Salivary microbiomes of indigenous Tsimane mothers and infants are distinct despite frequent premastication

Study design

Maternal and infant saliva samples were collected from 12 dyads in three villages. The participating dyads were part of an ongoing research project being conducted by Author 2 on Tsimane infant feeding transitions and health outcomes. All infants were reported by their mothers to still receive premasticated foods. For 10 of 12 infants, premastication in the two weeks prior to sample collection was also documented in routinely collected 24-hour dietary recalls (Table 1). All infants were currently breastfeeding, though measures of daily breast milk intake were not collected or estimated. Neither maternal nor infant subjects had taken antibiotics in the week prior to sample collection. Subjects were selected as a matter of convenience, as in their respective communities they lived in close proximity to the researcher’s field base.

Mothers and infants varied in their respective ages and dental health (Table 1). The following maternal subjects were sisters: Mother.1/Mother.3; Mother.10/Mother.11; Mother.13/Mother.15.

The four mothers from Village 3 who participated in the controlled experiment to collect premasticated food samples were also neighbors and frequently exchanged food or shared meals. As no dental records were available, we report the total number of teeth per subject rather than missing teeth. We assume that incomplete tooth counts for adult subjects (less than 32 teeth) primarily represent missing teeth due to dental disease and/or manual extraction, whereas incomplete tooth counts in infants (less than 20) primarily reflect different stages of tooth eruption.

Sample collection

DNA isolation

Following manufacturers’ protocols, a PSP Saliva Gene Kit (STRATEC Molecular GmbH, Berlin, Germany) was used to extract DNA from the saliva samples with added lysozyme for better lysis of Gram positive bacterial cells as suggested by the supplier. And a PowerLyzer^® UltraClean^® Microbial DNA Isolation Kit (MoBio Cat# 12255-50; MO BIO Laboratories, Carlsbad, CA, USA) was used to extract DNA from the food samples. DNA concentrations were measured with a Qubit Fluorometer (Life Technologies, Grand Island, NY, USA). Only two of the five control food samples were selected for DNA analysis, and did not yield sufficient DNA for PCR amplification. Two of the four premasticated food samples (corresponding to Mother.13 and Mother.14, Table 1) also failed to yield sufficient DNA for further analysis. PCR amplification and sequencing was therefore limited to the 24 saliva samples and only two premasticated food samples (FD12 and FD14, Table 1).

PCR amplification and sequencing

The V4 region of the bacterial 16S rRNA gene was amplified with a NEXTflex 16S rRNA gene V4 Amplicon-Seq Kit (with 48 barcodes) (BIOO Scientific, Austin, TX, USA) following the manufacturer’s instructions. The kit produces amplicons of 250 bases that can be sequenced with Illumina technology from both end with significant overlap. The PCR primers are V4_515f adaptor 5′AATGATACGGCGACCACCGAGATCTACACTAT GGTAATTGT GTGCCAGCMGCCGCGGTAA3′ and V4_806r adaptor 5′CAAGCAGAA GACGGCATACGAGAT XXXXXXXXXXXX AGTCAGTCAGCC GGACTACHVG GGTWTCTAA3′, with “X” indicating indexing sequencing capable to code up to 1,152 individual samples. Sequences at 5′ ends of the primers are adaptors for Illumina sequencing, those at 3′ ends are 16S rRNA gene primers. In total, 50 ng DNA was used as a template for each reaction. The amplified DNAs were verified on a gel and pooled at an equal molar ratio. The pooled DNA samples were purified with Agencourt AMPure beads. Sequencing was performed on a MiSeq (Illumina, San Diego, CA, USA).

16S rRNA gene sequence data processing

An analysis of the 16S rRNA gene sequencing data was performed using the mothur software package (Schloss et al., 2009). Mothur was selected for ease of installation, clear and brief instructions, and compatibility with other software (Nilakanta et al., 2014). Initially, the paired end sequences were assembled together, and any sequences with ambiguous bases (32% of them) were removed. To ease the sequence classification process, unique sequences were collected. The collection was further reduced by choosing representatives of near-identical sequences with fewer than 2 different bases. Chimeric sequences (∼18%) were removed with UCHIME (Schloss, Gevers & Westcott, 2011). Sequences left were then aligned using the SILVA databases supplied in mothur, and those sequences classified as Chloroplast, Mitochondria, Archaea, Eukaryota, or Unknown were removed (293 of them). The sequence yield and indexing codes for all samples analyzed are given in Table S1.

The unique sequence collection was clustered into operational taxonomic units (OTUs), with a similarity of 97%. OTU with single or double sequences were removed. Each OTU was assigned a taxonomic classification at all levels, from phylum to genus, using the reference Ribosomal Database Project (RDP) database provided in mothur. Alpha diversity was analyzed with Rarefraction curve deduced with tools from mothur. The count table was normalized with cumulative count scaling implemented in metagenomeSeq (Paulson et al., 2013). Distance ordination was done with normalized data. Total-sum scaling was followed for taxa comparison. The dissimilarity distances between samples were calculated in mothur using the dist.shared command. Thetayc distance matrices were used to compare phylogenetic distances among infants and mothers, and to examine differences in phylogenetic distance within and between dyads (Yue, Clayton & Lin, 2001) Weighted Unifrac distance metrics (Lozupone et al., 2011) were calculated to compare the microbial communities of mothers and infants, which result is also confirmed with ANOSIM (Anderson, 2001). Distance ordination was ploted with non-metric multidimensional scaling method that implemented in metagenomeSeq (Paulson et al., 2013). Finally, we attempted to identify some OTUs of interest at the species level using the HOMD library with the online 16S rRNA gene sequence identification service (Chen et al., 2010).

Statistical analyses

Student t-tests were used to compare alpha diversity, abundances at different phylogenetic levels, and phylogenetic distances between mothers and infants. The Shapiro–Wilk test was used to test the normality of the distribution of taxonomic data (the abundance of most common taxa). Of 91 data sets (12 samples each, infant and mother considered separately) 57% were normally distributed (p < 0.05). Mann–Whitney U tests were used for data sets that were not normally distributed. P-values were further adjusted for multiple comparisons using the false discovery rate method (Benjamini & Hochberg, 1995). All analyses were performed in R. Food samples were not statistically evaluated since only two samples were sequenced.

Ethics approval and consent

This study was approved by the Human Subjects Committee of the University of California Santa Barbara (ANTH-GU-MI-010-3U) and the Human Subject Research Review Board of Los Alamos National Laboratory (LANL-14-11 X). The Tsimane Health and Life History Project maintains annual agreements with the local municipal government of San Borja, the Hospital of San Borja, and the Tsimane governing council, El Gran Consejo Tsimane, to conduct research with the Tsimane. Maternal subjects gave verbal informed consent for themselves and their infants to participate prior to sample collection.

General characteristics of Tsimane mothers’ and infants’ oral microbiomes

The bacteria in the saliva and premasticated food samples belonged to 18 different phyla and 341 genera (Table S2). The main bacterial phyla were Firmicutes (with the predominant genus Streptococcus), Proteobacteria (Neisseria), Actinobacteria (Actinomyces), and Bacteroidetes (Prevotella). Other phyla included Fusobacteria, Spirochaetes, TM7, Tenericutes, SR1, and Verrucomicrobia. The relative abundance of each phylum was different between the saliva and food samples of mothers and infants (Figs. 1 and 2). Infant saliva samples contained more Firmicutes (p < 0.1) and less Fusobacteria and Spirochaetes (p < 0.5) than did maternal samples. The mean relative abundance of the most abundant families across all saliva samples are listed in Table S3.

Thirteen genera—Streptococcus, Neisseria, unclassified Pasteurellaceae, Gemella, Prevotella Leptotrichia , Actinomyces, Rothia, Veillonella, Sphingomonas, and unclassified genera from the Leptotrichiaceae, Flavobacteriaceae, and Lactobacillales families—were observed in all saliva samples. The most common genus, Streptococcus, has 4 OTUs (OTU0001, OTU0003, OTU0029 and OTU0084) and was dominant in 92% (22/24) of saliva samples. OTU0001 is identical to Streptococcus mitis, S. oralis, S. infantis and S. mitis bv.2based on comparison with the Human Oral Microbiome Database (HOMD) (Chen et al., 2010) We could not distinguish among these species, as the V4 regions of their 16S rRNA gene sequences are identical. It is possible that samples harbored more than one of these species. The average abundance of OTU0001 was 413% in infants’ saliva and 25.2% in mothers’ saliva (adjusted p < 0.01). OTU0003 is identical to 16S rRNA gene sequences of S. salivarius and S. vestibularis in HOMD. This OTU was about a third as abundant as OTU0001, and twice as abundant in infants (221%) as in mothers (11.5%) (p < 0.01). OTU0029 is identical to S. Intermedius. This OTU was more abundant in mothers’ saliva (0.8% vs 0.1% in infants), but the difference is not statistically significant. OTU0084 matches Streptococcus sp. oral taxon 487 in HOMD at 98.8% and was found in only two infant and four maternal samples. Overall, infants’ saliva contained significantly more Streptococcus (65.0% vs 36.9%, p < 0.01) and less Leptotrichia (0.2% vs 2.4%, p < 0.01) and Catonella (2 infants and 11 mothers have it, p < 0.01) than mothers’ saliva after adjusting for false discovery rate.

Microbial populations of premasticated foods in relation to donor salivary populations

Bacterial populations from the two premasticated food samples that could be sequenced were more similar to those of their donors than to those of other mothers, although this finding only trends toward statistical significance (p = 0.065), likely due to the small sample size. Differences in dissimilarity distances of the foods samples may also reflect individual differences in chewing “style” and duration. The sample that was chewed for 17 s had a smaller dissimilarity distance than that of the sample chewed for 8 s (0.061 vs 0.191), suggesting more donor bacteria may be incorporated in food with increased chewing time.

Despite within-subject similarities between premasticated food and donor saliva, 16S rRNA gene sequencing revealed distinct differences between chewed food and saliva samples (Table S4), which were both collected via expectoration. Sixty OTUs were enriched more than 5-fold in premasticated food compared with saliva samples, and 36 OTUs were only observed at >0.01% in premasticated food samples. Some bacteria in the food samples may have been incorporated through the surface water used in food preparation, as we identified bacteria associated with plants, soil, and water reservoirs in the premasticated food samples. However, we did not yield sufficient DNA from the original unchewed food for 16S rRNA gene sequencing in two attempts, suggesting cooking decreased the availability of bacteria or bacterial DNA in the food samples.

Relatedness of maternal and infant salivary microbiomes

We assessed relatedness of maternal and infant salivary microbiota by comparing alpha (intrasample) diversity and beta (intersample) diversity among saliva samples.

The internal sample alpha diversity was measured by rarefaction analysis (Fig. S1). On average, infant saliva samples had 41 OTUs when sampling 3148 sequences, whereas maternal samples had 87 (p < 0.01). The lower diversity of infants’ saliva paralleled greater dominance of Streptococcus. Intra-sample diversity was not correlated with age or number of teeth when analyzing mothers and infants separately, but was correlated with age when analyzing the two groups together (r = 0.75, p < 0.01). Alpha diversity differed among maternal subjects according to sample collection method:113 ± 24 for expectorated samples and 75 ± 22 for samples collected via buccal swab (p = 0.01).

Both ANOSIM and weighted Unifrac analysis indicated that mothers and infants had different salivary microbial community structures (p < 0.01). However, the difference was not significant when calculated with unweighted Unifrac analysis, which indicated that the abundance of shared OTUs contributed more to the weighted Unifrac calculation. The average distances among infant saliva samples (0.30 ± 0.18) were significantly shorter than those among maternal samples (0.48 ± 0.21) (p < 0.01), indicating greater similarity among infant microbial communities than among maternal communities (Fig. 3). Average distances calculated for related mother–infant pairs (0.45) were not significantly shorter than those of unrelated mother–infant pairs (0.47), suggesting community structures of related pairs were no more similar than those of unrelated pairs (Fig. 3). A non-metric multi-dimensional scaling ordination (NMDS) plot made using a Thetayc tree further shows that infants’ salivary microbiomes were more similar in community structure to one another than to maternal microbiomes (Fig. 4).

We also calculated the average distance between each infant with all possible combinations of mothers. Across infants, this average distance was negatively correlated with infant age (r =  − 0.64, p < 0.05, Fig. S3A). The average distance between each maternal sample and all infant samples, however, was not associated with maternal age (r =  − 0.33, p = 0.29) (Fig. S3B) Tsimane salivary microbial communities may therefore become more adult-like during late infancy (9–24 months), but there is little age-associated variation in community structure during adulthood (18–41 years) in this study group.

Oral pathogens identified in maternal and infant saliva samples

Chronic dental infections are common in Tsimane people, as evidenced by numerous caries and tooth loss in our maternal subjects (Table 1). We searched for pathogens associated with dental disease in maternal and infant samples, which could indicate vertical transmission. Using the HOMD library, OTU0120 matched S. mutans at 100%; OTU0143, T. denticola at 98%; OTU0068, T. forsythia at 97.2%; and OTU0033, P. gingivalis at 99.2% Because OTU0068 and OTU0143 matched target species at a relatively low level, we restricted our analyses to OTU0033 (P. gingivalis) and OTU0120 (S. mutans) and traced them across mother–infant dyads. However, these species-level identifications are limited because we did not perform confirmatory tests to identify these OTUs as pathogens.

In normalized results, OTU0033 (P. gingivalis) was observed in 11 mothers’ saliva samples (92%, MSA1, MSA8, MSA12, MSA13, MSA15), but not in any of the infants’ samples (p < 0.01). OTU0120 (S. mutans) was observed in 7 infants’ saliva samples (BSA1, BSA11, BSA12, BSA13, BSA14, BSA15, BSA8) and 8 mothers’ samples (MSA1, MSA10, MSA11, MSA12, MSA13, MSA14, MSA15, MSA9). The frequency of S. mutans did not significantly differ across maternal and infant samples (p = 0.24). Three pairs of mother–infant samples (BSA8 and MSA8, BSA9 and MSA9, BSA10 and MSA10) did not match each other in S. mutans status, and the remaining pairs (9) did. The occurrence and abundance of the two possible pathogens were not correlated with age, the number of teeth, dental caries, or the loss of teeth within each group. There were 754 S. mutans reads from infant saliva sample BSA14, compared with fewer than 5 in other samples, but the subject providing this sample had no observable dental caries at the time of sampling.

Fourteen of 22 genera associated with gingivitis were identified in 50% or more of maternal samples (Table S5). Seven of these 14 genera were identified in only 1/3 or less of infant samples, and no Eubacterium or Johnsonella were identified in any infant sample (Table S5). Fusobacterium, Leptotrichia, Porphyromonas, and Prevotella were identified in at least 83% of both maternal and infant samples (Table S5). All were more abundant in the maternal samples (Fig. S2). With the exception of Tannerella, the gingivitis-associated OTUs positively identified in infant samples were also identified in the corresponding maternal samples, but not vice versa.