Succession of the oral microbiome with the increasing severity of periodontitis

Periodontitis development is strongly associated with the succession of the oral microbiome. However, the knowledge about the succession of the oral microbiome in the development of periodontitis remains insufficient. In the present study, an analysis was conducted on the succession of tongue back, the saliva (Sal) microbiome, and gingival crevicular fluid (GCF) from healthy individuals and patients with mild (CPL), moderate (CPM), severe chronic (CPH), and generalized aggressive periodontitis (GAgP). The composition and structure of the oral microbiome gradually changed with the increasing severity of periodontitis, among which GCF showed the highest correlation with periodontitis. With an ecological preference, pathogens in the mouth varied with the development of periodontitis. In healthy and CPL patients, Sal‐derived microorganisms accounted for a large proportion of GCF but exhibited a decrease in patients with CPM, CPH, and GAgP. Permutation and time course sequencing analysis revealed that a variety of microorganisms changed with the severity of periodontitis. A majority of these microorganisms are closely related to clinical periodontal indices. Ecological analysis suggested that the composition of oral microbial communities at different stages of periodontitis is controlled by random processes. The comparison of microbial interaction networks demonstrated that a series of key microorganisms drive oral health to severe periodontitis. In this study, the relationship between the succession process of the oral microbiota and the development of periodontitis was revealed.


INTRODUCTION
Periodontitis, a common oral inflammatory disease, features the destruction of periodontium, such as gingiva, cementum, periodontal ligament, and alveolar bone.It is initiated by subgingival plaque microflora and specific pathogen-derived factors. 1 The initiation and progression of periodontitis are modulated by salivary planktonic ecology and the ecological dysbiosis of the subgingival biofilm attached to the tooth/root surface. 2 Diverse periodontitis-associated microbes have been identified in metagenomic, metatranscriptomic, and mechanistic studies.][10][11] Gingival exudates are closely bound up with periodontal tissue inflammation that results from oral microbial abnormalities.Therefore, understanding the association between the oral microbiome and periodontitis is an important proposition for the improvement of periodontal health.
Currently, it is widely accepted that the development of periodontitis is ascribed to the succession in the ecology of resident microbial communities.In healthy individuals, the oral microbiome stands for a well-balanced dynamic ecosystem that is generally inclined to keep within its typical values. 12Oral microorganisms are affected by random and deterministic factors, which results in the regular succession of microbiomes.In subgingival plaques, early colonizers, such as Streptococcus gordonii, Streptococcus mitis, Streptococcus oralis, and Streptococcus sanguinis, recognize saliva (Sal) complementary receptors in the membrane through bacterial surface adhesions.Additionally, they provide receptors for secondary colonizing bacteria including Actinomyces naeslundii, Campylobacter ochracea, Eikenella corrodens, Hemophilus parainfluenzae, Veillonella atypica, etc. 13,14 Next, transitional bacteria appear, with the typical bacteria being Fusobacterium nucleatum, which can aggregate with various early and late colonizing bacteria. 13Late colonizers, such as Porphyromonas gingivalis, Treponema denticola, and Tannerella forsythia, are named red complexes and closely linked with periodontal tissue destruction. 14,15 is considered that the development of periodontitis is initiated by the synergy and dysbiosis of microbial communities.With the development of periodontitis from mild to severe stages, the enrichment of pathogenic microbes has succeeded in certain patterns. 7The succession of the periodontal microbiota includes the involvement of various microbial species and the abundance variation of each bacterium.These bacteria play distinct roles in causing oral or extra oral tissue diseases in susceptible individuals. 16,17urthermore, oral pathogens can misdirect and subvert the immune response of hosts, which tips the balance from homeostasis to disease in the oral or extra oral niches of periodontitis patients. 18At present, the relationship between the succession of microbial ecology and the increased severity of periodontitis is not fully illuminated.
Here, the microbiota of gingival crevicular fluid (GCF), tongue back (TB), and Sal microbiome from healthy individuals and patients with mild (CPL), moderate (CPM), severe chronic (CPH), and generalized aggressive periodontitis (GAgP) was analyzed to address the above question.The definitions of the most abundant community members under each state were given.Meanwhile, the species that were enriched and those without changing in proportion (core species) were defined as well when a comparison was made between two periodontitis niches.In this study, how microbial communities were compositionally and spatially structured with periodontitis development in the distinct niches of the mouth was shown.

Human subjects and ethical approval
This study was approved by the Medical Ethical Committee of Stomatological Hospital, Shandong University (Protocol Number: GR201701).All subjects were informed about the research project and signed the written informed consent forms before participation according to the Helsinki Declaration of 1975.All subjects were met inclusion criteria with no antibiotic therapy or professional cleaning within the past 3 months, no diabetes or HIV or other systemic diseases.Healthy controls were confirmed with pockets with probing depth <3 mm and no bleeding on probing.Periodontitis patients were divided into four cohorts according to internationally accepted classification of periodontitis. 19

Sample collection and preparation
All subjects were asked to refrain from eating, drinking or performing oral hygiene within 2 h of sample collection.The oral and dental health status of each participant was evaluated by a calibrated dentist.First, all subjects were asked to rinse with tap water (10 mL) for 30 s and then expectorated at least 2 mL of unstimulated whole Sal into sterile tubes.Second, GCF samples were collected from each subject by inserting sterile paper points (size 30; two paper points per site) into the gingival sulcus of the upper right first molar, for 10 s, following isolation and supragingival plaque removal.Third, the tongue dorsum samples were collected by the sterile swabs.All samples were placed immediately into sterile micro-tubes containing 1 mL of phosphate-buffered saline (Solarbio) and frozen immediately on dry ice, and kept at −80 • C before further processing.

DNA extraction and sequencing of GCF, Sal, and TB
For high DNA purity, the cetyltrimethylammonium bromide (CTAB) method has been used, which briefly involves a proteinase K treatment, CTAB, and NaCl to precipitate proteins followed by a chloroform-isoamyl alcohol extraction. 20After extracting the genomic DNA of all samples in GCF, Sal, and TB, the V3-V4 hypervariable regions of the 16S rDNA gene were amplified by polymerase chain reaction and then sequenced on HiSeq 2500 platform.The paired-ended reads were merged to sequence tags according to the overlap relationship between the reads and passed the quality control test by UPARSE pipeline.Poor-quality reads and chimeras were filtered by vsearch software (maxee = 2.5, minlength = 400).After dereplication, the clean reads were clustered into different operational taxonomy units (OTUs) with similarity higher than 97%.The representative reads of 2750 OTUs were aligned to the Ribosomal Database Project (release 11.5) to obtain certain taxonomy of the microbiome with 80% similarity to the species of reference database.

Statistical analysis
For statistical analysis and figure generation, RStudio (R version 4.1.1)was predominantly employed.The alpha and beta diversity distance matrices were calculated using the QIIME (version 1.9.0), with a minimum depth threshold of 50,241.To explore the correlation between the host periodontal index and the microbiome of three oral niches, we employed Adonis2, Anosim, multiple response permutation procedure, and dbRDA from the "vegan" package.A statistically significant difference was defined as having a p-value of less than .05 in four methods.To assess differences in complexity based on Bray-Curtis distance, we conducted principal coordinate analysis (PCoA).Additionally, microbial source analysis of the GCF site was conducted using the SourceTracker.The site-specialist hypothesis analysis of ecological preferences.We first quantified the "core microbes" in the GCF, the metrics for quantifying the core microbiome included minimum occupancy value (80%) and abundance cutoff values (0.001). 21Next, we calculated the sitedistinctiveness ratio (SDR) for each of the "core microbes" in the GCF.The SDR is defined as the ratio of the mean relative abundance in one site to that in another one or to the average abundance across multiple sites.The criteria for categorizing site specialists and generalists were as follows: strong site specialists (SDR > 20), moderate site specialists (SDR between 5 and 20), and generalists (SDR < 5). 22ifferences among different cohorts and niches were tested using MaAsLin2 analysis, and we adjusted effects of age and gender.The remaining genera or species were selected if the false discovery rate (FDR)-adjusted p-value was less than .001.Additionally, k-means-based clustering analysis was performed to group the differently distributed differential microbes at the GCF.Those differently distributed microbial species at GCF were clustered into five modules according to the similarity of their abundantvariation pattern and were named M1-M5.Subsequently, Spearman correlation analysis was employed to establish the correlation network within each module (|Spearman correlation| ≥ 0.3, p-value < .01,and FDR-adjusted pvalue ≤ .25),as well as to determine the interrelationships between modules in different groups (|Spearman corre-lation| ≥ 0.55, p-value < .01,and FDR-adjusted p-value ≤ .25).To further characterize the microbial variation pattern along the severity of periodontitis, the expression mode clustering was analyzed and visualized using the time course sequencing ("TCseq") package (version 1.23.0).At the same time, the variation trend of differential microbes in some models was shown by line chart.
The correlation between the oral microbes and clinical phenotype was assessed using partial Spearman correlation analysis, and we adjusted effects of age and gender.The oral microbes here include the difference of microbes among healthy, CPL, CPM, CPH, and GAgP groups.Then, the random forest analysis (version 4.7-1.1)was used to test the classification effect of these different microbial.
The mediation analysis was conducted using the "mediation" package (version 4.5.0), and we adjusted effects of age and gender.The criteria for testing whether bacteria can influence the severity of periodontitis through clinical phenotype were set as FDR-adjusted p-value ACME < .05(ACME, average causal mediation effect) and FDRadjusted p-value Total effect < .05.
To examine the relative contribution of stochastic and deterministic processes in shaping oral microbes communities, we utilized the Sloan neutral model, which predicts the relationship between occurrence frequency and relative abundance of OTUs in different cohorts.The SparCC method was applied separately to the healthy group and the periodontitis group to estimate correlation coefficients between taxa.SparCC method was implemented using FastSpar (version 0.0.6)software (|SparCC correlation| ≥ 0.5, p-value < .05).To identify "driver" nodes in the network of microbial associations between healthy and diseased groups, the NetShift method available at https://web.rniapps.net/netshiftwas utilized.The complex changes in bacteria were visualized using a Sankey plot created with the "ggalluvial" package (version 0.12.3).Venn analysis, available at http://jvenn.toulouse.inra.fr/app/example.html,was employed to identify group interactions.The network diagram was drawn using Cytoscape (version 3.7.1)and Gephi (version 0.9.2).Additional figures were generated using the "ggplot2" (version 3.4.0)and the "pheatmap" (version 1.0.12).
We have uploaded the code to https://github.com/MFZhi/Periodontitis1 and the accession number of the submission is HRA006244.

Characteristics of the oral microbiome at different oral niches and different severities of periodontitis
A total of 419 samples were collected from GCF (123), Sal (147), and TB (149) from 30 healthy individuals and 44 CPL, 21 CPM, 35 CPH, and 19 GAgP patients according to the 1999 Classification of Periodontal Diseases and Conditions. 19The periodontal diagnosis 23,24 and detailed clinical information of each subject are presented in Table S1.The clean reads were clustered into 2750 OTUs, and annotated as 22 phyla, 385 genera, and 327 species taxa (Table S2).Adonis was used to probe into the correlation between host periodontal indices and the microbiome of three oral niches.It was found that the microbiome of GCF was associated mostly with the severity of periodontitis and significantly with all periodontal indices, followed by that of Sal.Nevertheless, the microbiome of TB was less related to the periodontal index (Figure 1A).
An analysis was first made of the main microbial composition at the genus level to reveal the characteristics of the oral microbiome with the increasing severity of periodontitis.As shown in Figure S1A, the main microbial composition of TB and Sal was relatively stable, whereas that of GCF gradually varied with the severity of periodontitis.At the genus level, for instance, Fusobacterium and Prevotella increased with the development of periodontitis, while Streptococcus exhibited a dramatic decline (Figure S1B).The analysis of α-diversity in the oral microbiome showed that the microbial diversity of GCF demonstrated a significant reduction in all periodontitis groups compared with healthy individuals.However, the microbial diversity of TB increased, while that of Sal showed no differences (Figures 1B and S1C).Of note, the comparison of α-diversity in healthy individuals showed significant differences among GCF, TB, and Sal.Among them, GCF was highest in α-diversity, followed by Sal and then TB (Figure S2A,B).In all periodontitis groups, the diversity of GCF was reduced and showed no differences from that of Sal, but remained significantly higher than that of TB.Subsequently, an investigation was conducted on the β-diversity of the oral microbiome to respond to periodontitis severity.PCoA and Bray-Curtis distance analysis showed a transitional shift along the severity of periodontitis at GCF and Sal (Figures S2C  and 1C).These results suggest that GCF is the most important action niche in periodontitis in comparison with Sal and TB, and changes continuously with the severity of periodontitis.
SourceTracker 25 was employed to calculate the proportion of Sal or TB-originated microbes to the GCF niche along periodontitis severity, to reveal whether the origin of microbes varied in GCF with the severity of periodontitis.It was observed that the contribution of Sal-originated microbiome accounted for a large proportion in healthy individuals and CPL patients, but reduced dramatically in CPM, CPH, and GAgP patients.In addition, microbes from unknown sources increased in CPM and CPH patients.TB-originated microbiome dramatically increased in the GCF niche of GAgP patients (Figure S2D).The distances between GCF and Sal and between GCF and TB samples were compared by Bray-Curtis distance analysis.It was found that GCF was more similar to Sal, which was consistent with the results of SourceTracker analysis (Figure S2E).

Variation and site preference of "core microbes" along the severity of periodontitis
The species-level "core microbiome" of each oral niche was analyzed and shown by Venn diagram.The diagram obtained 27, 24, and 14 "core microbes" at GCF, Sal, and TB, respectively (Figure 2A).Notably, several microbes recognized as classical periodontal pathogens were consistently over-represented in both healthy and periodontitis patients at GCF, such as P. gingivalis, T. forsythia, Campylobacter gracilis, Prevotella nigrescens, Parvimonas micra, and E. corroden.Most of these microbes were also "core microbes" at the Sal niche.As "core microbes" at TB, P. micra and Solobacterium moorei were classified as periodontal pathogens (Figure 2A).The core microbiome at the genus level was also analyzed and is shown in Figure S3A.
Next, k-means-based clustering analysis was performed to disclose the variation pattern of core microbes with the severity of periodontitis at GCF, Sal, and TB niches.The results show that P. gingivalis, T. forsythia, P. nigrescens, E. corroden, and other periodontal pathogens increased from healthy to CPH at both GCF and Sal niches.The taxa enriched in healthy individuals, such as Lautropia mirabilis and Neisseria subflava, decreased with the sever-ity of periodontitis at both GCF and Sal (Figure 2B).The variation pattern of the core microbiome at the genus level was analyzed and is shown in Figure S3B.
According to ecological theory, 22 a majority of residential microbes in the oral cavity of humans have a primary niche where they are most suitable to live.The SDR was calculated for each of GCF "core microbes" to analyze whether the primary habitat type of core microbes changes with periodontitis severity.SDR > 20 and between 5 and 20 were defined as strong and moderate site specialists, respectively, and SDR < 5 was defined as generalists in certain niches.A graphical representation of this habitatspecific microbial species was analyzed and is displayed in Figure 2C and Table S3.At the GCF niche, Treponema medium was the only strong site specialist in healthy individuals, while the number of specialists increased with the severity of periodontitis.Specifically, periodontal pathogen P. gingivalis was a GCF site generalist in healthy individuals and became a moderate GCF site specialist in CPL and CPM patients; T. forsythia was a moderate GCF site specialist in CPL, CPM, and CPH patients; and P. micra was a moderate GCF site specialist in all individuals except CPL patients (Figure 2D).Of Sal, GAgP harbored some strong and moderate site specialists, such as periodontal pathogens P. gingivalis, T. forsythia, and C. gracilis (Figure 2E).
At the genus level, the number of genus site specialists also increased with the severity of periodontitis at GCF and Sal niches.Campylobacter and Treponema were site specialists in healthy individuals (Figure S4A and Table S4).Selenomonas, Dialister, Aggregatibacter, and Cardiobacterium were GCF site generalists in healthy individuals and CPL, CPM, and CPH patients, respectively.In addition, they became moderate GCF site specialists in GAgP patients (Figure S4B).Of Sal, GAgP also harbored numerous strong and moderate site specialists classified as periodontal pathogens (Figure S4C).These results suggest that oral pathogens may have ecological preferences in the oral cavity and vary with the development of periodontitis.
Moreover, the Sankey diagram was adopted to reveal the niche preference of known periodontal pathogens and their distribution preference among different severities of periodontitis.The results show that microbes from "red complexes" accounted for a big part at GCF and were mainly harbored in patients with CPM and CPH.Healthy individuals could detect microorganisms from red, orange and green complexes whose abundance however was relatively low.These results indicate that periodontal pathogens were widely distributed in the whole population.Nevertheless, they were more abundant in periodontitis patients, especially in CPM and CPH patients, and tended to be enriched at the niche of GCF (Figure 2F).

Variation pattern of oral microbes along the severity of periodontitis
To figure out the microbes which varied significantly along the severity of periodontitis, a permutation test was conducted to compare the difference of microbes among healthy, CPL, CPM, and CPH groups.A total of 64, 30, and 13 microbial species at GCF, Sal, and TB niches were identified, respectively (Figures 3A and S5A,B and Table S5).The permutation test results at the genus level identified 101, 25, and 20 differently distributed genera at GCF, Sal, and TB niches, as shown in Figure S6A-C.Next, those differently distributed microbial species at GCF were clustered into five modules by the similarity of their abundant-variation pattern and named M1-M5.The M1-M5 modules represent the similarity of their abundant-variation pattern of differently distributed microbial species at GCF (Figure 3A).Microbes in M1, M2, and M3 including Lactobacillus san-franciscensis and Rhizobium rosettiformans, maintained a high abundance in healthy individuals and showed a continuous decrease from healthy individuals to CPH patients.Microbes in M4 preferred to harbor in healthy individuals and CPL patients.Microbes in M5, which involve a group of well-known periodontitis pathogens such as P. gingivalis, T. forsythia, and T. denticola, presented an increasing trend from healthy individuals to CPH patients.
Spearman correlation analysis showed that most microbes in each module had a positive correlation with each other (Figure 3B).The microbes in M1 and M5 had more positive correlation than M2 and M3.The microbes in M4 had the minimal correlation.
While the interrelationships between the five modules (M1-M5) varied with changes in oral health status, microbes in M1 were positively correlated with microbes from M2 and M3 but negatively correlated with microbes from M5 in health individuals.The interrelationships between the five modules (M1-M5) reduced with the increasing severity of periodontitis (Figure 3C).These results suggest that the relationship between oral microbes also changed continuously in addition to the changes in the abundance of individual microorganisms with the development of periodontitis.
To further characterize the variation pattern of microbes along the severity of periodontitis, TCseq analysis was performed on all microbial species in GCF, Sal, and TB, respectively.Of GCF, eight clusters were determined as the ideal grouping strategy, and the results are illustrated in Figure 4A.Microbes from cluster 1 showed an increasing pattern from healthy individuals to CPM patients.They involved "red complex" periodontitis pathogens P. gingivalis, T. forsythia, and T. denticola and other two oral microbes, namely Prevotella baroniae and Porphyromonas endodontalis, which were frequently reported in the research of periodontal diseases.Microbes in cluster 2 kept increasing from healthy individuals to CPH patients, which involved periodontitis pathogens P. micra, Anaeroglobus geminatus, and Treponema maltophilum.Cluster 7 exhibited a consistently decreasing pattern, which involved H. parainfluenzae and Abiotrophia defectiva (Figure 4A).
Of the Sal niche, eight clusters were also determined as the ideal grouping strategy, and the results are presented in Figure 4B.Periodontitis pathogens, including P. gingivalis, T. forsythia, and P. micra, were classified into cluster 1, which showed a consistent increasing pattern along periodontitis severity.Of the three periodontitis pathogens, P. gingivalis was the highest abundant microbe in this cluster.In cluster 4, the abundance of Catonella morbi kept increasing from healthy individuals to CPH patients, which was a novel bacterium reported to induce periodontitis (Figure 4B).Of the TB niche, periodontitis pathogens, including P. gingivalis and P. endodontalis showed a consistently increasing pattern along periodontitis severity, as shown in Figure S7.The TCseq results also identified eight clusters of differently distributed genera at GCF, Sal, and TB, as shown in Figure S8A-C.(Figures 5A and S9A).In GCF, 41 microbes negatively correlated with almost all of the seven clinical periodontal indices were identified.Most of those microbes were from modules 1-4 in Figure 3A and decreased with the severity of periodontitis.Also, 24 microbes in GCF were identified.

Correlation between oral microbes and periodontal clinical indices
Mainly from module 5 (Figure 3A), these microbes were positively correlated with most of the clinical periodontal indices.To be specific, L. sanfrancisensis and L. mirabilis were significantly negatively correlated with all clinical periodontal indices except pigment, while P. gingivalis, T. forsythia, T. maltophilum, and T. denticola were significantly positively correlated with most clinical periodontal indices (Figure 5A).In Sal, microbes positively related to clinical periodontal indices were also periodontal pathogens (Figure S9A).The microbes in Figure 5A were used as feature variables, and the random forest model at the GCF niche was adopted to determine the oral health status in any two of the five groups.It was shown that these microbes could effectively categorize individuals from different subgroups, with an area under curve value of over 90% in most cases (Figure 5B).The random forest model analysis at the Sal niche also showed the ideal power of distinguishing any two of the five groups (Figure S9B).
Then, mediation analysis was conducted to evaluate the potential causal relationships between oral microbes, clinical periodontal indices, and periodontitis severity.As a result, 170 mediation linkages in GCF (Figure 5C and Table S6) and 58 in Sal (Figure S9C and Table S6) were established.In specific terms, Eubacterium hallii was negatively correlated with the severity of periodontitis with the mediation of pocket depth, and Blautia wexlerae was negatively associated with the severity of periodontitis with the mediation of subgingival plaques.However, T. forsythia could promote the development of periodontitis by increasing pocket depth and P. gingivalis by increasing subgingival plaques in GCF (Figure 5D).More detailed mediation results between oral microbes and the degree of periodontitis are shown in Figure S10.

Changes of ecological communities in the development of periodontitis and driver microbes
To investigate the contribution of neutral processes to the assembly of oral microbial communities in periodontitis development, the neutral community model (NCM) was used for explaining the distribution of microbes along the severity of periodontitis development.The NCM succeeded in estimating a large fraction of the association between the occurrence frequency and relative abundance variations of OTUs (Figure 6A).Specifically, 83.6%, 87.4%, 87.1%, and 89.7% of explained community variances were for healthy individuals and patients with different severities of periodontitis.This indicates that stochastic processes were of importance to shape the oral microbiome across healthy individuals and patients with different severities of periodontitis.The m value was estimated between 0.0479 and 0.0716 in the oral microbiota from healthy individuals and groups with different severities of periodontitis.A high m value represented a high species dispersal ability.These NCM results indicated that oral microbiota species dispersal was highest in CPM, but low-est in CPL.This suggests that species dispersal ability was not homogeneous in healthy and periodontitis groups.
Some microbial taxa occurred more or less frequently than predicted by the model in consideration of their overall abundance in the metacommunity (points above in blue or below in orange, the line in Figure 6A).Specifically, points above the prediction represent taxa found more frequently than expected, which indicates higher migration ability, while those below the prediction represent taxa found less frequently than expected, which indicates lower dispersal ability.Given the difference of occurrence frequency between neutral model predictions, differently distributed microbes were categorized into neutral (as prediction) and non-neutral (above and below prediction) fractions (Figure 6B and Table S7).It was found that most of the healthy individuals enriched with oral microbes were neutral or below prediction, and microbes enriched in different periodontitis groups were above prediction.To be specific, P. gingivalis and P. endodontalis were as predicted in the healthy group but above the prediction in all periodontitis groups.T. forsythia was below prediction in the healthy group but above prediction in all periodontitis groups.This result indicates that the migration rate of periodontal pathogens may increase in all periodontitis groups.
Next, how the community structure varied with the dynamic balance of niches and neutrality was investigated by constructing the ecological co-occurrence networks of the oral microbiota.The results show that the network structure of periodontitis groups was relatively loose compared with that of the healthy control group (Figure S11A).Parameters of the network topology of microbial communities from healthy and all periodontitis groups were analyzed.It was shown that betweenness centralization, average path length, graph diameter, and node number decreased in periodontitis groups (Figure S11B).Inter-OTU associations were reduced in periodontitis groups compared with healthy habitats, which was thus more likely to trigger ecological imbalance.
To find out the key microbes that drove the development of periodontitis, NetShift was used for comparing the microbial interaction network among every two adjacent groups along the development of periodontitis (Figure 6C), and Table 1 shows the summary of node properties.N (control/case) refers to degree of the node in control/case, Exclusive refers to exclusive in "case," DelBet refers to the delta betweenness score from control to case, and COM refers to the community.
Impressively, T. forsythia, T. denticola, Selenomonas sputigena, etc., were identified to be critical nodes with high neighbor shift (NESH) scores from healthy to CPL groups.This means that these species contributed to the main differences between the two interaction networks.In addition, they were regarded as "drivers" of the variations between healthy individuals and CPL patients.
From CPL to CPM patients, P. gingivalis, T. maltophilum, and C. gracilis displayed a high NESH score and were viewed as "driver" species.Of CPM to CPH, S. sputigena, Selenomonas infelix, etc., contributed to major differences (Figure 6D).These microbes may provide an important contribution in the development of periodontitis.

DISCUCCION
The occurrence of periodontitis results from the regular succession of oral microecology influenced by the oral environment and external factors.As next-generation sequencing technology develops, the pathogenesis study of periodontitis has become a hot topic in research on the oral microbiome.Multiple studies revealed that the composition and characteristics of the oral microbiome changed in periodontitis patients.It was found that a large number of oral microbes were differentially distributed between periodontitis and healthy individuals. 27,28However, few studies evaluated the correlation between the regular succession pattern of the oral microbiome and the development of periodontitis.In this study, the focus was first placed on the succession pattern of the oral microbiome with the development of periodontitis in three oral niches.It was identified that periodontal pathogens such as P. gingivalis, T. fenticola, and T. forsythia may be the driver pathogens of periodontitis.0][31][32] In the current study, the importance of these pathogenic bacteria in the occurrence and development of periodontitis was further revealed.
Studies show that microbial communities in an assortment of oral niches are different due to environmental differences, including oxygen, pondus hydrogenii (pH), nutrition, etc.][35][36] Periodontitis gradually increases in severity with the development of subgingival plaques and the deepening of periodontal pockets, and these microorganisms initially colonize in other oral niches. 27In this study, the microbiome of GCF, TB, and Sal in healthy individuals and patients with different severities of periodontitis was analyzed, and the differences of microbes in different oral niches were revealed.SourceTracker was applied to predict the origin of periodontal pocket microbes.This study helped to find out the site preferences of oral microbes and revealed that the origin of microbes varied in GCF with the severity of periodontitis.
Next, the results of this study revealed the variation pattern of "core microbes" with the severity of periodontitis and showed that those periodontal pathogens were widely distributed in the whole population, but more abundant in periodontitis patients and tended to enrich in the niche of GCF.In a recent literature review, Welch et al. proposed the site-specialist hypothesis of the oral microbiome to summarize the following evidence: separate sites in the mouth, such as the teeth, tongue dorsum, and buccal mucosa, are profoundly different in the microbiota they support despite being bathed in the same Sal.The hypothesis is conducive to explaining the term "structure," which has been used in microbiome literature with compositional and spatial meanings.It was used for describing how microbial communities are structured both spatially and compositionally at distinct sites in the mouth.The hypothesis predicts that a microbe will actively localize at its preferred site, grow and divide there, and be detected with much lower abundance and display dramatically altered metabolism, gene expression, and spatial organization outside its preferred site. 22In the present study, an analysis was conducted to distinguish site specificity in the oral microbiome.A graphical representation of this habitat-specific microbial species was analyzed.Results of this study indicated that "core microbes" may have ecological preferences in the oral cavity and vary with the development of periodontitis, which validated the site-specialist hypothesis.
][39][40][41] The permutation test and TCseq analysis were applied to reveal the succession process of bacteria from healthy individuals to CPH patients, which showed the variation pattern of assorted periodontitis pathogens.However, some potential oral probiotics such as L. sanfranciscen-sis and R. rosettiformans maintained high abundance in healthy individuals and exhibited a consistent decrease from healthy individuals to CPH patients.Notably, not only the increase of single microbes but also the alterations in the overall structure of the microbial community were the main factors that induced periodontal diseases. 42To address this issue, community composition was compared with distribution across periodontitis development by a neutral assembly model.Niche-neutrality tradeoffs not only highlight the external indicators of microbial assemblages (species diversity and abundance distribution) but also affect the inner structures of communities such as the interactions or correlations between OTUs.Nicheand neutral-based theories constitute two important and complementary mechanisms for the understanding of microbial community assembly.In this study, the NCM successfully estimated a large fraction of the relationship between the occurrence frequencies of OTUs.Bohannan and coworkers found that the fit of the model to the observed gut microbial distribution decreased (R 2 from 0.81 to 0.39) with the development of zebrafish from larvae to adults.This suggests that the relative importance of microbe-microbe interactions, active dispersal, selection by hosts, and other non-neutral processes increases with the maturity of hosts. 43It was discovered that the fit of the model to the observed oral microbial distribution kept high and steady (R 2 all up 0.8, CPH was the highest up to 0.897) with the development of periodontitis from mildness to severity.Additionally, the differences in the interaction networks of the microbiome along with the increasing severity of periodontitis were compared.It was noticed that oral microbial communities may be affected by a higher neutrality stochastic process.
In ecology, species that influence the overall community disproportionally larger than the biomass they occupy are known as keystone or "driver" species. 44P. gingivalis is called a "keystone pathogen," namely a keystone species supporting and shaping a microbial community in ways that also promote disease pathogenesis. 5In this study, a method called "NetShift" was introduced to identify important microbial taxa serving as "drivers" from health to disease, and multiple important "driver" species which may get involved in the development of periodontitis were identified.

CONCLUSION
Overall, this study revealed the succession of the oral microbiome among multiple oral niches in patients with increasing severity of periodontitis.This helps improve the knowledge of ecological and spatial characteristics between oral microbiome communities and periodontitis development.

F I G U R E 2
The distribution of "core microbes" at gingival crevicular fluid (GCF), saliva (Sal), and tongue back (TB) with the severity of periodontitis.(A) The "core microbiome" at GCF, Sal, and TB (species level).(B) The abundance of core microbes at GCF, Sal, and TB with the severity of periodontitis.(C) The variation of preferred habitat with the severity of periodontitis.Colored dots represent species, the y-axis shows the ratio of the taxon's normalized abundance in Sal to its normalized abundance in TB; the x-axis shows the ratio of normalized abundance in GCF to the mean of the Sal and TB normalized abundance.The area of each dot is proportional to the mean abundance of the taxon in GCF site.(D) The heatmap of generalists and specialists vary with the severity of periodontitis at GCF. (E) The heatmap of generalists and specialists vary with the severity of periodontitis at Sal. Red represents strong site specialists, and blue represents moderate site specialists.(F) Sankey diagram of the known periodontal pathogens and their distribution preference with the severity of periodontitis.

F I G U R E 3
The succession pattern of oral microbes at gingival crevicular fluid (GCF), saliva (Sal), and tongue back (TB) with the severity of periodontitis.(A) The heatmap of differently distributed microbial species at GCF. (B) The networks of microbial species with similar distribution pattern (modules 1-5).The node colors represent the phylum.Lines between nodes represent correlations between the nodes connected by the lines, green representing positive correlation.(C) The interrelationships between the five modules with the severity of periodontitis.The node colors represent the five modules in (A).The node sizes represent the mean relative abundance.Lines between nodes represent correlations between the nodes connected by the lines, red representing negative correlation, and green representing positive correlation.F I G U R E 4 Time course sequencing (TCseq) analysis of all microbial species in gingival crevicular fluid (GCF) (A) and saliva (Sal) (B) along the severity of periodontitis.

Partial
Spearman correlation was utilized to analyze the correlation between oral microbial species and clinical periodontal indices in GCF and Sal, respectively

F I G U R E 5
The differently distributed microbes impact multiple clinical periodontal indices.(A) Spearman correlation of differently distributed microbes and clinical periodontal indices (gingival crevicular fluid [GCF] niche).(B) Random forest analysis in any two of the five groups (GCF niche).(C) Mediation analysis among oral microbes, clinical periodontal indices, and periodontitis severity (GCF niche).(D) The causal relationships between microbes, clinical periodontal indices, and periodontitis severity inferred by mediation analysis.The r Spearman and significance are labeled at each edge and the proportions of indirect effect (mediation effect) are labeled at the center of the ring charts.*FDR-adjusted p-value < .05,**FDR-adjusted p-value < .01,***FDR-adjusted p-value < .001,****FDR-adjusted p-value < .0001.

F I G U R E 6
Ecological characteristics of periodontitis microbiome succession.(A) Fit of the neutral community model (NCM) of community assembly.The solid blue line represents the fitting curve and the dashed blue line represents the 95% confidence interval.The coefficient of determination (R 2 ) was the goodness of fit of the neutral model.It ranged from ≤0 (no fit) to 1 (perfect fit).Operational taxonomy units (OTUs) that occur more frequently than predicted by the model are shown in blue while those that occur less frequently than predicted are shown in orange.(B) The distinct fitting proportions of microeukaryotic communities' OTU and sequence numbers by the Sloan neutral model.(C) The comparison of microbial interaction network between two adjacent groups along periodontitis development.Nodes of the common species are arranged on the periphery of the circle.All nodes are randomly assigned different colors.Node size shows the predicated "driver" scores, while the big and red nodes are particularly important "drivers."(D) The top five "drivers" species among along periodontitis development.

TA B L E 1 Summary of node properties. S_ID N (con- trol) N (case) Core (case) Union Intersect Exclusive Jaccard score NESH score DelBet COM
Abbreviations: COM, community; CPH, patient with severe chronic; CPL, patient with mild; CPM, patient with moderate; DelBet, delta betweenness score from control to case; Exclusive, exclusive in "case"; N (control/case), degree of the node in control/case; NESH score, neighbor shift score.