Subjects and pyrosequencing results
Saliva samples were taken on three occasions (1–3) from children with AOM, treated (A) or not treated (C) with amoxicillin. Patient characteristics are presented in Table 1 and in Supplementary material, Table S1.
Table 1. Patients' characteristics
| ||Antibiotic group (n = 15)||Control group (n = 18)||pa|
|Mean age (months ± SD)||23 ± 14||34 ± 18||0.07|
|Male sex||11 (61%)||6 (40%)||0.23|
|Antibiotic treatment in last 12 months||6 (33%)||3 (20%)||0.39|
A total of 212 852 filtered pyrosequencing reads were analysed. Trimmed sequences corresponded to the 16S rDNA hypervariable region V3 and a part of the conserved region between V2 and V3 (positions 257–514 in the E. coli 16S rRNA gene) and had an average length of 257 nucleotides. We extracted this region from the SILVA Bacterial reference set (14 956 16S rRNA gene sequences)  to assess the accuracy of taxonomic assignments. The proportion of such simulated 454 reads correctly and erroneously assigned at the genus level, using RDP Classifier at a bootstrap cut-off of 80%, reached 76% and 2%, respectively (see Supplementary material, Fig. S1).
The six phyla (Firmicutes, Proteobacteria, Bacteroidetes, Actinobacteria, TM7 and Fusobacteria), commonly found in salivary bacterial communities [11, 21-24], dominated the samples in our study (Fig. 1 and Supplementary material, Table S2). Averaging of the data from our study and published studies [11, 21-24] showed a trend of a higher proportion of Firmicutes and a decrease in the frequency of six other major phyla in saliva from children/adolescents in comparison to adults (Fig. 1).
Figure 1. Relative abundance of seven major phyla across different studies of the salivary microbiome. Data obtained for different individuals (and different time-points if available) were averaged for each study except that of Keijser et al.  in which individual non-bar-coded amplicon libraries were pooled in equimolar amounts before sequencing. ‘A’ and ‘C’ correspond to the treated and untreated group of children analysed in this study, respectively. ‘Children/adolescents AVG’ and ‘Adults AVG’ were obtained by averaging data from the relevant studies.
Download figure to PowerPoint
Two out of 81 genera identified, Amaricoccus and Moritella, have not been previously found in the salivary microbiota [11, 21-24], but each of them had only one occurrence. This supports the claim that the most frequent and widespread members of the human salivary microbiome in the western world have already been described . However, the analysis of samples from other geographic locations still identifies new relatively abundant genera .
The dataset was represented by 11 028 distinct sequences (100%-ID phylotypes) and 1 486 OTUs. In a recent study of the salivary microbiome, which included 264 samples from 107 individuals aged from 8 to 24 years, two OTUs belonging to the genus Streptococcus were found in all but one sample and corresponded to 10.5% of all sequence reads . Here, on a cohort including 33 subjects aged 1–6 years, 18 of which were exposed to the antibiotic, an OTU was identified across all time-point samples of all individuals (99 samples). It was also assigned to the genus Streptococcus and represented on average 44.6% of the total number of sequence reads per sample. Two additional OTUs from the genera Gemella and Granulicatella were shared by all (n = 45) samples of control subjects. After combining the sequence reads of the three time-point samples from the same individual into a single dataset, the ‘universal core’ of the entire cohort reached an average size of 69% and included 11 OTUs belonging to the genera Streptococcus (4), Granulicatella (2), Gemella (2), Veillonella (1), Coprococcus (1) and Fusobacterium (1).
The assessment of differences in relative taxon abundance at different visits (1, 2 and 3) revealed 39 significant changes within the antibiotic-treated (A) group and none within the control (C) group (Table 2). In all cases (31) where a statistically significant shift in the average taxon abundance was found between pre-treatment (A1) and end-of-treatment (A2) samples, we also observed the opposite direction of change between end-of-treatment and post-treatment (A3) samples, which points to a recovery from antibiotic treatment.
Table 2. Changes in taxa abundance at different visits
|Rank and taxon||Taxonomic information (phylum; class; order; family; genus)||Relative abundance||p||Number of subjects with increased/decreased abundancea|
|A1||A2||A3||A2 vs. A1||A3 vs. A1||A3 vs. A2||A2 vs. A1||A3 vs. A1||A3 vs. A2|
|Phylum Actinobacteria|| ||4.32||1.48||0.96||0.26||2.85||1.31||0.0070|| ||0.0478||3/15||8/10||14/4|
|Phylum Proteobacteria|| ||7.32||4.64||21.35||8.75||12.75||8.17|| ||0.0317|| ||11/7||13/5||8/10|
|Phylum TM7|| ||1.53||0.51||0.02||0.00||0.99||0.06||0.0019||0.0317||0.0478||0/17||2/15||9/1|
|Class Actinobacteria||Actinobacteria||4.32||1.48||0.96||0.26||2.85||1.31||0.0114|| || ||3/15||8/10||14/4|
|Class Clostridia||Firmicutes||6.24||4.61||4.21||2.34||2.38||1.75|| ||0.0142|| ||7/11||2/16||8/10|
|Class Erysipelotrichi||Firmicutes||1.33||0.14||0.04||0.00||0.17||0.00||0.0114|| || ||0/12||4/10||6/2|
|Order Actinomycetales||Actinobacteria; Actinobacteria||3.99||1.25||0.96||0.26||2.78||1.31||0.0214|| || ||4/14||9/9||14/4|
|Order Coriobacteriales||Actinobacteria; Actinobacteria||0.33||0.08||0.00||0.00||0.06||0.00||0.0146||0.0337|| ||0/12||1/12||5/0|
|Order Clostridiales||Firmicutes; Clostridia||6.24||4.61||4.21||2.34||2.38||1.75|| ||0.0205|| ||7/11||2/16||8/10|
|Order Erysipelotrichales||Firmicutes; Erysipelotrichi||1.33||0.14||0.04||0.00||0.17||0.00||0.0146|| || ||0/12||4/10||6/2|
|Order Burkholderiales||Proteobacteria; Betaproteobacteria||0.80||0.19||9.68||2.64||0.68||0.42||0.0146|| || ||13/5||9/8||5/12|
|Family Corynebacteriaceae||Actinobacteria; Actinobacteria; Actinomycetales||0.34||0.16||0.05||0.00||0.11||0.00||0.0140|| || ||2/11||5/9||7/5|
|Family Micrococcaceae||Actinobacteria; Actinobacteria; Actinomycetales||1.95||0.38||0.13||0.00||0.83||0.28||0.0133|| || ||2/13||8/9||14/1|
|Family Coriobacteriaceae||Actinobacteria; Actinobacteria; Coriobacteriales||0.33||0.08||0.00||0.00||0.06||0.00||0.0133|| || ||0/12||1/12||5/0|
|Family Aerococcaceae||Firmicutes; Bacilli; Lactobacillales||0.24||0.12||1.70||0.31||0.21||0.12||0.0183|| || ||13/4||7/9||6/10|
|Family Erysipelotrichaceae||Firmicutes; Erysipelotrichi; Erysipelotrichales||1.33||0.14||0.04||0.00||0.17||0.00||0.0133|| || ||0/12||4/10||6/2|
|Family Fusobacteriaceae||Fusobacteria; Fusobacteria; Fusobacteriales||0.96||0.59||0.13||0.00||0.64||0.23||0.0003|| || ||1/17||6/12||10/6|
|Genus Corynebacterium||Actinobacteria; Actinobacteria; Actinomycetales; Corynebacteriaceae||0.34||0.16||0.05||0.00||0.11||0.00||0.0170|| || ||2/11||5/9||7/5|
|Genus Rothia||Actinobacteria; Actinobacteria; Actinomycetales; Micrococcaceae||1.95||0.38||0.12||0.00||0.83||0.28||0.0170|| || ||1/13||8/9||14/1|
|Genus Atopobium||Actinobacteria; Actinobacteria; Coriobacteriales; Coriobacteriaceae||0.33||0.07||0.00||0.00||0.06||0.00||0.0170|| || ||0/11||2/11||5/0|
|Genus Prevotella||Bacteroidetes; Bacteroidia; Bacteroidales; Prevotellaceae||0.92||0.54||0.58||0.32||0.25||0.15|| ||0.0347|| ||5/13||3/15||8/10|
|Genus Abiotrophia||Firmicutes; Bacilli; Lactobacillales; Aerococcaceae||0.24||0.12||1.70||0.31||0.21||0.12||0.0192|| || ||13/4||7/9||6/10|
|Genus Solobacterium||Firmicutes; Erysipelotrichi; Erysipelotrichales; Erysipelotrichaceae||1.31||0.14||0.04||0.00||0.17||0.00||0.0170|| || ||0/12||4/10||6/2|
|Genus Fusobacterium||Fusobacteria; Fusobacteria; Fusobacteriales; Fusobacteriaceae||0.96||0.59||0.13||0.00||0.64||0.23||0.0003|| || ||1/17||6/12||10/6|
|OTU1357||Actinobacteria; Actinobacteria; Actinomycetales; Micrococcaceae; Rothia||0.28||0.19||0.03||0.00||0.12||0.00||0.0241|| || ||0/13||5/10||7/1|
|OTU1364||Actinobacteria; Actinobacteria; Actinomycetales; Micrococcaceae; Rothia||1.65||0.14||0.10||0.00||0.70||0.22||0.0378|| || ||1/11||9/7||13/1|
|OTU1354||Actinobacteria; Actinobacteria; Coriobacteriales; Coriobacteriaceae; Atopobium||0.32||0.07||0.00||0.00||0.06||0.00||0.0241|| || ||0/11||2/11||5/0|
|OTU502||Firmicutes; Bacilli; Bacillales; Staphylococcaceae; Gemella||1.40||1.16||0.26||0.08||1.21||0.50||0.0241|| || ||2/16||5/13||14/3|
|OTU609||Firmicutes; Bacilli; Lactobacillales; Aerococcaceae; Abiotrophia||0.22||0.10||1.54||0.25||0.17||0.11||0.0419|| || ||12/5||7/9||5/10|
|OTU382||Firmicutes; Bacilli; Lactobacillales; Streptococcaceae; Streptococcus||4.44||3.50||1.77||0.08||1.85||0.49||0.0401|| || ||3/14||4/13||10/6|
|OTU1026||Firmicutes; Erysipelotrichi; Erysipelotrichales; Erysipelotrichaceae; Solobacterium||1.29||0.14||0.04||0.00||0.17||0.00||0.0241|| || ||0/12||4/10||6/2|
|OTU1408||Fusobacteria; Fusobacteria; Fusobacteriales; Fusobacteriaceae; Fusobacterium||0.74||0.46||0.10||0.00||0.54||0.17||0.0011|| || ||1/17||8/10||10/6|
|OTU478||Proteobacteria; Betaproteobacteria; Burkholderiales||0.77||0.19||9.44||2.34||0.64||0.33||0.0241|| || ||13/5||9/8||5/12|
|OTU1502||TM7||0.44||0.10||0.00||0.00||0.38||0.00||0.0241|| || ||0/11||3/10||5/1|
|OTU1577||TM7||0.88||0.31||0.01||0.00||0.51||0.00||0.0241|| || ||0/13||2/12||8/0|
The end of treatment correlated among others with a reduced proportion of the phyla TM7 and Actinobacteria, genera Corynebacterium, Rothia, Atopobium, Prevotella, Solobacterium and Fusobacterium, as well as OTUs assigned to most of these genera but also to Streptococcus and Gemella. The proportion of two OTUs, assigned respectively to Abiotrophia and order Burkholderiales, increased at the end of treatment. Given the antibiotic spectrum of amoxicillin , the reduction in Streptococcus, Corynebacterium and Prevotella was expected. In contrast, genera Neisseria and Haemophilus (and OTUs assigned to them) were not consistently reduced at the end of treatment.
Among 31 taxa with statistically significant changes in the relative abundance between A1 and A2 samples, 13 had the same direction of change (increase or decrease) in all relevant individuals. For each of the remaining 18 taxa, the change in one direction prevailed over the other (Table 2). This was also the case with all the taxa showing significant changes in relative abundance in A1–A3 and A2–A3 sample comparisons.
At the post-treatment sampling point, Proteobacteria were significantly more abundant, whereas phylum TM7, class Clostridia, order Coriobacteriales and genus Prevotella had lower relative abundances compared with the baseline (A1) levels. It has been previously reported using the culture method that amoxicillin treatment reduced the number of Prevotella spp. isolates considered to interfere with the growth of potential pathogens in the nasopharynx .
We compared the saliva bacterial communities in terms of their phylogeny within the antibiotic-treated and control groups using UniFrac. Principal coordinate analysis of unweighted UniFrac distances in group C showed that samples taken from the same individual over time tend to cluster together (Fig. 2). In group A, the samples taken before and at the end of the amoxicillin treatment formed two relatively well separated clusters both of which partly overlapped with the cluster formed by the post-treatment samples (Fig. 2). The effect of antibiotic is clearly seen in the first two principal coordinates where 16 (of 18) end-of-treatment samples (A2) were displaced towards the left relative to the corresponding pre-treatment samples (A1) and all their post-treatment (A3) counterparts returned right. Similarly, 16 A2 samples were displaced downwards relative to A1 and 11 of them moved upwards at post-treatment.
Figure 2. Principal coordinate analysis of unweighted UniFrac distance matrices. Blue squares, orange circles and grey-contour triangles correspond to samples taken at the time of the first, second and third visit, respectively. The area comprising samples taken at the time of the first and second visit for treated (a) and control (b) subjects are blue and orange shaded, respectively.
Download figure to PowerPoint
For 12 individuals the A1–A2 and A2–A3 vectors on principal coordinate analysis had opposite directions. Of these, in ten cases (#A4, 5, 6, 15, 20, 21, 23, 24, 27, 32) the A2 points moved down and left relative to A1 (see Supplementary material, Fig. S2), whereas in two cases (#A7, A10) other patterns were observed. For the remaining six individuals (#A11, 12, 13, 16, 18, 28), the A1–A2 and A2–A3 vectors were not in opposite directions; the only five individuals who had the A1–A3 UniFrac distance greater than the A1–A2 one were found among these six subjects. Separation of pre-treatment and end-of-treatment samples was less clear using weighted UniFrac analysis (see Supplementary material, Fig. S3), although two-thirds of both A2 and A3 samples were located outside the area formed by the A1 samples.
Variations of salivary microbiota among time-points for the same control subject were smaller than those between different control subjects at the same time point (Fig. 3 and see Supplementary material, Fig. S4). In the antibiotic-treated group, however, the average within-individual variations were not significantly different from between-individual variations (Fig. 3). Clearly, this was a result of the perturbation of the microbiota by the antibiotic.
Figure 3. Average within-subject and between-subject unweighted UniFrac distances. ‘A’ and ‘C’ correspond to the treated and untreated groups of children. Standard error bars are given and the Mann–Whitney U-test significance is indicated as follows: *p <0.05; **p <0.01, *** p<0.001.
Download figure to PowerPoint
The results of permanova of unweighted UniFrac distances confirmed the marked effect of the antibiotic on the salivary microbiota (see Supplementary material, Table S3). Samples taken at the end of amoxicillin treatment (A2) were significantly different from the five other groups of samples defined by the treatment and medical visit (A1, A3, C1 C2, C3). Similar results were obtained with permanova of Bray–Curtis similarities. This test was based on square-root transformed abundance of OTUs and did not take into account the phylogenetic distance as was the case using UniFrac.
It has been shown that the oral microbiota is more stable compared with other human microbiota , but most important changes occur in childhood [23, 27]. Our study did not include children younger than 1 year, a period when dramatic microbiota changes occur . The result of permanova (see Supplementary material, Table S4) suggested that, at the time of the first visit, the salivary microbiota were not significantly different between: (i) children belonging to different age groups defined by 1-year increments (see Supplementary material, Fig. S5), except when 1–2-year-old children were compared with those who were 5–6 years old; (ii) genders; and (iii) patients prescribed (A1) and not prescribed (C1) amoxicillin. Previous AOM episodes, recorded in 11 patients, nine of which were treated by antibiotics, showed a weak effect on the salivary microbiota when a permanova test based on unweighted UniFrac distances was used (Pseudo-F = 1.5193, p 0.043).
Diversity of the salivary microbiota
Diversity and richness estimates are considered to be sensitive to the number of sequences analysed , so we normalized the data of each sample to 283 sequences, as it was the lowest number of sequences in any sample. Margalef richness and Shannon diversity indices were lower in 17 and 16 (of 18) A2 samples, respectively, in comparison to their A1 counterparts (see Supplementary material, Table S5). The average values for both indices in A3 samples were between values found in A1 and A2 samples (Fig. 4). In the control group, the average diversity and richness indices were somewhat reduced in samples taken at the time of the second visit in comparison to the baseline and third-visit samples, but these differences were not statistically significant. Decrease in Shannon diversity and Margalef richness indices, observed in 13 and 11 (of 15) control subjects, respectively, may be the result of some changes in diet, hygiene or immunological status linked to the illness.
Figure 4. Richness and diversity indices of the salivary bacterial communities. Margalef richness (d) and Shannon diversity (H‘) indices were calculated for the six groups of samples. Each sample was normalized to 283 sequences before analysis. Average values and standard error bars are presented. The statistical analysis was Wilcoxon signed-rank test.
Download figure to PowerPoint