Perio Part II
Perio Part II
Perio Part II
crossm
Population Structures (VAMPS) pipeline, to identify microbiota specific to health and Editor Andrew J. McBain, University of
Manchester
disease. Microbiota were visualized using CoNet and Cytoscape. Dysbiosis ratios, de- Copyright © 2017 American Society for
fined as the percentage of genera associated with disease relative to the percentage Microbiology. All Rights Reserved.
of genera associated with health, were calculated to distinguish disease from health. Address correspondence to Vincent Meuric,
Correlations between the proposed dysbiosis ratio and the periodontal pocket depth vincent.meuric@univ-rennes1.fr.
F.B.-H. and M.B.-M. contributed equally to this
were tested with a different set of data obtained from a recent study, to confirm the
work.
relevance of the ratio as a potential indicator of dysbiosis. Beta diversity showed sig-
nificant clustering of periodontitis-associated microbiota, at the genus level, accord-
ing to the clinical status and independent of the methods used. Specific genera
(Veillonella, Neisseria, Rothia, Corynebacterium, and Actinomyces) were highly preva-
lent (⬎95%) in health, while other genera (Eubacterium, Campylobacter, Treponema,
and Tannerella) were associated with chronic periodontitis. The calculation of dysbio-
sis ratios based on the relative abundance of the genera found in health versus peri-
odontitis was tested. Nonperiodontitis samples were significantly identifiable by low
ratios, compared to chronic periodontitis samples. When applied to a subgingival
sample set with well-defined clinical data, the method showed a strong correlation
between the dysbiosis ratio, as well as a simplified ratio (Porphyromonas, Treponema,
and Tannerella to Rothia and Corynebacterium), and pocket depth. Microbial analysis
of chronic periodontitis can be correlated with the pocket depth through specific
signatures for microbial dysbiosis.
IMPORTANCE Defining microbiota typical of oral health or chronic periodontitis is diffi-
cult. The evaluation of periodontal disease is currently based on probing of the peri-
odontal pocket. However, the status of pockets “on the mend” or sulci at risk of
periodontitis cannot be addressed solely through pocket depth measurements or
current microbiological tests available for practitioners. Thus, a more specific micro-
biological measure of dysbiosis could help in future diagnoses of periodontitis. In
this work, data from different studies were pooled, to improve the accuracy of the
results. However, analysis of multiple species from different studies intensified the
bacterial network and complicated the search for reproducible microbial signatures.
Despite the use of different methods in each study, investigation of the microbiota
at the genus level showed that some genera were prevalent (up to 95% of the sam-
July 2017 Volume 83 Issue 14 e00462-17 Applied and Environmental Microbiology aem.asm.org 1
Meuric et al. Applied and Environmental Microbiology
ples) in health or disease, allowing the calculation of bacterial ratios (i.e., dysbiosis
ratios). The correlation between the proposed ratios and the periodontal pocket
depth was tested, which confirmed the link between dysbiosis ratios and the sever-
ity of the disease. The results of this work are promising, but longitudinal studies
will be required to improve the ratios and to define the microbial signatures of the
disease, which will allow monitoring of periodontal pocket recovery and, conceiv-
ably, determination of the potential risk of periodontitis among healthy patients.
KEYWORDS chronic periodontitis, health, microbiota, dysbiosis ratio
RESULTS
Microbial community structure analysis. Using a matrix correlation analysis, the
possible clustering of microbiota according to the nature of the primers used, the site
of sampling, or the study investigated was explored. Despite various studies, the
analyzed data clustered into five groups according to the clinical status (healthy or CP)
or the sampling site, as shown by the three-dimensional (3D) principal-coordinate
analysis (PCoA) plots (Fig. 1). Healthy subgingival samples were primarily spread into
two main clusters; control samples were clearly separated into two other clusters,
corresponding to saliva and dentine caries/vagina, while the majority of CP samples
FIG 1 Different views of 3D PCoA plots illustrating the beta diversity of bacterial populations as a function of sampling site and diagnosis. Light blue,
supragingival samples; dark blue, healthy subgingival samples; green, outgroups (saliva, midvagina, and dentine caries samples); red, CP samples.
Percentages represent percent explained variance.
were found in a fifth cluster. Two-dimensional (2D) beta-diversity analysis showed the
precise distribution of the samples in the five clusters (Fig. 2). The search for an
association between clusters and primers and/or study type showed that the fourth
cluster was associated with V3V4 16S rRNA primers (correlation r ⫽ 0.537; P ⬍ 0.001)
and with the study by Kianoush et al. (10), which used those specific primers (correla-
tion r ⫽ 0.608; P ⬍ 0.001). No other correlations with primers were found. The two
healthy clusters (clusters 1 and 2) were characterized by subgingival and supragingival
samples in similar proportions (Fig. 2, light blue and dark blue sections). Focusing on
healthy subgingival samples, the main difference between the two healthy clusters
(clusters 1 and 2) was in the distribution of samples from the HMP study and from the
other studies in the clusters; 225/323 samples from the HMP study were clustered in
healthy cluster 2, while healthy cluster 1 was richer in samples derived from the other
studies (44/99 samples). Cluster 3 was characterized by saliva, as 91% of the saliva
samples (258/284 samples) were grouped in this cluster (correlation r ⫽ 0.892; P ⬍
0.001) (Fig. 2). Cluster 4 was characterized by dentine caries samples (73% [80/110
samples]; correlation r ⫽ 0.603; P ⬍ 0.001) and midvagina samples (100% [60/60
samples]; correlation r ⫽ 0.638; P ⬍ 0.001). Finally, cluster 5 contained 90% of the
CP samples (176/196 samples; correlation r ⫽ 0.708; P ⬍ 0.001).
It is interesting to note that 10% of the CP samples were found in the two healthy
clusters (19/196 samples) and contained similar microbiota (analyzed on the basis of
beta diversity), at the genus level, as dentine caries samples and/or midvagina samples
(1/196 samples). Conversely, 16% of the healthy subgingival samples (69/422 samples)
and 17% of the dentine caries samples (19/110 samples) were found in cluster 5.
Microbiota richness and alpha diversity in subgingival samples. A cluster com-
parison showed that the sampling depth (number of reads sequenced) was greater in
healthy subgingival clusters 1 and 2 than in cluster 5. Nevertheless, no significant
difference between healthy subgingival samples and CP samples of cluster 5 was found
(Fig. 3). The observed richness (S) was lower in the CP samples of cluster 5 than in the
samples of healthy clusters 1 and 2 and the healthy subgingival samples of cluster 5
(Fig. 3). However, the Shannon-Weaver diversity index values showed that the diversity
of healthy cluster 2 was significantly greater than the diversity of healthy cluster 1 and
that of all samples from cluster 5, which were similar.
Patterns of microbial communities in subgingival samples (genus level). Gen-
era that were present in at least 95% of all healthy subgingival samples or 95% of
the CP samples from cluster 5 are presented in Fig. 4A and B, respectively. Results
showed that healthy subgingival samples were dominated by 8 major genera, i.e.,
Fusobacterium, Actinomyces, Streptococcus, Neisseria, Capnocytophaga, Prevotella,
Corynebacterium, and Rothia, and 6 minor genera, i.e., Leptotrichia, Veillonella,
Porphyromonas, Granulicatella, Kingella, and Gemella. Associations were found be-
tween Fusobacterium and Prevotella, Actinomyces, and Rothia and between Lepto-
trichia and Porphyromonas. Common genera found in CP samples were less abun-
dant, with 4 major genera, Treponema, Porphyromonas, Prevotella, and Fusobacterium,
followed by Streptococcus, Eubacterium, Tannerella, and Campylobacter genera. Only one
association was found, between Eubacterium and Treponema, while Fusobacterium and
Treponema presented a negative correlation.
Calculation of dysbiosis ratios of bacteria. The dysbiosis ratios of the genera found
mainly in CP samples (Eubacteria, Campylobacter, Treponema, and Tannerella) to the
genera found mainly in healthy samples (Veillonella, Neisseria, Rothia, Corynebacterium,
and Actinomyces) were significantly different among the samples according to their
diagnosis. The dysbiosis ratios for healthy subgingival samples (from the HMP, n ⫽ 323,
FIG 4 Patterns of subgingival microbial communities. (A) Patterns of genera present in at least 95% of all healthy
subgingival samples. (B) Patterns of genera present in at least 95% of all CP samples from cluster 5. Edges represent
1 (thin line) or 2 or 3 (thick line) significant correlations between genera (green, positive; red, negative). Node colors
represent the numbers of partners, ranging from 1 (green) to 7 (dark orange). Node sizes represent the abundance
of each taxon.
ratio of 0.016; from the other studies, n ⫽ 99, ratio of 0.021) yielded a median ratio of
0.018; the samples from shallow sites had a ratio of 0.071, and the CP samples had a
ratio of 1.229 (P ⬍ 0.001) (Fig. 5A).
Although different clustering was achieved through beta-diversity analysis, no
significant difference in the ratios of clusters 1 and 2 according to the clinical status
(healthy, shallow, or CP) was found. Pooling of samples according to clinical status was
performed, and the resulting ratios were compared to the ratios for cluster 5, as shown
in Fig. 5B.
The dysbiosis ratio found for CP samples from cluster 5 (ratio of 1.510) was
significantly greater than the ratios for the majority of samples from clusters 1 and 2
(healthy subgingival samples, ratio of 0.015; shallow samples, ratio of 0.052; CP samples,
ratio of 0.088) and was also significantly greater than the ratio for healthy subgingival
samples (ratio of 0.184) from cluster 5 (P ⬍ 0.001). In clusters 1 and 2, the dysbiosis
ratios for CP samples were similar to the ratios for shallow sites. These two groups were
significantly different from the healthy subgingival samples (P ⬍ 0.05) in the same
cluster.
Healthy subgingival samples (n ⫽ 69) belonging to cluster 5 exhibited a dysbiosis
ratio (ratio of 0.184) significantly different from that for the other healthy subgingival
samples (ratio of 0.015) and also from that for the majority of the CP samples (cluster
5) (ratio of 1.510). These results confirmed the possible difference of these healthy
subgingival microbiota (P ⬍ 0.001) from those of healthy clusters 1 and 2. Their ratio
was not significantly different from the CP sample ratios in clusters 1 and 2, which could
be considered “on the mend.”
Validation of the dysbiosis ratio. A different data set, from Bizzarro et al. (11) and
containing well-described samples (pocket depths of 2 to 8 mm), was used as an
external control to confirm the relevance of the bacterial dysbiosis ratio. The dysbiosis
ratio at the genus level was correlated with the periodontal pocket depth (r ⫽ 0.655;
P ⬍ 0.001) (Fig. 6A). These results, based on data for 37 patients (147 samples collected
at different times, with different procedures for periodontal treatment), confirmed the
link between dysbiosis and the depth of the periodontal pocket. The simplified ratio of
Porphyromonas, Treponema, and Tannerella to Rothia and Corynebacterium showed a
similar correlation (r ⫽ 0.659; P ⬍ 0.001) (Fig. 6B).
DISCUSSION
Many studies have been published since the Human Microbiome Project in 2009,
increasing the volume of microbiota data available for the research community. How-
ever, comparisons between studies are challenging, at least at the species level,
because of the use of different methods. This issue is a real limitation to understanding
disease, as is the small number of samples in each study. Additionally, findings are more
complicated for healthy subgingival samples, which usually represent less than one-half
of the samples included in the studies (12, 13). This work is a taxon-based analysis, at
the genus level, of sequence reads from several studies. Studying a large number of
samples minimized individual variations and overcame technical variations by increas-
ing the effective sample size. Such an analysis had already been proposed in a recent
study of the microbiota in obesity (8). Studies with described healthy samples (sulci of
ⱕ3 mm) and CP samples (pocket depths of ⱖ5 mm) and available raw sequence data
in data banks were chosen. Data from the HMP resources (two different pairs of primers
used) were added to increase the number of healthy subgingival samples with available
microbiota data from 99 samples to 422 samples. The different microbiota clustered
either by sampling site, such as the outgroups used as controls for this study (saliva
samples in cluster 3 and dentine caries and vagina samples, which are both rich in
Lactobacillus, in cluster 4), or by clinical status, such as subgingival samples (healthy
samples in clusters 1 and 2 and CP samples in cluster 5). CP sites either can show
greater microbial diversity and observed richness, compared with healthy subgingival
sites (14, 15), or can present no significant difference in microbial diversity, as reported
for health versus periodontitis (16). Thus, the large number of samples surpasses the
technical variations, at least at the genus level, with the primers used in the different
studies, and the difference between periodontal health and disease is larger than the
technical variations, as described by Kirst et al. (16). No difference between the healthy
subgingival and supragingival samples was found when beta-diversity analysis was
performed at the genus level, as described previously (17). Ninety percent of the CP
samples were found in cluster 5. To define cluster 5 as a “periodontitis cluster” by beta
diversity was appealing. However, cluster 5 also contained healthy subgingival samples,
indicating that further investigations are necessary to understand and to develop
prediction markers for chronic periodontitis.
is a sampling issue between the top and base of the periodontal pocket (to be
discussed later). To study the hypothesis of microbiota on the mend and the calculated
dysbiosis ratios, a recent study presenting follow-up findings after treatment, with
well-defined depths of periodontal pockets, was performed (11). The study was con-
ducted with a different set of primers (for V5V7) and allowed testing of the dysbiosis
ratios at the genus level with a new set of primers that had not been used to determine
the ratios. Consequently, this comparative analysis can be considered a validation
experiment for the ratios. A strong correlation between the dysbiosis ratios and the
pocket depths was observed, thus highlighting the value of calculating the dysbiosis
ratio (using the selected genera of our study) as a microbial signature to evaluate the
microbiota of chronic periodontitis.
A major concern at the beginning of this work was the capacity to identify species
with multiple data sets. However, the V1V2 and V5V7 primers used in three studies
Kianoush et al. (10) (V3V4 regions; BioProject accession no. PRJEB5178) were used. One midvagina
microbiome data set from the HMP (V1V3 region, available in VAMPS) was used as an external mouth
control. Finally, the data set from the study by Bizzarro et al. (11), containing well-described sample
pocket depths (from 2 to 8 mm), was used to independently challenge the relevance of the dysbiosis
ratio of bacteria involved in periodontitis.
Ecology diversity and taxonomic identifications. Reads from the different data sets were analyzed
with VAMPS, using default parameters for taxonomic assignment to the genus level through the Global
Alignment for Sequence Taxonomy (GAST) process and using the Ribosomal Database Project (RDP)
classification to produce the best taxonomic assignment for each read. Reads identified as Archaea,
Eukarya, or organelle and unknown reads were excluded from further analysis. The frequency of each
taxonomic assignment in the data set was reported as a percentage (number of reads with the
taxonomic assignment relative to the total number of reads in the data set). Alpha diversity, as the
observed richness, and Shannon-Weaver index values were determined from the raw data sets. Differ-
ences between microbiota structures (beta diversity) were assessed using a 2D PCoA tree based on the
Bray-Curtis distances, through VAMPS. Samples were divided into five clusters (clusters 1 to 5); visual-
izations were performed using Figtree software (version 1.4.2), and 3D PCoA plots were generated using
Emperor software. Relative abundances were studied when the average abundance was above 1% in at
least one sample. Assessments of significant patterns of microbial cooccurrence or mutual exclusion at
the genus level were performed using Cytoscape 3.2.1 (42) and the CoNet plugin (43). Only genera found
in the great majority (at least 95%) of the healthy subgingival samples or the CP samples (from cluster
5) are represented.
Calculation of dysbiosis ratios of bacteria. To measure the dysbiosis, a first ratio, based on the
relative abundance of genera highly prevalent (⬎95%) in CP samples (Eubacterium, Campylobacter,
Treponema, and Tannerella) versus genera highly prevalent (⬎95%) in healthy microbiota (Veillonella,
Neisseria, Rothia, Corynebacterium, and Actinomyces), was calculated. The ratios were normalized between
samples using GraphPad Prism 6 software before comparisons. A second simplified ratio of Porphyromo-
nas, Treponema, and Tannerella versus Rothia and Corynebacterium was also tested. When no specific
genus was detected, a value of 0.1% was attributed (because “not detected” does not mean “absence”).
Statistical analysis. Normality tests for data distribution were performed. Data were studied with the
Spearman correlation test for correlations of biological origins, primers used, published sample origins,
and microbiota clusters. Observed richness (number of taxa per sample), Shannon-Weaver index values,
and dysbiosis ratios of the genera found in disease to the genera found in health were analyzed with a
Kruskal-Wallis test (nonparametric analysis of variance). Tests were carried out using GraphPad Prism 6
software and were considered significant with P values of ⬍0.05. The significant patterns of microbial
cooccurrence and mutual exclusion were analyzed as described by Faust et al. (43); a compilation of
statistical analyses (Spearman and Pearson correlations and Bray-Curtis and Kullback-Leibler dissimilarity
measures) was used, with a threshold set at 0.5. The data matrix was randomized by 100 row-wise
permutations. The P values were adjusted with the Benjamini-Hochberg false discovery rate (FDR)
correction for the number of tests, retaining only P values of ⬍0.05. Finally, the ratios of genera and
pocket depths were controlled for normality, followed by the Spearman correlation test.
ACKNOWLEDGMENTS
V.M., F.B.-H., and M.B.-M. conceived and designed the research, V.M. performed the
sampling, S.L.G.-D. performed the molecular biological analyses, V.M., S.L.G.-D., E.B.,
L.A.-A., B.M., S.B.F., and M.B.-M. performed the bioinformatic and statistical analyses and
wrote the manuscript, and M.B.-M. supervised the project.
We declare no competing financial interests.
REFERENCES
1. Teles R, Teles F, Frias-Lopez J, Paster B, Haffajee A. 2013. Lessons learned intestinal symptoms. PLoS One 6:e23035. https://doi.org/10.1371/
and unlearned in periodontal microbiology. Periodontol 2000 62: journal.pone.0023035.
95–162. https://doi.org/10.1111/prd.12010. 20. Laine ML, Moustakis V, Koumakis L, Potamias G, Loos BG. 2013. Modeling
2. Kilian M, Chapple IL, Hannig M, Marsh PD, Meuric V, Pedersen AM, susceptibility to periodontitis. J Dent Res 92:45–50. https://doi.org/10
Tonetti MS, Wade WG, Zaura E. 2016. The oral microbiome: an update for .1177/0022034512465435.
oral healthcare professionals. Br Dent J 221:657– 666. https://doi.org/10 21. Byrne SJ, Dashper SG, Darby IB, Adams GG, Hoffmann B, Reynolds EC.
.1038/sj.bdj.2016.865. 2009. Progression of chronic periodontitis can be predicted by the levels
3. Socransky SS, Haffajee AD, Cugini MA, Smith C, Kent RL, Jr. 1998. Microbial of Porphyromonas gingivalis and Treponema denticola in subgingival
complexes in subgingival plaque. J Clin Periodontol 25:134–144. https://doi plaque. Oral Microbiol Immunol 24:469 – 477. https://doi.org/10.1111/j
.org/10.1111/j.1600-051X.1998.tb02419.x. .1399-302X.2009.00544.x.
4. Hajishengallis G. 2015. Periodontitis: from microbial immune subversion 22. Haffajee AD, Teles RP, Socransky SS. 2006. Association of Eubacterium
to systemic inflammation. Nat Rev Immunol 15:30 – 44. https://doi.org/ nodatum and Treponema denticola with human periodontitis lesions.
10.1038/nri3785. Oral Microbiol Immunol 21:269 –282. https://doi.org/10.1111/j.1399
5. Hajishengallis G, Darveau RP, Curtis MA. 2012. The keystone-pathogen -302X.2006.00287.x.
microbiota. J Periodontol 81:89 –98. https://doi.org/10.1902/jop.2009 Biogeography of a human oral microbiome at the micron scale. Proc
.090397. Natl Acad Sci U S A 113:E791–E800. https://doi.org/10.1073/pnas
37. Zhou Y, Mihindukulasuriya KA, Gao H, La Rosa PS, Wylie KM, Martin JC, .1522149113.
Kota K, Shannon WD, Mitreva M, Sodergren E, Weinstock GM. 2014. 41. Zhou M, Rong R, Munro D, Zhu C, Gao X, Zhang Q, Dong Q. 2013.
Exploration of bacterial community classes in major human habitats. Investigation of the effect of type 2 diabetes mellitus on subgingival
Genome Biol 15:R66. https://doi.org/10.1186/gb-2014-15-5-r66. plaque microbiota by high-throughput 16S rDNA pyrosequencing. PLoS
38. Zijnge V, van Leeuwen MB, Degener JE, Abbas F, Thurnheer T, Gmur R, One 8:e61516. https://doi.org/10.1371/journal.pone.0061516.
Harmsen HJ. 2010. Oral biofilm architecture on natural teeth. PLoS One 42. Shannon P, Markiel A, Ozier O, Baliga NS, Wang JT, Ramage D, Amin N,
5:e9321. https://doi.org/10.1371/journal.pone.0009321. Schwikowski B, Ideker T. 2003. Cytoscape: a software environment for
39. Kigure T, Saito A, Seida K, Yamada S, Ishihara K, Okuda K. 1995. Distri- integrated models of biomolecular interaction networks. Genome Res
bution of Porphyromonas gingivalis and Treponema denticola in human 13:2498 –2504. https://doi.org/10.1101/gr.1239303.
subgingival plaque at different periodontal pocket depths examined by 43. Faust K, Sathirapongsasuti JF, Izard J, Segata N, Gevers D, Raes J,
immunohistochemical methods. J Periodont Res 30:332–341. https://doi Huttenhower C. 2012. Microbial co-occurrence relationships in the hu-
.org/10.1111/j.1600-0765.1995.tb01284.x. man microbiome. PLoS Comput Biol 8:e1002606. https://doi.org/10
40. Mark Welch JL, Rossetti BJ, Rieken CW, Dewhirst FE, Borisy GG. 2016. .1371/journal.pcbi.1002606.