Host 1 samples also exhibited some variation in the presence Prevotella OTU006, whereas Host 2 and Host 3 did not. Concern over the validity of PCA on relative abundance data prompted us to perform PCA again on the transforms of the relative abundance data. For this part of the analysis, we used both the centered-log-ratio transformation and the isometric log-ratio transformation. Figure 31 shows the PCA results on the CLR transform of the non-rarefied relative abundances. Compared to the clustering before CLR transformation , samples Hosts 2 and 3 spread out much more while samples from Host 1 clustered together more tightly . Color-coding samples by preservation conditions revealed no apparent trend , as was the case without CLR transformation . The Veillonella OTU still contributed greatly to the separation of Host 1 samples from Host 2 and Host 3 samples. However, after CLR transformation, Prevotella OTU006 appeared much more influential in accounting for the differences between Host 1 and Hosts 2 and 3. PCA on the ILR transformation of relative abundances yielded similar groupings, albeit with different scaling and different directions for the individual components . In the PCA results of the ILR transform, two components helped account for the differences observed between Host 1 samples from Host 2 and Host 3 samples while other components accounted for most of what was left of the sample variation. In the PCA results of both the ILR and CLR transformations, the first two principal components together accounted for 57% of the total sample variation.

As an additional quantitation of the extent of the influence exerted on the sample compositions by different preservation conditions and different hosts, blueberry containers we conducted ANOSIM significance tests with Bray-Curtis distance measures. We found that the relative abundance differences were not greatly influenced by preservation condition, as evidenced by the low correlation coefficients of 0.05505 and 0.1287 for non-rarefied and rarefied relative abundances, respectively. The compositional differences, as we had already observed in PCoA and PCA, were apparently influenced to a much greater extent by host differences, with correlation coefficients of 0.3365 and 0.3147 for non-rarefied and rarefied relative abundances, respectively. Hence, we numerically confirmed that host-based variation contributed the most to the observed differences across all samples.In the last two decades, more and more effort has been devoted to researching the effect of various preservation methods on complex microbiome samples. Particularly close attention has been paid to the gut microbiota, as the therapeutic potential of faecal microbiota transplantation has increased. Many ways of preserving both natural and artificial human gut microbiota have been investigated, including cryopreservation, lyophilization, and long-term storage in commercial storage media. Unlike the gut microbiome, however, the preservation of the oral microbiome does not seem to have received nearly as much attention despite our increasing understanding of its crucial role in human health and disease. Until five years ago, not much has been explored regarding the effects of storage methods on the stability of native human oral communities, let alone the stability of in vitro models.

It was one of our goals, in this set of experiments, to begin probing the effects of refrigeration and glycerol-assisted cryopreservation on communities derived from healthy hosts and generated in an in vitro environment. For this set of experiments, we chose an incubation time of 72 hours based on the results from the temporal experiments, where we observed a transition from the dominance by Streptococcus OTU genus to the dominance by the Veillonella genus. This incubation time seemed a good middle ground for capturing as many core members of the native oral bacterial community as possible without excessive internal contamination. Unlike in the temporal experiments, however, the relative abundances of the 72-hour cultures were not inclined toward Veillonella OTUs except for Host 1 . In fact, the Streptococcus OTU remained dominant in Hosts 2 and 3 throughout the preservation and propagation processes while dominance in Host 1 samples oscillated between Streptococcus and Veillonella OTUs except for one set of propagated samples that contained much higher proportions of Prevotella than others . It is not entirely clear whether preservation monotonically decreased or increased the relative abundance of any singleOTU. What is clear is that the combination of culturing and preservation procedures seemed to drive the community toward what we termed an “attractor” composition unless a substantial presence of the Veillonella genus already existed. In cultures with visible Veillonella presence, the relative abundances after preservation and propagation varied quite greatly, even within the same host and same preservation conditions. In terms of the effect of preservation on community composition, we observed that preservation alone did not lead to drastic changes in the relative abundances of the initial culture for any host. Members of the Streptococcus genus seemed to respond particularly well to glycerol-assisted cryopreservation as well as refrigeration, evidenced by the relatively minor changes in the abundances before and after preservation .

The Veillonella OTUs seemed to respond less well, as their relative abundances decreased upon propagation . Perhaps members of the Veillonella genus are less robust toward environmental changes, and the consequent decrease in the absolute biomass of the Veillonella OTUs in these experiments helped emphasize the increase in the relative abundances of Prevotella and Streptococcus OTUs. In any case, we clearly see in Figure 29 that in all hosts, community compositions pre- and post-preservation were remarkably similar. The preservation conditions we chose helped retain a substantial quantity of how OTUs were distributed in each sample. Thus, these conditions would be valuable for assessing community compositions in experiments where immediate processing may not be possible. On the other hand, the propagation of preserved cells seemed to preferentially select for Streptococcus OTUs, perhaps because this genus already occupied somewhat high proportions of initial cultures. Furthermore, it seemed that at least in Hosts 2 and 3, the Veillonella OTU did not respond as robustly to preservation as the Streptococcus OTU, hence contributing to the increase in relative abundance of at least one Streptococcus taxon. Another very plausible explanation for the shift to the Streptococcus genus is thatthe propagation was simply not long enough – we chose to incubate the preserved samples for 48 hours instead of 72 hours like the incubation for the initial cultures, and the difference of 24 hours might have allowed us to observe a rise in the relative abundance of Veillonella in the propagated cultures. However, we cannot conclude that increasing incubation time would indubitably allow us to see such a shift, especially in light of the observation that the relative abundances of the Veillonella OTUs had already begun to increase noticeably by the 48-hour mark in the temporal experiments for all three hosts . By contrast, only the propagated cultures in Host 1 showed observable presence of Veillonella OTUs, and only one pellet from the 1.5-week cryopreservation retained the presence of this genus. The differences between the relative abundances of the initial/preserved cultures on one hand and the propagated cultures on the other imply that different taxa respond differently to preservation, i.e. the number of viable cells after preservation differs for different OTUs, even if the cells were still intact and their DNA could be extracted. It is also possible that a few procedural changes contributed to the absence of Veillonella, including the step aspiration of approximately 1.5mL liquid from the wells during incubation, which we only introduced into the preservation experiments and not into the temporal experiments. This step, an attempt to minimize internal contamination, could have essentially removed a means by which the sedimented culture was re-inoculated during incubation. These factors and more may be worth investigating in future experiments should we aim to produce propagated communities with compositions that would be similar to those of the initial and preserved communities.

As for the composition of the attractor community and its relationship with different OTUs, two Streptococcus OTUs and one Veillonella OTU seemed to sit at the center of the attractor. Interestingly, the Prevotella and Alloscardovia taxa persisted through both preservation and propagation in Host 1, implicating their roles in a different attractor composition, perhaps one that is more developed than the attractor observed in Host2 and Host 3. The presence of these two taxa may hold special significance for human health given that members of both taxa have been linked to diseased states in the oral cavity. Perhaps the composition of the attractor community changes as the environment is primed for later colonization, potentially by pathogenic species. There has certainly been evidence that organisms of the Prevotella genus may be dependent on other organisms such as those in the Fusobacterium genus, which have also been shown to coaggregate with Veillonella and implicated in oral diseases. Whether the taxa unique to Host 1 cultures would truly compose part of a developed/separate attractor community would need to be investigated further in future experiments. As for the principal components that contribute to the total variation in the data set, best indoor plant pots neither of the log-ratio transformations eliminated underlying biological correlations or averaged out real biological differences. What the transformations did was mitigating some of the positive bias seen in the PCA results of untransformed relative abundance data. The log-ratio transformations yielded PCA results similar in kind to those from PCA of the untransformed relative abundance data. The separation of Host 1 samples from Hosts 2 and 3 persisted across both transformations, and the major components that contributed to host-based differences – the Streptococcus, Veillonella, and Prevotella OTUs – remained mostly the same before and after transformation. However, the degree of separation was much diminished post-transformation, and contributions from smaller but still important components, such as the Alloscardovia OTU and an additional Streptococcus OTU, surfaced upon transformation . Furthermore, the CLR-transformed PCA clearly showed that preservation conditions did not fundamentally influence compositional differences, whereas this lack of influence was not entirely evident in the untransformed PCA. Rather than preservation conditions, it may be the differential organismal responses to these conditions that exerted the greatest influence on sample variation, and the differential responses may well be connected to microbial interactions – perhaps ones similar to those between Veillonella and Strepto-coccus species or between Streptococcus and Actinomyces species – that affect the robustness of an organism toward low-temperature, desiccation, or nutrient depletion stresses, such as those occurring during preservation. The interaction-related responses would then fundamentally depend on the community composition before preservation, just as salivary Veillonella species depend on a specific strain of Streptococcus for coaggregation. We will attempt to investigate composition-based differential responses to preservation in the next phase of this project. Returning to a point made in a previous section, we observed in the relative abundances of the mock microbial community that the DNA extraction process seemed to generally favor Gram-negative bacteria , particularly E. coli and S. enterica, while the sequencing process seemed to favor B. subtilis at the cost of P. aeruginosa. These results underscore the importance of choosing proper bacterial strain should we ever revisit bacterial cell spike-ins for quantitation purposes. Ideally, the spikein organism would be non-oral, Gram-positive, and related to neither B. subtilis nor P. aeruginosa. Furthermore, even if we do not use a spike-in, we should strive to understand the biases in the extraction, amplification, and sequencing steps for different oral microbes. We may need to start the process by examining the extraction efficiencies of single-cell cultures, or in the absence of such a possibility, of co-cultures with known or easily characterizable strains. The results of these efficiencies could then be used to mathematically correct for relative abundance data, though ensuring the validity of this approach requires extensive proof of repeatability from one extraction-amplification sequencing trial to the next. The current dearth of research regarding the preservation of oral microbes may have originated from a perceived lack of need. Since facile identification of oral microbes has been difficult until high-throughput sequencing became viable, storing complex microbial communities for model-building and future study would not have been a reproducible or efficient approach. One of the few examples of examining the effect of preservation on oralbacteria compared both storage and transportation methods for the human supragingival dental plaque. The results showed that freezing dental samples in transport media without cryopreservation reagents led to no substantial differences between 48- and 72- hour storage for either S. sanguinis or S. mutans, though the survival rates of viable bacteria in frozen samples were predictably much lower than the those in samples stored at temperatures above freezing.