Abundance of most major phyla that was related to the NT-157 extraction procedure (Figure 1). In samples processed by enzymatic lysis, the relative abundance of sequence reads assigned to the phylum Firmicutes was significantly higher than in the mechanically-treated samples, whereas the phyla Actinobacteria, Spirochaetes, and TM7 showed the opposite pattern across all pipelines. The phyla Synergistetes and Tenericutes also had significantly higher relative abundance in mechanically-processed samples, except in Pipeline 5, in which members of this phylum were not identified. As shown previously [33], different members of the same taxon may present opposite directions of change in relative abundance (increase or decrease) when comparing one extraction method to the other. For instance, the relative abundance of the orders Burkholderiales and Neisseriales, both belonging to the class Betaproteobacteria, were higher and lower, respectively, in mechanically-lysed samples (Figure S1 and Table S2). Similarly, one Streptococcus OTU was present at a significantly higher proportion in samples treated mechanically, whereas the abundance of other OTUs from the same genus was significantly greater in enzymatically-lysed samples.Clustering of Bacterial CommunitiesSimilarities between bacterial communities in terms of their phylogeny were assessed by measuring MedChemExpress INCB039110 UniFrac distance [34]. This metric takes into account the relative abundance of OTUs (weighted data) or their presence/absence (unweighted data). The between-method weighted UniFrac distance was significantly higher (P,10215) than the within-method distance across all pipelines (Figure 3A). The within-method distances for the mechanical and enzymatic lysis methods were not significantly different. Comparisons of bacterial community membership using unweighted UniFrac (Figure 3B) showed that intra-method distances were significantly smaller for the mechanically-treated samples than for the enzymatically-treated samples in the pipelines in which OTUs were clustered using Mothur (Pipelines 1/5, 2 and 3). In these pipelines, as well as in Pipeline 4, UniFrac distances between the mechanically-lysed samples were also smaller than inter-method distances. In contrast, in Pipeline 6, a statistically significant decrease was observed when unweighted UniFrac distances among enzymatically-processed samples were compared with inter-method distances. Permanova of Bray-Curtis similarities confirmed the effect of the extraction method on the structure of the salivary microbiota (Table S3). This test was based on square-root transformed abundance of OTUs and did not take into account the phylogenetic distance, as this was the case when using UniFrac. Similarly, PCoA of the Bray-Curtis similarity matrix based on square-root-transformed OTU abundance showed that samples processed using the same extraction method clustered together (Figure 4). We used this non-phylogenetic approach to analyze smaller size libraries simulated by sampling a given number of OTUs from each sample. The PCoA plots show that as few as 100 randomly chosen OTUs per sample results in a clear separation of bacterial community profiles according to the extraction method (Figure 4). Further reducing this number to 50 still showed the between-method separation, although with some overlap.Taxonomic AssignmentsIn accordance with previous studies [32], seven phyla (Proteobacteria, Firmicutes, Bacteroidetes, Fusobacteria, Actinobacteria, TM7, and Spiroch.Abundance of most major phyla that was related to the extraction procedure (Figure 1). In samples processed by enzymatic lysis, the relative abundance of sequence reads assigned to the phylum Firmicutes was significantly higher than in the mechanically-treated samples, whereas the phyla Actinobacteria, Spirochaetes, and TM7 showed the opposite pattern across all pipelines. The phyla Synergistetes and Tenericutes also had significantly higher relative abundance in mechanically-processed samples, except in Pipeline 5, in which members of this phylum were not identified. As shown previously [33], different members of the same taxon may present opposite directions of change in relative abundance (increase or decrease) when comparing one extraction method to the other. For instance, the relative abundance of the orders Burkholderiales and Neisseriales, both belonging to the class Betaproteobacteria, were higher and lower, respectively, in mechanically-lysed samples (Figure S1 and Table S2). Similarly, one Streptococcus OTU was present at a significantly higher proportion in samples treated mechanically, whereas the abundance of other OTUs from the same genus was significantly greater in enzymatically-lysed samples.Clustering of Bacterial CommunitiesSimilarities between bacterial communities in terms of their phylogeny were assessed by measuring UniFrac distance [34]. This metric takes into account the relative abundance of OTUs (weighted data) or their presence/absence (unweighted data). The between-method weighted UniFrac distance was significantly higher (P,10215) than the within-method distance across all pipelines (Figure 3A). The within-method distances for the mechanical and enzymatic lysis methods were not significantly different. Comparisons of bacterial community membership using unweighted UniFrac (Figure 3B) showed that intra-method distances were significantly smaller for the mechanically-treated samples than for the enzymatically-treated samples in the pipelines in which OTUs were clustered using Mothur (Pipelines 1/5, 2 and 3). In these pipelines, as well as in Pipeline 4, UniFrac distances between the mechanically-lysed samples were also smaller than inter-method distances. In contrast, in Pipeline 6, a statistically significant decrease was observed when unweighted UniFrac distances among enzymatically-processed samples were compared with inter-method distances. Permanova of Bray-Curtis similarities confirmed the effect of the extraction method on the structure of the salivary microbiota (Table S3). This test was based on square-root transformed abundance of OTUs and did not take into account the phylogenetic distance, as this was the case when using UniFrac. Similarly, PCoA of the Bray-Curtis similarity matrix based on square-root-transformed OTU abundance showed that samples processed using the same extraction method clustered together (Figure 4). We used this non-phylogenetic approach to analyze smaller size libraries simulated by sampling a given number of OTUs from each sample. The PCoA plots show that as few as 100 randomly chosen OTUs per sample results in a clear separation of bacterial community profiles according to the extraction method (Figure 4). Further reducing this number to 50 still showed the between-method separation, although with some overlap.Taxonomic AssignmentsIn accordance with previous studies [32], seven phyla (Proteobacteria, Firmicutes, Bacteroidetes, Fusobacteria, Actinobacteria, TM7, and Spiroch.
