Microbiomes in Supragingival Biolms and Saliva of Adolescent Patients with Induced and Naturally Occurring Gingivitis Relative to Gingival Health

Background: Comparison of the microbiomes in supragingival biolm and saliva samples collected from juvenile patients developing induced or spontaneous gingivitis with healthy controls. Results: 36 supragingival biolm samples from 9 adolescent gingivitis patients wearing orthodontic appliances (induced gingivitis), 40 supragingival plaques from 10 patients having spontaneous gingivitis, and 36 control samples from 9 individuals without gingivitis in the same age group were analyzed by 16S rRNA gene amplicon sequencing. Salivary microbiomes of the same persons were characterized by shotgun metagenome sequencing to compare the sessile, i.e. biolm immobilized communities with planktonic microbiota. The amplicon and whole genome data sets were scrutinized using bioinformatics workows designed to minimize systemic biases. RDP and RefSeq reference databases were compared in the identication of microbiome members. The supragingival bacterial communities in 112 supragingival biolm samples were determined using next generation amplicon sequencing, in which a short, variable region of the 16S rRNA coding marker gene was amplied by PCR before sequencing. 36 samples derived from 9 adolescent patients diagnosed with induced gingivitis, 40 samples from 10 adolescent patients diagnosed with spontaneous gingivitis, and 36 samples from 9 age-matched healthy individuals without the symptoms of gingivitis were included in this study.

Although local plaque accumulation plays a decisive role in the induction of gingivitis, systemic factors may also modulate the gingival in ammatory response [19]. In an experimental gingivitis model, two subpopulations of healthy adults could be distinguished based on the individual variability of the plaqueinduced in ammatory response [20]. Thus, important differences in clinical parameters of gingival in ammation between "high responder" and "low responder" individuals were reported, which were characterized by similar amounts of dental plaque deposits [20,21]. It was concluded that metabolic, genetic and environmental factors may affect the bidirectional interactions between the gingival tissue and the bacterial bio lm [21]. Smoking suppressed gingival bleeding on proving, a clinical sign of gingival in ammation in a related study [22]. Other environmental factors may also contribute to the higher gingivitis susceptibility [23]. For example, salivary steroid hormones were also suggested as potential modulators of gingivitis prevalence, especially during adolescence [24].
The propensity to develop gingivitis increases gradually from early childhood to adult age [25,26].
Childhood gingivitis is typically characterized by gingival color change and swelling clinical symptoms, rather than bleeding [27]. Plaque-induced chronic gingivitis is highly prevalent among adolescents and even relative small amounts of plaque may elicit an in ammatory reaction in this age group (12)(13)(14)(15)(16)(17)(18)(19) years) [19,28]. It was also observed that the chance of periodontitis development following gingival in ammation is lower in adolescents, compared to adults [29].
Permanent orthodontic treatments are frequently associated with increased presence of oral pathogens, e.g. Porphyromonas gingivalis, Tannarella forsythia, Aggregatibacter actinomicetemcomitans, Prevotella intermedia, Treponema denticola, which triggers immune response [30]. In a related study P. intermedia, T. denticola, Fusobacterium nucleatum and Campylobacter rectus were identi ed in increased abundance in gingival plaques after 3 months of placement of the braces although the general rearrangement of the core gingival microbiota apparently did not take place [31].
A signi cant fraction of adolescents and adults may undergo orthodontic treatment to correct crowded, rotated, buried or prominent front teeth [32]. The application of permanent orthodontic appliances (braces) may facilitate plaque accumulation by creating new retention sites, which are di cult to access during teeth cleaning [33,34]. It was also suggested that the surface of orthodontic appliances may favor the build-up of bacterial communities, which differ in composition from the microbiomes of dental plaques causing spontaneous gingivitis [33]. Overall, the published depictions of microbial communities associated with gingivitis are incongruent, at least partly due to methodological differences applied in the various laboratories. Therefore, the utilization of up to date sequencing and bioinformatics work ows are warranted to minimize systemic oversights.
For this reason, and because of the paucity of metagenomics data related to the supragingival bio lm communities of adolescents, we analyzed the microbiomes in supragingival plaque and saliva samples of 15-18 years old patients displaying symptoms of either orthodontic appliance induced or spontaneous gingivitis, and compared them to the microbial communities of age-matched healthy controls. The study included amplicon sequencing of the 16S rRNA gene of dental plaque bio lms and whole genome sequencing of saliva samples of the same patients. The sequencing data were evaluated employing complex bioinformatics work ows. Both RDP and RefSeq reference databases were tested in the identi cation of microbiome community members.

Study design and recruitment of participants
The study protocol was approved by the Institutional Review Board of the University of Szeged, Szeged, Hungary. Signed informed consent was obtained from each adult participant enrolled into the study at the Department of Orthodontics and Pediatric Dentistry University of Szeged, Hungary. In case of study participants younger than 18 years old, signed informed consent was obtained from one of the parents.
Subjects were divided into 3 study groups: Group A: induced gingivitis patients wearing xed metal braces (9 patients, 3 males and 6 females, mean age 16.9 year, range: 15-18 years); Group B: spontaneous gingivitis patients diagnosed with plaque-induced gingivitis (10 patients, 7 males and 3 females, mean age 17 year, range: 16-18 years); Group C, healthy individuals without the symptoms of gingivitis (9 patients, 4 males and 5 females, mean age 17 year, range: 15-18 years). None of the participants suffered from known chronic systemic illness and none were treated with antibiotics at least 6 months prior to sampling. All subjects declared having no habitual smoking or drinking dependencies.
The patients' characteristics, including the Modi ed Gingival Indices (MGI) characteristic for the gingival condition [35,36], are summarized in Supplementary table 1. Fixed orthodontic appliances were placed on the labial tooth surfaces (more than 1.5 mm from the gingival margin) according to the professional rules with an acid-etch composite system (Transbond XT, 3M Unitek, Neuss, Germany), as in an earlier study [37].

Dental examination and sample collection
Each participant underwent dental examination by a quali ed dentist belonging to the staff of the Department of Orthodontics and Pediatric Dentistry. All patients were repeatedly instructed about proper oral hygiene during the orthodontic treatments. The gingival condition of each participant was assessed as described by Lobene et al., using the non-invasive Modi ed Gingival Index, MGI [35,36,38].
Unstimulated whole saliva samples were collected from the participants by the simple drooling method. The samples were aliquoted and stored at -80 C°. Supragingival plaque (bio lm) samples were taken using sterile paperpoints from the surface of 4 teeth with in amed gum of each participant diagnosed with gingivitis (Groups A and B) as well as from the surface of the same number of teeth of each healthy controls without the symptoms of gingivitis (Group C). The supragingival plaque samples were stored in sterile plastic tubes (Axygen, MCT-150-C 1.5 mL) at -80 C° until DNA isolation. DNA isolation, sequencing and data analysis DNA isolation from saliva and supragingival plaque samples Saliva samples were thawed and 3 mL of each was centrifuged at 13,000 rpm for 5 min. Supragingival plaque samples taken by sterile paperpoints were resuspended in 500 uL TE buffer (10 mM Tris, 1 mM EDTA, pH 8) and were also pelleted at 13,000 rpm for 5 min. DNA extractions were carried out from both sample types by using the Macherey-Nagel NucleoSpin Soil DNA kit (Macherey-Nagel: 740780.250). The lysis mixture contained 700 µL SL1 and 150 µL Enhancer SX lysis solutions. After lysis (bead beating: maximum speed for 5 min), the kit protocol was followed. The quantity of DNA was determined in a NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, USA) and a Qubit 2.0 Fluorometer (Life Technologies, Carlsbad, USA). DNA purity was tested by agarose gel electrophoresis and on an Agilent 2200 TapeStation instrument (Agilent Technologies, Santa Clara, USA).
Next generation-sequencing of supragingival bio lm and saliva samples Prokaryotic 16S rRNA gene ampli cation, puri cation and sequencing were performed as described in "Preparing 16S Ribosomal RNA Gene Amplicons for the Illumina MiSeq System" standard protocol provided by the supplier (Illumina). Brie y: the hypervariable V3-V4 region of the 16S rRNA gene was PCRampli ed by using the forward primer 5'TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGCCTACGGGNGGCWGCAG; and the reverse primer 5'GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGGACTACHVGGGTATCTAATC The PCR products were cleaned up by using the Macherey-Nagel NucleoSpin Gel and PCR Clean-up kit (Macherey-Nagel: 740609.50) and checked with Agilent TapeStation 2200. Library preparation was done following the instructions of NEBNext Ultra II DNA Library Prep Kit for Illumina (Cat. Num.: E7645L). DNA sequencing was carried out on an Illumina MiSeq machine using V2 sequencing chemistry (MiSeq Reagent Kit v2) (500 cycles). In vitro fragment libraries were prepared from the saliva total DNA samples for the whole genome sequencing (shotgun metagenome sequencing) using the NEBNext Ultra II DNA Library Prep Kit by Illumina. Paired-end reads were generated on an Illumina NextSeq sequencer using TG NextSeq 500/550 High Output Kit v2 (300 cycles).
Bioinformatics pipeline for amplicon sequence analysis of supragingival bio lm samples A novel bioinformatics pipeline consisting of ve modules was constructed to handle the amplicon sequencing data ( Figure 1A). In the Sequence preparation module, fastq interleacer was used to join paired-end fastq reads from two separate les, one with the left mates and one with the right mates, into a single le. The Trimming module employed Trimmomatic (v.0.36.5) to trimming of the raw sequences (sliding window: 4:20; minlen: 200; leading: 3; trailing: 3) [39]. Raw and processed sequence quality was checked by FastQC (v.0.11.8). Taxonomic classi cation of amplicon DNA reads was done in the Taxonomic annotation module by the sensitive and highly accurate Kraken 2 (v.2.0.8) program using the NCBI RefSeq (genome) and RDP (16S rDNA amplicon) databases [40]. In the Filtration and normalization module, both Kraken feature table outputs were ltered by Kraken 2 lter command (con dence threshold: 0.95 indicated ~95% precision to the lowest common ancestor). Normalization to the 16S rRNA gene copy numbers was done through the rrnDB (v.5.6) database [41]. Metagenomeseq (v.1.16.0) was used for cumulative sum scaling and to create normalized and scaled output of microbial abundance matrices [42]. The ltered and normalized microbial abundance matrix was converted to standard biom (JSON formatted, taxon table) format. In the Statistics and visualization module Megan6 (v.6.18.1) was used to scrutinize the microbial community and export data for statistic calculations [43]. Rarefaction estimation was performed by Megan6. The average composition of microbial taxa was visualized by the Krona (v.2.6.1) program [44]. The distribution of top 10 most frequent microbes between the three types of samples (A: induced gingivitis; B: spontaneous gingivitis; and C: control) was presented in Circos (v.0.63.9) [45]. For microbial core and diversity calculation MetaCoMET (Metagenomics Core Microbiome Exploration Tool), an interactive web tool, was used [46]. Shannon statistical method helped the estimation of alpha diversity. Emperor program (integrated to MetaCoMET) was employed for principal component analysis. Statistical Analysis of Metagenomic Pro les (STAMP v.2.0.9) computed the abundance differences in the case of whole microbiomes [47]. Dissimilar taxa were identi ed with two-sided t-test at 0.95 con dence intervals and the results with q-value (corrected p-value) of <0.05 were retained. Minimum difference between proportions was set to 0.3 in STAMP and Storey FDR (False Discovery Rate) was used in order to lter out false positive signi cant differences. UPGMA (Unweighted Pairwise Grouping Method with Arithmetic mean) with Bray-Curtis method was employed to cluster the samples (sample clusters: serious gingivitis (purple), moderate gingivitis (orange) and healthy (yellow) clusters. Signi cant differences and the UPGMA clustering of samples were visualized by interactive Tree of Life (iTOL v.5.3) online platform [48].
Bioinformatics pipeline of whole metagenome sequence analysis of saliva samples Two strategies were applied for the shotgun metagenome data analysis of saliva samples. Read-based metagenomics was combined with a genome centric approach. The work ow of the data analysis is summarized in Figure 1B.

Raw sequence ltering
Galaxy Europe server was employed to pre-process the raw sequences (www.usegalaxy.eu). Low-quality reads were ltered by Prinseq (v.0.20.4), using the following parameters: min. length: 150; min. score: 15; quality score threshold to trim positions: 20; sliding window used to calculated quality score: 1 [49]. The quality of raw and ltered sequences was checked with FastQC program.

Read-based metagenomics
For taxonomic analysis of read-based metagenomics data the Kraken 2 (v.2.0.8) program was employed using the NCBI RefSeq genome database. The Kraken 2 feature table output were ltered by Kraken 2 lter command (con dence threshold: 0.8 indicated 80% precision to the lowest common ancestor).

Statistical analysis of read-based metagenomics data
Microbial communities were investigated with MEGAN6 and data were exported for statistic calculations. The e ciency of sequencing was monitored by computing rarefaction curves based on the comparison of the reads with sequence data in RefSeq (NCBI Reference Sequence Database) as well as with rRNA sequences deposited in RDP (Ribosomal Database Project) database. Rarefaction estimation was performed by MEGAN6 [43] (data not shown). The average composition of microbial taxa was visualized via Krona [44]. The distribution of top 10 frequent microbes between the three types of samples was plotted in Circos.
The interactive web tool MetaCoMET (Metagenomics Core Microbiome Exploration Tool) calculated microbial core and diversity. Shannon statistical method was performed to estimate alpha diversity. The Emperor program (integrated in MetaCoMET) did the principal component analysis.
The adolescent supragingival bio lm microbiota (amplicon sequencing) The supragingival bacterial communities in 112 supragingival bio lm samples were determined using next generation amplicon sequencing, in which a short, variable region of the 16S rRNA coding marker gene was ampli ed by PCR before sequencing. 36 samples derived from 9 adolescent patients diagnosed with induced gingivitis, 40 samples from 10 adolescent patients diagnosed with spontaneous gingivitis, and 36 samples from 9 age-matched healthy individuals without the symptoms of gingivitis were included in this study.

Sequencing depth
In all cases, rarefaction curves reached their asymptotes (data not shown). This corroborated a su cient sequencing depth to cover all genera in the bacterial communities. In other words, more than enough sequences were obtained from each sample to extract the total taxonomic information of the microbial community in them. Alignment of the quality-ltered reads with sequences in the NCBI RefSeq database allowed the identi cation of 172 operational taxonomic units (OTUs) in the samples analyzed (Supplementary table 2).
The core microbial community The richness or alpha diversity of OTUs was lower in the supragingival bacterial communities of induced and spontaneous gingivitis patients relative to healthy controls, but the difference was not signi cant (data not shown). Comparison of the microbial communities using PCA (principal component analysis) also showed a general overlap between supragingival plaque samples from gingivitis patients and healthy controls (data not shown).
Slight differences were noted depending on the reference databases employed. Distinct differences were not evident between the study groups at higher taxonomy levels, i.e. genus and beyond. Characteristic alterations took place among the low-abundance members of the community, however, only very few sequence reads could be associated with those taxa, which made statistical comparisons uncertain and likely led to erroneous conclusions. Therefore, the low-abundant species and genera (i.e.: relative abundance <0.01%) were excluded from the subsequent analyses. Annotation of the genera by RDP (Ribosome Database Project) revealed 23 abundant genera (relative abundance 1%) and 17 rare genera (relative abundance <1%), while using the RefSeq database, 20 abundant species (relative abundance 1%) and 21 rare species (relative abundance <1%) were identi ed (Supplementary table 3).
The general similarity of the OTUs in the three study groups allowed the amalgamation of all data and determination of the global microbiota of adolescent supragingival plaque (Figure 2A). Veillonella parvula was the most abundant bacterial species in all samples. At genus level, Prevotella, Veillonella, Actinomyces, Capnocytophaga and Streptococcus predominated in the supragingival plaques ( Figure  2A).
The most abundant bacterial species in the supragingival bio lms of the distinct study groups The identi ed 10 most abundant bacterial species were con rmed by both RDF and RefSeq databases and represented 75 % of the whole community. The following bacterial species were the most prevalent ones in our study (Figure 2 Not surprisingly, the most ubiquitous species belonged in the predominant phyla of the human oral microbiota, i.e. Firmicutes, Fusobacteria, Actinobacteria, Proteobacteria and Bacteroidetes [65].

Pair-wise comparison of study groups
Although the composition of the microbial communities in the supragingival bio lm of patients diagnosed with the two types of gingivitis and healthy controls were similar, the ranking order of predominant bacterial species was different in each group ( Figure 2B). V. parvula dominated the microbiota in induced gingivitis patients, with a relative abundance higher than 57%, followed by C. gracilis and species in lower abundance, i.e. S. sanguinis, H. parain uenzae, C. concisus and F. nucleatum. In contrast, in the supragingival bio lm of patients with spontaneous gingivitis, the predominant V. parvula was followed by F. nucleatum (relative abundance higher than 21%), P. intermedia and C. gracilis. In healthy controls, the dominant V. parvula was accompanied by R. dentocariosa and H. parain uenzae and the moderately abundant species F. nucleatum and S. sanguinis ( Figure 2B).
In order to uncover the characteristic and signi cant differences between the three subject groups, the data sets were compared pairwise (Figure 3). Perhaps the most pronounced of these was the alterations between the diseased and healthy patients. We observed an increased relative abundance of the genera Fusobacterium, Accermansia, Treponema and Campylobacter in supragingival plaques of gingivitis patients versus controls. In contrast, the genera Lautropia, Kingella, Neisseria, Actinomyces and Rothia were substantially more abundant in controls than in either of the two groups of gingivitis patients ( Figure 3A). The genus Megasphaera shoved notable relative distribution changes between the control subjects and induced and spontaneous gingivitis patients. In addition, relative abundances between the two groups of gingivitis patients were also noticeable, which might deserve further studies on large cohorts of subjects.
At the species level, a signi cantly higher abundance of C. concisus was apparent in both gingivitis groups versus the controls ( Figure 3B). We also observed that the relative abundance of Candidatus Saccharibacteria oral taxon TM7x, R. dentocariosa, R. mucilaginosa, Lautropia mirabilis and H. parain uenzae was lower in supragingival bio lms of either gingivitis patients versus healthy controls. Interestingly, Fusobacterium periodonticum was detected in healthy samples only. Comparison of bacterial species in the two groups of gingivitis patients showed that the relative abundances of Candidatus Saccharibacteria oral taxon TM7x, R. dentocariosa and H. parain uenzae were signi cantly higher in patients with induced versus spontaneous gingivitis. Other species, including P. intermedia, F. nucleatum, Parvimonas micra, Dialister pneumosintes, C. concisus, C. curvus and Aggregatibacter segnis were less abundant in induced gingivitis versus spontaneous gingivitis group ( Figure 3B). Out of the 13 signi cantly different species seven were positively con rmed by both RDP (on the taxonomic level of genus) and RefSeq (on the taxonomic level of specie) databases, marked with stars in Figure 3B. These patterns may be useful for diagnostic and future targeted therapy point of view.

Pathogens in the gingivitis and healthy study groups
Similar to principal component analysis and alpha diversity, clustering of supragingival microbiota using UPGMA (unweighted pair group method with arithmetic mean) did not result in a clear separation the three study groups. Samples from induced gingivitis patients (A), spontaneous gingivitis patients (B) and controls (C) appeared intermingled with each other on the UPGMA tree (see the letters preceding the tooth position numbers in the innermost circle in Figure 4). The samples did not separate into distinct clusters according to the Modi ed Gingival Index (MGI) re ecting the severity of gingivitis either.
Nevertheless, a different clustering was apparent according to the pathogenic nature of the microbial taxa. Oral microbes have been classi ed on the basis of their roles in pathogenesis and have been arranged in color coded segments of the "Socransky pyramid" [66,67] (Supplementary gure 1). Species belonging in the purple Socransky complex (mainly V. parvula) are indicated by purple ID numbers, i.e. study group letter and tooth position numbers (Figure 4, innermost circle). This sample cluster comprised 52 of the 112 supragingival microbiota included in this study. F. nucleatum (25 out of 36 microbiota) predominated in the cluster highlighted in orange, together with other members of the Socransky orange complex [66]. The third cluster, highlighted in yellow, showed a balanced distribution of prevailing species, such as R. dentocariosa, H. parain uenzae and V. dispar. 12 out of 25 samples were assigned to this cluster ( Figure 4).
It is important to note, that no correlation between the pathogenicity of the bacterial members of the supragingival plaques and the health status of the patients (gingivitis or health) could be recognized. We have also investigated the potential relationship between oral health status and the patients' gender, age and orthodontic appliance wearing duration. Although, there was no signi cant correlation between any of these parameters, a tendency indicating better status of the female patients relative to the male ones and shorter duration of wearing braces versus long exposure was noted. These suggest that there are no substantial differences among the study groups in their general oral health status, hence gingivitis (induced or spontaneous) can be reversed with proper oral hygiene.

Salivary microbiota of adolescent subjects (whole metagenome sequencing)
In addition to the dental plaques, saliva samples were also collected from the same study participants. In these experiments we wanted to examine the importance of sampling site, i.e. tooth bio lm associated and planktonic microbiota, of gingivitis patients relative to the microbiota of healthy controls belonging in the same age group. Saliva samples contained enough DNA to allow whole metagenome sequencing, thereby we could eliminate the potential random systematic error implied in amplicon sequencing [68]. The analysis and evaluation of sequence data followed a work ow ( Figure 1B) similar to the one applied to the amplicon sequencing, although in this case both read-based and genome-based analyses became possible, which extended the information content considerably.

Read-based analysis of the whole metagenome data
The most abundant bacteria in the saliva of study groups The richness and evenness of the salivary microbial communities did not differ signi cantly between either gingivitis study groups and healthy controls according to the Shannon indices (data not shown). Principal component analysis (PCA) of the salivary microbiota did not reveal distinct clusters either (data not shown). In this respect the whole salivary microbiota exhibited the same pattern as those found in the dental plaque bio lms with amplicon sequencing. Therefore, we could combine all saliva microbiota to generate a global picture of the adolescent saliva microbial landscape ( Figure 5A).
The genera Prevotella, Rothia, Streptococcus and Veillonella predominated the salivary microbiota. It is noteworthy that Veillonella was not a principal genus in saliva, as in the cases of the supragingival bio lm samples (see Figure 2A and 5A).
The relative distribution of the 10 most abundant bacterial genera were determined next. In induced gingivitis, the relative abundance of Prevotella was higher than in the two other groups. Streptococcus was the most abundant genus in spontaneous gingivitis ( Figure 5B). In control saliva samples, the relative abundance of Rothia was comparable with that of Prevotella. In each study group, 14 genera accounted for >95% of the taxa identi ed at the genus level.

Genome-based evaluation of the saliva sequencing data (binning)
In addition to the read-based data, bioinformatics analysis of saliva samples was accomplished by genome-centric binning ( Figure 1B). In this approach, the ltered sequences were rst assembled in contigs, which were then distributed into virtual bins, based on their inherent sequence features. Inspection of the genetic content of the individual bins supplied detailed information about taxonomy from a viewpoint distinct of the read-based approach and in many cases about the genes coding for possible speci c metabolic pathways.
Genomic fragments belonging in the most abundant taxa of the healthy salivary microbiome [69] could be detected in the saliva of each study group. In line with the results of read-based metagenomics, most of the putative genomes identi ed by binning belonged to Prevotella species, which comprised 8 separate bins, whereas the putative genomes of species belonging in the genera Veillonella and Streptococcus comprised 4 separate bins, respectively. The genomes of Actinomyces and Rothia species were detected in 2 distinct bins, each. The genomes of 9 other species as well as the family Porphyromonadaceae and the taxon Candidatus Saccharibacteria TM7x occupied a single distinct bin ( Figure 6). All these genera, were present in both supragingival bio lms and planktonic saliva samples.
The overall similarity of the read-based and genome-based microbiota validated each other, starting from the same saliva sequencing databases the two distinct bioinformatics approaches gave comparable pictures of the microbial communities.

Discussion
Gingivitis is a non-destructive disease affecting both young and elderly worldwide [19,70]. Typically, the causative agent is the dental plaque and removal of the supragingival bio lm from the surface can revert the in ammatory response elicited by the bio lm-forming bacterial community [3,6,71,72]. When proper oral hygiene is not restored, gingival in ammation may persist and in susceptible individuals it can develop to periodontitis, a destructive in ammatory disease characterized with irreversible changes including loss of the tissues that attach the tooth to the alveolar structures, and alveolar bone loss [8][9][10]. Metagenomic analysis of the bacterial taxa in dental plaque samples collected from gingivitisaffected teeth and non-affected teeth bio lms revealed differences in the composition and community structure [11][12][13][14][15][16][17][18]. Most of the earlier sequencing studies analyzed pooled plaque samples of adult gingivitis patients or healthy adults. These observations suggested that the relative abundance of various bacteria in supragingival as well as subgingival bio lms may play a role in the progress of gingival in ammation [73].
We applied next-generation sequencing to characterize the bacterial communities in individual supragingival bio lms of adolescent patients diagnosed with either induced or spontaneous gingivitis. Similar samples were taken from the teeth of age-matched healthy controls. Four distinct supragingival bio lm samples were collected from the surface of four different teeth of each study participant and were analyzed one by one. Furthermore, in order to reveal whether the characteristic features of the supragingival microbes were re ected in the salivary microbiota of the participants, we also identi ed the predominant bacterial taxa in the non-stimulated saliva samples of the same subjects.
The overall composition and diversity of supragingival bio lm communities was similar in the three study groups (Figures 2A and 5A), although the ranking order of predominant bacterial species was unique for each group (Figures 2B and 5B). Veillonella parvula, an early colonizer of the tooth surface predominated the supragingival bio lm microbiota in all three study groups (Figure 2A), whereas Prevotella species were the most prevalent in the saliva samples ( Figure 5A), indicating their preference for planktonic lifestyle. V. parvula was abundant in supragingival plaques in an experimental gingivitis study, and it was suggested that via co-aggregation with other salivary bacteria, they played an important role in the formation and growth of multispecies microbial communities [7,74]. In other investigations Veillonella was found to be one of the most abundant genera both in supragingival and subgingival plaques of healthy adults [69].
One may envisage gingivitis and periodontitis as successive stages of an in ammatory cascade [75]. In this respect it is noteworthy that Veillonella species were implicated in the pathogenesis of periodontitis as members of the so called Socransky "purple complex" [66,67] (Supplementary gure 1). Although the Socransky complexes were denoted originally in subgingival plaque samples of periodontitis patients, the presence or absence of these bacterial complexes is also indicative of gingival health [67]. Early colonizing bacteria on the supragingival tooth surfaces are members of the "yellow", "green" and "purple" complexes, whereas members of the "orange" complex integrate into to the supragingival bio lm later [66,67,76]. Due to its elongated shape and diverse metabolic interactions Fusobacterium nucleatum, a member of the "orange" complex, contributes to the organization of supragingival bio lms [77]. Our observations are compatible with the previously suggested role of Fusobacterium, Campylobacter and Treponema in the pathogenesis of gingivitis, whereas the less abundant Catonella, Lachoanaerobaculun, Schwartzia and Akkermansia genera remain to be characterized in this respect [78,79]. Some genera, including Lautropia, Kingella, Neisseria, Actinomyces, Rothia and Megasphaera were extensively more abundant in the control samples. These genera have been associated with gingival health [15,34,67,[80][81][82][83]. Comparison of relative abundances of bacterial species revealed a substantially higher abundance of C. concisus in the two gingivitis groups versus controls (Figure 3). C. concisus, a member of the "green" complex, is ubiquitous in the oral cavity, it has been found both in periodontal in ammation and periodontal health associated microbiomes [67]. Species belonging in the "orange" complex, such as P. intermedia, F. nucleatum. P. micra, and "purple" complex (A. segnis) as well as the new periodontopathogenic species D. pneumosintes [84,85] (shown in the grey box in Supplementary gure 1) were more abundant in supragingival bio lms of the non-induced gingivitis group, but not in the induced gingivitis group (Figure 3) [66,67,76,85].
Comparisons at species level did not expose conceivable differences in the relative abundances of the Gram-negative anaerobic periodontal pathogens belonging in the "red" complex (Porphyromonas gingivalis, Treponema denticola and Tannerella forsythia), although the relative abundance of the genus Treponema was higher in gingivitis samples (Figure 3).
In contrast, the bacterial species Candidatus Saccharibacteria oral taxon TM7x, R. dentocariosa, R. mucilaginosa, L. mirabilis and H parain uenzae were detected in higher relative occurrence in supragingival plaques of the control group, relative to the gingivitis patients ( Figure 3). Candidatus Saccharibacteria oral taxon TM7x was implicated in periodontal disease, while the increased abundance of the other species in control samples is compatible with their classi cation as "Associated with Periodontal Health" [15,67]. It is noteworthy, that Fusobacterium periodonticum was identi ed in the control samples only although F. nucleatum occurred in all three study groups.
The development of gingivitis in patients undergoing orthodontic therapy has been linked to plaque accumulation at the new retention sites, which are di cult to access for oral hygiene [33,34]. The composition of the bacterial communities in induced gingivitis was also reported distinct from the microbiomes in naturally occurring gingivitis [33]. Our results, based on more advanced bioinformatics analysis, did not corroborate these ndings (Figure 4). In spite of the overall similarity, differences were detected at the species level resolution: Candidatus Saccharibacteria oral taxon TM7x, and healthassociated species R. dentocariosa and H. parain uenzae were markedly higher in patients with induced versus non-induced gingivitis ( Figure 3B), while F. nucleatum,P. intermedia,P. micra, C. concisus, C. curvus and A. segnis showed opposite tendencies ( Figure 3B).
Mapping the salivary bacterial community may be an easy and helpful diagnostic tool of distinct pathological processes [86]. Nevertheless, under the employed stringent evaluation conditions we did not observe major differences between the salivary microbiome of patients with gingivitis compared to controls.
This nding may be related to the salivary microbiome comprising the easily detachable mixture of diverse bacterial communities inhabiting the surface of the tongue, tonsils, and throat as well as the microbes of the supragingival plaque [69].

Conclusions
Our study indicated that the relative abundance of distinct bacterial taxa in supragingival bio lms may differ noticeably, although not extensively, between induced and non-induced gingivitis patients in spite of the global similarities.
The higher relative abundance of certain bacterial taxa in supragingival bio lms of gingivitis patients versus controls signal their involvement in the pathogenesis of the disease and may thus be of diagnostic value to prevent further escalation towards periodontitis.
Mapping salivary microbiome may obscure the variations between healthy and gingivitis status.
At any rate, a stringent and coherent bioinformatics work ow should be employed to detect the relatively small alterations in microbiome composition and taxonomic abundances.

Competing interests
The authors declare that the research was conducted in the absence of any commercial or nancial relationships that could be construed as a potential con ict of interest.     [35,36] are indicated as red, orange and green sectors in the next circle. Predominant bacterial taxa identi ed in each supragingival bio lm sample are shown as colored columns at the outer edge of the circular diagram above numbers indicating the age of the patients. Color-coding of the bacterial groups was based on their association with microbial complexes involved in oral pathogenesis, indicating their potential contribution to oral health [66,67,76] (Supplementary gure 1).  Bacterial taxa identi ed in saliva samples of induced and non-induced gingivitis patients and controls using genome-based evaluation of sequencing data. The distribution of contigs are plotted in the circular bands. The grouping of contigs was based on sequence-assignments of automated binning programs METABAT2, MAXBIN2 and CONCOCT as well as manually de ned bins as presented on the Anvi'o platform. The list of identi ed bins is given at the bottom part of the gure.