The theoretical distribution of the DNA standard shows that the Pseudomonas OTU was expected to have the lowest 16S relative abundance while the Bacillus OTU was expected to be the highest. The experimental results of sequencing, shown in the first 2 bars of Figure 47c, reflected the theoretical distribution in a few important aspects: the Pseudomonas OTU had the lowest relative abundance; the Bacillus and Staphylococcus OTUs were also abundant, with Bacillus being somewhat more abundant; the Enterococcus, Escherichia, and Salmonella OTUs were comparably abundant, though this held more true for Plate 2 than for Plate 3. The greatest deviations from the theoretical distribution in the sequencing results lay in the visibly increased abundance of the Escherichia OTU and the decreased abundance of the Lactobacillus OTU. Since we confirmed the lack of external contamination and were not expecting these organisms to appear in the cultures, the Escherichia genus being associated mostly with the gut microbiota and the Lactobacillus genus being associated largely with fermentation and dental lesions and caries, the differences between the theoretical and experimental relative abundances of these two particular OTUs were not of great concern. On the other hand, big plastic pots the observed distributions for the microbial standard deviated more from the theoretical distribution .

The Salmonella and Escherichia OTUs were by far the most abundant, with the Enterococcus OTU coming in third. The other five OTUs were much less abundant than in the expected distribution, with the possible exception of the Pseudomonas OTU, which seemed to be equally or even slightly more abundant. The Lactobacillus and Listeria OTUs were by far the most underrepresented in the microbial standard compared to the theoretical composition. The deviations between the extracted microbial community and the theoretical 16S distribution reflect how different types of bacteria respond differently to extraction: Many commercial kits extract DNA from Gram-negative bacteria, such as those belonging to the Escherichia and Salmonella genera, much more efficiently than they do from Gram-positive bacteria. Clearly, the bias came from different extraction efficiencies because the DNA standard yielded distributions not so different from the theoretical one, hence making the amplification and sequencing processes unlikely to be the source of these large biases toward Gram-negative organisms. In future experiments, possibly ones that require examination of diseased dental plaque, we would need to find an alternative extraction kit or process to minimize bias toward Gram-negative bacteria, including those that have been associated with dental caries and periodontal diseases such as organisms in the Porphyromonas, Trepomona and Tannerella taxa. We may also need to examine and correct for amplification/sequencing biases that affect the Lactobacillus taxa because of its implicated role in dental disease.

However, the DNA extraction bias was not a predominant issued for the preservation experiments because the organisms that would have been inaccurately labeled as abundant were absent from the cultures, which contained mainly Gram-positive bacteria from healthy hosts. In this case, sequencing the mock communities helped validate the amplification and sequencing processes while providing valuable information on some potential biases to be avoided in the future. As an initial validation of the microbial community results, we investigated which taxa were most prevalent in the sequencing reads . These included Actinobacteria, Bacteroidetes, Firmicutes, Fusobacteria, and Proteobacteria, which rank among the taxa shown in previous research to be the most prevalent in the oral microbiota, confirming the validity of the experimental and bio-informatics procedures.To investigate whether a correlation existed between read depth and diversity, we plotted three diversity indices – Shannon index, inverse Simpson’s index, and richness – against the number of reads per sample. We saw no correlation for the Shannon or inverse Simpson’s indices. However, a statistically significant weakly positive correlation was observed between read counts and richness . These results suggest that observed OTUs would increase with increasing sampling depths such that samples with higher read counts might yield apparently higher diversity that would really be artifacts of the higher sequencing depth. This inflation would interfere especially seriously with the preservation experiment because of the widely varied sequencing depths of the samples.

To address this problem, we examined the number of species-level OTUs before and after rarefying to the lowest number of read counts to investigate whether rarefaction could lead to more accurate comparisons. We observed a decrease in the number of OTUs for rarefied samples with more than 240 reads before rarefaction. This decrease was substantial for some samples ; for instance, in some of the microbial community standards , the number of OTUs decreased from 100 to approximately 10 after rarefaction. Given that only 8 organisms were expected from the community standards, rarefying to 240 reads yielded a number of OTUs that matched expected results much more closely. Furthermore, the OTUs thus identified after rarefaction match those expected from the theoretical composition . Therefore, we opted to use rarefaction as a corrective measure for accurately estimating the number of OTUs and making valid comparisons among samples.We examined the compositional differences of the original plaque, initial cultures, preserved cultures, and propagated cultures by plotting the relative abundances of the OTUs in these samples . Plaque samples from all three hosts included two OTUs from the Actinomyces genus, and at least one OTU from each of the Corynebacterium, Rothia, Streptococcus, and Veillonella genera. Host 2 plaque composition differed substantially from that of Hosts 1 and 3 in terms of evenness. In Host 2, Streptococcus OTU001 represented the single most dominant OTU, and Fusobacterium OTUs were absent, whereas Streptococcus OTU005 was much more prominent than OTU001 in Hosts 1 and 3 and a Fusobacterium OTU was present in these two hosts. Host 1 and Host 3 plaque also contained distinctively higher abundances of the Actinomyces, Corynebacterium, and Rothia OTUs than Host 2 plaque while the Veillonella OTU was more abundant in Host 2 than in Hosts 1 and 3. The Prevotella OTU seemed to only appear in Host 1, and the OTU from the uncultured F0332 genus was present only in Host 3 plaque. Even though these sampleshad low biomass, the sequencing process was clearly able to capture some key host-based differences. After anaerobically incubating plaque-inoculated SHI media in duplicate wells for 72 hours, we generated communities that we termed “initial cultures”. We observed that the community composition had changed in all three hosts compared to the composition of the plaque. The diversity of the community decreased in all cases. Members from the Actinomyces, Corynebacterium, F0332, Fusobacterium, and Rothia taxa were no longer observed; those from the Prevotella, Streptococcus, and Veillonella taxa became dominant . The composition of the Host 1 initial culture diverged from those of Hosts 2 and 3: Veillonella OTU002 was the most abundant in the Host 1 culture while Streptococcus OTU001 had clearly begun to dominate Host 2 and Host 3. Furthermore, the presence of Prevotella OTU006 remained in one duplicate of Host 1 but not in the other. An interesting but unexpected increase in the relative abundance of an Alloscardovia OTU was also observed in Host 1 after the 72-hour incubation, though only in one well . Generally, for all hosts, the initial culturing step visibly reduced the number of organisms in the plaque community, selecting for a small number of taxa. However, the selection was apparently not biased toward a rigidly defined set of organisms, as seen in the differential compositions of the initial cultures from different hosts despite identical culturing conditions. When the initial cultures were subjected to four different preservation conditions, the dominant taxon remained dominant . In Hosts 2 and 3, the high abundance of Streptococcus OTU001 was maintained while in Host 1, the high abundance of Veillonella OTU002 was maintained, growing berries in containers regardless of preservation conditions. The membership of the initial cultures of all three hosts also stayed the same regardless of preservation conditions. Preserved cultures in Host 1 still contained four major OTUs, including two Streptococcus OTUs and one Veillonella that were present in both initial culture wells and two other OTUs specific to individual wells ; preserved cultures in Host 2 and Host 3 consisted of two Streptococcus OTUs and one Veillonella OTU, as the initial cultures did. However, the relative abundances of these taxa changed for all three hosts and all four preservation conditions, to varying degrees.

In all hosts, preservation led to a decrease in the relative abundance of Veillonella OTU002 except for the three-day condition in Host 2. The minor Streptococcus OTU , OTU005, also seemed to decrease in relative abundance in most cases, though this decrease was not as consistent as the decrease in the Veillonella OTU, possibly because of the inevitable shifts of relative abundances with regards to each other. In Host 1, the relative abundance of the Prevotella OTU seemed affected only in the case of cryopreservation for one-and-a-half weeks while the relative abundance of the Alloscardovia OTU decreased for all preservation conditions. In Hosts 2 and 3, preservation led to a general maintaining or increasing of the abundance of Streptococcus OTU001 . When the preserved cultures were used to inoculate fresh media and then incubated anaerobically , both the membership and the relative abundances changed in all three hosts. In Hosts 2 and 3, Streptococcus OTU001 became unequivocally dominant in these propagated cultures, and in some cases, it became the only visibly present OTU. Prevotella continued to be absent from samples derived from Hosts 2 and 3; Streptococcus OTU005 largely disappeared from Host 2 samples after propagation while the presence of Veillonella OTU002 diminished in many propagated cultures in Host 3. In Host 1, propagation led to more variable outcomes. In most propagated samples, the relative abundance of Veillonella OTU002 decreased while that of Streptococcus OTU001 or Prevotella OTU006 increased. In addition to the greater variability seen in propagated Host 1 cultures, preserved and propagated Host 1 cultures appeared to be more diverse than those of Hosts 2 and 3, containing taxa such as Alloscardovia OTU011. Interestingly, the abundance of Alloscardovia OTU011 increased in many samples in Host 1, but not in all samples derived from the pellet with visible presence of the Alloscardovia.Overall, propagation of the preserved cultures tended to further reduce diversity, with communities shifting toward the Streptococcus, Veillonella, and occasionally Prevotella taxa.To visualize compositional similarities and differences, we performed Principal Coordinate Analysis with Bray-Curtis dissimilarity distances . In the preservation experiment, the first two principal coordinates explained more than 97% of the total variation in the samples. The most striking feature in the PCoA plot was that samples from Host 2 and Host 3, respectively, formed single clusters at low values of the first coordinate while samples from Host 1 formed multiple distinct clusters. Of the Host 1 clusters, two clusters – one containing a few propagated cultures and the other containing the initial and preserved cultures – did not appear at the position that the Host 2 and Host 3 clusters occupied while the third cluster – containing most of the propagated cultures – did. The similarity of the overlapping samples indicates that propagating the preserved cultures tended to produce a compositionally invariant community across samples obtained from different hosts, and this composition appears to act as an ‘attractor’ state for a more initially diverse community as the community undergoes in vitro culturing, preservation, and propagation. The two Host 1 clusters that did not overlap with Hosts 2 and 3 implicate that preservation without propagation did not produce communities with the ‘attractor’ composition. Furthermore, some propagated replicates gave distinct compositions.To determine the taxa with the highest contributions to the variation in the dataset and to the composition of the attractor cluster, we performed Principal Component Analysis on the relative abundances of the initial, preserved, and propagated cultures . The first principal component accounted for more than 93% of the total variance, containing large and opposing contributions from Veillonella OTU002 and Streptococcus OTU001. The second component accounted for about 6% of the total variance, with a major contribution from Prevotella OTU006. The ordination plot illustrates a cluster characterized by high Streptococcus OTU001, low Veillonella OTU002, and low Prevotella OTU006, and two other smaller, looser clusters, both characterized by high Veillonella OTU002. As expected from PCoA and the dot plot, most Host 1 samples sit separately from Hosts 2 and 3 samples, particularly along PC1, indicating that Veillonella OTU002 is strongly characteristic of Host 1 samples .