香京julia种子在线播放

Healthy, low risk, pregnant individuals delivering at term were recruited into the Maternal Microbiome LEGACY Project from three hospitals across British Columbia (BC), Canada between March 2018-March 2020: BC Women’s Hospital + Health Centre (Vancouver), Surrey Memorial Hospital (Surrey) and University Hospital of Northern BC (Prince George). These centres were chosen with the goal of enrolling a large, sociodemographically diverse, multi-ethnic study population. Participants were enrolled based on the following inclusion criteria: >18 years of age, ≥37 weeks gestational age at delivery, no known major foetal anomalies and singleton or twin gestation. Participants were excluded based on the following criteria: inability to provide informed consent, participation in drug or probiotic trials, triplet or higher order gestation, placenta previa at delivery, placental abruption, and emergency intrapartum complications that preclude time for engagement in research.

Informed written consent for study participation was obtained and ethics approval was granted by the University of British Columbia Children’s and Women’s Health Centre Research Ethics Board, harmonised with partner boards in the Fraser Health Authority and Northern Health Authority (Certificate no. H17-02253). Participants were recruited either during pregnancy or at the time of admission for delivery. Targeted recruitment of individuals with planned caesarean section and planned vaginal delivery ensured appropriate numbers of participants in the groups of interest. Target study population size was calculated to provide the best power to detect an interaction term in a multinomial regression model. Based on a maximum effect size of 0.41 from Human Microbiome Project data and a minimum of 0.2, a sample size of at least 600 would permit detection of effect sizes of at least 0.21 at power = 0.8, alpha = 0.05. Demographic and clinical data were collected by research staff via interview and medical chart review and entered into the Research Electronic Data Capture (REDCap) database securely hosted at BC Children’s Hospital Research Institute (Harris et al., 2019).

Maternal vaginal swabs of the posterior fornix and lateral vaginal wall were collected at first presentation to the labour/delivery assessment area, by a nurse, midwife, or clinician prior to examinations and procedures associated with the labour/delivery/birth process. Neonatal meconium, defined as the first collected stool specimen within 72 hours of birth, and two infant stool samples were collected at follow-up visits at 10 days ( ± 3 days) and 3 months ( ± 2 weeks) postpartum, respectively. Diapers with stool were collected by parents and kept refrigerated in sealed bags until collection by the study team at home visits either the same day or next day. Samples were transported to the lab on ice in a cooler, and stool was scraped from diapers using sterile spatulas in a biosafety cabinet and aliquoted into cryovials for storage at -80°C. Empty cryovials were processed in the same manner as samples to generate mock sample controls for assessment of contamination introduced during the handling process. Samples were shipped from the central collection lab at BC Women’s Health Research Institute to the University of Saskatchewan on dry ice and immediately stored at -80°C until DNA extraction.

On the day of DNA extraction, 1 mL of sterile phosphate-buffered saline (pH 7.4) was added to thawed vaginal swab storage tubes. Following vigorous vortexing, swab fluid was transferred to sterile 1.5 mL tubes; 100 µL was taken as input for DNA extraction and the remainder was stored at -80°C. Stool was removed from cryovials using disposable sterile spatulas in a biosafety cabinet and 200 mg was transferred to sterile 2 mL tubes for DNA extraction. Total genomic DNA was extracted with a MagMax DNA Ultra 2.0 kit (Applied Biosystems, Waltham, MA, USA) on an automated KingFisher Flex platform (ThermoFisher Scientific, Waltham, MA, USA) which employs a combination of mechanical and chemical lysis. Extractions of vaginal and stool samples were performed on different days to limit cross-contamination. Mock samples (empty tubes left open during sample processing in Vancouver) and extraction-negative controls (400 µL sterile molecular biology-grade water) were included with each batch of samples to assess contamination. Kit controls were also generated by separate extraction of kit reagents only.

The universal barcode region of the cpn60 gene was amplified as described previously (Fernando and Hill, 2021). A mixture of 20 plasmids harbouring cpn60 sequences from known vaginal microbiome species (mixed vaginal panel, MVP) mixed in equal proportions was used as a positive control (Dumonceaux et al., 2009). Amplicons were purified and indexed according to the standard Illumina 16S metagenomic sequencing library preparation protocol (Illumina, 2013). Indexed amplicons were quantified using Qubit fluorometry, normalised to 4 nM and pooled. Diluted 10 pM libraries (5% PhiX) were sequenced on an Illumina MiSeq platform using 500-cycle v2 reagent kits (401 R1, 101 R2). Only R1 sequence reads were used for analysis as read-pair overlap is not possible and only 150 bp of the cpn60 barcode region is required for accurate taxonomic identification (Vancuren et al., 2020).

Amplification primers were removed using cutadapt (Martin, 2011) and reads were trimmed to an average Phred score of Q30 over a 4 bp window and minimum length of 150 bp using Trimmomatic (Bolger et al., 2014). Variant calling was performed in the QIIME2 (Bolyen et al., 2019) package using DADA2 (Callahan et al., 2016) to generate 150 bp amplicon sequence variants (Vancuren et al., 2020) (ASVs) and the resulting feature table was converted to tsv format. Taxonomy was assigned by comparison to a non-redundant version of cpnDB (cpndb_nr_20210316) using wateredBLAST, and ASVs with <55% sequence similarity to a cpnDB hit were discarded (Schellenberg et al., 2009; Vancuren and Hill, 2019). ASVs aligning to the same cpnDB entry were collapsed into ‘nearest neighbours’ and their frequencies combined.

Feature tables were screened for potential contaminants in RStudio v.1.4.1717 [R v4.1.1 (R Core Team, 2022)] using the decontam (Davis et al., 2018) package. Any infant stool microbiomes with <1,000 reads were excluded from downstream analysis. Heatmaps were built from proportional abundance data of vaginal and stool microbiome profiles: all taxa were used for hierarchical clustering of samples (Ward linkage) based on Euclidean distance matrices calculated by the vegan (Oksanen et al., 2020) package. Clusters within vaginal, 10-day stool and 3-month stool microbiome profile datasets were defined algorithmically by silhouette indices using the NbClust (Charrad et al., 2014) package, and were bootstrapped to assess cluster support using the fpc (Hennig, 2007) package. Independence of maternal and infant microbiome clusters was examined using correspondence analysis (implemented by the FactoMineR (Lê et al., 2008) package) and Fisher’s exact test. Feature tables containing read counts for vaginal and stool samples underwent centre log-ratio transformation using the ALDEx2 (Fernandes et al., 2013) package prior to principal component analysis (PCA) and assessment of significant compositional differences using pairwise PERMANOVA (Anderson and Walsh, 2013). The aldex wrapper function was used to define differentially abundant taxa, while multivariable association analysis was conducted using the MaAsLin2 (Mallick et al., 2021) package. The Shannon-Weaver diversity index (H’) was calculated for 10-day and 3-month stool samples rarefied to 1,000 reads using the vegan (Oksanen et al., 2020) package, with differences between intrapartum antibiotic exposure and delivery mode groups investigated using Mann-Whitney U and Kruskal-Wallis tests in GraphPad Prism v9.3.1 Significant differences in the proportional abundance of differentially abundant infant stool taxa between antibiotic exposure groups were also investigated by Mann-Whitney U tests in GraphPad Prism v9.3.1, correcting for multiple testing by controlling the false discovery rate at <1%.

All sequence data associated with this study were deposited in the NCBI Sequence Read Archive (BioProject PRJNA824125). Annotated R code describing reproducible analysis of sequencing data, processed feature tables containing nearest neighbour read counts, and microbiome cluster metadata are available to download from the online version of this manuscript ( Supplementary Files S1–S3 ).

We enrolled 623 participants and their infants into the LEGACY study: 247 (39.6%) participants delivered vaginally, 221 (35.5%) had an elective C/S and 155 (24.9%) had an emergency C/S ( Supplementary Table S1 ), proportionally determined by our recruitment strategy. Mean age at delivery was 34.6 years, which is representative of our study hospitals’ delivery populations. Most participants self-identified as White (54.7%), 21.2% as Asian, 8.3% as South Asian, with lower percentages of other ethnicities. Twenty-one percent (21.3%) of participants tested positive for group B Streptococcus (GBS) on screening; 10.4% of participants had hypertension during pregnancy, and 16.5% were diagnosed with gestational diabetes (half of whom were treated with insulin). All participants undergoing elective C/S received intravenous antibiotic prophylaxis- usually cefazolin- approximately 1 hour prior to the procedure so that therapeutic concentrations are present in tissues at the time of skin incision (van Schalkwyk and Van Eyk, 2017) and those with GBS colonisation who were planning a vaginal delivery received intrapartum prophylaxis. The rate of prolonged duration of ruptured membranes (‗18 hours) was comparable between the vaginal and emergency C/S groups (9.3 vs 9.0%, respectively).

We profiled the microbiomes of 623 maternal vaginal swabs taken on admission to labour and delivery, 142 meconium samples, 581 stool samples from 10-day-old infants and 462 stool samples from 3-month-old infants. Data for LEGACY meconium samples have been described elsewhere, demonstrating lack of a detectable microbiome signature that could be differentiated from background signal and no evidence of an in-utero stool microbiome (Dos Santos et al., 2021). Following quality control, 71,826,602 reads were retained from 621 vaginal microbiomes, 570 10-day stool microbiomes, 460 3-month stool microbiomes and 176 controls. Complete microbiome data was available for 442 dyads (i.e., vaginal and both stool microbiome profiles). In line with recent recommendations for microbiome studies (Eisenhofer et al., 2019), a complete list of all taxa found in negative controls is shown in Supplementary File S3 . Cross-contamination of negative controls with common vaginal and stool organisms was prevalent, with species such as Lactobacillus crispatus and Escherichia coli detected in many controls ( Supplementary File S3 ). Exogenous contaminants were also apparent: Pseudomonas tolaasii, an environmental species, was detected in 631 samples and 89 controls and was particularly associated with certain sequencing batches. Moreover, decontam identified this species as a potential contaminant: approximately 20% of all P. tolaasii reads were detected in negative controls- the highest proportion of any taxon in the dataset. Accordingly, all reads aligning to this species were removed from the dataset prior to further analyses. Nineteen of twenty positive control taxa were detected across all sequencing runs, the exception being Streptococcus macedonius, which was not present in the plasmid library in the second batch of positive controls used in the study. With this exception, positive control composition was consistent across sequencing batches, indicating the consistency of PCR amplification.

Vaginal microbiomes were dominated by either one of several species of Lactobacillus, a single anaerobic species commonly associated with the vaginal environment, or were composed of a mixture of various anaerobes ( Supplementary Figure S1 ). The four Lactobacillus-dominated community state types (CSTs) commonly reported in studies of the vaginal microbiome were well represented. In accordance with other studies of pregnant individuals (Albert et al., 2015; Freitas et al., 2017; Freitas et al., 2018; Wells et al., 2020), profiles dominated by L. crispatus (CST I) were the single most abundant vaginal cluster (n = 263), followed closely by those dominated by various or mixed obligate anaerobes (CST IV; n = 241). Vaginal microbiomes dominated by L. gasseri (CST II; n = 37), L. iners (CST III; n = 45) and L. jensenii (CST V; n = 37) were markedly less common. Hierarchical clustering identified a total of 16 ‘cluster types’: the four commonly reported Lactobacillus CSTs above, plus a further 12 that divided CST IV into 12 separate clusters. Eleven of these represent a single dominant species (such as Megasphaera, Gardnerella and Bifidobacterium spp.), while the twelfth was comprised of vaginal microbiomes with mixed anaerobic populations. The distribution of maternal vaginal community state types was not significantly different between women undergoing rupture of membranes before or after sample collection (P = 0.238, Fisher’s exact test).

Among 10-day-old infants, diversity within stool microbiomes was remarkably low and most stool microbiomes were dominated by a single species (‗80% relative abundance). Enteric taxa often reported in gut microbiome studies were commonly detected, such as E. coli and various members of the genera Bifidobacterium, Enterococcus, Klebsiella, Bacteroides and Streptococcus (Yatsunenko et al., 2012; Shao et al., 2019) ( Figure 1A ). Stool samples clustered into 25 separate clusters, with 23 defined by a single dominant species, one defined by a mixture of E. coli, Bifidobacterium breve and Parabacteroides distasonis, and a final cluster of microbiomes typically composed of multiple species. The latter also included small sub-clusters of several samples where a single species was dominant (e.g., Bifidobacterium pseudocatenulatum or Enterobacter kobei).

At 3 months, the number of microbiome clusters defined decreased to 14, with 12 defined by a single species comprising at least half of the microbiome by proportion ( Figure 1B ). A further cluster was defined by a combination of E. coli and several species of Bifidobacterium, with a final cluster of mixed microbiome profiles, containing several small sub-clusters dominated by various species. Infant stool microbiomes also exhibited a significant increase in alpha diversity (Shannon-Weaver diversity index, Figure 2A ; P <0.0001), with far more samples composed of multiple species found together at more even relative abundances.

At both 10 days and 3 months of life, clusters dominated by E. coli represented the largest number of samples of any single cluster (n = 116 (20.4%) and 121 (26.3%), respectively). When considering all clusters for which E. coli was a dominant species, this increased to 138 (24.3%) and 213 (46.4%), respectively. Bifidobacterium spp.– widely recognised as keystone infant gut species (Gotoh et al., 2019)- and various species of Klebsiella were also prevalent in infant gut microbiome profiles at both timepoints. Clusters defined by these genera accounted for 111 (19.5%) and 117 (20.5%) samples from 10-day-old infants, and 168 (36.5%) and 74 (16.0%) samples from 3-month-old infants, respectively. Relative abundances of E. coli and Bifidobacterium spp. were significantly higher at 3 months of life ( Supplementary Figure 2B left & middle; P <0.0001 and <0.001 respectively), while no difference was observed for Klebsiella spp. ( Supplementary Figure S2B right, P = 0.942).

We next assessed if the maternal vaginal microbiome composition could explain differences in infant stool microbiome composition between birth modes. PCA of infant stool microbiomes showed no significant clustering of microbiome profiles at 10 days or 3 months by maternal CST ( Figures 2A, B , P ‗0.13 or ‗0.24 for all comparisons, respectively). Correspondence analysis also demonstrated no evident co-clustering of maternal and infant clusters at either timepoint ( Figures 2C–E and Supplementary Figure S3 ), and no differences were observed in the distribution of maternal vaginal microbiome clusters among infant stool microbiome clusters regardless of delivery mode ( Supplementary Figure S3D , P >0.22).To investigate this further, we assessed the transition of 442 mother-infant dyads with complete profiling data between microbiome clusters over time. Regardless of birth mode, no evident patterns in the flow of microbiomes from maternal to infant clusters were discernible ( Figure 3 ), with most infants in almost all 10-day clusters originating from mothers whose vaginal microbiome clusters were dominated by L. crispatus or various anaerobes (i.e., the two largest maternal clusters). Exceptions to this were exclusively seen in infant stool clusters comprised of a few or even single samples. Conversely, maternal vaginal clusters were distributed among stool clusters according to their frequency in the overall maternal population, indicating that the maternal vaginal microbiome holds no predictive value for infant stool microbiome composition. Lastly, we evaluated the stability of infant stool microbiomes over time and noted that over 75% of infants belonged to a cluster with a different ‘dominant’ species at 3 months compared to the 10-day sample, reflecting the dynamic nature of the gut microbiome in early life ( Supplementary Table S2 ). Of the seven pairs of twins enrolled in the LEGACY study, only two pairs shared the same stool microbiome cluster at 10 days of life. At 3 months, four of five pairs of twins (two lost to follow-up) shared the same microbiome cluster.

To confirm the lack of effect of maternal vaginal microbiome composition, we searched for taxa present in infant stool microbiomes at both timepoints which were significantly associated with each of the five canonical maternal CSTs. After correcting for multiple testing, MaAsLin2 identified two taxa- Lactobacillus gasseri and Actinomyces vaccimaxillae- which were significantly associated with infants born to mothers from CST II (L. gasseri-dominant vaginal microbiomes) at 10 days of life ( Supplementary File S3 ). However, this association did not hold when comparing the relative abundances of these taxa across stool microbiomes stratified by maternal CST ( Supplementary Figure S4 , P >0.137, all comparisons). On the contrary, this is likely the result of differences in the distribution of samples where the abundance of these two taxa is extremely low, but not zero, among CST groups. In 3-month-old infants, a single taxon, Hungatella hathewayi, was significantly associated with maternal CST II, though only 267 reads of this taxon were detected across 10 samples in the entire dataset, making this observation unreliable.

Given the assumption that maternal microbiome exposure drives differences in infant stool microbiomes between vaginally and C/S-delivered infants, we also assessed whether these differences could be seen in the LEGACY cohort. For 10-day-old infant stool microbiomes, principal component analysis (PCA) showed varying degrees of separation by birth mode, but appreciable overlap of all groups. However, many vaginally-delivered infants clustered separately from caesarean-born infants ( Supplementary Figure 5A , upper-right quadrant), and significant differences in microbiome composition by birth mode were detected for all pairwise comparisons (P <0.01). Scores across the first principal component, where most separation occurred at 10 days of life, were higher in vaginally-delivered infants compared to both elective and emergency C/S infants ( Supplementary Figure 5B ; P <0.0001) and were also higher in infants delivered by emergency rather than elective C/S (P <0.01). At 3-months of life, compositional differences in infant stool microbiomes were vastly reduced, with high amounts of overlap among all birth modes ( Supplementary Figure 5C ). Despite this, differences between vaginally delivered infants and both emergency and elective C/S deliveries remained statistically significant (P <0.05, both comparisons). No differences in alpha diversity were observed at either timepoint ( Supplementary Figures 5D, E , P >0.99, all comparisons).

We next attempted to identify other clinical factors that could explain differences in infant stool microbiomes commonly attributed to birth mode, including breastfeeding modality, intrapartum antibiotics, maternal ethnicity, infant sex and chorioamnionitis. Hierarchical clustering suggested that intrapartum administration of antibiotics to mothers was associated with specific stool microbiome clusters at 10 days of life ( Supplementary Figure S6 ). Stratifying infant stool microbiomes by maternal antibiotic administration during labour and delivery for any reason (e.g., caesarean delivery, GBS colonisation, suspected chorioamnionitis, etc.), PCA showed significant clustering of 10-day-old infant stool microbiomes based on antibiotic exposure ( Figure 4A , P <0.001), which was less evident, albeit still statistically significant at 3 months ( Figure 4B , P <0.05). PC1 scores for 10-day-old stool microbiomes were significantly lower in infants exposed to antibiotics when considering all infant stool microbiomes but also when restricting analysis to vaginally-delivered infants only ( Figures 4C, D ; P <0.0001 and <0.001, respectively). The similarity of these data to analyses grouped by delivery mode implies that delivery mode alone does not account for differences between infant stool microbiomes. To explore this further, we investigated the independent effects of delivery mode (comparing elective C/S infants with vaginally-delivered infants with antibiotic exposure) and intrapartum antibiotics (vaginally delivered infants with and without antibiotic exposure) on 10-day-old infant stool microbiomes. Both factors independently affected microbiome composition to a degree ( Supplementary Figure S7 ; delivery mode, P <0.05; antibiotics, P <0.001).

Although alpha diversity did not differ by antibiotic exposure ( Figures 4E, F ; P = 0.156 and 0.642, respectively), differential abundance analysis identified species driving antibiotic-mediated microbiome differences. In 10-day-old infants, Klebsiella michiganensis, Haemophilus parainfluenzae, Enterococcus faecalis and Streptococcus parasanguinis were significantly more abundant in antibiotic-exposed infants, whereas Escherichia. coli, Parabacteroides distasonis, Bacteroides vulgatus, and Bifidobacterium longum subsp. longum were overrepresented in unexposed infants ( Figure 4G ; Supplementary File S3 ). These findings were supported by generalised linear modelling, with MaAsLin2 demonstrating associations between antibiotic exposure and many of the same species in infant stool microbiomes ( Figure 4H ; Supplementary File S3 ). Additionally, Klebsiella pneumoniae was associated with perinatal antibiotic exposure, while Bifidobacterium bifidum and Collinsella aerofaciens were associated with a lack thereof. When grouping infant stool microbiomes by delivery mode, both ALDEx2 and MaAsLin2 identified many of the same differentially abundant species as above when comparing vaginal and elective C/S deliveries; however, differences between emergency C/S and vaginal deliveries were more subtle. ( Supplementary Figure S8 ). At 3 months, P. distasonis was the only differentially abundant taxon defined by ALDEx2 and was overrepresented in unexposed/vaginally-delivered infants, while MaAsLin2 identified an association only between elective C/S and Enterococcus faecalis. ( Supplementary Figure S8 ; Supplementary File S3 ). Differential abundance data were corroborated by comparing relative abundances of these taxa between antibiotic exposure groups. All abundances were significantly different between groups in 10-day-old infants, but only P. distasonis and E. faecalis exhibited significantly different abundances at 3 months ( Figures 4I, J ).

Assessment of maternal microbiome distribution among infant stool microbiome clusters by antibiotic exposure rather than birth mode did not alter the prior conclusion that the maternal vaginal microbiome is not predictive of infant stool microbiome composition ( Supplementary Figure S9 ). Finally, six infants in the LEGACY cohort underwent vaginal seeding: a practice whereby maternal vaginal secretions are transferred to CS-delivered infants in an effort to correct birth mode-related dysbiosis. Therefore, we examined their stool microbiome profiles to determine if the procedure had restored taxa associated with vaginally-delivered infants. Only 1/6 infants undergoing vaginal seeding showed an abundance of taxa associated with vaginal delivery at either timepoint: the remaining 5/6 were dominated by taxa associated with C/S ( Figure 5 ).