Multi-niche, multi-resolution study framework
A multi-niche, multi-resolution approach (Figure 1) was implemented to compare oral microbial communities of children and adults, to determine mother to child bacterial transmission within genetically related and unrelated families. Subjects recruited for this study included 50 adoptive mother-child- pairs and 55 biological mother-child pairs. The adoptive group included only children adopted at birth by a non-genetic relative. The biological group was matched on age (p=0.29) and socioeconomic status to the adoptive group. Additionally, an extended family dataset of samples was obtained from 23 fathers and 16 siblings of the children in the biological group, allowing comparison of microbial profile similarity among siblings, couples, and child-father pairs. Detailed meta-data including feeding and delivery mode, health measures, and demographics were also collected.
To comprehensively profile microbial communities from multiple niches within the oral cavity, one soft tissue/saliva swab sample, one supragingival plaque sample, and one subgingival plaque sample were collected from each subject, with the exception of predentate children, from whom no tooth-associated sample could be collected. Microbial communities from each sample were independently profiled at both the species and strain level. Species level resolution was achieved through sequencing of the 16S V1-V3 region and mapping to the OSU CORE12 reference database of oral bacteria, and strain level resolution was achieved by targeted sequencing of the 16-23S Intergenic Spacer Region (ISR) combined with high-resolution processing using DADA2, as shown previously5.
For strain-level community analysis, the ISR-amplicon library was sequenced using the Illumina HiSeq 2500 platform and processed with DADA213 to generate unique ISR amplicon sequence variants (ISR-ASVs) or ISR-type strains. Seven samples which did not generate sufficient sequences (cutoff 5000 reads) were excluded from analysis. A conservative approach was implemented in calling true biological strain variants by including in the analysis only those ASVs found in more than 1 sample. The final dataset consisted of 778 samples, representing 3,865 ASVs. This included 49 adoptive and 54 biological mother-child pairs for the saliva dataset, 46 adoptive and 53 biological pairs for the supragingival dataset, and 44 adoptive and 51 biological pairs for the subgingival dataset. A detailed breakdown of samples in each group is included in Supplementary Table ST1.
For the species-level community analysis, the same samples were used to generate a 16S V1-V3 amplicon library, sequenced on Illumina MiSeq platform, processed using Mothur11, and species level taxonomy was assigned to the reads using CORE12 oral 16S database. The final quality filtered dataset consisted of 709 samples, representing 581 species-level OTUs. This included 45 adoptive and 48 biological mother-child pairs for the saliva subset, 45 adoptive and 48 biological pairs for the supragingival subset, and 40 adoptive and 46 biological pairs for the subgingival subset. Supplementary Table ST2 lists the details of samples in each group for the 16S V1-V3 dataset.
Non-metric multidimensional scaling (NMDS) ordination, using Bray-Curtis (BC) dissimilarities based on community membership was used to compare beta-diversity of all the mother and child samples, both at the strain (ISR) and species (16S) levels (Figure 2). Profiling strain-level communities showed significantly greater separation between the samples compared to species-level communities (Supplementary Figure S1). No distinction between the adoptive and biological families in terms of beta diversity was observed (Supplementary Figure S2).
No genetic influence on acquisition of oral bacteria observed
Similarity between the microbial profiles of adoptive and biological mother-child pairs was quantified using a distance-based approach. BC dissimilarities were computed for each mother-child pair based on presence/absence of species/strain variants. This index ranges from 0-1, with samples having exactly identical microbial communities scored at 0 and absolutely different communities scored at 1. We compared the BC distances between mother-child at both species and strain levels (Figure 3) using the Wilcoxon rank sum test. Microbial profiles of biological and adopted children were equally similar to their mothers, for all three niches we sampled: saliva/soft tissue, supragingival plaque and subgingival plaque. This was true at both species and strain level resolutions. BC dissimilarities of all possible unrelated mother-child pairings among the samples were also computed. A group containing all possible distances between any unrelated mother-child pairs is often compared with distances between adoptive or biological mother-child pairs using a Wilcoxon test or t-test, but this violates a basic assumption of the tests by re-using the same observations multiple times. Therefore, as an alternative to the widely used Wilcoxon test or t-test approaches, we used a permutation-based method for statistical comparison between the unrelated group and the adoptive/biological groups (see Methods for details). At the level of strains, all mothers and their own children, regardless of genetic relationship, were significantly more similar to each other than unrelated mother-child pairs. This relationship was not as strong using the lower resolution species-level approach. Similar results were also observed using the Jaccard dissimilarity index (ISR soft tissue/saliva dataset shown in Supplementary Figure S3). Additionally, when using relative abundance measures in place of presence/absence, similar results for comparisons between adoptive/biological groups were observed (ISR soft tissue/saliva dataset shown in Supplementary Figure S4). Given that the 3 sites showed highly similar patterns, and the soft tissue swab/saliva provided the largest dataset for our comparisons because it included the predentate children, those samples were used as the primary dataset in subsequent analyses.
We calculated the average number of species and strains shared between mother-child pairs for the soft tissue/saliva samples (Figure 4). Both adoptive and biological groups shared 44% of their microbiota at species level, and 15% at strain level. As expected, the fraction of shared species was much higher than fraction of shared strains, and unrelated mother-child pairs shared 4 times as many oral species as oral strains. A list of the most widely shared species and their relative abundance in mother-child pairs for both the biological and adoptive families is provided in Supplementary Table ST3. Even though the set of species and strains shared between mothers and children made up 44% and 15% of the total number of species and strain variants, when considering relative abundance, they accounted for 93% and 48% of the total communities on average.
Ruling out possible confounders including feeding and delivery mode
Extensive demographic and clinical data were collected for all subjects and examined to determine if any of these variables significantly influenced mother-child dissimilarities. Targeted recruitment lead to close matching of age, health factors, and socio-economic status between the biological and adopted children (no significant differences). However, the following 9 potentially confounding factors for the mother-child dissimilarities were found to be significantly different between the two groups by Wilcoxon rank sum or Fisher’s exact test: child's feeding mode, child's gingivitis level, child's race, child's tongue biofilm level, mother-child race match, mother's age, mother's plaque level, mother's tongue biofilm level and mother’s gingivitis level. To assess the confounding effect, a univariate regression analysis with mother-child distance as the dependent variable and group as the independent variable was performed, and a multiple regression analysis with the unbalanced factors as additional independent variables was also performed. The estimated regression coefficient of the group variable changed substantially from the univariate model (beta= -0.017, p=0.433) to the multiple regression model (beta=-0.056, p=0.226). However, the group variable remained nonsignificant even after adjusting for the confounding effect in the multiple regression model, thereby ruling out any significant confounding effect on the mother-child dissimilarities. Within the biological group neither feeding mode nor delivery mode had a significant impact on mother-child distances for any niche (Supplementary Figure S5).
Extended family comparisons also fail to show influence of genetics
An extended dataset of the biologically related families that included fathers and siblings was analyzed to further explore the relative contributions of genetics and shared environment. Comparisons of distance between various related and unrelated pairs of subjects for the soft tissue/saliva dataset at strain level is shown in Figure 5. Cohabitating mother-father couples and sibling pairs showed the greatest oral microbiota resemblance of any pairing examined, sharing significantly more strains than mother-child or father-child pairs. Mothers and fathers were equally similar to their children, and both mother- and father-child pairs were significantly more similar than unrelated mother- and father-child pairings. Technical replicates consisting of samples that were processed through the ISR pipeline in duplicates were highly similar. Analysis of the extended family samples from the subgingival and supragingival dataset also produced similar results (Supplementary Figure S6). Taken together, these findings provide further evidence that shared environment and contact, rather than genetic background, is the primary determinant of microbial community structure.
Older children’s microbiota more similar to their mother’s
Child’s age is known to be a major determinant of oral microbiota composition, but targeted recruitment allowed us to ensure that the child’s ages in the adoptive and biologic group were balanced (Figure 6a). Since our results showed that the biological and adoptive children were equally similar to their mothers, the two groups were combined and examined to determine if child’s age had an effect on mother-child dissimilarities. A total of 101 children were considered for this analysis, with an age range of 3 months to 6 years. Contrary to our hypothesis, we saw that younger children’s microbial communities were less similar to that of their mothers compared to older children (Figure 6b). The strong negative association between child’s age and mother-child dissimilarities (Spearman correlation coefficient rho = -0.33, p=0.002) lead us to further explore this relationship. We hypothesized that increasing overall diversity of the oral microbial communities with age of the child may be responsible for increasing similarity of children’s microbiota to their mothers with age. Indeed, our results show that the Shannon Diversity Index, a measure of alpha diversity, increased sharply during the early years, reaching levels comparable to those of the mothers for most children by age 5 years (Fig 6c). Children from both the biological and adoptive groups showed the same pattern, as can be seen in Supplementary Figure S7.
No influence of genetics seen for individual species
While considering the entire bacterial community as a whole did not show any effect of genetic kinship, we wanted to explore whether individual species of bacteria showed differences in degree of strain matching between biologic and adoptive mother-child pairs. Our extensive database of oral ISR sequences allowed us to assign species level taxonomy to ISR-type strains, and a list of the most abundant species, along with the number of strains identified for each, is shown in Supplementary Table ST4. To determine if fidelity of transmission varied for individual oral bacteria species, we compared the adoptive and biologic mother-child dissimilarities from the saliva/soft tissue swab dataset for the ten most abundant species in this dataset (Figure 7). No species showed any significant difference between the adoptive and biologic groups, suggesting no differential heritability.