginal fluids. Before delivery (CS or natural delivery), a 7 × 5 cm four-layered piece of gauze was double-folded and soaked in sterile saline and then inserted into the lower vagina for at least 30 mins before the administration of prophylactic antibiotics, and then removed and kept at room temperature in a sterile collector. The gauze was then divided vertically into two equal parts. One part of the gauze was temporally stored in a 50 mL microcentrifuge tube for subsequent analyses of microbiome, while the other part was used to treat CS newborns (n = 16) as soon as they were born. The principle of ‘center-to-periphery’ was strictly followed during the swabbing procedure, which began on the lips, followed by the face, the thorax, the arms, the legs, and finally the back; the complete process took approximately 30 s for each newborn. Six vaginal-born and four CS-born babies were swabbed with gauze containing sterile saline solution and were used as positive or negative references. All treatments were performed in the delivery room. Transitional stool was first collected by sterile diapers between 36 h and 72 h after birth, and then 1-2 g of the sample was transferred into a 50 mL microcentrifuge tube. Approximately 1-2 g of fresh maternal faeces was collected into a 50 mL microcentrifuge tube at the last excretion before delivery. All samples were placed into containers of ice and transported to the laboratory within 1 h after collection and stored at -80°C for long-term storage until further processing.

Microbial DNA in all samples was extracted utilizing the MoBio Powersoil DNA Isolation Kit (QIAGEN, German) according to the manufacturer’s instructions. A NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies, USA) was used to determine the DNA concentration. The extracted DNA was stored at -20°C until further analysis. The hypervariable V3-V4 regions of the 16S rDNA gene were amplified with the specific primers 341F (5’-CCTACGGGNGGCWGCAG-3’) and 806R (5’-GGACTACHVGGGTATCTAAT-3’). PCRs were performed in triplicate in a 50 μL mixture containing 5 μL of KOD buffer (10 ×), 5 μL of dNTPs (2 mM), 3 μL of MgSO₄ (25 mM), 1.5 μL of 341F/806R primer (10 μM), 1 μL of KOD Polymerase, and 100 ng of template DNA. The PCR reagents were purchased from TOYOBO, Japan, and the PCR protocol was carried out for 30 cycles using the following parameters: 94°C pre-denaturation for 2 min, 98°C denaturation for 10 s, 50°C annealing for 30 s, 68°C annealing for 30 s, and hold at 4°C. After amplification, all of the PCR products were pooled, purified, and quantified using the AxyPrep DNA Gel Extraction Kit (Axygen Biosciences, USA) and the ABI Step-One-Plus Real-Time PCR System (Life Technologies, USA) according to the standard protocols.

The microbiome profiles were analyzed by 16S rRNA gene amplicon sequencing with the Illumina MiSeq 250PE platform (Illumina, San Diego, CA, USA). USEARCH software (version 8.1.1861) was applied to turn paired-end sequencing reads into merged, denoised, chimera-free, inferred sample sequences, and the sequence processing steps are described below in detail. Raw sequencing reads were filtered using FASTP software (version 0.18.0) to remove adapters and low-quality reads that would affect the subsequent assembly and analysis to obtain clean high-quality paired-end reads. Paired-end reads were merged as raw tags using FLASH (version 1.2.11), with a minimum overlap of 10 bp and mismatch error rates of 20% (Magoč and Salzberg, 2011), and further merged as raw amplicon sequence variants (ASVs) with a minimum overlap of 10 bp. Noisy sequences of raw tags were filtered under specific filtering conditions to obtain clean tags. The high-quality clean tags were clustered into OTUs of ≥ 97% similarity using the UPARSE (version 9.2.64) pipeline. All chimeric tags were removed using the UCHIME algorithm, and effective tags were finally obtained for the next analysis step. The tag sequence with the highest abundance was selected as the representative sequence within each cluster. The representative OTU sequences were classified into organisms by a naive Bayesian model using the RDP classifier (version 2.2) (Wang et al., 2007) based on the SILVA database (version 132) (Pruesse et al., 2007).

Because the number of subjects was less than 50, the Shapiro-Wilk test and Levene’s test were performed to assess the normality of distributions and the homogeneity of variance, respectively. One-way analysis of variance (ANOVA) was performed to compare normally distributed continuous variables, while the Kruskal-Wallis H test was conducted to compare unevenly distributed variables. Comparisons of Alpha and Beta diversity between any two groups were performed by utilizing Welch’s t test, whereas Kruskal-Wallis H test was used for the comparison among groups. Statistical analysis of the clinical data was achieved by Statistic Package for Social Science software (SPSS) (SPSS Inc., USA) (version 26.0). P < 0.05 was considered to be statistically significant.

Alpha-diversity was assessed by the observed species (Sobs) index, inverse Simpson index, Shannon index, and Pielou evenness index, while beta-diversity was calculated by the Bray-Curtis distance and illustrated by NMDS analysis. Alpha- and beta-diversity, as well as the NMDS were generated in the R project Vegan package (version 2.5.3) (Dixon, 2003) and plotted in the R project ggplot2 package (version 2.2.1) (Wickham et al., 2016). Circular layout representations of species abundance were graphed using circos (version 0.69-3) (Krzywinski et al., 2009). Between groups, Venn analysis was performed in the R project VennDiagram package (version 1.6.16) (Chen and Boutros, 2011). A ternary plot of taxa abundance was plotted using the R ggtern package (version 3.1.0) (Hamilton and Ferry, 2018).

A total of 111 mother-newborn dyads were initially recruited. As shown in Figure S2 , we excluded 85 pairs due to the lack of integrated bio-samples (n = 69), health check-up data (n = 9), and self-reported diseases information (n = 7). Finally, 26 mother-newborn dyads were ultimately included according to our inclusion and exclusion criteria, and they provided 78 samples for the subsequent analyses. Missing values of age (n = 2), enrolment age (n = 1), haemoglobin (n = 1), alanine aminotransferase (ALT) (n = 1), aspartate aminotransferase (AST) (n = 1), serum total bilirubin (STB) (n = 1), and serum conjugated bilirubin (SCB) (n = 1) were interpolated using the multiple imputation method. The basic information of the pregnant women is shown in Table 1 . The median maternal age was 26.50 (25.00-27.25) years, and the average body mass index (BMI) was 20.82 ± 2.45. No participants in the present research subjected active or passive smoking. Other demographic and clinical indicators among groups were roughly similar except for spouse age (SA) and AST (ANOVA, P _SA=0.033; Kruskal-Wallis H test, P _AST=0.022). Despite a significant difference in AST, other maternal biological factors were well homogenized among the different groups. In combination with the clinical indices, participants in this study were overall of a relatively good health status, and the maternal clinical statuses were comparable.

The diversity of the microbiota in a given habitat reflects the composition and relative abundance of the community. Approximately 460 bp of PCR products were generated by amplifying the V3-V4 region of the 16S rRNA gene to compare the bacterial diversity among samples. DNA sequencing after quality filtering yielded 8.29 million paired-end reads, which further merged into 7.21 million tags, with a minimum of 64253 tags per sample (average of 92448 ± 8822), and ultimately formed 27801 operational taxonomic units (OTUs).

Rarefaction curves evaluated the OTU richness and represented whether a reasonable sampling size (sequencing depth) was used. As shown in Figure S3 , the almost horizontal asymptotic curves indicated that the sequencing depth was sufficient for our research. The values of observed species (Sobs), Pielou evenness, Shannon, and inverse Simpson indices were calculated to thoroughly assess alpha-diversity. The Sobs index indicated the actual detected OTUs, while the Pielou evenness index referred to the species equitability within each group. The Simpson and Shannon indices measured the degree of species asynchrony and stability, and higher values represented higher richness, evenness, or both (Luo et al., 2017). As shown in Figure 1A , the highest Sobs value could be observed in the transitional stool samples of naturally delivered newborns (TSN), and the OTUs differed substantially between the transitional stool samples of caesarean-section newborns (TSC) and the TSN, which demonstrated that CS surgery reduced the actual number of species of newborn gut microbes compared with vaginal delivery. In contrast, the TSC group presented the lowest evenness according to the value of the Pielou evenness index ( Figure 1B ) and there was a significant difference between the TSC and TSN groups (Welch’s t test, P _Pielou =0.001). The above analyses revealed that CS surgery led to an obvious alteration in bacterial richness and evenness. Good agreement was exhibited in the Simpson index and Shannon index, which suggested that the gut microbial composition of CS newborns was less diverse (Welch’s t test, P _Simpson =0.017, P _Shannon <0.001), while a lesser degree of dominant bacteria and distribution homogeneity were observed in CS newborns with swabbing vaginal fluid intervention (I) ( Figures 1C, D ). The aggregate analyses of alpha-diversity indicated that the neonatal gut microbiota was affected by both the mode of delivery and the swabbing treatment. Beyond that, pregnant women who underwent different labour modes showed no significant differences in intestinal (Welch’s t test, P _Sobs =0.169, P _Pielou =0.207, P _Simpson =0.081, P _Shannon =0.186) and vaginal fluid microbiota (Welch’s t test, P _Sobs =0.152, P _Pielou =0.188, P _Simpson =0.246, P _Shannon =0.178).

We further investigated beta-diversity according to the Bray-Curtis distance to compare the microbial community structures among groups, as this method provided a model to describe the overall pattern of community composition based on OTUs. Interestingly, some results seem to be inconsistent with those obtained from alpha-diversity analyses ( Figure 2A ). Briefly, significant differences could be observed between the neonates of CS and I group (Adonis test, P= 0.001), as well as the CS and vaginally delivered newborns (Adonis test, P = 0.011). The gut microbiota of CS with vaginal seedings and vaginally delivered newborns seemed to show no substantial difference in terms of beta-diversity. The above-mentioned results hinted the microbial composition of neonates was highly dependent on the mode of delivery. Besides, the microbiome of vaginal fluids and stools samples were different between pregnant women who underwent different labour modes (Adonis test, P _V-VN =0.034, P _F-FN =0.047). Further, the visualization of principal coordinate analysis (PcoA) and non-metric multidimensional scaling (NMDS) analysis based on the Bray-Curtis distance all displayed clear ordinations that indicate the gut microbes of the CS newborns with swabbing treatment were more similar with those of the vaginally delivered newborns, rather than CS babies ( Figures 2B, C ). The above-mentioned results concluded that the infant gut microbiota shared more features with maternal vaginal fluids, and the microbial composition of transitional stool samples in CS-born babies with vaginal seedings tended to be more similar to that of TSN samples. The detailed information of statistical analyses was shown in Table 2 .

The analysis at the genus level indicated the microbial community of CS newborns consisted primarily of Bacteroides (12.10%), Lactobacillus (6.38%), and Escherichia-Shigella (6.03%) after the vaginal seedings intervention ( Table 3 ; Figures 3 ,  4 ), which was more closed to vaginally delivered babies. Among them, the genus of Bacteroides tended to be similar to those of vaginally delivered newborns (14.95%), which was significantly higher than CS-born neonates (0.15%). Statistical analysis further indicated that no obvious restoration pattern existed at the genus level, although the genus of Faecalibacterium, Enterobacter, and Akkermansia presented such trends ( Table 3 ).

The VENN plots confirmed the above results from another aspect, which more shared taxa were detected between vaginally delivered and CS newborns in the intervention group, than those with CS newborns ( Figure 5 ).

The main strength is that our present study is the first research to our best knowledge focusing on the microbial transfer and the intervention programs for CS-born neonates using the samples of transitional stools among Chinese population. The other strength is that the refined processes of sampling and experimental manipulations made the results more reliable and accurate.

There were several main limitations in our research. First, due to the difficulty in sampling, the number of subjects in CS group was relatively small, potentially leading to misinterpretation. Second, this study was based on cross-sectional data, the dynamic changes of microbiome could not be observed, which limited the generalizability of this research.