Common core microbial species in subgingival plaques
Overall, bacterial communities were found to be very specialized in the subgingival plaque samples. After preprocessing and profiling, an average of 47, 31 and 34 microbial species were detected in healthy, stable and progressing samples, respectively. This indicates a relatively small number of species when compared to all oral microbial phyla. The numbers are consistent with those of previous reports which found as few as 50 predominant species in subgingival plaques, irrespective of health or disease [40, 41]. These results showed that the subgingival plaques sampling procedure was carefully and conservatively performed to avoid possible contamination from the general oral environment. Based on the fact that such number has not substantially changed between their studies and ours, we conclude that the high-throughput, culture-independent methodology faithfully preserves the aboundance structure, even though it is now much more sensitive to the heterogeneity of microbiotas resident in host oral cavity.
Based on phylogenetic analysis (see Fig. 2), we identified such predominant microbial species as Streptococcus gordonii, Streptococcuss anguinis and Lactobacillus gasseri, which are consistent with those identified by Aas et al. and Paster et al. in subgingival samples [40, 41]. Since the oral cavity is the main portal through which most microorganisms enter human bodies, it is possible to detect many transient microbes in the oral environment through metagenomic techniques. Nonetheless, only a few core microbes were consistently found to inhabit subgingival plaques in both this study and those of A as et al. and Paster et al. [40, 41]. These results strongly suggest that periodontitis is induced by inflammatory response to bacterial challenge from the core microbes detected in subgingival biofilm [42]. Thus, the catalogue of these core microbes that persist in subgingival biofilms appears to represent the repertoire of pathogens responsible for disease onset.
In our analysis, we relied on reference genome and read mapping for composition and relative abundance estimation. It is possible that some rare species inhabiting in subgingival plaques were missed out due to low coverage of sampling procedure, low depth of read sequencing, mapping error and other random factors. However, these species’ abundance should be very low even if not truly zero. In this particular study, the microbial species with a relatively high abundance are more likely to be pathogen, because periodontitis is an inflammatory disease that human immune system have active confront with microbes in subgingival plaques. Since our statistics are mainly comparing highly differentiated species, zero abundance levels due to dropouts should not have an effect. In addition, we used standardized bioinformatics pipeline to avoid bias and to estimate the microbial abundance level as accurate as possible. Almost all of the reads got mapped to the provided reference set and therefore there is not much presence of de novo species. That is because human oral microbiota has been extensively studied by clone and culture sequencing in decades, which have generated a very comprehensive set of reference sequences.
Highly abundant microbial species in subgingival plaques
Among the microbial species discovered in healthy and periodontitis subgingival samples, the genus of Streptococcus was found in relative abundance. Many species of Streptococcus, such as Streptococcus gordonii, Streptococcus oligofermentans and Streptococcuss anguinis, were among the ten most prevalently abundant microbes at all status of periodontitis. This result suggests that pathogens from the genus Streptococcus may be among the most successful early colonizers to clean tooth surfaces in the human mouth by their adherence and metabolic capacities [43]. Based on their predominance in healthy samples, but decrease in periodontitis samples, their colonies might also serve as a source of biofilm adhesion for other colonizers [44].
Rothia dentocariosa was also found in high abundance in both healthy and periodontitis samples. In two of the progressing periodontitis samples, it held top abundance rank at 70.4% and 16.8%, respectively, as well as in stable periodontitis samples with relative abundance of 56.2% and 32%, respectively. Although Rothia species are often associated with oral health, these results are consistent with previous studies, which found that Rothia spp. can reduce oxygen levels around biofilm thus promoting the proliferation of inflammation-triggering anaerobes [17, 45].
Species like Atopobium parvulum, Lactobacillus gasseri, and Staphylococcus epidermidis are highly abundant in stable and progressing subgingival samples, and many of them have already been associated with periodontitis. The Atopobium genus, which is high in G + C-content and gram-positive, has previously been identified as prevalent in individuals with periodontitis, but not in healthy subjects. Lactobacillus was also found at high percentage in severe periodontitis subgingival samples [46], while Staphylococcus genus have only recently been identified as pathogens associated with periodontitis [47]. Our results further strengthen those findings.
Streptococcus mutans was also relatively abundant in our subgingival samples. This is particularly interesting because previous observation has shown that this microbial species can create a lactic acid-rich environment, a condition conducive to the development of symbiosis, thus making it possible for other pathogens to thrive [48].
Prevalent keystone species in healthy and periodontitis subgingival plaques
The keystone species Porphyromonas gingivalis previously identified by Orth et al. [19] was found in both healthy and diseased samples, albeit in rather small relative abundance, again hinting that this species may play a role as a keystone pathogen. A keystone pathogen typically remains functional in the background at low abundance, but promotes inflammatory state by altering host immune response to otherwise commensal microbiota. Thus, the existence of Porphyromonas gingivalis alone does not cause periodontitis. Rather, if accompanied by proper commensal microbiota, such as species from Lactobacillus, Staphylococcus and Streptococcus genera [49], in susceptible individuals, Porphyromonas gingivalis may be able to evade immune surveillance and mediate the breakdown of normal homeostatic balance in the oral environment [39].
Interestingly, Fig. 4 shows that Porphyromonas gingivalis was clustered in a small group with another three species: Haemophilus haemolyticus, Prevotella melaninogenica and Capnocytophaga ochracea. They shared similar patterns with Porphyromonas gingivalis, a widely accepted keystone pathogen, in their abundances in healthy and periodontitis samples, which indicates that these three species are likely to play a similar role similar to that of P. gingivalis in the oral community by significantly altering the number and community organization of oral commensal bacteria, possibly leading to periodontitis under specific conditions in susceptible bodies. However, more evidence and experimentation are needed to confirm this.
Differential representation of marker species between healthy and periodontitis subgingival plaques
Marker species are highly differentially represented in healthy and diseased subgingival plaques. As shown in Fig. 3, nine microbial species represent significant difference in relative abundance among the samples studied. In particular, Lactobacillus gasseri and Osenella uli were found to be significantly more abundant in periodontitis samples. Campylobacter showae was observed with significantly higher relative abundance in stable samples. In addition, six microbial species had significantly higher relative abundance in healthy samples compared to periodontitis samples, including Gemella morbillorum, Streptococcus sanguinis and Veillonella parvula, which were observed both in healthy and periodontitis samples, and Haemophilus parainfluenzae, Corynebacterium matruchotii and Neisseria flavescens, which were only detected in healthy samples. The p-values of their Dunn tests are shown in the Results section and Fig. 3.
These differentially represented species in subgingival plaques most likely compose the core microbiota directly associated with the etiology of periodontitis. As such, they could be considered markers of periodontitis. As shown in Fig. 4, when we used these nine marker microbial species as representatives of collected samples and clustered the samples based on Spearman correlations of their marker species profiles, the clustering result was nearly the same as their clinical classification (accuracy = 94.4%). This evidence strongly suggests that the combination of differentially abundant species is predictive of periodontitis at different states.
Moreover, unlike pathogen-specific diseases, the etiology of periodontitis is polymicrobial in nature and thus indicative of characteristics of the whole microbial community, or, at least, its core microbiota. It should be noted that the clustering was based on Spearman Rank-Order Correlation, rather than Euclidian distance, which gave us results consistent with clinical classification. This, in turn, means that relative hierarchy of microbial species is highly correlated with progressing periodontitis. Therefore, future studies should look for particular profile patterns of microbial composition that specifically corresponds to different states of periodontitis progression.
Ecological diversity as a prognostic marker of periodontitis disease
We then focused on the alpha-diversity of samples and its association with the progressing periodontitis. Figure 3 shows a number of microbial species with particularly high outlier abundances in both stable and progressing periodontitis subgingival samples, indicating ecological imbalance among host oral communities. When we compared alpha-diversity of samples between periodontitis and healthy samples, we found that healthy samples had the highest alpha-diversity, while stable periodontitis samples had the lowest with significant difference (see Fig. 3). Furthermore, the fitted univariate logistic model demonstrates the potential value of oral microbial alpha-diversity as a predictor of periodontitis.
Polymicrobial synergy and dysbiosis (PSD) model of periodontitis etiology
Based on recent metagenomics findings, Orth et al. proposed a Keystone-Pathogen Hypothesis which holds that keystone microbes like P. gingivalis may coordinate the onset and progress of periodontitis and these microbial species can alter host-microbial interactions in a manner that interrupts the homeostasis of microbial colonizers and results in inflammatory conditions that set the stage for the development of periodontitis [50]. Our results support this theory in the following way.
According to the PSD model, periodontitis is not caused by a specific microbial species. Instead, inflammation is thought to arise from immune response to the combined activity of core and keystone pathogens. In fact, our phylogenetic profile analysis identified core pathogens and marker species showing significant abundance shifts between healthy and periodontitis samples. For example, marker species Lactobacillus gasseri, Campylobacter showae and Olsenella uli was more abundant in periodontistis samples, while the marker species Gemella morbillorum, Veillonella parvula, Haemophilus parainfluenzae, Corynebacterium matruchotii, Neisseria flavescens, Campylobacter showae and Streptococcus sanguinis were significantly more abundant in healthy samples.
We also demonstrated potential keystone species as mediators of this process (see Fig. 6). Based on the PSD model, keystone species gain virulence through cooperating with accessory microorganisms to strengthen the biofilm by creating a powerful synergistic community. Meanwhile, however, previously commensal biofilm species loses homeostasis and drop out from the competition [39, 49, 51]. Indeed, we identified the keystone species Porphyromonas gingivalis, Haemophilus haemolyticus, Prevotella melaninogenica and Capnocytophaga ochracea, all prevalent at relatively low abundance across all samples and likely acting as mediators of the observed shifts in abundance profiles between periodontitis states [49].
Thus, while these keystone species gain virulence from pathogenic associations with, for example, Lactobacillus gasseri and Osenella uli, which show relative abundance and thrive in the shifting microenvironment, other species lose ground and disappear from the oral environment, such as Haemophilus parainfluenzae, Corynebacterium matruchotii and Neisseria flavescens. Others may decrease significantly in relative abundance, such as Gemella morbillorum, Streptococcus sanguinis and Veillonella parvula. Taken together, we found Shannon diversity for all but one healthy individual to be > 2, similar to the statistics for 10 healthy individuals reported by Bik et al. using 16S RNA sequencing data [52]. On the contrary, almost all periodontitis patients have Shannon diversity < 2 [49]. Importantly, using alpha-diversity as an indicator of microbial flux in a state of dysbiosis, we see a decreasing pattern in periodontitis samples in accordance with the PSD model [51].
The results of our network analysis also support the PSD model and the effects of decreased diversity. The microbial species (nodes) in periodontitis samples have much less interconnectivity (edges), in particular negative associations (blue dashed edges), representing a loss of check-and-balance. There are subnetworks consisting of > 5 correlational microbial species found only in periodontitis samples indicating the pathobionts and keystones thriving through synergistic interaction in the inflammatory state, as suggested in the preceding section. The component species in these subnetworks are different, indicating that the keystones and pathogens active in different states of periodontitis may be different. At this point, the alpha-diversity of subgingival samples has also reached the lowest level.
Implications of reduced ecological diversity in periodontitis
Irrespective of specific etiology, this and other studies have found oral microbial alpha-diversity to be strongly negatively correlated with periodontitis status [50].
Mutualism and commensalism are hallmarks of diversity among the aggregate of microorganisms resident in human, and such diversity is generally indicative of health [53]. This principle has been tested under conditions of dysbiosis in mucosal diseases of the gastrointestinal tract, such as inflammatory bowel disease and obesity, both considered to be associated with decreased diversity of microbial community [54]. Some researchers, however, have reported that the diversity of oral microbiota is higher in periodontitis compared to healthy subjects [55, 56].
We offered evidence in the present metagenomics study indicating that the principle microbiome diversity is also applicable to periodontal health. Indeed, a healthy oral microbiome is generally reflective of bacterial homeostasis, featuring both synergistic and antagonistic interactions between microbes and between microbes and host [14]. From our results, however, it is likely that polymicrobial synergy eventually leads to dysbiosis by the breakdown of ecological balance in periodontal microbial communities and that the disease state follows.
Accordingly, our results could potentially offer a novel diagnostic and prognostic paradigm based on establishing classification models to distinguish among different states of periodontitis using alpha-diversity and species composition as metrics, together with a sufficient database. However, since the results of the present study cannot pinpoint community diversity as either a cause or result of periodontitis, further study is required to determine whether alpha-diversity of subgingival plaques can, indeed, predict incipient periodontitis in individuals whose clinical symptoms have not yet manifested.
It should be noted that our study was limited in sample size and that we only used average abundances in network construction to reflect the general scope of the oral community in subgingival plaques under different states of periodontitis. However, if longitudinal metagenomics data were to be collected for each periodontitis subject in a larger cohort study, individual oral networks could be constructed, and the underlying correlations among those networks under particular states of periodontitis could be detected. This would provide a better understanding of the etiology of periodontitis.